Bradt R C Munz D Sakai M White K W Fracture Mechanics of Cer PDF

Fracture Mechanics
of Ceramics
Active Materials, Nanoscale Materials,
Composites, Glass and Fundamentals
Volume 1 Concepts, Flaws, and Fractography
Volume 2 Microstructure, Materials, and Applications
Volume 3 Flaws and Testing
Volume 4 Crack Growth and Microstructure
Volume 5 Surface Flaws, Statistics, and Microcracking
Volume 6 Measurements, Transformations, and High-Temperature Fracture
Volume 7 Composites, Impact, Statistics, and High-Temperature Phenomena
Volume 8 Microstructure, Methods, Design, and Fatigue
Volume 9 Composites, R-Curve Behavior, and Fatigue
Volume 10 Fracture Fundamentals, High-Temperature Deformation, Damage,
and Design
Volume 11 R-Curve Behavior, Toughness Determination, and Thermal Shock
Volume 12 Fatigue, Composites, and High-Temperature Behavior
Volume 13 Crack Microstructure Interaction, R-Curve Behavior, Environmental
Effects in Fracture, and Standardization
Volume 14 Active Materials, Nanoscale Materials, Composites, Glass and
Fundamentals
Fracture Mechanics
of Ceramics
Active Materials, Nanoscale
Materials, Composites, Glass
and Fundamentals
Edited by
R. C. Bradt
University of Alabama
Tuscaloosa, Alabama
D. Munz
University of Karlsruhe
Karlsruhe, Germany
M. Sakai
Toyohashi University of Technology
Toyohashi, Japan
and
K. W. White
University of Houston
Houston, Texas
Proceedings of the 8th International Symposium on Fracture Mechanics of Ceramics, held February
25 28, 2003, at the University of Houston, Houston, Texas
ISSN: 0197 2766
ISBN 10: 0 387 24134 5 Printed on acid free paper.
ISBN 13: 978 0387 24134 0
ß 2005 Springer ScienceþBusiness Media, Inc.
All rights reserved. This work may not be translated or copied in whole or in part without the
written permission of the publisher (Springer ScienceþBusiness Media, Inc., 233 Spring Street, New
York, NY 10013, USA), except for brief excerpts in connection with reviews or scholarly analysis.
Use in connection with any form of information storage and retrieval, electronic adaptation,
computer software, or by similar or dissimilar methodology now known or hereafter developed is
forbidden.
The use in this publication of trade names, trademarks, service marks and similar terms, even if they
are not identified as such, is not to be taken as an expression of opinion as to whether or not they are
subject to proprietary rights.
Printed in the United States of America.
9 8 7 6 5 4 3 2 1 (SPI/EB)
springeronline.com
CONTENTS
Preface ix
1. Contact Mechanics
Atomic Force Microscopy at Ultrasonic Frequencies 1
Walter Arnold, A. Caron, S. Hirsekom,
M. Kopycinska-Müller, U. Rabe, M. Reinstädtler
Indentation Size Effect on the Hardness of Zirconia Polycrystals 13
T. Akatsu, S. Numata, M. Yoshida,Y. Shinoda, F. Wakai
Indentation Fracture, Acoustic Emission and Modelling of the
Mechanical Properties of Thin Ceramic Coatings 21
S.J. Bull, I. Arce-Garcia, E.G.-Berasetegui
and T.F. Page
Nanoindentation, Nanoscratch and Nanoimpact Testing of
Silicon-Based Materials with Nanostructured Surfaces 43
Jinjun Lu, T. Xu, Q. Xue, B.D. Beake
Microstructural Control of Indentation Crack Extension under
Externally Applied Stress 57
Robert F. Cook
Instrumented Hardness Test on Alumina Ceramics and
Single Crystal with Spherical Indenter 69
Shuji Sakaguchi, K. Hirao, Y. Yamauchi, S. Kanzaki
2. Glass
Controlling the Fragmentation Behavior of Stressed Glass 77
R. Tandon, S.J. Glass
Elasto-Plastic Behavior of Glassy Carbon and Silica Glass
by Nanoindentation With Spherical-Tipped Indenter 93
N. Iwashita, M. Swain, J.S. Field
Scratch Test for Evaluation of Surface Damage in Glass 101
S. Yoshida, T. Hayashi, T. Fukuhara, K. Soeda,
J. Matsuoka, N. Soga
Shear-Driven Damage and Internal Friction in Indentation
Loading of a Glass-Ceramic 113
Anthony Fischer-Cripps
Indentation and Scratching of Glass: Load, Composition,
Environment and Temperature Effects 121
Jean-Christophe Sangleboeuf, Tanguy Rouxel
v
vi Contents
Effects of Tin on the Physical Properties and Crack Growth

in Soda-Lime-Silica Float Glass 135
Matthew H. Krohn, John R. Hellmann,
Carlo G. Pantano, Nathan P. Lower, and Richard K. Brow
Indentation Size Effects for Glasses: Yes, There is a Fracture 149
George D. Quinn
3. Fracture of Nano-Scale Materials

Slow Crack Propagation in Ceramics at the Nano- and Micro-
scale: Effect of the Microstructure 173
J. Chevalier, G. Fantozzi
Toughening and Strengthening Mechanisms in Nanocomposites
based on Dislocation Activity 191
Hideo Awaji, S-M Choi
4. Composites
Creep Behavior and Mechanisms for CMCs with
Continuous Ceramic Fibers 203
Jean-Louis Chermant, Gaëlle Farizy, Guillaume
Boitier, Séverine Darzens, Jean Vicens, and
Jean-Christophe Sangleboeuf
Strain Accumulation and Damage Evolution during Creep of
SiCf/SiC Composites 221
B. Wilshire, H. Burt
Modeling Multilayer Damage in Cross-ply Ceramics Matrix
Composites 233
M. Kashtalyan, H.W. Chandler
Quantification of Toughness Increase Due to Metal Particles
in Glass Matrix Composites 245
M. Kotoul, A.R. Boccaccini, I. Dlouhy
Fracture Resistance of Hybrid Glass Matrix Composite
and its Degradation Due to Thermal Aging and
Thermal Shock 263
I. Dlouhy, Z. Chlup, S. Atiq, A.R. Boccaccini
Creep Investigation on SiCf-SiBC Composites 275
Gaëlle Farizy, Jean-Louis Chermant, Jean Vicens,
Jean-Christoph Sanglebœuf
Fracture Toughness of BaTiO3 MgO Composites Sintered
by Spark Plasma Sintering 287
S. Rattanachan, Y. Miyashita
Contents vii
Fracture Toughness of BaTiO3 and BaTiO3-AI203 Composite

Under Applied Electric Field 297
S. Rattanachan, Y. Miyashita, Y. Mutoh
Room temperature Post-Fracture Tensile (PFT) Study of Two
Fiber-Reinforced Ceramic Matrix Composites with
a ‘‘Strong’’ Fiber/Matrix Interface 307
R.A. Longoria, N. Du Souich, K.W. White
An FE Model of Carbon/Carbon Composite Fracture 317
J. Ohgi, J.H.. Jackson, A.S. Kobayashi, K.W. White
5. Fracture Fundamentals
Tailoring the Composition of Self-Reinforced Silicon Nitride
Ceramics to Enhance Mechanical Behavior 327
P.F. Becher, G.S. Painter, H.T. line and M.K. Ferber
Nonlinear Fracture of a Polycrystalline Graphite: Size-Effect
Law and Irwin’s Similarity 337
Motosugu Sakai, R. Nonoyama
Fragmentation of Ceramics in Rapid Expansion Mode 353
S. Maiti, P. Geubelle, K. Rangaswamy
Experimental Approach to Microfracture Process of
Ceramics Under Thermal Shock 365
S. Wakayama, K. Nishino, K. Kubota, T. Wada, S. Kobayashi
Finite Element Analysis of Crack-Path Selection in a Brick and
Mortar Structure 375
R. Sarrafi-Nour, M. Monaharan, and C.A. Johnson
Investigation of Crack Growth Process in Dense
Hydroxyapatite using the Double Torsion Method 387
C. Benaqqa, J. Chevalier, M. Saâdaoui, and G. Fantozzi
Crack Propagation Behavior of Alumina with Different
Grain Sizes Under Static and Cyclic Fatigue 399
H. El Attaoui, M. Saadaoui, J. Chevalier, G. Fantozzi
Effects of Pore/Grain-Size Interaction and Porosity on the
Fracture of Electroceramics 411
C. Lu, R. Danzer, F. Fischer
6. Fracture of Coatings
Fracture Behaviour of Plasma Sprayed Thermal
Barrier Coatings 421
J. Malzbender, T. Wakui, E. Wessel, R.W. Steinbrech
viii Contents
Particle Impact Damage and Point Load-Induced Fracture

Behavior in Zirconia Plasma Spray Coating Film 437
Y. Akimune, K. Matsuo, T. Sugiyama, H. Yoshida,
S. Sodoeka, S. Shimizu
Mode I, Mode II, and Mixed-Mode Fracture of
Plasma-Sprayed Thermal Barrier Coating at Ambient
and Elevated Temperatures 451
Sung R. Choi, D. Zhu, R. Miller
7. Ferroelectric Materials
Modeling of Fracture in Ferroelastic Ceramics 471
C.M. Landis
Strength and Reliability of Lead Zirconate Titanate Ceramics 485
C.S. Watson
8. Reliability Prediction, Standardization and Design

Standard Reference Material 2100: Fracture Toughness of Ceramics 499
G.D. Quinn, K. Xu, R. Gettings, J.A. Salem, J.J. Swab
Measuring the Real Fracture toughness of Ceramics
ASTM C1421 531
J.A. Salem, G.D. Quinn, M.G. Jenkins
Predicting the Reliability of Brittle Material Structures Subjected
to Transient Proof Test and Service Loading 555
N.N. Nemeth, O.M. Jadaan, T. Palfi, and E.H. Baker
Estimation and Stimulation of Slow Crack Growth Parameters
from Constant Stress Rate Data 579
J.A. Salem, A.S. Weaver
On Integrity of Flexible Display 597
P.C.P. Bouten
Fracture of Conductive Cracks in Poled and Depoled PZT-4 Ceramics 609
Tong-Yi Zhang
Increasing Resistance to Low Temperature Ageing Degradation
of Y-TZP by Surface Modification 625
A. Feder, P. Morcillo, M.J. Anglada
Index 635
PREFACE
The 8th International Symposium on fracture mechanics of ceramics was held on

the campus of the University of Houston, Houston, TX, USA, on February 25-28,
2003. This marks three decades of progress in the science and technology develop-
ment in ceramics since the first meeting of this group, held at the Pennsylvania State
University in 1973, which was published in Volumes 1 and 2 of this series. The
meetings continued over approximately four year intervals, held subsequently in
Virginia Polytechnic Institute and State University, Japan Fine Ceramics Center,
Nagoya, Japan, and Kernforschungszentrum, Karlsruhe, Germany.
With the natural maturing of the fields of structural ceramics, the present
symposium has focused on nano-scale materials, composites, thin films and coatings
as well as glass. The symposium also addressed new issues on fundamentals of
fracture mechanics and contact mechanics, and a session on reliability and stand-
ardization.
The International Organizing Committee gratefully acknowledges the support
from the Cullen College of Engineering, and the dedicated help from the local
organizing committee. This includes Prof. L. Wheeler, Prashant Haldipur, Jennifer
Husmo, Sandra Luna, Charlotte Palm and Jonathan Claydon. Additionally, I wish
to thank the materials students who participated in the effort: Duo Liu, Mike Chelf,
Marcos Flores, Rafael Longo and Yi Fang.
Special thanks are extended to NASA Astronaut Dr. Bonnie Dunbar for her
wonderful banquet presentation.
Tuscaloosa, AL USA R.C. Bradt

Karlsruhe, Germany D. Munz
Toyohashi, Japan M. Sakai
Houston, TX USA K.W. White
ix
ATOMIC FORCE
MICROSCOPY AT
ULTRASONIC FREQUENCIES
W. Arnold, A. Caron, S. Hirsekorn,

M. Kopycinska-Müller, U. Rabe, and M. Reinstädtler*
I. INTRODUCTION
Atomic Force Microscopy (AFM) is a near-field technique to generate high-reso-

lution images of surfaces. A micro-fabricated elastic beam with an integrated sharp
sensor tip at its end is scanned over the sample surface. With various dynamic modes,
leading to Force Modulation Microscopy [1], Ultrasonic Force Microscopy [2],
Atomic Force Acoustic Microscopy (AFAM) [3 5], Microdeformation Microscopy
[6], Scanning Local Acceleration Microscopy [7] or Pulsed Force Microscopy [8],
images can be obtained in which the contrast depends on the elasticity of the sample
surface. However, quantitative determination of Young’s modulus of a sample
surface with AFM is a challenge, especially when stiff materials such as hard metals
or ceramics are encountered. In this presentation the basic idea of AFAM, i.e. the
evaluation of the cantilever vibration spectra at ultrasonic frequencies is discussed.
The AFAM technique can be used for imaging as well as to discern local elastic data
quantitatively. Nanocrystalline magnetic materials [9], multidomain piezoelectric
materials and silicon [10], diamond-like carbon layers [11], polymeric materials and
clay crystals [12] have been examined previously. The spatial resolution of the
AFAM technique is approximately 10 nm.
In an Atomic Force Acoustic Microscopy set-up [3-5] either the sample is
insonified or the cantilever suspension is excited at ultrasonic frequencies. The
vibration spectra of the cantilever depend on the local sample stiffness and hence
*
Fraunhofer Institute for Nondestructive Testing (IZFP), Bldg. 37, University, D 66123 Saarbrücken,
Germany
1
2 W. Arnold et al.
on the local elasticity of the sample constituents. Acoustic images can be obtained by
measuring the cantilever vibration amplitude at a frequency close to a contact
resonance or by monitoring the resonance frequency [13,14]. With this technique it
is possible to resolve the differences in the local tip-sample stiffness between phases
and/or grains having different crystal lattice orientation. In the case of ferroelectric
materials the presence of the domains influences the local elastic properties within
each individual grain. Depending on the method of sample preparation, for example
etching, one can make domain patterns visible on the sample surface using AFM [15].
In this case the contrast is caused by topography. A variety of AFM techniques have
been applied to image or modify the domains in ferroelectric single crystals as well
as in thin films [16,17]. The AFAM technique allows domain structure imaging in
single crystals and in ceramics, even when domains do not appear in the topography
image [18]. Coupling the AFAM technique with the so- called piezo-mode [16,18 21]
can additionally provide information about domain polarization and the orientation
of the grains within the crystal lattice [20].
2. AFAM EXPERIMENTAL TECHNIQUE
An AFM cantilever clamped on one side and free on the other can be described as a
micro-beam with a characteristic set of free resonance frequencies. The most common
modes excited are either flexural or torsional. Values of the free flexural or torsional
oscillations can be measured with e.g. an instrument such as a Dimension 3000
Nanoscope (Digital Instruments, Santa Barbara, Ca USA) or a Solver P47H (NT-
MDT, Zelenograd, Russia) which were employed for this study. When the cantilever
approaches the sample surface and finally touches it, the forces interacting between the
cantilever tip and the surface modify the boundary condition of the vibrating canti-
lever [4,5]. As a result, the resonance frequencies shift to higher values. These reson-
ance frequencies are called contact resonance frequencies (Fig. 1 and Fig. 2). The
values of free and contact resonance frequencies can be calculated as discussed
previously [4,5]. From the difference of the contact resonance frequencies relative to
the free resonances, one obtains the vertical or lateral contact stiffness [5, 22] which
itself depend on both the elastic moduli of the sample and the tip, and their Poisson
ratio, respectively [23,24]. As shown in Fig. 1, the tip contacts the sample surface over a
certain contact area which is inferred indirectly by using a calibration material of
known elasticity. Detailed information about tip-sample interactions and methods of
evaluating local tip-sample contact stiffness can be found elsewhere [9,10,25]. Under
certain circumstances, the local elastic modulus or the so-called indentation modulus
M [24] can be derived from the contact stiffness [9,10,18]. At present the contribution
of adhesion forces to the contact stiffness is made small in AFAM measurements by
applying a sufficiently high static force on the cantilever so that the Hertzian contact
forces are dominant [23]. When evaluating the vertical contact stiffness, the lateral
forces to k are neglected. The indented volume is about pr2h 3rh 500 nm3 where rh
is the contact radius, here about 10 nm or less.
The contact resonance frequencies are excited by an ultrasonic transducer,
which emits longitudinal waves or shear waves into the sample. Figure 2 displays
the normalized contact resonance spectra of the first (Fig. 2a) and second (Fig. 2b)
flexural mode measured on fused quartz, silicon, and sapphire. The spring constant
Atomic Force Microscopy At Ultrasonic Frequencies 3
Figure 1. Principle of Atomic Force Acoustic Microscopy. Using the shift of the contact resonance
frequencies relative to the free resonances, one can determine the local elastic constants of a material.
The adhesion forces contribute to the measured vertical contact stiffness as well. Their influence can be
neglected, provided the static forces are adjusted so that they are much larger than the adhesion forces.
Figure 2. Normalized contact resonance spectra of the first (Fig. 2a) and second (Fig. 2b) flexural mode
measured on samples with different elastic properties. The second mode on silicon shows a triple
resonance.
of the cantilever was 44 N/m and the free resonance frequencies were 164 kHz and
1023 kHz for the first and second mode, respectively (not shown). The values of the
indentation modulus of the reference samples varied from 77 (fused quartz), to
165 GPa (silicon), and to 423 GPa (sapphire). Very often double or triple resonance
appear which are caused either by mode-coupling [26] or by multiple contacts
between the tip and the surface.
In contact the resonance frequency varies during scanning and hence the amp-
litude and the phase of the cantilever vibration vary as well if the excitation frequency
of the transducer is kept constant. This amplitude and phase contain information
about the local tip-sample stiffness, and hence can be used as imaging quantities,
see Fig. 3. It shows an AFAM-amplitude image of a lead calcium titanate film
annealed at 6508 C. The size of the images is 2 2 mm2 . The cantilever material
was coated with a conductive diamond-like layer. The overall spring constant was
4 W. Arnold et al.
44 N/m and the free resonance frequency of the first flexural mode of the cantilever
was 188 kHz. The contrast seen in an amplitude image depends on the frequency of
the applied ultrasound relative to the contact resonance frequency. At a frequency of
733 kHz the softer grains appear brighter than the stiffer grains (Fig. 3a). Imaging at
a higher frequency, 791 kHz, the contrast of the image inverts (Fig. 3b) and the stiffer
grains appear now brighter than the more compliant grains. This contract inversion
has been known for a long time and has been noticed by several groups. Figure 4
shows an AFAM image of a nanocrystalline nickel sample based on monitoring the
local contact resonance. From this measuring quantity, the local stiffness k is
determined and then used as an imaging quantity. Values of the k ranged from
728 to 992 N/m. The tip used in this experiment was made from silicon, making it
susceptible to tip wear [11]. The spring constant of the cantilever employed was 48 N/
m and its free resonance frequencies were 166 and 1031 kHz for the first and second
mode, respectively. The size of the image is 1:5 1:4 mm2 . The image clearly shows
ensemble of grains in the nanocrystalline structure.
Figure 3. AFAM amplitude images of a lead calcium titanate thin film annealed at 6508 C. The contrast
is caused by a shift of the resonance frequency when the contact stiffness changes locally. This in turn
causes a change of the signal amplitude when the excitation frequency is kept constant.
Figure 4. AFAM image of a nanocrystalline nickel sample. The contrast of the image is based on the
contact stiffness comprising values between 728 N/m and 992 N/m. These values are determined from
the contact resonances. The size of the image is 1:5 1:4 mm2 .
In the so-called piezo-mode of atomic force microscopy an ac voltage is applied

to a conductive AFM cantilever while scanning the surface of a piezoelectric mater-
ial. The tip of the cantilever senses the local deformation of the surface caused by the
electric field between the tip and a counter electrode (Fig. 5b, see also Fig. 10).
Usually the ac frequency is far below the free resonance frequency of the AFM
cantilever [16,17,19,20]. In BaTiO3 , an image series based on vertical and torsional
cantilever vibration signals of the same surface area allowed the reconstruction of the
domain orientation using this mode [20].
The piezo-mode technique can also be performed at frequencies in the range of
the first free resonance of the cantilever [27]. In the images shown here, the ac
frequency was set equal to a contact resonance frequency [28] which entails resonance
amplification increasing the contrast and signal-to-noise ratio in an image. Figures 6a
and 6b show images in the AFAM and ultrasonic piezo-mode for comparison. In
contrast to AFAM amplitude images, piezo-mode images do not show contrast
inversion when the excitation frequency is varied. Furthermore it can be seen that
in the piezo-mode image (Fig. 5b) that there are areas with no piezo-activity (area
with question mark). Comparing Fig. 5a with Fig. 5b, one can notice that areas of
higher tip-sample contact stiffness correspond to areas where there is vertical piezo-
activity. This can only happen if the polarization of the domains is dominantly
vertical. The dark area in the AFAM image corresponding to low tip-sample contact
stiffness can either be caused by an in-plane orientation of the domains or because
the film is not piezoelectric at such locations and hence not ferroelectric. Further
investigations will follow.
As mentioned above rectangular AFM cantilever beams can also be forced to
torsional vibrations. In this case the experimental set-up is such that an ultrasonic
transducer emits shear waves into the sample causing in-plane surface vibrations. The
shear wave transducer is oriented so that the surface vibrations are polarized per-
pendicular to the long axis of the cantilever. If low excitation amplitudes (0.1 nm) are
applied and if the excitation frequency is set close to a contact resonance frequency,
the amplitude and the phase of the cantilever vibration contain now information
about the local lateral tip-sample stiffness. Used as imaging quantity, they yield
images of shear stiffness. By increasing the lateral excitation amplitude much above
Figure 5. AFAM (a) and ultrasonic piezo mode (b) images of the lead calcium titanate sample annealed
at 6508 C. The image in (a) is identical to the one in Fig. 3a.
6 W. Arnold et al.
0.1 nm, there is a change in the shape of the resonance curves. As shown recently [29]
stick-slip between tip and sample occurs. This can be identified by taking local lateral
spectra at one point of the sample: The maximum of the resonance curve increases
linearly with the amplitude of the lateral displacement only at low excitation. At a
certain threshold amplitude, the amplitude of the torsional vibrations does not
increase any more and the resonance curve flattens out indicating the onset of sliding
friction. Additionally, higher harmonics of the excitation signal appear which can be
used for imaging. Such an image is shown in Fig. 6. A silicon sample was imaged first
in the topography mode (Fig. 6a). At the excitation frequency of f 238 kHz, an
image was taken in the regime of stick-slip of the cantilever motion (Fig. 6b). Clearly
an image could also be taken at 3f, here 714 kHz (Fig. 6c). It displays information
related to the local friction coefficient. This new imaging mode is also at present the
object of ongoing research.
3. APPLICATION OF AFM AND AFAM TO CRACK IMAGING
AFM and AFAM can be used to image cracks in ceramics in a similar way as can be
done using SEM, Optical Microscopy or Scanning Acoustic Microscopy [30,31]. If
one fractures a surface of a brittle material with a Vickers indenter, it is well-known
and standard practice to infer from the length of the radial cracks the fracture
toughness of the material [32]. There is much experience needed to determine reliably
the length of the cracks with any of the microscopies. In this study the crack opening
displacement U(x) was measured with an AFM. Measured close to the crack tip away
from bridges in the wake of the crack, this would allow one to obtain information on
Ktip using on the equation [33]
p
U(x) ¼ 8x=pKtip =E’ (1)
Figs. 7a and 7b show AFM images of Si3 N4 ceramics where fracture was caused in
bending specimens of the size 3 4 25 mm3 by a 10 kg load using a Vickers
indenter. The size of the images is 30 30 mm2 . One can clearly see the cracks,
however, it was difficult to exactly determine the origin of the cracks. Using the
routine ‘‘section analysis’’ of the DI 300 AFM instrument, U(x) was measured as a
function of x. When evaluating the data, a possible offset for the exact position of
Figure 6. a) Topography of a Si sample, b) Torsional amplitude f fexcitation , c) Torsional amplitude

f 3 fexcitation .
Figure 7. AFM topography images of two Si3 N4 samples showing radial cracks caused by a Vickers
indenter. The COD along the length of the radial cracks (arrows) was measured using standard routines of
the DI instrument. A possible offset for the exact location of the tip was allowed in the analysis. The size of
the image is 30H30 mm2 and the topography scale was 120 nm (a) and 130 nm (b).
Sample # 12, Crack Nr. 1 Sample # 12, Crack Nr. 2

160 160
After few hours After few hours
Width U [nm]
Width U [nm]
120 After 2 days 120 After 2 days

80 80
40 Ktip = 6.69 MPa.m1/2 40 Ktip = 6.46 MPa.m1/2

Ktip = 6.85 MPa.m1/2 Ktip = 6.60 MPa.m1/2
0 0
0 3 6 9 12 15 0 3 6 9 12 15
Position (X) [mm] Position (X) [mm]
Figure 8. Crack opening displacement (COD) measured with an AFM in the topography mode. The
figures display the width of the COD measured a few hours after the indent was made. The measurement
was repeated after two days leading to a slight increase in the stress intensity factor. The data shown in
Fig. 8a were obtained on the sample shown in Fig. 7a, and Fig. 8b corresponds to Fig. 7b.
the crack tip was taken into account. The reduced modulus E’ for Si3 N4 is 290 GPa.
The value for Ktip was around 6:7 MPam0:5 (Figs. 8) within the range of published
data [32]. The high value indicates, however, that wake-effects played a role in the
measurements, also the fact that the profile measured was not parabolic, see Figs. 8,
as stipulated by Eq. (1)
We also studied in-situ crack propagation in an AFM by inducing cracks
electrically, see Fig. 9. Poled PIC 151 PZT samples were loaded electrically with dc
voltage of 900 V between the electrodes. Because the upper electrode was smaller
than the ground electrode, an inhomogeneous electric field developed which was
large enough to induce cracks due to the piezoelectric stresses.
Figure 10a shows a topography image of the environment of an individual crack
in a crack field generated this way and Fig. 10b shows the corresponding AFAM
image. Continued application of voltage led to additional cracking (Fig. 10c and 10d,
arrow). Note that the AFAM images show also the domain structure opening the
way to learn in-situ information on the role of the reorientation of domains in crack-
shielding etc. which was much discussed during FCM8. Further studies are under
way to clarify their role.
8 W. Arnold et al.
Conductive cantilever Plezo-Mode 7mm

Excitation of Cracks
Sample oscillation cantilever vibrations
13 mm
Electrodes
3 mm
Function
Specimen generator
0-10 Vss
Electrical insulating delay-line AFAM 10 kHz - 3 MHz Sample
Transducer
Figure 9. Configuration for the generation of in situ cracks in an AFM allowing one to observe these
cracks using the topography, the AFAM, and the ultrasonic piezo mode. The insulating delay line was
necessary in order to protect the ultrasonic transducer from the applied dc voltages. The cracks emanate
perpendicularly from the top electrode. The figure shows also the use of the electrodes for the ultrasonic
piezo mode.
Figure 10. Observation of in situ crack propagation after applying successive voltages up to 900 V to a
PZT sample PIC 151; Fig. 10a: Topography image; Fig. 10b: AFAM image. The domain structure
becomes visible. The images were taken with a cantilever having a spring constant of 47 N/m and at a
static load of 1,8 mN. Continued voltage application leads to additional crack formation and reorientation
of the domains (Fig. 10c and d, arrow). Image size of the individual images is 10H10 mm2 . The grey scale in
the topography images comprises 70 nm.
Finally we studied crack propagation using the lateral stiffness as contrast in the
images. Here the sensor tip vibrates in contact with the sample surface. An excitation
frequency close to a torsional contact resonance frequency is selected. The torsional
amplitude of the cantilever at this frequency is detected using the lateral channel
of the AFM and a lock-in amplifier, and is then displayed as a color-coded image.
A change in contact stiffness causes a shift of the contact resonance frequency and
hence a change of cantilever vibration amplitude in the same manner as the vertical
stiffness images, see above. Here also contrast inversion is observed. Cracks in a
nanocrystalline ZrO2 ceramic were generated by Vickers indents (Fig. 11). Fig. 11a
Figure 11. Topography (left) and lateral AFAM image (right) of a ZrO2 ceramic.
shows topography, Fig. 11b the corresponding torsional resonance images of the
identical surface areas. There are darker areas close to the crack flanks showing
regions of altered elastic properties. This could be caused for example by the crack
system below the surface and is examined further at present.
3. CONCLUSIONS
The AFAM and ultrasonic piezo-mode techniques described here lend itself for
studying materials properties in ceramics on a nanoscale. Whereas the various
imaging modes become more and more routine, the application of the AFM and
ultrasonic AFM technique to study crack propagation are still in the early stage.
However, the results presented here show that these techniques may provide much
wanted information on the elastic, piezoelectric, and fracture mechanical behavior of
ceramics during crack propagation.
4. ACKNOWLEDGEMENT
We gratefully acknowledge the financial support by the German Science Founda-

tion. One of us (M.K-M) was supported by the SFB 277 at the University of the
Saarland. Furthermore, it is a pleasure to thank D. Lupascu and J. Rödel for helpful
discussions.
REFERENCES
1. P. Maivald, H.T. Butt, S.A. Gould, C.B. Prater, B. Drake, J.A. Gurley, V.B. Elings, and P.K. Hansma,
Using Force Modulation to Image Surface Elasticities with the Atomic Force Microscope, Nanotech
nology 2, 103 106 (1991).
2. K. Yamanaka, H. Ogiso, and O. V. Kolosov, Ultrasonic force microscopy for nanometer resolution
subsurface imaging, Appl. Phys. Lett. 64, 178 180 (1994).
3. U. Rabe and W. Arnold, Acoustic Microscopy by Atomic Force Microscopy, in: Proc. 21st Int. Symp.
Acoustical Imaging, edited by J. P. Jones, Plenum Press, New York, (1995) 585 592.
10 W. Arnold et al.
4. U. Rabe, K. Janser, and W. Arnold, Acoustic Microscopy with Resolution in the nm Range, in: Proc.
24th Int. Symp. Acoustical Imaging, Eds. P. Tortoli, L. Masotti, Plenum Press, New York (1996)
669 676.
5. U. Rabe, K. Janser, and W. Arnold, Vibrations of Free and Surface Coupled Atomic Force Micro
scope Cantilevers: Theory and Experiment, Rev. Sci. Instrum. 67, 3281 3293 (1996).
6. P. Vairac and B. Cretin, Scanning microdeformation microscopy in reflection mode, Appl. Phys. Lett.
68, 461 463 (1996).
7. N.A. Burnham, G. Gremaud, A.J. Kulik, P.J. Gallo, and F. Oulevy, Scanning local acceleration
microscopy, J. Vac. Sci. Tech. B14, 794 799 (1996).
8. A. Rosa, E. Weilandt, S. Hild, and O. Marti, The simultaneous measurement of elastic, electrostatic
and adhesive properties by scanning force microscopy: pulsed force mode operation, Meas. Sci.
Technol. 8, 1333 1338 (1997).
9. E. Kester, U. Rabe, L. Presmanes, Ph. Tailhades, and W. Arnold, Measurement of Young’s Modulus
of Nanocrystalline Ferrites with Spinel Structures by Atomic Force Acoustic Microscopy J. Phys.
Chem. Solids 61, 1275 1284 (2000).
10. U. Rabe, S. Amelio, M. Kopycinska, S. Hirsekorn, M. Kempf, M. Göken, and W. Arnold, Imaging
and Measurement of Local Mechanical Material Properties by Atomic Force Acoustic Microscopy,
Interf. and Surf. Analysis 33, 65 70 (2002).
11. S. Amelio, A. V. Goldade, U. Rabe, V. Scherer, B. Bhushan, and W. Arnold, Measurements of
Elastic Properties of Ultra Thin Diamond Like Carbon Coatings using Atomic Force Acoustic
Microscopy, Thin Solid Films 392, 75 84 (2001).
12. M. Prasad, M. Kopycinska, U. Rabe, and W. Arnold, Measurement of Young’s Modulus of Clay
Minerals Using Atomic Force Acoustic Microscopy, Geophys. Res. Lett. 29, 13 16 (2002).
13. U. Rabe, S. Amelio, S. Hirsekorn, and W. Arnold, Imaging of Ferroelectric Domains by Atomic
Force Acoustic Microscopy, in: Proc. 25th Int. Symp. Acoustical Imaging, edited by M. Halliwell and
P.N.T. Wells, Kluwer Academic/Plenum Publishers, 253 260 (2000).
14. K. Yamanaka, Y. Maruyama, T. Tsuji, and K. Nakatomo, Resonance frequency and Q factor
mapping by ultrasonic atomic force microscopy, Appl. Phys. Lett. 78, 1939 1941 (2001).
15. A. L. Gruverman, J. Hatano, and H. Tokumoto, Scanning Force Microscopy Studies of Domain
Structure in BaTiO3 Single Crystals, Jap. J. Appl. Phys. 36, 2207 2211 (1997).
16. P. Güthner and K. Dransfeld, Local poling of ferroelectric polymers by scanning force microscopy,
Appl. Phys. Lett. 61, 1137 1139 (1992).
17. A. Gruverman, O. Auciello, and H. Tokumoto, Scanning force microscopy: application to nanoscale
studies of ferroelectric domains, Integrated Ferroelectrics 19, 49 83 (1998).
18. U. Rabe, M. Kopycinska, S. Hirsekorn, J. Muñoz Saldaña, G.A. Schneider, and W. Arnold, High
resolution characterisation of piezoelectric ceramics by ultrasonic scanning force microscopy tech
niques, J. Phys. D: Appl. Phys 35, 2621 2536 (2002).
19. K. Franke, J. Besold, W. Haessler, and C. Seegebarth, Modification and detection of domains on
ferro electric PZT films by scanning force microscopy, Surf. Sci. Lett. 302, 283 288 (1994).
20. L. M. Eng, H. J. Güntherodt, G. A. Schneider, U. Köpke, and J. Munoz Saldana, Nanoscale
reconstruction of surface crystallography from three dimensional polarization distribution in ferro
electric barium titanate ceramics, Appl. Phys. Lett. 74, 233 235 (1999).
21. S. Hong, J. Woo, H. Shin, J. U. Jeon, Y. E. Pak, E. L. Colla, N. Setter, E. Kim, K. No, Principle of
ferroelectric domain imaging using atomic force microscope, J. Appl. Phys. 89, 1377 1386 (2001).
22. V. Scherer, ‘‘Ultraschall Kraft Mikroskopie mit lateraler Anregung’’ Naturwisenschaftlich Technische
Fakultät III. PhD Thesis, University of the Saarland, (2002), unpublished.
23. K. L. Johnson, Contact Mechanics (Cambridge University Press, Cambridge, 1995).
24. G. M. Pharr, W. C. Oliver, F. Brotzen, On the generality of the relationship among contact stiffness,
contact area, and elastic modulus during indentation, J. Mat. Res. 7, 613 617 (1992).
25. K. Yamanaka, T. Tsuji, A. Noguchi, T. Koike, T. Mihara, Nanoscale elasticity measurement with in
situ tip shape estimation in atomic force microscopy, Rev. Sci. Instr. 71, 2403 2408 (2000).
26. M. Reinstaedtler, U. Rabe, V. Scherer, J.A. Turner, and W. Arnold, Imaging of flexural and
torsional resonance modes of atomic force microscopy cantilevers using optical interferometry,
Surface Science, (2003) in print
27. M. Labardi, V. Likodimos, and M. Allegrini, Force microscopy contrast mechanisms in ferroelectric
domain imaging, Phys. Rev. B 61, 14390 14398 (2000).
28. M. Kopycinska, U. Rabe, S. Hirsekorn, W. Arnold, Imaging of the Ferroelectric Domains Pattern in
the Ultrasonic Piezo Mode in: Proc. 26th Acoustical Imaging, edited by R. Maev Kluwer Academic
Publishers, New York, Acoustical Imaging, 26, 191 198 (2002).
29. M. Reinstädtler, U. Rabe, V. Scherer, U. Hartmann, A. Goldade, B. Bhushan, and W. Arnold, On
the nanoscale measurement of friction using atomic force microscope cantilever torsional resonances,
Appl. Phys. Lett. 82, 2604 2606 (2003).
30. W. Arnold, G. Weides, and S. Faßbender, Measurement of Elastic Properties Related to the
R Curve Behavior of Ceramics, in: Proc. of APCFS & ATEM ’01 (Asian Pacific Conference on
Fracture and Strength ’01 and International Conference on Advanced Technology in Experimental
Mechanics ’01), The Japan Society of Mechanical Engineers, Tokyo, 2001, pp 517 522.
31. T.J. Marrow, G.A.D. Briggs, and S.G. Roberts, In situ Acoustic Microscopy of Crack Bridging in
Alumina, J. Europ. Ceram. Soc. 14, 111 116 (1994).
32. B. Lawn, Fracture of Brittle Solids, in Cambridge Solid State Science Series (Cambridge University
Press, 1995) pp. 249 306.
33. J. Rödel, J.F. Kelly, and B.R. Lawn, ‘‘In situ Measurements of Bridging Crack Interfaces in the
Scanning Electron Microscope’’, J. Am. Soc. 73, 3313 3318 (1990).
INDENTATION SIZE EFFECT
ON THE HARDNESS OF
ZIRCONIA POLYCRYSTALS
Takashi Akatsu, Shingo Numata, Michiyuki Yoshida,

Yutaka Shinoda, and Fumihiro Wakai*
1. INTRODUCTION
The hardness of materials is one of the most important mechanical properties from
an engineering point of view. The hardening of metals is realized by the inhibition of
dislocation glide, which can be managed through microstructural design. The Hall-
Petch hardening H / d 1=2 , where H is the hardness and d is the grain size, is a well-
known and useful concept of the design. Conversely, the inverse Hall-Petch relation
is recently found out in a metal consisted of very fine grains.1,2 Also, it is revealed
that the H-value of metals does not depend on only d but indentation size, i.e. the
hardness increases with a decrease in indentation size. The indentation size depend-
ence is described through the building-up of geometrically necessary dislocations
under a concentrated stress field.3,4 In terms of ceramics, correlation between hard-
ness and microstructure has not been cleared, although a lot of research has been
done.5,6 For example, even the grain size dependence of the hardness has not been
generally described. The hardness of ceramic nanopolycrystals, which consists of
very fine grains with a diameter less than 1mm, has not been examined in detail.
Moreover, indentation size effect on the hardness has been scarcely reported. The
experimental and theoretical difficulty to estimate the plasticity of a hard elastic-
plastic material is one of the reasons for the obscurity in the hardness of ceramics.
The development and improvement of the nanoindentation technique helps us to
make an accurate evaluation of the plasticity of ceramics.
*
Takashi Akatsu, Shingo Numata, Michiyuki Yoshida, Yutaka Shinoda, and Fumihiro Wakai, Mater
ials and Structures Laboratory, Tokyo Institute of Technology, 4259 Nagatsuta, Midori, Yokohama
226 8503, Japan
13
14 Takashi Akatsu et al.
In this study, the nanoindentation on monoclinic-, tetragonal- and cubic-

zirconia polycrystals (referred as MZP, TZP and CZP, respectively) was carried
out with a home-made nanoindenter. Relationship between indentation load P and
penetration depth h of a diamond stylus was analyzed to derive the true hardness HT
as a function of the plastic depth hp , which is a measure of plasticity.7 The plastic
deformation of each zirconia at room temperature was discussed in detail through
the examination and the mutual comparison of indentation size dependence of HT .
2. EXPERIMENTAL
2.1. Samples
The zirconia polycrystals with different crystallographic structure (monoclinic, tetrag-

onal and cubic) were fabricated to be a fine structure. Three kinds of very fine zirconia
powders, TZ-0Y, TZ-3Y and TZ-8Y (Tosoh Co.ltd. Japan), were sintered with HIP
under 200MPa to make MZP, TZP and CZP, respectively.8 The average grain size of as-
sintered MZP, TZP and CZP, which was determined through SEM observation and the
intercept method, was 60, 60 and 110nm, respectively. In addition, heat treatment of
each zirconia was carried out to enlarge their grain size. Nanoindentation on the {100}
plane of a single crystal of cubic zirconia (YSZ(100), Shinkosha Co.ltd., Japan) was
also carried out in comparison. The mirror-finished surface of all samples was carefully
prepared by polishing with colloidal silica powders with a diameter less than 60 nm.
2.2. Nanoindentation
Precise h measurement is very important for the accurate evaluation of hardness

through the nanoindentation technique. External deformation due to the compliant
structure of a conventional nanoindenter must be eliminated from h measured, be-
cause h is usually given as the drive of a diamond stylus towards the surface of a
sample.9 Some calibrations to compensate the extrinsic deformation are convention-
ally carried out, but they are, in many cases, poor at reproducibility. That is one of the
reasons for the inaccurate hardness evaluation with the nanoindentation. In this study,
the accurate value of h was directly measured by using a home-made nanoindenter
with a unique system as shown in Fig.1. The difference between the drive and deflec-
tion of the beam fixed with the stylus (see Fig.1) definitely gives the accurate h without
doing any calibrations. The P-value was determined as a function of the deflection and
flexibility of the beam. In addition, vertical indentation within an error of 0:1 was
confirmed using a laser reflection displacement transducer (see Fig.1) and stages for
the tilt and horizontal rotation of a sample. The system shown in Fig.1 was made of
super invar alloy with a thermal expansion of 0:7 10 6 = K at room temperature to
decrease the error due to the temperature change during indentation. Nanoindenta-
tion at room temperature in air was actually carried out by using the nanoindenter
with the Berkovich-type diamond stylus at a penetration rate of 50nm/sec.
2.3. Analysis of P-h Curves
It is relatively hard to find ideal quadratic relation between P and h in the very
shallow indentation region. That is attributed to the blunt tip of a diamond stylus,
Capacitive DisplacementTransducer
Laser Reflection Displacement Transducer
hcap Dhlaser
h*cap
Laser
Sample Sample
Diamond Stylus Beam
Figure 1. A unique system of a home made nanoindenter to measure accurate P and h. The difference
between the drive Dhlaser and deflection Dhcap ( hcap hcap ) of the beam gives an accurate h. The P value is
given as a function of Dhcap and the flexibility of the beam.
which causes the overestimation of hardness in that region. We made a two-step

modification of P-h curves to avoid the overestimation at small h as follows: At first,
loading and unloading P-h curves were transferred by the distance Dh between the
real and ideal origins, which was determined by the extrapolation of linear relation
between P1=2 and h at relatively large h into h ¼ 0, in the positive direction of h (see
Fig. 2). Secondly, P-h data from h ¼ 0 to h > 2Dh for loading and from hr to
hr þ 2Dh for unloading were neglected, respectively, in the following analysis (see
Fig. 2). According to the Oliver and Pharr method,10 the Meyer hardness HM was
derived from the modified P-h curves as follows,
Pmax
HM ¼ (1)
ghc 2
1
h= P + hr
k2
1000
Penetration depth h /nm
2Δh
500 1
h= P
k1
>2Δh
0
Dh
Real origin
0 5 10 15
Ideal origin
1/2 1/2
Load P /N
Figure 2. Modification of P h curves to avoid the overestimation of hardness in the shallow indentation
region due to the blunt tip of a diamond stylus.
Pmax
hc ¼ hmax « (2)
S
where Pmax the maximum indentation load, hmax the maximum penetration depth of
the stylus, g the geometrical factor to be 24.5 for the Berkovich-type stylus, « the
constant to be 0.72 for conical stylus, and S the unloading slope at Pmax . An analysis
to derive HT and h 9p was also carried out for the modified p1=2 -h relationship as
follows,

k1 = g=gp 2
HT ¼ p 2 (3)
1 k1 =k2
s
Pg p 2
hp ¼ (4)
gHT
where k1 and k2 are the linear slopes of P1=2 -h relationship on loading and unloading,
respectively, and gp is the constant to be unity for the perfect plasticity.
2.4. Raman Spectroscopy

Highly concentrated stress field seems to be realized under an indentation. In the case
of TZP, stress-induced transformation from tetragonal into monoclinic phase can be
predicted to occur under the indentation. Raman spectroscopy by using a focused
ultraviolet laser with a diameter of 1mm was carried out at the bottom of an
indentation to confirm the phase transformation.
3. RESULTS AND DISCUSSION
3.1. Meyer Hardness of Zirconia Polycrystals

The hardness of materials is conventionally evaluated as HM . The HM -value is,
however, not a measure of the plasticity of ceramics,7 because it is definitely affected
by the elasticity, e.g. Young’s Modulus. Therefore, not only grain size dependence
but also indentation size dependence of the HM -value of MZP, TZP and CZP was
obscure as reported in previous studies.5,6 In this study, the evaluation of HT , a
measure of plasticity, for each zirconia was carried out, and then revealed the
following indentation size dependence.
3.2. True Hardness of CZP
The HT -value of CZP increased with a decrease in hp as shown in Fig.3. Also,

the grain size dependence of HT was hardly observed in CZP. Moreover, the HT -
value of a cubic zirconia single crystal as a function of hp was superimposed just
on that of CZP (see Fig.3). According to the theory of the strain gradient plasticity,3,4
in which the building-up of geometrically necessary dislocations under an indenta-
tion was modeled, the indentation size dependence of hardness is represented as
follows,
80
m=1
H0=24.7GPa
A=446nm
True hardness HT /GPa

60
40
d=110nm
20 d=290nm
d=4540nm
Single crystal{100}
0 500 1000 1500

Plastic depth hp /nm
Figure 3. The HT value of CZP as a function of hp . d is the average grain size of CZP. The broken line is
drawn by using Eq.(5) with m 1.
m
HT A
¼1þ (5)
H0 hp
where H0 is the hardness evaluated at hp ! 1, A is the constant associated with the

material characteristic length such as the Burgers vector of the dislocation, and m is
the constant to be 1 or 2. The hp dependence of HT shown in Fig.3 was well described
by using Eq.(5) with m ¼ 1. Good agreement between the experimental and theoret-
ical hp dependence of HT regardless of grain size indicates that the plastic deform-
ation of CZP at room temperature is simply due to the accumulation of dislocation
glide in each grain. The dislocations seem to be piled-up at grain boundary but are
not able to go through the boundary into the neighboring grains such as those in
metals because of the lack of the number of active slip system. In other words, the
contribution of grain boundary sliding to the plasticity of CZP seems to be negligibly
small.
3.3. True Hardness of MZP

The HT -value of MZP, which was much smaller than that of CZP, scarcely depended
on hp as shown in Fig.4. In addition, the HT -value of MZP remarkably increased
with a decrease in the grain size below 100 nm (see Fig.4). The softness of MZP at
room temperature is attributed to the plastic deformation caused by the twinning of
the monoclinic crystallographic structure.11 Twinning is quite different from disloca-
tion glide observed in CZP, which is obviously associated with the strain gradient
plasticity. That is the reason why hp dependence of the HT -value of MZP was hardly
observed. In terms of metals, the resistance to plastic deformation due to twinning
increases with a decrease in grain size such as the Hall-Petch hardening.12 Correlation
between the twinning and grain size has not been cleared in ceramics. However, the
resistance to the twinning of monoclinic zirconia may increase with a decrease in
grain size comparably with the size of the twin, which is of the order of 10 nm.11
50
d=60nm
d=140nm
40 d=410nm

30
20
10
0 500 1000 1500 2000

Figure 4. The HT value of MZP as a function of hp . d is the average grain size of MZP.
3.4. True Hardness of TZP
The HT -value and its hp dependence of TZP were classified into two categories as a
function of grain size. If the grains were larger than 100nm, the HT -value of TZP was
the same as that of MZP and scarcely depended on hp (see Fig.5). On the contrary,
the HT -value of TZP increased as large as that of CZP and significantly increased
with a decrease in hp if the grain size decreased below 100nm (see Fig.5). As a result of
Raman spectroscopy at the bottom of an indentation, the monoclinic phase was
observed in the former though only tetragonal phase was detected in the latter. This
indicates that the stress-induced transformation from tetragonal to monoclinic phase
occurred on loading in the former while it was inhibited in the latter. Therefore, the
HT -value of TZP consisted of grains with a diameter larger than the critical size of
the transformation coincided with that of MZP because of the plastic deformation of
the transformed monoclinic phase as mentioned above. The decrease of grain size
below the critical realizes the stabilization of tetragonal phase.13 That is the reason
why the HT -value and its hp dependence of TZP consisted of grains with a diameter
less than 100 nm was the same as those of CZP.
4. CONCLUSION
The HT -value of zirconia polycrystals increased with a decrease in hp . This is

attributed to the building-up of geometrically necessary dislocations under an inden-
tation. In addition, the grain size dependence of HT was hardly observed. These
results indicate that the plastic deformation of zirconia polycrystals at room tem-
perature is basically due to the accumulation of dislocation glide in each grain. In
other words, the contribution of grain boundary sliding to the plasticity was negli-
gibly small even if the grains were of the order of 100nm. The HT -value of MZP,
which was much smaller than that of CZP, scarcely depended on hp because of the
plastic deformation due to twinning. The HT -value and its hp dependence of TZP
80
d=60nm
d=100nm
d=300nm

60 CZP(d=110nm)
MZP(d=140nm)
40
20
0 500 1000 1500 2000

Figure 5. The HT value of TZP as a function of hp . d is the average grain size of zirconia polycrystals.
T
T T
T
T
A
Intensity /a.u.
d=60nm B
B
A
M
M
A
d=100nm
B
100 200 300 400 500 600 700
−1
Raman shift /cm
Figure 6. The spectrum of Raman spectroscopy measured at the bottom A of an indentation on TZP. B is
the place far from an indentation to obtain the reference spectrum. M and T are assigned to monoclinic and
tetragonal phases, respectively. d is the average grain size of TZP.
changed depending on whether the stress-induced phase transformation occurred or

not under an indentation, which was determined as a function of grain size.
ACKNOWLEDGEMENT
A part of this research work was supported by the Nippon Sheet Glass Foundation
for Materials Science and Engineering and the Grant-in-Aids for Scientific Research
(#13750622) of the Ministry of Education, Culture, Sports, Science and Technology,
Japan. The authors are grateful to Mr.Y.Teranishi for helping us to carry out Raman
spectroscopy.
REFERENCES
1. S.Takeuchi, The Mechanism of the Inverse Hall Petch Relation of Nanocrystals, Scripta Mater. 44,
1483 1487 (2001).
2. C.A.Schuh, T.G.Nieh, T.Yamasaki, Hall Petch Breakdown Manifested in abrasive wear resistance of
Nanocrystalline nickel, Scripta Mater. 46, 735 740 (2002).
3. W.D.Nix and H.Gao, Indentation Size Effects in Crystalline Materials: A Law for Strain Gradient
Plasticity, J.Mech.Phys.Solids 46[3], 411 425 (1998).
4. M.R.Begley and J.W.Hutchinson, The Mechanics of Size Dependent Indentation, J.Mech.Phys.
Solids 46[10], 2049 2068 (1998).
5. R.W.Rice, C.C.Wu, and F.Borchelt, Hardness Grain Size Relations in Ceramics, J.Am.Ceram.Soc.
77[10], 2539 2553 (1994).
6. A.Krell, A New Look at the Influences of Load, Grain Size, and Grain Boundaries on the Room
Temperature Hardness of Ceramics, Int’l.J.Ref.Met.Hard Mater. 16, 331 335 (1998).
7. M.Sakai, The Meyer Hardness: A Measure for Plasticity?, J.Mater.Res. 14[9], 3630 3639 (1999).
8. M.Yoshida, Y.Shinoda, T.Akatsu, and F.Wakai, Deformation of Monoclinic ZrO2 Polycrystals and
Y2 O3 Stabilized Tetragonal ZrO2 Polycrystals Below the Monoclinic Tetragonal Transition Tem
perature, J.Am.Ceram.Soc. 85[11], 2834 2836 (2002).
9. M.Sakai, S.Shimizu, and T.Ishikawa, The Indentation Load Depth Curve of Ceramics, J.Mater.Res.
14[4], 1471 1484 (1999).
10. W.C.Oliver and G.M.Pharr, An Improved Technique for Determining Hardness and Elastic Modu
lus Using Load and Displacement Sensing Indentation Experiments, J.Mater.Res. 7[6], 1564 1583
(1992).
11. I W.Chen, Implication of Transformation Plasticity in ZrO2 Containing Ceramics: II, Elastic Plastic
Indentation, J.Am.Ceram.Soc. 69[3], 189 194 (1986).
12. D.Hull, Effect of Grain Size and Temperature on Slip, Twinning and Fracture in 3% Silicon Iron,
Acta Metall. 9[3], 191 204 (1961).
13. F.F.Lange, Transformation Toughening Part 1 Size Effects Associated with the Thermodynamics of
Constrained Transformations, J.Mater.Sci. 17, 225 234 (1982).
INDENTATION FRACTURE,
ACOUSTIC EMISSION AND
MODELLING OF THE
MECHANICAL PROPERTIES
OF THIN CERAMIC
COATINGS
S.J. Bull, I. Arce-Garcia, E.G.-Berasetegui and T.F. Pagey
ABSTRACT
Fracture is an almost inevitable consequence of highly loaded contacts in ceramic-

coated systems. For relatively thick coatings fracture is often similar to that observed
in bulk samples of the coating but as the coating thickness is reduced the substrate
plays an increasing role in influencing or even controlling fracture behaviour. Both
through-thickness and interfacial fracture may be observed depending on the relative
toughness of the substrate, coating and interface. Through-thickness fracture is
exacerbated by plastic deformation in the substrate and therefore the load support
from the substrate is critically important in determining the type and extent of
fracture observed. In this paper, nanoindentation testing and post facto atomic
force and scanning electron microscopy is used to characterise the types of fracture
which occur for hard coatings on softer substrates and multilayer coatings on glass.
The effect of fracture on the nanoindentation load-displacement curves and the
hardness and Young’s Modulus values obtained from them is discussed and a simple
model to account for the observed behaviour is introduced. The use of acoustic
y
School of Chemical Engineering and Advanced Materials, University of Newcastle, Newcastle upon
Tyne, NE1 7RU, UK
21
22 S.J. Bull et al.
emission to monitor plasticity and fracture during the indentation cycle is also
discussed.
1. INTRODUCTION
The durability and performance of coatings in almost every application depend on

their as-deposited mechanical properties such as adhesion, hardness, toughness and
elastic modulus. These properties may not be the same as those of the same material in
bulk form due to different microstructural and defect states arising from the depos-
ition process. As the coating thickness is reduced it becomes increasingly difficult to
measure these properties by conventional methods and low load nanoindentation and
scratch testing have been developed to enable measurements to be achieved at scales
commensurate with the coating thickness. Continuously recording indentation test
methods are well established for the determination of Elastic Modulus and hard-
ness.1,2,3 but methods for assessment of fracture toughness and adhesion are much
less developed.4,5 In part the reason for this is that mechanism of fracture failure around
an indentation is complex and depends on the relative properties of coating, substrate
and interface and therefore a universal analysis method is unlikely to be produced.
However, in some cases it is possible to identify failure modes that are amenable to
analysis resulting in reasonable toughness data. Microscopy and analysis of fracture
paths is a key part of this approach, particularly as the indentations get smaller.
There are a number of different indentation fracture events which might occur in
a coating substrate system (Table 1), the occurrence of which will depend on the
relative toughness of coating/substrate and any interfaces present. These failure
modes will be altered by plastic deformation in the substrate which can lead to the
superposition of bending stresses on to those generated by the indentation.
This paper discusses the type of fracture events which occur for hard coatings on
soft and hard substrates and outlines the effect that fracture events have on other
mechanical properties measured by indentation (such as hardness and Young’s
Modulus). It also introduces the use of acoustic emission generation to monitor
fracture events in very thin coatings where it is difficult to identify them by conven-
tional microscopy (both scanning electron and atomic force).
2. EXPERIMENTAL
2.1. Sample Preparation
In this study thin ceramic coatings have been deposited onto both brittle and ductile
substrates to assess the fracture behaviour of the coating/substrate system. Niobium
Table 1. Fracture events associated with brittle (ceramic) coating/substrate systems.
Substrate Coating Fracture type
Ductile Brittle Through thickness fracture, brittle fracture in coating parallel

to interface, ductile interfacial fracture, microfracture in coating
Brittle Brittle Bulk chipping, through thickness fracture, brittle interfacial

fracture, microfracture in coating, microfracture in substrate
Indentation Fracture, Acoustic Emission and Modelling 23
nitride coatings were deposited in the same process run on a fully hardened powder
metallurgy high speed steel (ASP23) and annealed stainless steel (304) substrates by
cathodic arc evaporation using standard commercial deposition parameters.6 The
substrates were polished to a 3 mm diamond finish prior to coating and a layer of
approximately 4 mm was deposited on each. Fullerene-like CNx was deposited on a
range of substrates at thicknesses of 400 nm and 1 mm using unbalanced magnetron
sputtering.7
A range of oxide and oxynitride coatings was deposited onto float glass using
magnetron sputtering to 200 nm thickness. In addition a multilayer stack with a
coating architecture similar to current commercial low emissivity solar control
coatings was deposited to investigate the effect of internal interfaces in the coating.
Details of the coatings deposited are presented in Table 2.
2.2. Nanoindentation Testing

Nanoindentation experiments were performed using two different nanoindentation
systems to make indents at peak loads from 100 mN to 500 mN. Hardness and
Young’s Modulus were extracted from the load-displacment curves using the method
of Oliver and Pharr.8 All load displacement curves were corrected for thermal drift and
the indenter area functions were carefully calibrated following the Oliver and Pharr
approach using a fused silica standard sample prior to testing the coated samples.
A Nanoindenter II with a ‘blunt’ Berkovich tip (tip end radius 250 nm at the
time of making the experiments) was used for testing the NbN coated samples and
the CNx coated samples at loads in the complete range. Due to the fact that the
Table 2. Coating/substrate systems investigated in this study.
Substrate Deposition
Hardness Thickness Deposition Temperature
Coating Substrate (GPa) (nm) Technology (8C)
NbN ASP23 steel 9.5 3980 Arc evaporation 425
NbN 304 stainless steel 1.9 2810 Arc evaporation 425
CNx Silicon 11.5 400, 1000 Unbalanced magnetron 350

sputtering
CNx 1 mm SiC 35.0 SiC 400, 1000 Unbalanced magnetron 350

on silicon 11.5 Si sputtering
CNx Sapphire 32 400, 1000 Unbalanced magnetron 350

sputtering
TiOxNy Float Glass 6.5 200 Magnetron sputtering 200
SnO2 Float glass 6.5 200 Magnetron sputtering 200
Zr (O) Float glass 6.5 200 Magnetron sputtering 200
ZnO Float glass 6.5 200 Magnetron sputtering 200
Oxide Float glass 6.5 Total Magnetron sputtering 200

multi layer thickness 80 nm
24 S.J. Bull et al.
indenter is relatively blunt this approach is not suitable for measuring the hardness
and Young’s Modulus for the CNx coatings as they do not plastically deform before
the substrate.9 However, the substrate plastic deformation is limited at the higher
loads where coating fracture and interfacial detachment occur and this can be useful
for interface toughness determination. The constant loading rates used were
500 mN=s for indentations to maximum loads greater than 100 mN and 50 mN=s
for lower peak load indentations.
A Hysitron Trinboindenter fitted with a sharp Berkovich tip (tip end radius
50 nm) was used to test the CNx and coatings on glass. A fixed loading rate of
100 mN=s was used in all cases to minimise the effects of creep on the measured load-
displacement curves. This system can use the tip which is used to make the indent to
perform AFM imaging of the impression after the indentation cycle is complete.
Given the small size of the impressions made at low loads this is the only practical
way to achieve imaging of the impressions as the indent arrays are impossible to find
using off-line AFM or SEM even when precision locating stages are used.
Nanoindentations were also performed on the single and multi-layer coated
glass with the Hysitron Triboindenter fitted with a 1208 conical indenter (blunt tip,
radius 500 nm) instrumented for the recording of acoustic emission (AE) during
the indentation cycle.10 Loads from 1 mN to 10 mN were used in this analysis. AE
events were sampled at 4MHz; once a threshold signal is exceeded (determined from
noise measurements with the indenter in air) the system records the AE signal
waveform for subsequent analysis (Figure 1).
The energy in the acoustic event is then calculated from integration of the
amplitude of the AE waveform, A(t), with respect to time and is given by11:
ðt
EA (t) ¼ A(t)2 dt (1)
0
To calculate the net energy of the event, the linear background contribution (at) is
subtracted (Figure 2), assuming that the amplitude of the background contribution is
virtually constant during tests and has a mean amplitude a:
EANet (t) ¼ EA (t) at (2)
1500
Trigger event
1000
AE Amplitute
500
−500
−1000
0 0.04 0.08 0.12
Time, t (ms)
Figure 1. AE amplitude as a function of time after an acoustic emission event which occurred at 4000 mN
load in the testing of a 200 nm thick TiOxNy coating on float glass.
4
Background
3 Contribution
Raw Energy
Energy detargent 2
0 Net Energy
−1
0 0.02 0.04 0.06 0.08 0.1 0.12
Time (ms)
Figure 2. An example of the energy calculation for the event in Figure 1. The net energy is calculated by
subtracting background contribution from the raw energy.
The constant a is determined for each acoustic event by fitting the background
integrated energy a long time after the event as shown in Figure 2. It is assumed
that the background is linear which is reasonable over the AE pulse duration.
To discriminate between different events, not only the shape of the wave can be
analysed but also the parameter emission shape factor, defined as:
Amax
AESF ¼ (3)
ARMS
Previous studies have demonstrated that AESF differentiates between different types
of acoustic emission generation process.11,12 In addition, the attenuation time of the
signal, tat , may be used to differentiate between events.
2.3. Microscopy
Indentations produced at high loads using the Nanoindenter II could be located for
post facto imaging and were imaged by scanning electron microscopy (SEM) and
atomic force microscopy (AFM). Cracks were easily visible in the SEM and their
visibility was enhanced by using backscattered electron imaging. It proved to be
almost impossible to see cracks in the AFM images unless they displayed significant
residual (opening) displacements. This is probably due to the fact that water from the
environment is drawn into the fine cracks by capillary action and the AFM cantilever
rides over the open crack on the surface of a water film.
3.1. Ceramic Coating on Ductile Substrate: NbN on Steel
No evidence for interfacial detachment was found for any of the NbN coatings
investigated in this study but through thickness fracture was observed (Figure 3).
The type of fracture observed depends on the hardness of the substrate. For both
types of coating fracture initially occurs in the radial direction following the edges of
26 S.J. Bull et al.
Figure 3. Scanning electron micrographs of 500 mN indents in NbN coated (a) 304 stainless steel and
(b) ASP23 high speed steel.
the pyramidal indenter. For the harder ASP23 substrate the only other fracture
visible is picture frame cracking within the indent (Figure 3b). However, the coating
on the stainless steel substrate shows a circular through-thickness crack at the edge of
the contact (Figure 3a) which is some distance from the well defined indent at the
centre of the circle. Radial cracks defined by the indenter edges connect this central
indent to the edge of the contact but do not extend beyond it.
The difference in behaviour can be understood from the load-displacement
curves. Figure 4 shows a comparison between the load-displacement curves for the
NbN coatings on the two substrates tested at 500 mN. At low loads the load-displace-
ment curves superpose indicating that it is the coating properties which dominate the
measured response and the load is proportional to displacement squared as expected
for elastic-plastic indentation of a coating where the substrate provides good load
support. However, when the contact-induced shear stresses rise to trigger plastic
defomation in the substrate below the coating,13 plastic deformation starts in the
stainless steel substrate the substrate collapses beneath the coating which consequently
has to elastically bend to follow the deforming substrate. There is a change in slope of
4mm NbN on Different Steels

600
ASP23
500
304 Stainless
400
Load (mN)
300
200
100
0
0 500 1000 1500 2000 2500 3000
Displacement (nm)
Figure 4. Load displacement curves for NbN coatings on different steel substrates.
the load-displacement curve at this point. The size of the central indent is the same as
the size of an indent with peak load 100 mN (about 400 nm deep) which represents the
size of the indent in the coating at the onset of plastic deformation in the stainless steel
substrate. The impression caused by plastic deformation of the NbN coating stops
growing at this point and further plastic deformation is concentrated in the substrate.
The coating remains elastic but is bent into the impression created by substrate
deformation. To accommodate this deformation the radial cracks following the in-
denter edges extend as the plastic zone in the substrate extends. Tensile radial stresses
in the coating at the edge of the substrate plastic zone, in combination with these
bending stresses lead to the single circular through-thickness crack under such
circumstances the displacement of the indenter increases greatly as the disc of coating
is separated from the rest of the coated surface and membrane stresses supporting
some of the load drop to zero.
The fact that the load displacement curve shows no discrete discontinuities
(‘‘pop-ins’’ implies that there is a gradual extension of the radial cracks as the load
is increases and the through thickness ring crack occurs at close to the peak load. In
the absence of a recognisable pop-in event or a clear idea of the normal load at which
fracture occurs it is not possible to analyse the fracture toughness of the coating from
the available data.
For thinner coatings, through-thickness cracking may be exacerbated by pile-up
of substrate material, but pile-up is suppressed by the 4 mm NbN coatings investi-
gated here.
The fracture which occurs for the stainless steel substrate has a measurable effect
on the hardness and Young’s Modulus measured for the coating and must be taken
into consideration if the performance of the coated system is to be successfully
modelled. Recently some success in predicting the hardness of a coating substrate
system as a function of contact scale has been achieved using an energy based model
in which the contributions to the total work of indentation from coating, substrate
and interfaces can be summed and divided by the total deforming volume to deter-
mine the system hardness.14,15 The total deforming volume is assumed to be the
plastic zone associated with the impression and does not take into consideration any
plastic deformation associated with cracking which may occur outside this. From the
SEM images in Figure 3, where extensive cracking is not observed, this assumption is
reasonable. Two interfaces must be considered, thus

Vs H0 (s) þ Vf Ho ( f ) þ As gs þ Ai gi
H¼ (4)
V
where Vf and Vs are the deforming volumes of coating and substrate, H0 (f) and H0 (s)
are the bulk hardnesses of coating and substrate, As is the surface area and Ai is the
interfacial area within the plastic deforming region. The total deforming volume is
assumed to be hemispherical and the deforming volumes in coating and substrate are
given by appropriate slices through the hemisphere. The surface energy terms can be
used to account for the fact that the real deforming volume is unlikely to be hemispher-
ical. The radius of the deforming volume is related to the depth of the impression by16
1=2
E
R ¼ k2 d cot1=3 c (5)
H
28 S.J. Bull et al.
where k2 is a geometrical constant (k2 ¼ 3:64 for a Berkovich indenter), d is the

plastic indentation depth, E is the Young’s Modulus of the material and c is
the effective indenter angle (c ¼ 70:38 for a Berkovich indenter).
To include the effects of fracture in the model is relatively straightforward if we
assume that the coating within the cracked region does not contribute to the plastic
work of indentation after fracture occurs. Fracture usually occurs when the plastic
zone in the substrate is very well-established so we can consider the volume of coating
to be a cylindrical disc with the thickness of the coating. The area of the crack
produced is small compared to the volume it encloses so we also assume that the
surface energy of the crack plays only a minor part in the work of indentation. For a
radial cracked system the cracks have radius kc d where kc is a constant and d is the
plastic indentation depth. These form above a critical depth, dc . The area of material
bounded by radial cracks increases linearly with indentation size. In this case the
correction for the modelled hardness is given by

3 tH0F þ g s þ gi k2c d2
DH ¼
2R3
(6)
3tEf k2c d2
DE ¼
2R3
A simple estimate for kc is that it relates the depth to the radius of an equivalent cone
with the same volume as the impression. Thus, for a Berkovich indentation
r
24:5
pr ¼ 24:5d and therefore kc ¼
2 2
(7)
p
NbN on 304 Stainless Steel

50
Measured Hardness
40 Work Model
Work model with radial fracture
Hardness (GPa)
30
20
10
0
0 500 1000 1500 2000
Depth (nm)
Figure 5. Hardness of NbN on stainless steel as a function of contact depth comparing modelled data with
experiment. Correction for fracture is necessary to achieve a good fit. (Modelling parameters; Hf 24 GPa,
Ef 475 GPa, gf 15000J=m2 , H2 1:7 GPa, Es 210 GPa, gi 5000J=m2 and a 2:0).
Corrected hardness predictions are shown in Figure 5 compared to experimental

data. The quality of the fit is excellent (correlating coefficient >0.98) and much better
than can be achieved by any other hardness modelling approach investigated in this
study. An advantage of the model compared to existing hardness models is that it can
be used to make predictions of the behaviour of different coating/substrate systems
based on only a few scale-corrected physically-meaningful materials constants.14
3.2. Brittle Coating on Brittle Substrate: CNx on Silicon, Silicon Carbide

and Alumina
For brittle coatings on brittle substrates both through-thickness and interfacial
fracture can occur depending on the relative toughness of the coating, substrate
and interface. In the case of CNx coatings the coating and substrate toughness are
usually greater than the interface toughness and so detachment of the coating can be
produced by high load nanoindenation.
Loading a pyramidal indenter onto the surface of a coated sample creates
compressive stress in the coating (see Figure 6). Such stresses can cause the coating
to buckle and detach from the substrate around the indenter (see SEM micrograph in
Figure 6a). According to Thouless,17 once buckling starts the mode mixity of the
stress at the interfacial crack tip changes in such a way that the interfacial crack is
diverted to the surface forming a chip. This is registered in the load-displacement
curve, where a long step can be observed in its loading portion as the indenter
advances into the sample (see Figure 6b). In this case the buckled material has failed
catastrophically and the coating has been ejected from beneath the indenter since the
pop-in distance is the thickness of the coating.
If the coating has not undergone catastrophic fracture during loading, on
unloading the coating is still under compressive stress and around the indenter it is
already detached from the substrate (see Figure 7a). Towards the end of unloading
the coating buckles beneath the indenter pushing it up,5 which is associated with the
linear unloading observed in the P-d curve (see Figure 8a). During buckling the
outside of the coating is under tensile stress (see Figure 7b) and fracture may initiate
where the buckle curvature is highest (Figure 7c). This behaviour is observed for
500
A B.
400 Coating
fracture
Load, mN
300
200
100
0
0 500 1000 1500 2000 2500
(a) (b) Displacement, nm
Figure 6. (a) SEM micrograph showing coating detachment during loading of the indenter for a 1 mm
thick CNx coating on 3C SiC (001) substrate. (b) Step (A B) in the loading curve due to through thickness
cracking of the coating and spalling.
30 S.J. Bull et al.
Compressive Compressive
stress stress
(a)
Detached region
Coating fracture
Tensile stress
(b) (c)
Figure 7. Schematic diagram of the different steps followed by the coating fracture during unloading of
the indenter: (a) Coating under compressive stress and detached region (b) Buckling and tensile stress, and
(b) fracture of the coating.
Fracture during loading
500
400
Load (mN)
300 Linear
unloading
200
100
0
0 400 800 1200 1600
(a) Fracture during unloading (b) Displacement
Figure 8. (a) SEM micrograph for a 1 mm thick CNx coating on Si (001) showing coating fracture during
loading and unloading of the indenter at 500 mN applied load. (b) Corresponding P d curve showing linear
unloading at the bottom of the unloading curve associated with buckling of the coating.
1 mm thick CNx coatings on Si(001) substrate at 500 mN applied load (Figure 8b). In
this case fracture occurs both on loading and unloading.
In order that coating material be ejected from beneath the indenter, radial
fracture must occur along the indenter edges such that the coating may be pushed
laterally from the contact as the load increases. In such circumstances the size of
the pop-in is dictated more by coating thickness and the work of indentation
associated with the pop-in event is a complex convolution of the energy needed to
slide the coating beneath the indenter and the coating fracture energy. Any fracture
toughness measurements based on this failure mode therefore need to be treated with
caution.
3.2.1. Measurements of Fracture Toughness and Interface Toughness
Since both through thickness and interfacial failure occurs for CNx coatings on all of
the substrates investigated here it is possible to estimate the fracture toughness of
coating and interface. However, it must be born in mind that all current analysis
methods based on indentation provide only a qualitative assessment and more
research is needed to obtain reliable quantitative results.4
3.2.1a. Fracture Toughness of the Coating from Chipping

The nanoindentation fracture toughness of CNx films can be estimated by different
methods from the literature, the common basis of all of them being the calculation of
the energy dissipated in the fracture of the coating. This energy (Ufr ) is calculated by
determining the energy dissipation during chipping from the step observed in the
loading part of the load-displacement curve when through-thickness fracture of the
coating occurs during indentation.
Three main methods have been used in this study:-
(a) The method proposed by Li, Diao and Bhushan18,19 for which the fracture
toughness of the coating is given by,

12
E U
KIC ¼ (8)
(1 n2 )2pCR t
where KIC is the critical stress-intensity factor or the fracture toughness of the
coating; E is the Young’s modulus of the coating,; n is the Poisson’s ratio of
the coating; CR is the crack length,; U is the strain energy difference before and
after cracking; and t is the thickness of the coating. In this case the loading curve is
extrapolated from the start of the discontinuity to the depth of the end of the
discontinuity and U is given by the difference between the extrapolated and measured
curves. Toonder et al.4 introduced a slight variation to this expression to take into
account the number of chipped areas (N; generally 3 when using a Berkovich
indenter) and account for the fact that the crack does not propagate perpendicularly
to the substrate but at an angle d. Therefore it is necessary to consider the effective
coating thickness (t ¼ t=sin d; where d is the average angle of the chipping edge, see
Figure 9) instead of the actual thickness of the coating, t. This leads to the following
expression for the fracture toughness of the coating,

12
E U
KIC ¼ (9)
N(1 n2 )2pCR t0
(b) The method of Toonder et al.4 based on the calculation of the fracture energy,
G, from the energy dissipated during chipping, Ufr , and the area of the crack
formed in the coating by chipping. In this case the total irreversible work, Wir , is
determined by measuring the area between the loading and unloading curves
and determining the difference before and after chipping by extrapolating the
variation of Wir with load for chipped and unchipped samples. The fracture
energy is then given by
32 S.J. Bull et al.
S4500i153Profile
500
Height (nm)
−500 δ
−1000
−1500
−2000
0 10 20 30 40 50 60
(a) Distance (micron)
1micron CNx on AI2O3 P=500N i7

200
0
Height (nm)
−200
−400 δ
−600
−800
−1000
20 25 30 35 40 45 50 55
Distance (micron)
(b)
Figure 9. SEM micrographs and corresponding profiles of a 500 mN applied load on 1 mm CNx coating
deposited on (a) Si(001) and (b) Al2 O3 showing through thickness fracture and chipping of the coating at
an angle, d.
Ufr
G¼ (10)
N2pt0 CR
The fracture toughness of the coating is then calculated by its general equation,
r
p EUfr
KIC ¼ EG ¼ (11)
N2pt0 CR
where E is the elastic modulus of the coating and the other factors have been defined
previously.
(c) Finally, a lower and an upper limit for the fracture toughness of the coating can
also be determined by analysis of the load-displacement curve itself at the
moment of fracture.4 An upper and lower limit for the energy dissipated during
chipping of the coating can be calculated by taking into consideration the
following assumptions: (i) fully elastic (for the lower limit) or plastic (for the
upper limit) behaviour of the material before and after fracture, (ii) P is pro-
portional to d2 during loading and unloading20. This would lead to the follow-
ing limits for Ufr ,
2
Pfr (df d0 ) # Ufr # Pfr (df d0 ) (12)
3
where Pfr is the load at which the fracture of the coating is observed and d0 and df
are the indentation depths before and after the fracture of the coating. The limits for
the fracture toughness are calculated by substituting these limits in equations 10
and 11.
Table 3 shows the values for the fracture toughness of CNx deposited on
different substrates for the methods previously presented. There is reasonable agree-
ment between all the methods in the values of the fracture toughness of CNx except in
the case of the original uncorrected expression given by Li et al.18, 19 This is due to the
fact that for these samples through-thickness of fracture of the coating does not
happen perpendicularly to the substrate surface but at a certain angle, as SEM
micrographs and optical profilometry show in Figure 9.
3.2.1b. Fracture Toughness of the Interface
Two different expressions have been used to estimate the fracture toughness of the
interface between the CNx coating and different substrates. The first one is that
developed by Malzbender and de With21 which takes into account the residual stress
and the elastic properties of the substrate and is applied to the spalling of coatings
during scratch or indentation testing. For this method, the interfacial fracture
toughness for a coating possessing residual stress is given by,
v !2
u r
u p p t 52 s2res
Kint ¼
s t 0:397 sin b Eint E 2 þ t
Eint
(13)
L 2 E
where Ksint is the interfacial fracture toughness; b is the semi angle between two
cracks (for a Berkovich indenter b ¼ 60 ); t is the thickness of the coating; L is the
crack length (see Figure 10); sres is the residual stress in the coating; E is the Young’s
modulus of the coating,; and E int is the Young’s modulus of the interface, which can
be calculated from the following equation assuming equal contributions from coating
and substrate,22

1 1 1 1
¼ þ (14)
Eint 2 E Es
where, E and Es are the Young’s modulus of the coating and substrate, respectively.
Table 3. Comparison between the fracture toughness of 1 mm thick CNx coatings deposited
on different substrates based on the energy dissipated during chipping of the coating.
p
KIC , MPa m
Bhushan Bhushan Toonder Lower and Upper limits

Substrate et al.(1997) corrected et al.(2002) Toonder et al. (2002)
Si(001) 6.33 3.36 2.90 5.30 6.50
Al2 O3 5.10 2.82 2.44 2.53 3.10
3C SiC(001) 17.70 6.37 5.51 5.50 6.73
3C SiC(111) 13.80 5.66 4.90 5.68 6.95

34 S.J. Bull et al.
Radial crack
β
L
Chipped coating
segment
2a
Figure 10. Schematic diagram of the geometry of a chipped segment of coating.
The other expression used is that proposed by Rosenfeld et al.23 in which the
interfacial fracture toughness is related to the size of the delaminated area and the
load at which delamination of the coating occurs and is given by,
p
0:792H t(1 n2 )
KIC ¼ (15)
2(1 n)HCR2
1þnþ
Pdel
where H is the hardness of the coating; t is the coating thickness; n is the Poisson’s
ratio of the coating; CR is the length of the delamination crack; and Pdel is the load at
which delamination occurs.
The results obtained are presented in Table 4. The residual stress in the CNx
coatings deposited on Si(001) has been calculated from the change in radial crack
length around Vickers indentations after coating using the following expression
introduced by Lawn and Fuller,24
32 1
C0 2st 2
1 ¼ (16)
C Kc
where C0 and C are the crack length in the substrate and coated sample, respectively;
t is the coating thickness; s is the residual stress in the coating; and Kc is the fracture
toughness of the substrate calculated by,25
Table 4. Values of the interfacial fracture toughness for CNx

coatings (1 mm thick) deposited on different samples.
p
KIC , MPa m
Substrate Rosenfeld et al. (1990) Malzbender & de With (2002)
Si(001) 5.92 0.75
Al2 O3 4.67 0.68
3C SiC(001) 5.51 0.40
3C SiC(111) 5.58 0.41

12
E
Kc ¼ 0:016 PC 3=2
(17)
H
where E is the Young’s modulus; H is the hardness; P is the applied load; C is the
crack length in the uncoated sample. The value for the compressive residual stress in
1 mm thick CNx coating deposited on Si(001) substrate is 0.37GPa. This value has
also been considered for 1 mm thick CNx coatings deposited on 3C SiC(001) and 3C
SiC(111) since the thermal expansion coefficients for 3C SiC are similar to those for
Si(001). However, Al2 O3 substrate presents higher thermal expansion coefficients
than Si(001) and therefore the residual stress in CNx coatings deposited on this
substrate is around 0.6GPa.
The values for the fracture toughness of the interface obtained by the two
methods differ quite significantly (Table 4) although the trends for the different
substrates are the same. One of the reasons for this is that the equation proposed
by Rosenfeld et al.23 does not consider the residual stress in the coating after
deposition and the buckling of the coating during indentation, but assumes that
the coating has a constant hardness and this is directly proportional to the radial
stress. Furthermore, it has been previously seen that for these coated samples the
cracks propagate in the substrate/coating interface leading to chipping of the coating,
which implies that the fracture toughness of the interface is lower than that of the
coating. Thus, comparing Tables 3 and 4 it is expected that the method proposed by
Malzbender et al21 provides the most accurate data.
3.3. Brittle Coatings on Brittle Substrates: Oxide Coatings on Glass
In the previous section, clear through-thickness and interfacial fracture events were
observed form all of the coating/substrate combinations investigated. However,
although interfacial fracture is common in some systems it is by no means universal.
The requirement is that buckling of the coating occurs before there is excessive plastic
deformation of the substrate leading to through thickness cracking at the edge of the
impression prior to delamination during loading. In many cases this cannot be
achieved at practical indentation loads and scratch tests are necessary to cause any
coating detachment. However, the analysis of failure around a moving indenter is
more complicated than for a static indentation and there is no widely accepted
analysis method for interfacial toughness in the scratch test.
Oxide coatings on glass do show a range of though-thickness fracture features
which can be images by scanning electron microscopy if the indentation is large
enough (Figure 11). Radial cracking associated with the indenter edges is observed in
all cases (Figure 11a) and picture-frame cracking is also observed at higher loads
(Figure 11b). However, such cracking is invisible in smaller indents and it is not
always clear the conditions of load and indenter geometry necessary to initiate
fracture. For this reason there has been an interest in an on-line method for assessing
fracture in these small indentations and acoustic emission analysis has been devel-
oped with this in mind.
A range oxide materials which are used in low emissivity and solar control
coatings have been deposited on glass to 200 nm thickness and the acoustic emission
response monitored during nanoindentation testing. In addition a 200 nm layer on
zirconium metal was deposited on the float glass substrate and allowed to oxidise in air
36 S.J. Bull et al.
Figure 11. Scanning electron micrographs of high load nanoindentations in multilayer oxide coatings on
glass (a) 250 mN indent with radial splitting along the indenter edges and (b)500 mN indent showing a
combination of radial splitting and picture frame cracking.
prior to indentation testing. This is expected to have a higher toughness than the
deposited oxides. Figure 12 shows that the first load at which acoustic events are
observed above threshold is highest for this coating which would follow if the acoustic
events are associated with fracture. Also a multilayer optical stack ending in a 200 nm
thick TiOx Ny layer has been deposited for comparison with the single layer coatings;
results are comparable but not identical implying that the substrate is influencing the
fracture behaviour of the coating. Indentations in the uncoated substrate showed no
acoustic emission events over the complete load range investigated.
AE signals result from he sudden release of elastic energy or from surface
interactions such as indenter/surface friction or adhesion. In the case of static
indentation testing the release of elastic energy is the dominant mechanism for AE
formation; sudden release of elastic energy can arise due to unstable crack growth,
high speed phase transformations and plastic instabilities including the initiation of
new slip26 or twinning systems, the activation of dislocation sources27 of the sudden
acceleration or unpinning of existing dislocations.28 The substrate and coatings
investigated here are either amorphous and show poor crystallinity and it is unlikely
950
900
850
Load (μN)
800
750
700
650
600
550
TiOxNy
ZnO
Zr(O)
SnO2
Cap layer
TiOxNy
Figure 12. Graph of the load at which the first acoustic emission event is observed for the samples above
indicated.
200nm ZnO on 20nm TiOxNy on Float Glass

4000
Event 7
3500 P-δcurve
Load (μN) 3000 AE events
2500
Event 8
2000 Event 1
1500
1000 Event 12
500
0
0 50 100 150 200 250
Displacement (nm)
Figure 13. Load displacement curve for 4000 mN peak load nanoindentationin 200 nm ZnO coated glass
where the positions of the acoustic events are indicated.
that dislocation mechanisms will dominate behaviour, particularly as many events

occur on unloading as well as on loading (Figure 13). In addition there are no
acoustic events observed in the testing of uncoated glass at the same loads and the
first events occur well after plastic deformation has initiated and a permanent
impression is created. It is most likely that the acoustic emission arises as a result
of fracture and the ZnO and SnO2 layers would therefore appear to have the lowest
toughness if it is assumed that the defect distributions within these coatings do not
vary greatly. This is a reasonable assumption since they were all deposited in the
same coating system using very similar coating conditions.
Detailed characterisation of the acoustic emission waveform of each of the
events in terms of AEST and tat reveals that there are two different types of events
occurring, a fact which is mirrored for all the coated samples investigated here
(Figure 14). The event with the lower AESF always occurs first in the load-displace-
ment curve and always in groups of three whereas the other event is more random.
The fact that the event with the emission shape factor around two occurs in both
the loading and unloading curves is strong evidence that this event is associated with the
radial splitting of the coating under the indenter. Radial fracture is expected since
the hoop stresses in the material tested with a conical indenter are tensile and this
fracture can extend outside the impression on unloading due to the residual stresses
created by the plastic deformation which has occurred, akin to the mechanism for the
growth of radial cracks in microindentation testing of bulk glasses and ceramics29.
One other interesting observation is that the failures with the lower AESF tend to
occur in threes. This is not expected since the indenter is nominally a 1208 cone with
no edges. However, a detailed analysis of the indenter geometry reveals that it does
not have a circular cross-section but shows three distinct lobes (possibly related to the
effects of crystal structure on the ease of polishing) which are visible in the shape of
indents made at high load in softer materials. Such lobes will act as stress concen-
trations like the indenter edges of a Berkovich indenter and will therefore promote
fracture. This is added evidence that radial fracture is associated with this failure
mode. The higher emission shape factor failure occurs more randomly and at higher
loads and is less influence by the detailed indenter geometry as might be expected for
38 S.J. Bull et al.
1000
Amplitude (microV)
500
0
−500 AESF=2.13
−1000 Event 1
tat=0.053ms
−1500
0 0.02 0.04 0.06 0.08 0.1
(a) Time (ms)
1500
Amplitude (microV)
AESF=3.66
1000 Event 7
tat=0.067ms
500
0
−500
−1000
0 0.02 0.04 0.06 0.08 0.1
(b) Time (ms)
1000
Amplitude (microV)
500
0
−500
AESF=23.66
−1000 tat=0.045ms Event 8
−1500
0 0.02 0.04 0.06 0.08 0.1
(c) Time (ms)
Amplitude (microV)
800 AESF=3.392
tat=0.061ms Event 12
400
−400
0 0.02 0.04 0.06 0.08 0.1

(d) Time (ms)
Figure 14. Analysis of the acoustic function for events (a) 1, (b) 7, (c) 8 and (d) 12 in Figure 13. It is clearly
observed that event 1 and 8 are the same kind of event since the acoustic wave is similar and the emission
shape factor and attenuation time are of the same order of magnitude. The same happens for events 7 and
12, although the similarity is not so pronounced.
picture frame cracking. However, no evidence for fracture was observed in the in situ
AFM images obtained with the indenter after making the impressions so assignment
of the failure types must be treated with caution.
It is clear from the fact that a number of similar acoustic events are observed in a
single load displacement curve that the fracture occurring during indentation is
discontinuous and statistically determined depending on the defects in the coating
and substrate. If estimates of the fracture toughness of the coatings are to be achieved
using this approach then it is necessary to relate the acoustic events to the extra work
of fracture included in the irreversible work of indentation in a similar manner to that
outlined previously for CNx coatings. However, Figure 14 clearly shows that there
are no obvious pop-in events in the load displacement curve and any changes are too
subtle to discern in the raw data. To identify if such events exist the curve must be
processed to reveal these changes.
One approach to achieving this is to plot load against displacement squared30
(Figure 15). For an ideal material this ratio should be constant if elastic-plastic
deformation occurs but may vary gradually from one value to another as plastic
deformation extends from the coating material to the substrate. Any localised excur-
sions from this behaviour represent a change in deformation mechanism as might be
caused by fracture. The data from Figure 14 analysed in this manner in Figure 15 show
a large number of excursions occur in the loading curve for 200 nm ZnO on glass.
However, only seven acoustic events occur during loading and the correlation with the
excursions is poor. In general, the number of acoustic emission events increases as the
number of excursions increases. It is probable that there are many more fracture events
occurring during the indentation cycle than give rise to measurable acoustic emission.
The chances of relating a given acoustic event to the detailed changes in the load-
displacement curve are therefore very small. In addition, since the length of the radial
crack and how it grows are not known (though are probably related to indent size)
determining crack area and toughness by this method is not possible.
4. CONCLUSIONS
Fracture is critical in determining the performance of ceramic coatings in a range of

applications. Both through-thickness and interfacial fracture can occur but the
former is far more likely to be produced in indentation tests. Both through-thickness
200
150
100
Kexp (GPa)
50
−50
−100
0 50 100 150 200
Depth (nm)
Figure 15. Variation of Kexp ( P=d2 ) with displacement during loading for the loading curve in
Figure 13.
40 S.J. Bull et al.
radial fracture (associated with the indenter edges) and picture frame cracking are
observed for most coating/substrate systems. The radial fracture occurs at the lowest
loads and is the first fracture event to be observed in indentation tests. Radial crack
growth is discontinuous and occurs both on loading and unloading of the indenter.
Acoustic emission may be used to monitor fracture events in coated systems though
not all such events produce a measurable acoustic signal. Emission shape factors can
be used to differentiate between different types of events which are related to
different fracture modes.
Calculation of the fracture toughness of the coating and the interfacial tough-
ness can be undertaken in some cases where the deformation mechanisms can be
identified and the type of failure produced is amenable to analysis. However, the
analysis methods cannot be regarded as universal as the failure modes vary with
materials and are often more complex than the models allow for. As the thickness of
the coating is reduced the calculation of toughness is also made more difficult by the
need to obtain accurate measurements of crack area and differences in the work of
indentation which are beyond the resolution of current instrumentation.
ACKNOWLEDGEMENTS
The authors would like to thank Ion Bond, Linkoping University and Pilkington
Technology Centre for the provision of samples and Martin McGurk for the provi-
sion of some data.
REFERENCES
1. A.C. Fisher Cripps, Vacuum, 58 (2000) 569.
2. B. Bhushan, in Handbook of Micro Nanotribology, Ed. B. Bhushan, CRC Press, Boca Raton, 1995,
pp321 396.
3. S.V. Hainsworth, H. Chandler and T.F. Page, J. Mater. Res., 11 (1996) 1987.
4. J. den Toonder, J. Malzbender, G. de With and R. Balkenende, J. Mater. Res., 17 (2002) 224.
5. S.V. Hainsworth, M.R. McGurk and T.F. Page, Surf. Coat. Technol., 102 (1998) 97.
6. MrR. McGurk, PhD Thesis, University of Newcastle, (1997).
7. N. Hellgren M.P. Johansson, E. Broitman, P. Sandstrom, L. Hultman and J. E. Sundgren, Phys.
Rev., B59 (1999) 5162.
8. W.C. Oliver and G.M. Pharr, J. Mater. Res., 7 (1992) 1564.
9. S.V. Hainsworth, S.J. Bull and T.F. Page, Mat. Res. Soc. Symp. Proc. 522 (1998) 433.
10. A. Daugela, H. Kutomi and T.J. Wyrobek, Z. Metall., 92 (2001) 1052.
11. N.I. Tymiak, A. Daugela, T.J. Wyrobek and O.L.Warren, Mat. Res. Soc. Symp. Proc. Fall Meeting
(2002), in press.
12. J. Crepin, T. Bretheau, D. Caldemaison and F. Ferrer, Acta Materalia, 48, (2000) 505.
13. S.J. Bull and A.M. Korsunsky, Tribology Int., 31 (1999) 547.
14. S.J. Bull, J. Vac. Sci. Technol., A19, (2001) 1404.
15. S.J. Bull, Thin Solid Films, 398 399, (2001) 291.
16. B.R. Lawn, A.G. Evans and D.B. Marshall, J. Am. Ceram. Soc., 63, 574 (1980).
17. M.D. Thouless, Eng. Fract. Mech., 61 (1998) 75.
18. X.Li, D. Diao and B. Bhushan, Acta Mater., 44 (1997) 4453.
19. X.Li, D. Diao and B. Bhushan, Thin Solid Films, 315 (1998) 214.
20. J. Malzbender, G. de With and J.M.J. den Toonder, J. Mater. Res., 15 (2000) 1209.
21. J. Malzbender and G. de With, Surf. Coat. Technol., 154 (2002) 21.
22. J.W. Hutchinson and Z. Suo, Adv. Appl. Mech., 29 (1992) 63.
23. L.G. Rosenfeld , J.E. Ritter, T.J. Lardner and M.R. Lin, J. Appl. Phys., 67 (1990) 3291.
24. B.R. Lawn and E.R. Fuller, J. Mater. Sci., 19 (1984) 4061.
25. G.R. Anstis, P. Chantikul, B.R. Lawn and D.B. Marshall, J. Am. Ceram. Soc., 64 (1981) 533.
26. D. Rouby, P. Fleischman and C. Duvergier, Phil. Mag. A47 (1983) 671.
27. V.D. Natsik and K.A. Chishko, Sov. Phys. Solid State, 20 (1978) 1117.
28. J. Weiss and J.R. Grasso, J. Phys. Chem., 101 (1997) 6113.
29. B.R. Lawn and R. Wilshaw, J. Mater. Sci., 10 (1975) 1049.
30. M.McGurk and T.F. Page, J. Mater. Res., 14 (1999) 2283.
NANOINDENTATION,
NANOSCRATCH AND
NANOIMPACT TESTING OF
SILICON-BASED MATERIALS
WITH NANOSTRUCTURED
SURFACES
Jinjun Lu, Tao Xu, Qunji Xue, and Ben D. Beake*
1. INTRODUCTION
In recent years, nanomaterials receive extensive attention due to their unique prop-
erties. Nano-structured Si-based materials are considered as one of the key material
issues in microelectromechanical systems (MEMS).1 Many techniques have been
employed to fabricate Si-based materials with nano-structured surface.2–3 Nanograin
polysilicon ultra-thin film can be deposited on p-type silicon wafers by chemical
vapor deposition (CVD).2 Ion-beam sputtering (IBS) at low energy (typically
1.2 keV) has been used to fabricate nanodots on silicon wafers, which could meet
the key requirement for the development of future optoelectronic and electronic
devices.3 For the tribological consideration however, the bonding between nanograin
polysilicon ultra-thin film by CVD and its substrate might not strong enough.
Therefore, techniques like IBS should be studied to eliminate the interface between
the film and the substrate. Micromachine fabrication using ion implantation has the
*
Jinjun Lu, Tao Xu, Qunji Xue, State Key Laboratory of Solid Lubrication, Lanzhou Institute of
Chemical Physics, Chinese Academy of Sciences, Lanzhou, P. R. China 730000. Ben D. Beake, Micro
Materials Ltd. Unit 3, The Byre, Wrexham Technology Park, Wrexham, LL13 7YP, UK
43
44 Jinjun Lu et al.
advantage of good controllability of material properties.4 Compared with IBS, ion

implantation involves two procedures, implantation and sputtering. The most im-
portant thing is that implanting reactive species can modify the composition of the
implanted layer, which is beneficial for the mechanical and tribological performance.
Therefore, ion implantation is chosen as the means of fabrication.
The aims of our research are: 1) to explore the possibility of fabricating nano-
structured surfaces on single crystal Si and SiO2 by ion implantation and 2) to study
the nanomechanical and nanotribological behaviors of the ion beam modified layers
by nanoindentation, nanoscratch and nanoimpact testing. We found that nano-
structured surfaces can be obtained by N ion implantation at doses of 5 1016 and
1 1017 ions=cm2 , respectively.5 In the present paper, the nanomechanical and
nanotribological behaviors of N ion implanted single crystal SiO2 and Si were studied
using nanoindentation, nanoscratch and nanoimpact testing.
2. EXPERIMENTAL DETAILS
N ion implanted single crystal Si and SiO2 wafer with doses of 5 1015 ,
1 1016 , 5 1016 , 1 1017 and 5 1017 ions=cm2 respectively was prepared using
an ion implanter at 100 keV. Transmission electron microscopy (TEM) was used to
study the microstructure of N ion implanted single crystal SiO2 wafer.
The NanoTest system (Micro Materials, Wrexham, UK) was used for nanoin-
dentation and nanoscratch testing of ion implanted single crystal silicon and nitrogen
ion implanted single crystal SiO2 . For the nanoscratch testing a conical diamond with
25 mm end radius was used throughout. The scratch experiments involved three
sequential scans over the same 100 mm track all at 1 mm=s scan speed. In the first
topography scan the applied load was kept at 0.2 mN (no wear occurs at this load).
In the second (scratch) scan, the applied load was kept at 0.2 mN between 0---20 mm
and then was ramped at a constant rate of 1.77 mN/s so that the maximum load
reached at 100 mm was 142.5 mN. In the final scan the resultant topography was
observed by using a low applied load of 0.2 mN.
For the nanoindentation a Berkovich (3-sided pyramidal) indenter was used
throughout. Repeat indentations to 200 mN were performed with loading and
unloading rates set equal at 4.04 mN/s and an 0.05 mN initial load was applied.
There was a small (5s) dwell time at maximum load to allow for creep before
unloading, since continuing plastic deformation during unloading can distort the
shape of the unloading curve and lead to inaccurate values of hardness and modulus.
The aim of the nanoimpact testing was to investigate the impact wear-resistance
of the ion-implanted Si and SiO2 samples using the Impact Module of the Micro
Materials NanoTest system. The pendulum impulse technique was used. A solenoid
connected to a timed relay was used to produce the probe impacts on the surface,
as shown schematically Figure 1 (note: magnetic rather than mechanical means
was used to actuate the solenoid in this work). The probe was accelerated from
13 microns from the surface for each impact.
The samples were mounted on a standard NanoTest sample stub using super-
glue. They were cleaned with an ‘‘Air Duster’’ and thermally equilibrated in the
NanoTest environmental cabinet before the nano-impact testing. Two different types
of impact tests were set up (Table 1).
Nanoindentation, Nanoscratch and Nanoimpact Testing 45
sample in contact
with test probe
solenoid connected
to timed relay
and power supply
minature
stage
Ready for impact After impact
pendulum impulse impact testing
Figure 1. Illustration of nanoimpact testing.
Table 1. Description of two different types of impact tests.
Mode Test probe Impulse load Test duration
Preliminary testing with 25 mm spherical diamond (nominal radius 5.0 and 600 s
radius spherical probe 25 mm radius) 8.0 mN
Inducing fracture with sharp cube corner (90 degree) diamond 2.0 mN 150 s
cube corner indenter indenter
Impulse control cycle: repeat application of the impact load off for 2 s and then on for 2 s
This cycle corresponds to 1 impact every four seconds, with the probe spending 2 seconds on the sample surface between
successive impacts. Note that in all tests the sample was not moved during impacting so that the probe always impacts in
the same location on the surface.
3.1. Nanoscratch Testing on N Ion Implanted Si and SiO2

3.1.1. N ion Implanted Single Crystal SiO2
The critical load for the transition between plastic deformation and fracture failure
has been estimated from the inflexion point in the on-load scratch trace (scan 2). This
correlates well to the onset of cracking as seen in the final topography scan.
Samples 3 and 4 (5 1016 and 1 1017 ions=cm2 ) shows much improved per-
formance over the unimplanted SiO2 . Of the two, sample 3 has the higher critical
load (Lc 57.2 mN) for fracture failure. It is clear that ion-implantation does not affect
the sample roughness until an ion dose of 1 1017 ions=cm2 is reached. It is possible
therefore that the slight decrease in critical load on sample 4 compared to sample 3 is
due to the adverse influence of increasing roughness on the resistance to nanoscratch-
ing wear.
In addition, the nanoscratch experiments also clearly reveal the applied load
necessary for the transition between elastic and plastic behavior during scratching.
On unimplanted SiO2 there is a long period of elastic behavior (intuitively we expect
a single crystal to behave more elastically in compression than the damaged material)
before the deformation becomes plastic in nature. This is then followed by a short
period of plasticity before fracture (brittle) failure. The behavior on implanted
46 Jinjun Lu et al.
Table 2. Variation in Lc for plastic deformation to fracture failure transition from inflexion
in displacement vs. depth curve, surface roughness, and on load depth reached in scratch test
at maximum load (142.5 mN)
Implanting doses, N=cm2 Average Lc (mN) Roughness(Ra) nm Mean final depth/nm*
Sample 0, 0 42.7 3.09 1085 50

Sample 1, 5 1015 39.7 3.50 867 4
Sample 2, 1 1016 45.8 3.47 830 5
Sample 3, 5 1016 57.2 3.49 833 8
Sample 4, 1 1017 51.5 6.58; 8.40; 8.02 877 8
Sample 5, 5 1017 74.4; 67.7; 59.3 2150 200
Mean values shown 4 6 determinations. ‘‘ ’’ inflexion point not clear on this rough sample.
Ra values are over central region (33:3--66:6 mm) of a levelled 100 mm region of surface before scratch testing.
*
‘‘’’ represents the standard error in the mean, after leveling on the initial 0--20 mm (unworn) region of the track.
samples is different the critical load for the elastic-plastic transition is much lower,
and this is followed by a longer period of plastic behavior before brittle failure occurs
at higher applied load [as the polycrystallites can move more freely under shear].
Typically the elastic-plastic transition was at 27 mN on the implanted SiO2 and 6
12 mN on the implanted samples. The final on-load scratch depth is a composite
response of the sum of elastic þ plastic þ fracture deformation which occurs during
the scratch test, ( a measure of the resistance to scratching wear). It decreases with
increasing ion dosage. The high value on the unimplanted sample is due in part to the
increased elastic deformation on this material. The closeness of the values on the ion-
implanted samples 1 4 shows that at these doses the implantation produces a
material whose mechanical properties do not across the surface (i.e. suggestive of a
well-controlled implantation process).
3.1.2 N Ion Implanted Single Crystal Si
In contrast to what is observed on SiO2 , table 3 below shows that N ion implantation
actually decreases the critical load required for the onset of fracture failure in silicon.
The on-load scratch traces also show an inflexion at the elastic-plastic transition at
much lower load that correlates with final topography data as well. On sample 5 this
is at 10 mN. There is also catastrophic failure at a higher applied load than the Lc
values in Table 3.
The critical load for catastrophic failure occurs at a higher applied load than the
Lc values for the onset of fracture failure tabulated above. It also decreases markedly
with increasing N ion dosage, except on sample 2, which appears more resistant to
catastrophic failure (Table 3).
Although the on-load depth is lower on sample 4 this does not necessarily
represent improved tribological performance in this case. Examination of the curves
show that during the on-load scratch the probe is pushed out of the material by
cracking (e.g. by uplift at the edge of the scratch track which helps to support the
scratch load) which leads to a lower final depth. This material is pushed up over the
scratch track as observed in the final topography as shown in the figure 2b.
Micro-scale scratching of bulk Si by Bhushan and co-workers (1995) showed
similar behavior to our scratch results on the unimplanted Si wafer.6 Using a much
smaller probe (conical diamond with 1 mm end radius) they observed (from friction
Table 3. Variation in Lc for plastic deformation fracture failure transition from inflexion in
displacement vs. depth curve and on load depth reached in scratch test at maximum load
(142.5 mN).
Implanting doses, N=cm2 Mean Lc =mN Mean final depth/nm*
Sample 0, 0 82.1 4.4 490 8

Sample 1, 5 1015 70.7 5.6 467 5
Sample 2, 1 1016 71.5 6.3 561 7
Sample 3, 5 1016 45.7 0.4 524 5
Sample 4, 1 1017 40.9 4.6 390 6
Sample 5, 5 1017 572 26
Mean values from 4 6 determinations. * inflexion point not clear on this rough sample.
‘‘’’ represents the standard error in the mean, after levelling on the initial 0 -20 mm (unworn) region of the track.
and SEM data but with no accurate depth information) the elastic-plastic transition
occurred at 4mN and ploughing at higher load until the onset of fracture at 16 mN
and catastrophic failure at 18 mN. In view of the differences in contact pressure due
to the different tips used these results are in reasonable agreement with those of this
study. Post-scratch SEM shows that that much material is moved into/on-top of the
scratch track (particulate debris of sub-micron size). This is consistent with the large
increases in probe position in the final topography observed in our experiment.
It is clear from Table 3 that N ion implantation does not result in roughening of
the Si surface at any of the doses used. To generate sufficient damage to get rough-
ening it is necessary to use argon ions of larger cross-sectional area.
3.2. Pop-in and Pop-out Phase Changes of N Ion Implanted Si

Silicon is well-known as having unusual indentation behavior due to phase changes
which occur on loading and unloading. The pop-ins have been associated with
transition from semiconducting diamond cubic structure into metallic B-tin structure
(i.e. a semiconductor-to-metal) at critical pressure. Unloading causes a volume-
expanding reverse phase transformation from B-tin into an amorphous structure.7–8
Since these phase changes occur at critical pressures it is not possible to compare
456.4 393.3
350.0
400.0
300.0
inflexion
350.0 250.0
200.0
300.0 low final depth
150.0
due to volume uplift
250.0 100.0
Depth (nm)
Depth (nm)
onset of 50.0
200.0 fracture failure
0.0
150.0 -50.0
-100.0
100.0 dramatic
-150.0
fracture failure
50.0 -200.0
-250.0
0.0
-300.0
-40.7 -359.7
0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 90.0 99.9 0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 90.0 99.8
Distance (um) Distance (um)
(a) unimplanted surface (b) sample 4 (1 x 1017)

Figure 2. Typical behavior of N ion implanted Si.
48 Jinjun Lu et al.
Table 4. Pop in and pop out phase changes on ion implanted Si
Implanting doses, N=cm2 Pop in load/mN Pop out/mN
Sample 0, 0 170, 176 115.4, 123.8

Sample 1, 5 1015 178.3, 178.3 122.5, 126.3, 110.5, 111.1, 111.5
Sample 2, 1 1016 # 86.6, 82.4, 60.5
Sample 3, 5 1016 152.2 53.7
Sample 4, 1 1017 148.6 62.6, 57.9, 65.8
Sample 5, 5 1017 # 43.3, 52.2
*
critical load values in repeat tests are shown in italics. Note that the load given is that for the start of the
pop in or out transition. # not clear transition during loading.
exact values with other studies since our indenter geometry will be different (it is
likely to be blunter so that a greater applied load is necessary before the critical
pressure for phase transformation is reached).
The critical pressures on the Si sample implanted with the lowest dose are very
similar to those of the unimplanted Si (Figure 3). At greater pressures the pop-in phase
transformation occurs at lower load (i.e. is made energetically favourable) but the
unloading transformation occurs at lower load (made energetically unfavourable).
It has been shown previously that CNx films retard the unloading phase trans-
formation, i.e. that the coating keeps the Si in the dense metallic form.9 It has also
been shown that amorphous Si does not undergo the same phase transformations on
loading and unloading.10 It appears that the ion implanted samples act as if they were
partially amorphised.
The indentation curves also showed that Sample 5 shows a very soft near-surface
region revealed by shapes of loading curves i.e. plastic flow at low indentation
pressures, as has been seen on amorphous Si.
3.3. Comparison Between Si and SiO2
Various parameters such as the on-load depth and critical load for fracture can be
used to compare between the materials. Nanoindentation shows that unimplanted Si
is harder and less elastic (higher modulus) than SiO2 . The lower critical loads
1095 1105
1000 1000
pop-out event pop-out event
900 900
800 800
700 700
Depth (nm)
Depth (nm)
600 600
500 500
400 400
300 300
200 200
100 100
3 4
0.0 20.0 40.0 60.0 80.0 100.0 120.0 140.0 160.0 180.0 200.3 0.0 20.0 40.0 60.0 80.0 100.0 120.0 140.0 160.0 180.0 200.5
Load (mN) Load (mN)
Figure 3. Pop outs on sample 1 (5 10 ) and sample 4 (1 1017 ). 15

observed on SiO2 and greater maximum on-load scratch depth (by a factor of 2 3)
correlate with its lower hardness.
The H/E ratio is a key parameter determining the type of behaviour observed in
nanoindentation and nanoscratching wear. Si has a lower H/E ratio than SiO2 which
explains why the transition from elastic to plastic behaviour occurs at higher contact
pressure on this material.
It is interesting to speculate why ion implantation does not lead to a similar
obvious improvement in the nanotribological performance of Si as it does on SiO2 .
Being harder (and having a more closed crystal structure?) the Si may not be damaged
sufficiently by N ions at these doses to reach the fully-nanocrystalline state. Support-
ing evidence is (1) pop-ins and pop-outs are still observed during nanoindentation
(even if the critical pressures are changed) amorphous Si does not do this (2) even the
highest N dose does not lead to any roughening of the surface.
Bhushan (2001) has shown that Cþ ion implantation results in a hardening of
the surface layer which results in improved microscratching wear performance. With
N ion implantation do the competing mechanisms of 1) softening due to destruction
of the matrix 2) hardening due to the formation of Si3 N4 (or similar) phases, act
against each other resulting in no improvement in tribological properties?
3.4. Nanoimpact Testing of N Ion Implanted Single Crystal Si and SiO2
3.4.1. Preliminary Testing with 25 mm Radius Spherical Probe

It is not possible to use nanoindentation methods to obtain information about the
impact fracture and erosive wear performance. High hardness can confer improved
wear resistance but it is often associated with brittleness and poor toughness and
erosive wear resistance.
3.4.1a N Ion Implanted Single Crystal Si

The preliminary impact test results are shown in Figure 4a. The plots show the
evolution of impact-induced damage with time (i.e. plastic deformation, fatigue
wear and fracture) throughout the test. The height of the surface is recorded with
sub-nm precision when the probe is on the surface.
c:\Lanzhou\si wafer S2c

600
600
575
550 550 2, 4 - discrete failures
500 525
450
500 1, 3 - slow crack growth
475
400 450
350 425
Depth (nm)
400
300
375
Depth (nm)
250
350 4
200 325
300
150
275
3
100 250
50
Cursor Position
225
2
0
Depth (nm) 382 08 200 1
02 20 0 40 0 60 0 80 0 100 0 120 0 140 0 160 0 180 0 200 0 220 0 250 0 Time (s) 120 2 175
Time (s) 150
125
T me (s) Depth Ax s (nm) Scan Type Multiple Pendulum Impulse Exit 100
Gr d Auto scale Gr d Auto scale
Repeat number 1 Previous Next >>
75
Maximum 250 0 Maximum 600 00
Inve t 50
Minimum 0 2 Minimum 0 Pulse number 1 Previous Next >> 02 20 0 40 0 60 0 80 0 100 0 120 0 140 0 160 0 180 0 200 0
Time (s)
(a) (b)
50 Jinjun Lu et al.
The results show typical impact behavior for a brittle material. There are long
periods of damage accumulation (‘‘plateau periods’’) where small cracks grow sub-
surface and the surface position changes relatively little with time. These are separ-
ated by more rapid events where material is removed as the cracks propagate and
coalesce leading to these failures.
The instrument software allows for ‘‘zooming-in’’ on features of interest such as
these failure events. Figure 4b shows zoom to the beginning of the tests. The plot below
is very interesting as it shows volume uplift (or surface bulging) just before fractures.
This ‘‘volume uplift’’ is due to the formation of micro- and nano-cracks sub-
surface in the material. Because the damaged material is constrained from breaking
away from the surface by the (less damaged) material above it, the cracks cause the
material to swell up forcing the probe ‘‘out of the surface’’. This has also been seen on
DLC coatings and other brittle materials.
The behavior in extended impact tests (i.e. at longer impact times) shows a
similar trend to production of sub-surface damage and surface bulging before surface
failure etc.
As well as these very clear multiple fractures another common type of behavior
is illustrated below.
1125
Si wafer S3
1000
900 sub-surface cracks

push probe out
800
just before the
700 catastrophic failures
Depth (nm)
600
500
400
300
200
100
0
02 20 0 40 0 60 0 80 0 100 0 120 0 140 0 160 0 180 0 200 0 220 0 240 0 260 0 280 0 300 0
Time (s)
Figure 5. Volume uplift due to the formation of micro and nano cracks sub surface in the material.
400
380 Si wafer S1
360
340 short plateau period
320
300 nano-cracks
280 swelling surface
260
240
Depth (nm)
220
200
180
160
140
120
100
80
60
40
20
0
02 50 0 100 0 150 0 200 0 250 0 300 0 350 0 400 0 450 0 500 0 550 0 600 0
Time (s)
Figure 6. Multiple fractures.

There is a very short plateau period followed by 1 or more small cracking events
(the surface bulges before the cracks coalesce, and the process repeats). The first
inflexion between the flat plateau and faster wear/cracking occurs after ~30 50 s, (at
greater times for higher doses). The depth at this transition is (180 nm 20) for all
the samples.
3.4.1b N Ion Implanted Single Crystal SiO2

More common under these conditions than sharp fracture is the type of behaviors
shown below.
The probe penetrates ~70 nm into the surface under the static of load of 8 mN.
The first impact results in further 80 nm depth penetration (plastic deformation).
There is then a plateau period where the probe position levels off (i.e. some plastic
deformation occurs but surface not being worn away during this period, erosive wear
rate is very low). During this period the plastic deformation results in the initiation
and development of cracks underneath the contact surface. At ~54 s (shown by the
cursor in the plot) the depth begins to increase again (ie rate of wear increases
sharply). The time-at-inflexion is likely to be related to the fracture resistance, since
longer times presumably reflect increased barrier to coalescence/propagation of sub-
surface nano-/micro-cracks. These times vary on the SiO2 samples as shown below.
Results at 5 mN (squares) and 8 mN (triangles) show a trend to increasing time-
to-inflexion with the ion dose, reaching a maximum at dose 1 1016 ions=cm2 . The
inflexions are small.
From the results with the blunt spherical probe we can see that impact testing is
a statistical process, since it involves inducing plastic deformation, crack initiation
and coalescence/fracture [i.e. fatigue]. Several tests are necessary to get a complete
picture of the range of behaviour observed, as is common in fatigue tests.
c:\Lanzhou\s3 rep8mn
250
225
200
175
150
Depth (nm)
125
100
75
50
25
Cursor Position ...
Depth (nm) 161.56

0
02 20 0 40 0 60 0 80 0 100 0 120 0 140 0 160 0 180 0 200 0 220 0 240 0 260 0 280 0 300 0
Time (s) 54.3
Time (s)
Time (s) Depth Axis (nm) Scan Type Multiple Pendulum Impulse Exit
Grid Auto scale Grid Auto scale
Maximum 300 0 Maximum 250 00 Repeat number 1 Previous Next >>

Invert
Minimum 0 2 Minimum 0 Pulse number 1 Previous Next >>
Figure 7. Sharp fracture in N ion implanted single crystal SiO2 .

52 Jinjun Lu et al.
60
50
40
Time (s)
30
20
10
0
0 1 2 3 4
SiO2 sample nos
Figure 8. Variation in time to first failure (s) with sample.
Clearly the wear rate is not constant through the test and detailed examination
of the results allows individual failure events (shown by the changes in slope and
abrupt changes in depth) to be investigated.
There was evidence that the time at which the increase in wear rate occurs
correlated with ion dose for both the Si and SiO2 implanted samples (generally
greater dose ¼ longer time-to-failure).
3.4.2 Inducing Fracture with Sharp Cube Corner Indenter
A cube corner indenter was used to test these observations and to more easily induce
fractures. The cube corner indenter is sharper (higher contact pressure) and the
90 degree angle means that the instantaneous contact strain under the probe on
impact is much greater than for the spherical probe.
Failures are generally less sharp than on SiO2 . The fracture resistance of the samples
can be compared by: (1) wear rate at end of test, (2) depth at end of test, (3) presence or
absence of clear surface failure/fracture events, and (4) the time at which first failure/
fracture occurs. The type of behavior observed falls into distinct groups (Table 5).
The fracture toughness of sample 4, (i.e. resistance to propagation of the cracks
which would cause failure and removal of material) is high. This is seen in the final
impact depth data.
These impact depths have been compared with the final on-load scratch depth
data in the graph below (impact ¼ circles, scratch ¼ filled squares).
Table 5. Classified group on ion implanted Si.
Group Pop in load/mN
Sample 0, 0 Low wear rate, with little clear fracture

Sample 1, 2 and 5 All are very susceptible to surface damage, cracking/fracture (especially at short
times) and also show a high wear rate at 150 s. Short periods of surface bulging/
swelling are often followed by cracking, and fast wear rate
Sample 3 generally lower wear rate, with some failures at relatively short times
Sample 4 shows occasional small failures in some tests but mostly they are absent. The wear
rate soon drops to near zero and the predominant process is sub surface cracking
causing the slight bulging/swelling of the surface
1500 1500
1400 1400
1300 1300
1200 1200
1100 1100
1000 1000
900 900
Depth (nm)
Depth (nm)
800 800
700 700
600 600
500 500
400 400
300 300
200 200
100 100
0 0
02 10 0 20 0 30 0 40 0 50 0 60 0 70 0 80 0 90 0 100 0 110 0 120 0 130 0 140 0 150 0 02 10 0 20 0 30 0 40 0 50 0 60 0 70 0 80 0 90 0 100 0 110 0 120 0 130 0 140 0 150 0
Times (s) Times (s)
Si 0; unimplanted Si Si2
1500 1500
1400 1400
1300 1300
1200 1200
1100 1100
1000 1000
900 900
Depth (nm)
Depth (nm)
800 800
700 700
600 600
500 500
400 400
300 300
200 200
100 100
0 0
02 10 0 20 0 30 0 40 0 50 0 60 0 70 0 80 0 90 0 100 0 110 0 120 0 130 0 140 0 150 0 02 10 0 20 0 30 0 40 0 50 0 60 0 70 0 80 0 90 0 100 0 110 0 120 0 130 0 140 0 150 0
Times (s) Times (s)
Si3 Si4
Figure 9. Typical impact behavior of N ion implanted Si.
The correlation between the impact and scratch data is striking. A much higher
ion dose is necessary for tribological improvement on the Si than on the SiO2 .

Failure events on the SiO2 samples were clearer (greater change in depth) than on the
Si samples. Final depths were also larger which mimics the difference observed in the
scratch test data reported previously.
There was some variation in the time at which the first failure occurs and its size.
Although there was some variability (e.g. the samples did not always fracture within
1200 700
1000 600
scratch depth (nm)
impact depth (nm)
500
800
400
600
300
400
200
200 100
0 0
0 1 2 3 4 5
Si sample nos
Figure 10. Correlation between impact and scratch data for Si samples.
54 Jinjun Lu et al.
Table 6. Impact depth of ion implanted Si.
Implanting doses, N=cm2 Impact depth after 150 s (nm)
Sample 0, 0 823 30
Sample 1, 5 1015 939 48
Sample 2, 1 1016 1013 72
Sample 3, 5 1016 838 62
Sample 4, 1 1017 646 42
Sample 5, 5 1017 1067 27
Standard errors in the mean are from 10 measurements at different regions of the sample
surface.
150 s), 5 10 repeats for each sample were carried out so that clear trends in behavior
could be deduced.
Illustrative results on each of the samples are shown in the following graphs. S3
and s4 generally show much longer time-to-failure. The failure is more extreme on s4
and hence this sample shows the largest depth at 150 s.
We have previously impact tested a variety of hard coatings [DLC, TiN, alu-
mina etc]. The abrupt failures on S4 look quite similar to adhesion failures i.e. a
weak interface/interphase (between the ion implanted and unimplanted regions)
eventually gives way catastrophically. It may be that the final depth alone cannot
be used as a direct measure of fracture toughness for these SiO2 samples.
4. CONCLUSIONS
The tribological properties of N ion implanted single crystal Si and single crystal SiO2
surfaces have been evaluated by nanoscratch and nano-impact testing. Nanoscratch
data show the critical loads associated with (1) the elastic-to-plastic transition and (2)
the plastic deformation-to-fracture failure transition. The scratch results show that at
moderate ion doses the implantation process improves the nanotribological properties
of single crystal SiO2 , with similar ion dose of 1 1016 1 1017 N cm 2 also confer-
ring improved resistance to impact/fatigue wear. For the Si samples, the correlation
between the impact data and the on-load scratch data was striking. An ion dose of
5 1017 N cm 2 was necessary for improved resistance to impact/fatigue wear. Impact
wear appears to involve 3 processes (1) plastic deformation (2) formation of sub-surface
cracks and (3) coalescence of these cracks and failure (by material removal from contact
surface).
Table 7. Description of failure.

S0 generally short time to failure
S1 short time to first failure, multiple failures
S2 short time to first failure
S3 longer time to failure, which is often only minor
S4 quite long time to dramatic failure
S5 were not tested since previous tests under less strong conditions showed very poor performance on this sample the
near-surface layer is too weak that the impact probe punches straight through. Time-to-failure data show an increasing
trend S3>S4>S2~S1~S0
c:\Lanshou\s0 rep2mnx5cubeb c:\Lanshou\s0 rep2mnx5cubeb
2500 2500
2250 2250
2000 2000
2750 2750
2500 2500
Depth (nm)
Depth (nm)
2250 2250
1000 1000
750 750
500 500
Cursor Position Cursor Posit on
250 250
Depth (nm) 250 00 Depth (nm) 1250 00
0 0
0 2 10 0 20 0 30 0 40 0 50 0 60 0 70 0 80 0 90 0 100 0110 0120 0130 0140 0150 0 Time (s) 74 9 0 2 10 0 20 0 30 0 40 0 50 0 60 0 70 0 80 0 90 0 100 0110 0120 0130 0140 0150 0 Time (s) 74 9
Time (s) Time (s)
Time (s) Depth Axis (nm) Scan Type Mult ple Pendulum Impulse Ex t T me (s) Depth Ax s (nm) Scan Type Multiple Pendulum Impulse Exit
G id Auto scale Gr d Auto scale Grid Auto scale G id Auto scale
Max mum 150 0 Max mum 2500 00
Repeat number 1 Privious Next >> Max mum 150 0 Maximum 2500 00 Repeat number 1 Privious Next >>
Inve t Invert
Minimum 0 2 Min mum 0 Pulse number 1 Privious Next >> Minimum 0 2 M n mum 0 Pulse number 1 Privious Next >>
Figure 11. Typical impact behavior of N ion implanted SiO2 .
Sample 0, 0 1282 124

Sample 1, 5 1015 1315 203
Sample 2, 1 1016 1259 138
Sample 3, 5 1016 1138 95
Sample 4, 1 1017 1590 165
56 Jinjun Lu et al.
ACKNOWLEDGEMENT
The project is supported by Chinese Natural Science Foundation (No.50172052 and

No. 50023001).
REFERENCES
1. S. M. Spearing. Materials issues in microelectromechanical systems (MEMS), Acta Mater. 48,
179 196(2000).
2. S. Ecoffey, D. Bouvet, A. M. Ionescu, and P. Fazan. Low pressure chemical vapour deposition of
nanograin polysilicon ultra thin films, Nanotechnology 13, 290 293(2002).
3. R. Gago, L. Vazquez1, R. Cuerno, M Varela, C Ballesteros, and J M Albella. Nanopatterning of
silicon surfaces by low energy ion beam sputtering: dependence on the angle of ion incidence,
Nanotechnology 13, 304 308(2002).
4. S. Nakano, H. Ogiso, A. Yabe. Advanced micromachine fabrication using ion implanted layers,
Nucl. Instr. and Meth. in Physics Research B 155, 79 84(1999).
5. T Xu, J. J. Lu, J. Tian, and Q. J. Xue. The effect of nitrogen ion implantation on wear behavior of
single crystal SiO2 , J. of Phys. D: Appl. Phys. 33, 426 429(2000).
6. Bhushan B., Gupta B.K., and Azarian M. H., 1995, Wear 181: 743 758.
7. B. Roos, H. Richter, and J. Wollweber, Solid State Phenomena 47 48, 509 516(1996).
8. T. F. Page, W. C. Oliver and C. J. McHargue, J. Mat. Res. 7, 450(1992).
9. S. Bull, Applications of nanoindentation to industrial problems. Nanomechanical testing, VITO,
Belgium 5th Dec 2001
10. J. S. Williams, et al. Ultra microindentation of silicon and compound semiconductors with spherical
indenters, MRS 1999.
11. Bhushan B. Tribology on the macroscale to nanoscale of microelectromechanical system materials: a
review, Proc Instn Mech Engrs 215, 1 18(2001).
12. B.D. Beake, S.R. Goodes and J.F. Smith, Micro Impact testing: A new technique for investigating
thin film toughness, adhesion, erosive wear resistance and dynamic hardness, Surface Engineering 17,
187 193(2001).
13. B.D. Beake, M.J. Ibanez Garcia and J.F. Smith, Micro impact testing: a new technique for investi
gating fracture toughness, Thin Solid Films 398 399, 438 443(2001).
14. B.D. Beake, S.R. Goodes, J.F. Smith, A. Zhang and J. E. Micro impact testing: A novel nano /micro
tribological tool for assessing the performance of wear resistant coatings under impact/fatigue
conditions, Science in China (Series A) 44 Supp. 418 422 (Proc. 3rd Int. Symp. Tribology, Beijing,
Sept 2001)
MICROSTRUCTURAL
CONTROL OF INDENTATION
CRACK EXTENSION UNDER
EXTERNALLY APPLIED
STRESS
Robert F. Cook1
1. INTRODUCTION
In many ceramic and glass materials indentation half-penny cracks extend stably
under applied stress to over twice their original length before instability consistent
with an ideal identation response. Observations of such extension can be made with
custom-designed bending fixtures, allowing crack extension in indented specimens to
be observed with an optical microscope during controlled loading. The resulting
applied stress vs crack length traces exhibit a concave-down shape, with decreasing
derivative prior to instability at which the specimen breaks and the derivative
vanishes. Many materials, however, do not exhibit this behavior a consequence of
the influences of reactive environments, lateral cracking, microstructural toughening,
surface stresses, phase-transformation trapping and preferred crystallographic cleav-
age planes. These influences lead to subtantially suppressed or enhanced stable crack
extension, asymmetric extension and stick-slip extension. After a brief development
of the indentation fracture model, in-situ observations of indentation crack extension
in a set of cordierite glass-ceramic composites and a polycrystalline alumina are
presented and examined in terms of variability of microstructural interactions giving
rise to toughening.
1
Robert F. Cook, Materials Science and Engineering, University of Minnesota, Minneapolis, MN,
55455.
57
58 Robert F. Cook
2. INDENTATION FRACTURE MECHANICS
A fracture system comprising a crack in a brittle material generated by a (sharp)

elastic-plastic contact event and subsequently subjected to an applied stress is char-
acterized by the fundamental equilibrium equation1,2
Ka þ Kr ¼ Tbase þ Tm (1)
where Ka is the stress-intensity factor associated with the applied stress field, Kr is the
stress-intensity factor associated with the residual contact field, Tbase is the toughness
of the material in the environment of interest and Tm is a toughness amendment
associated with interactions between the crack and the microstructure of the material
or the component. By definition, Tbase is crack-length invariant and is related to the
equilibrium (fracture) surface energy of the material, g, by Tbase ¼ (2gE)1=2 , where E
is the material Young’s modulus: g depends on environment and takes a maximum
value of g 0 in an inert, non-reactive, environment, with an analogous maximum base-
toughness value of T0 . The remaining terms in Eq. (1) may all depend on the length
of the crack, c, and usually the applied stress-intensity factor is destabilizing,
dKa =dc > 0, the residual stress-intensity factor is stabilizing, dKr =dc < 0, and the
microstructural interaction leads to toughening, Tm > 0 and dTm =dc$0. Such micro-
structural toughening can be associated with the material, as in the cases of grain
related crack face frictional interlocking in polycrystalline alumina3 or metastable
phase transformation crack tip shielding in zirconia alloys.4 Or, toughening effects
can be imbedded in the component, as in the case of tempering-induced compressive
stress profiles generated in ion-exchanged glass parts.5 The equilibrium relation of
Eq. (1) has been expressed in the component frame of reference, in which the
microstructural interactions are associated with the toughess. Equilibrium can also
be expressed in the crack-tip frame of reference by setting Tm ¼ Km and collecting
all terms but Tbase on the left side of the equation, or in the applied loading frame of
reference by collecting all terms but Ka on the right side of the equation, thereby
treating Kr as a toughening element.
The variables in an indentation crack extension experiment are shown in Fig 1.
The indentation flaw consists of residual square contact impression (for a Vickers
diamond pyramid indenter) in the component surface with cracks perpendicular to
the surface emanating from the impression corners. The crack length, c, is charac-
terized by the surface trace, measured from the impression center. The applied stress,
sa , is imposed perpendicular to one set of cracks (for a uniaxial loading experiment)
The size of the contact impression, and the initial length of the cracks prior to
Figure 1. Schematic plan view diagram illustrating the measured parameters during an indentation crack
extension experiment: the crack surface trace length, c; and the applied stress, sa .
Microstructural Control of Indentation Crack Extension 59
application of any stress, is determined by the indentation load P used to generate the
flaw. The hardness, H, of the material is related to the load and the area of the
contact impression, A, by H ¼ P=A.
The driving forces for fracture during an extension experiment, in the context of
Eq. (1), are illustrated in Fig. 2. The residual stress-intensity factor, Kr , is modeled as
a point-source loading at the center of a circular crack, giving rise to a 1=c3=2 crack
length dependence. The loading arises from the reaction of the surrounding elastic
matrix to the strain mismatch imposed by the residual plastic contact impression.
The amplitude of the residual term is proportional the indentation load, modified by
an elastic-plastic expanding cavity term, x ¼ j(E=H):1=2:6
Kr ¼ xP=c3=2 (2)
The real geometry (finite extent of deformation zone, semi-elliptical crack on a free
surface and the presence of other crack systems) of the modeled crack system is
accounted for by the material-invariant dimensionless term j. The applied stress-
intensity factor, Ka , is modeled by a stress acting uniformly over the crack with a c1=2
dependence and a dimensionless term c accounting for crack-shape and free-surface
effects:
Ka ¼ csa c1=2 (3)
Ideal crack exentension behavior is obtained by combining the stress-intensity

factors of Eqs. 2 and 3 into the equilibrium equation condition of Eq. 1 under the
constraint of an invariant toughness, T ¼ Tbase þ Tm , and inverting to obtain
sa (c) ¼ (T xP=c3=2 )=cc1=2 (4)
Equation 4 can be solved for two equilibrium conditions: an initial stable equilibrium
configuration at zero stress,
cinit ¼ ( xP=T)2=3 , sa ¼ 0 (5)
Figure 2. Schematic cross section illustrating the crack driving forces acting during indentation crack
extension: the localized loading of the residual stress intensity factor, Kr , and the uniform loading of the
applied stress intensity factor, Ka . The semi elliptical surface crack is modeled as a circular crack in an
infinite body.
60 Robert F. Cook
Table 1. Indentation Crack Extension Behavior of Ceramic Materials.
Suppressed Enhanced Erratic Asymmetric

Crack Crack Crack Crack
Material Extension Extension Advance Advance
Cordierite Glass Ceramic

Polycrystalline Al2 O3 (small grain size)7
Si3 N84 (wrong shape)
Soda Lime Glass1 X
Tempered Glasses9 X
Tetragonal Zirconia Polycrystal10 X X
Polycrystalline Al2 O3 (large grain size) (wrong shape) X X X
Whiser Reinforced Glass Ceramic X X X
Al2 O3 =Al2 TiO75 X X X
and a final stable-unstable transition at the maximum sustainable stress,1,2
cm ¼ (4xP=T)2=3 , sm ¼ 4T=3cc1=2
m (6)
Provided the crack propagation is carried out under near equilibrium conditions, the
crack system follows the sa (c) trajectory given by Eq. 4 from the stable configuration
of Eq. 5, (cinit , 0), to the onset of instability characterized by Eq. 6, (cm , sm ).
Provided, also, that the amplitude terms, x and c, are invariant during propagation,
the extension ratio, cm =cinit ¼ 42=3 2:52. Under these conditions, an alternative
formulation of Eq. 4.
P=c3=2 ¼ (T=x)[1 (c=T)sa c1=2 ] (7)
allows the geometry terms to be evaluated easily from crack extension data, as a plot
of P=c3=2 vs sa c1=2 follows a straight-line trajectory of constant stress-intensity factor
(¼ toughness) with slope 1 and intercepts T=x and T=c.
In many materials, indentation crack extension behavior does not follow the
trajectory of Eqs. 4 or 7, and does not exhibit an extension ratio of 2.52. Table 1
summarizes the crack extension behavior of a number of ceramics. Three classes of
behavior are evident: (i) those materials that exhibit the ideal extension ratio and
trajectory; (ii) those that exhibit a suppressed extension ratio, < 2:52, often associated
with a non-ideal trajectory; and (iii) those that exhibit an enhanced extension ratio,
> 2:52, usually associated with erratic or asymmetric crack advance. It is the latter
group that is the focus here, as their crack extension characteristics are a reflection of
the effects of microstructure on toughness. That is, T in Eqs. 4-7 is not invariant.
Measurement of the sa (c) trajectory can then be used to discern the Tm (c) variation for
a material and its point-to-point variability within a microstructure.
3.1. Materials
Two materials are considered here to illustrate the effects of microstructure on crack
extension. The first is a cordierite (2MgO:2Al2 O3 :5SiO2 ) based glass-ceramic contain-
ing silicon nitride whiskers. The whiskers were approximately 1 15 mm and were
mixed in with the base glass-ceramic powder prior to tape-casting at 0, 5, 10 and 15
vol%. The whiskers significantly retarded densification during sintering, such that the
modulus and hardness of the composite materials were reduced; E and H were 120 and
8.3 GPa and 104 GPa and 3.1 GPa for the base material and 15 vol% composite
respectively. The second material is a large-grained polycystalline alumina containing
0.1 mol% MgO.2 The mixed powder was dry-pressed and then sintered in H2 at 1800 C
for 12 hours to reach > 99:5% relative density. The mode grain size was 15 mm.
3.2. Testing Method
Test specimens were cut into bar form, approximately 45 mm 6 mm 3 mm. Single
Vickers indentations, with one set of cracks aligned perpendicular to the bar axis,
were placed in the center of the largest face of the bar. In the case of the cordierite
composites, a single 100 N indentation load was used. For the alumina, loads in the
range 10 100 N were used. The bars were then placed in a custom-built testing
apparatus, illustrated in Fig. 3. The apparatus generated a tensile stress in four-
point bending on the indentation surface of the bar. The stress was maintained by
displacement control through a stepping motor and the load monitored by a resist-
ance load cell. The entire apparatus was mounted on an inverted optical microscope
so that the indentation crack lengths could be monitored during loading. All inden-
tation and testing was performed in air. During the crack extension on loading, the
load was incremented and the cracks allowed to come to rest (velocity less than
3 10 10 ms 1 ) before measuring the crack length and then re-incrementing the load.
The crack extension responses were thus characteristic of equilibrium fracture in the
air environment.
4. RESULTS
Figure 4 shows applied stress-crack length trajectories for the base cordierite glass-
ceramic material. Measurements on two opposing cracks on two specimens were
made; the trajectories are symmetric and relatively smooth for each specimen and the
entire set of data are well described by the (penny-crack based) ideal extension
behavior of Eq. 4, shown as the best fit solid line and associated instability crack
length and stress values. In this case it appears that the assumptions of extension
under conditions of quasi-equilibrium propagation and invariant material toughness
and crack geometry are fulfilled. Figure 5 replots the data of Fig. 4 according to the
scheme of Eq. 7, emphasizing the shift from residual stress-intensity dominance at the
beginning of the test to applied stress-intensity dominance at failure. A straight line
fit in these coordinates allows the combined toughness/geometry parameter combin-
ations to be obtained from intercepts as indicated on the Figure. These parameters
may be used to evaluate the toughness characteristics of similar materials, referenced
to this base material. Combining the full equilibrium expression of Eq. 1 with the
crack extension expression of Eq. 7 gives
T=Tbase ¼ (c=cbase )(cbase =Tbase )sa c1=2 þ (x=xbase )(x base =Tbase )P=c3=2 (8)
62 Robert F. Cook
Figure 3. Diagram of the apparatus used to measure indentation crack extension under applied stress.
which indicates that once a base material has been calibrated, the toughness vari-
ations with crack length of related materials can be determined from crack extension
measurements, once the applied and residual stress geometry transformations are
known.
Figure 6 shows the relative toughness behavior of the set of cordierite whisker
composites as a function of crack length, determined using crack extension data such
as those shown in Fig. 4 and transformed as given by Eq. 8. The applied geometry
term was taken as invariant with whisker volume fraction, c=cbase ¼ 1, and the
residual geometry term scaled with the elastic-plastic response, x=x base ¼ (E=H)1=2 =
(Ebase =Hbase )1=2 . As anticipated by Figs. 4 and 5, the base material response is invariant
with crack length. The responses of the whisker composites are also relatively invariant,
but clearly increase in their near steady-state toughness values with whisker volume
fraction. In addition the 15 vol% material exhibits a clear enhancement of crack
extension (to almost a factor of five) over that of the base material. The implication
Figure 4. Plot of the equilibrium indentation applied stress crack length, sa c, extension behavior for
cordierite glass ceramic in air. The data are well described by an ideal indentation response based on the
penny crack model of Fig. 2.
Figure 5. Plot of the equilibrium indentation crack extension behavior for cordierite glass ceramic in air
using parameters related to the residual and applied stress intensity factors. A straight line best fit to the
data allows the crack geometry terms to be evaluated from the intercepts.
of these data is that although the addition of the whiskers to the cordierite glass ceramic
increased the porosity and thus degraded the modulus and hardness (and presumably
the base toughness), the actions of the whiskers during crack propagation more than
compensated to lead to overall tougher materials. The crack length scale associated
with the attainment of steady-state toughness characteristics appears from Fig. 6 to be
approximately 200---300 mm.
64 Robert F. Cook
Figure 6. Plot of relative toughness as a function of crack length for cordierite glass ceramic containing
silicon nitride whiskers at the volume fractions indicated. Adding whiskers enhances the toughness over the
base material by over a factor of two, approximately 1 MP m1=2 for this material in air.
Figure 7 shows the applied stress-crack length trajectories for the alumina
material; measurements on two opposing cracks on a single specimen were made.
The trajectories exhibit behavior very different from that of the base glass-ceramic of
Fig. 4, as indicated in Table 1. For a given indentation, the crack traces are not of the
same length and the cracks do not extend similarly, leading to extremely asymmtric
crack extension. On increasing stress, the cracks exhibited erratic ‘‘stick- slip’’ exten-
Figure 7. Plot of the equilibrium indentation applied stress crack length, sa c, extension behavior for a
large grained polycrystalliune alumina in air. The data are not well described by an ideal indentation
response, exhibiting asymmetry, erratic and variable extension, and very small strength degradation over
the flaw size range, indicating a significant interaction between the cracks and the microstructure.
sion, sequences of fixed length for a large stress increase followed by an abrupt
increase in length for a small stress increment. This behavior was more prevalent
for small indentation flaws. Taken as a group, the indentations exhibited variable
crack extension ratios, from zero to approximately eight. And finally, the material
exhibited extreme flaw tolerance; over the range of a factor of 20 in indentation load,
the strength degraded by a factor of about 0.3, compared with a factor of 2.7
predicted by Eq. 6.
Figure 8 shows the relative toughness behavior of the alumina as a function of
crack length, determined using the crack extension data of Fig. 7 and transformed by
Eq. 8. The applied and residual geometry terms were taken from the experiments of
Braun et al.7 on an alumina of similar density and composition. This alumina had
much smaller grain size and exhibited an invariant toughness response, taken here as
representative of the base toughness of the large grain alumina, such that the values
of Tbase =c ¼ 3:68, and Tbase =x ¼ 36:0 determined for the small-grained material
were deemed applicable for the large-grained material. As both materials were
close to fully dense, the applied and residual geometry terms were assumed to be
unaltered by grain size, c=cbase ¼ 1 and x=xbase ¼ 1. As perhaps anticipated by Fig. 7,
the toughness response passes through range of values on crack extension and the
stick-slip crack extension is evident as increase-decrease sequences in toughness with
crack length. Overall, the toughness behavior of the alumina can be characterized as
lying within a wedge or triangle pointing to longer crack lengths, bounded by an
invariant upper value and with lower values significantly less than that of the base
material.
The data of Fig. 8 are very different from those of Fig. 6 and suggest a much
more variable interaction between an extending crack and the microstructure over
the range of crack extensions examined. The cordierite composite data suggest an
Figure 8. Plot of relative toughness as a function of crack length for a large grain size polycrystalline
alumina There is considerable variability in the toughness response, although it appears that the toughness
is increased by about a factor of 1.5 over fine grained material, approximately 1.35 MPa m1=2 for this
material in air.
66 Robert F. Cook
almost saturated toughening response with little variability, indicative of the influ-
ence of many microstructural interactions (whisker-related frictional interlocks)
accumulated over the crack face. The alumina data on the other hand suggest a
still-accumulating toughening response, with great variability, indicative of the point-
to-point stochastic nature of relatively few microstructural interactions (grain-related
frictional interlocks) over the crack face. An interpretation of the data is that the
toughness of the least tough interface in the alumina is about 1.5 MPa m1=2 , reflect-
ing grain boundary fracture. Most of the indentation cracks start out this way and
are arrested by grains in the crack path. As the applied stress is increased, the net
stress-intensity factor (and thus apparent toughness) increase at the arrested crack
length until the instability condition, about 4 MPa m1=2 , reflecting fracture of
the grains or obstacles, is exceeded and the crack extends until arrested again. The
sequence then repeats, but as the crack has extended there are now some toughening
interactions over the crack plane and the starting point is thus at a greater tough-
ness value than the previous sequence; the ending point is at about the same
toughness value, reflecting the maximum toughness achievable in the microstructure.
On continued sequential extension in this way, the value of the toughness at the
starting (arrest, stick) part of the sequence approaches the value of the ending
(extension instability, slip) part and the variability in the toughness with extension
decreases and settles on the upper-bound level associated with homogeneous, steady-
state microstructural interactions over the crack plane. Long cracks associated with
large indentation loads approach this behavior, as seen in Figs. 7 and 8.
5. CONCLUSIONS
Crack extension from indentation flaws under the influence of an applied stress
exhibits a great sensitivity to microstructural effects. This sensitivity is not observed
in inentation-strength tests or long crack toughness tests, as the crack extension
during failure or creation of the long crack, respectively, accumulate large numbers
of microstructural interactions that act to average or saturate the perceived tough-
ness value. Direct observation of indentation crack extension can thus be used to
ascertain the probable toughening mechanisms in a microstructure, the length scales
associated with toughening and the point-to- point variability in these effects, as well
as providing an easy method of evaluating material toughness with great specimen
economy.
The results presented here on cordierite composites and polycystalline alumina
typify the enhanced extension ratios (>2.52 predicted for an ideal indentation re-
sponse) and are reasonably interpreted within the established framework of the
cumulative toughening actions of obstacle (whisker or grain)- related crack face
frictional interlocking. Such toughening relies on the natural inhomogeneity of the
microstructure; an implication here is that significant variability of defect responses
to applied stress is inherent in such materials. However, explanations remain to be
found for the non-ideal extension trajectories and suppressed extensions exhibited by
many of the materials in Table 1 that have no such inhomogeneities (glass, Si3 N4 ,
TZP). Such non-idealities point to the need for examination of the crack driving force
K terms in Eq. 1 associated with the contact mechanics, rather than the T terms
associated with the toughening mechanics.
REFERENCES
1. D.B. Marshall, B.R. Lawn and P. Chantikul, Residual stress effects in sharp contact cracking. II.
Strength degradation., J. Mat. Sci., 14, 2225 35 (1979).
2. R.F. Cook, E.G. Liniger, R.W. Steinbrech, and F. Deuerler, Sigmoidal Indentation Strength Char
acteristics of Polycrystalline Alumina, J. Amer. Cer. Soc. 77 303 314 (1994).
3. P. L. Swanson, C. J. Fairbanks, B. R. Lawn, Y. W. Mai and B. J. Hockey, Crack interface grain
bridging as a fracture resistance mechanism in ceramics. I. Experimental study on alumina, J. Amer.
Cer. Soc., 70 279 89 (1987).
4. R.M. McMeeking and A. G. Evans, Mechanics of transformation toughening in brittle materials,
J. Amer. Cer. Soc., 65, 242 6 (1982).
5. R. Tandon, D.J. Green and R.F. Cook, Strength Variability in Brittle Materials with Stabilizing and
Destabilizing Resistance Fields, Acta metall. 41 399 408 (1993).
6. B.R. Lawn, A.G. Evans, and D.B. Marshall, Elastic/Plastic Indentation Damage in Ceramics: The
Median/Radial Crack System J. Amer. Cer. Soc., 63 574 581 (1980).
7. L. M. Braun, S. J. Bennison and B.R. Lawn, Objective evaluation of short crack toughness curves
using indentation flaws: case study on alumina based ceramics, J. Amer. Cer. Soc., 75, 3049 57
(1992).
8. D. B. Marshall, Controlled flaws in ceramics: a comparison of Knoop and Vickers indentation,
J. Amer. Cer. Soc., 66, 127 31 (1983).
9. R. Tandon and R.F. Cook, Indentation Crack Initiation and Propagation in Tempered Glass,
J. Amer. Cer. Soc., 76 885 889 (1993).
10. L. M. Braun and R.F. Cook, in Science and Technology of Zirconia V, edited by M.J. Bannister and
R.H.J. Hannink (Technomic Press, 1993), pp. 386 400.
INSTRUMENTED HARDNESS
TEST ON ALUMINA
CERAMICS AND SINGLE
CRYSTAL WITH SPHERICAL
INDENTER
Shuji Sakaguchi, Kiyoshi Hirao, Yukihiko Yamauchi

and Shuzo Kanzaki*
1. INTRODUCTION
Ceramic materials are expected to be applied to the structural parts, especially to

tribological application, as they have high hardness and high elastic constants. For
this application, it is important to know the properties of mechanical behavior on the
surface of ceramics. Instrumented indentation is one of the tests that can explain the
mechanical properties on the surface, as it expresses the relation between the pene-
tration depth of the indenter and the applied force.1,2 For metallic materials, the
standard test method in instrumented indentation is already published.3 We have
tried the instrumented indentation tests on ceramic materials. Figure 1 is a typical
example of the loading-unloading curve by this test on alumina ceramics.
If the test force is small enough, the indenter contacts the surface elastically, and
no indent shall be made and no energy is consumed for making indent. It means if we
increase the test force, the consumed energy ratio for making indent can increase. With
this assumption, we tried the measurement of instrumented hardness with conical
* Shuji Sakaguchi, Ceramic Research Institute, National Institute of Advanced Industrial Science and
Technology, Nagoya, Japan 463 8560. Kiyoshi Hirao, Yukihiko Yamauchi and Shuzo Kanzaki, Synergy
Materials Research Center, National Institute of Advanced Industrial Science and Technology, Nagoya,
Japan 463 8687.
69
70 Shuji Sakaguchi et al.
1.2 loading
unloading
1
1.1mm
0.8
Load (N) 0.6
0.4
0.2
0
0 0.5 1 1.5 2 2.5
Displacement (μm)
Figure 1. Typical result of instrumented indentation on alumina ceramics, measured with Vickers
indenter.
indenters (Berkovic, Knoop and Vickers), and we obtained the relation between the
applied test force and the energy consumption ratio for making indents. Figure 2
shows some results of the test force dependence of the consumed energy ratio. In some
results obtained with Knoop indenter, the results follow this expectation, but the test
force dependence of the energy consumption ratio was not clearly observed. The
scattering of the data is relatively large, as the applied test force was small. We could
not obtain the grain size dependence of the energy consumption clearly. It was
considered that the tip of the indenter was too keen to obtain the elastic contact on
these materials. Spherical indenters shall be used for this discussion.
In this study, we tried to use spherical diamond indenters with different tip
radius on the instrumented indentation tests. We tried to find the test force depend-
ence of the surface fracture with these indenters, and the grain size dependence of the
instrumented indentation results on ceramic materials.
2. EXPERIMENTS
2.1. Testing conditions

Recording hardness test machine supplied by CSIRO, Australia (Type: UMIS-2000)
was used for this study. Seven kinds of spherical indenters with different tip radius
(1, 2, 5, 10, 20, 50 and 100 mm) were used. Maximum test force was chosen as 20, 50,
100, 200 and 500 mN. Five tests were carried out in the same test condition, and
some scattered data that might be caused by some large fracture on the surface were
neglected. The data were averaged and the consumed energy was calculated from the
loading-unloading curve.
2.2. Specimens
The specimen was alumina sintered body with purity of 99.9%.4,5 The average grain
size of as-received sample was 1:1 mm. It was annealed in air in three conditions for
Instrumented Hardness Test on Alumina Ceramics 71
0.5 0.5
0.4
0.4
0.3
0.3 0.2
Ratio
Ratio
0.2 0.1
0
0.1
Berkovic −0.1 Knoop
1.1μm 1.1μm
0 −0.2
0.01 0.1 1 0.01 0.1 1
Force (N) Force (N)
0.5 0.5
0.4 0.4
0.3 0.3
Ratio
Ratio
0.2 0.2
0.1 0.1
Berkovic Knoop
5.7μm 5.7μm
0 0
0.01 0.1 1 0.01 0.1 1
Force (N) Force (N)
0.5 0.5
0.4 0.4
0.3 0.3
Ratio
Ratio
0.2 0.2
0.1 0.1
Berkovic Knoop
Sapphire Sapphire
0 0
0.01 0.1 1 0.01 0.1 1
Force (N) Force (N)
Figure 2. Examples of the relation between the test force and the energy consumption ratio on instru
mented indentation with Berkovic or Knoop indenter.
the grain growth. The specimens with average grain sizes of 1.8, 4.4 and 5:7 mm were
prepared. The mechanical properties of these specimens are shown in Table 1. The
Vickers hardness was not changed with the grain growth. For the comparison, the
c-plane of the alumina single crystal (sapphire) was also tested. The Vickers hardness
of the sapphire was 19.0 GPa with the standard deviation of 0.5 GPa. It was a little
harder than the sintered alumina.
Table 1. Four point bending strength, fracture toughness measured by single edge pre
cracked beam (SEPB) method, and Vickers hardness of the tested alumina specimens.
Standard deviations are in brackets.
Grain size (mm) Strength (MPa) KIC (MPa m1=2 ) HV (GPa)
1.1 652 (39) 3.00 (0.31) 18.3 (0.4)

1.8 555 (40) 3.20 (0.14) 18.7 (0.6)
4.4 415 (23) 3.95 (0.24) 17.3 (0.5)
5.7 369 (21) 4.21 (0.05) 17.6 (0.6)
3. RESULTS
Figure 3 shows the results of the energy consumption ratio of every specimen on the
indenter of every tip radius. On the indenters of 20 mm and 50 mm, the contacts were
almost elastically and no energy consumption was found. However, in the test with
20 mm indenter the scattering of the data became larger. On the indenters of 5
and 10 mm, energy consumption ratio increased continuously with increasing the
60 50
Energy Consumption Ratio (%)
1.1
50 40 1.8
4.4
5.7
40 30 Sap
30 20
1.1
1.8
20 4.4 10
5.7
Sap
10 0
Indernter R = 5 μm Indernter R = 10 μm
0 −10
10 100 1000 10 100 1000
Maximum Test force (mN) Maximum Test force (mN)
40 40
1.1 1.1
30 1.8 30 1.8
4.4 4.4
5.7 5.7
20 Sap 20 Sap
10 10
0 0
−10 −10
−20 −20
10 100 1000 10 100 1000
Figure 3. Energy consumption ratio of each test condition with spherical indenters of 5, 10, 20 and 50 mm
in radius (‘Sap’ is sapphire c plane).
maximum test force. The grain size dependence of the energy consumption was not
observed.
If the energy consumption increased abruptly at a certain test force, we can
estimate a limit of the surface damage. However, the energy consumption increased
gradually with increasing the test force. That means the surface damage is not made
at a certain stress but accumulated with increasing the stress.
We could not see the grain size dependence of the energy consumption. The
surface damage with compressive stress can be independent of the grain size, as it can
be expected from the results of the Vickers hardness.
4. DISCUSSION
For discussing all the data, the results were plotted as the relation between the
average contact stress of the indenter calculated with the equations for the elastic
contact, and the energy consumption ratio. Table 2 shows the average stress in the
contact circle from the equation of the elastic contact on every test condition. The
maximum stress, which appears at the center of the contact circle, is 1.5 times larger
than the average stress. Figure 4 shows the plot of all the data. The plots can be
divided into three regions. (1) Calculated stress is below 12 GPa. No energy con-
sumption was found. (2) Calculated stress is between 12 GPa and 18 GPa. The
energy consumption was still around zero, but the scattering of the data becomes
larger. (3) Calculated stress is over 18 GPa. The energy consumption increased
simply with increasing the applied stress.
These three stress regions correspond to the following conditions; (1)
The compressive stress is below the hardness value on all the contact area. (2) The
average compressive stress is smaller than the hardness value, but the maximum
stress is larger than the hardness. The contact area may partly damaged. (3) The
average stress is larger than the hardness. The indent size shall be larger than the
calculated elastic contact size.
With these explanations, we can assume the reason why the scattering of the
energy consumption became larger in the intermediate stress region. It is schematic-
ally shown in Figure 5. In this intermediate stress condition, the average stress is
smaller than the hardness. Then the indent with the size of the contact circle cannot
be made. However, the stress at the center of the contact circle is larger than the
hardness. The surface area at the center shall be damaged, or micro-fractured, by this
contact. It is not an elastic contact, but it cannot make a clear indent. That can be a
reason why the energy consumption was not clearly increased but only the scattering
of the data was increased.
Table 2. Calculated compressive stress (GPa) on elastic contact in every condition.
Test Force (mN)

Indenter Radius (mm) 20 50 100 200 500
5 16.1 21.9 27.6 34.8 47.2

10 10.2 13.8 17.4 21.9 29.7
20 (6.4) (8.7) 11.0 13.8 18.7
50 (3.5) (4.7) 5.9 7.5 10.2
60

50
40
30
20
10
1.1
0 1.8
4.4
−10 5.7
Sap
−20
0 10 20 30 40 50
Average stress (GPa)
Figure 4. All the data are plotted with the relation between the calculated contact stress and the
energy consumption. The results can be divided into three regions; (1) 0 12 GPa, (2) 12 18 GPa, (3) over
18 GPa.
Figure 5. Indentation damage under intermediate test force. The area around the center of the
contact circle is damaged. It is not an elastic contact, and it can make a scattering of the energy
consumption.
5. CONCLUSION
Instrumented indentation tests were carried out on alumina ceramics with different
average grain size, using spherical indenters.
(1) We cold not found the average grain size dependence of the energy consumption
for making indents on alumina specimens.
(2) The stress value for making surface damage is related to the Vickers hardness
value. It can be divided into three levels; 1. contacting elastically, 2. contact area
is partly damaged, 3. contact area is fully damaged and indent is made.
(3) The consumption energy ratio for making indents increased continuously with
increasing the maximum test force, if the applied stress is more than the Vickers
hardness value.
6. ACKNOWEDGMENTS
This work has been supported by NEDO, as part of the Synergy Ceramics Project
promoted by METI, Japan. The authors are members of the Joint Research Con-
sortium of Synergy Ceramics.
REFERENCES
1. M. Sakai and S. Shimizu, Elastic, plastic, and viscoelastic contact problems in axisymmetric indenta
tion, Ceramic Transactions (The American Ceramic Society), vol. 133, 105 114, (2002).
2. W. C. Oliver and G. M. Pharr, An improved technique for determining hardness and elastic modulus
using load and displacement sensing indentation experiments, J. Mater. Res., 7, 1564 1583, (1992).
3. ISO 14577 1, Metallic materials Instrumented indentation test for hardness and materials
parameters Part. 1: Test method, (2002).
4. S. Sakaguchi, The relation between the damage tolerance and the R curve behavior of alumina
ceramics, Fracture Mechanics of Ceramics (Kluwer Academic / Plenum Press), vol. 13, 229 234,
(2002).
5. S. Sakaguchi, R curve measurement of alumina using straight notched or indented bending bars,
Fracture Mechanics of Ceramics (Plenum Press), vol. 11, 39 51, (1996).
CONTROLLING THE
FRAGMENTATION
BEHAVIOR OF STRESSED
GLASS
Rajan Tandon and S. Jill Glassz
1. INTRODUCTION
Inducing compressive surface stress profiles in brittle materials is a well-known ap-

proach for strengthening. The compressive stress inhibits crack initiation and propa-
gation. The effect has been observed for tempered and ion-exchanged glasses,1–4 and for
oxide ceramics.5,6 While it is generally accepted that the magnitude of the stress and its
depth determine the strength response, it has recently been demonstrated that the shape
of the compressive stress profile can radically alter the strength distribution.7 For
tempered glasses, the role of the internal tensile stress in causing fragmentation is well
known,8 although it is not possible to predict the extent of fragmentation.9
Relatively less attention has been paid to the role of the internal compensating
tensile stress on the fracture process for ion-exchanged glasses. Zijlstra and Burg-
graaf10 described an approach to reduce the sharpness of fragments of ion-exchanged
glass. They found a powdering effect at the edges of the fragments when the
maximum in the compressive stress profile was shifted approximately 40 mm below
the surface, and the difference between the stress maximum and the surface stress was
> 147 MPa. This powdering effect rendered the fragments harmless. Bakioglu et al.11
theoretically considered the problem of a partially closed crack that penetrates
through the surface compression region. They presented fracture mechanics solutions
to analyze the spontaneous propagation of this crack in the internal tensile field aided
z
Rajan Tandon, S. Jill Glass, Ceramic Materials Department, Sandia National Laboratories,
Albuquerque, NM 87185, USA
77
78 Rajan Tandon et al.
by environmental effects. Donald and Hill12 reported the preparation and properties
of lithium magnesium alumino-silicate glasses. The lithium in this glass structure is
exchanged by sodium ions in solution, leading to compressive surface stresses.
Unusual properties, with the surface being in tension due to extreme stress relax-
ation, were reported. They also observed fragmentation of this glass after biaxial
flexure testing, with the higher strength samples possessing smaller fragments.
Bouyne and Gaume13 are developing a stress-intensity factor approach to predict
fragmentation, and have concluded that the elastic energy stored in the tensile stress
region is the parameter controlling the fracture.
The motivation of the present study is to identify the processing conditions, viz.,
exchange time, temperature, and glass size that produce a desired combination of
strength, fragmentation behavior, and reliability. Reliability here encompasses the
repeatability and predictability of the fragmentation process in addition to the
mechanical strength aspects. In this work, we attempt to relate the strength, inden-
tation, and central tension stress measurements of an ion-exchanged glass to its
fragmentation behavior. An energy analysis is proposed to relate the fragment size
to measurable parameters. The fracture surfaces of strengthened glasses are exam-
ined in order to better understand the fragmentation process.
2. EXPERIMENTAL PROCEDURE
2.1. Materials and Exchange Treatment
The material chosen for this study was an alumino-silicate glass (Corning Code
0317). The approximate composition of the glass (in wt%) was: 61.96% SiO2 ,
16.5% Al2 O3 , 11.88% Na2 O, 3.34% MgO, 3.41% K2 O, 1.75% Li2 O, 0.67% TiO2 ,
0.41% CaO and rest Fe2 O3 . Glass disks of nominal diameter of 2.54 cm, and three
different thickness values (1 mm, 1.8 mm, and 2.2 mm) were used for this study. The
process of ion-exchange was used to incorporate the residual stress profiles in the
glass. For Code 0317 glass composition, this process consists of exposing the glass to
a molten solution of potassium nitrate. Due to a chemical potential difference, the
Naþ ions in the glass are released into the molten salt, and the Kþ ions occupy their
place in the structure. The bath temperature was controlled to + 28C, and two
different temperatures of exchange, 4508C and 5008C were employed. Since both
these temperatures are below the glass transition temperature (strain point ¼ 5768C),
the Kþ is not accommodated in the structure of the glass, and the exchanged layer
tends to have a larger volume than the bulk. This leads to the development of a
surface compressive stress in the glass, and to a compensating tensile stress in the
interior. The ion-exchange was conducted for time intervals of 3, 6, 12, 24, 48 and 96
hours. The depth of exchange in some samples was measured using a microprobe,
and using the known value of the activation energy for diffusion,13 the depths for
various treatments were estimated. The depth of the Kþ ions into the glass is an
important parameter as the extent of the compressive stress profile is generally
between 0.9-0.99 times the exchange depth. Figure 1a provides a schematic of the
ion-exchange process, and Fig. 1b depicts the stress profiles expected in the process.
Fig. 1b also includes a stress profile that may be expected when the high surface stress
undergoes relaxation due to visco-elastic structural rearrangement.
Figure 1a. Schematic of the ion exchange process is shown. The large Kþ ions from the salt exchange
with the Naþ ions in glass.
Figure 1b. Schematics of the normal and relaxed stress profiles are shown. In the relaxed profile, the
maximum in the surface compression has moved away from the surface of the sample.
2.2. Central Tension Measurement

The measurement of compressive stress profiles in ion-exchange glasses is very tedious,
and was not attempted for this study. However, the values of the compensating tensile
stress in the bulk were obtained by using a set-up described in Ref. 14. A white light
source was viewed through a crossed polarizer and analyzer set to obtain extinction.
The samples were then introduced such that the light beam passed along the disk
diameter. The birefringence introduced due to the internal tension led to the extinction
condition being disturbed, and a particular fringe color was seen in the eyepiece. A
Soliel-Babinet compensator and additional quarter wave-plates were then introduced
in the light path to re-obtain extinction. Knowing the amount of retardation induced by
the compensator-wave-plate combination to attain extinction and the stress optic
coefficient of the glass, the central tension values in the samples were calculated.
2.3. Bi-Axial Strength Tests, Indentation Tests, and Fragmentation Observations

The strength values of the disks exchanged under various test conditions were measured
using a ring-on-ring biaxial fixture made in accordance with ASTM C1499. For each
test condition, one or two samples were fractured, and strength determined. In order to
establish a baseline, the strength of five un-exchanged glass disks was measured. To
probe the changes in the surface stress as a function of exchange time, Vickers inden-
tation experiments were conducted. A Vickers indenter was loaded onto the surface in
air at 40N and 50N peak applied loads, held for fifteen seconds, and then removed. The
indentation site was examined within an hour after load removal. Fragmentation
experiments on the test samples were conducted by progressively increasing the Vickers
indenter load till spontaneous fragmentation (or simple fracture) of the glass occurred.
There were instances when the median crack pop-in and propagation led to the sample
separating into two pieces. In most cases, however, fragmentation with multiple crack
branching was observed. No records of the load to cause fragmentation for particular
exchange conditions were kept. The fragmented pieces were held together by tape. The
fragmented pieces, and fracture surface of the fragments were examined using an
optical microscope. For comparison, the fracture surface of a tempered glass piece
that fragmented spontaneously was also observed. The magnitude of the stress in this
piece of glass prior to fracture is unknown.
3. RESULTS AND ANALYSIS
3.1. Central Tension Measurements

Under crossed polarized light, it was observed that the color in the central region of
the sample was fairly uniform, and extinction was obtained uniformly across the
sample cross-section. These observations indicate that the tensile stress was uniform
in the test samples. Samples that exhibited asymmetry in the tensile stress distribution
were removed from the set, and were not tested further. The values of central tension
measured for various exchange conditions and glass thickness values are shown in
Fig 2. Each point in the figure is an average value of three-four measurements made
on the same sample, and for two-three different samples. The scatter in the measured
stress values was <5%, and the lines drawn through the data are aids to the eye. The
open and solid symbols represent exchange temperature of 4508C and 5008C respect-
ively. For all conditions of exchange, the measured central tension increases with
increasing exchange time. For identical exchange conditions, the central tension (st )
Figure 2. Central tension values for various exchange times, temperatures, and glass thickness values
are shown.
values have the following distribution: st, 1 mm > st, 1:8 mm > st, 2:2 mm , i.e., the thinner
the sample, the higher the value of central tension. For the same thickness, the
exchange process at higher temperature yields the higher tension value.
3.2. Biaxial Flexure Strength Measurements
The results of the ring-on-ring biaxial flexure tests, as a function of increasing ion-
exchange time, are shown in Figure 3. The open and closed symbols represent the
exchange condition at 4508C and 5008C respectively, and the lines are aids to the eye.
Due to the large number of conditions that had to be studied, only one-two samples
per data point could be strength tested, and no statistical variation was calculated. It
can be observed that for equivalent time and thickness, the strength of any sample
exchanged at 4508C is higher than the sample exchanged at 5008C. Incidentally, the
strength values for all the samples exchanged at 4508C exceed those exchanged at
5008C. At any given temperature of exchange, the strength (sf ) follows the following
distribution: sf , 1 mm > sf , 1:8 mm > sf , 2:2mm , i.e., the thinner the sample, the higher its
strength. The strength values decrease with increasing time of exchange for both
exchange conditions, and for all sample thickness values. The unexchanged samples
have an average strength of 121 MPa.
3.3. Indentation Experiments
The results of observation of the Vickers indentation sites for a 40N peak load for
various times of ion-exchange are shown in Fig. 4. The indentation impression size
Figure 3. The bi axial strength results as a function of ion exchange times for various conditions
are shown.
Figure 4. Observation of indentation sites (40N) for various times of exchange for a sample exchanged
at 5008C.
(edge to edge) for this load is approximately 150 mm, and all the pictures are at the
same magnification. Previous work on the base glass16 has shown that for peak loads
exceeding 20 N, significant radial and lateral cracking may be expected at the
indentation site. However, for the 3 and 6 hr. exchange samples, no cracking is
detected. Traces of radial cracks emanating from the corners of the indenter can be
seen for the 48 and 96 hr. exchange samples. Sub-surface lateral cracks around the
indentation site are observed for the 12 96 hr. exchange samples. One quadrant of
the lateral crack has grown to intersect the surface, and caused a surface spall for the
96 hr. sample. The results of indentation testing for 50N load are not shown here;
however, it was observed that all the indentation sites developed large surface spalls
within an hour of load removal.
3.4. Fragmentation and Fracture Surface Observations

The central tension values for the 1.8 mm glass thickness as a function of the ion-
exchange time for both exchange temperatures are plotted in Figure 5a. As pointed
out in Section 3.1, for a given thickness, central tension values increase with increas-
ing time and temperature. These samples are then subjected to increasing indentation
loads, till failure occurs. The fragmentation observations are presented in Figure 5b.
For the 3 hr. exchange time, neither of the samples shows fragmentation, and one
pair of the indentation cracks extends through and across the sample. For 6 hr.
exchange, the 5008C samples exhibit branching from the indentation cracks. The
tension in the 4508C sample, however, is sufficient to only cause both pairs of
indentation cracks to propagate through the sample. Increasing the exchange time
further to 12 hr. leads to the onset of branching for the 5008C sample closer to the
origin of fracture, and it appears that further crack bifurcations occur at progres-
sively shorter distances. The 4508C sample exchanged for 12 hr. exhibits crack
branching only once from indentation crack pairs. Increasing the exchange times
Figure 5a. Central tension values as a function of exchange time for a 1.8 mm thick glass are shown.
Figure 5b. Observations of fragmentation behavior corresponding to the sample set in Fig. 5a.
further, and thereby increasing the central tension accentuates the branching process,
i.e., crack branching starts to occur closer to the origin, and the fragment size
distribution is smaller. For the 96 hr. exchange samples, the fragment size is almost
uniform, and is < 1 mm for both exchange temperatures. It was also noted that the
edges of many of the samples are powdered (fragment size< 0:1 mm) even when the
fragments adjacent to them were > 2 mm in size.
Figure 6a is a comparison of the central tension values as a function of ion-
exchange time for a 1.8 mm thick sample exchanged at 4508C, with that of a 2.2 mm
thick sample exchanged at 5008C. It is observed that for short exchange times, the
tension values are lower in the thinner glass, but approach that of the thicker glass
with increasing time. The fragmentation observations are presented in Figure 6b. For
the 3 hr. exchange samples, the tension value in neither sample is adequate to cause
branching. For the thicker sample, branching occurs for a 6 hr. exchange, whereas
the cracks in thinner samples start branching for the 12 hr. exchange condition. With
increasing tension in both samples, the degree of fragmentation increases. Longer
exchange times at the (different) exchange temperatures lead to conditions where the
central tension values are nearly equal. This is observed for the 48 hr. and 96 hr.
samples. The fragmentation behavior of these samples becomes very similar, more so
for the 96 hr. exchange samples.
The fracture surface observations for the tempered glass far from the origin of
failure are shown in Fig. 7. The thickness of the sample, t, is 6 mm, and the
photographs capture nearly the entire thickness. The dotted line on the picture
Figure 6a. Central tension values as a function of exchange time for a 1.8 mm, 4508C exchange sample
compared with 2.2 mm, 5008C exchanged sample.
500⬚ C
2.2 mm
450⬚ C
1.8 mm
3 hr. 6 hr. 12 hr. 24 hr. 48 hr. 96 hr.

