Professional Documents
Culture Documents
187
Journal of the American Ceramic Society-Evans Vol. 1 3 , No. 2
188
average microstructural features, such as fracture has been its contribution to our
grain size and porosity, whereas differ- understanding of the role of "toughness"
ences in strength between test methods in damage tolerance. Specifically, the
(e.g., bending and tension) were rational- tougher materials exhibit strengths after in-
ized on the basis of weakestlink statistics. dentation that vary less rapidly with inden-
Some of the first observations concern- tation load (crack size) than for brittle
ing the interactions of cracks with the mi- ceramics (Fig. 3). This behavior reflectsthe
crostructure were made within the next existence of a resistance curve. Indeed,
several years. Out of these observations this measure of toughness is probably the
came preliminary concepts for fracture re- most useful in terms of the practical ap-
sistance mechanisms and the models that plication and durability of ceramics.
emerged later. Many of these concepts Many of the indentation methods are
can be traced to the work of Lange and only approximate and do not provide the
of R. Rice. Also, work at Harwell Labora- quality of fracture resistance data needed
tories in the United Kingdom established to rigorously relate toughness to micro-
that very high toughness ceramics could structure. The surface flaw methods, in-
be made by incorporating carbon fibers. troduced first by Petrovic and Jacobson,
This body of work provided a strong moti- seem to be the most precise, providedthat
Fig. 1. Overall chronological trends in the vation to maintain a high level of activity residual stresses are eliminated by polish-
development of high-toughness ceramics. in the field and to seek a better under- ing out the plastic zone. However, none
standing of mechanisms and approaches of these methods can be used for the
for enhancing the reliability and damage highest toughness materials now
tolerance of ceramics. available.
Although these prospects were evident,
Materials by design progress was slow for two reasons. First- (4) Pivotal Role of Zirconia
ly, different methods for measuring frac- The discovery in 1976 that zirconia can
ture toughness often gave different results, exhibit high toughness initiated a remark-
leading to enquiries about the utility of the able decade of development,culminating
fracture toughness as a material parame- in materials having toughness on the or-
ter. Secondly, the most acceptable test der of 20 MPa.ml'2. A sequence of ma-
methods required moderately large test terial inventions in Australia, Germany, and
specimens, whereupon the data emerged the United States involving Garvie, Claus-
slowly and were confined to the relatively sen, Lange, and Gupta provided the in-
few ceramics amenable to the fabrication centive. Such materials elicited
of sufficiently large specimens. One con- outstanding characterizationresearch by
sequence of these factors was that the fo- Heuer, Ruhle, and Hannink and some
cus of much of the research in the field novel experiments originating with Swain
was on the characterization and implica- and Marshall. The results were then ration-
tions of slow crack growth, with minimal alized through a clear understanding of
attention given to the enhancement of the responsible mechanisms, established
touahness. by the micromechanics models of
T i e first of the above issues would later McMeeking. Budiansky, Hutchinson,
Fig. 2. Scheme involved in the materials-by- be resolved upon the revelation that many Evans, and Marshall. This process oc-
design concept. curred in an iterative manner built around
polycrystalline ceramics exhibit resistance-
curve behavior, as elaborated in Section the understanding that the process-zone
I (5). However, it is interestingto note that size was important and that resistance-
the resistance-curveeffects were first ob- curve effects are inherent to the
served in 1974, but the significance was mechanism.
not appreciated. The second issue was
addressed upon the introductionof inden- (5) Importance of Resistance Curves
tation methods discussed in the following All "tough" ceramics exhibit resistance
section. curves. However, this feature of their be-
havior did not become apparent until
(3) Influence of Indentation Fracture about 1980. The existence of resistance-
The introductionof approaches for frac- curve behavior was predicted by
ture toughness estimation by using vari- McMeeking and Evans to be an essen-
ous indentation methods allowed tests on tial feature of transformation toughening
small specimens and, thus, permitted rap- and soon verified experimentally by
id probing of the "damage tolerance" of Swain. These predictions placed empha-
many different ceramics. These ap- sis on the transformationwake. The more
proaches emerged from systematic pi- general importance of the wake was
oneering research by Lawn over more demonstrated by some clever experi-
-
than a decade. Of particular importance ments conducted by Steinbrecht, who re-
was the discovery that the cracks around vealed that the toughness of Al2O3 is
hardness indentationsformed on unload- diminished upon removing the wake by
Flaw size, a ing and, hence, were dominated by sawing. Continued research on this topic
residual fields Appreciation of this factor revealed that microcrack toughening ex-
Fig. 3. Effects of damage on the strength
of low and high toughness ceramics The tear permitted the 'dentificationOf the non- hibits wake effects and that crack bridg-
ing modulus IS the slope of the resistance dimensionalparameters that related ing by intact grain is a common feature
curve toughness to either the size of the cracks of crack extension in polycrystals and is
or the failure stress after indentation also an important contribution to the frac-
A more recent benefit of indentation ture resistance. Finally, and most recent-
February 1990 Perspective on the Development of High-Toughness Ceramics 189
ly, it has been demonstrated that and Isostatically Hot-PressedAl2O3Com-
toughening by fibers, whiskers, and met- posites," J. Am. Ceram. SOC.,66 [6]
al networks all exhibit wake-dominated 396-98 (1983); F. F. Lange and M. Met- Limiting Physical Characteristic
resistance-curvecharacteristics. calf, "Processing-RelatedFracture Origins:
When rigorously measured and inter- 11, Agglomerate Motion and Cracklike In-
preted, resistance-curvebehavior ration- ternal Surfaces Caused by Differential Sin-
alizes effects of specimen geometry on tering," ibid., 398-406; and F. F. Lange,
toughness, crack-size effects, and trends B. I. Davis, and I. A. Aksay, "Processing-
in strength. Additionally, in the highest Related Fracture Origins: Ill, Differential
toughness materials, linear elastic be- Sintering of Zr02 Agglomerates in
havior is violated for many of the common A1203/Zr02 Composite," ibid., 407-408.
test specimens. Consequently, nonlinear J. R. Rice; p. 191 in Fracture, Vol. 11,
approaches are needed to characterize Mathematical Fundamentals. Edited by H.
material behavior. Liebowitz. Academic Press, New York, Microstructure, Fabrication,
1968. micromechanics machining
(6) Discovery of Composites
R. W. Rice, "Mechanisms of Toughen- Fig. 4. Aspects for achieving high-reliability
High-work-of-fracture-fiber-reinforced ing in Ceramic Matrix Composites," ceramics.
