PDF

STP 1409
Fracture Resistance Testing

of Monolithic and Composite
Brittle Materials
J. A. Salem, G. D. Quinn, and M. G. Jenkins, editors
ASTM Stock Number: STPI409
ASTM International
100 Barr Harbor Drive
PO Box C700
West Conshohocken, PA 19428-2959
Printed in the U. S. A.
Library of Congress Cataloging-in-Publication Data
Fracture resistance testing of monolithic and composite brittle materials / J.A. Salem, G.D. Quinn,
and M.G. Jenkins, editors.
p. cm.
"ASTM stock number:. STP1409."
Includes bibliographical references and index.
ISBN 0-8031-2880-0
1. Brittleness---Congresses. 2. Fracture mechanics---Congresses. L Salem, J. A. (Jonathan A.),
1960- IL Quinn, G. D. (George D.) II1.Jenkins, Michael G., 1958-
TA418.16.F73 2002
620.1' t 26--<1c21
2001053953
Copyright 9 2002 ASTM International, West Conshohocken, PA. All rights reserved. This material
may not be reproduced or copied, in whole or in part, in any printed, mechanical, electronic, film, or
other distribution and storage media, without the written consent of the publisher.
Photocopy Rights
Authorization to photocopy items for internal, personal, or educational classroom use,

or the internal, personal, or educational classroom use of specific clients, is granted by the
American Society for Testing and Materials (ASTM) provided that the appropriate fee is paid to
the Copyright Clearance Center, 222 Rosewood Drive, Danvers, MA 01923; Tel: 978-750-8400;
online: http://www.copyright.com/.
Peer Review Policy
Each paper published in this volume was evaluated by two peer reviewers and at least one editor.
The authors addressed all of the reviewers' comments to the satisfaction of both the technical
editor(s) and the ASTM Committee on Publications.
To make technical information available as quickly as possible, the peer-reviewed papers in this
publication were prepared "camera-ready" as submitted by the authors.
The quaii~ of the papers in this publication reflects not only the obvious efforts of the authors and
the technical editor(s), but also the work of the peer reviewers. In keeping with long-standing
publication practices, ASTM maintains the anonymity of the peer reviewers. The ASTM Committee
on Publications acknowledges with appreciation their dedication and contribution of time and effort on
behalf of ASTM International.
Pdntedin Brk~geport,NJ
January20~
Foreword
This publication, Fracture Resistance Testing of Monolithic and Composite Brittle Materials, con-
tains papers presented at the symposium of the same name held in Orlando, Florida, on 14 November
2000. The symposium was sponsored by ASTM Committee C28 on Advanced Ceramics. The sym-
posium chairman was Jonathan A. Salem, NASA Glenn Research Center at Lewis Field, and the sym-
posium co-chairmen were George D. Quinn, National Institute for Standard and Technology, and
Michael G. Jenkins, University of Washington.
Contents
Overview vii
PLENARY SESSION
Fracture Mechanics of Brittle Ceramics - 30 Years of Progrem--R.C. BRADT
IMPLICATIONS FOR DESIGN AND TESTING
Failure from Large Grains in PolycrystaBine Ceramics: Transitions in Fracture

Toughness---s. FREIMAN 17
Stresses in Ceramic Plates Subjected to Loading Between Concentric Rings--

L.M. POWERS, J.A. SALEM, AND A.S. WEAVER 30
FRACTURETOUGHNESSSTANDARDIZATION
Development, Verification, and Implementation o f a National Full-Consensns

Fracture Toughness Test Method Standard for Advanced Ceramics--
M.G. mNKINS,J.A. SALEM,O.D. QUINN,ANDL BAR-GN 49
Does Anyone Know the Real Fracture Toughness? SRM 2100: The World's First
Ceramic Fracture Toughness Reference Material---O.D. QUtNN,K. •
R. J. GETI'INGS, J. A.SALEM, AND J. J. SWAB 76
Fracture Toughness of Ceramics using the SEVNB Method: From a Preliminary

Study to a Standard Test Method--J. J. KOBLER 93
The Fracture Toughness Round Robins in VAMAS: What We Have Learned--

G. D. QUINN 107
CRACKGROWTHRESISTANCE
R-Curve Measurement of Silicon Nitride Based Ceramics at Elevated Temperatures

with Single Edge Notched Beam Specimens---s. SAKAGUCm 129
Crack Deflection Toughening Mechanism in Brittle Materials---ft. AWAJI,
M. EBISUDANI, S.-M. CHO1, AND T. OHASHI 137
Application of Quantitative Fractography to the Characterization of R-Curve

Behavior--J. J. MECHOLSKY, JR., T. J. HILL, AND Z. CHEN 152
Fracture Testing of a Layered Functionally Graded Material--M.d. roLL,

R.D. CARPENTER, G.H. PAULINO, Z.A. MUNIR, AND J.C. GIBELING 169
UNIQUE MATERIALS AND ENVIRONMENTAL EFFF_~S
Environmental and Thermal Effects on the Toughness of TiCN-Based M a t e r i a l s - -

S. GUICCIARDI,C. MELANDRI, F. MONTEVERDE, A. BELLOS1, AND G. DE PORTU 187
Fracture Toughness Studies on Ceramics and Ceramic Particulate Composites

at Different Temperatures--G.A. C,OGOTSI 199
The Effect of Stress Rate on Slow Crack Growth Parameter Esfimates--J.A. SALEM
AND M.G. JENKINS 213
Overview
During the past decade, ASTM Committee C28 on Advanced Ceramics, along with its European
and Japanese counterparts, has made great progress in the development of new standards for the mea-
surement of fracture toughness, slow crack growth, and biaxial strength. These standards are de-
signed to result in quality measurements for engineering, research, and general characterization pur-
poses. They strive to strike a balance between accuracy and reasonable convenience. Although work
continues on improvement of existing standards and the development of new standards, members of
Committee C28 felt that the time had come to review and summarize the recent efforts by sponsor-
ing a symposium on fracture resistance testing of brittle materials. The symposium was held in
Orlando, Florida during November of 2000. Participants came from Europe, Asia, and the Americas.
Four themes were encompassed in the presentations: Implications for Design and Testing;
Fracture Toughness Standardization; Crack Growth Resistance; and Unique Materials and
Environmental Effects. Although three of the themes are relatively broad and papers representing a
variety of subtopics were presented, the session on Fracture Toughness Standardization was rela-
tively focused. The emphasis on this topic merits some explanation. Fracture toughness is a funda-
mental measure of a ceramic's ability to tolerate flaws, or conversely, its brittleness; however, there
are conflicting views on the importance of that property for design. Some design methodologies cur-
rently in use for ceramics (i.e., those based on strength statistics) do not employ fracture toughness,
despite its overwhelming importance in classical deterministic design techniques. However, it is the
fracture toughness and the flaws inherent in a ceramic material that control the strength measurements
used as the basis for such reliability methodologies. Furthermore, deterministic design methods that
employ fracture toughness are being used for ceramic component design.
Thus, quality fracture toughness measurements using a reasonable degree of similitude are needed.
Furthermore, since research on toughening of ceramics exists and continues, the techniques need to
be efficient in terms of the material used and the time required, Committee C28 chose three tech-
niques for development and standardization in the new fracture toughness standard C 1421-99. These
techniques are detailed in this special technical publication and, importantly, the three techniques
show convergence when good metrology is employed. In addition to standardized techniques, the
section Fracture Toughness Standardization discusses the single edged V-notched beam method that
is on a fast track for standardization in Europe.
The section on Implicationsfor Design and Testing contains papers on the analysis of plates for bi-
axial strength testing and the transition in measured fracture toughness from a value associated with
the properties of a single grain to the polycrystalline value. A number of papers presented in the other
sessions also had implications for design and testing. These included papers on the effect of stress rate
on slow crack growth parameters for design usage, and the application of quantitative fraetography
to the characterization of the R-curve behavior of a silicon nitride for bearing applications.
The section on Unique Materials and Environmental Effects includes papers on elevated tempera-
ture fracture toughness testing of particulate reinforced ceramic composites, thermal and environ-
mental effects on the fracture toughness of titanium carbonitrides for machining, and environmental
interactions that lead to rate effects in "dynamic fatigue" (i.e., stress corrosion) testing.
viii OVERVIEW
The section on Crack Growth Resistance includes papers on testing of functionally graded materi-
als, elevated temperature R-curve testing, and the study of a toughening mechanism. Although most
researchers applied classical mechanical techniques for the measurement of fracture toughness or
crack growth resistance, both theoretical and fractographic methods were also presented.
The papers are relevant in that they supply the background that determined many of the guidelines
used in current Committee C28 standards. This volume not only summarizes the latest standard meth-
ods for the measurement of fracture toughness, slow crack growth, and biaxial strength, but also in-
dicates new areas for fracture toughness test method development and standardization: testing of
complex materials, elevated temperature measurement, and R-curve measurement. Indications were
also given that test method development is needed in the areas of elevated temperature biaxial
strength measurement and slow crack growth by static loading. It is hoped that in coming years these
areas will be pursued fruitfully by Committee C28 on Advanced Ceramics.
Jonathan A. Salem
Life PredictionBranch
NASA Glenn Research Centerat Lewis Field
Cleveland,Ohio 44135
Symposiumco-chairand co-editor
George D. Quinn
Ceramics Division
National Institutefor Standardsand Technology
Gaithersburg,MD 20899
Michael G. Jenkins
Departmentof MechanicalEngineering
Universityof Washington
Seattle, WA 98195
Plenary Session
Richard C. Bradt 1
Fracture Mechanics of Brittle Ceramics - 30 Years of Progress
Reference: Bradt, R. C., ~Fracture Mechanics of Brittle Ceramics - 30 Years of

Progress, ~ Fracture Resistance Testing of Monolithic and Composite Brittle Materials,
ASTMSTP 1409, J. A. Salem, G. D. Qninn and M. G. Jenkins, Eds., American Society for
Testing and Materials, West Conshohocken, PA, 2002.
Abstract: The application offiacUae mechanics to advanced technical ceramics was initiated
nearly 30 years ago. It was heralded by an international conference at the Pennsylvania State
University in 1973. Unbridled optimism prevailed as many believed that the coupling of
fracture mechanics with ceramics would enable high temperature advanced ceramics to find
extensive structural applications. Unfortunately, there were a number of barriers to the use
of ceramics as structural materials, even with the application of fracture mechanics.
The use of brittle ceramics in structural applications faced major challenges. These
included: (1) the absence of a reliable data base of materials properties, (2) the lack of
appropriate standards for the determination o f the properties o f these materials, (3) a lack of
any rational, fracture mechanics based design methodology for brittle ceramics, and (4) the
high costs of producing and utilizing structural parts from advanced ceramics. These were
serious deficiencies that inhibited the achievement of successful applications of ceramics.
Fracture mechanics tests provided a basis for the design methodology and also provided
understanding of the basics of the fracture process in brittle ceramics. It informed the
engineering community that ceramic fi'acture toughnesses were low, < 10 MPa~/m and that
there were serious problems with slow crack growth and fatigue. The presence of rising R-
curve behavior was observed and related to the crack tip's following wake region. Cyclic
fatigue was confirmed to exist and related to the presence of a rising R-curve.
Today, the promise of structural ceramics remains mostly unfulfilled, except for a few
niche applications that are nonetheless encouraging. In addition, there have been significant
advances in the properties and integrity of structural ceramics over the past three decades.
Advanced structural ceramics are better than ever. This can be attributed to the use of
fracture mechanics concepts to identify the microstructural processing defects and then
improve the processing methodology to eliminate those defects. There have also been major
advances in the development of standards applied to ceramics in which ASTM has played a
major role. Unfortunately, we still have not learned to design with brittle ceramics and the
cost of their utilization in structural applications remains prohibitively high.
Keywords: brittle fracture, ceramics, design, fatigue, R-curves, wake region
1Department of Metallurgical and Materials Engineering, The University of Alabama,

Tuscaloosa, AL 35487-0202.
3
Copyright9 ASTM lntcrnational www.astm.org
4 FRACTURETESTING OF MONOLITHIC/COMPOSITE MATERIALS
Introduction
Progress in the fracture toughness testing of brittle ceramics has been intimately related
to the attempt to utilize advanced ceramic materials in structural applications, primarily at
elevated temperatures in heat engines. The potential for the utilization of ceramics in heat
engines funded much of the ceramic fracture mechanics studies during the past three decades.
For that reason, this assessment of the progress of the fracture toughness testing of brittle
ceramics over the past three decades can be addressed within that perspective, for the two
topics have been inseparably associated. It may not be the "pure" fracture mechanics
historical report that some expect from this article, but it necessarily couples the fracture
mechanics testing of ceramics with the main driving force for that testing, the promise of
utilization of advanced technical ceramics in structural heat engine applications.
When the field of fracture mechanics began in earnest in the 1950s, the structural heat
engine materials activity was focused toward cermets. Cermets were believed to be capable
of combining the high t ~ properties of ceramics with the toughness of metals. This
belief was never fulfilled. Rather, the composite cermet developments reproduced the high
temperature weaknesses of metals and the brittleness of ceramics, instead of the reverse.
Thus, into the 1960s and early 1970s, the technical ceramics field was receptive to any new
approach to utilize the attractive high temperature properties of ceramics. Once introduced,
fracture mechanics was immediately perceived to be the answer, for it appeared to address
the primary obstacle to the application of ceramics in structural applications, brittle fracture.
That briefly summarizes the background as to how the two topical areas became coupled and
why they are addressed together in this paper.
Considering the situation, it is not surprising that an international conference on the
fracture mechanics of ceramics was held at The Pennsylvania State University in 1973. It was
the first of a series of such conferences held approximately every four years. The proceedings
have been published by Plenum Publishing [1]. These conferences were among the most
successful topical conferences ever held on a technical ceramics topic. They have been
instrumental in the development of the understanding of the brittle fracture of advanced
ceramics for structural applicatons. Although the theme of these conferences has drifted from
purely ceramic fracture mechanics and now encompasses other related areas of mechanical
behavior, this series of symposia has nevertheless remained a focal point for the application
of fracture mechanics applied to brittle advanced structural ceramics.
Fracture of Ceramics in the 1960s
Because the field of ceramics is primarily chemistry and crystallography based, the
introduction of fracture mechanics concepts involving more mathematics was accepted only
with reservations at the start. The leaders in the 1960s were the British and the Japanese:
TattersaU, Tappin and Davidge in England and Nakayama in Japan. Buresch and Pabst in
Germany also made significant contributions. In the United States, it was Wiederhorn at NBS
with his basic studies on crack growth in glass that lead the way. There were difficulties with
the acceptance of the concept of fracture toughness with its unusual units and with the
relationship of strength to fracture toughness. Many of those original challenges are accepted
as the basics today. In spite of the fact that the analytical aspects offi'acture mechanics had
been firmly established for at least a decade, application to ceramics did not automatically
BRADT ON 30 YEARS OF PROGRESS 5
happen. It seemed that some of the fundamentals had to be rediscovered, sometimes more
than once, almost as if perhaps they did not apply to ceramics, when in fact with their brittle
elastic character, ceramics were ideally suited to the application of fracture mechanics.
In the e ~ arena, there were diificulties with the fracture mechanics tests, similar
to the ones that there are today, and always will be. For fracture toughness measurements,
the double cantilever beam and the single edge notched beam were the favorite geometries.
However, as the "crack" in the notched beam specimens was invariably introduced with a thin
diamond saw, there was a distinct lack of reproducibility in the results. Unlike metals, it was
not poss~le to introduce fatigue cracks for the fracture measurements. In 1970 there was
almost a complete lack of understanding about the frontal process zone at the crack tip of
brittle materials and absolutely no appreciation for the effects of the following wake region
behind the advancing crack tip. Many individuals believed that dislocations were of
paramount importance at the crack tips of brittle ceramics, just like in metals. It was also
difficult for some to accept that the conversions of the advanced technical ceramic fracture
toughness values to surface energies did not yield the thermodynamic surface free energies
that were expected for brittle ceramic materials, but it was a fact.
While there were many researchers attracted to the field, often making it difficult to give
anyone credit for being the first, one application stands out above all others. It is the work
of Nakayama in Japan with his development of the work-of-fracture test and its application
to improve the thermal shock damage resistance of refractory bricks, initially low quality
fireclays. Many, not familiar with the Japanese literature, have attributed the work-of-fracture
test to Tattersall and Tappin. They undeniably were working in the same area, but Nakayama
was the one who independently built his own testing machine, developed the test and applied
it to solve a major industrial problem for which Hasselman had previously advanced an energy
balance concept. To this day, 30 years later, no application of fracture mechanics to technical
ceramics has had such a major industrial impact. Even today, virtually all industrial refractory
microstructures are designed based on some variation of the work-of-fracture test developed
by Nakayama. The volume 0fthose products is at the level of millions of tons per year!
Entry into the decade of the 1970s presented the technical community with the general
belief of having found the answer for the utilization of technical structural ceramics in heat
engine applications. That answer was fracture mechanics. However, many unanswered
questions remained. As this symposium is addressing the testing situation, a couple of
problems specifically related to that issue are appropriate to note. Several were the
introduction of sharp cracks into test specimens, measurements of crack lengths and the
ability to obtain stable fracture conditions in work-of-fracture test specimens. The testing of
brittle ceramics was not nearly as advanced as it is today [2-4]. It was also puT~lirlg that the
total energy balance for the work-of-fracture test did not always agree with the derived
energy values from fracture mechanics crack initiation tests. In the late 1960s the flame of
hope for structural ceramics had been ignited and in the 1970s was burning vigorously, but
there were many very basic questions to be asked and equally as many to be answered.
That ceramic flame created a special era, one that is generally referred to as the ceramic
fever of the 1970s, 1980s and into the early 1990s. A level of unbridled optimism swept over
the worldwide technical ceramics community. It believed that many elevated temperature
applications of structural ceramics were imminent. Any day there would be ceramic engines
in all of our automobiles [5]. The discovery of ceramic superconductors in the middle 1980s
further fueled the optimism that ceramics are the answer. Yet, now in the year 2001, many
of the potential applications of advanced structural ceramics have not been realized. It is
appropriate to examine some oftbe reasons for this failure to achieve expectations. Not a
small portion of the reasom lies in the area of fracture mechanics testing and the development
of standards for those tests. Utilization of fracture mechanics concepts in the design process
for the application of brittle advanced ceramics remains a challenge to this day.
Shortcomings of Structural Ceramics from the Perspective of Hindsight
Now that the 21st century is upon us and we do not have ceramic engines in all of our
automobiles, it is perbaps appropriate to examine why this enthusiastic structural ceramics
venture failed, although there have been ceramic successes in other venues. It is the opinion
of this author that there are four major factors that were not adequate, or did not measure up
to the technical level which was required for the success of structural ceramics in heat engine
applications over the past couple of decades. Not necessarily in their order of importance,
these include: (1) the absence of a qualified data base for those materials properties that are
required by design engineers, (2) the lack of appropriate test standards for the determination
of those properties, (3) a weakness in the knowledge of fracture mechanics based design
concepts for brittle ceramic materials, and (4) the economic factor, the prohibitive costs of
producing structural parts from advanced ceramics. These four may not be the only reasons
for the lack of success. However, each of the four certainly assumed a paramount role in
delaying the utilization of advanced technical ceramics in structural applications during the
late 20th century. Each merits individual consideration.
The Absence of a Qualified Data Base of Materials Properties
Inthe 1960s and the 1970s there did not exist even a modest data base of the properties
of advanced ceramic materials. Mechanical engineers need a reliable data base for design with
structural ceramics. Yes, some properties were known, but only for the products of the
individual ceramic manufacturers. The values were the ones published by those producers in
their sales literature. Many of the producers only published their best results in this data, but
even the best were not very satisfactory by present today standards. For example, typical
bend strengths were in the range of 100 to 200 MPa (15000-30000 psi) and reliable Weihnll
moduli were not even available. Often only a few specimens were actually tested. We know
today that those Weibull moduli were probably only about 5 or 6, values hardly suitable for
high reliability designs! Fracture toughness values simply were not available and slow crack
growth and fatigue data were non existent. A tensile strength value was practically in the
category of a dream. It was only the continuing efforts of the government and industrial
structural ceramic programs directed towards the development of ceramics for heat engines
that lead to the eventual production of sufficient data bases suitable for design purposes.
Even now, the data bases for ceramics are not widely publicized and are very limited
compared with those existing for metals. Often these ceramic data bases are not generated
using consensus standards consisting of statistically significant sample sizes.
There was yet another impediment to the establishment of qualified material data bases for
designers to draw upon. It was the continuing development of new varieties of ceramic
materials by the ceramic producers. In an attempt to be the provider of the best material for
the anticipated structural applications that were just ahead, producers kept releasing new
ceramics with different properties. An interesting example o f this activity was silicon nitride,
which many engineers considered to be the primary candidate for the ceramic materials in heat
engine structural applications. This author did not keep a running count o f the number of
different silicon nitrides that were developed, but there were numerous reaction bonded
silicon nltrides with properties that were inferior to the equally numerous densely sintered
bodies and hot pressed silicon nitride formulations. As it was expensive to develop a data
base for any one of these materials, the need for many data bases resulted in the production
o f few data bases with significant reliability.
Lack of Appropriate Standards for Property Measurements
The title o f this section is a misnomer. There were no accepted standards for the
measurements o f the fracture mechanics parameters o f structural ceramics, or for other
mechanical properties o f technical ceramics 30 years ago. During the 1960s and 1970s,
fracture mechanics measurements were being made in universities and a few industrial
research laboratories by only a few individuals. Often, new entries into the field visited with
those scientists/engineers who were already making the measurements to learn how to apply
the experimental techniques. Without standards, it is not surprising that the data bases were
inadequate for engineers to produce rational designs. As late as 1984, Quinn [6] discussed
this at length in an excellent paper, which, unfortunately was not the wakeup call that it
should have been for manufacturers. It is only during the past decade that the national and
intemational technical communities have addressed the standards issue for advanced technical
ceramics. It is gratifying to see that entire symposia, such as this, are being devoted to the
presentation and development of fracture mechanics test standards for brittle ceramics.
Absence of a Design Methodology
The design process for incorporating brittle ceramics into structural applications as it
existed in the 1960s and 1970s left much to be desired. It was essentially nonexistent.
Engineers knew how to design ceramic microstructures, but they didn't know how to design
with those ceramic microstructures [7]. The first activities in this area were the one-for-one
substitutions. The candidate advanced ceramics were patterned after the metals that were
currently in use and then directly substituted into the turbine, or other engine. Catastrophic
failures occurred. Successes were nmastaed in minutes and hours o f operation before failure.
The message was loud and clear. Designing with brittle ceramics was not the same as that
with ductile metals. Unfortunately, this obvious lesson was not learned by the engineering
community. That shortcoming continued to impede progress with advanced ceramics.
Although practically unexplainable, instead o f refocusing and intensifying efforts to
develop and advance the design process, there was a redirection o f ceramic development
towards ceramic composites that would fail gracefully. It has always been a puzzle to this
author as to why anyone wants to use a material that fails gracefully. Isn't it more natural to
want to use a material that doesn't fail at all? Unfortunately, with the redirection o f some
attention to ceramic composites, the advanced ceramics technical community failed to step
back, recognize and accept the fact that the real problem was the design methodology, not
the material failure mode. Many o f the foolish mistakes that were made with monolithic
ceramics were destined to be repeated and have been repeated with ceramic composites.
Fortunately, a few individuals recogniTed the need for a fundamental brittle ceramic design
methodology. The design group at NASA Lewis Research Center (presently NASA Glenn
Research Center) with their enthusiastic proponent, Gyekenyesi, have made major advances
in the design methodology. They have incorporated fracture statistics and finite element
stress analysis into probabalistic design concepts. These applications were aided significantly
by the rapid development of computer systems over the same time period. These allowed the
engineers to make major advances in design methodology. In the opinion of this author,
designing with advanced structural ceramics requires a distinctly different approach than metal
designs. A major reason why structural ceramics have not been as successful as many
anticipated is that the design process has still not been perfected after nearly halfa century.
Perhaps some radically innovative ideas relative to design with ceramics are necessary?
Hindsight is always "20/20" and it is appropriate to apply a little of that vision at this time.
Although I am a materials engineer, it is obvious to me that from the very beginning there was
not a sufficient level of effort devoted to the development of design criteria and processes for
the use of brittle ceramics in structural applications. Perhaps this should not be surprising,
as most of the advanced structural ceramics programs were spearheaded by materials
engineers. However, the best materials in the world will not perform unless incorporated into
a fundamentally sound design. Needless to say, the first half of the 30 years of progress
reported here lacked sufficient design activity. An exception was a NASA sponsored
program led by Mueller at the University of Washington in Seattle. He tried to develop an
academic design program combining the talents of mechanical and materials engineers, but
it did not have the longevity to be successful It is in some ways disappointing that this lesson
still has not been learned nearly 20 years later. The next generation of designers are now in
the undergraduate classrooms. They are the ones who will be expected to finally solve the
brittle structural ceramic design problems with creative, innovative, radically new advances.
Unfortunately they are not being introduced to structural design with brittle ceramics early
enough in their careers.
It was noted earlier in this section that there has been a redirection of activities to the use
of ceramic composites as a replacement for monolithic ceramics in structural applications [8].
This class of materials will not substitute directly for metals either and obviously will have its
own design criteria when they are eventually developed. Furthermore, ceramic matrix
composites have several serious problems of their own. The thermal expansion and elastic
modulus mismatch between the fibers and matrix will generate high internal thermal stresses
on cycling, perhaps in the GPa range. These stresses will create internal cracking during
cycling, eventually leading to disintegration. The logic of graceful failure is also a bit
o f a mystery. Once the elastic limit is exceeded in most structural applications, failure has
occurred, as dimensions are not maintained and continued loading or thermal cycling is not
possible. Many present ceramic cong~ositcs exhibit matrix cracking at lower stress levels than
comparable monolithic ceramics. In addition, they bave problems with thermodynamic
stability and oxidation as well. Of course, a one time application is a different situation.
The above noted, there is no doubt in the mind of this author that the design criteria for
brittle ceramics in structural applications must be a fracture mechanics based one, one
involving fracture statistics and crack growth phenomena. It will be limited by the brittle
fracture process. For that reason alone, an integral part of the overall design process must
be the development of suitable test standards for the fracture of brittle ceramics. The results
of those standard tests will define the lifetime and the limit of the design.
Economic Factor, Prohibitive Costs
There are costs associated with everything and the costs of generating a data base for
brittle structural ceramics, developing fracture mechanics standards and perfecting the design
methodology have proven to be much greater than originally anticipated in the 1960s. The
funding for these items would never have become available if it weren't for the potential of
the utilization of ceramics in heat engines. Unfortunately, these factors do not include the
cost of manufacturing the ceramic components for the anticipated applications [9,10]. In the
final analysis, ceramics will never be used in structural applications if metals continue to
perform satisfactorily and the metals are more economical to produce. This is not an
insignificant consideration. It is a major barrier. It is the bottom line for the introduction of
technical ceramics into all structural applications, not only as components for heat engines.
At the present, structural ceramics are simply much too expensive to produce to find
extensive use in most heat engine applications. In some respects it is a chicken and egg
situation. If ceramics are to compete with metals then they must be produced in large
numbers (volumes) to benefit from the economy of scale. Thus, when all of the other factors
have proven to be positive, a command decision must still be made to convert to ceramic
components. It may not be an easy decision, for it will be one with considerable risk.
Contributions of Fracture Mechanics
It is evident that fracture mechanics has made major positive contributions to the
understanding of the failure of brittle ceramics. These have occurred in spite of the modest
advances that structural ceramics have made in the penetration of the heat engine market.
Often, the contributions have been indirect ones. However, many can be addressed in terms
of direct fracture related phenomena and improved ceramics themselves.
Fracture-Related Phenomena
Very little was understood about the fundamentals of the fracture of ceramics before the
appfication of fracture mechanics concepts to advanced structural ceramics. For that reason,
it is appropriate to take a global, somewhat historic viewpoint of the progress in this
particular area. The first item of significance was the determination of the fracture toughness
values for technical ceramics. For a point ofcalihration, polycrystalline aluminum oxide or
Al203 has a fracture toughness of about 4 MPa~/m, whereas aluminum metal is about 40
MPa~/rrt This is a factor of approximately 10x. This difference should have sent engineers
a message about the fracture toughnesses of advanced ceramics and that structural component
design with ceramics would be difficult. That design process might have to be different.
Unfortunately, the message did not register. Many researchers continued to strive to make
tougher ceramics [8]. There were numerous reports of higher toughness ceramics, some
values near to 20 MPa~/m. But, without the accompanying increases in strength it was
evident that most of these reports were of the salesman variety, namely "my ceramic is
tougher than your ceramic." Although there have been modest increases in the toughness
values for advanced ceramics through microstruetural design, it is now clear that for all
practical purposes the fracture toughnesses of most advanced technical ceramics are
10 FRACTURETESTING OF MONOLITHIC/COMPOSITEMATERIALS
somewhat less than 10 MPa~/m and in many instances considerably lower. These toughnesses
are much less than that of cast iron. Everyone knows how brittle cast iron is, as well as the
challenges of designing with cast iron. Yet, there is a lot of cast iron used in the world. I
suspect that valuable lessons could be learned from cast iron applications and their design.
Fracture toughness values are not the only revelations of the past 30 years of fracture
mechanics applications to ceramics. Much has been learned about events which occur in the
crack tip region, both in front ofthe crack tip and behind the crack front in the so-called wake
region. In the front of the crack, it is now finally accepted that dislocations are not very
important. Phenomena such as microcracking, phase transitions in toughened zirconias and
elastic/stress effects from higher modulus particles, or materials are most common. Buresch
and his student Pabst were the first to recognize the importance of these phenomena.
Concepts of crack deflection and in-situ toughening by elongated grains have also been
advanced as important frontal process zone toughening phenomena by Evans and Faber. It
is interesting that only now, decades later, it has finally been realized and accepted that
dislocation processes are not very important in the fracture of advanced technical ceramics.
It is, however, behind the advancing crack tip that many of the most interesting
phenomena have been discovered. It is these that have generally been the cause of rising R-
curve behavior. From the pioneering work of Steinbrech, we know that although the crack
has passed, that following wake region can still contribute significantly to the crack growth
resistance of advanced ceramics. Many interesting things happen behind a crack in the wake
region. Microstructural elements become wedged, grains bridge across the crack surfaces and
there are extensive frictional effects. These phenomena are active for some distance behind
the crack tip, until the crack opening displacement is too great for the new fracture surfaces
to continue to interact. It is these active phenomena in the following wake region that are
primarily responsible for rising R-curves in most advanced ceramics, the increasing crack
growth resistance with crack length. This phenomenon is now known to occur in most
ceramic materials except for the finest grain size ceramic bodies and glasses. It has created
serious problems in the accurate measurement of the fracture toughness. Some researchers
have inadvertently reported fracture toughnesses for the ascending portion of the R-curve,
or perhaps even the R-curve plateau in the most extreme instances. It is a cause of
erroneously high toughness values in the literature, particularly with the indentation toughness
measurement techniques. These fracture toughnesses cannot be applied for design purposes.
Unfortunately, they have been used indiscriminately by some researchers.
There is another major effect of rising R-curves in ceramics. It is the susceptibility to
fatigue from cyclic stresses. Unfortunately, it had been previously reported that brittle
ceramics were not suscept~le to cyclic fatigue from alternating stresses. However, it is now
well established that indeed ceramics are susceptible to fatigue and the reason is the opening
and the closing of the crack in the wake region. That cyclic process damages those elements
which interact across the newly formed fracture surfaces. This necessitates continual growth
ofthe main crack to maintain a stable overall crack tip region, both in the frontal process zone
and in the following wake region.
Fracture mechanics has also contributed to the understanding of slow, or environmentally
assisted crack growth. It has enabled that phenomenon to be incorporated within the design
process. Slow crack growth occurs whenever the stress intensity exceeds either the
threshold, I~, or the fatigue limit, I~o. Unlike metals, glasses and brittle advanced structural
ceramics are susceptible to environmentally assisted crack extension when under a constant
load or stress. This phenomenon has been referred to as static fatigue. It is different than the
cyclic fatigue phenomena originating from rising R-curve behavior in ceramics and dislocation
phenomena in metals. It is a major factor in many lifetime prediction schemes for Advanced
structural ceramics.
Improved Structural Ceramics
It was noted earlier that ceramics in the 1960s did not have very attractive properties in
terms of strengths, or Weibull moduli. In fact, they were rather poor by today's standards.
Presently, off-the-shelf advanced technical ceramics are available with average strengths
approaching one GPa and Weibull moduli in excess of 20. This is nearly an' order of
magnitude improvement in each of the parameters. Fracture mechanics was a major factor
in these ceramic material improvements as the combination of fracture toughness, fracture
statistics and fractography enabled the identification of critical flaws as processing defects and
subsequently the development of better ceramics. Bowen, Aksay and Lange have been active
in this segment of the ceramic processing field and refer to the flaws as processing related
fracture origins. It has often been possible to relate the first generation of critical flaws to
processing defects. Once these more serious defects were eliminated by improved processing
techniques, then the next level of flaws could be similarly identified and removed. These
successive defect eliminations have obviously been successful as ceramic strengths have
improved substantially. One could make the argument, and probably make it convincingly
that these processing improvements have been the major contribution of fracture mechanics
to advanced technical ceramics.
As a result o f this combination o f fracture mechanics concepts and improved ceramic
processing, today's advanced ceramics are better than ever. It is the opinion o f this author
that they are ofmafficiently high quality for more extensive applications in heat engines, if only
the designs were better and the manufacturing costs were reduced.
Ceramics in Structural Applications
With the significant improvements in the properties of technical ceramics, the advances
in design and the increased knowledge of fracture as a result of fracture mechanics studies
over the past 30 years, there have developed some utilization of advanced technical ceramics
in structural applications. Although not dominantly in heat engines, these applications are
meaningful structural uses of advanced ceramics. One of the most remarkable of these
applications is the use of ceramic turbochargers in passenger cars by both Nissan and Toyota
[11, 12]. Subjected to overload proof testing, there have not been any failures publicized to
date. One can imagine a ready, but perhaps not quite one-for-one translation of some of these
turbocharger design concept to similar applications in high temperature gas turbines. It is not
unreasonable for this to come of fruition in the next decade or two. Ceramic valves have also
made it into some reciprocating engines. Ceramic seals and ceramic ball bearings have also
made significant progress and are now commercial mainstays. It is evident, even obvious, that
there are a number o f niche markets for structural ceramics. As these continue to he
identified, there will be an increase in the use of ceramics and eventually their costs will he
significantly reduced. Hopefully this will occur in a timely, synergistic manner.
Summary and Conclusions
The connection of the application of fracture mechanics to brittle ceramic materials and
the promise of the utilization of ceramics as structural components in heat engines was
established in the 1970s. Contributions of fracture mechanics to understanding the fracture
of brittle ceramics were summarized and the attempted utiliTztion of ceramics in heat engines
was reviewed. Failure to achieve widespread use of ceramics in heat engines was examined.
Fracture mechanics has made major contributions to the understanding of fracture in
ceramics, including the failure criteria, slow crack growth and fatigue and their incorporation
into design methodologies for the utilization of ceramics. Fracture mechanics has also made
significant contributions to the fundamental understanding of the fracture processes in brittle
materials. Particularly relevant is the recognition that the following wake region of the crack
is respons~le for rising R-curves in ceramics and the presence of cyclic fatigue. However,
fracture mechanics was also a contributor to the identification of processing defects in
ceramics and indirectly responsible for many advances in ceramic processing that eliminated
those defects. Progress over the past three decades has been substantial.
Unfortunately, advanced technical ceramics have not achieved widespread use in structural
applications in heat engines. A series of reasons for this was advanced, including the lack of
data bases and appropriate standards for fracture testing, a weakness in the design
methodology for brittle ceramics and the cost of ceramic components. In spite of these
challenges, ceramics have been incorporated into automotive turbochargers, and also used in
seals and as bearings. It is evident that there exist niche applications where ceramics are the
materials of choice. It is anticipated that these types of applications will increase in number
and that ceramics will become utilized in more structural applications with time.
Acknowledgments
The author hesitates to acknowledge individuals for the fear of omitting some, but within
that perspective apologizes in advance to any he may omit. He will forever be indebted to
Dick Hasselman and Fred Lange who helped organized the first "Fracture Mechanics of
Ceramics" conference nearly 30 years ago. He also acknowledges his graduate students who
worked in the field of fracture of ceramics and glasses. He is especially grateful to Mototsugu
Sakai and Ken White with whom he continues to interact on the topic of the fracture of
ceramics and also to George Quinn, Mike Jenkins and Jon Salem who encouraged him to
write this paper and participate in this symposium. All three of them were students in his
classes. Finally, he would like to acknowledge the influence of Bob DeVries who introduced
him to the crystallography of minerals and was instrumental in creating the author's interest
in minerals, materials on which he currently spends much of his time studying the
phenomenon of single crystal cleavage.
References
[1] Bradt, R. C., Hasselman, D. P. H., and Lange, F. F., Fracture Mechanics of Ceramics
Vols. I - II, Plenum Publishing Co.; NY, NY (1974). (This series has > 10 volumes.)
[2] Sakai, M. and Bradt, R. C., "Fracture Toughness Testing of Brittle Materials",
International Materials Reviews, Vol. 38, 1993, pp. 53 - 78.
[3] Miller, J. H. and Liaw, P. K., "Fracture Toughness of Ceramics and Ceramic Matrix
Composites", in ASM Handbook, Volume 8, Mechanical Testing and Evaluation, ASM
International, Materials Park, Ohio, USA, 2000, pp. 654 - 664.
[4] Jenkins, M. G. and Salem, J. A., "Fracture Resistance Testing of Brittle Solids', inASM
Handbook, Volume 8, Mechanical Testing and Evaluation, ASM International, Materials
Park, Ohio, USA, 2000, pp. 665 - 678.
[5] Bradt, 1L C., "Ceramic Heat Engine Programs in the United States', in Proc. of the 2nd
European Symposium on Engineering Ceramics, London, 1987, edit. F. Riley, Elsevier
Pub. Co., London, England, 1989, pp. 229 - 240.
[6] Quinn, G. D., "Properties Testing and Materials Evaluation", Ceramic Engineering and
Science Proceedings, Vol. 5 No. 5-6, American Ceramic Society, Westerville, Ohio, 1984
pp. 298 -311.
[7] McLean, A. F. and Hartsock, D. L , "An Overview of the Ceramic Design Process',
Engineered Materials Handbook, Volume 4, Ceramics and Glasses, ASM
International, Materials Park, Ohio, USA, 1991, pp. 676-690.
[8] Evans, A. G., "Perspective on the Development of High-Toughness Ceramics", Journal
AmericanCeramic Society, Vol. 73, No 2, 1990, pp. 187 - 206.
[9] Savitz, M., "Commercialization of Advanced Structural Ceramics, Part I", American
Ceramic Society Bulletin, Vol. 78, No 1, 1999, pp. 53 -56.
[ 10] Savitz, M., "Commercialization of Advanced Structural Ceramics, Part II", American
Ceramic Society Bulletin, Vol. 78, No 3, 1999, pp. 52 -56.
[11] Matsui, M., Ishida, Y., Soma, T. and Oda, I. "Ceramic Turbocharger Rotor Design
Considering Long Terra Durability', Ceramic Materials and Components for Engines,
editors W. Bunk and I-I. Hausner, DKG Bad Honnef, 1986, pp. 1043-1062.
[ 12] Katano, Y., Ando, Itoh, and Sakai, M., "Applications of Ceramics to Turbocharger
Rotors for Passenger Cars', Journal Engineering far Gas Turbines and Power, Vol.
l15b, No. 1, 1993, pp. 9-16.
Implications for Design and Testing
9 1
Stephen Frelman
Failure from Large Grains in Polycrystalline Ceramics: Transitions in

Fracture Toughness
Reference: Freiman, S., "Failure from Large Grains in Polycrystailine Ceramics:

Transitions in Fracture Toughness" Fracture Resistance Testing of Monolithic and
Composite Brittle Materials, ASTM STP 1409, J. A. Salem, M. G. Jenkins, and G. D.
Quinn, Eds. American Society for Testing and Materials, West Conshohocken, PA, 2001.
Abstract: This paper addresses issues of fracture for materials in which large grains,
exist in a matrix of much smaller grains. Polycrystalline ZnS and ZnSe, used for optical
components, frequently fall into this category. For material microstructures of this
character failure can occur from flaws contained within the isolated large grains so that
the governing fracture toughness is that of a single crystal of the material, rather than that
of the polycrystalline matrix9 I point out that there are currently limitations and
uncertainties associated with the most popular experimental procedure available to
determine the fracture toughness for small crystals. As an alternative to testing, I discuss
the state of our ability to predict fracture toughness based solely on knowledge of lattice
parameters and elastic properties. The historical background and examples of such a
prediction technique are demonstrated. Second, the role of microstructure in governing
the transition from single crystal to polycrystalline values of toughness is shown9 I
postulate that mixed mode fracture mechanics expressions can be used to explain the
transition in fracture toughness9
Keywords: ceramics, fracture toughness, grain size, polycrystalline toughness, single

crystal toughness
Introduction
One important characteristic of ceramic microstructures is that grains are often similar
in size to the flaws that lead to failure. Therefore, depending on the depth of the surface
defects and the grain size, flaw environments can range from one crystal to many
surrounding grains, raising the issue that fracture toughness obtained with large cracks
may not be a valid predictor of the behavior of small flaws. Because typical techniques
for determining fracture toughness employ artificial cracks which can be many times the
size of the grains, the resistance to flaw growth in-service can be quite different than that
measured in the laboratory. Cracks, which encompass many grains, will be more
resistant to growth because of toughening mechanisms that do not apply to smaller flaws.
From the point of view of the user of the ceramic, it is the strength, especially the
minimum in the strength distribution of a component, not the polycrystalline fracture
i Ceramics Division, National Institute of Standards and Technology9Gaithersburg, MD, 20899.

17
Copyright9 ASTM International www.astm.org
toughness, which is important for reliability, making an understanding of the small flaw
behavior particularly crucial. In addition, there is data to indicate that while fracture
toughness can increase with grain size, strength can actually decrease [1-5].
While, fracture initiation and crack growth in ceramics can be either transgranular or
intergranular in nature, this paper will be limited to a discussion of flaws initiating wholly
within a grain, and propagating transgranularly. A number of polycrystalline materials,
e.g. ZnS and ZnSe used as optical components, have microstructures consisting of large
grains imbedded in a fine grain matrix. One of the implications of such a microstructure
is that the determinant of the strength of such components may be the fracture toughness
of one grain, i.e. a single crystal of the material. A number of experimental observations
of single-grain-controlled failure have been made [1-6].
Provided that sufficient size crystals are available, the fracture toughness, Kic, or
fracture energy, (, of such a grain can be experimentally determined in a number of ways.
In many instances, however, one may not have access to sufficient quantities of material
in single crystal form to use most, if not all of these techniques.
The one technique that is most widely applicable to small volumes of material, the
indentation-crack-length method [7], is subject to significant uncertainties, as will be
discussed later. As an alternative, I will attempt to show that models can be employed
which yield realistic predictions of the fracture resistance of single crystals of a wide
variety of brittle materials, and suggest that such models could be used in lieu of
available measurements.
In addition, I will discuss the mechanisms governing the transition from single crystal
to polycrystalline fracture toughness. As would be expected, measured fracture
toughness increases with increasing ratio of crack size/grain size until the level of
polycrystalline toughness is reached. While numerous papers have been written on the
subject of this so-called R-curve behavior, and its relationship to strength, what has been
less extensively discussed is the transition from purely single crystal to polycrystalline
failure, which also involves significant increases in toughness. In this paper I will put
forth a hypothesis as to a mechanism governing the shape and magnitude of the
toughness transition.
Failure from Large Grains in Polycrystalline Ceramics
As mentioned previously, strength is governed by failure from small flaws,

micrometers in extent, while most fracture mechanics procedures involve the use of large
cracks to determine fracture toughness. That results of such large crack tests can give
erroneous predictions of strength is shown by the following example.
Fractographic analysis shows that failure of chemically vapor deposited ZnSe is

initiated at machining flaws within large, isolated, grains of the material (Figure 1)[4].
Assuming that the only stresses acting on the flaws are those applied during testing, the
strength of the material is related to the flaw sizes and fracture toughness, through:
0 = Y-IKicc-'J2 (1)
FREIMAN ON LARGE GRAINS IN POLYCRYSTALLINE CERAMICS 19
Figure 1 Fracture surface of chemically vapor deposited, polycrystalline ZnSe showing

failure from isolated large grains (after Freiman et al. [4]). The arrows in (B)
point to the flaw boundaries.
where Y is a geometric constant, and c is the flaw depth. Or in terms of the fracture
energy, ~/:
o =r-'( 2er y'2c"2 (2)
where E is the elastic modulus 1. One can then plot failure stress as a function of the
inverse square root of the measured flaw size, generating the data shown in Figure 2. The
slope of this curve yields a value of Klc (or 3' ). It can be seen from Figure 2 that the
governing value of Ktc/T is that for a single crystal of ZnSe (~, --0.8 J/m2), and not that
for the polycrystalline material (3.4 j/m2). This difference arises because the critical
flaws are contained within one grain. For most materials, i.e., those in which Kic
(polycrystal) is < about 20x the Ktc (single crystal) [5]. the fact that the growing flaw later
encounters tougher, polycrystalline regions has no measurable effect on the strength. In
fact, I am unaware of experimental evidence to suggest that a crack will arrest at such a
boundary during mechanical loading.
1 The value of E used in the expression will depend on whether one is analyzing a random polycrystalline
or a single crystal environment, as will the relationship between ~/ and K~c.
20 FRACTURE TESTING OF MONOLITHIC/COMPOSITE MATERIALS
What is clear from these results is that despite the fact that the polycrystalline material
surrounding the large grain can be significantly more resistant to crack growth, if the flaw
achieves a critical size within the grain, this increased resistance will not prevent failure.
This statement appears to be true regardless of how close to the large grain boundary the
initial flaw exists. Any possible dynamic effects resulting from a rapidly moving crack
seem to be minimal.
There are two questions raised by the above discussion: 1) How does one determine
the fracture toughness of the large grain (single crystal) of the material, and 2) What are
the mechanisms leading to the single-to-polycrystalline transition? We will attempt to
address both of these questions in the remainder of this paper.
FLAW SIZE {/~m)

ecoo K~O 5O dO aO 20 10
I. . . . . . . I .... i i i I I 15
I00 9 T E N S I L E ~51JRFACE FAILURE, 14

o CORNER FAILURE 13
90 9 GRAIN SIZE AT FAILURE. ORIGIN
8O ~ ~
II
60 ~ 0
4O 9 o '~
O--I I I I I I I I I I I I I I I I I o
o ioo zoo 300
I/JFLAW SIZE t l / , / ~ )
Figure 2 Fracture strength of the ZnSe materials described in Figure 1 as a function of

the measured flaw sizes. Using Equations 1 or 2 one can calculate applicable
values of Ktc or ( ~ ). (after Freiman et al. [4]).
Single Crystal Fracture Toughness - Experiment
If one has access to large enough single crystals of the material of interest, there are
numerous fracture mechanics techniques available to determine Kic (or y ). However, it
is not likely that crystals of large enough size will be available for many materials. In
such a case one is constrained to using the indentation-crack-length_procedure. This
technique involves placing a Vickers indentation in a material and measuring the length
of the surface cracks emanating from the comers of the impression. The expression used
to calculate fracture toughness is given by [7]:
FREIMAN ON LARGE GRAINS IN POLYCRYSTALLINECERAMICS 21
E P
K,c = ( c o n s t a n t ) ( - - f.5 ( ~ ) (3)
By c
where Ktc is the fracture toughness, Hv is the Vickers hardness, P is the indentation load,
and c is the measured crack length.
While this procedure is used extensively, there are numerous problems associated with
it. First, there is the uncertainty over the length of the crack measured in the optical
microscope; any crack opening smaller than the resolution of the microscope will not be
accounted for. Second, the constant in Equation 3 is strictly empirical, determined by a
fit to data for polycrystalline ceramics [7]; uncertainties in its value approach 25%.
Thirdly, there are fundamental uncertainties in the procedure because of the anisotropic
nature of a crystal. The expression for the driving force on the flaw was obtained
assuming an isotropic elastic field. Since the actual elastic modulus in a crystal is very
directional, the driving force, and therefore the accuracy of this expression becomes very
uncertain. All of these factors contribute to a significant concern regarding the accuracy
of this technique, particularly as it applies to single crystals.
What then is the alternative?
Single Crystal Fracture Toughness - Prediction
In principle, if the relationship between bond stress and strain is known, one should be
able to calculate the energy required to separate two surfaces and, thereby, to predict the
fracture resistance of a material. Gilman [8] was the first to develop such a model. He
assumed a sinusoidal function for the attractive stress, o, between the surfaces, as shown
in Equation 5.
o = o msin/hi5 / (4)
a
om is a constant and ~/do is the bond strain, do the equilibrium spacing between the
cleavage planes, and ~5the difference between the strained bond length, d, and do; a was
defined by Gilman as the range of the attractive forces. Given this function, Gilman
derived the following expression for the energy needed to create the two surfaces:
E a 2
(5)
where, T is the fracture energy, i.e., cleavage energy, and E is Young's modulus. E is
obtained from the definition of Young's modulus: E = do/de*e=o, i.e., where the strain
approaches zero. A key issue is the value one assigns to a, i.e., where does complete
fracture of a bond occur? Gilman assumed a to have a value of the combined radii of the
atoms on the cleavage planes, claiming that this represented the extent of bonding
between the surfaces.
Nevertheless, Gilman used Equation 5 to obtain reasonable agreement between

calculated values of fracture energy for LiF, MgO, CaF2, etc., and those determined
experimentally. Using Gilman's methodology, Becher and Freiman [9] also showed
excellent agreement between measured fracture energies of KBr, KCI, NaCI, BaF2, SrF2,
and CaF2 and those calculated using Equation 5.
White etal. [10] calculated the fracture energy of a number of compounds having the
diamond cubic structure using Equation 5. However, rather than taking a to be the
atomic radii, they arbitrarily assumed that fracture would occur at a strain of 50%, i.e., a
= d d 2 . Using this approach, White et al. [10] also showed consistency between measured
and calculated values of T for a range of compounds. Finally, Freiman and Baker [ 11 ]
showed that similar calculations could be performed for heavy metal fluoride glasses,
which have no long-range order. In this case do was taken as the Zr - F distance in the
glass, with a ~ do~2. Again, good agreement between model and experiment was
achieved.
One fundamental question is what is the proper relationship between bond stress and
strain, and how will the use of alternative expressions to Equation 4 affect the prediction
ofT. It can be shown [12] that a general functional expression for y in terms of the force
law can be obtained in terms of the crystal structure and elastic modulus by making some
broad assumptions about the fracture process. These assumptions are (1) the process is
reversible, (2) no energy is lost to processes other than bond rupture, e.g., there is no
plastic deformation or phonon emission, (3) there are no environmental effects, and (4)
there are no microstructural contributions. The derived expression is:
E d o f= g ( e ) de. (6)
7= 2 Jo g'(O)
where e is the bond strain (/8~do) and g(e) is a load-displacement function for a bond on
the cleavage plane. All bonding functions will have the same general form. Namely, (1)
they will have an initial slope (g'(0)) which can be equated to Young--s modulus; (2)
there will be a maximum value of stress, ore; (3) there will be a range over which the
bonding forces extend, and; (4) there will be some area under the o-e curve, from which
we can calculate 7- For most bonding laws only 2 of the first 3 parameters are
independent. The equivalence of the initial slope with E leaves the maximum value of
the stress, o,,, and the stress range, a, undetermined. However, each of these two
parameters can be expressed in terms of the other.
Equation 6 can be recast as in terms of strain by taking
2Ee., (7)
o = sin( --- )
2E~
e. = 8~do and e.,,, = a/2do, i.e., em is the strain defining the attractive stress range.
Combining Equations 6 and 7, the following expression for 7 results.

(8)
7 =--~ Edo
which is identical in form to that derived by Gilman [8] (Equation 5). Other force laws
can be shown to lead to the same functional form for 7, i.e., they differ only in the value
of the constant [12].
To make calculations, one uses the actual values of E and do for the material of
interest; em is undetermined and must be evaluated independently. A value of em= 0.25
(equivalent to the value of a/do = 0.5 seems to give self-consistent results (Figure 3). The
experimental data was obtained using established fracture mechanics tests (usually
double cantilever beam) carried out on single crystals with the crack propagating along
the easiest cleavage plane. Despite the self-consistency in the predictive values of 7 the
process is not completely satisfying since there is an arbitrariness involved in choosing a
value for the strain at maximum stress. In addition, we do not know the reasons for the
discrepancies between model and experiment for materials such as Si. Nonetheless,
given the uncertainties in the currently available experimental procedures discussed
earlier, one might consider such a prediction as a reasonable alternative in the estimation
of crack growth resistance.
Comparison o f Measured a ~ c l C a l c u l a t e d
F~cture Ene:~ei
1.5o
1.00
G ~
r~ 0.50
~C1
0.(I(I '''' .... ' .... ' .... ' ....

0.00 0.50 1.00 1.50 2.00 2,50
M ~ {J/m~
Figure 3 Plot of measured and calculated values of(. The calculated ( was obtained
from Equation 8, using a value of em = .25.
Single to Polycrystalline Transition
I also wish to address the question of the mechanism(s) governing the transition to the
increased toughness of the polycrystalline material surrounding a large grain.
Experimentally, one observes that the transition from single to polyerystalline fracture
resistance can occur at flaw/grain size ratios ranging from 1 to 6, i.e., from immediately
at the boundary of the initiating grain to many times this value. As an example, Figure 4
shows the change in measured fracture toughness of ZnSe for a range of flaw sizes.
Transition from Single Crystal to Potycrysio~Sne Controlled Frocture

1.2 I f I I I I I I I
ZnSe o 3 POint Flexure
9 Frocture Uechenics Teat
1.0
P~176 ".j "..e . . . . . .
5
0.8
~ 0.6
0.5
0.2 0.25
0.0 t i
0 1 2 3 4 5 6 7 8 9
Row Size/Groin Size
Figure 4 Transition in fracture toughness in CVD ZnSe as a function of the ratio offlaw
size to grain size (after Rice et al., [1]).
Figure 5 shows transitions for a number of different materials. Another example of a

transition from single crystal to polycrystalline toughness is shown in Figure 6 for ZnS
[6]. This figure represents data from a controlled study employing indentation-induced
cracks. In this case the transition is very abrupt, occurring essentially at the boundary of
the large grain in the material. The cleavage plane(s) in grains surrounding the large
grain in which fracture initiated will be misoriented with respect to it as well as to one
another. Another factor will be the number of cleavage planes on which the crack will
grow, and their relative resistance to such growth.
O2 0.4 0.6 0.8 LO 12 S.4 t.6 t.8 2.0

, i , I I , , i v
FLAW O[PTH
GRAIN DI[PT H
Figure 5 Transition in fracture energy for a number of materials as a function offlaw

size~grain size (after Rice et al., I21).
1.4
A
1.2
<~A o
~ 1.0- o ,
# o oo
a ~ o o o
~ o
.g~ 0.8 a
D .
1
~ 0.6
o o LG: Non-Indent
o
o 0.4, r--e-~ a LG: Indent
o 08 o A SG: Non-Indent
0.2~- o 9 SG: Indent
0.1)
.01 .~ ; 1'0 loo 1000
c/G
Figure 6 Transition in fracture toughness of ZnS as a function of flaw size, C, and grain
size G. LG and SG refer to large grain size and small grain size material,
respectively, (after Yoder 16].
There are a number of models explaining the increased fracture toughness of

polycrystalline ceramics in terms of mechanisms such as crack deflection [ 13, 14] or
microcracking [l 5-16]. However, these models do not quantitatively predict the steep
rise in Ktc observed in the transition from single crystal to polycrystalline fracture such as
that observed for ZnS. From a crack perspective, we can view the transition
schematically as shown in Figure 7. When the crack is contained within the grain it
propagates along a cleavage plane. Once the growing crack reaches the large grain
Figure 7 Schematic showing the relationship between a flaw growing in a large grain
under stress, and the segments of the growing crack in the adjoining smaller
grains.
boundary, in order for it to continue to propagate on this plane in the polycrystalline

region, segments o f the crack must alter their orientation to conform to that of the grain
over that segment. The stress intensity factor needed to extend such misaligned segments
can be calculated fi'om mixed-mode models of crack growth. One such expression
assumes a coplanar extension o f the crack under mixed mode loading [17]. While the use
o f such an expression may be questioned for isotropic materials, its use under this
condition in which the crack is constrained to the cleavage plane seems reasonable. The
expression governing such a crack is:
r K 971~
(9)
where Klc is the fracture toughness and/s and K1; are the mode I and mode II stress
intensity factors whose values can be calculated from Equations 10a and b:
K; = ~ ; o v 1/2 cos 2 0 (10a)
(10b)
Kn = ar'2Hol71/2sin 0 c o s 0
In these expressions, cris the applied stress, c is the crack radius, f~; = 2m/n ~ =1.35, f);; =
4/n~(2-v) = 1.29, where m = 1.2 for semicircular cracks, and v is Poisson's ratio, taken to
be 0.25. 0 is the angle between the applied stress and the crack plane. Combining the
above equations and taking f2; ~f~;l. then:
Kzc = Ktc(sc)[l + tan 2 0] ~/2 (11)
where Ktc(so is the fracture toughness of a single crystal o f the material.

