An Examination of the

Mori-Tanaka Effective Medium

Approximation for Multiphase
Composites
The Mori-Tanaka method is considered in the context of both scalar thermal con-
ductivity and anisotropic elasticity of multiphase composites, and some general pro-
A. N. Morris perties are deduced. Particular attention is given to its relation to known general
Department of Mechanics and bounds, and to the differential scheme. It is shown that the moduli predicted by the
Materials Science, method always satisfy the Hashin-Shtrikman and Hill-Hashin bounds for two-phase
Rutgers University, composites. This property does not generalize to multiphase composites. A specific
Piscataway, N.J. 08855-0909 example illustrates that the method can predict moduli in violation of the Hashin-
Assoc. Mem. ASME Shtrikman bounds for a three-phase medium. However, if the particle shapes are all
spheres, then the prediction for the multiphase composite is coincident with the
Hashin-Shtrikman bounds if the matrix material is either the stiffest or the most
compliant phase. It is also shown that the generalized differential effective medium
method yields the same moduli as the Mori-Tanaka approximation if certain condi-
tions are satisfied in the differential scheme. Thus, it is required that at each stage in
the differential process, and for each phase j Q = l, 2, . . . , n) of new material, the
average field in the incrementally added phase j material must be the same as the
average field in the bulk phase j . For two phase media, n = l, this condition reduces
to the less stringent requirement that the ratio of the field in the incrementally added
material to the average field in the matrix material is the same as the dilute concen-
tration ratio. The cumulative findings of this paper, particularly those concerning
bounds, suggest that the Mori-Tanaka approximation be used with caution in
multiphase applications, but is on firmer ground for two-phase composites.

1 Introduction
Various approximate methods exist for predicting the ef- The Mori-Tanaka approximation is another method that
fective thermal, electrical, and mechanical properties of has received attention recently. It is based upon the original
composites. Among these are the self-consistent scheme, and work of Mori and Tanaka (1973), and has been used to advan-
the differential scheme (Cleary et al., 1980, McGlaughlin, tage by, for example, Taya and Mura (1981) and Taya and
1977). These effective medium approximations do not require Chou (1981). Weng (1984) applied the Mori-Tanaka method
detailed statistical information of the microstructure, but can to the effective medium problem for a two-phase composite
distinguish between different inclusion shapes. Therefore, with spherical inclusions. Further applications have been given
such schemes can be useful for statistically homogeneous com- by Benveniste (1986a,b; 1987a,b,c) for the thermal conductivi-
posites with known inclusion shapes. However, there is always ty and mechanical properties of two-phase and multiphase
some doubt as to their utility. For example, it is not obvious, a media. Unlike most other approximate methods which require
priori, whether the results will automatically satisfy known solving implicit equations numerically, the Mori-Tanaka
bounds on the moduli, such as those of Hashin and Shtrikman method yields explicit, closed-form answers for the effective
(1963). At the present time, several methods, including the dif- properties. As with all other effective medium methods, it
ferential scheme, are known to correspond to realizable hinges upon a mathematical approximation, explained in the
media, and hence satisfy the bounds (Avellaneda, 1987). following sections. A significant property was discovered by
Weng (1984), who showed that the Mori-Tanaka method with
spherical inclusions of the softer (harder) phase gives the
Contributed by the Applied Mechanics Division of THE AMERICAN SOCIETY OF
MECHANICAL ENGINEERS for publication in the JOURNAL OF APPLIED
Hashin-Shtrikman upper (lower) bounds for the bulk and
MECHANICS. shear moduli. Norris (1985) pointed out that randomly-
Discussion on this paper should be addressed to the Editorial Department, oriented disk-shaped particles of the softer (harder) phase
ASME, United Engineering Center, 345 East 47th Street, New York, N.Y. yields the lower (upper) bounds. Benveniste (1987c) has
10017, and will be accepted until 2 months after final publication of the paper recently proved, using a clever argument, that the bulk and
itself in the JOURNAL OF APPLIED MECHANICS. Manuscript received by ASME
Applied Mechanics Division October 27, 1987; final revision, February 5, 1988. shear modulus predicted by Mori-Tanaka for a two-phase