ceramics consisting of carbon fiber, rein- Ceram. Eng. Sci. Proc., 2 [7-81 661-701
forced glasses, and glass-ceramicswere (1981).
first demonstrated in 1972, accompanied R. W. Steinbrech and A. H. Heuer, "R-
by the first complete model of composite Curve Behavior and the Mechanical
behavior. However, activities on ceramic Properties of Transformation-Toughened
composites were not pursued vigorously ZrO2-ContainingCeramics," Mater. Res.
until 1983, when Prewo and Brennan an- Soc. Symp. Proc., 60,469-81 (1986).
nounced a Nicalont Sic-fiber-reinforced M. V. Swain, "Inelastic Deformation of
glass-ceramic composite. This announce- Mg-PSZ and Its Significance for Strength-
ment established the possibility that ce- Toughness Relationships of Zirconia-
ramic composites having high resistance Toughened Ceramics," Acta Metall., 33,
at elevated temperature might be possi- 2083-91 (1985).
ble. A major activity on ceramic-matrix A. V. Virkar and R. L. K. Matsumoto,
composites has ensued and is still in pro- "Ferroelastic Domain Switching as a
gress. Another related discovery was that Toughening Mechanism in Tetragonal Zir-
A1203 could be toughened by SIC conia," J. Am. Ceram. Soc., 69 110)
whiskers. Together, these materials have C-224-C-226 (1986).
provided the basis for the concepts of
toughening by brittle reinforcements, in- II. Toughening Mechanisms
volving the debonding and sliding proper-
ties of the reinforcementlmatrixinterfaces. (1) General Features
The objective of research concerned
(7) Bibliography
with structural ceramics is the generation f Crack advance, Aa
Actual references to the work alluded of materials having high reliability. To a0
to above can be located in Section 11, achieve this objective, there are two fun-
which covers the various mechanisms in damentally different approaches (Fig. 4): Fig. 5. Resistance-curvebehavior charac-
some detail. The reader is also referred to teristically encountered in tough ceramics K,
flaw control and toughening. The fiaw con- IS the fracture resistance and Aa IS the crack
the seven-volumeseries Fracture Mechan- trol approach accepts the brittleness of the
ics of Ceramics, edited by Bradt, Evans, advance
material and attempts to control the large
Hasselman, and Lange (Plenum Press, extreme of processing flaws. The tough-
New York, 1974-1983), which contains ening approach attempts to create micro-
much of the chronology, plus the follow- structures that impart sufficient fracture
ing basic references: resistance (Fig. 5) that the strength b e
A. G. Evans and A. H. Heuer, comes insensitive to the size of flaws (Fig.
"REVIEW-Transformation Toughening in 3). The former has been the subject of
Ceramics: Martensitic Transformations in considerable research that identifies the
Crack-Tip Stress Fields," J. Am. Ceram. most detrimental processingflaws, as well
SOC.,63 [5-61 241-48 (1980). as the processing step responsible for
R. C. Garvie, R. H. J. Hannink, and R. those flaws.'-3 The latter has emerged
T. Pascoe, "Ceramic Steel?," Nature (Lon- more recently, and has the obvious ad-
don), 258, 703-704 (1975). vantage that appreciable processing and
A. H. Heuer, "Transformation Toughen- postprocessingdamage can be tolerated
ing in Zr02-ContainingCeramics," J. Am. without compromisingthe structural relia
Ceram. SOC.,70 \lo] 689-98 (1987). bility.4.5
A. H. Heuer, F. F. Lange, M. V. Swain, The resistance of brittle solids to the
and A. G. Evans (eds.), "Transformation propagation of cracks can be strongly in-
Toughening," J. Am. Ceram. Soc., 69 [3] fluenced by microstructure and by the use
169-298 (1986); ibid., 69 [7] 511-84 of various reinforcements.The intent of the
(1986). present paper is to provide a succinct r e
J. W. Hutchinson, Harvard University view of the known effects of microstruc-
Report No. 58 (1974). ture and of reinforcements on fracture
F. F. Lange, "Processing-Related Frac- resistance. In most cases, toughening
ture Origins; I , Observations in Sintered results in resistance-curvecharacteristics
(Fig. 5), wherein the fracture resistance
tNippon Carbon Co., Tokyo, Japan svsternaticallv increases with crack exten-
190 Journal of the American Ceramic Society-Evans Vol. 13, No. 2
sion. The resulting material strengths then ever the length of the nonlinear zone is ap-
depend on the details of the resistance preciably larger than the spacing between
curve and the initial crack lengths,s,' such the relevant microstructuralentities. Such
that toughness and strength optimization conditions invariably exist when the ma-
usually involve different choices of micro- terial exhibits high toughness. The cou-
structure. The individual mechanisms in- pling between experiment and theory is
clude transformations, microcracking, another prevalent theme, because tough-
twinning, ductile reinforcements, fi- ening is sufficiently complex and involves
berlwhisker reinforcements, and grain a sufficiently large number of independent
bridging. variables that microstructureoptimization
An underlying principle concerns the only becomes practical when each of the
essential role of nonlinearity, as manifest important modes has been described by
in mechanisms of dissipation and energy a rigorous model, validated by experiment
storage in the material, upon crack propa- (Fig. 2).
gation. Consequently, the potent toughen- The known mechanisms can be con-
ing mechanisms can be modeled in terms veniently considered to involve either a
of stresddisplacement constitutive laws for process zone or a bridging zone (Fig. 7).
Fig. 6. Schematic diagram illustratingnon- representative volume elements (Fig. 6). The former category exhibits a toughen-
linear hysteretic elemental response and as- Furthermore, the toughening can be ex- ing fundamentally governed by a critical
sociations with enhanced toughness. plicitly related to stressldisplacement hys- stress for the onset of nonlinearity, $, in
teresis (Fig. 6), as will be elaborated for elements near the crack and by the as-
each of the important mechanisms. The sociated stress-free strain, ~ i . 8 - 1 1 The
general philosophy thus adopts the con- resulting stress-strain hysteresis of those
cept of homogenizing the properties of the elements within a process zone then yields
material around the crack and then for- a steady-state toughness given by11
mulating a constitutive law that character-
izes the three-dimensional material
response. Models that discretize micro- where AgCis the increase in the critical
structure details in two dimensions are strain energy release rate when the crack
typically less rigorous, because three- is long,$ f the volume fraction of the
dimensional interactions along the crack toughening agent, and h the width of the
front are not readily described, and be- process zone in steady state.9,QTransfor-
cause multiple calculations are needed to mation, microcrack, and twin toughening
eliminate artifacts of the discretization. The are mechanisms of this type.
