Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
SECOND EDITION
CERAMIC MATRIX COMPOSITES
SECOND EDITION
by
K. K. Chawla
University of Alabama at Birmingham
Birmingham, AL
U.S.A.
Brihadaranyaka Upanishad
Ch.II.3
To
1. Introduction 1
2. Ceramic Matrix Materials 11
3. Ceramic Reinforcements 47
4. Processing 109
5. Interface 139
6. Micromechanics: Elastic, Thermal,
and Physical Properties 169
7. Mechanical Behavior: Load Transfer;
Monotonic, Cyclic, and Fatigue Behavior 205
8. Thermal Stresses 263
9. Interface Mechanics 291
10. Macromechanics 355
11. Applications 395
Appendixes 425
Indexes 431
PREFACE TO THE SECOND EDITION
Charles Dickens starts his A Tale of Two Cities with the following words:
"It was the best of times, it was the worst oftimes, it was the age of wisdom, it
was the age of foolishness, it was the epoch of belief, it was the epoch of
incredibility, it was the season of Light, it was the season of Darkness, it was
the spring of hope, it was the winter of despair ... " These words provide a good
characterization of the state of affairs in the field of Ceramic Matrix
Composites (CMCs) during the last decade ofthe twentieth century and the first
decade ofthe twenty first century. New continuous ceramic fibers have become
available and new innovative techniques of processing CMCs have been
developed. The Achilles' heel of monolithic ceramics is their extreme
brittleness. Since the publication of the first edition of this book, substantial
progress has been made in processing of continuous ceramic fibers and their
composites. A reinforcing phase (particle, short fiber or continuous fiber) can
improve the toughness of these materials without sacrificing the well known
advantages of ceramic materials, viz., enhanced wear resistance, hardness,
corrosion resistance, and high-temperature capability. A range of ceramic
matrix composites (CMCs) has been developed that combine a matrix material
with a reinforcing phase of different composition (such as alumina and silicon
carbide) or the same composition (alumina/alumina or silicon carbide/silicon
carbide). Major applications of ceramic matrix composites include cutting tools
and wear resistant parts. Other minor but significant applications include hot
gas filters, high pressure and heat exchangers, and gas turbine seals. On the
other hand, persistent problems include lack of toughness and poor oxidation
resistance at high temperatures. High cost (both raw materials and processing)
continues to be a major obstacle to widespread application ofCMCs. Another
major obstacle is the lack of reliability in current materials. Some niche
applications involving the use of CMCs have appeared but the ceramic engine
remains a distant goal. All this points to a belief! have held for a long time.
That belief is that, just as in the economics, in research also there is no such
thing as a free lunch. Invariably, it is a road that involves a painstaking and
persistent effort to study various problems involved and find solutions to them.
Along the way, there will be advances, some incremental, some more than that.
Eventually, there will be results to show. Research and development work in
the areas of CMCs will continue to be of paramount importance for the
foreseeable future. Notable progress has been made yet more work needs to be
done because the challenges are indeed daunting. My effort at bringing out the
second edition of Ceramic Matrix Composites is a step in that direction.
In this revised second edition, I have updated the information in regard to the
material systems for ceramic matrix composites, the reinforcements and
matrices, processing, properties, and applications of CMCs. I have brought it
up to date on the work done in the decade since the first edition came out. I
have added new sections and new figures in all the chapters. Based on inputs
I received from various faculty members and students, I have tried to clarify
concepts where needed. I am grateful to all these people who talked to me
personally or sent email messages in this regard. I am grateful to my program
managers at the US Office of Naval Research (primarily, Dr. S. Fishman and
later Dr. A.K. Vasudevan) who thought it fit to fund my research work in this
area. Other people who have been helpful, in tangible or intangible ways, to
make this second edition see the light of the day include: (in alphabetical order)
are: A. Boccaccini, K. Carlisle, N. Chawla, Z. Chen, C. Coffin, G. Gladysz, G.
Gouadec, M. Koopman, R. Kulkarni, B.A. MacDonald, D. Mendoza, A.
Mortensen, K.L.Murty, B.R. Patterson, B.V. Patel, J.J. Petrovic, J.M. Rigsbee,
S. Scarritt, H. Schneider, N. Stoloff, Y.K. Vohra, and A. Tobin. Finally I am in
an everlasting debt of gratitude to my family members, Nivi, Nikhil, Anita, and
Kanika for their forbearance. Anita and Kanika cheerfully bore all the burden
of producing a camera-ready typescript.
K.K. Chawla
Birmingham, AL
August, 2002
PREFACE TO THE FIRST EDITION
And:
... the classical model of materials application has been inverted. We once
sought applications for materials. We now have applications driving the
creation of materials. We now design materials for what we need (Press, 1990).
well as the energy cost are critical in the final cost. A study by the US National
Research Council (1989) has emphasized in no uncertain terms the importance
of synthesis, processing, characterization and performance of materials for
success in the international marketplace. Add to this the ever-important public
demand for a clean and healthy environment, and one can easily realize how
important the whole material life cycle of a given component is. I wish to
emphasize the item of environmental impact of the new materials and
processes, including the recyclability of materials. Ultimately, the engineered
materials must last longer, reduce material waste, be more energy efficient. I
believe that composite materials can contribute to a safe and healthy
environment.
Although metals, ceramics, and polymers make the three legs of what might be
called the tripod ofMS&E, increasingly it is becoming evident that the lines of
demarcation between traditional disciplines such as metallurgy, ceramics and
polymers are getting quite blurry. Consider the following. Ceramic materials are
being made from polymeric precursors, metals are being produced with a glassy
rather than crystalline structure while semi-crystalline polymers are finding
commercial applications. This intermingling of materials is most evident in the
field of composite materials where one has the ultimate objective of tailoring
a material having a specific set of characteristics starting from components
having different characteristics. Tying together process and microstructural
control to the desired performance goals in the final component or product is
the ultimate goal. In fact it is now recognized on all hands that in order to meet
the diverse and exacting demands, materials must be engineered at every step.
In this regard, mother nature has an abundance of lessons for us. Materials in
nature are tailored over a very large spectrum of length scale, from atomic or
molecular level to micro to macroscopic dimensions. Some very interesting
examples of nature's work can be found in collagen-based composite materials
such as skin, cartilage, bone, sea shells, etc. Nature has designed these
composites for multifunctional applications requiring sometimes flexibility and
strength and at other times resistance to various environments.
The theme ofthis book is: processing, structure, properties and performance of
ceramic matrix composites. My definition of the ceramic matrix is rather broad
for the purposes of this book. It includes inorganic silica-based glasses,
crystalline ceramics, glass-ceramics, intermetallics, and that very special
material, in elemental form, called carbon. All of these have an implicit
unifying thread in that they are fairly high-temperature structural materials.
This, I believe, is the first dedicated text on the subject of ceramic matrix
composites. There are, of course, many conference proceedings or multi-author
books available on the subject, for example, Warren, 1991 and Mazdiyasni,
1990. In this book, I have excluded cement and similar building materials,
mainly because they are not high-temperature structural materials. For anyone
interested in the topic of fiber-reinforced cement-based composites, there are
books available on the subject (Balaguru and Shah, 1992; Benkur and Mindess,
1990).
xiii PREFACE
I have aimed this book at the final-year undergraduate and first-year graduate
students in materials science and engineering as well as the practicing engineer
or scientist. I have used portions of the material contained in this text for a
senior undergraduate course, a first-year graduate course, and in short courses
for engineers from the industry. The respohse was very heartening. I hope that
the book will also obtain the same response from a much broader audience.
Except for the very basic ideas of materials science and engineering, I have
assumed very little prior knowledge of any special kind on the part of the reader
that is required to follow the material. The book may thus appear to be rather
pedantic, at places, to the more experienced reader. I apologize for that. There
are plenty of references and suggested reading material for the reader who
wants to dig deeper. I have provided problems at the end of each chapter in the
hope that by solving them, the reader will add to her/his understanding of the
material in the text.
to these people for important feedback. I am also very thankful to the following
for their generous hospitality during my sabbatical in 1992-93: J. Cohen, Y.-P.
Chung, I. Daniel, K.T. Faber, M.E. Fine, S.P. Shah and J.R. Weertman (all at
Northwestern University, Evanston, IL), B. Ilschner and F .. Rezai-Aria (Ecole
Polytechnique, Lausanne, Switzerland).
Thanks are also due to my family members, Nivi, Nikhil and Kanika, for
understanding my compulsion to undertake such time-consuming ventures. That
is the intangible part. They also rendered more tangible help in sorting things
out, preparation of figures, copy-editing and indexing. As always, I can never
thank enough my parents, Manohar Lai and Sumitra Chawla, for their
selflessness, constant encouragement and inspiration. I have always considered
the book-writing assignments that I have undertaken as educational in nature.
This one was no exception. In this regard, I wish to record my appreciation of
N. Hancox and M. Dunn for inviting me to undertake this work.
REFERENCES
In this text, I have followed the standard American usage. This will be
immediately evident in the spellings of certain words such as fiber, center,
behavior, etc. rather than fibre, centre, behaviour, respectively. In general,
compound words are not hyphenated. Prefixes such as co, pre, semi are closed
up with the word they modify. For example, pushout rather than push-out,
preexisting rather than pre-existing, etc. However, I have left the spellings
unchanged in the journal title in a reference or a book title in a citation. While
I have bowed to the American usage in the language of the text, I have
rigorously followed the Systeme International (SI) units. This stems from my
belief that sooner or later the American scientists, engineers, and industry will
be using SI units. The widespread use of SI units in the US has suffered a long
delay. The scientific merits of the SI units are overwhelming. If not the
scientific merits of the SI units, then the force of economic necessity will do the
job. By using SI units, I am hereby contributing my mite in that direction. I am,
however, fully cognizant that one should be able to convert from one system of
units to another. Hence, detailed information on this topic is given in Appendix
B.
Cfiapter 1
INTRODUCTION
~
~ Lammate
The subject matter of this book is ceramic matrix composites. The term
ceramics covers a wide variety of non-metallic, inorganic materials, which are
frequently processed at high temperatures. In view of the advances made in the
last quarter of the twentieth century, it is convenient to categorize the ceramic
materials into the following two classes:
Table 1.1 lists some important high performance ceramic materials. These
advanced ceramics show unusually high resistance to heat, chemicals, and wear.
These very characteristics, however, make them difficult to fabricate in a simple
and economical way. It is this class of high performance ceramics that are
commonly used in making ceramic matrix composites. Monolithic high
performance ceramics combine some very desirable characteristics, e.g., high
strength and hardness, excellent high temperature capability, chemical
inertness, wear resistance, and low density. They are, however, not very good
3 INTRODUCTION
Processing
Performance
of
Composite
under tensile and impact loading, and unlike metals, they do not show any
plasticity and are prone to catastrophic failure under mechanical or thermal
loading (thermal shock). We characterize this difference in the behavior of
metals and ceramics by saying that metals are forgiving while ceramics are not
forgiving. The forgiving nature of metals has its source in the high mobility of
dislocations in them, which allows them to deform plastically before fracture.
Fig. 1.3 Service temperature limit (indicative) of polymers, metals, and ceramics.
examine this concept of fracture toughness and see how Mother nature might
have some interesting lessons for us in this area. As Hillig (1985) points out,
strong winds (50 -80 km/h) can snap trees and poles like matchsticks. A birch
tree, however, can get away with the loss of only a few leaves. The reason for
this is that its trunk is strong and resilient. Under the action of a gale, the tops
of young birch tree can bend as much as 45 0 • This high degree of resilience or
compliance in a birch tree results from the appropriate ply architecture that the
nature provides. This resilience is what we call toughness in a more scientific
parlance. The low toughness of ceramic materials also manifests itself in
another form, viz., strength of ceramics has a very high degree of sensitivity to
the presence and size of flaws such as cracks, voids, inclusions, etc.
2°Jt---------------jkw
10 --------------- ~o
---------------- Cu
8 ________________ Fe
---------------- TiB2
---------------- Ti
4 ________________ AhOJ
________________ SiC
---------------- AI
2 ---------------- S~N4
--------------- C
- - - - - - - - - - - - - - - - Polymers
_ _ _ _ _ _ 0 '--_ _ _ _ _..... _ _ _ _ _ __
is the cost of ceramics. Although, we shall not discuss this point in this book,
there is no gainsaying the importance of the cost factor. The great challenge is
to produce consistent and reliable ceramic components having superior
properties but without an excessive cost, i.e., they should be competitive on a
cost/performance basis with the materials they seek to replace. In this regard,
the use of some new processing methods (e.g., sol-gel, injection molding, hot
isostatic pressing, etc.) is very promising in our efforts to have an economical,
high volume production of less brittle ceramic materials.
24 -------
-------------------------------------------
_____________________________ Sn Al
__________________________________________ ~n
20
16 ----------------------NiAl, Cu
::::::::::::Nf--------------~g()
12
------------------------------------- ~
------------------------------ Soda glass AW"I
8 -----------------------------------------------
---------- Ti '~3
------------.--------- TiEl2
----------------------------.Si3N4
_________________ Lithium alummosilicate glass-ceramic
____________________ C (parallel to the fiber axis)
o
-4
Fig. 1.5 Coefficient of thermal expansion of some materials.
importance. Yet another point that the reader should note is that while the ratio
of the modulus of the reinforcement and the polymer or metal matrix is
generally between 10 - 100, this ratio for a CMC is rather low, and can
frequently be equal to one or even less. The high modulus ratio in PMCs and
MMCs allows an efficient load transfer from the matrix to the fiber via a strong
interface. However, unlike in PMCs and MMCs, the low modulus ratio in a
CMC means that the reinforcement and the matrix are not much different in
their load bearing capacity, i.e., a simple increase in strength of a ceramic
material is rarely the objective. The only exception to this is the case when in
a CMC we have a low modulus matrix material, such as a glass matrix.
Table 1.2 Typical fracture toughness (K Ic) values for some materials.
Material
Polymers
Polyethylene 1-2
Nylon 3
Epoxy, Polyester 0.5
Metals
100-300
Pure metals
(copper, nickel, aluminum)
As I have said above, CMCs represent the latest entry in the field of
composites. They also represent a most formidable set of problems and very
9 INTRODUCTION
characteristic in which the tungsten carbide/cobalt tool inserts excel the ceramic
tool inserts is toughness. Bearings made of advanced ceramics have been used
in missiles. The primary material in this area is silicon nitride. Materials
characteristics such as low density, high temperature capability, corrosion
resistance, and the ability to operate with little lubrication are desirable in
bearing applications.
REFERENCES
Hillig, W.B. (1985) in Tailoring Multiphase and Composite Ceramics, Plenum Press,
New York, p. 697.
Katz, R.N. (1985) Mater. Sci. & Eng., 71, 227.
SUGGESTED READING
Ceramics have mostly ionic bonding and some covalent bonding. In ionic
bonding there occurs a transfer of electrons between atoms that make the
compound. Generally, positively charged ions balance the negatively charged
12 CERAMIC MATRIX MATERIALS
ions to give an electrically neutral17 compound, for example, NaCl, where Na+
ions balance the Cl- ions. In covalent bonding, the electrons are shared between
atoms. The characteristic high strength as well as brittleness of ceramic
materials can be traced to these types of bonding which make the PeierIs-
Nabarro potential very high, i.e., inherent resistance to dislocation motion in the
lattice is very high. Thus, unlike metals, a stress concentration at a crack tip in
a crystalline ceramic cannot be relieved by plastic deformation, at least not at
low and moderate temperatures. This has led to attempts at toughening ceramics
by means other than large scale dislocation motion.
Most the ceramic crystal structures are variants of face centered cubic (FCC)
or hexagonal close-packed (HCP). Generally, metallic cations are smaller than
the nonmetallic anions. Thus, in crystalline ceramics, the metallic cations
occupy interstitial positions in a lattice made of nonmetallic ions.
Coefficient
Young's a/Thermal Fracture
Density, Melting Modulus, Expansion, Toughness,
p, Point, E, a, K/c,
glcm 3 °C GPa 1O-6/K MPam'h
1) Simple cubic: This cubic structure is also called cesium chloride (CsCI)
structure. Examples are CsCI, CsBr, CsLIt is not as common as other structures
listed below.
2) Close-packed cubic: This is a variant of the Face Centered Cubic (FCC)
structure and is also called sodium chloride (NaCl) structure. Examples ofthis
structure include CaO, MgO, MnO, NiO, FeO, BaO, etc. Oxygen ions occupy
the face centered positions while the metal ions occupy the interstices.
3) Hexagonal close-packed: Examples of this structure include AIP3' Fe20 3,
Cr20 3 , etc.
Noncrystalline or amorphous (i.e., without form) ceramics are not really solids
but supercooled liquids. Now, liquids flow under their own mass. However,
they can become very viscous at low temperatures. Very viscous liquids (for
14 CERAMIC MATRIX MATERIALS
Simple cubic (CsCI) type Face-centered cubic (NaCI type) Hexagonal close-packed
Fig. 2.1 Common crystal structures in ceramics.
example, honey in the winter time) behave solidlike although they maintain a
disordered structure characteristic of a liquid, i.e., they do not undergo a
transformation to a crystalline structure. Thus, noncrystalline ceramics such as
glasses may, in many respects, behave like solids but structurally they are
liquids.
super cooled
liquid
'"
E
o"
>
-="
·u
Q)
c.
Vl
T,
Temperature ---~
Fig. 2.2 Specific volume vs. temperature for a glassy and a crystalline material.
frozen-in structure of a liquid. However, unlike true liquids but like solids, it
shows resistance to shear forces and shows a Hookean behavior, i.e., strain
produced is linearly proportional to the applied stress at low temperatures.
As pointed out above, glasses are, in terms of their microstructure, not solids
but very viscous, supercooled liquids. They can be regarded as inorganic
thermoplastics inasmuch as they soften when heated. One should point out that
organic thermoplastics can have a large fraction of crystalline phase. The major
advantage of any silica-based glass is its ease of fabrication, which allows
processes such as melt infiltration and compression molding to be used. Glasses
have low elastic modulus, a low strain-to-failure and toughness. Table 2.3 gives
some important varieties of glass and their salient characteristics.
16 CERAMIC MATRIX MATERIALS
0 ........ 0 ' -
• ........ Si..
(a)
(b)
Fig. 2.3 (a) A tetrahedron made up of four large oxygen ions with a silicon ion at the
center; (b) a network of silica tetrahedra.
2.4 GLASS-CERAMICS
as TiO z or ZrOz are introduced during the melting operation in the glass to give
a controlled crystallization that results in a glass-ceramic. Two types of
nucleating agents can be used:
(i) Agents having low solubility in glass (e.g., Pt, Cu, Au, Ag.)
(ii) Agents having high solubility in glass (e.g., Ti0 2, Zr0 2 , pps; 1-20 wt.%)
The important crystalline phases in this system are B - spodumene (Li 20' Al z0 3 '
4Si02), B-eucryptite (LiP ·A120 3 ·2SiOz) or solid solutions of one or both of
18 CERAMIC MATRIX MATERIALS
1
I
!
Time •
Fig. 2.4 Schematic of processing of a glass-ceramic.
these. The crystallite size is very fine: 30 - 40 nm. Thermal shock resistant
glass-ceramics having characteristically low coefficient of thermal expansion
are frequently made from this system. Lithium aluminosilicate (LAS) with near
zero coefficient of thermal expansion is used for optical and telescopic mirrors.
Corningware, a line of thermal shock resistant cookware, is based on ~
- spudomene (Li0 2 - AlP3 - nSi0 2 or LAS) glass-ceramic. Other trade names
for such glass-ceramic materials include Zerodur and Ceran.
These are called fluormicas because they contain fluorine and they have a
layered silicate, "mica", as the main crystalline phase. The easily cleavable
mica phase makes for easy machinability of this glass-ceramic.
Glass 55-70 70
Glass-Ceramics 70-350 80-140
AI2 0 3 200-50 280-380
Steel 300-1400 210
20 CERAMIC MATRIX MATERIALS
Melting and casting or drawing of a molten ceramic into a desired shape is not
very common. The conventional processing of glass, however, frequently
involves the melting of raw materials, followed by casting. In the case of glass
fibers, the molten glass is drawn through electrically heated platinum bushings
(more about this in Chapter 3). Some of the important conventional methods of
processing ceramics are described below.
Material
loading ............-......~
w
m ~ c::tJ
~. m
Stripping of
/ Hot-I '0 "all, .,,,,'n,
Consolidated
Part container
Fig. 2.5 Hot isostatic pressing (HIP) involves the simultaneous application of uniform
pressure and elevated temperature.
22 CERAMIC MATRIX MATERIALS
2.5.5 Reaction-Forming
A direct melting process can result in a fully dense ceramic, but the
temperatures involved are extremely high. Also, quite frequently, the melt
viscosity at reasonable processing temperatures is too high to allow proper
homogenization. At times, the oxides show large differences in volatility and
a direct melting process can result in large losses of some of the constituents.
23 CERAMIC MATRIX MATERIALS
Thus, making ceramics and glasses via a chemical route assumes great
importance in such cases. A chemical process of making a ceramic or glass can
lead to a high degree of homogeneity on a molecular scale and consequently a
high purity product can be obtained. Among such novel processing techniques
developed mostly in the last quarter of the twentieth century, one of the most
innovative approach to ceramic and glass processing is the sol-gel technique.
We give a brief description of this process. The reader is referred to the list of
suggested reading at the end of the chapter for more detailed information on the
subject of sol-gel per se.
The sol-gel route of making any glass or ceramic involves the formation ofthe
appropriate glass or ceramic structure by chemical polymerization of suitable
compounds in the liquid state (sol) at low temperatures, followed by chemical
reactions such as hydrolysis or condensation at temperatures much lower than
those used in powder processing or direct melting. A sol is a colloidal
suspension in which the individual particles are so small that no sedimentation
occurs. The particle size in a sol generally varies between 1 - 100 nm. Thus,
the sol has a free-flowing form, which is a characteristic of the liquid state. It
can also be obtained by mixing a metal containing precursor (e.g., a metal
alkoxide), a solvent (e.g., an alcohol), a catalyst (e.g., an acid), and water.
Hydrolysis and polycondensation reactions make the sol viscosity increase until
a gelled state is obtained. This gel is like a wet solid. A gel is a suspension in
which the liquid medium is viscous enough to behave more or less like a solid.
A "precursor" material in the form of a gel is obtained from the sol. The "wet"
gel consists of a network of particles, alcohol, and water. The wet gel is dried
to remove any unwanted residues (water, organic compounds, etc.). A large
amount of shrinkage accompanies this step. The porous gel is then converted
into glass or ceramic by heating at temperatures much lower than those used in
direct melting processes or by conventional powder processes. In principle, the
desired glass or ceramic can be obtained in an appropriate form: powder, film,
fiber, etc. The gel is the starting material for conversion into glass or ceramic.
Figure 2.6a shows the sol-gel process flow diagram while Fig. 2.6b shows the
microstructural changes that occur during the process.
The sol-gel process really is not new. In 1846, a French chemist, Jacques
Ebelman reportedly formed a transparent solid that could be drawn into
fibers.The chemistry of the process became clearer in the 1920's. Essentially,
tetrachlorosilane (SiCI4) when reacted with ethyl alcohol gives the alkoxide
tetraethylorthosilicate, also called tetraethoxysilane [Si(OC2H s)4]' This
alkoxide, commonly known by the acronym TEOS, serves as the starting
material for making an inorganic polymer containing Si and 0 via the sol-gel
24 CERAMIC MATRIX MATERIALS
(.)
'®
liquid
Residual
porosity
Fig. 2.6 (a) The sol-gel process flow diagram; (b) microstructural changes occurring
during the sol-gel process.
25 CERAMIC MATRIX MATERIALS
route, i.e., the silica glass or ceramic, or even glass-ceramic. Most sol-gel
processing involves the formation of metal oxides (ceramics or glasses) from
metal alkoxides. A metal alkoxide has the chemical formula ofM(OR)n' where
M is a metal or metalloid, R is an alkyl group such as CH3, C2H s, etc., and n is
the valence of the metal atom. Hydrolysis of metal alkoxides provides a low
temperature route to produce ceramics. Most metal alkoxides react with water
to yield hydrous metal oxides which give the metal oxide on heating. Two or
more metal alkoxides can be reacted to produce a mixed-metal oxide which
yields binary or ternary oxides on hydrolysis.
The metal alkoxide route has been exploited largely to obtain oxide ceramic
systems. Among nonoxide ceramics, the notable examples are polycarbosilane
and aminosilane routes for SiC and Si3N4 production (Hench and Ulrich, 1984).
(a) Destabilization of silica sol (pure or with metal ions added to aqueous
solutions of salts) to obtain a homogeneous gel. When a sol is destabilized, the
resultant product can be precipitates, unaggregated particles, or a homogeneous
gel. It is the last one that we desire.
(b) Hydrolysis and polycondensation of organometallic compounds generally
metal alkoxides) dissolved in alcohols and a small amount of water.
Aging
strengthening of the network occurs. At the neck joining the particles there is
a negative radius curvature. Thus, local solubility at the neck is less than near
the particle surface. Therefore, transport and deposition of silica occur
preferentially to the neck region and neck thickening results. This results in a
strengthening ofthe particulate network, Fig. 2.7 (Zarzcyki et aI., 1982).
•
• • tf4r
•
• "II
•
• • ~
• • •
•
(i) Mix the appropriate metal alkoxides, water-soluble salts if possible, and
ethanol solutions to yield the desired cation ratio.
(ii) Hydrolyze the above mixture with a water based acid solution
which can be diluted with alcohol. This polymerizes the mixture and a gel is
formed as per the following reaction
where M is the metal and R is an alkyl group such as (C2HS). At this stage one
can choose the conditions to get the bulk, powder, or fiber form of the precursor
material.
(iii) Heat the gel slowly to obtain the oxide (glass or ceramic) through the
reaction
M(OH)n --+ MO n12 + 1/2 n HP
During the process of gel drying, the solvents (water and any residual organic
material) will be eliminated. Concomitantly, sintering of the porous oxide takes
place. Very controlled rate of drying is required because of the large shrinkage
that accompanies the drying process. At an intermediate stage, a solid with a
very large amount of porosity on a micro scale is obtained. This is called
xerogel. The resultant glass or ceramic is virtually the same density, thermal
expansion, refractive index, and mechanical characteristics as the material of
the same composition made by direct melting.
28 CERAMIC MATRIX MATERIALS
Microwave energy provides another new avenue for processing ceramics and
CMCs. In conventional sintering, heat is provided by an external source. In the
microwave process, just like in the simple home microwave oven, the heat is
generated internally in the material. Volumetric heating occurs because of
internal frictional losses. Microwaves generate electromagnetic waves in the
frequency range of 0.3 to 300 GHz (wavelength range from I m to I mm), and
they are coherent and polarized.
Many insulating ceramic materials such as A1 20 3, MgO, Si02, and many glasses
absorb microwaves at above a certain critical temperature, Terit • Large sections
can be heated uniformly, because of lower thermal gradients.
SHS products are rather porous. There are two sources for this porosity: (i) the
original mix of reactant powders has about 30-50% porosity, and (ii) there is a
29 CERAMIC MATRIX MATERIALS
Product
SiC
Reactants
Si+C
large difference in the molar volume of the reactants and the products. Yet
another source of porosity may be due to any adsorbed gases at the extremely
high temperatures generated. A modification of the technique to obtain
synthesis together with densification involves the application of high pressure
during the combustion or immediately after the completion ofthe reaction when
the product temperature is still quite high. Hot pressing, rolling , and shock
waves are some of the techniques to apply the pressure.
Many ceramics such as borides, carbides, nitrides, silicides, and sialons as well
as composites such as SiCw + Al20 3 have been synthesized by means of SHS.
Synthesis can be followed by explosive or dynamic compaction of ceramics.
The reader should consult some of the references listed under Suggested
Reading for greater details.
The characteristics such as high hardness, high strength, high modulus, wear
resistance, etc. are achievable only in a crystalline ceramic in a fully dense
condition. Generally, a fully dense state is easier to achieve in oxide ceramics
than in non oxide ceramics. Typically, a nonoxide ceramic such as reaction
bonded silicon carbide (RBSC) is highly porous and thus very weak. Even in
hot pressing of these ceramics, one must use additives such as magnesia or
yttria to attain the theoretical density. These additives generally have
deleterious effects on the creep and oxidation resistance ofthe ceramic.
Silicon carbide is a very hard and abrasive material. It has excellent resistance
to erosion and to chemical attack in reducing environments. In oxidizing
environments, any free silicon present in a silicon carbide compact will be
oxidized readily. Silicon carbide itself can also be oxidized at very high
temperatures.
Silicon carbide does not occur in nature, although silicon and carbon are two
of the most abundant elements on earth. Carborundrum is a trade name for
silicon carbide. It is a combination of the Latin words carbon and corundum.
Silicon carbide can be obtained in two forms: a-SiC (hexagonal) and P-SiC
(cubic). The hexagonal a-silicon carbide has many polytypes which differ in the
stacking sequence. Typical physical and mechanical properties of silicon
carbide are given in Table 2.5.
Silicon carbide is very hard to sinter and will not deform under usual conditions
of hot pressing. Additives such as MgO, B, C, or Al to either a or ~-SiC must
31 CERAMIC MATRIX MATERIALS
(ii) Sintering
Sintered silicon carbide does not melt but decomposes into a silicon-rich vapor
and a carbon-rich residue. It is also very creep resistant at high temperatures.
Such unusual characteristics make conventional sintering of silicon carbide
difficult. Prochaska (1974) used sub micrometer P-SiC powder with a low
oxygen content was used, with 0.5% Band 1% C as additives and succeeded
in pressureless sintering SiC to full density between 1950-21 00 °C in vacuum
or inert atmosphere.
Sintered SiC has the advantage that it can be shaped in the "green" state, i.e.,
when it is soft. The following points are very important in pressureless sintering
of SiC:
1. Use of a sub micrometer powder is important because this gives a large
surface area, short diffusion distances, and large thermodynamic driving force
for densification.
2. Chemistry of the starting powders is also very important. Carbon removes the
SiOz layer from the SiC particles. Addition of boron is thought to accelerate
volume and grain boundary diffusion. It is important to avoid a nonuniform
distribution of the C additive and an exaggerated grain growth of a-SiC.
A mixture of SiC powder, graphite, and a plasticizer is made into a plastic body.
This is pressed, extruded, and injection molded to obtain a green body. The
plasticizer is burned off by pyrolysis.
Silicon metal (liquid or vapor) is then infiltrated into the body. This reacts with
the carbon powder to form SiC in situ. Any excess silicon (2-12%) goes to fill
any voids, giving a reasonably nonporous body. Reasonable strength to melting
point of silicon (1400 0c) or beyond has been obtained, depending on the
amount of free Si retained.
There are two forms of silicon nitride, a and ~, and both are hexagonal, with the
c-axis of a-Si3N4 being about twice that of ~-Si3N4 (Messier and Croft, 1982).
Contamination ofSi3N 4 with oxygen is a perennial problem. A brief description
of the various processing techniques is given below.
Sintered silicon nitride has strength levels between those of hot-pressed silicon
nitride and reaction-bonded silicon nitride.
Hot
Pressed 690 620 330 317 3.0 30-15
(MgO
additive)
Sintered
(Yz03 655 585 275 236 3.2 28-12
additive)
Reaction-
bonded 210 345 380 165 2.8 6-3
(2.45
glcm3)
This process involves heating a-Si3N4 and some sintering aid under pressure.
Typically, the pressure is about 30 MPa, the temperature range is 1600-1800
DC, and the time involved is a few hours. MgO is a common sintering aid. It is
thought that the MgO combines with the Si02 (a common contaminant on the
Si3N4 powder surface) to form magnesium silicate liquid phases. This liquid
penetrates between Si3N4 grains, providing a high diffusivity path, and becomes
a glassy phase on cooling from the pressing temperature. a-Si3N 4 dissolves in
the liquid silicate and elongated B-Si3N4 grains precipitate out from this. After
the a ... Btransformation is complete, the microstructure ofHPSN is comprised
of B-Si3N4 bonded via a thin layer of the glassy phase at the boundary. This
glassy phase is the cause for poor strength at temperatures above 1200 DC.
Other possible sintering aids are Ce02, Y203, and Zr02 • Typically, 98% of
theoretical density can be obtained.
As with any hot-pressing process, the process is expensive and complex shapes
are difficult to form. Among the parameters that control the properties of the
final Si3N4 product are:
MgO is by far the best known densification additive used for Si3N 4. Typically,
because of the magnesium silicate grain boundary phase, creep resistance starts
falling off between 1200 and 1350 DC.
Silicon powder is compacted into a desired shape, say, by slip casting, and then
heated in a nitrogenous atmosphere (pure N2 or N2 + H 2) in the temperature
range of 1100-1400 °C. The following chemical reactions may occur in the
process of nitridation of Si:
A fully dense product can result, but complex shapes are difficult to obtain.
Typical properties of silicon nitride made by different modes are given in Table
2.6 (Messier and Croft, 1982).
Sintered
Property RBSN HPSN HPSN Si3N 4
(MgO) (Y,OJ (y,o,)
2.7.3 Alumina
0 Oxygen
A • AI
':; Vacant
C1
B
C2
C3
B
C1
'L a
Basal Plane
2.7.4 Mullite
Boron and nitrogen can form the following BN compounds which are
isostructural polymorphs of carbon (Lipp et aI., 1989).
The hexagonal variety of boron nitride shows natural lubricity because of easy
cleavage. The easy cleaage follows from its lamellar structure. It is frequently
used as a mold release agent during hot pressing of ceramics such as silicon
nitride. Boron nitride is applied generally as a paint, paste, or aerosol. Because
of its similarity with graphite, hexagonal BN is sometimes referred to as white
graphite. It should be mentioned, however, that graphite is a good electrical
conductor while BN is an insulator. Unlike graphite, BN also shows an
excellent resistance to oxidation. Boron nitride in a powder form is also used
as an additive for silicon nitride and alumina.
Boron carbide is another low density, high melting point, high hardness ceramic
(Thevenot, 1990). Boron carbide powder is produced by reacting briquettes of
B 20 3 and C mixture in an electric arc furnace. The overall reaction is:
2.7.7 Sialons
(i) W-sialon plus glass: This has a high room temperature strength (3-point bend
strength", 1000 MPa) but poor creep properties above 1000 °C because of the
glassy phase.
(ii) W-sialon plus Y AG: This has a ceramic oxide rather than a glassy phase as
the grain boundary phase and, consequently, has excellent creep and oxidation
resistance.
Sialons find applications in cutting tools for metals, seals, bearings and wear
components, and in molten metal handling.
2.7.8 Carbon
Carbon is used as a matrix with carbon fibers. Carbon can resist very high
temperatures but only in the absence of oxygen or air. Commonly carbon matrix
is obtained by chemical vapor deposition (CVD) or using on organic precursor
(a resin or pitch) to infiltrate a fibrous preform, followed by pyrolysis of the
precursor to convert it into carbon.
2.7.9 Intermetallics
The demand for materials that are strong, stiff, and ductile at high temperatures
has led to resurgence of interest in intermetallics. Westbrook defines
intermetallics as all compound phases (ordered or disordered) of two or more
metals (Westbrook, 1960). The field of intermetallics is, indeed, very vast.
39 CERAMIC MATRIX MATERIALS
For service temperatures up to 1600 DC, besides high strength, creep resistance,
fracture toughness, one needs characteristics such as oxidation resistance and
microstructural stability. Meschter and Schwartz (1989) surveyed some silicide
intermetallics for high temperature applications. Based on criteria such as
3,200...----..,....---------------------,
~
• -Si Compounds ~
3,000 p" EEl -Be Compounds ~
<!J) - AI Compounds
..
~ - Re Compounds
r
2,800 ~
O-Others
E 2,600
~ ~
SiC
~
Z EEl
.
'"
(j;
E 2,400
Zr 5 Si. ---t\
- (\ jZrSi:
:
~
Cl zrJSi2~.J M0 5 Si J ( •
-E
CD
2,200
.. : ... ~ •• 0
o
:2
.... T~5. ~I.'..~ ·~.~~~~OSi·;······~·· .•.. ·.. 6··· 0.. ······· ...········ .. ·· .. ··· .. ·T m· .. • •• ~
........ .
2,000
EEl •• : • • : e...
~ ~ : (fJ..J0 0
o
1,800
: .: <0
1,600
EEl, ~~~OO<!J)i <!J)0 8 0
0
0 o
2 4 6 8 10 12 14 16 18 20 22
Density (g/cmJ)
Fig. 2.12 A plot of melting point vs. density for intermetaIIics having 0.8Tm = 1600 °C
(after Meschter and Schwartz, 1989).
41 CERAMIC MATRIX MATERIALS
A typical creep curve is shown in Fig. 2.13. The curve is divided into three
stages: primary (stage I), secondary or steady state (stage II), and tertiary (stage
III) in which an accelerated damage leads to fracture.
The creep rate, e, in the steady state can be expressed by the following general
relationship:
e = (ADGb/kT) (b/dg)m (a/Gt
where A is a dimensionless constant, D is the diffusivity, G is the shear
modulus, b is the Burgers vector, k is the Boltzmann's constant, T is the
temperature in kelvin, dg is is the grain size, a is the applied stress, and m and
42 CERAMIC MATRIX MATERIALS
1
3
v-; ,
...L
I
I
2
As we have said repeatedly, ceramics have high strength but low fracture
toughness. The low fracture toughness has its origin in the extreme sensitivity
of ceramics to the presence of flaws in them.
Various flaw types can occur in ceramics. We can categorize these into three
broad types:
Porosity is one of the more common defects in ceramics. There exists a variety
of porosities in ceramics. Closed porosity is a serious problem in ceramic
bodies. Apparent porosity measures the interconnected porosity in a body, and
determines the permeability of the ceramic by gases and fluids. Careful
processing can eliminate many flaws.
a =
f
~ 2E1
a
where fij(9) is a function of the angle, 9. Fracture occurs when the stress
intensity factor K attains a critical value, K1c • K]c> also known as the plane strain
fracture toughness, is a material parameter. An equivalent approach to the stress
intensity approach is based on the energy viewpoint. According to this
44 CERAMIC MATRIX MATERIALS
approach, fracture occurs when the crack extension force, G, reaches a critical
value, Gle • For ceramic materials, Gle=2,),. It can also be shown that
(plane stress)
and
EG
(plane strain)
(1 -V)2
where v is the Poisson ratio. These expressions show that the K and G
approaches are equivalent.
Although, no one ceramic is going to meet all the requirements, one can make
a list of some desirable characteristics of a ceramic matrix. That is, what one
would like to see in an ideal matrix material. In a real matrix, one can only hope
for a large number of these characteristics. An ideal ceramic matrix material
should:
In air or oxygen, oxide ceramics are inherently more stable than nonoxide
ceramics. Alumina (AI 20 3 ) is one of the most stable ceramics in a variety of
environments, although its melting point (2050 0c) is not one of the highest.
For most crystalline ceramics, at ambient temperatures, the finer the grain size,
the smaller will be the flaw size, and the higher the strength. At high
temperatures, the strength of ceramics is mainly determined by the
characteristics of the glassy phase that is frequently present at the grain
boundaries.
45 CERAMIC MATRIX MATERIALS
x
Crack
Fig. 2.14 A crack tip and a point (r, 9) near the crack tip.
REFERENCES
Sakka, S. (1982) in Treatise on Materials Sci. and Tech., vol. 22, (M. Tomozawa and
R.H. Doremus, eds), Academic Press, New York, pp. 129-167.
Somiya, S., R.F. Davis, and J.A. Pask (1990) (eds.), Mullite and Mullite Matrix
Composites, Amer. Ceram. Soc., Westerville, Ohio.
Thevenot, F. (1990) J. Euro. Cer. Soc., 6, 205.
Villers, P. and L.D. Calvert (1985) (eds.), Pearson's Handbook o/Crystallographic
Data/or Intermetallic Phases, vol. 1,2,3, ASM, Metals Park, OH.
Washburn, M.E. and W.S. Coblenz (1988) Amer Cer. Soc. Bull., 67, 356.
Westbrook, J .H. (1960) (ed.), Mechanical Properties o/Intermetallic Compounds, John
Wiley, New York, p. 1.
Zarzycki, J. (1984) in Glass: Science and Technology, vol. 2, Academic Press, New
York, pp. 209-249.
Zarzycki, J., M. Prassas, and J. Phalippou (1982) J. Mater. Sci., 17,3371.
SUGGESTED READING
CERAMIC REINFORCEMENTS
I
There are three types of ceramic fiber fabrication methods: chemical vapor
deposition, polymer pyrolysis, and sol-gel process. The last two routes involve
rather novel techniques of obtaining ceramics from organometallic polymers.
Undoubtedly the great breakthrough in the area of ceramic fibers has been the
Discontinuous Fibers
(a) Whiskers SiC, TiB2' Al20 3
(b) Short Fibers Glass, A120 3, SiC, (AIP3+SiOJ,
vapor grown carbon fibers
Continuous Fibers
Oxide AIP3' (AI20 3+Si0 2), Zr0 2,
silica-based glasses
Nonoxide B, C, SiC, Si3N4 , BN
48 CERAMIC REINFORCEMENTS
3.1 DEFINITIONS
One can transform practically any material (polymers, metals, or ceramics) into
a fibrous form (Chawla, 1998). Nature makes extensive use of materials in a
fibrous form. Historically and traditionally, however, fibers (synthetic as well
as natural) have been used mostly in the textile industry. Thus, some common
terms used with fibers have their origin in textile technology terminology. We
examine some of the important terms below.
Fibers are the fundamental unit in making textile yarns and fabrics. There are
many naturalfibers, organic and inorganic. Organic natural fibers include silk,
wool, cotton, jute, sisal while asbestos and basalt are example of natural
inorganic fibers. Synthetic fibers include polymeric, metallic, and ceramic
fibers. Polymeric fibers such as nylon, polyethyleneterephthalate (PET),
polyethylene (PE), polyacrylonitrile (PAN), polytetrafluoroethylene (PTFE),
aramid" etc. are well established fibers. Metallic filaments such as steel,
aluminum, copper, tungsten, molybdenum, gold, silver, etc have been available
for a long time. Among ceramic and glass fibers, glass fiber for reinforcement
polymeric materials has been available since the 1940s; ceramic fibers such as
carbon, silicon carbide, alumina, etc. became available in the 1960s and
onward.