Figure 6b. Observations of fragmentation behavior corresponding to the sample set in Fig. 6a.
Figure 7. Fracture surface views of two fragments, A and B, for tempered glass.
shows the boundary between the two fragments, A and B. There is a region near each
of the surfaces where the fracture process appears to be different than in the core of
the sample. This region occupies 13% of the sample thickness on each side. For
normal tempering conditions, tempered glass has 46% (23% on each side) of the
thickness under compression. Wallner lines are visible on the fragment on the right,
and the crack propagation direction (marked by the arrow) is deduced from these
lines. In the center of the sample, a region where the crack front underwent the
mirror-mist-hackle transformations, as it accelerates under increasing stress inten-
sity, are clearly visible. For tempered glass, the central region is the location of
maximum tension. The fragments pictured were still adherent to each other across
the second fracture plane that is marked by the dotted line.
The fracture surfaces for ion-exchanged glass (5008C, 6hr., central ten-
sion ¼ 45 MPa) near (Figure 8a), and far (Figure 8b) from the fracture origin are
shown. The entire thickness, t ¼ 1:8 mm, is pictured in Figure 8a. At the bottom right,
the plastic deformation and the lateral and median vents from the indentation crack
that started the fragmentation process can be seen. Wallner lines at locations where the
crack front interacted with the lateral vents can be seen. A second, fainter, series of
Wallner lines which are semi-elliptical can be seen just further from the major lines.
The direction of the crack velocity is marked. At a distance of about 1.8 mm from the
median vent plane, the beginning of mist formation is observed. The mist area then
extends and spreads across the thickness of the sample. Figure 8b shows a continu-
ation of this fracture path. Here the mist region gradually changes into hackle. The
crack branched macroscopically at the left edge of this photograph. There appears to
be a very strong interaction of the crack front with the zone of compression as
evidenced by the sharp curvature of the Wallner lines at the top and bottom.
3.5. Analysis
A first estimate of the size of the fragments of ion-exchanged glass may be obtained
by equating the stored tensile strain energy in the body to surface energy of the new
area created. Warren9 has presented a similar analysis for thermally tempered glass.
We begin by making several assumptions: (a) that only the tensile strain energy is
used in the fragment process, (b) that all the tensile strain energy is used up in the
Figure 8 a, and b. Fracture surface views, near the origin (a), and far from the origin (b) for ion exchanged
glass.
Figure 9. Schematic of an idealized remnant shape used in the energy analysis.
process to create new surface area (i.e., effectively ignoring kinetic and sonic effects,
and implicitly assuming that the remnant is stress-free), and (c) that all remnants are
identical in size and shape. An idealized remnant is depicted in Figure 9. An
assumption with other geometrical cross sections (a hexagon for instance) does not
change the material dependences derived. Let st be the central tension, and let d be
the depth of the compressive profile. Hence the energy in a unit volume, dV, of the
fragment is DUE , which is given by
1X
DUE ¼ sij «ij DV (1)
2
Now since the state of tension is equi-biaxial, with other components assumed to be
zero, we obtain
E E
sxx ¼ syy ¼ «xx ¼ «yy ¼ st (2)
(1 v) (1 v)
where E, v are the elastic modulus and Poisson’s ratio. Substituting in Eq. 1, we
obtain
(1 v) 2
DUE ¼ st dxdydz (3)
E
Integrating to obtain the total tensile strain energy in the fragment, we obtain
Z ðZ2t dÞ Zx Zx
(1 v) 2
UE ¼ DUE ¼ st dxdydz (4)
E
dV ð2t dÞ 0 0
Equating UE to the total energy of the new surfaces created, US , we obtain
(1 v) 2 2
UE ¼ x st (t 2d) ¼ US ¼ 4xtg (5)
E
where g is the surface energy.
Simplifying, we obtain
2
2Eg 1 t K1c t
x¼2 ¼ 2(1 þ v) (6)
(1 v) s2t (t 2d) st (t 2d)
where K1c is the fracture toughness of the material. The resulting expression predicts
that the size of the fragment scales as tension 2 , and depends (weakly) on the
thickness of the sample.
4. DISCUSSION
4.1. Tensile Stress and Bi-axial Strength Measurements

The migration of the Kþ ions into the glass structure is diffusion-controlled, and the
depth of penetration is dependent on both time and temperature. The regions with
excess Kþ ions have a higher volume at equilibrium (a longer in-plane length) when
the process occurs below the glass transition temperature as in this case. The
chemical composition, and hence the equilibrium length of the core layers is un-
altered. Therefore, the actual state of the body is a length in between the two
described above. This sets up a state of residual compression in the Kþ rich layers,
and in response, an equi-biaxial state of compensating central tension to ensure force
equilibrium across the section. We find that the measured internal tension increases
with increase in exchange time for all conditions (Fig. 2). This implies that as more
Kþ ions diffuse into the sample, the net compressive force increases, leading to an
increase in the compensating tension. It is also observed that for identical conditions
of exchange viz., same temperature and time, the tension in the thinner samples is
highest. Similar conditions of exchange must lead to identical flux of Kþ ions, and
similar compressive stress profiles. Therefore, the thinner samples, which have the
least cross-sectional area, must develop a correspondingly higher tensile stress.
Interestingly, the bi-axial tests show a continuous drop in strength with increasing
exchange time. We expect a higher ion flux and deeper penetration with increasing time,
and therefore, a rise in the net compressive force. Since the compressive stress acts to
close the surface cracks, we expect an increase in strength rather than a decrease. These
results can be understood if we invoke the phenomena of stress relaxation17. The highly
stressed surface of the glass undergoes a structural relaxation and the values of the near
surfaces stress will fall, in some cases to near zero values. Since the flaws causing failure
are at or near the surface, this stress relaxation leads to a loss in strength. We observe
that the strength for samples treated at 5008C are significantly lower than those treated
at 4508C. Since the relaxation phenomena is activated by temperature and stress, higher
relaxation rates at higher temperature may be expected. However, it may be noted that
even the lowest strengths recorded (420 MPa) are three times the as-received values.
Estimates of the minimum surface compression values for these samples may be made
by assuming that the strength is equal to the as-received strength plus the value of the
surface compressive stress. For the 4508C, 3 hr. exchange, 1.8 mm thick sample, this
leads to a value of 780 MPa (-ve sign implies a compressive stress). We also observe
that for each condition, the thinner samples have the highest strength. This may be due
to the steeper drop-off of the applied flexure loading across the potentially failure
causing flaw in the thinner samples. A steeper drop-off would lead to a lower value of
net stress intensity factor acting on the crack for any given applied stress, and conse-
quently to higher required stress for failure.
Another interesting feature is that although the strength falls, and the peak
values of the compression relax for higher exchange times, the internal tension
increases monotonically. This implies that the rate of penetration of Kþ ions is
such that despite the relaxation, the summation of the compressive stress, across
the cross-section increases, i.e., the compressive force increased.
4.2. Indentation Response

Indentation on the base glass at a 40 N load leads to severe radial and lateral
cracks.16 The complete absence of any median, radial or lateral cracks at the inden-
tation site for the 3 and 6 hr. exchanged samples (Fig. 4) is indicative of the strong
effect of the residual compression on crack initiation. Unpublished work by the
authors has shown that the driving force for lateral cracks is unaffected by the resi-
dual compressive stress. Therefore the absence of lateral cracks for 3 and 6 hr.
exchange is surprising. However, lateral and small traces of radial cracks appear
for higher exchange times. This indicates that the surface stress relaxes with increas-
ing exchange time, and supports the hypothesis advanced for understanding the bi-
axial strength observations in Sec. 4.1. Observations of the 50 N indentation sites
reveal extremely severe lateral cracking and spalling, and will not be described here.
4.3. Fragmentation and Fracture Surface Observations

Fragmented samples, 1.8 mm thick, 5008C exchange for 6, 12 and 24 hr., held
together by tape are shown in Fig. 10. The tension values for these samples are
45 MPa, 55 MPa, and 75 MPa respectively. Observation of these samples and the
fracture surfaces lends insight into the fragmentation process. A crack (or pair of
cracks if the initiation is using an indenter) may penetrate the compressive zone, and
is then acted upon by the tensile stress. If the stress intensity factor exceeds the
material toughness, the crack propagates as an internal crack entirely in the zone of
tension. When the crack becomes sufficiently long, the net stress intensity factor
exceeds the branching constant, Kb , and macroscopic branching may occur. For the
6 hr. sample, two pairs of cracks have branched, and the number of crack branches
Figure 10. Comparison of crack paths in fragmented samples reveals details of the process.
range from 2-4. Since the 12 hr. sample has a higher value of tension, the value of Kb
is reached for shorter crack lengths, and branching occurs closer to the origin. This
process is further accentuated for the 24 hr. sample due to its highest tension value in
this set. It also appears that the second, third, and subsequent branching events occur
progressively closer, i.e., the branching length decreases with each successive branch.
The observed decrease in the fragment size as one translates from the origin to the
periphery of the sample could be due to this reduction in branching length. However,
it is possible that what appears to be a decrease in branching length is simply an
artifact of multiple crack propagation and intersection. Studies of larger samples
with low values of tension will be used to clarify this behavior.
The data presented in Fig. 5 and Fig. 6 underscore the importance of the central
tension value in order to define fragmentation. In Fig. 5, we observe that for samples
of identical thickness, higher tension values lead to higher fragmentation. In Fig. 6,
the differences in the central tension values are manifest in the fragmentation behav-
ior. For longer exchange times, the tension values become similar, and the fragmen-
tation behavior appears indistinguishable.
For Corning 0317 glass, the depth of Kþ ion diffusion for 24 hr. exchange at 5008C
is 180 mm.14 Using this value, and assuming that the diffusion distance (d) can be well
approximated by d 2(Dt)1=2 (where D is the diffusion coefficient, t is exchange time),
the value of D at 5008C was calculated. The value of the activation energy for diffusion
is known to be 113 kJ/mole,14 and the exponential relationship was used to calculate the
diffusion coefficient at 4508C. Using the value of D, the d for each condition was
obtained. These values, with the known value of tension, and the properties of the
glass (K1c ¼ 0:7 MPa:m1=2, n ¼ 0:25) were used in Eq. 6 to estimate the fragment size.
The results for the 1.8 mm thick glass series are shown in Table 1.
On comparing the predictions to the actual values of fragmentation, we find that
the calculation grossly underestimates the size of the fragments, especially for large
fragments. This is due to the assumption that all the energy in the tensile zone is
converted into surface energy, which overestimates the available energy. The assump-
tion may be accurate for very small fragments. This is because as the surfaces of a
fragment are stress free, the stress state may be expected in the volume as well
(somewhat analogous to the case of a thin material being in plane stress condition).
Indeed, the predicted values for the 48 and 96 hr. exchanged samples are very close to
the actual values (0.45 mm and 0.33 mm compared with 0.27 mm and 0.14 mm
predicted). However, for large fragments, complete energy release is extremely un-
likely, and observations of fragments in the polarized light reveal birefringence
indicative of residual stress.
Table 1. Predicted fragment size () using the energy balance approach.
Exchange time (hr.) st at 5008C (MPa) d500 (mm) (mm) st at 4508C (MPa) d450 (mm) (mm)
3 37 64 1.25 16 35 6.50
6 45 90 0.87 23 49 3.19
12 55 128 0.61 37 69 1.26
24 75 180 0.35 47 98 0.81
48 107 255 0.19 60 139 0.52
96 140 360 0.14 87 196 0.27
It was also noticed that the edges of many samples were powdered, even when
the adjacent fragments were rather large. This may be attributed to the stress wave
reflection from the edges interacting with the approaching crack front to magnify the
crack loading. It is also possible that ion-exchange from the edges leads to a localized
stress state with much higher tension than in the center, leading to powdering.
Observation of the fracture surface of the tempered glass (Fig. 7) indicates that
about 26% of the cross-section (13% each on the top and bottom) suffers a shear
fracture. Since the zone of compression occupies 46% of cross-section (23% at the top
and bottom), the cracks propagate not only in the tensile zone, but also in the area with
low compression. The shear fracture in the near surface region is due to the interaction
of the crack front with the compressive stress, and the introduction of mixed mode
loading. The main crack front undergoes the mirror-mist-hackle transition near the
center of the sample, which in tempered glass is the region of highest tension. The crack
branches in-plane to create the boundary between fragments A and B, and one portion
of the front continues to generate further misthackle features.
On the fracture surfaces of the exchanged glass (5008C, 6hr., central tension ¼
45 MPa) in Fig. 8 a, and b, the mirror-mist-hackle transitions are not localized to the
center portion. The mist starts from the far edge (away from the indentation), and
broadens to encompass almost the entire cross- section. This behavior is due to the
uniformity of the central tension. It is also observed that near-surface regions, nearly
300 mm on the near side, and 200 mm on the far side, are free of mist or hackle
features (this region is significantly larger than d 90 mm for this sample). The
Wallner lines clearly visible in this region are strongly curved, indicating that there
is a strong interaction between the crack front and the compressive zone. These
observations indicate that the compressive stress may be acting to pin the crack, and
its effectiveness may extend beyond its physical location in the sample.
5. CONCLUSIONS
For the glass system and the ion-exchange condition studied, the following conclu-
sions may be drawn from the data presented in this study:
1. Increasing the ion-exchange time leads to a decrease in the surface compressive
stress, but to an increase in the net compressive force via an increase in the
exchanged depth.
2. Compressive stress distributions raise the strength over the base glass by as much
as a factor of seven, but near surface stress relaxation processes lead to some
lowering of this enhanced value (to nearly three times of the base glass).
3. Compressive stresses are found to subdue median and radial cracking at inden-
tation sites, but the strain energy associated with the indentation leads to
enhanced lateral cracking.
4. The compensating tensile stress is found to increase monotonically with time of
exchange, and is higher for thinner samples under equivalent conditions of ex-
change. The tensile stress appears to be very uniform through out the cross- section.
5. Fragmentation in ion-exchanged glass is observed to be a multiple crack branch-
ing process yielding progressively smaller fragments further from fracture origin
6. In general, an increase in the tensile stress leads to higher fragmentation density,
and a more uniform fragment size.
7. An energy based analysis of the fragmentation process significantly underesti-

mates the observed fragment size.
8. Design of reliable, frangible glasses appears to be constrained by the following
factor: For increased fragmentation, a high value of tension is required. Higher
exchange times needed to obtain this value leads to strength loss due to surface
stress relaxation. This effect may be mitigated by using thinner sections. How-
ever, it must be ensured that the load bearing capacity of the system is adequate.
6. ACKNOWLEDGMENTS
Sandia is a multi-program laboratory operated by Sandia Corp., a Lockheed Martin

Co., for the United States Department of Energy, under Contract No. DE-AC04-
94AL85000.
REFERENCES
1. H. P. Kirchner, Strengthening of Ceramics, Treatments, Tests, and Design Applications. Marcel

Dekker, New York, (1979)
2. R. Gardon, Thermal Tempering of Glass, Glass: Science and Technology V. Academic Press, New
York. (1980)
3. S. Kistler, Stresses in Glass Produced by Nonuniform Exchange of Monovalent Ions, J. Am. Ceram.
Soc., 45 [2] 59 68 (1962)
4. M. E. Nordberg, E. L. Mochel, H. M. Garfinkel, and J. S. Olcott, Strengthening by Ion Exchange, J.
Am. Ceram. Soc., 47 [5] 215 219 (1964)
5. D. J. Green, A Technique for Introducing Surface Compression into Zirconia Ceramics, J. Am.
Ceram. Soc., 66 [9] C178 C179 (1983)
6. A. V. Virkar, J. L. Huang, and R. A. Cutler, Strengthening of Oxide Ceramics by Transformation
Induced Stresses, J. Am. Ceram. Soc., 70 [3] 164 170 (1987)
7. D. J. Green, R. Tandon, and V. M. Sglavo, Crack Arrest and Multiple Cracking in Glass through the
Use of Designed Residual Stress Profiles, Science, 283, 1295 1297 (1999)
8. J. M. Barsom, Fracture of Tempered Glass, J. Am. Ceram. Soc., 51 [2] 75 78 (1968)
9. P. D. Warren, Fractography of Glasses and Ceramics IV, 389 400, The American Ceramic Society,
Ohio (2001)
10. A. L. Zijlstra and A. J. Burggraaf, Strength and Fracture behavior of Chemically Strengthened Glass
in Connection with the Stress Profile, Part II, J. Non Cryst. Solids, 1, 163 185 (1969)
11. M. Bakioglu, F. Erdogan, and D. P. H. Hasselman, Fracture Mechanical Analysis of Self fatigue in
Surface Compression Strengthened Glasses, J. Mat. Sci., 11, 1826 1834 (1976)
12. I. W. Donald and M. J. C. Hill, Preparation and Mechanical Behavior of some Chemically Strength
ened Lithium Magnesium Alumino Silicate Glasses, J. Mat. Sci., 11, 1826 1834 (1976)
13. E. Bouyne and O. Gaume, Fragmentation of Thin Chemically Tempered Glass Plates, Proc. Int.
Congr. Glass Volume 2, Extended Abstracts, Scotland, (2001)
14. S. J. Glass, E. K. Beauchamp, C. S. Newton, R. G. Stone, W. N. Sullivan, R. T. Reese, S. D.
Nicolaysen, and R. J. Kipp, Controlled Fracture of Ion Exchanged Glass Rupture Disk, Sandia
National Laboratories Report, SAND2000 0828 (2000)
15. H. W. McKenzie and R. J. Hand, Basic Optical Stress Measurements, Society of Glass Technology,
Sheffield, U.K., 1999
16. R. Tandon, D. J. Green, and R. F. Cook, Surface Stress Effects on Indentation Fracture Sequences,
J. Am. Ceram. Soc., 73 [9] 2619 27 (1990).
17. A. Y. Sane and A. R. Cooper, Stress Buildup and Relaxation During Ion Exchange Strengthening of
Glass, J. Am. Ceram. Soc., 70 [2] 86 89 (1987)
ELASTO-PLASTIC BEHAVIOR
OF GLASSY CARBON AND
SILICA GLASS BY
NANO-INDENTATION WITH
SPHERICAL TIPPED
INDENTER
Norio Iwashita, Michael V. Swain

and John S. Field*
1. INTRODUCTION
Glassy carbons (GCs), obtained by pyrolysis and heat treatment of thermosetting

resins above 10008C, are hard and brittle materials that fail suddenly when loaded in
tension or compression.1–3 In commercially available GCs, the mean grain diameter,
evaluated from micrographs under a field-emission electron gun type scanning
electron microscopy, and the crystallite sizes, calculated from X-ray powder diffrac-
tion profiles, both slightly increased with the elevation of heat treatment tempera-
ture(HTT) from 10008C to 30008C.4
The Young’s modulus and the flexural strength values of the GC heat-treated at
30008C were smaller than those of the GCs heat-treated at 1000 and 20008C5.
However, the fracture toughness, KIc , of GCs slightly increased with the elevation
of HTT.5,6
*
Norio IWASHITA, National Institute of Advanced Industrial Science and Technology (AIST ), 16 1
Onogawa, Tsukuba, Ibaraki 305 8569 JAPAN. Michael V. SWAIN and John S. FIELD, Biomaterials
Science Research Unit, The University of Sydney, Suite G11, National Innovation Centre, Australian
Technology Park, Eveleigh NSW 1430 AUSTRALIA
93
94 Norio Iwashita et al.
GCs have excellent attributes including: isotropy, chemical stability, high hard-
ness, wear resistance, gas impermeability, and are non-contaminating.1–3,7,8 In
recent years, they have been used as essential materials for components of silicon
wafer processing equipment used in the manufacture of semiconductor devices and
other applications including fuel cell components, electrochemical electrodes etc.9
For such applications, the micro-metrological processing precision and micro-mech-
anics of GC surfaces are very important.
Sakai and co-workers have investigated the GCs produced by different densifi-
cation processes using a Vickers indenter.10 The indentation hysteresis behavior was
found on curves of indentation load versus indentation depth for all indentation
loads from 10 to 500 N. For GCs, it was also observed that ring/cone cracks were
induced by Vickers indentation instead of the median/radial crack systems more
usually observed in brittle materials under a sharp indenter. A superior feature of
GCs compared to ordinary brittle ceramics is their resistance to strength degradation
by contact with hard particles as demonstrated by anomalous ring/cone crack
formation.
Field and Swain have studied elasto-plastic response on a number of brittle
materials, including silicate glass, silicon single crystal and single crystalline sapphire,
by the indentation with small micron sized spherical tipped indenters.11,12 The
analysis of indentation stress-strain curve was performed from the indentation on
these materials.
Field and Swain found that the indentation force-displacement behavior for GC
exhibited almost complete recovery with a significant hysteresis between loading and
unloading when indented with 3 and 10 mm radius spherical tipped indenters.12–15
And also we have studied on elasto-plastic deformation of silica glass and glassy
carbons with different type of indenters16,17
In the present work, indentation of GCs heat-treated at different temperatures
was carried out using a spherical tipped indenter. For comparison, fused silica glass
was also used. Quantitative evaluation of ‘‘elasticity’’ and ‘‘plasticity’’ of elasto-
plastic deformation by the spherical tipped indentation on GC and silica was also
attempted. The dependence of hysteresis and elasto-plastic deformation of GCs on
their crystal structure are discussed from stress-strain curve.
2. EXPERIMENTAL
Glassy carbons (GCs) heat-treated at 1000, 2000 and 30008C (hereafter, sample
codes: GC1000, GC2000 and GC3000) were used in a series of work. For compari-
son, fused silica grass was used. The surface which was cut after carbonization of a
block of precursor thermosetting resin was used for the measurement of small radius
spherical tipped indentation. Previous work has shown that the surface of glass-like
carbons directly derived from the surface of thermosetting resin blocks exhibit
different crystal structure and micro-texture from their bulk.18 The surface of these
samples was polished using diamond slurry.
The following X-ray parameters were measured on the pulverized samples mixed
with high-purity silicon powder as an inner standard.19 Average interlayer spacing
(d002 ) and lattice constant along the a-axis (ao ) were determined from Bragg angles of
002 and 110 diffraction lines by referring to the diffraction positions of the inner
Elasto-Plastic Behavior of Glassy Carbon and Silica Glass 95
standard silicon, respectively. The crystallite sizes along the c-axis and the a-axis were
calculated from the half-maximum width of 002 and 110 diffraction peaks, denoted
as Lc (002) and La (110), respectively. The correction for the peak broadening due to
equipment was made by referring to silicon.19
Load cycle indentation was performed using a commercial ultra micro-
indentation system (UMIS-2000) with a spheroconical diamond indenter with 3 mm
radius. The radius of the indenter was calibrated using silica glass, of which the
elastic modulus and Poisson’s ratio were known. The initial contact force was
0.1 mN, and load controlled tests are performed. Each test was repeated ten times
at different positions on the polished surface of the specimen.
3. RESULTS & DISCUSSION
Bulk density and X-ray structural parameters of the used GCs are summarized in
Table 1. For comparison, the data of isotropic graphite block is shown in the same
table. It was characteristic that the bulk densities of the GCs decrease with the
elevation of heat treatment temperature (HTT). Commonly, GCs are non-graphitiz-
able carbon, so that the values of average interlayer spacing d002 of the present GC
samples are larger than 0.34 nm, even in the sample GC3000 heat-treated at 30008C.
The stacking of carbon hexagonal network layers is turbostratic. As compared with a
commercially isotropic graphite block, the crystallite sizes, La and Lc , are very small.
As the result that the crystallite structure of GCs is not almost grown by heat
treatment, it is difficult to evaluate the dependence of mechanical properties of GC
on heat treatment temperature by traditional tests, for instance, bending test. In
Figure 1, the force-penetration curves for the indentation on fused silica glass with
100 mN terminal force is shown. For the force-displacement response of a spherical
indenter on ‘‘elastic’’ surface, Hertz derived the basic relationship between penetra-
tion, ht , and applied force, P, namely20
2=3 1=3
3P 1
ht ¼ , (1)
4E R
where R is the radius of spherical indenter and E the composite modulus of the
indenter and specimen, which is defined by
Table 1. Bulk density and X ray structural parameters of the used GC samples and isotropic
graphite.
GC1000 GC2000 GC3000 isotropic graphite
Bulk density [g=cm3 ] 1.52 1.48 1.44 1.81
X ray structural parameter
d002 [nm] 0.3515 0.3482 0.3480 0.3362

ao [nm] 0.2412 0.2426 0.2436 0.2461
Lc (002) [nm] 1.7 2.6 3.5 55.4
La (110) [nm] 2.2 2.7 3.5 44.6
100 Fused silica glass
80
Force (mN)
60
40
20
0
0.0 0.2 0.4 0.6 0.8
Penetration (μm)
Figure 1. Load penetration curve of fused silica glass with 3 mm radius spherical tipped indenter.
1 1 n2s 1 n2i

¼ þ , (2)
E Es Ei
where n is Poisson’s ratio and the subscripts s and i mean specimen and indenter,
respectively. From Eq. (1), perfectly elastic curve can be drawn, showing as the
dotted line in the same figure. For the indentation with spherical indenter, thus, it
is characteristic that elastic deformation is observed at the initial of indentation, and
then elasto-plastic behavior is seen beyond a yield point.
Schematically illustrations of deformation on surface with spherical indenter are
shown in Figure 2. The geometrical factor for ‘‘elastic’’ surface deformation is equal
to 2 for spherical indenter.20 Thus, the contact circle penetration hp is described as
following:
ht
hp ¼ : (3)
2
Providing that ht << R, the radius of the contact circle is calculated using11:
p
a ¼ R hp : (4)
Thus, equation (1) can be rewritten as11:

P 4 a
2
¼ E : (5)
pa 3p R
The effective strain associated with a spherical tipped indenter may be written as the
ratio of radius of contact circle to the radius of indenter, a=R:21
As illustrated in Figure 2(b), ‘‘elasto-plastic’’ deformation occurs upon exceed-
ing a critical load for yielding. Thus, ht is larger than elastic component he calculated
using Eq. (1). The contact circle penetration hp is estimate from:11
he
h p ¼ ht : (6)
2
Figure 2. Schematically illustrations of deformation on surface with spherical indenter.
Figure 3. Plot of stress P=pa2 strain a/R of fused silica glass with 3 mm radius spherical tipped indenter.
The plot of P=pa2 versus a/R analyzed from the force-penetration curve of Figure
1 is shown in Figure 3. A value of elastic modulus E is estimated from a linear fitting
for the first several points within the linear region on the plots of P=pa2 versus a/R. A
non-linear relation is observed beyond the yielding point after the initial elastic
deformation on the plots of P=pa2 versus a/R. Since the secondary mechanism is
increasingly more curved at the higher terminal force, it maybe described by a simple
power law relationship, as following:11
P a n
2
¼ k: : (7)
pa R
The power law constants, n and k, were estimated from log-log plots of the non-linear
portion of the stress-strain curves.
Force-penetration curves of the indentation on GCs with 100 mN terminal force
are shown in Figure 4(a). In the case of inorganic glasses, the spherical tipped
indentation unloading is ‘‘elastic’’ deformation, as shown in Figure 1. A residual
penetration remains after fully unloading. For glass-like carbons, however, it is
characteristic that complete recovery of elasto-plastic deformation is observed on
unloading, that is, hysteresis loop. The measured penetration ht is smaller than he for
Figure 4. (a) Force penetration and (b) stress strain curves of the indentation on GCs with 100 mN
terminal force.
the recovery of elasto-plastic deformation. This hysteresis behavior was discussed in

the previous papaers.13,14
For the elevation of heat-treatment temperature of GCs, it is seen that the
penetration by the indentation is deeper and the hysteresis loop is larger. The area
of the hysteresis loop on the force-displacement curve corresponds to energy loss
during the deformation of the substrate by the indentation.
The loading and unloading elastic moduli, yield stress, yield strain and constants
associated with power law fitting of the indentation stress-strain curves on the
different GCs are summarized in Table 2. These values decrease with the elevation
of HTT of GCs.
In the present results, it was found that elasto-plastic deformation of the GCs by
indentation is very sensitive to their crystal structure and micro-texture. The elastic
modulus of the GCs reduced with the elevation of HTT (Table 2). The reduction of
the elastic modulus is predicted to result from not only the corresponding decrease in
bulk density of GC with the elevation of HTT, but also a decrease in shear modulus
(rigidity) of the carbon hexagonal network layers with development of the graphitic
structure.
The yield stress showed a similar trend to the elastic modulus of significantly
decreasing with development of graphitic structure of GCs. The constants of power
law fitting, n and k, for the stress-strain curves by the indentation on the different
GCs decreased with the elevation of HTT (Table 3). From the reduction of these
Table 2. Elastic modulus, yield stress, yield strain, constants of power law fitting and
indentation elasticity index obtained from stress strain curves with 3 mm radius spherical
tipped indenter on GCs and silica glass.
GC1000 GC2000 GC3000 Silica
E (loading) [GPa] 30.7 24.3 21.3 67.1

E (unloading) [GPa] 32.0 24.0 20.4 72.8
Yield Stress [GPa] 3.9 2.7 1.8 5.5
Yield Strain [ ] 0.30 0.26 0.21 0.19
n [ ] 0.87 0.71 0.63 0.72
k [GPa] 11.2 7.0 5.0 18.0
k=E [ ] 0.357 0.289 0.239 0.257
k/Y [ ] 2.87 2.60 2.75 3.27
parameters, the elasto-plastic deformation is more easily generated with developing

the graphitic structure and decreasing bulk density of GCs.
The value of k, which is the stress at strain a=R ¼ 1, corresponds to hardness
for spherical tipped indenter, although the extrapolation of the power law fitting to
such strains will invariably introduce errors into the present estimation of the
hardness.
Sakai has proposed that the ratio of hardness to elastic modulus E , that is, the
‘‘elasticity index’’ as a material parameter.22–24 The elasticity index k=E for each
the GC samples is also listed in Table 2. It is found that the GCs heat-treated at the
higher temperature showed the lower elasticity index k=E . Thus, elasticity of GC
decreases with the elevating HTT of GC.
It is assumed that the value of the ratio of hardness to yield stress, k/Y, is a
measure of plasticity. For GCs, the value of k/Y is nearly equal to 3, being independ-
ent of HTT. In comparison with fused silica glass, since GC shows slightly smaller
value of k/Y, the fused silica glass has a relatively high plasticity. On the other hand,
the elasticity index k=E of the fused silica glass is smaller than those of GC1000 and
GC2000 and larger than that of GC3000
4. CONCLUSION
As shown in the present work, elastic and elasto-plastic deformations derived by

spherical tipped indentation on glassy carbons depended strongly on their crystal
structure. The elastic modulus and the yield stress of glassy carbons decreased with
the elevation of HTT of glassy carbons.
The power law fitting constant k for the elasto-plastic deformation of the loading
stress-strain curves by the indentation and the indentation elasticity index k=E
decreased with development of the graphitic structure of glassy carbon.
In the present work, a variation of ‘‘elasticity’’ and ‘‘plasticity’’ of glassy carbons
with the development of the graphitic structure could be readily compared by the
parameter, indentation elasticity index k=E .
REFERENCES
1. Kobayashi, K., Sugawara, S., Toyoda, S. and Honda, H., Carbon, 6, 359 363 (1968).
2. Fitzer, E., Schaefer, W. and Yamada, S., Carbon, 7, 643 648 (1969).
3. Noda, T., Inagaki, M. and Yamada, Y., J. Non Cryst. Solids, 1, 285 302 (1969).
4. Yoshida, A., Kaburagi, Y. and Hishiyama, Y., Carbon, 29(8), 1107 1111 (1991).
5. Minnear, W.P., Hollenbeck, T. M., Bradt, R. C. and Walker, Jr. O. L., J. Non Cryst. Solids, 21,
107 115 (1976).
6. Zhao, J. X., Bradt, R. C. and Walker, Jr. P. L., Carbon, 24(1), 15 18 (1985).
7. Mehrota, B. N., Bragg, R. H. and Rao, A. S., J. Mater. Sci., 18, 2671 2671 (1983).
8. Bose, S. and Bragg, R. H., Carbon, 19(4): 289 295 (1981).
9. Suzuki, Y., Tanso, No. 190, 293 299 (1999). (Japanese)
10. Sakai, M., Hanyu, H. and Inagaki, M., J. Am. Ceram. Soc., 78(4): 1006 1012 (1995).
11. Field, J. S. and Swain, M. V., J. Mater. Res., 8, 297 306 (1993).
12. Swain, MV, Key Eng. Mater, 166, 41 46 (1999).
13. Field, J. S. and Swain, M. V., Carbon, 34(11), 1357 1366 (1996).
14. Iwashita, N., Swain, MV, Field, JS, Ohta, N and Bitoh, S, Carbon, 39(10), 1525 1532 (2001).
15. Iwashita, N, Field, JS and Swain, MV, Philosophical Magazine Part A, 82(10), 1873 1881 (2002).
16. Iwashita, N and Swain, MV, Molecular Crystals and Liquid Crystals, 386, 39 44 (2002).
17. Iwashita, N and Swain, MV, Philosophical Magazine Part A, 82(10), 2199 2205 (2002).
18. Park, S. M, Yasuda, E, Akatsu, T, Tanabe Y, Kakihana, M and Ozawa, K, Carbon, 33(10),
1377 1381 (1995).
19. Inagaki M, Tanso, No.36, 9 13 (1963). (Japanese)
20. Johnson, KL, ‘‘Indentation Contact Mechanics’’ Cambridge: Cambridge University Press, 1985.
21. Tabor, D, ‘‘Hardness of Metals’’ Oxford: Clarendon Press, 1951.
22. Sakai, M, Acta Metall. Mater., 41(6), 1751 1758 (1993).
23. Sakai, M, Shimizu, S and Ishikawa, T, J. Mater. Res., 14(2), 1471 1484 (1999).
24. Sakai, M, J. Mater. Res., 14(9), 3630 3639 (1999).
SCRATCH TEST FOR
EVALUATION OF SURFACE
DAMAGE IN GLASS
Satoshi Yoshida,* Takashi Hayashi,* Tetsuro Fukuhara,*

Kazuki Soeda,* Jun Matsuoka,* and Naohiro Soga**
1. INTRODUCTION
Scratch behavior in glass is very important to understand the interaction between the
abrasive particles and the glass product during grinding. In addition, the chipping
problem in glass often appears during the glass manufacturing process. A thin sheet
glass, such as an LCD(Liquid Crystal Display) or PDP(Plasma Display Panel) glass,
is susceptible to scratches, abrasions, rubbing, and chipping, which are mechanical
failures caused by a dynamical contact between the glass and a foreign body. In order
to suppress these contact problems and to develop a glass with high wearability, it is
essential to understand the mechanical response to a scratch as the cumulative
actions of a hard material. It is considered that a scratch test is one of the funda-
mental testing methods to evaluate these mechanical properties. However, only a few
papers have been published on scratch behaviors in glass.1–3 Especially, the origin of
scratch groove formation has little been understood. The formation of scratch
groove is probably related to microscopic fracture, plastic deformation, and material
removal. The scratch test surely gives us some important information on the dama-
geability of glasses with various compositions under various conditions.
It is well known that fracture properties in glass are affected by environment.
Stress corrosion cracking is believed to be the result from a stress-enhanced chemical
reaction between a chemical environment, usually water, and the Si-O-Si bonds.4
* Department of Materials Science, The University of Shiga Prefecture, Hikone, Shiga 522 8533, Japan
** National Institute of Advanced Industrial Science and Technology, Chiyoda Ku, Tokyo 100 8921,
Japan
101
102 Satoshi Yoshida et al.
In addition, Westwood and Huntington5 reported that both drilling rate and pendu-
lum hardness depended on the number of carbon atoms in n-alcohols and n-alkanes,
because these processes included fracture behaviors. Therefore, scratch behaviors of
glass also depend on these environments.
In this study, surface damage in glass is evaluated by using a scratch testing
equipment. The purpose of this study is twofold; (1) to evaluate the differences
between scratch hardness and conventional hardness, and (2) to evaluate the compos-
itional variation of the crack initiation load determined through a dynamic-loading
scratch test. It is considered that the crack initiation shows a fracture criterion at the
transition point from apparent plastic deformation to macroscopic fracture. The crack
initiation load will give some useful information on brittleness of glass.
2. EXPERIMENTAL PROCEDURES
A scratch hardness tester was assembled from two load cells, two moving stages, and a
Knoop indenter. A schematic of this measuring system is shown in Fig. 1. In this study,
a Knoop indenter was used as a scratching element. Normal load on the specimen was
detected with a load cell, which was contact with the indenter and driven by a stepping
motor capable of pulling and pushing at a constant speed. Normal load was controlled
by changing the position of the indenter. Horizontal load was measured by another
load cell connected with another moving stage. The stage was moved in the direction of
the long diagonal of Knoop indenter. In this study, the constant speed of the stage,
0.05 mm/s or 0.07 mm/s, was adopted for the scratch test. It has been reported
that scratch hardness depends on the scratching speed, especially at the speed of
>0.1 mm/s,1 so the scratch speed was fixed at the above value. Scratch hardness is
defined as the load per unit load-bearing area (Fig. 2)6 and calculated from the normal
load and the groove width using the following equation,
HS ¼ FN =A ¼ 0:5622 (FN =w2 ), (1)
where FN is the normal load, A the vertical projection of contact area, and w the
groove width. The constant in the right side can be determined from the geometry
of indenter. In the case of Knoop indenter, the angle between the opposite edges
is 172.58 and the other angle between the other two is 1308. By using the groove
width, w, the load-bearing area, A, becomes 1.779 w2 .
On the other hand, Jacobsson et al.6 also defined the plowing stress, HP , which is
the mean pressure opposing the sliding of the element due to the plowing process only.
HP ¼ FP =AP ¼ 1:124 (FP =w2 ), (2)
where FP is the horizontal load and AP the area of resultant groove cross-section. In
this study, however, because of a large scatter of FP data and its less dependence of
composition, only HS will be employed for discussions on the compositional vari-
ation of scratch hardness.
For the determination of scratch hardness of soda-lime glass(Matsunami 0050),
a normal load and a scratch speed were fixed at 0.98 N and 0.05 mm/s, respectively.
Water and some n-alcohols(1-pentanol, 1-hexanol, 1-heptanol, 1-octanol, 1-nonanol,
1-decanol, and 1-undecanol) were employed as environment. These liquids except for
Load cell FN: normal load

(normal load) Moving
stage Scratch direction
Indenter
Sample
A: vertical projection
Moving stage
w
Load cell of the contact area
Knoop indenter (horizontal load)
Figure 1. Schematic of a scratch tester. Figure 2. Schematic of a scratch groove.
water were dehydrated with a molecular sieve for at least one week. All the liquids were
dropped on the sample surface just before the scratch test. Scratch hardness in each
environment was determined from at least 10 runs. For a comparison, Knoop hardness
was also measured at the indentation load of 0.49, 0.98, 2.0, and 2.9N in the same liquids.
In the case of the measurements of crack initiation load for silica(Asahi AQ)
glass, Pyrex-type glass(Tempax) and soda-lime glass, a normal load was changing at a
rate of 0.03N/s. A scratch speed was fixed at 0.07mm/s. The crack initiation was
evaluated in water and n-heptane, which was also dehydrated. The width of scratch
groove and the existence of crack at the edge of groove were observed by using an
optical microscope after the scratch test. In some cases, the crack geometry was
observed by using a scanning electron microscope(SEM; Hitachi S-3200N). The
crack initiation load was determined from the position of an initial crack at the
edge of groove, the scratch speed, and the rate at which the normal load increased. At
least 30 specimens were used for the determination of crack initiation load.
3. RESULTS
3.1. Scratch Hardness
Figure 3 shows scratch hardness of soda-lime glass in water and in n-alcohols.

Scratch hardness increases with an increase in the number of carbon atoms in each
alcohol molecule, but shows a maximum at 1-decanol. Scratch hardness obtained in
1-decanol is 28% larger than that in 1-pentanol. Scratch hardness obtained in water is
comparable with that in 1-pentanol.
As a comparison, Knoop hardness in each environment was measured. The
result is shown in Fig. 4. A load dependence of Knoop hardness can be observed for
all the environments. This load dependence of hardness was also reported by Hirao
and Tomozawa,7 and will be discussed later. Knoop hardness at a given indentation
load increases by 8% with an increase in the length of alcohol chain. In addition, it is
found that Knoop hardness(4.35.0GPa) is over two times smaller than scratch
hardness (1015GPa).
3.2. Crack Initiation Load

The crack initiation test was performed for silica, Tempax, and soda-lime glasses in
water and in dehydrated heptane by using the scratch tester. Figure 5 shows an
Scratch hardness /GPa

15
10
water 5 6 7 8 9 10 11
Number of carbon atoms in a normal alcohol

Figure 3. Scratch hardness of soda lime glass in water and n alcohols. Error bars indicate 1 standard
deviations.
0.49N
0.98N
Knoop hardness /GPa
2.0N
5 2.9N
water 5 6 7 8 9 10 11
Figure 4. Knoop hardness of soda lime glass in water and n alcohols. The indentation loads are 0.49,
0.98, 2.0, and 2.9 N.
example of a relationship between a scratch position and a normal load for soda-lime
glass in hep-tane. Four points labeled A, B, C, and D correspond to the scratch areas
noted by the same letters in Fig. 6. The crack initiation can be observed in Fig. 6(B). As
for silica glass, a position-load curve obtained in heptane and photos of a scratch
groove are shown in Fig. 7 and Fig. 8, respectively. The crack initiation load for silica
glass is much smaller than that for soda-lime glass. The lower crack initiation load for
silica glass is also observed in water. The crack initiation loads for each glass can be
fitted well by a Weibull function. Weibull plots for the data obtained in heptane are
shown in Fig. 9. The averaged values of crack initiation loads and Weibull moduli for
all the glasses in heptane and in water are listed in Table 1. It is considered that Weibull
modulus describes the width of distribution in crack initiation load, and that it is much
affected by surface condition. In order to discuss the compositional variation of
3
C
Normal load /N
B
2 D
A
1 Crack initiation
0
0 1000 2000 3000 4000 5000
Position / μm
Figure 5. Relationship between normal load and position for soda lime glass in heptane.
30μm
Figure 6. Photos of a scratch groove after the crack initiation test for soda lime glass in heptane. Scratch
direction is from bottom to top in these photos.
C
Normal load /N
B
1
A
Crack initiation
0
0 1000 2000
Position / μm
Figure 7. Relationship between normal load and position for silica glass in heptane.
Weibull modulus, some information on the relationship between the detailed finishing
grade of glass surface and Weibull modulus will be needed. It is noteworthy that
Tempax has the highest crack initiation load in water, but not in heptane.
Figure 8. Photos of a scratch groove after the crack initiation test for silica glass in heptane. Scratch
4. DISCUSSION

The load-dependence of Knoop hardness in Fig. 4 is probably caused by a change in
surface energy of the glass in each environment. It has been reported that this load
dependence of hardness is attributed to the contribution of surface energy to the
indentation deformation.7 Following the works by Bernhardt8 and Frohlich et al.,9
Hirao and Tomozawa7 separated the indentation deformation into two energy
contributions, which were the surface energy contribution and the volume energy
contribution. They suggested that Knoop indentation load, L, can be expressed by
the following equation,
F = 1 exp [ (P/ξ)m]
Crack initiation probability, F /%
99
90
80
60
40 silica Tempax soda lime
m=6.5 m=6.3 m=11
20
10
0.5 1.0 2.0 3.0

Crack initiation load, P/N
Figure 9. Weibull plots of the crack initiation loads obtained in heptane.
Table 1. Averaged crack initiation loads and Weibull moduli for the glasses.
Glass Crack initiation load /N Weibull modulus, ma
Silica in heptane 0.63 6.5

in water 0.35 3.8
Tempax in heptane 1.39 6.3
in water 0.99 7.0
Soda lime in heptane 1.63 11
in water 0.56 5.6
a
F ¼ 1- exp [-(P=j)m ]
L ¼ a1 l þ a2 l 2 , (3)
Ð
where a1 and a2 are fitting parameters, and l is Knoop diagonal length. Because Ldl
may be related to the total energy of indentation deformation, a1 (l 2 =2) is related to
the energy to create new surface, and a2 (l 3 =3) to the energy for volume deformation.
Hence, a2 corresponds to the load-independent hardness, and a1 the parameter
related to the sur-face energy. On the other hand, from the values of L and l,
Knoop hardness, HK, is defined as the following equation,10
HK ¼ 14:23L=l 2 : (4)
From Eqs. (3) and (4), if the surface energy term can be negligible, Knoop hardness is
independent of the indentation load. Figure 10 shows some examples of the relation-
ship between L / l and l. From Eq. (3), the following expression can be obtained, and
a1 corresponds to the y-intercept, and a2 to the slope of the line in Fig. 10.
L=l ¼ a1 þ a2 l (5)
Figure 11 shows the fitting parameters, a1 and a2 , in all the environments. A more
distinct increase in a1 can be observed in this figure as compared with that in a2 .
40
1-decanol
water
30
20
L/l /103 Nm−1
10
0
0 20 40 60 80 100
l/ 10−6m
Figure 10. Relationship between L / l and l for soda lime glass in water and in 1 decanol.
4 400
3 300
a1 /kNm−1
a2 /MPa
2 200
1 100
0 0
water 5 6 7 8 9 10 11
Number of carbon atoms in a molecule
Figure 11. Fitting parameters, a1 and a2 in Eq. (5) obtained in water and in n alcohols.
It implies that the variation in chain-length of n-alcohol affects the surface state
between glass and liquid rather than hardness as a bulk property. The surface energy
is probably related to the polarity in n-alcohols because the surface charge of soda-
lime glass is in general negative. The larger the polarity in n-alcohol is, the less the
surface energy of glass in the liquid. Reported dielectric constants of n-alcohols
employed are listed in Table 2.
On the other hand, fracture toughness, KIc , is defined as the following equation,
KIc ¼ (2GE)1=2 , (6)
for the plane-stress condition. In this equation, G is the fracture surface energy, and
E is Young’s modulus. Although the surface energy is not coincident with the fracture
surface energy, the former covers a large part of the latter for ideal brittle materials.
Therefore, the surface energy must affect the fracture-related properties, such as
scratch hardness. It can be said that the scratch hardness is closely linked to the
surface-creating process, such as micro-cracking, chipping and so on. These energy
release processes surely reduce the energy for groove formation. This is the reason
Table 2. Dielectric constants of water and n alcohols.
Solvent Dielectric constant
Water 78.54(258C)a
1 pentanol 15.13(258C)b
1 hexanol 13.03(208C)b
1 heptanol 11.75(208C)b
1 octanol 10.30(208C)b
1 nonanol
1 decanol 8.1(208C)a
1 undecanol
a
Ref. [11]
b
Ref. [12]
why scratch hardness is much larger than Knoop hardness. The larger variation of
scratch hardness originates from a larger contribution of the surface-creating pro-
cess.
In Fig. 4, a longest-chain alcohol(1-undecanol) causes a decrease in scratch
hardness. Although the origin has been unclear, Westwood and Huntington5 pro-
posed one reasonable explanation that the hardness of glass can be markedly influ-
enced by the z-potential, which is likely to be similar to the surface charge of glass.
When z is negative, which is true in water and in the lower n-alcohols,13 this situation
must bring positive charge carriers, e.g. Naþ ions, into the near-surface region for
charge-counter balancing. This locally increased Naþ ions reduce hardness in this
region. When z is positive, on the other hand, the migration of OH ions is respon-
sible for softening of glass. Glass should be the hardest at its zero point of surface
charge. There is no evidence that the sign of z-potential is reversed by changing the
chain-length of n-alcohol. However, this surface compositional change would be one
possible origin of decrease in scratch hardness.

The cracks observed around the groove were mainly surface chips or radial(chevron)
cracks. There was no median crack beneath the groove. Figure 12 shows the cross
section of the scratch groove in soda-lime glass. For the other glasses, no median
crack can be also observed in the cross sections.
Water environment promotes crack initiation in silica, Tempax, and soda-lime
glasses (Table 1). This is because the crack initiation process is assisted by the
adsorption of water molecule. In the conceptual framework of this idea,14 a water
molecule is adsorbed at the strained bonds formed by a scratch process. After that,
the attacked Si-O-Si bond and the water molecule break one Si-O bond to leave new
two Si-OH bonds. West and Hench15 reported that the energy barrier of hydrolysis of
strained 3-fold rings is 97% smaller than that of fracture by water-free dilation. This
water-assisted bond-breakage process is the origin of lower crack initiation load
obtained in water.
The crack initiation load of silica glass is much lower than those of Tempax and
soda-lime glasses (Table 1). This lower crack initiation load cannot be explained in
terms of conventional mechanical properties, e.g. Young’s modulus and fracture
toughness. Table 3 shows some mechanical properties of silica, Tempax, and soda-
lime glasses. Here, Young’s modulus and Vickers hardness are the catalogue values,
5μm
Figure 12. SEM micrograph of a scratch groove of soda lime glass in air. Normal load 2.94N.
Table 3. Some mechanical properties of the glasses.
Young’s Vickers Fracture

Crack initiation Brittleness modulusa / hardnessa / toughnessb /
Glass load (in heptane)/N / mm1 GPa GPa MPam1=2
Silica 0.63 10.0 74.2 7.9 0.79

Tempax 1.39 8.5 63.0 5.5 0.65c
Soda lime 1.63 7.1 71.5 5.4 0.76
a
Catalogue value.
b
Ref. [16]
c
Ref. [17]
and fracture toughness the reported values. The lower crack initiation load is related
not to the weaker bond-strength(Young’s modulus), but to the resistance to defor-
mation(hardness). One part of the work done by an external force is used to deform
plastically. The rest part may be converted into a residual stress around the groove or
into a crack. This means that the decreased ability to deform leads to easy crack
initiation. Even if the averaged bond-strength of silica glass is comparable with that
of soda-lime glass, less deformability in silica glass causes the crack initiation load to
be lowered.
It is considered that the crack initiation load is a measure of brittleness (HV /
KIc ) of glass. In Table 3, brittleness of each glass is calculated from HV and KIc . The
higher brittleness of glass causes the lower crack initiation load. This relationship was
also reported by Le Houérou et al.3 It has been still unclear that Tempax shows the
maximum crack initiation load in water, but it is suggested that B2 O3 as a network
former plays an important role for deformation. Further experimental investigations
will be needed.
Sehgal and Ito18 have shown that the brittleness of glass can be determined from
the Vickers indentation load, the Vickers diagonal length, and the median/radial
crack length. However, their technique cannot be applied to the complicated crack
patterns by indentation, such as a cone crack in silica glass. It is concluded that the
crack initiation test by using a scratch tester is the simple and useful technique for
evaluating brittleness of various glasses.
5. CONCLUSIONS
The effect of environment on scratch hardness has been investigated for soda-lime
glass. Scratch hardness was much influenced by the chain-length of n-alcohol. The
variation of scratch hardness with the chain-length of n-alcohol was more conspicu-
ous than that of Knoop hardness. This result comes from the variation of surface
energy of glass in the liquids. From the measurement of load-dependent Knoop
hardness in various n-alcohols, it was found that the chain-length in n-alcohol
affected the parameter related to the surface energy rather than to the intrinsic
hardness. Because the surface energy is directly related to the fracture behavior,
scratch hardness including fracture process surely depends on environment.
The crack initiation at the groove-edge can be clearly observed for silica,
Tempax, and soda-lime glasses by using a scratch testing machine. The crack
initiation load obtained in water was much lower than that in heptane because of the
water-assisted bond-rupture. The crack initiation load for silica glass was lower than
those for Tempax and soda-lime glasses. This is attributed to less deformability in
silica glass. It is elucidated that the crack initiation load through the dynamic scratch
test is a measure of brittleness of glass.
REFERENCES
1. K. Li, Y. Shapiro, and J.C.M. Li, Scratch test of soda lime glass, Acta Mater. 46(15),
5569 5578(1998).
2. S. Yoshida, H. Tanaka, T. Hayashi, J. Matsuoka, and N. Soga, Scratch resistance of sodium
borosilicate glass, J. Ceram. Soc. Japan 109(6), 511 515(2001).
3. V. Le Houérou, J. C. Sangleboeuf, S. Dériano, T. Rouxel, and G. Duisit, Surface damage of soda
lime glasses: indentation scratch behavior, J. Non Cryst. Solids 316, 54 63(2003).
4. S.M. Wiederhorn, S.W. Freiman, E.R. Fuller Jr., and C.J. Simmons, Effect of water and other
dielectrics on crack growth, J. Mat. Sci. 17, 3460 3478(1982).
5. A.R.C. Westwood and R.D. Huntington, Adsorption sensitive flow and fracture behavior of soda
lime glass, Proc. Intl. Conf. Mechanical Behavior of Materials IV, Society of Materials Science, Kyoto,
Japan, 383 393(1972).
6. S. Jacobsson, M. Olsson, P. Hedenqvist, and O. Vingsbo, Scratch testing, ASM Handbook 18,
430 437(1992).
7. K. Hirao and M. Tomozawa, Microhardness of SiO2 glass in various environments, J. Am. Ceram.
Soc. 70(7), 497 502(1987).
8. E.O. Bernhardt, On microhardness of solids at the limit of Kick’s similarity law(in German),
Z. Metallkd. 33(3), 135 144(1941).
9. F. Frohlich, P. Grau, and W. Grellmann, Performance and analysis of recording microhardness tests,
Phys. Status Soldi 42, 79 89(1977).
10. Japanese Industrial Standards, Z 2251(1998).
11. Kagaku Binran, Kiso Hen II (in Japanese), p.1166, Ed. by the Chemical Society of Japan (Maruzen,
Tokyo, 1975).
12. D.R. Lide, Handbook of organic Solvents, p.254, 265, 370, 380 (CRC Press, Boca Raton, Florida,
1995).
13. M. DeSmet and J. Delfosse, Colloid Science, p.230, Ed. by H.R. Kruyt (Elsevier, New York, 1952).
14. T.A. Michalske and S.W. Freiman, A molecular interpretation of stress corrosion in silica, Nature
295, 511 512(1982).
15. J.K. West and L.L. Hench, Silica fracture Part II A ring opening model via hydrolysis, J. Mat. Sci.
29, 5808 5816(1994).
16. S.M. Wiederhorn, Fracture surface energy of glass, J. Am. Ceram. Soc. 52(2), 99 105(1969).
17. J. Sehgal and S. Ito, Brittleness of glass, J. Non Cryst. Solids 253, 126 132(1999).
18. J. Sehgal, Y. Nakao, H. Takahashi, and S. Ito, Brittleness of glasses by indentation, J. Mat. Sci. Lett.
14, 167 169(1995).
SHEAR DRIVEN DAMAGE
AND INTERNAL FRICTION IN
INDENTATION LOADING OF
A GLASS-CERAMIC
Anthony C. Fischer-Cripps*
1. INTRODUCTION
Indentation damage is a particularly important limitation in the use of brittle

materials in structural applications. Some nominally brittle materials have been
shown to exhibit plastic type yield in indentation testing.1-10 These materials are
characterized by a relatively coarse grain structure and weak grain boundaries.
A particularly convenient material for study is a mica-containing glass-ceramic. As
shown in Fig. 1, the microstructure consists of circular mica platelets in a glassy
matrix. The size of the platelets can be modified by simple heat treatment during
processing. The average platelet diameter and the aspect ratio increase with increas-
ing crystallization temperature. During indentation with a spherical indenter, a sub-
surface region of accumulated damage is observed to form in preference to the
characteristic Hertzian cone crack normally associated with this type of loading.
An indentation stress-strain response shows a shape similar to that obtained on
ductile materials.
Previous work indicates that11,12 the indentation response of mica glass-ceramic
materials is a function of the macroscopic uniaxial compressive yield stress, Y and a
strain-hardening characteristic. A range of behavior, from quasi-ductile to nominally
brittle, could be attained by simply varying the size and shape of the grains through
*
Anthony C. Fischer Cripps, CSIRO Division of Telecommunications and Industrial Physics, Bradfield
Rd, West Lindfield, NSW 2070 Australia.
113
114 Anthony C. Fischer-Cripps
Figure 1. Electron micrograph of microstructure of mica glass ceramic. Specimen was polished to 1 mm
finish and etched in HF for 10 s. Most of the platelets are viewed edge on. The occasional plan view of
a platelet can be seen.
heat treatment. Details of the microstructural properties for the glass-ceramic ma-
terials studied here are given in Table 1.
The dependence of the yield stress was subsequently found to be consistent with
a Mode II fracture mechanism within the microstructure that depended upon the
absolute value of the grain size (i.e. platelet diameter). The apparent strain hardening
arises from an elastic constraint on the deformation of grains imposed by the glassy
matrix and depends on the platelet aspect ratio.13 Further study showed that the
magnitude of the yield stress was consistent with a mechanism of friction on a
microstructural scale quite different to that of the familiar Amonton’s law.14 The
present work brings the previously reported work together into a cohesive account of
the nature of internal friction on the microstructural scale and its relevance to the
macroscopic behaviour of these types of materials. An understanding of these issues
assists in the design, at a microstructural level, of these nominally brittle materials to
suit the requirements of a particular application.
Table 1. Microstructural parameters for mica glass ceramic materials. For all materials, the
volume fraction is Vf 70%, the elastic modulus E 70 GPa and Poisson’s ratio n 0:26.
Platelet dimensions measured from sectioned and etched samples. Yield stress was determined
according to a procedure given in reference 12.
Platelet
4 hr heat treatment Maximum platelet thickness Yield stress Y
Micro structure temp. (8C) diameter (mm) (mm) (MPa)
Fine 1000 1.2 0.40 2160

Medium 1060 3.7 0.80 1230
Coarse 1120 10 1.2 750
Shear Driven Damage 115
2. MACROSCOPIC BEHAVIOUR
For the glass-ceramic materials considered here, the plastic zone resulting from
indentation with a spherical indenter is contained within an area bounded by the
circle of contact and can be best described in terms of an elastic constraint.11 Fig. 2
shows the results from a bonded-interface specimen for a glass-ceramic specimen
loaded with a spherical indenter. The top part of Fig. 2 shows the residual impression
in the surface and the bottom part is a section view showing the sub-surface accu-
mulated damage. A quantitative measure of plasticity is given by an indentation
stress-strain response as shown in Fig. 3 where it can be seen that the amount of
deviation of the expected linearly elastic behaviour depends upon the microstructure
of the material, the greater the grain size, the more pronounced the deviation from
linear elasticity. The size and shape of the damage zone and the shape of the
indentation stress-strain response is similar to that observed in ductile materials
under this type of loading.
Elastic-plastic behavior in these nominally brittle ceramics occurs as a result of
the relatively large hydrostatic component of stress in the indentation stress field and
is routinely observed by researchers in the rock-mechanics field. Sub-surface damage
occurs as a result of shear faulting along planes of weakness within the material
beneath the indenter. In the present case, cleavage planes within the mica appear to
be sources of weakness that lead to shear-driven damage. Finite element work12 has
demonstrated that there is a close connection between the macroscopic yield stress of
these materials and the absolute platelet size. A mismatch in elastic properties within
the microstructure leads to stress singularities resulting in yield stress which has an
inverse square root dependence on the grain size in the manner of a Mode II crack.
Evidently, the macroscopic response of the material depends on shear-driven failure
events on a microstructural scale, which in turn, are intimately connected with the
nature of friction at this scale of contact.
Figure 2. Optical micrograph of indented bonded interface mica containing glass ceramic specimen
showing half surface (top) and section (bottom) views. Indentations performed with WC sphere of radius
R 3:18 mm at indenter load P 2000 N.
Figure 3. Indentation stress strain relationships for glass ceramic. Dashed line shows the expected
Hertzian elastic response. Experimental results are shown for the coarse (10 mm grain size), medium
(5 mm) and fine (2 mm) microstructures. The larger the grains, the more deviation from the elastic response.
3. INTERNAL FRICTION AND YIELD
Amonton’s law has been shown to be a result of the random roughness of real
surfaces and arises as a result of the increasing area of contact between asperities
on both surfaces as they are bought into contact as the normal force is increased. It us
usually expressed in terms of the friction coefficient m such that:
F ¼ mN (1)
where N is the applied normal load and F is the sideways force required to initiate
tangential sliding between the two bodies. Contact thus only takes place at asperities
on each surface and friction forces arise from elastic deformation, shearing, welding,
ploughing, or plastic deformation of the asperities as they move past one another. If
we divide the terms in Eq. 1 by the real area of contact A, we find that the frictional
shear stress t is in direct proportion to the normal stress sN :
F mN
t¼ ¼ ¼ msN (2)
A A
Jaeger and Cook15 show that slip along a plane of weakness within a material
under compressive loading with confined pressure may occur when the Coulomb
criterion is satisfied:
jtj ¼ So þ msN (3)

Figure 4. Schematic of a sliding shear fault under hydrostatic stress.
In Eq. 3, t is the applied shear stress, So is the inherent shear strength of the
material. and sN is the normal stress on the plane under consideration as shown in
Fig. 4. Eq. 3 is equivalent to Eq. 2 at So ¼ 0.
In terms of the principal stresses s1 and s3 , where s3 is the confining pressure
and s1 is the applied normal stress, the Coulomb criterion becomes:
h 1=2 i h 1=2 i
s1 m 2 þ 1 m s3 m2 þ 1 þm ¼ 2So (4)
The macroscopic Tresca failure criterion (t ¼ Y=2) is thus a special case of the
Coulomb criterion with m ¼ 0 and So ¼ Y=2 where the angle at which sliding occurs
is 458.
Finite element analysis shows that12 for the glass-ceramic materials considered
here, differences in the microstructure can be accommodated by a change in the
macroscopic yield stress Y for each material in conjunction with the Tresca criterions
implying that the internal friction coefficient m plays a negligible role in these
materials. Further support for this contention is that the strong, but short range
Van der Waals forces at the grain boundary maintain the boundary intact under
shear loading and that failure occurs within the atomically smooth cleavage planes of
the mica platelets.16,17 As shown in Fig. 5, atomic force microscope examination of
the damaged regions clearly shows fissures within the platelets.
It is thus claimed that any resistance to internal sliding depends upon the
inherent shear strength of the mica platelets. This in turn depends upon the grain
size, which then impacts on the macroscopic yield strength of the material.
4. DISCUSSION
The task now is to substantiate the conclusions made above given the following
conclusions of previously reported work:
Figure 5. Atomic force microscope image of a cross section of a mica platelet in the shear driven damage
zone. Arrows show fissures in the platelet that are not visible in undamaged material.
1. Experiments on a macroscopic scale demonstrate the presence of internal yield

within the microstructure, the degree of which depends upon the physical char-
acteristics of the microstructure.
2. Finite element modelling on a macroscopic scale using the Tresca criterion
matches the observed size and shape of the plastic zone and also the shape of
the indentation stress-strain curves indicating that the physical characteristics of
the microstructure determine its yield strength. The use of the Tresca criterion
implies a zero coefficient of friction on the microstructural scale. The range of
macroscopic yield stresses for the materials studied here is 750 MPa to 2000 MPa.
3. Finite element modelling on a microstructural scale indicates that internal yield
has the appearance of a Mode II crack phenomenon (the shear strength of the
microstructure depends on the inverse of the square root of the platelet diameter
and is independent of the magnitude of the normal stress on the grain). There is a
correlation between the macroscopic yield stress Y used in (2) above with the
inverse square root of the grain size.
4. Atomic force microscope observations show that shear faulting occurs within the
atomically smooth planes of the mica platelets and that friction is likely to occur
under conditions in which the real and apparent area of contact coincide.
Can the observed behaviour on the macroscopic and microstructural scales be
reconciled with what we know about frictional sliding under these conditions? To
answer this question, we turn to other reported work with mica in which the real and
apparent areas of contact coincide. Johnson18 reports that experiments with an
atomic friction microscope (AFM), on mica in which the contact dimension is 2 to
10 nm, indicate a frictional shear stress of 1 GPa. Other measurements performed
with a surface force apparatus (SFA), in which the contact dimension is in the order
of 20-80 mm, indicate a frictional shear stress to be approximately constant at

20 MPa. The discrepancy is explained by the mechanism of plastic deformation at
the interface. According to Johnson, the physical mechanism of sliding friction is due
to the nucleation and propagation of dislocation-like defects through the interface
under the influence of the applied shear force. For small contacts (< 20 nm), disloca-
tion-like defects cannot be nucleated and atoms move relative to one another with a
shear strength equal to the theoretical strength of the material. For large contacts, the
rate of nucleation of dislocation-like continuities at the edge of the contact becomes
equal to the rate at which they disappear at the centre and the frictional stress reaches
a steady-state value. In the intermediate range, the shear stress required to nucleate
dislocation-like defects at the leading edge of the contact depends upon the square
root of the dimension of the contact.
The relevance to the present work is that the scale of ‘‘contacts’’ in the mica
platelets studied here is in the order of 1 to 10 mm. This is the intermediate zone
referred to by Johnson in which the frictional stress depends upon the inverse square
root of the grain size in the manner of a Mode II crack. It has been previously shown
that estimations of yield stress for a range of platelet sizes based on an inverse square
root law yielded predictions of macroscopic behavior that agreed well with experi-
mentally observed behavior and that the macroscopic yield stress was determined to
be in the order of 1 GPa. A yield stress in the order of 1 GPa gives a value of So in the
order of 500 MPa, precisely within the intermediate zone described above. The
nature of the observed Mode II behaviour on the microstructural scale is thus
explained.
What is the nature of these defects that show such a scale effect? Johnson
describes them as ‘‘dislocation-like’’ and shows that work involving conventional
dislocation mechanics is consistent with observations of the size effects for frictional
shear stress given above. The answer lies in molecular dynamics simulations of which
are limited by the number of atoms that can be accommodated and as such, usually
fall into the small contact regime. Simulations on a larger scale would evidently
elucidate the actual mechanism of deformation on the molecular scale. However,
from a material science perspective, the scale effects of frictional traction under
controlled conditions matches well with those existing at a macroscopic level and
thus offers a useful base from which macroscopic material properties may be
designed from microstructural variables in some detail. For example, for the mater-
ials studied here, if the platelets were made very large (> 50 mm), one would expect
the size dependence on yield to diminish and that, all other things being equal, the
shape of the indentation stress-strain curves would perhaps begin to be independent
on the grain size.
6. SUMMARY
The present work examines the role of internal friction in mica-containing glass-
ceramic materials subjected to indentation loading. It is proposed that shear-driven
damage beneath the specimen surface occurs via internal sliding along cleavage
planes within the mica platelets near the mica-glass boundaries. The sliding surfaces
in this case are considered to be atomically smooth such that the real and apparent
areas of contact coincide. The frictional shear stress is thus independent of the
normal forces arising from thermal mismatch stresses and only depends on the work
of adhesion of the interface and the scale of the contacts. The scale of contacts for
these materials lies within an intermediate zone in which the frictional shear stress
arises from the stress required to nucleate dislocation-like discontinuities within the
material. This leads to a size effect similar to that experienced by a crack in Mode II
loading and is in accordance with previous work in which a connection between such
a size effect and the macroscopic response of the material was identified.
REFERENCES
1. B. R. Lawn, N. P. Padture, H. Cai, and F. Guiberteau, Making ceramics ‘ductile’, Science, 263,
1114 16 (1994).
2. F. Guiberteau, N. P. Padture, H. Cai, and B. R. Lawn, Indentation fatigue. A simple cyclic Hertzian
test for measuring damage accumulation in polycrystalline ceramics, Phil. Mag. 68(5), 1003 16
(1993).
3. N. P. Padture, B. R. Lawn, Fatigue in ceramics with interconnecting weak interfaces: a study using
cyclic Hertzian contacts, Acta Metall. 43, 1609 17 (1995).
4. N. P. Padture, In situ toughened silicon carbide, J. Am. Ceram. Soc. 77(2), 519 23 (1994).
5. N. P. Padture and B. R. Lawn, Toughness properties of a silicon carbide with an in situ induced
heterogeneous grain structure, J. Am. Ceram. Soc. 77(10), 2518 22 (1994).
6. H. H. K. Xu, L. Wei, N. P. Padture, B. R. Lawn and R. L. Yeckley, Effect of microstructural
coarsening on Hertzian contact damage in silicon nitride, J. Mat. Sci. 30, 869 78 (1995).
7. F. Guiberteau, N. P. Padture, and B. R. Lawn, Effect of grain size on Hertzian contact damage in
alumina, J. Am. Ceram. Soc. 77(7), 1825 31 (1994).
8. H. Cai, M. A. Stevens Kalceff, B. R. Lawn, Deformation and fracture of mica containing glass
ceramics in Hertzian contacts, J. Mater. Res. 9(3), 762 70 (1994).
9. H. Cai, M. A. Stevens Kalceff, B. M. Hooks, B. R. Lawn, and K. Chyung, Cyclic fatigue of a mica
containing glass ceramic at Hertzian contacts, J. Mater. Res. 9(10), 2654 61 (1994).
10. D. G. Grossman, Machinable glass ceramics based on tetrasilicic mica, J. Am. Ceram. Soc. 55,
446 49 (1972).
11. A. C. Fischer Cripps, Elastic plastic behaviour in materials loaded with a spherical indenter, J. Mat.
Sci. 32, 727 36 (1997).
12. A. C. Fischer Cripps, A partition problem approach to microstructural modelling of a glass ceramic,
J. Am. Ceram. Soc. 82(7), 1835 43 (1999).
13. A. C. Fischer Cripps and B. R. Lawn, Indentation stress strain curves for ‘‘quasi ductile’’ ceramics,
Acta Mater. 44(2), 519 527 (1996).
14. A. C. Fischer Cripps, Role of internal friction in indentation damage in a mica containing glass
ceramic, J. Am. Ceram. Soc. 84(11), 2603 2606 (2001).
15. J. C. Jaeger and N. G. W. Cook, Fundamentals of Rock Mechanics, (Chapman and Hall, 1971).
16. Private communication, D. T. Smith, National Institute of Standards & Technology, 1995.
17. R. G. Horn and D. T. Smith, Contact Electrification and Adhesion Between Dissimilar Materials,
Science, 256, 362 64 (1992).
18. K. L. Johnson, The contribution of micro/nano tribology to the interpretation of dry friction, Proc.
Instn. Mech. Engrs, 214 Part C, 11 22 (2000).
EXPERIMENTAL APPROACH
TO MICROFRACTURE
PROCESS OF CERAMICS
UNDER THERMAL SHOCK
Shuichi Wakayama, Kousuke Nishino, Kouhei Kubota,

Toshiya Wada and Satoshi Kobayashi
1. INTRODUCTION
The thermal shock fracture behavior of ceramics has been investigated using trad-
itional testing methods such as the water quench in which the critical temperature
difference, where samples are subjected to severe damage, is used as the thermal
shock resistance.1,2 Recently, new experimental methods are proposed and applied to
the investigation of macroscopic crack propagation process during thermal shock
fracture, based on fracture mechanics.3–5
However the characterization of microfracture process under thermal shock as
well as transient thermal stress field is indispensable, since thermal shock fracture is
caused by the accumulation of microscopic damages such as microcrackings under
thermal stress. On the other hand, the authors investigated the microfracture process
during bending tests of ceramics using AE technique.6,7 Especially, the critical stress
for maincrack formation due to the growth and/or coalescence of microcracks was
evaluated by detecting the critical increase in AE signals.
In this study, new experimental technique for the investigation of thermal shock
fracture behavior, Disk-on-Rod test, is proposed. The temperature fields in the speci-
men were measured and used to calculate the 2 dimensional thermal stress field.
Shuichi Wakayama, Kousuke Nishino, Kouhei Kubota, Toshiya Wada and Satoshi Kobayashi, Depart
ment of Mechanical Engineering, Tokyo Metropolitan University, 1 1 Minami Ohsawa, Hachioji shi,
Tokyo 192 0397, JAPAN
365
366 Shuichi Wakayama et al.
Furthermore, microfracture process was characterized by AE measurement. The for-

mation of maincrack due to propagation and/or microcracks was focused in the present
paper. Especially, the critical stress for maincrack formation is evaluated from obtained
experimental data and compared with the results of mechanical biaxial bending tests.
2.1. Materials
The materials used in the present study were alumina ceramics (ADS-11) offered
from Toshiba Ceramics Co., Ltd. The relative density and mean grain size were
97.7% and 6 mm, respectively. Disk specimens were cut from the rod materials and
both surfaces were polished. The disk specimens with a diameter of 20 mm and a
thickness of 0.6 mm were then obtained. In order to avoid the influence of stress
corrosion cracking by water contents in materials, the specimens were dried in the
vacuum at 1508C for 2 hours.
2.2. Disk-on-Rod Tests

In this study, for characterizing the thermal shock fracture process, new experimental
technique, Disk-on-Rod test, was developed. An outline of the Disk-on-Rod test is
shown in Figure 1. A thin disk specimen was heated to the required temperature by
an infrared (IR) lamp and only the central part of disk was quenched by means of
contacting with a metal rod. Since the disk specimen has a diameter of 20 mm and a
thickness of 0.6 mm, and a contacting area has 4 mm diameter, 2-D thermal stress
field was obtained. In order to characterize the fracture process, AE signals during
thermal shock fracture were detected by an AE sensor attached on the bottom end of
a metal rod, thus it was used for both coolant and wave guide.
The apparatus of Disk-on-Rod test and AE measuring system are shown in
Figure 2, schematically. The heated specimen was brought down with a holder and
IR Heater
Disk Specimen
Metal Rod
AE Sensor
Heating Quenching
Figure 1. Outline of Disk on Rod Test.

Experimental Approach to Microfracture Process of Ceramics 367
IR camera
Disk Specimen IR Lamp
Rod
PC Pre-Amp
AE Sensor PC
Air Dumper
Figure 2. Schematic Diagram of Disk on Rod Test System.
contacted with a metal rod. For removing the generation of AE noise due to
contacting, contacting speed was controlled by an air damper. And temperature
distributions on disk surface were measured by a high-speed infrared camera (30
frames/s). Then thermal stress was calculated from the measured temperature distri-
bution using FEM analysis.
An AE sensor, in which amplifier is instrumented, with resonant frequency of
180 kHz was used, then the initiations of microcracks could be detected with excel-
lent sensitivity. The total gain of the AE system was 75 dB (main amplifier; 20 dB
and pre-amplifier with sensors; 55 dB) and the threshold level was 18 mV at the input
terminal of pre-amplifier.
2.3. Piston-on-Ring Tests

In order to obtain the strength under biaxial mechanical loading, Piston-on-Ring
tests were carried out. The disk specimen has a same geometry as Disk-on-Rod tests.
The specimen was supported by a supporting ring with a diameter of 16 mm and
loaded by a loading rod with a diameter of 4 mm. Then the maximum equi-biaxial
stress was obtained at the center region of specimen. According to the ASTM test
standard (F394-78), the maximum qui-biaxial stress, smax , can be calculated by the
following equation.
a 1 n 2
3P(1 þ n) b2 a
smax ¼ 1 þ 2 ln þ 1 2 (1)
4pt2 b 1þn 2a R2
where P is the load, n is the Poisson’s ratio, a is the radius of supporting ring, b is the
radius of loading rod, and t and R are the thickness and radius of the disk specimen,
respectively.
3. EXPERIMENTAL RESULTS AND DISCUSSION
3.1 Determination of Thermal Stress

The temperature on specimen surface was measured by 1/30 s using a high-speed
infrared camera system during Disk-on-Rod test and the thermal stress field was
computed from the obtained 2-dimensional temperature distribution using finite
element code (ANSYS; isoparametric structural shell, 2977 elements, 6082 nodes).
Young’s modulus (360 GPa), Poisson’s ratio (0.22) and the coefficient of thermal
expansion (7:8 10 6 =K) were assumed as constant in the FEM analysis. The
determined thermal stress fields are shown in Figure 3 (a) radial stress and (b)
tangential stress. The obtained stress field was axi-symmetric according to the
concentric temperature distribution. It is understood from the figure that the center
of the disk is subjected to the maximum equi-biaxial stress and the stresses decrease
250
3.0s
200
5.0s
Thermal Stress / MPa
150 1.0s
100
50 0.5s
−50
−100
−10 −5 0 5 10
Radius / mm
(a) Radial Stress
250
3.0s
200
1.0s
Thermal Stress / MPa
150
100 5.0s
50
0.5s
0
−50
−100
−10 −5 0 5 10
Radius / mm
(b) Tangential Stress
Figure 3. Thermal Stress Distribution during Disk on Rod Test (a) Radial Stress, (b) Tangential Stress.
to the outer region along the radial direction. It is also seen in the figure that the
radial stress is tension in the whole of disk specimen, while the tangential stress is
tension in the central region but compression in the outer region. Both stresses show
the maximum at 3 s and decrease gradually.
3.2 Disk-on-Rod Test

Figure 4 shows a typical result of Disk-on-Rod test. Figure 4 (a) shows the fractured
specimen. The crack path observed in the figure demonstrates that the maincrack was
formed at the center region, which was subjected to the maximum equi-biaxial stress,
and propagated to outer region associating with slight deflections.
Maximum thermal stress at the center of specimen, cumulative AE energy
and cumulative AE event are shown in Figure 4 (b). It is recognized in the figure
that both cumulative AE events and energy increase remarkably at 1.8 s, continue to
increase intermittently until 2.5 s and cease to generate.
(a) Fractured Specimen

350 500 200
300 th
400
150
250
Cumulative AE Energy/V2
Thermal Stress/MPa
Cumulative AE Event
200 Thermal Stress 300

Cumulative AE Energy
100
150
200
100 Cumulative AE Events

50
100
50
0 0 0
0 1 2 3 4 5
Time/s
(b)Thermal Stress and AE Behavior
Figure 4. Typical Results of Disk on Rod (a) Fractured Specimen (b) Thermal Stress and AE Behavior.
Since the formation of maincrack could be observed at the AE increasing point,

it is understood that the remarkable increase in AE corresponds to the main-
crack formation due to microdamage accumulation. Then the critical stress for
maincrack formation during thermal shock fracture, sth , can be determined by
Disk-on-Rod test. Those values ranged from 220 to 330 MPa. The instantaneous
crack path was not determined definitely because of its high growth velocity. There-
fore, the thermal stress analysis is no longer valid after maincrack formation because
the crack path could not be introduced into finite element models.
It can be also seen in Figure 4 (b) that the AE events increase intermittently after
the critical increasing. Therefore, it can be considered that the maincrack propagated
into the outer region to the final fracture at 2.5 s. Furthermore, the incubation period
was observed in AE generation from 1.9 to 2.1 s, which suggest the arrest and re-
propagation of maincrack occurred during thermal shock fracture. Consequently, it
can be concluded that the whole process associated with thermal shock fracture
presented in the unified theory of Hasselman2, i.e. crack initiation, arrest and
propagation, can be identified during a single Disk-on-Rod test.
3.3 Piston-on-Ring Test
In order to investigate the critical stress for maincrack formation under biaxial
mechanical loading, the Piston-on-Ring tests were carried out. Figure 5 shows a
fractured specimen (Figure 5 (a)), and the maximum biaxial stress and AE behavior
during Piston-on-Ring test (Figure 5 (b)). It is observed in Figure 5 (a) that the
maincrack was initiated at the center region of the specimen subjected to equi-biaxial
stress and propagated to the outer region.
In Figure 5 (b), the AE increasing point can be observed at 80 s before the final
fracture at 107 s. Similarly to the previous studies,6,7 it is understood that the critical
stress for maincrack formation under mechanical loading, sC , as well as bending
strength, sB , can be determined. The critical stresses, sC , were evaluated as
180 280 MPa; they were lower than the critical stress for maincrack formation
under thermal shock mentioned above.
3.4 Critical Stress for Maincrack Formation
For the quantitative comparison of the critical stress for maincrack formation under
thermal shock, sth , with mechanical loading, sC , the volume effect of strength should
be considered. According to Weibull statistics,8 the cumulative probability of failure
of brittle materials is written in the following simplified form.
P ¼ 1 exp½(sR =s0 )m VE (2)
The parameters m and s0 are the shape parameter (Weibull modulus) and the scale
parameter (characteristic strength), respectively. sR is the maximum stress in the
structure and VE is the effective volume.
ð
VE ¼ (s=sR )m dV (3)
V
From Eq. (2), the average strength, m: can be obtained as

(a) Fractured Specimen

400 150 120
B
350
100
300
Plane Bend stress/MPa
Cumulative AE Energy/V2
Cumulative AE Events
100 80
250 C
200 60
Plane Bend Stress

150
50 40
100
Cumulative AE Energy 20
50
Cumulative AE Events
0 0 0
0 20 40 60 80 100 120
Time/s
(b) Plane Bending Stress and AE Behavior
Figure 5. Typical Results of Plane Bending Test (a) Fractured Specimen, (b) Plane Bending Stress and
AE Behavior.
1=m
m ¼ s0 VE G((m þ 1)=m) (4)
where G is a gamma function. From Eq. (4), the volume effect of strength can be
derived as
m2 =m1 ¼ (VE1 =VE2 )1=m (5)
where m1 and m2 are the average strength of the structure with the volume of VE1 and
VE2 , respectively.
These statistical approaches can not be applied directly to the experimental results
in this study, because the fracture was occurred under biaxial stress in Disk-on-Rod
tests and Piston-on-Ring tests. The statistical approaches for multiaxial stress state
have been studied by several investigators.9,10 In order to determine the suitable
equivalent stress more precisely, further investigation is needed for the widely variable
stress states. However, it is considered that uniaxial statistical approach by eqs. (2) (5)
are available for the comparison of the critical stress for maincrack formation under
thermal shock, sth , with mechanical loading, sC , since the maincrack was formed at the
center of disk specimen subjected to equi-biaxial stress in both cases.
For the first assumption, the maximum principal stress was taken as the equiva-
lent stress in the present study. Inserting the experimental data into Eqs. (3) and (4),
the average value of critical stress for maincrack formation and effective volume were
calculated as 233 MPa and 4:1 mm3 for plane bending tests, and 271 MPa and
3:4 mm3 for thermal shock tests, respectively. On the other hand, considering the
volume effect (Eq. (5)), the critical stress under thermal shock can be estimated as
238 MPa from the experimental results of plane bending tests, which is lower than
the experimental results of thermal shock tests. Thermal shock tests were carried out
at high temperature while plane bending tests at room temperature in air. Therefore,
it is considered that the influence of stress corrosion cracking by water vapor resulted
in the deference in the estimated value and experimental results.
4. CONCLUSIONS
In the present study, new experimental technique, Disk-on-Rod test, was developed
and applied to the characterization of thermal shock fracture process in ceramics.
The thermal stress fields were computed from temperature distributions measured by
high-speed IR camera and fracture process was characterized by AE measurement,
both of which are significant for understanding the thermal shock fracture behavior.
During thermal shock fracture, the maincrack was formed at the center of disk
specimen subjected to maximum equi-biaxial stress and propagated into the outer
region with crack deflections. The critical stress for maincrack formation was suc-
cessfully determined by AE measurement and FEM analysis using the measured
temperature distribution. Furthermore, those were compared with the critical stress
under mechanical biaxial loading, considering the volume effect based on Weibull
statistics. It was then understood that the critical stress for maincrack formation
under thermal shock is higher than that under mechanical loading because of the
absence of stress corrosion cracking by water vapor.
Further investigation is needed for the application to complex multiaxial stress
states and the fracture mechanical analysis of crack arrest and propagation process.
However, new experimental technique of thermal shock fracture (Disk-on-Rod test),
by which the indispensable information for the structural application of ceramics at
high temperature, was established in the present study.
ACKNOWLEDGMENT
The authors would like to thank Toshiba Ceramics Co., Ltd. for their offer of
ceramic samples.
REFERENCES
1. W. D. Kingery, Factor Affecting Thermal Shock Resistance of Ceramic Materials, J. Am. Ceram. Soc.
38(1), 3 15 (1955).
2. D. P. H. Hasselman, Unified Theory of Thermal Shock Fracture Initiation and Crack Propagation in
Brittle Materials, J. Am. Ceram. Soc. 52(11), 600 604 (1969).
3. G. A. Schneider and G. Petzow, Thermal Shock Testing of Ceramics A New Testing Method, J. Am.
Ceram. Soc. 74(1), 98 102 (1991).
4. W. P. Rogers and A. F. Emery, Contact Thermal Shock Test of Ceramics, J. Mater. Sci. 27(1), 146
152 (1992).
5. Y. Mizutani, N. Nishikawa and M. Takatsu, Evaluation of Thermal Shock Damage in Ceramic
Disks Heated under Constant Heat Flux Conditions, J. Ceram. Soc. Jpn. 103(5), 494 499 (1995).
6. S. Wakayama, T. Koji and H. Nishimura, Evaluation of Critical Stress for Microcracking in Alumina
Ceramics by Acoustic emission, Trans. Jpn. Soc. Mech. Eng. 57(535), 504 511 (1991).
7. S. Wakayama and H. Nishimura, Critical Stress for Microcracking in Alumina Evaluated by
Acoustic Emission, Fracture Mechanics of Ceramics, 10, Plenum Press, New York, 59 72 (1992).
8. W. A. Weibull, Statistical Distribution Function of Wide Applicability, J. Appl. Mech. 18(3), 293 297
(1951).
9. S. B. Batdorf and J. G. Crose, A Statistical Theory for the Fracture of Brittle Structures Subjected to
Polyaxial Stress States, J. Appl. Mech. 41, 459 465 (1974).
10. J. Lamon, Statistical Approaches to Failure for Ceramics Reliability Assessment, J. Am. Ceram. Soc.
71(2), 106 112 (1988).
FINITE ELEMENT ANALYSIS
OF CRACK-PATH SELECTION
IN A BRICK AND MORTAR
STRUCTURE
Reza Sarrafi-Nour, Mohan Manoharan

and Curtis A. Johnson*
ABSTRACT
Many natural composite materials rely on organized architectures that span several
length scales. The structures of natural shells such as nacre (mother-of-pearl)
and conch are prominent examples of such organizations where the calcium carbo-
nate platelets, the main constituent of natural shells, are held together in an organized
fashion within an organic matrix. At one or multiple length scales, these organi-
zed arrangements often resemble a brick-and-mortar structure, with calcium carbon-
ate platelets acting as bricks connected through the organic mortar phase.
The structure of natural shells and their outstanding resistance to damage and
fracture have both attracted and inspired scientists from various disciplines, including
material science, for a number of years. It is believed that the organized nature of the
brick-and-mortar structure imparts a significant influence on the unusually high frac-
ture resistance and work-of-fracture of the natural shells. Hence, the interaction
between the growing cracks and the shell architecture is of central interest in any
attempt to mimic the structure and potentially design materials based on such arch-
itectures. This work presents a finite element analysis of crack-path selection in a simple
model brick-and mortar structure. The analysis relies on interface fracture mechanics
concepts to gain insight into the pathological behavior of cracks propagating in such
*
Ceramics and Metallurgy Technologies, General Electric Global Research Center, Niskayuna, NY
12309, USA. E mail: [email protected], Tel. (518) 387 5803, Fax (518) 387 7563
375
376 Reza Sarrafi-Nour et al.
structures. The ratio of the strain energy release rate between the interfacial and non-
interfacial crack-propagation directions is used as an indicator in evaluating the
tendency towards interfacial crack growth. The strain energy release rate ratio is then
evaluated as a function of elastic properties and the characteristic geometrical dimen-
sions of the brick and the mortar phase.
1. INTRODUCTION
Natural shells are known to exhibit unique fracture resistance characteristics, inspir-
ing materials scientists to study and to attempt mimicking their structures. Some
prominent examples of natural shells include nacre [1,2] and conch [3,4]. Both conch
and nacre are composed of more than 95% calcium carbonate, mostly aragonite
platelets, and less than 5% of a proteinaceous material which is believed to act as a
mortar or a binding matrix. In the microstructure of nacre, the calcium carbonate
platelets are highly tessellated and resemble a well-organized brick-and-mortar struc-
ture at a micrometer length scale. The microstructure of conch shell is far more
complex; it is laminated and consists of three distinct layers that resemble the
structure of plywood [3,4]. However, at lower length scales the microstructure of
conch shell also consists of lamellae and, further down, of high aspect ratio calcium
carbonate platelet arrangements. At both of these levels, the structure of conch also
resembles a brick and mortar structure.
It is believed that the organized nature of a brick-and-mortar structure imparts a
significant influence on the unusually high fracture resistance and work-of-fracture
of these natural shells. Hence, the interaction between growing cracks and these
structural arrangements is of central interest in any attempts to mimic and material
design based on such architectures. By examining the pathological behavior of cracks
as they propagate into a structure, one can gain fundamental understanding and
useful insight into such interactions. It is the goal of this article to gain such insights
on cracks propagating in a brick-and-mortar structure arrangement.
2. ANALYSIS METHODOLOGY
We employ the finite element method and interface fracture mechanics concepts to
parametrically examine interfacial crack path selection in a unit brick-and-mortar
structure. The analysis methodology, to be described below briefly, is based on the
analyses of cracks kinking into and out of a bi-material interface as discussed by He
and Hutchinson [5], He et al [6,7], and Suo and Hutchinson [8].
For a crack intersecting the bi-material interface, the basis for the analysis [6,7] is
the relative energy release ratio for crack propagation along the two competing
directions: the bi-material interface, Gi , and penetrating into the material across the
interface, Go . Comparison between the release rate ratio, Gi =Go and the ratio of
fracture energy, Gi =Go along the two competing directions would then establish a
criterion as which direction is favored. If Gi =Go > Gi =Go the condition for interfacial
crack propagation is met before that of penetration and hence cracks would propagate
along the interface. Therefore, Gi =Go can be viewed as a measure of the tendency for
interfacial path selection. When a crack is already at the bi-material interface, the
Finite Element Analysis of Crack-Path Selection 377
competition between further interfacial crack growth and the crack kinking out of the
interface could be assessed using a similar criterion. In this case, the crack propagation
process is controlled by Gi =Gd > Gi =Gd [5] where the index ‘‘d’’ refers to the crack
kinking out of an interface into the material on one side of the interface. Nevertheless,
complexities would arise in adopting a unified criterion to establish the direction of the
kink crack as discussed by He and Hutchinson [5]. For most material combinations
and loading conditions, both the well-known KII ¼ 0 and Gmax criteria have been
shown [5] to predict virtually the same direction for the kink crack. For certain
material and loading combinations (in bi-material terminology when a, the first
Dundurs’ parameter representing the mismatch between Young’s modulus of the
two materials, is less than 0:6 and the crack kinks into the stiffer material), however,
the two criteria were found to suggest substantially different kink angles, giving rise to
some ambiguities. He and Hutchinson [5] suggested that under such conditions the
selection of a criterion for kink directions had to be guided by experiments.
In this paper, we examine the pathological behavior of cracks propagating in
model brick-and-mortar structures using the interface fracture mechanics method-
ology described while adopting a KII ¼ 0 criteria for the crack kinking out of an
interface.
3. FINITE ELEMENT MODEL AND ANALYSIS
The interface crack propagation analysis was simulated in a two-dimensional brick-

and-mortar model structure consisting of six brick units along the length and six
brick units along the width of the model. FRANC2D [9] finite element code was used
for fracture mechanics analysis due to the ease of crack insertion and propagation,
and due to implementation of some interface crack propagation capabilities in the
code. The finite element realization of the model and the boundary conditions are
shown in Figure 1a. The finite element model was first created in ANSYS 6.0 with
proper material properties assigned to the brick and mortar regions, assumed to be
isotropic elastic. This model consisted of about 6,600 14,300 eight- and six-noded
isoparametric plane-strain elements (19,500 41,000 nodes) depending on the geo-
metrical ratios of the brick and mortar regions. The crack free model was next read
into FRANC2D and a crack was inserted into the model at the desired location. In
one subset of the analyses, the crack was inserted along the centerline of the vertical
mortar region at the mid point on one side of the model, as shown in Figure 1b, to
simulate the conditions when a crack terminated at the bi-material interface. In
another subset of the analyses, this crack was further propagated to the mid point
along the adjacent horizontal interface between the brick and mortar region as shown
in Figure 1c. (Details about crack insertion, propagation and fracture mechanics
analysis using FRANC2D can be found in contributions by Bittencourt et al [10] and
Wawrzynek and Ingraffea [11 13].) Further mesh refinements were employed in the
models during the course of crack insertion and propagation analyses to ensure a fine
mesh in the regions surrounding the crack tip. The crack tip itself consisted of a
rosette of eight quarter-point elements. After the crack insertion, the total number of
nodes in the models increased to some 25,000 60,000.
The energy release rates for the interface and kinked cracks were evaluated using
the equivalent domain version of elastic J-integral [14]. These values were further
Figure 1. Finite element model and the definitions of evaluated configurations of the interface cracks: a)
FE mesh with no cracks, b) the crack intercepting the interface between the brick and the mortar and c) the
crack at the interface between the mortar and the brick unit. The region marked with the white circle
corresponds to the location where the crack configurations (b) and (c) were analyzed.
used to calculate the Gi =Go or Gi =Gd depending on the crack configuration. The
implementation of the analysis of a crack kinking out of an interface in FRANC2D
has not been documented in the past. In short, the analysis proceeds as described in
the followings. When the parent crack is at an interface, the kink directions into the
materials on either side of the interface are determined based on the maximum
circumferential stress theory [15], a close approximation of KII ¼ 0 criterion, in
homogenous materials. Due to difficulties involved in re-meshing the region between
a shallow kink and the bi-material interface, a kink direction with an angle of less
than three degrees with the bi-material interface is neglected. Otherwise, a kink crack
with a length corresponding to a fixed fraction of the size of the singular elements at
the tip of the parent crack would be extended from the parent crack tip. The analysis
was then invoked to calculate the energy release rate of the new crack configuration.
This sequence was repeated for any other potential directions including the one along
the bi-material interface by recovering the original parent crack mesh configuration.
Another important quantity in interface fracture mechanics is mode mixity. The
crack tip field of an interface crack is intrinsically mixed mode [8] due to the
asymmetric elastic properties across the interface. Hence, mode mixity is required
to fully characterize the loading conditions at the crack tip. Furthermore, the fracture
toughness of an interface is known to depend on the mixed mode condition, typically
observed to rise with increasing mixed mode. The degree of the mode mixity can be
characterized using the definition of the complex interface stress intensity factor
K ¼ iK1 þ K2 after He and Hutchinson [5]:

1 Im[K]
w ¼ tan Eq:(1)
Re[K]
where w is the mode mixity or phase angle, and Im[K] and Re[K] are the imaginary
and real components of the stress intensity factor describing the stress field at the tip
of the interface crack:
K1 þ iK2 i«
s22 þ is12 ¼ p r Eq:(2)
2pr
In Eq. (2), s22 and s12 are normal shear stresses ahead of the crack tip, r is the
distance from the crack tip and e is the so-called oscillation index, which is related to
the second Dundurs’ elastic mismatch parameter b (see the appendix for the defin-
ition of Dundurs’ parameters):
1 (1 b)
«¼ ln Eq:(3)
2p (1 þ b)
Various techniques, such as interaction integral [16] and direct K solution using
enriched finite elements [17], have been used to calculate K2 =K1 ratio for an interface
crack. Here, we used the near-tip asymptotic displacement field solution of Hutch-
inson and Su [8] to calculate the K2 =K1 ratio, and hence w, from the displacements of
the quarter-point nodes behind the crack tip. The required derivations are brought in
the Appendix. An independent comparison between w results from a direct K solution
using enriched finite elements1 and those obtained by inserting the displacement
results from FRANC2D into the equation derived in the Appendix were found to
agree within 1 3%.
4. RESULTS AND DISCUSSIONS
Examples of the variation of energy release rate and mode mixity in reference to
crack-tip positions depicted in Figure 1b and Figure 1c are shown in Figure 2 for two
different ratios of mortar thickness, t, to brick with, W, at a constant aspect ratio of
the bricks, L/W. Two key observations could be made from the data. First, at the
onset of the crack propagation along the horizontal boundaries, there was a transi-
tion region where both Gi =Gd and the mixity of the interface crack passed through
minimums. A closer examination of the data indicated that the minima of both Gi =Gd
and the mixity were located at a distance of t=2 from initial crack tip position, the
mid-point of the vertical mortar region. Second, both the release rate ratio and mode
mixity along the interface tended to rise towards a plateau as the crack tip reached
the mid point of the horizontal boundary. In addition, while the release rate ratio
showed a strong dependence on t/W, the mixity appeared to be far less sensitive
to t/W (at least for the combination of the geometrical and material properties of
the models examined).
Any attempts to mimic such natural structural arrangements using ceramic
materials needs further understanding of how the geometrical ratios and material
properties may contribute/control the crack path selection process in such structures.
One could approach such evaluations through a full parametric study of the problem.
However, given the number of interations and numerical efforts involved, a response-
surface-methodology-based approach is more practical [18].
We employed such an approach to evaluate the effects of the aspect ratio of the
brick units, the thickness of the mortar phase and the elastic properties of each phase.
Guided by the preliminary analyses, the study was carried out for two particular
crack tip locations, i.e., the crack configuration shown in Figure 1b and after crack
propagation along the brick and mortar interface by 2x=L ¼ 0:5 (Figure 2). The
selection of this second configuration was primarily driven by the plateau behavior
1
FRAC2D Finite Element Code for 2D and Axisymmetric Fracture Analysis, General Electric Global
Research, Niskayuna, NY
Figure 2. Variation of energy release ratio and mixity as function of position of crack tip at the brick mortar
layer interface, Eb 320 GPa, Em 20GPa, nb nm 0:25. Note that the energy release ratio values at
2x=L 0 represent Gi =Go , the energy release ratio between interfacial and penetrated crack paths.
suggested by the mixity results (Figure 2) and by an apparent steady-state behavior of

Gi =Gd for small t/W values (Figure 2) at this location. The range of elastic properties
considered in the analysis was selected based on the properties of ceramic materials
of interest for structural applications. Table 1 summarizes the geometrical and
material properties considered.
A design-of-experiment software package2 was used to generate the design3
points that were suitable for a quadratic response surface analysis of the design
space with proper constraints on the variables. A constraint imposed while generat-
ing the design matrix was that the modulus of the brick had to be always greater than
the modulus of the mortar phase. The resulting design matrix consisted of 39
independent runs where the variables were changed systematically within their
selected range. Finite element analyses were carried out for the two crack configur-
ations using the sets of variables at each design point and the results (calculated
Table 1. Material and geometrical properties used in the parametric

study.
Parameter Range
Elastic modulus of brick, Eb , and mortar, Em 20 GPa 700 GPa

Poisson’s ratio of bricks, nb , and mortar, nm 0.1 0.3
Length to Width ratio, L/W, of bricks 2 10
Ratio of the mortar thickness to brick width, t/ W 0.05 0.5
2
Design Expert Version 6.0, Stat Ease Inc., Minneapolis, MN
3
D Optimal, Response Surface Design, Design Expert Version 6.0
energy release rate ratios and mixities) were analyzed using the same design-of-
experiment software package to obtain quadratic response surfaces for both the
energy release rate and mixity as a function of the design variables (Table 1). The
response surface functions obtained are illustrated in Figs. 3 5 for two different
combinations of elastic moduli of the brick and the mortar phases. It is to be kept
in mind that the response surface functions here are ‘‘best-fit’’ functions obtained by
regression analysis of the data with respect to the selected variables within their
specific domains. All the fit functions shown here had R-squared statistics of about
> 0.93, except for the fit function (not shown here) to the mixity results of the
interface crack tip at 2x=L ¼ 0:5.
As seen from the surface plots in Figure 3 and Figure 4, the energy release rate
ratio is strongly influenced by the aspect ratio of the brick and the mortar units. The
positive influence of the elastic mismatch across the interface on the energy release
rate ratio was expected based on the previous studies of interface cracks [5 8]. It was
also interesting to note that the results shown in Figure 3 suggested that a combin-
ation of high aspect ratio bricks with high elastic mismatch between the brick and
mortar would require a thin mortar layer in order to achieve high release rate ratio.
This could mean that for such combinations of geometry and material properties a
thin mortar layer would better facilitate crack deflection into the brick and mortar
interface. In contrast, when the crack has grown into the interface between the brick
and mortar layer, the energy release ratio appeared to become far less sensitive to the
thickness of the mortar layer and increased monotonically with increasing thickness
for all values of the aspect ratios of the bricks.
Figure 3. Variation of energy release ratio of interfacial versus penetrated crack as a function of the
aspect ratio of the brick and thickness ratio of the mortar units for crack configuration in Figure 1b. Upper
surface: Eb 400GPa, Em 20GPa. Lower surface: Eb 400GPa, Em 100GPa: nb nm 0:2
Figure 4. Variation of energy release ratio of interfacial versus deflected crack as a function of the
aspect ratio of the brick and thickness ratio of the mortar units for interface crack configuration
shown in Figure 1c. Upper surface: Eb 400GPa, Em 20GPa. Lower surface: Eb 400GPa,
Em 100GPa: nb nm 0:2
Figure 5. Variation of mixity of the interface crack at the onset of the deflection into the inter
face between the brick and mortar layers as a function of the aspect ratio of the brick and the
thickness ratio of the mortar units. Upper surface: Eb 400GPa, Em 20GPa. Lower surface:
Eb 400GPa, Em 100GPa: nb 0:2, nm 0:3
It is noteworthy that when the elastic mismatch between the bricks and the
mortar layer was low, the energy release ratio of the crack configuration deflecting
into an interface (Figure 3) gave a value of 0:2---0:3. This was consistent with earlier
suggestions related to the requirements for the interface toughness ratio in reinforced
ceramics. In earlier studies, Evans [19] suggested the toughness ratio between the
interface and matrix had to be #0:25 in order to achieve toughening; Becher [20]
suggested that the toughness ratio between the interface and the reinforcement
needed to be 1=6 [20] in order to ensure debonding at the interface between the
bridging ligament and the matrix.
The results shown in Figure 4, where the crack has already grown into the
interface between the brick and mortar layer, suggested that the energy release rate
ratio increased monotonically with increasing thickness of the mortar layer. How-
ever, the aspect ratio of bricks appeared to have a stronger influence on the energy
release ratio.
5. SUMMARY AND CONCLUSIONS
Based on interface fracture mechanics of brittle solids, the conditions of crack growth
into or out of an interface between two dissimilar materials can be assessed by
evaluating the ratio of energy release rate for crack propagating along the two
competing directions. A finite element implementation of this concept was employed
to evaluate the ratio of the energy release rate of cracks growing into and out of the
interface between two dissimilar materials in a brick-and-mortar structure. A para-
metric study was carried out to examine the effect of both the geometrical and elastic
properties of brick and mortar units on the energy release rate ratio between the
interface crack and the competing kink-out or penetrating cracks. Consistent with
earlier studies of interface cracks, elastic mismatch between the brick and mortar
units was found to influence the energy release ratio strongly. The results also
indicated a strong influence of geometrical characteristics. In the case of the crack
configuration at the brick-mortar interface, a combination of high aspect ratio of
bricks, low thickness of the mortar and high elastic mismatch appeared to achieve the
highest energy release ratio. The results for the crack configuration already grown
into the interface between the brick and mortar layer suggested a mild and mono-
tonic increase in energy release ratio with increasing thickness of the mortar layer.
However, in this case the energy release ratio appeared to be more sensitive to the
aspect ratio of the bricks.
5. ACKNOWLEDGEMENTS
Ahmet Kaya, GE Global Research is acknowledged for useful and stimulating

discussions, and for running comparative analysis in FRAC2D to verify the mixity
calculations. Bruce Carter, GE Power Systems, and Paul Wawrzynek, Cornell Frac-
ture Group, are acknowledged for their efforts to improve the interface fracture
routines and to increase the problem-size capability of FRANC2D in order to
accommodate analysis of large models.
REFERENCES
1. J.D. Currey, Proc. R. Soc. London, Ser, B, 196, 443 (1977)

2. R.Z. Wang, Z. Suo, I.A. Aksay and A. G. Evans, J. Mater. Res., 16 [9] 2485 2493 (2001)
3. S. Kamat, X. Su, R. Ballarini and A.H. Heuer, Nature, 405 [29] 1036 1040 (2001)
4. L.T. Kuhn Spearing, H. Kessler, E. Chateau, R. Ballarini, A.H. Heuer and S.M. Spearing, J. Mat.
Sci., 31, 6583 6594 (1996)
5. M Y. He, J.W. Hutchinson, Trans. ASME, 56, 270 278 (1989)
6. M Y. He, A. Bartlett, A.G. Evans and J.W. Hutchinson, J. Am. Ceram. Soc., 74 [4] 767 71 (1991)
7. M Y. He, A.G. Evans and J.W. Hutchinson, Int. J. Solids Struc., 31 [24], 3443 3455 (1994)
8. J.W. Hutchinson, Z. Suo, Adv. Appl. Mech., 29, 63 191 (1992)
9. P. Wawrzynek and A.R. Ingraffea, NASA Contractor Report 4572, Cornell University (1994).
10. T. Bittencourt, P.A. Wawrzynek and A.R. Ingraffea, Eng. Fract. Mech., 55 [2] 321 334 (1996)
11. P.A. Wawrzynek and A.R. Ingraffea, Theo. Appl. Fract. Mech., 8, 137 150 (1987)
12. Ibid, Eng. With Computers, 3, 13 20 (1987)
13. Ibid, Finit. Elem. Anal. Des., 5, 87 96 (1989)
14. R.H. Dodds, Jr. P.M. Vargas, Report, University of Illinois at Urbana Champaign, Dept. of Civil
Engineering (1988)
15. F. Erdogan and G. Sih, ASME J. Bas. Eng., 85, 519 527 (1963)
16. L. Banks Sills, N. Travitzky, D. Ashkenazi and R. Eliasi, Int. J. Fract., 99, 143 160 (1999)
17. H.F. Nied and A.C. Kaya, General Electric Technical Report, 92CRD028
18. R.H. Myers, D.C. Montgomery, ‘‘Response Surface Methodology: Process and Product Optimiza
tion using Designed Experiments,’’ Wiley Interscience Publication, New York (1995).
19. A.G. Evans, ‘‘Perspective on the Development of High Toughness Ceramics,’’ J. Am. Ceram. Soc., 73
[2] 187 206 (1990)
20. P.F. Becher, ‘‘Microstructural Design of Toughened Ceramics,’’ J. Am. Ceram. Soc., 74 [2] 255 69
(1991).
APPENDIX
A.1. DUNDURS’ ELASTIC PARAMETERS
A wide class of elastic plane problems related to bimaterials depends on only two
non-dimensional combinations of the elastic moduli known as Dundurs’ parameter a
and b:
G1 (K2 þ 1) G2 (K1 þ 1)
a¼ Eq:(A1)
G1 (K2 þ 1) þ G2 (K1 þ 1)
G1 (K2 1) G2 (K1 1)
b¼ Eq:(A2)
G1 (K2 þ 1) þ G2 (K1 þ 1)
where G and K are the shear and bulk moduli and the indices 1 and 2 refer to
materials 1 and 2 on either side of a bimaterial interface.
A.2. CALCULATION OF THE MIXITY AT THE TIP OF AN

INTERFACE CRACK
The near-tip asymptotic displacement field of a crack at the interface between two
isotropic elastic solids after Hutchinson and Suo reads:
r
8(K1 þ iK2 ) r i«
d2 þ id1 ¼ r Eq:(A3)#
E (1 þ 2i«) cos h(p«) 2p
where di ¼ ui (r, u ¼ p) ui (r, u ¼ p) are the crack surface displacements at dis-
tance r behind the crack. Replacing: 1 þ 2ie cos h(pe) ¼ 1 þ iA, rie ¼ Cos(e ln (r) )
þi Sin(e ln (r)), e ln (r) ¼ y and B ¼ (8=E )(r=2p)1=2 and rewriting Eq. A3 after
some manipulations gives:
d1 K1 (Sin( y) ACos( y) ) þ K2 (Cos( y) þ ASin( y) )

¼ Eq:(A4)
d2 K1 (Cos( y) þ ASin( y) K2 (Sin( y) ACos( y) )
Eq. A4 can be further simplified by dividing both the nominator and the denomin-
ator on the right-hand side by K2 [Sin(y) ACos(y)] to give:
K1 Cos( y) þ ASin( y)
þ
d1 K Sin( y) ACos( y)
¼ 2 Eq:(A5)
d2 K1 Cos( y) þ ASin( y)
1
K2 Sin( y) þ ACos( y)
Using the abbreviations g ¼ (Cos(y) þ ASin(y) )=(Sin(y) ACos(y) ) ¼ and f ¼
d2 =d1 , further rearrangement of Eq. A5 gives
K2 f :g 1
¼ Eq:(A6)
K1 f þg
and, hence, f ¼ tan 1 (K2 =K1 ).
Table A1 shows the verification of the mixity values calculated based on Eq. A6
using near-tip displacement solutions from finite element analysis. The crack config-
Table A1. Verification of Eq. A6 using finite elementEnriched Element Solution#

tan1 (K2 =K1 ) ¼ 1:0537 radian.
Distance from the crack tip, r DUx DUy tan1 (K2 =K1 ) [rad] from Eq. (A6)
FRAC2D*
0.005504 1.4809E 05 5.3778E 05 1.0660
0.011007 2.6683E 05 7.3379E 05 1.0702
0.023603 4.7940E 05 10.2779E 05 1.0744
0.037788 6.7124E 05 12.3337E 05 1.0848
FRANC2D (quarter point node)
0.002232 6.7927E 06 3.5325E 05 1.0860
0.003348 1.0014E 05 4.2355E 05 1.0843
0.008928 2.3735E 05 6.6502E 05 1.0871
# Eb ¼ 320GPa, Em ¼ 20GPa, nb ¼ nm ¼ 0:25

*
FRAC2D mesh consists of enriched elements at the crack tip and cubic elements elsewhere
#
It can be seen in Eq. A1 that the classical definition of interface stress intensity factors includes the
complex oscillatory term (length)ie . Therefore, the stress intensity factor unit in this case is not simply
(stress) (length)1=2 but (stress) (length)1=2 (length)ie . As suggested by Rice, the complex stress inten
sity factor K can be rotated in such a manner to reduce to the conventional form using K0 KLie . In this
case, L is a characteristic length (chosen arbitrarily) of the interface crack problem. This will allow one to
define a phase angle in the rotated coordinates by using f0 f þ e ln (L).
uration analyzed was the same as shown in Fig. 1c. The K2 =K1 at the crack tip was
calculated in FRAC2D through a direct solution of the stress intensity factors K1
and K2 using enriched finite elements.
FRACTURE BEHAVIOUR OF
PLASMA SPRAYED THERMAL
BARRIER COATINGS
Jürgen Malzbender, Takashi Wakui, Egbert Wessel, and

Rolf W. Steinbrech*
1. INTRODUCTION
Thermal barrier coatings (TBCs) of plasma sprayed yttria stabilised zirconia (YSZ)
are increasingly utilised for heat exposed components of advanced gas turbines1,2. An
important reason for the application of zirconia coatings is the low thermal conduct-
ivity of this ceramic material which is further diminished in a TBC by the high
concentration of spraying induced microstructural defects, e.g. crack-shaped defects
between and within the spraying splats. Thus with TBCs on gas cooled turbine
components stiff temperature gradients can be realised as an important prerequisite
for an increased thermal efficiency of the energy conversion process.
However, the same spraying defects also provide sources of TBC fracture under
mechanical loading and cyclic thermal exposure during the operation of the turbines.
In principle, segmentation and delamination fracture of the TBC can be distin-
guished with cracks orientated perpendicular and parallel to the lamellar microstruc-
ture.
The vertical segmentation cracks improve the in-plane strain tolerance of the
TBC and are at present deliberately introduced in some of the advanced industrial
coatings. Contrary, delamination is highly unfavourable since large scale spallation
of the TBC would terminate the service life of the component. Accordingly, any
further improvement of the plasma sprayed TBCs with respect to easier segmentation
and more difficult delamination requires in-depth understanding of the fracture
*
Jürgen Malzbender, Takeshi Wakui, Egbert Wessel and Rolf W. Steinbrech, Institute for Materials in
Energy Systems (IWV2), Research Centre Juelich, 52428 Juelich, Germany
421
422 Jürgen Malzbender et al.
behaviour. In particular, experimental methods to quantify the respective fracture

resistances and to study the microstructural crack growth mechanisms are of interest.
From the experimental point of view the toughness (resistance) domain is still
minimally encompassed3 due to the difficulty to analyse quantitatively the fracture
behaviour of TBCs in the actual thickness (few hundred microns) applied for gas
turbine components. Indeed, many of the fracture mechanics approaches that have
been carried out in recent years focus on long crack fracture behaviour in thick
coatings.
In this respect the fracture experiments of Damani and Lutz4, who measured the
toughness of air plasma sprayed bulk alumina prepared from free-standing, thick-
walled tubes, may be taken as a baseline. Using CT and SENB specimens they
obtained for as sprayed material long crack toughness values of 1:4 MPam1=2
and 0:4 MPam1=2 in segmentation and delamination orientation, respectively.
Essentially flat R-curve behaviour of the as sprayed material changed into a sharply
rising R-curve after heat treatment at 15508C.
Toughness values of segmentation cracks in the same order of magnitude
1 MPam1=2 were also measured by Choi et al.5 in SEVNB tests after separation
of thick (6mm) plasma sprayed TBC (ZrO2 -8wt% Y2 O3 ). The toughness was the
same at room temperature and 8008C. Long crack toughness data referring to
attached YSZ coatings (0.87 mm thick) have been reported by Thurn et al.6. Bending
tests with the un-notched TBC in tension revealed pronounced R-curve behaviour
during controlled propagation of segmentation cracks. In as sprayed condition the
segmentation toughness increased from 0.2 to 1:3 MPam1=2 over 500 mm of crack
extension, equivalent to a crack resistance range between 10 N/m and 180 N/m. With
increasing annealing temperature the R-curve shifted towards higher toughness.
Considering the maximum tensile stress in the specimen the authors defined a
fracture toughness value of 1:1 MPam1=2 .
Mechanical delamination studies with long cracks are more difficult to carry
out, since they bear the problem of appropriate loading of the TBC. Typically the
TBC is sandwiched by gluing on the TBC surface a metallic extension bar for load
application.
Thurn et al.6 used such symmetric composite specimens with intermediate TBC
in the shape of bend bars. The delamination crack was propagated from a notch in
the TBC adjacent to the interface with the bond coat. The work of fracture as
averaging crack resistance was about 80 N/m, which is 30% lower than the respective
value of segmentation. Anderson7 obtained with specimens in DCB geometry a
critical energy release rate of 17 N/m ( 0:68 MPam1=2 ). Shaw et al.8 used a
chevron notched sandwich specimen and determined from the maximum load of
the delamination test a critical energy release rate of 16 N/m. Failure occurred in all
studies within the TBC (white failure). Much higher crack resistance values were
reported from delamination studies with a testing geometry after Charalambides et
al.9, which provides a value for the steady state resistance of long cracks. Oechsner10
determined crack resistance values of 100 and 200 N/m for an interface roughness RA
of 4.5 and 12 mm, respectively. For the smoother interface the fracture resistance
remained essentially unaffected by heat treatment, i.e. the thickness of the thermally
growing oxide scale (TGO) had no influence, whereas in the case of a rougher
interface a decrease was measured with increasing TGO thickness.
Fracture Behaviour of Plasma Sprayed Thermal Barrier Coatings 423
Efforts to measure the resistance of short cracks in delamination orientation are

limited and typically rely on indentation cracking. Bartlett and Dal Maschina11 used
a Vicker‘s tip with 10 N load to generate near interface cracking in the TBC.
Unaffected by heat treatment the measured toughness was in the range of
1:5---1:9 MPam1=2 equivalent to 15 25 N/m with the reported elastic modulus of
130 GPa. Using the same indentation method Rabiei and Evans12 determined a
short crack toughness of 0:7 MPam1=2 , equivalent to a crack resistance of 50 N/m,
if a low TBC stiffness of 10 GPa is assumed.
Although a comparison of the results from the different groups has definitively
limitations, since different processing parameters and microstructures of the tested
TBCs must be assumed, some general trends can be deduced. It appears that
segmentation cracks experience a higher material resistance than the cracks causing
delamination. The segmentation cracks also tend to exhibit pronounced R-curve
behaviour when growing from the initial defect size to the long crack configuration.
The delamination cracks propagated in all the reported mechanical tests within the
TBC and the crack path was typically observed to be close to the interface with the
bond coat. However, controversial results exist with respect to long crack resistance
data and the R-curve behaviour proposed for delamination by Evans et al.3 has not
really been demonstrated. Also the initial stages of delamination due to growth of the
existing microstructural TBC defects have not been analysed in detail.
It is the aim of the present work to contribute to the still limited experimental
body of fracture characterisation of plasma sprayed TBCs with controlled crack
propagation studies which cover the size scale between the initial microcrack length
and the final length of long cracks. The presented fracture results are restricted to as
sprayed TBC at room temperature with particular focus on two aspects: i) crack
length dependent fracture resistance is determined to further elaborate the R-curve
behaviour of segmentation cracks as well as to demonstrate the same effect for
delamination and ii) the observed toughening is attributed to the microstructural
features of crack path selection and crack surface interaction. Bending geometries are
used to propagate the cracks in thick stand-alone and thin attached TBCs. As
exception, the long delamination cracks were propagated by wedge opening loading
(WOL). Thus gluing of extension bars on the TBC surface could be avoided. In order
to achieve a correlation with microstructure the crack growth is monitored in-situ
with optical and scanning electron microscopy.
2. EXPERIMENTAL
2.1. Material
Zirconia stabilised with 8 wt.% zirconia was used as coating material for processing
of the TBC specimens. All atmospheric plasma spraying (APS) was carried out by the
processing group of the Institute for Materials and Processes in Energy Systems
(IWV1) at the Research Centre Juelich.
Due to the plasma spraying process the TBC comprises a characteristic
microstructure, which is conveniently recognised on through-thickness fracture sur-
faces. The SEM micrograph of Fig. 1 reveals the lamellar alignment of the spraying
splats. Within the individual splats columnar, sub-micron diameter grains are visible.
1 μm
Figure 1. Fracture surface of air plasma sprayed TBC with lamellar microstructure. The spraying splats
consist of sub micron grains. Crack shaped defects exist between and within the spraying splats (inter and
intra splat cracks).
Moreover, crack-shaped defects exist between and within the splats, hereafter termed
as inter- and intra-splat cracks, respectively.
2.2. Specimen Preparation

The geometrical configuration of the TBC in the thermal barrier system of a gas
turbine component was adapted by spraying thin TBC ( 300 mm) on larger plates of
Ni-super alloy substrate with a NiCoCrAlY bond coat.
The multilayer composite plate was machined into bend bars (see Table 1) using
a diamond saw for the ceramic TBC and spark erosion for the metal.
Thick TBCs (2 3 mm) were obtained by plasma-spraying the YSZ on a mild
steel substrate plate. The substrate was then dissolved using hydrochlorid acid and
the remaining TBC layer was cut into bend bars and small rectangular blocks for
wedge opening tests.
In order to facilitate in-situ microscopic observation of the controlled crack
growth, one side face of each specimen was mirror polished (1 mm final diamond
paste).
Some of the stand-alone thick TBC bend bars as well as all WOL specimens were
notched using the razor-blade method13 to obtain a sharp notch tip.
2.3. Testing Procedure

The specimens were tested in miniaturised mechanical testing equipment mounted on
the stages of optical and scanning electron microscopes to monitor the crack propa-
gation in-situ on a side face of the specimens. The 4-point bending tests of attached
TBC were carried out either with the TBC under tension or under compression.
Stand-alone TBCs were tested in 4-point (un-notched) and 3-point (notched) bending
(see Table 1).
Table 1. Testing and specimen geometries used for fracture characterisation of APS TBC.
TBC
Testing Geometry Specimen Size Thickness State
(1:5 3 45)mm3 300 mm attached / tension

Segmentation
(2 3 45)mm3 2 mm separated / tension
(2 3 45)mm3 2 mm separated / notched / tension
(1:5 3 45)mm3 300 mm attached / compression

Delamination
(2:4 3 4:5)mm3 2.4 mm separated / wedge opening
Growth of cracks in segmentation orientation was achieved by bending tests

with the TBC under tension. For delamination cracking two different methods were
applied. The first stage of growth related with the extension of the initial inter-splat
cracks was studied in bending tests with the TBC under compression. Long delami-
nation cracks were propagated in the thick stand-alone TBCs by loading the notched
specimens with a wedge (17.58 angle). Details of the testing method will be reported
elsewhere14.
In all cases (bending and WOL tests) controlled crack growth could be obtained.
The load was applied stepwise to take micrographs of the resulting crack extension.
After the fracture tests a frame by frame comparison of the micrographs was
performed to correlated crack length with the mechanically measured load and
deflection.
The sequences of in-situ taken micrographs were also used to evaluate the
displacement occurring in the vicinity of the propagating crack. The determination
of the strain was based on commercial software UNIDAC (universal deformation
analysis by means of correlation, Chemnitzer Werkstoffmechanik GmbH). The
software program permits a comparison of the displacement and strain after various
increments of crack growth based on the analysis of the image contrast15. The grey-
scale analyses where performed in the following sequence: i) digital image acquisition
ii) image analysis of the digitised grey-scale picture and iii) interpretation of the
displacement vector field.
2.4. Crack Resistance Determination

The variety of testing geometries for the attached and separated TBCs as well as the
distinction between micro-cracks and long cracks (Table 1) made different methods
of crack resistance determination necessary.
2.4.1. Stand-alone TBC
The crack resistance R was determined from the balance with the critical crack
driving force (GIC ) using a compliance approach
P2 dC
GIC ¼ ¼R (1)
2B da
where P is the applied load, B the specimen width and C the compliance which
depends on crack length a.
The compliance functions (dC/da) used in the present study for segmentation
crack growth in stand-alone TBCs referred to Moon et al.16 (4-point bending) and
Knehans et al.17 (notched, 3-point bending).
In the case of long cracks propagated in delamination orientation by the WOL
procedure Th. Fett18 provided the compliance function:
8 a3
C¼ : l(a) (2)
B H3 E
with
l(a) ¼ 99:01a5 þ 332:24a4 446:1a3 þ þ303:23a2 107:21a þ 18:06 (3)
where a is the normalised crack length with respect to specimen height.
2.4.2. Attached TBC
Resistance of the segmentation crack in the attached TBC of the composite specimen
was derived from the general relationship for stress intensity
p
K1 ¼ Y s0 pa (4)
where s0 is the stress, which the crack experiences. The crack-shape parameter Y
depends on the ratio of crack length to specimen thickness and the loading config-
uration.
In principle, a crack in the TBC leads to a different specimen deflection at
a particular ratio of crack length compared to isotropic material behaviour in a
bending test. However, the analysis simplifies for short cracks since then Y is
essentially independent of specimen thickness and loading configuration. In the
case of the thermal barrier coatings analysed in this investigation the ratio of crack
length to total specimen thickness was 0:2. An additional factor that has to be
taken into consideration is the change in stress intensity as the segmentation crack
approaches the metallic bond coat and substrate, which can be evaluated using finite
element simulation19. Taking the actual crack configuration in the attached TBC into
account, Y is replaced by Fs , which depends on the ratio of the stiffness values.
Estimating ETBC =Esubstrate 0:1 the parameter Fs can be approximated by19
Fs ¼ 1:3 0:779(a=t) (5)
where t is the specimen thickness. Application of equation (4) requires also know-
ledge of the average stress s0 acting on the crack:
ð
amax
s0 ¼ sa da=amax (6)
a¼0
The stress sa can be determined from the sum of the residual stress20 and the applied
stress due to bending moment, where in the latter case the formulas for the bending
of multi-layered composites have to be used21. Finally, using the Irwin relationship
GI ¼ KI2 =E the toughness was converted into crack resistance.
Assumptions had also to be made with respect to the fracture mechanics
treatment of the inter-splat cracks which showed extension upon compressive loading
of the TBC in bending. As will be shown in section 3.2, favourably orientated inter-
splat cracks were observed to grow under the applied compressive bending stress
similar to the ‘‘wing-cracks’’ reported and analysed by Ashby and Hallam22. The
concept requires the existence of an initial crack inclined with respect to the com-
pressive stress. The resulting shear forces cause mode I crack opening in the ‘‘wing’’
portions of the crack, which thereby grow parallel to the compressive stress field.
Statistical evidence for this crack growth mechanism in TBCs still needs to be
demonstrated, but strong support for the effect is also found in the studies of Trice
and Faber23, who tested hollow cylindrical specimens of free-standing plasma
sprayed alumina under compression. The crack resistance of ‘‘wing-cracks’’ can be
computed from the initial crack length 2a and the extension of the wing portion Dc
under the influence of a compressive stress s according to22

2 Dc 1
R¼ D0 þ p s2 pa (7)
Ep2 a 2
The parameter D0 reflects the initial level of damage, D0 ¼ pa2 NA , where NA is the
number of cracks per unit area, NA ¼ ð2h ðDc þ aÞ Þ 1 and h is the height of the
splats. Note that again the contribution of the residual stress has to be taken into
account.
The two modes of macro-cracking that occur in bending tests in case of large
specimen deformation with attached TBC are shown in Fig. 2. With the TBC
under tension almost equidistant segmentation cracks develop (Fig. 2a). The ratio
of crack distance to TBC thickness is about G 2:25 which is in fair agreement with
TBC under tension
1 mm
a)
TBC under compression
b) 1 mm
Figure 2. Bending tests with attached TBC in tension (a) and compression (b). Micrographs from fracture
experiments carried out in SEM with in situ side face observation.
the theoretical value of G theo 2:5 reported in literature for coatings and thin surface
films 24.
Under compression damage by delamination cracking is prominent (Fig. 2b).
Fracture processes initiate near the free TBC surface where the highest compressive
stress exists. Damage and buckling then progresses by cracking on parallel planes in
an uncontrolled manner through the TBC towards the interface with the bond coat.
The complexity of the failure sequence and the dependence of new delamination from
the previous fracture events does not allow a quantitative toughness characterisation.
The WOL tests with the growth of a single delamination crack in a stand-alone TBC
provide an experimentally better defined alternative.
3.1. Segmentation Cracking
Micrographs of segmentation cracks in the three tested bending/material variants are

shown in Fig. 3. In each case the segmentation crack exhibits a tortuous crack path.
The crack resistance in the attached TBC of the composite specimen increases
significantly with crack extension (Fig. 4). The apparent R-curve reaches values up to
300 N/m. First crack extension is observed at low TBC resistance ( 10 N=m), which
is indicative of the weak bonding within and between the spraying splats. The
apparent crack resistance reduces, if the compressive residual stress in the TBC is
taken into account and subtracted from the measured values. For the given layer
geometry and the involved materials of the composite specimen the TBC possesses a
residual compressive stress in the order of 50 MPa20 . The slope of the R-curve
decreases by almost a factor of three. Having made the residual stress correction,
the R-curve result of the attached TBC can be compared with those of the stand-
alone TBCs (Fig. 5). All testing variants reveal a rising R-curve which follows the
same master-curve. Moreover the results agree well with the R-curve reported by
Thurn et al.6 for the free-standing APS-TBC.
100 μm
100 μm 50 μm
Figure 3. Controlled propagation of segmentation cracks in bending tests with attached (left) and stand
alone TBC.
350
no residual stress considered
300
crack resistance R, N/m
250
200
150
with residual stress
100
50 composite optical
composite SEM
0
0 0,2 0,4 0,6 0,8 1 1,2
crack length a, mm
Figure 4. Influence of compressive residual stress on resistance behaviour of segmentation cracks in

attached TBC.
Fig. 5 also displays a distinct difference in the initial resistance values of the
cracks developing from the microstructural defects in the TBC and those starting
from an artificial notch tip. The higher initial crack resistance of the notched TBC
might be related to the formation of a non-uniform crack front. Crack front seg-
ments with local shifts in delamination plane due to microstructural heterogeneities
may generate a spatially dispersed crack front which causes a higher apparent
toughness.
400
350
crack resistance R, N/m

300
250
200
composite (optical)
150 composite (SEM)
free-standing (with notch)
100
free-standing (nonotch)
50 free-standing (no notch)
Thurn et al. (2000)
0
0 0,5 1 1,5
crack length a, mm
Figure 5. R curve results for segmentation cracks in attached and stand alone TBCs. For comparison R
curve of attached TBC corrected from the influence of residual stress.
3.2. Delamination Cracking

In the bending tests with the attached TBC under compression, special attention was
given to fracture events along the interface between bond coat and TBC. According
to results of FEM modelling3,25 the undulated interface between bond coat and TBC,
in combination with the thermal mismatch of both materials, should give rise to local
tensile residual stresses perpendicular to the interface. However, the in-situ observa-
tions revealed no preferred growth of cracks along the interface or in the TBC next to
the interface (Fig. 6). But limited growth of some inter-splat cracks could be observed
in the TBC before ‘‘macroscopic’’ delamination damage occurred. The growth of the
inter-splat cracks happened unpredictable with respect to location and was only
detected by post experimental inspection of the sequence of SEM micrographs. As
an example Fig. 7 shows in comparison with the initial load-free situation the
extension of an inter-splat crack after applying a (horizontal compressive force of
150 N.
bond coat
APS -TBC
20 μm
Figure 6. Attached TBC under compression in bending test. No preferential delamination was observed
near interface to the bond coat.
10 μm
0N
150 N
Figure 7. Magnified section of TBC area from Figure 6. Extension of inter splat crack by horizontal
compressive bending stress. Arrows indicate displacements analysed from grey scale analysis. Crack
opening behaviour similar to ‘‘wing crack’’ model22.
Considering all analysed examples the stable crack growth of the inter-splat
cracks was in the order of 5---10 mm before the experiment had to be terminated, since
the higher applied bending stress near the TBC surface caused the above described
macroscopic delamination damage (Fig. 3). Examination of the load triggered dis-
placements in the vicinity of the inter-splat crack in Fig. 7 reveals shear effects and
crack closure in the central (inclined) portion as well as crack opening and extension
in the ‘‘wing’’ portion.
The average directions of displacement obtained from the grey-scale analysis are
visualised by arrows in Fig. 7. Evaluation of the mode I crack resistance based on
equation (7) yields a steep rising crack resistance behaviour with values between 3 N/m
at the beginning and 26 N/m after the limited amount of crack extension at the end of
the experiment. The analysis includes the supporting impact of the residual compres-
sive stress. Interestingly the average crack resistance 13 N/m is only slightly lower
than the short crack results reported by Bartlett and Dal Maschio12 from indentation
tests.
Fig. 8 displays the inter-splat crack resistance together with the results of the
long delamination cracks propagated in the WOL specimens14. The crack length
dependent resolution of the long crack resistance shows unambiguously R-curve
behaviour. Starting from R0 100 N=m the R-curve saturates in a plateau resistance
which is about 50% higher. Compared to the resistance level determined for the inter-
splat cracks (average value 13 N/m) a fairly large gap in the data exists. The
higher resistance of the long delamination cracks probably reflects a more tortuous
crack path consisting of inter- and intra-splat sections. In addition the high starting
value of the long crack R-curve can be explained again by assuming a non-planar,
kinked crack front formation from the notch tip. Interestingly the average resistance
values of the WOL tests are in the same order of magnitude like the long crack
delamination data reported by Thurn et al.6 (RWOF 90 N=m) and Oechsner10
( 150 N=m < R < 200 N=m).
Summarising, the fracture toughness values for segmentation and delamination
cracks obtained with the as sprayed TBC in the present study two results have to be
300
250
fracture resistance, N/m
200 long cracks
150
100
Crack 1
Crack 2
50 Crack 3
inter-splat cracks WOL1
WOL2
0
0,0 200,0 400,0 600,0 800,0 1000,0 1200,0
crack extension Δc, mm
Figure 8. Crack resistance of inter splat cracks and R curve from extension of long delamination cracks
in WOL tests14.
emphasised: i) both crack types showed R-curve behaviour and ii) although the R-
curves might be different in slope and plateau value, there was no pronounced
difference in the initial and average crack resistance level.
3.3. Crack Path Selection and Toughening Mechanisms
Fig. 9 displays the side-face of a notched stand-alone TBC specimen before loading
and after growth of the segmentation crack. Obviously the macro-crack selects and
combines essentially the suitable segments from the pre-existing pattern of inter- and
intra-splat cracks and thus the segmentation crack path comprises of segments of
these micro-defects26,27.
The same holds for the delamination crack which is also composed of segments
of inter- and intra-splat cracks. From the surface observation and crack path studies
it appears that only little additional fracture at the crack tip is required to obtain the
macroscopic segmentation and delamination configuration. However, the sequence
of inter- and intra-splat cracks generates a tortuous crack surface with sharp deflec-
tions on the microstructural level. Upon crack opening such serrated crack surfaces
frequently develop contact sites and friction28.
Examples of microstructural crack deflection are shown in Fig. 10 for the
different crack orientations27. In Fig. 10a an intact spraying splat deflects a segmen-
tation crack, in Fig. 10b the delamination crack bounces between the two interfaces
with the adjacent splats, leaving a serrated fracture surface.
Implications of cracks surface contact are observed also on a larger scale in the
crack wake. The crack bridging yields formation of a new side path behind the actual
tip of a segmentation crack. Obviously the tortuous crack surfaces provide sufficient
contact bridging to shield the apparent crack tip, a toughening mechanism frequently
observed in ceramics with R-curve behaviour28,29.
50 μm
0N 2.5 N
Figure 9. Tortuous segmentation crack path. Macro crack follows pre existing pattern of intra and inter
splat micro cracks.
a b
10 μm 10 μm
Figure 10. Crack deflection mechanisms in TBC. a) Segmentation crack deflects on intact spraying splat.
b) Delamination crack bounces between surfaces of thin spray splat. Crack propagation from top to
bottom.
However, it should be emphasised that grey scale analysis in the vicinity of

segmentation cracks also revealed displacement activities caused by side cracks and
crack branching, depending on the heterogeneities in the lamellar TBC microstruc-
ture.
4. CONCLUSIONS
The controlled crack growth studies of segmentation and delamination cracks in

plasma sprayed TBCs revealed an increase of crack resistance with crack extension.
This R-curve behaviour could be measured in the case of segmentation from short to
long crack configuration using un-notched, stand-alone TBCs. The observed per-
manent increase in crack resistance was attributed to the limited TBC thickness in
spraying direction. Propagation of long cracks in delamination orientation showed
also R-curve behaviour with plateau values in the order of 150 N/m.
The macro-crack path followed in both orientations the pre-existing pattern of

inter- and intra-splat cracks. The sharp deflections in the microcrack mode generate a
tortuous crack path which gives rise to crack surface interaction and thereby provides
a source of toughening. However, depending on the microstructural heterogeneities
of the APS-TBC and local weak adherence of spraying splats also crack branching
and side crack activities in a process zone around the macro-crack were observed.
Thus it seems premature to further quantify the measured R-curve behaviour
and to describe the toughening effect in a model. Yet an important conclusion can be
drawn from the present study. Tailoring plasma sprayed TBCs with low crack
resistance for segmentation cracks and high delamination resistance should focus
on a reduction of the density of inter-splat microcracks. Variation of the spraying
parameters has the potential to systematically change APS-TBCs with respect to the
ratio of inter- and intra-splat defects in the lamellar microstructure.
5. ACKNOWLEDGEMENTS
The authors would like to thank R. Vassen for providing the APS-TBC coatings,
J. Mönch for experimental support and B. Mukherjee for contributing WOL-results.
The encouraging and stimulating support of L. Singheiser throughout the work is
gratefully acknowledged.
6. REFERENCES
1. J.T. DeMasi Marcin, D.K. Gupta, Protective coatings in the gas turbine engine, Surface and
Coating Technology 68/69 (12) 1 9(1994)
2. N.P. Padture, M. Gell, E.H. Jordan, ‘‘Thermal barrier coatings for gas turbine engine applications’’,
Science, Vol. 296, 280 84 (2002)
3. A.G. Evans, D.R. Mumm, J.W. Hutchinson, G.H. Meier, F.S. Petit, Mechanisms controlling the
durability of thermal barrier coatings, Progress in Materials Science 46 (2001) 505 535
4. R.J. Damani, E.H. Lutz, Microstructure, strength and fracture characteristics of a free standing
plasma sprayed alumina, Journal European Ceramic Society 17 (1997) 1351 1359
5. S.R. Choi, D. Zhu, R.A. Miller, High temperature slow crack growth, fracture toughness and room
temperature deformation behaviour of plasma sprayed ZrO2 8wt%Y2 O3 , Ceram. Eng. Sci. Proc.,
Vol. 19, 1998, 293 301
6. G. Thurn, G.A. Schneider, H.A. Bahr, F. Aldinger, Toughness anisotropy and damage behaviour of
plasma sprayed ZrO2 thermal barrier coatings, Surface and Coatings Technology 123 (2000) 147 58.
7. C.A. Anderson, Thermal stress fracture of ceramic coatings, in Fracture Mechanics of Ceramics, Ed.
R.C. Bradt, D.P.H. Hasselman, F.F. Lange, A.G. Evans, Vol. 6, Plenum Press. New York, 1983,
497 510
8. L.L. Shaw, B. Barber, E.H. Jordan, M. Gell, Measurement of the interfacial fracture energy of
thermal barrier coatings, Scripta Materialia, Vol. 39, No. 10, pp. 1427 1434, 1998
9. P.G. Charalambides, J. Lund, A.G. Evans, R.M. McMeeking, A test specimen for determining the
fracture resistance of bimaterial interfaces, J. Appl. Mechanics, 56 (1989) 77 82
10. M. Oechsner, A contribution to life prediction of ceramic thermal barrier coatings, (Thesis, in
German), Fortschritt Berichte VDI Verlag, Duesseldorf, Series 18, No. 263, 2001
11. A.H. Bartlett, R. Dal Maschio, Failure mechanisms of zirconia 8wt% yttria thermal barrier coatings,
J. Am. Ceram. Soc., 78 [4] 1018 24, 1995
12. A. Rabiei, A.G. Evans, Failure mechanisms associated with the thermally grown oxide in plasma
sprayed thermal barrier coatings, Acta mater. 48 (2000) 3963 3976
13. J. Kübler, Fracture toughness of ceramics using the SEVNB Method: Preliminary results, Ceram.
Eng. Sci. Proc., 18, 1997, 155 162
14. B. Mukherjee, R.W. Steinbrech, Fracture behaviour of long delamination cracks in thermal barrier
coatings, to be published
15. E. Kieselstein, B. Seiler, M. Dost, E. Than, Characterization of composites by measuring the optical
displacement via UNIDAC, (in German) Technisches Messen 69 [10], 412 (2002)
16. R. Moon, K. Bowman, K. Trumble, J. Roedel, Comparison of R curves from single edge V notch
beam (SEVB) and surface crack in flexure(SCF) fracture toughness test methods on multilayered
alumina zirconia composites, J. Am. Ceram. Soc. 83 [2] 445 47 (2000)
17. F. Deuerler, R. Knehans, R.W. Steinbrech, Testing methods of R curve behaviour of Al2 O3 , Journal
de Physique, 1986, CI 617 621
18. Th. Fett, private communication
19. Th. Fett, D. Munz, Stress intensity factors and weight functions, Computational Mechanics Publ.,
Billerica, Massachusetts, 1997
20. G. Blandin and R.W. Steinbrech, ‘‘Thermomechanical behaviour of plasma sprayed multilayer
composites for thermal barriers’’ (in German), Juel Report 3954, Forschungszentrum Juelich, Ger
many (2002)
21. J. Malzbender, R.W. Steinbrech, Mechanical methods to determine the compliances within multi
layered composites, J. Mater. Res. 18 [6] (2003)
22. M.F. Ashby, S.D. Hallam, The failure of brittle solids containing small cracks under compressive
stress state, Acta metall. 34 (1986) 497 510
23. R.W. Trice, D.W. Prine, K.T. Faber, Deformation mechanisms in compression loaded stand alone
plasma sprayed alumina coatings, J. Am. Ceram. Soc., 83 [12] 3057 64 (2000)
24. S.R. Kim, J.A. Nairn, Fracture mechanics analysis of coating/substrate systems. Part I: Analysis of
tensile and bending tests, Eng. Fracture Mechanics, 65 (2000) 573 579
25. M. Ahrens, R. Vassen, D. Stöver, Stress distributions in plasma sprayed thermal barrier coatings as a
function of interface roughness and oxide scale thickness, Surface and Coatings Technology 161
(2002) 26 35
26. R.W. Steinbrech, ‘‘Thermomechanical behaviour of plasma sprayed thermal barrier coatings’’,
Ceramic Eng. Sci. Proc. 23 (2002) 379
27. E. Wessel, R.W. Steinbrech, Crack propagation in plasma sprayed ceramic, Key Engineering Ma
terials, Vol. 223 (2002) 55 60
28. R.W. Steinbrech, Toughening mechanisms for ceramic materials, J. Europ. Ceram. Soc. 10 (1992)
131 142
29. A.G. Evans, Perspective of the development of high toughness ceramics, J. Am. Ceram. Soc. 73 (990)
187 206
FRACTURE OF CONDUCTIVE
CRACKS IN POLED AND
DEPOLED PZT-4 CERAMICS
Tong-Yi Zhang*
ABSTRACT
The present work reports experimental results on the fracture behavior of conductive
cracks or deep conductive notches in poled and depoled PZT-4 ceramics. The
fracture tests were conducted with compact tension specimens under purely mech-
anical loading, purely electrical loading and combined mechanical and electrical
loading. Finite element calculations were carried out to obtain the energy release
rate, the stress intensity factor and the intensity factor of electric field strength of the
specimens. The results show that the critical energy release rate under purely either
electrical or mechanical load is a constant, independent of the ligament length of the
specimens. However, for both the poled and depoled ceramics, the electrical fracture
toughness in terms of the critical energy release rate is much higher than the
mechanical fracture toughness. The highly electrical fracture toughness arises from
the greater energy dissipation around the conductive sharp notch tip under purely
electrical loading, which is impossible under mechanical loading in the brittle ceram-
ics. Furthermore, the critical energy release rate under combined electrical and
mechanical loading depends on the weight of the electrical load in comparison
with the mechanical load. We normalize the critical stress intensity factor by the criti-
cal stress intensity factor under purely mechanical loading and normalize the critical
intensity factor of electric field strength by the critical intensity factor of electric field
strength under purely electrical loading. Then, a quadratic function describes the
relationship between the normalized critical stress intensity factor and the normal-
*
Department of Mechanical Engineering, Hong Kong University of Science and Technology, Clear
Water Bay, Kowloon, Hong Kong, China, E mail: [email protected]
609
610 Tong-Yi Zhang
ized critical intensity factor of electric field strength. This quadratic function may
serve as a failure criterion for conductive cracks in piezoelectric ceramics.
KEYWORDS
fracture, piezoelectric and dielectric ceramics, conducting cracks, dielectric discharge
1. INTRODUCTION
Ferroelectric ceramics have become preferred materials for a wide variety of elec-
tronic and mechatronic devices due to their pronounced dielectric, piezoelectric, and
pyroelectric properties. Aging, fatigue and electrical and/or mechanical breakdown
of the materials cause device failures, which are considered seriously in the design of
ferroelectric devices. Ferroelectric ceramics are brittle and susceptible to cracking at
all scales from domains to devices. Because of the importance of electrical and
mechanical reliability of these devices, there has been tremendous interest in studying
the fracture behavior of such materials. Zhang et al. (2002) provide an overview on
fracture of piezoelectric ceramics with summarizing current knowledge of the frac-
ture of piezoelectric ceramics. Their attention is confined to fracture mechanics
studies, yet experimental results are also examined for comparison with theoretical
predictions. Zhang (2002) reviewed the experimental observations on the effects of
static electric field on the fracture behavior of piezoelectric ceramics.
Internal electrodes have widely been adopted in electronic and electromechan-
ical devices made of ferroelectric ceramics. These embedded electrodes may naturally
function as pre-conductive cracks or notches, which may lead to failure of these
devices under electric and/or mechanical loads. Figure 1 schematically shows the
similarity between a conductive crack under electrical loading and a normal crack
under mechanical loading. To ensure that the electric field inside the conductive
crack remains zero, electric charges in the conductive crack surfaces must rearrange
themselves to produce an induced field that has the same magnitude as the applied
one but with the opposite sign. As a result, the charges in the upper and lower crack
surfaces near the crack tip have the same sign, as shown in Fig. 1. The charges with
the same sign repel each other and then have a tendency to propagate the crack. The
contour-independent J-integral used in fracture mechanics can also apply to con-
ductive cracks (Zhang et al., 1998; Garboczi, 1988; McMeeking, 1987) and the J-
2a 2a
Conductive crack
JM = CM σ2a J E = CEE 2a
Figure 1. A comparison of a normal crack loaded by uniform mechanical stress, s, and a conductive
crack loaded by uniform electrical field, E, where J M and J E denote, repectively, the mechanical and
electrical J integrals, and CM = p / Y * and CE pk=2.
integral results for the purely mechanical and electrical loadings are illustrated in
Fig. 1 correspondingly. It is therefore of practical importance and academic signifi-
cance to study the fracture mechanics and failure behavior of conductive cracks in
ferroelectric ceramics.
Lynch et al. (1995) carried out indentation fracture tests on electroded surfaces
submerged in electrically conducting NaCl solution and in distilled water. In both
cases, tree-like damage grew from the indented electrode under the cyclic electric
field. Heyer et al. (1998) studied the electromechanical fracture toughness of con-
ductive cracks in PZT-PIC ceramics. They conducted four-point bending tests on
pre-notched bars, in which the poling direction was toward the jig surface and the
notch was filled with NaCl solution to make the crack conducting. Wide scattering
results were obtained under a large applied electric field of jKE j > 50kV =m1=2 ,
where KE is the applied electric intensity factor. The critical stress intensity factor
increased as the applied intensity factor of the electric field strength changed from
30kV =m1=2 to 90kV =m1=2 . When the applied electric intensity factor was relatively
small, within the range of 15kV =m1=2 to 15kV =m1=2 , they could explain the
experimental data using a domain-switching-based model. Using compact tension
samples with conductive notches, Fu, Qian and Zhang (2000) have performed
extensive fracture tests on lead zirconate titanate (PZT-4) ceramics under purely
electrical or mechanical loading. The experimental results indicate that both the
purely electric and mechanical fields can propagate conductive cracks (notches) and
fracture the samples. Under purely electric loading, there exists a critical energy
release rate at fracture, which is named the electric fracture toughness. The electric
fracture toughness is much larger than the mechanical fracture toughness, which
is the critical energy release rate at fracture under purely mechanical loading. Like
the mechanical fracture toughness, the electric fracture toughness is a material
property, which is defined as the resistance of a material against fracture or
as the energy per unit area absorbed by the material as the crack propagates. Fu
et al. (2000) attributed the high electric fracture toughness to electrical plastic
deformation.
To clarify whether the failure of conductive cracks in poled PZT ceramics
was induced by the piezoelectric effect, Wang and Zhang (2001) thermally depoled
the ceramics and demonstrated that electric fields could fracture depoled lead
zirconate titanate (PZT-4) ceramics. Thus, the concepts of fracture mechanics can
be applied to the failure of conductive cracks in macroscopically dielectric ceramics
and the electrical facture toughness exists for dielectric ceramics. Like the case for
poled PZT-4 ceramics, the electrical fracture toughness for the depoled PZT-4
ceramics is also much higher than the mechanical fracture toughness of the same
depoled PZT-4 ceramics. The highly electrical fracture toughness arises from the
greater energy dissipation around the conductive crack tip under purely electric
loading, which is impossible under mechanical loading in the brittle depoled
ceramics.
The present work summarizes the experimental results of the failure behavior of
electrically conductive cracks in depoled (Wang and Zhang, 2001) and poled (Fu
et al., 2000) PZT ceramics under purely electrical or mechanical loading. Further-
more, the failure behavior of conductive cracks in depoled PZT ceramics under
combined electrical and mechanical loading and the charge-free zone model (Zhang
et al., 2003) are also briefly introduced here.
612 Tong-Yi Zhang
The material used in this study was lead zirconate titanate ceramics (PZT-4, Morgan
Matroc). Pre-notched or cracked compact tension (CT) samples were adopted in the
fracture tests under purely mechanical, purely electric and mixed mechanical and
electric loads. All samples had width 2w1 ¼ 10mm, height w0 ¼ 10mm and thickness
b ¼ 3mm. A pre-notch or crack was cut in each sample with a 0.15mm thick diamond
saw parallel to the poling direction for the poled ceramics, and further sharpened by
a 0.12 mm diameter wire saw. After the cutting, the samples were cleaned ultrason-
ically in distilled water for a few minutes. For conductive cracks, silver paint was
filled into the notch (crack) to make it function as an electrode. The silver paint is a
viscous solution and solidifies at room temperature quickly. The refilling technique,
which meant that filling was conducted again after the preceding filling and solidi-
fication, was adopted to ensure the notch was fully filled. The refill was continuing
until that no obvious shrinkage was observed by naked eyes. In this case, the
electrode in the sample was perfect, which was proofed by investigating the fracture
surface. Figure 2 schematically shows the sample geometry and the loading condi-
tion. Regarding electric loading, we applied a DC voltage and manually increased its
magnitude until the sample failed. For poled PZT ceramics, positive and negative
electric fields, which were respectively parallel and anti-parallel to the poling direc-
tion, were applied in the tests.
The fracture tests were carried out under a constant voltage mode or a constant
mechanical load mode with a homemade loading apparatus. In the constant voltage
mode, a constant voltage was applied first, then, the mechanical load was gradually
increased until the sample fractured, while in the constant mechanical load mode, a
constant mechanical load was applied first, following by a gradual increasing of the
applied electric voltage until the sample fractured. Positive and negative electric fields
were applied in both loading modes. To avoid electric sparking, the samples were put
into the silicone oil during tests under purely electric loading or covered by a thick
layer of silicone grease under combined loading. All tests were implemented at room
temperature. At least ten samples were repeatedly tested for each loading condition
under combined mechanical and electrical loading and 30 samples were tested under
the purely electrical loading.
2w1
conductive l
V
w0 w
P P
Figure 2. Sample geometry and loading condition.

