Вы находитесь на странице: 1из 32

Functionally graded metals and

metal-ceramic composites:
Part 2 Thermomechanical behaviour
S. Suresh and A. Mortensen

Following a review of the processing of In Part 1 (Ref. 1) of this two part series, attention
functionally graded metals and metal-ceramic
was focused on the different methods and processes
composites in Part 1; this Part 2 of the two part
series focuses on the thermomechanical by recourse to which functionally graded metals and
behaviour. The paper begins with an overview of metal-ceramic composites could be synthesised. Also
the fundamentals of thermoelastic and addressed in Part 1 were the overall principles under-
thermoplastic deformation in metal-ceramic lying the concept of functionally graded materials
composites. Various approaches, including the whereby gradual transitions in microstructure and/or
rule of mixture approximations, mean field composition, motivated by spatially varying func-
theories, crystal plasticity models, discrete
tional performance requirements within a single com-
dislocation models, and continuum finite element
ponent, could be engineered from currently available
Published by Maney Publishing (c) IOM Communications Ltd

formulations of the constitutive phases of the

composites, are discussed, and the significance materials and processing technologies. In the present
and limitations of these approaches are paper, Part 2, the micromechanics and physical
highlighted. Issues specific to the rationale for the purposeful introduction of controlled
thermomechanical analyses of graded materials gradients in the composition, microstructure, and
are then addressed. It is reasoned that the properties for the specific purpose of enhancing the
introduction of a new length scale to the problem thermomechanical performance of a structural mater-
due to compositional gradients inevitably calls for
detailed micromechanical analyses of the size,
ial are examined.
shape, continuity, and spatial dispersions of the When dissimilar materials, such as metals and
constituent phases of graded metal-ceramic ceramics, are joined or mixed to form a layered or a
composites. Models for the thermal, elastic, and composite structure, temperature changes induced
plastic deformation of graded multilayers are then during cooling from the joining/processing temper-
presented within the context of classic beam and ature and during subsequent service can generate
plate theories, in conjunction with strategies for high internal stresses as a result of unequal thermal
developing 'design diagrams' for
expansion or contraction between the constituent
thermomechanical performance. Methods to
identify the conditions governing the onset of
phases. Similarly, different local deformation fields
instability and a.brupt shape changes due to large are induced in the two materials during imposed
deformation in graded multilayers are also mechanical loading as a consequence of their differing
provided. The macroscopic continuum analyses elastic and plastic properties. Such an incompatibility
are followed by discussions of micromechanics or misfit in thermomechanical deformation can engen-
simulations of the real microstructural dispersions der internal stresses and strains, and locally elevated
by recourse to computational models which invoke values of triaxiality which, as schematically illustrated
von Mises type and crystal plasticity theories.
Experimental methods to assess the validity of
in Fig. 1, playa dominant role in the creation of such
such models are then examined, along with typical damage features as misfit or threading dislocations,
results of processing induced internal stresses and localised yielding with a high gradient in plastic
thermal stresses arising from temperature strains in the vicinity of the interfaces, interfacial
excursions in model systems with gradients in debonding, micro cracking, and brittle failure of the
metal-ceramic concentrations. It is demonstrated hard phase, or separation by void nucleation and
that stepwise or continuously graded metal- growth in and around the ductile phase. In many
ceramic composites can be designed to improve
cases, mitigating the severity of abrupt transitions in
interfacial bonding between dissimilar solids, to
minimise and optimally distribute thermal
composition or microstructure at both the micro-
stresses, to suppress the onset of plastic yielding, scopic and macroscopic scales by the introduction of
to mitigate the deleterious effects of singular fields a graded structure offers the possibility to control
at free edges of multilayers where interfaces and distribute damage and failure optimally.2-12 In
intersect free surfaces, to reduce the effective addition, when sharp interfaces between highly dis-
driving force for fracture, and to arrest cracks. similar solids, such as between a metal and a ceramic,
IMR/298 intersect a free surface, the corner at the point of
© 1997 The Institute of Materials and ASM International. intersection serves as a site of stress singularity in the
S. Suresh is R. P. Simmons Professor in the Department of vicinity of which elevated levels of a multi axial
Materials Science and Engineering and Professor of stress-strain field can promote the inception of flaws.
Mechanical Engineering at Massachusetts Institute of
Such 'edge effects', schematically shown in Fig. la,
Technology, Cambridge, MA 02139-4307, USA.
A. Mortensen is Professor, Laboratoire de Metallurgie can be suppressed, and consequently the damage
Mecanique, Departement des Materiaux, Swiss Federal tolerance enhanced, in some cases by introducing a
Institute of Technology, EPFL, CH-1015 Lausanne, smoother transition in the microstructure of the
Switzerla nd. interface between the metal and the ceramic over a
finite distance.
International Materials Reviews 1997 Vol. 42 No.3 85
86 Suresh and Mortensen Functionally graded metals: Part 2

State of in-plane normal stress only

Plastic zone due to

{ thermal and/or mechanical loading
Interface }
di slocation
Plastic zone during
~ { thermal excursion
Free edge
is a region of Sharp interface
singular fields Free edge where multiaxial
norma I and shear stresses
(0 ) prevail and cracks are likely

Constrained plastic 0 0 0
zone during thermal/ )'0 0 0
mechanica I loading /
0 o 0 0
0 0
/ 0 0 0
Ceramic particle /
Published by Maney Publishing (c) IOM Communications Ltd

Interface at which dislocations

are punched out to relieve
(b) misfit strains
1 Some characteristics of thermomechanical deformation in a metal-ceramic layered plates and b ceramic
particle reinforced metal matrixes

Functionally graded materials can, at least in prin- of diesel engines used in surface transportation
ciple, be designed to tailor thermomechanical charac- vehic1es.3o,31 The mechanical integrity of such a thick
teristics by the appropriate choice of gradients in coating can only be ensured by a stepwise increase in
composition or microstructure in view of the follow- the ceramic content from the interior to the outer
ing potential benefits: surface, by depositing a number of metal-ceramic
1. The magnitude of thermal stresses and the criti- composite sublayers of increasing ceramic content;
cal locations at which they act can be judiciously the deposition of the ceramic directly on to the metal
controlled.13-23 will in this case result in interfacial delamination even
2. The onset of plastic yielding and failure can be before the component enters service.
delayed for a given thermomechanicalloading.15-19 8. Gradients in the composition of surface layers
3. The severity of stress concentrations and can be tailored to suppress the singular fields which
singularities at free edges of interfaces can be arise at the root of sharp indentations on the surface32
suppressed.18,24 or to alter the plastic deformation characteristics
4. The strength of the interfacial bond between around the indentation.33 These processes are of
dissimilar solids, such as a metal and ceramic, can be considerable interest for the potential design of ther-
increased by the introduction of continuous or step- mal barrier, tribological, or impact resistant coatings
wise gradations in composition as compared to a (such as for turbine blades for aircraft jet engines
sharp interface.9,25 which are susceptible to domestic and foreign object
5. The density and kinetics of misfit/threading dis- damage), or of graded armour materials with
locations generated at the interfaces between dis- improved resistance to high strain rate deformation
similar solids can be altered.26 (This is indeed the and impact. It is also known34 that stepwise gradients
objective in some optoelectronic devices wherein, for in the alignment of fibres in continuously reinforced
example, a graded layer in InGaAs, produced by organic composites can be designed to improve sig-
molecular beam epitaxy or chemical vapour depos- nificantly their indentation resistance and to suppress
ition, is sandwiched between InGaAs and GaAs micro cracking damage.
layers.26) The present paper is arranged in the following
6. The driving force for crack growth through sequence. Since developing a complete understanding
and across an interface can be reduced by tailor- of the effective properties of composites is central to
ing the interface with gradients in mechanical the eventual success in the design of graded metal-
properties.24,27-29 . ceramic composites with optimised thermomechanical
7. Some applications necessarily involve the depos- performance, first there is a detailed discussion of
ition of a 'thick' brittle coating on a ductile substrate different approaches to composite modelling which
for protection against thermal exposure, environmen- include the simple rule of mixture formulations, mean
tal attack, contact failure, or wear. An example of field theories and their various adaptations, finite
such an application is the thermal spraying of a element unit cell models, and numerical modelling of
zirconia layer, more than 2 mm in thickness, on a a large sample of the real microstructure. This dis-
steel substrate for thermal protection in piston heads cussion is followed by analyses of the evolution of
International Materials Reviews 1997 Vol. 42 NO.3
Suresh and Mortensen Functionally graded metals: Part 2 87

stresses, strains, and curvatures in graded multilayered In this 'rule of mixtures', E is the isotropic Young's
structures. For this purpose, classic beam/plate theor- modulus, I the volume fraction of the phase, and the
ies of continuum mechanics are employed to develop subscripts 1 and 2 denote phase 1 and phase 2,
a theoretical foundation for elastic and elastoplastic respectively. Very small differences from the prediction
response of graded materials within the context of of equation (1) would be expected for the above equal
small strain, small deformation analyses. Departures strain case when the isotropic values of the Poisson
from small deformations during the thermomechan- ratios of the two phases are not equal.
ical deformation of graded multilayers are examined An alternative approach, often relevant to the
next in an attempt to identify the geometry and transverse loading of a unidirectionally reinforced
loading conditions which promote instabilities, shape continuous fibre composite, invokes the assumption
changes, and bifurcation. The descriptions of these that each phase in the composite carries an equal
deformation theories are then followed by a brief stress during the imposition of an external load. In
summary of available methods to monitor the evol- this case, the so called Reuss model,44 the overall
ution of stresses, strains, and curvatures in graded strain in the composite is the sum of the net strain
multilayers, and examples of such measurements in carried by each phase, and the effective composite
model systems are presented. Attention is then modulus is given by
directed to the fracture of graded materials, where
parameters to characterise the driving force for frac- E = (11 + 12)-1
e (2)
ture are examined. This is followed by a discussion £1 E2
of crack growth across graded interfaces. The paper
Published by Maney Publishing (c) IOM Communications Ltd

Various adaptations of equations (1) and (2) are

concludes with some overall strategies for alleviating
also commonly made, as shown in subsequent
the deleterious effects of sharp transitions in material
sections, to compute the effective shear moduli, bulk
composition and/or microstructure via gradation for
moduli, and Poisson ratios of the composite from a
the specific purpose of enhancing thermomechanical
knowledge of the corresponding values for the two
performance, and with some suggestions of areas in
phases and of the phase concentrations. Invoking
which future research might fruitfully be directed.
the principle of minimum potential energy, Hi1l4s
showed that the rule of mixtures provides an upper
Determination of effective properties bound for the elastic moduli for the composite.
of metal-ceramic composites
Modified rule of mixture approximations
A finite transition in materials spanning a range that
extends from one material or phase to another neces- In many practical situations, such as in two phase
sarily comprises an intermediate region which is a composites where one phase is discontinuous in a
composite of the two extremes if, as is generally the continuous matrix of the other phase, neither the
case, the extremes are not fully miscible. Therefore, Voigt model nor the Reuss model provides an accu-
the vast majority of graded structures are mostly rate description of the effective modulus of the com-
made of composite or multiphase materials. posite under both normal and shear loading. In these
Thermomechanical analyses of graded structures situations, empirically adjusted combinations of equa-
inevitably call for methods by which the effective tions (1) and (2) are commonly employed to describe
properties of the underlying dual phase or multiphase the experimentally observed stress-strain character-
structures can be determined. For this reason, in this istics of the composite. One well known expression
section analytical and computational methods which of this kind is the Halpin- Tsai46 equation for the
offer estimates of the composite properties of the transverse Young's modulus of a fibre reinforced
metal-ceramic graded structure are considered, in an composite which widely allows for increased load
attempt to establish a general basis on which quanti- carrying of the fibres
tative analyses for graded materials can be predicated.
E = {1- II [(~: -1) (~: + A) -IJ}-I
c E2
Rule of mixture approximations
The overall mechanical properties of a two phase
material are generally dependent on such factors as
the concentration, shape, and contiguity, and the (3)
spatial distribution of each phase.3s-42 Consider the Here the subscripts 1 and 2 refer to the properties of
simple case of a composite comprising two elastically the fibre and the matrix, respectively, and A is an
isotropic constituent phases, wherein the applied load empirical parameter whose magnitude is of the order
causes equal strains in each of the phases at all times. of unity.
Such a situation, commonly referred to as the Voigt Other modifications to the rule of mixture formu-
model,43 is encountered typically when a unidirec- lations have been employed for two phase cemented
tional continuous fibre reinforced composite is carbides47,48 and later for metal-eeramic graded com-
stressed along the fibre direction. In this case, the posites.1s,16,18,19 Here the composite is treated as
overall composite stress is the sum of the stresses isotropic (as, for example, may be envisioned for
carried by each phase; the composite modulus is then powder metallurgy processing), with its uniaxial stress
the weighted average of the moduli of the constituents and strain given, xespectively, by
. . . . (1) (J e = 11 (J 1 + /2 (J 2; 8e = /181 + /282 • ( 4)

International Materials Reviews 1997 Vol. 42 No.3

88 Suresh and Mortensen Functionally graded metals: Part 2

where subscripts 1 and 2 refer to the stresses, strains,

and volume concentrations of phases 1 and 2, respect-
ively. The ratio of the stress to strain transfer is then
defined by the parameter "./
/// Metal
. . . . . . . (5) .;:

Obviously, q ~ 0 and q ~ 00 refer to the equal stress o
and equal strain averaging schemes given in equations c:
(2) and (1), respectively. Combining equations (4)
and (5), one obtains

Ee =
[ 12 ( q
+ E1)
+ E2 + 11
Uniaxial strain

X [f2E2 (:: ~:) + flEl J . . . . (6)

2 Schematic representation of modified rule of
mixture formulation of uniaxial elastoplastic
deformation of metal-ceramic composite
For example, a value of q = 4500 MPa has been used
Published by Maney Publishing (c) IOM Communications Ltd

to model load transfer between the ferritic and mar-

fractions, the above formulations becomes complete.
tensitic phases in a dual phase stee1.49 The specific
For the case of a graded composite, additional inform-
choice of the empirical parameter q, however, widely
depends on the internal constraints which arise from ation regarding the variation of 11 or 12 with position
needs to be provided. Detailed discussions of plasticity
the particular spatial distribution of the phases in the
models for two phase materials can be found in
composite, the thermomechanical properties of the
Refs. 48-50.
two phases, the residual stresses induced during pro-
In addition to the foregoing approximations, the
cessing, and the connectivity of the phases which is
shear lag mode151 (which focuses on the transfer
sometimes defined as48
of tensile stresses between the phases via interfacial
No. of two phase shear processes) or one of its many adapt-
boundaries intersected ations36,37,39,4o,52,53 is commonly used to simulate the
Connectivity = N ft h . . (7) overall composite deformation in matrixes reinforced
0.0 wo p ase with aligned short fibres or whiskers.
boundaries analysed
The modified rule of mixture formulation has also Mean field theories
been extended to model the elastoplastic deformation The chief objective of the mean field theories for
of two phase composites (e.g. Refs. 15, 16, 18-20). composites is to assess the overall properties, such as
Figure 2 schematically shows the uniaxial elastoplas- the effective stiffness tensor E*, the effective com-
tic deformation of a metal-ceramic composite. Similar pliance tensor C*, and the effective thermal expansion
to the case of the elastic behaviour, equations (4)-( 6), coefficient (CTE) tensor a*, in terms of the corres-
the effective plastic response of the composite is ponding properties and concentrations of the consti-
assessed by invoking the parameter q in the following tutent phases.37,38,53-56 For this purpose, the phase
manner. The overall flow strength of the composite averaged stress and strain fields are employed.
corresponding to the onset of plastic yielding ye is (J
Consider first the case of linear thermoelastic
given by deformation. The overall stress-strain response can
be written as
lTyc=lTY2[f2+(::~:)~:flJ ..... (8) €* =c*a* + a*I1T
. . . . . . . . (10)
when the metal (phase 2) is continuous with yield a* =E*€* - E*a*I1T
strength (Jy2, and 11 = 1 - 12. The strain hardening of where the second term on the right hand side rep-
the composite, described by the uniaxial tangent resents the thermal strain or stress tensor due a
modulus, is uniform temperature change 11T with reference to
some initial stress free temperature. The symbol ''''
denoting a volume average is defined as

. . . . (9)
15 = ~
V Jvr p dV . . . . . . . . . . . (11)

where p is any spatially varying field quantity such

where H 2 is the instantaneous value of the elastoplas- as stress or strain, and V a representative control
tic tangent modulus of the metallic phase, as shown volume of the composite.
in Fig. 2. With a knowledge of the constitutive The total averaged stresses and strains within the
response of the individual phases and their volume two phases (i.e. the matrix phase, m, and the inclusion
International Materials Reviews 1997 Vol. 42 No.3
Suresh and Mortensen Functionally graded metals: Part 2 89

T I I I I 11
, I I , " I I I
T Tll..~JII-lr
.J.o'" """'looool
..........-1 11 ~,
'I II 11~
Q r I

I .-....rII...I
(a )
\ + / / (d)

, ..,
~, -'-- I Til

• -I --
-~-- t .•. /
", '"
t \

I I T1 .,. '-
\ £=0
~ --.
E= E I
1/ (c )
" .•. -•.
;;iii ". "
Transformation strain