homogenization approach rigorously The bridging category exhibits tough-
describes the toughening behavior when- ening governed by hysteresis along the
crack surface,'3~14induced by intact m a
Fig. 7. Schematic diagram illustrating both $Short cracks and cracks without fully developed terial ligaments
process-zoneand bridging-zonemechanisms process zones give smaller changes in toughness as
of toughening elaborated in the discussion of resistance curves U*
zone sizes measured in steady state20 (Fig. knowledge of the nucleation is not yet suffi-
11) has revealed that Eq. (5) predicts the cient to allow R and T to be expressed ex-
correct trends but consistentlv underesti- plicitly in terms of these variables.
mates the toughness by a factor of =2. It
(iii) Crack-Tip Stress Fields
has been surmised that the disparity arises
because shear effects have not been in- The general form of the crack-tip stress
corporated. One hypothesis regarding the field in the presence of a transformation
shear strain involves nonassociated flow.8 zone, depicted in Fig. 12, is characterized
Specifically, it is presumed that the shear by two stress intensity factors.9 Outside the
stress dominates the nucleation of the zone, for small-scale transformation (i.e.,
transformation, but that the residual strain a small zone compared with crack length
is predominantly dilatational, because of and specimen dimensions), the field is giv-
extensive twinning.21 This premise results en by the linear elastic solutions (h<r<a)
in a zone profile, in plane strain, with
diminished transformed material ahead of
OJ = (KJfZr) :, (7)
the crack. The resultant toughening ex- where K, is the stress intensity deter-
ceeds Eq. (5) and, furthermore, agrees mined by the applied loads and r the dis-
quite well with experimental data (Fig. 11). tance from the crack tip. Close to the tip,
Fig. 11. Comparison between theory and However, the necessary zone shapes are the transformation strain is saturated, and,
experiment for various partially stabilized zir not consistent with those observed by ex- hence, the material is again linear and can,
conia materials AKu refers to a zone shape periment.22 Some inconsistency thus re- consequently, be characterizedby anoth-
dictated by the equivalent stress, whereas mains to be addressed, and other shear er stress intensity factor, Ktlp,such that
AKd refers to a zone shape governed by the postulates, as well as the operation of ad-
mean stress ditional mechanisms, such as twin-
induced microcracking (see Section 11(2)),
may be involved.
The zone size, h, represents the major
The coefficient z,,,
which depends on the
polar angle, approaches that for elastic
microstructural influence on toughness. materials when f$N$ IS small
Clearly, h is governed by a martensite The transformation may thus be charac-
nucleation law. However, a fully validated terized by a stress intensity change, AK,
law does not yet exist. Consequently, con- defined as
nections between h and the microstruc-
ture still cannot be specified. Nevertheless, (9)
certain trends are apparent, based on the When Ktlp<K,,, the transformation zone
free energy of the fully transformed prod- shields the tip from the applied loads. The
uct. Specifically, h invariably decreases fracture behavior is governed by the
with increase in temperature and decrease values of Ktipand K, at the fracture criti-
in particle size. A temperature- and cality. Knowledge of the stress in the in-
particle-size-dependenttoughness is thus tervening regions is not required for
inevitable for this mechanism. analysis of the toughening.10 Specifically,
(ii) Constitutive Laws the near-tip field provides a plausible
Constitutive laws expressly relate the crack-extension criterion
components of the stress and trans- Ktip = KO (1 0)
formation strain tensors in the transform-
Fig. 12. Crack-tip stress fields in the ing solid. The stresses are most where KO is the fracture resistance of the
presence of a dilatational transformation zone conveniently expressed in terms of the material in the transformation zone im-
for a supercritical transformation. mediately ahead of the crack tip, where-
mean stress, 5, the deviatoric stresses,
s, =a,, - d,, 0, and/or the equivalent upon, the observed toughness is
stress, u e = v 2 , and their variation Kc = KO+ AK,
during transformation. Such relations are, (11)
in fact, the equivalent of the more familiar where AKc is the quantity - AK, evaluat-
constitutive laws used to describe plastic ed at the fracture criticality.
deformation and crack-tip fields in elas-
tic/plastic solids. A viable constitutive law (iv) Toughening and Resistance
must account for the effect of particle size Curves
and shape upon the incident of transfor- There are two equivalent analytical
mation by allowing only a certain fraction, methods for determining trends in tough-
6 of the solid to actually transform at a giv- ness: one based on stress intensity factors,
en imposed stress, ai. The simplest yield AK, and the other on conservation in-
criterion has the form23 tegrals A 9 . Both approaches predict the
same behavior, but have different utility
when various aspects of toughening re-
quire elucidation. Consequently, both ap-
proaches are briefly described. The
where R and T are experimentally deter- approach based on stress intensity factors
mined, normalizing parameters that de- indicates that the magnitude of AK clear-
Fig. 13. Change in normalized toughness pend on f , E~ and particle size ly depends on the shape of the zone and
as a function of normalized transformation Experiments per!ormed on partially stabi- on the components of the transformation
strain lized zirconia (PSZ) and tetragonal zirco- strain tensor. Initially, it is pertinent to con-
nra polycrystals (TZP) are consistent with sider a crack in an untransformedparent,
this transformationcondition 23 However, such that a frontal zone develops as the
February 1990 Perspective on the Development of High-Toughness Ceramics 193
load is imposed. Subsequently, a steady- Few toughened materials exhibit super-
statezone of uniform width over the crack critical transformation. Generally, there is
surfaces (Fig. 10) is examined. The AK lev- a gradient in the fraction of transformed
els are different for these two zone con- material with the transformation zone, with
figurations, resulting in R-curve effects.9 the largest fraction present near the crack
For the stationary crack, when the long plane. Such conditions are referred to as
range strain field of the transformed parti- subcritical. For this case, the toughness
cles is purely dilatationa1.f the mean can be obtained by integrating the meas-
stress dictates the shape of the transfor- ured transformed fraction over the width
mation zone, which then has the shape of the zone.g.10 Also, in some materials, the
depicted in Fig. 10. For this condition, transformation is reversible. Such transfor-
AK= 0. For the steady-stateconfiguration mations can produce toughening, provid-