Assuming that there is only one crack path, namely the single cleavage plane, then in a
randomly oriented polycrystalline microstructure an average 0 will be determined by the
grain size. For self-similar grain shapes, tbe area of fracture surface will be independent
o f grain size. However, finer grain size material will give rise to smaller values o f 2.
While other possible toughening mechanisms such as crack bridging, formation of
subsidiary cracks, grain boundary fracture, etc. could come into play, Equation 11
provides the minimum extra energy that would be required over and above that needed to
grow a crack through a single grain. Taking plausible values of 2, one can calculate
expected values o f Ktc/K/cesc) as shown in Table 1.
Table 1 Ratio of Kict Ktctsc~ as a function of misorientation angle, n, of cleavage plane

segments.
~) K_~ctsc~
0 1
30 1.2
60 2
70 2.9
80 5.7
It appears from the calculated values of Ktct Kictsc~ shown in Table 1 that mixed mode
fracture within each grain as expressed in Equation 11 can account for the observed
increases in toughness as a crack grows out of an isolated large grain.
Summary
In light of the fact that fracture of a number of important ceramic materials initiates
within one grain, the fracture toughness of a single crystal is the key parameter needed to
predict strength. Because of limitations to the availability of accurate experimental
techniques applicable to small single crystals, I suggest that the use of predictive models
might serve as guides to fracture resistance. I have shown that these models follow a
general form for the relationship between fracture energy and the fundamental crystal
structure and bonding laws for brittle materials independent of any specific stress-strain
relationship, which will allow an estimate of fracture energy to be made. Namely:
v = k eL Edo (12)
Where k is a force-law-specific constant. It is hypothesized that k will vary by an

amount less than the uncertainties in the experimental measurements themselves. I have
used this relationship to show that reasonable values of fracture energy can be calculated
for a number of single-crystal materials.
I have also put forth a model based upon mixed -mode fracture that can account for
the single crystal to polycrystalline transition in fracture resistance.
Acknowledgements
I would like to acknowledge the many helpful discussions with Jack Mecholsky and
Roy Rice as well as their editorial assistance in this manuscript. The helpful comments
of George Quinn are also appreciated.
References
[1] Rice, R. W., Freiman, S. W., Pohanka, R. C., Mecholsky, J. J., and Wu, C. C.,
"Microstructural Dependence of Fracture Mechanics Parameters in
Ceramics," Fracture Mechanics of Ceramics, 4, Edited by R.C. Bradt, D.P.H.
Hasselmann, and F.F. Lange, 1978, pp. 849-875.
[2] Rice, R. W., Freiman, S. W., and Mecholsky, J. J. "The Dependence of Strength-
Controlling Fracture Energy on the Flaw-Size to Grain-Size Ratio," Journal
of the American Ceramic Society, 63, No. 3-4, 1980 pp. 129-136.
[3] Rice, R. W., Freiman, S. W., and Becher, P. W., "Grain-Size Dependence of
Fracture Energy in Ceramics: I, Experiment," Journal of the American
Ceramic Society, 64, [6] 1981, pp. 345-50.
[4] Freiman, S. W., Mecholsky, J. J., Jr., Rice, R. W., and Wurst, J. C., "Influence of
Microstructure on Crack Propagation in ZnSe," Journal of the American
Ceramic Society, 58, No. 9-10, 1975, pp. 406-408.
[5] Singh, J. P, Virkar, V., Shetty, D. K, and Gordon, R. S., "Strength-Grain Size
Relations in Polycrystalline Ceramics," Journal of the American Ceramic
Society, 62, No. 3-4, 1979, pp. 179-183.
[6] Yoder, P. L., M.S. Thesis, Pennsylvania State University, 1989
[7] Anstis, G. R., Chantikul, B. R., Lawn, B. R., and Marshall, D. B., "A Critical
Evaluation of Indentation Techniques for Measuring Fracture Toughness: I,
Direct Crack Measurements," Journal of the American Ceramic Society, 64
[9], 1981, pp. 533-538.
[8] Gilman, J. J., "Direct Measurements of the Surface Energies of Crystals,"

Journal of Applied Physics, 31 (12) 1960, pp. 2208-2218.
[9] Becher, P. F. and Freiman, S. W., "Crack Propagation in Alkaline-Earth

Fluorides," Journal of Applied Physics, 49(7), 1978, pp. 3779-3783.
[10] White, G. S., Freiman, S. W., Fuller, E. R., Baker, T. L. "Effects of Crystal
Bonding on Brittle Fracture," Journal of Materials Research, 3, No. 3, 1980 pp.
491-497.
[11] Freiman, S. W. and Baker, T. L., "Effects of Composition and Environment on

the Fracture of Fluoride Glasses," J. of the American Ceramic Society, 71, (4)
1988, C214-216.
[12] Freiman, S. W., Fuller, E. R. Jr., and G. S. White, "Predicting the Fracture
Resistance of Brittle Materials," to be published.
[13] Faber, K. T. and Evans, A. G. "Crack Deflection Processes---l. Theory," Acta

Metallurgica, 31, No. 4, 1983, pp. 565-76.
[14] Faber, K. T. and Evans, A. G. "Crack Deflection Processes---ll, Experiment"

Acta Metallurgica, 31, No. 4, 1983, pp. 577-84.
[15] Fu, Y. and Evans, A. G., "Microcrack Zone Formation in Single Phase
Polycrystals," Acta Metallurgica, 36 1982, pp. 1619-25.
[16] Evans, A. G. and Faber, K. T., Crack-Growth Resistance of Microcracking

Brittle Materials," Journal of the American Ceramic Society, 67, [4] 1984, pp.
255-60.
[17] Paris, P. C. and Sih. G. C., "Fracture Toughness Testing and Its Applications,"
ASTM STP 381, 1965.
Lynn M. Powers, 1 Jonathan A. Salem, ~ and Aaron S. Weaver~
Stresses in Ceramic Plates Subjected to Loading Between Concentric Rings
Reference: Powers, L. M., Salem, J. A., and Weaver, A. S., "Stresses in Ceramic
Plates Subjected to Loading Between Concentric Rings," Fracture Resistance Testing
of Monolithic and Composite Brittle Materials, ASTM STP 1409, J. A. Salem, G. D.
Quinn and M. K Jenkins, Eds., American Society for Testing and Materials, West
Conshohocken, PA, 2002.
Abstract: Plates of both square and round geometries are tested to determine the strength
and fatigue properties o f ceramics under multiaxial stresses. The necessary tolerances
required for accurate testing are not well documented in published literature. The stresses
in square and round plates subjected to loading between concentric rings were
investigated via the finite element analysis method. In particular, the effects of load ring
concentricity and the amount of the plate overhanging the support ring were investigated.
Also, the errors generated by testing square plates between circular rings were
investigated. The analyses were performed in support o f an effort by the American
Society for Testing and Materials Committee C28 on Advanced Ceramics to standardized
multiaxial testing of ceramic plates.
Key Words: ceramics, multiaxial strength, stress, plates
The strength of brittle materials such as ceramics, glasses and semiconductors is a function
of the test specimen size and the state of applied stress [1]. Engineering applications of such
materials (e.g., ceramics as heat engine components, glasses as insulators, silicon and
germanium as semiconductors) involve components with volumes, shapes, and stresses
substantially different from those of standard test specimens used to generate design data.
Although a variety of models [2] exist that can use conventional, uniaxial test specimen data to
estimate the strength of large components subjected to multiaxial stresses, it is frequently
necessary to measure the strength of a brittle material under multiaxial stresses. Such strength
data can be used to verify the applicability of design models to a particular material or to
mimic the multiaxial stress state generated in a plate-like component during service.
1Research engineer, Case Western Reserve University, NASA Glenn Research Center, MS 49-7,
Cleveland, OH, 44135.
2Research engineer, Life Prediction Branch, NASA Glenn Research Center, MS 49-7, Cleveland, OH,
44135.
30
POWERS ET AL. ON STRESSES IN CERAMIC PLATES 31
Further, the strength of brittle materials is sensitive to machining and handling of the
test specimens. This can lead to spurious chips at the test specimen edges, which in turn
can induce failure not representative of the flaw population distributed through the
material's bulk. In the case of a plate subjected to bending, the stresses developed are
lowest at the edges, thereby minimizing spurious failure.
For components that are subjected to multiaxial bending, three different loading
assemblies, shown schematically in Figure 1, can be used to mimic biaxial stresses by
flexing circular or square plates: ball-on-ring (B-O-R), ring-on-ring (R-O-R), or pressure-
on-ring (P-O-R). The R-O-R and the P-O-R are
preferred because more of the test specimen
volume or surface area is subjected to larger
stresses. However, significant frictional or
wedging stresses associated with the loading ring
can be developed in the highly stressed regions
of the R-O-R specimen [3,4]. These stresses are
not generated in the P-O-R configuration and the
P-O-R configuration has been investigated for
testing of both isotropic [5-8] and anisotropic
materials [9,10]. Although the P-O-R has been
experimentally determined to produce results in
agreement with plate theory, it is not as
convenient as the R-O-R for elevated
temperature testing.
A wide variety of both square and round
R-O-R loaded plates are used to determine
fatigue and strength properties, however, the
necessary tolerance and geometric ratios for Figure l-Multiaxial bending test
accurate testing are not well-documented in the configurations a)ball-on-ring, b)
literature. The objective of this work was to ring-on-ring, c)pressure-on-ring.
investigate the stresses in square and round plates
subjected to loading between concentric rings. The following aspects were investigated
via the finite element method: the amount of the plate overhanging beyond the support
ring; the errors generated by testing square plates between circular rings; the effect of the
plate thickness on the wedging stresses; and the effect of concentricity between the load
and support rings. The analyses were performed in support of an effort by the American
Society for Testing and Materials Committee C28 on Advanced Ceramics to standardized
multiaxial testing of ceramic plates.
Theory
s
A schematic of the ring-on-
ring specimen is shown in Figure
2. The theoretical solution for the
deflection at the center of the
r
Figure 2-Geometry of a ring-on-ring test.
plate is given by [11]
w= 1+ 1+ In (1)
2nEt 3 2(I+v)R 2 J [
where P is the applied load, E is Young's modulus, v is Poisson's ratio, and the geometry
o f the disk in shown in Figure 2. The geometry of the R-O-R configuration consists o f its
radius, Rn, thickness, t, load ring radius, RL, and the support ring radius, Rs. This
geometry may also be described by the respective diameters, D, DL and Ds. The radial
and tangential stresses in the center section on the tensile surface are equal and may be
expressed as
3e r (2)
For all of the plates studied herein, the ratio of Rs to RL is 2.

Another important measure to consider when testing brittle materials is the
probability of failure of the test specimen. Since the plates are subjected to bending, the
stresses are highest at the surface. Thus, surface failure analysis was used. The probability
of failure for a surface element is
Pjs = 1-exp [ - BA] (3)
where Pys is the probability of failure, B is the risk o f rupture intensity, and A is the area
for that element. The probability o f failure for a component or plate is found by
integrating over all of the elements in a component or plate. The risk of rupture intensity
is written
(4)
27t do c,-
where trema~is the maximum effective stress, Ns is the crack density function, to is the length
of an angle, or, projected onto a unit radius circle in principal stress space containing all of the
crack orientations for which the effective stress is greater than or equal to the critical stress and
5". is the stress state. Since the stress state is equibiaxial inside the load ring, Y~= (tr,~r), Oremax
= tr, and to/2~r = 1. Equation 4 simplifies to
'5'
where kBs is the Batdorf crack density coefficient, eros and ms are the Weibull parameters. The
Weibull parameters are determined experimentally.
Finite Element Analysis
One o f the key elements of the ASTM draft standard on Monolithic Equibiaxial
Flexure Strength Testing of Advanced Ceramics at Ambient Temperatures is setting up
the guidelines for specimen geometry and load. The stresses and deflections generated by
these tests are subject to the limitations o f linear plate theory. Many o f these limitations
have been determined for ceramic equibiaxial plates through trial and error or experience.
This paper will confirm these ideas using finite element modeling of ceramic plates o f
various geometry and loads. The modeling in this study was completed using the ANSYS
program. The plates were assumed to have an elastic modulus of 300 GPa and a Poisson's
ratio of 0.2. The load rings for the contact analysis are steel with an elastic modulus o f
200 GPa and a Poisson's ratio of 0.3. The friction coefficient between the load and
support ring and the plate was 0.05.
The first issue addressed was to determine the load levels appropriate for linear plate
theory. To accomplish this, the contact between the load rings and the ceramic plate was
modeled. The mesh for this model is shown in Figure 3. An axisymmetric model was
used which contained 2518 8-node quadrilateral (PLANE82), 80 3-node target surface
(TARGET169) and 128 3-node contact (CONTACT172) elements. The boundary
conditions applied included the pressure on the load ring, the rigid support ring and the
coupled displacement on the load ring. As a result of this coupling, these nodes are forced
to take the same displacement in the specified nodal coordinate direction. Both the radial
and axial directions are coupled individually.
Figure 3-Axisymmetric mesh of the plate and ring assembly.
Using the model shown in Figure 3, the load vs. displacement curve was constructed for
the plate. The geometry of the plate was RD = 25mm, Rs = 20mm, RL = 10mm, and t =
2mm. The normalized load vs. displacement/thickness curve is shown in Figure 4. The dashed
and solid lines represent linear plate theory and the contact analysis, respectively. Linear plate
theory is assumed to the applicable where these two lines are coincident.
Where linear plate theory is not applicable, the stress state is no longer predictable
by equation (2). The maximum stress is not constant at the surface inside the load ring. The
stress is maximum under the load and has been shown to be as high as 50% larger than the
stress at the center [12]. Figure 5 shows the maximum stress in the plate as a function
Nm
12
r~
%, 10
X
Contact ~
"0
m 6
0
-1
'0| 4
....
m
E 2
0
z 0
~ "~- Theory
0 0.5 1 1.5
Deflection/Thickness
Figure 4-Normalized load as a function o f deflection.
1.5
1.4
E 1.3
~ 1.2
"~ 1.1
1
0 0.5 1 1.5
Deflection/Thickness
Figure 5-Ratio o f the maximum to center stress in the plate as a function o f deflection.
of the deflection. This figure is consistent with Figure 4. The maximum-stress to center-stress
ratio is a minimum and constant at deflections lower than '/, the thickness of the plate. All
subsequent analyses performed in this paper are done in the range of linear plate theory. This
is representative of deflection/thickness of less than '/4.
The mesh used for the modeling of the circular plates is shown in Figure 6. This model is
axisymmetric using PLANE82 (quadrilateral with midside nodes) elements. The boundary
conditions applied were the pressure load and the support ring. The mesh used for the square
plates is shown in Figure 7. A quarter symmetry model was used with SOLID95 (bricks with
midside nodes) elements. Boundary conditions applied were the pressure and support as shown
in Figure 7. The number of elements used for each model varied depending on the geometry of
the plate.
Figure 6-Axisymmetric mesh for circular plates.
Figure 7-Quarter symmetry mesh for the square plate.
Results
The radial stress distribution for the round and square plates are shown in Figure g.
The stress on the tensile surface is maximum and constant inside the load ring and
decreases toward zero at the support. The radial stress outside the support is nominally
zero. However, the tangential stress outside the support is not zero and is a concern at the
edge where flaws may cause failure. These flaws can be more severe than those inherent
on the surface of the disk. As a result, failure of the plate may occur at the edge instead o f
inside the load ring as expected.
To ensure that the plates ultimately fail from flaws originating from inside the rings,
test specimens should be designed so that the edge stresses are minimized. One of the
ways to accomplish this is to increase the amount o f material outside the support ring.
The effect of overhang on the edge stresses is shown in Figure 9 for round plates. As the
amount of overhang increases the tangential stress at the edge decreases. The stress
decreases as the ratio o f the support ring to thickness increases. For thin (Ds/t=-30) plates,
the edge stresses are lower than for thick (Ds/r=-lO) plates.
Figure8-Radial stress distribution in the a) round and b) square plates.
0.6
0.5
==
"= 0.4 DJr
L.
0.3
~Q2
"o
0.1 30
0 2 4 6
(D-Ds)12t
Figure9-Ratio of the tangential edge tO center stress as a function of overhang for

circular plates.
POWERSETAL.ONSTRESSESINCERAMICPLATES 37
0.6
0.5
W
m
.=
0.4 ~ . DJt
W
IL
0.3 10
O
0.2
Q " "" 30
0.1
0 i i
0 2 4 6
(D-DJI2t
Figure 1O-Ratio of the edge to center stress as a function of overhang for square plates.
1.04
W
=2 1.02
"o 0.98
0
.__.
E 0.96
0
z 0.94
0 0.2 0.4 0.6 0.8 1
~RL
Figure 11-Radial stress as a function of position and thickness for round plates.
The same is true for square plates as shown in Figure 10. As the amount o f overhang
increases the tangential stress at the edge decreases. The stress decreases as the ratio o f
the support ring to thickness increases. Once again, for thin (Ds/t=30) plates, the edge
stresses are lower than for thick (Ds/t=l O) plates.
38 FRACTURETESTINGOF MONOLITHIC/COMPOSITEMATERIALS
The effect of the plate thickness on the stress under the load ring is shown in Figure
11. The normalized stress is shown as a function of radial position for circular plates. At
or near the center of the plate (r/RL=O) the stress is equal to its theoretical value as given
from Equation 2. Inside the load ring at (r/RL=I), the stress is slightly higher. The
difference between the center and the stress inside the load ring changes with plate
thickness. As the thickness of plate increases, the maximum to center stress ratio in the
plate increases. The radial stress as a function of position and overhang is shown in
Figure 12. The ratio of the maximum to center stress is not a function of overhang.
1.04
(D-D3/2t
9 1.02 --1
w
--2
1.00
~3
--4
N 0.98
, n
~5
~6
~ 0.96
o
0.94
0.0 0.2 0.4 0.6 0.8 1.0
Figure 12-Radial stress as a function of position and overhang for round plates.
Another factor which may cause the stress distribution in circular plates to change is the
load ring concentricity. If the two rings, load and support, are not in perfect alignment, the
stress field contains perturbations. This becomes important when testing very small plates for
the electronics industry. The tolerances for the manufacture of testing equipment may be larger
than acceptable when compared to the size of the specimen. For example, consider a load ring
which is out of alignment with the support ring by 1% of the load ring radius. A schematic of
this geometry is shown in Figure 13 where the distance ~ is a measure of the shift between the
two rings.
The radial stress distribution inside the load ring for a plate with perfect alignment of the
rings and a misalignment of 0.01RL is shown in Figure 14. The Ds/t ratio for this plate was 20
and the ratio Ds/Dt = 2. Figure 14 a) shows stresses in a plate with concentric toad and
support rings. The stress distribution is axisymmetric. The stress distribution for a ~ of O.OIRt.
is shown in Figure 14 b). The stresses are no longer axisymmetric.
/ support ring ~ \
\
/' / load1 ring \
8
Figure 13-Schematic of the circular plate with misaligned load and support rings.
Ceramic plates are known to fail catastrophically from flaws or cracks inherent in the
material or from their processing. As a result, the risk of rupture distribution is crucial to
understanding the failure behavior of these plates. The risk of rupture plots are shown to
demonstrate that a small perturbation in the stress field may result in larger dispersions in
the risk of rupture intensity. These maps are made to indicate, in a specimen or
component, where failure is most likely to originate.
The risk of rupture for the stress distributions shown in Figure 14 is shown in Figures
15 and 16 for Weibull moduli of 10 and 20, respectively. The risk of rupture intensity is
determined on an element basis as described in Equation 4. The Weibull parameters used
were chosen to generate a risk of rupture intensity B=I for an equibiaxial stress state on
the surface of 100 MPa using Equation 5. These parameters are given in Table 1.
The stress and risk of rupture intensity distribution should be constant inside the load
ring. A 1% error in concentricity results in a 2% error in the stress within the load ring.
However, the risk of rupture is in increased by -30% near the load ring for m~=20 and
-15% for m~=10.
Table 1 - Weibull parameters[or the surface analysis.
BatdorfCrack Density Coefficient WeibullModulus, ms Scale Parameter, G0s,

MPam TM
4.063 10 115.05
5.625 20 i09.02
Figure 14-Radial stress distribution in the a) concentric and b) misaligned plate.
Figure 15-Risk of rupture for the a) concentric and b) misaligned plate with ms =10.
Figure 16-Risk of rupture for the a) concentric and b) misaligned plate with ms=20.
Square plates are often used in the place of circular ones in ceramic material testing.
The stress distribution in square plates has been proposed to be similar to that of round
plates with an adjustment for the specimen size. The effective diameter of a square plate
hag been proposed to be equal to the edge of the square plate or equal to the average of
the edge and diagonal length [13]. These stress distributions are shown in Figures 17 and
18, respectively. The radial stress distribution in thin square plates (Ds/t=30) as a function
of position and overhang is shown in Figure 17. The effective diameter is assumed to be
equal to the edge length of the square plate. This results in an underestimation of the
stresses in the thin square plates by 2 to 4%.
The radial stress distribution in thin square plates (Ds/t=30) as a function of position
and overhang is shown in Figure 18. For this case the effective diameter is equal to the
average of the plate edge length and the length of the diagonal. Using this correction
factor does not improve the results for thin square plates with a variety of overhang ratios.
1.04
W
E 1:02
(D-D~)/2t
m
m
"O 1.00
.m
"O 0.98 .-
O
m
0.96 1
O
r-
0.94
0.0 0.2 0.4 0.6 0.8 1.0
r/RL
Figure 17-Radial stress distribution in a thin square plate as a function of overhang for
an effective diameter equal to the edge length of the plate.
(D-D,)/2t
= 1.04
== 6
1.02 5
9~ 1 4
n- 0.98 3
"O
.~ 0.96 7
0.94
0
z 0.92
0 0.2 0.4 0.6 0.8 1
r/RL
Figure 18-Radial stress distribution in a thin square plate as a function of overhang for
an effective diameter equal to average of the edge length and the diagonal of the plate.
1.04
i
9~ 1.02 (D- ,)/2t
1.00
0.98
0.96
:~ 0.94
o
t,-
0 0.2 0.4 0.6 0.8 1
rto~
Figure 19-Radial stress distribution in a thick square plate as a function of overhang for
an effective diameter equal to the edge length o f the plate.
1.15
(D-Ds)/2t
u) 1.1 -6
.!
"o 1.05
"o
9 1
N
=m 0.95
Ik.
o
z 0.9
0 0.2 0.4 0.6 0.8 1
Figure 20:Radial stress distribution in a thick square plate as a function of overhang for
an effective diameter equal to average of the edge length and the diagonal of the plate.
The radial stress distribution in thick square plates (Ds/t=l 0) as a function of position
and overhang is shown in Figure 19. The effective diameter is assumed to be equal to the
edge !ength of the square plate. This results in a~aunderestimation of the stresses in the
thick square plates with little overhang. The peak stress in thick plates with a larger
overhang may be overestimated. The radial stress distribution in thick square plates
(Ds/t=-10) as a function of position and overhang is shown in Figure 20. For this case the
effective diameter is equal to the average of the plate edge length and the length of the
diagonal. Using this correction factor does not improve the results for thick square plates.
Errors as large as 12% result.
Conclusion
A biaxial test for ceramic materials is necessary for the understanding of the behavior
of these materials under multiaxial stress conditions. The ring-on-ring test generates an
equibiaxial stress inside the load ring. However, restrictions on this test are necessary in
order to ensure that the stress and displacement may be calculated using linear plate
theory. This theory is followed in general as long as the deflection is not greater than 88
the thickness of the plate.
Overhang and thickness also may affect the overall stress state. The stress state on the
tensile surface inside the load ring is in theory equibiaxial and constant. In practice,
contact stresses are present under the load ring. For round plates with Ds/t between 10
and 30 and (D-Ds)/t between 1 and 6, the stresses estimated with closed form solutions
are within of 2%. Also, to ensure that failure takes place inside the load ring and not at
the edge, the effect of overhang on the edge to center stress ratio was evaluated. The
tangential stresses at the edge of round or square plates decrease from -50% of the
maximum stress to -25 % as the overhang increases from 1 plate thickness to 6 plate
thicknesses.
An analysis of the square plates geometry reveals that the stress at the center of the
tensile surface may not be predictable. For thin plates (Ds/t = 30) with small overhang
((D-Ds)/t = 1), the square-to-round plate approximation results in overestimates of the
stresses by -4%. For thick plates (Ds/t = 10) with large overhang ((D-Ds)/t = 6), the
square-to-round plate approximation results in underestimates of the stresses by -10%.
The overall test setup is important for accurate measurement. A 1% error in
concentricity results in a 2% error in the stress within the load ring. However, the risk of
rupture is in increased by -30% near the load ring.
References
[1] Weibull, W., "A Statistical Theory of the Strength of Materials," lngeniors Vetenskaps
Akademien Handlinger, No. 151, 1939.
[2] Batdorf, S. B. and Crose, J. G., "A Statistical Theory for the Fracture of Brittle Structures
Subjected to Nonuniform Polyaxial Stresses." Journal of Applied Mechanics, Vol. 41,
No. 2, 1974, pp. 459-464.
[3] Adler, W. F. and Mihora, D. J., "Biaxial Flexure Testing: Analysis and Experimental
Results," in Fracture Mechanics of Ceramics, Vol. 10, R. C. Bradt, D. P. H.
Hasselman, D. Munz, M. Sakai, and V. Shevchenko, Eds., Plenum Press, New

York, 1991, pp. 227-246.
[41 Fessler, H. and Fricker, D. C., "A Theoretical Analysis of the Ring-On-Ring Loading
Disk Tests, "Journal of the American Ceramic Society, Vol. 67, No. 9, 1984, pp.
582-588.
[5] Rickerby, D. G., "Weibull Statistics for Biaxiai Strength Testing," in Fracture 1977,
Vol. 2, ICF4, Waterloo, Canada, June 19-24, 1977, pp. 1133 -1141.
[61 Shetty, D. K., Rosenfield, A. R., Duckworth, W. H., Held, P. R., "A Biaxial Test for
Evaluating Ceramic Strengths," Journal of the American Ceramic Society, Vol.
66, No. 1, 1983, pp. 36-42.
[7] Salem, J. A. and Jenkins, M. G., "A Test Apparatus for Measuring the Biaxial
Strength of Brittle Materials," Experimental Techniques, V 23, No 4, 1999, pp.
19-23.
[8] Salem J. A., and Adams, M., "The Multiaxial Strength of Tungsten Carbide,"
Ceramic Engineering and Science Proceedings, Vol. 20, No. 3, 1999, pp. 459-
466.
[9l Salem, J.A. and Jenkins, M. G., "Biaxial Strength Testing of Isotropic and
Anisotropic Monoliths" in Multiaxial Fatigue and Deformation: Testing and
Prediction, ASTMSTP 1387, S. Kalluri and P. J. Bonacuse, Eds., American
Society for Testing and Materials, West Conshohocken, PA, 2000, pp. 13-25.
[10] Salem, J. A., Manderscheid J. M., and Jenkins, M. G., "Multiaxial Strength Testing
Of Brittle Single Crystals," Ceramic Engineering andScience Proceedings, Vol.
21, Issue 3, 2000, pp. 83-90.
[I I] Vitman, F. F. and Pukh, V. P., "A Method for Determining the Strength of Sheet Glass,"
Aavodskaya Laboratoriya, 19, 1963, pp. 863-867.
[12] Kao, R., Perrone, N., and Capps, W., "Large-Deflection Solutions of the Coaxial-Ring-
Circular-Glass-Plate Flexure Problem," Journal of the American Ceramic Society, Vol.
54, No. 1 I, 1971, pp. 566-571.
[13] Nemeth, N. N., Powers, L. M., and Janosik, L. A., "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.
Fracture Toughness Standardization
Michael G. Jenkins, 1 Jonathan A. Salem, 2 George D. Quirm, 3 and Isa Bar-On 4
Development, Verification, and Implementation of a National Full Consensus

Fracture Toughness Test Method- Standard for Advanced Ceramics
Reference: Jenkins, M. G., Salem, J, A., Quinn, G. D., Bar-On, I., "Development,
Verification, and Implementation of a National Full Consensus Fracture
Toughness Test Method Standard for Advanced Ceramics," Fracture Resistance
Testing of Monolithic and Composite Brittle Materials, ASTM STP 1409, J.A. Salem,
G. D. Quinn, and M. G. Jenkins, Eds., American Society for Testing and Materials, West
Abstract: The dawn of a new age in understanding fracture of brittle material occurred in
1999 (eight years after an ASTM task group on fracture toughness of advanced ceramics
was formed) when a national full consensus fracture toughness test method standard was
approved as ASTM Test Methods for the Determination of Fracture Toughness of
Advanced Ceramics at Ambient Temperature (C 1421 ). C 1421 uses common flexure
test specimen geometries (typically 3 x 4 x 50 nun) in conjunction with three and four-
point flexure test configurations. Three different types of notch/crack geometries are
employed: single edge precracked beam (SEPB); surface crack in flexure (SCF); and
chevron-notched beam (CNB). Extensive experimental, analytical, and numerical
evaluations were conducted prior to and during the approval of the standard. Several
round robins (e.g., VAMAS) verified and validated the choice of dimensions and test
parameters. Two concurrent post-approval implementations of the standard are cited: i)
fracture tests of a single composition of an advanced ceramic led to the introduction of a
standard reference material (SRM) for fracture toughness and ii) fracture tests on two
different compositions of a ceramic system showed the statistically significantly greater
fracture toughness of one material for a wear application.
Keywords: advanced ceramic, fracture toughness, chevron notched beam (CNB), single
edge precracked beam (SEPB), surface crack in flexure (SCF)
1 Professor and Associate Chair for Academics, University of Washington, Department of Mechanical
Engineering, Box 352600, Stevens Way, Seattle, WA 98195-2600.
2 Materials Research Engineer,NASA Glenn Research Center, Structural Integrity Branch, 21000
Brookpark Road, MS 49-7, Cleveland, OH 44135.
3 Research Scientist, National Institute for Standards and Technology,Ceramics Division, Materials (223)
Room A359, 100 Bureau Drive, Stop 8521, Gaithersburg, MD 20899-8521.
4 Professor, WorcesterPolytechnicInstitute, Mechanical Engineering Deparanent, 100 Institute Road,
Worcester, MA 01609.
49
Introduction
Fracture toughness, while a conceptually simple parameter when discussed by Irwin

[1], has often posed a conundrum in the testing community. Originally, the term "fracture
toughness" referred to a material property expressed as the critical stress intensity factor
at which a crack initiated in the material [1].
However, because of complications in achieving consistent, repeatable, and
reproducible measures of fracture resistance, even the operational definition of fracture
toughness has been standardized depending on the procedures for the fracture tests. For
example, ASTM Test Method for Plane Strain Fracture Toughness of Metallic Materials
(E 399) defines fracture toughness as "a generic term for measures of resistance of
extension of a crack." Indeed, E 399 is even particular about the use of the symbol for
fracture toughness, Klc, for plane strain fracture toughness measured using the procedures
of E 399.
Fracture resistance of brittle materials such as glasses and advanced ceramics in
particular is of concern because these materials exhibit strength distributions with wide
variability. This wide variability is due to the brittle nature of advanced ceramics and
attendant catastrophic failure originating from inherent or induced critical flaws. These
critical flaws exhibit statistical distributions for size, orientation, and type. The strength
and flaw distributions are related through a form of the Griffith fracture criterion [1]
Xc (1)
O'f = y.,f-~
where crI is the critical stress at fracture (i.e. fracture strength), Kcis the "fracture
toughness" of the material, Y is the geometry correction factor for the stress intensity
factor, and a c is the critical crack length (flaw size). Because, Kc in Eq 1 is a material
property with limited variability, the wide scatter in strength observed for ceramics arises
from the variability in the flaw distributions.
Ceramics in general are brittle (K~<10 MPa ~/m) and therefore critical crack sizes at
fracture are small (on the order of 10 to 50 ~tm). Therefore, it is probably not reasonable
to think that designs of components comprised of ceramics would be analyzed using
fracture mechanics principles and the material's fracture toughness. Instead, the utility of
knowing a brittle material's fracture toughness might be in combination with other material
properties to define relative performance indices [2] such as the brittleness index [3,4]
HE
B=~ (2)
(K3 2
where B is the brittleness, E is the elastic modulus, and H is the hardness. In this case, in
order to decrease the tendency to brittle behavior (i.e., cracking rather than plastic
deformation), B needs to be decreased by decreasing H and E and/or increasing Kc. One
utility of the brittleness index, B, is its use for ranking materials for design selection.
Meaningful calculations of B require confidence in the measured material properties
(e.g., E, 11, K~) that make up the index. It is possible to argue that for performance
indices such as B, it may not be necessary to measure these material properties
accurately if the measurements are precise [5]. Of course, this argument assumes that for
ranking purposes based on performance indices, the material properties have been
JENKINS ET AL. ON CONSENSUS TEST METHOD STANDARDS 51
measured for all the materials using the same repeatable and reproducible test procedures
that can be easily carried out, even by a nonexpert [5].
In support of design needs such as performance indices for implementing advanced
c~mmics in structural applications, ASTM Committee C28 on Advanced Ceramics has
labored for over a decade to develop standard test methods for such mechanical properties
and performance as strength, elastic modulus, and hardness [6]. Standard test methods
for many of these properties typically have taken about one and a half to two years from
the appointment of the task group and attendant development of the first draft, to the
initial balloting, through revisions and finally on to approval as a full consensus standard
[6]. The ASTM standard for fiacture toughness of ceramics [ASTM Test Methods for
the Determination of Fracture Toughness of Advanced Ceramics at Ambient Temperature
(C 1421)] required eight years (approximately four times the "average" time) to complete
the same lirneline, which included the publication of a provisional standard [2 8] before
approval of the final full consensus standard [9]. While the three test methods contained
in C 1421 (see Figure 1) may have contributed to this extended timeline, it is also possible
to argue that the commonality of the basic test methods differentiated by the pre-crack or
notch configurations may have actually contributed to the synergism and harmony of the
test methods in producing a high-quality and technically rigorous full consensus standard.
The historical aspects of the development of C 1421 have been detailed elsewhere [7-
9]. However, explanations of certain critical technical aspects of C 1421 have not been
discussed fully, although limited explanations are contained in footnotes in the standard.
In this paper, explanations of some critical technical aspects of C 1421 are attempted.
The main body and aspects of C 1421 common to the three test methods are first
addressed. Then key aspects of each test method: single edge precracked beam (SEPB),
surface crack in flexure (SCF), and chevron-notched beam (CNB) are detailed. The utility
of C 1421 is demonstrated in a separate section on implementing the standard in assessing
Figure 1 -- Three test methods c o n t a i n e d in C 1421.

fracture toughness of several advanced ceramics. Finally, lessons learned and implications
of the development of C 1421 are contained in a summary and conclusion section.
Common Aspects of Test Methods
C 1421 contains three main parts: main body, annexes (mandatory information
specific to each test method), and appendices (nonmandatory information specific to the
SCF method). The main body of C 1421 includes information and sections common to all
three test methods. These sections include the following.
Scope
This section states that the test methods are intended for use with advanced ceramics
which are macroscopically homogeneous and isotropic. Note that this applicability may
include certain composites with randomly oriented whiskers or chopped fibers.
Terminology
This section includes pertinent definitions. These definitions are of two types:
previously-approved/generally applicable definitions and new definitions specific to
C 1421. It is in this section where the operationally dependent definitions of KIpb
(fracture toughness using the SEPB method), Klsc (fracture toughness using the SCF
method), and Klvb (fracture toughness using the CNB method) are contained.
Summary of Test Methods
This section provides an overview of the three test methods to provide initial
guidance for the user in selecting an appropriate method.
Significance and Use
This section gives specific uses of the test methods, including material development,
material comparison, quality assessment, and characterization. Specific differences
between the test methods are also indicated, including the SEPB and CNB methods for
measuring fracture resistance for large cracks and the SCF method for measuring fracture
resistance for small cracks.
Interferences
Interferences are factors that can influence the test results and affect the bias and
precision of the standard. This section discusses crack growth resistance (R-curve) and
slow crack growth (SCG) phenomena, and also has a subsection on "stability" and
controlled crack propagation. Later sections address how to avoid or accommodate these
phenomena by varying test rate or test conditions. Generally, for materials with fiat R-
curves and no sensitivity to SCG, all three methods give similar fracture toughness
results. However, because of differences in creation and propagation of the crack
materials with either rising R-curve or SCG sensitivity (or both) might be expected to
result in different fracture toughness results for the three methods.
Apparatus
This section gives specifics about the test machine, deflection measurement,
recording equipment, and test fixtures. Because the CNB test method requires stable
fxacture for a valid test and stable fracture is recommended for the SEPB method, it is
JENKINS El" AL. ON CONSENSUSTEST METHOD STANDARDS 53
necessary to provide a means of detecting this stability. One important method is to

record the force applied during the fracture test as a function of some other parameter.
This other parameter may include actuator displacement, load line displacement of the
test specimen, time, or even back face strain on the test specimen. Although the standard
allows the use of the remote actuator displacement, this displacement may not be
sensitive enough to detect stable crack extension thus creating two potential outcomes: i)
SEPB or CNB tests that are actually stable may be rejected or ii) stable crack growth in
the SEPB test may go undetected, and the incorrect crack length (shorter) may be used in
miscalculating a lower fracture toughness value than actually exists.
Backface strain (13FS) [I0] or similar displacement measurement on the test
specimen (e.g., center point deflection or crack mouth opening displacement) has been
shown to provide greater sensitivity to detecting stable crack growth and stability (see
Figure 2). The examples shown in Figure 2 illustrate that BFS measurements indicate that
the two tests are actually valid because of stable crack propagation whereas the actuator
displacement and time measurements indicate invalid tests.
Testing configurations are typically three and four-point flexure (see Figure 3).
Drawings of suggested test fLXtures are contained in an Annex. The most obvious
advantage of these testing configurations is the commonality of the flexure test
configuration for testing advanced ceramics. No less than four ASTM standards (ASTM
Test Method for Flexural Strength of Advanced Ceramics at Ambient Temperature
[C 1161]; ASTM Test Method for Flexuml Strength of Advanced Ceramics at Elevated
Temperature [C 1211]; ASTM Test Method for Determination of Slow Crack Growth
Parameters of Advanced Ceramics by Constant Stress-Rate Flexure1 Testing at Ambient
Temperature [C 1368]; ASTM Test Method for Determination of Slow Crack Growth
Parameters of Advanced Ceramics by Constant Stress-Rate Flexural Testing at Elevated
Temperature [C 1465]) use the flexure configuration as do, for example, five Comite' de
Europeen Normalization (CEN) standards (CEN Advanced Technical Ceramics-
Monolithic Ceramics-Mechanical Properties at Room Temperature-Determination of
Flexuml Strength [EN843-1], CEN Advanced Technical Ceramics-Monolithic Ceramics-
Mechanical Properties at Room Temperature-Determination of Subcritical Crack Growth
Parameters f~om Constant Stressing Rate Flexuml Strength Tests [ENV 843-3]; CEN
Advanced Technical Ceramics-Methods of Testing Monolithic Ceramics-Thermo-
Mechanical Properties-Determination of Flexural Strength at Elevated Temperatures
[ENV 820-1]; CEN Advanced Technical Ceramics-Methods of Testing Monolithic
Ceramics-Thermo-mechanical Properties-Determination of Self-loaded Deformation [ENV
820-2]; CEN Advanced Technical Ceramics-Methods of Testing Monolithic Ceramics-
Thermo-mechanical Properties-Determination of Resistance to Thermal Shock by Water
Quenching,[ENV 820-3]) one International Organization for Standardization (ISO)
standard ([SO Fine Ceramics (Advanced Ceramics, Advanced Technical Ceramics]-Test
Method for Flexuml Strength of Monolithic Ceramics at Room Temperature[FDIS
14704)] and four Japan Industrial Standards Committee (JISC) standards (JIS Testing
Method for Flexural Strength (Modulus of Rupture) of Fine Ceramics JR1601]; JIS
Testing Method for Flexural Strength (Modulus of Rupture) of Fine Ceramics at Elevated
Temperature [R 1604]; JIS Testing Methods for Bending Fatigue of Fine Ceramics at
Room Temperature [R1621]). This commonality is important in developing a standard
such as C 1421 because it builds on the research, experience and familiarity of the basic
flexural testing configuration [11-14] while allowing concentration on new issues such as
precracking, test rates and validity requirements.
The four-point flexure configuration is common to all three test methods of C 1421
with inner and outer spans of 20 mm and 40 ram, respectively, recommended. The rollers
400 . . . . . 40 . . . . . . 7
~300 .-.30
1
~ 25
200 ~ 20
100 10
5
00 200 400 600 800 1000 1200 00 5 10 15 20 25 30 35
Time (s) Time (s)
400 , , 40
35
.-.3O0 30
z ~25
r
.o200 ~ 2o
0
LL
100
5
00 0.04 0.08 0.12 0O 0.01 0.02 0.03 0.04
Actuator Displacement (mm) Actuator Displacement (mm)
~.300
v
,oo ....... t,0[ ~ 30
z 25
35
~
,,
200 o 20
~15
100 10
5
0o 5LO 1(~0 150
r 260 250 3()0 350 400 00 50 1(~0 150200 250300350 400
Backface Strain (p m/m) Backface Strain (IJ m/m)
a) Test results for the same SEPB test b) Test results for the same CNB test
Figure 2 -- Force vs time, force vs actuator displacement and force vs back face strain
plots for a) SEPB test and b) CNB test.
are fixed diameter of 4.5 nun but must be free to rotate to minimize frictional restraint
effects, which can cause errors of 3% to 20%. Although not used at all for SCF, and only
used by one CNB configuration, three-point flexure can be used over a wide range of
configurations for the SEPB method. Limitations have been placed on the span to depth
ratio of 4_~So/W<I0, the middle loading roller (fixed), and the tolerance on misalignment of
the notch or precrack with the middle loading roller (3% of So) [15].
J, ~. 4--'--
a) FouropointJ'lexure b) Three-pointflexure
Figure 3 -- Examples of testing configurations a) Four-point flexure and b) Three-

point flexure [C 1421].
Specimen Configurations, Dimensions, and Preparations
Conventional 3 x 4 x ---45 mm flexure test specimens are recommended for all three
methods (see Figure 4). However, the CNB does allow two configurations to differ from
these cross sectional dimensions and the SEPB method allows shorter test specimens to
20 mm in length. As with the testing configuration, there are several advantages to using
this test specimen configuration for fracture testing. Commonality of the flexure test
specimen configuration is a major plus. The same four ASTM standards, five CEN
standards, one 1SO standard and four JIS standards cited earlier use similar test specimen
geometries and configurations. Not only the base dimensions, but also the geometric
dimensioning and tolerancing (see Figure 4) have been established nsing analytical and
empirical research [12, 13] to the point that systematic errors in flexure testing are
reduced, and commercial machine shops can consistently and reliably produce single
flexure test specimen to the dimensions and tolerance required at costs as low as US$8
per test specimen (cost in year 2000 US dollars).
Machining aspects unique to the configurations of each test method are contained in
the appropriate annex. However, these unique aspects are generally restricted to notch or
precracking aspects of test specimen preparation.
3.00• mm
L_ ,.._J See DetailA
4.o~o.ls ~m - T . . ~ I ~ v . . . . . . ~
~ I ~ ,n,~ I
0.12+093 turn I~ ~ - -
Typical, 4X
0.15~0.05 mm R -,,~--~ ]
~~176 Typical, 4X /
Detail A Detail A-Alternate Method
Figure 4 -- Dimensions of a rectangular beam [C 1421].

General Procedures
This section addresses nuances specific to setting up and running fracture tests in
flexure. A validity subsection provides the conditions under which an individual test can
be considered as having been acceptably tested per C 1421. In short, "a valid test shall
meet general requirements in addition to the specific requirement of the particular test
method. One of the general requirements for C 1421 is that a minimum of four valid tests
be conducted for each material and test condition. However, recognizing that problems
may arise with testing particular materials or using certain test configurations, the
standard recommends the preparation of ten test specimens.
Test rates between 0.1 MPa~/m/s and 2.75 MPa~m/s are specified for normal testing.
If SCG is suspected, C 1421 recommends testing at two different rates (at least two to
three orders of magnitude difference). IfSCG occurs at the slower test rate, then a faster
rate must be chosen so as to eliminate the occurrence of SCG. Alternatively, parallel tests
should be conducted in laboratory ambient conditions and in an inert environment (e.g.,
dry nitrogen). If SCG does not occur in the inert environment, then the resulting fracture
toughness satisfies the conditions of the standard.
For the SEPB and SCF methods, the maxirnnm relevant force (i.e., force used in the
calculation of fracture toughness) is the maximum force occurring during the fracture test
regardless of the shape of the loading curve (see Figure 5a). In general, it is simple and
straight forward to use the maximum force in combination with the optically measured
precrack length to calculate the fracture toughness for these methods.
This simple approach is further justified for the SEPB method by its comparison
with two other approaches that use the force vs backface strain information (force at 2%
crack extension combined with the optically measured precrack length and maximum force
combined with the crack length determined from the relation of backface strain and crack
3
-J
i * i
Displacement Displacement
a) Relevant maximumforce for SCF b) Relevant maximumforce for SEPB and

and SEPB is absolute maximumforce CNB is the absolute maximumforce
l
/ = m i
Displacement
c) Relevant maximumforce of CNB less than absolute maximumforce
Figure 5 -- Examples of relevant maximumforce [C 1421].

length). Applying the three approaches to five different brittle materials, showed that all
three approaches gave similar results [I0]. However, because stability is desirable for the
SEPB method and stable crack propagation is a requirement for a valid CNB test,
nonlinear loading curves such as those shown in Figures 5b and 5c must be recorded so as
to indicate these conditions. In the case of the CNB test method, the maximum relevant
force may sometimes occur after the actual maximum force at crack pop-in (see Figure
5c). Unstable crack propagation (similar to Figure 5a) constitutes an invalid CNB test
and the test results cannot be used to calculate a fracture toughness value.
Stability is briefly discussed in the standard in relation to the effect of stable crack
extension on the measured fracture toughness value. Stability is the condition in which
only the stored energy in the test specimen is consumed by the fracture process.
Stability is manifested by controlled and stable crack propagation. Compliant test
systems may provide extemal to the energy to the fracture process and thereby promote
uncontrolled, catastrophic crack propagation. Therefore, certain test specimen geometries
and testing configurations are chosen to promote stable crack propagation and hence
stability [16, 17]. Specifics of the SEPB and CNB methods of C 1421 are intended to
promote stability.
Report
Proper reporting of the implementation of the C 1421 is a critical aspect of the
standard. Without proper reporting the test results may not be replicatable and the
resulting information may be useless for quality control, process development, material
selection, or design purposes. Suggested reporting templates for each test method are
included to provide guidance to the user in creating a complete report (e.g., see Figure 6).
Two levels of reporting are required: information for the test of each test specimen
and information for all the tests from a particular test method. For each test specimen,
the user is required to report information on the form of the product and materials
processing (if possible) including mean grain size. Environment of testing, relative
humidity, test temperature, and crack plane information are required. General test
specimen dimensions such as width, W, and breadth, B, are reported as well as test
specimen dimensions specific to each test method such as crack length, a, and notch
thickness, t, for the SEPB configuration; crack dimensions a and 2c for the SCF
configuration; and notch tip depth, ao, chevron length, al, and notch thickness, t, for the
CNB configuration.
Test fixture specifics to be reported include whether the test is three- or four-point
flexure and the outer and inner span dimensions (if applicable) for each test, So and Si,
respectively. Relevant maximum force is reported as well as the calculated fracture
toughness value and a statement of test validity. Additional information may need to be
reported as required in the appropriate annex depending on the test method used.
For all the tests from a particular test method, the mean and standard deviation for all
valid tests are reported. The number of test specimens used and the number of valid tests
also are reported.
Precision and Bias

An element unique to ASTM standard test methods is the precision and bias (P&B)
statement. The P&B statement provides the user with some guidance as to what to
K]sc DATA SHt:t: I
Material: Form: Processing Details:
Specimen ID: Test System: Elastic Modulus (GPa): "13
Date: Mach!ne Compliance m/N): Mean ~lrain size (10-6 m): o
Test Particulars Data Test Method Section Fracture Test Data Test Method Section C
~0
Ambient Environment 9.3, 9,.9 Crack Dimensions m
Relative Humidity (%RH) ).9 a(mm) A3.3.4, Figs. 5 & A3.4 '=I
m
Temperature (~ ).9 c (mm) A3.3.4, Fi~ls.5 & A3.4 60
.-~
i
.~pecimen/Crack 3.2.3 crack Details A3.5.1, A3.5.2,
Plane Orientation (valid crack?) G)
A3.5.3, A3.5.4 0"T1
Load Configuration 7.4 Tesi Rate (N/s) 9.7, A3.3.3 E
(4-poin!only) 0
Z
Outer span, So (mm) 7.4 0I'-
Inner span, SI (mm) 7.4 Test Record --I
Specimen Dimensions Stable Crack A3.5.5 -r"
Extension IY/N)
3 (mm) A3o1.1, Figs. 5 & Fractography •
A3.1 0
(SEM/Optical)
Initial (pre-polish) "D
A3.3.2.2, Figs. 5 &
W/ramI A3.1 0
--t
Precrackin~ )ata Test Method Section Calculation of Kt~c Data Method Section rrl
Indent Type Maximum Load, Prn~x(N)
ITest
9.8.1, A3.3.3
(Knoopor CantedVickers) A3.3.1, X3 Coefficients for Stress A3.4.1 m
Intensity Factor '11
Indent Diagonal, d (mm) ~.3.2.1 D~pem, Yd A3.4.1.1
Indent Depth, h (mm) A3.3.2.1 Surface, Y~ ~,3.4.1.2
Post-polish W (mm) A3.3.2.5, A3.3.2.9 Measured Fracture ~,3.4.2, A3.5.5
Toughness,
K/,W(MPa r ~ )
Amount removed by polish or Apparent Fracture ~.3.4.2, A3.5.5
hand grinding (turn)
Toughness,
Kbc* (MPa x/m)
ndentation Load iN) A3.3.1.2 /alid Klsc (Y/N) 9.2, A3.5 , A3.6
Table 1--Examples offracture toughnessfor SEPB, SCF and CNB test methodsfrom
C 1421for two materials (MPa ~lm).
Materials SEPB Method SCF Method CNB Method
t~-Silicon Carbide 5 2.58• (4) 2.76-a:0.08 (4) 2.61+0.05 (6)
Hot Pressed
Silicon Nitride 6 4.59-a:0.12 (11) 4.55+0.14 (14) 4.60-a:0.13 (8)
X.XX• (X) = mean a: one standard deviation (number of tests).
expect in terms of the accuracy and precision of test results obtained in accordance with
the standard. A statement on precision (i.e., closeness of agreement between test results
obtained under prescribed conditions) allows users of the test method to assess its
usefulness in proposed applications. Within-laboratory precision is known as
"repeatability" and between-laboratory precision is termed "reproducibility." A
statement of accuracy includes reference to bias, which is a systematic error that
contributes to the difference between a mean of a large number of test results and an
accepted reference value. A key aspect of C 1421 is that the three methods should give
fracture toughness xalues in agreement with each other for a material with a flat R-curve
and negligible sensitivity to SCG. Examples of such agreement a sintered a-silicon
carbide 5and a hot-pressed silicon nitride 6 are included in the standard's P&B section.
Agreement is generally good within individual studies and reasonable across several
studies when considering batch to batch variations (responsible for much of the
variability). Table 1 shows examples of fracture toughness values extracted from the
P&B section of C 1421 along with standard deviation and number of tests for the two
different materials. Note that although these values were not necessarily obtained using
all the requirements of C 1421, they do indicate the utility of the test methods.
Table 2 [8] contains statistically-derived precision estimates for the SEPB and SCF
methods for which data from two Versailles Advanced Materials and Standards
(VAMAS) round robin projects were available [18-21]. The uncertainties were
calculated in accordance with ASTM Practice for Conducting an Interlaboratory Study to
Determine the Precision of a Test Method (E 691). E 691 gives "repeatability standard
deviation" and the "reproducibility standard deviation" that are normalized in Table 2 as
coefficients of variation.
Table 2 -- Examples of precision for SCF and SEPB test methods

Test Method Material Coefficients of Variation Coefficients of Variation
for Repeatability for Reproducibility
(within laboratory) (between laboratories)
SEPB gas pressure sintered 4.6% 7.1% or 8.8%
[18, 19] silicon nitride (results depend on test rate)
SCF hot pressed silicon 5.4% 6.8%
[20, 21] nitride 6
SHexoloy SA, Carborundum Co, Niagara Falls, NY (vintages to 1977) (now Saint Gobain Industrial
Ceramics)
6 NC132, Norton, Worcester, MA (vintages to 1975) (now Saint Gobain Industrial Ceramics)
Annexes
There are four mandatory annexes in C 1421: one on suggested test fixtures and one
for each of the three test methods. Individual subsections address each of the annexes as
follows.
Suggested Test Fixture Schematics

In this annex, drawings of two suggested flexure fixtures (four-point and three-point
flexure) are given (see Figure 3). General requirements for the flexure test fixtures are
indicated including critical requirements for load and support pin sizes (4.5 + 0.5 mm
diameter) and rolling action.
Special Requirements for SEPB Method

As memioned, the basic test specimen geometry for the SEPB method is the 3x4 mm
cross section of the bend bar (see Figure 4). The length may vary from 20 to 50 mm to
accommodate different testing configurations ranging from 16-mm outer span of three-
point flexure to 40-mm outer span of four-point flexure (20-mm inner span). The test
specimen may oriented such that B is either 3 mm or 4 mm depending on the needs of the
user and the particular material although B=3 mm is recommended.
Procedurally, the SEPB involves introducing either a precrack (using a row of Vickers
or Knoop indentations) or a shallow notch. A bridge compression fixture is then used to
"pop-in" a crack for the actual fracture test. Once the crack is introduced, the fracture
test is then conducted in a manner nonunique to the SEPB method.
Post-test inspections are required to establish the validity of the individual test as
follows. The length of the crack at fracture is measured at three different crack from
locations. A 10% or less difference between the average crack length and the minimum
measured crack length is required for a valid test. The average crack length should also be
in the range of 0.35W to 0.60W for a valid test. The plane of the final crack shall be
parallel to the test specimen dimensions, B and 1I, within :~5~ for three point flexure and
within +10 ~ for four point flexure. The loading curve can provide information on the
degree of "stability" of the test where lack of nonlinearity in the curve indicates unstable
fracture.
Finally, depending on the testing configuration fracture toughness for the SEPB
method, Klpb, is calculated using the relation
Fem~,SolO-'T 3(,~lW) '1~ q
for three-poim flexure and
s 3(a/w)''2 ]
Kleb = ] L B-Vr _]_-2(-i__a T - ~ , 2 (4)
for four-point flexure, where Pmax is the relevant maximum force for the SEPB method, So
is the outer test span, Si is the inner test span, a is the crack length, g is the stress
intensity factor coefficient for three-point flexure that is the well known relation of
Srawley [22] for So/W = 5 but is a wider ranging relation for 5_<So/W_<I0 [17], and f is the
JENKINS ET AL. ON CONSENSUSTEST METHOD STANDARDS 61
6.5 I I I I I
-~---~ 6.0 Treat
~ M~ 5.5 ~ "" . . " I As Indented

Lr~ : HeatTreat then
Indentation
5.0 , , , I ~t ,
0 100 200 300 400 500 600 700
Indentation Force (N)
Figure 7 -- Plot offracture toughness vs. indentationforce without and with heat
treatmentsfor SEPB method of a hot pressed silicon nitride [after Ref 24].
relation for the stress intensity factor coefficient for four-point flexure relation of Srawley
and Gross [23]. The utility of the SEPB method is reflected in its use in two other test
standards: JIS Testing Methods for Fracture Toughness of Fine Ceramics (R1607) and
ISO Fine Ceramics (Advanced Ceramics, Advanced Technical Ceramics)-Test Method for
Fracture Toughness of.Monolithic Ceramic at Room Temperature by Single Edge
Precracked Beam (SEPB) Method (DIS 15732).
While many technical issues involving the SEPB method could be elaborated on here,
only two will be addressed for the sake of brevity. These two issues are: the limit of 100
N on the maximum indentation force for the introducing a precrack and effect crack twist
and hook in the limits of 5~ for three-point flexure and 10 ~ for four point flexure.
The limit on the indentation force of 100 N is intended to minimize potential residual
tensile stresses that could influence the fracture. Indentation forces greater than I00 N
have been shown to produce fracture toughness results for the SEPB that are less than
those obtained for ihdentation forces of 100 N or less (see Figure 7) [24] because the
residual tensile stresses superpose an additional stress intensity factor on the stress
intensity factor applied during the fracture test, thereby decreasing the calculated fracture
I 2,xo/C k
\
| d Precrack
Figure 8 --Illustration of angular allowance of the crack plane [C 1421].