Linear Density. This term gives us an idea ofthe yam size. Denier, commonly
used in the U.S, is the weight in grams of a yam 9000 m long. Tex, commonly
used in Europe, is the weight in grams of a yam 1000 long. Strictly speaking,
these linear density terms can be applied to single fibers as well. The basic idea
is that the larger the linear density (denier or tex), the heavier the yam. For
example, an individual fiber may be 1.5 den, while the fiber yam, may be
available in 200 den, 400 den, or all the way to 1500 den. Clearly, as the denier
(or tex) increases, the number of individual fibers in the yam increases.
Number of Fibers. This gives the number of individual filaments in the yam.
This parameter is really fixed by the number of holes in the spinneret.
50 CERAMIC REINFORCEMENTS
Twist in the Yarn. Two types of twists can be given to the yarn: a
counterclockwise twist or S twist, and a clockwise twist or Z twist. We can
make a ply yarn by using reverse twist directions which serves to balance out
any residual stresses. We can also twist two or more plies together to make a
cord.
Fibers have some unique features which stem mainly from their small cross-
section and large aspect ratio:
The long length also makes it imperative, in most cases, to incorporate them in
some medium, called matrix, to hold them together, i.e., to make a fiber
reinforced composite. It should be emphasized that this, by no means, is the sole
purpose of matrix in a composite.
Warp
Fill or Weft
independent of its form or size and is generally a material constant for a given
chemical composition and fully dense material. Thus, for a given composition
and density, the flexibility of a material is determined by its shape, or more
precisely by its diameter. The product of the bending moment (M) and the
radius of curvature (R) is called flexural rigidity. We can use the in verse of this
flexural rigidity as a measure of flexibility. From elementary strength of
materials, we have the following relationship for a beam bent to a radius R:
MIl =EIR
where E is Young's modulus of the material from which the beam is made and
the other terms are defined above. For a beam or a fiber of diameter, d, we have
I = d4/64. Substituting this in the above expression and rearranging the above
expression, we can write
MR = EI = E1td4/64
or,
IIMR = 641E1td4 (3.1)
where d is the equivalent diameter and I is the moment of inertia of the beam
(fiber). Equation (3.1) indicates that (lIMR), a measure of flexibility, is a very
sensitive function of diameter, d. Given a sufficiently small diameter, it is
possible to produce, in principle, a fiber as flexible as any from a polymer, a
metal, or a ceramic. In other words, one can make very flexible fibers out of a
ceramic such as silicon carbide or alumina provided one can make a fine
enough diameter. Producing a fine diameter ceramic fiber, however, is a
problem in ceramic processing.
52 CERAMIC REINFORCEMENTS
Carbon is a very versatile element. It is very light, and has a theoretical density
of2.27 g cm-l, and exists in a variety of forms. Two most important forms of
carbon are diamond and graphite. Among the relatively new forms of carbon
one should mention the Buckminster Fullerene, also known as Buckyball, and
the drawn out version of Buckyballs, called variously as nanotubes or
nanowires. It is the graphitic form of carbon that is important in understanding
the carbon fibers. The hexagonal structure of carbon in the graphitic form is
shown in Fig. 3.2.
Carbon fiber is a generic name representing a family of fibers. Over the years,
it has become one of the most important reinforcement fibers in all kinds of
composites. It is unfortunate fact that the terms carbon and graphite are used
interchangeably in commercial practice as well as in some scientific literature.
Rigorously speaking, graphite fibers are a form of carbon fiber that is obtained
after heating to a temperature greater than 2400 °C. This process, called
graphitization, results in a highly oriented, layered crystallographic structure,
which, in turn, leads to significantly different chemical and physical properties
from non-graphitic forms of carbon. An extreme case of the graphite structure
is, of course, a graphite single crystal. Such a single crystal has a hexagonal
symmetry and consequently anisotropic characteristics. In particular, its elastic
properties will be transversely isotropic in the layer plane. A hexagonal
symmetry requires 5 independent elastic constants. The Young's modulus E(9)
as a function of angle e (i.e., the angle between the c-axis and the stress axis)
for a hexagonal crystal is given, in terms of the compliances S, by the following
expression:
Table 3.2 gives the compliance (S) and stiffness (C) values ofa graphite single
crystal (Kelly, 1981). If we plug in these values in Eq. (3.3), we can obtain a
curve of the calculated tensile Young's modulus E as a function ofe (angular
displacement from the "a" axis) for a single crystal of graphite, Fig. 3.3. This
figure shows clearly the highly anisotropic nature of a graphite single crystal.
It should be clear to the reader that the more aligned the basal planes in a
carbon fiber, i.e., the more graphitic the structure, the higher the modulus in the
direction of the axis.
1200
a
w
0
a.
c.9
II)
::l
~
1000
800
1111
Basal Planes
Lc
::l
"'CJ 600
0
~
II)
~o> 400
C
::l
?- 200
0
0 20 40 60 80
Orientation, e degree
Fig. 3.3 Calculated tensile Young's modulus E as a function of e (angular displacement
from the "a" axis) for a single crystal of graphite.
54 CERAMIC REINFORCEMENTS
Table 3.2 Compliance (S) and stiffness (C) values for a graphite single crystal (after
Kelly, 1981)
S, GPa-1 C,GPa
SIl 0.00098
= C II = 1060
S33 0.0275
= C33 = 36.5
S44 0.25
= C44 =4
SI2 = -0.00016 C I2 = 180
SI3 = -0.00033 C 13 = 15
There are certain essential steps common to all processes of carbon fiber
manufacture:
The flexible polymeric chain structure of the PAN precursor fiber is shown in
Fig. 3.4a while the same polymer after stretching but without stabilization is
shown in Fig. 3.4b. On load removal, the structure shown in Fig. 3.4b collapses
to that shown in Fig. 3.4a. The more rigid, ladder structure obtained after
stabilization treatment is shown in Fig. 3.4c. A flow diagram showing the
55 CERAMIC REINFORCEMENTS
f
(a)
(b) (e)
Fig. 3.4 Structural changes in PAN (a) precursor fiber (b) stretched precursor without
stabilization (c) after stabilization.
various processing steps for PAN-based carbon fiber is shown in Fig. 3.5. The
PAN precursor has a flexible polymer chain structure, but it has an all carbon
backbone chain, Fig. 3.6a, containing polar nitrile groups. During the
stabilization treatment, the PAN precursor fiber is heated to 200 - 220°C under
tension. During this treatment, the nitrile groups react to for a ladder structure,
which is a rigid and thermally stable structure, Fig. 3.6b. Also, when PAN is
56 CERAMIC REINFORCEMENTS
Fig. 3.5 Flow diagram for the fabrication of PAN-based carbon fiber.
heated in air at 220 0 C, the absorbed oxygen cross links the chains, the fibers turn
black, and a stable ladder structure is formed. This treatment, done under
tension, helps in maintaining the orientation of the ring structure. During the
carbonization treatment between 1000 - 1500 °C, well developed hexagonal
networks of carbon form and a considerable amount of gases are evolved. This
gas evolution is partly responsible for some crack formation in the carbon fiber,
resulting in a lower tensile strength.
CN--CH2
/
CN--CH
'/" CH 2
CN--CH
'/" CH 2
'/" CH2
(a)
o 0
CH, C
I CH, C
I CH,
~/~/~/~/
CH CH CH CH
I
C
I
C
I
C
I
C
/ ~N/ ~N/ ~N/ ~
(b)
Fig. 3.6 (a) The carbon backbone chain structure of PAN (b) the ladder structure of
PAN after stabilization.
58 CERAMIC REINFORCEMENTS
Pitches form an important and low cost raw material for producing carbon fibers.
There are three common sources of pitch:
• petroleum asphalt
• coal tar
• polyvinyl chloride (PVC)
Pitches are thermoplastic in nature and are difficult to carbonize without being
first stabilized against fusion during pyrolysis. A schematic of the process of
making carbon fibers from a pitch is shown in Fig. 3.7. It involves the following
steps:
Spinnability of the pitch and its conversion into a nonfusible state are the most
important steps (Diefendorf and Tokarsky, 1975). These characteristics depend
on the chemical composition and molecular weight distribution of the pitch. The
molecular weight (MW) of the pitch controls the viscosity and the melting range,
i.e., MW controls the temperature and speed of spinning. The pitch composition
depends on its source. In fact, there is tremendous variability in the composition
of a pitch because it can be a mixture of hundreds of different species, varying
with the source of crude oil and the process conditions in the refinery. Indeed,
this is one of the main problems in the manufacture of pitch-based carbon fibers.
Generally, the pitches are made of different organic compounds containing
groups of condensed benzene-ring systems separated by and carrying alkyl
chains. Petroleum and coal-tar pitches may also contain large amounts of sulfur.
PVC pitch is also a mixture of several different organic compounds - polynuclear
aromatic compounds. These compounds have 3 or 4 aromatic nuclei, i.e.,
consisting of 3 or 4 rings per molecule. One suggested formula is C62 H52 which
is obtained by thermal degradation of PVC at 400°C under nitrogen for 30
minutes, the yield being 20%. The melting range is 150-200 °C and the
molecular weight is between 700-800.
Refine
Adjust Mol. Wt and
Chemical Composition
stress
graphitization
Table 3.3 presents property data for three different types of PAN-based carbon
fibers, while Table 3.4 gives the properties of mesophase pitch-based carbon
fibers. Notice the high density and high modulus of pitch-based fibers compared
to PAN-based fibers. The very high temperature treatment of graphitization (to
a temperature as high as 3000 0c) increases the degree of order in carbon fibers.
This is accompanied by a large increase in the elastic modulus ofthe fiber. The
tensile strength of PAN-based fibers when subjected to high temperature
treatment, however, falls, Fig. 3.8 (Watt 1970). This is attributed to the presence
of discrete flaws on the fiber surface and within it. Most of the volumetric
defects in carbon fibers originate from:
Table 3.3 Properties of PAN-based carbon fiber (strand data) (after Riggs, 1985)
• Thomel T-300, T-500, T-600, T-700; Celion 3000, 6000,1200; AS2, AS4, AS6,
IM6
b Thomel T-50, Celion G-50, HMS
c Celion GY-70
61 CERAMIC REINFORCEMENTS
Table 3.4 Properties of mesophase pitch-based carbon fiber (after Singer, 1981)
These defects get transformed during the high temperature treatment into diverse
imperfections. Basal-plane cracks called Mrozowski cracks are perhaps the most
important flaw type that limit the tensile strength of carbon fibers. These occur
as a result of anisotropic thermal contractions within the ribbon structure on
cooling from high temperature treatment (> 1500 DC). These cracks are generally
aligned along the fiber axis. Their presence lowers the tensile strength of the
fiber by providing easy crack nucleation sites. The fiber elastic modulus,
however, is unaffected because the elastic strains involved in the modulus
400
o 0
1200 1400 1600 1800 2000 2200 2400 2600
Heat trea1ment temperature, °c
Fig. 3.8 Elastic modulus and tensile strength of carbon fiber as a function of high
temperature treatment (after Watt, 1970).
62 CERAMIC REINFORCEMENTS
measurement are too small. Surface flaws can also limit the tensile strength of
the carbonized fibers. Oxidation treatments tend to remove the surface defects
and thus increase the strength levels ofthe fiber.
Many researchers have studied the structure of carbon fibers; see, for example,
Perret and Ruland, 1970; Bennett and Johnson, 1978, 1979; Inal et aI., 1980;
Reynolds and Sharp, 1974; Fourdeux,et aI., 1971; Riggs, 1985; Singer, 1981;
Bennett et aI., 1983. A scanning electron micrograph of pitch-based carbon
fibers is shown in Fig. 3 .9a while an atomic force microscope picture of the same
fibers is shown in Fig. 3.9b. Note the surface striations and the roughness on a
microscopic scale. When viewed in a transmission electron microscope, the
heterogeneous microstructure of carbon fibers becomes clear. In particular, the
rather pronounced irregularity in the packing of graphitic lamellae from the fiber
surface inward can be seen. The basal planes are much better aligned in the near
surface region of the fiber. In very general terms, the graphitic ribbons are
oriented more or less parallel to the fiber axis with random interlinking oflayers,
longitudinally and laterally. Figure 3.1 Oa shows a two-dimensional
representation of this lamellar structure. Deurberger and Oberlin (1991) studied
the structure of PAN-based carbon fibers obtained from several stabilized
precursors and carbonized in under various conditions and temperatures. An
example of the structure ofa carbon fiber from their work is shown in Fig. 3.10b.
This is a high resolution electron micrograph of a longitudinal section of a fiber
carbonized at 1330 °C. The micrograph shows 002 lattice fringes which
represent the graphitic planes of the carbon fiber. The lamellar structure of the
carbon fiber is clear. A schematic of the three-dimensional structure of PAN-
based carbon fiber is shown in Fig. 3.11 (Fourdeux et aI., 1971). Note the
distorted carbon layers and the rather irregular space filling. The degree of
alignment of the basal planes increases with the final heat treatment temperature.
Fig. 3.9 (a) SEM micrograph of pitch-based carbon fibers (b) An atomic force
micrograph (AFM) ofa pitch-based carbon fiber (courtesy ofK. Upadhya).
(a'
(b)
Fig. 3.1 0 (a) Two-dimensional representation of the lamellar structure of carbon (b) high
resolution electron micrograph of a carbon fiber showing lattice fringes (courtesy of A.
Oberlin).
65 CERAMIC REINFORCEMENTS
Fig. 3.11 Schematic of the three dimensional structure of PAN-based carbon fiber (after
Fourdeux et ai., 1971).
One should note that a low firing or sintering temperature will give a smaller
grain size but at the same time lead to an unacceptable level of residual porosity.
Should one go to higher processing temperatures, one can eliminate porosity but
excessive grain growth will result because of the high temperatures involved.
One can avoid this dilemma by introducing a second phase that restricts grain
boundary mobility while the porosity is removed at high temperatures. In
principle, it is possible to choose the type and amount of the second phase that
inhibits the grain growth at the service temperature. One way to use oxides of
silicon, phosphorus, boron, or zirconium as grain growth inhibitors but they
generally lower the working temperature. Among the examples are: a-alumina
plus 15-20% zr02 , 8-alumina plus about 4% of Si02 , and a-alumina plus 0.4%
Fe20 3 and 0.25% Si02 • The reader should note that for high creep resistance, one
requires a large grain size because grain boundary related phenomena such as
grain boundary sliding can lead to a rather large creep strain in a fine-grained
material.
Minnesota Mining and Manufacturing Co. (3M Co.) has developed a series of
alumina +silica fibers via sol-gel route. In particular, a pure a-alumina fiber has
been commercialized under the trade name Nextel 610. The sol-gel process of
making fibers involves the following steps common to all sol-gel processing:
(a)
(b)
Fig. 3.13 (a) Microstructure ofan (a-alumina + zirconia) or PRD-166 fiber as seen in a
scanning electron microscope. Zirconia particles (white) are located mostly at the grain
boundary triple points. (b) Microstructure of PRD-166 fiber as seen in a transmission
electron microscope.
69 CERAMIC REINFORCEMENTS
Specifically, in the case ofthe 3M process of making AIP3 fiber, the following
steps are involved:
Many other alumina or alumina-silica type fibers are available, most of which
are made by the sol-gel process. Sumitomo Chemical company produces a fiber
that is a mixture of alumina and silica. The flow diagram of this process is
shown in Fig. 3.15. Starting from an organoaluminum (polyaluminoxanes or a
mixture of polyaluminoxanes and one or more kinds of Si containing
compounds), a precursor fiber is obtained by dry spinning. This precursor fiber
is calcined to produce the final fiber. The fiber structure consists of fine
70 CERAMIC REINFORCEMENTS
Sol Reservoir
Filter
Pump
Spinneret
Pyrolysis Furnace
Fig. 3.14 3M process of making AlP3 fiber.
crystallites of spinel. Si02 serves to stabilize the spinel structure and prevents it
from transforming to a-AIP3 (Chawla, 1998). Mention has been made of the
Nextel series of fiber produced by the 3M Company. These are mainly Al 20 3+
Si02 and some B 20 3. The composition and properties of these fibers are given
in Table 3.5. The sol-gel manufacturing process used by 3M Co. has metal
alkoxides as the starting materials. Metal alkoxides are M(OR)n type compounds
where M is the metal and n is the metal valence, and R is an organic compound.
Selection of an appropriate organic group is very important. It should provide
sufficient stability and volatility to the alkoxide so M - OR bonds are broken and
MO- R is obtained to give the desired oxide ceramics.
Hydrolysis of metal alkoxides results in sols which are spun and gelled. The
gelled fiber is then densified at relatively low temperatures. The high surface
free energy available in the pores of the gelled fiber allows for densification at
a relatively low temperature. The sol-gel process provides a close control over
solution composition rheology of fiber diameter. The disadvantage is that rather
large dimensional changes must be accommodated and fiber integrity conserved.
71 CERAMIC REINFORCEMENTS
Tensile Young's
Fiber Type Composition, Diameter, Density, Strength, Modulus,
wt.% Jjm g cm- 3 MPa GPa
In the process used by 3M Co. for making the Nextel oxide fibers, aluminum
acetate [Al(OH)2 (OOCCH3) ' 1I3H3B03], e.g., Niaprooj, from Niacet Corp., is
the starting material (Sowman, 1988). Aluminum acetate having an A120/B 20 3
ratio of3 to 1 becomes spinnable after water removal from an aqueous solution.
In the fabrication of3M continuous fibers, a 37.5 % solution of basic aluminum
acetate in water is concentrated in a rotating flask partially immersed in a water
bath at 32 - 3 6 °C. After concentration to an equivalent A120 3 content of28.5 %,
a viscous solution with viscosity, 11, between 100-150 Pa s is obtained. This is
extruded through a spinneret having 130 holes of 100 flm diameter under a
pressure of 800-1 000 kPa and a shiny, colorless fibers are obtained on firing to
1000 °C. The microstructure shows cube-shaped and lath-shaped crystals. The
boria addition lowers the temperature required for mullite formation and retards
the transformation of alumina to a-A120 3. One needs boria in an amount
equivalent to or greater than a 9A120 3:2B 20 3 ratio in A1203-B203-Si02
compositions to prevent the formation of crystalline alumina.
A technique called Edge-defined Film-fed Growth (EFG) has been used to make
continuous, monocrystalline sapphire (AI 20 3) fiber(LaBelle and Mlavsky, 1967;
LaBelle, 1971; Pollack, 1972; Hurley and Pollack, 1972; Gasson and
Cockayne,1970). LaBelle and Mlavsky (1967) were the first ones to grow
sapphire (AIP3) single crystal fibers using a modified Czochralski puller and
radio frequency heating. In 1971, these authors devised a growth method, called
the edge- defined, film-fed growth (EFG) method. Figure 3.16 shows a
schematic of the method. Growth rates as high as 200 mm/min have been
attained. The die material must be stable at the melting point of alumina;
molybdenum die is used. A sapphire seed crystal is used. A capillary supplies a
73 CERAMIC REINFORCEMENTS
IOrganoalumino compound I
•
Alkyl aluminum
or
AIkoxy aluminum
(AlR3)
t1-ot
Polymerization
AIR, + H20 -
•
Precursor fiber
(organoaluminum polymer
and alkyl silicate)
I Calcination I
•
Fig. 3.15 Flow diagram of the Sumitomo process for making a mixture of alumina and
silica fiber.
constant liquid level at the crystal interface. Molten alumina wets both,
molybdenum and alumina. The crystal grows from a molten film between the
growing crystal and the die. The crystal shape is defined by the external shape
of the die rather than the internal shape. Perhaps, the most important item about
the Saphikon fiber is its single crystal nature, i.e., there are no grain boundaries.
Thus, under creep conditions involving grain boundary related phenomena such
cavitation and grain boundary sliding will not be operative and one can expect
74 CERAMIC REINFORCEMENTS
a creep resistant fiber. However, single crystal alumina can undergo dislocation
creep on basal plane. It would thus appear that if one can orient the basal plane
such that basal slip does not occur then, one can obtain a creep resistant fiber_
Such will be the case for a single crystal fiber with its c-axis parallel to the fiber
aXIs.
Temperalllte
M ell ing
Point
Fig. 3.16 Schematic of edge-defined, film-fed, growth process of making a single crystal
alumina fiber.
combined direct current (250 rnA) and very high frequency (VHF ~ 60 MHz) to
obtain an optimum temperature profile. To obtain a 100 flm to SiC
monofilament, it generally takes about 20 s in the reactor. The filament is wound
on a spool at the bottom ofthe reactor. The exhaust gases (95% ofthe original
mixture + HCI) are passed around a condenser to recover the unused silanes. An
efficient reclamation of the unused silanes is very important for a cost effective
production process. Such CVD processes result in composite monofilaments
which have built in residual stresses. The process is, of course, very expensive.
Methyltrichlorosilane is an ideal raw material for this process as it contains one
silicon and one carbon atom, i.e., a stoichiometric SiC will be deposited. The
chemical reaction is:
A series of surface modified silicon carbide fibers, called SCS fibers, has been
used in a variety of ceramic matrix composites. These special fibers have a
complex through the thickness gradient structure. SCS-6, for example, is a thick
fiber (diameter = 142 flm) and is produced by chemical vapor deposition of
silicon and carbon containing compounds onto a pyrolytic graphite coated
carbon core. The pyrolytic graphite coating is applied to a carbon monofilament
to give a substrate of3 7 flm. This is then coated with SiC by CVD to give a final
monofilament of 142 Ilm diameter. The surface modification of the SCS fibers
consists of the following. The bulk of the 1 flm thick surface coating consists of
C doped Si. Figure 3.18 shows schematically the cross-sections of the two SCS
type silicon carbide fibers and their characteristic surface compositional
gradients. In SCS-6 fiber, zone I at and near the surface is a carbon rich zone. In
zone II, Si content decreases, and in zone ill, the Si content increases back to the
stoichiometric SiC composition. Thus, the SCS-6 silicon carbide fiber has a
surface graded outward to be carbon rich. Mann et al. (1999) verified the
dramatic changes in the mechanical characteristics such as Young's modulus and
hardness of SCS-6 fiber as function of fiber radius using nanoindentation
techniques. Besides SCS-6 there are two other varieties of CVD silicon carbide
fibers on a carbon core that are available commercially. Typical properties of
77 CERAMIC REINFORCEMENTS
WFilament
@ Hydrogen
supply
r
IScrubber I
--
l
l LI Silane
~I vaporizer
r l
l
IRefrigeration ~---,
r
I Silane l~_--,
supply I
IDistillation I
Photo-
Gases/ • optical
for - - diameter
recuperation~ sensor
Waste
by-products ~iC/w
Fig. 3.17 A schematic of the process of making silicon carbide monofilament by CVD.
(The trade name is sigma fiber and the substrate is tungsten.)
SCS-type silicon carbide fibers are given in Table 3.7. Another SiC fiber made
by CVD is called the sigma fiber; this has core of tungsten filament. Such fibers
made by CVD on a heated substrate are a kind of composite fibers, and can show
unusual behavior. Lara-Curzio and Sternstein(1993) examined the behavior of
such composite fibers when subjected to thermomechanicalloading. Their main
conclusion was that one must take into account the strains experienced by the
substrate during the CVD process of making such composite fibers. Not doing
78 CERAMIC REINFORCEMENTS
As pointed out above, the SiC fibers obtained via CVD are very thick and not
very flexible. Work on alternate routes of obtaining fine, continuous, and
flexible fiber had been in progress for sometime when in the mid-nineteen
seventies the late Prof. Yajima and his colleagues (1976, 1980) in Japan
developed a process of making such a fiber by controlled pyrolysis of polymeric
precursor. This method of using silicon based polymers to produce a family of
ceramic fibers having good mechanical properties, good thermal stability, and
oxidation resistance has an enormous potential. Figure 3 .19 shows a general flow
79 CERAMIC REINFORCEMENTS
Pyrolytic graphite
coated carbon core ~
- --
""-
Inner zone: ""-~
D
--- carbon-rich ~ - SiC
Outer zone:
Stoichiometric ~ - SiC
~
~_
/
/
/
Carbon-rich surface /
coating (0-4!Am) .-/
SCS-6 fiber
u 1.0
::;:..
C".f.l
~-SiC
Distance from
the surface, !-tm
diagram for making ceramic fibers from a polymeric precursor. Note the
similarity with the process of making a carbon fiber from a polymeric precursor.
The various steps involved in this polymer route can be listed as follows (Wax,
1985):
(a) polymer characterization (yield, molecular weight, purity, etc.)
(b) melt spin polymer into a precursor fiber
(c) cure the precursor fiber to crosslink the molecular chains, making it infusible
during the subsequent pyrolysis.
(d) pyrolyze the precursor fiber under controlled conditions to obtain the ceramic
fiber.
80 CERAMIC REINFORCEMENTS
Polymeric Precursor
Melt or so lution
spinni ng
"
Precursor Fiber I
Curin 9
"
Cured or Stabilized
Fiber
Controlled Pyrolysis
"
Ceramic or Glass
Fiber
Fig. 3.19 Flow diagram for making ceramic fibers from a polymeric precursor.
Specifically, the Yaj ima process of making SiC fiber from a polymeric precursor
fiber involves the following steps and is shown schematically in Fig. 3.20.
Polycarbosilane, a high molecular weight polymer, containing Si and C is
synthesized. This involves the use of a commercially available material, viz.,
dimethylchlorosilane. Solid polydimethylsilane is obtained by dechlorination of
dimethylchlorosilane by reacting it with sodium. Polycarbosilane is obtained by
thermal decomposition and polymerization of polydimethyl silane. This is carried
out under high pressure in an autoclave at 470°C in an argon atmosphere for
8-14 h. A vacuum distillation treatment at up to 280°C follows. The average
molecular weight ofthe resulting polymer is about 1500. This is melt spun from
a 500 hole nozzle at about 350°C under N2 gas to obtain the so called pre-
ceramic continuous, precursor fiber. The precursor fiber is quite weak (tensile
strength ~ 10 MPa). This is converted to inorganic SiC by curing in air, heating
to about 1000 °C in N2 gas, followed by heating to 1300 °C in N2 under stretch.
During the pyrolysis, first stage of conversion occurs at around 550°C when
cross-linking of polymer chains occurs. Above this temperature, the sidechains
containing hydrogen and methyl groups decompose. Fiber density and
mechanical properties improve sharply. The conversion to SiC is occurs above
about 850°C.
81 CERAMIC REINFORCEMENTS
CH, / Q
Si
CH3/ '0
Dichlorodimethylsi1ane
I Dechlorination
I
t-B
with Na (to NaCl)
. CH3 n
Polydimethy)silane
Polymerization at
I 470 OC in autoc:Iaft
CH3 H
t~-?1CH3 H n
Polyalrbosilane
Melt spinning at
I 3SOOC(N2l
Polycarbosi1ane
fiber
Curing 190 OC in air
r or Rf in ozone
Polycarbosi1ane fibers
with molecular c:ross-
linking by oxygen to
awid subsequent
melting
Pyrolysis heating to
I 1300 OC or less on in
vacuum (1000 0CIh)
SiC fiber
Amorphous or micro-
CI')'Stalline rt-SiC
Fig. 3.20 The Yajirna process of making fme diameter SiC from a polycarbosilane.
There are many silicon carbide type fibers, with elemental compositions ofSi-C,
Si-N-C-O, Si-B-N, Si-C-O, and Si-Ti-C-O, that are commercially available.
Almost all of them are made from polymeric precursors. A multifilament silicon
carbide fiber, called Tyranno (Yamamura et aI., 1988) is made by pyrolysis of
poly (titano carbosilanes) and it contains between 1.5 - 4 wt. % titanium. A textile
grade silicon carbide fiber made, called Syiramic, has, according to the
manufacturer, a nanocrystalline structure (crystallite size 0.5 Ilm), a density of
3.0 g cm- 3, a tensile strength of 3.15 GPa and an elastic modulus of 405, GPa.
Fig. 3.21 High resolution transmission electron micrograph of Nicalon type SiC
produced in a laboratory, indicating the amorphous nature of the SiC produced by the
Yajima process (courtesy ofK. Okamura).
Many of the "SiC" type fibers contain free carbon. One can analyze this by the
shift (S") in the 1580 cm- 1 Raman band of carbon as a function of applied strain.
This band is characteristic of aromatic!graphitic forms of carbon. A compressive
strain results in a positive S" while a tensile strain results in a negative S". Figure
3.23 shows the shift in Raman wavenumber per percent of applied strain as a
function of inverse Young's modulus square root for a series ofNicalon fibers,
Tyranno fiber, PAN and pitch-based carbon fibers and two organic fibers,
aramid and PBZT (Gouadec and Colomban, 2001). Hi and Hi-S grades of
Nicalon do not show any shift because carbon does not form a continuous
amorphous network but is in the form of isolated nanoprecipitates.
IS o Aramid
mpBZf
"'Pilch
~PAN
10 OTyranno
• NLM-NlCalon
+ Hi-Nicalon
A Hi-S-NlCalon
,-.
S
~
;:;-
·s
~ 0
~ o l70
+
1000 x E-11l
-5
(GPa11l)
-10
•
-15
Fig. 3.23 Shift in Raman wavenumber per percent of applied strain as a function of
inverse Young's modulus square root for a series ofNicalon fibers, Tyranno fiber, PAN
and pitch-based carbon fibers and two organic fibers, aramid and PBZT (Gouadec and
Colomban, 2001).
86 CERAMIC REINFORCEMENTS
~ 100
--- SiC
-o-Carbon
J
.S 80
1 60
~
•e 40
0
20
Z 0
0 10 20 30 40 50 60 70
0
400 ' ,
40
~ 300 30 ~
=- =-
~ 200 20 ~
~
10 =
0 0
0 10 20 30 40 50 60 70
Fig. 3.24 Correlation between the Raman spectra along the radius of an SCS- 6 fiber
(Gouadecand Columban, 2001) and Young's modulus and hardness (Mann eta!., 1999).
3.6 WHISKERS
Whiskers are monocrystalline, short fibers with extremely high strength. This
high strength, approaching the theoretical strength, comes about because of the
absence of crystalline imperfections such as dislocations. Being mono crystalline,
there are no grain boundaries either. Typically, whiskers have a diameter of a
few /lm and a length ofa few mm. Thus, their aspect ratio (length/diameter) can
vary between 50 and 10000. Whiskers, however, do not have uniform
dimensions or properties. This is perhaps their greatest disadvantage, i.e., the
spread in properties is extremely large. Handling and alignment of whiskers in
a matrix to produce a composite are other problems.
87 CERAMIC REINFORCEMENTS
Whiskers are normally obtained by vapor phase growth. Early in the nineteen
seventies, a new process was developed, starting from rice hulls, to produce SiC
particles and whiskers (Milewski et aI., 1974; Lee and Cutler, 1975). The SiC
particles produced by this process are very fine in size. Rice hulls are a waste by-
product of rice milling. For each 100 kg of rice milled, about 20 kg of rice hull
is produced. Rice hulls contain cellulose, silica, and other organic and inorganic
materials. Silica from soil is dissolved and transported in the plant as monosilicic
acid. This is deposited in the cellulosic structure by liquid evaporation. It turns
out that most of silica ends up in hull. It is the intimate mixture of silica within
the cellulose that gives the near ideal amounts of silica and carbon for silica
carbide production. Raw rice hulls are heated in the absence of oxygen at about
700°C to drive out the volatile compounds. This is called coking. Coked rice
hulls, containing about equal amounts of Si02 and free C, are heated in inert or
reducing atmosphere (flowing N2 or NH3 gas) at a temperature between 1500 and
1600 °C for about 1 h to form silicon carbide as per the following reaction
3C + Si02 -+ SiC + 2CO
Figure 3.25 shows a schematic ofthe process. When the above reaction is over,
the residue is heated to 800°C to remove any free C. Generally, both particles
and whiskers are produced together with some excess free carbon. A wet process
is used to separate the particles and the whiskers. Typically, the average aspect
ratio of the as-produced whiskers is 75.
The vapor-liquid-solid (VLS) process of growing whiskers has been known for
sometime (Lindemanis, 1983; Milewski et aI., 1985; Petrovic et aI., 1985). The
acronym VLS stands for vapor feed gases, liquid catalyst, and solid crystalline
whiskers. The catalyst forms a liquid solution interface with the growing
crystalline phase while elements are fed from the vapor phase through the liquid-
vapor interface. Whisker growth takes place by precipitation from the
supersaturated liquid at the solid liquid interface. The catalyst must take in
solution the atomic species of the whisker to be grown. For SiC whiskers,
transition metals and iron alloys meet this requirement. Silicon and carbon are
supplied in the form ofSiO and CH4 gases, respectively. The SiO gas is obtained
by carbothermal reduction ofSi02. Figure 3.26 shows the process chemistry of
the VLS whisker process. Generally a range of whisker morphologies is
obtained. The tensile strength values have been reported to range from 1.7 to
23.7 GPa in 40 tests (Milewski et aI., 1985; Petrovic et aI., 1985). Whiskers
lengths were about 10 mm and the equivalent circular diameter averaged 5.9 /lm.
Their average tensile strength and modulus were 8.4 GPa and 581 GPa,
respectively. The VLS process is extremely slow.
88 CERAMIC REINFORCEMENTS
Rice hull
Carbon tube
reactor
Whisker/carbon separation
Carbon oxidation
I SiC whiskers I
Fig. 3.25 Process of making silicon carbide whiskers from rice hulls.
SiC in particulate form has been available quite cheaply and abundantly for
abrasive, refractories, and chemical uses. In this conventional process silica in
89 CERAMIC REINFORCEMENTS
CO(g)
SiO(g) + ~ = Si + CO(g)
Generator
Fig. 3.26 Chemistry of the Vapor-Liquid-Solid (VLS) process of making SiC whiskers.
the form of sand and carbon in the form of coke are made to react at 2400 °C in
an electric furnace. The SiC produced in the form of large granules is
subsequently comminuted to the desired size.
Pysher et al. (1989) measured high temperature strengths of various oxide and
nonoxide ceramic fibers. High temperature strength of the SiC type fibers was
limited by oxidation and internal void formation while in the case of oxide fibers
intergranular glassy phase led to softening. Thermal stability of ceramic fibers
derived from polymeric precursors is of special interest, mainly because they
frequently have some undesirable phases present in them. Polycarbosilane-
derived SiC type fibers, such as Nicalon or Tyranno, involve a thermal oxidation
curing process as described above and can contain as much as 10 mass% oxygen.
Such fibers decompose at temperatures above 1200 °C in a nitrogen or argon
atmosphere with SiO and CO gas evolution:
Polycrystalline Oxide
Carbon Single Crystal Oxide
Single Crystal
Oxide
b '~-+-_Polycrys!alline
Oxide
T-
Strength drops at Intermediate
Poor OxidatIOn Resistance Temperatures for Polycrystanine
C+o.- Co. Oxides
T-
Resonable Oxidation
Resistance
Fig. 3.27 Schematic of strength vs. temperature for some ceramic fibers.
91 CERAMIC REINFORCEMENTS
Mah et al. (1984) studied the degradation ofNicalon fiber (SiCxO y ) after heat
treatment in different environments. The strength of this fiber at temperatures
greater than 1200 °C degraded because of CO evaporation from the fiber as well
as P-SiC grain growth. Another unusual feature observed with ceramic fibers
made via pyrolysis of polymeric precursors, especially with compositions Si-C-O
and Si-N-C-O, is that they have lower densities than the theoretical values. The
theoretical density, Pt, value can be calculated by using the relationship
Pt=PYi
where p is the density, V is the volume fraction, the subscript i indicates the ith
phase, and summation over all the phases present is implied. Lipowitz et al.
(1990) used x-ray scattering techniques to show that porosity present is such
fibers was due to globular pores of nm size and that the pore fractions ranged
from 5% to 25%. According to these authors, nanochannels form during the
early states of pyrolysis when rather large volumes of gases are given out. In the
later stages of pyrolysis, of gases are given out. In the later stages of pyrolysis,
during densification these nanochannels suffer a viscous collapse and nanopores
are formed. The volume fraction of nanopores decreases with increasing
pyrolysis temperature. The reader should note that a higher density and a lower
void fraction will lead to a higher elastic modulus of these ceramic fibers.
I I
3.0r- Oxygen
content
(mass%)
• <0.5
0=4.0
.. " 1.9
t:.. = 10.B
100
O~ __~__~I__~__~IJ 1 I 1
1400 1600 1800 1400 1600 1800
Temperature, K Temperature, K
Fig. 3.28 Tensile strength and Young's modulus as a function of temperature for
polycarbosilane derived SiC fiber with different oxygen amounts (after Okamura et aI.,
1992).
fiber with the basal plane is perpendicular to the fiber axis. It is easy to see that
in this orientation the basal slip during creep will not occur and the fiber will
show good creep resistance. It turns out, however, that ifthe fiber orientation is
off by more than 3 degrees, easy basal slip comes into play and some of the
creep resistance will be lost. Single crystal YAG (yttrium aluminum garnet) is
another oxide fiber that shows excellent creep resistance. YAG has a complex
cubic structure and is less anisotropic than the hexagonal sapphire. Single crystal
YAG fiber can be grown by either laser floating process or edge-defined film-fed
growth (EFG) method. The topic of creep of CMCs is further discussed in
chapter 7.
Two most important characteristics of any reinforcement material are its strength
and Young's modulus. First of all, we note that all these high performance fibers
have very low density values. Given the general low density of these fibers, the
best of these fibers group together in the top right hand corner of Fig. 3.29. The
reader will also recognize that the elements comprising these fibers pertain to the
93 CERAMIC REINFORCEMENTS
0
a..
(!) 4
-
I
..c:
0'1 3
c
-
(J)
"-
en
2
(J) Glass I:~::::::::: ::::::::::~:~I
(/) Mullite
c
~ g~-Pitch.
IsotropIc
oL-~~--~~--~----~--------~
o 100 200 300 400 500
Young's Modu Ius, GPo
Fig. 3.29 Strength vs. Young's modulus for some ceramic fibers.
first two rows ofthe periodic table. Also to be noted is the fact that, irrespective
of whether in compound or elemental form, they are mostly covalently bonded
which is the strongest bond. Generally, such light strong, and stiff materials are
very desirable in most applications, but particularly so in aerospace field, land
transportation, energy related industry, housing and civil construction, etc. Fiber
flexibility is associated with the Young's modulus and the diameter (see Sec.
3.1). In the general area of high modulus (i.e., high E) fibers, the diameter
becomes the dominant parameter controlling the flexibility. For a given E, the
smaller the diameter the more flexible it is. Fiber flexibility is a very desirable
characteristic if one wants to bend, wind, and weave a fiber in order to make a
complex shaped final product.
Weibull distribution gives the failure probability, F(O"), ofa fiber at an applied
stress 0" by the following expression:
where a and pare statistical parameters. F( 0") is called the cumulative frequency
distribution function. The parameter Pis referred to as the Weibull modulus. It
is a measure of the variability in the strength of the fiber, the higher the value of
Weibull modulus, the higher the uniformity of strength values. The probability
density function for a brittle fiber oflength L that obeys Weibull distribution can
be obtained by differentiating the above expression and takes the following form
.1(0") = Lap O"~ -) exp ( - L a O"~) (3.3)
(3.4)
For a fiber showing the strength distribution given by Eq. (3.3), we can obtain
an expression for the mean strength as follows:
= Jexp( - L a (J~) .
~ (L a (J~)13
I · d(L a (J~) (3.6)
o (L 0,)13
~-I
cr = j exp( - x)· (x) ~ I dx (3.7)
o (La)13
Now, letting ~+ 11~ = n, we can convert Eq. (3.7) to the following simple
form:
Jexp( -x)· x
1 ~
cr = (L a) -13 n - 1 dx
(3.8)
o
The integral portion of Eq. (3.7) is called the gamma function of order nand
denoted by [(n). Thus, the mean strength is given by
1
cr = (L a) -13 . [(n) (3.9)
[(0) = 00
[(112) =.[7t
[(1) = 1
[(2) = 1
[(3) = 2!
97 CERAMIC REINFORCEMENTS
-- 1
a = Ml 1 r [ 1 + f3 ]
= (L a) P. (3.10)
1
S = [M2 -Mn2 (3.11)
S2 = 1-12 = M2 - (J2
The expression for standard deviation, s, for the Weibull distribution under
consideration can be evaluated as follows.
(3.12)
Let us evaluate the first integral in this expression. The second integral is given
by Eq. (3.1). Substituting for f( 0'), we get for the mean strength
(L a (13)2/13
= Jexp( - LaoP),
00
. d(Lao P)
o (La)21 p
dx
n
The expression for s, Eq. (3.12), then becomes
s
fl
For fibers characterized by Weibull distribution and having P> 1, we have the
following general relationships:
For glass fibers, fl '" 0.1, P= 11
For alumina, carbon, boron, silicon carbide type fibers, fl = 0.2-0.4, P=2.7- 5.8.
cr = (aL r I1~ r ( 1 + *)
For a unit length of fiber (L = 1)
where
For p> 1, we have 0.88 ~ k ~ 1.0. This means that we can regard the quantity
a- 1/6 as the reference level strength.
We can find the statistical mode a:, the most probable strength value, as
follows:
j(cr) = Lapcr(~-l) exp(-Lacr~)
For large~,
~ -1
- - ==>
~
and
-.!.
0* = (aL) P
where Pis the Weibull modulus and a is a scale parameter. Pis a measure of the
scatter in the strength data. Rearranging Eq. 3.12, we get
From Eq. 3.14, we can easily obtain a and Pgraphically. One generally arranges
the tensile strength values of the single filaments in an ascending order and
assigns a probability of failure using an estimator given by
where F(a j ) is the probability of failure corresponding to the ilhe strength values
and N is the total number of fibers tested. Substituting Eq. 3.15 in Eq. 3.14, we
have
Eq. 3.16 says then that a plot of In [(N + 1)/(N + 1 -i)] vs. a j on a log-log graph
will be a straight line ifthe tensile strength data follows a Weibull distribution.