The depoling was achieved by thermally annealing the PZT-4 ceramics at 400
8C for 40 minutes and then cooling them down in the furnace to room temperature
(Wang and Zhang, 2001). After depoling, the ceramics lost piezoelectricity and
became macroscopic dielectrics that were isotropic in elasticity and dielectricity.
No detectable piezoelectric constants in the depoled ceramics were measured with
the piezo d33 tester with 1% accuracy (Pennebaker, Model 8000). The isotropic
dielectric constants were measured using an impedance analyzer (Hewlett Packard,
4192A Lf) at the standard frequency of 1M Hz. The mean value of the dielectric
constants with the standard deviation was k ¼ 9:794 0:205 10 9 C=Vm. The
elastic constant and Poisson ratio were measured by compressive tests using strain
gauges on a universal testing machine (MTS Sintech 10/D). The means of the elastic
constant and the Poisson ratio were, respectively, Y ¼ 25:3 0:34 GPa and
n ¼ 0:38 0:006 for the depoled ceramics.* Pre-notched (or pre-cracked) compact
tension samples were adopted for fracture tests under purely mechanical and purely
electrical loads. The depoled samples were prepared in the same way as that for
the poled sample such that the sample geometry and the loading condition for the
depoled samples were also the same as that for the poled samples, as shown in
Fig. 2.
3. ENERGY RELEASE RATE AND INTENSITY FACTORS
Finite element calculations were conducted to have the energy release rate for the
used sample geometry and loading condition. The energy release rate is equivalent to
the contour-independent J-integral in the linear electro-mechanics (Cherepanov,
1979; Zhang et.al., 2002), which is given by
ð
G ¼ J ¼ (hn1 sij nj ui,1 þ Di E1 ni )dG, i, j ¼ 1, 2, 3, (1)
G
where h ¼ sij «ij =2 Di Ei =2 is the electric enthalpy per unit volume, G is an integra-
tion contour around the crack tip, and n is the unit out normal vector to the contour.
In the finite element calculation of J-integral, we used the commercial software
ABAQUS and the eight-node plane strain piezoelectric elements.
For the poled PZT ceramics, the following material constants were used in the
calculations.
Elastic compliance(10 12
m2 =N):
s11 ¼ 12:37, s12 ¼ 3:97, s13 ¼ 5:52, s33 ¼ 16:11, s44 ¼ 39:06;
10
Piezoelectric constants(10 m=V):
d31 ¼ 1:35, d33 ¼ 3:00, d15 ¼ 5:25;
Dielectric constants(10 8 C=Vm):
K11 ¼ 1:31, K33 ¼ 1:15;
*
The late test results showed that the average Young’s modulus and the average Poisson ratio were
Y 51 GPa and n 0:43, respectively, for the depoled ceramics (Zhang et al., 2003). The discrepancy
might be caused by the mechanically nonlinear behavior of the depoled ceramics.
614 Tong-Yi Zhang
where the subscript ‘‘3’’ denotes the poling direction and N, m, V, and C represent,
respectively, Newton, Meter, Volt, and Coulomb. For the depoled ceramics, the
material constants used in the calculations were mentioned above. The conductive
crack is modeled by using a large dielectric constant (109 times of K33 ) of the medium
inside the crack. The FEA results provide two equations to calculate the energy
release rate, GM E
I or GI , under the purely mechanical or electrical loading for the
samples with ligaments ranging from 2.5 mm to 4.5 mm:
" 2 3 4 #
P2 l l l l
GI ¼ 2
M 4:28735:944 þ118:629 178:729 þ102:704 103 ,
b w1 Y11 w w w w
(2)
" 2 3 4 #
2
K33 V l l l l
GEI ¼ 10:27863:719 þ185:213 266:316 þ151:755 ,
w w w w w
(3)
where Y11 ¼1=s11 , b is the sample thickness and, G, P, and V are, respectively, in units
of N/m, N and kV, and all geometry parameters are in meters.
We calculated the intensity factors from the J-integral by assuming that the
relationship between the intensity factors and the J-integral for the used CT samples
was the same as that in an infinite body (Wang, 2003). In this case, the intensity
factors are expressed in terms of
P a V a
KIs ¼ p f1 , K E ¼ p f2 : (4)
b w w w w
We regressed seven sets of calculated results corresponding to the different ligaments

and then had the geometry functions f1 and f2 in the following forms,
f1 (t) ¼ 5:608 þ 66:5151t 43:373t2 139:197t3 þ 226:173t4 ,

(5)
f2 (t) ¼ 10:500 81:457t þ 254:772t2 338:301t3 þ 167:191t4 ,
where 0:33#t#0:72, and t ¼ a=w, a ¼ w l, w ¼ 9 10 3 m, as shown in Fig. 2.
For the depoled PZT-4 dielectric samples with ligament, l, ranging from 3.2 to
5.2 mm, we had the following energy release rates:
P2
I ¼
GM 38:664 274:69s þ 757:44s2 940:75s3 þ 439:32s4 102 (6)
b2 w1 Y
for purely mechanical loading, and
kV 2
GE ¼ 26:231 213:07s þ 681:45s2 962:53s3 þ 499:49s4 (7)
w
for purely electrical loading, where s ¼ l=w, Y ¼ Y =(1 n2 ), and k is the dielectric
constant of the material.
4. RESULTS
Figure 3(a) shows the mechanical fracture toughness of the depoled ceramics as a
function of the ligament length, where a solid circle represents an experimental
datum (hereafter, the same symbol is used in the remaining figures without notation).
The mechanical fracture toughness has a mean value of GM IC ¼ 30:4 3:9 N=m in
terms of the critical energy release rate, which is indicated by the solid line in Fig. 3(a).
Clearly, the mechanical fracture toughness is a material property that can serve as a
mechanical fracture criterion. To illustrate the mechanical fracture behavior statis-
tically, we plot the failure probability in Fig. 3(b), which follows the Weibull distri-
bution of
" 8:69 #
GM
F (GIC ) ¼ 1 exp
M IC
and GM
IC is in units of N=m, (8)
32:1
where F is the cumulative failure distribution function. Equation (8) indicates that
the depoled PZT-4 ceramics under purely mechanical loading has a Weibull modulus
of 8.69, which is more or less the same as the modulus in engineering ceramics. Using

IC ¼ (KIC ) =Y , the mechanical fracture toughness can be
M 2
the relationship of GM
expressed in terms ofp the critical stress intensity factor and it is
M
KIC ¼ 0:947 0:063 MPa m under purely mechanical loading.
50
40
GICM(N/m)
30
20
(a) Mechanical loading
Exp. data
10 Mean
0
2.8 3.0 3.2 3.4 3.6 3.8 4.0 4.2 4.4
Ligament(mm)
Figure 3(a). The mechanical fracture toughness vs. the ligament.
99
90
Probility of failure (%)
80
60
(b) Mechanical load ng
40
30
20
10
5
2
16 20 24 28 32 36 40
GICM(N/m)
Figure 3(b). Weibull distribution of the mechanical fracture toughness.

616 Tong-Yi Zhang
Figure 4(a) shows the electrical fracture toughness as a function of the ligament
length with the solid line representing the mean value. The mean of the critical energy
release rate is GEC ¼ 263 35 N=m. The experimental results clearly indicate that the
depoled ceramics have an electrical fracture toughness, which is also a material
property and can serve as a failure criterion for conductive cracks in dielectrics
under purely electrical loading. Similarly, we plot the failure probability of the
depoled ceramics under purely electrical loading in Fig. 4(b), which follows the
Weibull distribution,
" 3:35 #
GEC 150
F (GC ) ¼ 1 exp
E
and GEC is in units of N=m: (9)
125:6
It is interesting to note that the depoled ceramics has a Weibull modulus of 3.35
under purely electrical loading. The lower the Weibull modulus is, the greater the
variability of the fracture toughness. The results indicate that the experimental data
for electric fracture toughness are more scattered than those for mechanical fracture
toughness. Using the relationship of GEC ¼ k(KCE )2 =2, the electrical fracture toughness
400
300
GCE(N/m)
200
(a) Electrical loading

100 Exp. data
Mean
0
2.5 3.0 3.5 4.0 4.5 5.0 5.5
Ligament(mm)
Figure 4(a). The electrical fracture toughness vs. the ligament.
99
90
80
Probility of failure(%)
60
40
30 (b) Electrical loading
20
10
5
2
1
30 60 90 120 150 180 21
GCE-G0(N/m)
Figure 4(b). Weibull distribution of the electrical fracture toughness.

in terms of the critical pelectric field intensity factor is given by

KCE K E0 ¼ 0:23 0:022MV = m under purely electrical loading.
The electrical fracture toughness is about 9 times higher than the mechanical
fracture toughness because electrical plastic deformation, including electrical dis-
charge and domain switching, occurs at the tip of the conductive crack, forming an
electrical plastic zone. Actually, dielectric breakdown accompanies fracture under
purely electrical loading. Figure 5 shows fractographs of the samples fractured under
the purely mechanical or purely electrical loading. The fracture surfaces are flat for
samples fractured under the mechanical loading, while the electrical loading yields
rough fracture surfaces, as shown in Fig. 5. In the electrically fractured samples, any
discharge may locally burn the sample, as indicated by the dark area on the right
fractograph of Fig. 5. At high magnification, clear grain morphology is revealed on
the mechanical fracture surface, whereas some regions of the electrical fracture
surface appear to be melted and resolidified. It is the electrical plastic deformation
that consumes more energy and thus leads to the high electrical fracture toughness.
Fu et al. (2000) and Wang et al. (2001) performed fracture tests on conductive
cracks in poled PZT-4 piezoelectric ceramics and the poling direction was parallel to
the pre-notch. The compact tension samples were prepared in the same way as that
used by Wang and Zhang (2001), as described above. The following three types of
tests were conducted: 1) purely mechanical loading, 2) purely electric loading, and 3)
combined mechanical and electrical loading.
Figure 6(a) shows the critical energy release rate versus the sample ligament
under purely mechanical loading. It is clearly seen that the linear regression of the
plot is very approximate to the horizontal line, indicating that the critical energy
release rate is a material constant independent of the sample ligament. The mechan-
ical fracture toughness can also be expressed in p terms of the critical stress intensity
factor, which is KICM
K s0 ¼ 0:934 0:06 MPa m under purely mechanical load-
ing. Figure 6(b) shows that the failure probability of the PZT-4 ceramics under
purely mechanical loading follows the Weibull distribution
10μm 10μm
Point A Point B
Mechanical Electrical
A B
1mm 1mm
Figure 5. Lower: optical images; Upper: SEM images from the corresponding regions of A and B
in the lower optical images.
618 Tong-Yi Zhang
20
Critical energy release rate

Regr.
15 Mean
(N/m)
10
0
2.0 2.5 3.0 3.5 4.0 4.5
Ligament (mm)
Figure 6(a). The critical energy release rate versus the ligament under purely mechanical loading.
99
Probability of failure(%)
90
80
60
40
30
20
10
5
2
6.0 7.0 8.0 9.0 10.0 11.0 12.0
GICM
Figure 6(b). Weibull distribution of the critical energy release rate under purely mechanical loading.
" 9:6435 #
GM
F (GM
IC ) ¼ 1 exp IC
with GM
IC in units of N=m: (10)
9:0103
Under purely mechanical loading, the PZT-4 ceramics has a Weibull modulus of
9.6435, more or less the same as other engineering ceramics.
In the same way, Fig. 7(a) illustrates the critical energy release rate versus the
sample ligament under purely electrical loading. Again, the linear regression of the
plot is almost a horizontal line. This feature, in analogy with the mechanical case,
indicates the existence of the electric fracture toughness under purely electric loading.
The electric fracture toughness is a material property and has the value
GEC ¼ 223:7 17:0N=m in terms of the critical energy release rate, which is about
25 times larger than GM IC ¼ 8:7 0:4N=m. This is because electrical discharge and
domain switching may occur at the tip of the conductive crack, forming an electric-
ally plastic zone. Actually, electric discharge was observed during the tests. This kind
of electrical plastic deformation accompanies crack propagation and consumes
energy, thus leading to the high electrical fracture toughness. The failure probability
of the PZT-4 ceramics under purely electric loading, as shown in Fig. 7(b), follows
the Weibull distribution of
" 2:331 #
GEC 130
F (GC ) ¼ 1 exp
E
with GEC in units of N=m: (11)
110:25
500
Critical energy release rate

Regr.
400 Mean
300
(N/m)
200
100
0
2.0 2.5 3.0 3.5 4.0 4.5
Ligament (mm)
Figure 7(a). The critical energy release rate versus the ligament under purely electrical loading.
99
Probability of failure(%)
90
80
60
40
30
20
10
5
2
15 40 70 100 160 190 250
GCE-G0(N/m)
Figure 7(b). Weibull distribution of the critical energy release rate under purely electrical loading.
Comparing Eq. (11) with Eq. (10) indicates that a three parameter Weibull distribu-
tion has to be used to describe the failure probability under purely electrical loading,
while a two parameter Weibull distribution is sufficient to describe the failure
probability under purely mechanical loading. Equation (11) is valid only when GEC
is larger than 130 N/m. When GEC is smaller than 130 N/m, no failure will occur under
purely electrical loading. The PZT-4 ceramics under purely electric loading has a
Weibull modulus of 2.331, as shown in Eq. (11), which is much smaller than that
under purely mechanical loading, indicating again a larger variability of the electric
fracture toughness.
Under combined electrical and mechanical loading, we normalize the critical
stress intensity factor by the critical stress intensity factor under purely mechanical
loading and normalize the critical intensity factor of electric field strength by the
critical intensity factor of electric field strength under purely electric loading. Figure 8
shows the relationships of the normalized intensity factor of electric field strength
versus the normalized stress intensity factor. The experimental data can be approxi-
mately described by the equation
E E 2 s M 2
K =KC þ K =KIC ¼ 1, (12)
where KCE denotes the critical intensity factor of electrical field strength under purely
electrical loading. Equation (12), shown as the circular curve in Fig. 8, may serve as a
620 Tong-Yi Zhang
1.5
1.0
KE/KCE
0.5
0.0
0.0 0.5 1.0 1.5
Kσ/KICM
Figure 8. Experimental results for the failure of electrically conductive cracks in the depoled PZT 4
ceramics under combined mechanical and electrical loading.
failure criterion for the depoled PZT ceramics under combined electric and mechan-
ical loading. Based on the experimental results, Zhang et al. (2003) developed a
Charge-Free Zone (CFZ) model, which successfully explains the experimental obser-
vations.
5. THE CHARGE-FREE ZONE MODEL
The charge-free zone (CFZ) model (Zhang et al. 2003) was proposed to understand
the failure behavior of conductive cracks in dielectric ceramics under electrical and/or
mechanical loading. The CFZ model treats dielectric ceramics mechanically brittle
and electrically ductile. Charge emission and charge trapping were assumed to occur
in the CFZ model, which consume more work and thus lead to a high value of the
electric toughness. In the CFZ model, the local electric intensity factor has a non-zero
value and consequently there is a non-zero local electric energy release rate, which
contributes to the driving force to propagate the conductive crack. The merit of the
CFZ model lies in the ability to apply the Griffith criterion directly to link the local
energy release rate to the fracture toughness in a completely brittle manner. As a
result, an explicit failure criterion results from the CFZ model to predict the failure
behavior of conductive cracks in dielectric ceramics under electrical and/or mechan-
ical loading and the theoretical predictions agree perfectly with the experimental
observations.
To simplify the analysis, charges were treated as line charges and the trapped
charge cloud in front of a conductive crack was modeled as a charge strip in the CFZ
model, as shown in Fig. 9, where ob denotes the CFZ size and the region of ba is the
charge strip. The trapped charges shield the conductive crack tip from the applied
electric field, while the charge-free zone reflects that the shielding is only partial.
Therefore, the local value of the intensity factor of electric field intensity, which is
Ec
Conductive
crack
0 b a x
Figure 9. The field distribution in front of a conductive crack, wherein ob is the size of the charge free
zone and ba denotes the charge zone.
called the electric intensity factor for convenience, has a nonzero value. Based on the
CFZ model, Zhang et al. (2003) derived the local electric intensity factor, KE(l) , which
was directly proportional to the applied electric intensity factor, KE(a) , KE(l) ¼ VKE(a) .
For dielectric ceramics there is no interaction between the electrical and mechanical
fields. Thus, the local energy release rate for a dielectric ceramic is expressed in terms
of the intensity factors as
2 k 2
Gl ¼ Ksa =Y þ KEl : (13)
2
Under purely mechanical loading or purely electrical loading, an application of the
Griffith criterion to Eq. (13) yields

M 2

KIC
¼
GlC ¼ G, (14)
Y
2 2
k KEl k KCE V2
GlC ¼ ¼ ¼ G, (15)
2 2
where G represents the material resistance to failure. Eq. (14) indicates that the value
of G is identical to the mechanical toughness, GM IC . Then, the value of V was
calculated from Eq. (15) with the value of KCE . Under combined loading, applying
the Griffith criterion to Eq. (13) yields
2 k l 2
GlC ¼ Ksa,C =Y þ KE ,C ¼ G (16)
2
The subscript ‘‘C’’ is labeled to indicate the fracture critical condition. Then, using
Eqs. (14) and (15) leads to the failure criterion, Eq. (12), which has been observed
experimentally.
Very recently, Zhang et al., (2004a) further developed the CFZ model for poled
ceramics and verified the model experimentally (Zhang et al., 2004b). Although the
coupling between the electrical and mechanical fields makes the mathematic deriv-
ation complex, the physical picture remains the same as that described above.
Interested readers may refer the original publications (Zhang et al., 2004a and
2004b).
622 Tong-Yi Zhang
6. CONCLUDING REMARKS
In summary, the experimental results demonstrate that electric fields can fracture the
depoled and poled ceramics and that the electric facture toughness exists under
purely electrical loading, which can be expressed in terms of the critical energy release
rate or the critical intensity factor of electric field strength. The significance of the
existence of electrical fracture toughness is that it enables us to use concepts from
fracture mechanics to understand electrically induced failures, and it provides de-
signers of electronic and electromechanical devices with a useful material property.
The advantage of applying the concepts of fracture mechanics to dielectric failure lies
in the ability to predict the critical electric field at which a dielectric ceramic material
containing a conductive crack or an internal electrode fails. The critical electric field
is a function of the crack dimension or the length of the electrode, while the electrical
fracture toughness is a material constant. Thus, one can predict the critical electric
field when information on the sample geometry and the electric fracture toughness is
available.
ACKNOWLEDGMENTS
This work is supported by a grant from the Research Grants Council of the Hong
Kong Special Administrative Region, China. The author thanks Drs. M.H. Zhao, R.
Fu, T.H. Wang, C.F. Qian and Y. Wang and Mr. G.N. Liu for their contributions to
the present work.
REFERENCES
Cherepanov, G.P., 1979, Mechanics of Brittle Fracture, McGraw Hill, New York, p. 317.
Fu, R., Qian, C.F. and Zhang, T.Y., 2000, Electrical fracture toughness for conductive cracks driven by
electric fields in piezoelectric materials, Appl. Phys. Lett. 76:126.
Garboczi, E.J., 1988, Linear dielectric breakdown electrostatics, Phys. Rev. B 38:9005.
Heyer, V., Schneider, G.A., Balker, H., Drescher, J., and Bahr, H.A., 1998, A fracture criterion for
conducting cracks in homogeneously poled piezoelectric PZT PIC 151 ceramics, Acta mater. 46:6615.
Lynch, C.S., Yang, W., Collier, L., Suo, Z., and McMeeking, R.M., 1995, Electric field induced cracking in
ferroelectric ceramics, Ferroelectrics 166:11.
McMeeking, R.M., 1987, On mechanical stresses at cracks in dielectrics with application to dielectric
breakdown, J. Appl. Phys. 62:3116.
Wang, T.H. and Zhang, T.Y., 2001, Electrical fracture toughness for electrically conductive deep notches
driven by electric fields in depoled lead zirconate titanate ceramics, Appl. Phys. Lett. 79:4198.
Wang, T.H., Fu, R. and Zhang, T.Y., 2001, Fracture behavior of conductive cracks in PZT 4 piezoelectric
ceramics, Proceedings of the 10th International Conference on Fracture, 2 6 December 2001, Hawaii,
USA, CD ROM Proceedings of ICF10, ICF100475OR.
Wang, T.H., 2003, Fracture Mechanics Studies of Failures of Lead Zorconate Titanate Ceramics Under
Mechanical and/or Electrical Loadings, PhD Dissertation, Hong Kong University of Science and
Technology.
Zhang, T.Y., Qian, C.F., and Tong, P., 1998, Linear electro elastic analysis of a cavity or a crack in a
piezoelectric material, Int. J. Solids Struct. 35:2121.
Zhang, T.Y., Zhao, M.H. and Tong, P., 2002, Fracture of piezoelectric ceramics, Advances in Applied
Mechanics 38:147.
Zhang, T.Y., 2002, Effects of static electric field on the fracture behavior of piezoelectric ceramics, Acta
Mechanica Sinica 18:537.
Zhang, T.Y., Wang, T.H., and Zhao, M.H., 2003, Failure behavior and failure criterion of conductive
cracks (deep notches) in thermally depoled PZT 4 ceramics, Acta Mater. 51:4881.
Zhang, T.Y., Zhao, M.H., and Liu, G.N., 2004a, Failure behavior and failure criterion of conductive
cracks (deep notches) in piezoelectric ceramics I the charge free zone model, Acta Mater. 52:2013.
Zhang, T.Y., Liu, G.N., and Wang Y., 2004b, Failure behavior and failure criterion of conductive cracks
(deep notches) in piezoelectric ceramics II experimental verification, Acta Mater. 52:2025.
FRACTURE RESISTANCE OF
HYBRID GLASS MATRIX
COMPOSITE AND ITS
DEGRADATION DUE TO
THERMAL AGEING AND
THERMAL SHOCK
Ivo Dlouhý, Zdeněk Chlup, Shabbar Atiq,

Aldo R. Boccaccini*
1. INTRODUCTION
In brittle matrix composites reinforced by continuous ceramic fibres, the favourable

fracture behaviour is provided by the presence of weak fibre/matrix interfaces, which
lead to the fibre pullout effect [1]. The thermal stability and high temperature
mechanical properties of silicate matrix composites reinforced by carbon and SiC
based fibres in oxidising environments have been investigated quite extensively in the
past by conducting thermal aging and thermal cycling experiments over a wide range
of temperatures [2 5]. A common result of investigations conducted at temperatures
in the range 500 7008C is that there is a decrease of tensile and flexural strength of
the composites. It has been shown that this is the consequence of oxidation of the
fibres, in case of carbon fibre reinforced composites, or of degradation of the fibre/
*
Ivo Dlouhý (corresponding author: [email protected]), Zdeněk Chlup, Institute of Physics of Materials,
Academy of Sciences of the Czech Republic, 61662 Brno, Czech Republic. Shabbar Atiq, Aldo
R. Boccaccini, Department of Materials, Imperial College London, London SW7 2BP, UK.
263
264 Ivo Dlouhý et al.
matrix interphase, which is in fact a carbon-rich nanometric interfacial layer, in SiC

fibre reinforced composites [2 5].
Hybrid composites are created incorporating simultaneously fibres, particles
and/or whiskers as reinforcing elements. The few hybrid glass and glass-ceramic
matrix composites described in the literature include: barium magnesium alumino-
silicate glass-ceramic matrix composites containing both SiC-Nicalon1 fibres and
SiC whiskers [6,7], cordierite glass-ceramic matrix composites containing SiC mono-
filament and SiC whiskers [8], borosilicate glass containing SiC-Nicalon1 fibres,
carbon fibres and various ceramic particle fillers (i. e. alumina, zirconia or carbon)
[9] and aluminosilicate glass containing SiC-Nicalon1 fibres and SiC particles [10].
Recently, fused silica matrix composites containing Si3 N4 particles and chopped
carbon fibres have been developed [11]. In hybrid glass matrix composites, the fibres
(e. g. chopped or continuous carbon or SiC fibres) are used to impart fracture
resistance and toughness by exploiting well-known fracture mechanisms such as
fibre pullout and crack deflection. The particulate phase is used to improve other
engineering properties relevant for the intended application of the materials, such as
thermal shock resistance, wear resistance or impact strength.
Since very little research has been conducted in the area of hybrid glass matrix
composites so far, there is need for further investigations, in particular regarding the
fracture behaviour of the material after thermal aging and thermal cycling in oxidis-
ing environments. Clearly, an in-depth understanding of the materials degradation
and microstructural damage occurring upon long-term exposure at high temperat-
ures is a prerequisite to consider new application areas for these composites.
In the present investigation, the chevron notch specimen technique was used to
detect the degradation of the flaw resistance and fracture toughness of hybrid glass
matrix composites after thermal ageing in air and thermal shock. The focus of the
study was to apply the three-point bending chevron-notch test technique to evaluate
fracture toughness (KIc ) of the composites in as-received condition and after thermal
exposition. Despite its several advantages [12,13], the chevron notched specimen
method has not received wide applicability for KIc determination in fibre reinforced
brittle matrix composites. Indeed the authors are not aware of previous work focused
on the application of the chevron notch technique to determine KIc in hybrid com-
posites, the single edge notch beam specimen method has been mainly used [7,11].
Thus another objective of this article is to examine the applicability of the chevron-
notch technique to monitor possible changes in the fracture toughness due to micro-
structural damage in this class of composite materials.
2. EXPERIMENTAL
The material investigated was a hybrid glass matrix composite supplied by Schott Glas
(Mainz, Germany). The composite contained short carbon fibres (24 wt%) as well as
ZrO2 particles (30 wt%) dispersed in a borosilicate glass (DURAN1, Schott Glas)
matrix. The density of the composites was 2:62 g=cm3 . The composites were prepared
by a proprietary method. Details of the fabrication procedure are given in the litera-
ture [9]. The samples were received in the form of plates of thickness of about 4 mm.
Rectangular test bars of nominal dimensions 4 mm 3 mm 50 mm were cut
from the plates. Thermal aging for 24 h was carried out in air using an electric
Fracture Resistance of Hybrid Glass Matrix Composite 265
furnace at temperatures of 500, 600 and 7008C. These temperatures were chosen in
agreement with those used in separate studies on thermal ageing of unidirectional SiC
fibre reinforced composites having the same borosilicate matrix [14]. Another set of
specimens was exposed to repeated thermal shocks produced by quenching samples
from 5008C into a water bath kept at room temperature. After each quench, the
samples were heated again for 15 minutes. Specimens were further characterised after
5,10,20 and 24 cycles. The temperatures chosen for this investigation cover the
medium and upper temperature ranges at which the composites are likely to find
applications [15]. After thermal treatment, the mass and dimensions of the samples
were measured, and the outer surfaces inspected carefully for the appearance of any
macroscopic damage, such as change of dimensions, shape, colour, delamination and
fibre debonding/protrusion effects.
The microstructure of the as received samples was observed by scanning electron
microscopy (SEM) using both secondary electron and backscattered electrons image
modes. Selected aged samples were also polished for SEM observations.
Thermally aged samples and samples in the as-received condition were charac-
terised using the chevron-notched specimen technique for fracture toughness deter-
mination. Chevron notches with angles of 908 were cut carefully at distances of about
15 20 mm in each test bar using a thin (0.15 mm) diamond wheel. Because the span
of the three-point bend test was 16 mm, several notches were introduced in each bar
(of 50 mm length), i.e. several fracture toughness tests were carried out using each
bar. This approach (one test bar with several notches, keeping the same loading
geometry) enables to conduct more tests than in case of different short samples and
so a more efficient usage of the available material is possible. Each bar with notches
was then placed in the three-point bend fixture and loaded up to fracture initiation.
After each test the bar was moved to place another chevron notch into position in the
three point bending rig. A constant crosshead speed of 0.01 mm/min was employed
for loading. The tests were carried out at room temperature in normal atmosphere.
Graphs of load vs. time were recorded and the maximum force was determined from
each trace. The fracture toughness value, KIc , was calculated from the maximum load
(Fmax ) and the corresponding minimum value of geometrical compliance function
(Ymin ) according to the equation [16]:
Fmax Ymin
KIC ¼
BW1=2
where B and W are the thickness and height of the specimens, respectively. The
calculation of the geometric function Ymin for chevron-notched bend bars was based
on the use of Bluhm’s slice model [17]. The calculation procedure used for the
purposes of this investigation has been described in detail elsewhere [16]. The
chevron-notch depth a0 , needed for determining Ymin , was measured after testing
from optical micrographs of fractured specimens.
Additionally, the acoustic emission technique was used during the test. Traces
of cumulative number of acoustic emission (AE) events were obtained in the same
time scale as the load vs. time plots. This technique allows for an accurate detection
of the onset of microcracking at the chevron notch tip, which occurs when a sharp
increase in the number of AE events is observed. Valid measurements for computing
KIc were those in which this increase of AE events coincided with the end of the linear
part of the force vs. time trace. For all specimens tested, the events producing the AE
response increased continuously, practically from the beginning of the test. Some
more substantial increase of AE events occurred at two stages during loading:
(i) during microcracking development at the crack tip, and (ii) when the maximum
force was reached and an unstable jump occurred.
Fracture surfaces of selected broken samples both in as-received condition and
after thermal ageing were observed by optical stereomicroscopy and SEM.
A more comprehensive description of other methods applied to characterise the
hybrid composites has been presented elsewhere [18].
3. RESULTS
3.1. As-received Material

A SEM micrograph showing the microstructure of an as-received sample is shown in
Figure 1a. Figure 1b shows the same area in backscattered electrons revealing (due to
different phase contrast) the distribution of ZrO2 particles (white phase) in the glass
matrix.
The as received material exhibits KIc values in the range 2:6---6:4 MPam1=2 , as
determined by the chevron notched specimen test [18]. These values are higher than
KIc values quoted for unreinforced borosilicate glass (0:77 MPam1=2 [13]) and for
Al2 O3 platelet (30 vol%) reinforced borosilicate glass composites (1:92 MPam1=2
[19]). The values are similar to data reported in the literature for hybrid glass matrix
composites reinforced by unidirectional SiC fibres and SiC whiskers [7] and for silica
matrix composites containing Si3 N4 particles and C fibres [11], but they are much
lower than KIc data of unidirectional SiC fibre reinforced composites containing 40
vol% fibres (19---26 MPam1=2 [13]).
The scatter of data in the present composites was however found to be extremely
high [18]. This is connected with different orientations of chopped fibre bundles
below the chevron notch tip, as discussed in a previous report [18].
Figure 1. SEM micrographs of a polished section of an as received sample showing (a) fibres, ZrO2
particles and absence of porosity; and (b) the same region of the sample in back scattered electrons (BSE)
mode, identifying the (bright) ZrO2 particles.
The fracture surface of an as-received sample is shown in Figure 2. The fracture

surface exhibits the characteristic fiber pull-out typical of ‘‘quasi-ductile’’ glass
matrix composite materials [13,15]. However, inspection of the fracture surfaces
reveals that the average pull-out lengths are not uniform across the composite section
but depend on the relative orientation of the fibre bundles and the fracture propa-
gation plane. Areas exhibiting fewer fibres are observed when these were oriented
parallel to the fracture surface. This behaviour explains qualitatively the lower KIc
values determined in this material in comparison to unidirectional fibre reinforced
composites, as mentioned above, where all fibres contribute equally to toughening by
the pull-out mechanism [13,15].
3.2. Thermally Aged Material
The fracture toughness values determined by the chevron notched specimen tech-
nique are strongly dependent on the ageing condition, as shown in Figure 3. In the
figure, also data of the as received material are shown.
Figure 2. Typical fracture surface morphology of a broken test specimens (material in as received
condition), showing fibre pull out.
Figure 3. The effect of ageing conditions on fracture toughness of the hybrid composite.
The poorest mechanical response of the samples aged for 24 h at 700 8C is

evident from comparison of data in Figure 3. It should be noted that microstructural
damage accompanying the thermal ageing has produced localised macroscopic shape
and volume changes in the samples, with the extent of this damage depending on
fiber bundle orientation. This was most evident for the above-mentioned condition of
ageing (24 hrs at 7008C), as described elsewhere [18], and it is most probably related
to softening of the glass matrix, as discussed below. The measured fracture toughness
value for this composite was KIc ¼ 0:4 MPam1=2 .
The microstructures of thermally aged samples (areas of composite sections
orientated perpendicularly to the fibres) are shown in Figure 4 and 5 for the ageing
conditions 5008C/24 h and 7008C /24 h, respectively. It is evident that at 5008C the
carbon fibres remain in the composite, however the starting of fibre oxidation leads
to the loss of contact between fibre and matrix (Figure 4). The image also shows that
there is no development of porosity and the overall microstructural damage is
limited. Microstructural damage is severe in the composite aged for 24 h at 7008C
(Figure 5). Comparing to as-received specimens (Fig. 1), high porosity in the aged
samples is evident, which can be explained by the carbon fibres decomposition due to
Figure 4. SEM micrograph of a polished section of a sample showing the microstructure after ageing for
24 h at 5008C.
Figure 5. SEM micrograph of a polished section of a sample showing microstructural degradation after
ageing for 24 h at 7008C.
oxidation effects. Moreover, softening of the glass matrix at elevated temperature

may have contributed to the formation of cavities.
The fracture resistance of the hybrid composite substantially decreased due to
the above-described microstructural changes. The progressive change of fracture
behaviour with ageing condition can be appreciated from the load deflection traces,
shown in Figure 6, and analysing fracture features, as discussed elsewhere [18]. The
traces for the as received composite and for the material after thermal ageing (Fig. 6)
show a significant decrease in the maximum value of the fracture load (Fmax ) for
similar chevron notch and specimen geometry. In addition to these quantitative
characteristics, directly related to a decrease in fracture toughness values, a change
in the slope of the linear part of the traces can also be seen in Figure 6.
The decrease in slope with increasing exposure temperature corresponds to a
decrease of specimen compliance that can be explained by poorer fibre/matrix
interfacial bond and, for the most severe thermal ageing, by matrix damage due to
softening and porosity formation as a result of carbon fibre oxidation. A decrease in
Young’s modulus is a direct consequence of this type of microstructural damage, as
also discussed in the literature for other glass composites [14].
For all aged samples, brittle fracture behaviour was observed based on fracto-
graphic analysis of the broken specimens, as carried out in the previous investigation
[18], whilst the evidence of fibre pull-out mechanism was less frequent with increasing
ageing temperature. Moreover X-ray diffraction (XRD) analyses carried out on
thermally aged samples showed the formation of a cristobalite phase [18]. This was
the sole new crystallisation peak detected and thus XRD confirmed that under the
aged conditions investigated there has been no formation of other crystalline phases
due to reactions of the ZrO2 particles and the glass.
3.3. Material exposed to repeated thermal shock
The fracture toughness values determined by the chevron notched specimen tech-
nique are strongly dependent on the condition of thermal shock, in particular for the
first ten thermal shock cycles, as shown in Figure 7. In this figure, the fracture
Figure 6. Typical load displacement plots obtained in chevron notch tests for samples in as received
condition and after thermal ageing.
toughness values are represented by two kinds of points. Filled triangles represent the
values obtained in those specimens where fibers are randomly oriented in relation to
the chevron notch plane. The empty triangles are for the fracture toughness values
representing data from specimens with fibre bundles parallel to the chevron notch
plane. It is clearly evident how the fracture resistance and thus fracture toughness
strongly depends on the fibre bundle orientation as mentioned above and also
discussed in detail in our previous work [18].
It was noted that the highest number of thermal shock cycles that was applied in
this experiment (24 cycles) led to specimen failure without any further mechanical
loading: the specimen broke upon quenching into water. Fracture plane analysis
showed that the crack had developed along the interface between two fibre bundles.
Figure 8 shows a SEM micrograph of the fracture surface of a sample thermally
shocked for 20 cycles. Very limited pull-out is detected in this sample, which is similar
to observations on samples thermally aged at 5008C for 24 hours, as discussed above
and reported in detail elsewhere [18].
Figure 7. The effect of thermal shock cycling on fracture toughness of the hybrid composite.
Figure 8. SEM micrograph showing a typical fracture surface of a composite after 20 thermal shock
cycles
In Figure 7, also data of the as received material (filled circles) and material aged
at 5008C for 24 hrs (empty rhombi) are shown for comparison. It is seen that the
most significant part of composite damage occurred during the first 10 thermal shock
cycles. However, the total exposition time at 500 8C during this treatment was
comparably shorter than in case of thermal ageing analysed in previous chapter.
The more extensive and rapid damage during thermal shock cycling is thus ascribed
to the quenching effect, in particular to the effect of cumulative thermal stresses
produced during repeated thermal shocks.
The investigation of load-deflection traces provides also very valuable informa-
tion for the assessment of damage development in thermally shocked samples. The
representative curves are shown in Figure 9. Similarly as in case of specimens after
thermal ageing, the load-deflection curves and fracture toughness values are strongly
affected by the orientation of fibre bundles with respect to the chevron notch plane.
In case of parallel orientation, the remarkable drop of fracture resistance is associ-
ated with the significant change of the load-deflection trace as it can be seen, as
example, comparing the curves for specimens after 10 thermal shock cycles in Figure
9: specimen 2 3 displayed parallel orientation of fibers and fracture plane, while
specimen 2 1 displayed random orientation of fibres.
After the first five thermal shock cycles the fracture behaviour is very similar to
that of as received composites (compare the corresponding curves in Figures 7 and 10).
For higher number of thermal shock cycles two typical features arise on the load
deflection traces. The first one is a marked pop-in behaviour at loads of about 10 N
(marked by circles in Figure 10). The second phenomenon observed is increasing
specimen deflection at fracture (maximum) load, as evident when comparing the
curves for 10 and 24 cycles in Figure 10. This can be caused by the change of crack tip
behaviour affected by damage development at the matrix/fibre interfaces.
4. DISCUSSION
As discussed in the previous investigation [18], the fracture toughness values of the
hybrid composites investigated are in the range 2:6---6:4 MPam1=2 and they exhibit an
Figure 9. Typical load displacement plots obtained in chevron notch tests for samples in as received
condition and after thermal ageing.
extremely large scatter. This was associated with different orientations of short fibre
bundles in relation to the chevron notch plane as discussed in detail on the basis of
microstructural and fracture surface observations [18]. In general, the as-received
material exhibit relatively high KIc values, in comparison with other hybrid glass
matrix composites [7,11,19], which suggests that an effective part of the toughening
effect may be provided by the zirconia particles added. The addition of tetragonal
zirconia particles to glass matrices should contribute to toughening by the tetragonal-
to-monoclinic transformation mechanism, as investigated in the past in a variety of
glass and glass-ceramic matrices [20 23]. However in the present composite a mono-
clinic zirconia particulate phase was added, thus ruling out transformation toughen-
ing. In fact the presence of a particulate zirconia phase dispersed in a brittle matrix may
further contribute to toughening and strengthening by other mechanisms, mainly by
crack deflection and by residual thermal stresses [22,23]. In the present composites, a
favourable residual thermal stress field (hoop compressive stresses in the matrix)
should develop upon cooling from the fabrication temperature as a result of the
thermal expansion mismatch between the zirconia particles (10:3 10 6 1=8C [24])
and the borosilicate glass matrix (3.25 10 6 1/8C [13]). However, the relative contri-
bution to toughening due to the fibre pull-out mechanism and to zirconia particle
related mechanisms cannot be quantitatively determined from the results obtained in
this study. In general, however, fibre pull-out is thought to be the dominant and more
effective toughening mechanism in these hybrid composites [18].
For the most severe thermal ageing conditions (7008C, 24 h, Figure 3), the
KIc values dropped to 0:4 MPam1=2 . This value is even lower than KIc of monolithic
borosilicate glass (0:77 MPa m1=2 [19]), which can be explained by the very high
porosity developed in the aged sample as consequence of carbon fibre oxidation,
as discussed elsewhere [18]. Thus, for the conditions of thermal ageing investigated
here, degradation of carbon fibres due to oxidation occurred, this being more
severe with increasing temperature, and therefore the apparent fracture toughness
and flaw tolerant resistance of the composites decreased. For the highest temperature
tested (7008C), enhanced porosity formation in the matrix was detected, which is
thought to be the result of softening of the glass. Indeed the aging temperature
(7008C) was well above the transformation temperature of the borosilicate glass
matrix used (Tg ¼ 5258C for borosilicate DURAN1 glass [15]). Fractographic ob-
servations support well this discussion as reported previously [18]. It seems
that oxidation of the carbon fibres starts at 5008C, leading to degradation of
the fibre matrix interface. As shown in Figure 4 for a polished sample, the interface
boundary between the glass matrix and fibres appears to be degraded comparing
to the initial state (Figure 1). The combination of matrix and matrix-fibre inter-
face degradation and the damage to the boundary region between neighbouring
fibre bundles, as detected by scanning electron microscopy observations (Figures 4
and 5), supplied a more favourable condition for fracture initiation and this
resulted in a decrease of both KIc values and the data scatter in thermally aged
materials.
A similar detrimental effect on KIc values was measured in samples after
repeated thermal shocks from 5008C. KIc dropped to 1:3 MPam1=2 for the maximum
number of thermal shocks applied (24 cycles), as shown in Figure 7. The figure also
indicates that the fracture toughness values of thermally shocked samples lie between
those of as received samples and those of specimens thermally aged at 5008C for 24
hours. This is related to the shorter exposure time at high temperature in the
thermally shocked samples in comparison to the aged samples.
The actual contribution of the thermal shocks to composite degradation is hard to
quantify and the samples have undergo degradation also during the holding time at
temperature. There is an important effect that has been detected however when com-
paring the load-deflection traces of specimens thermally aged (Fig. 6, see the curve for
5008C) and specimens after thermal shock cycling (Fig. 9, curves for 10 an 24 cycles).
Occurrence of pop-in effect at loads of about 10 N appeared to be typical for specimens
after thermal shock treatment. An attempt has been made to estimate the stress
intensity factor (fracture toughness value) corresponding to this level of load for
particular specimens. This is possible assuming that the dependence of the geometrical
function on crack length is known for the particular chevron notch depth. The values
obtained are represented by crosses in Fig. 8; they correspond well to the fracture
toughness of the borosilicate glass matrix used in the composite (0:7 MPam1=2 [13]).
One possible explanation is that whereas high temperature exposition (aging) leads to
carbon fiber and fiber matrix interface degradation, the thermal shock loading causes
additional damage to the matrix. This can be due to residual thermal stresses developed
in the matrix after increasing number of thermal shocks, which causes premature
matrix microcracking at the chevron notch tip in the early stages of specimen loading.
This indicates that cumulative degradation of the composite due to increasing number
of thermal shocks is more detrimental, in terms of fracture toughness, compared to
simple thermal exposition (aging) at the same temperature.
5. CONCLUSIONS
The fracture behaviour of hybrid glass matrix composite samples before and after
thermal ageing and/or thermal shock cycling in air has been analysed and evaluated
for the first time. The fracture toughness values measured were in the range
2:6---6:4 MPam1=2 for the as received material and they were significantly reduced
by thermal ageing and thermal shock treatments. The KIc values dropped to
0:4 MPam1=2 for ageing at 7008C for 24 h due to fibre oxidation and porosity
formation in the matrix as a result of softening of the glass. The KIc values decreased
to 1:3 MPam1=2 for the maximum number of thermal shock cycles from 5008C
(24 cycles). The damage in the samples due to thermal shock is based on fibre
oxidation and porosity formation, as in case of thermal ageing, and, in addition,
due to superposed residual thermal stresses in the matrix produced by repeated
thermal shocks. The present results are useful to define parameters and conditions
for the high temperature application of the composites.
ACKNOWLEDGEMENTS
This research was funded from grant of NATO (Nr. PST.CLG.977558). The experi-
mental part of the research was financially supported by grant Nr. A2041003 of the
Grant Agency of the Academy of Sciences, Czech Republic. The authors gratefully
acknowledge Prof. W. Beier and Mr. R. Liebald of Schott Glas, Mainz, Germany,
for supplying the samples.
REFERENCES
1. A.G. Evans, F.W. Zok FW. Review. The Physics and Mechanics of Fibre Reinforced Brittle Matrix
Composites. J. Mat. Sci., 29, 3857 3896 (1994).
2. S. Sutherland, K.P. Plucknett, and M.H. Lewis, High Temperature Mechanical and Thermal Stabil
ity of Silicate Matrix Composites. Comp. Eng., 5, 1367 1378 (1995).
3. W. Sinkler, M. Monthioux, V. Bianchi, P. Goursat, E. Menessier, Carbon Fibre Reinforced (YMAS)
Glass Ceramic Matrix Comp. II. Struct. Changes in the Matrix with Tem. J. Europ. Ceram. Soc., 19,
305 316 (1999).
4. Y.M. Sung, S. Park, Thermal and Mechanical Properties of Graphite Fibre Reinforced Off Stochio
metric BaO:Al2 O3 :2SiO2 Glass Ceramic Matrix Composites. J. Mat. Sci. Lett., 19, 315 317 (2000).
5. K.M. Prewo, J.A. Batt, The Oxidative Stability of Carbon Fibre Reinforced Glass Matrix Compos
ites. J. Mat. Sci., 23, 523 527 (1988).
6. N. Chawla, K.K. Chawla, M. Koopman, B. Patel, C. Coffin, J.I. Eldridge, Thermal Shock Behaviour
of a Nicalon Fiber Reinforced Hybrid Glass Ceramic Composite. Comp. Sci. Technol., 61, 1923
1930 (2001).
7. Lewinshon LA. Hybrid Whisker Fiber Reinforced Glass Matrix Composites with Improved Trans
verse Toughness. J. Mat. Sci. Lett., 12, 1478 1480 (1993).
8. S.B. Haug, L.R. Dharani, D.R. Carroll, Fabrication of Hybrid Ceramic Matrix Composites. Appl.
Comp. Mater., 1, 177 181 (1994).
9. R. Reinicke, K. Friedrich, W. Beier, R. Liebald, Tribological Properties of SiC and C Fibre Re
inforced Glass Matrix Composites. Wear; 225 229, 1315 1321 (1999).
10. K.P. Gadkaree, Particulate Fibre Reinforced Glass Matrix Hybrid Composites, J. Mat. Sci.,
27, 3827 3834 (1992).
11. D.C. Jia, Y. Zhou, T.C. Lei, Ambient and Elevated Temperature Mechanical Properties of Hot
pressed Fused Silica Matrix Composites, J. Europ. Ceram. Soc., 23, 801 808 (2003).
12. J.S. Ha, K. K. Chawla, Mechanical Behaviour of Mullite Composites Reinforced with Mullite Fibres.
Materials Science and Engineering; A203, 171 176 (1995).
13. I. Dlouhy, A.R. Boccaccini, Reliability of the Chevron Notch Technique for Fracture Toughness
Determination in Glass Composites Reinforced by Continuous Fibres. Scripta Mat., 44 (3), 531 537
(2001).
14. A.R. Boccaccini, A.J. Strutt, K.S. Vecchio, D. Mendoza, K.K. Chawla, C.B. Ponton, and D.H.
Pearce, Behavior of Nicalon Fiber Reinforced Glass Matrix Composites Under Thermal Cycling
Conditions. Composites A, 29, (11), 1343 1352 (1998).
15. A.R. Boccaccini, D.H. Pearce, J. Janczak, W. Beier, and C.B. Ponton. Investigation of the Cyclic
Thermal Shock Behaviour of Fibre Reinforced Glass Matrix Composites Using a Non Destructive
Forced Resonance Technique. Mat. Sci. Technol., 13, 852 859 (1997).
16. I. Dlouhy, M. Holzmann, J. Man, and L. Válka, The Use of Chevron Notched Specimen for Fracture
Toughness Determination. Metallic Materials, 32, 3 13 (1994).
17. J.I. Bluhm, Slice Synthesis of a Three Dimensional ‘‘Work of Fracture Specimen’’. Engng. Fract.
Mech.; 7: 593 604 (1975).
18. I. Dlouhy, Z. Chlup, S. Atiq, A.R. Boccaccini, Fracture Behaviour of Hybrid Glass Matrix Com
posites: Thermal Ageing Effects, Composites Part A, submitted (2003).
19. A.R. Boccaccini, V. Winkler, Fracture Surface Roughness and Toughness of Al2 O3 Platelet
Reinforced Glass Matrix Composites. Composites A, 33, 125 13 (12002).
20. Z.B. Yu, D.P. Thompson, A.R. Bhatti, Synergistic Role of Carbon Fibres and ZrO2 Particles in
Strengthening and Toughening Li a sialon Composites. J. Europ. Ceram. Soc., 22, 225 235 (2002).
21. J.K. Guo, Z. Mao, C. Bao, T. Wang, D.S. Yan, Carbon Fibre Reinforced Silicon Nitride Composite.
J. Mat. Sci.; 17, 3611 3316 (1982).
22. M. Nogami, and M. Tomozawa, ZrO2 Transformation Toughened Glass Ceramics Prepared by the
Sol gel Process from Metal Alkoxides. J. Am. Ceram. Soc., 69, 99 102 (1986).
23. R.D. Sarno, and M. Tomozawa, Toughening Mechanisms for a Zirconia Lithium Aluminosilicate
Glass Ceramic. J. Mat. Sci.; 30, 4380 4388 (1995).
24. D.J. Green, An Introduction to the Mechanical Properties of Ceramics. Cambridge University Press,
Cambridge, UK, 1998.
FRACTURE TOUGHNESS OF
BaTiO3 AND BaTiO3-Al2O3
COMPOSITE UNDER
ELECTRIC FIELD
Sirirat Rattanachan, Yukio Miyashita

and Yoshiharu Mutoh*
ABSTRACT
In order to investigate the effect of electric polarization on fracture toughness of

monolithic BaTiO3 and 5 mol% BaTiO3 -Al2 O3 composite (5B95A), indentation
fracture tests were conducted on unpoled and poled samples under various applied
electric fields. From the results, it is found that applied electric fields can increase or
decrease fracture toughness in both monolithic BaTiO3 and 5B95A composite de-
pending on direction of electric field. For unpoled and poled samples under the
positive (in the same direction to the poling direction), fracture toughness parallel to
the poling direction increased, while that perpendicular to the poling direction
decreased. Under the negative (in the opposite direction to the poling direction),
fracture toughnesses both parallel and perpendicular to the poling direction for poled
monolithic BaTiO3 and 5B95A composite were decreased with increasing applied
electric field. This behavior of 5B95A composite was consistent with that of mono-
lithic BaTiO3 . It was concluded that polarization switching of piezoelectric BaTiO3
particles phase under electric fields has a significant influence on fracture toughness
of the present composite.
*
The author address: Department of Mechanical Engineering, Nagaoka University of Technology,
1603 1 Kamitomioka, Nagaoka shi 940 2188 Japan, e mail: [email protected]
297
298 Sirirat Rattanachan et al.
1. INTRODUCTION
Many approaches to improve the fracture toughness of ceramics have been devel-
oped. As a novel approach for toughening of ceramics, composites with piezoelectric
secondary phase were proposed, where energy dissipation due to piezoelectric effect
and domain switching were suggested as the main toughening mechanisms1. Various
composites such as BaTiO3 =Al2 O23 , BaTiO3 =3Y-TZP3 , Nd2 Ti2 O7 =Al2 O43 , Sr2 Nb2 O7
=3Y-TZP5 and BaTiO3 =ZrO62 have been studied. High fracture toughness has been
recently achieved for a BaTiO3 toughened Al2 O3 system7, where the fracture tough-
ness of 5:5 MPam1=2 attained for an Al2 O3 matrix composite with 5 mol% BaTiO3
sintered by a spark plasma sintering, while that of a monolithic Al2 O3 was about
4 MPa:m1=2 .
In the previous work, Al2 O3 based composites with 3 and 5 mol% BaTiO3
secondary phases were fabricated by the pressureless sintering method8 The Inden-
tation Fracture (IF) method was employed to evaluate fracture toughness before and
after polarization. Toughening mechanisms of BaTiO3 =Al2 O3 composites were dis-
cussed. It was verified that an polarization induced distinct anisotropy in fracture
toughness of BaTiO3 =Al2 O3 composites between parallel and perpendicular direc-
tions to the poling direction. Crack deflection around BaTiO3 particles was often
observed in the poling direction. Anisotropy in residual stress was also observed after
polarization. It was suggested that the polarization induced the domain reorientation
along the poling direction and consequently resulted in the anisotropy in residual
stress8.
Several researches are available on anisotropic behavior of fracture toughness
and toughening mechanisms of monolithic piezoelectric materials. Pak and Tobin9
performed the IF tests on the isotropic and anisotropic planes of poled PZT samples.
They observed that the presence of the electric field in the direction of poling assisted
crack growth and retarded it in reversed field. Lynch10 studied crack growth in a
ferroelectric and a relaxor composition of PLZT ceramic using Vickers indentation
under electric field. He found that the effect of residual stress was much larger than
other mechanisms that interact with crack propagation. Fu and Zhang11 conducted
IF tests and compact-tension fracture toughness tests to study the effect of applied
electric field on fracture toughness of poled commercial PZT ceramics. Their results
were inconsistent with the experimental results by Pak and Sun12 and also Pak and
Tobin9. Kishimoto and Seo13 reported a change in strength of barium titanate/
zirconia composite by polarization treatment. The results exibited that bending
strength of BaTiO3 -8YSZ composite increased by longitudinal poling and decreased
by transverse poling compared to the untreated sample. Both strengthened and
weakened specimens tended to return almost to the original strength by heating
over Tc of barium titanate.
The research works on toughening mechanisms and fracture toughness of
ceramic composites with piezoelectric phases have been limited, while those for
monolithic piezoelectric/ferroelectric materials have been available, as mentioned
above. Therefore, details of toughening behavior and mechanisms in composite
materials with piezoelectric phases have been still not clear.
In the present study, fracture toughness of 5 mol% BaTiO3 =Al2 O3 composite
for both before and after polarization was investigated under various applied electric
field to discuss the toughening mechanisms. The generated cracks in the parallel or
Fracture Toughness of BaTiO3 and BaTiO3-Al2O3 Composite 299
perpendicular direction to poling were observed in detail, and compared to those for
monolithic BaTiO3 .
EXPERIMENTAL PROCEDURE
Commercial barium titanate (BT-05, Sakai chemical industrial Co. Ltd.) with an
average particle size of 0:5 mm and high purity alumina (Sumitomo Sekitan Kougyo,
KK) powders with an average particle size of 0:2 mm were used as the starting
materials. Al2 O3 powders with 5 mol% BaTiO3 powders (as denoted 5B95A com-
posite in this study) were mixed by a ball milling machine with alumina balls in
ethanol for 24 h. The wet slurry was then dried by a rotary evaporator. Dried
powders were milled again and sieved through a 150 mm mesh screen. The mixed
powders were put into a cylindrical graphite die with a diameter of 25 mm, which was
on the vibration table for homogeneous packing of the powders, and then sintered
with graphite punches on both sides by using a spark plasma sintering machine (Dr.
sintering, SPS 1030, Sumitomo Coal Mining Co., Ltd.). The sintering was conducted
at a sintering temperature of 13008C under an applied load of 38 MPa in vacuum.
The temperature was increased up to the sintering temperature at a rate of 1008C/
min. After holding of 3 5 min at the sintering temperature, the d.c. power of SPS
machine was shut off to let the sample rapidly cool down to 6008C in 30 min. During
sintering, the linear change in shrinkage was recorded by monitoring the displace-
ment of the sample along the pressing direction. The temperature was measured by
means of an optical pyrometer focused at the middle of the graphite die surface.
Rectangular bar specimens were cut from the sintered sample with a diamond saw.
The specimen surfaces were then ground and polished with a 600# SiC abrasive
paper and a finer diamond paste to yield a mirror-like surface suitable for the
subsequent indentation test. The final dimensions of the specimens were
4 mm 3 mm 20 mm: BaTiO3 monolithic ceramics were also fabricated by the
same sintering procedure at 11008C.
The relative density of the monolithic BaTiO3 and 5B95A composite sintered by
SPS were 99.99 and 94%, respectively. From the results of XRD analysis, some
intermediate phases such as BaAl13:2 O20:8 , BaAl6 TiO12 , as well as BaTiO3 and
Al2 O3 phases were found. Figures 1(a) and (b) show microstructures of monolithic
BaTiO3 and 5B95A composite sintered at 13008C by SPS, respectively. Al2 O3 ,
intermediate and BaTiO3 grains in the SEM micrographs (Back scattering electrons
image) are black, gray and white regions, respectively. The agglomerated BaTiO3
grains and the aggregated intermediate phases were formed at grain boundaries.
2.2. Polarization and Fracture Toughness Test

In order to remove carbon contamination on sintered samples, 5B95A composite was
annealed at 11008C before polarization. 3 35 mm surfaces of the specimen were
polished by using an 800 mesh SiC paper and then applied silver paste as an electrode
and dried in air oven at 1208C for 10 min. Polarization treatment was conducted in
silicon oil at 808C for 10 min with electric field of 1.25 kV/mm. The indentation
Figure 1. Scanning electron micrographs of thermal etched surface of (a) monolithic BaTiO3 and (b)
5B95A composites (A:Al2 O3 , B:BaTiO3 , BAO/BATO: intermediate phases).
surface was parallel to the poling direction for poled specimens. The indenter was
loaded on the surface perpendicular to the electrode surfaces. In this experiment, the
cracks introduced by a Vickers indenter were perpendicular or parallel to the applied
electric field (see Fig. 2).
The combination of electrical and mechanical loading was achieved by applying
an electrical potential across the electroded specimen during indenting. Figure 3
shows sample setup for Vickers indentation test under applied electric field. To avoid
additional mechanical constraints on the specimen, the electric fields were applied to
the electrode surfaces by the welding contacts. Static electric fields were applied in
either the same or opposite direction to the poling direction using a high voltage
power supply (model HDV-40 K2 SU, Pulse Electronic Engineering, Co., Ltd.). The
electric field was applied before the indenter contacted the specimen and was turned
off after the indenter was lifted off the specimen.
The Vickers indentation test was conducted at room temperature using loads of 9.8
and 49 N for BaTiO3 and 5B95A composite, respectively, for a constant duration of
15s. After unloading, the crack lengths were measured immediately. Fracture tough-
Figure 2. Schematics of directions of applied electric field and cracks
Figure 3. Schematic view of sample setup in Vickers indentation test under applied electric field.
ness was evaluated using the equation proposed by Tanaka14. At least 20 measurements
were taken for each data, and the average and standard deviation were calculated.
3.1. Fracture Toughness of Unpoled and Poled Samples

In absence of applied electric field, fracture toughness of unpoled BaTiO3 and 5B95A
composite were isotropic, where crack lengths in two orthogonal directions were
almost identical. After poling, anisotropy in fracture toughness was found for both
monolithic BaTiO3 and 5B95A composite, where the crack length in perpendicular
direction was longer compared to that in parallel direction (see Fig. 4). Similar
behavior have been observed in some other works15–17.
3.2. Fracture Toughness under Applied Electric Field
Polarization switching is known as a unique feature of piezoelectrics/ferroelectrics.

A switching of 1808 causes little strain and is activated by the electric field, whereas a
switching of 908 results from the electric and / or stress fields and induced a sizable
strain of fixed amount, with contracting along the previous polarization direction
and elongation along the current one. Therefore, anisotropy of fracture toughness
for ferroelectric may result from the polarization switching induced by the crack tip
stress. It has been reported that fracture toughnesses along the poling direction and
the normal to it are different6,10,15–19.
In the present composite, since BaTiO3 structure extends in the poling direction
whereas Al2 O3 matrix maintains its crystal structure and orientation, even after
poling the residual stress would be generated in Al2 O3 matrix. In the pervious
work8, residual stresses were observed in the same composite as the present study.
From Fig. 5 (a) and (b), it was found that since the difference between fracture
toughnesses of poled and unpoled BaTiO3 were not significant, the relationships
Figure 4. Fracture toughness of unpoled and poled samples in the absence of applied electric field
(a) monolithic BaTiO3 and (b) 5B95A composite.
Figure 5. Fracture toughness of monolithic BaTiO3 as a function of the applied electric field (a) Unpoled
BaTiO3 under applied electric field. (b) Poled BaTiO3 under positive applied electric field (c) Poled BaTiO3
under negative applied electric field.
between fracture toughness and applied electric field for unpoled and poled samples
were also almost identical. Fracture toughness in the poling direction increased with
increasing strength of applied electric field, while that perpendicular to the poling
direction decreased with increasing strength of applied electric field. These behavior
can be explained due to polarization switching: the compressive stress and tensile
stress are induced by the switching in the parallel and perpendicular direction to the
poling direction, respectively. The same behavior was observed in the 5B95A com-
posite, as can be seen in Figs. 6(a) and (b).
Under negative applied electric field, both the fracture toughnesses parallel and
perpendicular to the poling direction were decreased with increasing strength of
applied electric field for both poled BaTiO3 and 5B95A composite, as shown in
Figs. 5(c) and 6(c). These behavior can be explained as follows: The negative applied
electric field induced polarization switching of 1808, which reduced the compressive
residual stress in the poling direction, which was introduced by the previous poling
treatment, and increases the tensile residual stress perpendicular to the poling direc-
tion.
Improvement of fracture toughness due to in-situ application of electric field
was significant compared to that due to the poling treatment. This may be explained
as follows:
Figure 6. Fracture toughness of 5B95A composite as a function of the applied electric field (a) Unpoled
5B95A composite under applied electric field. (b) Poled 5B95A composite under positive applied electric
field (c) Poled 5B95A composite under negative applied electric field
The polarization switching situation corresponding to the certain applied elec-

tric field might be maintained during crack propagation under applied electric field,
while it is possible to degrade during crack propagation for the poled sample without
applied electric field. Whether the electric field impedes or enhances crack propaga-
tion is still a debatable issue. From Zeng and Rajapakse’s work21, the effects of
anisotropic properties and electromechanical coupling were investigated. Base on
their numerical results, it agreed with the experimental observation of this study.
Another factor to be considered will be the interaction between non-uniform stress
and electric field distributions near the crack tip20,22. Domain switching behavior
may be different between with and without applied electric field.
CONCLUSION
Fracture toughness of unpoled and poled monolithic BaTiO3 as well as Al2 O3 -based
composite with 5 mol% BaTiO3 was evaluated under various applied electric fields.
The positive applied electric field increased the fracture toughness in the parallel
direction to the poling direction and decreased it in the perpendicular to the poling
direction. The negative applied electric field degraded the fracture toughness in the
parallel direction and enhanced it with perpendicular direction. Improvement of
fracture toughness under positive applied electric field was significant compared to
that for poled sample without applied electric field. These behavior can be explained
based on the polarization switching and the resultant induced residual stress.
ACKNOWLEDGEMENT
The authors wish to thank Macoho Co. for conducting spark plasma sintering, and
Prof. Uematsu’s laboratory in the department of chemistry, Nagaoka University of
Technology, for preparing the mixed powders.
REFERENCES
1. Chen, X.M. and Yang, B.A., A new approach for toughening of ceramics, Mater. Lett., 1997, 33,
237 240.
2. Chen, X.M. and Yang, B., A new approach for toughening of ceramics. Mat. Lett., 1997, 33,
237 240.
3. Yang, B., Chen, X. M. and Liu, X. Q., Effect of BaTiO3 addition on structures and mechanical
properties of 3Y TZP ceramics. J. Eur. Ceram. Soc., 2000, 20, 1153 1158.
4. Yang, B. and Chen, X.M., Alumina ceramics toughened by a piezoelectric secondary phase. J. Eur.
Ceram. Soc., 2000, 20, 1687 1690.
5. Chen, X. M., Liu, X. Q., Liu, F. and Zhang, X. B., 3Y TZP ceramics toughened by Sr2 Nb2 O7
secondary phase. J. Eur. Ceram. Soc., 2001, 21, 477 481.
6. Seo, S. and Kishimoto, A., Effect of polarization treatment on bending strength of barium titanate/
zirconia composite. J. Eur. Ceram. Soc., 2000, 20, 2427 2431.
7. Rattanachan, S., Miyashita, Y., and Mutoh, Y., Microstructure and fracture toughness of a spark
plasma sintered Al2 O3 based composite with BaTiO3 particulates, Journal of the European Ceramic
Society, 2003, 23[8], 1269 1276
8. Rattanachan, S., Miyashita, Y., and Mutoh, Y., Effect of Polarization on Fracture Toughness of
BaTiO3 =Al2 O3 Composites, Journal of the European Ceramic Society, in submitted.
9. Pak, Y.E., and Tobin, A., On electric field effects in fracture of piezoelectric materials, ASME
Mechanics of Electomagnetic Materials and Structures, Lee, J.S., Mangin, G.A., and Shindo, Y.,
eds., AMD vol161, 1993, MD vol. 42, 51 62.
10. Lynch, C.S., Fracture of Ferroelectric and Relaxor Electro Ceramics: Influence of Electric field, Acta
Mater., 1998, 46, 599 608.
11. Fu, R., and Zhang, T. Y., Effect of Electric field on the Fracture Toughness of Poled Lead Zirconate
Titanate Ceramics, J. Am. Ceram. Soc., 2000, 83, 1215 1218.
12. Park, S.B., and Sun, C.T., Effect of Electric field on Fracture of Piezoelectric Ceramics, Int. J. Fract.,
1995, 70, 203 216.
13. Kishimoto, A., and Seo, S., Strength control of a ceramic composite by electric field, Smart Materials,
Proceeding of SPIE, 2001, 4234, 321 327.
14. Tanaka, K., Elastic/Plastic Indentation Hardness and Indentation Fracture Toughness: the Inclusion
Core Model, J. Mater. Sci., 1987, 22, 1501.
15. Okazaki, K., Mechanical Behavior of Ferroelectric Ceramics, Bull. Am. Ceram. Soc., 1984, 63(9),
1150 1157.
16. Pisarenko, G.G., Chushko, V.M., and Kovalev, S.P., Anisotropy of Fracture Toughness of Piezo
electric Ceramics, J. Am. Ceram. Soc., 1985, 68(5), 259 265.
17. Yamamoto, T., Igarashi, H., and Okazaki, K., Internal Stress Anisotropics Induced by Electric Field
in Lanthanum Modified PbTiO3 Ceramics, Ferroeletrics, 1983, 50, 273 278.
18. Wang, H., and Singh, R.N., Crack propagation in piezoelectric ceramics: Effects of applied electric
fields, J. Appl. Phys., 1997, 81[1], 7471 7479
19. Singh, R.N., and Wang, H., Crack Propagation in Piezoelectric Materials Under Combined Mech
anical and Electrical Loadings: An Experimental Study, Adaptive Material Systems ASME, 1995,
AMD Vol. 206/MD Vol. 58, 85 95.
20. Schneider, G.A., and Heyer, V., Influence of the Electric Field on Vickers Indentation Crack Growth
in BaTiO3 , J. Euro. Ceram. Soc., 1999, 19, 1299 1306.
21. Zeng, X., and Rajapakse, R.K.N.D., Domain switching induced fracture toughness variation in
ferroelectrics, Smart Mat. Struc., 2001, 10, 203 211.
22. Zhu, T. and Yang, W., Toughness variation of ferroelectrics by polarization switch under non
uniform electric field, Acta mater., 1997, 45[11], 4695 4702.
BaTiO3-MgO COMPOSITES
SINTERED BY SPARK
PLASMA SINTERING
Sirirat Rattanachan, Yukio Miyashita

and Yoshiharu Mutoh*
ABSTRACT
BaTiO3 -MgO composites with various compositions were fabricated by using a Spark
Plasma Sintering (SPS) method. BaTiO3 -MgO composites with very high density were
successfully sintered by SPS method. From the experimental results of as-sintered
composites, Vickers hardness decreased and fracture toughness increased with in-
creasing BaTiO3 content. The higher fracture toughness of BaTiO3 -MgO composite
with 10 vol% BaTiO3 content was achieved compared to monolithic MgO. In order
to investigate the effect of piezoelectric BaTiO3 particles dispersed in MgO matrix,
fracture toughness of the polarized composites was also evaluated. After polariza-
tion, fracture toughness of the BaTiO3 -MgO composites was improved and higher
than that of monolithic MgO, while polarization induced distinct anisotropy in frac-
ture toughness between parallel and perpendicular directions to the poling direction.
1. INTRODUCTION
In the field of engineering ceramics, ceramic composites have been increasingly

studied for overcoming brittle nature of monolithic ceramics. Ceramic composites
* The author address: Department of Mechanical Engineering, Nagaoka University of Technology,

1603 1 Kamitomioka, Nagaoka shi 940 2188 Japan, e mail: [email protected]
287
with perovskite-type-ferroelectrics have been developed for introducing ferroelectri-

city into structural ceramics. Ferroelectric ceramic has the sensing ability of crack
propagation using electromotive force1. Fracture toughness of ferroelectric ceramics
is influenced by electric field.
Polarization of ferroelectric materials induces anisotropy of fracture toughness. It
is known that this phenomena is caused by domain switching. Yang and Zhu2 have
proposed a model of stress-assisted 908 polarization switching to explain the toughen-
ing behavior. Many experimental and theoretical studies (Mehta and Virkar 19903,
Lynch et al 19954, etc) showed that the intensified stress and electric fields at a crack tip
in ferroelectrics can trigger domain switching. If ferroelectric particles are introduced
into the structural ceramics, the particles in the composites will also exhibit domain
switching and then improve the fracture resistance. This behavior also suggests that
the composite will have some intelligent functions such as fracture sensing ability5.
The microstructure, grain size and density as well as flaws play an important role
to the mechanical properties of ceramics. Spark plasma sintering (SPS) is one of the
solid consolidation processes similar to hot pressing, where powders are pressed
uniaxially in a graphite die and simultaneously the direct current pulse voltage is
applied. Rapid sintering in SPS realizes short soaking time at high sintering tem-
perature and then high density with fine grains6–8.
In the present study, piezoelectric/ferroelectric composites, BaTiO3 -MgO, were
fabricated by using an SPS method at 13008C in vacuum atmosphere with an applied
pressure of 38 MPa. Microstructure, phases and fracture toughness of the compo-
sites were also studied.
2.1. Composite Preparation

Commercial Barium Titanate, BT-05 (Sakai chemical industrial Co. Ltd.) and high
purity magnesium oxide (MgO) powders were used as the starting materials. BaTiO3
and MgO powders with various contents (1,3,5 and 10 vol% BaTiO3 ) were mixed by
a ball milling machine in ethanol for 24 h. The milled slurry was dried by an rotary
evaporator, and then was milled again and sieved through a 150 mm mesh screen. The
mixed powders were put into a cylindrical graphite die with a diameter of 25 mm,
which was on the vibration table for homogeneous packing of the powders, and then
sintered with graphite punches on both sides by using a spark plasma sintering
machine at 13008C under an applied load of 38 MPa in vacuum. The temperature
was increased up to a certain sintering temperature at a rate of 1008C/min. After
holding about 5 min at the sintering temperature, the d.c. power of the SPS machine
was shut off to let the sample rapidly cool down to 6008C in 30 min. During
sintering, the linear change in shrinkage was recorded by monitoring the sample
length along the pressing direction. Rectangular bar specimens were cut with a
diamond saw from the sintered samples. The specimen surfaces were then ground
and polished with a 600# SiC abrasive paper and finer diamond paste to yield a
mirror-like surface suitable for the subsequent indentation test. The final dimensions
of the specimen were 4 mm 3 mm 20 mm. The MgO monolithic ceramics was
also fabricated by the same sintering procedure.
Fracture Toughness of BaTiO3-MgO Composites Sintered 289
2.2. Characterization
The bulk density was determined by Archimedes’ method in water. To determine the
phases of sintered composites, the X-ray diffraction (XRD) analysis was conducted
using Shimadzu XRD 6100 with nickel-filtered Cu Ka radiation. The average grain
size and pore size of the samples were determined from scanning electron micro-
graphs of the polished and etched samples.
2.3. Polarization
In order to remove carbon contamination on sintered samples, BaTiO3 -MgO com-

posites were annealed at 10008C before polarization. The 3 35 mm surfaces of
specimens were polished by using an 800 mesh SiC paper and then applied silver
paste as electrode and dried in air oven at 1208C for 10 min. The poling was
conducted under an electric field of 1.25 kV/mm at 808C for 10 min.
2.4. Fracture Toughness Test
The indentation fracture (IF) method for evaluating fracture toughness was adopted.
The Vickers indentation test was conducted at room temperature using a load of
49 N for a constant duration of 15 s. The indentation surface was parallel to the
poling direction for poled specimens. After unloading, the crack lengths were meas-
ured. c? and c== were denoted to the crack lengths in the perpendicular and parallel
to the poling direction, respectively. Fracture toughness was evaluated using the
equation proposed by Tanaka9. At least 20 measurements of the crack length were
taken for each data, and the average and standard deviation were calculated.
3.1. Relative Density and Phases of Sintered Composites
The relative density of BaTiO3 -MgO composites sintered at 12008C decreased with
increasing BaTiO3 content, as shown in Fig. 1. At 13008C, the relative density of
100
Relative density (%)
90
80
1300⬚C
1200⬚C
70
0 5 10
BaTiO3 content (vol%)
Figure 1. Relative density of the BaTiO3 MgO composites sintered at 12008C and 13008C as a function of
BaTiO3 content.
composites was significantly improved and the reduction with increasing BaTiO3
content was not significant.
X-ray diffraction patterns of as-sintered BaTiO3 -MgO composites sintered at
13008C are shown in Fig. 2. All the peaks were assigned to MgO or tetragonal
BaTiO3 , and no reaction phase between MgO and BaTiO3 was detected. This
contrasts with the case of BaTiO3 -Al2 O3 composites, in which the third reaction
phases are obviously observed10. Thus, MgO was found to be compatible with
BaTiO3 for avoiding the reaction phase. Nagai et. al.11 reported that BaTiO2:977
phase was found for as-hot-pressed sample of MgO with 10 vol% BaTiO3 compos-
ites. They suggested that the appearance of the reduced BaTiO3 phase was due to the
reaction atmosphere during hot pressing. In this study, due to lower sintering
temperature and shorter time of the SPS sintering compared to the hot-pressing,
no BaTiO2:977 phase was found. However, from the high angle X-ray diffraction
analysis of BaTiO3 dispersions for BaTiO3 -MgO composites in the range of 2u
between 90.5 and 93 degree that Nagai et. al.12 also found, the peaks showed the
tetragonal distortion of BaTiO3 dispersoids. To confirm the tetragonal phase of
BaTiO3 dispersoids, Raman spectra of the nanocomposites were measured. They
concluded that the BaTiO3 dispersoids were tetragonal and ferroelectric phase.
Another their work13 by using Raman Spectroscopy exhibited the existence of
orthorhombic phase at room temperature for 0.5 and higher mol% MgO-doped
BaTiO3 (99.86 vol% BaTiO3 ) ceramics. In the present study, however, less than 10
vol% BaTiO3 content was dispersed in MgO matrix so that it was expected that there
would be tetragonal BaTiO3 and no orthorhombic phase in the present sintered
composites.
Figure 2. X ray diffraction patterns of various BaTiO3 MgO composites sintered by SPS at 13008C.
6
BaTiO3 grain
Average grain size (μm)

5 MgO grain
4
0
0 5 10
BaTiO3 addition (vol%)
Figure 3. Average sizes of BaTiO3 particles and MgO grains of the composites sintered at 13008C.
Figure 4. SEM photographs for the thermal etched surfaces of BaTiO3 MgO composites sintered by SPS
at 13008C: (a) pure MgO, (b) 1, (c) 3, (d) 5 and (e) 10 vol% BaTiO3 MgO.
3.2. Microstructure
Typical SEM images of the BaTiO3 -MgO composites are shown in Fig. 3. The white
BaTiO3 particles homogenously dispersed within the MgO matrix grains and at grain
boundaries. The average grain size of MgO matrix varied from 5:65 mm for the
monolithic MgO to less than 5 mm for the 10 vol% BaTiO3 added MgO, as shown
Figure 5. Hardness of the sintered composites as a function of BaTiO3 addition.
in fig. 3. The presence of BaTiO3 in MgO matrix seems to prevent grain boundary
migration and grain growth. However, amount of intergranular pores was increased
with increasing BaTiO3 addition.
With further increase in BaTiO3 content, some of BaTiO3 particles, which were
initially located at MgO grain boundaries, were agglomerated into larger particles
(fig. 4).
3.3. Hardness and Fracture Toughness of Sintered Composites

From figure 5, hardness of the BaTiO3 -MgO composites decreased with increasing
BaTiO3 content.
Fracture toughnesses of the MgO-based composites with BaTiO3 contents up to
3 vol% were low, compared with MgO monolithic. This may result from the decrease
in MgO grain size, which will be discussed in the next section based on the observa-
tions of crack path. With increasing BaTiO3 content higher than 5 vol%, higher
fracture toughness of BaTiO3 -MgO composites was achieved.
After polarization, fracture toughness of BaTiO3 -MgO composites was im-
proved, compared with that of the same samples before polarization. In addition,
fracture toughness parallel to the poling direction was higher than that perpendicular
to the poling direction for all of the composites (fig. 6). These results indicate that
piezoelectric effect of BaTiO3 plays a role in enhancement of toughness.
3.4. Toughening Mechanism

To investigate toughening mechanisms in the BaTiO3 secondary phase dispersed
MgO ceramic composite, indented crack paths were examined in detail. For mono-
lithic MgO ceramics, most of crack paths were along grain boundaries of MgO grains
and some of them cut across the grains, that is, intergranular manner of crack
propagation was dominant with some extent of transgranular manner, as shown
in Fig.7(a). The SEM micrographs of crack propagation paths for 1 and 3 vol%
BaTiO3 -MgO composites are shown in Fig.7(b) and (c), respectively. It can be seen
that cracks propagate dominantly in intergranular manner and also cut through
BaTiO3 particles. Rice13 reviewed the overall fracture mode of ceramics. He pro-
posed that fracture mode was mainly intergranular for finer grain ceramics and then
transferred to transgranular with increasing grain size for a broad range of stress
Figure 6. Fracture toughness of the sintered composites before and after polarization.
Figure 7. SEM micrographs of crack paths of (a) MgO monolithic and (b), (c) and (d) 1, 3 and 10 vol%
BaTiO3 MgO composites sintered at 13008C, respectively.
conditions. He noted that transgranular fracture of the matrix was dominant for
many particulate and whisker composites and also frequently of the dispersed
phase13. Further, transgranular fracture preferentially occurred in the larger grains,
as in monolithic ceramics, and the degree of transgranular fracture in the composite
matrix was typically higher than in the monolithic one with the same grain size13.
This was clearly shown by the work of Baek and Kim14, where the monolithic
alumina with grain size of around 6 mm predominantly showed intergranular frac-
ture, but with 20% SiC whiskers the alumina matrix with the same grain size showed
dominant transgranular fracture. Therefore, the low fracture toughness of 1 and
3vol% BaTiO3 -MgO composites resulted mainly from the finer grain size and the
lower hardness of the these composites. With further increase in BaTiO3 content,
transgranular crack path at or near BaTiO3 particle were increasingly observed, as

shown in Fig. 7(d). This change of crack path might contribute to higher fracture
toughness. This toughening mechanism was also found in LiTaO3 -Al2 O3 15 and
LaNbO4 -Al2 O3 composites16.
The existence of domain structures inside grains of LiTaO3 and LaNbO4 ce-
ramics confirmed the possibility of domain switching toughening mechanism15–16.
Domain structure like ripples was distinctly detected near the crack tip. Domain
switching was suggested as a new toughening mechanism in the ceramic composites.
In this work, fracture toughness of all the composites increased after polarization.
This indicated that domain switching improves fracture toughness of BaTiO3 -MgO
composites. The larger the BaTiO3 content was, the greater the fracture toughness
was, as expected. This toughening mechanism was decompensated by the decrease in
hardness and intergranular fracture mode for the composites with 1 and 3 vol%
BaTiO3 .
Experimental and theoretical analyses2,15–21 indicated that domain switching
plays an important role in improvement of toughness of ferroelectrics. The intensi-
fied stress in the vicinity of a crack-like flaw in the composite leads to BaTiO3
domain. The switched domains induce incompatible strain under the constraint of
unswitched matrix material, and consequently alter the stress distribution near the
flaw. Thus, the apparent toughness of BaTiO3 -MgO composites is improved due to
domain switching.
4. CONCLUSION
MgO based composites with addition of BaTiO3 up to 10 vol% were fabricated by a

spark plasma sintering method (SPS) with an applied load of 38 MPa. Nearly full
densification and high fracture toughness of BaTiO3 -MgO composites were achieved
by SPS method at 13008C. A high fracture toughness of around 1:81 0:26 MPa:m1=2
could be obtained by the addition of 10 vol% BaTiO3 before polarization. Moreover,
after polarization, fracture toughness of BaTiO3 -MgO composites remarkably in-
creased, since a domain switching toughening mechanism was induced in the present
composites. Domain switching of piezoelectric BaTiO3 particles may improve fracture
toughness of the BaTiO3 -MgO composites.
ACKNOWLEDGEMENT
The authors wish to thank the Macoh Co. for conducting spark plasma sintering,
and Prof. Uematsu’s laboratory, chemical engineering, Nagaoka University of Tech-
nology for preparing the mixed powders.
REFERENCES
1. T., Noma, S., Wada, M., Satake, T., Otsuka and T., Suzuki, Indentation Fracture of Poled Barium
Titanate Ceramics, in Proceedings of the Annual Meeting of the Ceramic Society of Japan (Yokohama,
Japan, April 1996). The Ceramic Society of Japan, Tokyo, Japan, pp.551(1996)
2. W., Yang and T., Zhu, Switch toughening of ferroelectrics subjected to electric fields. J. Mech. Phys.
Solids, 46 [2], 291 311 (1998)
3. K., Mehta and A. V., Virkar, Fracture mechanism in ferroelectric ferroelastic lead zirconate titanate
(Zr:Ti 0.54:0.46) ceramics, J. Am. Ceram. Soc., 73, 567 574 (1990)
4. C. S., Lynch, W., Yang, L., Collier, Z., Suo and R.M., McMeeking, Electric field induced cracking in
ferroelectric ceramics, Ferroelectrics, 166, 11 30 (1995)
5. T., Takagi, A Concept and Perspective of Intelligent Materials, Bull. Ceram. Soc. Jpn., 28 [6], 539 544
(1993)
6. M.P., Hammer and R.J., Brook, Fast firing microstructural benefits, J. Br. Ceram. Soc., 80, 147 149
(1981)
7. M., Tokita, Trends in advanced SPS spark plasma sintering system and technology. J. Soc. Powder
Technol., Jpn., 30 [11], 790 804 (1993)
8. L., Gao, Z., Shen, H., Miyamoto and M., Nygren, Superfast densification of oxide/oxide ceramic
composites. J. Am. Ceram. Soc., 82[4], 1061 1063 (1999)
9. K., Tanaka, Elastic/Plastic Indentation Hardness and Indentation Fracture Toughness: the Inclusion
Core Model, J. Mater. Sci., 22, pp.1501 (1987)
10. S., Rattanachan, Y., Miyashita and Y., Mutoh, Microstructure and fracture toughness of a spark
plasma sintered Al2 O3 based composite with BaTiO3 particulates, Journal of the European Ceramic
Society, 23[8], 1269 1276 (2003)
11. T., Nagai, H.J., Hwang, M., Yasuoka, M. Sando and K., Niihara, Preparation of a Barium Titanate
Dispered Magnesia Nanocomposite, J. Am. Ceram. Soc., 81[2], 425 428 (1998)
12. T., Nagai, H.J., Hwang, M., Sando and K., Niihara, Preparation of Ceramic Nanocomposite with
Perovskite Dispersoid, Mat. Res. Soc. Symp. Proc., 457, 375 380 (1997)
13. T., Nagai, K., Iijima, H.J., Hwang, M., Sando, T., Sekino and K., Niihara, Effect of MgO Doping on
the Phase Transformations of BaTiO3 , J. Am. Ceram. Soc., 83[1], 107 112 (2000)
14. R.W., Rice, Ceramic fracture mode Intergranular vs. Transgranular Fracture. Ceramic Transec
tions, 64: Fractography of glasses and Ceramics III (JR Varner, VD Frechette, and GD Quinn, eds).
Am. Cer. Soc., Westerville, OH, 1996, p.1 53.
15. Y.K., Baek and C.H., Kim, The Effect of Whisker Length on the Mechanical Properties of Alumina
SiC Whisker Composites, J. Mat. Sci., 24, 1589 1593(1989)
16. Z., Zhang, L., Zhou, Y., Hu and L., Jiang, Preparation and characterization of Al2 O3 LaNbO4
composites, Scripta Materialia, 47, 637 641 (2002)
17. Y.G., Liu, Y., Zhou, D.C., Jia, Q.C., Meng and Y.H., Chen, Domain switching toughening in a
LiTaO3 dispersed Al2 O3 ceramic composite, Scripta Materialia, 47, 63 68 (2002)
18. G.G., Pisarenko, V.M., Chushko and S.P., Kovalev, Anisotropy of Fracture Toughness of Piezo
electric Ceramics, J. Am. Ceram. Soc., 68[5], 295 265 (1985)
19. S.B., Park and C.T., Sun, Effect of electric field on fracture of piezoelectric ceramics, Inter. J. Fract.,
70, 203 216 (1995)
20. A., Kishimoto and S., Seo, Strength control of a ceramic composite by electric field, Smart Materials,
4234, 321 327 (2001)
21. W., Lu, D.N., Fang and K.C., Hwang, Micromechanics of ferroelectric domain switching behavior
Part I: Coupled electromechanical field of domain inclusions, Theor. App. Fract. Mechan., 37, 29 38
(2001)
FRAGMENTATION OF
CERAMICS IN RAPID
EXPANSION MODE
Spandan Maiti, Philippe H. Geubelle,

and Krishnan Rangaswamy1
1. INTRODUCTION
The study of the fragmentation process goes back to more than a century, motivated
primarily by problems related to mining and ore handling (Grady and Kipp, 1985).
Various theories have been proposed to predict the fragmentation stress and the
fragment size and distribution. But the investigations are generally case specific and
relate to only a narrow set of fragmentation processes. A number of theoretical
studies of dynamic fragmentation in a rapidly expanding body can be found in the
literature. For example, the study summarized in (Grady, 1982) presents a model
based on a simple energy balance concept between the surface energy released due to
fracture and the kinetic energy of the fragments. Subsequent refinements of the
energy balance model have been proposed by (Glenn and Chudnovsky, 1986),
which take into account the strain energy of the fragments and specify a threshold
stress below which no fragmentation occurs. These models assume that the fracture
events are instantaneous and occur simultaneously. Evidently, these assumptions are
quite restrictive and these models cannot take into account the transient nature of the
fragmentation process after the onset of fracture in the material. A more recent
model proposed by (Miller et al., 1999) however takes into account this time-
dependent nature of the fragmentation event and the distribution of flaws of various
strengths in the original material.
1
Department of Aeronautical and Astronautical Engineering, University of Illinois at Urbana
Champaign, Urbana, USA.
353
354 Spandan Maiti et al.
Attempts to model the fragmentation phenomenon numerically are quite recent

(Camacho and Ortiz, 1996, Miller et al., 1999), and some success has been reached
through the use of cohesive modeling. Although the cohesive finite element models
used so far have been shown to be effective in simulating the dynamic crack propa-
gation in the material, they do not take into account the effect of the microstructure
inherent to ceramic materials. An exception can be found in the recent work by
(Espinosa et al., 1998), although the emphasis of this work is less on the fundamental
problem of fragmentation and more on the numerical modeling of rod on rod impact
experiments. Also, the cohesive models are sensitive to the spacing and size of the
cohesive surfaces making convergence difficult to achieve.
In this paper, we study the fundamental fragmentation problem of a specimen
with an inherent granular microstructure that is subjected to a rapid expansion. Of
special interest are the effects of the average grain size, initial strain rate and cohesive
failure model parameters (strength and toughness) on the fragmentation process.
2. NUMERICAL SCHEME
The key elements of the numerical scheme used in this study are its ability to
incorporate the granular microstructure of the ceramic material, to simulate the
spontaneous initiation, propagation and branching of intergranular cracks and
subsequent fragmentation of the body, to account for inertial and finite kinematics
effects and to capture the complex contact events taking place between the fragments.
To simulate the spontaneous dynamic motion of cracks, we use a 2-D plane
strain cohesive/volumetric finite element (CVFE) scheme that has shown great suc-
cess in the simulation of various dynamic fracture events in brittle media (Xu and
Needleman, 1994, Camacho and Ortiz, 1996, Geubelle and Baylor, 1998). As its name
indicates, the numerical method is based on the combination of conventional (volu-
metric) elements used to model the mechanical response of the ceramic material
and interface (cohesive) elements introduced to simulate the intergranular failure pro-
cess. Figure 1 schematically illustrates the CVFE concept. The volumetric elements
Figure 1. Schematic of the grain based CVFE scheme, showing three grains discretized with 6 node
volumetric elements and linked with 6 node cohesive elements placed along their boundary. The cohesive
elements are shown in their deformed configuration: they initially have no thickness and adjacent nodes are
superposed.
used in the present study are six-node triangular elements, with a constitutive re-
sponse described by a linear isotropic relation between the second Piola-Kirchhoff
stress tensor S and the Lagrangian strain tensor E (Geubelle and Baylor, 1998). A
nonlinear kinematic description is used here to allow for possible large rotations
associated with the fracture event. The cohesive elements placed along the grain
boundaries are characterized by a bilinear rate-independent failure law relating the
displacement jump vector (D) to the cohesive traction vector (T ) acting across the
cohesive surfaces Gc :
S Dn S Dt
Tn ¼ smax , Tt ¼ t max , (1)
1 S Dnc 1 S Dtc
where the subscripts n and t denote normal and tangential components, respectively,
smax and t max are the tensile and shear failure strengths, while Dnc and Dtc are the
critical opening and shear displacement jumps. The evolution of the damage process
is quantified by the monotonically decreasing damage parameter S defined as (Geu-
belle and Baylor, 1998, Bi et al., 2002)

S ¼ min Smin , max(0, 1 kD~ k) , (2)
~ denotes the normalized displacement jump vector

where D

~¼ D ~n Dn =Dnc
D ~t ¼ : (3)
D Dt =Dtc
As the grain boundary fails, the value of S gradually decreases from an initial value
close to unity to zero, point at which complete failure is acheived. The coupling
between normal and tangential failure in mixed-mode situations is captured by
defining S in terms of the L2 -norm of the normalized displacement jump vector D ~,
as indicated by Equation 2 and illustrated in Figure 2.
To prevent overlapping of adjacent grains, the damage parameter S is kept to its
initial value Sinit close to unity when the normal displacement jump Dn becomes
negative. As illustrated in Figure 2(a), this approach results in very high repulsive
normal traction along the contacting surfaces. This simple explicit contact enforce-
ment scheme along with intermittent use of more sophisticated optimization-based
Figure 2. Cohesive failure model: variation of the normal (a) and tangential (b) components of the
cohesive traction vector T with respect to the normal (Dn ) and shear (Dt ) crack opening displacements,
showing the coupling between tensile and shear failure.
contact algorithm designed for angular fragments (Kane et al., 1999) is appropriate
for the present study dedicated to fragmentation due to tension. Once the fragmen-
tation process is complete, the overlap of the fragments does not affect the phenom-
ena we are investigating, so no contact algorithm is invoked at this stage.
The finite element implementation is based on the principle of virtual work.
Since the capture of the dynamic fracture process requires very small time steps, an
explicit central difference time stepping scheme is used in this investigation, with the
time step size Dt chosen as 4% of the limiting (Courant) value.
To generate the granular microstructure, a two-step discretization process is
adopted. The grains are first created through a Voronoi tessellation of the ceramic
specimen, leading to a relatively uniform size distribution. The grains are then
discretized with 6-node triangular elements by Delaunay triangulation and 6-node
cohesive elements are inserted along the grain boundaries. For more detail about the
mesh generation, see (Maiti and Geubelle, 2003). To calculate the number of frag-
ments, we have generated a graph with each grain in the domain as a node, and have
computed the connected components of the graph. A breadth-first algorithm for
finding the spanning forest of a graph (Nijenhuis and Wilf, 1978) is utilized for this
purpose.
3. PROBLEM DESCRIPTION
Figure 3 depicts the specimen used in our fragmentation study. The specimen is
rectangular with the upper and lower boundaries subjected to a constant velocity V.
We assume that the specimen is in a state of rapid expansion just after an impact and
thus have a constant initial strain throughout the domain. For this reason, we subject
all the points of the computational domain to a linear initial velocity distribution
with the magnitude equal to V at the ends of the specimen, as illustrated in Figure 3.
The material used for all the simulations is alumina, whose properties are as
follows: Young’s modulus E ¼ 400 GPa, Poisson’s ratio n ¼ 0:27 and density
Figure 3. Geometry and initial and boundary conditions of the test specimen used in the fragmentation
study.
r ¼ 3800 kg=m3 . The mode I and mode II fracture toughnesses are taken to be the
same: 69:5 J=m2 . The cohesive strength in the normal (smax ) and tangential (t max )
directions are always chosen to have the same magnitude. For most of the simula-
tions presented hereafter, this value is varied between 0.1 and 4 GPa. The initial value
of the damage parameter Sinitial is chosen as 0.98. Five average grain sizes are chosen
for this study: 120, 100, 80, 60, and 40 mm, yielding a total of 187, 227, 369, 636 and
1406 grains, respectively. Different grain distributions for a single average size are
also examined. Figure 4 shows a sample of these different grain size distributions.
In the discussion that follows, we repeatedly use the term damage index (DI).
This parameter is defined as the ratio of failed cohesive elements to the total number
of cohesive elements and is expressed in percent. As the length of all cohesive
elements is almost the same, this parameter gives a good indication of the extent of
fracture in the specimen.
4.1. General Features
Let us start the result discussion by studying the general features of the fragmenta-
tion process for a specimen with an average grain size of 60 mm and a failure strength
of 1 GPa subjected to an initial strain rate 2_ of 0:5 105=s. The evolution of the DI is
presented in Figure 5 for the grain distributions shown in Figure 4(b), (c) and (d). As
apparent in Figure 5, the fragmentation process is not instantaneous but initiates at
about 0:165 ms to end at approximately 0:25 ms. While some differences are noted
between the three grain distributions giving some measure on the variability of the
Figure 4. Granular microstructure of the sample with an average grain size of (a) 40 mm, (b d) 60 mm and
(e) 80 mm. For figures (b), (c) and (d), the average grain size is the same but the distributions are different.
Figure 5. Evolution of the damage index with time for a specimen with an average grain size of 60 mm
with the three distributions shown in Figures 4(b) (d).
results on the microstructure, the overall behavior is very similar between the three
cases. In the remainder of this paper, we use the microstructure shown in Figure 4(b).
As shown in Figure 6(a), the loading rate has a very strong influence on the
initiation of fragmentation and on the final value of the damage index, i.e., on the
extent of fragmentation. As 2_ increases, the fragmentation process starts sooner and
leads to more extensive damage. The dependence of initiation time on the initial
strain rate is summarized in Figure 6(b) and is qualitatively similar to that presented
in (Miller et al., 1999).
The evolution of the fragmentation process for a specimen with an average grain
size of 60 mm and a cohesive strength of 1 GPa subjected to 2_ ¼ 3 105 =s is illus-
trated by the four snapshots shown in Figure 7. Initially, the fragment size is
relatively large and, for values of DI less than 50%, there are only a few fragments.
As time progresses, however, the cracks coalesce and the number of fragments
increases rapidly.
Figure 6. Effect of the initial strain rate 2_ on the evolution of the damage index (DI) (a) and on the
initiation time (b) for a ceramic specimen with an average grain size of 60 mm and a cohesive strength of
1 GPa.
Figure 7. Snapshots of the fragmentation process for a specimen with an average grain size of 60 mm at (a)
t 0:0116 ms(DI 5%), (b) t 0:0122 ms(DI 20%), (c) t 0:0168 ms(DI 63%) and (d) t 0:1985
ms(DI 79%). The cohesive strength is 1 GPa and 2_ 3 105 =s.
To quantify the evolution of the fragment size during the fragmentation process,
we present in Figure 8 the cumulative distribution of fragment areas at three instants
during the failure process for which DI ¼ 63,70 and 79%. As apparent in that figure,
the curves shift to the left and become steeper as fragment size decreases. The
numerical distribution curve corresponding to DI ¼ 79% is compared in Figure 9
to two models available in the literature: the so-called Mott distribution (Mott, 1947)
and the linear exponential distribution (Grady and Kipp, 1985), showing better
Figure 8. Cumulative distribution of fragment areas at (a) DI 63%, (b) DI 70% and (c) DI 79% for
the case shown in Figure 7.
Figure 9. Fit of linear exponential distribution and Mott distribution for the curve DI 79% shown in
Figure 8.
agreement with the latter. This agreement with the linear exponential distribution is
observed throughout the fragmentation process after the specimen starts breaking
into multiple fragments.
4.2. Effect of Cohesive Parameters

To study the effect of the cohesive strength on the fragmentation process, a series of
simulations have been performed with a range of cohesive strength values (from 0.10
to 4 GPa) but keeping the fracture toughness unchanged in all the cases. The average
grain size is 60 mm and we assume the tensile (smax ) and shear (t max ) cohesive
strengths to be equal.
Figure 10 shows how the cohesive strength affects the final value of the damage
index, i.e., the final extent of fragmentation. Note that the failure of all cohesive
Figure 10. Variation of final damage index with cohesive strength for different initial strain rates. The
average grain size is 60 mm.
surfaces is observed for high values of the strength (superior 2 GPa). It can thus be
concluded that, for truly brittle materials with low fracture toughness and high
fracture initiation stress, the fragmentation does not depend on the initial strain
rate (Glenn and Chudnovsky, 1986) as the fragmentation process is always complete.
The results also indicate that, as the cohesive strength is reduced, the relative portion
of failed cohesive elements experiences first a sharp drop followed by a plateau for
small value of the strength (less then 0.5 GPa). Also, as expected, final DI increases
as the initial strain rate is increased, though the curves retain their general trend. The
presence of the drop is to be expected: as the cohesive strength is decreased, the
material reaches the fracture initiation point at earlier time for the same initial strain
rate. The material has thus less time to store energy to be expended during the
fracture event. But it is interesting to note the existence of a strain rate dependent
lower bound for the final damage index for lower cohesive strengths.
The variation of the initiation time of the fragmentation process (i.e., time at
which the first cohesive failure is detected) with respect to the cohesive strength is
presented in Figure 11 for constant values of the fracture toughness. For higher
values of the cohesive strength, the initiation time decreases with decreasing cohesive
strength since the material points in the specimen reach the fracture initiation stress
sooner. For smaller cohesive strengths, however, the trend is reversed and the
initiation time increases with decreasing strength reflecting the corresponding in-
crease of the critical displacement jump values Dnc and Dtc . As shown in Figure 11, an
increase in «_ results in a decrease of the initiation time while keeping the nature of the
curves same.
4.3. Effect of Grain Size
In this section we study the effect of the average grain size on the fragmentation
process. For that purpose we have chosen specimens with five different grain sizes:
40, 60, 80, 100 and 120 mm. First, we present the effect of average grain size on the
Figure 11. Variation of fragmentation initiation time with cohesive strength for different initial strain
rates. The average grain size is 60 mm and the fracture toughness is kept constant.
initiation time of fragmentation. As observed in Figure 12, the grain size does not
affect the initiation time, especially for high values of the initial strain rate «_ . As
the grain size decreases, the density of cohesive elements increases. But since the
cohesive strength is kept constant, the time required for the cohesive traction to reach
the critical value and initiate the failure process does not depend significantly on the
number of available fracture paths.
The average grain size has however a strong influence on the evolution and
extent of the fragmentation process, as illustrated in Figure 13, which presents the
evolution of the damage index for the same five specimens as those used in Figure 12.
As noted earlier, the initial phase of the fragmentation process is independent of the
microstructure. However, the grain size affects the final value of the damage index,
which increases as the grains become larger. This trend can be explained by the fact
that the ratio of energy required for fragmentation to the energy available varies with
the inverse of grain size, and hence the extent of final fragmentation increases with an
Figure 12. Variation of the initiation time with initial strain rate for different average grain sizes. The
cohesive strength smax tmax 1:0 GPa.
Figure 13. Evolution of damage index for different average grain size of the material. Cohesive strength
smax tmax 1:0 GPa and 2_ 2 105 =s.
increase of average grain size. Note that, for higher cohesive strengths (in excess of
2 GPa), the damage index reaches 100% for all different grain sizes as the strain and
kinetic energy stored in the system prior to the failure process is higher due to
increased fragmentation initiation time.
5. CONCLUSIONS
In this paper we have studied the fragmentation process of a rapidly expanding bar
using a numerical scheme relying on a grain-based mesh generation technique and a
cohesive/volumetric finite element method which is able to capture dynamic inter-
granular fragmentation events. We have observed that the fragmentation process is
not instantaneous and that the initiation time decreases drastically with the increase
in initial strain rate. Also, we have shown that the cumulative frequency of the
fragment area can be estimated reasonably by a linear exponential distribution
throughout the fragmentation process.
6. ACKNOWLEDGEMENTS
This research project is funded by NSF through the Career Award CMS-9734473.
The grain-based finite element discretizations have been obtained with the help of the
mesh generation software TRIANGLE originally developed by J. R. Shewchuk.
REFERENCES
Bi, X., Li, Z., Geubelle, P. H., and Lambros, J. (2002). Dynamic debonding and frictional push out in
model composite systems: numerical analysis. Mech. of Mater., 34:433 446.
Camacho, G. T. and Ortiz, M. (1996). Computational modeling of impact damage in brittle materials. Int.
J. Solids Structures, 33(20 22):2899 2938.
Espinosa, H. D., Zavattieri, P. D., and Dwivedi, S. K. (1998). A finite deformation continuum/discrete
model for the description of fragmentation and damage in brittle materials. J. Mech. Phys. Solids,
46(10):1909 1942.
Geubelle, P. H. and Baylor, J. (1998). Impact induced delamination of composites: a 2 D simulation.
Composites B, 29:589 602.
Glenn, L. A. and Chudnovsky, A. (1986). Strain energy effects on dynamic fragmentation. J. Appl. Phys.,
59:1379 1380.
Grady, D. E. (1982). Local inertial effects in dynamic fragmentation. J. Appl. Phys., 53:322 325.
Grady, D. E. and Kipp, M. E. (1985). Geometric statistics and dynamic fragmentation. J. Appl. Phys.,
58:1210 1222.
Kane, C., Repetto, E. A., Ortiz, M., and Marsden, J. E. (1999). Finite element analysis of nonsmooth
contact. Comp. Meth. Appl. Mech. Engrg., 180:1 26.
Maiti, S. and Geubelle, P. H. (2003). Mesoscale modeling of dynamic fracture of ceramic materials. To
appear in Comp. Meth. Engr. Sci.
Miller, O., Freund, L. B., and Needleman, A. (1999). Modeling and simulation of dynamic fragmentation
in brittle materials. Int. J. Fracture, 96:101 125.
Mott, N. F. (1947). Fragmentation of shell cases. Proc. Royal Soc., A189:300 305.
Nijenhuis, A. and Wilf, H. S. (1978). Combinatorial Algorithms For Computers and Calculators. Academic
Press, 2nd edition.
Xu, X. P. and Needleman, A. (1994). Numerical simulation of fast crack growth in brittle solids. J. Mech.
Phys. Solids, 42:1397 1434.
INCREASING RESISTANCE
TO LOW TEMPERATURE
AGEING DEGRADATION
OF Y-TZP BY SURFACE
MODIFICATION
Adrian Feder, Paulina Morcillo, and Marc J. Anglada*
1. INTRODUCTION
The high strength and fracture toughness of yttria doped tetragonal zirconia poly-
crystals (Y-TZP) make these materials very interesting for some engineering appli-
cations. These properties depend on the stability of the different polymorphs of
zirconia, which are very sensitive to the quantity and type of stabilizing oxide that
is used to decrease the temperature of the tetragonal to monoclinic (t-m) phase
transformation. This transformation is martensitic in nature and is accompanied
by an increase in volume of nearly 4%. When such transformation becomes active
under the high stress field that exists in front of the crack tip, the constrained
transformation strain leads to a significant increase in the fracture toughness.1
The extent of the transformation at the crack flanks is an important parameter,
which depends on temperature, the amount and type of stabiliser, and the grain size
of tetragonal zirconia. The maximum capability of a material to experience trans-
formation toughening can be expressed as a shielding stress intensity factor, Ks ,
which is a function of the transformation zone shape, as evaluated by the parameter
a, and the zone size, h0 . It may be expressed as
*
Department of Materials Science and Metallurgical Engineering, Universitat Politcnica de Catalunya,
ETSEIB, Avda. Diagonal 647, 08028 Barcelona, Spain.
625
626 Adrian Feder et al.
p
KS ¼ aeT Efv h0 =(1 n) (1)
where eT is the unconstrained dilatational transformation strain, E Young’s modu-

lus, n Poisson’s ratio, and fv the volume fraction of transformed tetragonal phase
within the zone.2
Unfortunately, this transformation may also take place spontaneously with the
presence of water vapour during ageing at low temperatures (from about 65 to 400
8C) and the phenomenon is referred to as low temperature ageing degradation of
Y-TZP. The volume expansion that accompanies the transformation can generate
microcracks in the transformed surface, thereby degrading the strength and surface
properties of the material [see, for instance, ref. 3].
It is now well established that low temperature ageing degradation is character-
ised by: a) the transformation proceeds from the surface to the interior; b) when
degradation occurs in air, from all constituents of air, only water vapour is respon-
sible; c) both, increasing the amount of oxide stabiliser and decreasing the grains size,
can retard the degradation.
Although much research has been performed, the exact mechanism responsible
for the spontaneous transformation is not still clear. However, one strategy to
increase the resistance of Y-TZP to low temperature ageing degradation is to act
upon the microstructure of the material. Along these lines, different methods have
been developed such as reduction of grain size, introduction of inert materials, and
the increase of either the cation-stabiliser content, or the introduction of an anion-
stabiliser such as occurs in nitridation.4 The operating mechanism in surface nitrida-
tion is the formation of vacancies in the unit-cell of zirconia by high temperature in
the presence of nitrogen. This element stabilises the cubic phase of zirconia (c-ZrO2 )
by virtue of a partial substitution of oxygen ions (O2 ) by nitrogen ions (N3 )
causing in this way the desired vacancies.5 This mechanism is identical to the one
that allows stabilising the tetragonal and cubic phases of ZrO2 by the addition of bi-
and trivalent cations of, for instance, Mg2þ and Y3þ .
The strength, fracture toughness and stability of nitrided 2.5 mol% of Y-TZP
has been recently analysed in air, and also after being in contact with water at 100 8C
for long times6. The main finding of this work is that a protective surface nitrided
layer of non-transformable t’-phase is formed, which is resistant to degradation by
water. However, below the t’ layer, large tetragonal grains are stabilised by nitrogen
and these are easily transformed into the monoclinic structure. The result is that
nitrided Y-TZP suffers spontaneous transformation to monoclinic structure by
heating to temperatures of about 600 8C in air. This has been associated to the lost
of nitrogen, so that the large tetragonal grains become unstable.6
In the present paper, it has been studied the feasibility of protecting zirconia
from the environment by means of an alumina protective layer produced via a high
temperature solid state reaction between zirconia and aluminium nitride in a nitrogen
environment. It is shown that the coated Y-TZP that results has an excellent com-
bination of toughness and strength, and also a much higher resistance to environ-
mental slow crack growth. Finally, it is verified that the coated Y-TZP does not
suffer from low temperature ageing degradation at 250 8C in air after times of
exposure longer than 750 hours.
Increasing Resistance to Low Temperature Ageing Degradation 627
The starting material examined in this work is a zirconia stabilised with 2.5% mole of
Y2 O3 (2.5Y-TZP). Its microstructure consists of tetragonal grains with an average
size of 0:3 mm. The original cylindrical bars of 8 mm in diameter were cut in slices of
about 1 mm in thickness. The resulting discs were polished with diamond pastes of 30
and 6 mm and then heat treated at 1650 8C during 2 hours either in nitrogen with the
surface covered with a fine-grain powder of AlN, or just only in air, and the resulting
materials are referred to as 2HA and 2H, respectively. The heating and cooling rates
applied were both of 20 8C/min.
The microstructural evolution resulting from such treatments was assessed
through X-ray diffraction, scanning electron microscopy, and Raman microprobe
spectroscopy. The fraction of monoclinic phase in the surface of the samples was
estimated by the first technique, and the last one was used to calculate the monoclinic
fraction present in different places of the fracture surfaces with the help of an
equation developed in prior work by calibration with X-ray data.7
Microhardness tests were conducted to evaluate the variation in hardness with
depth by using an indentation load of 100 g. In addition, Vickers indentations were
performed to induce cracks on the surface of the discs by using a load of 294 N and
with the objective of estimating the fracture toughness. In prior work, it has been
demonstrated that the geometry of cracks introduced by Vickers indentation in 2.5 %
Y-TZP has a Palmqvist-type shape.8 Therefore, the fracture toughness KIC was
evaluated by using the following equation,9
p
KIC ¼ A(E=HV )2=5 (P=d l ) (2)
where is the Young’s modulus, HV the hardness, P is the applied indentation load, l
is the length of the crack, d the semi-diagonal of the indentation imprint, and A is a
constant equal to 0.0076.
The mechanical behaviour of the discs has been studied in a miniaturised ball-
on-three-balls flexure-testing device. The strength of the discs, smax , can be obtained
by using the equation proposed by Börger et al:10
smax ¼ ( fPmax )=t2 (3)
where Pmax is the fracture load, t the thickness of the discs, and f is a calibration
factor given by the following expression:
f(t=R, Ra =R, n)¼c0 þ{[c1 þc2 (t=R)þc3 (t=R)2 þc4 (t=R)3 ]=[1þc5 (t=R)]}[1þc6 (Ra =R)]
(4)
10
The parameters ci (i¼1...6) depend on the coefficient of Poisson , R is the diameter
of the discs and Ra is the diameter of the circle defined by the centres of the three
balls. In the present case, R and Ra are 4 and 3 mm, respectively, and the diameter of
the balls is 2 mm. The tests were carried out in air at two different constant load rates
of 200 y 10 N/s with the aim to assess the influence of the laboratory environment on
the strength of the heat treated materials. In order to evaluate the resistance to low
temperature degradation, the discs were subjected to ageing in air at 250 8C for times
longer than 750 hours.
3 RESULTS AND DISCUSSION
3.1. Microstructure
The microstructures of the starting material, 2.5Y-TZP, and of 2H are depicted in
Fig. 1. The analysis by XRD (Fig. 2) confirms that the original material is formed by
only tetragonal grains, while in 2H there is a duplex microstructure of tetragonal and
cubic grains with average size much larger than the original tetragonal grains (see
Table 1). For 2HA, it appears a surface layer of approximately 3:9 mm in thickness
(Fig. 3) as a result of the high temperature reaction of Y-TZP with AlN powder. The
analysis by XRD of the surface of 2HA shows the existence of Al2 O3 , as well as t and
c-ZrO2 .
According to Bayern and Mocellin11, a solid state displacement reaction takes
place in the ZrO2 ---AlN system at the temperature studied:
c
1μm 1μm
(a) (b)
Figure 1. Micrographs showing microstructures of Y TZP material (a) and Y TZP heat treated at 1650
8C during 2 hours in air (b); t and c refer to tetragonal and cubic grains respectively.
33000
a : Al2O3
t: tetragonal - ZrO2
32000
c : cubic - ZrO2
Intensity (arbitrary counts)
31000
2HA
30000 a
a a
a a a aa a
c+t
29000
c+t
c+t
c+t
2H
c+t
10000
c+t
5000
t
2.5Y-TZP
t
0
tt t
t tt t tt
20 30 40 50 60 70 80
2θ(⬚)
Figure 2. X ray diffraction patterns of the surface of the materials investigated.

Table 1. Layer thickness and grain size of the materials studied.
Material Layer Thickness (mm) Dg Surface (mm) Dg Interior (mm)
2.5Y TZP 0.3 0.1 0.3 0.1
2H 1.8 0.1 t ZrO2 1.8 0.1 t ZrO2

4.4 0.1 c ZrO2 4:4 0:1c ZrO2
2HA 3.9 0.4 16 4 1.8 0.3 t ZrO2

4.8 0.1 c ZrO2
t+c – ZrO2 Al2O3
2μm
Figure 3. Microstructure of a cut perpendicular to the surface of 2HA.
3ZrO2 þ 4AlN ¼ 3ZrN þ 2Al2 O3 þ 0:5N2 (1)
However, we have not detected ZrN, but only Al2 O3 , and t- and c-ZrO2 . For times of
reaction of 1 hour, from X-ray diffraction measurements we have observed the
presence of ZrAlON, ZrN, Al2 O3 , and ZrO2 , but for 2 hours, the first compound
could not be detected. It should be added that ZrAlON, was also detected by Toy and
Savrun12, which proposed a different reaction between ZrO2 and AlN:
4ZrO2 þ 5AlN ¼ 3ZrAlON þ Al2 O3 þ ZrO2 þ N2 (2)
The reason for these different findings with respect to our results may lie in the fact
that in the present experiments there is a large amount of unreacted zirconia that
may be a large source of oxygen to oxidise either ZrN or ZrAlON after long times at
1650 8C.
The microstructure of 2HA below the surface layer is practically identical to that
originated at the same temperature but only in contact with air (2H). This has been
previously described in detail by Casellas et al8. It is worthwhile noticing that under
these treatments it is possible to retain grains of t-ZrO2 of average size as large as
3 mm at room temperature.
3.2. Hardness and Fracture Toughness

The microhardness along the depth of the specimens does not change, and it is very
similar for all materials. On the surface, the influence of the alumina layer on the
value of the microhardness measured is negligible, because of the relatively high

load used and the small thickness of the layer. In the interior of the material, the
microhardness for 2H and 2HA are identical, as expected because of similar micro-
structures.
From Table 2 it can be seen that the length of the indentation cracks nucleated
at the surface of 2H and 2HA are much shorter that in Y-TZP, and this corresponds
to an increase in KIc of more than 30 % in the former materials. This value of fracture
toughness is related to the microstructure developed below the surface layer in 2HA.
That is, KIc is not significantly affected by the possible influence of residual stresses
in the layer, because similar value of fracture toughness is obtained when the surface
thin layer is removed by polishing. Then, the increase in fracture toughness in 2HA
with respect to Y-TZP is also due to the more transformable duplex microstructure
that develops during the high temperature reaction. This is confirmed by Raman
spectroscopy analysis of the fracture surfaces where there are zones largely trans-
formed (up to 50 % of m-ZrO2 ), while in 2.5Y-TZP the amount of transformation is
very limited.
The nominal expected residual stress in the alumina layer can be estimated by
considering that it is very much thinner than the substrate of stabilised zirconia, so
that a simplified expression for the biaxial stress (sA ) developed in the film can be
used:
sA ¼ EA (aA aZ )DT=(1 nA ) (5)
where the average expansion coefficients of alumina and zirconia, aA and aZ , are taken
as 11.0 and 8:91 10 6 K, respectively, and the bulk elastic modulus, EA , and Poisson’s
ratio, nA , of alumina used here are 400 GPa and 0.23, respectively. By using these bulk
values for these properties, the value of sA induced by a change in temperature from
1650 8C to room temperature can be estimated, and this gives a value of about 1.6 GPa.
Since this residual tensile stress is concentrated in a very thin layer at the surface, its
contribution to the residual stress intensity factor of an indentation semicircular crack,
which is comparatively much larger, can be estimated as,
p
KR bsA s= l (6)
where b is a factor close to unity, and s is the thickness of the alumina layer. One can
see that KR is appreciably smaller than the fracture toughness, since equation (3)
gives values for KR of about 0.5 MPa m1=2 . This explains that not significant
differences are found for the fracture toughness of 2HA and 2H.
Table 2. Crack lengths (l) and fracture toughness (KIC ) measured by indentation at the
surface and in the interior of the discs.
Material 1 (mm) surface KIC (MPa m1=2 ) surface 1 (mm) interior KIC (MPa m1=2 ) interior
2.5Y TZP 174 2 4.9 0.1
2HA 98 8 6.5 0.2 102 4 6.4 0.1
2H 110 10 6.2 0.4

3.3. Biaxial Fracture Strength
The biaxial fracture strength (smax ), measured by a miniaturised ball-on-three-balls

flexure-testing device and by using equation (3) and (4), is presented in Fig. 4 for two
different loading rates, of 10 and 200 N/s. It can be noticed that the strength for
the highest loading rate is hardly different for the different materials. However,
for the lower loading rate there is a very clear decrease in strength for Y-TZP as
well as for 2H. On the contrary, 2HA keeps its strength constant for both loading
rates and this is an indication of a stronger resistance to environmental slow crack
growth, which may be attributed to the alumina surface layer.
Fractographic analysis of the broken specimens shows important differences in
the fracture path for the different conditions. Thus, the crack path is completely
intergranular in Y-TZP, while in 2H and 2HA there is a mixture of intergranular
(tetragonal grains) and transgranular (cubic grains) crack paths.
It should be noticed that the strengths of 2H and 2HA are similar to that of
Y-TZP in spite of the much smaller grain size of the latter. A simple explanation for
this can be given by looking to the values of strength and fracture toughness of Y-TZP,
2H and 2HA. Using the values for these parameters given in Table 2 and in Fig.4, if R-
curve effects are neglected and the plateau toughness is used in the calculations, it is
possible to estimate that the critical crack sizes for Y-TZP and 2H are about 15 mm and
25 mm, respectively. This difference in the critical defect sizes may exist since the grain
size has increased during exposure at 1650 8C by a factor of near 10. However, at least
in Y-TZP, it is clear that the origin of fracture are processing defects, such as porous.
Then, if this was still true for 2H, so that no significant increase in defect sizes occurred
at 1650 8C, there would be only two possibilities to explain the larger critical defect size
in 2H. One is that natural critical cracks are too small for developing plateau fracture
toughness before fracture, and the other possibility is that transformation takes place
locally before fracture at stress concentrations.
Another point worthwhile considering is the influence of residual stresses in the
surface layer on the strength of 2HA. Because of the difference in thermal expansion
coefficients of alumina and zirconia, tensile stresses should be present in the alumina
layer. However, it seems that these do not have an important effect on the strength.
1400
200 N/s
10 N/s
1200
σmax(MPa)
1000
800
600
400
Y-TZP 2H 2HA
Figure 4. Biaxial fracture strength of different materials investigated.

The reason lies in that possible defects in the alumina layer are necessarily much
smaller than the critical defect size for 2H so that the residual stress should be
extremely high to have a significant effect on the strength. It is important to
emphasise that in 2HA, besides modifying the surface by coating, there is an increase
in the fracture toughness whose origin is related to an increase in the fracture
toughness of the bulk material.
4. LOW TEMPERATURE AGING DEGRADATION
The change in biaxial strength of discs of 2.5Y-TZP, 2HA and 2H is shown in Fig. 5
in terms of time for ageing at 250 8C in air. It can be observed that for both uncoated
materials there is a reduction of smax with ageing time, and this is very strong for 2H.
On the contrary, the strength of 2HA does not change at all during the range of time
studied, in spite that 2HA, in the bulk, has a similar microstructure as 2H. It is then
very clear that the alumina surface layer protects the base material to environmental
low temperature degradation. The fact that 2H is more susceptible to low tempera-
ture ageing degradation can be attributed to the higher transformability of the
tetragonal grain because of its larger grain size.
By the above high temperature reaction of AlN with Y-TZP it is possible to
produce a layer that protects Y-TZP from the environment, making the material
inert to low temperature ageing degradation. In addition, the resulting material,
2HA, does not only have the same biaxial strength as Y-TZP, but also a higher
fracture toughness.
In order to correlate the decrease in strength with the appearance of monoclinic
phase at the surface, the presence of this phase was checked by X-ray diffraction
during ageing. From Fig. 6 it can be noticed that there is an increase in the volume
fraction of the monoclinic phase on the surface of Y-TZP and 2H during ageing while
this phase is not detected on the surface of 2HA. This is consistent with the variation
of the biaxial strength observed in Fig. 5. Finally in Fig 7 (a-c) it can be noticed the
1400
1200
1000
σmax(MPa)
800
600 2HA
2.5Y-TZP
400 2H
200
0
0 100 200 300 400 500 600 700 800
Time (hours)
Figure 5. Change in biaxial strength with ageing time at 2508C.

complete cracked surface of a specimen of 2H after ageing for 150 hours at 250 C.
The easy activation of the t-m transformation at the surface, because of the high
transformability of the large t-ZrO2 grains, has led to the formation of micro and
macro-cracks which have propagated to the interior affecting very strongly the
strength. It can be seen that the microcracks propagate trough the tetragonal grains,
since these are responsible for the expansion that takes place with the transformation
(Fig. 7b).
It has been investigated the resistance to low temperature aging degradation of 2.5Y-
TZP as well as of the materials that result after heat treatment at 1650 8C in air, and
after formation of a thin layer of alumina by reaction with AlN powder in a nitrogen
atmosphere. In both cases, the strength hardly changes and the fracture toughness
increases because it is controlled by the high transformability of the coarser tetrag-
onal grain size that it is induced in the bulk. The larger tetragonal grain size makes
the material heat treated in air more sensitive to low temperature ageing degradation
0.5
Y-TZP
2HA
Volume fraction ofm-ZrO2
0.4 2H
0.3
0.2
0.1
0.0
0 100 200 300 400 500 600 700 800
Ageing time(h)
Figure 6. Change in volume fraction of m ZrO2 with time at 250 8C.
1 mm 7μm 100μm
(a) (b) (c)

Figure 7. Micrographs of the sample 2H aged at 250 8C: surface of the disc (a); micro and macro cracks
formed due to t m phase transformation (b) and fracture surface showing micro and macro cracks formed
near the surface.
than the starting Y-TZP. By contrast, the surface layer of alumina that is formed
after high temperature reaction has a large resistance to t-m transformation during
ageing at 250 8C in air, as well as an increased resistance to environmental subcritical
crack growth. Then, the final result is a material with high strength, high fracture
toughness and inert to low temperature ageing degradation.
ACKNOWLEDGEMENTS
The authors gratefully acknowledge for its financial support to the MCYT under
grant No. MAT-99-0781, to DURSI (2001SGR-00229), and also to CIDEM for
supporting CIEFMA (XARXA IT).
REFERENCES
1. D.J. Green, R.H.J. Hannink, M.V.Swain, Transformation Toughening of Ceramics. Florida: CRC
Press Inc., (1989).
2. R.M. McMeeking and A.G. Evans, Mechanics of transformation toughening in brittle materials,
J. Am. Ceram.Soc. 65 242 246 (1982)
3. F.F. Lange, G.L. Dunlop and B.I. Davis, Degradation during aging of transformation toughened
ZrO2 Y2 O3 materials at 250 8C, J. Am. Ceram. Soc. 69 [3] 237 40 (1986).
4. T.J.Chung, J.S. Lee , D.H. Kim and H.Song, Surface nitridation of yttria doped tetragonal zirconia
polycrystals (Y TZP): Microstructural evolution and Kinetics, J. Am. Ceram. Soc. 82 [11] 3193 99
(1999).
5. N. Claussen, R. Wagner, L. J. Gauckler and G. Petzow, Nitride stabilized cubic zirconia, J. Am.
Ceram. Soc. 61 [7 8] 369 70 (1978).
6. A. Feder, J. Alcalá, L. Llanes and M. Anglada, On the stability of nitrided Y TZP at high tempera
ture, to be published in J. Eur. Ceram. Soc (2003).
7. D. Casellas, F.L.Cumbrera, F.Sánchez Bajo, W. Forsling, L. Llanes and M. Anglada, On the
transformation toughening of Y ZrO2 ceramics with mixed Y TZP/PSZ microstructures, J. Eur.
Ceram. Soc. 21 765 77 2001.
8. D.Casellas, A. Feder, L. Llanes and M. Anglada, Fracture toughness and mechanical strength of Y
TZP/PSZ ceramics, Scripta Mater. 45 213 20 (2001).
9. K. Niihara, R. Morena, D. P. H. Hasselman, Evaluation of KIC of brittle solids by the indentation
method with low crack to indent ratios, J. Mat. Sci. Lett. 1 13 16 (1982).
10. A.Borger, P. Supancic and R. Danzer, The ball on three balls test for strength testing of brittle discs:
stress distribution in the disc, J. Eur. Ceram. Soc. 22 1425 1436 (2002).
11. G. Bayer and A. Mocellin, Displacement reaction sintering in the system ZrO2 AlN, Rev. Chim. Min.
23 80 87 (1986).
12. C. Toy and E. Savrun, Novel composites in the aluminum nitride zirconia and hafnia systems,
J. Eur. Ceram. Soc. 18 23 29 (1998).
INDENTATION AND
SCRATCHING OF GLASS:
LOAD, COMPOSITION AND
TEMPERATURE EFFECTS
Jean-Christophe Sanglebœuf and Tanguy Rouxel*
1. INTRODUCTION
The response of glass to mechanical contact has been the subject of numerous
publications. However, most of the investigations were dedicated to standard win-
dow glasses, although there is a need for the characterization of functional glasses
such as chalcogenide glasses and of newly developed structural glasses, including
oxynitride and oxycarbide glasses. Besides, most investigators focused on the phe-
nomenology and on the mechanics of contact damage and the incidences of the glass
composition, the environment and the temperatures were little studied and thus
remain poorly understood.
In this paper, we intend to show the importance of the glass composition
through several examples. Silicon oxycarbide and magnesium-containing soda-
lime-silica glasses provide interesting illustrations of the effect of the network poly-
merization degree and of the compactness on the indentation behavior. A series of
soda-lime-silica glasses with different silica contents was studied to get insight into
the role of moisture and a pronounced indentation-creep phenomenon was noticed in
chalcogenide glasses from the Ge-Se system.
*
Jean Christophe Sanglebœuf, Tanguy Rouxel, Laboratoire de Recherche en Mécanique Appliquée de
l’Université de Rennes 1, LARMAUR UPRES JE 2310, Université de Rennes 1, campus de Beaulieu,
35042 Rennes cedex, France.
121
122 Jean-Christophe Sanglebœuf et al.
2. EXPERIMENTAL TECHNIQUES
The glass density (d) was measured on powdered samples by the pycnometry method
with an accuracy of 0:005 g cm 3 . Young’s modulus (E) and Poisson’s ratio (n)
were measured by the pulse-echo technique on 4 mm thick rectangular specimen.
Vickers indentations (load of 9.8 N applied during 15 s), were performed to estimate
both Meyer’s hardness (H) and the ‘‘indentation’’ fracture toughness1 (Kc ) by means
of the following equations:
H ¼ P=2a2 (1)
and
Kc ¼ 0:016(E=H)1=2 P=c3=2 (2)
where P is the load, a is half the mean size of the two diagonals, E is Young’s
modulus, H is Meyer’s hardness, and c is half the mean length of the two radial
cracks (tip to tip).
An apparatus allowing for a variable loading cycle during a single experiment
was designed, and placed into a glove box in order to achieve a good moisture control
(+ 2%) in the range 0% (< 50 ppm) 100%, so as to study the effect of the
environment2 (Figure 1).
The horizontal displacement of the indenter (to produce a groove) is obtained by
a bearing slide, while a piezoelectric actuator controls the vertical movement. Both
the penetration and the tangential forces are continuously monitored. The main
characteristics of the apparatus are a maximal normal load of less than 50 N with
an accuracy of 0.05 N, a vertical displacement limited to 100 mm, and a bearing slider
allowing for a displacement accuracy better than 1 mm and a maximum speed of
Figure 1. Experimental set up designed for variable loading scratch experiments in a moisture & tem
perature controlled atmosphere (gloves are removed on the picture).
Indentation and Scratching of Glass 123
5 mm/s. The speed and acceleration of the bearing slider is computer controlled, and
displacements and environmental parameters are continuously monitored. The
scratch resistance (tangential load) was followed through constant normal load and
low rate loading-unloading experiments. Such experiments allow first to study a wide
load range within a single experiment and second, to reveal any possible dynamical
effect, by studying the symmetry of the scratch pattern. A Vickers penetrator (inverse
pyramidal, 1368 face to face angle) was used with a leading edge orientation. The
samples and the indenters are cleaned with ethanol before each test. Four different
grooves were made on each glass and a good reproducibility was observed.
An optical microscope allowing for interferometrical observations and an
atomic force microscope (AFM, Digital Instrument, Nanoscope IIIA) operated
both in tapping and in contact modes, were used to observe and to analyze the
topology of the damaged surfaces. The optical observations were systematically
made 1 hr after the mechanical testing, so that the possible incidence of fatigue-
corrosion phenomena is about the same for all the specimens.
The specific glass transition temperature (Tg) of each glass was determined by
Differential Scanning Calorimetry (DSC) at a heating rate of 108C min 1 and each
batch was then annealed at the corresponding Tg temperature for 30 min and cooled
again to room temperature at 18C min 1 . Note that the onset of the first endothermal
peak was used and an error of 28C was estimated. The thermal expansion coeffi-
cients (a) of the glasses were determined from the 50 3008C portions of the dilato-
metric curves obtained during heating at 58C min 1 .
3. MATERIALS
Silicon-oxycarbide glasses are of great interest because of their remarkable mechan-

ical properties and refractoriness, which clearly correlate to the increase of the
average coordination number in the glass network.3–8 Interest in oxycarbide glasses
is hence obvious, as far as carbon atoms form C-Si covalent bonds, as it is mainly the
case for the presently studied glasses in which most carbon atoms get into C(Si)4
four-fold coordinated The glass preparation has already been previously reported in
detail.5,6,9 In its mainlines, a (DH ,TH ) gel is first synthesized from hydrolysis con-
densation reactions of a mixture of methyldiethoxysilane, HMeSi(OEt)2 , and,
triethoxysilane, HSi(OEt)3 , where DH and TH stand for HSiO2 and HSiO3 units
respectively. The gel is then pyrolysed at 10008C under inert atmosphere, to obtain
black glass rods, 1 to 2 mm thick, and over 30 mm long, consisting in i) an oxycar-
bide phase with SiCx O2(1 x) stoichiometric composition, where carbon atoms form
C(Si)4 tetrahedra; and ii) free carbon microdomains (which give the black color to the
glass). The DH TH 0:5 and DH TH 1 grades were studied in this work. In addition, a
standard soda-lime-silica glass (float glass) and a vitreous silica (hydroxyl content ¼
30 ppm) were characterized for sake of comparison. The glassy phase theoretical
compositions of the DH TH 0:5 and DH TH 1 grades, are SiC0:33 O1:33 and SiC0:375 O1:25
respectively. Free carbon contents of 0.5 and 4 vol. % were estimated for the
DH TH 0:5 and DH TH 1 grades respectively.
Oxide glasses with compositions based on the soda-lime-silica system were synthe-
sized by mixing in oxide (SiO2 , Al2 O3 ) and carbonate (NaHCO3 ,K2 CO3 ,Mg(OH)2
(MgCO3 )4 ,CaCO3 ) powders.10 The powder mixtures was melted in a 200 cm3 Pt/
Rh10% crucible at about 16008C for 3.5 hrs, including a 1 hr long mixing treatment by
means of device with two platinum helices called a guinand. The melt was further poured
out in a pre-heated (5008C) graphite mould and annealed for 30 min at this temperature,
and further cooled to room temperature at 28 C min 1 . A serie of glasses with more than
70 mol% SiO2 (starting powder) was synthesized (Table 1). This ‘‘SLS’’ series consists of
glasses with Na2 O=CaO ¼ 1:52, and with rising SiO2 contents from 71 to 80 mol%.
Ge-Se glasses were synthesized by introducing the appropriate quantities of
germanium (Ge) and selenium (Se) in silica ampoules sealed under vacuum, by
further heating this mixture at high temperature, typically between 750 and 9508C,
and by subsequent quenching (see ref. 11 for details). The physical and mechanical
properties of the studied glasses are reported in Table 2.
4.1. Normal versus Anomalous Behavior
Concerning the indentation fracture patterns, two different behaviors may be distin-
guished: the so-called normal glasses tend to develop the radial-median crack
Table 1. Composition (mol %) and thermal properties of the oxide glasses.
Series SiO2 Na2 O K2 O MgO CaO Al2 O3 Tg (8C) a( 106 8C1 )
SLS1 71 17.5 0 0 11.5 0 558 10.9

SLS2 74 15.7 0 0 10.3 0 559 10.1
SLS3 77 13.9 0 0 9.1 0 560 9.1
SLS4 80 12.1 0 0 7.9 0 561 7.7
Error / / / / / / + 28C + 0.1
Table 2. Density, elastic moduli, indentation hardness and toughness of Ge Se glasses. n.d.:
non determined. 1)Partially devitrified. !Planilux, Saint Gobain Co.
p
Series d E (GPa) m (GPa) n H (GPa) Kc (MPa: m) Tg (8C)
SiO2 2.2 73 31.8 0.148 7.2 0.73 1190

DH TH 0:5 2.23 104 n.d. n.d. 8.63 0.7 1350
DH TH 1 2.2 110 n.d. n.d. 10.6 0.57 >1350
SLS1 2.53 70.2 29.2 0.202 5.35 0.71 558
SLS2 2.50 69.3 29.2 0.188 5.11 0.70 559
SLS3 2.47 68.6 29.0 0.181 5.15 0.76 560
SLS4 2.43 68.1 29.1 0.172 5.19 0.82 561
Float glass! 2.55 72.0 30.0 0.21 5.30 0.72 562
Se1) 4.28 10.25 3.74 0.322 0.39 0.16 41
Ge5 Se95 4.31 11.05 4.18 0.316 0.57 0.09 68
Ge10 Se90 4.34 12.08 4.59 0.307 0.77 0.12 92
Ge15 Se85 4.36 13.8 5.33 0.295 1.05 0.22 110
Ge20 Se80 4.37 14.73 5.72 0.286 1.38 0.28 162
Ge25 Se75 4.36 16.05 6.25 0.281 1.72 0.22 228
Ge30 Se70 4.32 17.9 6.82 0.264 2.02 0.20 300
Ge40 Se60 4.36 22.38 8.80 0.273 2.35 0.16 340
p
Error + 0.05 + 1 GPa + 1 GPa + 0.01 + 0.01 GPa + 0.01 MPa. m +2
system from the corners of the indents, whereas cone-cracks are observed in the
anomalous glasses.12,13 Pure fused silica is a model example of the anomalous
behavior (Figure 2). In this case, the permanent deformation beneath the indenter
can be mainly attributed to a pressure-driven densification process, and this process
originates in the flexibility of the Si-O-Si bond angle. The anomalous indentation
behavior of gel-derived SiOC glasses was already evidenced in a previous work.8 The
different indentation-fracture patterns observed in the studied glasses are summar-
ized in Table 3, for three different loads.14
Whereas well developed radial-median cracks form in the soda-lime-silica glass,
none are observed in pure silica. Instead, cone cracks appear, which intersect the
surface to produce ring cracks, some of which having diameters larger than the
indentation diagonal (Figure 2). The glass starts to be shivered for loads higher
than 4.9 N. In the DH TH 0:5 grade, the same anomalous behavior is observed, but
with the cone-cracks intersecting the surface at the indentation corners. For loads
equal or higher than 9.81 N, several secondary cone cracks appear, but no radial-
median crack. The behavior of the DH TH 1 grade is quite similar to the one of the
DH TH 0:5 grade, but differs by the occurrence of radial-median cracks as soon as the
load exceeds 4.9 N. At higher loads, chipping comes into play and the specimens tend
to be shivered. Therefore, it seems that the higher the carbon content is and the more
normal the SiOC glass behaves. It is suggested that covalently-bonded carbon atoms
reduce the mobility of the residual vitreous silica network (the matrix) at the mo-
lecular scale. It follows that deformation becomes less and less displacive (densifica-
tion) and proceeds more and more by shear.
Noteworthy, an examination of the fine details of the deformation-cracking
pattern by AFM reveals the presence of an array of microcracks almost concentric
with respect to the load axis, which is well resolved near the indentation corners
(Fig. 2). Of particular interest is the fact that these cracks follow the small turbostratic
Figure 2. AFM image (contact mode) of Vickers indentations performed on pure vitreous silica and on a
DH TH 0:5 specimen with a 2.94 N indentation load.
Table 3. Vickers indentation patterns obtained for various indentation loads.
Indentation load (N) Float glass Vitreous silica DH TH 0:5 DH TH 1
2.94
4.91
9.81
graphite segments. The ultimate stage of this process is the coalescence of the cracks
to form Hertzian-type cone-cracks, the main of which intersecting the indent corners.
These observations show that the free carbon plays a major role on the indentation
deformation/fracture behavior. The orientation of the graphite domains is likely to
be responsible for the indentation anisotropy, and these domains provide easy
fracture initiation sites and low energy fracture paths (likely through cleavage
along basal planes) and cause the surface ring cracks to initiate just outside of the
contact zone, i.e. almost along the indentation edges. Thus, in addition to the
densification process, flow beneath the indenter presumably proceeds by the collaps-
ing of the material by successive shear of hemispherical shells or strata. The largest
surface ring crack corresponds to the last stratum sheared.
4.2. Elastic Recovery and Time-Dependent Hardness
Chalcogenide glasses from the Ge-Se system exhibit relatively low glass transition
temperatures (from 41 (Se) to 3408C (Ge40 Se60 ) )11,15 and low activation enthalpy for
flow (from 285 to 592 kJ/mol).16 As a consequence, they experience a time or rate
dependent hardness at room temperature, especially for chalcogen-rich composi-
tions17 (Fig. 3). Therefore, on one hand, these glasses offer a unique opportunity to
study viscoelasticity at room-temperature, as was already mentioned earlier,18 but on
the other hand, structural parts such as lens or fibres for applications in night
visibility optical devices or in thermometry, are likely to suffer from creep deform-
Figure 3. AFM image (contact mode) of Vickers indentations performed on glassy selenium under 0.1 N
for 5 (left) and 50 s (right).
ation in service. The higher the germanium content and the larger the elastic
contribution.
In terms of the vertical displacement of the tip of the indenter, one can write:
u ¼ uel þ ufl (3)
where u is the total displacement, and uel and ufl are the elastic and flow (or creep)
components respectively. The creep displacement was directly measured from the
AFM profiles (Fig. 3) and the corresponding elastic contribution was estimated by
extrapolating lines from the indentation corners, at an angle of 748 (half of the edge
to edge angle) from the vertical. On the basis of a linear viscous behaviour, using
fluid-mechanics, the following expression is obtained for the apparent viscosity
coefficient:
ptana
h¼ tH(t) (4)
4
where a is half the face to face angle (for a Vickers indenter a 688), Eq. (4) gives:
h ¼ 1:943 tH(t) (5)
H is far from being simply proportional to 1/t, as it would be expected. Indeed a

pronounced shear-thinning trend is observed, with a more than one order of magni-
tude decrease of h with rising hardness.17 The higher the germanium content is and
the higher the viscosity becomes. Indeed, as most glasses studied so far at relatively
high stresses or strain-rates chalcogenide glasses seem to exhibit a pronounced
shear-thinning behaviour below Tg and for sufficiently high applied stresses (or
strain-rates). It was observed that the behaviour becomes more and more shear-
thinning as the germanium content is raised.). An excellent description of the time-
dependence of the hardness was obtained on the basis of a non-Newtonian indenta-
tion flow mechanism, using the standard Norton-Arrhenius law for creep
(de=dt ¼ Asn ). The following expression was then derived for the time-dependent
hardness:

1=n
2 DGa
H(t) ¼ t 1=n
exp þ (6)
pA(1 þ v)tana nRT
It is noteworthy that the time dependence in this expression is similar to the

phenomenological expressions proposed by Han et al.19 in a study of indentation
creep in soda-silica glasses, and more recently by Grau et al..20 Interestingly n
increases with the germanium content from 11 to 62, whereas Poisson’s ratio de-
creases (with the exception of the Ge40 Se60 grade). It means that for a given normal
strain, the volume change is larger and larger as the Ge content increases. It is
suggested that this volume change favours the alignment of the structural units
toward the stress axis, inasmuch it occurs at low temperature with respect to the
Tg , i.e. in a temperature range where the elastic contribution is still important. These
trends of h and n can hence be regarded as correlated to the glass transition
temperature: the higher the Tg of the glass is and the more pronounced the non-
Newtonian flow behavior becomes.
4.3. Load Dependence
Ge-Se glasses behave normally with respect to indentation, i.e. radial-median cracks
tend to form under Vickers indentation. For a given Ge-Se glass composition, the
indentation cracking behavior depends much on the applied load. There are typically
three major microcracking regimes:21 1) below a critical indentation load, hereafter
noted P , no crack is observed; 2) above P , two cracks show up, symmetrical with
respect to the indention, which indeed correspond to the intersect of a half-penny shape
crack with the surface (as the load increases, the number of cracks connected to the
indentation corners increases to three and four); and 3) above a critical load, Pc , chipping
occurs. Mostly, Pc coincides with the load necessary for the four crack pattern to form.
Note that a detailed analysis of the indentation corners reveals the nucleation of several
small cracks, less than 10 mm in length, for indentation loads at or just below P . The
load ranges for P and Pc are plotted as a function of the Ge content in Fig. 4.
The sensitivity of the glass to crack formation or to chipping increases signifi-
cantly with the germanium content. Lawn et al.22 and Hagan23 suggested that the
critical load for crack initiation from a pre-existing subsurface flaw be proportional
to Kc 4 =H3 . According to our experimental data, the proportionality factor would be
about 500.103, whereas coefficient of 22.103 and 885 were proposed by Lawn and
Hagan respectively, from data obtained either on oxide glasses and on ceramic
materials. Furthermore, the former dependence of P on Kc and H would predict a
linear increase of P with the germanium content up to 20% germanium whereas the
opposite trend is observed.
In a scratching experiment, the scratch pattern also strongly depends on the level
of the normal load, and for low loads the different kinds of micro-cracks which form
during, or with the scratch process;2 they are namely: i) the median cracks, ii) the
radial cracks and iii) the lateral cracks. This last kind of cracks induces chipping
when crack propagation toward the specimen surface occurs. The radial cracks are
significantly inclined toward the sliding direction, and propagate in a curved shape
usually ending perpendicular to the scratch direction when there is no combination
with another damage phenomenon (i.e. lateral cracks).
Indeed three different regimes appear during a typical scratch experiment at a
constant loading rate: i) the first regime is associated with a permanent groove with
Figure 4. Load ranges for the various microcracking regimes as a function of the germanium content.
Figure 5. Scratch behavior of SLS glasses (monotonic loading).
eventually the formation of sub-surface lateral cracks, and corresponds to a micro-

ductile regime, ii) the second one, so-called micro-cracking regime, is characterized
by an important damage (lateral cracks intersecting the surface and radial cracks)
and iii) the third one is a micro-abrasive regime and gives birth to many debris. SLS
glasses generally fall in a narrow range of compositions along the boundary
between devitrite (Na2 O 3CaO 6SiO2 ) and tridymite (SiO2 ) in the Na2 O
CaO SiO2 phase diagram. Shifting away from this narrow range affect various
important characteristics including the scratchability. A micro-abrasive regime
appears in the low-load domain for the SLS3 and SLS4, with high silica contents,
whereas lateral chipping occurs for the SLS1 and SLS2 compositions. Glasses from
the devitrite phase field are sensitive to chipping and glasses with silica-like net-
works appear to be much resistant to both crack propagation and chipping during
scratch experiments. At the atomic or molecular scale, the resistance of the high
silica content glasses to chipping or micro-cracking is due to the opened structure of
the glass network, allowing both for the network flexibility and for the flowdensi-
fication process.2 On the other hand, glasses with high modifying cation content
(typically > 25%), such as SLS1 and 2, contain more non-bridging oxygens which
provide easy paths for fracture and slip. Note that a similar conclusion can be
drawn from the indentation-scratching behavior of silicon-oxycarbide glasses,14
where the covalently bonded carbon in 4-fold coordination are found to induce a
more normal behavior.
4.4. Role of Humidity

Figure 6 shows the scratches conducted on the standard float glass at different
humidity levels. Monotonic loading/unloading cycles were conducted in order to
screen all the possible damage events in a single scratch experiment.2
Figure 6. Pictures of scratches conducted under different hygrometric levels with a Vickers penetrator
following a slope load unload process.
As expected from the background studies on the effect of humidity on the

fracture behavior of soda-lime-silica glasses,24,25 the scratch patterns significantly
depend on the hygrometric level: the more important the hygrometric level is, the
earlier (with regards to the normal load) the different damage processes appear.
Another observation concerns the scratch obtained with 0% of humidity. The dam-
age is characterized by a large lateral crack remaining under the surface of the
sample. A similar but shorter crack occurs during the unload process. This phenom-
enon fades rapidly as the humidity level is raised.
The peculiar morphology of the scratch made at 0% of humidity is interesting.
We can notice that there is almost a total lack of radial cracks while lateral cracks,
which never reach the surface, extend over a large distance. This observation put in
light the mechanism of the chipping process occurring during scratching: a chip is the
combination of two radial cracks and a lateral one which then intersects the specimen
surface. Two consecutive radial cracks located on the same side of the scratch track
may intersect each other as a result of a deviation process resulting from the
attractive forces between two close cracks. The coupling with a lateral crack leads
to a chip and thus to matter removal. When only one radial crack frames the lateral
one, this latter crack stay under the surface.
Subsurface lateral cracks, which propagate after the experiment, were found
to be insensitive to the hygrometric rate and to remain under the surface of the
material.
4.5. Role of Temperature
Indentation behaviour in temperature has been investigated in three different glasses