Journal of Applied Mechanics MARCH 1989, Vol. 56/83

Copyright © 1989 by ASME
Copyright © 1998 by ASME
Downloaded From: http://appliedmechanics.asmedigitalcollection.asme.org/ on 04/06/2014 Terms of Use: http://asme.org/terms
 composite with randomly-oriented ellipsoidal particles will lie
within the Hashin-Shtrikman bounds. This benign property
had previously been noted from the numerical results of Tan-
don (1986) and Maewal and Dandekar (1987). However, as we
show in this paper, the Mori-Tanaka method can give results
for multiphase media that are in violation of the Haskin-
Shtrikman bounds. It appears, therefore, that two-phase com-
posites are a special case for the method.
The purpose of this paper is to examine the connection be-
tween the Mori-Tanaka approximation and known bounds.
We also relate the approximation to the generalized differen-
tial scheme. One common consequence from both studies is
that the two-phase medium is a special case. Both thermal and
elastic properties are considered. The analysis is simpler for
the scalar thermal conductivity problem, and is presented in
Section 2. The theory for the elastic moduli is presented in Sec-
tion 3, where the major results concerning bounds are derived.
The connection with the differential scheme is explored in Sec-
tion 4.
tained from equation (5) by assuming that the ratio H'J)/Hi0)
is equal to the ratio for a single, isolated grain of phase j
embedded in uniform matrix material of infinite extent.
Equivalently, the ratio HU) /ff0) is taken to be the ratio per-
taining in the dilute limit of c« 1, where c is the total volume
fraction of the added phases,
c «c = 1 c
= E ; - o-
The ratio H^/H*® for randomly-oriented ellipsoidal par-
ticles of phase j is described by three depolarization coeffi-
cients jS], 0 2 , and 0 3 for each of the major axes (see Landau
and Lifshitz (1960) for explicit formulae). These coefficients
are between 0 and 1, and satisfy 0j + 0 2 + 0 3 = 1. They each
equal 1/3 for spherical particles; long circular cylinders
(needles) have 0 i = O , 0 2 = 0 3 = l/2, and thin, circular disks
have 0! = 0 2 = 0, 0 3 = 1. In general,

2 The Effective Thermal Conductivity of a Multiphase ""'"•--l-l; ('•-&£•)"'• »

Isotropic Composite Benveniste (1986a) observed that if k0 is smaller (larger) than
2.1 General Equations and Definitions. Consider an n + 1 all of the other kj, j=\, 2 n, then the Mori-Tanaka
phase composite made of isotropic constituents with thermal method with Hu)/H^0) for spherical particles j= 1,2,. . . , n,
conductivities kh i=0, 1, 2, . . . , n, and occupying total gives the Hashin-Shtrikman lower (upper) bound on the effec-
n tive conductivity.
volume fractions c,-, such that E ct = \. Phase i=0 cor- On the other hand, if all the particles are in the shape of
/=o thin, circular disks, the Mori-Tanaka scheme gives, from
responds to the matrix material. The temperature field <j>(x) equations (5) and (6),
and the normal component of the heat flux, q«n, where q is
the flux and n the unit normal, are both continuous across in-
terfaces between the constituents. The heat flux and
temperature field in phase i are related by
where
q(/>=/t,H<i», (1) n n