The intercept on the y-axis will then be a and the slope will be p. Figure 3.30
shows such a double log plot for some Nextel fibers of (gage length = 25 mm)
(Wilson and Visser, 2001). An example of cumulative probability of failure of
Nexte1312 fiber as function of fiber strength is shown in Fig. 3.31 (Kerr et ai.,
2002). The Weibull mean tensile strength, cr, standard deviation, s, and
coefficient of variation (CV) can be obtained as follows:
102 CERAMIC REINFORCEMENTS
(3.17)
(3.18)
where
CV = 100 ( ~l
2,---------------------------~----------------~
Nextel610: Me . . = 3.40 GPa,m = 10.1
Nextel6SO: Mean = 2.60 GPa,rn= 6.8
Nextel720: Mean = 1.98 CPa, m = 7.6
-1
~
.-<
-2
.5
.5
-3
~
Nutel650
"
Nu:td 610
-4
" :>C
-5
" >:
-6
0.2 0.4 0.6 0.8 1.2 1.4
Fig. 3.30 Weibull distribution of some Nextel fibers (after Wilson and Visser, 2001).
103 CERAMIC REINFORCEMENTS
-e
...... •
0.8
••
u..
~
.a ••
•
••
'iii
....0
II. 0.6
•
~ •••
:aC\'I
..0
0.4 •
~
a.
0.2
0
500 1000 1500 2000 2500 3000
T ensile Strength (MPa)
Fig. 3.31 Cumulative probability of failure of Nextel 312 fiber (after Kerr et aI., 2002).
REFERENCES
Bennett, S.C. and D. J. Johnson (1978) in 5th Int. Carbon and Graphite Con!, Soc.
Chern. Ind., London, p. 377.Bennett, S.C. and D. J. Johnson (1979) Carbon, 17,25.
Bennett, S.C., D.J. Johnson, and W. Johnson (1983) J Mater. Sci., 18,3337.
Birchall, J.D., J.A.A. Bradbury, and J. Dinwoodie (1985) in Strong Fibres, Handbook
o/Composites, vol. 1, North-Holland, Amsterdam, p. 115.
Chawla, KK (1998) Fibrous Materials, Cambridge University Press, Cambridge.
DeBolt, H.E., V. 1. Krukonis, and F. E. Wawner (1974) in Silicon Carbide 1973, Univ.
of S. Carolina Press, Columbia, SC, p. 168.
Deurbergue, A. and A. Oberlin, (1991) Carbon, 29,691.
Dhingra, A.K (1980) Phil. Trans. Roy. Soc., London, A294, 41l.
DiCarlo, lA. (June 1985) J 0/ Metals, 37, 44.
Diefendorf, R.J. and E. Tokarsky (1975) Polymer Eng. & Sci., 15, 150.
Dresher, W.H. (April 1969) Journal 0/ Metals, 21, 17.
Ezekiel, H.N. and R. G. Spain (1967) J Polymer Sci. C., 19, 27l.
Fourdeux, A., R. Perret, and W. Ruland (1971) in Carbon Fibres: Their Composites and
Applications, The Plastics Inst., London, p .57.
Gasson, D.G. and B. Cockayne (1970) J Mater. Sci., 5,100.
Gouadec, G. and P. Colomban (2001) 1 Euro. Ceram. Soc., 21, 1249.
104 CERAMIC REINFORCEMENTS
SUGGESTED READING
Bunsell, A.R. (ed.) (1988) Fibre Reinforcements for Composite Materials, Elsevier
Amsterdam.
Chawla, K.K. (1998) Fibrous Materials, Cambridge University Press, Cambridge.
Elices, M and J. LLorca (eds.) (2002) Fracture in Fibers, Elsevier, Oxford.
Peebles, L.H. (1995) Carbon Fibers, CRC Press, Boca Raton, FL.
Cliapter4
• melting point
• volatility
• density
• elastic modulus
• coefficient of thermal expansion
• creep characteristics
• strength
• fracture toughness
• compatibility between fiber and matrix
•• chemical compatibility
•• thermal compatibility (should be able to withstand high
temperature excursions)
•• compatibility with the environment, internal as well as external.
The external compatibility mainly involves oxidation and evaporation
characteristics.
Figure 4.1 (a) shows a schematic of the slurry infiltration process. The first stage
involves some degree of fiber alignment, in addition to incorporation of the
reinforcing phase in the matrix slurry. The slurry typically consists of the
matrix powder, a carrier liquid (water or alcohol), and an organic binder. The
109 PROCESSING OF CERAMIC MATRIX COMPOSITES
As mentioned above, the slurry infiltration process is well suited for glass or
glass-ceramic matrix composites, mainly because the processing temperatures
for these materials are lower than those used for crystalline matrix materials.
The hot pressing process does have the limitation of not being able to produce
complex shapes. Application of a very high pressure during hot pressing can
also easily damage the fibers and decrease the strength of the composite. The
fibers may also be damaged by mechanical contact with refractory particles of
a crystalline ceramic by or from reaction with the matrix 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
undesirable. To this end, it is important to completely remove the fugitive
binder and use a matrix powder particle smaller than the fiber diameter. The hot
pressing operational parameters are also important. Precise control within a
narrow working temperature 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/matrix interfacial reaction, along with its detrimental effect on the bond
strength, are difficult to avoid in hot pressing. As with any processing
technique, the fibers should suffer little or no damage during handling.
Whisker reinforced CMCs are generally made by mixing the whiskers with a
ceramic powder slurry, dried, and hot-pressed. Sometimes hot isostatic pressing
rather than uniaxial hot pressing is used. Whisker agglomeration in a green
body is a major problem. Mechanical stirring and adjustment of pH level of the
110 PROCESSING OF CERAMIC MATRIX COMPOSITES
Stack of matrix
impregnated
fiber tapes
Fibers Pressure
Binder bumouot
Composite sod'c
Hot pressing
800-92S·C Graphite die
(a)
(b)
Fig. 4.1. Slurry infiltration process followed by hot pressing: (a) Schematic of process,
(b) Optical micrograph of a transverse section of a unidirectional alumina fiber/glass
matrix composite (some porosity can be seen in this micrograph), and (c) pressure and
temperature schedule used during hot pressing of this composite (see on p. 110).
111 PROCESSING OF CERAMIC MATRIX COMPOSITES
- - - Temperature
6
100 ------ Pressure
5
800
600
r-----------,
400
,I
I 2
200
,,
I
II
OL---~~--+---~~--~~O
o 50 100 15 200
Time, min
(c)
Reaction bonding processes similar to the ones used for monolithic ceramics
can also be used to make ceramic matrix composites. These have been used
mostly with silicon carbide or silicon nitride matrices. Among the advantages
of this method are:
Nicalon fiber
Fig. 4.2 SEM micrograph of a hybrid composite, consisting ofNicalon fibers and SiC
whiskers in a glass-ceramic matrix (Chawla et aI., 2001).
One great disadvantage of this process is that high porosity is hard to avoid.
Hot-press to consolidate
Composite
(a)
Fig. 4.3. (a) Flow diagram of the reaction bonding process for processing SCS-6 fiber/
Si3N4 composites (after Bhatt, 1990) and (b) Microstructure of SCS-6 fiber/ Si3N4
composite showing uniform fiber distribution and small amounts of residual porosity
around the periphery of the large diameter fibers (courtesy ofR.T. Bhatt).
114 PROCESSING OF CERAMIC MATRIX COMPOSITES
E
:t
E
::t
o·
o
It)
(b)
Fig 4.3 - contd
115 PROCESSING OF CERAMIC MATRIX COMPOSITES
furnace between 1100 and 1400 °C. Typically, the silicon nitride matrix has
about 30% porosity, which is not unexpected in reaction bonded silicon nitride.
Note also the matrix density variations around fibers in Fig. 4.3(b).
4.4 INFILTRATION
Piston
o Melt
o
o
o
o
Prefonn
o h~~mm-::l
o
o
oo 0Or-Heating
coils
o
o 0
o 0
o ~ __________ ~ 0
Fig. 4.4. Schematic of liquid infiltration processing. The technique is very similar to
liquid polymer or liquid metal infiltration.
h I yrtcosEl
~ 2n
where r is the radius of the cylindrical channel, t is the time, y is the surface
energy of the infiltrant, e is the contact angle, and 11 is the viscosity. Note that
the penetration height is proportional to the square root of time and inversely
proportional to the viscosity of the melt. Penetration will be easier if the contact
angle, e is low (i.e., better wettability), and the surface energy (y) and the pore
radius (r) are large. However, if the radius, r, ofthe channel is made too large,
the capillarity effect will be lost.
Growth barrier
Preform fabricatioJl
Composite {
( MaMx.~ ...",
~ CMCpllrf
Fig. 4.5. Schematic of the directed metal oxidation process (DIMOX™). (Courtesy of
Lanxide Corp.)
Chemical vapor infiltration (CVI) has been used extensively for processing near
net shape CMCs. The first attempts at using CVI as a processing technique
were in densifyingporous graphite bodies with carbon (Bickerdike et aI., 1962).
In fact, about half of the commercially available carbon-carbon composites
today are made by CVI (Lowden et aI., 1993). CVI can be thought
of as a bulk form of chemical vapor deposition (CVD), which is widely used in
depositing thin coatings. The process involves deposition of the solid matrix
over an open volume, porous fibrous preform by the reaction and
decomposition of gases. An example of a CVI reaction is the deposition of
titanium diboride, which has a melting temperature of 3225 °C but can be
deposited at 900°C via CVI:
The HCe by-product is very common in such reactions. The solid materials are
deposited from gaseous reactants onto a heated substrate. A typical CVD or
CVI process would require a reactor with the following parts:
Figure 4.7 shows such a reactor in its simplest form. One can synthesize a
variety of ceramic matrixes such as oxides, glasses, ceramics, intermetallics by
CVD. There are two main variations of the CVI process. Isothermal chemical
vapor infiltration (ICVI) relies on diffusion for deposition (Naslain et aI., 1983;
Naslain, 1992). The preform is maintained at a uniform temperature while the
reactant gases are allowed to flow through the furnace and deposit the solid
species. To obtain a uniform matrix around the fibers, deposition is conducted
at low pressures and reactant concentrations. When the CVI process is carried
out isothermally, however, surface pores tend to close first, restricting the gas
flow to the interior of the preform. This phenomenon, sometimes referred to as
FILAMENT WOUND
NICALON
TUBE BRAIDED
NEXTEL
TUBE
CVD/CVI SIC
FILAMENT WOUND
N ICALON TUBE CVD/CVI SiC
BRAIDED NEXTEL
TUBE
Fig. 4.6. Filament wound Nicalon tube and a braided Nextel tube before and after being processed by Chemical Vapor Infiltration
(CVI). (Courtesy of Thermo Electron Corp.)
(') "!rJ
00000000
N
i o;Q'
.... 01:>0.
-
(') 1-
~ -..!
_.. Q
~
< CI}
~ (') ""0
o"'1 ::r
(1)
~
o '"
0
s·= ""...... ~
,..,. 'i:l
t:n ~ ~{{~~{{{{~*~~{~~~{{{~
q(')a. ~:r~~~~~~~~~~~~~~l\ ~
\r'..r-,.r-,.r-...r-n."\r',.r.,r-TIo.n.,..n.1"Ion.,.r-,..n.~n.'\r'n.n.ri", C':l
~. a, tr:I
g ~ $$~}$~~~~~~~$$~f:§$~ en
en
,.....(') ~~~~~~$$~~j:~~;:::~~~
"'n.n.ri",..n.n.n."n.ri"...r-n.n.ri".....n.,.r-n.n.n.n.~n.fton.,.r-n.n...,r;
~ \o.."'\r'~n.n.n.ri"ri"n.n.~,..n. n.'I'\on...........
..... ri"T\n.n.n.ri"'l"tr.n.,.r,
"'n.n...r'...r-n.n.TIo.n.ri",.r-,ro..,r-n.n.,.rn.n.n.n.n.n.,r-n.n.ri\r'n.~
tl ~
'Tr'ri",.r,r-,.r~ri"ri"n.~n.n.,.r,.rn.n.n.n.,.r-n.n.l'It.'"
~i
........
"'" (') "'n,'I"Io...:"ri"TIon."r-,r-"r-,.r,.r1'lon.,.r,.r,..,1"II..,r-,.,."'"r-"r-,.r","; ~
'n.n.TIo.n.~,.r-,{"n.n.,.r""n.n.n.,.r..r'n.n.,{"n.n.n.n.n.n.ri"n."';
§ e,
Q.< :J:J ~~~~~~~~*~~~~~~
\r',.r-,.r:...,n.n.n.,{",.rn.n.,.r..... n.,..,n.,.r,.r,.r".n.~,r,
cr~ ~,.r,.rn.n.n.n.,.r,.r,.r,.rn.~ri"ri",r-,..,n.~~~
"",g ':...'.~n.n.,.rn.n,{",.r,.r,.rn.n.,."'.,.r,.rn.n.,.r,.r,.rn.n.n.1"II.n.,
00-1 'n.1"lo.n.n.,.r.....n.,.r,.r,.r.,{"ri".....n."r-,r-.....n.,.r'l"lo""TIo.n.y\,n..".n.,.r,
c;> .... m>
e ~~-
OO Z
::;] :e ~
n
d ::s -I 'O"Ij ~
(1) ::n "'I ......
Q.q 00'
~
g. ~. 0'8
"'I~ c:l
(1) 0
S'" ><
= ::s C':l
~rn ~
-<
< .... 000000
~ "'" ~
en
"'1
'8'0a
tr:I
s· (')
::ng (1)
-...,
en
- rn n~
J:t,.....
a~
0"'-'
::s tn·
\: o~.
~~
""(1Q
"il 0
(")st-
~~
""'e,
122 PROCESSING OF CERAMIC MATRIX COMPOSITES
The deposition process in CVI is governed primarily by mass and heat transfer,
as well as chemical kinetics (Carlsson, 1990; Gokoglu, 1992). 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
fibrous 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 reach the hot zone, they
decompose and deposit on and between the fibers to form the matrix.
As the matrix material gets deposited in the hot portion of the preform, the
preform density increases; and the hot zone moves progressively from the top
of the preform toward the bottom. When the composite is formed completely
at the top and is no longer permeable, the gases flow radially through the
preform, exiting from the vented retaining ring. To control deposition, the rate
of deposition must be maximized while minimizing density gradients.
Deposition reaction and mass transport are competing factors, so very rapid
deposition results in the exterior of the preform being well infiltrated, while
severe density gradients and a large amount of porosity are present within the
preform. Very slow deposition rates, on the other hand, require large times and
are not economically feasible. A balance between the two factors is required for
optimum infiltration.
The vapors of SiC deposit as solid phases on and between the fibers in a
freestanding preform to form the matrix. The CVI process is very slow because
it involves diffusion ofthe reactant species to the fibrous substrate, followed by
outflow of the gaseous reactant products. The CVI process of making a ceramic
matrix is, indeed, a kind of low stress and low temperature CVD process, and
thus avoids some of the problem associated with high temperature ceramic
processing. Using CVI, when processing CFCMCs, one can deposit the
interfacial coating on the fibers as well as the matrix in situ. For example, for
Nicalon/SiC composites with a carbon interface, the carbon layer is deposited
first, and then the SiC matrix is infiltrated without changing the preform
conditions. The fibrous preforms are stacked layer by layer between perforated
plates, through which the gases pass during infiltration. The carbon coating is
typically deposited by means of a hydrocarbon gas at around 1000°C and
reduced pressure, to protect the fibers.
The graphitic coating on the fibers has a characteristic aligned structure of the
basal planes. These basal planes are parallel to the fiber direction, but
perpendicular to the incoming crack front, so deflection of cracks at the weakly
bonded basal planes takes place instead of fracturing the fibers. The softer c-
axis of the graphite is also aligned in the perpendicular direction, to
accommodate the thermal residual stresses that arise from processing. The
matrix consists of a nucleation zone in a small region at the coating/matrix
interface. After this, long columnar grains are seen which are perpendicular to
the surface of the fiber. The preferred orientation is such that the (111) planes
are aligned parallel to the fibers. The grains are composed predominantly ofb-
SiC with a cubic structure with small disordered regions of a-SiC. For CVI
composites reinforced with woven fiber fabrics, the nature of the porosity is
trimodal. Macroporosity is found between fiber bundles and between layers of
fabric, with pore sizes less than 100 mm. Microporosity occurs between fibers
in the fiber bundle, and the pore size is usually on the order of 10 mm. Lowden
et al. (1993), found that 70% of the pore volume was in the form of
microporosity within the fiber bundle, 25% between the cloth layers, and 5%
as holes between layers of the fabric.
This variant of CVI which combines forced gas flow and temperature gradient
avoids some of the problems mentioned earlier. Under these modified
conditions, 70-90% dense SiC and Si3N4 matrixes can be impregnated in SiC
and Si3N4 fibrous preforms in less than a day. Under conditions of plain CVI,
it would take several weeks to achieve such densities, i.e., one can reduce the
processing time from several days to less than 24 hours. One can also avoid
using binders in this process with their attendant problems of incomplete
removal. The use of a graphite holder simplifies the fabrication of the preform
and the application of a moderate pressure to the preform can result in a higher
than normal fiber volume fraction in the final product. The final obtainable
124 PROCESSING OF CERAMIC MATRIX COMPOSITES
density in a ceramic body is limited by the fact that closed porosity starts at
about 93-94% oftheoretical density. It is difficult to impregnate past this point.
The term reaction bonding is used rather loosely in literature. Some researchers
use it to encompass all processes not involving hot pressing. We shall restrict
the term to processes involving chemical reaction(s) between components to
produce the desired end product. Reactive consolidation or liquid phase
sintering are the other more descriptive terms of such processes. Some
commercial success has been obtained in making silicon carbide based-
composites by reaction bonding. Siliconized silicon carbide is the name give to
a composite of SiC grains in a silicon matrix. Commercially, such composites
are available under different designations (e.g., K- T, Refel, and NC-435) from
different producers.
Polymers are used to bond the preforms containing carbon in the form of carbon
and silicon carbide, followed by pyrolysis of the polymer to give more carbon
(Forrest et aI., 1972; Willermet et aI., 1978). One such CMC, called Refel
(Forrest et aI., 1972), is formulated to give a minimum of free silicon, which in
practice is limited to about 10 vol. %. The infiltration is done under reduced
pressure at temperatures up to 1700 °C and for hours. Hillig et al. (1975) used
a low modulus carbon fiber instead of carbon powder in the preform. Molten
silicon reacts with carbon fibers to form SiC. The original geometry of the
carbon fibers is retained. A big advantage of SiC/Si composite is that the
constituents are in chemical eqUilibrium and they have closely matched thermal
expansion coefficients. Carbon fiber in the form of cloth, tow, felt, or matte is
used as a precursor. A preform is made of carbon fiber and infiltrated with
liquid silicon. Silicon reacts with carbon fibers to form SiC fibers in a Si
matrix. Typical composition of the resultant composite is Si (30 - 50 %) + SiC
fiber. Silicon matrix limits the use temperature to about 1400 OC.
125 PROCESSING OF CERAMIC MATRIX COMPOSITES
4.7 SOL-GEL
Sol-gel techniques, which have been used for making conventional ceramic
materials, can also be used to make ceramic matrix materials in the interstices
of a fibrous preform. We have described the sol-gel technique in chapter 2.
Very briefly, a solution containing metal compounds, e.g., a metal alkoxide,
acetate, or halide is reacted to form a sol. The sol is converted to a gel, which
in turn is subjected to controlled heating to produce the desired end-product: a
glass, a glass-ceramic, or a ceramic. Characteristically, the gel to ceramic
conversion temperature is much lower than that required in a conventional
melting or sintering process. A schematic of a typical sol-gel process for
processing CMCs is given in Fig. 4.8. Some of the advantages of these
techniques for making composites are the same as the ones for monolithic
ceramics, viz., lower processing temperatures, greater compositional
homogeneity in single phase matrices, potential for producing unique
multiphase matrix materials, etc. Specifically, in regard to composite material
fabrication, the sol-gel technique allows processing via liquids oflow viscosity
such as the ones derived from alkoxides. Covalent 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 CVD processes. Among the
disadvantages of sol-gel are high shrinkage and low yield compared to slurry
techniques. The fiber network provides a very high surface area for the matrix
to gel on. Consequently, the shrinkage during the drying step, frequently, to
results in a large density of cracks in the matrix. Generally, repeated
impregnations are required to produce a substantially dense matrix. It is easy
to see that many of the polymer handling and processing techniques can be used
for sol-gel as well. Impregnation of fibrous preforms in vacuum and filament
winding are two important techniques. In filament winding, fiber tows or
rovings are passed through a tank containing the sol and the impregnated tow
is wound on a mandrel to a desired shape and thickness. The sol is converted
to gel and the structure is removed from the mandrel. A final heat treatment
then converts the gel to a ceramic or glass matrix.
The sol-gel technique can also be used to prepare prepregs by the slurry
126 PROCESSING OF CERAMIC MATRIX COMPOSITES
o
o .....t - - --
o
o
~t---- Hea ting Coils
o
HcOl. to Con\Iertlhc Gel
into Gla s o rCeramic
infiltration method. The sol in the slurry acts as a binder and coats fibers and
glass particles. The binder burnout step is thus eliminated because the binder,
being of the same composition of the matrix, becomes part of the glass matrix.
Hurwitz et aI., 1989). These include complex shape forming and fabrication.
Furthermore, by processing and pyrolyzing at lower temperatures (compared to
sintering and hot-pressing, for example) one can avoid fiber degradation and the
formation of unwanted reaction products at the fiber/matrix interface. French
(1996) lists some desirable characteristics in a preceramic polymer:
Typically, one chooses the range of the molecular mass of the polymer,
followed by shaping of the product (Greil, 1995). The polymer is then
crosslinked and finally pyrolyzed in an inert or reactive atmosphere (e.g., NH3)
at temperatures between 1000-1400°C. The pyrolysis involves the following
three steps. In the first step, between 550-880 °C, an amorphous hydrogenated
compound of the type Si(C.ObNcBd) is formed. The second step involves
nucleation of crystalline precipitates such as SiC, Si3N4 , and Si02 at
temperatures between 1200 and 1600°C. Grain coarsening may also result from
consumption of any residual amorphous phase and reduction in the amount of
128 PROCESSING OF CERAMIC MATRIX COMPOSITES
oxygen due to vaporization ofSiO and CO. Porosity is typically of the order of
5-20 vol.% with pore sizes of the order of 1 to 50 nm. It should be noted that the
average pore size and volume fraction of pores decreases with increasing
pyrolysis temperature, since the amount of densification (and shrinkage)
becomes irreversible at temperatures above the maximum pyrolysis
temperature.
The main disadvantage of PIP is the low yield of ceramic during the polymer-
to-ceramic transformation and the resulting shrinkage, which typically causes
cracking in the matrix during fabrication (Greil, 1995; French, 1996; Hurwitz
et aI., 1989; Lundberg et aI., 1990; Sirieix et aI., 1990; Sato et al. 1995; Gonon
et aI., 1995). Because of the shrinkage and weight loss during pyrolysis,
residual porosity after a single impregnation is of the order of 20-30%. To
reduce the amount of residual porosity, multiple impregnations are needed.
Reimpregnation is typically conducted with a very low viscosity prepolymer,
so that the slurry may wet and infiltrate the micropores existing in the preform.
Usually, reimpregnation is done by immersing the part in the liquid polymer in
a vacuum bag, while higher viscosity polymers require pressure impregnation.
Typically, the amount of porosity will reduce from 35 % to less than 10 % after
about 5 impregnations.
Significant gas evolution also occurs during pyrolysis (French, 1996). Thus, it
is advisable to allow these volatile gases to slowly diffuse out of the matrix,
especially for thicker parts. Typically, pyrolysis cycles ramp to 800-1400 °C
over periods of 1-2 days, to avoid delamination. Recall that pyrolysis must be
done at a temperature below the crystallization temperature of the matrix (or
large volume changes will occur) and below the degradation temperature of the
reinforcing fibers. The pyrolysis atmosphere is most commonly argon and
nitrogen, although in ammonia a pure amorphous silicon nitride with low
amounts of free carbon, can be obtained (Burns and Chandra, 1989). Such an
atmosphere may also lead to the formation of nitrides from the reaction of filler
particles (Erny et aI., 1993). With the formation of the ceramic, the gaseous by-
product reaction can be written as (Greil, 1995):
where P is the polymer, C is the ceramic, and G is the gaseous by-product. With
the loss of the volatile gaseous products, the ceramic yield, a is the ratio of
ceramic formed and the initial amount of polymer:
a = m(C) = 1- meG)
m(P) m(P)
129 PROCESSING OF CERAMIC MATRIX COMPOSITES
P(P)
{3= P(C)
'V = a~ - 1
If the volume is constrained (Le., 'V = 0) then shrinkage does not occur, but a
large amount of residual porosity is present. The maximum amount of porosity
can be written as:
1t =1- a~
Fitzer and Gadow (1986) used repeated infiltration and in situ thermal
decomposition of porous reaction-bonded ceramics; for example infiltration of
porous preforms of silicon carbide and silicon nitride with silazanes and
polycarbosilanes, to obtain Si3NiSiC composites. The following steps are taken
in processing the composites:
(i) Porous SiC or Si3N4 fibrous preform with some binder phase is prepared.
(ii) Fibrous preform is evacuated in an autoclave.
(iii) Preform is infiltrated with molten precursors, silazanes or polycarbosilanes,
at high temperature (780 K) and the argon or nitrogen pressure is slowly
increased from 2 to 40 MPa. The high temperature results in a transformation
of the oligomer silane to polycarbosilane and simultaneous polymerization at
high pressures.
(iv) Infiltrated samples are cooled and treated with solvents.
(v) Samples are placed in an autoclave and the organosilicon polymer matrix
is thermally decomposed in an inert atmosphere at a high pressure and at
temperature in the 800-1300 K range.
130 PROCESSING OF CERAMIC MATRIX COMPOSITES
(vi) Steps (ii) through (v) are repeated to attain an adequate density. To produce
an optimum matrix crystal structure, the material is annealed in the l300- 1800
K range.
• Reduce and disrupt the formation of matrix cracks that form during
shrinkage of the polymer.
• Enhance ceramic yield by forming reaction products during pyrolysis.
• Strengthen and toughen the weak amorphous matrix and increase the
interlaminar shear strength ofthe composite.
The filler must be submicrometer in size in order to penetrate the tow bundle
and the filler's coefficient of thermal expansion must match that of the
polymeric matrix. It should be noted that the filler must not be used in very high
fractions and the slurry should not be forced into the reinforcing fibers since
abrasion of the fiber fabric may take place. This is especially true with hard,
angular fillers or ceramic whiskers. Typically, the volume fraction of filler is
15-25% of the matrix volume fraction. High filler loading may result in an
increase in interply spacing and lower volume fraction of fibers. When an
"active" filler phase is added to the polymer, it reacts with solid or gaseous
decomposition products to form new carbide phases (Greil, 1995):
where T is the active filler, and M is the carbide phase formed. The other
symbols have the same meaning as described in the previous expression. The
maximum volume change of a precursor containing an active filler, \/f*, can be
expressed as:
131 PROCESSING OF CERAMIC MATRIX COMPOSITES
where V T' is the critical volume fraction of filler that determined maximum
particle packing density of the reacted filler phase in the pyrolyzed product, a ™
and WM describe mass change of the filler phase and density ratio. For a™ and
~TM > I, volume expansion of the filler phase may compensate for polymer
shrinkage during pyrolysis.
• •• •
•• •
• •
Filler Network
(a)
(b)
Fig. 4.9 (a) Schematic oftiller-free versus active filler pyrolysis and (b) microstructure
of a woven Nicalon/Si-C-O-N matrix composite with a SiC filler, showing small
amounts of residual porosity and a clear filler free-network in the matrix of the
composite (Chawla et al., 1998).
133 PROCESSING OF CERAMIC MATRIX COMPOSITES
the compact at the top in a cold-walled vessel. A combustion wave will pass
through the compact, giving titanium carbide.
Many ceramics such as borides, carbides, nitrides, silicides, and sialons as well
as composites such as SiCw + Al2 0 3 have been synthesized by means of SHS.
The SHS process gives a weakly bonded compact. Therefore, the process is
generally followed by breaking the compact, milling, and consolidation by
some technique such as HIPing. Explosive or dynamic compaction can result
in a relatively dense product. A good example of an SHS process to make
composites is the proprietary process of Martin Marietta Corp., USA, called the
XD™ Process, wherein exothermic reactions are used to produce multiphase
alloy powders. These are hot pressed at 1450 °C to full density. Reinforcement
in the form of particles, whiskers, and platelets can be added to the master alloy
to make a composite. A good example is that of TiB2 particles, about 1 !lm
diameter, distributed in intermetallic matrixes such as TiAI, TiAI + Ti3AI, NiAI,
etc.
The phenomenon of electrophoresis has been known since the beginning of the
nineteenth century, but its applications in are of processing of ceramics and
ceramic composites is relatively recent. Electrophoretic deposition (EPD)
should not be confused with electroplating. In electroplating, ions are the
moving species and they undergo ion reduction on deposition. In EPD, on the
other hand, solid particles migrate with no charge reduction on deposition.
Also, the deposition rate in EPD is» 1 mmlmin while in electroplating it» 0.1
mmlmin. Electrophoretic deposition is a relatively simple and inexpensive
technique, which can be profitably exploited for infiltration of tightly woven
fibre preforms (Illston et aI., 1993). EPD makes use of nanoscale ceramic
particles in a stable non-agglomerated form (such as in a sol or colloidal
suspension) and exploits their net surface electrostatic charge characteristics
while in suspension. On application of an electric field the particles will
migrate toward and deposit on an electrode. If the deposition electrode is
replaced by a conducting fibrous preform, the suspended particles will be
attracted into and deposited within it, providing an appropriate means of
effectively infiltrating densely packed fibrous bundles. A schematic diagram of
the basic EPD cell is shown in Fig. 4.10a while an SEM micrograph of a
134 PROCESSING OF CERAMIC MATRIX COMPOSITES
Nicalon fiber mat infiltrated with a mixed sol of mullite composition using
electrophoretic deposition is shown in Fig. 4.1 Ob (Kaya et aI., 2000). Note the
high-level of ceramic matrix infiltration achieved by this technique.
The movement of ceramic sol particles in an aqueous suspension within an
electric field is governed by the field strength, and the pH, ionic strength and
viscosity of the solution (Illston et aI., 1993). The electrophoretic mobility of
charged particles in a suspension is given by the following equation called the
Smoluchowski equation (Brown and Salt, 1965):
Boehmite sol
Fig. 4.10 (a) Schematic of a basic electrophoretic deposition (EPD) cell (after Kaya et
aI., 2000) (b) Transverse section of a Nicalon fiber mat infiltrated with a mixed sol of
mullite composition by electrophoretic deposition (Boccaccini, et aI., 2001).
135 PROCESSING OF CERAMIC MATRIX COMPOSITES
(b)
4.11 SUMMARY
Table 4.1 Summary ofCMC fabrication processes and examples of typical composite
systems fabricated by these processes.
Process Examples
Slurry infiltration SiC/Glass-Ceramic, CarboniGlass-
(ply stacking and hot pressing) ceramic, C/glass, Mullite/Glass
Powder processing and hot pressing SiCIAIP3' AIP/AlP3
Gas-liquid metal reaction SiC/AI 20 3, SiC/SiC
(Lanxide)
Sol-gel C/glass, Mullite/Mullite
(infiltration and sintering/hot pressing)
Chemical vapor infiltration SiC/SiC, C/SiC
(infiltration of a woven preform)
Polymer conversion C/C, C/SiC, SiC/Si-C-N
(infiltration and pyrolysis)
l36 PROCESSING OF CERAMIC MATRIX COMPOSITES
REFERENCES
Barclay, S.J., J.R. Fox, and H.K. Bowen (1987) J. Mater Sci., 22, 4403.
Bhatt R. T. (1986), National Astronautics and Space Administration, NASA TN-88814.
Bhatt R.T. (1990) J. Mater. Sci., 25, 3401.
Bickerdike, R.L., A.R.G. Brown, G. Hughes, and H. Ranson (1962) Proc. o/the Fifth
Conference on Carbon, S. Mrosowski, M.C. Studebaker, and P.L. Walker (eds.),
Pergamon Press, Buffalo, NY, p. 575.
Boccaccini, A.R., C. Kaya, and K. K. Chawla (2001) Composites Part A, 32,997.
Bordia, R.K.. and R. R~ (1988) J. Am. Ceram. Soc., 71, 302.
Brown, D.R. and F.W. Salt (1965)J. App. Chem., 15,40.
Brennan, J.J. and K.M. Prewo (1982) J. Mater. Sci., 17,2371.
Burkland, C.V., W.E. Bustamante, R. Klacka. and J.-M Yang (1988) in Whisker- and
Fiber-Toughened Ceramics, ASM IntI., Materials Park, Ohio, p. 225.
Bums, G.T. and G. Chandra (1989) J. Am. Ceram. Soc., 72,334.
Caputo, A.J., D.P. Stinton and R.A. Lowden.(1987) Am. Ceram. Soc. Bull., 66,1987,
368.
Carlsson, J.O. (1990) Thin Solid Films, 168, 19.
Chawla, K.K. (1987) Composite Materia/s, 2 nd ed.,Springer-Verlag, New York.
Chawla N., K.K. Chawla, M. Koopman, B. Patel, C.C. Coffm, and J.I. Eldridge (2001)
Compo Sci. Tech., 61, 1923.
Chawla, N., Y.K. Tur, J.W. Holmes, J.R. Barber, andA. Szweda(1988)J. Am. Ceram.
Soc., 81, 1221.
Chawla, N. (1997), Metall. & Mater. Trans. A, 28A, 2423.
Claussen, N., T. Le, and S. Wu (1989) J. Eur. Ceram. Soc., 5, 29.
Claussen, N., S. Wu, and D. Holtz (1994) J. Eur. Ceram. Soc. ,14, 209.
Cornie, J.A. , Y.-M. Chiang, D.R. Uhlmann, A. Mortensen, and J.M. Collins (1986)
Am. Ceram. Soc. Bull., 65, 293.
De Jonghe, L.C., M.N. Rahaman, C.H. Hseuh (1986) Acta Met., 39, 1467.
Erny, T., M. Seibold, O. Jarchow, and P. Greil (1993) J. Am. Ceram. Soc. 76,207.
Fitzer, E. and D. Hegen (1979) Angew. Chem., 91, 316.
Fitzer, E. and J. Schlichting (1980) Z. Werkstofftech., 11, 330.
Fitzer, E. and R. Gadow (1986) Am. Ceram. Soc. Bull., 65, 326.
Forrest, C. W., P. Kennedy, andJ. V. Shennan (1972) Special Ceramics, British Ceramic
Research Association, Stoke-on-Trent, U.K., vol. 5, p.99.
French, J.E. (1996) in Handbook o/Continuous Fiber Ceramic Composites, American
Ceramic Society, Westerville, OH,), p. 269.
Gokoglu, S.A.(1992) Mater. Res. Soc. Symp. Proc. vol. 250, p.17.
Gonon, M.F., G. Fantozzi, M. Murat, and J.P. Disson (1995) J. Eur. Ceram. Soc. 15,
185.
Greil, P. (1995) J. Am. Ceram. Soc., 78, 835.
Homeny, J., W.L. Vaughn, and M.K. Ferber (1987) A mer. Cer. Soc. Bull., 67, 333.
Hurwitz, F.I., J.Z. Gyekenyesi, and P.J. Conroy (1989) Ceram. Eng. Sci. Proc., 10,750.
Hurwitz, F.J. (1992) NASA Tech. Memo, 105754.
Illston, T.J, C.B. Ponton, P.M. Marquis, E.G. Butler (1993) Third Euroceramics, vol.
l37 PROCESSING OF CERAMIC MATRIX COMPOSITES
1, P. Duran and J.F. Fernandez (eds.), Faenza Editirice Iberica, Madrid, pp. 419-424.
Kaya,C., A.R Boccaccini., and K.K. Chawla (2000) J. Am. Ceram Soc., 20, 1189.
Kristofferson, A., A. Warren, J. Brandt, and R. Lundberg (1993) in Proc. Int. Cont
HTCMC-I, (ed. R. Naslain et al.),Woodhead Pub., Cambridge, UK,., p. 151.
Lipowitz, J., lA. Rabe, L.K. Frevel, and RL.Miller (1990) J. Mater. Sci., 25, 2118.
Liu, H.Y., N. Claussen, M.J. Hoffmann, and G. Petzow (1991)J. Eur.Ceram. Soc. 7,
41.
Lowden, RA, D.P. Stinton, and T.M. Besmann (1993) in Handbook o/Continuous
Fiber Ceramic Matrix Composites, Amer. Ceram. Soc., Westerrville, 0 H, p. 205.
Lundberg, R., R. Pompe, and R. Carlsson (1990) Compo Sci. Tech. 37, 165.
Naslain, R et al. (1983) Euro-CVD-Four, The Centre, Eindhoven, p. 293.
Naslain, R. (1992) in Ceramic Matrix Composites, R. Warren, ed., Chapman and Hall,
London, p. 199.
Phillips, D.C. (1983) in Fabrication o/Composites, North-Holland, Amsterdam, p. 373.
Prewo, K.M.(1982)J. Mater. Sci., 17,3549.
Prewo, K.M.(1986) in Tailoring Multiphase and Composite Ceramics, Materials
Science Research, Plenum Press, New York, vol. 20, p. 529.
Prewo, K.M. and J.l Brennan (1980) J. Mater. Sci., 15,463.
Prewo K.M., J.J. Brennan, and G.K. Layden (1986) Am. Ceram. Soc. Bull., 65, 305.
Rahaman, M.N. and L.C. De Jonghe (1987) J. Am. Ceram. Soc., 70, C-348.
Raj, R.and R.K Bordia (1989) Acta Met., 32, 1003.
Riedel, R., G. Passing, H. Schonfelder, and R.J. Brook (1992) Nature, 355, 355.
Sacks, M.D., H.W. Lee, and O.E. Rojas (1987) J. Am. Ceram. Soc., 70, C-348.
Sambell, R.A.J., D.C. Phillips, and D.H. Bowen (1974) in Carbon Fibres: Their Place
in Modern Technology, The Plastics Institute, London, p. 16/9.
Sato, K., H. Morozumi, A Tezuka, O. Funayama, and T. Isoda (1995), in High
Temperature Ceramic-Matrix Composites II, American Ceramic Society, Westerville,
OH,p.199.
Shalek, P.D., J.J. Petrovic, G.F. Hurley, F.D. Gac (1986) Am. Ceram. Soc. Bu/l., 65,
351.
Sirieix, F., P. Goursat, A. Lecomte, and A Dauger (1990) Camp. Sci. Tech., 37, 7.
Stinton, D.P., A.J. Caputo, and R.A. Lowden (1986) Am. Ceram. Soc. Bull., 65, 347.
Stinton, D.P., AJ. Caputo, R.A Lowden, and T.M. Besmann (1986) Ceram. Eng. Sci.
Proc., 7, 983.
Urquhart, AW. (1991) Mater. Sci. Eng., A144, 75.
Wu, S. and N. Claussen (1994) J. Amer. Ceram. Soc., 77,2898.
Yang, M. and R. Stevens (1990) J. Mater. Sci., 25, 4658.
SUGGESTED READING
Krenkel, W., R Naslain, and H. Schneider (eds.) (2001) High Temperature Ceramic
Matrix Composites, Wiley-VCH, Weinheim, Germany.
Kroke, E., Y.-L. Li, C. Konetschny, E. Lecomte, C. Fasel, and R. Riedel (2000)
138 PROCESSING OF CERAMIC MATRIX COMPOSITES
Silazane derived ceramics and related materials, Mater. Sci. Eng., R26, 97.
Narula, C.K. (1995) Ceramic Precursor Technology and its Applications, Marcel
Dekker, New York.
National Materials Advisory Board (1991) High Temperature Metal and Ceramic
Matrix Compositesfor Oxidizing Atmosphere Applications, NMAB-376, Washington,
DC.
Cliapter 5
INTERFACE
The interface region has a great deal of importance in determining the ultimate
properties of a given composite. There are two main reasons for this importance
of the interface region:
(i) the interface occupies a very large area per unit volume in composites (see
Sec. 5.1 below), and
(ii) in general, the matrix and reinforcement form a system that is not in
thermodynamic equilibrium. This can lead to a series of interaction at the
interface.
We can define an interface between any two phases (for example, between a
reinforcement and a matrix) as a surface across which there occurs a
discontinuity in one or more material parameters or properties, such as lattice
parameter, density, elastic modulus, thermal expansion coefficient, strength,
fracture toughness, etc. This definition of an interface treats it as a
bidimensional entity over which the discontinuity occurs rather abruptly. It
would, however, be more realistic to treat it as an interface zone or region
having a certain thickness across which the discontinuity in property occurs
gradually.
Consider a fiber reinforced composite. It can readily be shown that, for a given
fiber volume fraction, the total area of the fiber/matrix interface increases as the
fiber diameter decreases. Consider a composite bar containing N continuous,
unidirectionally aligned fibers, each of diameter d. Let the length, width, and
height of the composite be~, w, and h, respectively. Then, for the fiber volume
fraction, we can write
Equation (5.3) says that the fiber/matrix interfacial area, for a given fiber
volume fraction, varies inversely with the fiber diameter. Let us take the volume
of the composite to be 1 m3, and in order to simplify Eq. (5.3), let the fiber
volume fraction be 0.25. Then, Eq. (5.3) reduces to
(5.4)
d
where IA is now the interfacial area per unit volume in a composite containing
25% fibers by volume. Figure 5.1 shows a plot of the interfacial area as a
function of lid, where we have indicated some of the commercial fibers. As one
goes from the extremely tiny whiskers to rather large diameter CVD type fibers
such as SiC, B, etc., the interfacial area decreases as lid.
10'
10 1
10°
10- 1
10-'
10- 4
10° 10 1 10' 103
a,l'm
Fig. 5.1 Interfacial area, IA (in m for a fiber volume fraction of 0.25 as a function of
2)
fiber diameter, d (in ~m). For a given fiber volume fraction, the finer the fiber, the
higher the interfacial area.
volume of particles
volume of composite 1
1.5
(5.5)
d
Equations (5.4) and (5.5) show similar inverse dependence of interfacial area
on the diameter of the fiber or the particle.
142 INTERFACE
The type of bonding at the interface, which controls the interface strength, is of
obvious importance in ceramic matrix composites. In ceramic matrix
composites made by melt-infiltration, there is another aspect that assumes
importance, viz., wettability. We briefly discuss these concepts first and then
discuss the types of interface bonding, methods of interface testing, etc.