and when the transformation does not re- ed that hysteresis is involved.8
verse in the wake, as well as when all par- The equivalent approach for cracking
ticles within h are transformed, the plane toughness, based on conservation in-
strain zone width for small-sca/edilatational tegrals, dictates that, for the frontal zone,
transformation is9,IO the volume elements within the zone do
not experience unloading. Consequently,
6 (1 + v ) ~K,
h= 12n (z) the path-independent J integral applies
and the relation
The corresponding extent of crack shield- J = (1 - Y*)K~/€ (1 7)
ing is given by Eq. (5) as
pertains for all line contours around the
AK= -0.22E f$\/T; /(I -u) (13) crack tip. Furthermore,because the elas-
tic properties of the transformed and un-
The supercritical plane strain toughness transformed materials are essentially the
can thus be expressed in the following same, contours around the tip (giving
forms: Ktip)and remote from the tip (giving K,)
AKc = 0.22E f$fi/(
1 - u) yield identical values of K whereupon
Ktip=K, and A K = O , as before. Con-
AKJKC = ( O / I2R)(1 f U)EfE: / F ( I - U) versely, when a fully developed zone ex-
(1 4) ists (Fig. l o ) , the material within the zone,
behind the crack tip, has experienced un-
KJKo=[1 -(fi/1211)(1 +u)
loading, and a path-independentJ does
x EFT/ 3(1 - u p (1 5) not apply. In this case, the appropriate
I f the transformation is not activated by the conservation integral, I , has the same form
hydrostatic crack-tipfield, but instead oc- as J at the tip
curs in shear bands inclined at =n/3 to the
crack plane (R>> in Eq. (6)), all of the
deleterious transformations in front of the but remote from the tip
crack are excluded. Then, if the transfor-
h
mation strain is still hydrostatic, the super-
critical AKc increases to l = [(l - u2) K $ E ] - 2 IU Q
0
dy (19)
fl
A& = 0.40EO~ (24)
whereas, for a critical mean stress
AK, = 0.32EO~fi (25)
where OT is now the misfit strain caused
by microcracks (analogous to f.9. By
contrast, the toughening imparted by the
modulus reduction depends on zone
shape, but not on zone size. Following
Hutchinson,% the asymptotic modulus
contribution is34
(1 - v)AKJKC= (kf - $)(GIG- 1)
+ (k2+ $)@GIG- V )
(26)
where kl and k2 depend on the micro-
cracking criterion. Values of AKc/Kc cap
be obtained by simply inserting GIG
and V into Eq.(26). Typical results for the
steady-statetoughening are, for a critical
normal stress and pemy-shaped micro-
cracks
AKJK, =1.42q (27) Fig. 14. Ratio of near-tipto remote stress intensity factors. The dashed line is the asymptotic
result for small E:
with q being a measure of the microcrack
density, defined following Eq. (30). In
general, the AKc due to dilatation and
modulus effects are not additive: interac- microcracked bodies establish the modu-
tion terms are involved. lus effect34 as
strain in the first two is limited by elastic the work hardening rate. These simple a,
c
strains in the elastic network. Consequent- results already provide the important in-
sight that a peak stress exists at small
E
ly, experience and analysis confirm that F
crack openings, such that the plastic dis- m
crack bridging is usually the most potent Q
mechanism when the ceramic phase is sipation is dominated by the large strain
continuous, because the only ductile (necking) regime (Fig. 23). The importance
regions which experience extensive plas- of interface debonding thereby becomes
tic strain are those segments that stretch apparent, because debonds reduce the
between the crack surfaces in the bridg- constraint, but increase the plastic stretch
ing zone. The plastic dissipation in this prior to failure. Numerical solutions ob-
bridging zone can be large and can pro- tained with prescribed initial debonds es-
vide a major increase in toughness. Con- tablish the salient trends (Fig. 23). Notably,
the plastic dissipation increases systemat- Distance from crack plane, y (pm)
versely, a composite with a continuous
metal network is subject to nonlinear pow- ically as d/R increases. Furthermore, if the Fig. 20. Trends in the microcrack density
er law deformation and can experience debond evolves during crack opening, the parameter, '1=N<c3>, with distance y from the
substantial plastic strain within a plastic dissipation is further enhanced. The above crack surface.
zone. Consequently, the potential for ap- predictions are generally similar to ex-
preciable plastic dissipation within a plas- perimental measurements, which confirm
tic process zone is much greater for this large stresses at small u and increased
class of microstructure. dissipation when debonding occurs. Bridging Grains
Understanding of the toughness gener- The toughness attributed to bridging,
based on Eq. (3),can be reexpressed in Interfaces
ated by ductile ligaments is contingent
upon the law that characterizes the nondimensional form by noting that the Residual strains
stress/stretchrelation, flu).Insights regard- flow stress scales with the uniaxial yield
ing the parametersthat affect this relation strength, Y, and that the plastic stretch is
can be gained from simplified analytical proportional to the radius of the cross sec-
models. Complementary numerical solu- tion of the reinforcing ligaments, R; con-
tions then allow determination of specific sequently,the asymptotic toughness is4041
trends. Budiansky and CO-workers13.52 A g C I f YR = x
and Rose53 analyzed srnal/-%a/ebridging,
(34)
in which bridge length is small in relation where x is a "work-of-rupture''parameter
to crack length, specimen dimensions, that depends on the critical plastic stretch,
and distances from the crack to the speci- u, (or ductility), of the reinforcement and
men boundaries. The stresslstretch rela- on the extent of the interface debonding,
tion depends strongly on the mode of d. Values of x have been obtained both
failure of the ductile ligaments. A small- by calculation and by experiment40 based
scale yielding analysis indicates that t on determinations of the area under the
Fig. 21. Ligament formation allowed by
residual stress.
should increase rapidly with initial crack stresdstretchcurve for the reinforcing liga-
opening. A large-strain necking analysis ments. For a wekbonded interface (d=O)
then reveals that, without debonding, the and for ductile ligamentsthat fail by neck-
stress attains high initial levels because of ing to a point, the resultant trend in x with
198 Journal of the American Ceramie Society-Evans Vol. 73, No. 2
wherein the precipitates locally block the
passage of dislocation, causing the flow
strength to be inversely proportional to the
precipitate spacing. Inserting the meas-
ured precipitate spacing ( k l 00 nm), the
yield strength is Y=220 MPa. This value
is comparableto typical values quoted for
Al alloys precipitation hardened with Cu.
Stereo measurements have been used to
evaluate the plastic stretch to failure, u,.
The normalized plastic stretch, u, lR,
varied appreciably between ligaments.
