Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                

Recovered PDF 17204 PDF

Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

Acta Materialia 53 (2005) 28452855

www.actamat-journals.com

Phase-eld microelasticity theory and micromagnetic simulations


of domain structures in giant magnetostrictive materials
J.X. Zhang *, L.Q. Chen
Department of Materials Science and Engineering, The Pennsylvania State University, 305 Steidle Building, University Park,
State College, PA 16802, United States
Received 17 December 2004; received in revised form 27 February 2005; accepted 2 March 2005
Available online 7 April 2005

Abstract
A computational model is proposed to predict the stability of magnetic domain structures and their temporal evolution in giant
magnetostrictive materials by combining a micromagnetic model with the phase-eld microelasticity theory of Khachaturyan. The
model includes all the important energetic contributions, including the magnetocrystalline anisotropy energy, exchange energy, magnetostatic energy, external eld energy, and elastic energy. While the elastic energy of an arbitrary magnetic domain structure is
obtained analytically in Fourier space, the LandauLiftshitzGilbert equation is solved using the ecient GaussSeidel projection
method. Both Fe81.3Ga18.7 and Terfenol-D are considered as examples. The eects of elastic energy and magnetostatic energy on
domain structures are studied. The magnetostriction and associated domain structure evolution under an applied eld are modeled
under dierent pre-stress conditions. It is shown that a compressive pre-stress can eciently increase the overall magnetostrictive
eect. The results are compared with existing experiment measurements and observations.
 2005 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
Keywords: Phase-eld models; Simulation; Domain structure; Magnetostriction

1. Introduction
Magnetostriction is a phenomenon in which a material changes its physical dimensions when there is a
change in its magnetization. The change in magnetization can be induced by the application of a magnetic
eld or a stress eld. Magnetostriction is observed, to
diering degrees, in all ferromagnetic materials. While
typical magnetostrictive strains of magnetic materials
are of the order of 105106, some compounds containing rare earth elements have magnetostriction as
high as 103. Among these, Terfenol-D (TbxDy1  xFe2,
x = 0.3) has been extensively investigated for use in sensor and actuator devices because of its high magnetostriction to anisotropy ratio [1,2]. It has been recently
*

Corresponding author. Tel.: +1 814 865 0389; fax: +1 814 865 2917.
E-mail address: jzz108@psu.edu (J.X. Zhang).

found that large magnetostriction (over 200 ppm) at


room temperature also occurs in FeGa alloys with disordered body-centered cubic a-Fe structure [35].
Giant magnetostrictive materials often display complex domain structures that arise from the competition
of the magnetocrystalline anisotropy energy, exchange
energy, magnetostatic energy, external eld energy,
and elastic energy. The domain evolution under applied
elds is directly responsible for the overall strain response of the materials. Therefore, the domain structures and their evolution have been extensively studied
experimentally [615]. To explain the observed magnetic
behavior, there have also been a number of theoretical
developments for magnetostrictive materials, including
the works by Brown [16], James and Kinderlehrer
[17,18], and DeSimone and James [19]. However, due
to the rather complicated elastic solutions associated
with an arbitrary domain structure, there has been

1359-6454/$30.00  2005 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.actamat.2005.03.002

