4 Metal Matrix Composites

Krishan K. Chawla

Department of Materials and Metallurgical Engineering,

New Mexico Institute of Mining and Technology, Socorro, NM, U.S.A.

List of Symbols and Abbreviations 122

4.1 Introduction 125
4.1.1 Types of Metal Matrix Composites 125
4.1.2 Important Metallic Matrix Materials 126
4.2 Processing 128
4.2.1 Liquid State Processes 128
4.2.2 Solid State Processes 129
4.2.3 Deposition Techniques 134
4.2.4 In Situ Processes 135
4.3 Interfaces in Metal Matrix Composites 137
4.3.1 Importance of the Interface in Composites 138
4.3.2 Crystallographic Nature of the Fiber/Matrix Interfaces 138
4.3.3 Interfacial Bonding in Metal Matrix Composites 139
4.3.3.1 Mechanical Bonding 139
4.3.3.2 Chemical Bonding 140
4.3.4 Thermal Stresses in Metal Matrix Composites 141
4.3.5 Microstructure of the Interface Region 144
4.3.5.1 Alumina/Metal Composites 145
4.3.5.2 Silicon Carbide/Metal Composites 147
4.3.5.3 Carbon Fiber/Metal Matrix Composites 150
4.3.5.4 Interaction Between Interfacial Phenomena 150
4.4 Properties 151
4.4.1 Importance of the Metallic Matrix 151
4.4.2 Modulus 154
4.4.3 Strength 156
4.4.4 Matrix Aging Response in Composites 157
4.4.5 Toughness 160
4.4.6 Thermal Properties 162
4.4.7 Fatigue 163
4.4.8 Creep 173
4.4.9 Tribological Properties 174
4.5 Applications 176
4.6 References 179

Materials Science and Technology

Copyright © WILEY-VCH Verlag GmbH & Co KGaA. All rights reserved.
122 4 Metal Matrix Composites

List of Symbols and Abbreviations

a crack length
a radius of fiber
A fatigue crack propagation parameter
b radius of matrix shell
d diameter of fiber, density
D diffusivity
E elastic constants (general)
E Young's modulus
Eo modulus of uncracked material
Et transverse Young's modulus
f subscript: fiber
g acceleration due to gravity
G shear modulus
h height of composite
IA area of fiber/matrix interface
Jc critical current density (of superconductors)
k Boltzmann constant
K bulk modulus
AK cyclic stress intensity factor
AK th threshold stress intensity
Klc plain strain fracture toughness
(1) liquid
I length of composite, length of shaft
m fatigue crack propagation exponent
m subscript: matrix
n creep stress exponent
N number (of cycles)
Nc critical speed (rpm)
P pressure between interface and matrix
P property
P subscript: particle
P load
r geometric parameter
R solidification rate
R stress ratio
R{ inner radius
Ro outer radius
(s) solid
s aspect ratio of discontinuous reinforcement
S stress amplitude
S,Sf precipitates in A l - C u - M g system
t time
T temperature
 List of Symbols and Abbreviations 123

AT temperature interval
Tc critical temperature (of superconductors)
Tm melting temperature
V volume fraction
w subscript: whisker
w width of composite
x thickness of the reaction zone
Y a constant, a geometric factor

a linear expansion coefficient

Aa thermal mismatch
s creep strain
e strain
0'90" precipitates in A l - C u system
X interfiber spacing
v Poisson's ratio
a strength
o stress
Aa cyclic stress range
G0 friction stress
cra cyclic stress amplitude
<r m mean stress
GX stress in radial direction
G0 stress in tangential direction
oz stress in direction of fiber

ARALL aramid aluminum laminate

C-PFZ carbon-precipitation free zone
CMC ceramic matrix composite
CTE coefficient of thermal expansion
CVD chemical vapor deposition
CVI chemical vapor infiltration
DC direct chill
FP polycrystalline a-alumina fiber (trade name)
GP Guinier-Preston zones
HIP hot isotopic pressing
HREM high resolution electron microscopy
LEFM linear elastic fracture mechanics
MMC metal matrix composite
MRI magnetic resonance imaging
NASP National Aerospace Plane (U.S.)
OPIT oxide-powder-in-tube method
PMC polymer matrix composite
PRD-166 Alumina + Zirconia fiber (trade name)
PVD physical vapor deposition
124 4 Metal Matrix Composites

SCS-6 monofilament silicon carbide fiber (trade name)

SEM scanning electron microscopy
SHS self propagating high temperature synthesis
TEM transmission electron microscopy
XD proprietary process to produce a composite

4.1 Introduction
A composite material can be defined
(Chawla, 1987a) as a material consisting of
two or more physically and/or chemically
distinct, suitably arranged or distributed
phases. It has characteristics that are not
depicted by any of the components in iso-
lation. Generally, one component acts as a
matrix in which the reinforcing phase is
distributed. The matrix component is,
thus, the continuous phase. When the ma-
trix component is a metal, we call such a
composite a metal matrix composite
(MMC). The reinforcement can be in the
form of particles, whiskers, short fibers, or
continuous fibers.
There are three entities that determine
the characteristics of a composite, viz., re-
inforcement, matrix, and interface. Earlier
on, the role of matrix was considered to be
that of a medium or binder to hold the
strong and stiff fibers or other types of
reinforcement. This view was particularly
prevalent in the polymer matrix composite
area. Over the years, however, it has been
realized that the matrix microstructure
and consequently its mechanical proper-
ties have a considerable influence on the
overall performance of a composite. This
is particularly true of the metal matrix
composites because the very act of incor-
porating a reinforcement - be that fiber,
whisker, or particle - can result in
change(s) in the microstructure of the
metallic matrix and, consequently, in their
structure-sensitive properties such as
strength and toughness.
In this article we first briefly describe
some of the important metal matrix mate-
rials and some commonly used reinforce-
ments. This is followed by a description of
the various processing techniques for mak-
ing metal matrix composites, the structural
characteristics of the interface region in a
4.1 Introduction 125
variety of MMCs, and properties of differ-
ent metal matrix composites. Finally, we
indicate various applications of metal ma-
trix composites.
4.1.1 Types of Metal Matrix Composites
There are three kinds of metal matrix
composites (MMCs): particle reinforced
MMCs, short fiber or whisker reinforced
MMCs, and continuous fiber or sheet rein-
forced MMCs. Figure 4-1 (a and b) shows
typical microstructures of particle (SiC
particle/Al) and continuous fiber rein-
forced (Al2O3/Al-Li) composites.
Particle or discontinuously reinforced
MMCs (one uses the term discontinuously
reinforced MMCs to indicate metal matrix
composites having reinforcements in the
form of short fibers, whiskers, or particles)
have assumed special importance for the
following reasons:
- Particle reinforced composites are inex-
pensive vis a vis continuous fiber rein-
forced composites. Cost is an important
and essential item for large volume
usage.
- Conventional metallurgical processing
techniques such as casting or powder
metallurgy, followed by conventional
secondary processing by rolling, forg-
ing, and extrusion can be used.
- Higher use temperatures than the unre-
inforced metal.
- Enhanced modulus and strength.
- Increased thermal stability.
- Better wear resistance.
- Relatively isotropic properties com-
pared to the fiber reinforced composites.
Within the broad category of discontin-
uously reinforced composites, the cast
metal matrix composites are cheaper to
produce than the powder metallurgy com-
posites. There two types of cast metal ma-
trix composites:
126 4 Metal Matrix Composites
(a) cast composite having local rein-
forcement.
(b) cast composite in the form of a billet
having general reinforcement with a
wrought alloy matrix. Such composite bil-
lets are forged and/or extruded, followed
by rolling or other forming operations.
(a)
(b)
Figure 4-1. Typical microstructures of some metal
matrix composites: (a) particle (SiC particle/Al) and
(b) continuous fiber reinforced (Al2O3/Al-Li) com-
posites.
4.1.2 Important Metallic Matrix Materials
A variety of metals and their alloys can
used as matrix materials. We describe the
important characteristics of some of the
more common ones:
Aluminum alloys
Aluminum alloys, because of their low
density and excellent strength, toughness,
and resistance to corrosion, find important
applications in aerospace field. Of special
mention in this regard are the Al-Cu-Mg
and Al-Zn-Mg-Cu alloys, very important
precipitation hardenable alloys. The latest
and by far the most important develop-
ment in the precipitation hardenable alu-
minum alloys is a series of aluminum-lithi-
um alloys. Lithium when added to alu-
minum as a primary alloying element has
the unique characteristics of increasing the
elastic modulus and decreasing the density
of the alloy. Understandably, the aerospace
industry has been the major target of this
development. Al-Li alloys are precipitation
hardenable much like the Al-Cu-Mg and
Al-Zn-Mg-Cu alloys mentioned above. The
precipitation hardening sequence in Al-Li
alloys is, however, much more complex
than that observed in conventional precip-
itation hardenable aluminum alloys. Gen-
erally, these alloys contain, besides lithi-
um, some copper, zirconium, and magne-
sium. Vasudevan and Doherty (1989)
provide an account of the various alu-
minum alloys.
Rapidly Solidified Aluminum Transition
Metal Alloys
Aluminum alloys capable of competing
with titanium alloys on a specific strength
basis up to 375 °C can be obtained by
strengthening with transition metal inter-
metallic compounds. Allied-Signal Corpo-

ration has developed a series of Al-Fe-V-Si
alloys using rapid solidification (cooling
rates > 1 0 6 K s ~ 1 ) having an excellent
combination of mechanical properties.
The high cooling rate results in an ex-
tremely fine grain size which, in turn, re-
sults in excellent property combinations
unobtainable by conventional metallurgy
route. Allied-Signal uses what it calls a
planar flow casting process to produce
rapidly solidified aluminum-based quater-
nary alloys having enhanced solubility of
iron, vanadium, and silicon with good
room temperature strength and ductility.
The silicon addition results in a ther-
mally stable, 40-50 nm spherical silicides
(Al12(Fe,V)3Si).
Titanium alloys
Titanium is one of the important
aerospace materials. It has a density of
4.5 gem""3 and a Young's modulus of
115 GPa. For titanium alloys, the density
can vary between 4.3 and 5.1 g cm" 3 while
the modulus can have a range of 80-
130 GPa. High strength/weight and modu-
lus/weight ratios are important. Titanium
has a relatively high melting point
(1672°C) and retains strength to high tem-
peratures with good oxidation and corro-
sion resistance. All these factors make it an
ideal material for aerospace applications.
Titanium alloys are used in jet engine (tur-
bine and compressor blades), fuselage
parts, etc. It is, however, an expensive ma-
terial.
At very high speeds, the skin of an air-
plane heats up so much that aluminum al-
loys become weak or even melt. Titanium
alloys must be used at such high tempera-
tures. In the U.S. National Aero Space
Plane (NASP) program, the plane will fly
at speeds over Mach 2 and the tempera-
tures will be even higher than what titani-
4.1 Introduction 127
um alloys can withstand. Titanium alu-
minides are one of the candidate materials
in this case.
Titanium has two polymorphs; alpha (a)
titanium has an h.c.p. structure and is sta-
ble below 885 °C, and beta (p) titanium
which has a b.c.c. structure and is stable
above 885 °C. Aluminum raises the a -• p
transformation temperature, i.e., alu-
minum is an a stabilizer. Most other alloy-
ing elements (Fe, Mn, Cr, Mo, V, Nb, Ta)
lower the a -> P transformation tempera-
ture, i.e., they stabilize the phase. Thus,
three general alloy types can be produced,
viz., a, a+(3, and P titanium alloys. The
Ti-6%-Al-4%-V, called the work horse Ti
alloy of the aerospace industry, belongs to
the (oc + P) group. Most titanium alloys are
not used in a quenched and tempered con-
dition. Generally, hot working in the
(a + p)-region is carried out to break up the
structure and distribute the a phase in an
extremely fine form.
Titanium has a great affinity for oxygen,
nitrogen, and hydrogen. Parts per million
of such interstitials in titanium can change
mechanical properties drastically, produc-
ing embrittlement. That is why welding of
titanium by any technique requires protec-
tion from the atmosphere. Electron beam
techniques, in vacuum, are frequently
used.
Magnesium alloys
Magnesium and its alloys form another
group of very light materials. Magnesium
is one of the lightest metals, its density
being 1.74 g cm" 3 . Magnesium alloys, es-
pecially castings, are used in aircraft gear-
box housings, chain saw housing, electron-
ic equipment, etc. Magnesium, being a
hexagonal close packed metal, is difficult
to cold work.
128 4 Metal Matrix Composites
Copper
Copper has a face centered cubic struc-
ture. Its use as an electrical conductor is
quite ubiquitous. It has good thermal con-
ductivity. It can be cast and worked easily.
One of the major applications of copper in
a composite as a matrix material is in nio-
bium based superconductors.
Intermetallic Compounds
There are many other intermetallics, not
necessarily ordered, available. Ordered in-
termetallic alloys possess structures char-
acterized by long range ordering, i.e., dif-
ferent atoms occupy specific positions in
the lattice. Because of their ordered struc-
ture, dislocations in intermetallics are
much more restricted than in disordered
alloys. This results in retention (and, in
some case, even an increase) of strength at
elevated temperatures, a very desirable
feature. For example, nickel aluminide ex-
hibits a marked increase in strength with
temperature up to 800 °C. An undesirable
feature of intermetallics is their extremely
low ductility. Among the explanations put
forth to explain this are:
(a) Availability of an insufficient num-
ber of slip systems because of the low sym-
metry of the structure;
(b) Weak grain boundaries.
Attempts at ductility enhancement in in-
termetallics have involved a number of
metallurgical techniques. Rapid solidifica-
tion is one method. Another technique
that has met great success is the addition of
boron to Ni 3 Al. Boron segregates to the
grain boundaries, strengthening them.
With extremely small amounts of boron
(0.06 wt.%), the ductility increases from
about 2% to about 50%. Long range order
also has significant effects on diffusion
controlled phenomena such as recovery,
recrystallization, and grain growth. The
activation energy for these processes is in-
creased, and these processes are slowed
down. Thus, ordered intermetallic com-
pounds tend to exhibit a high creep resis-
tance. Enhanced toughness by making
composites with intermetallic matrix mate-
rials is a potential possibility.
Chief among the disordered intermetal-
lics is molybdenum disilicide (MoSi2) (Va-
sudevan and Petrovic, 1992). It has a high
melting point and shows good stability at
temperatures greater than 1200 °C, in oxi-
dizing atmospheres. It is commonly used
as a heating element in furnaces. The high
oxidation resistance comes from a protec-
tive SiO2 film that it tends to form at high
temperatures.
4.2 Processing
Many processes for fabricating metal
matrix composites are available. For the
most part, these processes involve process-
ing in the liquid and solid state. Some pro-
cesses may involve a variety of deposition
techniques or an in situ process of incorpo-
rating a reinforcement phase. We provide a
summary of these fabrication processes.
4.2.1 Liquid State Processes
Casting or liquid infiltration involves in-
filtration of a fiber bundle by liquid metal.
Alternatively tows of fibers can be passed
through a liquid metal bath, where the in-
dividual fibers are wet by the molten met-
al, wiped of excess metal, and a composite
wire is produced. A bundle of such wires
can be consolidated by extrusion to make
a composite. Relatively few MMCs have
been made by simple liquid phase infiltra-
tion, mainly because of difficulties with fi-
ber wetting by the molten metal. When the
infiltration of a fiber preform occurs read-

ily, reactions between the fiber and the
molten metal can significantly degrade fi-
ber properties. Fiber coatings applied pri-
or to infiltration, which improve wetting
and control reactions, have been devel-
oped and can produce impressive results.
In this case, however, the disadvantage is
that the fiber coatings must not be exposed
to air prior to infiltration since surface ox-
idation will alter the positive effects of
coating (Katzman, 1987).
Squeeze casting or pressure infiltration
involves forcing the liquid metal into a fi-
brous preform. Pressure is applied until
the solidification is complete. Figure 4-2
shows a schematic of this process. By forc-
ing the molten metal through small pores
of a fibrous preform, this method obviated
the requirement of good wettability of the
reinforcement by the molten metal. Com-
posites fabricated with this method have
minimal reaction between the reinforce-
ment and molten metal, and are free from
common casting defects such as porosity
and shrinkage cavities. Infiltration of a fi-
brous preform by means of a pressurized
inert gas is another variant of liquid metal
infiltration technique. Net-shaped compo-
nents can be produced inexpensively
(Cook and Warner, 1991). The process is
conducted in the controlled environment
of a pressure vessel and rather high fiber
volume fractions and complex shaped
structures are obtainable (Cook and
Warner, 1991). Alumina fiber reinforced
intermetallic matrix composites (e.g., TiAl,
Ni 3 Al, and Fe3Al matrix materials) com-
posites have been prepared by pressure
casting or squeeze casting (Nourbakhsh
et al. 1990). The matrix alloy is melted in a
crucible in vacuum while the fibrous pre-
form is heated separately. The molten ma-
trix material (at about 100 °C above the
Tm) is poured onto the fibers and argon gas
is introduced simultaneously. Argon gas
129
Molten Metal
Heater
mmk
MBS*
WHO*
Fiber preform mum
Mold
Figure 4-2. Squeeze casting or pressure infiltration
process. The liquid metal is forced into a fibrous pre-
form. Pressure is applied until the solidification is
complete.
pressure forces the melt to infiltrate the
preform. The melt generally contains addi-
tives to aid wetting of the fibers.
4.2.2 Solid State Processes
Diffusion Bonding
Diffusion bonding is a common solid
state welding technique for joining similar
or dissimilar metals. Interdiffusion of
atoms from clean metal surfaces in contact
at an elevated temperature leads to weld-
ing. The major advantages of this tech-
nique are: (i) ability to process a wide vari-
ety of matrix metals, (ii) control of fiber
orientation and volume fraction. Among
the disadvantages are: (i) processing times
of several hours, (ii) cost of high process-
ing temperatures and pressures, (iii) only
objects of limited size can be produced.
130 4 Metal Matrix Composites

There are many variants of the basic diffu-
sion bonding process; however, all of them
involve a simultaneous application of pres-
sure and high temperature. Matrix alloy
foil and fiber arrays, or composite wire, or
monolayer laminae are stacked in a prede-
termined order. Figure 4-3 shows a sche-
matic of one such diffusion bonding pro-
cess, while Fig. 4-4 shows the microstruc-
ture of a transverse section of silicon car-
bide fiber/aluminum composite made by
diffusion bonding. Vacuum hot pressing is
a most important step in the diffusion
bonding processes for metal matrix com-
posites. Hot isotopic pressing (HIP), in-
stead of uniaxial pressing, can also be
used. Hipping involves gas pressure push-
ing against a can which consolidates the
composite piece contained inside the can.
With HIP it is relatively easy to apply high
pressures at elevated temperatures over
variable geometries. Hot roll diffusion
bonding is another variant of diffusion
bonding technique which can be used to
produce a composite consisting of differ-
ent metals in the sheet form. Such com-
posites are called sheet laminated metal
matrix composites.

(1) Apply metal foil

(5) Heat to fabrication temperature

(2) Cut to shape

ULL
:T H K H n n n n.

(3) Lay up desired plies

(6)Apply pressure and hold for

consolidtion cycle

M M U M )l il Li~i

(4) Vacuum encapsulate

OOOOOOOO
OOOOOOOO
OOOOOOOO
OOOOOOOO
Figure 4-3. Schematic of a
(7) Cool, remove, and clean part diffusion bonding process.

