Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
www.elsevier.com/locate/ijengsci
Received 24 April 2004; received in revised form 15 November 2004; accepted 21 January 2005
Abstract
Coupling between the electric eld, magnetic eld, and strain of composite materials is achieved when
electro-elastic (piezoelectric) and magneto-elastic (piezomagnetic) particles are joined by an elastic matrix.
Although the matrix is neither piezoelectric nor piezomagnetic, the strain eld in the matrix couples the
electric eld of the piezoelectric phase to the magnetic eld of the piezomagnetic phase. This three-phase
electro-magneto-elastic composite should have greater ductility and formability than a two-phase compos-
ite in which the electric eld and the magnetic eld are coupled by directly bonding two brittle materials. A
nite element analysis (FEA) and micromechanics based averaging of a representative volume element
(RVE) are performed in this work to determine the eective dielectric, magnetic, mechanical, and cou-
pled-eld properties of an elastic matrix reinforced with piezoelectric and piezomagnetic bers as functions
of the phase volume fractions, the ber arrangements in the RVE, and the ber material properties with
special emphasis on the poling directions of the piezoelectric and piezomagnetic bers. The eective mag-
neto-electric moduli of this three-phase composite are found to be less than the eective magneto-electric
moduli of a two-phase piezoelectric/piezomagnetic composite, because the elastic matrix is not sti enough
to transfer signicant strains between the piezomagnetic and piezoelectric bers.
2005 Published by Elsevier Ltd.
*
Corresponding author.
E-mail addresses: jboyd@aero.tamu.edu, jboyd@mems.tamu.edu (J.G. Boyd IV), lagoudas@aero.tamu.edu,
lagoudas@smart.tamu.edu (D.C. Lagoudas).
1. Introduction
Eorts are under way to develop composite materials with improved product properties [1],
which are created by the interaction between the constituent phases. Those product properties,
which are absent in constituent phases, are obtained by combining dierent active materials.
An active material is a continuum whose mechanical constitutive response is coupled with a
non-mechanical eect. Thermoelasticity, piezoelectricity, piezomagneticity, electrostriction, mag-
netostriction and hygroelasticity are a few examples of active phenomena [2]. A two-phase com-
posite consisting of a thermoelastic phase and a piezoelectric phase can exhibit pyroelectricity
even though neither of the phases is pyroelectric [3]. The composite consisting of a piezoelectric
phase and a piezomagnetic phase exhibits a magneto-electric coupling eect that is absent in each
of the phase. The magneto-electric coupling eect has recently attracted attention due to the
extensive applications for broadband magnetic eld probes, electronic packaging, acoustic
devices, hydrophones, medical ultrasonic imaging, sensors, and actuators [48].
In 1974 Van Run et al. [9] reported that the magneto-electric eect obtained in a BaTiO3
(piezoelectric phase)CoFe2O4 (piezomagnetic phase) composite was two orders of magnitude
larger than that of the single-phase magneto-electric material Co2O3. Bracke and Van Vliet [10]
reported a broad band magneto-electric transducer with a at frequency response using composite
materials. Since then, much of the theoretical and experimental work for investigation into the
magneto-electric coupling eect has been carried out in papers published by Harshe [11], Harshe
et al. [12], Avellaneda and Harshe [13], Nan [1], Benveniste [14], Huang and Kuo [15], Li and
Dunn [4], Wu and Huang [16], Huang et al. [17] and Aboudi [5]. They obtained expressions for
the eective magneto-electric coecient and showed numerical results with practical examples
using BaTiO3CoFe2O4 two-phase composite.
Harshe et al. [12] treated the magneto-electric eect of piezoelectricpiezomagnetic two-phase
composites in terms of a simple geometrical model by assuming the particles embedded in the
matrix were small cubes. Then, they solved the elds in one cube for which the boundary value
problem is tractable. This simple cubes model is lacking in theoretical rigor.
Nan [1] proposed a non-self-consistent (NSC) approach to calculate the eective properties of
piezoelectricpiezomagnetic two-phase composites, consisting of aligned cylindrical bers embed-
ded in the matrix, based on a Greens function method and perturbation theory, which have been
widely used to treat the general, linear-response properties of inhomogeneous media. However,
Benveniste [14] derived exact connections between the eective moduli of piezoelectricpiezomag-
netic two-phase composites, which were independent of the details of the microgeometry and of
the particular choice of the averaging model, using the uniform eld concept. The NSC approx-
imation of Nan failed to satisfy the exact connections between dierent components of the eec-
tive moduli obtained by Benveniste.
Li and Dunn [4] developed a micromechanics approach to analyze the eective moduli for two-
phase piezoelectricpiezomagnetic composites. They derived explicit expressions for the general-
ized Eshelby tensors and used the MoriTanaka method to obtain closed form expressions for the
eective electro-magneto-elastic moduli of circular cylinder brous and laminated two-phase
composites.
The MoriTanaka method was also employed by Wu and Huang [16] in order to obtain a
closed form solution for eective moduli of piezoelectricpiezomagnetic two-phase composites
792 J. Lee et al. / International Journal of Engineering Science 43 (2005) 790825
for four practical inclusions: elliptical cylinder, circular cylinder, penny shape, and ribbon. Huang
et al. [17] analytically obtained the optimized ber volume fraction for maximum magneto-electric
coupling eects in the piezoelectricpiezomagnetic two-phase composite by dierentiating their
previous closed form formulae with respect to the ber volume fraction and setting zero. The opti-
mized ber volume fraction is a function of the elastic properties of the constituents, but not a
function of the magnetic and electric properties.
Li [18] analyzed the electro-magneto-elastic eld in a multi-inclusion, which means there is an
inclusion containing another inclusion, embedded in an innite matrix and developed a numerical
algorithm to evaluate electro-magneto-elastic Eshelby tensors.
Aboudi [5] employed a homogenization micromechanical method to predict the eective
moduli of electro-magneto-thermo-elastic multi-phase composites. The homogenization method
assumes that elds vary on multiple spatial scales due to the existence of a microstructure and
the fact that the microstructure is spatially periodic. It was seen that the homogenization theory
and MoriTanaka results of Li and Dunn [4] for the electro-magneto-elastic two-phase composite
were very similar.