(b) ( e)
Published by Maney Publishing (c) IOM Communications Ltd

3 Eshelby method61
for estimating stress state in metal matrix reinforced with misfitting ellipsoidal
inclusion (After Ref. 37)

phase, i) of the composite can be defined as where I is the identity tensor. Expressions similar to

emE' =~i m Vm
for the matrix . . . . (12)
equation (17) can be obtained for the volume aver-
aged stresses in each phase by recourse to the so
called stress concentration tensors,45.54 The stress and
strain concentration tensors and their components
=~rV JVm
udV generally depend on the geometry and spatial distri-
bution of each phase.57 Expressions which link the
concentration tensors to the elastic stiffness and com-
and pliance tensors and the tensor of CTE have been
derived58-60 by invoking the principle of virtual work.
1 K
1 Vi
edT/:' For the particular case of an ellipsoidal inclusion
of arbitrary aspect ratio in an infinite continuous
for the inclusion . . . . (13)
matrix, Eshelby61 showed that the stresses are uniform
a. =~
1 Vi JV
r adV throughout the inclusion. Consider that an ellipsoid
i shaped elastic inclusion, made from the same material
Here Vm and Vi are the volumes of the matrix and as the surrounding infinite matrix (i.e. taken out from
inclusion phase, respectively, and V = Vm + Vi. The an ellipsoidal region of an infinite matrix), is permitted
total averaged thermoelastic stresses in the matrix to undergo an unconstrained (i.e. stress free) trans-
and the inclusion can then be written as formation strain eT (i.e. shape change or eigenstrain,
such as that produced by pure thermal expansion or
am=Emem-Emam~1; ai=Eiei-Eiai~T (14) martensitic transformations which involve no changes
If the composite is subjected to a homogeneous in elastic constants of the inclusion and which may
applied stress aapp, the resultant overall thermoelastic be anisotropic), as shown in Fig. 3. Since the stress
strain is given by state in the constrained inclusion is uniform, Ui = G'i.
The inclusion is now forced to assume its original
e* = fiei + fmem = e app + a*~ T = c*aapp
. . . . . . . . (15)
+ a*~T shape by the imposition of surface tractions (Fig. 3c)
and forced back to fill exactly the ellipsoidal cavity
in the matrix. Once placed back in its original position
where fi and fm (= 1 - fi) are the volume fractions of
inside the matrix, the interfaces between the matrix
the inclusion and the matrix, respectively. The overall
and the inclusion are such that no interfacial sliding
elastic stress is given by
occurs, and the surface tractions constraining the
. (16) inclusion to its original shape are released. As equi-
librium is attained between the inclusion and the
The total phase averaged strains and stresses are then matrix, the inclusion develops a constrained strain eC
related to the overall strains and stresses via the strain relative to its initial pretransformation shape. The
concentration tensors A and a, which are defined as constrained strain is related to the stress free trans-
em = Ame* + am~ 1; and ei = Ai e* + ai ~ T ( 17) formation strain by the so called Eshelby tensor, S

where the subscripts i and ,n denote inclusion and . (19)

matrix, respectively. Note that
where the subscript In refers to the matrix. Since
fiAi + fmAm = I; and fiai + imam = 0 . . (18) (eC - eT) is the uniform elastic strain in the inclusion

International Materials Reviews 1997 Vol. 42 NO.3

90 Suresh and Mortensen Functionally graded metals: Part 2

which causes internal stresses, the stress within the Pedersen56 has proposed that the average matrix
inclusion can be obtained by Hooke's law to be stress due to the inclusions be treated as if it were an
externally applied stress, an approach which is
O'i = Ei(e C
- eT) . . • • (20)
expected to give lower bound estimates for composite
A particularly useful feature of the Eshelby tensor S properties. Similar methods which estimate directly
is that it enables the determination of the uniform the elasticity tensor of the composite have been
stress and strain in the inclusion, and obviates the formulated by Tandon and Weng.66 Wakashima
need for a full knowledge of the complex matrix stress et al.67 have proposed an approach for computing
field.37-39,61,62If the matrix elastic deformation exhib- the image stresses by employing the Eshelby method.
its isotropic material symmetry, Sm depends only on They postulate that, since the volume average total
the Poisson ratio of the matrix and on the aspect matrix stress am is essentially the same as the applied
ratio of the ellipsoidal inclusion. stress aapp in the dilute case, the non-dilute concen-
Now consider the case of an 'inhomogeneous tration effects be accounted for by replacing the
inclusion', i.e. one which is made of a material different applied stress with the volume averaged total matrix
from the matrix. In this case, the transformation stress. This method has also been extended to graded
strain eT in the inhomogeneous inclusion can be metal-ceramic composites.68
modelled by considering a homogeneous 'equivalent' During plastic flow, the back stresses may arise
or 'reference' inclusion (with the same elastic prop- from the pile up of dislocations (such as the Orowan
erties as the matrix) which is permitted to undergo a loops) around brittle inclusions in a ductile matrix
suitable equivalent eigenstrain q in such a way that e: which may oppose continued forward deformation
Published by Maney Publishing (c) IOM Communications Ltd

the inhomogeneous inclusion. and the equivalent while aiding reverse deformation (this can be viewed
inclusion attain the same uniform stress state as one of the mechanistic origins of the Bauschinger
O'i = Ei(eC
- eT
) = Em(e C
- e:q) • • (21)
effect).63,69 The magnitude of the Bauschinger effect
scales with the area fraction of the Orowan loops or
Since equation (19) applies equally to both a homo- the unsheared area of the glide plane. The dislocation
geneous and an inhomogeneous inclusion (with eT or loops can also promote additional inhomogeneous
e: q as the transformation strain, respectively), equa- stresses in the matrix since the presence of one loop
around a particle may repel successive loops which
tion (21) can be written as
are being formed around the same particle and also
O'i = Ei(Sme:q - eT) = Em(Sm - I)e:q . . (22) the effective spacing between the particles and hence
This equation can be solved to obtain the equivalent the matrix volume undergoing plastic flow are
transformation strain as a function of the stress free reduced. This latter term is commonly referred to as
transformation strain, the elastic constants for the the 'source shortening stress',56,69-73 and has been
matrix and the inclusion, and the tensor Sm. used to explain the strengthening and uniaxial ten-
When the concentrations of both phases of the sion--eompression asymmetry of metal matrix com-
composite are non-dilute, interactions of the fields posites reinforced with ceramic particles, whiskers, or
from other inclusions will be expected to influence continuous fibres. A variety of energy minimisation
the evolution of the average fields in the matrix and schemes and relaxation micromechanisms are also
the reinforcement. One way to account for such effects incorporated in such models to account for the onset
is to estimate an average matrix stress arising from and spread of inelastic deformation.37,38
the back stresses or image stresses from all the other A commonly discussed iterative method for estimat-
reinforcements, and to incorporate this back stress in ing the effective properties of a composite with a non-
the computation of the fields within and around a dilute concentration of a second phase is to envision
particular inc1usion.63,64 the composite volume as comprising an inclusion that
Irrespective of the particular mechanistic origins of is surrounded by an average medium whose effective
the average matrix stress, approaches which employ properties are the properties of the composite. These
the image stress or mean stress concept are commonly composite properties are not known a priori, but are
referred to as mean field theories, the Mori- Tanaka solved iteratively to obtain a self-consistent result. In
approaches, or the 'equivalent inclusion-average the generalised self-consistent schemes, the inclusion
stress' methods. For isothermal stressing of a matrix is embedded within a matrix material which, in turn,
reinforced with inclusions65 is surrounded by an effective composite medium. A
number of unit cell based analytical and semi-analyti-
cal models, wherein continuous fibres are periodically
arranged in a matrix, have also been considered by
Bapp = Bi [( 1 - h)I + h(AddiluteJ [(AddiluteJ-1
different researchers.74,75
where the subscript 'dilute' refers to the value of the
subscripted quantity for the dilute concentration of
the inclusion. From equation (23a), it is readily shown Computational models
that All of the foregoing analyses of two phase composites,
with very rare exceptions, do not accurately capture
(Adnon-dilute = (Ai)dilute [( 1 - fdI + 11 (AddiluteJ-1 the geometrical effects of the shape and spatial distri-
. . (23b)
bution of one phase in the matrix of another phase.
A number of different variations of the mean field This limitation raises certain key issues.
theory have been proposed for non-dilute concen- 1. As noted above, the Eshelby method61 provides
trations of a second phase in a matrix. For example, elegant analytical tools for quantifying the stresses in
International Materials Reviews 1997 Vol. 42 NO.3
Suresh and Mortensen Functionally graded metals: Part 2 91

an elastic two phase composite reinforced with ellip- (a)

soidal inclusions of arbitrary aspect ratio. However, 6 0.3
when the inclusions are of arbitrary shape with sharp
corners (such as in whiskers), precipitous gradients in
the fields emerge in the vicinity of the sharp corners,
and complete solutions can only be gathered by L

4 10
recourse to computational models, such as the finite
element method. b 0.2
2. As noted in a number of studies,76-s6 composite ~
models which invoke assumptions of spatially uniform
Ib 3
phase dispersions and idealised shapes (such as
spheres, ellipsoids, or cylinders) for one phase in the 2 0.5
matrix of the other may, in some cases, provide I
erroneous predictions of the effective elastic moduli I
for most real composites where the shapes and spatial o 0.1 0.2 0.3 0.4
distributions of the phases are seldom uniform. Inclusion volume fraction (fj)
3. For some reinforcement shapes and large
reinforcement concentrations, averaging schemes38,87 (b)
(such as those based on the Eshelby method)
may fall outside the elastic bounds of Hashin
Published by Maney Publishing (c) IOM Communications Ltd

and Shtrikman.88 The self-consistent method esti- Packet morphology

mates89,90 may also deviate significantly from finite
element predictions81 at large concentrations. 4

Detailed micromechanical analyses81,82 have revealed
that the dependence of the effective elastic moduli of
two phase composites on the shape of the reinforcing b
phase can be traced to the differences in the extent of ~
load transfer between the phases. Ib 3 Aligned
4. The effective CTE of a two phase composite discs
depends strongly on the overall bulk modulus of the
composite. For composites exhibiting macroscopically
isotropic elastic behaviour, the effective CTE is of
the form
(1/Q) - (1/Q2)
0: = 0:2 + (a1 - 0:2) (1/Qd _ (1/Q2) . (24)
2 +
where a and Q = E/[3( 1- 2v)] are the CTE and bulk orientated
modulus, and the subscripts 1 and 2 refer to the discs
ductile and brittle phases, respectively. The variables 0.1 0.2 0.3 0.4
without the subscripts denote the effective properties Inclusion volume fraction (fj)
of the composite. Where E is strongly influenced by
the spatial dispersion of the constituent phases as a 4 Normalised limiting value of overall composite
consequence of differing levels of constrained shear- tensile strength plotted as function of volume
ing, Q is essentially insensitive to phase distribution fraction of a aligned ellipsoidal inclusions and
effects. As a result, from equation (24), a is also b aligned and randomly oriented discs and
unaffected by spatial distribution (see, for example, needles (After Ref. 78)
Refs. 87 and 91). This result has important impli-
cations for the evolution of thermal stresses in graded
composites. rounding matrix within the cell are considered to
5. The effects of particle shape and spatial distri- deform identically to every other particle and matrix
bution on composite deformation become more pro- neighbourhood in the composite, and the particles
nounced when constrained plastic deformation, are assumed to populate the composite in a perfectly
accompanied by high triaxial stress and plastic strain periodic arrangement.
gradients, occurs in one of the two phases.76,78,82,83 One study of this type was published by Bao
To address these issues, a number of studies have et al.,78 who analysed the effects of particle morph-
dealt with detailed finite element simulations of the ology and orientation on composite elastoplastic
micromechanics of deformation within the context of deformation. Figure 4a shows the strengthening ratio
unit cell models.76-86 In these models, the concen- iiy/(Jy for an elastic-perfectly plastic matrix reinforced
tration of the reinforcing phase in the cell is the same with ellipsoid shaped isotropic elastic inclusions plot-
as that in the composite, and all the reinforcement ted against their volume fraction Ii for different ratios
particles are assumed to have the same shape, orien- of the particle aspect ratio atfaL, which was taken to
tation, and size. The unit cell comprising the particle be the same as the cell aspect ratio AIfAL. The matrix
in the ductile matrix is generally approximated by an is assumed to conform to a Mises type yield behaviour
axisymmetric geometry with periodic boundary con- whereby the effective stress (Je = [(3sijSij)/2] 1/2= (Jy,
ditions. Furthermore, the reinforcement and its sur- where sij are the deviatoric stress components and
International Materials Reviews 1997 Vol. 42 No.3
92 Suresh and Mortensen Functionally graded metals: Part 2

(a) ~.,

/'" Square
0120 ~/
a. / ..,.,-..,.,-


~ 80 /,
,.,. ", ... ......-- --- ---- --_.-.---

,.( /"/:,.....
~ ...
... --"'~
...""-Square diagonal
~ 40
Transverse tension

o 0.2 0.4 0.6 0.8 1.0
Tensi Ie strai n, 0/0

Sa transverse tensile deformation of 6061-0 AI-Zn-Mg alloy reinforced with 46 vol.-o/o of boron fibres
that are distributed in different periodic or random arrangements and b finite element unit cell showing
Published by Maney Publishing (c) IOM Communications Ltd

random arrangement of 60 fibres in matrix (After Ref. 94)

(Jy is the tensile yield strength of the matrix. iiy is the only recourse for quantitative modelling of the com-
limiting value of overall tensile yield stress ii of the posite micromechanical behaviour.
composite. This overall limit yield stress is independ- Finite element models of the dependence of spatial
ent of the elastic properties of the reinforcement and distribution of the two phases on the overall inelastic
the matrix, and is not influenced by the size of the response have been studied by a number of research-
reinforcement. Figure 4a shows that aligned elongated ers.76-83,92-100 The most systematic among such simu-
(atfaL> 1) or disc shaped (atfaL < 1) inclusions impart lations involve the effects of phase geometry on the
significantly more strengthening than spheres (adaL = transverse tensile deformation of metal matrix com-
1). Figure 4b is a plot of iiy/(Jy versus fi for randomly posites reinforced with a high concentration of uni-
oriented elongated ellipsoidal needles (atfaL = 0'1) and directionally reinforced fibres.83,93,94 Figure 5a shows
discs or platelets (atfaL = 10) reinforcing an elastic- the transverse tensile stress-strain curve for a 6061-0
perfectly plastic matrix. (Here the grainlike packets Al-Zn-Mg alloy reinforced with 46 vol.-% of boron
are oriented in such a way that the overall composite fibres. When the fibres are packed in a square array
behaviour is isotropic. These calculations of Bao and transverse tensile loading is applied along the
et al.78 involved (a) a three-dimensional cell model to edge of the square (termed 'square edge packing'
extract the multi axial limit yield surface of the grain- arrangement), the composite exhibits the greatest
like packets and (b) averaging over all orientations resistance to both elastic and plastic deformation.
with respect to the tensile axis, by recourse to a When the applied tensile loading is along the diagonal
Bishop-Hill procedure, to obtain an upper bound for of the square (termed 'square diagonal packing'
iiy•92) Also superimposed in Fig. 4b are the results for arrangement), the most compliant response is seen.
aligned inclusions of the same shape from Fig.4a. For fibres packed in a triangular or hexagonal per-
The randomly oriented particles, as anticipated, are iodic array, an inbetween stress-strain curve results
not as effective as strengtheners as the aligned ones. between the two bounds for the square packing case.
While the results shown in Fig. 4 exhibit macro- Figure 5b shows a random arrangement of 60 fibres
scopic trends, the micromechanics of plastic deforma- packed in a unit cell, which is repeated in the finite
tion in the metallic phase is subject to further element model with periodic boundary conditions
complexities. When one of the two phases deforms involving a generalised plane strain model. For this
plastically due to thermal and/or mechanical loading, geometrical arrangement of 'random fibres' (which
the local deformation undergoes severe non-pro- can be created by imaging and then discretising a
portional loading and becomes highly inhomo- typical microstructure in the real composite), the
geneous with steep gradients in plastic strains and stress-strain curve falls inbetween the bounds pre-
hydrostatic stresses, especially in the vicinity of sharp dicted by the square edge and square diagonal pack-
corners.76 The overall deformation of the composite ing models in Fig.5a. Note that the random fibre
then becomes increasingly more sensitive to the load- model predictions are noticeably different from those
ing path and to the shape and spatial distribution of of the periodic hexagonal array predictions (which
the phases. The onset and spread of local failure produces a transversely isotropic response during
processes are also influenced markedly by these fac- elastic deformation only). Brockenbrough et al. 83 and
tors. As a result, analytical methods become increas- Nakamura and Suresh94 showed that the effects of
ingly incapable of tracking the evolution of inelastic fibre distribution (for a fixed volume fraction) on the
deformation, damage, and failure in the composite, plastic response of the composite directly correlated
and computational models, albeit time consuming with the extent of hydrostatic stresses and constrained
and numerically cumbersome, appear to provide the plastic flow developed in the matrix. Consequently,
International Materials Reviews 1997 Vol. 42 NO.3
Suresh and Mortensen Functionally graded metals: Part 2 93

shear bands can be seen in the OFE Cu matrix

reinforced with 30 vo1.- % of unidirectional W fibres.
Figure 6b is an optical micrograph showing partial
recrystallisation of the deformed region around the
fibres after annealing at 300°C for 30 min.
Since the evolution of matrix stress and stress fields,
as well as the onset of shear localisation (such as the
one shown in Fig. 6) are strongly influenced by the
geometry, concentration, and spatial arrangement of
one phase in another, the evolution of damage and
failure is also expected to be highly sensitive to these
factors. In addition, experimental results by Embury
and co-workersl02 and Lewandowski and co-
workersl03 have documented how externally im-
posed constraints, in addition to local microstructural
constraints, influence the deformation and failure of
two phase composites. Extensive studies of failure in
metal-ceramic composites have identified a variety
of conditions which govern the following princi-
pal mechanisms: (a) fracture of the brittle phase,
Published by Maney Publishing (c) IOM Communications Ltd

(b) ductile failure by the nucleation, growth, and

coalescence of voids within the metallic phase,
and (c) delamination and crack growth along the
interface between the brittle and the ductile phases.
Detailed reviews of the mechanics and micromech-
anisms of these failure processes can be found
elsewhere. 37,39,41,79,82,86,104

Issues of size scales and discrete

microstructural deformation processes
The approaches outlined in the preceding sections on
the mechanics and micro mechanics of deformation in
metal-ceramic composites are predicated on the
premise that the size of each phase, i.e. the size of the
reinforcement and the size of the matrix region sur-
rounding it, are large enough to be amenable to be
modelled by continuum mechanics. That is, these
dimensions are large compared to the characteristic
microstructural size scales, such as the grain size or
b the dislocation cell size in the ductile phase. This
6 a backscattered electron micrograph of requirement may not be satisfied when the volume
deformed grid, etched on specimen surface by fraction of one phase in the matrix of the other is
electron beam lithography, showing presence large. Furthermore, none of the continuum models
of shear bands in OFE Cu matrix reinforced with discussed above can account for the important role
30 vol.·% of unidirectional W fibres and b optical of size scale in influencing the deformation and failure
micrograph showing partial recrystallisation of composites because the constitutive model for the
of deformed region around fibres after anneal· matrix material or the reinforcement has no length
ing at 300°C for 30 min (Reproduced with scale built into it. For example, it is known from
permission from Ref. 101)
experiments that the strengthening of particle
reinforced composites and the propensity for the
the order in which the constraint increases in the fracture of the brittle reinforcement can be strongly
matrix corresponds to the order in which different influenced by the size of the reinforcing particle; these
spatial arrangements of fibres increase the overall are features which are completely left out of the scope
deformation resistance of the composite. Systematic of continuum formulations. There have been recent
experimental measurements of localised strain distri- developments in the modelling of strain gradient
bution in plane strain compression (derived from plasticitylOS which, seek to address such size effects.
changes in the shapes of fiducial grids), matrix texture In addition, finite element models which examine the
development, recrystallisation behaviour, and damage details of micro mechanical failure processes, such as
evolution in the ductile matrix, for hexagonal and brittle reinforcement fracture,86 or ductile failure by
square periodic arrangements of tungsten fibres in a void growth in the metallic matrix,79 or along the
copper matrix have been reported by Poole et al.lOO,lOl interface,86,104 appear to capture some aspects of
Figure 6a is a backscattered electron micrograph of the effects of reinforcement size on the overall
a deformed grid, etched on the specimen surface by deformation and failure of the composite. However,
electron beam lithography, where the presence of general methods for the quantification of composite