8.0 t 1 1 l
~ 7.5
13 [3
~- 6.5
~v
~= -~ 6.0
u_ 5.5
5.0 I I I ]
0 10 20 30 40 50
Maximum Hook Angle (~
Figure 9 -- Fracture toughness (SEPB) vs. "hook"angle (i.e. deviation of crack from
plane) for silicon carbide whisker-reinforced silicon nitride7 [l OJ.
toughness. C 1421 recommends that if residual stresses from the indentation are
suspected, the indentations may be removed by polishing, hand grinding or grinding after
the crack has been formed using the compression anvil. Annealing via a heat treatment
may be used provided that the crack tip is not blunted nor the crack tip/planes are
"healed." The effects of such annealing (heat treatment) are shown in Figure 7.
Limits of 5~ for three-point flexure and 10~ for four point flexure on the deviation on
the crack plane (see Figure 8) have been suggested in a recent study [10]. In this study,
the "hook" angle was defined as the angle between the lines drawn tangent to the initial
crack plane and the crack tip. For a single material, there is an indication that the
calculated fracture toughness decreases with increasing hook angle (i.e., angle of the crack
plane) as shown in Figure 9. Twisting of cracks has been noted to increase the measured
Klpb and reduces the tendency for stable crack extension.
Special Requirementsfor SCF Method

As mentioned, the basic test specimen geometry for the SCF method is the 3x4 mm
cross section of the bend bar (see Figure 4). Because the only testing configuration
allowed for the SCF method is four-point flexure with 20 mm and 40 mm inner and outer
test spans, respectively, the test specimen length may only vary from 45 to 50 mm. The
test specimen may also be used with B=3 mm or B--4 mm depending on the needs of the
user and the particular material.
Procedurally, the SCF method involves introducing a precrack using a single Knoop
indentation made in the polished surface at 24 to 49 N with the test specimen tilted 1/2~
from the flat. After measuring the diagonal length, d, of the Knoop indentation, the
indentation zone containing residual stresses is eliminated by using hand polishing or
machine grinding to remove material to a depth of 4.5 h to 5.0 h (where h=d/30). Once
the precrack has been properly introduced and prepared, the fracture test is conducted in
a manner nonurdque to the SCF method.
Post-test inspections are required to establish the validity of the individual test as
r Kryptonite,Japan Metals and ChemicalsCo., Tokyo,Japan (20% vol silicon carbide whiskers)
follows. The shape of the precrack should be semi-elliptical with dimensions a and 2c for
a valid test (note that the stress intensity factor coefficients used to calculate the fracture
toughness are valid only for a/cL:_l).
Finally, the fracture toughness for the SCF method, Klsc, is calculated as:
Pm~x(So- S,)104
K,,c= ~ ~-B-W~ 14-d (5)
where Pmax is the relevant maximum force for the SCF method, So is the outer test span,
Si is the inner test span, a is the crack depth, and Y is the the greater of either Yd
calculated at the deepest point of the precrack or Ys is calculated at the intersection of the
precrack with the surface per the solution proposed by Newman and Raju [25].
While many technical issues involving the SCF method could be elaborated on here,
only three will be addressed for the sake of brevity. These three issues are: the rationale
for the depth of material removal to eliminate the indentation and residual stresses; the
introduction and detection of the semi-elliptical precrack; and the application of the stress
intensity factor shape factors and coefficients.
After the indentation to form a precrack, the indentation damage zone and its
attendant residual stress field must be removed by polishing or hand grinding. Although
polishing and/or annealing can be done, polishing is preferred because annealing carries the
risk of crack blunting or healing. The polishing or hand grinding may be done with silicon
carbide papers for most ceramics, but diamond grinding may be necessary for very hard or
"tough" materials. A hand micrometer with a resolution of 0.002 mm may be used to
monitor the amount removed during hand polishing or grinding. The depth that must be
removed has been set at 4.5 h to 5.0 h (h is the depth of the indentation) as shown in
Figure 10. Early work on the SCF method [21, 26] empirically demonstrated that as little
as 3 h was sufficient to remove the residual stresses. This conclusion was confirmed by
additional work [27]. A greater amount of removed material has been specified in C 1421
because in addition to removing the residual stresses, it was also desired to remove any
traces of lateral cracks that could interfere with the stress field in the vicinity of the
primary median crack (see Figure 10). In addition, the removal of additional material has
the effect of changing the influence of the maximum stress intensity to be at the deepest
portion of the precracks, away from surface irregularities and surface polishing of
machining damage. After polishing or hand grinding, the crack shape is actually a
truncated semicircle or semi-ellipse, and the semi-elliptical crack stress intensity shape
factors are an approximation. The approximation is satisfactory if the maximum stress
intensity is at the deepest portion of the precrack, but is less satisfactory if the maximum
Yis at the surface.
Precrack size is measmxxl on the fracture surface. The 1/2~ precrack tilt during the
indentation step helps discern the precrack on the fracture surface. Optical and scanning
electron microscopy may be used. In some materials the precrack is easy to discern, even
with a low power stereo binocular microscope. In other materials, a careful examination
with a scanning electron microscope may be necessary. A variety of features, including
color, contrast, brightness, ridge or rim lines, halos, and micro hackle line redirection help
to discern the precracks (see Figure 11) [27, 28]. 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 proposed [27].
remove 4.5h to 5.Oh/' ~ ~h- d-'~3_O____ 1 _

--~--
Figure 10 -- Illustration of the depth of material removalfor SCF method to eliminate

Knoop indentation [C 1421].
The stress intensity shape factors, Y, for semicircular and semi-elliptical surface
cracks in bending are from the empirical equation developed by Newman and Raju [25]
(see Figure 12). Y depends upon the ratio of a to c, and their sizes relative to the
specimen cross section. Ytakes into account the bending stress fields. The Yfactor 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 promoted 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 on the fracture surfaces) it
was possible to confirm that fracture did initiate from the point of the periphery where Y
was maximized. It is also important to calculate the Y maximum for each and every
precrack, rather than using a single average value [29]. The Newman-Raju formulas [25]
for Yare widely accepted and have estimated accuracies to within a few percent.
One notable conclusion of the VAMAS round robin [18] was that the computed
fracture toughness was insensitive to the exact precrack boundaries marked, because the
computed fracture toughness was not sensitive to the precrack size measurement! This is
Figure 11 -- Example of a view of a precrack for SCF Method (99.9% sintered alumina as
photographed in the scanning electron microscope) [C 1421].
St M
I I I
~S b * ~'
.~t2
2h
~--- 2c -- t
I
20
St M
111 TeruliOn. (hi I k ~ l r ~ g .
Figure 12 -- Surface Crack in a finite plate [24] a)tension and b) bending. Note that
the SCF precracks in ceramics are very small relative to the plate size [25].
partly due to the dependence of the fracture toughness on the square root of the crack
size. Thus, a 10% uncertainty (or error) in crack size is diminished to a 5% uncertainty in
fracture toughness. A serendipitous outcome was the offsetting influence of Y on the
errors due to crack size measurements. For example, if the crack depth is overestimated,
the corresponding value of Yis underestimated. The shape of the ellipse affects the stress
intensity shape factor, Y, which ranges from 1.28 to 1.99 for shallow semicircular and
semi-elliptical surface cracks in bending.
The SCF method is also the subject of international standardization 8. For consistency
with the other ISO standard for fracture toughness of advanced ceramics, the current SCF
draft follows the format of ISO/DIS 15732 on the SEPB method. The current schedule
for the ISO SCF standard is for approval as an international standard by 2002.
Special Requirements for CNB Method
The basic test specimen geometry for two of the four test specimen geometries
allowed in the CNB method is the 3 x 4 mm cross section bend bar tested in four-point
flexure. A third geometry is a square cross section of 6.35 x 6.35 mm tested in three-
point flexure with a fourth geometry having a 3 x 6 mm cross section is tested in four
point flexure. The minimum length for all four geometries as shown in Figure 13 is 45
mm. Specific dimensional ranges on the depth of the chevron notch tip, ao, and the length
of the chevron notch section, al, are given for each geometry.
Procedurally, the CNB method involves fabricating the test specimen with the
properly dimensioned chevron notch. Once the notch has been introduced, the fracture
test is then conducted in a manner nonunique to the CNB method.
a WorkingGroup, WG16/TC206, ISO/WD 18766 "Determinationof Fracture Toughness of Monolithic
Ceramics at Room Temperature by Surface Crack in Flexure (SCF) Method," ISO, Geneva, Switzerland.
. - ~ t~).25 W=4D0 • -.~ ~ B--3.00 :t-0.13
L= 1_=45 (min) J ~ T
I- -]
alland a12 = 0.95W - 1.00 W
Configuration A
t t~'25 W=6.35 • ~ 1 ~ B=6.35 •
L
I
L=,5 (min) J
I alzand a12= 0.95W-1.00 W
Configuration B
9-=lq---- t ~).25 W=6.00 • _.~ ~ B-_3.00 •
I_
F" L=45 (m~) _1
"7
iX Tp T
a 1land a 12=4.20~O.07
Configur~lion C
w=,oo
I
I_
i
L=45 (rain) -J
i-t T
a11and a12= 0.95~/ - 1.00W
Configuration D
Notes 1) All dimensions in mm.

2) "lqpsof chevrons on transverse centerline within 0.02 B
3) Planes on either side which form chevrons shall meet within 0.3t
Figure 13 -- Illustrations of chevron notchflexure test specimen geometries [C 1421].
Post test inspections are required to establish the validity of the individual test as
follows. The tip of the chevron should be within 0.02 B of the center and the center line
of the notch grooves on either side of the tip should meet within 0.03 t where t is the
notch thickness for a valid test. The notch thickness should be <0.25 mm for a valid test.
The "crack follow" should be predominantly in the chevron notch plane. The test record
should exhibit a smooth nonlinear transition through the maximum force for a valid test
(see Figures 5b and 5c).
Finally, the fracture toughness for the CNB method, Klvb, is calculated using the
relation:
9 Pm~(So -- S~)IO -6
(6)
where Prnax is the relevant maximum force for the CNB method, So is the outer test span,
Si is the inner test span, and Y'rain is the minimum stress intensity factor coefficient that
occurs theoretically for the chevron notch at the maximum force, Pmax.
While several technical issues involving the CNB method could be elaborated on here,
only three will be addressed for the sake of brevity. These three issues are: test specimen
dimensions, dimensional tolerances, machining issues; minimum stress intensity factor,
Y'rain, and detection of stable crack propagation.
The four test specimen geometries and test configurations shown in Figure 13 were
selected because they had been utilized in several investigations and were therefore
reasonably well-studied and analyzed. In addition to its common 3 x 4 mm cross section
bend bar test specimen, Geometry A has been used in various industrial processing and
life prediction programs funded by the U.S. Department of Energy (US-DOE) and a
significant data base exists [30]. Geometry B and its simple three-point flexure
configuration has also been used in industry and academia in various US-DOE programs
and a data base on a variety of materials exists, especially at elevated temperatures [31].
Geometry C has been reported to the most likely to arrest an unstable pop-in fracture
from the chevron tip [32] and has been used in studies of whisker-reinforced silicon
nitrides [33]. Similar to Geometry A, Geometry D uses the 3 x 4 mm cross section bend
bar tested in four-point flexure. However, the chevron notch tip is deeper for Geometry
D than Geometry A. Geometry D was the subject of an international round robin [34].
Comprehensive parametric studies of chevron notched test specimens have not been
completed although studies targeted to specific experimental issues [35-37] have been
used in detenuining certain limits on dimensions. However, in general, the dimensional
tolerances contained in C 1421 and shown in Figure 13 were determined from post
fracture measurements of valid tests conducted using the CNB method of C 1421 [38].
These test specimens had been machined in commercial machine shops using conventional
tooling and procedures per the drawings supplied from C 1421. It should be noted that
proper and consistent machining of chevron notches is critical to producing repeatable and
consistent fracture toughness results. Two recommendations are the use of thin diamond-
grit saw blades (~0.154 mm thick) laterally stiffened with stabilizers to minimize wobble
and runout, and machining fixtures (see Figure 14) for simultaneously machining multiple
test specimens to promote consistent material removal [39]. Variability in fracture
toughness results due to inconsistent and out of tolerance machining has been indicated in
round robins using the CNB method [34].
Determination of the minimum geometry correction factor, Y'rain was essential for
producing accurate fracture toughness results for the CNB method. Common fracture
toughness test specimens such as the single edge notched beam (SENB) have stress
intensity factors that can be analyzed using two-dimensional planar models [22, 23]
because the crack front has a fixed breadth (B) as the crack propagates through the test
specimen in W direction. For a CNB test, the breadth of the crack front changes as the
crack propagates through the triangularly shaped chevron in the W direction and therefore
the stress intensity factor for CNB test specimens must be analyzed to account for this
three-dimensionality. Stress intensity factors for CNB tests can be determined from
numerical [finite element analysis (FEA)] and experimental analyses of models with
. . . . . 1 st cut
SAW BLADE - - - - 2nd cut
/---.OLO-OOW.
Figure 14 -- Fixture for machining chevron notch test specimens [39].
particular dimensions [35, 36]. However, more wide ranging and and dimensionally
techniques such as the slice model [40] and the straight-through crack assumption
(STCA) [41, 42] have been proposed to simplify the analysis of stress intensity factors.
Once the stress intensity factor has been determined as a function of crack length for a
particular geometry, the minimum value of the geometry correction factor, Y'rain, must be
determined for use in Eq 5. A comprehensive investigation was undertaken to compare
Y'rain results for the four geometries of C 1421 as determined from FEA, boundary
element analysis (BEA), slice and STCA methods [38]. The allowed tolerances on the
dimensions of the chevron notch geometries shown in Figure 13 required surface fitting
polynomial functions of Y*mm as functions of al/W and ao/W. The STCA method
showed better agreement with the FEA and BEA results (1 to 3% maximum differences)
for Geometries A, B, and D because this method works well for the al/W=l of these
geometries. The slice method showed better agreement with the FEA and BEA results
&5% maximum difference) for Geometry C because this method works well for the
al/W<! of this geometry. The polynomial functions for Y'rain are contained in C 1421 as
Eqs A4.2 to A4.5 with claims of less than 1% error (compared to the curve fit) over the
allowed ranges of the chevron notch dimensions.
As discussed in the section on the "Common Aspects of the Test Methods," a
condition for a valid CNB test is stable crack propagation as indicated in nonlinear loading
curves of Figures 5b and 5c. A linear loading curve accompanied by catastrophic fracture
does not constitute a valid test because the premise of the relation for fracture toughness
using CNB (Eq 6) is that the maximum force, Pmax, occurs at the minimum geometry
correction factor, Y'rain. This premise requires the crack to initiate at the tip of the
chevron and then propagate in a stable manner across the chevron section. Depending on
the method for measuring the test specimen displacement, the decrease of stiffness of the
test specimen may or may not be detected as shown in Figure 2b. Note that plots of
force vs time and force vs actuator displacement do not indicate stable crack propagation,
whereas force vs back face strain is sufficiently sensitive to indicate stable crack
propagation for the same test. Measurement of center point deflection on the test
specimen or even crack (notch) mouth opening displacement are also preferred methods
to remote displacement measurement for detecting stable crack propagation (hence
indicating a valid test) [38].
The CNB method is also the subject of international standardization 9. For

consistency with the other ISO standard for fracture toughness of advanced ceramics, the
current CNB draft follows the format of ISO/DIS 15732 on SEPB method. The current
schedule for the ISO CNB effort is for elevation to working group status in 2001.
As a final point, it is interesting to note that an advantage of the CNB method is the
in situ crack initiation, which eliminates problems of crack "healing" that might occur for
the SEPB and SCF methods during elevated temperature fracture tests [43]. This
advantage was cited in a VAMAS round robin [34] for fracture toughness testing of a
silicon nitride using Geometry D and was exploited to test over a range of temperatures
(20~ to 1400~ and a variety of materials (spinel to whisker-reinforced alumina) using
Geometry B [31].
Implementation of C 1421
The justification for the existence of any standard is its applicability to solving
problems. In the case of C 1421, two examples of the implementation of the standard for
determining fracture toughness of advanced ceramics are described as follows.
Standard Reference Material 2100

A new standard reference material (SRM) for fracture toughness has been prepared
by the National Institute of Standards and Technology using PS070, the immediate
predecessor of C 1421 [44]. 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 which
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 x
158 mm x 20 to 30 mm thick. SRM 2100 is a very consistent material from a mechanical
properties standpoint [44] owing to it uniform microstructure (-3 ~tm grain size) and
small amount of sintering aid (~l wt% magnesium oxide). The material is fully dense with
a density slightly greater than theoretical for ~ silicon nitride because of very free
tungsten carbide inclusions.
In developing the SRM data base, a large sample set of representative flexure test
specimens (3 mm x 4 mm x 47 mm) from each of three billets of the material (Billets C, G
and D) were prepared in accordance with the procedures o f C 1161 and C 1421. The test
specimens were cut from the billets with their long axis perpendicular to the hot pressing
direction. The certified fracture toughness of SRM 2100 is for cracks propagated parallel
to the 3 x 4 mm cross section. One 4 mm wide face was given a 900 grit finish to serve as
a flat surface for the Knoop indentation for the SCF test method of C 1421. The edges of
the test specimens were chamfered in accordance with the SCF and SEPB test methods of
C 1421.
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 and SEPB test specimens. In some
9 Preliminary Work Item 5 "Determinationof Fracture Toughness of Monolithic Ceramics at Room
Temperature by Chevron Notch Beam (CNB) Method," ISO/TC 206 "Fine Ceramics," ISO, Geneva,
Switzerland.
lo NC 132, Norton Company, Worcester, MA.(now Saint Gobain Industrial Ceramics)
Table 3 --Fracture toughness results for SRM 21OOfrom billets C, D, G.

Billet Method Mean Standard Valid Tests
Deviation
(MPa~/m) (MPaqm)
C SCF 4.56 0.13 17
C CNB 4.58 0.12 10
C 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
instances, broken halves of test specimens were retested in three- and four-point flexure
per the SEPB method. A limited number of CNB tests were conducted on test
specimens from Billet C using Geometry A and the Y'rain relations contained in C 1421.
Table 3 shows the summary of the fracture toughness results for the three billets.
Fifty-eight test specimens from Billet C were tested, sometimes multiply, for a total of
75 fracture tests from which a statistical sample set of 46 valid tests was obtained.
Thirty-six test specimens from Billet G were tested, giving a statistical sample set of 32
valid tests. Twenty-five test specimens from Billet D were tested, giving a statistical
sample set of 37 valid tests.
Within a given billet, the fracture toughness results were in good agreement (1% to
2% regardless of test method). Between billets there are distinct differences with the
results for Billets C and D being statistically significantly different form Billet G [41].
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.
Material Selection for a Design Application

Another example of the implementation of C 1421 was by the manufacturer of fluid
control valves for abrasive environments. Due to the industrial-proprietary nature of the
application, complete details of the application or the design cannot be described here.
However, it is possible to briefly describe the use of C 1421 to provide information for
selecting a "better" material (i.e., longer-lived and more cost-effective) for the valve seats.
Two silicon carbides were compared: a well-known commercial a-silicon carbide"
(the current valve seat material) and a commercial, proprietary silicon carbide (the
candidate material). Representative billets of each material were first obtained from their
respective manufacturers from which ten flexure test specimens, 3 x 4 x 50 mm were
machined using the procedures of C 1421. Chevron notches per Geometry D from the
CNB test method of C 1421 were introduced into five test specimens of each material.
Back face strain gages were applied to the test specimens before testing in four-point
flexure (20 mm inner and 40 mm outer spans) at 0.05 mm/min under ambient conditions.
The results of the fracture tests are shown in Figure 15. Note that all five tests for
11 Hexoloy SA, Carborundum Co, Saint Gobain Industrial Ceramics, Niagara Falls, NY.
the candidate material were valid and four out of the five tests for the current material
were valid. Since four valid tests are the minimum number required for valid test series
per C 1421, the mean and standard deviation of fracture toughness for both materials
were validly measured per C 1421. An analysis of variance (ANOVA) at the 95% and
99% significance levels showed that the fracture toughness of the candidate material was
not only different but also statistically significantly greater than that of the current
material. Obviously, additional tests of other material properties need to be conducted
before a fmal design decision can be made. However, implementation of C 1421 again
shows that the reputation of ASTM standards for being technically rigorous and high
quality make them highly useful to industry for confidently assessing materials and
components in important engineering applications.
S u m m a r y and Conclusions
C 1421 is a full consensus standard approved in 1999, eight years after the formation
of the task group on fracture toughness testing within ASTM Committee C28 on
Advanced Ceramics. C 1421 is a refinement of the provisional standard PS070 sanctioned
in 1997. More than 125 changes were incorporated into PS070 to create C 1421.
C 1421 is composed of a main body, four annexes and three appendices. Three test
methods, SEPB, SCF, and CNB are contained in the standard. Ostensibly, for a fiat R-
curve material with no sensitivity to SCG, the three methods should give the same
fracture toughness results. When interferences arise (such as R-curve or SCG issues),
guidance is given in C 1421 addressing these issues.
E 6.0
5.0
v
\
4.0 \ 2",
e-
l-
3.0
x
\ -- \h
O 2.0 ~-~
I-- P..~4
-'1 -., \
1.0 ~-~ 9-., \ ~-a -.%
N."t ,) #
LL kN
\ "-,
0.0 ~"~ ~ xX
i
'13
"O "O "O "O
c e- r- e- O
O O
Figure 15 -- Results for two materials fracture toughness tested per C 1421(Geometry D of
CNB method in ambient air and room temperature at O.05 mm/min) Note." error bars
represent one standard deviation
Implementation of C 1421 in at least two cases has shown the robustness of the
standard in not only helping establish credible fracture toughness values but also the
ability to discern real differences in fracture toughness from billet to billet of the same
material, but also between two different materials.
In conclusion, C 1421 provides the ceramic community with not only a precise test
method, but also a test method with which results can be obtained that are accurate for
design purposes. The obvious implication of the success of C 1421 is the extension of
the test methods contained in this national standard to international standards. Efforts in
this regard have already been successful for the SEPB method, with progress under way
for the SCF method and activity initiated for the CNB method. As these international
efforts progress, revisions and refinements to C 1421 will be made to reflect new lessons
learned. Finally, logical extensions of C 1421 is a fracture toughness standard for elevated
temperature and a standard to assess R-curve behavior.
Acknowledgments
The authors wish to thank the members of ASTM Committee C28 on Advanced
Ceramics for their support during the development of C 1421 within the voluntary
standardization process of ASTM. The constructive criticism, both technical and
literary, while at times trying eventually lead to the high quality, technically rigorous
document which even now is forming the basis of forthcoming fracture toughness
standards nationally and internationally.
References
[1] Irwin, G. R., "Fracture I" Handbuch derPhysik, Vol 6, Springer-Verlag Berlin, 1958,
pp. 558-590.
[2] Jenkins, M. G. and Salem, J. A., "Fracture Resistance Testing of Brittle Solids,
"Mechanical Testing and Evaluation, ASM Handbook Vol 8, ASM International, Metals
Park, OH, 2000, pp. 665-678.
[3] Quinn, J. B. and Quinn, G. D., "Hardness and Brittleness of Ceramics," Ceramic
Engineering andScience Proceedings, Vol. 17, No. 3, 1996, pp. 59-68.
[4] Quinn, J. B. and Quinn, G. D., "Indentation Brittleness of Ceramics: A Fresh
Approach," Journal of Materials Science, Vol. 32, 1997, pp. 4331-4346.
[5] Freiman, S. W., "The Reasons Behind Mechanical Test Procedures for Ceramics or
Why We Measure What We Measure," Mechanical Testing Methodology for Ceramic
Testing and Reliability, D.C. Cranmer and D. W. Richerson, Eds., Marcel Decker
Publishers, New York, NY, 1998, pp. 1-16.
[6] Quinn, G. D., "ASTM Committee C-28 Advanced Ceramics: The First Ten Years,"
American Ceramic Society Bulletin, Vol. 75, No. 12, 1996, pp. 45-46.
[7] Quinn, G. D., Salem, J. A., Bar-on, I., and Jenkins, M. G., "The New ASTM Fracture
Toughness of Advanced Ceramics Standard: PS 070-97," Ceramic Engineering and
Science Proceedings, Vol. 19, No. 3, 1998, pp. 565-578.
[8] Quinn, G. D., Jenkins, M. G., Salem, J. A., and Bar-On, I., "Standardization of
Fracture Toughness Testing of Ceramics in the United States," Journal of Korean
Ceramic Society, Vol. 4, No. 4, 1998, pp. 311-322.
[9] Bar-On, I, Quinn, G. D., Salem, J. A., and Jenkins, M. G., "Fracture Toughness
Standard Test Method C 1421-99 for Advanced Ceramics," Fatigue and Fracture
Mechanics." 32nd Volume, ASTM STP 1406, R. Chona, Ed., American Society for Testing
and Materials, West Conshohocken, PA, 2001, pp. 315-335.
[10] Salem, J. A., Ghosn, L. J, and Jenkins, M. G., "Back-face Strain as a Method for
Monitoring Stable Crack Extension," Ceramic Engineering and Science Proceedings, Vol.
19, No. 3-4, 1998, pp. 587-594.
[11] Hoagland, R. G., Marschall, C. W., and Duckworth, W. H., "Reduction of Errors in
Ceramic Bend Tests," Journal of the American Ceramics Society, Vol. 59, No. 5-6, 1976,
pp. 189-t92.
[12] Baratta, F. I., "Requirements of Flexure Testing of Brittle Materials," U.S. Army
Materials Technology Laboratory, Technical Report, TR 82-20, Watertown, MA, 1987.
[13] Baratta, F. I, Quirm, G.D.,and Matthews, W.T., "Errors Associated with Flexure
Testing of Brittle Materials," U.S. Army Materials Technology Laboratory, Technical
Report, TR 87-35, Watertown, MA,1987.
[14] Quinn, G. D., "Twisting and Friction Errors in Flexure Testing," Ceramic
Engineering andScience Proceedings, Vol. 13, No. 3-4, 1992, pp. 319-330.
[15] Baratta, F. I and Fett, T., "The effect of load and crack misalignment on stress
intensity factors for bend-type fracture toughness specimens,"Journal of Testing and
Evaluation, Vol. 28, No. 2, 2000, pp. 96-102.
[16] Baratta, F. I., and Dunlay, W. A., "Crack Stability in Simply Supported Four-Point
and Three-Point Loaded Beams in Brittle Materials," Mechanics of Materials, Vol. 10,
1990, pp. 149-159.
[17] Bar-on, I., Baratta, F. I., and Cho, K., "Crack Stability and its Effect on Fracture
Toughness of Hot Pressed Silicon Nitride Beam Specimens," Journal of American
[18] Awaji, H., Kon, J., and Okuda, H., "The VAMAS Fracture Toughness Test Round
Robin on Ceramics," VAMAS Report #9, Japan Fine Ceramic Center, Nagoya, Japan,
1990.
[19] Quinn, G., Salem, J., Bar-on, I., Chu, K., Foley, M., and Fang,, H., "Fracture
Toughness of Advanced Ceramics at Room Temperature," Journal of Research of
National Institute of Standards and Technology, Vol. 97, No. 5, 1992, pp. 579-607.
[20] Quinn, G. D., Ktibler, J. J. and Gettings, R. J., "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.
[21] Quinn, G. D., Getting,s, R. J., and Ktibler, J. J., "Fracture Toughness of Ceramics by
the Surface Crack in Flexure (SCF) Method," Fracture Mechanics of Ceramcs, 1Iol11, R.
C. Bradt, D.P.H. Hasselman, D. Munz, M. Sakai, and V. Yashevechenko, Eds., Plenum
Press, New York, NY, 1996, pp. 203-218.
[22] Srawley, J. E., "Wide Range Stress Intensity Factore Expressions for ASTM E399
Standard Fracture Toughness Specimens," International Journal of Fracture Mechanics,
Vol. 12, 1976, pp. 475-485.
[23] Srawley, J. E., and Gross, B., "Side-Cracked Plates Subject to Combined Direct and
Bending Forces," Cracks and Fracture: Proceedings of the Ninth National Symposium on
Fracture Mechanics, ASTMSTP 601, J. L. Swedlow and M. L. Williams, Eds., American
Society for Testing and Materials, West Conshohocken, PA, 1976, pp. 559-579.
[24] Murayama, N., Sakaguchi, S., and Wakai, F., "Evaluation of Fracture Toughness
using SENB Specimens with Precrack Produced by Growing Microcrack Around Vicker's
Indent," Journal of Ceramic Society of Japan, International Editions, Vol. 95,1987, pp.
980-982.
[25] Newman, J. C. Jr. and Raju, I. S., "An Empirical Stress-Intensity Factor Equation for
the Surface Crack," Engineering Fracture Mechanics, Vol. 15, No. 1-2, 1981, pp. 185-
192.
[26] Petrovic, J. J., Dirks, R. A., Jacobson, L. A., and Mendiratta, M. G., "Effect of
Residual Stress on Fracture from Controlled Surface Flaws," Journal of American
Ceramic Society, Vol. 59, No. 3-4, 1976, pp. 177-178
[27] Quinn, G. D., Gettings, R. J. and Kiibler, J. J., "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.
[28] Swab, J. J., and Quinn, G. D., "Investigation of 'Halos' Associated with Fracture
Toughness Precracks," Ceramic Engineering and Science Proceedings, Vol. 18, No. 4,
1997, pp. 173-182.
[29] Tracy, C. A. and Quinn, G. D. "Fracture Toughness by the Surface Crack in Flexure
(SCF) Method," Ceramic Engineering and Science Proceedings, Vol. 15, No. 5, 1994, pp.
837-845.
[30] Wereszczak, A. A., Ferber, M. K., Sanders, R. R., Jenkins, M. G. and Khandelwal,
P. "Fracture Toughness (KIc and ~tWOF)of a HIPed Si3N4 at Elevated Temperatures,"
Ceramic Engineering andScience Proceedings, Vol. 14, No. 3-4, 1993, pp. 101-112.
[31] Jenkins, M. G., Ferber, M. K, Ghosh, A, Peuusa, J. T., Salem, J. A., "Chevron-
Notched Flexure Tests for Measuring the Elevated Temperature Fracture Resistance of
Structural Ceramics," Chevron-Notch Fracture Test Experience: Metals and Non-Metals,
ASTMSTP 1172, K. R. Brown and F.I. Baratta, Eds., American Society for Testing and
Materials, West Conshohocken, PA, 1992, pp. 159-177.
[32] Calomino, A. C. and Ghosn, L. G., "Optimal Configuration for Chevron-Notched
Four-Point Bend Specimen," International Journal of Fracture, Vol. 72, No. 4, 1995, pp.
113-116.
[33] Yell, H., Solidum, E, Schienle, J., Karasek, K., and Salem, J., "Processing and
Mechanical Behavior of Whisker-Toughened Silicon Nitride," Proceedings of the 28th
Annual SAE Automotive Technology Development Contractors Coordination Meeting,
P243, Society of Automotive Engineers, Warrendale, Pennsylvania, 1990, pp. 255-259.
[34] Mizuno, M and Okuda, H., "VAMAS Round Robin on Fracture Toughness of
Silicon Nitride at high Temperature," VAMAS Technical Report 16, Japan Fine Ceramics
Center, Nagoya, Japan, 1993.
[35] Jenkins, M. G, Kobayashi, A. S., White, K. W. and Bradt, R. C., "A 3-D Finite
Element Analysis of a Chevron-Notched, Three-Point Bend Fracture Specimen for
Ceramic Materials," International Journal of Fracture, Vol. 34, 1987, pp. 281-295.
[36] Joch, J., Zemankova, J. and Kazda, J., "Analysis of a Chevron-Notch Four-Point
Bend Specimen by Three-Dimensional Finite Element Method," Communications of the
American Ceramic Society, Vol. 71, No. 3, 1988, pp. C154-C155.
[37] Kohle, R., Hui, C. Y., and Zehnder, A. T., "Effects of Finite Notch Width on the
Fracture of Chevron-Notched Specimens, International Journal of Fracture, Vol. 94, No.
2, 1998, pp. 189-198.
[38] Salem, J. A., Ghosn, L. J., Jenkins, M. G. and Quinn, G., "Stress Intensity Factor
Coefficients for Chevron-Notched Flexure Specimens and Comparison of Fracture
Toughness Methods," Ceramic Engineering and Science Proceedings, Vol. 20, No. 3,
1999, pp 503-512.
[39] Jenkins, M. G., Chang, T., and Okura, A, "A Simple Machining Jig for Chevron-
Notched Specimens," Experimental Techniques, Vol. 12, No. 8, 1988, pp. 20-22.
[40] Bluhm, J. I., "Slice Synthesis of a Three-Dimensional 'Work of Fracture' Specimen,"

Engineering Fracture Mechanics, Vol. 7, 1975, pp. 593-604.
[41] Munz, D., Bubsey, R. T., Shannon, J. L. Jr., "Fracture Toughness Determination of
A1203 Using Four Point Bend Specimens with Straight Through and Chevron Notches,"
Journal of American Ceramic Society, Vol. 63, No. 5-6, 1980, pp. 300-305.
[42] Munz, D., Bubsey, R. T., Srawley, J. E., "Compliance and Stress Intensity
Coefficients for Short Bar Specimens with Chevron Notches," International Journal of
Fracture, Vol. 16, No. 4, 1980, pp. 359-374.
[43] Salem, J. A., Shannon, J. L. Jr., and Jenkins, M. G., "Some Observations in Fracture
Toughness Testing with Chevron-Notched Specimens," Chevron-Notch Fracture Test
Experience: Metals andNon-Metals, ASTMSTP 1172, K.R. Brown and F.I. Baratta, Eds.,
American Society for Testing and Materials, West Conshohocken, PA, 1992, pp. 9-25.
[44] Quinn, G. D., Xu, K, Gettings, R., Salem, J. A., and Swab, J., "Standard Reference
Material 2100: Ceramic Fracture Toughness," Fatigue and Fracture Mechanics: 32nd
Volume, ASTM STP 1406, R. Chona, Ed., American Society for Testing and Materials,
West Conshohocken, PA, 2001, pp. 336-350.
George D. Quinn, 1 Kang Xu, 2 Robert J. Oettings, s Jonathan A. Salem, 4 and Jeffrey J.
Swab 5
Does Anyone Know The Real Fracture Toughness?

SRM 2100: The World's First Ceramic Fracture Toughness Reference Material
Reference: Quinn, G. D., Xu, K., Gettings, R. J., Salem, J. A., Swab, J. J. "Does
Anyone Know the Real Fracture Toughness? SRM 2100: the World's First
Ceramic Fracture Toughness Reference M a t e r i a l , " Fracture Resistance Testing of
Monolithic and Composite Brittle Materials. ASTM STP 1409. J. A. Salem, G. D. Quinn,
and M. G. Jenkins, Eds., American Society for Testing and Materials, West
Abstract: Standard Reference Material (SRM) 2100 is the first reference material in the
world for the property fracture toughness. The SRM was developed in tandem with the
new ASTM test method C 1421-99. The SRM also supports two International Standards
Organization (ISO) draft standards now under development in ISO Committee TC 206,
Fine Ceramics. The SRM is a set of five silicon nitride test specimens. The SRM may be
used with any credible ceramic fracture toughness test method, but has been optimized
for test methods that use beams in bending. Identical results are obtained with single-
edge precracked beam, chevron-notched beam, and surface crack in flexure methods.
SRM 2100 should dramatically improve fracture toughness testing and enhance data
quality around the world. This paper describes SRM 2100 and its creation.
Keywords: fracture toughness, ceramics, surface crack in flexure, chevron notch,

precracked beam, reference material, round robin, VAMAS
Introduction and Background
SRM 2100 is intended for verification of ceramic fracture toughness testing procedures
and is an ideal complement to ASTM Test Method for Determination of Fracture
Toughness o f Advanced Ceramics at Ambient Temperature (C 1421-99) as well as three
draft ISO standard test methods that are under development in ISO Technical Committee
Staff Scientist, Ceramics Division, Stop 8521, NIST, Gaithersburg, MD, 20899.
2 Engineer,Inland Steel, R+D Center, MC 9-00, 3001 E. Columbus Dr., East Chicago, IN 46312
3 Staff Officer, Standard Reference Materials Program Office, National Institute of Standards and
Technology, Gaithersburg, MD 20899
4 Materials Research Engineer, Life Prediction Branch, Structures Division, MS 49-7, NASA-Glenn
Research Center, Cleveland, OH 44135
s Team Leader, Weapons and Materials Research Directorate, U.S. Army Research Laboratory,
Aberdeen Proving Ground, MD, 21005.
76
QUINN ET AL. ON SRM 2100-REFERENCE MATERIAL 77
TC 206, Fine Ceramics: ISO DIS 15732, Fine Ceramics (Advanced Ceramics, Advanced
Technical Ceramics) - Test Method for Fracture Toughness at Room Temperature by
Single-edged Precracked Beam (SEPB) Method; ISO CD 14704, Test Method for
Fracture Toughness at Room Temperature by Surface Crack in Flexure (SCF) Method;
and ISO new work item Test Method for Fracture Toughness at Room Temperature by
Chevron Notched Beam (CNB) Method.
The SRM is a set o f five unprecracked ceramic test specimens cut from a single plate
(billet) o f hot-pressed silicon nitride (HPSN), Norton grade NC 1326.7 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. The SRM may
be used with any fracture toughness test method, but has been optimized for beam
bending tests. The SRM data base was generated with the same three methods that were
adopted in C 1421. The three procedures, illustrated in Figure 2, 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 o f
C 1421 and its precursor, Provisional Standard PS 070, s are documented elsewhere [1,2].
Work on the test method standard commenced in Committee C-28 in 1991 whereas
work on the reference material began at NIST in May 1997 shortly after PS 070 was
approved. SRM 2100 was completed in June 1999, just about the same time (April 1999)
that the full consensus C 1421 was approved. The simultaneous development of the
reference material and the test method was mutually beneficial. For example, SRM
results supported a change in the stress intensity factor coefficients for two of the four
chevron-notched beam geometries in C 1421-99 as described in Reference[3]. 9
A fracture toughness reference material was first contemplated in late 1994 after the
completion o f a very successful Versailles Advanced Materials and Standards (VAMAS)
international round robin [4,5]. The round robin featured the SCF method. Twenty
laboratories obtained good results on specimens from a NC 132 single billet (designated
"E") [6,7,8]. A grand average fracture toughness o f 4.59 MPa~/m with a standard
deviation o f 0.37 was obtained from 107 experiments. The test method precision for 102
valid outcomes was evaluated using ASTM E 691-92, Practice for Coordinating an
Interlaboratory Study to Determine the Precision o f a Test Method. The within-laboratory
precision (repeatability standard deviation) was a low 0.24 MPa~/m (coefficient o f varia-
6 Norton Co., Worcester, MA. NC 132 is no longer available from the manufacturer.
7 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.
PS 070-97, which had the same title as C 1421-99, was a Provisional Standard that was adopted in
April 1997 by Committee C-28. It was converted to the full consensus standard C 1421 in April 1999.
Over 100 changes were made to the former to create the latter.
9 Thishighlights one of the important differences between PS 070 and C 1421. PS 070 had slice model
Yma.*solutions for specimenchevron configurationsA and D. It is now believed that the straightthrough
crack assumption (STCA) model is more accurate than the slice model for these two geometries[3]. C
1421 includes the new STCA formulationfor Yn" for test configurationsA, B, and D.
SRM 2 1 0 0
Ceramic Fracture Toughness
Figure 1-Standard Reference Material 2100, Ceramic Fracture Toughness.
Figure 2-The three test methods in C 1421 used to generate the SRM data base. The
SCF specimens may be tested in either o f two orientations. The specimen cross-section
size is 3 mm x 4 mm.
QUINN ETAL. ON SRM 2100-REFERENCEMATERIAL 79
tion of only 5.4 %). The between-laboratory precision (reproducibility standard

deviation) was 0.31 MPa~/m (coefficient of variation of only 6.8 %). These are very
consistent results for ceramic fracture toughness. Some laboratories tried other methods
such as the CNB test on spare specimens furnished as part of the round robin and they
obtained results that agreed with the SCF outcomes. Indeed, it was shown that the round
robin results concurred with 25 years of other NC 132 data from many laboratories that
used credible test methods. Furthermore, results by the three test methods in C t421
converged in the range of 4.5 MPa~/m to 4.9 MPa~/m. The uniformity of NC 132 even
extended to its creep and static fatigue resistance [9,10]. In light of the successful VAMAS
SCF round robin, NIST decided to investigate the suitability of NC 132 HPSN as a standard
reference material.
The Material
Silicon nitride grade NC 132 is a 1970 - 1980s vintage material [11] that was hot-pressed
with a small amount (~ 1 wt%) of a magnesia-based sintering aid. It is composed of slightly
elongated beta phase silicon nitride grains with a maximum size of about 3 micrometers as
shown in Figure 3. 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 from pick up during ball
milling of the starting powders. The material is fully dense and has a density 3.22 x 10+3
kg/m 3 - 3.26 x 10+3 kg/m 3, 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 m m x 158 nun x 20 nun - 30
mm thick. Beam specimens 45 m m - 47 mm long were sliced from the billet such that their
3 m m x 4 mm cross section planes in which cracks would later be initiated and propagated
were parallel to the hot-pressing direction. (Planes that are perpendicular to the hot
pressing direction have lower fracture toughness, due to a preferred orientation of the beta
silicon nitride grains that tend to lie flat in the plane, normal to the hot-pressing direction.)
NC 132 fractures in a mixed transgranular and intergranular mode at room temperature. It
has a flat R-curve (crack growth resistance is independent of crack size) and is highly
resistant to environmentally assisted slow crack growth at room temperature. Hence,
different test methods should produce identical fracture toughness results.
NC 132 is no longer made by the original manufacturer, so 582 specimens were prepared
from six billets that we had on hand. After extensive testing, it was determined that
specimens from only three billets, labeled "C", "G", and "D" for short, would be used for
the SRM. SRM 2100 comprises specimens exclusively from billet C. When SRM 2100 is
expended, specimens from billets G and D will be used for SRM 2100A and SRM 2100B,
respectively. Three additional billets, "H," "A," and "4," were also evaluated but ultimately
were rejected for the SRM. Billets A and 4 had excessive variability in toughness that was
traced to a material inhomogeneity that manifested itself on the specimen surface as a faint
patchwork mottling. Such mottling, commonly detected in hot-pressed silicon nitfides with
rare earth oxide sintering additives, has even been reported in magnesia-doped NC 132. In
Figure 3-The microstructure of the silicon nitride. (a) shows a fracture surface and is
from the inside of a SCF precrack. (b) is a polished section.
the case of NC 132 it is associated with elemental silicon from thermal decomposition of
the silicon nitride [12].
One might well ask why we used such an old, obsolescent material for the SRM. NC
132 silicon nitride was one of the most thoroughly studied, advanced ceramic materials of
all time. We felt it would be wiser to work with a reliable, consistent material for which
an enormous experience base existed, rather than risk using a newer transient material.
NC 132's fracture toughness is representative of many structural ceramics whose fracture
toughness values range from 1.0 MPa~/m to 10.0 MPa'Jm. Commercial derivatives of NC
132 are used in seal and bearing applications around the world.
Test Methods and Preliminary Findings
A large sample set of flexural specimens from each of the billets were tested at NIST by
the SCF and SEPB methods in accordance with ASTM C 1421 or PS 070 unless otherwise
noted. In addition, a few specimens from billets C and H were tested by the CNB method at
NIST and NASA-Glenn. As we will show, fracture toughness was independent of test
method. A substantial number of additional experiments using nonstandardized
procedures were conducted but were not used for the SRM data base. Data from these
supplemental experiments were very important for establishing the robustness of the test
methods and for confirming the tolerances and procedures in the ASTM and ISO
standards.
One of the advantages of the three methods in the C 1421 fracture toughness standard is
that they use the same genericflexural strength specimen. Each 3 mm x 4 mm x 45 mm -
47 mm bar specimen was prepared in accordance with the "standard" preparation procedure
specified in the flexuml strength standards: MIL STD 1942 MR "Flexuml Strength of High
Performance Ceramics at Ambient Temperature," U. S. Army Materials Technology
Laboratory, Watertown, MA, Nov. 1983; and ASTM Standard Test Method for Flexural
Strength of Advanced Ceramics at Ambient Temperature, C 1161. One 4 nun x 45 mm
wide face was given a 900 grit finish (finer than is normally required) to produce a very
flat surface for the Knoop indentation for the SCF test method. The four long edges of
the specimens were chamfered in accordance with requirements for the SCF and SEPB
test methods.
Groups of five specimens were precracked and then fractured per test session. Fracture
testing was in laboratory ambient conditions (20~ - 25 ~C; relative humidities, 30 % - 72
%), with the exception that some specimens were tested in dry-nitrogen gas in order to
ascertain whether environmentally-assisted slow crack growth might have been active.
Slow crack growth was not detected in billets C, G, or D specimens, but it was active in
some billet H specimens. This billet had a slightly greater concentration of glassy
boundary phase than the other billets. Conventional flexure strength (bend) fixtures with
20 m m x 40 mm spans usually were used. Some broken halves were retested in the SEPB
configuration in either three- or four-point flexure strength fixtures with a 20 mm outer
span. Fixtures were either semi- articulating or fully-articulating and were in accordance
with the well-known standard flexural strength test standards ASTM C 1161 and MIL STD
1942 which have critical requirements for fixture alignment and articulation, load pin and
specimen alignment, and rolling pin action to eliminate frictional constraint.
SCF Precracking and Testing
The SCF method, also known as the "controlled flaw" method, is a development of early
work by Petrovic and colleagues [13,14]. This method utilizes a Knoop hardness indenter
to create a single surface micro precrack. The indentation causes a very deep semi-elliptical
crack to open up underneath the hardness indentation in brittle ceramic materials and
glasses. The crack is much deeper than the indentation. Crack orientation may be
controlled by orienting the Knoop indentation axis. An essential step is the removal of the
indentation and its residual stresses by polishing or hand grinding after indentation. The
specimen is then fractured in a conventional flexural strength test fixture. Fractographic
techniques are used to detect and measure the precrack after specimen fracture. Fracture
toughness is computed from the maximum stress (or maximum load) at fracture, the crack
size, and the appropriate stress intensity shape factor. Precrack size must be measured for
each individual specimen. The stress intensity shape factor should be individually
calculated for each specimen. Additional details are in the literature [6-&15].
This method should not be confused with other less rigorous indentation test methods
that are widely used by the ceramics research community. The SCF method has the
ingredients of a good fracture mechanics test: a well-defined crack, a good stress intensity
factor solution, and a well-controlled fracture. Many of the alternative indentation methods
do not meet these basic requirements, are notoriously unreliable, and often produce inacur-
Figure 4-,4 precrack on the fracture surface of a surface crack in flexure test specimen.
rate or imprecise results. Many of these crude methods use a Vickers indenter to create a
so-called controlled flaw. In reality, the Vickers precrack site has a complex network of
radial, median, and lateral cracks as well as significant residual stresses from a plastically-
deformed and shear-faulted damage zone. "Controlled-flaw" is certainly a misnomer in
these alternative methods.
Most of the SCF database was generated with a 24.5 N indentation force. Cracks were
usually put into a 4 mm x 45+ nun flat face that had been finished with the 900 grit wheel.
The smoother than normal surface facilitates uniform indentations and precracks. The
average semi-elliptical precrack size after indentation and hand-polishing removal of the
residual stress surface zone was approximately 50 ~m deep x 150 ~tm wide as shown in
Figure 4. One sample set was also evaluated with the indentation implanted into the
narrower 3 mm wide face, but there was no significant difference in toughnessP Hand
grinding was done with silicon carbide papers or with a diamond impregnated disk} 1
A few specimens from billet C had the residual stress damage zone machined off by a
surface grinding machine rather than by hand grinding. The identical mean fracture
toughness was obtained, but the standard deviation was considerably larger. 12 The
precracks were more irregular than those from hand ground specimens. Apparently the
machine grinding applied forces that altered the indentation precracks and possibly
propagated portions of the precrack. It was also difficult to guarantee that each individual
specimen had the correct amount of material removed with diamond wheel grinding
because specimens were mounted together and ground simultaneously.
Only four-point loading was used for SCF fracture tests because it would be very
difficult to align the tiny precrack with the middle load roller in three-point loading.
The long edges were chamfered to eliminate the chance that a chip or edge machining
cracking would cause fracture. Although small, these chamfers reduce the cross section
area and the second moment of inertia of the cross section about the specimen neutral axis.
The fracture toughness was corrected (0.7 % to 1.0 %) for the change in the moment of
inertia as directed in C 1421.
Every semi-elliptical precrack was photographed with a scanning electron microscope
(SEM), often with stereo pairs, at magnifications of 200X to 500X. If there was any
doubt about an interpretation, the second fracture half was also examined and
photographed. A SEM length calibration standard was photographed with each set of
~0 Six specimens from billet D with the indentationplaced into the narrow face produced an average
fracture toughness of 4.56 MPax/mwith a standard deviationof 0.16 MPa~m as compared to the grand
average fracture toughness of 4.50 MPa~/m_+0.23 MPa~Jm for outcomes for billet D. SCF billet E
specimens tested flat or on edge also produced the same outcomes as reported in the VAMAS round
robin final report [6].
~l One of the great misunderstandingsabout the SCF method is that meticulousand delicate polishingis
required. This is not true. Handgrinding or polishingmay be aggressive. The goal is to remove the
indentationand the damage zone underneath it and the final specimen finish is of no consequence.
Hand grinding with dry 220 to 320 grit silicon carbide abrasive papers on a 203 mm diameterrotary
grinding-polishingmachinetook no more than 2 to 3 minutesper specimen.
~2 The average fracture toughness of nine machined SCF specimenswas 4.63 MPa~m as compared to a
billet average fracture toughness of 4.59 MPax/m. The standard deviationswere 0.26 MPa~/mfor the
machined SCF specimens,but only 0.13 MPa~/mfor seventeenhand-ground SCF specimensand 0.15
MPa~/m for all data for the billet.
specimens under examination (usually five at a time) in order to correct for the
magnification distortion or error in the SEM. The magnification correction was usually
of the order o f 3 % to 5 %, but sometimes as large as 8 %.
It is beyond the scope of this paper to cover all experimental details, but a remarkable
interaction between the stress intensity shape factor and the measured crack size was
beneficial in mitigating any errors or uncertainties in the crack size interpretation or
measurements [6-8]. Our success rate in conducting the SCF experiments was of the
order 90% or more. That is to say, for each set of five specimens tested, we usually
obtained four or five valid outcomes. An invalid test usually occurred when the precrack
was too irregular or too difficult to measure.
PB (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 slit or indentation flaw [16]. A conventional bridge
precracker was used to pop in a precrack that was between 0,45W to 0.55W deep, where the
specimen thickness, W, was 4 mm. A single 98 N (10 kgf) Knoop indentation was used as
a starter flaw in a ground 3 nun wide face. The indented 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 was used to detect the pop-in which could be heard as a faint
snap.
Four-point loading with 20 mm x 40 mm spans was used exclusively with the 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 preloading the
specimens. No correction for specimen chamfers was made. After fracture, precrack sizes
were measured with a precision traversing stage with a resolution of 0.001 mm under a
stereo binocular microscope at 16X to 40X. The crack front was quite obvious on this fine-
grained, fully dense material, and dye penetration was unnecessary. Figure 5 shows a
typical fracture surface. A few specimen halves (from PB or SCF trials) were retested on
10 mm x 20 mm four-point fixtures, or on three-point fixtures with 20 mm spans. There
was no difference in outcomes for full-size specimens versus outcomes from the specimen
halves/3 The reject rate for the three-point configuration was greater, however, due to crack
twisting and alignment problems. It is also difficult to line up precisely the precrack in
three point fixtures with a short span. Specimens were rejected if either the precraek or the
final fracture crack were tilted too much. j4 86 % of billet C specimens were accepted, 8 1 %
from billet G, and 100% from billet D. Every specimen passed the precrack depth evenness
criteria in C 1421and ISO 15732.
is Five short four-point specimens from billet C had an average fracture toughness of 4.58 MPax/m-+
0.05 MPax/mand three three-point specimens had an average fracture toughness of 4.56 MPa~m _+
0.09 MPa~/mwhich concurred nicely with the billet average fracture toughness of 4.58 MPa'~m+0.15
MPax/m for 55 specimens. Short SEPB specimens from billet D also produced results consistent with
full-length SEPB and SCF outcomes.
14 C 1421-99 and 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 criteria.
QUINN ET AL. ON SRM 2100-REFERENCEMATERIAL 85
Figure 5-Fracture surfaces of a SEPB specimen from billet C. Low angle vicinal
illumination helps delineate the precrack boundary.
CNB testing
A limited number of specimens from billets C and H were tested by the chevron notch
method [17_]. SRM 2100 specimens may be used with C 1421 configurations "A" or "D."
For the SRM data base, "A" type notches were prepared by a professional machine shop to
the tolerances in C 1421. The notch width was less than 0.25 mm and was carefully
machined to ensure that each side of the notch lined up well. The notch tip in the A
configuration is 0.80 mm from the specimen end. The A configuration was chosen since it
has the most amount of relative 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 completely stable fracture had occurred.
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. For the two unstable specimens, instability was due to
a chevron notch-tip having been chipped during machining in one case; in the other case,
instability was due to glue accidentally being in place in the notch during strain gaging.
Several specimens were tested in three-point loading, but no meaningful difference in
three- versus four-point results was detected) 5 The NIST and NASA-Glenn results
concurred. Stress intensity factor coefficients (Ym~,') for the four-point specimens were
calculated on the basis of a straight-through crack assumption (STCA) model.
Two successful three-point outcomeshad an average fracture toughness of 4.52 MPadm + .05 MPa~/m
compared to 4.60 MPa~/m _+.13 MPa~m for the eight four-point chevron notch results and 4.58
MPax/m +.15 MPa~/mfor all accepted data for billet C.
20-
19-
18-
SRM 2100
17-
BilletC j~
16-
15-
14- average = 4.572 ~'/A
std dev. =0.112
13-
12-
r n=55
11-
10-
9-
LL
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.
Fracture toughness MPa~/m
Figure 6-Histogram of all 46 valid test results (shaded)for billet C, SRM 2100. An
additional 9 SCF data points, based upon specimens with the residual stress
removed by a surface grinder are superimposed (light bars).
10
9 SF~ 21008
BilletD
8 n = 37 datapo~s
auerage =4,485
stddev. =0.232
7
\\
6 \\
\\
~\,\\
4 x\,\\
\\~ \\
3 ~x.\\
.\,\\
2
0 ~
,\\\-
L I i I i i i
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 7-Histogram of all 37 valid test results for billet D, SRM 2100B.
Results
Table 1 shows the results by test method for the three billets. Fifty-eight specimens from
billet C were tested (sometimes multiply) for a total of 75 experiments. The SRM 2100
database uses 46 outcomes from this billet C data. An additional 234 specimens were
available for SRM 2100 distribution as 46 kits of 5 specimens. Thirty-six specimens from
billet G were tested and yielded a database of 32 valid experiments leaving 215 specimens
available for the SRM 2100A. Twenty-five specimens from billet D were tested and
produced 37 test results for SRM 2100B. 16 Only 90 specimens remain for the SRM since
portions of this billet were used for other work.
Within a given billet, the results for all data usually were in extraordinary agreement. As
noted above, we usually conducted tests in sets of five specimens, the same number in a
SRM kit. The mean fracture toughness for five specimen sets usually concurred within 1%
to 2 %, irrespective o f test method. Even the standard deviations for the five specimen sets
were very consistent. Fracture toughness varied very little within a billet. Figure 6 shows
the splendid results distribution for billet C. Billet D data, shown in Figure 7, have a little
more variability.
The mean fracture toughness varied between the billets, however. The billets C and D
toughness results are statistically significantly different than billet G. The source of the
variability between billets is subtle and could not be traced to any obvious density,
chemistry, or microstructural difference. This remarkable data consistency confirms that
the C 1421 methods have been optimized to the extent that genuine material variability
(billet-to-billet, or batch-to-batch) can now be discerned.
The SRM kits of five specimens will be prepared with specimens from a single billet on
account of the between-billet variability.
Table 1 -- Fracture toughness data for hot-pressed silicon nitride billets C, D, and G by the three
test methods in C 1421-99. n is the number of experiments
Billet Method Mean Std. Dev. n

MPmJm MPm/m
C SRM 2100 SCF 4.56 0.13 17
SCF (4.58) (0.16) (26)*
C SRM 2100 CNB 4.58 0.12 10
C SRM 2100 SEPB 4.58 0.10 19
G SRM 2100A SCF 4.29 0.19 15
G SRM 2100A SEPB 4.27 0.16 17
D SRM 2100B SCF 4.53 0.22 19
D SRM 2100B SEPB 4.46 0.24 18
* SCF values in parenthesis include 9 specimens wherein the damage

zone and residual stresses were removed by machine grinding. These
were not used in the master database.
,6 Someof the specimen halves were retested.

88 FRhCTURETESTING OF MONOLITHIC/COMPOSITEMATERIALS
SRM 2100 Uncertainty
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. The mean
fracture toughness and the scatter in results as measured by the three test methods were
statistically indistinguishable and the data therefore were pooled. The certified average
fracture toughness in Table 2 is the grand average of the pooled NIST database. The
uncertainty U~ is a 95 % prediction uncertainty for a single future observation and is based
on the results of the NIST independently and randomly selected observations from the same
normally distributed population [18]. The uncertainty U~ is a 95 % confidence interval for
the mean o f five future observations, also based on the results of the NIST independently
and randomly selected observations. The expanded uncertainties U represent the combined
effects of material inhomogeneity and within-laboratory components o f uncertainty and
were calculated from
l~z
U=t(l+n) S
where s is the grand NIST sample standard deviation, m = 1 for a single specimen, or m = 5
for the SRM kit sample size, n is the NIST sample size and t is the student's t parameter
(~2.) for a 95 % confidence interval and the appropriate degrees of freedom, (n-l). Table 3
shows that the smallest uncertainties are associated with SRM 2100 billet C since the
database was larger and the results less scattered than for the other two billets.
Table 2 - 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.
Grand Grand Uncertainty, Ul Uncertainty, Um

Billet Mean Std. Dev. Single specimen Mean of 5 specimens
MPa~tm MPa~/m MPa~/m MPa ~/m
C SRM 2100 4.572 0.112 0.228 (5.0 %) 0.106 (2.3 %)
G SRM 2100A 4.268 0.148 0.307 (7.2 %) 0.145 (3.4 %)
D SRM 2100B 4.495 0.232 0.477 (10.6 %) 0.224 (5.0 %)
An Application of the SRM: Accuracy of the new SEVNB Method
A popular new nonstandardized t e s t 17 is the single-edge V notched beam (SEVNB).