(a Zr-Cu-Al-Ni bulk metallic glass, a Ge-As-Se chalcogenide glass and the standard
float glass (Table 2) ) and was revealed to be complex.26 In fact, when temperature
increases, glass response becomes time-dependent (as discussed in § 4.2), and in the
vicinity of Tg increased permanent deformation is left by the indenter. However, all
glasses showed so-called ‘strong’ character with a relatively low decrease of hardness
around Tg 26 (Fig. 7).
Using high indenting loads, brittle behaviour could be generated in standard
sodalime-silica glass and chalcogenide glass. The behaviour was revealed to differ
strongly from one glass to the other. While standard float glass showed enhanced
apparent toughness at elevated temperatures almost no change in apparent toughness
was revealed in GeAsSe glass. However, comparison between chevron-notched bars
loading and indented samples showed that toughness determination is changed dra-
matically around transition temperature when the loading rate is changed27 (Fig. 8).
5. CONCLUSION
The indentation and scratching behavior of glasses from different glass forming
systems was investigated and was found to depend much on the glass composition.
Results were discussed in the light of the glass network structures, with a special
1.2
1.0
0.8
Hv/Hvo
0.6
0.4 ZrCuAINi
GeAsSe
0.2 soda-lime
Soda-lime (5)
0.0
0.2 0.7 1.2
T/Tg
Figure 7. Reduced hardness as a function of the reduced temperature Vickers indentation under 0.98 N
for the GeAsSe glass and under 4.91 N for the other glasses, with a loading duration of 30 s26 .
Figure 8. Temperature dependence of the apparent fracture toughness of a standard window glass.27
focus on the normal versus anomalous behaviors, on the critical load for micro-
cracking, on the indentation creep phenomenon and on the role of humidity. High
silica contents promote flow-densification process beneath the indenter. On the
contrary, carbon increases the polymerisation degree and the stiffness of the glass
network and lead toward a more normal behavior. Chalcogenide glasses, which can
be regarded as low Tg glasses with structures comparable to those of polymeric
glasses, exhibit indentation-creep at room temperature, although they are very brittle
and sensitive to indentation damage. The scratch resistance can be interpreted in the
light of the indentation features. For instance, the role of moisture, which is well
documented in indentation studies, is even exemplified during scratching.
REFERENCES
1. G.R. Anstis, P. Chantikul, B.R. Lawn and D.B. Marshall, ‘‘A critical evaluation of indentation
techniques for measuring fracture toughness: I, direct crack measurements’’, J. Am. Ceram. Soc., 64
[9] 533 538 (1981).
2. V. Le Houérou, J C. Sangleboeuf, S. Dériano, T. Rouxel, G. Duisit, Surface damage of soda lime
silica glasses: indentation and scratch behavior, J. Non cryst. Sol., 316 54 63 (2003).
3. J. Homeny, G.G. Nelson and S.H. Risbud, ‘‘Oxycarbide glasses in the Mg Al Si O C system’’, J. Am.
Ceram. Soc., 71 [5] 386 390 (1988).
4. H. Zhang and C.G. Pantano, ‘‘Synthesis and characterization of silicon oxycarbide glasses’’, J. Am.
Ceram. Soc., 73 [4] 958 963 (1990).
5. F. Babonneau, G.D. Soraru, G. D’Andrea, S. Dire and L. Bois, ‘‘Silicon oxycarbide glasses from sol
gel precursors’’, Mat. Res. Soc Symp. Proc., 271 789 94 (1992).
6. G.D. Soraru, ‘‘Silicon oxycarbide glasses from gels’’, J. Sol Gel Science and Tech., 2 843 48 (1994).
7. G.M. Renlund, S. Prochazka and R.H. Doremus, ‘‘Silicon oxycarbide glasses: Part II. Structure and
properties’’, J. Mater. Res., 6 [12] 2723 34 (1991).
8. G.D. Soraru, E. Dallapicola and G. D’Andrea, ‘‘Mechanical characterization of sol gel derived
silicon oxycarbide glasses’’, J. Am. Ceram. Soc., 79 [8] 2074 80 (1996).
9. G.D. Soraru, G. D’Andrea, R. Campostrini and F. Babonneau and G. Mariotto, ‘‘Structural
characterization and high temperature behaviour of silicon oxycarbide glasses prepared from Sol
Gel precursors containing Si H bonds’’, J. Am. Ceram. Soc., 78 379 387 (1995).
10. S. Dériano, Conception chimique de verres silicatés à hautes performances mécaniques, doctorate
thesis, University of Rennes, December 2002.
11. I. Melscoët, ‘‘Synthèse, caractérisation et optimisation de verres optiques pour des applications
infrarouges dans la fenêtre 8 12 mm’’, doctorate thesis, University of Rennes, December 2000.
12. A. Arora, D.B. Marshall, B.R. Lawn and M.V. Swain, ‘‘Indentation deformation/fracture of normal
and anomalous glasses’’, J. Non cryst. Sol., 31 415 428 (1979).
13. C.R. Kurkjian, G.W. Kammlott, M.M. Chaudhri, ‘‘Indentation behavior of soda lime silica glass,
fused silica, and single crystal quartz at liquid nitrogen temperature’’, J. Am. Ceram. Soc., 78 3
737 744 (1995).
14. T. Rouxel, J C. Sangleboeuf, J P. Guin, V. Keryvin, Surface damage resistance of gel derived
oxycarbide glasses: hardness, toughness, and scratchability, J. Am. Ceram. Soc., 84 [10] 2220 24
(2001).
15. V.F. Kokorina, Glasses for Infrared Optics, Clarendon Press (1996).
16. See for example: U. Senapati and A. K. Varshneya, ‘‘Viscosity of chalcogenide glass forming liquids:
an anomaly in the ‘‘strong’’ and ‘‘fragile’’ classification’’, J. Non cryst. Sol. 197 (1996) 210 218.
17. J P. Guin, T. Rouxel, V. Keryvin, J C. Sangleboeuf, I. Serre, J. Lucas, Indentation creep of Ge Se
chalcogenide glasses below Tg: elastic recovery and non Newtonian flow, J. Non cryst. Sol., 298,
260 269 (2002).
18. D. Lewis III, W.C. LaCourse and D.B. Hardy, ‘‘Strength of selenium glass in the transformation
range’’, J. Am. Ceram. Soc. 60 (3 4) (1977) 107 110.
19. W.T. Han and M. Tomozawa, ‘‘Indentation creep of Na2 O 3SiO2 glasses with various water
contents’’, J. Am. Ceram. Soc. 73 (12) (1990) 3626 32.
20. P. Grau, G. Berg, H. Meinhard and S. Mosch, ‘‘Strain rate dependence of the hardness of glass and
Meyer’s law’’, J. Am. Ceram. Soc. 81 (6) (1998) 1557 64.
21. J P. Guin, T. Rouxel, J C. Sangleboeuf, Hardness, toughness, and scratchability of germanium
selenium chalcogenide glasses, J. Am. Ceram. Soc, 85 [6] 1545 52 (2002).
22. B.R. Lawn and D.B. Marshall, ‘‘Hardness, toughness and brittleness: an indentation analysis’’, J.
Am. Ceram Soc., 62, 7 8, 347 350 (1979).
23. J.T. Hagan, ‘‘Micromechanics of crack nucleation during indentations’’, J. Mat. Sci., 14, 2975 2980
(1979).
24. V. H. Bulsara, ‘‘Scratch formation in brittle solids and its application to polishing’’, PhD Thesis,
Perdue University, USA (1997).
25. S.M. Wiederhorn, in Fracture Mechanics vol 2 Plenum Press New York, (1974) 613.
26. E. Le Bourhis, T. Rouxel, Indentation response of glass with temperature, J. Non cryst. Sol., 316,
153 159 (2003).
27. T. Rouxel, J C. Sangleboeuf, The brittle to ductile transition in a soda lime silica glass, J. Non cryst.
Sol., 271, 224 235 (2000)
INDENTATION FRACTURE,
ACOUSTIC EMISSION AND
MODELLING OF THE
MECHANICAL PROPERTIES
OF THIN CERAMIC
COATINGS
S.J. Bull, I. Arce-Garcia, E.G.-Berasetegui and T.F. Pagey
ABSTRACT
Fracture is an almost inevitable consequence of highly loaded contacts in ceramic-

coated systems. For relatively thick coatings fracture is often similar to that observed
in bulk samples of the coating but as the coating thickness is reduced the substrate
plays an increasing role in influencing or even controlling fracture behaviour. Both
through-thickness and interfacial fracture may be observed depending on the relative
toughness of the substrate, coating and interface. Through-thickness fracture is
exacerbated by plastic deformation in the substrate and therefore the load support
from the substrate is critically important in determining the type and extent of
fracture observed. In this paper, nanoindentation testing and post facto atomic
force and scanning electron microscopy is used to characterise the types of fracture
which occur for hard coatings on softer substrates and multilayer coatings on glass.
The effect of fracture on the nanoindentation load-displacement curves and the
hardness and Young’s Modulus values obtained from them is discussed and a simple
model to account for the observed behaviour is introduced. The use of acoustic
y
School of Chemical Engineering and Advanced Materials, University of Newcastle, Newcastle upon
Tyne, NE1 7RU, UK
21
22 S.J. Bull et al.
emission to monitor plasticity and fracture during the indentation cycle is also
discussed.
1. INTRODUCTION
The durability and performance of coatings in almost every application depend on

their as-deposited mechanical properties such as adhesion, hardness, toughness and
elastic modulus. These properties may not be the same as those of the same material in
bulk form due to different microstructural and defect states arising from the depos-
ition process. As the coating thickness is reduced it becomes increasingly difficult to
measure these properties by conventional methods and low load nanoindentation and
scratch testing have been developed to enable measurements to be achieved at scales
commensurate with the coating thickness. Continuously recording indentation test
methods are well established for the determination of Elastic Modulus and hard-
ness.1,2,3 but methods for assessment of fracture toughness and adhesion are much
less developed.4,5 In part the reason for this is that mechanism of fracture failure around
an indentation is complex and depends on the relative properties of coating, substrate
and interface and therefore a universal analysis method is unlikely to be produced.
However, in some cases it is possible to identify failure modes that are amenable to
analysis resulting in reasonable toughness data. Microscopy and analysis of fracture
paths is a key part of this approach, particularly as the indentations get smaller.
There are a number of different indentation fracture events which might occur in
a coating substrate system (Table 1), the occurrence of which will depend on the
relative toughness of coating/substrate and any interfaces present. These failure
modes will be altered by plastic deformation in the substrate which can lead to the
superposition of bending stresses on to those generated by the indentation.
This paper discusses the type of fracture events which occur for hard coatings on
soft and hard substrates and outlines the effect that fracture events have on other
mechanical properties measured by indentation (such as hardness and Young’s
Modulus). It also introduces the use of acoustic emission generation to monitor
fracture events in very thin coatings where it is difficult to identify them by conven-
tional microscopy (both scanning electron and atomic force).
2. EXPERIMENTAL
In this study thin ceramic coatings have been deposited onto both brittle and ductile
substrates to assess the fracture behaviour of the coating/substrate system. Niobium
Table 1. Fracture events associated with brittle (ceramic) coating/substrate systems.
Substrate Coating Fracture type
Ductile Brittle Through thickness fracture, brittle fracture in coating parallel

to interface, ductile interfacial fracture, microfracture in coating
Brittle Brittle Bulk chipping, through thickness fracture, brittle interfacial

fracture, microfracture in coating, microfracture in substrate
nitride coatings were deposited in the same process run on a fully hardened powder
metallurgy high speed steel (ASP23) and annealed stainless steel (304) substrates by
cathodic arc evaporation using standard commercial deposition parameters.6 The
substrates were polished to a 3 mm diamond finish prior to coating and a layer of
approximately 4 mm was deposited on each. Fullerene-like CNx was deposited on a
range of substrates at thicknesses of 400 nm and 1 mm using unbalanced magnetron
sputtering.7
A range of oxide and oxynitride coatings was deposited onto float glass using
magnetron sputtering to 200 nm thickness. In addition a multilayer stack with a
coating architecture similar to current commercial low emissivity solar control
coatings was deposited to investigate the effect of internal interfaces in the coating.
Details of the coatings deposited are presented in Table 2.
2.2. Nanoindentation Testing

Nanoindentation experiments were performed using two different nanoindentation
systems to make indents at peak loads from 100 mN to 500 mN. Hardness and
Young’s Modulus were extracted from the load-displacment curves using the method
of Oliver and Pharr.8 All load displacement curves were corrected for thermal drift and
the indenter area functions were carefully calibrated following the Oliver and Pharr
approach using a fused silica standard sample prior to testing the coated samples.
A Nanoindenter II with a ‘blunt’ Berkovich tip (tip end radius 250 nm at the
time of making the experiments) was used for testing the NbN coated samples and
the CNx coated samples at loads in the complete range. Due to the fact that the
Table 2. Coating/substrate systems investigated in this study.
Substrate Deposition
Hardness Thickness Deposition Temperature
Coating Substrate (GPa) (nm) Technology (8C)
NbN ASP23 steel 9.5 3980 Arc evaporation 425
NbN 304 stainless steel 1.9 2810 Arc evaporation 425
CNx Silicon 11.5 400, 1000 Unbalanced magnetron 350

sputtering
CNx 1 mm SiC 35.0 SiC 400, 1000 Unbalanced magnetron 350

on silicon 11.5 Si sputtering
CNx Sapphire 32 400, 1000 Unbalanced magnetron 350

sputtering
TiOxNy Float Glass 6.5 200 Magnetron sputtering 200
SnO2 Float glass 6.5 200 Magnetron sputtering 200
Zr (O) Float glass 6.5 200 Magnetron sputtering 200
ZnO Float glass 6.5 200 Magnetron sputtering 200
Oxide Float glass 6.5 Total Magnetron sputtering 200

multi layer thickness 80 nm
24 S.J. Bull et al.
indenter is relatively blunt this approach is not suitable for measuring the hardness
and Young’s Modulus for the CNx coatings as they do not plastically deform before
the substrate.9 However, the substrate plastic deformation is limited at the higher
loads where coating fracture and interfacial detachment occur and this can be useful
for interface toughness determination. The constant loading rates used were
500 mN=s for indentations to maximum loads greater than 100 mN and 50 mN=s
for lower peak load indentations.
A Hysitron Trinboindenter fitted with a sharp Berkovich tip (tip end radius
50 nm) was used to test the CNx and coatings on glass. A fixed loading rate of
100 mN=s was used in all cases to minimise the effects of creep on the measured load-
displacement curves. This system can use the tip which is used to make the indent to
perform AFM imaging of the impression after the indentation cycle is complete.
Given the small size of the impressions made at low loads this is the only practical
way to achieve imaging of the impressions as the indent arrays are impossible to find
using off-line AFM or SEM even when precision locating stages are used.
Nanoindentations were also performed on the single and multi-layer coated
glass with the Hysitron Triboindenter fitted with a 1208 conical indenter (blunt tip,
radius 500 nm) instrumented for the recording of acoustic emission (AE) during
the indentation cycle.10 Loads from 1 mN to 10 mN were used in this analysis. AE
events were sampled at 4MHz; once a threshold signal is exceeded (determined from
noise measurements with the indenter in air) the system records the AE signal
waveform for subsequent analysis (Figure 1).
The energy in the acoustic event is then calculated from integration of the
amplitude of the AE waveform, A(t), with respect to time and is given by11:
ðt
EA (t) ¼ A(t)2 dt (1)
0
To calculate the net energy of the event, the linear background contribution (at) is
subtracted (Figure 2), assuming that the amplitude of the background contribution is
virtually constant during tests and has a mean amplitude a:
EANet (t) ¼ EA (t) at (2)
1500
Trigger event
1000
AE Amplitute
500
−500
−1000
0 0.04 0.08 0.12
Time, t (ms)
Figure 1. AE amplitude as a function of time after an acoustic emission event which occurred at 4000 mN
load in the testing of a 200 nm thick TiOxNy coating on float glass.
4
Background
3 Contribution
Raw Energy
Energy detargent 2
0 Net Energy
−1
0 0.02 0.04 0.06 0.08 0.1 0.12
Time (ms)
Figure 2. An example of the energy calculation for the event in Figure 1. The net energy is calculated by
subtracting background contribution from the raw energy.
The constant a is determined for each acoustic event by fitting the background
integrated energy a long time after the event as shown in Figure 2. It is assumed
that the background is linear which is reasonable over the AE pulse duration.
To discriminate between different events, not only the shape of the wave can be
analysed but also the parameter emission shape factor, defined as:
Amax
AESF ¼ (3)
ARMS
Previous studies have demonstrated that AESF differentiates between different types
of acoustic emission generation process.11,12 In addition, the attenuation time of the
signal, tat , may be used to differentiate between events.
2.3. Microscopy
Indentations produced at high loads using the Nanoindenter II could be located for
post facto imaging and were imaged by scanning electron microscopy (SEM) and
atomic force microscopy (AFM). Cracks were easily visible in the SEM and their
visibility was enhanced by using backscattered electron imaging. It proved to be
almost impossible to see cracks in the AFM images unless they displayed significant
residual (opening) displacements. This is probably due to the fact that water from the
environment is drawn into the fine cracks by capillary action and the AFM cantilever
rides over the open crack on the surface of a water film.
3.1. Ceramic Coating on Ductile Substrate: NbN on Steel
No evidence for interfacial detachment was found for any of the NbN coatings
investigated in this study but through thickness fracture was observed (Figure 3).
The type of fracture observed depends on the hardness of the substrate. For both
types of coating fracture initially occurs in the radial direction following the edges of
26 S.J. Bull et al.
Figure 3. Scanning electron micrographs of 500 mN indents in NbN coated (a) 304 stainless steel and
(b) ASP23 high speed steel.
the pyramidal indenter. For the harder ASP23 substrate the only other fracture
visible is picture frame cracking within the indent (Figure 3b). However, the coating
on the stainless steel substrate shows a circular through-thickness crack at the edge of
the contact (Figure 3a) which is some distance from the well defined indent at the
centre of the circle. Radial cracks defined by the indenter edges connect this central
indent to the edge of the contact but do not extend beyond it.
The difference in behaviour can be understood from the load-displacement
curves. Figure 4 shows a comparison between the load-displacement curves for the
NbN coatings on the two substrates tested at 500 mN. At low loads the load-displace-
ment curves superpose indicating that it is the coating properties which dominate the
measured response and the load is proportional to displacement squared as expected
for elastic-plastic indentation of a coating where the substrate provides good load
support. However, when the contact-induced shear stresses rise to trigger plastic
defomation in the substrate below the coating,13 plastic deformation starts in the
stainless steel substrate the substrate collapses beneath the coating which consequently
has to elastically bend to follow the deforming substrate. There is a change in slope of
4mm NbN on Different Steels

600
ASP23
500
304 Stainless
400
Load (mN)
300
200
100
0
0 500 1000 1500 2000 2500 3000
Displacement (nm)
Figure 4. Load displacement curves for NbN coatings on different steel substrates.
the load-displacement curve at this point. The size of the central indent is the same as
the size of an indent with peak load 100 mN (about 400 nm deep) which represents the
size of the indent in the coating at the onset of plastic deformation in the stainless steel
substrate. The impression caused by plastic deformation of the NbN coating stops
growing at this point and further plastic deformation is concentrated in the substrate.
The coating remains elastic but is bent into the impression created by substrate
deformation. To accommodate this deformation the radial cracks following the in-
denter edges extend as the plastic zone in the substrate extends. Tensile radial stresses
in the coating at the edge of the substrate plastic zone, in combination with these
bending stresses lead to the single circular through-thickness crack under such
circumstances the displacement of the indenter increases greatly as the disc of coating
is separated from the rest of the coated surface and membrane stresses supporting
some of the load drop to zero.
The fact that the load displacement curve shows no discrete discontinuities
(‘‘pop-ins’’ implies that there is a gradual extension of the radial cracks as the load
is increases and the through thickness ring crack occurs at close to the peak load. In
the absence of a recognisable pop-in event or a clear idea of the normal load at which
fracture occurs it is not possible to analyse the fracture toughness of the coating from
the available data.
For thinner coatings, through-thickness cracking may be exacerbated by pile-up
of substrate material, but pile-up is suppressed by the 4 mm NbN coatings investi-
gated here.
The fracture which occurs for the stainless steel substrate has a measurable effect
on the hardness and Young’s Modulus measured for the coating and must be taken
into consideration if the performance of the coated system is to be successfully
modelled. Recently some success in predicting the hardness of a coating substrate
system as a function of contact scale has been achieved using an energy based model
in which the contributions to the total work of indentation from coating, substrate
and interfaces can be summed and divided by the total deforming volume to deter-
mine the system hardness.14,15 The total deforming volume is assumed to be the
plastic zone associated with the impression and does not take into consideration any
plastic deformation associated with cracking which may occur outside this. From the
SEM images in Figure 3, where extensive cracking is not observed, this assumption is
reasonable. Two interfaces must be considered, thus

Vs H0 (s) þ Vf Ho ( f ) þ As gs þ Ai gi
H¼ (4)
V
where Vf and Vs are the deforming volumes of coating and substrate, H0 (f) and H0 (s)
are the bulk hardnesses of coating and substrate, As is the surface area and Ai is the
interfacial area within the plastic deforming region. The total deforming volume is
assumed to be hemispherical and the deforming volumes in coating and substrate are
given by appropriate slices through the hemisphere. The surface energy terms can be
used to account for the fact that the real deforming volume is unlikely to be hemispher-
ical. The radius of the deforming volume is related to the depth of the impression by16
1=2
E
R ¼ k2 d cot1=3 c (5)
H
28 S.J. Bull et al.
where k2 is a geometrical constant (k2 ¼ 3:64 for a Berkovich indenter), d is the

plastic indentation depth, E is the Young’s Modulus of the material and c is
the effective indenter angle (c ¼ 70:38 for a Berkovich indenter).
To include the effects of fracture in the model is relatively straightforward if we
assume that the coating within the cracked region does not contribute to the plastic
work of indentation after fracture occurs. Fracture usually occurs when the plastic
zone in the substrate is very well-established so we can consider the volume of coating
to be a cylindrical disc with the thickness of the coating. The area of the crack
produced is small compared to the volume it encloses so we also assume that the
surface energy of the crack plays only a minor part in the work of indentation. For a
radial cracked system the cracks have radius kc d where kc is a constant and d is the
plastic indentation depth. These form above a critical depth, dc . The area of material
bounded by radial cracks increases linearly with indentation size. In this case the
correction for the modelled hardness is given by

3 tH0F þ g s þ gi k2c d2
DH ¼
2R3
(6)
3tEf k2c d2
DE ¼
2R3
A simple estimate for kc is that it relates the depth to the radius of an equivalent cone
with the same volume as the impression. Thus, for a Berkovich indentation
r
24:5
pr ¼ 24:5d and therefore kc ¼
2 2
(7)
p
NbN on 304 Stainless Steel

50
Measured Hardness
40 Work Model
Work model with radial fracture
Hardness (GPa)
30
20
10
0
0 500 1000 1500 2000
Depth (nm)
Figure 5. Hardness of NbN on stainless steel as a function of contact depth comparing modelled data with
experiment. Correction for fracture is necessary to achieve a good fit. (Modelling parameters; Hf 24 GPa,
Ef 475 GPa, gf 15000J=m2 , H2 1:7 GPa, Es 210 GPa, gi 5000J=m2 and a 2:0).
Corrected hardness predictions are shown in Figure 5 compared to experimental

data. The quality of the fit is excellent (correlating coefficient >0.98) and much better
than can be achieved by any other hardness modelling approach investigated in this
study. An advantage of the model compared to existing hardness models is that it can
be used to make predictions of the behaviour of different coating/substrate systems
based on only a few scale-corrected physically-meaningful materials constants.14
3.2. Brittle Coating on Brittle Substrate: CNx on Silicon, Silicon Carbide

and Alumina
For brittle coatings on brittle substrates both through-thickness and interfacial
fracture can occur depending on the relative toughness of the coating, substrate
and interface. In the case of CNx coatings the coating and substrate toughness are
usually greater than the interface toughness and so detachment of the coating can be
produced by high load nanoindenation.
Loading a pyramidal indenter onto the surface of a coated sample creates
compressive stress in the coating (see Figure 6). Such stresses can cause the coating
to buckle and detach from the substrate around the indenter (see SEM micrograph in
Figure 6a). According to Thouless,17 once buckling starts the mode mixity of the
stress at the interfacial crack tip changes in such a way that the interfacial crack is
diverted to the surface forming a chip. This is registered in the load-displacement
curve, where a long step can be observed in its loading portion as the indenter
advances into the sample (see Figure 6b). In this case the buckled material has failed
catastrophically and the coating has been ejected from beneath the indenter since the
pop-in distance is the thickness of the coating.
If the coating has not undergone catastrophic fracture during loading, on
unloading the coating is still under compressive stress and around the indenter it is
already detached from the substrate (see Figure 7a). Towards the end of unloading
the coating buckles beneath the indenter pushing it up,5 which is associated with the
linear unloading observed in the P-d curve (see Figure 8a). During buckling the
outside of the coating is under tensile stress (see Figure 7b) and fracture may initiate
where the buckle curvature is highest (Figure 7c). This behaviour is observed for
500
A B.
400 Coating
fracture
Load, mN
300
200
100
0
0 500 1000 1500 2000 2500
(a) (b) Displacement, nm
Figure 6. (a) SEM micrograph showing coating detachment during loading of the indenter for a 1 mm
thick CNx coating on 3C SiC (001) substrate. (b) Step (A B) in the loading curve due to through thickness
cracking of the coating and spalling.
30 S.J. Bull et al.
Compressive Compressive
stress stress
(a)
Detached region
Coating fracture
Tensile stress
(b) (c)
Figure 7. Schematic diagram of the different steps followed by the coating fracture during unloading of
the indenter: (a) Coating under compressive stress and detached region (b) Buckling and tensile stress, and
(b) fracture of the coating.
Fracture during loading
500
400
Load (mN)
300 Linear
unloading
200
100
0
0 400 800 1200 1600
(a) Fracture during unloading (b) Displacement
Figure 8. (a) SEM micrograph for a 1 mm thick CNx coating on Si (001) showing coating fracture during
loading and unloading of the indenter at 500 mN applied load. (b) Corresponding P d curve showing linear
unloading at the bottom of the unloading curve associated with buckling of the coating.
1 mm thick CNx coatings on Si(001) substrate at 500 mN applied load (Figure 8b). In
this case fracture occurs both on loading and unloading.
In order that coating material be ejected from beneath the indenter, radial
fracture must occur along the indenter edges such that the coating may be pushed
laterally from the contact as the load increases. In such circumstances the size of
the pop-in is dictated more by coating thickness and the work of indentation
associated with the pop-in event is a complex convolution of the energy needed to
slide the coating beneath the indenter and the coating fracture energy. Any fracture
toughness measurements based on this failure mode therefore need to be treated with
caution.
3.2.1. Measurements of Fracture Toughness and Interface Toughness
Since both through thickness and interfacial failure occurs for CNx coatings on all of
the substrates investigated here it is possible to estimate the fracture toughness of
coating and interface. However, it must be born in mind that all current analysis
methods based on indentation provide only a qualitative assessment and more
research is needed to obtain reliable quantitative results.4
3.2.1a. Fracture Toughness of the Coating from Chipping

The nanoindentation fracture toughness of CNx films can be estimated by different
methods from the literature, the common basis of all of them being the calculation of
the energy dissipated in the fracture of the coating. This energy (Ufr ) is calculated by
determining the energy dissipation during chipping from the step observed in the
loading part of the load-displacement curve when through-thickness fracture of the
coating occurs during indentation.
Three main methods have been used in this study:-
(a) The method proposed by Li, Diao and Bhushan18,19 for which the fracture
toughness of the coating is given by,

12
E U
KIC ¼ (8)
(1 n2 )2pCR t
where KIC is the critical stress-intensity factor or the fracture toughness of the
coating; E is the Young’s modulus of the coating,; n is the Poisson’s ratio of
the coating; CR is the crack length,; U is the strain energy difference before and
after cracking; and t is the thickness of the coating. In this case the loading curve is
extrapolated from the start of the discontinuity to the depth of the end of the
discontinuity and U is given by the difference between the extrapolated and measured
curves. Toonder et al.4 introduced a slight variation to this expression to take into
account the number of chipped areas (N; generally 3 when using a Berkovich
indenter) and account for the fact that the crack does not propagate perpendicularly
to the substrate but at an angle d. Therefore it is necessary to consider the effective
coating thickness (t ¼ t=sin d; where d is the average angle of the chipping edge, see
Figure 9) instead of the actual thickness of the coating, t. This leads to the following
expression for the fracture toughness of the coating,

12
E U
KIC ¼ (9)
N(1 n2 )2pCR t0
(b) The method of Toonder et al.4 based on the calculation of the fracture energy,
G, from the energy dissipated during chipping, Ufr , and the area of the crack
formed in the coating by chipping. In this case the total irreversible work, Wir , is
determined by measuring the area between the loading and unloading curves
and determining the difference before and after chipping by extrapolating the
variation of Wir with load for chipped and unchipped samples. The fracture
energy is then given by
32 S.J. Bull et al.
S4500i153Profile
500
Height (nm)
−500 δ
−1000
−1500
−2000
0 10 20 30 40 50 60
(a) Distance (micron)
1micron CNx on AI2O3 P=500N i7

200
0
Height (nm)
−200
−400 δ
−600
−800
−1000
20 25 30 35 40 45 50 55
Distance (micron)
(b)
Figure 9. SEM micrographs and corresponding profiles of a 500 mN applied load on 1 mm CNx coating
deposited on (a) Si(001) and (b) Al2 O3 showing through thickness fracture and chipping of the coating at
an angle, d.
Ufr
G¼ (10)
N2pt0 CR
The fracture toughness of the coating is then calculated by its general equation,
r
p EUfr
KIC ¼ EG ¼ (11)
N2pt0 CR
where E is the elastic modulus of the coating and the other factors have been defined
previously.
(c) Finally, a lower and an upper limit for the fracture toughness of the coating can
also be determined by analysis of the load-displacement curve itself at the
moment of fracture.4 An upper and lower limit for the energy dissipated during
chipping of the coating can be calculated by taking into consideration the
following assumptions: (i) fully elastic (for the lower limit) or plastic (for the
upper limit) behaviour of the material before and after fracture, (ii) P is pro-
portional to d2 during loading and unloading20. This would lead to the follow-
ing limits for Ufr ,
2
Pfr (df d0 ) # Ufr # Pfr (df d0 ) (12)
3
where Pfr is the load at which the fracture of the coating is observed and d0 and df
are the indentation depths before and after the fracture of the coating. The limits for
the fracture toughness are calculated by substituting these limits in equations 10
and 11.
Table 3 shows the values for the fracture toughness of CNx deposited on
different substrates for the methods previously presented. There is reasonable agree-
ment between all the methods in the values of the fracture toughness of CNx except in
the case of the original uncorrected expression given by Li et al.18, 19 This is due to the
fact that for these samples through-thickness of fracture of the coating does not
happen perpendicularly to the substrate surface but at a certain angle, as SEM
micrographs and optical profilometry show in Figure 9.
3.2.1b. Fracture Toughness of the Interface
Two different expressions have been used to estimate the fracture toughness of the
interface between the CNx coating and different substrates. The first one is that
developed by Malzbender and de With21 which takes into account the residual stress
and the elastic properties of the substrate and is applied to the spalling of coatings
during scratch or indentation testing. For this method, the interfacial fracture
toughness for a coating possessing residual stress is given by,
v !2
u r
u p p t 52 s2res
Kint ¼
s t 0:397 sin b Eint E 2 þ t
Eint
(13)
L 2 E
where Ksint is the interfacial fracture toughness; b is the semi angle between two
cracks (for a Berkovich indenter b ¼ 60 ); t is the thickness of the coating; L is the
crack length (see Figure 10); sres is the residual stress in the coating; E is the Young’s
modulus of the coating,; and E int is the Young’s modulus of the interface, which can
be calculated from the following equation assuming equal contributions from coating
and substrate,22

1 1 1 1
¼ þ (14)
Eint 2 E Es
where, E and Es are the Young’s modulus of the coating and substrate, respectively.
Table 3. Comparison between the fracture toughness of 1 mm thick CNx coatings deposited
on different substrates based on the energy dissipated during chipping of the coating.
p
KIC , MPa m
Bhushan Bhushan Toonder Lower and Upper limits

Substrate et al.(1997) corrected et al.(2002) Toonder et al. (2002)
Si(001) 6.33 3.36 2.90 5.30 6.50
Al2 O3 5.10 2.82 2.44 2.53 3.10
3C SiC(001) 17.70 6.37 5.51 5.50 6.73
3C SiC(111) 13.80 5.66 4.90 5.68 6.95

34 S.J. Bull et al.
Radial crack
β
L
Chipped coating
segment
2a
Figure 10. Schematic diagram of the geometry of a chipped segment of coating.
The other expression used is that proposed by Rosenfeld et al.23 in which the
interfacial fracture toughness is related to the size of the delaminated area and the
load at which delamination of the coating occurs and is given by,
p
0:792H t(1 n2 )
KIC ¼ (15)
2(1 n)HCR2
1þnþ
Pdel
where H is the hardness of the coating; t is the coating thickness; n is the Poisson’s
ratio of the coating; CR is the length of the delamination crack; and Pdel is the load at
which delamination occurs.
The results obtained are presented in Table 4. The residual stress in the CNx
coatings deposited on Si(001) has been calculated from the change in radial crack
length around Vickers indentations after coating using the following expression
introduced by Lawn and Fuller,24
32 1
C0 2st 2
1 ¼ (16)
C Kc
where C0 and C are the crack length in the substrate and coated sample, respectively;
t is the coating thickness; s is the residual stress in the coating; and Kc is the fracture
toughness of the substrate calculated by,25
Table 4. Values of the interfacial fracture toughness for CNx

coatings (1 mm thick) deposited on different samples.
p
KIC , MPa m
Substrate Rosenfeld et al. (1990) Malzbender & de With (2002)
Si(001) 5.92 0.75
Al2 O3 4.67 0.68
3C SiC(001) 5.51 0.40
3C SiC(111) 5.58 0.41

12
E
Kc ¼ 0:016 PC 3=2
(17)
H
where E is the Young’s modulus; H is the hardness; P is the applied load; C is the
crack length in the uncoated sample. The value for the compressive residual stress in
1 mm thick CNx coating deposited on Si(001) substrate is 0.37GPa. This value has
also been considered for 1 mm thick CNx coatings deposited on 3C SiC(001) and 3C
SiC(111) since the thermal expansion coefficients for 3C SiC are similar to those for
Si(001). However, Al2 O3 substrate presents higher thermal expansion coefficients
than Si(001) and therefore the residual stress in CNx coatings deposited on this
substrate is around 0.6GPa.
The values for the fracture toughness of the interface obtained by the two
methods differ quite significantly (Table 4) although the trends for the different
substrates are the same. One of the reasons for this is that the equation proposed
by Rosenfeld et al.23 does not consider the residual stress in the coating after
deposition and the buckling of the coating during indentation, but assumes that
the coating has a constant hardness and this is directly proportional to the radial
stress. Furthermore, it has been previously seen that for these coated samples the
cracks propagate in the substrate/coating interface leading to chipping of the coating,
which implies that the fracture toughness of the interface is lower than that of the
coating. Thus, comparing Tables 3 and 4 it is expected that the method proposed by
Malzbender et al21 provides the most accurate data.
3.3. Brittle Coatings on Brittle Substrates: Oxide Coatings on Glass
In the previous section, clear through-thickness and interfacial fracture events were
observed form all of the coating/substrate combinations investigated. However,
although interfacial fracture is common in some systems it is by no means universal.
The requirement is that buckling of the coating occurs before there is excessive plastic
deformation of the substrate leading to through thickness cracking at the edge of the
impression prior to delamination during loading. In many cases this cannot be
achieved at practical indentation loads and scratch tests are necessary to cause any
coating detachment. However, the analysis of failure around a moving indenter is
more complicated than for a static indentation and there is no widely accepted
analysis method for interfacial toughness in the scratch test.
Oxide coatings on glass do show a range of though-thickness fracture features
which can be images by scanning electron microscopy if the indentation is large
enough (Figure 11). Radial cracking associated with the indenter edges is observed in
all cases (Figure 11a) and picture-frame cracking is also observed at higher loads
(Figure 11b). However, such cracking is invisible in smaller indents and it is not
always clear the conditions of load and indenter geometry necessary to initiate
fracture. For this reason there has been an interest in an on-line method for assessing
fracture in these small indentations and acoustic emission analysis has been devel-
oped with this in mind.
A range oxide materials which are used in low emissivity and solar control
coatings have been deposited on glass to 200 nm thickness and the acoustic emission
response monitored during nanoindentation testing. In addition a 200 nm layer on
zirconium metal was deposited on the float glass substrate and allowed to oxidise in air
36 S.J. Bull et al.
Figure 11. Scanning electron micrographs of high load nanoindentations in multilayer oxide coatings on
glass (a) 250 mN indent with radial splitting along the indenter edges and (b)500 mN indent showing a
combination of radial splitting and picture frame cracking.
prior to indentation testing. This is expected to have a higher toughness than the
deposited oxides. Figure 12 shows that the first load at which acoustic events are
observed above threshold is highest for this coating which would follow if the acoustic
events are associated with fracture. Also a multilayer optical stack ending in a 200 nm
thick TiOx Ny layer has been deposited for comparison with the single layer coatings;
results are comparable but not identical implying that the substrate is influencing the
fracture behaviour of the coating. Indentations in the uncoated substrate showed no
acoustic emission events over the complete load range investigated.
AE signals result from he sudden release of elastic energy or from surface
interactions such as indenter/surface friction or adhesion. In the case of static
indentation testing the release of elastic energy is the dominant mechanism for AE
formation; sudden release of elastic energy can arise due to unstable crack growth,
high speed phase transformations and plastic instabilities including the initiation of
new slip26 or twinning systems, the activation of dislocation sources27 of the sudden
acceleration or unpinning of existing dislocations.28 The substrate and coatings
investigated here are either amorphous and show poor crystallinity and it is unlikely
950
900
850
Load (μN)
800
750
700
650
600
550
TiOxNy
ZnO
Zr(O)
SnO2
Cap layer
TiOxNy
Figure 12. Graph of the load at which the first acoustic emission event is observed for the samples above
indicated.
200nm ZnO on 20nm TiOxNy on Float Glass

4000
Event 7
3500 P-δcurve
Load (μN) 3000 AE events
2500
Event 8
2000 Event 1
1500
1000 Event 12
500
0
0 50 100 150 200 250
Displacement (nm)
Figure 13. Load displacement curve for 4000 mN peak load nanoindentationin 200 nm ZnO coated glass
where the positions of the acoustic events are indicated.
that dislocation mechanisms will dominate behaviour, particularly as many events

occur on unloading as well as on loading (Figure 13). In addition there are no
acoustic events observed in the testing of uncoated glass at the same loads and the
first events occur well after plastic deformation has initiated and a permanent
impression is created. It is most likely that the acoustic emission arises as a result
of fracture and the ZnO and SnO2 layers would therefore appear to have the lowest
toughness if it is assumed that the defect distributions within these coatings do not
vary greatly. This is a reasonable assumption since they were all deposited in the
same coating system using very similar coating conditions.
Detailed characterisation of the acoustic emission waveform of each of the
events in terms of AEST and tat reveals that there are two different types of events
occurring, a fact which is mirrored for all the coated samples investigated here
(Figure 14). The event with the lower AESF always occurs first in the load-displace-
ment curve and always in groups of three whereas the other event is more random.
The fact that the event with the emission shape factor around two occurs in both
the loading and unloading curves is strong evidence that this event is associated with the
radial splitting of the coating under the indenter. Radial fracture is expected since
the hoop stresses in the material tested with a conical indenter are tensile and this
fracture can extend outside the impression on unloading due to the residual stresses
created by the plastic deformation which has occurred, akin to the mechanism for the
growth of radial cracks in microindentation testing of bulk glasses and ceramics29.
One other interesting observation is that the failures with the lower AESF tend to
occur in threes. This is not expected since the indenter is nominally a 1208 cone with
no edges. However, a detailed analysis of the indenter geometry reveals that it does
not have a circular cross-section but shows three distinct lobes (possibly related to the
effects of crystal structure on the ease of polishing) which are visible in the shape of
indents made at high load in softer materials. Such lobes will act as stress concen-
trations like the indenter edges of a Berkovich indenter and will therefore promote
fracture. This is added evidence that radial fracture is associated with this failure
mode. The higher emission shape factor failure occurs more randomly and at higher
loads and is less influence by the detailed indenter geometry as might be expected for
38 S.J. Bull et al.
1000
Amplitude (microV)
500
0
−500 AESF=2.13
−1000 Event 1
tat=0.053ms
−1500
0 0.02 0.04 0.06 0.08 0.1
(a) Time (ms)
1500
Amplitude (microV)
AESF=3.66
1000 Event 7
tat=0.067ms
500
0
−500
−1000
0 0.02 0.04 0.06 0.08 0.1
(b) Time (ms)
1000
Amplitude (microV)
500
0
−500
AESF=23.66
−1000 tat=0.045ms Event 8
−1500
0 0.02 0.04 0.06 0.08 0.1
(c) Time (ms)
Amplitude (microV)
800 AESF=3.392
tat=0.061ms Event 12
400
−400
0 0.02 0.04 0.06 0.08 0.1

(d) Time (ms)
Figure 14. Analysis of the acoustic function for events (a) 1, (b) 7, (c) 8 and (d) 12 in Figure 13. It is clearly
observed that event 1 and 8 are the same kind of event since the acoustic wave is similar and the emission
shape factor and attenuation time are of the same order of magnitude. The same happens for events 7 and
12, although the similarity is not so pronounced.
picture frame cracking. However, no evidence for fracture was observed in the in situ
AFM images obtained with the indenter after making the impressions so assignment
of the failure types must be treated with caution.
It is clear from the fact that a number of similar acoustic events are observed in a
single load displacement curve that the fracture occurring during indentation is
discontinuous and statistically determined depending on the defects in the coating
and substrate. If estimates of the fracture toughness of the coatings are to be achieved
using this approach then it is necessary to relate the acoustic events to the extra work
of fracture included in the irreversible work of indentation in a similar manner to that
outlined previously for CNx coatings. However, Figure 14 clearly shows that there
are no obvious pop-in events in the load displacement curve and any changes are too
subtle to discern in the raw data. To identify if such events exist the curve must be
processed to reveal these changes.
One approach to achieving this is to plot load against displacement squared30
(Figure 15). For an ideal material this ratio should be constant if elastic-plastic
deformation occurs but may vary gradually from one value to another as plastic
deformation extends from the coating material to the substrate. Any localised excur-
sions from this behaviour represent a change in deformation mechanism as might be
caused by fracture. The data from Figure 14 analysed in this manner in Figure 15 show
a large number of excursions occur in the loading curve for 200 nm ZnO on glass.
However, only seven acoustic events occur during loading and the correlation with the
excursions is poor. In general, the number of acoustic emission events increases as the
number of excursions increases. It is probable that there are many more fracture events
occurring during the indentation cycle than give rise to measurable acoustic emission.
The chances of relating a given acoustic event to the detailed changes in the load-
displacement curve are therefore very small. In addition, since the length of the radial
crack and how it grows are not known (though are probably related to indent size)
determining crack area and toughness by this method is not possible.
4. CONCLUSIONS
Fracture is critical in determining the performance of ceramic coatings in a range of

applications. Both through-thickness and interfacial fracture can occur but the
former is far more likely to be produced in indentation tests. Both through-thickness
200
150
100
Kexp (GPa)
50
−50
−100
0 50 100 150 200
Depth (nm)
Figure 15. Variation of Kexp ( P=d2 ) with displacement during loading for the loading curve in
Figure 13.
40 S.J. Bull et al.
radial fracture (associated with the indenter edges) and picture frame cracking are
observed for most coating/substrate systems. The radial fracture occurs at the lowest
loads and is the first fracture event to be observed in indentation tests. Radial crack
growth is discontinuous and occurs both on loading and unloading of the indenter.
Acoustic emission may be used to monitor fracture events in coated systems though
not all such events produce a measurable acoustic signal. Emission shape factors can
be used to differentiate between different types of events which are related to
different fracture modes.
Calculation of the fracture toughness of the coating and the interfacial tough-
ness can be undertaken in some cases where the deformation mechanisms can be
identified and the type of failure produced is amenable to analysis. However, the
analysis methods cannot be regarded as universal as the failure modes vary with
materials and are often more complex than the models allow for. As the thickness of
the coating is reduced the calculation of toughness is also made more difficult by the
need to obtain accurate measurements of crack area and differences in the work of
indentation which are beyond the resolution of current instrumentation.
ACKNOWLEDGEMENTS
The authors would like to thank Ion Bond, Linkoping University and Pilkington
Technology Centre for the provision of samples and Martin McGurk for the provi-
sion of some data.
REFERENCES
1. A.C. Fisher Cripps, Vacuum, 58 (2000) 569.
2. B. Bhushan, in Handbook of Micro Nanotribology, Ed. B. Bhushan, CRC Press, Boca Raton, 1995,
pp321 396.
3. S.V. Hainsworth, H. Chandler and T.F. Page, J. Mater. Res., 11 (1996) 1987.
4. J. den Toonder, J. Malzbender, G. de With and R. Balkenende, J. Mater. Res., 17 (2002) 224.
5. S.V. Hainsworth, M.R. McGurk and T.F. Page, Surf. Coat. Technol., 102 (1998) 97.
6. MrR. McGurk, PhD Thesis, University of Newcastle, (1997).
7. N. Hellgren M.P. Johansson, E. Broitman, P. Sandstrom, L. Hultman and J. E. Sundgren, Phys.
Rev., B59 (1999) 5162.
8. W.C. Oliver and G.M. Pharr, J. Mater. Res., 7 (1992) 1564.
9. S.V. Hainsworth, S.J. Bull and T.F. Page, Mat. Res. Soc. Symp. Proc. 522 (1998) 433.
10. A. Daugela, H. Kutomi and T.J. Wyrobek, Z. Metall., 92 (2001) 1052.
11. N.I. Tymiak, A. Daugela, T.J. Wyrobek and O.L.Warren, Mat. Res. Soc. Symp. Proc. Fall Meeting
(2002), in press.
12. J. Crepin, T. Bretheau, D. Caldemaison and F. Ferrer, Acta Materalia, 48, (2000) 505.
13. S.J. Bull and A.M. Korsunsky, Tribology Int., 31 (1999) 547.
14. S.J. Bull, J. Vac. Sci. Technol., A19, (2001) 1404.
15. S.J. Bull, Thin Solid Films, 398 399, (2001) 291.
16. B.R. Lawn, A.G. Evans and D.B. Marshall, J. Am. Ceram. Soc., 63, 574 (1980).
17. M.D. Thouless, Eng. Fract. Mech., 61 (1998) 75.
18. X.Li, D. Diao and B. Bhushan, Acta Mater., 44 (1997) 4453.
19. X.Li, D. Diao and B. Bhushan, Thin Solid Films, 315 (1998) 214.
20. J. Malzbender, G. de With and J.M.J. den Toonder, J. Mater. Res., 15 (2000) 1209.
21. J. Malzbender and G. de With, Surf. Coat. Technol., 154 (2002) 21.
22. J.W. Hutchinson and Z. Suo, Adv. Appl. Mech., 29 (1992) 63.
23. L.G. Rosenfeld , J.E. Ritter, T.J. Lardner and M.R. Lin, J. Appl. Phys., 67 (1990) 3291.
24. B.R. Lawn and E.R. Fuller, J. Mater. Sci., 19 (1984) 4061.
25. G.R. Anstis, P. Chantikul, B.R. Lawn and D.B. Marshall, J. Am. Ceram. Soc., 64 (1981) 533.
26. D. Rouby, P. Fleischman and C. Duvergier, Phil. Mag. A47 (1983) 671.
27. V.D. Natsik and K.A. Chishko, Sov. Phys. Solid State, 20 (1978) 1117.
28. J. Weiss and J.R. Grasso, J. Phys. Chem., 101 (1997) 6113.
29. B.R. Lawn and R. Wilshaw, J. Mater. Sci., 10 (1975) 1049.
30. M.McGurk and T.F. Page, J. Mater. Res., 14 (1999) 2283.
FOR GLASSES: YES, THERE IS
A FRACTURE CONTRIBUTION
George D. Quinna
ABSTRACT
The influence of cracking upon apparent Knoop hardness was investigated for five
glasses. The Knoop hardnesses decreased with increasing load in accordance with the
classic indentation size effect (ISE). Cracking commenced at moderate loads and, in
three of the five glasses studied, altered the indentation sizes and the ISE trends.
Cracked indentations were as much as 10 mm longer than uncracked indentations
made under identical conditions. Cracking around the indentation enabled the
indenter to penetrate deeper into the material and larger indentations were produced.
1. INTRODUCTION
Does cracking influence hardness? This question was raised by M. Swain and
M. Wittling at the 1995 Fracture Mechanics of Ceramics conference in Karlsruhe,
Germany with their paper titled: ‘‘Indentation Size Effects for Brittle Materials: Is
There A Simple Fracture Mechanics Explanation?’’1 They contended that median
cracks beneath an indentation could enable an indenter to sink deeper into the
material and thereby produce larger indentations than if there were no cracking.
They suggested that Knoop indentations should be particularly susceptible, but they
furnished no Knoop data to support their hypothesis.
This paper will show that cracking does affect Knoop hardness. Figure 1 shows
side-by-side Knoop indentations that were made by the identical procedure in an
a
George D. Quinn, Ceramics Division, Stop 852, National Institute for Standards and Technology,
Gaithersburg, MD 20899, [email protected].
149
150 George D. Quinn
Figure 1. Dark field views of 49 N indentations made in the 1723 aluminosilicate glass. The cracked
indentations were 9:0 mm longer on average at this load. Despite the intensive cracking on the sides of the
indentation on the right, the tips were undisturbed and the overall diagonal length could easily be read.
aluminosilicate glass. The cracked indentation has a longer diagonal length. This
paper reports results of a study to evaluate effect of cracking upon conventionally
measured Knoop hardness and the indentation size effect trend.
Hardness tests characterize a brittle material’s resistance to deformation, densi-
fication, displacement, and fracture. Whether cracking affects hardness or not de-
pends upon the indentation load. Cracking usually is not a factor at low loads. At
moderate loads, incidental cracking has little effect upon apparent hardness. At large
loads cracking can become so extensive that the indentation is so shattered or spalled
that diagonal lengths cannot be read at all. This is more of a problem for Vickers
indentations than for Knoop indentations. This paper concerns the effects of crack-
ing in and around Knoop indentations at loads commonly used by conventional
microindentation hardness testing machines (1 N to 50 N). A discussion of whether
cracked indentations are valid is deferred to the discussion section. We begin first
with a brief review of the Knoop indentation size effect.
The indentation size effect (ISE) is a trend wherein hardness decreases with
increasing indentation size or indentation load as shown schematically in Figure
2.1,2,3,4,5,6,7,8,9,10,11,12,13 A plateau Knoop hardness is reached at loads from 5 N to
20 N in glasses. The ISE exists for both conventional Knoop and Vickers hardness,
but usually with different trends due to different amounts of deformation, densifica-
tion, displacement, displacement rate, and fracture induced by the two indenters. The
ISE has been variously attributed to test procedure artifacts, frictional forces, envir-
onmental effects, or various material responses including elastic recovery, densifica-
Indentation Size Effect For Glasses 151
Figure 2. The indentation size effect (ISE) trend whereby hardness decreases with increasing load. Knoop
(HK) and Vickers (HV) hardness trends are similar but HK and HV values cannot be interchanged. In very
brittle materials, the HV curve may reach a plateau hardness at a distinct transition point.
tion, plastic flow, viscous flow, strain rate effects, or cracking. The author believes
that they all play a role. Although there have been many studies of the effects of
hardness upon fracture (e.g., References 14,15,16), much less has been written on the
effects of cracking on hardness. Cracking may manifest itself as median, radial, or
lateral cracks. There may also be generalized crushing or micro fissuring beneath the
indentations. Micro fissuring and shear band fracture under an indentation are
important deformation components.16,17,18
Frederick Knoop and his colleagues developed the elongated pyramidal in-
denter as an alternative to the square base pyramidal Vickers indenter, in part to
overcome the cracking in brittle materials with the latter.19,20 They carefully selected
their diamond geometry and a definition of hardness (load normalized by the
unrecovered projected area) so that hardness of a variety of metals exhibited a
negligible ISE. Nonetheless, they observed a small ISE for flint glass between 1 N
and 9.8 N.20 Their 9.8 N flint glass indentations showed no signs of cracking, but as
we will show, this is just below the load where cracking affects hardness. Subse-
quent to their early work, it became apparent that many glasses and crystalline
ceramics do have an ISE.
Ideally, one should measure the entire hardness-load curve, but in practice users
often chose one reference or standard load for comparisons between materials. Is
there an optimum load for a standard test method? Indentations should be as large as
possible to reduce measurement uncertainties, yet not so large as to induce excessive
cracking that interferes with the measurement or destroys the indentation altogether.
152 George D. Quinn
The current world standard for glass hardness (e.g. ASTM C 730,21 DIN 52333,22
and ISO 938523) specify a 0.98 N indentation load. This load is rather small and
evidently was chosen to avoid any cracking at all. It is in a steep load dependent
portion of the ISE curve where hardness changes rapidly with load and machine or
measurement errors can have a pronounced effect upon the measured hardness. The
0.98 N indentation is far from universal and many researchers have used a variety of
indentation loads up to 9.8 N. In their review of the hardness of vitreous silica, Li
and Bradt4 pointed out the widely divergent results and indentation loads that have
been reported over the years.
During a recent study24 of the fracture toughness of an optical glass using
Knoop indentations for fracture toughness experiments, it was noticed that cracked
19.6 N indentations were as much as 11 mm longer than uncracked indentations
created under identical conditions. The present study was conducted to further
investigate the effects of cracking upon apparent Knoop hardnesses and the ISE
over a broad indentation load range.
2. MATERIALS
Table 1 lists the five commercial glassesb that were utilized as well as the manufac-
turers’ values of density, r; elastic modulus, E; and HK.1 hardness at :98 N.c Fracture
toughness, KIc , was from Reference 25. Test coupons approximately 50 mm 50 mm
in size were cut from 5 mm to 10 mm thick plates, slabs, or disks.
The fused silica was an optical grade, pure silica made by flame hydrolysis.
A polariscope examination through the polished test coupon faces and 6.3 mm
thickness showed no residual strains.
A polariscope examination of the 6.3 mm thick borosilicate test coupon revealed
some slight strains. These were eliminated by reannealing at 5758C for 1 hour
followed by slow 28C cool down per minute to 108C below the strain point and
then slow cooling to room temperature. This glass has a composition of 81 w% SiO2 ,
13 w% B2 O3 , 4 w% Na2 O, and 2 w% Al2 O3 .
The aluminosilicate glass was cut from a 9.5 mm polished plate and the polari-
scope examination revealed no residual strains. This glass is a heat shock resisting
glass with the ability to withstand higher temperatures than the borosilicates.26 The
glass has 57 w% SiO2 , 16 w% Al2 O3 , 10 w% CaO, 7 w% MgO, 6 w% BaO, and 4 w%
B 2 O3 .
The borosilicate crown glass test piece was a 76 mm diameter5:6 mm thick
polished disk cut from a larger block. This optical grade glass was specifically
annealed by the manufacturer to eliminate residual strains. Its composition was 69
w% SiO2 , 11 w% B2 O3 , 10 w% Na2 O, 7 w% K2 O, 2 w% BaO, 1 w% CeO, and 0.2 w%
CaO. This glass has much greater network modifier content than the borosilicate
glass, yet has a similar annealing point.
b
Certain commercial materials or equipment are identified in this paper to specify adequately the
experimental procedure. Such identification does not imply endorsement by the National Institute of
Standards and Technology nor does it imply that these materials or equipment are necessarily the best
for the purpose.
c
HK.1 is the customary designation for Knoop hardness at 0.1 kgf, or 0.98 N indentation load.
Table 1. Properties of the glasses studied. The manufacturer’s data for HK.1 for the first
three glasses are very different from our results, probably since the former were adjusted for
optical resolution limits in accordance with ASTM C 730.
r Annealing E HK.1 HK.1 KIC

p
(Mg=m3 ) Point, 8C (GPa) (GPa)* (GPa)** (MPa m)
Glass * * * Manuf. This study ***
Fused silica Corning 7940 2.20 1084 72.5 4.79 5.88 + .13 .741
Borosilicate Corning 7740 2.23 560 62.7 4.10 4.94 + .08 .760
Aluminosilicate Corning 1723 2.64 710 86.2 5.04 5.94 + .09 .846
Borosilicate Crown Schott BK 7 2.51 559 82. 5.98 5.90 + .10 .862
Soda lime silicate 2.5 514 70 5.20 + .11 .75
* Manufacturer’s data. HK.1 data for the first three glasses probably has a 7l/(2NA) correction.
** Average (GPa), + standard deviation (GPa) for 12 indentations. No correction for optical resolution limits.
*** With the exception of the soda lime silicate, the fracture toughness data are from double cantilever beam experiments
in vacuum by Wiederhorn and Roberts, Reference 25.
The soda lime coupon was cut from a 6 mm thick plate. Typical soda lime glass
properties are shown in Table 1. Residual strains were eliminated by annealing at
5208C for 1 hour followed by a slow 28C cool down to 108C below the strain point
and then a slow cool to room temperature. An examination with a black light failed
to detect tin on either surface.
3.1. Hardness Testing Machines

Special care was used in this study to obtain accurate and precise readings, otherwise
experimental errors or data scatter could mask or obscure the cracking effect.
Thibault and Nyquist’s masterful 1947 paper13 addressed many of the experimental
problems in making Knoop hardness measurements in brittle materials. Anyone who
has ever made and measured Knoop indentations will certainly appreciate their
comments:
As might be expected with Knoop indentations which, even under conditions of maximum
resolution, seem to fade gradually away instead of ending at any definite place, differences in the
visual acuity of different individuals may greatly affect the determined length of impression and
hence the Knoop number. So elusive are the ends of the indentations, particularly in transparent,
low index substances, we sometimes wonder to what extent imagination enters into the determin
ation. Of this we are certain: The measurement of Knoop indentations, made in hard substances
under low loads, is the most exasperating type of microscopy which we have ever experienced.
Harassed by the indistinctiveness of the indentation terminations which actually seem to vacillate
while being examined, and the realization that the utmost precision must be obtained, we have
never been able to spend more than a very few hours a day measuring Knoop impressions.
Having prepared the world’s first standard reference material for Knoop Hard-
ness of Ceramics, SRM 2831,d the author fully appreciates the metrological and
practical problems posed by the Knoop hardness method.27 A broad range of
d
Standard Reference Material 2830, Knoop Hardness of Ceramics, National Institute of Standards and
Technology, Gaithersburg, MD 20899.
154 George D. Quinn
indentation loads was used in the present study order to explore the full ISE curve.
Two indentation hardness machines were used. Initially there were problems ensur-
ing that both machines produced comparable results, but eventually we determined
that both machines were working properly and produced consistent indentations and
readings. Indentations made by both machines concurred within 0:5 mm to 1:0 mm
Some early difficulties that we had in comparing indentations made by the two
machines were traced to the use of a low power objective microscope lens for
measuring very large indentations as will be discussed below. Standard reference
materials, certified length standards, and load cells were used frequently to verify the
procedures. A full description of the experimental procedure is described elsewhere.28
A tabletop machinee was used for indentation loads from 0.98N to 19.6 N.
A larger floor mount machine was used for loads from 19.6 N to 98 N.f All inden-
tation diagonal lengths were measured with the optical system on the tabletop
machine. All hardnesses were computed from:
HK ¼ 14:229P=d2 (1)
where P is the indentation load and d is the long diagonal length.

The microindentation table, which was used to measure all indentations made
by both machines, had a 50X objective lens with a 0.5 numerical aperture. A green
filter was used in conjunction with a 10X digital filar eyepiece for a total magnifica-
tion of 500X to measure most of the diagonal lengths. An optical stage micrometer
with 10 mm and 100 mm divisions that had been calibrated by NIST to an accuracy
better than 0.05% was used to establish the correct scalar factors for the 50X lens and
the digital readout device. The repeatability of making a single diagonal length
reading was 0:4 mm-0:5 mm (standard deviation) or better. This was determined by
repeatedly reading single 9.8 N indentations that had been made in a soda lime
silicate specimen and a steel reference block. There was no need to correct for the
optical resolution limitations of the 50X objective lens. Load was applied carefully to
the indenter via a dashpot and care was taken to not touch the machine or the table it
was mounted on during an indentation cycle. The initial diamond clearance and
contact time were set in accordance with the manufacturer’s instructions, and hence
the indenter velocity was of the order of 0:042 mm=s---0:091 mm=sec depending upon
the applied load. Dwell time was set for 15 seconds, also in accordance with the
manufacturer’s instructions.
Larger indentations were made with the floor mounted hardness-testing ma-
chine that had a 98 N capacity. This dead weight loading machine applied force via a
dashpot controlled lever beam and was set in compliance with ASTM E 384 Micro-
hardness of Materials29 and ASTM C 1326.30 The initial diamond clearance and
contact time were set in accordance with the manufacturer’s instructions, and pro-
duced an average indenter velocity of 0.025 mm/s. Dwell time was 15 s 20 s after
which the machine automatically extracted the indenter. Indentations made with this
floor mounted larger capacity machine were measured within 10 minutes with the
tabletop machine in order to eliminate machine optics as a source of variability.
There was one important difference in procedure, however, that initially caused us
e
Miniload II, Wetzlar, Germany.
f
Tukon 300 floor model, Wilson Division of Instron, Canton, MA.
some difficulties. Indentations made at 29.6 and 49 N were so large (> 300 mm) that
they could not be measured with the tabletop machine’s 50X objective lens since they
extended beyond the field of view. Consequently, the large indentations were meas-
ured with the tabletop machine’s lower power 10X, 0.18 numerical aperture lens.
This lens and the digital filar scalar factors were also carefully checked with the NIST
calibrated optical stage micrometer. Unfortunately, it was more difficult to detect the
full length of the large indentations with the 10X low power objective lens due to
optical resolution limitations.
3.2. Measuring Knoop Diagonal Lengths: Problems Due to Resolution Limits
Problems in assessing the exact tip location have long bedeviled those who have read
Knoop indentation diagonal lengths, especially in glasses. Optical microscope reso-
lution limitations are a problem for Knoop indentations due to their slender tapered
tips.13,31,32,33 Thibault and Nyquist13 devoted seven full pages to this topic in their
1947 paper. Gahm33 presented compelling evidence that diagonal length readings are
strongly dependent upon the lens numerical aperture and errors of up to 4 mm are
possible with low power lenses. The resolution error in underestimating the true
Knoop diagonal length has been estimated as 7l=2(NA) where l is the wavelength
of light and NA is the objective lens numerical aperture. The error is independent of
indentation size, so the percentage error is greater for smaller indentations. So, for
example, the 50X objective lens used on the Leitz Miniload machine in the present
study had a 0.50 NA and green filtered light (l ¼ 0:55 mm). The calculated correc-
tion was 3:9 mm. This is a huge correction for small ( 50 mm) indentations made at
0.98 N in glasses, but is proportionately less important for larger indentations.
Tarasov and Thibault32 suggested that the entire ISE may be due to this correction
factor, since if a constant correction length is applied to diagonal lengths for a range
of indentation loads, it will produce a trend similar to the ISE.
A 7l=2(NA) correction is incorporated in the ASTM Knoop standard C 730 for
glass, but is not used in the master microindentation hardness of materials standard E
384, nor in the advanced ceramic standards C 1326 and ISO 14705, nor in the glass
standards DIN 52333 or ISO 9385. This variability between the standards has created
considerable confusion in hardness testing and probably accounts for a lot of
inconsistent results in the literature.
We investigated this matter further as described in detail in reference 28. With
good microscopy technique and proper illumination and diaphragm control, meas-
urement errors due to optical resolution limitations with 40X or 50X objective lenses
may be only 0:4 mm to 1:5 mm depending upon the observer and material. The full
7l=2(NA) correction factor is excessive. In contrast, a correction factor for reso-
lution limitations in our work was essential for readings made with the low power
10X objective lens. All glass diagonal length readings made with this lens had to be
corrected by þ5 mmg as discussed in detail in Reference 28.
g
The length differentials are much less with opaque ceramics. The same experiment was preformed on
the black silicon nitride SRM 2830 reference hardness block. Three observers repeatedly measured the
same indentation with 10X, 20X and 40X objectives lenses. They obtained differentials of only 0:3 mm to
1:5 mm when switching from the 10X to 40X objectives.
156 George D. Quinn
3.3. Evaluation of Cracking and Changes of the Indentations with Time
Cracking assessments were made with the optical microscope while measuring the
indentation size on the indented surface.h Lateral cracks were easily detected, as well
as generalized cracking around and underneath the indentation. Ideally, the inden-
tations should have been monitored through the glass specimen during the indenta-
tion process, as demonstrated by Cook and Pharr34 or Wilantewicz and Varner,35 but
this was beyond the scope of the present short study. The indentations were observed
in the same manner most researchers would with their conventional hardness testing
machines.
We noticed that cracks grew with time around some of the indentations. Lateral
crack shards popped off some large indentations. Nevertheless, the general assess-
ment of whether an indentation was cracked or not did not change over the course of
a few weeks. Specific comments about the nature and extent of the cracking are given
in the results section. Some indentation diagonal lengths shrank with time. A de-
crease of the indentation length of as much as several microns over the course of a
few days was not uncommon, so it was important to measure the indentation lengths
within a few minutes after they were made.
3.4. Modeling the ISE Trend

Various functions have been proposed to model the indentation size effect data. The
Meyer law is:
P ¼ cd n (2)
where c is a constant and n is the Meyer or logarithmic index. For ceramics and
glasses, n < 2, and hardness decreases with increasing load. An alternative expression
(which often fits glass or ceramic hardness data better) that is attributed to Bern-
hardt38 that is commonly used is:4,7,8,36,37,38,39
P ¼ a1 d þ a 2 d 2 (3)
where a1 and a2 are constants. The first term accounts for the ISE trend at low loads,
whereas the second term accounts for the constant hardness plateau at larger inden-
tation loads. Combining equations 1 and 3:
0
HK ¼ 14:229(a1 =d) þ 14:229a2 ¼ a01 =d þ a2 (4)
where 14.229 is the indenter constant for the Knoop indenter and correlates the long
diagonal length, d, to the projected surface area. If hardness is load independent
(a01 ¼ 0), then:
HK ¼ a02 (5)
h
The choice of the objective lens sometimes altered the cracking assessment. Lower power lenses with
greater depths of field revealed more cracking damage. For the sake of consistency, the assessments in this
study were made with the lens that was used for the diagonal length measurements.
The parameter a02 is considered by some to be the ‘‘true hardness’’ or the hardness
due to volumetric deformation processes in the absence of surface effects.2,7,8,12
4. RESULTS
4.1. Hardness Versus Indentation Load and Cracking Propensities

All five glasses exhibited the classic ISE of decreasing hardness with increasing load.
Cracking did not affect the ability to measure the Knoop indentation diagonal
lengths in any of these glasses since the cracking damage was along the sides or
underneath the indentation and there was very little or no damage or displacement at
the tips. Faint, very tight median cracks sometimes emanated from the indentation
tips, but since these often extended at a slight angle to the indentation axis, it was not
difficult to distinguish them from the indentation itself. This is in sharp contrast to
Vickers indentations, wherein tip cracking or spalling renders length measurements
difficult or impossible.
Figures 3 and 4 show the outcomes for the aluminosilicate and borosilicate
crown glasses wherein cracking had a distinct effect upon the apparent hardness.
Cracked indentations were longer than uncracked indentations by as much as 10 mm.
The scatter in diagonal lengths of uncracked indentations was quite small in either
case, whereas cracked indentations had much greater variability.i Although a 10 mm
length differential is quite pronounced and easy to measure, the percentage difference
in length was small. The difference in hardness is magnified, however, since hardness
depends on the square of the diagonal length. The difference in hardness between
uncracked and cracked indentation was 5%.
Cracking occurred in some aluminosilicate glass specimens at loads as small as
9.8 N, just as the ISE trend approached the plateau. Figure 1 shows a dark field view of
indentations. Dark field illumination sometimes revealed that the ‘‘uncracked’’ inden-
tation had some localized damage and possible short lateral cracks. The fully cracked
indentations were 9:0 mm longer on average for 49 N. Cracking gradually increased
for larger loads. Figure 3 shows two data sets at 19.6 N that were made by the two
different hardness machines. The average diagonal size and hardness are in good
agreement. A clear distinction between cracked and uncracked lengths was observed
with both machines. The differences in average diagonal lengths were only 2:3 mm to
4:2 mm at 19.6 N, but these differences are both statistically significant.j Even at the
largest loads, there were some indentations that had negligible cracking, however, and
a dichotomy in hardness trends occurred as shown in Figure 3b. Indentations were
sharp and had straight side edges and bottom edges at loads to 29.4 N. The indenta-
tion edges became more ragged at 49 N as extensive crushing commenced. Neverthe-
less, only faint median cracks could be detected from the tips even at these high loads.
Cracking also affected the hardness of the BK-7 borosilicate crown glass as
shown in Figure 4. No median, radial, or lateral cracks were detected at 4.9 N.
i
A few cracked indents had lengths similar to uncracked indentations, but usually the cracked indenta
tions were appreciably longer.
j
The two tailed student’s t statistic for small sample sizes was 8.96 for HK and 8.96 for d for data from
the floor mounted machine, and 5.70 and 5.81, respectively, for data from the tabletop machine. These are
greater than the critical t ratio for 10 degrees of freedom at the .001 probability level (t 4.59).
158 George D. Quinn
Figure 3. Diagonal length (a) and Knoop hardness (b) for the alumino silicate glass. Cracked indenta
tions were as much as 10 mm longer than uncracked indentations.
Indentations were sharp and had straight edges at loads up to 49 N. Lateral and
other cracks commenced at 9.8 N at the point where the ISE curve approached the
plateau and there was an immediate effect upon the diagonal lengths. Upon cracking,
Figure 4. Borosilicate crown BK 7 data. (a) shows diagonal length. (b) shows hardness, and (c) shows P/d
versus d.
the data bifurcated into two sets, both with small scatter. At 19.6 N, the cracked
indentations were 7:0 mm longer than the uncracked indentations. The hardness of
the uncracked indentations was 0.27 GPa (5.9 %) greater. Both of these differences
are statistically significant.k The cracked indentations had extended lateral and
localized cracking, whereas the uncracked indentations had only a slight disturbance
on the side of the impression. At loads of 29 N and above, nearly all indentations
were cracked. Figure 5 illustrates adjacent cracked and uncracked indentations made
k
The two tailed student’s t statistic for small sample sizes was 5.93 for HK and 5.59 for d. Both of these
are greater than the critical t ratio for 17 degrees of freedom at the .001 probability level (t 3.97).
160 George D. Quinn
Figure 5. Uncracked and cracked Knoop indentations in the BK 7 borosilicate crown glass at 19.6 N.
Indentations should be spaced further apart for most testing purposes in order to avoid interference. (a) is a
bright field and (b) dark field optical microscope illumination. The dark field photo reveals that some
localized cracking occurs under the left indentations, but the extent of cracking is dramatically different.
The cracked indentation is 7 mm longer.
at 19.6 N. The 7 mm length difference is not obvious from the photos, but is readily
detected with the hardness machine microscope. Figure 4c shows a P/d versus d
graph, which in accordance with equation 3 should have a slope of a2 and an
intercept of a1 . The cracked and uncracked data sets have different lines, but if the
data were combined into one set, there is no obvious inflection point since the
transition of uncracked to cracked indentations occurs gradually over a range of
loads.
Figure 6 shows that in contrast to the above results, cracking had negligible
effect on hardness for the soda lime and borosilicate glasses. Lateral cracks devel-
oped in a few soda lime silicate indentations at 14.7 N and grew progressively more
Figure 6. Soda lime silicate (a) and borosilicate 7740 (b) hardness data. Cracking had no effect upon
hardness for either material.
162 George D. Quinn
prominent at larger loads. Significant lateral cracking and crushing occurred at

29.4 N and larger loads. All of the indentations had straight, sharp, and clear
edges and bottoms. Shear bands could be discerned through the sides of the inden-
tation. Despite all this damage, no median or radial cracks were detected on the
indentation surface at loads to 49 N.
Cracking also had no effect upon apparent hardness for the borosilicate 7740
glass (Figure 6b). Very localized cracking was observed as close parallel circular arcs
on the sides of the borosilicate glass indentations at loads of 4.9 N and above
(Figure 7). These may be cone cracks, or possibly shear band cracks due to the
large material displacement and uplift along the middle sides of the Knoop inden-
tation. No median cracks were detected. Lateral cracks only formed at the largest
loads (49 N) by which time the hardness-load curve had reached a plateau. One
curious aspect of these indentations was a significant curvature and pinching in
along the sides as shown in Figure 7. Significant elastic spring back probably caused
the stiletto appearance.
The fused silica hardness was sensitive to cracking (Figure 8), but this glass was
different than the other glasses. Extensive ring-like cone cracks formed on the
indentation sides at 9.8 N and above. At larger loads numerous multiple concentric
ring cracks appeared. The ring crack damage was much more severe than that
detected in the 7740 borosilicate glass. No cracking was detected at the indention
tips. The difference in the average diagonal lengths between cracked and uncracked
indentations was small: 1:1 mm and 1:9 mm at 4.9 N and 9.8 N indentation loads,
respectively. A greater than normal number of indentations were made at these two
loads to confirm this difference. Fragments around the indentation sides spalled off
Figure 7. Indentations in the 7740 borosilicate glass. Minor cone or ring cracks form along the sides of
the 4.9 N indentations, which have a noticeable pinching in of the sides.
Figure 8. Hardness of fused silica. Cracking did affect hardness in this glass.
at loads of 14.7 N and greater. Major portions of the indentation were lost in some
indentations at 19.6 N yet the tips were still intact on the surface. Three separate data
sets were generated by the two hardness machines at 19.6 N for a total of 32
measurements. The results concurred. Some of these indentations still had intact
indentation bottoms (the cross was still intact), but others had shattered or crushed
bottoms. At 29.4 N, the indentation cores were shattered and extensive crushing
around the indentation bottom had occurred. In a few indentations, a remnant of the
indentation bottom edge was still visible, but skewed relative to the initial orientation
of the indentation, attesting to massive displacement and uplift of the indentation
bottom. Large scallop cracks popped off the indentation sides. Despite all this
damage, the indentation tips were undisturbed and easy to read. No median cracks
were detected even for the 49 N load. Cracked indentations were curved or pinched
in the middle and did not have the well-defined pyramid shape, evidently a conse-
quence of major elastic spring back of the indentation sides. Furthermore, unlike all
the other glasses in this study, equation 4 did not fit the data quite so well and the
data did not reach a distinct plateau by 15 N to 20 N.
4.2. Hardness at the Standard Load of 0.98 N
Table 1 has our HK values for 0.98 N (the customary load in the world hardness
standards) as well as the manufacturers’ own data. The BK-7 borosilicate crown
hardness at 0.98 N concurs since neither the manufacturer nor we applied a correc-
tion factor for optical resolution limits. In contrast, our hardnesses do not agree with
the manufacturer’s listings for the first three glasses, possibly since the manufacturer
164 George D. Quinn
applied correction factors for the optical resolution limits in accordance with ASTM
C 730.
4.3. Indentation Size Effect Parameters

The ISE curve parameters for equation 4 are given in Table 2. The plateau hardness
at large loads is a02 . Single entries are shown for the two glasses for which cracking
had no effect on hardness. Two entries are shown for the three glasses that were
sensitive to cracking. The uncracked borosilicate crown data did not fit equation 4
very well, so no a01 parameter is listed and only an approximate estimate of a02 is
shown.
The hardness of the fused silica at 1.96 N (5:42 GPa 0:04 GPa) almost exactly
matches the 5.41 GPa reported by Hirao and Tomozawa11 for a 20 second dwell time
in air. Li and Bradt4 analyzed Hirao and Tomozawa’s results that were obtained in a
variety of environments and concluded that the best estimate for the load independ-
ent hardness, a02 , for fused silica was 5.30 GPa. Individual data sets for tests in
different environments produced estimates from 4.88 to 5.30 GPa with uncertainties
(one standard deviation) of 10%. Our new estimate of a02 for uncracked indenta-
tions (4.74 GPa) is a little below this range. This is because our estimate is based on
data from a much broader range of larger indentation loads than utilized in the
former studies that had a maximum of only 2 N.
5. DISCUSSION
These findings confirm that cracking affects the hardness response of some but not
all glasses. Figure 9 shows the general behavior. Cracking begins to affect HK at
about the load where the ISE curve reaches a plateau. Swain and Wittling1 suggested
that median crack opening could account for larger penetrations. Our work suggests
that lateral material displacement and lateral cracking can also facilitate greater
indentation penetration. The cracking effect on HK can be easily detected with
careful but routine testing, but is small and probably has been ignored or lost in
data scatter in earlier studies. Many of the earlier studies utilized microhardness
Table 2. Effect of cracking, ISE curve parameters, and brittleness.
Did
cracking Primary Brittleness,
0 0
affect Crack a1 a2 B
Glass HK? Type Data set (GPa N) (GPa) (108 m1 )
Fused silica 7940 Yes Ring / Cone uncracked 1.16 4.74 6.28
cracked 2.90 4.32
Borosilicate 7740 No Cone / Other 1.00 3.93 4.27
Aluminosilicate 1723 Yes Lateral uncracked 1.77 5.13 6.12
cracked 2.16 4.94
Borosilicate uncracked 4.9 5.43
Crown BK 7 Yes Lateral cracked 2.45 4.51
Soda lime silicate No Lateral 0.89 4.34 5.40
Not determined, or not applicable.

machines with load limits of only 9.8 N or 19.6 N and thus missed the onset of
cracking. Many didn’t reach the hardness plateau which usually occurs at 10 N or
greater. Some investigators arbitrarily disregarded cracked indentations.
Several factors could explain why cracking affected the hardness of some glasses
and not others. The soda lime and borosilicate glasses were the two softest glasses in
the study and were not affected by cracking. They also had the lowest elastic moduli.
Evidently the harder and more rigid the glass is, the more likely that cracking affects
Knoop hardness.
The timing of the crack formation may also determine whether the cracking
affects hardness. Cracks that form during load application probably increase the
local compliance of the material and enable the indenter to sink in deeper. Cracks
which form during load removal probably do not affect the Knoop indentation tip
positions and hence the indentation diagonal length. Cook and Pharr33 observed
Vickers indentation cracking sequences in various glasses and ceramics. Their study
and ours have four glasses in common. If the same cracking sequences they detected
for Vickers indentations are the same as for Knoop indentations, then the timing of
the crack formation would account for our results in only two of the four cases,
however. Cracking during loading in the fused silica could account for our longer
indentations. Cracking during unloading could account for why the soda lime glass
cracking did not affect the diagonal lengths and hardness. The Vickers cracking
sequences do not match the observed effect on Knoop hardnesses for the other two
glasses in common with our study (aluminosilicate and borosilicate). Perhaps the
difference in indenter geometries causes different fracture sequences and effects upon
apparent hardness.
Arora et al.40 characterized glasses into so-called ‘‘normal’’ and ‘‘anomalous’’
categories based upon the behavior during Vickers indentation. ‘‘Anomalous’’
glasses are silica rich and densify during Vickers indentation. Once they do crack,
they form ring cracks and steep, deeply penetrating cone cracks. ‘‘Normal’’ glasses
Figure 9. Cracking may (a) affect the Knoop ISE trend, or as shown in (b), may not affect the ISE.
166 George D. Quinn
have greater network modifier content and tend to form shear deformation bands
and shear micro fissures. Of the two ‘‘anomalous’’ glasses in our study (fused silica
and borosilicate), only the former glass exhibited Knoop hardness sensitivity to
cracking. Of the three ‘‘normal’’ glasses in this study, two were sensitive to cracking
(borosilicate crown, aluminosilicate) and one (soda lime silicate) was not. Thus the
normal and anomalous Vickers categorization scheme does not correlate with the
present Knoop results
An energy model may account for the different behaviors. Equation 3 may be
expressed with units of energy by multiplying both sides by d:
Pd ¼ a1 d 2 þ a2 d 3 (8)
Ð
The term Pd is proportional to the external work done by the indenter, Pd‘, where
‘, the indenter penetration depth, is proportional to d. A number of investigators
have attempted to correlate surface energy or fracture processes to the a1 d2
term.1,2,4,5,6,8,11,36,37 The a2 d3 term, on the other hand, is considered to be the
‘‘work of permanent deformation’’5,12,37 or the ‘‘volume energy of deformation.’’2,4,5
In their work on the Knoop hardness of fused silica, Hirao and Tomozawa11 showed
a very strong ISE and argued that the Knoop a1 term was related to the propensity of
the glass to crack. Cracking propensity was estimated from a few Vickers experi-
ments at loads only up to 1.2 N, however, and Knoop indentation cracking was not
actually evaluated. Table 2 shows that the three glasses in which cracking altered
0
Knoop hardness did indeed have larger a1 values.
A handful of other studies have detected an effect of cracking on the ISE.
Fröhlich et al.39 noted distinct cracking events effects during recording Vickers
hardness experiments in germanium single crystals, but their work was limited to a
maximum load of 2 N. The load versus displacement traces showed discrete load
drop offs when cracks popped in and they also associated the a1 term to cracking and
the surface energy of the indentation. Most instrumented (depth sensing) hardness
experiments do not discern the onset of cracking during loading. Gahm33 obtained a
1
subtle shift in a peak P 2 versus final d curves for two glasses once cracking com-
menced, but these were at very light loads of 0.5 N. At much greater loads up to
40 N, Zeng and Rowcliffe41 observed a significant increase in the elastic compliance
of Vickers indentations during unloading once significant cracking had occurred.
They even proposed a damage parameter based upon the extent of the compliance
change. Yoo et al.42 reported a fascinating example of how cracking contributes to
the ISE in MgO single crystals. At low loads, deformation was controlled by
dislocation slip and pileups. At larger loads, cracks acted as dislocation sinks that
relieved the pileups around the indentation. This led to enhanced plasticity and lower
hardness.
A study of Vickers hardness of polycrystalline ceramics revealed that cracking
may cause critical transition points in the Vickers ISE trends.8 The transition point
was associated with extensive cracking in and around the indentation and a shift in
the energy balance during indentation. Different ratios of the indentation work are
expended on volumetric deformation and surface fracture processes above and below
the transition point. The transition point was very distinct for brittle materials such
as silicon carbide. The Vickers hardness transition point was related to a new index of
ceramic brittleness defined as:8
HE H
B¼ 2
/ (9)
KIc 2gf
where H is hardness at the plateau, E is the elastic modulus, KIc is the fracture
toughness and gf is the fracture surface energy. Hardness has units of energy per unit
volume and may be considered to be the work to deform a unit volume, whereas gf is
the work to create unit surface area. B is a ratio of volume deformation energy to the
surface fracture energy. Table 2 lists values of B for the five glasses using the
uncracked Knoop plateau hardness value a02 for H in equation 9. With the exception
of the borosilicate crown glass, the more brittle the glass (higher B), the more likely
cracking alters the hardness.
Is a cracked indentation a valid indentation? Most researchers believe that
cracked indentations do not yield the correct hardness values. Most standard test
methods call for the rejection of cracked indentations. ASTM C 1326 and ISO 14705
for ceramics43 allow limited cracking and furnish guidance on interpretation.l ISO
9385 for glass cautions against excessive fracture. On the other hand, to the extent
that hardness is deemed a ‘‘measure of resistance of penetration by a prescribed
indenter under prescribed conditions,’’ then cracked indentations are legitimate and
simply manifest a different material response to indenter penetration. Indeed, sec-
tional views of indentations shown by Hagan and Swain17,18 clearly show that micro
fissuring and cracked shear bands underneath indentations are important compon-
ents of deformation even at low loads. One remarkable aspect of the present study is
that despite the dramatic cracking at large loads, in no case did it interfere with the
ability to see and measure the Knoop tip-to-tip distances since the tips extend well
beyond the damage zones. Median and radial cracks from the tips were only detected
in a few specimens at the largest loads and were usually easily distinguished from the
indentation.
Fifty-five years ago Thibault and Nyquist13 called attention to the effect of
cracking on hardness in their masterful paper on the metrology of Knoop hardness
measurements. They listed six categories of cracking from no cracking (class I) to
pronounced cracking such that large portions of the specimen were chipped away
(class 6). They relied on direct examination of the indented surface as was done in the
present study. Cracking had no measurable effect on the Knoop hardness of silicon
carbide or topaz for up to class 3 severity: large amount of subsurface cracking, but
none on the surface; or minor subsurface cracking and slight surface cracking.
Cracking altered results for topaz once it reached class 4 severity: a large amount
of subsurface cracking accompanied by slight to moderate surface cracking. Cracked
indentations were longer than uncracked indentations and the apparent hardness was
consequently lower. They concluded that Knoop hardnesses based on class 1-3
indentations were not affected by fracturing, class 4 and 5 indentations should be
avoided, and class 6 indentations should not be used at all.
Notwithstanding claims to the contrary, the evidence is compelling that the ISE
is real and not due to a single phenomenon. Other studies have presented strong
evidence that experimental errors (e.g., optical resolution limitations), frictional
l
Limited cracking was acceptable provided that it did not alter the indentation size compared to
uncracked indentations, did not remove portions of the indentation, nor obscure or broaden the indenta
tion tips to the extent that tip location could not be reliably assessed.
168 George D. Quinn
forces, environmental effects, or various material responses including elastic recov-

ery, densification, plastic flow, and viscous flow contribute to the ISE. This paper has
demonstrated that cracking also can alter the hardness versus load trends. If the
cracking is ignored and combined data analyzed, the increased scatter will shift the
plateau hardness value by up to 5 %.
The standard test methods for Knoop indentation hardness of glass have
specifications that warrant some comment. The correction factor for diagonal length,
7l=2NA, for optical resolution limitations that is included in ASTM C 730 may need
review. This correction factor adds 3 mm or more to the length of an indentation
irrespective of its size and can drastically change the hardness number. Uncertainty
over whether to add the correction factor or not has caused confusion in reported
glass hardness values. For example, the manufacturer of the borosilicate crown glass
used in this study formerly published dual hardness tables for their glasses.44 Hard-
ness values for the borosilicate crown glass were 5.88 GPa without the 3 mm correc-
tion and 5.29 GPa with the correction, a difference of 11%! Our limited
investigations28 suggest that a 7l=2NA correction is not needed with 40X or 50X
objective lenses, but a correction is needed for lower power lenses especially for
glasses.
The formal glass hardness standard test methods all recommend 0.98 N (100 gf)
as the standard test force. Although this is a small load it produces moderate-sized
( 50 mm) indentations since glasses are not especially hard. Nevertheless, it is merely
one point on the very steep portion of the ISE curve. Hardness at a single load may
have appeal for routine purposes, but clearly much more useful information is over-
looked. Perhaps the glass hardness standards should recommend measurements over
a broader load range. ISO 9385 takes a step in this direction by recommending
multiple indentation loads.
One last matter deserves a few comments. Cracking appears to affect Vickers
and Knoop indentations in different ways as illustrated in Figure 10. Extensive
cracking underneath Vickers indentations seem to cause the ISE to have a transition
point to a constant hardness plateau as described by Quinn and Quinn.8 The effect is
most pronounced for very brittle ceramics. On the other hand, the present study
shows that cracking around Knoop indentations causes the ISE trend to shift to
lower hardness values, but only in some materials. A possible explanation for the
differences may be made with reference to the schematics in Figure 10.
At low Knoop indentation loads, cracking damage may be confined to a small
zone underneath the indentation. Once significant cracking occurs, material is dis-
placed from underneath the indentation. The indenter sinks deeper into the material,
but then becomes supported on its ends like a bridge. The indentation develops a
longer diagonal length (2Dd) than if the cracking had not occurred.
Alternatively, for Vickers indentations, at low loads cracking may also be
limited to microcracking underneath the indentation. With increasing load, the
extent and volume of the cracking increases. Above a critical load, the micro fissuring
damage underneath the indentation may become saturated and reach up to the
surface along the sides of the indentation. Material displacement and shear may
cause significant uplift on the surface. At this point the balance of indentation energy
expended on volumetric and surface deformation damage processes may change. Van
der Swagg and Hagan45 showed pictures of just such a progression of damage
underneath conical indentations as indentation load increased in fused silica and
soda lime glass. Figure 11 shows just how pronounced the uplift may be around
indentations.
Hagan’s method of cleaving through an indentation is an effective way to study
damage beneath.17,18,45,46 Future work could apply his approach to help correlate
cracking damage to shifts in the ISE trends for Knoop and Vickers hardness.
Figure 10. The effect of cracking on the ISE trends for Knoop (top row) and Vickers (bottom row)
hardness.
Figure 11. Illustration of the massive uplift on the sides of 98 N and 74 N Vickers indentions in silicon
nitride. The specimen is strongly tilted and was photographed by a stereo microscope. The magnification
bar is only approximate.
170 George D. Quinn
6. CONCUSIONS
Cracking around the indentation affects the Knoop hardness and ISE in some
glasses. Cracked indentations are larger than uncracked indentations by as much as
10 mm, which reduces the Knoop hardness by 5%. Harder, more rigid glasses are
more apt to be affected by cracking, but the timing of the crack formation or the
brittleness of the glasses may also be contributing factors. Cracking is not the sole
cause of the ISE, but does contribute to the trend and affects the plateau hardness.
Measurements of Knoop hardness over a broad load range are preferable to a
measurement at a single load. Care must be taken to account for optical resolution
limitations and diagonal length underestimates when using low power optical micro-
scope objective lenses.
REFERENCES
1. M. Swain and M. Wittling, Indentation Size Effects for Brittle Materials: Is There A Simple Fracture
Mechanics Explanation, in Fracture Mechanics of Ceramics, Vol 11, eds. R. C. Bradt, D. P. H.
Hasselman, D. Munz, M. Sakai, and V. Y. Shevchenko, (Plenum Press, NY, 1996), pp. 379 388.
2. H. Li and R. C. Bradt, The Microhardness Indentation Size Load Effect (ISE) in Hard Ceramic
Materials, J. Hard Mat., 3 (3 4) 403 419 (1992).
3. H. Li and R. C. Bradt, The Effect of Indentation Induced Cracking on the Apparent Microhardness,
J. Mat. Sci., 31 1065 1070 (1996).
4. H. Li and R. C. Bradt, The Indentation Load/Size Effect and the Measurement of the Hardness of
Vitreous Silica, J. Non Cryst. Sol., 146 197 212 (1992).
5. T. Sakai, A. Ghosh, and R. C. Bradt, The Indentation Fracture Resistance of Self Reinforced
Mullites, in Fracture Mechanics of Ceramics, Vol. 10, ed. R. C. Bradt et al., (Plenum Press, NY,
1992), pp. 119 132.
6. H. Li and R. C. Bradt, The Microhardness Indentation Load/Size Effect in Rutile and Cassiterite
Single Crystals, J. Mater. Sci., 28 917 926 (1993).
7. H. Li, A. Ghosh, Y. H. Han, and R. C. Bradt, The Frictional Component of the Indentation Size
Effect in Low Hardness Testing, J. Mater. Res., 8 (5) 1028 1032 (1993).
8. J. B. Quinn and G. D. Quinn, Indentation Brittleness of Ceramics: A Fresh Approach, J. Mater. Sci.,
32 4331 4346 (1997).
9. A. L. Yurkov and R. C. Bradt, Load Dependence of Hardness of SIALON Based Ceramics, in
Fracture Mechanics of Ceramics, Vol 11, Eds. R. Bradt et al., (Plenum Press, NY, 1996), pp. 369 378.
10. P. M. Sargent, Use of the Indention Size Effect on Microhardness for Materials Characterization, in
Microindentation Techniques in Materials Science and Engineering, ASTM STP 889, eds. P. J. Blau
and B. R. Lawn, (ASTM, West Conshohocken, PA, 1986), pp. 160 174.
11. K. Hirao and M. Tomozawa, Microhardness of SiO2 Glass in Various Environments, J. Amer.
Ceram. Soc., 70 (7) 497 502 (1987).
12. G. H. Frischat, Load Independent Microhardness of Glasses, in Strength of Inorganic Glasses, ed. C.
R. Kurkjian, (Plenum Press, NY, 1986), pp. 135 145.
13. N. Thibault and H. Nyquist, The Measured Knoop Hardness of Hard Substances and Factors
Affecting Its Determination, Trans. A.S.M., 38 271 330 (1947).
14. B. R. Lawn and D. B. Marshall, Hardness, Toughness, and Brittleness, An Indentation Analysis, J.
Am. Ceram. Soc., 62 (7 8) 347 350 (1979).
15. D. B. Marshall and B. R. Lawn, Indentation of Brittle Materials, in Microindentation Techniques in
Materials Science and Engineering, ASTM STP 889, eds. P. J. Blau and B. R. Lawn, (ASTM, West
Conshohocken, PA, 1986), pp. 26 46.
16. B. R. Lawn, A. G. Evans, and D. B. Marshall, Elastic/Plastic Indentation Damage in Ceramics:
Median/Radial Crack System, J. Am. Ceram. Soc., 63 (9 10) 574 581 (1980).
17. J. T. Hagan, Deformation and Cracking Modes Around Plastic Indentations in Glasses, Verres
Refract., 35 (2) 306 314 (1981).
18. J. T. Hagan and M. V. Swain, The Origin of Median and Lateral Cracks Around Plastic Indents in
Brittle Materials, J. Phys. D: Appl. Phys., 11 2091 2102 (1978).
19. F. Knoop, C. Peters, and W. Emerson, A Sensitive Pyramidal Diamond Tool for Indentation
Measurements, J. Res. Nat. Bur. Stds., 23 39 61, July 1939.
20. C. G. Peters and F. Knoop, Resistance of Glass to Indentation, Glass Ind., 20, 174 176, May 1939.
21. ASTM C 730 98, Standard Test Method for Knoop Indentation Hardness of Glass, ASTM Annual
Book of Standards, Vol 15.02, ASTM, West Conshohoken, PA, 1999.
22. DIN 52333, Knoop Hardness Testing, Glass and Glass Ceramic, German Institute for Standards,
Berlin, 1987.
23. ISO 9385, Glass and Glass Ceramics, Knoop Hardness, International Organization for Standards,
Geneva, 1990.
24. G. D. Quinn and J. A. Salem, Effect of Lateral Cracks Upon Fracture Toughness Determined by the
Surface Crack in Flexure Method, J. Am. Ceram. Soc., 85 (4) 873 80 (2002).
25. S. M. Wiederhorn and D. E. Roberts, Fracture Mechanics Study of Skylab Windows, NBS Report 10
892, National Bureau of Standards, Washington, 1972.
26. D. C. Boyd and D. A. Thompson, Glass, in Encyclopedia of Chemical Technology, Vol. 11, 3rd ed.,
(Wiley, NY, 1980), pp. 807 880.
27. R. Gettings, G. Quinn, W. Ruff, and L. Ives, New Hardness Standard Reference Materials (SRM’s)
for Advanced Ceramics, Ceram. Eng. and Sci. Proc., 15 (5) 717 826 (1994).
28. G. D. Quinn, P. Green, and K. Xu, Cracking and the Indentation Size Effect for Knoop Hardness of
Glasses, accepted by J. Amer. Ceram. Soc., Nov 2002.
29. ASTM Standard E 384 89 (1997), Standard Test Method for Microindentation Hardness of Mater
ials, ASTM Annual Book of Standards, Vol. 3.01, ASTM, West Conshohoken, PA, 2001.
30. ASTM Standard C 1326 89 (1996), Standard Test Method for Knoop Indentation Hardness of
Advanced Ceramics, ASTM Annual Book of Standards, Vol. 15.01, ASTM, West Conshohoken, PA,
2001.
31. B. W. Mott, Micro Indentation Hardness Testing, (Butterworth, London, 1955).
32. L. Tarasov and N. Thibault, ‘‘Determination of Knoop Hardness Numbers Independent of Load,’’
Trans. A.S.M., 38 331 353 (1947).
33. J. Gahm, Neuere Erkenntnisse zur Mikro Härte, VDI Berichte, Nr. 160, 25 41 (1972).
34. R. F. Cook and G. M. Pharr, Direct Observation and Analysis of Indentation Cracking in Glasses
and Ceramics, J. Am. Ceram. Soc., 73 (4) 787 817 (1990).
35. T. Wilantewicz and J. R. Varner, The Effect of Contact Parameters on the Crack Initiation Behavior
of Glasses, in Fractography of Glasses and Ceramics, IV, Eds. J. R. Varner and G. D. Quinn, Ceramic
Transactions Vol. 122, (American Ceramic Society, Westerville, OH, 2001), pp. 403 416.
36. H. Bückle, Mikrohärteprüfung, (Berliner Union Verlag, Stuttgart, 1965).
37. R. Mitsche, Über die Eindringhärte Metallischer Fest und Lockerkörper, Osterr. Chem. Z., 49 186
(1948).
38. E. O. Bernhardt, On Microhardness of Solids at the Limit of Kick’s Similarity Law, Z. Metallkunde,
33 135 144 (1941).
39. F. Fröhlich, P. Grau, and W. Grellmann, Performance and Analysis of Recording Microhardness
Tests, Phys. Stat. Sol. (a), 42 79 89 (1977).
40. A. Arora, D. B. Marshall, B. R. Lawn, Indentation Deformation/Fracture of Normal and Anomal
ous Glasses, J. Noncryst. Sol., 31 415 428 (1979).
41. K. Zeng and D. Rowcliffe, Analysis of Penetration Curves Produced by Sharp Indentation on
Ceramic Materials, Phil. Mag. A, 74 (5) 1107 1116 (1996).
42. K. C. Yoo, R. G. Rosemeier, W. L. Elban, and R. W. Armstrong, X ray Topography Evidence for
Energy Dissipation at Indentation Cracks in MgO crystals, J. Mat. Sci. Lttrs., 560 562 (1984).
43. ISO 14705 (2000) Fine Ceramics (Advanced Ceramics, Advanced Technical Ceramics) Test Method
for Hardness of Monolithic Ceramics at Room Temperature, International Organization for Stand
ards, Geneva, 2000.
44. Optical Glass, Schott Glass Technologies, Duryea, PA, Catalog 3111/1e USA III, 1984.
45. S. Van der Zwagg and J. T. Hagan, Deformation Processes in Silica and Different Soda Lime Glasses
under Conical Indentations, in Strength of Inorganic Glass, ed. C. Kurkjian, (Plenum Press, NY,
1986) pp. 147 151.
46. J. T. Hagan, Shear Deformation Under Pyramidal Indentations in Soda Lime Glass, J. Mat. Sci., 15
1417 1424 (1980).
ON THE HARDNESS OF
ZIRCONIA POLYCRYSTALS
Takashi Akatsu, Shingo Numata, Michiyuki Yoshida,

Yutaka Shinoda, and Fumihiro Wakai*
1. INTRODUCTION
The hardness of materials is one of the most important mechanical properties from
an engineering point of view. The hardening of metals is realized by the inhibition of
dislocation glide, which can be managed through microstructural design. The Hall-
Petch hardening H / d 1=2 , where H is the hardness and d is the grain size, is a well-
known and useful concept of the design. Conversely, the inverse Hall-Petch relation
is recently found out in a metal consisted of very fine grains.1,2 Also, it is revealed
that the H-value of metals does not depend on only d but indentation size, i.e. the
hardness increases with a decrease in indentation size. The indentation size depend-
ence is described through the building-up of geometrically necessary dislocations
under a concentrated stress field.3,4 In terms of ceramics, correlation between hard-
ness and microstructure has not been cleared, although a lot of research has been
done.5,6 For example, even the grain size dependence of the hardness has not been
generally described. The hardness of ceramic nanopolycrystals, which consists of
very fine grains with a diameter less than 1mm, has not been examined in detail.
Moreover, indentation size effect on the hardness has been scarcely reported. The
experimental and theoretical difficulty to estimate the plasticity of a hard elastic-
plastic material is one of the reasons for the obscurity in the hardness of ceramics.
The development and improvement of the nanoindentation technique helps us to
make an accurate evaluation of the plasticity of ceramics.
*
Takashi Akatsu, Shingo Numata, Michiyuki Yoshida, Yutaka Shinoda, and Fumihiro Wakai, Mater
ials and Structures Laboratory, Tokyo Institute of Technology, 4259 Nagatsuta, Midori, Yokohama
226 8503, Japan
13
In this study, the nanoindentation on monoclinic-, tetragonal- and cubic-

zirconia polycrystals (referred as MZP, TZP and CZP, respectively) was carried
out with a home-made nanoindenter. Relationship between indentation load P and
penetration depth h of a diamond stylus was analyzed to derive the true hardness HT
as a function of the plastic depth hp , which is a measure of plasticity.7 The plastic
deformation of each zirconia at room temperature was discussed in detail through
the examination and the mutual comparison of indentation size dependence of HT .
2. EXPERIMENTAL
2.1. Samples
The zirconia polycrystals with different crystallographic structure (monoclinic, tetrag-

onal and cubic) were fabricated to be a fine structure. Three kinds of very fine zirconia
powders, TZ-0Y, TZ-3Y and TZ-8Y (Tosoh Co.ltd. Japan), were sintered with HIP
under 200MPa to make MZP, TZP and CZP, respectively.8 The average grain size of as-
sintered MZP, TZP and CZP, which was determined through SEM observation and the
intercept method, was 60, 60 and 110nm, respectively. In addition, heat treatment of
each zirconia was carried out to enlarge their grain size. Nanoindentation on the {100}
plane of a single crystal of cubic zirconia (YSZ(100), Shinkosha Co.ltd., Japan) was
also carried out in comparison. The mirror-finished surface of all samples was carefully
prepared by polishing with colloidal silica powders with a diameter less than 60 nm.
2.2. Nanoindentation
Precise h measurement is very important for the accurate evaluation of hardness

through the nanoindentation technique. External deformation due to the compliant
structure of a conventional nanoindenter must be eliminated from h measured, be-
cause h is usually given as the drive of a diamond stylus towards the surface of a
sample.9 Some calibrations to compensate the extrinsic deformation are convention-
ally carried out, but they are, in many cases, poor at reproducibility. That is one of the
reasons for the inaccurate hardness evaluation with the nanoindentation. In this study,
the accurate value of h was directly measured by using a home-made nanoindenter
with a unique system as shown in Fig.1. The difference between the drive and deflec-
tion of the beam fixed with the stylus (see Fig.1) definitely gives the accurate h without
doing any calibrations. The P-value was determined as a function of the deflection and
flexibility of the beam. In addition, vertical indentation within an error of 0:1 was
confirmed using a laser reflection displacement transducer (see Fig.1) and stages for
the tilt and horizontal rotation of a sample. The system shown in Fig.1 was made of
super invar alloy with a thermal expansion of 0:7 10 6 = K at room temperature to
decrease the error due to the temperature change during indentation. Nanoindenta-
tion at room temperature in air was actually carried out by using the nanoindenter
with the Berkovich-type diamond stylus at a penetration rate of 50nm/sec.
2.3. Analysis of P-h Curves
It is relatively hard to find ideal quadratic relation between P and h in the very
shallow indentation region. That is attributed to the blunt tip of a diamond stylus,
Capacitive DisplacementTransducer
Laser Reflection Displacement Transducer
hcap Dhlaser
h*cap
Laser
Sample Sample
Diamond Stylus Beam
Figure 1. A unique system of a home made nanoindenter to measure accurate P and h. The difference
between the drive Dhlaser and deflection Dhcap ( hcap hcap ) of the beam gives an accurate h. The P value is
given as a function of Dhcap and the flexibility of the beam.
which causes the overestimation of hardness in that region. We made a two-step

modification of P-h curves to avoid the overestimation at small h as follows: At first,
loading and unloading P-h curves were transferred by the distance Dh between the
real and ideal origins, which was determined by the extrapolation of linear relation
between P1=2 and h at relatively large h into h ¼ 0, in the positive direction of h (see
Fig. 2). Secondly, P-h data from h ¼ 0 to h > 2Dh for loading and from hr to
hr þ 2Dh for unloading were neglected, respectively, in the following analysis (see
Fig. 2). According to the Oliver and Pharr method,10 the Meyer hardness HM was
derived from the modified P-h curves as follows,
Pmax
HM ¼ (1)
ghc 2
1
h= P + hr
k2
1000
Penetration depth h /nm
2Δh
500 1
h= P
k1
>2Δh
0
Dh
Real origin
0 5 10 15
Ideal origin
1/2 1/2
Load P /N
Figure 2. Modification of P h curves to avoid the overestimation of hardness in the shallow indentation
region due to the blunt tip of a diamond stylus.
Pmax
hc ¼ hmax « (2)
S
where Pmax the maximum indentation load, hmax the maximum penetration depth of
the stylus, g the geometrical factor to be 24.5 for the Berkovich-type stylus, « the
constant to be 0.72 for conical stylus, and S the unloading slope at Pmax . An analysis
to derive HT and h 9p was also carried out for the modified p1=2 -h relationship as
follows,

k1 = g=gp 2
HT ¼ p 2 (3)
1 k1 =k2
s
Pg p 2
hp ¼ (4)
gHT
where k1 and k2 are the linear slopes of P1=2 -h relationship on loading and unloading,
respectively, and gp is the constant to be unity for the perfect plasticity.
2.4. Raman Spectroscopy

Highly concentrated stress field seems to be realized under an indentation. In the case
of TZP, stress-induced transformation from tetragonal into monoclinic phase can be
predicted to occur under the indentation. Raman spectroscopy by using a focused
ultraviolet laser with a diameter of 1mm was carried out at the bottom of an
indentation to confirm the phase transformation.
3.1. Meyer Hardness of Zirconia Polycrystals

The hardness of materials is conventionally evaluated as HM . The HM -value is,
however, not a measure of the plasticity of ceramics,7 because it is definitely affected
by the elasticity, e.g. Young’s Modulus. Therefore, not only grain size dependence
but also indentation size dependence of the HM -value of MZP, TZP and CZP was
obscure as reported in previous studies.5,6 In this study, the evaluation of HT , a
measure of plasticity, for each zirconia was carried out, and then revealed the
following indentation size dependence.
3.2. True Hardness of CZP
The HT -value of CZP increased with a decrease in hp as shown in Fig.3. Also,

the grain size dependence of HT was hardly observed in CZP. Moreover, the HT -
value of a cubic zirconia single crystal as a function of hp was superimposed just
on that of CZP (see Fig.3). According to the theory of the strain gradient plasticity,3,4
in which the building-up of geometrically necessary dislocations under an indenta-
tion was modeled, the indentation size dependence of hardness is represented as
follows,
80
m=1
H0=24.7GPa
A=446nm

60
40
d=110nm
20 d=290nm
d=4540nm
Single crystal{100}
0 500 1000 1500

Figure 3. The HT value of CZP as a function of hp . d is the average grain size of CZP. The broken line is
drawn by using Eq.(5) with m 1.
m
HT A
¼1þ (5)
H0 hp
where H0 is the hardness evaluated at hp ! 1, A is the constant associated with the

material characteristic length such as the Burgers vector of the dislocation, and m is
the constant to be 1 or 2. The hp dependence of HT shown in Fig.3 was well described
by using Eq.(5) with m ¼ 1. Good agreement between the experimental and theoret-
ical hp dependence of HT regardless of grain size indicates that the plastic deform-
ation of CZP at room temperature is simply due to the accumulation of dislocation
glide in each grain. The dislocations seem to be piled-up at grain boundary but are
not able to go through the boundary into the neighboring grains such as those in
metals because of the lack of the number of active slip system. In other words, the
contribution of grain boundary sliding to the plasticity of CZP seems to be negligibly
small.
3.3. True Hardness of MZP

The HT -value of MZP, which was much smaller than that of CZP, scarcely depended
on hp as shown in Fig.4. In addition, the HT -value of MZP remarkably increased
with a decrease in the grain size below 100 nm (see Fig.4). The softness of MZP at
room temperature is attributed to the plastic deformation caused by the twinning of
the monoclinic crystallographic structure.11 Twinning is quite different from disloca-
tion glide observed in CZP, which is obviously associated with the strain gradient
plasticity. That is the reason why hp dependence of the HT -value of MZP was hardly
observed. In terms of metals, the resistance to plastic deformation due to twinning
increases with a decrease in grain size such as the Hall-Petch hardening.12 Correlation
between the twinning and grain size has not been cleared in ceramics. However, the
resistance to the twinning of monoclinic zirconia may increase with a decrease in
grain size comparably with the size of the twin, which is of the order of 10 nm.11
50
d=60nm
d=140nm
40 d=410nm

30
20
10
0 500 1000 1500 2000

Figure 4. The HT value of MZP as a function of hp . d is the average grain size of MZP.
3.4. True Hardness of TZP
The HT -value and its hp dependence of TZP were classified into two categories as a
function of grain size. If the grains were larger than 100nm, the HT -value of TZP was
the same as that of MZP and scarcely depended on hp (see Fig.5). On the contrary,
the HT -value of TZP increased as large as that of CZP and significantly increased
with a decrease in hp if the grain size decreased below 100nm (see Fig.5). As a result of
Raman spectroscopy at the bottom of an indentation, the monoclinic phase was
observed in the former though only tetragonal phase was detected in the latter. This
indicates that the stress-induced transformation from tetragonal to monoclinic phase
occurred on loading in the former while it was inhibited in the latter. Therefore, the
HT -value of TZP consisted of grains with a diameter larger than the critical size of
the transformation coincided with that of MZP because of the plastic deformation of
the transformed monoclinic phase as mentioned above. The decrease of grain size
below the critical realizes the stabilization of tetragonal phase.13 That is the reason
why the HT -value and its hp dependence of TZP consisted of grains with a diameter
less than 100 nm was the same as those of CZP.
4. CONCLUSION
The HT -value of zirconia polycrystals increased with a decrease in hp . This is

attributed to the building-up of geometrically necessary dislocations under an inden-
tation. In addition, the grain size dependence of HT was hardly observed. These
results indicate that the plastic deformation of zirconia polycrystals at room tem-
perature is basically due to the accumulation of dislocation glide in each grain. In
other words, the contribution of grain boundary sliding to the plasticity was negli-
gibly small even if the grains were of the order of 100nm. The HT -value of MZP,
which was much smaller than that of CZP, scarcely depended on hp because of the
plastic deformation due to twinning. The HT -value and its hp dependence of TZP
80
d=60nm
d=100nm
d=300nm

60 CZP(d=110nm)
MZP(d=140nm)
40
20
0 500 1000 1500 2000

Figure 5. The HT value of TZP as a function of hp . d is the average grain size of zirconia polycrystals.
T
T T
T
T
A
Intensity /a.u.
d=60nm B
B
A
M
M
A
d=100nm
B
100 200 300 400 500 600 700
−1
Raman shift /cm
Figure 6. The spectrum of Raman spectroscopy measured at the bottom A of an indentation on TZP. B is
the place far from an indentation to obtain the reference spectrum. M and T are assigned to monoclinic and
tetragonal phases, respectively. d is the average grain size of TZP.
changed depending on whether the stress-induced phase transformation occurred or

not under an indentation, which was determined as a function of grain size.
ACKNOWLEDGEMENT
A part of this research work was supported by the Nippon Sheet Glass Foundation
for Materials Science and Engineering and the Grant-in-Aids for Scientific Research
(#13750622) of the Ministry of Education, Culture, Sports, Science and Technology,
Japan. The authors are grateful to Mr.Y.Teranishi for helping us to carry out Raman
spectroscopy.
REFERENCES
1. S.Takeuchi, The Mechanism of the Inverse Hall Petch Relation of Nanocrystals, Scripta Mater. 44,
1483 1487 (2001).
2. C.A.Schuh, T.G.Nieh, T.Yamasaki, Hall Petch Breakdown Manifested in abrasive wear resistance of
Nanocrystalline nickel, Scripta Mater. 46, 735 740 (2002).
3. W.D.Nix and H.Gao, Indentation Size Effects in Crystalline Materials: A Law for Strain Gradient
Plasticity, J.Mech.Phys.Solids 46[3], 411 425 (1998).
4. M.R.Begley and J.W.Hutchinson, The Mechanics of Size Dependent Indentation, J.Mech.Phys.
Solids 46[10], 2049 2068 (1998).
5. R.W.Rice, C.C.Wu, and F.Borchelt, Hardness Grain Size Relations in Ceramics, J.Am.Ceram.Soc.
77[10], 2539 2553 (1994).
6. A.Krell, A New Look at the Influences of Load, Grain Size, and Grain Boundaries on the Room
Temperature Hardness of Ceramics, Int’l.J.Ref.Met.Hard Mater. 16, 331 335 (1998).
7. M.Sakai, The Meyer Hardness: A Measure for Plasticity?, J.Mater.Res. 14[9], 3630 3639 (1999).
8. M.Yoshida, Y.Shinoda, T.Akatsu, and F.Wakai, Deformation of Monoclinic ZrO2 Polycrystals and
Y2 O3 Stabilized Tetragonal ZrO2 Polycrystals Below the Monoclinic Tetragonal Transition Tem
perature, J.Am.Ceram.Soc. 85[11], 2834 2836 (2002).
9. M.Sakai, S.Shimizu, and T.Ishikawa, The Indentation Load Depth Curve of Ceramics, J.Mater.Res.
14[4], 1471 1484 (1999).
10. W.C.Oliver and G.M.Pharr, An Improved Technique for Determining Hardness and Elastic Modu
lus Using Load and Displacement Sensing Indentation Experiments, J.Mater.Res. 7[6], 1564 1583
(1992).
11. I W.Chen, Implication of Transformation Plasticity in ZrO2 Containing Ceramics: II, Elastic Plastic
Indentation, J.Am.Ceram.Soc. 69[3], 189 194 (1986).
12. D.Hull, Effect of Grain Size and Temperature on Slip, Twinning and Fracture in 3% Silicon Iron,
Acta Metall. 9[3], 191 204 (1961).
13. F.F.Lange, Transformation Toughening Part 1 Size Effects Associated with the Thermodynamics of
Constrained Transformations, J.Mater.Sci. 17, 225 234 (1982).
INSTRUMENTED HARDNESS
TEST ON ALUMINA
CERAMICS AND SINGLE
CRYSTAL WITH SPHERICAL
INDENTER
Shuji Sakaguchi, Kiyoshi Hirao, Yukihiko Yamauchi

and Shuzo Kanzaki*
1. INTRODUCTION
Ceramic materials are expected to be applied to the structural parts, especially to

tribological application, as they have high hardness and high elastic constants. For
this application, it is important to know the properties of mechanical behavior on the
surface of ceramics. Instrumented indentation is one of the tests that can explain the
mechanical properties on the surface, as it expresses the relation between the pene-
tration depth of the indenter and the applied force.1,2 For metallic materials, the
standard test method in instrumented indentation is already published.3 We have
tried the instrumented indentation tests on ceramic materials. Figure 1 is a typical
example of the loading-unloading curve by this test on alumina ceramics.
If the test force is small enough, the indenter contacts the surface elastically, and
no indent shall be made and no energy is consumed for making indent. It means if we
increase the test force, the consumed energy ratio for making indent can increase. With
this assumption, we tried the measurement of instrumented hardness with conical
* Shuji Sakaguchi, Ceramic Research Institute, National Institute of Advanced Industrial Science and
Technology, Nagoya, Japan 463 8560. Kiyoshi Hirao, Yukihiko Yamauchi and Shuzo Kanzaki, Synergy
Materials Research Center, National Institute of Advanced Industrial Science and Technology, Nagoya,
Japan 463 8687.
69
1.2 loading
unloading
1
1.1mm
0.8
Load (N) 0.6
0.4
0.2
0
0 0.5 1 1.5 2 2.5
Displacement (μm)
Figure 1. Typical result of instrumented indentation on alumina ceramics, measured with Vickers
indenter.
indenters (Berkovic, Knoop and Vickers), and we obtained the relation between the
applied test force and the energy consumption ratio for making indents. Figure 2
shows some results of the test force dependence of the consumed energy ratio. In some
results obtained with Knoop indenter, the results follow this expectation, but the test
force dependence of the energy consumption ratio was not clearly observed. The
scattering of the data is relatively large, as the applied test force was small. We could
not obtain the grain size dependence of the energy consumption clearly. It was
considered that the tip of the indenter was too keen to obtain the elastic contact on
these materials. Spherical indenters shall be used for this discussion.
In this study, we tried to use spherical diamond indenters with different tip
radius on the instrumented indentation tests. We tried to find the test force depend-
ence of the surface fracture with these indenters, and the grain size dependence of the
instrumented indentation results on ceramic materials.
2. EXPERIMENTS
2.1. Testing conditions

Recording hardness test machine supplied by CSIRO, Australia (Type: UMIS-2000)
was used for this study. Seven kinds of spherical indenters with different tip radius
(1, 2, 5, 10, 20, 50 and 100 mm) were used. Maximum test force was chosen as 20, 50,
100, 200 and 500 mN. Five tests were carried out in the same test condition, and
some scattered data that might be caused by some large fracture on the surface were
neglected. The data were averaged and the consumed energy was calculated from the
loading-unloading curve.
2.2. Specimens
The specimen was alumina sintered body with purity of 99.9%.4,5 The average grain
size of as-received sample was 1:1 mm. It was annealed in air in three conditions for
0.5 0.5
0.4
0.4
0.3
0.3 0.2
Ratio
Ratio
0.2 0.1
0
0.1
Berkovic −0.1 Knoop
1.1μm 1.1μm
0 −0.2
0.01 0.1 1 0.01 0.1 1
Force (N) Force (N)
0.5 0.5
0.4 0.4
0.3 0.3
Ratio
Ratio
0.2 0.2
0.1 0.1
Berkovic Knoop
5.7μm 5.7μm
0 0
0.01 0.1 1 0.01 0.1 1
Force (N) Force (N)
0.5 0.5
0.4 0.4
0.3 0.3
Ratio
Ratio
0.2 0.2
0.1 0.1
Berkovic Knoop
Sapphire Sapphire
0 0
0.01 0.1 1 0.01 0.1 1
Force (N) Force (N)
Figure 2. Examples of the relation between the test force and the energy consumption ratio on instru
mented indentation with Berkovic or Knoop indenter.
the grain growth. The specimens with average grain sizes of 1.8, 4.4 and 5:7 mm were
prepared. The mechanical properties of these specimens are shown in Table 1. The
Vickers hardness was not changed with the grain growth. For the comparison, the
c-plane of the alumina single crystal (sapphire) was also tested. The Vickers hardness
of the sapphire was 19.0 GPa with the standard deviation of 0.5 GPa. It was a little
harder than the sintered alumina.
Table 1. Four point bending strength, fracture toughness measured by single edge pre
cracked beam (SEPB) method, and Vickers hardness of the tested alumina specimens.
Standard deviations are in brackets.
Grain size (mm) Strength (MPa) KIC (MPa m1=2 ) HV (GPa)
1.1 652 (39) 3.00 (0.31) 18.3 (0.4)

1.8 555 (40) 3.20 (0.14) 18.7 (0.6)
4.4 415 (23) 3.95 (0.24) 17.3 (0.5)
5.7 369 (21) 4.21 (0.05) 17.6 (0.6)
3. RESULTS
Figure 3 shows the results of the energy consumption ratio of every specimen on the
indenter of every tip radius. On the indenters of 20 mm and 50 mm, the contacts were
almost elastically and no energy consumption was found. However, in the test with
20 mm indenter the scattering of the data became larger. On the indenters of 5
and 10 mm, energy consumption ratio increased continuously with increasing the
60 50
1.1
50 40 1.8
4.4
5.7
40 30 Sap
30 20
1.1
1.8
20 4.4 10
5.7
Sap
10 0
0 −10
10 100 1000 10 100 1000
40 40
1.1 1.1
30 1.8 30 1.8
4.4 4.4
5.7 5.7
20 Sap 20 Sap
10 10
0 0
−10 −10
−20 −20
10 100 1000 10 100 1000
Figure 3. Energy consumption ratio of each test condition with spherical indenters of 5, 10, 20 and 50 mm
in radius (‘Sap’ is sapphire c plane).
maximum test force. The grain size dependence of the energy consumption was not
observed.
If the energy consumption increased abruptly at a certain test force, we can
estimate a limit of the surface damage. However, the energy consumption increased
gradually with increasing the test force. That means the surface damage is not made
at a certain stress but accumulated with increasing the stress.
We could not see the grain size dependence of the energy consumption. The
surface damage with compressive stress can be independent of the grain size, as it can
be expected from the results of the Vickers hardness.
4. DISCUSSION
For discussing all the data, the results were plotted as the relation between the
average contact stress of the indenter calculated with the equations for the elastic
contact, and the energy consumption ratio. Table 2 shows the average stress in the
contact circle from the equation of the elastic contact on every test condition. The
maximum stress, which appears at the center of the contact circle, is 1.5 times larger
than the average stress. Figure 4 shows the plot of all the data. The plots can be
divided into three regions. (1) Calculated stress is below 12 GPa. No energy con-
sumption was found. (2) Calculated stress is between 12 GPa and 18 GPa. The
energy consumption was still around zero, but the scattering of the data becomes
larger. (3) Calculated stress is over 18 GPa. The energy consumption increased
simply with increasing the applied stress.
These three stress regions correspond to the following conditions; (1)
The compressive stress is below the hardness value on all the contact area. (2) The
average compressive stress is smaller than the hardness value, but the maximum
stress is larger than the hardness. The contact area may partly damaged. (3) The
average stress is larger than the hardness. The indent size shall be larger than the
calculated elastic contact size.
With these explanations, we can assume the reason why the scattering of the
energy consumption became larger in the intermediate stress region. It is schematic-
ally shown in Figure 5. In this intermediate stress condition, the average stress is
smaller than the hardness. Then the indent with the size of the contact circle cannot
be made. However, the stress at the center of the contact circle is larger than the
hardness. The surface area at the center shall be damaged, or micro-fractured, by this
contact. It is not an elastic contact, but it cannot make a clear indent. That can be a
reason why the energy consumption was not clearly increased but only the scattering
of the data was increased.
Table 2. Calculated compressive stress (GPa) on elastic contact in every condition.
Test Force (mN)

Indenter Radius (mm) 20 50 100 200 500
5 16.1 21.9 27.6 34.8 47.2

10 10.2 13.8 17.4 21.9 29.7
20 (6.4) (8.7) 11.0 13.8 18.7
50 (3.5) (4.7) 5.9 7.5 10.2
60

50
40
30
20
10
1.1
0 1.8
4.4
−10 5.7
Sap
−20
0 10 20 30 40 50
Average stress (GPa)
Figure 4. All the data are plotted with the relation between the calculated contact stress and the
energy consumption. The results can be divided into three regions; (1) 0 12 GPa, (2) 12 18 GPa, (3) over
18 GPa.
Figure 5. Indentation damage under intermediate test force. The area around the center of the
contact circle is damaged. It is not an elastic contact, and it can make a scattering of the energy
consumption.
5. CONCLUSION
Instrumented indentation tests were carried out on alumina ceramics with different
average grain size, using spherical indenters.
(1) We cold not found the average grain size dependence of the energy consumption
for making indents on alumina specimens.
(2) The stress value for making surface damage is related to the Vickers hardness
value. It can be divided into three levels; 1. contacting elastically, 2. contact area
is partly damaged, 3. contact area is fully damaged and indent is made.
(3) The consumption energy ratio for making indents increased continuously with
increasing the maximum test force, if the applied stress is more than the Vickers
hardness value.
6. ACKNOWEDGMENTS
This work has been supported by NEDO, as part of the Synergy Ceramics Project
promoted by METI, Japan. The authors are members of the Joint Research Con-
sortium of Synergy Ceramics.
REFERENCES
1. M. Sakai and S. Shimizu, Elastic, plastic, and viscoelastic contact problems in axisymmetric indenta
tion, Ceramic Transactions (The American Ceramic Society), vol. 133, 105 114, (2002).
2. W. C. Oliver and G. M. Pharr, An improved technique for determining hardness and elastic modulus
using load and displacement sensing indentation experiments, J. Mater. Res., 7, 1564 1583, (1992).
3. ISO 14577 1, Metallic materials Instrumented indentation test for hardness and materials
parameters Part. 1: Test method, (2002).
4. S. Sakaguchi, The relation between the damage tolerance and the R curve behavior of alumina
ceramics, Fracture Mechanics of Ceramics (Kluwer Academic / Plenum Press), vol. 13, 229 234,
(2002).
5. S. Sakaguchi, R curve measurement of alumina using straight notched or indented bending bars,
Fracture Mechanics of Ceramics (Plenum Press), vol. 11, 39 51, (1996).
INVESTIGATION OF CRACK
GROWTH PROCESS IN DENSE
HYDROXYAPATITE USING
THE DOUBLE TORSION
METHOD
C. Benaqqa1,2, J. Chevalier1, M. Saâdaoui2,

and G. Fantozzi1
ABSTRACT
In this work, double torsion tests were performed to investigate slow crack growth
behavior of dense hydroxyapatite materials. Crack rate, V, versus stress intensity
factor, KI , laws were obtained for different environments and processing conditions.
Stress assisted corrosion by water molecules in oxide ceramics is generally responsible
for slow crack growth. The different propagation stages obtained here could be
analyzed in relation to this process. The presence of a threshold defining a safety
range of use was also observed. Hydroxyapatite ceramics appear to be very sensitive
to slow crack growth, crack propagation occurring even at very low KI . This can be
explained by the fact that their contain hydroxyl groups (HAP: Ca10 (PO4 )6 (OH)2 ),
favoring water adsorption on the crack surface and thus a strong decrease of surface
energy in the presence of water. This study demonstrates that processing conditions
must be carefully controlled, specially sintering temperature, which plays a key role
on V -KI laws. Sintering at 508C above or below the optimal temperature, for
example, may shift the V -KI law towards very low stress intensity factors. The
influence of ageing is finally discussed.
1
G.E.M.P.P.M. (U.M.R. C.N.R.S. 55 10) I.N.S.A. de Lyon, 69621 Villeurbanne Cedex France
2
L.E.R.S.I.M. Ecole Mohammadia d’Ingénieurs, Rabat, Morocco
387
388 C. Benaqqa et al.
1. INTRODUCTION
Bio active hydroxyapatite ceramics (HAP), are widely used as coatings on femoral
stems or acetabular cups and as bone substitutes, because they induce growth and
bonding of the bone on their surface [1]. For these ceramics, biological properties are
of prime interest. However, their mechanical properties must not be neglected. De-
pending on the site where bio active ceramics are implanted, non negligible stresses
may appear in the biomaterial. In addition to the sudden failure of the ceramics, which
takes place for a stress intensity factor reaching the toughness, KIC , it is essential to
consider also the so-called ‘‘Subcritical Crack Growth’’ which corresponds to a crack
propagation for stress intensity factors, KI , lower than KIC in particular in corrosive
environments. The propagation of the preexisting natural defects occurs at low rates
(slow crack growth), and causes delayed failure of ceramics when the flaw size reaches a
critical value. The sensitivity of ceramics to the slow crack growth is one of their major
limitations. It is thus very important to understand the mechanisms which are involved
and to determine with precision the crack propagation laws in order to predict the
lifetime of these materials for a given application.
Few interest has been paid to this behavior in hydroxyapatite ceramics, despite
its importance for long term success. Some investigations have shown that hydro-
xyapatite ceramics were sensitive to slow crack growth and fatigue, mainly by so
called static or dynamic fatigue tests [2,3]. In these studies, Hydroxyapatite was
shown to present reduced lifetime under stress. Static and dynamic fatigue are
indirect methods, since they do not allow the real measurement of crack rates for a
given applied load. They only show the evidence of fatigue effects. In contrast, the
double torsion is used here as a direct method of crack growth characterization
because it can allow the real determination of the crack growth rate, V, versus the
applied stress intensity factor, KI , over a wide range of crack velocities, from 10 12 to
10 2 m=s [4].
The entire V ---KI diagram should be used for lifetime prediction, and the aim of
this study is to characterize the V ---KI law of this material using the double torsion
method. Emphasis is given to the role of processing conditions and of ageing on the
entire V -KI law, which to the knowledge of the authors has not been studied in
previous works.
2.1. Material
A pure atomized hydroxyapatite powder was supplied by Coating Industry (France).

Specimens were processed by uniaxial pressing at 20 MPa followed by cold isostatic
pressing at 200 MPa. The green compacts reached about 55% of the theoretical
density (3:16 g=cm3 ). They were then sintered in air at different temperatures (from
10008C to 14008C) for three hours.
Grain size and morphology were examined by Scanning Electron Microscopy
(SEM) and the mean grain size was evaluated from the linear intercept method
according to the ASTM E112 standard1. Densities were determined by the Archime-
1
American Society Testing and Materials, West Conshohocken, PA
Investigation of Crack Growth Process in Dense Hydroxyapatite 389
des method using water as the immersion medium. The evolution of the density and
the microstructure versus sintering temperature are shown in figure 1. In agreement
with previous studies [2, 3, 5, 6], full density is hardly obtained by natural sintering in
air. However, 12008C allows to reach 98% of theoretical density. The evolution of the
grain size versus sintering temperature initially shows a slight increase, followed by a
strong acceleration above 12508C. Hardness measurements were conducted at low
loads (200 g) for the same range of temperatures. They are shown in figure 2 and
compared to density values. The two curves clearly show the same behavior, with
an increase until 12008C before reaching a plateau. 12008C can be considered as the
optimum sintering temperature, since densification is nearly achieved and grain size
is limited.
From this preliminary results, three model materials were chosen for crack
propagation studies: sintered 3 hours at 12008C (a priori optimum properties),
11508C (lower density) and 12508C (coarsening of the grain size).
Figure 1. Evolution of density and grain size of HAP with sintering temperature.
Figure 2. Evolution of relative density and vickers hardness of HAP with temperature.
2.2. Double Torsion Technique
The double torsion specimens, consisting of plates of dimensions 40 20 3 mm3 ,

and the loading configuration, are shown in figure 3. Specimens were thicker than we
use generally for testing monolithic ceramics (2 mm) [4,p7–9]. Indeed, hydroxyapatite
is very brittle, with a toughness on the order of 1 MPa m [10], thus 3 mm thickness
was necessary to work under reasonable loads. The tensile surface of sintered samples
was polished with a series of diamond pastes down to a final polish of 3 mm in order
to observe the crack with a precision of 2 mm. A notch of length a0 ¼ 8 mm and
root r ¼ 0:1 mm was machined with a diamond saw and an indentation was per-
formed at low load (300 g) in order to initiate a small natural crack. Subsequent
precracking was performed by loading the specimens at low rate in order to induce a
sharp crack of initial length ai ¼ 12 mm.
Generally, it is admitted that the stress intensity factor, KI , is independent of the
crack length, and given by:

Wm 3(1 þ v) 1=2
KI ¼ 2 P (1)
U W c
where P is the load, Wm the span, U and W the width and the thickness of the
specimen, v the Poisson ratio (taken here as equal to 0.3), and c a calibration factor
[11].
To obtain accurate crack growth measurement of the V ---KI curves, it has been
recently argued that a correction factor should be introduced in the conventional
expression of KI [4]. Thus, the following more realistic expression of KI was used:
0:17
Wm 3(1 þ v) 1=2 a
KI ¼ P (2)
U2 W c a0
where a0 and a are respectively the notch and the crack lengths.
Crack velocity versus stress intensity factor (V -KI ) curves were obtained by two
general tests, so called ‘‘constant loading’’ and ‘‘relaxation’’ tests [4]. In the load
relaxation method, the pre-cracked specimen was subjected to fast loading, followed
by subsequent stopping of the crosshead at a certain displacement. Hence, the crack
propagation resulted in a load relaxation and the load versus time curve allows the
determination of the V -KI curve with a single experiment, for crack rates down to
10 7 m=s [4]. The stress intensity factor can be calculated from expression (2) pro-
Figure 3. Double Torsion specimen and loading configuration.

vided the crack length is known. Both crack length and crack rate can be calculated
from the compliance and the compliance rate at each time of the load relaxation
curve. More details concerning this method can be found elsewhere [12–14].
To obtain lower crack rates, an alternative method (so called constant loading
test) was used. The specimen is subjected to a static load under a prescribed duration,
Dt. The crack length is measured via optical microscopy, with a precision of 2 mm,
and V is defined as the ratio of crack increment, Da, to the duration, Dt (V ¼ Da=Dt).
By conducting experiments under different loads, it is therefore possible to obtain
different points of the V -KI diagram. This technique presents the advantage of
allowing very low velocities in order to investigate the presence of a threshold stress
intensity factor value below which crack propagation does not occur [9, 12].
2.3. Influence of Environment and Ageing on Slow Crack Growth
In order to understand the crack propagation mechanisms in HAP, and in particular

the role of water, relaxation tests were conducted on the same specimen at room
temperature in three different media, namely in ambient air, liquid water and silicon
oil (Rhodorsil 710, Rhône Poulenc, France). To obtain reproducible results in liquid
environments, precracked specimens were cleaned by ultrasonic cleaner in acetone
and then in ethyl alcohol for 5 minutes to insure a perfectly clean crack path. They
were subsequently put in a desiccator under 10 2 hPa vacuum for 1 hour and
dropped into the test liquid environment while vacuum was maintained for an
additional hour. They were further transferred to the testing jig without any contact
with ambient atmosphere. This procedure insures perfect wettability of the liquid
along the crack path. In order to increase the significance of the results, three
successive relaxation curves were conducted on the same specimen, first in air, then
in water and oil. This ensures that no variation occurs from one sample to another.
Ageing was performed in a thermostated bath at 378C. Specimens were left
during 6 weeks, then tested in air by means of the relaxation technique. 6 weeks
may correspond roughly to the time necessary for bone healing to occur in-vivo [15].
Although these ageing tests are conducted in-vitro, they could give a first idea of the
stability of HAP properties in an immersed media.
3.1. V---KI Diagram

The V -KI diagram determined in air for a dense hydroxyapatite with a grain size of
2 mm (sintered at 12008C for 3 hours) is given in figure 4 for crack velocities from
10 10 m=s to 10 2 m=s. The results show p that crack propagation takes place well
below the toughness (about 0:9 Mpa m), over a wide range of crack velocity:
Hydroxyapatite is thus very sensitive to crack growth. The overall curve presents
three different stages, which are generally observed in oxide ceramics [12]. Each stage
can be fitted by a single relation, between the crack speed, V, and the stress intensity
factor, KI , [7, 8, 16, 17], on the form:
V ¼ AKIn (3)
Figure 4. V KI diagram of HAP in air obtained by: . Relaxation test and

8 Constant loading test.
where A and n are constants which correspond to each stage. They are given in
table I. They can be used for lifetime prediction. In particular, stage I has a primor-
dial practical importance since low crack rates (V < 10 4 m=s) rule the lifetime of the
pieces in service. This region is attributed to the reaction kinetic between water and
the ceramic bonds [12] : this is the rate of the reaction kinetic which drives the crack
velocity. Stage II occurs in air, when the transport of moisture from the crack mouth
to the crack tip noticeably reduces the crack velocity: in this stage, the crack rate is
controlled by the diffusion of the water molecules to the reaction site. The crack
velocity is a weak function of the stress intensity factor (low n value). A third stage
occurs for even higher crack rates (region III), when the crack rate exceeds the
molecules transport rate. In this case, the fracture of ceramic bonds occurs without
reaction with environmental species. Thus, this stage corresponds to the vacuum
conditions. Region II of crack propagation is often considered as a connecting
branch between fracture of ceramic bonds by water (region I) and fracture under
vacuum conditions (region III).
It must be noted that stage I was not completely described by relaxation tests
and that no threshold could be obtained. This was done by constant loading tests.
The threshold, KI0 , is the most important parameter in practice because it defines a
perfectly sure field of work. It is seldom defined clearly in the case of polycrystalline
ceramics, due to the experimental difficulty and duration of tests. Our results showp
that crack rates tend to zero for a stress intensity factor around KI0 ¼ 0:6 Mpa m.
The ratio of threshold to toughness KI0 =KIC is low, showing that hydroxyapatite is
very sensitive to subcritical crack growth.
Table 1. Crack growth parameters for hydroxyapatite.
Stage I Stage II Stage III

p 4 4
A(KI in MPa m) 3:3 10 3:6:10 0.8
n 55 7 52
The presence of three stages, noted I, II and III on the figure and of a threshold
stress intensity factor, strongly suggests that stress corrosion by water molecules is
the key mechanism for crack propagation [12, 18]. Its spreading over a large range of
KI shows the detrimental effect of stress corrosion by water on the crack resistance.
Hydroxyapatite ceramics contain hydroxyl groups (chemical formulae: Ca10 (PO4 )6
(OH)2 ), favoring water adsorption on the crack surface and thus a strong decrease of
surface energy in the presence of water. Some surface energy measurements are
available in the literature [3]. They show that the ratio of surface energy in the
presence of water to that under vacuum condition is low, which agrees with our
measurements, since KI0 and KIC refer respectively to the surface energy in air and in
vacuum [12].
Results of successive relaxation tests on the same specimen in different media are
shown in figure 5. Only one stage is present in both liquids. For oil, this unique stage
is superimposed to the third stage of the V ---KI law in the air, which proves that this
stage corresponds well to the conditions of vacuum. In the case of water, the slope of
the V ---KI diagram is very close to that of the primary stage in air. This proves the
effect of water activity and concentration on crack rate and surface energy. From
these results, it
p can indeed be anticipated that the threshold in water should be less
than 0:4 MPa m. This can already question about the use of this material under
significant loading. Using the well known equation:
p
KI ¼ s p a (4)
One obtains that a defect of only 20 mm will propagate slowly under a stress of
50 MPa and lead to delayed failure.
Some results were already available in the literature concerning the stress
exponents of equation (3). The effects of environment on dynamic fatigue resistance
of hydroxyapatite ceramics were investigated by Raynaud et al. [6]. They found that
the crack propagation exponent decreased from 22:5 2 in air to 10 4 in Ringer’s
solution for dense materials. Wakamatsu et al. [19] gave a constant fatigue parameter
n ¼ 19 for sintered hydroxyapatite in 378C distilled water. Values obtained with these
indirect methods (e.g. dynamic fatigue) are only average values, whereas double
Figure 5. V KI diagram of HAP in three media obtained on the same specimen.

torsion test (direct method) allows measurement of the real parameters A and n for
each propagation stage. These averages take into account the three stages of crack
propagation, but the importance of stage I is often minor and difficult to control.
Unfortunately, in industrial applications, where the components have to support
stresses for long duration, stage I is much more important than stages II and III.
Therefore, such indirect methods cannot be used to predict the lifetime of pieces.
3.2. Microstructure–Mechanical Properties Relation

The results of crack propagation in air for the three selected sintering temperatures
(11508C, 12008C and 12508C) obtained by the relaxation method are shown in
figure 6. In agreement with our first preliminary results, 12008C is the optimal
sintering temperature, since it corresponds to the highest crack resistance. Few
degrees above or below this optimal sintering temperature lead to a shift of the
V -KI diagram towards weak stress intensity factor. This is particularly crucial for
the material sintered at 12508C, which exhibits a very poor crack resistance.
The mode of fracture can easily explain these results: figure 7 shows that 12008C
corresponds clearly to a transition from inter- to trans-granular failure. In ceramics,
this transition corresponds generally to a strong decrease of crack resistance. How-
ever, it occurs often for grain sizes coarser than several dozen microns. In the case of
HAP, this transition corresponds to a very low grain size (about 2 mm). The fact that
the material sintered at 11508C exhibited a slightly lower crack resistance than that
sintered at 12008C is explained by the fact that complete densification is not reached.
Table II shows Young’s modulus values for these three materials, measured by a
resonance frequency method. As reported in the literature [3], 12008C corresponds
again to a maximum. Below this temperature, Young’s modulus is lower, due to the
presence of residual porosity. Above, the fall of Young’s modulus can traduce micro-
cracking [20]. Indeed, during cooling, residual stresses due to the anisotropy of grains
can be sufficient to promote crack growth, since both toughness and threshold
are low.
Figure 6. V KI laws of hydroxyapatite in air for three different sintering temperature, obtained by the
relaxation test.
Figure 7. SEM images of HAP sintered at different temperatures ((a) 11508C, (b) 12008C, (c) 12508C):
etched surfaces (1) and fracture surfaces (2).
Table 2. Young’s modulus evolution with different sintering temperature of HAP.

Sintering temperature (8C, 3h) 1150 1200 1250
Young’s modulus (GPa) 104 2 112 2 106:5 2
3.3. Influence of ageing on V KI laws

In order to investigate the influence of ageing on crack resistance of HAP, two
relaxations were performed on the same specimen before and after ageing, as
shown in figure 8. The results show a strong decrease of the toughness and a shift
of the overall V -KI curve towards lower stress intensity factors. Such decrease of the
mechanical properties of hydroxyapatite has been already observed in the literature.
Raynaud et al. [2] argued that ageing occurred by a progressive attack of the grain
boundaries by water and apparition of pores. This is confirmed in our study: ageing
was followed by a decrease of the density from 98% down to 95%. It is interesting to
Figure 8. V KI laws and SEM micrographs of HAP before (1) and after (2) ageing water at 378C during
6 weeks.
note that the second stage is much more limited after ageing than before. It must be
kept in mind that stage II corresponds to a saturation of crack velocity limited by the
diffusion of water molecules. In the case of aged materials, water can be adsorbed at
the surface, giving rise to higher diffusion rate to the crack tip.
4. CONCLUSION
This study shows the strong sensitivity of hydroxyapatite to slow crack growth, due
to stress corrosion by water, resulting in a spectacular fall of crack resistance in the
presence of humidity or after ageing in water. The presence of a threshold
p is observed
in air. However, in water, this threshold is lower than 0:4 MPa m, which is excep-
tionally low. Another important issue raised here is the narrow range of sintering
temperatures giving acceptable mechanical properties. In particular, sintering tem-
peratures above 12008C lead to a dramatic decrease of crack resistance due to trans-
granular failure and micro-cracking.
ACKNOWLEDGMENTS
Coating Industries and Région Rhône Alpe are acknowledged for their help and
financial support (‘programme MIRA’ and ‘Thématiques prioritaires’).
REFERENCES
1. Hardouin P, Chopin D, Devyver B, Flautre B, Blary MC, Guigui P, Anselme K. J Mater Sci Mater
Med 1991; 3:212 218.
2. Raynaud S, Champion E, Bernache Assolant D. Bioceramics 1997; 10:75 78.

3. De With G, Van Dijk J. A, Hattu N, Prijs K. J Mater Sci 1981; 16:1592 1598.
4. Chevalier J, Saädaoui M, Olagnon C, Fantozzi G. Ceramics Inter 1996; 22:171 177.
5. Muralithran G, Ramesh S. Ceramics Int 2000; 26:221 230.
6. Raynaud S, Champion E, Bernache Assolant D, Tetard D. J Mater Scien : Mater. in medicine 1998;
9:221 227.
7. Chevalier J, Olagnon C, Fantozzi G, J Am Ceram Soc 1999; 82, n811:3129 3138.
8. Ebrahimi M. E, Chevalier J, Fantozzi G. J Mater Res 2000; 15, n81:142 147.
9. Chevalier J, Olagnon C, Fantozzi G, Cales B. Ceramics International 1997; 23:263 266.
10. Gautier S, Champion E, Bernache Assolant D. 4th Euroceramics 1995; 8:201 208.
11. Plekta B. G, Fuller E. R, Koepke B. G. ASTM Spec tech publ 1979; 678:19 38.
12. Lawn B. Fracture of Brittle Solids : 2nd Edition, Cambridge University Press 1993:378 p.
13. Ciccotti M, Gonzato G, Mulargia F. Int J of Rock Mechanics & Mining Sciences 2000;
37:1103 1113.
14. Williams D. P, Evans A. G. J. Test Eval 1973;1, n84:264 270.
15. Poitout D. Biomécanique orthopedique. Paris : Masson 1987:203 p.
16. Rhanim H, Olagnon C, Fantozzi G, Torrecillas R, J Europ Cer Soc 1997; 17:85 89.
17. Fett T, Munz D. Fracture mechanical of ceramics 1992; 9:219 233.
18. Wan K, Lathabai S, Lawn B. J Europ Ceram Soc 1990; n86:259 268.
19. Wakamatsu N, Goto T, Adachi M, Imura S et al. J Japanese Soc for Dental Mater And Devices 1990;
9:178 188.
20. Halouani R, Bernache Assolant D, Champion E, Ababou A. J Mater Sci 1994; 5:563 568.
MEASURING THE REAL
CERAMICS: ASTM C 1421
Jonathan Salem,1 George Quinn,2 and Michael Jenkins3
ABSTRACT
ASTM C 1421 ‘‘Standard Test Methods for Determination of Fracture Toughness

of Advanced Ceramics at Ambient Temperature’’ is a high-quality, technically-
rigorous, full-consensus standard that may have finally answered the question,
‘‘What is the ‘real’ fracture toughness of ceramics?’’ This document was eight years
in the actual standardization process (although an estimated two decades of prepar-
ation work may have preceded the actual standardization process). Three different
types of notch/crack geometries are employed in flexure beams: single edge pre-
cracked beam (SEPB); chevron-notched beam (CNB), and surface crack in flexure
(SCF). Extensive experimental, analytical, and numerical evaluations were con-
ducted in order to mitigate interferences that frequently lower the accuracy of
fracture toughness test results. Several round robins (e.g., Versailles Advanced
Materials and Standards {VAMAS}) verified and validated the choice of dimensions
and test parameters included in the standard. In addition, the standard reference
material NIST SRM 2100 was developed and can be used in concert with ASTM C
1421 to validate a fracture toughness test setup or test protocol.
1
Materials Research Engineer, NASA Glenn Research Center at Lewis Field, Life Prediction Branch,
21000 Brookpark Road, MS 49 7, Cleveland, OH 44135. [email protected].
2
Research Scientist, National Institute for Standards and Technology, Ceramics Division, 100 Bureau
Drive, Stop 852, Gaithersburg, MD 20899 852.
3
Professor and Chair, Department of Mechanical Engineering, University of Detroit Mercy, Engineering
Bldg, 4001 W. McNichols Rd, P.O. Box 19900, Detroit, MI 48219 0900.
531
532 Jonathan Salem et al.
1. INTRODUCTION
A wide variety of test configurations and specimen geometries have been used to
determine the fracture toughness of ceramic materials. A variety of definitions exist
for fracture toughness: for example, ‘‘a generic term for measures of resistance to
crack extension’’ [1], or ‘‘a measure of the ease of crack propagation through a
material’’ [2]. In terms of standardization, the definitions typically boil down to a
well defined operational procedure as stated in the discussion following one defin-
ition: ‘‘ . . . . the value of stress-intensity factor designated KIc as measured using the
operational procedure specified in this test method’’ [3]. The need for an operational
rather than theoretical definition is used in standardization in order to avoid
interferences and pitfalls that can occur in many popular and widely used test
specimens.
Interferences must be minimized if accurate, repeatable results are to be ensured.
Standards developed by full consensus bodies, such as the American Society for
Testing and Materials (ASTM) and ISO (International Organization for Standard-
ization), not only put data measurement on a common basis, but attempt to improve
the metrology of the result by controlling such interferences. A standard sets up
rigorous guidelines that will identify and avoid the occurrence of interferences during
data generation. For fracture toughness testing, interferences include the lack of
stable crack extension, stress corrosion, crack growth resistance, stress intensity
factor solution accuracy, crack length measurement, crack shape, residual stresses,
and alignment of the crack with the fixture. In addition, fracture toughness tests
performed by using beams tend to encounter the interferences associated with
flexural strength testing, such as nonparallelism, twist, friction, and bevel effects.
The severity of the interferences varies with the particular test method.
Recently, a full consensus standard for the ‘‘Determination of Fracture Tough-
ness of Advanced Ceramics at Ambient Temperature’’ was developed and published
under the auspices of ASTM [4]. During development of the standard test method,
typical interferences were addressed, and when possible each error was kept to 2%
or less. In addition to development of the standard, several round robin exercises
were run [5 9], and a standard reference material was developed [10]. Additional
work on fracture toughness standardization should include a method(s) for elevated
temperature testing, very thin shapes such as fuel cell electrolyte membranes, and
miniaturized test specimens.
2. THE STANDARD METHODS
The standard allows three tests methods: the chevron notch beam (CNB), the single-
edge-precracked-beam (SEPB), and the surface-crack in flexure (SCF). Three test
methods are allowed because no one method provides valid test results for the variety
of ceramics being developed. All the test methods use 3- or 4-point flexure to test
rectangular cross sections such as the common 3 mm by 4 mm test section used in
flexural strength testing of ceramics [11]. More importantly, all have a well defined
loading configuration, a well defined crack, and an accurate stress intensity factor
coefficient solution. Some allowances are made for other cross sections because
adequate results can be obtained from such sections.
Measuring the Real Fracture Toughness of Ceramics: ASTM C 1421 533
Despite the commonality of bend testing a rectangular cross section, the

methods differ substantially in terms of the size, shape and development of the cracks
used, as illustrated in Figure 1. The SEPB uses a large, straight-through precrack that
2c
(a)
SCF
a
h Remove 4.5h to 5h
SEPB CNB
a1= a1 /W
a =(a1+a2+a3)/3W
ao= ao /W
W
a1
a1 a2 a3
ao
PUSH
(b) ROD
BALL
Strain Gage
TEST Si
SPECIMEN
So
Displacement
Transducer
SUPPORT
ROD
Figure 1. (a) Crack configurations allowed in ASTM C 1421. (b) Example of a test specimen in four point
loading.
is rapidly formed from a small Vickers indentation or a shallow machined notch. The
CNB uses a 3-dimensional crack that is generated stably during fracture testing of the
section. The SCF uses a semi-elliptical precrack generated by removing a portion of a
Knoop indentation precrack.
All the methods fulfill the fundamental requirement that a sharp, real crack be
used to detect the onset of crack extension. Ideally this extension should be stable,
and it generally has been shown to be for both the CNB and SEPB methods. The
definitions of sharp and real differ somewhat from those used for metals. For metals,
a sharp, real crack is generated with low stress intensity cyclic fatigue [3] and a
minimal amount of plasticity. Because plasticity is generally minimal in ceramics,
sharp cracks can usually be developed with monotonic loading.
In this paper, rather than recite the specifications in C 1421, we have chosen
instead to focus on some of the key interferences in fracture toughness testing and
how they were dealt with in C 1421.
3. THE INTERFERENCES
3.1. Stability
A basic requirement for a valid experiment with the CNB specimen is stable crack
extension. Unstable crack extension, which may occur in test pieces with poorly
prepared notches, typically causes the test piece to be overloaded and leads to
overestimates of fracture toughness. Generally, notch width between 0.25 and
0.30 mm at any point on the specimen surface is available from commercial machine
shops. Larger notch thickness is acceptable provided that stable crack extension
occurs. The offset between the notch planes is particularly critical. For alpha silicon
carbide, notch widths of 0.225 to 0.250 mm with an offset less than 0.030 mm were
consistently stable. For sapphire, notch widths of 0.300 to 0.325 mm with notch
offsets of 0.075 to 0.110 mm were consistently unstable. However, notch widths of
0.250 to 0.300 mm with offsets of 0.025 to 0.075 mm were stable. Wider notches may
allow for more total offset.
On the other hand, there may be instances when seemingly unstable fractures
are in fact valid stable fractures. The inability to adequately detect stable crack
extension results in the appearance of an unstable test and the presumption of an
invalid result. A variety of methods have been used to verify stability such as load-
point displacement (LPD) as determined from the linear variable differential trans-
ducer (LVDT) used to control the test system actuator; center-point displacement
(CPD) as determined using an LVDT or capacitance extensometer placed in or near
the flexure fixture; a time record as generated with an XY recorder; or strain as
monitored with a strain gage centered about the crack plane on the compressive or
back-face of the test specimen [12,13]. In addition, laser extensometers have also been
used [14].
Generally, actuator displacement or time-sweeps are only marginally adequate
for detecting specimen stability. Success depends upon the choice of scales, the test
system ‘‘noise’’ levels, and the fracture behavior of the material. Locally placed
extensometers and strain gages are usually excellent. Figure 2.2 shows load as a
function of various detection methods for a-SiC.
400 40
35
(a) (a)
300 30
Force (N)
25
Force (N)
200 20
15
100 10
5
0 0
0 0.04 0.08 0.12 0 0.01 0.02 0.03 0.04
Actuator Displacement (mm) Actuator Displacement (mm)
400 40
(b) 35 (b)
300 30
Force (N)
Force (N)
25
200 20
15
100 10
5
0 0
0 200 400 600 800 1000 1200 0 5 10 15 20 25 30 35
Time (s) Time (s)
400 40
35 (c)
(c)
300 30
Force (N)
Force (N)
25
200 20
15
100 10
5
0 0
0 50 100 150 200 250 300 350 400 0 50 100 150 200 250 300 350 400
Backface Strain (μ m/m) Backface Strain (μ m/m)
2.1 - SEPB test of silicon nitride. 2.2 - CNB test of SiC.
Figure 2. Plots of (a) force vs. time, (b) force vs. actuator displacement, and (c) force vs. back face strain
for silicon nitride tested using the SEPB (2.1) and silicon carbide tested using the CNB (2.2). In both cases
stable crack extension occurred as evidenced by the nonlinearity near maximum load in the BFS plots
2.1(c) and 2.2(c). The other methods did not detect the crack extension.
For the SEPB test specimen, stability is not required, however, the lack of
stability can [15] result in overloading and slightly increase the measured fracture
toughness. Actuator displacement is inadequate for detecting test stability even with
a digitally controlled electro-mechanical system (e.g. Instron 85624 ) using relatively
sophisticated zero shift suppression and range (amplification) capability. Locally
placed extensometers are only marginally adequate whereas back face strain (BFS)
is excellent for detecting stability, as shown in Figure 2.1.
In addition, excessive stability in the SEPB can create problems because the
precrack length is used to calculate the fracture toughness. If the crack extends
significantly, the measured maximum load might not correspond reasonably with
the precrack length, and lower, apparent fracture toughness will result. This is in part
due to the current operational procedure of C 1421, which uses maximum load and
precrack length to calculate the SEPB fracture toughness, KIpb . Ideally, the calcula-
tion should be based on the load and the crack size corresponding to the onset of
crack extension. Generally, most ceramic materials exhibit sufficient but not exces-
sive stability when tested in four-point flexure [13]. The only notable exception is
when the precracks are very twisted [16]. Such cracks tend to exhibit complete
instability or alternatively, jumps followed by unstable failure as detected with
back-face strain. The stability exhibited in four-point flexure probably results be-
cause of factors such as stress corrosion, crack shape, and crack growth resistance.
However, it is notable that limited experimental work indicated that four-point
flexure is unstable and results in an overestimate of the SEPB fracture toughness as
compared to three-point results [17]. That finding may merely have been due to a
failure to detect stability (the actuator displacement was used), and excess extension
or other differences in the comparative three-point tests. A comparison of fracture
toughness calculated with the C 1421 procedure (maximum load and precrack
length) and the use of the precrack length and load corresponding to 2% crack
extension, as done in ASTM E 399, indicated little practical difference for four-
point flexure [13], as shown in Table 1. Although the standard allows user to choose
either configuration, it is recommended that four-point flexure be used because it
avoid errors associated with aligning the crack and load-point.
Stability is not normally an issue with the SCF test. The fractures usually are
unstable unless unusual precautions are taken, or the material has a very steep
R-curve.
3.2. Stress Corrosion

Many ceramics and glasses exhibit slow crack growth or stress corrosion. This
interferes with fracture toughness measurements because the results become a func-
tion of the test environment used and stress-intensity rate applied. Slow crack growth
occurs at stress intensities less than the critical stress intensity, and any such crack
extension may cause the measured fracture toughness to underestimate the fracture
toughness obtained in an inert environment. An example of the significance of the
effect for an alumina is shown in Table 2 [16]. To mitigate the interference, the choice
Table 1. Stress intensity factors and standard deviations calculated with various operational
p
procedures [13]. Units are MPa m and uncertainties are one standard deviation.
Maximum load and 2% extension load and

Material4 optical precrack length optical precrack length Maximum load and crack
{number of tests} (C 1421 procedure) (E 399 procedure) length estimated from BFS
Glass {4} 0.70 0.01 0.70 0.01 0.70 0.04

Si3 N4 (GN 10) {6} 5.42 0.17 5.38 0.20 5.73 0.12
Si3 N4 =SiCWiskers {3} 6.62 0.17 6.56 0.16 6.48 0.03
Si3 N4 (AS800) {4} 7.52 0.34 7.38 0.33 7.53 0.49
Si3 N4 (NKK) {4} 10.3 0.6 10.2 0.6 10.1 0.4
4
Standards and Technology or the National Aeronautics and Space Administration nor does it imply
that these materials or equipment are necessarily the best for the purpose.
Table 2. Effect of test environment on the measured fracture toughness one standard
deviation for ALSIMAG 614 alumina [16]. The number of tests is given in parentheses.
Test Environment
Test Specimen
Configuration Water Air Silicone Oil or Dry N2
p
KIvb (A)MPa m 2.75 0.01 (4) 3.19 0.07 (7) 3.37 0.05 (4)
p
KIsc MPa m 3.18 0.10 (5) 3.47 0.12 (4)
of test environment should be tied to the application if possible. For example, in the
calculation of slow crack growth parameters from ‘‘dynamic’’ fatigue or ‘‘static’’
fatigue data, the fracture toughness in an inert environment is combined with the
inert strength to indirectly estimate the initial flaw size distribution of the test
specimens. However, for field applications, a value measured in an environment
such as air may be more realistic, depending upon the loading conditions and
crack velocity expected at failure.
Very fast loading rates have been used to eliminate environmental effects in
strength testing of ceramics. However, the use of fast rates for SEPB and CNB
specimens is generally impractical because the specimens are more likely to be over-
loaded and unstable. An inert environment should be used instead to eliminate the
environmental effects. In order to determine the presence and significance of envir-
onmental effects, C 1421 recommends testing either in several environments (e.g. oil
and water) or at very different rates. One caution in using slow rates for testing is that
the correspondence between the optically measured precrack and the maximum load
may be lost, and although an effect will be indicated, the measured value may not be
accurate.
SCF precracks are especially sensitive to stable crack extension from environ-
mental effects. Reference 18 shows examples for alumina and several other ceramics.
C 1421 recommends that the precracks be examined for fractographic evidence of
stable extension and if it is detected, the critical crack size should be used for the
fracture toughness calculation. Testing in inert atmospheres is effective in eliminating
the interference in SCF testing.
3.3. Crack Growth Resistance
Ceramics such as coarse grained aluminas, in situ toughened silicon nitrides, and
transformation toughened zirconias exhibit an increase in the measured stress inten-
sity factor as a function of crack extension. This behavior is commonly referred to as
crack growth resistance or R-curve behavior. This phenomenon affects fracture
toughness measurements because many components are expected to fail with rela-
tively short crack lengths (< 0:5 mm) whereas many fracture toughness measure-
ments are performed with longer crack extensions and/or precrack lengths. Thus the
point of determination of fracture resistance on the crack growth curve does not have
similitude with the applications. Another disconnect between applications and many
methods is the existence of cyclic loading during formation of the crack. Cyclic
loading tends to break down the toughening mechanisms exhibited by some materials
and thus renders the mechanisms less effective in applications.
To address these interferences, test specimens could be prepared with fatigue

precracking and the measurements made at small, defined crack extensions. How-
ever, the additional effort is significant and was considered too substantial for the
first version of the standard. Instead, the effort was focused upon consistency and
accuracy between several methods for materials exhibiting little crack growth resist-
ance. C 1421provides warnings to alert the user to this interference.
It should be noted that both the CNB and SEPB test methods do produce very
similar results for in situ toughened silicon nitrides despite the differences in the
formation, growth and size of the crack. The CNB grows a crack monotonically
0:86 mm during the test whereas as the SEPB uses a 2 mm precrack that is
subsequently failed with monotonic loading that typically involves 5% crack
extension [13]. Table 3 compares results from these relatively different methods for
several in situ toughened silicon nitrides. The agreement is surprisingly good. The
SCF method is difficult to apply to coarse grained materials because the precrack
becomes difficult to delineate. This is particularly unfortunate because the SCF
method should provide fracture resistance values more akin to those developed in
real applications.
For alumina with a 5 mm average grain size, the effects of compressive or tensile
fatigue precracking versus monotonic precracking via large chevron-notch short bar
(CNSB) specimens shows the fatigued test specimens to have lower values that were
not a function of crack length for very long cracks [21]. Comparison of the data using
student’s t distribution indicated a statistically
p significant difference between mono-
tonically grown long cracks (3:93p MPa m for an extension of 16 mm) and fatigue
induced long cracks (3:56 MPa m for extensions between 3 and 30 mm) at better
than 99% confidence, implying that the best approach is to keep crack length as short
as possible while employing good metrology, regardless of the precracking method,
as shown in Table 4. Thickness effects, which play a role in the plasticity and degree
of plane strain developed, could not be detected in testing of relatively thick alumina
test specimens, as shown in Table 5. Thus the small section sizes used in ASTM C
1421 are likely reasonable, however, additional testing of very thin sections should be
performed.
Table 3. Fracture toughness of in situ toughened silicon nitrides as measured by the methods
specified in ASTM C 1421. Results in any row are for the same billet and have identical test
orientation except as noted. The number of tests is given in parenthesis. Exceptions to criteria
for validity are given in the footnotes.
p p p
Material KIvb (A)1 MPa m KIpb MPa m KIsc MPa m
NKK 10.43 0.48 (5)2 10.29 0.63 (3) Unmeasurable3

AS440 7.16 0.27 (5) 7.3 0.3 (7)4
AS800 (light) 7.40 (1) 7.60 0.20 (4)
AS800 (dark) 6.90 0.09 (4) 7.17 0.20 (5)
PY6 6.30 0.03 (5) 6.19 0.01 (2)5
SRBSN 147 31 N5 5.28 0.17 (5) 5.59 0.22 (14) 5.58 0.24 (16)
1. Geometry A of ASTM C 1421.

