CHAPTER 7

Ceramic Matrix Composites

Ceramic materials in genera l have a very att ractive packa ge of properties:

high stren gth and high stiffness a t very high tem peratliTeS, chemical inert-
ness.jew density, and so on. ntis att ractive pac kage is marred by one deadly
flaw, namely, an utt er lack of toughness. They a re prone to catastrophic
fa ilures in the p resence of flaws (surface or internal). They are extremely
susceptible to thermal shoc k a nd are easily dama ged during fabrication an dl
or service. It is therefore understandable that an overriding co nsideration in
ceramic matrix composites (CMClI) is to to ughen the ceram ics by incorpo-
rating fibers in them and thus exp loit the a ttra ctive high-temperature stre ngth
and environmental resistance o f ceram ic materials witho ut risking a cata-
strophic failure. It is worth pointing out at the very outset that there are
certain basic differences between CMCs an d other composites_The general
philoso phy in nonceramic matrix co mposites is to have the fiber bear a
greater proportion of the a pplied load. This load pa rtitioning depends on the
ratio of fiber and matrix clastic moduli, E[ I E",. In nonceramic matrix com-
posites, this ratio can be very high , while in CMC s, it is rather low and ca n
be as low as unity. Another d istinctive point regarding C MC s is that because
of limited mat rix ductility and generally high fabrica tion temperature, ther-
mal mismatch between compo nents has a very impo rta nt bearing on CMC
performance. The problem of chemical co mpa tibility between components
in CMC s has ramifications simila r to those in, say, M MCs. We first describe
some o f the processing techniques for CMCs, followed by a description of
some salient cha racteristics of CMCs regarding interface a nd mechanical
properties a nd, in pa rticular, the va rious possible toughness mecha nisms,
and 6nally a description of some applications of C MC s.

7.1 Processing of CMCs

Ceramic matrix composites (C MCs) can be processed either by con ventional
po wder processing techniques used for mak ing polycrystalline ceramics or by
some new techniques de veloped specitically for ma king C MC s. We describe
below some of the impo rta nt processing techniq ues for CMCs.
 7.1. Prcces..ing of CMCs 213

7. 1.1 Cold ~ng lUId Sinterlng

Cold pressing of the matrix powder and fiber followed by sintermg is a ca rry-
over from conventi onal processing of ceram ics. Generally, in the sintering
step , the matrix shrinks considerably and the resulting composite ha s many
cracks. In addition to thi s general problem of shrinka ge associa ted with
sintering of an y ceramic, certain other problems arise when we put high-
aspect ra tio (Iengthfdiameter) reinfo rcemen ts in a glass or ceramic matrix
mat erial and try to sinter. Fibers and whiskers ca n form a network that may
inh ibit the sintering process. Depending on the difference in thennal expa n-
sion coefficients of the reinforcement a nd matrix , a hydrostatic '''nsil" stress
may develo p in the matrix o n cooling, which will counter the driving force
(surface energy minimization) for sintering ( Raj and Bordia, 1989; Kellet
and Lan ge, 1989 ). Thus the densification rate or tbe matrix will, in general,
be reta rded in the presence of reinforcements ( Bordia and Raj, 1988; De
Jonghe er al., 1986; Sacks et a l., 1987; Rahama n and De Songhe, 1987;
Prewc , 1986). Whiskers or fibers may also give rise to the phenomenon of
bridging, which is a fu nction of the orientation an d aspect ratio of the
reinforcement.

7.1.1 Hot ~11j(

Some form of hot pressing is frequ ently resorted to in the consolidation stage
of CMCs. This is because a simultaneous applica tion of pressure and high
tempera ture can accelerate the rat e of densification an d a pore-free and
fine-grained compact can be obtained. A common varia nt, called the slurry
injillrali01l process, is perhaps the most impo rta nt tech nique used to produce
conti nuo us fiber reinforced glass and glass-ceramic composites ( Phillips,
1983; Cornie et al., 1986; Prewo an d Brenna n, 1980; Brenna n and Prewo ,
1982; Sam bell et al., 1974 ). The slurry infi ltration process involves two
stages:
I. Incorpora tion of a reinforcing phase into an unconsolidated matrix.
2. Matrix consolida tion by bot pressing.
Figure 7.1 sho ws a schematic of this process. In addition to incorporation
of the reinfo rcing phase , the fi rst stage involves some kind o ffiber alignment.
A fiber tow or a fiber preform is impregnated with a matrix-containing slurry
by passi ng it through a slurry tank . The impregna ted fiber tow or preform
sheets are simila r to the prepregs used in po lymer ma tri x com posites. The
slurry consists of the matrix powder, a carrier liq uid (water or alcohol), and
an orga oic binder. The organic binder is burned out prior to consolidation.
Wetting agen ts may be added to ease the infiltration o f the fiber tow or pre-
form. The impregnated lOW or prepreg is wound on a drum and dried . This
is follo wed by CU lling an d stacking of the prepregs a nd consolidation in a
ho t press. The p rocess bas the advantage that, as in PMCs, the prepregs ca n
be arranged in a: variety o f stacking of sequences, e.g. , unidirectiona l, cross-
21 4 7. Ceramic Matri1 Composites

-=
-
o;- -~.

!f'J
mv
-
~

AI- 7.1. Scllnnalic of lhe slurry im~gnalion process.

plied (CY' /9(1' / CY' /9(1' etc.), or angle-plied (+0/- 0/ + 0/ -0 etc.). Figure 7.2a
shows an optical micrograph of a transverse section of a unidirectional
Nica lon fiber/glass matri x composite. Some porosity can be seen in this pie-
ture . Figure 7.2b shows the pressure and tempe rature schedule used during
hot pressing of a typical C MC.
The slurry infiltra tion process is well suited for glass or glass-ceramic
matrix composi tes, mainly because the processing temperatu res for these
materials are lower than those used for crystalline matrix materials. Any hot
pressing proa:ss has certain limitations on producing com plex shapes. The
fibers should suffer little or no da mage during handling. Application of a
very high pressure can easily damage fibers. Refracto ry particles of a crys-
talline ceramic can damage fibers by mechanical contact. The reinforcement
can also suffer damage from reaction with the malri1 at very high processing
temperatures. The matrix should have as little porosity as possible in the
final product as porosity in a structural ceramic material is highly undesir-
able. To thi. end, it is important to completely remove the fugitive binder
and use a matrix powder particle smaller than the fiber diameter. The bOI
pressing operational parameters are also importa nt. Precise control within a
narrow working tempe rature range, minimization of the processing time,
and utilization of a pressure low enough to avoid fiber damage are important
factors in this final consolidation part of the process. Fiber damage and any
fiber/malri x interfacial reaction , along with its detrimental effect on the bond
strength, are unavoidable attributes of the hot pressing operation.
In summary, the slurry infiltration process generally results in a fairly
uniform fiber dislribution; low porosity and high strength values can be
 7.l.l"roceosingofCMCs 215

,.-
----- •
••
,

,
,
, ,---::-'-~:--~
' 00 150 _
o-J'
200

• """
Ftc. 7~ . AD optical nUcrocrapb at. ITaJlIWftW Illttioft at. unidirectioll&I Nicaloa
liber/lJlul matri. composile. ~ ~ POd lempentvn IICbedvk used dIJriq hoc
pm;sin, at. rypiW composile.
216 7. Ceramic Matrix Composites

obtained. T he ma in disadvantage of this process is that one is restricted to

relative ly low melting or low softening point matrix mate rials.
Whisker reinforced CMCs are generally made by mixing the whiskers
with a ceramic powder slurry , then drying and hot-pressing. Sometimes hot
isostatic pressing ( HlPing) rat her than uniaxial hot p ressing is used. Whisker
agglo mera tion in a green body is a major problem . Mechanical stirring and
adjustment of pH level of the suspension [ma trix powder-s-whiskers in water)
can be of help in this rega rd. Addition of whiskers to a slurry can result in
very high viscosity. Also, whiskers with large aspect ratios ( > 50) tend to
form b undles and clumps (Liu et aI. , 1991). Obtaining well-separated and
deagglomerated whiskers is of great importance for reason ably high-density
composi tes. Use of orga nic dispersants ( Barcla y et a t , 1987), techniques
such as agitation mixing assisted by an ultrasonic probe, and deflocculation
by a proper p H contro l (Yang a nd Stevens, 1990) can be usefully employed.
Most whisker reinforced composites are made at temperatures in the 1500-.
1900'C range and pressures in the 20--40 MPa range ( Homeny et a l., 1987;
Shalek et a I. , 1986).

