07.11.2014 Views

2006 Magneto-elastic.. - UCLA Engineering

2006 Magneto-elastic.. - UCLA Engineering

2006 Magneto-elastic.. - UCLA Engineering

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

ARTICLE IN PRESS<br />

Journal of the Mechanics and Physics of Solids<br />

54 (<strong>2006</strong>) 975–1003<br />

www.elsevier.com/locate/jmps<br />

<strong>Magneto</strong>-<strong>elastic</strong> modeling of composites containing<br />

chain-structured magnetostrictive particles<br />

H.M. Yin a,b , L.Z. Sun a,c, , J.S. Chen d<br />

a Department of Civil and Environmental <strong>Engineering</strong> and Center for Computer-Aided Design,<br />

The University of Iowa, Iowa City, IA 52242, USA<br />

b Department of Civil and Environmental <strong>Engineering</strong>, University of Illinois at Urbana-Champaign,<br />

Urbana, IL 61801, USA<br />

c Department of Civil and Environmental <strong>Engineering</strong>, University of California, Irvine, CA 92697, USA<br />

d Department of Civil and Environmental <strong>Engineering</strong>, University of California, Los Angeles, CA 90095, USA<br />

Received 2 May 2005; received in revised form 21 September 2005; accepted 22 November 2005<br />

Abstract<br />

<strong>Magneto</strong>-<strong>elastic</strong> behavior is investigated for two-phase composites containing chain-structured<br />

magnetostrictive particles under both magnetic and mechanical loading. To derive the local magnetic<br />

and <strong>elastic</strong> fields, three modified Green’s functions are derived and explicitly integrated for the<br />

infinite domain containing a spherical inclusion with a prescribed magnetization, body force, and<br />

eigenstrain. A representative volume element containing a chain of infinite particles is introduced to<br />

solve averaged magnetic and <strong>elastic</strong> fields in the particles and the matrix. Effective magnetostriction<br />

of composites is derived by considering the particle’s magnetostriction and the magnetic interaction<br />

force. It is shown that there exists an optimal choice of the Young’s modulus of the matrix and the<br />

volume fraction of the particles to achieve the maximum effective magnetostriction. A transversely<br />

isotropic effective <strong>elastic</strong>ity is derived at the infinitesimal deformation. Disregarding the interaction<br />

term, this model provides the same effective <strong>elastic</strong>ity as Mori–Tanaka’s model. Comparisons of<br />

model results with the experimental data and other models show the efficacy of the model and<br />

Corresponding author. Department of Civil and Environmental <strong>Engineering</strong>, University of California, Irvine,<br />

CA 92697, USA. Tel.: +1 949 824 8670; fax: +1 949 824 2117.<br />

E-mail address: lsun@uci.edu (L.Z. Sun).<br />

0022-5096/$ - see front matter r 2005 Elsevier Ltd. All rights reserved.<br />

doi:10.1016/j.jmps.2005.11.007


976<br />

ARTICLE IN PRESS<br />

H.M. Yin et al. / J. Mech. Phys. Solids 54 (<strong>2006</strong>) 975–1003<br />

suggest that the particle interactions have a considerable effect on the effective magneto-<strong>elastic</strong><br />

properties of composites even for a low particle volume fraction.<br />

r 2005 Elsevier Ltd. All rights reserved.<br />

Keywords: Microstructure; Elastic material; Particulate reinforced composites; <strong>Magneto</strong>striction; <strong>Magneto</strong>-<strong>elastic</strong><br />

modeling<br />

1. Introduction<br />

In the past few years, applications of magnetic particle-filled composites as a class of<br />

smart materials have attracted a good deal of attention from engineers and researchers (Jin<br />

et al., 1992; Sandlund et al., 1994; Duenas and Carman, 2000; Borcea and Bruno, 2001).<br />

Composites manufactured by brittle magnetic particles and a strong matrix allow the<br />

magnetic properties of the particles to be maintained while keeping overall mechanical<br />

performance adjustable (Sandlund et al., 1994). Composites made of ferromagnetic<br />

particles and a soft matrix belong to another specific class of smart materials where the<br />

mechanical properties can be changed under different magnetic environments. These<br />

composites have been applied to automotive components for isolation of car noise,<br />

vibration, and harshness (Ginder et al., 2000).<br />

In recent years it has been found that fabricating the composite under an applied electric<br />

or magnetic field transforms the randomly dispersed particles into chains parallel to the<br />

direction of the applied field due to the interaction force (Halsey, 1992; Sandlund et al.,<br />

1994; Ginder et al., 1999). Upon curing the matrix material, the chain structure is locked<br />

into place so that the chain-like microstructures in the composites are obtained. When a<br />

magnetic field is further applied, an effective magnetostriction has been observed due to<br />

the particle’s magnetostriction and the magnetic interaction force between particles.<br />

In general, magnetostriction is relatively small for most typical ferromagnetic materials<br />

such as iron, nickel, and cobalt, and saturation magnetostriction is of the order 10 5 –10 6<br />

(Bozorth, 1951; Bednarek, 1999). In the 1960s, researchers were surprised by the<br />

measurement of giant magnetostriction in the range of 0.2–0.6% for some rare earth<br />

crystals like Terfenol-D measured at a variety of temperatures (McKnight, 2002).<br />

Sandlund et al. (1994) measured the effective magnetostriction, <strong>elastic</strong> moduli, and<br />

magneto-mechanical coupling factor for Terfenol-D composites. Herbst et al. (1997)<br />

fabricated magnetostrictive composites consisting of SmFe 2 embedded in an Fe or Al<br />

matrix and proposed a single-sphere model to predict overall magnetostriction of the<br />

composites. Chen et al. (1999) examined Terfenol-D particles filled in various kinds of<br />

matrices and investigated the effect of the matrix’s <strong>elastic</strong> modulus on effective<br />

magnetostriction. Guo et al. (2001) also studied the effect of the matrix’s <strong>elastic</strong> modulus<br />

on the magneto-mechanical coupling and effective magnetostriction of Terfenol-D<br />

composites. Duenas and Carman (2000) evaluated the magnetostrictive response of<br />

Terfenol-D composites and evaluated the effective <strong>elastic</strong>ity by Voight and Reuss<br />

approximations (Mura, 1987). Anjanappa and Wu (1997) and Friedmann et al. (2001)<br />

showed some unique applications of magnetostrictive particulate composites as actuators<br />

or sensors. Nan (1998) and Nan and Weng (1999) developed an analytical model based on<br />

the Green’s function technique. Since this model is based on the solution for one particle<br />

with magnetostriction embedded in the matrix, they are unable to take into account either


ARTICLE IN PRESS<br />

H.M. Yin et al. / J. Mech. Phys. Solids 54 (<strong>2006</strong>) 975–1003 977<br />

particle distribution or interaction between particles. Armstrong (2000) studied composites<br />

containing a large number of well-distributed ellipsoidal magnetostrictive particles and<br />

presented an analysis of magneto-<strong>elastic</strong> behavior based on Eshelby’s theory. Feng et al.<br />

(2003) extended the double-inclusion model to predict magneto-<strong>elastic</strong> behavior of these<br />

composites.<br />

In the above-mentioned models, magnetic force between particles is not considered, and<br />

only the magnetostriction of magnetostrictive particles is accounted for in the effective<br />

magnetostriction. Thus, these models are only applicable to composites whose matrix<br />

moduli are sufficiently high or whose magnetic permeability ratio of the particle phase to<br />

the matrix phase is very small. However, for some magnetostrictive composites like<br />

Terfenol-D filled composites where the matrix <strong>elastic</strong> modulus may not so be high enough<br />

and the particles’ relative magnetic permeability can also reach as high as 15 times as the<br />

matrix (McKnight, 2002), the magnetic forces between particles cannot be simply<br />

disregarded since they may play an important role on the effective magneto-<strong>elastic</strong><br />

behavior of the magnetostrictive composites.<br />

To study the effect of magnetic force on magneto-mechanical behavior of these<br />

composites, Davis (1992) applied the classical point dipole force (Coulson, 1961) to<br />

consider pair-wise interaction between particles. Gast and Zukoski (1989) and Klingenberg<br />

and Zukoski (1990) used the multipole expansion method to derive the interaction force<br />

for the general case in which the center–center line is at an arbitrary angle with respect to<br />

the external field, and this approach provided an approximate solution with three<br />

dimensionless functions generated by curve fitting. Chen et al. (1991) numerically solved<br />

local interaction for a chain of spherical particles in electrorheological fluid from<br />

Rayleigh’s theory (Rayleigh, 1892). The numerical results showed that this model has good<br />

agreement with experimental data but is obviously different from Klingenberg and<br />

Zukoski’s (1990) model. In addition, Davis (1992) used the finite element method to solve<br />

the interaction force and also discovered this difference.<br />

The objective of this paper is to investigate effective magneto-<strong>elastic</strong> behavior of<br />

composites containing chain-structured magnetostrictive particles. Here, the particles are<br />

assumed to be ideal soft magnetic materials, i.e. the residual magnetic field is always zero<br />

when the external magnetic field is removed. For simplicity, we assume that applied<br />

magnetic and <strong>elastic</strong> loading is in the linear range, both the particle and matrix phases have<br />

isotropic magnetic and <strong>elastic</strong> properties, and the magnetic permeability for each phase is<br />

assumed to be independent to the mechanical loading such that the magnetic permeability<br />

and <strong>elastic</strong>ity of the particles and matrix are treated constant. For the magnetostatic theory<br />

to be applicable, magnetic loading is assumed to be at a low frequency and the size of the<br />

particles is much smaller than the wavelength of the magnetic field. Given magnetic and<br />

mechanical loading, we can solve the averaged stress and strain in the particle and matrix<br />

phases. From the relation between mechanical responses and magnetic or mechanical<br />

loading, we can derive the effective magnetostriction and <strong>elastic</strong>ity of composites.<br />

The rest of this paper is organized as follows. In Section 2, Green’s functions for<br />

magnetic and <strong>elastic</strong> problems are reviewed, and the modified Green’s functions for the<br />

homogeneous infinite medium including a spherical inclusion with prescribed magnetization,<br />

body force, and strain are derived. Using the explicit integral of these Green’s<br />

function, the local magnetic and <strong>elastic</strong> fields can be obtained. In Section 3, Green’s<br />

functions and Eshelby’s equivalent inclusion method are employed in micromechanical<br />

analysis of chain-structured composites. A representative volume element (RVE)


978<br />

ARTICLE IN PRESS<br />

H.M. Yin et al. / J. Mech. Phys. Solids 54 (<strong>2006</strong>) 975–1003<br />

containing a chain of infinite particles is used to solve the averaged magnetic fields and<br />

<strong>elastic</strong> fields for composites under magnetic and mechanical loading with consideration of<br />

particle interactions. The effective magnetostriction due to the particle’s magnetostriction<br />

and the magnetic interaction force are shown in Section 4. Under the infinitesimal<br />

deformation, effective <strong>elastic</strong>ity is proved to be transversely isotropic. Some comparisons<br />

of numerical solution with experimental data and other models and parametric analysis are<br />

also presented in this section. Concluding remarks are given in Section 5.<br />

2. Basic theories of micromechanics<br />

Micromechanics encompasses mechanics related to microstructures of composites and it<br />

studies the effective behavior of composites. The method employed is a continuum theory<br />

based on local fields at the microscale. Due to material inhomogeneities, the local fields are<br />

significantly disturbed even under a macroscopically uniform loading. Green’s functions<br />

are widely used to solve this class of problems. Green’s function is an integral kernel that<br />

has been used to solve an inhomogeneous differential equation with boundary conditions.<br />