International Materials Reviews 1997 Vol. 42 NO.3

94 Suresh and Mortensen Functionally graded metals: Part 2

strengthening in plastically deforming matrixes into plastically deforming matrixes from the ends of
remain far from complete at this time. reinforcing particles has been documented during
The analysis techniques discussed in the foregoing thermomechanical loading, using the photoplastic
sections were also predicated on isotropic elastic or Agel as a model matrix material and SiC or A1203
inelastic constitutive models. These approaches do as reinforcements.118,119 These observations illustrate
not allow for the discreteness of crystallographic slip the significant role of discrete dislocations and loops
in the metallic grains of the composite, and the in not only relieving misfit strains during thermome-
ensuing anisotropy of microscopic deformation. This chanical loading, but also in influencing subsequent
limitation can at least partly be overcome by resorting deformation.
to continuum formulations of crystal plasticity at the
single grain level. Such models, which build on the
works of Taylor106 and Hill and Rice,107 have been
Some critical issues in modelling of
developed into a framework amenable for detailed graded composites
computational simulations.los-115 Modelling of the Modelling of the elastoplastic deformation character-
plastic deformation can be accomplished to account istics of compositionally graded composite micro-
for both the discreteness of straining by slip and the structures requires the incorporation of additional
randomness of grain orientations in a microstructural factors and length scales into the analyses, over and
ensemble, by allowing for the shearing along crystallo- above the aforementioned factors for two phase com-
graphic planes in preferential slip directions and posites without gradients in phase mix, in view of the
summing the plastic shear flow over all active slip following considerations:
Published by Maney Publishing (c) IOM Communications Ltd

systems in different crystal structures. It has been 1. The existence of a gradient in phase mix intro-
found that while the predictions of the crystal plas- duces a new length scale to the problem, i.e. the
ticity models for the deformation of metal-ceramic variation of the relative concentrations of the two
composites qualitatively mirror the trends derived phases with distance. In addition to the phase shape,
from continuum Mises type (J 2 flow theory) calcu- connectivity, and dispersion, the gradient in compos-
lations, the discreteness of the slip process has been ition generates additional geometrical effects in the
found to identify more pronounced shear localisation deformation of a composite.
effects which can have a major influence on failure 2. In a unidirectionally graded composite layer
processes. The crystal plasticity theories, however, are sandwiching two homogeneous materials, discrete
also incapable of predicting the effects of phase dimen- sublayers can be found within which the composite
sions (such as the size of the reinforcement particle) has essentially uniform concentration of the two
on deformation. A more detailed discussion of con- phases. For example, if a graded metal-ceramic com-
tinuum crystal plasticity theories, as they specifically posite layer is synthesised by sintering techniques by
apply to graded metal-ceramic composites, is pre- stacking sublayers of different relative proportions of
sented in the section 'Crystal plasticity models' below. the two materials, the thickness of each sublayer
The discreteness of microscopic deformation can introduces another microstructural length scale. In
be considered at yet smaller length scales by invoking order for continuum analyses to be valid, the thickness
the concept of discrete dislocations within the ductile of each such sublayer must be significantly greater
phase. Such dislocation based plasticity models have than the characteristic microstructural unit size, such
a built in characteristic length scale, which is the as a grain size or a particle size. When the graded
Burgers vector. Cleveringa et al.116 have presented an composite layer is discretised into sublayers for com-
analysis of plastic flow arising directly from the putational modelling, each sublayer should comprise
collective motion of a large number of dislocations in a sufficiently large number of elements for numerical
a composite material subject to simple shear. A full convergence and accuracy, and must be sufficiently
boundary value problem is considered with a two- larger than the microstructural unit size. At the same
dimensional array of line defects in an isotropic linear time, the size of each unit must be small enough to
elastic solid. Deriving from earlier studies117 (see capture the local fluctuations in stresses between
Ref. 116 for additional references on the subject of sublayers. This issue is illustrated with examples in
dislocation modelling of deformation), the stresses the section 'Experimental measurement of stresses
and strains are cast as the superpositions of the fields and deformation in graded multilayers' below.
from discrete dislocations and the complementary 3. In a metal-ceramic composite graded continu-
fields which account for the interactions with hard ously or discretely from an all metal layer to an all
inclusions and enforce the boundary conditions. ceramic layer, the concentration changes from a dilute
Continuum elasticity fields describe the long range mix of a brittle phase in a ductile matrix at one end,
interactions between dislocations, and an additional to a high fraction of a brittle (ductile) phase in a
set of constitutive rules account for dislocation drag, ductile (brittle) matrix near the middle, to a dilute
interactions with obstacles, as well as dislocation mix of a ductile phase in a brittle matrix at the other
nucleation and annihilation. It is shown that such an end. The phase shape, connectivity, and dispersion
approach would predict the dependence of the local may also change continuously through the thickness
stress and deformation fields in the metallic phase of of the graded layer.
the composite on the size of the brittle phase, as well 4. Differential shrinkage during sintering and poss-
as on the spatial dispersion and geometry of the ible differences in the evolution of pores and other
two phases. defects across the thickness of the compositionally
This section is concluded by noting that direct graded layer engender residual strains which can have
experimental evidence for the punching of dislocations a marked effect on subsequent thermomechanical

International Materials Reviews 1997 Vol. 42 NO.3

Suresh and Mortensen Functionally graded metals: Part 2 95

performance. These effects need to be adequately

captured in the models for graded materials. Some
techniques to alleviate these phenomena were
addressed in Part 1 of this series.1

Layerl, 2El,vpal
5. Since the constraint due to the brittle phase on
Graded layer (GL) 20
the ductile matrix changes with distance, constitutive
models for deformation and failure must be designed y hi Layer 2, E2, v2' a2
to capture the differing constrained plastic flow and
damage evolution in the different regions within the .• Lx

graded layer. That is, the pressure sensitivity of

7 Schematic of layered plate with graded
deformation, damage evolution, and fracture in each interlayer and associated nomenclature
of the constituent phases must be known.

difference between the different layers and that

Thermoelastic deformation of
between the phases within the graded layer causes a
graded multilayers variation in thermal stress to evolve along the thick-
The most convenient analytical and computational ness of the plate. The plate begins to curve because
models of functionally graded materials, which are of the through-thickness strain gradient, thereby
representative of many practical applications, involve accommodating the thermal stresses. When the geo-
formulations within the framework of small deforma- metrical conditions of the plate (which is isotropic
Published by Maney Publishing (c) IOM Communications Ltd

tion, small strain plate/beam theories of classical in-plane) are such that the strain is allowed to be a
continuum mechanics (e.g. Ref. 120). These formu- function of z only, the small strain compatibility
lations, which ignore dynamic effects, thermal gradi- equations lead to the result that the in-plane normal
ents, and stress relaxation mechanisms, provide exact strain C = Cxx (= Cyy for the biaxial stress state in
results within the realm of small strain analyses in Fig. 7) is linear; it can be expressed as
regions away from free edges where the multiaxial
fields due to edge effects and singularities need to be c(z) = Cxx = Co + KZ . • . . •. .• (25)
examined. For the purpose of illustration, consider where Co is the normal strain at z = 0 and K the
the layered structure schematically shown in Fig. 7 curvature (i.e. the inverse of the radius of curvature)
which comprises two homogeneous layers, 1 and 2, of the plate in its plane. The only non-zero stress is
between which a graded layer of thickness 2a is O"(z) = O"xx(z) = O"yy(z), which is given, for the equal
placed. The in-plane shape of this layered solid is a biaxial stress state, by
rectangle whose length and width are Lx and Ly along
the x and y axes, respectively. The plate is assumed E(z)
O"(z) = 1 _ v(z) [c(z) - ex(z)~ T(z)]
to have a uniform thickness everywhere, and for the
purpose of simple analytical formulation, the entire
multilayer is assumed to have a uniform temperature = 1 _ v(z) [co + KZ - ex(z)~ T(z)] . . . (26)
everywhere at all times during thermal excursions
(although simple heat conduction across the layers
where ex is the thermal expansion coefficient, and ~ T
can easily be modelled on the basis of the present
represents the change in temperature from the initial
formulation using a personal computer). The
stress free state, and in general, they are all functions
geometry of the layered structure shown in Fig. 7 is
of z. (For simplicity of illustration, isothermal con-
such that the problem can be treated as one-dimen-
ditions are assumed in the following discussion, i.e.
sional. When the geometry of the layers is such that
~ T is independent of z.) In the general case involving
Ly» (h1 + h2»> Lx in Fig. 7, plane stress beam con- externally imposed thermal and/or mechanical loads,
ditions prevail, and the parameters of interest for
the resultant force and the resultant moment of the
thermomechanical performance depend only on the
stress distribution O"(z) along the height z from equa-
vertical position z. The general three-dimensional
tion (26) must be equal to the applied axial
stress state, which prevails over a distance from the force Fap and the applied bending moment Map,
edge which is approximately equal to the layer thick-
ness, alters the plane stress results only near the edges.
The analyses for plane stress can easily be extended
to the more realistic situation involving thermoelastic
response in the equibiaxial stress state (corresponding
to the geometrical condition of a plate that
fh: +a a(z) dz

+ t:a

a(z) dz

Ly »(h1 + h2) and that Lx »(h1 + h2) by simply replac- +

J O"(z) dz = Fap . . . . . . . (27)

ing Young's modulus E by the corresponding biaxial
modulus E = E/( 1 - v) where v is the Poisson ratio.
Consider an initially perfectly planar uncon-
strained, layered and/or graded plate with a graded
t~:+a a(z)z dz + a(z)z dz

interlayer, as shown in Fig. 7, which is subjected to a 1

uniform thermal excursion ~ T from some stress free + Ja+h O"(z)z dz = Map . . . . . . (28)
reference temperature (such as the processing temper-
ature, diffusion bonding temperature, or softening These two conditions, i.e. force balance and moment
temperature). The thermal expansion or contraction balance, lead to a linear system of equations in Co

International Materials Reviews 1997 Vol. 42 NO.3

96 Suresh and Mortensen Functionally graded metals: Part 2

and K. The solutions,17,18,23 in the absence of buckling,

are given by z/h

- 1 (J 0 + pap)
2 + 11 (J 1 + Map)
80 = -----------
Ii - 1012
I1(JO + pap) - IO(J1 + Map) h
K = -------- .... (29)
Ii - 1012 :{ a

Ii = fh 2

ziE(z) dz, i = 0, 1, 2


Ji = f:, Zi()(z)~T(z)E(z) dz, i = 0, 1 . . . (30)

- h1 and h2 are the extreme values of z (Fig. 7). Note

that no singular values for 80 and K can be encoun-
tered since Ii =I-1012. The value of Ii depends only on
the geometry and the profile (variation across the
8 Example of thermal stress distribution in
Published by Maney Publishing (c) IOM Communications Ltd

height h) of the Young's modulus. Additionally, if ~ T

Ni-GL-AI203 trilayer beam which is subject to a
is uniform and pure thermal loading is applied, then
temperature drop of 100 K from initial stress
Jo and J1 are proportional to ~T. This, together with free temperature
pap = 0 and Map = 0 for pure thermal loading, implies
that 80 and K are also proportional to ~ T. From the
above four ~quations, the thermal stress profile In equations (31) and (32), the symbol ~ preceding
through the thickness of the multilayered material a property refers to the change in that property for a
with a compositional gradient can be deduced. If E change in temperature ~T. From equations (25)-(30),
and a~ T vary continuously with z, then (J is contin- the strains, curvatures, and stresses in the multilayered
uous too. Furthermore, if E or a~ T is discontinuous, plate can be deduced. In an attempt to develop a
then (J is also discontinuous which is the case for a quantitative assessment of stress evolution, consider
sharp interface between a metal and a ceramic. The the model system of aNi-graded layer-Al203 trilay-
force and moment balance equations (27) and (28) ered system. The thermoelastic properties within the
can be incorporated into simple softwares amenable inbetween graded layer (GL) vary linearly with dis-
for use with a personal computer in such a way that tance z, according to equations (31) and (32). The
the general thermomechanical analysis of a multi- spatial variation of the elastic thermal stress (Jxx with
layered material with or without a gradient in com- distance z in the Ni-GL-AI203 system (in plane
position in one or more layers can be performed; in stress) subjected to a temperature reduction of 100 K
this way, the overall curvature of the plate, the stresses (from an initial stress free reference temperature) is
and the strains at any thickness location in any of plotted in Fig. 8 for the particular geometrical con-
the three layers can be computed numerically. Such dition that h1 = h2 = h and that a/h = 0·6. Also indi-
an approach121 is additionally capable of incorporat- cated in Fig. 8 are the corresponding thermal stress
ing non-uniform temperatures, in-plane orthotropy in profiles of a Ni-AI203 bilayer without a graded
elastic properties in one or more layers or graded interlayer (shown by the broken lines), for the same
sublayers, plasticity with isotropic hardening, steady temperature drop from the same initial stress free
state power law creep, and simple criteria for brittle temperature. A comparison of the behaviour of the
or ductile failure initiation at critical locations of the bilayer (with a sharp interface) and the graded trilayer
m ul tila yer. reveals the following trends: (1) when the interface
It is illustrative at this point to examine the impli- between the metal and the ceramic is abrupt, large
cations of compositional gradation, in the context of stresses develop in the interior of the layered solid;
equations (25)-(30). For this purpose, assume that (2) for cooling from the stress free temperature, the
the biaxial Young's modulus E and coefficient of near interface region of the metal develops a tensile
thermal expansion in the graded layer (- a < z < a) stress, while the near interface region in the ceramic
in Fig. 7 and isothermal conditions such that develops a compressive stress - there is an abrupt
change in the sign of the stress at the interface; (3) the
E=Eo-~E=Eb -h1~z~-a
magnitude of thermal stress in the interior can be
- - -z reduced and the abrupt change in sign of the stress
E = Eo + ~E -, -a ~ z~ a at the interface can be erased by introducing a graded
interlayer, as shown in Fig. 8; (4) the stress at the
E=Eo+~E=E2, a~z~h2 .... (31) reference plane (z = 0) is eliminated by the introduc-
a=aO-~a=a1, -h1~z~-a tion of the graded layer for the particular geometry
z and gradient profile shown in Fig. 8. (In general, the
a = ao + ~a - , - a~ z~ a interior location where the residual stress is made to
vanish can be tailored by the appropriate choice of
a = ao + ~a = a2, a~ z~ h2 . (32) the gradient in the metal-ceramic mix of the
International Materials Reviews 1997 Vol. 42 No.3
Suresh and Mortensen Functionally graded metals: Part 2 97

interlayer.16,18) Note in Fig. 8 that the stresses vary where C* is the global compliance tensor, the
linearly with distance within the metallic and ceramic subscript 0 denotes the matrix phase and the subscript
layers, and parabolically within the compositionally i denotes the inclusion phase which in the most
graded layer (when the elastic modulus and thermal general case is considered to be of an ellipsoidal
expansion coefficient vary linearly with distance). geometry of arbitrary major and minor axis ratio,
Table 1 provides an example calculation of the and orientation. Br denotes the phase concentration
elastic stresses in the Ni-GL-AI203 layered material factor tensor, and the components of Br are functions
for different values of Pa = a/h and a Ni-AI203 of the ellipsoidal parameters as well as of the elastic
bimaterial with a sharp interface (a/h = 0), subjected constants of the matrix and the inclusion. Porosities
to a temperature drop of 805 K from the initial stress in the matrix, whose concentration may vary as
free temperature.18 For small Pa, the largest stress in functions of position (as, for example, due to differen-
the Ni layer is at the free surface, and it occurs at the tial shrinkage), can be considered as inclusions with
alumina/FGM interface for larger values of Pa' The no stiffness. In view of the possibility that the matrix
residual stresses vanish as Pa ~ 1, as expected from and inclusion phases may not be interchangeable and
the theory, equations (25)-(30). Table 1 also illus- that there exists an uncertain or 'fuzzy' transition in
trates that by compositionally grading an interface microstructure between the two phases, the effective
between a metal and a ceramic, the abrupt transitions property within the graded layer (which comprises
in residual stress gradients seen at a sharp interface two materials involving different microstructural
are eliminated, and that the magnitude of the residual combinations i = 1 to I, where I is the total number
stresses is markedly reduced through the majority of of microstructures) has been taken to be123
Published by Maney Publishing (c) IOM Communications Ltd

the layer thickness.