However, there was no systematic depen-
dence on either the ligament dimension,
R, ortheaspect ratio: the mean stretch
was u,/R= 1.6. The plastic stretch to
failure indicates that x is in the range 2 to
3.5. Substitutingthe above values of Y and
x into Eq. (35) and noting that f=0.2 and
R=2.0 pm, A g , is calculated as 170
to 300 J * m-2, Consequently, there is ac-
ceptable agreement with the measured
value of 150 to 200 J. m-2, Such behavior
is consistent with the interpenetratingnet-
work microstructure.
For cemented carbide materials, the
Fig. 22. Schematic indicating the various contributions to the steady-state toughness. contribution from bridging deduced from
the measured zone sizes is found to be
-
relatively sma1154,55 (AgC=40J m-2) for
reasonable choices of the flow stress of
the Co alloy, whereas the overall meas-
ured toughness is g?,=400 J . m-2.
work-hardening rate n indicates that x is Consistent with this interpretation, meas-
in the range 0.3 to 1. Less-ductile liga- urements of plastic zones indicate zone
ments that rupture prematurely by profuse sizes in the range h=40 pm.42 It is sup-
hole nucleation have correspondingly posed that dissipation is appreciable in this
smaller values of x. Systems subject to de- zone and that this may be the
bonding exhibit larger x and approach 8 predominant contribution to the toughness
for large dlR. Experimental results have in- of these materials.
dicated that trends in Y and d are reflect-
ed in a plot of x with rei‘ative plastic stretch (B) FiberMhisker Reinforcement
u,/R (Fig. 24). (i)Basic Features
(ii) Toughness Correlations Practical ceramic-matrix composites
reinforced with continuous fibers exhibit
The basic nondimensional solution (Eq. the failureldamage behaviors sketched in
34) can be used both to rationalize tough- Fig. 26. The composite properties are
ness measurements and to develop a known to be dominated by the interface,
I IcI.7LI” *LIFL”II, _ I a
predictive capability. For purposes of the and bounds must be placed on the inter-
former, Y and n can be inferred either from face debonding and sliding resistance to
Fig. 23. Nondimensional stress versus a TEM characterization of the microstruc-
stretch behavior and associated “work-of- have a composite with attractive mechan-
rupture” x for various debond lengths.
ture50 or from microhardness measure- ical properties. The strong dependence of
ments (Elliott ef a/.a),while R, u,, and ceramic-matrix composite properties on
d can be determined by quantitative SEM the mechanical properties of the interface
of the fracture surface.50 Comparison be- generally demands consideration of fiber
tween theory and experiment is most coatings andlor reaction product layers.
readily achieved by reexpressing Eq. (34) Residual stresses caused by thermal ex-
in the form
pansion differences are also very im-
A q , = f YRx (u, IR) (35) portant.
The specific microstructural parameters
To use this result, Y is first o b t a i n e m - that govern Mode I failure are the relative
lowed by evaluation of x , using u, IR. fibedmatrix interface debond toughness,
Thereafter, A q c is examined for con- rjlrf,the misfit (thermal expansion) strain
sistency by comparison with the ex- between fiber and matrix, E ; the friction
perimentally determined value. For coefficient at the debonded interface, p,
Lanxide composites of A1203 reinforced the statistical parameters that characterize
with Al, the experimental information is as the fiber strength, So and rn, the matrix
follows: The yield strength has been evalu- toughness, rm,and the fiber volume frac-
Fig. 24. Work of rupture as a function of ated by noting that the alloy contains Al- tion, f. The prerequisite for toughness is
plastic stretch measured for various ductile Cu precipitates which behave as im- that ri /I-,; 114 to allow crack-front de-
reinforcements. penetrable obstacles (Fig. 25). Conse- bonding (Fig. 27). Subject to this require-
quently, their influence on yielding should ment, the misfit strain must be small
be represented by Orowan hardening, (E: 2 3 x 10-3) and preferably negative,
February 1990 Perspective on the Development of High-Toughness Ceramics 199
such that the interface is in tension. sites reinforced with SIC fibers reveal that
Furthermore, for a continuous fiber, the materials with a C interlayer satisfy de-
friction coefficient along the debonded in- bonding requirements (Fig. 29) and also
terface should be small. The ideal fiber have small T and thus demonstrate exten-
properties include a high median strength sive pullout.60-62 Conversely, composites
(large So)and large variability (small m), having a continuous SiOn layer between
as needed to encourage large pullout the matrix and fiber exhibit matrix crack
lengths. When r j Irf and p are both small, extension through the fiber without
experience has indicated that the tensile debonding62.63 (Fig. 30). Experiments on
stress-strain behavior illustrated in Fig. these composites and on whisker-
26(A) is obtained. Three features of this reinforced composites also confirm the
curve are important: matrix cracking at a strong influence of T. Notably, systems
stress uo. fiber-bundlefailure at uu, and which debond but do not slide readily (be-
pullout. Conversely, larger r,/rfand p cause of either a high friction coefficient
cause the stress-strain curve to become or morphological irregularity) exhibit small
linear (Fig. 26(!3)). The ultimate strength pullout lengths and moderate toughness.
then coincides with the propagation of a Such behavior is exemplified by
single dominant crack. LASttlSiC composites with oxide inter-
Present understanding of debonding is phases and by various whisker-reinforced
consistent with the following sequence of materials, respectively.
events during matrix crack propagation. Tough composites can be obtained by
Initial debonding along the interface at the creating the appropriate interphases
crack front requires that r i / r f be small between the fiber and matrix, either by
enough to lie within the debond zone coating or, in situ, by segregationhter-
depicted in Fig. 27.56 Furthermore,the ex- diffusion. The most common approach is Fig. 25. Bright-fieldTEM view of the precipi-
tates in the Al alloy that cause hardening.
tent of debonding is typically small when the use of a dual coating: the inner coat-
residual compression exists at the inter- ing satisfies the debonding and sliding re-
face, but can be extensive when the in- quirements, while the outer coating
terface is in residual tension and the fibers provides protection against the matrix dur-
are smooth. Further debonding is usually ing processing. However, the principal
induced in the crack wake57 (Fig. 28). The challenge is to identify an inner coating
extent of this debonding is governed that has the requisite mechanical proper-
largely by the residualfield. Residual radial ties while also being thermodynamically
tension results in unstable conditions and stable in air at elevated temperatures.