7. 1.3 ReadioD SoDding ~

Reaction bonding processes similar to the ones used for monolithic ceramics
can be used to ma ke ceramic ma trix composites. A reaction bonding process
has the grea t advantage that problems with matri x shrinkage during densi-
fication are avoided. The other advantages a re:
Ra ther la rge volume fract ions o f whiskers o r fiber can be used.
• Multidirectional , continuous fiber prefonns can be used.
• The reaction bonding temperatures for most systems are generally lower
than the sintering tempera tures, so that fiber degrada tion can be a voided.
One great disadvantage of this p rocess is that high porosity is difficult to
a void.
A hybrid process involving a combi na tion of hot pressing a nd the reac-
tion bond ing technique can also be used ( Bhatt , 1986; Bhatt, 1990). Silicon
cloth is prepared by att rition milling a mixture of silicon powder, a polymer
binde r, and an organic solvent to obtain a dough of proper consistency.
Th is dough is then rolled to make a silicon cloth of desired thickness. F iber
mats a re made by filament winding of silicon carbide with a fugitive binder.
Th e fiber ma ts and silicon cloth are stac ked in an alternate sequence, debin-
derized, and hot pressed in a molybdenum die in a nitrogen or vacuum
environment. The temperature and pressure are adjusted to produce a
ha ndleable preform. At this stage, the silicon ma trix is convened to silicon
nitride by tra nsferring the composite to a nitrid ing furnace between 1100 and
1400 ·C. Typically, the silico n nitride matrix has about 30"10 porosity, which
is not unexpected in reaction bonded silicon nit ride.
 7.1 . Processing of CMCa 217

F1z. 7.3. Schem.a~ of the melt infil_

lnltion process. [Af\er Comic et Ill.
(1986), used with permission.]

0 '--'- : ;; j'--I o r - l nfiltro nl

o
o
o
o
o -'UJ- Pr ~orm

Heating
o I---ro""" coils
o
o
o
o
o
o
o
0L-_ _- - ',

7.1.4 Infiltration
Infiltra tion o r a preform mad e or a reinforcem ent can be don e with a ma tri~
material in solid, liquid , or gaseous form.

Liquid Infiltra tion

This technique is very simila r 10 liq uid polymer or liquid metal infiltra tion
( F ig. 7.3). Proper control or the Duidity or the liquid matrix is, or course , the
key 10 this technique. It yields a high-dcnsity matrix. i.e., no pores in the
ma trix. Almost any reinforcement geometry ca n be used to produce a vir-
tually flaw-rree composi te. The temperatures involved, however, are much
higher than those encountered in polymer or metal precessing. Pr ocessing at
such high tem perat ures ca n lead to deleteri ous chemica l react ions between
the reinforcement and the ma trix. Thermal expansion mismatch between the
reinforcemen t and the ma trix, the rather large temperature interval between
the processing tem perat ure and room temperature, and the low stra in to
failure or ceramics can add up to a formidable set or problems in prod ucing
a creek-free CMC. Viscosities or ceramic melts a re generally very high, which
ma kes the infiltration or preforms rather difficult. Wcttability or the rein-
forcement by the molten ceramic is anothe r item to be considered. Hillig
(1988) has d iscussed the melt infiltration processing or ceramic matrix com-
po sites in regard to chemical reactivi ty, melt viscosity, and welti ng or the
reinforcemen t by the melt. A preform made or reinforcement in an y form
Ifo r example, fiber, whiske r, or pa rticle ) having a network or pores ca n
be infiltrated by a cera mic melt by using capillary pressure. Application or
pressure or processing in vacuum can aid in the infiltration process. Assum-
218 7. Cerami<: Matril Compooiln

ing that the preform consists of a bundle of regularly spaced, parallel chan-
nels, one can use Poissuelles's equation to obtain the infiltration height, h:

h '" l rr;:s(J
where r is the radius of the cylindrical channel, I is the time, ,. is the surface
(J
energy of the infi.Itrant, is the contact angle, and" is the viscosi ty. The
penet ration height is proponi onal to the square root of time, and the time
required to penetrate a given height is inversely proponionalto the viscosity
of the melt. Pene tration will also be easier if the coruace angk: is low {i.e.,
better wettability), and the surface energy, )', and the pore radius, r, are
la rge. If the radius, r, of the channel is made too large , the capillarity effect
will be lost.
Infiltra tion of a fibro us preform by a molten intermeta llic matrix mate rial
under pressure has been succc:ssfully done [Nour bakhsh and Margolin, 1989).
Alwnina fiber reinforced intenneta l1ic matrix composites (e.g., TW, Ni]Al,
and Fe)Al matrix materials) have been prepared by prrssurr casting, also
called YpU'pu casting (N ourbakhsh and Margolin, 1989; Nourbakhsh et al.,
1990). The matrix a lloy is melted in a crucible in a vacuum while the fibrous
preform is heated separa tely. The mol ten matrix material (at about IOO ·C
above the melting temperature, T..l is poured onto the fibers and argon gas
is introduced sim ultaneously. A rgon gas press ure forces the melt to infiltrall:
the preform. The melt general1)' contains additives to aid wetting of the
fibers.
We may summ arize the advantages and disadvantages of differen t melt
infiltration tech niques lUI follows. Th e advantages are:
• The matrix is formed in a single processing step.
• A homogeneous ma trix can be obtained.
Th e disadvantages of infiltra tion techniques are :
• High melting points of ceramics mean a greater likelihood of reaction
between the melt and the rein forcement.
• Ce ramics have higher melt viscosities tha n metals; therefore, infiltration of
preforms is relativel y diflkult.
• The matrix is likely to crac k because of the differen tial shrinkage between
t he matrix and the rein forcem ent on solidification. This can be mini ·
mized by choosing components with nearly equal coefficients of thermal
ellpalUion.

7. 1.5 Dlndf'd O xidalion, or the LanxideT1"l ProceM

Yet another version of liquid inll1 tra tion is the directed oxidation process , or
the Lanxjde" process, developed by Lanxide Corp. ( Ijrquhan, 1991). One
of the Lanxide processes is called DIMO XlM , which stands for directed
 7.1. ProcessingofCMCs 219

G n>wt h b.rrlu

C<l1ll.p<ll ite {

CM C ptUt

Fig. 7.4. Schematic of the directed metal o~ ida ti on J)l"OCXSs of LaMidt: Corp. (Cour-
tesy of Lenxide Corp.)

metal oxidation process. A schematic of the directed metal oxidation process

is sho wn in Figure 7.4. The first step in this process is to make a preform ,
which in the case of a particulate composite can be a ceramic green body. In
the case of a fibro us composite, filament winding or a fab ric lay-up may be
used to make a preform. A barrier to stop growth o f the matrix ma terial is
placed on the preform surfaces. In this method, a molten metal is subjected
to directed oxidation, i.e., the desired reaction product forms on the surface
o f the molten metal an d grows outwa rd. The metal is supplied conti nuously
a t the reaction front by a wicking action thro ugh channels in the oxidation
product. Fnr exa mple, molten al uminum in air will get oxidized to a lumi -
num oxide. If one wants to form aluminum nitride, then molten aluminum is
reacted with nitrogen . The reaction ca n be represented as follows:
Al + air ..... AIlO j
AI + nitrogen ..... AIN
The end product in th is process is a three-d imensiona l, interconnected net-
work of a ceramic material plus abo ut 5-30% of unreected metal. When
filler particles are placed ne xt to the molten meta l surface, the ceramic net-
work forms arouod these panicles. As we said earl ier, a fabric made of a
220 , . Ceramic Matrix e."mposi!ei

contin uous fiber can also be used . The fa bric is coa ted with a proprietary
coating to protect th e fiber from highly reducing aluminum and to provide a
weak inte rface, which is desira ble fo r enhanced to ughness . Some aluminum
(6-7 " 1. "/0) remains at the end of th e process. This must be removed if t he
composi te is to be used at tem pera tures a bove the melting point of alumi-
num (660 · C). On the other hand, the presence of a resid ual metal can be
exploited to provide some fract ure toughness in these composites.
Proper cont ro l of the reaction kinetics is of great importance in th is pro-
cess. The process is potentially a low-cost process because near-net shapes
are possible. A lso, good mechanical properties (such as strength and tough-
ness) have been reponed ( U rq uhart, 1991 ). Figure 7.5 shows some fiber
reinforced ceramic com ponents made by Lam lde Corp. Figure 7.Sa shows
some fiber reinforced ceramic composites for applications in high-temperat ure
gas tur bine engine components, while Fig. 1.Sb shows heat exchanger and
radian t burner tu bes, flame tu bes, and other high-temperat ure furnace parts
made of particle reinforced ceramic composites.
The main d isadvan ta ges of the Lanxjde processes are:
• It is difficult to contro l th e chemist ry a nd prod uce an all-ceramic matrix by
th is method. There is al wa ys some residual me ta l, which is not easy to
remove com plet ely.
• It is di fficult to envision the use of suc h tec hniq ues for large, com plex
parts, such as those required, say, for aerospace application s.

7, 1,6 In S itu Chemical Reaction Techniques

In sit u chemical reaction techniques to produce C MCs are extensions of
those used to produce monolith ic ceramic bodies. We descri be below som e
of th e more important tec hniques, viz., chemical vapor deposition (C VD)
and chem ical vapor infiltration (CVI) and different types o f reaction bond-
ing tec hniques.