2846

J.X. Zhang, L.Q. Chen / Acta Materialia 53 (2005) 28452855

essentially no simulation of macroscopically inhomogeneous domain patterns and their evolution under external elds in giant magnetostrictive materials.
One type of approach assumed that the magnetizations of domains are constrained by the anisotropy
energy to lie along the easy magnetic axes. The ferromagnetic constitutive behavior was then described by
the distribution of a xed quantity of available magnetic
domains within materials among all the possible directions. The model assumed that each domain is magnetized independently without being aected by the
magnetization states of neighboring domains. This type
of approach has been used in many works to study the
magnetostriction under an external magnetic eld [20
24]. But this model is not suitable for describing a material with low magnetocrystalline anisotropy or the
system states close to saturation when the magnetic eld
is applied along directions far away from those of easy
axes. Zhu et al. [25] studied the inuence of external
stress on the cohesive eld in magnetic thin lms. In
their micromagnetic model, without solving the total
compatible strain, they use the stress-free magnetostrictive strain caused by the magnetization instead; thus the
intrinsic stress resulting from the elastic incompatibility
of magnetostrictive strains is not taken into account.
The intrinsic stress is critically important for materials
with large magnetostriction [26]. Fabian and Heider
[27] have calculated the magnetostrictive self-energies
of a magnetic particle imbedded in a non-magnetic matrix using the nite element method and the continuum
theory of defects. Recently, Shu et al. [26,28] developed
a modied boundary integral formalism to calculate the
intrinsic stress induced by incompatible magnetostrictive
strains, but their method is conned to some simple twodimensional (2D) cases due to the complexity of solving
the elastic equilibrium equations.
The main purpose of this paper is to describe a
computational approach to model the stability and evolution of magnetic domain structures in giant magnetostrictive materials. It combines the micromagnetic model
for magnetic domain evolution [29] and the phase-eld
microelasticity theory [30,31] for the elastic solutions
in a magnetostrictive material with arbitrary distributions of magnetic domains. The phase-eld microelasticity theory has been extensively used in computer
simulations of structural phase transformations and
microstructure evolution in both bulk systems, see, for
example, Refs. [3234], and for more extensive references [35], and thin lms [3638]. The model is able to
predict the detailed domain structure and their evolution
under an applied eld without a priori assumptions on
domain morphologies. The main limitation of the present model as well as essentially all prior phase-eld models for solid state phase transformations is the
assumption of periodic boundary conditions, and hence
the results are only valid when the simulation system size

is much smaller than the actual sample size. It will be


shown that the domain shapes and domain wall conguration in magnetostrictive materials predicted from the
model automatically satisfy the condition of magnetic
and elastic compatibilities. The domain structure and
magnetostriction evolution under a magnetic eld will
also be studied for samples with dierent magnitudes
of applied pre-stress. The results predicted from our simulations will be compared with existing experimental
observation and theoretical predictions.
2. Micromagnetic model and phase-eld elasticity theory
In a micromagnetic model, the domain structure is
described by the spatial distribution of the local magnetization M(r). The temporal evolution of the magnetization conguration, thus the domain structure, is
described by the LandauLifshitzGilbert (LLG)
equation
1 a2

oM
ca
c0 M  Heff  0 M  M  Heff ;
ot
Ms

where Ms is the saturation magnetization, c0 is the gyromagnetic ratio, a is the damping constant, and He is the
eective magnetic eld, which can be represented as a
variational derivative of the total free energy of the system with respect to magnetization
1 oE
;
2
Heff 
l0 oM
where l0 is the permeability of vacuum, the total free energy of magnetostricitve materials is given by
E Eanis Eexch Ems Eexternal Eelastic ;

where Eanis, Eexch, Ems, Eexternal, Eelastic are the magnetocrystalline anisotropy energy, exchange energy, magnetostatic energy, external eld energy, and elastic
energy, respectively.
The magnetocrystalline anisotropy energy of a cubic
crystal is
Z


Eanis
K 1 m21 m22 m21 m23 m22 m23 K 2 m21 m22 m23 dV ;
4
where mi are the components of the unit magnetization
vector, m = M/Ms. K1 and K2 are the anisotropy
constants.
The exchange energy is determined solely by the spatial variation of the magnetization orientation and can
be written as
Z
Eexch A m21;1 m21;2 m21;3 m22;1 m22;2 m22;3
m23;1 m23;2 m23;3 dV ;

where A is the exchange stiness constant. In this paper,


a comma in a subscript stands for spatial dierentiation,

J.X. Zhang, L.Q. Chen / Acta Materialia 53 (2005) 28452855

for example, mi,j = omi/oxj, where xj is the jth component of position vector in the Cartesian coordinates.
The magnetostatic energy of a system can be written
as
Z
1
Ems  l0 M s Hd  m dV ;
6
2
where Hd is the stray eld that is determined by the
long-range interaction among the magnetic moments
in the system. The stray eld Hd is governed by [29]
H d1;1 H d2;2 H d3;3 M s m1;1 m2;2 m3;3 ;

where Hdi are the components of Hd. By introducing


magnetic scalar potential /,
H di /;i :