Figure 4-4. Microstructure of a transverse section of
silicon carbide fiber/aluminum composite made by
diffusion bonding.
Deformation processing
Deformation processed metal/metal
composites are formed during mechanical
processing (by swaging, extrusion, draw-
ing, or rolling) of a ductile two-phase ma-
terial. The two phases co-deform, causing
the minor phase to elongate and become
fibrous in nature within the matrix. These
materials are sometimes referred to as in
situ composites. The properties of a defor-
mation processed composite depend large-
ly on the characteristics of the starting ma-
terial which is usually a billet of two-phase
alloy that has been prepared by casting or
powder metallurgy methods. Deformation
processed composites are limited to start-
ing materials in which both phases are duc-
tile and have similar flow stresses so that
co-deformation occurs and large total de-
formation strains may be employed.
Niobium based conventional filamen-
tary superconductors in a copper matrix
and the oxide based high-Tc superconduc-
tors in a silver matrix constitute a special
4.2 Processing 131
class of deformation processed metal ma-
trix composites. Conventional niobium
based superconductors, niobium/titanium
filaments in a copper matrix and Nb 3 Sn
filaments in a copper matrix, constitute
perhaps the largest commercial market for
continuous fiber MMCs. High-Tc super-
conductors were hailed in 1987 as a revolu-
tionary discovery. It was pointed out at
that time (Chawla, 1987 a, p. 164) that
eventually some kind of filamentary route
will have to be used for making useful
components out of these materials. There
are two main problems with high-Tc super-
conductors that any process technique
must contend with. These are:
(i) High-Tc superconductors are diffi-
cult to make into desirable forms such as
wires or films, especially wires.
(ii) They have a very low current carry-
ing capacity. This has been attributed to
grain boundaries as weak links and flux
creep.
We shall come back to the high-r c su-
perconductors shortly. We first describe
the processing of niobium based copper
matrix composite superconductors (Foner
and Schwarz, 1981; Hillmann, 1981; Hill-
mann et al., 1989). There are two main
types of the conventional niobium based
superconductors: Nb-Ti/Cu and Nb 3 Sn/
Cu. Niobium-titanium (about 50-50)
form a ductile system. Rods of Nb-Ti are
inserted in holes drilled in a block of cop-
per, evacuated, sealed, and subjected to a
series of drawing operations interspersed
with appropriate annealing treatments to
give the final composite superconducting
wire. This process is fairly straight for-
ward. In the case of Nb 3 Sn/Cu, an inge-
nious process called the bronze route is
used to make this composite. Nb 3 Sn, being
an A-15 type, brittle intermetallic, cannot
be processed like Nb-Ti. Instead, one
starts with a bronze (Cu-13 % Sn) matrix.
132 4 Metal Matrix Composites
Pure niobium rods are inserted in holes
drilled in bronze, evacuated, sealed, and
subjected to wire drawing operations as in
the case of Nb-Ti/Cu. The critical step is
the heat treatment at about 700 °C that
drives out the tin from the bronze matrix
to combine with niobium to form stoichio-
metric, superconducting Nb 3 Sn, leaving
behind copper matrix.
In view of the highly important and pub-
licized discovery of high temperature su-
perconductivity, made in 1987, we digress
briefly to provide some background on the
subject and then come to the processing
problem of these superconductors. Oxide
superconductors, (Y, Ba, Cu)-oxide or the
so-called 1-2-3 compound ceramic super-
conductors, with a transition temperature
around 90 K were discovered in February,
1987. Such a high transition temperature
enables the use of a cheap and easily avail-
able coolant, viz., liquid nitrogen (boiling
point 77 K). Since the discovery of these
high-Tc superconductors, a lot of hyper-
bole has been made of the potentialities
and possibilities of these superconductors.
There is no gainsaying the fact that the
discovery of the high temperature super-
conductivity does represent an extremely
important discovery. In all fairness, how-
ever, it should be recognized that these
high-Tc superconductors have a number of
problems to be ironed out before they
come anywhere close to the conventional
superconductors in terms of applications.
One of the major problems is that these
superconductors carry very low current
densities. The electric resistance goes to ze-
ro at Tc« 90 K, but the supercurrents seem
to have a great difficulty in going from one
grain to another. This may be due to impu-
rities or misshaped grains. The trouble-
some fact remains that these materials lose
their superconductivity at very modest
current densities of about a few 100 A/
cm2. Also, their performance deteriorates
as the applied magnetic field increases. At
a magnetic field of 1 tesla, the Jc«1 -10 A/
cm2, whereas one would like to have a
/ c ^ 105 A/cm2. The inherent brittleness of
ceramics is another problem in these oxide
superconductors. Due to the phenomenon
of Lorentz force, extremely high tensile
and shear forces develop in dipole magnets
that would shatter the brittle ceramic ox-
ides. This is particularly true for large
magnets which are like pressure vessels
wherein the pressure on the superconduc-
tor due to the Lorentz force goes as square
of the magnetic field. It is easy to see that
in such a system the mechanical character-
istics of the composite superconductor will
dictate the design.
Critical current density, / c , of a super-
conductor is a function of material pro-
cessing. The standard process for making
(Y, Ba, Cu)-based superconductors is to
take oxide powders of these elements, cold
compact them and then sinter. All that
such a material can be used for is to
demonstrate the phenomenon of high tem-
perature superconductivity. The main rea-
sons are the ones given above, viz., low
current carrying capacity and brittleness.
Jin (1991) used a technique called melt tex-
tured growth technique which gave impres-
sive current density values. This technique
involves melting at 1300°C and cool back
to room temperature via a proprietary reg-
imen, which results in needle-like crystals
(about 100 \im long). These needle-like
grains are aligned (locally) in the grain
conduction direction. 1-2-3 supercon-
ductors are highly anisotropic with their
ability to superconduct. Their current den-
sity results are as follows:
/ c « 7400 A/cm2 at £ =0
/ c « 1000 A/cm2 at B=\T

In a manner similar to the so-called
bronze route of making conventional su-
perconductors described above, where the
superconductor is a very brittle intermetal-
lic, Nb 3 Sn, one must avoid the synthesis of
the brittle ceramic until the very end. The
sintering step should be carried after the
formation of the desired shape. This tech-
nique involves loading the superconduct-
ing powder in the elemental form into a
metal tube, draw into a fine wire, and wind
into a coil before sintering. For example, in
one technique, the metallic elements (a
rare earth, Ba, Cu) are melted and solidi-
fied as a ternary metal alloy by melt spin-
ning. Melt spinning gives a rapidly solidi-
fied alloy in a ribbon form. The ribbon is
then shaped and thereupon heated in oxy-
gen to convert into an oxide superconduc-
tor. (Jin, 1991; Singh and Narayan, 1991;
Sandhage e t a l , 1991).
A process that is increasingly gaining
importance is called the Oxide-Powder-In-
Tube (OPIT) method (Sandhage etal.,
1991). In this process, the oxide powder of
appropriate composition (stoichiometry,
phase content, impurities, etc.) is packed
inside a metal tube (generally silver). This
assembly is first deformation processed in-
to the desired geometry (round wire or flat
tape) and then given an annealing treat-
ment to obtain the superconductor com-
posite. Intermediate anneals are given dur-
ing deformation processing. This OPIT
process is very similar to the bronze route
of making Nb 3 Sn/Cu composite supercon-
ductors.
Silver is the most common metal tubing
material for the following reasons:
(i) relatively high oxygen diffusivity
(ii) good thermal conductivity
(iii) relatively low reactivity with bis-
muth cuprate-based superconducting ox-
ides.
4.2 Processing 133
The oxide powder is packed in silver
tube (4-7 mm inside diameter and 6-
11 mm outside diameter). The silver tube is
sealed after packing and degassed. The de-
formation processing techniques common-
ly used are swaging and drawing for wires,
and rolling is used for tapes. Heat treat-
ments, intermediate and/or subsequent to
deformation are given to form the correct
phase, promote grain interconnectivity,
crystallographic alignment of the oxide,
and to obtain proper oxygenation (Sand-
hage, etal., 1991).
Powder Processing
Powder metallurgy methods involving
cold pressing and sintering, or hot pressing
can be used to fabricate metal matrix com-
posites. A number of powder processing
techniques are used for making particle or
whisker reinforced MMCs. The matrix
and the reinforcement powders are blend-
ed to produce a homogeneous distribu-
tion. Most of the information on this criti-
cal step is considered to be proprietary.
The blending stage is followed by cold
pressing to produce what is called a green
body, which is about 80% dense and can
be easily handled. The cold pressed, green
body is canned in a sealed container and
degassed to remove any absorbed moisture
from the particle surfaces. The final step is
hot pressing, uniaxial or isostatic, to pro-
duce a fully dense composite. The hot
pressing temperature can be either below
or above the matrix alloy solidus.
Continuous filament reinforced com-
posites can also be made by this technique.
Fiber tows are infiltrated by dry matrix
powder, followed by hot isostatic pressing.
Alternatively, matrix alloy powder that
has been dry blended with short whiskers,
chopped filament, or particles is hot
pressed to produce a composite material
134 4 Metal Matrix Composites
with a random fiber orientation. This ma-
terial can become partially oriented by hot
mechanical deformation, usually extrusion
or hot forming. Cold pressing and sinter-
ing is a less desirable process since the high
pressure needed to press the powders to
the required density can break the fibers,
or the sintering process can degrade the
fibers.
4.2.3 Deposition Techniques
Deposition techniques for metal matrix
composite fabrication involve coating in-
dividual fibers in a tow with the matrix
material needed to form the composite,
followed by diffusion bonding to form a
consolidated composite plate or structural
shape. Since the composite is composed of
identical units, the microstructure will be
more homogeneous than cast composites.
The main disadvantage of using deposition
techniques is that they are time consuming.
However, there are several advantages:
- The degree of interfacial bonding is eas-
ily controllable; interfacial diffusion
barriers and compliant coatings can be
formed on the fiber prior to matrix de-
position, or graded interfaces can be
formed.
- Filament wound thin, monolayer tapes
can be produced which are easier to han-
dle and mold into structural shapes than
other precursor forms; unidirectional or
angle-plied composites can be easily fab-
ricated in this way.
Several deposition techniques are avail-
able: immersion plating, electroplating,
spray deposition, chemical vapor deposi-
tion (CVD), and physical vapor deposition
(PVD). Dipping or immersion plating is
similar to infiltration casting except that
fiber tows are continuously passed
through baths of molten metal, slurry, sol,
or organometallic precursors. Electrode-
position or electroplating produces a coat-
ing from a solution containing the ion of
the desired material in the presence of an
electric current. Fibers are wound on a
mandrel, which serves as the cathode, and
placed into the plating bath with an anode
of the desired matrix material. The advan-
tage of this method is that the tempera-
tures involved are moderate and no dam-
age is done to the fibers. Problems with
electroplating involve void formation be-
tween fibers and between fiber layers, ad-
hesion of the deposit to the fibers may be
poor, and there are limited numbers of al-
loy matrices available for this processing.
A spray deposition operation typically
consists of winding fibers onto a foil coat-
ed drum and spraying molten metal onto
them to form a monotape. The source of
molten metal may be powder or wire feed-
stock which is melted in a flame, arc, or
plasma torch. The advantages of spray de-
position are easy control of fiber alignment
and rapid solidification of the molten ma-
trix. In a CVD process, a vaporized com-
ponent decomposes or reacts with another
vaporized chemical on the substrate to
form a coating on that substrate. The pro-
cessing is generally carried out at elevated
temperatures. The product phase can be
homogeneously nucleated if it occurs in
the vapor phase or heterogeneously nucle-
ated if it occurs on a substrate surface.
Metallic, oxide, carbide, or nitride coat-
ings or films can be produced with CVD.
These films can be grown amorphous, sin-
gle crystal, or polycrystalline form. When
the matrix material is developed on a rein-
forced preform, such as a woven mat, it is
called chemical vapor infiltration (CVI).
The three basic PVD processes for creating
metal matrix composites differ mainly in
how the vapor is generated: evaporation,
ion plating, and sputtering. Evaporation
methods for matrix materials under high
vacuum, generally contained in a crucible,
include resistive heating, electron beam
evaporation, arc evaporation, radiation
heating, or laser ablation. Ion plating in-
volves passing the vaporized component
through an argon gas glow discharge
around the substrate which ionizes some of
the vapor molecules. These ionized atoms
subsequently deposit onto the substrate.
Sputtering techniques involve vaporiza-
tion of the coating material (the target or
cathode) via a momentum transfer from
an ionized argon gas molecule in a glow
discharge around the cathode. The prima-
ry advantage of the PVD is versatility in
the compositions and micro structure of
the coating produced. These processes can
produce coatings with superior bonding to
the substrate and excellent surface finishes.
In addition, no chemical reactions by
products are produced by these processes.
The disadvantage is that the processing
equipment is complex and expensive.
Spray-forming of Particulate MMCs
A relatively new and promising devel-
opment for making particle reinforced
MMCs deserves a separate description. It
involves the use of spray techniques that
have been used for some time to produce
monolithic alloys (Willis, 1988). One par-
ticular example of this, a cospray process,
is shown in Fig. 4-5. This process uses a
spray gun to atomize a molten aluminum
alloy matrix, into which heated (to dry
them) silicon carbide particles are injected.
An optimum particle size is required for
transfer. Whiskers, for example, are too
fine to be transferred. The preform pro-
duced in this way can be up to about 97%
dense. The metal matrix composite is sub-
jected to scalping, consolidation, and sec-
ondary finishing processes. The product is
a wrought material. The process is totally
Manufactyre
of Billets
1 Tundish
2 Gas atomi/<?r
3 Particle injection
4 Spray chamber
5 Product
6 Exhaust i?ar> . """"" ' 6 '
Figure 4-5. Cospray process of making an MMC. A
spray gun atomizes a molten alloy matrix, into which
heated ceramic particles are injected.
computer controlled and is quite fast. It
should be noted that the process is essen-
tially a liquid metallurgy process. One
avoids the formation of deleterious reac-
tion products because the time of flight is
extremely short. Silicon carbide particles
of an aspect ratio (length/diamter) be-
tween 3-4 and volume fractions up to
20% have been incorporated into alu-
minum alloys. Alcan can cast 200 kg in-
gots at a rate of 6-10kg/min (Lloyd,
1992).
A great advantage of the process is the
flexibility that it affords in making differ-
ent types of composites. For example, one
can make in situ laminates using two
sprayers. One can have selective reinforce-
ment. This process, however, is more ex-
pensive because of the expensive capital
equipment.
4.2.4 In Situ Processes
In these techniques, as the name sug-
gests, one forms the reinforcement phase
136 4 Metal Matrix Composites
in situ. The composite material is pro-
duced in one step from an appropriate
starting alloy, e.g., Co-NbC, Ni-TiC, etc.,
thus avoiding the difficulties inherent in
combining the separate components as
done in a typical composite processing.
Controlled unidirectional solidification of
a eutectic alloy is a classic example of in
situ processing. Unidirectional solidifica-
tion of a eutectic alloy can result in one
phase being distributed in the form of
fibers or ribbon in the other. One can con-
trol the fineness of distribution of the rein-
forcement phase by simply controlling the
solidification rate. The solidification rate
in practice, however, is limited to a range
a of 1 - 5 cm/h because of the need to main-
tain a stable growth front. The stable
growth front requires a high temperature
gradient.
Some of the newer processes for produc-
ing metal matrix composites include Lanx-
ide's PRIMEX, pressureless metal infiltra-
tion processes, and Martin Marrietta's XD
self propagating reaction process. A brief
description of these follows.
The Lanxide Process, called PRIMEX,
involves a pressureless infiltration of metal
into a ceramic preform. The ceramic pre-
form can be prepared by tape casting,
molding, or machining of a green body.
The metal or alloy ingot is placed on the
ceramic preform in a graphite mold and
the assembly is heated in a nitrogen atmo-
sphere, so that the molten metal sponta-
neously infiltrates the ceramic preform. It
appears that some specific aluminum-mag-
nesium alloy compositions must be used in
a nitrogen atmosphere in order to get wet-
ting and infiltration of silicon carbide pre-
form by the molten alloy.
The XD process is another in situ pro-
cess which uses an exothermic reaction be-
tween two components to produce a third
component. Sometimes such processing
techniques are referred to as Self Propagat-
ing High Temperature Synthesis (SHS)
processes. Specifically, the XD Process is a
proprietary process developed by Martin
Marietta Corp. to produce ceramic parti-
cle reinforced metallic alloy. Generally, a
master alloy containing rather high vol-
ume fraction of reinforcement is produced
by the reaction synthesis. This is mixed
and remelted with the base alloy to pro-
duce a desirable amount of particle rein-
forcement. Typical reinforcements are SiC,
TiC, TiB 2 , etc. in an aluminum, nickel, or
an intermetallic matrix. Figure 4-6 gives a
schematic of this process, while Fig. 4-7
shows the microstructure of a 15%-vol-
ume fraction TiC/Al composite made by
this process, followed by extrusion.
Matrix and reinforcement precursor materials
A+B+C+X+Y+Z
Mixing Heating
(^Exothermic reaction"^
Metallic or intermetallic
matrix with intermetallic
or ceramic reinforcements,
as particles or whiskers
A + B + A B + C + BC + XY + Y Z + . . .
Reinforcements
Figure 4-6. Schematic of an SHS process of making
an MMC.
 4.3 Interfaces in Metal Matrix Composites
Figure 4-7. Microstructure of a 15% volume fraction
TiC/Al composite made by the XD process, followed
by extrusion.
4.3 Interfaces in Metal Matrix
Composites
One of the objectives of this article is to
highlight the importance of the interface in
metal/ceramic composites, discuss the dif-
ferent types of interfacial bonding, mi-
crostructural characterization of the inter-
face region, and the effects of interface on
composites properties. We shall do this by
giving some specific examples from repre-
sentative composites.
The interface region in a given com-
posite has a great deal of importance in
determining the ultimate properties of the
composite. An interface is, by definition, a
bidimensional region through which there
occurs a discontinuity in one or more ma-
terial parameters. In practice, there is al-
ways some volume associated with the in-
terface region over which a gradual transi-
tion in material parameter(s) occurs. The
importance of the interface region in com-
posites stems from two main reasons: (i)
the interface occupies a very large area in
composites, and (ii) in general, the rein-
forcement and the matrix will form a sys-
137
tem that is not in thermodynamic equilib-
rium. In this seciton, I give examples of
interface microstructure in different metal
matrix composite systems (particle and fi-
ber reinforced) and discuss the implica-
tions of various interface characteristics on
the resultant properties of the composite.
We can define an interface as a bound-
ing surface between two phases where a
discontinuity in some material parameters
occurs. Among the important discontinu-
ities are: elastic moduli, thermodynamic
parameters such as chemical potential, and
the coefficient of thermal expansion. The
discontinuity in chemical potential is likely
to cause chemical interaction, leading to
an interdiffusion zone or a chemical com-
pound formation. The discontinuity in the
thermal expansion coefficient means that
the interface will be in equilibrium only at
the temperature which the reinforcement
and the matrix were brought into contact.
At any other temperature, there will be
present biaxial or triaxial stress fields be-
cause of the thermal mismatch between the
components of a composite. Thermal
stresses due to a thermal mismatch will
generally have an expression of the form
<j=f(E,a,b,r)A(xAT (4-1)
where/(ii, a, b, r) is a function of the elastic
constants, E, and geometric parameters, a,
b, and r to be defined later, Aa is the differ-
ence in the expansion coefficient of the
components, and AT is the temperature
change. The term (Aa AT") is, of course, the
thermal strain. It what follows, I describe
the importance of the interface, its crystal-
lographic nature, bonding type and mi-
crostructure of the interface in some im-
portant composites.
138 4 Metal Matrix Composites
4.3,1 Importance of the Interface
in Composites
The importance of the interface in com-
posites is recognized on all sides. The fol-
lowing simple exercise will help bring
home this importance. For a given volume
fraction, the total area of fiber/matrix in-
terface increases as the fiber diameter de-
creases. We can visualize this in a very sim-
ple manner. Consider a composite bar con-
taining N continuous, unidirectionally
aligned fibers. Let the length, width, and
height of the composite be /, w, and /z,
respectively, and let the fiber diameter be
d. Then, we can write
volume of fibers Nnd2
Vf = (4-2)
volume of composite /wh
and area of the fiber/matrix interface
IA = Nndl (4-3)
From Eqs. (4-2) and (4-3), we get
u-A-^y, (4-4)
Let the volume of composite be 1 m3. Let
us take the fiber volume fraction to be 0.25.
Then, Eq. (4-4) reduces to
/.-i(m') (4-5)
Equation (4-5) says that the fiber/matrix
interfacial area, for a given fiber volume
fraction, varies inversely with the fiber di-
ameter. We can do a similar calculation for
a particle reinforced composite. Assume N
spherical ceramic particles embedded in a
spherical shell of the matrix. Let the total
volume of the composite be 1 m3. Then, we
can write
volume of particles
p
volume of composite
N(n/6)d3
(4-6)
i faces, i.e., the lattice mismatch between the
j A = particle/matrix interface area =
= Nnd2 (4-7)
J (4-8)
d
Taking Vp = 0.25, we have
(4-9)
Comparing Eqs. (4-5) and (4-9), we find
that for a given diameter and volume frac-
tion of reinforcement, a particulate com-
posite will have a larger interfacial area
than a fibrous composite. However, the
more important conclusion to be drawn
from Eqs. (4-5) and (4-9) is that the in-
terfacial area increases with decreasing re-
inforcement diameter. It is easy to visual-
ize that the interfacial area can indeed be-
come very large for reinforcements less
than 10-20 ^im in diameter. Since chemi-
cal and/or mechanical interactions be-
tween reinforcement and matrix occur at
interfaces, an extremely large area of inter-
face has a tremendous importance in deter-
mining the final properties and perfor-
mance of a composite.
4.3.2 Crystallographic Nature
of the Fiber/Matrix Interface
In crystallographic terms, ceramic/metal
interfaces in composites are, generally, in-
coherent and high-energy interfaces. Ac-
cordingly, they can act as very efficient va-
cancy sinks, provide rapid diffusion paths,
segregation sites, sites of heterogeneous
precipitation, as well as sites for precipitate
free zones. Among the possible exceptions
are the eutectic composites (Chawla,
1987 b) and the newer XD type particulate
composites (Wang and Arsenault, 1991;
Mitra et al., 1991). The in situ or eutectic
composites do show semi-coherent inter-
 4.3 Interfaces in Metal Matrix Composites 139