While the foregoing analyses deal with continuum - Ibif:U~
p= b·f' . . (35)
descriptions of elastic deformation, it is worth noting 1 m

that micromechanical formulations of elastic or elas-

where ~ is effective property estimated from the mean
toplastic deformation predicated on the mean field
field approach, and bi a function (linear, quadratic,
theories discussed above have also been applied
etc.) which characterises the variation of P as a
to characterise the thermomechanical response of
function of z. Further details of these approaches can
graded multilayer plates. Hirano, Wakashima, and
co_workers68.122,123 have extended the mean field be found in Refs. 122-125.
formulations of Wakashima et al.67,124 to graded
metal-eeramic composites in the following manner. Onset and progression of plastic flow
The fractional volume of the constituent phases in in graded multilayers
the graded region is characterised by the distribution
function When plastic flow and/or other types of permanent
inelastic deformation commence in one or more layers

f:(z) = (1- fm: Ji) (1- fp); of the graded multilayer, the stresses begin to redis-
tribute in a manner that is different from the descrip-
tions provided in equations (25)-(30). A simple and
convenient way to estimate the temperature change
f:n = ( 1 - f!: Ji) (1 - fp) . (33) at which plasticity begins, (~T)pb is to equate the
von Mises effective stress (Je to (J(z) under plane stress
where fp is the volume fraction of the micropores in or equal biaxial stress states, and to invoke the
the graded region and z is the coordinate along the criterion that plastic yielding begins at a location
thickness of the graded region. The effective prop- where (Je first reaches the yield strength of the material
erties, such as the compliance tensor, in the graded (Jy.* Closed form solutions17,18,126 and results of com-

region are obtained by the following formulation putational simulationsI5.18.127,128 are available for the
plastic deformation of graded metal-eeramic compos-
(34) ites as a function of different layer geometries and
gradient profiles. Also known are the conditions
Table 1 Calculation of elastic stresses in which govern the location for the onset of plastic
Ni-GL-AI203 layered material for different flow in a general metal-eeramic graded multilayer,18
values of Pa = a/ hand Ni-AI203 bimaterial such as the one illustrated in Fig. 7.
with sharp interface (a/ h = 0), subjected General 'plasticity diagrams' which map the onset
to temperature drop of 805 K from the of yielding in the graded metal-eeramic layered solid
initial stress free temperature18 for any arbitrary combinations of the individual thick-
a/h = 0'1 a/h=0'4 a/h = 0'9-1'0
nesses of the homogeneous and graded layers can be
developed on the basis of the above simple criteria.
z FGM* Sit FGM* Sit FGM* Sit Figure 9 is a diagram for the Ni-GL-AI203 model
h -1'91 -1'80 -1'80 -1'80 ~O -1'80 system.18,129 In this diagram, the three independent
a 4·20 4'26 2'07 2·23 ~O -1'80 layer dimensions, i.e. a, hb and 112, can be represented
0+ 0'40 4'93 0'12 4·93 ~O 4'93
0- 0'40 -7'61 0'12 -7'61 ~O -7'61
-a -6'54 -6'44 -3'47 -2'80 ~O 4·37 * Under plane strain conditions, E or E should be replaced by
-h 4·24 4'37 3'44 4'37 ~O 4'37 E/(l- v2) and a by a(l + v). The von Mises criterion for plane
strain becomes
* Functionally graded.
t Sharp interface. . (36)

International Materials Reviews 1997 Vol. 42 NO.3

98 Suresh and Mortensen Functionally graded metals: Part 2

TvaroC commences at relatively low temperature changes

......, 1,
AI203 thin film on Ni ;---_.j 8 600 because of the sharp interface between the metal and
1.0 ..-r
7 300 the ceramic. It is also seen that when the graded
6 200 interlayer between the metal and the ceramic is thick,
5 160 i.e. for Pa > 0·8, plasticity is essentially suppressed for
0.8 4 140 the range of temperature changes considered here.
...--C\J 3 120 For Pa ~ 1, there is no residual stress in the entire
..c 2 100 graded layer as long as ad T varies linearly with z as
+ 0.6 1 50 anticipated from the small strain thermoelasticity
/p theory.130
..c Also plotted in Fig. 9 is the region, bounded by
II 0.4
the broken line and the line connecting the points
(Pa, Ph) = (0, 1) and (1, 0'5), within which plastic flow
initiates at the Ni free surface. Outside this region are
o hI geometrical combinations for which yielding initiates
<i: at the interface between the homogeneous Ni layer
and the graded layer.
0.0 0.2 0.4 0.6 0.8 1.0

Micromechanics of plastic
Published by Maney Publishing (c) IOM Communications Ltd

9 Contours of constant temperature change

necessary for onset of plastic flow in The continual spatial variations in the concentrations
Ni-GL-AI203 trilayer system for any arbitrary of the constitutent phases of a graded composite
combination of relative thicknesses of each of engender new 'length scales' and other complexities
layers; see text for details (After Ref. 18) in the computational modelling of the micromechan-
ics of deformation and failure. When the volume
fraction of one phase in the matrix of another is in
in terms of two non-dimensional geometric para- the range 30-700/0 and when there exists a transition
meters as, for example, Pa = 2a/(h1 + h2) and Ph = from a ceramic matrix to a metallic matrix, the simple
h2/(h1 + h2). Any two-dimensional 'performance dia- 'unit cell' computational models discussed above fail
gram' can then be drawn for a given graded material to provide accurate descriptions of the thermomech-
system on the basis of these two parameters for all anical behaviour without appropriate modifications
possible combinations of the thicknesses of each of and refinements. To overcome these limitations, newer
the three layers. The ordinate in this figure represents classes of computational models have been developed
different geometrical combinations of a Ni-AI203 for graded composites.131,132
bilayer with a sharp interface. The three coordinate
points represented by (Pa, Ph) = (0, 0), (0,1), and
(1,0'5) represent the three corners of a triangle within Mises type continuum plasticity models
which all thickness combinations of the three layers Figure lOa is a diagram which illustrates a planar
can be mapped. The first of these three corners beam geometry comprising a unit cell model (dark
denotes the location of a thin Ni film on a thick shaded region) for a graded multilayer.131 The
Al203 substrate, the second denotes a thin Al203 film deformed and undeformed configurations of the unit
on a thick Ni substrate, and the third, marked as cell are also indicated in that figure. For pure thermal
point P in Fig. 9, represents the condition where the loading of the graded material in response to an
entire material is a single compositionally graded increase or decrease in temperature by dT (without
layer extending from all Ni at one end to all Al203 any thermal gradient through the thickness), the
at the other. following boundary conditions are imposed
Figure 9 shows the contours of constant critical
temperature variation, (d T)pl, at which plasticity ux(O, y) = 0; ux(O, y) = 0 )
begins in a Ni-GL-AI203 trilayer (shown sche-
matically in Fig. 7) for all possible combinations of
uxO¥' y): uxCW; 0) + A[ux(W; P) = ux(W; 0)] (37)
the thicknesses of the Ni, A1203, and linearly graded uy(Jt: y) - uy(Jt: 0) + A[uy(Jt: P) uy(Jt: 0)]
layers. Here the elastic modulus and CTE variations
within the graded layer are assumed to follow equa- where ux(l, 2) and uy(l, 2) are the displacements in
tions (31) and (32). For the purpose of simple illus- the x and y directions, respectively, at locations (1, 2),
tration, plane stress conditions are assumed and the Wand P are the width and height, respectively, of
yield strength of Ni is taken to be 100 MPa (which is the unit cell, and A is a parameter, symbolising the
a typical value at 527°C). For a Ni thin film on an vertical position as a fraction of the total height,
Al203 substrate, i.e. at the origin of the figure, a which can vary between zero and unity. The displace-
temperature change (d T)pl = 57 K, is needed to yield ments uy(O, y) were left unconstrained. Physically,
the metal film fully. On the other hand, a bilayer equation (37) implies that free deformation allows
comprising a thin Al203 film on a thick Ni substrate the top and bottom sides of the unit cell to deform
does not lead to any plastic flow in the Ni, even for freely, the left side of the unit cell is precluded from
(d T)pl > 600 K. For the geometric conditions rep- any deformation in the horizontal direction, and the
resented by points along the ordinate, plastic flow right side of the unit cell is forced to remain as a
International Materials Reviews 1997 Vol. 42 NO.3
Suresh and Mortensen Functionally graded metals: Part 2 99

Deformed plate with unit cell

y. ...\
0.12 Al20J Unit cell
Graded \
~ 0.57 layer
Undeformed (solid)
\ Deformed (broken)
0.37 Ni \

... - X
Published by Maney Publishing (c) IOM Communications Ltd

(a) (b) (c) (d)

10 a unit cell and nomenclature of trilayered beam with graded interlayer for finite element simulation
of thermomechanical response, b hexagonal packing arrangement of graded metal-ceramic composite
layer; c and d examples of unit cells with random distributions of metal-ceramic phases in graded
layer (After Ref. 131)

straight line during all deformation though it can employed to analyse the thermomechanical response
rotate. The axial and bending strains, respectively, of the graded layer. The finite element unit cell
are defined as comprises a homogeneous metal layer, a homo-
geneous ceramic layer, and an intermediate metal-
1 ceramic composite layer within which the metal
ea = 2W [ux(Jt: 0) + ux(Jt: P)J
and ceramic concentrations are graded, from 10 to
90 vol.- 0/0, in nine steps corresponding to a linear
eb = W [ux(JtJt: 0) - ux(Jt: P)J . . . (38) gradation in composition. In the vicinity of equal
volume concentrations of the metal and the ceramic,
a switch from a metal matrix to a ceramic matrix (or
For deformation induced solely by thermal excur-
vice versa) is needed here for numerical accuracy, as
sions, the curvature of the plate K as well as the
these types of models. represent only a matrix-
instantaneous coefficients of thermal expansion for
inclusion structure.
the unit cell for axial loading and bending, Lia and Lib,
The geometrical arrangement of the phases in
respectively, are defined as*
Fig. lOb represents a highly idealised periodic struc-
eb dea deb ture, whereas real graded microstructures exhibit
K = P; Lia = d T; Lib = dT . . (39) phase dispersions which are significantly more irregu-
lar and non-periodic. These departures from ideal
The above arrangement can be analysed by arrangements can be captured, within the context of
employing a two-dimensional periodic unit cell for- two-dimensional numerical simulations, by consider-
mulation wherein each phase is envisioned as an ing random microstructural unit cells for finite
isolated particle in the matrix of the other phase, with element analyses which can be created by discretising
the volume fraction of the particle adjusted to reflect optical or scanning electron micrographs of typical
the appropriate spatial distribution of the concen- graded microstructures at the appropriate magnifi-
tration of that phase. Weissenbek et al.131 have pro- cation. Figure 10c and d shows two examples of
posed, for this purpose, a hexagonal packing random dispersions of a graded metal-eeramic com-
arrangement, whose finite element discretisation is posite which is sandwiched between the homogeneous
shown in Fig. lOb. The spacing between the inclusions metal and ceramic layers; here the dark grains rep-
is computed on the basis of the hexagonal packing resent the ceramic. The hexagonal packing arrange-
arrangement, since the gradations in composition do ment of Fig. lOb or the random packing arrangements
not allow symmetry in the unit cell along the thickness shown in Fig. 10c and d represent repeating periodic
of the graded layer. In the particular configuration unit cells that are subject to the boundary conditions
shown in Fig. lOb, nine different sublayers are in equation (37). The constitutive properties of the
two phases can then be introduced within the context
of continuum formulations, either by invoking the
* While lXa is based on the standard definition for linear CTE,
equation (39) shows a similar definition for the bending CTE. By von Mises plastic yield behaviour for the metal and
these definitions, the CTE values represent instantaneous values lXi isotropic elastic response for the ceramic,131 or by
at any given temperature, and not a secant or total value.131 employing single crystal plasticity models for the

International Materials Reviews 1997 Vol. 42 No.3

100 Suresh and Mortensen Functionally graded metals: Part 2



1.00 _.- Modified rule of mixture
e:1 -- Experiment 20- 350°C
--- Hexagonol
..•. Random arrangement (d)
- Rondom orrangement (c)

a 200 400 600 800

11 Comparison of model predictions of variation

of CTE in bending with temperature against ':., .,
experimental observations (After Refs. 19 and :.:


metallic grains and isotropic elastic response for the

Metal, vol.-%
ceramic grains. 132
Published by Maney Publishing (c) IOM Communications Ltd

The predictions of the foregoing finite element

models which employ a hexagonal unit cell (Fig. lOb)
and random grain dispersions (Fig. 10c and d) are
compared with experimental measurements and with
the predictions of the modified rule of mixture
approximation, equations (4)-(9), in Fig. 11 for the
t X2

Ni-AI203 graded composite. (For these particular

simulation and experiments, the relative thicknesses
::I!; ~ L-XI
of the Ni layer, graded layer, and Al203 layer were 12 a illustration of decomposition of total
0'12, 0'57, and 0'37, respectively.) Figure 11 shows the deformation gradient into thermal, plastic, and
variation of bending CTE, C(b, with temperature T. lattice components, and modelling of metal in
Also shown in this figure are the experimental values graded layer as c plastic single crystal rather
of C(b versus T for the Ni-AI203 graded trilayer than as b von Mises continuum (After Ref. 134)
obtained by Finot et al.19 for a temperature excursion
of 20 ~ 350°C. While the modified rule of mixture and lattice components, FT, FP, and F*, respectively,
matches the numerical predictions in the elastic as shown in Fig. 12. If u is the displacement vector
regime, it severely overestimates the temperature for and X is the material position vector with respect to
the onset of plasticity in Ni. The finite element models the undeformed reference state
estimate that the temperature for the initiation of au
plastic flow in Ni is around 150°C, which is in F=F* ·FT• FP = I +- . (40)
accordance with experimental observations. The ax
range of variation in C(b, arising from plastic flow in where I is the second order identity tensor. Plastic
the multilayer, over the temperature range 100-250°C, deformation in individual metal grains is modelled as
as well as the flatness of the initial portion of the simple shearing along crystallographic slip planes
C(b-T curve and the plateau region (from ~ 250 with ~nit normals mp and along slip directions with
to 300°C) after the initiation of plasticity, are also unit vectors sp, where the index f3 refers to the slip
predicted reasonably well by the computational system. The velocity gradients for the plastic and
simulations. 131 Note that finite element formula- thermal parts are given, respectively, by
tions, which incorporate adaptations131-133 of the
jp. (FP)-l = LYpspmp and iT. (FT)-l = 'Fa
Mori- Tanaka mean field method64 in an incremental p
form, have also been used to obtain computational . . .. . (41)
results similar to the ones shown in Fig. 11.
where Yp is the slip rate on the slip system f3 with the
summation done over all active slip systems, T is
Crystal plasticity models the temperature, and a = ~i ~j C(ijaiaj is the tensor of
While the foregoing analyses discretise the graded thermal expansion coefficients with components C(ij
layer into continuum sublayers of averaged properties, defined with respect to the time independent Cartesian
as in Fig. lOb, or into microstructural units wherein base vectors iii' In cubic crystals, the vectors iii are
the effective thermoelastoplastic properties are aligned with the cube axes. The slip rate y p generally
volume averaged or continuum formulations, as in is a function of temperature, stress state, and material
Fig. 10c and d, the plastic deformation of the metal state. Full details of the constitutive response for
grains within the graded layer has also been modelled crystal plasticity are available elsewhere for homo-
at the single crystal level using crystal plasticity geneous metals,107-111and metal-ceramic compos-
theories.134 In this approach, the total deformation ites without compositional gradients112,113and with
gradient F is decomposed into the thermal, plastic, graded composites.134
International Materials Reviews 1997 Vol. 42 NO.3
Suresh and Mortensen Functionally graded metals: Part 2 101

It has been shown that such continuum crystal Onset of inelastic

plasticity models for composite response provide pre- deform\n
dictions of thermomechanical response which are
qualitatively similar to those predicted by the corres- Elastic
ponding von Mises type models, with the exception unloading
that the localisation of intense plastic strains along Elastic
certain directions is much more pronounced in the
former case than in the latter. This difference may
Onset of reverse
highlight some distinctions in the predictions of failure inelastic deformation
processes for the two classes of models. curvature
Other approaches have also been proposed for the
thermomechanical analyses of graded microstructures
by employing higher order theories for partial homo-
genisation which explicitly couple microscopic and
macroscopic responses.135,136 In this approach, the
generic unit cell is not viewed as the representative I~TI-
volume element whose effective properties are cap- 13 Effect of elastoplastic deformation on
tured through homogenisation. Instead, the represen- curvature evolution during thermal cycling
tative volume element is composed of an entire
column of such cells that are stacked in the direction
Published by Maney Publishing (c) IOM Communications Ltd

of compositional gradation, with the response of each mechanical loading. A number of accurate methods,
cell coupled to the response of the entire column. including strain gauges, linear variable differential
transformer (LVDT), and scanning laser techniques,
can be used to measure the in situ evolution of
Experimental measurement of curvature K in the plate during externally imposed
thermal and/or mechanical cycling.19,143-148The cur-
stresses and deformation in
vature change can then be used, in conjunction with
graded multilayers equations (25)-(30), to estimate the variation of
X-ray and neutron diffraction methods137,138 have biaxial stresses through the thickness of the multi-
traditionally been used to assess the internal stresses layer. Figure 13 shows schematically the variation of
in layered materials and coatings. Here the residual curvature with temperature for a layered and graded
strains are computed from experimental measure- plate (such as the one shown in Fig. 7) which is
ments of shifts in lattice spacing, and then are used subjected to thermal cycling. Several features of ther-
to derive residual stresses from known elastic con- momechanical deformation in a graded multilayer
stants of the materials. Residual stress measurements can be extracted from such a plot: '
by X-rays are generally restricted to a shallow region, 1. When all the layers of the plate undergo only
typically of the order of several micrometres, beneath thermoelastic deformation, the curvature changes lin-
a free surface. Although neutrons are capable of early with temperature.
penetrating deeper into the material, both these 2. If inelastic deformation (such as plasticity or
diffraction methods are amenable to uncertainties in creep), phase transformation or dilatational strains
the estimation of internal stresses for the following (such as those arising from martensitic transform-
reasons: (1) it is generally difficult, and in some cases ations in metals and ceramics or swelling due to
impossible, to perform in situ evolution of stresses moisture absorption in polymers), damage (such as
using these techniques during thermomechanical micro cracking in brittle layers or degradation due to
loading of a graded material; (2) when extensive sunlight in organic films), cracking or interfacial
plastic strains occur in the layered structure, the delamination occurs in one or more layers, the I K I
estimation of thermal stresses on the basis of lattice versus I~ TI plot deviates from linearity. This point
constant changes becomes highly inaccurate; and of deviation from a linear variation has been used to
(3) fluctuations and spatial variations in composition quantitatively measure the onset of inelastic deforma-
in a graded layer, over and above the normal micro- tion or damage in graded metal-eeramic
structural variations in the composite layer, can cause multilayers.18,19,126
pronounced errors in the interpretation of results. 3. The direction in which this deviation occurs
In addition to the above diffraction techniques, from the initial linear response can be used to gauge
destructive methods such as layer removal and hole the location within the multilayered solid at which
drilling,139-141 as well as optical fluorescence tech- inelastic deformation is initiated.19
niques142 have been used to estimate internal stresses 4. Reducing the temperature causes a residual cur-
in layered materials. These techniques are not gener- vature to evolve due to permanent deformation in
ally amenable to provide in situ information about one or more layers as a consequence of dislocation
stresses, strains, and geometry changes in a graded plasticity, creep cavitation, grain boundary sliding,
multilayer. irreversible phase transformations, or fracture. The
The evolution of curvature in a layered plate with magnitude of this residual curvature provides an
or without gradients in compositions, when subjected indication of the extent of permanent inelastic
to thermal and/or mechanical loads, also provides a deformation or cracking.
means to assess the evolution of stresses and strains Figure 14a shows a comparison of a numerically
across the thickness of the plate during thermo- predicted variation of curvature with experimental