Unfortunately, piezoelectric and piezomagnetic materials are usually brittle ceramics, therefore
a composite consisting of those two phases would be susceptible to brittle fracture. A three-phase
electro-magneto-elastic composite consisting of piezoelectric and piezomagnetic phases separated
by a polymer matrix would have greater ductility and formability. Recently, Boyd et al. [19] pre-
sented a method for using arrays of micro-electro-mechanical systems (MEMS) electrodes and
electromagnets to achieve microscale positioning of piezoelectric and piezomagnetic particles in
liquid polymers, which would then be solidied to make a polymer matrix magneto-electric com-
posite. The ability to pattern the particles into microscale unit cells would reduce concentrations
of stress, electric eld, and magnetic eld, thereby increasing eective threshold properties such as
strength, electric breakdown eld, and magnetic saturation eld. Furthermore, one can hope that
someday arrays of MEMS can be used to control the microscale distribution of the poling direc-
tion of piezoelectric and piezomagnetic particles.
Lee et al. [20] introduced three-phase electro-magneto-elastic composites consisting of an elastic
matrix reinforced with piezoelectric and piezomagnetic bers. By employing the nite element
method (FEM), they obtained the eective properties of a three-phase electro-magneto-elastic
composite and found that the three-phase electro-magneto-elastic composite contained a non-zero
magneto-electric coecient.
Motivated by the work of Boyd et al. [19], in the present paper we calculate the eective mag-
neto-electric coecients of three-phase composites consisting of piezoelectric and piezomagnetic
phases separated by an elastic matrix. Although the matrix is neither piezoelectric nor piezomag-
netic, the strain eld in the matrix couples the electric eld of the piezoelectric phase to the mag-
netic eld of the piezomagnetic phase. The nite element method is used because, unlike closed
form micromechanics methods, the nite element method reveals the concentration of stress, elec-
tric eld, and magnetic eld within the unit cell. These eld concentrations determine the eective
strength, electric breakdown eld, and magnetic saturation eld of the composite. Volumetric
averaging is then applied to the FEA solution to determine the eective electro-magneto-elastic
properties. In addition to FEA, a mean-eld approach based on the MoriTanaka approximation
is used in this work to connect with previous work in the literature and also with the proposed
FEA. There are two novel aspects in the present work. First, a three-phase electro-magneto-elastic
J. Lee et al. / International Journal of Engineering Science 43 (2005) 790825 793
composite consisting of the elastic matrix reinforced with the piezoelectric bers and the piezo-
magnetic bers is introduced. Second, the magneto-electric coupling eects are studied when
the poling directions of the piezoelectric phase and the piezomagnetic phase change. The commer-
cially available software FEMLAB is used for the computational analysis.
The basic equations and notation used in this paper are presented in Section 2.1. In Sections 2.2
and 2.3, respectively we presented the FEA and volumetric averaging method of a unit cell and the
MoriTanaka method for three-phase electro-magneto-electric composites. In Section 3.1, the re-
sults of a nite element analysis are compared to the MoriTanaka results of Li and Dunn [4] for
a two-phase composite consisting of a piezomagnetic matrix with piezoelectric bers. For a three-
phase composite, the nite element results are presented in Section 3.2. In Section 3.3, parametric
studies are presented with special emphasis on the phase volume fraction and the poling directions
of the piezoelectric and piezomagnetic bers.
where r, e, g, l and k are the stress, strain, dielectric permittivity, magnetic permeability and mag-
neto-electric second order tensors; e and q are the piezoelectric and piezomagnetic third order ten-
sors; C is the fourth order elasticity tensor. D, E, B and H are electric displacement, electric eld,
magnetic induction and magnetic eld vectors. For specic cases of the piezoelectric phase, the
piezomagnetic phase, and the elastic matrix, the constitutive equations can be expressed with
Eq. (1) by deleting coupling coecient tensors e, q and k accordingly. For the piezoelectric phase,
the piezomagnetic phase, and the elastic matrix, the constitutive equations are thus given by Eqs.
(2a)(2c) below, respectively:
Piezoelectric constitutive equations
r Ce eT E
D ee gE 2a
B lH
Piezomagnetic constitutive equations
r Ce qT H
D gE 2b
B qe lH
Elastic matrix constitutive equations
r Ce
D gE 2c
B lH
It should be noted that, even though the same symbols are used, the material coecient tensors
are in general dierent for the dierent phases.
The balance of linear momentum (quasistatic conditions), Gauss law, and the condition of no
magnetic poles are given by
r r 0; r D 0; rB0 3ac
We assume small displacement gradients and employ the linearized straindisplacement
relations given by
1
e ru ruT 4
2
where u is a displacement vector.
By the theory of electrostatics and magnetostatics, the electric eld E and the magnetic eld
strength H are the gradient of an electric potential Ue and a magnetic potential Um:
E rUe ; H rUm 5ab
Combining the governing dierential equations (3)(5) with the constitutive equations (2a)(2c)
results in the following eld equations (6a)(6c) for u, Ue and Um in the piezoelectric phase, piezo-
magnetic phase and elastic matrix:
J. Lee et al. / International Journal of Engineering Science 43 (2005) 790825 795
electric elds of the piezomagnetic and piezoelectric particles are coupled through the matrix
strain eld. Note that the eective moduli L* is diagonally symmetric as shown in Appendix B.
2.2. Representative volume element (RVE) and nite element analysis (FEA)
2.2.1. RVE
The RVE consists of two types of bers, i.e. a piezoelectric and a piezomagnetic, separated by
an elastic matrix. Various types of RVEs (periodic unit cells) are possible for unidirectionally ber
reinforced composites. Figs. 2 and 3 show a square periodic unit cell and a hexagonal periodic
unit cell frequently employed in the literature [21]. An important characteristic of these compos-
ites is the material symmetry of the eective properties in the transverse plane. Although a square
periodic unit cell is more frequently used than a hexagonal periodic unit cell in the literature, a
square periodic unit cell lacks the transverse isotropy that most unidirectional composites, consist-
ing of a matrix reinforced with randomly distributed bers, possess owing to the random distri-
bution of bers in the matrix in the transverse plane. If the transverse isotropy is not a
characteristic to be preserved through the idealization of the periodic microstructure, then advan-
Fig. 2. (a) Square periodic arrangement of bers and (b) periodic unit cell for square arrangement.
Fig. 3. (a) Hexagonal periodic arrangement of bers and (b) periodic unit cell for hexagonal arrangement.