The Green’s function for the infinite domain is a scalar function for magnetic problem and<br />

a second-order tensorial function for <strong>elastic</strong> problem (Kro¨ner, 1990).<br />

Modified Green’s functions are defined as the derivatives of the conventional Green’s<br />

functions. Although Green’s functions have been one of the most powerful tools in the past<br />

century to solve temperature, electric, magnetic and <strong>elastic</strong> fields (Morse and Feshbach,<br />

1953; Mura, 1987; Jackson, 1999) in infinite medium with point sources, the method was<br />

not widely used in material science until Eshelby (1957, 1959) obtained an explicit solution<br />

for an ellipsoidal inhomogeneity within an isotropic matrix. Since then, Mura (1963, 1987),<br />

Kro¨ner (1972, 1990), and Mazilu (1972) have done extensive work in this field. Willis<br />

(1965), Indenbom and Orlov (1968), Mura and Kinoshita (1971), andPan and Chou<br />

(1976) have extended Green’s functions to anisotropic materials. Recently, Wang (1992),<br />

Chen (1993), Nan (1994), Dunn (1994), Michelitsch (1997), Huang et al. (1998), Gao and<br />

Fan (1998), and Li (2002) have obtained Green’s functions for piezomagnetic and<br />

piezoelectric solids.<br />

In this work, we consider magnetostrictive inhomogeneities distributed in a nonmagnetic<br />

matrix. Under magnetic and mechanical loading, the inhomogeneities are magnetized,<br />

magnetic forces are induced between the particles, and the <strong>elastic</strong> fields are also<br />

significantly disturbed due to the stiffness mismatch between the particles and matrix. In<br />

this section, modified Green’s functions are derived for three conditions: (1) a magnetic<br />

field caused by a magnetization, (2) a displacement field caused by a body force (magnetic<br />

force), and (3) a displacement field induced by a prescribed in<strong>elastic</strong> strain (magnetostriction),<br />

which is also called eigenstrain. The integrals of modified Green’s functions over a<br />

spherical domain can be explicitly derived and these integrals can be employed to obtain<br />

the local magnetic and <strong>elastic</strong> fields.<br />

2.1. Magnetic field caused by magnetization<br />

Green’s theorem reads (Jackson, 1999):<br />

Z<br />

Z <br />

ður 2 v vr 2 uÞ dr ¼ u qv v qu <br />

dr, (1)<br />

qn qn<br />

V<br />

qV


where u and v are functions with continuous second order derivatives in a region V with<br />

surface qV. The above equation transfers an integral over a domain to one over the<br />

boundary. We know for function v ¼ 1=jr r 0 j that<br />

r 2 v ¼ 4pdðr r 0 Þ. (2)<br />

Thus, using Eqs. (1) and (2), we can write<br />

uðr 0 Þ¼ 1 Z<br />

r 2 u<br />

4p V jr r 0 j dr þ 1 Z <br />

1<br />

u<br />

4p qV jr r 0 j 2 þ 1 <br />

qu<br />

dr. (3)<br />

jr r 0 j qn<br />

For linear isotropic magnetic materials, the magnetic problem is reduced to finding a<br />

solution to the following Poisson’s equation with boundary conditions (Reitz et al., 1979):<br />

r 2 UðrÞ ¼ qM kðrÞ<br />

, (4)<br />

qr k<br />

where U is the scalar magnetic potential, and M k is the magnetization. For numerable<br />

magnetic particles filled in the infinite nonmagnetic domain, the magnetic field falls off<br />

faster than 1=jrj for r !1 (Jackson, 1999). Considering the identity of Eq. (3) in the<br />

magnetic potential governed by Eq. (4) with the boundary integral vanished, the magnetic<br />

potential can be expressed as<br />

Z<br />

UðrÞ ¼ Gðr; r 0 Þ qM kðr 0 Þ<br />

V qr 0 dr 0 , (5)<br />

k<br />

where Green’s function Gðr; r 0 Þ describes the magnetic potential at point r due to the<br />

magnetization at point r 0 in the infinite domain. Here, we have<br />

Gðr; r 0 1<br />

Þ¼<br />

4pjr r 0 j . (6)<br />

Because M i ðrÞ ¼0 for r !1, Eq. (5) can be rewritten as<br />

Z<br />

qGðr; r 0 Þ<br />

UðrÞ ¼<br />

M k ðr 0 Þ dr 0 , (7)<br />

V qr k<br />

where qGðr; r 0 Þ=qr 0 k ¼ qGðr; r0 Þ=qr k is used.<br />

The magnetic field is related to the magnetic potential as<br />

H i ¼ U ;i . (8)<br />

Then, the substitution of Eq. (7) into Eq. (8) yields<br />

Z<br />

H i ðrÞ ¼ G m ik ðr; r0 ÞM k ðr 0 Þ dr 0 , (9)<br />

V<br />

where the modified Green’s function is<br />

ARTICLE IN PRESS<br />

H.M. Yin et al. / J. Mech. Phys. Solids 54 (<strong>2006</strong>) 975–1003 979<br />

G m ik ðr; r0 Þ¼ q2 Gðr; r 0 Þ<br />

. (10)<br />

qr i qr k<br />

Consider one spherical magnetic particle with radius ‘‘a’’ occupying domain O with<br />

uniform magnetization M 0 i embedded in the infinite nonmagnetic domain V. Outside O,<br />

the magnetization is zero. Then, from Eq. (9), the local magnetic field reads<br />

H i ðrÞ ¼D m ik ðr r0 ÞM 0 k . (11)


980<br />

ARTICLE IN PRESS<br />

H.M. Yin et al. / J. Mech. Phys. Solids 54 (<strong>2006</strong>) 975–1003<br />

Here D m ik ðr r0 Þ¼ R O Gm ik ðr; r0 Þ dr 0 can be written in explicit form as<br />

(<br />

1<br />

D m ij ðr r0 3<br />

Þ¼<br />

r3 ðd ij 3n i n j Þ for r4a;<br />

1<br />

3 d (12)<br />

ij for rXa;<br />

where r ¼jr r 0 j, r ¼ a=r, and n i ¼ðr i r 0 i Þ=r, r0 is the center of the particle. For a linear<br />

isotropic magnetic particle with magnetic permeability m m 1 filled in the infinite nonmagnetic<br />

domain under a far field H 0 i , the magnetic constitutive law provides<br />

M i ðrÞ ¼ mm 1 m m 0<br />

m m H i ðrÞ, (13)<br />

0<br />

where the magnetic permeability in vacuum reads<br />

m m 0 ¼ 4p 10 7 Ns 2 =C 2 . (14)<br />

Here N, s and C are units for force, time and electric charge, respectively. It is noted that<br />

the magnetic permeability for nonmagnetic materials is the same as that in vacuum. Using<br />

Eshelby’s equivalent inclusion method (Eshelby, 1957, 1959), we can simulate the material<br />

mismatch by introducing a distributed magnetization on the particle domain. The total<br />

domain is therefore treated as a homogeneous material with thermal conductivity m m 0<br />

subjected to a uniform field and a prescribed magnetization on the particle domain. The<br />

magnetization is solved based on the equivalent condition that the magnetic induction in<br />

the real particle with permeability m m 1 is same as that in the equivalent particle with<br />

permeability m m 0 . This method has been used to solve the thermal conduction problem<br />

(Hatta and Taya, 1986; Yin et al., 2005). From Eqs. (11) and (13), we can solve<br />

magnetization of the particle as<br />

M i ¼ 3 mm 1 m m 0<br />

m m 1 þ H 0<br />

2mm i (15)<br />

0<br />

and the magnetic field can be expressed in two parts: the far field and the perturbed field<br />

from the magnetization of the particle, i.e.,<br />

H i ðrÞ ¼H 0 i þ D m ij ðr r0 ÞM j , (16)<br />

which is same as the solution obtained by the variable separation method (von Hippel,<br />

1954).<br />

2.2. Strain field caused by body force<br />

For linearly isotropic materials, the mechanical problem is reduced to finding a solution<br />

to the following equation with boundary conditions:<br />

C ijkl u k;lj ðrÞþf i ðrÞ ¼0, (17)<br />

where C ijkl ¼ ld ij d kl þ m e ðd ik d jl þ d il d jk Þ is the isotropic stiffness tensor with the Lame<br />

constants l and m e , u k ðrÞ is the displacement field and f i ðrÞ is the body force. The <strong>elastic</strong><br />

Green’s function tensor reads (Mura, 1987)<br />

G ij ðr; r 0 Þ¼ 1<br />

4pm e jr<br />

d ij<br />

r 0 j<br />

1<br />

16pm e ð1<br />

q 2 jr r 0 j<br />

, (18)<br />

nÞ qr i qr j


which satisfies<br />

C ijkl G kp;lj ðr; r 0 Þ¼ d ip dðr r 0 Þ. (19)<br />

Here, n is Poisson’s ratio. Using this Green’s function, we can find that the displacement<br />

for an infinite homogeneous isotropic domain with local body force is written as<br />

Z<br />

u i ¼ G ij ðr; r 0 Þf j ðr 0 Þ dr 0 . (20)<br />

V<br />

Here, we use a boundary condition that the far field strains are zero. Based on the<br />

kinematic relation between strains and displacements, strains can be written as<br />

Z<br />

e ij ¼ G f ijk ðr; r0 Þf k ðr 0 Þ dr 0 , (21)<br />

V<br />

where the modified Green’s function is<br />

G f ijk ðr; r0 Þ¼ 1 2 ½G ik;jðr; r 0 ÞþG jk;i ðr; r 0 ÞŠ. (22)<br />

Consider an infinite domain V with shear modulus m e and Poisson’s ratio n containing a<br />

spherical domain O with a body force f 0 i , the local strain field is<br />

e ij ðrÞ ¼D f ijk ðr r0 Þf 0 k , (23)<br />

where D f ijk ðr<br />

D f ijk ðr<br />

r0 Þ¼ R O Gf ijk ðr; r0 Þ dr 0 . This integral can also be written in explicit form as<br />

8<br />

" #<br />

r 2 a ð5 10n þ 3r 2 Þðd ik n j þ d jk n i Þ<br />

><<br />

60m e ð1 nÞ þð 5 þ 3r 2 Þd ij n k þ 15ð1 r 2 for r4a;<br />

Þn i n j n k<br />

r<br />

>:<br />

30m e ð1 nÞ ½ð4 5nÞðd ikn j þ d jk n i Þ d ij n k Š for rXa:<br />

r0 Þ¼<br />

Note that the strain in O is linearly distributed.<br />

2.3. Strain field caused by eigenstrain<br />

The concept of eigenstrain was first introduced by Eshelby (1957) and was initially<br />

named by Mura (1987) to describe local in<strong>elastic</strong> strain, such as thermal expansion, phase<br />

transformation, initial strains, plastic strains and misfit strains. <strong>Magneto</strong>striction is also a<br />

kind of eigenstrain. In the following, Eshelby’s solution for a spherical inclusion with<br />

eigenstrain embedded in the infinite domain is reviewed.<br />

For linearly isotropic <strong>elastic</strong> material with local magnetostriction, the mechanical<br />

problem without body force is reduced to finding a solution to the following equation with<br />

boundary conditions:<br />

C ijkl ½u k;lj ðrÞ e kl;jðrÞŠ ¼ 0. (25)<br />

Using the same <strong>elastic</strong> Green’s function tensor Eq. (18), we can obtain the displacement<br />

field as<br />

Z<br />

u i ¼ C jlmn e mn ðr0 ÞG ij;l ðr; r 0 Þ dr 0 , (26)<br />

V<br />

ARTICLE IN PRESS<br />

H.M. Yin et al. / J. Mech. Phys. Solids 54 (<strong>2006</strong>) 975–1003 981<br />