Method [19,20]. How accurate or precise are the results from this method? In the
SEVNB method, a conventional (saw cut) single-edge notched beam specimen is
prepared with a diamond wheel slit. The notch is sharpened with a common single edge
or shaving razor blade and fine diamond paste. The root of the notch is a "V" shape. The
achievable notch tip radius depends upon the material's microstructure, the diamond
paste grit, the condition of the razor blade and the razor- polishing machine set up, but
radii as small as 1 Ixm are possible. The notched specimen is fractured in four-point
bending. This method has widespread appeal since the "precracking" step is relatively
simple, but doubts remain as to whether the notch tip is genuinely sharp or whether stable
crack extension or whether irregular crack damage at the notch root cause complications.
This method was the focus of a recent major VAMAS international round robin.J21,22]
We tested five SRM 2100 billet C specimens by this method and obtained an average
toughness of 4.34 MPa~/m. This is 5.0 % less than the SRM 2100 certified value of 4.572
MPax/m (Table 3). K~ibler [20] independently measured five SEVNB specimens from
billet E and obtained an average fracture toughness of 4.36 MPa~/m. This outcome is also
exactly 5.0 % less than the 4.59 MPa~/m obtained in the large VAMAS round robin with
billet E specimens [5-7]. The mean SEVNB outcomes from billets E and C are very
consistent, but both are 5.0 % less than results from the mature standard methods. The
differences in the means are statistically significant at the 99.9 % level using the t-test.
SEVNB notch tip issues mentioned in the paragraph above probably account for the
discrepancy. Other experience with the SEVNB method confirms that SEVNB results
are usually within 5 % to 10 % of results obtained with sharp crack, standardized fracture
mechanics test methods [22].
Table 3 - Comparison of SEVNB results to standard test methods results, n is the number of
experiments.
SEVNB ASTM C 1421 standard methods

Billet Mean Std. Dev. n ref. method Mean Std. Dev. n ref.
MPaqm MPaqm MPa~/m MPa~/m
C 4.34 0.19 5 this i SCF,SEPB 4.57 0.11 46 this paper
SRM 2100 paper i CNB
E 4.36 0.08 6 [201i SCF 4.59 0.37 107 [6-8]
VAMAS
~7 A draft European Standard Committee(CEN) standard is in preparation in TC 184 that will include the
SEVNB as well as the SCF, SEVNBand CNB methods.
Conclusions
The first reference material in the world for the property fracture toughness has been
created in tandem with ASTM C 1421. It will also support emerging ISO standards.
C 1421 and SRM 2100 were completed within months of each other in mid-1999. The
finding that multiple refined test methods produce virtually identical results strongly
substantiates C 1421 and furthermore proves that a single fracture toughness exists for a
given billet of the SRM material. Methods for measuring fracture toughness have been
optimized to the extent that genuine material differences may now be discerned. This is one
goal of any standard: to enable the desired property to be measured with good accuracy and
precision.
The price per kit is (as this was written) was approximately $350 and covers the cost of
fabricating the specimens, SRM project overhead costs, as well as a small fraction of the
preliminary research costs. The SRM database (115 outcomes) is supported by years of
work on the same material. A further 125 experiments from the rejected billets A, H, and 4,
although not used for the database, nonetheless furnished valuable information. A total of
85 additional preliminary NIST experiments to support the development of C 1421 and set
up the VAMAS SCF round robin were invaluable. The 220 specimens tested by other
laboratories during the round robin were crucial in demonstrating that many laboratories
could obtain consistent results for the silicon nitride. All told, 582 fracture toughness
experiments on this particular material have been conducted over eight years in our ceramic
fracture toughness standardization program.
Eventually the ceramic community will begin using the new standard methods
although they do require somewhat more effort than the rudimentary indentation
methods. Although simpler to perform, the latter produce data of dubious value.
Hopefully the ceramic community will recognize that standard test methods generate high
quality data. The standards and SRM will restore some credibility to ceramic fracture
toughness testing.
C 1421 and SRM 2100 are based on quality test methods for quality data. Quality
methods are practical test methods, based on sound principles and good metrology, that
produce accurate and precise data.
Acknowledgments
Nien-Fan Zhang of the NIST Statistical Engineering Division assisted with the statistical
interpretations. Michael Foley of St. Gobain/Norton contributed with helpful consultations.
The U. S. Department of Energy, Office of Transportation Technologies, Heavy Vehicle
Propulsion Systems Materials Program is acknowledged for their support of the SCF test
method development under contract DE-AC05-g40R21400 with Martin Marietta Energy
Systems, Inc.
References
[1] Quiun, G. D., Salem, J. A., Bar-On, I., and Jenkins, M. G., "The New ASTM
Fracture Toughness of Ceramics Standard: PS 070-97," Ceramic Engineering and
Science Proceedings, Vol. 19, No. 3, 1998, pp. 565-578.
[2] Jenkins, M. G., Bar-On, I., Quinn, G. D., and Salem, J. A., "Development,
Verification, and Implementation of A National Full-Consensus Fracture Toughness
Test Method Standard for Advanced Ceramics," Fracture Resistance Testing of
Monolithic and Composite Brittle Materials. ASTM STP 1409, J. A. Salem, G. D.
Quinn, and M. G. Jenkins, Eds., American Society for Testing and Materials, West
[3] Salem, J. A., Ghosn, L., Jenkins, M. J., and Quinn, (3. D., "Stress Intensity Factor
Coefficients for Chevron Notched Flexure Specimens and a Comparison of Fracture
Toughness Methods," Ceramic Engineering and Science Proceedings, Vol. 20, No.
3, 1999, pp. 503-512.
[4] Quinn, G. D., "VAMAS After Twelve," Bulletin of the American Ceramic Society,
Vol. 78, No. 7, 1999, pp. 78-83.
[5] Quinn, G. D., "The Fracture Toughness Round Robins in VAMAS: What We Have
Learned," in Fracture Resistance Testing of Monolithic and Composite Brittle
Materials, ASTMSTP 1409. J. A. Salem, G. D. Quinn, and M. G. Jenkins, Eds.,
American Society for Testing and Materials, West Conshohocken, PA, 2002.
[6] Quinn, G. D., Kiibler, J. J., and Gettings, R.J., "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.
[7] Quinn, G. D., Gettings, R. J., and KiJbler, J. J., "Fracture Toughness of by the Surface
Crack in Flexure (SCF) Method: Results of the VAMAS Round Robin," Ceramic
Engineering andScience Proceedings, Vol. 15, No. 5,1994, pp. 846-855.
[8] Quinn, G. D., Gettings, R. J., and Klibler, J. J., "Fracture Toughness of Ceramics by
the Surface Crack in Flexure (SCF) Method;" Fracture Mechanics of Ceramics,
Vol.ll, eds. R. C. Bradt, D. P. H. Hassehnan, D. Munz, M. SakaJ, and V.
Yashevchenko, Plenum, NY, 1996, pp. 203 - 218.
[9] Quinn, G. D., "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.
[10] Quinn, G. D., and Quinn, J. B., "Slow Crack Growth in Hot Pressed Silicon Nitride;"
Fracture Mechanics of Ceramics 6, eds. R. Bradt, A. Evans, D. Hasselman, F. Lange,
Plenum Press, NY, 1983, pp. 603 - 636.
[11] Torti, M. L., "Processing Hot Pressed Silicon Nitride for Improved Reliability: HS
110 to NC 132," Ceramicsfor High Performance Applications, Ill Reliability, eds.,
E. Lenoe, R. N. Katz, and J. J. Burke, Plenum, NY, 1983, pp. 261-273.
[12] Clarke, D. R., "A Large Scale Processing Inhomogeneity in Silicon Nitride
Ceramics and Its Effect on Oxidation," Materials Science Forum, Vol. 47, Trans
Tech. Publ., Switzerland, 1989, pp. 110 -118.
[13] Petrovic, J. J., Jacobson, L. A., Talty, P. K., and Vasudevan, A. K., "Controlled
Surface Flaws in Hot-Pressed Si3N," Journal of the American Ceramic Society,
Vol. 58, No. 3-4, 1975, pp. 113 - 116.
[14] Petrovic, J. J. and Mendiratta, M. G., "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.
[15] Quinn, G. D., Get-tings, R. J., and Ktibler, J. J., "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.
[16] Nose T., and Fuji, T., "Evaluation of Fracture Toughness for Ceramic Materials by
a Single-Edge-Precracked-Beam Method, Journal of the American Ceramic Society,
Vol. 71, No. 5, 1988, pp. 328-333.
[17] Salem, J. A. and Shannon, Jr., J. L., "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. Brown and F.
Baratta, American Society for Testing and Materials, West Conshohoken, PA,
1992, pp. 9-25.
[18] Hahn, G. and Meeker, W., Statistical Intervals: A Guide for Practitioners, Wiley,
NY, 1991.
[19] KObler, J. J., "Fracture Toughness of Ceramics Using the SEVNB Method: From a
Preliminary Study to a Standard Test Method," Fracture Resistance Testing of
Monolithic and Composite Brittle Materials. ASTM STt' 1409. J. A. Salem, G. D.
Quinn, and M. G. Jenkins, Eds., American Society for Testing and Materials, West
[20] KUbler, J. J., "Fracture Toughness of Ceramics Using the SEVNB Method:
Preliminary Results," Ceramic Engineering and Science Proceedings, Vol. 18, No. 4,
1997, pp. 155-162.
[21] Ktibler, J. J., "Fracture Toughness of Ceramics Using the SEVNB Method: First
Results of a Joint VAMAS/ESIS Round Robin," ibid, Vol. 20, No. 3, 1999, pp. 495-
502.
[22] KObler, J. J., "Fracture Toughness of Ceramics Using the SEVNB Method; Round
Robin," VAMAS Report No. 37, Swiss Federal Laboratories for Materials Testing
and Research, Diibendorf, Switzerland, 2000.
Jakob J. Kfiblerm
Fracture Toughness of Ceramics using the SEVNB Method: From a Preliminary

Study to a Standard Test Method
Reference: Ktibler, J. J., "Fracture Toughness of Ceramics using the SEVNB

Method: From a Preliminary Study to a Standard Test Method," Fracture Resistance
Testing of Monolithic and Composite Brittle Materials, ASTM STP 1409, J. A. Salem, G.
D. Quinn, and M. G. Jenkins, Eds., American Society for Testing and Materials, West
Abstract: Numerous methods are currently in use to measure the fracture toughness of
ceramic materials. Methods based on widely accepted theory are often difficult to realize,
unreliable, or expensive. Recently, the single-edge-V-notched beam (SEVNB) method,
where a saw cut is tapered to a sharp V-notch using a razor blade sprinkled with diamond
paste, was reintroduced. The fracture toughness was measured with this method on five
advanced technical ceramics in an international round robin. It showed that the
repeatability and reproducibility of this method was very good. Further, the method
proved to be forgiving and robust with respect to notch preparation for ceramics having a
major microstructural feature greater than about 1 pm in size. Most participants had no
difficulties conducting the measurements and rated the SEVNB method as user-friendly,
easy and cheap to conduct, reliable, accurate, and worthwhile for standardization.
Therefore, a recommended practice has been written that is now serving as basis for a
new CEN standard.
Keywords: Fracture toughness, advanced technical ceramics, alumina, silicon nitride,

silicon carbide, zirconia, single-edge-V-notched beam, razor blade, diamond paste,
flexure
Introduction
If the fracture toughness of an advanced technical ceramic is to be measured, scientists

can choose from over a dozen test methods and uncountable variations thereof. In Japan
the single-edge-precracked beam (SEPB) and indentation fracture (IF) have been
LResearch scientist, Department High Performance Ceramics, EMPA, Swiss Federal

Laboratories for Materials Testing and Research, Ueberlandstr. 129, CH-8600
Diibendorf, Switzerland.
93
standardized. In the United States, the three methods have been standardized: surface
crack in flexure (SCF), chevron notch (CN), SEPB. During the standardization process at
least a half-dozen round robins were conducted in the last decade. They showed that in
general the currently standardized methods have some limitations with regard to user-
friendliness, and that the IF method is unreliable.
Standardized Methods
S C F - The SCF method is the only method which uses a sharp crack in the size range
of natural cracks. Due to the removal of the damaged zone around the hardness indent
from which the sharp cracks emanate, the test results are not influenced by residual stress
fields. Quinn, Gettings, and Ktibler [1] demonstrated in a Versailles Project of Advanced
Materials and Standards (VAMAS) round robin that accurate fracture toughness values
can be measured with the SCF method. However, they also showed that making the small
cracks and finding their crack fronts after the test can, depending on the material, range
from being simple to very challenging, if not impossible.
C N - The CN test is very simple to conduct because there is no need to measure a
crack size on the fractured surface of the test specimens. On the other hand, the
machining of the chevron notch is difficult if the stringent size and quality requirements
are to be fulfilled (e.g., ASTM Standard Test Methods for the Determination of Fracture
Toughness of Advanced Ceramic Materials at Ambient Temperatures C 1421-99). A
necessary criteria for the validity of a CN test is that evidence of stable crack extension
(stability) e.g. in the load vs. displacement or load vs. time plots must be seen. With
respect to stable crack extension, Primas and Gstrein [2] reported that in an European
Structural Integrity Society (ESIS) round robin only one laboratory had submitted valid
tests for MgPSZ and none had succeeded with HPSN, SSiC or Y-TZP. The poor success
rate may not have been a fundamental defect of the CN method, but instead of the
extremely vague guidelines to prepare the notches given in the round robin instructions.
Recently, Salem, Ghosn, Jenkins, and Quinn showed consistent stability if the notches
were narrow (< 0.3 ram) and well-machined [3].
SEPB - T h e rather simple and easy to conduct SEPB method, with its long and sharp
crack, was used in one ESIS and three VAMAS round robins. In general, a high
variability in the test results could be seen. Nishida, Hanaki, and Pezzotti [4] concluded
that practical problems with the SEPB method make the fracture toughness determination
difficult to apply and even unsuitable for some ceramics. Quinn [5] mentioned that the
SEPB method was accurate and reliable for laboratories that tested in three-point flexure
and had experience with this method. On the other hand, he states that even after all these
round robins, many participants outside Japan are still having great difficulty obtaining
the valid precracks necessary.
I F - The IF method is well-liked because only very small samples and a hardness
indenter are necessary to determine a toughness value. Unfortunately, as for example
Evans [6] states, the method and its many variations are only approximate and do not
provide the quality of fracture resistance data needed to rigorously relate toughness to
microstructure. Further, he believes that none of the indentation-based methods can be
used for the highest toughness materials now available.
KOBLER ON FRACTURETOUGHNESS USING THE SEVNB METHOD 95
Therefore, many questionable, inconsistent, and even erroneous results can be found
in published data. As a result, many scientific and industrial users of ceramics do not
trust fracture toughness testing and avoid measuring or applying such data. In order to
combat this distrust, a simple, user-friendly, reliable, accurate, widely accepted and
standardized fracture toughness test method is needed and being sought after.
SEVNB Method
The single-edge-V-notched beam (SEVNB) method is a further development of the

simple and user-friendly single-edge-notched beam (SENB) method. The latter method
can be influenced by the notch width, as e.g. Primas and Gstrein [2] discovered in an
ESIS round robin. In a more detailed analysis, Damani, Gstrein, and Danzer [7] suggested
that the stress field around the notch tip is responsible for the notch width effect. They
showed that the notch width must be on the order of the size of the relevant
microstructural or machining-induced defects. Nishida, Hanaki, and Pezzotti [4] recently
reintroduced an interesting technique described earlier by Le Bac [8] to taper a saw cut to
a sharp V-notch using a razor blade sprinkled with diamond paste. This method, known
as the SEVNB method, relates also to basic work conducted by Awaji and Sakaida [9].
With the aim to examine whether the SEVNB method is user-friendly, reliable, and
most important, comparable with other recognized methods, a preliminary study with
eight ceramics was conducted by the author [10] (Si3N4: NC-132, hot pressed Si3N4, hot-
isostatic pressed Si3N4, Si3N4 with SiCw, Y-TZP, SSiC, Mg-PSZ, and A1203). These
materials had all been used in previous international fracture toughness round robin tests.
After this promising study, detailed instructions were written and validated in a mini-
round robin with two laboratories (National Physical Laboratory, Teddington, UK;
Montan University, Leoben, A). Finally, more than 30 companies and institutes from
Europe, the United States, Japan, Australia, and Brazil participated in a fracture
toughness round robin conducted jointly under the ESIS and VAMAS organizations
(participants are listed in the acknowledgments).
Materials
In the round robin, five ceramic materials with varying fracture toughness
measurability were used. Each participant was required to test an alumina-998 and a gas-
pressure sintered silicon nitride (GPSSN) (Table 1). An alumina-999, a silicon carbide
(SSiC) and a yttria-stabilized tetragonal zirconia (Y-TZP) were optional.
Experimental Procedure
Each participant received a package containing five 3 x 4 x 45 nun alumina-998 and

GPSSN bend bars and instructions detailing how to conduct the round robin. Some
participants also received bend bars from the other three ceramics. All bend bars were
prepared in accordance with CEN (Comit6 Europ6en de Normalisation) standard
Monolithic Ceramics - Mechanical Properties - Part 1: Determination of Flexurai
Strength (EN 843-1). The participants had to measure the fracture toughness of these
bend bars with the SEVNB method. Further, participants who volunteered received
additional specimens of the aluminas and the GPSSN on which they were asked to
measure the fracture toughness using their preferred method (SCF, SEPB, CN, etc.).
Table 1 - Processing parameters and material properties.
alumina-998 alumina-999 GPSSN SSiC Y-TZP
Powder > 99.8 % > 99.9 % Bayer N3208 Tosoh TZ-3Y

pressed CIP 2 CIP 2 HP 3 CIP 2
sintered air air gas-pressure air
post hipped gas- gas-pressure
pressure
Grain size >10 lam -1.7 lam < 1 lam, elong. 7 lam 0.45 lam
Density 3.86 g/cm 3 3.97 g/cm 3 3.23 g/cm 3 3.15 g/cm 3 6.03 g/cm 3
Strength 342 MPa ~ 350 MPa ~ > 920 MPa 1 > 750 MPa 1
x Four-point bending
2 Cold isostatic pressed
3 Hot pressed
The fracture toughness with the SEVNB method had to be measured in four-point
bending tests with spans of 40 / 20 mm. Before testing, each participant was required to
cut the V-notches at the center of each specimen's tensile surface (Figure 1). The V-
notches could be produced either by hand or by automated means. The general procedure
could be performed in three steps:
1) Mount five specimens parallel and side by side with their compression surface down
on a plate;
2) With a thin diamond wheel cut a straight slot to a depth of about 0.5 mm; and
3) Polish a second deeper notch into this slot with a razor blade sprinkled with diamond
paste.
Afterwards, the participants had to compute the fracture toughness Kic using Equation 1
F S1- S2 3qt-~
Klc= B~/-W W " 2( 1 - 0t)1"5 Y• (1)
with
Y* = l . 9 8 8 7 - 1 3 2 6 a - ( 3.49-O.68a +135a2 ) a ( 1 - a ) ( l + a) -2 ( A S T M C 1421 -A2.5)
where
KOBLER ON FRACTURE TOUGHNESS USING THE SEVNB METHOD 97
F, Sx, a, B = fracture load, span, notch depth, specimen width, respectively;

W, a, Y* = specimen height, a/W, stress intensity shape factor, respectively.
Figure 1 - a) Schematic of V-notch geometry; b) V-notch polishing by hand.
Results and Discussion
Table 2 lists the grand population average (G.P.Avg.) and standard deviation
(G.P.Std.Dev.) for all materials tested.
Table 2 - Grand population average and standard deviation.
Material Participants Total # of G.P.Avg. G.P.Std.Dev.

(total) specimens MPa ~/m MPa x/m
alumina-998 28 (31) 135 3.57 _+0.22

alumina-999 21 (28) 102 3.74 _+0.40
GPSSN 27 (30) 129 5.36 _+0.34
SSiC 12 (12) 56 2.61 -4-0.18
Y-TZP 7 (7) 35 5.34 +_0.65
For a first validity check of the round robin test results, graphs as shown in Figure 2
were used. No significant influence of the notch width below 30 lam could be seen.
Therefore, a notch width of 30 lam was chosen as the criterion to accept or reject a result.
No results were rejected on the basis of either an incorrect V-notch geometry or the use
of a test procedure other than 4-point bending. The success rate of the accepted
participants conducting the SEVNB method was better than 93 % (4.7 specimens out of 5
per participant) for all the ceramics tested.
Notch Quality, Width, and Depth
Twelve participants polished the V-notches by hand and nineteen used some form of
machine. Typical polishing times for five specimens were 0.5 h for the aluminas and 1.5
h for the GPSSN and SSiC. To polish the notches, the participants used razor blades with
thicknesses ranging from 100 to 250 lam. The polishing stroke length varied from about 4
to 20 mm and the stroke frequencies from 1 to 15 Hz. The loads on the razor blades
ranged from 1 to 5 N and the diamond pastes used had grain sizes between 0.5 and 3 lam.
An analysis of all the information furnished by the participants did not give a clear
picture of the influence of these parameters. The general impression is that it is easy to
polish V-notches with a width between 20 and 30 lam, but the parameters need to be
optimized for each combination of polishing method (i.e. machine or hand), razor blade
type, stroke length, frequency and load in order to obtain notches smaller than 10 lam.
Alumina-998 "~ I Ab~mtna-999

=-4
I
~6.0 ,.,., 6.0 S
9 '
4.0 40 -ff~,--:-o--_-----~-.~-,-:--_-
?_--_-_?-
o.o
0
.... I . . . . . . .
I0 20 30
I, i:ii 40
:
50
~
~" o.o |
~
0
.... t ....
I0
I ....
20 30
--- G.P.Std.Dev.
i .... t ....
40 50
N o t c h W i d t h [pro] Notch Width [pm]
~ 8.0
GPSSN 34 8 25 ._~ 8.0 -t S S i C
6.0 z-L- ....... i.- ..... :r . . . . . . . -t- . . . . . . . . . ~6"0 t

. . . .
4.0 40 17 30
2.0 9 Lab. Avg. 2.0 t- /" 9 Lab. A~g.

- - G.P.Avg. d ] 7 - - G.P.Avg.
--- G.P.Std.Dev. J --- G.P.Std.Dev,
0.0 .... I .... I .... I .... i .... t~O.O .... I .... I .... I .... i ....
0 10 20 30 40 50 0 10 20 30 40 50
Notch Width [urn] Notch W i d t h [pm]
8.0
22 Y-TZP
6.0
~ 4.0
~ 2.0 9 Lab. Avg.

- - G.P.Avg.
--- G.P.Std.Dev.
~ o.o .... I .... I .... f .... t ....
10 20 30 40 50
Notch W i d t h [pm]
Figure 2 - Master result graphs. The bold dashed lines represent the grand population
average and the fine dashed lines the grand population standard deviation. Results from
tests with notches > 30 pm were not used to compute the G.P.Avg. and G.P.Std.Dev. The
numbers "identify" the participant responsible for that result.
KUBLER ON FRACTURETOUGHNESS USING THE SEVNB METHOD 99
That the fracture toughness measured on edge-notched bend bars can be influenced
by the notch width is well-known. From a practical point of view this had been
demonstrated, for example, by Primas and Gstrein [2] in an ESIS round robin. Munz and
Fett [11] found that the measured fracture toughness increases above a critical notch
width. Munz, Bubsy, and Shannon [12] reported a good agreement between critical stress
intensity values measured on an alumina by the CN and SENB methods in bending,
provided the notch width was less than 70 lam. Ktibler [10] found empirically that
fracture toughness values could be measured which were comparable with other methods
if the notch width was less than about twice the size of a major microstructural feature,
e.g., the average grain size. Interesting theoretical work by Fett [13], a study by Damani,
Gstrein, and Danzer [7] and an analysis by Fett [14] seem to confirm the criterion S <
2"amfs (S = acceptable notch width; a~s = major microstructural feature size) to estimate
an acceptable notch width for a material.
In order to provide additional support for this criterion, more GPSSN bend bars with
varying notch widths were tested to analyze the influence of the effective notch width
under identical test conditions. An increase in the measured fracture toughness was
noticeable even at very small notch widths. Figure 3 shows that an acceptable notch
width of 7 ttm calculated from the relation S _<2"amfs (amfs = 3.5 lam - length of elongated
grains) is in good agreement with the test data. Furthermore, the data also shows that the
measured fracture toughness does not drop out of the grand population standard deviation
band up to a notch width S of approximately 20 lam.
8.0 -g 8-0 4
CrSSN t "~ ~ GPSSN round robin
"g 6.0
~ 4.0 9 Avg. of 5 tests

--- G.P.Avg.
4,0-1
I'
;:
/is R 1607 (SEPB)
- - G.P.Std.Dev.
~ 2.0 2.0 ~ " l.,ab. Avg,
,3 ] - - G.P.Avg.
S in p m ---) 7 10 20 30 40 4 --- G.P.Std.Dev. + 3-point bendinl
OpO . . . . . ~ , , "r: ~] = , , 4 Ii ~ , p ,: ~ , r: , r I t~ 0.0/ ............................
0 l 2 3 4 5 0.0 0.1 0.2 0.3 0.4 0.5 0.6

(0.5 x S) ~ [pro ~ a / W [-]
Figure 3 - Influence of V-notch width S and depth ratio a/W on fracture toughness.
The participants had been asked to polish their notches to a depth between 0.8 and
1.2 ram, resulting in an a/W-ratio between 0.2 and 0.3. Apart from reducing the polishing
times, this narrow range of shallow crack depths was chosen to minimize the potential
influence of different notch depths on the measured fracture toughnesses. Schindler [15]
had shown that the measured fracture toughness can increase with an increasing crack
(notch) length. Additional bending tests conducted for the aluminas and the GPSSN with
a/W-ratios between 0.1 and 0.55 showed no significant influence of the notch depth
(Figure 3). Therefore, it is assumed that between a/W = 0.2 and 0.5 none or only a small
notch depth dependence of the measured fracture toughness exists.
Repeatability and Reproducibility
In order to determine the repeatability (within-laboratory) and reproducibility

(between-laboratories) of the SEVNB method, the ISO (International Organization for
Standardization) standard Accuracy (trueness and precision) of Measurement Methods
and Results (ISO 5725-2) was used. The statistical results are shown in Table 3 and
compared with results from an earlier VAMAS round robin conducted with the SCF
method [1]. The precision and bias statements for the SCF method were computed on the
same basis, but in accordance with ASTM Standard Practice for Conducting an
Interlaboratory Study to Determine the Precision of a Test Method (E 691), which
computes repeatability and reproducibility in the identical manner. The data in Table 3
show that the repeatability and reproducibility of the SEVNB method compares very well
with the SCF method.
Table 3 - Repeatability and reproducibility.
Repeatability Reproducibility
(within-lab) (between-lab)
Total Total Std.Dev. 2 CV 3 Std.Dev. 2 CV )
Material Method Part. Spec. MPa "4m % MPa ~/m %
alumina-998 SEVNB 28 135 0.17 4.6 0.22 6.1

alumina-999 SEVNB 21 102 0.23 6.2 0.40 10.7
GPSSN SEVNB 27 129 0.28 5.3 0.34 6.3
SSiC SEVNB 12 56 0.12 4.5 0.18 6.8
Y-TZP SEVNB 7 35 0.33 6.2 0.68 12.7
hot pressed Si3N 4 SCF 1 19 102 0.24 5.4 0.31 6.8

hot iso-pressed Si3N4 SCF l 15 100 0.38 7.7 0.45 8.9
1 earlier VAMAS round robin

2 standard deviation
3 coefficient of variance
In a next step the results of the participants were assessed with respect to stragglers
(test statistic between the 5 % and i % critical values) and outliers (test statistic greater
than the 1% critical value) in accordance to Mandel's h and k statistics, as explained in
ISO 5725-2. The specimens and test technique of participants rated as stragglers or
outliers were analyzed more carefully.
Alumina-998 - The within- and between-laboratory consistencies of participant 31
(Figure 2) were so poor that they were rated outliers. The analysis showed that two
toughness values were very low. On request, the participant analyzed the tested
specimens fractographically, but could not find any signs of suberitical crack growth
(scg) or other defects responsible for the low values.
Alumina-999 - Participant 26 reported only two very low fracture toughness values,
and the remaining three specimens failed during handling. The participant was asked to
KOBLER ON FRACTURETOUGHNESS USING THE SEVNB METHOD 101
analyze the fractured surfaces fractographically and both specimens were then shown to
have large precracks. Because of the precracks the values were discarded.
The within-laboratory consistencies of participants 8 and 22 and also the between-
laboratory consistencies of participants 15 and 31 were rated stragglers. None were rated
outliers. The reason for the stragglers is not yet clear. It might be significant that
participants 8 and 22 had problems resetting the razor blade properly into the notch and
therefore polished some small additional notches at the notch root. Furthermore,
participant 31 explained that two specimens had failed while polishing the V-notches.
This, together with the fact that this particular participant had already measured two very
low fracture toughness values for the alumina-998, could indicate that some or all of the
participant's alumina specimens had been precracked during preparation.
Compared with the alumina-998, GPSSN and SSiC, the coefficients of variation for
the repeatability and especially the reproducibility are rather high for the alumina-999.
This could be due to an environmental influence, e.g. subcritical crack growth. Therefore,
two additional sets of five bend bars each were tested under nitrogen atmospheres to
suppress scg and in water to promote scg. Interestingly, the measured fracture
toughnesses were ranked Kair < Kwater = Knitrogen (3.54 _+0.17 MPa ~/m < 3.97 _+0.20 MPa
~/m = 4.00 _+0.21 MPa ~/m). Therefore, if scg influences the measured fracture toughness,
liquid water must have been prevented from reaching the notch tips by polishing paste
residues or by some other mechanism which did not impede the passage of humid air.
G P S S N - The within-laboratory consistency of participant 8 and the between-
laboratory consistency of participant 25 were so poor that they were rated outliers. The
analysis showed that participant 25 had used a four-point bending jig with a span of only
30 / 10 mm instead of the required 40 / 20 mm and that the supporting rollers were set in
grooves and thus not free to roll. Constrained rollers can lead to an overestimate of the
bending strength by up to 5 % and could explain the high values measured. No
irregularities in the test technique of participant 8 could be found.
Further, the between-laboratory consistencies of participants 6 and 34 were in the
straggler range. The advancing route of the V-notch tip on the specimens of participant
34 was approx. 20 ~ which is about twice the permitted value set by JIS (Japanese
Industrial Standard) standard Testing Methods for Fracture Toughness of High
Performance Ceramics (R 1607-1990) for the crack tip if tested with the SEPB method.
This will lead to an overestimate of the bending strength and might explain the high
toughness values. The test technique of participant 6 did not show any irregularities.
SSiC - The within-laboratory consistency of participant 17 was rated an outlier and the
one of participant 30 a straggler, even though their standard deviations were only 0.26
MPa ~/m and 0.20 MPa ~/m, respectively. Further, the between-laboratory consistency of
participant 7 was in the range of a straggler. This is not too surprising since the absolute
values of the within- and also the between-laboratory consistencies were the smallest of
all five ceramics tested. It might be of interest that precracks (pop-in) were recognized on
the fractured surfaces of two rejected specimens from participant 17 and that participants
7 and 30 reported specimens which failed before testing.
Y - T Z P - A 12.7 % between-laboratory coefficient of variation was the largest for the
five ceramics tested. This is not surprising because of the fine grain size and notch width
dependence discussed earlier. No stragglers and outliers were found with respect to the
between-laboratory consistency, but interestingly one outlier was declared with respect to
the within-laboratory consistency. Participant 22, who already had difficulties testing the
alumina-999 and GPSSN, reported this outlier.
Comparison with Other Methods
All fracture toughness values measured with the SEVNB method, excluding those for
alumina-998, compared well with values determined by other methods like SCF, CN and
SEPB (Figure 4). Regarding the alumina-998 discrepancy, additional SEVNB tests
showed no dependency on environmental factors (dry N2, silicon oil), test speed (100X
slower test speed, dynamic fatigue tests with SEVNB specimens using test rates of 0.005
and 50 mm/min) or notch depth (increased notch depth of 2.1 mm).
The discrepancy might be explained with a study by Nishida, Pezzotti, Mangialardi,
and Paolini [16] performed with SEVNB specimens to detect the near-tip R-curve
behavior and to discern it from the long-crack R-curve behavior. In a coarse-grained
(18 }am) alumina with a microstructure similar to the one of the alumina-998, they
detected and quantified the increase in the measured fracture toughness as 1.5 MPa ~m
within the first 10-20 grains just beyond the notch tip.
-g 8.0 80
" ~ . . . . 9 other methods, iadiv, avg.
Aluminu-998 9 other methods, indiv, avg.J
--- SEVNB; G.P.Avg. ! "~ t /~,dumma-~ --- SEVNB; G.P.Avg,
~ 6.0
- - SEVNB; G.P.Std.Dev. 6 0 ~- - - SEVNB; G.P.Std.Dev.
9 ~ r ~ ~ ~ .__
4.0 ............... ~--- - " - ~ - _~..-_-_4f_2_-
:.E---_7-E-E---E-E~_E-:--- ---: ~--_-._:._7-E-E-_--_=-_-.~.E-:.2
2.0 ~-2.0+ ~ ~ ~ ~ ~ ~ ~ .,~,~

~ ~ ~ ~. ~ > >
0.0 0.0 . . . . . . . . . .
M e t h o d ( P a r t i c i p a n t / N u m b e r of Specimens) M e t h m i ( P a r t i c i p a n t / N u m b e r of S p e ~ a e a s )
8.0- 8.0 .~ r- ~ ~ ~
"~ 9 other methods, indiv, avg.
--- SEVNB; G.P.Avg.
- SEVNB; G.P.Std.Dev.
:------_------~---~---:_-~-----~-----7_---~----:_ 6.0 +
-
6.0 ~ ~ ~ ~ ~ ~
4.0- GPSSN
g g .p_z_~.z._.~.__:._.~_z_~__:.~:___ ~_:_-__-~_:._..:
2.0 + ~ ~. - - -
_ q other methods, indiv, avg.
b ~ ~ uz r~ 0.01
t ~SiC
--- . .
--- SEVNB; G.P.Avg
.- - SEVNB; G:P.Std.I~v.
0.~
M e t h o d ( P a r t i c i p a n t / N u m b e r of Specimen) M e t h o d ( P a r t i c i p a n t s / N u m b e r of Specimens)
Figure 4 - Comparison of fracture toughness values measured with other test methods
with values from the SEVNB method (alumina-999: SCF( I O/5), and SCF +halo( l O/5) are
calculated from the same test set. SSiC: "ESIS" [2]; "SEPB-JFCC", and "SEVNB-Lb"
[101; Non-participant X: CN geometry (A ) in accordance with ASTM C 1421).
Another possibile explanation for the alumina-998 discrepancy could be stable crack
growth or a pop-in during all SEVNB tests. An undetected crack growth of = 250 ]am
would increase the fracture toughness from 3.6 to 4.2 MPa ~/m and would reach the value
measured with the SCF method. One participant, an experienced fractographer, saw
evidence of a 50 to 150 lam crack "initiation" region at the root of the notches. A closer
look at digitally recorded load-time curves showed evidence of stable crack growth
(Figure 5). The load-time curves of the alumina-998 display a departure from linearity
before failure while GPSSN, for example, maintains linear behavior up to the point of
fracture.
110 100
AInmina-998
.9 9o 80
81) 70
4.5 5.0 5.5 6.0 6.5 3.5 4.0 4.5 5.0 5.5 6.0 6.5
Time, relative [s] Time, relative [s]
Figure 5 - Load-time curves (--)first deviation from linear behavior).
Conclusions
With the SEVNB method, very consistent fracture toughness results were obtained for
the coarse-grained alumina-998 and the sintered silicon carbide. Consistent and
reasonably consistent results were obtained for the gas pressure sintered silicon nitride
and the fine-grained alumina-999, respectively. As predicted, the results for the yttria-
stabilised tetragonal zirconia polycrystal were less consistent due to its grain size being in
the submicron range. Only the mean of the coarse-grained alumina-998 did not compare
well with results obtained using other test methods. A combination of a high sensitivity to
stable crack growth and pop-in of small cracks to form a crack "initiation" region might
be responsible for the observed discrepancy.
The repeatability and reproducibility of the method are equal or better than those of
other methods such as, for example, SEPB and SCF. "Good" fracture toughness values
can be measured for ceramics with an average grain size or major microstructural feature
size greater than about lmm. The V-notch depth had no influence on the measured
fracture toughness over a wide range. Participants unfamiliar with the SEVNB method in
general had no difficulty performing the measurements. The method proved to be
forgiving and robust with respect to notch preparation and notch quality. Participants
rated the method user-friendly, easy, reliable, and accurate and are interested in a SEVNB
standard.
Outlook
The VAMAS / ESIS round robin proved that the SEVNB method is forgiving and
robust with respect to the notch preparation and notch quality, user-friendly, easy to
conduct, reliable and accurate. Therefore, as a fast, first step ESIS / V A M A S prepared the
Procedure for Determining the Fracture Toughness of Ceramics using the Single-Edge-V-
Notched Beam (SEVNB) Method (ESIS P5-00 / VAMAS). The procedure was published
in Summer 2000 (Ed. Schwalbe, K.-H-, ESIS Procedures and Documents, Geesthacht,
Germany). At present, ENV Test Method for the Determination of Apparent Fracture
Toughness - Part 5: Single-Edge-V-Notched Beam Method (pre-standard ENV xxx-5) is
in preparation. The ENV should be ready for a formal vote at the end of 2001 or at the
beginning of 2002. An ENV (= pre-standard) has a maximum lifetime of three years,
after which it must be converted into an EN or terminated, thus the final EN will probably
not be ready before 2005. A final step might be the conversion of the EN or even the
ENV into an ISO standard.
An other aspect of the successful round robin had been that researchers are using the
method now too to measure the fracture toughness of materials they had difficulties like
dental ceramics, epoxy resins, and ceramic matrix composites. For example, Moon,
Bowman, Trumble, and R6del [17] report that the ability to control the V-notch tip
position within the material, combined with the short-crack extensions, enabled this R-
curve measuring technique to probe the influence of specific microscopic features on the
fracture toughness KR for both monolithic and multilayered alumina-zirconia composites.
Interestingly, they found that the SEVNB measured R-curve was consistent with the R-
curve measured using the SCF testing method.
Acknowledgments
The author wishes to thank all participants for joining in this successful round robin
exercise and Metoxit, Switzerland, for supplying the alumina-999. The participating
laboratories and corresponding members are listed below.
M.J. Anglada Univ.Politic. Catalunya, Spain H.A. Lindner CFI, R/Sdental,Germany

S. Barinov Acad. Sciences, Moscow, Russia F. Meschke Lehigh Univ., Bethlehem, USA
K. Breder ORNL, Oak Ridge, USA R. Morrell NPL, Teddington, UK
S.R. Choi NASA, Cleveland, USA Y. Nagano JFCC, Nagoya, Japan
Da Costa Neto Univ.Fed. Rio de Janeiro, Brazil J.Y. Pastor Univ. Politic. Madrid, Spain
R. Damani Montanuniv.Leoben, Austria G. Quinn NIST, Gaithersburg,USA
G. De Portu CNR-IRTEC,Faenza, Italy G. Rauchs Univ. Karlsruhe, Germany
P. Descamps B.C.R.C.,Mons, Belgium M. Reece Queen M&W Col., London, UK
J. Dusza Acad. Sciences, Kosice, Slovakia T. Reich FhG-IKTS, Dresden, Germany
G.A. Gogotsi Acad.Sciences, Kiev, Ukraine H. Richter CeramTech, Plochingen, Germany
M. Hoffman Univ.New South Wales, Australia J. R/Sdel TH Darmstadt, Germany
K. Keller BASF, Ludwigshafen, Germany E. Rudolph BAM, Berlin, Germany
V. Knoblauch Bosch,Stuttgart, Germany S. Sakaguchi NIRIN, Nagoya, Japan
M. Kuntz Univ. Bremen, Germany M. Steen EC-JRC Petten, Netherlands
M. Labanti ENEA-CRNM,Faenza, Italy R. Westerheide FhG-IW , Freiburg, Germany
Last but not least, the author wishes to thank his employer, the Swiss Federal
Laboratories for Materials Testing and Research (EMPA), for the opportunity to organise
the round robin together with the financial support (EMPA contract 165D13, 168'611 /
840'129), his co-worker R. B~ichtold for the many additional tests performed, and M.
Wegmann for reviewing the paper.
References:
[1] Quinn, G.D., Gettings, R.J., and Ktibler, J.J., "Fractography and the Surface Crack in
Flexure (SCF) Method for Evaluating Fracture Toughness of Ceramics,"
Fractography of Glasses and Ceramics III: Ceramic Transactions, Vol. 64, J.R.
Varner, V.C. Fr6chette, and G.D. Quinn, Eds., American Ceramic Society,
Westerville, Ohio, 1996, pp. 107-144.
[2] Primas, R.J. and Gstrein, R., "ESIS TC 6 Round Robin on Fracture Toughness,"
Fatigue and Fracture of Engineering Materials and Structures, Vol. 20, No. 4,
1997, pp. 513-532.
[3] Salem, J.A., Ghosn, L., Jenkins, M.G., and G.D. Quinn, "Stress Intensity Factor
Coefficients For Chevron-Notched Flexure Specimens and A Comparison
Fracture Toughness Methods," Ceramic Engineering and Science Proceedings,
Vol. 20, No. 3, 1999, pp. 503-512.
[4] Nishida, T., Hanaki, Y., and Pezzotti, G., "Effect of Notch-Root Radius on the
Fracture Toughness of a Fine-Grained Alumina," Journal of the American
[5] Quinn, G.D., "VAMAS after Twelve," The American Ceramic Society Bulletin, July
1999, pp. 78-83.
[6] Evans, A.G., "Perspective on the Development of High-Toughness Ceramics,"
Journal of the American Ceramic Society, Vol. 73, No. 6, 1987, pp. C133-C135.
[7] Damani, R., Gstrein, R., and Danzer, R., "Critical Notch-Root Radius Effect in
SENB-S Fracture Toughness Testing," Journal of the European Ceramic Society,
16, 1996, pp. 695-702.
[8] Le Bac, "Verfahren zum Feinkerben yon keramischen Ktirpern," Patentschrift 146416,
Deutsche Demokratische Republik - Amt fur Erfindungs- und Patentwesen, 1979-
1981.
[9] Awaji, H. and Sakaida, Y., "V-Notch Technique for Single-Edge Notched Beam and
Chevron Notch Methods," Journal of the American Ceramic Society, Vol. 76, No.
11, 1990, pp. 3522-23.
[10] KUbler, J., "Fracture Toughness Using the SEVNB Method: Preliminary Results,"
Ceramic Engineering and Science Proceedings, Vol. 18, Issue 4, 1997, pp. 155-
162.
[11] Munz, D , and Fett, %, "Ceramics; Mechanical Properties, Failure Behaviour,
Materials Selection," Materials, Vol. 36, Springer-Verlag, Berlin, Germany,
1999.
[12] Munz, D., Bubsey, R.T., and Shannon, J.L.,jr., "Fracture Toughness Determination
of A1203Using Four-Point-Bend Specimens with Straight Through and Chevron
Notches," Journal of the American Ceramic Society, Vol.63, 1980, pp. 300-305.
[13] Fett, T., "Notch Effects in Determination of Fracture Toughness and Compliance,"
International Journal of Fracture, 72, 1995, pp. R27-R30.
[14] Fett, T., "Estimated Stress Intensity Factors for Semi-Elliptical Cracks in Front of
Narrow Circular Notches," Engineering Fracture Mechanics, 64, 1999, pp. 357-
362.
[15] Schindler, H.J., "Size Effects on Fracture toughness of Brittle Materials,"
International Journal of Fracture, 82, 1996, pp. R3-R10.
[16] Nishida, T., Pezzotti, G., Mangialardi, T., and Paolini, A.E., "Fracture Mechanics
Evaluation of Ceramics by Stable Crack Propagation in Bend Bar Specimens,"
Fracture Mechanics of Ceramics, Vol. 11, R.C. Bradt et al., Eds.,, Plenum Press,
New York, 1996, pp. 107-114.
[17] Moon, R., Bowman, K., Trumble, K., and RSdel, J., "Comparison of R-Curves from
Single-Edge-V-Notched-Beam (SEVNB) and Surface-Crack-in-Flexure (SCF)
Fracture-Toughness Test Methods on Multilayered Alumina-Zirconia
Composites," Journal of the American Ceramic Society, Vol. 83, No. 2, 2000,
pp. 445-447.
George D. Quiun 1
The Fracture Toughness Round Robins in VAMAS: What We Have Learned
Reference: Quinn, G. D., "The Fracture Toughness Round Robins in VAMAS:

What We Have Learned," Fracture Resistance Testing of Monolithic and Composite
Brittle Materials, ASTMSTP 1409, L A. Salem, G. D. Quinn, and M. G. Jenkins, Eds.,
American Society for Testing and Materials, West Conshohocken, PA, 2002.
Abstract: Over the last 13 years the Versailles project on Advanced Materials and
Standards (VAMAS) has conducted five full fledged, international round robins on
ceramic fracture toughness characterization. As many as forty laboratories have done
thousands of experiments on silicon nitride, silicon carbide, zirconia, alumina, zirconia-
alumina composite, and whisker-reinforced alumina. Test methods have included single-
edged precracked beam, single-edged V-notched beam, chevron-notched beam, surface
crack in flexure, indentation strength, and indentation crack length. Highlights of these
projects are presented along with some lessons learned about round robins in general.
Keywords: fracture toughness, ceramics, surface crack in flexure, chevron notch,

precracked beam, single-edge precracked beam, reference material, round robin, VAMAS
Introduction and Background: VAMAS
The Versailles project on Advanced Materials and Standards (VAMAS) is a forum for
prestandardization research that has spread awareness of new testing and characterization
methods around the world. VAMAS was launched in 1982 following an economic
summit meeting held in Versailles by the group of seven (G-7) Heads of State and
representatives from the European Communities. Several collaborations including
Technical Working Area (TWA) 3, Ceramics, were already under way by the time a
Memorandum of Understanding was signed in April 1987. TWA 3 has contributed to
many national and international standards. TWA 3 in 2000 had almost one hundred
volunteer members from government, academia, and industry. A history of VAMAS
TWA 3 and the first twelve round robins conducted since 1986 is presented in Ref. 1.
Figure 1 schematically summarizes the current total of 14 projects. Over 12,000
experiments have been done in these projects.
Round robins are an important tool in the development of standards, but it is important
that they be carefully planned. TWA 3 has project guidelines shown in Table 1. These
1 Staff Scientist, Ceramics Division, Stop 8521, NIST, Gaithersburg, MD, 20899.
107
Copyright9 ASTMlntcrnational www.astm.org
Figure 1 - - The fourteen VAMAS TWA 3 round robins.
facilitate successful completion o f a project and minimize the chances a project will drag
on for years or not even finish at all. Projects usually are planned to take one year to run
from first distribution o f test specimens to final report publication, but in practice a more
realistic timetable is for one and a half to three years depending upon the complexity o f
the program. Participants are encouraged to recommend improvements, alternatives to a
test procedure, and alternative interpretations o f the data. Participants furnish "reality
checks" on a method and they often find errors or loopholes in the project instructions or
they are unable to do the work strictly according to the instructions. Sometimes the
procedure is impractical outside a research laboratory. Weaknesses such as these need be
identified prior to formal standardization.
Table 1 - - G u i d e l i n e s for TWA 3 Projects
1. Projects must have focused objectives.

2. Round-robins shall be limited to 1-I V2man weeks of effort per participant.
3. Reports and data release must be made in a timely fashion.
4. All participants shall have the opportunity to review and correct a draft version of
a final report.
5. Extra material should be furnished to allow participants to try alternative
procedures.
QUlNN ON ROUND ROBIN IN VAMAS 109
This article is an overview of the five fracture toughness projects and over 4,500
experiments and the lessons learned about conducting international round robins in
general. As many as 40 laboratories have participated in some of these round robins,
which is remarkable since VAMAS does not fund projects and participation is completely
voluntary. The results of this work have directly contributed to the new ASTM Standard
Test Method for the Determination of Fracture Toughness of Advanced Ceramics (C 1421-
99) as well as several Japanese Industrial Standards (JIS), a draft European Community
(CEN), and several draft International Organization for Standards (ISO) standards. A
NIST certified reference material is now available as a direct consequence of one of the
VAMAS round robins.
Fracture Toughness Round Robins
A myriad of test procedures have been used to measure this critical property and, to be
candid, the literature is infested with inconsistent and erroneous test results. Standardized,
refined, and validated procedures were desperately needed. VAMAS aggressively
addressed this problem. This brief summary article cannot cover the five fracture toughness
projects in detail, but presents a few highlights to give a sense of each project and its key
findings. The publications listed in the Bibliography (Appendix I) may be consulted for full
details on any particular project.
SEPB, IS, and IF Methods
Fracture toughness project I, organized by the Japan Fine Ceramic Center (JFCC) in
1989, featured three methods: single-edged precracked beam (SEPB), indentation strength
(IS), and indentation fracture (IF) as illustrated in Figs. 1 and 2. The SEPB method was
highly favored in Japan and was on the verge of being standardized by JIS as Testing
Methods for Fracture Toughness of High Performance Ceramics (R 1607). A variant on the
older single-edged notched beam (SENB) method, the SEPB method uses an innovative
precmcking procedure (compression bridge-anvil loading) to produce sharp precracks [2].
A saw cut or indentation was placed in the narrow side of a flexure specimen. The
specimen then was placed into a compression-anvil precracking apparatus and loaded
carefully until a precrack popped in. Next, the specimen was fractured in a common flexure
fixture. Participants tested duplicate sets of specimens at different crosshead displacement
rates. Crack size was measured on the fracture surface.
SEPB IS
+
IF
Figure 2 - - The three test method in the fracture toughness I round robin.
1 10 FRACTURETESTING OF MONOLITHIC/COMPOSITE MATERIALS
For comparison purposes, the round robin also included two indentation methods. The
indentation crack length method [3,4], also known as the indentation fracture (IF) method,
is also in JIS R 1607 but as a deprecated fracture toughness estimator ("I~", not "K~c").
This method, no longer held in high esteem, involves placing a Vickers indentation on a
polished surface and then measuring the lengths of the comer cracks. Two indentation
loads were prescribed. The IS method, also known as the indentation strength in bending
(ISB) procedure [5], uses a Vickers indenter to precrack a bend specimen which
subsequently is fractured. Fracture toughness is calculated from the beam fracture force, the
indentation load, the hardness and elastic modulus of the material, and a calibration
constant. Two indentation loads were specified in the round robin. How well did these
simpler indentation methods fare compared to the more rigorous SEPB method?
The round robin ran over a period of three years from 1989 through 1992. After one and
a half years, the organizers had received results from 15 of the 23 laboratories that signed
up. An initial final report was prepared by JFCC. Within a year, five additional
laboratories from the United States finished their work and a follow-on final report was
prepared by NIST.
More than 1000 silicon nitride and zirconia-alumina composite (ZAC) specimens were
used for more than 1500 experiments. Results were mixed. SEPB and IS results were
reasonably consistent, although the IS method results depended on indentation load. IF
results were very inconsistent, varying by as much as a factor of 2 between laboratories!
The silicon nitfide used in the project evidently had a flat R-curve and was not
susceptible to slow crack growth at room temperature. On the other hand, slow crack
growth and R-curve phenomena were active in the oxide composite and results were much
more variable.
The SEPB method was more difficult to do than the other two methods, but participants
had greater confidence in the results. A special bridge precracker had to be used.
Participants who made their own apparatus experienced difficulties whereas the standard
Japanese design worked very well. The project underscored the need for a carefully aligned
and well-designed bridge precracker. Compression precraeking required a large testing
machine and large forces to pop in the precrack. There was considerable risk of overload to
the machine and its load cell if operated in displacement control. Precrack inspection after
fracture was essential in order to verify that the crack was reasonably straight and met the
validity requirements for this method. The correct crack length is an important parameter in
the fracture toughness calculation. Dye penetration and fraetographic methods may be
applied. The project confirmed that the dye penetration from a "green" felt tip pen worked
well for the white oxide ceramic. The dye was less effective with the black silicon nitride,
but microscopic inspection with low-angle incident lighting worked well. Three-point
ftexural testing, while allowed by the test procedure and included in JIS R 1607, was more
problematic than four-point flexure testing since exact alignment of the very tight precrack
with the middle loading roller in three-point flexure is very difficult. Unlike the silicon
nitride, the oxide composite had a pronounced rate sensitivity that varied dramatically from
laboratory to laboratory. We now realize that much of this variability was due to stable
crack extension during the test but the original instructions overlooked this key point.
The IS method was much easier to do and therefore was popular with some participants,
but fracture toughness showed a disappointing dependence upon indentation load,
QUINN ON ROUND ROBIN IN VAMAS 111
especially for the oxide composite. Many participants were uneasy as to whether valid
precracks were obtained and whether slow crack growth affected the outcomes. The 490 N
(50 kgf) precrack load in the oxide probably created large precracks relative to the
specimen's cross section size, a factor not considered in the derivation of the equations for
this method. This is a weak point of the method and contributed to experimental error.
The IF method was the simplest of the three methods with the key exception that
measuring the true crack length posed formidable problems. Fracture toughness depends
upon the crack length raised to the 1.5 power, so any uncertainty in measuring the crack
length was magnified in the toughness calculations. Between-laboratory consistency was
very poor. For the silicon nitride, average results varied from 5 MPa~/m to 7 MPa~/m.
Results were even more variable for the oxide composite: from 5 MPa~/m to over 9
MPa~m with dramatically different trends of apparent toughness versus indentation load
as shown in Figure 3. The IF method is not suitable for standardization.
11
IF method
10 Material: ZAC
E
(1. . 5 o
usa
G)
r
t-
8
O)
o
"-t 7 ! .,, USA 2
LI. ~ 9" " ~ 1..13
.
;2A24~ ;_._-_-_
"........... J.-;u~sA 4
USA 2
. c/a < 2.3 I c/a >
2.3
Indentation Load, N
Figure 3 - - The IF results for the zirconia-alumina composite exhibited excessive scatter
underscoring the unreliability of this method Apparent fracture toughness is shown versus
indentation load Each data point corresponds to the average toughness from a particular
lab and the numbers are the laboratory codes. Each lab used the two prescribed loads.
The crack length to indentation half diagonal length ratio is c/a.
The round robin identified topics for further study. R-curve and/or slow crack growth
behavior affected the zirconia-alumina composite SEPB data. Such stable crack extension,
which is often overlooked, should be factored into the analysis of fracture toughness by this
method. Some participants wondered whether the SEPB results were affected by the initial
Vickers indentations used to start the precrack. Overall, the participants felt that this round
robin required a lot of work since it featured three methods, two materials, and different
parameters for each test. Nonetheless, the participants felt the quality of the results and the
lessons learned warranted the effort.
This round robin had a significant influence on standardization programs in the United
States (ASTM Committee C-28 on Advanced Ceramics) and in Europe (Committee CEN
TC 184) and ISO, the International Standards Organization (Technical Committee TC 206).
The SEPB method became widely accepted in the fracture mechanics and ceramics
community in large part since it is a traditional fracture mechanics configuration that now
has a good record for producing accurate and precise results. The IF and IS methods were
not considered further within Committee C-28 after this VAMAS round robin.
High Temperature Fracture Toughness: SEPB, SENB, CNB Methods
Fracture toughness round robin II, organized by the JFCC in Japan in 1990, was a high-
temperature project but included some room temperature testing for comparison. Three
methods were tried: single-edge precracked beam (SEPB), chevron notch beam (CNB), and
single-edge V-notched beam (SEVNB) as illustrated in Figure 4. Only a single material
was used in this round robin. A sintered silicon nitfide with an elongated grain structure
was tested at both room temperature and 1200~ in air and inert atmospheres. Would the
laboratory atmosphere have an effect upon toughness? Would the cracks heal? Would
there be test rate effects? How would the test results from the different methods compare?
The SEVNB specimen is a variant of the older single-edged notched beam (SENB)
method (wherein a saw blade creates a straight-through notch) except that a tapered
diamond wheel is used to create a "sharp" notch.
Fewer laboratories enrolled in this project due to the greater difficulty of elevated-
temperature testing, equipment limitations, or a desire to finish first VAMAS fracture
toughness round robin described above. Of the 12 laboratories that enrolled in 1990,
eight finished their work by 1993. Several laboratories had great difficulty obtaining
diamond wheels with the requisite shaped V-tip for the SEVNB method.
0 0 , 0 :i 41 r 1
I o ' o ' o 3
SEPB, SEVNB, CNB
Figure 4 - - Three methods at room and elevated temperature were used in the fracture
toughness II project. Magnified specimen cross sections are shown on the right.
Dimensions are in mm.
QUINN ON ROUND ROBIN IN VAMAS 1 13
Fracture toughness was not sensitive to crosshead rate, but oxidation of the silicon
nitride at 1200~ in air increased the apparent fracture toughness when measured by the
SEVNB and SEPB methods. This was probably due to healing of machining damage at the
V-root in the SEVNB method or main crack healing in the SEPB method. Hence, these two
methods are not suitable at elevated temperature in air if the material oxidizes. In contrast,
SEPB and SEVNB fracture toughness values at 1200~ in nitrogen were similar to values at
room temperature. This suggests that inert atmosphere testing is safer for evaluating
fracture toughness at elevated temperature for materials that oxidize.
Most laboratories obtained similar CNB results at room temperature and 1200~ for both
inert and air atmospheres. This indicates there was little or no variation of fracture
toughness with temperature or environment in this particular material. Hence, CNB testing
may be done at elevated temperatures in air or inert atmosphere, even for materials that may
oxidize. This is a major advantage of the CNB test over the SEVNB or SEPB methods. On
the other hand, the CNB results at 1200~ varied considerably between participants,
probably due to differences in the machining procedures for the notches. The measured
CNB results were somewhat greater than the SEPB or SEVNB results. This is not
surprising for this material that probably has a rising R-curve since it had an elongated
silicon nitride grain microstructure.
Proper CNB specimen preparation requires careful machining by an experienced
laboratory or machine shop. Notch variability probably accounted for the large between-
laboratory variability in this project. Poorly prepared notches are likely to cause unstable
fracture. Crack propagation must be stable for the result to be valid. Casual notch preparation
by a student with a cut-offwbeel in a laboratory will not suffice. The organizing laboratory
tried a variation on the CNB test which had the chevron notch made by two sharp V-
grooves. They obtained low scatter in their results, irrespective of crosshead speed,
temperature, or atmosphere. This "CVNB" method was the only one that measured fracture
toughness without influence of machining conditions.
After this project, Japan adopted Testing Method for Fracture Toughness of Fine
Ceramics at Elevated Temperature (JIS R 1617-1994) based on the SEPB method.
SCF Method
Fracture toughness project III was organized in 1992 by NIST, USA and the Swiss
Federal Laboratories for Materials Testing and Research (EMPA), Switzerland and featured
the surface crack in flexure (SCF) method, Figure 5. The SCF method (previously known
as the controlled surface flaw method ~) was one of the three eventually incorporated in
ASTM C 1421, ISO Fine Ceramics (Advanced Ceramics, Advanced Technical Ceramics),
Determination of Fracture Toughness of Monolithic Ceramics by the Surface Crack in
Flexure (SCF) Method (ISO 18766), and the new draft CEN standard, Advanced Technical
Ceramics-Monolithic Ceramics-Test Methods for Determination of Apparent Fracture
2 The name of this method was changed in order to avoid confusion with other so-called
controlled flaw methods and to make the name consistent with fracture mechanics conventions
such as used in Practice for Fracture Testing with Surface-CrackTension Specimens (E 740).
Figure 5 The SCF method was featured in the fracture toughness III project. The photo
shows a typical precrack in the hot-pressed silicon nitride. The middle insert shows the
specimen cross section with dimensions in mm. The semielliptical precrack is exaggerated
in size relative to the cross section size.
8.0
7.0
E
6.0
ft.. 7 7 8 3 5 5 8 $ 3 43 10
5 5 5 3 4 5 66 "[" -r"
5.0
I -I- I -i
4.0
r-
.-I
o 3.0
m
"6 2.o * SEM
m
LL o Optical
1.0
0.0
0 i 2 3 4 5 6 7 ~ 9 1~01=1 121=3141=516 11718 192021 ~_22324
Laboratory
Figure 6 - - SCF results for hot-pressed silicon nitride, The laboratory average, standard
deviation, and number of specimens are shown for each laboratory. The dashed line is the
grand average o f 98 outcomes. There was no significant difference in outcomes whether
optical or electron microscopy was used to measure the precrack sizes
Toughness. A Knoop indenter was used to create a tiny semi-elliptical surface flaw in a
bend specimen [6]. After indentation, a prescribed amount of material was hand ground or
polished away in order to eliminate the residual stress--damage zone associated with the
indentation. The goal was to create a simple semi-elliptical precrack, unencumbered by
surface damage or residual stresses. Fractography was used to measure the precrack after
fracture. Could participants make, find, and reliably measure such a small crack? Would
fractography be too difficult or too subjective?
Two silicon nitrides and one zirconia were evaluated and 20 of 24 laboratories finished
their work in a little more than 1 year. There was excellent data consistency with the hot-
pressed silicon nitride, Figure 6, and all laboratories had some degree of success. The hot-
isopressed silicon nitride and sintered zirconia were more difficult. Preparation and fracture
testing of the specimens were easy, but the steps of hand polishing then detecting and
characterizing the precracks were time consuming and problematic for some laboratories.
SCF precrack detection is difficult in some materials and dye penetration methods are under
development. Nonetheless, the VAMAS round robin showed that some members who had
never tried the method before and who had limited fractographic experience obtained good
results.
Indeed, one of the astonishing findings of the VAMAS project was that the outcomes are
not sensitive to the crack size measurement. It didn't matter whether an optical microscope
or a scanning electron microscope were used. Before the round robin, it had not been
appreciated that there is a fortuitous interaction between the crack size measurement and the
stress intensity shape factor coefficient, Y, in the formula for fracture toughness. An error
in one is mitigated by a compensating effect in the other. In other words, an overestimate in
crack size, a, is matched by an underestimate in the computed Y. For some crack
configurations, a 10% error in the crack size led to as little as a 1% error in fracture
toughness!
SCF results are representative of the critical fracture toughness that natural flaws
experience when loaded to fracture, in contrast to results from the large-crack methods. For
brittle ceramics that have negligible R-curve behavior, the SCF method gives identical
results as the SEPB and CNB methods. The unexpectedly good results for the hot-pressed
silicon nitride in this VAMAS round robin and corroborative data by other sound methods
led NIST to prepare Standard Reference Material SRM 2100, the world's first certified
reference material for fracture toughness as is described elsewhere in this volume [7].
Fracture Toughness of Composites: SEPB, SEVNB, SENB Methods
Fracture toughness project IV was organized by JFCC in 1994 and used three test
methods: single-edged precracked beam (SEPB), single-edged "V-Notched" beam
(SEVNB), and single-edged notched beam (SENB), on a single material (Fig. 7). The
composite ceramic was a SiC whisker-reinforced SiaN4. Each participant was required to
test 10 specimens by SEPB, but then could test 20 additional specimens by SEVNB or
SENB. The SEVNB specimens were notched with a tapered diamond wheel. The SENB
specimens were straight-through notched with a simple cut-off-wheel. Either three- or
four-point loading was permitted. Fifteen of sixteen laboratories had completed their
0 I I o I o~ I 4 lit
I cO ' 0 s 3
SEPB, 8EVNB, 8ENB
Figure 7 - - Three types o f specimen were used in the fracture toughness IV project on a
ceramic whisker-reinforced composite. An oversized specimen cross section with
dimensions in mm is on the right.
work by late 1997.