encourages the extensive debonding of
smooth fibers. Residual compression (ii, Constitutive Laws
andlor an irregular fiber morphology The mechanical properties of uniaxial- Fiber bundle failure, o,,
cause stable debonding,Se with the extent ly reinforced composites are largely
determined by the friction coefficient and governed by the relationshipsbetweenthe
the roughness of the debonded interface. opening of a matrix crack, u, and the
Subsequent fiber fracture involves the stresses, t, exerted on the crack by the in-
statistics of fiber failure,59 subject to an ax- tact bridging fibers and the failed fibers as
ial stress governed by the sliding they pull out. The traction t(u)is well-known
resistance of the debonded interface. The for composites in which debonding occurs
above sequence suggests that, although easily (very small r j / r f )and which also
debonding is a prerequisitefor high tough- slide easily along the debonded interface
ness, the properties of the composite are (small T).For other cases, reliable t(u)func-
dominated by the sliding resistance of the tions have yet to be elucidated. Easy E
debonded interface, which dictates the debonding and sliding provide a crack- (a) 'Tough' composite
major contribution to toughness caused by opening function that depends on the sign
pullout. The locations of fiber failure that of the misfit strain, as well as the rough-
govern the pullout distributions can be de- ness of the debond interface,through their
termined from the stresses on the fibers, effect on the sliding resistance. For in-
using concepts of weakest-link statistics. stance, sliding can be described by a
Analysis of this phenomenon has been Coulomb friction law
performed for composites having debond-
ed interfaces subject to a constant sliding T=Pq (36)
stress, 7.59 The magnitude of T governs the where p is the friction coefficient and q the
load transfer from the fiber to the matrix. nominal residual compression normal to
Large values of T cause the fiber stress to the interface. At the simplest level, q is set
vary rapidly with the distance from the by the residual misfit strain and the
matrix crack and induce fiber failure close fiberkoating morphology, such that T is
to the crack, leading to small pullout essentially invariant. In this case, prior to E
lengths, h,. Conversely, small T results in the incidence of fiber failure, t and u are (b) 'Brittle' composite
large h,. related by17.18
Various observations of crack interac- Fig. 26. Schematic illustrating the range of
tions with fibers are supportive of the stress-strain characteristics exhibited by
above rules. In particular, experiments on (37) ceramic-matrix composites.
glass and glass-ceramic-matrix compo-
Equation (37) has been used to describe
the mechanical behaviors that are ob-
Kithiurn aluminum silicate tained while the fibers are largely intact.
200 Journal of the American Ceramic Society-Evans Vol. 73, No. 2
At another extreme, when all of the fibers fiber and matrix surface separate, leading
have failed, the traction on any fiber is59.63 to unrestricted crack opening. Additional
complexity is involved when debonding
ti = 2f~(hj
- u)/R (38) and sliding occur simultaneously. Some
where hi is the distance from the matrix preliminary results for such behavior ex-
1.o
crack plane at which that fiber failed. ist, but are not addressed here. it is sim-
e
G More difficult problems to address con- ply noted that, in such cases, the fracture
failure cern the incidence and location of fiber energy of the coating becomes another
0.5
failure. This has been regarded as a parameter of interest.
problem in weakest-linkstatistics,a where- The location of fiber failure vis-a-visthe
0 upon the function t(u) depends upon the matrix crack plane is of critical importance
I I I I I
statistical parametersSo and m in addition because this location governs the pullout
-1.0 -0.5 0 0.5 1.0
to the variables contained in Eq. (37).The length h,. Both theory59 and experi-
Elastic mismatch, a
expressions are unwieldy and are not menF suggest that, for aligned reinforce-
reproduced here, but can be located in ments, h, is governed by weakest-link
the article by Thouless and Evans.59 Some statistics. For the simplest case, wherein
trends, expressed in nondimensional form, the sliding stress T remains constant, an
are summarized in Fig. 31. The small dis- expression for the mean pullout length has
placement behavior is dominated by the been derived as59
intact bridging fibers, whereas the long tail
(2h,/R)m" = (1/2rr(m+ l)m](AdRz)
is governed by the pullout of failed fibers.
Another level of complexity is involved x (so/T)mr[(m+ 2)/(m+ l ) ]
when the interface stress q~ varies with (40)
-1.0 -0.5 0 0.5 1.0
Elastic mismatch, a the loads t on the fiber. A simplified result where r in this expression is the gamma
obtained using a modified shear lag ap- function and A. is a reference area for the
Fig. 27. Fracture energy requirements for proach suggests the following features58 fibers (usually set equal to 1 m2). Conse-
crack-frontdebonding. (Ef = Em= E and vf= urn= u): quently, for aligned fibers, it is evident that
U/R=(fJ/p)[f(l+F)/(l - uf)] h R is essentially governed by S ~ Thigh :
x(l -utuf)/(l-9 (39) g'
fi er "strength" and low sliding resistance
encourage large pullout lengths. Cor-
where responding trends for inclined fibers are
unknown.
f = t/f$E
(iiq Matrix Cracking Stress
with being the misfit strain which
causes the interface to be in residual com- The stress a. at which matrix cracking
pression. Equation (39),although approx- occurs has been the most extensively
imate, has several salient features. In studied behavior in ceramic-matrix com-
particular, as expected, crack opening is posites. For composites in which the
inhibited by large values of the friction residual stress normal to the interface, qN,
coefficient. Furthermore, as f + l / v , the is tensile and the interface properties can
be effectively represented by a unique
sliding stress, T , Eq. (37) may be used to
derive the lower-bound, steady-state
matrix cracking stress."
00 a* P
E-E Em
where
stress exhibits a maximum 60 given by17 toughness. The above effects are indica-
tive of resistance-cuwe behavior, because
6,x +
= (2/3)[2fpr,,,/(i E/E,)EA~’~ (42) each contribution is only fully realized
This result has evident implications for when the fibers fail and pull out. The full
material design. details of the increase in the fracture
More detailed diagnosis of matrix crack- resistance can be calculated from the
ing confirmsthat Eq. (41)is a lower bound crack surface tractions t(u)by applying Eq.
for the onset of cracking. Further crack- (3). A useful simplification for further dis-
ing occurs as the stress is raised above cussion is the peak (or asymptotic) tough-
go, resulting in a periodic crack array (Fig. ness that is obtained when each
33).’6 The crack spacing reaches a satu- mechanism exerts its maximal contribu-
ration value, s, when the stress every- tion. At the simplest, physically relevant
where in the “matrix blocks” between level, this toughness is given by66
cracks becomes smaller than that applied
stress. The magnitude of the saturation Agczfd[S2/E- E ( E ; ) ~
crack spacing is governed by the sliding +
+ 4rj/R(1 - f ) ] 2~fhE/R (45)
stress, T,such that T and s are related b p
The first term is a bridging contribution that
derives from the stored strain energy dis-
sipated as acoustic waves, with S being
Consequently, the matrix cracking stress- the reinforcement “strength.” The second
es can be expressed in terms of the crack term is the loss of residual strain energy
spacing as caused by matrix crack extension and d e
bonding. The third term reflects the new
“surface area” caused by debonding, Fig. 30. Fracture surface indicating fracture
and the fourth term is the pullout contri- through the fiber in a silicalSiC-fibercomposite.