matrix phase and the reinforcement phase

is accommodated by creating a network of
dislocations. Figure 4-8 shows an example
of a network of interface dislocations be-
tween NiAl-matrix and Cr-rod in a direc-
tionally solidified NiAl-Cr eutectic (Cline
etal., 1971).
The XD process gives particulate com-
posites with stable and clean interfaces
(Lewis, 1991). Wang and Arsenault (1991)
used HREM to study interfaces in some
XD composites. They observed semicoher- 10011
WPHji
ent interfaces between OC-A1203 and NiAl,
but incoherent interface between NiAl and .._„-.-~IO1O)
TiB2.
0.4 fim
4.3.3 Interfacial Bonding Figure 4-8. Network of interface dislocations between
in Metal Matrix Composites NiAl-matrix (light area) and Cr-rod in a directionally
solidified NiAl-Cr eutectic (Cline et al., 1971).
Some bonding must exist between the
ceramic reinforcement and the metal ma-
trix for load transfer from matrix to fiber
trix. As it happens, these factors frequently
to occur. Neglecting any direct loading of
overlap, and it may not always be possible
the reinforcement, the applied load is
to isolate these effects.
transferred from the matrix to the rein-
forcement via a well-bonded interface.
4.3.3.1 Mechanical Bonding
Here, it is important to distinguish be-
tween wettability and bonding. Wettability Mechanical keying effect between two
refers to ability of a liquid to spread on a surfaces can lead to bonding. Hill e t a l .
solid. Frequently, the contact angle is tak- (1969) confirmed this experimentally for
en as a measure of wettability, a contact tungsten filaments in an aluminum matrix
angle of 0° indicating perfect wetting, and while Chawla and Metzger (1978) ob-
a contact angle of 180° indicating no wet- served mechanical gripping effects at
ting. Wettability just tells about the possibil- A1 2 O 3 /A1 interfaces. Hill et al. etched
ity of attaining an intimate contact between tungsten wires along a portion of their
a liquid or solid. Good wetting is a neces- length to produce a rough interface. The
sary but not sufficient condition for strong tungsten filaments were incorporated into
bonding. One needs a good wetting even an aluminum matrix by the liquid metal
for purely mechanical bonding or weak infiltration technique in vacuum. They
van der Waals bonding, otherwise voids evaluated three interface conditions by
may form at the interface. Besides wet- longitudinal tensile tests of composites. In
tability, there are other important factors the case of a smooth surface, a chemical
such as chemical, mechanical, thermal, reaction between aluminum matrix and
and structural that will affect the nature of tungsten led to a high strength. In the case
bonding between reinforcement and ma- of a smooth surface with a graphite layer,
140 4 Metal Matrix Composites

the graphite barrier layer prevented the re- Polished substrate

action from taking place, that is, there was
no chemical bonding, and because the in- Steep sided pits
terface was smooth (no roughness), there 106 per cm2
was also very little mechanical bonding.
The resultant strength was, therefore, very Gently sloping pits
low. In the case of the etched wires with a 103 per cm2
70 pm
graphite layer, there was no reaction bond-
ing, but there was a mechanical keying ef-
fect because of the rough surface produced 100
by etching. This mechanical bonding re-
stored the strength to the level achieved 90
with chemical reaction at the interface.
Chawla and Metzger (1978) artificially
produced interfacial roughness by etchpit-
ting and microhardness indentation. Fig-
ure 4-9 shows plots of number of cracks
per mm (crack density) vs. percent strain.
Cracks in the alumina film first appeared
at about the same strain and initially the
crack density rose at about the same rate
for a rough and a smooth surface. Howev-
er, for a smooth interface, the crack densi-
ty remained constant above 8% while for
a rough interface, the crack density contin-
ued to increase beyond this value, i.e., in
the case of a rough interface the load trans-
fer continued to higher strain values than 5 10 15 20
Elongation (%)
in the case of a smooth surface. Similar
mechanical bonding effects due to inter- Figure 4-9. Number of cracks per mm (crack density)
vs. percent strain in an alumina/aluminum composite
face roughness have been observed in bab- for different degrees of roughness (Chawla and Metz-
bitt-bronze composites (Liaw et al., 1990). ger, 1978).
It should be mentioned that in fiber rein-
forced composites mechanical bonding
will be effective mostly in the longitudinal able kinetic conditions (which in practice
or fiber direction. Transverse to the fiber, it means a high enough temperature or long
will provide only a minor effect. enough time), diffusion and/or chemical
reactions will occur between the compo-
nents. The interface layer(s) formed be-
4.3.3.2 Chemical Bonding
cause of such a reaction will generally have
Most metal matrix composite systems characteristics different from those of ei-
are nonequilibrium systems in the thermo- ther one of the components. It should be
dynamic sense; that is, there exists a chem- pointed out, however, that at times, some
ical potential gradient across the fiber/ma- controlled amount of reaction at the inter-
trix interface. This means that given favor- face may be desirable for obtaining strong
 4.3 Interfaces in Metal Matrix Composites 141

bonding between the fiber and the matrix, inforcement and the matrix. This thermal
but, too thick an interaction zone will ad- mismatch can lead to thermal stresses large
versely affect the composite properties. enough to cause plastic deformation in the
Ceramic/metal interfaces are generally matrix. Plastic deformation in the matrix
formed at high temperatures. Diffusion means introduction of defects such as dis-
and chemical reaction kinetics are faster at locations, vacancies, etc. in the matrix, es-
elevated temperatures. One needs to have pecially in the interface region. The intro-
knowledge of the chemical reaction prod- duction of such defects can and do affect
ucts and, if possible, their properties. It is the phenomena responsible for chemical
thus imperative to understand the thermo- reactions at the interface as well as charac-
dynamics and kinetics of reactions such teristics of the matrix such as the precipita-
that processing can be controlled and opti- tion kinetics. A brief review of the thermal
mum properties can be obtained. stress effects is therefore presented in the
As pointed out above, interfacial bond next section.
degradation can occur by an adverse
chemical reaction between the reinforce- 4.3.4 Thermal Stresses
ment and the matrix. Therefore, it is im- in Metal Matrix Composites
portant to characterize the fiber/matrix in-
In general, ceramic reinforcements
terfacial reaction zone. One generally does
(fibers, whiskers, or particles) have a coef-
this by studying a diffusion couple made
ficient of thermal expansion smaller than
of the components of the composite with
that of most metallic matrices. This means
a planar interface. Diffusion controlled
that when the composite is subjected to a
growth in an infinite diffusion couple with
temperature change, thermal stresses will
a planar interface follows the relationship
be generated in both the components.
x2 = Dt (4-10) A simple two-element model shown in
Fig. 4-10 can be used to make a thermo-
where x is the thickness of the reaction
elastic analysis. It can be shown that for
zone, D is the diffusivity, and t is the time.
For a composite containing cylindrical
fibers of small diameter, the diffusion dis-
tance is very small. However, to a first ap-
proximation, we can write
x2 = kt (4-11)
where k is a material constant and has the
dimensions of diffusivity (m 2 s~ 1 ). Such
parabolic relationships between the reac-
tion zone thickness and time have been
observed by many researchers (see for ex-
ample DiCarlo, 1983; Schulyer et al.,
1990).
A factor that ends up complicating the
simple chemical incompatibility problem Figure 4-10. Simple two-element model for thermo-
at the interface is the ever present thermal elastic analysis, a and b are the radii of the fiber and
expansion incompatibility between the re- matrix, respectively.
142 4 Metal Matrix Composites

this unit fiber composite, using cylindrical of the cylinder, b is the radius of the matrix
coordinates, we have the following princi- shell, and the subscripts m and f denote the
pal stresses in the matrix (Poritsky, 1934; matrix and the fiber, respectively. Figure
Hull and Burger, 1934; Chawla, 1987 c): 4-11 shows the thermoelastic stress distri-
bution for an alumina fiber/magnesium
b2 composite with 35% volume fraction of
(4-12)
fibers as a function of a normalized dis-
tance, r/a, with r/a = 1 representing the fi-
-AU+Z (4-13) ber/matrix interface. Note that or and ae
fall off rapidly with the distance, r. It turns
= B (4-14)
2
Em(am-al)AT(a/b)
A (4-15)
• - [ • l+(a/b)2(i-2v)[(b/a)2-l]EJE
1 + (a/b)2 (1 - 2 v) + (g/b)2 (1 - 2 v) [(b/a)2 - 1 ] g j ^ l
D
(4-16)
l+[(ft/a) 2 -l]£ m /£ f J
where a is the coefficient of linear expan-
sion, v is the Poisson ratio (assuming out that or and oe are relatively insensitive
vf = vm = v), E is the Young's modulus, AT to fiber volume fraction while az depends
is the temperature interval, a is the radius on the fiber volume fraction.

1600 1 1 1 1 1 I ' I • J

1400 ,/Mg

1200 AT= -

1000

800

600

400

200
Figure 4-11. Stress distri-
bution for an alumina
0
fiber/magnesium composite
with 35% volume fraction
-200
of fibers as a function of a
-400
normalized distance, r/a,
with r/a = 1 representing
-600 the fiber/matrix interface.
Note that or and oe fall off
-800 rapidly with the distance, r.
0 I.I 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2 . 0
 4.3 Interfaces in Metal Matrix Composites 143

If we regard a ceramic particle in a metal matrix and a higher dislocation density.

matrix as more or less an elastic sphere
embedded in an infinite elastic continuum,
then it can be shown from theory of elas-
ticity (e.g., Brooksbank and Andrews,
1972) that an axially symmetrical stress
distribution around the particle will result.
The pressure at the interface between the
particle and the matrix due to the thermal
mismatch is given by Eq. (4-17), where a is
the coefficient of linear expansion, v is the
Poisson ratio, E is the Young's modulus,
AT is the temperature interval, a is the ra-
dius of the particle, b is the radius of the
matrix, Vp is the particle volume fraction,
and the subscripts m and p denote the
matrix and the particle, respectively. Fig-
ure 4-12 shows, schematically, the stress
distribution in a particulate composite.
Chawla and Metzger (1972) showed, in a
definitive matter, the importance of ther-
mal stresses in metal matrix composites.
They used a tungsten fiber/copper single
crystal matrix and by employing a disloca-
tion etch-pitting technique showed that
near the fiber the dislocation density was
much higher in the matrix than the disloca-
tion density far away from the fiber. The
enhanced dislocation density in the copper
matrix near the fiber arises because of the
plastic deformation in response to the ther-
mal stresses generated by the thermal mis-
match between the fiber and the matrix.
The situation in the as-cast composite can
be depicted as shown schematically in
Fig. 4-13 where a section along a primary
plane of the single crystal copper matrix of
the composite is shown (Chawla, 1975 a).

(am-ap)AT
P = 0.5(l+v ) (4-17)
m (l-2v m )F p (l-2v p )
V) E

The principal elastic stresses in the ma-

trix, the radial and tangential components,
in terms of p are:

(4-18)
1- K
-P (4-19)
l-VpUr3
The stresses in the particle are:
GrV = P (constant) (4-20)
°e? ~ P (constant) (4-21)
It can be readily shown from these ex-
pressions that the thermal stresses generat-
ed in a composite, fibrous or particulate,
for rather small temperature changes will
exceed the yield stress of the metal matrix Figure 4-12. Schematic of the stress distribution in a
and, thus, cause plastic deformation of the particulate composite.
144 4 Metal Matrix Composites

TUNGSTEN FIBER

COPPER MATRIX

DISLOCATION
ETCH PITS Figure 4-13. Schematic of a
section along the primary
plane of the single crystal
copper matrix containing
tungsten wires. There exists
a hard zone (high disloca-
tion density) around each
fiber and a soft zone (low
dislocation density) away
from the fiber.

^PRIMARY SLIP PLANE

There exists a hard zone (high dislocation been observed by many researchers (see for
density) around each fiber and a soft zone example Mahon etal., 1990). We discuss
(low dislocation density) away from the this subject of aging behavior of an alloy
fiber. Later on, the existence of a plastical- matrix in the presence of a reinforcement
ly deformed zone containing high disloca- in Sec. 4.4.5.
tion density in the metallic matrix in the
vicinity of the reinforcement was con-
4.3.5 Microstructure of the Interface Region
firmed by transmission electron microsco-
py by a number of researchers, both in In this section, we describe the micro-
fibrous and particulate metal matrix com- structural features in some important com-
posites (Arsenault and Fisher, 1983; posites.
Suresh et al., 1989; Christman and Suresh,
1988a; Rack, 1987; Chawla et al., 1991).
Figure 4-14, a TEM micrograph, shows
dislocations in aluminum matrix near an
SiC particle (Chawla etal., 1991). Such
high densities of defects in the metal ma-
trix will, of course, lead to a different set of
properties of the matrix. In nonprecipita-
tion hardening metals, this will simply
cause a strengthening due to a higher dislo-
cation density. In precipitation hardenable
matrix alloys such as aluminum-copper,
one would expect faster aging kinetics (Ar-
senault and Fisher, 1983; Suresh et al.,
1989; Christman and Suresh, 1988 a; Rack,
1987; Chawla etal., 1991). Preferential Figure 4-14. TEM micrograph showing dislocations
precipitation at the reinforcement/matrix in aluminum in the region near a silicon carbide par-
interface in an age-hardenable matrix has ticle (SiCp) (Chawla et al., 1991).
 4.3 Interfaces in Metal Matrix Composites 145

4.3.5.1 Alumina/Metal Composites

Wetting is rather poor between different
forms of alumina and molten aluminum,
with contact angles ranging from 180° (no
wetting) at the melting point to greater
than 60° at 1800 K. In particular, a-Al 2 O 3
(FP) based systems may involve the forma-
tion of spinels at the fiber/matrix interface,
and they enhance the chemical bond at the
interface (Levi et al., 1978). Alloying Al
matrix by Li to promote wetting of poly-
crystalline A12O3 (FP) fibers is a very good
Figure 4-15. Interface in a continuous fiber a-Al2O3
and successful example of the matrix mod- (FP)/Mg alloy (ZE41A) matrix showing the reaction
ification approach to obtain enhanced zone (RZ) (Pfeifer et al., 1990). BF denotes bright field
wettability and bonding. Thus, Al/2-3%- TEM micrograph
Li matrix would wet the A12O3 fibers in
vacuum and the technique of metal infil-
tration of a fiber bundle can be used. X-ray
diffraction studies of A12O3 fibers extract-
ed from matrix showed the presence of
LiA102 on the fiber surface (Champion
et al., 1978). The following chemical reac-
tion has been proposed for the metallurgi-
cal bonding in FP fiber/Al-Li matrix
(Dhingra, 1980):
6 Li + A12O3 -» 3LiO 2 + 2A1
LiO 2 + A12O3 -> 2LiAlO 2
Electron diffraction analysis showed the
presence of Li2O • 5A12O3 spinel at the fi- Figure 4-16. Dark field (DF) TEM micrograph show-
ber surface (Prewo, 1977). ing deformation twins in the matrix in the same sys-
tem as in Fig. 4-13. These deformation twins form
Interface in a continuous fiber a-Al 2 O 3 because of thermal stresses in the matrix during cool
(FP)/Mg alloy (ZE41A) matrix made by down. M and F indicate matrix and fiber regions,
casting is shown in Fig. 4-15 (Pfeifer et al., respectively (Chawla, 1987).
1990). The reaction zone consists of MgO
and is about 100 nm thick. The chemical
reaction occurring at the interface can be the matrix during cool down. Not unex-
represented by the reaction: pectedly, the thermal mismatch gives a
zone of the matrix around each fiber that
Al2O3(s) + 3 Mg(l) -> 3 MgO(s) + 2 Al(l)
suffers plastic deformation during solidifi-
Figure 4-16 shows deformation twins in cation. Figure 4-17 shows the microhard-
the magnesium matrix in the alumina/ ness profiles of a-Al 2 O 3 (FP)/Al-Li and
magnesium composite. These deformation a-Al 2 O 3 (FP)/Mg alloy composites from
twins form because of thermal stresses in the fiber through the reaction zone to the
146 4 Metal Matrix Composites

Fiber -FP/Al
^Fiber surface o FP/Mg
1300 -^&^
1000- '-Interface reaction zone

~500^
x
Matrix
<D
c
x 100

50-

25
0 1 2 3 4 5
Distance (r/a)
Figure 4-17. Microhardness profiles of a-Al2O3
(FP)/Al-Li and a-Al2O3 (FP)/Mg alloy composites
from the fiber through the reaction zone to the matrix.
Note the high hardness of the matrix in the vicinity of
the interface, indicating a workhardened matrix.