SEPB results were accurate and reliable for laboratories that tested in three-point flexure
and had experience with the method, although there were some notable exceptions. SEPB
specimens tested on very short 16 mm span three-point fixtures overestimated fracture
toughness compared with data from longer 30 mm or 40 mm spans. It was difficult to
precisely align a sharp SEPB precrack with the middle roller in short, three-point flexure
spans. There was high unaccounted variability in SEPB four-point test results. This was
the third VAMAS fracture toughness round robin with the SEPB method, yet some
participants outside Japan still had difficulty obtaining valid precracks.
SENB results inevitably overestimated the fracture toughness, even for notch widths
down to 60 txm. Apparent fracture toughness correlated strongly with notch width. This
underscores the weak point of the SENB method: an ordinary saw blade notch does not
simulate a true sharp crack. The SENB method is not under consideration for
standardization.
SEVNB results converged towards but were slightly less than the SEPB results if notch
tip radii were smaller than 30 txm. The SEVNB notches were easy to align in three-point
fixtures, unlike the tight and hard to detect SEPB precracks.
In summary, the SEPB and SEVNB methods were deemed suitable for whisker-
reinforced ceramic composites. The SEVNB precracks (which have all material behind the
precrack cut out) may generate a fracture toughness from the small crack length portion of a
R-curve, whereas the SEPB method (which may have whisker or grain bridges behind the
crack tip) may produce a greater fracture toughness from a point further along a rising R-
curve.
SEVNB Method
Fracture toughness project V was organized by the EMPA, Switzerland, and was based
on a new variant of the single-edged V-notched beam method (SEVNB). The project was a
joint VAMAS / European Structural Integrity Society (ESIS) collaboration. The innovation
in this project was that the V-notch was created with a simple reciprocating razor blade and
diamond paste as shown in Figure 8. This method was not on any schedule or plan for
standardization, but its potential simplicity had widespread appeal. The project organizer
was eager to demonstrate its value.
A common diamond wheel was used to make a thin slot in the specimens. The notch was
then sharpened with the reciprocating razor blade and diamond paste, either by hand (a
tedious process) or by machine. The notch tip radius may be made sharper than 10 ~tm with
Figure 8 - - The single-edged V-notched beam method used in project V. The right side
photo shows a gang of five specimens mounted beneath a razor blade during the notch
preparation step.
suitable care.
Many laboratories prefer this scheme to the SEPB method, since they feel that bridge
precracking is too difficult or they feel that the SEPB precrack may not necessarily
produce a short crack fracture toughness in rising R-curve materials. Thirty-six
laboratories enrolled in this very popular project. All participants received ten specimens
o f a coarse-grained alumina and a silicon nitride. Five specimens of each material were
slated for the SEVNB method. The remaining five specimens were available for any
other test method or variation of the participant's choice.
Each laboratory also had the opportunity to test five specimens of any one of three
additional materials: a fine-gained alumina, a sintered silicon carbide, or a yttria-stabilized
zirconia.
Thirty-one laboratories sent back results. Overall, there was good agreement of results
provided that the notches were narrower than a critical size. Every participant had to send
sample photos of their tips to the organizer. This round robin confirmed that the critical
notch width is related to the microstructure. A few problem data sets or individual
outcomes were traced to specific problems such as too wide notches, excessive cracking at
the notch root during notch preparation, or notch depths that were out of specifications.
Figure 9 shows the best results of the round robin which were obtained with the coarse-
grained alumina. This data set had the best between-laboratory reproducibility and results
concurred with other test method data. All but three of the laboratories (3, 9, 29) obtained
notch widths narrower than the 30 lam that was required for this material. Laboratory 31
had greater scatter due to two atypically low outcomes. Their notch widths were within
specifications.
The yttria-stabilized zirconia outcomes had somewhat greater scatter, and gave lower
fracture toughness values than those obtained earlier in the SCF round robin which used the
same material. The SEVNB values were also less than SEPB and CNB results furnished by
four other participants. These low SEVNB outcomes are probably due to stable crack
extension. The organizing laboratory double checked their digitally recorded force-time
curves and did find some nonlinearity near fracture, confirming that stable crack extension
8.0
9 individual Lab avg.
E 7.0
- - -- grand, population avg.
n ..... grand, population dev.
6.0
~" 5.0
| 9
~o) 4.0
o
~e
.-I
3.0
-- 2.0
1.0
0.0 = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =
Pa~oi~nt #
Figure 9 - - SEVNB results for the coarse- grained alumina. The average and standard
deviation for each laboratory set (typically five specimens) are shown. The grand
population mean and standard deviation of a l l 135 specimens are shown as dashed lines.
caused the low SEVNB outcomes.

Excellent convergence of results from the different methods (with the exception of IF
and IS data which were deviant) was noted for the fine-grained alumina, the silicon nitride,
and the silicon carbide. This variant of the SEVNB method is a promising candidate for
standardization and the European Structural Integrity Society (ESIS)/VAMAS has already
prepared a recommended practice: Procedure for Determining the Fracture Toughness of
Ceramics Using the Single-Edge-V-Notched Beam (SEVNB) Method, (ESIS P5-
O0/VAMAS).
General Conclusions on the Methods
The VAMAS round robins showed that between-laboratory reproducibilities of 4 % to

10 % (coefficient of variation) are achievable with the SEPB, SEVNB, SCF, or CNB
methods. As is usual with any method, a little experience enhances the chances of success
with a method. Many of the laboratories were attempting a particular method for the first
time. The uncertainties would be even less if the laboratories were to repeat the
experiments.
Besides furnishing general uncertainty assessments, the round robin furnished valuable
information about weak points or nuances of the methods. The natural instinct for an
organizing laboratory was to hope that the participants would return consistent results and
thereby vindicate the particular method. On the other hand, aberrant or atypical results
inevitably prompted further review. Some of the best information and analysis came when
conscientious project organizers studied the outliers to find the source of the discrepancy.
Discrepancies could be caused by misunderstandings, equipment or experience limitations,
poor or misconstrued instructions, or unexpected material phenomena. Conscientious
project organizers began to appreciate the divergent laboratory results as much as the
mainstream data. Valuable input like this helped refine the procedures prior to
standardization.
A few general remarks about the individual test methods are in order. First we note the
obvious. Every specimen tested in the five round robins was a bend bar, underscoring the
strong preference of the ceramic community for this versatile specimen type.3
After spending years to develop the SEPB method for domestic standardization
purposes, the Japanese vigorously advanced the method internationally via three VAMAS
round robin projects. They were successful since the method is now included in ASTM C
1421, ISO 15732, and will be included in a European CEN standard. Nevertheless, some
weaknesses of the test method were identified by the round robins. Most laboratories are
now comfortable with the method provided that they have a good bridge precmcker. Stable
crack extension must be monitored and accounted for in the analysis, a point overlooked in
the original JIS R 1607.
The CNB method was used in only one round robin. Positive results from the high
temperature round robin showed that this method is eminently suitable for high temperature
testing and avoids many of the problems in other methods. Good notch preparation remains
a critical issue and is the root of most of the between-laboratory reproducibility problems.
Quality results require a quality chevron notch. 4 Although the CNB method was not
featured in the other four round robins, participants conducted parallel CNB experiments
with their extra specimens. This attests to the popularity of the method and a new chevron
notch VAMAS round robin would be welcome.
The SCF method yielded surprisingly consistent results particularly considering that
fractographic analysis was required. The project was as much a fraetography round robin
as it was a fracture toughness exercise. Most participants concluded that the method
required more work than they would like to do but that the fracture toughness outcomes
were technically rigorous. The SCF method remains the only method capable of reliably
ascertaining the fracture toughness for cracks of the order of size of real material flaws. It
has the best potential for evaluating apparent R-curve behavior of small flaws. The draft
ISO standard 18766 is venturing into this realm, ahead of any national standard in the
world. The most pressing need for this method is a simple dye penetration scheme to make
precrack detection easier.
The IF or indentation crack length method has fallen into disrepute, in part due to the
wildly variable results in the first VAMAS round robin. The metrological problems
associated with the measurement of the cracks lengths, the uncertainty associated with
3 Even the indentation crack length experiments in the first round robin were performed on
bend bar fragments.
4 An earlier European Structural integrity Society round robin also learned this lesson the hard
way.
which fracture toughness equation to use, and the uncertainty with the "calibration
constants" underscore the shortcomings of this method.
The IS method produced somewhat more consistent results and has appeal due to the
simplicity of the procedure, yet some of the same underlying doubts apply here as well.
The crack system is poorly defined and is very complicated. It is a great misnomer to call
this a "controlled-flaw" method. Although test results often have low scatter, the mean test
results often differ from those obtained from other methods. Outcomes often vary with
indentation load and, while this is commonly attributed to R-curve behavior, in reality it is
often a purely geometric or stress gradient effect. Measured toughness values are probably
affected by slow crack growth as well as the requisite stable crack extension from the
residual stress fields. Cracks shapes may evolve dramatically and may even grow to
become a major fraction of the cross section size of the specimen. The "calibration
constant" which has a large uncertainty also led to unease. This method is not a candidate
for ASTM standardization.
The SEVNB method was featured in two primary variants. Two Japanese coordinated
projects used custom-made diamond wheels to produce the sharp V-notch. Some
participants had difficulty obtaining such wheels. There were also some doubts about
wheel wear causing the notches to vary from specimen to specimen. In contrast, the V-
notching by razor blade and diamond paste procedure used in the Swiss led round robin
met with widespread acceptance. This method is on a fast track for European
standardization. The VAMAS round robins did identify some limitations, however.
Sometimes the V-notch simulates a sharp crack and sometimes it does not. Oxidation at
elevated temperature can affect the notch root. Attention to the crack initiation region at
the root o f the notch is warranted, even at room temperature. Stable crack extension can
occur and erroneous results will be obtained if it is not included in the calculation.
Benefits of Round Robins
Round robins can be very valuable, but should not be undertaken lightly since they are a
lot of work, they may raise more questions than they answer, and they may even backfire! 5
In some cases, "round robin fatigue" may set in whereby a project is organized, run, but a
final report was never written since the organizers cannot summon the energy to finish the
project or they have moved onto other topics. After 13 years, we may step back and reflect on
some of the benefits of round robins.
Several non-VAMAS projects have failed for one reason or another. Valuable lessons
sometimes may be learned by studying the causes of unsuccessful projects. In the mid 1980s
one ASTM Sub committee E24.07 round robin on fracture toughness and another on
fractography were not finished nor any results ever published, even though many participants
completed their work. A mid 1990s European fracture toughness round robin based on dye
penetration of SCF specimens collapsed when the dye mysteriously failed to work in the
participating laboratories, despite having worked reliably in the organizing laboratory! A
1994-1995 German project to identify candidate materials for ceramic hardness reference
blocks ended with less than optimum results that were never published. The specimens had
variable microstructures and surface polish qualities. It could be argued that the latter two
projects were in a sense successful, since they identified problem areas.
QUlNN ON ROUND ROBIN IN VAMAS 121
1. Round robins determine whether the specified procedure was sound and whether
anything was overlooked
Round robins may determine whether a procedure can be done at all by non-expert
laboratories. Round robins address equipment limitations, instruction problems, or
practicality issues. What are the special tricks or secrets? What are the weak points? What
are the alternatives? Round robins determine the robustness and ruggedness of a procedure,
since the participants inevitably try small variations or make slight departures (intentional or
unintentional) from the specified procedures. For example, several weaknesses in the SEPB
method were highlighted. SEPB is widely used in Japan, but European laboratories had
much greater difficulty due to precracking problems. The problem was traced to the
precracking fixture. A high quality precracker with good alignment must be used. One U.
S. A. laboratory detected a weakness in the first SEPB fracture toughness project when they
did not follow the directions. They measured the critical crack length and not simply the
initial precrack size. Analysis revealed that this laboratory obtained correct results. Stable
crack extension must be monitored and accounted for [8].
2. Round robins identify needs for future prestandardization research.

For example, the round robins demonstrated that it is difficult to align the crack in
miniature three-point SEPB in which case fracture toughness is overestimated. This
prompted one team to conduct new analytical work furnished guidance to refine the
tolerances for precrack and fixture alignment [9].
3. Round robins furnish data for uncertainty statements in standards. Are the results
accurate and precise?
ASTM precision and bias statements enable users to understand what they should expect
if they use the standard. Provided that the round robin procedure matches the standard
procedure, the within-laboratory and between-laboratory precision estimates can easily be
computed using ASTM Practice for Conducting an Interlaboratory Study to Determine the
Precision of a Test Result (E 691) or the analogous ISO Accuracy (Trueness and Precision)
of Measurement Methods and Results (ISO 5725).
The strong precision and bias statement in C 1421 uses precision estimates from the
VAMAS SCF and SEPB round robins.
4. Round robins identify needs for reference materials.

The intensive work both within and outside of VAMAS on fracture toughness prompted
NIST to prepare SRM 2100, the world's first fracture toughness reference material.
5. Round robins spread awareness of test methods favored in one country or group of
countries.
A round robin may also expose or "sell" the ceramic community on a new procedure.
Sometimes a round robin is set forth as a challenge: "Method A is better than method B and
the round robin will prove it!" Round robins may meet domestic or international political
needs. Potential barriers to trade may be reduced or eliminated by early prestandardization
cooperation prior to national or international standardization.
6. Round robins induce laboratories to try new methods.

A VAMAS round robin was often the first time a laboratory tried a particular new
method. The most recent round robin on the fracture toughness by the SEVNB method has
encouraged many laboratories to try razor blade sharpening of notched beam specimens, a
method devised as recently as 1994. The SEVNB method had been perceived as yet
another variant of the rudimentary and error prone SENB method. The positive results
with the razor sharpening refinement paved the way for fast track standardization of this
method.
The SEPB method, although popular in Japan due to the adoption of JIS R 1607 in
1990, was not widely used elsewhere due to perceptions about the difficulty of the method.
These perceptions were dispelled in many quarters after the VAMAS round robins.
The SCF round robin was the first time many laboratories had tried that method.
7. Round robins expand scientific and technical knowledge.

The surprising effects of stable crack growth on SEPB fracture toughness results were
under appreciated until the first VAMAS fracture toughness round robin was finished.
Oxidation related crack healing on very tight SEPB precracks was identified as a major
limitation of the SEPB method in the elevated temperature round robin.
The SCF fracture toughness round robin revealed how the interaction of the stress
intensity shape factor coefficient with the crack size measurement mitigates error
associated with the crack size measurement. This remarkable aspect of this method
heretofore had not been realized. New fractographic techniques were discovered that help
detect and characterize the tiny precracks.
The razor-paste SEVNB round robin corroborated other research that showed the
maximum notch tip width for valid results can be directly related to the microstructure of
the material.
The first VAMAS hardness round robin [1] identified critical problem areas in
conventional hardness testing. This prompted NIST to investigate the matter more
carefully, to prepare standard reference hardness blocks, and in conjunction with fracture
toughness research, to discover an new brittleness parameter as described in Ref. 10.
8. Round robins foster good communications between standards researchers around the
world
VAMAS projects have enhanced "grass-roots" collaborations between scientists,
engineers, industrialists, and standards writers in a neutral environment at a
prestandardization level. Experts argue about technical details at an earlier rather than later
stage. International standardization can be expedited since many of the methods have been
tried out at an earlier stage on a non-confrontational basis. Many of the VAMAS project
leaders are task group leaders in ASTM, CEN, and ISO.
Table 2 - - Quinn's rules of thumb for round robins.
1. Have a specific, focused objective. Do not undertake one "just for the fun of it!"
2. Run a 2 to 3 laboratory mini round robin first.
3. Keep it to less than one-man week of work per participant.
4. Make certain that the material is consistent and uniform.
5. Give the participants extra specimens or material. Keep extra material as a reserve.
6. Write the instructions very, very carefully.
7. Don't add too many "interesting" side issues.
8. Start with no fewer than 6 to 8 labs.
9. Expect the unexpected. Murphy's laws are very active!
10. Consult a statistician, if there is any doubt, both before and after running the project.
Keys to Running Successful Round Robins
Having organized or participated in thirteen VAMAS and five non-VAMAS round

robins, and observed several dozen others from the sidelines, the author has tabulated a set
of Rules of Thumb for round robins in Table 2. Rule 4 on ensuring material consistency is
very important and often the most difficult to verify with ceramics.
Rule 2 on pre-round robins is prudent since organizers may ferret out gaps in the
procedure, misunderstandings, or errors in the instructions. Rule 5 on extra specimens is
important if some prescribed experiments are invalid and need to be redone. Participants
are less fearful o f inadvertently ruining a few specimens or experiments. The extra
specimens are also an enticement for many participants since they often try their
alternative favored procedures. The round robins inevitably raise some interesting
technical questions and the extra specimens enable some eager laboratories to pursue
these issues. The extra material or specimens give flexibility to a project and creates
enthusiasm among the participants enhancing the chances o f obtaining responses from
them.
During the data review and preparation o f the final report, the project organizers
should comprehensively interpret the data rather than merely recount numerical results.
It is not unusual for a VAMAS TWA 3 final report to include participants' qualitative
assessments such as: "This method produces an accurate scientific result, but is too
difficult to be used on a practical basis." Such assessments may be just as valuable as
the test data!
Conclusions
The VAMAS TWA 3 program has been a key to standardizing good fracture
toughness methods. One may hope that the ceramic community will begin using the new
formal standards but this probably will be a gradual process. The allure of the
rudimentary indentation methods continues to draw users away from better test methods,
no matter how dubious the quality of the indentation method data. 6 The formal standards
will restore credibility to fracture toughness testing of ceramics. They do require a little
more work, but the quality of the data warrants the extra effort. A little experience goes a
long way towards alleviating concerns about the difficulty of a particular method.
The VAMAS program has emphasized quality test methods for quality data and
quality standards. Quality methods produce accurate and precise data and are practical test
methods that are based on sound principles and good metrology.
Acknowledgments
Support for this program was furnished by the Propulsion Systems Materials Program,
U. S. Department of Energy, Office of Transportation Technologies, through the Heavy
Vehicle Propulsion System Materials Program. The author wishes to thank the members of
TWA 3 who so enthusiastically contributed to this program during the ten years the author
was chairman.
Appendix I Bibliography
Each project had a comprehensive final report that is usually available from the authors.
Open literature summary papers for each project are also listed in this bibliography.
Fracture Toughness Project I, SEPB, IS, IF
a. Awaji, H., Kon, J., and Okuda, H., "The VAMAS Fracture Toughness Test Round-Robin
on Ceramics," VAMAS Report #9, Dec. 1990, (Available from the authors at the Japan
Fine Ceramics Center, Nagoya).
b. Awaji, H., Yamada, T., and Okuda, H. "Results of the Fracture Toughness Test Round
Robin on Ceramics, VAMAS Project," Journal of the Ceramic Society of Japan, Int.
Edition, Vol. 99, 1991, pp. 403-408.
C, Quinn, G. D, Salem, J. A., Bar-On, 1., Chu, K., Foley, M. J., and Fang, H., "Fracture
Toughness of Advanced Ceramics at Room Temperature," Journal or Research of the
National Institute of Standards and Technology, Vol. 97, No. 5, 1992, pp. 579-607.
d. Quinn, G. D., "Fracture Toughness of Advanced Ceramics at Room Temperature: A
VAMAS Round Robin," Ceramic Engineering andScience Proceedings, Vol. 14, No. 7-8,
1993, pp. 75-83.
Inevitably we hear that the advantages of the indentation methods are that they use very little
material, are easy to do, and furnish a comparative value of fracture resistance that is "good
enough." Years of conflicting and poor results, and dozens of "refined indentation
equations" and calibration constants underscore the unsoundness of these attitudes.
Fracture Toughness Prqject II, High Temperature, SEPB, CN
Mizuno, M. and Okuda, H., "VAMAS Round Robin on Fracture Toughness of Silicon
Nitride at High Temperature," VAMAS Technical Report #16, December 1993. (Available
from the authors at the JFCC, Nagoya)
b. Mizuno, M., Murata, H., and Okuda, H., "VAMAS Round Robin on Fracture Toughness of
Silicon Nitride at High Temperature," pp. 585-591 in New Horizons for Materials
Proceedings of the 8th World Ceramic Congress, CIMTEC, Florence, Italy, ed. P.
Vincenzini, Techna, Faenza, 1995.
C. Mizuno, M. and Okuda, H., "VAMAS Round Robin on Fracture Toughness of Silicon
Nitride," Journal of the American Ceramic Society, Vol. 78, No. 7, 1995, pp. 1793-1801.
Fracture Toughness Project IlI, SCF
a. Quinn, G. D., Kiibler, J. J., and Gettings, R. J., "Fracture Toughness of Advanced Ceramics
by the Surface Crack in Flexure (SCF) Method: A VAMAS Round Robin," VAMAS
Technical Report # 17, June, 1994, (Available from the authors at NIST or EMPA)
b. Quinn, G. D., Kiibler, J. J., and Gettings, R. J., "Fracture Toughness of Advanced Ceramics
by the Surface Crack in Flexure (SCF) Method: Results of the VAMAS Round Robin,"
Ceramic Engineering andScience Proceedings, Vol. 15, No.5, 1994, pp. 846-855.
C. Quinn, G. D., KiJbler, J. J., and Gettings, R. J., "Fracture Toughness of Advanced
Ceramics: A New VAMAS Round Robin," in New Horizons for Materials, Proceedings of
the 8th World Ceramic Congress, CIMTEC, Florence, Italy, ed P. Vincenzini, Techna,
Faenza, 1995, pp. 577-584.
Fracture Toughness Project IV, Composites
a~ Mizuno, M., and Kon, J., "VAMAS Round Robin on Fracture Toughness Measurement of
Ceramic matrix Composite," VAMAS Report #32, JFCC, Nagoya, September 1997.
b. Mizuno, M., Nagano, Y., Cao, J., and Kon, J., "VAMAS Round Robin on Fracture
Toughness Measurement of Ceramic Matrix Composite," pp 895-902 in Ceramics: Getting
into the 2000's, Part D, ed P. Vincenzini, Techna Srl, 1999 (proceedings of CIMTEC
1998.)
Fracture Toughness Project V, SEVNB
Kiibler, J. J., "Fracture Toughness of Ceramics Using the SEVNB Method: Initial Results
for Si3N4 of a Joint VAMAS/ESIS Round Robin," presented at the 9th CIMTEC
Conference, Florence, Italy, June 14-19, 1998.
b. Kiibler, J. J., "Fracture Toughness of Ceramics using the SEVNB Method: First Results
of a Joint VAMAS/ES1S Round Robin," Ceramic Engineering & Science Proceedings,
Vol. 20, No. 3, 1999, pp 495-502.
C. Kithler, J. J., "Round Robin on Fracture Toughness of Ceramics Using the SEVNB
Method," VAMAS Report #37, ES1S # D2-99, EMPA, Dubendorf, Switzerland, 1999.
References
[1] Quinn, G. D. "VAMAS After Twelve," Bulletin of the American Ceramic Society,
Vol. 78, No. 7, 1999, pp. 78-83.
[2] Nose, T. and Fuji, T., "Evaluation of Fracture Toughness for Ceramic Materials by a
Single-Edge-Precracked-Beam Method," Journal of the American Ceramic Society,
Vol. 71, No. 5, 1988, pp. 328-33.
[3] Evans, A. G. and Charles, E. A., "Fracture Toughness Determination by
Indentation," ibid, Vol. 59, No. 7-8, 1976, pp. 371-2.
[4] Anstis, G, R., Chantikul, P., Lawn, B. R., and Marshall, D. B., "A Critical
Evaluation of Indentation Techniques for Measuring Fracture Toughness: I, Direct
Crack Measurements," ibid, Vol. 64, No. 9, 1981, pp. 533-8.
[5] Chantikul, P., Anstis, G. R., Lawn, B. R., and Marshall, D. B., "A Critical
Evaluation of Indentation Techniques for Measuring Fracture Toughness: II,
Strength Method," idem, pp 539-43.
[6] Petrovic, J. J. and Mendiratta, M. G., "Fracture from Controlled Surface Flaws," in
Fracture Mechanics Applied to Brittle Materials, ASTM STP 678, S. W. Freiman,
ed, American Society for Testing and Materials, 1979, pp. 83-102.
[7] Quinn, G. D., Xu, K., Gettings, R. J., Salem, J. A., and Swab, J. J. "Does Anyone
Know The Real Fracture Toughness? SRM 2100: The World's First Ceramic
Fracture Toughness Reference Material, "this volume.
[8] Salem, J. A., Ghosn, L. J., and Jenkins, M. G., "Back Face Strain as a Method for
Monitoring Stable Crack Extension," Ceramic Engineering and Science
Proceedings, Vol. 19, No. 3-4, 1998, pp. 587-594.
[9] Baratta, F. I. and Fett, T., "The Effect of Load and Crack Misalignment on Stress
Intensity Factors for Bend-Type Fracture Toughness Specimens," Journal of
Testing and Evaluation, Vol. 28, No. 2, 2000, pp. 96-102.
[10] Quinn, J. B. and Quinn, G. D., "Indentation Brittleness of Ceramics: A Fresh
Approach, "Journal of Materials Science, Vol. 32, 1997, pp. 4331-4346.
Crack Growth Resistance
Shuji Sakaguchi'
R-Curve Measurement of Silicon Nitride Based Ceramics at Elevated

Temperatures with Single Edge Notched Beam Specimens
Reference: Sakaguchi, S., "R-Curve Measurement of Silicon Nitride Based

Ceramics at Elevated Temperatures with Single Edge Notched Beam Specimens,"
Fracture Resistance Testing of Monolithic and Composite Brittle Materials, ASTM STP
1409, J. A. Salem, G. D. Quinn, and M. G. Jenkins, Eds., American Society for Testing
and Materials, West Conshohocken, PA, 2002.
Abstract: In order to understand the fracture behavior of the structural ceramics at

elevated temperatures, the R-curve behavior of silicon nitride based ceramics was
measured by using single edge notched beam (SENB) specimens, at room and elevated
temperatures. It is difficult to obtain stable fracture with the SENB method, but if we
cut the straight notch deep enough, we can obtain the stable fracture in bending, even
though the specimen is brittle and shows no rising R-curve behavior. This technique
is easy to apply in high temperature testing. Three kinds of silicon nitride based
ceramics were tested up to 1200 ~ We could measure the R-curve behavior, if the
specimen did not show creep deformation. This technique is quite useful for
discussing fracture behavior of ceramics at elevated temperatures.
Keywords: fracture resistance, silicon nitride, ceramics, elevated temperature, bending
Introduction
Silicon nitride ceramics are expected to be used for mechanical components,

because they have excellent mechanical properties at elevated temperatures. Because
ceramic materials are typical brittle materials, it is important to understand their
fracture behavior [1-6]. R-curve behavior is the relation between the crack length
and the critical stress intensity factor, and it indicates the fracture resistance of a
material. The single edge notched beam (SENB) is one of the test methods for
measuring fracture toughness of ceramics [7]. In regular SENB tests, the specimens
break instantaneously, and we could not discuss the relation between the crack length
and the stress intensity factor. However, with an analysis, we show that stable fracture
can be obtained in a SENB test if the notch is cut deep enough. The SENB test is then
easily applied for high temperature testing. We have tried to measure the R-curve
behavior of several silicon nitride based ceramics at elevated temperatures up to
1200~
1 Team Leader, Materials Evaluation and Characterization Team, Synergy Materials

Research Center, National Institute of Advanced Industrial Science and Technology,
2268-1, Shimo-Shidami, Moriyama-ku, Nagoya 463-8687 Japan.
129
t30 FRACTURETESTING OF MONOLITHIC/COMPOSITE MATERIALS
Calculation
The stress intensity factor on a SENB specimen tested with four-point bending is
calculated using Equations 1 to 3 [8]. These equations are valid in the condition of
O<a/W< l .
P S+-$2
K+ = B W +/2 ( W:--) Yea)
(1)
3 /~M 0r 1/2
Y 6o) . . . . .
2 (1- e ) 1/2
(2)
( 3 . 4 9 - 0 6 8 a +1.35 a 2) a (1- a )
FM =1.9887- 1.326 c~ . . . . . . . . . .
(l + a) 2
(3)
where
K1 = stress intensity factor,

P = applied load,
B = width of the bending bar,
W = height of the bending bar,
$1 = outer span of the 4-point bending,
$2 = inner span of the 4-point bending, and
a = dimensionless depth of the notch (= a/W, where 'a' is the notch depth).
The compliance of the bending bar, as a function of the crack extension is

calculated using Equations 4 and 5
1 $1-$2 a
CcO)=E.B ( ~ )2 f 2Y2(x)ak:+Co
0
(4)
1 S,,-$2 2 S++2Sz
CO=E,B (-W) ( 4 W '---)
(5)
where
E ' = E / (1 - v 2) (E: Young's modulus, I,:.Poisson's ratio)
We obtained an approximate polynomial equation for the calculation of the

compliance, instead of Equation 4, as it includes an integral calculation
1 St -$2
C= E'B ~W -) 2D+C~
(6)
SAKAGUCHI ON SILICON NITRIDE BASED CERAMICS 131
a (001362 +830450l -2L664 a 2 +30486 a s-23.668 a 4 +7.5197a s

D =
(1 - a) 2 (7)
For the case of three-point bending whose span-to-height ratio (=S/W) is 4,

Equation 3 shall be changed to Equation 8
1.99- a (1-oL) ('215-3.93 a +27 a 2)

FM--
(1 +2 ce) (8)
In this case, Equation 7 shall be changed to Equation 9
a ( 0 . 0 2 2 1 5 + 7 . 9 0 2 0 ce - 2 3 . 8 0 4 ce 2 + 3 8 8 1 2 a 3_ 33.028 a 4 + 11.090 a ~
D=
(1- ce) 2 (9)
If we give fixed values of stress intensity factor and the elastic constants, we can
obtain the change of the deflecting displacement of the bending bar by changing the
notch depth. Figure 1 shows the relation of the fracture load, fracture displacement,
and the notch depth by applying the parameters of typical silicon nitride (K~c= 5.0 MPa
m ~/2, E = 310 GPa, v = 0.25) to four-point bending with 30-10 mm spans. The height
and the width are 4 mm and 3 mm, respectively. The numbers on the curve indicate
the dimensionless depth of the Jnotch.
This figure shows that if the dimensionless notch depth is 0.2, a displacement of
the bending bar of about 14 btm is required for starting fracture. The displacement
does not decrease in a usual bending test. Thus, the stress intensity factor increases
with crack extension, bringing about an instantaneous fracture. On the other hand, if
the dimensionless notch depth is 0.8, a displacement of the bending bar of about 12 ~m
is required for starting the fracture. In this case, once the crack extends, it requires an
200
150 1
z
"~
o, 100 I
i
i
50 i
i
0
0 5 10 15 20 25 30
Displacement (~, m)
Figure 1 - Calculated load-displacement curve f o r four-point bending o f a typical

silicon nitride. The numbers on the curve are the dimensionless notch depth ct (= a/W).
increasing displacement to maintain the value of the stress intensity factor. This
means stable fracture proceeds even if the specimen is brittle and shows no rising R-
curve.
Experiments
Three kinds of silicon nitride based ceramics were tested [9]; (1) hot pressed
silicon nitride (HPSN), (2) hot pressed silicon nitride containing 20wt% silicon carbide
whiskers (SCwSN), and (3) gas pressure sintered silicon nitride (GPSN). All
specimens were commercially supplied, and contained yttria and alumina as sintering
additives. Figure 2 shows typical fracture surfaces of the three kinds of specimens,
which were observed with scanning electron microscope. It shows that the SCwSN
has the smallest grain size and GPSN has the largest grain size among these specimens.
Test pieces were bending bars of 4 mm height and 3 mm width. The depth of the
notch was 3.5 mm for HPSN, 3.6 mm for SCwSN and 3.0 mm for GPSN. The width
of the notch was 0.1 mm. Bending tests were performed in four-point bending with
spans of 30-10 mm. The test fixture was made of silicon carbide. Crosshead speed
of the bending test was 1 ~tm/min. R-curves were obtained at room temperature,
IO00~ llO0~ and 1200~ in air. Three to five specimens were tested for each
Figure 2 - Scanning electron micrographs of the fracture surface of the tested silicon
nitride based ceramics: (a) HPSN, (b) SCwSN, (c) GPSN.
condition. The crosshead position and the applied force during the bending tests were
recorded, and the displacement of the bending bar was calculated from the recorded
values and the Young's modulus o f the specimen. The values o f Young's modulus for
this calculation were obtained in a previous work [10].
Results
Figure 3 shows the R-curve o f HPSN at room temperature, 1000~ ll00~ and
1200~ It showed no rising R-curve at room temperature, but it showed a rising R-
curve at elevated temperatures. The scatter o f the data was similar for the room
temperature and the high temperature tests. Figure 4 shows the R-curve o f SCwSN at
each temperature. We succeeded in obtaining an R-curve for only one specimen at
room temperature. However, all the specimens broke stably in elevated temperature
tests. It showed no rising R-curve either at room or at elevated temperatures. Figure
5 shows the R-curve o f GPSN at each temperature. GPSN showed a rising R-curve
both at room and elevated temperature. The data was invalid in the tests at 1200~
because creep deformation was observed. The measured load did not become zero,
but kept a constant value while the displacement was increased. This was caused by
10 ! i
8 8
iltllilAiAI
6 -
,,L,t
v&l,olp i .
#.
~4 i
I
4 I-211=*"
,i
2 --R.T. 5 specimens 2 10oo~ 5 specimens
t I
0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
Crack Extension (turn) Crack Extension (rnm)
10 10
8 r ~
8
i T l l * * ' . '_~__lliiO i it,Ill, z''

Q. ilr-
~4i j i ! ~4
l
2 1i 00~ 5 specimens - - 1200~ - - - 5 specimens-
o
I:
I 0 t
I
0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
Crack Extension (ram) Crack Extension (mm)
Figure 3 - R-curve behavior of hot pressed silicon nitride (HPSN) at room temperature,
IO00~ llO0~ and 1200~ A rising R-curve can be seen in the results at elevated
temperatures.
10 I 10 i
I
8
I 8
i
C
} i I
~Oo OiO 9 i
Zlilitltl II l', 9 I
4
v,
2 ~.m. 1 specimen 2 -----iO00oc - .5 specimens
1 0
I
0
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
Crack Extension (mm) Crack Extension (ram)
10
t I1
8
0 .L '
0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 04 0.5
Crack Extension (ram) Crack Extension (ram)
Figure 4 - R-curve behavior of hot pressed silicon nitride containing 20wt% silicon
carbide whiskers (SCwSN). No rising curves were observed either at room or
elevated temperatures.
creep deformation. The calculation of the crack length from the displacement of the
bending bar was not correct. The increase of the apparent compliance was not caused
by the crack extension, but by the creep deformation.
Discussion
The rise in the R-curve became larger with ascending temperature for HPSN,
because grain boundary fracture was likely to occur at high temperature. We could
not see a rising R-curve for SCwSN specimens, even at elevated temperature.
SCwSN has the smallest grain size among these materials, as shown in Figure 2. The
silicon carbide whiskers inhibited the grain growth of silicon nitride matrix. The
crack deflection within the SCwSN specimens must be less than that in HPSN or GPSN
specimens because of the fine microstructure. This implies that the rising R-curve is
influenced by the grain size.
The initial fracture toughness of each specimen decreased with ascending
temperatures. It may be related to the subcritical crack growth or softening of the
grain boundary. A crosshead speed of 1 l.tm/min was used in these tests. The
influence of the subcritical crack growth needs to be considered, as we could observe
12 12
10 !',= e 10
= 8
~. 6
v
_~ 4 ~4
2 R.T. 4 specimens 1000~ .... 4 specimens
I I
0.2 0.4 0.6 0.8 0 02 0.4 0.6 0.8
Crack Extension (mm) Crack Extension (ram)
12 2C
48
10
~=15 I 9 9 ~
I
I *
~ ~
,t
~- ~o ,200~176 t
~4
5~ ~- 5
2 ~ioooc - - 4 specimen9
0 I 0 / i
0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8
Crack Extension (ram) Crack Extension (ram)
Figure 5 - R-curve behavior o f gas pressure sintered silicon nitride (GPSN). Rising R-
curve can be observed at room temperature, 1000~ and 1100~ The influence o f the
creep deformation was observed on the result at 1200~
the influence of subcritical crack growth on the R-curve behavior of alumina ceramics
[11]. If the crack extension occurs easily at elevated temperatures by subcritical crack
growth, the initial fracture toughness becomes lower.
The fracture resistance for long cracks was not changed so much with ascending
temperatures. This may suggest that the fracture resistance at longer crack lengths is
influenced by crack deflection and by grain bridging. The effects of crack deflection
and grain bridging for the fracture do not depend on the temperature, as they are
determined by the geometrical conditions of the grains.
Conclusion
The R-curve behavior of three kinds of silicon nitride based ceramics was
measured at room and elevated temperatures by using SENB specimens. We could
obtain stable fracture of SENB specimens if the machined notch was deep enough.
This technique is quite useful for discussing the fracture behavior of ceramics at
elevated temperatures. However, the analysis is not valid if the specimens show creep
deformation, as the crack length is calculated from the deflection of the bending bars.
References
[1] Nakayama, J., Abe, H. and Bradt, R. C., "Crack Stability in the Work-of-
Fracture Test: Refractory Applications," Journal of the American Ceramic
Society, Vol. 64, 1981, pp. 671-675.
[2] Krause Jr., R. F., "Rising Fracture Toughness from the Bending Strength of
Indented Alumina Beams," Journal of the American Ceramic Society, Vol. 71,
1988, pp. 338-343.
[3] Steinbrech, R. W., Riedel, A. and Schaarwaechte, R. W., "R-Curve Behavior of
Long Cracks in Alumina," Journal of the American Ceramic Society, Vol. 73,
1990, pp. 2009-2015.
[4] Ramachandran, N. and Shetty, D. K., "Rising Crack-Growth-Resistance (R-
curve) Behavior of Toughened Alumina and Silicon Nitride," Journal of the
American Ceramic Society, Vol. 74, 1991, pp. 2634-2641.
[5] Fett, T. and Munz, D., "Subcritical Crack Growth of Macro-cracks in Alumina
with R-curve Behavior," Journal of the American Ceramic Society, Vol. 75,
1992, pp. 958-963.
[6] Nishida, T. and Hanaki, Y., "Measurement of Rising R-curve Behavior in
Toughened Silicon Nitride by Stable Crack Propagation in Bending," Journal of
the American Ceramic Society, Vol. 78, 1995, pp. 3113-3116.
[7] Pabst, R. F., Steeb, J. and Clausen, N., "Microcracking in a Process Zone and Its
Relation to Continuum Fracture Mechanics," Fracture Mechanics of Ceramics,
Vol. 4, R. C. Bradt, D. P. H. Hasselman, and F. F. Lange, Eds., Plenum Press,
New York, 1977, pp. 821-833.
[8] Munz, D., Bubsey, R. T. and Shannon Jr., J. L., "Fracture Toughness
Determination of A1203 Using Four-Point-Bend Specimens with Straight-
through and Chevron Notches," Journal of the American Ceramic Society, Vol.
63, 1980, pp. 300-305.
[9] Sakaguchi, S., Murayama, N., Kodama, Y. and Wakai, E, "R-Curve
Measurement of Silicon Nitride Ceramics Using Single-Edge Notched Beam
Specimens," Journal of Materials Science, Vol. 29, 1994, pp. 5183-87.
[10] Sakaguchi, S., Murayama, N., Kodama, Y., and Wakai, F., "The Poisson's Ratio
of Engineering Ceramics at Elevated Temperature," Journal of Materials
Science Letters, Vol. 10, 1991, pp. 282-84.
[11] Sakaguchi, S., "R-Curve Measurement of Alumina Using Straight Notched or
Indented Bending Bars," Fracture Mechanics of Ceramics, Vol. 11, R. C. Bradt,
D. E H. Hasselman, D. Munz, M. Sakai, and V. Ya. Shevchenko, Eds., Plenum
Press, New York, 1996, pp. 39-51.
Hideo Awaji, 1 Masaki Ebisudani, l Seong-Min Choi, t and Tsuguhiro Ohashi I
Crack Deflection Toughening Mechanism in Brittle Materials
Reference: Awaji, H., Ebisudani, M., Choi, S-M., and Ohashi, T., "Crack Deflection
Toughening Mechanism in Brittle Materials," Fracture Resistance Testing of
Monolithic and Composite Brittle Materials, ASTM STP 1409, J. A. Salem, G. D. Quirm,
and M. G. Jenkins, Eds., American Society for Testing and Materials, West
Abstract: Toughening mechanisms in polyerystalline ceramics and ceramic-based

composites were discussed based on Griffith's energy equilibrium and classified into
three groups: (A) the frontal process-zone toughening mechanism, (B) the crack-bridging
mechanism, and (C) the macroscopic crack-deflection mechanism. The increase in
fracture toughness caused by enforced macroscopic crack-deflection was estimated using
the energy equilibrium between the energy-release rate and the fmettae,energy rate. The
analytical results revealed that the fracture toughness under Mode I loading was
enhanced by twice the normal value when the crack propagates in the direction 90~ from
its original plane. Under Mode II loading the fracture toughness is increased 1.2 times
the normal value when the crack propagates in its original plane. Crack-deflection tests
were carried out by using wooden, thin rectangular plates with a central crack. The
increase in the fracture toughness caused by crack deflection was confirmed and the
experimental data was coincident with the calculated results.
Keywords: toughening mechanism, crack deflection, self-reinforced ceramics, Griffith

criterion, multilayered composites
Introduction
tProfessor, and graduate students, respectively, Nagoya Institute of Technology,

Gokiso-cho, Showa-ku, Nagoya, 466-8555 Japan.
137
Over the past decade, the most attractive research area in toughened ceramics was
nanocomposites. Niihara [1] developed a new concept of nanocomposites and showed a
drastic improvement in the toughness and strength of ceramics. The microstructure of the
nanocomposites was composed of second-phase nano-size particles dispersed within the
matrix grains. Thermal expansion-mismatch between the matrix and the second-phase
particles produced the remarkable improvement in fracture strength by additional
annealing, which was due to the further development of sub-grain boundaries around the
second-phase particles. Although a certain skepticism about the nano-toughening
mechanism was proposed by several researchers [2-4], Nawa and co-workers [5] recently
developed a 3Y-TZP/Mo nanocomposite material that showed a maximum strength of
2.1 GPa (three-point bending) and a fracture toughness 11.4 MPa" m 1/2 a s measured by
the SEVNB (single-edge V-notched beam) technique [6-8]. This technique can be used
to estimate the intrinsic fracture toughness, K~, of the material with rising R-curve
behavior, where the crack resistance of the material, KR, is expressed as
K R(Act) = K, + AK R(Aa) (1)
where d KR(A a) represents the extrinsic increase in the crack resistance.

Another attractive, new material system is ceramic multilayered composites, which
exhibit significant improvements in fracture resistance and flexural strength [9-23].
Clegg et al. [9] clarified the toughening mechanism of the multilayered system and
revealed that the apparent fracture toughness increased more than fourfold for the system
of SiC layers with weak graphite interfaces, relative to the monolithic SiC material. The
key feature of this mechanism is the presence of macroscopic crack-deflection between
the layers [10,11,14]. Based on the concept, several multilayered ceramic systems for
enhancing structural reliability were fabricated, such as ZrO2/AI203 [13-17], Si3N4
dense/porous layers [18], and AI203/Ni [19], and these materials showed the significant
increase in both strength and toughness. On the other hand, Hirao and coworkers [20-22]
proposed a new technique for controlling orientation of elongated seed particles of
self-reinforced Si3N4 with a bimodal microstructure composed of small matrix grains and
large elongated grains, and obtained highly toughened and strengthened ceramics. Kondo
et al. [23] also proposed a new Si3N4 with highly anisotropic microstructure having
two-dimensionally aligned rod-shaped grains produced by superplastic deformation.
These dominant mechanisms of enhanced fracture resistance are, however, basically
founded on a macroscopic crack-deflection behavior, like the multilayered composites. It
is, therefore, essential to understand why fracture toughness in these systems can he
AWAJI ET AL. ON CRACK DEFLECTIONTOUGHENINGMECHANISM 139
improved so significantly over their monolithic counterparts.

Based on the above descriptions of toughened ceramics, we classified the
toughening mechanisms in polycrystalline ceramics into three groups [24]. (A) The
frontal process zone toughening mechanism, which is related to the intrinsic fracture
energy of a material and creates a damaged zone ahead of the crack tip. This mechanism
generates intrinsic fracture toughness, K~, in materials. (B) The crack-bridging and/or
interlocking toughening mechanism working in a process-zone wake, which produces an
extrinsic increase in crack resistance, A KR(A a), after a certain extension from the
initial crack. (C) The macroscopic crack-deflection mechanism, which decreases the
energy-release rate and, consequently, the stress-intensity factor at the crack tip.
The toughening mechanism in nanocomposite materials is related to mechanism
(A), in which sub-grain boundaries around the second-phase particles, observed by
Niihara [1], are expected to create many microcracks around the crack tip, increase the
frontal process-zone size, and then enhance the intrinsic fracture toughness of the
material.
Particle dispersion is the most popular toughening mechanism in ceramics.
Dispersed particles can be elongated grains, ductile particles, platelets, whiskers or short
fibers [25]. This toughening can be explained as a crack deflection or pinning in a frontal
process zone, and bridging in a wake. Therefore, the former belongs to the frontal
process zone toughening mechanism (group A), and the letter the crack-bridging
mechanism (group B).
The present paper gives a basic explanation of the crack-deflection mechanism of
(C), in which the fracture toughness of a material increases by controlling the direction
of crack growth. The fracture toughness for macroscopically deflected crack growth is
analyzed using the energy equilibrium between the energy-release rate and the
fracture-energy rate, known as Griffith's energy equilibrium [25-27]. The present results
reveal that controlling crack growth in the direction 90 ~ from the original crack surface
under Mode I loading produces an apparent increase in the fracture toughness of about
twice the normal value, Klc, for a crack extending in the direction of the maximum
energy release rate.
Theory
Griffith's energy equilibrium for brittle crack extension is equivalent to the

maximum energy release rate theory, which states that a crack will extend in the direction
of the maximum energy release rate, when the maximum energy release rate reaches the
fracture energy rate o f the material. For Mode I loading, Crriitith's energy equilibrium is
expressed using the Irwin's notation, as follows
r'~ = 2r, (2)

E'
where :),1is the Mode I fracture energy per unit cracked surface area, E ' = E for the
plane stress condition and E ' =E/(1-v 2) for the plane strain condition, where E is
Young's modulus, and v is Poisson's ratio. The left side of Eq (2) indicates the critical
value of the maximum energy release rate, G = Go, and the fight side the fracture-energy
rate of the material, R.
The attractive point of the Crriffith concept is that the equation for the energy
equilibrium is composed of two discrete parts. One part is the critical energy-release rate,
Go, as expressed on the left side of Eq (2). The energy-release rate is not dependent on
the nonlinear stress state in the process zone area in front of a crack tip. In other words,
the energy-release rate is precisely determined by the stress intensity factor which
depends only on the boundary conditions, namely, the applied loading and specimen
geometry, as long as the condition of small scale yielding is applicable. The second part
is the fracture energy rate, as expressed on the right side of Eq (2), which is directly
related to the damage process zone, represents the resistance of the material that must be
overcome for crack growth to occur and ideally is a material constant.
Referring to Eq (2), it is recognized that straightforward toughening of brittle
materials is achieved by increasing the fracture energy of the material, ~'/. Actually,
much effort has been directed toward improving the fracture energy of ceramics in ways
such as nanometer- and micrometer-size dispersions of ceramic or metallic particles,
whisker reinforcement in the matrix, and transformation toughening [25]. However, all of
these mechanisms are classified into the frontal process zone toughening mechanism
(Group A) and the crack-bridging mechanism (Group B) described above.
Equation (2) also suggests that there is an alternative method for increasing
fracture toughness, which involves decreasing the energy-release rate, as shown on the
left side of Eq (2), rather than increasing the fracture-energy of a material. If the direction
of crack extension can be controlled and the crack deflects in the non-maximum
energy-release rate ~ t i o n , the crack-deflection leads to a reduction of energy-release
rate, which will generate apparent increase of the fracture toughness of the material.
In the present paper, we assume that the material with a crack is homogeneous and
brittle, and that the fracture-energy is not dependent on the direction of crack extension.
AWAJI ET AL. ON CRACK DEFLECTIONTOUGHENINGMECHANISM 141
x K~P',~K m
KT
Figure 1 - Kinked crack under combined Mode 1-11loading.
Under combined Mode I-II loading, a crack extends in non-eoplanar directions, as shown
schematically in Fig. 1. The energy-release rate was analyzed by several researchers for
an infinitesimally kinked crack extension under combined Mode I-II loading. Hussain et
al. [28] proposed the following expression for the energy-release rate, using the stress
intensity factors at the kinked crack tip
G(8) = 1 (K~ + K~k) (3)

E'
where G(R) represents the energy-release rate when the kinked crack extends in the @
direction, Klk and Km Mode I and Mode II stress intensity factors defined at the kinked
crack tip, respectively, as shown in Fig. 1. The relationship of Eq (3) is also derived and
confirmed by other researchers, such as Wu [29], Kageyama and Okamura [30,31], and
Ichikawa and Tanaka [32,33]. Here, we use the result of Kageyama and Okamura [31],
which was confirmed by Ichikawa [33]
6(o) -- ~ [h, (a)K,~ + 2h,, (O)K, K,, + h, (O)K,~] (4)

/2,'
where K1 and Kn represent Mode I and Mode II stress-intensity factors at the original
crack tip, respectively. The factors, hi, h12 and h2 in Eq (4), are expressed as functions of
kinked crack orientation as follows [31]
/% : l e o s 2 0~(1+ cos8)(1 - 0.003r 2 + 0.041r +)

z 2
/%2 = -cos2 O--sinO(1 - 0.003~ 2 + 0.027~')
2 (5)
h2 = leos2 ~--(5- 3cosO)[1 + sin 2 0(0.168 + 0.02cos30)]
2 2
= 0 / 90 (deg), or 20 / ~r (radian)
For Mode I loading, Eq (4) is simply expressed as
~(o) = ~/% K~ (6)

15
where h; is a nondimensional energy-release rate. The relationship between hi and 8 is

shown in Fig. 2.
The energy equilibrium for the kinked crack extension under Mode I loading is
derived
~K~
E' 2Yl (7)
-- --,----~-- Mode II [ .... /
O(
3'''-60'"-3(j'"~ . . . . 3'0 . . . . 6 ' 0 " '
0 (degree)
Figure 2 - Relationships between the direction of enforced crack deflection and

nondimensional energy release rates for Mode I and Mode II loading.
AWAJI ET AL. ON CRACK DEFLECTIONTOUGHENING MECHANISM 143
t.,jl....... ' I .... I .... I ' ' ' ' I .... I ....
L)
r...) 1.96 ( - 9 0 ~ ) 1.96 (90 ~ )---..._~
r.) Mode I ]
. . . . . . Mode II I
0.~
'(J" "60" '-36 ' ' ' 0 . . . . 3'0. . . . 6'0 . . . . 90
0 (degree)
Figure 3 - Relationships among the direction of the enforced crack deflection, Kl oc/Klc
and Kll oc/Kllc.
where Klac is the critical stress intensity factor at the original crack tip when the til:
growth is enforced in the 0 direction. Hence, from Eqs (2) and (7), the following
relationship between Kic and Klac is derived as
Kl~c= 1 (8)
r,c
Figure 3 shows the relationship between Klac/Klc and 8. It can be seen that when the
crack growth is enforced in the direction 0 = 90 ~ , the apparent fracture toughness is
highest, becoming 1.96 times higher than Klc, the normal fracture toughness.
Figure 4 illustrates the relationships among the maximum energy-release rate, Grnax
= G ( 8 =0) the energy-release rate, G = G(@), for enforced crack-deflection, and the
remote stress on the specimen, a, where Gc represents the critical maximum
energy-release rate, ac the related critical remote stress, and ~ the critical remote
stress when the crack deflects in any directions. It is clear that a reduction of the
energy-release rate leads to the increase of the critical remote stress, and hence the
apparent improvement of fracture toughness.
The same treatment is possible in the case of Mode II loading. Griffith's energy
equilibrium under Mode II loading for a crack extending in the direction of the maximum
energy-release rate, namely -76 ~ from the original crack (see Fig. 2), is expressed as
G~ =-~2/G p_.,'
~C O'ec
remote stress
Figure 4 - Relationships among the maximum energy release rate, Gma. the energy
release rate for enforced crack deflection, G, and remote stress, a.
1.521K~c = 2r. . (9)

E'
Here, Kuc is the Mode II fracture toughness and y u the fracture energy per unit area of
a crack under Mode II loading. The energy-release rate, G(@), for an infinitesimally
kinked crack with an arbitrary angle, O, under Mode II loading is expressed as
G(O) = --~ h, K~ (1 O)
The relationship between h2 and 0 for Mode II loading also is shown in Fig. 2. Then,
the energy equilibrimn for the kinked creek extension is expressed as
h~K~sc = 2y , (11)
E'
where Kntc is the critical Mode II stress-intensity factor when crack growth is enforced
in an arbitrary Odireetion. From Eqs (9) and (11), the following relation is derived
Kna2 _
(12)
AWAJI ET AL. ON CRACK DEFLECTION TOUGHENING MECHANISM 145
The relationship between Kusc/Kllc and 0 is shown in Fig. 3, which indicates that the
apparent fracture toughness is 1.23 times higher than Knc, when the crack growth is
enforced in the direction g = 0 ~
Experimental Procedure
To confirm the analytical results of an apparent increase in fracture toughness

caused by restriction of the propagation direction of the crack, we carried out crack
extension tests using orthotropic rectangular wooden specimens with a central artificial
crack; where the crack extends along the grains (the natural arrangement of the fibers) of
the wooden specimens, as shown in Fig. 5. The wooden specimens were used for
scientific purposes, because ceramics are difficult to control the crack direction.
Specimens 30 by 30 by 3 mm in size were machined from a plate and the central crack
was made at several angles to the grain of wooden specimens. A slit with
chevron-shaped notches at the tips was machined using a thin round diamond saw of 13
mm in diameter and 0.15 mm in thickness. The notch was then extended using a thin
razor blade to roughly alW = 0.7, where 2a is the crack length and 2Wthe width of the
plate. Each rectangular specimen was glued onto soft PVC (polyvinyl chloride) sheets
with narrow slits, on both the top and bottom to reduce misalignment of the loading
system. All tests were performed with CHS (cross-head speed) of 200 mm/min using a
testing machine (Instron 5582).
Figure 5 - Enforced crack deflection test using an orthotropic rectangular wooden specimen.
The stress intensity factor for a central crack in a homogeneous and isotropic
rectangular plate subjected to uniaxial tensile stress is expressed as
/(, = F : r d - ~ 03)
where FI represents the shape factor and ~r the applied remote stress. The shape factors
of the specimens were taken from a handbook [34]. The stress intensity factors for a
crack in a rectangular plate with anisotropie elasticity were analyzed by Bowie and
Freese [35]. Referring to their results, we modified the stress intensity factors in relation
to the anisotropy in the Young's modulus of the orthotropic wooden specimen. The
relationship between the ratio of the Young's moduli, EffE.v, of the wooden plate and the
grain angle was measured using strain gages, where Ex denotes the Young's modulus in
the direction of the load and Ey the value in the vertical direction of the load. Because the
values Ex and Ey are the same in the wooden plate with 45" grain angle, the experimental
results were modified assuming that the wooden plate with 45" grain is an isotropic
material, then normalized using the modified Ktc value. The experimental results and
their modified and normalized values are shown in Table 1, where 0 indicates the grain
angle, Pc the critical load, Ksc~pp the apparent fracture toughness calculated from Eq (13),
9 ~
Kscm~ the fracture toughness modified for the anisotropy, Kt ac normalized critical stress
intensity factors, which is defined as Ks ac" = Ks oc/Ksc, where Kt ac indicates the critical
stress intensity factor when the crack propagates in the 0 direction and Klc = Ks ace e=o)
= 0.70 MPa.m it2, the modified (actual) fracture toughness of the wooden specimen. Ks*
in Table 1 represents correction factors for the anisotropy in elasticity defined as
K~ =____H (14)
F,
where His correction factors analyzed by Bowie and Freese [35].