5.3 WETTABILITY
Ysv
YIv
From the equilibrium of forces in the horizontal direction in Fig. 5.3, we have;
(5.6 )
For e = 0°, we have perfect wetting, while for e = 180°, we have no wetting.
For 0 °< e < 180 0 , there will be partial wetting. The contact angle for a given
system can vary with temperature, hold time, presence of any adsorbed gases,
145 INTERFACE
It is important to realize that wettability and bonding are not synonymous terms.
Wettability describes the extent of intimate contact between a liquid and a
solid. It does not necessarily mean a strong bond at the interface. One can have
excellent wettability and a van der Waals type bonding, i.e., a weak or low
energy bond. A low contact angle, meaning good wettability, is a necessary but
not sufficient condition for strong bonding.
1. Mechanical bonding
2. Chemical bonding
Mechanical keying between two surfaces can lead to bonding between them
and can be quite important in ceramic matrix composites. Figure 5.4a shows an
ideal, smooth interface while Fig. 5.4b shows a real interface which is rough,
which will allow for interlocking. Imagine, for example, a situation where the
matrix in a composite radially shrinks more than the fiber on cooling from a
high temperature. This would lead to a gripping of the fiber by the matrix even
in the absence of any chemical bonding, Fig. 5.5. The matrix penetrating the
crevices on the fiber surface, by liquid or viscous flow or high temperature
diffusion, can also lead to some mechanical bonding. In Fig. 5.5, we have a
radial gripping stress, 0"" which can be related to the interfacial shear strength,
'ti> as
Ideal Real
fiber fiber
.~: ~~y
;; matrix
::x.
(a) (b)
Fig. 5.4 (a) An ideal, smooth interface; (b) a real, rough interface.
T·1
~--ar
Matrix Fiber
Fig. 5.5 A frictional or mechanical interface with a transverse gripping force (Le., in
the radial) direction.
147 INTERFACE
where x is the thickness of the reaction zone, D is the diffusivity, and t is the
time. The diffusivity, D, depends on the temperature in an exponential manner
D = A exp( - i\Q/kT)
where A is a pre-exponential constant, i\Q is the activation energy for the rate
controlling process, k is the Boltzmann's constant, and T is the temperature in
kelvin.
failure mode.
Ifwe examine the key issues involved in developing ceramic matrix composites
that show damage-tolerant or enhanced toughness characteristics, it soon
becomes clear that by far the most important area in this regard is the
reinforcement/matrix interface. In particular, fiber/matrix deb onding, crack
deflection and fiber pullout can contribute to toughness enhancement. The
sequence of failure of a CMC as a function of the interfacial bond strength is
depicted in Fig. 5.6. If the interfacial bond between the fiber and ceramic matrix
is very strong, then a crack starting at any point in the composite will rapidly
run through the composite cross section and result in a planar fracture, Fig.
5.6a. In other words, the presence of a strong fiber/matrix bond does not allow
any extra expenditure of energy in the fracture process, i.e., the fracture process
in a CMC having a strongly bonded reinforcement/matrix interface will be a
low energy fracture process, not unlike that encountered in monolithic
ceramics. In fact, ifthe modulus of the fiber is not very high compared to that
of the matrix, then in a strongly bonded CMC, virtually, no gain in toughness
is obtained. However, if the interface is weak, then it can act as a crack arrestor,
by deflecting the propagating crack in the matrix in a direction perpendicular
to its original direction, Fig. 5.6b.
r r
Fibe
"'.... ~
>
)
;:;;=
, ,
/~ ,
Matrix
~
Strong
~
Weak
interfac interface
(a) (b)
Fig. 5.6 Failure of a CMC as a function of interfacial bond strength: (a) strong
interfacial bond; (b) weak interfacial bond.
149 INTERFACE
y Interface
)-x
y
~----------------------.x
Distance from notch
Fig. 5.7 Schematic of stress distribution at a crack tip (after Cook and Gordon,
1964).
151 INTERFACE
thermal stresses due to the thermal mismatch between SiC and alumina
(Angelini et aI., 1987). We treat the subject of thermal stresses in composites
in chapter 8.
(5.7)
where a is the net section compressive stress (i.e. load/minimum area), E is the
Young's modulus, v is the Poisson's ratio, and the subscripts f and m denote
fiber and matrix, respectively. One can measure the net section stress, a,
corresponding to interface debonding and compute the interfacial tensile
strength from the above expression. There are some important points that
should be taken into account before using this test. One needs a special mold
to prepare the specimen and a very precise alignment of the fiber along the axis
152 INTERFACE
Fig. 5.8 A curved neck specimen of a composite containing a single fiber along its
central axis.
Bend tests are easy to perform but they do not give a true measure of the
interfacial strength. We describe below some variants bend tests.
P/2 P/2
j-,
)
(a)
P/2 Pf2
(b)
Fig. '5.9 Transverse three-point bend tests: (a) fibers aligned perpendicular to the
loading direction; (b) fibers aligned parallel to the loading direction. Qdenotes the fiber
direction. In both cases, the fiber/matrix interface in the outermost layer will be put in
tension.
configurations, the fracture will occur on the outermost surface of the specimen
which is under a tensile stress and thus will put the fiber/matrix interface under
tension, which will be a measure of tensile of the fiber/matrix interface. The
transverse strength is given by
154 INTERFACE
3PS
cr =-- (5.8)
where P is the applied load, S is the load span, b is the specimen width, and h
is the specimen height.
This test is also known as the InterLaminar Shear Strength (ILSS) test. In this
test, the fibers are aligned parallel to the length ofthe three-point bend bar, Fig.
5.10. In such a test, the maximum shear stress, 't, occurs at the midplane and is
given by
3P
't = - - (5.9)
4bh
The maximum tensile stress occurs at the outermost surface and is given by Eq.
5.8. Dividing Eq. 5.9 by Eq. 5.10, we get
't h
(5.10)
cr 2S
Equation 5.10 says that if we make the load span, S, very small, we can
maximize the shear stress, 't, such that the specimen fails under shear with a
crack running along the midplane.
The reader should bear in mind that the interpretation of this test is not
straightforward. Clearly, the test becomes invalid if the fibers fail in tension
before shear induced failure occurs. The test will also be invalid if shear and
tensile failure occur simultaneously. It is advisable to make an examination of
the fracture surface after the test and make sure that the crack is along the
interface and not through the matrix.
This test was originally developed for use with isotropic materials (Iosipescu,
1967). It is a special type of bend test that can be sued to measure interfacial
155 INTERFACE
....-- - - - S - - - - - . I
PI2 P/2
~,
Fig. 5.10 ILSS (InterLaminar Shear Strength) test. 1 denotes the fiber direction.
shear strength parallel and perpendicular to the fiber directions. In this test a
double-edged notched specimen is subjected to two opposing force couples.
This is a special type of four-point bend test in which the rollers are offset, as
shown in Fig. 5.11., to accentuate the shear deformation. A state of almost pure
and constant shear stress is obtained across the section between the notches by
selecting a proper notch angle and notch depth (90 0 and 22% of full width).
The average shear stress is given by
P
't =- (5.11)
bh
where P is the load, b is the breadth and h is the height of specimen. The most
attractive point of this test that a large region of uniform shear is obtained vis
a vis other tests. However, there can be a substantial stress concentration near
the notch tip in orthotropic materials (not so in isotropic materials) such as fiber
reinforced composites. The stress concentration is a function of the fiber
volume fraction and the fiber orientation.
Single fiber pullout and pushout tests have been devised to measure interfacial
characteristics. Such tests give load vs. displacement curves, with a peak load
corresponding to fiber/matrix debonding and a frictional load corresponding to
the fiber pullout from the matrix. The mechanics and interpretation of these
tests is rather involved and knowledge of the underlying assumptions is
important in order to get a useful information from such tests. Some of these
aspects related to interfaces are discussed in Chapter 9. The main simplification
that is frequently made in the analysis of such tests involves averaging the load
values over the entire interfacial surface area to get the interface debond
strength and/or frictional strength. Analytical and finite element analyses show
that the shear stress is a maximum close to the fiber end and falls rapidly within
a distance of a few fiber diameters. Thus, one would expect the interface
debonding to start near the fiber end and progressively propagate along the
embedded length. We describe the salient features ofthese tests.
Single fiber pullout tests are frequently used in PMCs and to a lesser extent in
MMCs. These tests can provide useful information about the interface strength
157 INTERFACE
in model composite systems. They are not very helpful in the case of
commercially available composites. One must also carefully avoid any fiber
misalignment and introduction of bending moments. The mechanics of the
single fiber pullout test is rather complicated (Cham is, 1974; Penn and Lee,
1989; Marshall et aI., 1992; Kerans and Parthasarthy, 1991). In all the variants
of such test, the fiber is pulled out of the matrix in a tensile testing machine and
a load vs. displacement record is obtained.
Figure 5.12a shows the experimental setup for such a test. A portion of fiber,
length Q, is embedded in a matrix and a pulling tensile force is applied as
shown. If one measures the stress required to pull the fiber out of the matrix as
a function of the embedded fiber length, the plot shown in Fig. 5.12b is
obtained. The stress required to pull the fiber out without breaking it increases
linearly with the embedded fiber length, up to a critical length, Qc ' For an
embedded fiber length, Q~Qc, the fiber will fracture under the action of the
tensile stress, a, acting on the fiber. The tensile stress, a, on the fiber results in
a shear stress, 't, at the fiber/matrix interface. A simple force balance along the
fiber length gives
For e < ec, the fiber is pulled out and the interfacial shear strength is given by
ar
't = - (5.12)
2e
One measures the load required to debond as a function of the embedded fiber
length. Then, we can write
P = 2nre't (5.13)
and the interfacial shear strength, 't, can be calculated from the slope of the P
vs. Q plot. For e > ec, fiber failure rather than pullout occurs. The interfacial
shear strength is a function of the coefficient of friction, Jl, and any normal
compressive stress at the interface, i.e., 'tj = Jla... A common source of radial
compressive stress is the shrinkage of the matrix during cooling from the
processing temperature. The most doubtful assumption in this analysis is that
158 INTERFACE
Matrix
Pulling force
Fiber
(a)
I
Fiber --"+01.. 1----- Fiber fracture
pullout 1
Embedded length, 1 -
(b)
Fig. 5.12 (a) Experimental setup for fiber pullout test; (b) stress required to pull fiber
out vs. embedded fiber length.
The fabrication of the single fiber pullout test sample is often the most difficult
part; it entails embedding a part of the single fiber in the matrix. A variation of
this simple method is to embed both ends of the single fiber in the matrix
material, leaving the center region of the fiber uncovered.
The peak load corresponds to the initial debonding of the interface. This is
followed by frictional sliding at the interface, and finally by the fiber pullout
from the matrix, during which a steady decrease in the load with displacement
is observed. The steady decrease in the load is attributed to the decreasing area
of the interface as the fiber is pulled out. Thus, the test simulates the fiber
159 INTERFACE
pullout that may occur in the actual composite, and more importantly, provides
the bond strength and frictional stress values.
The effect of different Poisson contractions of fiber and matrix can lead to a
radial tensile stress at the interface. The radial tensile stress will no doubt aid
the fiber/matrix debonding process. The effect of Poisson contraction together
with the problem that the imposed shear stress is not constant along the
interface complicates the analysis of fiber pullout test. Fiber pullout tests have
been used to determine the interfacial shear strength in a variety of fiber
reinforced ceramic or glass matrix composites.
indenter into the sample. The schematic of the indenting mechanism of one
such instrument is shown in Fig. 5.13a while the indentation system and the
corresponding indenter movement and corresponding voltage output are shown
in Fig. 5.13b. The indenter is a triangular pyramid and its position is determined
(a)
Load Voltage
Moving Plate
Displacement Voltage
(b)
Fig. 5.13 (a) Schematic ofa nanoindentation machine; (b) indentation system and the
corresponding indent movement and voltage.
161 INTERFACE
In the pushout test, one pushes a fiber out, by means of an indenter, in a CMC
sample with the fibers aligned perpendicular to the viewing surface. Such a
fiber pushout test can give the frictional shear stress, 't, acting at the
fiber/matrix interface. The pushout test uses a thin specimen (1- 3 mm). In the
pushout test, a three-region curve is obtained. An example of such curve
obtained from a fiber pushout test done on a Nicalon fiber reinforced glass-
ceramic matrix composite is shown in Fig. 5.14. (Chawla et aI., 2001). In the
first region, the indenter is in the contact with the fiber and the fiber sliding
length ~ is less then the specimen thickness t. This is followed by a horizontal
region in which fiber sliding length is greater than or equal to the sample
thickness. In the third region, the indenter comes in contact with the matrix. In
the first region, we can determine the interfacial shear stress, 't from Eq. (5.18).
The specimen thickness should be much greater that the fiber diameter for this
relationship to be valid. In the horizontal region, the interfacial shear stress is
given by:
't = F/21t r t
where t is the specimen thickness. In the third region of Fig. 5.14, the value of
the interfacial shear stress cannot be determined because the indenter comes in
contact with the matrix.
It has been observed that the fiber will slide along the interface over a distance
162 INTERFACE
20
15
'.,.,"
n.
~ 10
~
Indenter hits
~ matrix
en
5
+-
0
-1 0 2 3 4 5 6
Displacement (u ml
Fig. 5.14 An example of a force vs. displacement curve obtainable from a pushout test
done on a Nicalon fiber reinforced glass-ceramic matrix composite. (Chawla et aI.,
2001).
that is dependent on the load applied by the indenter. In this model, the load on
the indenter is assumed to be balanced by the frictional stress at the interface,
and the effect of radial expansion during indentation is neglected. The fiber is
elastically compressed by the indenter load over the debonded length, which is
assumed to be dependent on the interfacial friction.
Some other variations on these techniques have been proposed. For example
Wang et al. (Wang et aI., 1992) used a dimpler, commonly used for specimen
preparation for transmission electron microscopy, to grind and polish a CMC
specimen to the desired thickness. Their reasoning for adopting this variation
was that in the conventional thin slice fiber pushout configuration, bending may
occur under indentation loading. Such bending of the specimen will result in an
inaccurate measurement of the interfacial strength. Their specimen preparation
technique results in thin, damage-free areas suitable for fiber pushout tests. A
nanoindenter instrument can then be used on such a sample to obtain the
163 INTERFACE
Load
interface shear strength. Another technique, called thin slice compression test,
involves taking a slice of an aligned fiber composite cut normally to the fibers,
put between two plates and subject it to a compression test. The top plate is of
a soft metal such as aluminum while the bottom plate is made of rigid materials,
for example, silicon nitride. The slice of composite is loaded to a desired peak
stress and then unloaded. At the top surface the elastic mismatch is maximum
and as the interface deb onding and sliding occurs, the top plate retains a
negative image of the fiber protrusion (Shafry et aI., 1989; Kagawa and Honda,
1991).
where cr'm is the stress in the matrix at which cracking begins, r is the fiber
radius, 't' is the shear stress at the interface, and Vr and V m are the volume
fractions of the fiber and the matrix, respectively. In terms of the composite
stress, cr e, the spacing X is given by
where Ee and Em are the elastic moduli of the composite and matrix,
respectively.
Matrix crack
Fig. 5.16 Multiple cracking in matrix with crack spacing betweenXand 2X.
test and obtain uniformly spaced, multiple matrix cracking, then we can
estimate the interfacial fiber/matrix shear strength from these two equations.
In practice, interfacial shear strength obtained from matrix crack spacing
measurements may not always agree with the experimental values obtained by
direct measurements such as a fiber pushout tests. This may be, in part, due to
any variation in various parameters such as processing conditions, interfacial
roughness, etc.
One may, however, question the validity of these equations because the ACK
model assumes no bonding at the interface. This is particularly true in
composite systems that show extensive chemical bonding between fiber and
matrix or if there is large degree of radial gripping of the fiber by the matrix
because of residual thermal stress or roughness induced clamping by the matrix.
Later, Aveston and Kelly (1973) modified the ACK model to include partially
bonded and fully bonded interfaces. The principle difference between the AK
model and the ACK model is that the shear stress along the interface is uniform
in the latter. The two models are in reasonable agreement except when
fiber/matrix bond shear strength is much greater than the matrix failure stress.
Gupta et al. (1990, 1992) devised a laser spallation technique to determine the
166 INTERFACE
tensile strength of a planar interface between a coating (thickness> 0.5 11m) and
a substrate. Figure 5.17 shows their experimental set up. A collimated laser
pulse impinges on a thin film sandwiched between the substrate and a confining
plate. This plate is made of fused quartz which is transparent to Nd:YAG laser
(wavelength = 1.06 11m). An aluminum film works as the laser-absorbing film.
Absorption of the laser energy in the gold or aluminum film causes a sudden
expansion of the film. The expansion of the constrained film produces a
compressive shock wave in the substrate and moves toward the
coating/substrate interface. When the compression pulse hits the interface, part
of it is transmitted into the coating. This compressive pulse is reflected into a
tensile pulse at the free surface of the coating. If this tensile pulse is of a
sufficient magnitude, it will remove the coating from the substrate. Gupta et al.
later used a laser Doppler displacement interferometer to record the time rate
of change displacement of the coating free surface as the compressive pulse is
reflected. By means of a sophisticated Tektronix SeD 1000 digitizer, they could
obtain a time resolution of about 0.5 ns for recording displacement fringes. This
information is then related to the stress pulse history at the interface. A direct
recording of the stress pulse makes their technique useful for interface systems
involving ductile components. We describe some results obtained by this
technique in chapter 9.
Substrate:
Energy Si crystal, pilCh-55
absorbing ribbon, etc. SiC
film (Au Or AI)
Surface
Wavelength = 1.06 J1ffi acceleral ions
Maximum energy = 850 mJ measured
onnal pul e duration = 8 n
Fig. 5.17 Schematic of the laser spallation technique (after Gupta et aI., 1990).
167 INTERFACE
REFERENCES
Angelini, P., W. Mader, and P.F. Becher (1987) in Mat. Res. Soc. Symp., vol. 78,
Materials Research Society, Warrendale, PA, p. 24l.
Aveston, J., G.A. Cooper, and A. Kelly (1971) in The Properties ofFibre Composites,
IPC Science and Technology Press, Survey, England, 15.
Aveston, J. and A. Kelly (1973) J Mater. Sci., 8, 129.
Bright, I.D., S. Danchaivijit, and D.K. Shetty (1991) J Amer. Cer. Soc., 74, 115.
Broutman, L.J. (1969) in Interfaces in Composites, ASTM STP #452,34.
Chamis, C.C. (1974) in Composite Materials, Vol. 6, Academic Press, New York, 32.
Chawla, N., K.K. Chawla, M. Koopman, B. Patel, C. Coffm, and J.I. Eldridge (2001)
Composite Sci. & Tech., 61, 1923.
Cook, I. and I.E. Gordon (1964) Proc. Roy. Soc. London, A228, 508.
Cranmer, D.C. (1991) in Ceramic and Metal Matrix Composites, Pergamon Press, New
York, p. 157.
Doerner, M.F. and W.D. Nix (1986) J Mater. Res., 1, 60l.
Ferber, M.K., A.A. Wereszczsak, L. Riester, R.A. Lowden, and K.K. Chawla, (1993)
Ceram. Eng. Sci. Proc., 13,168.
Gao, Y.C., Y.W. Mai, and B. Cotterell (1988) SAMPE, 39, 550.
Gupta, V., A.S. Argon, J.A. Cornie, and D.M. Parks (1990) Mater. Sci. Eng., A126,
105.
Gupta, V., A.S. Argon, J.A. Cornie, D.M. Parks (1992) J Mech. Phys. Solids, 4, 14l.
He, M.Y. and Hutchinson, J.W. (1989) J App. Mech., 56,270.
Iosipescu N. (1967) J Mater., 2, 537.
Kagawa, Y. and K. Honda (1991) Ceram. Eng. Sci. Proc., 12, 1127.
Kerans, R.J. and T.A. Parthasarathy (1991) J Amer. Cer. Soc., 74, 1585.
Maheshwari, A., K.K. Chawla, T.A. Michalske (1989) Mater. Sci. & Eng., AI07, 267.
Mandell, J.F., D.H. Grande, T.H. Tsiang, and F.I. McGarry (1986) in Composite
Materials: Testing & Design, ASTM STP 327, ASTM, Philadelphia, 87.
Mandell, J.F., K.C.C. Hong, and D.H. Grande (1987) Ceram. Eng. Sci. Proc. 8, 937.
Marshall, D.B. (1984) J Amer. Ceram. Soc., 67, C259.
Marshall, D.B. and W.C. Oliver (1987) J Amer. Ceram. Soc., 70, 542.
Marshall, D.B., M.C. Shaw, and W.L. Morris (1992) Acta Met., 40, 443.
Parthasarathy, T.A, N.J. Pagano, and R.I. Kerans (1989) Ceram. Eng. Sci. Proc., 10,
872.
Penn, L.S. and S.M. Lee (1989) J Compo Tech. & Res., 11,23.
Shafiy, N., D.G. Brandon, and M. Ferasaki (1989) Euro-Ceramics, 3,.453
Shetty, D.K. (1988)J Amer. Cer. Soc., 71, C-107.
Takaku, A. and A.G.C. Arridge (1973) J Phys. D: Appl. Phys., 6, 2038.
Wang, L. and R.J. Arsenault (1991) Met. Trans. A, 22A, 3013.
Wang, S.-W., A. Khan, R. Sands, and A.K. Vasudevan (1992) J of Materials Sci.
Letters, 11, 7739.
Weihs, T.P. and W.D. Nix (1991)J A mer. Ceram. Soc., 74, 524.
Cliapter6
Hooke's law relating a second rank stress tensor, O"ij' and a second rank strain
tensor, ekb can be written as
(6.1)
where C ijkl is a fourth-rank elastic stiffness tensor and the indices iJ,k take
values of 1,2, and 3. The stress and strain are second rank tensors; they have
32 = 9 components each. The stiffness tensor, Cijkl has 3 4 = 81 components. It
turns out that the stress, strain and the stiffness tensors are symmetric, i.e., O"ij
= O"jj, e kl = elk and Cjjkl = Cklij := Cjikl := Cijlk • These result, in the most general case,
in a reduction of the maximum number of independent elastic constants from
81 to 21. Equation (6.1) can be simplified and rewritten in a contracted notation
more appropriate for matrix operations as
(6.2)
Both Cij and Sij are 6 x 6 symmetric matrices, i.e., Cij = C ji and Sij = Sji' Also, the
stiffness and the compliance matrices are inverse of each other, i.e.,
We can write Hooke's law in an extended form for cubic and hexagonal
symmetry as follows:
Cubic
0'1 C II C I2 C I2 0 0 0 £1
0'2 C II C I2 0 0 0 £2
0'3 C II 0 0 0 £3
0'4 C44 0 0 £4
O's C44 0 £s
0'6 C44 £6
Hexagonal
0'1 C II C I2 C I3 0 0 0 £1
0'2 CII C I3 0 0 0 £2
0'3 C 33 0 0 0 £3
0'4 C44 0 0 £4
0'5 C44 0 £s
0'6 (C II -C I2 )/2 £6
171 MICROMECHANICS: ELASTIC, THERMAL, AND PHYSICAL PROPERTIES
In engineering usage, for an isotropic material, any two of the following four
engineering elastic constants will suffice; Young's modulus (E), Poisson's ratio
(v), shear modulus (G), and bulk modulus (K). This is so because of the
following relationships that exist among these four constants:
E = 3K(1 - 2v)
G = E/2(1 +v)
K = E/3(1 - 2v)
and v = (E/2G) - 1
Thus, one may measure all the four constants given above for an isotropic
material, but only two are independent.
Figure 6.1 shows cross sections of some fiber reinforced composites with fibers
arranged in a square (or rectangular), a hexagonal, or a random array. A
unidirectional fiber reinforced composite with fibers arranged in a hexagonal
or random manner, in the transverse section, is said to be transversely isotropic,
i.e., if we consider the fibers to be along direction 1, then there is no preferred
direction in the 2 - 3 plane. Such an arrangement of fibers gives us the same
number of independent elastic constants as the hexagonal crystal, i.e., one needs
five elastic constants to fully describe the elastic behavior of such a composite.
A square or rectangular arrangement of fibers in the transverse section, Fig. 6.1,
gives an orthotropic symmetry, i.e., three mutually perpendicular planes of
symmetry. Such an orthotropic symmetry requires nine independent elastic
172 MICROMECHANICS: ELASTIC, THERMAL, AND PHYSICAL PROPERTIES
--Ji-I:e I: ••
!r------.- i i •
•
•
I I
I I
. I eie
I I
I I
~b~
·0
(a)
ee
• e
••• e
•• e.• e (b)
•• e
• e
• e •
(c)
• •
Fig. 6.1 Various fiber arrays in a matrix: (a) rectangular (a*b) or square (a = b); (b)
hexagonal; (c) random. Hexagonal and random arrays are transversely isotropic.
.;.. x
Fig. 6.2 A unidirectional fiber reinforced composite loaded in tension parallel to the
fibers (isostrain or action in parallel).
(6.4)
where E is the strain, M is the change in length, eis the original length, and the
subscripts f, m, and cl indicate fiber, matrix, and composite in the longitudinal
direction, respectively.
For components behaving elastically, from Hooke's law we can write the
uniaxial stress acting on the fiber and the matrix as
where a is the stress, E is the Young's modulus, and the subscripts have the
meanings given above.
The applied load on the composite, Pc is partitioned between the fiber and the
matrix, i.e.,
Pc=Pr+Pm
or a el Ae = ar Ar + am Am
= Er Eel Ar + Em Eel Am
= (Er Ar + Em Am) Eel
ael = Eel Eel = (ErAIAe + Em Am/AJ Eel
(6.5)
(6.6)
174 MICROMECHANICS: ELASTIC, THERMAL, AND PHYSICAL PROPERTIES
where V is the fiber volume fraction. The variation of the longitudinal Young's
modulus, Eel (or Ell) as a function of fiber volume fraction, Vr, as per Eq. (6.5)
is shown in Fig. 6.3.
(6.7)
where the subscript ct denotes the composite in the transverse direction and the
other subscripts have the significance given earlier.
We can write for the total displacement, .M, of the composite in the thickness
direction
400
350
300
~
o;l
250 Ed
c.:>
'"
;:l
"3
200
""
0
::;s 150
E"
100
50
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Vr
Fig. 6.3 Variation of longitudinal (Eel) and transverse modulus (Eet ) with fiber volume
fraction (Vr)'
175 MICROMECHANICS: ELASTIC, THERMAL, AND PHYSICAL PROPERTIES
where ~1m and ~4 are the displacements in the matrix and fiber, respectively.
Dividing throughout by te, the gage length, we get the strain in the transverse
direction
or
(6.8)
(6.10)
In the case of longitudinal or principal shear modulus, both the fiber and matrix
are subjected to the same shear stress, Fig. 6.5. Shear strains in the matrix and
fiber are given by
176 MICROMECHANICS: ELASTIC, THERMAL, AND PHYSICAL PROPERTIES
r---- -----~----
L L_ _ _ "--------{
t
Fig. 6.5 A unidirectional fiber reinforced composite loaded in shear parallel to the
fibers.
177 MICROMECHANICS: ELASTIC, THERMAL, AND PHYSICAL PROPERTIES
d=yt (6.11)
where 'Y is the average shear of the composite and t is t9tal thickness of the
composite. We can write the total shear displacement in terms of the shear
displacements of the components as
or
(6.12)
(6.13)
Writing shear strains as shear stress divided by shear modulus in Eq. (6.13), we
get
or
(6.14)
In an anisotropic material, there are two Poisson's ratios. The principal or major
Poisson ratio (V I2 ) gives the transverse strain caused by an axially applied stress
while the minor Poisson ratio (V21 ) gives the axial strain caused by a
transversely applied stress. These two Poisson's ratios in an anisotropic material
are shown in Fig. 6.6. In an isotropic material, V 12 and V 21 are equal and we have
only one Poisson's ratio. In a transversely isotropic fibrous composite, V12 and
V 21 are not equal, but the two quantities are related: v 12 E22 = V 21 E II • The elastic
constants for a transversely isotropic composite in terms ofthe elastic constants
ofthe two components are summarized in Table 6.2 (Hash in and Rosen, 1964).
In this case, we have the 2-3 plane to be isotropic, i.e., the properties in
directions 2 and 3 are identical. We treat the matrix as an isotropic material and
treat the fiber as an anisotropic material. Thus, E and v are the two constants
required for the matrix while five constants (En, Ea , Gn , Ga , and Vf) are
required for the fiber. Oftentimes; composite structures are fabricated by
stacking thin sheets of unidirectional composites called plies in an appropriate
orientation sequence dictated by elasticity theory. It is of interest to know the
properties of a ply; for example, its elastic constants and strength. In particular,
it is useful to be able to predict the characteristics of a composite lamina if we
know the individual component characteristics and their geometrical
disposition. Later in the macromechanical analysis we treat a ply as a
homogeneous but thin orthotropic material. In the case of thin sheets or plies,
it is not unreasonable to ignore the elastic constants in the thickness direction,
179 MICROMECHANICS: ELASTIC, THERMAL, AND PHYSICAL PROPERTIES
leaving 4 independent elastic constants, viz., Ell> E 22 , V l2 and G 12 , i.e., one less
than the number for a thick but transversely isotropic material. The missing
constant is G 23 , the transverse shear modulus in the 2-3 plane normal to the fiber
axis.
-----Ld---- -
V12
t
IT]
Fig. 6.6 The two Poisson's ratios in an anisotropic material: the principal one (v 12) gives
the transverse strain caused by an axially applied stress and the secondary one (v21 )
gives the axial strain caused by a transversely applied stress.
180 MICROMECHANlCS: ELASTIC, THERMAL, AND PHYSICAL PROPERTIES
Table 6.2 Elastic moduli and stress-strain relationships for a transversely isotropic
fibrous composite (after Rashin and Rosen, 1964).
Elastic moduli
1
1 Cll-E '2
= -(--)
2 K23
4K23 V2
<p=I+--
E
Stress-Strain Relationshi s
Kp
Km+iG (V K + Vmm
3mpp
K )
K=
VK+VK+iG
p m m p 3 m
where K, G, and V indicate the bulk modulus, shear modulus, and volume
fraction, respectively, and the subscripts p and m refer to the particle and the
matrix, respectively. Hashin and Shtrikman (1963) and Rosen (1973) have also
analyzed macroscopically isotropic, particulate composite with an arbitrary
internal phase geometry and only phase volume fractions being specified. The
bounds of Hashin and Shtrikman on the shear modulus, G and the bulk
modulus, K are given below:
182 MICROMECHANICS: ELASTIC, THERMAL, AND PHYSICAL PROPERTIES
(1- Vp )
Kp+--------~3-V----
(K -K
m p
+ 3K
p
+~Gp )
where ~ > ~ and Gp > Gm • For K,11 < Kp and Gp < Gm the inequalities will be
reversed. Treating the particulate composite as an isotropic material, we can
obtain the bounds on the Young's modulus of the composite, Ee , by using the
following relationship:
9KG
E =
(3K +G)
For 0.5 < E/E m < 3, the bounds are close enough to give us a value within
~10% of the true modulus.
constants that are analogous to those due to Hashin and Rosen (1964) and
Rosen (1973). He put rigorous bounds on the longitudinal Young's modulus, E,
in terms of the bulk modulus in plane strain (~), Poisson's ratio (v), and shear
modulus (G) of the two phases without making any restrictions on fiber form
or packing geometry. The bulk modulus in plane strain, ~, is the modulus for
lateral dilation with zero longitudinal strain and is given by
~ = E/[2(1-2v)(1+V)]
We can readily verify from Eq. (6.15) that the deviations from the rule of
mixtures for Ell are quite small. If we substitute some values for practical
composites such as carbon or silicon carbide fibers in an alumina matrix we
shall find that the deviations in Ell from the rule of mixtures are less than 2%.
Notice that the deviation from the rule-of-mixtures value comes in from the (vf
- vrn)2 factor. Also, for Vf = vrn , we have Ell given precisely by the rule of
mixtures.
For the Poisson's ratio of a unidirectionally aligned fiber composite, Hill (1965)
also showed that
and
(6.16)
If Vf < Vrn and E f » Em' then, Vl2 will be less than that predicted by the rule of
mixtures (= vrYr+vrnV rn). It is easy to see that the bounds on V12 are not as close
as the ones on Ell. This is because (ve - vrn) appears in the case ofvl2 (Eq. 6.16)
while (ve - vrn)2 appears in the case of Ell (Eq. 6.15). If(vf -vrn) is very small, the
bounds will be close enough to allow us to write
(6.17)
on the results of Hashin and Rosen, 1964 and Hill, 1965) are given below:
where ~ and kr are the plane-strain bulk moduli of the matrix and fiber
respectively. The plain-strain bulk modulus is defined as:
1
k = K+-G
3
For most practical purposes the last term is negligible in the above expression.
In this case, the upper and lower bounds are not coincident.
Lower bound:
185 MICROMECHANICS: ELASTIC, THERMAL, AND PHYSICAL PROPERTIES
Upper bound:
where
Y+~m
a =--
y-l
~m = - - -
3 -4Vm
Halpin, Tsai, and Kardos (Halpin and Tsai, 1967; Kardos 1971; Halpin and
Kardos, 1976) devised an empirical approach to obtain some generalized
equations which readily give quite satisfactory results for unidirectional
composites compared to the more complicated micromechanical equations.
These equations contain adjustable fit parameters and work quite well with low
fiber volume fractions. They can also provide useful estimates of properties of
composites containing discontinuous fibers or whiskers oriented in the loading
direction. The adjustable parameters must be obtained from experimental data
or must conform to some analytical solution. One uses a single equation of the
form
(6.18)
186 MICROMECHANICS: ELASTIC, THERMAL, AND PHYSICAL PROPERTIES
Table 6.3 Bounds for transversely isotropic fiber reinforced composites (after Chamis,
1983).
Longitudinal modulus
Transverse modulus
Shear modulus
Shear modulus
Poisson's ratio
and
(6.19)
where p represents composite moduli, e.g., Ell, E22 , G 12 or G 23 ; Pr and Pm are the
corresponding matrix and fiber moduli, respectively; Vr is the fiber volume
fraction; and ~ is a measure of reinforcement which depends on boundary
conditions (fiber geometry, fiber distribution, and loading conditions). The term
~ is a fit parameter that is used to make Eq. 6.13 conform to the experimental
data. The function 1'\ in Eq. (6.19) is constructed in such a waytbatp=Pm when
V r = 0 and p = Pf when Vr = 1. Furthermore, the form of 1'\ is such that
187 MICROMECHANICS: ELASTIC, THERMAL, AND PHYSICAL PROPERTIES
and
These two extremes provide the bounds, not necessarily tight, on the composite
properties. Thus, values of ~ between 0 and 1 will give an expression for p
between these extremes. Some typical values of ~ are given in Table 6.4. For
example, for the transverse modulus, E22 , we can write from Eqs. (6.18) and
(6.19)
and
11 = (E/Em - 1)/(E/Em + ~)
where
11 = (P/Pm - 1)/(P/Pm + ~)
The maximum packing factor, Fmax is equal to 0.907 for hexagonal packing and
0.785 for square packing. A suggested value ofFmax for random packing is 0.82.
Modulus (
En 2(@/d)
E22 0.5
0 12 1.0
- 0 21 0.5
K 0
188 MICROMECHANICS: ELASTIC, THERMAL, AND PHYSICAL PROPERTIES
Thermal energy is ever present and is responsible for the atomic or molecular
vibration about a mean position in any material. As the temperature of the
materials is increased, the amplitude of thermal energy induced vibrations is
increased and the interatomic or intermolecular spacing increases, i.e., an
expansion of the body occurs. Most materials show such an expansion with
increasing temperature.
In general, the thermal expansion of a material is greater in the liquid state than
in the crystalline state, with the transition occurring at the melting point. In the
case of a glassy material, such a transition occurs at what is called the glass
transition temperature. Over a certain range of temperature, one can relate the
temperature interval and thermal strain by a coefficient, called the coefficient
of thermal expansion. In the case of a linear strain, the linear thermal expansion
coefficient, a, is a second-rank symmetric tensor, and is related to the strain
(6.20)
(6.21)
where V is the volume and T is the temperature. For small strains, it can be
easily shown that
189 MICROMECHANICS: ELASTIC, THERMAL, AND PHYSICAL PROPERTIES
(6.23)
As we said above, only over some specified range of temperature can the
coefficient of thermal expansion be treated as a constant. Consider a
temperature range ll.T over which a is a constant. Then we can write Eq. (6.20)
in an extended form as
61 a1
62 <lz
63 Cl3
ll.T (6.25)
64 Cl4
6s as
66 a6
Ifwe know the coefficient of thermal expansion along the fiber direction and
perpendicular to it, then we can find the coefficient of thermal expansion in any
arbitrary direction [hkl] as follows. Let the arbitrary direction [hkl] have
190 MICROMECHANICS: ELASTIC, THERMAL, AND PHYSICAL PROPERTIES
direction cosines n l, n 2, and n3, then we can write for the linear coefficient of
thermal expansion, (1hkl' in that direction as
(6.26)
(6.27a)
(6.27b)
where e is the angle between direction [hkl] and the fiber axis.
6.2.1 Expressions for Coefficients of Thermal Expansion of Composites
Various equations have been proposed for obtaining the coefficients thermal
expansion of a composite, knowing the material constants of the components
and their geometric arrangements (Turner 1946; Kerner 1956; Schapery 1968).
Different equations predict very different values of expansion coefficient for
a given composite. Almost all expressions, however, predict expansion
coefficient values different from that given by a simple rule of mixtures (= arVr
+ am Vm)' This is because these equations take into account the important fact
that the presence of a reinforcement, with an expansion coefficient different
from that of the matrix, introduces a mechanical constraint in the system matrix.
A fiber will cause a greater constraint on the matrix than a particle. One can
regard a particulate composite as a homogeneous material in a statistical sense,
191 MICROMECHANICS: ELASTIC, THERMAL, AND PHYSICAL PROPERTIES
i.e., assuming a uniform distribution of the particles in the matrix, Fig. 6.8. Let
us denote the volume fractions of the two phases making a particulate
composite by V 1 and V2 (= 1- V I)' Various researchers have derived bounds and
given expressions for the coefficients of expansion and other transport
properties such as thermal conductivity. Kerner (1956) developed the following
expression for the volumetric expansion coefficient of a composite consisting
of spherical particles dispersed in a matrix:
(6.28)
am VmKm + ap VpKp
(6.29)
VpKp + VmKm
where the symbols have the significance given above. Turner's expression,
generally, gives an expansion coefficient much lower than the rule-of-mixtures
value.
(a) the bonding between the fiber and matrix is perfect and mechanical in
nature, i.e., no chemical interaction is allowed.
(b) the fibers are continuous and perfectly aligned.
(c) the properties of the constituents do not change with temperature.
Schapery (1968) used energy methods to derive the following expressions for
expansion coefficient of a fibrous composite, assuming Poisson's ratios of the
components are not very different. The longitudinal coefficient of expansion for
the composite is
(6.30)
(6.31 )
and
(6.32)
where a c1 is given by Eq. (6.30) above. For low fiber volume fractions, V r < 0.2
193 MICROMECHANICS: ELASTIC, THERMAL, AND PHYSICAL PROPERTIES
Figure 6.9 shows the variation of acl and act for carbon fibers in a glass matrix.
The following numerical values for the properties of matrix (indicated by
subscript 1) and fibers (indicated by subscript 2) were used to used to construct
this figure:
Note the marked anisotropy in the coefficient of thermal expansion for this
system. This anisotropy is accentuated by the inherent anisotropy of expansion
in the carbon fiber. The reader should note that at Vr equal to 1, we have the
two values of the coefficient of expansion of carbon fiber itself. Alumina fiber
in a glass type matrix can have the situation where the fiber has a higher
coefficient of expansion than the matrix'. Polycrystalline alumina fiber is
isotropic in expansion. Figure 6.10 shows a plot of the longitudinal and
transverse coefficient of thermal expansion of the composite: alumina fiber (Ur
= 8.5 X 10-6 K- I , E f = 380 GPa) in a glass matrix (am = 4.75 x 10- 6 K- I , Em = 60
GPa).
7r-~~~~~~~~~-'
6
5
I
~ 4
to
I
o 3
t:J 2
a '" (6.34)
3
where act and act are given by Eqs. (6.30) and (6.31).
It should be pointed out the length of fiber as well as fiber orientation have an
effect on the coefficients of thermal expansion It would appear that the fiber
length has a more sensitive effect on the expansion characteristics because the
constraint on the matrix is highest when the fibers are continuous. An extreme
case of a low constraint is one of a matrix containing spherical particles, i.e.,
9.5
9.0
8.5
8.0
i 7.5
~
'"I 7.0
.....
0
~~
E-<
6.5
U
6.0
5.5
5.0
4.5
4.0
10.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Vr
Fig. 6.10 Variation of longitudinal and transverse coefficient of thermal expansion for
an alumina fiber/glass matrix composite. Alumina fiber is isotropic in expansion.
195 MICROMECHANICS: ELASTIC, THERMAL, AND PHYSICAL PROPERTIES
aspect ratio = 1. Continuous fibers with an aspect ratio of infinity represent the
other extreme in constraint. The effect of the fiber length of the thermal
expansion can be incorporated into Schapery's equation as (Marom and
Weinberg, 1975; Vaidya and Chawla, 1994)
k=-
e (e<e c' O<k<O.5)
2e c
ec
k=l--e (e>e c ,O<k<l)
2
where e is the length and Qc is the critical length of the fiber (for more about
critical fiber length, see chapter 7).
Chamis (1984) used a simple force balance to derive expressions for a. t and~.
Rosen and Hashin (1970) used a plane strain model to derive the following
expressions for a. t and a.2 :
196 MICROMECHANICS: ELASTIC, THERMAL, AND PHYSICAL PROPERTIES
where a:= acYr + amVm' K.: is the bulk modulus of the composite, Ee is the
elastic modulus of the composites, and k. is the plane strain transverse bulk
modulus of the composite.