matrix. Note the high hardness of the ma-

trix in the vicinity of the interface, indicat-
ing a workhardened matrix.
PRD-166 fiber is a continuous fiber with
the composition of a-Al 2 O 3 with about
15-20 wt.% of ZrO 2 . The interface in
a composite of PRD-166/Ti-48.4%-Al-
1%-Mn made by high temperature pres-
sure casting showed ZrAl 3 at the fiber pe-
riphery. There was a zone just below the
surface free of ZrO 2 , and the fibers con-
tained some entrapped ZrAl 3 (Nour-
bakhsh etal., 1990).
Discrete facetted particles of ZrAl 3 were (b)
observed at the interface in a PRD-166/
Figure 4-18. (a) Discrete facetted particles of ZrAl3
high purity aluminum made by liquid met- were observed at the interface in a PRD-166/high
al infiltration, Fig. 4-18 (Isaacs etal., purity aluminum made by liquid metal infiltration, (b)
1991). These particles formed and grew in- Zr mapping of the region shown in (a) (Isaacs et al.,
to the matrix when the processing was 1991).
done above the melting point of the matrix
for thirteen minutes or more. This is as
expected from thermodynamic calcula-
tions. However, when the fibers were infil-
 4.3 Interfaces in Metal Matrix Composites
trated with the preform temperature below
the melting point of aluminum and with a
hold time of less than one minute, no reac-
tion products were visible in an SEM, indi-
cating thereby that by controlling the reac-
tion kinetic conditions, one can control the
extent of interfacial reaction.
4.3.5.2 Silicon Carbide/Metal Composites
Silicon carbide in a particulate or fi-
brous form has been incorporated in a
number of metallic matrix materials. Sili-
con carbide is unstable in titanium alloys.
Depending on the specific titanium alloy
composition, a series of compounds can
form in the interface zone. Generally, a
very brittle interface region forms in titani-
um-based composites containing SiC, fiber
or composite (Lorreto and Konitzer, 1990;
Warren and Andersson, 1984).
Das (1990 a, b) has characterized the mi-
crostructure of silicon carbide fiber (SCS-
-2 0 2
Fiber > Reaction Zone j f y t a t r i x
Distance (nut)
147
6) reinforced a-Ti, (5-Tx, Ti3Al, and TiAl
matrix composites. He examined the mi-
crostructure of these composites in the as-
fabricated state and after thermal expo-
sure. SCS-6 silicon carbide fiber has an
outer diameter of 142 |im and a core of
about 37 }^m diameter of a carbon fila-
ment. The SiC is produced by chemical
vapor deposition onto this core (Chawla,
1987 a). A surface coating rich in carbon is
applied to heal the "intrinsic" surface
flaws due to the chemical vapor deposition
technique and to promote compatibility
with the matrix. This surface layer consists
of carbon-rich nonstoichiometric amor-
phous SiC. There occurs a local change in
the chemical composition of the matrix in
the vicinity of the fiber/matrix reaction
zone. Figure 4-19 shows a TEM mi-
crograph of the reaction zone in SCS-6/
near a-Ti composite after exposure to
975°C for 4 hours (Das, 1990a). Recrys-
tallized grains and voids can be seen in the
Figure 4-19. TEM mi-
crograph of the reaction
zone in SCS-6/near a-Ti
composite after exposure to
975 °C for four hours
(Das, 1990).
148 4 Metal Matrix Composites
reaction zone. The bottom part of this fig-
ure shows the concentration profiles of
various elements as determined by energy
dispersive analysis. Knoop indentations
on the matrix in the vicinity of the reaction
zone of the same composite after exposure
FIBER
MATRIX COATING
\ (Chou etal., 1991). The arrows indicate
REACTION
ZONE
CRACK
25jim
Figure 4-20. Knoop indentations on the matrix in the
vicinity of the reaction zone of the same composite
after exposure to 975 °C for 16 hours (Das, 1990). A
crack around the indentation near the fiber can be
seen in the brittle matrix.
Figure 4-21. Cross-section of the reaction zone in SiC/Ni after six hours at 1100 °C. C-PFZ next to nickel is the
carbon precipitate free zone (Chou et al., 1991).
to 975 °C for 16 hours are shown in Fig.
4-20 (Das, 1990 b). A crack around the in-
dentation near the fiber can be seen. This
is because the matrix around the fiber is
more brittle than far away.
Chou et al. (1991) studied interfacial re-
actions between some refractory metals
(Co, Ni, and Pt) and silicon carbide at
temperatures between 800 and 1200 °C.
They observed a series of silicides and free
unreacted carbon precipitates in the reac-
tion zones. A cross-section of the reaction
zone in SiC/Ni after 6 hours at 1100°C is
shown in Fig. 4-21 (Chou et al., 1991). Ad-
jacent to Ni, there is a carbon precipitation
free zone (C-PFZ). Between C-PFZ and
SiC, random and modulated carbon pre-
cipitates appear. Quantitative electron mi-
croprobe profiles of nickel, silicon, and
carbon across the reaction zone after 6
hours at 1100°C are shown in Fig. 4-22
the relative nickel concentrations for Ni3Si
and Ni 5 Si 2 . The layered regions of the un-
reacted nickel, carbon precipitate free
zone, and carbon precipitate zone are also
indicated.
These authors observed that the thick-
ness of the reaction zone on the metal side
was always greater than on the SiC side.
 4.3 Interfaces in Metal Matrix Composites
100
90
*—.•.... Ni .
- ^ 80
§: 70
^T 60
o tion zone after six hours at
•"S 50
40
8 Ni3Si and Ni 5 Si 2 . The layered re-
30
20
Si
Ni 3 Si
10
0 (Chou et al., 1991).
Distance (20jum/Division)
They also did not observe a parabolic reac-
tion zone between the thickness of the SiC
reaction zone and the annealing time. This
was attributed to the reaction being con-
trolled by the decomposition of SiC, rather
than by interdiffusion of Si or metal. Fig-
ure 4-23 shows a series of cross-sectional
views of the reaction zone in SiC/Ni after
4 hours at 1170°C (Chou etal., 1991).
Modulated carbon precipitation, random
carbon precipitation, as well as voids are
seen in this figure. Porosity results from
melting and solidification. The formation
of low melting point metal silicides as well
as a large enthalpy of mixing between the
metals and silicon at the interface lead to
melting.
Krishnamurthy (1990) studied interfa-
cial reactions in TiAl-based alloys contain-
ing SCS-6 fibers and single crystal alumina
fibers. The former showed significant reac-
tion at the fiber/matrix interface while the
latter did not show any noticeable reac-
tion.
Many researchers have examined the in-
terface region in silicon carbide particle or
whisker/aluminum alloy composites made
by powder metallurgy or casting routes
(see for example Arsenault and Fisher,
1983; Suresh etal., 1989; Christman and
149
Ni
Ni5Si2 Figure 4-22. Quantitative electron
microprobe profiles of nickel, sili-
con, and carbon across the reac-
1100°C. The arrows indicate the
relative nickel concentrations for
gions of the unreacted nickel, Ni,
S
'Ni 5 Si 2 carbon precipitate free zone,
C-PFZ, and carbon precipitate
zone, CPZ, are also indicated
Suresh, 1988 a; Rack, 1987; Chawla et al.,
1991; Mahon etal., 1990; Nutt and Car-
penter, 1990; Lagace and Lloyd, 1989). In
magnesium containing aluminum alloys,
MgO is frequently observed at the inter-
face (Nutt and Carpenter, 1990). In the
case of liquid metal infiltration route,
molten aluminum can react with SiC to
form aluminum carbide at the interface as
per the reaction given below:
4Al + 3SiC -> A14C3 + 3 Si
Besides giving a brittle constituent, A14C3
at the interface, such a reaction will inject
silicon into the matrix, i.e., a changed alloy
composition of the matrix will result.
The casting route of making ceramic
particle/metal matrix composites involves
higher temperatures than those in the pow-
der metallurgy route. Severe interfacial re-
actions between ceramic particle and metal
matrix can occur. Extensive reaction be-
tween the ceramic particle and the molten
metal matrix can result in degradation of
the reinforcement, formation of brittle
constituents at the interface, and/or alter-
ation in the matrix microstructure. Wang
etal. (1992) used electron microscopy
techniques to study the interface in a cast
SiCp/Al alloy (A356-7 wt.% Si, 0.3 wt.%
150 4 Metal Matrix Composites
>.;" r ? ;**T.;.'*~ '•
a "
Figure 4-23. Series of cross-sectional views of the reaction zone in SiC/Ni after four hours at 1170 °C. Modulated
carbon precipitation, random carbon precipitation, as well as voids are seen (Chou et al., 1991).
Mg, and the rest Al). They observed mag-
nesium aluminate, MgAl 2 O 4 , a spinel, at
the interface.
4.3.5.3 Carbon Fiber/Metal
Matrix Composites
Carbon fiber reinforced metals have
been investigated by a number of re-
searchers (Fuet al., 1986; Katzman, 1987).
Fu et al. did a detailed analysis of the inter-
face region in carbon fiber/aluminum com-
posites by TEM and observed fine-grained
y-Al 2 O 3 and some A14C3 at the interface.
An interesting example of interface tailor-
ing comes from the C/Mg system. Molten
Mg does not wet and thus does not bond
C fibers. Codeposition of Ti-B on carbon
fiber before incorporating it into an alu-
' b
minum matrix is commonly resorted to.
The Ti-B coating, however, oxidizes in
air, and molten Mg does not wet the oxi-
dized coating. The option of not exposing
the coated fibers to air is not very feasible.
Katzman (1987) used a silicon dioxide
coating from an organometallic precursor
solution. SiO2 coating facilitates melting
and bonding of C fibers by molten Mg.
Such innovative barrier coatings between
reinforcement and matrix can be a power-
ful means of designing a composite with
appropriate properties.
4.3.5.4 Interaction Between
Interfacial Phenomena
Various processing parameters such as
time, temperature, pressure combined with

the thermodynamic, kinetic, and thermal
data can be used to get an optimum set of
interface characteristics in a given metal
matrix composite. Such a control of inter-
face characteristics can result in an in-
crease in a given mechanical property of
the composite. Figure 4-24 shows this idea
in a schematic manner. This figure is some-
what simplistic inasmuch as many of the
thermal, mechanical, chemical, and struc-
tural factors frequently overlap and it is
not always easy, in practice, to separate the
effects of these factors. For example, con-
sider the interaction between chemical in-
teraction and thermal mismatch. A chemi-
cal reaction will produce one or more
chemical compounds at the interface. It is
of great importance to take into account
the presence of any new species at the in-
terface when computing the thermal stress-
es. In other words, if it is known that a
compound forms at the fiber/matrix inter-
face, then it is pointless to use a two-ele-
ment cylindrical model to compute the
thermal stress distribution because the new
species is very likely to have a different
coefficient of thermal expansion than ei-
ther one of the original components. Even
li
Processing Parameter — •
(temperature, time, pressure)
Figure 4-24. Schematic of correlation between pro-
cessing parameter and property of a composite.
4.4 Properties 151
a thermally matched composite can have
thermal stresses if an interfacial product
forms.
The extent of interfacial region in com-
posites can be very large. Mechanical as
well as chemical bonding can contribute to
the bond strength. A controlled interfacial
reaction between a metal and a ceramic
reinforcement can provide good adhesion.
A diffusion barrier coating between the re-
inforcement and the matrix is frequently
employed to control the extent of interfa-
cial reaction. It is therefore important to
understand the thermodynamic, kinetic,
mechanical, and microstructural aspects of
a metal/ceramic interface in order to use
that information to optimize the process-
ing to obtain a desirable set of characteris-
tics in the composite. In particular, such an
understanding is important if the com-
posite is to have an ability to sustain high
temperatures during processing as well as
long time exposures at operating tempera-
tures.
4.4 Properties
4.4.1 Importance of the Metallic Matrix
The differential thermal expansion be-
tween the reinforcement and the metal ma-
trix introduces a high dislocation density
in the matrix, especially in the near inter-
face region of the matrix. This high matrix
dislocation density as well as the reinforce-
ment/matrix interfaces can provide high
diffusivity paths in a composite, an impor-
tant topic that we discuss in Sec. 4.4.5 on
the aging behavior of particle reinforced
MMCs. Most reinforcements for use in
metals are essentially elastic right up to the
fracture point. Once we know their physi-
cal and elastic properties (elastic constants,
expansion coefficient, etc.), we have pretty
well defined their behavior. The interface
152 4 Metal Matrix Composites
region, however, can suffer changes just as
the matrix does during processing and/or
service. The importance of the interface
region, however, has been recognized on
all hands almost from the very beginning,
and we have described some of these in
Sec. 4.3 above. The recognition of the im-
portance of the matrix is more recent. The
metallic matrix is much more than a mere
medium or glue to hold the reinforcement,
be that fibers, whiskers, or particles
(Chawla, 1985, McKimpson and Scott,
1989). It should be chosen after due con-
sideration of its chemical compatibility
and thermal mismatch with the reinforce-
ment. Processing induced chemical reac-
tion and/or thermal stresses can cause
changes in the matrix microstructure of the
matrix. These microstructural changes in
the matrix, in turn, can and do affect the
mechanical and physical behavior of the
composite. Typically, commercial produc-
ers have taken off-the-shelf wrought or
casting alloy compositions for the matrix.
Let us consider an aligned fiber rein-
forced metal matrix composite under a
load Pc in the direction of the fibers. This
load is distributed between the fiber and
the matrix. Thus,
Pe = PmVm + P( V{
where P m and P f are the loads on the ma-
trix and the fiber, respectively. This equa-
tion can be converted to the following rule-
of-mixtures relationship under conditions
of isostrain:
<7f V( (4-22)
where a is the stress, V is the volume frac-
tion, and the subscripts c, f, and m denote
the composite, the fiber, and the matrix,
respectively.
Equation (4-22) is commonly referred to
as the Rule of Mixtures. This equation
says that the strength of the composite is a
volume weighted average of the strength of
the fiber and the matrix. On the face of it,
there is nothing wrong with this statement.
The problem, however, is that one does
not have a very good idea of the <7m and af
values, i.e., the in situ values of the stress
on the fiber and matrix. If the fiber re-
mains essentially elastic up to the point of
fracture, then o{9 the fiber strength in the
composite, is the same as that determined
in an isolated test of the fiber. This is par-
ticularly true of ceramic fibers. The same,
however, cannot be said for the metallic
matrix. The matrix strength in the com-
posite (the in situ strength) will not be the
same as that determined from a test of an
unreinforced matrix sample in isolation.
The metal matrix is likely to suffer several
microstructural alterations during process-
ing and consequently changes in its me-
chanical properties. Let us consider the
process of liquid metal infiltration of a
bundle of ceramic fibers. After the fiber
bundle is impregnated by the liquid ma-
trix, the composite must cool down to
room temperature. In view of the fact that,
in general, ceramic fibers have a coefficient
of thermal expansion less than that of most
metallic matrices, thermal stresses will be
generated in both the components, fiber
and matrix. A series of events can take
place in response to the thermal stresses:
(i) plastic deformation of the ductile met-
al matrix (slip, twinning, cavitation,
grain boundary sliding and/or migra-
tion),
(ii) cracking and failure of the brittle fi-
ber,
(iii) failure of the fiber/matrix interface.
Some very fundamental work has been
done on a simple system, tungsten fila-
ment/copper matrix, to elucidate the role
of matrix in MMCs. The tungsten/copper
system is an ideal system for studying pure
fiber reinforcement. The two components

are mutually insoluble in the liquid as well
as solid states. Thus, there is no interfacial
reaction between the fiber and the matrix.
In the case of a single crystal matrix, mi-
crostructural changes induced in the ma-
trix by the mere act of incorporating fibers
into it can be studied. Polycrystalline but
pure matrix is the next step in complica-
tion while an alloy matrix will be even
more complex. We provide a summary of
these results (Chawla and Metzger, 1972;
Chawla, 1973 a, b, 1975 a, b, 1976 a, b).
Single crystal copper matrix is amenable
to etch-pitting techniques for dislocation
density and distribution studies (Liv-
ingston, 1962). Low volume fraction (0 to
15%) and large diameter fibers (228 |im)
were used to ensure the availability of large
matrix areas for direct structural observa-
tions. For the etch-pitting technique, the
copper surface has to be within 3° of (1 1
1) plane.
Dislocation density profiles in the cop-
per matrix for different fiber volume frac-
tions of tungsten filament are shown in
Fig. 4-25. Close to the fiber, the curves
tend to merge. The differences in the dislo-
cation density were greater away from the
fiber (the plateau region). Near the fiber
7^7.551
10 9
crn" 2
/ i o1°
/ ° ° ! n0.5 1
density
o
CL
107
~~i
106
i (the plateau region).
L0 80 120 160 jim 200
Normal distance from interface
4.4 Properties 153
the dislocation cell structure was better de-
fined. The dislocation densities can be
thought of as a combination (not necessar-
ily a linear one) of two contributions:
(i) a fiber volume fraction independent
contribution (near the interface)
(ii) a fiber volume fraction dependent con-
tribution (throughout the matrix).
In order to understand the origin of
this dislocation distribution, Chawla and
Metzger (1972) used the simple two-ele-
ment model shown in Figure 4-10. For the
W/Cu system, a was 114 jam and the radius
b of the equivalent cylinder of matrix var-
ied from 1590 urn at 0.5 % Vf to 290 (im at
15% Vf. The enhanced dislocation densi-
ties in the copper matrix arise owing to the
plastic deformation in response to the ther-
mal stresses generated because of the ther-
mal mismatch between the fiber and the
matrix.
According to the Tresca yield criterion
[or — ae] = Y, the matrix yielding will occur
first at the interface and then spread out
with continued cycling. The dislocation
density gradient will tend to even out with
continued thermal cycling.
In the as-fabricated state, the copper
matrix had a cell structure with a diameter
15%
10 % , Figure 4-25. Dislocation density profiles
in the copper matrix for different fiber
7.5%" volume fractions of tungsten filament.
2.5%
'• ! 0.5 %
Close to the fiber, the curves tend to
0% merge. The differences in the dislocation
density are greater away from the fiber
154 4 Metal Matrix Composites
10 Jim. The cell structure develops due to
the high temperature, low (thermal) stress
creep that the matrix sufffers during cool-
ing from 1200 °C to room temperature. On
thermal cycling to low peak temperatures,
the subgrain size and the cell diameter did
not change, and the cell diameter did not
decrease significantly. The cell walls, how-
ever, thickened and the cell structure be-
came better defined. It would appear that
the dislocation cell walls formed during the
composite fabrication served as obstacles
to dislocations generated on subjecting the
composite to either mechanical or thermal
cycling.
Mortensen and coworkers (1991, 1992)
as well as others have focussed their atten-
tion on matrix microstructural control in
cast MMCs. The final matrix microstruc-
ture is a function of fiber type, fiber diame-
ter, fiber distribution as well as conven-
tional solidification parameters. For ex-
ample, in an SCS-2 silicon carbide fiber/
Al-4.5 %-Cu matrix, they observed that in
the unreinforced region of the matrix, nor-
mal dendritic structure could be seen while
in the reinforced region the dendritic mor-
phology was controlled by the fiber distri-
bution. Second phase appeared preferen-
tially at the fiber/matrix interface or in the
narrow interfiber spaces.
Porosity is a critical defect that is likely
to be present in the matrix, especially in
cast materials. Porosity can be very delete-
rious to the overall performance of cast
MMCs. The main sources of porosity are
any gas evolution and shrinkage occur-
ring on solidification. A high volume frac-
tion of reinforcement may impede the
flow of the interdendritic liquid and inhibit
any "bulk movement" of the semisolid
metal.
In short, the micro structure of the
metallic matrix in the fiber composite is
likely to differ significantly from that of
the unreinforced metal processed in an
identical manner.
In what we have discussed in this section
so far, we have implicitly assumed that the
fibers are uniformly distributed in the ma-
trix and that the fiber cross sectional ge-
ometry does not have any effect on the
strength and plastic deformation of the
composite. In practice, there invariably
does occur some fiber clustering and the
fiber cross sectional shape may also vary.
Christman et al. (1989) used a plane strain
model to study the effect of fiber clustering
on the plastic deformation of silicon car-
bide whisker reinforced aluminum. They
found that when the whiskers were clus-
tered, either horizontally or vertically, the
flow stress and strain hardening of the
composite decreased significantly com-
pared to the situation where the whiskers
were arranged in a perfectly periodic man-
ner. Brockenbrough and Suresh (1990) an-
alyzed continuous boron fiber reinforced
aluminum 606 l-O T using the ABAQUS
finite element code. The fiber packing ar-
rangement as well as the cross sectional
geometry influenced the plastic deforma-
tion behavior when the composite was
loaded in the transverse direction. The ef-
fect of fiber packing arrangement on strain
hardening response, however, was more
pronounced than that of fiber geometry.
Both these variables did not have much
effect when the composite was loaded par-
allel to the fibers. Suresh and Brocken-
brough (1993) provide a review of the var-
ious continuum models for deformation of
continuous fiber reinforced metal matrix
composites.
4.4.2 Modulus
Unidirectionally reinforced continuous
fiber reinforced metal matrix composites
show a linear increase in the longitudinal
 4.4 Properties 155

Figure 4-26. Example of

modulus increase as a func-
tion of fiber volume frac-
tion for alumina fiber rein-
forced aluminum-lithium
alloy matrix (Champion
et al., 1978).
0 0.2 (U 0.6 0.8

175 n
Young's modulus of the composite as a
function of the fiber volume fraction. Fig-
ure 4-26 shows an example of modulus in-
crease as a function of fiber volume frac- tf)
l-
r

tion for alumina fiber reinforced alu- 1125;

minum-lithium alloy matrix (Champion O

et al., 1978). The increase in the longitudi- 1100 ; L

nal Young's modulus is in agreement with
the rule-of-mixtures value while the modu-
lus increase in a direction transverse to the
fibers is very low. Particle reinforcement 10 20 30 40
also results in an increase in the modulus Fiber volume fraction (%)
of the composite; the increase, however, is Figure 4-27. Modulus increase as function of short
much less than that predicted by the rule of fiber volume fraction (Girot et al., 1987).
mixtures. This is understandable inasmuch
as the rule of mixtures is valid only for
continuous fiber reinforcement. Figure Continuous fibres •
Whiskers
4-27 shows modulus increase as function o_
o /
of short fiber volume fraction (Girot et al., in
120-
1987). Note the rather large variation from D
D
_ / ^ '
-o 100- Particle
the rule-of-mixtures type linear variation o
with fiber volume fraction. A schematic of E
o
sti

80-
the modulus increase in a composite with o
LU
reinforcement volume fraction for differ- 60-
ent reinforcement type, continuous fiber, 0 5 10 15 20 25 30 35
whisker, or particle, is shown in Fig. 4-28. Volume fraction (%)

This schematic shows the loss of reinforce-
ment efficiency as one goes from continu-
ous fiber to particle. For discontinuous re-
inforcement, one may use a Halpin-Tsai-
Figure 4-28. Schematic of modulus increase in a com-
posite with reinforcement volume fraction for differ-
ent reinforcement type, continuous fiber, whisker, or
particle. Note the loss of reinforcement efficiency as
one goes from continuous fiber to particle.
156 4 Metal Matrix Composites

14
Kardos type expression to obtain the mod-
ulus (Chawla, 1987 a):
12
l+2sqVp
Pc = Pm 1 _ y

where p stands for the property of interest,

for example, the transverse Young's modu-
lus, Ev s is the aspect ratio of the discontin-
uous reinforcement, and