Figure 6 shows the relationship between the direction of the crack extension and
the normalized critical stress intensity factors, K#sc ~ The theoretically calculated
values of 1(1,c/Klc using Eq (8) are also shown using a broken line. A good coincidence
is exhibited between the experimental and theoretical results.
Clegg et al. [9] reported that the crack of the multilayered SiC/graphite composites
propagated along the weak graphite interface, and the apparent critical stress intensity
factor increased more than four times the normal fracture toughness value of monolithic
Table 1- Results o f the enforced crack extension tests.
0 P, cr, K~'** Kr"*d

degree N a/W Ex/F~ Kl* Kle c*
= MPa MPam1/2 MPam1/z
0 181 0.815 1.55 0.58 21,5 0.840 0.58 0.89
144 0.872 1.49 0.60 0.71 1.01
139 0.669 1.45 0.59 0.70 1,00
123 0.722 1.25 0.57 0.88 0.97
20 147 0.670 1.52 0.62 6.44 0.880 0.72 1.03
147 0.676 1.55 0.64 0.74 1.06
_ 140 0.722 1.45 0.64 0.74 1.06
30 143 0.732 1.50 0.68 3.28 0.899 0.76 1.08
163 0.675 1.70 0.69 0,77 1.10
155 0,893 1.65 0.70 0.78 1.11
138 0,711 1.47 0.64 0.71 1.01
162 0,877 1.67 0.69 0.77 1.10
45 215 0,641 2.16 0.85 1.00 1.00 0.85 1.21
180 0,719 1,82 0.84 0.84 1.20
183 0,721 1.89 0.83 0.83 1.19
182 0.707 1.88 0.82 0.82 1.17
190 0.685 1.94 0.82 0.82 1.17
60 283 0,643 2.87 1.15 0.307 1.22 0,94 1.35
274 0.877 2.82 1.16 0.95 1.36
314 0.643 3.22 1.24 1.02 1.46
265 0.681 2.70 1.14 0.93 1.33
70 377 0.668 3.87 1.57 0,155 1.45 1,08 1.55
325 0.697 3.36 1.45 1.00 1.43
SiC, whereas our results indicate that the critical value is twice in maximum. The three
reasons for the discrepancy will be considered. One is residual stresses in the
multilayered composites. If the thermal expansion coefficients of the layers composed
are different, each layer of the composites has residual stresses with different signs at
i I I I I I
..... -theoritical values [

U 9 experimental values I
12
J-"
I...,.d
0"-0
' 1'0' 2'0' 30
' ' 40' '5'0'6'0
0 [degree]
Figure 6 - Experimentally and theoretically obtained toughening enhanced by
macroscopic crack-deflection.
room temperature. Another is an anisotropic feature of the multilayered composites. The

values of stress intensity factor at a crack tip in an anisotropic material vary with the
anisotropie properties as analyzed by Bowie and Freese [35]. The other reason is the
difference of the definition of the critical stress intensity factor between Clegg etal. and
this work. Clegg etal. calculated the critical stress intensity factor using the ultimate load
the specimen withstood, whereas we used the load when the crack just initiated
propagation.
Conclusions
Toughening mechanisms in polyerystalline ceramics and ceramic-based

composites were discussed based on the Griffith energy criterion. The toughening
mechanisms were classified into three groups. (A) The frontal process-zone toughening
mechanism, which increases the intrinsic fracture-energy rate. (B) The crack-bridging
mechanism, which produces the extriusie increase of fracture energy rate after a certain
extension from the initial crack length. (C) The macroscopic crack-deflection toughening
mechanism, where the crack is forced to deflect at the crack tip. The deflection decreases
the energy-release rate, and consequently increases the critical stress-intensity factor at
the crack tip. The toughening mechanisms of materials with multilayered structures, such
as combinations of tougldweak materials, dense/porous materials and structure controlled
self-reinforced composites belong to this group. In the present work, we clarified
mechanism (C) of enforced macroscopic control of the direction of crack growth. The
analytical results based on Gfiffith's energy equilibrium revealed that the critical stress
intensity factor under Mode I loading was enhanced by twice the normal fracture
toughness when the crack propagates in the direction 90* from its original plane.
Experimental verification was performed by using wooden, thin rectangular plates with a
central crack, where the grains of wooden specimens control the direction of the crack
extension.
References
[1] Niihara, K., "New Design Concept of Structural Ceramics Ceramic

Nanoeomposites -," Journal of the Ceramic Society of Japan, Vol. 99, No. 10,
1991, pp. 974-982.
[2] Pezzotti, G. and Sakai, M., "Effect of a Silicon Carbide "Nano-Dispersion" on the
Mechanical Properties of Silicon Nitride," Journal of the American Ceramic
Society, Vol. 77, No. 11, 1994, pp. 3039-3041.

[3] Carroll, L., Sternitzke, L. C. M., and Derby, B., "Silicon Carbide Particle Size
Effects in Alumina-Based Nanocomposites," Acta Materialia, Vol. 44, No. 11,
1996, pp. 4543-4552.
[4] Perez-Rigueiro, J., Pastor, J. Y., Llorca, J., Elices, M., Miranzo, P. and Moya, J. S.,
"Revising the Mechanical Behavior of Alumina/Silicon Carbide
Nanocomposites," Acta Materialia, Vol. 46, No. 15, 1998, pp. 5399-5411.
[5] Nawa, M., Yamazaki, K., Sekino, T., and Niihara, K., "Microstructure and
Mechanical Behaviour of 3Y-TZP/Mo Nanocomposites Possessing a Novel
Interpenetrated Intragranular MicrostrucaLre," Journal of Material Science, Vol. 31,
1996, pp. 2849-2858.
[6] Awaji, H. and Sakaida, Y., "V-Notch Technique for Single-Edge Notched Beam
and Chevron Notch Methods," Journal of the American Ceramic Society, Vol. 73,
No. 11, pp. 3522-3523.
[7] Awaji, H., Watanabe, T., Yamada, T., Sakaida, Y., Tamiya, H., and Nakagawa, H.,
"Evaluation of Fracture Toughness by a Single-Edge V-Notched Beam Method,"
Transactions of the Japan Society of Mechanical Engineers, Vol. 56, No. 525,
1990, pp. 1148-1153 (in Japanese).
[8] Awaji, H., Watanabe, T., Sakaid~ Y., and Nakagawa, H., "Fracture Toughness
Measurements of Ceramics by V Notch Technique," Ceramics International, Vol.
18, 1992, pp. 11-17.
[9] Clegg, W. J., Kendall, K., Alford, N. McN., Button, T. W., and Birchall, J. D.,
"A Simple Way to Make Tough Ceramics," Nature, Vol. 347, 1990, pp. 455-457.
[10] Folsom, C. A., Zok, F. N., and Lange, F. F., "Flexural Properties of Brittle
Multilayer Materials: I, Modeling," Journal of the American Ceramic Society, Vol.
77, t994, pp. 689-696.
[11] Folsom, C. A., Zok, F. N., and Lange, F. F., "Flexural Properties of Brittle
Multilayer Materials: II, Experiments," Journal of the American Ceramic Society,
Vol. 77, 1994, pp. 2081-2087.
[12] Harmer, M. P., Chart, H. M., and Miller, G. A., "Unique Opportunities for
Microstructural Engineering with Duplex and Laminar Ceramic Composites,"
Journal of the American Ceramic Society, Vol. 75, 1992, pp. 1715-1728.
[13] Requena, J., Moreno, R., and Moya, J. S., "Alumina and Alumina/Zirconic
Multilayer Composites Obtained by Slip Casting," Journal of the American
Ceramic Society, Vol. 72, 1989, pp. 1511-1513.
[14] Marshall, D. B. and Ratto, J. J., "Enhanced Fracture Toughness in Layered
Microcomposites of Cr and A1203," Journal of the American Ceramic

Society, Vol. 74, 1991, pp. 2979-2987.
[15] Nicholson, P. S., Sarkar, P., and Haung, X., "Electrophrcfic Deposition and its Use
to Synthesize ZrO2/AI203 Micro-Laminate Ceramic/Ceramic Composites,"
Journal of Materials Science, Vol. 28, 1993, pp. 6274-6278.
[16] Prokash, O., Sarkar, P., and Nicholson, E S., "Crack Deflection in
Ceramic/Ceramic Laminates with Strong Intcrfac,es," Journal of the American
Ceramic Society, Vol. 78, 1995, pp. 1125-1127.
[17] Tomaszewski, H., Strzeszewki, J., and Gebicki, W., "The Role of Residual
Stresses in Layered Composites of Y-ZrO2 and A1203," Journal of the European
Ceramic Society, Vol. 19, 1999, p. 255-262.
[18] Ohji, T., Shigegaki, Y., Miyajima, T., and Kanzaki, S., "Fracture Resistance
Behavior of Multilayered Silicon Nitfide," Journal of the American Ceramic
Society, Vol. 80, 1997, pp. 991-994.
[19] Chen, Z. and Mecholsky, J. J. Jr., "Toughening by Metallic Lamina in
Nickel/Alumina Composites," Journal of the American Ceramic Society, Vol. 76,
1993, pp. 1258-1264.
[20] Hirao, K., Nagaoka, T., Brito, M. E., and Kanzaki, S., "Microstructure Control of
Silicon Nitfide by Seeding with Rodlike /~-Silicon Nitride Particles," Journal of
the American Ceramic Society, Vol. 77, 1994, pp. 1857-1862.
[21] Hirao, K., Ohashi, M., Brito, M. E., and Kanzaki, S., "Processing Strategy for
Producing Highly Anisotropic Silicon Nitride," Journal of the American Ceramic
Society, Vol. 78, 1995, pp. 1687-1690.
[22] Becher, P. F., Sun, E. Y., Hsueh, C. H., Plucknett, K. P., Kim, H. D., Hirao, K., and
Brito, M. E., "Toughening Behavior in Ceramics and Cermets," Key Engineering
Materials, Vols, 161-163, Edited by Niihara et al., Trans. Tech. Publications,
Switzerland, 1999, pp. 285-290.
[23] Kondo, N., Suzuki, Y., and Ohji, T., " Vickers Indentation Crack in
Superplastically Compressive Deformed Silicon Nitride with Highly Anisotropic
Microstructure," Journal of the Ceramic Society Japan, Vol. 107, 1999, pp.
300-302.
[24] Awaji, H., Choi, S-M., Ebisudani, T., and Jayaseelan, D. D., "Toughening
Mechanisms of Structural Ceramics," Journal of the Ceramic Society of Japan,
Vol. 108, No. 6, 2000, pp. 611-613 (in Japanese).
[25] Green, D. J., An Introduction to the Mechanical Properties of Ceramics,
Cambridge Univ. Press, Cambridge, 1998, pp. 248-266.
[26] Awaji, H., Kato, T., Honda, S., and Nishikawa, T., "Criterion for Combined Mode
I-II of Brittle Fracture," Journal of the Ceramic Society of Japan, Vol. 108, No. 6,
2000, pp. 611-613 (in Japanese).
[27] Awaji, H and Kato, T., "Griffith Criterion for Mode II Fracture of Ceramics,"
Experimental Mechanics, Ed. I. M. Allison, Balkema, Rotterdam, 1998, pp.
1199-1204.
[28] Hussain, M. A., Pu, S. L., and Underwood, J., "Strain Energy Release Rate for a
Crack Under Combined Mode I and Mode II," ASTM STP 560, American Society
for Testing and Materials, West Conshohocken, PA, 1974, pp. 2-28.
[29] Wu, C. H., "Fracture Under Combined Loads by Maximum-Energy-Release-Rate
Criterion," Journal of Applied Mechanics, Vol. 45, 1978, pp. 553-58.
[30] Kageyama, K. and Okamura, H., Transaction of the Japan Society of Mechanical
Engineering, Vol. 48, 1982, pp. 783-791 (in Japanese).
[31] Kageyama, K. and Okamura, H., Stress Intensity Factors Handbook, Vol. 1, Ed. Y.
Murakami et al., Pergamon Press, Oxford, 1987, p.369.
[32] Ichikawa, M. and Tanaka, S., "A Critical Analysis of the Relationship Between the
Energy Release Rate and the Stress Intensity Factors for Non-Coplanar Crack
Extension Under Combined Mode Loading," International Journal of Fracture,
Vol. 18, 1982, pp. 19-28.
[33] Ichikawa, M., "Proposal of An Approximate Analytical Expression of Maximum
Energy Release Rate of A Mixed Mode Crack in Relation to Reliability Evaluation
of Ceramic Components," Journal of the Society of Materials Science, Japan, Vol.
40, 1991, pp. 224-227.
[34] Kitagawa, H. and Yuuki, R., Stress Intensity Factors Handbook, Vol. 2, Ed. Y.
Murakami et al., Pergamon Press, Oxford, 1987, pp.909-911.
[35] Bowie, O. L. and Freese, C. E., "Central Crack in Plane Orthotropic Rectangular
Sheet," International Journal of Fracture Mechanics, Vol. 8, No. 1, 1972, pp.
49-58.
John J. Mecholsky, Jr., 1 Thomas J. Hill,2 and Zheng Chen3
Application of Quantitative Fractography to the Characterization of R-Curve

Behavior
Reference: Mecholsky, J. J., Jr., Hill, T. J., and Chen, Z., "Application of
Quantitative Fractography to the Characterization of R-Curve Behavior,"
Fracture Resistance Testing of Monolithic and Composite Brittle Materials, ASTM
STP 1409, J. A. Salem, G. D. Quinn and M.G. Jenkins, Eds., American Society of
Testing and Materials, West Conshohocken, PA, 2002.
Abstract: The objective of this work is to demonstrate the applicability of quantitative
fractography and fractal analysis to fracture processes in materials that exhibit P-curve
behavior. In-situ toughened silicon nitride and baria-silica glass ceramics are used as
model materials to demonstrate the effect of changes in morphology on toughness.
These two tools, i.e., quantitative fracture surface analysis and fractal geometry,
provide a new way of quantitatively assessing the mechanisms of toughening due to
elongated grains and changes in microstrueture.
For example, in silicon nitride the g-curve behavior only has an effect when
the crack sizes are larger than about 30 micrometers. In applications where small
cracks are expected, i.e., in bearing applications and for other well-polished surface
finishes, R-curve materials may not be desirable.
In another example, 3BaO*5SiO2 glass ceramics were fabricated with crystals
of aspect ratios (AR) of 1.4, 3.6 and 8.1. Based on fractal analysis, fractography and
other accumulated data, it was shown that there was a statistical difference in the
toughening behavior between materials with A_R= 1.4 & 3.6, as compared to the
material with AR = 8.1. The fractal dimensional increment was used to provide a
method for determining the amount of toughening increase over that due to crack
deflections as 0.4-0.6 MPam v2. The implication of these results for the fracture
process of toughened materials and the role of microstructure are discussed and
explained in terms of material design.
Key words: Fracture mechanisms, fractography, fractal geometry, R-curve, fracture

mechanics, toughness.
~Professor, and 2Graduate Student, Departments of ~'ZMaterials Science and

Engineering, and ZDental Biomaterials, University of Florida, Gainesville, FL 32611.
3 Process Engineer, GE Superabrasives, Worthington, Ohio 43085.
152
Copyright9 ASTMInternational www.astm.org
MECHOLSKY ET AL. ON QUANTITATIVEFRACTOGRAPHY 153
Introduction
The strength, o, of most materials which fail in a brittle manner, depends on the
crack size, c, and toughness and can be obtained using fracture mechanics [1]
YKIc
o- ~c (1)
where Y is a well-documented parameter of crack, load and specimen geometry, Kxc,

is the critical stress intensity factor, i.e., a measure of the toughness of the material,
and c = (ab) 1/2 where a is the depth and 2b is the width of a semi-elliptical crack [2].
When an applied stress on a material in which a crack has length c results in a stress
intensity factor K~ at the crack tip larger than KIc, the material will fracture instantly.
For semi-circular cracks that are small relative to the thickness, Y = 1.24 [2]. In the
cases considered here, the depth of the cracks are less than 2.5 % of the thickness, and
so Y = 1.24 was used. According to Equation (1), for isotropic, monolithic materials,
the fracture strength shows a decrease with increase in c. Many ceramics and glass
ceramics exhibit this strength sensitivity to crack size. However, for many two-phase
materials, the toughness increases with the critical crack size and displays a behavior
termed "R-curve," "T-curve," or resistance curve [3]. This behavior indicates that
the resistance to crack growth increases with increase in the critical crack size.
Several studies have proposed possible mechanisms responsible for R-curve
behavior in brittle materials [4-6]. Phase transformation, residual stress due to
microstructure, elastic grain-bridging, frictional grain-pullout and transition from
relatively large to small grains have been proposed as possible mechanisms for rising
crack resistance. Many investigators have studied this phenomenon using direct
observation of cracks in controlled crack growth studies [4], post-mortem
examination of the fracture surface [5], and other observation-based techniques [7].
Only limited studies have been performed on the study of R-curve behavior by using a
quantitative measure of the microstructure and topography of the material on the
fracture surface.
Fractal geometry is a non-Euclidean geometry that exhibits self-similarity, or
self-affinity, and scale invariance and is characterized by the fractal dimension [8].
The fractal dimension has been used to characterize quantitatively irregular fracture
surfaces and relate these measurements to mechanical properties such as fracture
toughness
Kk = E 80 I/2 D *1/2 (2)
where E is the elastic modulus, a0 is a characteristic size and D* is the fractal

dimensional increment. Note that D* ~: 0 for primary bond fracture [9]. The larger
values of D* correspond to greater degrees of tortuosity, which, in turn, correspond to
greater values of toughness. Hence, the rougher the appearance of the fracture
surface, the tougher the material. In addition, it has been shown that D* can be related
to characteristic fractographic features.
The fracture surface of materials which fail in a brittle manner leaves characteristic
markings on the surface which have been observed for many decades [10]. An
example of the markings on a silicon nitride fracture surface is shown in Figure 1.
From the fracture origin, c, in Equation (1), the crack propagates in a relatively smooth
plane (the mirror region) to the boundary, r, and progressively gets rougher by
deviating slightly out of plane in a region that resembles mist (the mist region),
between r~ and r2. Finally, the crack deviates locally from the main plane of fracture
(the hackle region), between r2 and r3, getting very rough and finally branching into
two or more cracks at r3. Beyond macroscopic crack branching, the process can repeat
itself on each branch of the propagating crack, if there is sufficient energy (Figure 1).
These regions are all related to the applied (far field) stress at fracture, of [9]
O f rj I/2 = constant = KBj / Y'(O) (3)
where rj = rt, r2 or r3 correspond to the different regions in an analogous equation to

Equation (1). KBj is the crack branching stress intensity and Y'(O) is a crack border
correction factor where O is the angle from the surface to the interior of the crack
along the crack front, i.e. O ~ 0 to 90 ~ [11].
MECHOLSKY ET AL. ON QUANTITATIVE FRACTOGRAPHY 155
Figure 1 - Fracture surface of Si3N4 showing mirror, mist and

hackle regions. White arrow indicates origin, black arrows indicate
the mirror-mist transition (1) and the mist-hackle transition (2).
The crack-to-mirror size ratio is related to the fractal dimensional increment

[I21
c/rt = D* (4)
Notice that D* is a constant for a given material. What this implies, and has been
verified experimentally, is that although the regions of mirror, mist and hackle appear
to, and in fact do, have different roughness values, they have the same relative
tortuosity, i.e., same horizontal to vertical change in dimensions.
Two materials will be discussed in this paper to illustrate the use of
quantitative fractography and fractal analysis to study R-curve behavior in brittle
materials: Si3N4 polycrystalline ceramic and 3BaOe5 SiO2 glass ceramic. The
microstructure of the Si3N4 selected for this study consists of rod-like or "whisker-
shaped" grains. Therefore, "R-curve behavior is expected. The parent glass of the
3BaOe5SiO2 glass ceramic can be heat treated to produce various aspect ratios of
grains within the microstructure. Elongated grains of up to 8:1 can be produced.
Because the apparent toughness of "R-curve" materials is dependent on the critical
crack size, it is impossible to use a constant value to describe the material fracture
resistance for material selection. The intent of this paper is to investigate whether the
crack resistance of the material with "R-curve" behavior can be described by a
constant quantity.
Experimental Procedures
Si3N4
The silicon nitride material was manufactured by Toshiba Co. (TSN-03H) and
is widely used in bearing industries. The composition is silicon nitride with primarily
Y203 and A1203 as sintering aids. The geometry of the specimens was provided by the
supplier from bearing material and was 16 mm long, 12 mm wide, and 6 mm thick.
The samples were polished and their surface finish roughness was 1 ~un. A scanning
electron micrograph of the silicon nitride microstructure after plasma etching is shown
in Figure 2. The 13phase grains have a rod-like structure of about 0.5~n in diameter
and are about 2.5 ~m long.
Figure 2 - SEM Image of Plasma Etched Si3Na. View shows both 2-D and 3-
D views of microstructure. Note the elongated grains. White bar = 1 Inn.
The indentation-strength technique was used to test the silicon nitride bars.
The initial flaws were introduced using a Vickers diamond pyramid. Indentation loads
were 0.1, 0.2, 0.5, 1, 2, 5, 7, 24 kg, respectively. Applying multiple indentations
created larger initial flaws in the surface of specimens. A seven-kilogram indentation
load was used to make 5, 7, 9, 12, 15, and 25 impressions in different specimens,
respectively, along the middle line of the tensile surface perpendicular to the tensile
stress direction. The distance between each crack tip was kept within 10 Ima.
Indented samples were then mounted in a three-point fixture whose span distance was
10 mm. Cyclic loading in a form of a half sine wave was applied to the samples. The
cyclic tensile stress ranged from 100 MPa to 0 MPa on the indented surface. Each
indented specimen experienced 300 loading cycles at 1 Hz frequency. The purpose of
this pre-cyclic loading was to release residual stress around the crack tip, which was
induced during indentation, and to link multiple indent cracks into a large single crack.
Fractographic observation showed that the individual cracks from the indents did link
up to form one elongated crack. Afterwards, the specimens were fractured using a
three-point bend test fixture in an Instron test machine. Details of the experimental
setup can be found elsewhere [13].
3BaO.5Si02
Barium silicate glass containing 37.5 mol% BaO and 62.5 mol% SiO2 was
melted at 1550~ for 2 h and poured into graphite molds to form plates 5.1 cm x 12.7
cm x 2.1 cm. The glass was prepared at the National Institute of Standards and
Technology (Gaithersburg, MD). From these plates, bars approximately 24.0 mm
long, 3.5 mm thick, and 6.0 mm wide were cut using a gem-cutting saw (Model AL-
P10S, Raytech Industries, Middlefield, CT). The glass transition temperature range
and exothermic peaks reported by Freiman et al. [14] were confirmed using
differential thermal analysis (DTA). Twenty-four glass bars were nucleated and
crystallized to produce each glass-ceramic microstructure. The treatment times and
temperatures used to produce these microstructures are shown in Table 1.
Aspect ratios greater than 8.1 could not be produced in this system because of
two factors. First, the barium silicate system is extremely temperature and time
sensitive. Any small deviation in temperature causes large changes in microstructure.
Second, crystal growth occurs until the crystals impinge on each other. Heat
treatments leading to further crystal growth produce high internal stresses that are
undesirable. Freiman et al. [15] demonstrated this result by producing specimens with
internal cracking associated with larger crystals.
The glass-ceramic bars were ground and polished following the respective
crystallization heat treatments. The purpose of the final polishing step was to ensure
that the surfaces of the bars were parallel and to reduce the number and severity of
surface defects. The glass-ceramic bars were ground and polished through 600-grit
silicon carbide abrasive. Only the surface that subsequently sustains tension during
flexural testing was polished through 1200-grit silicon carbide and 1 prn alumina
paste. The specimen edges along the tensile surface were then rounded using 1200-
grit silicon carbide abrasive to reduce these sites as stress concentrators. All polishing
was performed using a hand lapper (South Bay Technologies, San Clemente, CA).
The finished flexure bars were cleaned ultrasonically in ethanol for 5 min, dried, and
stored in closed specimen trays until flexure testing. All specimens were annealed at
600~ for 2 h following surface finishing and indentation to remove residual stresses.
The flexure bars were fractured after indentation at 2.94N, 4.9N and 9.8N,
respectively. More detail of the experimental arrangement can be found in the
literature [16].
Table 1 - Heat-treatment schedules for desired aspect ratios in 3BaOo5SiO2 glass
ceramics.
Crystal
Nucleation Crystallization Size Aspect
Temp (~ Time (h) Temp (~ Time (h) Range (~m) Ratio
675 24 775 1.5 2-5 1.3
700 0.75 825 1.5 1.5 - 11 3.6
700 1 825/1050 24/24 0.5 - 15 8.1


We present two examples that illustrate the application of fractographic
principles in analyzing R-curve behavior. The first will discuss R-curve behavior in
silicon nitride material used in bearing applications. The second will discuss the effect
of microstructure on R-curve behavior in baria-silica glass ceramics. These glass
ceramics have been shown to be an excellent model material for analyzing particulate
composites [16].
Figure 3 - Fracture toughness as a function of critical crack size for TSN-O3H

Silicon Nitride.
Si3N4
Toughness values were obtained from crack size measurements on the fracture
surface applying Equation (l). The fracture toughness, in terms of Kic, versus the
crack size obtained from the fracture surface of the flexure bars is shown in Figure 3.
Notice that this silicon nitride exhibits R-curve behavior, as expected.
Several studies [4-7] have proposed possible mechanisms responsible for R-
curve behavior in brittle materials. The R-curve behavior in silicon nitride is usually
attributed mostly to grain bridging. Silicon nitride bearing material consists of rod-like
or "whisker-shaped" grains. These grains grow in randomly oriented directions and
are interwoven with one another. This interweaving results in an additional

mechanical interlock force causing them to remain in close contact. Most of the
fracture in silicon nitride is through the grain boundaries. When the cracks start to
propagate through the grain boundaries, the crack immediately faces an interlock force
among the grains, which resists in the grains pulling apart during crack opening. This
resistant force is also called a crack bridging force.
In order for bridging to occur, a growing crack propagates locally mostly in the
grain boundaries, avoiding the grains and leaving the grains unbroken, hence bridging
the macroscopic crack. The bridging changes the strain energy release rate AG by
exerting a closure (bridging) stress on the crack surfaces.
From Figure 3, we see that Klc stays at about 3-4 MPam I/2 for a crack size of c =
6 om until the crack gets larger than about 30 ~ n . The manufacturer reports a K~c
value of 6 MPam t/2 [13]. The maximum bridging stress can be estimated from
Equation (1) and the load cell for all cracks greater than 30 microns using Kic = 3.4
MPam 1/2 and Y=1.24. The difference between the stress calculated from Equation (1)
and the stress recorded from the load cell provides an estimate of the bridging stress at
~ 155 MPa [13], in reasonable agreement with the estimates of Li [18] of 105 MPa.
For c = 30 prn to 335 ~rn, the Kk rises at a very steep rate to reach a value of about 7
MParn 1/2. From then on, Kic stays at about 7 MPam l/z for a wide range of c. The
values of toughness calculated here are in general agreement with literature values
[13,18,19]. The R-curve behavior also indicates that the grain bridging zone size is
about 335 ~rn. This is because when the crack is less than 30 pro, less than 10% of
the grain bridging effective zone acts as a closure stress on the crack surface and no
significant bridging occurs to increase K~c. After the crack increases to greater than
335 ~un, which is the maximum bridging zone size, Kxc reaches its maximum. Further
increase in the crack size does not result in an increase in a bridging force, and hence
there is also no increase in Kic.
Fracture surface analysis can be used to study the relationship between crack
size, crack branching size, and fracture energy. In this study, we found that the ratio
of the critical crack size c and mirror boundary size, r, increases with an increase in c
(Figure 4). For isotropic materials, e.g., silicate glasses, the values of c/r would be
expected to remain constant as a function of crack size [2,10]. This is because Kk is a
constant for glasses and most monolithic ceramics. The present results show that the
c/r ratio varies as a function of the crack size because the toughness value increases
with crack size.
~O
0.3
0.25
0.2
0.15
0.1
0.05
//
0
200 400 600 800
C r a c k S i z e (c) M i c r o n s
Figure 4 - Crack size-to-mirror size ratio as a function of crack size.

Note the R-curve behavior that mimics the fracture toughness data in
Figure 3. D* should be proportional to c/r as suggested by Equation (4).
Previously [13, 17], it was shown that there is an elastic grain bridging stress
on the crack surfaces when a crack propagates in silicon nitride. This bridging stress
balances part of the applied stress at the crack tip. Extra stress is needed to reach the
critical stress intensity factor K1c, at which the crack propagation starts. Therefore, a
longer crack has a higher bridging force in suppressing crack opening and hence a
higher critical strain energy is needed to drive crack propagation. The grain bridging,
therefore, results in K~c as a function of the initial crack size c as shown in Equation
(1). The stress intensity criterion for crack branching, KB, is given by Equation (3).
The mirror boundary is the point at which the crack front at some locations starts to
deviate visibly from its apparent original crack plane and create more than one
macroscopic crack path. Many criteria were developed to explain the cause of crack
branching: (1) a velocity criterion [20] in which branching occurs when the crack
velocity reaches a maximum value (close to the sound velocity); (2) a critical stress
intensity factor criterion [21,22] in which branching occurs when the stress intensity
factor reaches a critical value; (3) a strain intensity criterion [23] requiring that the
strain energy available at the crack tip must be sufficient to form two cracks; and (4)
an energy criterion [24]. All these criteria have different explanations but they all
indicate that crack branching needs a critical energy level to occur. The branching
starts at the maximum strain energy release rate or at the maximum crack velocity.
Since the stress wave velocity is a constant for an identical material, the crack velocity
in the material is dependent on the material and the stress level. Therefore, KB should
j i
be a material constant based on Equation 3. The curve of normalized KB ( K a / K B)
versus mirror size r is plotted in Figure 4. K~ is one of 15 crack branching stress
intensity constant measurements obtained from 15 different fracture surfaces, where
the superscript j is equal to 1, 2, etc. up to 15. K~ is one of the K~ values measured

from a pair of of I and r 1 values (Equation 3). The curve in Figure 5 shows a straight
line with essentially a zero slope. The result indicates that K~is a material constant
regardless of the initial crack size. From Equations 1 and 3, we have
c
-=
rr, K: = y rcO:C
2 2
(5)
~" 2 rc 2
r --B KB
where Yrc=Yr/Yc is a constant (Yr is about 1.24 when r is measured along the tensile
surface for small cracks). Since KB is a constant for the material, the ratio of c over r
increases with c. The c/r ratio totally depends on KIc, which is a function of the initial
crack size, c. Thus, we can use quantitative fractography to determine the local
influence of the microstructure on the toughness values.
3BaO*5Si02
At least 80% of the increased toughening for glass-ceramic microstructures over the
glass structure in baria-silicate glass-ceramics was previously attributed to crack
deflection [19]. These morphological changes lead to a more tortuous path for a
propagating crack to travel than in the glass. The crack deflection mechanism was
also observed in Si3N4 polycrystalline ceramics with needle-like grains, [13,17,19,25]
and canasite glass-ceramics with tubular grains [26].
1.6
1.2
N '%*§ § 9
0.8
0.4
I I I I I I I I
soo lOOO 15o0 2000 2500 3000 3500 4000
2r (pm)
Figure 5 - Critical stress intensity at branching as a function o f the

mirror diameter, The data are essentially independent o f r. The solid
(horizontal) line is placed as an aid to the eye.
Another toughening mechanism shown in materials with high crystal aspect

ratios is crack bridging. Evidence of this mechanism is suggested in specimens with
an aspect ratio of 8.1. Two indications of the presence of this mechanism observed are
micro-branching into the fracture surface, resulting in multiple crack paths and
propagation around a crystal without fracture of the crystal [16]. There were no
similar observations for the glass, or glass-ceramics with AR values of 1.4 or 3.6.
Fracture Toughness - Most of the mean fracture toughness values correlate
well with the theoretical predictions made by Faber and Evans [27] for crack
deflection of disc shape particles (Fig. 6). The values of fracture toughness for the
materials with a lower aspect ratio are equivalent to the predicted toughness values for
discs with aspect ratios of 1.4 and 3.6. Only the value for an aspect ratio of 8.1 is
much higher than the predicted value and is equivalent to the toughness for a disc with
an aspect ratio of 12. This increase may be explained by the presence of another
toughening mechanism occurring in the glass-ceramic, such as crack bridging.
Fractography in the form of fractal analysis can provide a means by which we can
distinguish between the contributions of one toughening mechanism and another. In
the next section, we describe the use of the fractal dimension in analyzing fracture
behavior. The fractal dimension is a quantitative measure of the tortuosity of a
fracture surface.
Fractal Dimensional Increment - The fractal dimension is a number such as
2.D* where 2 is the topological dimension and D* is the fractal dimensional increment
which provides the surface structural information. If D* is small, say, 0.06, the
surface is relatively smooth. If D* is large, say, 0.35, then the surface is relatively
tortuous. The relationship between fractal dimensional increment and fracture
toughness has been noted for several brittle materials such as the lithia disilicate glass-
ceramics [28-30], canasite glass-ceramics [26], a tetrasilicate fluoromica glass-ceramic
[31], brittle intermetallics [32], and various other ceramics [33,34]. These data along
with the present data for barium silicate are plotted in Figure 6. Each line in the plot
of Kc versus D* represents a specific class of material in which fracture toughness
increases as the fractal dimensional increment increases. These classes tend to fall
within a pattern according to their type of microstructure as indicated in Figure 6.
Note that the line for glass-ceramics includes Ocala Chert, Dicor| glass-ceramic,
lithia disilicate glass-ceramic, silica glass, barium-silicate glass, and barium-silicate
glass-ceramics [29, 33, 34].
Equation (2) is graphed in Figure 7 with representative data, including the
3BaOo5SiO2 glass ceramic with differing aspect ratios. [Note that the solid lines in
Figure 7 extended to D* = 0 pass very near the origin, indicating that an intercept
value of Ktc is not needed in Equation (2)] Note also that the data point for an aspect
ratio of 8.1 deviates significantly from the expected line for glass-ceramics. This
deviation can be either a true deviation or merely another point on a linear plot for all
of the present data, i.e., glass, and glass-ceramics with AR values of 1.4, 3.6, and 8.1.
If we consider the latter first, we see that a linear relationship would be interpreted as
having a Klc value which is negative at D* = 0. This negative intercept has no
physical meaning and is considered unreasonable. However, if the three data points,
i.e., glass andglass-ceramics with aspect ratios of 1.4 and 3.6 are considered part of the
glass ceramic data shown in Figure 7, then the value of Klc= 0 corresponds to a value
of D* = 0. Although we do not expect the data to extend to K~c = 0, this value for the
intercept is more reasonable than a non-zero intercept and can be explained as
discussed above. Using the latter interpretation of the data, we conclude that the data
point for an AR of 8.1 deviates from the expected trend for other glass ceramics. One
of the reasonable interpretations of this deviation is that it is caused by a change in the
fracture process associated with the toughening mechanism. The change from
primarily crack deflection (intergranular) to a mixture of crack deflection and another
mechanism, probably crack bridging (transgranular, pullout and then fracture of
elongated grains in the wake of the crack front) is consistent with the observed data.
The amount of vertical deviation from the line could be an indication of the additional
toughening of approximately 0.7 MPa,m 1/2 contributed from the other mechanism.
Rod (AR = 12)
AR =8.1
Disc (AR = 12)
J
(AR ~ 3)
" - "~ - , ' , "~ - ' - ~ AR = 3.6
-- Disc(AR = 3)
2
AR= 1.4
1 I I I i Glass
0.0 0.1 0.2. 0.3 0.4
Volume fraction of ~k-~'ticle.s
Figure 6 - Relative toughness as a function of volume

fraction of reinforcing particles or grains.
Analysis of Combined Data - There are several independent aspects of the data
reported that strongly suggest a change in the fracture process from the materials with
the lower two aspect ratios to the material with an AR value of 8.1. (i) Fracture
surface analysis shows a change from intergranular fracture for groups with aspect
ratios of 1.4 and 3.6 compared to transgranular fracture for the group with an aspect
ratio of 8.1. (ii) SEM images reveal examples of crack bridging in the group with an
aspect ratio of 8.1 while no evidence was observed on any of the fracture surfaces of
groups with aspect ratios of 1.4 or 3.6. (iii) All materials appear to exhibit fracture as
predicted by the Faber and Evans model [19,27] (Figure 6) except for the specimens
with an aspect ratio of 8.1, whose mean toughness was greater than would be
predicted by crack deflection alone. (iv) The strength, fracture toughness, and fractal
dimension for the glass-ceramic with an aspect ratio of 8.1 were statistically
significantly greater (p < 0.05) compared with corresponding values for the other
aspect ratio specimens. (v) The deviation of the value of fracture toughness versus the
square root of the fractal dimensional increment (~ 0.7 MPaom 1/2) in Figure 7 is
consistent with a change in the surface topography observed, i.e., the change from
primarily intergranular for an AR value of 1.4 and 3.6, to primarily transgranular for
AR = 8.1 (~ 0.4 MPa~
K ic
(MPa-m 112)
0 I I I
0.2 0.3 0.4 0.5 0.6
(D*) 1/2
Figure 7 - Fracture toughness as a function of the fractal dimensional

increment for brittle materials. The 3BaOe5Si02 glass ceramics are
included to demonstrate the effect of R-curve behavior.
Although any one of the above observations would not be sufficient to suggest an
additional toughening mechanism for an AR value of 8.1 material, the five
observations cited taken together are consistent in suggesting that the change in
behavior is caused by an additional mechanism. Further, the change in morphology
from intergranular to transgranular, the observation of elongated grains that have
bridged the crack, the agreement of data for groups with AR values of 1.4 and 3.6 with
the crack deflection model but the poor agreement of the data points for the group with
an AR value of 8.1 with the crack deflection model and the deviation of this point
from the KIC versus D *1/2 line, are all consistent in suggesting that the additional
toughening mechanism is crack bridging. If this interpretation is correct, then the
results suggest that the fractal method can be used to quantify the amount of
toughening caused by bridging. Obviously, additional research is needed to determine
if this theory applies for other materials. We suggest that the use of fractal analysis
represents a new approach for quantitative fractography which allows for the
distinguishing between multiple mechanisms.
Conclusions
Quantitative fractography can be used to interpret fracture behavior in
materials that exhibit R-curve behavior. Two examples with rod-like grain structure
were described in this paper: SiaN4 and 3BaOo5SiO2.
TSN-03H Si3N4 has a rising resistance to crack growth (R-curve behavior) due
to crack bridging. However, the R-curve behavior only has an effect when the crack
size, c, is greater than 30 Inn. For small cracks, which are usually found in bearing
elements, the R-curve behavior has little effect. Therefore, a high aspect ratio of rod-
like grain structure may result in a decrease in material strength when small defects or
flaws are present in the material due to localized residual stresses and thermal
expansion anisotropy. The present research questions whether it is recommended to
develop material with R-curve behavior to be used in bearing applications as a trade-
off for material homogeneity.
As anticipated, the fracture toughness of 3BaOo5SiO2 glass-ceramics increased
with increasing aspect ratio from 0.7 + 0.1 MPa~ 1/2 for glass to 2.2 + 0.6 MPa~ 1/2
for the glass-ceramic with an aspect ratio (AR) of 8.1. The fractal dimensional
increment, D*, also increased with crystal aspect ratio. The fracture surface
morphology changed from intergranular for glass-ceramics with AR values of 1.4 and
3.6 to transgranular for a glass-ceramic with an AR value of 8.1. The toughness
versus fractal dimensional increment data were interpreted to provide a method for
calculating the amount of increased toughening (0.4-0.7 MPa,m I/2) compared with 9
that associated with crack deflection alone. This additional toughening increment can
be explained, for the group with an AR value of 8.1, as most likely due to crack
bridging.
Acknowledgments
The authors thank Dr. S. W. Freiman of NIST for providing 3BaO*5SiO2 glass
and for useful discussions. We also thank Dr. Herb Chin of Pratt and Whitney for
much help in understanding the role of materials in bearing systems. Part of the silicon
nitride research was performed under contract and we wish to acknowledge the
support of DARPA and the USAF under Air Force Contract # F33615-96-2-5263.
References
[1] Hertzberg, R.W., "Deformation and Fracture Mechanics of Engineering

Materials," Wiley, New York, 1989.
[2] Mecholsky, J. J., Jr., Freiman, S.W. and Rice, R. W., "Effect of Grinding on Flaw
Geometry and Fracture of Glass," J. American Ceramic Society., Vol. 60,
1977, pp. 114-117.
[3] Lawn, B. R., "Fracture of Brittle Solids," Cambridge Solid State Science Series,
1996.
[4] McMeeking, R., and Evans, A. G., J. American Ceramic Society, Vol. 65, No.5,
1982 pp.242-46.
[5] Evans, A. G., Faber, K., J. American Ceramic Society, Vol. 67, No. 4, 1984,
255-60.
[6] Mai, Y., and Lawn, B. R., J. American Ceramic Society, Vol.70, No. 4,1987,pp.
289-94.
[7] Okada, A., and Hirosaki, N., J. American Ceramic Society, Vol. 73, No.7,1990, pp.
2095-96.
[8] Russ, J.C., Fractal Surfaces, Plenum Press, New York, 1994.
[9] Mecholsky, J. J., "Fractography, Fracture Mechanics and Fractal Geometry: An
Integration," In: Advances in Ceramics, J.C. Varner and J.D. Frechette, editors,
The American Ceramic Society, Westerville, Ohio (1996).
[10] Mecholsky, J. J., Jr., "Quantitative Fractography: An Assessment." In: Ceramic
Transactions, Vol. 17, Fractography of Glasses and Ceramics 1. Frechette VD,
Varner J, editors. Westerville: American Ceramic Society, 1991 ,pp. 413-451.
[11] Newman, J. C., and Raju, I. S., "Analysis of Surface Cracks in Finite Plates
Under Tension and Bending Loads," NASA Technical Report 1578, 1979.
[12] Mecholsky, J. J., Jr., and Freiman, S. W., J. American Ceramic Society, Voi. 74,
No.12, 1991, p. 3136.
[13] Chen, Zheng, Wear Mechanisms of Silicon Nitride Bearing Materials Under
Contact Fatigue Stresses, Ph.D. Dissertation, University of Florida, 1995.
[14] Freiman, S.W., Onoda, G.Y., Pincus, A.G., "Controlled Spherulitic
Crystallization in 3BaO-5SiO2 Glass" J. American Ceramic Society, Vol. 55,
1972, pp. 354-359.
[15] Freiman, S.W., Onoda, G.Y., Pincus, A.G., "Mechanical Properties of 3BaO-
5SIO2," J. American Ceramic Society, Vol. 57, No. 1, 1974, pp. 8-12.
[16] Hill, T. J., Mecholsky, J. J., Jr. and Anusavice, K. J., "Fractal Analysis of
Toughening Behavior in 3BaOe5SiO2 Glass Ceramics", J. American Ceramic

Society, Vol. 83, No.3, 2000.
[17] Chen, Zheng, Cuneo, J., Mecholsky, J. J., Jr., and Hu, S., "Damage Process of
Si3N4Bearing Material Under Contact Loading," Wear, Vol. 198, 1996, pp.
197-207.
[18] Li, C., Lee, D., and Lui, S., J. American Ceramic Society, Vol. 75, No.7, 1992,
pp.177-85.
[19] Faber, K.T., Evans, A.G., "Crack Deflection Processes - II. Experiment," Acta
Metallurgica, Vol. 31,1983, pp. 577-584.
[20] Joffe, E., Philosophical Magazine 112, 1951, pp. 739-50.
[21] Clark, A. and Irwin, G. R., Experimental Mechanics 10, 1978, pp. 321-30.
[22] Kalthoff, J., "On the propagation direction of bifurcated cracks", pp. 449-58 in
Dynamic Crack Propagation, ed. G. Sih, Noordhoff International Publishing.
[23] Kirchner, H. P., Engineering Fracture Mechanics 10, 1978, pp. 283-88.
[24] Y. Tsai, Experimental and Molecular Dynamics Determination of Fractal
Fracture in Single Crystal Silicon, Ph.D. Thesis, The University of Florida,
1993.
[25] Chen, Zheng, Mecholsky, J.J., Jr, Joseph, T., Beatty, C.L., "The Fractal
Geometry of Si3N 4 Wear and Fracture Surfaces," Journal of Materials
Science, Vol. 32, 1997, pp. 6317-23.
[26] Chart, J.Y., Mechanical Degradation of Canasite Glass-Ceramic Under Dynamic
and Cyclic Fatigue; Master's Thesis, University of Florida, Gainesville,
Florida, 1996.
[27] Faber, K.T. and Evans, A.G., "Crack Deflection Processes - I. Theory," Acta
Metallurgica., Vol. 31,1983, pp. 565-576.
[28] Naman, A. Mechanical Property and Fractal Dimension Determination in
Li20.2Si02 Glass-Ceramics; Master's Thesis.University of Florida,
Gainesville, Florida, 1994.
[29] Thompson, J.Y., Characterization of Mechanical Behavior Based on
Microstructural Phenomena in Li20-SiO 2 Glass-Ceramics; PhD Thesis.
University of Florida, Gainesville, Florida, 1995.
[30] Thompson, J. Y., Mecholsky, J. J., Anusavice, K. J., Balasubramanian, K. J.,
"The Effect of Microcracking On The Fracture Toughness and Fractal
Dimension of Lithia-based Glass-Ceramics," J. American Ceramic Society,
Vol. 78, 1995, pp. 3045-49.
[31 ] Thompson, J.Y., Cuneo, J.C., Naman, A., Mecholsky, J.J., Anusavice, K.J.,
"Relationship Between Fracture Toughness and Fractal Dimension of a Dental
Glass-Ceramic," Transactions of the Academy of Dental Materials, Vol. 7,
Abstract P8, 1994, p. 87.
[32] Fahmy,Y., Russ, J.C., Koch, C.C. "Application of Fractal Geometry
Measurements to the Evolution of Fracture Toughness of Brittle
Intermetallics," Journal of Materials Science, Vol. 6, 1991, pp. 1856-1861.
[33] Mecholsky, J.J., Passoja,D,E., Feinberg-Ringel, K.C., "Quantitative Analysis of
Brittle Fracture Surfaces Using Fractal Geometry," J. American Ceramic
Society, Vol. 72, 1989, pp. 60-65.
[34] Mecholsky, J.J. Mackin T.J. and Passoja, D.E., Self-Similar Crack Propagation
in Brittle Materials, Advances in Ceramics, 22; Fractography of Glasses and
Ceramics, The American Ceramic Society, 1988.
Michael R. Hill, I R. Doug Carpenter, 1 Glaucio H. Paulino,2 Zuhair A. Munir, t and
Jeffery C. Gibeling I
Fracture Testing of a Layered Functionally Graded Material
Reference: Hill, M. R., Carpenter, R. D., Paulino, G. H., Munir, Z. A., and
Gibeling, J. C., "Fracture Testing of a Layered Functionally Graded Material,"
1409, J. A. Salem, G. D; Quinn, and M. G. Jenkins, Eds., American Society for Testing
Abstract: This paper describes measurements of the Ki R-curve for a layered ceramic-
metallic functionally graded material (FGM) composed of Ti and TiB phases. Single edge
notch bend specimens were fabricated for crack propagation perpendicular to the graded
layers from the brittle to the ductile side of the FGM. The precracking method and
residual stresses affected the measured toughness. A new reverse bending method
produced a sharp precrack without damaging the material. A representative sample
indicates R-curve behavior rising from K,~,~ = 7 MPa.m I/2 in the initial crack-tip location
(38% Ti and 62% TiB) to 31 MPa.m 1/2 in pure Ti. Residual stresses acting normal to the
crack-plane were measured in the parent plate using the crack compliance technique. The
residual stress intensity factor in the SE(B) specimen determined from the weight
function was -2.4 MPa.m In at the initial precrack length. Thus, the actual material
toughness at the start of the test was 34% less than the measured value.
Keywords: Functionally graded material, residual stress, fracture toughness, R-curve
Introduction
A functionally graded material (FGM) is a composite containing at least two distinct

phases that vary in relative proportions along at least one dimension of the solid. The
classic example is that of a metallic/ceramic FGM that is predominantly metallic on one
side of a plate and continuously becomes more ceramic through the thickness. The
primary goal in designing such materials is to take advantage of the desirable properties
1 Assistant Professor, Department of Mechanical and Aeronautical Engineering;

Graduate student, Professor and Professor, respectively, Department of Chemical
Engineering and Materials Science, University of California, One Shields Avenue,
Davis, CA 95616.
2 Assistant Professor, Department of Civil and Environmental Engineering,
University of Illinois, Urbana IL 61801-2352.
169
Copyright9 ASTMInternational www.astm.org
of the ceramic, such as high hardness, corrosion resistance, and high melting temperature,
while retaining the benefits of the metallic phase including good fracture toughness,
machinability, and weldability. When discrete layers of metal and ceramic are used, the
difference in coefficients of thermal expansion can lead to high internal stresses. Thus,
another advantage of a graded composition lies in avoiding problems associated with
high thermal and residual stresses. To achieve these design goals, it is necessary to
understand and control the variation in properties and the gradient in composition.
FGMs have been produced by a variety of methods, including chemical vapor
deposition [I], powdered metal sintering [2], high-temperature plasma spraying [3], self-
propagating combustion synthesis (SHS) [4, 5], thermo-chemical diffusion, and
sedimentation of slurried powders [6]. As a practical matter, synthesis of materials with a
continuous variation in the relative volume fractions of the two constituents may be
difficult. Thus, numerous investigators have created discontinuous or layered FGMs.
When the material is cooled from the synthesis temperature, residual stresses generally
develop at the interfaces between layers. Although the residual stresses can be minimized
by varying the composition in smaller steps, their presence can nonetheless limit the
utility of a FGM.
As noted above, one motivation for developing functionally graded materials is to
avoid problems associated with the low toughness of many ceramics. Thus, fracture
toughness is a limiting property in the development of FGMs. For that reason, a number
of authors have presented mathematical treatments of the fracture resistance of
functionally graded materials [7-9]. These studies have shown that the crack tip fields for
general non-homogeneous materials are identical to those in a homogenous material
provided the properties of the former are continuous and piece-wise continuously
differentiable [8]. Under these conditions it is possible to extend the concept of the stress
intensity factor to study fracture and fracture toughness of FGMs. Jin and Batra [9, 10]
have presented a theoretical prediction of rising R-curve behavior in a metallic/ceramic
FGM based on specimen size, loading conditions, and metal particle size. They used a
simple linear rule of mixtures for Young's modulus, shear modulus, and Poisson's ratios
and noted that "the micro-mechanical models developed for macro-homogeneous
composites are only approximately valid for FGMs."
In contrast, there is a relative scarcity of experimental fracture studies on FGMs due
primarily to the lack of materials of size sufficient to produce standard fracture testing
specimens. However, recent advances in processing methods have resulted in the
production of larger FGMs. For the present study, we acquired a Ti/TiB FGM with
dimensions 150mm x 150mm and 16mm thick from Cercom, Inc. The dimensions of this
plate are sufficient to permit single edge notch beams to be machined for standard
fracture experiments.
Thus, the goal of the present paper is to provide experimental measurements of the
fracture toughness of functionally graded materials and to extend the concepts embodied
in fracture mechanics testing standards to this new class of materials. We address
problems associated with precracking, crack length measurement, and data reduction. We
demonstrate that rising R-curve behavior is observed in a layered Ti/TiB FGM when the
crack propagates from the more brittle side to the more ductile composition. In addition,
we quantitatively account for residual stresses and show that they can have a significant
effect on the measured fracture toughness values.
HILL ET AL. ON LAYERED FUNCTIONALLYGRADED MATERIAL 171
Material and Experimental Procedure
Material
The material tested in this study was a layered, functionally graded titanium/titanium
monoboride plate. This material was prepared by Cercom, Inc. using a commercially pure
titanium metal base. Tape cast layers composed of various mixtures of titanium and
titanium diboride powders were placed on top of this plate. The assembled laminate was
hot pressed at 1578K at a pressure of 13.8 MPa. In order to facilitate densification at this
temperature, a proprietary sintering aid containing nickel was added to the starting
powders. This material created a liquid phase at 1215K that also catalyzed the reaction of
titanium and TiB2 to form TiB with virtually no residual TiB2. The resulting FGM was
composed of seven layers ranging from pure Ti on one side to 85 volume% TiB on the
other. The compositions of each of the layers of this FGM are listed in Table 1.
Table 1 - Composition and elastic properties o f the Ti/TiB FGM
Layer # vol% Ti vol% TiB d (mm)* E (GPa) v
1 15 85 2.5 312.67 0.14