where k = E cJkJand * ^ = E cikr' •

H(i» = -V</> (;) . (2)
For a two-phase composite, « = 1, in which kQ<kx (k0>kl),
Assuming an isotropic distribution of grains, the effective the prediction of equation (7) corresponds to the Hashin-
conductivity is isotropic and equal to &(eff), defined by the Shtrikman upper (lower) bound on the effective conductivity.
macroscopic relation This result does not generalize to the multiphase composite,
n> 1. In fact, equation (7) can violate the Hashin-Shtrikman
q = /fc<eff>H. (3) bounds for n > 1. For example, k of (7) exceeds the upper
an c c 4
An overbar denotes the spatial average of a quantity : Thus, H bound if n = 2, k{ = 2k0, k2 = 3fc0 d i = 2 = 0- - This viola-
(
is the average of H over the entire composite, and H " is the tion indicates that the Mori-Tanaka scheme is not always
average of H ( / ) in phase /. The average H could be imposed, realizable for multiphase composites.
for example, by the boundary condition that <j> = — H«x on the It is possible to show for two-phase composites that the
exterior surface of the composite. Under the assumption of Mori-Tanaka k of equations (5) and (6) satisfies the Hashin-
macroscopic isotropy, the vectors H, H ( / ) , can be replaced by Shtrikman bounds. To see this, consider equation (5) with
ii 0
scalar quantities H, fH'"1, where -&(,) is the component of H ( , ) « = 1 and <x = H '>/H<- K Then,
in the direction of H. The effective conductivity follows from
equations (1) and (3) as
n
Thus, k is maximum (minimum) for a maximum (minimum).
kmH= E k,cfiw. (4) Now consider a. of equation (6) as a function of 0!, 02> a n d 03
(=0 constrained to the interior and surface of the tetrahedron
0!+0 2 + 05 = 1. It is easily shown that a(0!, 0 2 , 03) attains
Equation (4) can be rewritten stationary values at the four vertices, corre-
n sponding to plate-like particles, at (1/3, 1/3, 1/3), which is a
E (kj-k0ycjHM/m sphere, and at the points (1/2, 1/2, 0), (1/2, 0, 1/2), and (0,
km =k0+ >=1 . (5) 1/2, 1/2) which are needles. If kl>k0, then a(sphere) <
a(needles) < a(plate). But the sphere and plate values of a
c0+£>,.#«>/#<» correspond to the Hashin-Shtrikman bounds on k, therefore,
i= l
all other a give intermediate results.
This is an exact equation for the effective conductivity £*ff),
but is complicated by the determination of the ratios Hu) /H^ 3 The Effective Elastic Moduli of a Multiphase
for each of the added phases j= 1, 2, . . . , «. Composite
2.2 The Mori-Tanaka Scheme and Hashin-Shtrikman 3.1 The Mori-Tanaka Approximation. Let L be the effec-
Bounds. The Mori-Tanaka effective conductivity k is ob- tive elastic modulus tensor for a multiphase composite, each

84/Vol.56, MARCH 1989 Transactions of the ASME

Downloaded From: http://appliedmechanics.asmedigitalcollection.asme.org/ on 04/06/2014 Terms of Use: http://asme.org/terms

 phase, j = 0, 1, 2, . . . , n characterized by a fourth-order spherical particles coincides with the lower (upper) Hashin-
modulus tensor L,-. The anisotropic Hooke's law for each Shtrikman bounds on K and n if KJ>K0, and ^,->/t0> j= 1, 2,
phase is . . . , n (KJ<KQ, Hj<n0, j=\, 2, . . . , ri). This equivalence
has been previously noted by Weng (1984) for a two-phase
ff,=L,e,-, (9) composite.
where e,= l/2[Vu,+ (Vu,-)r] is the strain. Let the average While still on the topic of isotropic constituents, we note
strain in phases / and 0 be related by that if all the particles are randomly-oriented disks, then T,0
can still be expressed in the form of equation (12), but Sj0 now
e,=f, 0 e 0 , #=1,2, . . . ,n, (10) depends upon L ; ,
(eff)
then an exact expression for the effective moduli L follows (17)
in a manner analogous to the derivation of equation (6), as S / , = (I+L;L 0 - 1 )-
The Mori-Tanaka effective moduli K and /i reduce in the
L(ef0
= Lo + ( t C)(h~LoJtjo) [c„I + t c/t/c
special case of a two-phase composite to the Hashin-
(11)
V=i ' 1=1 L Shtrikman lower (upper) bounds if K, <K0 and nt </t0 (jq > K0,
Here I is the fourth-order isotropic indentity tensor, I = (l, 1) /i! >n0). As for the thermal conductivity, this result does not
in the concise notation of Hill (1965). In the same notation, an generalize to multiphase composites.
isotropic stiffness tensor is L = (3 K, 2JX), where K and pi are bulk If all the constituents are aligned, transversely isotropic
_1
and shear moduli, the compliance tensor is M = L = (1/3K, phases, the moduli tensors can be represented succinctly in the
1/2^,), and tensor products are LLL2 = (9K1K2, 4/u.1/Lt2)- notation of Walpole (1969) as L = (2Ar, /, n, 2m, 2p),
1
In the Mori-Tanaka method T,0 is approximated by the M = L - = l / 2 ( n / 7 , -l/y, 2k/y, 1/m, l/p), where
analogous quantity for an isolated particle of phase i in an in- y=kn — P- = kE, and E is the axial Young's modulus. For ex-
finite matrix of phase 0. Equivalently, the dilute limit value of ample, if x3 is the symmetry axis, then k = Cn - C66, /= C13,
t, 0 is taken. Then equation (1) provides an explicit equation « = C33, m = C66, andp = C44. Positive definiteness requires
for the Mori-Tanaka effective moduli L. In particular, if the that k,m,p, and n — P/k are each positive. The bounds of Hill
particles of phase / are ellipsoidally shaped and aligned. (1964) and Hashin (1965) apply to composites made of aligned
Eshelby's results provide t, 0 in the simple form cylindrical fibers of arbitrary transverse geometry. These
bounds have been phrased succinctly by Walpole (1969) for
f,.0 = [I + S;oL0-1(L,.-L0)]- (12) multiphase composites. In this case the tensor L0* is not well-
defined, so it is necessary to work with the compliance tensor
where S,0 depends only upon L0 and the aspect ratios of the M. The dual equation to (14) is
particle of phase / (see, for example, Mura (1982)). The Mori-
Tanaka method as defined by equation (11) and the approx-
imation (12) has been called the "direct approach" by M v (18)
Benveniste (1987c). He showed it is identical for two-phase y=o '
composites to the usual "equivalent inclusion-average stress" where Mg = L§ , when the latter is defined. If (M,-M 0 ) is
formulation of, for example, Weng (1984). The equivalence of positive (negative) definite for all i= 1, 2, . . . , n, then the
the two formulations for multiphase composites is lower (upper) bound, M*, on M(eff) is given by equation (18),
demonstrated in the Appendix. with M8 = M0*, where M0* is
3.2 Bounds on the Elastic Moduli. If all particle shapes 1 (/ 1 „_„. 1 2 1 \
are identical, and the particles are aligned, then the Eshelby Mi (19)
2 (\m0 ,0,0, m0 + K_0 _ , p_0-) .
tensor S,0 is the same for each phase /= 1, 2 n. Define Note that Walpole's\m(1969) n Momncontains
kn a ptypographical
n/ er-
L°0 b y ror: The correct expression follows from Laws (1974).
T 0 _ T
u
•L'n
0 —M — J C -
• >o°°l/00 ^0- (13) The Eshelby tensor S,0 corresponding to equation (19) is
that of a circular, cylindrical particle. Therefore, the Hill-
Then the Mori-Tanaka effective moduli become Hashin bounds for a multiphase tranversely isotropic fibrous
composite of arbitrary transverse geometry corresponds to the
(14) Mori-Tanaka approximation with circularly cylindrical par-
ticles. This equivalence has been noted by Tandon (1986) for a
Walpole (1966) obtained lower (upper) bounds on L in two-phase composite.
(eff)