^+2s
Pm

It would appear that the increase in elastic 298 K

modulus of a metal matrix composite as a
function of the reinforcement volume frac-
tion is fairly straight forward. The modu-
lus of the composite is fairly independent
of the reinforcement packing arrangement
as long as all the fibers are parallel. For
discontinuous reinforcement also the mod-
ulus is quite independent of the particle
clustering, etc. The modulus does show a
dependence on temperature, which essen-
tially comes in from the matrix modulus
dependence on temperature, the ceramic
modulus being quite independent of tem-
perature for moderate temperatures in-
volved with MMCs.
4.4.3 Strength
Commonly, a metal matrix composite
has a ceramic reinforcement, however,
metal reinforcement in a metal matrix or
metal/metal laminated composites are not
uncommon. Stress-strain curves of the cop-
permatrix, the Nb-Ti filaments, and the
resultant composite consisting of Nb-Ti
filaments in a copper matrix are shown in
Fig. 4-29 (Reed et al., 1977). This is an un-
usual composite in that both the compo-
nents, matrix and fiber, are ductile. An in-
teresting aspect of the stress-strain curve
0.02 0.03 0.04 0.05
STRAIN
Figure 4-29. Stress-strain curves of the copper ma-
trix, the Nb-Ti filaments, and the resultant composite
consisting of Nb-Ti filaments in a copper matrix
(Reed et al., 1977).
shown in Fig. 4-29 is the rather high frac-
ture strain of the composite. The strength
of a metal matrix composite is difficult to
predict by a rule of mixtures type relation-
ship because the strength is a very struc-
ture sensitive property. Processing induced
thermal stresses and/or chemical reaction
at the interface can make it difficult to pre-
dict the strength by a simple rule of mix-
tures type relationship. In Sec. 4.2.1, we
pointed out the importance of taking into
account the microstructural changes oc-
curring in the metal matrix during process-
ing. If we take due account of these pro-
cessing induced microstructural changes in
the metallic matrix, then it would appear
that shear-lag, which involves load trans-
fer from the matrix to the fiber via shear on
planes parallel to the fiber, would explain
the fiber strengthening phenomenon. For
a description of this fiber strengthening
model, the reader is referred to one of the

references given in the list of suggested
reading at the end of the chapter. The
shear-lag model would also appear to
work for whisker reinforced metal matrix
composites. It becomes of rather doubtful
validity when one considers particulate
composites. Clearly, the matrix alloy and
the processing induced micro structural
changes in the matrix will influence the
final properties of the composite much
more in this case. The strengthening of a
precipitation hardenable aluminum alloy
by incorporating particles of, say, SiC (as-
pect ratio about unity) is not as significant
as in the case of a fiber or whisker of the
same material. It appears that dislocation
density increase, and the concomitant mi-
crostructural alterations in the matrix are
an important source of strengthening in
particulate composites. This would appear
to explain why the composite shows a
strength higher than that predicted by con-
tinuum mechanics theories. At least in the
case of aluminum matrix composites, the
increase in strength would appear to be
mainly due to the much higher dislocation
density and smaller subgrain size in the
matrix than in the unreinforced aluminum
alloy. The increase in elastic modulus, as
discussed above, can be appreciable. Mc-
Danels (1985) evaluated the effects of rein-
forcement type SiC-whisker, -nodule, or
-particle reinforcement in various matrix
alloys, 6061, 2024/2124, 7075, and 5083 Al
matrices. His results showed that in these
cases, the elastic modulus of the com-
posites was independent of both reinforce-
ment type and the matrix alloy. The rein-
forcement content was the dominant fac-
tor in the improvement of elastic modulus
in these composites. However, the yield
and ultimate tensile strength of the com-
posites were primarily controlled by the
intrinsic yield/tensile strengths of the ma-
trix alloys and by reinforcement content.
4.4 Properties 157
The yield strength and ultimate tensile
strength of aluminum alloys can be in-
creased to as much as 60% more than
those of the unreinforced matrix. Howev-
er, it is important to emphasize that for a
given reinforcement content, yield strength
and ultimate tensile strength are controlled
by the type and temper of the matrix alloy.
The precipitate-hardenable aluminum al-
loy matrix has a much more important role
to play in these composites than a non-
hardenable matrix material. A consider-
able strength increment results due to the
age hardening treatments. The particulate
or whisker type reinforcements such as
SiC, B4C, A12O3, etc. are unaffected by the
aging process. These particles, however,
can affect the precipitation behavior of the
matrix quite significantly, as we discuss be-
low.
The strength, <r, of in situ metal matrix
composites made by directional solidifica-
tion of eutectic alloys is given by a relation-
ship similar to the Hall-Petch relationship
used for grain boundary strengthening of
metals,
where a0 is a friction stress term, k is a
material constant and X is the spacing be-
tween rods or lamellae. It turns out that
one can vary the interfiber spacing, X,
rather easily by controlling the solidifica-
tion rate, R, because X2 R = constant.
Thus, one can easily control the strength of
the resultant composite.
4.4.4 Matrix Aging Response
in Composites
We have pointed above that the mi-
crostructure of a metallic matrix is modi-
fied by the presence of ceramic reinforce-
ment (particle, whisker, or fiber). In partic-
ular, a higher dislocation density in the
158 4 Metal Matrix Composites
matrix metal or alloy than that in the unre-
inforced metal or alloy has been observed.
The higher dislocation density in the ma-
trix has its origin in the thermal mismatch,
Aa, between the reinforcement and the
metal matrix. For example, the thermal
mismatch in the case of SiC/Al has a high
value of 21 x 10" 6 K" 1 , which will lead to
thermal stresses high enough to deform the
matrix plastically and thus leave the matrix
workhardened. One would expect that the
high dislocation density will also affect the
precipitation kinetics in a precipitation
hardenable matrix such as 2XXX series
aluminum alloys.
Aging Process in Al-Cu-Mg Alloys
The general sequence of precipitation
from a supersaturated Al-Cu-Mg alloy
when aged at a temperature above room
temperature is:
supersaturated alloy -> GP zones -•
This aging behavior is commonly analyzed
in terms of a hardness or strength vs. aging
time curve as shown in Fig. 4-30. The
shape of the aging curve can be explained
as follows: Immediately after quenching,
only solid solution hardening is present.
As GP zones form, hardness or strength
increases because extra stress is needed to
Equilibrium
precipitates
Aging time (particle size)
Figure 4-30. Aging curve, hardness vs. aging time, of
a precipitation hardenable aluminum alloy.
have dislocations shear the coherent zones.
The hardness increases as the GP zone size
increase with time, making it ever more
difficult for the dislocations to shear them.
The peak hardness or strength is associat-
ed with a critical dispersion of coherent or
semicoherent precipitates. With still fur-
ther increase in time, equilibrium precipi-
tates with incoherent interface start ap-
pearing, and the mechanism of Orowan
bowing of dislocations around the parti-
cles becomes operative. The stress required
for dislocation bowing varies inversely as
the interparticle spacing, which decreases
with increasing aging time. Continued
aging results in an increase in precipitate
size and interparticle distance, and a lower
strength results as dislocation bowing be-
comes easier. It is well known that me-
chanical deformation prior to aging can
accelerate the formation of S' at the ex-
pense of GP zones. The S' precipitate has
the same chemical composition as the
equilibrium phase S, orthorhombic struc-
ture with [10 0], [010], and [001] direc-
tions of S' being parallel to the [100],
[0 2 1], and [0 1 2] directions of the matrix.
The S' precipitates nucleate on dislocations
and grow as laths on {2 1 0} planes along
[1 0 0] directions.
One expects that the quenching from the
solutionizing temperature to room temper-
ature, a AT of about 450°C, will result in
a large zone of matrix plastically deformed
around each ceramic particle in which the
dislocation density will be very high. One
would expect that this high dislocation
density will tend to accelerate the aging
kinetics. Nieh and Karlak (1984) showed
that the presence of B4C particles in 6061
Al made the matrix reach the peak hard-
ness in 3 hours at 450 K while a control
powder metallurgy processed alloy took 10
hours. This accelerated aging behavior was
attributed to the high diffusivity character-

istics of the composite. A considerable
strength increment can be due to the age
hardening treatments in a precipitation
hardenable aluminum alloy composite.
The particle or whisker type reinforce-
ments such as SiC, B4C, A12O3, etc. are
unaffected by the aging process. These
particles, however, can affect the precipita-
tion behavior of the matrix quite signifi-
cantly. Arsenault and Fisher (1983) used
transmission electron microscopy to ob-
serve a high density of dislocations
(4x 109 cm" 2 ) clustered around the ends
of SiC whiskers (SiCw) and forming low
angle cell boundaries in a 20%-SiCw/Al
6061 composite. They also observed very
fine precipitates along the dislocations.
Mahon et al. (1990) observed decoration
of dislocations by precipitates in SiC parti-
cle reinforced aluminum composites. In
situ observation of dislocations generated
due to thermal mismatch have been made
by using a hot stage in a high voltage trans-
mission electron microscopy (Vogelsang
etal., 1986). These authors also observed
that the presence of SiC particles hindered
the subgrain growth in the aluminum ma-
trix. It was also observed that the disloca-
tion density in the matrix was smaller
around small, equiaxial SiC particles. As
the particle size increased from 1 |j,m to
5 jim, the dislocation density in the matrix
around a particle increased. Such an effect
of the matrix particle size on the size of the
matrix plastic zone around the particle was
verified by Kim et al. (1990). These dislo-
cations form heterogeneous nucleation
sites for the precipitates in the matrix dur-
ing subsequent aging treatments. This in
turn results in the alteration of the precip-
itation kinetics in the matrix of the com-
posite as compared to those of the unrein-
forced material (Nieh and Karlak, 1984;
Arsenault and Shi, 1986; Rack, 1987;
Christman and Suresh, 1988 a; Papazian,
4.4 Properties 159
1988; Dutta and Bourell, 1988; Christman
etal., 1989; Suresh etal., 1989; Chawla
et al., 1991; Dutta et al., 1991). Dutta and
Bourell (1989) considered the effect of elas-
tic strain field induced in the matrix by the
ceramic particles to cause an enhanced ma-
trix diffusivity and accelerated aging kinet-
ics. Dutta et al. (1991) made a theoretical
analysis and experimental observations
leading them to conclude that precipitate
growth rate was accelerated because of
short circuit, pipe-diffusion, through dislo-
cations.
An interesting aspect of the effect of ce-
ramic particles on aging kinetics of a ma-
trix is that the aging kinetics are a very
sensitive function of the aging tempera-
ture. Specifically, the aging behavior of
the matrix in SiCp/Al 2014 (Chawla et al.,
1991; Esmaeili et al., 1991) and in Al 2 O 3p /
Al 2014 (Chawla et al., unpublished) com-
posites made by casting route, was exam-
ined at four different aging temperatures
and compared with the unreinforced Al
2014. The time to peakage for the Al 2014
matrix and the unreinforced Al 2014 as a
function of aging temperature is shown in
Fig. 4-31. As the aging temperature in-
creased, the difference in peak aging time
between Al 2014 matrix and the unrein-
forced Al 2014 became larger. Note at
150°C the difference in peak aging time
between the reinforced and unreinforced
material was rather negligible. At 195 °C
this difference was about 7 hours. The rel-
atively large difference between thermal
expansion coefficient of SiC and Al result-
ed in substantial stresses which generated
dislocations at and near the interface on
quenching. During aging at 195 °C, high
density of dislocations would provide a
short-circuit path for heterogenous nucle-
ation and fast growth of S' precipitates re-
sulting in accelerated aging of the matrix.
As the aging temperature was lowered, the
160 4 Metal Matrix Composites
165 180 195
Temperature (°C)
Figure 4-31. Time to peakage for Al 2014 matrix and
Al 2014 as a function of aging temperature. As the
aging temperature increases, the difference in time to
peakage between Al 2014 matrix and the unreinforced
Al 2014 increases.
difference in aging kinetics diminished.
This was attributed to the homogeneous
nucleation of GP zones and a decrease in
growth rate of precipitates at lower aging
temperature. It is well known (Lorimer,
1978; Porter and Easterling, 1981) that the
formation of GP zone and its transition to
Q" and & in Al-Cu system is only possible
when the alloy is aged at a temperature
below the GP zone solvus. If, for example,
the aging temperature is above 0" but be-
low 9' solvus, then there will be no GP
zones and the first precipitate will be 0',
heterogeneously nucleated at dislocations.
At high aging temperatures, the S' precipi-
tates nucleate heterogeneously on disloca-
tions and grow as laths on (2 1 0)a in [0 01]
directions (Porter and Easterling, 1981).
At lower temperatures the first precipitates
to nucleate are the coherent Cu-rich and
GP zones. It is important to mention that
the nucleation of GP zones at low temper-
ature is mostly homogeneous, and excess
vacancies introduced by quenching are
thought to play an important role in their
formation. Upon quenching the reinforced
composite, the soft matrix undergoes a
plastic deformation due to the substantial
thermal stresses resulting from the CTE
difference between Al matrix and SiC par-
ticles. The plastic deformation of the ma-
trix will result in a large number of disloca-
tions and vacancies. These defects would
appear to accelerate the aging kinetics of
the matrix at high aging temperatures.
Such effects do not seem to be effective at
low aging temperatures. Differential scan-
ning calorimetry confirmed the presence of
GP zones and S' precipitates at 150°C, and
only S' precipitates at 195 °C in the 2014
matrix (Chawla et al., 1991).
Most of the metal matrix composite
work has involved off-the-shelf metallic al-
loys, especially in the case of particle rein-
forced metal matrix composites. It is im-
portant to bear in mind that in such cases,
using the standard heat treatment prac-
tices given in the manuals and handbooks
for the unreinforced alloys can lead to
drastically different results. For a thor-
ough review of the matrix aging behavior
in particle reinforced metal matrix com-
posites, see the review by Suresh and
Chawla (1993).
4.4.5 Toughness
Toughness of a given metal matrix com-
posite depends on
- the matrix alloy composition and mi-
crostructure,
- the reinforcement type, size, and orien-
tation, and
- processing insofar as it affects mi-
crostructural variables (e.g., distribution
of reinforcement, porosity, segregation,
etc.).
The general range of the plane strain
fracture toughness, Klc, values for particle
reinforced aluminum type MMCs is be-
tween 15-30 MPa m 1/2 , while short fiber
 4.4 Properties 161

or whisker reinforced MMCs show a range Table 4-1. Toughness of particle reinforced MMCs
of about 5-10 MPa m 1/2 . The reader is vis-a-vis unreinforced alloy (Friend, 1989).
cautioned that a lot of data in the literature Material Klc (MPa m1/2)
has been obtained from invalid fracture
toughness tests! A summary of fracture Al 6061 (T6) 37
toughness results for some particle and A12O3/A1 6061 15
SiC/Al 6061 7-18
whisker reinforced composites with alu-
minum 6061 matrix and unreinforced alu-
minum 6061 are given in Tables 4-1
Table 4-2. Toughness of Al 6061 and SiC whisker and
(Friend, 1989) and 4-2 (Hasson and
particle reinforced composites (Hasson and Crowe,
Crowe, 1985). Table 4-2 shows that by sub- 1985).
stituting silicon carbide whiskers with par-
ticles, one can increase the toughness of Material Klc (MPa m1/2)
the composite. 20 vol. % SiCw/A16061 T6 7.1
Continuous fiber reinforced metal ma- 25 vol. % SiCp/A16061 T6 15.8
trix composites show anisotropy in tough- Al 6061 T6 37.0
ness just as in other properties. Table 4-3
shows the results of impact energy for the
B/Al system in the longitudinal and trans- Table 4-3. Anisotropy in impact fracture energy
verse directions. Note that, for a given ma- (kJ m~2) of B/Al (McDanels and Signorelli, 1976).
trix alloy, the longitudinal fracture energy
System Longitudinal Transverse
is higher than the transverse energy. Table
4-4 gives the results of 0°, 90°, and 0°/90° B/Al 1100 200-300 0.53
reinforcement of aluminum by boron B/Al 5052 90 0.85
fibers. The 0° and 90° arrangements of B/Al6061 80 0.27
fibers result in two extremes of toughness B/Al2024 40 0.31
while the 0°/90° arrangement, a sort of
pseudo-random arrangement, reduces the
degree of anisotropy (Sun and Prewo, Table 4-4. Toughness as function of ply orientation in
1977). Using fibers in the form of a braid a B/Al composite (Sun and Prewo, 1977).
can make the crack propagation toughness Ply angle K IC (MPam 1/2 )
increase tremendously due to extensive
matrix deformation, fiber bundle debond- 0° 100
ing, and pullout. Table 4-5 shows this for 90° 34
0°/90° 61-63
alumina/Al-2.5 % -Li composite (Majidi
and Chou, 1987). Note the increase in the
transverse toughness on three-dimensional
Table 4-5. Anisotropy in toughness reduced by 3 D
braiding of the fibers. braiding in an alumina FP (36 vol.%)/Al-2.5 Li com-
Matrix alloy composition can also have posite (Majidi and Chou, 1987).
significant effect on toughness of a com-
iC lc (MPam 0 - 5 )
posite. Table 4-6 shows the effect of differ-
ent aluminum matrix alloys on Charpy en- Longitudinal Transverse
ergy of boron/aluminum alloy composites
Unidirectional 32 15-17
(McDanels and Signorelli, 1976). The 3 D braided 32.3 21.1
tougher the matrix, the tougher will be the
162 4 Metal Matrix Composites

Table 4-6. Effect of matrix alloy on Charpy energy of ficients of thermal expansion of com-
B/Al alloy composites (McDanels and Signorelli, 1976). posites (Kerner, 1956; Turner, 1946;
Composite Charpy impact energy (kJ m 2) Schapery, 1968), determined experimental-
ly these coefficients and analyzed the gen-
Thin sheet Standard eral thermal expansion characteristics of
20-300 200-300
metal matrix composites (Vaidya and
B/1100
B/5052 90 170 Chawla, 1991; Premkumar etal., 1992;
B/6061 80 126 Zweben, 1992). Figure 4-32 shows the vari-
B/2024 40 120-170 ation of thermal expansion coefficient of
SiCp/Al composite as a function of particle
volume fraction. The solid line represents
Table 4-7. Effect of fiber diameter on impact fracture the Turner model prediction of the expan-
energy (kJm~ 2 ) of B/A11100 (McDanels and Sig- sion coefficient as
norelli, 1976).

Fiber Diameter (um) Energy (kJ m 2) a = VmK VpKp

100 90
140 150
200 200-300
composite. The effect of fiber diameter on
the toughness of a metal matrix can be
seen in Table 4-7. For a given Vf9 the larger
the diameter of the fiber, the tougher is the
composite. This is because the larger the
fiber diameter, for a given fiber volume
fraction, the larger is the amount of tough,
metallic matrix in the inter/fiber region that
can undergo plastic deformation and thus
contribute to the toughness.
4.4.6 Thermal Properties
We discussed in the interface section the
importance of thermal expansion mis-
match between the reinforcement and the
matrix. Thermal mismatch is indeed some-
thing that is difficult to avoid in any com-
posite. By the same token, however, one
where a is the coefficient of thermal expan-
sion, Fis the volume fraction, Kis the bulk
modulus, and the subscripts p and m rep-
resent the particle and the matrix. In
Fig. 4-32, we have taken the coefficient of
expansion of SiCp to be 4.5xlO~ 6 K~ 1
while that of aluminum i s 2 3 x l O ~ 6 K ~ 1 .
Values of the expansion coefficients of
some common electronic substrate materi-
als are also included. Turner equation
seems to predict the values very well in
these particulate composites. In terms of
25
Turner model
prediction
10
— AI 2 O 3
GaAs-
AlN.

can control the overall thermal expansion 0

characteristics of a composite by con- 0 0.2 (U 0.6 0.8 1.0
Volume fraction of SiC
trolling the proportion of reinforcement
Figure 4-32. Variation of thermal expansion coeffi-
and matrix and the distribution of the rein- cient of SiCp/Al composite as a function of particle
forcement in the matrix. Many researchers volume fraction. The solid line represents the Turner
have proposed models to predict the coef- model prediction.
 4.4 Properties 163

matching the expansion coefficient of the dominant at higher particle volume frac-
composite with that of a given substrate, tions.
all one needs to do is to use the appropriate Thermal conductivity is another thermal
particle volume fraction. Varying the alu- property of great significance. Unidirec-
minum alloy composition can allow some tionally aligned carbon fibers in an alu-
additional control of thermal expansion minum matrix can have very high thermal
coefficient of the composite. Figure 4-33 conductivity along the fiber direction. The
shows the predicted influence of the silicon conductivity transverse to the fibers is
content in the matrix on the coefficient of about two-thirds that of aluminum. In
thermal expansion of the composite as a fact, in the temperature range of-20°C to
function of silicon carbide volume fraction 140 °C, the thermal conductivity of the
(Premkumar etal., 1992). Understand- composite along the fiber direction is
ably, the influence of SiCp becomes more greater than that of copper. On a per unit
weight basis, the composite is about four
12 1
1
times as conductive as pure aluminum and
'
2.6 times as conductive a Al 6061. Figure
11 AI-5%S1 matrix " 4-34 compares the thermal conductivity of
10 X A M 0%SI matrix
...... AI-20%SI matrix - copper and carbon fiber/aluminum (Sharp
AI-25%SI matrix
N and Loftin, 1992). Such a C/Al composite
- ^
i: VS.
-
can find applications in heat transfer appli-
cations where weight reduction is an im-
S4 portant consideration. For example, in
high-density, high-speed integrated-circuit
6 -
packages for computers and in base plates
5 - for electronic equipment. Another possi-
1 1 i i i bility is to use this composite to dissipate
0.3 0.4 0.5 0.6 0.7 0.8 0.9
Volume fraction of SIC
heat from the leading edges of wings in
high-speed airplanes.
Figure 4-33. Predicted influence of silicon content in
the matrix on the coefficient of thermal expansion of
the composite as a function of silicon carbide volume 4.4.7 Fatigue
fraction (Premkumar et al., 1992). The influence of
SiCp becomes more dominant at higher particle vol- We can define fatigue as the phe-
ume fractions. nomenon of mechanical property degrada-