The heat flow in a material is proportional to the temperature gradient, and the
constant of proportionality is called the thermal conductivity. Thus, in the most
general form, using indicial notation, we can write
where qj is the heat flux along X; axis, dT/dxj is the temperature gradient across
a surface that is perpendicular to the x.; axis, and kij is the thermal conductivity.
As should be evident from the two indexes, thermal conductivity is also a
second-rank tensor. Although, kjj not a symmetric tensor in the most general
case, it is a symmetric tensor for most crystal systems. For an isotropic material,
kij reduces to a scalar, i.e., a number, k. For an orthotropic material, we have
three constants along the three principal axes, viz., kll' k22' and k33 • For a
transversely isotropic material such as a unidirectionally reinforced fibrous
composite, there will two constants: thermal conductivity in the axial direction,
k.b and that in the transverse direction, keto The thermal conductivity in the axial
direction, k.1 can be predicted by a rule-of-mixtures type expression (Behrens
1968)
(6.35)
where krl is the thermal conductivity of the fiber in the axial direction and ~
is that of the isotropic matrix, and V r and Vm are the volume fractions of the
fiber and the matrix, respectively.
197 MICROMECHANICS: ELASTIC, THERMAL, AND PHYSICAL PROPERTIES
'~ 80~--~--~--~--~--~--~---r---r--~---'
'E 70
3_ 60
>.
."§; 50
g 40
"0
§ 30
u
_ 20
c
~ 100 r~~;:==~======~==:c:::C:::I:::i::~:::J
f=. t
o 0.1 0.2 0.3 0.40.50.60.7 0.80.9 1.0
Vf
Fig. 6.11 Longitudinal and transverse thermal conductivity as a function of fiber volume
fraction for a carbon fiber/glass matrix composite.
198 MICROMECHANICS: ELASTIC, THERMAL, AND PHYSICAL PROPERTIES
and
11 = [(kf2/~) -1] /[(kf2/~) + 1]
where we have taken l; equal to I in the Halpin-Tsai-Kardos equation.
We can find the thermal conductivity kx and ky in two arbitrary but mutually
perpendicular directions, x and y if we know the thermal conductivity in the
principal directions 1 and 2" respectively, by the following equations:
kx = k( cos2 e + k2 sin2 e
~ = k( sin2 e + k2 cos2 e
kxy = (k2 - k() sine cose
where e is the angle measured from the x-axis to the I-axis and k"y can be
considered to be a thermal coupling coefficient.
6.4 DENSITY
Density is one physical property for which the rule of mixtures works
absolutely and irrespective of the geometry of the phases involved. The total
mass of a composite is the sum total of the masses of the reinforcement and the
matrix, that is,
m c = m r +m m (6.37)
(6.38)
Dividing Eq. (6.37) by me and Eq. (6.38) by Vc and denoting the mass and
volume fractions by~, M m, and V" V m' V v, respectively, we can write
and
Or
mr mm
-+-+vv
P, Pm
200 MICROMECHANICS: ELASTIC, THERMAL, AND PHYSICAL PROPERTIES
Pc = - - - - -
M
_r
M v
+ --2!!. + _v
Mr Mm Vv
-+-+-
Pr Pm PcVc
Mr Mrn Vv (6.39)
-+-+-
Pr Pm Pc
We can use Eq. (6.39) to obtain indirectly the volume fraction of voids in a
composite. Rewriting Eq. (6.39), we obtain
(6.40)
we shall discuss one property, viz., dielectric constant. Dielectric materials are
used to increase the capacitance of electrical capacitors. They are also used as
electrical insulators. The dielectric constant of a material can vary with
temperature and frequency. The dielectric constant of a material, also known
as the relative permittivity, E" can be defined as the ratio of the permittivity of
the material (e) to that of vacuum (eo):
(6.41)
A
C = -s (6.42)
d
where A is the plate area, d is the plate spacing, and e is the permittivity of
vacuum.
Dielectric
Positive Plate
~~----{-U
+ +
Negative Plate
Fig. 6.12 Dielectric of relative pennittivty, sr' inserted in a parallel plate capacitor.
202 MICROMECHANICS: ELASTIC, THERMAL, AND PHYSICAL PROPERTIES
A
C = -B
d 0 Br = C0 Br
where Co is the capacity without the dielectric material in between the plates.
1 VI v2
= -+-
Bes BI B2
REFERENCES
SUGGESTED READING
Besides holding the fibers together, the matrix has the important function of
transmitting the applied load to the fibers. Many researchers (Cox, 1952;
McCartney, 1989; Gresczuk, 1969; Lawrence, 1972; Lee and Daniel, 1992)
have contributed to this subject of load transfer from the matrix to the fiber.
Let us focus our attention on a fiber embedded in a matrix. Figure 7 .1a shows
the situation prior to the application of an external load, where we have drawn
imaginary lines running through the fiber/matrix interface in a continuous
manner in the unstressed state. Let us load this composite axially. We assume
that no direct loading of the fibers occurs. Then, the fiber and the matrix will
experience locally different axial displacements because of the different elastic
moduli ofthe components. Under such circumstances, our imaginary vertical
lines in the unstressed state will become distorted as shown in Fig. 7.1 b. The
resultant displacement in the fiber, u, and in the matrix, v, are shown in Fig.
7.2. The larger the difference between the elastic modulus of the fiber and the
matrix, the larger will be the difference in their axial displacements. Different
axial displacements in the fiber and the matrix mean shear strains are produced
in the matrix on planes parallel to the fiber axis and in a direction parallel to
the fiber axis, i.e., displacements and stresses do not vary with angle about the
206 MECHANICAL BEHAVIOR
fiber axis. Thus, the transfer of the applied load to the fiber occurs by means
of these shear strains in the matrix. Frequently, this model has been referred
to in the literature as the shear-lag model.
(a) There exists a perfect bond between fiber and matrix, i.e., there is no
sliding between the two.
(b) Poisson's ratios of fiber and matrix are equal, which implies an absence of
transverse stresses when the load is applied along the fiber direction.
(c) All load transfer occurs via shear in the matrix and at the fiber/matrix
interface. This shear stress is assumed to be constant. There is no direct
loading of the fibers.
(d) Stress concentrations at the fiber ends are ignored.
207 MECHANICAL BEHAVIOR
I
" 2R
I
1
Fig. 7.2 A single fiber embedded in a matrix and the assembly subjected to an axial
strain, e, and the resultant displacements in the fiber (u) and the matrix (v).
Let the displacement of a point at a distance x from one extremity of the fiber
be u, in the presence of a fiber, and v, in the absence of a fiber, Fig. 7.2. The
load transfer from matrix to fiber occurs because of the difference in these two
axial displacements, u and v. For the transfer of load from matrix to fiber, we
can write the following expression:
dP/dx = B (u - v) (7.1)
where P is the load on the fiber, and B is a constant that depends on geometric
arrangement of fibers, matrix type, and moduli of fiber and matrix.
Differentiating Eq. (7.1), we get
The axial strain in the matrix away from the fiber is e = dv/dx while that in the
fiber, using the Hooke's law, is du/dx = PIEr Ar, where P is the axial load on
the fiber, Ar is the cross-sectional area of the fiber, and Er is the fiber modulus.
P = Er Ar e + S'cosh ~x + T'sinh ~x
O'r= P/A r = Ere + S cosh ~x + T sinh ~x
208 MECHANICAL BEHAVIOR
Of = 0 at x = 0 and x = e
This boundary condition follows from the fact that there is no direct loading
of the fiber from the two ends. The condition P = 0 at x = 0 gives
o = e Ef + S
S = -eEf
while the condition P = 0 at x = e gives
o = eEf+Scosh~Q+Tsinh~Q = eEf-eE(cosh~e+Tsinh~Q
T = e E [cosh ~ Q- 1] = e E tanh ~
f sinh~Q/2 f 2
o = E e [1 _cosh ~(Q/2 - x) ]
f f cosh ~(Q/2)
(7.4)
Q
for 0 < x < -
2
The maximum possible value of strain in the fiber is the imposed strain, e, and,
thus, the maximum stress is eEr' Thus, shear-lag analysis predicts that if the
fiber is long enough, then the stress in the fiber will increase from the two ends
to a maximum value, Of = Ere. Thus, the average stress in the fiber of length Q
is given by
209 MECHANICAL BEHAVIOR
- = +
~
fQ crdx = E e [1 -
I
tanh(~Q/2)
~e/2
(7.5)
o
We can obtain the variation of shear stress, 't, along the fiber/matrix interface
by considering the equilibrium of forces acting over an element dx of fiber,
radius, rr, and acting in the fiber direction, Fig. 7.3. We can write from Fig. 7.3
(7.6)
Now P, the tensile load on the fiber, is equal to nr/crr. Substituting this in Eq.
(7.6), we get
(7.7)
't =
EIrfe A
p •
~
sinh ( i-x)
2
(7.8)
2 cosh~
2
Figure 7.4 shows the variation of crr and 't with distance x along the fiber. The
maximum shear stress in Eq. (7.8) can be either the shear yield stress of the
matrix or the shear strength of the fiber/matrix interface; whichever is smaller
should be used in Eq. (7.8). In ceramic matrix composites, the shear strength
of fiber/matrix interface would be expected to be lower than the matrix failure
stress in shear, there being no realistic possibility of yielding in shear of the
ceramic matrix. Thus, in CMCs, one can safely take 'tmax to be the shear
strength of the fiber/matrix interface, 'ti.
L---L
p p + t.. p
<If---+--'
" L
.. dx - - - - - - 4.....1
Fig. 7.3 Equilibrium offorces acting over an element dx of fiber (radius, rr).
at a distance r from the axis as -r(r). Then, at the fiber surface (r = rf)' we can
write
dP
dx
B = _ 21trf -r(rf )
(7.9)
(u -v)
(7.10)
(7.11 )
For the shear stress in the matrix, we can write from Hooke's law, -r(r) = Gm'Y,
where G mis the matrix shear modulus and 'Y is shear strain given by
211 MECHANICAL BEHAVIOR
dw = "C(r) = "C(r)rj
y= dr G G r (7.12)
III m
(7.13)
But, by definition
T,(rj)rj
(7.15)
u-v
B = (7.16)
and from Eq. (7.13), one can obtain an expression for the load transfer
parameter, ~,
(7.17)
The term (Rlrf ) in this expression brings in the fiber packing factor. For a
square array of fibers, In (Rlrf) = 112 In (7tlVr), while for a hexagonal packing,
In (Rlrf ) = 112 In (21t!V3Vf). We can define the packing term, in a more general
way, in the following manner: In (Rlrf) = 112 In (<f>maxNf)' where <f>max is the
212 MECHANICAL BEHAVIOR
41tG m
The greater the value of the quantity (GmlEf ), the greater is the value of Band
the more rapid is the load increase in the fiber from either end.
It follows from the shear-lag analysis that in order to load a fiber to its
maximum strength, the shear strength given by Eq. (7.8) must be large. In a
metal matrix composite, the relevant shear strength is the matrix yield stress
in shear, because the metallic matrix will deform plastically in shear rather
easily in response to the high shear stresses developed. Of course, if the
fiber/matrix interface shear strength is less than the matrix yield stress in
shear, then the interface will fail first. We expect this to be the likely situation
in ceramic and polymer matrix composites, because in these brittle matrix
composites, frictional sliding at the interface is more likely than plastic flow
of the matrix. Therefore, in the case ofCMCs, the limiting shear stress will be
the interface strength in shear, "Cj • If the differential shrinkage between the
matrix and the fiber during processing or any other source of mechanical
interlocking at the interface results in a radial pressure p on the fibers, then "Cj
will be equal to fJP, where fJ is the coefficient of sliding friction between fiber
and matrix. Considering once again the equilibrium offorces over an element
#12 of the fiber
1td 2 Q
a-- 't.1td-
'j 4 I
2
or
Q aj
-
d 2't;
Here, we consider #12 and not #because the fiber is being loaded from both
ends. Given a sufficiently long fiber and a strong enough interface, it should
be possible to load the fiber to its breaking stress, afu, by means of load
transfer through the matrix. Let (#Id)c be the minimum fiber length to diameter
213 MECHANICAL BEHAVIOR
ratio necessary to accomplish this. We call the ratio (Q/d), the aspect ratio of
a fiber, and (lJd)e, the critical aspect ratio necessary to attain the fracture
strength stress of the fiber, (fji,. Thus, we can write
(7.18)
For a given fiber diameter, d, we can think of a critical fiber length, 1,., thus,
(7.19)
Over a length 1,12 from either end, the load in the fiber builds up. Strain builds
in a likewise manner. Beyond f/2 from the two ends, (i.e., in the middle
portion of fiber) the local displacements in matrix and fiber are the same, and
the fiber carries the major load while the matrix carries only a minor portion
of the applied load. It is instructive to decipher the implications ofEq. (7.19).
First of all, it tells us that the fiber length fmust be equal or greater than ec for
the fiber to be loaded to its maximum stress, (fji/. Secondly, the load transfer
length of a fiber in a matrix is independent of the total length of the fiber, i.e.,
the ratio, f/f, decreases as the fiber length increases, Fig. 7.5. An examination
of Fig. 7.5 shows that even for (fld) > (lJd)c the average stress in the fiber will
be less than the maximum stress to which it is loaded in its central region. In
fact, we can write for the average fiber stress as follows:
+fa;Jx
Q
af = = +[ap-ec)+Pa}c] = +[a}-ec(af-pa)]
o
or
af (1 -l.=f)
Q/Q
c
(7.20)
214 MECHANICAL BEHAVIOR
atrix
y._.f})-i._--
2rr ;
p p+t..P
(-j ~i_b~_<
:: ~'t :: \
:/J2 : :/J2: \
;..--.: ;..--.:
0:
Of max :
't :
O L
. ~ _ _--_~
Fig. 7.4 Variation of tensile stress, 0, and shear stress, 't, along the fiber/matrix
interface.
l-~- -~-~--~
Fig. 7.5 Load transfer as a function of fiber length. Fiber length !must be equal to or
greater than "" for the fiber to be loaded to its maximum stress, 0fu' ~ is the critical
length of fiber for load transfer.
215 MECHANICAL BEHAVIOR
where par is the average stress in the fiber over a portion ~/2 of its length at
both the ends. Thus, one may regard pas a load transfer function.
Flexure tests are commonly used to study the mechanical behavior of ceramics
at low and high temperatures. This is because the technique is easy to use and
inexpensive. However, the interpretation of flexure test data is not straight
forward. This is especially so in fiber reinforced ceramics, mainly because
they are highly anisotropic. These considerations have led to the development
of tensile testing of ceramics and ceramic matrix composites (Kandil and
Dyson, 1988; Fett et aI., 1988; Carroll and Tressler, 1988, 1989; Carroll et aI.,
1989; Wiederhorn and Hockey, 1991; Wiederhorn et aI., 1988; Davidge and
Briggs, 1989; Ohji, 1988; Liu and Brinkman, 1985; Seshadri and Chia, 1987;
Vaidyanathan et aI., 1988; Jenkins, et aI., 1990). Actually, as Davidge and
Briggs (1989) point out, flexure tests are easy to perform but difficult to
interpret while tensile tests are difficult to do but easy to interpret.
216 MECHANICAL BEHAVIOR
Bend test is one of the easiest to do and is very commonly used, especially in
the evaluation of brittle materials. There are two basic governing equations for
a simple beam elastically stressed in bending:
M
---
E
(7.21)
J R
and
M 0"
- - (7.22)
J Y
where M is the applied bending moment, I is the second moment of area of the
beam section about the neutral plane or the moment of inertia, E is the Young's
modulus of elasticity of the material, R is the radius of curvature of the bent
beam, and 0" is the tensile or compressive stress on a plane distance y from the
neutral plane.
For a beam of a uniform, circular cross section, the moment of inertia I of the
beam is given by
rrd 4
J=- (7.23)
64
bh 3
J=- (7.24)
12
Bending takes place in the direction of the depth, i.e., hand yare measured in
the same direction. Also, for a beam with symmetrical section with respect to
the neutral plane, replacing hl2 (or d/2) for yin Eq. (7.2) gives the stress at the
beam surface.
217 MECHANICAL BEHAVIOR
Figure 7.6 shows the elastic normal stress distribution through the thickness
of a bent beam. The stress and strain vary linearly with distance across the
section, with the neutral plane representing the zero level. The material on the
outside or above the neutral plane of the bent beam is stressed in tension while
that on the inside or below the neutral plane is stressed in compression. In the
elastic regime, the stress and strain are related by
cr=EE (7.25)
From equations (7.21), (7.22) and (7.25) we can obtain the following simple
relationship valid in the elastic regime,
(7.26)
We have given the expressions for three-point and four-point bend strength in
chapter 5, which we recall here. The maximum tensile stress in a three-point
bend test occurs on the outermost surface under the middle loading point and
is given by
r---1-----~--c--~--~
I
3PS
0" " ' - - (7.27)
where P is the applied load, S is the load span, b is the specimen width, and h
is the specimen height.
In such a test, the maximum shear stress, 't, occurs at the midplane and is given
by
3P
't " ' - - (7.28)
4bh
The ratio of the maximum tensile stress to shear stress is thus given by
't h
(7.29)
°" 2S
It is easy to see that the beam under test will fail in tension ifthe span to depth
ratio, Slh, is made very large. Typically, for S/h values greater than 20, failure
in tension will occur, while for Sih less than 5, failure in shear will result. The
test will also be invalid if shear and tensile failure occur simultaneously. It is
advisable to make an examination of the fracture surface after the test and
make sure that the fracture has taken place in the presumed failure mode. For
an elastic beam bent in four-point, the bending moment in the central position
is given by
20"max I EI
M= (7.30)
112 R
where I is the moment of inertia, tl2 is the distance from the neutral axis to the
outer surface, O"max is the maximum normal stress on a transverse section, and
R is the radius of curvature.
Figure 7.7 shows the variation in stress with specimen length for different
tests. Actually, what we plot Fig. 7.7 (a and b) is the variation of bending
moment, M, but for a given set of specimen dimensions, this corresponds to
the stress variatio~ as well. In a three-point bend test, Fig. 7.7a, the maximum
219 MECHANICAL BEHAVIOR
stress occurs along a plane in the center of the specimen span length. In a four-
point bend test, Fig. 7.7b, a constant stress acts over the inner span length of
the specimen. Four-point bending is also called pure bending since there are
no transverse shear stresses on the cross-sections of the beam between the two
loading points, i.e., the inner span length. In a tensile test, the whole gage
length of the specimen is subjected to a uniform stress, see Fig. 7.7c. We
discuss the tensile test situation below at some length because of the increasing
importance of tensile testing for ceramics.
t t f
I
t
t t t
·:~;'
.. ..
____ (f~
··· .. ..
(a)
(b)
-c-:-::=J-- .. ..
j r-----------~..
SJ----
.
(e)
Fig. 7.7 Variation in stress with specimen length: (a) three-point bend test; maximum
stress occurs along a plane in the center of the span length; (b) four-point bend test;
a constant stress acts over the inner span length; (c) tensile test; the whole gage length
of the specimen is subjected to a uniform stress.
220 MECHANICAL BEHAVIOR
Fiber pullout
1
Strain -->
(a)
300
Nextel 480/BN/giass
250
«I 200
Q.
::::E
i.. 150
tiS 100
50
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Apparent deflection, mm
(b)
Fig. 7.8 (a) A schematic stress-strain curve ofa continuous fiber reinforced ceramic
matrix composite in a three-point bend test done under displacement control. The
curve is linear elastic up to <Jo when the first matrix crack appears. From <Jo onward,
the curve becomes nonlinear until the ultimate stress, <Ju is reached. At <Ju, fiber
fracture occurs. Straining past this point results in fiber pullout. (b) A three-point bend
test curve for a BN coated Nextel 480 fiber in a glass matrix. For the sake of
comparison, a similar curve of the uncoated Nextel 480 fiber/glass matrix is also
shown. Note the noncatastrophic failure mode in the case of BN coated fiber
composite. There is also an increase in <Jo of the composite vis a vis that of the
unreinforced matrix.
222 MECHANICAL BEHAVIOR
Tensile tests are important for evaluating inherent flaw distributions and the
resultant strength distributions (Davidge and Briggs, 1989; Dyson et aI., 1989;
Ohji, 1988; Liu and Brinkman, 1985). One major problem in tensile testing
of ceramic and ceramic composites is the bending moments that can be easily
introduced during specimen alignment, gripping, and testing. Such accidental
bending moments are referred to as parasitic bending moments. Various grip
designs have been developed (Ohji, 1988; Liu and Brinkman, 1985) to take
care ofthis problem of parasitic bending moments. One such development, due
to Liu and Brinkman (1985), involves a hydraulic, self-aligning grip system
that results in a condition of near-zero bending moment during a tensile test.
The specimen grip is self-aligning and has a built-in flex coupling consisting
of eight tiny hydraulic pistons. These pistons reduce bending in the specimen
to near-zero, less that 1% bending in the gage length. With proper attention to
the gripping of sample, tensile tests on ceramics can be performed at
temperatures up to 1300 °C without much problem (Ohji,1988).
If the fiber and the matrix have the same failure strain, then we can write for
the ultimate tensile strength of a composite (creu) as
where crfu and crmu are the ultimate tensile strengths of the fiber and the matrix,
respectively. This condition of the fiber and the matrix having an identical
223 MECHANICAL BERAVIOR
where or' is the stress on the fiber corresponding to the failure strain of the
matrix. The condition of equality in the above expression serves to define a
critical fiber volume fraction, V erit • For fiber volume fractions greater than Verit,
a series of cracks will form in the matrix and it will fracture into thin slices.
Assuming that the fiber strain to failure is greater than that of the matrix, a
schematic tensile stress-strain curve of a unidirectional Nicalon fiber
reinforced SiC composite is shown in Fig. 7.10. The curve is linear elastic up
to 0 0 when the first matrix crack appears. On continued loading, additional
load is thrown on to the fibers. Assuming that the fibers can withstand this
additional load and that fiber/matrix interface remains intact, then additional
load will be transferred back to the matrix as per shear lag analysis described
above. This will result in, more or less, parallel cracks in the matrix, which
will be perpendicular to the tensile axis. Eventually, the specimen will be
traversed by parallel cracks, perpendicular to the fiber axis, and distance x and
224 MECHANICAL BEHAVIOR
Single fracture
Efu Emu
Strain - v,-
(a)
elu>emu
Uiu Single: Multiple matrix
fracturel fracture elu
I
u;
T T
- - -- --
gj
--
CIl
CIl
--
~
--
~
CI) CI)
----\
emu u",=u;V,+umu (1-V,)
When a crack in CMC is small in length, the stress required for its growth
depends on its length. When the crack becomes sufficiently long and when it
is bridged by fibers, we obtain a state of steady state cracking. Under this
condition the crack grows at a constant stress. At some distance away from the
leading edge of the crack, an equilibrium crack opening develops and the load
is fully supported by the fibers. The stress intensity associated with matrix
cracking then becomes independent of crack length. According to Marshall et
aI. (1985), a steady state cracking will exist when the crack length exceeds a
critical value, called the characteristic crack length. A CMC showing such
steady state cracking will have very desirable damage tolerant characteristics
and significant nonlinearity before reaching the maximum stress value.
We have repeatedly said that a mere increase in elastic modulus and strength
is generally not the objective in ceramic matrix composites because the
monolithic ceramics have very high strength. However, such an increase in
elastic modulus and strength can be a welcome attribute for low modulus
matrix materials, for example, glasses, glass-ceramics, and some crystalline
ceramics such as MgO, mullite, etc. The major driving force, however, for the
use of continuous fibers in ceramic or glass matrix materials is to toughen
them. This idea of using continuous fibers in a brittle matrix to obtain
enhanced toughness can be traced to the 1960s (Bowen, 1968; Evans, et aI.,
1961; Tattersall and Tappin, 1966). Various researchers have used different
measures of toughness. One convenient term in this regard is the work of
fracture. Work of fracture can be defined as the amount of energy required to
produce a unit area of fracture. The concept of work of fracture can be used
as a qualitative measure of toughness. It should be pointed out, however, that
the concept of work of fracture has value only as a comparative tool; it is not
a design parameter.
Bowen (1968) has cited a 1961 report by Evans et aI. (1961) wherein it was
pointed out that a practical way to obtain an enhanced toughness, i.e., an
increased work offracture in ceramics, may be via obtaining multiple cracking
226 MECHANICAL BEHAVIOR
such that no single crack could propagate unhindered and cause failure.
Tattersall and Tappin (1966) analyzed the work of fracture and observed that,
for a wide variety of materials, the work of fracture increased as the toughness
increased.
Various glass matrix compositions reinforced with carbon fibers have been
shown to have strength and modulus increase with fiber volume fraction, in
accord with the rule-of-mixtures type behavior (Prewo et aI., 1986; Dawson
et aI., 1987). An example of such a linear increase in the strength of the
composite as a function of fiber volume fraction was observed by Dawson et
al. (1987) in Nicalon fiber reinforced borosilicate glass, Fig. 7.11. A linear
increase in strength of unidirectionally reinforced carbon fiber/glass
composites with fiber volume fraction Ve up to about 55% was observed
(Phillips et aI., 1972), Fig. 7.12. Beyond 55% Vf> increased matrix porosity led
to a drop in strength. The Young's modulus also increased linearly with Ve,
Fig. 7.13, but at higher Vr it deviated from linearity owing to matrix porosity
and possible fiber misalignment (Phillips et aI., 1972). The dashed line in Fig.
7.13 corresponds to the rule-of-mixtures prediction. The experimental values
are consistently below the rule-of-mixtures prediction line because of the
porosity in the glass matrix. Phillips et al. (1972) made a correction for the
effect of matrix porosity on its Young's modulus by using the following
expression
where Em is the modulus of the matrix with pores, Emo is the modulus of the
matrix with zero porosity, and P is the porosity in the matrix. Application of
this correction to the measured modulus values gives a better agreement with
the rule-of-mixtures prediction.
300r---------------------~--~~--~
250
~200
~
~ 150
(jj
100
50
3000
~
./
0 ./
~ 2000
..
./
./
./
... .-
=5 1500 ./
01 I'
c ./
Q)
L 1000
-+--'
(f) .~.,..
500 ./
./
./
B.o 1.0
Fig. 7.11 Linear increase in strength of the composite as a function of fiber volume
fraction in Nicalon fiber-reinforced borosilicate glass (after Dawson et aI., 1987).
229 MECHANICAL BEHAVIOR
60 80 100
Lee and Daniel (1992) used a modified shear lag analysis to predict stress-
strain and stress vs. matrix crack density curves and obtained a reasonable
agreement with experimental results for silicon carbide/calcium aluminosilcate
glass-ceramic composite subjected to monotonic loading.
Rule of Mixtures
'\v/
300 -
.
/
/
/-'0
p
200 -
/.
/-6
If •
/~. • Measured
100 - /
//
o Calculated,
Corrected for
Matrix Porosity
I I I
o 10 20 30 40 50 60 70 80 90 100
Vr. %
Fig. 7.13 Linear increase in Young's modulus with percentage fiber volume in a
unidirectional, continuous carbon fiber/glass composite (after Davidge and Briggs,
1989).
that the stability of SiC whiskers as well as fracture toughness depended on the
amount of additives used in the processing.
7.3 FATIGUE
0
a... 1000
~
I
I- 800
(9
z
W
0:
I- 600
(j)
W C 10 vol %
0:
=>
l-
400
e. 5 vol % I ±~~
u I i l
c::r
0:
LL
200 ---""'1
0 400 800 1200
TEST TEMPERATURE ( 0 C)
Fig. 7.14 Strength increase in silicon carbide whisker/alumina composites as a
function of the whisker volume fraction and test temperature.
turbine to withstand over 30,000 cycles offatigue (low cycle fatigue) (Helms
and Haley, 1989). Yet another problem in this regard is that the application of
conventional approaches to analyze the fatigue behavior of composites, for
example, the stress vs. cycles (S-N) curves or the application oflinear elastic
fracture mechanics (LEFM), is not straightforward. This problem is, of course,
common to all composites. The main reasons for this are the inherent
heterogeneity and anisotropic nature of the composites, especially in fiber
reinforced composites. These inherent characteristics lead to damage
mechanisms in composites that are very different from those encountered in
conventional, homogeneous, or monolithic materials.
The fatigue behavior of a material can be conveniently divided into two stages:
crack initiation and propagation. Quite frequently, a rule-of-thumb approach
is used, viz., increase the monotonic strength of the material which
concomitantly results in an increase in its cyclic strength. This rule-of-thumb
assumes that the ratio of fatigue strength/tensile strength is fairly constant. It
is generally true that the maximum efficiency in terms of stiffness and strength
gains in fiber reinforced composites occurs when the fibers are continuous,
uniaxially aligned and the properties are measured parallel to the fiber
direction. As we go off-angle, the role ofthe matrix becomes more important.
One major drawback of studies offatigue behavior of a material using this S-N
233 MECHANICAL BEHAVIOR
~
~
I-
~
lIE
c.
::l L
.
\oJ
II:
I-
Fig. 7.15 S-N curve from conventional fatigue testing. S is the stress amplitude and
N is the number of cycles to failure. Curve A has an endurance or fatigue limit SL
while curve B does not.
approach is that no distinction can be made between the crack initiation phase
and the crack propagation phase.
where da/dN is the crack growth per cycle and ~K is the alternating stress
intensity factor.
CTmOK
U)
U)
I.&J
...a:
U) 0 TIME
O"min
Crack growth rate, da/dN, is related to the cyclic stress intensity factor range,
L\K, according to the power law relationship (Paris, Gomez, and Anderson,
1961; Paris and Erdogan, 1963):
daldN = A(L\K)m
where A and m depend on the material and test conditions. The applied cyclic
stress intensity range is given by
L\K = YL\aJ"a
where Y is a geometric factor, L\a is the cyclic stress range, and a is the crack
length.
The major problem in this kind of test is to make sure that there is one and
only one dominant crack that is propagating. It was mentioned above that the
complexities in composites lead to the presence of many modes of damage,
such as matrix cracking, fiber/matrix debonding, crack deflection along the
interface leading to multidirectional cracking, ply delamination in the case of
a laminated composite, fiber fracture, etc. Some of these modes appear rather
early in the fatigue life of composites. It thus becomes important to have a tool
to characterize such a complex state of damage. Progressive loss of stiffness
has been found to be a very important and useful characteristic that can be
used to study the fatigue of polymeric laminate composites. Schematics of the
different types of damage zones formed in an isotropic material (e.g., a metal,
or a ceramic, or a polymer) and a fiber reinforced composite, which is an
anisotropic material, are shown in Figs. 7.17 and 7.IS, respectively. In the case
of an isotropic material, a single crack propagates in a direction perpendicular
to the cyclic loading axis (mode I loading), see Fig. 7.17. In a fiber reinforced
composite, on the other hand, a variety of subcritical damage mechanisms lead
to a highly diffuse damage zone as shown in Fig. 7.1S (a,b,c).
in the composite are shown in Fig. 7.19. We come back to this problem later
in this section. First we describe some results obtained with CMCs using
conventional approaches to fatigue.
Prewo and Brennan (1980) and Brennan and Prewo (1982) studied the tensile
fatigue behavior of Nicalon-type silicon carbide fiber reinforced lithium
aluminosilicate (LAS) glass-ceramic composite. They used two different types
::::
.... ::::
.:.: ~:~:
:.:. :::: ::::
:::: =:!: ;~~~
::~: ::::
:.:.
.:.: :jj
~~~
~ 1~ ~ ~
~~ m~ -0:
Fig. 7.18 Diffuse damage zone in a fiber reinforced composite (anisotropic): (a) fiber
breakage and local debonding, (b) matrix cracking, deflection of the principal crack
along a weak fiber/matrix interface and (c) deflection of the principal crack.
236 MECHANICAL BEHAVIOR
Q)
.~ Composites
til
~
~ Metals
Q ({f---In-iti-ati-on---+--p~ropagation
Initial
imperfections
Life
(cycles or time)
glass at the whisker tips (white regions marked by arrows) in the SiCjAlP3
composite subjected to a constant load fracture at 1400 °C in air. More
information about such oxidation of SiC whiskers is provided in the next
section on creep. What we wish to emphasize here is that there is a
difference in deformation and failure mechanisms under static and cyclic
loadings. When the composite shown in Fig. 7.21 was subjected to cyclic
loading, such large scale oxidation of whiskers was not observed. Instead,
cyclic loading resulted in multiple cracking of whiskers and the formation
of glass at these crack sites, see Fig. 7.22. A careful examination of Fig. 7.22
shows the meniscus of the molten glassy phase that has flowed inside the
cracked whisker.
Fig. 7.21 Formation of glass pockets at the whisker tips (white regions marked by
arrows) in the SiCjAlz03 composite subjected to a constant load fracture at 1400
°C in air. (Courtesy of S. Suresh).
Fig. 7.22 When the composite shown in Fig. 7.21 was subjected to cyclic loading,
large scale oxidation of whiskers was not observed. Instead, multiple cracking of
whiskers and the formation of glass at these crack sites occurred. Note the meniscus
of the molten glassy phase that has flowed inside the crackedwhisker. (Courtesy of
S. Suresh).
241 MECHANICAL BEHAVIOR
Nt (cycles to failure)
7.4 CREEP
'i!E.'"
1'01
....
::
IS
.,.,
0.9
"
~
tll
0.8 O!---!-.50~--:'1OO!:-:---:1~50~~2~OO:::--"::"25~O~~300
N (cycles)
Fig. 7.24 Schematic of the progression of cyclic fatigue damage in under two
different interfacial conditions (after Ramakrishnan and layaraman,1993).
the matrix will decrease and that in the fiber will increase until the creep
rates of the two become equal.
The basic equation given above describing the creep behavior is independent
of the sign of the applied stress, i.e., it is immaterial whether tensile and
compressive stress is applied. Wiederhorn and Hockey (1991) have analyzed
the creep behavior ofCMCs, both particle and whisker reinforced. In two-
phase ceramics, creep rate in tension is faster than in compression for
identical stress and temperature conditions. At first sight, this would appear
to be attributable to the ease of cavitation and microcracking during tension
than in compression because tensile stresses assist cavitation while
compressive stresses tend to close the cavities and microcracks. However,
the results of Wiederhorn and Hockey's (1992) on creep of siliconized
silicon carbide in tension and compression at 1300°C were unusual. Their
siliconized SiC composite consisted of SiC (about 67% by volume) in a
matrix of silicon. It showed a highly asymmetric creep behavior between
tension and compression, much more creep in tension than in compression.
As we said above, such asymmetry in creep behavior has been observed in
many two-phase ceramics. In single phase ceramics the asymmetry vanishes.
The interesting observation of Wiederorn and Hockey was that in their
siliconized SiC composite cavitation did not occur at low tensile stresses so
the cavitation alone could not explain the creep asymmetry in their material.
They used a model from the area of soil mechanics to explain their results.
As per this model, sand with 38% voids shows dilatant behavior (expands
by about 3%) during the initial stages of axial strain. Their siliconized SiC
had 33% silicon by volume, which Wiederhom and Hockey treated as voids.
In short, silicon flows but the movement of packets of SiC grains determines
the creep rate; such movement of packets of grains is more difficult in
compression than in tension.
creep curve at 1400 °C for the composites shows a marked change in the n
value at a stress level of about 125 MPa indicating a change in the rate
controlling process. A stress exponent value of about 2 is generally thought
to be due to grain boundary sliding. The higher stress exponent and the
higher creep rate at 1400 °C was attributed by the authors to extensive
cavitation.
10- 5
1200°C
10- 6
I
(J)
~ 1300°C
10- 7
4-Point
Flexure in air
10- 9
50 100 200 300 400
.a(MPa)
Fig. 7.25 Creep strain rate as a function of stress at 1200, 1300, and 1400 °C. For
tests at 1200 and 1300 °C, creep resistance is improved due to a retardation of grain
boundary sliding by SiCw present at the grain boundaries. The creep curve at 1400
°C for the composites shows a marked change in n at about 125 MPa indicating a
change in the rate controlling process (after Lin and Becher, 1990).
246 MECHANICAL BEHAVIOR
size, presence of glassy phases, etc. (Routbort et aI, 1990; Lin and Becher,
1990; Pysher et aI., 1989; Bender et aI., 1991). In the case of creep of
ceramic matrix composites, one needs to consider the intrinsic creep
resistance of the fiber, matrix, and interface region. We have described the
creep characteristics of various ceramic fibers in chapter 3. Oxide fibers are
fine-grained and generally contain some glassy phase. Nonoxide fibers are
also fine grained, multiphasic (with some glassy phase) and susceptible to
oxidation (Lessing and Gordon, 1977). Nonoxide fiber/nonoxide matrix
composites, such as SiC fiber/SiC and SiC fiber/Si3N4' generally show good
low temperature strength, but oxidation resistance is the major limitation.
Strength ofNicalon type SiC fiber is very sensitive to temperature above
1200 °C and its environment (Mah et aI., 1984). Nonoxide fiber/oxide matrix
composites or oxide fiber/nonoxide matrix composites, such as carbon/glass,
SiC/glass, SiC/alumina and AI20/SiC, generally do not possess high
oxidation resistance because the permeability constant for the diffusion of
oxygen is high, resulting in rapid oxygen permeation through the oxide
matrix (Mah et aI., 1984; Okamura, 1987; Hermes and Kerans, 1989). Prewo
et aI. (1986) observed that the glass matrix did not prevent the degradation
of carbon fiber caused by oxidation. When hot pressed SCS-6 fiber/Si3N4
matrix composite was subjected to creep, a complex glassy layer (Si02 -
Y 20 3-MgO) formed on the surface (Holmes, 1991). Figure 7.26 shows an
example of glass layer on sample subjected to creep at 1350 °C and at 110
MPa. The glass layer formed by oxidation of silicon nitride and sintering
aids. Holmes also observed that the extent of fiber pullout decreased as the
applied stress increased in a creep test. Rather pronounced separation along
the fiber/matrix interface was observed after low stress (70 MPa) creep.
Figure 7.27 shows an example of such debonding. Severe degradation at
temperatures above 1000 °C has been observed in many nonoxide systems,
for example, in a CVI-processed silicon carbide fiber reinforced SiC matrix
due to the evolution of CO, SiO, and other gases (Lin and Yang, 1991).
Fig. 7.26 FOlmation ofa glass layer on sample of SiCI Si3N4 composite to subjected
to creep at 1350 °C and at 110 MPa. The glass layer forms by oxidation of silicon
nitride and sintering aids (courtesy of lW. Holmes).
7.27 Separation along the SCS-6 fiberlSi 3N4 matrix interface after low stress (70
MPa) creep (Courtesy of J.W, Holmes).
248 MECHANICAL BEHAVIOR
to that of the cross-plied composite because the fibers in the 90° direction
contribute less to creep resistance than the fibers in the 0° direction.
-2.0
g m
-2.6
-3.2
·W o
bIl -3.8 ~o
.3 0 o
[0/90J
o
-4.4 o 00 [OJ
o
-5.0
-s·hs 2.75
Log 0'
Fig. 7.28 Steady state creep rate of monolithic silicon nitride matrix, [0] and [0/90]
cross-plied SCS-6 fiber/Si3N4 composites as a function of applied stress at 1200 °C
(Yang and Chen, 1992).Creep resistance of the composite is superior to that of the
monolithic silicon nitride, while the creep resistance ofthe unidirectional composite
is superior to that of the cross-plied composite.
249 MECHANICAL BEHAVIOR
10-3
10-4
.....
-
I
UJ 1Q"5
ai
e! 1Q-6
0-
Q)
~ 10-7
U
10-8
10-9
10 100
Stress, MPa
Fig. 7.29 Creep resistance of alumina and mullite can be improved significantly by
incorporation of SiC whiskers.
250 MECHANICAL BEHAVIOR
however, are limited by plastic flow under creep conditions. It has been
estimated that there are at least 38 oxides with their melting temperature
above 2000 °C (Routbort et aI., 1990). Commercially available oxide fibers
include polycrystalline alumina, mullite, and single crystal filament of
sapphire. Polycrystalline fibers show large strength reduction at 1200 °C and
Fig. 7.31 The same sample as in Fig. 7.30 after etching. The etchant removed the
oxidized part of the whisker, leaving the remnants of SiCw with a reduced cross
section and irregular shape. (Courtesy of A. Wereszczak).
252 MECHANICAL BEHAVIOR
almost a complete loss of strength at 1300 °C. Single crystal oxide fibers
such as alumina and YAG can withstand temperatures above 1300 °C.
Most ceramic oxides have rather poor creep resistance. This susceptibility
of oxides to creep can be limiting factor on the lifetime of oxide/oxide
composites at high temperatures. One can estimate the allowable creep rate
for an application by using a Monkman-Grant type curve (Parthasarathy et
aI., 1991), in which we plot the logarithm of creep rate as a function ofthe
logarithm of rupture time. If we assume a failure strain of 1% for ceramics,
then we can estimate the allowable creep rate for a given lifetime. Let us say
that we wish to have creep rates of 10- 8 S-1 or less for extended service
applications. Figure 7.32 shows creep data for some oxide matrix materials
(Warshaw and Norton, 1962; Penty and Hasselman, 1972; Evans, 1970;
Seltzer and Talty, 1975; Lessing et aI., 1975; Pysher and Tressler, 1992).
All data are plotted for an applied stress of 100 MPa. For a creep rate of
1O- 8s- 1 at a temperature above 1400 °C, it would appear that there is no
oxide matrix material capable of providing adequate creep strength. To
achieve an acceptable creep rate for extended service applications, a good
creep resistant oxide fiber would be required to reinforce a less creep
resistant matrix. Figure 7.33 shows the creep resistance data for some oxide
fibers (Jakus and TUlluri, 1989; Johnson et aI., 1987; Garboriaud, 1981;
Firestone and Heuer, 1976; Corman, 1991; Bullock et aI., 1977). Note that
polycrystalline oxide fibers possess limited creep resistance, while single
crystal fibers, sapphire and yttria-alumina-gamet (YAG), can maintain good
creep resistance above 1400 °C. The creep performance of a composite
system may be estimated as follows (Monkman and Grant, 1956). Assuming
the creep of fiber and matrix phases can be described by a common power
law expression, we can write
where cr denotes the stress, c denotes the strain, cr fo ' cfo' crmo ' and Cmo are
constants obtained from empirical stress-strain rate relationships, and
subscripts, f and m refer to the fiber and matrix, respectively. The strain rate,
cc' in the composite at a stress, crc, is weighted according to the volume
fraction of the two phases, V f and V m' assuming a fully bonded interface and
both fiber and matrix being creep at the same rate,
253 MECHANICAL BEHAVIOR
1Q-3
YAG.3"",
10-4
AI2O,.7"",
/
~ 10-5
1i ~ Mullite. 1.2 "'" ~
e!