2. Geometry D tested at 0.1 mm/min.
3. Not measurable due to a coarse grain structure that masked the crack size.
4. See ref. 19. Inner and outer spans of 10 and 18 mm were used.
5. Inner and outer spans of 10 and 18 mm were used. 6. See ref. 20.
Table 4. Effect of precracking method on the measured fracture toughness of ALSIMAG 614
alumina [21].
Precrack Size Number Fracture

or Extension of Toughness
p
Test Method Precracking Method mm Tests MPa m
ASTM C 1421 Methods:
CNB Monotonic 0.86 7 3.19 0.07
SEPB Monotonic 1 3 13 3.09 0.17
SCF Monotonic 0.21 0.42 5 3.18 0.10
Long Cracks by Monotonic Crack Extension:
CNSB1 Monotonic 6.3 20 3.67 0.05
CNSB Monotonic 15 3 3.93 0.04
DT 2
Monotonic 3.84 0.05
Long and Short Cyclic Fatigue Precracks:
CNSB3 Tension/Tension 3 30 17 3.56 0.12
CNSB 3
Tension/Tension Dry N2 7 9 5 3.90 0.08
SNSB4 Compression/Compression 0.1 0.5 11 3.52 0.07
1. Chevron-notch short bar (CNSB).

2. Double torsion. See ref. 22.
3. Fatigued as a CNSB to the extension listed, and then machined into a straight-notch short bar (SNSB) and fractured.
4. Fatigued in compression as a straight-notch short bar (SNSB).
Table 5. Effect of test specimen thickness on the measured fracture toughness of ALSIMAG
614 alumina. The number of tests is given in parentheses [21].
Tension Tension Fatigue Precracked, Tested in N2 :
Test Specimen Thickness, mm 7.6 12.7 17.8

p
Fracture Toughness, MPa m 3.92 0.13 (2) 3.90 (1) 3.88 0.10 (2)
Compression Compression Fatigue Precracked, Tested in Air:
Test Specimen Thickness, mm 5 10

p
Fracture Toughness, MPa m 3.53 0.08 (6) 3.51 0.07(5)
Crack growth resistance curves measured with short and long crack techniques
might be expected to measure different ranges of the R-curve depending the exten-
sion and area of the cracked region. Figure 3a shows R-curves measured with the
indentation strength technique [23] and with chevron-notch short bar test specimens
[24]. For comparison, the values of KIpb , KIvb , and KIsc are included in the plot. The
R-curves generated with short and long cracks not only cover the same range of
fracture resistance, but start just below the fracture toughness values generated with
C 1421. The IS data show a sharply increasing R-curve trend, which is not reflected in
the other data. The IS trend is probably due to an artificial stabilizing influence of the
5
ALSMAG 614 Alumina, Air
Fracture Toughness or Resistance
4
MPa√m
3 SCF KIsc
CNB KIvb (A)

SEPB KIpb
2 IS R-curve
CNSB R-curve B=10.1 mm
CNSBR-Curve B=24.4 mm
1
0 5 10 15 20
Crack Extension or Precrack length, mm
Figure 3a. Crack growth resistance curves from indentation strength (IS) [23] and chevron notch short
bar methods [24] compared to fracture toughness values via C 1421. For the PB and SCF methods, the
precrack length is plotted. For the CNB method, the crack extension to maximum load is plotted.
5
ALSIMAG 614 Alumina, Air
Fracture Toughness MPa√m
3
Compression-Compression Fatigue
Tension-Tension Fatigue
CNSB B=10.1 mm
2 CNSB B =24.4 mm
SEPB KIpb
SCF KIsc
CNBKIvb(A)
1
0 5 10 15 20 25 30
Crack Extension or Precrack Length, mm
Figure 3b. Fracture toughness measured after fatigue precracking compared to fracture toughness values
via C 1421. For the PB, SCF, and fatigue methods, the precrack length is plotted. For the CNB method,
the crack extension to maximum load is plotted.
residual stress field associated with the Vickers indentation. Some recent studies [25,
26] have shown that by varying the SCF precrack size by using a range of indentation
loads, it may be possible to gain a sense of whether the calculated fracture toughness
depends on crack size and whether R-curve phenomena are active.
Table 6. Effect of texture on the measured fracture toughness of ALSIMAG 614 alumina.
The number of tests is given in parentheses [24].
Test Orientation:
Processing
Method RC LC RL
Extruded 4.94 0.37 3.63 0.21 3.99 0.12
Isopressed 3.69 0.05 3.64 0.03 3.68 0.02
Interestingly, the fracture toughness measured with fatigue precracked speci-

mens appears to be higher than that measured with C 1421 techniques, as shown in
Figure 3b. Thus C 1421 provides values on the lower end of the R-curve without
fatigue precracking.
Texture due to the manufacturing process can also influence the crack growth
resistance and fracture toughness as shown in Table 6 [24]. Alignment of pores and
elongated grains produces both crystallographic texture and macro-textures that
influence the measured properties.
3.4. Stress Intensity Factor Coefficients
Accurate stress intensity factor coefficients (SCIF) are particularly of concern for the
CNB test specimen. This is mainly a result of its 3-dimensional nature and the use of
approximate solutions such as the Slice and straight-through-crack-assumption
(STCA). In order to determine the best SIFC function, the Slice and STCA models
were compared to finite element analyses (FEA) and boundary element analyses (BEA)
solutions [16]. The STCA is a good approximation to the BEA for geometries (see
Figure 1) with a1 ¼ 1:0 (which include configurations A, B, D in C 1421), but a poor
approximation for geometries with a1 < 1 (such as configuration C in C 1421) as shown
in Figure 4 and summarized in Table 7. For geometries with a1 < 1 the Slice is a better
but less than ideal approximation, and FEA results are currently the best solution.
For the SEPB test specimen, good SIFC’s are available for configuration similar
to those used in testing of metallic materials. However, ceramics are frequently tested
in shorter sections that alter the span-to-depth ratio and thereby affect the SIFC. In
order to allow a wider range of span-to depth ratios, previously unpublished solu-
tions generated at the Materials Technology Laboratory, Watertown, Massachusetts
[27] were reviewed and added to C 1421.
For the SCF, the SIFC, Y, for semicircular and semi-elliptical surface cracks in
bending are from the empirical equations developed by Newman and Raju [28]. The
Newman-Raju formulas [28] for Y are widely accepted and have estimated accuracies
of a few percent. The SIFC depends upon the ratio of a to c, the cross section size relative
to a and c, and the presence of bending. The Y factor is not constant and varies along the
precrack boundary. It is important to use the maximum value of Y, which will either be
at the deepest point of the precrack, or where the precrack intersects the surface.
Fracture from the deepest point of the precrack is preferred and can be pro-
moted by polishing sufficient material off the test specimen so as to make the
precrack a shallow semi-ellipse. In some materials (e.g., those with clear markings
8
Stress Intensity Factor Coefficient, Y*

ao / W a1 / W
7
0.2 1.0 BEA
0.2 1.0 Slice
6
0.2 1.0 STCA
ao / W a1 / W
2
0.2 0.7 FEM
1 0.2 0.7 Slice
0.2 0.7 STCA
0
0.2 0.3 0.4 0.5 0.6 0.7
Normalized Crack Length, a/ W
Figure 4. Stress intensity factor coefficient Y for CNB test geometries A and C of ASTM C 1421 as
calculated with the Slice, STCA, and FEA/BEA methods.

Table 7. Comparison of minimum stress intensity factor coefficients, Ymin , from BEA and
FEA methods, and the Slice and STCA approximations.
Numerical % %
ao a1 B/W Si mm So mm Methods Slice Diff. STCA Diff.
0.2 0.7 0.5 0 40 2.850 (FEA) 2.695 5:4 3.216 12.8
0.2 0.7 0.5 20 40 2.979 (FEA) 2.801 6:0 3.347 12.3
0.2 1.0 0.5 20 40 4.240 (BEM) 4.549 7.3 4.233 0:2

1
0.25 1.0 1.0 20 40 4.719 (FEA ) 5.061 7.3 4.690 0:6
2
0.37 0.8 1.0 0 7.3 4.540 (FEA ) 4.527 0.3 4.962 9.3
0.44 1.0 1.0 0 38 6.910 (FEA3) 7.514 8.7 7.147 3.4
1. See Ref. 29. 2. See Ref. 30. 3. See Ref. 31.
on the fracture surfaces) it has been possible to confirm that fracture initiated from
the point of the periphery where Y was maximized. One notable conclusion of one of
the VAMAS round robins [5] was that the computed fracture toughness was insensi-
tive to the exact precrack boundaries marked, because the computed fracture tough-
ness was not sensitive to the precrack size measurement! This is partly due to the
dependence of the fracture toughness on the square root of the crack size. Thus, a
10% error in crack size is diminished to a 5% error in fracture toughness. A seren-
dipitous outcome was the offsetting influence of Y on the errors due to crack size
measurements. For example, if the crack depth, a, is overestimated, the correspond-
ing calculated Y is underestimated.
3.5. Residual Stresses
Residual stresses affect the fracture toughness tests in C 1421 in different ways. For
the CNB test specimen, residual stresses associated with grinding of the chevron
notch have been anecdotally associated with unstable fracture from the notch tip.
For the SEPB test specimen, residual stresses associated with indentation used to
form the precrack have been demonstrated to lower fracture toughness meas-
urements because the crack is wedged open by an additional, unmeasured stress
intensity. Figure 5 illustrates the effect for a silicon nitride, and implies that
indentation loads should be less than 100 N for silicon nitrides [32]. For a SiC,
which is relatively hard, the removal of 50 N indentations by lappingpresulted in
comparablepvalues to those measured without lapping (2:52 0:11 MPa m vs: 2:58
0:08 MPa m) [33].
C 1421 recommends that if residual stresses from the indentation are suspected,
the indentations may be removed by polishing or hand grinding. Annealing may
be used provided that the crack tip is neither blunted nor the crack tip/planes
‘‘healed.’’
For the SCF test method, residual stresses around the Knoop precrack affect the
measured fracture toughness and thus have been studied in some detail. The stresses
are minimized or eliminated by removing the indentation and surrounding material
by using hand polishing or machine grinding to a depth of 4.5 h to 5.0 h (where
h ¼ d=30 in Figure 1). Most studies on silicon nitride imply that as little as 3h is
sufficient [34 36], however for SiC, recent studies comparing SCF, CNB, and SEPB
results indicate that more removal may be necessary to eliminate lateral cracks, as
shown in Figure 6 [37].
3.6 Crack Length Measurement

The measurement of crack length has been identified as a major source of error in
fracture testing of metallic materials [38]. In ceramics, crack length measurement is
particularly difficult for a number of reasons: the small specimens typically used; the
6.5
Fracture Toughness
6.0 Indentation
then Heat Treat
MPa√m
5.5 As Indented
Heat Treat then
Indentation
5.0
0 100 200 300 400 500 600 700
Indentation Force (N)
Figure 5. SEPB fracture toughness of for hot pressed silicon nitride as a function of indentation force
without and with heat treatments [32].
4.0
ASTM C1421
Original Range
recommendation X = 4.5 - 5.0h
X > 3h Other Data
by
Fracture Toughness, MPa√m
3.0 ASTM C 1421

CN-b
CN-a
CN-a
SEPB
SEPB
Lateral crack Lateral cracks CN-c
remnants eliminated
2.0 Residual present
tensile stress
1.0
0.0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
n, Number of Indentation Depths Removed
Figure 6. SCF, CNB and SEPB fracture toughness test results for specimens from one lot of alpha silicon
carbide [37]. SCF fracture toughness values overestimate the correct value if the amount removed, at
X nh, is insufficient to remove lateral cracks that interfere with the main precrack.
semi-opaque nature of many glass-based ceramics; the lack of plastic deformation at

the crack tip which results in poor definition of the crack front; and the high elastic
moduli and low fracture toughnesses that result in small crack opening displacements.
The CNB test specimen does not require crack length measurement, thereby
providing some convenience to the user. However the SEPB and SCF test methods
require quality measurements if accurate fracture toughness estimates are expected.
For the SEPB test specimen, the precrack front is very difficult to delineate unless it
leaves a crack arrest line. An arrest line is easily detected if the crack tip turns out
of plane or is hooked, thereby making the precrack and fast fracture plane non-
coplanar, as shown in Figure 7. In one study, some cracks were so planar that the
front could not be detected with conventional optical microscopy [12]. In such cases,
the use of dye penetrants have been successful as shown in Figure 8, however, in some
cases penetrants tend to bleed on the fracture surface and the measurements must be
made immediately. Dye penetrants must be used with caution as they may cause
environmentally assisted crack growth.
Another simple and sensitive technique that can be used to measure crack length
and extension in both the CNB and SEPB techniques is the back-face strain gage
(BFSG). It has been used previously in metallic compact tension specimens [39], and
in fracture toughness testing of ceramic flexure specimens [12, 13]. Figure 9 shows a
strain gage mounted on the back-face of a flexure specimen and Figure 10 show the
Figure 7. Effect of precrack shape on optical visibility: (a) and (b) are side and top views of a coplanar
precrack and failure plane; (c) and (d) are views of a precrack with a hooked tip [12].
correlation between measurements made with BFS and optical microscopy for glass,
silicon nitrides, and a - SiC.
The SCF method uses a very small precrack introduced by a Knoop indenter.
This method of relying on the material’s brittleness to cause small surface precracks
to form underneath an indention is unique to ceramics. The precrack is small and on
the order of real flaws in ceramics. Thus, fractographic techniques, and some care
and skill, are needed to see and measure the size of the precrack. Both the cracks
depth and the width must be measured for the SCF precrack. Crack detection and the
size measurement is aided by purposely tilting the test specimen 1=28 relative to the
indenter so that the precrack and fast fracture regions are not coplanar. As a result,
the precrack in some materials is easy to discern, even with a low power stereo
binocular microscope. In other materials, a careful examination with a scanning
1.0
SLOPE = 1
a/ W Without Penetrant 0.8 GN10 Cycled

NKK Cycled
0.6
0.4 Ideal
Kryptonite
GN10
0.2 NKK
AS 800
SiC
0.0
0.0 0.2 0.4 0.6 0.8 1.0
a/ W With Penetrant
Figure 8. Normalized crack length in SEPB test specimens as measured with and without penetrant [12].
Data points on the x axis were not measurable without using dye penetrant.
P/2 P/2
Si
So
Figure 9. Illustration of a strain gage mounted on the back face of a four point flexure specimen.
electron microscope may be necessary. A variety of features, including color, con-

trast, brightness, ridge or rim lines, halos, and micro hackle line redirection help to
discern the precracks, as shown in Figure 11 [40,41]. Stereo scanning electron
microscope (SEM) photo pairs may be the best solution in very difficult materials.
Dye penetration techniques may also be useful. A detailed and well-illustrated
discussion of fractographic techniques to detect SCF precracks has been presented
[36] and C 1421 has a detailed appendix on this topic. As discussed above in the Stress
Intensity Factor Coefficient section, the calculated SCF fracture toughness is sur-
prisingly insensitive to the precision of the crack size measurement.
1.0
GLASS SLOPE =1
GN10
a/ W via Back-Face Strain 0.8 KRYPTONITE
AS800
NKK
SiC
0.6
0.4
0.2
GN10 Cycled
NKK Cycled
0.0
0.0 0.2 0.4 0.6 0.8 1.0
Normalized Precrack Length,a/ W
Figure 10. Comparison of precrack lengths in SEPB test specimens as determined by optical measure
ments and by the initial compliance as measured with back face strain [13].
Figure 11. Example of a SEM photograph of a precrack generated in 99.9% sintered alumina with the
SCF method.
3.7. Crack Shape

Crack shape can be a problem in the SEPB test method if the precracks hook or twist
excessively because the fixture or indentation was not accurately aligned. The effect
of hook angle on measured toughness is shown in Figure 12. Little trend can be seen.
Currently, the standard recommends keeping the angle below 108 , however, this may
be conservative based on currently available data.
7.5
Fracture Toughness, MPam 7.0
6.5
6.0
ε
5.5
5.0
0 10 20 30 40
Maximum Hook Angle, degrees
Figure 12. Fracture toughness as a function of hook angle for the SEPB test on silicon nitride [13].
As discussed in the section on Stability, crack twist appears to effect crack growth
and stability. For example, during testing of ADS96R alumina, two test specimens with
relatively twisted cracks exhibited nopstable crack extension and greater p fracture
toughness values (3:96 and 4:00 MPa m vs. 3.63, 3.68 and 3:83 MPa m for the
planar cracks). Similar behavior was exhibited by the a-SiC test specimens: Twisted
cracks exhibited crack jumps followed
p by unstable failure
p and greater fracture tough-
ness values (2:74 and 2:72 MPa mvs:2:58 0:08 MPa m for four tests). More sys-
tematic study of such effects are needed. Currently, the standard mitigates the issue by
recommending that twist angle be less than 108 .
A semi ellipse is used to model the SCF precrack created by Knoop indentation.
In reality, the subsequent hand polishing or grinding to eliminate residual stress
truncates the semicircle or semi-ellipse precrack shape. Approximating the crack by a
semi ellipse is satisfactory if the maximum stress intensity is at the deepest portion of
the precrack, but is less satisfactory if the maximum Y is at the surface. The initial
test piece tilt 1=28 ) during indentation causes precracks to pop in at angles from 0 to
2 38 off perpendicular and greatly enhances the visibility of the precracks on the
fracture surfaces. Specimen tilt angles greater than 28 are unsatisfactory, however,
because the precracks often pop in 58 or more off perpendicular. This is felt to be
excessive, especially since Mode II loading may be introduced to the crack and the
stress intensity factor coefficients, Y might be affected as discussed in appendix 5 of
the VAMAS Round Robin report [8]. For precrack tilt angles up to 58, the influence
on Y is negligible.
C 1421 recommends that SCF precrack be prepared to force the maximum Y
factor to be at the deepest part of the precrack. Complications arise if, on the other
hand, the maximum Y is at the intersection of the precrack with the test piece surface.
The complications include local perturbations of the crack front, uncertainties about
the ellipse approximation to the truncated semicircle or semi ellipse, possible inter-
actions with surface damage, and uncertainties associated with the Newman-Raju
solutions at that site. For most ceramics, the amount of material that must be
removed to eliminate the residual indentation stresses is only about three times the
indentation depth. This usually leaves a nearly semicircular precrack, however, which
still has a maximum Y at the test piece surface. Removal of 4.5 to 5.0 times the
indentation depth creates a shallower precrack ( 0:6 0:8 aspect ratio, depth to
half width) and shifts the maximum Y to the deepest part of the precrack as desired.
3.8. Method to Method Variation
Because the cracks used in the standardized methods are so different, substantially
different results might be expected, even for fairly well behaved materials. Table 8
compares results from the three methods and demonstrates that very comparable
results can be obtained with good metrology. In most cases, the results are within a
few percent, and in the cases of discrepancy, the variations are usually the result of
different billets being used. In the case of a-SiC, the presence of lateral cracks has
also caused differences. If the lateral cracks are not removed, then p the measured
fracture toughness increased from 2:76 0:08 and 2:64
p 0:05 MPa m for the JAS
and UW billets to 2:86 0:03 and 3:01 0:35 MPa m, respectively [37].
Table 8. Fracture toughness and standard deviation measured by the methods specified in
ASTM C 1421. Results in any row are for the same billet and have identical test orientation
except as noted. The number of tests is given in parenthesis. Exceptions to criteria for validity
are given in the footnotes.
p p p
Material KIvb (A)MPa m KIpb MPa m KIsc MPa m
a SiC (JAS) 2.61 0.05 (6) 2.58 0.08 (4)1 2.76 0.08 (4)2
a SiC (UW) 2.62 0.06 (6) 2.54 0.20 (3)1 2.64 0.05 (4)
ADS96R 3.56 0.03 (5) 3.71 0.10 (3)
ALSIMAG 614 4
3.19 0.06 (7) 3.09 0.17 (13)1,3 3.18 0.10 (5)
ALSIMAG 6146 3.13 0.03 (4)5 2.98 0.06 (4)1,7
NC1328 SRM 2100 4.58 0.12 (10) 4.58 0.10 (19) 4.56 0.13 (17)
NT154 5.18 0.11 (5) 5.21 0.02 (5) 5.80 0.239 (5)
SN260 5.19 0.06 (5) 5.13 0.15 (4)10
SiAlON 2.45 0.09 (6) 2.55 0.05 (4)11
1. Different billet from the same material batch as the CNB geometry A specimens.
2. See ref. 37.
3. See ref. 21
4. Processed by isopressing.
5. B ¼ 4:06mm, W =5:08mm,ao ¼ 0:2 and a1 ¼ 1 with inner and outer spans of 12.7 and 38.1 mm.
6. Processed by drypressing mild texture was present.
7. A single 30 kg indentation was used as the PB starter crack.
8. Billet ‘‘C’’ of SRM 2100 in ref. 10.
9. Different material lot. Strong lot to lot variation has been reported.
10. A single 20 kg indentation was used as the PB starter crack. Test spans of 10 and 18 mm.
11. Two of the data points were measured by Westmorland Mechanical Testing, Youngstown, PA.
4. STANDARD REFERENCE MATERIAL
For researchers with little experience in measuring fracture toughness via the
methods in C 1421, valuable experience and the adequacy of a test setup can be
verified by performing tests on material of a know fracture toughness. Such a
material has been prepared and made available by the National Institute of Stand-
ards and Technology [10].
Standard Reference Material (SRM) 2100 is the first reference material in the
world for the property of fracture toughness for any class of material. Although
SRM 2100 may be used with any credible fracture toughness test method, it was
optimized for those that use beams in bending such as in C 1421.
A package of SRM 2100 consists of a set of five hot pressed silicon nitride beams
cut from a single billet. The billet was originally fabricated in the form of plates
158 mm 158 mm 20 to 30 mm thick. SRM 2100 is a very consistent material from
a mechanical properties standpoint, owing to its uniform microstructure ( 3 mm
grain size) and small amount of sintering aid ( 1 % mass fraction magnesium oxide).
The material is fully dense with a density slightly greater than theoretical for b silicon
nitride because of very fine tungsten carbide inclusions.
In developing the SRM data base, a large sample set of representative flexure
test specimens (3 mm 4 mm 47 mm) from each of three billets of the material
were prepared in accordance with the procedures of C 1161 [11] and C 1421. The test
specimens were cut from the billets with their long axis perpendicular to the hot
pressing direction, and thus the certified fracture toughness of SRM 2100 is for
cracks propagated parallel to the 3 4 mm cross section.
Fracture testing was in laboratory conditions using semi to fully-articulating
flexure fixtures. Four-point flexure fixtures with 20 mm and 40 mm inner and outer
spans, respectively were used to test the full length SCF, SEPB and CNB test
specimens. In some instances, broken halves of test specimens were retested in
three- and four-point flexure per the SEPB method.
Table 8 summarizes the fracture toughness results for one billet. Within a given
billet, the fracture toughness results were in good agreement. The consistent results
confirm that the test methods of C 1421 have been refined to the extent that genuine
material variability (e.g., billet to billet, batch to batch) can be discerned, thus
fulfilling one goal of a good standard: to provide tools to measure the desired
property with good accuracy and precision.
ASTM C 1421 provides detailed guidance for a quality determination of the fracture
toughness of ceramic materials. The sources of error addressed in the standard
include stability, stress corrosion, crack growth resistance, stress intensity factor
solution accuracy, crack length measurement, crack shape, residual stresses, and
alignment of the crack with the fixture. Many of these have been sufficiently ad-
dressed as demonstrated by the comparable results for a wide variety of ceramics.
For situations in which the interferences are problematic, C 1421 gives recommenda-
tions and a choice of three methods to address the issue. The contribution of any
individual interference to the measured fracture toughness has generally been kept
below 2%. Use of the standard should avoid significant errors in the measurement
of fracture toughness.
ACKNOWLEDGEMENT
The authors thank the Low Emission Alternative Power Project for funding.
REFERENCES
1. ASTM C 1145 01, ‘‘Standard Terminology of Advanced Ceramics,’’ Annual Book of Standards, Vol.
15.01 (American Society for Testing and Materials, West Conshohocken, Pennsylvania, 2001).
2. A. Dodd and D. Murfin, Dictionary of Ceramics (The Institute of Materials, London, 1994).
3. ASTM E 399 90, ‘‘Standard Test Method for Plane Strain Fracture Toughness of Metallic Mater
ials,’’ Annual Book of Standards, Vol. 03.01 (American Society for Testing and Materials, West
Conshohocken, Pennsylvania, 2001) pp. 434 464.
4. ASTM C 1421, ‘‘Standard Test Method for Determination of Fracture Toughness of Advanced
Ceramics at Ambient Temperatures’’ Annual Book of Standards, Vol. 15.01 (American Society for
Testing and Materials, West Conshohocken, PA, 2001) pp. 641 672.
5. H. Awaji, J. Kon, and H. Okuda, ‘‘The VAMAS Fracture Toughness Test Round Robin on
Ceramics,’’ VAMAS Report #9, Japan Fine Ceramic Center, Nagoya, Japan, 1990.
6. G. Quinn, J.A. Salem, I. Bar on, K. Chu, M. Foley, and H. Fang, ‘‘Fracture Toughness of Advanced
Ceramics at Room Temperature,’’ Journal of Research of National Institute of Standards and Tech
nology, Vol. 97, No. 5, 579 607 (1992).
7. M. Mizuno, and H. Okuda, ‘‘VAMAS Round Robin on Fracture Toughness of Silicon Nitride at
High Temperature,’’ VAMAS Technical Report 16 (Japan Fine Ceramics Center, Nagoya, Japan,
1993).
8. G.D. Quinn, J.J. Kübler, and R.J. Gettings, ‘‘Fracture Toughness of Advanced Ceramics by the
Surface Crack in Flexure Method, A VAMAS Round Robin,’’ VAMAS Technical Report #17
(National Institute of Standards and Technology, Gaithersburg, MD, 1994).
9. G. D. Quinn, ‘‘The Fracture Toughness Round Robins in VAMAS: What We Have Learned,’’ in
Fracture Resistance Testing of Monolithic and Composite Brittle Materials, ASTM STP 1409, edited
by J. A. Salem, G. D. Quinn, and M. G. Jenkins (ASTM, West Conshohocken, PA, 2002) pp. 107
126.
10. G.D. Quinn, K. Xu, J.A. Salem, and J.J. Swab, ‘‘SRM 2100: The World’s First Fracture Toughness
Reference Material,’’ Published in this proceeding.
11. ASTM C 1161 96, ‘‘Standard Test Method for Flexural Strength of Advanced Ceramics at Ambient
Temperature,’’ Annual Book of Standards, Vol. 15.01 (American Society for Testing and Materials,
West Conshohocken, Pennsylvania, 1999) pp. 309 315.
12. J. A. Salem, L. J. Ghosn, and M. G. Jenkins, ‘‘A Strain Gage Technique to Measure Stable Crack
Extension in Ceramics,’’ in Post Conference Proceedings of the 1997 SEM Spring Conference on
Experimental Mechanics (Society for Experimental Mechanics, Bethel, Connecticut, 1997) pp. 1 8.
13. J.A. Salem, L.J. Ghosn, and M. Jenkins, ‘‘Back Face Strain as A Method for Monitoring Stable
Crack Extension,’’ in Ceramic Engineering and Science Proceedings, Vol. 19, Issue 3 4, (Am. Ceramic
Soc, Westerville, OH, 1998) pp. 587 594.
14. M.G. Jenkins, M.K. Ferber, A. Ghosh, and J. Peussa and J.A. Salem ‘‘Chevron Notch, Flexure Tests
For Measuring the Elevated Temperature Fracture Resistance’s of Structure Ceramics,’’ in Chevron
Notch Test Experience: Metals and Non Metals, ASTM STP 1172, edited by K. R. Brown & F. I.
Baratta (American Society for Testing and Materials, Philadelphia, Pennsylvania, 1992) pp. 159 177.
15. F.I. Baratta, and W.I. Dunlay, ‘‘Crack Stability in Simply Supported Four Point and Three Point
Loaded Beams of Brittle Materials,’’ Mechanics of Materials, 10, 149 159 (1990).
16. J.A. Salem, L.J. Ghosn, M.G. Jenkins, and G.D. Quinn, ‘‘Stress Intensity Factor Coefficients for
Chevron Notched Flexure Specimens and Comparison of Fracture Toughness Methods,’’ Ceramic
Engineering and Science Proceedings, Vol. 20, No. 3, 503 512 (1999).
17. I. Bar on, F.I. Baratta, and K. Cho, ‘‘Crack Stability and its Effect on Fracture Toughness of Hot
Pressed Silicon Nitride Beam Specimens,’’ Journal of American Ceramic Society, Vol. 79, No. 9, 2300
2308 (1996).
18. J. J. Swab and G. D. Quinn, ‘‘Effect of Precrack ‘‘Halos’’ on KIc Determined by the Surface Crack in
Flexure Method,’’ ibid., 81 [9] 2261 2268 (1998).
19. V. Tikare and S.R. Choi, ‘‘Combined Mode I and Mode II Fracture in Monolithic Ceramics,’’ J. Am.
Ceram. Soc., 76 [9], 2265 72 (1993).
20. G. D. Quinn, J. J. Swab, and M. J. Motyka, ‘‘Fracture Toughness of a Toughened Silicon Nitride by
ASTM C 1421,’’ ibid., 86 [6] pp. 1043 1045 (2003).
21. J.A. Salem, M.G. Jenkins, M.K. Ferber, and J.L. Shannon, Jr., ‘‘Effects of Pre cracking Method on
Fracture Properties of Alumina,’’ in Proceedings of Society of Experimental Mechanics Spring
Conference on Experimental Mechanics, Milwaukee, Wisconsin, June 10 13, (Society for Experimen
tal Mechanics, Bethel, Connecticut, 1991) pp. 762 769.
22. G.K. Bansal and W. Duckworth ‘‘Comments on ‘‘Subcritical Crack Extension and Crack Resistance
in Polycrystalline Alumina,’’ J. Mat. Sci. Letters, No. 13, pp. 215 216, (1978).
23. J.A. Salem, J.L. Shannon, Jr., and M.G. Jenkins, ‘‘Some Observations in Fracture Toughness Testing
with Chevron Notched Specimens,’’ in Chevron Notch Fracture Test Experience: Metals and Non
Metals, ASTM STP 1172, edited by K.R. Brown and F.I. Baratta (American Society for Testing and
Materials, West Conshohocken, PA, 1992) pp. 9 25.
24. J.A. Salem, J.L. Shannon, Jr. and R.C. Bradt, ‘‘Crack Growth Resistance of Texture Alumina,’’
Journal of the American Ceramic Society, 72, [1], 20 27 (1989).
25. T. Fett, D. Munz, and G. Thun, ‘‘Fracture Toughness and R Curve Behavior of PZT,’’ Research
Center, Karlsruhe, Technical Report KZKA 6058, July 1998.
26. K. Yasuda, T. Taguchi, J. Tatami, and Y. Matsuo, ‘‘Estimation of Short Crack R Curves of
Polycrystalline Ceramics by Surface Crack in Flexure Method,’’ pp. 115 120 in Improved Ceramics
Through New Measurements, Processing and Standards, Ceramic Transactions, vol. 133, (2002).
27. C.P. Freese, Unpublished Research, ARL, Watertown, MA.
28. J.C. Newman, Jr., and I.S. Raju, ‘‘An Empirical Stress Intensity Factor Equation for the Surface
Crack,’’ Engineering Fracture Mechanics, Vol. 15, No. 1 2, 185 192 (1981).
29. J. Joch, J. Zemankova and J. Kazda, ‘‘Analysis of a Chevron Notch Four Point Bend Specimen by
the Three Dimensional Finite Element Method,’’ Comm. Am. Ceram. Soc., C154 C155, March
(1988).
30. R. Kolhe, C.Y. Hui, A.T Zehnder, Effects of Finite Notch Width on the Fracture of Chevron
Notched Specimens,’’ International Journal of Fracture, 94 (2): 189 198 (1998).
31. M.G. Jenkins, A.S. Kobayashi, K.W. White and R. Bradt, ‘‘A 3 D finite Element Analysis of a
Chevron notched, Three Point Bend Fracture Specimen for Ceramic Materials,’’ Int. J. Fract. 34,
281 295 (1981).
32. N. Murayama, S. Sakaguchi, and F. Wakai, ‘‘Evaluation of Fracture Toughness using SENB
Specimens with Precrack Produced by Growing Microcrack Around Vickers Indent,’’ Journal of
Ceramic Society of Japan, International Editions, Vol. 95, pp. 980 982 (1987).
33. J.A. Salem, unpublished work.
34. G.D. Quinn, R.J. Getting, and J.J. Kübler, ‘‘Fracture Toughness of Ceramics by the Surface Crack in
Flexure (SCF) Method,’’ in Fracture Mechanics of Ceramics, Vol. 11, edited by R.C. Bradt, D.P.H.
Hasselman, D. Munz, M. Sakai, and V. Yashevechenko (Plenum Press, New York, NY, 1996),
pp. 203 218.
35. J.J. Petrovic, R.A Dirks, L.A. Jacobson, and M.G. Mendiratta, ‘‘Effect of Residual Stress on
Fracture from Controlled Surface Flaws,’’ J. of Am. Ceram. Soc., 59 [3 4] 177 178 (1976).
36. G.D. Quinn, R.J. Gettings, and J.J. Kübler, ‘‘Fractography and the Surface Crack in Flexure (SCF)
Method for Evaluating Fracture Toughness of Ceramics,’’ Fractography of Glasses and Ceramics,
Ceramic Transactions, Vol. 64, American Ceramic Society, Westerville, OH, 1996, pp. 107 144.
37. G.D. Quinn and J.A. Salem, ‘‘Effect of Lateral Cracks on Fracture Toughness Determined by the
Surface Crack in Flexure Method,’’ J. Am. Ceram. Soc., 85 [4] 873 880 (2002).
38. W.G. Clark, and S.J. Hudak, J. Testing and Eval., Vol. 3, No. 6, 454 476 (1975).
39. W.F. Deans, and C.E. Richards, ‘‘A Simple and Sensitive Method of Monitoring Crack Extension
and Load in Compact Fracture Mechanics Specimens Using Strain Gages,’’ J. Testing and Eval., Vol.
7, No. 3, 147 154 (1979).
40. G.D. Quinn, R.J. Gettings, and J.J. Kübler, ‘‘Fractography and the Surface Crack in Flexure (SCF)
Method for Evaluating Fracture Toughness of Ceramics,’’ Fractography of Glasses and Ceramics,
Ceramic Transactions, Vol. 64, (American Ceramic Society, Westerville, OH, 1996), pp. 107 144.
41. J.J. Swab, and G.D. Quinn, ‘‘Investigation of ‘Halos’ Associated with Fracture Toughness Pre
cracks,’’ Ceramic Engineering and Science Proceedings, Vol. 18, No. 4, 1997, pp. 173 182.
42. A. Ghosh, M.G. Jenkins, K. W. White, A.S. Kobayashi, and R. Bradt, ‘‘Elevated Temperature
Fracture Resistance of a Sintered a Silicon Carbide,’’ J. Am. Ceram. Soc., Vol. 77, No. 2, 242 247,
(February 1989).
43. D. Munz, R.T. Bubsey, and J.E. Srawley, ‘‘Compliance and Stress Intensity Coefficients for Short
Bar Specimens with Chevron notches,’’ Int. J. Fract., Vol. 16, No. 4 359 374 (1980).
MICROSTRUCTURAL
CONTROL OF INDENTATION
CRACK EXTENSION UNDER
EXTERNALLY APPLIED
STRESS
Robert F. Cook1
1. INTRODUCTION
In many ceramic and glass materials indentation half-penny cracks extend stably
under applied stress to over twice their original length before instability consistent
with an ideal identation response. Observations of such extension can be made with
custom-designed bending fixtures, allowing crack extension in indented specimens to
be observed with an optical microscope during controlled loading. The resulting
applied stress vs crack length traces exhibit a concave-down shape, with decreasing
derivative prior to instability at which the specimen breaks and the derivative
vanishes. Many materials, however, do not exhibit this behavior a consequence of
the influences of reactive environments, lateral cracking, microstructural toughening,
surface stresses, phase-transformation trapping and preferred crystallographic cleav-
age planes. These influences lead to subtantially suppressed or enhanced stable crack
extension, asymmetric extension and stick-slip extension. After a brief development
of the indentation fracture model, in-situ observations of indentation crack extension
in a set of cordierite glass-ceramic composites and a polycrystalline alumina are
presented and examined in terms of variability of microstructural interactions giving
rise to toughening.
1
Robert F. Cook, Materials Science and Engineering, University of Minnesota, Minneapolis, MN,
55455.
57
58 Robert F. Cook
2. INDENTATION FRACTURE MECHANICS
A fracture system comprising a crack in a brittle material generated by a (sharp)

elastic-plastic contact event and subsequently subjected to an applied stress is char-
acterized by the fundamental equilibrium equation1,2
Ka þ Kr ¼ Tbase þ Tm (1)
where Ka is the stress-intensity factor associated with the applied stress field, Kr is the
stress-intensity factor associated with the residual contact field, Tbase is the toughness
of the material in the environment of interest and Tm is a toughness amendment
associated with interactions between the crack and the microstructure of the material
or the component. By definition, Tbase is crack-length invariant and is related to the
equilibrium (fracture) surface energy of the material, g, by Tbase ¼ (2gE)1=2 , where E
is the material Young’s modulus: g depends on environment and takes a maximum
value of g 0 in an inert, non-reactive, environment, with an analogous maximum base-
toughness value of T0 . The remaining terms in Eq. (1) may all depend on the length
of the crack, c, and usually the applied stress-intensity factor is destabilizing,
dKa =dc > 0, the residual stress-intensity factor is stabilizing, dKr =dc < 0, and the
microstructural interaction leads to toughening, Tm > 0 and dTm =dc$0. Such micro-
structural toughening can be associated with the material, as in the cases of grain
related crack face frictional interlocking in polycrystalline alumina3 or metastable
phase transformation crack tip shielding in zirconia alloys.4 Or, toughening effects
can be imbedded in the component, as in the case of tempering-induced compressive
stress profiles generated in ion-exchanged glass parts.5 The equilibrium relation of
Eq. (1) has been expressed in the component frame of reference, in which the
microstructural interactions are associated with the toughess. Equilibrium can also
be expressed in the crack-tip frame of reference by setting Tm ¼ Km and collecting
all terms but Tbase on the left side of the equation, or in the applied loading frame of
reference by collecting all terms but Ka on the right side of the equation, thereby
treating Kr as a toughening element.
The variables in an indentation crack extension experiment are shown in Fig 1.
The indentation flaw consists of residual square contact impression (for a Vickers
diamond pyramid indenter) in the component surface with cracks perpendicular to
the surface emanating from the impression corners. The crack length, c, is charac-
terized by the surface trace, measured from the impression center. The applied stress,
sa , is imposed perpendicular to one set of cracks (for a uniaxial loading experiment)
The size of the contact impression, and the initial length of the cracks prior to
Figure 1. Schematic plan view diagram illustrating the measured parameters during an indentation crack
extension experiment: the crack surface trace length, c; and the applied stress, sa .
application of any stress, is determined by the indentation load P used to generate the
flaw. The hardness, H, of the material is related to the load and the area of the
contact impression, A, by H ¼ P=A.
The driving forces for fracture during an extension experiment, in the context of
Eq. (1), are illustrated in Fig. 2. The residual stress-intensity factor, Kr , is modeled as
a point-source loading at the center of a circular crack, giving rise to a 1=c3=2 crack
length dependence. The loading arises from the reaction of the surrounding elastic
matrix to the strain mismatch imposed by the residual plastic contact impression.
The amplitude of the residual term is proportional the indentation load, modified by
an elastic-plastic expanding cavity term, x ¼ j(E=H):1=2:6
Kr ¼ xP=c3=2 (2)
The real geometry (finite extent of deformation zone, semi-elliptical crack on a free
surface and the presence of other crack systems) of the modeled crack system is
accounted for by the material-invariant dimensionless term j. The applied stress-
intensity factor, Ka , is modeled by a stress acting uniformly over the crack with a c1=2
dependence and a dimensionless term c accounting for crack-shape and free-surface
effects:
Ka ¼ csa c1=2 (3)
Ideal crack exentension behavior is obtained by combining the stress-intensity

factors of Eqs. 2 and 3 into the equilibrium equation condition of Eq. 1 under the
constraint of an invariant toughness, T ¼ Tbase þ Tm , and inverting to obtain
sa (c) ¼ (T xP=c3=2 )=cc1=2 (4)
Equation 4 can be solved for two equilibrium conditions: an initial stable equilibrium
configuration at zero stress,
cinit ¼ ( xP=T)2=3 , sa ¼ 0 (5)
Figure 2. Schematic cross section illustrating the crack driving forces acting during indentation crack
extension: the localized loading of the residual stress intensity factor, Kr , and the uniform loading of the
applied stress intensity factor, Ka . The semi elliptical surface crack is modeled as a circular crack in an
infinite body.
60 Robert F. Cook
Table 1. Indentation Crack Extension Behavior of Ceramic Materials.
Suppressed Enhanced Erratic Asymmetric

Crack Crack Crack Crack
Material Extension Extension Advance Advance
Cordierite Glass Ceramic

Polycrystalline Al2 O3 (small grain size)7
Si3 N84 (wrong shape)
Soda Lime Glass1 X
Tempered Glasses9 X
Tetragonal Zirconia Polycrystal10 X X
Polycrystalline Al2 O3 (large grain size) (wrong shape) X X X
Whiser Reinforced Glass Ceramic X X X
Al2 O3 =Al2 TiO75 X X X
and a final stable-unstable transition at the maximum sustainable stress,1,2
cm ¼ (4xP=T)2=3 , sm ¼ 4T=3cc1=2
m (6)
Provided the crack propagation is carried out under near equilibrium conditions, the
crack system follows the sa (c) trajectory given by Eq. 4 from the stable configuration
of Eq. 5, (cinit , 0), to the onset of instability characterized by Eq. 6, (cm , sm ).
Provided, also, that the amplitude terms, x and c, are invariant during propagation,
the extension ratio, cm =cinit ¼ 42=3 2:52. Under these conditions, an alternative
formulation of Eq. 4.
P=c3=2 ¼ (T=x)[1 (c=T)sa c1=2 ] (7)
allows the geometry terms to be evaluated easily from crack extension data, as a plot
of P=c3=2 vs sa c1=2 follows a straight-line trajectory of constant stress-intensity factor
(¼ toughness) with slope 1 and intercepts T=x and T=c.
In many materials, indentation crack extension behavior does not follow the
trajectory of Eqs. 4 or 7, and does not exhibit an extension ratio of 2.52. Table 1
summarizes the crack extension behavior of a number of ceramics. Three classes of
behavior are evident: (i) those materials that exhibit the ideal extension ratio and
trajectory; (ii) those that exhibit a suppressed extension ratio, < 2:52, often associated
with a non-ideal trajectory; and (iii) those that exhibit an enhanced extension ratio,
> 2:52, usually associated with erratic or asymmetric crack advance. It is the latter
group that is the focus here, as their crack extension characteristics are a reflection of
the effects of microstructure on toughness. That is, T in Eqs. 4-7 is not invariant.
Measurement of the sa (c) trajectory can then be used to discern the Tm (c) variation for
a material and its point-to-point variability within a microstructure.
3.1. Materials
Two materials are considered here to illustrate the effects of microstructure on crack
extension. The first is a cordierite (2MgO:2Al2 O3 :5SiO2 ) based glass-ceramic contain-
ing silicon nitride whiskers. The whiskers were approximately 1 15 mm and were
mixed in with the base glass-ceramic powder prior to tape-casting at 0, 5, 10 and 15
vol%. The whiskers significantly retarded densification during sintering, such that the
modulus and hardness of the composite materials were reduced; E and H were 120 and
8.3 GPa and 104 GPa and 3.1 GPa for the base material and 15 vol% composite
respectively. The second material is a large-grained polycystalline alumina containing
0.1 mol% MgO.2 The mixed powder was dry-pressed and then sintered in H2 at 1800 C
for 12 hours to reach > 99:5% relative density. The mode grain size was 15 mm.
3.2. Testing Method
Test specimens were cut into bar form, approximately 45 mm 6 mm 3 mm. Single
Vickers indentations, with one set of cracks aligned perpendicular to the bar axis,
were placed in the center of the largest face of the bar. In the case of the cordierite
composites, a single 100 N indentation load was used. For the alumina, loads in the
range 10 100 N were used. The bars were then placed in a custom-built testing
apparatus, illustrated in Fig. 3. The apparatus generated a tensile stress in four-
point bending on the indentation surface of the bar. The stress was maintained by
displacement control through a stepping motor and the load monitored by a resist-
ance load cell. The entire apparatus was mounted on an inverted optical microscope
so that the indentation crack lengths could be monitored during loading. All inden-
tation and testing was performed in air. During the crack extension on loading, the
load was incremented and the cracks allowed to come to rest (velocity less than
3 10 10 ms 1 ) before measuring the crack length and then re-incrementing the load.
The crack extension responses were thus characteristic of equilibrium fracture in the
air environment.
4. RESULTS
Figure 4 shows applied stress-crack length trajectories for the base cordierite glass-
ceramic material. Measurements on two opposing cracks on two specimens were
made; the trajectories are symmetric and relatively smooth for each specimen and the
entire set of data are well described by the (penny-crack based) ideal extension
behavior of Eq. 4, shown as the best fit solid line and associated instability crack
length and stress values. In this case it appears that the assumptions of extension
under conditions of quasi-equilibrium propagation and invariant material toughness
and crack geometry are fulfilled. Figure 5 replots the data of Fig. 4 according to the
scheme of Eq. 7, emphasizing the shift from residual stress-intensity dominance at the
beginning of the test to applied stress-intensity dominance at failure. A straight line
fit in these coordinates allows the combined toughness/geometry parameter combin-
ations to be obtained from intercepts as indicated on the Figure. These parameters
may be used to evaluate the toughness characteristics of similar materials, referenced
to this base material. Combining the full equilibrium expression of Eq. 1 with the
crack extension expression of Eq. 7 gives
T=Tbase ¼ (c=cbase )(cbase =Tbase )sa c1=2 þ (x=xbase )(x base =Tbase )P=c3=2 (8)
62 Robert F. Cook
Figure 3. Diagram of the apparatus used to measure indentation crack extension under applied stress.
which indicates that once a base material has been calibrated, the toughness vari-
ations with crack length of related materials can be determined from crack extension
measurements, once the applied and residual stress geometry transformations are
known.
Figure 6 shows the relative toughness behavior of the set of cordierite whisker
composites as a function of crack length, determined using crack extension data such
as those shown in Fig. 4 and transformed as given by Eq. 8. The applied geometry
term was taken as invariant with whisker volume fraction, c=cbase ¼ 1, and the
residual geometry term scaled with the elastic-plastic response, x=x base ¼ (E=H)1=2 =
(Ebase =Hbase )1=2 . As anticipated by Figs. 4 and 5, the base material response is invariant
with crack length. The responses of the whisker composites are also relatively invariant,
but clearly increase in their near steady-state toughness values with whisker volume
fraction. In addition the 15 vol% material exhibits a clear enhancement of crack
extension (to almost a factor of five) over that of the base material. The implication
Figure 4. Plot of the equilibrium indentation applied stress crack length, sa c, extension behavior for
cordierite glass ceramic in air. The data are well described by an ideal indentation response based on the
penny crack model of Fig. 2.
Figure 5. Plot of the equilibrium indentation crack extension behavior for cordierite glass ceramic in air
using parameters related to the residual and applied stress intensity factors. A straight line best fit to the
data allows the crack geometry terms to be evaluated from the intercepts.
of these data is that although the addition of the whiskers to the cordierite glass ceramic
increased the porosity and thus degraded the modulus and hardness (and presumably
the base toughness), the actions of the whiskers during crack propagation more than
compensated to lead to overall tougher materials. The crack length scale associated
with the attainment of steady-state toughness characteristics appears from Fig. 6 to be
approximately 200---300 mm.
64 Robert F. Cook
Figure 6. Plot of relative toughness as a function of crack length for cordierite glass ceramic containing
silicon nitride whiskers at the volume fractions indicated. Adding whiskers enhances the toughness over the
base material by over a factor of two, approximately 1 MP m1=2 for this material in air.
Figure 7 shows the applied stress-crack length trajectories for the alumina
material; measurements on two opposing cracks on a single specimen were made.
The trajectories exhibit behavior very different from that of the base glass-ceramic of
Fig. 4, as indicated in Table 1. For a given indentation, the crack traces are not of the
same length and the cracks do not extend similarly, leading to extremely asymmtric
crack extension. On increasing stress, the cracks exhibited erratic ‘‘stick- slip’’ exten-
Figure 7. Plot of the equilibrium indentation applied stress crack length, sa c, extension behavior for a
large grained polycrystalliune alumina in air. The data are not well described by an ideal indentation
response, exhibiting asymmetry, erratic and variable extension, and very small strength degradation over
the flaw size range, indicating a significant interaction between the cracks and the microstructure.
sion, sequences of fixed length for a large stress increase followed by an abrupt
increase in length for a small stress increment. This behavior was more prevalent
for small indentation flaws. Taken as a group, the indentations exhibited variable
crack extension ratios, from zero to approximately eight. And finally, the material
exhibited extreme flaw tolerance; over the range of a factor of 20 in indentation load,
the strength degraded by a factor of about 0.3, compared with a factor of 2.7
predicted by Eq. 6.
Figure 8 shows the relative toughness behavior of the alumina as a function of
crack length, determined using the crack extension data of Fig. 7 and transformed by
Eq. 8. The applied and residual geometry terms were taken from the experiments of
Braun et al.7 on an alumina of similar density and composition. This alumina had
much smaller grain size and exhibited an invariant toughness response, taken here as
representative of the base toughness of the large grain alumina, such that the values
of Tbase =c ¼ 3:68, and Tbase =x ¼ 36:0 determined for the small-grained material
were deemed applicable for the large-grained material. As both materials were
close to fully dense, the applied and residual geometry terms were assumed to be
unaltered by grain size, c=cbase ¼ 1 and x=xbase ¼ 1. As perhaps anticipated by Fig. 7,
the toughness response passes through range of values on crack extension and the
stick-slip crack extension is evident as increase-decrease sequences in toughness with
crack length. Overall, the toughness behavior of the alumina can be characterized as
lying within a wedge or triangle pointing to longer crack lengths, bounded by an
invariant upper value and with lower values significantly less than that of the base
material.
The data of Fig. 8 are very different from those of Fig. 6 and suggest a much
more variable interaction between an extending crack and the microstructure over
the range of crack extensions examined. The cordierite composite data suggest an
Figure 8. Plot of relative toughness as a function of crack length for a large grain size polycrystalline
alumina There is considerable variability in the toughness response, although it appears that the toughness
is increased by about a factor of 1.5 over fine grained material, approximately 1.35 MPa m1=2 for this
material in air.
66 Robert F. Cook
almost saturated toughening response with little variability, indicative of the influ-
ence of many microstructural interactions (whisker-related frictional interlocks)
accumulated over the crack face. The alumina data on the other hand suggest a
still-accumulating toughening response, with great variability, indicative of the point-
to-point stochastic nature of relatively few microstructural interactions (grain-related
frictional interlocks) over the crack face. An interpretation of the data is that the
toughness of the least tough interface in the alumina is about 1.5 MPa m1=2 , reflect-
ing grain boundary fracture. Most of the indentation cracks start out this way and
are arrested by grains in the crack path. As the applied stress is increased, the net
stress-intensity factor (and thus apparent toughness) increase at the arrested crack
length until the instability condition, about 4 MPa m1=2 , reflecting fracture of
the grains or obstacles, is exceeded and the crack extends until arrested again. The
sequence then repeats, but as the crack has extended there are now some toughening
interactions over the crack plane and the starting point is thus at a greater tough-
ness value than the previous sequence; the ending point is at about the same
toughness value, reflecting the maximum toughness achievable in the microstructure.
On continued sequential extension in this way, the value of the toughness at the
starting (arrest, stick) part of the sequence approaches the value of the ending
(extension instability, slip) part and the variability in the toughness with extension
decreases and settles on the upper-bound level associated with homogeneous, steady-
state microstructural interactions over the crack plane. Long cracks associated with
large indentation loads approach this behavior, as seen in Figs. 7 and 8.
5. CONCLUSIONS
Crack extension from indentation flaws under the influence of an applied stress
exhibits a great sensitivity to microstructural effects. This sensitivity is not observed
in inentation-strength tests or long crack toughness tests, as the crack extension
during failure or creation of the long crack, respectively, accumulate large numbers
of microstructural interactions that act to average or saturate the perceived tough-
ness value. Direct observation of indentation crack extension can thus be used to
ascertain the probable toughening mechanisms in a microstructure, the length scales
associated with toughening and the point-to- point variability in these effects, as well
as providing an easy method of evaluating material toughness with great specimen
economy.
The results presented here on cordierite composites and polycystalline alumina
typify the enhanced extension ratios (>2.52 predicted for an ideal indentation re-
sponse) and are reasonably interpreted within the established framework of the
cumulative toughening actions of obstacle (whisker or grain)- related crack face
frictional interlocking. Such toughening relies on the natural inhomogeneity of the
microstructure; an implication here is that significant variability of defect responses
to applied stress is inherent in such materials. However, explanations remain to be
found for the non-ideal extension trajectories and suppressed extensions exhibited by
many of the materials in Table 1 that have no such inhomogeneities (glass, Si3 N4 ,
TZP). Such non-idealities point to the need for examination of the crack driving force
K terms in Eq. 1 associated with the contact mechanics, rather than the T terms
associated with the toughening mechanics.
REFERENCES
1. D.B. Marshall, B.R. Lawn and P. Chantikul, Residual stress effects in sharp contact cracking. II.
Strength degradation., J. Mat. Sci., 14, 2225 35 (1979).
2. R.F. Cook, E.G. Liniger, R.W. Steinbrech, and F. Deuerler, Sigmoidal Indentation Strength Char
acteristics of Polycrystalline Alumina, J. Amer. Cer. Soc. 77 303 314 (1994).
3. P. L. Swanson, C. J. Fairbanks, B. R. Lawn, Y. W. Mai and B. J. Hockey, Crack interface grain
bridging as a fracture resistance mechanism in ceramics. I. Experimental study on alumina, J. Amer.
Cer. Soc., 70 279 89 (1987).
4. R.M. McMeeking and A. G. Evans, Mechanics of transformation toughening in brittle materials,
J. Amer. Cer. Soc., 65, 242 6 (1982).
5. R. Tandon, D.J. Green and R.F. Cook, Strength Variability in Brittle Materials with Stabilizing and
Destabilizing Resistance Fields, Acta metall. 41 399 408 (1993).
6. B.R. Lawn, A.G. Evans, and D.B. Marshall, Elastic/Plastic Indentation Damage in Ceramics: The
Median/Radial Crack System J. Amer. Cer. Soc., 63 574 581 (1980).
7. L. M. Braun, S. J. Bennison and B.R. Lawn, Objective evaluation of short crack toughness curves
using indentation flaws: case study on alumina based ceramics, J. Amer. Cer. Soc., 75, 3049 57
(1992).
8. D. B. Marshall, Controlled flaws in ceramics: a comparison of Knoop and Vickers indentation,
J. Amer. Cer. Soc., 66, 127 31 (1983).
9. R. Tandon and R.F. Cook, Indentation Crack Initiation and Propagation in Tempered Glass,
J. Amer. Cer. Soc., 76 885 889 (1993).
10. L. M. Braun and R.F. Cook, in Science and Technology of Zirconia V, edited by M.J. Bannister and
R.H.J. Hannink (Technomic Press, 1993), pp. 386 400.
MODE I, MODE II, AND
MIXED-MODE FRACTURE OF
PLASMA-SPRAYED
THERMAL BARRIER
COATINGS AT AMBIENT AND
ELEVATED TEMPERATURES
Sung R. Choi, Dongming Zhu, and Robert A. Miller*
The mixed-mode fracture behavior of plasma-sprayed ZrO2 -8 wt% Y2 O3 thermal

barrier coatings was determined in air at 25 and 13168C in asymmetric four-point
flexure with single edge v-notched beam (SEVNB) test specimens. The mode p I
fracture toughness was found to be KIc ¼ 1:15 0:07 and 0:98 0:13 MPa m,
respectively, at 25 and 13168C. The respectivep mode II fracture toughness values
were KIIc ¼ 0:73 0:10 and 0:65 0:04 MPa m. Hence, there was an insignificant
difference in either KIc or KIIc between 25 and 13168C for the coating material,
whereas there was a noticeable distinction between KIc and KIIc , resulting in
KIIc =KIc ¼ 0:65 at both temperatures. The empirical mixed-mode fracture criterion
best described the coatings’ mixed-mode fracture behavior among the four mixed-
mode fracture theories considered. The angle of crack propagation was in reasonable
agreement with the minimum strain energy density criterion. The effect of the
directionality of the coating material in on KIc was observed to be insignificant,
while its sintering effect at 13168C on KIc was significant.
*
Sung R. Choi, Ohio Aerospace Institute, Brook Park, Ohio 44142. Dongming Zhu, U.S. Army
Research Laboratory, Glenn Research Center, Cleveland, Ohio 44135. Robert A. Miller, NASA Glenn
Research Center, Cleveland, Ohio 44135.
451
452 Sung R. Choi et al.
1. INTRODUCTION
Thermal barrier coatings (TBCs) have attracted increasing attention for advanced gas
turbine and diesel engine applications due to their ability to provide thermal insulation
to engine components.1–3 The merits of using the ceramic thermal barrier coatings are
well recognized and include the potential increase in engine operating temperature
with reduced cooling requirements, resulting in significant improvement in thermal
efficiency, performance, and reliability. Plasma-sprayed zirconia-based ceramics are
one of the most important coating materials in light of their low thermal conductivity,
relatively high thermal expansivity, and unique microstructure as a result of the
plasma spraying process. However, the limited durability of thermal barrier coatings
under severe thermal and mechanical loading conditions encountered in heat engines
remains one of the major problems. As a result, the development of thermal barrier
coatings requires a better understanding of both thermal and mechanical behavior of
the coating materials to ensure life and reliability of the related components.
It has been suggested that the important limiting factor encountered in thin
plasma-sprayed thermal barrier coatings is the relatively low fracture energy of the
coating in planes close to and parallel to the interface. Various efforts have been
made to determine mode I ‘interfacial’ fracture toughness of the coatings in the
vicinity of the interface using various techniques such as the indentation method,
debonding technique, three- or four-point flexure delamination technique, and com-
pact tension test method.4 –6 Failure of the thick thermal barrier coatings has been
observed within the bulk of the coating material, independent of any delamination
that typically occurs at the interface of a thin coating and a substrate.7 There have
been efforts to determine mode I fracture toughness of thick thermal barrier coatings
at ambient and elevated temperatures. In fact, the majority of data on fracture
toughness of both thin and thick coating materials have been determined under
mode I loading. However, rarely are structural components or coatings subject to
pure mode I loading. This is particularly true for thermal barrier coatings that
encounter complex thermal and mechanical loading in engine operations. Recently,
Callus and Berndt8 used a pure-shear technique to determine the interfacial critical
mode II strain energy release rate of some thin coatings. Their data, however, were
only for ambient temperature. Mixed-mode data on either thin or thick coatings are
rarely available in the literature at elevated temperatures, despite an important fact
that mechanical behavior of coating materials should not be based solely on ambient-
temperature properties. Dense oxide ceramics (such as zirconia) have exhibited
significant changes in mechanical properties such as strength, fracture toughness,
and slow crack growth when temperature is increased above 800 to 10008C.9
The objective of this work was to determine modes I and II fracture toughness
and mixed-mode fracture behavior of free-standing thermal barrier coatings of
plasma-sprayed ZrO2 -8 wt% Y2 O3 at an elevated temperature of 13168C in air. The
choice of this temperature was based on a typical target temperature of aerospace gas
turbine applications. These same properties were also determined at ambient tem-
perature (258C). An asymmetric four-point flexure test technique was used at both
temperatures in conjunction with single-edge-v-notched beam (SEVNB) test speci-
mens, which yielded simplicity in both specimen and crack preparation and in test
procedure. Fracture locus, ranging from mode I to mode II, was determined. Several
mixed-mode fracture criteria were analyzed based on experimental data on mode I
Mode I, Mode II and Mixed-Mode Fracture of Plasma-Sprayed 453
and mode II stress intensity factors as well as on crack propagation angles. The
mixed-mode fracture behavior of the coatings was compared with that of typical
advanced structural ceramics in terms of mixed-mode fracture theories. The effects of
sintering and the material’s directionality on mode I fracture toughness at ambient
temperature were also characterized.
2.1. Material
The ZrO2 -8 wt% Y2 O3 powder with an average particle size of 60 mm, was first
plasma-sprayed on a graphite substrate measuring 150 by 100 by 6.5 mm to a
thickness of about 6 mm, using a Sulzer-Metco ATC-1 plasma coating system with
an industrial robot. The plasma-spray conditions can be found elsewhere.10 A free
standing, plasma-sprayed ceramic billet was then obtained by burning away the
graphite substrate at 6808C in air for 24 h. The billet was machined into the final,
rectangular flexure test specimen with nominal dimensions of 3 by 4 by 50 mm,
respectively, in width, depth, and length. The 3-mm-wide face of flexure test speci-
mens corresponded to the plane perpendicular to the plasma spraying direction.
Major physical and mechanical properties of as-sprayed coating material including
hardness, fracture toughness, and strength have been determined previously at
ambient temperature11,12 and are presented in Table 1. Figure 1 shows a typical
fracture surface and a polished surface showing the microstructure of as-processed
coatings, in which large amounts of microcracks and pores are characterized in
conjunction with a unique platelet structure.
2.2. Mixed-Mode Fracture Testing

2.2.1. Preparation of Sharp Precracks
Sharp v notches were introduced in flexure test specimens, using the single-edge-v-
notched-beam method.13 This method utilizes a razor blade with diamond paste to
Table 1. Typical Physical and Mechanical Properties of Plasma Sprayed Zro2 8 Wt% Y2 O3
Thermal Barrier Coatings at Ambient Temperature.11,12
Strength
a
Elastic modulus Fracture
(in compression), toughness,b Number of Average
p
Density, g=cm3 E, GPa KIc , MPa m Type of test specimens strength,b MPa
5.22 34 1.0(0.1) Tension 10 10 (2)

Trans thickness tension 10 11(1)
compression 10 324(72)
Four point flexure 20 32(6)
Biaxial flexure 10 40(4)
a
The value of E is estimated from the starting point of the stress-strain curve.11
b
The number in parenthesis represents the + 1.0 standard deviation.
Figure 1. Typical microstructures of plasma sprayed ZrO2 8 wt% Y2 O3 thermal barrier coatings.
(a) Polished surface. (b) Fracture surface. Arrow indicates plasma spray direction.
introduce a sharp root radius by tapering a saw cut. Sharp v-notch radii ranging from
4 to 6 mm have been successfully obtained for alumina, glass ceramic, silicon nitride,
zirconia, and silicon carbide ceramics.13 A starter straight-through notch 0.6 mm
deep and 0.026 mm wide was made on the 3-mm-wide face of the test specimens.
A steel razor blade was put into the starter notch sprinkled with diamond paste with
a particle size of 9 mm. Typically, a load of about 10 N was applied through the razor
blade with a stroke rate (back-and-forth motion) of 1 Hz and a stroke distance of
about 13 mm in a specially designed polishing machine.11 The final notch depth and
root radius were 2.0 mm and 20 to 50 mm, respectively, resulting in the crack size-to-
specimen depth ratio of 0.5 (¼ a=W ). The coating material resulted in less sharpness
in root radius due to its porous and microcracked nature, compared to typical dense
ceramics. However, it has observed that the sharpness ranging from 20 to 50 mm was
sufficient to give a consistent and accurate value of fracture toughness of the coating
material.11 Note that the through-the-thickness sharp notches thus prepared were
aligned parallel with respect to the plasma spraying direction.
Other methods to generate sharp cracks to estimate fracture toughness, such as
the single-edge-precracked-beam (SEPB)14 and the indentation techniques, were not
feasible for the coating material: The indentation response was poor because of the
material’s significant porosity, microcracks, and ‘softness,’ so that well-defined in-
dentation cracks were not achieved. This indicates that microindentation techniques
may not be appropriate for plasma-sprayed coating materials whose pore and
microcrack sizes in some cases are much greater than microindentation cracks, and
as a result, a continuum approach is no longer valid.11 Also note that cracks
produced by this SEVNB method were very uniform in their crack-front geometry
and very consistent in size with high reproducibility from specimen to specimen. This
gives rise to very little scatter in or a small coefficient of variation in the values of
fracture toughness, a notable advantage of the SEVNB method over other methods.
2.2.2. Asymmetric Four-Point Flexure Testing
The sharp v-notched flexure test specimen was loaded in asymmetric four-point
flexure as shown in figure 2. The ratio of mode I to mode II stress intensity factor
Figure 2. Schematic of asymmetric four point flexure geometry with accompanying shear force and
bending moment diagrams.
(SIF), KI =KII , was varied by varying the distance of the precrack from the center
plane, s, as shown in the figure. When the precrack was centered with respect to the
loading point (s ¼ 0), the precrack was subjected to pure mode II loading. As s was
increased, KI =KII also increased. The stress intensity factors, KI and KII , are ex-
pressed as
a
KI ¼ s(pa)1=2 FI (1)
W
a
KII ¼ t(pa)1=2 FII (2)
W
where s is the applied (remote) normal stress, t is the applied (remote) shear stress, a
is the crack size, and W is the specimen depth. FI and FII are crack geometry factors
in modes I and II, respectively. The normal and shear stresses are given from the
elementary beam theory by

A B 6sP
s¼ (3)
A þ B bW 2

AB P
t¼ (4)
A þ B bW
where A and B are distances from a load point as defined in figure 2, b is the specimen
width, and P is the applied force. The values of A ¼ 10 mm and B ¼ 5 mm were
typically used, but in some cases (KIIc at RT) values of A ¼ 12 and B ¼ 6 mm were
used for comparison and verification. The test specimen’s cross section was
W ¼ 4 mm and b ¼ 3 mm.
Several different expressions of FI and FII were suggested by Suresh et al.,15

Wang et al.,16 and He and Hutchinson17 for the case of through-the-thickness crack,
and their respective results of FI and FII are illustrated in figure 3. As seen from the
figure, no significant difference in FI and FII between the solutions was found,
particularly when a/W is between 0.35 and 0.50. The solution by He and Hutchin-
son17 that provided a convenient polynomial expression was used in here, and is
given as follows:
a a a 2 a 3 a 4
FI ¼ 1:122 1:121 þ 3:740 þ3:873 19:05 þ
W W W W W
a 5 a
(5)
22:55 for #0:7
W W
a a a 2
FII ¼ 7:264 9:37 þ 2:74 þ1:87
W W W
a 3 a 4 a
(6)
1:04 for 0# #1
W W W
Note that Eq. (5) was quoted from Murakami’s18 and was almost identical to Srawly
and Gross’ solution19 up to a=W ¼ 0:7, as seen in the figure. Also note that the
difference in FI and FII between solutions in the range of a=W ¼ 0:4 to 0.5 was
negligible, resulting in
FI FII (7)
Figure 3. Various solutions of crack geometry factors FI and FII applied in the asymmetric four point
flexure configuration.
Hence, in the case of a=W ¼ 0:4 to 05, the KII =KI ratio (or called ‘mixity’) can be
simplified from Eqs. (1) (4) and (7) as follows:
KII WFII W
¼ (8)
KI 6sFI 6s

KII W
b ¼ tan 1 tan 1 (9)
KI 6s
where b is the mixity angle. The angle b was also called the ‘‘equivalent crack angle’’
by Maccagno and Knott20 for situations which do not actually use the inclined crack
configuration. The ratio of crack size to specimen width used in this work was chosen
to be a=W ¼ 0:5 so that Eq. (8) or (9) could be conveniently used. The values of s ¼ 0
to 3.6 mm were selected to give uniformly distributed b between pure mode I(b ¼ 0 )
and pure mode II(b ¼ 90 ) based on Eq. (9). Pure mode I fracture toughness KIc was
determined using a symmetric four-point flexure fixture with 10-mm inner and
20-mm outer distances.
All testing was performed in displacement control using a SiC flexure fixture
with an actuator speed of 0.5 mm/min of an Instron electromechanical test frame
(Model 8562, Instron, Canton, MA). Testing was conducted at 25 and 13168C in air.
Each test specimen in elevated-temperature testing was held for thermal equilibration
for about 20 min prior to testing. At room temperature, four and nine specimens
were tested for KIc and KIIc , respectively, whereas at 13168C, four specimens were
examined for each test. There were nine mixity values associated with the specimens
tested at 258C, and four mixity values associated with those tested at 13168C.
Typically, two or three specimens of each mixity value were tested. After testing,
the crack size of each tested specimen was determined optically from its fracture
surface based on the three-point measurements in accordance with test standard
ASTM C1421.14 The crack propagation angles were also determined.
2.3. Effects of Sintering and Material Directionality
In order to determine the effect of sintering on mode I fracture toughness of the

coating material, additional fracture toughness testing was carried out at ambient
temperature using the test specimens annealed at 13168C in air for 5, 20, 100, and
500 h. The sharp v-notches were introduced in the annealed specimens using the
Figure 4. (a) A schematic geometry of double cantilever beam (DCB) test specimens used to determine
KIc in the plane perpendicular to the plasma sprayed (PS) direction. (b) ‘Regular’ fracture toughness
testing by the SEVNB method with crack planes parallel to the plasma spraying direction is shown for
comparison.
same technique described in section 2.2.1. The test fixture, test frame, and test
procedure used were the same as those used in the aforementioned ‘regular’ fracture
toughness testing. Fracture surfaces were also examined via SEM with respect to
annealing time to see any change in microstructure and morphology.
Mode I fracture toughness (KIc ) of as-sprayed coating material was also evalu-
ated at ambient temperature in the plane perpendicular to the plasma-sprayed
direction (that is, within the ‘layered’ direction) using the double cantilever beam
(DCB) method. A sharp v notch was introduced in the center plane of the 4-mm-wide
face of each flexure test specimen with a crack length of about 6 mm (see fig. 4).
Fracture toughness was calculated based on the formula by Murakami.18 A total of
three DCB specimens were tested. This additional fracture testing was to determine
directionality of the material in response to fracture toughness relative to the plasma-
spraying direction.
3.1. Test Results
The results of the mode I, mode II, and mixed-mode fracture testing for the coating
material at both 25 and 13168C are presented as KII versus KI in figure 5. Each point
represents a single datum, and the values KIc and KIIc represent the average of four
and eight measurements, respectively, at 258C and the average of four measurements
for each at 13168C. The effect of test distance onpKIIc at 258C was found to be
insignificant: KIIc ¼ 0:74 0:1 and 0:69 0:09 MPa m, respectively, for A/B values
of 10/5 and 12/6 mm (see fig. 2). Hence, the 10/5-mm distances were exclusively used
Figure 5. A summary of KII versus KI for mode I, mode II, and mixed mode fracture of plasma sprayed
ZrO2 8 wt%Y2 O3 thermal barrier coatings at both 25 and 13168C in air.
throughoutp the mixed-mode test program. At 258C, KIc ¼ 1:15 0:07 and KIIc ¼ p 0:73
0:10 MPa m, and at 13168C, KIc ¼ 0:98 0:13 and KIIc ¼ 0:65 0:04 MPa m.
Note that the values of KIIc were 37 and 34 percent lower than those of KIc at 25 and
13168C, respectively. It has been reported that for some dense ceramics KIIc was
greater than KIc , presumably attributed to the frictional interaction between the two
crack planes.21 However, the previous studies on advanced (dense) ceramics includ-
ing silicon nitrides, alumina and zirconia22, 23 showed a different result that KIIc was
almost identical to KIc for a given material, indicative of an insignificant frictional
effect on KIIc by either coarse grained or fine grained ceramics. The coating material,
however, did not exhibit a similar value in both KIIc and KIc but rather yielded a
lower value in KIIc than in KIc at both 25 and 13168C. It is believed that because of
the porous, microcracked, ‘soft’ nature the coating material, there would not have
been any significant frictional contribution to KIIc . Furthermore, the coating material
would be easier to shear (mode II) than to cleave (mode I), probably because of its
unique platelet structure formed by plasma-spraying process, which results in low
KIIc compared with KIc . The difference between KIIc and KIc for various ceramics and
brittle materials will be described in more detail in a next section.
As shown in figure 5, because KIIc < KIc , the overall KII versus KI relation is
represented by an ellipse at either 25 or 13168C with KIc being a major axis. For a
given mixity angle ranging from b ¼ 0 to 908, KEQ [ ¼ (KI2 þ KII2 )1=2 ] was lower at
13168C than at 258C, probably because of a somewhat softening effect at elevated
temperature. The values of KIc and KIIc were decreased by 15 and 11 percent,
respectively, when temperature was increased from 25 to 13168C. However, this
decrease was marginal, as can be seen in a summary of KIc as a function of
temperature in figure 6, in which the KIc data previously determined11 at 25 and
Figure 6. Fracture toughnesses of KIc and KIIc as a function of temperature for plasma sprayed
ZrO2 8 wt%Y2 O3 thermal barrier coatings. The previous KIc data11 were included. Error bars represent
1:0 standard deviation.
8008C by the SEVNB method were included with the current p data. KIc remains
statistically invariable with an average of KIc ¼ 1:0 0:1 MPa m, regardless p of
test temperature up to 13168C. The same would be true for KIIc ( ¼ 0:69 MPa m)
as depicted in the figure, notwithstanding the lack of intermediate-temperature data.
The insensitivity of KIc to temperature was also observed in a similar but dense (hot-
pressed) 10 mol% yttria-stabilized zirconia (10-YSZ) in which fracture toughness of
the material evaluated
p in air by the same SEVNB method was almost unchanged
(KIc ¼ 1:6-1:8 MPa m) from 25 to 10008C.9
Figure 7 shows typical examples of crack propagation with different mixities
ranging from pure mode II (KI =KII ¼ 0) to pure mode I (KI =KII ¼ 1). The angle of
crack propagation uc was defined as an angle of the crack plane propagated with
respect to the original precrack plane. The direction of crack plane propagated was in
most cases straight with some exceptions such as tortuous or not-straight paths. The
angle of crack propagation was greatest in pure mode II, and decreased with
increasing KI =KII reaching to uc ¼ 0 in pure mode I. A summary of experimental
data on the angle of crack propagation as a function of KI =KII determined at both 25
and 13168C is depicted in figure 8. The overall trend of the angle of crack propaga-
tion was that uc initially decreased quickly with increasing KI =KII and then decreased
monotonically at KI =KII > 3. No significant difference in uc between 25 and 13168C
was observed. It is noted that the scatter in uc , for example, uc ¼ 45 to 708 at
KI =KII ¼ 0 (pure mode II), seemed more appreciable in the coating material than
in dense advanced ceramics,22 which exhibited a typical scatter of about 158.
Analysis of fracture surfaces revealed no explicit difference in the fracture morph-
ology of the pure mode I, mixed-mode, or pure mode II specimens. However, it should
be mentioned that unlike dense ceramics the coating material has provided a challen-
ging subject on fractography because of its unique open and microcracked platelet
Figure 7. Typical examples of specimens showing angles of crack propagation with respect to KI =KII of
plasma sprayed ZrO2 8 wt% Y2 O3 thermal barrier coatings: (a) pure mode II; KI =KII 0, (b)
KI =KII 2:7, (c) KI =KII 5:5, and (d) pure mode I; KI =KII 1.
Figure 8. Crack propagation angles as function of KI =KII of plasma sprayed ZrO2 8 wt% Y2 O3 thermal
barrier coatings determined in asymmetric four point flexure at both 25 and 13168C in air.
microstructure. The origin or nature of strength-controlling flaws in the coating

material has rarely been identified with any traditional means of fractography (optical
and/or SEM) that would be otherwise sufficient for the case of dense ceramics.12 An
improved means of fractography is required to better distinguish fracture morpholo-
gies in response to different modes of fracture of the coating material.
3.2. Consideration of Mixed-Mode Fracture Criteria
3.2.1. Considered Criteria

As shown from the crack propagation angles in figure 8, mixed-mode fracture does
not take place in the same plane as the original crack. This noncoplanar crack
propagation requires different fracture criteria, typically a combination of the driving
forces KI and KII together with KIc and KIIc . The mixed-mode fracture criteria for an
isotropic, homogeneous material can be broadly categorized in four areas and will be
briefly described then compared with experimental results.
3.2.1a. maximum principal stress criterion.24 this criterion assumes that a crack at
instability propagates in the direction normal to the maximum principal stress
direction. the instability condition is given by the following relationship:
uc uc uc
KIc KI cos3 þ 3KII cos2 sin ¼ 0 (10)
2 2 2
The angle of crack propagation uc is a function of KI =KII and is given by
8 " #1=2 9
<K 1 K 2 =
I I
uc ¼ 2 tan 1 þ8 (11)
:4KII 4 KII ;
In case of pure mode II, the following relations can be obtained from Eqs. (10) and
(11):
KIIc
uc ¼ 70:538 , ¼ 0:8662 (12)
KIc
3.2.1b. minimum strain energy density criterion.25 This criterion assumes that
crack propagation would take place in the direction along which the strain energy
density is minimum. this assumption gives rise to the following mixed-mode fracture
criterion
1
KIc2 [a11 KI2 þ 2a12 KI KII þ a22 KII2 ] ¼ 0 (13)
2(k 1)
where
a11 ¼ (1 þ cos uc )(k cos uc )

a12 ¼ sin uc (2 cos uc k þ 1)
a22 ¼ (k þ 1)(1 cos uc ) þ (1 þ cos uc )(3 cos uc 1)
with
k ¼ 3---4n for plain strain

k ¼ (3 n)=(1 þ n) for plain stress
where n is Poisson’s ratio. The angle of crack propagation, uc , is given by

KI KI
(k 1) sin uc 2 tan 1 2 sin 2 uc tan 1 sin 2uc ¼ 0 (14)
KII KII
In case of pure mode II, the following relations can be obtained from Eqs. (13)
and (14):

k1 KIIc 2(k 1) 1=2

uc ¼ cos 1
, ¼ (15)
6 KIc a22
3.2.1c. maximum energy release rate criterion. this criterion assumes that a
crack with an infinitesimally small kink at an arbitrary direction propagates in the
direction along the maximum strain energy release rate.26,29 many investigators keeps
that the strain energy release rate, g, can be expressed in terms of KI and KII as
follows:
G ¼ c11 KI2 þ c12 KI KII þ c22 KII2 (16)
where c11 , c12 , and c22 are coefficients. Hyashi and Nemeat-Nasser28 determined
numerically the coefficients and the direction of maximum G as a function of KI =KII .
Palaniswamy and Knauss27 proposed an approximated mixed-mode fracture cri-
terion on the maximum energy release rate as follows (that was also used by Suresh
et al.15)

KI 3 KII 2
þ ¼1 (17)
KIc 2 KIc
where for pure mode II the approximation leads to
KIIc
uc ¼ 77:48 , ¼ 0:817 (18)
KIc
The fracture criterion and the crack propagation angle on the maximum G were used
based on Eqs. (16) and (17) in this work.
3.2.1d. empirical criterion.30,31 The empirical fracture criterion can be expressed

by the following generalized, simplified form
p
KI KII q
þ ¼1 (19)
KIc gKIc
where p and q are parameters (typically p,q ¼ 1 2) to be determined from experi-

mental data through a curve fitting, and
KIIc
g¼ (20)
KIc
with 0 < g < 2, an experimentally determined value. Despite the lack of its theoret-
ical frame work in some cases, the empirical fracture criterion can still provide a
much simplified representation (as database, too) of mixed-mode behavior of a
material particularly when other (preceding) fracture criteria do not describe reason-
ably well the actual material’s behavior under mixed mode. The maximum G criter-
ion (Eq. 15) or the coplanar crack propagation criterion,32 in fact, is one particular
type of the empirical fracture criterion in its form.
3.2.2. Comparison with Experiment
Figure 9 compares the afore-reviewed four different mixed-mode fracture criteria

with the experimental data (that were presented in fig. 5), where KII normalized with
respect to KIc , KII =KIc , was plotted as a function of KI =KIc for both 25 and 13168C
data. The prediction for the minimum strain energy density criterion was made using a
value of n ¼ 0:2. Neither the maximum principal stress nor the minimum strain energy
density nor the maximum energy release rate criterion was in reasonable agreement
with the experimental data. Only the empirical criterion was in good agreement with
both ambient and elevated-temperature data, resulting in the following parameters:
p ¼ 1;q ¼ 2;g ¼ 0:65 for 258C

(21)
p ¼ 1;q ¼ 1:3;g ¼ 0:65 for 13168 C
Previous studies on dense ceramics such as silicon nitrides and alumina22 and ceria-
doped ziconia23 showed that the minimum strain energy density criterion resulted
in best agreement among the mixed-mode criteria mentioned above, as that was the
case in another study of alumina by Suresh et al. as well.15 Hence, there exists a
Figure 9. Plots of KII =KIc as a function of KI =KIc for mixed mode fracture criteria and experimental data
determined for plasma sprayed ZrO2 8 wt% Y2 O3 thermal barrier coatings. n is Poisson’s ratio.
remarkable contrast in mixed-mode behavior between the dense and the porous TBC
ceramics. The unique nature of the coating material possessing porosity, micro-
cracks, and platelet microstructure might have been responsible for the ability to
distinguish its unique mixed-mode fracture pattern, as this is not the case for dense
ceramics. As also seen from figure 9, the three mixed-mode fracture criteria the
maximum principal stress, minimum strain energy density, and maximum energy
release rate criteria predict the values of KIIc =KIc ¼ 0:8 to 1.1. The poor agreement
with these three fracture criteria was, therefore, due to the lower values of KIIc =KIc
that were around 0.65 for the coating material at both temperatures. Hence, the value
of KIIc =KIc is the parameter with the largest influence on the degree of agreement
with the prediction of the fracture pattern. In case of KIIc =KIc 1, statistically
reasonable agreement would be found in any of the major mixed-mode fracture
criteria mentioned. The dense ceramics examined in previous studies22,23 all exhibited
a value close to KIIc =KIc 1. Comparison of KIIc =KIc among other brittle materials
will be presented and discussed below.
The data on crack propagation angle shown in figure 8 were compared with the
predictions made with the three mixed-mode fracture criteria, and the results are
presented in figure 10. The minimum strain energy density criterion overall seemed to
yield a better prediction compared with the other two criteria. However, because of
somewhat significant scatter in uc as well as little difference in prediction between the
criteria, it is difficult to state which criterion gives the best agreement with the
experimental data. Note that a significant discrepancy was found particularly in
the region close to pure mode II, e.g., KI =KII < 1. This discrepancy predominant
in the region of KI =KII < 1 was also observed in the previous studies with silicon
nitrides and alumina22 and other grade alumina by Suresh et al.15 By contrast, this
discrepancy seemingly unique in ceramics was not found in amorphous glassy
polymer PMMA (polymethylmetha-crylate) mixed-mode tested in either asymmetric
Figure 10. Plots of crack propagation angle as a function of KI =KII for mixed mode fracture criteria and
experimental data determined for plasma sprayed ZrO2 8 wt% Y2 O3 thermal barrier coatings.
four-point flexure23 or uniaxial tension25 configuration. A good agreement of propa-

gation angle with the fracture criteria (as well as little scatter in uc ) in a wide range of
KI =KII close to pure mode II was also found in steel tested in asymmetric four-point
flexure at 1968C to induce brittle fracture.33
As previously mentioned, the overall locus of mixed-mode fracture (in KII versus
KI , or KII =KIc versus KI =KIc relation) is mainly controlled by the value of KIIc =KIc ;
hence, the ratio of mode II fracture toughness to mode I fracture toughness is a very
important measure of a material in terms of its response to mixed-mode fracture. With
this in mind, a summary of data on KIIc =KIc for various brittle materials compiled by
Munz and Fett31 as well as data on dense ceramics from previous studies22,23 was used
and compared with the TBC data obtained in this work, as shown in figure 11. The
values of KIIc =KIc for the dense ceramics two silicon nitrides (one with fine and one,
elongated grains), coarse grained alumina,22 and ceria-doped tetragonal zirconia
polycrystal (Ce-TZP),23 all tested in asymmetric four-point flexure using naturally
sharp precracks, are invariably around KIIc =KIc ¼ 1 and are much greater than those
of the TBC (KIIc =KIc 0:65). Except for this comparison, a large variation of KIIc =KIc
varying from 0.6 to 2.0 is noted depending on materials, test specimens, notch or initial
crack preparations, and even investigators, etc. As a result of this significant variation,
a reasonably unified trend on the value of KIIc =KIc cannot be made for the materials
provided. However, it should be noted that the value of KIIc =KIc for the TBC evaluated
in this work is still at the lower end of the data pool. One thing to note is that test
technique must be technically sound, particularly in the areas of specimen configur-
ation and precrack preparation. A diametral compression technique, for example,
would be in danger to overestimate greater KIIc as a result of frictional constraint
between the two crack faces by the existence of compressive stress therein.22 Use of
sharp precracks using an appropriate method such as single edge precrack beam
(SEPB) method14 or SEVNB method as used in this work is also crucially important
to obtain as accurate fracture toughness as possible.
Figure 11. Comparisons of KIIc =KIc of plasma sprayed ZrO2 8 wt%Y2 O3 thermal barrier coatings with
other advanced monolithic ceramics and brittle materials at ambient temperature. (a) Comparison with
previous studies22,23. (b) Comparison with data compiled by Munz and Fett31.
3.3. Effect of Directionality

To determine a possible effect of the material’s directionality on fracture toughness,
mode I fracture toughness testing for the coating material was performed by the DCB
method in the direction perpendicular to the plasma-sprayed direction, as mentioned
in the Experimental section.
p The value of fracture toughness was found to be
KIc ¼ 1:04 0:05 MPapm. This value of KIc was very similar to that of
KIc ¼ 1:15 0:07 MPa m determined in the direction parallel to plasma-spraying
direction. Hence, the directionality of the coating material in response to mode I
fracture toughness was insignificant. The previous studies showed that even the tensile
strength of the coatings with different vintages was almost independent whether the
plasma-spraying direction which was either perpendicular or parallel. Therefore, a
notion that mechanical properties of TBC would be strongly dependent on the direc-
tion of plasma spraying may not be supported and generalized, based on the fracture
toughness testing in this study and the strength testing in previous studies.12
3.4. Effect of Sintering
Figure 12 shows the results of mode I fracture toughness testing for the coating
material annealed at 13168C in air with annealing times ranging from t ¼ 0 (as-
sprayed) to 500 h. Fracture toughness increased significantly at t ¼ 5 h, increased
monotonically to t ¼p100 h, and then reached a plateau at t ¼ 500 h with a value of
KIc ¼ 2:6 0:2 MPa m, about a 120 percent increase from t ¼ 0. This increase in
KIc was attributed to a sintering effect.6 Figure 13 shows typical fracture surfaces of
specimens tested, subjected to annealing for t ¼ 0 and 100 h. The sintering effect was
manifest by the evidence of significant grain growth for the specimen annealed for
t ¼ 100 h. It has been shown that sintering gives rise to an increase not only in
fracture toughness but in elastic modulus, strength, and thermal conductivity of the
coating material.34,12,6 Although not performed in this work, it is expected that KIIc
in response to sintering would follow a trend similar to KIc . Therefore, evaluation of
KIIc as a function of annealing time at 13168C is an immediate task for future study.
Since a change in mechanical properties (and physical properties as well) would occur
inevitably upon sintering, particularly in a short period of time, a continuous use of
as-sprayed mechanical data in elevated-temperature applications would give an over-
conservative estimate, leading to erroneous results in reliability and/or life predic-
tions of components.
3.5. Other Considerations

Finally, it should be mentioned that the coating material has shown nonlinearity and
hysteresis in its stress-strain relation, irrespective of loading configuration: tension,
uniaxial or biaxial flexure, or compression.12 A typical result showing such nonli-
nearity and hysteresis exhibited in compression by the as-sprayed ZrO2 -Y2 O3 coat-
ings is presented in figure 14. This nonlinearity in constitutive relation was due to the
coatings’ unique microstructure (porosity, microcracks, and platelets) and dimin-
ished upon annealing to a degree depending on annealing time and temperature.
Figure 12. Result of fracture toughness as a function of annealing time for plasma sprayed
ZrO2 8 wt% Y2 O3 thermal barrier coatings annealed at 13168C in air.
Figure 13. Comparison of fracture surfaces of fracture toughness tested specimens of plasma sprayed
ZrO2 8 wt%Y2 O3 thermal barrier coatings. (a) As sprayed. (b) Annealed in air at 13168C for 100 h.
Aquestion arises as to whether linear elastic fracture mechanics (LEFM) approach

could be applicable to this nonlinear, elastic, as-sprayed coating material. The degree
of nonlinearity was more significant when the level of applied stress was increased. In
most cases, however, fracture force employed in fracture toughness testing was
relatively low (with about one tenth of the peak force in figure 14), and in this case
the nonlinearity became negligibly small. As a result, an LEFM approach (a con-
tinum approach as well) would be justified at least in fracture toughness testing for
the as-sprayed coating material.
4. CONCLUSIONS
A full range of mixed-mode fracture behavior covering mode I and mode II for
plasma-sprayed ZrO2 8 wt% Y2 O3 thermal barrier coating was determined in air at
Figure 14. Typical stress strain curve for plasma sprayed ZrO2 8 wt% Y2 O3 thermal barrier coatings in
compression11.
25 and 13168C in asymmetric four-point flexure in conjunction with the single edge
v-notched beam (SEVNB) method. The following conclusions were made:
1. The mode I fracturep toughness was found to be KIc ¼ 1:15 0:07 and
0:98 0:13 MPa m, respectively, at 25 and 13168C. The respective p mode II
fracture toughness was KIIc ¼ 0:73 0:10 and 0:65 0:04 MPa m. Hence, the
coating material exhibited an insignificant difference in either KIc or KIIc between
25 and 13168C, whereas it exhibited noticeable difference between KIc and KIIc ,
resulting in KIIc =KIc ¼ 0:65 at both temperatures.
2. The empirical mixed-mode fracture criterion was in best agreement with the
coating’s mixed-mode fracture behavior among the four mixed-mode fracture
theories considered. The angle of crack propagation was in reasonable agreement
with the minimum strain energy density criterion.
3. The mode I fracture toughness of the coating material at 258C was almost
irrespective of the plane either parallel (typical) or perpendicular to plasma
spraying direction so that the coating’s directionality in response to KIc was
insensitive.
4. Sintering at 13168C in air showed a significant influence on mode I fracture
toughness, giving rise to an increase in KIc at the plateau region by 120 percent
when the coatings were annealed for the duration of 500 h.
REFERENCES
1. R. A. Miller, ‘‘Current Status of Thermal Barrier Coatings An Overview,’’ Surface and Coating
Technology, 30, 1 11 (1987).
2. R. A. Miller, ‘‘Thermal Barrier Coatings for Aircraft Engines History and Direction,’’ pp. 17 34 in
NASA CP 3312 (Ed. W.J. Brindley), National Aeronautics and Space Administration, Glenn Re
search Center, Cleveland, OH (1995).
3. T. M. Yonushonis, ‘‘Thick Thermal Barrier Coatings for Diesel Components,’’ NASA CR 187111,
National Aeronautics and Space Administration, Glenn Research Center, Cleveland, OH (1991).
4. Y. C. Tsui, T. W. Clyne, Proc. 9th Nat. Thermal Spray Conf., Cincinnati, OH (1996).
5. L. L. Shaw, B. Barber, E. H. Jordan, and M. Gell, Scr. Mater., 39, 1427 1434 (1998).
6. G. Thurn, G. A. Schneider, H. A. Bahr, and F. Aldinger, ‘‘Toughness Anisotropy and Behavior of
Plasma Sprayed ZrO2 thermal Barrier Coatings,’’ Surf. Coat. Tech., 123, 147 158 (2000).
7. K. F. Wesling, D. F. Socie, and B. Beardsley, ‘‘Fatigue of Thick Thermal Barrier Coatings,’’ J. Am.
Ceram. Soc., 77[7] 1863 1868 (1994).
8. P. J. Callus and C. C. Berndt, ‘‘Relationship between the Mode II Fracture Toughness and Micro
structure of Thermal Spray Coatings,’’ Surf. Coat. Tech., 114, 114 128 (1999).
9. S. R. Choi and N. P. Bansal, ‘‘Strength and Fracture Toughness of Zirconia/Alumina Composites for
Solid Oxide Fuel Cells,’’ Ceram. Eng. Sci. Proc., 23[3] 741 750 (2002); ‘‘Processing and Mechanical
Properties of Various Zirconia/Alumina Composites for Fuel Cells Applications,’’ NASA/TM
2002 211580, National Aeronautics and Space Administration, Glenn Research Center, Cleveland,
OH (2002); also presented at CIMTEC 2002 Conference, paper no. G1:P03 (to be published in the
proceedings), June 14 18, 2002, Florence, Italy.
10. D M. Zhu and R. A. Miller, ‘‘Influence of High Cycle Thermal Loads on Thermal fatigue Behavior
of Thick Thermal Barrier Coatings,’’ NASA Technical paper 3676 (also in Army Laboratory
Technical Report ARL TR 1341), National Aeronautics and Space Administration, Glenn Research
Center, Cleveland, OH (1997).
11. S. R. Choi, D M. Zhu, and R. A. Miller, ‘‘High Temperature Slow Crack Growth, Fracture
Toughness and Room Temperature Deformation Behavior of Plasma Sprayed ZrO2 8 wt%
Y2 O3 ,’’ Ceram. Eng. Sci. Proc., 19[4] 293 301 (1998).
12. S. R. Choi, D M. Zhu, and R. A. Miller, ‘‘Deformation and Strength Behavior of Plasma Sprayed
ZrO2 8 wt% Y2 O3 Thermal Barrier Coatings in Biaxial Flexure and Trans Thickness Tension,’’
Ceram. Eng. Sci. Proc., 21[4] 653 661 (2000).
13. J. Kübler, (a) ‘‘Fracture Toughness of Ceramics Using the SEVNB Method: Preliminary Results,’’
Ceram. Eng. Sci. Proc., 18[4] 155 162 (1997); (b) ‘‘Fracture Toughness of Ceramics Using the
SEVNB Method; Round Robin,’’ VAMAS Report No. 37, EMPA, Swiss Federal Laboratories for
Materials Testing & Research, Dübendorf, Switzerland (1999).
14. ASTM C 1421 ‘‘Test Method for Determination of Fracture Toughness of Advanced Ceramics at
Ambient Temperature,’’ Annual Book of ASTM Standards, Vol. 15.01, American Society for Testing
and materials, West Conshohocken, PA (2002).
15. S. Suresh, C. F. Shih, A. Morrone, and N. P. O’Dowd, ‘‘Mixed Mode Fracture Toughness of
Ceramic Materials,’’ J. Am. Ceram. Soc., 73[5] 1257 1267 (1990).
16. K. J. Wang, H. C. Lin, and K. Hua, ‘‘Calculation of Stress Intensity Factors for Combined Mode
Bend Specimens,’’ pp. 123 133 in Advances in Research on the Strength and Fracture of Materials,
Vol. 4, Edited by M. D. R. Taplin, ICF4, Waterloo, Canada (1977).
17. M. Y. He and J. W. Hutchinson, ‘‘Asymmetric Four Point Crack Specimen,’’ J. Appl. Mech., 67,
207 209 (2000).
18. Y. Murakami (ed.), Stress Intensity Factors Handbook, Vol. 1, p. 16, Pergamon Press, New York
(1987).
19. J. E. Srawley and B. Gross, ‘‘Side Cracked Plates Subjected to Combined Direct and Bending
Forces,’’ pp. 559 579 in Cracks and Fracture, ASTM STP 601, American Society for Testing and
Materials, Philadelphia (1976).
20. T. M. Maccagno and J. F. Knott, ‘‘The Fracture Behavior of PMMA in Mixed Modes I and II,’’ Eng.
Fract. Mech., 34[1] 65 86 (1989).
21. D. K. Shetty, A. R. Rosenfield, and W. H. Duckworth, ‘‘Mixed Mode Fracture of Ceramics in
Diametral Compression,’’ J. Am. Ceram. Soc., 69[6] 437 443 (1986).
22. V. Tikare and S. R. Choi, ‘‘Combined Mode I and Mode II Fracture of Monolithic Ceramics,’’
J. Am. Ceram. Soc., 76[9] 2265 2272 (1993).
23. V. Tikare and S. R. Choi, ‘‘Combined Mode I Mode II Fracture of 12 mol % Ceria Doped Tetrag
onal Zirconia Polycrystalline Ceramic,’’ J. Am. Ceram. Soc., 80[6] 1624 1626 (1997).
24. F. Erdogan and G. C. Sih, ‘‘On the Crack Extension in Plates under Plane Loading and Transverse
Shear,’’ J. Basic Eng., 85, 519 527 (1963)
25. G. C. Sih, ‘‘Strain Energy Density Factor Applied to Mixed Mode Crack Problems,’’ Int. J. Fracture,
10[3] 305 321 (1974).
26. R. J. Niusmer, ‘‘An Energy Release Rate Criterion for Mixed Mode Fracture,’’ Int. J. Fracture, 10[3]
305 321 (1974)
27. K. Palaniswamy and W. G. Knauss, ‘‘On the Problem of Crack Extension in Brittle Solids under
General Loading,’’ pp. 87 148 in Mechanics Today, Vol. 4, Edited by S. Nemat Nasser, Pergamon
Press, New York (1978).
28. K. Hyashi and S. Nemat Nasser, ‘‘ ’’ J. Appl. Mech., 48, 520 524 (1981).
29. Y. Ueda, K. Ikeda, T. Yoa, and M. Aoki, ‘‘Characteristics of Brittle Fracture under General
Combined Modes Including Those under Biaxial Tensile Loads,’’ Eng. Fract. Mech., 18[6]
1131 1158 (1983).
30. D. K. Shetty, ‘‘Mixed Mode Fracture Criteria for Reliability Analysis and Design with Structural
Ceramics,’’ J. Eng. Gas Turbine Power, 109[7] 282 289 (1987).
31. D. Munz and T. Fett, Ceramics, pp. 45 49, Springer Verlag, Berlin, Germany (1999).
32. P. C. Paris and G. C. Sih, ‘‘Stress Analysis of Cracks,’’ ASTM ATP 381, pp. 30 81, American Society
for Testing and Materials, Philadelphia, PA, 1965.
33. T. M. Maccagno and J. F. Knott, ‘‘The Low Temperature Brittle Fracture Behavior of Steel in Mixed
Modes I and II,’’ Eng. Fract. Mech., 38[2 3] 111 128 (1991).
34. D Z. Zhu and R. A. Miller, ‘‘Thermal Conductivity and Elastic Modulus Evolution of Thermal
Barrier Coatings under High Heat Flux Conditions,’’ NASA/TM 1999 209069, National Aeronaut
ics and Space Administration, Glenn Research Center, Cleveland, OH (1999).
MODELING OF FRACTURE IN
FERROELASTIC CERAMICS
Chad M. Landis*
1. INTRODUCTION
Ferroelectric ceramics are used in applications ranging from actuators and sensors to
ultrasonic motors. A significant shortcoming of these materials in structural appli-
cations is their inherent brittleness. Specifically,
p most ferroelectrics have fracture
toughness levels on the order of KIc ¼ 1 MPa m. The characteristic of ferroelectric
ceramics that makes them useful as smart materials is their ability to deform and
change polarization irreversibly by the mechanism of domain switching. In a manner
akin to transformation toughening, domain switching leads to R-curve behavior and
toughness enhancement during crack growth in ferroelectrics. Hence, an understand-
ing of the dissipation due to domain switching and the fracture mechanics governing
these materials is crucial for the efficient design of ferroelectric devices.
As an initial investigation, this chapter will focus on fracture in unpoled ferro-
electric ceramics under mechanical loading. Note that in the absence of electrical
loading, unpoled ferroelectrics remain unpoled. Hence, unpoled ferroelectrics loaded
mechanically exhibit purely ferroelastic response; i.e. irreversible straining as a result
of applied stress. Experimental investigations on unpoled ferroelectric ceramics by
Meschke et al. (2000) and Oates et al. (2003) have found toughness enhancements in
the range of DKIc ¼ 40 100% of the initiation toughness, which corresponds to DGc ¼
100 300%. Here DKIc represents the difference in the steady state and initiation levels
of fracture toughness during crack propagation. Generally, this toughness enhance-
ment has been attributed to ferroelastic domain switching near the crack tip.
A number of theoretical investigations of switch toughening have been carried
out (Zhu and Yang (1997), Yang and Zhu (1998), Reece and Guiu (2002), and
Kreher (2002) ) which follow the approximate methods used to analyze transform-
*
Department of Mechanical Engineering and Materials Science, MS 321, Rice University, P.O. Box
1892, Houston, TX 77251, [email protected]
471
472 Chad M. Landis
ation toughening developed by McMeeking and Evans (1982) and Budiansky et al.
(1983). With the exception of the work of Kreher (2002), the models applied above
assume that domain switching near the crack tip occurs as a discrete transformation.
More specifically for the case of pure ferroelasticity, switching is assumed to occur at
a well defined level of stress. Once this level of stress is achieved the material
accumulates a finite pre-determined level of irreversible strain. The switch is all or
none, and there is no partial transformation allowed. Finally, once the switch occurs
the irreversible strain cannot be reverse-switched or reoriented; it is locked into the
material. This is an approximation to the actual ferroelastic constitutive behavior. In
reality, switching takes place over a range of stress levels, transformation is limited by
kinematic and equilibrium constraints, and strain reorientation can occur. The
previously stated approximations have been used for at least two reasons. First, for
most switching criteria the model can be solved analytically. Second, until very
recently, no multi-axial incremental constitutive law for ferroelastic switching has
been available to analyze this problem.
In this work the newly developed constitutive law of Landis (2003a) for ferro-
elastic switching is applied to the ferroelastic cracking problem. Some characteristics
of the ferroelastic behavior of ferroelectric ceramics include asymmetry in the levels
of attainable irreversible strain in tension versus compression, Fett et al. (1998) and
Fett and Thun (1998), a Bauschinger effect upon reverse loading, and reorientation
of the irreversible strain with reorientation of the applied stress. Due to the highly
non-proportional loading experienced by material elements near a passing crack tip,
the effects of unloading and strain reorientation are significant in the determination
of the quantitative effects of domain switching on the toughness of ferroelastic
materials. The constitutive law to be used here is cast in incremental form and
captures all of the features described above.
The remainder of the chapter is organized as follows. Section 2 presents the
constitutive law used to describe polycrystalline ferroelastic materials. Section 3
presents the fracture model, including the finite element method implemented to
determine the fields near a steadily growing crack and the crack tip energy release
rate. Results for the toughness enhancement predicted by the model will also be
presented in this section. Finally, Section 4 will be used to discuss the results and their
comparison to experimental observations.
2. CONSTITUTIVE LAW FOR FERROELASTIC SWITCHING
The nonlinear constitutive response of ferroelastic ceramics is a result of the mech-

anism of domain switching. The crystal structure of many ferroelectrics of techno-
logical interest is tetragonal. Single crystals of ferroelectric material are divided into
domains where each domain has a single crystallographic variant. Within a single
variant the orientation of the c-axis of the unit cells is identical. Domains of different
variants are separated by domain walls, and domain switching proceeds in an
incremental fashion by the motion of these domain walls. Since the volume of the
variants on either side of a domain wall is identical, domain switching can only result
in irreversible shear strains. Note here that we are concerned with ferroelasticity and
hence 908 degree switching only, since 1808 degree switching does not lead to
irreversible strains. A more detailed discussion of domain switching and the consti-
tutive behavior of single crystals can be found in Huber et al. (1999), Kamlah (2001),
or Landis (2003a). Here we are interested in the behavior of ferroelastic polycrystals
and a phenomenological constitutive description that can be readily implemented
within finite element computations.
In order to maintain simplicity, isotropic elastic response and linear kinematics
will be assumed such that the stress-strain relations can be written as
1þv v
«ij ¼ sij skk dij þ «rij (2:1)
E E
Here, «ij and sij are the Cartesian components of the infinitesimal strain tensor and
Cauchy stress tensor, v is the Poisson’s ratio and E is the Young’s modulus. Also, «rij
are the Cartesian components of the remanent strain tensor. The remanent strain is
the irreversible strain and can also be referred to as the plastic strain. In all cases
discussed in this paper, the datum for remanent strain is the state of the material as
cooled from above the Curie temperature. In such a state, all possible domain
orientations are equally likely.
The purpose of the nonlinear constitutive law is to provide the evolution of the
remanent strain history given the stress or total strain history. Consistent with the
facts that domain switching gives rise to deviatoric strains and ferroelastic ceramics
exhibit kinematic hardening effects, it is assumed that the material responds elastic-
ally within a switching (yield) surface F described by
3
F¼ sij sBij sij sBij s20 ¼ 0: (2:2)
2
If the stress state is on the switching surface and the load increment induces remanent
straining, then the remanent strain increment is normal to the switching surface such
that

«_ rij ¼ l sij sBij (2:3)
Here sij are the components of the stress deviator such that sij ¼ sij dij skk =3, sBij are
the components of the deviator of the back stress tensor sBij , s0 is the initial switching
(yield) strength of the material in tension or compression, and l is the as yet
undetermined plastic multiplier.
The back stress tensor leads to kinematic hardening and must be used to enforce
the remanent strain saturation conditions. The approach used to determine the back
stresses is based on the assumption that the internal state of the ferroelastic material
is completely characterized by the components of the remanent strain tensor, Cocks
and McMeeking (1999) and Landis (2002). This assumption leads to the identifica-
tion of a remanent potential, Cr «rij , such that the back stresses are derived from the
potential in the following manner
@Cr
sBij ¼ (2:4)
@«rij
In order to complete the constitutive theory, the form of Cr must be specified.

As previously noted, ferroelectric ceramics exhibit an asymmetry in the attainable
474 Chad M. Landis
levels of remanent strain in tension versus compression. Specifically, micromechani-

cal simulations by Frölich (2001) and Landis (2003a) suggest that the ratio of
remanent strain possible in tension to that possible in compression is 1.37:1. Fur-
thermore, this theoretical value is in the same range as the ratio of the maximum
poling remanent strain to the maximum compressive remanent strain due to stress
depolarization measured in PLZT by Lynch (1996), and in PZT by Fang and Li
(1999). Unfortunately, knowledge of the saturation strains in axisymmetric tension
and compression do not provide the complete range of saturation conditions for
other multiaxial remanent strain states.
By implementing the micromechanical model of Huber et al. (1999), Landis
(2003a) was able to map out the full range of remanent strain saturation states.
Figure 1 illustrates the remanent strain states that are possible in a ferroelastic
polycrystal comprised of unpoled randomly oriented single crystal grains with tet-
ragonal crystal structure. Remanent strain states that are below the curve are possible
in the material, while those above the curve are unattainable. The following proced-
ure has been proposed by Landis (2003a) to determine when saturation is ap-
proached. First, since the remanent strain due to domain switching is volume
conserving, the following two remanent strain invariants can be used to describe
any multi-axial remanent strain state.
1=2
2 r r 4 r r r 1=3
J2e ¼ e e and J3e ¼ e e e (2:5)
3 ij ij 3 ij jk ki
Here, erij is the remanent strain deviator, erij ¼ «rij dij «rkk =3. Due to volume conser-
vation, the introduction of the remanent strain deviator appears to be unnecessary.
However, the potential Cr will be described with these two invariants, and the
Figure 1. The remanent strain saturation curve dividing remanent strain space into regions that are
attainable and unattainable by a polycrystal assembled from randomly oriented tetragonal single crystals.
Only remanent strain states below the curve are attainable by such a material. The dots are numerical
results from Landis (2003a) obtained using a micromechanical self consistent model, and the line is one
divided by the function f given in Eqs. (2.7) and (2.8). The remanent strain invariants J2e and J3e are defined
in Eq. (2.5) and the results are normalized by the saturation strain in axisymmetric compression «c .
derivatives of Cr with respect to «rij will be affected by the distinction between the
remanent strain and its deviator. With the definition of these two invariants, a full
range of remanent strain saturation states can be probed by allowing the ratio of
J3e =J2e to vary from 1 (axisymmetric contraction) to 0 (pure shear) to 1 (axisym-
metric extension).
Now, define a strain-like variable « as

« ¼ J2e f J3e =J2e (2:6)
where
e e 3 e 6 e
J3 J3 J3 J3
f ¼ 0:0965 þ0:01 þ0:8935, for <0 (2:7)
J2e J2e J2e J2e
and
e e 3 e 6 e 21 e
J3 J3 J3 J3 J3
f ¼ 0:1075 0:027 0:028 þ0:8935, for $0 (2:8)
J2e J2e J2e J2e J2e
Here, f is a functional fit to the numerical results obtained from the micromechanical
computations described in Landis (2003a). This fit is illustrated in Figure 1. When the
remanent strain level characterized by « reaches the compressive saturation level, «c ,
the remanent strain will be saturated. In other words, the only possible remanent
strain states in the material are those that satisfy «#«c . Hence, in order to prohibit
remanent strain states characterized by « > «c , the remanent potential Cr must
increase without bound as « ! «c .
Two final considerations were used to establish the functional form for Cr . First,
the potential should yield smooth stress versus strain behavior as the remanent strain
goes through zero. Second, the functional form should be simple so that a parametric
study can be readily performed within the fracture model. These criteria have been
implemented to make the following choice for the remanent potential,
e
2
1 J m
Cr ¼ H0 «c 2 exp (2:9)
2 «c 1 «=«c
Here, H0 is a characteristic level of back stress that primarily affects the initial slope
of the uniaxial stress versus remanent strain curve, and m is another hardening
parameter that controls how abruptly the strain saturation conditions are reached.
Figure 2a illustrates the predictions of the effective stress versus the effective reman-
ent strain from the constitutive law for uniaxial compression, pure shear strain, pure
shear stress and uniaxial tension. It is interesting to note that the shear strain and
shear stress curves do not coincide. This feature is due to the fact that the material
can strain more in tension than in compression, and has been confirmed in micro-
mechanical simulations. Figure 2b illustrates the uniaxial stress versus remanent
strain hysteresis curves for two sets of the material parameters H0 and m.
With the remanent potential specified, it is possible to derive the back stresses
and finally solve for the plastic multiplier yielding the incremental form of the
constitutive law. For the implementation within the finite element model presented
476 Chad M. Landis
Figure 2. (a) Effective stress versus effective remanent strain curves for the model material described in
Section 2 in uniaxial compression, pure shear strain, pure shear stress and uniaxial tension tests. (b)
Uniaxial stress versus remanent strain hysteresis loops for the model material illustrating the effect of
the hardening parameter H0 . In both cases notice the asymmetry in the remanent strains that can be
achieved in tension versus compression.
in the next section, a backward Euler integration scheme was developed such that the
switching surface of Eq. (2.2) and the flow rule of Eq. (2.3) are satisfied at the end of
the increment.
3. THE FRACTURE MODEL
Throughout this paper it is assumed that applied loading is small enough such that
the height of the switching zone around the crack is small compared to the length of
the crack or any other specimen length dimension. Furthermore, only stress fields
symmetric with respect to the crack plane will be analyzed, i.e. mode I loading. As
discussed by Rice (1968), under these conditions a small scale switching analysis can
be performed on a semi-infinite crack with the remote applied stress field being that
of the mode I asymptotic elastic K-field,
8 9 8 9
< s11 = >
< 1 sin u2 sin 3u
2 >
=
KI u
s22 ! p cos 1 þ sin u2 sin 3u as r ! 1 (3:1)
: ; 2>:
2 >
3u ;
s12 2pr u
sin 2 cos 2
Here KI is the remote applied mode I stress intensity and r and u are polar coordin-
ates centered on the crack tip. Under plane strain conditions, KI is related to the
remote applied energy release rate G by
1 v2 2
G¼ KI (3:2)
E
Consider stable crack growth conditions such as a double cantilever beam
specimen loaded by forcing a wedge along the crack growth direction. Initially the
crack tip is stationary. As the loading is increased from zero, a localized zone of
domain switching surrounds the crack tip. The characteristic size of this switching
zone, Rs , is given as

1 KI 2 1 GE
Rs ¼ ¼ : (3:3)
3p s0 3p (1 v2 )s20
The crack tip will remain stationary and the switching zone size will continue to grow
until some criterion for crack growth is met. In this paper, it is assumed that crack
growth occurs when the crack tip energy release rate, Gtip , reaches a material specific
critical level of G0 .
Given this crack growth criterion and assuming that any non-proportional load-
ing effects near a stationary crack in these materials can be neglected, the path-
independence of the J-integral implies that Gtip ¼ G prior to crack growth and crack
growth will commence when G ¼ G0 . This point corresponds to the crack growth
initiation toughness on the R-curve for the material in consideration. Then, when the
crack advances, material elements near the crack tip will unload or load in a non-
proportional manner resulting in dissipated energy that must be supplied by the
external loading in order to maintain crack growth. Hence, after initiation KI and G
must continue to increase until steady growth conditions are met. This behavior
corresponds to the rising portion of the R-curve and occurs over a crack growth
range of a few times the maximum switching zone height. After a considerable amount
of crack growth, the switching zone and wake of switched material become well-
developed and crack growth continues at a constant level of KI or G. This corresponds
to the plateau on the R-curve where steady state crack growth occurs. The constant
level of applied energy release rate during steady crack growth will be denoted as Gss .
The analysis presented in this paper will focus only on the toughening due to
domain switching during the steady crack growth conditions described above. Under
steady growth conditions, all increments of field quantities can be related to deriva-
tives with respect to the x1 coordinate direction by
@f
f_ ¼ a_ (3:4)
@x1
Here, f is any scalar field quantity such as a Cartesian component of stress or strain,
and ȧ is the increment of crack advance in the x1 direction. Finally, Hutchinson
(1974) derived the following path-independent integral that can be use to determine
the crack tip energy release rate Gtip during steady crack growth,
ð
Gtip ¼ I (Un1 sij nj ui,1 )ds (3:5)
G
where G is a counterclockwise directed contour encircling the crack tip, ni are the
components of the unit normal directed to the right along the contour, ui are the
components of the displacement vector, and U is the history dependent stress work
density at a material point defined by
ð«
U ¼ sij d«ij (3:6)
0
478 Chad M. Landis
In order to determine the amount of switch toughening that can occur in a given
material, the stress and strain fields near a steadily growing crack must be computed.
Due to the complex nonlinear character of the ferroelastic constitutive law described
in Section 2, these fields cannot be determined analytically; instead, they will be
computed numerically with the finite element method. The numerical model to be
used here was originally developed by Dean and Hutchinson (1980) and has been
used by numerous authors to investigate steady crack growth in dissipative materials
including conventional elastic-plastic, strain-gradient plastic, and visco-plastic ma-
terials among others. The basic idea of the formulation is that the finite element mesh
is fixed to the crack tip and the material moves through the mesh form x1 ¼ 1 to
x1 ¼ 1. The virtual work expression used to derive the finite element method is
ð ð ð
d«ij Cijkl «kl dV ¼ dui Ti dS þ d«ij Cijkl «rkl dV (3:7)
V s V
where S is the boundary of the volume V, Cijkl are the Cartesian components of the
isotropic elastic stiffness tensor that can be written in terms of E and v, and the
tractions acting on the boundary S are given as Ti ¼ sji nj . These tractions are
determined from the stress field of Eq. (3.1).
After insertion of the appropriate finite element interpolations, the left-hand
side of this equation yields a standard linear elastic stiffness matrix dotted with the
vector or unknown nodal displacements. The first term on the right is a vector of
applied forces, and the last term is a residual body force vector that accounts for the
remanent strain in the volume. An iterative procedure is applied to solve the non-
linear steady crack growth problem. Initially it is assumed that the remanent strain at
every point in the body is zero. Hence, the body force vector due to the remanent
strains is zero and the linear elastic asymptotic crack tip field solution is recovered.
With this solution, the constitutive law of Section 2 is integrated along streamlines of
constant height above the crack plane from the right edge of the finite element mesh
to the left. This integration yields a new approximation for the distribution of
remanent strains. The residual body force vector is recomputed and a new nodal
displacement solution is calculated. Again, the constitutive law is integrated along
streamlines to obtain a new distribution of remanent strains and the procedure is
repeated until the solution achieves a suitable level of convergence.
Within this model for fracture in ferroelastic materials, dimensional analysis
dictates that all field quantities, for example the stresses, will be of the form

sij x y «c E H 0
¼ sij , ; , ,m,v (3:8)
s0 Rs Rs s0 s0
where sij are dimensionless functions of the normalized spatial coordinates para-
meterized by the normalized material quantities. Also note that here Rs is the
characteristic size of the switching zone given by Eq. (3.3) when G ¼ Gss . Further-
more, the relative toughening ratio will only depend on the dimensionless material
parameters, i.e.

G Gss «c E H 0
¼ ¼G , ,m,v (3:9)
Gtip G0 s 0 s0
Equation (3.9) reinforces that in this model, crack growth occurs when the crack tip
energy release rate, Gtip , reaches some material specific critical level G0 .
3.1. Results
Prior to presenting results for the toughening ratio, some features of the distributions
of stress and remanent strain fields will be given. Due to the mode I symmetry of the
problem only the upper half plane needs to be analyzed. Figure 3 is a contour plot of
the effective strain distribution near a steadily growing crack in a ferroelastic mater-
ial. As a material point approaches the crack tip from the right, switching commences
as it enters the active switching zone. Very close to the crack tip the remanent strain
of such a material point will accumulate, reorient as the crack tip passes and
eventually the material point will unload elastically. The active switching zone is
the region where remanent strain is accumulating or reorienting, while in the
unloaded wake the remanent strain distributions are uniform in the x-direction.
The shape of the active switching zone is very similar to the shape of the active
plastic zone near a steadily growing crack in an elastic-perfectly plastic material,
Dean and Hutchinson (1970), Landis et al. (2000). One significant difference between
the active zones in these two types of materials is that in the ferroelastic material the
unloaded elastic sector behind the crack tip actually lies along the crack faces, while
the same feature in the perfectly plastic material is tilted away from the crack faces.
Another difference between the ferroelastic and perfectly plastic material is that
remanent strains in the ferroelastic material are bounded, while plastic strains in
the perfectly plastic material are unbounded.
Due to this strain saturation effect, the stresses near the crack tip in the ferroe-
lastic material increase severely. In fact,
p the numerical results suggest that very close to
the crack tip the stresses have a 1= r radial dependence. Hence the crack tip stress
intensity factor KItip can be defined such that on the plane ahead of the crack tip
Figure 3. The distribution of effective remanent strain near a growing crack in a ferroelastic material. The
active switching zone, elastically unloaded wake, and unloaded elastic sector behind the crack tip are each
denoted on the illustration. The material law is given in Section 2, and the material parameters for this
computation are «c E=s0 5, H0 s0 , m 0:01 and v 0:25.
480 Chad M. Landis
p
syy ! KItip = 2pr as r ! 0. However, the results also suggest that angular dependence
of the stress field asymptotically close to the crack tip is not equivalent to the asymp-
totic field for an isotropic elastic solid. The discrepancy between the ferroelastic
asymptotic field and the isotropic elastic asymptotic field is most pronounced for
low values of H0 =s0 and large values of «c E=s0 . Figure 4a plots the angular depend-
ence of the stresses near a growing crack tip in a ferroelastic material with
p properties
«c E=s0 ¼ 3, H0 ¼ 0:1s0 , m ¼ 0:01 and v ¼ 0:25. Notice that the 1= r radial de-
pendence of the stresses has been accounted for in the normalization. Furthermore,
since the stresses do not follow the isotropic asymptotic field, the relationship between
Gtip and KItip is not given by Eq. (3.2). This then implies that G=Gtip 6¼ (KI =KItip )2 .
Specifically, for these material parameters, G=Gtip ¼ 3:02 whereas KI =KItip ¼ 2:25.
For comparison, if the hardening parameter is changed to H0 ¼ 10s0 , then the stress
field is very close to the isotropic asymptotic field (the sxx component falls noticeably
but just slightly below the isotropic curve) and G=Gtip ¼ 1:29 with KI =KItip ¼ 1:20.
A similar approach of the asymptotic ferroelastic stresses to the isotropic elastic
asymptotic fields occurs as «c E=s0 ! 0, which is to be expected.
Figure 4b plots the angular dependence of the remanent strains near the growing
crack tip in the same material described above. Note on Figure 4b that the saturation
strain variable « is very close to «c over the entire angular range, denoting that the
material is in fact near saturation close to the crack tip. However, the remanent strain
components are not uniform, indicating that remanent strain reorientation does
occur around a growing crack.
The primary goal of this fracture model is to determine how the steady state
toughness enhancement in ferroelastic materials, Gss =G0 , depends on the material
properties. Eq. (3.9) identifies the material properties in question and ranks them in
order of significance. Poisson’s ratio n will be shown to have a very weak influence
over the toughness enhancement. The two hardening parameters, H0 =s0 and m, have
a much stronger influence on Gss =G0 , and it will be shown that as the hardness of the
Figure 4. (a) The angular distribution of the stresses very close to the crack tip. The dots are numerical
results for a material with «c E=s0 3, H0 0:1s0 , m 0:01 and v 0:25, and the solid lines are the
p
result for an isotropic elastic solid. Note that the 1= r radial dependence has been accounted for within the
normalization. (b) The angular distribution of remanent strains very close to the crack tip from the same
computation. The numerical results plotted in both figures are for all integration stations within the radial
range 8 103 Rs < r < 1:3 102 Rs .
material increases then the toughness enhancement decreases. Finally, the most
fundamental material parameter that has been identified in all transformation tough-
ening models is «c E=s0 . Explicitly, this is the ratio of the irreversible strain to the
characteristic
elastic strain in the material. However, if this ratio is rearranged as
ðs0 «c Þ= s20 =E , then it can also be interpreted as the ratio of the characteristic level of
dissipated to stored energy in the material. Results for the dependence of Gss =G0 on
«c E=s0 will be given presently.
Figure 5 plots Gss =G0 versus «c E=s0 for a range of initial hardening values H0 =s0
with m ¼ 0:01 and n ¼ 0:25. The first observation to make about this plot is that the
toughness enhancement increases monotonically as the saturation level of irreversible
remanent strain increases. Larger saturation strains imply that the material can
dissipate more energy, which in turn implies higher fracture toughness. A second
feature illustrated on this plot is the dependence on the hardening parameter H0 =s0 .
As shown in Figure 2b, increasing levels of H0 =s0 induce higher stresses in the material
that are required to reach given levels of remanent strain. Figure 5 illustrates that there
is a relatively strong dependence of the toughness enhancement on the initial hardness.
Hence, the actual shape of the stress strain curve in these materials is important when
making a quantitative determination of the fracture toughness.
Finally, it is worth noting that these toughness curves are relatively straight at
low levels of «c E=s0 . Therefore, the initial slope of these curves when «c E=s0 ¼ 0
provides a useful measure of the toughness enhancement. Specifically, the toughen-
ing parameter a is defined as

s0 @Gss
a¼ (3:10)
G0 E @«c «c ¼0
Figure 6 plots the toughening parameter a as a function of H0 =s0 for ranges of

Poisson’s ratio and the second hardening parameter m. The insert plots hysteresis
Figure 5. The toughness enhancement, Gss =G0 , during steady crack growth in a ferroelastic material as a
function of the saturation strain level, «c E=s0 , for a range of initial hardening values, H0 =s0 .
482 Chad M. Landis
Figure 6. The initial slope of the toughness ratio versus the relative saturation strain, a, as a function of
the initial hardening H0 =s0 , for a range of Poisson’s ratio and m. The insert plots hysteresis loops in order
to illustrate the dependence of the stress strain behavior on m.
loops for a few values of m and a single value of H0 =s0 ¼ 1. First notice that a is only
weakly dependent on Poisson’s ratio, and this dependence is very weak for the range
of 0:25 < n < 0:4. The toughness enhancement has a stronger dependence on the
hardening parameter m. However, note from the insert that increasing m drastically
increases the stress levels near strain saturation. Effectively, large m values reduce the
saturation strain of the material. It is likely that most materials are characterized
within the range of 0 < m < 0:1.
In summary, results from numerical computations of the stress and strain fields
and the toughness enhancement during steady crack growth in ferroelastic materials
have been presented. The computations illustrate a few interesting features and
confirm some intuitive p hypotheses about the solution. First, the near tip stresses
appear to recover a 1= r singular form, however the radial dependence of these
stresses is not the same as those for an isotropic elastic material. It was also shown
that the distributions of remanent strain are not trivial and do reorient as the crack
tip passes. Lastly, as would be expected, the steady state toughness of the material
increases as the relative saturation strain increases, and decreases as the hardness of
the material increases.
4. DISCUSSION
The steady state fracture toughness enhancement in ferroelastic ceramics has been
computed accurately by implementing a micromechanically verified phenomeno-
logical constitutive law to describe the continuum material response. The constitutive
law is able to account for most of the features of ferroelastic deformation, including the
asymmetry in tensile versus compressive remanent strains and the Bauschinger effect.
This constitutive law was implemented within the finite element method to determine
the fields near a steadily growing crack. Crack growth was assumed to occur at a
critical level of the crack tip energy release rate. The most significant material features
that control the level of toughening in these materials have been identified as the
relative saturation strain level and the shape of the uniaxial stress-strain curve.
It is worth noting that the assumptions implemented within transformation
toughening-type analyses have not been used in this model. Specifically, the remanent
strain is allowed to vary and change within the switching zone near the crack tip.
Figure 4b illustrates how the remanent strain components vary near the crack tip and
well into the region were strain saturation is achieved. This figure emphasizes the fact
that the stress fields near a growing crack tip act to reorient the remanent strain in the
material. This reorientation of the remanent strain is a dissipative process and influ-
ences the toughness enhancement due to switching. Transformation toughening
models that assume a ‘‘locked in’’ switching strain simply cannot account for this fea-
ture of the problem. While these transformation toughening models can give estimates
of the toughness enhancement due to switching, material features such as the initial
hardening ratio H0 =s0 do not enter into these types of models. Hence, qualitative
determinations of the effects of such parameters cannot be ascertained from trans-
formation toughening models. A more detailed discussion of transformation toughen-
ing switching models, their limitations, and a comparison to the results from the finite
element computations performed in this work are included in Landis (2003b).
Due to the inherent brittleness of ferroelastic ceramics, there is a dearth of detailed
experimental measurements on their full range of uniaxial stress-strain behavior. In
tension, the material fractures before strain saturation can be achieved. Hence, most
experiments in the literature are carried out in compression. This is unfortunate since
even a partial hysteresis loop would be very useful for the determination of the switching
strength. Specifically, the width of the hysteresis loop is approximately 2s0 . In order to
obtain information from the uniaxial compressive stress-strain curve, the following
observations are made. The elastic modulus should be taken as the initial slope of the
unloading portion of the stress-strain curve. The compressive saturation strain is at least
the strain extrapolated back to zero stress using this initial unloading slope. If the stress-
strain behavior is measured on an electrically poled sample, then the nonlinear strain
range is approximately 2:37«c . Finally, the slope of the switching portion of the stress-
strain curve is approximately equal to H0 =«c . Applying these considerations to the data
available in the literature, it was found that PZT materials have properties in the range
«c E=s0 3 7 and H0 =s0 1, Schäufele and Härdtl (1996), and PLZT has proper-
ties near «c E=s0 5 and H0 =s0 1, Lynch (1996). Applying these material properties
to the model computations of Section 3 the predicted steady state toughness enhance-
ment for these materials is in the range of Gss =G0 2:5 5:8 or DK1 60 140%. As
mentioned in the Introduction, experimental investigations on unpoled ferroelectric
ceramics by Meschke et al. (2000) for barium titanate and Oates et al. (2003) for PZT,
have found toughness enhancements in the range of DKIc ¼ 40 100% of the initiation
toughness. Hence, the model yields prediction in the appropriate range. A more critical
experimental comparison can only be made if both fracture data and detailed stress-
strain measurements are obtained for the same material.
ACKNOWLEDGEMENT
The author would like to acknowledge support from the Army Research Office
contract DAAD19-02-1-0241.
484 Chad M. Landis
REFERENCES
Budiansky, B., Hutchinson, J. W. and Lambropoulos, J. C., 1983, Continuum theory of dilatant trans
formation toughening in ceramics, International Journal of Solids and Structures 19:337 355.
Cocks, A.C.F. and McMeeking, R.M., 1999, A phenomenological constitutive law for the behavior of
ferroelectrics, Ferroelectrics 228:219 228.
Dean, R.H. and Hutchinson, J.W., 1980, Quasi static steady crack growth in small scale yielding. In
Fracture Mechanics, ASTM STP 700, 383 405.
Fang, D. and Li, C., 1999, Nonlinear electric mechanical behavior of a soft PZT 51 ferroelectric ceramic,
Journal of Materials Science 34:4001 4010.
Fett, T., Müller, S., Munz, D., Thun G., 1998, Nonsymmetry in the deformation behavior of PZT, Journal
of Materials Science Letters 17:261 265.
Fett, T. and Thun, G., 1998, Nonsymmetric deformation behavior of lead zirconate titanate determined in
bending tests, Journal of the American Ceramic Society 81:269 272.
Frölich, A., 2001, Mikromechanisches modell zur ermittlung effektiver materialeigenschaften von piezo
elektrischen polykristallen, Dissertation Thesis, Universität Karlsruhe, Germany.
Huber, J.E., Fleck, N.A., Landis, C.M. and McMeeking, R.M., 1999, A constitutive model for ferroelec
tric polycrystals, Journal of the Mechanics and Physics of Solids 47:1663 1697.
Hutchinson, J.W. 1974, On steady quasi static crack growth, Harvard University Report, Division of
Applied Sciences, DEAP S 8.
Kamlah, M., 2001, Ferroelectric and ferroelastic piezoceramics modeling and electromechanical hyster
esis phenomena, Continuum Mechanics and Thermodynamics 13:219 268.
Kreher, W., 2002, Influence of domain switching zones on the fracture toughness of ferroelectrics, Journal
of the Mechanics and Physics of Solids 50:1029 1050.
Landis, C.M., Pardoen, T. and Hutchinson, J.W., 2000, Crack velocity dependent toughness in rate
dependent materials, Mechanics of Materials 32:663 678.
Landis, C.M., 2002, Fully coupled, multi axial, symmetric constitutive laws for polycrystalline ferroelectric
ceramics, Journal of the Mechanics and Physics of Solids 50:127 152.
Landis, C.M., 2003a, On the strain saturation conditions in polycrystalline ferroelastic ceramics, to appear
in Journal of Applied Mechanics.
Landis, C.M., 2003b, On the fracture toughness of ferroelastic materials, submitted to Journal of the
Mechanics and Physics of Solids.
Lynch, C.S., 1996, The effect of uniaxial stress on the electro mechanical response of 8/65/35 PLZT, Acta
Materialia 44:4137 4148.
McMeeking, R.M. and Evans, A.G., 1982, Mechanics of transformation toughening in brittle materials,
Journal of the American Ceramic Society 65:242 246.
Meschke, F., Raddatz, O., Kolleck, A., and Schneider, G.A., 2000, R curve behavior and crack closure
stresses in barium titanate and (Mg, Y) PSZ ceramics, Journal of the American Ceramic Society
83:353 361.
Oates, W.S., Lynch, C.S., Rödel, J., Lupascu D. and Aulbach, E., 2003, Investigation of subcritical crack
growth in lead zirconate titanate, submitted to the Journal of the American Ceramic Society.
Reece, M.J. and Guiu, F., 2002, Toughening produced by crack tip stress induced domain reorientation in
ferroelectric and/or ferroelastic materials, Philosophical Magazine A82:29 38.
Rice, J.R., 1968, A path independent integral and the approximate analysis of strain concentration by
notches and cracks, Journal of Applied Mechanics 35:379 386.
Schäufele, A.B. and Härdtl, K.H., 1996, Ferroelastic properties of lead zirconate titanate ceramics,
Joiurnal of the American Ceramic Society 79:2637 2640.
Yang, W. and Zhu, T., 1998, Switch toughening of ferroelectrics subjected to electric fields, Journal of the
Mechanics and Physics of Solids 46:291 311.
Zhu, T. and Yang, W., 1997, Toughness variation of ferroelectrics by polarization switch under non
uniform electric field, Acta Materialia 45:4659 4702.
MODELLING MULTILAYER
DAMAGE IN CROSS-PLY
CERAMIC MATRIX
COMPOSITES
M Kashtalyan and HW Chandler*
1. INTRODUCTION
Compared with monolithic ceramics, continuous fibre reinforced ceramic matrix

composites (CMCs) have higher fracture toughness, better fatigue resistance and
damage tolerance, retaining their load carrying capacity in the presence of damage.
Over the years, a large number of experimental and theoretical studies have been
conducted on damage evolution and failure mechanisms of CMCs, see review by
Birman and Byrd (2000). Much of this work has been concerned with unidirectional
materials. However, if CMCs are to be applied under multi-axial stress conditions,
multidirectional laminated or woven composites containing fibre reinforcement at
more than one orientation need to be employed.
Matrix cracking has long been recognised as a first damage mode observed
in unidirectional CMCs under mechanical, thermomechanical or thermal loading.
A number of studies of cross-ply CMC laminates under both quasi-static and cyclic
uniaxial tensile loading have shown that matrix cracks accumulate both in the 908
and 08 plies. Matrix cracking in the 908 plies leads to formation of macrocracks, or
tunnelling cracks, while in the 08 plies matrix cracks are bridged by the fibres and are
accompanied by fibre-matrix debonding. Macrocracks, or tunnelling cracks, are
perpendicular to the direction of loading and run parallel to the fibres in the 908
layer. The onset of matrix cracking in cross-ply CMCs occurs at lower strain levels
than in unidirectional materials, with 908 ply cracks initiating before the 08 ply
*
School of Engineering and Physical Sciences, University of Aberdeen, AB24 3UE, UK
233
234 M Kashtalyan et al.
cracks, although the latter may reach a higher saturation density (Pryce and Smith,
1994; Erdman and Weitsman, 1998).
Matrix cracking results in reduction in the laminate stiffness and strength.
Stiffness reduction in CMCs is greater than that observed in polymer matrix com-
posites due to lower ratio between fibre and matrix moduli. It also affects coefficients
thermal expansion and vibration frequencies. A number of models have been sug-
gested to estimate the effect of transverse macrocracks in the 908 plies and matrix
cracks bridged by fibres in the 08 plies on the mechanical properties of cross-ply
CMC laminates (Pryce and Smith, 1994; Daniel and Anastassopoulos, 1995; Lu and
Hutchinson, 1995; Erdman and Weitsman, 1998; Birman and Byrd, 2001; Yasmin
and Bowen, 2002; Birman and Weitsman, 2003).
When a cross-ply laminate is subjected to biaxial loading, matrix cracking may
lead to formation of macrocracks both in the 908 and 08 plies. The microscopic
damage evolution in cross-ply CMCs under biaxial loading has not been well char-
acterised yet. In this paper, stiffness-damage behaviour of cross-ply [0m =90n ]s CMC
laminates with macrocracks both in the 908 and 08 plies is predicted theoretically
using the Equivalent Constraint Model (ECM) of the damaged lamina (Fan and
Zhang, 1993; Kashtalyan and Soutis, 1999). To capture simultaneous accumulation
of damage both in the 908 and the 08 plies, two ECM laminates are analysed
simultaneously, as a coupled problem, instead of the original one. In each ECM
laminate, one ply group or layer contains damage explicitly, while the other is
replaced with equivalent homogeneous layer(s) with reduced stiffness properties.
Following analysis of stresses in the explicitly damaged layer(s) of ECM laminates,
closed form expressions for the reduced stiffness properties of the damaged laminate
are derived representing them as functions of crack densities in the 908 and 08 plies.
Numerical results for SiC/CAS cross-ply laminates of different lay-ups are presented
and discussed.
2. EQUIVALENT CONSTRAINT MODEL
The Equivalent Constraint Model (ECM) was introduced by Fan and Zhang (1993)
with the aim to analyse the in-situ constraint effects on damage evolution in a
particular lamina within a multidirectional laminate. In this model, all the laminae
below and above the chosen lamina are replaced with homogeneous layers having the
equivalent constraining effect. It is assumed that the in-plane stiffness properties of
the equivalent constraint layers can be calculated from the classical lamination
theory, provided stresses and strains in them are known.
Figure 1 shows a [0m =90n ]s cross-ply laminate damaged by transverse and
longitudinal tunnelling cracks (macrocracks). The cracks in each ply are assumed
to be spaced uniformly and span the full thickness and width of the ply. Spacings
between longitudinal macrocracks are denoted 2s1 , and between transverse macro-
cracks 2s2 The laminate is subjected to biaxial tension (s
s11 and s22 ) and shear loading
s12 ).
(s
Application of the ECM to the laminate with transverse and longitudinal
macrocracks described above means that instead of considering the damaged con-
figuration shown in Fig. 1, the following two ECM laminates will be analysed. In the
ECM1 laminate, the 08 lamina (layer 1) contains damage explicitly, while the 908
Modelling Multilayer Damage in Cross-Ply Ceramic Matrix Composites 235
Figure 1. Cross ply laminate under general in plane loading damaged by transverse macrocracks in the
908 layer and longitudinal macrocracks in the 08 layer.
lamina (layer 2), damaged by transverse cracks, is replaced with a homogeneous layer
with reduced stiffness properties. In the ECM2 laminate, the 908 lamina (layer 2) is
damaged explicitly, while the damaged 08 laminae are replaced with homogeneous
layers with reduced stiffness properties. Representative segments of the ECM2 and
ECM1 laminates, containing a single crack (pair of ) crack(s) are shown in Fig. 2.
The purpose of the analysis of an ECM laminate is to determine the reduced
stiffness properties of the explicitly damaged layer. Since the reduced stiffness prop-
erties of the equivalent constraint layer(s) used in the analysis of an ECM laminate,
are assumed to be determined from the analysis of the other ECM laminate, prob-
lems for ECM1 and ECM2 laminates are coupled. The advantage of the ECM-based
approach is that it avoids cumbersome consideration of a repeated laminate element
defined by the intersecting pairs of macrocracks used by other researchers (Hashin,
1987; Tsai and Daniel, 1992; Henaff-Gardin et al, 1996a,b).
3. STRESS ANALYSIS
The ECM1 and ECM2 laminates are referred to the same co-ordinate system x1 x2 x3
(Fig. 2) and assumed to be under generalised plane strain in the planes Ox1 x3 and
Ox2 x3 , respectively. Due to symmetry, only a quarter of each segment needs to be
considered.
Figure 2. Equivalent Constraint Model of a cross ply laminate with multilayer damage.
The in-plane stresses in the explicitly damaged mth layer of the ECMm laminate
(m ¼ 1 or 2) can be can be determined from the equilibrium equations
(m)
d (m,k) tj
~jm þ ( 1)k
s ¼ 0, j,m,k ¼ 1,2 (1)
dxm hk
(m,k)
Here s ~ij are the in-plane microstresses (i.e. stresses averaged across the layer
thicknesses and the width of the laminate) in the kth layer of the ECMm laminate
and t (m)
j are the interface shear stresses at the (0/90) interface of the ECMm laminate.
The in-plane microstresses are related to the total stresses sij applied to the laminate
by the equations of global equilibrium
(m,1) (m,2)
~ij
xs þs
~ij ¼ (1 þ x)sij x ¼ h1 =h2 , i, j ¼ 1,2 (2)
The constitutive equations for both layers in terms of microstresses and microstrains are
{s ^ (m) ]{~
~(m,m) } ¼ [Q «(m,m) }, (3a)
~(m,k) } ¼ [Q(k) ]{~
{s «(m,k) } (3b)
Here [Q^ (m) ] is the in-plane stiffness matrix of mth layer of the ECMm laminate, and
[Q ] is the modified in-plane stiffness matrix of the equivalent constraint layer (kth
(k)
layer). Crack surfaces are assumed to be stress-free, so that

(m,m)
~jm
s jxm ¼0 ¼ 0, j,m ¼ 1,2 (4)
The interface shear stresses t (m)

j can be expressed in terms of the in-plane displace-
(m,k)
ments uj , j ¼ 1,2 as
(m,1) (m,2)
t (m)
j ¼ Kj(m) (~
uj ~uj ) (5)
(m,1)
where Kj are the shear lag parameters. If out-of-plane shear stresses sj3 , j ¼ 1,2,
vary linearly with x3 and in the 08 lamina this variation is restricted to the region of
about one ply thickness t, the shear lag parameters will be
^ (1) G
3G ^ (2)
j3 j3
Kj(m) ¼ Kj ¼ , h ¼ t=h1 , j ¼ 1,2 (6)
^ (1) þ (1 þ (1 h)=2)hh1 G
h2 G ^ (2)
j3 j3
Here, G ^ (k) , k ¼ 1,2 are the out-of-plane shear moduli of the kth layer; they are
j3
unaffected by the presence of matrix cracks. Equations (1) (6) provide full set of
equations and boundary conditions, required to determine the in-plane microstresses
(m,m)
~jm j,m ¼ 1,2 in the representative segment of the ECMm laminate. The above-
s
mentioned set is reduced to two single ordinary differential equations
(m,m)
~mm
d2s (m,m)
L(m)
1 s ~mm þ V(m) (m)
11 s11 þ V22 s22 ¼ 0 (7a)
dx2m
(m,m)
~12
d 2s (m,m)
L(m)
2 s ~12 þ V(m)
12 s 12 ¼ 0 (7b)
dx2m
where L(m) (m) (m) (m) (m)

1 , L1 , V11 , V22 , V12 are the laminate constants depending on the layer
compliances S ^ (m) , S(k) , k 6¼ m, shear lag parameters Kj and the layer thickness ratio
ij ij
x ¼ h1 =h2 . In detail, they are presented elsewhere (Kashtalyan and Soutis, 2000). Given
the boundary conditions at the macrocrack surfaces, Eq. (4), solutions to Eqs. (7) are
0 q 1
(m)
(m,m) 1 B cos h[ L1 (xm sm )]C (m) (m)
~mm
s ¼ (m) @1 q A(V11 s11 þ V22 s22 ) (8a)
L1 (m)
cos h[ L1 sm ]
0 q 1
(m)
(m,m) 1 B cos h[ L2 (xm sm )]C (m)
~12 ¼ (m) @1
s q AV12 s12 (8b)
L2 cos h[ L(m) s m ]
2
Stress analysis presented here does not take into account residual thermal stresses,
generated during fabrication due to mismatch in thermal properties of the CMC
constituents. In multidirectional laminates, residual thermal stresses at the lamina
level can be estimated using the classical lamination theory (Daniel and Ishai, 1994).
For example, residual stresses in the 08 and 908 plies of a cross-ply [0=90]s SiC/CAS
laminate were found to be of magnitude 5 MPa for an assumed temperature difference
DT ¼ 10008C. At that, stresses in the fibre direction are tensile and those in the
direction transverse to the fibres are compressive in each ply. The properties of unidir-
ectional SiC/CAS material used in the calculations were as follows (Daniel and Ishai,
1994): longitudinal modulus EL ¼ 121 GPa, transverse modulus ET ¼ 112 GPa,
inplane shear modulus GLT ¼ 44 GPa, major Poisson’s ratio vLT ¼ 0:2, minor Pois-
son’s ratio vTL ¼ 0:18, longitudinal thermal expansion coefficient aL ¼ 4:05 10 6 =8C,
transverse thermal expansion coefficient aT ¼ 4:15 10 6 =8C. Shear modulus in the
plane perpendicular to the fibres (plane of isotropy) was estimated as GTT ¼ 45 GPa,
and single ply thickness was taken t ¼ 0:2 mm.
4. STIFFNESS ANALYSIS
The reduced stiffness properties of the damaged mth layer can be determined by
applying the classical lamination theory to the ECMm laminate after replacing the
explicitly damaged layer with an equivalent homogeneous one. Constitutive equations
of the homogeneous layer, equivalent to the explicitly damaged one, have the form
s(m,m) } ¼ [Q(m) ]{««(m,m) }

{s (9)
The lamina macrostresses {s s(m,m) } in the equivalent homogeneous layer can be

obtained by averaging the in-plane microstresses, Eqs. (8), across the length of the
representative segments as
q
(m,m) 1
smm ¼ (m) [1 tan h( L(m) (m) (m)
1 sm )](V11 s11 þ V22 s22 ) (10a)
L1
q
(m,m) 1
~12 ¼ (m) [1 tan h( L(m)
s (m)
2 sm )]V12 s12 (10b)
L2
The lamina macrostrains {« «(m,m) } are assumed to be equal to those in the equivalent
(m,m)
constraint layer, i.e. {«
« } ¼ {« «(m,k) }, as well as to the applied laminate strain, and
are calculated from the constitutive equation for the equivalent constraint layer, Eq.
(3b). The modified in-plane stiffness matrix [Q(m) ] of the homogeneous layer equiva-
lent to the mth layer of the ECMm laminate is related to the in-plane stiffness matrix
^ (m) ] of the undamaged layer as [Q(m) ] ¼ [Q
[Q ^ (m) ] [R(m) ], where
2 3
^ (m) L(m)
Q ^ (m) L(m)
Q 0
11 22 12 22
6 ^ (1) )2 (m) 7
6 (m) (m) (Q 7
[R ] ¼ 6
(1) ^ 12 7,
6 Q12 L22 ^ (1) 22
L 0 7 (11a)
4 Q 11 5
0 0 ^ (m) L(m)
Q 66 66
2 3
^ (2) )2 (m)
(Q 12 ^ (m) L(m)
Q 0
6 ^ (2) L22 12 22 7
6 Q 7
[R(2) ] ¼ 6 7
22
6 ^ (m) (m) 7 (11b)
4 Q12 L22 ^ (m) L(m)
Q 0 5
22 22
0 0 ^ (m) L(m)
Q 66 66
Here L(m) (m) th

22 , L66 are the In-situ Damage Effective Functions for the m layer of the
ECMm laminate
(m,m)
smm
L(m)
22 ¼ 1 ^ (m,m) (m,m)
Q1m «1 þQ ^ (m,m) «(m,m)
2m 2
" #" # 1 (12a)
Dmcm l(m) (m) D mc
m l (m)
¼ 1 1 (m) tan h 1mc 1 þ a1 (m) tan h 1mc , m ¼ 1,2
l1 Dm l1 Dm
(m,m)
s12
L(m)
66 ¼ 1 , (m,m)
^
Q
(m m)
66 «6
" #" # 1 (12b)
Dmc
m l(m) Dmc
m l(m)
¼1 1 tan h 2mc 1þ a(m)
2 tan h 2mc , m ¼ 1,2
l(m)
2
Dm l(m)
2
Dm
Here Dmc mc
1 , D2 are damage parameters associated with longitudinal and transverse
macrocracks, respectively. The constants l(m) (m)
i ,ai ,i ¼ 1,2 depend solely on the layer
compliances S ^ , S , k 6¼ m, shear lag parameters Kj and the layer thickness ratio x.
(m) (k)
ij ij
Through the modified compliances Sij(k) , k 6¼ m of the equivalent constraint layer (kth
layer) of the ECMm laminate, the functions L(m) (m) th
22 , L66 for the m layer of the ECMm
laminate implicitly depend on the damage parameter Dk ¼ hk =sk for the kth layer.
mc
The In-situ Damage Effective Functions for both layers L(k) qq , q ¼ 2,6 k ¼ 1,2 form a
system of 4 simultaneous nonlinear algebraic equations, which is solved computa-
tionally by a direct iterative procedure. Consequently, all four functions and there-
fore the reduced stiffness properties of the layers are determined as functions of
Dmc mc
1 , D2 . Once reduced stiffness properties of all layers are know, the extensional
stiffness matrix of the damaged laminate can be obtained as
X
A] ¼
[A [Q(k) ]hk (13)
k
If the interaction between the transverse and longitudinal cracks is neglected, it

is possible to derived a closed-from expression for reduction in the shear modulus
caused by transverse and longitudinal macrocracks
" #
(1,1) (2,2)
Gxy Dmc Dmc l2 l2
¼ 1 (1,1) (2,2) tanh mc tanh mc
1 2
^ xy
G l2 l2 D1 D2
" # 1
(1,1) (2,2) (1,1) (2,2)
Dmc l 1 D mc
l D mc
D mc
l l
1þx (11,1) tanh 2 mc þ (22,2) tanh 2 mc þ (11,1) (22,2) tanh 2 mc tanh 2 mc (14)
l D1 xl D2 l l D1 D2
2 2 2 2
Here Gxy and G ^ xy are the shear moduli of the damaged and intact laminate, respect-
ively. This expression, Eq. (14), is similar to the semi-empirical expression suggested
by Tsai and Daniel (1992) on the basis of the ‘superposition’ of solutions for a single
set of cracks
" # 1
(1,1) (2,2)
Gxy Dmc l 1 D mc
l
¼ 1 þ x (11,1) tan h 1 mc þ 2
tan h 2 mc (15)
^ xy
G l2 D1 x l(2,2) D2
2
However, Eq. (14) contains some additional terms in comparison to Eq. (15). If
longitudinal cracks are absent, Eqs. (14) and (15) are reduced to the same expression.
The ECM has been already successfully applied to predict reduced stiffness proper-
ties (axial and shear moduli and the Poisson’s ratio) in polymer matrix composites
with transverse and longitudinal macrocracks (Kashtalyan and Soutis, 1999, 2000).
Tables 1 and 2 show reduction of the longitudinal modulus in glass/epoxy cross-ply
[0=90]s and [0=903 ]s laminates damaged by transverse cracks only and transverse and
longitudinal cracks, respectively. Predictions based on the ECM (Kashtalyan and
Soutis, 2000) are compared to those of Hashin (1987), who applied variational
method to stress analysis and considered a repeated laminate element defined by
intersecting pairs of transverse and longitudinal cracks. The results are presented in
Table 1. Normalised longitudinal modulus for glass/epoxy [0=90]s and [0=903 ]s laminates with
transverse macrocracks only.
[0=90]s [0=903 ]s
Dmc
2 (Dmc
1 0) Hashin (1987) ECM Hashin (1987) ECM
0.02 0.990 0.992 0.980 0.985

0.05 0.976 0.981 0.951 0.963
0.1 0.953 0.963 0.907 0.929
0.2 0.910 0.928 0.830 0.867
0.33 0.859 0.889 0.745 0.799
0.5 0.813 0.851 0.661 0.727
1.0 0.775 0.801 0.548 0.613
2.0 0.770 0.795 0.524 0.553
Table 2. Normalised longitudinal modulus of glass/epoxy [0=90]s and [0=903 ]s laminates with
transverse and longitudinal macrocracks.
[0=90]s [0=903 ]s
Dmc
1 Dmc
2 Hashin (1987) ECM Hashin (1987) ECM
0.02 0.990 0.992 0.980 0.985

0.05 0.975 0.980 0.951 0.962
0.1 0.951 0.961 0.906 0.927
0.2 0.907 0.925 0.829 0.863
0.33 0.853 0.882 0.743 0.792
0.5 0.804 0.841 0.658 0.718
1.0 0.762 0.787 0.542 0.602
2.0 0.757 0.765 0.516 0.541
terms of the longitudinal modulus of the damaged laminate, normalised by its value
for the intact laminate. Damage parameters Dmc mc
1 , D2 vary from 0.02 to 2.0. This
corresponds to the variation in longitudinal crack density C1 ¼ (2s1 ) 1 from 0.5 to
5.0 cracks per millimetre in 08 layers of both laminates. Corresponding variation in
the density of transverse cracks C2 ¼ (2s2 ) 1 is from 0.5 to 5.0 cracks/mm for a
[0=90]s laminate and from 0 16 to 1.6 cracks/mm for a [0=903 ]s laminate. Predictions
of both models are found to be in good agreement.
Table 3 shows comparison of predictions for the shear modulus reduction in
graphite/epoxy [03 =903 ]s laminate with transverse and longitudinal macrocracks. The
results are presented in terms of the shear modulus of the damaged laminate,
normalised by its value for the intact laminate. In most of cases, predictions based
on Eq. (15) are within 10% of those obtained from Eq. (14). However, in some cases
the difference can be as big as 20%.
Figure 3 shows stiffness reduction in a [0=90]s SiC/CAS laminate with transverse
cracks in the 908 layer without any damage in the 08 plies. The properties of a
unidirectional material used in the calculations are again taken from (Daniel and
Ishai, 1994). In Fig. 3, stiffness properties of the damaged laminate are normalised by
their values for the intact laminate and plotted as a function of transverse crack
density C2 , which varies from 0 (no damage) to 50 cracks/cm. The maximum crack
Table 3. Normalised shear modulus of carbon/epoxy [03 =903 ]s laminates with transverse and
longitudinal macrocracks.
Tsai and Daniel (1992), ECM model,

Dmc
1 Dmc
2 Eq. (15) Eq. (14)
0. 0.788 0.805
0.2 0.711 0.706
0.39 0.4 0.647 0.624
0.6 0.596 0.557
0.8 0.557 0.503
0. 0.692 0.713
0.2 0.631 0.613
0.66 0.4 0.581 0.532
0.6 0.540 0.465
0.8 0.508 0.410
Figure 3. Normalised stiffness properties of a [0=90]s SiC/CAS cross ply damaged by transverse macro
cracks as function of transverse cracks density C2 (no longitudinal cracks).
density corresponds to crack spacing of 2s2 ¼ 0:2 mm, or one ply thickness. Minimal
crack spacing observed in CMC cross-ply laminates under uniaxial loading is com-
parable to the layer thickness (Daniel and Anastassopoulos, 1995).
It may be seen from Fig. 3 that, as transverse crack density increases, all stiffness
properties of the laminate are significantly reduced. Longitudinal and transverse
moduli of the undamaged laminate, calculated from the classical lamination theory,
are 166.5 GPa, shear modulus 44 GPa, Poisson’s ratio 0.19. When transverse crack-
ing in the 908 layer reaches saturation, the laminate longitudinal and shear moduli
are predicted to lose more than 45% of their value. Inclusion of tensile residual
stresses into the analysis would lead to even more significant reduction in the
longitudinal modulus and Poisson’s ratio, but reduction in shear modulus would
remain the same. Predictions for a [0=902 ]s SiC/CAS laminate are shown in Fig. 4.
Stiffness reduction in a [0=90]s SiC/CAS laminate with macrocracks both in 908
and 08 layers is shown in Fig. 5. Crack densities of transverse and longitudinal cracks
are assumed to be equal C1 ¼ C2 . Laminate stiffness properties are normalised by
their values for the intact laminate and plotted as a function of the transverse crack
density C2 . As one would expect, the longitudinal modulus is not affected by the
presence of longitudinal macrocracks. However, the shear modulus and Poisson’s
ratio undergo further reduction as the density of longitudinal cracks increases. The
most significant reduction due to longitudinal cracks is observed for the shear
modulus. At C2 ¼ 20 cracks/cm its values for the laminate with (C1 ¼ C2 ) and
without (C1 ¼ 0) longitudinal cracks are 27.8 GPa and 17.1 GPa, respectively,
while the shear modulus of the undamaged laminate was 44 GPa.
Predictions of stiffness reduction due to transverse and longitudinal macro-
cracks in a [0=902 ]s SiC/CAS laminate are shown in Fig. 6. Laminate stiffness
properties are normalised by their value for the intact laminate and plotted as a
function of the longitudinal crack density C1 . Transverse crack density is C2 ¼ 10
cracks/cm. As longitudinal cracks initiate, and their density increases, the transverse
Figure 4. Normalised stiffness properties of a [0=902 ]s SiC/CAS cross ply damaged by transverse macro
cracks as function of transverse cracks density C2 (no longitudinal cracks).
Figure 5. Normalised stiffness properties of a [0/90], SiC/CAS cross ply damaged by transverse and
longitudinal macrocracks as function of transverse crack density C2 . Longitudinal crack density C1 C2 .
and shear modulus and Poisson’s ration undergo further reduction, but the longitu-
dinal modulus remains unchanged.
6. CONCLUSIONS
In this paper, the Equivalent Constraint Model (ECM) of the damaged lamina was
applied to predict theoretically stiffness-damage behaviour of cross-ply [0m =90n ]s
Figure 6. Normalised stiffness properties of a [0=902 ]2 SiC/CAS cross ply damaged by transverse and
longitudinal macrocracks as function of longitudinal crack density C1 . Transverse crack density C2 10
cracks/cm.
CMC laminates with macrocracks both in the 908 and 08 plies. To capture simultan-
eous accumulation of damage both in the 908 and the 08 plies, two ECM laminates
were analysed simultaneously, as a coupled problem, instead of the original one.
Following analysis of stresses in the explicitly damaged layer(s) of ECM laminates,
closed form expressions for the reduced stiffness properties of the damaged laminate
were derived representing them as functions of crack densities in the 908 and 08 plies.
Residual thermal stresses were neglected in the stress analysis, but their value was
estimated from the classical lamination theory.
Numerical results for SiC/CAS cross-ply laminates of different lay-ups have
shown that transverse and longitudinal cracks cause significant reduction of all
laminate stiffness properties. Reduction in the longitudinal modulus occurs mainly
due to transverse macrocracks, while the shear modulus appears to be the most
affected by the presence of longitudinal macrocracks.
REFERENCES
Birman, V., and Byrd, L. W., 2000, Review of fracture and fatigue in ceramic matrix composites, Applied
Mechanics Reviews 53:147.
Birman, V., and Byrd, L. W., 2001, Matrix cracking in transverse layers of cross ply beams subjected to
bending and its effect on vibration frequencies, Composites Part B: Engineering 32:47.
Birman, V., and Byrd, L. W., 2003, Effect of matrix cracking in cross ply ceramic matrix composite beams
on their mechanical properties and natural frequencies, International Journal of Non Linear Mechanics
38:201.
Daniel, I. M., and Ishai, O., 1994, Engineering Mechanics of Composite Materials, Oxford University Press,
New York.
Daniel, I. M., and Anastassopoulos, G., 1995, Failure mechanisms and damage evolution in crossply
ceramic matrix composites, International Journal of Solids and Structures 32:341.
Erdman, D. L., and Weitsman, Y., 1998, The multi fracture response of cross ply ceramic composites,
International Journal of Solids and Structures 35:5051.
Fan, J., and Zhang, J., 1993, In situ damage evolution and micro/macro transition for laminated compos
ites’’, Composites Science and Technology 47:107.
Hashin, Z., 1987, Analysis of orthogonally cracked laminates under tension, Transactions ASME Journal
of Applied Mechanics 54:872.
Henaff Gardin, C., Lafarie Frenot, M.C., and Gamby, D., 1996a, Doubly period matrix cracking in
composite laminates Part 1: General in plane loading, Composite Structures 36:113.
Henaff Gardin, C., Lafarie Frenot, M.C., and Gamby, D., 1996b, Doubly period matrix cracking in
composite laminates Part 2: Thermal biaxial loading, Composite Structures 36:131.
Kashtalyan, M., and Soutis, C., 1999, Application of the Equivalent Constraint Model to investigate
stiffness properties of transversally cracked and split FRP laminates, Advanced Composites Letters
8:205.
Kashtalyan, M., and Soutis, C., 2000, Stiffness degradation in cross ply laminates damaged by transverse
cracking and splitting, Composites Part A: Applied Science and Manufacturing 31:335.
Lu, T. J., and Hutchinson, J. W., 1995, Thermal conductivity and expansion of cross ply composite with
matrix cracks, Journal of the Mechanics and Physics of Solids 43:1175.
Pryce, A. W., and Smith, P. A., 1994, Matrix cracking in crossply ceramic matrix composites under static
and fatigue loading, Acta Metallica et Materialia 42:861.
Tsai, C. L., and Daniel, I. M., 1992, Behavior of cracked cross ply composite laminate under shear loading,
International Journal of Solids and Structures 29:3251 3267.
Yasmin, A., and Bowen, P., 2002, Fracture behaviour of cross ply Nicalon/CAS II glass ceramic matrix
composite laminate at room and elevated temperatures, Composites Part A: Applied Science and
Manufacturing 33:1209.
NANOINDENTATION,
NANOSCRATCH AND
NANOIMPACT TESTING OF
SILICON-BASED MATERIALS
WITH NANOSTRUCTURED
SURFACES
Jinjun Lu, Tao Xu, Qunji Xue, and Ben D. Beake*
1. INTRODUCTION
In recent years, nanomaterials receive extensive attention due to their unique prop-
erties. Nano-structured Si-based materials are considered as one of the key material
issues in microelectromechanical systems (MEMS).1 Many techniques have been
employed to fabricate Si-based materials with nano-structured surface.2–3 Nanograin
polysilicon ultra-thin film can be deposited on p-type silicon wafers by chemical
vapor deposition (CVD).2 Ion-beam sputtering (IBS) at low energy (typically
1.2 keV) has been used to fabricate nanodots on silicon wafers, which could meet
the key requirement for the development of future optoelectronic and electronic
devices.3 For the tribological consideration however, the bonding between nanograin
polysilicon ultra-thin film by CVD and its substrate might not strong enough.
Therefore, techniques like IBS should be studied to eliminate the interface between
the film and the substrate. Micromachine fabrication using ion implantation has the
*
Jinjun Lu, Tao Xu, Qunji Xue, State Key Laboratory of Solid Lubrication, Lanzhou Institute of
Chemical Physics, Chinese Academy of Sciences, Lanzhou, P. R. China 730000. Ben D. Beake, Micro
Materials Ltd. Unit 3, The Byre, Wrexham Technology Park, Wrexham, LL13 7YP, UK
43
44 Jinjun Lu et al.
advantage of good controllability of material properties.4 Compared with IBS, ion

implantation involves two procedures, implantation and sputtering. The most im-
portant thing is that implanting reactive species can modify the composition of the
implanted layer, which is beneficial for the mechanical and tribological performance.
Therefore, ion implantation is chosen as the means of fabrication.
The aims of our research are: 1) to explore the possibility of fabricating nano-
structured surfaces on single crystal Si and SiO2 by ion implantation and 2) to study
the nanomechanical and nanotribological behaviors of the ion beam modified layers
by nanoindentation, nanoscratch and nanoimpact testing. We found that nano-
structured surfaces can be obtained by N ion implantation at doses of 5 1016 and
1 1017 ions=cm2 , respectively.5 In the present paper, the nanomechanical and
nanotribological behaviors of N ion implanted single crystal SiO2 and Si were studied
using nanoindentation, nanoscratch and nanoimpact testing.
N ion implanted single crystal Si and SiO2 wafer with doses of 5 1015 ,
1 1016 , 5 1016 , 1 1017 and 5 1017 ions=cm2 respectively was prepared using
an ion implanter at 100 keV. Transmission electron microscopy (TEM) was used to
study the microstructure of N ion implanted single crystal SiO2 wafer.
The NanoTest system (Micro Materials, Wrexham, UK) was used for nanoin-
dentation and nanoscratch testing of ion implanted single crystal silicon and nitrogen
ion implanted single crystal SiO2 . For the nanoscratch testing a conical diamond with
25 mm end radius was used throughout. The scratch experiments involved three
sequential scans over the same 100 mm track all at 1 mm=s scan speed. In the first
topography scan the applied load was kept at 0.2 mN (no wear occurs at this load).
In the second (scratch) scan, the applied load was kept at 0.2 mN between 0---20 mm
and then was ramped at a constant rate of 1.77 mN/s so that the maximum load
reached at 100 mm was 142.5 mN. In the final scan the resultant topography was
observed by using a low applied load of 0.2 mN.
For the nanoindentation a Berkovich (3-sided pyramidal) indenter was used
throughout. Repeat indentations to 200 mN were performed with loading and
unloading rates set equal at 4.04 mN/s and an 0.05 mN initial load was applied.
There was a small (5s) dwell time at maximum load to allow for creep before
unloading, since continuing plastic deformation during unloading can distort the
shape of the unloading curve and lead to inaccurate values of hardness and modulus.
The aim of the nanoimpact testing was to investigate the impact wear-resistance
of the ion-implanted Si and SiO2 samples using the Impact Module of the Micro
Materials NanoTest system. The pendulum impulse technique was used. A solenoid
connected to a timed relay was used to produce the probe impacts on the surface,
as shown schematically Figure 1 (note: magnetic rather than mechanical means
was used to actuate the solenoid in this work). The probe was accelerated from
13 microns from the surface for each impact.
The samples were mounted on a standard NanoTest sample stub using super-
glue. They were cleaned with an ‘‘Air Duster’’ and thermally equilibrated in the
NanoTest environmental cabinet before the nano-impact testing. Two different types
of impact tests were set up (Table 1).
sample in contact
with test probe
solenoid connected
to timed relay
and power supply
minature
stage
Ready for impact After impact
pendulum impulse impact testing
Figure 1. Illustration of nanoimpact testing.
Table 1. Description of two different types of impact tests.
Mode Test probe Impulse load Test duration
Preliminary testing with 25 mm spherical diamond (nominal radius 5.0 and 600 s
radius spherical probe 25 mm radius) 8.0 mN
Inducing fracture with sharp cube corner (90 degree) diamond 2.0 mN 150 s
cube corner indenter indenter
Impulse control cycle: repeat application of the impact load off for 2 s and then on for 2 s
This cycle corresponds to 1 impact every four seconds, with the probe spending 2 seconds on the sample surface between
successive impacts. Note that in all tests the sample was not moved during impacting so that the probe always impacts in
the same location on the surface.
3.1. Nanoscratch Testing on N Ion Implanted Si and SiO2

3.1.1. N ion Implanted Single Crystal SiO2
The critical load for the transition between plastic deformation and fracture failure
has been estimated from the inflexion point in the on-load scratch trace (scan 2). This
correlates well to the onset of cracking as seen in the final topography scan.
Samples 3 and 4 (5 1016 and 1 1017 ions=cm2 ) shows much improved per-
formance over the unimplanted SiO2 . Of the two, sample 3 has the higher critical
load (Lc 57.2 mN) for fracture failure. It is clear that ion-implantation does not affect
the sample roughness until an ion dose of 1 1017 ions=cm2 is reached. It is possible
therefore that the slight decrease in critical load on sample 4 compared to sample 3 is
due to the adverse influence of increasing roughness on the resistance to nanoscratch-
ing wear.
In addition, the nanoscratch experiments also clearly reveal the applied load
necessary for the transition between elastic and plastic behavior during scratching.
On unimplanted SiO2 there is a long period of elastic behavior (intuitively we expect
a single crystal to behave more elastically in compression than the damaged material)
before the deformation becomes plastic in nature. This is then followed by a short
period of plasticity before fracture (brittle) failure. The behavior on implanted
46 Jinjun Lu et al.
Table 2. Variation in Lc for plastic deformation to fracture failure transition from inflexion
in displacement vs. depth curve, surface roughness, and on load depth reached in scratch test
at maximum load (142.5 mN)
Implanting doses, N=cm2 Average Lc (mN) Roughness(Ra) nm Mean final depth/nm*
Sample 0, 0 42.7 3.09 1085 50

Sample 1, 5 1015 39.7 3.50 867 4
Sample 2, 1 1016 45.8 3.47 830 5
Sample 3, 5 1016 57.2 3.49 833 8
Sample 4, 1 1017 51.5 6.58; 8.40; 8.02 877 8
Sample 5, 5 1017 74.4; 67.7; 59.3 2150 200
Mean values shown 4 6 determinations. ‘‘ ’’ inflexion point not clear on this rough sample.
Ra values are over central region (33:3--66:6 mm) of a levelled 100 mm region of surface before scratch testing.
*
‘‘’’ represents the standard error in the mean, after leveling on the initial 0--20 mm (unworn) region of the track.
samples is different the critical load for the elastic-plastic transition is much lower,
and this is followed by a longer period of plastic behavior before brittle failure occurs
at higher applied load [as the polycrystallites can move more freely under shear].
Typically the elastic-plastic transition was at 27 mN on the implanted SiO2 and 6
12 mN on the implanted samples. The final on-load scratch depth is a composite
response of the sum of elastic þ plastic þ fracture deformation which occurs during
the scratch test, ( a measure of the resistance to scratching wear). It decreases with
increasing ion dosage. The high value on the unimplanted sample is due in part to the
increased elastic deformation on this material. The closeness of the values on the ion-
implanted samples 1 4 shows that at these doses the implantation produces a
material whose mechanical properties do not across the surface (i.e. suggestive of a
well-controlled implantation process).
3.1.2 N Ion Implanted Single Crystal Si
In contrast to what is observed on SiO2 , table 3 below shows that N ion implantation
actually decreases the critical load required for the onset of fracture failure in silicon.
The on-load scratch traces also show an inflexion at the elastic-plastic transition at
much lower load that correlates with final topography data as well. On sample 5 this
is at 10 mN. There is also catastrophic failure at a higher applied load than the Lc
values in Table 3.
The critical load for catastrophic failure occurs at a higher applied load than the
Lc values for the onset of fracture failure tabulated above. It also decreases markedly
with increasing N ion dosage, except on sample 2, which appears more resistant to
catastrophic failure (Table 3).
Although the on-load depth is lower on sample 4 this does not necessarily
represent improved tribological performance in this case. Examination of the curves
show that during the on-load scratch the probe is pushed out of the material by
cracking (e.g. by uplift at the edge of the scratch track which helps to support the
scratch load) which leads to a lower final depth. This material is pushed up over the
scratch track as observed in the final topography as shown in the figure 2b.
Micro-scale scratching of bulk Si by Bhushan and co-workers (1995) showed
similar behavior to our scratch results on the unimplanted Si wafer.6 Using a much
smaller probe (conical diamond with 1 mm end radius) they observed (from friction
Table 3. Variation in Lc for plastic deformation fracture failure transition from inflexion in
displacement vs. depth curve and on load depth reached in scratch test at maximum load
(142.5 mN).
Implanting doses, N=cm2 Mean Lc =mN Mean final depth/nm*
Sample 0, 0 82.1 4.4 490 8

Sample 1, 5 1015 70.7 5.6 467 5
Sample 2, 1 1016 71.5 6.3 561 7
Sample 3, 5 1016 45.7 0.4 524 5
Sample 4, 1 1017 40.9 4.6 390 6
Sample 5, 5 1017 572 26
Mean values from 4 6 determinations. * inflexion point not clear on this rough sample.
‘‘’’ represents the standard error in the mean, after levelling on the initial 0 -20 mm (unworn) region of the track.
and SEM data but with no accurate depth information) the elastic-plastic transition
occurred at 4mN and ploughing at higher load until the onset of fracture at 16 mN
and catastrophic failure at 18 mN. In view of the differences in contact pressure due
to the different tips used these results are in reasonable agreement with those of this
study. Post-scratch SEM shows that that much material is moved into/on-top of the
scratch track (particulate debris of sub-micron size). This is consistent with the large
increases in probe position in the final topography observed in our experiment.
It is clear from Table 3 that N ion implantation does not result in roughening of
the Si surface at any of the doses used. To generate sufficient damage to get rough-
ening it is necessary to use argon ions of larger cross-sectional area.
3.2. Pop-in and Pop-out Phase Changes of N Ion Implanted Si

Silicon is well-known as having unusual indentation behavior due to phase changes
which occur on loading and unloading. The pop-ins have been associated with
transition from semiconducting diamond cubic structure into metallic B-tin structure
(i.e. a semiconductor-to-metal) at critical pressure. Unloading causes a volume-
expanding reverse phase transformation from B-tin into an amorphous structure.7–8
Since these phase changes occur at critical pressures it is not possible to compare
456.4 393.3
350.0
400.0
300.0
inflexion
350.0 250.0
200.0
300.0 low final depth
150.0
due to volume uplift
250.0 100.0
Depth (nm)
Depth (nm)
onset of 50.0
200.0 fracture failure
0.0
150.0 -50.0
-100.0
100.0 dramatic
-150.0
fracture failure
50.0 -200.0
-250.0
0.0
-300.0
-40.7 -359.7
0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 90.0 99.9 0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 90.0 99.8
Distance (um) Distance (um)
(a) unimplanted surface (b) sample 4 (1 x 1017)

48 Jinjun Lu et al.
Table 4. Pop in and pop out phase changes on ion implanted Si
Implanting doses, N=cm2 Pop in load/mN Pop out/mN
Sample 0, 0 170, 176 115.4, 123.8

Sample 1, 5 1015 178.3, 178.3 122.5, 126.3, 110.5, 111.1, 111.5
Sample 2, 1 1016 # 86.6, 82.4, 60.5
Sample 3, 5 1016 152.2 53.7
Sample 4, 1 1017 148.6 62.6, 57.9, 65.8
Sample 5, 5 1017 # 43.3, 52.2
*
critical load values in repeat tests are shown in italics. Note that the load given is that for the start of the
pop in or out transition. # not clear transition during loading.
exact values with other studies since our indenter geometry will be different (it is
likely to be blunter so that a greater applied load is necessary before the critical
pressure for phase transformation is reached).
The critical pressures on the Si sample implanted with the lowest dose are very
similar to those of the unimplanted Si (Figure 3). At greater pressures the pop-in phase
transformation occurs at lower load (i.e. is made energetically favourable) but the
unloading transformation occurs at lower load (made energetically unfavourable).
It has been shown previously that CNx films retard the unloading phase trans-
formation, i.e. that the coating keeps the Si in the dense metallic form.9 It has also
been shown that amorphous Si does not undergo the same phase transformations on
loading and unloading.10 It appears that the ion implanted samples act as if they were
partially amorphised.
The indentation curves also showed that Sample 5 shows a very soft near-surface
region revealed by shapes of loading curves i.e. plastic flow at low indentation
pressures, as has been seen on amorphous Si.
3.3. Comparison Between Si and SiO2
Various parameters such as the on-load depth and critical load for fracture can be
used to compare between the materials. Nanoindentation shows that unimplanted Si
is harder and less elastic (higher modulus) than SiO2 . The lower critical loads
1095 1105
1000 1000
pop-out event pop-out event
900 900
800 800
700 700
Depth (nm)
Depth (nm)
600 600
500 500
400 400
300 300
200 200
100 100
3 4
0.0 20.0 40.0 60.0 80.0 100.0 120.0 140.0 160.0 180.0 200.3 0.0 20.0 40.0 60.0 80.0 100.0 120.0 140.0 160.0 180.0 200.5
Load (mN) Load (mN)
Figure 3. Pop outs on sample 1 (5 10 ) and sample 4 (1 1017 ). 15

observed on SiO2 and greater maximum on-load scratch depth (by a factor of 2 3)
correlate with its lower hardness.
The H/E ratio is a key parameter determining the type of behaviour observed in
nanoindentation and nanoscratching wear. Si has a lower H/E ratio than SiO2 which
explains why the transition from elastic to plastic behaviour occurs at higher contact
pressure on this material.
It is interesting to speculate why ion implantation does not lead to a similar
obvious improvement in the nanotribological performance of Si as it does on SiO2 .
Being harder (and having a more closed crystal structure?) the Si may not be damaged
sufficiently by N ions at these doses to reach the fully-nanocrystalline state. Support-
ing evidence is (1) pop-ins and pop-outs are still observed during nanoindentation
(even if the critical pressures are changed) amorphous Si does not do this (2) even the
highest N dose does not lead to any roughening of the surface.
Bhushan (2001) has shown that Cþ ion implantation results in a hardening of
the surface layer which results in improved microscratching wear performance. With
N ion implantation do the competing mechanisms of 1) softening due to destruction
of the matrix 2) hardening due to the formation of Si3 N4 (or similar) phases, act
against each other resulting in no improvement in tribological properties?
3.4. Nanoimpact Testing of N Ion Implanted Single Crystal Si and SiO2
3.4.1. Preliminary Testing with 25 mm Radius Spherical Probe

It is not possible to use nanoindentation methods to obtain information about the
impact fracture and erosive wear performance. High hardness can confer improved
wear resistance but it is often associated with brittleness and poor toughness and
erosive wear resistance.