Chtmical Va por I>qlosItion and Chemical Va por ImJl"'ll:nation

When the chemical va por depositio n (C VD) technique is used to im pregn ate
rather large amoun ts of matrix material in fibrous preforms, it is called
cht mical oopor imp" gM lion (CVI). Common cera mic matrix materials used
are SiC, Sh N 4 , and H re. The CVI method has been used s~fully by
several resea rchers to im pregnate fibro us prefo rm s ( Fitzer and Hegen, 1979;
Fitze r and Schlic ht ing, 1980; Fitzer and Gadow, 1\186; Stinton et al., 1986;
Burkl and et al., 1\188). The prefonns can consist of yams, woven fabrics, or
three-dimensional shapes.
In very sim ple terms, in the CVI process a solid material is deposited fro m
gaseous reactan ts onto a hea ted substrate . A typical CV D or C VI process
would require a reactor with t he following parts:
 7.\. Processing of CMCs 211

•
~111.7.5. Some commercially a...ailable liber-fcinfort:e<l <,:et'IlJnic componenlll made
by Lanxide Corp. • Fiber reinforced QCrami<.: composites for applications in high-
tem perature gas lurbine engine component&. b Heat excbang<:r a nd radiant burner
tubes, !lame tubes, and otlxr high-tcmpcratun: furnace patti made or particle-
remjoreed QCramie composites. (Courtesy of unxil!e Corp.)
222 7. Cera mi<:: Mat ri~ Composile!l

Fib rous
preform
Gases out '.;::=:;-l
o o Heating
o ~ coils

o o
o o
o o
o o
o o
o o
Gases in -=~~J
•
AI- 7.6. Schema tic of an i!iOtherm.a1chemical vapor infiltration JH"OC'""S-

I. A vapor feed sy~tem.

2. A CYD reactor in which the substra te is heated and gaseous reactants a re
fed.
3. An effluent system where exhaus t gases a re handled.
The chemistry for making ceramic fibers by CYD was given in Chapter 2.
The basic chemistry to make a bulk ceramic matrix in and around fibers in a
preform remains the same. One can synthesize a variety of ceramic ma trixes
such as oxides, glasses. ceramics, and in termetallics by CYD. Commonly,
the process involves an isothermal decomposition of a chemical compound
in the vapor form to yield the desired ceramic matrix on and in between the
fibers in a preform . Figure 7.6 Sh OW5 a schematic of such an isothermal
process. For exam ple, methyltrichlorosilane (CH,SiCh ), the starting mate-
rial to obtain SiC, is decomposed a t between 1200 and 1400 K:
CH10,Si (g) ..... SiC{s) + 3 HO (g)
The vapors of SiC deposit as solid phase on and between the fibers in a free-
standing preform to form the ma trix. An exam ple of the microstructure of a
SiC/SiC composite obtained by CYI is shown in Figure 7.7. The CYI process
is very slow because it involves diffusion of the reactant species to the fibrous
substrate, followed by outflow of the gaseous reacta nt products. The CYI
process of making a ceramic matrix is, indeed, a kind of le w-stress and low-
 7.1. Processing of CMCs 223

Fill. 7.7. An e~ample of the microstructure of a NicaJon fiber/SiC rnatri~ composite

obtained by CVT. (Courtesy of R.H. Jones.)

temperat ure CVD process, an d thus avoids some o f the p roblems associated
with high-temperature cerami c processing. However, when the CVI process
is carried o ut isothermally, surface po res tend to close first, restricting the
gas flow to the interior of the preform. This necessitates multiple cycles of
impregnation, surface machining, a nd reinfiltration to obtain an adequate
de nsity. One ca n avoid some of these p roblems to some extent by using a
forced gas flow and a temperature gradient. A schematic of one version of
this process is shown in Figure 7.8 (Stinton et a I., 1986). A graphite holder
in contact with a water-cooled metallic gas distributor holds the fibrous
preform. The bottom and side surfaces thus stay cool while the top of the
fi brous preform is exposed to the hot zone, creating a steep thermal gradient.
The reactant gaseous mixture passes unreacted through the fibrous preform
because of the low temperature. When these gases reac h the hot zo ne, they
decompose a nd deposit on and between the fi bers to form the matrix. As the
matrix material gets deposited in the hot po rtion of the preform , the preform
density and thermal co nductivity increase an d the hot zone moves progres-
sively from the top of the preform toward the bo ttom. When the composite
is formed completely at the top and is no longe r permeable, the gases flow
radially through the preform, exiting from the vented retaining ring.
This va riant o f CVI, which co mbines forced gas flow a nd temperature
gradient, avoids some of the pro blems mentioned ea rlier. Under these modi-
fied co nd itions, 70 to 90% dense SiC and Si lN4 matrices can be impregnated
224 7. Cerami<: Matrix Composi tes

Exl>atJ s! gas

Infiltrote<!
~sit e

Fibrous
pre form

,
."
Coat,ng

Fil- 7.8. Schematic o f a chemical vapor infiltrat ion process with pressure and tem·
perature gradients. [After Stinlon el al. (1986).]

in SiC and SijN 4 fibrous p reforms in less than a day. Unde r condition s of
plain CVI, il would take several weeks to achieve such densities, i.e., one ca n
reduce the processing time from several days to less tha n 24 hours. One can
also avo id using binders in this process with their attend ant problems of
incomplete removal. The use of a graphit e holder simplifies the fa brica tion
of the preform , and the applica tion of a moderate press ure to the preform
can result in a higher-tha n-no nnal fiber volum e fraction in the final prod uct.
The final obtaina ble densi ty in a ceramic body is limited by the fact tha t
closed porosity starts at about 93- 94% of theoretical density. It is difficult to
impregnate past this po int.
Advanta ges of a CVI technique or an y variant thereof include:
• Good mecha nical properties at high tem peratures.
• Large, complex shapes ca n be produced in a near-net shape.
• Considerable flexibility in the fibers and matrices that ca n be used (oxide
and nonox ide).
Among the disadvantages, one should mention that the process is slow and
expensive.

Reacth e COftSOlid. tioDor Liquid-Phase SiDtering

SillcQlt;;oed silinm curbi~ is the name give to a composite of SiC grai ns in
a silicon matrix. Molten silicon reacts with ca rbo n fibers to form SiC. The
original geomet ry ofthe ca rbon fibers is retai ned. Carbo n fiber in the fo nn of
cloth, lOW, felt, or matte is used as a precursor. A preform is made of ca rbon
fiber and infiltrated with liq uid silico n. Silicon reacts with carbon fibers to
 7.1. Proce..inll of eMe s US

form SiC fibers in a Si ma trix. Typical composition of the resultant compo-

site is Si ( 30~ SO'%) + SiC fiber. The silicon matrix limits the use temperature
to abo ut 1400 "C. A big advantage of Siq Si composites is that the con-
stituents are in chemical equilibrium and they have dosely matched thermal
expansion coefficients,
In another version of this process, a liquid phase forms as a result of
an exothermic reaction between elementa l powders. A good exampl e is that
from the field of imermetallics, e.g., nickel aluminides. The following steps
are involved:
I. Mix nickel and aluminum in stoichiometric proportions.
2. Cold isostatic press to 70% theoretical density to obtain a green body.
3. Vacuum encapsulate the green body in a 304 stai nless-steel can.
4. Subject the can ned material to reactive ho t isosta tic pressing.

7,1,7 SoI~I.nd Polymer Pyrolysis

Sol-gel and polymer pyrolysis techniques, which ha ve been used to ma ke
conventional ceram ic materials, can also be used to make cera mic matrix
materials in the interstices of a fibrous preform. We described the sol-gel
technique in Chapter 2. Very briefly, a solution conta ining metal com-
pounds, e.g., a metal alkoxide, acetate, or halide, is reacted to form a sol.
The sol is convened to a gel, which in tum is subjected to controlled heating
to produce the desired end product: a glass, a glass-ceramic, or a ceramic.
Cha racteristically, the gel-to-ceramic conversion temperature is much lower
than that required in a conventional melting or sintering process. Some of
the advantages of these techniques for making composites are the same
as the o nes for monol ithic ceramics, viz., lower processing temperatures,
greater composi tional homogeneity in single-phase ma trices, potential for
prod ucing unique multiphase matri x materials a mong ot hers. Specifically, in
regard to composite material fa bricatio n, the sol-gel technique allows pro-
cessing via liquids of low viscosi ty such as the ones derived from alko xldcs.
Covale nt ceramics, for example, can be produced by pyrolysis of polymeric
precursors at temperatures as low as 1400"C and with yields greater than
those in cve processes. Among the disadvantages of sol-gel are high
shrinkage and low yield compared to slurry techniq ues. The fibe r network
provides a very high surface area to the matrix gel. Consequently, the
shrinkage during the d rying step, frequently, results in a la rge density of
cracks in the matrix. Generally, repeated impregnations are required to
prod uce a substantially dense mat rix.
It is easy to see that many of the polymer handli ng a nd processing tech-
niques can be used for sol-gel as well. Impregnation of fibrous preforms in
a vacuum and fi lament winding are two importa nt techniques. In filament
winding, fiber tows or rovings a re passed thro ugh a tank containing the sol,
and the impregnated tow is wound on a mandrel to a desired shape and
226 7. Ceramic Matrix Composi tes