International Materials Reviews 1997 Vol. 42 NO.3

102 Suresh and Mortensen Functionally graded metals: Part 2

a comparable in the theory and the experiments.19

(a) Figure 14a also reveals that the initiation of plastic
deformation at ~ I50aC shifts the curvature to more
negative values compared to the extrapolation of the
'7 -0.5 initial elastic region. This shift, also predicted by the
model, is indicative of the onset of yielding at the Ni
~ free surface.19 Had plasticity commenced either at the
o> Ni/GL interface or within the graded layer, the curva-
~ ture would have changed slope in the opposite direc-
3 -1.0
tion. Figure 14a thus serves to demonstrate how
~ Experimental result quantitative information on the thermomechan-
- - - - Prediction: smooth profi Ie ical deformation can be obtained from curvature
••••••• Prediction: discrete profi Ie measurements.
a 100 200 300 400
It is worth noting here that while curvature
measurements provide a quantitative measure of the
overall deformation, they are relatively insensitive to
150 local fluctuations in stresses. This point is now illus-
Ni GL trated with the results shown in Fig. 14b, where the
100 equal biaxial normal stress, (Jxx = (Jyy (see Fig. 7 for
Discrete profile the orientation of the coordinate axes), is plotted as
(13 steps)"
Published by Maney Publishing (c) IOM Communications Ltd

a function of position along the layer thickness direc-

c tion in the Ni-AI203 graded trilayer for the same
a.. thermal cycle as in Fig. I4a. The differences between
~ .. a
x the predictions of thermal stress from the discrete
b layer model and smooth gradient model are apparent,
-50 though no differences in curvature predictions were
noted between the two models. The continuous profile
model does not accurately capture the fluctuations in
stresses at the interfaces between the sublayers,
whereas the discrete sublayer model is sensitive to
(b) even the small jumps in compositions at the sublayer
interfaces. (It was also found19 that the discrete sub-
14 Variation of a predicted and measured
layer model exhibited a greater plastic zone at the
curvature as function of temperature and
b predicted stress distribution in Ni-GL-AI203 Ni/G L interface with the corresponding reduction in
trilayer with either smooth or discretely the plastic zone at the Ni free surface.) Despite these
sublayered graded region (After Ref. 19) differences, a number of studies have shown (e.g.
Refs. 15, 18-20, 131) that, for the dimensions consid-
ered in these examples, 10 or more sublayers within
measurements19 (involving an in situ scanning laser the graded layer provide essentially convergent results
method) during one thermal cycle over the range for the thermo mechanical characterisation.
20-475-20aC in a Ni-AI203 graded trilayer (whose
geometry is identical to the one considered in Figs.
10 and 11). For predictions, two different models are
Large deformation of graded
considered: one in which the graded layer is regarded
as continuous with smoothly varying thermal and multilayers
mechanical properties, and the other in which the The thermoelastic and thermoplastic characteristics
microstructure and properties are incremented within discussed thus far on the basis of the classic beam
the graded layer in 13 equal discrete steps. Both and plate theories pertain to small strain, small
models use the modified rule of mixture approxi- deformation behaviour. As reviewed by Hyer149 and
mation (Fig. 2) for the computation of effective elas- Finot and Suresh,127 the small deformation elastic
toplastic response in the graded layer. Only curvature analyses are valid when the following assumptions
changes, instead of the absolute magnitude of the are compatible with real behaviour:
curvature, are considered here in an attempt to cir- 1. The out of plane deflections of the layered solid
cumvent the uncertainties in the intrinsic stresses (and are small compared with the thickness and the
hence any prior curvature) arising from processing or through-thickness stresses in all the layers are small
specimen preparation. It is apparent in Fig. 14a that compared with the in-plane stresses.
little difference exists between the discretely and con- 2. The normals to the interfaces remain unde-
tinuously graded cases through the entire range of formed during deformation of the multilayer.
temperatures. Both models also capture experimental 3. The displacements vary continuously across the
trends well, especially in the elastic region. The experi- interfaces between the layers.
mental results show that the plastic deformation in 4. The strains vary linearly with displacements.
one of the layers commences at f1T;::;j 150 K, which is 5. The free edges, where multi axial stress states
evident from the deviation of the 1(- T plot from prevail, are confined to a small region whose dimen-
linearity. The hysteresis in the curvature after one sions are small compared with the in-plane dimen-
cycle stems from plastic deformation, which is also sions of the layered solid.

International Materials Reviews 1997 Vol. 42 NO.3

Suresh and Mortensen Functionally graded metals: Part 2 103

flat plate lC x = lCy = 0 sphere 1Cx = 1Cy to be constant in the plate. The six degrees of
lCxy =0 1Cxy =0 freedom for the small deformation case are: the two
normal curvatures, Kxx = Kx and Kyy = Ky in the x
/7 and y directions, respectively, the twist curvature,
Kxy' and the components of strain at z = 0 which are
° °
8xx' 8yy, an dOTh
Yxy' e M 0 h r clrc
. Itt'
e represen a Ion 149,150
can be used to visualise the shapes' and curvatures
of the layered plate, as shown in Fig. 15. This circle,
drawn in the Kx or Ky versus Kxy space, is centred at
(Kx + Ky)/2 and has a radius equal to (Kx - Ky)/2.
When the out of plane displacement w becomes
comparable with the plate thickness (while still being
small compared to the in-plane dimensions), the strain
and curvature throughout the layered plate are no
longer uniform.127,149-152 The relation between the
saddle shape lCx = -lCy midplane (z = 0) strains 83 and the displacements
cylinder 1C x=0 J 'ICy i:- 0 along the x and y directions (see Fig. 1), UO and va,
1Cxy i:- 0
(or) 1C y = 0, lex i:- 0 respectively, are given by
lC xyi:- 0

° _ auo !(aw)2
Published by Maney Publishing (c) IOM Communications Ltd

15 Mohr's circle representation of shapes and 8xx - ax + 2 ax

curvatures of layered plates (After Ref. 149)

° _ avo !(aw)2
While the small deformation theory is generally 8yy - ay + 2 ay
adequate for analysing the thermomechanical
deformation of layered and graded materials, there
exist many practical situations where the foregoing
= auo avo (aw) (aw)
a +a + a a . . (43)
y x x y
assumptions break down and large deformations
(albeit small strains) needs to be accounted for. These with the assumption
examples include: unsymmetric laminated tapes of
sintered ceramics and unsymmetric laminates of poly-
w(x, y) = - t (KxxX2 + Kyyy2 + 2KxyXY) . (44)
meric composites with or without stepwise gradients In conjunction with the above formulation, the
in composition, and thin film-substrate systems used in-plane displacements u and v can now be chosen to
in such applications as microelectronics, optoelectron- be polynomials of an appropriate order in x and y.
ics, and thermal barrier coatings which may contain Hyer149,150 used polynomials of order two in his
homogeneous thin films on substrates with a sharp study of large deformation in unsymmetric poly-
or graded interlayer. Furthermore, curvature meric composite laminates, while Masters and
measurements (such as those discussed above), which Salamon151,152 chose a polynomial of order six for
are widely used to assess experimentally the internal large deformation of thin films on substrates and
stresses in layered and graded coatings, can be per- incorporated in-plane shear. With these quantities,
formed with a greater degree of precision if the film- the curvatures can be computed by minimising the
substrate system is designed to undergo large total strain potential energy of the system ~E' and
deformation. examining the stability of various solutions
Consider the layered plate schematically shown in
Fig. 7. When this plate is subject to thermal loading
~E = f Wd d(volume);
from an initial stress free temperature, the relevant Jvol.
total strain components can be written as
. (45)

where Cijkl are the components of the elasticity stiff-

ness tensor.
Finot and Suresh127 have compared a variety of
analytical solutions for small and large deformation
with detailed three-dimensional finite element analy-
ses of Ni-AI203 bilayer plates with sharp and graded
interlayers. Figure 16 shows the evolution of a nor-
. (42) malised curvature with normalised temperature
change for the Ni-AI203 bilayer with a sharp interface
(where each layer in Fig. 7 is of the same thickness,
i.e. h1 = h2) due solely to thermoelastic deformation
arising from a change in temperature from an initial
where G and K represent three-dimensional vectors, stress free temperature. This normalisation was
and w is the out of plane displacement. For small chosen so that the results can be presented independ-
deformation, i.e. when w is small compared with the ently of the in-plane dimensions (which were chosen
total thickness of the multilayered plate (see Fig. 7), to be at least 50 times the total plate thickness, Lx (=
the curvatures and strains at z = 0 can be assumed Ly) > 50). The small deformation equal biaxial stress
International Materials Reviews 1997 Vol. 42 NO.3
104 Suresh and Mortensen Functionally graded metals: Part 2

~-directiOn) evolution of bifurcation. It was shown that unequal

50 Small de.formation theory
(Generalised plane strain) ,./'
,./<// /~ ,,'"
in-plane dimensions of the plate suppress abrupt
shape changes, but instead produce a more gradual
40 ~ ,/' ,"'\ transition from one shape to another. The effect of
Ni-AI203 .' //. . ~~all deformation in-plane dimensional asymmetry, however, on the
Elastic response ,./'" theory
N 30 Lx: Ly // ,"':/' (Biaxial stress state)
evolution of plastic flow in Ni was significantly less
+ h, - h2 ,./' ,,' / pronounced.
A noteworthy trend identified in Ref.127 is that the

20 /~~' -- / introduction of compositional gradients does not have
.'.,. ,. ~ .
.,~ Large deformation theory
any major influence on the critical temperature
./," (Analytical method) change at which bifurcation occurs or on the magni-
.,:,/./ ' Kx(x-direction) tude of curvature corresponding to the onset of
0 bifurcation. This was particularly true for graded thin
films on substrates. Figure 17 shows contours of
0 10 20 30
constant normalised values of temperature change,
~T ~o:L~/(h,+h2)2
L11b, at which bifurcation and abrupt shape changes
16 Variation of normalised curvature, predicted occur in graded multilayers of Ni-AI203 for all poss-
by different analytical and computational ible combinations of the relative thicknesses of the
models for different stress states, as function Ni, gradient, and Al203 layers. The results of this
of normalised temperature change in Ni-AI203 figure have the important practical implication that
square plate (After Ref. 127)
Published by Maney Publishing (c) IOM Communications Ltd

in coatings and structures, the controlled introduction

of a graded interlayer is not likely to offer any benefit
from the viewpoint of avoiding abrupt shape changes,
state produces a linear variation of curvature with though it might suppress thermal stresses, onset of
temperature, as expected. Here the curvature produces yielding and cracking, and· edge effects.
a spherical shape (Kx = Ky and Kxy = 0) of the plate
due to thermal stresses, for the entire range of temper-
atures considered. The small deformation generalised Processing induced stresses in
plane strain state leads to a larger curvature through- graded composites
out, and for this case, the shape of the bilayer is
cylindrical. The discussion up to this point has centred around
The analytical formulation given in equations the analytical, computational, and experimental
(41)-( 43) predicts curvature variations which are characterisation of the thermomechanical response of
shown in Fig. 16 by the solid line. At low values of graded metal-ceramic composites. A key feature
temperature changes, a linear dependence of curvature of such analyses is that the current state of the
on temperature is seen. However, with increasing material, as characterised by the distributions of
thermal stress at increasing temperatures, a pro- stresses, strains, displacements, and damage through
nounced departure from linearity is clearly seen. In the thickness, is a strong function of the initial con-
this non-linear regime, the large deformation theory dition of the material. While most of the foregoing
predicts a spherical curvature which is smaller than analyses envision the initial state as the 'stress free'
that for small deformation. At a critical value of condition at which sintering, diffusion bonding, or
normalised temperature change, which is roughly 11·5 spray deposition of the material is accomplished, it is
in Fig. 16, a bifurcation in the solution occurs. (The widely recognised that essentially all processing
bifurcation was instigated in the simulation by methods produce 'intrinsic' or 'quench' stresses, over
imposing a small moment about the free edge of the and above the 'thermal stresses' induced during tem-
plate.) Beyond this point, three different equilibrium perature excursions from the stress free processing
shapes are predicted: (a) a stable equilibrjum ellip- temperature to the service temperature. * The magni-
soidal shape with a large curvature in the x direction tude and spatial distribution of these internal stresses
and a much smaller one in the y direction, which at depend on the specific processing methods employed
large values of temperature change, approaches a in addition to the choice of geometry and materials,
cylindrical shape; (b) another ellipsoidal shape where and are markedly influenced by such factors as non-
the x and y axes are reversed (this shape is not shown uniform and differential sintering of the material
in Fig. 16 for reasons of clarity of presentation); and through the thickness, non-equilibrium cooling of the
(c)an unstable equilibrium involving a spherical shape different phases, the rapid solidification of a molten
which is an extrapolation of the initial non-linear droplet on a substrate, and epitaxial misfit. In
curve representing large deformation (broken line). addition, the fabrication of such sublayer structures
The three-dimensional finite element analysis127 pre- may also give rise to 'interface mixing' and uncertaint-
dicts overall trends similar to those of the analytical ies in composition.
model. The differences between the two arise from
the fact that the numerical simulations account for * If the intrinsic stress profile is known following processing, and
the non-uniform curvatures along the in-plane axes, if the magnitude of the stresses everywhere is such that only elastic
with larger magnitudes of curvatures which evolve at deformation occurs due to the internal stresses, then this known
stress field can be superimposed on the subsequent externally
the free edges and spread inward. Finot and Suresh 127 applied loads to obtain an improved analysis using any of the
also studied the effects of a number of different methods discussed in the section 'Thermoelastic deformation of
geometric parameters and plastic deformation on the graded multilayers' above.

International Materials Reviews 1997 Vol. 42 NO.3

Suresh and Mortensen Functionally graded metals: Part 2 105

• Residual stress at 20°C
200 A Thermal mismatch stress
Elastic response 20.0
for T = 130°C
O.B L.=L.,
~ 150 • Quench stress at 150°C
en" 100

0.4 Ci5 50

Gradient 2a h1
NI.(AI,O,),.. h, o

0.0 0.2 0.4 0.6 O.B 1.0 Volume fraction of AI2031 0/0
Pa = 2a1(h1+h2) 18 Variation of total processing induced internal
17 Contours of constant normalised values of stress at room temperature, the ther-
temperature change L\ Tb at which bifurcation mal mismatch stress, and internal stress
Published by Maney Publishing (c) IOM Communications Ltd

and abrupt shape changes occur in graded at processing temperature, estimated for
multilayers of Ni-AI203 for all possible different monolayer coatings of Ni-AI203
combinations of relative thicknesses of Ni, composites (After Ref. 148)
gradient, and AI203 layers (After Ref. 127)
processing induced internal stress at room temper-
The various techniques discussed above137,138 for ature estimated for a number of different monolayer
the experimental measurement of internal stresses are coatings. One of the monolayer coatings comprised a
prone to considerable error in the estimation of single layer of Ni-5 wt-O/oAI on a steel substrate
processing induced stresses especially when gradients (which corresponds to the data point at 0°10 alumina).
in microscopic concentrations of different phases The remaining data points refer to single layer graded
occur and inelastic deformation and damage at the coatings where the content of Al203 was changed in
microscopic level strongly control the overall thermo- 20°10 increments through the thickness of the coating
mechanical response. Consequently, no 'standard' on the steel substrate, with one layer ranging from 0
method of proven reliability is available for the esti- to 200/0Al203 in Ni-5AI, and the others ranging from
mation of internal stresses. A key drawback of such 20 to 40, 40 to 60, 60 to 80, and 80 to 1000/0Al203
paucity of understanding of internal stresses is that in Ni-5Al. In Fig. 18, the results for these specimens
the 'initial mechanical state' of the material, a quanti- are plotted at the average composition of 10, 30, 50,
tative description of which is vital to the success of 70, and 90 respectively. The maximum internal

design for thermomechanical performance, remains stress at room temperature is 200 MPa for the 1"'o‫ס‬.I

mostly unknown. This situation is further com- Ni-5AI layer on the steel substrate (which is close to
pounded by the fact that the internal stresses gener- the yield strength at room temperature), and an
ated in graded microstructures have been largely increase in Al203 content causes a reduction in the
unexplored despite the obvious need for research. internal stress. An average internal stress of
A method, which provides an estimate of processing -11 MPa is seen for the single graded coating with
induced stresses, thermal mismatch stresses, as well an average Al203 content of 900/0 and a range of
as in-plane Young's modulus and CTE as a function 80-100 The results in Fig. 18 were obtained by

of temperature, is proposed in Ref. 148 and applied estimating the effective elastic modulus for Ni-AI203
to graded Ni-AI203 plasma sprayed coatings on steel composite coatings which were prepared by thermal
substrates. Here use is made of a number of identical spraying (and whose effective· stiffness can be five
specimens of the substrates in the spray chamber for times smaller than that of the corresponding fully
the simultaneous coating of surface layers with fixed dense microstructure). * From the experimentally
or stepwise graded compositions (in increments of determined values of the effective CTE of the
sublayer thicknesses of the order of 100 Jlm). Coated composite coatings, Kesler et al.148 also estimated
specimens are periodically removed from the depos- the quench stress at the processing temperature of
ition chamber such that different thicknesses of coat- 150 C. The difference between the quench stress at
ings could be obtained on identical substrates under 150aC and the residual stress at room temperature
the same spray conditions or different compositions is plotted as the thermal mismatch stress for a
of the metal-eeramic graded coating can be produced temperature excursion of 130 C.
for the same coating thickness. By performing a priori
and in situ strain or curvature measurements during * Similar results can also be derived for thin graded films on thick
the spray deposition, and subsequent four-point bend substrates by recourse to measurements of curvature changes before
tests and thermal loading at different temperatures, and after processing and by employing the classic Stoney153
approximation. Here the average stress in the graded thin film is
the distribution of quench stresses and thermal mis- given by l(jl = Es1z;Kj(61zfilm). The calculation of stress does not
match stresses through the thickness of the coatings involve the properties of the film, but is based solely on curvature
is obtained. Figure 18 shows the variation of the total changes, the substrate stiffness and the thicknesses of the layers.