The preliminary impact test results are shown in Figure 4a. The plots show the
evolution of impact-induced damage with time (i.e. plastic deformation, fatigue
wear and fracture) throughout the test. The height of the surface is recorded with
sub-nm precision when the probe is on the surface.
c:\Lanzhou\si wafer S2c

600
600
575
550 550 2, 4 - discrete failures
500 525
450
500 1, 3 - slow crack growth
475
400 450
350 425
Depth (nm)
400
300
375
Depth (nm)
250
350 4
200 325
300
150
275
3
100 250
50
Cursor Position
225
2
0
Depth (nm) 382 08 200 1
02 20 0 40 0 60 0 80 0 100 0 120 0 140 0 160 0 180 0 200 0 220 0 250 0 Time (s) 120 2 175
Time (s) 150
125
T me (s) Depth Ax s (nm) Scan Type Multiple Pendulum Impulse Exit 100
Gr d Auto scale Gr d Auto scale
Repeat number 1 Previous Next >>
75
Maximum 250 0 Maximum 600 00
Inve t 50
Minimum 0 2 Minimum 0 Pulse number 1 Previous Next >> 02 20 0 40 0 60 0 80 0 100 0 120 0 140 0 160 0 180 0 200 0
Time (s)
(a) (b)
50 Jinjun Lu et al.
The results show typical impact behavior for a brittle material. There are long
periods of damage accumulation (‘‘plateau periods’’) where small cracks grow sub-
surface and the surface position changes relatively little with time. These are separ-
ated by more rapid events where material is removed as the cracks propagate and
coalesce leading to these failures.
The instrument software allows for ‘‘zooming-in’’ on features of interest such as
these failure events. Figure 4b shows zoom to the beginning of the tests. The plot below
is very interesting as it shows volume uplift (or surface bulging) just before fractures.
This ‘‘volume uplift’’ is due to the formation of micro- and nano-cracks sub-
surface in the material. Because the damaged material is constrained from breaking
away from the surface by the (less damaged) material above it, the cracks cause the
material to swell up forcing the probe ‘‘out of the surface’’. This has also been seen on
DLC coatings and other brittle materials.
The behavior in extended impact tests (i.e. at longer impact times) shows a
similar trend to production of sub-surface damage and surface bulging before surface
failure etc.
As well as these very clear multiple fractures another common type of behavior
is illustrated below.
1125
Si wafer S3
1000
900 sub-surface cracks

push probe out
800
just before the
700 catastrophic failures
Depth (nm)
600
500
400
300
200
100
0
02 20 0 40 0 60 0 80 0 100 0 120 0 140 0 160 0 180 0 200 0 220 0 240 0 260 0 280 0 300 0
Time (s)
Figure 5. Volume uplift due to the formation of micro and nano cracks sub surface in the material.
400
380 Si wafer S1
360
340 short plateau period
320
300 nano-cracks
280 swelling surface
260
240
Depth (nm)
220
200
180
160
140
120
100
80
60
40
20
0
02 50 0 100 0 150 0 200 0 250 0 300 0 350 0 400 0 450 0 500 0 550 0 600 0
Time (s)
Figure 6. Multiple fractures.

There is a very short plateau period followed by 1 or more small cracking events
(the surface bulges before the cracks coalesce, and the process repeats). The first
inflexion between the flat plateau and faster wear/cracking occurs after ~30 50 s, (at
greater times for higher doses). The depth at this transition is (180 nm 20) for all
the samples.

More common under these conditions than sharp fracture is the type of behaviors
shown below.
The probe penetrates ~70 nm into the surface under the static of load of 8 mN.
The first impact results in further 80 nm depth penetration (plastic deformation).
There is then a plateau period where the probe position levels off (i.e. some plastic
deformation occurs but surface not being worn away during this period, erosive wear
rate is very low). During this period the plastic deformation results in the initiation
and development of cracks underneath the contact surface. At ~54 s (shown by the
cursor in the plot) the depth begins to increase again (ie rate of wear increases
sharply). The time-at-inflexion is likely to be related to the fracture resistance, since
longer times presumably reflect increased barrier to coalescence/propagation of sub-
surface nano-/micro-cracks. These times vary on the SiO2 samples as shown below.
Results at 5 mN (squares) and 8 mN (triangles) show a trend to increasing time-
to-inflexion with the ion dose, reaching a maximum at dose 1 1016 ions=cm2 . The
inflexions are small.
From the results with the blunt spherical probe we can see that impact testing is
a statistical process, since it involves inducing plastic deformation, crack initiation
and coalescence/fracture [i.e. fatigue]. Several tests are necessary to get a complete
picture of the range of behaviour observed, as is common in fatigue tests.
c:\Lanzhou\s3 rep8mn
250
225
200
175
150
Depth (nm)
125
100
75
50
25
Cursor Position ...
Depth (nm) 161.56

0
02 20 0 40 0 60 0 80 0 100 0 120 0 140 0 160 0 180 0 200 0 220 0 240 0 260 0 280 0 300 0
Time (s) 54.3
Time (s)
Time (s) Depth Axis (nm) Scan Type Multiple Pendulum Impulse Exit
Grid Auto scale Grid Auto scale
Maximum 300 0 Maximum 250 00 Repeat number 1 Previous Next >>

Invert
Minimum 0 2 Minimum 0 Pulse number 1 Previous Next >>
Figure 7. Sharp fracture in N ion implanted single crystal SiO2 .

52 Jinjun Lu et al.
60
50
40
Time (s)
30
20
10
0
0 1 2 3 4
SiO2 sample nos
Figure 8. Variation in time to first failure (s) with sample.
Clearly the wear rate is not constant through the test and detailed examination
of the results allows individual failure events (shown by the changes in slope and
abrupt changes in depth) to be investigated.
There was evidence that the time at which the increase in wear rate occurs
correlated with ion dose for both the Si and SiO2 implanted samples (generally
greater dose ¼ longer time-to-failure).
3.4.2 Inducing Fracture with Sharp Cube Corner Indenter
A cube corner indenter was used to test these observations and to more easily induce
fractures. The cube corner indenter is sharper (higher contact pressure) and the
90 degree angle means that the instantaneous contact strain under the probe on
impact is much greater than for the spherical probe.
Failures are generally less sharp than on SiO2 . The fracture resistance of the samples
can be compared by: (1) wear rate at end of test, (2) depth at end of test, (3) presence or
absence of clear surface failure/fracture events, and (4) the time at which first failure/
fracture occurs. The type of behavior observed falls into distinct groups (Table 5).
The fracture toughness of sample 4, (i.e. resistance to propagation of the cracks
which would cause failure and removal of material) is high. This is seen in the final
impact depth data.
These impact depths have been compared with the final on-load scratch depth
data in the graph below (impact ¼ circles, scratch ¼ filled squares).
Table 5. Classified group on ion implanted Si.
Group Pop in load/mN
Sample 0, 0 Low wear rate, with little clear fracture

Sample 1, 2 and 5 All are very susceptible to surface damage, cracking/fracture (especially at short
times) and also show a high wear rate at 150 s. Short periods of surface bulging/
swelling are often followed by cracking, and fast wear rate
Sample 3 generally lower wear rate, with some failures at relatively short times
Sample 4 shows occasional small failures in some tests but mostly they are absent. The wear
rate soon drops to near zero and the predominant process is sub surface cracking
causing the slight bulging/swelling of the surface
1500 1500
1400 1400
1300 1300
1200 1200
1100 1100
1000 1000
900 900
Depth (nm)
Depth (nm)
800 800
700 700
600 600
500 500
400 400
300 300
200 200
100 100
0 0
02 10 0 20 0 30 0 40 0 50 0 60 0 70 0 80 0 90 0 100 0 110 0 120 0 130 0 140 0 150 0 02 10 0 20 0 30 0 40 0 50 0 60 0 70 0 80 0 90 0 100 0 110 0 120 0 130 0 140 0 150 0
Times (s) Times (s)
Si 0; unimplanted Si Si2
1500 1500
1400 1400
1300 1300
1200 1200
1100 1100
1000 1000
900 900
Depth (nm)
Depth (nm)
800 800
700 700
600 600
500 500
400 400
300 300
200 200
100 100
0 0
02 10 0 20 0 30 0 40 0 50 0 60 0 70 0 80 0 90 0 100 0 110 0 120 0 130 0 140 0 150 0 02 10 0 20 0 30 0 40 0 50 0 60 0 70 0 80 0 90 0 100 0 110 0 120 0 130 0 140 0 150 0
Times (s) Times (s)
Si3 Si4
Figure 9. Typical impact behavior of N ion implanted Si.
The correlation between the impact and scratch data is striking. A much higher
ion dose is necessary for tribological improvement on the Si than on the SiO2 .

Failure events on the SiO2 samples were clearer (greater change in depth) than on the
Si samples. Final depths were also larger which mimics the difference observed in the
scratch test data reported previously.
There was some variation in the time at which the first failure occurs and its size.
Although there was some variability (e.g. the samples did not always fracture within
1200 700
1000 600
scratch depth (nm)
impact depth (nm)
500
800
400
600
300
400
200
200 100
0 0
0 1 2 3 4 5
Si sample nos
Figure 10. Correlation between impact and scratch data for Si samples.
54 Jinjun Lu et al.
Sample 0, 0 823 30
Sample 1, 5 1015 939 48
Sample 2, 1 1016 1013 72
Sample 3, 5 1016 838 62
Sample 4, 1 1017 646 42
Sample 5, 5 1017 1067 27
surface.
150 s), 5 10 repeats for each sample were carried out so that clear trends in behavior
could be deduced.
Illustrative results on each of the samples are shown in the following graphs. S3
and s4 generally show much longer time-to-failure. The failure is more extreme on s4
and hence this sample shows the largest depth at 150 s.
We have previously impact tested a variety of hard coatings [DLC, TiN, alu-
mina etc]. The abrupt failures on S4 look quite similar to adhesion failures i.e. a
weak interface/interphase (between the ion implanted and unimplanted regions)
eventually gives way catastrophically. It may be that the final depth alone cannot
be used as a direct measure of fracture toughness for these SiO2 samples.
4. CONCLUSIONS
The tribological properties of N ion implanted single crystal Si and single crystal SiO2
surfaces have been evaluated by nanoscratch and nano-impact testing. Nanoscratch
data show the critical loads associated with (1) the elastic-to-plastic transition and (2)
the plastic deformation-to-fracture failure transition. The scratch results show that at
moderate ion doses the implantation process improves the nanotribological properties
of single crystal SiO2 , with similar ion dose of 1 1016 1 1017 N cm 2 also confer-
ring improved resistance to impact/fatigue wear. For the Si samples, the correlation
between the impact data and the on-load scratch data was striking. An ion dose of
5 1017 N cm 2 was necessary for improved resistance to impact/fatigue wear. Impact
wear appears to involve 3 processes (1) plastic deformation (2) formation of sub-surface
cracks and (3) coalescence of these cracks and failure (by material removal from contact
surface).
Table 7. Description of failure.

S0 generally short time to failure
S1 short time to first failure, multiple failures
S2 short time to first failure
S3 longer time to failure, which is often only minor
S4 quite long time to dramatic failure
S5 were not tested since previous tests under less strong conditions showed very poor performance on this sample the
near-surface layer is too weak that the impact probe punches straight through. Time-to-failure data show an increasing
trend S3>S4>S2~S1~S0
c:\Lanshou\s0 rep2mnx5cubeb c:\Lanshou\s0 rep2mnx5cubeb
2500 2500
2250 2250
2000 2000
2750 2750
2500 2500
Depth (nm)
Depth (nm)
2250 2250
1000 1000
750 750
500 500
Cursor Position Cursor Posit on
250 250
Depth (nm) 250 00 Depth (nm) 1250 00
0 0
0 2 10 0 20 0 30 0 40 0 50 0 60 0 70 0 80 0 90 0 100 0110 0120 0130 0140 0150 0 Time (s) 74 9 0 2 10 0 20 0 30 0 40 0 50 0 60 0 70 0 80 0 90 0 100 0110 0120 0130 0140 0150 0 Time (s) 74 9
Time (s) Time (s)
Time (s) Depth Axis (nm) Scan Type Mult ple Pendulum Impulse Ex t T me (s) Depth Ax s (nm) Scan Type Multiple Pendulum Impulse Exit
G id Auto scale Gr d Auto scale Grid Auto scale G id Auto scale
Max mum 150 0 Max mum 2500 00
Repeat number 1 Privious Next >> Max mum 150 0 Maximum 2500 00 Repeat number 1 Privious Next >>
Inve t Invert
Minimum 0 2 Min mum 0 Pulse number 1 Privious Next >> Minimum 0 2 M n mum 0 Pulse number 1 Privious Next >>
Figure 11. Typical impact behavior of N ion implanted SiO2 .
Sample 0, 0 1282 124

Sample 1, 5 1015 1315 203
Sample 2, 1 1016 1259 138
Sample 3, 5 1016 1138 95
Sample 4, 1 1017 1590 165
56 Jinjun Lu et al.
ACKNOWLEDGEMENT
The project is supported by Chinese Natural Science Foundation (No.50172052 and

No. 50023001).
REFERENCES
1. S. M. Spearing. Materials issues in microelectromechanical systems (MEMS), Acta Mater. 48,
179 196(2000).
2. S. Ecoffey, D. Bouvet, A. M. Ionescu, and P. Fazan. Low pressure chemical vapour deposition of
nanograin polysilicon ultra thin films, Nanotechnology 13, 290 293(2002).
3. R. Gago, L. Vazquez1, R. Cuerno, M Varela, C Ballesteros, and J M Albella. Nanopatterning of
silicon surfaces by low energy ion beam sputtering: dependence on the angle of ion incidence,
Nanotechnology 13, 304 308(2002).
4. S. Nakano, H. Ogiso, A. Yabe. Advanced micromachine fabrication using ion implanted layers,
Nucl. Instr. and Meth. in Physics Research B 155, 79 84(1999).
5. T Xu, J. J. Lu, J. Tian, and Q. J. Xue. The effect of nitrogen ion implantation on wear behavior of
single crystal SiO2 , J. of Phys. D: Appl. Phys. 33, 426 429(2000).
6. Bhushan B., Gupta B.K., and Azarian M. H., 1995, Wear 181: 743 758.
7. B. Roos, H. Richter, and J. Wollweber, Solid State Phenomena 47 48, 509 516(1996).
8. T. F. Page, W. C. Oliver and C. J. McHargue, J. Mat. Res. 7, 450(1992).
9. S. Bull, Applications of nanoindentation to industrial problems. Nanomechanical testing, VITO,
Belgium 5th Dec 2001
10. J. S. Williams, et al. Ultra microindentation of silicon and compound semiconductors with spherical
indenters, MRS 1999.
11. Bhushan B. Tribology on the macroscale to nanoscale of microelectromechanical system materials: a
review, Proc Instn Mech Engrs 215, 1 18(2001).
12. B.D. Beake, S.R. Goodes and J.F. Smith, Micro Impact testing: A new technique for investigating
thin film toughness, adhesion, erosive wear resistance and dynamic hardness, Surface Engineering 17,
187 193(2001).
13. B.D. Beake, M.J. Ibanez Garcia and J.F. Smith, Micro impact testing: a new technique for investi
gating fracture toughness, Thin Solid Films 398 399, 438 443(2001).
14. B.D. Beake, S.R. Goodes, J.F. Smith, A. Zhang and J. E. Micro impact testing: A novel nano /micro
tribological tool for assessing the performance of wear resistant coatings under impact/fatigue
conditions, Science in China (Series A) 44 Supp. 418 422 (Proc. 3rd Int. Symp. Tribology, Beijing,
Sept 2001)
NONLINEAR FRACTURE
OF A POLYCRYSTALLINE
GRAPHITE – SIZE-EFFECT
LAW AND IRWIN’S
SIMILARITY
M. Sakai and R. Nonoyama
ABSTRACT
Nonlinear fracture of a polycrystalline graphite is examined through the studies for the
test specimen size-effect on fracture toughness parameters. The fracture toughness of
polycrystalline graphite significantly depends on the dimensions and the geometries of
test specimens. A finite non-negligible frontal process zone dictates this size-effect law.
The Bažant’s theoretical approach modified by the present authors is successfully
applied to the nonlinear fracture of polycrystalline graphite, allowing an estimate of
the size of frontal process zone, as well as the theoretical prediction for the size-effect
law of fracture toughness. The Irwin’s similarity relationship (the equivalence between
the stress-intensity derived toughness, Kc 2 =E 0 and the potential energy-derived tough-
ness Rc ;Kc 2 =E 0 ¼ Rc ) is also studied in the present nonlinear fracture mechanics
regime; Rc is always larger than Kc 2 =E 0 , and the excess energy (Rc Kc 2 =E 0 ) becomes
progressively significant with the increase in the critical load Pc for the onset of main
crack extension, independent of the notch/crack length of test specimen. This fact
suggests that the excess energy, Rc Kc 2 =E 0 , is consumed in the region away of the
Department of Materials Science

Toyohashi University of Technology
Tempaku cho, Toyohashi 441 8580, JAPAN
337
338 M. Sakai et al.
localized area including the main crack and its adjoint frontal process-zone. The
microscopic considerations are made of the characteristic processes and mechanisms
for the excess energy consumption in polycrystalline graphite materials.
1. INTRODUCTION
The standard test methods for the stress-intensity-derived fracture toughness Kc are
very restrictive with respect to the size requirements of test specimen to assure the
plane-strain condition, where the fracture process zone in front of the tip of a critical
crack is extremely small in size compared to the geometrical dimensions of test
specimen (thickness, width, height, and crack/notch length). Linear Elastic Fracture
Mechanics (LEFM) only refers to this mechanical situation of small-size process zone.
The plane-strain fracture toughness Kc thus defined in these standards is independent
of the dimensions of test specimen, being supposed to be a characteristic material
parameter for the resistance to failure. In LEFM-regime, Irwin demonstrated the
similarity between Kc and the potential energy derived fracture toughness Rc through
the relation of Rc ¼ Kc 2 =E 0 in terms of the elastic modulus E 0 [ ¼ E=(1 n2 )] (E:
Young’s modulus, n: Poisson’s ratio). Rc is defined by the potential energy consump-
tion associated with creating a unit fracture surface area at crack-tip [1 3].
Once the size of frontal process zone is not fit to the criterion of LEFM, the stress-
intensity factor KQ (apparent fracture toughness) for the critical crack significantly
depends on the geometries and dimensions of test specimen. Furthermore, a finite
amount of residual offset in the displacement is often observed on unloading in the
load P vs. load-point displacement u diagram, due to the irreversible displacement
enhanced by nonlinear mechanical processes and mechanisms in the frontal process
zone [4 6]. In this non-linear fracture regime, not only the work-of-fracture gwof (total
work for a static failure divided by the total areas of resultant fracture surfaces), but
also the energy fracture toughness Rc do not fulfill the Irwin’s similarity
[Rc ( ¼ 2gwof ) ¼ Kc 2 =E 0 ]. These energy-derived toughness parameters (Rc and 2gwof )
are always larger than the stress-intensity-derived value of Kc 2 =E 0 . Phase-transform-
ation-toughened zirconia ceramics [6], numbers of refractory materials [4,5], and fiber-
reinforced ceramic composites [7], by way of examples, are not subjected to the Irwin’s
similarity.
The test specimen size-effect on Kc and Rc is examined in the present work for a
polycrystalline graphite. In particular, the excess energy consumption, Rc Kc 2 =E 0 ,
will be related to the size of test specimen as well as to the test conditions. It will be
concluded that this excess energy is intimately related to the processes and mechan-
isms of microcracking and slippage of graphite basal planes that occur in the
locations away of the crack-tip process zone.
2.1. Test Material
An isotropic polycrystalline graphite (IG-110U, Toyo Tanso Co., Osaka, Japan) was
obtained for study. The microstructural characteristics of the polycrystalline graphite
Nonlinear fracture of a Polycrystalline Graphite 339
are a mean size of petroleum coke grains of 15 mm, a mean pore size of 5 mm, a
porosity of 15%, and a bulk density of 1:76 g=cm3 . Some mechanical characteristics
include the young’s modulus (E ) of 9.0 GPa, three-point p flexural strength (sf ) of
35 MPa, and the fracture toughness (Kc ) of 0:85 MPa m. This graphite material
shows rising R-curve (crack growth resistance KR vs. crack extension Da curve)
behavior duringp crack extension, yielding the eventual plateau toughness of
Kc ¼ 1:17 MPa m after a crack extension Da of about 3 mm. Grain bridging in
the crack-face contact region gives rise to this rising R-curve behavior [4].
2.2. Fracture Mechanics Tests
Single-edge-notched beam (SENB) specimens were used for fracture mechanics tests.
The geometrical notation is shown in Figs. 1(a) and 1(b). The test specimen geom-
etries and the dimensions employed in the present work were classified into the
following two categories: (1) geometrically similar specimens with different sizes of
S, W, B and a0 , but with the same ratios of a0 =W ¼ 0:5, and S=W ¼ 4:0 [Fig. 1(a)],
and (2) geometrically nonsimilar specimens with fixed dimensions of S ¼ 50:0 mm,
W ¼ 12:5 mm(S=W ¼ 4:0), and L ¼ 60:0 mm, besides different depths of initial
notch a0 (0:1 # a0 =W # 0:9) and different thickness values B [see Fig. 1(b)]. The
notch was machined with a 0.3mm-thick diamond wheel saw, and then extended
by about 1 mm using a very thin saw made from a razor blade, yielding a resultant
notch-tip radius less than 10 mm. It has been confirmed in the previous study that this
tip-acuity is sufficient enough for providing the fracture toughness Kc as a charac-
teristic material parameter [4].
Fracture mechanics tests were conducted on an Instron-type displacement con-
trolling test machine (Sanwa Testing Machine Co., Toyohashi, Japan) under a
crosshead speed of 0.05 mm/min. All the fracture tests were made in three-point
flexural modes. A stable crack extension was observed for all the specimens with the
relative notch depth a0 =W $ 0:4. However, the crack extension was always unstable
for a0 =W < 0:4. The apparent fracture toughness, KQ , at the onset of crack initiation
for the critical load Pc was determined by
Figure 1. Test specimen geometries and dimensions for (a) geometrically similar configuration
(S 10:0 80:0mm, W 2:5 20 mm, S=W 4:0, L S þ 10 in mm, a0 =W 0:5, and B 5:0 and
10:0 mm), and (b) geometrically nonsimilar configuration (S 50:0 mm, W 12:5 mm, S=W 4:0,
L 60:0 mm, a0 =W 0:1 0:9, and B 3:0, 5:0, and 10:0 mm).
340 M. Sakai et al.
Pc
KQ ¼ p Y (a0 =W ) (1)
B W
using the shape factor Y (a0 =W ) of the ASTM E399 formula for SENB geometry [8].
A schematic diagram for determining the potential energy-derived fracture
toughness, Rc , is illustrated in Fig. 2, utilizing two specimens with different initial
notch lengths a1 and a2 (a1 < a2 ).
The area, Dp, of the diagram (the hatched area in Fig. 2) enclosed by their
loading lines (load P vs. load-point displacement u lines) to the respective critical
load Pc indicates the critical energy to create an incremental crack surface
DA[ (a2 a1 )B]. Accordingly, the energy fracture toughness Rc for the onset of
crack extension is defined by the following formula [1 3,5];
Dp
Rc ¼ (2)
DA
A generalization of the concept of energy fracture toughness for linear elastic
materials to that for nonlinear elastic materials leads to the path independent
J-integral. Some details are given in APPENDIX.
In all the test specimens, the critical load Pc for the onset of crack extension is
some-what smaller than the peak load Pmax (see Fig. 3). The crack extension from Pc
to Pmax is always stable. Accordingly, the determination of the critical load Pc is
made in a precise manner by monitoring the notch-tip on the polished side surface of
each specimen through a traveling microscope (50).
In the present fracture tests, the difference of notch depths, a2 a1 , was con-
trolled (1:0 0:1) mm, and then the energy fracture toughness Rc obtained from the
procedures in Fig. 2 was assigned to the toughness of test specimen with the initial
notch depth of a0 ¼ (a1 þ a2 )=2. A precise measurement of load-point displacement u
Figure 2. Procedures for determining Rc by the use of two test specimens with different notch lengths
a1 and a2 .
Figure 3. An example for the load P vs. load point displacement u relation of a polycrystalline graphite.
The arrow indicates the critical load for the onset of crack extension.
is substantial for the Rc -determination in the preceding procedure (Fig. 2). The load-
point displacement u in three-point flexural configuration was determined in the
present work by measuring the relative displacement between the inner loading point
and the outer supporting points of flexure by the use of a twinning eddy-current gap
detector.
3. RESULT
An example of P-u curve during a stable crack extension is shown in Fig. 3

(a0 =W ¼ 0:4, W ¼ 12:5 mm, B ¼ 5:0 mm, and S ¼ 50:0 mm). The crack extension
is very stable. The critical load Pc for the onset of crack extension is marked by the
arrow, indicating that the crack initiates propagation before the load P reaches the
peak value. Furthermore, it must be noticed that the loading P-u line is curved (non-
Hookean) at loads even smaller than the critical Pc value, implying the presence of
inelastic deformation processes in loading [4].
The dependence of the apparent fracture toughness KQ for geometrically similar
test specimens [refer to Fig. 1(a)] is shown in Fig. 4 as a function of the width W. KQ -
values increase pwith the increase in W, eventually reaching a plateau value of
KQ 0:85 MPa m. In contrast to the geometrically similar test specimens, the KQ -
values of geometrically nonsimilar test specimens [refer to Fig. 1(b)], as shown in Fig.
5, decrease with the increase in the relative notch depth, a0 =W . However, in the
extreme of a0p=W # 0, the KQ -values attain the same plateau value of
KQ 0:85 MPa m as that observed in Fig. 4.
The dependence of the energy-derived fracture toughness Rc for geometrically
similar test specimens [Fig. 1(a)] is plotted in Fig. 6 as a function of W, along with the
values of KQ 2 =E 0 for comparison. The similar plots for geometrically nonsimilar test
342 M. Sakai et al.
Figure 4. Dependence of the apparent fracture toughness KQ on the width W of test specimens with
geometrically similar configuration.
Figure 5. Dependence of the apparent fracture toughness KQ on the relative notch depth a0 =W of test
specimens with geometrically nonsimilar configuration.
specimens [Fig. 1(b)] are given in Fig. 7. A careful inspection for Figs. 6 and 7 indicates;
(1) Rc is always larger than KQ 2 =E 0 , (2) both the toughness parameters, KQ2 =E 0 and Rc ,
are significantly dependent on W and a0 =W ; increasing with the increase in W and
with the decrease in a0 =W , and (3) the excess energy, Rc KQ 2 =E 0 , tends to diminish
and then to satisfy the Irwin’s similarity with the increase in a0 =W , and with the
decrease in W.
Figure 6. Dependence of the energy fracture toughness parameters, KQ 2 =E 0 and Rc , on the width W of
test specimens in geometrically similar configuration.
Figure 7. Dependence of the energy fracture toughness parameters, KQ2 =E 0 and Rc , on the relative notch
depth a0 =W in geometrically nonsimilar configuration.
4. DISCUSSION
4.1. Stress Intensity Derived Fracture Toughness KQ -The Size-Effect Law-

The failure strength sc of brittle materials is given by the following general descrip-
tions in a LEFM-regime [9 12]:
Kc
sc ¼ N W p (3a)
W
344 M. Sakai et al.
Kc
sc ¼ N a p (3b)
a0
In these expressions, W is a characteristic length (width) of the test specimen, a0 is a

characteristic length of a crack as a fracture origin, a0 is defined as a0 =W , and both
Y (a0 ) and F (a0 ) are the dimensionless stress intensity
p factors (shape factors) of the
crack. NW is defined by cn =Y (a0 ) and Na by 1=[ pF (a0 )]. The parameter cn is a
characteristic coefficient of the test specimen geometry. Equation (3a) has been
utilized by Bažant, et. al. for the considerations of size-effect law of geometrically
similar test specimens [9,10]. One of the present authors (MS) has applied Eq. (3b) to
the discussions of the size-effect law for geometrically nonsimilar test specimens with
different notch/crack lengths, and then to the discussions of R-curve studies [11,12].
It must be noticed in a fracture mechanics sense that both Eqs. (3a) and (3b) are
equivalent, as pointed out by the present author [11,12].
A ‘‘finite nonnegligible fracture process zone’’ in front of the crack tip gives rise
a deviation from the size-effect law prescribed in Eqs. (3a) and (3b). When the
characteristic dimension of the frontal process zone cf becomes significant compared
to the characteristic dimensions of W and a0 , Eqs. (3a) and (3b) must include the
nonlinear contributions of the microscopic deformation and fracture in the process
zone, and be modified through the replacement of the crack length a0 with its
effective value aeff ( a0 þ cf ). This modification in Eqs. (3a) and (3b) results in the
following formulae of failure strength in their intrinsic expressions;
KQ
tc ¼ p (4a)
w
p
with the relation of KQ ¼ b=(1 þ b)Kc , and
KQ
^tc ¼ p (4b)
a0
q
with the relation of KQ ¼ b ^ =(1 þ b
^ )Kc . In these equations and relationships,
tc ( ^tc ) / sc , a0 / a0 , w / W , b ¼ w=cf , and b ^ ¼ a0 =cf . The details of the defin-
itions and their fracture mechanics meanings were given in the literature [9 12]. The
intrinsic failure strength tc was first introduced by Bažant, et. al [9]. It must be
noticed in Eqs. (4a) and (4b) that expressions of tc and ^tc , the brittleness numbers
b and b ^ , and the characteristic dimensions a0 and w are, respectively, completely
equivalent in a fracture mechanics sense [11]. Accordingly, the differences in the
intrinsic strengths tc [Eq. (4a)] and ^tc [Eq. (4b)] are only apparent [11,12].
A fracture mechanics estimate of the characteristic size of frontal process zone,
cf , can be made by recasting Eqs. (4a) and (4b) into the following equations [9 12]:
1 1 cf
¼ wþ 2 (5a)
tc 2 K c 2 Kc
for geometrically similar test specimens, and
1 1 cf
¼ a0 þ 2 (5b)
tc 2 Kc 2 Kc
for geometrically nonsimilar specimens. In Eqs. (5a) and (5b), use has been made of
the equivalence of ^tc tc . Equations (5a) and (5b) indicate an important advantage
in the experimental estimate of the plane-strain fracture toughness Kc and the frontal
zone size cf , respectively, from the slope and from the intercept of the linear
regression of 1=t2c vs. w and 1=t2c vs. a0 ; their linear regressions are shown in Fig. 8
for both geometrically similar and nonsimilar specimens. All of the data fall on a
single regression
p line with some experimental variations; yielding Kc ¼ 0:85
0:1MPa m and cf ¼ 0:8 0:1 mm. The in situ observation of the frontal micro-
cracking zone of the present graphite material with a scanning electron microscope
supports this fracture mechanics estimate of cf [4].
Once the cf -value was estimated, it is possible to calculate the brittleness number
b( a0 =cf , w=cf ), and then KQ can be universally plotted against b for all the test
specimens with geometrically similar and nonsimilar configurations, as illustrated in
Fig. 9 in its normalized
p form, where the solid line indicates the theoretical prediction
of KQ =Kc ¼ b=(1 þ b).
4.2. Potential Energy-Derived Fracture Toughness, Rc -The Irwin’s Similarity-

4.2.1. Nonlinear Deformation and Fracture of Polycrystalline Graphite
The stress vs. strain curves (S-S curves) of various types of carbon and graphite
materials have been studies in tensile, compressive, and flexural tests [13]. Most of
these S-S curves are nonlinear (non-Hookean), and have a permanent residual strain
after complete unloading, resulting in a hysteresis loop during a loading/unloading
cycle. The apparent elastic moduli (the tangent of the loading S-S curves at the
respective strain values) are always greater at small strains than at large strains.
Two of the major microscopic origins of the mechanical non-linearity of carbon and
,
Figure 8. Linear regression of the intrinsic failure strength tc and the characteristic dimensions, a and W
of test specimen.
346 M. Sakai et al.
Figure 9. The universal plot of normalized fracture toughness KQ =Kc as a function of the brittleness
number b.
graphite materials include (1) dislocation-related plastic slip along graphite basal
planes, and (2) stress induced microcracking [13]. Due to the significant crystallo-
graphic anisotropy of carbon and graphite materials, numerous microcracks (Mro-
zowsky microcracks) are induced in the carbonization and graphitization processes
of these materials. These microcracks pre-existing in the material is very suscept-
ible to applied external stresses/strains, leading to sizable stress-induced microcrack-
ing [4].
The plastic-slip always results in a nonlinear S-S curve with a finite residual
strain after complete unloading. In contrast, the microcracking-induced nonlinear
S-S loading curve and the subsequent unloading hysteresis do not leave a permanent
residual strain after complete unloading, unless the stress-induced microcracks sig-
nificantly relieve residual stresses. Crack-face bridging and interlocking also contrib-
ute significantly to the permanent residual strain after complete unloading.
4.2.2. Non-Linear Fracture with Irreversible Deformations

Burns and Swain [6], and then Sakai [4,5] examined the energy principle for the
fracture with a finite amount of irreversible deformation that is associated with crack
extension, yielding a residual load-point displacement duir after an incremental crack
extension of da.
Figure 10 schematically illustrates two different types of crack extension prob-
lems for a linear elastic body (material A) and a nonlinear inelastic body (material B)
with irreversible load-point displacement (duir ) after an infinitesimal crack extension
da at a constant critical load Pc . The total amount of load-point displacement du
induced by this virtual crack extension da of the inelastic material B is given by the
sum of the elastic and irreversible displacements, due and duir , i.e., du ¼ due þ duir .
This fracture mechanics consideration for residual load-point displacement affords
the following modification of the Irwin’s similarity [4 6]:
Figure 10. Load P vs. load point displacement u diagrams for (a) a linear elastic material A, and (b) a
nonlinear inelastic material B with a finite offset displacement duir .
KQ 2
Rc ¼ (1 þ 2h) (6)
E0
where h is defined by
duir =da duir

h¼ ( ) (7)
due =da due
Accordingly, h stands for the fraction of irreversible load-point displacement duir

relative to the elastic displacement due during a virtual crack extension of da. The
Irwin’s similarity must be modified by a factor of 1 þ 2h for an inelastic body with a
finite residual offset load-point displacement; Eq. (7) necessarily becomes the Irwin’s
LEFM expression when h ¼ 0 (that is, duir ¼ 0).
4.2.3. Deviation from the Irwin’s Similarity

In LEFM-regime, the stress-intensity-derived fracture toughness Kc (or KQ ) and the
potential-energy derived fracture toughness Rc is related by the Irwin’s similarity of
Rc ¼ Kc 2 =E 0 or ¼ KQ 2 =E 0 . The deviations from Irwin’s similarity are well demon-
strated in Fig 6 for geometrically similar specimens and in Fig. 7 for geometrically
nonsimilar specimens. The deviation is significantly dependent on the geometries and
the dimensions (a/W and W) of test specimen. It is not well understood in the present
study what kinds of microscopic processes and mechanisms dictate the size-effect
laws through which quantitatively described is the deviation from the Irwin’s simi-
larity. Burns and Swain related the finite amount of h-value of zirconia ceramics to
the compressive irreversible deformations via martensitic phase-transformation in
the crack-wake region behind a propagating crack [6]. In the present study for a
critical crack (not propagating crack) of polycrystalline graphite, however, the wake-
effect is not the origin for providing the finite discrepancies between Rc and Kc 2 =E 0
observed in Figs. 6 and 7.
348 M. Sakai et al.
Figure 11. The nonlinear parameter h plotted against the critical load per thickness Pc =B for all the test
specimens in geometrically similar and nonsimilar configurations.
As discussed in the preceding section, the deviation from the Irwin’s similarity
can be quantitatively expressed in a phenomenological manner by the dimensionless
parameter h. The h-values of all the test specimens examined in this study are plotted
in Fig. 11 against the critical load per unit thickness of test specimen, Pc =B. As clearly
demonstrated in Fig. 11, there exists a strong correlation between h and Pc =B,
though some experimental variations are included. This correlation implies that the
deviation from the Irwin’s similarity is directly enhanced by the critical load Pc
applied to the specimen, being independent of the fracture mechanics requirement,
‘‘how large or small the notch/crack length is’’. This fact suggests that the deviation is
not related to the presence of a notch or a crack itself, but is related to the stress/
strain filed in the locations away of the crack-tip process zone.
Examples of the stress distributions [the normal stresses (tensile/compressive
stresses) along the longitudinal direction (x-direction) of specimen] in three-point
flexural configuration are shown in Fig 12 for test specimens with three different
relative notch/depths (a0 =W ¼ 0:1, 0.5, and 0.9) at their respective critical loads. The
numerical calculations were made in a finite element method (FEM) (ANSYS, Ver.
5.3, ANSYS, Inc., IL., USA). As clearly seen in Fig. 12, the stress-values away of the
notched area are rather large both in tensile and compressive stresses for the speci-
men with a relatively small notch depth of a0 =W ¼ 0:1, resulting from the rather
large value of Pc . In contrast to the small notch depth, the specimen with a relatively
large notch depth of a0 =W ¼ 0:9 has none of large stresses outside the notch-tip
region. This is resulted from the very smaller critical load Pc for test specimens with a
larger value of a0 =W . The significant discrepancy between Rc and KQ 2 =E 0 for the
smaller a0 =W -values and its diminishing nature for a0 =W > 0:7 observed in Fig. 7
have a good correlation to the stress distributions of test specimens demonstrated in
Fig. 12. These FEM-results combined with the considerations for the characteristic
non-linear microscopic deformation/fracture processes given in Sec. 4.2.1. may lead
to a rather good understanding for the deviation from the Irwin’s similarity observed
in the present polycrystalline graphite. Accordingly, it may be well concluded that the
Figure 12. Numerical results of finite element analysis for the stress distribution profile (normal stresses,
sx ) of the specimens with relative notch depths of a0 =W 0:1, 0.5, and 0.9 at the respective critical loads
Pc , illustrating a significant difference in the tensile/compressive distribution profiles in the locations away
of the cracked region.
excess energy (Rc KQ 2 =E 0 ) is consumed in the area outside the cracked region,
where the stress/strain field is large enough for operating inelastic micro-deform-
ations and micro-fractures (i.e., plastic slippage along graphite basal planes and
stress-induced microcracking).
350 M. Sakai et al.
5. CONCLUSION
Nonlinear fracture of a polycrystalline graphite was examined. Fracture mechanics

test specimens (3-point flexural rectangular bars) with geometrically similar as well as
non-similar configurations were utilized to measure the stress intensity-derived frac-
ture toughness (Kc , KQ ) and the potential energy-derived fracture toughness Rc . The
dependences of these toughness parameters on the geometries and dimensions
allowed an estimate of the size of crack-tip frontal process zone through the consid-
erations of the size-effect law of fracture toughness.
The Irwin’s similarity relationship of Rc ¼ Kc 2 =E 0 and the deviation from this
relationship in nonlinear fracture mechanics regime were studied as functions of the
dimensions and geometries of test specimens. In nonlinear fracture regime, Rc is
always larger than Kc 2 =E 0 (or KQ 2 =E 0 ). The excess energy, Rc Kc 2 =E 0 , was experi-
mentally expressed by an increasing function of Pc =B, independent of the magni-
tude of notch/crack depth, and being attributed to the energies consumed in the
regions away of the cracked area. This excess energy consumption was intimately
related to the characteristic microcracking in polycrystalline graphite, as well as to
the intrinsic plastic slippage of graphite basal planes. These microscopic processes
and mechanisms leading to the excess energy consumptions may not contribute to
the resistance to main crack extension in a dead load (failure for constant loading).
However, in mechanical conditions of fixed grips (failure under a constraint dis-
placement) and of thermal strains, these microscopic deformations and fractures
operating outside the cracked region may significantly enhance the mechanical
relaxation of local stresses and strains at crack-tip, and then result in a sizable
reduction of the driving force for the extension of main crack, contributing to
toughening a cracked body.
REFERENCES
1. D. Broek, Elementary Engineering Fracture Mechanics (Martinus Nijhoff, Dordrecht, 1986), Chap. 5.
2. R.W. Hertzberg, Deformation and Fracture Mechanics of Engineering Materials (Wiley, New York,
1983), Chap. 8.
3. M.F. Kanninen and C.H. Popelar, Advanced Fracture Mechanics (Oxford University Press, New
York, 1985), Chap. 3.
4. M. Sakai, J. Yoshimura, Y. Goto, and M. Inagaki, J. Am. Ceram. Soc., 71[8] 609 616(1988).
5. M. Sakai and R.C. Bradt, Int. Mater. Rev., 38[2] 53 78(1993).
6. S.J. Burns and M.V. Swain, J. Am. Ceram. Soc., 69[3] 226 230(1986).
7. S.M. Barinov and M. Sakai, J. Mater, Res., 9[6] 1412 1425(1994).
8. Stress Intensity Factors Handbook, Vol. 1, edited by Y. Murakami (Pergamon, Oxford, 1987), pp. 13
15.
9. Z.P. Bažant and P.A. Pfeiffer, Am. Concr. Inst. Mater. J., Nov.-Dec., 463 480(1987).
10. Z.P. Bažant and M. Kazemi, J. Am. Ceram. Soc., 73[7] 1841 1853(1990).
11. M. Sakai and H. Kurita, Fracture of Brittle Disordered Materials: Concrete, Rock and Ceramics,
edited by G. Baker and B.L. Karihaloo (E&FN Spon, London, 1995), pp. 227 245.
12. M. Sakai and H. Kurita, J. Am. Ceram. Soc., 79[12] 3177 3184(1996).
13. B.T. Kelly, Physics of Graphite (Applied Science, London, 1981), Chap. 3.
Figure 13. One to one correlation between Jc and Rc .
APPENDIX
Rice was the first to recognize the potential use of J-integral in elastic-plastic fracture
mechanics, after Eshelby was apparently the first to derive the integral [1 3]. The
J-integral of a ‘‘critical’’ crack, Jc , is actually equivalent to Rc defined in Eq. (2)
combined with the schematic procedure in Fig. 2. For the case of a plate containing a
deep notch and subjected to pure bending, Rice found that [1 3]
Z uc
2
Jc ¼ Pdu (8)
bB 0
where b is the length of the remaining un-notched ligament. The correlation between
the Rc -values estimated in Eq. (2) and the Jc -values thus determined in Eq. (8)
is shown in Fig. 13, demonstrating a rather good one-to-one correlation with
some experimental variations. Accordingly, the discussions for the deviation from
the Irwin’s similarity in terms of Rc , and for the dependence of h on Pc =B made in the
present study are made without any essential modifications, even if we utilize
the critical J-integral, Jc , instead of Rc .
ON INTEGRITY OF FLEXIBLE
DISPLAYS
Piet C.P. Bouten*
1. INTRODUCTION
Nowadays two display types are dominant in the display market: the bulky cathode
ray tube (CRT) and liquid crystal displays (LCD). Both types use glass as substrate
material. The LCD display is the dominant player for mobile applications, in for
instance mobile phones and portable computers. In the development of displays and
their applications a clear interest exists to replace the rigid rectangular display cells
by free-shaped, curved or even roll-up cells. These types of applications require
flexible displays.
Different types of flexible displays have been presented recently.1 Section 2
gives a short introduction to these display types. To realize the flexible displays,
flexible substrate materials such as thin glass,2 metal foil,3 and/or polymer are used as
substrate materials instead of rigid (rectangular) glass sheets. Requirements for
(polymer) substrates, used to realize flexible displays are treated in section 3.
To fulfill the requirements, functional brittle layers are applied on the polymer
substrates.
Various types of failures encountered in flexible displays are presented in
section 4. Dominant failure types are related to the application of thin brittle layers
on polymer substrates. An example of such a failure type, the tensile failure of a
brittle ITO layer on a polymer substrate, is presented in more detail in section 5. This
example shows that formalisms, used in scaling laws and reliability predictions for
the (standard) glass displays are not (completely) valid for flexible displays. In the
final discussion the need for scaling laws and reliability models for flexible displays is
emphasized.
*
Piet C.P. Bouten, Philips Research Laboratories, 5656AA Eindhoven, The Netherlands.
597
598 Piet C.P. Bouten
2. TYPES OF FLEXIBLE DISPLAYS
Slikkerveer1 defines a flexible display as ‘‘a flat panel display made using thin, flexible
substrates, where the substrate can be bent to a radius of curvature of some centi-
metres without loss of functionality.’’ An example of a flexible display is given in
figure 1. This display, a flexible CTLC (Cholesteric Texture Liquid Crystal) cell, is
only 250 mm thick and can be bent down to a radius of two centimeter while
operating.4 The size of the presently shown display cell is 150 125 mm2 .
Four application areas are recognized for flexible displays:
1. Flat displays. In this area the main points are the ruggedness of the cell, the low
weight and the freedom to use non-rectangular devices.
2. Curved displays. The display is used at a constant, predefined, curvature, for
instance on a (curved) surface of a handheld device.
3. Displays on flexible devices. The display bends as the substrate bends. An
example is a display in a smartcard.
4. Roll-up displays. These displays are frequently rolled and unrolled during appli-
cation.
In application areas 2 4 the display cells are mechanical loaded on purpose. In
the second application area it is loaded once when bending the cell in its final shape.
Areas 3 and 4 require more loading cycles. Not only functional requirements on
the display performance (the image quality), but also structural requirements on the
display construction and the materials are needed.
Many presently used flat panel displays with glass substrates, such as LCD cells
in a mobile phone, are used in constructions that prevent loading of the display cell.
Excessive loading of these cells leads to glass failure. In mobile phones, for instance,
display cells are always incorporated at a certain depth in the body to prevent glass
fracture in a drop test. At this point the flexible displays differ clearly from the
Figure 1. A flexible liquid crystal display.

On Intergrity of Flexible Displays 599
presently used glass-based flat panel displays. To bend the display, a mechanical load
has to be applied on the cell. This load leads to a stress distribution in the display.
Upon bending the inner surface is loaded in compression and the outer surface in
tension. Knowledge about the acceptable stresses and/or strains of the various
display parts is required for reliability modeling.
Flexible displays require not only thin, flexible substrates, but also thin display
effects. In most flat panel display the display effect is sandwiched between two
substrates. Patterned electrodes on both substrates are used for localized addressing.
Three thin display effects are mentioned here:
(3) The liquid crystalline material in a LCD display acts as a light switch. The
thickness of this layer is 1---10 mm. LCD displays require accurate cell gap
control. This can be achieved with flexible substrates (fig. 1).4
(3) The OLED (organic light emitting) displays use light emitting organic mol-
ecules. They are incorporated in a very thin (< 1 mm) functional stack. The
functional display layer is extremely sensitive to water and oxygen. This sets
tough requirements to water and oxygen permeation rates of the substrates.5
(3) Electrophoretic displays, for instance an E-ink display,3 use colored charged
particles and a colored liquid. They are encapsulated in an about 50 mm thick
polymer film.
High performance displays, such as color displays on mobile phones and computer
screens, require high image refresh rate in combination with high display brightness.
The patterned electrode structures in these ‘‘active matrix’’ devices4 contain elec-
tronic switches in the display area. High performance active matrix devices are
silicon-based.
3. SUBSTRATES
Various types of flexible displays sets different requirement to the substrates. In

general, the display substrates need to be transparent and of good optical quality.
During display processing, they must withstand a number of chemicals, such as
organic solvents and/or strong acids/bases. Although frequently the process tempera-
ture can be lowered relative to that of glass, the substrates should be able to
withstand temperatures between 1508C and 3008C. The maximum temperature de-
termines to a large extend the selection of the base substrate material. To give an
example: the deposition of polycrystalline silicon for high performance active matrix
displays requires a maximum process temperature of about 2758C.6 This temperature
requirement limits the substrate selection seriously.
Properties of base polymer films are often not appropriate for display produc-
tion processes and display applications. Important parameters, like permeability and
solvent resistance are inferior to those of glass. To improve the performance of the
polymer sheets, several coatings are applied on a base polymer film (fig. 2). Func-
tional coatings like a solvent-resistant and/or scratch resistant coating are either
polymer-based or a combined organic/inorganic system. These layers are a few
microns thick. Other functional coatings are thin (< 150 nm) fully inorganic layers.
So is the transparent conductor ITO (Indium Tin Oxide) used as electrode material.
A hermetic transparent permeation barrier is required to guarantee sufficient lifetime
transparent conductor (ITO)
and/or hermetic coating
base substrate material
solvent resistant coating

and/or scratch resistant coating
Figure 2. A schematic representation of the cross section of a flexible display substrate.
for the OLED cells. The low water permeation rate (< 10 6 gr=m2 =day) can only
be realized when (transparent) inorganic layers are applied on the polymer substrate.
To achieve sufficient flexibility for the device, the selected substrate thickness is in
the 50---200 mm range. Due to thermal expansion differences of the layers, a stress
distribution will be present over the thickness of the multi-layer substrate. This leads to
warping of the freestanding film. Processing of strongly warped films is not preferred.
The thin (< 150 nm) inorganic (brittle) layers have in general a significantly
smaller failure strain than the base substrate material. They limit seriously the
allowed deformations of the substrates and the cells. Failure of these brittle layers
is a serious point of concern in the discussion on integrity of flexible displays. In the
next section attention will be paid to several failure mechanisms, observed in flexible
display cells and substrates. In section 5, some aspects on the failure of an ITO layer
are described in more detail.
4. FAILURE TYPES
A complex interplay between residual stress distribution, cohesive and adhesive

properties of the various layers in the stack, and the location of a critical defect
determine how a crack proceeds in a product. Some examples of evolving cracks will
be given.
In the previous section it is indicated that the basic substrates require processing
at temperatures above 1508C. The substrates therefore should have a high glass
transition temperature. The highly transparent substrates are amorphous or semi-
crystalline. At ambient temperatures, they are in the glassy state. This can also be
seen from the failure behavior. Figure 3 shows a strongly bent flexible LCD display.
In the LCD display a cell gap between the two substrates is filled with the liquid
crystalline material. A crack has started at the lower cutting edge of the upper
substrate and runs in the strongly curved substrate, loaded in tension. Around the
crack the gap between the substrates increased, and the LC material is withdrawn.
The crack was initiated at an edge defect. This mechanism is similar to that observed
in glass sheets. The quality of the cutting edge determines the failure strain. The
sample shows a dominant brittle failure behavior for the polymer substrate.
At ambient conditions usually a biaxial compressive stress is present in the
brittle top layer of the substrate. At high biaxial stresses a characteristic delamination
pattern, the ‘‘telephone cord’’ structure is frequently observed.7 Figure 4 shows an
AFM image of this structure, observed for an ITO layer on a polymer substrate.
From the characteristic sizes an estimate for the adhesion quality can be obtained.
Figure 3. Edge failure of the upper substrate of a flexible LCD cell upon bending.
80 μm
40 μm
0 μm
0 μm 40 μm 80 μm
Figure 4. ‘‘Telephone cord’’ type delamination pattern due to biaxial compressive stress in the top layer.
A seal line is used to connect both substrates in the construction of flexible

display cells, for instance at the circumference of the cell. Upon bending serious stress
concentrations occur at the mechanically loaded seal line. This may lead to prefer-
ential failure near this seal line. of the some layers on the substrate. Figure 5 shows
crack patterns in the ITO layer on top of a substrate at a location where the seal line
is removed. The ITO layer is partly removed with the seal line.
Figure 5. Crack patterns in the ITO layer at the location of the removed seal line (left).
A more detailed experimental study on the propagation of cracks in the multi-

layer structure of model substrates for flexible displays8 clearly shows an interaction
of the cracks in the brittle layer and cracks in a polymer hard-coat layer.
5. ITO TESTING
We will now focus on the failure of the transparent conductor (ITO). A tensile test is
used in combination with resistance measurements9, 10 to determine the critical failure
strain of the brittle transparent conductor (ITO) on the polymer substrate. The strain,
determined directly from the displacement of the clamps9 is systemically too high. To
circumvent this problem the displacement can be measured on the sample.10
The two point bending test,11 schematically represented in figure 6, has no
clamping problems. A rectangular sample of the film (length 80 mm, width 15 mm)
is placed between two parallel plates. At the location with the lowest curvature radius,
Figure 6. Geometry of the two point bending test.

Rmin , the largest tensile strain is present at the outer surface. For a symmetrical sample
the strain at this location is emax ¼ 1:198 h=L, where h is the sample thickness and L the
distance between the neutral lines. The faces of the clamps isolate the sample electric-
ally from the machine frame. In the test the resistance of the ITO layer is determined as
a function of the distance Lo between the plates (Lo ¼ L þ h). The crosshead velocity
of the universal testing machine (Instron 5566) during the test is 10 mm/min. The
relative resistance increase DR=R is represented as a function of maximum tensile
strain emax in figure 7. The curves, represented in this figure are measured on samples of
the same width (15 mm). One sample has a uniform ITO layer over the full width. On
the other samples a single ITO line of a defined width is etched.
Up to a certain strain level the resistance is constant. Above this strain an
increase in the resistance is observed. The strain corresponding to a 10% resistance
increase is used as a critical strain. The significant increase of the resistance is due to
the development of cracks in the ITO layer normal to the conductive path. The shape
of the curve in figure 7 depends on the ITO line width. A sudden resistance increase is
observed for narrow lines, a more gradual increase for the wider lines.
Figure 8 shows the edges of the ITO layer. The cracks, visible in the upper
sample surface (uniform ITO layer on top), are due to the cutting process. The cracks
are present in a zone near the sample edges that is approximately 200 mm wide. The
Figure 7. The increase in resistance of ITO layers as a function of the maximum applied strain. ITO lines
of different width are tested.
Figure 8. ITO edge of cut sample (left) and etched ITO line (right).
etched ITO layer shows no edge cracks. The crack onset strain for the etched lines
and the sample with the uniform ITO layer is nearly similar, about 1.2% for the tested
samples. The presence of the edge cracks has no significant effect on the on the
critical failure strain of the thin layers.
To get information on the failure strain distribution, larger series of samples
are tested in three batches. For one substrate type, samples with a uniform ITO
layer and samples with etched narrow ITO lines are measured. For a second sample
type only a uniform ITO layer is investigated. Details of the tests are given in
table 1.
Experiments on narrow single ITO lines show at a critical failure strain a sudden
increase of the resistance, in most cases to infinity. This effect is used to reduce
the number of tested samples: Six samples with ten parallel 0.75 mm wide ITO
lines at 1.0 mm pitch are etched. The resistance of these ten parallel lines is mea-
sured as a function of the plate distance Lo . Failure of a single ITO line is recorded as
a step in the resistance-strain curve. From the six samples 60 line failures are
recorded.
The data are represented in a two parameter Weibull distribution: P ¼ 1- exp
(e=eo )m . Figure 9 shows the experimental data of substrate II in a Weibull graph. The
experimental failure strain emax and the failure probability P are represented on
the axes. The slope m and the characteristic strain eo (at 63 % failure probability)
are obtained from a fit on the experimental data. The fit constants for the 3 tests are
included in table 1.
Table 1. Characteristic failure strains obtained from a Weibull analysis on 3 sets of samples.
Sample type h [mm] ITO layer # samples Weibull
eo [%] m
I 130 Lines, 0.75 mm 60 1:17 0:02 20

I 130 Uniform 30 1:24 0:02 20
II 100 Uniform 36 1.46 0.02 40
Figure 9. Weibull graph for the critical failure strain of sample II.
6. DISCUSSION
6.1 Failure of Brittle Layers

In the sections 4 and 5 a variety of failure modes for flexible displays is presented.
Knowledge of at least the dominant failure modes is required for the development of
reliability models and scaling rules. These types of models, applied to brittle mater-
ials, like glass and ceramics, use weakest link statistics, in many cases Weibull
statistics.
In the previous section Weibull statistics is used to describe the critical failure
strain of a brittle conductive layer (ITO) on a polymer substrate. During the test the
polymer substrate behaves completely elastic11. The results, obtained by tensile
loading of the brittle layer on the substrate differs significantly from the behavior
of (bulk) brittle materials:
(3) The presence of large edge cracks does not affect the critical strain of the thin
layer. Nearly identical failure strains are obtained for the smooth etched ITO
lines with straight crack-free edges and uniform ITO layers with heavy damaged
ITO edges.
(3) The critical failure strain of the uniformly coated sample (1.24 %) is even slightly
higher than that of the ITO lines (1.17%). This is not expected when a distribu-
tion of weakest spots in the surface area is responsible for the failure behavior.
The tested ITO area is larger for the stronger sample.
(3) The length of the ITO edges, tested in the two point bending test, is similar for
the uniform ITO layer and the etched ITO line. The difference in crack density
at the ITO edges is not responsible for the larger critical failure strain of the
uniform ITO layer.
(3) Identical strength distributions (m ¼ 20) are obtained for the same substrate
type (I, table I). For the other substrate type a significantly higher failure strain
(1.46 %) and a Weibull modulus m ¼ 40 are reported. These differences reflect
the different deposition conditions.
(3) High Weibull moduli (m 20) are only obtained in bulk materials characteriza-
tion with well-defined process conditions. They are observed, for instance, when
well defined surface damage is applied, for instance with powder blasting.12
A Weibull modulus m 40 is only obtained for quite deterministic stress levels,
for instance with the intrinsic strength of glass fibers.13
Crack initiation and crack propagation are two clearly different phenomena in
the mechanics of thin layers on substrates7. The experiments on the ITO layer
indicate that not the crack initiation but the crack propagation dominates the critical
failure strain of the ITO layers. The fact that the poor quality of the ITO edge of
the uniformly coated sample does not affect the failure strain clearly supports the
propagation-controlled mechanism. A well-defined crack propagation threshold
leads to a high Weibull modulus (m 20). The level of this threshold is determined
by layer properties, like internal stress, layer thickness and adhesion to the substrate.
Leterrier14 et al. studies the crack evolution of brittle silicon oxide coatings on
polymer substrates with the fragmentation test. The number of cracks is counted as a
function of the applied strain. The evolution of the crack density in a sample just
above the crack onset strain is described with Weibull statistics. Weibull moduli
2.5 m 12 are reported. This is significantly lower than the m 20 obtained from our
tests with, using resistance measurement on ITO. It is noted in this respect that the
phenomena might be slightly different. Leterrier reports on the appearance of cracks
within a limited area, observed with a microscope. The cracks may have a limited
length. A significant increase of the resistance of an ITO layer is expected either with
a large amount of small cracks in the layer, or a crack over nearly the full layer width.
The differences in Weibull moduli might be related to differences in crack initiation or
crack propagation limits.
For narrow lines a sudden increase of the resistance is observed; for wider lines a
more gradual increase is measured. In a model description9 of the resistance increase
in the tensile test, the presence of conducting material in the crack is assumed to be
responsible for the gradual resistance increase. This model does not account for the
presently observed line width dependence of the resistance increase.
In the present set of experiments, with 0.75 mm wide ITO lines, the sudden
increase in resistance and the well-defined, propagation controlled critical failure
strain is obtained. In many applications in electronics industry significantly smaller
details are present. For active matrix addressed displays, for instance, transistors and
connection lines significantly smaller than 100 mm are applied. When the size of the
cracks, initiated for instance around a defect, have a stable size, comparable to the
size of the component, crack initiation-controlled mechanisms and not the propaga-
tion control may govern the resistance of the lines.
Cracks in a (oxygen and/or water) barrier coating degrade the performance of the
hermetic coating. For this reason, crack initiation is expected to be a dominant failure
mechanism for the performance of OLED displays, requiring hermetic coatings.
6.2 Reliability Models

Glass is the dominant substrate material in the display industry. The brittle failure
behavior of glass is exploited in production processes, and taken into account in
applications. Knowledge of the failure behavior, and the mechanisms leading to
failure, is quite important in the production of reliable products. An appropriate
combination of stress distributions, failure statistics and crack extension models is
used in reliability models for brittle materials.
This type of modeling is also required for new products, entering the market.
Flexible displays are nowadays quite appealing, both due to their flexibility and the
robustness of the cell. From fracture mechanical point of view, the base polymer
substrate has certainly a higher failure strain than a glass substrate. The limiting
strain for the substrate can be related to the onset to non-elastic deformations or to
brittle failure behavior, thus the presence of large cracks.
The substrate is a multi-layer stack. Cohesive and/or adhesive failure of one of
the layers might lead to cell failure. Section 5 describes the loss of conduction of a
tensile loaded ITO layer. It is strongly related to the crack evolution in the brittle
layer. Brittle functional layers seriously limit the applicable deformations of the
flexible display cell. The allowable deformation of the presently tested ITO layer is
slightly above 1%.
The critical strain of the presently tested brittle ITO layer is determined with
relatively small samples in a short lasting test in ambient conditions. For the predic-
tion of the reliability of a real product, tests under more severe conditions and at
longer timescales are required. This requires knowledge of the behavior of different
materials as a function of time and temperature. For polymeric materials, incorpor-
ation of time-dependent deformation behavior, like creep and relaxation or plastic
deformation, might be necessary. For thin brittle layers phenomena like slow crack
might contribute to a lowering of the critical strain.
Apart from these phenomena, stress concentrations due to display constructions
might seriously lower the applicable safe deformation. Near a seal line, for instance,
preferential failure might occur (figure 4). All these factors result in a different, and
probably significant lower, safe critical strain.
A large variety of mechanisms is responsible for the failure of a flexible displays.
Production and market introduction of complete new products like flexible displays
requires knowledge of the critical failure mechanisms. This knowledge is necessary to
development reliability models.
7 CONCLUSIONS
Different failure modes for flexible display substrates are presented. Brittle layers,
such as inorganic permeation barriers and transparent conductors (ITO) fail at strain
levels of about 1%. They predominantly limit the allowable substrate deformation.
The critical failure strain of the conducting ITO layer does not depend on the
presence of cracks in this layer. This is not in agreement with models used to predict
the reliability of (bulk) ceramics.
During production, application and use of flexible displays, mechanical load is
exerted on the substrates and the display cell. A good understanding of the different
possible failure mechanisms is required in the development of flexible displays
towards a reliable product. Appropriate reliability models are presently not available
for flexible displays.
8. ACKNOWLEDGEMENT
The author acknowledges Judith de Goede and Jeroen Alberts for their contributions
to the experimental work. Part of the work is done within the IST program (IST-
2001-34215 FlexLED) of the European Community.
REFERENCES
1. P.J. Slikkerveer, Bending the Display Rules, Options and Challenges for Flexible, Proc. EuroDisplay
2002, Nice, 1 4 oct. 2002, pp 273 276.
2. A. Weber, S. Deutschbein, A. Plichta and A. Habeck, Thin Glass polymer Systems as Flexible
Substrates for Displays, SID 02 digest, Boston, 21 23 may 2002, pp. 53 55
3. P. Kazlas, A. Ritenour, J. Au, Y. Chen, J. Goodman, R. Paolini and H. Gates, Card size Active
matrix Electronic Ink Display, Proc. EuroDisplay 2002, Nice, 1 4 oct 2002, pp 260 262.
4. P. Slikkerveer, G. Nisato, N. Kooyman, P. Cirkel and P. Bouten, A Fully Flexible, Cholesteric LC
Matrix Display, SID ’02 digest, Boston 21 23 may 2002, pp 27 29.
5. G. Nisato, P.C.P. Bouten, P.J. Slikkerveer, W.D. Bennett, G.L. Graff, N. Rutherford and L. Wiese,
Evaluating High Performance Diffusion Barriers: the Calcium Test, Proc. IDW ’01, Nagoya, 16 19 oct
2001, pp 1435 1438.
6. N. Young, M.J. Trainor, D.T. Murley, D.J. McCulloch and R.J. Wilks, LPTS for AMLCD on Glass
and Polymer Substrates, Proc. IDW ’99, 1999
7. J.W. Hutchinson and Z. Suo, Mixed Mode Cracking in Layered Materials, Adv. Appl. Mech., 29
(1992) pp. 63 191.
8. Y. Leterrier, C. Fischer, L. Médico, F. Demarco, J. A. E. Månson, P.C.P. Bouten, J. de Goede,
G. Nisato and J.A. Nairn, Mechanical Properties of Transparent Functional Thin Films for Flexible
Displays, Proc. Soc. Vacuum Coaters, 46th annual conference, San Francisco, may 2003, to be
published.
9. D. R. Cairns, R. P. Witte II, D. K. Sparacin, S. M. Sachsmann, D. C. Paine, and G. P. Crawford,
Strain dependent Electrical Resistance of Tin doped Indium Oxide on Polymer Substrates, Appl.
Phys. Lett. 76 (2000) 1425 1427
10. Y. Leterrier, L. Médico, F. Demarco, J. A. E. Månson, M. Escola Figuera, M. Kharazzi Olson and
U. Betz, Mechanical Integrity of Transparent Conductive Oxide Film for Flexible Polymer based
Displays, submitted to Thin Solid Films.
11. P.C.P. Bouten, Failure test for brittle conductive layers on flexible display substrates, Proc. EuroDis
play 2002, Nice, 1 4 oct 2002, pp 313 316.
12. P.J. Slikkerveer, P.C.P. Bouten, F.H. in ‘t Veld and H. Scholten, Erosion and damage by sharp
particles, Wear 217 (1998) 237 250.
13. C.R. Kurkjian and U.C. Paek, Single valued Strength of ‘‘Perfect’’ Silica Fibers, Appl. Phys. Lett. 42
(1983) 251 253.
14. Y. Leterrier, J. Andersons, Y. Pitton and J. A. E. Månson, Adhesion of Silicon Oxide Layers on
Poly(ethylene terephthalate). II: Effect of Coating Thickness on Adhesive and Cohesive Strengths,
J. Polym. Phys. 35 (1997) 1463 1472.
PARTICLE IMPACT DAMAGE
AND POINT LOAD-INDUCED
FRACTURE BEHAVIOR IN
ZIRCONIA PLASMA SPRAY
COATING FILM
Yoshio Akimune, Kazuo Matsuo, Tatsuo Sugiyama, Hiro

Yoshida, Satoshi Sodeoka, Satoshi Shimizu*
ABSTRACT
This study concerns the method for examining the load conditions and evolving
stress related to debonding at the coat-substrate interface to identify the responsible
factors, by indenting a ceramic coated specimen with a spherical indenter either
through impact or quasi-statically under static loading. The morphology of cracks
induced by the impact of a foreign object on to specimens of inconel substrate with
thermo-resistive coating of zirconia sprayed by two different methods was studied in
comparison with the indentation test results by a spherical indenter under static
loading. Consequently, the load conditions for coat fracture and debonding at the
coat-substrate interface were successfully identified in a plasma spray specimen.
INTRODUCTION
The heat-resistant coating technology has been expected to be utilized for protecting
the surface of metal blade in the gas turbine. The gas turbine is one of main devices in
National Institute of Advanced Industrial Science and Technology

1 1 1 Umezono, Tsukuba city, 305 8568Japan
*
1, Natsushima, Yokosuka city, 237 0061 Japan
437
438 Yoshio Akimune et al.
the thermal power generation, to be used with various kinds of fuel, such as heavy oil
and kerosene, and in the presence of combustion residuals and corrosive atmosphere.
Troubles in the coated turbine blade involve point load-induced damages resulting
from the impact of oxide scale which cause coating delamination and substrate
fractures. This problem is to be resolved as quickly as possible1.
Works to evaluate mechanical properties of ceramics protective coating for
turbine blade include damage analysis through the finite element method and
experimental results by A.S. Kobayashi et al.1, particle impact experiment with
ceramics film deposited on silicon nitride through CVD by Y. Akimune et al.2,
analysis of crack development after the indentation with a spherical indenter by
J.B. Davis et al.3. Evaluation of damage morphology with stress- strain curve of
coating film in a quasi-static testing for point load by A. Pajares et al.4–5,
S. Wuttiphan et al.6, M. Swain and J. Mencik7 as well. In these papers, analysis of
relationship between the mechanical behaviors of film and substrate under high
speed and quasi-static point loads are not dealt with. The theoretical analysis
has been conducted by Hertz8, and followed by those by S. P. Timoshenko and
J. N. Goodier9 and by K. L. Johnson10. Studies on damages in ceramic materials
using a spherical indenter are reviewed by B.R. Lawn11. On the other hand, it seems
to be important to set up a quasi-static evaluation technique to assess the fracture
events caused by foreign object damage in the actual turbine blade at the time of
depositing protective film.
The present study concerns a test for the purpose of proposing the optimum
indentation test method for assessing foreign object damage in heat resistant coating
film prepared through two spraying methods: atmospheric pressure plasma spray
and pressurized plasma spray. In the experiment, the film fracture and surface
delamination were observed after the high speed impact with particles, and quasi-
static indentation test with a spherical indenter, for a specimen consisting of inconel
600 substrate with a bond coat on the surface and an additional deposit of zirconia
spray film, prepared through either of two methods.
2.1. Specimens
The specimen was composed of inconel 600 substrate, 8wt% stabilized zirconia
(METECO204NS-G) as spray material, and NiCoCrAlY alloy (AMDRY-1) as
intermediate bond coat. The top coat is sprayed either with atmosphere or with
pressurized argon (200 kPa Ar), to form zirconia film of 200 mm thickness.
Specimen designation:
A: atmospheric pressure plasma spray (film thickness: 200 mm)
H: Ar pressurized plasma spray (film thickness: 200 mm)
Inconel 600 (substrate), thickness: 3 mm
2.2. Plasma Spray Conditions
The heat resistant coat was sprayed either by atmospheric spraying or by Ar pres-
surized spraying. Conditions for the experiment are summarized in Table 1.
Particle Impact Damage and Point Load-Induced Fracture Behavior 439
Table 1. A summary of spray conditions.
Atmospheric pressure Pressurized plasma

Bond Coat plasma Spray Spray
Top coat materials NiCoCrAlY 8 wt% Y2 O3 8 wt% Y2 O3

(AMDRY365 1) Stabilized ZrO2 Stabilized ZrO2
(METCO204NS G) (METCO204NS G)
Atmosphere Argon Atmosphere Argon
Pressure 10 kPa Atm. pressure 200 kPa
Spray distance 300 mm 100 mm 100 mm
Gas flow Ar: 50 L/min. Ar: 42 L/min. Ar: 50 L/min.
H2 : 9 L/min. H2 : 10 L/min. H2 : 5 L/min.
He: 10 L/min.
Powder feed 25 g/min. 4.8 g/min. 5.2 g/min.
2.3. Testing Method
(1) Foreign Object Impact Test12,13

A high-speed impact test system with an electron thermal (ET) gun is illustrated in
Fig. 1. First, a substrate with coating film is mounted on a specimen holder. The
specimen is to be mounted mechanically in line with the ET gun and the solenoid
coils. Then, a spherical magnet of samarium-cobalt (Sm2 Co17 , density: 8:4 g=cm3 ,
hardness: 5.4 GPa, Young’s modulus: 151 GPa, strength 118 MPa) is mounted in a
plastic sabot as a simulated foreign object.
The sabot is blasted off by the pressure of aluminum vapor, generated explo-
sively by applying 10 kV/50 kA power (electrical energy: 2.5 kJ) to an aluminum foil.
The sabot is halted at a stopper located at the muzzle, and the samarium-cobalt
magnet alone is shot through the two coils (100mm separation) to hit the target
specimen. The flight speed of the magnet sphere is calculated from the time difference
Figure 1. An overall view of the high speed impact testing system.

in rise-time of electromotive force induced as it passes through the coils. The impact
velocity in the experiment was around 500 to 600 m/s.
(2) Static Indentation Test14
A spherical indenter is pressed into the film at a constant speed of 0.05mm/min and
released. With load (P) and impression depth (h) measured, the stress-strain curve
was calculated by using a theoretical formula. The loading and unloading session was
repeated three times at a same rate. Owing to the mechanical design for reversing the
crosshead of the testing machine, the indenter was held for 1 second or so at the top
and bottom positions.
The specimen was processed by cutting to a size of about 15 mm square and
polished with abrasive (Buehler’s Colloidal Silica). With a point load indentation
device mounted on an INSTRON strength test machine(#4487), as shown in Fig. 2, a
spherical indenter (materials: WC, indenter diameter: 1, 2 and 4 mm) was loaded and
unloaded continuously up to the maximal load of 1 kN at a rate of 0.05 mm/min at
the room temperatures to measure the indent depth.
(3) Damage Observation
The impact impression of the shot was observed under an optical microscope. After
the macro-indentation test, a part of specimen surrounding the indentation was cut
out, embedded in epoxy-resin, sectioned and had the sectional face ground and
polished with abrasive (Buehler’s Colloidal Silica). The surface was observed under
the optical microscope in the bright field and through the differential interference, a
scanning electron microscope (SEM:Hitachi s-4000), and an electron probe surface
analyzer (EPSA:Elionix ERA8000).
Figure 2. An overall view of quasi static indentation testing system.

3.1. Foreign Object Impact Test

(1) Surface Observation
In the impact test, a spherical particle was shot onto the surface of atmospheric
pressure plasma spray coat specimen. At shooting speeds of 550 m/s and 620 m/s, no
radial crack was seen at the point of impact unlike the data in the literature2, though
spray coating film fell out of the surface immediately below the impact point (Fig. 3).
The observation with an EPSA, shown in Fig. 4, gave a removal of spray materials
immediately below the impact point and a buildup in its periphery. In case of
substrate only, similar morphology of crater and ridge was observed due to plastic
deformation4–5.
Figure 3. Surface fracture on atmospheric plasma spray specimen.
Figure 4. 3D analysis of surface fracture on atmospheric plasma spray specimen.

When a spherical particle was shot onto the surface of a specimen coated by
pressurized plasma spray, the removal of spray coating film from the surface was
seen just below the impact point for both 550 m/s and 540 m/s, as in case of
atmospheric pressure spray coating film, but none of radial cracks were recognized
from the impact point (Fig. 5). The observation under the EPSA, shown in Fig. 6,
revealed slight crater immediately below the impact point and buildup surrounding
it, definitely out of rough surface of spray coating.
(2) Cross-Sectional Observation

(i) atmospheric pressure plasma spray specimen
With a specimen using inconel 600 substrate, the cross-sectional view showed crater
and slight ridge caused by plastic deformation immediately below the impact site
(Fig. 7). In case of impact with 1 mmf particle, minor cracks and separation from
Figure 5. Surface fracture on pressurized plasma spray specimen.
Figure 6. 3D analysis of surface fracture on pressurized plasma spray specimen.

Figure 7. Cross section of an atmospheric plasma spray specimen.
bond coat were observed with very little plastic deformation of substrate. The
separation of coating film may be attributed to change in the direction of stress at
the time of unloading following the fracture of the spherical particle. As the particle
is made from sintered intermetallic compounds, with low strength, the fracture of the
sphere was thought to occur when the stress within the sphere exceeded its strength at
the moment of indenting the spray coat film. Consequently, the impact force was
dissipated by fracturing, before inducing plastic deformation in the substrate and the
bond coat. In this process, the vertical stress direction from the coating to the
substrate was reversed to cause lateral crack due to by stress direction change15.
(ii) pressurized plasma spray specimen
In case of pressurized plasma spray specimen, different kinds of damage were

observed, such as Hertz cone crack caused by impact loading in the spray coating
film, lateral cracks caused by unloading due to fracture of sphere, and debonding
(Fig. 8) from the bond coat extending radialy from the tip of Hertz cone crack8–10,
Figure 8. Cross section of a pressurized plasma spray specimen.

Figure 9. A crack vs debonding diagram (x R/h).
and a distinct plastic deformation of the substrate. These effects are a combination of
Hertz cone cracks generation with debonding. In the crack map proposed by Davis et
al.3, where normalized energy release rate x ¼ 1:30=p2 tan2 b is plotted against
normalized crack length, that is, cone crack radius R divided by coating film
thickness h, the experimental data, x ¼ 0:0127 and 0.305(left side crack and right
side crack, respectively) and R/h ¼ 3.2 and 2.08, (left side crack and right side crack,
respectively), falls in an area where Hertz cone cracks coexist with debondings
running from the cone crack tip in the radial direction and an area where Hertz
cone crack, in agreement with data in the reference2, (Fig. 9). The fracture proceeds
in the following way: the load brought about by the impact of the particle induces
cone cracks, which grow in the radial direction along the bond coat interface. As the
load is eliminated by the fracture of the sphere, debonding occurs immediately below
the cone crack15. This debonding is caused by the reversal of stress direction, through
a mechanism entirely different from that for debonding in the growing phase of cone
cracks3.
The difference in impact fracture behavior in two coating methods is thought to
be related to the different yielding process in two coating films. The residual pene-
tration depth in the coating films revealed impact damage behavior change which
resulted in stress change under impact site.
3.2. Quasi-Static Indentation Test

(1) Continuous Load/Unload Test14
An load/unloading curve with an atmospheric plasma spray-coated specimen (A) is

shown in Fig. 10, and a stress-strain curve by the indentation method in Fig. 11. In
the latter curve, the initial slope, corresponding to the Young’s modulus of materials,
was different when using a spherical indenter of diameter 1 mm from that using
Figure 10. Load unload curve with spherical indenter.
Figure 11. Stress strain curve for the indentation testing.
indenters of other sizes, for all the specimens. This may be attributed to the fluctu-
ation in the thickness of bond coat and top coat, and it seems necessary to take
adequate caution when interpreting the results with 1 mmf indenter.
It is evident that the yield stress (Pm ) represented by a point of inflection from
the initial slope tends to increase with the diameter of indenter. The similar tendency
is recognized with the substrate (Inconel 600), suggesting that the difference in the
elastic-plastic behavior or fracture morphology of spray film is responsible, depen-
dent on the plastic deformation of substrate. While there is little difference in yield
stress between atmospheric pressure plasma spray specimen (A) and argon pressur-
ized plasma spray specimen (H), the subsequent behavior of the latter specimen is
more inferior, suggesting that the pressurized plasma spray gives coating of poorer
tolerance to fractures.
Note that the calculation of Pm by equation (1).
Pm ¼ Py =pRh (1)
P ¼ ke h3=2 : relation of pressure to indent depth (2)

p
Ke ¼ 4=3 (RE )1=2 composite Youngs modulus (3)
1=E ¼ (1 n2 )=E þ (1 n2i )=Ei
where Pm :Yield contact pressure, Py :yield load, hy :penetration dept, R:sphere

radius, Ei :Young’s modulus of indenter(WC:534 GPa), vi : Poisson’s ratio of inden-
ter(WC:0.28)
(2) Surface Observation
Examples of surface observation of the impression after the load/unload test with
atmospheric plasma spray and pressurized plasma spray coating are shown in Figs.
12, respectively. In specimens with thin spray coat film, a lot of ring cracks are seen
on the surface, while in those with thick film, the number of cracks is not so many. In
the pressurized plasma spray coat specimen, circular depression is seen not only at
the indent impression but also in its surrounding. This may be interpreted in this way:
in consideration of after-yield behavior in the indentation test, if the spray film is
subjected to a force greater than the yield stress of materials in the process of the
load/unload test, micro-cracks grow to enlarge apparent radius of the indenter,
reducing the contact pressure and making the after-yield behavior flatter.
(3) Cross-Sectional Surface

The cross-sections of impression for different combination of specimen and indenter
size are shown in from Figs. 13 to Fig. 16. For the observation of plastic zone
immediately below the impression4,5. While there was little distinction in the fracture
of spray film and the formation of plastic zone in different specimens, they presented
clear differences depending on the indenter size. In case of indenter 1 mmf, the
fracture of Hertz cone crack8 occurred rather than interfacial debonding4,5. As the
indenter size increased, the fracture tended to turn closer to the interfacial debond-
ing. With smaller indenter, the indentation pressure is greater, to produce plastic
deformation in the substrate. With 1 mmf indenter, plastic deformation occurred in
proportion to the size of damage area, while with 4 mmf indenter, the plastic
Figure 12. Impression (pressurized plasma spray, indenter 4 mmf).

Figure 13. Cross section of impression (atmospheric plasma spray, indenter 1 mmf).
deformation occurred in a limited area only. On the basis of the observation of

fracture and debonding, it was shown that both specimens had hard coat film,
presenting elastic response with Hertz cone cracks in the earlier phase, and later,
shifting to elastic-plastic behavior. For this reason, when unloaded, the stress field
changed so as to cause lateral cracks and debonding, at the top coat/bond coat
interface15. That is, loading accumulates elastic strain energy at the bond coat and
the substrate, which is released as repulsive force, and as the stress acts vertically to
the surface of bond coat, debonding occurs at the interface between the top coat and
the bond coat.
3.3. Comparison of Spray-Coating Method on Static Indentation

With specimens prepared in different ways, the results of quasi-static testing using
spherical indenters of different diameters and those of fracture observation after
impact tests using 1 mmf spherical particle were compared.
When an atmospheric spray specimen was subjected to a quasi-static test with an
indenter of 1 mmf, both the substrate and the bond coat were affected with marked
plastic deformation, and in particular, the substrate underwent so drastically that the
limit of elastic-plastic deformation was stepped over to present a porous zone with
concentrated cracks. As the indenter size grew, the area of plastic deformation in the
substrate and the bond coat was reduced, to present Hertz cone cracks in the cross-
section. In addition to Hertz cone cracks, a number of cracks were seen in parallel to
the substrate, which were thought to have occurred when the load was eliminated.
With an indenter of 4 mmf, the area of plastic deformation was reduced further, and
lateral cracks were observed as the load was eliminated. Debonding between the
bond coat and the substrate was observed only with the 4-mm indenter. Comparing
to impact tests results(Fig. 7), qusi-static tests(Fig. 14) can reproduce similar fracture
behavior.
In case of pressurized plasma spray specimens, marked plastic deformation
occurred in the substrate and the bond coat with 1 mmf indenter, and in the
substrate, a porous zone with concentrated cracks as well as lateral cracks were
observed, stepping over the region of elastic-plastic deformation. As the indenter
size grew, the area of plastic deformation in the substrate and the bond coat was
Figure 14. Cross section of impression (atmospheric plasma spray, indenter 4 mmf).
Figure 15. Cross section of impression (pressurized plasma spray, indenter 1 mmf).
reduced, and Hertz cone cracks were seen in the cross-section. As in case of atmos-
pheric spray specimen, a number of lateral cracks occurred besides Hertz cone
cracks, which were thought to have been produced when the load was eliminated.
When a 4 mmf indenter was used, the area of plastic deformation was reduced
further, but cone cracks under the load and lateral cracks without load were ob-
served. Comparing to impact tests results(Fig. 8), quasi-static tests(Fig. 16) can
reproduce Hertz cracks and debonding under impact site but debonding from
Hertz cone crack tip.
4. CONCLUSION
The morphology of cracks induced by the impact of a foreign object on to specimens

of inconel substrate with thermal-resistive coating of zirconia sprayed by two differ-
ent methods was studied in comparison with the indentation test results by a
spherical indenter under static loading.
Figure 16. Cross section of impression (pressurized plasma spray, indenter 4 mmf).
In the impact test, for an atmospheric pressure plasma spray specimen, in case of
impact with 1 mmf particle, Hertz cone crack and separation from bond coat were
observed, with very little plastic deformation of substrate. On the otherhand for
pressurized plasma spray specimen, different kinds of damage were observed, such as
Hertz cone crack caused by loading in the spray coat, lateral cracks caused by
unloading due to fracture of impact sphere, and debonding from the bond coat
extending radially from the tip of Hertz cone crack at impact loading, and a minor
plastic deformation of the substrate.
On the basis of the observation of fracture and debonding for quasi-static
indentation, crack morphologies in an atmospheric pressure plasma spray specimen
were reproduced with 4 mm indenter. In the case of pressurized plasma spray
specimen, crack morphologies were also reproduced with 4 mm indenter except
debonding from cone crack tip.
ACKNOWLEDGEMENT
This work was performed as a part of the Nanostructure Coating Project carried out
by the New Energy and Industrial Technology Development Organization.
REFERENCES
1. B. M. Liaw, A. S. Kobayashi, A. F. Emery, and J. J. Du, ‘‘Impact Damage Model of Ceramic
Coating’’; pp. 187 96 in Fracture Mechanics of Ceramics. Vol. 7. Edited by R. C. Bradt, A. G.
Evans, D. P. H. Hasselman, and F. F. Lange. Plnum Press, New York, 1986.
2. Y. Akimune, T. Akiba, N. Hirosaki, and T. Izumi, ‘‘Impact damage behavior of CVD coated silicon
nitride for gas turbine,’’ J. Mat. Science.29,3243 47(1994).
3. J. B. Davis, H. C. Cao, G. Bao, and A. G. Evans, ‘‘The Fracture Energy of Interfaces: an Elastic
Indentation Technique,’’ Acta metall. mater. Vol. 39. No. 5. pp. 1019 1024. (1991).
4. A. Pajares, L. Wei, and B. R. Lawn, ‘‘Contact Damage in Plasma Sprayed Alumina Based Coatings,’’
J. Am. Ceram. Soc., 79 [7] 1907 14 (1996).
5. A. Pajares, L. Wei, B. R. Lawn, N. P. Padture, C. C. Berndt, ‘‘Mechanical Characterization of

Plasma Sprayed Ceramic Coatings on Metal Substrates by Contact Testing,’’ Materials Science and
Engineering., A208 158 165 (1996).
6. S. Wuttiphan, A. Pajares, B. Lawn, C. C. Berndt, ‘‘Effect of substrate and bond coat on contact
damage in zirconia based plasma sprayed coatings,’’ Thin Solid Films, 293 251 260 (1997).
7. M. V. Swain, J. Menčı̄k, ‘‘Mechanical Property Characterization of Thin Films Using Spherical
Tipped Indenters,’’ Thin Solid Films, 253 204 211 (1994).
8. H. R. Hertz, Hert’s Miscellaneous Papers; Chs. 5 and 6., Macmillan, London, 1896
9. S. P. Timoshenko, J. N. Goodier, ‘‘Theory of Elasticity third edition ,’’ McGraw Hill Book
Company, 409 415.
10. K. L. Johnson, ‘‘Contact Mechanics’’ Cambridge University Press, 84 106.
11. B. R. Lawn, ‘‘Indentation of Ceramics with spheres: A Century after Hertz,’’ J. Am. Ceram. Soc.,
81[8]1977 94 (1998).
12. H. Yoshida, S. Kano, Y. Hasegawa, T. Shimamori, and M. Yoshida, ‘‘Particle impact phenomena of
silicon nitride ceramics,’’ Philosophical Magazine A, 1996, Vol74, No.5, 1287 1297.
13. H. Yoshida, Y. Hoshi, K. Uematsu, and Y. Kitazawa, ‘‘A single, small particle launch system by
electro thermal gun and micro sabot,’’ Rev. Scientific. Instrument, 68(1), (1997)
14. S. Shimizu, PhD Dissertation, Toyoshashi Institute of Technology (in Japanese)(2001)
15. D. B. Marshall, B. R. Lawn, and A. G. Evans, ‘‘Elastic/Plastic Indentation Damage in Ceramics : The
Lateral Crack System,’’ J. Am. Ceram.Soc.,65 [11] 561 566(1982).
PREDICTING THE
RELIABILITY OF BRITTLE
MATERIAL STRUCTURES
SUBJECTED TO TRANSIENT
PROOF TEST AND SERVICE
LOADING
Noel N. Nemeth,1 Osama M. Jadaan,2 Tamas Palfi,3

and Eric H. Baker4
1. INTRODUCTION
Brittle materials today are being used, or considered, for a wide variety of high tech
applications that operate in harsh environments, including static and rotating turbine
parts, thermal protection systems, dental prosthetics, fuel cells, oxygen transport
membranes, radomes, and MEMS. Designing brittle material components to sustain
repeated load without fracturing while using the minimum amount of material
requires the use of a probabilistic design methodology. The NASA CARES/Life1
(Ceramic Analysis and Reliability Evaluation of Structures/Life) code provides a
general-purpose analysis tool that predicts the probability of failure of a ceramic
component as a function of its time in service. This capability includes predicting
the time-dependent failure probability of ceramic components against catastrophic
1
Research Engineer, Life Prediction Branch, NASA Glenn Research Center, Brookpark, OH, 44135
2
Professor, College of Engineering, Mathematics, and Science, University of Wisconsin Platteville, WI,
53818
3
Ph. D. Student, Budapest University of Technology and Economics, Budapest, Hungary
4
Research Engineer, Connecticut Reserve Technologies, 2997 Sussex Court, Stow, OH, 44224
555
556 Noel N. Nemeth et al.
rupture when subjected to transient thermomechanical loads (including cyclic loads).

The developed methodology allows for changes in material response that can occur
with temperature or time (i.e. changing fatigue and Weibull parameters with tem-
perature or time). For this article an overview of the transient reliability methodology
and how this methodology is extended to account for proof testing is described. The
CARES/Life code has been modified to have the ability to interface with commer-
cially available finite element analysis (FEA) codes executed for transient load
histories. Examples are provided to demonstrate the features of the methodology
as implemented in the CARES/Life program.
NOMENCLATURE
A ¼ slow crack growth coefficient (A1 power law, A2 Walker law)

Ae ¼ effective area
a ¼ crack length
B ¼ slow crack growth material parameter
C ¼ Shetty shear sensitivity constant
CARES ¼ Ceramics Analysis and Reliability Evaluation of Structures
fc ¼ cyclic frequency
FEA ¼ Finite Element Analysis
Ki ¼ Stress intensity factor (i ¼ I, II, III for modes I, II, or III,
respectively)
KIc ¼ critical mode I stress intensity factor (fracture toughness)
KIeq ¼ equivalent mode I stress intensity factor from applied effective stress
KIeq,cyclicmax ¼ maximum mode I equivalent stress-intensity factor over a cycle
kB ¼ normalized Batdorf crack density coefficient
m ¼ Weibull modulus (scatter parameter)
N ¼ slow crack growth exponent
Pfa ¼ attenuated probability of failure from proof test and service load
Pf ¼ probability of failure (Pf ¼ 1 Ps )
PS (tf ) ¼ probability of survival at time t ¼ tf
PSa ¼ attenuated probability of survival from proof test and service load
PSi ¼ probability of survival from proof test and service load
PSp ¼ probability of survival from proof test
Q ¼ Walker law R-ratio sensitivity exponent
R ¼ R-ratio; ratio of minimum stress divided by maximum cyclic stress
S ¼ surface area based property (indicates surface flaw analysis)
SCG ¼ Slow Crack Growth
T ¼ temperature
t ¼ time
tmin ¼ service time after proof test where no failure should occur
tmin,1,eff ¼ effective service time after proof test where no failure should occur
for an applied static stress of sIeq,1,T max ,p based on the properties of
time step 1
tp,1,eff ¼ effective proof test time for an applied static stress of sIeq,1,T max ,p
based on the properties of time step 1
Perdicting the Reliability of Brittle Material Structures 557
tq,1,eff ¼ effective time for an applied static stress of sIeq,1,T max ,p based on the
properties of time step 1 for all time steps (service and proof test)
tp ¼ proof test time
tq ¼ total time in proof testing and service
V ¼ volume, or volume based property (indicates volume flaw analysis)
Ve ¼ effective volume
Y ¼ crack geometry correction factor
x,y,z ¼ represents a location in the body of the structure
Z ¼ number of load blocks or number of cycles
Ze ¼ number of service load blocks or number of cycles
Zp ¼ number of proof test load blocks or number of cycles
D ¼ increment or difference
Dtj ¼ duration of time step j
sf ¼ represents the peak stress in a component
sIeq ¼ equivalent mode I far-field stress
sIeq,j ¼ equivalent mode I far-field stress during the step j
sIeq,k, T max ¼ maximum transformed stress over k time steps (using the properties
of step k)
sIeq,k,T max ¼ maximum transformed stress over k steps (using the properties of
step k)
sIeq,z,T max ¼ maximum transformed stress over z time steps (using the properties
of step z)
sIeq,1,T max ,e ¼ maximum transformed stress over the service load time steps (using
the properties of step 1)
sIeq,1,T max ,p ¼ maximum transformed stress over the proof test time steps (using
the properties of step 1)
sIeq,k,T max ,e ¼ maximum transformed stress over the proof test time steps (using
the properties of step k)
sIeq,z,T max ,e ¼ maximum transformed stress over the service load time steps (using
the properties of step z)
sn ¼ applied far-field stress normal to a crack face
s0 ¼ Weibull scale parameter
s0B ¼ Weibull scale parameter incorporating the effect of kB
su ¼ characteristic strength (value of sf where 63.21% of specimens
fail)
t ¼ applied far-field shear stress on a crack face
C ¼ represents a location (x,y,z) and crack orientation (a,b)
dV ¼ sin adadb
2. BACKGROUND
Ceramic materials display a stochastic strength response because of the combination

of brittleness and the random nature of the size, orientation, and distribution of
inherent microscopic flaws. In addition, the ability of ceramic structures to sustain
loads degrades over time due to a variety of effects such as slow crack growth (SCG),
cyclic fatigue, creep, and oxidation. Life prediction codes such as CARES/Life,1
CERAMIC/ERICA2, and STAU3 are available and have been demonstrated to
be successful in predicting the probability of ceramic components failing from

spontaneous catastrophic rupture when these components are subjected to multiaxial
loading and SCG conditions. For the CARES/Life code this includes a generalized
approach that enables the calculation of component failure probability when loading
and temperature varies over time. This capability is referred to as transient reliability
analysis and can be used to predict component reliability (probability of survival) for
situations such as thermal shock, startup and shutdown conditions in heat engines,
and cyclic loading. This computational methodology is an extension to the work of
Paluszny and Nichols,4 Stanley and Chau5 and Ziegler.6
For the CARES/Life program the transient reliability analysis methodology is
developed with the following capabilities:
a) Fast-fracture transient analysis (reliability analysis without SCG).

b) SCG transient modeling with SCG and Weibull parameters (including the
shape parameter, m) allowed to vary with temperature.
c) Computationally efficient algorithm to compute reliability for cyclic loading.
d) Cyclic fatigue modeling using a combined SCG and Walker fatigue law.
e) Transient proof test capability.
Three example problems are shown; 1) A disk in thermal shock to illustrate fast-
fracture transient analysis, 2) a diesel engine exhaust valve to contrast predictions for
cyclic loading and proof testing, and 3) the hypothetical effect of changing Weibull
and fatigue parameters over time on the reliability of alumina flexure bars in static
fatigue.
3. METHODOLOGY
The detailed development of the transient reliability methodology for SCG has been
described previously7,8 and will not be repeated here. Instead a highlight of these
equations will be presented as well as a proposed relation for transient proof testing.
The CARES/Life software describes the probabilistic nature of material strength
using the Weibull cumulative distribution function.9 For uniaxially stressed compon-
ents the 2-parameter Weibull distribution for volume residing flaws describes the
component fast-fracture failure probability, PfV , as
ð
1
Pf V ¼ 1 exp s(x,y,z) mV
dV (1)
s0V mV V
where V is the volume, terms that are a function of volume have V in the subscript,
s(x,y,z) is the uniaxial stress at a point location in the body, and m and s0 are the
shape and scale parameters of the Weibull distribution, respectively. The shape
parameter is a (unitless) measure of the dispersion of strength while the scale
parameter is the characteristic strength of a unit volume of material in uniaxial
tension and has units of stress volume1=m . An analogous equation based on surface
area can be shown for flaws that exclusively reside on the component surface.
However, for the sake of brevity this paper is restricted to showing volume flaw
based relations only.
Determination of the Weibull parameters comes from rupture experiments of

specimens in simple tension or flexure. Regression techniques such as least squares
and maximum likelihood have been developed that can determine these parameters
from a simplified form of Eq. (1);
Z
s(x, y, z) mV s f mV
PfV ¼ 1 exp dV
V sf s0V
mV

(2)
sf s f mV
¼ 1 exp ¼ 1 exp Ve
suV s0V
where sf is the peak stress in the specimen, su V is the specimen characteristic

strength, and Ve is known as the effective volume.
To predict component reliability for multiaxial stress states the Batdorf the-
ory10,11 is used. Batdorf theory combines the weakest link theory with linear elastic
fracture mechanics. It includes the calculation of the combined probability of the
critical flaw being within a certain size range and being located and oriented so that it
may cause fracture.
Slow crack growth refers to the stable extension of a crack over time. It results
from the combination of stress at the crack tip and chemical attack such that
chemical bonds break and the crack tip extends. At high temperatures it can also
occur from the action of stress and the loosening of viscous phases such that material
at the crack tip displaces. The crack length, a, as a function of time, t, can be
expressed as a power law12 with the following form
da(C, t) N(x, y, z, t)
¼ A(x, y, z, t)KIeq (C, t) (3)
dt
where KIeq is the equivalent mode I stress intensity factor from the applied effective
stress, (C, t) denotes a term that is a function of C (indicating a location (x,y,z) and
crack orientation described by angles a, and b) and t, and A and N are material
parameters that depend on the temperature and environment and hence become a
function of time and location but not orientation.
To take into account the time dependence of loading and material response, the
stress history for each finite element i is discretized into short time steps (Dtj ) during
which the stress and material parameters are assumed to remain constant over that
time step interval. For a specific time step j, the applied equivalent stress in element i is
given by sIeq, j , the temperature Tj , the scale parameter s0Vj , the Weibull modulus
mVj , the fatigue constant BVj , and the fatigue exponent NVj . At time Ztk (for k time
steps per cycle over Z number of cycles) the component survival probability, using the
Batdorf approach, can be expressed in it’s most computationally efficient form as8
Xn ð
Vi sIeq,k,T max NVk
PSV (Ztk ) ¼ exp {- [ [ . . . [ [( ) 2þ
i¼1
4p s0BVk
V
N
sN sN
Vj mVj (NVi 2)
Ieq; k ZDtk Ieq,1 ZDt1
Vk V1 mV 1
sIeq; j ZDtj
mVk (NVj 2)
m (NVj 2)
þ þ ... þ
N 2
NVk 2
]k mVj (NVk 2)
N 2
]j Vi NV 1 2
]1 V 1 dV]i }
BVk 1 BV 1
Vj
s0BVk s0BVj BVj s0BV
(4)
where, in Eq. (4) j ¼ (k-1), i ¼ (j-1) ¼ (k-2) and

0 1
s0Vj
s0BVj ¼ @ 1=mVj A (5)
kBVj
2
BVj ¼ (6)
Aj Y2 K IC,j
NVj 2 (N
Vj 2)
Most of the terms in these equations are functions of C and t. The fatigue parameter B
has units stress2 time. Y is the crack geometry correction factor, KIC is the critical
mode I stress intensity factor, and dV ¼ sin adadb. The term sIeq,k,T max represents the
effective stress examined over k time steps (entire load history) which yields the highest
fast-fracture failure probability5–8. It can be computed by

sIeq,k,T max ¼ Max sIeq,1,k , sIeq,2,k , . . . , sIeq,i,k , . . . , sIeq,k,k (7)
where
mmVi
sIeq,i Vk
sIeq,i,k ¼ s0BVk (8)
s0BVi
and sIeq, i, k is the transformed stress during time step i, using the properties of time
step k.
For non-cyclic loading where the entire load history is made up of one transient
cycle, Z is set equal to 1 in Eq. (4). In this case no approximation is used in the analysis
and the solution yields the most accurate results. Equation (4) was developed based on
an approach of using flaw strength and maintaining compatibility of failure probabil-
ity between discrete time steps. This methodology allows for the introduction of a
variable Weibull modulus as a function of time or temperature. The effective stress
sIeq, j represents an equivalent normal stress on the crack face caused by the combined
action of the normal stress sn, j and the shear stress tj on the crack face, oriented
normal to angles a and b which define the orientation of the flaw. A mixed-mode
fracture criterion and an assumed crack shape define the relationship for the equiva-
lent stress. A parameter, C, known as the Shetty shear-sensitivity coefficient13, is used
in CARES/Life to weight the contribution of tj relative to sn,j for the equivalent stress.
Shetty13 has found this parameter to range between 0:80 # C # 2:0. As C increases, the
response becomes progressively more shear insensitive. The term kBVj in Eq. (5) is the
normalized Batdorf crack density coefficient for volume flaws. It is used in the
reliability equation for compatibility purposes. It insures that the multiaxial Batdorf
theory collapses to the basic uniaxial Weibull equation (Eq. 1) when a uniaxial stress
state is applied for a single time step solution. It needs to be clarified that for increased
numerical accuracy CARES/Life further subdivides each of the elements into their
constituent Gaussian integration points and performs calculations in Eqs. (4)-(8)
based on these values.
3.1. Combined Fatigue and SCG Behavior
Equation (4) for component reliability is based on the power law shown in Eq. (3).
This equation is affected by the shape of the cyclic wave form and is predicted to be
less damaging than static loading - when the static load is the peak stress of the wave
form applied over the same service time as the cyclic load. Glasses seem to follow this
behavior well, while polycrystalline materials may show more complex behavior
under cyclic loading. Damage from cyclic loading can be caused by a variety of
effects, such as debris wedging or the degradation of bridging ligaments, but essen-
tially it is based on the accumulation of some type of irreversible damage that tends
to enhance the crack growth. To empirically account for cyclic effects the CARES/
Life program has implemented the Walker law14, which traditionally has been used
for metal fatigue, superimposed with the power law15
da(C, t)
¼ A1V (x, y, z, t)KIeq (C, t)NV (x, y, z, t) þ
dt
fc A2V (x, y, z, t)KIeq, cyclicmax (C, t)NV (x,y,z,t) QV (x,y,z,t)
DKIeq (C, t)QV (x, y, z, t)
(9)
where A1 , A2 , N, and Q are material constants which depend on temperature,

KIeq (C, t) is mode I equivalent stress-intensity factor at location/orientation
C, KIeq, cyclicmax (C, t) is the maximum mode I equivalent stress-intensity factor (i.e.,
the maximum value of KIeq over the cycle at location/orientation C), DKIeq (C, t) is
the range of mode I equivalent stress-intensity factor at C, and fc is the frequency.
A methodology for estimating these parameters from cyclic fatigue specimen rupture
data is described in Rahman15. The values of A1 , A2 , N, and Q in Eq. (9) depend on
location x,y,z and time t. It is assumed that da(C, t)=dt is continuous at fractions of a
cycle, that is, it is assumed that da(C, t)=dt could be computed for non-integer (real
number) cycle counts. The advantage Eq. (9) has over the Walker law for ceramic
materials is that when the range of the stress intensity factor DKIeq is zero, crack
growth is not predicted to be zero in the presence of a static load. Following the same
derivation methodologies from Jadaan et al.7 and Nemeth et al.8 an equation similar
to Eq. (4) can be constructed
Xn ð
Vi
PSV (Ztk ) ¼ exp { [ [...[
i¼1
4p
V
2 3mmVk (NVj 2)
sIeq,k,TMax NVk Vj (NVk 2)

NVk
sIeq,k 1 þ A
fc A12
(1 R) QVk
ZDtk
6 sIeq,k,TMax NVk 2 sIeq,k 7
6 þ
Vk
7
4 s0BVk N Vk
s0BVk 2
BVk 5
k
NVj 3mmViVj (N
(NVi 2)
2)
N sIeq,j,T Max Vj
sIeqVj,j 1 þ fc A2
A1 Vj (1 R)QVj ZDtj
sIeq,j 7
þ 7 þ
NVj 2 5
s0BVj BVj
j
3(NmV1V1 2)
sIeq,1,TMax NV1
sN V1
1þ A2
f c A1 (1 R) QV1
ZDt1
Ieq,1 sIeq,1 7
þ
V1
7 dV]i }
NV1 2
s0BV1 BV1 5
1
(10)
where R is the R-ratio, shown below as a function of C and t
sIeq,cyclicmin (C, t)
R(C, t) ¼ (11)
sIeq,cyclicmax (C, t)
There are limitations associated with Eq. (10) and it’s implementation in CARES/
Life. The first is that the equation represents an averaging of properties over the
cycle. There is no weighting given to whether the time step is at the peak or trough of
the cycle all time steps are weighted equally regarding their damage contribution.
Thus, Eq. (10) is not a true thermomechanical fatigue methodology for brittle
material transient reliability analysis. It is however accurate when parameters are
constant over the cycle. In CARES/Life R(C, t) is calculated by examining all the
time steps.
3.2. Proof Testing

Prior to placing a ceramic component in service, confidence that it will perform
reliably is usually demonstrated through proof testing. Ideally, the boundary condi-
tions applied to a component under proof testing simulate the conditions that the
component would be subjected to in service. Additionally the proof test loads are
designed to be appropriately greater in magnitude over a fixed time interval tp . After
proof testing, the survived component is placed in service with greater confidence in
its integrity and a predictable minimum service life, tmin , where no failure should
theoretically occur.
The attenuated probability of failure, PfaV , of a component surviving proof
testing over time tp and subjected to operational (service) loading over a time interval
te is
PSV (tq )
PfaV (tq ) ¼ 1 PSaV (tq ) ¼ 1 (12)
PSV (tp )
where tq ¼ (tp þ te ). The term PSV (tp ) is the probability of survival of a component
subjected to a proof test over a time interval denoted by tp . The term PSV (tq ) is the
probability of survival of a component subjected to a proof test over time interval tp
and service loading over time interval te . The reliability of the survived component
increases as the ratio of the proof test stress to the service stress increases.
In the CARES/Life program the computation associated with Eq. (12) is not
performed at the component level, but rather at C - a given x,y,z location and flaw
orientation a, b. This allows that the proof test loading does not have to exactly
simulate the service loading. For example, the attenuated failure probability can be
computed for the case when proof test load and the service load are applied in
different directions. Further details of this methodology are described in Nemeth
et al.1 and the extension to transient reliability analysis is subsequently described
herein. This methodology is developed by taking the transient loading condition (at
C) and re-casting it as an equivalent static stress applied over an equivalent time. The
equivalent static stress is set equal to the peak stress found over the duration of the
proof test. The weakest flaw that could survive the proof test (at C) is assumed to be
of strength just greater than the peak proof test stress (i.e. the strength of the weakest
Load
Proof Test Service Load
Figure 1. Schematic diagram of proof test and service loading versus time.
flaw at time t ¼ tp just exceeds the peak stress seen over the proof test). Therefore this
methodology does not consider a circumstance where the weakest flaw that could
survive the proof test could be lower in strength than the peak proof test load at the
end of the proof test (at time t ¼ tp ). This limitation should be acceptable since
typically the proof test is of short duration (such that SCG is minimized) and
unloading is performed quickly after peak load is attained (again so that SCG is
minimized) so that a component surviving the proof test with a weakest flaw (at C)
with a strength (significantly) less than the peak stress is not possible.
The loading history for the proof test is denoted by Zp , which represents the
number of proof test cycles, and for the service history Ze represents the number of
service load cycles. The equation for the survival probability for the proof test,
PSV (tp ), is straightforward and is identical to Eq. (4), except shown here with different
subscripts representing the proof test conditions. The component survival probability
PSV (tp ) for the proof test condition is
Xn ð mV 1
Vi
PSV (tp ) ¼ exp {
N 2
[ [Xp (C)]1 V 1 dV]i } (13)
i¼1
4p
V
where (in it’s most computationally efficient form)

( 2 NV k mV k (NV (k 1) 2)
sIeq,k,T max NV k sIeq ,k Zp Dtk mV (k 1) (NV k 2)
Xp (C) ¼ [ . . . [ [ þ N k2 ]k
s0BV k s V BV k 0BVk
9
N m
sIeqV,(k(k1) 1) Zp Dt(k 1) mVV (k(k 1) (NV (k 2) 2)
sN =
Ieq,1 Zp Dt1
V1 mV 1 (NV k 2)
2) (NV (k 1) 2)
þ NV (k 1) 2 þ ::::::: þ
mV k (NV 1 2)
]k 1 ]1 (14)
s0BV (k 1) BV (k 1) N 2
;
1 BV 1
V1
s0BV 1
The proof test time steps are from 1 to k for the first cycle of loading and a total of
kZp time steps are involved with the proof test.
The service loading time steps for the first cycle of service loading are counted
from k þ 1 to z. The authors chose this numbering scheme because they assume that
from the transient finite element analysis the first k time steps are for the proof test and
that the service loading time steps number from k þ 1 to z. Running more than one
cycle (or load block) of loading for the proof test and service loading in the transient
finite element analysis is unnecessary (its redundant). Therefore a total of (z k)Ze
time steps are involved with the service loading. The overall component survival
probability, PSV (tq ), for the combination of the proof test and the service loading is
Xn ð mV 1
Vi
PSV (tq ) ¼ exp {
N 2
[ [Xq ((C)]1 V 1 dV]i } (15)
i¼1
4p
V
where (in its most computationally efficient form)

( ( NV z N mV z (NV (z 1) 2)
2 sIeqV,zz Ze Dtz
sIeq,z,T max m 1) (NV z 2)
Xq (C) ¼ [...[[ [...[[ þ N 2
]z V (z
s0BV z Vz
s0BV z BV z
N mV (z 1) (NV (z 2) 2) N mV (kþ1) (NVk 2)
sIeqV,(z(z1) 1) Ze Dt(z 1) mV (z 2) (NV (z 1) 2) sIeqV,(kþ1)
(kþ1) Ze Dt(kþ1) m (N 2)
þ NV (z 1) 2
]z 1 þ ::::::: þ NV (kþ1) 2
Vk V (kþ1)
](kþ1)
s0BV (z 1) BV (z 1) s0BV (kþ1) BV 1
N
sN
mV k (NV (k 1) 2) mV (k 1) (NV (k 2) 2)
Ieq,k Zp Dtk
Vk
m 1) (NV k 2)
sIeqV ,(k(k1) 1) Zp Dt(k 1) m 2) (NV (k 1) 2)
þ ]kV (k þ ]kV (k1
sN Vk 2
0BV k BV k
N
V (k 1)
s0BV
2
(k 1) BV (k 1)
)
sN
Ieq,1 Zp Dt1
V1
þ::::::: þ NV 1 2
]1 (16)
s0BV 1 BV 1
The proof test time steps are from 1 to k and the service time steps are from k þ 1 to
z. Substituting Eq. (13) and Eq. (15) into Eq. (12) yields the expression for the
attenuated survival probability
Xn ð mV 1
Vi mV 1
PSaV (tq ) ¼ exp { [ [Xq (C)]NV 1 2 [Xp (C)]NV 1 2 H(C)dV]i } (17)
i¼1
4p
V
The Heaviside function H(C) in Eq. (17) was originally used in Nemeth et al.1 for
static (non-time-varying loads) loads and constant Weibull and fatigue parameters.
Its use here requires expressing variables in terms of equivalent peak static loads and
equivalent times. Equivalent time is the comparable time interval for a static applied
load that has the same amount of crack growth (or same reliability) as the transient
load (load varying with time) situation. The Heaviside function H(C) in Eq. (17) is
used where
H(C) ¼ 1; sIeq, 1,T max ,e (C)$sIeq, 1,T max , p (C)

Or, if sIeq, 1,T max , e (C) < sIeq, 1, T max , p (C) then
(18)
H(C) ¼ 1; tmin , 1,eff (C) < tq,1,eff (C) tp, 1,eff (C)
H(C) ¼ 0; tmin , 1,eff (C)$tq,1,eff (C) tp,1,eff (C)
The Heaviside function accounts for tmin ,1,eff (C), the minimum effective service time
interval in which reliability can not decrease based on the properties of time step 1. It
is obtained by satisfying the condition PSV (tq ) ¼ PSV (tp ) locally at C. For the whole
component tmin ,1,eff denotes the minimum value of tmin ,1,eff (C) when evaluated for
all C. If at any location the component proof test stress level is less than the service
stress level, then an assured minimum effective lifetime tmin ,1,eff does not exist and the
component can not be assumed to survive for any given time during service loading.
Note that tmin as used here would denote the real time equivalent of tmin ,1,eff for the
transient loading condition. The equations for tmin ,1,eff are described in the subse-
quent sections. tp,1,eff is the effective proof test time for an applied static stress of
sIeq,1,Tmax,p based on the properties of time step 1, and tq,1,eff is the effective time for
an applied static stress of sIeq,1,Tmax,p based on the properties of time step 1 for all
time steps (service and proof test).
The subscript 1 in Eq. (18) indicates values transformed to equivalent values
based on the properties of time step 1. sIeq,k,Tmax is the maximized fast-fracture stress
for the first kZp,Total time steps (the proof test time steps) expressed in terms of the
properties of time step k. The maximization procedure is identical to that described
previously (see Eq. 7 and Eq. 8) except that the maximization is performed only over
the proof test time steps 1 to k. sIeq,1,Tmax,p can be computed by
sIeq,1,T max ,p ¼ Max(sIeq,1,1 ,sIeq,2,1 , . . . , sIeq,i,1 , . . . ,sIeq,k,1 ) (19)
where
mmVi
sIeq,i V1
sIeq,i,1 ¼ s0BV 1 (20)
s0BVi
and sIeq, i, 1 is the transformed stress during time step i, using the properties of time
step 1. Similarly for the service loading time steps;

sIeq,1,T max,e ¼ Max sIeq,(kþ1),1 ,sIeq,(kþ2),1 , . . . ,sIeq,(kþi),1 , . . . ,sIeq,z,1 (21)
To solve for tp,1,eff (C) the proof test equation (Eq. 13) is re-expressed as an equiva-
lent static load sIeq,1,T max ,p (C) applied over an equivalent time tp,1,eff (C)
Xn ð NV 1
Vi sIeq,1,T Max ,p NV 1 2 sIeq,1,T Max ,p tp,1,eff NmV 1 2
PSV (tp ) ¼ exp { [ [[( ] þ NV 1 2
] V 1 dV]i } (22)
i¼1
4p s0BV 1 s0BV 1 BV 1
V
And equating Eq. (22) and Eq. (13) results in

" #
NV 1 2
s0BV 1 BV 1 sIeq,1,TMax ,p NV 1 2
tp,1,eff ¼ NV 1 Xp (C) (23)
sIeq,1,T Max ,p s0BV 1
To solve for tq,1,eff (C) the combined proof test and service load reliability equation
(Eq. 16) is re-expressed as an equivalent static load sIeq,1,T max ,p (C) applied over an
equivalent time tq, 1, eff (C)
Xn ð NV 1
Vi sIeq,1,TMax ,e NV 1 2 sIeq,1,T ,p tq, 1, eff NmV 1 2
PSV (tq ) ¼ exp { [ [ [( ) þ 2 ] V 1 dV]i } (24)
i¼1
4p s0BV 1 sNV 1 BV 1
0BV 1
V
And equating Eq. (24) and Eq. (15) assuming sIeq,1,T max ,p > sIeq,1,T max ,e , results in
" NV 1 2 #
NV 1 2
s0BV 1 BV 1
tq,1,eff ¼ NV 1 Xq (C) sIeq,1,TMax ,e (25)
sIeq,1,TMax ,p s0BV 1
It is important to point out that the term for final strength in Eq. (24) is sIeq,1,T max ,e
and not sIeq,1,T max ,q . This condition is necessary since the final strength must be
sIeq,1,T max ,e when sIeq,1,T max ,p > sIeq,1,T max ,e where sIeq,1,T max ,p > sIeq,1,T max ,e is
a stipulation of Eq. (18) in order for tmin ,1,eff (C) to exist. However, this violates the
fast-fracture condition (when there is no SCG) that failure probability can not
decrease over time. Equating Eq. (24) and Eq. (22) results in
" #
BV1 sIeq,1,T max ,e (C) NV1 2
tmin ,1,eff (C) ¼ 2 1 (26)
sIeq,1,T max ,p (C) sIeq,1,T max ,p (C)
where
tmin ,1,eff (C) ¼ tq,1,eff (C) tp,1,eff (C) (27)
To make Eq. (26) and Eq. (27) true (solving for the equlity in Eq. 27) requires
iteratively finding the value of the service time te that satisfies the equality, however,
this is computationally intensive. Fortunately, in order to perform the attenuated
reliability analysis for the component only the sign of the inequality in Eq. (18) needs
to be established locally at C. This means that iteration is not necessary. Instead Eq.
(26) is used to establish tmin ,1,eff (C), which is used on the left-hand side of the
inequality in Eq. (18). The right-hand side of the inequality in Eq. (18) is determined
by computing tq,1,eff (C) from Eq. (25) and tp,1,eff (C) from Eq. (23).
For computational efficiency in the CARES/Life program the term for Xq in Eq.
(25) is evaluated with sIeq,1,T max ,p instead of sIeq,1,T max ,e . This is conservative and
avoids having to evaluate Xq twice - once for Eq. (16) using sIeq,1,T max , p and once
for Eq. (25) using sIeq,1,T max ,e .
4. EXAMPLES
4.1. Example 1: Thermal Shocked Disk In Fast-Fracture
In this example the fast-fracture reliability response of a laser induced thermal

shocked disk made of silicon nitride is examined. The purpose is to see if the strength
response of the disks from rupture data of simple beams in uniaxial flexure can be
predicted. Also the transient fast-facture reliability predictions of the disk versus the
instantaneous fast-fracture reliability predictions of a given time step will be com-
pared in order to verify the correctness and accuracy of the numerical algorithm. This
example is derived from an international study16 involving laboratories from Ger-
many, Japan, and the United States. Phase I of that study worked to develop and
verify thermal up-shock techniques in which a disk specimen was centrally heated to
fracture by an appropriate heating source, including laser, quartz lamp, shaped
heating element, and gas torch. Phase II of the study involved a round-robin between
the three countries testing thermal fracture stress for two silicon nitrides, Allied-
Signal’s AS800 and Kyocera’s SN282. For this example disk results from the Siemens
AG organization, as provided by Rettig,17 for the SN282 material was used. These
disks were tested using the laser irradiation technique described by Kirchhoff18 and
Rettig.19 Three-point flexure bar data that was provided by Ferber20 was also used.
Thin disks 20 mm in diameter and 0.3 mm thick were centrally heated by a
800 W laser working in continuous wave mode. A large centrally heated area and a
steep temperature gradient near the edge was created, which yielded high tensile
stresses near the edge. The specimens were rapidly heated so that fracture would
occur in less than a second. The temperature versus time response across the disk was
measured with a fast scanning pyrometer. The heating time and the thinness of the
disk were chosen such that through-thickness temperature gradients (and hence
bending stresses) were negligible. Further details regarding the experimental set-up
are found in Ferber et al.,16 Kirchoff et al.,18 and Rettig.19
A total of 15 disks were fractured. The time of rupture and the radius corre-
sponding to the location of fracture was recorded for each specimen. Fracture
stresses were computed by the study participants using the temperature profile at
the instant of fracture, temperature dependent elastic modulus and thermal expan-
sion coefficient, and integral equations from standard elasticity theory. Some of the
disks were cut into three-point flexure specimens in order to independently quantify
the strength characteristics. Fourteen of these flexure specimens were tested at room
temperature with average size of 0.30 mm thick, 3.25 mm width, 15.00 mm long with
a support span of 9.44 mm. The specimen edges were not beveled.
In order to perform the CARES/Life reliability analysis an ANSYS finite
element model of the disk was prepared. The model comprised a 900 slice of the
disk and spanned 1⁄2 the thickness (1/8 of the disk is modeled). Solid elements were
used in the model. The disk was not constrained (other than to prevent rigid body
motion) and thus was freely allowed to expand. Temperature dependent thermal
expansion and elastic modulus values from Ferber et al.16 were used. The thermal
loading profiles versus time for two specimens (designated as #3 and #9) as supplied
by Rettig17 were used for the thermal loading. The temperatures were assumed
constant through the thickness. Specimen #3 had the highest time to failure of 0.65
seconds and, hence, the highest reported fracture strength of 430 MPa from Ferber.16
Specimen #9 had an intermediate fracture strength of 340 MPa from Ferber.16
Figure 2 shows the transient thermal profile for specimen #3. Figure 3 shows
the specimen #3 transient stress analysis results from the FEA. This figure shows
the tangential (circumferential) stresses versus distance from the disk center. The
tangential stresses are compressive near the disk center and become tensile towards
the disk edge. The radial stresses are compressive over the entire disk and are not
shown. The FEA analysis for specimen #3 consisted of 27 time steps ranging from
0.0 to 0.65 seconds, while the analysis for specimen #9 consisted of 15 time steps
ranging from 0.0 to 0.35 seconds. The FEA transient tangential stress results com-
pared very well with the numerical calculations performed by the Siemens group.
For the CARES/Life reliability analysis the Weibull parameters obtained from
the three-point flexure bars were used to predict the strength response of the disks.
Utilizing Eq. (2) with maximum likelihood analysis and assuming volume flaws a
Weibull modulus mV ¼ 11:96, a characteristic strength suV ¼ 612:7 MPa, and
a Weibull scale parameter s0V ¼ 453:8 MPa mm3=m was obtained for the flexure
1200
1000
Temperature [C]
800 0.65
600
400
200
0
0 2 4 6 8 10
Distance from Center [mm]
Figure 2. Disk specimen #3 transient thermal profile. Time steps range from 0.0 to 0.65 seconds. Not all
time steps are shown.
450
Tangential stress [MPa]
0.65
350
250
0.0
150
50
−50
−150
−250
−350
−450
0 2 4 6 8 10
Distance from Center [mm]
Figure 3. Disk specimen #3 tTransient tangential stress profile. Time steps range from 0.0 to 0.65
seconds. Not all time steps are shown.
bars tested at room temperature. Using these values of mV and s0V and the results of
the FEA with Eq. (4), the transient reliability response of the disk was calculated. For
fast-fracture transient analysis, the time increment terms Dt can be set to zero and
only sIeq,k,T max has to be determined. Since only one set of Weibull parameters at one
temperature are used, sIeq,k,T max in this case is simply the maximum effective stress at
C of all time steps used in the analysis.
Material strength of ceramics (and hence Weibull and fatigue parameters) are
known to be temperature dependent. Ferber16 shows this relationship for SN282,
where average strength gradually lowers as temperature increases. However, this
dependency is not considered in this analysis since only room temperature results
were available for the bars cut from the disks. The authors believe this is still
satisfactory since from Figure 2 the temperatures near the edge of the disk (where
fracture is most likely to occur) are relatively low such that deviation of material
properties from room temperature values should be limited.
Figure 4 shows the predicted failure probability versus time for specimens #3
and #9. These results are obtained using volume based analysis with the Batdorf
multiaxial methodology, a Griffith crack and C ¼ 0:82. The plot shows straight line
segments connecting the failure probability predictions for the various time steps.
Each time step is based on analysis results from the experimentally measured tem-
perature profile. The solid line shows the results of the transient analysis from Eq. (4),
1.00
0.90 Disk #3
Probability of Failure 0.80

0.70
0.60
0.50
Disk #9
0.40
0.30
0.20
0.10
0.00
0.10 0.20 0.30 0.40 0.50 0.60 0.70
Time [seconds]
Figure 4. Failure probability as a function of time for disk specimens #3 and #9. Solid line is transient
fast fracture prediction and dotted line is fast fracture prediction for a particular time step.
while the dotted line is results from fast-fracture analysis of the individual time steps.
Notice that the dotted line occasionally shows a lower failure probability than a
previous time step, while the solid line for the transient analysis correctly does not
show this trend. Also, there is a close correlation between the transient fast-fracture
results (solid line) and the single-time-step fast-fracture results (dotted line), increas-
ing confidence of the validity of the transient solution algorithm. The solid line and
the dotted line of disk #9 are virtually coincident. Disk #3 and disk #9 truncate at
different failure probabilities because each failed at a different maximum stress sf .
Another interesting observation is that disk #3 and disk #9 appear to have somewhat
different failure probability responses versus time.
Figure 5 shows the predicted failure probability response of the disk versus the
maximum stress sf in the disk. The experimentally obtained fracture stresses are
overlaid for comparison. The curve for the three-point flexure bar results represents a
line of best fit to the data as previously described (mV ¼ 11:96, and
suV ¼ 612:7 MPa) that was used to obtain the Weibull parameters used for the
disk reliability analysis (mV ¼ 11:96, and s0V ¼ 453:8 MPa mm3=m ). The solid
curve for the disk represents predictions based on the analysis of disk #3, while the
more difficult to see dotted curve is the prediction from disk #9. Notice that the disk
#9 results truncate around Pf ¼ 0:45 consistent with Figure 4. The curves for disk #3
and #9 follow nearly the same path in Figure 5, unlike the results shown in Figure 4.
A striking observation about figure 5 is the difference in median strength between the
disk and the three-point flexure bar. This primarily represents the ‘‘Weibull size-
effect’’ meaning that a component with a larger amount of volume under high stress
will have a lower average strength than a component with a smaller amount of
volume under high stress, and is a direct consequence of Eq. (2).
Another interesting item worth commenting on is the significant difference
between the Weibull modulus mV for the three-point flexure bar rupture data
(mV ¼ 11:96, suV ¼ 612:7 MPa) and that of the thermal shocked disk
(mV ¼ 6:91, suV ¼ 345:9 MPa) experimental rupture data (shown in Figure 5) as
determined by CARES/Life maximum likelihood parameter estimation. Least
squares regression (using an Excel spreadsheet) on the CARES/Life disk predictions
1
Probability of Failure 0.9 Disk 3-Point Bar

0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0 200 400 600 800
Stress [MPa]
Figure 5. Predicted failure probability of disk versus stress, sf , using the Weibull parameters estimated
from the 3 point flexure bar data. Solid line is for specimen #3 and dotted line is specimen #9. Experi
mental rupture data also shown.
curves shown in Figure 5 yields a Weibull modulus mV ¼ 8:72. This result is worthy
of note because under usual circumstances the Weibull modulus for the test specimen
and the designed component are presumed to be the same given that they are
sampling similar flaw populations and that the effective volume Ve stays constant.
Meaning, under usual circumstances, the Weibull modulus obtained from regression
of the predicted disk failure probability curve versus strength would be 11.96 - the
same value as the three-point flexure specimen data. The fact that in this case the
Weibull modulii between the predicted desk response curve and the flexure bar data
are significantly different can be understood by examining Figure 3. In the figure
notice that as time increases, the amount of volume under high tensile stress signifi-
cantly decreases. This has the effect of decreasing the effective volume Ve with time,
and through the size effect, increasing predicted failure stress sf . The overall effect of
this changing effective volume over time is to decrease the observed (apparent)
Weibull modulus for the thermally shocked disks based on sf . The fact that
CARES/Life predicts that the Weibull modulus obtained from the failure probability
versus sf curve reduces to 8.7 compared to the flexure bar value of 12.0, compares
favorably to the to the experimental disk result of 6.9. In other words, some of the
discrepancy in the Weibull modulus values between the experimental disk and flexure
bar results can be explained as a consequence of the transient thermal loads and how
they influence the stress distribution in the disk as a function of time. The difference
therefore is not necessarily indicative of some error or inconsistency. The remainder
of the difference can be explained as natural statistical variation (within 90% confi-
dence bounds).
The excellent correlation in Figure 5 to experimental results must be considered
within the context of the underlying statistics, given that the effective volumes Ve
between the three-point bend bar and the disk is large and sensitive to the value
chosen for the Weibull modulus mV . Because of the relative small number of samples
tested (in this case 14 flexure specimens) and the large size effect, using Weibull
parameters based on 90% confidence intervals from the three-point flexure bar data
could shift the disk predictions significantly to the left and right of the experimental
data. In general a good design practice would be to avoid large size-effect scalings
between specimens and components unless experimental data exists for both speci-
men and component such that data pooling practices can be taken advantage of to
obtain a set of best-fit Weibull parameters.
4.2. Example 2: Diesel Exhaust Valve In Cyclic Loading
This example, involving a heavy-duty diesel ceramic exhaust valve (Corum et al.21),
was selected to contrast failure probability predictions for power law, Walker law
and a proof test condition. The valves were made of NT-551 silicon nitride material.
Table 1 contains a summary of the Weibull and SCG parameters obtained from four-
point flexure bars at three different temperatures (Andrews et al.22). CARES/Life
was used to obtain these parameters using the raw data listed in Andrews et al.22.
Data to obtain values for the Walker law parameters did not exist, so assumed values
for QV and A2 =A1 were used.
Fifteen valves were engine tested without failure. These valves consisted of seven
longitudinally machined valves and eight transversely machined valves. The trans-
versely machined valves had been engine tested for 1000 hours while the longitudin-
ally machined valves had been engine tested for 166 hours. These valves were
subsequently tested in fast-fracture in order to examine their retained strength. The
mode of failure for both valve-machining orientations was found to be volume
induced.
Since all engine-tested valves failed due to volume flaws, the valve’s transient
reliability was based on volume analysis. Figure 6 shows the pressure variation as a
function of time during a typical combustion cycle of 0.0315 seconds. The pressure is
applied to the valve’s face and other exposed surfaces within the cylinder. The
maximum attained pressure during the combustion cycle was estimated to be
15.85 MPa (Corum et al.21). A 445 N (100 lb) force due to spring preload is applied
to the valve stem when it is in the open position. At the moment the valve closes an
impact force of 1335 N (300 lb) is applied to the valve stem. In addition, thermal
stresses due to the temperature distribution in the valve are superposed to the
mechanical stresses.
Figure 7 shows the approximate mean thermal profile in the valve. Steady-state
thermal analysis using ANSYS FEA code was conducted to compute these temper-
atures. This figure shows that the temperature is maximum near the valve face and
decays towards the valve seat and stem.
Transient reliability analysis (using Eq. 4 and Eq. 10) was conducted by dividing
the load history into 29 time steps. During each step, the load was assumed constant.
The loads corresponding to these time steps were modeled within the ANSYS FEA
Table 1. NT551 fast fracture and SCG material properties.
T(0 C) mV s0V (MPa mm3=m ) su (Mpa) NV BV (MPa2 sec) QV A2 =A1
20 9.4 1054 806 31.6 5.44e5 3.2 0.65
700 9.6 773 593 86.5 1.12e4 3.2 0.65
850 8.4 790 577 18.5 1.13e6 3.2 0.65

2500
2000
Pressure (psi)
1500
1000
500
0
0 0.01 0.02 0.03 0.04
Time (sec)
Figure 6. Pressure variation applied to the face of a ceramic valve during a typical engine combustion
cycle.
Figure 7. Mean thermal profile of the valve.
program, which yielded the stress results for these 29 time steps (stress history).
Figure 8 highlights the first principal thermomechanical stress distribution in the
valve at the moment of maximum applied pressure (at time step 6). From the figure it
is apparent that the maximum stress location is at the valve radius, which is in
agreement with the FEA results of Corum et al.21.
The valve’s stress history and other relevant terms (temperature, volume, ma-
terial properties, element number, etc.) were subsequently read into CARES/Life.
Figure 9 shows the predicted failure probability as a function of the number of cycles
for various scenarios. None of the tested valves failed, 8 of the 15 of which were
tested to 1000 hours (which corresponds to approximately 1:1 108 cycles). Thus the
failure rate of the tested valves was less than 1 out of 8 at 1000 hours of operation.
From Figure 9 it can be seen that in fast-fracture (at one cycle with no SCG) about 5
out of 100,000 valves are predicted to fail from the loading. With SCG and the power
law about 20 out of 100,000 valves fail after 1000 hours operation. If we conserva-
tively assume that a worst-case load (time step 6) is applied over the whole engine
cycle (static loading), then about 40 out of 100,000 valves fail after 1000 hours. Using
the combined Walker power law with the hypothetical parameters from Table 1 this
rate increases to 70 out of 100,000 valves at 1000 hours. All of these rates were well
Figure 8. First principal stress distribution in the valve at the moment of maximum applied pressure
(MPa).
0.01
Probability of Failure
Combined Walker & Power Law

0.001
Static Analysis (Step 6)
Power Law
0.0001
0.00001
100 101 102 103 104 105 106 107 108 109 1010 1011
Cycles
Figure 9. Failure probability versus cycles for proof test (dotted lines) and non proof test (solid lines)
conditions for a static load, power law, and combined law.
below what was experimentally observed. These three scenarios were presented in
order to contrast the predictions. The power law with cyclic loading predicts the least
damage. The static loading scenario is predicted to be more damaging at double the
rate of the cyclic loading. The combined Walker-power law predicts the most damage
due to the enhanced cyclic fatigue effect.
Since the effect of cyclic loads is unknown, a more conservative assumption is to
assume static loading, however even this assumption can be un-conservative.
To reduce the predicted failure rate even further, proof testing can be applied to
prevent the weakest components (those with the highest likelihood of failing) from
being placed into service. Figure 9 also shows the predicted results for the three
various scenarios for an applied proof test of 10,000 cycles at a factor of 1.1 of the
service loading (dotted lines in the figure). Increasing the proof test load reduces the
attenuated probability of failure even further.
4.3. Example 3: Alumina In Static Fatigue - Material Properties Changing

With Time
In this example it is examined how the reliability response of a vitreous bonded

alumina is hypothetically affected by Weibull and fatigue parameters that change
over time. The two scenarios that are presented are strictly based on curve fitting the
data. The authors do not have the sufficient information necessary to physically
specify which parameters are really responsible for the change in material response
with time. However, the point of this analysis is to present an analytical model
which potentially can explain the nonlinear and changing rupture behavior of
the material given that the physical parameters that are responsible for the changing
material behavior are known. This capability is potentially useful in modeling
materials with changing composition (or changing physics of crack growth), includ-
ing oxidation and crack blunting/healing phenomena.
The data for the example was obtained from Quinn23 and consists of rupture
lives of alumina four-point flexure bars in static fatigue (loading at a constant stress
level over time). The specimens had average dimensions of 2.2 mm height, 2.8 mm
width, a load span of 19.0 mm, and a support span of 38.0 mm. This data set was
chosen from the literature because of it’s ‘‘non-ideal’’ behavior and the careful
experimental technique reported by the author, which reduces the likelihood that
equipment and measurement errors significantly affected the results. The rupture
data for the individual specimens is shown in Figure 10. Testing was performed at
10008C. The report stated that very little creep deformation was detected, and
although fractography was attempted, it did not reveal the source of the strength
limiting flaws.
Two trends make this data non-ideally behaved: 1) dramatically increased
scatter at lower applied stresses, and 2) pronounced nonlinear behavior in the stress
rupture data as the applied stresses became lower (the strength data shifted to the
right which signified that the lives became longer than would have been predicted if
the standard power law was applicable). An ideally behaved data set would show a
100
Specimen Data
Runout Data
Maximum Failure Stress (MPa)
90
80
70
60
10% 50% 90%
50 0
10 101 102 103 104 105 106 107 108
Time-to-Failure (Seconds)
Figure 10. Affect on failure probability of changing fatigue exponent, N, with the log of time from Table
2 on static fatigue rupture data of an alumina at 10008 C.
straight line trend indicating that the fatigue exponent is constant, also the scatter
band would appear with a constant width versus the applied stress indicating the
Weibull modulus is constant.
A confirmatory piece of evidence that material properties were changing with
time was obtained by performing static rupture experiments on specimens that were
annealed (at no load) for either 1 or 24 hours at a temperature of 10508C. The
outcome of these treatments (not shown herein) was that life dramatically increased
versus exposure time for an applied stress level (Quinn23). Quinn23 indicated that the
observed behavior is probably due to the partial devitrification of the material’s
glassy phase, resulting in an increased viscosity of the phase and a material with
improved creep and stress rupture behavior. Quinn cites Wiederhhorn et al.24 to
support this, although he concedes that crack blunting, healing, or residual stress
changes could also be operative.
The data of Quinn defies the conventional modeling approach since the material
properties could be changing over time. To account for this the authors show what
happens when Weibull and fatigue parameters change with time for two cases; (1) chan-
ging the fatigue exponent N and the Weibull modulus m over time, and (2) changing the
Weibull parameters (m and s0 ) over time. Tables 2 and 3 show the sets of parameters
chosen to demonstrate these scenarios. Given the fact that no established parameter
estimation techniques currently exist for this type of non-ideally behaved data, these
parameters were iteratively selected to fit the data. Table 2 contains a set of parameters
versus time where the fatigue exponent, N, and the Weibull modulus, m, were varied to
yield an improved fit to the data. Table 3 shows a set of parameters where the Weibull
modulus, m, is varied while the fatigue exponent, N, is kept relatively constant. Note
that in both cases the scale parameter s0 is dependent on the Weibull modulus, and B is
dependent on the fatigue exponent as well as the Weibull parameters. When applying
these parameter sets within a reliability analysis, the Weibull and fatigue parameters are
linearly interpolated with the log of time within the time spans listed in the tables and
are held constant outside of the time span.
Because of the simplicity of the four-point specimen loading and geometry,
results from FEA were not needed for the reliability analysis. Instead, a closed
form expression for the effective area, Ae , was used to evaluate the integral of Eq.
Table 2. Weibull and fatigue parameters associated with Figure 10.
t (seconds) mS (s0S )(MPa mm2=m ) NS BS (MPa2 sec)
1.6 29.4 156.8 6.7 2711.1
31.6 15.8 152.7 13.2 9707.7
1.0e5 13.1 127.3 36.4 2276.2
Table 3. Weibull and fatigue parameters associated with Figure 11.
t (seconds) mS (s0S )(MPa mm2=m ) NS BS (MPa2 sec)
1.6 29.4 165.8 6.7 2711.1
31.6 7.4 263.3 8.0 2395.9
316.2 4.5 870.1 9.0 10,389.0

(2) for a surface flaw failure mode (Nemeth et al.25). An effective area of 58:0 mm2
was calculated based on a Weibull modulus value of 7.7. This value was chosen
because sensitivity analysis using Eq. (2) indicated that the maximum error in
strength for a given failure probability would be 1% or less for Weibull modulus
values ranging between 5.0 and 30.0. This level of error is negligible for the illustra-
tive purposes of this example. To model the transient material response 10 time steps
are used per decade of the log of the time. Hence 10 time steps are used between 1 and
10 seconds, while 20 time steps are used between 1 and 100 seconds. The time steps
are log increments of time that is, they would appear as equally spaced increments
in Figure 10 with 10 steps per decade.
Figure 10 shows the predicted 10%, 50%, and 90% failure probability iso-lines
for the parameters shown in Table 2, while Figure 11shows 1%, 10%, 30%, 50%, 70%,
90%, and 99% failure probability iso-lines for the parameters shown in Table 3.
Neither plot represents an optimized set of fitted parameters. As previously stated no
established parameter estimation techniques currently exist for this type of non-
ideally behaved data. Parameters were obtained by segregating portions of the data
and performing parameter estimation as described in Nemeth et al.1 for the four-
point bending bar specimen as well as further refinements by trial-and-error. Figure
10 shows that the curvature in the data can be captured by changing the fatigue
exponent, N, however, accounting for the change in scatter still required modifying
the Weibull modulus, m. Figure 11 is interesting because a satisfactory fit to the data
can be obtained primarily by changing the Weibull modulus and scale parameter
only (B is changing mainly in response to these parameters). The ‘‘kink’’ shown in the
failure probability iso-lines between 10 and 100 seconds is not purposely modeled
that is, parameter values were not specifically selected to obtain this response. The
‘‘outlier’’ rupture data (data that appears to be not part of, or not consistent with, the
main body of data) at long times to failure at 70 MPa and the ‘‘outlier’’ data at short
times to failure at 60 MPa are better accounted for in this model compared to Figure
10. That is, data that visually appears as ‘‘outliers’’ are actually consistent with the
100
Specimen Data
Maximum Failure Stress (MPa)
90 Runout Data
80
70
60 1% 10% 30% 50% 70% 90% 99%
50 0
10 101 102 103 104 105 106 107 108
Time-to-Failure Seconds)
Figure 11. Affect on failure probability of changing Weibull modulus, m, with the log of time from Table
3 on static fatigue rupture data of an alumina at 10008C.
flaw population failure probability response in other words they are predicted to be
there.
The modeling assumptions that produced the results in Figure 11 reasonably
capture all of the trends in the data as well as or better than the approach used in
Figure 10. Also, both approaches required changing the Weibull modulus m to
account for the range of scatter. A changing Weibull modulus could be indicative
of new flaw generation, or it could be indicative of changes of the physics of crack
growth associated with R-curve behavior. On the other hand a changing fatigue
exponent with time is consistent with crack blunting/healing phenomena, where the
crack velocity relationship in Eq. (3) is changing with time. In this case further tests
on the material would be needed to understand the underlying mechanism that was
driving the SCG behavior.
5. CONCLUSIONS
A methodology for computing the transient reliability in ceramic components sub-

jected to fluctuating thermomechanical loading was developed and incorporated into
the CARES/Life code. This enables CARES/Life to be used to predict component
reliability for situations such as thermal shock, startup and shutdown conditions in
heat engines, and cyclic loading. The methodology accounts for varying material
response, whether due to temperature or environmental changes by allowing Weibull
and fatigue parameters to vary over the loading history. Examples demonstrating the
viability of the technique for fast-fracture, cyclic loading, and proof testing, were
presented.
6. ACKNOWLEDGEMENTS
We would like to thank Dr. Uwe Rettig and Dr. Matt Ferber for providing data from
the IEA study on thermally shocked disks, and Dr. John Gyekenyesi for his support
and technical direction over the years. We also thank Debbie Haught, DOE office of
Industrial Technologies, and Carol Ginty, manager, NASA Hot Propulsion Com-
ponents program.
REFERENCES
1. N. N. Nemeth, L. M. Powers, L. A. Janosik, and J. P. Gyekenyesi, Time Dependent Reliability

Analysis of Monolithic Ceramic Components Using the CARES/LIFE Integrated Design Program,
Life Prediction Methodologies and Data for Ceramic Materials, ASTM STP 1201, C. R. Brinkman, and
S. F. Duffy, Eds., American Society for Testing and Materials, Philadelphia, 1993, pp. 390 408.
2. A. D. Peralta, D. C. Wu, P. J. Brehm, J. C. Cuccio, and M. N. Menon, Strength Prediction of Ceramic
Components Under Complex Stress States, Allied Signal Document No. 31 12637, 1995.
3. A. Heger, Stau Programm Berech,’’ Ph. D. Thesis, Karlsruhe University, Karlsruhe, Germany, 1991.
4. A. Paluszny, and P. F. Nicholls, ‘‘Predicting Time Dependent Reliability of Ceramic Rotors, Ceramics
for High Performance Applications II, J. Burke, E. Lenoe, and N. Katz, Eds., Brook Hill, Chesnut Hill,
Massachuestts, 1978.
5. P. Stanley, and F. S. Chau, A Probabilistic Treatment of Brittle Fracture Under Non monotonically
Increasing Stresses, International Journal of Fracture, Vol. 22, 1983, pp. 187 202.
6. C. Ziegler, Bewertung der Zuverlassigkeit Keramischer Komponenten bei zeitlich veranderlichen

Spannungen und bei Hochtemperaturbelastung, Ph.D. Thesis, Karlsruhe University, Karlsruhe,
Germany, 1998.
7. O. M. Jadaan, and N. N. Nemeth, Transient Reliability of Ceramic Structures, Fatigue & Fracture of
Engineering Materials and Structures, Vol. 24, 2001, pp. 475 487.
8. N. N. Nemeth, and O. M. Jadaan, Transient Reliability of Ceramic Structures For Heat Engine
Applications, Proceedings of the 5th Annual FAA/Air Force/NASA/Navy Workshop on the Appli
cation of Probabilistic Methods to Gas Turbine Engines, June 11 14, 2001, Westlake Ohio.
9. W. A. Weibull, A Statistical Theory of the Strength of Materials, Ingenoirs Vetenskaps Akadanien
Handlinger, No. 151, 1939.
10. S. B. Batdorf, and J. G. Crose, A Statistical Theory for the Fracture of Brittle Structures Subjected to
Nonuniform Polyaxial Stresses, Journal of Applied Mechanics, Vol. 41, No. 2, June 1974, pp. 459
464.
11. S. B. Batdorf, and H. L. Heinisch Jr., Weakest Link Theory Reformulated for Arbitrary Fracture
Criterion. Journal of the American Ceramic Society, Vol. 61, No. 7 8, 1978, pp. 355 358.
12. S. M. Wiederhorn, Subcritical Crack Growth in Ceramics, Fracture Mechanics of Ceramics, R. C.
Bradt, D. P. H. Hasselman, and F. F. Lange, Eds., Vol. 2, Plenum Press, New York, NY, 1974,
pp. 613 646.
13. D. K. Shetty, Mixed Mode Fracture Criteria for Reliability Analysis and Design with Structural
Ceramics, Journal of Engineering Gas Turbines & Power, Vol. 109, No. 3, July 1987, pp. 282 289.
14. K. Walker, Effects of Environmental and Complex Load History on Fatigue Life, ASTM STP 462,
American Society for Testing and Materials, Philadelphia, PA., 1970, pp. 1.
15. S. Rahman, N. N. Nemeth, and J. P. Gyekenyesi, Life Prediction and Reliability Analysis of Ceramic
Structures Under Combined Static and Cyclic Fatigue, 1998 ASME Turbo Expo Land, Sea, and Air,
Stockholm Sweden, June 2 5, 1998, Paper No. 98 GT 569.
16. M. Ferber, G. Kirchhoff, T. Hollstein, R.Westerheide, U. Bast, U.Rettig, and M. Mineo, Thermal
Shock Testing of Advanced Ceramics Subtask 9. International Energy Agency Implementing
Agreement For a Programme of Research and Development on High Temperature Materials for
Automotive Engines, prepared for The Heavy Vehicle Propulsion System Materials Program Oak
Ridge National Laboratory for the U.S. Department of Energy, M00 107208, March 2000.
17. U. Rettig, Personal communication, August, 2002.
18. G. Kirchoff, M. Holzherr, U. Bast, and U. Rettig, Thermal Shock and Thermal Cycling Behavior of
Silicon Nitride Ceramics, Proceedings of the International Conference on Silicon Nitride Based
Ceramics, October 4 6, 1994, Stuttgart, Germany. Published by Trans Tech Publications Ltd,
Switzerland, in Key Engineering Materials, Vol. 89 91, 1994, pp. 605 610.
19. U. Rettig, Laserschock Mebmethode Ein Quantitatives Verfahren Zur Charakterisierung Der
Temperature Wechselbestandigkeit Hochfester Keramischer Werkstoffe, Eingereicht an der Fakultat
Maschinenwesen der Technischen Universitat Dresden zur Erlangung der Wurde Eines Doktors der
Ingenieurwissenschaften (Dr. Ing.), October 16, 1997.
20. M. Ferber, Personal communication, July, 2002.
21. J. Corum, R. Battiste, R. Gwaltney, and C. Luttrell, Design Analysis and Testing of Ceramic Exhaust
Valve for Heavy Duty Diesel Engine, Oak Ridge National Laboratory, ORNL/TM 13253, 1996.
22. M. A. Andrews, A. A. Wereszczak, T. P. Kirkland, and K. Breder, Strength and Fatigue of NT551
Silicon Nitride and NT551 Diesel Exhaust Valves, Oak Ridge National Laboratory, ORNL/
TM 1999/332, 1999.
23. G. D. Quinn, Delayed Failure of a Commercial Vitreous Bonded Alumina, Journal of Materials
Science, Vol. 22, 1987, pp. 2309 2318.
24. S. M. Wiederhorn, B. Hockey, K. Krause Jr., and K. Jakus, Creep and Fracture of a Vitreous
Bonded Aluminum Oxide, Journal of Material Science, Vol. 21, 1986, pp. 810.
25. Nemeth, N. N., Powers, L. M., Janosik, L. A., and Gyekenyesi, J. P., ‘‘Ceramics Analysis and
Reliability Evaluation of Structures Life Prediction Program Users and Programmers Manual.’’
NASA Glenn Research Center, 1993.
QUANTIFICATION OF
TOUGHNESS INCREASE DUE
TO METAL PARTICLES IN
GLASS MATRIX COMPOSITES
M. Kotoul, A. R. Boccaccini and I. Dlouhy1
1. INTRODUCTION
A number of examples exist that indicate the potential for increasing the toughness of
brittle matrix by dispersing a ductile phase. The basics of the mechanism of fracture
toughness enhancement were amply analyzed in the literature. Several models were
developed which commonly predict the increase of toughness DGc as
ðc
D
DGc ¼ f sp (D)dD, (1)

0
where sp (D) is the T/S law traction, with the integral representing the plastic stretch-
ing work to rupture of the particle and f is the volume fraction (area fraction on
fracture plane) of the particles. It was found that toughening improves with increas-
ing bridging length. However, it was also proved that increasing bonding imposes
lateral constraints on the particles thus prohibiting the full utilization of the particles
inherent ductility (see also Kotoul and Urbis, 2001). As a result, the critical crack
opening displacement is significantly reduced due to the influence of the stress
triaxiality factor. Plastic strains develop only in a thin layer as it can also be seen
from FEM simulation of the deformation of the particle constrained by elastic
1
M. Kotoul, Dept. of Mechanics of Solids, Brno University of Technology, 616 69 Brno, Czech Republic
(corresponding author [email protected]), A. R. Boccaccini, Department of Materials, Imperial
College London, London SW72BP, UK, I. Dlouhy, Institute of Physics of Materials ASCR, 61662 Brno,
Czech Republic
245
246 M. Kotoul et al.
matrix. The bridging zone length is significantly reduced and the assumption, that
bridging zone contains a large number of particles, ceases to hold. In such a case the
action of the system of discretely distributed particles cannot be replaced by the
action of smeared forces over the bridging zone length, the critical value Dc becomes
very uncertain and the model in Eq. (1) fails. Particularly in the limit, when the crack
is bridged just by one row of particles distributed along the crack front - so called
crack trapping, the details of an interaction of the crack front with the particles
become of growing importance for an assessment of the toughness increment. The
change of the toughening mechanism from crack bridging (extrinsic toughening) to
crack trapping (intrinsic toughening) closely relates to the extent of debonding which
depends both on internal factors like debonding toughness, particle size and/or
volume fraction, and temperature. The paper is focused on a calculation procedure
that is capable of discerning subtle changes in fracture toughness behavior of very
brittle material reinforced by ductile metallic particle. Rice’s perturbation analysis is
used to the configuration of the front of a planar crack that is trapped against
forward advance by a periodical array of closely spaced bridges. The solution is
lifted by analogy from known solutions of a periodic array of through microcracks
along the z-axis collinear with the undisturbed crack front and an assessment of the
toughness increase is provided.
2. EXPERIMENTAL FACTS
Dlouhy et al. (1997) observed that experimental values of the fracture toughness of
borosilicate glass reinforced by vanadium or molybdenum particles, their shape can
roughly be treated as spherical one, have fallen far behind to the theoretical expect-
ations of the enhancement of fracture toughness according to the crack bridging
view. Specifically, the experimentally found toughening ratio (composite toughness/
matrix toughness) was about 1.4 for the volume fraction f ¼ 0:1 of vanadium
particles while the theoretical prediction based on the crack bridging model accord-
ing to Eq. (1) is more than one order higher. When the observed particle cleavage in
about 85 vol% of the metal inclusions was taken into account, the theoretical
prediction of the toughening ratio somewhat decreases but still amounts about
5 times higher value. It should be noted that the observed particle cleavage indicates
a decrease in particle plasticity induced due to two causes: (i) constraints imposed by
the rigid matrix and (ii) embrittlement due to the high-temperature fabrication
process. Uncleaved particles delivered some toughening due to crack bridging, but
extensive stretching was not detected and, as a consequence, the bridging zone length
was significantly reduced, see Fig. 1.
Choe et al. (2001) reported in their study on Mo-12Si-8.5B alloy that crack
trapping by Mo-particles appears to provide the main source of (intrinsic) toughen-
ing at lower temperatures. As the vast majority of the a-Mo particles intersected by
the crack did not remain intact in the crack wake, there was insignificant crack
bridging and hence little crack-growth toughening. The failure of the Mo phase
was quite brittle, whereas at 13008C, the Mo particles debonded from the
Mo3 Si=Mo5 SiB2 matrix with significant amounts of plastic stretching. Such in-
creased ductility of the Mo particles did result in some degree of wake bridging in
the Mo-12Si-8.5B alloy at elevated temperatures.
Quantification of Toughness Increase Due to Metal Particles 247
Figure 1. a) Crack in inclusion and partial debonding; b) crack in small particle.
3. ENERGY CONDITION FOR DEBONDING
Before describing a trapped crack front by a row of particles distributed along the
crack front, some place will be devoted to the analysis of debonding along the
particle/matrix interface. The aim is to estimate the length h of debonded interface
of a particle in the vicinity of crack tip under assumption that h R (average radius
of particles), i.e. e. for a relatively strong interface. Due to mechanical constraint the
plastic strain then develops only in a thin layer of thickness 2h as it can also be seen
from FEM simulation of the deformation of the particle constrained by elastic
matrix, see Fig. 2. The length h will be estimated from energy considerations. Enclose
the debonded layer of particle by a surface J which forms the surface of a cylinder
of the height 2h and the radius R. As the thickness of the layer 2h increases, the part of
the surface J migrates in the y-direction accompanied by axisymmetric debonding
of particle/matrix interface. The aim is to calculate the potential energy loss associ-
ated with the growth of thickness of the debonded layer. To do so, we apply the
general result derived by Eshelby (1970) concerning the global change of the Gibbs
potential of prestressed body due to the possible migration of the interface J between
the bulk and a part of material which undergoes any kind of transformation. Eshelby
has shown that a virtual change of the Gibbs potential dP associated with a virtual
displacement dji at each point of J can be conveniently expressed in terms of the
energy momentum tensor of the elasticity Pij
Figure 2. FEM simulation of the elastic plastic particle constrained by elastic matrix illustrating a layer
undergoing a nonzero axial plastic strain increment D«yy .
Z
dP ¼ djj [Pij ] ni dJ (2)
J
where ni dJ is an oriented surface element and [Pij ] is the jump across the interface J
of the Eshelby tensor component, Pij
Pij ¼ gdij ski uk,j : (3)
In Eq. (3) g stands for the Helmholtz free energy density in the case of isothermal
change, dij is the Kronecker symbol, skj is the Cauchy stress tensor components, uk
denotes the displacement vector components and ukj ¼ @uk =@xj is the deformation
gradient. In the context of the problem investigated, P reduces to the potential
energy and g to the strain energy density. In the cylindric coordinate system (r,w,y)
originated in the particle centre Eq. (2) reduces to the following form
8 9
>
<Z Z >
=
dP ¼ djr [g srr ur,y syr uy,y ]nr dJ þ djy [g sry ur,y syy uy,y ]ny dJ
>
: >
;
J1 J2
(4)
where J1 stands for the area of the cylinder (where debonding takes place) and J2
stands for the foots of the cylinder, see Fig. 3. The virtual displacement djy dh
describes a virtual change of the layer thickness while the virtual displacement
@uM (y;h)
djr ¼ r@h dh, where uM r (y;h) is the radial displacement of the debonded matrix
face. The jump across the interface J of the Eshelby tensor component, Pij is defined as
[Pij ] ¼ PM
ij Pij
P
(5)
where the superscript ‘M’ refers to limiting values of Pij as the interface J is
approached along ni from the matrix whereas the superscript ‘P’ refers to limiting
values of Pij as the interface J is approached along ni from the particle. Note that ni
is oriented from the bulk to the particle. The boundary J1 is traction-free. Thus
Figure 3. Scheme of debonding model.