Eq. (7) is thus rewritten as,


D/ M s m1;1 m2;2 m3;3 :

The solution of the potential for Eq. (9) is given in Fourier space by,
M s m1 nn1 m2 nn2 m3 nn3 
;
10
n21 n22 n23
p
where i 1; ni are the coordinates in Fourier space,
and /(n), mi(n) are the Fourier transforms of / and mi,
respectively. The value of / in real space can be obtained through an inverse Fourier transform of /(n).
The value of Hd can then be calculated by Eq. (8).
Note that the origin of the Fourier space
(n1 = n2 = n3 = 0) is a singularity point in Eq. (10) and
its contribution corresponds to the demagnetization
eld caused by the average magnetization of the system.
This can easily be seen by writing the local magnetization eld as
 dMr;
Mr M
11
/n i

 is spatially independent average magnetization


where M
and dM is the spatially dependent heterogeneous part of

the magnetization
R eld. The average eld M is dened in
such a way that dMr dV 0. Since the contribution
from (n1 = n2 = n3 = 0) is excluded in Eq. (10), it only includes the contribution from the heterogeneous part of
the magnetization eld dM. Therefore, in order to consider the demagnetization eld caused by the average
magnetization in this work, we approximate the demagnetization eld by
 N M;

Hd M

12

where N is the demagnetizing factor, which depends


only on the shape of the specimen. It should be pointed
out since we assume periodic boundary conditions for
the heterogeneous magnetization, Eq. (12) is an approximation for incorporating the eect of sample shape on
domain structures. Such an approximation is, in principle, only valid if the simulation system size is much

2847

smaller than the real sample size to be simulated.


Although the demagnetization factor, N, is known for
certain specic shapes, in general, it has to be computed
numerically.
The eect of an external applied magnetic eld Hex on
the system energy can be taken into account through the
interaction between the magnetization and the external
eld
Z
Eexternal l0 M s Hex  m dV :
13
For a cubic magnetostrictive material, the deformation associated with the local magnetization is described
by the stress-free strain:


3
1
3
e011 k100 m21  ; e012 k111 m1 m2 ;
2
3
2


3
1
3
14
e022 k100 m22  ; e013 k111 m1 m3 ;
2
3
2


3
1
3
e033 k100 m23  ; e023 k111 m2 m3 ;
2
3
2
where k100 and k111 are the magnetostrictive constants of
a cubic crystal.
To accommodate the local deformations arising from
the magnetostrictive eect, the elastic strains eij and thus
elastic energy Eelastic are generated in a magnetic domain
structure,
eij eij  e0ij ;

15

where eij is the total strain. The corresponding elastic energy can be expressed as
Z
Z
1
1
cijkl eij ekl dV
cijkl eij  e0ij ekl  e0kl dV ;
Eelastic
2
2
16
where cijkl is the elastic stiness tensor. The summation convention for the repeated indices is employed
and i, j, k, l = 1, 2, 3. For a cubic material with its
three independent elastic constant c11, c12 and c44 in
the Voigts notation, the elastic energy can be rewritten as

Eelastic

Z 

1
c11 e211 e222 e233
2

c12 e11 e22 e22 e33 e11 e33



2c44 e212 e223 e213 dV
Z 
1
2
2
2
c11 e11  e011 e22  e022 e33  e033 

2
c12 e11  e011 e22  e022 e22  e022 e33  e033
e11  e011 e33  e033  2c44 e12  e012

0 2
0 2
e23  e23 e13  e13  dV

17

2848

J.X. Zhang, L.Q. Chen / Acta Materialia 53 (2005) 28452855

which can be separated into three contributions,


Eelastic Eelastic1 Eelastic2 Eelastic3
with
Eelastic1

Z 

18

1
c11 e211 e222 e233
2

c12 e11 e22 e22 e33 e11 e33



2c44 e212 e223 e213 dV ;