thickness. The sol is convened ( 0 gel and the struc ture is removed from the
mandrel. A final heat treatmen t then converts the gel to a ceramic or glass
matrix.
The sol-gel technique can also be used to p repa re prepregs by the slurry
infiltration metbod. The sol in the slurry acts as a binder and co ats fi bers and
glass particles. The binder burnout step is thus eliminated because the
binder, being of the same co mposition as the matrix, becom es part of the
glass ma trix. An advantage of this sol-gel-based slurry method is that con-
solidation ca n be done at lower temperatures . Amo ng the problems, one
sho uld mention that the coating layer on the fiber is porous, carbon-rich, and
nonuniform .
Polymeric precursors can also be used to form a cera mic matrix in a
composite. The matrix selection criteria include:
• High cha r yield
• Low shrinkage
• Good mech anical properties
• Easy fabrica tion
To make a silicon ca rbide matrix, polycar bosilane would appear to be a
natural choice as a precursor to produce a silicon ca rbide matrix, in view of
its successful use to mak e Nicalon (SiC) fiber. Polysilane is another possi-
bility. Its char yield is about 60"10. One can use fillers to reduce the shrinka ge.
Repeated infiltra tion and in situ thermal decomposition of po rous
reaction-bonded ceramics, such as silicon ca rbide an d silicon nitrate with
silaza nes an d po lycarbosilanes, give ShN. /SiC composi tes. The organ ic sili-
con co mpo und is thermally decomposed in situ. Fitzer and Gadow (1986),
for exa mple, followed these steps given to mak e such co mposites:
I. Porous SiC o r SiJN. fibrous preform with some binder phase is prepared .
2. Fibrous preform is evacuated in a n a utoclave.
3. Samples are infiltrated with mol ten precursors, silazaees, or po lyca rbo-
silanes, at a high temperature (780 K) an d the argon or nitrogen press ure
is slowly increased from 2 to 40 MPa . The high temperature results in a
transformation of the o ligomer sila ne to po lycarbosilane a nd simulta-
neo us po lymerization at high pressures.
4. Infiltrated samples are coo led an d treated with solvents.
5. Samples are placed in an autoclave and the organosilicon polymer ma trix
is thermally decomposed in an inert atmosphere a t a high pressure and at
a tempera ture in the 800-1300 K range.
6. Steps 2 through 5 a re repeated to attain an ad equate density.
7. To prod uce an optimum matrix crystal struc ture, the ma terial is annealed
in the 1300- 1800 K range.
In all these methods involving use of polymeric precursors, one must
reso rt to repeated impregna tions to imp rove the density. Typically, the
amount of porosity will reduce from 35% to less than 10% after abo ut five
impregnations.
 1.2. Interface in CMC. 227

7.1.8 SeIf-PropaCating Hlgb-Tem.-at\lR S)'lItheslf (SHS)

The SHS technique involves synthesis of compounds without a n external
source of energy. One exploits exothermic reactions to synthesize ceramic
compounds, which are difficult to fabricate by conventional techniques. For
example, one ca n mix titanium powder a nd carbon black, cold press the
mixture, an d ignite the compact at the top in a cold-walled vessel. A com-
busn on wa ve will pass through the compact, giving titanium carbide.
Among the salient features of SHS a re:
• High combustion temperature (up to 4OOO ' C).
• Simple, low-cost equipment.
• Good control of chemical composition.
• Different shapes and forms can be obtained.
This technique can be used to produce a variety of refractory materials. The
main disad vantage is that SHS products are very porous, because of the
fairly large porosity in the original mix of reactants a nd because of the large
volume change that results when the reactants tra nsform to the products.
Any adsorbed gases at the eleva ted temperatures used d uring this process
can also add to the porosity of the final product. Synthesis concomitant with
densificatio n can improve the situa tion to some extent. This involves appli-
cation of high pressure during the combustion or immediately after the
completion of the combustion reaction, when the product temperature is still
quite high. Hot pressing, rolling, and shoc k waves a re some of the techniques
used to ap ply the necessary pressure.
Many ceramics, suc h as bo rides, ca rbides, nitrides, silicides, and sialons,
and composites, suc h as Sic.. + Ah O), have been synthesized by means of
SHS. The SHS process gives a weakly bonded compact. Therefore, the pro-
cess is generally followed by breaking the compact, milling, and consolida-
tio n by some technique such as HII>ing. Explosive or dynamic compaction
can result in a relati vely dense product. A good example of a n SHS process
to make composites is the proprietary process of Martin Ma rietta Ccrp.,
called the XDTM process, wherein exothermic reactions are used to produce
multiphase alloy po wders. These are hot pressed at -1 4SO oC to full density.
Reinforcement in the fonn of particles, whiskers, and pla telets ca n be added
to the master alloy to make a composite. A good example is that of TiB1
pa rticles, about I urn in diameter, distri buted in intermetallic matrixes such
as TW, TW + T iJAI, an d NiAl.

7.2 Interface in CMCs

In general, for C MCs one must oatisfy the following compa tibility requi re-
ments: thermal expa nsion compatibility an d chemical compatibility. Ceram-
ics have a limited du ctility, and in lhe fabricatio n of CMC. one uses high
tempera tures. Thll5, thennal mismatch on cooling from high temperatures
228 7. Cerami<:: Matri~ Composit..

can cause matrix (or fiber) cracking. Thermal strai n in composites is pro-
portional to 6'A11T, where &l = '1.[ - (1"" where '1.[ and 'I..., arc the linear
expansion ooefficienu of the fiber and matrix, respectively, and A T is the
tem perature interval. There is, of course, another complication, namely,
tiber expansion coefficients are also sometimes not equal in the axial and
radial directions. Carbon fiber in pa rticular has the following ax ial and
rad ial coefficients:

If A:I is positive, the matrix is compressed on cooling, which is beneficial

because it leads to an increase in the tensile stress at which ma trix cracking
will occur. Conversely, if 6 01 is negative, the matrix experiences tension ,
which, if A T is sufficiently large, can cause matrix crack ing. Tn the radial
direction , if AOl , is positive, the fibers tend to shrink away from the matri x on
cooli ng, which results in a reduced interfacial bond strength. If, ho wever,!1:l
is negative, the fiber matrix bond strength can even be improved.
Ma trix cracking resulting from thermal mismatch is a more serious
problem in short fiber composites tha n in continuous fiber composites. The
reason is that in a ceramic matrix containing aligned ccminuous fibers,
transverse microcracks appear in the matrix, bu t fibers continue to hold the
various matrix blocks together an d the composit e can still display a reaso n-
able amount of strength. Tn a randomly oriented short fiber composite,
owing to increased stress at the fiber ends, matrix cracking occurs in all
directions and the composite is very weak. We can define a thermal expan-
sion mismatch parameter for axial and radial directions in an a ligned fiber
composite as (Sambell et al., 1972):

Table 7.1 makes a compa rison of damage resulting from thermal expansion
mismatch in some carbon fiber reinforced ceramic matrix composites. Note
that only glass and glass-ceramic matrices show no damage.
Chemical compatibility between the ceramic matri x and the fibe r involves
the same thermodynamic and kinetic considerations as with other composite
types. Quite frequently, the bo nd between fiber and ceramic matrix is simple
mechanical interloc king. During fabrication (by hot pressing) or during
subsequent heat treatments, the tiber(matrix bond could be affected by the
high temperatures attained because of any chemical reaction between the
fiber and matrix or because of any phase cha nges in eithe r one of the com-
ponents. Sambel1 et al. (1972) studied the zirco nia-reinforced magnesia
composite system in which there occurs a chemical reaction at 1600' C, At
temperatures less than l600 "C, the composites showed a weak fiber/matrix
interface and fiber pullout ooxun-ed during me<:hanical polishing. Upon heat
 7.2. Interface in CMCs 229

Tillie 7.1 A comparison of damage resulting from the thennal e~panliOll. mismatch in
some carbon fiber-reinforced SYltcm.s"

~. ,.• E
"- • ,:
rq ,:
..
M al rix (1O-a"C' ) (O Pal (M Pa) Dama,.

MoO
Al,O, 180'10 dcnoe)
Soda _lim< II!uo
1l<>roSlicalC g\us
Gl......,.ramic
•., ,...
13.6

•••
""
"'00

... .
,' ID
see
,~

sc
'00
zce
'00
'00
'00 ""
B

"" ,• "'"' . m

- l. ~
- 6,S
~crack ing
seven: cracking
I.ocaliud """kf
U n<:rackod
Uncrac ked

"Type I ca rbon tiben a" '" 0, Ill,. '" 8 ~ 10-' K ' .