J. Lee et al. / International Journal of Engineering Science 43 (2005) 790825 797
tage can be taken of the square periodic arrays simplicity by comparing it with other types of peri-
odic unit cells. The advantage of the hexagonal periodic unit cell over the square periodic unit cell
is that the hexagonal periodic unit cell preserves transverse isotropy. However, the hexagonal peri-
odic unit cell is substantially more complicated than a square periodic unit cell [2123]. For this
reason, we chose the square periodic unit cell of Fig. 2.
To choose the smallest computational domain of the periodic unit cell for the FEA, we have to
consider the symmetry in terms of the geometry, the boundary conditions and the poling direc-
tions of the piezoelectric and piezomagnetic phases. The periodic unit cell shown in Fig. 2(b) is
the smallest unit cell of the periodic structure in Fig. 2(a) that can model all the dierent poling
cases. We next describe the boundary conditions for the square periodic unit cell of Fig. 2(b).
Fig. 4. Schematic of the periodic unit cell used for the periodic boundary conditions.
Similarly, for the planes at x2 = a, the periodic boundary condition equations (9)(11) reduce
to
ou1
u1 x1 ; a; x3 u1 x1 ; a; x3 2a 13a
ox2
ou2
u2 x1 ; a; x3 u2 x1 ; a; x3 2a 13b
ox2
ou3
u3 x1 ; a; x3 u3 x1 ; a; x3 2a 13c
ox2
e
oU
Ue x1 ; a; x3 Ue x1 ; a; x3 2a 13d
ox2
m
m m oU
U x1 ; a; x3 U x1 ; a; x3 2a 13e
ox2
In the x3 direction the elds vary linearly with x3, which can be expressed as periodic boundary
conditions similar to Eqs. (9)(11) for the planes x3 = a, i.e.,
ou1
u1 x1 ; x2 ; a u1 x1 ; x2 ; a 2a 14a
ox3
ou2
u2 x1 ; x2 ; a u2 x1 ; x2 ; a 2a 14b
ox3
ou3
u3 x1 ; x2 ; a u3 x1 ; x2 ; a 2a 14c
ox3
ou3
u3 x1 ; x2 ; a u3 x1 ; x2 ; a 2a 14d
ox3
e
oU
Ue x1 ; x2 ; a Ue x1 ; x2 ; a 2a 14e
ox3
J. Lee et al. / International Journal of Engineering Science 43 (2005) 790825 799
Due to the special dependence on x3, the FEA can be performed using special 2D elements [24].
However, a 3D analysis has been conducted in this work and the thickness 2a has been arbitrarily
chosen for the use in FEA with the commercial software FEMLAB.
The boundary conditions are simply applied to the 3D square periodic unit cell using the peri-
odic boundary condition equations (12)(14) [2528]. For example, by applying a uniaxial strain
state, hou1/ox1i = he11i = constant and all of the other averaged terms to be zero, one can con-
struct one of the independent boundary value problems. Returning to the engineering formulation
of the constitutive response for the eective composite, the implementation of this boundary con-
ditions would be as follows:
8 9 8 9
>
> hr11 i >
> >
> he11 i >
>
>
> >
> >
> >
>
>
> >
> >
> >
>
>
> hr22 i >
> >
> 0 >
>
>
> >
> >
> >
>
>
> >
> >
> >
>
>
> hr i >
> >
> 0 >
>
>
>
33 >
> >
> >
>
>
> >
> >
> >
>
>
> hr i >
> >
> 0 >
>
>
> 23 >
> >
> >
>
>
> >
> >
> >
>
>
> > >
> >
> hr31 >
i > 2 3> 0 >
>
>
> >
> C eT q T >
> >
>
>
> > >
> >
< hr12 i >
= 6 7< 0
>
=
6
4e g k 5T 7
() hRi1 L hZi1 15
>
> hD1 i >> > 0 >
> >
>
> > > >
>
>
>
>
> q k l > >
>
>
>
>
>
> hD i >
> >
> 0 >
>
>
> 2 >
> >
> >
>
>
> >
> >
> >
>
>
> >
> >
> >
>
>
> hD 3 i >
> >
> 0 >
>
>
> >
> >
> >
>
>
> > >
> >
> hB1 i >
> >
> >
> 0 > >
>
>
> >
> >
> >
>
>
> >
> >
> >
>
>
> hB i >
> >
> 0 >
>
>
>
2 >
> >
> >
>
>
: >
; >
: >
;
hB3 i 0
where the superscript indicates the results from a certain input state.; [hRi1 hRi12] and
[hZi1 hZi12] are 12 12 matrices, which are represented as
800 J. Lee et al. / International Journal of Engineering Science 43 (2005) 790825
2 31212
hr11 i hr11 i
6 hr i hr22 i 7
6 22 7
6 7
6 hr33 i hr33 i 7
6 7
6 7
6 hr23 i hr23 i 7
6 7
6 hr i hr31 i 7
6 31 7
6 7
6 hr12 i hr12 i 7
6 7
6 .......... 7
6 hD1 i hD1 i 7
6 7
6 7
6 hD2 i hD2 i 7
6 7
6 hD i hD3 i 7
6 3 7
6 7
6 hB1 i hB1 i 7
6 7
6 7
4 hB2 i hB2 i 5
hB3 i hB3 i
2 3
he11 i 0 0 0 0 0 0 0 0 0 0 0
6 0 he22 i 0 0 0 0 0 0 0 0 0 0 7
6 7
6 7
6 0 0 he33 i 0 0 0 0 0 0 0 0 0 7
6 7
6 7
6 0 0 0 2he23 i 0 0 0 0 0 0 0 0 7
6 7
6 0 0 0 0 2he31 i 0 0 0 0 0 0 0 7
6 7
6 7
6 0 0 0 0 0 2he12 i 0 0 0 0 0 0 7
6 7
L 6 7
6 0 0 0 0 0 0 hE1 i 0 0 0 0 0 7
6 7
6 7
6 0 0 0 0 0 0 0 hE2 i 0 0 0 0 7
6 7
6 0 0 0 0 0 0 0 0 hE3 i 0 0 0 7
6 7
6 7
6 0 0 0 0 0 0 0 0 0 hH 1 i 0 0 7
6 7
6 7
4 0 0 0 0 0 0 0 0 0 0 hH 2 i 0 5
0 0 0 0 0 0 0 0 0 0 0 hH 3 i
17
From this equation, the 12 12 eective matrix of three-phase electro-magneto-elastic composite
L* is easily determined to be
L hRi1 hRi12 hZi1 hZi12 1 18
In this section we introduced the generalized expression for the eective properties of N-phase
electro-magneto-elastic composites and the MoriTanaka method for the three-phase electro-
magneto-elastic composites which is used for comparison with FEA.