the form of equation (14) under the assumption that (L,- — L0) We next develop general results relating the Mori-Tanaka
is positive (negative) definite for all i=l, 2 n. The method to the bounds discussed above. The procedure
adopted is a generalization of Benveniste's (1987c). Returning
bounding modulus tensor is L*, defined by (14), with Lg = L0*, to the general assumption that the particles are all identically
where L0* depends upon the particular type of anisotropy of shaped and aligned, then equation (14) is correct to first-order
the composite. in c in the dilute limit of c« 1. Thus,
If all the constituents are isotropic, then the well-known
Hashin-Shtrikman (1963) bounds for a macroscopically n

isotropic composite are defined by L<eff)~L = L 0 +i;c y .(L ; -L 0 )(L y + Lg)-1(L0 + L«) + 0(c2).
L0* = (3K0*,2/i0*), (15)
(20)
where
For a variation SLjj of Lg in this equation, the corresponding
(16a) variation in L is
• / * ; >

n
V-j
5L=£c J .(L ; -L 0 )(L,. + L«)-15L8(Ly + L»)-'
V-I = (16ft)
L K; + 2lt; J
X(L,.-L 0 ) + 0(c2). (21)
The corresponding Eshelby tensor S,0 is for a spherical parti-
cle. Thus, we have that the Mori-Tanaka approximation for Thus, a positive (negative) definite change in L§ yields a cor-