~
f1O-S
()
10-7 /
5~
Y-PSZ. 41 "'"
MUlllt••
10-"
4.5 5.0 5.5 6.0 6.'5
Temperature.104/K
Fig. 7.32 Creep rate as function of inverse temperature for some oxide matrix
materials for an applied stress of 100 MPa. For a creep rate of 1O- 8/s at a
temperature above 1400 °C, it would appear that there is no oxide matrix material
capable of providing adequate creep strength.
It is worth pointing out that properties of a single crystal are a function of its
orientation. This is true of single crystal fiber as well. Figure 7.34 compares
the creep rate as a function of inverse temperature for Y AG and alumina
single crystal fibers. Note that the YAG single crystal exhibits creep rate
dependence on crystal orientation than alumina single crystal. Rather large
differences occur in the flow stress for the three slip systems of an alumina
single crystal. In particular, an off-axis loading can significantly affect the
fiber and composite properties. It can be seen from Fig. 7.34, shown above,
that the flow stress of such a material at 00 orientation can be more than 10
times greater than that at an off-axis orientation at high temperature. As we
pointed out above, Y AG, a cubic compound, is less orientation dependent
than alumina.
254 MECHANICAL BEHAVIOR
Case 1
Fiber: YAG Single crystal, f= 116 (110)*
Matrix: Alumina Polycrystalline, m = 111.6 *
Stress = 34 MPa
Case 2
10-5
.-
m 10-6
.sf
~
C.
Q) 10-7
~
()
I
10-9
Temperature. °C
1800 1500 1250 1000
\ Sapphire (42°)
\
YAG (100) \
\
\
Sapphire (T 0001) \
I' \
\
\
VAG (110)
Fig. 7.34 A comparison of creep rate as a function of inverse temperature for YAG
and alumina single crystal fibers. Note that the YAG single crystal exhibits a lower
creep rate dependence on crystal orientation than alumina single crystal.
3r---~----~--~~--~----r----'
T = 1773K in air
0=140MPa
0=150 MPa
o 234 5 6
Time. S (10 4 )
Fig. 7.35 Tensile creep curves at 1773 K under different stresses for AI20/Y3AlsOI2
eutectic composites (after Hirano et aI., 2001). Note primary and secondary stages.
10.2 r--...-_..........,....,......,r---...-..............................
AbOJIYAG Sintered Composites
10') (polycrystalline) ....
Compression in argon ....... ".04 Compression
1873K ............. ' in argon
•• .&: ..........
Tension
1(t6
,~JI
10.1
Fig. 7.36 Steady state creep curves for the single and polycrystal samples of
Al203N3AlsOl2 eutectic in argon showed a power-law relationship (After Hirano
et aI., 2001). The creep resistance is better in compression than in tension and for
the single crystal than for the polycrystaI. (After Hirano et aI., 2001)
257 MECHANICAL BEHAVIOR
tensile creep rates are faster than in compression, although the environments
are also different. Note also that the single crystal eutectic composite has
better creep resistance than the polycrystalline eutectic. Finally, we present
a comparison of tensile creep rate at 1200 °C for 100 h of some CMCs with
a nickel-based oxide-dispersioned superallOY, MA 754, see Fig. 7.37
(Holmes and Wu, 1994). The creep rate ofCMCs is lower than that ofMA
754. But does this mean that the CMCs can replace the superalloy in ajet
engine? Not quite. Most of the results such as those shown in Fig. 7.37 they
are obtained over short time periods and generally in inert atmospheres.
Then there is problem of compatibility between requirements for toughness
such as ease of interface sliding with the requirements for enhanced creep
reisistnace. The message of the results compiled in Fig. 7.37 is simply that
the CMCs do have a great potential as high temperature materials.
Time for 1% £
10-6 0.00032 years
Inconel MA 754 1200 0 C (2.7 h)
Stephens & Nix
NicaloniCAS
(40v/o), IOppm O.
Abbe &
Chermant ~
.....
I
10-8 0.032 years
UJ (270 h)
aJ~ \ #0
~ •
#
"'"
fr
e
U
SCS-6IHPSN
(3Ov/o), air
1 year
3.2 years
10- 10 (27000 h)
10 years
SCS-6/RBSN
(24v/o)
10 ppm 0z
0th -60MPa
10- 12 320 years
30 100 300
Stress, MPa
Fig. 7.37 Comparison of tensile creep rate of some CMCs with a nickel-based
oxide-dispersion hardened superalloy, MA 754 at 1200 °C for 100 h. (After Holmes
and Wu, 1995).
258 MECHANICAL BEHAVIOR
REFERENCES
Aveston, J., G.A. Cooper, and A. Kelly (1971) in Properties ofFibre Composites,
IPC Science and Technology Press, Guildford, p.l5.
Becher, P.F., C.-H. Hsueh, P. Angelini, and T.N. Tiegs. (1988) J. Amer. Ceram.
Soc., 71,1050.
Becher, P.F. and T.N. Tiegs, (1988) Adv. Ceram. Mater., 3, 148.
Bowen, D.H.(l968) Fibre Sci. & Tech., 1, 85.
Bender, B.A., J.S. Wallace, and DJ. Schrodt (1991) J. Mater. Sci., 12,970.
Brennan JJ and KM. Prewo (1982) J. Mater. Sci., 17, 1201.
Bullock, E., M. McLean, and D.E. Miles,(1977) Acta. Metall., 25, 333.
Carroll, D.F. and RE Tressler (1988) J. Amer. Ceram. Soc., 71, 472.
Carroll, D.F. and R.E. Tressler (1989) J. A mer. Ceram. Soc., 72, 49.
Carroll, D.F., S.M. Wiederhom, and D.E. Roberts (1989) J. Amer. Ceram. Soc., 72,
1610.
Chokshi, A.H. and lR. Porter (1985) J. Amer. Ceram. Soc., 68, C144.
Corman, G.S. (1991) Ceram. Eng. Sci. Proc., 12,1745.
Cox, H.L. (1952) Brit. J. App. Phys., 3,72.
Crivelli-Visconti, I. and G. Cooper (1969) Nature, 221, 754.
Davidge, R.W. and A. Briggs (1989) l Mater. Sci., 24, 2815.
Dawson, D.M., RF. Preston, and A. Purser (1987) Ceram. Eng. Sci. Proc., 8, 815.
Dyson, B.F., R.D. Lohr, and R. Morrel (eds.) (1989) Mechanical Testing of
Engineering Ceramics at High Temperatures, Elsevier Applied Sci., London
Evans, C.C.,J.E. Gordon, D.M. Marsh, and N.J. Parrott (1961) Tube Investments
Research Laboratories, Report No. 133, as cited in Bowen, 1968.
Evans, P.E. (1970) J. Am. Ceram Soc. 53, 365.
Fett, T., K Keller, and D. Munz (1988) J. Mater. Sci., 23, 467.
Firestone, R.F. and A.H. Heuer, (1976) J. Amer. Ceram. Soc., 59,24.
Garboriaud, R.l (1981) Phil. Mag., 44A, 561.
Gresczuk, L.B. (1969) in Interfaces in Composites, ASTM STP 452, Amer. Soc.
for Testing and Materials, Philadelphia, p. 42.
Hahn, H.T. and R.Y Kim (1976) J. Composite Materials, 10, 156.
Han, L.x. and S. Suresh (1989) J. Amer. Ceram. Soc., 72, 1233.
Helms, H.E. and PJ. Haley (1989) in Ceramic Materials and Components for
Engines, V.l Tennery (ed.), Amer. Ceram. Soc., Westerville, Ohio, p. 1347.
Hermes, E.E. and R.J. Kerans. (1989) in Mat. Res. Soc. Symposium Proceedings,
MRS volume 125, Materials Res. Soc., Warrendale, PA, p. 93.
Highsmith, A.L. and KL. Reifsnider (1982) in Damage in Composite Materials,
ASTM STP 775, Amer. Soc. Testing & Mater., Philadelphia, p. 103.
Hirano, K, Y. Harade, and T. Suzuki (2001) in High Temperature Ceramic Matrix
Composites, W. Krenkel, R Naslain, and H. Schneider (eds.), Wiley-VCH,
Weinheim, Germany, p. 709
Holmes, J.W. (1991)J. Mater. Sci., 26,1808.
Holmes, lW. and C Cho (1992) J. Amer. Ceram. Soc., 75, 929.
259 MECHANICAL BEHAVIOR
Phillips, D.C., RA.J. Sambell, and D.H. Bowen (1972) J. Mater. Sci., 7, 1454.
Porter, J.R., F.F. Lange, andA.H. Chokshi (1987) Amer. Ceram. Soc. Bull., 66, 343.
Prewo, K.M. and J.J. Brennan (1980) J. Mater. Sci., 15,463.
Prewo, K.M., J.J. Brennan,and G.K. Layden (1986) Amer. Ceram. Soc. Bull., 65,
305.
Pysher, D.J., K.C. Goretta, R.S. Hodder, Jr.,and R.E Tressler (1989) J. A mer.
Ceram. Soc., 72, 284.
Pysher, D.J. and R.E. Tressler (1992) J. Mater. Sci., 27, 423.
Ramakrishnan V. and N. Jayaraman (1993) J. Mater. Sci., 28, 5580.
Routbort, J.L., K.C. Goretta, A. Dominguez-Rodriguez, and A.R de Arellano-
Lopez (1990) J. Hard Materials, 1,221.
Sambell, R.A.J., D.H. Brown, and D.C. Phillips, J. Mater. Sci., 7 (1972) 663.
Seltzer, M.S. and P.K. Talty (1975) J. Amer. Ceram. Soc., 58, 124.
Seshadri, S.G. andK-Y. Chia(1987)J. Am. Ceram. Soc., 70, C242.
Shuler, S.F., J.W. Holmes, X.Wu,. and D. Roach (1993) J. Amer. Ceram. Soc., 76,
2327.
Shih, CJ., J.-M. Yang, and A. Ezis (1992) Composites Sci. & Tech., 43, l3.
S0rensen B. and J.W. Holmes (1995) Scripta Met. et Mater., 32,1393.
Steif, P. and A. Trojnacki.(1994) J. A mer. Ceram. Soc., 77, 221.
Suresh, S. (1991) J. Hard Materials, 2, 29.
Suresh, S., L.X. Han, and J.J. Petrovic (1988) J. A mer. Ceram. Soc., 71, c158.
Talreja, R. (1985) Fatigue of Composite Materials, Technical University of
Denmark, Lyngby, Denmark.
Tattersall, H.G.T. and G. Tappin (1966)J. Mater. Sci. 1,296-301.
Vaidyanathan, R, J. Sankar, and V.S. Avva(1988) Ceram. Eng. Sci. Proc., 9, l383.
Venkatesh, Rand K.K. Chawla (1992) J. Mater. Sci. Lett., 11,650.
Wang, Y.R. and D.S. Liu (1989) Ceram. Eng. Sci. Proc., 10, 1154.
Wang, Z., C. Laird, Z. Hashin, B.W. Rosen, andC.F. Yen (l991)J. Mater. Sci., 26,
5335.
Warshaw, S.l. and F.H. Norton (1962) J. A mer. Ceram. Soc., 45,479.
Wei, G.C. and P.F. Becher (1985) Amer. Ceram. Soc. Bull., 64, 333.
Wiederhom, S.M. and BJ. Hockey (1991) Ceramics IntI., 17,243
Wiederhom, S.M., W. Liu, D.F. Carroll, and T.-J. Chuang (1988)J. Amer. Ceram.
Soc., 12,602.
Yang, J.-M. and Chen, S.T. (1992) Adv. Composites Lett., 1,27.
Zwissler, J.G., M.E. Fine, and G.W. Groves (1977) J. Amer. Cemm. Soc., 60, 390.
SUGGESTED READING
THERMAL STRESSES
Thermal stresses are very important in all composite materials, but particularly
so in ceramic matrix composites because, in general, a ceramic matrix has a
very low strain to fracture. Ceramics, in general, also have a lower coefficient
ofthermal expansion than polymers and metals. Figure 8.1 shows linear thermal
expansion as a function of temperature for some polymers, metals, and
ceramics. Note the rather low coefficient of expansion of glass-ceramic (LAS)
and fused silica. Thermal stresses are internal stresses that arise when there
exists a constraint on free dimensional change of a body (Chawla, 1973a,
1973b). In the absence of this constraint, the body can experience free thermal
strains without any accompanying thermal stresses. The constraint can have its
origin in:
2.0
1.6
~
1.2
"
0
'"
"0.
0
x
w
'0 O.B
E
Q)
.c
f- Silicon Carbide
~
Q)
"
:.:J
0.4
tt.~~~======::1111~:""--- Fused
o II Silica
LAS Glass-Ceramic (LiAISi 20 S )
-0.2 L - - _ - ' - _ - L _ - - ' - _ - l ._ _L-_.L.-_-L-_-.L..:-:--_-'
Temperature, 0 C
8.1 Linear thermal expansion as a function of temperature for some polymers, metals,
and ceramics.
where Cp is the specific heat, p is the density, and k is the thermal conductivity.
Metals generally have high thermal diffusivity; therefore, any thermal gradients
that might develop are dissipated rather quickly. Ceramics, on the other hand,
have low thermal diffusivity and, thus, poor thermal shock resistance. It is easy
to see that the presence oftemperature gradients in CMCs can make an already
265 THERMAL STRESSES
Thermal stresses arising from a thermal mismatch between the two components
of a composite are a function of a number of variables:
where E represents the elastic moduli, ~a. is the thermal mismatch, ~T is the
temperature change, a and b the radii of the components, and V represents the
volume fractions of the components. The specific geometric distribution of one
component in the other one can have very important effects. We derive below
the analytical expressions for thermal stress components in two types of
composites: a central particle surrounded by its associated spherical shell of
matrix and a central fiber surrounded by its cylindrical shell of matrix.
ar> aa = a'll
cr> ca = c'll
ur = u, independent of e or <p
dar 2
+ -(a - aa) = 0 (8.2)
dr r r
267 THE~STRESSES
~--------~--- y
,,
, " : p'
,
'
6..
"
x
Fig. 8.3 Spherical coordinate system.
268 THERMAL STRESSES
E =-
du
r dr
u
(8.3)
r
Substituting Eq. (8.3) in Eq. (8.2), we get the governing differential equation
for our problem:
(8.4)
u = Ar + -
c
r2
cr,(a) = -p
When we apply these boundary conditions, we find that the stresses in the
ceramic particle are
,(Jrm
P
- Vp [a'-r3 - V
p 1 (8.6)
p
(Jem -
- Vp [ 2 r3 a'
1- + V
p 1 (8.7)
p
( 8.8)
whereVp is the particle volume fraction, a is the particle radius, b is the matrix
radius, and other symbols have the significance given earlier. Figure 8Aa shows
the three-dimensional stress distribution in a particulate composite. The particle
is under a uniform pressure, P, while the matrix has radial and tangential stress
components. The radial and tangential components in the matrix vary with
distance as shown in Fig. 8Ab. The radial component goes to zero at the free
surface, r = b, as per our boundary conditions. The tangential component has
a nonzero value at the free surface.
b
MATRIX
(a)
(b)
The axial symmetry means that we can treat the problem in terms of the
principal stresses that are independent of e. We derive expressions for the
thermal stresses in a three-element cylindrical composite. The two-element,
uncoated fiber composite will then be a special case, wherein the coating is the
same material as the matrix. For the axisymmetric case under consideration, the
radial and circumferential stresses will have expressions of the following form
(Poritsky, 1934; Hull and Berger, 1934; Chawla and Metzger, 1972):
B.
air =A.----! (8.9a)
, r2
B.
a,lI = A.+-'
, r2
(8.9b)
The subscript i represents fiber (t), coating (c), and matrix (m) and the constants
Ai and Bi are determined by the boundary conditions. Also note that air + a ie =
2~.
For the case of a coated fiber with a finite matrix region we can write the
following expressions for tangential and axial strains in the fiber (r = 0 to a),
coating (r = a to b), and matrix (r = b to c):
272 THE~STRESSES
Matrix
:c:
," ,
~ , ,
b' :
I , ,
, ,
h h
I ,
(a) (b)
Fig. 8.5 (a) Basic fiber/matrix unit consisting of a central fiber (radius, a) surrounded
by its sleeve of matrix (radius, b), (b) three-element unit consisting of the central fiber
surrounded by two concentric sleeves of the coating and the matrix material.
E;z = ~.[cr;z-v;(cr;r+cr;6))+(l;LlT
I
(S.10b)
~. [cr;z - v;2A;] + (l;LlT
I
Note that in the case ofthe fiber, Br = 0, otherwise the radial stress will become
infinite at r = O. This leads to the special condition that crre = crfr = Af
273 THERMAL STRESSES
Let the radial stresses at the fiber/coating and coating/matrix interfaces be 0".
and O"b' respectively. At r = c, the free surface, O"mr (= O"c) is zero, because the
free surface does not support a normal stress. Using this boundary condition,
we can write
at r = c o = Am
Am
1 1
ame =Bm -+- (S.11a)
c 2 r2
a me = Am [1 + ~l
r
2
This gives
B
A = A __C
f c
a2
274 THE~STRESSES
_1
E [( A c + Bc)_
c a
2
vc[( A c - Be)
2
a
II
+ cr cz + uc!l. T
(In,-ac )I1T
A _
= _c +B
_
EC
/b'1.
c_
V
- _c (A
EC C C
B c2 +b2
-B /b2) _ ~
E 2b2
m
l C
j
- v
~O"
EC cz
0" V
v
+ ~ + -.!!!...
E 111 E 111
I-B [22b2b 211
m
C-
C
E [0"C-v
_1
C
ZCC
2A ] + a C I1T E [0" mzm
= _1 -v 2 Bm2
m C
I+ a m I1T
0" - ---.!!!!..
(a -a )I1T = ...5.. 0" - _c 2A c _ ( A __
v __ B -.-!!!.. -,-c_~
(2 b 2)
C
)2V
m C Ec Em Ec C b2 E m (2
C
-b2)
(8.13)
After combining the equations obtained from the application of the boundary
conditions, we can write
276 THERMAL STRESSES
(8.14)
(8.15)
(8.16)
(8.17)
-2v
= __f
K6
a 2E
'/
1
K7 - -
Ec
Ec c
c 2-+b+
-1( l - v ) +1- ( -
Em
2
v
2_b 2 m
1
C
K -
14 -
2
b2
VIII (
Em
b2 1
c 2 -b 2
278 THERMAL STRESSES
K 15
Ec
1
K 16 -
Em
Equations 8.14 through 8.17 can be solved for Ae, Be' O"rz' O"m" and O"w This
thermoelastic solution can provide information about the magnitude of the
elastic stresses involved and if the elastic range will be exceeded or not. In the
case of a metallic matrix, it is likely to deform plastically in response to these
thermal stresses (Chawla, 1973a; Chawla, 1973b; Arsenault and Fisher, 1983;
Christman and Suresh, 1988). In the case of a ceramic matrix, cracking can
occur to relieve the high level of thermal stresses. Figure 8.6 shows the
distribution of the radial, tangential, and axial stresses in a silicon carbide
fiber/alumina matrix composite for a temperature change of - 1 °C for two fiber
volume fractions, 5 and 35 %. In this figure we have used the data for a CVD
SiC fiber (see chapter 3) and a fully dense alumina (see chapter 2). We can
draw some important inferences from Figure 8.6:
(i) Axial stress is uniform in fiber and matrix, although its magnitude is
different in the two and depends on the respective elastic constants.
(ii) In the fiber, i.e., the central component f, O"rf and O"er are equal in magnitude
and sense. In the matrix, i.e., the sleeve m, O"nn and O"am vary as [1 - (b2/r2)] and
[1 + (b 2/r2)], respectively.
(iii) When either the temperature difference or the expansion coefficient
difference goes to zero, the thermal stresses vanish, as expected.
Figure 8.6 also shows that the change in the axial stress magnitude with a
change in fiber volume fraction is much more than in radial and tangential
components. Also to be noted is the compressive radial stress at the fiber/matrix
interface in this system, a situation not very favorable for interfacial debonding.
If we incorporate a coating between the fiber and the matrix, the stress
distribution will change. This is commonly done to prevent a strong chemical
bond between the matrix and the fiber and thus provide a somewhat weak
mechanical bond which is conducive to fiber pullout, a desirable feature for
enhanced toughness. An example involving a coating is the alumina fiber/silica-
279 THERMAL STRESSES
'0.70
0.56 SiC/AI 2 0 3
0.42 fl T= -1°C
Vf =0.05
10.28
vi '0.14'
<Il a~m
....~
C/) 0.00
co
E -0.14
Q;
..c -0.28:
I-
-0.42
-0.56
-0,70 '--_ _...._ _....I.._ _...I._ _....&_ _ ~
based glass matrix composite. Commonly, a carbon coating is used such that
one has an alumina fiber/carbon coating/glass matrix composite. Alumina and
silica-based glass form a strong chemical bond; the carbon coating prevents that
and provides a weak mechanical bond. It can be shown that the radial stress in
alumina/glass and alumina/carbon/glass is tensile. Even though the radial stress
is tensile in alumina/glass, the strong chemical bonding between silica-based
glass and alumina inhibits debonding. In alumina/carbon/glass composites, the
tensile radial stress combined with weak mechanical bonding due to the carbon
coating aids in debonding. This has been verified by a number of researchers
(Fitzer and Schubert, 1979; Brun and Singh, 1988; Jurewicz et aI., 1989;
Feldman, 1986). We discuss this subject at some length in chapter 9.
Residual stresses are elastic stresses that remain in a body after all external
loads have been removed. It is important to realize that, for a body in
equilibrium, residual stresses must form a balancedforce system in a body. The
magnitudes of tensile and compressive stress components, however, need not
be equal. The most important source of residual stresses in CMCs, as discussed
above, is the thermal mismatch between the components. These can be relieved,
to some extent, in a CMC by micro cracking of the matrix. That, however, may
not always be the case. For example, in a three-dimensional carbon/carbon
composite, Feldman (1986) observed warping in thin slices of the composite.
As we pointed out in chapter 3, carbon fiber itself is highly anisotropic in
thermal expansion. The high processing temperatures involved make the
residual stress problems particularly severe. Feldman cut thin slices of different
281 THERMAL STRESSES
The most important effect of the thermal expansion mismatch between the
reinforcement and matrix is that residual stresses build up in the composite
when the composite is subject to any temperature change, say, during cooling
from the processing temperature to room temperature, or during service. The
residual stresses have significant effects on both interface sliding stress and
matrix cracking stress (Evans and Marshall, 1989) as well as ultimate strength
(Hutchinson and Jensen, 1990). The residual axial stress will directly
superimpose on the applied load in the fiber direction, and large tensile residual
stress will induce cracks in either fiber or matrix. The radial tensile forces at the
interface will encourage fiber debonding and slippage, which in turn result in
high toughness, high work of fracture, and a fibrous appearance on the fracture
surface. The interfacial sliding resistance can be expressed as (Hutchinson and
Jensen, 1990)
where 't is the shear stress acting at the interface, 0"2 is the stress normal to the
interface, Jl is the friction coefficient and 'to is the sliding resistance when 0"2 is
positive (tensile).
2d hkl sin9 = A
The generalized Hooke's law relating stress and strain components for an
isotropic solid, in Cartesian (x, y, z) coordinates, can be written as
1 1
-[0" -v (0" +0" )], E._ -[0" -v (0" +0")]
E Y Z x E Z x Y
where s is the strain, 0" is the stress, E is the Young's modulus, and v is the
Poisson's ratio.
If a large plate has a thennal gradient across its thickness and it is cooled from
both surfaces, then we have a state of biaxial stress at any point. The two stress
components are given by
aEl!.T
(I-v)
284 THERMAL STRESSES
uEf).T
°max
(I-v)
hr"
k
The surface heat transfer coefficient, h, depends on the quenched material and
the surroundings. One can incorporate Biot' s modulus into a function,f@} such
that ATc
(J (I-v)
max 1{P)
uE
where 0max is the maximum surface stress that develops in a quenched body, A Tc
is the critical temperature difference that will initiate cracks in the sample,
generally at the surface, and the parameter /(/3) is given by:
4.67 51
j{P) = 1.5 + - - 0.5 exp(--)
p p
For a severe quench, /3 is very large, which makes/@} approximately equal to
1.5.
285 THERMAL STRESSES
R = I1Tc
R'
where we have introduced the thermal conductivity, k, of the material. For slow
cooling rate, the parameter f(fJ) is about 1I0.31~, which gives the critical
temperature difference as
R'
toughness (K,c):
286 THERMAL STRESSES
2
R /11/ = Ey K 1c
In view of the fact that CMCs are likely to find major applications at high
temperatures, it is of interest to study their behavior under conditions of
isothermal exposure as well as under conditions of thermal cycling. We should
mention that more realistic testing for thermal shock resistance involves rather
expensive equipment such as burner rigs or other hot gas testing facilities and
combustion chambers (Eckel et aI., 1997). Many researchers have contributed
in this area (Blissett et aI., 1997; Boccaccini et aI., I997a, I997b; Bouchetou
et aI, 1993; Bhatt, 1992; Bhatt and Palczer, 1997; Case et aI., 1993; Chawla et
aI., 2001; Kagawa et aI., 1993; Kim et aI., 1990; Lamicq et aI., 1986; Sinha et
aI., 1994, Tiegs and Becher, 1987; Wang et aI., 1996; Wang and Chou, 1991;
Wang and Singh, 1994; Wetherhold and Zawada, 1991). We provide a summary
of the results. Wetherhold and Zawada (1991) studied the behavior of ceramic
grade Nicalon fiber in an aluminosilicate glass matrix under isothermal and
thermal cycling. In the 650 - 700°C, isothermally exposed as well as thermally
cycled samples showed rapid oxidation and loss in strength. Oxidation behavior
overshadowed any thermal cycling effect for these test conditions. The
embrittlement was attributed to oxygen infiltration from the surface, which
destroyed the weak carbon-rich interface in this composite. At 800°C, however,
less embrittlement was observed and the fiber toughening effect remained. This
decreased embrittlement at higher temperatures was attributed to smoothening
of the sample surface by glass flow and slow oxygen infiltration. Thermal shock
resistance of glass matrix containing aluminum titanate particles was improved
as evidenced by an increase in ~ Tc from about 100 to 400 K for 30% aluminum
titanate (Boccaccini and Pfeifer, 1999). Boccaccini et al. (1997a, 1997b)
studied the thermal shock behavior ofNicalon fiber (NL 202) reinforced Duran
glass (a borosilicate glass) matrix composite, a system that has almost zero
thermal mismatch between the fiber and the matrix. Thermal shock quenching
in water (22°C) from 620 °C led to matrix microcracking. The evolution ofthis
287 THERMAL STRESSES
damage was studied by measuring the change in Young's modulus and internal
friction; a decrease in the modulus and a simultaneous increase in internal
friction occurred. The change in internal friction was more sensitive to
microstructural changes, especially in the early stages of damage development,
than the change in Young's modulus. Crack healing occurred in the glass matrix
composites when cycled above the glass transition temperature of the matrix.
Lamicq et al. (1986) used a water quench technique to test the thermal shock
resistance of SiC fiber/SiC composite materials made by chemical vapor
infiltration. The composite showed a loss in flexural strength gradually over the
range of i1T = 300-750° C. Increasing i1T to 1200° C resulted in no further
decrease in flexural strength. Bhatt and Phillips (1990) also used water
quenching on a reaction bonded SCS-6 SiC fiber/Si3N4 composite and observed
a decrease in its flexural strength at i1T greater than 600°C: however, tensile
strength remained unchanged! They attributed this fact to a loss in interply
integrity alone, with shear delamination occurring in flexural tests but not in
tensile tests.
Kagawa et al. (1993) incorporated Nicalon fibers into two different matrix
materials-borosilicate glass and lithium alumino silicate (LAS) - for which the
fiber/matrix interfacial bond is very weak. Unidirectional composites were
made by hot pressing. Water quench thermal shocks with i1T values of250 to
1000 ° C produced multiple cracking of the matrix perpendicular to the fiber
axis for i1T greater than 600° C. This cracking occurred only in the outermost
surface of the material. Flexural strength decreased by less than 5% at i1T = 800
°C. Neither elastic modulus nor flexural strength were affected at i1T < 600°C.
Also, the depth of crack penetration was independent of any increase in i1T, and
only the increase in crack density at the surface was responsible for the
decrease in elastic modulus. For the LAS matrix composite, however, the cracks
formed parallel to the fiber axis, but again they formed only at the surface.
Flexural strength remained virtually unaffected by thermal shock. Elastic
modulus, however, decreased as a result of matrix microcracking for i1T > 600
°C. At 800°C, a phase change occurred in the LAS matrix, which could be a
possible source of microcracking.
• Fiber reinforcement changes both the crack initiation and the crack
propagation behavior of a ceramic.
• The mismatch of properties between the fibers and matrix (e. g. the
Young's modulus, coefficient of thermal expansion, thermal
conductivity, or symmetry of properties) causes additional thermal
stresses within the material.
Wang and Chou (1991) summarize the following conclusions from their work
on SiC fibers in a borosilicate glass matrix:
REFERENCES
SUGGESTED READING
Noyan, I.C. and J.B Cohen (1987) Residual Stress, Springer-Verlag, New York.
Schneider, G.A. and G. Petzow (eds.) (1993) Thermal Shock Behavior and Thermal
Fatigue ofAdvanced Ceramics, Kluwer Acad. Pub., Dordrecht, The Netherlands.
Cliapter 9
Once the matrix and the reinforcement of a composite are chosen, it is the set
of characteristics of the interface region that determines the final properties of
the composite. In this regard a thorough microstructural characterization of the
interface region assumes a great deal of importance. Many sophisticated
techniques are available to characterize the interface region. We give examples
from some important CMCs that bring home two points: firstly, there are tools
available that allow us to make an in-depth characterization and secondly, the
interface region in real composites can be immensely complex and variable.
The chemical composition of the components, the crystallographic as well as
chemical nature of the interface region, the effect of processing conditions, etc.
can affect the properties and performance of the composite to a very large
extent.
Fig. 9.1 An example of the highly disordered interface region in a rice-hull derived
silicon carbide whisker/alumina composite made by hot pressing (courtesy of A.
Weresczczak). Note also the stacking faults in the whiskers.
293 INTERFACE MECHANICS AND TOUGHNESS
shows an example of the interface region in a rice hull derived silicon carbide
whisker/alumina composite made by hot pressing. One can see the stacking
faults in the SiC whisker and the disordered region at the interface. Das
Chowdhury et al. (1 992a, 1992b) observed discontinuous and continuous
interfacial layers in silicon carbide whisker reinforced silicon nitride
composites by high resolution electron imaging and analytical microscopy.
Examples of their observations and analyses of interfaces in these composites
are shown in Fig. 9.2 and 9.3. Figure 9.2 shows a bright field TEM micrograph
of the interface in a SiCwf'Si3N4 doped with Y203 and A1 20 3. The circled region
was analyzed by high resolution electron microscopy, see Fig. 9.3b and d. A
high resolution electron microscopy image of the interface, Figure 9.3a, shows
an amorphous region at the interface. The structural width of the interface
region is 1.5-2 nm. The arrow indicates the scan direction. Figure 9.3b shows
the variation of oxygen concentration across the interface as a function of
distance. A sharp change in oxygen concentration occurs at the interface. The
reader will also notice rather large difference in chemical and structural widths
of the interface. Figure 9.3c shows a high resolution electron image of another
interface region ofthe same composite while Fig. 9.3d shows the oxygen profile
across the interface. In this case no abrupt change in oxygen concentration at
the interface was detected. One also notes the high oxygen concentration in the
silicon nitride matrix and the large difference between the chemical and
structural width of the interface.
Fig. 9.2 A bright field TEM micrograph of the interface in a SiCjSi3N 4 doped with
Y203 and A120 3. The circled region was analyzed by high resolution electron
microscopy (courtesy ofK. Das Chowdhury).
294 INTERFACE MECHANICS AND TOUGHNESS
(a)
1.2
O-K
1.0
0.8
;: 0.6
x
E
~
0
U 0.4,
0.2
0.0
0 200 400 600 800 1000 1200
Distance A
(b)
Fig 9.3 (a) HREM image ofthe interface showing an amorphous region atthe interface.
The structural width ofthe interface region is 1.5 - 2 nm. The arrow indicates the scan
direction. (b) Variation of oxygen concentration across the interface as a function of
distance. Note the sharp change in oxygen concentration at the interface. (c) HREM
image of another interface region of the same composite (d) oxygen profile across the
interface in (c); in this case no abrupt change in oxygen concentration at the interface
was detected. Note the high oxygen concentration in the silicon nitride matrix and the
large difference between the chemical and structural width of the interface. (Courtesy
ofK. Das Chowdhury).
295 INTERF ACE MECHANICS AND TOUGHNESS
(c)
0.5
O-K
0.4
'" S
>< 0.3
~::>
0
U 0.2 SiC Silicon Nitride
~ interface
0.1
0 200 400 600 800 1000 1200
Distance A
(d)
The transformation at the crack tip results in dilation of the transformed zone
around a crack. This increase in volume is opposed by the surrounding
untransformed material leading to compressive stresses which tend to close the
crack. This makes crack propagation difficult leading to an increased toughness.
Figure 9.4 shows this transformation toughening mechanism. Transformation
in the wake of a crack can result in a closure force that tends to resist the crack
opening displacement. Crack deflection at zirconia particles can also contribute
to toughness. A good survey of transformation toughening in particulate CMCs
can be found in Green et al. (1989). Kriven (1995) have exploited the phase
transformation concept in fiber reinforced CMCs containing a fiber coating
which undergoes a phase transformation involving a volumetric change; we
discuss this below.
It can be seen from the above analysis that in all these mechanisms, control of
the crack propagation, i.e., crack path, plays a key role in the improvement of
crack growth resistance or fracture toughness of CMCs. There has been some
work to quantify the crack path, and relate it to the mechanical behavior of the
materials (Bhargava and Patterson, 1997; Wu et aI., 2000, 2001). They analyzed
the influence of crack path on fracture toughness of particle reinforced CMCs
such as Al 20 3 particles in a cubic zirconia matrix, silicon carbide or alumina
particles in a glass matrix. They defined a stereo logical parameter, called
degree of contact, R, to characterize interactions between crack and reinforced
particle. As one would expect, they found that the thermal and elastic misfits
between the matrix and the particulate reinforcement, the loading and/or
unloading rate, and the test environment all significantly influenced the crack
path. The crack propagation velocity was directly related to the variation in
loading rate. The amount of particle-crack interaction significantly influenced
the fracture toughness of ceramic composites. For a composite in which the
matrix had higher thermal expansion coefficient (n) and lower Young's
modulus (E) than the particle, increasing interaction between particles and
cracks decreased fracture toughness. For a composite in which the matrix had
lower n and lower E than the particles, decreasing interaction between particles
and cracks decreased the fracture toughness. This work showed that, excluding
the environmental effect, crack-particle interactions played a key role in
influencing the fracture behavior of the composites. This degree of contact
parameter provides a simple and useful tool for analysis of the influence of
variables on particle-crack interaction and influence of such interaction on
fracture behavior of ceramic composites.
reinforced CMCs as well. Whisker or fiber reinforcements can bring into play
a series of energy absorbing mechanisms in the fracture process of CMCs and,
thus, make them damage tolerant.
Mechanism Requirement
5. Phase transformation The crack tip stress field in the matrix can cause
toughening the second phase particles (fibers) at the crack tip
to undergo a phase transformation causing
expansion (il V> 0). The volume expansion can
squeeze the crack shut.
The final failure of the composite is not the result of the passage of a single
crack, i.e., a self-similar crack propagation does not occur. Thus, it is difficult
to define an unambiguous fracture toughness value, such as a K1c value
(Marshall et aI., 1985). Under some circumstances, matrix cracking stress, 0"0
can be independent of the preexisting flaw size and is a material property. 0"0
is a function of matrix fracture toughness, interfacial shear strength, fiber
volume fraction, fiber radius, and the elastic constants of the fiber and matrix.
For fully bridged cracks, the 0"0 value can be independent of crack length
(Marshall et aI., 1985).
Although, the appearance of the first matrix crack does not signify a complete
failure of the composite, it does result in a reduced slope of the stress-strain
curve, i.e., a decreased modulus. It also implies an easy access path for any
aggressive environmental species that might be present. Matrix microcracking
will cause a reduction of strength and modulus and might lead to internal
oxidation and spalling as well. Thus, in practical terms, any microcrack
toughening is not likely to be very high and the reliability of such composites
containing microcracks also will not be very high.
(a) (b)
Fig. 9.5 Typical stress-strain curves obtained with continuous fiber reinforced ceramic
matrix composites. (a) in tension (b) in flexure.
301 INTERFACE MECHANICS AND TOUGHNESS
• Matrix fracture
• Debonding at the crack tip, followed by crack deflection
• Debonding in the crack wake, followed by crack deflection
• Frictional sliding between the fiber and matrix
• Fiber failure
• Fiber pullout
It has been amply demonstrated (Donald and McMillan, 1976; Ramualdi and
Basto, 1963; Majumdar, 1970; Sambell et aI., 1972; Aveston et aI., 1973;
Aveston, 1973; Vaidya et aI., 1992) that reinforcement with continuous fibers
such as carbon, alumina, silicon carbide, mullite fibers in cement, glass, and
glass-ceramic matrix can result toughening. It is not necessary for all these
failure mechanisms to operate simultaneously for a given fiber/matrix system
and often in many composite systems only one or two of these toughness
contributions will dominate the total fracture toughness. Let us examine these
mechanisms in some detail.
It would appear that the debonding (sometimes the term delamination is also
used, but, we prefer debonding) of the fiber/matrix interface is a prerequisite
for phenomena such as crack deflection, crack bridging by fibers, and fiber
pullout. Cook and Gordon (1964) analyzed the crack deflection at an interface
between materials of identical elastic constants, i.e., the same material joined
at an interface, see Sec 5.5 and Figs. 5.6 and 5.7 in chapter 5.
Two materials that meet at an interface are more than likely to have different
elastic constants. The modulus mismatch causes shearing of the crack surfaces.
This results in a mixed-mode stress state in the vicinity of a crack tip at an
302 INTERFACE MECHANICS AND TOUGHNESS
interface involving both the tensile and shear components. This, in turn, results
in a mixed-mode fracture, which can occur at the crack tip or in the wake of the
crack. Figure 9.6 shows this crack front and crack wake debonding in a fiber
reinforced composite. What this means in practical terms is that instead of a
simple, one parameter description by the critical stress intensity factor Kle , one
needs a more complex formalism of fracture mechanics to describe the
situation. The parameter K under such a situation becomes scale sensitive, but
the critical strain energy release rate, Gle , is not a scale sensitive parameter. G
is a function of the phase angle, 'P, which is a function of the normal and shear
loading. One needs to specify G and 'P to analyze the debonding at the
interface. G and 'P are related, through the so-called Dundurs parameters, a and
~, as defined below:
1 [G\(l-2v2) - Gi l - 2v\)]
(9.1)
2 [G\(l-v 2) + G2(l-v\)]
where
- b
E=--
1 - v2
The expression for the phase angle, 'P, in terms ofthe elastic coefficients of the
two media, radius r from the crack tip and the displacements, u and v, at the
crack tip is as follows:
303 INTERFACE MECHANICS AND TOUGHNESS
Such expressions have been used by many researchers (He and Hutchinson,
198~; Evans and Marshall, 1989; Ruhle and Evans, 1988; Gupta, 1991; Gupta
et aI., 1993) to analyze the conditions for fiber/matrix debonding in terms of the
energy requirements. One result of this body of work is shown in Fig. 9.7 in
terms of a chart of G/Gf vs. a, where Gi is the mixed-mode interfacial fracture
energy of the interface, Gr is the mode I fracture energy of the fiber, and a is the
measure of elastic anisotropy as defined above. For opening mode or mode I,
'P = 0° while for mode II, 'P = 90°. The plot in Fig. 9.7 shows the conditions
under which the crack will deflect along the interface or propagate through the
interface into the fiber. For all values of G/G f below the dashed line area,
interface debonding is predicted. For the special case of zero elastic mismatch,
i.e., for a = 0, the fiber/matrix interface will debond for G/Gf less than about
0.25. Conversely, for G/G r greater than 0.25, the crack will propagate across
the fiber. In general, for the elastic mismatch, a greater than zero, the minimum
interfacial toughness required for interface debonding increases, i.e., high
modulus fiber tend to favor debonding. One shortcoming of this analysis is that
it treats the fiber and matrix as isotropic materials. This is not always true,
especially for the fiber.
Il) r III
t r III
1
r1
~ -
t I
'-------
I ,
r ,)
-r
-
frl l ll/
.--,
~
d¢bondi n" ~ dcbon ling
Fig. 9.6 Crack front and crack wake debonding in a fiber reinforced composite.
304 INTERFACE MECHANICS AND TOUGHNESS
1.0
<!>
.....
0.5
o ~--~----~--~--~
-1.0 -0.5 o 0.5 1.0
Elastic mismatch, a
67.5
..,..'" 45.0
~
22.5
-1.0 -0.5 0 0.5 1.0
Elastic mismatch, a
Fig. 9.7 Charst of G/Gr vs. elastic mismatch, a and phase angle, 'I' as function of a
(after Evans and Marshall, 1989).
Gupta et al. (1993) derived strength and energy criteria for crack deflection at
a fiber/matrix interface for several composite systems, taking due account ofthe
anisotropic nature of the fiber. Their laser spallation technique was described
in chapter 5. This technique allows one to measure the tensile strength of a
planar interface. They have tabulated the required values of the interface
strength and fracture toughness for delamination in number of ceramic, metal,
intermetallic, and polymer matrix composites. We discuss their results briefly.
Figure 9.8a shows a crack terminating normal to the interface between two
orthotropic materials. Both materials are treated as homogeneous, linear elastic
solids. Fibers are anisotropic while the coatings are isotropic. Gupta et al.