. ... ;1 the milrix tbcnnal •• pansiOll codlicient.
< T, iI tbc lCmpenl l ure bel"", which hltle II..... relantioD ca n occur.
4,,_ iI the maori. IItr=gth.
• ~. and~• • re tbc tbcrm.ol expom;.on mismateh po.ramttcrI,
.s:o..r.v: Adapled wil h permi""'" from Phillipo (1 983).

treating at 1600 °C, however, beca use of the interfacial reaction and the
resultant improved bonding, no dam age was observed upon mechanical
polishing. Heat treatment at 1700 °C resulted in the complete destruction of
the zirconia fibers and the distri bution of zirconia 10 grain boundaries in
magnesia . Thus, as we noted in the case of MM e s (Cha p. 6), it is of the
utmost importance to be able to control the interfacial bond by means of
controlled chemical reactions between components. Chokshi and Porter
( 1985) also observed a reaction layer on the fiber surface after creep testing
in air. Auger electron microscopy analysis showed a mullite layer wit h large
glassy phase region s along grain boundaries. The following interfacial reac-
tion was proposed :
2 SiC + 3 AI10 J + 4 0 = AI6Si10 u +2C
and reaction kinetics were modeled by an equation of the form
~ :>: Dt

where x is the reaction zone thickness, D = D.eJlp(-QlkT), Q is the acti-

vatio n energy, k is Boltzmann's constant, T is the temperature in kelvin, and
t is the time in seconds.
The nature of the bond between fiber and ceramic matrix is thought to be
predominantly mechanical. In carbon fiber reinforced glass or glass-ceramics
there is little or no chemical bonding ( Da vidge, 1979; Phillips et al., 1972;
Prewo, 1982 ). Evidence for this is the low transverse strength o f these com -
posites a nd the fac t tha t one does not sec matri x material adhering to the
fibers on the fracture surface. The bond is thought to be entirely mechan ical
with the ceramic matrix penetrating the irregularities present on the carbon
surface . Shear strength da ta on carbon fiber in borosilicate glass and lithium-
al uminosilicate glass-ceramics ( LAS) show tha i the borosilica te glass com-
no 1. Cera mic Mat rix Composites

posites have double the shear stre ngth of LAS composites. The reason for
this is the different radial shrinkage of fibers from the matrix du ring cooling.
Ca lculated radial contractions of fiber from the matrix are 2.4)( 10- ' m for
the LAS ceramic and 0.9 )( 1 0~' m for the glass...:eramic composite ( Phillips,
1983). Shrinkage reduces the mecha nical interlocking and thus the fiberl
matrix bond.
In ceramic matrix composites, interfacial roughness-ind uced radial stress
will affect the interface debonding, the slidi ng friction of debcnded fibers,
and the fiber pullout length. Fiber pullout is one of the importa nt energy
dissipati ng fracture processes in fiber reinforced ceramic or glass matrix
com posites. An absence of strong chemical bond an d a purely lne(;hanical
bond at the fiber/matrix interface is highly desirable for the fiber pullout to
occur. In regard 10 the mechanical bonding, a number of researchers have
pointed out the importance of interfacial roughness in ceramic matrix corn-
posites (C MCs) (Jere , I'Wll; Jero and Kerans, 1'Wll; Carter et al., 1991; Jero
er aI., 1991; Kerans and Panhllsarathy, 1991; Mackin er al., 1992; Murnrn
and Faber, 1992; Yenkatesh and Chawla, 19'92; Chawla et et., 19'9 3a, 19'93b;
Sorensen, 19'93). As shown by Chawla and coworkers ( Venkatesh and
Chawla, 1992; Chawla el al., 1993a , 1993b), even when the coefficient of
therma l expansions of the coating, fiber, and matrix. are such that a radial
tensile stress exists at the fiber/coating interface after cooling from an ele-
vated processing tem perat ure, fiber pullout may not occur becau se of a
stro ng mechanical bonding d ue: to a rollgbness--induo:d clam ping a t the
fiberfmatrix. Thus, a tensile therm al stress in the radial at the interface is
desira ble factor. A rad ial tensile stress at the interface will encourage fiber
debonding and slippage, which in turn result in high toughness and high
work of fracture. The sliding resistance has been expressed as (Hutchinson
and J enson, 1'Wll)

t = t~ - 1'(1. {(I < OJ

(17. e 0)
where 17. is the stress normal to the interface, I' is the coefficient of friction,
and t . is the sliding resistance wilen (12 is positive (tensile). Kerans and
Parthasa rathy (1991) included interface surface roughness in a detai led
treatment of fiber debo nding an d sliding during both pU$hout and pullout
experiments. They modeled the effect of fiber surface roughness as an
increase in the interfacial strain mismatch, lead ing to an increase in the
interfacial normal pressure. The effective normal stress 17. at the interface
due to residual thermal stresses and roughness-induced stresses can be
written as

U. = (I, + U.

where (I, and iT, a re thermal and ro ughness-ind uced st~, respectively.
 7.3. Properties of CMCa 23 1

Th e thermal stress and roughness-induced stress can be expressed as

tI, = fJ(A~ T)
a. = fJ(A/ r)
where A:E is the difference in thermal expansion coefficients between ma trix
and fiber, AT is the temperature differential ( T~ - Til, fJ is a term co ntaining
elastic co nstants, A is the amplitude o f ro ughness, and , is the fiber radius.

7.3 Properties of CMCs

As mentioned earlier, the relative elastic modulus values of fiber and matrix
a re very important in C MCs. Th e ratio Ed E... dete rmines the extent of
matrix rnicrocracking. Typically, the strain- tc-Iracture value of a cerami c
matrix is very low. Th us, in MM Cs and in thermoplastic PMCs, the ma trix
failure strain (t",) is considerably greater than that of fibers. Most unrein-
forced metals show t", > 10% while most polymers fail between 3 and 5%
strain. Th us, in both MM Cs and PMCs fiber failure strain co ntrols the
composite failure strai n. Typically, fibers such as boron, carbon, and silicon
carbide show failure strain values of _ 1%. Compare this with the failure
strains of less than 0.05% for most cera mic matrix ma terials. The situation
in regard to fiber a nd/or matrix failure is shown in a simplified manner in
Figure 7.9. The crack-free original situatio n is shown in Figure 7.9a. In the
case: of MMCs and PMCs, fibers faU lirst at various wealr. points distributed
along their lengths. The composite will fail along a section that has a large
number of fiber fractures; see Figure 7.9b. In a stro ngly bonded CMC, fiber

j I 1
• b c d

Fig. 1.9. a Original CJ1ICIr.. frcc ";Iuatioll. b In the case of M MCI and PMC&, fibers fall
lint al variOU'l weak painlS distributed alon g their lengths. The composile will fa il
a100g a section thai has a large nwnber of fiber fractures. e In a strongly bonded
C MC, fiber an d matrix would fail simul taneously a l matrix fa ilLn strain. • In a
",eakly bonded C MC, hO'>Ve\'er, the matrix will start o;rao;kiog finl a nd the fibers will
be bridgi ng the matrix blocks.
232 7. Ceramic Matrix Composites

and matrix would fail sim ultaneously a t matrix fa ilure st ra in and a sit ua tion
similar to tha t shown in Figure 7.9c will prevail. In a weakly bonded CMC,
howev er, the matrix will start cracking fint a nd the fibers will be bridgi ng t he
ma trix blocks ( F ig. 7.9d). Thus, from a toughness poi nt of view, in general,
we do not wan t too strong a bond in a C MC because it would make a crack
ron thr ough the specime n. A weak interface , however, would lead to fiber-
bridgi ng of ma trix microcracks. Consider the sim ple isost rain model that
pre dicts a rule-of-mixt ures type relationship (see Chap. 10) for a unid irec-
tional composite. Let the composite be subjected to a strain ee' The isostrain
cond ition implies:
Ce = Cj = C",

where subscripts c. [ , and m denote composite, fi ber, and matrix , respec-

tively. This results in a ro le-of-mixtures relationship for st rength in the
longi tudinal d irection (see Chap. 10), namely,
f1, =(1j VI +(1., V..
where f1 is the stress, V is the volume fraction, and the subscripts have the
same meaning as given earlier. Not e tha t VI + V., = I. Then, assuming
elastic behavior for both the matrix and the fiber, we can write

stress carried by fiber _!!1.!L _ Ef VI

stress carried by matrix - 0 ", V., - E", V.,

As the composite is loaded, ma trix failure strain being smaller than that o f
the fiber, it will start showing microcracks a t some stress 0", as shown in
Fig ure 7. 10. We can write
(10 = Of Vj + ,,_( I - VI )
where 0 .... is the matrix stress at its breaking st ra in.

/ , - d.
Fiber bundle
f• failure
.~
•
•
o ' do
Mat rix
microerocking

Te ns ile st rO in
Fit. 7.1l). TenJiJe stress-strain <:wvc of an al igned CMC in the longitudinal dircdion
showing a damage-tolerant behavior. [Evans ( 1985), used with permission). As the
composite is loaded, matrix failure main being smaller than that of the fiber, micro-
c rack. appear in the matrix ar some ,tress 00, followed by fiber bundle fai lure and
pullout
 7.3. Propernes of CMCs 233

Tabk 7.2 Theoretical matri~ cra<.:king stresses for a bor-

osilicate glass and magnesia reinforced with 60"/. of high·
modulus carbon fiben (E = 360 OPal
e,
'-

-'"
Ma l<rial {G Pa) {MPa)

Borosilicate g1.... ec ' 00

Magn e.i a

So ~IU:
'"' '"
Adapted ooith pmni";"" from Phillips (l 9113l.