International Materials Reviews 1997 Vol. 42 NO.3

106 Suresh and Mortensen Functionally graded metals: Part 2

interfaces intersect free boundaries, however, there

exists a general multi axial stress state with non-zero
c.5 components of shear stresses and out of plane normal
~150 stresses. Even when there are no cracks present in the
co layered material, these critical locations serve as sites
~ of singular stresses whose magnitudes are given by154

co 50·
<A <2

.~ 0'---+--1- (i, j = x, y); 0 (46)

cc 10 20
Volume fraction of A1203, % where rand 8 are the polar coordinates centred
at the point of .singularity, A a constant signifying
the strength of the singularity which depends on
the mismatch in mechanical properties between
the materials on either side of the interface, Bij(8) are
non-dimensional universal functions of 8 and the
material mismatch, and K a scalar measure of the
intensity of the singular field. Abrupt discontinuities
in material properties across the interface produce
positive values of A. When the interface exhibits
Published by Maney Publishing (c) IOM Communications Ltd

continuously varying mechanical properties through

a finite thickness as a result of compositional gra-
dation, it can be shown24,154 that the stresses become
finite and that A becomes zero.
The effectiveness of a graded interlayer in mitigating
the singular stress fields at edges and sharp corners
a estimated variation of total internal stress in is illustrated here with a specific example due to
continuously graded Ni-AI203 graded coating Erdogan and co-workers.24,155 Figure 20a-c shows
produced by plasma spraying and b growth of schematics of a Rene 41 Ni-base superalloy substrate
crack from surface where high tensile stresses which is coated with a single layer (100% zirconia),
are induced during processing (After Ref. 148) two layers (which comprise 50% zirconia-50%
Rene 41 and 100% zirconia), and four layers (which
comprise 25, 50, 75, and 1000/0zirconia), respectively,
Figure 19a shows the estimated variation of total for thermal protection.24 Figure 20d shows the vari-
internal stress in a continuously graded Ni-AI203 ation of Young's modulus with thickness of coatings
coating produced by plasma spraying.148 The speci- of the same overall thickness which are continuously
mens used to obtain these results comprised steel graded from the superalloy substrate to an all zirconia
substrates on which a Ni-5Allayer was deposited, as outer surface with three different gradient profiles:
well as those containing coatings of Ni-AI203 com- linear, metal rich, and ceramic rich; (Since the coating
posites with Al203 volume concentration terminating considered here is deposited by thermal spraying
at 0, 20, 40, 60, 80, and 1000/0.Figure 19a reveals that methods, the elastic modulus of the ceramic is signifi-
the stress in the graded coatings is mostly tensile, and cantly lower than that of bulk zirconia, and is also
that it first decreases, and then increases from the Ni lower than that of the substrate.) The dimensions of
rich end to the alumina rich end. The highest value the coatings and the substrates, and the orientation
of tensile residual stress, ~ 200 MPa, is at the surface of the reference coordinate axes for subsequent dis-
of the Al203 deposit. Since this value exceeds the cussion of stress state are also marked in Fig. 20.
tensile strength of sprayed alumina (I"V 200-250 MPa), Figure 21a shows the shear stress (Jxy (arising from a
the surface layer of the fully graded coating is expected thermal excursion L\ T) plotted as a function of
to crack under a state of equal biaxial stress, which location along the width, which varies from the centre
is expected to advance normally from the free surface of the substrate width (x = 0) to the free edge (x =
to the interior. An optical micrograph of such a Lx/2), for the superalloy coated with a single layer of
surface crack is shown in Fig. 19b for the graded zirconia and with two or four discrete layers of the
coating. superalloy-zirconia composite. The shear stress is
normalised here by (Js = Esll.sL\T, where the subscript
s refers to the properties of the substrate. In all three
Edge effects and singular fields cases, the shear stresses assume non-zero values over
When plates of dissimilar materials are bonded a distance along the x axis from the free edge which
together with no cracks or macroscopic defects along is roughly equal to the total thickness of the layered
their interfaces, a state of equal biaxial loading exists solid. The interior of the substrate-coating near the
in the regions ·of the plate away from the free edges. centre (x = 0) is free of shear stresses, and near the
In this biaxial field, the materials undergo normal free edges an r - A stress singularity is predicted for all
stresses only in the in-plane directions and are free of three discrete stepwise coatings. The normal stresses
shear stresses or out of plane stresses. Under these (Jyy are also unbounded at the free edge in this case
conditions, the plate theory formulations specified in (not shown in the figure). The corresponding predic-
equations (25)-(30) hold. At locations where the tions are plotted in Fig. 21b for the three graded
International Materials Reviews 1997 Vol. 42 NO.3
Suresh and Mortensen Functionally graded metals: Part 2 107

T.....-. Z_i_r_c_o_n_ia ~

T -0.05
o --o~---'=--,----

Sharp interface

1.45cm Rene 41 substrate 1.25cm o Four layer coating

l~_----,....:.. ! tf

Two layer coating
One layer coating

(a) 1- 4-- L)( = 4cm ----- -I

2 layer coating {A singularity near the
free edge of Interface
o 0.004 0.008 0.012 0.016 0.020
x, m
y o - ------~

Published by Maney Publishing (c) IOM Communications Ltd

4 layer coating

Graded interface
o Metal rich gradient
6. Linear gradient
<) Ceramic rich gradient
(c) -0.15

240r (b)
I (d)
o 0.004 0.008 0.012 0.016 0.020
x, m
o 200 21 Normalised shear stress, arising from thermal
a.. excursion, plotted as function of location along
width, which varies from centre of substrate
w.. 180 width (x= 0) to free edge (x= Lx/2), for
Linear superalloy coated with a discrete layers and
gradien1 b graded layers (After Refs. 24, 155)

140 Since the sharp interfaces produce the largest

I. 20 I. 25 1.30 1.35 1.40 1.45 1.50 stress concentration in the vicinity of free edges, they
also promote early onset of plastic flow or cracking
y, em compared with the graded interfaces. Consequently,
20 Rene 41 Ni-base superalloy substrate which is the vicinity of free edges serve as potent sites for
coated with a single layer (100% zirconia), the nucleation of dominant cracks, as a result of
b two layers (50% zirconia-500/0 Rene 41 and brittle tensile failure mechanisms on the ceramic
100% zirconia), and c four layers (25, 50, 75, side of the metal/ceramic interface, or ductile failure
and 100% zirconia), for thermal protection; aided by plastic strain accumulation (due to thermal
d shows variation of Young's modulus with cycling) or void growth on the metal side, or
thickness of coatings of same overall thickness interfacial debonding. Suresh et aI.,17 Yang and
which are continuously graded from superalloy M~nz,28,156 and Arai et al.157 have performed compu-
substrate to all zirconia outer surface with
tational simulations of the evolution of plastic zone
three different gradient profiles: linear, metal
rich, and ceramic rich (After Refs. 24, 155)
at the free edges along metal/ceramic interfaces,
whereas creep zones at free edges have been analysed
by Williamson et aI.158 It has also been shown by
coatings shown in Fig. 20d. It is evident that the stress Williamson and Rabin 159 that the reduction in the
singularity is eliminated by the smooth gradation in magnitude of a certain stress component at a critical
composition across the thickness of the coating for location in a graded material may in fact lead to an
all three gradient profiles, and that the shear stress increase in the magnitude of another stress com-
vanishes at the free edge, where the normal stress (Jyy ponent. Thus optimisation of graded compositions
(not shown in the figure) has a finite value. The metal to mitigate edge effects is strongly dictated by the
rich gradient profile shows the smallest edge effect choice of materials, geometry, and the specific failure
whereas the ceramic rich gradient has the largest. mechanisms.
International Materials Reviews 1997 Vol. 42 NO.3
108 Suresh and Mortensen Functionally graded metals: Part 2

y crack tip (in the vicinity of x = c) vary as

O"ij(r, 8) = exp [r(B1 cos 8 + B2 sin 8)]
k1 -I k2_11
x (2nr)1/2 O"ij(8) + (2nr)1/2 O"ij(8) (49)
Here, the stress intensity factors at the crack tip are
1---2c~ defined as
k1 (c) = lim [2n(x - C)]1/20"yy(X, 0);

k2(c) = lim [2n(x - C)]1/20"XY(X, 0) . . . . (50)

22 Crack centrally located in graded medium x~c

Note that the universal angular functions in equations

(47) and (49) are identical. The asymptotic stress
Fracture and fatigue of states for the inhomogeneous cracked body in Fig. 22
graded materials and the corresponding homogeneous solid are the
same only at the crack tip (r ~ 0).
A key feature of fracture which distinguishes graded The crack driving force for the fixed grip loading
materials from homogeneous ones is that the resist- case (at the right side crack tip, x = + c) is the strain
ance of the former to fracture and damage tolerance
Published by Maney Publishing (c) IOM Communications Ltd

energy release rate G

varies spatially. The design of graded materials against
failure is, therefore, inevitably linked to the specific G = n(l + 1]) (k21 + k22) (51)
geometry of the material vis-a-vis the choice of gradi- 8 fl( c, 0) . . . . . . . .
ents in mechanical properties. Consequently, analyses
of fracture in graded materials are considerably more where 1] = 3 - 4v in plane strain and 1] = (3-v)/(l +v)
complex than in the corresponding homogeneous case in plane stress. For brittle fracture in a graded material
of the same specimen and crack geometry subjected which is subject only to local mode I loading (i.e.
to the same loading conditions. Despite these compli- k2 = 0 in equation (51)), one may invoke the criterion
cations, some basic features of fracture mechanics in that quasistatic fracture initiation occurs when
inhomogeneous solids in general and their impli- G~ Gc, where Gc is a measure of the fracture initiation
cations for failure can be extracted from available toughness. Unlike the case of a homogeneous solid,
results. Gc varies with location in the graded material.
Fracture mechanics analyses providing stress intensity
calibrations for a variety of crack geometries, grad-
Fracture mechanics for inhomogeneous
ient profiles, and loading modes can be found
solids elsewhere.24,161-165
Consider the plane problem of a linear elastic, infinite
medium which contains a centrally located line crack Crack driving force in graded materials
of length 2c, as shown in Fig. 22. This crack is located The effect of compositional gradients in reducing the
along y = 0 and - c < x < c, and is subjected to effective driving force for fracture is illustrated with
arbitrary loading. If the cracked body were to have the example sketched in Fig. 23a. This figure shows
homogeneous elastic properties throughout, with two symmetrically formed edge cracks located at the
spatially invariant shear modulus fl and Poisson ratio interface between a Rene 41 Ni-base superalloy sub-
v, the crack tip stress fields will be characterised by strate and a coating. Eight different possibilities are
the following asymptotic response according to linear considered for the choice of this thermal barrier
elastic fracture mechanics for homogeneous solids160 coating, with all the geometric dimensions fixed. The
first of these is a homogeneous zirconia coating
K1 -I KII_II directly on to the superalloy substrate. The seven
O"ij(r, 8) = (2nr)1/2 O"ij(8) + (2nr)1/2 O"ij(8), i, j = r, 8 remaining possible choices are coatings which involve
. . (47) gradient profiles of the superalloy and the ceramic
that include two sigmoidal variations of the graded
where K1 and KII are the mode I and mode II stress composition through the thickness of the coating, two
intensity factors, respectively, at the tip of the crack metal rich gradient profiles, two ceramic rich gradient
under general plane loading, rand 8 the polar coordi- profiles, and a linear profile. The elastic modulus of
nates centred at the crack tip, and atj(8) and al](8) the composite, normalised by the modulus of the
known non-dimensional universal functions of 8. substrate, plotted as a function of the thickness of the
Now consider the cracked body to be an inhomo- coating is shown in Fig. 23b for each of these seven
geneous medium (Fig. 22) with a constant value of graded .coatings. Figure 24 shows the variation of
Poisson ratio v and a shear modulus whose spatial the strain energy release rate G, normalised by the
variation is given by corresponding value for the substrate Gs =
(1 - v;)Es nb(as~ T)2 (where b is the edge crack length
fl(X, y) = flo exp(B1x + B2y) . . . . . . (48)
in the substrate), as a function of the relative crack
where flo, B1, and B2 are known parameters. For this depth cll for the homogeneous and inhomogeneous
case of non-uniform material property, Konda and coatings shown in Fig. 23a and b, in response to
Erdogan161 have shown that the singular fields at the thermal excursions. Since the homogeneous single
International Materials Reviews 1997 Vol. 42 NO.3
Suresh and Mortensen Functionally graded metals: Part 2 109

Ceramic rich I


rich 2

Vl ------~oida-I~---
<.9 •• ,...------- ------ .••."'-
o x
<3 0.02 ~/' Linear ~

Metal rich I

0,01 Metal rich 2

••... o
o 0.2 0.4 0.6 1.0

~ 0.90
c /1
24 Variation of the strain energy release rate G,
normalised by corresponding value for
Published by Maney Publishing (c) IOM Communications Ltd

.~ 0.80 substrate Gs' as function of relative crack
"0 depth ell for homogeneous and inhomo-
o geneous coatings (After Ref. 24)
One layer ical damage and contact fatigue failure.172-174While
0.85 0.90 0.95 1.00
there exists a vast amount of practical data on the
fatigue of layered structures and weldments, there is
(b) Normalised thickness,y/{hl+h2)
a surprising paucity of fundamental information from
23 a two symmetrically formed edge cracks which conclusive trends could be extracted.
located at interface between Rene 41 Ni-base In an attempt to rationalise results from a study of
superalloy substrate and a coating and b eight fatigue crack propagation across ferritic/austenitic
different possibilities for choice of thermal steel interfaces,175,176and to assess the driving force
barrier coating, with all geometric dimensions for fracture across interfaces with homogeneous and
fixed. First is a homogeneous zirconia coating graded properties, Sugimura et al.I77,178 and Kim et a1.29
directly on to superalloy substrate; other seven have conducted detailed finite element simulations of
coatings involve gradient profiles of super- near tip fields. These simulations consider three
alloy and ceramic (two sigmoidal variations
different interfaces oriented normally to the plane of
of graded composition through thickness of
coating, two metal rich gradient profiles,
the crack between two materials of the same elastic
two ceramic rich gradient profiles, and a and thermal expansion properties and isotropic strain
linear profile) for which elastic modulus hardening response, but different yield strengths: an
of composite, normalised by modulus of idealised, sharp interface of zero thickness, a homo-
substrate, is plotted as function of thickness geneous interlayer of finite thickness (whose elastic
of coating (After Ref. 24) properties are the same as the two materials it separ-
ates, but whose yield strength, uniform through the
interlayer thickness, is the arithmetic mean of that of
layer zirconia coating on the alloy substrate produces
the two materials), and a graded interlayer of finite
the largest thermal mismatch stress, it exhibits the
thickness within which the yield strength varies lin-
highest G for all ell. For the choices of gradient
early with distance across the interlayer from one
profiles considered here, it is apparent that a reduction
material to the other. It has been found that the crack
of as much as a factor of four can be gained in the
tip 'driving force', as characterised by the near tip
strain energy release rate by the proper choice of a
J integral,179 Jtip, is altered vis-it-visthe applied driv-
compositional gradient, which signifies a doubling of
the apparent fracture toughness. ing force J app' When the crack approaches the
interface from the weaker material, Jtip becomes
smaller than J app' as the plastic zone spreads
Fatigue crack growth across homogeneous across the interface (Fig. 25a). (In this figure,
and graded interlayers Jtip/Japp is plotted against K/((Jy VI), where K =
Fatigue failure, arising from fluctuations in thermal [Japp(l- v2)/E]I/2, (Jy the yield strength of the plas-
and/or mechanical loads during service, is one of the tically weaker material, and L the distance from the
most potent mechanisms of failure in a number of crack tip to the centre of the interlayer, as shown in
practical applications in which layered and graded Fig. 25b.) In other words, the crack is shielded from
materials are used, or are candidates for potential the applied loads as the plastically stronger material
applications. These applications primarily involve further ahead of the crack tip sustains a greater
cyclic damage and cracking in thermal barrier coat- magnitude of opening stress than the weaker material
fatigue in the heat affected zone of directly in front of it. The homogeneous interlayer
welded structures,l71 and coatings to prevent tribolog- exhibits the greatest shielding effect, and the sharp
International Materials Reviews 1997 Vol. 42 NO.3
110 Suresh and Mortensen Functionally graded metals: Part 2

- - - - - - - homogeneous interlayer
bimaterial interfaces wherein criteria for the growth
1.3 ---.- ..----. graded interlayer of the cracks are formulated in terms of the relation
-- sharp interface between the crack opening displacement and the
12 traction acting across the crack surfaces.1Sl In
addition, criteria for different brittle failure modes
1.1 have been formulated for elastic cracks approaching

~~::--...a 7 8 9
sharp interfaces and those with homogeneous prop-
erties,182-18Sand fracture mechanics solutions for


--_:::~~~~::~:::::)_"' cracks intersecting perpendicularly oriented interfaces
between two elastic solids have been developed.18s-187
The foregoing results clearly reveal that the proper
design of a layered or graded interface can potentially
result in the deceleration or even arrest of the crack.
(a) soft~hard
This point has considerable implications for the
fatigue resistance of layered and graded coatings, as
illustrated in the following example. Plasma sprayed
chromia (Cr203) coatings, typically 100-500 Jlm in
thickness, are commonly used to enhance resistance
to contact fatigue in a number of structural com-
ponents.188 When such coatings are deposited directly
Published by Maney Publishing (c) IOM Communications Ltd

E,v E,v E,v on to the steel substrate, and the coated material is
cry" N, cry, N cry2, N2 subjected to fatigue, the crack, once initiated on the
free surface of the coating runs rapidly through the

__ X'l{
thickness of the brittle coating; the interface, oriented
normally to the surface crack, easily delaminates and
a new fatigue crack initiates in a nearby location
and fractures the substrate. Figure 26a shows an
example176of such a fatigue fracture in a 0·45 wt-%C
~ I x, t
steel spray coated with Cr203' In order to improve
the bond strength, a plasma sprayed interlayer or
~L-- bond coat, typically of a ductile alloy such as
(b) (L=I+; ) Ni-5 wt-%AI of about 50-100 Jlm thickness, is sand-
wiched between the Cr203 and steel layers. If the
25 a predicted variation of
Jtipl Japp as function of
thickness of the bond coat is significantly greater than
for different interlayer geometries
the size of the plastic zone (for a given service loading
and properties and b schematic of interlayer
geometry (After Ref. 29)
condition) when the crack tip approaches the interface
between the ceramic and the bond coat, the propensity
for the crack to jump across the bond coat will be
interface exhibits the least shielding effect, with the suppressed. In this case, the results of experiment
graded interlayer exhibiting an inbetween trend. shown in Fig. 25 would suggest that the crack would
When the crack approaches the interface from the advance through the interface unimpeded if this
stronger material, the near tip driving force is ampli- interface remains well bonded. However, when the
fied, and hence Jtip/Japp > 1 for all three cases. Recall crack advances through the bond coat and its plastic
that for a homogeneous material, J = J app' since tw zone begins to interact with the interface with the
the J integral is path independent.17 Kim et al.29 steel substrate, i.e. the crack is approaching an
have shown that the extent of shielding and amplifi- interface from a weaker material, crack tip shielding
cation is strongly influenced by the direction of crack is expected. (Note that the elastic moduli of Ni-5AI
advance relative to the interface, the distance from and steel are comparable.) In this case, the crack tip
the crack tip to the interface, the thickness of the would be expected to arrest before reaching the bond
interface and the gradient in the properties within the coat/substrate interface. This is indeed seen exper-
interface. imentally, as shown in Fig. 26b.176 The arrest of the
If now an elastic mismatch is superimposed on the crack enhances the overall fatigue crack growth resist-
plastic mismatch between the two materials, the extent ance of the substrate-coating system. If one compares,
of shielding and amplification can be altered in Fig. 26c, the stress-life (S-N) fatigue curves for the
depending on the relative magnitudes of the two uncoated steel, steel coated only with the ceramic
mismatches. An additional complicating effect in the outer coating, and steel coated with both a bond coat
analyses of near tip driving force for fracture at and the ceramic outer coating, the beneficial effect of
interfaces is the thermal mismatch. Numerical simu- the interlayer is evident. By virtue of its hardness, the
lations180of crack driving force for a ferritic-austenitic ceramic surface coating improves the resistance to
bilayer have shown that the thermal mismatch aids fatigue crack initiation at the free surface where the
both the shielding and amplification effects, especially cracks nucleates189 compared with the behaviour of
when the crack tip is close to the interface. the steel substrate. The S-N curve for the former
While the numerical results given in Fig. 25 do not system is, therefore, higher than that of the substrate
invoke any crack growth laws, cohesive zone models alone. However, when the bond coat is introduced
have been developed for cracks terminating at between the ceramic and the substrate, the ensuing
International Materials Reviews 1997 Vol. 42 NO.3
Suresh and Mortensen Functionally graded metals: Part 2 111

tailoring stepwise or continuously graded interlayers

for a variety of practical applications, such as in
thermal barrier coatings, where both the crack
initiation and crack growth resistances could be suit-
ably altered by reducing the effective driving force
for fracture.