Journal of Applied Mechanics MARCH 1989, Vol. 56/85

Downloaded From: http://appliedmechanics.asmedigitalcollection.asme.org/ on 04/06/2014 Terms of Use: http://asme.org/terms

 responding positive (negative) definite change in L. By It is preferable to work with the incremental volume frac-
assumption, the dilute limit effective moduli satisfy the tions dcj, rather than with the incremental volumes dvj. Since
bounds for the particular type of microstructure considered. the total volume Vremains fixed, it is possible to show (Norris
Therefore, if L* represents the lower (upper) bounds, then et al., 1985) that
both L - L* and L§ - L0* are positive (negative) definite ten- dv dc
sors of the same anisotropy class. - = dC; + C, (25)
y J J
Now consider the finite concentration case. Variation of Lg
in equation (14) yields and hence,

h + c ^]
n
«L = (L + L8)^c,[(L]/+L8)-1 dk= L (kj-k)
J + - (26)
y=o p[ H(t)
- ( L + L8)-1]fiL8[(L/ + L 8 ) - 1 - ( L + Lg)- 1 ](L + L8), (22) This becomes an ordinary differential equation by introducing
a parameter t to describe the evolution of the composite from
which is positive (negative) definite if Sh\ is positive (negative) homogeneous phase 0 with initial conditions k(0) = k0,
definite. It then follows from the dilute limit result for L§ - L0* Cj(0) = 0,7 = 1, 2 n. The volume fraction c could be used
that L of equation (14) satisfies the bounds for finite concen- as the parameter t, for example.
trations. We have thus derived the result that the Mori-Tanaka A rigorous justification for the differential equation (26)
approximation for multiphase composites, with all particles of has been given by Avallaneda (1987). An explicit equation can
the same shape and aligned, satisfies the appropriate Hashin- be obtained, if, for example, the particles are assumed to be in
Shtrikman or Hill-Hashin bounds. This result is not as general the shape of randomly-oriented ellipsoids. Then H^/H is
as it may seem at first, since the only particle shapes that given by the right-hand side of equation (6) with k0 replaced
satisfy these criteria for the Mori-Tanaka approximation are by k(t). Bruggeman's (1935) scheme is contained in (26) as the
spheres for isotropy, and circular cylinders for fiber-rein- special case of a two phase composite, n=l. The effective
forced transverse isotropy. Also, these particular materials medium approximation for multicomponent composites is the
correspond to the Hashin-Shtrikman and Hill-Hashin bounds. limit of (26) as c— 1. Discussions of these and other limiting
Although there is no additional information here concerning cases are contained in Norris et al. (1985) and Norris (1985).
multiphase composites, these results do have significant ap-
plication to the particular case of two-phase composites. 4.1 The Connection With the Mori-Tanaka Approxima-
The Mori-Tanaka approximation for a two-phase com- tion. The field ratio in equation (26) can be written
posite (« = 1) can be written in the form of equation (14) with
HU)/ffO)
LQ — (LI — L 0 )(T 10 —I) (23) (27)
H
where T 10 is the strain concentration ratio for dilute par-
ticulate concentration. Note that L§ of equation (23) is in-
dependent of c 1( and holds for any T 10 . Therefore, by the
same arguments as before, since the dilute limit moduli must Note that the averages HU} and Hu) are generally quite
satisfy the lower (upper) bounds on L (eff) , it follows that distinct and unrelated. The former is the average field in the
LQ - L0* is positive (negative) definite. This in turn implies that incrementally added particles of phased, while the latter is the
L - L * is positive (negative) definite at all concentrations, average field in the entire volume of phase j in the composite.
0 < c , < l . Hence, the Mori-Tanaka moduli for two-phase Define A and A by
composites satisfy the Hashin-Shtrikman or Hill-Hashin
bounds, as appropriate. This general result is unique to two-
phase composites, since it is not generally possible to write the A(t)=—— Y^Cjif^/m, (28)
\-c j=i
Mori-Tanaka approximation for multiphase composites in the
form of equation (14).
Av)=—— Y^CjH^/m\ (29)
4 The Mori-Tanaka Approximation in Terms of the l-c 7=1
Differential Scheme An alternative form of the differential scheme follows from
4.1 The Differential Scheme. The generalized differential equations (27)-(29),
scheme as understood here is a generalization of Bruggeman's
(1935) method to multiphase composites. The present develop-
ment is similar to that of Norris et al. (1985), where it was ap-
[ n

7=1
-.