Rearra nging, we have

Table 7.2 shows some relevant parameters for borosilicate glass and ma g-
nesia matrix materials containing 60"/0 VI of carbon fibers (Phillips, 1983).
We note that in high matrix modulus composites (magnesia in the current
case), matrix cracking will be expected to occur at much lower stresses.
Th us, it would appea r that low-modulus glasses a nd ceramics o ffer some
advantages over high-modulus cera mic matrices. A CMC with even a micro-
crac ked ceramic matrix can retain some rea sonable strength (0'< "" 0"1 VI )
and there a re applications, such as bushings, where such a damage-tolerant
characteristic would be very valuable because in the absence of fibers brid g-
ing the cracks, the mo nolithic matri" would simply disintegrate. The dis-
advantage, of course, is that matnx microcracking provides an easy path for
environmental attack of the fibers and the fiber/matri" interface. Let us focus
attention on the tensile stress-strain curve of an aligned C MC in the longi-
tud inal direction, as shown schematically in Figure 7.10 (Evans, 1985). This
figure shows that C MCs ha ve damage-toleran t cha racteristics in uniaxial
tension. At a stress 0"0 , the stress-str ain curve shows a dip, indicating the
incide nce of periodic matrix cracking. Because the fibers have enough
strength to support the load in the presence of a damaged matrix (a very
desirable feature indeed), the stress-stra in curve co ntinues to rise until, at a
stress marked 0'. , the fiber bundle fails. At this point, the phenomenon of
fiber pullou t sta rts. The extent of this fiber pullou t region depends critically
on the interfacial frictional resistance. The fiber/matrix interface has a lot to
do with the form of the stress-strain curve. If the bonding is too strong,
ma trix cracking will be accompanied by a small amount of fiber pullout,
which is an undesirable characteristic from the toughness viewpoint, as we
shan see in Section 7.4. Both 0'0 and 0'. are insensitive to specimen or com-
ponent size because both strength levels are independent of ma trix flaws
( Evans, 1985). Thi s is in distinct co ntrast to the behavior of monolithic
cera mic materials, which show a significa nt size dependence. Increased stiff-
234 7. Ceramic Matri~ Composites

ness and strength were o bserved in a unidircctionally aligned com posite

consisting of conti nuous car bon fibers {50% V/ ) in a g1ass matrix compared
to the unn:i nforced glass matrix. {Da vidge, 1979). But more importantly, a
large increase in the work o f fracture occurred. The increased work of frac-
ture is a result of the controlled fracture behavior o f the composite, while the
unreinforced matrix failed in a catastrophic manner.
Fiber length, or more predsely, the fiber aspect ratio (lengthfd iameter),
fi ber orientation, rela tive strengt hs a nd moduli of fiber and matrix, thennal
expan sion mismatch, ma trix po rosity, and fiber Haws are the important vari-
ables tha t control the performa nce of CMCs. Sambell er al. (1974) showed
tha t, fo r ceramic matrix materials containing sho rt, randomly distribu ted
ca rbon fibers, a weak ening effect occurred rather than a stren gthening effect.
Th is was attribu ted to the stress concentration effect at the extremities of
randomly distributed short fibers a nd thermal expa nsion mismatch . Aligned
continu ous fiben do lead to a real fiber reinforcement effect. The stress con-
centration a t fiber ends is minimized an d higher fiber vol ume fractions can
be obtained. At very high fiber volume fra ctions, however, it beco mes diffi-
cult to remove matrix porosity. Figure 7.11 shows a linea r increase in stre ngth
with fiber volume fraction V/ up to -55% (Phillips et at , 1972). Beyon d
55% V/ , the matrix poros ity increased. The Young's modulus also increased

100 0
MP0 I
0 o MQKimum or
80 0 minimum 01
Q set

•,•
~ 60 0
i
0
I I stcooc re
error 01 .
0
the me a n
0
0

0
0
j
200

00 30 50 90 v/ o 120
Fibe r votu ree tra ct io n
Fl,. 7.11. A linear increase in strength with fiber volume fraction V/ up to _ SSY•.
[Af\cr Phill ips et al. (I972), used with p:rmiMion.)
 7.4. Toughness ofCMCs 23S

300 ,
GPo /
RUlle oj mixltureS-, /

•
-3 200 ,L."
"I . /
•m
5 10 01--... .- .
'" ~

~ v.. . "• Measured

0 Calculated, cor rect ed
for matr ix porosity

00 20 40 60 80 Yo \00
0

Fiber volume Ircction

~1&,7.12. A linear increase in Young's modulus with fiber volume fraction, VI . At

higher VI> il devia tes from linear owing 10 matrix porosity and possible fiber mis-
alignmerlt. [Phillip!! er a!., (11172).]

linear with VI (see Fig. 7.1 2), but at higher VI it deviated from linear owing
to matrix porosity and possible fiber misalignment ( Phillips et al., 1912).
Silicon carbide SiC whisker reinforced alumina is also more creep-resistant
than pclycrystalline alumina (Chokshi a nd Porter, 1985); see Figure 7,1 3.
Nole that the composite has a higher-stress exponent than the unreinforced
ma trix. Th e higher-stress exponent indicated a cha nge in the operating creep
mechanism. Th e exponent for polycrysta1line alumina is about 2, and this is
rationalized in terms of some kind of diffusion creep being the controlling
mechanism . A stress exponent of about S, which was observed for the corn-
posite, iii indicative of a dislocation creep mechanism being in operation.
Observation of the specimens deformed in creep in transmission electron
microscope showed dislocation activity. Figure 7.14 shows d islocations
ema nating from the whisker tips, probably resulting from high stress con-
centrations at these sites.

7.4 Toughness or CMCs

Many concepts have been proposed for augmen ting the toughness o f
ceramic matrix materials (see Chawla, 1993, for a summary). Table 7.3 lists
some of these concepts and gives the basic req uirements for the models to
be valid. Clearly, more than one toughness mechanism may be in operation
at a given time. Ma trilt microcracking, fiber/ma trix debon ding leading to
crack deflection and fi ber pullou t, and phase tra nsformation toughening
 7. Ceramic Malri~ Composi tes
'" ,
'0· Fla. 7.1l. Silicon carbide SiC
s , 0 Al 101 . 15 % SiC.. alwnina i.
whisk....- m nf<>r<:cd
0 AI O more I:lttp-resi.stanl than poly.
crystalline a lwnina (Chokshi and
'0 ·
, T:I 77J K
/ Porter, 1985). Note that the
composite has a higher stress

:'V/,
n po nent than the un reinforced
matri~.

•
5.2
ow

I
10· •

,
10
$. '0' • ,
1 10 d _ _10 MPo 10)

are all basically energy-dissipati ng processes that can result in an increase

in to ughness or work of fracture. Figure 7.15 shows schematically some of
these toughness mechanisms o r energy-dissipating mechani sms thai can be
brought 10 play in C MCs. Impressive work on ca rbon and SiC fiber rein-
forced glass and glass-ceramic composites has been done by Prewo a nd
coworkers ( Brennan a nd Prewo, 1982; Prewo et al., 1986; Fitzer and Hegen,

FIg. 7.t -ll. Dioloution. omanating

from the SiC wh isker tips, probably
raullins from high stresa concen-
trations al these !.ltes: TEM. (Cour-
tesy of A,H. Chotshi.)
 7.4. Toughness of CMCs 237

T. bIe 7.3. CmImie matrix composite toughening med lanisms

Mochaniom
, Compreui..e prestn:ssib'
, of the matrix
Cf'lOI;:k impodi"l
ReqwtemC1lt
"f > '"'" wiU rault in 1.1I alia! compraoive prnIressinl of
the rna,," af\er fabrica tion.
Ff'IIoCture ~ of the lICI:OfId phase (6bcn llI" panicles)
it. Jrealer than \hat of the ma,," locaUy. Cr.d< is either

, Fiber (.... ",hioker)

pullout
• Cf'lOI;:k dtfIcctioo.
arrested llI" _ ""'- ( ~ ... tcmiob droct).
Fihe.. ox whisk... havillJ hip Ira......... fraocturt: lOOP-
will ca.... failure a.Ioq fiber/l1Ill,," interfaoo leadin, 10
6ber pullout 011 further OltainillJ.
Weak 6ber/_ tria interfaca defIoct the PfOIIIlptin, <:rack

, Pbaoo traruformatiOll
a..ay from the princiiWol dirocUon.
The cra<k tip.trea lldd in the ....tria C&J\ ca.... tho ~
touRbenin, p/Iue p..-tideo (fiben) ltthe <:rack tip to lII>deJJo a p/Iue
lra nsfonnation cauan, .. panoion ( A- V > 0). The voI\1n>O
..pansiOll can oq_ the <:rack shut.

1979; Prewc, 1982). Their work showed that tough and strong CMCs could
be made. Extensive fiber pullou t and a controlled fracture behavior of the
CMC [i.e., a damage-tolerant fracture behavior) were the marked charac-
teristics. A scan ning electron microgr aph of a hot-~ssed A Ba-Si-AI-O-N
glass-ceramic containing Nicalon fibers composite is shown in F igure 7. 16.
( Herron and Risbud, 1986). The fracture surface of this composite ( Fig.
7.17) showed the phe nomenon of fiber pullout, indicating a weak fiberl
ma trix bond. The fibers were bonded to the matrix by an amorphous layer
whose cha racteristics changed with hea t treatment; a carbon-rich layer was
also obse rved on the fiber surface (Herron and R isbud, 1986). Rem nants of
the interfacial amorphous layer adhe ring to the Nicalon fibers can be seen in
Figure 7. 17. Another illustration of Nicalcu fiber pullout in a Nica1on/
pyrolitic carbon coating/SiC compos ite is sho wn in Figure 7.18 (Chawla et
aI., 1994). Among oxide cera mic matrix materials, al umina and mull ite have
a ttracted the most attention. In particular, SiC whisker reinforced alumina
com posites (20- 30"10 by volume of SiC whi skers in al umina, made by hot
pressing) showed im pressive gains in toughness and strengt h ( Becher and
Wei, 1984; Wei and Becher 1984; T iegs and Becher, 1986); see Figures 7. 19
and 7.20. A typical fine-grained monol ithic alumina has a toughness (X/c) of
4---5 MFa m 1f2 and a f1e:o;u ral strength between 350 and 450 MPa. AI~OJ
containing 20"/. by volume of SiC whiskers showed a Kk of8 -8.5 MPa m ill
and a flexural strength of 650 M Pa; these levels were maintained up to abou t
IOOO ·C.
If the crack gro wth can be im peded by some means, then a higher stress
would be req uired to make it move. F ibers (metallic or ceramic) can play the
role of to ughening agents in ceramic matrices. Melallic fiber- reinforced
cera mics will clearly be restric ted to lower temperatures than ceramic fiber-
reinfo rced ceramic matrices. Glass and glass-ceramic matrices contai ning
carbon fibers ( Phillips et at., 1972; Davidge, 1979; Prewo and Bren na n,
238 7. Ceramic Matrix Composites

F"tI:. 7.15. Schematic of some

tou8hnl:SS or energy.<Jissipat-
ing m«hanillml that can be
brousbl to play in CMCs.