Concluding remarks
The development of graded metals and metal-eeramic
composites can offer a number of distinct advantages
for thermomechanical properties, over those seen in
conventional structural metals and composites. These
include: reduction and optimal distribution of stresses,
suppression of free edge effects and of the attendant
singular fields at interfaces, enhanced interfacial bond
strength, the feasibility of depositing 'thick' coatings
on substrates, a greater resistance to damage by sharp

indentors or abrasive failure processes, and a
reduction in the driving force for fracture. It is demon-
Published by Maney Publishing (c) IOM Communications Ltd

a strated in this review that the thermomechanical

300J.lrn characteristics of a graded material are strongly influ-
enced by a number of length scales and geometrical
A Steei blasted with alumina powder
and annealed (no coatinCJs)
parameters. Examples of such length scales and geo-
metrical considerations include spatial gradients in
Steel coated with
concentration or, equivalently, the (macroscopic)
o 430,um Cr203 graded layer dimensions, their relation to micro-
• 400,u.m Cr203 + 50,u.m Ni- 5AI inter layer structural length scales such as particle dimensions,
800 distribution or connectivity, and/or the Burgers vector
associated with misfit dislocations, as well as ratios
a.. 700 • of surface energy and surface stress to volumetric
~ stored energy.190 Among the various approaches
600 o
• reviewed in this paper for the thermomechanical
modelling of graded materials, the choice of a partic-
• ular approach inevitably rests on the length scale that
0- 500 A 0 •
E A 0 is relevant for a given geometry and application.
o A
400 The theoretical frameworks for metal-eeramic
(/) A
composite analysis which comprise such diverse
300 approaches as the rule of mixture formulations, mean
en field theories, continuum finite element models, crystal
plasticity models, and discrete dislocation models,
provide a starting point for the development of
thermomechanical models for graded composites.
Number of cycles to failure (N )
f There are, however, a number of fundamental
issues specific to the graded material which warrant
26 Fatigue crack growth characteristics of
additional considerations:
chromia coated steel a with and b without
1. What is the local value of an effective property
interlayers; c S-N fatigue curves for steel
without any coating, with only chromia given a continuously changing composition or micro-
coating, and with chromia coating and Ni-AI structure? What size scales should experiments aim
interlayer (After Ref. 176) at in order to pro be such effective properties?
2. What is an effective 'layer size' in a smoothly
graded microstructure, and how does it compare with
crack arrest elevates the resistance to fatigue crack the characteristic microstructural length scale such as
propagation in addition to the crack initiation resist- a grain size, particle size, or dislocation cell size?
ance. As a result, additional improvements are seen 3. Since all processing methods invariably give rise
in the S-N curve.176 to a stepwise variation rather than a perfectly smooth
The design of the bond coat is usually based on variation in composition or microstructure, how
considerations of interfacial strength and environmen- small, relative to the total graded layer thickness,
tal resistance (as in the case of the Ni-Cr-AI-Y bond must the step size be before the potential advantages
coat used between the Ni-base superalloy substrate of a smooth variation for thermomechanical perform-
and the zirconia thermal barrier coating in aircraft ance can be realised?
jet engines). The results of Fig. 26 additionally show 4. Are the graded microstructures designed for
that interlayers and bond coats could be designed for optimal thermomechanical performance thermo-
improvements in thermomechanical fatigue resistance. dynamically, chemically, and morphologically stable
This particular feature offers many possibilities for under service temperatures and loads to realise the

International Materials Reviews 1997 Vol. 42 NO.3

112 Suresh and Mortensen Functionally graded metals: Part 2

potential benefits of compositional gradation? How

do the spatial variations in desired properties compare Appendix
with those arising from normal microstructural segre-
gation intrinsic to a processing method? Lower limits List of symbols
in size scales may occur in many systems, such as a one half of thickness of graded layer
Ti-Al, Cu-Nb, and Fe-Pt, where one plane adopts a al minor axis diameter of axisymmetric
non-equilibrium structure due to the interface or .ellipsoidal inclusion
epitaxy stresses. In addition, small scale layered struc- aL major axis diameter of axisymmetric
tures may exhibit other flow processes such as large ellipsoidal inclusion
driving forces for interdiffusion. ii Cartesian base vectors for crystal
5. Given that the fracture and fatigue resistance of plasticity
graded materials are influenced strongly by the a thermal strain concentration tensor
geometry and gradients in properties, what general A empirical constant
methods of broad interest can be developed for Al diameter of axisymmetric unit cell
assessing the toughness and damage tolerance? AL height of axisymmetric unit cell
6. Since graded composites contain a large number A strain concentration tensor
of microscopic and macroscopic interfaces, the micro- b depth of edge crack in substrate
mechanisms of fundamental and practical interest can hi function which characterises variation of
extend over a broad range of size scales. While atomic i{ with z
level modelling is obviously necessary for a complete Bb B2 material constants
Published by Maney Publishing (c) IOM Communications Ltd

understanding of microscopic interfaces, such as those Br phase concentration factor tensor

between the grains of a metal and a ceramic, under c half crack length
what conditions do macroscopic failure analyses CijkI components of stiffness tensor
require interface separation laws which may warrant Ci compliance tensor of inclusion phase
a knowledge of the force-displacement relations at Co compliance tensor of matrix phase
an atomic level, in addition to micromechanical and C* effective compliance tensor
continuum modelling. E Young's modulus
Developing the necessary numerical and experi- ~E change in Young's modulus with
mental tools to address the above issues will certainly temperature
provide challenging opportunities for both funda- Es Young's modulus of substrate
mental research and practical applications involving E biaxial Young's modulus = E/(l- v)
functionally graded materials. Beyond these challenges Eo biaxial Young's modulus at temperature
lie those which arise from the need to integrate the 10
fundamentals of processing (addressed in Part 1, ~Eo change in biaxial Young's modulus
Ref. 1) with those of thermomechanical behaviour E* effective shear tensor
(addressed here), to achieve the optimisation that I volume fraction
underlies the entire approach of functional grading of Ip volume fraction of micropores
materials. This optimisation must, for each process Ii fraction of inclusion phase in graded
and material type, lead to the best possible perform- region
ance predicted by thermomechanical analysis, within I'm fraction of matrix phase in graded region
the constraints of processing. It is believed that, though pap in-plane normal force applied on layered
tangible successes have been achieved in this direction, plate
the definition of one, or several, integrated gradation F total deformation gradient
approaches, which cut across the wide spectrum of FP plastic component of deformation
materials science and engineering, is still in its early gradient
stages. This exploration remains perhaps the most FT thermal component of deformation
challenging and most interesting within this relatively gradient
new arena of scientific and industrial pursuit. F* lattice component of deformation
G strain energy release rate
Acknowledgments Gc critical value of strain energy release rate
The authors acknowledge many helpful discussions Gs strain energy release rate for substrate
and technical collaborations on the topic of metal- hfilm thickness of film on substrate
ceramic graded materials with J. D. Embury, hs thickness of substrate
M. Finot, A. E. Giannakopoulos, O. Kesler, A. S. Kim, hi, h2 distance from reference plane to free
S. Sampath, and C. F. Shih, and on the topic of metal surface of multilayer
matrix composites with Y.-L. Shen and A. Needleman. H elastoplastic tangent modulus
Professor Suresh further acknowledges support Ih Ji constants defined in equation (30)
of his research in the general area of graded materials I identity tensor
through Grant N00014-94-1-0139 from the Office J app remote value of J integral
of Naval Research, Grant DE-FG02-93ER45506 Jtip near tip value of J integral
from the Basic Science Division of the US Department k1, KI mode I stress intensity factor in graded
of Energy, and Grant N00014-94-1-0832 from the and homogeneous materials
Office of Naval Research and the Advanced Research k2, Kn mode II stress intensity factor in graded
Projects Agency. and homogeneous materials
International Materials Reviews 1997 Vol. 42 NO.3
Suresh and Mortensen Functionally graded metals: Part 2 113

l edge crack length fl shear modulus

L distance from crack tip to centre of flo reference value of shear modulus
interlayer v Poisson ratio
in-plane dimension of layered plate Q E/[3( 1- 2v)J = bulk modulus
unit vector normal to slip plane in slip Pa normalised thickness of graded layer, a/h
s)T'Stemf3 a stress
bending moment applied on layered ae von Mises effective yield stress
plate as stress in substrate
p any quantity ay yield strength
p volume average value of p aij component of stress tensor
P height of unit cell fi overall tensile yield stress of composite
P effective property within graded layer fiy overall limit yield stress of composite
~ effective property estimated from mean 6ij functions of the polar angle 8;
field approach superscripts I and II denote modes I
q ratio of stress to strain transfer and II
r distance in cylindrical coordinates 8app homogeneous applied stress
deviatoric stress components 8* effective volume averaged stress
unit vector along slip direction in slip 8 polar angle
system f3
S Eshelby tensor Subscripts
Published by Maney Publishing (c) IOM Communications Ltd

T temperature Unless stated otherwise, the subscripts connote the

I1T tern pera ture change following meaning when applied to these variables:
11 {ux, uy, uz} = {u, V, w} = displacement a, A, E, E, H, S, ~ a, €, v, u, and fi.
vector e composite
V control volume i inclusion
w out of plane displacement 111 matrix
W width of unit cell 1, 2 phases 1 or 2
~ strain potential energy per unit volume
~E total strain potential energy
x,y,z Cartesian coordinate axes References
X material position vector
1. A. MORTENSEN and s. SURESH: 1nt. Mater. Rev., 1995, 40,
isotropic thermal expansion coefficient 2. M. YAMANOUCHI, M. KOIZUMI, T. HIRAI, and I. SHIOTA (Eds):
(CTE) Proc. 1st Int. Symp. on 'Functionally gradient materials',
CTE for axial loading Sendai, Japan, 1990, Functionally Graded Materials Forum
and the Society of Non-Traditional Technology.
CTE for bending
CTE of substrate R. M. SPRIGGS, and Y. KAIEDA: Ceram. Eng. Sci. Proc., 1992,
change in thermal expansion coefficient 13, 392-409.
with temperature 4. S. SURESH, A. MORTENSEN, and H. McMANUS (Eds): MIT-ONR
tensor of thermal expansion coefficients Workshop on 'Functionally graded structural materials',
Cambridge, MA, 1994; Materials Processing Center;
index which refers to slip system Cambridge, MA, Massachusetts Institute of Technology.
shear strain 5. B. ILSCHNER and N. CHERRADI (Eds): 3rd Int. Symp. on
midplane shear strain at centre of 'Structural and functional gradient materials', Lausanne,
multilayer plate Switzerland, 1994; Lausanne, Presses Poly techniques et
Universitaires Romandes.
slip rate on slip system f3
6. B. H. RABIN and I. SHIOTA (Eds): 'Functionally graded materials'
strain (Special issue), MRS Bull., 1995, 20, (1).
axial strain 7. M.-J. PINDERA, J. ABOUDI, S. M. ARNOLD. and w. F. JONES (Eds):
bending strain 'Use of composites in multi-phased and functionally graded
normal strains along x and y directions, materials' (Special issue), Compos. Eng., 1995,5, 743-974.
8. A. NEEDLEMAN and s. SURESH (Eds): 'Mechanics and physics
respectively of layered and graded materials (Special issue) J. Mech. Phys.
normal strains along x and y directions Solids, 1996, 44, 643-825.
at midplane 9. Y. MATZUSAKI: Doctoral thesis, Kyoto University, Japan, 1994.
normal in-plane strain in layered plate at 10. Proc. of 4th Int. Symp. on 'Functionally graded materials',
z=o Tsukuba City, Japan; 1996, Tokyo, Functionally Graded
Materials Forum, in press.
constrained strain 11. A. J. MARKWORTH, K. S. RAMESH, and w. P. PARKS, Jr: J. Mater.
stress free transformation strain Sci., 1995, 30, 2183-2193.
equivalent transformation strain 12. T. HIRAI: in 'Processing of ceramics, Part 2: Materials science
tensor of midplane strains and technology, a comprehensive treatment', (ed. R. W. Cahn
et al.), Vol. 17B, 293-341, 1996; Weinheim, Germany, VCR
effective volume-averaged strain Verlagsgesellschaft mbR.
(Kxx, Kyy, Kxy) = (Kx, Ky, Kxy) = curvature 13. J. B. HOLT, M. KOIZUMI, T. HIRAI, and z. A. MUNIR (Eds): Pmc.
curvature tensor 2nd Int. Symp. on 'Functionally gradient materials', San
constant signifying strength of singularity Francisco, CA, Nov. 1992, Ceramic Transactions, Vol. 34;
at crack tip 1993, Westerville, OR, The American Ceramic Society.
14. M. KOIZUMI: Ceram. Eng. Sci. Proc., 1992, 13, 333-346.
A vertical position as function of total 15. R. L. WILLIAMSON, B. H. RABIN, and J. T. DRAKE: J. Appl. Phys.,
height of layer 1993,74, 1310-1320.

International Materials Reviews 1997 Vol. 42 NO.3

114 Suresh and Mortensen Functionally graded metals: Part 2

16. J. T. DRAKE, R. L. WILLIAMSON, and B. H. RABIN: J. Appl. Phys., 55. o. B. PEDERSEN: Z. ang. Math. Mech., 1978,58,227-228.
1993, 74, 1321-1336. 56. o. B. PEDERSEN: Acta Metall., 1983, 31, 1795-1808.
17. s. SURESH, A. E. GIANNAKOPOULOS, and M. OLSSON: J. Mech. 57. R. HILL: J. Mech. Phys. Solids, 1965, 13, 213-222.
Phys. Solids, 1994, 42, 979-1018. 58. Y. BENVENISTE and G. J. DVORAK: in 'Micromechanics and
18. A. E. GIANNAKOPOULOS, s. SURESH, M. FINOT, and M. OLSSON: inhomogeneity', (ed. G. 1. Weng et al.), 65-82, 1990; New
Acta Me tall. Mater., 1995,43, 1335-1354. York, Springer-Verlag.
19. M. FINOT, s. SURESH, C. BULL, and s. SAMPATH: Mater. Sci. Eng., 59. G. J. DVORAK and Y. BENVENISTE: Proc. R. Soc., 1992, A437,
1996, 205A, 59-71. 291-310.
20. K. s. RAVICHANDRAN: Mater. Sci. Eng., 1995, A201, 269-278. 60. Y. BENVENISTE, G. J. DVORAK, and T. CHEN: J. Mech. Phys.
21. J. ABOUDI, M.-J. PINDERA, and s. M. ARNOLD: Int. J. Solids Solids, 1991, 39, 927-946.
Struct., 1994, 31, 1393-1428. 61. J. D. ESHELBY: Proc. R. Soc., 1957, A241, 376-396.
22. L. B. FREUND: J. Mech. Phys. Solids, 1996, 44, 723-736. 62. J. D. ESHELBY: in 'Progress in solid mechanics', (ed. I. N.
23. L. B. FREUND: J. Cryst. Growth, 1993, 132, 341-344. Sneddon and R. Hill), 89-140; 1961, New York, Pergamon
24. F. ERDOGAN: Compos. Eng., 1995, 5, 753-770. Press.
25. S. SAMPATH, H. HERMAN, N. SHIMaDA, and T. SAITO: MRS Bull., 63. L. M. BROWN and w. M. STOBBS: Phi/os. Mag., 1971, 23,
1995, 20, 27-31. 1185-1199.
26. E. A. FITZGERALD, Y.-H. XIE, D. MONROE, P. J. SILVERMAN, J.-M. 64. T. MaRl and K. TANAKA: Acta Metall., 1973, 21, 571-574.
KUO, A. R. KORTAN, F. A. THIEL, B. E. WEIR, and L. C. 65. Y. BENVENISTE: Mech. Mater., 1987, 6, 147-157.
FELDMAN: J. Vac. Sci. Technol. B, 1992, 10, 1807-1819. 66. G. P. TANDON and G. J. WENG: Polym. Compos., 1984,5,327-
27. and F. ERDOGAN: 'Fracture of functionally graded
Y. D.~ LEE 333.
materials', Report no. F49620-93-1-0252, 1994, Bethlehem, 67. K. WAKASHIMA, H. TSUKAMOTO, and B. H. CHOI: in 'The Korea-
PA, Lehigh University. Japan metals symp. on composite materials', 1988, 102-115;
28. Y. YANG and D. MUNZ: in 'Fracture mechanics', (ed. W. G. Seoul, The Korean Institute of Metals.
Reuter et al.), STP 1256; 1997, Philadelphia, PA, American 68. K. WAKASHlMA and H. TSUKAMOTO: in Proc. 1st Int. Conf. on
Society for Testing and Materials, in press. 'Functionally gradient materials', (ed. M. Yamanouchi et al.),
Published by Maney Publishing (c) IOM Communications Ltd