(24)


982<br />

ARTICLE IN PRESS<br />

H.M. Yin et al. / J. Mech. Phys. Solids 54 (<strong>2006</strong>) 975–1003<br />

where the boundary condition with the far field strain being zero is used. Using the minor<br />

symmetry of C ijmn ¼ C jimn , Eq. (26) can be rewritten as<br />

u i ¼ 1 Z<br />

C jlmn e mn<br />

2<br />

ðr0 Þ½G ij;l ðr; r 0 ÞþG il;j ðr; r 0 ÞŠ dr 0 . (27)<br />

V<br />

From the kinematic relation, strains can be written as<br />

Z<br />

e ij ¼ G e ijmn ðr; r0 ÞC mnkl e kl ðr0 Þ dr 0 , (28)<br />

V<br />

where the modified Green’s function is<br />

G e ijmn ðr; r0 Þ¼ 1 4 ½G im;njðr; r 0 ÞþG in;mj ðr; r 0 ÞþG jm;ni ðr; r 0 ÞþG jn;mi ðr; r 0 ÞŠ. (29)<br />

In contrast to Kro¨ner’s (1990) original definition, which only has minor symmetry, we<br />

can find that this modified Green’s function has both minor and major symmetries, i.e.<br />

G e ijmn ðr; r0 Þ¼G e jimn ðr; r0 Þ¼G e mnij ðr; r0 Þ. As can be seen in the subsequent chapters, this change<br />

provides convenience in numerical operations as well as carrying a clear physical meaning.<br />

If in the infinite domain V with shear modulus m e and Poisson’s ratio n there is only a<br />

spherical domain O with an eigenstrain e 0<br />

ij , the local strain field can be written as<br />

e ij ðrÞ ¼ D e ijmn ðr r0 ÞC mnkl e 0<br />

kl , (30)<br />

where D e ijkl ðr r0 Þ¼ R O Ge ijkl ðr; r0 Þ dr 0 . This integral can also be written in explicit form<br />

(Mura, 1987; Ju and Sun, 1999) as<br />

D e ijkl ðr<br />

8<br />

r0 Þ<br />

r 3<br />

>< 60m e ð1<br />

¼<br />

2<br />

ð5 3r 2 Þd ij d kl ð5 10n þ 3r 2 3<br />

Þðd ik d jl þ d il d jk Þ<br />

15ð1 r 2 Þðd ij n k n l þ d kl n i n j Þ<br />

nÞ 15ðn r 2 6 Þðd ik n j n l þ d il n j n k þ d jk n i n l þ d jl n i n k Þ 7<br />

4<br />

5<br />

þ15ð5 7r 2 Þn i n j n k n l<br />

for r4a;<br />

1<br />

>:<br />

30m e ð1 nÞ ½d ijd kl ð4 5nÞðd ik d jl þ d il d jk ÞŠ for rpa:<br />

We can find that the strain field in the particle domain is still uniform.<br />

3. Micromechanical analysis<br />

Consider a two-phase chain-structured particulate composite containing magnetostrictive<br />

particles and the nonmagnetic matrix with isotropic permeability m m 1 and m m 0 , and<br />

isotropic <strong>elastic</strong>ity C 1 and C 0 , respectively, as seen in Fig. 1(a). Here m m 0 is defined in Eq.<br />

(14); C 1 may be written as C 1 ijkl ¼ l 1d ij d kl þ m e 1 ðd ikd jl þ d il d jk Þ with the Lame constants l 1<br />

and m e 1 ; C0 is also with the Lame constants l 0 and m e 0<br />

. The chains are randomly dispersed in<br />

the composite and have the same direction. The radius of particles is a, and the distance<br />

between two neighboring particles in one chain is assumed to be equal, denoted by b. Due<br />

to a bunching effect when fabricating the composite under a high magnetic field, the<br />

distance between two neighboring chains is generally much larger than b and r 0 ¼ a=b is<br />

close to 1 2 . A uniform magnetic field H0 i and a uniform stress field s 0 ij applied on the<br />

ð31Þ


ARTICLE IN PRESS<br />

H.M. Yin et al. / J. Mech. Phys. Solids 54 (<strong>2006</strong>) 975–1003 983<br />

H 0 X 0<br />

D<br />

(a)<br />

Ω 8<br />

Ω 6<br />

D<br />

Ω 4<br />

Ω 2<br />

x3<br />

a<br />

Ω 0<br />

Ω 1<br />

b<br />

x 2<br />

x 1<br />

Ω 3<br />

Ω 5<br />

(b)<br />

Ω 7<br />

Fig. 1. Chain structured composite under a uniform external magnetic field. (a) Overall microstructure; (b) RVE<br />

of the neighborhood of the material point X 0 .<br />

boundary of the composite are considered in this investigation of the magneto-<strong>elastic</strong><br />

response. Because the particle spacing in the same chain is much smaller than the distance<br />

between chains, the particle interactions in the same chain is dominant. Thus, we only


984<br />

ARTICLE IN PRESS<br />

H.M. Yin et al. / J. Mech. Phys. Solids 54 (<strong>2006</strong>) 975–1003<br />

consider the interactions of particles from the same chain and disregard those from<br />

particles in other chains. It should be noted that, when the volume fraction of particles is<br />

much higher, the effect of the chain spacing may become an important factor and should<br />

be considered in the future. In this section, we will study the averaged magnetic field, stress<br />

field and strain field considering the particle interactions.<br />

3.1. Averaged magnetic field<br />

From Maxwell’s equations and boundary conditions, the magnetostatic field in the<br />

composite satisfies<br />

B i;i ðrÞ ¼0 for r 2 D, (32)<br />

and<br />

H i ðrÞ ¼H 0 i for r 2 qD (33)<br />

½B þ i ðrÞ B i ðrÞŠl i ¼ 0 for r 2 qO p , (34)<br />

where superscripts + and denote the outside and inside surface on the interface, l i is the<br />

outward unit normal vector, the subscript D is the overall domain of the composite, and O p<br />

is the domain of the pth particle ðp ¼ 1; 2; ...; NÞ. Here the total number of particles N can<br />

be infinite. In the following, we use subscripts O and M to denote the domains of the<br />

particle and matrix phases, respectively. The overall domain D ¼ O [ M. The local<br />

magnetic field and induction satisfy<br />

B i ðrÞ ¼mðrÞH i ðrÞ, (35)<br />

where mðrÞ ¼m m 1 for r on particles and mðrÞ ¼mm 0 for r on the matrix. Thus, Eq. (33) can be<br />

rewritten as<br />

B i ðrÞ ¼m m 0 H0 i for r 2 qD. (36)<br />

Using the following identity<br />

B i ðrÞ ¼ðr i B j ðrÞÞ ;j B j;j ðrÞr i , (37)<br />

along with Eq. (32), we can write the averaged internal induction as<br />

2<br />

3<br />

hB i i D ¼ 1<br />

Z<br />

4 <br />

V D<br />

Zq P r<br />

N i B j ðrÞl j dr þ r i B j ðrÞl j dr5. (38)<br />

D<br />

p¼1 O p<br />

q P N<br />

p¼1 O p<br />

Considering the boundary conditions in Eqs. (33) and (34), the averaged internal<br />

induction is expressed as<br />

hB i i D ¼ m m 0 H0 i . (39)<br />

Alternatively, from the definition of volume average, we can write<br />

hB i i D ¼ fm m 1 hH ii O þð1 fÞm m 0 hH ii M (40)<br />

and<br />

hH i i D ¼ fhH i i O þð1 fÞhH i i M . (41)


To obtain a particle’s magnetic field, an RVE containing a chain of particles seen<br />

in Fig. 1(b) is introduced in the neighborhood of a material point X 0 marked in gray in<br />

Fig. 1(a). Without any loss of generality, a local coordinate is employed with the origin at<br />

X 0 . The x 3 -axis is along the direction of the chains. The particle centered at the origin is<br />

numbered the 0th particle, and the other particles are ordered from 1st to Nth. The<br />

number of particles N in the chain can be very large or infinite. Each particle is centered at<br />

r i . Because particles have a much smaller size than the composite, we can assume that the<br />

chain of the particles in the RVE is embedded in the infinite domain. Magnetic loading is<br />

transported from the boundary to the RVE through the matrix in the similar way to a far<br />

field, which is assumed to be the averaged magnetic field of the matrix. Thus, the magnetic<br />

field on the particle can be simulated by the solution for a chain of particle in the infinite<br />

domain under a far field strain hHi M . Similarly to Eq. (16), the magnetic field over the<br />

RVE is written in two parts:<br />

H i ðrÞ ¼hH i i M þ XN<br />

p¼0<br />

Z<br />

O p<br />

G m ij ðr; r0 ÞM j ðr 0 Þ dr 0 . (42)<br />

We use the averaged field on the 0th particle to represent a particle’s averaged field. By<br />

performing volume average of Eq. (42) over the domain of the 0th particle and by<br />

considering the same averaged magnetization in each particle, the following equation is<br />

reached:<br />

hH i i O ¼hH i i M þ mm 1 m m 0<br />

m m 0<br />

1 X N<br />

4=3pa 3<br />

p¼0<br />

ZO 0<br />

Z<br />

Using Eq. (12), we can simplify the above equation as<br />

in which<br />

O P<br />

G m ij ðr; r0 Þ dr 0 dr hH j i O . (43)<br />

hH i i M ¼ T ij hH j i O , (44)<br />

T ij ¼ 1<br />

m m 1 m m 0<br />

4pa 3 m m 0<br />

ARTICLE IN PRESS<br />

H.M. Yin et al. / J. Mech. Phys. Solids 54 (<strong>2006</strong>) 975–1003 985<br />

X N<br />

p¼0<br />

ZO 0<br />

Z<br />

O P<br />

G m ij ðr; r0 Þ dr 0 dr ¼ 1 a ð1 þ 2:4042b d I37:2126bÞd ij ,<br />

where a ¼ 3m m 0 =ðmm 1 þ 2mm 0 Þ and b ¼½ðmm 1 m m 0 Þr3 0 Š=ðmm 1 þ 2mm 0 Þ. During the derivation, the<br />

identity P 1<br />

i¼1 1=i3 ¼ 1:2021 is used. It is noted that Mura’s (1987) tensorial indicial<br />

notation is followed in the above equation; i.e., upper-case indices have the same<br />

representation as the corresponding lower-case ones but are not summed.<br />

Combining Eqs. (39)–(44) yields the averaged magnetic fields in the particles, matrix and<br />

overall composite as<br />

1<br />

hH i i O ¼ f mm 1<br />

m m d ij þð1 fÞT ij H 0 j , (46)<br />

0<br />

(45)<br />

1<br />

hH i i M ¼ f mm 1<br />

m m T ij 1 þð1 fÞd ij H 0 j (47)<br />

0


986<br />

and<br />

" # 1<br />

fÞT ij<br />

hH i i D ¼ d ij f mm 1 m m 0<br />

m m 0<br />

ARTICLE IN PRESS<br />

H.M. Yin et al. / J. Mech. Phys. Solids 54 (<strong>2006</strong>) 975–1003<br />

f mm 1<br />

m m d ij þð1<br />

0<br />

H 0 j . (48)<br />

Due to the higher magnetic permeability of particles, from the continuity condition in Eq.<br />