._-
L
I
---
~-
•

1980; Prewc , 1982; Brennan and Prewo, 1982; Prewc et al., 1986) have been
shown to hav e fiber pullout as the dominant toughening mechanism. Basi-
cally, fiber pullout requires that the strength transferred to the fiber during
the ceramic matrix fract ure he less than the fiber ultimate strengt h, tI/o, and
l!utl an interfacial shea r stress be developed that is greater than the fiberl
matrix interracial strength, 1,; tha t is the interface must fail in shea r. For a
given fiber of radi us ' , we have the axial tensile stress in the fiber (see Chap.
10)

tll = 211(~)
where /, is the critica l fiber length and " I < "/0' The tensile stress increases
from a minimum al both fiber ends and at tains a maxim um along the central
ponion of the fiber (see F ig. 10.13). Fibers that bridge the fract ure plane and
whose ends terminate within 1./ 2 from the fract ure plane will suffer fiber
pullout, while those with ends further away will fracture when til = "/0. A
 7.4, ToughTlC'Sll or CMC. 239

2 0 ~m

fI&. 7.16. Scanningelectron micrograph of a hOI-pnssed Ba-5i-Al-Q-N glass-

cmlmic matri~fNicalon fibers ccmpceue. [From HeTT011 and Rishud (19&6), IIS<'<l
with permission.]

1.0~m

FIg. 7.17. Fract~ surface of SiC(Nicalon)/ Ba.Si.A I-Q-N cornpOOte showing fi~ r
nullnlll . lM.,..,.nn an" Rid",,! ~ 1 9l1~~ """,, wirh .....rmi ...;nn ~
240 7. Ceramic M atri~ Composill':ll

~111: .
7.18. Fiber pullout in a Nicalonlpyrolitic carbon coaling/SiC matri~ composite.
lCo urtl':lly of N. Chawla.)

crac k deflection mecha nism also requires a weak fiberfmatrlx bo nd so that as

a matri x crack reaches the interface. it gets deflected along the interface
rather than passi ng straight thr ough the fiber. This is illustra ted in Figure
1.21 ( Ha rris, 1980). The original sta te involving frictional gripping of the
fiber by the matrix is sho wn in Fig. 7.2 Ia . On stressing the composite, a
crack initia tes in the matrix and stans propagati ng in the matrix norm al to
the interface. As it approaches the interface, the crack is momentarily halted
by the fiber; see Figure 7.2 Ib. If the fiber/matrix interface is weak, then
interfacia l shear and lateral contraction of fiber an d matrix will result in
debonding and crack deflection away from its principal direction (norma l to
the interface); see Figure 7.2 Ic. A fwther increment of crack extension in the
242 7. Ceram ic Matrix Composilel

l~, 1

•
I=u.b
1rLL.L
c
• •
fl&. 7.21. "The cra<.:k deflcction mo:.:hanism mjuil'C5 a weal< fiber/rnatri;,; bond so that
as a rnatrix cra<.:k naches the intelfaot, il gen deflocted along the interface rather lhan
pas$ing .tra ight through the fiber. ThU: i. illUlltnited in after Harri . (1980)• • "The
original ltate involving frictional gripping of the fiber by the matri;,;; ~ the crack in the
mauU i. momentarily halted by the fiber; e inletfacial shear and lateral contraction
of fiber and rnauU mult in debonding and crack deflection along the interface; d
funller debonding, fiber failun at a weak point, and further extension, f broken fiber
fnd5 an pulled out agaill3l frictional ruimnce of interface, leading to tetal separa-
tion. [Rtprinted from lI arris (1981), used with pmni55ion.j

would appea r to be an elemen tary and straightforwa rd recommendation. If

processing results in large fla ws in the maw, the composite fracture strain
will be low. In this respect, fiber bndging of cracks in a CMC will result in a
reduced flaw size in the ma trix. This, in tum , w ill help achieve higher applied
strains before crack propagation in the matrix than in an uereinrorced,
monolithic ce ramic (A veston et al., 191 1). A weak interfacial bond, as

o,~~x Distance from not ch

Fig. 7.n . Stre.. distribution at" crack tip. [After Cook and Gordon (1964}.)
 7.4. Toughness ofCMQI 243

pointed out earlier, leads to crack deflection at the liberlmatrix interface

andlnr tiber pullout. Use of a high-volume fraction of continuo us fibers
stiffer than the matrix will give increased stiffness, which results in a higher
stress level being needed to producc matrix microcracking and a higher
composite ultimate tensile stress, as well as a high creep resistance. A high-
volume fraction and a small fiber diameter also provide a sufficient number
of fibers for crack bridging and postponing crack propagation to higher
strain levels. A small-diameter fiber also translates into a small I" the critical
length for effective load transfer from matrix to fiber (see Chap, 10).
Although a weak interface is desirable from a toughness point of view, it
provides a short circuit for environmental attack. An ability to maintain a
high strength level and high inertness at high temperatures and in aggressive
atmospheres is highly desirable.

Cr ack l>rileetlon Criteria

Cook and Gordon (19M ) analyzed the phenomenon of crack deflection or
the formation of secondary cracks at a weak interface in terms of the state of
stress at the crack tip. Let us consider a fiber/matrix interface, perpendicular
10 the main advancing crack. Cook and Gordon estimated the strength of
the interface necessary to cause a diversion of the crack from its original
direction when both fiber and matrix have the same elastic constant. At the
tip of any crack, a triaxial state of stress (plane strain) or a biaxial stress
(plane stress) is present. Figure 7.22 shows schematically the stress distri-
bution at a crack tip. The main applied stress component, ay, has a very high
value at the crack tip, and decreases sharply with distance from the crack tip.
The stress component acting nonn al to the interface, o», is zero at the crack
tip; it rises to a maximum value at a small distance from the crack tip and
then falls off in a manner similar to a, . Now, it is easy to visualize that if
the interface tensile strength is less than the maximum value of o», then the
interface will fail in front of the crack tip. According to the estimates of
Cook and Gordon, an interfacial strength of 1/ 5 or less than that of the main
stress component, a" will cause the opening of the interface in front of the
crack tip.
More sophisticated analyses of crack interaction with an interface ha ve
been proposed {He and Hutchinson, 1999; Evans and Marshall, 1989; Ruhle
and Evans, 1988; Gupta , 1991 ; Gupta et al., 1993). He and Hutchinson's
results give the conditions for liberlmatrix debonding in tenns of the energy
requirements; see Figure 7.23. A plot is made of G;/G/ vs. IX, where G1 is the
mixed-mode interfacial fracture energy of the interface, G/ is the mode I
fracture energy of the fiber, and e is a measure of elastic anisotropy as
defined in the Fig. caption. The plot in Figure 7.23 shows the conditions
under which the crack will deflect along the interface or propagate through
the interface into the fiber. For all values of GdGr below the dashed line
area, interface debonding is predicted. For the special case of zero elastic
244 7. Cerami<:: Matri~ Composites

1.. r------r-------,

t .•
No debond ing

...
• •1.0 - ... •
• - (E,: - El l! tE,: + El l
...
De bonding

1••

fit. 7.13. Fiber/mat"" debondiog criterion in terms of the energy requimnmt:s. Gt is

tl>o: mixed-mode interfacial fracture energy of the interface, G, il the llI<lde I fracture
energy of the fiber, and '" is • measl,II'e of elastic llPisotropy_ [AI'tcr He llPd Hut-
chin""" (1989).] E' .. E/ (I _ . 2).

mismatch, i.e., for ,:II: = 0, the fiber/mat rix interface will debond for GI/G,
less than about 0.25. Conversely, for GdG, greater than 0.25, the crac k will
propagat e across t he fiber. In general, for the elastic mismatch, with :l
greater than zero , the minimum inte rfacial toughness requi red for interface
debondi ng increases, i.e., high modul us fiber tends to favor debonding. On e
shortcoming of t his a nalysis is that it treats the fiber and matrix as isotropic
materials. This is not always true, especially for the fi ber. G upta et a t (1993)
have derived strength and energy criteria for crack dellection at a fiberl
matrix interface for several composite systems, taking due account of the
ani!lOtropic na ture of tne fiber. 11leir experimental technique, laser spa llation
experiment using a laser Doppler displacement interferometer , was described
in Chapter 4. By this tco.::hniq ue they can measure the tensile strength of a
planar interface . They ha ve ta bula ted the requi red values o f the inte rface
strength and fracture toughness for delamination in number of ceramic,
metal , intenneta llic, and polymer matrix composites.