Composite
5.000
Pure copper /^3
2.000

2.000
C
o
o 1.000 Figure 4-34. Comparison of thermal
2
I 1.000
conductivity of copper and carbon
fiber/aluminum as a function of tem-
perature.
-200 -100 0 100 200
Temperature (°C)
164 4 Metal Matrix Composites

tion leading to failure of a material or a
component under cyclic loading. Under-
standing of the fatigue behavior of com-
posites has lagged that of other aspects
such as the elastic stiffness or strength. An
understanding of the fatigue behavior of
composites of all kinds (polymer matrix
composites - PMCs, metal matrix com-
posites - MMCs, and ceramic matrix com-
posites - CMCs) is of vital importance,
because without such an understanding it
would be virtually impossible to gain ac-
ceptance of the design engineers. It is well
known that a high proportion of failures in
metals can be traced to the phenomenon of
fatigue. Many high volume applications of
composite materials involve cycling load-
ing situations, e.g., automobile compo-
nents.
In general, it would appear that the fa-
tigue resistance of a given material can be
enhanced by reinforcing it by short discon-
tinuous fibers (or whiskers) or continuous
fibers. Not unexpectedly, the improvement
is the greatest when the fibers are aligned
parallel to the stress direction.
While conventional approaches such as
stress vs. cycles (S-N) curves or fatigue
crack propagation tests based on the frac-
ture mechanics concepts can be used for
comparative purposes and for obtaining
information on the operative failure mech-
anisms, they do not provide information
useful to designers. Novel approaches as
that epitomized by the measurement of
stiffness reduction as a function of cycles
in polymer matrix composites are desir-
able. Since many applications of com-
posites involve temperature changes, it is
important that thermal fatigue characteris-
tics of composites be evaluated in addition
to their mechanical fatigue characteristics.
A major difficulty in this regard is that
the application of conventional approach-
es to fatigue of composites, for example,
the stress vs. cycles (S-N) curves or the
application of linear elastic fracture me-
chanics (LEFM), is not straightforward.
The main reasons for this being the inher-
ent heterogeneity and anisotropic nature
of the composites. This results in damage
mechanisms in composites being very dif-
ferent from those encountered in conven-
tional, homogeneous, or monolithic mate-
rials. Despite these limitations, conven-
tional approaches have been used, and we
describe them briefly before describing
some novel approaches to the problem of
fatigue in composites. In particular, we
shall describe some of the techniques that
have been used to analyze the fatigue be-
havior of PMCs, and are beginning to be
applied to MMCs as well. Following this,
we briefly touch upon the important prob-
lem of thermal fatigue in composites.
S-N Curves
A very popular conventinal fatigue test-
ing technique used with metals involves de-
termination of the so-called S-N curves,
where S is the stress amplitude and N is the
number of cycles to failure (Fig. 4-35). In
general, for ferrous metals, one obtains a
Cycles to failure A/f
Figure 4-35. S-N curve obtained in conventional fa-
tigue testing. S is the stress amplitude, and N is the
number of cycles to failure. Curve A shows an en-
durance or fatigue limit, SL, while curve B does not.

fatigue limit or endurance limit, curve A in
Fig. 4-35. For stress levels below this en-
durance limit, theoretically, the material
can be cycled infinitely. If, however, such
an endurance limit does not exist, for ex-
ample for nonferrous metals (curve B in
in
CD
Fig. 4-35), then one can arbitrarily define a
certain number of cycles, say 106, as the
cutoff value. It would be useful at this
point to define some important parameters
which are used in the discussion of the fa-
tigue phenomenon. Based on Fig. 4-36, we
can define the following parameters:
cyclic stress range,
cyclic stress amplitude,
mean stress,
stress ratio,
r>
Fatigue behavior of metals can be con-
veniently divided into two stages: crack
initiation and propagation. In high cycle-
low stress fatigue of metals, most of the
property changes occur in what has been
called the small process zone at the crack
tip. The rest of the material away from the
crack does not seem to be affected serious-
ly by the phenomenon of fatigue. In the
low cycle-high stress fatigue of metals, the
phenomena of workhardening and work-
softening are observed. Quite frequently,
however, a rule-of-thumb approach in
metal fatigue is to increase its monotonic
strength which concomitantly results in an
increase in its cyclic strength. This rule-of-
thumb assumes that the ratio fatigue
4.4 Properties 165
\
\ /
/
/
'V/
\ / \v / Time
Figure 4-36. Various fatigue parameters.
strength/tensile strength is about constant.
Maximum efficiency in terms of stiffness
and strength gains in fiber reinforced com-
posites occurs when the fibers are continu-
ous, uniaxially aligned and the properties
are measured parallel to the fiber direc-
tion. Incorporation of fibers certainly im-
proves the fatigue resistance of fiber rein-
forced composites in the fiber direction.
High strength, brittle fibers such as carbon
or boron would not be expected to show
fatigue behavior similar to that of metals.
Not surprisingly, therefore, composites
containing these fibers, aligned along the
stress axis and in large volume fractions,
will show high monotonic strength values
which are translated into high fatigue
strength values as well. Typically, the S-N
curves of such composites are very flat.
Figure 4-37 shows the S~N curves in ten-
sion-tension for unidirectionally rein-
forced boron (40 vol.%)/Al 6061, alumi-
na-FP (50vol.%)/Al, and alumina-FP
(50vol.%)/Mg composites (Champion
et al., 1978). The cyclic stress is normalized
with respect to the monotonic ultimate
tensile stress. Note the rather flat S-N
curves in all the cases and the fact that the
unidirectional MMCs show better fatigue
properties than the matrix when loaded
parallel to the fibers. For example, at 107
cycles, the fatigue to tensile strength ratio
of the composite is about 0.77, almost dou-
ble that of the matrix.
166 4 Metal Matrix Composites

I.O
40 v/o B/6061 Al - Room Temp.
if) •»,a;ss^s
- — ,, - ° ~"~
- ———-
? —
n
~u u ^} Figure 4-37. S-N curves in
3======== tension-tension for unidi-
0.5 - o 50 v
/o rectionally reinforced bo-
* 50 v
/o FP/Mg-RT
FP/AI-RT ron (40 vol.%)/Al 6061, alu-
v
• 50 /o FP/Mg-260°C Mg Alloys -RT—*|
b 6061-T6 AI-RT-^XL mina-FP (50 vol.%)/Al, and
7079-T6 A I - R T ^ alumina-FP (50 vol.%)/Mg
Al Alloy- 260°C—"" composites (Champion
I etal., 1978).
I02 I0 3 I04 I0 5 I0 7
CYCLES TO FAILURE

McGuire and Harris (1974) studied the
fatigue behavior of tungsten fiber rein-
forced aluminum-4%-copper alloy under
tension-compression cycling (R= — 1).
They found that increasing the fiber vol-
ume fraction from 0 to 24% resulted in
increased fatigue resistance. This was a di-
rect result of increased monotonic strength
of the composite as a function of the fiber
volume fraction. It should be mentioned
that due to the highly anisotropic nature of
the fiber reinforced composites in general,
the fatigue strength of off-axis MMCs, just
like that of any fibrous composite, decreas-
es with increasing angle between the fiber
axis and the stress axis.
Fatigue crack propagation studies have
been done on aligned eutectic or in situ
composites as well. Since many of these in
situ composites are meant for high temper-
ature applications in turbines, their fatigue
behavior has been studied at temperatures
ranging from room temperature to
1100°C. The general consensus is that the
mechanical behavior of in situ composites,
static and cyclic strengths, is superior to
that of the conventional cast superalloys
(Stoloff, 1978).
A major drawback of studies of fatigue
behavior of a material using this S-N ap-
proach is that no distinction can be made
between the crack initiation phase and the
crack propagation phase.
Fatigue Crack Propagation Tests
Such tests are generally conducted in an
electrohydraulic closed-loop testing ma-
chine on notched samples. The results are
presented as log (da/dN) (crack growth per
cycle) vs. log AK (the alternating stress in-
tensity factor). Crack growth rate, da/dN,
is related to the cyclic stress intensity fac-
tor range, AK, according to the power low
relationship first formulated by Paris and
Erdogan (1963):
where A and m depend on the material and
test conditions. The applied cyclic stress
intensity range is given by
AK= YAa(a)1/2
where Y is a geometric factor, ACT is the
cyclic stress range, and a is the crack
length. The major problem in this kind of
tests is to make sure that there is one and
only one dominant crack that is propagat-
ing.
It was mentioned above that the com-
plexities in composites lead to the presence
of many modes of damage, such as matrix
 4.4 Properties 167

cracking, fiber fracture, delamination,

debonding, void growth, multidirectional
cracking, etc. And these modes appear
rather early in the fatigue life of com-
posites.
Progressive loss of stiffness during fa-
tigue of composites is a very important
characteristic, and it is very different from
the behavior of metals in fatigue. The dif- Figure 4-38. Damage zone in a conventional, homo-
ferent types of damage zones formed in an geneous, monolithic material (isotropic).
isotropic material such as a metal, or ce-
ramic, or polymer and anisotropic materi-
al such as a fiber reinforced composite are
shown in Figs. 4-38 and 4-39, respectively.
In the case of an isotropic material, a single
crack propagates in a direction perpendic-
ular to the cyclic loading axis (mode I load-
ing), Fig. 4-38. In the fiber reinforced com- Figure 4-39. Diffuse damage zone in a fiber reinforced
posite, on the other hand, a variety of sub- composite (anisotropic): (a) fiber break and local
critical damage mechanisms, Fig. 4-39 (a, debonding, (b) matrix cracking, deflection of the prin-
cipal crack along a weak fiber/matrix interface.
b, c), lead to a highly diffused damage
zone. Figure 4-40 shows schematically a
comparison of damage accumulation in a
composite and in monolithic, homoge-
neous materials under constant stress am-
Composites
plitude fatigue. We plot damage ratio
against cycle ratio. The damage ratio is the
current damage normalized with respect to
the damage at final failure. The cycle ratio,
similarly, is the number of cycles at a given
instant divided by the number of cycles to
failure. Life ^
In a homogeneous material, the term (cycles or time)
damage simply represents the crack length, Figure 4-40. Comparison of damage accumulation in
and not surprisingly it increases monoton- a fiber reinforced composite and a monolithic mate-
ically with cycling (Fig. 4-40). In the case rial.
of a composite, we do not have a simple
and unambiguous manifestation of dam-
Fatigue Behavior of Whisker and Particle
age, such as a crack length. Instead, dam-
Reinforced Composites
age means the crack density. Note that,
unlike in homogeneous materials, the Williams and Fine (1985 a) studied the
damage in composites accelerates at first microcrack initiation and growth in fa-
and then decelerates with cycling (Fig. tigue in silicon carbide whisker (20 vol-%)
4-40). This distinctive behavior is very im- reinforced 2124 aluminum alloy under fa-
portant. tigue. Fatigue crack initiation sites were
168 4 Metal Matrix Composites
the unbonded silicon carbide whiskers as
well as non-SiC intermetallies. Not unex-
pectedly, a reduction in the clustering of
SiC particles and the number and size of
the intermetallics led to an increased fa-
tigue life. The composite showed a slower
crack growth rate than the unreinforced
matrix. However, the unreinforced 2124-
T6 matrix was able to support cracks as
long as 500 \xm at the peak cycle stress
close to its yield stress while the composite
could support only a crack 30 \xm long at
60% of its yield strength.
Silicon carbide particle reinforced alu-
minum alloy composites have been studied
by researchers (Crowe and Hasson, 1982;
Williams and Fine, 1985 a, b; Logsdon and
Liaw, 1986; Shang etal., 1988; Davidson
1989; Kumai etal., 1990; Bonnen etal.,
1990; Williams and Fine, 1987; Christman
and Suresh, 1988 a, b). In general, in terms
of S-N curve behavior, the composite
shows an improved fatigue behavior vis a
vis the reinforced alloy. Such an improve-
ment in stress controlled cyclic loading or
high cyclic fatigue is attributed to the
higher stiffness of the composite. Howev-
er, the fatigue behavior of the composite,
evaluated in terms of strain amplitude vs.
cycles or low cycle fatigue, was observed to
be inferior to that of the unreinforced alloy
(Bonnen et al., 1990). This was attributed
to the generally lower ductility of the com-
posite compared to unreinforced alloy.
Fatigue Crack Initiation
Particles on short fibers can provide
easy crack initiation sites. The detailed be-
havior can vary depending on the volume
fraction of the reinforcement, shape, size,
and most importantly on the reinforce-
ment/matrix bond strength (Williams and
Fine, 1987). For example, they observed
fatigue crack initiation at the poles of SiC
whiskers in 2124 aluminum. They also ob-
served arrest of short cracks at the
whisker/Al interfaces.
Frequently in aluminum matrix com-
posites, especially those made by casting
route, there are present particles other
than SiC, such as CuAl2, (Fe,Mn) 3 SiAl 12 ,
and Cu 2 Mg 5 Si 6 Al 5 (Kumai and Knott,
1991). The phenomenon of particle push-
ing ahead of the solidification front results
in SiC particles and the so-called constitu-
ent particles decorating the cell boundaries
in the aluminum alloy matrix. The slow
cooling rate in conventional castings pro-
cess leads to rather coarse constituent par-
ticles and SiC segregating to the interden-
dritic region. Some possible reinforcement
and crack tip interactions are shown in
Fig. 4-41 while Fig. 4-42 shows schematic
representations of crack growth vs. cyclic
intensity factor (Kumai et al., 1990).
It would appear that choosing an opti-
mum particle size and volume fraction to-
gether with a clean matrix alloy will result
in a composite with improved fatigue char-
acteristics. Shiang et al. (1988) examined
the effect of particle size on fatigue crack
propagation as a function of cyclic stress
intensity in a silicon carbide particle rein-
forced aluminum. They observed that for
fine particle size, the threshold stress inten-
sity, AKth, for the composite was less than
that for the unreinforced alloy, i.e., initial
fatigue crack growth resistance of the com-
posite was less than that of the unrein-
forced alloy. For coarse particles, the
threshold intensity values were about the
same for the two, while at very high values
of the cyclic stress intensity, the fatigue
crack growth rate of the composite was
less than that of the unreinforced alloy.
As was pointed out above, laminated
fibrous composites can sustain a variety of
subcritical damage (cracking of matrix, fi-
ber/matrix decohesion, fiber fracture, ply
 4.4 Properties 169

Promoted by coarse particles, high cracking, delamination, etc.). For exam-

strength matrix ple, the cracking of a ply will result in a
relaxation of stress in that ply, and with
continued cycling no further cracking oc-
curs in that ply. Ply cracking generally in-
volves cracking in the matrix and along the
fiber/matrix interface but rarely any fiber
fracture. Other damage accumulating
mechanisms include the growth of existing
cracks into interfaces leading to ply delam-
Promoted by strong interfaces,
high modulus particles ination. The delamination of a ply results
in a reduction of stress concentration on
the neighboring plies. As can be seen, the
subcritical damage can accumulate rather
rapidly on cycling.
The various types of subcritical damage
mentioned above result in a reduction of
the load-carrying capacity of the laminate
composite, which in turn manifests itself as
Promoted by weak interfaces a reduction of laminate stiffness and
(precipitation, oxidation, reaction)
strength (Hahn and Kim, 1976; Highsmith
and Reifsnider, 1982; Talreja, 1985;
O'Brien and Reifsnider, 1981; Ogin et al.,
1985). Many researchers have experimen-
tally related the stiffness changes in the
laminated composites to the accumulated
damage under fatigue (Hahn and Kim,
1976; Highsmith and Reifsnider, 1982; Tal-
Figure 4-41. Some possible interactions between par- reja, 1985; O'Brien and Reifsnider, 1981).
ticulate reinforcement and a crack tip. It can safely be said that the change in
stiffness values is a good indicator of the
damage in composites. The modulus re-
duction rate — 1/EO dE/dN, at a given val-
ue of E/Eo, is the tangent to curve Ev at
that value of E/Eo where E is the current
modulus, Eo is the initial modulus of the
Monolithic uncracked material, and N is the number
alloy
of cycles. This rate of modulus reduction
can be described over a certain range of
peak fatigue stress by the expression (Ogin
etal., 1985)

Figure 4-42. Schematic representations of crack growth i d£ _

vs. cyclic intensity factor. (Kumai et al., 1990.) £diV
£odiV
170 4 Metal Matrix Composites
where A and n are constants and amax is the
peak fatigue stress. Stiffness loss in conti-
nous fiber reinforced MMCs and be a use-
ful parameter for detecting fatigue damage
initiation and growth (Johnson, 1989).
Laminated MMCs
There are two type of laminated MMCs:
(i) Metallic matrix containing fibers ori-
ented at different angles in different
layers. This is similar to that of poly-
meric laminates,
(ii) Two or more different metallic sheets
bonded to each other.
In the case of the fibrous laminate,
Johnson (1982) observed a significant loss
of stiffness in boron/aluminum laminate
when cycled in tension-tension. Also, in
manner similar to that in the laminated
PMCs, the ply stacking sequence affected
the fatigue behavior. For example, the 90°
surface plies in a 90°/0° sequence would
develop damage more rapidly than the 0°
plies. In the case of laminated composites
made out of metallic sheets, e.g., stainless
steel and aluminum, enhanced resistance
against fatigue crack propagation was ob-
served than in either one of the compo-
nents in isolation (Chawla and Liaw,
1979).
Hybrid Composites
Composites containing more than one
type of fibers are called hybrid composites.
Such composites, by using two or more
type of fibers, extend the idea of tailor-
making a composite material to meet
specific property requirements. Partial re-
placement of expensive fibers by cheaper
but adequate fiber types is another attrac-
tive feature of hybrid composites. An in-
teresting type of hybrid composite is made
of alternating layers of high strength alu-
minum alloy sheets and layers of unidirec-
tional aramid fibers in an epoxy matrix.
It is called aramid aluminum laminates
(ARALL). Improved fatigue resistance of
ARALL over that of monolithic alu-
minum structures is the main attractive
feature. Cracks can grow only a short dis-
tance before being blocked by the aramid
fibers spanning the crack tip. Figure 4-43
shows the slow fatigue crack growth char-
acteristics of ARALL laminates compared
to two monolithic aluminum alloys
(Mueller and Gregory, 1988). Applications
of ARALL are envisaged in tension domi-
nated fatigue structures such as aircraft fu-
selage, lower wing, and tail skins. It is ex-
pected that the use of ARALL will result in
15-30% weight savings over conventional
construction.
Structural Observations on MMCs
Subjected to Fatigue
Structural observations on metal matrix
composites after subjecting them to fatigue
can reveal a lot of important information
on the operating deformation mecha-
nisms, in particular such observations can
2024
l
1OU 10b 10b
Cycles
Figure 4-43. ARALL composite shows slow fatigue
crack growth characteristics compared to the mono-
lithic aluminum alloys (after Mueller and Gregory,
1990).