Eq. (4) rewrites

8 9
>
< Z Z >
=
dP @uM (y;h)
¼ gM dJ þ þ (sM u M
ry r,y s P P
u
ry r,y þ s M M
u
yy y,y s P P
u
yy y,y ) dJ ,
dh > : @h >
;
J1 J2
(6)
where the strain energy density was already assumed to be continuous across J2 due
to the continuity of tractions and displacements across J2 . The first term represents
@
R M
the rate of the change in the strain energy @h g dV of the material volume V
V
enclosed by the surface S [ J1 [ J2 due to the motion of the boundary J1 . In the
case of an elastostatic boundary-value problem associated with the material con-
tained within the surface S [ J for which external loading is imposed only by
~ on S, the rate of the change in the elastic strain energy can be inferred
traction T
from the Clapeyron theorem of elasticity
Z Z Z
@ @ 1 1 @ui
g dV ¼
M
Ti ui dS ¼ Ti dS: (7)
@h @h 2 2 @h
V (h) S S
It allows to estimate the first term in Eq.(6). Similarly as in Budiansky et al. work
(1985) we suppose that stresses in the vicinity of the crack tip debond the particle/
matrix interface just ahead of advancing crack for a distance h on each side of the plane
of the crack. Due to this symmetry with respect to the plane y ¼ 0 it is sufficient to
investigate only the upper part of the cylinder. Assume that axisymmetric debonding is
produced by an axisymmetric distribution of load applied to the matrix cylinder that
jackets the particle. The magnitude of this loading is conservatively taken as the stress
KIC
srr (y) ¼ p , (8)
4 pjyj
which is just the asymptotic distribution of horizontal tension just above the crack-
tip and KIC is the critical stress intensity factor of the matrix. The stress distribution
from Eq. (8) is applied at the mean radius Rm of the matrix jacket
1
Rm ¼ (l þ R), (9)
2
where l is half of the interparticle distance. The cylindrical surface of the radius Rm
just corresponds to the surface S in Eq. (7). The deformation of the matrix will be
analyzed on the basis of thick-cylinder theory, in which only transverse shear stresses
and circumferential tension resist radial displacement. In terms of the matrix cylinder
thickness
t¼lR (10)
and the mean radius Rm the differential equation governing the radial displacement
r (y) at r ¼ Rm is
uM
Em t d 2 uM
r Em t KIC
þ 2 r ¼ p
uM , (11)
2(1 þ nm ) dy 2 Rm (1 nm )
2
4 pjyj
where Em and nm is the Young modulus and Poisson’s ratio of matrix respectively. The
r ( h) ¼ 0 can be written in the form
solution of (11) with the boundary conditions uM
3=2 1=4
1 KIC (1 n2m ) Rm 2 y
uM (y;h) ¼ p ~u H;H , (12)
r
4 p Em t 1 nm h
where
1=2
h 2
H¼ (13)
Rm 1 n m
and
y
y ZhH ZH
sinh( yh H h) cosh yh H sinh(H h)
~
u H;H ¼ p dh þ p dh: (14)
h h coshH h
0 0
The solution (12) allows to reduce Eq.(7) to the form
Z Z Zh
@ 1 @uM
r KIC @uM
r
g dV ¼
M
srr (y) dS ¼ 2pRm p dy ¼
@h 2 @h 4 py @h
V (h) S 0
(15)
1=2 ZH
GIC R2m 2 1 @~u
¼ p ds,
8 t 1 nm s @H
0
where GIC ¼ KIC2

(1 n2m )=Em is the critical matrix crack extension energy release
rate. Because R=2l ¼ (3f =4p)1=3 and
ZH
@~
u cosh yh H coshh
¼ p dh, (16)
@H cosh2 H h
0
the change in the elastic strain energy finally becomes

h i2
Z 1 4p 1=3
( ) þ 1 1=2
1 @ GIC 2 3f 2
g dV ¼
M h i Q(H)2pR, (17)
2 @h 128p 1 ( 4p )1=3 1 1 nm
V (h) 2 3f
where
8H 92
> R >
> p dh>
coshh
>
< h >
=
0
Q(H) ¼ : (18)
>
> coshH >
>
>
: >
;
and the symmetry with respect to the plane y ¼ 0 was included. The terms sM M
ry ur, y
M M
and syy uy, y in the second integral in Eq.(6) can be derived from the asymptotic
distribution of vertical tension, shear and displacement just above the crack-tip.
Including the above-mentioned symmetry we obtain after some algebra
Z p
1 GIC (3 4nm ) 1 2
(sry ur, y þ syy uy, y )dJ ¼
M M M M h i 2pR: (19)
2 4p(1 nm ) 3=2 1
( ) þ1 H
4p 1=3
J2 2 3f
Remaining terms in the second integral in Eq. (6) refer to the particle. Also the energy
absorbed in the process of plastic deformation within the plastic layer is a matter of
interest. For h R it seems reasonable to approximate the stress-strain fields in the
debonded plastic layer by Prandtl’s solution for thin circular plastic layer connected
with rigid blocks. Prandtl’s solution, outlined in Appendix A, gives
uPr, y (h) ¼ uPy, y (h) ¼ 0, from where it follows that the terms in the second integral in
Eq. (6), referring to the particle, are zero. By inspection of Eqs. (6), (17) and (19)
there is observed the loss in the potential energy of the loaded matrix jacket dP dh < 0
due to debonding. The energy supply is partly dissipated by plastic work in the
debonded layer and partly is absorbed in the mechanism of particle/matrix interface
debonding. The virtual plastic work dWp exerted in the half of the layer is
ZR hZþdh
dWp ¼ 2p (sPrr d«Prr þ sPww d«Pww þ sPyy d«Pyy þ 2sPry d«Pry )rdrdy ffi
0 0
(20)
ZR Zh
ffi 2p (sPrr d«Prr þ sPww d«Pww þ sPyy d«Pyy þ 2sPry d«Pry )rdrdy,
0 0
@«
where d«ij ¼ @hij dh and a linear approximation with respect to the virtual displace-
ment dh was adopted. At this point is useful to introduce a dimensionless composite
parameter k.
r
KIC p
k¼ (21)
2sy 2R
which combines the fracture toughness of the matrix KIC with the yield strength sy
and the radius of particles R. The parameter k allows to handle further analysis in
dimensionless form. Using the results in Appendix A, Eq. (13) and the definition of
the parameter k, one can write Eq. (20) as follows
ZR Zh " 0 00
#
dWp psy @uy @uy
1 GIC (3 4nm )
¼p (h y )
2 2
yr 2
p drdy ¼ 2pR
dh 3h @h @h k 8 6 1 nm
0 0
8 9
>
> " #1=2 3=4 >
>
>
<14 1 n 1=4 1 4p 1=3 1 8 >
=
m 3 1 n 1
þ1 H þ :
1=2 m
>
3=2 H 3=2 >
(22)
>15 8 2 3f 1=3 >
>
: 1 4p
þ1 >
;
2 3f
Denote by Gd the critical energy release rate for debonding of the particle/matrix
interface; then the energy release rate relation
dWp dP
2pRGd þ ¼ (23)
dh dh
provides the condition governing the debonding length h. The results are presented
by the pair of parametric equations
" #
h 1 4p 1=3 1 nm 1=2
¼ þ1 H,
R 2 3f 8

2
1=3
1 4p
2 3f þ1 1=2
Gd 1 2
¼
Q(H)þ (24)
GIC 128p 1 4p 1=3 1 nm
2 3f 1
p
2(3 4nm ) 1 1 1 1 (3 4nm )
þ
p R(H),
4p(1 nm ) 3=2 1=3 H k 8 6 1 nm
1 4p
2 3f þ1
h Gd
which relate R to GIC , where
" #1=2
14 1 nm 1=4 1 4p 1=3
R(H) ¼ þ1 H 1=2 þ
15 8 2 3f
3=4
1 8 1 1
þ
3=2 3=2 :
3 1 nm 1=3 H
1 4p
2 3f þ1
see Fig. 4. Note that equations (24) imply that Rh is not a single-valued function of GGICd .
The upper part of each curve represents stable debonded lengths, but for debonding
to occur at all an energy barrier associated with the region between the horizontal
axis and each lower branch must be overcome.
Gd
The results in Fig. 4 indicate that there exist threshold values
GIC of the ratio
of debonding to fracture toughness
such that for Gd
GIC > Gd
GIC debonding can not
Gd
occur. The threshold value GIC depends on the composite parameter k and on the
Figure 4. Debond length vs. debond toughness/matrix toughness ratio a) for several values of the
composite parameter k and the particle volume fraction f 0:15, b) for several values of the particle
volume fraction f and the composite parameter k 2.

Gd
Figure 5. Threshold value GIC as function of the composite parameter k for several values of the
particle volume fraction f.

particle volume fraction f. Fig. 5 summarizes threshold values GGICd as a function of
the composite parameter k for several values of f that would prevent debonding.
Note that for a constant value of the matrix fracture toughness the composite
parameter k increases with decreasing
particle yield stress and/or particle size. As a
consequence, threshold values GGICd increases with a decrease of the particle
yield stress and/or particle size. The results
can be envisaged also in another way -
specifically, assume that the ratio GGICd is virtually constant while the particle yield
stress decreases with increasing temperature. Apparently, a composite system can
then pass from no debonding region to debonding region.
h
This is a matter of interest to calculate
the threshold debond lengths R
corresponding to the threshold values GGICd along the curves GGICd vs. k shown
in Fig. 5. The results of these calculations are plotted in Fig. 6. It is interesting to note
that for a composite system with the particle volume fraction f $0:30 the threshold

debond length Rh dramatically changes due to a small change of k within the range
about of 1:75 1:85.
Figure 6. Threshold debond lengths ( Rh ) corresponding to the threshold values ( GGICd ) .

4. TRAPPED CRACK FRONT MODEL
Assume that the front of a planar crack is trapped against forward advance by a
periodical array of closely spaced bridges. As a result, the location of that front
differs from a straight line. Following the paper by Rice (1985) consider a half plane
crack in the y ¼ 0 plane, growing in the direction x and having a straight front along
x ¼ a0 , see Fig. 7a. Denote the stress intensity factor along the straight crack front
parallel to the z axis by KI0 (z;a0 ) which is given by
KI0 (z;a0 ) ¼ KIN (a0 ) KIR (z) (25)
where KIN (a0 ) is the remote stress intensity factor, KIR (z) is the restraining stress
intensity factor for periodical array of closely spaced bridges and can be readily
found, see Appendix B. If the crack front is not straight but lies along the arc x ¼ a(z)
in the plane y ¼ 0, then Rice (1985) has shown that, provided a(z) a0 is small such
that in an average sense the crack front is still straight and the variation of KI0 (z;a0 )
with a0 , i.e. @KI0 =@a0 is much smaller that KI0 itself, the stress intensity factor
KI (z;a(z) ) along the moderately curved crack front is
Z1 da(z0 )
KI (z;a(z) ) KI0 (z;a0 ) 1
¼ 0
dz0
dz0 (26)
0
KI (z;a0 ) 2p z z
1
to the first-order deviation of a (z) from a0 , i.e. to the first order in da (z) / dz. Next
consider the situation depicted in Fig 7c. The crack front interacts with physical
ligaments of some given distribution. Then a (z) is known along the bridged parts of
the crack front Lbrid but KI (z) is unknown there. Conversely, KI (z) ¼ KIC is known
along the penetration zones Lpen (i.e. the matrix part) but the depth of penetration
a (z) is unknown there. These two conditions together reduce (26) to the singular
integral equation
Z da(z0 ) Z da(z0 )
KI0 (z;a0 ) KIC 1 01
dz0
0
dz ¼ dz0
dz0 (27)
0
KI (z; a0 ) 2p zz 2p z z0
Lbrid Lpen
for all z included in Lpen . Once (27) is solved for a (z) along Lpen , KI (z) along Lbrid can
be found from (26). The solution of (27) can be obtained by analogy from known
solutions in two-dimensional crack problems and is shown in Appendix C. The
average value of KI (z; a(z) ), denoted by hKI (z)i over the bridged crack front
(l R < jzj < l þ R) is
x x x
2A
a0 a(z)
Lbird Lpen
z z z
(a) (b) (c)
Figure 7. Scheme of the trapped crack front by a periodical array of closely spaced bridges.
l[KIN hKIR (z 2 Lpen )i] (l R)KIC

hKI (z)i ¼ : (28)
R
The limiting value of KI0 denoted as before (KIN )eff corresponds to the instant at
which the local stress intensity factor along the bridged crack front hKI (z)i hKIR
(z 2 Lbrid )i just attains the value KIC . Eq. (28) then provides
N
KI R hKIR (z 2 Lbrid )i þ hKIR (z 2 Lpen )i
¼1þ : (29)
KIC eff l KIC
As appropriate approximations for hKIR (z 2 Lbrid )i and hKIR (z 2 Lpen )i respectively

we choose the value of the restraining stress intensity factor computed at the centre of
ligaments given by Eq. (B5) and the value of the restraining stress intensity factor
computed in between the ligaments given by Eq. (B7) respectively. When the ratio R/l
is transformed to the particle volume fraction f and the parameter k defined through
Eq. (21) is introduced, one can put Eq. (29) into the form
r " #
KIN R 2 1 8 1 3f 1=3
¼1þ f þ (p þ 4 arctan(3) 2 arctan(2) ln (5) ) , (30)
KIC eff h 3 3k 3 p 4p
5. DISCUSSION
Consider a specific normalized value of the critical energy release rate for debonding
Gd
GIC . Then, as Fig. 5 shows, the transition from debonding to no-debonding regime
takes place for a set of combination of values of the parameter k and the particle
volume fraction f. Each of these combinations relates to a particular value of the
h
debond length
R, see Fig. 6. Choosing f as the independent variable, the toughening
KIN
ratio KIC in Eq. (30) can be plotted vs. f for various values of GGICd , see Fig. 8. It is
eff
seen that the contribution from the crack trapping diminishes with increasing f when
the bridging mechanism progressively takes over. It would appear that the transition
in dominant mechanisms depends on GGICd ; while for GGICd ¼ 0:18 it occurs at f 0:15,
for GGICd ¼ 0:65 it occurs at f 0:38.
Now compare the experimental toughening data reported by Dlouhy et al.
(1997) with theoretical predictions of the model proposed. Note that borosilicate
Figure 8. Toughening ratio at the threshold from debonding to no debonding regime.

p
glass matrix possesses a typical fracture toughness value of only KIC ¼ 0:7 MPa m.
The average size of the vanadium inclusions was 2R 6 mm and their yield strength
sy 450 MPa. When these values are used in the definition of the composite par-
ameter k in Eq. (21), the value of k 0:6 is obtained. The results shown in Fig. 5
indicate that a debonding toughness that is a quite small fraction of the matrix
fracture toughness suffices to suppress debonding for such a low value of k and the
volume fraction f ¼ 0:1. However, if debonding do takes place then the results
shown in Fig. 6 indicate that a significant extent of debonding can be expected in
this case. Apparently, statistical changes in both k and the debonding toughness from
particle to particle result in a random occurrence of (i) complete debonding, (ii)
partial debonding and (iii) no debonding. This may explain the observed fracture
features in the glass composites investigated by Dlouhy et al. (1997): -interfacial
debonding (corresponds to the case (i)), -limited crack bridging (corresponds to the
case (ii)) and -particle cleavage (corresponds to the case (iii)). It seems that the
investigated composite systems were in the regime of transition from extrinsic tough-
ening to intrinsic toughening. Fig. 5 indicates that for k ¼
1 and f ¼ 0:1 the thresh-

Gd
old value of the ratio of debonding to fracture toughness GIC is about of 0.17. The
KIN
lowest curve in Fig. 8 shows the plot of the toughening ratio KIC eff against f for a
very close value of Gd
¼ 0:18. It is seen that the calculated toughening ratio for
GIC
f ¼ 0:1 is in a surprisingly good agreement with the experimental value 1.4. The
agreement is even better if the fraction of particles effectively interacting with the
crack (i.e. taking into account only uncleaved particles) is reduced down to
0:02 0:03.
6. CONCLUDING REMARKS
The model proposed is capable to explain experimental observations concerning the

transition from intrinsic toughening controlled by second phase crack trapping to
extrinsic toughening controlled by second phase crack bridging even when the ratio GGICd
is virtually constant. The transition relates to the particle yield stress which decreases
with increasing temperature. It was also
found that the threshold
value of the ratio of

debonding to fracture toughness GIC , such that for GIC > GIC debonding cannot
Gd Gd Gd
occur, increases with the particle volume fraction

p f. Thus,
p for a fixed value of the
dimensionless composite parameter k ¼ KIC p= 2sy 2R and a fixed value of GGICd
particles more readily debond at a higher particle volume fraction f, see Fig. 5. This
may explain why the contribution from the crack trapping diminishes with increasing f
when the bridging mechanism progressively takes over, see Fig. 8. The diminishing
contribution from the crack trapping is mathematically incorporated in Eq. (30) via
the dependence of the threshold debond lengths ( Rh ) on f, see Fig. 6.
ACKNOWLEDGMENTS:
The support through the grant GAČR 101/02/0683 is gratefully acknowledged.

REFERENCES
Budiansky, B., Hutchinson, J.W., Evans, A.G., 1985, Matrix cracking in fiber reinforced ceramics,
J. Mech. Phys. Solids 34:167.
Dlouhy, I., Reinisch, M., Boccaccini, A.R. and Knott, J.F., 1997, Fracture characteristics of borosilicate
glasses reinforced by metallic particles, Fatigue Fract. Engng. Mater. Struct. 20:1235.
Choe, H., Chen, D., Schneibel, J.H. et al., 2001, Ambient to high temperature fracture toughness and
fatigue crack propagation behavior in a Mo 12Si 8.5B intermetallic, Intermetallics 9:319.
Eshelby, J.D., 1970, The energy momentum tensor in continuum mechanics. In: Inelastic Behavior of
Solids, M.F. Kaninen et al., ed., McGraw Hill, New York, pp.26 33.
Kotoul M. and Urbis, R. 2001, On the crack stability in metal particulate reinforced brittle matrix
composites, In: Proc. of ICF10, A.T. Yokobori, R.O. Ritchie, K. Ravi Chandar and B.L. Karihaloo,
eds., Elsevier, Oxford, CD.
Rice, J.R., 1988, Cracks fronts trapped by arrays of obstacles: solution based on linear perturbation
theory, In: Analytical, Numerical and Experimental Aspects of Three Dimensional Fracture Process,
A. Rosakis, K. Ravi Chandar and Y. Rajapakse, eds., 91, ASME, p. 175.
Tada, H., 1985, Handbook for Stress Analysis of Cracks, 2nd ed., Del. Research Corp., 1985.
Venkateswara Rao K.T., Soboyejo, W.O., Ritchie, R.O., 1992, Ductile phase toughening and fatigue
crack growth in Nb reinforced molybdenum disilicide intermetallic composites, Metallurgical Trans
actions A 23A:2249.
APPENDIX A
Assume a thin ideally plastic layer connected with two rigid blocks. Cross sections of
the layer at y ¼ 0, y ¼ h stay planar. Since the layer is thin, it is reasonable to
accept the hypothesis of planar cross sections, i.e. uPy ¼ uPy (y). The equation of
incompressibility
@uPr uPr @uPy

þ þ ¼0 (A1)
@r r @y
provides the solution for uPr which matches the condition uPr jr¼0 ¼ 0
1 0
uPr ¼ uPy (y)r (A2)
2
Resorting to the deformation theory of plasticity we have
@uPr uPr uPr @uPy @uPy @uPr @uPr @uPy
@r r r @y @y @r 1 @y þ @r
¼ ¼ ¼ (A3)
sPrr sPww sPww sPyy sPyy sPrr 2 sPry
and using (A2) we obtain

0 0 0
0 3 uPy 3 P
uy 1 uPy r
¼ P 2 P ¼ P2 ¼ : (A4)
srr sww sww syy syy srr
P P P 4 sPry
From (A4) it follows sPrr ¼ sPww everywhere in the layer. Equilibrium equations now
reduce to the form
@sPrr @sPry
þ ¼0
@r @y
(A5)
@sPry sPry sPyy
þ þ ¼0
@r r @y
and von Mises condition
(sPyy sPrr )2 þ 3(sPry )2 ¼ s2y : (A6)
Boundary conditions at r ¼ R will be fulfilled in Saint-Venant’ sense
Zh Zh
sPry dy ¼ 0, sPrr dy ¼ 0: (A7)
h 0
(A7)1 requires sPry to be an odd function of y and in view of relation h R the

solution for sPry is sought in the form
sy y
sPry ¼ S( r) p , (A8)
3h
where S( r) is an unknown function. Now analyze (A4) at y ¼ h. From (A6) one gets
for y ¼ h in the case of tension
p
sPyy sPrr ¼ sy 1 S2 (r): (A9)
Substitute (A9) into (A4) and obtain

p P0 00
3 uy uPy r
p ¼ : (A10)
1 S 2 (r) 2S(r)
Introduce an arbitrary parameter C

00
!
1 uPy
p ¼C (A11)
2 3 uPy 0
y¼h
and find the distribution of shear stress along y ¼ h as
Cr
S¼p : (A12)
1 þ C 2 r2
The equilibrium equation (A5)1 then gives the solution for sPrr as
ZR
sy
sPrr ¼p Sdr þ b ( b ¼ const), (A13)
3h
r
where b can be found from the boundary condition (A7)2 . At the outset of plastic
deformation C ¼ 0 which leads to S ¼ 0. For small values of C the function S(r) is
approximately linear and so is the contact shear stress sPry (r, h). As C increases, S(r)
deflects from the linear distribution and, in the limit C ! 1, which corresponds to
the fully plasticpconditions, S ¼ 1. The contact shear stress sPry approaches the shear
yield stress sy = 3 and Eqs. (A7)2 , (A9) and (A13) provide
sy
sPrr ¼ p (R r) ¼ sPyy (h): (A14)
3h
The average normal stress sP acting over the cross section of the layer is
Z2p ZR
1 sy R
s ¼
P
sPyy rdrdw ¼ p : (A15)
pR2 3 3 h
0 0
APPENDIX B
Assume a periodical array of bridging particles along the crack front and replace
each particle by square shaped ligament with side length of 2R. Assume a constant
cohesive stress sP acting over the bridging particles and use the solution (Tada, 1985)
for the stress intensity factor K for point forces of magnitude P applied to both faces
of the crack at a horizontal distance z and a vertically downwards distance l from the
crack front
p p
P 2 l
K¼ p : (B1)
p p(l2 þ z2 )
Then we obtain for the total restraining stress KIR at the centres of the bridged zones
of crack front (z ¼ ( 12 þ R þ 2nl ), n ¼ 0,1::)

p 1 Z
2R p p P ZR
l 4 2s P R X ldl 2s
KIR z ¼ þ R þ 2nl ¼ p þþ p
2 p p n¼1 l þ (2nl)
2 2 p p
0 R
0 1
Z2R p
ldl A
@ dz, (B2)
l þ z2
2
0
where the first term describes the restraining effect of other ligaments while the
second term describes the restraining effect of the very ligament being contiguous
with the place of evaluation, i. e. along R < z < R. Changing the order of integra-
tion and summation in the first term allows (B2) to be expressed as

Zx " 3=2 # p
l 2sP R cothx 0 1 0 2sP R
KIR z ¼ þ R þ 2nl ¼ p p 0 dx þ p [p
2 p l x x0 p p
0
þ 4 arctan(3) 2 arctan(2) ln (5)], (B3)

where x ¼ pRl #1. The integral in (B3) must be treated in the sense of principal value.
Thus, to the order O (x7=2 ) finally obtain

p
l p 2R 2 p 2sP R
KIR z ¼ þ R þ 2nl ¼s p
P
R þ p [p þ 4 arctan(3)
2 3l p p : (B4)
2 arctan(2) ln (5)]
A cohesive stress sP can be estimated from the Prandtl’s solution for thin circular
plastic layer connected with rigid blocks as it is shown in Appendix A, see Eq. (A15)
In a similar fashion one can evaluate the restraining stress intensity factor in
between the ligaments, i.e. at z ¼ 2nl, n ¼ 0, 1.. obtaining
p 1 Z
2R p
4 2s P R X ldl
KIR [z ¼ 2nl ¼ p
p p n¼1 l þ [(2n 1)l]2
2
0
(B5)
Zx
2sP R tanhx 0 0
¼ p p 0 dx ,
p l x
0
which to the order O(x 5=2 ) gives

2
p 2R p
KIR [z ¼ 2nl] ¼ 3s P
p R: (B6)
3l
APPENDIX C
Consider a two-dimensional medium in (y, z) plane containing an array of cracks.

The medium is loaded remotely in mode I such that the stress syy ¼ sappl corresponds
to KIN and the opening gap between y ¼ 0þ and y ¼ 0 is 2uy (z) ¼ d(z). Then syy (z)
along z axis is obtained as
Z1 dd(z0 )
E 1
syy (z) ¼ s appl
þ 0
dz0
dz0 : (C1)
1 n 4p
2 z z
1
Eq. (C1) may be rearranged to coincide exactly with (26) if one makes the identifi-
cations
1 n2 KI (z) KI0
2 syy (z) sappl !
E KI0 : (C2)
d(z) ! a(z)
In this analogy, along Lbrid the opening displacement is d(z) ¼ a(z), whereas along
K K0
Lpen the crack faces sustain the stress syy (z) ¼ sappl þ 1 En2 IC2K 0 I . The bridges with
I
centre-to-centre spacing 2l have a gap 2(l R) between them, where the crack front
can propagate. This configuration is reminiscent of a periodic array of through
cracks consisting of cracks of length 2(l R) separated by unbroken ligaments of
length 2R. Its response to a mode I stress field can be determined analytically (see,
e.g. Tada et al. 1985), with the y-axis displacement between y ¼ 0þ and y ¼ 0 given
by
q
þ 2 p(l R)
2l cos
pz
8l(1 n )s
2 appl cos 2l cos2 pz 2l
d(z) ¼ ln : (C3)
E cos p(l2l R)
So with identifications made in (C2) and realizing that syy (z) ¼ 0 along Lpen receive
for (l R) < z < (l R)
q
cos pz þ cos2 pz cos2 p(l R)
4l KIC 2l 2l 2l
a(z) ¼ d(z) ¼ 1 0 ln p(l R)
: (C4)
p KI cos 2l
substituting (C4) into (27) one obtains along the bridged crack front
Lbrid (l R < jzj < l þ R) the following stress intensity factor
j sin pz
2l j
KI (z, a(z) ) ¼ KIC þ (KI0 KIC ) q , (C5)
p(l R)
cos2 2l cos2 pz
2l
which is singular at the borders of the bridged zone. The average value hKI (z)i is then
given by Eq. (28) in the main body of text.
ROOM TEMPERATURE
POST-FRACTURE TENSILE
(PFT) STUDY OF TWO
FIBER-REINFORCED
CERAMIC MATRIX
COMPOSITES WITH A
‘‘STRONG’’ FIBER/MATRIX
INTERFACE
Rafael A. Longoria, Nicolas Du Souich,

Kenneth W. White
1. INTRODUCTION
Elevated temperature applications require materials that can maintain good mech-
anical properties such as strength and hardness. Ceramics have good mechanical
properties at high temperature and, thus, appear to be good candidates for elevated
temperature applications. However, due to their brittle nature, monolithic ceramics
are unsuitable for many applications where reliability is a critical issue. In the last few
years, a new class of ceramic materials has been developed and studied. It is under-
stood that two brittle materials can show non-brittle behavior if they are properly
mixed. Fiber-reinforced ceramic matrix composites (CMCs) exhibit pseudo-plastic
behavior at room temperature, as well as in an elevated temperature environment.
Since the fiber and the matrix are made of ceramic material, creep behavior and
hazardous emissions are reduced considerably.
307
308 Rafael A. Longoria et al.
The first CMCs to be developed consisted of three major components: a

ceramic matrix, fibers embedded in the matrix, and a tailored interface between
the fiber and the matrix.1 Although these materials show damage tolerance and
non-brittle behavior, the non-oxide materials that compose the CMCs are prone to
oxidation, especially when matrix cracks are present.2 Lately, the development of
an all-oxide CMC has captured the researchers’ attention. In these oxide/oxide
composites, fracture toughness is achieved through crack deflection inside the
matrix. A controlled level of matrix porosity will provide suitable conditions for
crack deflection while inherently impeding oxidation during high temperature
service.3
2. MATERIALS
Two all-oxide composites were studied. A sol-gel derived alumina matrix composite
produced by COI Ceramics that is reinforced with Nextel 720 (3M Company) fibers.
The matrix porosity level is 40% and the fiber volume fraction is 47%.4 The
processing route is shown in Figure 1. A sintering process between 1800 2100 8F
controls porosity levels. The second ceramic composite was a mullite/alumina matrix
composite developed by the University of California at Santa Barbara with the same
reinforcing fibers. The porosity level for this material is 40% with a fiber volume
fraction of 37%.5 The processing route is shown in Figure 2. A repeated precursor-
impregnation process is responsible for the porosity.
Initial damage (crack) was introduced by using a Double Cantilever Beam (DCB)
configuration. Aluminum fixtures were used to avoid delamination of the composites
during the DCB test. Grooves on both sides of the DCB specimen assured symmetric
Figure 1. Processing route for the Alumina matrix composite.4

Room Temperature Post-Fracture Tensile (PFT) 309
3-D Ceramic Mullite Alumina

ALCL(OH)5 Woven Fiber Powder Powder
Solution Preform Cloth (−1μm) (−0.1μm)
Cutting &
De-sizing Slurry Preparation
Lay-up
Slurry Infiltration
Vacuum-Assisted
Precursor Matrix Bonding

Impregnation Drying
Treatment (900°C)
Pyrolysis Sintering Composite

(900°C) Treatment (1200°C) Product
Figure 2. Processing route for the mullite/alumina matrix composite.5
Figure 3. a) DCB testing configuration. b) PFT specimens obtained from DCB specimen.
crack propagation during the test. The DCB test was deliberately interrupted before
the crack reached the opposite side of the specimen. The DCB configuration provides
a wide range of crack openings. PFT specimens were obtained from the cracked DCB
specimen. The initial COD characterizes each PFT specimen.6,7 Figure 3 shows a
schematic for both the DCB configuration and the PFT specimen distribution.
PFT specimens are numbered from 1 to 8. The PFT with the smallest COD
becomes PFT-1. COD measurements were performed using optical microscopy.
A Static PFT test (monotonic loading) was performed on different PFT specimens.
The crosshead speed was 200 10 6 inch/min. Cyclic PFT test required a sinusoidal
loading profile with relatively small amplitude. A suitable frequency and amplitude
for the oscillation was obtained by using a piezo actuator. The increase in COD was
tracked with a video extensometer for both tests.
4. RESULTS
A family of PFT curves can be obtained by plotting the Static PFT data of PFT
specimens with different initial COD. The influence of characteristic (initial) COD
becomes noticeable by using this type of graph. An inversely proportional bridging
load with respect to initial COD is clearly observed. Furthermore, all the curves show
a good agreement in the strain-softening region, proving the universality of the PFT
test. Figures 4 and 5 show a family of PFT curves for the alumina composite and the
mullite/alumina matrix composite, respectively.
The overlapping of curves for final stages of the separation event validates the
existence of a universal strain-softening behavior (exponent), independent of the
initial condition of the specimen. This observation can be corroborated by a load-
ing-reloading static PFT test on the same specimen.
Cyclic PFT gave a better insight about the failure mechanism. Due to the nature
of the PFT test, where only tensile loading is possible, the energy required to close a
cycle loop arises solely from the stored elastic energy. Therefore, the existence of
enough stick points with sufficient strain energy will return the crack to its initial
state. This behavior is observed in Figure 6, where a PFT-2 specimen of the alumina
matrix composite was subjected to one loading cycle. The maximum load was
controlled to 0.9 kg.
Although, the hysteresis loop is fully close for cycle one of a PFT-2 specimen,
gross slip was observed for later stages of the cyclic test. Figure 7 shows the
comparison of cycle 1 between a PFT-2 specimen and a PFT-3. As the initial
compliance depends on the total number of contact points, a nearly constant number
of points remain between PFT-2 and PFT-3.
5
Nominal Stress (MPa)
0
0 200 400 600 800 1000
Displacement (micrometers)
Figure 4. Family of PFT curves for the Alumina matrix composite.

12
10
Bridging Stress (MPa)
0
0 200 400 600
Displacement (mm)
Figure 5. Family of PFT curves for the Mullite/alumina matrix composite.
0.9
0.8
0.7
Kilogrammes
0.6
0.5
0.4
0.3
0.2
0.1
0
0 1 2 3 4 5 6 7 8 9
Micrometers
Figure 6. First cycle for a PFT 2 of the alumina matrix composite.
Although the initial compliance remains similar between the two specimens, the
area fraction increases significantly. The slip-to-stick contact point ratio increases
considerably for the level of damage in PFT-3. The appearance of a sharp transition
point indicates that the number of stick points converted to slip contacts is large
enough to induce gross slip at that load. The insufficient number of stick contact
points at the maximum load yields to a non-recoverable displacement.
Figure 8 shows three subsequent cycles for a PFT-7 specimen of the mullite/
alumina matrix composite. The transition point between partial slip and gross slip
0.9
0.8
0.7
C2 = 4.5 microns/kg C3 = 4.8 microns/kg
Kilogrammes
0.6
0.5
0.4
0.3
0.2 PFT 2
PFT 3
0.1
0
0 5 10 15 20
Micrometers
Figure 7. Comparison of cycle 1 between PFT 2 and PFT 3 specimen of the alumina matrix composite.
0.132
0.13
0.128
kilograms
0.126
0.124
0.128
0.12
0.118
0 2 4 6 8 10
micrometers
Figure 8. Three subsequent cycles for a PFT 7 specimen of the mullite/alumina composite.
occurs at a higher load at each cycle. This is an indication of a rearrangement of

contact points leading to a different bridging configuration. Since new portions of
fiber are sliding after gross slip, each cycle brings a slightly increased friction
coefficient due to fiber/matrix interaction. There is an increasing load during gross
slip for this material. This suggests an increasing fiber/matrix interaction during
sliding.
Long-term fatigue behavior was also studied. The hysteresis loop after each
cycle was observed to behave stochastically. As the crack opening displacement
grows, the hysteresis loop opens and closes, suggesting that after gross slip takes
place, an appropriate configuration of stick points can be formed. Although inter-
facial wear might take place, the new portions of fiber sliding after gross slip
apparently provide enough interfacial friction and contact points to fully close the
hysteresis loop. This new configuration will remain active until wear diminishes the
contact surfaces enough to induce gross slip again. Therefore, interfacial properties
play a significant role in fiber pullout under both monotonic and cyclic loading.
SEM analysis led to the identification of different microstructural features
between the two composites. The alumina matrix composite exhibited single fiber
pullout after complete fiber pullout, while fiber bundle pullout dominated in the
mullite/alumina composite. Comparing Figures 4 and 5 in this work, one can see a
larger displacement to total failure for the alumina matrix composite than for the
mullite/alumina, suggesting that fiber bundle pullout represents a more severe failure
mode. Figure 9 and 10 show the fiber pullout mode for the alumina matrix and for
the mullite/alumina matrix composite, respectively.
5. CONCLUSIONS
Room temperature fiber pullout under monotonic and cyclic loading was success-
fully investigated by using the PFT test configuration. Static (monotonic) PFT test
led to the identification of a common strain-softening exponent between PFT speci-
Figure 9. Single fiber pullout observed for the alumina matrix composite.
Figure 10. Fiber bundle pullout observed for the mullite/alumina matrix composite.
mens. This suggests the existence of a displacement-controlled failure mechanism for

this new type of fiber-reinforced ceramic matrix composites with a strong fiber/
matrix interface. An inversely proportional maximum bridging stress with respect
to COD was found for both composites. Gross slip under cyclic loading was observed
for both composites, suggesting a reduced number of contact points due to interfacial
wear. However, the stochastic behavior of the hysteresis loops during long-term
fatigue experiments suggested a reconfiguration of the fiber/matrix interacting sur-
face after gross slip. In addition, SEM analysis led to the identification of different
pullout mechanisms between composites. The data suggests bundle pullout is a more
severe pullout mode.
REFERENCES
1. T. Radsick, B. Saruhan, and H. Schneider, ‘‘Damage tolerant oxide/oxide fiber laminates composites,’’
Journal of the European Ceramic Society 20, 545 550 (2000).
2. K. K. Chawla, Z. R. Xu, J. S. Ha, M. Schmucker, and H. Schneider, ‘‘Effect of BN coating on the
strength of a mullite type fiber,’’ Applied Composite Materials 4, 263 (1997).
3. W. Tu, F. F. Lange, and A. G. Evans, ‘‘Concept of a Damage tolerant Ceramic Composite with
‘‘Strong’’ Interfaces,’’ Journal of the American Ceramic Society 79, 417 424 (1996).
4. http://www. coiceramics.com/cmcdevelopment.htm COI Ceramics, San Diego, California, USA.
5. C. G. Levi, J. Y. Yang, B. J. Dalgleish, F. W. Zok, and A. G. Evans, ‘‘Processing and Performance of
an All oxide Ceramic Matrix Composite,’’ Journal of the American Ceramic Society 81, 2077 2086
(1998).
6. R. D. Geraghty, J. C. Hay, and K. W. White, ‘‘Fatigue Degradation of the Crack Wake Zone in
Monolithic Alumina,’’ Acta Materialia 47, 1345 1353 (1999).
7. J. C. Hay, and K. W. White, ‘‘Crack Face Bridging Mechanisms in Monolithic MgAl2 O4 Spinel
Microstructures,’’ Acta Metallurgica and Materialia 40, 3017 3025 (1992).
SCRATCH TEST FOR
EVALUATION OF SURFACE
DAMAGE IN GLASS
Satoshi Yoshida,* Takashi Hayashi,* Tetsuro Fukuhara,*

Kazuki Soeda,* Jun Matsuoka,* and Naohiro Soga**
1. INTRODUCTION
Scratch behavior in glass is very important to understand the interaction between the
abrasive particles and the glass product during grinding. In addition, the chipping
problem in glass often appears during the glass manufacturing process. A thin sheet
glass, such as an LCD(Liquid Crystal Display) or PDP(Plasma Display Panel) glass,
is susceptible to scratches, abrasions, rubbing, and chipping, which are mechanical
failures caused by a dynamical contact between the glass and a foreign body. In order
to suppress these contact problems and to develop a glass with high wearability, it is
essential to understand the mechanical response to a scratch as the cumulative
actions of a hard material. It is considered that a scratch test is one of the funda-
mental testing methods to evaluate these mechanical properties. However, only a few
papers have been published on scratch behaviors in glass.1–3 Especially, the origin of
scratch groove formation has little been understood. The formation of scratch
groove is probably related to microscopic fracture, plastic deformation, and material
removal. The scratch test surely gives us some important information on the dama-
geability of glasses with various compositions under various conditions.
It is well known that fracture properties in glass are affected by environment.
Stress corrosion cracking is believed to be the result from a stress-enhanced chemical
reaction between a chemical environment, usually water, and the Si-O-Si bonds.4
* Department of Materials Science, The University of Shiga Prefecture, Hikone, Shiga 522 8533, Japan
** National Institute of Advanced Industrial Science and Technology, Chiyoda Ku, Tokyo 100 8921,
Japan
101
In addition, Westwood and Huntington5 reported that both drilling rate and pendu-
lum hardness depended on the number of carbon atoms in n-alcohols and n-alkanes,
because these processes included fracture behaviors. Therefore, scratch behaviors of
glass also depend on these environments.
In this study, surface damage in glass is evaluated by using a scratch testing
equipment. The purpose of this study is twofold; (1) to evaluate the differences
between scratch hardness and conventional hardness, and (2) to evaluate the compos-
itional variation of the crack initiation load determined through a dynamic-loading
scratch test. It is considered that the crack initiation shows a fracture criterion at the
transition point from apparent plastic deformation to macroscopic fracture. The crack
initiation load will give some useful information on brittleness of glass.
A scratch hardness tester was assembled from two load cells, two moving stages, and a
Knoop indenter. A schematic of this measuring system is shown in Fig. 1. In this study,
a Knoop indenter was used as a scratching element. Normal load on the specimen was
detected with a load cell, which was contact with the indenter and driven by a stepping
motor capable of pulling and pushing at a constant speed. Normal load was controlled
by changing the position of the indenter. Horizontal load was measured by another
load cell connected with another moving stage. The stage was moved in the direction of
the long diagonal of Knoop indenter. In this study, the constant speed of the stage,
0.05 mm/s or 0.07 mm/s, was adopted for the scratch test. It has been reported
that scratch hardness depends on the scratching speed, especially at the speed of
>0.1 mm/s,1 so the scratch speed was fixed at the above value. Scratch hardness is
defined as the load per unit load-bearing area (Fig. 2)6 and calculated from the normal
load and the groove width using the following equation,
HS ¼ FN =A ¼ 0:5622 (FN =w2 ), (1)
where FN is the normal load, A the vertical projection of contact area, and w the
groove width. The constant in the right side can be determined from the geometry
of indenter. In the case of Knoop indenter, the angle between the opposite edges
is 172.58 and the other angle between the other two is 1308. By using the groove
width, w, the load-bearing area, A, becomes 1.779 w2 .
On the other hand, Jacobsson et al.6 also defined the plowing stress, HP , which is
the mean pressure opposing the sliding of the element due to the plowing process only.
HP ¼ FP =AP ¼ 1:124 (FP =w2 ), (2)
where FP is the horizontal load and AP the area of resultant groove cross-section. In
this study, however, because of a large scatter of FP data and its less dependence of
composition, only HS will be employed for discussions on the compositional vari-
ation of scratch hardness.
For the determination of scratch hardness of soda-lime glass(Matsunami 0050),
a normal load and a scratch speed were fixed at 0.98 N and 0.05 mm/s, respectively.
Water and some n-alcohols(1-pentanol, 1-hexanol, 1-heptanol, 1-octanol, 1-nonanol,
1-decanol, and 1-undecanol) were employed as environment. These liquids except for
Load cell FN: normal load

(normal load) Moving
stage Scratch direction
Indenter
Sample
A: vertical projection
Moving stage
w
Load cell of the contact area
Knoop indenter (horizontal load)
Figure 1. Schematic of a scratch tester. Figure 2. Schematic of a scratch groove.
water were dehydrated with a molecular sieve for at least one week. All the liquids were
dropped on the sample surface just before the scratch test. Scratch hardness in each
environment was determined from at least 10 runs. For a comparison, Knoop hardness
was also measured at the indentation load of 0.49, 0.98, 2.0, and 2.9N in the same liquids.
In the case of the measurements of crack initiation load for silica(Asahi AQ)
glass, Pyrex-type glass(Tempax) and soda-lime glass, a normal load was changing at a
rate of 0.03N/s. A scratch speed was fixed at 0.07mm/s. The crack initiation was
evaluated in water and n-heptane, which was also dehydrated. The width of scratch
groove and the existence of crack at the edge of groove were observed by using an
optical microscope after the scratch test. In some cases, the crack geometry was
observed by using a scanning electron microscope(SEM; Hitachi S-3200N). The
crack initiation load was determined from the position of an initial crack at the
edge of groove, the scratch speed, and the rate at which the normal load increased. At
least 30 specimens were used for the determination of crack initiation load.
3. RESULTS
Figure 3 shows scratch hardness of soda-lime glass in water and in n-alcohols.

Scratch hardness increases with an increase in the number of carbon atoms in each
alcohol molecule, but shows a maximum at 1-decanol. Scratch hardness obtained in
1-decanol is 28% larger than that in 1-pentanol. Scratch hardness obtained in water is
comparable with that in 1-pentanol.
As a comparison, Knoop hardness in each environment was measured. The
result is shown in Fig. 4. A load dependence of Knoop hardness can be observed for
all the environments. This load dependence of hardness was also reported by Hirao
and Tomozawa,7 and will be discussed later. Knoop hardness at a given indentation
load increases by 8% with an increase in the length of alcohol chain. In addition, it is
found that Knoop hardness(4.35.0GPa) is over two times smaller than scratch
hardness (1015GPa).

The crack initiation test was performed for silica, Tempax, and soda-lime glasses in
water and in dehydrated heptane by using the scratch tester. Figure 5 shows an
Scratch hardness /GPa

15
10
water 5 6 7 8 9 10 11

Figure 3. Scratch hardness of soda lime glass in water and n alcohols. Error bars indicate 1 standard
deviations.
0.49N
0.98N
Knoop hardness /GPa
2.0N
5 2.9N
water 5 6 7 8 9 10 11
Figure 4. Knoop hardness of soda lime glass in water and n alcohols. The indentation loads are 0.49,
0.98, 2.0, and 2.9 N.
example of a relationship between a scratch position and a normal load for soda-lime
glass in hep-tane. Four points labeled A, B, C, and D correspond to the scratch areas
noted by the same letters in Fig. 6. The crack initiation can be observed in Fig. 6(B). As
for silica glass, a position-load curve obtained in heptane and photos of a scratch
groove are shown in Fig. 7 and Fig. 8, respectively. The crack initiation load for silica
glass is much smaller than that for soda-lime glass. The lower crack initiation load for
silica glass is also observed in water. The crack initiation loads for each glass can be
fitted well by a Weibull function. Weibull plots for the data obtained in heptane are
shown in Fig. 9. The averaged values of crack initiation loads and Weibull moduli for
all the glasses in heptane and in water are listed in Table 1. It is considered that Weibull
modulus describes the width of distribution in crack initiation load, and that it is much
affected by surface condition. In order to discuss the compositional variation of
3
C
Normal load /N
B
2 D
A
1 Crack initiation
0
0 1000 2000 3000 4000 5000
Position / μm
Figure 5. Relationship between normal load and position for soda lime glass in heptane.
30μm
Figure 6. Photos of a scratch groove after the crack initiation test for soda lime glass in heptane. Scratch
C
Normal load /N
B
1
A
Crack initiation
0
0 1000 2000
Position / μm
Figure 7. Relationship between normal load and position for silica glass in heptane.
Weibull modulus, some information on the relationship between the detailed finishing
grade of glass surface and Weibull modulus will be needed. It is noteworthy that
Tempax has the highest crack initiation load in water, but not in heptane.
Figure 8. Photos of a scratch groove after the crack initiation test for silica glass in heptane. Scratch
4. DISCUSSION

The load-dependence of Knoop hardness in Fig. 4 is probably caused by a change in
surface energy of the glass in each environment. It has been reported that this load
dependence of hardness is attributed to the contribution of surface energy to the
indentation deformation.7 Following the works by Bernhardt8 and Frohlich et al.,9
Hirao and Tomozawa7 separated the indentation deformation into two energy
contributions, which were the surface energy contribution and the volume energy
contribution. They suggested that Knoop indentation load, L, can be expressed by
the following equation,
F = 1 exp [ (P/ξ)m]
Crack initiation probability, F /%
99
90
80
60
40 silica Tempax soda lime
m=6.5 m=6.3 m=11
20
10
0.5 1.0 2.0 3.0

Crack initiation load, P/N
Figure 9. Weibull plots of the crack initiation loads obtained in heptane.
Table 1. Averaged crack initiation loads and Weibull moduli for the glasses.
Glass Crack initiation load /N Weibull modulus, ma
Silica in heptane 0.63 6.5

in water 0.35 3.8
Tempax in heptane 1.39 6.3
in water 0.99 7.0
Soda lime in heptane 1.63 11
in water 0.56 5.6
a
F ¼ 1- exp [-(P=j)m ]
L ¼ a1 l þ a2 l 2 , (3)
Ð
where a1 and a2 are fitting parameters, and l is Knoop diagonal length. Because Ldl
may be related to the total energy of indentation deformation, a1 (l 2 =2) is related to
the energy to create new surface, and a2 (l 3 =3) to the energy for volume deformation.
Hence, a2 corresponds to the load-independent hardness, and a1 the parameter
related to the sur-face energy. On the other hand, from the values of L and l,
Knoop hardness, HK, is defined as the following equation,10
HK ¼ 14:23L=l 2 : (4)
From Eqs. (3) and (4), if the surface energy term can be negligible, Knoop hardness is
independent of the indentation load. Figure 10 shows some examples of the relation-
ship between L / l and l. From Eq. (3), the following expression can be obtained, and
a1 corresponds to the y-intercept, and a2 to the slope of the line in Fig. 10.
L=l ¼ a1 þ a2 l (5)
Figure 11 shows the fitting parameters, a1 and a2 , in all the environments. A more
distinct increase in a1 can be observed in this figure as compared with that in a2 .
40
1-decanol
water
30
20
L/l /103 Nm−1
10
0
0 20 40 60 80 100
l/ 10−6m
Figure 10. Relationship between L / l and l for soda lime glass in water and in 1 decanol.
4 400
3 300
a1 /kNm−1
a2 /MPa
2 200
1 100
0 0
water 5 6 7 8 9 10 11
Number of carbon atoms in a molecule
Figure 11. Fitting parameters, a1 and a2 in Eq. (5) obtained in water and in n alcohols.
It implies that the variation in chain-length of n-alcohol affects the surface state
between glass and liquid rather than hardness as a bulk property. The surface energy
is probably related to the polarity in n-alcohols because the surface charge of soda-
lime glass is in general negative. The larger the polarity in n-alcohol is, the less the
surface energy of glass in the liquid. Reported dielectric constants of n-alcohols
employed are listed in Table 2.
On the other hand, fracture toughness, KIc , is defined as the following equation,
KIc ¼ (2GE)1=2 , (6)
for the plane-stress condition. In this equation, G is the fracture surface energy, and
E is Young’s modulus. Although the surface energy is not coincident with the fracture
surface energy, the former covers a large part of the latter for ideal brittle materials.
Therefore, the surface energy must affect the fracture-related properties, such as
scratch hardness. It can be said that the scratch hardness is closely linked to the
surface-creating process, such as micro-cracking, chipping and so on. These energy
release processes surely reduce the energy for groove formation. This is the reason
Table 2. Dielectric constants of water and n alcohols.
Solvent Dielectric constant
Water 78.54(258C)a
1 pentanol 15.13(258C)b
1 hexanol 13.03(208C)b
1 heptanol 11.75(208C)b
1 octanol 10.30(208C)b
1 nonanol
1 decanol 8.1(208C)a
1 undecanol
a
Ref. [11]
b
Ref. [12]
why scratch hardness is much larger than Knoop hardness. The larger variation of
scratch hardness originates from a larger contribution of the surface-creating pro-
cess.
In Fig. 4, a longest-chain alcohol(1-undecanol) causes a decrease in scratch
hardness. Although the origin has been unclear, Westwood and Huntington5 pro-
posed one reasonable explanation that the hardness of glass can be markedly influ-
enced by the z-potential, which is likely to be similar to the surface charge of glass.
When z is negative, which is true in water and in the lower n-alcohols,13 this situation
must bring positive charge carriers, e.g. Naþ ions, into the near-surface region for
charge-counter balancing. This locally increased Naþ ions reduce hardness in this
region. When z is positive, on the other hand, the migration of OH ions is respon-
sible for softening of glass. Glass should be the hardest at its zero point of surface
charge. There is no evidence that the sign of z-potential is reversed by changing the
chain-length of n-alcohol. However, this surface compositional change would be one
possible origin of decrease in scratch hardness.

The cracks observed around the groove were mainly surface chips or radial(chevron)
cracks. There was no median crack beneath the groove. Figure 12 shows the cross
section of the scratch groove in soda-lime glass. For the other glasses, no median
crack can be also observed in the cross sections.
Water environment promotes crack initiation in silica, Tempax, and soda-lime
glasses (Table 1). This is because the crack initiation process is assisted by the
adsorption of water molecule. In the conceptual framework of this idea,14 a water
molecule is adsorbed at the strained bonds formed by a scratch process. After that,
the attacked Si-O-Si bond and the water molecule break one Si-O bond to leave new
two Si-OH bonds. West and Hench15 reported that the energy barrier of hydrolysis of
strained 3-fold rings is 97% smaller than that of fracture by water-free dilation. This
water-assisted bond-breakage process is the origin of lower crack initiation load
obtained in water.
The crack initiation load of silica glass is much lower than those of Tempax and
soda-lime glasses (Table 1). This lower crack initiation load cannot be explained in
terms of conventional mechanical properties, e.g. Young’s modulus and fracture
toughness. Table 3 shows some mechanical properties of silica, Tempax, and soda-
lime glasses. Here, Young’s modulus and Vickers hardness are the catalogue values,
5μm
Figure 12. SEM micrograph of a scratch groove of soda lime glass in air. Normal load 2.94N.
Table 3. Some mechanical properties of the glasses.
Young’s Vickers Fracture

Crack initiation Brittleness modulusa / hardnessa / toughnessb /
Glass load (in heptane)/N / mm1 GPa GPa MPam1=2
Silica 0.63 10.0 74.2 7.9 0.79

Tempax 1.39 8.5 63.0 5.5 0.65c
Soda lime 1.63 7.1 71.5 5.4 0.76
a
Catalogue value.
b
Ref. [16]
c
Ref. [17]
and fracture toughness the reported values. The lower crack initiation load is related
not to the weaker bond-strength(Young’s modulus), but to the resistance to defor-
mation(hardness). One part of the work done by an external force is used to deform
plastically. The rest part may be converted into a residual stress around the groove or
into a crack. This means that the decreased ability to deform leads to easy crack
initiation. Even if the averaged bond-strength of silica glass is comparable with that
of soda-lime glass, less deformability in silica glass causes the crack initiation load to
be lowered.
It is considered that the crack initiation load is a measure of brittleness (HV /
KIc ) of glass. In Table 3, brittleness of each glass is calculated from HV and KIc . The
higher brittleness of glass causes the lower crack initiation load. This relationship was
also reported by Le Houérou et al.3 It has been still unclear that Tempax shows the
maximum crack initiation load in water, but it is suggested that B2 O3 as a network
former plays an important role for deformation. Further experimental investigations
will be needed.
Sehgal and Ito18 have shown that the brittleness of glass can be determined from
the Vickers indentation load, the Vickers diagonal length, and the median/radial
crack length. However, their technique cannot be applied to the complicated crack
patterns by indentation, such as a cone crack in silica glass. It is concluded that the
crack initiation test by using a scratch tester is the simple and useful technique for
evaluating brittleness of various glasses.
5. CONCLUSIONS
The effect of environment on scratch hardness has been investigated for soda-lime
glass. Scratch hardness was much influenced by the chain-length of n-alcohol. The
variation of scratch hardness with the chain-length of n-alcohol was more conspicu-
ous than that of Knoop hardness. This result comes from the variation of surface
energy of glass in the liquids. From the measurement of load-dependent Knoop
hardness in various n-alcohols, it was found that the chain-length in n-alcohol
affected the parameter related to the surface energy rather than to the intrinsic
hardness. Because the surface energy is directly related to the fracture behavior,
scratch hardness including fracture process surely depends on environment.
The crack initiation at the groove-edge can be clearly observed for silica,
Tempax, and soda-lime glasses by using a scratch testing machine. The crack
initiation load obtained in water was much lower than that in heptane because of the
water-assisted bond-rupture. The crack initiation load for silica glass was lower than
those for Tempax and soda-lime glasses. This is attributed to less deformability in
silica glass. It is elucidated that the crack initiation load through the dynamic scratch
test is a measure of brittleness of glass.
REFERENCES
1. K. Li, Y. Shapiro, and J.C.M. Li, Scratch test of soda lime glass, Acta Mater. 46(15),
5569 5578(1998).
2. S. Yoshida, H. Tanaka, T. Hayashi, J. Matsuoka, and N. Soga, Scratch resistance of sodium
borosilicate glass, J. Ceram. Soc. Japan 109(6), 511 515(2001).
3. V. Le Houérou, J. C. Sangleboeuf, S. Dériano, T. Rouxel, and G. Duisit, Surface damage of soda
lime glasses: indentation scratch behavior, J. Non Cryst. Solids 316, 54 63(2003).
4. S.M. Wiederhorn, S.W. Freiman, E.R. Fuller Jr., and C.J. Simmons, Effect of water and other
dielectrics on crack growth, J. Mat. Sci. 17, 3460 3478(1982).
5. A.R.C. Westwood and R.D. Huntington, Adsorption sensitive flow and fracture behavior of soda
lime glass, Proc. Intl. Conf. Mechanical Behavior of Materials IV, Society of Materials Science, Kyoto,
Japan, 383 393(1972).
6. S. Jacobsson, M. Olsson, P. Hedenqvist, and O. Vingsbo, Scratch testing, ASM Handbook 18,
430 437(1992).
7. K. Hirao and M. Tomozawa, Microhardness of SiO2 glass in various environments, J. Am. Ceram.
Soc. 70(7), 497 502(1987).
8. E.O. Bernhardt, On microhardness of solids at the limit of Kick’s similarity law(in German),
Z. Metallkd. 33(3), 135 144(1941).
9. F. Frohlich, P. Grau, and W. Grellmann, Performance and analysis of recording microhardness tests,
Phys. Status Soldi 42, 79 89(1977).
10. Japanese Industrial Standards, Z 2251(1998).
11. Kagaku Binran, Kiso Hen II (in Japanese), p.1166, Ed. by the Chemical Society of Japan (Maruzen,
Tokyo, 1975).
12. D.R. Lide, Handbook of organic Solvents, p.254, 265, 370, 380 (CRC Press, Boca Raton, Florida,
1995).
13. M. DeSmet and J. Delfosse, Colloid Science, p.230, Ed. by H.R. Kruyt (Elsevier, New York, 1952).
14. T.A. Michalske and S.W. Freiman, A molecular interpretation of stress corrosion in silica, Nature
295, 511 512(1982).
15. J.K. West and L.L. Hench, Silica fracture Part II A ring opening model via hydrolysis, J. Mat. Sci.
29, 5808 5816(1994).
16. S.M. Wiederhorn, Fracture surface energy of glass, J. Am. Ceram. Soc. 52(2), 99 105(1969).
17. J. Sehgal and S. Ito, Brittleness of glass, J. Non Cryst. Solids 253, 126 132(1999).
18. J. Sehgal, Y. Nakao, H. Takahashi, and S. Ito, Brittleness of glasses by indentation, J. Mat. Sci. Lett.
14, 167 169(1995).
SHEAR DRIVEN DAMAGE
AND INTERNAL FRICTION IN
INDENTATION LOADING OF
A GLASS-CERAMIC
Anthony C. Fischer-Cripps*
1. INTRODUCTION
Indentation damage is a particularly important limitation in the use of brittle

materials in structural applications. Some nominally brittle materials have been
shown to exhibit plastic type yield in indentation testing.1-10 These materials are
characterized by a relatively coarse grain structure and weak grain boundaries.
A particularly convenient material for study is a mica-containing glass-ceramic. As
shown in Fig. 1, the microstructure consists of circular mica platelets in a glassy
matrix. The size of the platelets can be modified by simple heat treatment during
processing. The average platelet diameter and the aspect ratio increase with increas-
ing crystallization temperature. During indentation with a spherical indenter, a sub-
surface region of accumulated damage is observed to form in preference to the
characteristic Hertzian cone crack normally associated with this type of loading.
An indentation stress-strain response shows a shape similar to that obtained on
ductile materials.
Previous work indicates that11,12 the indentation response of mica glass-ceramic
materials is a function of the macroscopic uniaxial compressive yield stress, Y and a
strain-hardening characteristic. A range of behavior, from quasi-ductile to nominally
brittle, could be attained by simply varying the size and shape of the grains through
*
Anthony C. Fischer Cripps, CSIRO Division of Telecommunications and Industrial Physics, Bradfield
Rd, West Lindfield, NSW 2070 Australia.
113
Figure 1. Electron micrograph of microstructure of mica glass ceramic. Specimen was polished to 1 mm
finish and etched in HF for 10 s. Most of the platelets are viewed edge on. The occasional plan view of
a platelet can be seen.
heat treatment. Details of the microstructural properties for the glass-ceramic ma-
terials studied here are given in Table 1.
The dependence of the yield stress was subsequently found to be consistent with
a Mode II fracture mechanism within the microstructure that depended upon the
absolute value of the grain size (i.e. platelet diameter). The apparent strain hardening
arises from an elastic constraint on the deformation of grains imposed by the glassy
matrix and depends on the platelet aspect ratio.13 Further study showed that the
magnitude of the yield stress was consistent with a mechanism of friction on a
microstructural scale quite different to that of the familiar Amonton’s law.14 The
present work brings the previously reported work together into a cohesive account of
the nature of internal friction on the microstructural scale and its relevance to the
macroscopic behaviour of these types of materials. An understanding of these issues
assists in the design, at a microstructural level, of these nominally brittle materials to
suit the requirements of a particular application.
Table 1. Microstructural parameters for mica glass ceramic materials. For all materials, the
volume fraction is Vf 70%, the elastic modulus E 70 GPa and Poisson’s ratio n 0:26.
Platelet dimensions measured from sectioned and etched samples. Yield stress was determined
according to a procedure given in reference 12.
Platelet
4 hr heat treatment Maximum platelet thickness Yield stress Y
Micro structure temp. (8C) diameter (mm) (mm) (MPa)
Fine 1000 1.2 0.40 2160

Medium 1060 3.7 0.80 1230
Coarse 1120 10 1.2 750
2. MACROSCOPIC BEHAVIOUR
For the glass-ceramic materials considered here, the plastic zone resulting from
indentation with a spherical indenter is contained within an area bounded by the
circle of contact and can be best described in terms of an elastic constraint.11 Fig. 2
shows the results from a bonded-interface specimen for a glass-ceramic specimen
loaded with a spherical indenter. The top part of Fig. 2 shows the residual impression
in the surface and the bottom part is a section view showing the sub-surface accu-
mulated damage. A quantitative measure of plasticity is given by an indentation
stress-strain response as shown in Fig. 3 where it can be seen that the amount of
deviation of the expected linearly elastic behaviour depends upon the microstructure
of the material, the greater the grain size, the more pronounced the deviation from
linear elasticity. The size and shape of the damage zone and the shape of the
indentation stress-strain response is similar to that observed in ductile materials
under this type of loading.
Elastic-plastic behavior in these nominally brittle ceramics occurs as a result of
the relatively large hydrostatic component of stress in the indentation stress field and
is routinely observed by researchers in the rock-mechanics field. Sub-surface damage
occurs as a result of shear faulting along planes of weakness within the material
beneath the indenter. In the present case, cleavage planes within the mica appear to
be sources of weakness that lead to shear-driven damage. Finite element work12 has
demonstrated that there is a close connection between the macroscopic yield stress of
these materials and the absolute platelet size. A mismatch in elastic properties within
the microstructure leads to stress singularities resulting in yield stress which has an
inverse square root dependence on the grain size in the manner of a Mode II crack.
Evidently, the macroscopic response of the material depends on shear-driven failure
events on a microstructural scale, which in turn, are intimately connected with the
nature of friction at this scale of contact.
Figure 2. Optical micrograph of indented bonded interface mica containing glass ceramic specimen
showing half surface (top) and section (bottom) views. Indentations performed with WC sphere of radius
R 3:18 mm at indenter load P 2000 N.
Figure 3. Indentation stress strain relationships for glass ceramic. Dashed line shows the expected
Hertzian elastic response. Experimental results are shown for the coarse (10 mm grain size), medium
(5 mm) and fine (2 mm) microstructures. The larger the grains, the more deviation from the elastic response.
3. INTERNAL FRICTION AND YIELD
Amonton’s law has been shown to be a result of the random roughness of real
surfaces and arises as a result of the increasing area of contact between asperities
on both surfaces as they are bought into contact as the normal force is increased. It us
usually expressed in terms of the friction coefficient m such that:
F ¼ mN (1)
where N is the applied normal load and F is the sideways force required to initiate
tangential sliding between the two bodies. Contact thus only takes place at asperities
on each surface and friction forces arise from elastic deformation, shearing, welding,
ploughing, or plastic deformation of the asperities as they move past one another. If
we divide the terms in Eq. 1 by the real area of contact A, we find that the frictional
shear stress t is in direct proportion to the normal stress sN :
F mN
t¼ ¼ ¼ msN (2)
A A
Jaeger and Cook15 show that slip along a plane of weakness within a material
under compressive loading with confined pressure may occur when the Coulomb
criterion is satisfied:
jtj ¼ So þ msN (3)

Figure 4. Schematic of a sliding shear fault under hydrostatic stress.
In Eq. 3, t is the applied shear stress, So is the inherent shear strength of the
material. and sN is the normal stress on the plane under consideration as shown in
Fig. 4. Eq. 3 is equivalent to Eq. 2 at So ¼ 0.
In terms of the principal stresses s1 and s3 , where s3 is the confining pressure
and s1 is the applied normal stress, the Coulomb criterion becomes:
h 1=2 i h 1=2 i
s1 m 2 þ 1 m s3 m2 þ 1 þm ¼ 2So (4)
The macroscopic Tresca failure criterion (t ¼ Y=2) is thus a special case of the
Coulomb criterion with m ¼ 0 and So ¼ Y=2 where the angle at which sliding occurs
is 458.
Finite element analysis shows that12 for the glass-ceramic materials considered
here, differences in the microstructure can be accommodated by a change in the
macroscopic yield stress Y for each material in conjunction with the Tresca criterions
implying that the internal friction coefficient m plays a negligible role in these
materials. Further support for this contention is that the strong, but short range
Van der Waals forces at the grain boundary maintain the boundary intact under
shear loading and that failure occurs within the atomically smooth cleavage planes of
the mica platelets.16,17 As shown in Fig. 5, atomic force microscope examination of
the damaged regions clearly shows fissures within the platelets.
It is thus claimed that any resistance to internal sliding depends upon the
inherent shear strength of the mica platelets. This in turn depends upon the grain
size, which then impacts on the macroscopic yield strength of the material.
4. DISCUSSION
The task now is to substantiate the conclusions made above given the following
conclusions of previously reported work:
Figure 5. Atomic force microscope image of a cross section of a mica platelet in the shear driven damage
zone. Arrows show fissures in the platelet that are not visible in undamaged material.
1. Experiments on a macroscopic scale demonstrate the presence of internal yield

within the microstructure, the degree of which depends upon the physical char-
acteristics of the microstructure.
2. Finite element modelling on a macroscopic scale using the Tresca criterion
matches the observed size and shape of the plastic zone and also the shape of
the indentation stress-strain curves indicating that the physical characteristics of
the microstructure determine its yield strength. The use of the Tresca criterion
implies a zero coefficient of friction on the microstructural scale. The range of
macroscopic yield stresses for the materials studied here is 750 MPa to 2000 MPa.
3. Finite element modelling on a microstructural scale indicates that internal yield
has the appearance of a Mode II crack phenomenon (the shear strength of the
microstructure depends on the inverse of the square root of the platelet diameter
and is independent of the magnitude of the normal stress on the grain). There is a
correlation between the macroscopic yield stress Y used in (2) above with the
inverse square root of the grain size.
4. Atomic force microscope observations show that shear faulting occurs within the
atomically smooth planes of the mica platelets and that friction is likely to occur
under conditions in which the real and apparent area of contact coincide.
Can the observed behaviour on the macroscopic and microstructural scales be
reconciled with what we know about frictional sliding under these conditions? To
answer this question, we turn to other reported work with mica in which the real and
apparent areas of contact coincide. Johnson18 reports that experiments with an
atomic friction microscope (AFM), on mica in which the contact dimension is 2 to
10 nm, indicate a frictional shear stress of 1 GPa. Other measurements performed
with a surface force apparatus (SFA), in which the contact dimension is in the order
of 20-80 mm, indicate a frictional shear stress to be approximately constant at

20 MPa. The discrepancy is explained by the mechanism of plastic deformation at
the interface. According to Johnson, the physical mechanism of sliding friction is due
to the nucleation and propagation of dislocation-like defects through the interface
under the influence of the applied shear force. For small contacts (< 20 nm), disloca-
tion-like defects cannot be nucleated and atoms move relative to one another with a
shear strength equal to the theoretical strength of the material. For large contacts, the
rate of nucleation of dislocation-like continuities at the edge of the contact becomes
equal to the rate at which they disappear at the centre and the frictional stress reaches
a steady-state value. In the intermediate range, the shear stress required to nucleate
dislocation-like defects at the leading edge of the contact depends upon the square
root of the dimension of the contact.
The relevance to the present work is that the scale of ‘‘contacts’’ in the mica
platelets studied here is in the order of 1 to 10 mm. This is the intermediate zone
referred to by Johnson in which the frictional stress depends upon the inverse square
root of the grain size in the manner of a Mode II crack. It has been previously shown
that estimations of yield stress for a range of platelet sizes based on an inverse square
root law yielded predictions of macroscopic behavior that agreed well with experi-
mentally observed behavior and that the macroscopic yield stress was determined to
be in the order of 1 GPa. A yield stress in the order of 1 GPa gives a value of So in the
order of 500 MPa, precisely within the intermediate zone described above. The
nature of the observed Mode II behaviour on the microstructural scale is thus
explained.
What is the nature of these defects that show such a scale effect? Johnson
describes them as ‘‘dislocation-like’’ and shows that work involving conventional
dislocation mechanics is consistent with observations of the size effects for frictional
shear stress given above. The answer lies in molecular dynamics simulations of which
are limited by the number of atoms that can be accommodated and as such, usually
fall into the small contact regime. Simulations on a larger scale would evidently
elucidate the actual mechanism of deformation on the molecular scale. However,
from a material science perspective, the scale effects of frictional traction under
controlled conditions matches well with those existing at a macroscopic level and
thus offers a useful base from which macroscopic material properties may be
designed from microstructural variables in some detail. For example, for the mater-
ials studied here, if the platelets were made very large (> 50 mm), one would expect
the size dependence on yield to diminish and that, all other things being equal, the
shape of the indentation stress-strain curves would perhaps begin to be independent
on the grain size.
6. SUMMARY
The present work examines the role of internal friction in mica-containing glass-
ceramic materials subjected to indentation loading. It is proposed that shear-driven
damage beneath the specimen surface occurs via internal sliding along cleavage
planes within the mica platelets near the mica-glass boundaries. The sliding surfaces
in this case are considered to be atomically smooth such that the real and apparent
areas of contact coincide. The frictional shear stress is thus independent of the
normal forces arising from thermal mismatch stresses and only depends on the work
of adhesion of the interface and the scale of the contacts. The scale of contacts for
these materials lies within an intermediate zone in which the frictional shear stress
arises from the stress required to nucleate dislocation-like discontinuities within the
material. This leads to a size effect similar to that experienced by a crack in Mode II
loading and is in accordance with previous work in which a connection between such
a size effect and the macroscopic response of the material was identified.
REFERENCES
1. B. R. Lawn, N. P. Padture, H. Cai, and F. Guiberteau, Making ceramics ‘ductile’, Science, 263,
1114 16 (1994).
2. F. Guiberteau, N. P. Padture, H. Cai, and B. R. Lawn, Indentation fatigue. A simple cyclic Hertzian
test for measuring damage accumulation in polycrystalline ceramics, Phil. Mag. 68(5), 1003 16
(1993).
3. N. P. Padture, B. R. Lawn, Fatigue in ceramics with interconnecting weak interfaces: a study using
cyclic Hertzian contacts, Acta Metall. 43, 1609 17 (1995).
4. N. P. Padture, In situ toughened silicon carbide, J. Am. Ceram. Soc. 77(2), 519 23 (1994).
5. N. P. Padture and B. R. Lawn, Toughness properties of a silicon carbide with an in situ induced
heterogeneous grain structure, J. Am. Ceram. Soc. 77(10), 2518 22 (1994).
6. H. H. K. Xu, L. Wei, N. P. Padture, B. R. Lawn and R. L. Yeckley, Effect of microstructural
coarsening on Hertzian contact damage in silicon nitride, J. Mat. Sci. 30, 869 78 (1995).
7. F. Guiberteau, N. P. Padture, and B. R. Lawn, Effect of grain size on Hertzian contact damage in
alumina, J. Am. Ceram. Soc. 77(7), 1825 31 (1994).
8. H. Cai, M. A. Stevens Kalceff, B. R. Lawn, Deformation and fracture of mica containing glass
ceramics in Hertzian contacts, J. Mater. Res. 9(3), 762 70 (1994).
9. H. Cai, M. A. Stevens Kalceff, B. M. Hooks, B. R. Lawn, and K. Chyung, Cyclic fatigue of a mica
containing glass ceramic at Hertzian contacts, J. Mater. Res. 9(10), 2654 61 (1994).
10. D. G. Grossman, Machinable glass ceramics based on tetrasilicic mica, J. Am. Ceram. Soc. 55,
446 49 (1972).
11. A. C. Fischer Cripps, Elastic plastic behaviour in materials loaded with a spherical indenter, J. Mat.
Sci. 32, 727 36 (1997).
12. A. C. Fischer Cripps, A partition problem approach to microstructural modelling of a glass ceramic,
J. Am. Ceram. Soc. 82(7), 1835 43 (1999).
13. A. C. Fischer Cripps and B. R. Lawn, Indentation stress strain curves for ‘‘quasi ductile’’ ceramics,
Acta Mater. 44(2), 519 527 (1996).
14. A. C. Fischer Cripps, Role of internal friction in indentation damage in a mica containing glass
ceramic, J. Am. Ceram. Soc. 84(11), 2603 2606 (2001).
15. J. C. Jaeger and N. G. W. Cook, Fundamentals of Rock Mechanics, (Chapman and Hall, 1971).
16. Private communication, D. T. Smith, National Institute of Standards & Technology, 1995.
17. R. G. Horn and D. T. Smith, Contact Electrification and Adhesion Between Dissimilar Materials,
Science, 256, 362 64 (1992).
18. K. L. Johnson, The contribution of micro/nano tribology to the interpretation of dry friction, Proc.
Instn. Mech. Engrs, 214 Part C, 11 22 (2000).
SLOW CRACK PROPAGATION
IN CERAMICS AT THE
NANO- AND MICRO-SCALE:
EFFECT OF THE
MICROSTRUCTURE
J. Chevalier1 and G. Fantozzi1
A review of Slow Crack Growth (SCG) results obtained for different oxide and non-
oxide ceramics at ambient temperature, under different environments, is presented.
They are analyzed on the basis of their crack velocity (V) versus stress intensity factor
(KI ) diagrams. The aim of this paper is to consider mechanisms acting at the crack tip
(i.e. at the nano-scale) and microstructural mechanisms occurring in the crack wake
or at the crack front (i.e. at the micro-scale) to rationalize the approach of SCG.
Fine grained, alumina, zirconia, silicon nitride and silicon carbide ceramics are
first compared as model materials with low amount of toughening. Results are all
consistent with a stress corrosion mechanism acting at the crack tip. They show that
the higher the ionic to covalent ratio, the higher the susceptibility to stress corrosion
by water. Some materials doped with amounts of silica rich phases are analyzed.
They exhibit different behavior depending on the location of a glassy phase at grain
boundaries or triple points. An interesting aspect on a practical point of view is the
presence of a threshold stress intensity factor below which no propagation occurs.
The effect of microstructure on V-KI laws can be investigated by modifying the
grain size of non transformable ceramics (e.g. alumina) and the grain size or the
amount of stabilizing agent in the case of transformable ceramics (e.g. zirconia or
zirconia toughened alumina). The effect of bridging or transformation toughening on
V-KI curves is discussed. If the reinforcement at the microscopic scale is correctly
1
G.E.M.P.P.M. (U.M.R. C.N.R.S. 55 10) I.N.S.A. de Lyon, 69621 Villeurbanne Cedex France
173
174 J. Chevalier et al.
taken into account, intrinsic V-KItip curves, only dependent of the chemical bounds
broken at the crack tip, are obtained.
INTRODUCTION
Delayed fracture of brittle materials was first reported by Grenet [1] in glass. Later,
Orowan [2] reported such a sub-critical crack growth by observing a strength reduc-
tion of a glass in air compared to that measured under vacuum which he attributed to
a reduction of fracture energy. Following this work, different systematic studies have
been conducted on glass, alumina single crystals and mica [3 6]. As a general trend,
they have shown that an unique relation could be established, for a given environ-
ment, between the velocity (V) and the stress intensity factor (KI ) with three different
stages attributed to distinct mechanisms. In the first stage (Region I, for low veloci-
ties), crack propagation is limited by the thermally activated reaction kinetic between
the ceramic and corrosive molecules. These molecules were clearly identified in the
case of silica as H2 O [7]. Region II, at intermediate velocities, depends on environ-
ment, but is much less sensitive to KI . This region is often attributed to gaseous
diffusion along the crack interface. Region III, for high velocities represents the
fracture of ceramic bonds without the need of environmental species. Although
mentioned by Wiederhorn [8] in glass, the occurrence of a threshold has not always
been clearly identified. Recently, Wan et al.[9,10] considered this point and the
presence of a threshold was definitely shown in mica. These authors and Maugis
[11] considered that the threshold corresponds to an equilibrium with a null crack
velocity. In this respect, crack healing should occur below the threshold, which was
partly shown in the case of mica [10]. The threshold, in this sense should correspond
to the Griffith criterion for crack propagation. However, the presence of a threshold
is still in debates in the ceramic society.
To the authors knowledge there is no such detailed study for polycrystalline
ceramics, partly because of the scattering caused by the microstructure, which leads
to a more distributed fracture. However, a similar general trend with 3 propagation
stages has been observed in alumina, mullite, silicon nitride, zirconia etc. [12]. Thus,
slow crack growth in polycrystalline ceramics is also attributed to stress corrosion,
despite a lack of experimental evidence. The presence of a threshold is even less
clarified. In polycrystalline ceramics, R-curve, which traduces an increase of crack
resistance with crack extension, can also play a significant role on the sub-critical
crack propagation and complicates the slow crack growth analysis [13 15]. The
influence of an R-curve behavior has been investigated by different authors. This
leads to a progressive shift of the crack propagation to lower velocities as crack
resistance increases. Depending on the crack length measurement with respect to the
rising part of the R-curve, either a complete shift or a higher slope can be obtained
[13,14]. Fett et al. [15] showed that a crack velocity decrease could be observed at the
beginning of the propagation of small cracks in alumina. Different microstuctural
mechanisms can play a role on crack velocities, the most important in ceramics being
crack bridging and transformation toughening.
This is the aim of this paper to consider mechanisms acting at the crack tip (i.e.
at the nano-scale) and microstructural mechanisms occurring in the crack wake or at
the crack front (i.e. at the micro-scale) to rationalize the approach of Slow Crack
Slow Crack Propagation in Ceramics at the Nano- and Micro-Scale 175
Growth (SCG). The knowledge of the respective influence of nano- and micro-scale
mechanisms during SCG allows to define strategies for better SCG resistant
materials.
1. SCG AT THE NANO-SCALE LEVEL : STRESS CORROSION

AT THE CRACK TIP
2.1. Stress Corrosion Equations
Stress corrosion by water at the crack tip is the likely mechanism responsible for
crack propagation in ceramics. In this respect, it is interesting to develop briefly the
model proposed by Lawn [16] for single crystals or glasses. This model is based on a
thermally activated reaction of water and ceramic to form complex. For low crack
velocity (stage I), V is given by :

E0I aI (G 2g)

VI ¼ 2 n0 a0 exp sinh (1)
kT kT
where no is the fundamental reaction frequency (no ¼ kT=h, with h Planck constant
and k Boltzman constant), EoI the activation energy of the reaction, G* the mech-
anical energy due to the stresses at the crack tip, ao is a characteristic bond separ-
ation, aI is an activation area and ge is the fracture surface energy in the presence of
water molecules. It is to note that G* can be easily related to the stress intensity factor
at the crack tip, KItip , by : G ¼ K Itip 2 =E, where E is the Young’s modulus.
In gaseous environment, crack velocity can be limited by a transport of corrosive
molecules to the crack tip and is given by:
64 G a30 pE
VII ¼ p (2)
l (2 p m kT )
3 p E a0 ln
a0
where E is the Young modulus for zirconia ceramic (220 GPa), pE the water
pressure, m the mass of water molecules and l is the crack opening displacement.
Gaseous diffusion and reaction occur in series so the overall crack velocity is given by:
1 1 1
¼ þ (3)
V VI VII
For higher crack velocities, fracture occurs under vacuum conditions and the crack
speed is given by :

E0III aIII (G 2gs)

VIII ¼ 2 n 0 a 0 exp sinh (4)
kT kT
with EoIII the activation energy for the fracture of zirconia bonds, aIII an activation
area and gs is the fracture surface energy in the absence of water molecules (under
vacuum).
Equations (1) and (4) are consistent with the basic Griffith theory or its gener-
alization with the Dupré work of adhesion. A crack propagates when:
G ¼ R ¼ 2gs under vacuum (5a)

G ¼ Re ¼ 2ge in a given environment (5b)
with g e < gs . Thus the equilibrium of a given crack may be changed either by
variation of the loading or by the environmental conditions.
The initial Griffith criteria corresponds therefore to this point of equilibrium.
When we move above or below this equilibrium point, the crack velocity increases
(backward or forward). This is the theoretical meaning of a threshold in polycristal-
line ceramics, below which no propagation occurs (ideally below which a crack
heals).
Figure (1) represents examples of V-KI curves in a polycristalline zirconia
ceramic, which exhibit the typical pattern of SCG due to stress corrosion at the
crack tip, including the presence of thresholds under vacuum, water and air condi-
tions. The threshold, KI0 , often observed on macro-cracks was questioned because of
the differences observed between the behavior of micro- and macro-cracks in poly-
crystalline ceramics [15]. The results obtained in figure (2) for the same polycrystal-
line zirconia show that the threshold obtained for macro- (i.e. by Double Torsion)
and microcracks (indentations) are in agreement. It is of course possible that
reinforcement mechanisms, discussed below, can change significantly the value of
the threshold if they lead to a R-Curve behavior.
2.2. Influence of the Ionic/Covalent Ratio on Stress Corrosion
With the aim of clarifying the processes involved in stages I and II an idealised
representation of a proposed reaction between water and a strained ceramic bond at
the crack tip has been proposed [7,16]. Reaction steps involve: (1) adsorption of
water to ceramic bond, (2) reaction involving simultaneous proton and electron
transfer, and (3) formation of surface hydroxyls. The susceptibility of a given ceramic
to the mechanism of stress assisted corrosion by water, hence to SCG, lies in its
ability to react with water and to dissociate. Glass was studied in the past as a model
ionic material with a high susceptibility to SCG, because stressed Si-O-Si bounds
Crack rate (m/s)

10−2
water II
10−4 air
Silicon oil
vacuum
10−6 III
I
10−8
10−10
10−12
3 4 5 6
KI(MPa√m)
Figure 1. V KI diagram of a polycrystalline zirconia ceramic (from [17]).

10−2
V (m/s)
10−4
Double torsion
10−6
10−8 Indentations
10−10
10−12
2,5 3 3,5 4 4,5 5
KI (MPa√m)
Figure 2. Threshold stress intensity factor determined by two different methods on polycrystalline
zirconia (from ref. [17]).
Table 1. Ionic fraction and susceptibility to SCG for different ceramics: in

general, the higher the ionic fraction, the higher the intrinsic susceptibility to SCG.
Zr02 SiO2 Si3 N4 SiC
Ionic fraction 0,6 0.51 0,30 0,11

KI0air =KI0vacuum 0,6 0,6 0,9 1
could be easily broken by the combined effect of stresses and adsorption of water.
This is also the case of zirconia, very prone to SCG [17]. Figure (3) represents
schematically the susceptibility of different ceramics to SCG in a V-KI =KI0vacuum
diagram. Very fine microstructures were chosen to avoid any R-Curve effect which
would perturb the analysis. Table (1) represents the ionic fraction for different
ceramics compared to the ratio of threshold obtained in air to that obtained under
v(m/s) mullite
10−2
Al2O3
10−3
ZrO2 / SiO2
10−4
10−5
SiC
10−6
Si3N4
10−7
100 1 1.5
KI/KI0vacuum
Figure 3. Susceptibility to SCG of different fine grained ceramics. SiO2 shown for comparison.
vacuum conditions. From these results, it can be argued that the intrinsic suscepti-
bility of a given ceramic to SCG is related to its atomic structure.
2.3. Effect of a Glassy Phase on SCG in Ceramics

The presence of a glassy phase can have a profound effect on the properties of
polycrystalline ceramics. Such glassy phases in ceramics sometimes result from silica
rich impurities but they can also be added during processing route in order to obtain
special physical or mechanical properties. For instance, the addition of glassy phases
in yttria doped zirconia ceramics is a way to increase their superplastic behavior
[18,19] or their resistance to ageing [20,21]. The common feeling is that the presence
of glassy phases in ceramics leads to a decrease of their crack resistance, because silica
is brittle by nature and sensitive to SCG. In other words, it would mean that glassy
phase leads to a shift of the V-KI diagram towards low KI values.
In a recent study, it was shown that silica addition in zirconia materials did not
lead necessary to glassy phase at grain boundaries [21,22]. Special processing condi-
tions are required to obtain glassy phase at triple points or grain boundaries. An
addition of small amounts of silica and alumina leads more easily to glassy phase at
grain boundaries. Figure (4) represents conventional and High Resolution Transmis-
sion Electron Microscopy images of three model polycrystalline zirconia materials
processed without any silica addition, with 0.5 wt. % silica (the same pattern would
be obtained with 2.5 wt.% addition) and 0.5 wt. % silica þ alumina addition. The
structure at triple points and grain boundaries is different at a nano-scale level.
Undoped zirconia do not show any evidence of vitreous phase neither at grain
boundaries nor at triples junctions, confirming the high purity of the starting powder
and the absence of any contamination during the process. Silica doped zirconia
shows an interesting microstructure with always clean grain boundaries (among
more than 30 edge-on grain boundaries studied), but with the presence of a silica
rich amorphous phase at triple joints (details can be found in [21,22]). This is in
contrast with a generally accepted idea of a continuous vitreous phase in silica doped
ceramics. Grains are rounder in comparison to the undoped material, due to the
presence of vitreous phase at triple joints. Silica þ alumina doped zirconia exhibits
another type of microstructure with the presence of a silica and alumina rich phase
both at triple joints and, at least partially, at grain boundaries. This means that the
surface energies between the amorphous phase and the crystalline grains can be
modified by the composition of the silica film.
Is their a detrimental effect of silica addition to zirconia on SCG? The answer is
not necessarily. Figure (5) shows that the presence of a glassy phase located at triple
points does not induce any important change in SCG. A slight acceleration of the
crack rates in stage I is observed for the material doped with 2.5 wt. % silica, since
intergranular cracks have to propagate through these glassy pockets. Nevertheless,
KI0 and KIC are identical in all batches. KI0 is related to the surface energy of grain
boundaries in the presence of water. Since there is no glassy phase at grain bound-
aries even with an addition of silica, the surface energy can be considered as constant
in average whatever the silica content. The same argument can be used for KIC ,
which is related to the fracture energy under vacuum conditions.
A grain boundary glassy phase has a small effect, particularly on the threshold
stress intensity factor KI0 and on stage I (fig.6). In this case, it can be argued that the
Figure 4. Microstructures of polycrytalline zirconia ceramics : (a) undoped, (b) doped with 0.5 wt.% silica
and (c) doped with 0.5 wt.% silica þ alumina.
10−2
V (m/s)
10−4
10−6
Un-doped
Doped SiO2, 0.5%
10−8
Doped SiO2, 2.5%
10−10
KI (MPa.m1/2)
10−12
2.5 3 3.5 4 4.5 5 5.5 6
Figure 5. Effect of silica addition on V KI curves of zirconia

V (m/s)
10−4
10−6
10−8
Doped SiO2/Al2O3
10−10 Undoped
10−12
2.5 3.5 4.5 5.5 6.5
KI (MPa√m)
Figure 6. Effect of silicaþalumina addition on V KI curves of zirconia
behavior of ceramic corresponds to that of the silica rich phase, since crack propa-
gation is still intergranular. However, since the sensitivity of silica to stress corrosion
is similar to that of zirconia, the effect is far from detrimental. The conclusions could
be different if a glassy phase was present in a ceramic intrinsically less sensitive to
stress corrosion.
2. SCG at the Micro-Scale Level : Effect of the Microstructure
The preceding section has shown that SCG was intrinsically related to the bounds
that are subjected to the combined effect of stresses and water. However, the
microstructure plays a major role on the V-KI curves obtained experimentally. For
polycrystalline zirconia for example, the thresholds obtained in air and vacuum give
a crack resistance of 30 J=m2 and of 50 J=m2 respectively (they can be calculated from
figure (1) ). In contrast to the case of glass and single crystals^, these values are far
above the surface energy that can be measured by the heat of vaporisation in
ceramics (on the order of 1 J=m2 ). This means that the crack resistance R (or Re ) is
not only related to the surface energy but also to additional energy dissipation. The
two main mechanisms contributing to energy dissipation in polycrystalline ceramics
are phase transformation toughening and crack bridging. The equations (5) must be
transformed as :
R ¼ 2gs þ Rf under vacuum 6(a)

Re ¼ 2g e þ Rf in a given environment 6(b)
where Rf is the contribution of microstructural related reinforcement to the crack

resistance. This contribution in the case of Y-TZP is about ten times that of the
^
Even in the case of single crystal or glass there is a slight discrepancy between the surface energy
measured by vaporisation and the crack resistance, owing to the dissipation unavoidably associated to
mechanical fracture.
surface energy. The same order of magnitude is observed for most polycristalline
ceramics. This means that only a portion of the applied stress intensity factor is used
to propagate the crack, the other portion being dissipated by reinforcement mech-
anisms. In terms of stress intensity factors, the stress intensity factor at the crack tip,
which rules crack growth, is lower to the applied one, because of reinforcement. This
leads to the well known equation:
KItip ¼ KI DKr (7)
where KI and KItip represent the applied stress intensity factor (by the external loads)
and the stress intensity factor at the crack tip respectively and DKr the reinforcement
due to the microstructure.
The following sections aim at showing the effect of the microstructure on DKR
and consequently on V-KI curves.
3.1. Effect of the Grain Size on the Slow Crack Growth in Alumina
Slow crack growth in polycrystalline alumina ceramics was first reported by Freiman
et al. [23]. The propagation rates are strongly influenced by the grain size, the large
grain materials offering more resistance to crack growth.
It is well known that crack propagation properties are related to bridging effect
and this effect is more pronounced in coarse-grained materials [24 27]. Coarse-grain
alumina shows an R-curve behavior which is characterized by an increase in crack
resistance with crack extension, due to crack surface interaction in the crack wake
[28 29]. This R-curve behavior is due to the stresses required to overcome the
traction of interlocking grains and pullout of unbroken ligaments in the wake of
the crack.
The influence of an R-curve behavior on slow crack growth in ceramics corres-
ponds to a shift toward higher stress intensity factor values with crack extension
[13,14]. Fett and Munz [15] have shown that a crack velocity decrease could even be
observed at the beginning of the crack propagation of small macrocracks in alumina.
Crack propagation tests were conducted by Ebrahimi et al. [30] on polycrystalline
alumina with average grain sizes d between 2 and 13 mm and equiaxed grain morph-
ology. The tests were performed on a high-purity alumina (> 99,9 wt % purity)
doped with 500 ppm MgO (Baikowski, SM8). High density and uniform equiaxed
structures were available after sintering because the addition of MgO. The slow crack
growth behavior was studied by using the double torsion method which allows to
obtain the V-KI curves. In order to accurately consider the influence of crack
resistance curves on slow crack growth, R-curves were also measured with the double
torsion technique on pre-cracked specimens.
The results of crack propagation in air for different grain sizes obtained by the
relaxation method under constant displacement are shown in Fig.7. The cracking
rate da/dt is very sensitive to the stress intensity factor K below 2 10 5 m=s for the
three alumina ceramics (region I). For da/dt faster than these values, K dependence
becomes moderated (region II). However, the slope of region I increases with the
grain size and region II of crack propagation is much more limited for higher grain
size. Above 10 4 m=s, a clear increase of crack rate is observed for fine-grained
alumina (region III). It is less visible in the case of coarse-grained alumina.
10−2
10−3
Crack Velocity(m/s) 10−4
10−5
13 μm
5 μm
10−6
2 μm
10−7
3 3.5 4 4.5 5
KI (MPam)
Figure 7. V KI curves of three alumina with different grain sizes, obtained by the Double Torsion method
(from [30]).
The KIO , defined as the minimum stress intensity factor that leads to measurable
crack propagation during relaxation, is much higher ( 25%) in coarse-grained
alumina (d ¼ 13 mm) than in fine-grained alumina (d ¼ 2 mm). Thus, the slow
crack growth resistance becomes higher for the higher grain size. This result can be
attributed to crack-bridging R-curve effects which increase with crack extension for
coarse-grained alumina.
The R-curves for the alumina, measured with the double torsion technique on
pre-cracked specimens, as a function of crack length increment are illustrated in Fig .8
which depicts the DKr (increase of toughness after pre-cracking) versus Da (crack
extension after pre-cracking) plot. The crack resistance increases with crack growth
and this behavior becomes more remarkable when the grain size increases. The
results show that increasing the grain size increases both the ‘‘steady-state’’ toughness
and the toughening rate (initial slope of the R-curve). This increasing toughness is
commonly attributed to ‘‘wake’’ effects which shield the crack tip from the externally
applied stress.
Crack shielding in non-transforming ceramics occurs mainly by bridging,
branching and deviation and by friction of the crack interfaces. The typical pattern
1.2
ΔKr (MPa√m)
0.8
0.4 13 μm
5 μm
2 μm
0
0 5 10 15
Crack length increment (mm)
Figure 8. R Curve behavior of the three alumina ceramics (from [30]).

of an intergranular crack propagation was observed,

p even for the largest grain size
(13 mm). Increases in toughness up to 1 MPa m are observed between the crack
initiation toughness and the final plateau toughness of coarse-grained alumina
(d ¼ 13 mm).
The dependence of toughness upon crack extension may be written as:
KI appl: (a) ¼ KItip þ DKr (a) (7b)
where KI appl is the total stress intensity factor calculated from the external load by
the DT method, KItip the stress intensity factor acting at the crack tip, DKr the
toughness contribution from the wake (R-curve) and a the crack length.
According to this explanation and by consideration of the results in Figs.7 and 8,
the stress intensity factor acting at the crack tip at various crack lengths can be
calculated. Fig.9 shows the V KItip for the three aluminas. The KItip in the three
aluminas is similar and independent of grain size. This definitively shows that the
crack growth mechanism, stress corrosion, is the same at the crack tip for the three
ceramics. In particular, a steep dependence of V with KItip is observed for each
material (region I), followed by a saturation in crack velocity on a so-called region II
of slow cracking. The statement that the V KItip is independent of the grain size
underlines the fact that atomic bonds broken during crack growth are intrinsically
the same.
Other results were obtained with a higher purity alumina (> 99.99 wt %, AKP
50 Sumitomo), without addition of MgO (Fig.10) [31]. The mean grain size ranges for
AKP aluminas were 6---53 mm for the same range of sintering temperatures, because
of the absence of MgO . The grain size distribution in AKP aluminas was broader,
with the appearance of elongated grains for highest grain sizes (10% and 50% of
abnormal grains for the mean grain sizes 35 and 53 mm respectively).
For grain sizes ranging from 6 to 35 mm, the conclusions were the same than
before : the larger the grain size, the larger the crack resistance. However, for the
largest grain size (53 mm), a marked shift of the V KI curve towards low stress
intensity factor was observed (Fig.10). In this case, transgranular rupture is prevalent
as pointed out by Fig.11 which shows the crack propagation paths in AKP aluminas.
For the fine-grained AKP alumina (6 and 9 mm) the crack propagation is mostly
10−2
10−3
Crack Velocity (m/s)
10−4
10−5 13 μm
5 μm
10−6 2 μm
10−7
3 3.5 4 4.5 5
KItip (MPa√m)
Figure 9. Intrinsic V KItip curves for the three alumina ceramics.

10−1
6 μm
Crack velocity (m/s) 9 μm
10−3 35 μm
53 μm
10−5
10−7
2 3 4 5
KI (MPa√m)
Figure 10. V KI curves of four alumina with different grain sizes, up to 53 mm.
10 μm 20 μm
(a) (b)
Figure 11. Mode of fracture for an AKP alumina, for 6 mm grain size (a) and 53 mm grain size (b).
intergranular whereas in the alumina with a 53 mm mean grain size, the crack path is
mostly transgranular. Thus the shift of the V KI curve in the coarse-grained
alumina is due to the appearance of transgranular fracture, which leads to a decrease
of the R-curve effect.
The experimental results clearly show that the stress corrosion is the key mech-
anism for crack propagation in alumina ceramics and that a unique crack velocity-
crack tip stress intensity factor law can be obtained by taking into account the crack
resistance curve in coarse-grain ceramics.
3.2. Effect of Transformation Toughening on V--- KI Curves.
The case of alumina is relatively simple since at least as first approximation, the
reinforcement depends only on the crack length. Other cases are more complex, and
the reinforcement can also depend on the applied stress. This is the case of the
mechanisms acting at the crack front, such as microcracking and phase transform-
ation toughening. A good example of the last case is the reinforcement in zirconia
based ceramics. It is based on the tetragonal to monoclinic transformation of
particles at the crack tip due to the local high stresses. The associated volume
expansion leads to stress shielding. It can be shown that transformation zone size is
related to the stresses applied to the ceramic (the higher the stresses, the larger the
transformation zone). It can be demonstrated that shielding stress intensity factor in
toughened ceramics is given by [32]:
Kr ¼ Csh KI (8)
p !
0:214 E Vf e (1 þ n)
T
3
with Csh ¼ (9)
(1 n) scm 12p
where E is the Young modulus, Vf the volume fraction of transformed particles, eT is

the volume dilatation associated to the transformation, n the Poisson ratio, smc is
the local stress for phase transformation and KI the applied stress intensity factor. Eqn
(8), which suggests that the shielding stress intensity factor is proportional to the applied
one, has been verified experimentally by Chevalier et al.[17]. Reporting Eqn (8) in Eqn
(7), the stress intensity factor at the crack tip, KItip , can be expressed as:
KItip ¼ (1 Csh )KI (5)
with Csh increasing with the material transformability. If we assume that the intrinsic
mechanism of SCG in a given ceramic is given by:
V ¼ A0 KItip n (6)
Then, the analysis means that the experimentally determined V-KI law in the pres-
ence of transformable zirconia particles is:
V ¼ Ao (1 Csh Þn K I n ¼ A K I n (7)
These results mean that the effect of transformation toughening on SCG is to shift the
V-KI law towards high stress intensity factors, preserving the n exponent of the SCG
power law. This can be applied to zirconia ceramics as well as zirconia based compos-
ites. Csh can be modified by changing the stability of the tetragonal phase. In zirconia,
it can be easily modified by changing the grain size. Figure (12) shows V-KI laws of
yttria stabilized zirconia ceramics with different grain size, in a log-log plot. The results
are compared to a zirconia single crystal The higher the grain size, the higher the
transformation toughening, thus the larger the crack resistance. At the same time, the
curves run parallel, as a signature of the ‘intrinsic’ zirconia n exponent.
In this paper, we have analyzed to major mechanisms acting separately. In a
certain number of ceramics, the two mechanisms can act at the same time. It is for
example the case of Ceria stabilized zirconia ceramics, where the reinforcement is
function of both the crack length and the applied stress intensity factor. In this
particular case, the analysis of the results is more difficult.
3.3. An Application : Slow Crack Growth in a Zirconia-Toughened Alumina

Nanocomposite.
Zirconia-toughened alumina (ZTA) ceramics with high transformation toughening

ability can be obtained if a high portion of tetragonal phase with the possibility to
transform under applied stress is retained at room temperature. This requires that the
Single crystal
10−2 0,3 μm
0,5 μm
1 μm
10−4
Crack rate (m/s)
10−6
10−8
10−10
10−12
1 2 3 4 5 6 7
KI (MPa√m)
Figure 12. Effect of Grain size on V KI laws in yttria stabilized zirconia, comparison with a single cristal.
zirconia particles must have a size distribution ranging between the critical size for
spontaneous transformation on cooling to room temperature after sintering, Dc , and
0
the critical size for stress induced transformation, Dc [33,34]. Dc size is in the range
0
of 0:5---0:8 mm for unstabilized zirconia in an alumina ceramic matrix and Dc is in the
order of 0:1 mm.
Colloidal processing may allow to obtain a narrow size distribution of zirconia
particles homogeneously dispersed in the alumina matrix [35].
Indeed, as shown by Fig.13, the composites obtained by the colloidal processing
route show nano-sized ZrO2 particles (D50 ¼ 0:4 0:1 mm) homogeneously distrib-
uted at Al2 O3 grain boundaries (D50 ¼ 1:2 0:4 mm) without agglomerates and a
minimum number of microstructure flaws. The composites obtained by conventional
powder-mixing processing route show large ZrO2 particles less homogeneously dis-
tributed in the alumina matrix.
Figure 13. Scanning Electron Microscopy image (Back scattered electrons mode) of a zirconia toughened
alumina nano composite, showing the narrow distribution of well dispersed zirconia particles in an
alumina matrix [39].
The alumina zirconia composites processed by the colloidal route present a

0
narrow distribution that is roughly between Dc and Dc . This leads to a more efficient
transformation toughening in the ceramics processed by the colloidal route and to a
shift of the crack velocity versus stress intensity factor curve towards higher KI values
for the composites obtained by the colloidal route compared to the composites
obtained by a powder-mixing technique [35].
The measured V KI diagrams of alumina and the composites prepared with
the colloidal route and different volume fraction of unstabilized zirconia are plotted
in Fig.14: Al2 O3 , A7.5Z, A10Z and A15Z correspond to nanocomposites with 0, 7.5,
10 and 15 vol % ZrO2 respectively and A7.5Z-2Y is a composite with 7.5 vol % ZrO2 ,
stabilized by 2 mol % Y2 O3 . Subcritical crack growth laws were measured using the
double torsion technique and the relaxation method.
The curves exhibit a trend to shift towards higher KI values up to a maximum
before dropping with increasing volume fraction above 10%. These results are in
agreement with previously reported values on the toughness of ZTA composites [36].
The increasing crack resistance of the composites can be explained by the stress
induced martensitic transformation of the tetragonal zirconia particles (t-ZrO2 ) to
the monoclinic form (m-ZrO2 ). As shown by Claussen [36], the amount of stress in
the matrix increases with the content of zirconia and, beyond a maximum, the matrix
is not able to retain the tetragonal zirconia as a metastable phase. Thus, the KIC value
decreases and the overall V-KI curve is shifted towards lower KI values for high
zirconia content (A15Z).
For the composite processed with 2 mol % of Y2 O3 (A7.5Z-2Y), the V-KI curve
is slightly shifted to lower KI values compared with the ZTA A7.5Z without yttria.
This lower crack resistance corresponds to a lower effective transformation toughen-
ing, as really observed, yttria stabilizing the t-ZrO2 .
Fig.14 shows that the ZTA processed by colloidal route always exhibit higher
crack resistance than pure alumina, even for A15Z and A7.5Z-2Y where little or non
transformation occurs. Transformation toughening is thus not the only mechanism
acting to reinforce alumina in ZTA and crack deflection and microcracking also
3 4 5 6
10−2
Al2O3
10−3 A7.5Z
A15Z
V (m/s)
1x10−4
1x10−5
A10Z
10−6
A7.5Z-2Y
3 4 5 6
KI (MPa m1/2)
Figure 14. SCG behavior of zirconia toughened alumina nano composites, with different compositions
(from 35).
occur. Finally, the influence of residual stresses in these nano-composites can be

significant. In any case, the material exhibiting the highest amount of phase trans-
formation during crack extension (A10Z) presents the highest resistance. The load
relaxation method does not allow measurement at very low velocity (#10 7 m=s) but
present the advantage of being fast and a full curve can be recorded in a single
experiment. Measurement of crack velocity under constant load presents the advan-
tage of allowing the measurement of very low velocities. Such measurements were
made on alumina, ZTA processed by colloidal route (A15Z) 3Y-TZP (3mol % yttria
stabilized zirconia) [37]. The whole crack velocity diagrams are compared in fig.15.
The overall curves obtained in air exhibit three distinct stages that can be fitted by a
power law:
V ¼ AKI n
The values of parameter n for the first stage I are reported in table 2. Thresholds
below which no crack propagation occurs (KIo ) were observed at values indicated in
table 2. Fig 13 shows that the 3Y-TZP exhibits a higher toughness than alumina but
the thresholds of both materials are lower than of the ZTA nanocomposite, meaning
that the necessary stress to initiate crack growth is lower in alumina and 3Y-TZP.
Additionally, the ZTA nanocomposite has a toughness close to that of zirconia but a
higher threshold, a consequence of the steeper slope of its V-KI diagram.
Slow crack growth in oxyde ceramics is attributed to stress assisted corrosion at
the crack tip. This is indeed the combined effect of high stresses at the crack tip and
the presence of water, temperature that induce crack propagation in a subcritical
manner. Alumina has lower susceptibility to water and thus to stress assisted corro-
sion. Therefore, the V-KI curve of alumina presents a higher slope than the curve
corresponding to zirconia (see Table 2). From an atomistic point of view, this means
that the fracture energy of zirconia is lower in the presence of water, because the
zirconia bonds are prone to chemisorption of the polar water molecules. The add-
ition of a small amount of zirconia has two main advantages. First ZTA exhibit crack
propagation primarily through the alumina matrix and have a low susceptibility to
KIC KICKIC
1x10−2
III
II III
II
1x10−4 Al2O3
1x10−6 II
V (m/s)
I
I
1x10−8 3Y-TZP I
1x10−10 Al2O3- 10 vol % ZrO2
KI0 KI0 KI0

10−12
2 3 4 5 6 7
KI (MPa m1/2)
Figure 15. SCG in ZTA composites, compared to alumina and zirconia monoliths (for details, see ref.
[35,39]).
Table 2. Threshold (KIo ) and Toughness (KIc ) of Al2 O3 , A10Z and 3Y TZP ceramics.
Material Thresholds (KIo ) MPa:m1=2 Toughness (KIc ) MPa:m1=2 n 1st stage
Al2 O3 2.5 0.2 4.2 0.2 56.8

A10Z 4.0 0.2 5.9 0.2 53.5
3Y TZP 3.5 0.2 6.1 0.2 31.5
stress assisted corrosion from water. Second, these composites are reinforced by the
presence of transformable zirconia particles, shifting the V-KI diagram towards
higher KI values while preserving the slope of the curve (table 2).
In any case, the thresholds KIo represents the key parameter as far as durability
is concerned [38] (as for orthopedic implants, witch must survive in vivo for more
than 10 years).
The KIo of the best ZTA composite (A10Z) is twice that of pure alumina. This
means that this nanocomposite is expected to work at loads two times greater than
monolithic alumina without any delayed fracture. So, these ZTA composites may
offer to improve the lifetime of ceramic and to improve the safety range of use of
point prostheses or orthopedic implants [39].
4. CONCLUSION
SCG in polycrystalline ceramics can be analyzed at two major scales. At a nano-scale

level, the susceptibility of a given ceramic is ruled by the intrinsic response of the
atomic bounds to stress assisted corrosion. Ceramics with a high ionic fraction are
particularly sensitive to stress corrosion. This is the case of zirconia, silica and
hydroxyapatite for example. Non oxide ceramics appear to be almost insensitive to
this phenomenon at ambient temperature.
At a microscopic level, the SCG behavior of a given ceramic is dependent on the
amount of reinforcement. Thus, for a same type of ceramic, i.e. for the same V-KItip
intrinsic response, V-KI curves can be influenced by the microstructure. Two major
toughening mechanisms have been discussed: crack bridging and phase transform-
ation toughening. Other mechanisms are postulated to occur: micro-cracking and
residual stresses for example. They might also influence V-KI curves.
REFERENCES
1. L. Grenet, Bull. Soc. Encour. Ind. Nat., 4, 838 48 (1889).

2. E. Orowan, Nature, 154, 341 49 (1944).
3. R.J. Charles, W.B. Hilling, pp.511 27 in symposium sur la résistance mécanique du verre et les moyens
de l’améliorer, Union Scientifique Continentales du Verre, 24, Charleroi, Belgium, 1962.
4. S.M. Wiederhorn, J. Am. Ceram. Soc., 50, 407 14 (1967).
5. A.I. Bailey, S.M. Kay, Proc. Roy. Soc. Lond., Public. N8A301, 47 (1967).
6. S.M. Wiederhorn, in NBS special publication, mechanical and thermal properties of ceramics, Vol. 303,
217 41 (1969).
7. T.A. Michalske , S.E. Freiman, J. Am. Ceram. Soc., 66 [4] 284 88 (1983).
8. S.M. Wiederhorn, P.R. Townsed, J. Am. Ceram. Soc., 53, 486 89 (1970).
9. K.T. Wan, S. Lathabais, B.R. Lawn, J. Europ. Ceram. Soc, 6 [4] 259 68 (1990).
10. K.T. Wan, N. Aimard, S. Lathabais, J. Mater. Research, 5 [1] 172 82 (1990).
11. D. Maugis, J. Mater. Sci., 20, 3041 73 (1985).
12. F. Sudreau, PhD. Thesis, INSA de Lyon, France, 272 pages (1992).
13. F. Mignard, C. Olagnon, G. Fantozzi, pp. 565 76 in Fracture Mechanics of Ceramics, Vol. 11. 1996.
14. A. Osaka, N. Hirosaki, M. Yoshimura, J. Am. Ceram. Soc., 73 [7] 2095 96 (1990).
15. T. Fett, D. Munz, J.Am.Ceram.Soc., 75 [4] 958 63 (1992).
16. B.R. Lawn, Fracture of brittle solids, Second edition, Cambridge University Press, 378 pages, 1993.
17. Chevalier, J., Olagnon, C., Fantozzi, G, J. Am. Ceram. Soc., 82, 11, p.3129 3138 (1999).
18. K. Kajihara, Y. Yishizawa, T. Sakuma, Acta metall. mater. vol. 43 p. 1235 2 (1995).
19. K. Hiraga, H.Y. Yasuda, Y. Sakka, Mater. Sci. Eng. A vol. 234 236 p. 1026 9 (1997).
20. M.L. Mecartney, J. Am. Ceram. Soc. Vol. 70 p. 54 58 (1987).
21. L. Gremillard, T. Epicier, J. Chevalier et al., Acta Mater. vol. 48 p. 4647 52 (2000).
22. L. Gremillard, J. Chevalier, T. Epicier, G. Fantozzi, J. Am. Ceram. Soc. vol. 85 n82 p.401 07 (2002).
23. S. W. Freiman, K.R . McKinney and H.L. Smith, in Fracture Machanics of Ceramics, Vol.2, pp.659
676, 1973.
24. R. Knehans and R.W. Steinbrech, in Science of Ceramics, edited by P.Vincenzini (Research Institute
for Ceramics Technology, Faenza, Italy, 1984), Vol, pp.613 619.
25. P. Chantikul, S. Bennison and B.R. Lawn, J.Am.Ceram. Soc. 73, 2419 (1990).
26. G. Vekinis, M.F. Ashby and P.W.R. Beaumont, Acta Metall.Mater. 38, 1151 (1990).
27. G. Fantozzi, J. Chevalier and M. Saadaoui, in Fracture Mechanics of Ceramics, Vol.13, pp.213 228,
2002.
28. P. Swanson, C. Fairbanks, B.R. Lawn, Y. Mai and B. Hockey, J.Am.Ceram. Soc. 70, 279 (1987).
29. G. Pezzotti, O. Sbaizero, V. Sergo, N. Muraki, K. Maruyama and T. Nishida, J.Am. Ceram. Soc. 81,
187 (1998).
30. M.E. Ebrahimi, J; Chevalier and G. Fantozzi, J.Mater.Res. 15, 142 (2000).
31. M.E. Ebrahimi, PhD. Thesis, INSA de Lyon, France, 220 pages (2000).
32. R.M. McMeeking, A.G. Evans, J. Am. Ceram. Soc., 65 [6] 242 46 (1982).
33. N. Claussen and M. Ruhle, in Advances in Ceramics, Vol.3, Science and Techonology of Zirconia I,
1981, pp. 137 163.
34. R.C. Garvie, in Advances in Ceramics, Vol.24A, Science and Technology of Zirconia III, 1998,
pp. 55 69.
35. A.H.De Aza, J. Chevalier, G. Fantozzi, M. Schell and R. Torrecillas, J.Am.Ceram.Soc. 86, 115
(2003).
36. N. Claussen, J.Am.Ceram.Soc. 59, 49 (1976).
37. J. Chevalier, C. Olagnon and G. Fantozzi, Composites, Part A, 30,525 (1999).
38. G. Willmann, Adv.Eng.Mater. 2, 114 (2000).
39. J. Chevalier, A.H. De Aza, G. Fantozzi, M. Schell and R. Torrecillas, Adv. Mater. 12, 1619 (2000).
STANDARD REFERNCE
MATERIAL 2100: FRACTURE
TOUGHNESS OF CERAMICS
George D. Quinn,a Kang Xu,b Robert Gettings,c Jonathan

A. Salem,d and Jeffrey J. Swabe
ABSTRACT:
Standard Reference Material (SRM) 2100 is the first reference material in the world
for the property fracture toughness for any class material. The SRM is for fracture
toughness of ceramics and may be used with any ceramic fracture toughness test
method, but is best suited for methods that use beams in bending. The SRM
complements ASTM, ISO, CEN and national fracture toughness standards. This
paper describes SRM 2100 and its creation.
1. INTRODUCTION
Standard Reference Material (SRM) 2100 is intended for verification of ceramic

fracture toughness testing procedures. It may be used with any fracture toughness
test procedure. It may also be used in conjunction with standard test methods from
American Society for Testing and Materials (ASTM), the International Organization
a
Ceramics Division, Stop 852, National Institute of Standards and Technology, Gaithersburg, MD 20899.
b
Praxair Technology Center, 175 East Park Dr. , Tonawonda, NY 14150.
c
Formerly with the Standard Reference Materials Program Office, National Institute of Standards and
Technology, Gaithersburg, MD 20899.
d
Life Prediction Branch, Structures Division, MS 49 7, NASA Glenn Research Center, Cleveland, OH
44135.
e
Weapons and Materials Research Directorate, U.S. Army Research Laboratory, Aberdeen Proving
Ground, MD, 21005.
499
500 George D. Quinn et al.
Figure 1. SRM 2100 comprises groups of five hot pressed silicon nitride test pieces.
for Standards (ISO), the European Committee on Standardization (CEN) or Japan-

ese Industrial Standards Association (JIS). The SRM is a set of five unprecracked
ceramic test specimens cut from a single billet (plate) of hot-pressed silicon nitride
(HPSN), Norton grade NC 132f,g as shown in Figure 1. A considerable experience
base exists for this particular grade of silicon nitride and it was featured in an
international fracture toughness round robin described below.
ASTM C 1421 for advanced ceramics1 includes three related test method pro-
cedures as illustrated in Figure 2. The three methods are the surface crack in flexure
(SCF), precracked beam (PB) [also known as single-edge precracked beam (SEPB)],
and chevron notch in bending (CNB) methods. The content and development of C
1421 and its precursor, ASTM PS 070-97h are documented elsewhere [2,3,4,5]. The
SRM may also be used with three ISO standard test methods that are under
development in Technical Committee TC 206, based on the SEPB,6 SCF,7 and
CNB8 methods. It may also be used with a draft CEN standard9 that features
SCF, CNB, SEPB, and single edge-V-notched beam methods, as well as JIS R 1607
which has the SEPB method.10
f
St. Gobain Norton Co., Worcester, MA. NC 132 is no longer available from the manufacturer.
g
Standards and Technology nor does it imply that these materials or equipment are necessarily the best
for the purpose.
h
ASTM PS 070 97 was a Provisional Standard adopted in April 1997 by Committee C 28. It was
converted to the full consensus standard C 1421 in April 1999.
Standard Reference Material 2100 501
Figure 2. The three test methods in ASTM Standard C 1421. JIS R 1607 has only the SEPB, whereas ISO
and CEN will include all three methods.
2. BACKGROUND
A substantial amount of work on fracture toughness testing methodologies and

materials characterization was performed in the 1970’s, much of it motivated by
potential ceramic heat engine applications. Nevertheless, there was considerable
confusion and conflicting fracture toughness data in the 1980’s. A 1981 report
associated with a project to prepare a handbook of ceramic property data stated:11
‘‘Although a limited amount of fracture toughness information is present in the ceramic
literature, it was decided . . . not to include such data. The reason for this is the nebulous and
sometimes controversial nature of this property and the test methods used to obtain it.’’
A 1980 National Materials Advisory Board report even had discussion of whether
KIc was an intrinsic, bulk property of a ceramic.12 One 1984 mechanical properties
review paper concluded:13
‘‘Fracture toughness (KIc ) is a critical property of interest to a designer since it is a direct
measure of the intrinsic resistance to crack propagation or fracture. There is a respectable
amount of published data on fracture toughness, but much of it is contradictory and unreliable.
. . . ..{Results from} large cracked fracture mechanics specimens often do not relate well to
component failure from small natural defects . . . .. Unfortunately, no analysis or technique is
emerging as a leading candidate and there appears to be little chance a standard method will be
available in the near future.’’
The advent of the Vickers indentation crack length techniques in the mid to late
1970’s was a major distraction. It was not until 1991 that a concerted effort to refine
and standardize technically rigorous fracture toughness test methods was resumed in
ASTM Committee C-28, Advanced Ceramics. Work on the reference material began
at National Institute of Standards and Technology (NIST) in May 1997 shortly after
the ASTM Provisional Standard PS 070-97 was approved in April 1997.
We first contemplated creating a fracture toughness reference material in late
1994 after the completion of a very successful Versailles Advanced Materials and
Standards (VAMAS) international round robin. VAMAS is a prestandardization
collaboration program between the G-8 countries and the European Community.
Technical Working Area 3, Structural Ceramics, has conducted thirteen major round
robins over the course of 13 years, including five round robins and 4,500 experiments
on fracture toughness alone.14 The 1993 to 1994 round robin featured the surface
crack in flexure (SCF) method. Twenty laboratories around the world obtained very
consistent results on test specimens from a single billet (designated
p ‘‘E’’) of the p
NC
132 HPSN.15,16,17 A grand mean fracture toughness of 4:59 MPa m 0:37 MPa m
was obtained from 107 experiments from all laboratories as shown in Figure 3.
Eliminating
p results from one laboratory improved p the data to a mean of
4:56 MPa m with a standard deviation of 0:32 MPa m for 102 valid experiments.
The test method precision was evaluated using ASTM E 691-92, Practice for Coord-
inating an Interlaboratory Study to Determine the Precision of a Test Method. p The
within-laboratory precision (repeatability standard deviation) was 0:24 MPa m, or a
coefficient of variation of only 5.4 %. Thepbetween-laboratory precision (reproduci-
bility standard deviation) was 0:31 MPa m, or a coefficient of variation of only
6.8 %. Some laboratories tried other methods such as the chevron notch in bending
test on spare test specimens furnished as part of the round robin and they obtained
results that agreed with the SCF outcomes.
The VAMAS round robin results matched other NC 132 silicon nitride data
from many laboratories over the course of 25 years. Table 1 shows the tabulation of
thirty-eight values of fracture toughness for this material using methods as diverse as
double cantilever beam, double torsion, SCF, SEPB, CNB, short bar chevron notch,
and fractographic evaluations using NC 132’s natural flaws. Data by the Vickers
indentation crack length and single-edge notched beam are not included since such
data is unreliable.
p The p SCF, CNB and SEPB methods usually yielded results from
4:3 MPa m to 4:8 MPa m. NC 132’s uniform creep and static fatigue resistance had
been noted previously.19,42
20
19
NC 132
18
VAMAS 1994
17
SCF Round Robin
16
n =107
15
14
13
12
frequency
11
10
9
8
7
6
5
4
3
2
1
0
2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0
Fracture toughness MPa√m
Figure 3. Histogram of all 107 results obtained for the HPSN in the VAMAS SCF round robin.
p
Table 1. Fracture toughness (MPa m) for NC 132 silicon nitride.
Mean KIc S. dev. Method n Source (Ref.)
4.59 0.37 SCF 107 VAMAS Round Robin (15,16)
4.65 0.10 SCF Petrovic et al.** (18)
4.64 0.25 SCF 4 Quinn and Quinn** (19)
4.48 0.07 SCF 4 "
4.33 0.37 SCF 3 "
4.39 0.19 SCF, lab ambient 3 Swab and Quinn (20)
4.32 0.29 SCF, dry nitrogen 5 "
4.64 0.4 SCF 5 Tikare and Choi (21)
5.25*** SCF 4 Gonczy and Johnson (22)
4.67 0.3 SEPB 7 Tikare and Choi (21)
4.54 0.12 SEPB 5 Bar on, Baratta, and Cho (23)
4.21 0.12 SEPB 10 "
4.5 0.4 SEPB 3 Salem, Shannon and Jenkins (24)
4.60 0.30 SEPB Salem and Choi (25)
4.68 y 0.19 CN Short Bar 35 Salem and Shannon (26)
4.85 y CNB 4 Salem, Shannon and Jenkins (24)
4.64 y 0:2 CN Short Bar 13 Bubsey, Shannon, and Munz * (27)
4.72 CN Short Bar 7 "
4.71? CN Short Bar 9 "
4.85? CNB 2 "
4.85? CNB 2 "
4.42 0.14 CNB 2 VAMAS Round Robin (15)
4.5 5.0 Natural Flaws 6 Quinn and Quinn (19)
4.58 y 0.10 Vickers Indent. Strength 4 Kübler (28)
4.9 y Vickers Indent. Strength Salem and Choi (29)
4.1 y Vickers Indent. Strength Tikare and Choi (30)
5.2 Double torsion 4 Annis and Cargill (31)
4.9 Double torsion Evans and Charles (32)
4.1 0.21 Double torsion 4 Govila (33,34)
5.8 0.74 Double torsion 3 Quinn (35)
(continued)
p
Table 1. Fracture toughness (MPa m) for NC 132 silicon nitride (Cont’d).
Mean KIc S. dev. Method n Source (Ref.)
4.24 0.30 Double torsion Bansal and Duckworth (36,37)
3.9 5.0 Double torsion Bansal and Duckworth (39)
4.80 0.09 Double torsion Tressler, Yonushonis, Meiser (38)
4.93 0.31 Double Torsion Bourne and Tressler (39)
4.20 0.15 Fracture Mirror Analysis Bansal and Duckworth (36,37)
5.4 Fracture Mirror Analysis Bansal and Duckworth (40)
4.0 ? 5.0 Double Cantilever Beam Bansal and Duckworth (40)
4.0 ? Double Cantilever Beam 30 Freiman et al. (41)

*
Several chevron geometries and orientations. - Not reported
**
Several annealing conditions (air or inert atmospheres). ? Perpendicular to the HP direction
***
Annealed in air and cracks were probably healed. y Flat R-curve
These positive results prompted us to investigate the NC 132 further. Fracture

toughness test specimens were prepared from six separate billets of NC 132. One
thousand one hundred and thirty-one test specimens were prepared. Two hundred
thirty-seven experiments were conducted. Eventually five hundred and thirty-seven
specimens from three billets (C, G, D) were accepted for SRM 2100, and four
hundred sixty-four specimens from billets (A, H, and 4) were rejected and set aside
as will be discussed below.
3. THE SRM MATERIAL
This 1970 1980’s vintage material43 which was hot-pressed with a small amount 1 %
mass fraction of a magnesia-based sintering aid is composed of slightly elongated
beta phase silicon nitride grains with a maximum size of about 3 mm as shown in
Figure 4. A small amount of amorphous second phase resides in small triple point
pockets or as very thin layers between the silicon nitride grains. Very fine tungsten
carbide or tungsten silicide inclusions are dispersed throughout the bulk and were
picked up during ball milling of the starting powders. The material is fully dense and
has a density 3:22 g=cm3 to 3:26 g=cm3 , which is slightly greater than the theoretical
density of silicon nitride because of the tungsten compounds in the material. This
HPSN was originally fabricated in the form of billets (plates) of nominal size
158 mm 158 mm 20 mm to 30 mm thick as illustrated in Figure 5. NC 132 frac-
tures in a mixed transgranular and intergranular mode at room temperature. It has a
flat R-curve (whereby crack growth resistance is independent of crack size)24, 26–30
and is highly resistant to environmentally-assisted slow crack growth at room tem-
perature. Consequently, different test methods should produce identical fracture
toughness results.
NC 132 is no longer made by the manufacturer, so test specimens were prepared
from six billets that were on hand. From the six billets, only three labeled ‘‘C’’, ‘‘G’’,
Figure 4. Microstructure of the silicon nitride. (a) shows a fracture surface and is from the inside of a SCF
precrack; (b) is a polished and plasma etched section.
Figure 5. All test pieces were cut from the billet in the same orientation. The test specimen cross section
plane is parallel to the direction of hot pressing. The small arrows on the cross section planes show the
direction of crack propagation. The fracture toughness is certified only for crack planes parallel to the hot
pressing direction.
and ‘‘D’’ for short, were actually used for the SRM. The test specimens in each SRM
kit are all randomized from one billet whose identity is listed on the accompanying
certificate and scribed onto an end face of each test specimen. SRM 2100 is made up
exclusively of test specimens from billet C. When SRM 2100 is expended, test
specimens from billets G and D will be used for SRM 2100A and SRM 2100B.
Three additional billets, ‘‘H,’’ ‘‘A,’’ and ‘‘4’’ were also evaluated but were
ultimately rejected for the SRM. Billets A and 4 had excessive variability in fracture
toughness that was traced to a material inhomogeneity that manifested itself on the
test specimen surfaces as a faint patchwork mottling. Such mottling, commonly
detected in HPSN’s with rare earth oxide sintering additives, has even been reported
in magnesia-doped NC 132. In the case of NC 132, it is associated with elemental
silicon from a thermal decomposition of the silicon nitride.44 Billet A also had an
aberrant amount of aluminum detected by x-ray fluorescence analysis in one chem-
ical analysis sample. Billet H was rejected since, unlike the other billets, there was
evidence of environmentally-assisted slow crack growth in the SCF and SEPB
test specimens. Billet H also had a blistered appearance on one hot-pressed surface.
Specimens tested in laboratory air produced different lower KIc ’s than specimens
tested in dry-nitrogen gas environment. Halos around the SCF precracks, which are
signs of stable crack extension,45 were detected in the SCF specimens. The source of
the environmental sensitivity in billet H was traced to a slightly greater glassy
boundary phase content as detected on polished micrographs, but strangely, not
from the chemical analysis.
Tables 2 and 3 show chemical and density data for the billets. Carbon was
determined by a direct combustion infrared detection method, and oxygen by a direct
inert gas fusion-thermal conductivity method.i The other elements were measured by
x-ray fluorescence at NIST. X-ray diffraction revealed only b-Si3 N4 , some Si2 N2 O,
and WSi2 . No a-Si3 N4 was detected. The manufacturer’s densities were taken either
from the billet identification label or a chalk written marking the manufacturer
applied to the as-pressed billets. We also measured the density of five or more bend
bars taken at random after the billets were cut up. Density was computed from the
mass (measured to 0.001 g) and the geometric dimensions of these well-defined
prisms. Bar lengths were measured to 0.01 mm with a digital caliper and the cross
section dimensions to within 0.002 mm with a vernier micrometer. All four chamfers
sizes were measured and a correction made to the bar geometry. In this manner, the
density of a bar could be measured to within 0:002 g=cm3 . Table 3 shows that the
densities were very uniform. On the other hand a 2 mm thickness taper was detected
in billet D and the variability in bend bar density is greater.
NC 132 silicon nitride was one of the most studied advanced ceramic materials
of all time. Its fracture toughness is representative of many ceramics whose fracture
Table 2. Chemical analysis (mass fraction in percent) for the several billets.
Billet C O Al Ca Co Fe Mg W
C .45 2.8 .3 .02 .13 .22 .5 2.0
G .45 3.0 .4 ND .12 .31 .5 1.9
H .41 2.4 .4 .04 .13 .27 .5 1.7
A .39 4.6 6.7 .01 .15 .25 .4 2.3

*
E 1.1 3.3
Not measured
ND Measured, but not detected
*
VAMAS SCF Round Robin
Table 3. Densities (g/cm3) of the NC 132 billets.
Billet Manufacturer’s density NIST density
C 3.225 0.002
D 3.233 0.004
G 3.23 3.207 0.002
H 3.233 0.002
A 3.21 3.188 0.002
4 3.194 0.002
E 3.227 0.001
*
VAMAS SCF round robin
Not available
i
Tests performed by Leco Corp., St. Joseph, MI.
p p
toughness values lie in the range of 1:0 MPa m to 10:0MPa m. Although it is
obsolete, the material is relevant since derivatives are produced commercially today
for use in seal and bearing applications around the world. NC 132 has the fracture
toughness of a fully dense, fine beta silicon nitride grain material with slight grain
elongation, and only a small amount of boundary phase, and which has mixed trans-
and intergranular fracture mode.
Some consideration was given to using sintered a-silicon carbidej as a fracture
toughness reference material. It also has a flat R-curve and does not experience slow
crack growth at room temperature.46 It was a mature engineering grade structural
ceramic that was the focus of many studies. One p team went so far as to declare in
1989 its true fracture toughness was 3:0 MPa m.46 We did not use it for several
reasons. Tabulations of fracture toughness data for a-SiC in the 1990’s still showed
some variability. It is now known that some of the variability was in the test methods
and not the material. On the other hand, we have now confirmed that there were
batch-to-batchpvariations in fracture
p toughness.5,47 The fracture toughness varied
from 2:6 MPa m to 2:9 MPa m depending upon the batch. There was also some
concern that chevron notch beam testing may have been a little more difficult for
many users, since it is more difficult to achieve stable crack extension due to a-SiC’s
high elastic modulus and low fracture toughness. The decisive factor was our famil-
iarity with NC 132 from the international round robin which gave us an experience
base to expand upon on.
4. TEST METHODS USED FOR THE SRM
4.1. Specimen Preparation
A large sample set of representative flexural test specimens from each of the billets was
tested at NIST by the SCF and PB (SEPB) methods in accordance with C 1421 unless
otherwise noted. In addition, a few test specimens from billets C and H were tested by
the CNB method at NIST and at National Aeronautics and Space Administration
(NASA)-Glenn Center. As we will show, fracture toughness was independent of test
method. A substantial number of additional experiments using nonstandardized pro-
cedures were conducted but were not used for the SRM database. Data from these
dozens of supplemental experiments were very important for establishing the robust-
ness of the individual test methods and for confirming the tolerances and procedures in
the ASTM and ISO standards.
One of the advantages of the three methods in C 1421 is that they all use a simple
bend bar. Each 3 mm 4 mm 45 mm to 47 mm bar test specimen was prepared in
accordance with the generic (‘‘standard’’) preparation procedure specified in the
flexural strength standards MIL STD 1942 MR,48 ASTM C 1161,49and ISO
14704.50 One 4 mm wide face was finished with a 900 grit wheel (finer than is
normally required) to produce a flat and smooth surface for the Knoop indentation
in the SCF test method. The four long edges of the test specimens were chamfered in
accordance with requirements for the surface crack in flexure (SCF) and precracked
beam (SEPB) test methods.
j
Hexoloy SA grade, St. Gobain Carborundum, Co., Niagara Falls, NY.
All test specimens were cut with their 45 mm length axis perpendicular to the hot
pressing direction as shown in Figure 5. The certified fracture toughness is valid only
for cracks in planes that are parallel to the hot-pressing direction. Cracks may be
initiated or propagated from either the 3 mm wide or the 4 mm wide test specimen
surfaces. The fracture toughness is not certified for cracks propagated on planes
parallel to the billet’s 152 mm 152 mm faces, since these other planes may have
lower fracture toughness, due to a preferred orientation of the elongated beta silicon
nitride grains that tend to be aligned in the plane, normal to the hot-pressing direction.
The outer 6 mm of the billet sides was sliced off and not used since it became evident
that the fracture toughness was a little greater in these regions. (For
p example, in p billet
C, the average fracture toughness for 5 SCF tests was 4:90 MPa m 0:16 MPa m.)
X-ray radiographs of the billets revealed slight density gradients in the edges.
Groups of five test specimens were precracked and then fractured in a single
session. Fracture testing was in laboratory ambient conditions (20 8C to 25 8C;
relative humidities, 30 % to 72 %). Some sets were tested in dry-nitrogen gas for
comparison, in order to ascertain whether environmentally-assisted slow crack
growth might have been active. No such slow crack growth was detected in test
specimens from billets C, G, or D and environment had no effect. Slow crack growth
was active in some test specimens from billet H.
Conventional flexural strength (bend) fixtures with 20 mm 40 mm spans usu-
ally were used. Some broken halves were retested in the PB (SEPB) configuration in
either three- or four-point flexure strength fixtures with a shorter 16 mm or 20 mm
outer span. Fixtures were of either the semi-articulating or fully-articulating styles
and were in accordance with C 1421 and the well-known standard flexural strength
test standards C 1161 and MIL STD 1942. These standards have requirements for
fixture alignment and articulation, and loading rollers that are free to rotate in order
to eliminate frictional errors.
4.2. Surface Crack in Flexure (SCF) Precracking and Testing
The SCF method, also known as the ‘‘controlled flaw’’ method, is a development of
early work by Petrovic and colleagues.18,51 This method utilizes a Knoop hardness
indenter to create a single micro surface precrack such as shown in Figure 6. In brittle
ceramic materials and glasses, the indentation causes a very deep semi-elliptical crack to
open up underneath the hardness indentation. The crack is much deeper than the
indentation. Crack orientation may be controlled by orienting the Knoop indentation
axis. The indentation and its residual stresses must be removed by polishing or hand
grinding after indentation. The specimen is then fractured in a conventional flexural
strength fixture. Fractographic techniques are used to detect and measure the precrack
on the fracture surface after test specimen fracture. Fracture toughness is computed
from the maximum stress (or maximum load) at fracture, the crack size, and the stress
intensity factor solution. The precrack size must be measured for each test specimen.
The stress intensity shape factor should be individually calculated for each test speci-
men. Additional details are in the literature15–17,52 and ASTM C 1421 and ISO 18756.
Most of the SRM 2100 SCF database was generated with a 24.5 N (2.5 kgf )
indentation force. This is less than the 49.0 N recommended in C 1421, but the
VAMAS round robin showed that the smaller force was suitable. Less material had
to be removed by hand grinding around the smaller precrack. Cracks were usually
Figure 6. SEM photograph of SCF precrack at about 500X original magnification. The boundary is clear
on the top and right side (arrows), but can be easily extrapolated along the left side. Some precracks were
easier than this one to detect, and some were harder. The precrack is tilted at a slight angle relative to the
final fracture plane which accounts for the lighter contrasting features around the top.
put into a 4 mm 45 mm flat face. Semielliptical precracks were approximately

50 mm deep 150 mm wide after hand grinding removal of the indentation and
residual stress zone. The hand polishing/grinding was done in stages in order to
remove 23 mm to 25 mm of material, the necessary amount for these 24.5 N indenta-
tions. The first 10 mm to 15 mm of material was removed with a dry 30 mm diamond-
impregnated abrasive grit wheel on a low speed 200 mm diameter rotary polishing-
grinding machine with light finger pressure. The final 10 mm 15 mm was removed
by more delicate hand sanding-grinding with round sheets of dry 240 grit silicon
carbide abrasive paper, also mounted on the low speed rotary polishing machine. The
removal was done in many small steps and the test specimen thickness dimension W
was closely monitored with a hand micrometer with a resolution of 0.002 mm. Care
was taken not only to remove the prescribed material, but also to keep the cross
section as even as possible. One sample set was also evaluated with indentations
implanted into the narrower 3 mm wide face, but the outcomes were the same. A few
test specimens from billet C had the indentation and residual stress damage zone
machined off by surface grinding with a diamond wheel rather than by hand grinding.
SCF test specimens were only done in four-point loading since it is very difficult
to align the tiny precrack with the middle load roller in three-point loading.
The long edges were chamfered 0.08 mm to 0.13 mm in accordance with C 1421
in order to eliminate the chance that a chip or edge cracks would cause fracture.
Although small, these chamfers reduce the cross-sectional area and the second
moment of inertia of the cross section about the test specimen neutral axis. All
four chamfers were measured for each test specimen using a stereo binocular micro-
scope with a precision traversing stage with a resolution of 0.001 mm. An average
chamfer size was computed for each group of five test specimens and a single correc-
tion factor computed for the set. The fracture toughness was corrected (0.7 % to
1.0 %) for the change in the moment of inertia as directed in C 1421.
After fracture every precrack was photographed with a scanning electron micro-
scope (SEM), often with paired stereo photos, at magnifications from 200 X to 500 X.
If there was any doubt about an interpretation, the second fracture half was also
examined and photographed. An SEM length calibration standard was photo-
graphed in both the horizontal and vertical orientations with each set of test speci-
mens under examination (usually five at a time) in order to correct for the
magnification distortions or errors. The magnification correction was usually 3 %
to 5 %, but sometimes as large as 8 %. Users of the SRM 2100 need not resort to these
precautions and simpler SEM or optical microscopy techniques are effective in
finding and measuring the precracks, with only a small loss of precision or accuracy,
but magnification checks are a wise precaution, even in routine work.
It is beyond the scope of this paper to cover all details, but a remarkable
interaction between the calculated stress intensity shape factor and the measured
crack size was beneficial in mitigating errors or uncertainties in the crack size inter-
pretation or measurements.15,17 The stress intensity shape factor for a semi-elliptical
surface crack depends upon the precrack depth and width, and the test specimen cross-
section size. The Newman-Raju stress intensity factors (Y) for surface cracks in beams
in bending53 were used in accordance with C1421. For the SRM test specimen
geometries, it was discovered that a slight measurement error of the precrack size
(depth or width) was offset by a compensating error in the computed Y factor. That is
to say, a þ10 % error in a precrack depth measurement also led to a 5 % error in the
calculated stress intensity shape factor. The computed fracture toughness depends
upon the square root of the crack size and linearly with the Y factor.
The success rate in conducting the SCF experiments was of the order 90 % or
more. That is to say, for each set of five test specimens, we usually obtained four or
five valid outcomes. An invalid test usually corresponded to the precrack being too
irregular or too difficult to mark precisely.
4.3. Precracked Beam (SEPB) Precracking and Testing
This variant of the single-edge notched beam test uses a bridge precracker to pop in a
sharp precrack from a starting notch or indentation flaw.54 A bridge precracker was
used to pop in a 0.45 W to 0.55 W deep precrack, where the test specimen thickness,
W, was 4 mm. Figure 7 shows some specimens. A single 98 N (10 kgf) Knoop
indentation was used to produce a starter flaw in a ground 3 mm wide face. The
indented test specimen was carefully centered in the bridge anvil. With a 6 mm bridge
anvil gap, pop-in loads were between 8,000 N and 11,000 N. A simple stethoscope
attached to the bridge precracker was used to detect the pop-in, which was heard as a
faint snap in this material.
Four-point loading with 20 mm 40 mm spans was usually used with exceptions
noted below. Fracture loads were small (50 N to 120 N) and depended upon the
testing configuration and precrack size, so care was exercised in handling and pre-
loading the test specimens. No correction for test specimen chamfers is necessary or
appropriate for the SEPB test. After fracture, precracks were measured with a
precision traversing stage with a resolution of 0.001 mm under a stereo binocular
Figure 7. SEPB specimens. (a) shows fracture surfaces for one specimen. (b) and (c) show side views of
two different specimens. In each case the crack ran from bottom to top. Specimen height was 4 mm in each
case.
microscope at 16X to 40X magnification. The precrack boundary was quite obvious
on the fracture surfaces of this fine-grained, fully-dense material as shown in Figure
7a. Dye penetration was unnecessary.
Some test specimen halves from SEPB or SCF trials were retested. They were
precracked again and tested on 10 mm 20 mm four-point flexure fixtures, or on
three-point flexure fixtures with a 16 mm span. Three-point loading with 16 mm
span with a 4 mm tall specimen is a traditional configuration that is a scaled down
version of the notched beam test in ASTM E 399 for fracture toughness of metals.55
Only a very few tests were done in our work with this configuration since it was very
difficult to align the SEPB precrack with the middle roller in the bend fixture. As will
be shown below, there was no difference in outcomes for full-size specimens versus
their halves. The reject rate for the three-point configuration was greater, however,
due to crack twisting and alignment problems.
Precrack pop in and final fracture angles were measured on both sides of every
specimen. We used the angle from the side with the greatest deviation from a straight
crack to determine whether the fracture met the requirements of the ASTM, JIS, or
ISO standards. Figure 7b shows a specimen with acceptable precrack and final
propagation angles. Figure 7c shows a specimen that just barely passed the specifica-
tions, but which was rejected due to the sharp jog at the intersection of the precrack
and the final crack plane. Test specimens outcomes were rejected if either the precrack
or the final fracture crack was tilted too much.k Eighty-six percent of billet C test
specimens were accepted, eighty-one percent from billet G, and one hundred percent
from billet D. Every test specimen passed the precrack depth evenness criterion of
standards ASTM C 1421, JIS R 1607, and ISO 15732.
4.4. Chevron Notched Beam (CNB) Testing

A limited number of specimens from billets C and H were tested by the chevron notch
method.56 SRM 2100 specimens may be used with C 1421 configurations ‘‘A’’ or
‘‘D.’’ Specimens with ‘‘A’’ type notches were prepared by a professional machine
shop to the tolerances in C 1421. Figure 8 shows an example. The notch width was
less than 0.25 mm and was carefully machined to ensure that each side of the notch
lined up well as shown in Figure 8b. The notch tip in the A configuration is 0.80 mm
from the specimen end. This configuration was chosen since it has the most amount
of stable crack extension and the lowest crack velocity, and hence, stability can be
more readily detected. Two specimens from billet C were tested at NIST, but we
initially were uncertain as to whether stable fracture had occurred. We later con-
firmed that these two tests were indeed valid. Ten other billet C specimens were tested
at NASA-Glenn and eight valid outcomes were obtained. Stability was verified either
by back face strain gages or by displacement transducers contacting the specimen.
Instability was due to a chevron notch-tip having been chipped during machining in
one case; and residual glue from strain gaging accidentally being in the notch. Several
specimens were tested in three-point loading, but no meaningful difference in three-
versus four-point results was detected. The NIST and NASA-Glenn results con-
k
ASTM C 1421 and the ISO 15732 differ on this point. The former specifies an allowable final fracture
misalignment, whereas the latter specifies an allowable precrack misalignment. We rejected a datum for the
SRM if it failed either criterion.
Figure 8. Chevron notch specimen, configuration ‘‘A’’. (a) shows matching halves. (b) shows a close up
of one tip with the piece tilted to show how well the two sides of the chevron meet.
curred. Stress intensity factor coefficients (Ymin ) for the four-point specimens were
calculated on the basis of a straight-through crack assumption (STCA) model.
5. RESULTS
Table 4 and Figure 9 show the results by test method. The results for all data usually
were in extraordinary agreement within a given billet. An analysis of variation
analysis confirmed that the mean fracture toughness varied between billets, however.
Fracture toughness results for billets C and D are statistically significantly different
than the results for billet G. The source of the differences between billets is subtle, but
may be due to slight variations in the densities of the billets and to slight variations in
the secondary phase content.
Table 4. Fracture toughness data for hot pressed silicon nitride billets C, D, and G by the
three test methods in ASTM C 1421 99. SRM 2100 uses test specimens from billet C.
p p
Billet Method Mean MPa m Std. Dev. MPa m Number of outcomes
C (SRM 2100) SCF 4.56 (4.58)a 0.13 (0.16)a 17 (26)a
C (SRM 2100) CNB 4.58 0.12 10
C (SRM 2100) SEPB 4.58 0.10 19
D SCF 4.53 0.22 19
D SEPB 4.46 0.24 18
G SCF 4.29 0.19 15
G SEPB 4.27 0.16 17
a
SCF values in parenthesis include nine SCF test specimens wherein the damage zone and residual stresses
were removed by machine grinding and not by hand.
Fifty-eight specimens from billet C were expended. Some halves were retested as
SEPB experiments with shorter span flexure fixtures. Altogether, 75 experiments
were conducted on billet C test specimens from which the SRM 2100 database of
46 outcomes was constructed. An additional 234 test specimens were available for
SRM distribution from which 46 kits of 5 test specimens were prepared.
A discussed previously, nine additional SCF test specimens from billet C had the
indentation and residual stresses removed by a surface grinder as opposed to the
usual hand grinding with abrasive papers. These experiments confirmedpthat al-
though almost identical mean fracture toughness was obtained
p (4:63 MPa m), the
standard deviation was considerably larger (0:26 MPa m). The effect of adding
these nine test specimens to the other forty-six SCF outcomes is shown in the
bracketed data entries in the first line of Table 4 as well as in the open bars in
Figure 9. The precracks were more irregular than those from hand grinding. Appar-
ently machine grinding altered the indentation precracks and possibly propagated
portions of the crack front. It was also difficult to guarantee that each test specimen
had the correct amount of material removed with machine grinding since test
specimens were mounted together and ground simultaneously.
Thirty-six test specimens from billet G were tested and produced a database of
32 valid experiments, leaving 215 test specimens available for the SRM. The billet G
specimens had a 6.4 % lower fracture toughness than billet C, but the scatter was very
small. Testing G specimens in dry nitrogen had no effect. Table 3 shows that this
billet had a slightly lower density (0.5 % to 0.8 %) than billets C and D.
Twenty-five test specimens from billet D were tested and produced 37 test
results. Many SCF and SEPB fragments halves were retested as short SEPB speci-
mens to minimize the number of specimens expended from this smaller than normal
batch. Billet D had somewhat greater variability than the other two billets, yet on
average it had a similar fracture toughness as billets C and billet E (used in the
VAMAS round robin). The cause for the increased variability in billet D could not be
conclusively identified,
p but two of the
p SCF test pieces that had low fracture tough-
nesses (4:02 MPa m and 4:19 MPa m) also produced low fracture toughnesses
when their halves were retested as miniature SEPB specimens (as described below).
3.7 3.8 3.9 4.0 4.1 4.2 4.3 4.4 4.5 4.6 4.7 4.8 4.9 5.0 5.1 5.2
20
19
18
SRM 2100
17 Billet C
16
15
14
13
Frequency 12
11
10
9
8
7
6
5
4
3
2
1
0
10
9 SRM 2100 A or B
Billet D
8
7
Frequency
0
15
14 SRM 2100 B or A
13 Billet G
12
11
10
Frequency
9
8
7
6
5
4
3
2
1
0
3.7 3.8 3.9 4.0 4.1 4.2 4.3 4.4 4.5 4.6 4.7 4.8 4.9 5.0 5.1 5.2
Fracture toughness MPa√m
Figure 9. Results for billets C, D and G.