J. Lee et al. / International Journal of Engineering Science 43 (2005) 790825 801
where the subscript m denotes the matrix and i is used for numbering the dispersed phases. The
matrix and dispersed phases are characterized by their volume fractions cm and ci. Applying an
analogous result for hRi, and using the constitutive equations for each phase and the composite
yields
X
N1
L Z0 Lm Z0 ci Li Lm hZi i 22
i1
Finally, by simple manipulations, the general expression for the eective properties of perfectly
bonded N-phase composites yields
X
N 1
L Lm ci Li Lm Ai 23
i1
where Ai is the concentration factor that relates the average strain and potential gradients in phase
i to that in the composite, namely,
hZi i Ai Z0 24
The estimation of Ai is the key to predicting the eective electro-magneto-elastic moduli L*.
hZi i Adil
i hZm i 25
where Adil
i is the dilute concentration tensor that can be obtained from electro-magneto-elastic
Eshelby tensors Si, which are composed of one fourth rank tensor, four third rank tensors, and
four second rank tensors. Explicit expressions for the electro-magneto-elastic Eshelby tensors
given by Li and Dunn [29] and Huang et al. [30] are tabulated in Appendix C. For ellipsoidal
inclusions, they are functions of the shape of the inclusion and the electro-magneto-elastic moduli
of the matrix.
From the average strain theorem and the denition of the concentration factor, the dilute
concentration factor Adil
i and MoriTanaka concentration factor Ai
MT
can be shown as
1 1
Adil
i I Si Lm Li Lm 26
!1
X
N 1
AMT
i Adil
i cm I ci Adil
i 27
i1
3. Results
Table 1
Material properties of BaTiO3 and CoFe2O4
BaTiO3 (piezoelectric phase) CoFe2O4 (piezomagnetic phase)
10
C11 (Pa) 16.6 10 28.6 1010
C12 (Pa) 7.7 1010 17.3 1010
C13 (Pa) 7.8 1010 17.05 1010
C33 (Pa) 16.2 1010 26.95 1010
C44 (Pa) 4.3 1010 4.53 1010
g11 (C2/N m2) 112 1010 0.8 1010
g33 (C2/N m2) 126 1010 0.93 1010
l11 (N s2/C2) 5 106 590 106
l33 (N s2/C2) 10 106 157 106
e31 (C/m2) 4.4 0
e33 (C/m2) 18.6 0
e15 (C/m2) 11.6 0
q31 (N/A m) 0 5.803 102
q33 (N/A m) 0 6.997 102
q15 (N/A m) 0 5.5 102
The basic assumption of the MoriTanaka method used by Li and Dunn [4] to determine the
eective properties of a two-phase electro-magneto-elastic composite is that all bers are parallel,
even though they are randomly distributed in the transverse direction. This means that the eec-
tive moduli determined by the MoriTanaka method preserve the transverse isotropy. Thus, the
proper unit cell for the FEM to compare with the MoriTanaka method by Li and Dunn [4] is the
hexagonal unit cell provided all material properties are transversely isotropic. However, we con-
sider the square unit cell for a two-phase electro-magneto-elastic composite shown in Fig. 5
because it derives from the unit cell for a three-phase composite, as explained above.
Because the direct comparison is not proper, to be able to compare with the MoriTanaka
method, we consider a rotational average value of the FEA result for all possible rotational var-
iation with x3 the axis of rotation. The rotational average method and the verication of the trans-
verse isotropy of the averaged properties of the FEA results are explained in Appendix D.
The eective moduli of the brous two-phase electro-magneto-elastic composite were evalu-
ated for two ber volume fractions, vf = 0.4 and vf = 0.6. The averaged results for the rotational
804 J. Lee et al. / International Journal of Engineering Science 43 (2005) 790825
variation are compared with the predictions of Li and Dunn [4] in Figs. 611, in which the eec-
tive moduli determined by the FEA are represented by dots. The FEA results are in excellent
agreement with the MoriTanaka method results of Li and Dunn [4]. As mentioned by Li and
Dunn [4], the magneto-electric coupling eect is present in the two-phase composite (Fig. 11),
which does not exist in either of the individual phases.
The number of elements used for the FEA with the ber volume fractions vf = 0.4 and vf = 0.6,
which satises the convergence test, are 16,207 and 27,804, respectively. The analysis is carried out
using a linear iterative solver.
Material properties of the piezoelectric phase and the piezomagnetic phase are shown in
Table 1.
x 10
11 C(elastic moduli)
3
C
11
2.5
C
33
2
C12
C(Pa)
1.5
C
13
C44
0.5
0
0 0.2 0.4 0.6 0.8 1
f
x 10 Dielectric permittivity
1.4
1.2
33
1
(C2/Nm )
2
0.8
0.6
0.4
0.2 11
0
0 0.2 0.4 0.6 0.8 1
f
x 10 Magnetic permeability
2
33
(Ns /C )
2
2
11
Piezoelectric constants
20
e
15 33
10
e(C/m2)
e15
0
e31
Piezomagnetic constants
700
600
q33
500
q31
q(N/Am)
400
q15
300
200
100
0
0 0.2 0.4 0.6 0.8 1
f
Magnetoelectric constant
9 33
x 10
3
2.5
2
33(Ns/VC)
1.5
0.5
0
0 0.2 0.4 0.6 0.8 1
f
Fig. 11. Eective magneto-electric constant k33 of a brous two-phase composite vs vf.
applied. <E1> induces the E1 distribution of Fig. 13, which, in turn, results in the shear strain dis-
tribution of Fig. 14. The strain in the piezomagnetic ber results in the magnetic eld distribution
of Fig. 15. Thus, the three-phase composite exhibits the magneto-electric eect by coupling the
piezomagnetic and piezoelectric bers through the strain in the elastic matrix.
The number of elements used for the FEA, which satises the convergence test, are 21,824,
when the phase volume fractions are 0.3, 0.3 and 0.4 for the piezomagnetic ber, the piezoelectric
ber and the elastic matrix, respectively. The analysis is carried out using a linear iterative solver.