29. A. S. KIM, S. SURESH, and c. F. SHIH: Int. J. Solids Struct., 1997, 19-26; 1990, Tokyo, FGM Forum.
in press. 69. 1. D. ATKINSON, L. M. BROWN, and w. M. STOBBS: Phi/os. Mag.,
30. B. BEARDSLEY: in Proc. of Workshop on 'Coatings for advanced 1974, 26, 1247-1280.
engines', Castine, ME, 1990, (ed. E. W. Gregory and E. E. 70. L. M. BROWN and D. R. CLARKE: Acta Metall., 1977,25,563-570.
Wells), US Department of Energy, Washington, DC, USA, 71. H. LILHOLT: Acta Metall., 1977, 25, 571-585.
11-53-11-56. 72. o. B. PEDERSEN: Acta Metall., 1990, 38, 1201-1219.
31. R. C. NOVAK, A. P. MATARESE, R. P. HUSTON, A. 1. SCHARMAN, 73. P. J. WITHERS, W. M. STOBBS, and o. PEDERSEN: Acta Metall.,
and T. M. YONUSHONIS: Mater. Manuf Proc., 1992,7, 15-30. 198~3~3061-308~
32. A. E. GIANNAKOPOULOS and s. SURESH: Int. J. Solids Struct., 74. J. ABOUDI: Appl. Mech. Rev., 1989, 42, 193-221.
1997,34, (19), 2357-2428. 75. S. NEMAT-NASSER, T. IWAKUMA, and M. HEJAZI: Mech. Mater.,
33. S. SURESH, A. E. GIANNAKOPOULOS, and J. ALCALA: Acta Metall. 1982, 1, 239-267.
Mater., 1997, 45, (5), 1307-1321. 76. T. CHRISTMAN, A. NEEDLEMAN, and s. SURESH: Acta Metall.,
34. O. J0RGENSEN, A. E. GIANNAKOPOULOS, and s. SURESH: Int. J. 198~3~3029-305Q
Solids Struct., 1997, in press. 77. v. TVERGAARD: Acta Metall. Mater., 1990, 38, 185-196.
35. D. HULL and T. W. CLYNE: 'An introduction to composite 78. G. BAa, R. M. McMEEKING, and J. W. HUTCHINSON: Acta Metall.
materials', 2nd edn; 1996, Cambridge, Cambridge University Mater., 1991,39, 1871-1882.
Press. 79. J. LLORCA, A. NEEDLEMAN, and s. SURESH: Acta Metall. Mater.,
36. K. K. CHAWLA: 'Introduction to composites'; 1992, Berlin, 1991, 39, 2317-2335.
Springer- Verlag. 80. A. LEVY and J. M. PAPAZIAN: Metall. Trans., 1991, 22A, 411-
37. T. W. CLYNE and P. J. WITHERS: 'An introduction to metal 420.
matrix composites'; 1993, Cambridge, Cambridge University 81. Y.-L. SHEN, M. FINOT, A. NEEDLEMAN, and s. SURESH: Acta
Press. Metall. Mater., 1994, 42, 77-97.
38. T. MURA: 'Micromechanics of defects in solids'; 1987, The 82. Y.-L. SHEN, M. FINOT, A. NEEDLEMAN, and s. SURESH: Acta
Hague, The Netherlands, Nijhoff. Metall. Mater., 1995, 43, 1701-1722.
39. M. TAYA and R. J. ARSENAULT: 'Metal matrix composites: 83. J. R. BROCKENBROUGH, S. SURESH, and H. A. WIENECKE: Acta
Thermomechanical behaviour'; 1989; Oxford, Pergamon Metall. Mater., 1991, 39, 725-752.
Press. 84. c. L. HOM: J. Mech. Phys. Solids, 1992,40,991-1008.
40. A. KELLY: 'Strong solids'; 1966, Oxford, Oxford University 85. M. B. BUSH: Mater. Sci. Eng., 1992, A154, 139-152.
Press. 86. M. FINOT, Y.-L. SHEN, A. NEEDLEMAN, and s. SURESH: Metall.
41. s. SURESH, A. MORTENSEN, and A. NEEDLEMAN (Eds): Mater. Trans, 1994, 25A, 2403-2420.
'Fundamentals of metal matrix composites'; 1993, Stoneham, 87. R. M. CHRISTENSEN: 'Mechanics of composite materials'; 1979,
MA, Butterworth-Heinemann. New York, Wiley.
42. J. D. EMBURY and J. L. DUNCAN: J. Met., 1982,34, (4), 24-29. 88. z. HASHIN and s. SHTRIKMAN: J. Mech. Phys. Solids, 1962,
43. w. VOIGHT: 'Lehrbuch der Kristallphysik'; 1928, Stuttgart, 10, 335-342.
Germany, B. G. Teubner. 89. E. KRONER: Z. Phys., 1958, 151, 504-509.
44. A. REUSS: Z. ang. Math. Mech., 1929,9,49-54. 90. B. BUDIANSKY: J. Mech. Phys. Solids, 1965, 13, 223-238.
45. R. HILL: J. Mech. Phys. Solids, 1963, 11, 357-372. 91. Y.-L. SHEN, A. NEEDLEMAN, and s. SURESH: Metall. Mater. Trans.,
46. J. C. HALPIN and s. w. TSAI: 'Details of composites' Report 1994, 25A, 839-850.
AFML- TR-67-423, 1967, Dayton, OH, Air Force Materials 92. J. W. HUTCHINSON and R. M. McMEEKING: in 'Micromechanics
Laboratory. of defects in solids', (ed. S. Suresh et al.), Chap. 9, 158-173;
47. Y. TOMOTA, K. KUKORI, T. MaRl, and I. TAMURA: Mater. Sci. 1993, Boston, MA, Butterworth - Heinemann.
Eng., 1976, 24, 85-91. 93. s. SURESH and J. R. BROCKENBROUGH: in 'Micromechanics of
48. K. CHO and J. GURLAND: Metall. Trans., 1988, 19A, 2027- defects in solids', (ed. S. Suresh et al.), Chap. 10, 174-190;
2040. 1993, Boston, MA, Butterworth - Heinemann.
49. H. FISCHMEISTER and B. KARLSSON: Z. Metallk., 1977, 68, 94. T. NAKAMURA and s. SURESH: Acta Metall. Mater., 1993, 41,
311-327. 1665-1681.
50. S. ANKEM and H. MARGOLIN: Metall. Trans., 1986, 17A, 95. H. J. BOHM and F. G. RAMMERSTORFER: Mater. Sci. Eng., 1991,
2209-2223. A135, 185-190.
51. H. L. cox: Br. J. Appl. Phys., 1952, 3, 73-79. 96. E. WEISSENBEK and F. G. RAMMERSTORFER: Acta Metall. Mater.,
52. v. C. NARDONE and K. M. PREWO: Scr. Metall., 1986,20,43-48. 1993, 41, 2833-2843.
53. T. W. CLYNE: Mater. Sci. Eng., 1989, A122, 183-192. 97. T. SIEGMUND, E. WERNER, and F. D. FISCHER: J. Mech. Phys.
54. H. J. BOHM: 'Thermoelastic deformation of composites', Report Solids, 1995, 43, 495-515.
108, Vienna, Austria, The Christian Doppler Laboratory, 98. M. H. POECH, H. F. FISCHMEISTER, D. KAUTE, and s. SCHMAUDER:
1994. Compos. Mater. Sci., 1993, 1, 213-230.

International Materials Reviews 1997 Vol. 42 NO.3

Suresh and Mortensen Functionally graded metals: Part 2 115

99. J. TEPLYand G. J. DVORAK:J. Mech. Phys. Solids, 1988, 36, 136. J. ABOUDI,M.-J. PINDERA,and s. M. ARNOLD:Compos. Eng.,
29-58. 1995, 5, 771-792.
100. w. J. POOLE,1. D. EMBURY, S. MacEWEN,and u. F. KOCKS:Philos. 137. I. C. NOYANand J. COHEN: 'Residual stresses'; 1985, New
Mag., 1994, 69, 645-665. York, Elsevier.
101. w. J. POOLE,J. D. EMBURY,S. MacEWEN,and u. F. KOCKS:Philos. 138. J. LU and M. R. JAMES(Eds): 'Handbook of measurement of
Mag., 1994,69, 667-687. residual stresses'; 1996, Lilburn, GA, Fairmount Press.
102. A. K. VASUDEVAN, O. RICHMOND,F. ZOK, and J. D. EMBURY: 139. D. J. GREVING,E. F. RYBICKI,and J. R. SHADLEY:in Proc. 7th
Mater. Sci. Eng., 1989, A107, 63-69. National Thermal Spray Conf., Boston, MA, 647-653; 1994,
103. J. J. LEWANDOWSKI,C. LIU, and w. HUNT: Mater. Sci. Eng., Thermal Spray Society.
1989, A107, 241-255. 140. P. BIALUCKI,W. KACZMAR,and J. GLADYSZ:in 'Advances in
104. A. NEEDLEMAN,S. R. NUTT, S. SURESH,and v. TVERGAARD:in thermal spray processing', 837-842; 1986, New York,
'Micromechanics of defects in solids', (ed. S. Suresh et al.), Pergamon Press.
Chap. 13, 233-250; 1993, Boston, MA, Butterworth - 141. M. K. HOBBSand H. REITER:in 'Advances in coatings tech-
Heinemann. nology', 285-290; 1988, Metals Park, OH, American Society.
105. N. A. FLECKand J. W. HUTCHINSON:J. Mech. Phys. Solids, 1993, for Metals.
41, 1825-1848. 142. s. C. GILLand T. W. CLYNE:Metall. Trans., 1990, 21B, 377-385.
106. G. I. TAYLOR:J. Inst. Met., 1938,62, 307-324. 143. T. W. CLYNEand s. c. GILL: J. Therm. Spray Te c111l 0 1., 1996,
107. R. HILLand 1. R. RICE:J. Mech. Phys. Solids, 1972,20,401-424. 5,401-435.
108. R. J. ASAROand J. R. RICE: J. Mech. Phys. Solids, 1977, 144. c.-c. CHIU and E. D. CASE:Mater. Sci. Eng., 1991, A132, 39-
25, 309-330. 50.
109. R. J. ASARO:J. Appl. Mec11., 1983, 50,921-936. 145. c.-c. CHIU: Mater. Sci. Eng., 1992, A150, 139-147.
110. D. PEIRCE,R. J. ASARO,and A. NEEDLEMAN: Acta Metall., 1983, 146. T. P. WEIHS,S. HONG,J. C. BRAVMAN,and w. D. NIX:in 'Stresses
31, 1951-1968. in thin films', (ed. M. Doernen et al.), Mater. Res. Soc. Symp.
111. s. V. HARREN,H. DEVE,and R. J. ASARO:Acta Metall., 1988, 36, Proc., 1989, 130, 87-92.
Published by Maney Publishing (c) IOM Communications Ltd

2435-2455. 147. P. A. FLINN, D. S. GARDNER,and w. D. NIX: IEEE Trans.

112. P. E. McHUGH,R. J. ASARO,and c. F. SHIH: Acta Metall. Mater., Electron Devices, 1987, 34, 689-696.
1993, 41, 1461-1500. 148. O. KESLER,M. FINOT,S. SURESH,and s. SAMPATH:Acta Metall.
113. A. NEEDLEMANand v. TVERGAARD:J. Appl. Mech., 1993, 60, Mater., 1997, in press.
70-76. 149. M. w. HYER:J. Compos. Mater., 1981, 15, 175-194.
114. A. M. CUITINOand M. ORTIZ: Model. Simul. Mater. Sci. Eng., 150. M. w. HYER:J. Compos. Mater., 1982, 16, 318-340.
1992, 1, 225-250. 151. c. B. MASTERSand N. J. SALAMON:Int. J. Eng. Sci., 1993,
115. J. BASSANI:Adv. Appl. Mech., 1994, 30, 191-239. 31, 915-925.
116. H. H. M. CLEVERINGA,E. Van der GIESSEN,and A. NEEDLEMAN: 152. N. J. SALAMONand c. B. MASTERS:Int. J. Solids Struct., 1994,
Acta Metall. Mater., 1997, in press. 32, 473-481.
117. E. Van der GIESSENand A. NEEDLEMAN: Model. Simul. Mater. Sci. 153. G. G. STONEY:Proc. R. Soc., 1909, A82, 172-175.
Eng., 1995, 3, 689-699. 154. v. L. HEIN and F. ERDOGAN:Int. J. Fract. Mech., 1971, 7,
118. D. C. DUNANDand A. MORTENSEN: Acta Metall. Mater., 1991, 317-330.
39, 1405-1416. 155. F. ERDOGAN:MRS Bull., 1995,20, 43-44.
119. E. EARHARTand A. MORTENSEN:Unpublished results, MIT, 156. D. MUNZand Y. Y. YANG:Int. J. Fract., 1993,60, 169-177.
Cambridge, MA, 1997. 157. Y. ARAI,A. E. GIANNAKOPOULOS, and s. SURESH:Unpublished
120. s. P. TIMOSHENKO and s. WOINOWSKY-KRIEGER:'Theory of plates results, MIT, Cambridge, MA, 1997.
and shells'; 1959, New York, McGraw-Hill. 158. R. L. WILLIAMSON,B. H. RABIN, and G. E. BYERLY:Compos.
121. M. FINOTand s. SURESH:'MultiTherm, Version 1.4, A personal Eng., 1995, 5, 851-863.
computer software for the thermomechanical analysis of the 159. R. L. WILLIAMSONand B. H. RABIN: Ceram. Trans., 1993,
layered and graded materials', Copyright The Massachusetts 34,55-65.
Institute of Technology, Cambridge, MA, 1994. 160. M. L. KANNINENand K. POPELAR:'Advanced fracture mech-
122. T. HIRANO,J. TERAKI,and T. YAMADA:in Proc. 1st Int. Symp. anics'; 1985; Oxford, Oxford University Press.
on 'Functionally gradient materials', Send ai, Japan, 1990, (ed. 161. N. KONDA and F. ERDOGAN:Eng. Fract. Mech., 1994, 47,
M. Yamanouchi et al.), Functionally Gradient Materials 533-545.
Forum and the Society of Non-Traditional Technology, 5-10. 162. F. ERDOGAN:J. Appl. Mech. (Trans. ASME), 1985,52,823-828.
123. T. HIRANOand K. WAKASHIMA: MRS Bull., 1995, 20, 40-42. 163. F. DELALEand F. ERDOGAN:J. Appl. Mech. (Trans. ASME),
124. K. WAKASHIMA,M. OTSUKA, and s. UMEKAWA:J. Compos. 1983, 50, 609-614.
Mater., 1974, 8, 391-400. 164. F. DELALEand F. ERDOGAN:J. Appl. Mech. (Trans. ASME),
125. J. TERAKI,T. HIRANO,and K. WAKASHIMA: Ceram. Trans., 1993, 1988, 55, 317-324.
34,67-74. 165. F. ERDOGAN,A. C. KAY A, and P. F. JOSEPH:J. Appl. Mech. (Trans.
126. Y.-L. SHENand s. SURESH:J. Mater. Res., 1995, 10, 1200-1215. ASME), 1991, 58, 400-418.
127. M. FINOT and s. SURESH:J. Mech. Phys. Solids, 1996, 44, 166. F. KROUPA,Z. KNESL,and J. VALACH:Acta Tech. CSAV, 1993,
683-721. 38,29-74.
128. B. H. RABIN, R. H. WILLIAMSON,and s. SURESH:MRS Bull., 167. Y. R. TAKEUCHIand K. KOKINI: J. Eng. Gas Turbines Power
1995, 20, 37-39. (Trans. ASME), 1994, 116,266-271.
129. M. FINOT,S. SURESH,C. BULL,and s. SAMPATH:in Proc. 3rd Int. 168. K. KOKINI and Y. R. TAKEUCHI:in 'Fracture mechanics'
Symp. on 'Structural and functional gradient materials', (ed. (ed. F. Erdogan and R. J. Hartranft), STP 1220, 249-263;
B. Ilschner and N. Cherradi), Lausanne, Switzerland, 223-228, 1995, Philadelphia, PA, American Society for Testing and
1994; Lausanne, Presses Poly techniques et Universitaires Materials.
Romandes. 169. R. A. MILLER and c. c. BRENDT: Thin Solid Films, 1984,
130. B. BOLEYand J. H. WEINER:'Theory of thermal stresses'; 1960, 119, 195-202.
New York, Wiley. 170. R. C. BRINK:J. Eng. Gas Turbines Power (Trans. ASME), 1989,
131. E. WEISSENBEK,H. PETTERMANN,and s. SURESH:Acta Me ta ll. 111, 570-577.
Mater., 1997, in press. 171. T. R. GURNEY:'Fatigue of welded structures'; 1968, Cambridge,
132. A. C. GAVAZZIand D. c. LAGOUDAS:Compos. Mech., 1990, Cambridge University Press.
7,13-20. 172. w. CHENG,H. S. CHENG,T. MURA,and L. M. KEER:J. Tribology
133. H. E. PETTERMANN,H. J. BOHM,and F. G. RAMMERSTORFER: In (Trans. ASME), 1994, 116,2-8.
Proc. of 'General workshop in materials science and engineer- 173. K. L. JOHNSON: 'Contact mechanics'; 1985, Cambridge,
ing', Sept.-Oct. 1996, (ed. M. Rappaz), European Commission, Cambridge University Press.
in press. 174. A. KLEEMANN:'Literature review of contact fatigue, Part II:
134. M. DAO, P. GU, A. MAEWAL,and R. J. ASARO:Acta Metall. Calculation models', Report no.IM-3224, Swedish Institute
Mater., 1997, in press. for Metal Research, Stockholm 1993.
135. 1. ABOUDI:'Mechanics of composite materials - a unified 175. S. SURESH,Y. SUGIMURA,and E. K. TSCHEGG:Ser. Metall. Mater.,
micro mechanical approach', 1991, New York, Elsevier. 1992, 27, 1189-1194.

International Materials Reviews 1997 Vol. 42 NO.3

116 Suresh and Mortensen Functionally graded metals: Part 2

176. S. SURESH, Y. SUGIMURA, and T. OGAWA: Scr. Metall. Mater., 184. M. Y. HE, P. STAHLE, C. F. SHIH, N. T. ZHANG, and R. M.
1993, 29, 237-242. McMEEKING: Unpublished results, University of California,
177. Y. SUGIMURA, P. G. LIM, C. F. SHIH, and s. SURESH: Acta Metall. Santa Barbara, CA, 1996.
Mater., 1995, 43, 1157-1169. 185. A. R. ZAK and M. L. WILLIAMS: J. Appl. Mech. (Trans. ASME),
178. Y. SUGIMURA, L. GRONDIN, and s. SURESH: Scr. Metall. Mater., 1963, 30, 142-143.
1995, 33, 2007-2012. 186. F. ERDOGAN and v. BIRICIKOGLU: Int. J. Eng. Sci., 1973, 11,
179. J. R. RICE: J. Appl. Mech. (Trans. ASME), 1968, 35, 379-386. 745-766.
180. A. S. KIM: Doctoral thesis, Brown University, Providence, RI, 187. J. L. BEUTH, Jr: Int. J. Solids Struct., 1992, 29, 1057-1069.
USA, 1996. 188. K. TOKAJI, T. OGAWA, H. SHIBATA, and Y. KAMIYA: Trans. Jpn
181. A. ROMEO and R. BALLARINI: J. Appl. Mech. (Trans. ASME), Soc. Mech. Eng., 1991, 57, 2293-2296 (in Japanese).
in press. 189. s. SURESH: 'Fatigue of materials'; 1991, Cambridge, Cambridge
182. M. Y. HE and J. W. HUTCHINSON: Int. J. Solids Struct., 1989, 25, University Press.
1053-1069. 190. J. D. EMBURY,L. B. FREUND, A. NEEDLEMAN, C. F. SHIH, F. SPAEPEN,
183. J. W. HUTCHINSON and z. suo: Adv. Appl. Mech., 1992, 29, and s. SURESH: J. Mech. Phys. Solids, 1996, 44, 823-825.
Published by Maney Publishing (c) IOM Communications Ltd

International Materials Reviews 1997 Vol. 42 NO.3