(1993) find that the stress and deformation fields depend on two material
parameters, ').. and p, for the two media and two bimaterial constants, a and B.
For plane stress condition, these parameters depend only on the elastic
compliances, Sij of the two media. These parameters are defined as follows:
(9.2a)
305 INTERFACE MECHANICS AND TOUGHNESS
(9.2b)
rss
Y = V~22~11' n= ~
~~ (9.2c)
where the subscripts 1 and 2 outside the parentheses in Eq. 9.2b refer to the two
media, with the subscript 2 designating the cracked medium. Note that the
parameters A. and p become unity for an isotropic medium. We use these
parameters in describing the important results of Gupta et al. (1993).
According to this criterion, the crack will deflect along the interface if
where o*j is the interface strength, O*f is the fiber strength, Oyy{QO) is the stress
at the interface, and oxx{9QO) is the stress in the fiber. Figure 9.8b depicts this
strength criterion for delamination. Knowing the elastic constants of the fiber
and matrix, we can calculate a. and p. For these values of a. and p, the ratio
o*/o*c can be obtained from Fig. 9.8b. We can obtain the fiber strength, o*c,
from the manufacturer or determine it experimentally. Knowing these three
quantities, we can obtain the maximum permissible interface strength value for
delamination.
The energy release rate criterion for crack deflection considers two possible
directions of crack propagation at the interface: along the interface or across the
interface. If the ratio of energy release rate for crack growth along the interface,
Gd, to growth perpendicular to the fiber, Gp, is less than the ratio of work of
delamination of the interface, Gci , to the work offracture across the fiber, Gcft,
then the crack will deflect along the interface, i.e., the crack will deflect for
306 INTERFACE MECHANICS AND TOUGHNESS
Fiber (2)
-
Matrix
y,2
Reaction -.l;~;;:I@~F==-T--,-
product Gcft
wedge
(a)
0.2
(b)
Fig. 9.8 (a) A crack tenninating nonnal to the interface between two orthotropic
materials (b) strength criterion for interface delamination. (After Gupta et ai., 1993)
The main crack may be deflected singly or doubly, Fig. 9.9 (a,b). According to
Gupta et aI., for most pairs of materials, the energy release rate is higher for the
doubly deflected crack than that for the singly deflected case. Figure 9.10 plots
G/Gp for a doubly deflected crack as a function of the bimaterial parameter a
for a series of 1-,1 values.
ofGiGp for a few chosen interface systems from metal, ceramic, and polymer
matrix composites. Table 9.2 gives the information for some ceramic systems.
The first column is the reinforcement (medium 1), the second column is either
a coating or a matrix. Columns 3 and 4 give the values of a and ~, respectively,
for the material pairs in columns 1 and 2. Columns 5 and 6 give the calculated
energy release rate ratios for the singly deflected and the doubly deflected
cracks, respectively. Column 7 gives the maximum of the two values and thus
provides a bound on the ratio GjGeft for interface delamination. Column 8
gives the ratio of the mode 2 to mode 1 components of the stress intensity
factors at the deflected crack tip. Note that almost all interfaces are governed
by the doubly deflected crack.
2
°max.
m f m f
( (
\ .
(a) (b)
Fig. 9.9 Crack deflection at the interface (a) singly deflected and (b) doubly deflected.
where h is the unstressed separation distance between the planes meeting at the
308 INTERFACE MECHANICS AND TOUGHNESS
interface, Eo is the tensile modulus of the interfacial layer, and other symbols
have the same significance given earlier. This kind of information does provide
some fundamental input into the behavior of a given pair of components of a
composite and the interface thereof.
The strain value at which the first matrix crack appears in a unidirectionally
reinforced, fibrous CMC and the corresponding first matrix crack stress value
can be important parameters. Aveston, Cooper, and Kelly (1973) made an
energy balance analysis of a composite consisting of unidirectionally aligned,
continuous fibers, in a matrix whose failure strain is less than that of the fibers.
The ACK model makes the following assumptions:
* AI= 0.2
+ AI= 1.0
X AI= 2.0
1.5
o AI= 5.0
Lnrerface
O.S
Fig. 9.10 Plot of G/Gp for a doubly deflected crack as a function of the bimaterial
parameter a for a series of A) values (after Gupta et aI., 1993).
Table 9.2 Energy criterion for interface crack deflection for various metallic, ceramic, intermetallic and polymer matrix composites
(adapted from Gupta et a1., 1993)
• the fiber and the matrix both have well-defined stress - strain curves and they
are elastic to fracture.
• the fiber/matrix interface is under a constant, frictional shear stress.
• the interface debonding energy is neglected.
• the situation of isostrain prevails.
• the strain to failure of the brittle matrix in the composite is less than that of
the fiber.
In such a composite, with only frictional bonding between the fiber and the
brittle matrix, the ACK analysis shows that the matrix failure strain is enhanced
by the presence of fibers and that the matrix will be traversed by a series of
more or less parallel cracks and perpendicular to the fibers. The crack
separation will be between x and 2x. Consider the loading of such a composite
in tension. The brittle matrix will crack first. The ACK model makes another
assumption at this point, to wit, the matrix crack propagates across the whole
matrix, causing a planar fracture in the matrix and the broken halves of the
matrix are bridged by the fibers. When the first matrix crack appears, the stored
elastic energy in the matrix is released. At the same time, an additional load is
thrown onto the fibers in the vicinity of the crack. An energy balance done by
ACK indicates that the matrix failure strain of such a CMC will be enhanced
when the following inequality holds
(9.3)
where Emu is the failure strain of the unreinforced ceramic; 't is the interfacial
frictional bond strength; Ym is the matrix fracture energy; r is the fiber radius;
Yrand Ymstand for fiber and matrix volume fraction, respectively; Er, Em, and
Ec are the Young's moduli of the fiber, matrix, and composite, respectively.
If the fiber radius, r, is large and we have a small fiber volume fraction, Yr, the
right hand side in Eq.(9.3) will be small and Emu will be greater than the right
hand side in Eq.(9.3). In this case, the matrix strain in the composite Emuc will
be equal to the matrix strain of the unreinforced ceramic, Emu' When we have a
small fiber radius, r, and a large fiber volume fraction, Yr, then Emu will be less
than the value given by the equality in Eq.(9.3) and we will have Emuc> Emu'
We now obtain an expression for the crack spacing, x. Assuming only sliding
311 INTERFACE MECHANICS AND TOUGHNESS
friction between the fiber and matrix, i.e., physical or chemical bonding,
consider a single crack in the matrix. Over some characteristic distance x from
the crack, the load on the intact fiber will be greater than the average load on
the fiber. This distance x depends on the shear strength of the interface, t. Let
us say there are N fibers per unit area, then
The load transferred from a fiber to the matrix over a distance x is 21trxt. This
must equal the tensile load carried by the matrix V mam' i.e.,
where am is the tensile stress in the matrix at a distance x from the crack face.
At some point along the fiber length, the fiber and matrix will experience the
same strain.
2V
_f'CX
I
r
(9.5)
With increasing applied load, the load on the central region of the matrix
increases until it fractures. Of course, the matrix fracture will occur only if
enough load transfer length is available for the matrix fracture strength to be
attained. Let us say that at x = XI' am = a mu' the fracture strength of the matrix.
Then, from Eq.(9.4), we have The ACK model does have some problems.
Consider Eq. (9.3). It indicates that the matrix strain to failure goes to zero at
V f = 0, which clearly not tenable. The model also predicts that Cmuc increases
with interfacial strength. Thus, a high interfacial strength would be desirable.
This ignores the fact that in the case of a strong interface, fiber/matrix
debonding may not even occur.
the steady state cracking (SSC) in CMCs including frictional and residual
stresses. The SSC condition is when crack grows with no increase in load. The
BHE model considers two types of interfaces:
(i) Unbonded fibers, held in the matrix under the action of thermal or other
mismatches, but subject to frictional sliding.
(ii) A weak fiber/matrix interface which becomes debonded under the action of
stresses at the crack tip in a matrix.
Both ACK and BHE models predict that above a certain fiber volume fraction,
the matrix cracking strain (and stress) will be higher than that of the
unreinforced material. It is worth pointing out that both these models consider
a very special case of composite geometry, loading and failure mode. Details
of crack growth are ignored and an instantaneous planar crack formation in
matrix is assumed. The so-called SSC condition does not always exist in
composites. It is not unusual to observe small but distinct cracks in the matrix
well before extensive, large-scale cracking occurs.
For fiber pullout, on the other hand, the above inequality will be reversed.
1 2 crt'u
(9.6)
d 4 't,
The aspect ratio (~/d) corresponding to the equality in Eqs. (9.6) is called the
critical aspect ratio. One can write
(9.7a)
(9.7b)
Once the inequality condition for fiber pullout is satisfied, the fiber will be
pulled. The work done in this pullout process is discussed in the next section.
Consider the situation depicted in Fig. 9.12. Let us say that the fiber breaks at
some distance k below the main crack plane, such that 0 < k < ~/2. There is
local debonding of the fiber/matrix interface. When the fiber is pulled out of the
matrix, an interfacial frictional shear stress, 'ti will be generated. Recall that in
our simple analysis here, we have assumed that this shear stress, 'ti, resisting the
fiber sliding is a constant. More complex treatments involving a Coulomb
friction law governing the fiber sliding resistances (Shetty, 1988; Gao, et aI.,
1988) and treatments taking into account residual stresses as well as a Coulomb
friction law are available in the literature (Cox, 1990; Hutchinson and Jensen,
1990; Kerans and Parthasarathy, 1991). We derive a simple expression for the
work done in the process of fiber pullout.
314 INTERFACE MECHANICS AND TOUGHNESS
Let the fiber be pulled out a distance x. The interfacial shear force opposing this
motion of fiber is 'rjnd(k-x). If the fiber is pulled out through a small distance
dx, then the work done by the interfacial shear force is 'rjnd(k-x)dx. The total
work done in the fiber-pullout process over the length k is obtained by
integrating.
k k k
f'rind(k-x)dx 'r;Ted fkdx - fXdx
o o 0
(9.8)
The length of the fiber pulled out will vary between 0 and Q/2. The average
work done in the fiber pullout process is
0<
Wfp _
- (-1 ) JT,xd k' dk
Q/2 2
r
c 0
~ (~,:2) [T,;d ~3
2 2
'rind Qc 'rind Qc
--- (9.9)
6 4 24
315 INTERFACE MECHANICS AND TOUGHNESS
Matrix
}I'
Pull
Fig. 9.11 A fiber of radius r embedded in a matrix and being pulled out. The
longitudinal tensile force on the fiber generates shear at the fiber/matrix interface.
-T
Principal crack
1
Fig. 9.12 Fiber pullout
This equation assumes that all of the broken fibers are pulled out. Experimental
observations, however, show that only fibers with broken ends within a distance
V2 from the principal fracture plane undergo the pullout process. Thus, we
should expect a fraction (e,/O of fibers to pullout. Therefore, the average work
done per fiber in fiber pullout can be written as
Wjp (9.10)
e 24
316 INTERFACE MECHANICS AND TOUGHNESS
Ifthere are N fibers in the area of the crack, A, then we can write the following
for the fiber volume fraction, V f
(9.11 )
Pullout work done per unit area of the crack can then be obtained from Eqs.
(9.8) and (9.10) as
Rearranging,
2
- N 1td 2 T;~c
n
W=---
A 4 6d
2
T·~c
V-'- (9.12)
f 6d
It follows from Eq. (9.12) that in order to maximize the fiber pullout work, we
should have a large fiber volume fraction, high interfacial strength, high ~c, and
small fiber diameter. Here we have tacitly assumed that fiber/matrix debonding
has already occurred and that the fiber pullout work is due to sliding at the
interface. Recalling from Eq. (9.6) that for a fiber being loaded from both ends,
the critical aspect ratio is given by
(9.13)
317 INTERFACE MECHANICS AND TOUGHNESS
Using Eq. (9.12), we can eliminate the critical length, and write the pullout
contribution to toughness per unit area of the crack as
d2
-.2. (J/u
2
W = V.
f 6d 4 '[2
(9.14)
Finally, p. is the critical axial load on the fiber at which the Poisson contraction
balances the residual normal stress and the roughness effects such that no
frictional resistance remains to sliding. p. is related to radial stress, O'r at the
interface
(9.16)
Parthasarathy et ai. (1991) showed how one can extract the key interface
parameters from their model and a single fiber pullout test. It should be
recognized, however, that the large scatter in the data poses problems in
obtaining the interface parameters with high reliability (Jero, et aI., 1992).
Mumm and Faber (1992) used a special single fiber pullout experiment on SiC
monofilament (SC S-6)/soda-borosilicate glass composite system to evaluate this
progressive debonding model. Their single fiber sample was notched to give a
reduced cross section at the midpoint of the sample. They successfully used the
Kerans and Parthasarathy analysis to obtain interfacial parameters from the
progressive debonding portion of load-displacement curves. The calculated
interfacial parameters were consistent with independent measurements
available in literature.
1
!l.K =-
wr 2
For a very brittle matrix, we can take Km '" 0, and rewrite .1Kwr as
Vwr Ee Gm
(9.17)
6(1 -v 2) Ew G;
It should be clear to the reader by now that a proper control of the interface
characteristics is the key to obtain a tough ceramic composite. The components
of the composite must be chemically stable at the processing temperatures as
well as at the service temperatures. Any chemical interaction between the
matrix and the reinforcement during processing will influence the interface
bond strength, and consequently the ability to debond at the fiber/matrix
interface and cause fiber pullout.
(i) proper selection of the fiber and matrix such that they form a
thermodynamically stable interface during processing and service.
Interfacial coating provides an easy and effective way to tailor the interface
region. The coating on fiber could also protect the fiber from mechanical
degradation during processing and handling. Ideally, one would like to have an
interfacial coating that does not interact with either the fiber or the matrix and
is stable at the fabrication temperature and time as well as during the service
conditions of time and temperature. Clearly, coating thickness will be an
important parameter. Coating thicknesses in the range of 0.1 to 1 Ilm are
commonly used.
321 INTERFACE MECHANICS AND TOUGHNESS
(a)
SiC
Whisker
(b)
Fig. 9.13 (a) Schematic of crack bridging by whiskers. (b) A silicon carbide whisker
bridging a crack in alumina matrix. (Courtesy of P.F. Becher). (c) Toughness
enhancenment obtained by silicon carbide whisker reinforcement of different matrix
materials (after Becher, 1990).
322 INTERFACE MECHANICS AND TOUGHNESS
S
?d ....0 .iKfb+.iKPO
p.,
~
.,<Ii' 00
=
<U
.c:
01)
;::!
0 -.a
1-<
....
<U
....<.>
;::!
..,.
....d
.... .iKfb
.~ N
'"
0
Q.
S
0 0
U
0 10 20 30 40
SiC Whisker Content, Volume %
(c)
A number of researchers (Bender et aI., 1984, 1986; Lowden and Stinton, 1988;
Brun and Singh, 1989; Ha and Chawla, 1993; Ha et aI., 1993; Chawla et aI.,
1993, 1997,2000; Lewis, 2001; Lowden, 1991; Morgan and Marshall, 1993;
Naslain et aI., 1991) have pursued this route to obtain tough ceramic
composites. Bender et ai. (1984) observed that BN-coated (Nicalon) SiC fibers
in a zirconia based matrix (Zr0 2 + 50 mol% Si02 and zr02 + 50 mol% Ti0 2)
resulted in substantial improvements in strength and toughness of the
composites. The BN coating prevents the strong bonding between the silica
reaction layer in the matrix and the fiber. Composites made with uncoated
fibers showed a very brittle behavior. Improvement in room temperature
toughness of a Nextel 480 (essentially mullite) in a glass matrix by
incorporating a BN coating at the interface was observed by Vaidya et
aI.(1992). Figure 9.16 (a,b,c) taken from their work shows the difference in the
fracture behavior of the composites containing uncoated Nextel480 (mullite)
324 INTERFACE MECHANICS AND TOUGHNESS
Interphase Incorporation
Characterization of Interlaces
------"---
I'ropeliy I:valuation of the
Lngineered Composite
SiC COATING
(a)
Fig. 9.15 Examples of coated fibers: (a) carbon coated (Nicalon) fiber (courtesy ofR.
Lowden) (b) tin dioxide (white layer) coated alumina (Saphikon) fiber, slightly oval in
shape.
326 INTERFACE MECHANICS AND TOUGHNESS
(b)
monoclinic and hexagonal polymorphs of the celsian matrix and free silicon in
the SiC region. It was proposed that the following reaction occurred at the
interface:
BN + SiC -» Si + BN-C
It should be pointed out that what is important for high temperature use is the
stability of the coating at the service temperature. In this regard, carbon
coatings in air and at moderately high temperatures will not be suitable. Boron
nitride coating also tends to oxidize in air at about 600°C or so, i.e., at
temperatures slightly above those for carbon. Silicon carbide coatings tend to
form silica at temperatures above 1200 °C. Lowden (1991) studied the effect of
different thicknesses of graphitic carbon interfacial coating in NicaloniSiC
composites. Carbon coatings were deposited isothermally from an argon and
327 INTERFACE MECHANICS AND TOUGHNESS
(a)
(b)
(c)
2000
0
-4 -2 0 2 4 6
300
250
200
...... -
~
0 150
....._-.\
«
0
-'
OXIDIZED
...
"'\ TO 0.13
100
'.... -...... .-........
'---
50
0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.Q7
DISPLACEMENT (em)
Fig. 9.20 Flexure load-displacment curves for Nicalon/SiC composites with varying
carbon thickness. The uncoated fiber composite showed poor strength and a brittle
fracture devoid of any fiber pullout. The oxidation treatment resulted in low strength.
Incorporation of carbon coating resulted in a graceful failure (Lowden, 1991).
332 INTERFACE MECHANICS AND TOUGHNESS
.
500
0
0 2 4 8 10 12
Fig. 9.21 The inverse relationship between interfacial shear stress and the carbon
coating thickness for the NicalonlSiC composites (Lowden, 1991).
(a)
(b)
Fig. 9.22 (a) A longitudinal section of the composite showing the rough nature of the
PRD-166 fiberlSn0 2 interface; (b) A longitudinal section of the composite showing the
relatively smooth Saphikon fiberlSn0 2 interface. F, C, and M indicate fiber, coating,
and matrix respectively. SEM.
335 INTERFACE MECHANICS AND TOUGHNESS
Fig. 9.23 A schematic of the interface roughness (periodic) with an amplitude A. f and
m denote the fiber and the matrix, respectively.
TA
..... l
where
•
L = length of roughness curve
f(x) =roughness curve
Rmax = Maximum Roughness
difference between the highest and lowest point
Rz =Ten-Point Average
Fig. 9.24 Root mean square and peak to valley roughness amplitude, A, of the fiber.
336 INTERFACE MECHANICS AND TOUGHNESS
The values obtained for both PRD-166 fiber/tin dioxide/glass matrix and
Saphikonltin dioxide/glass matrix composites are shown in Table 9.5. It is
interesting to note that in PRD-166 fiber/tin dioxide/glass matrix composites the
compressive radial strain induced due to fiber roughness is more than an order
of magnitude greater than the thermal strain. In Saphikon/tin dioxide/glass
matrix composites, the roughness induced strain is only three times greater than
the thermal strain. This is advantageous since some bonding is required to
maintain the integrity of the composite.
Fracture surface features help elucidate the effects of interfacial roughness and
thermal stresses in these composites (Venkatesh et aI., 1992; Chawla et aI.,
1993). The fracture surface of a PRD-166 fiber/glass composite, Fig.9.26,
shows a characteristically flat, brittle fracture. Figure 9.27 shows the fracture
surface ofPRD-166 fiber/tin dioxide/glass matrix composites. Note the partial
debonding of the fibers from the matrix and the characteristic rough fiber
surface that results in strong mechanical bonding, see Fig. 9.27. Figure 9.28
(a,b) shows the fracture surface of Saphikonltin dioxide/glass matrix
composites. Note the fiber/coating debonding and the neat fiber pullout at the
alumina/tin dioxide interface. This is as expected since alumina and tin dioxide
show no solid solubility which precludes the chemical bonding but more
conspicuously the smooth interface results in a weak mechanical bonding.
Major toughening mechanisms in case system include crack deflection,
fiber/matrix debonding, and fiber pullout. Similar effects of interfacial
roughness were verified in Saphikon fiber/ lanthanum phosphate/alumina
matrix composites (Chawla et aI., 2000). We digress here a bit to mention the
usefulness of atomic force microscope in monitoring the fiber surface
morphology and to obtain the roughness profiles (Chawla et aI., 1993; Chawla
et aI., 1995). Very briefly, in an AFM, the probe tip in the form of a
microcantilever is placed a few nm from the sample surface. The cantilever tip
is deflected by the interatomic force between the atoms on the tip and the atoms
on the sample surface. A laser beam measures the deflection of the tip. The
deflection of the cantilever tip monitored dn used in a feedback loop. An
example of the surface profile ofNicalon fiber is shown in Fig. 9.29.
Saphikon
1 .. m
Fig. 9.25 SEM micrographs and surface roughness profiles of PRD-166, Nextel 610,
and Saphikon fibers.
338 INTERFACE MECHANICS AND TOUGHNESS
Fig. 9.26 Fracture surface of uncoated PRD-166 fiber/glass composite showing a flat,
brittle fracture.
Table 9.4 Roughness amplitude (A), fiber radius (r), and the ratio (Air) for PRD-166
and Saphikon fibers.
PRD-166 Sapphire
A, pm 0.185 0.0077
r, /.lm 10 40
Air 0.0185 0.0002
Table 9.5 Comparison of roughness induced strain (Air) and thermal strain (~a ~T)
in PRD-166 fiberlSn0 2 and SaphikonlSn02 interface
(a)
(b)
Fig. 9.28. Fracture surfaces ofSaphikon/SnO/glass composite. Note the fiber/coating
debonding and extensive fiber pullout at the fiber/coating interface.
341 INTERFACE MECHANICS AND TOUGHNESS
200
(nm) 100
o o
2
(jim)
Fig. 9.29 Surface texture of a Nicalon fiber. AFM. (Courtesy ofN. Chawla)
342 INTERFACE MECHANICS AND TOUGHNESS
120
100
l PRD-166 FIBER
e
til
80
til
~ 60
1;;
....l
~ 40
Q
;,2 20 COATING
o 5 ~
(a)
100
..•••••........••••••....
~ 0~
c..
~
;; -100 ~
til
~
PRD-166 FIBER -- ~
MATRIX
c=::
ti; -200 ~
f-<
Z
~ -300 "-
(j
Z COATING- ~
~ -400 ~
-500
II I I
o 5 10 15 20 25
Fig. 9.30 Distribution of thermal stresses in alumina (PRD-166) fiber/glass matrix and
alumina (PRD-166) fiber/SnO/glass matrix composites, for a AT = - 450°C. (a) Radial
stresses (b) Tangential stresses (c) Axial stresses.
343 INTERFACE MECHANICS AND TOUGHNESS
400
6
<Zl
............... _-- ...........
<Zl
~
40
~
f-;
<Zl
....l
<t:
......
Q 20
<
~ COATING
o 20 100
-- ---
---
0-
(;j'
Cl... .
6
CI)
-100 -
CI)
~ SAPHIKON FIBER MATRIX
~
f0-
CI)
....
...... -200 -
-<
fo-
Z
~
"-<
Z
fo-
-300 - COATING---t
-400
o 20 40 60 80 100
300~~~~~~~~~n---------------------~
200-
(;j' 100 ~
Cl...
6 SAPHIKON FIBER MATRIX
CI)
CI)
Or-
~
~ -1oo~
CI)
....
-< -200r-
~ COATING---t
-3001-
_4ooL-______~I--------~I-=~----~I~-------~I------~
o 20 40 60 80 100
DISTANCE FROM THE CENTER OF THE FIBER (/lm)
(c)
These coatings also offer advantages difficult to obtain with other interphase
types described above. One unique difference is the ability to use a coating with
strong bonding to the fiber and matrix, and thus retain overall good composite
strength, but possessing the ability to convert to localized weak bonding in front
of a crack tip. An example of such a coating is enstatite (MgSi03) (Zhu and
Kriven, 1996). The polymorphs of enstatite and the conditions for
transformation between them are shown in Fig. 9.33. This material undergoes
a shear-induced proto to clino destructive transformation at 96 MPa
accompanied by a molar unit cell volume shrinkage of 5.5% and
crystallographic unit cell shape change of 18.3 0 • Such composites were
fabricated by a co-extrusion method to produce multi-filament arrangements of
fibrous titania (Ti0 2) surrounded by enstatite and then polycrystalline titania.
Three-point flexure tests showed a work of fracture of 1000 J/m2 for the work
of fracture associated with transformation-weakened interfaces. Fractography
revealed pullout of fibrous regions as well as significant fracture along the
transformation weakened interface regions in the annealed specimens.
Magnetoplumbites are yet another class of materials that may provide some
suitable coatings (Cinibulk, 1994). As mentioned earlier, a layered
microstructure with weak bond between layers allows for easy cleavage.
Common examples are, of course, pyrolytic carbon and hexagonal boron
carbide. Compounds having p-alumina or magnetoplumbite structure have such
an anisotropic layered oxide structure. These layered oxides can be regarded as
346 INTERFACE MECHANICS AND TOUGHNESS
MATRIX
IH li'l~
TRANSFORMABLE
OXIDE INTERPHASE
CRACK
Fig. 9.32 Under shear stress from an approaching crack tip, a transformation is induced
in the coating, causing a volume or shape change at the fiber/matrix interface (after
Kriven, 1995)
Protoenltatite PE
(orthorhombic>
~ 865"C
fait quenching
4V .. ·5.591.
104~
slow cooling
.~ clinoenstatite C E
(monoclinic)
shearing under
high pressures and
temperatures {2:SOOoC>
Orthoenstatite 0 E
(orthorhombic)
Fig. 9.33 Polymorphs of enstatite and the conditions for transformation between them
(after Kriven, 1995).
347 INTERFACE MECHANICS AND TOUGHNESS
Fiber coatings of phyllosilicates or micas have also been tried. They are,
however, unstable above 700°C because the hydroxyl (OH-) groups in their
crystal lattice decompose. The thermal stability of phyllosilicates can be
improved by eliminating the OH- groups. A modification of phyllosilicates,
called phyllosiloxides, having a mica like layered structure but without the OH-
groups has been suggested. Schematic descriptions of mica and a phyllosiloxide
are shown in Figure 9.35.
Yet another method to create multiple weak layers at the fiber/matrix interface
involves interfacial reactions during processing. Unlike the earlier described
interphase materials, in which reactions between the interphase and fiber or
matrix are not desirable, this method actually utilizes reactions to form
multiple, weak, and stable interface layers between the fiber and matrix. An
348 INTERFACE MECHANICS AND TOUGHNESS
Fig. 9.34 Indentations in bulk KC~NbPlo showing cracks parallel to the basal planes.
(Courtesy of S. Sambasivan)
602- 602 -
2K+ 2K+
602 - 602 -
3Si4 + + lA1 3+ 4Si 4+
.........................
402- + 20H- 602 -
------------------ 6Mg 2+
------------------ 4Mg2+ + 2Ae+
40 - + 20H-
2 60 -
2
602- 602-
KMg3(ShAl)OIO(OH)Z KMgzAlSi40 12
(a) (b)
Fig. 9.35 Structure of (a) micaphlogopite and (b) related phyllosiloxide (after Reig et
aI., 1995).
349 INTERFACE MECHANICS AND TOUGHNESS
Porosity has also been used in the matrix material of ceramic matrix composites
to improve damage tolerance (Levi et aI., 1999). The notch insensitivity ofthese
porous composites is claimed to be similar to that of conventional ceramic
matrix composites, but they have a failure strain greater than that of
conventional monolithic ceramics. Processing is easier because a high density
is not required and no interphase is needed between the fiber and matrix. Such
composites with a porous matrix have a failure mode different form
conventional ceramic matrix composites. Matrix cracking occurs throughout the
entire loading range with the tensile stress/strain behavior almost linear to
failure.
REFERENCES
Aksay, LA. and J.A. Pask (1975), J. Amer. Ceram. Soc., 58, 507.
Aveston, J., G.A. Cooper, and A. Kelly (1973) in The Properties o/Fibre Composites,
Conference Proceedings, National Physical Laboratory, IPC Sci. & Tech. Press,
Guildford, Surrey, p. 15.
352 INTERFACE MECHANICS AND TOUGHNESS
Gouadec, G., P. Colomban, and N.P. Bansal, (2001) J Amer. Ceram. Soc., 84,1121.
Green, D.J., R.H.I. Hannink, and M.V. Swain (1989) Transformation Toughening of
Ceramics, CRC Press, Boca Raton, FL .
Gupta, V. (1991) MRS Bulletin, XVI-4, 39.
Gupta, V., J. Yuan, and D. Martinez (1993) J Amer. Ceram. Soc., 76, 305.
Ha, J-S. and K.K. Chawla (1993) J Mater. Sci. Letters, 12,84.
Ha, J.-S., K.K. Chawla, and R. Engdahl (1992) Mater. Sci. and Eng., A161, 303.
He, M. Y. and J. W. Hutchinson (1989) J App. Mech., 56, 270.
Hutchinson, J.W. and H.M. Jensen (1990) Mech. Mater., 9,139.
Janney, M. (1987) Bull. Amer. Ceram. Soc., 66,322.
Jero, P.D., Parthasarathy, T.A, and Kerans, R.J. (1992) Ceram. Eng. Sci. Proc., 13, 54.
Kerans, R.J. and T.A. Parthasarathy (1991) J Amer. Ceram. Soc., 74,1585.
Kanka, B. And Schneider (2000) J Eur. Ceram. Soc., 20, 619.
Kriven.W.M. (1995) J Physique IV 5, C8-101.
Krstic, V.V., P.S. Nicholson, andR.G. Hoagland (1981)J Amer. Ceram. Soc., 64,499.
Lange, F.F. (1970) Phil. Mag., 22, 983.
Levi, C.G., F.W. Zok, J.-y' Yang. M. Mattoni, and J.P.A. L5fvander (1999) Z.
Metallkunde, 90, 12.
Lewis, M.H. (2001) in High Temperature Ceramic Matrix Composites, Wiley-VCH,
Weinheim, Germany, p. 136
Lowden, R.A. (1991) inAdv. Camp. Mater., vol. 19, Amer. Ceram. Soc., Westerville,
OH, p. 619.
Lowden, R.A. and D.P. Stinton (1988) Ceram. Eng. Sci. Proc., 9, 705.
Maheshwari, A., K.K. Chawla, and Michalske, T.A. (1989) Mater. Sci. & Eng., AI07,
269.
Majumdar, A.J (1970)Proc. Roy. Soc. Land., A319, 69.
Marshall, D.B., B.N. Cox, and A.G. Evans (1985) Acta Met., 33, 2013.
Mendelson, M.1. and M.E. Fine (1974) J A mer. Ceram. Soc., 57, 154.
Michalske, T.A. and J. Hellmann (1988) J Amer. Ceram. Soc., 71, 725.
Morgan, P.E.D. and D.B. Marshall, (1993) Mater. Sci. and Eng., A162, 15.
Mumm, D.R. and K.T. Faber (July-Aug., 1992) Ceram. Eng. & Sci. Proc., 13,218.
Naslain, R., O. Dugne, A. Guette, C.R. Sevely, J. Brosse, J.-P. Rocher, and J. Cotteret
(1991)'/' Amer. Ceram. Soc., 74, 2482.
Parthasarathy, T.A., P.D. Jero, and R.I. Kerans (1991) Scripta Met., 25, 2457.
Ramualdi, J.P. and G.B. Baston (1963) Proc. ASCE, J Eng. Mech. Dn., 89,147.
Reig, P., G. Demazeau, andR. Naslain (1995) Eur. J SolidStateInorg. Chem., 32, 495.
Ruhle, M. and A.G. Evans (1988) Mater. Sci. and Eng., AI07, 187.
Rose, J.H., J.R Smith, and J. Ferrante (1983) Phys. Rev. B, 28, 1835.
Rose, J.H., J.R Smith, F. Guinea, and J. Ferrante (1984) Phys. Rev. B, 29, 2963.
Sambell, R.A.J., et. al (1972) J Mater. Sci., 7, 676.
Schoenlein, L.H., R.H. Jones, C.H Henager, Jr., C.R. Schilling, and F.C. Gac (1988)
Proc.: Materials Res. Soc. Symp., Spring 1988, April 5-9, Reno, NV.
Sambasivan, S., T. A. Parthasarathy, F. J. Scheltens, and R. J. Kerans (1994) Ceram.
Eng. Sci. Proc., 15,972.
354 INTERFACE MECHANICS AND TOUGHNESS
SUGGESTED READING
MACROMECHANICS
S22 S23 0 0 0
S33 0 0 0
[Sii] = (10.1)
S44 0 0
S55 0
S66
A lamina being a thin orthotropic material, we can take the through thickness
components to be zero. Then, we can write the stiffness matrix for an
orthotropic lamina by eliminating the terms involving the z- or 3-axis
(10.2)
Using this [Sjj] we can write Hooke's law for a thin orthotropic lamina, with
natural and geometric axes coinciding, as follows
£) S)) S)2 0 a)
£2 = S)2 S22 0 a2 (10.3)
£6 0 0 S66 a3
357 MACROMECHANICS
0'2 = Q 12 Q 22 0 E2 (l0.4)
0'6 0 0 Q66 E6
Here, we have used the symbol Qij rather than Cij for thin materials because of
convention. Qij are called reduced stiffnesses. Qij and Sij are related as follows
S22
QIl = 2
SII S 22 - S 12
S12
Q12 = 2
S11 S 22 - S 12
(10.5)
SII
Q22 = 2
S11 S22 - S 12
1
Q66 =S
66
(ij = 1,2,6)
Equations (10.3) and (10.4), showing terms with indices 16 and 26 to be zero,
represent a case of special orthotropy when the principal material axes of
symmetry coincide with the geometric axes, i.e., the fiber direction (1) and the
direction transverse to it (2) coincide with the principal loading directions. If
the material symmetry axes and the geometric axes do not coincide, then we
have the case of general orthotropy, a fully populated elastic constant matrix,
i.e., 16 and 26 terms are nonzero. These two cases are illustrated in Fig. 10. 1.
The constitutive equation in terms of general orthotropy can be written as
O'x Q 11 Q 12 Q16 Ex
L x,1
(a)
y
2
(b)
Fig. 10.1 Two cases of orthotropy: (a) special orthotropy when the material symmetry
axes (1-2) and the geometric axes (x-y) do not coincide, (b) general orthotropy when
the principal material axes of symmetry (1-2) coincide with the geometric or principal
loading directions (x-y).
where the Qij matrix is called the transformed reduced stiffness matrix because
it is obtained by transforming Qij (specially orthotropic) to Qij (generally
orthotropic), In order to obtain Qij from Qij' the transformation of axes can be
performed as shown in Fig. 10.2. This shows a unidirectional lamina with axes
in 1-2 system not coinciding in the x-y system, Ifwe know the properties in the
1-2 system, we can determine them in the x-y system or vice versa. This
transformation of axes is carried out easily in the matrix form. A
counterclockwise rotation is taken to be positive. The transformation equation
for stress, in matrix form, is
(10.7)
359 MACROMECHANICS
2
y !
X
Fig. 10.2 Transformation of axes to obtain Qij from Qij (counterclockwise rotation is
positive).
(l0.8)
es
where [TJa and [TJ. are the transformation matrices for stress and strain
transformations, respectively, and are given by
m2 n2 2mn
n2 m2 -2mn (10.9)
-mn mn m 2 -n 2
mn
-mn (10.10)
This method of using different transformation matrices for stress and strain
transformations is convenient because it avoids the need of putting the factor
112 before the engineering shear strains to convert them to tensorial strain
components suitable for transformation.
m2 n2 -2mn
[Ttl = n2 m2 2mn (10.11)
where m = cose and n = sine. Stresses in the x-y system are then
(10.12)
(10.13)
(10.14)
where
361 MACROMECHANICS
(10.15)
[Q] is the stiffness matrix for a generally orthotropic lamina whose components
in expanded form are written as follows (m = cose, n = sine):
The reader should note that although Qij is a completely filled matrix, only four
of its components are independent: QI6 and Q26 are linear combinations of the
four, Qij.
a 1 =amx
2
(l0.17a)
a 2 =anx
2 (l0.17b)
a 6 = -axmn (10.17c)
(10.18a)
(10.18b)
(10.18c)
e
1
= ax (m2
E
- v ~l
1 E (10.19a)
1 1
e=a(~-vm21
2 E 2 E
x
(l0.19b)
2 2
ax mn (10.19c)
e =---
6 G6
where we have used the relationship v/E t = viE2 derived in the next section.
Since we have the strain transformation given by Eq. (10.8), we can write the
inverse ofEq. (10.8) as
363 MACROMECHANICS
ex = m 2e 1 + n 2a 2 - mna6
(10.20a)
ay = n 2a 1 + m 2a2 + mna6 f
(10.20b)
a,I' = 2(a 1 - a2)mn + e6(m 2 - n 2) (10.20c)
(10.21a)
(10.21b)
(10.22)
(10.23)
Or
(10.24)
Taking 0'. to be the only nonzero stress component, and noting that the in-plane
shear strain 8. = O'iGs' and applying Hooke's law, we obtain the following
relationship for in-plane shear modulus as a function of angle e.
1 + v2 - _I ] (10.26)
E2 G6
and
365 MACROMECHANICS
e,I'
mx =--
and
Figure 10.3 shows, schematically, the variations of Ex, Ey, G., VX (= VXy) and Vy
(= vyx), and mx and my with fiber orientation 9. Various researchers have
examined laminated composites made by hot pressing of plies of Nicalon
fabric/glass matrix (Prewo and Brennan, 1982) or monofilament silicon carbide
fiber (SCS-6 type CVD fiber) in a reaction bonded silicon nitride (RBSN)
matrix (Bhatt, 1991; Bhatt and Phillips, 1990). Bhatt and Phillips (1990), in
particular, used the laminate theory and modified constitutive equations to
account for weak bonding in their SiC/RBSN composites, to predict elastic
properties of unidirectional, cross-ply, and angle-ply composites.
Consider the thin lamina shown in Fig. 10.4 with its natural axes coinciding
with the geometric axes. The conventional engineering constants in this case are
Young's moduli in directions 1, EI and direction 2, E 2 ; the principal shear
modulus, G6 ; and the principal Poisson ratio, VI. In an isotropic material, there
is only one Poisson ratio, because the contractions in the two principal
transverse directions are equal. In the case of an orthotropic lamina, there will
366 MACROMECHANICS
o 45 90
(a)
Ply angle, (10
o 45 90
Ply angle, (r
(b)
Fig. 10.3 Variation (schematic) of various elastic constants with fiber orientation e for
a unidirectional fiber composite. (a) Ex and Ey, (b) G" (c) Vx and vy' and (d) mx and my.
367 MACROMECHANICS
o
Ply angle, 0°
(c)
t
E
o 90
Ply angle, 0°
(d)
be two Poisson's ratios, as shown schematically in Fig. 10.5. When the lamina
is strained in direction I, we get equal the Poisson's ratio Vb equal to -Ez!E 1.
When the lamina is strained in direction 2, we get the Poisson's ratio v2, equal
to -E/E2•
0'"6
Ii;
- f- ;\:
):
l - F--
- r- r--- 0'"2
'"
t--
~ \
- f-
:f, .:1:
..
'--- 2
- i1 ~
,.
r- :~
t-- i---
-, l-
\~
f;
.~-;,
;i - ~
~~:
Fig. 10.4 A thin orthotropic lamina with its natural axes coinciding with the geometric
axes.
369 MACROMECHANICS
2
Ql1 Q22 - QI2
EI (10.27)
Q22
c2 Q I2
VI -- (10.28)
f,1 Q22
Now let us repeat the above procedure with (J2 as the only nonzero strain
component in Eq.(lO.14). Then
2
(J2 Q ll Q22 - QI2
f,2 Qll (10.29)
(10.30)
~
---[d---------
....... .
--- -----------------
.....
'L t.
[]
1
·
:
·:
:.
:
I·· I..
:---------!
Similarly, if we consider that (j6 is the only nonzero component, we can get
(j6
G6 = - = Q66
E6 (10.31)
VI E2 = v2 E I (10.32)
EI VI
(10.33)
E2 v2
Similarly, it can also be shown that the relationships between compliances and
engineering constants are as follows:
1
S =-
II E
I
// [0]8
// [0/90/+ 45/ - 45],
Fig. 10.6 Two stacking sequences, an eight-ply laminate made of eight unidirectionally
aligned (0°) and stacked in parallel mode, denoted by [0]8 and another eight-ply laminate
with plies stacked at 0°, 90°, +45°, and -45°, denoted by [0/90/+45/-45],.
in the top half and the bottom half being the mirror image of the top, denoted
by [0/90/+45/-45]., where the subscript s indicates that the stacking is
symmetrical with respect to the geometrical midplane of the laminate (z = 0).
OU ou
- +
OV
Ex Exy
Ox oy Ox
Ov Ou ()W
Ey - Exz - + - (10.34)
oy oz ()x
()W ()V OW
Ez Eyz - +-
oz oz ()y
For i = j, Eij are the normal strain components while, for i *' j, Eij represent the
engineering shear strain components equal to twice the tensorial shear
components. According to Kirchhoffs assumption, the in-plane displacements
being linear functions of the thickness coordinate, z, we write
T /y,v
~x,u
+h12
- h/2
f
midplane (z=O)
Fig. 10.7 A laminated composite. The geometric axes, x-y, of the laminate are the in-
plane axes while z is the thickness direction. z = 0 is the geometric midplane of the
laminate and h is the thickness of the laminated composite. Also shown are the
displacements U,V, and w along x,y, and z axes, respectively.
373 MACROMECHANICS
U = Uo(X,y) + z F) (x,y)
(10.35)
()W
€yz = F2 (x ,y) +- =0
()y
()W
and F2 (x,y) =- (10.36)
()y
The strain in the thickness direction, €z' is negligible, thus we can write
w =w(x,y)
That is, the vertical displacement of any point does not change in the thickness
direction.