Eelastic2

Z ("


2c44

m21 m22

3
k111
2

m22 m23

2

K 31
19


2 #
3
 c11  c12 k100
2
)

m21 m23

dV ;

21
Following Khachaturyans theory [31], the total
strain eij(r) may be represented as the sum of homogeneous and heterogeneous strains:
eij r eij gij r:
22
The homogeneous strain is dened in such a way so that
Z
23
gij r dV 0:
The homogeneous strain represents the macroscopic
shape change of a system generated due to the formation
of a domain structure. The heterogeneous strain does
not aect the macroscopic shape of a system.
The equilibrium heterogeneous strain satises the
mechanical equilibrium condition given by the Euler
equation with respect to the elastic displacement
24

where rij are the stress components and are given by


rij cijkl ekl cijkl ekl  e0kl .
For the case of homogeneous modulus approximation, the equilibrium heterogeneous strain can be calculated by solving the equation Eq. (24) in Fourier space.
First, we introduce a set of displacements ui(r),
gij 12ui;j uj;i :

25

The equations of equilibrium (24) are thus rewritten as


cijkl uk;lj cijkl e0kl;j :

26

The general solution of the displacement eld for


Eq. (26) is given in Fourier space by
ui n X j N ij n=Dn;

K 33

Dn l2 k 2l vn6
lv2k 2l vn2 n21 n22 n21 n23 n22 n23
v2 3k 3l vn21 n22 n23 ;

29

N 11 n l2 n4 lk l vn2 n22 n23

Z 

rij;j 0;

K 32

and D(n) is the determinant of matrix K[n]. Note that


Kki = ckjilnjnl.
For cubic crystals, the explicit expressions of D(n)
and Nij(n) are:

20

3
 k100 c11  c12 e11 m21 e22 m22 e33 m23
2

6k111 c44 e12 m1 m2 e23 m2 m3 e13 m1 m3 dV :

Eelastic3

where X i icijkl e0kl nnj ; ui n and e0kl n are the Fourier transforms of ui and e0kl , respectively, Nij(n) are
cofactors of a 3 3 matrix K[n],
2
3
K 11 K 12 K 13
6
7
28
Kn 4 K 21 K 22 K 23 5

27

v2k 2l vn22 n23 ;


N 12 n k ln1 n2 ln2 vn23
and other components are obtained by the cyclical permutation of 1, 2, 3, where
k c12 ;

l c44 ;

v c11  c12  2c44 ;

and n2 ni ni :
30

The displacement eld ui(r) in the real space can be


obtained through an inverse Fourier transform of
ui(n). Consequently, the heterogeneous strain can be calculated by Eq. (25).
The value of the homogeneous strain depends on the
boundary conditions. For a clamped boundary condition, the system as a whole is not allowed to deform.
Therefore, the homogeneous strain is zero. When a system is subject to a constant homogeneous applied strain
eaij , the homogeneous strain is simply equal to the applied strain. If there is no external stress applied and
the system is unconstrained with respect to the macroscopic deformation, the homogeneous strain is obtained
by minimizing the total elastic energy respect to the
homogeneous strain. Substituting Eq. (22) into the total
elastic energy given in Eq. (16) and using the fact of
Eq. (23), we obtain
Z



1
Eelastic
cijkl eij gij r  e0ij ekl gkl r  e0kl dV
2
Z
Z
V
0
cijkleijekl  cijkleij ekl dV cijkleij gkl r dV
2
Z



1
cijkl gij r  e0ij gkl r  e0kl dV

2
Z
V
cijkleijekl  cijkleij e0kl dV
2
Z



1
cijkl gij r  e0ij gkl r  e0kl dV ;
31

2
where V is the total volume of the system.