The crack spacing clearly provides an ap- bution, dissipated by frictional sliding of the
proach for estimating T and for establish- interfaces.
ing a self-consistent description of matrix Experience indicates that the residual
cracking. strain term is small in systems of practical
utility and can often be neglected. The lar-
(iv) Toughness gest potential for toughness resides in the
In reinforced ceramics which fracture by
the growth of a single dominant flaw in
pullout term, provided that hdR is large.
An extreme range of pullout behaviors is - 0.6
Mode I, there are four effects which in-
fluence toughness.66 Debonding gener-
ates new surface and contributes
apparent among the available range of
fiber- and whisker-reinforced ceramics,
resulting in wide variations of toughness.
- 0.4
2 200 resilience, such as Sic, tend to fail at the supported by fractured fibers by means
.- matrix crack plane when inclined to the of stress transfer from the matrix through
E, 100
crack,67.68 whereas C fibers may still pull- interface fraction, gives
z
out over large distances69 Fiber alignment
- 1 i
0 [1 - (1 - TS/R$m*l]
0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 issues associated with pullout thus depend o,=fS exp -
Engineering strain (%) on fiber properties. +
(m 1)[I - (I - T S / R ~ ) ~ ]
When fiber pullout does not contribute
Fig. 32. Tensile stress-strain curves for a to toughness, as in many whisker- (48)
range of glass- and glass-ceramic-matrixcom- reinforced ceramics, the elastic energy with
posites reinforced with Sic (Nicalon) fibers. and debonding energy terms tend to
govern the fracture resistance, such that (R&.S)m+l =( & / ~ x R L ~ ) ( ~ ? S ~ / T S ) ~
x [l -(1 -Ts/RS)m-q
AQczfdS*/f+ 4r; f (d/R)/(I- f ) (46)
where L, is the specimen gauge length.
To interpret this expression, it is essential The effect of the sliding stress on u,, ap-
to appreciate that the debond length d de- pears directly, as well as through its effect
pends on the interface fracture energy Ti, on the crack spacing s, while the effect of
misfit strain E: and friction coefficient. The residual stress is present through its effect
associated relationships are unknown, but on T. The ultimate strength is also expect-
dimensional analysis suggests that57.58 ed to be influenced by the residual stress.
d/R = H(f,R(&,T)*/r,,S/f&,T,1/p) Specifically, in systems for which the fiber
(47) is subject to residual compression,the ax-
where H is a function. The important point ial compression should suppress fiber
is that there remains much scope for con- failure and elevate the ultimate strength to
trolling toughness by manipulating the in- a level in excess of that predicted by Eq.
terface debonding and sliding properties (48). This effect may be estimated by
and by maximizing the fiberlwhisker regarding the matrix as clamping onto the
strength. fiber and, thus, simply superposing the
A s already noted, resistance-curveef- residual stress onto S.
Fig. 33. Periodic matrix cracks formed upon loading above the steady-state matrix cracking stress.
February 1990 Perspective on the Development of High-Toughness Ceramics 203
(vi) Propefly Transition are shown in Fig. 35.62 Limited experience
Nonlinear macroscopicmechanical be- has indicated that the ratio of median
havior on tension is most desirable for strengths scales with the ratio of compo-
structural purposes. Analysis of the tran- site ultimate strengths.
sition between this regime and the linear Based on the above experimental fea-
regime is important, and involves compar- tures that C and BN coatings are effec-
ison of the basic trends in the steady-state tive in terms of satisfying crack-front
matrix cracking stress, oo, and in the debonding and also providing a low slid-
asymptotic fracture resistance, Age. Most ing resistance, whereas amorphous ox-
significantly, uo increases but A g C ides are incapable of debonding, it may
decreases as T increases. These oppos- be concluded that C and BN are the
ing trends with T suggest the existence of preferred coatings. However, both of
an optimum T that permits good matrix these materials are susceptible to rapid
cracking resistance while still allowing high oxidationa.75 and mostly incapableof per-
toughness. More specifically, a property forming their required debonding and slid-
transition is expected when the matrix ing functions at elevated temperatures in
cracking stress attains the stress needed oxidizing environments. A variety of other
for fiber-bundle failure. One estimate on coatings are thus under investigation.
the property transition can be obtained by When such coatings have been deve-
simply allowing oo to exceed the ultimate loped, some of the more complex crack-
strength, ou, whereupon a nondimension- opening and pullout constitutive laws
al parameter p which governs the transi- presented may be needed to describe the
tion when T is small is mechanical properties of the composite. t
(C) Bridging Grains Fig. 34. Schematic of the pullout process for
(49) debonded fibers indicating the roles of fiber
As already noted, bridging grains can
Specifically, when p exceeds a critical morphology, misfit strain, .and friction
be induced in two ways:5.76 grain- coefficient.
value, brittle behavior initiates. This esti- boundary debonding and residual stress.
mate has not been tested, and, further- Low fracture energy grain boundaries can
more, alternate parameters may be allow debonding, as in whisker-reinforced
conceived. materials, such that toughening involves
the same terms described by Eq. (45). In-
(vii) Experimental Correlations deed, a trend toward an increase in
Uniaxially reinforced ceramic- and steady-state toughness is found in various
glass-matrix composites with either C or polycrystalline A1203materials as the grain-
BN interlayer generally satisfy crack-front boundary fracture energy decreases.77
debonding requirements, and the materi- Also, elongated grains, which allow larg-
als exhibit axial and transverse mechani- er debond lengths, could lead to higher
cal properties that accord well with the toughness, as noted over a decade ago
above expressions for the matrix cracking for both Si3N478 and AI2O3.79 It is also
stress and the ultimate strength.62 To known that, when the grain-boundary frac-
achieve these comparisons, each of the ture energy is not low, as in sintered Sic,
parameters in Eqs. (41) and (48) has been elongated grains do not lead to higher 8-
independently ascertained. As already toughness. However, detailed application
mentioned, T can be measured using var- of Eq. (45) has not been attempted.