Applying the remaining 11 independent boundary conditions and solving for the eective prop-
erties of the composite as previously discussed, we obtain the 12 12 eective moduli (refer back
to Eq. 29 and Table 2) of the three-phase composite from the FEA results. Note that the magneto-
J. Lee et al. / International Journal of Engineering Science 43 (2005) 790825 807
Fig. 12. Electric potential distribution of a brous three-phase electro-magneto-elastic composite when <E1> is
applied.
Fig. 13. Electric eld E1 distribution of a brous three-phase electro-magneto-elastic composite when <E1> is applied.
Fig. 14. Shear strain e31 distribution of a brous three-phase electro-magneto-elastic composite when <E1> is applied.
808 J. Lee et al. / International Journal of Engineering Science 43 (2005) 790825
Fig. 15. Magnetic eld H1 distribution of a brous three-phase electro-magneto-elastic composite when <E1> is
applied.
Fig. 16. Magnetic potential distribution of a brous three-phase electro-magneto-elastic composite when <E1> is
applied.
Table 2
L* for three-phase composite
J. Lee et al. / International Journal of Engineering Science 43 (2005) 790825 809
electric coupling eect is present in the three-phase composite (Figs. 1216), which does not exist
in either of the constituent phases and the eective magneto-electric moduli of the three-phase
composite are about twenty times less than those of the two-phase composite (Fig. 11 and Table
2). This is due to the fact that the stiness of the elastic matrix used in the three-phase composite is
two orders of magnitude less than those of the piezomagnetic and piezoelectric phases. Thus,
although the matrix of the three-phase composite is strained, it lacks sucient stiness to transfer
these strains to the piezomagnetic ber.
10
x 10
Effective stiffness C (m =0.4)
10
MT
9 C FEM
33
6
C (Pa)
2
C
11
1 C13
C44
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Volume fraction of Piezomagnetic phase (f)
7
C 66 (Pa)
1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Volume fraction of Piezomagnetic phase (f)
Note that all the zero terms of the eective matrix generated by the FEM are not exactly zero.
The terms which are over three orders of magnitude less than the other terms are assumed to be
numerical error.
8
e (C/m )
2
2
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Volume fraction of Piezomagnetic phase (f)
160
140
q15 MT
120 q MT
31
q MT
33
100
q (N/Am)
FEM
80
60
40
20
20
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Volume fraction of Piezomagnetic phase (f)
The eective dielectric, magnetic, mechanical, and coupled-eld properties are functions of the
phase volume fractions, the ber shapes, the ber material properties, the poling directions of pie-
zoelectric phase and piezomagnetic phase and the ber arrangements in the unit cell. The FEM is
used to determine the eective properties with special emphasis on the phase volume fractions and
the poling directions of the piezoelectric and piezomagnetic bers. The results were compared with
those of the MoriTanaka method. Because the eective moduli predicted by the MoriTanaka
method are functions of the phase volume fractions, the ber shapes and the phase material prop-
erties, the MoriTanaka method cannot recognize particular arrangements of the bers. Thus, the
x 10
9 Effective dielectric permittivity (m =0.4)
8
11 MT
7 33 MT
FEM
6
(C /Nm )
5
2
4
2
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Volume fraction of Piezomagnetic phase (f)
4
x 10 Effective magnetic permeability (m =0.4)
1
0.9
0.8
MT
0.7 11
MT
33
(Ns /C )
0.6
2
FEM
2
0.5
0.4
0.3
0.2
0.1
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Volume fraction of Piezomagnetic phase (f)
direct comparison between the two is not reasonable because the FEM uses a specic periodic
arrangement of bers (Fig. 2) that is not recognized by the MoriTanaka method. We represent
the FEM results without the rotational averaging (that was performed for the two-phase compos-
ite) because the rotational averaging of the three-phase composites in which the piezoelectric and
piezomagnetic phases have dierent poling directions is of little use. Nevertheless, the predictions
of the MoriTanaka method are a useful reference.
13
x 10
Effective magnetoelectric constants (m=0.4)
11
3
MT
FEM
2.5
2
11 (Ns/VC)
1.5
0.5
0.5
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Volume fraction of Piezomagnetic phase (f )
x 10
11 Effective magnetoelectric constants (m=0.4)
33
12
MT
FEM
10
8
(Ns/VC)
4
33
2
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Volume fraction of Piezomagnetic phase (f)
is denoted by vf. The eective elastic, dielectric, magnetic, piezoelectric, piezomagnetic and mag-
neto-electric coupling moduli of a three-phase composite are generated for 0 6 mf 6 0.6. The FEM
results are compared with the MoriTanaka method results in Figs. 1728, in which the eective
moduli determined by the FEA are presented by triangles.
The eective elastic modulus C12 is not presented, because C12 is very close to C13. In some
cases, the FEM results and MoriTanaka method results coincide, while in other cases slight devi-
ations are observed. The eective stiness constants in Figs. 17 and 18 generated by the Mori
Tanaka method show an almost linear variation with the volume change. On the other hand,
the eective stiness determined by the FEM, which recognizes the specic arrangement of the
10
x 10
Effective stiffness C11 (p.e:p.m=1:1)
2.5
MT
FEM
2
C11 (Pa)
1.5
1
0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6
Volume fraction of elastic matrix ( m)
x 10
9 Effective stiffness C12 (p.e:p.m=1:1)
14
MT
13 FEM
12
11
C12 (Pa)
10
5
0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6
Volume fraction of elastic matrix ( m )
x 10
9 Effective stiffness C (p.e:p.m=1:1)
13
14
MT
13 FEM
12
11
(Pa) 10
13
9
C
5
0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6
Volume fraction of elastic matrix (m)
x 10
10 Effective stiffness C33 (p.e:p.m=1:1)
10
MT
9.5 FEM
8.5
C33 (Pa)
7.5
6.5
5.5
0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6
Volume fraction of elastic matrix (m)
bers, shows nonlinear behavior. This may be due to a number of factors other than volume frac-
tion. Particular size, size distribution and inter-particular spacing have been suggested as addi-
tional parameters which can aect the predictions of eective moduli. These factors are
expected to become signicant as the volume fractions of the ber phases become large [31]. Large
dierences are observed for a few constitutive constants, especially the transverse shear modulus
C66. This is due to the fact that the transverse shear modulus is the most sensitive to the ber
packing [32]. The circular dot shown in Fig. 18 represents the rotational averaging value when
the phase volume fractions are 0.3, 0.3 and 0.4 for the piezoelectric ber, the piezomagnetic ber,
and the elastic matrix, respectively. It shows that the rotational averaging leads to better agree-
ment with the MoriTanaka method. Figs. 23 and 24 indicate that the magneto-electric coupling
J. Lee et al. / International Journal of Engineering Science 43 (2005) 790825 815
constants are maximum when the piezoelectric volume fraction and the piezomagnetic volume
fraction are approximately in a 1:1 ratio.