J.X. Zhang, L.Q. Chen / Acta Materialia 53 (2005) 28452855

Minimizing it with respect to the homogeneous


strain, i.e.,
Z
oEelastic
0 Vcijklekl  cijkl e0kl dV ;
32
oeij
we have
Z
1
ekl
e0kl dV :
V

33

Using the stress-free strain given in Eq. (14), the


homogeneous strain is given by:


3
1
3
2
e11 k100 m1  ; e12 k111 m1 m2 ;
2
3
2


3
1
3
e22 k100 m22  ; e13 k111 m1 m3 ;
34
2
3
2


3
1
3
e33 k100 m23  ; e23 k111 m2 m3 ;
2
3
2
m2i

where
and mi mj represent the volume average of
m2i and mi mj over a system containing a domain structure, respectively.
When a system is subject to a homogeneous applied
stress raij , the total potential energy is given by the sum
of elastic energy of the system and the potential of the
mechanical loading,
Ep Eelastic  V raijeij :

oEp oEelastic V raijeij


oeij
oeij
oeij
Z
Vcijklekl  cijkl e0kl dV  V raij 0:
Therefore
Z
1
ekl
e0kl dV sijkl raij ;
V

The total strain can be calculated by Eq. (22). Consequently, the elastic energy can be obtained for a magnetostrictive material with an arbitrary magnetic domain
structure by Eq. (16).

3. Simulation results and discussion


The magnetic domain structures are obtained by
the numerically solving the LLG equation using the
GaussSeidel projection method in this work [39]. The
LLG equation can be rewritten in dimensionless form,
by employing the following set of dimensionless
variables:
M M s m;
Heff M s heff ;

39

1 a2
s:
t
c0 M s
The equation becomes,
om
m  heff  am  m  heff :
os

35

Minimizing the total potential energy with respect to the


homogeneous strain, we have

2849

40

We can rewrite the eective magnetic eld as


1
Heff1 Heff2
Ms


 

1
1 oEexch
1 oE0



Ms
l0 oM
l0 oM

0
1
1
oE
A Dm

Ms
l0 oM

heff

36

41

A Dm hm;
37

where sijkl is the elastic compliance tensor. Using Eq.


(14), we obtain the homogeneous strain:


3
1
a
a
a
2
e11 s11 r11 s12 r22 r33 k100 m1  ;
2
3
1
3
e12 s44 ra12 k111 m1 m2 ;
2
2


 a
3
1
a
a
2
e22 s11 r22 s12 r11 r33 k100 m2  ;
2
3
38
1
3
a
e13 s44 r13 k111 m1 m3 ;
2
2


3
1
a
a
a
2
e33 s11 r33 s12 r11 r22 k100 m3  ;
2
3
1
3
e23 s44 ra23 k111 m2 m3 ;
2
2
where s11, s12, and s44 are the three independent compliance constants for a cubic material in the Voigts notion.

where E 0 is the total free energy excluding the exchange


energy part. A* is dened as A 2A=l0 M 2s l2d , where ld is
the cell size in the model.
Following Ref. [39], we solve the equation
om
m  A Dm hm  am  m  A Dm hm
os
42
in three steps:
Step 1:
1

gni 1  A DsD mni Dshi mn ;


1

gi 1  A DsD mi Dshi mn ;


1
1 0 n
m1 gn2 mn3  gn3 mn2
m1
C
B C B n
@ m2 A @ m2 gn3 m1  g1 mn3 A:

i 1; 2; 3:

43

m3

mn3

g1 m2

g2 m2

44

2850

J.X. Zhang, L.Q. Chen / Acta Materialia 53 (2005) 28452855

Step 2:
0  1 0 
1

m1 aDsA Dm
m1
1 h1 m 
B  C B 

 C
@ m2 A @ m2 aDsA Dm
2 h2 m  A:
m
3

m3

aDsA

Dm
3

Step 3:
0 n1 1
0  1
m1
m1
1 B  C
B n1 C
@ m2 A  @ m2 A:
jm j
m
mn1
3
3

45

h3 m 

46

We solve Eqs. (43) and (45) by employing the Fourier


transform. For example, by Fourier transforming both
sides of Eq. (43), the equation becomes