(34), the particle magnetic field should be lower than the applied magnetic field on the<br />

boundary, so the overall averaged magnetic field in Eq. (48) is smaller than the applied<br />

magnetic field.<br />

3.2. Averaged stress field<br />

In the composite, the internal stress should satisfy the equilibrium and boundary<br />

conditions, as follows:<br />

and<br />

s ij;j ðrÞþt i ðrÞ ¼0 for r 2 D (49)<br />

s ij ðrÞ ¼s 0 ij for r 2 qD, (50)<br />

½s þ ij ðrÞ s ij ðrÞŠl j ¼ 0 for r 2 qO p , (51)<br />

where t i is the body force. Taking into account the following formula<br />

s ij ðrÞ ¼ðs ik ðrÞr j Þ ;k s ik;k ðrÞr j (52)<br />

along with Eq. (49), we can rewrite the averaged stress as<br />

2<br />

3<br />

hs ij i D ¼ 1<br />

Z<br />

Z<br />

4 <br />

V D<br />

Zq P s<br />

N ik ðrÞl k r j dr þ P s<br />

D O N ik ðrÞl k r j dr þ t i ðrÞr j dr5.<br />

p¼1<br />

p<br />

q O p¼1<br />

p<br />

D<br />

(53)<br />

Because the gravity only produces an initial deformation, the contribution to the<br />

constitutive behavior can be disregarded. Only magnetic force is taken into account. With<br />

the consideration of boundary conditions in Eqs. (50) and (51), the averaged internal stress<br />

can be obtained as<br />

hs ij i D ¼ s 0 ij þ sm ij , (54)<br />

where s m ij ¼ 1=V D<br />

RD t iðrÞr j dr is the stress caused by magnetic force, and t i ðrÞ is the local<br />

magnetic body force at r. The magnetic force vanishes in the matrix, whereas in a particle<br />

there is an interaction magnetic force induced by any other particles. For any pair of<br />

particles, the magnetic interaction force can be simulated by the averaged field of two<br />

particles embedded in the matrix. The local magnetic field for two particle embedded in the<br />

matrix is solved by Eshelby’s equivalent inclusion method (Yin and Sun, 2005a). For the<br />

inhomogeneous magnetic field, the magnetic body force can be solved by (Pao, 1978)<br />

f i ðrÞ ¼m m 0 M kðrÞH e i;kðrÞ, (55)


where H e i denotes the external magnetic field. With a lengthy but straightforward<br />

derivation, we can obtain<br />

f i ðr I<br />

r J Þ¼ 3mm 0 ð11mm 1 þ 4mm 0 Þb2 ~r 4<br />

ð2m m 1 þ 3mm 0 Þ a ðd ij ~n k þ d ik ~n j þ d jk ~n i<br />

5 ~n i ~n j ~n k ÞhH j i M hH k i M<br />

þ 3mm 0 ð11mm 1 þ 4mm 0 Þb3 ~r 7<br />

ð2m m 1 þ 3mm 0 Þ a ðd ij ~n k þ d ik ~n j 2d jk ~n i 8 ~n i ~n j ~n k ÞhH j i M hH k i M<br />

þ Oð~r 10 Þ.<br />

Here f i ðr I ; r J Þ is the interaction force of the Ith particle due to the Jth particle, hH i i M is the<br />

matrix’s averaged magnetic field in Eq. (47), ~r ¼ a=jr I r J j,and ~n i ¼ðr I i r J i Þ=jrI r J j.<br />

When I ¼ J, f i ðr I ; r J Þ¼0. Here, we can find that the magnetic interaction force is rapidly<br />

attenuated with the order Oðjr I r J j 4 Þ. Considering that the distance between chains is<br />

much larger than b, we disregard the interaction with the particles of other chains. By<br />

ergodicity of particles in Fig. 1(b), we can write<br />

s m ij<br />

¼ V O<br />

V D<br />

X N<br />

X N<br />

I¼0 J¼0<br />

ð56Þ<br />

f i ðr I r J Þr I j , (57)<br />

where N is the largest number of the particle of the RVE in Fig. 1(b), and it can be<br />

infinitely large. Considering f i ðr I r J Þ¼ f i ðr J r I Þ, we can write<br />

s m ij<br />

¼ V O<br />

2V D<br />

X N<br />

X N<br />

I¼0 J¼0<br />

f i ðr I r J Þðr I j r J j Þ. (58)<br />

Based on geometric symmetry of a particle located in the chain containing infinite<br />

particles, we can rewrite the above equation as<br />

s m ij ¼ f 2<br />

X N<br />

J¼0<br />

ARTICLE IN PRESS<br />

H.M. Yin et al. / J. Mech. Phys. Solids 54 (<strong>2006</strong>) 975–1003 987<br />

f i ðr 0 r J Þðr 0 j r J j Þ, (59)<br />

where f ¼ NV O =V D is used. Finally, substituting Eq. (56) into Eq. (59) and truncating the<br />

higher order terms yields<br />

s m ij ¼ 3fmm 0 ð11mm 1 þ 4mm 0 Þðmm 1 m m 0 Þ2<br />

ð2m m 1 þ 3mm 0 Þðmm 1 þ hH k i M hH l i M<br />

2mm 0<br />

2<br />

Þ2<br />

1:2021r 3 0 ðd 3<br />

il ~n j ~n k þ d ik ~n j ~n l þ d kl ~n i ~n j 5 ~n i ~n j ~n k ~n l Þ<br />

6<br />

<br />

þ mm 1 m m 7<br />

4<br />

0<br />

m m 1 þ 1:0173r 6<br />

2mm 0 ðd 5, ð60Þ<br />

il ~n j ~n k þ d ik ~n j ~n l 2d kl ~n i ~n j 8 ~n i ~n j ~n k ~n l Þ<br />

0<br />

where the identities P 1<br />

i¼1 1=i3 ¼ 1:2021 and P 1<br />

i¼1 1=i6 ¼ 1:0173 are used. It is noted that the<br />

stress s m ij may be an asymmetric tensor when the magnetic field is not parallel to the chain’s<br />

direction due to a distributed angular momentum in the composite caused by the magnetic<br />

force (Brown, 1966; Pao, 1978). Consequently, the stress s m ij can be decomposed into two<br />

parts:<br />

s m ij ¼ s m ðijÞ þ sm ½ijŠ , (61)


988<br />

ARTICLE IN PRESS<br />

H.M. Yin et al. / J. Mech. Phys. Solids 54 (<strong>2006</strong>) 975–1003<br />

where the first is the symmetric part that induces a strain in the composite, and the second<br />

is the antisymmetric part which is balanced by the external mechanical loading s 0 ij . Thus,<br />

the overall averaged stress is still symmetric.<br />

Alternatively, from the definition of volume average, we can write<br />

hs ij i D ¼ fC 1 ijkl ðhe kli O e ms<br />

kl Þþð1 fÞC0 ijkl he kli M (62)<br />

and<br />

he ij i D ¼ fhe ij i O þð1 fÞhe ij i M , (63)<br />

where e ms<br />

ij is the particle’s magnetostriction, which depends on the particle’s magnetization.<br />

Especially, when the applied magnetic field is large enough, the magnetostriction is<br />

saturated on the particle and can be regarded as a material constant (Bozorth, 1951). In<br />

this paper, we assume magnetostriction to be uniform in each particle. When<br />

magnetostriction is less than the saturation level, for simplicity, it is assumed to vary<br />

linearly with the particle’s averaged magnetization (Cullity, 1972). The detailed discussions<br />

will be given later.<br />

3.3. Averaged strain field<br />

To solve the particle’s averaged strain, the RVE in Fig. 1(b) is again used. External<br />

loading is transferred to the RVE through matrix material in the similar way as a far field<br />

strain, which is assumed to be the averaged strain of the matrix. Thus, the strain field in the<br />

particle can be obtained by the solution for a chain of particles in the infinite medium<br />

under a far field strain he ij i M . If the particles have the same <strong>elastic</strong>ity as the matrix, the<br />

strain field should come from three parts: (1) the initial strain in the matrix, which can be<br />

assumed to be the averaged strain in the matrix, denoted by hei M , (2) the strain induced by<br />

the magnetic force from other particles, denoted by e mp<br />

ij<br />

particles, denoted by e ms<br />

ij , namely,<br />

ē ij ðr 0 Þ¼he ij i M þ e mp<br />

ij ðr 0 Þ<br />

X N<br />

K¼0<br />

D e ijmn ðrK<br />

, and (3) the magnetostriction of<br />

r 0 ÞC 0 mnkl ems kl , (64)<br />

where the strain e mp<br />

ij ðr 0 Þ, which is caused by all the magnetic forces between particles, can be<br />

written as<br />

e mp<br />

ij ðr 0 Þ¼ XN X N<br />

D f ijk ðrI r 0 Þf k ðr I r J Þ. (65)<br />

I¼0 J¼0<br />

Based on the same procedure as that in Eqs. (57) and (58), e mp<br />

ij ðr 0 Þ can be written as<br />

e mp<br />

ij ðr 0 Þ¼ 1 X N X N<br />

½D f ijk<br />

2<br />

I¼0 J¼0<br />

r 0 Þ D f ijk ðrJ r 0 ÞŠf k ðr I r J Þ. (66)<br />

Because P N<br />

I¼0 Df ijk ðrI r 0 Þ¼ P N<br />

J¼0 Df ijk ðrJ r 0 Þ are two absolutely convergent series for<br />

N !1, the summation in Eq. (66) is zero. Physically, for a chain containing infinite<br />

particles, the local strain field caused magnetic forces is only affected by the neighboring<br />

particles due to the absolute convergence of the series. Based on the symmetry of particle<br />

distribution, the resultant magnetic force on each particle vanishes, which further makes


the local strain field zero. Eq. (64) can then be rewritten as<br />

ē ij ðr 0 Þ¼he ij i M<br />

X N<br />

K¼0<br />

D e ijmn ðrK<br />

r 0 ÞC 0 mnkl ems kl . (67)<br />

However, the particles have different <strong>elastic</strong>ity property from the matrix. Based on<br />

Eshelby’s equivalent inclusion method (Mura, 1987), the strain at point r in the 0th particle<br />

due to the material mismatch can be obtained by introducing an eigenstrain in the particle<br />

domain, which satisfies<br />

C 1 ijkl ½ē klðrÞ e ms<br />

kl þ e 0 kl ðrÞŠ ¼ C0 ijkl ½ē klðrÞ e ms<br />

kl þ e 0 kl ðrÞ e klðrÞŠ, (68)<br />

where the perturbed strain e 0 ij ðrÞ is caused by the distributed eigenstrain e ijðrÞ, i.e.<br />

e 0 ij ðrÞ ¼<br />

XN<br />

K¼0<br />

D e ijmn ðrÞC0 mnkl e kl . (69)<br />

Taking the volume average of Eq. (68) and combining Eqs. (67)–(69), we can solve the<br />

averaged eigenstrain as<br />

he ij i O ¼<br />

e ms<br />

ij ðC 0 ijmn Þ 1 ðDC 1<br />

mnkl<br />

þðC 0 ijmn Þ 1 ðDC 1<br />

mnst<br />

D 0 mnst<br />

D 0 mnkl<br />

¯D mnkl Þ 1 he kl i M<br />

¯D mnst Þ 1 ðDC 1<br />

stpq þ C0 stpq Þ 1 C 0 pqkl ems kl ,<br />

where D 0 ijkl describes the perturbed strain caused by 0th particle itself as<br />

D 0 ijkl ¼ De ijkl ð0Þ ¼ 1<br />

30m e 0 ð1 n 0Þ ½d ijd kl ð4 5nÞðd ik d jl þ d il d jk ÞŠ (71)<br />

and ¯D ijkl renders the interaction from all other particles in the same chain as<br />