Numerical calculations for four dierent elastic matrix volume fraction (vm) cases are obtained
for a xed volume fraction of 1:1 ratio of the piezoelectric phase to the piezomagnetic phase. The
ratio of 1:1 is chosen at which the magneto-electric coupling constants show their maximum value.
In this case, the eective elastic, dielectric, magnetic, piezoelectric, piezomagnetic and magneto-
electric coupling moduli of a three-phase composite are generated for 0.3 6 mm 6 0.6. The FEM
results are compared with the MoriTanaka method results in Figs. 2536, in which the eective
moduli determined by the FEA are presented by triangles.
x 10
9 Effective stiffness C44 (p.e:p.m=1:1)
9
MT
FEM
8
7
(Pa)
6
44
5
C
2
0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6
Volume fraction of elastic matrix ( m)
x 10
9 Effective stiffness C66 (p.e:p.m=1:1)
12
MT
11 FEM
10
8
(Pa)
7
66
C
2
0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6
Volume fraction of elastic matrix ( m)
e (C/m2)
4
1
0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6
Volume fraction of elastic matrix (m)
60
40
20
20
0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6
Volume fraction of elastic matrix ( m)
The eective magneto-electric constants of the three-phase composite decrease as the matrix
volume fraction increases.
The dierences between the two methods decrease as the volume fraction of the elastic matrix
increases. The FEM results and MoriTanaka method results coincide when the matrix volume
fraction is 0.6.
(C2/Nm2) 3
2.5
1.5
0.5
0
0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6
Volume fraction of elastic matrix (m)
4
(Ns2/C2)
0
0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6
Volume fraction of elastic matrix (m)
phase volume fractions are 0.3, 0.3 and 0.4 for the piezoelectric ber, the piezomagnetic ber, and
the elastic matrix, respectively. Recall that the bers are aligned in the x3 direction. Table 4 shows
the magneto-electric coecient determined by the FEM for each case.
Arrows shown in Fig. 37 denote the poling directions of each phase tabulated in Table 3.
In Case (1) and Case (4), where the piezoelectric phase and the piezomagnetic phase are poled
to the same direction, <E1> is coupled with <H1> by the magneto-electric constant k11. In the
same way, <E2> and <H2> are coupled, <E3> and <H3> are coupled by k11 and k33, respectively.
In Case (2) and Case (3), by the eect of dierent poling directions, <E1> couples with <H3> and
<E3> couples with <H1>. In Case (4), the magnitude of k33, which is coupled through e23, is one
order larger than that of Case (1) which is coupled through e31.
818 J. Lee et al. / International Journal of Engineering Science 43 (2005) 790825
11 (Ns/VC) 5
0
0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6
Volume fraction of elastic matrix (m)
2
(Ns/VC)
33
0
0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6
Volume fraction of elastic matrix ( m)
Table 3
Poling directions of each case
Piezoelectric phase poling direction Piezomagnetic phase poling direction
Case (1) (0, 0, 1) (0, 0, 1)
Case (2) (1, 0, 0) (0, 0, 1)
Case (3) (0, 0, 1) (1, 0, 0)
Case (4) (0, 1, 0) (0, 1, 0)
J. Lee et al. / International Journal of Engineering Science 43 (2005) 790825 819
Table 4
Magnetoelectric coupling coecients for four dierent poling cases
k11 (Ns/VC) k22 (Ns/VC) k33 (Ns/VC) k13 (Ns/VC) k31 (Ns/VC)
Case (1) 2.689E13 2.700E13 1.169E10 0 0
Case (2) 0 0 0 1.664E11 2.928E12
Case (3) 0 0 0 3.472E12 3.543E11
Case (4) 5.719E13 6.887E14 1.846E9 0 0
Fig. 37. Schematic of poling directions for electro-magneto-elastic composites shown in Table 3.
4. Conclusions
The FEM was used to solve an electro-magneto-elastic boundary value problem within a peri-
odic unit cell of a brous composite and to determine the eective properties using periodic
boundary conditions. The eective properties were calculated for a three-phase electro-mag-
neto-elastic composite consisting of piezoelectric and piezomagnetic bers separated by an isotro-
pic linear elastic matrix. Although the piezoelectric, piezomagnetic and elastic phases do not
exhibit magneto-electric coupling, the resulting composite has non-zero magneto-electric eective
response as a result of the interaction between the piezoelectric and piezomagnetic bers through
the elastic matrix.
As a special case of the three-phase composite, the eective properties of a two-phase composite
consisting of a piezomagnetic matrix reinforced with piezoelectric bers are also determined using
the FEM. The rotationally averaged results were in excellent agreement with the predictions of the
MoriTanaka method of Li and Dunn [4].
The eective magneto-electric moduli of the three-phase composite were about 20 times less
than the magneto-electric moduli of the two-phase composite because the elastic matrix used in
the numerical simulation had low stiness to represent a typical epoxy, and was unable to transfer
signicant strains between the piezomagnetic and piezoelectric bers.
In addition to the FEM, the extension of the MoriTanaka method has been implemented for
the three-phase composite following the approach of Li and Dunn [4]. Through the parametric
study, it was seen that the eective magneto-electric coupling moduli are maximum when the
piezoelectric volume fraction and the piezomagnetic volume fraction are in a 1:1 ratio and de-
crease as the matrix volume fraction increases. It was also demonstrated that the magneto-electric
820 J. Lee et al. / International Journal of Engineering Science 43 (2005) 790825
coupling can be controlled by changing the poling directions of the piezoelectric and piezomag-
netic bers. Future work may include a systematic study of the eects of the following variables:
the type of periodic unit cell, the shape of the phase and the stiness of the elastic matrix.