7.5 Thermal shock resistance

Thermal shock resistance is a vet)' important characteristic wit h cera mics

and ceramic com posites tha t a re meant to be used at high tempera tures and
 7.5. Thermal shock resistance 245

must undergo thermal cycling. We define a thermal shock resistance (TSR )

parameter as
uk
Thermal S hock Resistance = £z

where (J is the fracture strength, k is the thermal conductivity, E is the

Young's modulus, and z is the coefficient of thermal expansion. Most ceo
ramics have low thermal conductivity which is one problem . But high a co-
efficient of thermal expansion, iX, will aggravate the situation further. Thus,
common soda-lime glass and alumina have a high z, about 9.10- 6 K_I. This
makes them very poor in TSR. If we reduce CaD and Na20 in the common
glass, and add B20 3 to form a borosilicate, we get a special glass that has
an e = 3.10- 6 K _ I . Such a glass will show superior TSR and is commerc ially
available under the trade names, Pyrex or Dura n glass. Glass-ceramics such
as lithium aluminosilicates (LAS) have an :x close to zero and thus excellent
TSR. LAS are used to ma ke the Corningwa re which can be ta ke direc tly
from the freezer to the cooking range or oven without causing it to shatter!
Boccaccini et a1. (1997a, 1997b) studied the cyclic thermal shock behavior
of Nicalon fiber reinforced Duran glass matrix composites. The thermal
mismatch between the fiber and the matrix in this system is almost nil. A
decrease in Young's mod ulus a nd a simultaneous increase in internal friction
as a function of thermal cycles were observed. The magnitude of internal
friction was more sensitive to microstructura l damage than Young's modulus.
An interesting finding of theirs involved the phenomenon of crack healing
when the glass matrix composite was cycled above the glass transition tem-
perature of the matrix. We discuss this topic further in Chapter 13.

7.6 Applications of CMCs

Ceramic matrix composites find applications in many areas. A convenient

classification of the a pplications of C MCs is aerospace a nd nonaerospace.
Ma terials-related d rivers for a pplications of C MCs in the aerospace field
are:
• High specific stiffness and strength leading to a weight reduction, and ,
consequently, decreased fuel consumption.
• Reduction in fabrication and maintenance cost.
higher operating temperatures leading to a grea ter therm al efficiency.
• longer service life.
• signa ture reduction.
CMCs can lead to improvements in aerospace vehicles including aircra ft,
helicopters, missiles, a nd reentry vehicles. Projected skin tempera tures in
future hypersonic aircraft a re higher than I600 C. Other parts, such as
Q

radc rnes, nose tips, leading edges, and con trol surfaces, will experience only
248 7. Ceramic Mat rix ComJlO5ilcs

Fig. 7.H. A can dle-type 61ler consisting of Nex tel™ 312 cera mic fibers in a sili~on
carbide matrix. TIle fil ter is 1.3 m long. S"",h tiltcn an: used (0 remove particula te
ma lter from high-temperalun: gas slre ams up 10 lOoo"C. (Courtesy of 3M Co .)

FIt- 7.16. An example of a support made of fiber reinforced glass. (Counesy of Scholl
G lIlowerke.)
250 7. Ceramic Matrix Composites

M .Y. H e and I .W . H utchinson (1989). J. App. Mn:It. , 56, 270.

M . Herron and S.H. Risbud (1986). Am. Crranl, Soc. Bull" 65, 342.
W.B, H illig (1988). J. A.....r , u rtlm. Soc., 7 1, C·96.
J. Hom eny, W.L. Va ughn, and M.K. Ferber (1987). Amt'r, Ct!" Soc. Bull , 61, 313.
J. W. H utchinson and H.M . Jemm (1990) . Mt!dt. Mark , 9,139-1 63.
P.D. Jere (1990). Am. CPam. Soc. Bull., 69, 484.
P.O. Jero a nd R.I. Kenns (1990). &ripra. Mt!I<lll , 24, 231 S- 2318.
P.D. Jero, R.J. Kerans, a nd T .A. Parthasara th y (199 1). J. Am, Cuam. Soc., 104,
2793-280 1.
B, Kellett a nd F.F. LaD~ (1989). J. Am. U ranL s«, 61, J69.
R.I . Ke ran s and T .A. Parthasarathy (199 1). J. Am. Ct!ranLsoc. 14 , IS85- 1596.
D, Lewis (1991). In Mt!ral Malrix Compo.<ilt!s: Pro«.JSing <JNJ Jorrt!,faus, Academic
Press, Boston, p . 121.
H ,Y . uc, N. Clausxn, M .J. H offmann, and G. Petrow (199 1). J. Eur.Crrant. Soc. ,
1, 4 1.
T.J, MackiD, P.O . Warren, and A.G. EvaDs (1992). A~la Mt!lall. Mau ,., 40. 12SI -
1151.
D.R. Mumm and K.T. Fa ber (1992). Ct!rtlm. Eng. d Sd P,oc., 1- 8. 70--n.
S. NourbaklLsh, F.L. Liang, and H . Margolin (1 990). Mt!I. Tr<UIJ. A, 21A, 213-
S. Nourbai:1Lsh and H. Margolin (1 990). Mi!/. Trwu. A, lOA, 2159.
D .C. Pllillips (1983 ). in Fahrirotio'l of eomposilt!s, North-Holla nd , Amsterdam,
p. 373.
D .C. Phillips, R .A.J. SambeU, and D.H. Bowen (1972). J Mal"'- Sci , 1,1454.
K.M . Prewo (1982). J. Mat"r. Set , 17, 3549.
K, M. Prewo (1986). In Tailoring Mllltip~ and Composiu C",omks, Vol. 20,
Ma teria ls Science Resoan:h, Plen um Prns, New Yori:, p. 529.
K ,M. Prewo a nd J,I . Brennan (1980). J. Mal" . Set , 15, 463.
K ,M. Prewo, 1.1. Brennan, and G ,K , La yden (1986). Am. C",om, Soc. Bull., 65, lOS,
M.N. Ra ha man a nd L.C. De Jon ghe (1987). J. Am. Crrant. Soc., 10, C-348.
R. Raj and R, K Bordia (1989). ACla M i!/. , 32. 100 3.
M. R uh le and A.G. Evans (1988). Mat" . Se( and £"11. , AI01, 187.
M.D. Sacks, H .W. Lee, and O.E. Rojas (1987). J. Am. C"tIm. Soc., 10 . C·34li.
RAJ. Sambdl, D .H. Bowen, and D.C. Pbillips (1912). J. Mau,. Sct , 7, 713.
RA J. Sambell, D.C. Phillips. an d D.H. ~ (1974). In Co.,boor Fih,,,s: "TMi, PIa~"
ill MtHkm Tn:MoIogy, the Pla stics Insti tute, London , p. 16f9.
L.J. Schioler and jJ . Stiglich (1986). Am. Ct!rtlm. Soc. Bull., 65, 289.
P.O. Shalek, 1.1. hlrovic, G .F. H urley, a nd f ,D , Gac (1986). A,....,.. Cnam, Soc.
BIIII., 65, 3SI.
B.F. Sorensen (1993). Scripta Metal'. el Maler., 28. 43S·439.
D .P. Stinton, A.J. Caputo , a nd R.A. Lowden (1986). Am. Ctfram. Soc, Bull., 65 , 347.
T.N. T iegs and P. F . Becber (1986). In Tai/o'illfl Multip~ and Composire u ,anria ,
Plen um Press, New Yori:o p. 639.
A.W. Urquhart ( 199 \). Mal"'- Sct big" A I" . 7S.
R. Veni:ate:sh an d K.K . Chawla (1992). J. Malrr. Sci , ll, 6S0----6S2.
o.c Wei and P.F. Becher (1984). Am. C",tIm. Soc. Bull., 64, 298.
P.A. Willennet, R .A. Pelt, a nd T.J . Whal en (I978). " Deve lopment and Processing
of Injection · Molda ble Reaction-Sintered SiC Composi tions," A....... CtfTam. Soc,
Bull" 57, 744,
M. Yan g and R , Stevens (1990). J. Ma rer, Set , 25, 46S8.
 Su~ted Reading 2j I

Suggested Reading

K .K . Chawla (1993). Cn amic MaIm: Compo.Jires, Chapman &. H aU, London.

K.S. Mazdiyami ( Ed.) (1990). CPamic FiMr hiJlf(H~ Composilrs, Noyes Pub .,
Pa rk Ridge, NJ.
D .C. Phillips (1983). " F iber Reinforced Ce:ramics," In FabricaliOll of u ramics, vol. 4
of Handbook <Jf Comporilrs, N<Jrth-Holland, Amsterdam, p. 373.
R. Warrell (Ed.) (199 1). Crramic Ma" ix Composilrs, Blac kie 4 Sons, Glasgow , U K .