() U = () U 0 _ Z ()2 W =€ 0 + zK (10.37a)
()X ()X ()X2 x x
()V
€ =- (10.37b)
y ()y
374 MACROMECHANICS
ou +-
ov OV" 02w
- 2z - - = s " + zK
oy ox oy ox oxoy xy xy
sake of consistency, we shall denote Sxy by Ss and K,.y by Ks, we can rewrite the
expression for Sxy above as
(1O.37c)
The components sox, SOy, and so, are the midplane strains while Kx, Ky, and K,
are the plate curvatures. We can represent these quantities in a compact form
as follows
(10.38)
and
o-w
--
ox 2
Kx
02w
Ky -- (10.39)
oy2
1\\. 202W
--
~ ~
(10040)
Consider again the laminated composite shown in Fig. 10.7. The geometric
axes, x-y, of the laminate are the in-plane axes while z is the thickness
direction. z = 0 is the geometric midplane of the laminate and h is the thickness
of the laminated composite. Also shown are the displacements u, v, and w along
x,y, and z axes, respectively. The following assumptions are involved:
Then, we can write, for the kth layer, the following constitutive relation
(10041 )
(10042)
(10043)
376 MACROMECHANICS
(10.44)
As can be seen from the above expressions, these stress resultants have the
dimensions of force per unit length and are positive in the same direction as the
corresponding stress components. These resultants give the total force per unit
377 MACROMECHANICS
(10.45)
h
2
~y - M.. fa .. zdz
h
'2
This system of three stress resultants (Eq. 10.44) and three moment resultants
(Eq. 10.45) is statically equivalent to actual stress distribution through the
thickness of the composite laminate.
-v
x M,
x N,
From Eq.(1 0.43), we can write for the stress resultants as a summation over the
n plies
Q I2 QI6 Kx
Q22 Q26 Ky z dz
Q26 Q66 Ks
Since [SO] and [K] are not functions of z and in a given ply, and [0] is not a
function of z, these can be taken outside the integral sign to simplify the above
expression to
~ t::
1111 1112 11 16 Kx
= (10.47)
~. ~16
where
n
Aij = L (Qij)k(hk -hk_l )
k=1
(10.48)
and
379 MACROMECHANICS
(10.49)
where
The stress and moment resultants with respect to midplane strains and plate
curvatures can be put in a compact form
where
and Qij denote the plane stress reduced stiffnesses for the individual plies and
h is the total thickness of the laminate.
A quick examination of Eqs. (10.47) through (10.51) will show that a tensile
stress resultant is a function of the midplane tensile strains (COx and COy),
midplane shear (eo.), bending curvatures (I<.,. and Ky), and twisting (K.). This is
a much more complex situation than that observed in a homogeneous plate
where tensile loads result in only tensile strains. In a laminated plate we have
coupling between tensile and shear, tensile and bending, and tensile and
twisting effects. Specifically, the terms A 16 and A26 bring in the tension-shear
coupling while the terms B 16 and B26 represent the tension-twisting coupling.
The D 16 and D26 terms in a similar expression for Mx represent flexure-twisting
coupling. Under certain special conditions, the stress and moment resultants
become uncoupled.
Figure 10.9 shows that when the deformation is parallel to the material axes in
orthotropic material, then there is no coupling. This will be the case when the
fibers at 0° or 90° degrees to the stress axis. When the fibers are at an angle to
the stress axis, Fig. 10.9, the coupling terms will come into play and even when
a normal stress is applied, the response field will include normal, shear, and
bending strains. We can also examine the expressions for [A], [B], and [D]
terms and arrive at some conclusions. For example, the Aij terms are the sum of
ply Oij times the ply thickness (Eq. 10.40). Thus, Aij will be zero if the positive
contribution of some laminae is canceled by the negative contribution of others.
Now, the Qij terms of a ply are derived from orthotropic stiffness and, due to the
form of transformation equations (Eqs. 10.25), the components 011' 012' 022'
and 066 are always positive, so the components Aw A 12, A22 , and A66 are always
positive. Also, 016 and 026 are zero for 9 = 0 0 and 90 0 , and are positive or
negative for 0 0 < 9 < 90 0 • The components 016 and 026 are odd functions of9.
This means that
016(+9) = -016(-9)
and
026(+9) + -026(-9)
Thus, iffor each +9 ply, we have another identical ply of the same thickness at
-9, then we shall have what is called a specially orthotropic laminate with
respect to in-plane stresses and strains, i.e., A 16 = A26 = O. Relative position of
such plies in the stacking sequence is immaterial.
The Bij terms are sums of terms involving Oij and differences of the square of
the z terms of top (hk ) and bottom (hk_1) of each ply. Thus, Bij terms are even
381 MACROMECHANICS
functions ofhk> which means that they will be zero if the laminate composite is
symmetrical with respect to thickness. In other words, Bij will be zero if we
have for each ply above the midplane, a ply identical in properties and
orientation and at an equal distance below the midplane. Such a laminate is
called a symmetric laminate and will have Bjj identically equal to zero. The
absence of bending-stretching coupling in symmetric laminates eliminates the
problem of warping encountered in non-symmetric laminates when in-plane
forces are induced by thermal mismatch during processing. Symmetric
laminates will only experience thermal strains at the mid plane but no flexure.
The reader should observe that the origin of [B] matrix lies not in the intrinsic
orthotropy of laminae, but in the heterogeneous (non-symmetric) stacking
sequence of the plies. Thus, if were to make a two-ply composite consisting of
different isotropic materials such as alumina and glass sheets, such laminate
will have a nonzero [B].
---
~
~--
Fig. 10.9 The coupling phenomenon in an off-axis lamina. The coupling is zero in an
on-axis lamina.
382 MACROMECHANICS
The bending matrix Dij terms are defined in terms of Qij and the difference
between hk and hk_,. The geometrical contribution (h3k - h3k-I) will always be
positive. Thus, as explained ab~e for~, values ofD w D 12 , D22 , and D66 will
always be positive. Recall that QI6 and Q26 are odd functions ofe. DI6 and D26
will therefore be zero for all plies oriented at 0° or 90° as these plies have QI6
= Q26 = O. DI6 and D26 can also be made zero if for each ply oriented at +e and
at a given distance above the midplane, we have an identical ply at an equal
distance below the midplane but oriented at -e. This follows from the property
of odd functions of e, i.e., QI6 (+e) = - Q16( - e), Q26(+e) =Q26( - e) and (hk- hk_,)
is the same for both plies. Such a laminate, however, will not have a midplane
of symmetry, i.e., Bij 'f. O. In fact, DI6 and D26 are not zero for any midplane
symmetric laminate except for unidirectional laminates (0° or 90°) and cross-
plied laminates (0°/90°). We can, however, make DI6 and D26 arbitrarily small
by using a large enough number of plies stacked at +e. This is because the
contributions of +e plies to 0 16 and D 26 will be opposite in sign to those of - e
plies, and although their location will be at different distances from the
midplane, they will tend to cancel each other.
Ifwe wish to have an isotropic laminate, we can stack identical plies oriented
in such a way that the angle between any two adjacent layers is 21t1n. Such a
laminate will have [A] independent of orientation in the plane. We call such a
stacking sequence quasi-isotropic, as [B] and [D] are not necessarily isotropic.
Aij goes to zero only if the positive contributions are canceled by the negative
contributions.
Figure 10.10 shows the tensile stress-strain curves of [0]8' [90/0ls, and [90]8
SiCIRBSN laminates as well as of the unreinforced RBSN (Bhatt, 1991). The
[0]8 laminate curve shows three regions: a linear region, a nonlinear region
containing serrations, and a linear region again. Other laminates show only
linear region fracture. The elastic analysis described above had an important
assumption, viz., the bonding between the fiber and the matrix is perfect. That
assumption is not valid in the case of weakly bonded SiCIRBSN composites.
Accordingly, Bhatt et al. modified the above relationships to obtain the
383 MACROMECHANICS
600
500 - [O]B
m
0...
{ =, Specimen
~
failure
'"
en
<])
tJ
en
400
~
en
.::
<])
+-' 300
"@ [90;OJ,
.~
~
<])
200
+-'
S
0
0
~
100
Axial strain, %
Fig. 10.10 Tensile stress-strain curves of [0]8, [90/0]8' and [90]8 SiCIRBSN laminates
as well as ofthe unreinforced RBSN (after Bhatt, 1991).
When we stress a laminate composite in a given direction, the ply with the least
strength in that direction will fail first. This is commonly referred to as the first-
ply-failure (FPF). In a [0/90] laminate, a 90 ply cracks first while the 0 ply,
having a higher strength than the 0 ply, will be uncracked. If we unload this
laminate and reload, the stress-strain curve will follow the line "partially
degraded" which shows a reduced stiffness of laminate. At higher strains, the
o ply will also crack. The crack density in 0 and 90 plies increases with
384 MACROMECHANICS
c
Q.
(!)
200f\
175
~
......
~
)(
)( 150
LLI
UI 125
:::J
:::J
't:J
0
100
E
-u
UI
c
LLI
75
50
25
0
0 20 40 60 80 100
e, deg.
Fig. 10.11 Variation of Young's modulus as a function of fiber orientation in
SiC/RBSN laminate (after Bhatt, 1991).
increasing strain until a saturation crack spacing is attained. Figure 10.14 shows
crack density as a function of applied strain for different laminates. As we
pointed out in chapter 7, such cracking results in a loss of stiffness which can
used as a measure of damage in the laminate. Figure 10.15 shows the variation
in normalized stiffness (0.02% secant modulus) as a function of applied strain.
The stiffness reduction results from the combined effect of cracking in 90 and
oplies.
Ply delamination is an important failure mode in laminated composites, Fig.
10.16. Delamination cracking in [0/90] laminated ceramic composites,
consisting ofNicalon fibers/LAS-II glass-ceramic matrix made by hot pressing,
has been analyzed (Sbaizero et a!., 1990). This material has a thin layer of
carbon at the fiber/matrix interface which imparts ease of debonding and sliding
and consequently toughness in the longitudinal directions (Prewo, 1986). It was
found that crack growth initiation resistance was dominated by the critical
strain energy release rate of the matrix. Bridging effects associated with intact
fibers and segments of matrix led to an increased resistance as the crack
extension progressed, i.e., a rising R-curve type behavior was observed.
385 MACROMECHANICS
900
800
o Matrix Crocking
Stress
c 700
Q.
~ • Ultimate Tensile
~
~
600 Strength
><
b><
500
.c
C.
c::
400
Q)
!:
III 300
~
'iii 200
c::
~
100
0
0 20 40 60 80 100
e, deg
Fig. 10.12 Variation of room temperature strength with fiber orientation for an
SiCIRBSN laminate (after Bhatt, 1991).
400.-----------------------------------------~-,
350
.,
a.
300
~ 250
...,
If)
-.-
~
iii
200
___.- (0/90)5
~ 150
Q. ....._..--_... (0/90)35
a.
<{
..................-_......._........... __ ._..._ ..- ..-- (02/904)5
.>-/,.------------
100
50
01+-----r----,----~----_r----~----~----~---4
o 0.1 0.2 0.3 0.4 0.5 0.6 0.7 O.B
Strain (%)
Fig. 10.13 Stress-strain curves of Nica10n fiber/calcium aluminosilicate matrix
laminates (after Pryce and Smith, 1991). Note the characteristic knee in the curves
indicating the beginning of matrix cracking.
386 MACROMECHANICS
8 I I I I I I
• • Q -
7
_.
'-
6
• b- b-
I
E
I--
• • •
E • ¢ ¢
-
-
5 r- b-
>-
C/) 4 I--
• • -
"0
c:
Q)
3 r- ¢
b-
~
• (0)12 -
~
c
u
2 I--
•• • (0/90)S
-
u
~
b-
/:).
(02 /904)S
~
II-- ¢¢.b- <> (0/90)3S -
0 ~ I L I I I
1.0 AJrA
-b-
E»
0.9 rP
0 (0)12
C/) -rP A (02 /90 4 )S
-- •
C/)
.~
Q)
c: 0.8 (0190 )s
•.eb-
eb-
C/)
"0
0.7 .ib- b-
•i
Q)
N b- b-
e
c 0.6 b- b-
E • •
• •
~
0 iii
z
0.5 e
0.4
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Strain, %
Fig. 10.15 Variation in normalized stiffness (0.02% secant modulus) as a function of
applied strain. (After Pryce and Smith, 1991).
387 MACROMECHANICS
Central
delamination
Edge Edge
- -
delamination / delamination
~ .-
~
33
032
3 °23
°12
1 °11
Free
edge
o
Load
14 .1
Laminate
thickness
Fig. 10.18 Stresses in the boundary layer of a laminate. Note the stress in the thickness
direction.
389 MACROMECHANICS
Laminated ceramic composites can also be made from sheet materials, i.e.,
without using fibers. Such laminated multilayer ceramic composites are used
extensively in the electronic industry as well as, to a lesser extent, in other
areas. Laminated glass composites, for example, a composite made of a sheet
of polyvinyl butyryl polymer sandwiched in between two transparent glass
sheets provides high optical quality and safety features and are extensively used
in automotive and architectural windows. Virkaretai. (1987, 1988) made three-
layer composites by dry-pressing powders to obtain surface layers of
unstabilized zirconia and magnesia or alumina and a core layer of matrix oxide
(magnesia or alumina) and stabilized zirconia. When cooled from the
processing temperature, the unstabilized zirconia transformed to monoclinic
zirconia, putting the surface layers in compression. This decreased the
sensitivity to large surface flaws as long as the compressive layer thickness was
greater than the flaw size. Similar ideas can be used to produce high
temperature and corrosion resistance surfaces in nonfibrous laminated
composites. Tape casting technique can be used to produce multilayer laminates
(Amateu, 1990; Amateau and Messing, 1990, Gladysz et aI., 1999). Amateau
and coworkers applied the c1assicallaminate plate theory given above to layers
containing different volume fractions of silicon carbide whiskers to tailor the
residual stress in the SiCwi'Al2 0 3 and SiCwI'mullite laminates. They were able to
exploit the thermal expansion and elastic modulus mismatch between layers to
produce tailored laminates having residual compressive surface stresses and a
reduced tensile stress in the core. Charalambides (1991) also used the laminated
plate theory to analyze the steady-state mechanics of delamination cracking in
symmetric laminated notched beams subjected to an applied moment and an
axial force. Clegg and coworkers (Clegg et aI., 1990; Phillips et aI., 1992a;
Clegg and coworkers (Clegg et aI., 1990; Phillips et aI., 1992a; Clegg and
390 MACROMECHANICS
Seddon, 1992; Phillips et aI., 1992b; Phillips et aI., 1992c) used the concept of
toughening brittle materials by interleaving the laminae with weak interfaces.
The basic idea is a very simple one. They produced SiC laminates with weak
graphitic interlayers by viscous processing. The typical laminate thickness was
about 100 - 300 !lm with graphite layer thickness being about 10 !lm. Figure
10.20 shows load vs. apparent displacement in a
three-point bend test for notched monolithic SiC and for the laminated
composite. The most characteristic feature of these laminates was the
phenomenon of crack deflection at the interfaces between laminae. This
resulted in apparent work of fracture of 6 kJ m- 2 vis it vis 28 J m- 2 of the bulk
SiC produced by the same route.
Load Edge
delamination
Fig. 10.19 Schematic of delaminations at the free edge.
250
200
--. 150
6 Laminate
p..
~~
100
0
~
50
REFERENCES
Phillips, A.J., Clegg, W.J., and Clyne, T. W. (March 31-April2, 1992a) in Proc. Fatigue
and Fracture 0/Inorganic Composites, Cambridge.
Phillipps, A.J., Clegg, W.J., and Clyne, T.W., submitted to Acta Met. et Mater., 1992b.
Phillipps, A.J., Clegg, W.J., and Clyne, T. W., submitted to Acta Met. et Mater., 1992c.
Pipes, R.B., Kaminski, B.E., and Pagano, N.J. (1973) in Analysis o/the Test Methods
for High Modulus Fibers and Composites, ASTM STP 521, ASTM, Philadelphia, p.
218.
Pipes, R.B. and N.J. Pagano (1974) J. Appl. Mech., 41, 668.
Prewo, K.M. (1986) J. Mater. Sci., 21, 3590.
Prewo, K.M. and U. Brennan (1982)J. Mater. Sci., 17,1202.
Pryce, A.W. and P.A. Smith (1991) in Proc.: Inti. Con! on CompOSite Mater.l8
(JCCM/8), Hawaii, p. 24-A-1.
Sbaizero, 0., P.G. Charalambides, and A.G. Evans (1990) J. Amer. Ceram. Soc., 73,
1936.
Velamakanni, B.V. and F.F. Lange (1991) J. Amer. Ceram. Soc., 74, 166.
Virkar, A., J.L. Huang, and R.A. Cutler (1987) J. Amer. Ceram. Soc., 70, 164.
Virkar, A., J. Jue, J. Hansen, andR.A. Cutler (1988) J. Amer. Ceram. Soc., 71, C148.
SUGGESTED READING
APPLICATIONS
'"
(a)
Fig. 11.1 (a) Ceramic matrix composite turbine blades made by sol-gel vacuum
impregnation of fibrous preforms, (b) a filament wound tube of CMC, (c) hammering
a nail through a CMC shows the damage tolerance, and (d) an oxyacetylene flame
impinging on the wall of a tube made of CMC shows the ability to withstand high
temperatures (courtesy of A.R. Hyde, 1989).
396 ApPLICATIONS
(b)
(c)
Fig. 11.1 - contd
397 ApPLICATIONS
Cd)
Fig. 11.1 - contd
Mach number
2.4 2.8 3.2 3.8 4.5 6.0 10.0
.
c:
3.0
In what follows, we first describe some general applications ofCMCs and then
examine, in detail, some specific applications, highlighting the salient features
regarding processing, microstructure, and properties of these composites.
plane programs (Desnoyer, 1991). Besides the space plane, other applications
for ceramic matrix composites include a variety of high speed airplane, various
defense related projects such as Advanced Tactical Fighter (ATF), many
existing fighters, missiles, hypersonic radomes, hard armor and turbine engines.
It is expected that a number of propulsion related components in the high speed
airplane will use a variety of CMCs. Requirements such as high thrust, high
temperatures, low airport noise, reduced NO x emissions, etc. mean that CMCs
will have to be used. As a rule, aerospace applications require high performance
materials in large, complex shapes.
While the ultimate temperature requirement may exceed 2500 °C in some rare
applications, one can list the following three temperature categories where
different materials can meet the requirements:
For ceramic matrix composites the last two categories are relevant, i.e;, the
service temperature range is 800 -1650 °C. This temperature range which
represents the extremes for a turbine engine exhaust under normal and after-
burner operations and structural elements for some of the conceptual vehicles.
A structural fiber reinforced composite with balanced in-plane tensile and
compression strength in the range of 175-350 MPa and moduli in the range of
100-175 GPa throughout the temperature range would offer a substantial
payoff in turbine engine, space satellite and transport, and hyper velocity
vehicle technology.
surprising that for applications related to wear, cutting tool inserts, and heat
engines, there are commercially available dense, wear-resistant, particle- and
whisker-reinforced ceramic matrix composites. Successful inroads into cutting
tool inserts market have been made by CMCs such as TiC particle reinforced
Si3N4 and AIP3' and SiCjAI20 3. The competition in this case is mainly
tungsten carbide/cobalt, which accounts for about 50 % of the u.S. metal-
cutting tool market of $650 million. The main advantages of ceramic cutting
tools include chemical stability, superior hardness, and ability to operate at
higher temperatures which translate into higher cutting speeds.
In the area of wear resistant parts, ceramics offer high hardness, low friction,
superior abrasion resistance, good elevated-temperature capabilities, and good
mechanical performance at high speeds. Applications include vanes, seals,
nozzles, bearings, and wear guides. Among wear resistant parts, toughened
zirconia, and whisker and continuous fiber reinforced composites find
applications.
In the energy related areas involving heat transfer, storage, and recovery, heat
exchangers and recuperators made from monolithic ceramic materials can
operate at higher temperatures and in a more aggressive environments than
metals. However, a major drawback of monolithic ceramics is their poor
thermal shock resistance. Here continuous fiber reinforced ceramic matrix
composites can be used profitably. For example, continuous fiber reinforced
hot-gas filters for particle control purposes can be very efficient in hot, coal-
derived gas streams. Such gas streams can be found in pressurized fluidized-bed
combustors, direct coal-fueled combustors, and combined gasification systems.
The use ofCMC components in a variety of applications makes economic sense
because it will result in higher operating temperature, higher efficiency,
benefits in the energy and environmental (reduced emission) areas, longer
service life, etc. We describe below some of these applications
Hot-gas filters
These are porous, closed-end tubes used to remove fine particles in a variety of
operations, such as fly ash in coal gasification and in pressurized fluidized-bed
401 ApPLICATIONS
Combustor liners
The use of CMC as combustor liners in gas-turbine power systems will result
in reduced emissions of nitrous oxides, commonly indicated by NOx, and fuel
savings because of operation at higher inlet temperatures.
Automotive area
temperature and thereby increase fuel efficiency. The requirements for ceramics
in heat engine applications are good heat insulation, high strength, low abrasion
and corrosion, high toughness, low coefficient of friction, low density, good
thermal shock resistance and low cost.
Table 11.1: Potential industrial applications for continuous fiber reinforced ceramic
composites (after Sheppard, 1990).
It is worth pointing out that the most important parameter in determining the
thermal characteristics of a carbon fiber is the degree of crystalline orientation
along the fiber axis. The higher the degree of crystalline orientation along the
fiber axis, the higher will be the modulus along the fiber, the lower will be the
coefficient of thermal expansion, and the higher will be the thermal
conductivity. Some pitch-based carbon fibers such as the series K fibers and
another form of carbon fiber, called vapor grown carbon fiber (VGCF), have
much more ordered crystalline structure than PAN-based carbon fibers and,
consequently, the former can show extraordinarily high thermal conductivity
values (> 1000 W/mK).
• oxidation resistance
• low oxygen permeability and volatility
• good adhesion to carbon
• chemical compatibility with carbon
• thermal expansion match with carbon
Figure 11.5a shows the critical factors that should be considered in an oxidation
protection system for carbon/carbon composites (Strife and Sheehan, 1988),
while Fig. 11.5b shows, schematically, a typical multilayer oxidation protection
system on a C/C substrate. Luthra (1988) has considered various rate limiting
steps such as gas phase diffusion, interface reaction, diffusion through liquid
B2 0 3 , and diffusion through cracks in the barrier layers. He calculated the
405 ApPLICATIONS
25
-- - -- ,
- - -- - - --
2D Carbon/Carbon __ / '
/
20
- - - -- - --
15 ~ Nickel based Superalloy
a LL.L..L...L.l-l-L.L..L I II II ! I ! I I I ! I ! ) ! 11~~LL~~~
Temperature, °C
Fig. 11.3 A comparison of high temperature capabilities of carbon/carbon composites
with metallic alloys, in terms of strength to weight ratio as a function of temperature
(Strife and Sheehan, 1988). Carbon/carbon composites show superior specific strength
at higher temperatures in inert atmospheres.
2.2
2.0
HIPIC
r<>
1.8 Process
I
E
t> 1.6
Ol
~ 1.4
>-
+-
·in 1.2
c
Q)
1.0
0
0.8
0.6
0.4
0 234 5 6 7
Cycles
virtual maximum rates for these steps and compared them with the acceptable
oxidation rates for short term (20 hours) and long term (2000 hours)
applications. His main conclusions are that the oxidation of unprotected
carbon/carbon composites below 500°C is controlled by interface reaction, at
higher temperatures the rate limiting step changes to a mixed gas phase
diffusion - interface reaction. Finally, above 700 °C, depending on the gas flow
conditions, the reaction is limited by gas phase diffusion. Inhibitors can be
beneficial at temperatures up to 700°C for long term and 850 °C for short term.
His calculations on oxygen transport rates through 100 11m layer of B20 3
indicate that for systems relying on it exclusively, the temperature limits are
about 600°C for short term and 850 °C for long term.
The two approaches are frequently combined in that one uses diffusion barriers
and internal glass forming inhibitors and sealants. Silicon-based ceramics, such
as SiC and Si3N 4, are commonly used as the primary oxygen barriers. Sealants
have the added advantage that they also seal thermal stress cracks originating
because of the thermal mismatch between C/C composite and silicon carbide
coating. Tetraethylorthosilicate (TEOS) is commonly used as an agent to
impregnate such cracks (McKee, 1986; McKee, 1987). TEOS leaves oxidation
resistant silica in the cracks. Ehrburger et al. (1986) studied the oxidation
inhibition by boria, B2 0 3. Their results are shown as weight loss of untreated
and B2 0 3 treated samples as a function of oxidation time at 980 K. The
presence ofBP3 significantly decreased the oxidation rate. McKee (1986) used
a ammonium borate - boric acid solution to protect the carbon/carbon
composite. Figure 11.6 shows the weight loss curves vs. time for the untreated,
2% B20 3 and 7% BP3 - treated carbon/carbon in flowing air at 1000 °C
(McKee, 1987). The borate-coated material showed a very small weight loss.
Extended oxidation protection of carbon/carbon composites can be obtained by
using boria containing dense Si3N4 or SiC coatings and boron-rich inner
coating. Borate glasses act as sealants and fill up the cracks that form in the
outer coatings because ofthermal expansion mismatch. An oxidation protection
system used for extended protection under thermal cycling conditions, with
peak temperatures reaching 1400 °C, is shown in Fig. 11.7 (Strife and Sheehan,
1988). Moisture sensitivity of borate glasses, corrosion of outer coatings by
glasses, glass volatilization, and an unacceptably high oxygen permeability are
some of the negative factors. Alvey and George (1991) used multilayer Zr/Pt
coating that formed ZrPt3 on annealing on carbon/carbon composites. Their
results showed that this coating protected carbon and a carbon fiber/phenolic
resin composite from a 2800 °C flame. They attributed this performance to the
oxidation resistance of ZrPt3 and to the reflectivity of the coating, which
reduced the thermal flux on the substrate to a short time.
407 ApPLICATIONS
Oxygen In Volatility
Adherence:!:;:
t;~~~~~~~~~~~~~~~~r-Multilayer
Coating
CtC
Substrate
(a)
Borosilicate Overglaze
(optional)
CVD SiC
Oxidation Barrier
Boride Sealant
. CjC Substrate
(b)
Fig. 11.5 (a) Critical factors that should be considered in an oxidation protection system
for carbon/carbon composites (after McKee, 1987), (b) schematic of a typical multilayer
oxidation protection system on a C/C substrate.
408 ApPLICATIONS
40 Untreated
I
I
I
I
on
on 2%B20 3 I
0
....l 30 I
I
.....
..c I
.....
bI) I 7%B20 3
~.....
I
/
1=1 20 I
<l) I
~
<l)
I
I
0..
I
10 /
/
,,
,
/ ,
/ "
/ "
//
-::-- -
-','
00 5 10 15 20
Time, h
Fig. 11.6 Weight loss curves vs. time for the untreated, 2% BP3 and 7% BPl - treated
carbon/carbon in flowing air at 1000 °C. The borate-coated material shows a very small
weight loss (after McKee, 1986).
Inner Coating of
Glass Formers Cracked Outer
Coating
/
7
o.. . o
..
·0 •
·0 •
..
.. . o
•• 0
Continuous
Carbon Fibers
Glass Formers in
Carbon Mattix
Fig. 11.7 A schematic of the protection system for extended protection ofC/C under
thermal cycling conditions, with peak temperatures reaching 1400 °C. Borate glasses act
as sealants and fill up the cracks that form in the outer coatings because of thermal
expansion mismatch (after McKee, 1986).
409 ApPLICATIONS
The brake design must consider the following items (Awasthi and Wood,
1988):
When the brakes are applied, multiple-disk brakes with alternating rotors and
stators are forced against adjacent members by hydraulic pressure. Friction
between rotating and stationary. disks causes them to heat up to 1500 °C
(surface temperature can be as high as 3000 0c), so good thermal shock
resistance is required. In view of the requisites listed above, any braking
material must be a good structural material, an efficient heat sink, and must
have excellent abrasion resistance. Let us consider an example from the aircraft
field which will help illustrate the formidable demands made on a braking
material. The Concorde supersonic plane and the Boeing 767 are among the
civilian aircraft that use carbon/carbon composite brakes. Let us take Boeing
767 as an example to help illustrate the formidable demands made on a braking
material. It has a mass of about 170,000 kg. Let us say that it has a take-off
velocity of about 320 kmIh. This will give us a kinetic energy at take-off of 670
MJ. In the event of an aborted take-off, this energy must be dissipated in about
30 seconds by the eight brakes on the aircraft. A rejected take-off (RTO) is,
indeed, the worst case scenario but then the braking material must be able to
meet such requirements. It is estimated that about 3000 RTOs occur every year
in the US.
410 ApPLICATIONS
Fig. ll.S An example of a carbon/carbon brake assembly used on a Boeing 767 plane
(Courtesy of AlliedSignal Corp.).
• Abrasion resistance
• Thermal shock resistance
• Strength
• Fracture toughness
• Thermal conductivity
Ceramic tool bits retain high strength and hardness to elevated temperatures.
This allows the use of ceramic tool bits at much higher machining speeds than
is possible with tungsten carbide based tool bits. However, ceramic tool bits
have a toughness lower than that of tungsten carbide tool bit materials. This
restricts the use of ceramic tool bits to turning and milling operations. Table
11.2 gives a summary of room temperature properties of different cutting tool
412 ApPLICATIONS
Fig. 11.10 Concept of thermal protection (after Strife and Sheehan, 1988).
Table 11.2 Room Temperature Properties of Cutting Tool Materials (after Burden et
aI., 1988)
Ceramic oxide matrix composites are used as cutting tool inserts, wear parts
such as wire drawing or extrusion dies, valve seats, high precision balls,
bearings for corrosive environments, and plungers for chemical pumps.
Partially stabilized zirconia is by far the largest of the ceramic material market
in wear and aerospace applications. Silicon carbide whisker or fiber reinforced
alumina, titanium carbide particle reinforced alumina, partially stabilized
zirconia are examples of ceramic composites used for machining. Typically,
SiC)Al20 3 cutting tools contain 30-45 % SiC whiskers. Cutting ofInconel718
with SiCjAI 20 3 composite tools shows a performance 3 times better than
conventional ceramic tools and 8 times better than cemented carbides. Typical
physical properties and performance parameters of AI 2 0 3 and SiC)AIP3
composites used for cutting tool inserts are given in Table 11.3 (Billman et aI.,
1988). During metal removal, a large amount of shear stress is required to
plastically deform and shear the metal, which results in high temperatures. As
much as 80% of the heat generated in cutting is due to this action (Whitney and
Vaidyanathan, 1988; Smith, 1986). This large amount of heat can cause failure
of the cutting tool and thus necessitates a reduction in the cutting speed of the
tool made of conventional materials. SiC whisker reinforced Al 20 3 composites
can withstand high temperatures while maintaining reasonable strength and
toughness. With these tool materials, the heat generated in the shear zone ahead
of the tool can be exploited to obtain optimum speeds at temperatures greater
than 1400 °C. In light finishing of nickel based alloys, it is claimed that a chip
415 ApPLICATIONS
is not produced, the metal being removed is so hot that it oxidizes to a powder
(Whitney and Vaidyanathan, 1988; Smith, 1986).
(a)
Fig. 11.12 (a) Ceramic tools made by hot pressing of SiC whiskers in an AIP3 matrix
(courtesy of Greenleaf Corp.); (b) microstructure of this material.
416 ApPLICATIONS
(b)
Fig 11.12 -contd
Under low pressure and temperature conditions (Naslain and Langlais, 1986;
Fitzer et aI., 1980), chemical reaction kinetics control the matrix growth and
good matrix can be obtained. CVI techniques have been successfully used to
417 ApPLICATIONS
deposit C, SiC, Si3N 4 , TiC, B 4C, BN, A120 3, Zr0 2 as matrix materials (see
Chapter 4). Also, invariably, multiple impregnation cycles combined with
intermediate surface removal are required to obtain an adequate density. The
process, however, is very slow. The French aerospace company, SNECMA, has
specialized in using flexible, small diameter fibers, such as carbon or Nicalon,
in a fabric form to make any complex shaped preform. The carbon fiber/silicon
carbide matrix (trade name SEPCARBINOX), silicon carbide fiber/silicon
carbide matrix (trade name CERASEP), and carbon fiber reinforced carbon
(trade name SEPCARB) are some of the commercial varieties made by
SNECMA. A two-dimensional preform is made of balanced (0°/90°) woven
fabric of carbon (T300 high strength fiber) or silicon carbide (Nicalon) fiber.
Three dimensional preforms, fibers woven manually or in a machine in three
orthogonal directions have also been used. Table 11.4 gives the physical
properties of various C/SiC and SiC/SiC composites made by SNECMA. Table
11.5 gives mechanical and thermophysical properties of two-dimensional C/SiC
and SiC/SiC and three-dimensional C/SiC composites at room temperature,
1000 °C, and 1400 °C (Lacombe and Rouges, 1990). Exceptionally high fracture
toughness have been obtained in these composites. Tensile strength of C/SiC
and SiC/SiC as a function of temperature is shown in Fig. 11.13, while thermal
Table 11.3 Properties and Perfonnance of A1 20 3-based Cutting Tools (after Billman et
aI., 1988)
Table 11.4 Physical properties of two dimensional (2D) C/SiC and SiC/SiC composites
and three dimensional (3D) C/SiC composites (after Lacombe and Rouges, 1990).
2D C/SiC 45 2.1 10
2D SiC/SiC 40 2.5 10
3D C/SiC 24 2.3 12
conductivity and thermal diffusivity of these composites in the plane and in the
thickness direction are shown in Figs. 11.14 and 11.15, respectively. Note the
high degree of anisotropy in two-dimensionally reinforced composites.
These CMCs are mainly meant for use in rocket or jet engines:
• Turbojet parts
These are meant mainly for the afterburner area and have the benefits of lower
weight, high thermal shock resistance and less maintenance compared to the
conventional materials. Actual flights have been performed on the Mirage 2000
using components made of these composites (Lamicq et aI., 1991).
Panels having aerodynamic shapes and dimensions as large as 1.8 x 0.8 x 0.3
m, have been formed with these CMCs. The main limitation on the dimensions
comes from the size of the reactors available. One major application of
SEPCARBINOX (C/SiC) will be in making of thermal structures in the
European space plane, Hermes, which will experience high mechanical loading
and surface temperatures as high as 1300 °C. These panels will be used in the
Hermes space plane on the external surfaces, where the thermal fluxes are very
419 ApPLICATIONS
450
400
2D C/SiC /
--
~
350 -
tr:: 300
1l
IZl 250
~r:: 2D SiC/SiC
200 - - - - - - -
~
150
3D Novollex
C/SiC Composile
100
Temperature, °C
Fig. 11.13 Tensile strength ofC/SiC and SiC/SiC as a function of temperature (after
Lacombe and Rouges, 1990).
25
In Plane
20
1
-5
~
15
~
o~
0p
U
.g
I':
0 10
U
] 5
Through Thickness
Temperature, DC
Fig. 11.14 Thermal conductivity, in-plane and in thickness direction, of C/SiC and
SiC/SiC composites (after Lacombe and Rouges, 1990)0
420 ApPLICATIONS
high. It is claimed that the overall weight of this is lower that the system used
in the US space shuttle where light weight but thick tiles are used (Lamicq, et
aI., 1991). Carbon/carbon thermal structures and protection are planned to be
used in the Huygens probe that will explore the moon Titan. The Titan
atmosphere is nonoxidzing, therefore, carbon/carbon are appropriate for this
mISSIon.
Materials for turbines, land-based stationary gas turbine (CSGT) for power-and-
steam cogeneration operation or for service in aerospace must withstand long
421 ApPLICATIONS
1. 20 C/SiC In Plane
2. 20 C!SiC Through Thickness
12 3. 3D C!SiC In Plane
4. 30 C{SiC Through Thickness
5. 20 SIC/SiC In Plane
6. 20 SiC/SiC Through Thickness
10
\ \
\ '- \
\
8 \\ \
\\ \
\
\', , \.,
6 \. \', " "-
" '-
\., ""
, ", '- '-
4 "
--
~, - -- ·3
--------. 5
-4
2 6 -------------
Temperature, °C
Fig. 11.15 Thennal diffusivity, in- plane and in thickness directions, of CISiC and
SiC/SiC composites (after Lacombe and Rouges, 1990).
REFERENCES
Mater., 3rd Euro. Conf. on Composite Mater (ECCMl3)., Elsevier Sci. Pub., London"
p.99.
Desnoyer, D. (1991) Mater. Tech. (paris) 79, 57.
Ehrburger, P., Baranne, P., and Lahaye, J. (1986) Carbon, 24, 495.
Faber, K.T. and Evans, A.G. (1983) Acta Metall., 31, 565.
Fitzer, E., Hegen, D., and Strohmeier, H. (1980) Rev. Int. Hautes Temp. Refrac., 17,23.
Giannakopoulos, A.E. and Breder, K. (1991) J. Amer. Ceram. Soc., 74, 194.
Hyde, A.R.(1989) GEC J. Res. 7, 65.
Korb, L.J., Morant, C.A., Calland, R.M., and Thatcher, C.S. (I981)Amer. Ceram. Soc.
Bull, 60, 1188.
Lacombe, A. and Bonnet, C. (October 1990) in AIAA '90, Second Inti Aerospace Planes
Conference '90, Orlando, FL" Amer. Inst of Aero. and Astro., Washington, DC.
Lacombe, A. and Rouges, J.M. (September 1990) in AIAA'90, Space Program and
Technologies Conference '90, Huntsville, AL, Amer. Inst. of Aero. and Astro.,
Washington, DC, AIAA-90-3837.
Lamicq, P., Bonnet, C., and Chateigner, S. (1991) in Designing with Structural
Ceramics, Elsevier App. Sci., London, p.251.
Luthra, K.L. (1988) Carbon, 26, 217.
McAllister, L.E. and Lachman, W.L. (1983) in Fabrication of Composites, North-
Holland, Amsterdam 143.
McKee, D.W. (1986) Carbon, 24, 737.
McKee, D.W. (1987) Carbon, 25,551.
Meetham, G.W. (1989) Mater. and Design, 10, 197.
Miriyala N., J. Kimmel, J. Price, K. More, P. Tortorelli, H. Eaton, G. Linsey, and E.
Sun (2002) in Proc.: 47th ASME IGTI ASME Turbo Expo 2002: Land, Sea, and Air,
GT-2002-30585, Amsterdam, The Netherlands.
Naslain, R. and Langlais, F. (1986) Mater. Sci. Res., 20,145.
Sheppard, L.M. (1990) Amer. Ceram. Soc. Bull., 69, 667.
Smith, K.H. (1986) Carbide and Tool Journal, 8:5, 8.
Strife, J.R. and Sheehan, J.E. (1988) Amer. Ceram. Soc. Bull., 67, 369.
Taverna, AR.and L.E. McAllister (1971) inAdv. Materials: Composites and Carbon,
Amer. Ceram. Soc., Columbus, OH, p.199.
Whitney, E.D. and Vaidyanathan, P.N. (1988) Amer. Ceram. Soc. Bull., 67,1010.
APPENDIX A: WEIBULL STATISTICS
Let P(Vo) be the survival probability, i.e., P(Vo) is the fraction of identical
samples, each of volume Yo, that survives when loaded a stress o. According
to a distribution named after a Swedish engineer. Weibull, this survival
probability is given by
(A.I)
where 0 0 and ~ are constants. Figure A.1 shows a plot ofP(Vo) vIS. o. At 0 = 0,
i.e., no load on the sample, all samples survive and P(V0) ;: 1. As 0 increases,
more samples fail and P(Vo) decreases. Eventually, P(Vo) ~ 0, as 0 ~ 00, Le.,
all samples fail at very high loads. It is easy to see from Eq. (A. I ) that for 0 =
1
P (V) = -; = 0.37 (A.2)
i.e., 0 0 is the stress corresponding to a survival probability of 37%. Pis the so·
called Weibull modulus. The smaller the value of p, the greater the variability
in strength. In other words, the value of ~ tells us how rapidly the strength falls
as we approach 0 0 , Figure A.2 shows schematically the strength distributions
for a series of ~ values.
------1---------
t
1
1
1
.5 ~-~-~-~-~-~
--
.-
~
~ UL-______~____~__
a ---
Fig A.I. Survival probability, P(Vo) as function of stress, 0.
426 ApPENDIX A
t
~=5
OL-__________~~~~__
a --..
Fig. A.2. Strength distributions for a series of Weibull modulus (~ )values.
CA3)
(A.4)
Equation (A4) represents a straight line on a double log plot with the slope, ~
and intercept on the y-axis of VN o' Table Al gives some typical ~ values.
Figure A3 shows how the Weibull plot, for a given material, changes with
improvement in, say, processing such that there are smaller defects, which
would result in a higher mean strength but that the distribution of flaws remains
the same, i.e., we have the same value of ~. On the other hand, if the
distribution of flaws changes but the size does not, we shall have a higher value
of ~ but the same value of mean strength. In chapter 3, we have used a
expression for the Weibull distribution that includes the dependence of strength
on fiber length. We can include the volume dependence of material in the
Weibull distribution in the following manner:
427 ApPENDIX A
v (-)P]
P(V)=exp[-- 0 ° (A.S)
o Vo 00
In P (V) = - -v °
(-)P0
Vo 00
In In [_1_] = In ~ + ~ In ( ~) (A.6)
P(V) Vo 00
Original
i
....
Q)
i
....
Q)
distribution
I
I
I
I
.= ::l I
--
0 /
~ 00-
I
.
/
00-
-
0 0
>. >. I
I
..0 :0 h /
/
0 0
! Higher
/t1
..0 ..0
....
0 0
.... I
• 13 ,
I
0. 0.
I
same
·
0>
0
0>
0
I
• 0:- ISame {1 ,
...J ...J I
• " higher (f
Fig A.3. Effect of improvement in processing results in a higher mean strength but that
the distribution of flaws remains the same, i.e., the same value of ~. If the distribution
of flaws changes but the size does not, we have a higher value of ~ but the same value
of mean strength.
428 ApPENDIX A
Table A.I: Typical values of the Weibull modulus for some materials".
Material (3
Steel 100
• indicative values only.
APPENDIX B: SOME IMPORTANT UNITS AND
CONVERSION FACTORS
B.2 DENSITY
B.3 VISCOSITY