tell us how the metal matrix composites
behave under cyclic loading vis a vis the
monolithic metallic materials. Chawla
(1975 a) examined the dislocation struc-
tures produced in tungsten fiber reinforced
single crystal copper matrix composites.
As pointed out above, in single crystal cop-
per one can use etch-pitting technique to
study the dislocation structure in the bulk
material unlike the TEM techniques which
give information about an extremely small
region. The fiber/matrix thermal coeffi-
cient mismatch results, in cooling through
a temperature range of more than 1000 °C,
in a very heterogeneous dislocation distri-
bution in the matrix. Recall that the dislo-
cation density in the matrix rose to very
high value near the fiber/matrix interface
and decreased with distance from the inter-
face (see Figs. 4-13 and 4-25). There was a
high dislocation density, hard zone
(marked H) immediately next to the fiber
and a low dislocation density, soft zone
(marked S) away from the fiber. All struc-
tural observations on fatigue cycling were
made in the plateau or soft region, i.e., at
distances greater than 100 |im from the fi-
ber. At distances less than that the cell
structure was well developed to start with
and the only difference observed on cy-
cling was a reduction in cell size from the
original 10 jim diameter (approximate) to
a final 5 to 6 |im.
The dislocation structure produced after
the first cycle up to a stress of 15 MPa, and
which give a permanent strain of
8.6xlO~ 6 , showed dislocation pileups in
the copper matrix against well-defined
sub-boundary obstacles and less well-de-
fined cell wall obstacles. Near the fiber the
dislocation distribution, after the first cy-
cle, was indistinguishable from that exist-
ing in the as-prepared state. It would thus
appear that plastic deformation started in
this soft region away from the fiber while
4.4 Properties 171
the high dislocation region immediately
next to the fiber stayed relatively elastic at
this stage. The preexisting sub-boundaries
and, to a lesser extent some cell walls, act-
ed as obstacles to dislocation motion. The
dislocation structure produced on further
cycling to the same stress amplitude
showed that more plastic deformation
took place in this soft matrix region, lead-
ing to a greater number of pileups against
the above-mentioned obstacles and also
thickening of the sub-boundaries by col-
lection of dislocations in their vicinity.
Williams and Fine (1985 a, b, 1987) used
TEM technique to study the structural
changes occurring in SiC whisker rein-
forced aluminum matrix composites. The
unfatigued foil showed a nonuniform dis-
location distribution. High initial disloca-
tion density near the pole of a whisker pro-
moted fatigue crack initiation.
Thermal Fatigue
As pointed out above, thermal stresses
in composite materials are very important
because, in general, there is a large differ-
ence in the thermal expansion coefficient
of the ceramic and the metallic compo-
nents. Thus, even a small change in tem-
perature can cause rather large thermal
stresses. The presence of these thermal
stresses of large magnitude makes it imper-
ative that we understand the microstruc-
tural changes and the mechanical behavior
of composites under thermal cycling con-
ditions. In particular, structural damage
can occur on thermal cycling, which
should be taken into account in any appli-
cation involving cyclic temperature
changes. At the risk of repetition, we wish
to reiterate that thermal stresses arise in
composite materials due to the large differ-
ences in the thermal expansion coefficient,
a, of the fiber and the matrix. It should be
172 4 Metal Matrix Composites

emphasized that thermal stresses in com-

posites will arise even if the temperature
change is uniform throughout the volume
of the composite. Such thermal stresses
can be introduced in composites during
cooling from high fabrication or annealing
or curing temperatures or during any tem-
perature excursions (inadvertent or by de-
sign) during service. Turbine blades, for
example, are very much susceptible to
thermal fatigue. The magnitude of thermal
stresses in composites is proportional to
the thermal strain, Aa • AT", where Aa is the
difference in the expansion coefficient of
the two components and AT is the ampli-
tude of the thermal cycle (see Sec. 4.3.4).
In MMCs, the matrix generally has a
much higher coefficient of thermal expan-
sion than the fiber. Therefore, rather large
Figure 4-44. Multiple slip in copper single crystal
internal stresses can result when fiber rein- matrix in a W/Cu composite thermally cycled in the
forced composites are heated or cooled chamber of an SEM. The polished, undeformed cop-
through a temperature range. When this per matrix shows slip lines due to the plastic deforma-
happens in a repeated manner, we have tion (Chawla, 1973).
what is called the phenomenon of thermal
fatigue, because the cyclic stress is thermal
in origin. Thermal fatigue can lead to per matrix in a W/Cu composite (Fig.
cracking in a brittle matrix, e.g., a poly- 4-45 a), while grain boundary sliding and
meric or a ceramic matrix. In an MMC, migration occurred on cycling to very high
plastic deformation of a ductile metallic temperatures (Fig. 4-45 b). Besides cavita-
matrix is more likely (Chawla, 1973 a, b, tion in the matrix, fiber/matrix debonding
1976 a, b; Chawla, 1974, 1976; Garmong, can also occur due to thermal fatigue in
1975; Koss and Copley, 1971; Kwei and composites. Li et al. (1989) observed cavi-
Chawla, 1992; Li et al., 1989; Yoda et al., tation at the interface in an alumina fiber
1978). Some examples of microstructural reinforced magnesium alloy matrix that
damage that can occur on thermal cycling was subjected to 2500 thermal cycles be-
of metal matrix composites are given in tween room temperature and 300 °C.
Figs. 4-44 and 4-45. Figure 4-44 shows the In general, one can reduce the damage in
formation of slip lines (multiple slip) in a the matrix by choosing a matrix material
tungsten/copper single crystal matrix ther- that has a high yield strength and a large
mally cycled in the chamber of an SEM. strain to failure (i.e., ductility). The even-
The polished, undeformed copper matrix tual fiber/matrix debonding can only be
shows slip lines due to the plastic deforma- avoided by choosing the components such
tion. In a polycrystalline copper matrix, on that the difference in the thermal expan-
cycling to high temperatures, void forma- sion characteristics of the fiber and the ma-
tion occurred at grain boundaries in cop- trix is low.
 4.4 Properties 173

ulus, AQ is the activation energy for creep,

k is the Boltzmann's constant, and Tis the
temperature in kelvin. The power law ex-
plains the behavior at low to intermediate
stress levels and high temperatures while
the exponential relationship is generally
applicable at high stress levels and relative-
ly low temperatures. A typical creep curve
showing the three stages, primary, sec-
ondary, and tertiary is shown in Fig. 4-46.
Kelly and Tyson (1966) showed that incor-
poration of tungsten wires in silver can re-
duce the steady state creep rate, Fig. 4-47.

0.05

0.04
tertiary
. 0.03 -
i steady state stage
:
' 0.02

0.01 -primary stage

10 20 30
Time (h)
Figure 4-45. Microstructural changes from thermal Figure 4-46. Typical creep curve showing the pri-
cycling, (a) Void formation occurred at grain mary, secondary, and tertiary stages.
boundaries in a polycrystalline copper matrix in a
W/Cu composite, (b) Grain boundary sliding oc-
curred on further cycling.
• matrix
3 A composite
io-
4.4.8 Creep
Creep behavior of metals and alloys can
be conveniently described by combining
two types of expressions, a power and an
exponential relationship (Meyers and 8
Chawla, 1984). The basic governing equa-
tion of creep can be written in the follow-
io- 5
ing form 0.5 1 10
n
e = A (a/G) exp ( - AQ/kT) Stress, MPa
Figure 4-47. Incorporation of tungsten wires in silver
where e is the strain rate, A is a constant, can reduce the steady state creep rate (Kelly and
a is the applied stress, G is the shear mod- Tyson, 1966).
174 4 Metal Matrix Composites
T = 350°C / o
/
5.5- O/
^=6.3
/ was lower than that of unreinforced
n=6.0-
7 /
a -6.5-
/ /
0
/ n=io.oy
/ regime where the metal matrix shows a
E --7.5-j
/ AISi7Cu3
AISi7Cu3Mg
o AISi7Cu3-F
/ A AISi7Cu3Mg-F / / * - n = 1 0.2
J
A
-8.5- /
r /
log (stress/MPa)
Figure 4-48. Minimum creep rate as a function of ap-
plied stress for Saffil short fiber reinforced aluminum
alloys and unreinforced aluminum alloys (Dlouhy
et al., 1993). Note that the creep rate for the composite
is lower than that of the control alloy, but the slope
of the composite line is higher than that of the unrein-
forced alloy.
Figure 4-48 shows the minimum creep rate
as a function of applied stress for Saffil
short fiber reinforced aluminum alloys and
unreinforced aluminum alloys (Dlouhy
etal., 1993). Note that the creep rate for
the composite is lower than that of the
control alloy, but the slope of the com-
posite line is characteristically much higher
than that of the unreinforced alloy. Many
researchers have verified the high stress ex-
ponent, n around 20, for a discontinuously
reinforced aluminum vis a vis a stress ex-
ponent of n around 4 for the unreinforced
aluminum (Nieh, 1984; Nardone and
Strife, 1987; Morimoto etal., 1988; Drag-
one and Nix, 1992). Also, the activation
energy for these composite has generally
been observed to be very high, in the range
of 225-400 kJ/mol while the activation
energy for self diffusion in aluminum is
about 150kJ/mol.
Vasudevan et al. (1991) and Sadananda
etal. (1991) examined the creep behavior
of monolithic and SiC reinforced MoSi2,
in compression, and in the temperature
range of 1100-1400 °C. They observed
that the overall creep rate of the composite
/ molybdenum silicide. Some efforts have
/ been made at modeling the creep behavior
// of MMCs. The main problem is that in the
/
/ jf
/
steady state creep, the fibers will generally
/ / be deforming elastically.
4.4.9 Tribological Properties
Wear or tribological characteristics are
system properties rather than material
properties, i.e., operating wear mecha-
nism^) depends on the contact type
(rolling, sliding, etc.), the operating condi-
tions, environment, the material character-
istics of the test material and the mating
material. In the case of metal matrix com-
posites, the hard ceramic constituents such
as silicon carbide will be more wear resis-
tant than the metal matrix, the soft matrix
keeps the reinforcement in place and trans-
fers the applied load to the reinforcement.
Many of the conventional materials, gen-
erally not considered to be composites, do
have the characteristics of particulate com-
posites because they have hard, brittle in-
termetallics or carbides distributed in a
soft metallic matrix (Fine, 1991).
There is a variety of tests available for
wear resistance testing. Lubricated rolling
and sliding wear are two common ones.
For rolling contact tests, or roller-on-cyl-
inder is used, the roller is made of the
MMC test material and the cylinder is
made of cast iron. The sliding test, called
block-on-ring test, involves a stationary
test block (MMC) and a rotating steel ring.
The wear test is done at specific applied
normal load and rotational speed while a
jet of lubricant is sprayed on the steel ring.
The wear of the test material (MMC) is
 4.4 Properties 175

measured by measuring weight loss as a 160

2124-Al
function of sliding distance. 140 h
As in other areas, one would expect that I

continuous fiber reinforced composites 3120

in
will show anisotropy in their wear behav- o 100
ior as well (Eliezer et al., 1979; Greenfield 80
and Vignaud, 1985). The wear rate is high S
when the fibers are parallel to the wear €,60
surface but perpendicular to the sliding di- £ 40
rection. Ceramic particle (alumina, silicon
20 SiCw/2124 P
carbide, graphite, etc.) reinforced alu- T
minum alloys show excellent abrasive re- 0 L,
sistance with a variety of mating surfaces 20 40 60 80 100 120
(a) Sliding distance (km)
(Pan et al., 1990; Hosking et al., 1982; Ro-
hatgi, etal., 1990, 1992). Figure 4-49a
shows the weight loss vs. sliding distance
curves for different orientations of silicon 50

carbide whisker (20% by volume)/alu-

minum 2124 composites and unreinforced
aluminum 2124 under lubricated contact,
168 N load and 2 m s " 1 speed while Fig. §30
4-49 b shows the behavior under dry con- (/>
tact, 52 N load and l m s " 1 speed (Pan JO

etal., 1992). Analyses of the wear tracks 2 20

a)
and subsurface micro structure by SEM
and TEM showed that the adhesive wear 10
occurred through plastic deformation un- SiCw/2124

der dry sliding condition and abrasive

wear through plowing and polishing oc-
curred under lubricated conditions. An in-
teresting aspect of precipitation harden-
able aluminum matrix particulate com-
posites is the finding that the aging charac-
teristics of the matrix can affect the wear
behavior of the composite. Pan et al.
(1992) observed a decrease in the wear rate
of powder metallurgy alloy 2124 contain-
ing silicon carbide particles in the overaged
condition because of an easier crack prop-
agation along incoherent precipitate/ma-
trix interface than along SiC/matrix inter-
face. This resulted in less pullout of SiC
particles due to the change of the fracture
path from the SiC/matrix interface in the
underaged and peakaged to the grain
10 20 30 40
(b) Sliding distance (102m)
Figure 4-49. (a) Weight loss vs. sliding distance curves
for different orientations of silicon carbide whisker
(20 vol.%)/aluminum 2124 composites and unrein-
forced aluminum 2124 under lubricated contact,
168 N load and 2 ms" 1 speed, (b) Behavior under dry
contact, 52 N load and l m s " 1 speed (Pan et al.,
1992).
boundaries in the overaged condition. In
short, proper heat treatment can reduce
the extent of SiC particle pullout and thus
reduce the wear damage.
176 4 Metal Matrix Composites
4.5 Applications
It is convenient to divide the applica-
tions of metal matrix composites into
aerospace and non-aerospace categories.
In the category of aerospace applications,
low density coupled with other desirable
features such as a tailored thermal expan-
sion and conductivity, high stiffness and
strength, etc. are the main drivers. Perfor-
mance rather than cost is an important
item. Inasmuch as continuous fiber rein-
forced MMCs deliver superior perfor-
mance than particle reinforced com-
posites, the former are frequently used in
aerospace applications. In the non-
aerospace applications, cost and perfor-
mance are important, i.e., an optimum
combination of these items is required. It is
thus understandable that particle rein-
forced MMCs are increasingly finding ap-
plications in non-aerospace applications.
We provide below a brief description of
various applications of MMCs.
Aerospace Applications
Reduction in the weight of a component
is a major driving force for any application
in the aerospace field. Also, it should be
recognized that space environment is not
benign. Among the hazardous items that
may be encountered in space are orbital
debris, meteorites, atomic oxygen, etc. It
appears that metal matrix composites can
withstand the space environment better
than polymer matrix composites. In the
Hubble telescope, pitch-based continuous
carbon fiber reinforced aluminum was
used for waveguide booms because this
composite is very light, has a high elastic
modulus, E, and a low coefficient of ther-
mal expansion, a.
Other aerospace applications of MMCs
involve replacement of light but toxic
beryllium. For example, in the US Trident
missile, beryllium has been replaced by
SiCp/Al composite. Silicon carbide parti-
cle/aluminum composite is replacing car-
bon/epoxy in aircraft electronic equipment
racks.
Under the US National Aerospace
Plane (NASP) program, various US gov-
ernment agencies have funded perhaps the
world's first plant devoted solely to the
production of SCS type fiber reinforced
titanium matrix composites. The NASP
will experience very high skin tempera-
tures, especially at the leading edges.
Non-Aerospace Applications
One of the important applications of
MMCs in the automotive area is in the
Diesel piston crowns (Donomoto et al.,
1983). This application involves incorpo-
ration of short fibers of alumina or alumi-
na + silica in the crown of the piston. The
conventional Diesel engine piston has an
Al-Si casting alloy with crown made of a
nickel cast iron. The replacement of the
nickel cast iron by aluminum matrix com-
posite resulted in a lighter, more abrasion
resistant, and cheaper product.
Particulate metal matrix composites, es-
pecially with light metal matrix composites
such as aluminum, magnesium, also find
applications in automotive and sporting
goods. In this regard, the price per kg be-
comes the driving force for application. If
we take the price of direct chill (DC) alu-
minum ingot as unity, then the price of cast
SiCp/Al or A12O3/A1 composite will be
about 1.5 while the powder metallurgy
processed composite will be between 30-
35, and a composite made by co-spray will
be about 10. It is worth pointing out some
of the cost drivers. In the production of an
extrusion billet, the cost factors are only
material and sawing costs. The billets are
generally subjected to extrusion to pro-

duce a wrought alloy product. The cost
factors in the extrusion process are the
speed of extrusion, die cost, die wear, etc.
Cutting, machining, forming, welding, sur-
face finishing, and transportation are the
other cost drivers before a finished product
is available. Further items that should be
included in this list are recycling and recla-
mation, particularly so in aluminum ma-
trix composites because recycling has been
extremely successful in the aluminum in-
dustry. The two terms, recycling and recla-
mation, are used advisedly for composites.
Recycling means reuse of a worn out prod-
uct as a composite while reclamation in-
volves a recovery of the individual compo-
nents, i.e., aluminum and ceramic particles
separately (Lloyd, 1993). Cast MMCs can
be remelted and reused as a composite.
Although some of the issues such as parti-
cle/molten metal reaction are the same in
remelting, it should be appreciated that
scrap sorting and melt cleanliness and de-
gassing can be rather tricky problems in
MMCs. Special fluxing and degassing
techniques have been developed at Dural-
can (Lloyd, 1993). It would appear the
powder processed MMCs will also involve
remelting. The reclamation route also in-
volves remelting. According to Lloyd
(1993), using combined argon and salt
fluxing, silicon carbide particles can be re-
moved from the melt, and 85-90% of the
aluminum can be reclaimed.
Hoover (1992) provides a good example
of a potential commercial application of
particle reinforced aluminum composite to
make automotive driveshafts. In the de-
sign of a driveshaft, one needs to consider
the speed at which it becomes dynamically
unstable. This critical speed (in rpm) is giv-
en by
0.5
15K
N = 2 g(R2o-R?)
L \d
4.5 Applications 177
where E is the Young's modulus, L is the
length of shaft, g is the acceleration due to
gravity, d is the density, Ro and R{ are the
outer and inner radii, respectively. This ex-
pression says that in terms of geometric
parameters, a shorter shaft length and
larger diameter will give a higher critical
speed, Nc, while in terms of material
parameters, the higher the specific stiffness
(stiffness/density), the higher will be Nc.
The specific stiffness of two common ma-
terials used for shafts, steel and aluminum,
are about equal. Thus, the driveshafts
made out of these materials will have quite
similar dynamic properties. Hoover ana-
lyzed the use of a 6061 aluminum alloy
reinforced with 20 vol. % alumina parti-
cles. This particulate MMC had a specific
modulus about 40% higher than that of
steel or aluminum. The measured critical
speed, Nc was over 6000 rpm. Driveshafts
of identical dimensions but made of unre-
inforced aluminum have a critical speed of
5200 rpm. One can also make changes in
the driveshaft geometry (increase in length
or a reduction in the diameter) while main-
taining a constant critical speed. A de-
crease in the driveshaft diameter can be
important because of under-chassis space
limitations.
Copper based composites having Nb,
Ta, or Cr as the second phase in a discon-
tinuous form are of interest for certain ap-
plications requiring high thermal conduc-
tivity and high strength. Sometimes we
refer to these as Cu-X composites, where
X is insoluble in copper at room-tempera-
ture and forms the second phase. One
specific example is high heat flux appli-
cation in rocket engine thrust chambers.
Cu-X systems are very useful for process-
ing such composites. The second phase ap-
pears in a dendritic form at room tempera-
ture, which can be converted into a fila-
mentary or ribbon form by mechanical
178 4 Metal Matrix Composites
working. It should be pointed out that the
ribbon morphology is thermodynamically
unstable at high temperatures because the
ribbons tend to spherodize as a function of
time and temperature.
Carbon fiber reinforced copper can find
applications in the aerospace as a very high
thermal conductivity material. The prob-
lem with carbon fibers and metallic ma-
trices is that the surface energy consider-
ations preclude wetting of the carbon
fibers by the molten metals. The surface
energy of carbon fiber is about 100 mJ/m 2
while most molten metals have surface en-
ergy of about 1000 mJ/m 2 . Wetting occurs
when the surface energy of the substrate
(fiber) is higher than that of the molten
metal. Thus, one does not expect the wet-
ting of carbon fibers by molten copper.
Two possible routes around this difficulty
are: fiber surfaces modification or matrix
modification.
Filamentary Superconductors
Conventional commercial superconduc-
tors are referred to as niobium-based su-
perconductors because Nb-Ti and Nb 3 Sn
are the superconducting materials. It
should be recognized that these conven-
tional superconductors became commer-
cially viable after techniques were devel-
oped to put the superconducting material
in the form of extremely fine filaments in a
copper matrix (Chawla, 1987 a, p. 165),
i.e., the final product is a copper matrix
composite in the real sense of the word.
There are two reasons for using the super-
conducting material in a filamentary com-
posite form. Both of them have to do with
stabilization of the superconductor. Pin-
ning of magnetic flux lines is an absolute
necessity because a flux jump due to any
reason (magnetic, thermal, or mechanical)
can lead to quenching of the superconduc-
tor. Dividing the superconductor material
into smaller volume units in the form of
ultrathin (diameter in the jum range) fila-
ments is helpful because in this way the
heat capacity of the material is high
enough to limit the temperature increase.
This is called adiabatic stabilization.
Putting a high conductivity, normal mate-
rial such as high purity copper can dampen
the rate of flux change and can serve as an
efficient heat sink. This is called dynamic
stabilization (Foner and Schwarz, 1981;
Luhman and Dew-Hughes, 1979; Krauth,
1991). Krauth (1991) shows that for a
magnetic field perpendicular to the fila-
ment axis in the case of Nb-Ti supercon-
ductors at 4.2 K, a filament diameter of
less than 100 jim is required. Note that this
requirement of fine diameter filaments will
hold for the high temperature supercon-
ductors as well. If we can use them at 77 K,
then larger diameter filaments may be per-
missible because at 77 K, the specific heat
capacity values will be higher than at
4.2 K. Because the current carrying capac-
ity of ultrathin filaments will be rather low,
a large number of them are incorporated in
a normal conductor matrix such as copper.
Copper matrix, of course, provides the dy-
namic stabilization.
Niobium-titanium superconductors are
used in the magnetic resonance imaging
(MRI) techniques for medical diagnostics.
The superconducting solenoid made from
Nb-Ti/Cu composite wire, is immersed in
a liquid helium cryogenic Dewar. The liq-
uid helium is consumed at about 4 mL per
hour, and each refill of the Dewar lasts
about three months. Nb-Ti/Cu supercon-
ducting composites are also used in vari-
ous high-energy physics applications, such
as particle accelerators. They are also like-
ly to find applications in magnetically levi-
tated train project of the future. They are
being used in experimental test beds of