¯D ijkl ¼ 1 X N Z<br />

4=3pa 3 G<br />

p¼1<br />

ZO e ijkl ðr; r0 Þ dr 0 dr. (72)<br />

0 O P<br />

Considering the geometry of the microstructure, we can explicitly write the above equation<br />

as<br />

¯D ijkl ¼ R 1 IK d ijd kl þ R 2 IJ ðd ikd jl þ d il d jk Þ, (73)<br />

where<br />

ARTICLE IN PRESS<br />

H.M. Yin et al. / J. Mech. Phys. Solids 54 (<strong>2006</strong>) 975–1003 989<br />

R 1 IK ¼ 1:202r3 0<br />

6m e 0 ð1 n 0Þ ½1 1:035r2 0 3ð1 1:725r 2 0 Þðd I3 þ d K3 Þþð15 36:23r 2 0 Þd I3d K3 Š<br />

ð70Þ<br />

ð74Þ<br />

R 2 IJ ¼ 1:202r 3 0<br />

6m e 0 ð1 n 0Þ ½1 2n 0 þ 1:035r 2 0 þ 3ðn 0 1:725r 2 0 Þðd I3 þ d J3 ÞŠ. (75)<br />

Here, the identities P 1<br />

i¼1 1=i3 ¼ 1:2021 and P 1<br />

i¼1 1=i5 ¼ 1:0369 are used.<br />

Substituting Eq. (70) into Eq. (69), we can solve a particle’s averaged strain as<br />

he ij i O ¼ DCijmn 1 ðDC mnkl<br />

1<br />

¯D mnkl Þ 1 he kl i M<br />

D 0 mnkl<br />

ðD 0 ijmn þ ¯D ijmn ÞðDC 1<br />

mnst<br />

D 0 mnst<br />

¯D mnst Þ 1 DC 1<br />

stpq C1 pqkl ems kl .<br />

ð76Þ


990<br />

ARTICLE IN PRESS<br />

H.M. Yin et al. / J. Mech. Phys. Solids 54 (<strong>2006</strong>) 975–1003<br />

Combining Eqs. (62), (63), and (76) provides the averaged strain of the composites as<br />

he ij i D ¼ T 1 ijkl ðC0 klmn Þ 1 hs mn i D þ T 2 ijkl ems kl , (77)<br />

where<br />

and<br />

T 1 ijkl ¼fI ijkl þ fðC 0 ijmn Þ 1 ½DC 1<br />

mnkl ð1 fÞðD 0 mnkl þ ¯D mnkl ÞŠ<br />

T 2 ijkl ¼ f½fDC ijpq 1 C1 1<br />

pqmn þð1 fÞðDCijpq<br />

Considering Eq. (54), we can obtain<br />

s 0 ij ¼ C0 ijkl ðT 1 klmn Þ 1 ðhe mn i D<br />

D 0 ijpq<br />

1 g 1 (78)<br />

¯D ijpq ÞC 0 pqmn Š 1 DC 1<br />

mnst C1 stkl . (79)<br />

T 2 mnpq ems pq Þ sm ij . (80)<br />

4. Results and discussion<br />

In the last section, we solved the averaged magnetic fields and <strong>elastic</strong> fields for the<br />

particles, matrix, and overall composites. In this section, we investigate the effective<br />

magneto-<strong>elastic</strong> behavior of chain-structured composites for a given magnetic and<br />

mechanical loading.<br />

4.1. Effective magnetostriction<br />

If a magnetic field H 0 i is applied with the absence of mechanical load, there exists an<br />

effective stress s m ij in the composite as seen in Eq. (60), and an effective strain or<br />

magnetostriction can be derived from Eq. (77) as<br />

e m ij ¼ T 1 ijkl ðC0 klpq Þ 1 s m pq þ T 2 ijkl ems kl , (81)<br />

where Eq. (54) has been used. From the above equation, we can see that the effective<br />

magnetostriction e m ij is induced by the magnetic interaction force between particles as well<br />

as the particle magnetostriction. As we know, the effective magnetic stress s m ij in Eq. (60) is<br />

in general asymmetric except in some special cases where the magnetic field is parallel or<br />

perpendicular to the chain’s direction. Note that an asymmetric stress in a free body is not<br />

physical, and in fact, this situation will never happen. For instance, when a free chainstructured<br />

composite presents in a magnetic field with a small angle between the directions<br />

of chains and the magnetic field, the magnetic force will cause the composite to rotate until<br />

the chains have the same direction as the magnetic field to reach equilibrium. That is why<br />

this type of materials is also called ‘‘active’’ materials. Thus, a chain-structured composite<br />

will not have an asymmetric magnetic stress unless a balancing external mechanical<br />

loading exists. By virtue of this fact, we will only consider the effective magnetostriction for<br />

the composite with the chain’s direction parallel to the magnetic field.<br />

For pure magnetostrictive materials, magnetostriction is experimentally measured by the<br />

fractional change in length l ¼ Dl=l (Clark, 1980). The value of l measured at magnetic<br />

saturation is called the saturation magnetostriction l s . Typically, the saturated<br />

magnetostriction is a material constant. However, it may change with ambient<br />

temperatures and stress states in a nonlinear manner. It is noted that, in this work, we<br />

do not consider the effect of temperatures and stress on particle’s magnetostriction and


ARTICLE IN PRESS<br />

H.M. Yin et al. / J. Mech. Phys. Solids 54 (<strong>2006</strong>) 975–1003 991<br />

magnetization. We assume that particle’s magnetostriction changes linearly with the<br />

particle’s averaged magnetization (Cullity, 1972) before the magnetostriction is saturated<br />

at the fixed saturated magnetostriction. In the following, we set the external magnetic field<br />

in the same direction as the chains (x 3 direction). For isotropic magnetostrictive particles,<br />

the magnetostriction due to magnetic field in the x 3 direction has the form (Yin et al., 2002)<br />

lðyÞ ¼ 3 2 lðhH 3i O Þðcos 2 1<br />

y<br />

3Þ, (82)<br />

where y denotes the angle between the magnetization direction and the measurement<br />

direction and lðhH 3 i O Þ is measured in the direction of magnetic field. From the linear<br />

assumption, this can be written as<br />

lðhH 3 i O Þ¼ ls<br />

H s jhH 3i O j, (83)<br />

where H s is particle’s magnetic field when its magnetostriction is saturated. Then, the<br />

magnetostrictive eigenstrains are<br />

l s<br />

e ms<br />

33 ¼ ls<br />

H s jhH 3i O j; e ms<br />

11 ¼ ems 22 ¼ 1 2 H s jhH 3i O j; e ms<br />

ij ¼ 0 for iaj. (84)<br />

From Eq. (84), we can see that the volumetric strain of particles caused by<br />

magnetostriction is zero, which yields an incompressible deformation. Substituting the<br />

above equation into Eq. (81), we get the effective magnetostriction of the composite.<br />

Fig. 2 shows the effective magnetostriction changing with the magnetic loading and the<br />

comparison with Duenas and Carman’s (2000) experimental data. In the experiment,<br />

700.00<br />

λ s =0.00075, H s =9.5*10 4 , µ m 0 =4π*10 -7 , E 0 =3*10 9 , v 0 =0.45<br />

µ m 1 =3µ m 0 , E 1 =35*10 9 , v 1 =0.25, φ=0.4, ρ 0 =0.499<br />

600.00<br />

Effective magnetostriction (ppm)<br />

500.00<br />

400.00<br />

300.00<br />

200.00<br />

100.00<br />

0.00<br />

Experimental data<br />

Proposed model<br />

-0.4 -0.2 0.0 0.2 0.4<br />

Flux density (T)<br />

Fig. 2. Effective magnetostriction in the direction of magnetic field vs. flux density in comparison to the<br />

experimental data (Duenas and Carman, 2000).


992<br />

ARTICLE IN PRESS<br />

H.M. Yin et al. / J. Mech. Phys. Solids 54 (<strong>2006</strong>) 975–1003<br />

a magnetostrictive composite containing chain-structured Terfenol-D in an epoxy matrix is<br />

used and the Young’s moduli and Poisson’s ratios for the particles and matrix are<br />

E 1 ¼ 35 GPa, n 1 ¼ 0:25, E 0 ¼ 3 GPa, and n 0 ¼ 0:45, respectively. The volume fraction of<br />

particles is 40% and particle’s center–center distance is very close and thus we use<br />

r 0 ¼ 0:499. The relative magnetic permeability of Terfenol-D is in the range of 3–15<br />

(McKnight, 2002). Here we use m m 1 ¼ 3mm 0 . Based on Wun-Fogle and colleagues’ (1999)<br />

measurements, the saturation magnetostriction of Terfenol-D is 750 ppm when the applied<br />

magnetic field reaches 2000 Oe. Then, the particle’s magnetic field is calculated as 1200 Oe<br />

by Eq. (16), and l s ¼ 0:00075 and H s ¼ 9:5 10 4 A=m are also obtained. To solve the flux<br />

density, the effective permeability of the composite in the chains’ direction is calculated as<br />

1:689m m 0 (Yin, 2004). In Fig. 2, we find that the effective magnetostriction is still<br />

comparable to particle’s saturation magnetostriction in this microstructure. Since it is<br />

assumed that the magnetostriction changes linearly with the particle’s magnetic field up to<br />

the fixed saturated magnetostriction and that Terfenol-D has a constant magnetic<br />

permeability during the magnetic loading, it is shown that the proposed model provides a<br />

nearly linear prediction of the effective magnetostriction before it is saturated. Although<br />

this is a reasonable approximation under small magnetic fields, a rigorous consideration of<br />

the nonlinear magnetic behavior of Terfenol-D is ultimately needed for more general cases.<br />

In addition, the particles’ magnetostriction is also sensitive to stresses. In the<br />

experiments of Wun-Fogle et al. (1999), it is found that Terfenol-D has a considerably<br />

larger saturation magnetostriction under a compressive stress. For instance, the saturated<br />

magnetostriction can reach 1200 ppm under the prestress at 5.3 MPa. When a composite is<br />

subjected to a magnetic field, the particle stress comes from the constraint of matrix to the<br />

particle magnetostriction and the magnetic force between particles. The particle stress<br />

should be compressive, so particle’s saturated magnetostriction is underestimated.<br />

Consequently, it is shown in Fig. 2 that the proposed model underestimates when<br />

compared with the experimental results for the case of high magnetic flux density (larger<br />

than 0.2 T). If the magnetostriction and magnetic permeability of particles changing with<br />

stress is exactly characterized, the proposed linear model is still applicable in an<br />

incremental way, in which the loading is divided into many small steps and in each step the<br />

linear behavior is considered with updated magnetostriction and magnetic permeability.<br />

In Fig. 2, due to the very high modulus of the matrix, and the small magnetic<br />

permeability in Terfenol-D, the magnetic force does not provide a considerable<br />

contribution to the effective magnetostriction. Fig. 3 illustrates the effect of the Young’s<br />

modulus of the matrix and the magnetic permeability of the particles on the effective<br />

magnetostriction. In Fig. 3(a), the Young’s modulus of the matrix varies in the range of<br />

3 MPa to 30 GPa sampled by five values. Since magnetic force causes a considerable<br />

contractive deformation when the Young’s modulus of the matrix is small, we see that the<br />

effective magnetostriction decreases when the Young’s modulus is reduced from<br />

300–3 MPa. On the other hand, when the Young’s modulus is large, the matrix provides<br />

a strong constraint for the particles’ magnetostriction, and thus the effective magnetostriction<br />

also decreases as the Young’s modulus is increased from 300 MPa to 30 GPa. Thus,<br />

there exists an optimal choice of the matrix Young’s modulus that maximizes effective<br />

magnetostriction of the composite. However, this is contrary to the prediction of Herbst et<br />

al. (1997) in that the matrix’s stiffness has no effect on the overall magnetostriction. In<br />

addition, when the Young’s modulus of the matrix is at 3 MPa, the effective<br />

magnetostriction no longer changes linearly with the applied magnetic field. This is