ious techniques: matrix crack spacing,l6,6* Residual stress-induced bridging occurs
indentation push-through stress,70,71 and because circumferential compression
matrix crack opening hysteresis.61 The causes the crack to circumvent local, high-
misfit strain $can also be determined us- ly stressed regions. Furthermore, it has
ing a variety of methods: X-ray or neutron been possible to rigorously demonstrate
line shifts, offset stresses in the crack open- that such a bridging process increases the
ing, and residual displacements obtained steady-state toughness in accordance
from nanoindenter tests. Accurate values with37
of the matrix fracture energy rm and of
AKc = 2.5fEAaATcR 0 1 2 3 4 5
the constituent elastic properties are also Fiber fracture stress, S (GPa)
needed. Furthermore, it is noted that the where / is the volume fraction of highly
comparison between theory and experi- stressed grains that result in bridging liga- Fig. 35. Fiber strength distributions ascer-
ment for the matrix cracking stress is con- ments and 2R the grain diameter. This in- tained on two composites (refer to Fig. 32 for
tingent upon having accurate values of the crease in toughness arises because the corresponding stress-strain curves) using
above parameters. ligament failure occurs unstably and ener- fracture mirror measurements
Correlations of trends in the ultimate gy is dissipated in this process as acous-
strength are primarily contingent upon tic waves, partly negated by some
knowledge of the in situ mechanical reduction in residual strain energy (c.f.. the
properties of the fibers. Appreciation for first two terms in Eq. (47)). For this process,
these properties can be gained by ex- elongated grains are not obviously
amining and measuring fracture mirrors beneficial.
on the fiber fracture surfaces in the com-
posite. Specifically, the distributionof mir- (0) Multiple Mechanisms
ror radii can be used to evaluate the axial The preceding sections have described
stress S on the fiber of the fracture loca- microstructural issues concerned with
tion.72-74 Some typical results for two toughness optimization when a single
materials with different ultimate strength mechanism operates. In practice, more
Journal of the American Ceramic Society-Evans Vol. 73, No. 2
204
2R. W. Rice, Treatise On Materials Science and
than one mechanism may exist. Conse-
Technology, Vol. 11; pp. 191-381. Academic Press,
quently, interactions between mechanisms New York, 1978.
must be considered, In some instances, 3A. G. Evans, "Structural Reliability:A Processing-
the interactions may be highly beneficial Dependent Phenomenon," J. Am. Ceram. Soc., 65
and produce synergism between [3] 127-37 (1982).
mechanisms. Such synergism has been 4A. G. Evans, "New Opportunities in the Process-
illustrated to exist when both bridging and ing of High-Reliability Structural Ceramics"; pp.
process-zone mechanisms operate simul- 989-1 010 in Ceramic Transactions, Vol. 1 , Ceramic
taneously.15 Conversely, some interaction Powder Science. Edited by G. L. Messing, E. R.
effects can be deleterious and reduce the Fuller, Jr., and H. Hausner. American Ceramic so-
ciety, Westerville, OH, 1989.
efficacy of the individual mechanisms.
5R. F. Cook, 6. R. Lawn, and C. J. Fairbanks,
Synergism is most likely when bridging "Microstructure-Strength Properties in Ceramics: I,
and process-zone mechanisms interact. Effect of Crack Size on Toughness," J. Am. Ceram
Multiplicative interactions are evident in this SOC, 68 [ l l ] 604-15 (1985).
case because the crack surface tractions 6D. B. Marshall, "Strength Characteristics of
caused by bridging can expand the Transformation-ToughenedZirconia." J. Am. Cefam.
process-zonewidth h in the crack wake, SOC.,69 [3] 173-80 (1986).
causing an additional increase in shield- 7D. Shetty and J. Wang, "Crack Stability and
ing, proportional to this increase in h. Strength Distribution of Ceramics That Exhibit Ris-
ing Crack-Growth-Resistance [R-Curve] Behavior,"
Straightforward logic indicates that mul-
J. Am. Ceram. Soc.. 72 [7] 1158-62 (1989).
tiplicative toughening between bridging eA. G. Evans and A. M. Cannon, "Toughening of
and process-zone effects should occur Brittle Solids by Martensitic Transformations," Acfa
when the ratio of the bridging-zone size, Mefall., 34 [5] 761 -800 (1986).
L, to the process-zone width, h, is small, OR. M. McMeeking and A. G. Evans, "Mechanics
because bridging generates a new effec- of Transformation-Toughening in Brittle Materiais,"
tive crack-tip toughness, that causes the J. Am. Ceram. SOC.,65 [5] 242-47 (1982).
process-zone size to further increase. 106. Budiansky, J. W. Hutchinson, and J. C. Lam-
TIT. P. Wiehs. C. M. Dick, and W. D.Nix, "The Fric- 75E. Y. Luh and A. G. Evans, "High-Temperature
tional Resistanceto Sliding of a Sic Fiber in a Brittle Mechanical Properties of a Ceramic Matrix Compo-
Matrix," Mater. Res. Soc. Symp. Proc., 120, 247-52 site," J. Am. Ceram. Soc., 79 [7] 466-69 (1987).
(1989). 7 W R. Lawn, "Physics of Fracture," J. Am. Cer-
)2J. J. Mecholsky and S. R. Powell (eds.), Frac- am. SOC.,66 [2] 83-91 (1983).
tography of Ceramic and Metal Failures, STP 827; 77R. F. Cook, "Segregation Effects on the Fracture
pp. 187-95. American Society for Testing and of Brittle Materials"; to be published in Acta Mefall.
Materials, Philadelphia, PA, 1984. 7aF. F. Lange, "Fracture Toughness of Si3N4as
l3J. F. Jamet, D. Lewis, and E. Y. Luh, "Charac- a Function of Initial a-Phase Content," J. Am. Cer-
terization of Mechanical Behavior and Fractograph- am. Soc., 62 428 (1979).
ic Observations on Compglas SiC/lAS Composites," 79A. G. Evans, A. H. Heuer, and D. L. Porter, "The
Cefam. Eng. SCI.Proc., 5 [7-81 625-42 (1984). Fracture Toughness of Ceramics"; pp. 529-56 in
l4A. J. Eckel and R. C. Bradt. "Strength Distribu- Fracture 77, Advances in Research on the Strength
tion of Reinforcing Fibers in a Nicalon Fiber/Chemi- and Fracture of Materials, Fourth International Con-
cally Vapor Infiltrated Silicon Carbide Matrix ference on Fracture, Vol. 1. Edited by D. M. R. Taplin
Composite," J. Am. Ceram. Soc.,72 [3] 455-58 Pergamon. New York, 1978.
(1989). 0