plied to a special type of three-phase medium in which one of Yjkj-^—H^/H^A

n
the added phases was identical to the original matrix material. + J^ {k-kj) —i—d[HW/ii<®\. (30)
In order to understand the scheme, consider the scalar conduc- ;= i 1—C
tivity problem for a composite at some finite concentrations of
the added p h a s e s 7 = 1 , 2 , . . . , n._ Let Vbe the total volume of This equation is integrable if it is assumed that
the composite, and as before, H is the average of the field HW> W/ffV (t) =H^(0)/HW(Q), 7= 1, 2 , . . . . n, (31)
H(x) in V. Infinitesimal, discrete volumes of the homo-
geneous composite material are then removed and replaced by and
homogeneous amounts of phases 7 = 1 , 2, . . . , n. Thus,
volume dVj is replaced by phase j , such that the replaced H<J)=HV\ 7=1,2, . . . ,n. (32)
volume is perfectly bonded to the composite, and dvj contains Then A=A, and integration of equation (30) subject to the in-
a representative quantity of the existing composite of conduc- itial conditions fc(0) = /c0, c,(0) = 0 , 7 > 1 , gives precisely equa-
tivity k. If H^ is the average of H in dvj, the incremental tion (5) with HW/ff® equal to its dilute concentration value.
change in the effective conductivity is Thus, the differential scheme yields the Mori-Tanaka effective
conductivity if equations (31) and (32) hold.
vV,, , HU) dv. Equations (31) and (32) combined imply that the ratio
(24)
H^ (t)/H^ (t) remains constant and equal to the dilute value

86/Vol. 56, MARCH 1989 Transactions of the ASME

Downloaded From: http://appliedmechanics.asmedigitalcollection.asme.org/ on 04/06/2014 Terms of Use: http://asme.org/terms

 throughout the process. Thus, at each stage of the process, the 5 Discussion and Conclusions
effective k is given by the Mori-Tanaka method for that con-
centration. Equation (32) requires, in addition, that the Our major findings concern the relationship of the Mori-
average field in the incrementally added volumes be the same Tanaka approximation to the Hashin-Shtrikman and Hill-
as the bulk average for that phase. It is interesting to note that Hashin bounds. These bounds will always be satisfied when
(32) is not required a priori for two-phase composites, « = 1, the approximation is used for two-phase composites.
but follows as a consequence of assuming (31). To see this, However, this result does not generalize to multiphase media,
consider the differential equation (26) for the two-phase com- as demonstrated by a counter example in Section 2.
posite, which becomes using (27), We have shown that the Mori-Tanaka method can be for-
mulated in terms of the generalized differential scheme pro-
vided certain conditions are satisfied by the average fields in
the latter. The condition for two-phase media is that the ratio
However, it follows from equation (4) with k(tt!) = k, and of the field in the incrementally added particles to that in the
H=(l-cl)Hi0) + CiH^l\ that bulk matrix phase remains constant as the concentration of
the added phase goes from zero to its final, finite value. The
Mori-Tanaka method, in its "direct approach" formulation,
requires that the ratio of the field in the bulk added phase to
Substituting from equation (34) into equation (33) gives that in the matrix equals its dilute concentration value. It is
perhaps surprising that the condition on the incremental field
ratio produces the same results as the Mori-Tanaka condition
for two-phase media. This is particularly so since the same
This equation can be integrated directly if (31) holds fory'= 1. condition does not suffice to yield the Mori-Tanaka results for
The equality Hil'> = Hil) then follows by substituting the multiphase media.
resulting k into (34). The relationship of the Mori-Tanaka method with the dif-
ferential scheme offers an alternative way of looking at the
former. However, it is not clear whether the relevant condi-
tions, (31)-(32) or (37)-(38), can be realized by specific
4.3 Differential Effective Medium Theory for Multiphase microgeometries. The answer is, in general, no, since we have
Elastic Media. The differential equation for the effective shown that the Mori-Tanaka method for multiphase media
moduli L(t) can be derived in a manner similar to the deriva- can give moduli outside the limits of the Hashin-Shtrikman
tion of equation (26). Thus, bounds. That it may be possible in two-phase composites is
suggested by the fact that the Mori-Tanaka approximation
rfL=E(Ly-L)fyJ(l-c)I satisfies the bounds, and also because the differential scheme
condition (31) or (37) is simpler than that for multiphase
media. It remains as an interesting and worthwhile challenge
i / dc to provide a realization of the method for two phases.
£ c,T;o] (,dc: + c, (36)
1- ) •
Acknowledgments
where the initial conditions are L(0) = L 0 j c,(0) = 0, y = l , 2,
. . . , « . The strain concentration tensors tJ0,j= 1, 2, . . . , n Thanks to G. Milton and G. Weng for helpful discussions.
in equation (36) compare the strain in the currently added This work was supported by the National Science Foundation
phase y to the strain in phase 0 throughout the composite. The through Grant No. MDM 85-16256.
tensors T,0 compare the bulk strain in phase y to the bulk
strain in phase 0. References
In the same way that the differential equation (26) was
shown to be integrable if equations (31) and (32) hold, so it Avellaneda, M., 1987, "Iterated Homogenization, Differential Effective
Medium Theory and Applications," Communications in Pure and Applied
can be shown that equation (36) is exactly integrable if both Mathematics, Vol. 40, pp. 527-554.
Benveniste, Y., 1986a, "On the Effective Thermal Conductivity of
tjo(0=f„(0), y=l,2,. ,n, (37) Multiphase Composites," Journal of Applied Mathematics and Physics
(ZAMP), Vol. 37, pp. 696-713.
and Benveniste, Y., 1986b, "On the Mori-Tanaka's Method in Cracked Bodies,"
Mechanics Research Communications, Vol. 13, pp. 193-201.
T,o — T, 0 , y — 1, 2, (38) Benveniste, Y., 1987a, " A Differential Effective Medium Theory with a
Composite Sphere Embedding," ASME JOURNAL OF APPLIED MECHANICS, Vol.
The case of a two-phase composite is again the exception. 54, pp. 466-468.
The strain concentration tensor T 10 can then be expressed in Benveniste, Y., 1987b, "The Effective Thermal Conductivity of Composites
terms of the effective moduli as with a Thermal Resistance Between the Constituents: Non-Dilute Case," Jour-
nal of Applied Physics, Vol. 61, pp. 2840-2843.
Benveniste, Y., 1987c, " A New Approach to the Application of Mori-
-(^K Lo)(L,-L)-1,