magnetically levitated trains. The Nb 3 Sn/
Cu superconducting composites are used
for magnetic fields greater than 12 T. Such
high fields are encountered in a thermonu-
clear fusion reactor, representing a sizeable
fraction of the capital cost of such a fusion
power plant.
Electronic Grade MMCs
Metal matrix composites can be tailored
to have optimized thermal and physical
properties to meet the requirements of
electronic packaging systems (e.g., cores,
substrates, carriers, and housings). We can
list the following as the desirable attributes
for electronic grade applications:
- Dimensional stability. This can be ob-
tained by controlling the following char-
acteristics of a composite:
- Coefficient of thermal expansion. For
example, the very high coefficient of
thermal expansion of aluminum,
(24 x 10 ~6 K~ x ) can be easily lowered
to a desirable value by adding an ap-
propriate amount of ceramic particle
such as SiC or alumina (for isotropy)
or carbon fibers (for anisotropy).
- high thermal conductivity for efficient
heat transfer from the substrate and
to even out any hot spots generated
on components.
- Lightness or low density is an important
consideration for airborne and space ap-
plications.
- Surface quality (as smooth as possible).
Use of MMCs for electronic packaging
must be regarded as a major application
that has come about in the last quarter of
the twentieth century. Many MMCs have
been produced that meet these require-
ments. Examples include: SiCp/Al, SiCw/Al,
carbon fiber (pitch-based)/Al, etc. A con-
trolled thermal expansion space truss, i.e.,
one having a high precision dimensional
4.6 References 179
tolerance in space environment was devel-
oped from carbon fiber (pitch-based)/Al
composite.
Continuous boron fiber reinforced alu-
minum composites made by diffusion
bonding have been used as heat sinks in
chip carrier multilayer boards. Lanxide
Electronic Components makes silicon car-
bide reinforced aluminum by their pres-
sureless infiltration process for high pow-
er, high density electronic circuit enclo-
sures, with a specific gravity about lA of
Kovar. Many other companies make such
components by different methods, for ex-
ample, DWA makes SiC particle or carbon
fiber reinforced powder metallurgy alu-
minum alloy composites, Advanced Com-
posite Materials Corporation makes SiC
whisker reinforced powder metallurgy alu-
minum alloy composites, while PCast
Equipment Corp. makes SiC/Al, A1N/A1,
C/Al, C/Cu, TiB2/Al composites by pres-
sure infiltration casting using an inert gas
to force liquid metal into a fibrous pre-
form.
4,6 References
Arsenault, R.J., Fisher, R.M. (1983), Scr. MetalL 17,
67-71.
Arsenault, R.J., Shi, N. (1986), Mater. Sci. Eng. 81,
175-186.
Bonnen, J.J., You, C.P., Allison, J.E., Jones, XW.
(1990), Proc. Int. Conf. on Fatigue, pp. 887-892.
Brooksbank, D., Andrews, K.W. (1972), /. Iron Steel
Inst. 4, 246-255.
Brockenbrough, J.R., Suresh, S. (1990), Scr. MetalL
Mater. 24, 325-330.
Champion, A.R., Krueger, W.H., Hartman, H.S.,
Dhingra, A.K. (1978), in: Proc. Int. Conf. Com-
posite Materials (ICCM/2), New York: TMS-
AIME, pp. 883-904.
Chawla, K.K. (1973 a), Metallography 6, 155-169.
Chawla, K.K. (1973 b), Phil. Mag. 28, 401-413.
Chawla, K.K. (1974), Microstructural Science, Vol. 2,
New York: Elsevier, p. 115.
Chawla, K.K. (1975 a), Fibre Sci. Technol. 8, 49-56.
Chawla, K.K. (1975 b), in: Grain Boundaries in Engi-
neering Materials, Proc. 4th Bolton Landing Conf.
Baton Rouge, LA.: Claitor's Publ. Div., p. 435.
180 4 Metal Matrix Composites
Chawla, K.K. (1976a), /. Mater. Sci. 11, 1567-1569.
Chawla, K.K. (1976 b), in: Proc. Int. Conf. Composite
Materials (ICCM/75), New York: TMS-AIME,
p. 535.
Chawla, K.K. (1976 c), J. Mater. Sci. 11, 1567-1569.
Chawla, K.K. (1985), /. Met. 37, 25-28.
Chawla, K.K. (1987 a), Composite Materials: Science
and Engineering. New York: Springer.
Chawla, K.K. (1987 b), Composite Materials: Science
and Engineering. New York: Springer, p. 110.
Chawla, K.K. (1987c), Composite Materials: Science
and Engineering, New York: Springer, p. 191.
Chawla, K.K., Liaw, P.K. (1979), J. Mater. Sci. 14,
2143-2150.
Chawla, K.K., Metzger, M. (1972), J. Mater. Sci. 7,
34-39.
Chawla, K.K., Metzger, M. (1978), in: Advances in
Research on Strength and Fracture of Materials,
Vol. 3, New York: Pergamon, p. 1039.
Chawla, K.K., Esmaeili, A.H., Datye, A.K., Vasude-
van, A.K. (1991), Scr. Metall. Mater. 25, 1315—
1319.
Chou, T.C., Joshi, A., Wadsworth, J. (1991), /. Mater.
Res. 6, 796-809.
Christman, T., Suresh, S. (1988 a), Acta Metall. 36,
1691-1704.
Christman, T., Suresh, S. (1988 b), Mater. Sci. Eng.
102'A, 211-216.
Christman, T., Needleman, A., Suresh, S. (1989), Ac-
ta Metall. 37, 3029-3047.
Cline, H.E., Walter, E.F., Koch, E.F., Osika, L.M.
(1971), Ada Metall. 19, 405-414.
Cook, A.J., Warner, P.S. (1991), Mater. Sci. Eng.
A144, 189.
Crowe, C.R., Hasson, D.F. (1982), in: Proc. 6th Int.
Conf. on the Strength of Metals and Alloys. Oxford:
Pergamon, pp. 859-865.
Das, G. (1990 a), in: Metal and Ceramic Matrix Com-
posites, Warrendale, PA: The Minerals, Metals and
Materials Soc, p. 413.
Das, G. (1990b), Metall. Trans. 21A, 1571-1578.
Davidson, D.L. (1989), Eng. Fract. Mech. 33, 965-
977.
DiCarlo, J.A. (1983), in: Mechanical Behavior of Met-
al Matrix Composites. New York: TMS-AIME,
p.l.
Dhingra, A.K. (1980), Phil. Trans. R. Soc. Lond.
A294, 559-564.
Dickson, R.F., Jones, C.J., Harris, B., Leach, D.C.,
Moore, D.R. (1985), /. Mater. Sci. 20, 60-70.
Dlouhy, A., Merk, N., Eggeler, G. (1993), unpub-
lished work.
Donomoto, T, Miura, N., Funatani, K., Miyake, N.
(1983), SAE Tech. Paper No. 83052, Detroit, MI.
Dragone, T.L., Nix, W.D. (1992), Ada Metall. Met.
40, 27'81.
Dunand, D.C., Mortensen, A. (1991), Mater. Sci.
Eng. A144, 179.
Dutta, I., Bourell, D.L. (1989), Mater. Sci. Eng.
A112, 67-77.
Dutta, I., Allen, S.M., Hafley, XL. (1991), Metall.
Trans. 22A, 2553.
Eliezer, Z., Khanna, V.D., Amateau, M.F. (1979),
Wear 53, 387-389.
Esmaeili, A.H., Chawla, K.K., Datye, A.K., Vasude-
van, A.K., Mei, Q., (1991), in: Proc. Int. Conf. on
Composite Materials VIII.
Fine, M.E. (1991), in: Advances in Engineering Tribol-
ogy. Soc. of Tribologists and Lubrication Engi-
neers, pp. 151-162.
Foner, S., Schwarz, B.B. (Eds.) (1981), Superconduct-
ing Materials Science. New York: Plenum.
Friend, CM., Luxton, S.D. (1988), /. Mater. Sci. 23,
3173.
Friend, C M . (1989), Mater. Sci. Technol. 5, 1-7.
Fu, L.-X, Schmerling, M., Marcus, H.L. (1986), in:
Composite Materials: Fatigue and Fracture, ASTM
STP 907. Philadelphia: Am. Soc. for Testing Ma-
ter, p. 51.
Garmong, G. (1975), Metall. Trans. 6A, 1179.
Girot, F , Quenisset, J.M., Naslain, R. (1987), Rev.
Compos. Sci. Technol. 30, 155-184.
Greenfield, I.G., Vignaud, R.R. (1985), in: Advances
in Composites: Proc. Metals Park, ASM, OH:
pp. 213-221.
Hahn, H.T., Kim, R.Y. (1976), /. Composite Mater.
10, 156.
Hancock, J.R., (1974), in: Composite Materials: Frac-
ture and Fatigue, Vol. 5. New York: Academic,
p. 371.
Hasson, D.F., Crowe, C.R. (1985), in: Strength of
Metals and Alloys. Oxford Pergamon, pp. 1515-
1520.
Highsmith, A.T., Reifsnider, K.L. (1982), in: Damage
in Composite Materials, ASTM STP 775, Philadel-
phia: Am. Soc. for Testing Mater., p. 103.
Hill, R.G.R.P., Nelson, R.P., Hellerich, C.L (1969),
in: Proc. 16th Refractory Working Group Meeting,
Oct., Seattle, WA.
Hillmann, H. (1981), in: Superconductor Materials
Science: Foner, S., Schwarz, B.B. (Eds.). New
York: Plenum, p. 275.
Hillmann, H., Rudinger, K., Heisterkamp, F , Wag-
ner, K.P., Hulka, K., Sibum, H. (1989), /. Miner.
Met. Mater. Soc, Oct. 20-22.
Hoover, W. (1992), Mater. Eng., 30-32.
Hosking, F.M., Portillo, S.V., Wunderlin, R., Mehra-
bian, R. (1982), J. Mater. Sci. 17, 411-498.
Hull, A. W, Burger, E. E. (1934), Physics 5, 384-405.
Isaacs, J.A., Taricco, F , Michaud, V.X, Mortensen, A.
(1991), Metall. Trans. 22A, 2855-2862.
Jin, S. (March, 1991), J. Miner. Met. Mater. Soc. 43,
7-12.
Johnson, W.S. (1982), in: Damage in Composite Mate-
rials, ASTM STP 775, Philadelphia: Am. Soc. for
Testing Mater., p. 83.
Johnson, WS. (1989), in: Metal Matrix Composites:
Testing, Analysis, and Failure Modes, ASTM STP
1032: Johnson, W.S. (Ed.). Philadelphia: Am. Soc.
for Testing Mater, p. 194.

Katzman, H.A. (1987), J. Mater. Sci. 22, 144-148.
Kelly A, Lilholt, H. (1969), Phil. Mag., 311-328.
Kelly, A., Tyson, W.R. (1966), /. Mech. Phys. Solids
14, 177.
Kerner, E.G. (1956), Proc. Phys. Soc. 69, 808-813.
Kim, C.T., Lee, J.K., Plichta, M.R. (1990), Metall.
Trans. 21A, 673.
Krishnamurthy, S. (1990), in: Interfaces in Metal-Ce-
ramics Composites. Warrendale, PA: The Mineral,
Metals and Materials Soc, p. 75.
Koss, D.A., Copley, S.M. (1971), Metall. Trans. 2,
1557-1560.
Krauth, H. (1991), in: Superconducting Technology,
Fossheim, K. (Ed.). Singapore: World Scientific,
p. 149.
Kumai, S., Knott, IF. (1991), Mater. Sci. Eng. A146,
317-326.
Kumai, S., King, J.E., Knott, J.F. (1990), Fatigue
Fract. Eng. Mater. Struct. 13, 511-524.
Kwei, L., Chawla, K.K. (1992), J. Mater. Sci. 27,
1101-1106.
Lagace, H., Lloyd, D.J. (1989), Can. Met. Quarterly
28, 145-152.
Levi, C.G., Abbaschian, G.X, Mehrabian, R. (1978),
Met. Trans. 9A, 696.
Lewis, D. (1991), in: Metal Matrix Composites; Pro-
cessing and Interfaces. San Diego: Academic,
p. 121.
Li, C.-S., Chawla, K.K., Rigsbee, J.M., Pfeifer, M.
(1989), in: Cast Metal Composites. Materials Park,
OH: ASM Int., p. 301.
Liaw, P.K., Ijiri, Y, Taszarek, B.J., Frohlich, S., Gun-
gor, M.N. (1988), Metall. Trans. 19A, 2215-2224.
Livingston, ID. (1962), Ada Metall. 10, 229.
Lloyd, D.J. (1992), in: Proc. 3rd Int. Conf. on Alu-
minum Alloys, June 22-26,1992, Trondheim, Nor-
way.
Lloyd, D.J. (1994), Int. Mater. Rev., to be published.
Logsdon, W.A. (1990), Metall. Trans. 21A, 529-538.
Logsdon, W.A., Liaw, P.K. (1986), Eng. Fract. Mech.
24, 737-751.
Lorimer, G.W. (1978), Precipitation Processes in
Solids. Warrendale, PA. Metallurgical Society,
p. 88.
Lorreto, M.H., Konitzer, D.G. (1990), Metall. Trans.
21A, 1579-1587.
Luhman, T., Dew-Hughes, D. (Eds.) (1991), Metal-
lurgy of Superconducting Materials. New York:
Academic.
Mahon, G.J., Howe, J.M., Vasudevan, A.K. (1990),
Ada Metall. Mater. 38, 1503-1512.
Majidi, A.P., Chou, T.W (1987), Proc. ICCM VI,
Vol. 2, pp. 422-430.
McDanels, D.L. (1985), Metall. Trans. 16A, 1105-
1115.
McDanels, D.L., Signorelli, R.A. (1976), NASA TN
D 8204.
McGuire, M.A., Harris, B. (1974), /. Phys. D: Appl.
Phys. 7, 1788-1802.
4.6 References 181
McKimpson, M.G., Scott, T.E. (1989), Mater. Sci.
Eng. A107, 93-106.
Meyers, M.A., Chawla, K.K. (1984), Mechanical
Metallurgy: Principles and Applications. Engle-
wood Cliffs, NJ: Prentice-Hall.
Mitra, R., Chiou, W.A., Weertman, J.R., Fine, M.E.,
Aikin, R.M. (1991), Scr. Metall. Mater. 25, 2689-
2694.
Morimoto, T, Yamaoka, T., Lilholt, H., Taya, M.
(1988), /. Eng. Mater. Technol. 110, 71.
Mueller, L.R., Gregory, M. (1990), Paper presented
at I Annual Metals and Metals Processing Conf. of
SAMPE Aug., Cherry Hill. NJ.
Nardone, V.C., Strife, J.R., Metall. Trans. 18A, 109-
114.
Nieh, T.G. (1984), Metall. Trans. 15A, 139-146.
Nieh, T.G., Karlak, R.F (1984), Scripta Metall. 18,
25.
Nourbakhsh, S., Liang, R.L., Margolin, H. (1990),
Metall. Trans. 21 A, 213-228.
Nutt, S.R., Carpenter, R.W (1990), Mater. Sci. Eng.
76, 169-177.
Pan, Y.M., Cheng, H.S., Fine, M.E. (1990), in: Funda-
mental Relationship between Microstructure and
Mechanical Properties of Metal Matrix Com-
posites. Warrendale, PA: The Mineral, Metals and
Materials Soc, p. 637-653.
O'Brien, T.K., Reifsnider, K.L. (1981), /. Composite
Mater. 15, 55.
Ogin, S.L., Smith, P.A., Beaumont, P.WR. (1985),
Compos. Sci. Technol. 22, 23-31.
Pan, Y.M., Fine, M.E., Cheng, H.S. (1992), Tribol.
Trans. 35, 482-490.
Papazian, J.M. (1988), Metall. Trans. 19A, 2945-
2953.
Paris, P.C., Erdogan, F. (1963), J. Basic Eng. (Trans.
ASME, Ser. D) 85, 528-534.
Pfeifer, M., Rigsbee, J.M., Chawla, K.K. (1990), /.
Mater. Sci. 25, 1563-1567.
Poritsky, H. (1934), Physics 5, 406-411.
Porter, D.A., Easterling, K.E. (1981), Phase Transfor-
mation in Metals and Alloys, New York: Van Nos-
trand Reinhold, p. 308.
Premkumar, M.K., Hunt, W.H., Sawtell, R.R. (1992),
/. Miner. Met. Mater. Soc. July 24-28.
Prewo, K.W. (1977), United Technologies Res. Cen-
ter. Rep. R77-912245-3.
Rack, H.J. (1987), in: Proc. 6th Int. Conf. on Com-
posite Materials. New York: Elsevier, p. 382.
Reed, R.P., Mikesell, R.P., Clark, A.F. (1977), Adv.
Cryog. Eng. 22, 463.
Rohatgi, P.K., Blau, P.J., Yust, C.S. (Eds.) (1990),
Tribology of Composite Materials. Materials Park,
OH: ASM International.
Rohatgi, P.K., Ray, S., Liu, Y. (1992), Int. Mater.
Rev. 37, 129.
Sadananda, K., Jones, H., Feng, C.R., Petrovic, I I ,
Vasudevan, A.K. (1991), in: Ceram. Eng. Sci. Proc.
Westerville, OH: Am. Ceram. Soc
182 4 Metal Matrix Composites
Sandhage, K.H., Riley, Jr., G.N., Carter, W.L.
(1991), J. Miner. Met. Mater. Soc. March, 21-25.
Schapery, R.A. (1968) J. Compos. Mater. 2, 380-404.
Schulyer, D.R., Sohi, Mahapatra, R. (1990), in: Inter-
faces in Metal-Ceramics Composites. Warrendale,
PA: The Minerals, Metals and Materials Soc, p. 5.
Shang, J.K., Yu, W., Ritchie, R.O. (1988), Mater. Sci.
Eng. A102, 181-192.
Sharp, G.R., Loftin, T.A. (1992), Personal communi-
cation from DWA Speciality Composites Co., CA.
Singh, R.K., Narayan, J. (1991), /. Miner. Met. Ma-
ter. Soc, March, 13-20.
Stoloff, N.S., in: Advances in Composite Materials.
London: Applied Sci. Publ., p. 247.
Sun, C.T., Prewo, K.M. (1977), J. Compos. Mater. 11,
164.
Suresh, S., Brockenbrough, J.R. (1993), in: Metal
Matrix Composites, Suresh, S., Needleman, A.,
Mortensen, A. (Eds.) Stoneham, MA: Butter-
worth-Heinemann.
Suresh, S., Chawla, K.K. (1993), in: Fundamentals of
Metal Matrix Composites: Suresh, A., Needleman,
A., Mortensen, A. (Eds.). Stoneham, MA: Butter-
worth-Heinemann.
Suresh, S., Christman, T., Sugimura, Y. (1989), Scr.
Metall 23, 1599-1602.
Talreja, R. (1985), Fatigue of Composite Materials,
Lyngby: Technical Univ. of Denmark.
Turner, P.S. (1946), J. Res. Natl. Bur. Stand. 37, 239-
250.
Vaidya, R.U., Chawla, K.K. (1991), in: Developments
in Metal and Ceramic Matrix Composites: Upad-
hya, K. (Ed.). Warrendale, PA: The Minerals,
Metals, and Materials Society, pp. 253-272.
Vasudevan, A.K., Doherty, R.D. (Eds.) (1989), Alu-
minum Alloys-Contemporary Research Applica-
tions. Boston: Academic.
Vasudevan, A.K., Petrovic, J.J. (Eds.) (1992), High
Temperature Structural Silicides, Amsterdam:
North-Holland.
Vasudevan, A.K., Petrovic, J.J., Sadananda, K.
(1991), in: Metal Matrix Composites-Processing,
Microstructure and Properties. Roskilde, Denmark;
Ris0 Natl. Lab., p. 707-712.
Vogelsang, M., Arsenault, R.J., Fisher, R.M. (1986),
Metall. Trans., 17A, 379-389.
Wang, L., Arsenault, RJ. (1991), Metall. Trans. 22A,
3013-3021.
Wang, N., Wang, Z., Weatherly, G.C. (1992), Metall.
Trans., 23A, 1423-1430.
Warren, R., Andersson, C.-H. (1984), Composites 15,
101-111.
Williams, D.R., Fine, M.E. (1987), Proc. 6th Int.
Conf on Composite Materials, Vol. 2. London: El-
sevier, pp. 113-118.
Williams, D.R., Fine, M.E. (1985), Proc. 5th Int.
Conf Composite Materials., pp. 275-290.
Williams, D.R., Fine, M.E. (1985), in: Proc. 5th Int.
Conf. Composite Materials (ICCM/V), Warren-
dale, PA: TMS-AIME, p. 639.
Willis, T.C. (1988), Met. Mater., Aug., 485.
Yoda, S., Kurihara, N., Wakashima, K., Umekawa,
S. (1978), Metall. Trans. 9A, 1229-1236.
Zweben, C. (1992), /. Miner. Met. Mater. Soc, July,
15-23.
General Reading
Arsenault, R.J., Everett, R. (Eds.) (1990), Metal Ma-
trix Composites. Orlando, FL: Academic.
Chawla, K.K. (1987), Composite Materials: Science
and Engineering. New York: Springer.
Suresh, S., Mortensen, A., Needleman, A. (Eds.)
(1993), Fundamentals of Metal Matrix Composites.
Stoneham, MA: Butterworth-Heinemann.
Taya, M., Arsenault, RJ. (1989), Metal Matrix Com-
posites. Oxford: Pergamon.