2 21 79 1.7 303.37 0.152
3 38 62 1.8 289.38 -
4 53 47 1.4 (262.64) t
5 68 32 1.8 227.37 0.238
6 85 15 2.1 170.66 0.278
7 100 0 3.4 106.87 0.34
* Approximate layer thickness

t Estimated by interpolation
An example of the resulting microstructures is shown in Figure 1, which is a

backscatter scanning electron microscope image illustrating the interface between Layers
5 (32 volume% TiB) and 6 (15 volume% TiB). The excellent bonding of the layers and
lack of a distinct interface are apparent in this figure. In addition, the two distinct
morphologies of the TiB are shown. The blocky and needle shaped TiB particles (darkest
phases) are similar to those previously observed by Tsang, et al. [11]. A semi-continuous
grain boundary network of the NiTi reaction product (lightest phase) from the sintering
aid can also be observed in these two layers.
Elastic modulus and Poisson's ratio data were measured by slicing thin beams from
the FGM. Specifically, bilayer beams were made from Layers 1 and 2 and from Layers 6
and 7. Trilayer beams were cut from Layers 3, 4 and 5. Longitudinal and transverse strain
gages were bonded to one side of each beam. These beams were loaded in four-point
bending on one side, then flipped over, re-gaged and again loaded. The modulus and
Poisson values calculated from these bending experiments using composite beam theory
are listed above (Table 1).
Figure 1 - Scanning electron microscope image of the layer 5

(32% TiB) and 6 (62% TiB) interface.
Residual Stress Measurements
Residual stresses were measured in the FGM in order to account for their influence
on measured toughness. Residual stresses were found in 6.4 mm wide samples removed
from the full thickness of the parent FGM plate. The measurements followed the general
procedures of the compliance method [12], where strains are measured as a slot is
incrementally cut through the thickness of the sample. Residual stress normal to the slot
is found as a function of slot-depth, computed from the strain versus slot-depth
measurements and based on elastic stress release. The calculation scheme employs elastic
finite element solutions that account for the variation in material properties present in the
FGM.
Slotting proceeded from the TiB-rich side of the FGM. Two strain gages were used,
as shown in Figure 2, one located next to the start of the slot, and one opposite this
location, but centered on the slot-plane. Coordinates used are also shown in the figure,
indicating that x corresponds to the slotting direction. Cutting was performed using
electric-discharge machining (EDM) to ensure that as little material was disturbed a
possible. Slotting was performed in increments of 0.127 mm (0.005 inch) for the first six
increments, and then in increments of 0.508 mm (0.020 inch) for the remainder.
Stresses were computed by representing the unknown residual stress in the Legendre
polynomial basis, and using released strain to find coefficients of the basis functions. It
was assumed that the variation in material properties in the FGM is sufficiently smooth
not to produce large stress discontinuities at the layer interfaces, so that residual stresses
can be expressed as a polynomial in the thickness coordinate. For through-thickness
residual stress measurements, the constant and linear terms in the polynomial series are
not included because they do not satisfy equilibrium. Higher order Legendre polynomial
terms do satisfy equilibrium. Using these assumptions, the unknown residual stress,
crRs(x), was written as a sum of Legendre polynomial terms, Pi(x), each with a
corresponding amplitude, Ai
HILL ET AL. ON LAYEREDFUNCTIONALLYGRADEDMATERIAL 173
O'Rs(X)= ~ A~PI(x) (1)

i=2,rtl
In this equation, m represents the order of the highest term in the series. A finite element
model was constructed to relate the stress given by a particular basis function, with unit
amplitude, to strain release at the location of the strain gages. For a particular basis
function, Pj(x), and slot depth, ai, the finite element model was used to find the
corresponding strain
Co - e(ai)l,,=pj(.) (2)
Repeating this analysis for all members of the basis functions and all slot depths resulted
in a linear system relating basis function amplitudes to strain as a function of slot depth
E(a,) = ~,~CoA ~ (3)
j=2,m
or, using matrix notation
e_= CA. (4)
Given this system, and strains measured experimentally during cutting, the amplitudes of
the stress expansion are found by inversion of Equation 4 in a least squares sense
A = (C" C)' (5)
Once these amplitudes are found, stress existing in the FGM prior to slotting is obtained
from Equation 1.
Finite element calculations were carried out using a commercial finite element
package [13]. All computations were preformed assuming plane strain. The model had
2620 quadratic (serendipity family) elements and 8163 nodes and exploited the symmetry
of the residual stress state about the crack plane, therefore assuming the shear stress on
the slotting plane was negligible. The mesh was sufficiently refined that doubling the
mesh density in both directions had a negligible effect on the strain versus depth profiles
in Equation 2. Material variations were handled by specifying spatially-dependent
material constants at the integration points of the elements, as discussed in [14]. Strains
were computed by differencing nodal displacements at the boundaries of the gage and
dividing by the undeformed length, and therefore account for strain gage averaging.
Data reduction and error analysis were performed using available software. Matrix
computations were performed using a commercial package [15]. Custom software was
developed to route the finite element output to the input of the matrix processor. Stresses
were computed for several assumed orders of the polynomial series for stress (m in
Equation 1). The optimal order of series expansion was selected by plotting the root mean
square of the error between measured strain and the fitted strain of Equation 4. Low-order
polynomials generally produce high error due to lack of fit, and increasing order
kl- 2.4 mm
Y~ + ~ I" (0.096in)
=17 mm TiB rich
I (0.66 in) CP Ti F-- t.., ~ centerof

x~,a" gage
Figure 2- Typical gage location for slotting.

decreases error only for a limited number of terms. The numerical condition of the
inverted matrix in Equation 5 also becomes poor with increasing number of terms, so that
experimental errors can have significant impact on residual stress measurements when an
unduly large number of terms are taken in the stress expansion. Therefore, an optimal
number of terms must be selected to ensure a good fit to the measured strain while not
amplifying experimental errors. In this work, the optimal order of polynomial series was
determined by plotting the root mean square error versus order number, and selecting the
order at which the error reached a plateau. This has been shown to be a reliable method of
selecting the polynomial order in Equation 1 [16]. Once the order is selected, error
bounds were found for the residual stresses using Monte Carlo. Both slot depth and
measured strain were assumed to be normally distributed random variables, with mean
equal to the measured value and standard deviation of 0.001 inch for slot-depth and 3 Ixe
for strain.
The effect of residual stress on measured toughness was found by superposition.
Residual stresses were measured in samples of larger dimension those of the SE(B)
samples used for fracture testing. Residual stress redistribution resulting from removal of
the SE(B) specimen from the larger FGM was found by finite element modeling. The
analysis involved distributing an initial stress field, corresponding to the measured
residual stress, within a model of the SE(B) geometry, and finding the new equilibrium
stress state accounting for the removed material. The contribution of the redistributed
residual stress field to the total stress intensity factor during fracture testing was found
using a weight function for a homogeneous single edge cracked panel [17]. The total
stress intensity factor at any point during fracture testing was computed by adding the
contributions from residual stress and applied loading. Since the analysis assumes linear
elastic behavior, the residual stress correction is valid when the crack-tip is in a brittle
region.
Fracture Testing
Fracture testing was conducted using single-edged-notched bend specimens (SE(B))

cut from the FGM using EDM. This material was expected to exhibit rising R-curve
behavior during propagation of a crack from the brittle side (containing 85 volume% TiB)
toward the ductile titanium layer. For that reason, the specimens were oriented with the
machined notch on the side containing the highest concentration of TiB. The machined
specimens had a width of 14.73 mm, a thickness of 7.37 mm and a length of 79.50 mm,
with a notch 5.08 mm deep (initially ao/W = 0.345) with integral clip gage knife-edges in
accordance with ASTM Test Method for Measurement of Fracture Toughness (E 1820-
96). Prior to fracture testing, one face of each specimen was carefully polished to a mirror
finish using diamond and silicon carbide media. The polished surface facilitated
measurement of the crack length during crack growth using a traveling microscope. The
precracking and fracture experiments were conducted at room temperature on a
servohydraulic load frame under computer control. Crack mouth opening displacements
were measured using a cantilever type clip gage, as suggested in E 1820-96.
It is well-established that fracture toughness testing requires the generation of a
short, sharp precrack at the root of the machined notch. Although the procedures
appropriate for metals are well-accepted and a number of methods have proven
successful for brittle ceramics [18-21], there are no established precracking methods for
functionally graded materials. In addition to the usual difficulties of controlling the crack
length in a relatively brittle material, the varying properties of the FGM present unique
challenges in developing a suitable precracking technique.
As a starting point, precracking was attempted using the method described in E 1820-
96 by loading the beams in three-point bending under load control. The initial maximum
load was 490 N, the load ratio was 0.1 (minimum/maximum load) and the loading span
was 63.5 mm. Crack growth was monitored by crack mouth opening displacement
(CMOD) compliance. After 75 000 cycles, no change in compliance had been detected, so
the maximum load was increased to 540 N. At this load level the crack popped-in to a
length of 6.29 mm (a/W = 0.427), as detected by CMOD compliance based on the initial
effective modulus. This crack was longer than the desired starting crack length of 5.28
mm (a/W = 0.358) and prevented the fracture toughness testing of two layers in the FGM
specimen. Further, a similar approach with another specimen resulted in unstable crack
growth to a total length of 14.65 mm (a/W = 0.995). Thus, it was concluded that this
method did not provide sufficient control to reliably generate sharp precracks in this
relatively brittle material.
Suresh and Brockenbrough have shown that axial compression of ceramic materials
leads to precrack initiation by residual tensile stresses during compressive unloading [18].
This method was further developed by Ewart and Suresh [19]. The layer of the Ti/TiB
FGM at the root of the machined notch is brittle, hence it is reasonable to expect that this
method of precracking may be effective in functionally graded materials of this type. In
order to explore this possibility, SE(B) specimens 50mm in length were machined in the
manner described earlier. These shorter specimens and a hemispherical loading fixture
were used to prevent buckling and shear failure of the beams in axial compression. A
gross stress of 212 MPa was found to be necessary to initiate precracking. This value is
only slightly below the yield strength of the CP titanium layer (280 MPa). Short cracks
were observed at the root of the machined notch, predominantly oriented in the direction
of the loading axis rather than in the desired perpendicular direction. However, the high
stress necessary to initiate cracking in the relatively brittle layer also resulted in plastic
deformation of the titanium-rich layers. These stresses resulted in microstructural damage
(cracking and debonding of the TiB particles from the matrix in the titanium-rich layers)
which was observed by scanning electron microscopy. In addition, the average elastic
modulus calculated from compliance in three-point bending changed from 145 GPa prior
to axial compression loading to 96 GPa after precracking. Due to this microstructural
damage, the axial compression technique was deemed unsuitable for use in this study.
Based on limitations of the methods described above, a new precracking method was
developed for FGMs and is described in detail in a related paper [22]. In order to create
compressive stresses at the crack tip and simultaneously avoid tensile deformation in the
pure titanium layer, the SE(B) specimens were cyclically loaded in reverse four-point
bending, as shown in Figure 3, at a 6 800 N maximum load. The load was applied with a
load ratio R = 0.1 at a cyclic frequency of 5Hz. These loading conditions were selected
with the goal of developing the precrack in approximately 5 000 cycles. In order to
achieve the desired a/W = 0.358 for the fracture experiments, the target precrack length
was 0.2 mm. The starting notch and resulting sharp precrack developed by reverse four-
point bending fatigue are shown by optical micrographs in Figure 4. The effective
P/2 P/2
l: 16.05:1
15% Ti/85% TiB
X
14.73
CP Ti
l
Note:All dimensionsare in millimeters.
79.38
Figure 3 - Seven-layer Ti/TiB reverse four-point bending specimen.
Figure 4 - (a) EDM notch in Ti/TiB FGM and (b) precrack resulting from reverse four-
point bending fatigue.
modulus was 145 GPa before and after precracking, indicating that, in contrast to the
axial compression method, reverse four-point bending did not damage the overall
microstructure of the FGM. After precracking, the specimens were heat tinted at 400~
and cooled overnight to room temperature prior to fracture testing.
Fracture tests of the precracked beams were conducted in general accord with the
provisions ofASTM E 1820-96. The outer loading span for these tests was 58.93 mm
(S = 4W). The load on each specimen was initially cycled twenty times at 2 Hz between
22.6 to 184.4 N to seat the clip gage. The load versus CMOD compliance of the specimen
was determined by loading three times from 89.2 to 266.7 N. The specimen was then
loaded to 356.0 N at a rate of 22.26 N/sec, after which the compliance was again
measured three times by reducing the applied load by 177.5 N and reloading.
Subsequently, the force was increased by the amount necessary to increase the crack
mouth opening displacement by 0.0005 mm and the compliance was again measured by
the above procedure. After each such increment, the specimens were held for 30 seconds
to accommodate any time dependent crack extension. Because the elastic modulus of the
FGM varies as the crack extends, the conventional compliance calibration could not be
used to determine crack length. Accordingly, independent crack length measurements
were taken after each increment of crack extension using a traveling microscope to
observe the polished specimen surface.
The fracture tests were continued until the cracks had propagated into the final layer
except for two tests that were terminated prematurely to study the mechanisms of crack
interaction with the TiB particles in Layers 5 and 6, respectively. After testing, the
specimens were removed from the load frame and heat tinted at 400~ in air for 15
minutes and cooled to room temperature. The initial and final cracks were measured in
accordance with ASTM E 1820-96 using a nine-point average. The initial crack lengths
ranged from 5.1 to 5.23 mm (a/W = 0.34 to 0.37). The final crack fronts were straight to
within 5% deviation.
Results
Residual Stress Results
Residual stresses were measured in samples removed from four locations of the
FGM plate. Released strain for one particular experiment is shown as a function of slot
depth in Figure 5. Error in the strain fit is plotted in Figure 6 versus the highest order of
polynomial assumed in the stress expansion, indicating an optimal order of 10 (nine terms
in the series). The strain fit resulting from Equation 3 and a 10th order polynomial is
shown in Figure 5. The residual stress at all four measurement locations is shown in
Figure 7, with the optimal order of stress expansion given in the legend for each location.
Redistributed residual stress and stress intensity were computed from residual stress
measured at location 8-1. Figure 8 shows the original and redistributed residual stress
together with the residual stress intensity factor, KRs.
These results indicate that the residual stresses have a negative contribution to the
crack driving force for crack size a > 2.7mm (a/W > 0.2). Since laboratory fracture
measurements are based on measured load, they assume that K = Ko~t. The actual state at
the crack tip is more correctly indicated by K = Kappt+ KRs. In the absence of residual
stress, one would measure the true material toughness, K,~,,,~,u. When residual stresses are
present, and have been measured, the laboratory measurements, K~a~, can be corrected by
the known KRs using
K,~,e~ = Kmeas+ KRs. (6)
Therefore, the results shown in Figure 8 indicate that fracture testing, without
correction for residual stress, will over-estimate material toughness for a/W > 0.2.
Fracture Results
Fracture toughness was calculated from the load versus optically-measured crack
+ Gage1 (Top) I I o Gage2 (Bottom) I.

20 ..... Fit ( T o p ) . . . . . Fit ( B o t t o m ) | 1200
0 ~ ......................................................................................... 1000
m
-20 ---~.... ! ............. ~ . . . . . . ~............ ? .......... ~ .......... ~.............. ~ - - 800 O
-40 , i i i i ~
O
3
...... ~ ! ....... ! ......... ? ........ ~............. I ............. I ........ : ..... 600
03 I i I I I I !
(9 '~i i i i i i I
400
Q. t~
o
I- 200
-8o I ............. I)- t~~.e:i+
+.........
- - T iii...........
.... ~ ...........
/ - :-. . . . . .i . . . i .............. ..... i::
-100 ~ - - + -~-- ~:--:~- -~:- 0
-120 , , ,~_, ,, I , , , , , , , , , , , , ~_~_, , -200

0 2 4 6 8 10 12 14
Slot Depth, a (mm)
Figure 5 - Strain release and numerical fit for slotting location 8-1.
1000 i , ~ ~ , , ~ i , i I ' ' ' , , , , , t i , ,
A "lit i i
f~ __ :', i [ * RMS error | i
4.. 100 . . . . . . . . . . . . . . . . i ...... -,...-i ...... - - A- - Max error ..i...........
I ~, i "! t I I
"~ 10 . . . . . . . . . . . . . . . . . . . . ~ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Location 8-1 i i :.
0 2 4 6 8 10 12 14
Highest Order Polynomial Term
Figure 6 - RMS and maximum error in strain fit for measurement location 8-1.
length data using the standard formula for linear elastic, isotropic materials. These
measured values are designated Kmeo,because it is not possible to completely verify the
validity of the fracture toughness values for this functionally graded material. An example
of the results for specimens precracked by the three different methods is illustrated in
Figure 9(a). These data reveal rising K behavior as the crack grows into successive layers
with increasing metal content. The locations of the interfaces between the layers are
noted on the plot.
The beam without a precrack exhibited rapid, unstable crack growth when loads
were increased to the point of crack initiation and gave limited useful fracture toughness
HILL ET AL. ON LAYERED FUNCTIONALLY G R A D E D MATERIAL 179
60 . . . . . . . . ! .... ! .... ] . . . . . . . . . . . . . . . .
.~ : : , ,
.../~.~'~'~_ .............. i I o 8-100th) I ................................................

40 ~ _ il ......... 8-2(8th) I i i i
=. ~' " ~ ' , ~1 - x.- - 7-1 (lOth) | is- ~x..~
20 ~"........7 ~ - ![- -~ 7-3 (8th.) I ~ = . = ......
~1 0 --,i............. i,............. i.. . . . . . . . . . . ..-.o

ia 4 ~i ........
*. , . . . . . . . . . . . . . . . . i . ............
-40 .........................................................................................
n- -60 ........................................ i............ i . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
: i
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8

Normalized Slot Depth (x/t)
Figure 7 - Measured residual stress for all four locations.
Error bars shown for location 8-1.
......... FGM slice .........
............................... !.......................... ! ............................
~ i''... i xx i i..... .
~ 1.0 ~
r 20 i .~'. o~...,
0 "v
=.
-m ' "~ i :. ~ i,: i ~ 3
-20 .... ~----i ....... ": ....... ~- ......... ,'-:"...............................
...............
! ...................... i .... 1.o
" -40 ~ , , iv ~ i ~ -2o
-60 ,,I ......... I ......... I .......... 3.0

0 5 10 15
Distance from SE(B) Front-face (mm)
Figure 8 - Original and redistributed residual stress with KRsfor location 8-1.
data. The specimen precracked by uniform axial compression showed stable crack growth
with no pop-in during the fracture test indicating that a sharp crack was initiated during
compressive loading. The load values and the resultant K values, however, are only 70%
of those obtained by the specimen that was precracked by reverse four-point bending. In
addition, the initial slope of the load versus CMOD curve for this specimen was
significantly lower than the slopes of the data for the other two specimens. The lower
values for the specimen precracked by axial compression are a consequence of the
extensive microcrack damage introduced by this precracking method.
Focussing on the results for the specimen precracked in reverse bending, the present
results demonstrate that K rises dramatically, from approximately 7 MPa'm 1:2(62% TiB)
in Layer 3 to over 30 MPa'm ~a in Layer 6 (15% TiB). This increase in toughness is as
expected based on the increasing metal content as the crack propagates toward the Ti rich
layer and confirms the general concepts predicted by Jin and Batra [9, 10]. Figure 9(b)
illustrates the effect of the residual stresses on the measured K values for the specimen
precracked by the reverse four point bending method. Since the residual stresses are
negative throughout layers in which the crack propagated, the corrected fracture
toughness, K~,enat, is less than the nominally measured value, Kmea~.This difference is
approximately constant at 2 MPa.m 1:2throughout the test.
Figure 9 - (a) Measured KI R-curves for Ti/TiB FGM specimens precracked by three
different methods and (b) measured, residual and resultant material Kt data for fracture
test of specimen precracked by reverse four-point bending.
Discussion
Accurate measurement of fracture toughness in brittle and ductile materials requires

a sharp precrack at the machined notch. In general, the precracking methods that have
been developed for homogeneous metallic and ceramic composite materials are not
appropriate for FGMs. Since the FGM studied in the present investigation contains both
ductile and brittle layers, it was necessary to develop a new method to introduce a sharp
precrack in a controlled manner. Conventional fatigue loading such that the notch root
experiences tension resulted in unstable crack propagation. Uniform axial compression
methods developed for ceramic materials introduced general microstructural damage in
the metal-rich layers resulting in lower than expected fracture toughness. In order to
combine these loading conditions to subject the notch root to cyclic compression while
the metal-rich layers remain in tension, a novel reverse four-point bending method was
developed. This new method provided a reliable means by which a controlled, sharp,
short precrack was developed at the machined notch in metallic/ceramic FGM beams,
allowing accurate fracture toughness measurements. However, it is important to
recognize that this method is limited in its applicability to ductile/brittle FGMs.
Depending on the properties of the components of a particular FGM, other precracking
methods may be more appropriate.
The data for crack propagation in the Ti/TiB functionally graded material clearly
reveal rising R-curve results when the crack propagates from the more brittle layers
toward the tougher layers. However, the shape of the curve differs substantially from that
normally observed in materials that exhibit rising KI R-curves because of differences in
the toughening mechanisms. Although extrinsic toughening mechanisms such as crack
bridging and crack deflection are evident in scanning electron microscope images of the
crack tip region [23], the fracture resistance of the FGM is dominated by increases in
intrinsic toughness as the volume fraction of TiB decreases with crack extension. The
mechanisms contributing to the intrinsic toughness of the Ti/TiB mixtures include
particle cracking and particle/matrix debonding. As the crack propagates through the
FGM, however, the volume fraction of Ti increases and these mechanisms become
relatively less important.
Because of the significant residual stresses present in the FGM tested in this
investigation, the initial measured toughness is significantly higher than the toughness
indicated by superposition. This result demonstrates the beneficial effect of the residual
stresses in this material. We note, however, that the present results are limited by the fact
that the calculated stress intensity values are based on the standard expressions for
isotropic, linear elastic solids. Although this approach provides a useful initial estimate
and may be warranted in continuously graded materials [8], further refinement of this
approach is needed for discrete, layered FGMs. 3
Near completion of this manuscript, a weight function for a single edge cracked
non-homogeneous plate was published [24]. This recent study suggests that for
the variation in modulus present in the Ti/TiB FGM, stress intensity factors
presented here are reduced by 5-10% relative to what would be calculated using
the results presented in [24].
Conclusions
The results of fracture testing of a layered Ti/TiB functionally graded material

suggest that both the precracking method and residual stresses have a significant effect on
the measured toughness. Single edge notch bend (SE(B)) specimens were fabricated so
that the crack propagation direction was perpendicular to the graded layers and advancing
from the brittle to the ductile side of the FGM. Reverse bending (crack tip in
compression) of specimens was found to produce a sharp precrack without damaging the
material, and appears to be appropriate for ductile/brittle FGMs. In contrast, uniaxial
compression damaged the specimen prior to fracture testing and conventional bending
(crack tip in tension) led to unstable and uncontrollable crack propagation.
Rising R-curve behavior for the FGM, starting at Km~s= 7 MPa'm za in the initial
crack-tip location (38% Ti and 62% TiB) was observed in a representative sample. As the
proportion of Ti increased with crack advance, toughness increased to Km~=
31 MPa'm ~z. Residual stresses acting normal to the crack-plane in the SE(B) specimens
were measured in the parent FGM plate using the crack compliance technique. The
remaining residual stresses in SE(B) specimens machined from the plate were found
computationally, and the weight function was used to determine their influence on
fracture toughness. Peak residual stresses were 30 to 50 MPa and for a representative
stress profile corresponded to a residual stress intensity factor (KRs) of -2.4 MPa-m la at
the initial precrack length. As a consequence, the actual material toughness at the start of
the test was calculated to be K,~,er~= Kme~+ KRs = 4.6 MPa'm la, or 34% less than that
measured by fracture testing. The residual stress intensity factor remained nearly constant
with crack extension, hence its relative contribution decreased to approximately 6.5% of
the maximum measured toughness.
Acknowledgments
This paper is based on work supported by the National Science Foundation, Grant
CMS 9713798. The authors wish to thank M. B. Prime, W. W. Liang, and E. M. Carrillo-
Heian for helpful discussions and Cercom Incorporated for the Ti/TiB FGM material
used in this study, The work of W.-Y. Lin for executing many of the residual stress
measurements is also acknowledged.
References
[1] Sasaki, M., Wang, Y., Hirano, T. and Hirai, T., "Design of SiC/C Functionally
Gradient Material and its Preparation by Chemical Vapor Deposition," Journal of
the Ceramic Society of Japan, Vol. 97, 1989, pp. 539-543.
[2] Kawasaki, A. and Watanabe, R., "Powder Metallurgical Fabrication of the Thermal-
Stress Relief Type of Functionally Gradient Materials," in Sintering, Tokyo, Japan,
1987, Vol. 2, Elsevier, London, pp. 1197-1202.
[3] Fukushima, T., Kuroda, S. and Kitahara, S., "Gradient Coatings Formed by Plasma
Twin Torches and Those Properties," in Proceedings of the First International
Symposium, FGM, Sendai, Tokyo, Japan, 1990, Functionally Gradient Materials
Forum and the Society of Non-traditional Technology, pp. 145-150.
HILL ET AL. ON LAYERED FUNCTIONALLY GRADED MATERIAL 183
[4] Sata, N., Sanada, N., Hirano, T. and Niino, M., "Fabrication of a Functionally
Gradient Material by Using a Self-propagating Reaction Process," Proceedings of
the First International Symposium on Combustion and Plasma Synthesis of High-
Temperature Materials, VCH Publishers, Inc., 1990, pp. 195-203.
[5] Miyamoto, Y., Nakanishi, H., Tanaka, I., Okamoto, T. and Yamada, O., "Gas
Pressure Combustion Sintering of TiC-Ni FGM," in Proceedings of the First
International Symposium, FGM, Sendai, Tokyo, 1990, Functionally Gradient
Materials Forum and the Society of Non-traditional Technology, pp. 257-262.
[6] Tomasi, A. P., Saiz, E., Ishibashi, H., Diaz, M., Requena, J. and Moya, J. S.,
"Powder Processing of Mullite/Mo Functionally Graded Materials," Journal of the
European Ceramic Society, Vol. 18, 1998, pp. 1365-1371.
[7] Erdogan, F. and Wu, B. H., "Analysis of FGM Specimens for Fracture Toughness
Testing," Functionally Gradient Materials, Ceramic Transactions, American
Ceramic Society, Westerville, OH., 1993, pp. 39-46.
[8] Jin, Z.-H. and Noda, N., "Crack-tip Singular Fields in Nonhomogeneous Materials,"
Journal of Applied Mechanics, Vol. 61, No. 3, 1994, pp. 738-740.
[9] Jin, Z. H. and Batra, R. C., "R-curve and Strength Behavior of a Functionally
Graded Material," Materials Science and Engineering A, Vol. 242, No. 1-2, 1998,
pp. 70-76.
[10] Jin, Z. H. and Batra, R. C., "Some Basic Fracture Mechanics Concepts in
Functionally Graded Materials," Journal of the Mechanics and Physics of Solids,
Vol. 44, No. 8, 1996, pp. 1221-1235.
[11] Tsang, H. T., Chao, C. G. and Ma, C. Y., "Effects of Volume Fraction of
Reinforcement on Tensile and Creep Properties of in-situ TiB/Ti MMC," Scripta
Materia!ia, Vol. 37, No. 9, 1997, pp. 1359-1365.
[12] Prime, M. B., "Residual Stress Measurement by Successive Extension of a Slot: A
Literature Review," Technical Report LA- 13283-MS, Los Alamos National
Laboratory, 1997.
[13] ABAQUS/Standard, Version 5.8, Hibbitt, Karlsson, and Sorenson, Inc., Providence,
RI, 1999.
[14] Giannakopoulos, A. E., Suresh, S., Finot, M. and Olsson, M., "Elastoplastic
Analysis of Thermal Cycling: Layered Materials with Compositional Gradients,"
Acta Metallurgica et Materialia, Vol. 43, No. 4, 1995, pp. 1335-1354.
[15] Matlab, Version 5.3.0, The MathWorks, Inc., Natick, MA, 1999.
[16] Lin, W.-Y., "Measurement of Residual Stress Distribution Through the Thickness of
Functionally Graded Material Ti/TiB2," Master of Science, Mechanical and
Aeronautical Engineering Department, University of California, Davis, 1999.
[17] Glinka, G. and Shen, G., "Universal Features of Weight Functions for Cracks in
Mode I," Engineering Fracture Mechanics, Vol. 40, No. 6, 1991, pp. 1135-1146.
[18] Suresh, S. and Brockenbrough, J. R., "Theory and Experiments of Fracture in
Cyclic Compression: Single Phase Ceramics, Transforming Ceramics and Ceramic
Composites," Acta. Metallurgica, Vol. 36, No. 6, 1988, pp. 1455-1470.
[19] Ewart, L. and Suresh, S., "Dynamic Fatigue Crack Growth in Polycrystalline
Alumina under Cyclic Compression," Journal of Materials Science Letters, Vol. 5,
1986, pp. 774-778.
[20] ASTM PS 070-97, American Society for Testing and Materials, West
[21] Sakai, M. and Bradt, R. C., "Fracture Toughness Testing of Brittle Materials,"
International Materials Reviews, Vol. 38, No. 2, 1993, pp. 53-78.
[22] Carpenter, R. D., Paulino, G. H., Munir, Z. A. and Gibeling, J. C., "A Novel
Technique to Generate Sharp Cracks in Metallic/Ceramic Functionally Graded
Materials By Reverse 4-Point Bending," Scripta Materialia, Vol. 43, 2000, pp. 547-
552.
[23] Carpenter, R. D., Liang, W. W., Paulino, G. H., Munir, Z. A. and Gibeling, J. C.,
"Fracture Testing and Analysis of Layered Functionally Graded Ti/TiB in 3-Point
Bending," Acta Materialia, to be published.
[24] Fett, T., Munz, D. and Yang, Y. Y., "Direct Adjustment Procedure for Weight
Functions of Graded Materials," Fatigue and Fracture of Engineering Materials
and Structures, Vol. 23, No. 3, 2000, pp. 191-198.
Unique Materials and Environmental
Effects
Stefano Guicciardi, 1 Cesare Melandri, l Frederic Monteverde, l Alida Bellosi, l
and Goffredo de Portu 1
Environmental and Thermal Effects on the Toughness of TiCN-Based Materials
Reference: Guicciardi, S., Melandri, C., Monteverde, F., Bellosi, A., and De Portu, G.,
"Environmental and Thermal Effects on the Toughness of TiCN-Based Materials,"
1409, J. A. Salem, G. D. Quinn, and M. G. Jenkins, Eds., American Society for Testing
Abstract: Titanium carbonitrides (TiCN) are very hard materials with increasing
technological importance. They are mainly applied in composites with various metal
carbides and/or metallic binders for cutting tools. In these applications, high
temperatures are usually involved and, sometimes, chemical reactions. Being composed
of a metallic-binder phase, in which a hard ceramic phase is dispersed, the performance
of these materials can be largely affected by temperature and environment. In this work,
nearly fully-dense TiCN based cermets have been produced starting from the following
composition (wt%): TiC0.sN0.5+15.3(WC-5 %Co)+6.2Ni+2.1 Co. Some of the mechanical
properties have been routinely measured, such as hardness, Young modulus, fracture
toughness and flexural strength. The main effort has been devoted to the study of the
effects of temperature and environment on the fracture toughness. Chevron notch tests
have been carried out at room temperature, 600 ~ 800 ~ and 1 000 ~ in static air,
flowing air and flowing nitrogen atmosphere. The influence of the interaction between
temperature, environment and microstructure on the fracture behavior has been
analyzed.
Keywords: cermets, fracture toughness, high temperature, environmental effects,

mechanical properties
Introduction
Titanium carbonitrides (TiCN) are very hard materials with increasing technical
importance. They are mainly applied in composites with various metal carbides and/or
~Research scientist, Research Institute for Ceramics Technology (IRTEC-CNR),

Via Granarolo 64,1-48018 Faenza, Italy.
187
binder metals, the so-called cermets, for metal-cutting operations [1-4]. The typical
compositions and microstructure of the cermets are very complex and heterogeneous,
due to the variation of type and amount of the ceramic compounds and of the
metal-binder systems: cermets, in fact, can be described as the combination of hard
particles (carbides, borides, nitrides, carbonitrides, or a combination of these) bonded
together with a metallic phase such as cobalt (Co), nickel (Ni), iron (Fe), or a
combination of them [1, 3, 5, 6]. The high hardness and wear resistance are provided by
the hard particles, while the ductility, toughness, and thermal shock resistance are
provided by the metallic-binder phase. For many applications, including metal cutting,
the ideal material should exhibit high hardness and high toughness. However, as
hardness and toughness are generally antagonistic properties, either a compromise has to
be found between these two properties or one property should be favored on the basis of
the specific application. TiCN-based materials are characterized by high creep and
diffusion-wear resistance, due to the formation of a rim around the TiCN core: it
enhances wetting of the carbonitrides to the metal binder and inhibits the coalescence
and growth of the carbonitride grains during sintering [7]. The microstructure of
TiCN-based cermets can range from typical ceramic configuration, where metal binders
are in low amounts as a grain boundary phase, to metal-ceramic composites where the
metal or alloy phases constitute a matrix in which ceramic particles are dispersed. The
alloy composition of the binder phase depends on the overall stoichiometry and nitrogen
and carbon content of the hard phase [1, 8-10]. The physical and mechanical properties
of these cermets can be adjusted within certain limits to meet the requirements of the
application, through design of the compositions and choice of raw materials and
processing parameters [10-19]. Like most of the cermets, performances of TiCN-based
materials are strongly dependent on bulk and surface characteristics [11, 12];
compositional gradients can be introduced in order to improve the properties for specific
applications [13], while surface oxidation or corrosion can negatively affect some of the
mechanical properties.
In this work, the fracture characteristics of a TiCN-based material are studied by
measuring the effects that temperature and environment have on the fracture toughness.
Chevron notch fracture tests were carried out at room and high temperatures in static air,
flowing air, and flowing Nz.
Experimental
Starting from commercial powders: TiC0.sN0.5 (HCST-Starck, Grade E), WC-5%

Co (Nanodyne, Nanocarb), Ni (Carbonyle) and Co (Sherrit), the following powder
mixture was prepared (amounts in wt%): TiC0.sN0.5 + 15.3 (WC-5%Co)+ 6.2 Ni+2.1 Co.
More details on processing and characteristics of the dense material are reported
elsewhere [18, 19]. The powder mixture was sintered through hot pressing (applied
pressure 5 MPa, sintering temperature 1 650 ~ holding time 30 rain). On the dense
material, the following microstructural characteristics were assessed: identification and
measurement of the lattice parameters of crystalline phases (X-ray diffraction), grain
size, morphology, core/rim structure, porosity, defects (scanning electron microscopy).
The Young's modulus (E) was measured using the resonance frequency method.
Vickers microhardness (HV) was measured using a load of 9.81 N (loading time 15 s).
Flexural strength (o) was measured on a four-pt bending fixture, 20 mm lower span, and
GUICCIARDI ET AL. ON TOUGHNESS OF TiCN-BASED MATERIALS 189
10 mm upper span, on bars 25 x 2.5 x 2 mm 3 (length x width x thickness, respectively,

where the width is perpendicular to the axis of bending) at room and high temperature
(in air). Fracture toughness (Kic) was evaluated using the chevron-notched beam (CNB)
in flexure. The bars, 25 x 2 x 2.5 mm 3 (length x width x thickness, respectively, where
the width, 2 mm in this case, is perpendicular to the axis of bending) were notched with
a 0.08 mm diamond saw; tt0 and tzl were about 0.12 and 0.80, respectively. The flexural
tests were performed on a semi-articulated alumina four-pt jig with a lower span of 20
mm and an upper span of l0 mm in a high temperature testing furnace mounted on an
universal screw-type testing machine. For the relative tests, air and N2 were introduced
into the furnace-chamber with a flow of 41/min just after closing the chamber and for
the whole duration of the test, including cooling. The specimens, five for each
combination of temperature and atmosphere, were deformed with a crosshead speed of
0.05 mm/min. To ensure thermal equilibrium, the test was started 20 min after the
desired temperature was reached. The "slice model" equation of Munz et al. [20] was
used for the calculation of fracture toughness. SEM analysis was carried out on the
fracture surfaces.
Microstructure
The microstructure of the material is shown in Figure 1. Some of the salient

microstructural characteristics are reported in Table 1.
Figure l - SEM micrograph from a polished surface of the material. Labels identify the
phases (see text).
The microstructure of the material shows grains with the typical core-rim type
structure (Figure 1); the Z-contrast imaging mode distinguishes the core, i.e., original
TiCN particles, from the rim, i.e. the outer shell of TiCN particles that reacted with WC.
Ti(C,N), labeled as phase et in Figure 1 and Table 1, is a solid solution between the TiC
and TiN phases; (Ti,W)(C,N), labeled as phase 3, in Figure 1 and Table 1, is a solid
Table 1 - Microstructural parameters.
-Density, g/cm3 Crystalline phases Phase area fraction Meangrain size, Inn Lattice parameter, nm
5.82 (98 ~) ct, 13,~/2 0.37 (0.58) 3 0.75 0.429 (0.354) '
i % th.d.; 2ct=Ti(C,N), []~Ni-Co-Ti-W,~/=(Ti,W)(C,N);3or-phase(~phase)i 4T-phase(I]-phase).
solution containing also W ions. Ti(C,N) and (Ti,W)(C,N) have a cubic symmetry with
almost the same lattice parameter. The Ti/W atomic ratio in this phase is about 6 and the
area fractions occupied by phases ct and y are 0.37 and 0.58, respectively. Also, the
binder phase, labeled as ~, is characterized by a cubic symmetry and its stoichiometry is
in the system Ni-Co-Ti-W; its composition was assessed on mirror-like flat surfaces
where the binder could be easily localized [19].
Mechanical Properties
The values of some of the mechanical properties, mean+_l standard deviation when
appropriate, are shown in Table 2.
Table 2 - Mechanical properties.
Young's modulus, GPa Hardness, GPa Flexural strength, MPa
R.T. 600 ~ 800 ~ 900 ~ 1 000 ~

445 16.8_+0.3 I 286_+98 1 064+151 858+1 6 3 3 + 3 3 339+_56
The high values of Young's modulus and hardness, in comparison to similar

cermets, are the result of the relatively high amount of ceramic phase in the composition
of this material. For the same reason, the flexural strength at room temperature is lower
than that of other TiCN-based cermets [1]. The progressive degradation of the flexural
strength with the temperature is due to the softening of the metal phase and to the
oxidation attack on the specimen surface. The load-displacement curves are already non-
linear at 800 ~ and this behavior is more and more evident at higher temperatures [12],
Fracture Toughness
The fracture toughness results are summarized in Table 3 and Figure 2. At room
temperature, the fracture toughness tests were performed only in static air as it was
supposed that there was no influence by the surrounding environment at this

temperature.
Typical load-displacement curves are shown in Figut:e 3; they were almost
Table 3 - Fracture toughness values.
Temperature, ~ Fracture toughness,MPa.m~
Static air Flowin~gair Flowingnitrogen

R.T. 6.9_+0.4* . . . . . .
600 6.9_+0.4 6.9-2-0.5 6.9_+0.4
800 10.5_+1.4 13.3_+2.0 I 1.7_-/-0,8
1 000 8.0-2-0.8 7.8_+0.7 8.7_+0.5
*mean+l standarddeviation
independent of the atmosphere. In this figure, as already observed during the flexural
strength tests, the progressive softening of the material due to the metallic intergranular
phase is clearly evident. The arrows indicate where the load-displacement curve departs
from a linear behavior which is due to the contribution of the metallic phase softening
and the stable crack extension which is a characteristic feature of the chevron notch test
bars [21].
The SEM analysis on the fracture surfaces of the specimens broken at room
temperature revealed that the material failed mainly intergranularly, see Figure 4. The
same behavior was observed at 600 ~ and at 800 ~ see Figure 5, while at 1 000 ~
the surfaces of all the specimens were so heavily altered that it was not possible to
distinguish any salient feature; see Figure 6.
17
tr~
c5 ~ Static air [
15 Flowingnitrogen
Flowinga r
~., 13
~d
2
T
5
0 200 400 600 800 1000 1200
Temperature, ~
Figure 2 - Fracture toughness as a function o f temperature and environment.

(b)
Displacement --) Displacement ---)
T I (c) ",.
,d
Displacement Displacement
Figure 3 - Load-displacement curves: (a) room temperature, (b) 600 ~ (c) 800 ~
and (d) 1 000 ~
In order to ascertain if the apparent differences of the fracture toughness values as

a function of the various temperatures and atmospheres were statistically significant, the
Figure 4 - SEM micrograph of a fracture surface. CNB specimen broken at room

temperature. Picture taken in the stable crack growth region.
Figure 5 -SEM micrographs from a

fracture surface of CNB specimens
tested at 800 ~ in static air (a), flowing
air (b) and flowing nitrogen (c). Pictures
taken in the stable crack growth region.
analysis of variance (ANOVA) was carried out. Unfortunately, the A N O V A test could
not be applied on the whole data set as the variances of the data groups were correlated
to their means, see Figure 7, and this violates a critical assumption for the A N O V A
analysis [22]. Therefore, the A N O V A test, specifically the Scheff6 test, which is the
most conservative, was performed twice in order to reduce the heteroscedasticity effects.
The first time, each atmosphere was considered separately and the temperature was

fracture surface of CNB specimens
tested at I 000 ~ in static air (a),
flowing air (b) and flowing nitrogen (c).
Pictures taken in the stable clack growth
region.
2.5-
d
E
2.0- ............---';
. .~176
1.5
1.0
"O
0.5
0.0
6 8 10 12 14
Mean, MPa.m ~
Figure 7 - Plot of the fracture toughness standard deviations vs the respective means.
Coefficient of correlation of the data: r=0.87. Dashed lines are 95% confidence
bounds.
taken as a factor; the second time, the test was applied at each temperature point with
the atmosphere as a factor. The results of the ANOVA analysis indicated that 1) in static
air, there was no difference between the fracture toughness at room temperature and at
600 ~ 2) independently from the atmosphere, the fracture toughness values at 800 ~
and at 1 000 ~ were different from that at 600 ~ with the fracture toughness at 800 ~
proving the highest; 3) independently from the atmosphere, the fracture toughness at
1 000 ~ was higher than that at 600 ~ 4) at 600 ~ and 1 000 ~ there was no
influence of the atmosphere on the fracture toughness; 5) at 800 ~ the fracture
toughness measured in flowing air was higher than that measured either in static air or in
flowing nitrogen which, in turn, were statistically the same.
That the material at hand exhibits a fracture toughness peak at high temperature, in
this case at 800 ~ is typical of microstructure with a softening grain boundary phase.
Such behavior has already been observed in ceramics like silicon-infiltrated SiC [23] or
Si3N4 [24] and explained by the blunting of the crack tip: initially, it is coupled with a
still very viscous binder phase and this promotes the fracture toughness increase, while,
at very high temperature, the binder phase becomes less and less viscous and this
promotes the fracture toughness degradation. In our material, the softening of the
metallic binder starts between 600 ~ and 800 ~ and is still effective at 1 000 ~ as
indicated by the ANOVA results
More complex is the effect of the atmosphere. In Si3Na-based ceramics, for
example, no difference was found between the high-temperature fracture toughness in
air or in vacuum [25], but, for the silicon-infiltrated SiC reported in [24], the
high-temperature fracture toughness in air was higher than in vacuum. In our case, the
only temperature at which the atmosphere seemed to play a role is 800 ~ At this
temperature, especially for the specimens tested in static or flowing air, the oxidation
product was clearly observable both on the fracture surfaces, see Figure 5, and on the
cross section of the specimens, see Figure 8. In flowing air, the thickness of oxidation
product scale formed on the specimen surface was almost double that formed in static

polished cross section showing the
surface scale of the specimens tested at
800 ~ in static a& (a) and flowing air
(b). In the specimen tested in flowing
nitrogen (c), the scale was not observed.
air, while there was no evidence of such a scale when the test was performed in flowing
nitrogen, Figure 8. Notwithstanding that the fracture surface of the specimens broken in
flowing nitrogen had almost the same appearance of those broken in static or flowing
air, see Figure 5, the thickness of the scale formed in flowing nitrogen was hardly
detectable when observed in cross section, see Figure 8. From the fracture point of view,
the presence of the scale was not a crucial factor for the fracture toughness: the values
obtained in flowing nitrogen were in fact not different from those obtained in static air.
Only under flowing air were the fracture toughness values influenced.
To understand this behavior better, flexural tests were carded out at 800 ~ with
bars where three indentations (load: 98.1 N) were placed on the tensile surface with a
Vickers indenter. The cracks perpendicular to the applied tensile stress emanating from
the comers of the indentations were in this way allowed to propagate stably, the failure
of the specimen being due to only one of the three indentations. Examples of crack path
during the stable propagation are shown in Figure 9 for each atmosphere. As can be
seen, the crack path was almost straight in the case of flowing nitrogen (Figure 9c),
while in the case of static and flowing air, the crack path was characterized by crack
arrests and restarts from different positions (Figures 9a and 9b). In the case of flowing
air, the crack path was slightly more tortuous than in the other two conditions for each
crack portion (Figure 9b). The crack arrest and the crack restart from a different position
can be due to the oxidation, which interacts with the microstructure affecting the crack
propagation in a very complex way. Regarding the crack stop, oxidation can reduce the
stress intensity factor of the advancing crack below the critical value for the propagation
by crack blunting or healing [26-28] and/or by increasing the fracture energy [29,30];
for the restart, oxidation can create damaged zones, for example, by localized grain
boundary embrittlement or oxidation pits, from which another main crack can be
triggered and propagated. The situation resembles that of creep crack growth already
observed in ceramics with glassy intergranular phase [31]. The more tortuous crack
Figure 9 - SEM micrographs of crack

path (arrowed) in specimens tested at
800 ~ in static air (a), flowing air (b)
and flowing nitrogen (c). Pictures taken
in the stable crack growth region. Crack
propagation from left to right.
path, in the case of flowing air, can be promoted by a more severe grain boundary
embrittlement that derives from the rapid oxidation of the metallic intergranular phase in
presence of a larger amount of oxygen availability. The higher fracture toughness
measured in flowing air should be therefore attributed to a sort of crack deflection
mechanism, which seems to be effective when the oxygen availability in the atmosphere
is above a certain threshold.
Conclusions
The fracture toughness of a titanium carbonitride-based material was measured at

high temperature in different atmospheres, i. e., static air, flowing air and flowing
nitrogen. As a general behavior, the fracture toughness, independently from the
atmosphere, showed a toughness peak at 800 ~ due to the softening of the intergranular
metallic phase. Generally, no influence of the atmosphere was found as a function of
temperature except at 800 ~ where the toughness measured in flowing air was higher
than that in the other two conditions. It is proposed that this effect is due to a deflection
toughening mechanism caused by microstructural damage due to an enhanced oxidation
process caused, in turn, by a higher oxygen availability.
References
[1] Ettmayer, P., Kolaska, H., Lengauer, W., and Dreyer, K., "Ti(C,N) Cermets-
Metallurgy and Properties," International Journal of Refractory Metals and
Hard Materials, Vol. 13, No. 6, 1995, pp. 343-351.
[2] Zhang, S., "Material Development of Titanium Carbonitride-Based Cermets for
Machining Applications," Key Engineering Materials, Vol. 138-140, T r P
Switzerland, 1998, pp. 521-543.

[3] Watanabe, T., Dotsu, T., and Nakanishi, T., "Sintering Properties and Cutting-Tool
Performance of Ti(C,N)-Based Ceramics," Key Engineering Materials Voi.
114, TTP Switzerland, 1996, pp. 189-266.
[4] Clark, E. B. and Roebuck, B., "Extending of Application Areas for Titanium
Carbonitiide Cermets," Refractory Metals and Hard Materials, Vol. 11, No.
1, 1992, pp. 23-33.
[5] Gustafson, P. and Ostlund, A., "Binder-Phase Enrichment by Dissolution of Cubic
Carbides," International Journal of Refractory Metals and Hard Materials,
Vol. 12, No. 3, 1993-1994, pp. 129-136.
[6] Holleck, H. and Kleykamp, K., "Constitution of Cemented Carbide Systems,"
Refractory Metals and Hard Materials, Vol. 1, No. 3, 1982, pp. 112-116.
[7] Laoui, T., Zou, H., and Van der Biest, O., "Analytical Election Microscopy of the
Core/Rim Structure in Titanium Carbonitride Cermets," International Journal
of Refractory Metals and Hard Materials, Vol. 11, No. 4, 1992, pp. 207-212.
[8] Ehira, M. and Egami, A., "Mechanical Properties and Microstiucture of Submicron
Cermets," International Journal of Refractoy Metals and Hard Materials,
Vol. 13, No. 5, 1995, pp. 313-319.
[9] Rynemark, M., "Investigation of Equilibria in theTi-W-C-N System at 1750~ ''
Refractory Metals andHard Materials, Vol. 10, No. 4, 1991, pp. 185-193.
[10] Laoui, T. and Van der Biest, O., "Effect of TiC additions on the Microstructure
and Properties of Ti(C,N)-WC-Co-Ni Cermet," Journal of Materials Science
Letters, Vol. 13, No. 21, 1994, pp. 1530-1532.
[11] Xingzhong, Z., Jiajun, L., Baoliang, Z., Hezhuo, M., and Zhenbi, L., "Wear
Mechanisms of Ti(C,N) Ceramic in Sliding Contact with Stainless Steel,"
Journal of Materials Science, Vol. 32, No. 11, 1997, pp. 2963-2968.
[12] Bolognini, S., Feusier, G., Mari, D., Viatte, T., and Benoit, W., "High Temperature
Mechanical Behaviour of Ti(C,N)-Mo-Co Cermets," International Journal of
Refractory Metals and Hard Materials, Vol. 16, No. 4-6, 1998, pp. 257-268.
[13] Zackrisson, Z., Rolander, U., Weine, G., and Andren, H. O., "Microstiucture of the
Surface Zone in a Heat Treated Cermet Material," International Journal of
Refractory Metals and Hard Materials, Vol. 16, No. 4-6, 1998, pp. 315-322.
[14] Liuying, M., Huayi, L., Hengzhen, B., and Gouchun, L., "Influence of CHdN2
Ratio on Microstructure and Mechanical Properties of PCVD-(T(CxNt.x)
Coating," International Journal of Refractory Metals and Hard Materials,
Vol. 13, No. 6, 1995, pp. 369-374.
[ 15] Pastor, H., "Titanium Carbonitiide-Based Hard Alloy for Cutting Tools,"
Materials Science and Engineering, Vol. A 105/106, 1988, pp. 401-409.
[16] Zhang, S., Qin, C. D., and Lim, L. C., "Solid Solution ofWC ad TaC in Ti(C,N) as
Revealed by Lattice Parameter Increase," International Journal of Refractory
Metals and Hard Materials, Vol. 12, No. 6, 1993-94, pp. 329-333.
[17] Park, D.-S. and Lee, Y.-D., "Effect of Carbides on the Microstructure and
Properties of Ti(C,N)-Based Ceramics," Journal of the American Ceramic
Society, Vol. 82, No. 11, 1999, pp. 3150-3154.
[18] Bellosi, A., Medri, V., and Monteverde, F., "processing and Properties of Ti(C,N)-
WC-Based Materials," accepted for publication on Journal of American
Ceramic Society with letter dated 25th March 2001.
[19] Monteverde, F., Medri, V., Melandri, C., Dalle Fabbriche, D., and Bellosi, A.,
"Microstructure and Properties of Titanium Carbonitride-Based Certamics,"
in Proceedings EURO PM "Advances in Hard Materials Production",
European Powder Metallurgy Association, 1999, pp. 253-260.
[20] Munz, D. G., Shannon, J. L., Jr., and Bubsey, R. T., "Fracture Toughness
Calculation from Maximum Load in Four Point Bend Tests of Chevron Notch
Specimens," International Journal of Fracture, Vol. 16, 1980, pp. R137-
R141.
[21] Sung, J., and Nicholson, P. S., "Valid Ktc Determination via In-Test Subcritical
Precracking of Chevron-Notched Bend Bars," Journal of the American
[22] STATISTICATM for Windows, Volume I: General Conventions and Statistics I, 2 "d
Edition, StatSoft Inc., Tulsa, USA, 1995, p. 1711.
[23] Pabst, R. F. and Popp, G., "Critical and Subcritical Crack Extension in A1203 and
SiC as Functions of Temperature, Environment and Loading Rate," Fracture
Mechanics of Ceramics, Vol. 5, R. C. Bradt, A. G. Evans, D. P. H.
Hasselman, and F. F. Lange, Eds., Plenum Press, New York, 1983, pp. 305-
315.
[24] Gosh, A., Jenkins, M. G., Ferber, M. K., Peussa, J. and Salem, J. A., "Elevated-
Temperature Fracture Resistance of Monolithic and Composite Ceramics
Using Chevron-Notched Bend Tests," Fracture Mechanics of Ceramics, Vol.
10, R. C. Bradt, D. P. H. Hasselman, D. Munz, M. Sakai, and V. Ya.
Shevchenko, Eds., Plenum Press, New York, 1992, pp. 89-107.
[25] Mutoh, Y., Yamaishi, K., Miyahara, N., and Oikawa, T., "Brittle-to-Ductile
Transition in Silicon Nitride," Fracture Mechanics of Ceramics, Vol. 10, R.
C. Bradt, D. P. H. Hasselman, D. Munz, M. Sakai, and V. Ya. Shevchenko,
Eds., Plenum Press, New York, 1992, pp. 427-440.
[26] Easler, T. E,, Bradt, R. C., and Tressler, R. E., "Effects of Oxidation under Load
on Strength Distributions of Si3N4," Journal of the American Ceramic Society,
Vol. 65, No. 6, 1982, pp. 317-320.
[27] Carroll, D. F. and Tressler, R. E., "Time-Dependent Strength of Siliconized Silicon
Carbide under Stress at 1000 ~ and 1100~ '' Journal of the American Ceramic
Society, Vol. 68, No. 3, 1988, pp. 143-146.
[28] Ogasawara, T., Hori, T., and Okada, A., "Threshold Stress Intensity for Oxidative
Crack Healing in Sintered Silicon Nitride," Journal of Material Science
Letters, Vol. 13, No. 6, 1994, pp. 404-406.
[29] Maugis, D., "Subcritical Crack Growth, Surface Energy, Fracture Toughness,
Stick-Slip and Embrittlement," Journal of Material Science, Vol. 20, No. 9,
1985, pp. 3041-3073.
[30] Lawn, B. R., Roach, D. H., and Thomson, R. M., "Thresholds and Reversibility in
Brittle Cracks: an Atomistic Surface Force Model," Journal of Material
Science, Vol. 22, No. 11, 1987, pp. 4036-4050.
[31] Jakus, K., Wiederhorn, S. M., and Hocket, B. J., "Nucleation and Growth of
Cracks in Vitreous-Bonded Aluminum Oxide at Elevated Temperatures,"
Journal of the American Ceramic Society, Vol. 69, No. 10, 1986, pp. 725-731.
George Gogotsi ~
Fracture Toughness Studies on Ceramics and Ceramic Particulate Composites at

Different Temperatures
Reference: Gogotsi, G. A., "Fracture Toughness Studies on Ceramics and

Ceramic Particulate Composites at Different Temperatures," Fracture Resistance
Testing of Monolithic and Composite Brittle Materials, ASTM STP 1409, J. A. Salem,
G. D. Quinn, and M. G. Jenkins, Eds., American Society for Testing and Materials,
West Conshohocken, PA, 2002.
Abstract: The fracture toughness of alumina, zirconia, silicon nitride, and silicon
carbide ceramics as well as silicon nitride and silicon carbide particulate composites
were studied by the single edge V-notched beam (SEVNB) method over a wide
temperature range. Manual and mechanical procedures for V-notch polishing-out and
three- and four- point flexure tests were developed. Load-deflection diagrams for V-
notched specimens contributed to better understanding of the deformation behavior o f
ceramics at room temperature and 1300 - 1400 ~ SENB (single edge notched beam)
and SEPB (single edge precracked beam) as well as micro-Raman spectroscopy data
were used to analyze the SEVPB results. The ratio between SEVNB and SENB results
is about 0.6 for elastic materials and over 0.9 for inelastic ones. The results for
zirconia, in the case o f a V-notch root radius less than 5-7 pro, are in good agreement
with SEPB results. The polishing-out of a V-notch does not lead to the tetragonal-
monoclinic phase transformation in these materials.
Key words: ceramics, ceramic particulate composites, fracture toughness, SEVNB-

method, flexure test, temperature, micro-Raman spectroscopy
Introduction
Ceramics and ceramic composites are promising materials having rather high
strength characteristics but quite low crack resistance properties at the same time. This
is one o f the major factors hindering the wide-scale application o f these materials in
various fields of human activities. The crack resistance is critical not only for ceramic
products operating under extreme mechanical and thermal loads but also for structural
components whose brittle fracture is intolerable even under arbitrary loads.
Professor, Department of Strength of Ceramic Materials, Institute for Problems of