which when eliminated from equation (36) gives,

(39)
dc
dL = ( L - L , ) f 10(L, - L 0 ) - , ( L - L I ) ' . (40)
This differential equation depends only upon T 10 , and can be
directly integrated if it is constant. The resultant moduli when
substituted into equation (39) give T 10 = T 10 , and the moduli
themselves are identical to those of the Mori-Tanaka method.
It has come to my attention that laws (1980) also derived the
Mori-Tanaka effective medium equation for a two phase com-
posite as a hybrid of the differential scheme and the self-
consistent method.
Tanaka's Theory in Composite Materials," Mechanics of Materials, Vol. 6, pp.
147-157.
Bruggeman, D. A. G., 1035, "Berechnung Verschiedener Physikalischer
Konstante von Heterogene Substanze," Annalen der Physik, Vol. 24, p. 636.
Budiansky, B., 1965, "On the Elastic Modulus of Some Heterogeneous
Materials," Journal of the Mechanics and Physics of Solids, Vol. 13, pp.
223-227.
Cleary, M. P., Chen, I. W., and Lee, S. M., 1980, "Self-Consistent Tech-
niques for Heterogeneous Solids," ASCE Journal of Engineering Mechanics,
Vol. 106, pp. 861-887.
Hashin, Z., 1965, "On Elastic Behavior of Fibre Reinforced Materials of Ar-
bitrary Transverse Phase Geometry," Journal of the Mechanics and Physics of
Solids, Vol. 13, pp. 119-134.
Hashin, Z., and Shtrikman, S., 1963, "A Variational Approach to the Elastic
Behavior of Multiphase Materials," Journal of the Mechanics and Physics of
Solids, Vol. 11, pp. 127-140.
Hill, R., 1964, "Theory of Mechanical Properties of Fibre-Strengthened

Journal of Applied Mechanics MARCH 1989, Vol. 56/87

Downloaded From: http://appliedmechanics.asmedigitalcollection.asme.org/ on 04/06/2014 Terms of Use: http://asme.org/terms