The SCF precracks for these two specimens may have had slight halos suggesting
that slow crack growth had occurred in these two specimens. Such halos were not
detected in the other billet D specimens. It is possible that the variability may be from
some regions in the billet that have slight sensitivity to slow crack growth. Although
this billet had a high density, it was tapered and had a slightly greater density
variation than the other billets. Only ninety specimens remain for the SRM since
parts of this billet had been used for other studies.
As previously noted, groups of five specimens were usually tested, the same
number that is included in a SRM 2100 kit. The mean fracture toughness for sets of
five test specimens also usually agreed within 1 % to 2 %, irrespective of test method.
Even the standard deviations for groups of five test specimens were consistent.
We investigated the data consistency further by conducting SEPB tests on the
fragments of some broken SCF and SEPB specimens from billets D and C. The
halves were tested on shorter 10 mm 20 mm four-point or 16 mm span three-point
fixtures. Table 5 shows the results. Most of the miniature specimen outcomes
Table 5. Results for repeat tests using the halves of previously broken billet D and C
specimens. In some instances, both halves were retested and double entries are noted in the
right column. All tests were in four point loading unless otherwise noted.
Full Sized Specimen Half Sized Specimen

p p
Billet Test Method, KIc , MPa m Test Method, KIc MPa m
D SCF 4.02 SEPB 4.10
D SEPB 4.47 SEPB 4.39 and 4.29
D SEPB 4.60 SEPB 4.60 and 4.75
C SEPB 4.39 SEPB 4.56 and 4.50
C SEPB 4.69 SEPB 4.60 and 4.63
C SEPB 4.46 SEPB 4.61 and 4.55*
C SEPB 4.72 SEPB 4.46
*
3 point flexure
Figure 10. Matching fracture surfaces for a single SRM 2100 test specimen. The specimen was first tested
by the SCF method and produced the fracture surface on the right. It is common to put the semielliptical
surface flaw into the wider face. The precrack is just barely visible (arrow) at this magnification within the
fracture mirror. This fracture surface has all the normal characteristics of a strength test fracture surface.
One half of the broken specimen was retested in the SEPB configuration as shown on the left. The precrack
extends half way into the test piece. The lower facture load for the SEPB configuration creates a much
soother fracture surface.
p
matched the full sized beam outcome within 0:1 MPa m. Figure 10 shows two
fracture surfaces from one such test set and illustrates the dramatic difference in
precrack styles and fracture modes. The SCF full sized specimens on the right side
has a tiny 51 mm deep semielliptical crack that is just barely visible in the middle of
the fracture mirror. One half of the specimen was retested in the SEPB configuration
with a 2 mm deep precrack shown on the left.
The remarkable within-billet and between-method data consistency confirms
that the C 1421 test methods have been optimized to the extent that genuine material
variability (billet-to-billet, or batch-to-batch) can now be discerned.
The comprehensive certificate that accompanies each SRM 2100 set lists the
fracture toughness for the set as well as the uncertainty associated with the estimate.
For each billet the mean fracture toughness and the scatter in results as measured by
the three test methods were statistically indistinguishable. The data were therefore
pooled for each billet. The certified average fracture toughness in Table 6 is the grand
mean of the pooled NIST database. The uncertainty U1 (with the subscript 1) is a
95 % prediction uncertainty for a single future observation and is based on the results
of the NIST observations from the same normally-distributed population.57 The
uncertainty Um (where the subscript m denotes the mean) is a 95 % confidence
interval for the mean of five future observations, also based on the results of the
NIST independently and randomly selected observations. The expanded uncertain-
Table 6. Certified fracture toughness values. The uncertainties are at the 95% confidence
interval. The numbers in parenthesis is the uncertainty expressed as a percentage of the mean
fracture toughness.
Uncertainty, U1 , Uncertainty, Um ,
Grand Mean Grand Std. Dev. Single specimen Mean of 5 specimens
p p p p
Billet KIc MPa m MPa m n MPa m (%) MPa m (%)
C (SRM 2100) 4.572 0.112 46 0.228 (5.0 %) 0.106 (2.3 %)
G 4.268 0.148 32 0.307 (7.2 %) 0.145 (3.4 %)
D 4.495 0.232 37 0.477 (10.6 %) 0.224 (5.0 %)
ties, U, represent the combined effects of material inhomogeneity and within-labora-

tory components of uncertainty and were calculated from:

1 1 1=2
U ¼t þ s (1)
m n
where s is the grand NIST sample standard deviation, m ¼ 1 for a single test
specimen, or m ¼ 5 for the SRM kit sample size, n is the NIST sample size and t is
the Student’s t parameter (¼ 2) for a 95 % confidence interval and the appropriate
degrees of freedom, (n 1). Table 6 shows that the smallest uncertainties are asso-
ciated with billet C since the database was larger and the standard deviation smaller
than for the other two billets. Once all the SRM 2100 kits have been distributed, then
SRM’s 2100a and 2100b will be available using the billet G and D test specimens.l
6. REFINEMENTS TO C 1421
Findings from the SRM 2100 project were helpful during the drafting of ASTM C
1421. In the paragraphs below, we give examples where lessons learned from the
SRM work aided setting the specifications and tolerances for the test methods.
A key issue with any fracture test method is the choice of proper stress intensity
factor coefficients. Accurate stress intensity factor coefficients are particularly of
concern for the CNB test specimen. This is mainly a result of its 3-dimensional
nature and the use of approximate solutions from the Slice and Straight-through-
crack-assumption (STCA) models. One key parameter is a1 , which is the normalized
dimension for the location of the end of the chevron on both sides of the specimens.
An a1 of 1.0 (configurations A, B, D in C 1421) means that the ends of the chevron
merge exactly with the side face and the back face of the test pieces. The Provisional
Standard PS070-97 used the stress intensity factor coefficients from the Slice model
for configurations A, B, C and D. During the conversion of this standard to the full
consensus standard C 1421, this matter was investigated more carefully. In order to
determine the best stress intensity factor coefficient function for ASTM C 1421, the
Slice and STCA models were compared to finite element analyses (FEA) and bound-
ary element analyses (BEA) solutions.58 It was determined that the STCA is a good
l
The price per kit is $350 in the year 2003. The price covers the cost of fabricating the specimens, SRM
project overhead costs, as well as a small fraction of the preliminary research costs.
approximation to the BEA for geometries with a1 ¼ 1:0), but a poor approximation
for geometries with a1 < 1 (e.g., configuration C in C 1421). In the latter case the
Slice model is a better but less than ideal approximation. (FEA results are currently
the best solution.) It is now believed that the STCA model is more accurate (by 5 - 10
percent) than the slice model for configurations A, B, and D. C 1421 changed the
formulation for Ymin to the STCA model for test configurations A, B, and D, and
retained the Slice model solution for configuration C. This was an important change.
The decision was reinforced by our SRM 2100 data. We compared calculated
fracture toughness results with both solutions for the SRM 2100 ‘‘A’’ configuration
specimens. Better agreement with the SCF and SEPB data was achieved with the
STCA model, thus reinforcing the analytical findings.
One key issue with the SEPB test is the straightness of the precracks and final
fracture planes. Ideally the precracks should pop in reasonably straight and then
should propagate directly to the back specimen face during final fracture. In practice,
precracks often pop in at an angle as illustrated in Figures 7 and 11. Many precracks
are curved. We initially measured the angle at the tip, uprecrack,tip , to check the
deviations, but we realized this would be impractical for routine work. Instead we
used a chord angle, uprecrack , as shown in Figure 11a. Final fracture may progress
along a different angle, ufinal . The slight shift in precrack and final fracture angle
actually is helpful in determining the depth of the precrack on the fracture surface.
How much of an angular departure from the ideal plane is acceptable? ASTM C 1421
differs from ISO 15732 and JIS R 1607 on this point. ASTM C 1421 sets an allowable
final fracture misalignment, whereas the latter two specify an allowable precrack mis-
alignment. JIS 1607 allows a maximum 108 precrack deviation for three-point load-
ing. ISO 15732 allows 108 for four-point loading, but only 58 for three-point loading.
(The tightening of this tolerance in ISO 15732 as compared to JIS 1607 was driven by
fear that a 108 tilted crack would be very hard to properly line up in a three-point
bend fixture, especially one with a 16 mm span.) In contrast, C 1421 sets a 108 limit
on the final fracture plane for four-point fracture and only 58 for three-point loading.
All these tolerances were chosen on the basis of prior work, but the SRM 2100 data
provided a useful check. The precrack and final crack angles were measured to within
0.58 on both sides of each SRM 2100 specimen. We rejected a datum for the SRM
database if it failed either a precrack or final fracture tolerance. Figure 11 shows the
results for all SEPB tests for billets C, D, and G and includes data rejected for SRM
2100. The fracture toughness values were normalized by the billet mean value to put
the data on a common basis. Most of the data is from four-point flexure. Several
points may be made. First, we usually had no trouble obtaining good precracks and
straight crack propagation and most cracks were well below the 108 limits. Second,
the trends do not show a dramatic change at 108 for either angle. Data may be
slightly more scattered at > 10 and may have a slight trend of increasing fracture
toughness. In summary, our SRM 2100 work indicates the tolerances chosen for the
world standards are reasonable and conservative.
Lessons learned from the SCF experiments also aided the drafting of the test
method standards ASTM C1421 and ISO 18756. We confirmed that it did not matter
whether the tiny Knoop precrack was implanted in the narrow 3 mm face or the wider
4 mm face. The formal test standards allow both indentation and loading configur-
ations. Slight (0.5 % 1.0 %) discrepancies in our SCF outcomes early in the project
were traced to the size of the chamfers in some of the preliminary bend bar sets. Bend
(b) (c)
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
1.5 1.5
1.4 1.4
Normalized Fracture Toughness

Normalized Fracture Toughness
1.3 1.3
1.2 1.2
1.1 1.1
1.0 1.0
0.9 0.9
0.8 0.8
0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0.0 0.0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
Crack Angle Crack Angle
Figure 11. The effect of SEPB crack angles upon the fracture toughness for all billet C, D, and G data.
Fracture toughness is normalized for the billet average in each case. (b) shows the effect of the precrack pop
in angle, uprecrack ; and (c) shows the effect of the final crack angle, ufinal .
bars and fracture toughness specimens often have chamfers applied to the edges in
order to eliminate chipping or machining damage. This must not be overdone, how-
ever, lest the cross section of the beam be altered too much. The relationship is not
linear: removing 1 % of the material from the cross section by using 458 chamfered
edges has a 2.8 % effect on stresses due to the change of the second moment of inertia of
the cross section. Hence the flexural strength standards have tolerances on the size of
the chamfers. The maximum 458 chamfer size is 0.15 mm maximum for a
3 mm 4 mm bend bar. The SCF method is in essence a strength test with a well-
defined tiny flaw. Once the bending stresses were corrected for the chamfers, we were
amazed to see an appreciable diminishment in the SRM 2100 data variability. The
small discrepancies diminished or disappeared altogether. ASTM C 1421 and ISO
18756 have tolerances for allowable chamfer sizes. The latter standard even includes
correction factors for specimens with oversized chamfers.m
Although at NIST we find it easy to remove the Knoop SCF indentation and it’s
residual stress damage zone with a diamond abrasive disk or abrasive SiC papers,
some users prefer to do this step via surface grinding with a grinding machine. Our
SRM 2100 work showed this is acceptable, except that it is difficult to verify each
specimen has had the correct amount of material removed. There also is a risk that
the precrack may be enlarged or disturbed by the machine grinding. There are some
hard and tough ceramics where it would be impractical to remove the requisite
material by hand grinding or polishing and it is valuable to know that machine
grinding is acceptable if precautions are taken. With practice with the SRM 2100
specimens, we became adept at controlling the exact amount of material that was
removed and ensuring that the hand ground surface remained reasonably flat and
m
Such chamfers corrections are not appropriate for the SEPB or CNB tests, however.
not crowned. It is routine to control the amount removed to within 2 mm to 4 mm. It

was even possible to tailor the shape of the semielliptical precrack so that the
maximum stress intensity shape factor was either at the deepest part of the crack,
or was approximately constant around the flaw periphery. The Y factors for the
deepest point were usually 1.3 to 1.5. At the surface, the Y factors were 1.25 to 1.30.
Precracking is an important step in most fracture toughness tests and it was reassur-
ing that we had good control and were the ‘‘masters’’ of the SCF precracks.
Anyone who has ever tried the SCF method has worried about how well they
need to discern the SCF precrack. The more Knoop precracks we prepared, the more
we recognized some of the nuances of their shapes and boundaries. In some cases the
full boundary was not clear (e.g. the left side of the precrack in Figure 6), but we
realized that reasonable extrapolations of the precrack boundary could be made with
no sacrifice in data quality. On the basis of the SRM 2100 work as well as studies on
other materials, we determined that reasonable estimates of the crack size and shape
may be made if two-thirds of the boundary is visible. A specification to this effect is
incorporated in C 1421. Users also often worry about how precisely they need to
measure the tiny precrack. We discovered to our delight that measurements do not
have to be especially precise! We frequently made a best estimate of precrack size and
computed toughness, but then checked what the effect would be if we added 10 mm to
a dimension. Much to our surprise, the effect on computed toughness was often
negligible. This is due to the dependence of fracture toughness on the square root of
crack size, so uncertainties in the size measurement are diminished in the calculated
fracture toughness. Furthermore, as discussed previously, there is a serendipitous
interaction of the computed Y factor with errors in the crack size measurements:
errors in the latter are offset by compensating errors in the former. On the other
hand, it is not sufficient to measure an average stress intensity shape factor Y for a
group of specimens. Each specimen should have its precrack measured and its own Y
calculated. Both C 1421 and ISO 18756 are explicit on this point.
So in summary, detailed ‘‘nuts and bolts’’ work with the SRM 2100 specimens
for these methods led us to appreciate which details are important for ensuring
accurate and precise fracture toughness measurements, and which details are not
important.
7. APPLICATIONS OF SRM 2100
7.1. General
SRM 2100 may be used a variety of ways. The most obvious usage will be to help
engineers, researchers, and technicians confirm that their SEPB, SCF, or CNB
procedures are sound. Our use of NC 132 in the 1994 to 1995 VAMAS SCF round
robin exemplifies this. Most users obtained good results on specimens from billet E.
In the few instances where problems occurred, and data was atypical or deviant, it
was usually possible to review the testing procedure and find the origin of the
discrepancy.
We do not recommend that the five specimens in an SRM 2100 kit be given
immediately to a novice. Less expensive practice specimens should be used first until
the tester has some familiarity with the test method. Once the tester is familiar with
the test procedure and some of its nuances, then the SRM 2100 specimens may be
used to check for accuracy and precision.
As time goes on, we hope the user community will benefit from the test method
refinements. More and more examples of consistent fracture toughness data are now
starting to appear in the literature. References 59 and 47 show remarkable consist-
ency of SCF, CNB and SEPB data for a toughened silicon nitride and a sintered
silicon carbide, respectively. SRM 2100 is available to support ISO and the European
Community fracture toughness standards. The reference material should facilitate
specifications writing and trade. The paragraphs below give some examples of the
utility of SRM 2100.
7.2. Facilitating the refinement and adoption of new methods
SRM 2100 should facilitate the introduction of new alternative test methods. For
example, a popular method that may be adopted in the European Community
standard is the single-edge V-notched beam (SEVNB)60,61 procedure. SEVNB refines
the older SENB method by sharpening the saw cut slit with a common single edge or
shaving razor blade and fine diamond paste.62 NIST and the Swiss Federal Research
Laboratory (Dubendorf, Switzerland) both tested sets of SRM 2100 specimens with
the results shown in Table 7. They confirm that the SEVNB results were consistent
but underestimated the certified fracture toughness by 5 %, a difference that was
statistically significant at the 99.9 % level using the Student’s t-test to compare the
means. It is believed that notch root issues account for the difference, and indeed,
fractographic examinations of the notch root of this and other ceramics often show
that the notch root has a ‘‘start up zone’’ of the order of 100 mm. Addition of this
to the initial crack lengths brings the SEVNB results into better agreement with the
certified values. Although the SEVNB notch is very sharp, it is still not as sharp as a
bona fide crack. Other experience with the SEVNB method confirm that SEVNB
results are usually within 5 % to 10 % of results obtained with sharp crack, stand-
ardized fracture mechanics test methods. This 5 % to 10 % discrepancy should be
resolved before the SEVNB method is formalized.
7.3. Materials Specifications

New ceramic material specifications may incorporate fracture toughness character-
ization. One example is the new ASTM Standard F 2094 01 for silicon nitride ball
bearings.63 Three grades of bearing balls are specified in this standard and fracture
Table 7. Comparison of SEVNB results to standard test methods results. Each entry has the
p
mean and standard deviation fracture toughness in MPa m, and the number of test specimens
is in parenthesis.
Billet SEVNB data C 1421 data Methods
C (SRM 2100) 4.34 0.19 (5) This paper 4.57 0.11 (46) SCF, SEPB, CNB This paper
E 4.36 0.08 (6) Ref. 62 4.59 0:37 (107) SCF VAMAS round robin, Refs. 15 17
toughness is one of the key determinants whether a bearing grade material is Class I
(KIc $6:0), Class II (KIc $5:0), or Class III (KIc $5:0).n
A time goes on, more specifications like this may be crafted, possibly for
biological materials such as surgical implants. Several ASTM standards for alumina
and zirconia materials for surgical implants are on the books, but only require
flexural strength, elastic modulus, and hardness at the moment.
8. WHAT IS THE REAL FRACTURE TOUGHNESS?
The SRM 2100 results are reassuring in the sense that they are extraordinarily
consistent and independent of test method. Claims of ceramic fracture toughness
measurements that were accurate and precise to within a few percent were unheard of
in the 1980’s and early 1990’s. It would seem that the experimental errors now have
been ‘‘wrung out’’ of the individual test methods to the extent that the methods are
reliable and the test method uncertainties are of the order of a few percent. The
scatter of the SRM data base (Table 6) is smaller than the scatter in the VAMAS
round robin (Table 1). This is not surprising since the round robin scatter included
material variability, within-laboratory test method uncertainty, plus the between-
laboratory method uncertainties.
Our SRM results are also consistent with 30 years of test data from a variety of
sources (Table 1). Indeed, our experience with the six billets studied for the SRM as
well as billet E, and data from the old stress rupture and fracture map work cited
earlier, suggests that NC 132 actually did have billet-to-billet variability after all.
Other investigators have commented on variability between and within billets, but
often furnished no concrete data or stated that the variation could not be correlated
with any observable microstructural variations or inhomogeneities.40,64,65,66,67,68
p
We conclude that the fracture toughness for NC 132 is 4:6 MPa m for fully
dense plates. This fracture toughness was attained in the billets that had the full
density of 3:23 g=cm3 . Although billet D had this density, it also had greater scatter
since the billet had a small taper after hot pressing and some density gradients. Billet
G had a slightly lower density ( 0:5 %) and, in turn, a 6.4 % reduced fracture
toughness compared to billet C. We also detected a slightly greater boundary phase
content from the microstructural analysis, but this was not confirmed by the chem-
ical analysis. So in summary, it seems that slight density, microstructural, and
chemical variations have measurable effects upon fracture toughness, and that the
refined fracture toughness tests methods can now discern the effect.
Is the fracture toughness of hot-pressed silicon nitride, grade NC 132 independ-
ent of crack size? Does a 40 mm deep crack have the same fracture resistance as a
2.5 mm long crack? Our data suggests the fracture toughness is indeed invariant over
this range. Tiny 40 mm deep SCF precracks showed no significant signs of stable
crack extension on the fracture surfaces,o nor did the larger 2.5 mm cracks in SEPB
n
The standard unfortunately also included a Vickers indentation crack length method as an alternative
to estimate fracture resistance. The measured property in that instance is deemed as ‘‘indentation fracture
resistance’’ and not a genuine fracture toughness.
o
An important exception was for billet H specimens wherein cracks did have stable crack extension, but
from environmentally assisted slow crack growth and not from a rising R curve behavior. Specimens
tested in inert atmosphere did not have any stable extension.
specimens. The natural inherent flaws in this material are of the order of 15 mm to
35 mm in size (depth of a surface crack or radius for an internal flaw). Perhaps there is
an ultimate short p toughness (crack size 0 mm) which may be as
p crack, tip fracture
low as 1:5 MPa m to 2:5 MPa m for tiny cracks in dense polycrystalline silicon
nitrides. If the toughness is this low, then the R-curve must come up very quickly
with increasing crack size or extension. Any other trend could not account for our
observations and those of other investigators over the years who have deduced this
material does indeed have a flat R-curve.24,26–30
Our work has shown that for all practical purposes the fracture toughness of this
simple MgO-doped hot-pressed, dense silicon nitride, with mixed trans- and inter-
granular fracture, and with slightly elongated grains of the order p of 1 mm to 3 mmpin
size, is invariant with crack size (40 mm to 2.5 mm) and is 4:3 MPa m to 4:6 MPa m
depending on the billet and its density.
It is ironic that fracture toughness test methods for ceramics, which are bona fide
brittle materials, have come so late to the scene. JIS R 1607 arrived in 1990, ASTM C
1421 in 1999, and the ISO and CEN standards in 2003. This is 20 - 33 years after ASTM
E 399 was first adopted for metals in 1970. So although ceramists were late, we need not
include in our standards the singularly pessimistic statement in E 399:p
‘‘There is no accepted ‘‘standard’’ value for the plane strain fracture toughness of any material.
In the absence of such a value, any statement concerning bias is not meaningful.’’
Our reply is simple. There now is a reference material with a certified fracture
toughness. The following quote is taken directly from the NIST Certificate that
comes with every SRM 2100 kit:
‘‘This SRM is intended for the verification of fracture toughness testing procedures and consists
of a set of 5 hot pressed silicon nitride flexure specimens . . . The certified fracture toughness is
p
4:572 MPa m and the uncertainty at the 95% confidence interval for a single specimen is
p p
0:228 MPa m, and for the mean of five specimens, is 0:106 MPa m.’’
9. CONCLUSIONS
The first reference material in the world for the property fracture toughness has been
created in tandem with ASTM C 1421. Both were completed within months of each
other in mid-1999. The fact that multiple test methods produce virtually identical
results strongly validates the ASTM and ISO standard test methods for ceramics.
Methods for measuring fracture toughness have been optimized to the extent that
genuine material differences may now be discerned. This meets one goal of any
standard: to enable the user to measure the desired property with good accuracy
and precision.
The 115 valid results in Table 2 (out of 152 total experiments) that constitute the
SRM database on billets C, G, and D are supported by years of work and hundreds
of additional fracture toughness experiments on NC 132. Altogether, 582 fracture
toughness experiments on this particular material have been conducted over the last
eight years in our ceramic fracture toughness standardization program. One hundred
and seven kits are available for sale. Forty-six are from billet C for SRM 2100.
p
Precision and Bias paragraph 11.3.
Eighteen and forty-three kits from billets D and G, respectively, will be used for
SRM’s 2100a and 2100b.
It is hoped that the ceramic community will begin using the new standard
methods, but this will probably be a gradual process. Rudimentary indentation
methods will continue to have some allure, no matter how dubious the quality of
the data. Hopefully the ceramic community will recognize the value of refined
standard test methods and the methods will restore some long overdue credibility
to ceramic fracture toughness data.
ACKNOWLEDGMENTS
Nien-Fan Zhang of the NIST Statistical Engineering Division assisted with the
statistical interpretations. John Sieber of the Chemistry Division furnished the x-
ray fluorescence data. Michael Foley of St. Gobain/Norton contributed with helpful
consultations. The U. S. Department of Energy, Office of Transportation Technolo-
gies, Heavy Vehicle Propulsion Systems Materials Program is acknowledged for its
support of the SCF test method development under contract DE-AC05-840R21400
with Martin Marietta Energy Systems, Inc.
REFERENCES
1. ASTM C 1421 99, Standard Test Method for Determination of Fracture Toughness of Advanced
Ceramics at Ambient Temperature,’’ Annual Book of Standards, Vol. 15.01, ASTM, West Consho
hocken, PA, 1999.
2. G. D. Quinn, J. A. Salem, I. Bar On, and M. J. Jenkins, The New ASTM Fracture Toughness of
Ceramics Standard: PS 070 97, Ceramic Engineering and Science Proceedings, 19 (3) 565 578 (1998).
3. G. D. Quinn, I. Bar On, M. G. Jenkins, and J, A. Salem, Fracture Toughness Standards Test Method
C 1421 99 for Advanced Ceramics, Korean Journal of Ceramics, 4 (4) 311 322 (1998).
4. I. Bar On, G. D. Quinn, J. A. Salem, and M. J. Jenkins, Fracture Toughness Standard Test Method C
1421 99 for Advanced Ceramics, pp. 315 335 in Fatigue and Fracture Mechanics, Vol. 32, ASTM
STP 1406, ed. R. Chona, American Society for Testing and Materials, West Conshohoken, PA, 2001.
5. J. A. Salem, G. D. Quinn, and M. J. Jenkins, Measuring the Real Fracture Toughness of Ceramics:
ASTM C1421, this volume.
6. ISO 15732, Fine Ceramics (Advanced Ceramics, Advanced Technical Ceramics) Test Method for
Fracture Toughness at Room Temperature by Single Edge Precracked Beam (SEPB) Method,
International Organization for Standards, Geneva, 2003.
7. ISO 18756, Fine Ceramics (Advanced Ceramics, Advanced Technical Ceramics) Determination of
Fracture Toughness of Monolithic Ceramics at Room Temperature by the Surface Crack in Flexure
(SCF) Method, ISO Geneva, 2003.
8. ISO xxxxx, draft, Fine Ceramics (Advanced Ceramics, Advanced Technical Ceramics) Test Method
for Fracture Toughness of Monolithic Ceramics at Room Temperature by Chevron Notched Beam
(CNB) Method, ISO, Geneva, 2003.
9. PrENV xxxx, European Standard, Advanced Technical Ceramics Monolithic Ceramics Fracture
Toughness Parts 1 5, European Committee for Standardization, Brussels, 2003.
10. Japanese Industrial Standard R 1607 1990, Japanese Standards Association, Tokyo, February, 1990.
11. D. C Larsen, J. W. Adams, and S. A. Bortz, Survey of Potential Data for Design Allowable MIL
Handbook Utilization for Structural Silicon Based Ceramics, final report on IITRI/AMMRC Con
tract DAAG46 79C 0078, U.S. Army Materials and Mechanics Research Center, Watertown, MA,
Dec. 1981.
12. Reliability of Ceramics for Heat Engine Applications, National Materials Advisory Board Report,
NMAB 357, Nat. Acad. of Sci., Washington, DC, 1980.
13. G. D. Quinn, Properties Testing and Materials Evaluation, Ceramic Engineering and Science Pro
ceedings, 5 (5 6) 298 311 (1984).
14. G. D. Quinn, VAMAS After Twelve, Bulletin of the American Ceramic Society, 78 (7) 78 83 (1999).
15. G. D. Quinn, J. J. Kübler, and R. J. Gettings, Fracture Toughness of Advanced Ceramics by the
Surface Crack in Flexure (SCF) Method: A VAMAS Round Robin, VAMAS Technical Report #17,
National Institute of Standards and Technology, Gaithersburg, MD 20899, June, 1994.
16. G. D. Quinn, R. J. Gettings, and J. J. Kübler, Fracture Toughness of by the Surface Crack in Flexure
(SCF) Method: Results of the VAMAS Round Robin, Ceramic Engineering and Science Proceedings,
15 (5) 846 855 (1994).
17. G. D. Quinn, R. J. Gettings, and J. J. Kübler, Fracture Toughness of Ceramics by the Surface Crack
in Flexure (SCF) Method, Fracture Mechanics of Ceramics, Vol.11, eds. R. C. Bradt, D. P. H.
Hasselman, D. Munz, M. Sakai, and V. Yashevchenko, (Plenum, NY, 1996), pp. 203 218.
18. J. J. Petrovic, L. A. Jacobson, P. K. Talty, and A. K. Vasudevan, Controlled Surface Flaws in Hot
Pressed Si3 N4 , J. Am. Ceram. Soc., 58 (3 4) 113 116 (1975).
19. G. D. Quinn and J. B. Quinn, Slow Crack Growth in Hot Pressed Silicon Nitride, in ‘‘Fracture
Mechanics of Ceramics, Vol. 6, eds. R. C. Bradt, A. G. Evans, D. P. H. Hasselman, and F. F. Lange,
(Plenum, NY, 1983) pp. 603 636.
20. J. J. Swab and G. D. Quinn, unpublished results, billet P, 1997.
21. V. Tikare and S. R. Choi, Combined Mode I and Mode II Fracture of Monolithic Ceramics, J. Am.
Ceram. Soc., 76 (9) 2265 2272 (1993).
22. S. T. Gonczy and D. L. Johnson, Impact Fracture at High Temperature, in Fracture Mechanics of
Ceramics, Vol. 3, eds. R. C. Bradt, D. P. H. Hasselman, and F. F. Lange, (Plenum, NY, 1978)
pp. 495 506.
23. I. Bar on, F. I. Baratta, and K. Cho, Crack Stability and Its Effect on Fracture Toughness of Hot
Pressed Silicon Nitride Beam Specimens, J. Am. Ceram. Soc., 79 (9) 2300 2308 (1996).
24. J. A. Salem, J. L. Shannon, Jr., and M. Jenkins, Some Observations in Fracture Toughness and Fatigue
Testing with Chevron Notched Specimen, in Chevron Notch Fracture Test Experience: Metals and Non
Metals, ASTM STP 1172 eds. K. R. Brown and F. I. Baratta, (ASTM, Phil., PA, 1992) pp. 9 25.
25. J. A. Salem and S. R. Choi, Toughened Ceramics Life Predictions Project, in Propulsion System
Materials Program Bimonthly Technical Progress Report, DOE Office of Transportation Technolo
gies, Dec. 1995 Jan. 1996, ORNL, Oak Ridge, TN, 1996, pp. 180 187.
26. J. A. Salem, and J. L. Shannon, Jr., Fracture Toughness of Si3 N4 Measured With Short Bar Chevron
Notched Specimens, J. Mat. Sci., 22 321 324 (1987).
27. R. T. Bubsey, J. L. Shannon, Jr. and D. Munz, Development of Plane Strain Fracture Toughness
Test for Ceramics Using Chevron Notched Specimens, in Ceramics for High Performance Applica
tions III, Reliability, (Plenum, NY, 1983), pp. 753 771.
28. J. J. Kübler, Unpublished research.
29. J. A. Salem and S. R. Choi, Toughened Ceramics Life Prediction, pp. 220 234 in Ceramics Technol
ogy Project Bimonthly Progress Report, Oct. Nov. 1991, Oak Ridge National Laboratory, Oak Ridge,
TN, 1991.
30. V. Tikare and S. Choi, Influence of the Si3 N4 Microstructure on its R curve and Fatigue Behavior,
Cer. Eng. Sci. Proc., 12 (7 8) 1437 1447 (1991).
31. C. G. Annis and J. S. Cargill, Impact Fracture of Ceramics at High Temperature, in Fracture
Mechanics of Ceramics, Vol. 4, eds. R. C. Bradt, D. P. H. Hasselman and F. F. Lange, (Plenum,
NY, 1978), pp. 737 744.
32. A. G. Evans and E. A. Charles, Fracture Toughness Determinations by Indentation, J. Am. Ceram.
Soc., 59 (7 8) 371 372 (1976).
33. R. K. Govila, Material Parameters for Life Prediction in Ceramics in Ceramics for High Performance
Applications III, Reliability, (Plenum, NY, 1983), pp. 535 567.
34. R. K. Govila, Indentation Precracking and Double Torsion Methods for Measuring Fracture
Mechanics Parameters in Hot pressed Si3 N4 , J. Am. Ceram. Soc., 63 (5 6) 319 326 (1980).
35. G. D. Quinn, Unpublished results.
36. G. K. Bansal and W. H. Duckworth, Effects of Specimen Size on Ceramic Strength, in Fracture
Mechanics of Ceramics, Vol. 3, eds. R. C. Bradt, D. P. H. Hasselman, and F. F. Lange, (Plenum, NY,
1978), pp. 189 204.
37. G. K. Bansal and W. H. Duckworth, Fracture Toughness of Hot Pressed Si3 N4 , Am. Ceram. Soc.
Bull., 14 254, 1984.
38. R. E. Tressler, T. M. Yonushonis, and M. D. Meiser, Crack Propagation in Molten Salt Corrosion of
Si3 N4 , private communication, 1976.
39. W. C. Bourne and R. E. Tressler, ‘‘Alteration of Flaw sizes and KIc ’s of Si3 N4 Ceramics by Molten
Salt Exposure,’’ in Fracture Mechanics of Ceramics, Vol. 3, ed. R. C. Bradt, D. P. Hasselman, and
F. F. Lange, (Plenum, NY, 1978), pp. 113 124.
40. G. K. Bansal and W. H. Duckworth, Fracture Surface Energy Measurements by the Notched Beam
Technique, in Fracture Mechanics Applied to Brittle Materials, ASTM STP 678, ed. S. W. Freiman,
(ASTM, Philadelphia, PA, 1979), pp. 38 46.
41. S. W. Freiman, A. Williams, J. J. Mecholsky, Jr., and R. W. Rice, Fracture of Si3 N4 and SiC, in
Ceramic Microstructures 1976, eds. R. M. Fulrath and J. A. Pask, (Westview Press, Boulder, Co,
1977), pp. 824 834.
42. G. D. Quinn, Fracture Mechanism Maps for Advanced Structural Ceramics, Part I, Methodology
and Hot Pressed Silicon Nitride, Journal of Materials Science, Vol. 25, 1990, pp. 4361 76.
43. M. L. Torti, Processing Hot Pressed Silicon Nitride for Improved Reliability: HS 110 to NC 132,
Ceramics for High Performance Applications, III Reliability, eds., E. Lenoe, R. N. Katz, and J. J.
Burke, (Plenum, NY, 1983), pp. 261 273.
44. D. R. Clarke, A Large Scale Processing Inhomogeneity in Silicon Nitride Ceramics and Its Effect on
Oxidation, Materials Science Forum, 47 (Trans Tech. Publ., Switzerland, 1989), pp. 110 118.
45. J. J. Swab and G. D. Quinn, Effect of Precrack ‘‘Halos’’ on KIc Determined by the Surface Crack in
Flexure Method, J. Am. Ceram. Soc., 81 (9) 2261 2268 (1998).
46. A. Ghosh, M. G. Jenkins, K. W. White, A. S. Kobayashi, and R. C. Bradt, Elevated Temperature
Fracture Resistance of a Sintered a Silicon Carbide, J. Am. Ceram. Soc., 72 [2] 242 47 (1989).
47. G. D. Quinn and J. A. Salem, Effect of Lateral Cracks Upon Fracture Toughness Determined by the
Surface Crack in Flexure Method, J. Am. Ceram. Soc., 85 (4) 873 80 (2002).
48. MIL STD 1942 (MR) Flexural Strength of High Performance Ceramics at Ambient Temperature,
U. S. Army Materials Technology Laboratory, Watertown, MA, Nov. 1983.
49. ASTM C 1161, Test Method for Flexural Strength of Advanced Ceramics at Ambient Temperature,
Annual Book of Standards, Vol. 15.01, ASTM, West Conshohocken, PA, 2002.
50. ISO 14704, Fine (Advanced, Advanced Technical) Ceramics Determination of Flexural Strength at
Room Temperature, International Organization for Standards, Geneva, 2002.
51. J. J. Petrovic, and M. G. Mendiratta, Fracture from Controlled Surface Flaws, Fracture Mechanics
Applied to Brittle Materials, ASTM STP 678, ed. S. Freiman, (ASTM, West Conshohocken, PA,
1979), pp. 83 102.
52. G. D. Quinn, R. J. Gettings, and J. J. Kübler, Fractography and the Surface Crack in Flexure (SCF)
Method for Evaluating Fracture Toughness of Ceramics, Fractography of Glasses and Ceramics,
Ceramic Transactions, 64, (American Ceramic Society, Westerville, OH, 1996), pp. 107 144.
53. J. C. Newman, Jr., and I. S. Raju, An Empirical Stress Intensity Factor Equation for the Surface
Crack, Engineering Fracture Mechanics, 15 (2) 185 92 (1981).
54. T. Nose and T. Fuji, Evaluation of Fracture Toughness for Ceramic Materials by a Single Edge
Precracked Beam Method, Journal of the American Ceramic Society, 71 (5) 328 333 (1988).
55. ASTM E 399 90, Standard Test Method for Plane Strain Fracture Toughness of Metallic Materials,
Annual Book of Standards, Vol. 3.01, ASTM, West Conshohocken, PA, 1990.
56. J. A. Salem, J. L. Shannon, Jr., and M. G. Jenkins, Some Observations in Fracture Toughness and
Fatigue Testing with Chevron Notched Specimen, Chevron Notch Fracture Test Experience: Metals
and Non Metals, ASTM STP 1171, eds. K. R. Brown and F. I. Baratta, (American Society for
Testing and Materials, West Conshohoken, PA, 1992), pp. 9 25.
57. G. Hahn, and W. Meeker, Statistical Intervals: A Guide for Practitioners, Wiley, NY, 1991.
58. J.A. Salem, L.J. Ghosn, M.G. Jenkins, and G.D. Quinn, Stress Intensity Factor Coefficients for
Chevron Notched Flexure Specimens and Comparison of Fracture Toughness Methods, Ceramic
Engineering and Science Proceedings, 20 (3) 503 512 (1999).
59. G. D. Quinn, J. J. Swab, and M. J. Motyka, Fracture Toughness of a Toughened Silicon Nitride by
ASTM C 1421, to be publ. J. Am. Ceram. Soc., 2003.
60. J. J. Kübler, Fracture Toughness of Ceramics Using the SEVNB Method: First Results of a Joint
VAMAS/ESIS Round Robin, Ceramic Engineering and Science Proceedings, 20 (3) 495 502 (1999).
61. J. J. Kübler, Fracture Toughness of Ceramics Using the SEVNB Method; Round Robin, VAMAS
Report No. 37, Swiss Federal Laboratories for Materials Testing and Research, Dübendorf, Switz
erland, 2000.
62. J. J. Kübler, Fracture Toughness of Ceramics Using the SEVNB Method: Preliminary Results,
Ceramic Engineering and Science Proceedings, 18 (4) 155 162 (1997).
63. ASTM F 2094 1, Standard Specification for Silicon Nitride Bearing Balls, Annual Book of Stand
ards, Volume 14.02, ASTM, West Conshohocken, PA, 2001.
64. K. D. McHenry, T. Yonushonis, and R. E. Tressler, Low Temperature Subcritical Crack Growth in
SiC and Si3 N4 , J. Am. Ceram. Soc., 59 (5 6) 262 263 (1976).
65. K. Jakus, J. E. Ritter, Jr., and J. P. Fahey, Effect of Billet Difference on the High Temperature
Strength of Hot Pressed Silicon Nitride, J. Am. Ceram. Soc., 65 (9) C143 C144 (1982).
66. M. E. Gulden and A. G. Metcalfe, Stress Corrosion of Silicon Nitride, J. Am. Ceram. Soc., 59 (9 10)
391 396 (1976).
67. D. G. Miller, C. R. Booher, Jr., R. J. Bratton, ‘‘Brittle Materials Design, High Temperature Gas
Turbine Stator Vane Development and Static Rig Tests, Westinghouse Final Report, U. S. Army
Materials and Mechanics Research Center, Watertown, MA, Technical Report, AMMRC CTR 76
32, Volume II, page 224, 1976.
68. D. W. Richerson and T. M. Yonushonis, Properties of Silicon Nitride Rotor Blade Materials,
pp. 193 217 in Ceramic Gas Turbine Demonstration Program Review, Castine MA, Metals and
Ceramics Research center, Battelle Columbus Technical Report, MCIC 78 36, March 178.
STRAIN ACCUMULATION
AND DAMAGE EVOLUTION
DURING CREEP OF SiCf =SiC
COMPOSITES
Brian Wilshire and Howard Burt*
1. INTRODUCTION
For many high-performance applications, worldwide research efforts continue to be

focussed on ceramic-fibre-reinforced ceramic-matrix composites (CFCMCs), with
numerous studies featuring SiC-fibre-reinforced SiC-matrix materials (termed
SiCf =SiC type products). In particular, because these CFCMCs are being considered
for components which must operate for long periods without failure under load in
hostile high-temperature environments, special attention has then been directed to
characterization of their creep and creep fracture behaviour. In turn, many of these
studies have been concerned with clarification of the damage processes which cause
creep failure, aiming to acquire the understanding needed for future product devel-
opment and component design.
Creep strain/time curves have been determined for a range of SiCf =SiC type
composites,1–6 for various types of SiC fibre7–9 and for other CFCMCs.5, 6, 10 How-
ever, the factors controlling the creep properties of different composites have usually
been discussed by reference to only a few standard parameters, such as the minimum
creep rate (_em ), and the creep rupture life (tf ). For this reason, the present project seeks
to demonstrate that evidence derived from analyses of creep curve shape is relevant to
interpretation of the detailed manner in which specific material variables influence
strain accumulation and damage evolution during tensile creep of CFCMCs.
* Brian Wilshire and Howard Burt, Materials Research Centre, School of Engineering, University of
Wales Swansea, Singleton Park, Swansea SA2 8PP, UK
221
222 Brian Wilshire et al.
To illustrate the dependence of composite performance on the properties of

the fibres and matrices, creep data comparisons and creep curve shape analyses
have been undertaken for three SiC-matrix products reinforced with either silicon
carbide or alumina fibres. These materials were selected to assess the effects of
variations
(a) in matrix composition for composites having nominally identical fibre config-
urations and fibre/matrix interfaces and
(b) in fibre type for composites produced with essentially the same matrices, fibre/
matrix interfaces and fibre architectures.
The materials studied, termed standard SiCf =SiC,6 enhanced SiCf =SiC,3 and
Al2 O3f =SiC composites,10, 11 were all reinforced with 40vol% of 15mm diameter
fibres incorporated as bundles of 500 fibres interwoven to obtain 2D layers of
fabric. The plain woven layers or plys were then aligned and stacked to produce
multilayer preforms having balanced 0/908 architectures. Carbon interfaces
( 0:5mm thick) were deposited onto the fibres by decomposition of a hydrocarbon
gas, before the preforms were densified by chemical vapour infiltration to give SiC
matrices with porosities of $ 15%. The matrices contained large pores (referred to as
macropores) between the plies or at yarn intersections within the plies, with small
pores in the matrix regions within the fibre bundles (Fig. 1). Yet, while the three
composites have many common features, two major differences are relevant to the
present analyses.
(a) The standard SiCf =SiC and enhanced SiCf =SiC products were fabricated with
Nicalon2 NLM202 fibres (Nippon Carbon Co.). Hence, both composites had
nominally identical fibre types, volume fractions and architectures, as well as
similar carbon interfaces. However, the SiC matrix of the enhanced SiCf =SiC
Figure 1. Scanning electron micrograph showing (a) the porous matrix within fibre bundles and (b) the
fibre bundle configuration of the standard SiCf =SiC composite before creep testing.
Strain Accumulation and Damage Evolution 223
samples contained boron-based additives which form a sealant glass to limit

oxygen penetration into the material during creep exposure at high-temperat-
ures.
(b) The standard SiCf =SiC testpieces were reinforced with Nicalon2 NLM202
fibres, whereas the Al2 O3f =SiC material was made with alumina fibres contain-
ing 85% g-alumina and 15% amorphous silica (Sumitomo Chemicals). Even
so, both composites were produced with virtually identical matrices and fibre/
matrix interfaces, as well as with the same fibre volume fractions and fibre
architectures.
Tensile creep tests were carried out using specimens with gauge lengths ma-
chined such that the stress axis was parallel (08) to one of the 0/908 fibre directions.
With the two SiCf =SiC samples, the tests were conducted in air at 1573K.3,6 How-
ever, with the Al2 O3f =SiC product, the alumina fibres transform to mullite above
1400K, so tests were completed at a maximum temperature of 1373K in air and
vacuum.10,11
3. EXPERIMENTAL OBSERVATIONS
For clarity, the observed creep behaviour patterns are first analysed for the standard
SiCf =SiC and enhanced SiCf =SiC products, offering a foundation for subsequent
comparisons of data sets recorded for the SiCf =SiC type materials and the
Al2 O3f =SiC composite.
3.1. Data Comparisons for Standard SiCf =SiC and Enhanced

SiCf =SiC Composites
For a variety of SiCf =SiC type composites, it has been stated that the SiC matrices
are more creep resistant than the Nicalon2 fibres.2–4 This conclusion was reached on
the grounds that the creep rates reported for silicon carbide produced by chemical
vapour deposition appeared to be lower than the rates predicted by extrapolation of
results obtained for Nicalon2 fibres. However, it is a simple matter to demonstrate
that the fibres rather than the matrices control the creep strength of both the
standard SiCf =SiC and enhanced SiCf =SiC composites.
The dominant rôle of the fibres becomes evident from plots of the variations of
the minimum creep rate (_em ) with stress (s) for the standard SiCf =SiC,6 enhanced
SiCf =SiC,3 and Nicalon2 NLM202 fibres7 at 1573K (Fig. 2a). Clearly, stresses about
five times higher must be applied to the Nicalon2 NLM202 fibres to obtain creep
rates comparable with those reported for the Nicalon2-fibre-reinforced composites.
Both SiCf =SiC type materials contained 40vol% fibres, so the longitudinal fibres
parallel to the tensile stress axes occupied one fifth of testpiece cross-sectional areas.
Hence, the creep strength of the composites must be determined by the creep
resistance of their longitudinal fibres, with the brittle matrices contributing little to
the overall creep strengths. However, with the fibres controlling the creep strengths,
it then becomes necessary to explain why the standard SiCf =SiC material is less creep
resistant than the enhanced SiCf =SiC product (Fig. 2a), when both composites were
produced with identical fibre types, volume fractions and architectures.
Figure 2. Stress/minimum creep rate relationships for (a) SiCf =SiC type composites and Nicalon2
NLM202 fibres at 1573K and (b) the Al2 O3f =SiC material and alumina fibres at 1373K.
The relative creep strengths of the standard SiCf =SiC and enhanced SiCf =SiC
samples (Fig. 2a) can be rationalized in a straight-forward manner6 by reference to
the creep curve shapes recorded for these materials. Over the stress ranges covered at
1573K, both composites displayed continuously decaying ‘primary’ curves, during
which the creep rate (_e) decreased gradually with time (t), reaching a minimum rate
(_em ) when fracture occurred, ie little or no period of accelerating tertiary creep was
apparent before failure. Under uniaxial tension, the creep rupture life (tf ) can be
defined conveniently as the time taken for the accumulated creep strain (e) to become
equal the limiting creep ductility (ef ). Continuously decaying creep curves therefore
terminate at t ¼ tf and e_ ¼ e_ m when e ¼ ef (Fig. 3).
Except for the small amounts of boron-based additives in the enhanced matrix,
the standard SiCf =SiC and enhanced SiCf =SiC samples are very similar, with the
fibre types, volume fractions and architecture being virtually identical. Because the
longitudinal fibres control the rates of creep strain accumulation, the initial portions
of the creep strain/time curves are therefore indistinguishable. However, the point of
Figure 3. Creep strain/time curves for the standard SiCf =SiC and enhanced SiCf =SiC composites tested in
air at 1573K under a stress of 90 MPa.
curve termination differs markedly for the two materials under the same test condi-
tions, as evident from the results shown in Fig. 3.
With the standard SiCf =SiC material, the total creep strains to failure were low
over the entire stress range covered at 1573K (with ef ffi 0:002). Consequently, the
decaying creep curves terminate early, giving high e_ m and low tf values (Fig. 3). In
contrast, with the enhanced SiCf =SiC specimens, creep must continue for much
longer times when failure occurs at higher creep ductilities. As a result, the minimum
creep rates determined for the enhanced SiCf =SiC samples are over an order of
magnitude slower (Fig. 2a) and the times to fracture are almost two orders
of magnitude longer than the values observed for the standard SiCf =SiC product
under equivalent test conditions (Fig. 3). Even so, the damage processes governing
creep fracture must then be identified in order to account for the differences in the
creep ductilities recorded for the standard SiCf =SiC and enhanced SiCf =SiC com-
posites.
3.2. Creep Fracture of Standard SiCf =SiC and Enhanced SiCf =SiC Composites
For both SiCf =SiC type composites, the results presented in Fig. 4 demonstrate that
the rupture life is almost inversely proportional to the minimum creep rate, such that
:
em :tf ¼ constant (1)
showing that the rate of creep damage development and the time to eventual failure
are determined by the rate of creep strain accumulation ie fracture is strain con-
trolled. Thus, with the fibres governing the measured creep rates, creep of the
longitudinal fibres must also determine the fracture times (tf ).
With the standard SiCf =SiC product, creep of the longitudinal fibres was
accompanied by crack formation in the brittle matrix. Microcracks originated at
pores within the transverse and longitudinal fibre bundles (Fig. 5a), with some longer
cracks (Fig. 5b) forming parallel to the stress axis (presumably because of the
complex stress states introduced by straightening of the interwoven fibre bundles
during tensile creep). However, these types of crack rapidly became arrested, with the
Figure 4. The dependences of the creep lives on the minimum creep rates recorded for standard SiCf =SiC
and enhanced SiCf =SiC in air at 1573K and the Al2 O3f =SiC composite under vacuum at 1373K.
Figure 5. Micrographs showing matrix cracks developing (a) normal and (b) parallel to the tensile stress
axis (s) during creep of standard SiCf =SiC samples.
fractured testpieces showing a very low crack incidence. Instead, fracture was a
consequence of the development of a dominant crack nucleated at inter-bundle
macropores at the specimen surface (Fig. 6a). As this dominant crack grew normal
to the tensile stress, on penetrating into the longitudinal fibre bundles, the crack faces
became bridged by unbroken fibres. The creep strength of the longitudinal fibres
therefore governs both the overall creep rates and the crack growth rates, accounting
for the dependence of the rupture life on the rates of creep strain accumulation (Fig.
4). Unfortunately, as the surface-nucleated dominant crack grew, direct oxygen
penetration along the opening crack oxidized the vulnerable carbon interfaces,
promoting failure of the bridging fibres. These events produced fracture surfaces
with near-planer crack growth zones, characterized by oxidation-assisted in-plane
failure of the longitudinal fibres, together with regions where sudden failure occurred
by fibre pull out (Fig. 6b).
The zone characterized by fibre pull out was confined to only one region of the
fracture surface, with this zone surrounded by the region of near-planar crack
Figure 6. Scanning electron micrographs showing (a) crack nucleation and link up from macropores and
(b) fracture surface of standard SiCf =SiC samples tested in air at 1573K.
development with negligible pull out (Fig. 6b). Moreover, with the standard
SiCf =SiC composite, the fractional area (A) of the final testpiece cross-section
occupied by the near-planar crack increased from around 0.25 to about 0.75 as
the applied stressp decreased from 80 to 47 MPa at 1573K. Thus, ultimate failure
occurred when s A reached a critical value ( 40 MPa), ie when the dominant crack
grew to the size required to cause sudden fracture by fibre pull out.
With the standard SiCf =SiC composite, direct ingress of oxygen along the
dominant crack promotes rapid failure of the crack-bridging fibres. High crack
growth rates then lead to low ductility failure, resulting in poor creep and creep
fracture resistance (Fig. 3). Yet, while surface-nucleated macrocracks also form from
inter-bundle macropores in the enhanced SiCf =SiC samples, the boron-based matrix
additions react with oxygen to form a sealant glass restricting oxygen penetration.3
Although the initial rates of creep strain accumulation are the same for the standard
and enhanced SiCf =SiC testpieces (Fig. 3), limiting oxygen penetration reduces the
crack growth rates. Hence, fracture of the enhanced SiCf =SiC composite is post-
poned to higher creep strains (Fig. 3), giving creep rates which are lower (Fig. 2a) and
creep lives which are longer (Fig. 4) than the values recorded for the standard
SiCf =SiC product.
3.3. Data Comparisons for SiCf =SiC and Al2 O3f =SiC Composites
The alumina fibres reinforcing the Al2 O3f =SiC composite are markedly less creep
resistant than Nicalon2 NLM202 fibres.8 Even so, as shown in Fig. 2b, a near five
fold difference in creep strength is evident for the Al2 O3f =SiC samples and the
alumina fibres, with the longitudinal fibres occupying one fifth of the testpiece
cross-sectional areas. Although it has been claimed that the SiC matrix is more
creep resistant than the alumina fibres,10, 11 the results in Fig. 2b confirm that the
longitudinal fibres control the creep properties of the Al2 O3f =SiC composite. Fur-
thermore, the dependence of the creep life on the minimum creep rate (Fig. 4)
indicates that the longitudinal fibres also govern the rates of crack development
and the times to rupture of the Al2 O3f =SiC product.
As with the SiCf =SiC type materials, the rates of creep strain accumulation and
crack growth are therefore determined by the longitudinal alumina fibres in the
Al2 O3f =SiC samples. However, unlike the various forms of SiC fibre, alumina fibres
might not be expected to be prone to oxidation-assisted fibre failure. Yet, as matrix
cracking proceeds, it appears that oxygen penetration into the Al2 O3f =SiC testpieces
can destroy the carbon interfaces, accelerating fibre failure and reducing creep
ductility. This view is supported by the fact10, 11 that, at 100 MPa, the creep ductility
of the Al2 O3f =SiC material decreased from 0.10 to 0.01, with a corresponding
decrease in creep rupture life from 2 105 s to 5 102 s, when the test atmosphere
was changed from vacuum to air (Fig. 7).
These observations suggest that the processes of creep strain accumulation and
damage development are broadly similar for the SiCf =SiC and Al2 O3f =SiC compos-
ites, even though very different creep curve shapes are displayed (Figs. 3 and 7). Thus,
decaying curves are recorded for the standard and enhanced SiCf =SiC products (Fig.
3), with decaying curves having no discernible tertiary stages also found for Nicalon2
NLM202 and other types of SiC fibres,7–9 as well as for sintered silicon carbide.12
In contrast, with the Al2 O3f =SiC testpieces tested in air or vacuum, decaying
Figure 7. Creep strain/time curves recorded for the Al2 O3f =SiC composite tested in (a) vacuum and (b) air
under a stress of 100 MPa at 1373K.10,11
primary stages of relatively short duration were followed immediately by extended

tertiary stages (Fig. 7). Creep curves exhibiting definite tertiary stages have also been
reported for various alumina fibres9 and even for monolithic alumina.13 On this basis,
it appears that the reinforcing fibres determine not only the rates of creep strain
accumulation and damage evolution but also the creep curves shapes displayed by
CFCMCs.6
4. DISCUSSION
With the SiCf =SiC type composites now considered, the creep strain rates are
governed by the creep resistance of the longitudinal fibres. However, the creep
strengths and diameters of individual fibres vary, causing the scatter in the data
shown in Fig. 1a for Nicalon2 NLM202 fibres.7 Because the weakest fibre regions
within the longitudinal bundles deform most easily, the creep rate decays continu-
ously as the load is transferred to the stronger fibres. Although matrix cracks
nucleate and grow as the longitudinal fibres extend, microstructural studies have
shown that the crack incidence is very low. Because the brittle matrices are weak,
crack development therefore has little effect on the decaying creep rate (Fig. 3).
4.1. Creep Curve Shape Analysis for SiCf =SiC Type Composites
The shapes of continuously decaying creep curves (Fig. 3) can be analysed most easily
by describing the accumulation of creep strain (e) with time (t) as
e ¼ atx (2)
where a and x are constants, with x ffi 0:5 for the standard SiCf =SiC material at
1573K.6 With this form of curve, fracture occurs when t ¼ tf , e ¼ em and e ¼ ef .
Hence, differentiation of Eq. (2) gives (putting t ¼ tf )
:
em: tf ¼ Xef (3)
with x ¼ X. Eq. (3) then reduces to Eq. (1) when the creep ductility (ef ) does not vary
significantly with changing test conditions, as is the case with the standard SiCf =SiC
:
composite for which em: tf ffi 0:001 and ef ffi 0:002, so x ¼ X ¼ 0:5 over the full stress
range studied at 1573K.6 In contrast, with the Al2 O3f =SiC10, 11 and seemingly with
the enhanced SiCf =SiC,3 as well as with other CFCMCs such as SiC-fibre-reinforced
Al2 O3 -matrix samples,6 Eq. (3) rather than Eq. (1) applies when ef increases with
increasing test duration.6 Hence, tf increases by more than the decrease in e_ m , so that
:
em: tf increases when higher creep ductilities are recorded (Eq. 3).
4.2 Creep Damage Tolerance of SiCf =SiC Type Composites
While the matrices contribute little to the overall creep strengths of the SiCf =SiC type
products, the composition and properties of the matrices affect the creep property
values recorded for the composites by governing the rates of oxygen ingress. In turn,
with fibres and fibre/matrix interfaces prone to oxidation, the rates of oxidation-
assisted failure of the crack-bridging fibres influences the creep ductility and there-
:
fore the measured em and tf values (Fig. 3). For this reason, ‘self healing’ approaches
such as the use of enhanced SiC matrices reduce crack growth rates, improving creep
lives by postponing failure to higher creep strains. However, with both the standard
SiCf =SiC and enhanced SiCf =SiC materials, fracture eventually occurs when a dom-
inant macro-crack grows to a size sufficient to cause sudden failure by fibre pull out
(Fig. 6b). Particularly during long-term creep exposure under low stress conditions,
large cracks must then develop before failure takes place. Hence, the present
SiCf =SiC materials can be considered as crack tolerant, but this criterion may not
be sufficient for selection of these products for safety-critical aerospace applications.
Because the Nicalon2 NLM202 and other SiC fibres display continuously
decaying creep curves,7–9 similar curve shapes are exhibited by SiC-fibre-reinforced
composites (Fig. 3). In the absence of distinct tertiary stages, the primary creep strain
(ep ) equals the total creep strain to failure (ef ), so the tertiary strain (ef ep ) is zero.
Consequently, near-zero values are also obtained for the creep damage tolerance
parameter (l), given by14, 15
:
l ¼ (ef ep )=em: tf (4)
The creep damage tolerance value is important in practical situations when materials
must withstand local strain concentrations, say, in regions where a change in com-
ponent cross-section leads to stress concentrations. Values of l in the range 5 to 10
are then thought to ensure that the strain concentrations encountered during service
will not lead to premature failure.14 Hence, with the standard and enhanced SiCf =SiC
samples, l values near to zero may represent a severe design constraint.
4.3. Creep Damage Tolerance of the Al2 O3f =SiC Composite
With the Al2 O3f =SiC composite, creep of the longitudinal alumina fibres is also
accompanied by crack formation in the brittle SiC matrix. Moreover, for tests carried
out under vacuum, matrix cracking continues throughout the decaying primary and
accelerating tertiary stages, although fibre failure begins only when the alumina
fibres reach their limiting creep ductility under the test conditions imposed. Fracture
of the Al2 O3f =SiC samples therefore takes place by progressive fibre failure and
eventual fibre pull out, but only after extended tertiary stages have been recorded
(Fig. 7a).
By exhibiting clearly-defined tertiary stages, l values of 5 or more are found for
the Al2 O3f =SiC material, even when low ductility failure occurs during tests con-
ducted in air (Fig. 7b). Consequently, Al2 O3f =SiC type composites are both crack
tolerant and creep damage tolerant. Hence, the requirement exists for high-strength
high-stability fibres of weaveable diameter ( 15mm) displaying creep curves with
extended tertiary stages, allowing development of high-performance CFCMCs pos-
sessing the crack tolerance and creep damage tolerance needed for safety-critical
aerospace applications.
5. CONCLUSIONS
For three SiC-matrix composites reinforced with interwoven bundles of silicon

carbide or alumina fibres, the longitudinal fibres are shown to control the rates of
creep strain accumulation and crack development, with oxygen ingress through the
matrices affecting the strains and times to fracture. The fibres also govern the shapes
of the creep curves displayed by the composites which, in turn, determine whether the
creep damage tolerance levels required for safety-critical engineering applications can
be attained.
REFERENCES
1. A. G. Evans and C. Weber, Creep damage in SiC/SiC composites, Mat. Sci. Eng. A208, 1 6 (1996).
2. S. Zhu, M. Mizano, Y. Kagawa, J. Cao, Y. Nagano and H. Kaya, Creep and fatigue behaviour of SiC
fibre reinforced SiC composite at high temperatures, Mater. Sci. Eng. A225, 69 77 (1997).
3. S. Zhu, M. Mizano, Y. Nagano, J. Cao, Y. Kagawa and H. Kaya, Creep and fatigue behaviour in an
enhanced SiC/SiC composite at high temperature, J. Am. Ceram. Soc. 81, 2269 2277 (1998).
4. S. Zhu, M. Mizano, Y. Kagawa, J. Cao, Y. Nagano and H. Kaya, Creep and fatigue behaviour in Hi
Nacolon2 fibre reinforced silicon carbide composites at high temperatures, J. Am. Ceram. Soc. 82,
117 128 (1999).
5. G. Fantozzi, J. Chevalier, C. Olagnon and J. L. Chermant, Creep of Ceramic Matrix Composites, in
Comprehensive Composite Materials, edited by A. Kelly and C. Zureben, (Elsevier, London 2000)
115 162.
6. B. Wilshire, Creep property comparisons for ceramic fibre reinforced ceramic matrix composites, J.
Eur. Ceram. Soc. 22, 1329 1337 (2002).
7. G. Simon and A. R. Bunsell, Creep behaviour and structural characterization at high temperatures of
Nicalon SiC fibres, J. Mater. Sci. 19, 3658 3670 (1984).
8. A. R. Bunsell and M H. Berger, Fine diameter ceramic fibres, J. Eur. Ceram. Soc, 20, 2249 2260
(2000).
9. J. A. DiCarlo, H. M. Yun and J. B. Hurst, Fracture mechanisms for SiC fibres and SiC/SiC
composites under stress rupture conditions at high temperatures, in. Proc. Tenth Inter. Conf. on
Fracture, edited by K. Ravi Chandra, B. L. Karihaloo, T. Kishi, R.O. Ritchie and A. T. Yokabori Jr.
(Elsevier, London 2001) Publication No. ICF 100586 (C.D.Rom).
10. F. Lamouroux, M. Steen and J. L. Valles, Damage of a 2D Al2 O3 =SiC composite during uniaxial
creep, Comp. Sci. Tech. 56, 825 830 (1996).
11. J. N. Adami. Comportement en fluage uniaxial sous vide d’un composite 2D Al2 O3 =SiC, Thèse,
Ecole Polytechnique de Zürich, 1992.
12. B. Wilshire and H. Jiang, Deformation and failure processes during tensile creep of sintered silicon
carbide, Br. Ceram. Trans. 93, 213 218 (1994).
13. C. K. L Davies and S. L. Sinha Ray, High temperature creep deformation of polycrystalline alumina
in tension, in Special Ceramics Vol. 5, edited by P. Popper (Br. Ceram. Res. Assoc., Stoke on Trent,
1972) 193 207.
14. I. N. Goodall, R. D. H. Cockcroft and E. J. Chubb, An approximate description of the creep rupture
of structures, Int. J. Mech. Sci. 17, 351 359 (1975).
15. M. F. Ashby and B. F. Dyson, Creep damage mechanisms and mechanics, in Advances in Fracture
Research, edited by S.R. Valluri et al (Pergamon Press. Oxford, 1984), 3 30.
STRENGTH AND
RELIABILITY OF LEAD
ZIRCONATE TITANATE
CERAMICS
Chad S. Watsonz
ABSTRACT
The influences of temperature and processing conditions (unpoled or poled-depoled)

on strength, Weibull parameters and the stress-strain behavior of tin-modified lead
zirconate titanate (PSZT) were evaluated in four-point bending. PSZT exhibited
temperature-dependent non-linear and non-symmetric stress-strain behavior. A re-
sult of temperature dependent non-linearity was an apparent reduction in the flexural
strength of PSZT as temperature increases. At room temperature the average stress in
the outer-fiber of bend bars was 84 MPa, whereas, for specimens tested at 1208C the
average failure stress was only 64 MPa. The load-carrying capacity, however, did not
change with temperature, but the degree of deformation tolerated by PSZT prior to
failure increased with temperature.
1. INTRODUCTION
Partially electroded ferroelectric devices that are heated to enhance dipole alignment
under the influence of an electric field develop differential strains at the electrode
edge. Strain incompatibilities at the electrode boundary can result in stresses large
enough to initiate cracks from nearby flaws.1–3 Similarly, multilayer actuators, which
z
Chad S. Watson, Ceramic Materials Department, Sandia National Laboratories, Albuquerque, NM
87185
485
486 Chad S. Watson
are composed of several layers of partial internal electrodes, can develop stresses
during switching operations that are high enough to generate cracks at electrode
boundaries, thus limiting the effective lifetime of the device.4 The reliability of
partially electroded ferroelectrics can be improved by using thinner layers, which
effectively reduces the volume of material exposed to high tensile stresses around the
electrode edge during poling.1,2 Another reliability enhancement technique includes
the use of a pre-conditioning step prior to polarizing the electroded volume. By
polarizing and then depolarizing the entire ferroelectric element, the domains in the
material, instead of being randomly oriented, are now aligned ‘‘parallel’’ to the
applied field. As a result, the strain differential between the electroded and unelec-
troded area can be reduced; for example, from 800 me to only 200 me in the e31
direction for tin-modified lead zirconate titanate.3
Ferroelectric devices are frequently exposed to elevated temperatures during
processing or while in operation, which could potentially degrade component integ-
rity. For instance, the mechanical properties of ferroelectrics can vary as temperature
is increased due to phase transformations, domain wall movement, residual stress
changes, microcracking or other mechanisms. Research relating the effect of tem-
perature on the mechanical properties of ferroelectrics indicates a reduction in
strength5,6 and fracture toughness7 as the applied temperature approaches the
Curie temperature. The potential for temperature-dependent mechanical degradation
of ferroelectrics needs to be addressed to establish a more complete understanding of
component reliability. To this end, the effects of temperature on the stress-strain
behavior and flexural strength of tin-modified lead zirconate titanate (PSZT) have
been examined. In addition, Weibull strength parameters were determined for
unpoled and poled-depoled specimens at room temperature.
Sintered PSZT [Pb0:99 (Sn0:13 Zr0:82 Ti0:05 )0:98 Nb0:02 O3 ] ceramics were prepared using
conventional mixed oxide techniques (Alliant Corporation, New Hope, MN). Bend
bars were sliced and ground from sintered billets and finished with a 600 grit diamond
wheel to dimensions of 3 mm 4 mm 45 mm. Analogous to the procedures used to
fabricate PSZT partially electroded components, all of the bend bars were thermally
cleaned at 8258C. At this point, the bend bars were in the unpoled state, i.e., the domains
in the material were assumed to be randomly oriented. Some of these unpoled bend bars
were saved after thermal cleaning for mechanical property measurements. The remain-
ing bend bars were processed through the pole-depole state. A temporary silver elec-
trode was applied to the top and bottom surfaces of the bend bars and dried at 718C.
After the temporary electrode was dried and the bend bar cooled to room temperature, it
was subjected to a pole-depole process. For this process, the bend bars were polarized in
dielectric fluid (Fluorinert FC77, 3M Corp., Minneapolis, MN) at room temperature to
32 kV/cm; whereupon, the electric field was reversed until negligible charge was stored.
After pole-depole the temporary electrode was removed by rinsing in acetone.
A schematic illustrating the experimental configuration of the bend bars with respect
to the poling-depoling direction and loading conditions is shown in Fig. 1.
The stress-strain behavior and flexure strength of PSZT were measured in
four-point bending at 23,75,86,100 and 1208C. For the 75, 86 and 1008C runs, the
Figure 1. Schematic of bend bar geometry, polarization direction and strain gage placement.
samples were taken to 1058C and held at temperature for 15 min and then cooled to
their respective test temperature at 38C/min and held at temperature for 15 min prior
to testing to minimize thermal gradients in the material. These heating and cooling
cycles (38C/min ramp rates and 15 min holds) are analogous to those used for hot
poling PSZT components. An MTS 810 Universal Testing Machine and a MTS
Environmental Chamber (MTS Systems Corporation, Minneapolis, MN) were
used to mechanically load the specimens at temperature. Strength was measured
according to ASTM C1161-02c.8 Weibull parameters for unpoled and poled-depoled
PSZT were determined from room temperature strength measurements.
To evaluate the stress-strain behavior of PSZT, load-strain and load-deflection
measurements were performed at each temperature condition. Tensile and compres-
sive strains that developed while loading to failure were determined for two speci-
mens at each temperature with the use of strain gauges with pre-attached leads (Type
CEA-06-062UW-350, Measurements Group, Inc.; Raleigh, NC). The gauges were
affixed to the compression and tension face of the bend bars with adhesive (M-Bond
600; Measurements Group, Inc.; Raleigh, NC) and cured at 1008C for four hours.
Deflection measurements were made on eight specimens at each temperature utilizing
an extensometer (MTS Systems Corporation, Minneapolis, MN). The deflection
data was converted to strain («),
6Wh
«¼ (1)
L2
where W is the magnitude of bar deflection, h is bar thickness and L is the distance
between the outer supports.
Representative SEM micrographs of a polished and chemically etched PSZT
specimen and of a typical fracture surface are shown in Figs. 2a & b, respectively.
Figure 2. SEM images of (a) polished and chemically etched PSZT and (b) of a typical fracture surface.
488 Chad S. Watson
PSZT microstructures consist of equiaxed grains with a bimodal grain size distribu-
tion. Some intragranular porosity within the larger grains is evident. The grain
pullout is an artifact of polishing. PSZT fractures in both an intergranular and
transgranular mode. Typically, the larger grains fail transgranularly while smaller
grains fail intergranularly.
3.1. Non-Linear Stress-Strain Behavior and True Flexural Strength
Based on the load-strain and load-deflection measurements, PSZT exhibits non-linear

stress-strain behavior. A plot of linear-elastically computed stress (or engineering
stress) versus strain for poled-depoled specimens tested at room temperature, 75, 86,
105 and 1208C is shown in Fig. 3. Deviations from linear-elastic behavior initiate at a
nominal stress level of approximately 20 30 MPa for specimens tested at room tem-
perature and 10 20 MPa for specimens tested at an elevated temperature. Further-
more, the extent of non-linearity increases as the testing temperature increases.
Conversion of the load-strain data to the true stress-strain behavior was achieved by
implementing the approach first described by Nadai9 and adapted by Chen et al.10 The
true compressive (sc ) and tensile stresses (st ) were calculated as follows:
2M(«t «c ) @M 1
sc ¼ (2)
«c bh2 @r «c bh
2M(«t «c ) @M 1
st ¼ (3)
«t bh2 @r «t bh
where M is the applied bending moment, «c and «t are the compressive and tensile
strains respectively, b is the width of the bend bar, and r is radius of curvature given by
h
r¼ (4)
«t «c
Figure 3. Tensile stress strain curves to failure for PSZT four point bend specimens at 23, 75, 86, 100 and
1208C assuming linear elasticity.
Figure 4. Conversion of load strain data to the true tensile stress strain curves to failure for PSZT four
point bend specimens at 23, 75, 86, 100 and 1208C.
The true stress versus strain curves are shown in Fig. 4. These curves are derived from
the same load-strain data as the engineering stress-strain curves in Fig. 3. It is evident
that as the temperature increases, the amount of deformation supported by the
material to failure also increases. For example, the specimen tested at room tem-
perature failed at a strain of 800 me, whereas, the specimen tested at 1208C
deformed twice as much at failure ( 1600 me). This additional deformation is
most likely the result of a higher percentage of domains switching in response to
the applied load. The reproducibility of the stress-strain measurements is shown in
Fig. 5 where the true tensile stress-strain traces of two strain-gauged specimens at
each temperature follow similar curves. Assuming the remaining eight bend bars that
were tested to failure without a strain gauge at each temperature exhibit a similar
stress-strain response, then the true failure stress for each specimen can be deter-
mined. Because strain-gauge measurements can be cumbersome, strain development
can also be assessed from deflection measurements using an extensometer. However,
extensometers do not provide the same sensitivity as strain-gauges. In addition,
contributions from non-symmetry are not measured with an extensometer. Nonethe-
less, the true strength can be ascertained, but non-symmetry contributions are
assumed negligible and strain resolution is compromised. The true failure stress of
specimens loaded to failure without strain-gauges was determined by converting
load-deflection measurements made with an extensometer to the true failure stress
following Gogotsi,11

3a « dP
st ¼ 2 P þ (5)
bh 2 d«
where a is the distance between the inner supports, P is the load, and « is the tensile
strain (calculated using Eq. 1) experienced by the material during flexure. A small
portion of the load-deflection data (three out of eight poled-depoled specimens at
238C and one out of eight poled-depoled specimens at 1208C) was taken from
measurements made with the LVDT of the mechanical testing machine. In addition,
the strains measured with the two strain-gauge experiments at each temperature were
also included in the extensometer data set, as a result, a total of ten specimens were
490 Chad S. Watson
Figure 5. True tensile stress strain curves to failure of two different strain gauged PSZT four point bend
specimens at (a) 23, (b) 75, (c) 86, (d) 100 and (e) 1208C.
averaged to determine the true failure stress using Eq. 5. The results are given in Table 1
and shown in Fig. 6. In addition, the average load at failure, linear elastic or engin-
eering stress, and the average true failure stress based on the stress-strain response of
the two strain-gauged specimens, 10 specimens per temperature, is shown in Table 1
and Fig. 6. The true failure stresses, as calculated from the strain-gauge and displace-
ment techniques, are typically within 3%. As shown in Fig. 6, the average true stresses
for both measurement techniques follow the same trend: the true failure stress de-
creases with increasing temperature. At room temperature, the true failure stress as
calculated using the strain gauge technique is 84.3 MPa and decreases significantly to
Table 1. Mechanical Property Measurements with 95% Confidence Intervals.
Failure Engineering True Failure Stress True Failure

Load Failure (MPa) Strain Stress (MPa)
Temperature (N) Stress (MPa) Gauges Extensometer
238C* 103:9 6:3 86:0 5:3 83:9 6:7 82:2 4:6

238C 106:2 8:3 88:0 6:9 84:3 6:1 82:3 5:1
758C 110:9 5:7 91:7 4:5 82:1 3:1 82:0 3:5
868C 105:7 5:6 87:7 4:6 73:9 2:3 76:8 2:9
1008C 108:0 3:6 89:5 2:8 71:6 0:7 72:7 1:8
1208C 100:7 4:3 83:8 3:5 63:7 0:7 65:4 2:1
*
Unpoled Specimen
Figure 6. Average engineering and true (strain gauge and displacement technique) failure stress with 95%
confidence intervals from room temperature to 1208C.
63.7 MPa by 1208C. However, if we were to assume that the material behaved linearly
elastically, the true failure stress would have been overestimated by 4% at room
temperature and 24% at 1208C. The increase in the difference between the engin-
eering stress and the true failure stress as temperature increases is because non-
linearity increases with temperature. Although the true failure stress decreases with
temperature, the load-carrying capacity of PSZT does not differ significantly with 95%
confidence. The average load at failure is between 100.7 and 110.9 N over the entire
temperature range. Thus, the material is able to support similar loads prior to failure
while also being able to support a higher degree of deformation as temperature
increases. Intuitively, this suggests that at elevated temperatures PSZT is able to
alleviate mechanically induced stresses by deforming.
At room temperature, soft PZT-based materials can exhibit non-linear and non-
symmetric behavior.12 To fully complement strength data, it is recommended that
when flexural strength measurements of ferroelectrics are reported that the stress-
strain behavior of the material also be included. Otherwise, without full disclosure of
stress-strain behavior, the predicted strength of the material, as that calculated from
linear-elastic fracture mechanics can lead to a systematic over-estimation of strength.
Indeed, the most appropriate method for reporting strength data without strain-
gauge or deflection measurements is load at failure. With this disclosure, at least the
load-carrying capacity of the material is identified.
492 Chad S. Watson
3.2. Stress-Strain Behavior of Unpoled and Poled-Depoled PSZT
A distinct difference in the stress-strain response of unpoled and poled-depoled PSZT,

as measured with strain gauges, is shown in Fig. 7. Poled-depoled PSZT exhibits
additional non-linearity. This trend would most likely continue until the Curie tem-
perature is reached. Presumably, the difference in the non-linearity is the result of the
formation of 1808 domains parallel to the applied electric field (or as close as crystal-
lographically possible) during the polarization and depolarization step. Because do-
mains attempt to align parallel to the direction of maximum principal stresses in
tension, which is perpendicular to the polarization-depolarization direction, a higher
percentage of domains in the poled-depoled specimen are available for switching. For
the unpoled specimen, domains are randomly oriented, thus certain percentages are
already parallel or nearly parallel to the direction of maximum principal stresses;
therefore, fewer domains are available for switching during loading.
3.3. Non-Symmetrical Stress-Strain Behavior
PSZT, as with similar soft ferroelastics, exhibits non-symmetric stress-strain behav-

ior.12 Non-symmetry was negligible at room temperature for PSZT but at elevated
temperatures non-symmetry was evident (Fig. 8a-e). However, because of strain-
gauge misalignment, poor strain-gauge adhesion, or statistical variations in data,
not all of the elevated temperature stress-strain tests revealed significant non-symmet-
rical behavior.13 Non-symmetry refers to deviations in the stress-strain response in
tension as compared to compression. In response to tension, domains align parallel to
the principal stress direction, and in compression domains align perpendicular to the
principal stress direction. As a result, at the same loading level, above which non-
linearity has developed, the resulting strain in a specimen poled perpendicularly to the
maximum principal direction (or for an unpoled specimen) is greater in tension than in
compression. As a consequence the outer-fiber stress of a perpendicularly poled or
unpoled specimen tested in flexure is greater in compression than in tension. Compar-
isons of non-symmetry obtained through flexure testing to measurements made from
uniaxial compression and tension tests follow the same trends.12 But by applying strain
gauges on the compression and tension faces of bend bars, non-symmetry can be
Figure 7. True tensile stress strain behavior (strain gauge technique) of unpoled and poled depoled PSZT
at room temperature.
Figure 8. Non symmetry is evident by comparing the tension and the compressive stress strain curves to
failure for PSZT four point bend specimens at (a) 23, (b) 75, (c) 86, (d) 100 and (e) 1208C.
ascertained.14 In doing so, the intrinsic variability of material properties can be

avoided because both the compression and tension response can be determined for
the same test specimen instead of two independent test specimens needed for com-
pression and tension tests.
3.4. Weibull Statistics
Ceramic components are susceptible to failure at stresses much lower than predicted
from theoretical strength calculations because of stress concentrations at surface or
494 Chad S. Watson
volume flaws, where the location and size of these strength-limiting flaws are ran-
domly distributed. Flaw types and their distributions are processing and handling
history dependent, and can vary significantly for ceramic components processed
from nominally the same source. Weibull statistics, which are based on weakest
link theory, can be employed to estimate the probability of failure for a particular
material at a given stress level. An indirect measurement of component reliability can
be obtained from the Weibull modulus: a descriptor relating the strength distribution
(thus failure-initiating flaw distribution) within the material. A low Weibull modulus
is indicative of a material that exhibits a wide range of failure strengths, and thus has
a higher probability of failure at a lower stress level than a material exhibiting a
higher Weibull modulus. Knowledge of PSZT Weibull parameters is needed to have
a more intimate understanding of the effect processing and fabrication conditions
have on flaw distributions and ultimately component reliability. As such, Weibull
strength parameters15 of unpoled and poled-depoled PSZT were determined.
A total of 44 specimens was tested, per processing condition (unpoled or poled-
depoled), to failure in 4-point bending. The true failure stress was determined based
on the stress-strain results of unpoled and poled-depoled specimens. The strength
distribution was characterized using a two-parameter Weibull distribution and max-
imum likelihood estimators for Weibull modulus (m) and characteristic strength (s0 )
m
s
Pf ¼ 1 exp (6)
s0
where Pf is the probability of fracture (Fig. 9). Fractographic examinations of a

portion of the fracture surfaces indicated that both unpoled and poled-depoled PSZT
contained multiple flaw distributions; however, Weibull strength parameters were
not estimated with the use of censored data analysis techniques.15 Consequently,
significant statistical bias in the estimated Weibull strength parameters is possible.
The Weibull modulus for the specimens tested in the unpoled condition was 17 with a
characteristic strength of 90.9 MPa. The strengths ranged from 70.9 to 99.8 MPa
with an average strength of 88:2 6:2 MPa (one standard deviation). The poled-
depoled specimens exhibited a Weibull modulus of 13.2, a characteristic strength of
92.1 MPa, and an average strength of 88:4 8:8 MPa (one standard deviation)
where the strength values ranged from 67.0 to 102.7 MPa. The apparently lower
Figure 9. Weibull plots of strength of unpoled and poled depoled PSZT. Measurements were made at
room temperature on 44 specimens per processing condition.
Weibull modulus of poled-depoled PSZT is indicative of a material containing a

wider flaw distribution compared to unpoled specimens, thus, the pole-depole pro-
cess could potentially reduce component reliability. If there is a true difference in
Weibull moduli it may be attributed to the additional handling required for the pole-
depole step. This includes (i) the application of temporary silver electrodes to the
top and bottom surfaces of the bend bars, (ii) drying of these electrodes at 718C,
(iii) polarization and depolarization in dielectric fluid, (iv) removal of the temporary
electrodes with acetone and (v) transportation of specimens between each step. Also,
the polarization and then depolarization step could potentially degrade the mechan-
ical properties16–18 of some specimens as a result of the high electric fields (32 kV/cm)
required for this procedure, by creating new flaws or by making intrinsic failure-
initiating flaws more severe. Despite the additional handling and processing steps
experienced by poled-depoled specimens, the Weibull moduli are remarkably similar.
Fractography revealed that the lowest strength specimens, unpoled or poled-depoled,
typically failed from similar flaws (Fig. 9). Perhaps a higher number of these gross
flaws happened to be sampled for the poled-depoled specimens. Accordingly, statis-
tical sampling may have a significant effect on the calculated Weibull moduli and
care should be taken when interpreting the results. From this analysis, the poled-
depoled specimens have a slightly lower Weibull modulus than unpoled PSZT.
The calculated difference in the Weibull moduli may be attributed to the statistical
nature of sampling or the additional processing steps experienced by poled-depoled
PSZT.
3.5. Fractography
Fracture surfaces were examined to determine failure origins and the type of flaws that
initiated failure at low stress levels for both unpoled and poled-depoled specimens
(Fig. 9). Examination of fracture surfaces revealed that the majority of failures
originated from large grain clusters (Fig. 10a), surface or near-surface pores
(Fig. 10b) or large grain clusters and surface or near-surface pores (Fig. 10c). The
weakest specimens, however, failed from very large surface or near-surface pores (Fig.
10d), agglomerates (Fig. 10e) or from large pre-existing cracks located near chamfers.
The reliability of ceramic materials is highly dependent on the presence of
handling or machining damage and/or processing flaws. Several of the low strength
specimens failed from pre-existing cracks at the chamfer that may have formed as the
result of handling or machining damage. Care must be taken to ensure that the
handling and machining of PSZT specimens is innocuous, thus reducing the prob-
ability of a large pre-existing crack inducing failure, and increasing the reliability of
the component. The reliability of ceramic materials, however, will always be subject
to processing and fabrication history as well as handling damage. For instance, the
microstructure of ceramics, which is processing history dependent, can be the
strength-limiting factor. Microstructural subtleties, such as, abnormally large grains,
porosity, agglomerates and/or the presence of second phases can have a significant
influence on failure probabilities. Both Figs. 10d & 10e show microstructural features
(failure origins) representative of low strength samples. Closer examination of
Fig. 10d reveals that, besides the large pore, there is a difference in grain size around
the pore compared to that found in the bulk of the material (Fig. 11a). This micro-
structural feature can certainly be a contributor to the low failure stress. In addition,
496 Chad S. Watson
Figure 10. SEM images of fracture surfaces showing (a) clusters of large grains in combination with
possible machining damage as the failure origin (fractured at 91.5 MPa), (b) a near surface pore in
combination with possible machining damage (fractured at 86.1 MPa), (c) a failure origin consisting of a
combination of large grains and a surface pore (92.5 MPa), (d) a fracture surface where a large near
surface pore initiated failure at a stress of 72.1 MPa, and (e) an agglomerate where failure initiated at a
stress level of 69.2 MPa.
the failure-initiating flaw shown in Fig. 10e appears to be a large agglomerate. At

higher magnification, there is noticeable separation between the agglomerate and the
bulk (Fig. 11b). The cracks associated with the agglomerate led to a very low failure
stress (69.2 MPa). Both of these flaws are processing dependent and can be eliminated
by modifying processing parameters. Thus, with the use of Weibull statistics the flaw
size distribution can be determined and in combination with fractography, failure-
initiating flaws can be identified and processing modifications can be implemented to
Figure 11. SEM images of (a) a fracture surface showing a distinct boundary between small and large
grains where the small grains surround a large pore (Fig. 10d) and (b) a large processing flaw (agglomerate)
that was responsible for failure (Fig. 10e).
reduce or eliminate these severe flaws. In addition, the reliability of PSZT compon-
ents can be estimated using computer programs, such as CARES/LIFE19 or FAIL-
PROB,20 which incorporate Weibull strength statistics into the predicted stress
distribution of the component.
4. CONCLUSIONS
Failure stress and the stress-strain behavior of PSZT were determined as a function
of temperature (23,75,86,100 and 1208C). The strength of PSZT decreased with
temperature. Strength measurements were carried out with the use of strain gauges
and an extensometer in order to evaluate the true stress-strain behavior. PSZT
underwent non-linear and non-symmetric stress-strain behavior, which was more
predominant at elevated temperatures. A comparison of the strain gauge and extens-
ometer techniques indicates only a slight difference (3%) in the calculated true
failure stress. Both measurement techniques indicate a significant reduction in the
outer-fiber stress at failure as temperature increased. The implications of this are that
if the assumption of linear-elasticity was used a systematic overestimation in the
failure stress would result. The load-carrying capacity of PSZT did not change as a
function of temperature, but the extent of deformation did increase with tempera-
ture. The Weibull parameters of PSZT indicated that poled-depoled specimens have
a slightly higher probability of failure at a particular stress level (thus lower reliabil-
ity) than unpoled PSZT. Thorough examinations of the fracture surfaces suggested
that the lower Weibull modulus of poled-depoled PSZT may be attributed to sam-
pling or differences resulting from additional processing steps.
5. ACKNOWLEDGMENTS
Technical contributions by Michael Hutchinson, George Burns, Bonnie Mckenzie,

John Laing, Roger Moore, Pin Yang and Jill Glass are gratefully acknowledged.
Sandia is a multiprogram laboratory operated by Sandia Corporation, a Lockheed
498 Chad S. Watson
Martin Company, for the United States Department of Energy’s National Nuclear
Security Administration under Contract DE-ACO4-94-AL85000.
REFERENCES
1. S. L. dos Santos e Lucato, D. C. Lupascu, M. Kamlah, J. Rödel, and C. S. Lynch, ‘‘Constraint
Induced Crack Initiation at Electrode Edges in Piezoelectric Ceramics,’’ Acta Mater., 49[14]
2751 2759 (2001).
2. S. L. dos Santos e Lucato, D. C. Lupascu and J. Rödel, ‘‘Crack Initiation and Crack Propagation in
Partially Electroded PZT,’’ J. Eur. Ceram. Soc., 21 1425 1428 (2001).
3. R. H. Moore, G. R. Burns, S. N. Burchett, S. J. Glass, M. A. Hutchinson, T. W. Scofield, M. E.
Stavig, C. S. Watson, and P. Yang, ‘‘Polarization Induced Cracking in Partially Electroded PSZT
Ceramic,’’ Ceram. Trans., 136 105 116 (2003).
4. A. Furuta and K. Uchino, ‘‘Dynamic Observation of Crack Propagation in Piezoelectric Multilayer
Actuators,’’ J. Am. Ceram. Soc., 76 [6] 1615 1617 (1993).
5. R. F. Cook, S. W. Freiman, and B. R. Lawn, ‘‘Fracture of Ferroelectric Ceramics,’’ Ferroelectrics, 50
267 272 (1983).
6. R. Fu and T. Y. Zhang, ‘‘Influences of Temperature and Electric Field on the Bending Strength of
Lead Zirconate Titanate Ceramics,’’ Acta Mater., 48 [8] 1729 1740 (2000).
7. K. Mehta and A. V. Virkar, ‘‘Fracture Mechanisms in Ferroelectric Ferroelastic Lead Zirconate
Titanate [Zr Ti 0:54 0:46] Ceramics,’’ J. Am. Ceram. Soc., 73 [3] 567 574 (1990).
8. ‘‘Standard Test Method for Flexural Strength of Advanced Ceramics at Ambient Temperatures,’’
ASTM Designation C1161 02c, American Society for Testing and Standards, West Conshohocken,
Pennsylvania.
9. A. Nadai, Theory of Flow and Fracture of Solids, Vol. 1; Ch. 22. McGraw Hill, New York, 1950.
10. W. Chen, D. C. Lupascu, J. Rödel, and C. S. Lynch, ‘‘Short Crack R curves in Ferroelectric and
Electrostrictive PLZT,’’ J. Am. Ceram. Soc., 84 [3] 593 597 (2001).
11. G. A. Gogotsi, ‘‘The Problem of the Classification of Low Deformation Materials Based on the
Features of Their Behavior Under Load,’’ Strength Mater. (Engl. Transl.), 9 [1] 77 83 (1977).
12. T. Fett, S. Muller, D. Munz, and G. Thun, ‘‘Nonsymmetry in the Deformation Behaviour of PZT,’’
J. Mater. Sci. Lett., 17 [4] 261 265 (1998).
13. C. S. Watson, ‘‘Mechanical Behavior, Properties and Reliability of Tin Modified Lead Zirconate
Titanate,’’ Sandia National Laboratories Report, SAND2003 2422 (2003).
14. T. Fett, D. Munz, and G. Thun, ‘‘Nonsymmetric Deformation Behavior of Lead Zirconate Titanate
Determined in Bending Tests,’’ J. Am. Ceram. Soc., 81 [1] 269 272 (1998).
15. ‘‘Standard Practice for Reporting Uniaxial Strength Data and Estimating Weibull Distribution
Parameters for Advanced Ceramics,’’ ASTM Designation C1239 00, American Society for Testing
and Standards, West Conshohocken, Pennsylvania.
16. A. G. Tobin and Y. E. Pak, ‘‘Effect of Electric Fields on Fracture Behavior of PZT Ceramics,’’ Proc.
SPIE Int. Soc. Opt. Eng., 1916 78 86 (1993).
17. K. D. McHenry and B. G. Koepke, ‘‘Electric Field Effects on Subcritical Crack Growth in PZT’’; pp.
337 352 in Fracture Mechanics of Ceramics, Vol. 8. Edited by F. F. Lange. Plenum Press, New York,
1983.
18. S. B. Park and C. T. Sun, ‘‘Fracture Criteria for Piezoelectric Ceramics,’’ J. Am. Ceram. Soc., 78 [6]
1475 1480 (1995).
19. N. N. Nemeth, L. M. Powers, L. A. Janosik, and J. P. Gyekenyesi, ‘‘CARES/Life Ceramics Analysis
and Reliability Evaluation of Structures,’’ Users and Programmers Manual, NASA TM 106316.
20. G. W. Wellman, ‘‘FAILPROB A Computer Program to Compute the Probability of Failure of a
Brittle Component,’’ Sandia National Laboratories Report, SAND2002 0409 (2002).
TAILORING THE
COMPOSITION OF
SELF-REINFORCED SILICON
NITRIDE CERAMICS TO
ENHANCE MECHANICAL
BEHAVIOR
P. F. Becher, G. S. Painter, H. T. Lin, and M. K. Ferber*
1. ABSTRACT
Studies have shown that seeding allows for greater control of the microstructure of
self-reinforced beta-silicon nitride based ceramics, which when combined with tailor-
ing of the sintering additives can results in significant improvements in fracture
toughness and strength. Similar behavior was noted in alpha-SiAlONs. In beta-
Si3 N4 ceramics in which alumina serves as one of the additives, the larger elongated
reinforcing grains result from the epitaxial deposition of Si6-z Alz Oz N8-z layers on the
Si3 N4 core. Structural models show that the bond strength of the interface between
the SiA1ON and the amorphous intergranular film increases as the A1 and O
contents of the SiA1ON increase. This is consistent with experiments that revealed
interfacial debonding was promoted when the z-value of the SiA1ON epitaxial layer
was reduced. Compositional tailoring that leads to enhanced interfacial debonding
and an increase in the fracture toughness.
For a number of applications, creep resistance is also required in these ceramics.
Recent results have been obtained in self-reinforced silicon nitride ceramics where
*
P. F. Becher, G. S. Painter, H. T. Lin, and M. K. Ferber, Metals and Ceramics Division, Oak Ridge
National Laboratory, P. O. Box 2008, Oak Ridge, Tennessee 37831.
327
328 P. F. Becher et al.
controlled increases in the fraction of larger elongated reinforcing grains were found
to reduce the creep rates. However, viscous flow in the amorphous intergranular
films (IGF) in silicon nitride-based ceramics significantly impact their creep behavior
and mechanical reliability at elevated temperatures. Recent studies of bulk Si-based
oxynitride glasses have shown that increasing the nitrogen to oxygen (N:O), rare
earth (RE): aluminum (A1) ratios of the glass and using smaller RE ions raise the
glass viscosities. Creep models suggest that this will improve the creep resistance of
the silicon nitride ceramics, which is supported by the results of creep studies. The
above findings are providing a framework for designing the next generation of silicon
nitride-based ceramics by microstructural and compositional tailoring.
2. BACKGROUND
As a class of materials, ceramics present an unusually complex set of both intellectual

and technological challenges. One of the key issues has been the need to make cera-
mics tougher by devising approaches to increase the energy required to grow cracks.
This would be reflected in increased fracture toughness where traditional ceramics
exhibit values that are only 2 to 10% of those of metallic alloys. Even so, ceramics,
such as silicon nitride, continue to find important and demanding applications
including high performance bearings, turbocharged rotors, fuel igniters and machine
tool inserts. The most recent important milestones achieved include the incorpor-
ation of silicon nitride ceramic as (1) hot section components for auxiliary power
units produced by AlliedSignal for aircraft and industrial auxiliary turbo-power units
and various components for aircraft turbine engines1 and (2) valves in selected auto-
motive diesel engines by Daimler-Benz.2
These are quite remarkable considering the brittle nature of ceramics and the
dearth of research on high tech ceramics until 30 years ago. Two major advances
have given rise to these subsequent successes. These include: (1) substantial improve-
ments in the technology to reliably fabricate ceramic components including those
with complex shapes, and (2) improvements in the mechanical properties due to the
evolution of a class of tougher silicon nitride ceramics that are more resistant to
damage and attendent strength losses. This evolution must continue for future
applications and requires a scientific basis for tailoring a wider range of tough,
high strength ceramics for both room and elevated temperature applications.a The
goal must be to improve the reliability of silicon nitride components whether limited
by brittle fracture, crack growth at intermediate and low temperatures, or creep- or
oxidation/corrosion-induced damage that reduce the load bearing capacity or the
lifetime of components. Depending on the application, all of those factors controlling
the mechanical reliability may have to be considered in the tailoring of the compos-
ition and microstructure of a silicon nitride ceramic. Here we consider some of those
factors that can influence the fracture and creep responses.
a
Note that the issue of the brittleness of ceramics restricting their application is not limited to structural
applications. For instance, the brittleness of superconducting ceramics had to be resolved by employing
them either as thin films on ductile metal wires or embedded within ductile metal casings. Thus the
understanding of how to defeat the brittleness of the ceramics impacts their use in a host of applications.
Tailoring the Composition of Self-Reinforced Silicon Nitride Ceramics 329
3. TAILORING THE FRACTURE TOUGHNESS AND STRENGTH
For high strength-high toughness response, one now has a basis for tailoring the
microstructure of silicon nitride-based ceramics. In the case of beta-silicon nitride
(and b-SiA1ON) ceramics, it has been established that this can be accomplished
through the controlled introduction of larger elongated grains combined with the
tailoring of the sintering additive composition.3,4 The larger elongated grains that
form serve as the reinforcing elements much like whiskers in whisker-reinforced
ceramics4,5 or fibers in fiber-reinforced ceramics.6 Results using these principles
show that self-reinforced silicon nitride (SRSN) ceramics can have high strengths
(1 to 1.4 GPa) combined with high fracture toughness (e.g., up to 11 MPam1=2 or at
least two to three times that of commericial materials), Figure 1. Higher fracture
toughness values are reported in the literature, but these often have been obtained
using notched beam or related techniques, which are subject to overestimating the
toughness. More rigorous measurements (e.g., double cantilever beam and similar
test geometries), as used here, provide more conservative toughness values and allow
one to obtain R-curve behavior.b
However, these findings also reveal that the presence of larger elongated grains
in silicon nitride is not a sufficient condition for achieving high strength and high
toughness. As seen in Figure 1, careful control of the microstructure is a key here as
the toughening achieved by crack bridging processes is dependent upon the size and
volume fraction of these larger elongated grains. In self-reinforced ceramics, the
reinforcements are generated during the sintering of the ceramic, which greatly
simplifies the processing and avoids the use of expensive ceramic fibers or whiskers.
The larger elongated beta-phase grains form in silicon nitride ceramics due to the
presence of some beta phase in the alpha phase silicon nitride powders that are
typically used. Unfortunately in this case, it is difficult, if not impossible, to regulate
the size and number of the larger reinforcing grains as well as to obtain a more
uniform spatial distribution of them. In fact, each of these factors must be regulated
in order to achieve both high toughness and high strength. Recent studies have
shown that the microstructural control can be achieved by the incorporation of a
small fraction (2%) of rice-shaped beta seeds into the silicon nitride powder-sinter-
ing additive mixture. The beta seeds used to date are typically # 4 mm long by # 1 mm
in diameter; however, the optimum size (and fraction) of such seeds has not been
determined as yet. The beta seeds, then, serve as the nuclei for generating well
dispersed elongated reinforcing grains with controlled size(s).3,9 Without the use of
such a seeding method, control of the microstructure to reproducibly obtain high
strength and toughness is quite difficult. Indeed, fracture strengths can be severely
limited by the clustering of larger grains or formation of excessively large grains.3 In
addition, clustering and lack of control of the formation of the reinforcing grains
typically results in intermediate toughening effects as indicated in Figure 1.3 An
important extension of these concepts has been reported in recent studies p that
revealed substantially increased fracture toughness (in the range of 10 MPa m)
values were also achieved in a0 -SiA1ON ceramics using a similar seeding approach.10
b
The recent advent of methods to introduce very sharp notches and, hence, very short (#50 mm) initial
crack lengths (e.g., ref. 8) will also allow us to obtain a greater understanding of the R curve response at
scales more consistent with natural flaws.
Figure 1. Very high strength and toughness values are achieved by controlled generation of large
elongated b Si3 N4 reinforcing grains into a fine grained b Si3 N4 matrix.3 The average room temperature
fracture strength values, sf , shown above for each material were obtained in four point flexure using
ASTM C1161B method.7
Both the growth of the reinforcing grains during sintering and their subsequent
ability to act as reinforcements are controlled by the intergranular material found in
silicon nitride ceramics. These intergranular films (typically < 2 nm thick) are a result
of the reaction between the native oxygen in silicon nitride powders and the specific
chemical additives used to promote sintering. This reaction results in the formation
of a viscous glass that allows liquid phase sintering to occur and acts as a media for
the growth of the elongated reinforcing beta grains. In addition, the resultant inter-
granular films (IGFs) affect the debonding of the reinforcing grains from the matrix
material in the crack tip region.11 This debonding process prevents the crack from
cutting through the reinforcing grains and allows them to bridge across the crack.
This is similar to what occurs in fiber-reinforced ceramics. Recent studies are reveal-
ing that the composition of the sintering additives and, hence, the IGF can be
tailored to optimize the toughening effects of the elongated reinforcing grains in
silicon nitride ceramics. These studies revealed that the fracture toughness of SRSN
ceramics with 7 wt.% additives of yttria plus alumina could be raised by 50%
simply by substituting more yttria in place of the alumina.12 Theoretical calculations
reveal that energetics favor A1 migration to the interface from the silicate intergra-
nular film (IGF) in polycrystalline Si3 N4 . This migration of A1 (and O) to the
interface, then, leads to the growth of an epitaxial SiA1ON layer on the Si3 N4 .13
These calculations also show that the presence of an interfacial SiA1ON layer
increases the number of strong interfacial bond pairs and results in strengthening
the interface, consistent with the experimental results. The tailoring of the additive
composition plus the microstructure is now recognized as the critical factors in
obtaining both high fracture toughness and high fracture strength.
What can we expect out of such toughened ceramics? The object must be to
generate robust materials with improved mechanical properties and resistance to
damage during handling or in-service. As an example, the fracture strength of
toughened ceramics may become more resistant to degradation due to the severity
of surface damage. One can simulate such conditions by the damage introduced by
more severe (i.e., coarser diamond grit size) surface grinding. A very significant
benefit of tailoring the self-reinforced silicon nitrides can be seen in Figure 2 where
the high strengths are unaffected by the severity of the surface damage.14 This itself
suggests that in self-reinforced silicon nitride ceramics with controlled microstruc-
tures (e.g., eliminate wide variations in microstructures, extensive clustering of larger
elongated grains and formation of excessively large grains) it should be possible to
improve the flaw size distributions and, thus, the strength distributions. Indeed
recent studies of tensile strength distributions of a self-reinforced silicon nitride
ceramic with both a high toughness and a strong R-curve response give support for
this.15 In this case, a two parameter Weibull analysis was p conducted using 93 tensile
strength results obtained with a toughened ( 8:5 MPa m) silicon nitride ceramic.
This yielded a very high Weibull modulus of 33 by the maximum likelihood criterion
(or 45 by linear regression analysis) compared to typical Weibull modulus values
of 10.
4. FACTORS CONTRIBUTING TO CREEP RESISTANCE
The mechanical reliability of silicon nitride ceramics at intermediate and elevated

temperatures can be dominated by both the microstructure and the intergranular
phases present. For instance, recent creep studies indicate that the creep resistance
increases in self-reinforce silicon nitride ceramics and can be raised by increasing the
fraction of the larger elongated reinforcing grains by the beta-seeding method.16 The
elongated grains apparently restrict grain boundary sliding and grain rotation during
Figure 2. The self reinforced silicon nitride ceramics exhibit fracture strengths that are not degraded by
increasing the severity of surface damage while the unreinforced silicon nitrides undergo substantial
strength loss. Note that the x axis represents the size in microns of the diamond particles used in the
grinding wheel used to finish the tensile surface of the four point flexure bars. In this case, the depth of the
residual grinding grooves and associated damage zone as well as the size of the surface cracks generated all
increase as the size of the diamond particles in the grinding wheel increases. Data for toughened and non
toughened silicon nitride are taken from Tajima.14 Data for microstructurally designed silicon nitride from
ORNL studies.
creep by forming a rigid network that constrains the motion of the fine grained
matrix.
In addition to the affects of reinforcing grains, amorphous intergranular films can
have a substantial impact on the creep deformation of silicon nitride ceramics at
elevated temperatures.17,18 Thus, studies of the properties of oxynitride glasses com-
prising these IGF films can provide directions for tailoring the sintering additives for
silicon nitride to improve the mechanical reliability of these ceramics at elevated
temperatures, as well as at low and intermediate temperatures. For instance, studies
have shown that the glass transition temperature, which reflects a viscosity in the range
of 1012 Pa.s, can be increased by increases in the N:O ratio, decreases in aluminum
content, and substitution of smaller rare earth ions in SiREA1 oxynitride glasses.19 In
order to better illustrate this, it is informative to assess how the viscosity of bulk Si-
based oxynitride glasses is influenced by changes in the glass composition. The trends
in the shifts in the temperature-dependent viscosity with composition from recent
studies of silicon oxynitride glasses containing aluminum and various rare earths20
are summarized in Figure 3. As shown here, the viscosity of a 45Si30La25A1 glass
(composition in equivalent percent) is increased by over three orders of magnitude by
replacing some of the oxygen by nitrogen. An additional ten fold increase can be
obtained by replacing the La ion by the smaller Gd ion, while reducing the A1 content
by raising the Si (or RE) level can result in a further 100-fold increase in the glass
viscosity. In this simple illustration, upwards of a million-fold increase in viscosity
Figure 3. Modifying their composition as summarized here can raise the viscosity of SiREA1 oxynitride
glasses. By replacing some of the oxygen by nitrogen, the viscosity of a 45Si30La25A1 glass (composition
in equivalent percent) is increased by over three orders of magnitude (line 1 to line 2). Replacing the La by
Gd increases the viscosity by one order of magnitude (line 2 to line 3). Reducing the A1 content by
increasing the Si (or RE) content raises the viscosity by another two orders of magnitude.
could be achieved in the bulk glass by tailoring the composition, which should allow us
to impact the creep resistance of silicon nitride ceramics.
To examine the effects that might be achieved in the ceramic by simply altering
the viscosity of the IGF, a ‘‘brick wall’’ type model can be used to characterize the
effective viscosity of the silicon nitride ceramic composed of rigid, non-deforming
grains surrounded by a viscous media.17 For the simple case of an array of cubic rigid
grains with a size of G with an amorphous film of a thickness (t) and a viscosity (ho ),
the effective viscosity (heff ) of the silicon nitride ceramic is:
heff ¼ A ho (t=G) 3
(1)
where A is a constant. For the typical 1 to 2 nm thick viscous films in silicon nitride,
the calculated heff values of a ceramic with a grain size of one micron can be as much
as six orders of magnitude greater than the ho values for the bulk glasses. In addition,
the equilibrium thickness of the intergranular films in silicon nitride ceramics has
been shown to decrease with the substitution of rare earths with smaller ionic size as
densification additives.21 Thus, the improvement in creep resistance of the ceramic
brought on by increases in bulk glass viscosities with the use of smaller rare earth
additions can be augmented by the formation of thinner IGFs in the ceramic. Studies
of the creep of silicon nitrides that had been fabricated using silica plus rare earth
oxides revealed a decrease in creep rates with the use of smaller rare earths Sm > Gd
> Dy.22 Recent studies of the creep response of commercial silicon nitride ceramics
illustrate a similar trend for decreasing creep rates when smaller rare earths are used
as densification aids with lutetium being the most effective.23 The creep resistance of
self-reinforced silicon nitrides has been found to increase by about 10-fold when the
yttria to alumina ratio in the densification additives (fixed total amount) is in-
creased,16 which is known to result in amorphous intergranular films with increased
Y: Al ratios.12a This is consistent with the observations of increases in glass viscosity
brought about by increases in the Y: Al ratio. Obviously, other microstructural
factors (e.g., presence of elongated grains, phase changes and reactions involving
the test environment, and migration and crystallization of the amorphous films) will
also influence the creep response of silicon nitride ceramics. However, there is
considerable potential for improving the creep resistance of silicon nitride ceramics
by the judicious selection of the sintering additives to control the viscosity of the
amorphous intergranular films that are formed.
5. SUMMARY
Self-reinforced silicon nitride ceramics exhibiting high toughness and strength are
obtained by tailoring the size and number of the bridging grains. In each material,
bridging mechanisms rely on debonding of the reinforcing grains from the matrix to
increase toughness. Interfacial debonding can be influenced by the nature of the
intergranular films, whose composition and properties are a function of the sintering
aids. In silicon nitride ceramics that use alumina as one of the sintering aids, the
amorphous intergranular films surround the Si3 N4 grains on which epitaxial SiAlON
layers form. In this case, interfacial debonding is dependent on the composition of
the SiAlON layer, becoming more difficult as the Al (and the O) content of the
SiAlON increases. Theoretical calculations show this results from the formation of
strong Si-O, Si-N, and Al-O bonds across the glass-crystalline interface. In addition
to the influence of intergranular films on fracture, the amorphous IGFs containing
Si, Al, O, N, and RE will also influence the temperature dependent viscosity of the
amorphous phase. Measurements of the temperature dependent viscosities of bulk
glasses reveal that the viscosity increases with increase in the nitrogen to oxygen
ratio, the substitution of smaller rare earth ions, and a reduction of the Al content in
the film. Increasing the viscosity of glasses similar to those in the amorphous inter-
granular films offers a means to substantially improve the creep resistance of the
ceramic and, thus, enhance its life at elevated temperatures. It is apparent that
sintering additives used to promote the densification of self-reinforced ceramics can
play an important role in improving both their fracture and creep resistances.
6. ACKNOWLEDGEMENT
The research was sponsored by the Division of Materials Sciences and Engineering,
U. S. Department of Energy, under Contract No. DE-AC05-00OR22725 with
UT Battelle, LLC.
REFERENCES
1. M. Savitz, ‘‘Commercialization of Advanced Structural Ceramics, Part I: Patience is a Necessity,’’
Am. Ceram. Soc. Bulletin 78(1), 53 56 (1999).
2. K. Moerganthaler, private communication.
3. P. F. Becher, E. Y. Sun, K. P. Plucknett, K. B. Alexander, C H Hsueh, H T Lin, S. B. Waters, C. G.
Westmoreland, E S Kang, K. Hirao, and M. Brito, ‘‘Microstructural Design of Silicon Nitride with
Improved Fracture Toughness, Part I: Effects of Grain Shape and Size,’’ J. Am. Ceram. Soc., 81(11)
2821 30 (1998).
4. P. F. Becher, C. H. Hsueh, K. B. Alexander, and E. Y. Sun, ‘‘The Influence of Reinforcement
Content and Diameter on the R Curve Response in SiC Whisker Reinforced Alumina,’’ J. Am.
Ceram. Soc., 79(2) 298 304 (1996).
5. P. F. Becher, ‘‘Microstructural Design of Toughened Ceramics,’’ J. Am. Ceram. Soc., 74(2) 255 69
(1991).
6. A. G. Evans, ‘‘Perspective on the Development of High Toughness Ceramics,’’ J. Am. Ceram. Soc.,
73(1) 187 206 (1990).
7. C1161 02c Standard Test Method for Flexural Strength of Advanced Ceramics at Ambient Tem
perature, Developed by Subcommittee: C28.01, ASTM Book of Standards, Volume: 15.01, ASTM
International, West Conshohocken, PA, 2003.
8. H. Kawamoto, Y. Takigawa, K. Hiramatsu, and A. Okada, ‘‘Behavior of Crack Grow Resistance in
Toughened Anisotropic Silicon Nitride Ceramics,’’ in Key Engineer Materials, Vols. 175 176: Engin
eering Ceramics, edited by P. Sajgalik and Z. Lences, (Trans Tech Publications, Zurich, 2000),
pp. 241 252.
9. a. K. Hirao, T. Nagaoka, M. E. Brito, and S. Kanzaki, ‘‘Microstructure Control of Silicon Nitride by
Seeding with Rodlike b Silicon Nitride Particles,’’ J. Am. Ceram. Soc., 77, 1857 62 (1994). b. K.
Hirao, T. Nagaoka, M. E. Brito, and S. Kanzaki, J. Ceram. Soc. Jpn. 104: 55 59 (1996).
10. J. Kim, A. Rosenflanz, and I. W. Chen, ‘‘Microstructure Control of In Situ Toughened a’ SiAlON
Ceramics,’’ J. Am. Ceram. Soc., 83(7) 1819 21 (2000).
11. a. P. F. Becher, E. Y. Sun, C. H. Hsueh, K. B. Alexander, S. L. Hwang, S. B. Waters, and C. G.
Westmoreland, ‘‘Debonding of Interfaces Between Beta Silicon Nitride Whiskers and Si Al Y
Oxynitride Glasses,’’ Acta. Metall., 44(10) 3881 93 (1996). b. E. Y. Sun, P. F. Becher, C. H.
Hsueh, G. S. Painter, S. B. Waters, S. L. Hwang, and M. J. Hoffmann, ‘‘Debonding Behavior

Between b Si3 N4 Whiskers and Oxynitride Glasses With And Without An Epitaxial b SiAlON
Interfacial Layer,’’ Acta. Mater., 47(9) 2777 85 (1999).
12. a. E. Y. Sun, P. F. Becher, C H Hsueh, S. B. Waters, K. P. Plucknett, K. Hirao, and M. Brito,
‘‘Microstructural Design of Silicon Nitride with Improved Fracture Toughness, Part II: Effects of
Additives,’’ J. Am. Ceram. Soc., 81(11) 2831 40 (1998). b. P. F. Becher, G. S. Painter, E. Y. Sun,
C. H. Hsueh, and M. J. Lance, ‘‘The Importance of Amorphous Intergranular Films in Self
Reinforced Si3 N4 Ceramics,’’ Acta. Mater., 48(12) 4493 99 (2000).
13. G. S. Painter, P. F. Becher, and Ellen Y. Sun, ‘‘Bond Energetics at Intergranular Interfaces in
Alumina Doped Silicon Nitride,’’ J. Am. Ceram. Soc., 85(1) 65 67 (2002).
14. Y. Tajima and K. Urashima, ‘‘Improvement of Strength and Toughness of Silicon Nitride Ceramics,’’
in Tailoring of Mechanical Properties of Si3 N4 Ceramics, edited by M. J. Hoffmann and G. Petzow,
(Kluwer Academic Publishers, Dordrecht, Netherlands, 1994), pp. 101 109.
15. C. W. Li, J. Wimmer and J. J. Nick, ‘‘Tensile Strength of an In Situ Reinforced Silicon Nitride
Exhibiting R Curve Behavior,’’ ECD meeting, January 2003, submitted to Ceramics Transactions.
16. H. T. Lin, unpublished results.
17. a. D. Wilkinson, ‘‘Creep Mechanisms in Multiphase Ceramic Materials,’’ J. Am. Ceram. Soc., 81(2)
275 99 (1998). b. J. R. Dryden and D. S. Wilkinson, ‘‘Three Dimensional Analysis of the Creep Due
to a Viscous Grain Boundary Phase,’’ Acta. Mater. 45(3) 1259 73 (1997).
18. W. E. Luecke and S. M. Wiederhorn, ‘‘A New Model for Tensile Creep of Silicon Nitride,’’ J. Am.
Ceram. Soc. 82(10) 2769 78 (1999).
19. P. F. Becher, S. B. Waters, C. G. Westmoreland, and L. Riester, ‘‘Influence of Composition on the
Properties of SiREAl Oxynitride Glasses: RE La, Nd, Gd, Y or Lu,’’ J. Am. Ceram. Soc., 85(4)
897 902 (2002).
20. P. F. Becher and M. K. Ferber, ‘‘The Temperature Dependent Viscosity of SiREAl Based Glasses As
a Function of N:O and RE:Al Ratios. Where RE La, Gd, Y and Lu,’’ to be published.
21. C. M. Wang, X. Pan, M. J. Hoffmann, R. M. Cannon, and M. Rühle, ‘‘Grain Boundary Films in
Rare Earth Based Silicon Nitride,’’ J. Am. Ceram. Soc., 79(3) 788 92 (1996).
22. M. K. Cinibulk, G. Thomas and S. M. Johnson, ‘‘Strength and Creep Behavior of Rare Earth
Disilicate Silicon Nitride Ceramics,’’ J. Am. Ceram. Soc. 75(8) 2050 55 (1992).
23. a. S. M. Wiederhorn and M. K. Ferber, ‘‘Silicon nitride for gas turbines,’’ Current Opinion in Solid
State and Mater. Sci., 5: 311 16 (2001). b. F. Lofaj, S.M. Wiederhorn, G.G. Long, P.R. Jemian and
M.K. Ferber, ‘‘Cavitation Creep in the Next Generation Silicon Nitride,’’ in Ceramic Materials and
Components for Engines, edited by J.G. Heinrich and F. Aldinger (Wiley VCH 2001), pp. 487 to 493.
TOUGHENING AND
STRENGTHENING
MECHANISMS IN
NANOCOMPOSITES BASED
ON DISLOCATION ACTIVITY
Hideo Awaji and Seong-Min Choi*
1. INTRODUCTION
Ceramics have low fracture toughness because dislocations are difficult to move in
ceramics especially at room temperature. It is, therefore, hypothesized that a frontal
process zone (FPZ) ahead of a crack tip is composed of many nano-cracks rather
than dislocations as in metals.1,2 To overcome the inherent brittleness of ceramics, a
new microstructural design concept must be developed. The design concept of
‘‘nanocomposites’’ is a new, and significantly improved strengths are achieved with
moderate enhancement in fracture toughness.3 The typical microstructure of nano-
composites consists of second-phase nano-size particles dispersed within the matrix
grains. Thermal expansion mismatch between the matrix and second-phase particles
improves several mechanical properties of nanocomposites.
Remarkable characteristics of nanocomposites observed by many researchers
are summarized:4,5 Drastic change of fracture mode is observed from intergranular
fracture of monolithic ceramics to transgranular fracture of nanocomposites,
especially in alumina-based nanocomposites.3–13 Several mechanical properties
were also improved, such as fracture strength,3,4,7,8,9,14 fracture toughness,3,4,7,9,12
thermal shock resistance,3 creep resistance,3,4,15 hardness,3,9,12,14 and wear resistance.4
* Nagoya Institute of Technology, Gokiso cho, Showa ku, Nagoya, Japan 466 8555 awaji@mse.
nitech.ac.jp, & Fax.: 052 735 5276
191
192 Hideo Awaji et al.
Moderate enhancement of fracture toughness with improved fracture strength is

typical in nanocomposites, in spite of the well-known trade-off relation between
the fracture strength and fracture toughness of monolithic ceramics.
In this paper, strengthening and toughening mechanisms in alumina/silicon
nitride nanocomposites are discussed, emphasizing on dislocation activities in an
alumina matrix. The strengthening mechanism is expressed as: Dislocations will be
generated around dispersed nano-particles at elevated temperatures caused by ther-
mally induced residual stresses, release intrinsically existed residual stresses in sin-
tered alumina grains, and then improve the strength of alumina. The toughening
mechanism is expressed as: Dislocations in alumina grains become sessile disloca-
tions at room temperature, act as nano-crack nuclei under highly stressed area such
as a main crack tip, expand the frontal process zone at the crack tip, and enhance the
fracture toughness.
2. RESIDUAL STRESS
Microstructural characteristics of intra-type nanocomposites are such that dispersed

nano-size particles are embedded within matrix grains. Many researchers have ob-
served this microstructure in nanocomposites,3,7,8,9,14,17 and this microstructure gen-
erates thermally induced residual stresses after sintering due to thermal expansion
mismatch between the dispersed particles and matrix18 and generate dislocations.
We analyze residual stresses around dispersed particles using a simplified model
consisting of a spherical particle within a concentric matrix sphere with axial sym-
metry shown in Fig. 1 to clarify dislocation activities in nanocomposites. Existence of
a surrounding interlayer (glass-phase) on the second-phase particle is also considered
for analyzing the effects of the interlayer on the stress distributions.
General solutions of thermal residual stresses and displacement in a homoge-
neous hollow sphere under a constant temperature difference are derived as:19,20
In the second-phase sphere,
2 ap E p u0 Ep
srp ¼ sup ¼ þ c1 , (1)
3 1 np 1 2np
αp, Ep, νp dg
a b r
αm, Em, νm
αg, Eg, νg
Figure 1. A spherical particle with a glassy phase within a concentric sphere of a matrix grain.
Toughening and Strengthening Mechanisms in Nanocomposites 193
1 1 þ np
up ¼ ap u0 r þ c1 r, (2)
3 1 np
in the interlayer,
2 ag Eg a3 Eg 2Eg c3
srg ¼ u0 (1 3 ) þ c2 , (3)
3 1 ng r 1 2ng 1 þ n g r3
1 ag Eg a3 Eg Eg c3 Eg
sug ¼ u0 (1 3 ) þ c2 þ ag u0 , (4)
3 1 ng r 1 2ng 1 þ n g r3 1 n g
1 1 þ nEg a3 c3
ug ¼ ag uo (r 3 ) þ c2 r 2 , (5)
3 1 ng r r
and in the matrix hollow sphere,

( )
2 am Em (a þ dg)3 Em 2Em c5
srm ¼ u0 1 þ c4 , (6)
3 1 nm r3 1 2nm 1 þ n m r3
( )
1 am Em (a þ dg)3 Em Em c5 Em
sum ¼ u0 1 þ c4 þ am u 0 , (7)
3 1 nm r3 1 2nm 1 þ n m r3 1 n m
( )
1 1 þ nEm (a þ dg)3 c5
um ¼ am u0 r þ c4 r 2 , (8)
3 1 nm r2 r
where sr and su are stresses, u displacement in radial direction, r coordinates, c1 c5

unknown constants, a thermal expansion coefficient, E Young’s modulus, n Poisson
ratio, u0 temperature difference, a radius of the second phase sphere, b outer radius of
the matrix sphere, dg thickness of the interlayer, and the properties with suffix ‘‘p’’,
‘‘g’’, and ‘‘m’’ denote the properties of the second-phase particle, interlayer, and
matrix, respectively.
Equations (1) to (8) are normalized by taking the following nondimensional
variables
r sr su t max u
j ¼ , sr ¼ , su ¼ , t max ¼ , u ¼ , (9)
a a m E m u0 am Em u0 am Em u0 am au0
where t max represents the maximum shear stress. The unknown constants in Eqs. (1)
to (8) can be determined by boundary conditions.
The analyzed residual stress distributions in the matrix and second-phase par-
ticle with no interlayer are shown in Fig. 2, where the ratio of the thermal expansion
coefficients of dispersed particle and matrix (ap =am ) is 1/2, the ratio of Young’s
moduli (Ep =Em ) is 1/1, the ratio of the inner and outer radii (a/b) is 1/2, and the
Poisson ratios of both the matrix and particle are approximated to be 0.2. In the
figure, the thin solid and thin dotted lines indicate the following simple solutions to
an infinite body with a spherical second-phase particle with no interlayer analyzed by
Weyl21 and Selsing.22
a3 (ap am )u0
srm ¼ 2sum ¼ P 3 , P ¼ : (10)
r 1 þ nm 1 2np
þ
2Em Ep
Figure 2 indicates that the residual stresses have the highest values at the particle/
matrix boundary and reduce quickly as distance from the boundary increases.
Therefore, this highly stressed area is limited in the vicinity of the particle in the
case of the nano-sized particles, and if these stresses are high enough to generate
dislocations in the matrix, only dislocations will be generated and locate in the
vicinity of the boundary.
Figure 3 shows the residual stress distributions of su and t max in the matrix for
the case where a/b is 1/5, 1/2, and 1/1.25. The matrix is assumed to be Al2 O3 and the
second-phase SiC. The material properties of polycrystalline Al2 O3 and SiC are
shown in Table 1. Figure 3 suggests that the matrix with smaller a/b value has higher
stress on the particle/matrix boundary, but the distributions of the maximum shear
stress are almost independent of a/b.
Figure 4 shows the stress distributions in the particle, interlayer, and matrix for
the Al2 O3 =SiC system for supposed values of ag =am , under the assumption of
0.4
αp/αm=1/2
σ*θ
Nondimensional stresses
0.2 Ep/Em=1/1
−0.2 Finite sphere

σ*r (a/b=1/2)
Infinite sphere
−0.4
particle
−0.6
0 1 2 3 4 5
Nondimensional coordinates (ξ)
Figure 2. Comparison of stresses around a dispersed particle without a interlayer in matrices of finite and
infinite spheres.
0.4
a/b=1/5
a/b=1/2
0.2 a/b=1/1.25
-0.2 σθ*
− τ*max
-0.4
0 1 2 3 4 5
Nondimensional coordinates, ξ
Figure 3. Stress distributions around a second phase particle embedded within a matrix grain, calculated
for the Al2 O3 =SiC system.
Table 1. Residual stresses along the particle/matrix boundary in Al2 O3 =SiC nanocomposites
under the assumption of DT ¼ 1,3258C anda=b ¼ 1=5.
Stresses on the boundary

am =ap 106 (K1 ) Em =Ep nm =np
(GPa) su (GPa) smax (GPa)
8.8/4.7 380/490 0.21/0.19 0.98 1.44
αg/αm=1.5
1
αg/αm=0.8
αg/αm=0.3
σr*
0.5 σθ*
−0.5
0.9 1 1.1 1.2 1.3

particle glass matrix
Nondimensional coordinates, ξ
Figure 4. Stresses in a particle, interlayer, and a matrix for the Al2 O3 =SiC system, where dg=a 0:1,
a=b 1=2, and Eg Em.
dg=a ¼ 0:1, a=b ¼ 1=2, and Eg ¼ Em . It is noted that the existence of interlayer on the
spherical particle has significant effects on the stress distributions in the matrix.
Residual stresses calculated on the particle/matrix boundary with no interlayer
for the Al2 O3 =SiC system are shown in Table 1, where u0 ¼ 13258C anda=b ¼ 1=5
are assumed. There is large maximum shear stress on the particle/matrix boundary,
which is expected to generate dislocations around the dispersed particles within
alumina grains.
3. DISLOCATION ACTIVITY
The temperature dependency of the critical resolved shear stresses (CRSS) for a
single a Al2 O3 crystal was estimated by Lagerlöf et al. experimentally.23 They
showed that the relation between temperatures and CRSS of both basal slip and
prism plane slip in a single a Al2 O3 crystal could be described by a simple
logarithmic law over a wide range of temperature;
ln t c ¼ ln t 0 B T, (11)
where t c represents the CRSS, T temperature [K], and t 0 ¼ 109 GPa or 9 GPa and
B ¼ 0:0052 or 0.0026 for basal slip or prism plane slip, respectively. Equation (11)
indicates that at room temperature (T ¼ 300 K), t c is calculated to be 22.9 or
4.1 GPa for basal slip or prism plane slip, respectively, which is higher than the
maximum residual shear stress on the grain boundary listed in Table 1, and also
higher than the theoretical strength of a single a Al2 O3 crystal.
Figure 5 shows the temperature dependencies of the CRSS for basal slip and
prism plane slip in a single a Al2 O3 crystal and the residual shear stress on the
Al2 O3 =SiC boundary in nanocomposites, where t cb and t cp represent the CRSS for
basal slip and prism plane slip, respectively, and t res is the maximum residual shear
stress on the grain boundary of Al2 O3 =SiC system calculated under the assumption
that the residual stress is linearly related to temperature. It is noted that dislocation
movement in Al2 O3 grains is possible at temperature ranging from 7008C to 12008C,
suggesting that this temperature range is quite important in generating dislocations in
the Al2 O3 matrix during the cooling process after sintering or annealing. Therefore,
control over the cooling rate and pressure during the cooling process is required for
generating dislocations. Further appropriate annealing will lead to the development
of dislocation networks as a result of dislocation rearrangement. It is worthy of note
that dislocations in a single a Al2 O3 crystal can move at elevated temperatures due
to residual stresses, whereas these dislocations cannot move at room temperature and
become sessile dislocations.
The mismatches in thermal expansion coefficients and Young’s moduli between
the matrix and dispersed particles yield highly localized residual stresses around the
particles. The stress distribution reduces quickly as distance from the boundary
increases because of the nano-sized particles, which can generate small defects such
as dislocations in close vicinity of the particles at elevated temperatures during the
cooling process, as shown in Fig. 6(A). Large-scale cracks or other large defects will
be difficult to create in nanocomposite systems due to the residual stresses. Con-
versely, if cracks instead of dislocations are created, the FPZ toughening mechanism
mentioned later will not operate effectively. Only dislocations can disperse in matrix
grains at elevated temperatures and operate as nano-crack nuclei at room tempera-
ture. Several researchers have observed embryonic dislocations in the matrix
grains,8,9,14,16,17 and only Niihara observed highly developed dislocation networks
in the annealed Al2 O3 =SiC nanocomposites.3 This fact suggests that appropriate
annealing after sintering is important to disperse dislocations and make dislocation
τcb : Basal slip

τcp : Prism plane slip
Shear Stress (GPa)
2 τcp τres: Residual stress
τres
τcb
0
0 200 400 600 800 1000 1200
Temperature (°C)
Figure 5. Temperature dependencies of the critical resolved shear stresses in a Al2 O3 and the residual
shear stress on the boundary of Al2 O3 =SiC in nanocomposites.
networks around the particles, shown in Fig. 6(B), which will enhance fracture
toughness of the nanocomposites.
4. TOUGHENING MECHANISM
Polycrystalline ceramics usually exhibit rising R-curve behavior. A schematic dia-

gram explaining a cracked surface in the polycrystalline ceramics is shown in Fig. 7.
This figure indicates that there are two kinds of fracture energy to prevent crack
propagation in ceramics; one is the fracture energy for formation of the FPZ at the
crack tip, and the other is the mechanical energy consumed at the bridging in the
wake. Therefore, the Griffth-Irwin formula can be expressed for mode I crack
extension24 as
2
KIC
¼ 2gi þ 2gR (12)
E0
Dislocation
Subgrain
boundary
Nano-size particle
(A) After sintering (B) After annealing

Figure 6. Dislocation creation by sintering, and dislocation dispersion by annealing.
σ
wake
precracked surface critical

bridging
FPZ
Figure 7. Schematic drawing of a frontal process zone and bridging in a wake for polycrystalline ceramics
with R curve behavior.
where the left side of Eq. (12) indicates the critical energy release rate, the right side is
the fracture energy rate, KIC the fracture toughness of the material with rising R-
curve behavior, gi mode I fracture energy per unit area of the cracked surface to
create the critical FPZ size, gR fracture energy per unit area to be consumed at the
bridging, and E 0 ¼ E for plane stress, E 0 ¼ E=(1 n2 ) for plane strain. Equation (12)
is derived on the premise that the FPZ is small enough to satisfy the condition of
small scale yielding.
Referring to Eq. (12), we recognize that toughening of brittle materials can be
achieved by increasing the fracture energy of the materials; the intrinsic fracture
energy for formation of the FPZ and the extrinsic fracture energy consuming in the
bridging. Much effort has been directed toward increasing the intrinsic energy
consumed in the FPZ in such ways as phase-transformation toughening and nano-
composites which will be mentioned later, and increasing the extrinsic energy con-
sumed at the bridging by means of whisker, fiber, and platelet reinforcement in the
matrix. The former is the FPZ toughening mechanism and the later is the crack-face
bridging mechanism. The toughening mechanism of nanocomposites belongs to the
FPZ toughening mechanism, where dislocation activities in matrix grains expand the
FPZ size and increase the intrinsic fracture energy.24
Figure 8 shows a schematic illustration of the FPZ at a main crack tip in the
annealed Al2 O3 =SiC nanocomposites. In a matrix grain, sub-grain boundaries with
dislocations or dislocation networks are generated around the dispersed SiC nano-
particles by means of annealing procedure. Therefore, when a main crack tip reaches
this area, the sessile dislocations in the ceramic matrix at room temperature will
operate as nano-crack nuclei in the highly stressed area. Then the FPZ are expanded
owing to nano-crack formation and consequently enhance the fracture toughness of
the material.25
4. STRENGTHENING MECHANISM
Nano-size dispersed particles drastically change the fracture mode from intergranu-
lar fracture to transgranular fracture, and also improve the fracture strength of
ceramics markedly, especially after annealing, mentioned previously. To explain
these phenomena, we consider a strengthening mechanism in nanocomposites
Dislocation
Nanocrack
Matrix grain Dispersed particle
Sub-grain boundary
Main crack tip Frontal Process
Zone
Figure 8. Schematic description of the toughening mechanism in nanocomposites.

based on dislocation activities. A schematic explanation of the mechanism is shown

in Fig. 9. A sintered monolithic polycrystalline alumina has tensile residual stresses in
grains and at grain boundaries intrinsically because of anisotropc thermal expansion
and elastic modulus along the crystal axes, shown in Fig. 9(A). Therefore, it is
conceivable in sintered polycrystalline alumina that synergetic effects of both residual
stresses and processing defects, such as cavities, create the largest crack along the
grain boundary in the material, and the crack can be equivalent to the grain size of
the material. The fracture toughness of grain boundaries is usually lower than that
within the grains. Hence, polycrystalline alumina exhibit mainly intergranular frac-
ture mode, as schematically shown in Fig. 9(A). The longest crack generated along
the boundaries will be the Weibull’s weakest crack that dominates the strength of the
specimen. Thus, the weakest crack length will become comparable with the grain size.
It is widely accepted that the strength of the polycrystalline materials shows depend-
ence on the grain size, which is known as Petch’s relation.26 However, the nano-
cracks created by dislocations in the matrix are expected to be small compared with
the grain size, as shown in Fig. 9(B), and the dislocations reduce the residual stresses
at the grain boundaries. Therefore, the weakest crack length dominating the strength
is considered to be equal to the grain size in monolithic ceramics, and to the size of
the largest cavity in the nanocomposites. This is a source of highly improved strength
of nanocomposites.
On the contrary, dislocations generated around the dispersed particles in the
nanocomposites relieve the tensile residual stresses in the matrix, and consequently
reduce the defect size at the grain boundaries, shown in Fig. 9(B). Also, the disloca-
tions are difficult to move in ceramics at room temperature, serve as origins of stress
concentrations, and create small nano-cracks around the main crack tip. The nano-
cracks slightly reduce the transgranular strength of the alumina matrix. Then the
reduction of both the defect size along the grain boundaries and the transgranular
strength in the matrix are attributable to a change in the fracture mode from
Remote stress
Residual
stress Crack path
Nanocrack
Crack path Process

zone wake
Cavity The weakest
crack
Dispersed
particle
Grain
boundary Dislocation
(A) Monolithic alumina (B) Annealed nanocomposites

Figure 9. Schematic description of the strengthening mechanism in nanocomposites.
intergranular fracture in monolithic alumina shown in Fig. 9(A) to transgranular

fracture in nanocomposites, shown in Fig. 10(B). Also, the fracture surface of the
transgranular mode of nanocomposites is not a simple planar cleavage plane. Many
steps are observed on the surface,6,10,12,13 which can be evidence of nano-cracking
around the second-phase particles in the FPZ.
Relieving tensile residual stresses in the matrix grains due to dislocations results
in improvement of several mechanical properties of nanocomposites, such as thermal
shock resistance, wear resistance, creep resistance, and hardness.
5. CONCLUSIONS
The toughening and strengthening mechanisms in alumina-based nanocomposites

were discussed, emphasizing dislocation activities in alumina. The residual stresses
around a spherical second-phase nano-particle within a concentric sphere of matrix
grain were analyzed to clarify the effects of residual stresses on the dislocation
activities. The results revealed that the residual stresses were large enough to generate
dislocations at elevated temperatures, and these dislocations become sessile disloca-
tions at room temperature. The generated dislocations relieve tensile residual stresses
intrinsically existing in alumina grains, reduce the defect size at the grain boundary,
and consequently improve the strength of alumina markedly. Conversely, the sessile
dislocations at room temperature serve as origins of stress concentrations, operate as
nano-crack nuclei at a main crack tip, expand the frontal process zone at a crack tip,
and enhance the fracture toughness of alumina.
REFERENCES
1. R. G. Hoagland and J. D. Embury, A Treatment of Inelastic Deformation Around a Crack Tip due
to Microcracking, J. Am. Ceram. Soc., 63, 404 410 (1980).
2. A. G. Evans and K. T. Faber, Toughening of Ceramics by Circumferential Microcracking, J. Am.
Ceram. Soc., 64, 394 398 (1981).
3. K. Niihara, New Design Concept of Structural Ceramics Ceramic Nanocomposites , The Centennial
Memorial Issue of Ceram. Soc. Japan, 99, 974 981 (1991).
4. R. W. Davidge, R. J.Brook, F. Cambier, M. Poorteman, A. Leriche,D. O’Sullivan, S. Hampshire,
and T. Kennedy, Fabrication, Properties, and Modelling of Engineering Ceramics Reinforced with
Nanoparticles of Silicon Carbide, British Ceramic Trans., 96, 121 127 (1997).
5. M. Sternitzke, Review: Structural Ceramic Nanocomposites, J. Euro. Ceram. Soc., 17, 1061 1082,
(1997).
6. A. Nakahira and K. Niihara, Sintering Behaviors and Consolidation Process for Al2 O3 =SiC Nano
composites, J. Ceram. Soc. Jpn, 100, 448 453 (1992).
7. J. Zhao, L. C. Stearns, M. P. Harmer, H. M. Chan, G. A. Miller, and R. F. Cook, Mechanical
Behavior of Alumina Silicon Carbide ‘‘Nanocomposites’’, J. Am.Ceram. Soc., 76, 503 510 (1993).
8. C. E. Borsa, S. Jiao, R. I. Todd, and R. J. Brook, Processing and Properties of Al2 O3 =SiC
Nanocomposites, J. Microscopy, 177, 305 312 (1995).
9. L. Carroll, M. Sternitzke, and B. Derby, Silicon Carbide Particle Size Effects in Alumina Based
Nanocomposites, Acta mater., 44, 4543 4552 (1996).
10. H. Tan and W. Yang, Toughening Mechanisms of Nano Composite Ceramics, Mechanics of Mater
ials, 30, 111 123 (1998).
11. J. Pérez Rigueiro, J. Y. Pastor, J. Llorca, M. Elices, P. Miranzo, and J. S. Moya, Revisiting the
Mechanical Behavior of Alumina/Silicon Carbide Nanocomposites, Acta Mater., 46, 5399 5411
(1998).
12. L. Gao, H. Z. Wang, J. S. Hong, H. Miyamoto, K. Miyamoto, Y. Nishikawa, and S. D. De la Torre,
Mechanical Properties and Microstructures of Nano SiC Al2 O3 Composites Densified by Spark
Plasma Sintering, J. Euro. Ceram. Soc., 19, 609 613 (1999).
13. H. Z. Wang, L. Gao, and J. K. Guo, The Effect of Nanoscale SiC Particles on the Microstructure of
Al2 O3 Ceramics, Ceramics Int., 26, 391 396 (2000).
14. M. Sternitzke, B. Derby, and R. J. Brook, Alumina/Silicon Carbide Nanocomposites by Hybrid
Polymer/Powder Processing: Microstructures and Mechanical Properties, J. Am. Ceram. Soc., 81, 41
48 (1998).
15. A. M. Thompson, H. M. Chan, and M. P. Harmer, Tensile Creep of Alumina Silicon Carbide
‘‘Nanocomposites,’’ J. Am. Ceram. Soc., 80, 2221 2228 (1997).
16. S. Jiao, C. E. Borsa, and C. N. Walker, The Microstructures of Alumina Ceramics Containing
Nanoparticles of Silicon Carbide or Titanimun Nitride, Silicate Industriels, 7 8, 211 214 (1995).
17. I. Levin, W. D. Kaplan, D. G. Brandon, and T. Wieder, Residual Stresses in Alumina SiC Nano
composites, Acta metal. Mater., 42, 1147 1154 (1994).
18. T. Sekino and K. Niihara, Fabrication and Mechanical Properties of Fine Tungsten Dispersed
Alumina Based Composites, J. Mater. Sci., 32, 3943 49 (1997).
19. H. Awaji, S M. Choi, and E. Yagi, Mechanisms of Toughening and Strengthening in Ceramic Based
Nanocomposites, Mech. Mat., 34, 411 422 (2002).
20. S. P. Timoshenko and J. N. Goodier, Theory of Elasticity, (McGraw Hill, New York 1970). V. D.
Weyl, Influence of Internal Strains on Texture and Mechanical Strength of Porcelains, (in German),
Ber. Deut. Keram. Ges., 36, 319 324 (1959).
21. D. Weyl, Über den Einfluss innerer Spannungen auf das Gefüge und die Mechanische Festigkeit des
Porzellans, Ber. Deut. Keram. Ges., 36, 319 324 (1959).
22. J. Selsing, Internal Stresses in Ceramics, J. Am. Ceram. Soc., 44, 419 (1961).
23. K. P. D. Lagelöf, A. H. Heuer, J. Castaing, J. P. Rivière, and T. E. Mitchell, Slip and twinning in
sapphire (a Al2 O3 ), J. Am. Ceram. Soc., 77, 385 97 (1994).
24. H. Awaji, S M. Choi, T. Ebisudani, and D. D. Jayaseelan, Toughening Mechanisms of Structural
Ceramics, J. Ceram. Soc. Jpn., 108, 611 613 (2000), (in Japanese).
25. H. Awaji, S M. Choi, and D. D. Jayaseelan, Indirect Estimation of Critical Frontal Process Zone
Size Using a Single Edge V Notched Beam Technique, J. Ceram. Soc. Jpn., 109, 591 595 (2001).
26. N. J. Petch, The Cleavage Strength of Polycrystals, Iron & Steel Inst., 174, 25 28 (1953).
INDEX
Acoustic emission, 21 Glass, 60, 77 101, 121, 135, 149, 597

Alcohols, 104 Glassy carbon, 93
Alumina, 29, 65, 69, 78, 123, 177, 185, 199, Glass composites, 240, 245, 263
227, 298, 356, 366, 399, 401, 540, 574 Granular microstructure, 354
ASTM, 499, 515 Graphite, 337
Atomic force microscopy, 1
Hardness, 13, 69, 103, 126, 149, 192,
Biaxial strength, 79 629
Bio active, 388 Hertz crack, 443
Bi material interface, 376 Hydoxyapatite, 387
Brick and mortar structure, 375
Brittleness, 246 Impact, 437, 439
Indentation, 13, 21, 43, 57, 79, 93, 103,
Carbon, 31 113, 121, 149, 289, 440
Carbon composite, 31 Indentation fracture, 21, 57, 88, 113, 121,
Ceramic coatings, 28, 437 297, 440
Ceramic matrix composites, 203, 221, 233, Indentation size effect (ISE), 13,
287, 297, 307, 375 149
Cohesive, 360 Instrumented hardness, 69
Chevron notch, 513 Interface toughness, 31
Compact tension, 612 Internal fraction, 113
Conductive cracks, 609 Ion implanting, 57
Crack bridging, 310, 399
Crack growth, 8, 135, 173, 387, 404, 537, 579 Knoop indentation, 149
Crack imaging, 6
Crack path, 375, 139 Lateral crack, 447
Crack velocities, 408
Creep, 203, 221, 275, 331 Metal particles, 245
MgO, 287
Damage, 211, 280, 357 Micro fracture, 365
Disk on rod test, 366 Micromechanical simulation, 474
Double torsion, 390, 401 Mixed mode, 451
Monte Carlo Simulation, 584, 625
Elasto plastic, 93
Electric field effects, 297 Nano composites, 191
Electro ceramic, 1, 411 Nano impact, 43
Elevated temperature, 121, 451 NASA Cares, 555
Nano indentation, 13, 23, 43, 93
Fatigue, 142, 233, 399, 558 Nano scratch, 43
Ferro elastic, 471 Nanostructure, 43
Film, 438 Natural shells, 375
Finite elements, 317, 377, 414, 613 Numerical model, 354
Flexible displays, 597
Flexure test, 61, 81, 339, 453, 488, 501, 532 Piston on ring test, 367
Fragmentation, 77, 353 Plasma spray, 437, 451
635
636 Index
Porosity, 427 Thermal ageing, 263, 625

Post fracture tensile test, 307 Thermal barrier coating, 451
Proof testing, 555 Thermal shock, 263, 365, 566
PSZT, 485, 492 Thermal stress, 368
PZT, 609 Thin films, 21
Threshold stress intensity factor, 392
R curves, 182, 197, 403 Titanates, 4, 60, 287, 297, 471, 485, 609, 625
Reliability, 485, 555 Trapped crack model, 254
Residual stress, 192, 543 True hardness, 17
Scratch test, 101, 121 Ultrasonics, 1

Silica, 44, 153
Silicon, 29, 43 V K diagram, 391, 399, 184
Silicon carbide, 29, 177, 211, 221, 275, 413, Vickers indentation, 149
444, 531
Silicon nitride, 327, 413, 444, 499, 536 Weibull, 104, 144, 370, 413, 485, 558, 579,
Size effect, 13, 149, 337, 412 604, 615
Slow crack growth, 175, 558, 579
Standard reference material, 499, 531 Y TZP, 625
Strength distribution, 412, 558
Stress corrosion, 175, 536 Zirconia, 13, 60, 177, 266, 437, 451, 485, 627
Sub critical crack growth, 388 ZnO, 411, 37

Bradt R C Munz D Sakai M White K W Fracture Mechanics of Cer PDF

Uploaded by

Copyright:

Available Formats

Bradt R C Munz D Sakai M White K W Fracture Mechanics of Cer PDF

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Bradt R C Munz D Sakai M White K W Fracture Mechanics of Cer PDF

Uploaded by

Copyright:

Available Formats

Fracture Mechanics

Effects of Tin on the Physical Properties and Crack Growth

3. Fracture of Nano-Scale Materials

Fracture Toughness of BaTiO3 and BaTiO3-AI203 Composite

Particle Impact Damage and Point Load-Induced Fracture

8. Reliability Prediction, Standardization and Design

The 8th International Symposium on fracture mechanics of ceramics was held on

Tuscaloosa, AL USA R.C. Bradt

W. Arnold, A. Caron, S. Hirsekorn,

Atomic Force Microscopy (AFM) is a near-field technique to generate high-reso-

2. AFAM EXPERIMENTAL TECHNIQUE

In the so-called piezo-mode of atomic force microscopy an ac voltage is applied

3. APPLICATION OF AFM AND AFAM TO CRACK IMAGING

Figure 6. a) Topography of a Si sample, b) Torsional amplitude f fexcitation , c) Torsional amplitude

Sample # 12, Crack Nr. 1 Sample # 12, Crack Nr. 2

120 After 2 days 120 After 2 days

40 Ktip = 6.69 MPa.m1/2 40 Ktip = 6.46 MPa.m1/2

Conductive cantilever Plezo-Mode 7mm

We gratefully acknowledge the financial support by the German Science Founda-

Takashi Akatsu, Shingo Numata, Michiyuki Yoshida,

In this study, the nanoindentation on monoclinic-, tetragonal- and cubic-

The zirconia polycrystals with different crystallographic structure (monoclinic, tetrag-

Precise h measurement is very important for the accurate evaluation of hardness

2.3. Analysis of P-h Curves

which causes the overestimation of hardness in that region. We made a two-step

2.4. Raman Spectroscopy

3. RESULTS AND DISCUSSION

3.1. Meyer Hardness of Zirconia Polycrystals

3.2. True Hardness of CZP

The HT -value of CZP increased with a decrease in hp as shown in Fig.3. Also,

True hardness HT /GPa

0 500 1000 1500

where H0 is the hardness evaluated at hp ! 1, A is the constant associated with the

3.3. True Hardness of MZP

True hardness HT /GPa

0 500 1000 1500 2000

3.4. True Hardness of TZP

The HT -value of zirconia polycrystals increased with a decrease in hp . This is

True hardness HT /GPa

0 500 1000 1500 2000

changed depending on whether the stress-induced phase transformation occurred or

S.J. Bull, I. Arce-Garcia, E.G.-Berasetegui and T.F. Pagey

Fracture is an almost inevitable consequence of highly loaded contacts in ceramic-

The durability and performance of coatings in almost every application depend on

2.1. Sample Preparation

Table 1. Fracture events associated with brittle (ceramic) coating/substrate systems.

Substrate Coating Fracture type

Ductile Brittle Through thickness fracture, brittle fracture in coating parallel

Brittle Brittle Bulk chipping, through thickness fracture, brittle interfacial

2.2. Nanoindentation Testing

Table 2. Coating/substrate systems investigated in this study.

NbN ASP23 steel 9.5 3980 Arc evaporation 425

NbN 304 stainless steel 1.9 2810 Arc evaporation 425

CNx Silicon 11.5 400, 1000 Unbalanced magnetron 350

CNx 1 mm SiC 35.0 SiC 400, 1000 Unbalanced magnetron 350

CNx Sapphire 32 400, 1000 Unbalanced magnetron 350

TiOxNy Float Glass 6.5 200 Magnetron sputtering 200

SnO2 Float glass 6.5 200 Magnetron sputtering 200

EANet (t) ¼ EA (t) at (2)

Sample 0, 0 42.7 3.09 1085 50

Sample 0, 0 82.1 4.4 490 8