Strength, 2 T imiryazevskaya Street, K iev, 01014 Ukraine.
199
For many years, the performance o f ceramics has been evaluated on the basis o f
full-scale tests. However, their fracture toughness characteristics have not always
been the object of scientific interest. Wide-scale fracture toughness investigations
were started only in the late 1980s. International prestandard studies [Round Robin on
Fracture Toughness (RRFT)] aimed at the assessment o f the accuracy and reliability
of the data obtained by commonly accepted test methods were important steps in this
field (Table 1). Leading specialists from different countries engaged in those
investigations and performed statistical analyses to compare the test methods
summarized in Table 2. They were helpful for creating the basis o f more reliable
determination of the fracture toughness characteristics o f ceramics. For a long time,
no particular preference was given to any o f the methods considered because each o f
them had appreciable shortcomings, which restricted their application [7, 8].
Recently, much attention has been focused on the SEVNB method [9], which is a
further development of the simple and widely used single edge notched beam (SENB)
method (Table 2), which is suitable only for rough fracture toughness evaluations.
The test data obtained by the SEVNB method were found to be comparable to those
obtained by the SEPB method, which is usually considered as providing the real value
of the critical stress intensity factor (fracture toughness). Unfortunately, the SEPB test
is often difficult to realize in practice [10]. To evaluate the potential o f the SEVNB
method as a standard method for testing ceramics, an analysis was performed [6]. By
participating in this work (RRFT'97 in Table 1), we verified the efficiency of our
Table ! - International investigations of fracture toughness of ceramics and ceramic

composites
Name Date Our index Methods

Fracture Toughness Round Robin
Test (RRT) at Room Temperature 1989 - 92 RRFT'89 [1] SEPB, IS, IF
VAMAS Round Robin on Fracture
Toughness o f Silicon Nitride at High SEPB, SEVNB,
Temperature 1990 - 93 RRFT'90 [2] CNB
Fracture Toughness o f Advanced
Ceramics by the Surface Crack in
Flexure (SCF) Method: A VAMAS
Round Robin 1992 - 93 RRFT'92 [3] SCF
VAMAS Round Robin on Fracture
Toughness Measurement o f Ceramic SEPB, SENB,
Matrix Composite 1994 - 9 8 RRFT'94 [4] SEVNB
ESIS TC 6 Round Robin on Fracture SEPB, CNB, IS,
Toughness 1995 - 9 6 RRFT'95 [5] SENB, SEVNB, IF
VAMAS TWA # 3/ESIS TC 6
Round Robin on Fracture Toughness
of Ceramics Using the SEVNB
Method (Single-Edge-V-Notched
Beam) 1997 - 99 RRFT'97 [61 SEVNB
GOGOTSI ON CERAMIC PARTICULATE COMPOSITES 201
Table 2 - Conventional*fracture toughness test methods
Method and Method and

Specimen fracture surface of Specimen fracture surface of
a specimen a specimen
~ w SEPB ~ . . ~v ~ , SCF
(Single Edge ~ (Surface Crack in

Precracked Beam) Flexure)
r ~ ' ~ / w (Single Edge (Indentation

Notched Beam) Strength)
~ w CNB ~'~fff~') IF
(Indentation
(Chevron Notched Fracture)
Beam)
/
*The SVNB method is not included in the table because it has not found wide
application as yet.
procedures and obtained additional experimental data used in the preparation of this
paper. These data demonstrate the applicability of the SEVNB method to different
ceramic materials.
Experimental
Materials
Hot isostatically pressed silicon nitride ceramics (Si3N4)[11] and sintered

partially-stabilized zirconia ceramics (Y-PSZ) [12] were the major materials used in
the process of verifying the experimental procedures. For high-temperature
investigations, reaction-bonded and sintered ShN4 + 30% SiC + 3% MgO [13] and
hot-pressed (in graphite molds) SiC + 50%ZrB2 + 10% B4C [14] ceramic particulate
composites were also used. Gas-pressure sintered silicon nitride (GPSSN) as well as
sintered alumina (AHO3-998) and sintered silicon carbide (SSiC) [6] analyzed in the
RRFT'97 also became objects of the experiment. Some data on these ceramics are
given in Table 3.
Test Specimens
Rectangular beams (3mm x 4 mm cross-section) V-notched (Figure la) in two

stages were used as test specimens. A group of 2 to 6 test specimens (3 in Figure lb)
were glued onto a ceramic plate (1 in Figure lb) and then V-notched. Two additional
specimens (2 in Figure lb) were fixed on both sides of the test group to protect the
edges from the edge effects due to notching. A preliminary notch 0.8 mm deep
(specimens for four-point flexure) or 1.6 mm deep (specimens for three-point flexure)
was cut in the group by a 100 - 300 ~m thick diamond disk. In the next stage, the V-
notch was manually (convenient for comparatively "soft" ceramics) or mechanically
(necessary for "hard" ceramics) polished-out with a 100 to 300 p.m thick razor blade
(4 in Figure lb) after filling the preliminary notch with diamond paste of grain sizes
ranging from 2 to 7~tm. In the case of manual polishing-out, the walls of the
preliminary notch (1.6 mm deep) served as the razor blade guides. The loading
configuration determined the number o f notches (one or three) cut in a 45-mm long
specimen (Figure 2). The criterion for the proper V-notch sharpness is its width, S,
which is equal to two root radii [6]. The notch length and radius were measured with
optical and scanning electron microscopes (x 500 or better).
Table 3 - Characteristics o f tested ceramics
Densit~r Strength at 20 ~ Hardness Method of

Material (kg/cm~) (MPa) (GPa) processing
Si3N4 3 140 700 14.1 HIP
GPSSN I 3 230 >920 13.5 Gas-pressure
sintered
SSiC2 3 150 - Sintered
A1203-9981 3 860 350 19.3 Sintered
Y-PSZ 6 050 425 12. l IP + sintered
Materials of RRFT'97.
z Materials of RRFT'96 and RRFT'97.
Figure 1 - V-notch with a 5 IBn root radius on a Y-PSZ specimen (a) and its manual
polishing-out (b) : ceramic plate (1), additional specimens (2),test specimen group (3),
and razor blade (4)
GOGOTSI ON CERAMIC PARTICULATECOMPOSITES 203
(a) (b) (c)

Figure 2 - SEVNB specimens: a, b - three-point flexure; c - four-point flexure
Procedures
The specimens were tested in three-point (16 mm span betwoan the bearing rollers)
or four-point flexure (20 to 40 mm spans between the bearing rollers) on a custom-
made test fixture [15], mounted on a testing machine that was fitted with a hard load
cell, a system for precision displacement of a loading rod, and loading supports. The
latter are equipped with an attachment for the rotation o f the bearing rollers during
specimen loading. Three-point flexure required the precise alignment of a specimen
on the hearing rollers, with the axis o f the central bearing roller and the radius o f the
V-notch root being in the same plane as the applied load. In high-temperature tests,
we used a test fixture [16] which is similar to that used in room temperature tests. For
the recording o f the load--deflection curves, a high-sensitivity LVDT-based
deflectometer was suspended (Figure 3) on the test specimen and was in no way
connected to the testing arrangement [17]. With this LVDT system, there is no
necessity to recalibrate the deflectometer before each test. The LVDT was located
outside the heating ehamher in the high-temperature tests. In all the experiments, the
speed o f the testing machine crosshead was constant and equal to 0.5 ram/rain. Load
cell and deflectometer readings were recorded with a coordinate potentiometer.
Load Deflectometer
Sp~
Figure 3 - Loading scheme for three-point flexure

Fracture toughness values were calculated in accordance with ASTM Standard

Test Method for the Determination o f Fracture Toughness of Advanced Ceramics at
Ambient Temperature (C 1421-99) for the three-point flexure and in accordance with
DIN Testing o f Advanced Technical Ceramics; Determination of Fracture Toughness
(51 109) for the four-point flexure.
Additional procedures for evaluating fracture toughness and other mechanical
characteristics are described elsewhere [8 and 15].
A micro-Raman imaging microscope with a 514.5 nm excitation line o f an argon
laser (x 100 objective and a ~l-tun diameter spot) was used to examine the surfaces
of fractured Y-PSZ specimens by a procedure described in Ref. [18].
The results o f comparative three- and four-point flexure tests are summarized in
Table 4, where the data obtained within the RRFT'97 program are also cited. The
variation o f fracture toughness values as a function o f notch root radius was studied
for Y-PSZ specimens annealed at 1400 ~ and Si3N4 specimens (Figure 4). The crack
nucleation in Y-PSZ specimens is shown in Figure 5. Load-deflection diagrams for
V-notched specimens are presented in Figure 6. The results ofmicro-Raman analysis
are given in Figure 7. The comparative data on SEVNB and SENB results are
summarized in Table 5. High-temperature test results are cited in Table 6.
Table 4 - Comparison of fracture toughness test results (MPa ~nl/z)
Three-point F o u r - p o i n t flexure
Test method flexure (a/We0.2-0.3)
(a/W~0.5) our results RRFT'97 results
Si3N4 5.50+0.07 (5)" 5.35+0.16 (5) -
GPSSN 5.30+0.04 (5) 5.20• (5) 5.36• (129)
Si3N4+30%SiC+3%MgO 2.27• (5) 2.40• (5)
SSiC 2.45 (1) 2.66-+0.20(4) 2.61• (56)
SiC+50%ZrB2+I0%B4C 3.70• (3) 3.53• (3)
AI203-998 3.50-+0.05 (5) 3.60-+0.06(5) 3.57• (135)
Y-PSZ 5.70+0.17 (5) 5.90:t:0.19 (5) -
The number ofspeci'mens tested (in parentheses).
It is useful to start the analysis with emphasis on the fracture toughness values
obtained in four-point flexure (Table 4). The data presented in this table demonstrate
good agreement between our results and the average results o f RRFT'97 [6], which
confirms sufficient accuracy o f our test procedures. The comparison of the data
presented in Table 4 also points to the fact that essential differences between three-
and four-point flexure results are absent. A similar conclusion was also made
elsewhere [4] for ceramic matrix composites. Consequently, both test methods might
be considered identical. Moreover, four-point flexure can be more easily appfied in
practice because it does not require a precise placement o f specimens on the bearing
rollers, which is difficult to achieve without a trained operator. On the other hand,
10 100 10 1~
. . . . . . , . . . . . . . . i
11 . . . . . . . . . . . . . . . . . . . . . .
~10~
81 /
7 . " ""
. . . . . . ~ . . . . . . . . .
10 100 4 10 1~
V-notchrootradius(~m) V~rootradim(~rn)
Figure 4 "Effect of V-notch root radii on thefracture toughness valuesfor Si3N4(a)

and Y-PSZ (b) ceramics
three-point flexure tests can utilize smaller specimens, which is advantageous for
materials science research.
The analysis o f results (Figure 4) shows that a decrease in the V-notch radius o f a
specimen leads to a decrease in the fracture toughness values for Si3N4and Y-PSZ
ceramics (a similar relation was also observed for other materials [7]). It is interesting
to note that the fracture toughness-notch root radius curve for Si3N4(Figure 4a)
flattens in the vicinity o f a small notch root radius as compared to that of the Y-PSZ
(Figure 4b). Such an effect is probably determined by different sensitivity o f these
materials to stress concentrations because o f differences in their grain sizes (-4 ~m
for Si3N4 and > l ~ for Y-PSZ) as was mentioned elsewhere [20]. It should be
emphasized that only for notch root radii less than 5 to 71~'n [11] were the fracture
toughness values for Y-PSZ ceramics in agreement with SEPB data (similar behavior
is also typical o f fine-grained alumina [19]). A similar conclusion follows from [6],
where an attempt was made to relate the V-notch radius values required for the correct
determination of fracture toughness values to the averaged values o f this parameter
obtained in the RRFT'97 studies. As follows from Figure 4a, the fracture toughness
values for Si3N4ceramics decrease with the decrease in notch radii down to about 30
iJ,m, which does not contradict the results ofsintered SiC whisker-reinforced Si3N4 [4]
tests. A similar tendency is also typical of SiC ceramics [20].
Examination of the specimens fractured in the tests (their results are presented in
Figure 4b) revealed a fracture crack (e.g. Figure 5) that propagated from the points
where "additional" stress concentrators were present. This observation also confwms
the assumption that the fracture o f a loaded ceramic specimen starts from a small
crack ahead of a machined notch root [20]. In this context, it is interesting to mention
that fracture toughness magnitudes are influenced by the sharpness of a notch root
rather than by its shape [11].
Figure 5 - Crack nucleation patterns for Y-PSZ specimens with V-notch root radii
19 Inn (a) and l OOl~n (b)
Table 5 - Fracture toughness data obtained by the SEVNB and SENB methods
Fracture toughness, MPa m ~ Coefficient Brittleness Type of

Material measure flexure
SEVNB SENB ~ X (points)
Y-PSZ 5.14_+0.29 (5) 9.54_+0.47(5) 0.538 1 four
Industrial glass 0.66+0.07 (6) 1.18-+0.09 (7) 0.562 I four
SiC+50%ZrB2+ 3.70+0.08 (8) 6.24_+0.37(6) 0.593 1 three
10%B4C
Si3N4 5.17_+0.06 (5) 9.12_+0.29(5) 0.567 1 three
SSiC 2.61+0.18 (5) 4.42_+0.232 (5) 0.590 I four
Si3N4+30%SiC+ 2.27_+0.14 (4) 2.49_+0.16(4) 0.920 0.88 three
3%MgO
Mg-PSZ 9.44_+0.12 (5) 10.2_+0.27 (5) 0.925 0.41 four
(TS-Grade)
La0.2C80.8Co033 2.2 (2) 2.2 (2) 1 0.25 three
Width of notch is 0.2-0.3 mm.
2 Data from [5].
3 Data from [21].
The SEVNB method (see, e.g., [6]) is usually compared with the SEPB, SCF, and
CNB methods. But it is also feasible to compare SEVNB and SENB data (Table 5),
paying attention to the fact [22 and 23] that the evaluation o f inelastic materials gives
fracture toughness results which practically coincide with those for the specimens
with sharp and blunt stress concentrators. Table 5 summarizes the index o f sensitivity
to stress concentrations, q~, equal to the ratio o f the fracture toughness values obtained
by the SEVNB and SENB methods, and the inelasticity o f ceramics described by the
brittleness measure X [23], the latter being equal to the ratio o f the specific elastic
energy accumulated in ceramics at the moment o f fracture to the total energy spent for
its deformation. Brittleness measurement values were determined from stress-strain
curves obtained in four-point flexure o f solid (without stress concentrators)
specimens. The analysis o f these data confirms the existence o f a relationship between
q~and g. In the tests of elastic materials (x=l), q~ is about 0.6, and for inelastic
materials (~<1), r exceeds 0.9.
Almost all the studies on the deformation behavior o f V-notched ceramic
specimens with a q~value o f about 0.6, produced linear load-deflection diagrams or
diagrams with small nonlinearity (e.g., curves 1 and 3 in Figure 6), which is
associated with comparatively little crack growth that is permissible in accordance
with ISO Test Method for Fracture of Monolithic Ceramics at Room Temperature by
Single Edge Precracked Beam (SEPB) Method (DIS 15 732).
But in some tests of composite particulate ceramics, the situation was different
[24], especially if the sharpness o f a V-notch was not uniform (V-notch mot radii on
the opposite sides of the specimen are not equal). Several load-deflection diagrams
exhibited an unusual shape because fracture initiated in the vicinity o f the notch root
with a smaller radius, where higher stress concentrations were present.
a
60- 60- b
Z 40 ~40
t~
O
~ 21) ,.z 20-
ction
specimen .
2 4 6 8 10 12 14 16 18 0 2 4
Deflection (gin) Deflection (p.m)
Figure 6 - Load-deflection diagrams for notched Si3N4+30% SIC+3% MgO

(a) and SIC+50% ZrB2+ 10% B4C (13)specimens tested at room temperature (1, 3)
and at 1400 qC (2, 4)
To complete the analysis o f Table 5, we should note that in fracture toughness tests
ofTS-grade zirconia ceramics, the index ~pwas lower than unity. At the same time,
the comparison of SENB and SEPB data for these ceramics (if the annealing o f
specimens was not performed at temperatures exceeding the boundary o f monoclinic-
tetragonal transformation) could give a ratio of fracture toughness values more than
unity. For example, this ratio was 1.24 [25], which was probably caused by the phase
transformation in the area o f the initial crack nucleation, when the specimen was
prepared for SEPB. Therefore, it was interesting to investigate the phase state on the
fracture surfaces o f zireonia specimens tested by the SEVNB method. The phase state
was analyzed by micro-Raman spectroscopy (Figure 7). Photo micrographs of
fracture surfaces were obtained by an environmental microscope (Figure 8). R is clear
from this figure that the surface o f a V-notch is smoother than others. As the analysis
o f Raman spectra (Figure 7) demonstrates, in contrast to fracture or saw notch
surfaces (Figure 7a) or the area of indentation (Figure 7b), a polished V-notch surface
(4 in Figure 7a) does not contain the monoclinic phase (there is no effect inducing the
phase transformation). This is an important advantage o f the SEVNB method in
Table 6 - High - temperature fracture toughness test results

Fracture toughness by the SEVNB method,
Material MPa.m v2
20 ~ 1300 ~ 1400 ~
Si3N4 5.5_+0.1 (5) 42_-_H).3(5)
Si3N4 [4fl 5.6_+0.52 5.0-20.4
SiC+50%ZrBE+10%B4C 3.4_+0.1 (8) 3.6_+0.3 (3) 3.7_+0.1 (3)
Si3N4+30%SiC+3%MgO 2.3-+0.1 (4) - 2.7_+0.1 (4)
The notches were produced by a diamond saw with a V-shaped tip.
2 The quantity of tested specimens is not known.
300 b
30-
250
g
x 200
20"
O
150 ,3
10"
100
I
I. I I I
14"0 16"0 180 20() 200 460
Raman shift (cm"l) Raman shift (cm"l)
Figure 7 - Raman spectra of Y-PSZ specimens fractured by the SEVNB method (a)
and by indentation (b) : nonfractured specimen surface (1), diamond saw surface (2),
fracture surface (3), razor blade surface (4), indentation edge (5), and indentation
center (6)
Figure 8 - Micro-photos of a Y-PSZ specimen (see also Figure 7) fractured by the

SEVNB method: diamond saw surface (2), fracture surface (3), and razor blade
surface (4)
comparison with other f~acture toughness test methods for zirconia and similar
ceramics.
The comparison o f high-temperature l~acture toughness measurements for Si3N4
ceramics, presented both in Table 6 and in Ref. [4], reveals their similarity. However,
in our case, in contrast to [4], damage o f the V-notch surface did not occur and we did
not observe active oxidation in the vicinity of the notch. It is noteworthy that in SEPB
tests of similar ceramics [4], oxidation induced the blunting (healing) o f the initial
sharp crack and, as a consequence, the fracture toughness results varied.
It is necessary to emphasize the increase in the l~acmre toughness o f ceramic
particulate composites with test temperatures (Table 6). The q~value varies
moderately with this increase (0.81 for Si3N4+ 30% SiC + 3% MgO and 0.71 for SiC
+ 50% Zrl~ + 10% B4C). The changes in the load-deflection diagrams of these
ceramics with test temperatures are not so pronounced (Figure 6) and are connected
with a certain increase in their inelastic deformation. It should be noted that the load-
deflection diagrams for Si3N4 + 30%SIC + 3% MgO unnotched specimens and
specimens with wide notches were linear. The oxidation layers on fractured
specimens are not strong and cannot affect the fracture toughness o f the ceramics
studied in high-temperature SEVNB tests. And, probably, the blunting of a stress
concentrator does not occur as in [4], where in SEPB tests crack healing and a
considerable increase in fracture toughness values, even at 1200 ~ were observed.
Conclusions
The test data confirm that the SEVNB method can be easier applied in practice and
can be used for the majority of advanced ceramics and ceramic particulate composites
at different temperatures and in an oxidizing environment. The SEVNB data are
independent of the flexural loading configuration and exhibit small scatter. For this
reason, it commands the attention of engineers involved in both certification testing
and materials science research. It was found that the ratio between SEVNB and SENB
data equaled about 0.6 for elastic materials and over 0.9 for inelastic ones. The
polishing-out of a V-notch root does not damage the surface layer formed on a
material; therefore, there are no residual stresses, phase transformations or other
effects, which often accompany the formation of stress concentrators by other
procedures.
Acknowledgments
Thanks are expressed to Mr. V.Galenko and Mr. B. Ozersky (IPS, Ukraine) for their
assistance in performing experiments. The author is indebted to Dr. G. Quirm (NIST,
USA) for his proposition to take part in such an investigation. The ceramics were
kindly supplied by Dr. A. Vasiliev and Dr. O. Grigoriev.(IPMS, Ukraine), and Mr. J.
Kiibler (EMPA, Switzerland) who also helped with the methodology of the
preparation of V-notched specimens.
References
[1] Quinn, G. D., Salem, J., Bar-on, I., Cho, K., Foley, M., and Fang H., "Fracture
Toughness of Advanced Ceramics at Room Temperature", dournal of Research
of the National Institute of Standards and Technology, Vol. 97, 1992, pp. 579-
607.
[2] Mizuno, M. and Okuda, H., "VAMAS Round Robin on Fracture Toughness of
Silicon Nitride," Journal of the American Ceramic Society. Vol. 78, 1995, pp.
1793-1801.
[3] Quinn, G. D., Ktlbler, J. J., and Gettings, R. J., "Fracture Toughness of
Advanced Ceramics by the Surface Crack in Flexure (SCF) Method: a VAMAS
Round Robin," VAMAS Technical Report No. 17, National Institute of
Standards and Technology, Gaithersburg, biD, June 1994.
[4] Mizuno, M., Nagano, Y., Cao, J.-W., and Kon, J.-l., "VAMAS Round Robin on
Fracture Toughness Measurements o f Ceramic Matrix Composite," in
h . . .
Abstracts, 9t Internatsonal Conference on Modem Materials & Technologzes
(CIMTEC'98), Florence, Italy, June, 1998, p. 108.
[5] Primas, R. J. and Gstrein, R., "ESIS TC 6 Round Robin on Fracture
Toughness," EMPA Report No. 155'088, EMPA, Swiss Federal Laboratories
for Materials Testing and .Research, Dubendors Switzerland, October 1995.
[6] Ktibler, J., "Fracture Toughness of Ceramics Using the SEVNB Method: Round
Robin," VAMAS Report No.37/ESIS Document D2-99, EMPA, Swiss Federal
Laboratories for Materials Testing and Research, Dubendor~ Switzerland,
September 1999.
[7] KUbler,J., "Fracture Toughness of Ceramics using the SEVNB Method:
Preliminary Results", Ceramic Engineering & Science Proceedings, J.P. Singh,
Ed., American Ceramic Society, Vol.18, Issue 4, ISSN 0196-6219, 1997,
pp.155-162.
[8] Gogotsi, G. A. and Galenko, V. I., "Comparative Analysis of Fracture
Toughness Tests Methods for Ceramics and Crystals at Room and Lower
Temperatures," Strength of Materials, Vol. 29, 1997, pp. 287-297.
[9] Awaji,H. and Sakaida, Y., "V-Notch Technique for Single-Edge Notched
Beam and Chevron Notch Method," Journal of the American Ceramic Society,
Vol. 76, 1990, pp. 3522-3523.
[10] Dameny, D.J. and Danzer, A., "Method for Fracture Toughness Testing of
Ceramics- Ready for Standardisation", Fracturefrom Defects, M.W. Brawn,
E.R., L.Riosand, and K.J.Miller, Eds., Emas Publishing, 1998, p 491-496.
[ll] Gogotsi, G. A, "Fracture Toughness of Ceramics and Ceramic Matrix
Composites (SEVNB Method)," Refractories and Technical Ceramics, Vol.
11-12, 1998, pp. 7-13.
[12] Gogotsi, G. A., "Fracture Toughness Tests of V-Notched Specimens," Strength
of Materials, Vol. 32, 2000, pp. 170-177.
[ 13] Gogotsi, G. A. "Several Experimental Results of High Performance Ceramics
used in Heat Engine Components", lPS AN USSR, Kiev, 1983 (in Russian).
[14] Grigoriev, 0. N., Gogotsi, G. A., Gogotsi, Yu. G., Suhbotin, V. I., and
Brodnikovskiy, N. P., "Development and Properties of SiC-MeB Ceramics,"
Journal of the European Ceramic Society (submitted for publication).
[15] Gogotsi,G. A., "Deformation Behaviour of Zirconia," Strength of Materials,
Vol. 30, 1998, pp. 638-644.
[16] Gogotsi, G. A. and Ostrovoy, D. Yu., "Deformation and Strength of
Engineering Ceramics and Single Crystals," Journal of the European Ceramic
Society, Vol. 15, 1995, pp. 271-281.
[17] Gogotsi, G.A., Zavada, V.P., and Nerodenko, V.I., USSR Patent 1,224,551, !5
Des. 1984.
[18] Gogotsi, G.A. and Lomonova, E.E., "Micro-Raman Spectroscopy Study of
Indentation-Induced Phase Transformation in Zireonia Crystals," Refractories
and Industrial ceramics, V.41, 2000, pp. 191-195.
[19] Nishida, T., Hanaki, Y., and Pezzotti, G., "Effect of Notch-Root on the
Fracture Toughness of a Fine-grained Alumina," Journal of the American
Ceramic Society, Voi. 77, 1994, pp. 606-608.
[20] Damani,R., Gstrain, 1L, and Danzer, R., "Critical Notch-Root Radius Effect in
SENB-S Fracture Toughness Testing," Journal of the European Ceramic
Society, Vol. 16, 1996, pp. 695-702.
[21 ] Klevelend K., Orlovskaya N., Grande T., Moe A.-M., Einarsrud M.-A., Breder
K., and Gogotsi G., "Non-Elastic Behaviour of LaCoO3-Based Ceramics,"
Journal of the American Ceramic Society (submitted for publication).
[22] Davidge, R. W.and Tappin, G., "The Effective Surface Energy of Brittle
Materials," Journal of Materials Science, Vol. 3, 1968, pp. 165-173
[23] Gogotsi, G. A., "Deformational Behaviour of Ceramics," Journal of the
European Ceramic Society, VoL 7, 1991, pp. 87-92.
[24] Gogotsi, G. A., "Fracture Toughness of Ceramics and Ceramic Composites

(SEVNB Method)," in Abstracts, Sixth Conference & Exhibition of the
European Ceramic Society, Brighton, UK, July 1999, p. 225.
[25] Gogotsi, G. A, Drozdov, A. V., Zavada, V. P., and Swain, M. V., "Comparision
of the Mechanical Behaviour of Partially Stabilised Zirconia With Yttria and
Magnesia," Journal of the Australian Ceramic Society, Voi. 27, 1991, pp. 37-
49.
J. A. Salem I and M. G. Jenkins 2
The Effect of Stress Rate on Slow Crack Growth Parameter Estimates
Reference: Salem, J. A., and Jenkins, M.G., " T h e Effect of Stress Rate on Slow Crack
Growth Parameter Estimates," Fracture Resistance Testing of Monolithic and
Composite Brittle Materials," ASTM STP 1409, J. A. Salem, G. D. Quinn, and M. G.
Jenkins, Eds., American Society for Testing and Materials, West Conshohocken, PA,
2002.
Abstract: Dynamic fatigue or constant stress-rate testing is a convenient method for

generating SCG (slow crack growth) data on ceramics and glasses. The tests are simple,
inexpensive, and accomplished quickly. However, the use of dynamic fatigue testing to
generate slow crack growth data for SCG parameter estimation assumes that the test
specimen lifetime is dominated by a single region of the SCG curve. This assumption
breaks down when excessive stressing rates are used and significant portions of the crack
growth occurs in all three regions of the SCG curve. The resultant averaging of data
representing all three regions of the SCG curve results in an error in the estimated SCG
parameters n and D, and such data may not be accurate for predicting the lifetime of
industrial components.
Dynamic fatigue tests covering five orders of magnitude well below the inert strength
were performed on an alumina. The tests were performed using ASTM C 1161 "B"
flexure specimens in accordance with ASTM Test Method for Determination of Slow
Crack Growth Parameters of Advanced Ceramics by Constant Stress Rate Testing (C
1368-97). Statistical analysis of the experimental results indicates a statistically
significant effect of stress rate on the SCG parameter n for stress rates above several
MPa/s. The experimental results are in agreement with analytical results of Sudreau et al.
[1] that show significant effects for rates greater than approximately one MPa/s.
Recommendations for dynamic fatigue testing are given with regard to the current
procedures in ASTM C 1368-97.
Keywords: dynamic fatigue, slow crack growth, ceramics, strength, life prediction, stress
rate, stress corrosion, reliability
Materials Research Engineer, Life Prediction Branch, NASA Glenn Research Center, MS 49-7,
21000 Brookpark Rd., Cleveland, OH, 44135.
2 Professor, University of Washington, Box 352600, Seattle, WA, 98195.
213
Nomenclature
A Slow crack growth material constant

B Slow crack growth material constant
D Slow crack growth parameter
DOF Degrees of freedom
F Statistic for comparing standard deviations
Kl Stress intensity factor
Ktc Fracture toughness
N Number of specimens tested
P Applied load
i, Applied load rate
So I.~ad span
si Support span
SDi Standard deviation of the ith variable
b Specimen thickness
ms Weibull modulus (surface flaws)
n Slow crack growth exponent
t Statistic for comparing means
V Crack velocity
W Specimen width
Ol Slope of log fracture stress versus log stress rate curve
Intercept of log fracture stress versus log stress rate curve
Inert strength
a: Measured fracture strength
cro Characteristic strength
(r Applied stress rate
Mean strength
Introduction
Slow crack growth or "static fatigue" of ceramics and glasses is an environmentally

induced, chemically activated atomic reaction that is enhanced by stress and temperature.
Crack growth occurs by breakage of the bonds at the crack tip by dissociative
chemisorption or ionic solvation [2]. In particular, the introduction of water to the
environment can cause crack growth in ceramics and glasses subjected to stress, and
variations in the moisture in air are sufficient to affect the SCG parameters. Macroscopic
observations of the behavior of cracks extending due to environmentally induced SCG
indicate three regions of growth, as exhibited in Figure 1. In Region I the active species
has ready access to the crack tip. In Region H the crack growth rate depends on the
diffusion rate of the active species to the crack tip, and is thus strongly dependent on
concentration changes in the environment and mechanical effects
SALEM AND JENKINS ON SLOW CRACK GROWTH PARAMETERS 215
i
i
RegioIIIn i
O
o RegionII //J. -~~
O
Average
cu o
~/}
L)
I:))
O
.-I
,o0e ~Regio
........ e o o n eS
,n
I Environment
K,h Kk
Log Stress Intensity Factor, K1
Figure 1 - Generalized crack velocity as a function of stress intensity factor on a

logarithmic scale for ceramics and glasses exhibiting environmentally induced slow
crack growth.
that limit exposure of the crack tip, such as grain bridging and crack size. In Region IIl
the crack growth rate becomes sufficiently fast that it is independent of the environment
chemistry. Conceptually the maximum fracture resistance and "inert" strength (i.e., the
strength under noncorrosive conditions) of the material has been reached. The existence
of a slope in Region I]] implies that the strength in an inert environment is a function of
stress rate, and that stable extension occurs in the absence of stress corrosion. Stable
crack extension in the absence of stress corrosion is dependant on crack shape and
loading geometry, and thus Region I]/is likely to be test specimen and method dependant.
Region I dominates the time-to-failure during long lifetimes, and design and test
methodologies attempt focus on it.
A variety of test methods exist for generating SCG data on ceramics and glasses.
The methods can be classified into strength based and fracture mechanics based
approaches. Strength-based methods employ smooth test specimens, such as flexural
beams or tensile specimens, and estimate SCG material properties from strengths
measured over different time intervals. Loading is generally done in a static fashion (i.e.,
"static fatigue") or in a continuously increasing fashion (i.e., "dynamic fatigue"). These
methods are thus practical, as the tests are simple, inexpensive, and accomplished
quickly.
Strength based methods directly sample the preexisting flaw distribution within or
on the surface of the test specimens. Thus the cracks develop from at least some of the
same sources that are expected to cause failure in a component manufactured from the
Inert
Environment
6 TM
t
tE Average RegionlIl
,= Curve . S
.= / ~ ' RegionII
I.l.
~ / , ~ RegionI Dominated
O)
O
._1
Sub-Threshold Slope= l/(n+l)
Region
Log Stress Rate, d
Figure 2 - Generalized fracture strength as a function of stress-rate on a logarithmic

scale for ceramics and glasses exhibiting environmentally induced slow crack growth.
same material. National and international standardization of slow crack growth and
cyclic fatigue of ceramics has revolved around the use of smooth specimens (e.g.
Determination of Subcritical Crack Growth Parameters from Constant Stressing Rate
Flexural Strength Tests (ENV 843-3), and Testing Method for Bending Fatigue of Fine
Ceramics (JIS R 1621-95)), and the dynamic technique has been standardized by A S T M
and CEN (European Committee for Standardization).
The disadvantage of strength-based methods is that the results are subject to the
scatter inherent in the strength distribution of the material, and thus a relatively large
number of tests are required for accurate parameter estimation. Further, strength based
methods do not measure the actual crack size, but infer it from the measured strength. As
a result, all three regions of the slow crack growth curve are averaged in the experimental
measurements [1,3] as shown in Figures 1 and 2. The current standards either assume
that all the crack growth can be described by a Region I model, or they do not give
guidance on calculating the SCG parameters.
Sudreau et al. [1] demonstrated that the averaging of the SCG regions could lead to
large errors in the SCG parameters A and n when estimated from dynamic fatigue data
measured with fast stress rates. This is particularly the case for material/environmental
combinations that result in a significant Region I / o f the SCG curve. They used a fracture
mechanics test method to determine SCG parameters for all three regions of the SCG
curve and thereby were able to simulate the effect of stress rate on the slope of the
dynamic fatigue curve. Their predictions indicate a shift in the slope at 1 MPa/s for a
mullite. Actual dynamic fatigue testing of the mullite at excessive rates (0.04 to 30
MPa/s) results in estimates ofA = 0.3 x 10 .8 and n = 19, whereas the actual Region I
parameters, as measured with the double torsion method, are A = 0.3 x lff 15 and n = 41.
For a silicon nitride the measured dynamic fatigue parameters were A = 6.8 x 10-96 and n
= 100, whereas the actual Region I parameters, as measured with a fracture mechanics
technique, are A = 4.4 x 10-63 and n = 67.
The large differences, though conservative, are troubling from the standpoint of
standardization because dynamic fatigue results may not only be incorrect, but will be a
function of the particular stress rates used by a particular laboratory. The results also
imply that the use of excessive preloads will shorten the time the crack spends in Region I
of the SCG curve, further skewing the SCG parameters.
Currently, the Test Method for Determination of Slow Crack Growth Parameters of
Advanced Ceramics by Constant Stress Rate Testing (C 1368-97) allows stress rates less
than 2000 MPa/s and gives no cautions or recommendations regarding the interaction
between rates and environment. The objective of this work was to determine if a
statistically significant effect of stress rate on the dynamic fatigue curve could be detected
in data run in accordance with A S T M C 1368-97. The experimental demonstration of a
change in the slope of the dynamic fatigue curve was considered important as fracture
mechanics techniques and strength based techniques can give different slow crack growth
parameters for reasons besides the averaging of the S C G curve regions (e.g., long crack
versus short crack effects). More than 120 dynamic fatigue tests covering five orders of
magnitude in stress rate were performed for the study.
Experimental Procedures
Specimen Removal and Preparation
The flexural test specimens were cut from two alumina RF (radio frequency)
windows that were manufactured for used in advanced accelerator systems. The test
specimens were machined in accordance with Test Method for Flexural Strength of
Advanced Ceramics at Ambient Temperature (C 1161), and cut so that the tensile surfaces
were the "as-received" window surfaces; however, the bevels were longitudinally ground
per ASTM C 1161.
These test specimens were part of a larger study of three circular RF windows [4].
For that study, test specimens were removed in both radial and tangential orientations in
order to estimate the effects of grinding orientation, which was nominally in the radial
direction, on the strength and SCG properties of the windows. In addition to testing
specimens with the as-received surface of the windows, specimen sets from the windows
were subjected to grit blasting, which has been noted to improve window life, to ascertain
any change in the strength or SCG parameters. These results have been reported
elsewhere [4]. The specimens were identified with the pattern illustrated below:
I D2TAS1 I
Table 1 - Strength statistics and 95% confidence bands for uncensored flaw populations.
The strengths were measured per ASTM C 1161 "B" specimens in four-pointflexure.
Window Group Number Mean Strength, 5 ,

Number Designation Tested + Standard Deviation
Lower Estimated Upper
95% Value 95%
Band MPa Band
D1TAS 13 227 237 + 16 247
D 1TGB 20 228 234 + 13 240
D2TAS 15 230 239 + 17 248
Table 2 - Weibull strength statistics and 95% confidence bands for uncensored flaw
populations.
Group Number
Designation Tested Weibull Modulus, ms Characteristic Strength, o0
Lower Estimated Upper Lower Estimated Upper
95% Value 95% 95% Value 95%
Band MPa Band
D1TAS 13 9 13 20 234 245 255
D1TGB 20 13 18 25 231 237 243
D2TAS 15 12 18 27 239 246 253
where the first two digits (e.g., "D2") give the window number (i.e., window 2), the third
digit (e.g., "T" or "R") gives the specimen orientation (i.e., tangential or radial), the forth
and fifth digits (e.g., "AS" or "GB") give the tensile surface condition (i.e., as-received or
grit-blasted) and the last digit gives the test specimen number.
The mean strengths and Weibull statistics of the two billets used in this study are
compared Tables 1 and 2. No statistically significant difference between the strengths
and Weibull moduli of the windows could be determined for the as-received condition.
Also, no statistically significant difference in strength, Weibull moduli or SCG
parameters could be determined between the as-received condition as compared to the
grit-blasted condition. As such is the case, the billets are not considered to be not
statistically significantly different for the purposes of this study.
Elastic Modulus
The elastic moduli of some of the test specimens were determined with the impulse
excitation of vibration method in accordance with ASTM Standard Test Method for
Dynamic Young's Modulus, Shear Modulus, and Poisson's Ratio for Advanced Ceramics
by Impulse Excitation of Vibration (C 1259-94). The mean and standard deviation of 21
test specimens was 330 + 3 GPa.
SCG Testing and Parameter Estimation
Testing of the "B" specimens was conducted in four-point flexure with 20 m m load
and 40 nun support spans in ambient laboratory air at stress rates ranging from 0.0003 to
30 MPa/s. The fastest rate nominally corresponds to that specified by ASTM C 1161 for
measuring the materials flexure strength, and the slowest rate is two orders of magnitude
lower than the slowest typically used in dynamic fatigue testing. The specimen preloads
were less than 12% of the inert strength. This corresponded to about 20% of the failure
load for the slowest stress rate. The inert strength was determined by testing specimens
in silicon oil after baking at 115~ for more than 1 hour. In order to ensure a minimal
effect of adsorbed water on the inert strength, the specimens and oil were heated
simultaneously in the oven. After drying, the specimens were immediately placed in the
oil container while in the warm oven, and tested immediately on cooling. In addition to
using the inert environment, the stress rate was increased to 100 MPa/s to minimize
Region ID stable crack extension.
The stress rate as a function of load rate is
(r = 3 P(So - Si) (1)

2 bw 2
where 15 is the load rate, b is the specimen thickness, w is the width, and So and Si are the
outer and inner spans. The use of different stress rates provides flaws in the material time
to grow to different sizes via stress corrosion prior to specimen failure, thereby indicting
the sensitivity of the material to the test environment. For ceramics and glasses, the slow
crack growth rate under Mode I loading above the SCG limit is generally expressed by the
following power-law relation [5]
da K1 ]n
v = -- = A [ (2)
dt K1c
where v, a, and t are crack velocity, crack size, and time, respectively. A and n are the
material/environment dependent SCG parameters, and KI and Klc are, respectively, the
Mode I stress intensity factor and the critical stress intensity factor or fracture toughness of
the material under Mode I loading. Dynamic fatigue testing employs a constant load or
stress rate. 6-, to fail ,test specimens, and the corresponding SCG fracture strength,a/, is
expressed as a function of stress rate as follows [5]
o : = [B(. + "+' (3)
where ~ is the inert strength, and B is a parametex associated with A, n, fracture toughness,
crack geometry and loading configuration. The SCG parameter n can he determined from a
plot of log ~ a s a ftmction of log 6- with Eq. 3 written as
1
log t7 / -- ~n + 1 l o g 6- + log D (4)
where
log D = 1 log[B(n+ l)aT-'] (5)
n+l
Once the slope and intercept of Eq. 4 are estimated, the parameters n, D, B and A,
and their standard deviation SD,, etc., are estimated from [6]
n = I_ I (6)
a
SD~, (7)
SD. ~2
D = 1013 (8)
so,, 2.3o2 (so , X o' ) (9)
B = ~ (lO)
alla'3
lO[J/a SD~ i
SDn = 1_-----~-.J(ln
- 10) 2 SD2# + (1 - 3or)2 ~ - T + ( Or- fl lnlO+lntri)2 SD2ot2 (11)
(o,)~
1
---3
2K~ (O'i) ~ 2K2
A = lOlJ/a ( 1 - 3 a ) y 2 = B( n - 2 ) y 2 (12)
I4SDZxlc + (lnlO~ SD 2 + ( 1 3 )2 SD2,
soTC__ c,_ , <,3,

~ - fl ln l O + ln ~ i - -3
where o~and fl are the slope and the intercept estimated by linear regression of Eq. 4, and
Y is the geometry correction factor for the stress intensity factor.
In order to mitigate any undetected effects of window-to-window variation, test
specimens from both windows were tested at both the ultra-slow and the typical stress
rates. This effectively blocks any billet-to-billet influences on the estimated slope. To
minimize orientation effects, only test specimens of radial orientation were used in this
study. In addition, to confirm repeatability of the results, test were repeated with different
load frames and load cells.
Results
SCG Parameters
Fracture strength as a function of stress rate is plotted in Figure 3. The data

generated at rates less than -0.3 MPa/s appears to fall on one line while the data
generated at rates greater than 0.3 MPa/s falls on a steeper line. The SCG parameters, as
determined with linear regression, are complied in Table 3 for data generated above and
below 0.3 MPa/s.
The statistics for comparing the significance of the differences in the slopes of the
SCG curves and the significance of the differences in standard deviations of slope are
calculated from [7]
~ , -~21 (14)
t~ 4SOZal + $D22
SD2I (15)
F~ SO~2
where SD~I is the greater of the two standard deviations. These statistics are complied in
Table 4. The statistic F is significant at more than 99% confidence, implying that the
standard deviations of the slopes in the two regions are significantly different and that the
DOF (degrees-of-freedom) cannot be pooled for determining the t statistic for any
significance level. The effective DOF needs to be calculated via the Aspin-Welch [7]
formula
Strain Rate, 1/s

3 e-9 3 e-8 3 e-7 3 e-6 3 e-5 3 e-4
300
o D1RAS
i i i i
' b
275 zx D2RAS
250
v
o
D2RGB
D2RAS Oil
/'~ (5)
-- Regression Line<_.0.3MPa/s
0_
.... Regression Line> 0.3 MPa/s (5) /
225
(20) /~ 0/ V
0
200
175 (19)
(5) ~'/ (10)
o~
O o a
LI. 0
O
150 V
125 ' ' ' '""1 . . . . . . . . i , , , ,,1111 , i 1, .... i . . . . . . . . i . . . . . . . . i
0.0001 0.001 0.01 0.1 1 10 100
Stress Rate, d , M P a / s
Figure 3 - Experimentally measured failure stress as a function o f stress rate f o r

alumina. The number o f specimens tested is given in parenthesis.
T a b l e 3 - Slow crack growth parameters estimated from data measured in water with
typical rates and in air with stress rates above and below 0.3 MPa/s. Values following +__
are one standard deviation.
Air Air Water

Parameter
Rates < 0.3 M P a / s Rates > 0.3 M P a / s Rates of 0.03 to 30 MPa/s
Ot 0.0207 + 0.0020 0.0620 + 0.0031 0.0243 + 0.0025
fl 2.2747 + 0.0048 2.2950 + 0.0034 2.2451 + 0.0037
n 47.3 + 4.8 15.1 + 0.8 40.1 + 4.2
D 188.2 + 2.1 197.2 + 1.6 175.8 + 1.5
DOF 52 48 18
(c 2 (1-4)-I
DOFf,= N I - 2 + N 2 - 2
(16)
where
SD~I/N,~j
c = (17)
so ,/uoj + soL/u 2
and N is the number of test specimens. This results in a somewhat lower degrees-of-
freedom for t than direct pooling of the DOF's for the rates above and below 0.3 MPa/s
(i.e., 80 instead of 100). Despite this, the t statistic of 2.07 at 99% confidence is
substantially smaller than that associated with the data (11.13), implying a statistically
significant difference in slope at much better than 99% confidence.
Effect of Stressing Rate on Strength Variation
Another aspect of using linear regression to calculate the SCG parameters is the
requirement that the data is normal and identically distributed. The transformation of
slow crack growth data into a log-log space should accomplish this requirement. This can
be examined by observing the standard deviation of the strength as a function of stress
rate in log space. Table 5 summarizes the standard deviations along with F statistics for
comparing the deviations.
The standard deviations are relatively similar and the F statistics small. No
statistically significantly difference can be detected at reasonable confidence levels (i.e., >
90%), implying little change in standard deviation of strength with stress rate in log-log
space.
Effect of Test Environment
If tests are conducted in a manner that minimizes or eliminates Region 11 of the

SCG curve, then Region I is extended toward Region I / / a n d the errors induced by
averaging can be minimized while stress rates approaching those specified in ASTM C
1368 are used. This will allow the timesaving benefit of dynamic fatigue to be realized
along with accurate parameter estimates for component life prediction.
This effect can be observed in the data of Sudreau: dynamic fatigue of mullite in
water at rapid rates results in A = 0.5 x 1015 and n = 36, whereas the actual Region I
parameters, as measured with the double torsion relaxation method, are A = 0.3 x 1015
and n = 41 for air andA = 0.5 x 10 15 and n = 43 for water. The dynamic fatigue results
for water are in good agreement with the actual Region I parameters. However, the
dynamic parameters generated in air (A = 0.3 x 10-8 and n = 19) are not because a
significant Region II exists and is averaged into the results.
Thus by using a 100% concentration of the corrosive medium (water), the diffusion
rate is increased, Region rl is minimized and more realistic SCG parameters are
determined for efficient stress rates. In order to experimentally verify the effect of
concentration for the alumina, additional tests were conducted in distilled water at rates
0.03 and 30 MPa/s. The resultant data is shown in Figure 4 and the calculated SCG
parameters are listed in Table 3. Note the relatively good agreement between the standard
deviations and the slopes of the ultra-slow rate data generated in air and the data
generated in water at typical rates. The significance of the small differences can be
determined by using Eqs. 14 and 16. The statistics are listed in Table 6 and imply that the
null hypotheses of equivalent standard deviations and equivalent slopes cannot be
rejected. Thus testing in water at typical rates results in a slope not significantly different
from that measured in air at ultra-slow rates for reasonable confidence levels.
Although no significant difference can be detected between the slopes, the level of
the curve for water appears to be lower (about 7% at 0.3 MPa/s). This is probably the
result of a longer Region I causing more crack growth and a larger crack size on
beginning Region Ill. The crack growth associated with stress intensities for Region 11 in
air is increased in water due to the higher concentration of corrosive medium.
Table 4 - Statistics for comparison of slopes of SCG curves estimated from data
measured in air with stress rates above and below 0.3 MPa/s. The null hypotheses for
the statistics F and t can be rejected with high confidence.
F 90% 95% 99%

Parameter DOF (Data) Confidence Confidence Confidence
Level Level level
F 48/52 2.29 1.60 1.75 2.10
t 80 11.13 1.66 1.99 2.64
Table 5 - Comparison of standards deviations in log space for various stress rates. The
null hypotheses for the statistics F cannot be rejected with high confidence.
Stress Rate Standard Rates F F F

MPa/s Deviation Compared DOF (Data) 95% 90%
36 0.0206 36 vs 0.00036 24/18 1.20 2.50 2.15
0.36 0.0211 0.36 vs 0.00036 19/18 1.26 2.58 2.20
0.00036 0.0188 0.36 vs 36 19/24 1.05 2.35 2.04

Strain Rate, 1Is
3 e-9 3 e-8 3 e-7 3 e-6 3 e-5 3 e-4

300 i t i i t
o D1RAS
275 (2s) ......
v D2RGB O .:
~2
zx D2RAS (5)
250 9 D2RAS Water
ca 9 D1RAS Water
fl_
225 9 D2RGB Water
o D2RAS Oil
200
.e
(5)
03
175
(19)
-q
.m
ca
ii
150 v
(10)
125 ' ' ''""l ........ i ' '''""l ' '''""l ' ''" . . . i. . . . . . . . i
0.0001 0.001 0.01 0.1 1 10 100
Stress Rate, d, MPa/s
Figure 4 - Experimentally measured failure stress as a function of stress rate for

alumina. The number of specimens tested is given in parenthesis.
Implications for Preloading
Preloading is a useful tool for reducing the test time required to fail a specimen [8].
The drawback is that large preloads may compound errors caused by excessively fast
stress rates because the starting stress intensity factor may be above that for Region I and
the resultant time spent in Region I of the SCG curve is thus truncated (see Figure 1). For
the data of Sudreau, Region I of the mullite ends at -65% of the maximum stress intensity
factors measured (i.e., the fracture toughness). For the silicon nitride, Region I ends at
-90% of the maximum stress intensity factor measured. Apparently the degree of
acceptable preloading depends on the shapes and sizes of the SCG regions. Experimental
examination of this effect is left to further study.
Table 6 - Statistics for comparison of slopes of dynamic fatigue curves estimated from
data measured in air with ultra-slow stress rates (i.e. <-0.3 MPa/s) and in water with
typical stress rates. The null hypotheses for the statistics F and t cannot be rejected with
high confidence.
F 90% 95% 99%

Parameter DOF (Data) Confidence Confidence Confidence
Level Level Level
F 18/52 1.47 1.81 2.02 2.51
t 70 1.14 1.67 2.00 2.66
Summary and Conclusions
The fatigue strength of an alumina was measured in air at stress rates between
0.0003 MPa/s and 30 MPa/s. The slope of the log trf versus log ct curve is statistically
significantly different above and below rates of -0.3 MPa/s, implying that multiple
regions of the SCG curve are active in dynamic fatigue measurements. As a result of the
averaging of the regions of the SCG curve in dynamic fatigue measurements, a significant
error in the estimated parameters for this alumina and a mullite [1] will occur if stress
rates greater than - 1 MPa/s are employed. This occurs because all three SCG regions
become significant at higher stress rates. A similar effect has been measured for a silicon
nitride [11.
However, if the concentration of corrosive medium, in this case airborne water, is
increased sufficiently to minimize Region 13, a substantial improvement in the estimated
parameters can be attained when stress rates greater than - 1 MPa/s are employed.
However, additional crack extension will occur prior to Region I11, and a slightly lower
strength level will result. Additional measurements and statistical tests are needed to
determine the significance of the differences in the intercepts for tests run in air at ultra-
slow rates and in water at typical rates. Conceptually, the intercepts should approach for
very slow stress rates. This is left to future work.
The strength as measured by ASTM C 1161 corresponds to a value in the upper end
of Region II or in Region III of the SCG curve. For this alumina the strength as measured
with ASTM C 1161 was within ~10% of the inert strength measured in oil after baking.
Currently, ASTM C 1368-97 allows stress rates and environments that will result in
significant errors in the fatigue slope as define by Eqs. 1 and 3 for a single region SCG
model. Based on the data in this study and the data of Sudreau [1], the highest allowable
rates in ASTM C 1368-97 are nearly two orders of magnitude too high for testing in air.
For accurate design of components that are expected to have a long time-to-failure (i.e.,
greater than several hours) in air, it is recommended that either stress rates below -1
MPa/s or a water test environment be used.
References
[ 1] Sudreau, F., Olaguon, C., and Fantozzi, G., "Lifetime Prediction of Ceramics:
Importance of Test Method," Ceramics International, 10, 1994, pp. 125-135.
[2] T.A. Michalske, B.C. Bunker and S.W. Freiman, "Stress Corrosion of Ionic and
Mixed Ionic/Covalent Solids," Journal of the American Ceramic Society, Vol. 69,
No. 10, 1986, pp. 721-724.
[3] Rockar E.M., and Pletka, B.J., "Fracture Mechanics Of Alumina In A Simulated
Biological Environment," in Fracture Mechanic of Ceramics Vol. 4, R. C. Bradt et al.,
Eds., Plenum Press, NY, 1978, pp. 725-735.
[4] Salem, J.A. "Fatigue Parameter Estimation For Alumina RF Windows Used In
Advanced Accelerator Systems" Report For Fulfillment Of Space Act Agreement 3-
184, NASA GRC, 1998; also K. A. Cummings, "Theoretical Predictions and
Experiment Assessment of the Performance of Alumina RF Windows," Los Alamos
National Laboratory LA-13466-T, July 1998.
[5] Ritter, J.E. "Engineering Design and Fatigue Failure of Brittle Materials," in Fracture
Mechanics of Ceramics, Vol. 4, R. C. Bradt, D. P. H. Hasselman, and F. F. l_ange, Eds.,
Plenum Publishing Co., NY, 1978, pp. 661-686.
[6] Salem, J.A., Jenkins M.G., and Keller D., "Estimating Standard Deviations Of Fatigue
Parameters For Ceramics Exhibiting Slow Crack Growth," accepted Journal of
Materials Science Letters, 2000.
[7] Acheson, D.J., Quality Control and Industrial Statistics, forth edition, Richard D.
Irwin, Inc., Homewood, Illinois, 1974, pp. 835-838.
[8] Choi, S.R., and Gyekenyesi, J.P., "Fatigue Strength as a Function of Preloading in
Dynamic Fatigue Testing of Ceramics and Glasses," ASME Journal of Engineering
For Gas Turbines and Power, Vol. 119, 1997, pp. 493-499.

PDF

Загружено:

Сведения о документе

Оригинальное название

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

PDF

Загружено:

Авторское право:

Доступные форматы

STP 1409

Fracture Resistance Testing

J. A. Salem, G. D. Quinn, and M. G. Jenkins, editors

ASTM Stock Number: STPI409

Authorization to photocopy items for internal, personal, or educational classroom use,

Peer Review Policy

Fracture Mechanics of Brittle Ceramics - 30 Years of Progrem--R.C. BRADT

IMPLICATIONS FOR DESIGN AND TESTING

Failure from Large Grains in PolycrystaBine Ceramics: Transitions in Fracture

Stresses in Ceramic Plates Subjected to Loading Between Concentric Rings--

Development, Verification, and Implementation o f a National Full-Consensns

Fracture Toughness of Ceramics using the SEVNB Method: From a Preliminary

The Fracture Toughness Round Robins in VAMAS: What We Have Learned--

R-Curve Measurement of Silicon Nitride Based Ceramics at Elevated Temperatures

Application of Quantitative Fractography to the Characterization of R-Curve

Fracture Testing of a Layered Functionally Graded Material--M.d. roLL,

UNIQUE MATERIALS AND ENVIRONMENTAL EFFF_~S

Environmental and Thermal Effects on the Toughness of TiCN-Based M a t e r i a l s - -

Fracture Toughness Studies on Ceramics and Ceramic Particulate Composites

Fracture Mechanics of Brittle Ceramics - 30 Years of Progress

Reference: Bradt, R. C., ~Fracture Mechanics of Brittle Ceramics - 30 Years of

Keywords: brittle fracture, ceramics, design, fatigue, R-curves, wake region

1Department of Metallurgical and Materials Engineering, The University of Alabama,

Fracture of Ceramics in the 1960s

Shortcomings of Structural Ceramics from the Perspective of Hindsight

The Absence of a Qualified Data Base of Materials Properties

Lack of Appropriate Standards for Property Measurements

Absence of a Design Methodology

Economic Factor, Prohibitive Costs

Contributions of Fracture Mechanics

Improved Structural Ceramics

Ceramics in Structural Applications

Summary and Conclusions

Failure from Large Grains in Polycrystalline Ceramics: Transitions in

Reference: Freiman, S., "Failure from Large Grains in Polycrystailine Ceramics:

Keywords: ceramics, fracture toughness, grain size, polycrystalline toughness, single

i Ceramics Division, National Institute of Standards and Technology9Gaithersburg, MD, 20899.

Failure from Large Grains in Polycrystalline Ceramics

As mentioned previously, strength is governed by failure from small flaws,

Fractographic analysis shows that failure of chemically vapor deposited ZnSe is

Figure 1 Fracture surface of chemically vapor deposited, polycrystalline ZnSe showing

o =r-'( 2er y'2c"2 (2)

FLAW SIZE {/~m)

I00 9 T E N S I L E ~51JRFACE FAILURE, 14

Figure 2 Fracture strength of the ZnSe materials described in Figure 1 as a function of

Single Crystal Fracture Toughness - Experiment

Single Crystal Fracture Toughness - Prediction

Nevertheless, Gilman used Equation 5 to obtain reasonable agreement between

Combining Equations 6 and 7, the following expression for 7 results.

0.(I(I '''' .... ' .... ' .... ' ....

Single to Polycrystalline Transition

Transition from Single Crystal to Potycrysio~Sne Controlled Frocture

Figure 5 shows transitions for a number of different materials. Another example of a

O2 0.4 0.6 0.8 LO 12 S.4 t.6 t.8 2.0

Figure 5 Transition in fracture energy for a number of materials as a function offlaw

There are a number of models explaining the increased fracture toughness of

boundary, in order for it to continue to propagate on this plane in the polycrystalline

K; = ~ ; o v 1/2 cos 2 0 (10a)

Kzc = Ktc(sc)[l + tan 2 0] ~/2 (11)

where Ktc(so is the fracture toughness of a single crystal o f the material.

Table 1 Ratio of Kict Ktctsc~ as a function of misorientation angle, n, of cleavage plane

Where k is a force-law-specific constant. It is hypothesized that k will vary by an

[6] Yoder, P. L., M.S. Thesis, Pennsylvania State University, 1989