 Materials: I. Elastic Behaviour," Journal of the Mechanics and Physics of a+5 = L0(e0 + e). 042)
Solids, Vol. 12, pp. 199-212.
Hill, R., 1965, "Continuum Micromechanics of Elastoplastic Polycrystals," The additional perturbed stress and strain in phase i, ;>0, are
Journal of the Mechanics and Physics of Solids, Vol. 13, pp. 89-101. of' and ef", where
Landau, L. D., and Lifshitz, E. M., 1960, Electrodynamics of Continuous
Media, Pergamon Press, Oxford. a + a + af =L,(e 0 + f+tf'),
Laws, N., 1974, "The Overall Thermoelastic Moduli of Transversely
Isotropic Composites According to the Self-Consistent Method," International
Journal of Engineering Science, Vol. 12, pp. 79-87.
= L0(e0 + e + 6f'-ei*), 043)
Laws, N., 1980, "A Note on the Prediction of Overall Moduli for Composite and e* is the transformation strain in phase i. Taking the
Materials," Quarterly Journal of Mechanics and Applied Mathematics, Vol. 33,
pp. 43-45.
average strain throughout the composite, and using 041),
Maewal, A.,andDandekar, D. P., 1987, "Effective Thermoelastic Properties gives
of Short-Fiber Composites," Acta Mechanica, Vol. 66, pp. 191-204. n
McGlaughlin, R., 1977, "A Study of the Differential Scheme for Composite
Materials," International Journal of Engineering Science, Vol. 15, pp. 237-244. ' " = ! > , . (*!*-<?')• 044)
Mori, T., and Tanaka, K., 1973, "Average Stress in Matrix and Average ;'=1
Elastic Energy of Materials with Misfitting Inclusions," Acta Metallurgica, Vol.
231, pp. 571-574. Equation 044) then implies that the average strain is
Mura, T., 1982, Micromechanics of Defects in Solids, Martinus Nijhoff, The n
Hague, Netherlands.
Norris, A. N., 1985, " A Differential Scheme for the Effective Moduli of * = «o + X^,e,*. 045)
Composites," Mechanics of Materials, Vol. 4, pp. 1-16. /=i
Norris, A. N., Callegari, A. J., and Sheng, P., 1985, "A Generalized Dif-
ferential Effective Medium Theory," Journal of the Mechanics and Physics of The effective moduli are defined by <r = Le, or from 041) and
Solids, Vol. 33, pp. 525-543. 045),
Tandon, G. P., 1986, "Micromechanical Determination of the Elasticity and
Plasticity of Composites," Ph.D. Thesis, Rutgers University.
L
Taya, M., and Mura, T., 1981, "On Stiffness and Strength of an Aligned (fo+ £c,«,*)=L 0 e 0 . 046)
Short-Fiber Reinforced Composite Containing Fiber-End Cracks Under Uniax-
ial Applied Stress," ASME JOURNAL OF APPLIED MECHANICS, Vol. 48, pp.
361-367. Some relations between t* and e0 are necessary to solve (A6)
Taya, M., and Chou, T. W., 1981, "On Two Kinds of Ellipsoidal In- forL.
homogeneities in an Infinite Elastic Body: An Application to a Hybrid Com-
posite," International Journal of Engineering Science, Vol. 17, pp. 553-563.
For each /= 1, 2, . . . , n let
Walpole, L. J., 1966, "On Bounds for the Overall Elastic Moduli of In-
homogeneous Systems—II," Journal of the Mechanics and Physics of Solids,
ef = S„«;. 047)
Vol. 14, pp. 289-301.
Walpole, L. J., 1969, "On the Overall Elastic Moduli of Composite
The average strain in the matrix phase 0 is, from 042)
Materials," Journal of the Mechanics and Physics of Solids, Vol. 17, pp.
235-251.
e0 = e0 + e. 048)
Weng, G. J., 1984, "Some Elastic Properties of Reinforced Solids with
Special Reference to Isotropic Ones Containing Spherical Inclusions," Interna-
Equations 043), 047), (AS), and (12) imply, for each /= 1, 2,
tional Journal of Engineering Science, Vol. 22, pp. 845-856. . . . ,n,
ef=-L 0 -'(L,-L 0 )f ; 0 6 0 . (A9)
APPENDIX Then, equations 044), 047)-049), and (12) give
Alternative Formulation of the Mori-Tanaka Method
*o = (c0I + E c<'Lo" lL / T /o) «o- 0410)
For an average stress in the composite of a, define the cor- x
;=i '
responding strain e0 in pure matrix material, Equations 049) and 0410) can then be substituted into (Ad) to
a = L0e0. (AX) given an explicit expression for L that is identical to equation
(11). The Mori-Tanaka assumption is that S,0 is equal to the
The perturbed stress and strain in phase 0 in the composite are corresponding Eshelby tensor for a single inclusion of phase i
a and t, where in phase 0.

88/Vol. 56, MARCH 1989 Transactions of the ASME

Downloaded From: http://appliedmechanics.asmedigitalcollection.asme.org/ on 04/06/2014 Terms of Use: http://asme.org/terms