ARTICLE IN PRESS<br />

H.M. Yin et al. / J. Mech. Phys. Solids 54 (<strong>2006</strong>) 975–1003 993<br />

Effective magnetostriction (ppm)<br />

600<br />

500<br />

400<br />

300<br />

200<br />

100<br />

λ s =0.00075, H s =9.5*10, µ 0 m =4π*10 -7 , v 0 =0.45<br />

4<br />

µ m 1 =3µ m 0 , E 1 =35*10 9 , v 1 =0.25, φ =0.4, ρ 0 =0.499<br />

E 0 =3MPa<br />

E 0<br />

=30MPa<br />

E 0<br />

=300MPa<br />

E 0<br />

=3GPa<br />

E 0 =30GPa<br />

(a)<br />

0<br />

0 40 80 120 160<br />

Applied magnetic field (KA/m)<br />

λ s =0.00075, H s =9.5*10, µ 0 m =4π*10 -7 , v 0 =0.45<br />

4<br />

E 0 =3*10 8 , E 1 =35*10 9 , v 1 =0.25, φ =0.4, ρ 0 =0.499<br />

Effective magnetostriction (ppm)<br />

600<br />

500<br />

400<br />

300<br />

200<br />

100<br />

µ m m<br />

1<br />

=3µ 0<br />

µ m m<br />

1<br />

=6µ 0<br />

µ m m<br />

1<br />

=9µ 0<br />

µ m m<br />

1<br />

=12µ 0<br />

µ m m<br />

1<br />

=15µ 0<br />

(b)<br />

0<br />

0 40 80 120 160<br />

Applied magnetic field (KA/m)<br />

Fig. 3. Effective magnetostriction in the direction of magnetic field vs. applied magnetic field for (a) the different<br />

matrix’s Young’s moduli and (b) the different particles’ magnetic permeabilities.<br />

because the magnetic force has a second order effect on the effective magnetostriction as<br />

seen in Eq. (60). In Fig. 3(b), the particles’ relative magnetic permeability varies in the<br />

range of 3–15 sampled by five values while the Young’s modulus of the matrix is fixed at<br />

300 MPa. It is apparent since the effective magnetostriction reduces as the particles’


994<br />

ARTICLE IN PRESS<br />

H.M. Yin et al. / J. Mech. Phys. Solids 54 (<strong>2006</strong>) 975–1003<br />

magnetic permeability increases due to the increasing magnetic force between particles.<br />

However, it is expected that the effect of particles’ magnetic permeability on the effective<br />

magnetostriction will not be so significant when the matrix is stiffer.<br />

Fig. 4 shows the effect of the volume fraction and the center–center distance of particles<br />

on the effective magnetostriction of composites. In Fig. 4(a), we observe that when the<br />

particles’ volume fraction is larger than 20%, the effective magnetostrictions for four cases<br />

are very similar before the effective magnetostriction is saturated. Especially, 30% and<br />

40% volume fraction composites produce the largest magnetostriction in the linear range,<br />

which is contrary to the general understanding that the greater magnetostrictive particles’<br />

volume fraction will lead to a greater effective magnetostriction of the composite. Duenas<br />

and Carman (2000) also observed this interesting phenomenon in their experiment, which<br />

was interpreted to be caused by thermal residual stresses during the curing process (Duenas<br />

and Carman, 2000). In the current work, it is found that the phenomenon may still exist in<br />

magnetostrictive composites even without thermal residual stresses. In Fig. 4(b), the<br />

volume fraction is set at 0.10. With the increase of the ratio of a particle’s radius to its<br />

center–center distance, the chain structure in the composite becomes more apparent. We<br />

find that the effective magnetostriction is significantly larger for r 0 ¼ 0:5 than that for<br />

r 0 ¼ 0:3. It is noted that because the topological structure of composites can greatly<br />

change the magnetic field on the particle, the effective magnetostrictions are saturated at<br />

the different applied magnetic fields for both figures.<br />

4.2. Effective <strong>elastic</strong>ity<br />

When a mechanical loading is applied with the absence of a magnetic loading, from the<br />

relation between the loading and the response as hs ij i D ¼ ¯C ijkl he kl i D , we can derive the<br />

effective <strong>elastic</strong>ity of the composite. From Eq. (77), we have<br />

¯C ijmn ¼ C 0 ijkl ðT 1 klmn Þ 1 . (85)<br />

Substituting Eq. (78) into Eq. (85) yields<br />

¯C ijkl ¼ C 0 ijkl þ f½DC ijkl 1 ð1 fÞðD 0 ijkl þ ¯D ijkl ÞŠ<br />

1 . (86)<br />

From Eqs. (86) and (73), it is apparent that effective stiffness depends on the material<br />

properties of the particles and matrix, as well as the volume fraction of the particles and<br />

the ratio r 0 . The term ¯D includes the interactions from all other particles in the same chain.<br />

If we drop this term, Eq. (86) is reduced to<br />

¯C ijkl ¼ C 0 ijkl þ f½DC ijkl 1 ð1 fÞD 0 ijkl Š 1 , (87)<br />

which is the same as the result of Mori–Tanaka’s model (Nemat-Nasser and Hori, 1999).<br />

We can see that Mori–Tanaka’s model does not consider the effect of microstructure and<br />

direct interaction between particles.<br />

Although chains have the same direction, they can be randomly distributed in the<br />

composite. The microstructure has transversely isotropic symmetry along the chain’s<br />

direction. We can show that the effective <strong>elastic</strong>ity tensor ¯C in Eq. (86) is also transversely<br />

isotropic. The proof, along with the explicit form of the components of ¯C, is given in the<br />

Appendix.


ARTICLE IN PRESS<br />

H.M. Yin et al. / J. Mech. Phys. Solids 54 (<strong>2006</strong>) 975–1003 995<br />

4<br />

λ s =0.00075, H s =9.5*10, µ m 0 =4π*10 -7 , v 0 =0.45<br />

E 0 =3*10 8 , µ m 1 =3µ m 0 , E 1 =35*10 9 , v 1 =0.25, ρ 0 =0.499<br />

Effective magnetostriction (ppm)<br />

700<br />

600<br />

500<br />

400<br />

300<br />

200<br />

100<br />

φ=0.1<br />

φ=0.2<br />

φ=0.3<br />

φ=0.4<br />

φ=0.5<br />

(a)<br />

0<br />

0 40<br />

80<br />

120 160<br />

Applied magnetic field (KA/m)<br />

λ s =0.00075, H s =9.5*10, µ m 0 =4π*10 -7 , v 0 =0.45<br />

E 0 =3*10 8 , µ m 1 =3µ m 0 , E 1 =35*10 9 , v 1 =0.25, ρ 0 =0.1<br />

4<br />

Effective magnetostriction (ppm)<br />

ρ 0<br />

=0.3<br />

300 ρ 0 =0.4<br />

ρ 0<br />

=0.5<br />

200<br />

100<br />

(b)<br />

0<br />

0<br />

40<br />

80<br />

120 160<br />

Applied magnetic field (KA/m)<br />

Fig. 4. Effective magnetostriction in the direction of magnetic field vs. applied magnetic field for (a) the different<br />

particle volume fractions and (b) the ratios of a particle’s radius to its center–center distance.<br />

To assess the validity of the proposed model, Fig. 5 illustrates the comparisons to<br />

Duenas and Carman’s (2000) experimental data as well as Mori–Tanaka’s model, Voigt<br />

and Reuss’ (Mura, 1987) upper and lower bounds. Mechanical properties of materials are<br />

the same as those in Fig. 2. InFig. 5, we see that the upper bound provides a higher


996<br />

ARTICLE IN PRESS<br />

H.M. Yin et al. / J. Mech. Phys. Solids 54 (<strong>2006</strong>) 975–1003<br />

E 0 =3GPa, v 0 =0.45, E 1 =35GPa, v 1 =0.25, ρ 0 =0.499<br />

Effective <strong>elastic</strong> moduli (GPa)<br />

20<br />

18<br />

16<br />

14<br />

12<br />

10<br />

8<br />

6<br />

The upper bound<br />

The proposed model<br />

Experimental data<br />

Mori-Tanaka's model<br />

The lower bound<br />

4<br />

0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50<br />

Volume fraction φ<br />

Fig. 5. Effective <strong>elastic</strong> modulus in the direction of chains vs. volume fraction in comparison to the experimental<br />

data (Duenas and Carman, 2000), Mori–Tanaka’s model, Voigt and Reuss’ bounds.<br />

E 0 =3GPa, v 0 =0.45, E 1 =35GPa, v 1 =0.25, ρ 0 =0.499<br />

Effective <strong>elastic</strong> moduli (GPa)<br />

20<br />

16<br />

12<br />

8<br />

4<br />

C 1111<br />

C 1122<br />

C 1133<br />

C 3333<br />

C 2323<br />

0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50<br />

Volume fraction φ<br />

Fig. 6. Five independent <strong>elastic</strong> constants vs. volume fraction.<br />

estimate, the lower bound has an underestimation, and the proposed model yields the best<br />

prediction compared to the experimental data. Because Mori–Tanaka’s model does not<br />

consider particle interaction, it only provides an isotropic <strong>elastic</strong>ity, which significantly<br />

underestimates effective <strong>elastic</strong> properties even when the particle’s volume fraction is quite<br />

small because particle orientations are so close in a chain.


ARTICLE IN PRESS<br />

H.M. Yin et al. / J. Mech. Phys. Solids 54 (<strong>2006</strong>) 975–1003 997<br />

The overall material properties are shown in Fig. 6 where five independent material<br />

constants for the transversely isotropic effective <strong>elastic</strong>ity of the chain-structured<br />

composite are displayed. A comparison between C 3333 and C 1111 confirms that the <strong>elastic</strong><br />

modulus in the chain’s direction is larger than the one perpendicular to the direction. Thus,<br />

this microstructure can greatly improve the effectiveness of materials in applications where<br />

the primary loading direction to the composites is known. Thus, in the following we will<br />

focus on the <strong>elastic</strong> properties of composites in the direction of the chains. It is noted that<br />

when the particle’s volume fraction approaches 50%, because the distance between chains<br />

is also small, the chain structure will be fuzzy and the advantage of this microstructure<br />

cannot be fully achieved.<br />

Although a chain-structured composite has a transversely isotropic symmetry with five<br />

independent <strong>elastic</strong> constants, one can still experimentally define the Young’s modulus<br />

and the shear modulus in the corresponding directions by the properties measured from a<br />

x 3 - directional uniaxial test and a simple shear test in the x 1 2x 3 or x 2 2x 3 plane,<br />

respectively, as<br />

Ē ¼ C 3333<br />

2C 2 1133<br />

C 1111 þ C 1122<br />

; ¯m ¼ C 2323 . (88)<br />

Fig. 7 shows how the effective Young’s modulus and shear modulus vary with volume<br />

fraction. With an increase of Young’s modulus in the particles, the effective <strong>elastic</strong> moduli<br />

also increase. Special cases are illustrated for rigid or void particles. In fact, when<br />