Acknowledgement
This paper is based upon work supported by the Texas Advanced Research (Technology) Pro-
gram under Grant No. 000512-0005-2001. This paper is also based upon work supported by
NASA award number NCC-1-02038 and the Texas Institute for Intelligent Bio-Nano Materials
and Structure (TiiMS). Any opinions, ndings, and conclusions or recommendations expressed
in this material are those of the authors and do not necessarily reect the views of the National
Aeronautics and Space Administration.
Appendix A. Explicit form of eld equations for piezoelectric and piezomagnetic phases
The partial dierential eld equations for the piezoelectric phase and the piezomagnetic phase
with x3 as a symmetry axis are given by the following expressions:
Piezoelectric
o2 u1 o2 u1 o2 u1 o2 u2 o2 u3 o2 Ue
C 11 C 66 C 44 C 12 C 66 C 13 C 44 e 15 e31 0
ox21 ox22 ox23 ox1 ox2 ox1 ox3 ox1 ox3
o2 u2 o2 u2 o2 u2 o2 u1 o2 u3 o2 Ue
C 66 2 C 11 2 C 44 2 C 12 C 66 C 13 C 44 e15 e31 0
ox1 ox2 ox3 ox1 ox2 ox2 ox3 ox2 ox3
o2 u3 o2 u3 o2 u3 o2 u1 o2 u2
C 44 C 44 C 33 C 13 C 44 C 13 C 44
ox21 ox22 ox23 ox1 ox3 ox2 ox3
o 2 Ue o2 Ue o 2 Ue
e15 e 15 e33 0
ox21 ox22 ox23
o2 u3 o2 u3 o2 u3 o2 u1 o2 u2
e15 e15 e33 e31 e15 e31 e15
ox21 ox22 ox23 ox1 ox3 ox2 ox3
o 2 Ue o 2 Ue o2 Ue
g11 g 11 g 33 0
ox21 ox22 ox23
o 2 Um o 2 Um o 2 Um
l11 l11 l33 0
ox21 ox22 ox23
A:1
J. Lee et al. / International Journal of Engineering Science 43 (2005) 790825 821
Piezomagnetic
o2 u1 o2 u1 o2 u1 o2 u2 o2 u3 o 2 Um
C 11 C 66 C 44 C 12 C 66 C 13 C 44 q15 q31 0
ox21 ox22 ox23 ox1 ox2 ox1 ox3 ox1 ox3
o2 u2 o2 u2 o2 u2 o2 u1 o2 u3 o 2 Um
C 66 C 11 C 44 C 12 C 66 C 13 C 44 q15 q31 0
ox21 ox22 ox23 ox1 ox2 ox2 ox3 ox2 ox3
o2 u3 o2 u3 o2 u3 o2 u1 o2 u2
C 44 C 44 C 33 C 13 C 44 C 13 C 44
ox21 ox22 ox23 ox1 ox3 ox2 ox3
o2 Um o2 Um o2 Um
q15 q 15 q 33 0
ox21 ox22 ox23
o2 u3 o2 u3 o2 u3 o2 u1 o2 u2
q15 q 15 q 33 q31 q 15 q 31 q15
ox21 ox22 ox23 ox1 ox3 ox2 ox3
o2 Um o2 Um o2 Um
l11 l 11 l 33 0
ox21 ox22 ox23
o2 Ue o 2 Ue o2 Ue
g11 g11 g 33 0
ox21 ox22 ox23
A:2
The eld equations for the linear elastic matrix take the usual form (Lame equations).
The eective moduli of the three-phase electro-magneto-elastic composite are diagonally sym-
metric. In Appendix B, the diagonal symmetry of the eective moduli is proved. R, L and Z are
(12 1), (12 12) and (12 1) matrices respectively. This matrix notation can be expressed with
index notation as
Ri Lij Z j ; i; j 1; 2; 3; . . . 12 B:1
Each phase of the multi-phase composite satises the symmetry conditions
n n
Lij Lji ; i; j 1; 2; 3; . . . 12 B:2
In this paper, local elds will be denoted by an argument (x), and quantities without an argu-
ment will refer to averages.
Suppose that a representative volume V of a heterogeneous medium is subjected to two dierent
states of uniform overall Zj and Z 0j , the actual local elds are denoted as Ri(x), Zj(x) and R0i x,
Z 0j x. They satisfy the following relationships.
Z Z
1 1
hRi xi Ri xdV Lij xZ j xdV B:3
V V
Z Z
1 1
hR0i xi R0i xdV Lij xZ 0j xdV B:4
V V
822 J. Lee et al. / International Journal of Engineering Science 43 (2005) 790825
By the denition
hRi xi Lij hZ j xi B:5
hR0i xi Lij hZ 0j xi B:6
Because Lij(x) = Lji(x)
Ri xZ 0i x Lij xZ j xZ 0i x Lji xZ 0i xZ j x R0j xZ j x B:7
so Z Z
1 0 1
Ri xZ i xdV R0j xZ j xdV B:8
V V
hRi xZ 0i xi hR0j xZ j xi B:9
The HillMandel condition can be used to rewrite Eq. (B.9) as
hRi xihZ 0i xi hR0j xihZ j xi B:10
Lij hZ j xihZ 0i xi Lji hZ 0i xihZ j xi B:11
Lij Lji B:12
Thus the eective moduli are diagonally symmetric.
Appendix D. Rotational averaging and verifying the transverse isotropy of rotationally averaged
eective properties of a two-phase composite
where Rij is the direction cosines for a rotation about the x3 axis. Then each of the tensors is
remapped to the engineering notation.
The stiness matrix of transversely isotropic materials with x3 the axis of symmetry has six dif-
ferent components, C11, C12, C13, C33, C44 and C66. However, there exist only ve independent
components by the relation between the components as following:
C 11 C 12
C 66 0 D:7
2
To verify the transverse isotropy, the relation between the normalized components for the rota-
tional variation about the x3 axis is plotted in Fig. D.1.
avg
3
x 10 ((C C )/2C66 )/C )
11 12 11
1.5
1
((C C )/2C66 )/Cavg(Pa)
11
0.5
average value
0
12
0.5
11
1.5
0 45 90 135 180 225 270 315 360
(rotation angle)
Fig. D.1. Dierence distribution for rotational variation for a two-phase composite.
824 J. Lee et al. / International Journal of Engineering Science 43 (2005) 790825
The rotationally averaged components of the stiness properties satisfy the following equation:
C avg avg
11 C 12
C avg
66 0 D:8
2
Thus, the rotationally averaged stiness matrix Cavg of the two-phase composite obtained by the
FEM shows the transverse isotropy.