E 1 =E 0 o0:1, the effective <strong>elastic</strong> moduli do not change much so that particle’s can be<br />

approximated as void. Similarly when E 1 =E 0 4100, the effective <strong>elastic</strong> moduli are so<br />

similar that particle’s can be assumed rigid.<br />

To study the effect of the ratio of particle’s radius to the particle center–center distance<br />

r 0 , we fix the particle’s volume fraction as f ¼ 0:05, and vary r 0 from 0.2 to 0.5. Fig. 8<br />

shows that the effective Young’s moduli rapidly increase along with r 0 but the effective<br />

shear moduli slowly decrease. Thus, although composites become harder in the direction of<br />

the chains, as particles move closer, the capability to resist shear is reduced. In Fig. 8(a),we<br />

can see that when E 1 =E 0 gets larger, the increment in the effective Young’s moduli with the<br />

increase of r 0 gets larger. For E 1 =E 0 ¼ 100, the effective Young’s moduli increase as high<br />

as 1 3 for a r 0 that increases from 0.2 to 0.5. In contrast, for E 1 =E 0 p2, the effective Young’s<br />

moduli only have a very marginal increase associated with r 0 .InFig. 8(b), the change of<br />

the effective shear modulus is not as significant as that of the effective Young’s modulus.<br />

It is noted that when the matrix is a very compliant material like rubber and when the<br />

particle’s relative magnetic permeability is high, the magnetic interaction forces between<br />

particles change sensitively with respect to the mechanical loading, and it provides a<br />

contribution to the effective <strong>elastic</strong>ity. Here we do not consider this extreme case for<br />

magnetostrictive composites, because generally magnetostrictive materials have a small<br />

relative magnetic permeability. Furthermore, to effectively use the particles’ magnetostriction,<br />

as seen in Fig. 3(a), we should not use matrix material with high compliant. In this<br />

case, the magnetic interaction forces will not change severely to a mechanical loading that<br />

lead to infinitesimal deformation, and the effective <strong>elastic</strong>ity of magnetostrictive<br />

composites will stay constant. On the contrary, for ferromagnetic composites,<br />

ferromagnetic particles typically have relative magnetic permeability as high as 10 4 and<br />

the matrix is made of rubber-like materials. The magneto-mechanical behavior is fully


998<br />

ARTICLE IN PRESS<br />

H.M. Yin et al. / J. Mech. Phys. Solids 54 (<strong>2006</strong>) 975–1003<br />

Effective Young's modulus E/E 0<br />

5.0<br />

4.5<br />

4.0<br />

3.5<br />

3.0<br />

2.5<br />

2.0<br />

1.5<br />

1.0<br />

rigid particles<br />

E 1<br />

/E 0<br />

=100<br />

E 1<br />

/E 0<br />

=10<br />

E 1<br />

/E 0<br />

=1.1<br />

E 1<br />

/E 0<br />

=0.1<br />

voids<br />

v 0 =0.45, v 1 =0.3, ρ 0 =0.45<br />

(a)<br />

0.5<br />

0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40<br />

Volume fraction φ<br />

Effective shear modulus µ/E 0<br />

0.8<br />

0.6<br />

0.4<br />

0.2<br />

rigid particles<br />

E 1<br />

/E 0<br />

=100<br />

E 1 /E 0 =10<br />

E 1 /E 0 =1.1<br />

E 1<br />

/E 0<br />

=0.1<br />

voids<br />

v 0 =0.45, v 1 =0.3, ρ 0 =0.45<br />

(b)<br />

0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40<br />

Volume fraction φ<br />

Fig. 7. Effective <strong>elastic</strong> moduli vs. volume fraction. (a) Young’s modulus; (b) the shear modulus.<br />

coupled and magnetic-dependent <strong>elastic</strong>ity behavior is commonly observed (Yin and Sun,<br />

2005b).<br />

5. Conclusions<br />

A three-dimensional micromechanics-based <strong>elastic</strong> model is proposed to investigate magneto<strong>elastic</strong><br />

behavior of two-phase composites containing chain-structured magnetostrictive particles


ARTICLE IN PRESS<br />

H.M. Yin et al. / J. Mech. Phys. Solids 54 (<strong>2006</strong>) 975–1003 999<br />

Effective Young's modulus E/E 0<br />

1.4<br />

1.3<br />

1.2<br />

1.1<br />

1.0<br />

v 0 =0.45, v 1 =0.3, φ=0.05<br />

E 1 /E 0 =100<br />

E 1<br />

/E 0<br />

=10<br />

E 1<br />

/E 0<br />

=5<br />

E 1<br />

/E 0<br />

=2<br />

E 1<br />

/E 0<br />

=0.5<br />

E 1 /E 0 =0.1<br />

(a)<br />

0.9<br />

0.20 0.25 0.30 0.35 0.40 0.45 0.50<br />

Ratio of radius to center-center distance ρ 0<br />

Effective shear modulus µ/E 0<br />

0.42<br />

0.40<br />

0.38<br />

0.36<br />

0.34<br />

0.32<br />

v 0 =0.45, v 1 =0.3, φ=0.05<br />

E 1 /E 0 =100 E 1<br />

/E 0<br />

=10<br />

E 1 /E 0 =5<br />

E 1 /E 0 =2<br />

E 1 /E 0 =0.5 E 1 /E 0 =0.1<br />

(b)<br />

0.20 0.25 0.30 0.35 0.40 0.45 0.50<br />

Ratio of radius to center-center distance ρ 0<br />

Fig. 8. Effective <strong>elastic</strong> moduli vs. ratio of a particle’s radius to its center–center distance. (a) Young’s modulus;<br />

(b) the shear modulus.<br />

under both magnetic and mechanical loading. To derive the local magnetic and <strong>elastic</strong><br />

fields, the Green’s function technique is used. Three modified Green’s functions are derived<br />

and explicitly integrated for the infinite domain containing a spherical inclusion with a<br />

prescribed magnetization, body force, and eigenstrain. Introducing the representative<br />

volume element containing a chain of infinite particles, we derive the averaged magnetic<br />

field, stress field and strain field in both phases of a chain-structured magnetostrictive<br />

particle filled composite. From the relation between the overall averaged strain and<br />

the magnetic/mechanical loading, we derive effective magnetostriction and <strong>elastic</strong>ity


1000<br />

ARTICLE IN PRESS<br />

H.M. Yin et al. / J. Mech. Phys. Solids 54 (<strong>2006</strong>) 975–1003<br />

properties. It is found that there exists an optimal choice of the matrix Young’s modulus<br />

and the particle volume fraction that maximizes effective magnetostriction. In this<br />

composite model, a transversely isotropic effective <strong>elastic</strong>ity is obtained at the infinitesimal<br />

deformation. Disregarding the interaction term, this model provides the same effective<br />

<strong>elastic</strong>ity as Mori–Tanaka’s model. The numerical simulations and the comparisons to<br />

experiments and other models show the efficacy of the proposed model for prediction of<br />

effective magneto-<strong>elastic</strong> properties of chain-structured composites. Our results also<br />

suggest that the composite with a smaller particle center–center distance offers a<br />

considerable advantage in amplifying the effective magnetostriction and strengthening the<br />

effective Young’s modulus in the chain’s direction.<br />

Acknowledgements<br />

This work is sponsored by the National Science Foundation under Grant number CMS-<br />

0084629 and DMR-0113172. The support is gratefully acknowledged.<br />

Appendix. Details of effective <strong>elastic</strong>ity<br />

Substituting Eq. (73) along with Eqs. (74) and (75) into Eq. (86), we can explicitly write<br />

effective <strong>elastic</strong>ity ¯C as<br />

where<br />

¯C ijkl ¼ðl 0 þ L IK Þd ij d kl þðm e 0 þ M IJÞðd ik d jl þ d il d jk Þ,<br />

(A.1)<br />

and<br />

K ¼ f D<br />

0<br />

d 2 2 d 1 ðd 3 þ 2d 4 þ 4d 5 Þ d 2 1<br />

2 d 1 ðd 3 þ 2d 4 þ 4d 5 Þ 2d 4 ðd 1 þ d 2 Þ<br />

d 2 2 d 1 ðd 3 þ 2d 4 þ 4d 5 Þ d 2 2 d 1 ðd 3 þ 2d 4 þ 4d 5 Þ 2d 4 ðd 1 þ d 2 Þ<br />

B<br />

C<br />

@<br />

A<br />

d<br />

2d 4 ðd 1 þ d 2 Þ 2d 4 ðd 1 þ d 2 Þ 2d 2 2 d 1 ðd 3 þd 4 Þ d 4 ðd 2 þd 3 Þ<br />

4 d 4þ2d 5<br />

ðA:2Þ<br />

0<br />

1<br />

1 1 1<br />

d 4 d 4 d 4 þ d 5<br />

M ¼ f 1 1 1<br />

4 d 4 d 4 d 4 þ d , (A.3)<br />

5<br />

B<br />

C<br />

@ 1 1 1 A<br />

d 4 þ d 5 d 4 þ d 5 d 4 þ 2d 5<br />

in which<br />

2<br />

d 1 þ 2d 4 d 1 d 1 þ d 2<br />

3<br />

6<br />

D ¼ det4<br />

d 1 d 1 þ 2d 4 d 1 þ d 2<br />

7<br />

5,<br />

d 1 þ d 2 d 1 þ d 2 d 1 þ 2d 2 þ 2d 4 þ 4d 5


ARTICLE IN PRESS<br />

H.M. Yin et al. / J. Mech. Phys. Solids 54 (<strong>2006</strong>) 975–1003 1001<br />

d 1 ¼<br />

l 1 l 0<br />

2ðm e 1<br />

m e 0 Þ½2ðme 1<br />

m e 0 Þþ3ðl 1 l 0 ÞŠ<br />

1 f<br />

30m e 0 ð1 n 0Þ<br />

ð1 1:035r 2 0Þð1 fÞw,<br />

d 2 ¼ 3ð1 1:725r 2 0Þð1 fÞw,<br />

d 3 ¼ ð15 36:225r 2 0Þð1 fÞw,<br />

1<br />

d 4 ¼<br />

4ðm e 1<br />

m e 0 Þ þ ð1 fÞð4 5n 0Þ<br />

30m e 0 ð1 n 0Þ<br />

d 5 ¼ 3ðn 0 1:725r 2 0Þð1 fÞw,<br />

þð1 2n 0 þ 1:035r 2 0Þð1 fÞw,<br />

w ¼ 1:202r3 0<br />

6m e 0 ð1 n 0Þ .<br />

A tensor with the form of X ijkl ¼ c 1 IK d ijd kl þ c 2 IJ ðd ikd jl þ d il d jk Þ has the transversely<br />

isotropic symmetry along x 3 axis if and only if it should satisfy two conditions (Cowin and<br />

Mehrabadi, 1995): major and minor symmetries,<br />

g 1 IJ ¼ g1 JI ; g2 IJ ¼ g2 JI (A.4)<br />

and rotational symmetry,<br />

g 1 11 þ 2g2 11 ¼ g1 22 þ 2g2 22 ; g1 13 ¼ g1 23 ; g2 13 ¼ g2 23 ; g2 12 ¼ 1 2 ðg1 11 þ 2g2 11 g 1 12Þ. (A.5)<br />

Substituting Eqs. (A.1)–(A.3) into Eqs. (A.4) and (A.5), we can see that the effective<br />

<strong>elastic</strong>ity satisfies the requirements and is a transversely isotropic tensor. Here, the term w<br />

signifies the interaction between particles. With w ¼ 0, Eq. (A.1) is reduced to<br />

Mori–Tanaka’s model.<br />

References<br />

Anjanappa, M., Wu, Y.F., 1997. <strong>Magneto</strong>strictive particulate actuators: configuration, modeling and<br />

characterization. Smart Mater. Struct. 6, 393–402.<br />

Armstrong, W.D., 2000. The non-linear deformation of magnetically dilute magnetostrictive particulate<br />

composites. Mater. Sci. Eng. A A285, 13–17.<br />

Bednarek, S., 1999. The giant magnetostriction in ferromagnetic composites within an elastomer matrix. Appl.<br />

Phys. A 68, 63–67.<br />

Borcea, L., Bruno, O., 2001. On the magneto-<strong>elastic</strong> properties of elastomer-ferromagnet composites. J. Mech.<br />

Phys. Solids 49, 2877–2919.<br />

Bozorth, R.M., 1951. Ferromagnetism. Von Nostrand, New York.<br />

Brown Jr., W.F., 1966. <strong>Magneto</strong><strong>elastic</strong> Interactions. Springer, New York.<br />

Chen, T., 1993. Greens functions and the non-uniform transformation problem in a piezoelectric medium. Mech.<br />

Res. Commun. 20, 271–278.<br />

Chen, Y., Sprecher, A.F., Conrad, H., 1991. Electrostatic particle–particle interactions in electrorheological fluids.<br />

J. Appl. Phys. 70, 6796–6803.<br />

Chen, Y., Snyder, J.E., Schwichtenberg, C.R., Dennis, K.W., Falzgraf, D.K., McCallum, R.W., Jiles, D.C., 1999.<br />

Effect of the <strong>elastic</strong> modulus of the matrix on magnetostrictive strain in composites. Appl. Phys. Lett. 74,<br />

1159–1161.<br />

Clark, A.E., 1980. <strong>Magneto</strong>strictive rare earth-Fe 2 compounds. In: Wohlfarth, E.P. (Ed.), Ferromagnetic<br />

Materials, vol. 1. North-Holland, Amsterdam, pp. 531–589.<br />

Coulson, C.A., 1961. Electricity. Wiley-Interscience, Edinburgh.<br />

Cowin, S.C., Mehrabadi, M.M., 1995. Anisotropic symmetries of linear <strong>elastic</strong>ity. Appl. Mech. Rev. 48, 247–285.


1002<br />

ARTICLE IN PRESS<br />

H.M. Yin et al. / J. Mech. Phys. Solids 54 (<strong>2006</strong>) 975–1003<br />

Cullity, B.D., 1972. Introduction to Magnetic Materials. Addison-Wesley, Reading, MA.<br />

Davis, L.C., 1992. Polarization forces and conductivity effects in electrorheological fluids. J. Appl. Phys. 72,<br />

1334–1340.<br />

Duenas, T.A., Carman, G.P., 2000. Large magnetostrictive response of Terfenol-D resin composites. J. Appl.<br />

Phys. 87, 4696–4701.<br />

Dunn, M.L., 1994. Electro<strong>elastic</strong> Green’s functions for transversely isotropic piezoelectric media and their<br />

application to the solution of inclusion and inhomogeneity problems. Int. J. Eng. Sci. 32, 119–131.<br />

Eshelby, J.D., 1957. The determination of the <strong>elastic</strong> field of an ellipsoidal inclusion, and related problems. Proc.<br />

R. Soc. A 241, 376–396.<br />

Eshelby, J.D., 1959. The <strong>elastic</strong> field outside an ellipsoidal inclusion. Proc. R. Soc. A 252, 561–569.<br />

Feng, X., Fang, D., Soh, A., Hwang, K.-C., 2003. Predicting effective magnetostriction and moduli of<br />

magnetostrictive composites by using the double-inclusion method. Mech. Mater. 35, 623–631.<br />

Friedmann, P.P., Carman, G.P., Millott, T.A., 2001. <strong>Magneto</strong>strictively actuated control flaps for vibration<br />

reduction in helicopter rotors—design considerations for implementation. Math. Comp. Model. 33,<br />

1203–1217.<br />

Gao, C.F., Fan, W.X., 1998. Green’s functions for generalized 2D problems in piezoelectric media with an elliptic<br />

hole. Mech. Res. Commun. 25, 685–693.<br />

Gast, A.P., Zukoski, C.F., 1989. Electrorheological fluids as colloidal suspensions. Adv. Colloid Interface Sci. 30,<br />

153–202.<br />

Ginder, J.M., Nichols, M.E., Elie, L.D., Tardiff, J.L., 1999. <strong>Magneto</strong>rheological elastomers: properties and<br />

applications. Proc. Ser. SPIE Smart Struct. Mater. 3675, 131–138.<br />

Ginder, J.M., Nichols, M.E., Elie, L.D., Clark, S.M., 2000. Controllable-stiffness components based on<br />

magnetorheological elastomers. Proc. Ser. SPIE Smart Struct. Mater. 3985, 418–425.<br />

Guo, Z.J., Busbridge, S.C., Piercy, A.R., Zhang, Z.D., Zhao, X.G., Wang, B.W., 2001. Effective magnetostriction<br />

and magnetomechanical coupling of Terfenol-D composites. Appl. Phys. Lett. 78, 3490–3492.<br />

Halsey, T.C., 1992. Electrorheological fluid. Science 258, 761–766.<br />

Hatta, H., Taya, M., 1986. Equivalent inclusion method for steady state heat conduction in composites. Int.<br />

J. Eng. Sci. 24, 1159–1172.<br />

Herbst, J.F., Capehart, T.W., Pinkerton, F.E., 1997. Estimating the effective magnetostriction of a composite: a<br />

simple model. Appl. Phys. Lett. 70, 3041–3043.<br />

Huang, J.H., Chiu, Y.H., Liu, H.K., 1998. <strong>Magneto</strong>-electro-<strong>elastic</strong> Eshelby tensors for a piezoelectric–piezomagnetic<br />

composite reinforced by ellipsoidal inclusions. J. Appl. Phys. 83, 5364–5370.<br />

Indenbom, V.L., Orlov, S.S., 1968. Construction of Green’s function in terms of Green’s function of lower<br />

dimension. J. Appl. Math. Mech. 32, 414–420.<br />

Jackson, J.D., 1999. Classical Electrodynamics. Wiley, New York.<br />

Jin, S., Tiefel, T.H., Wolfe, R., Sherwood, R.C., Mottine, J.J., 1992. Optically transparent, electrically conductive<br />

composite medium. Science 255, 446–448.<br />

Ju, J.W., Sun, L.Z., 1999. A novel formulation for exterior-point Eshelby’s tensor of an ellipsoidal inclusion.<br />

J. Appl. Mech. 66, 570–574.<br />

Klingenberg, D.J., Zukoski, C.F., 1990. Studies on the steady-shear behavior of electrorheological suspensions.<br />

Langmuir 6, 15–24.<br />

Kröner, E., 1972. Statistical Continuum Mechanics. CISM Courses and Lectures No. 92. Springer, Berlin.<br />

Kröner, E., 1990. Modified Green functions in the theory of heterogeneous and/or anisotropic linearly <strong>elastic</strong><br />

media. In: Weng, G.L., Taya, M., Abé, H. (Eds.), Micromechanics and Inhomogeneity. Springer, New York,<br />

pp. 197–211.<br />

Li, J.Y., 2002. <strong>Magneto</strong>electric Green’s functions and their application to the inclusion and inhomogeneity<br />

problems. Int. J. Solids Struct. 39, 4201–4213.<br />

Mazilu, P., 1972. On the theory of linear <strong>elastic</strong>ity in statistically homogeneous media. Rev. Roum. Math. Pures<br />

Appl. 17, 261–273.<br />

McKnight, G.P., 2002. [112] Oriented Terfenol-D composites. Ph.D. Dissertation, <strong>UCLA</strong>.<br />

Michelitsch, T., 1997. Calculation of the electro<strong>elastic</strong> Green’s function of the hexagonal infinite medium. Z. Phys.<br />

B 104, 497–503.<br />

Morse, P.M., Feshbach, H., 1953. Methods of Theoretical Physics. McGraw-Hill, New York.<br />

Mura, T., 1963. Continuous distribution of moving dislocations. Philos. Mag. 8, 843–857.<br />

Mura, T., 1987. Micromechanics of Defects in Solids, second ed. Kluwer Academic Publishers, Dordrecht.<br />

Mura, T., Kinoshita, N., 1971. Greens functions for anisotropic <strong>elastic</strong>ity. Phys. Stat. Sol. B 47, 607–618.


ARTICLE IN PRESS<br />

H.M. Yin et al. / J. Mech. Phys. Solids 54 (<strong>2006</strong>) 975–1003 1003<br />

Nan, C.W., 1994. <strong>Magneto</strong>electric effect in composites of piezoelectric and piezomagnetic phases. Phys. Rev. B<br />

50, 6082–6088.<br />

Nan, C.W., 1998. Effective magnetostriction of magnetostrictive composites. Appl. Phys. Lett. 72, 2897–2899.<br />

Nan, C.W., Weng, G.J., 1999. Influence of microstructural features on the effective magnetostriction of composite<br />

materials. Phys. Rev. B 60, 6723–6730.<br />

Nemat-Nasser, S., Hori, M., 1999. Micromechanics: Overall Properties of Heterogeneous Materials, second ed.<br />

North-Holland, Amsterdam.<br />

Pan, Y.C., Chou, T.W., 1976. Point force solution for an infinite transversely isotropic solid. J. Appl. Mech. 43,<br />

608–612.<br />

Pao, Y.H., 1978. Electromagnetic forces in deformable continua. In: Nemat-Nasser, S. (Ed.), Mechanicas Today,<br />

vol. 4, pp. 209–306.<br />

Rayleigh, L., 1892. On the influence of obstacles arranged in rectangular order upon the properties of a medium.<br />

Philos. Mag. 34, 481–502.<br />

Reitz, J.R., Milford, F.J., Christy, R.W., 1979. Foundations of Electromagnetic Theory, third ed. Addison-<br />

Wesley, Reading, MA.<br />

Sandlund, L., Fahlander, M., Cedell, T., Clark, A.E., Restorff, J.B., Wun-Fogle, M., 1994. <strong>Magneto</strong>striction,<br />

<strong>elastic</strong> moduli, and coupling factors of composite Terfenol-D. J. Appl. Phys. 75, 5656–5658.<br />

von Hippel, A.R., 1954. Dielectrics and Waves. Wiley, New York.<br />

Wang, B., 1992. Three-dimensional analysis of an ellipsoidal inclusion in a piezoelectric material. Int. J. Solids<br />

Struct. 29, 293–308.<br />

Willis, J.R., 1965. The <strong>elastic</strong> interaction energy of dislocation loops in anisotropic media. Quart. J. Mech. Appl.<br />

Math. 18, 419–433.<br />

Wun-Fogle, M., Restorff, J.B., Leung, K., Cullen, J.R., 1999. <strong>Magneto</strong>striction of Terfenol-D heat treated under<br />

compressive stress. IEEE Trans. Magnet. 35, 3817–3819.<br />

Yin, H.M., 2004. Micromechanics-based magneto-<strong>elastic</strong> constitutive modeling of particulate composites. Ph.D.<br />

Dissertation, The University of Iowa.<br />

Yin, H.M., Sun, L.Z., 2005a. Overall magnetic properties of composites with randomly dispersed magneticparticle<br />

interactions. Phys. Rev. B 72, 054409.<br />

Yin, H.M., Sun, L.Z., 2005b. <strong>Magneto</strong>-<strong>elastic</strong>ity of chain-structured ferromagnetic composites. Appl. Phys. Lett.<br />

86, 261901.<br />

Yin, H.M., Sun, L.Z., Chen, J.S., 2002. Micromechanics-based hyper<strong>elastic</strong> constitutive modeling of<br />

magnetostrictive particle-filled elastomers. Mech. Mater. 34, 505–516.<br />

Yin, H.M., Paulino, G.H., Buttlar, W.G., Sun, L.Z., 2005. Effective thermal conductivity of functionally graded<br />

composites. J. Appl. Phys. 98, 063704.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!