References
[1] C.W. Nan, Magnetoelectric eect in composites of piezoelectric and piezomagnetic phases, Phys. Rev. B50 (1994)
60826088.
[2] D.C. Lagoudas, J.G. Boyd, Z. Bo, Micromechanics of active composites with SMA bers, J. Eng. Mater. Tech. 116
(1994) 337345.
[3] J.Y. Li, On micromechanics approximation for the eective thermoelastic moduli of multi-phase composite
materials, Mech. Mater. 31 (1999) 149159.
[4] J.Y. Li, M.L. Dunn, Micromechanics of magnetoelectroelastic composite materials: Average elds and eective
behavior, J. Intell. Mater. Syst. Struct. 9 (1998) 404416.
[5] J. Aboudi, Micromechanical analysis of fully coupled electro-magneto-thermo-elastic multiphase composites,
Smart Mater. Struct. 10 (2001) 867877.
[6] J.H. Huang, W.S. Kuo, The analysis of piezoelectric/piezomagnetic composite materials containing ellipsoidal
inclusions, J. Appl. Phys. 81 (1997) 13781386.
[7] J.Y. Li, M.L. Dunn, Anisotropic coupled-eld inclusion and inhomogeneity problems, Philos. Mag. A 77 (1998)
13411350.
[8] J.Y. Li, M.L. Dunn, Analysis of microstructural elds in heterogeneous piezoelectric solids, Int. J. Eng. Sci. 37
(1999) 665685.
[9] A.M.J.G. Van Run, D.R. Terrell, J.H. Scholing, An in situ grown eutectic magneto-electric composite material,
J. Mater. Sci. 9 (1974) 17101714.
[10] L.P.M. Bracke, R.G. Van Vliet, A broadband magneto-electric transducer using a composite material, Int. J.
Electron. 51 (1981) 255262.
[11] G. Harshe, Magnetoelectric eect in piezoelectric/magnetostrictive composites, Ph.D. dissertation, The Pennsyl-
vania State University, University Park, USA, 1991.
[12] G. Harshe, J.P. Dougherty, R.E. Newnham, Theoretical modeling of 30/03 magneto-electric composite, Int. J.
Appl. Electromagn. Mater. 4 (1993) 161171.
[13] M. Avellaneda, G. Harshe, Magnetoelectric eect in piezoelectric/magnetostrictive multilayer (22) composites,
J. Intell. Mater. Syst. Struct. 5 (1994) 501513.
[14] Y. Benveniste, Magnetoelectric eect in brous composites with piezoelectric and piezomagnetic phases, Phys. Rev.
B51 (1995) 1642416427.
[15] J.H. Huang, Analytical predictions for the magnetoelectric coupling in piezomagnetic materials reinforced by
piezoelectric ellipsoidal inclusions, Phys. Rev. B58 (1998) 1215.
[16] T.L. Wu, J.H. Huang, Closed-form solutions for the magneto-electric coupling coecients in brous composites
with piezoelectric and piezomagnetic phases, Int. J. Solids Struct. 37 (2000) 29813009.
[17] J.H. Huang, H.K. Liu, W.L. Dai, The optimized ber volume fraction for magneto-electric coupling eect in
piezoelectricpiezomagnetic continuous ber reinforced composites, Int. J. Eng. Sci. 38 (2000) 12071217.
[18] J.Y. Li, Magnetoelectroelastic multi-inclusion and inhomogeneity problems and their applications in composite
materials, Int. J. Eng. Sci. 38 (2000) 19932011.
[19] J.G. Boyd, D.C. Lagoudas, C.S. Seo, Arrays of micro-electrodes and electromagnets for processing of electro-
magneto-elastic multifunctional composite materials, in: SPIE 10th Annual International Symposium on Smart
Structures held at San Diego, CA March 26, vol. 5055, 2003, pp. 268277.
J. Lee et al. / International Journal of Engineering Science 43 (2005) 790825 825
[20] J. Lee, J.G. Boyd, D.C. Lagoudas, Three-phase electro-magneto-elastic multifunctional composite materials, in:
SPIE 10th Annual International Symposium on Smart Structures held at San Diego, CA March 26, vol. 5053,
2003.
[21] S. Li, General unit cells for micromechanical analyses of unidirectional composites, Compos. Part, A.: Appl. Sci.
Manufact. 32 (2001) 815826.
[22] C.T. Sun, R.S. Vaidya, Prediction of composite properties from a representative volume element, Compos. Sci.
Tech. 56 (1996) 171179.
[23] J.D. Achenbach, H. Zhu, Eect of interphases on micro and macromechanical behavior of hexagonal-array ber
composites, ASME J. Appl. Mech. 57 (1990) 956963.
[24] J. Noe, J. Whitcomb, Eect of various parameters on the eective properties of a cracked ply, J. Compos. Mater.
35 (2001) 689712.
[25] M. Hori, S. Nemat-Nasser, On two micromechanics theories for determining micromacro relations in
heterogeneous solids, Mech. Mater. 31 (1999) 667682.
[26] L. Pan, D.O. Adams, F.J. Rizzo, Boundary element analysis for composite materials and a library of Greens
functions, Comput. Struct. 66 (1998) 685693.
[27] J.D. Whitcomb, C.D. Chapman, X. Tang, Derivation of boundary conditions for micromechanics analyses of plain
and satin weave composites, J. Compos. Mater. 34 (2000) 724747.
[28] M. Kaminski, Homogenized properties of periodic n-component composites, Int. J. Eng. Sci. 38 (2000) 405427.
[29] J.Y. Li, M.L. Dunn, Anisotropic coupled-eld inclusion and inhomogeneity, Problems Philos. Mag. A 77 (1998)
13411350.
[30] J.H. Huang, Y.H. Chiu, H.K. Liu, Magneto-electro-elastic Eshelby tensors for a piezoelectricpiezomagnetic
composite reinforced by ellipsoidal inclusions, J. Appl. Phys. 83 (1998) 53645370.
[31] P. Kwon, C.K.H. Dharan, Eective moduli of high volume fraction particulate composites, Acta Metall. Mater. 43
(1995) 11411147.
[32] C. Yan, On homogenization and de-homogenization of composite materials, Doctorial thesis, Drexel University,
2003.