gni n

mni n Ds~
hi mn 
;
2
2
1 n1 n3 n23 A Ds

47

where gni n; mni n; and ~


hi mn  are the Fourier transn
n
forms of gi ; mi ; and hi mn , respectively. The value of
gni in real space can be obtained by doing an inverse
Fourier transform of gni n. Eq. (45) can be solved in a
similar way.
3.1. Domain structures and magnetostriction in FeGa
We rst consider Fe81.3Ga18.7 as an example for the
numerical simulations. The corresponding material
parameters are from Refs. [35]: Ms = 1.432 106 A/m,
K1 = 2 104
J/m3,
K2 = 4.5 104
J/m3,
and
4
k100 = 2.64 10 , k111 = 0. The bulk cubic elastic constants are c11 = 1.96 1011 N/m2, c12 = 1.56 1011 N/
m2, and c44 = 1.23 1011 N/m2. To save the computational time and since the magnetic easy axes are along
1 0 0 directions, we performed the simulations with
(512 512 1 or 256 256 1) discrete cells, i.e., essentially 2D systems. Periodic boundary conditions are applied along the x1, x2 and x3 axes. The magnetization is
assumed to be uniform in each cell, but it is allowed to
rotate in three dimensions (3D). For the damping constant, a value of a = 0.5 is used in the simulation. In reduced variables, the exchange stiness constant is chose
be A* = 0.0625, and the time step is Ds = 0.1.
3.1.1. Energetic contributions and domain structures
Giant magnetostrictive materials usually exhibit
rather complicated domain structures resulted from
competing energetic contributions such as elastic energy
and magnetostatic energy. To understand the relation of
each energetic contribution to the resulting domain
structure, we performed simulations taking into account
dierent energy contributions.
The domain structures obtained are shown in Figs. 1,
35. Fig. 1 shows the temporal evolution of a domain
structure in the presence of elastic energy but without

Fig. 1. Domain evolution in the presence of elastic energy but the


absence of magnetostatic energy: (a) 4000 steps; (b) 40,000 steps; (c)
200,000 steps; (d) 400,000 steps. Black = [1 0 0] and 1 0 0 domains,
light gray = [0 1 0] and 0 1 0 domains, and dark gray = [0 0 1] and 0 0 1
domains.

the magnetostatic energy. Mechanical clamped boundary conditions were used, which means that the overall
dimensions of the system do not change. At the beginning, the random initial conguration evolves to domains as shown in Fig. 1(a). All the six dierent kinds
of orientation domains determined by the cubic anisotropy energy, including the domains with magnetization
along the positive x1 direction ([1 0 0] domain), negative
x1 direction (1 0 0 domain), positive x2 direction ([0 1 0]
domain), negative x2 direction (0 1 0 domain), positive
x3 direction ([0 0 1] domain), and negative x3 direction
(0 0 1 domain), are present with almost equal fractions.
Fig. 1(b) illustrates the domain structure after 40,000
steps of evolution, indicating domain growth and coarsening. The domains are separated by 180 or 90 domain
walls. A schematic diagram showing the 180 domain
and 90 domain walls separating the domains is given
in Fig. 2. The 90 domain walls tend to align along
the [1 1 0] or 1 1 0 directions separating [1 0 0] domain
(or 1 0 0 domain) and [0 1 0] domain (or 0 1 0 domain).
The alignment becomes increasingly strong at the later
stage of evolution as shown in Fig. 1(c) and (d). Since
in this case, the magnetostatic energy is not included
and the exchange energy is isotropic, the domain wall
alignment must be entirely due to the anisotropic elastic
interactions. For 90 domain walls, the condition of
elastic compatibility between the associated domain
pairs requires that the tangential components of the

J.X. Zhang, L.Q. Chen / Acta Materialia 53 (2005) 28452855

(a)

(b)

Fig. 2. Schematic diagram of the 180 domain wall and 90 domain
wall separating the domains: (a) 180 domain wall; (b) 90 domain
wall.

stress-free strain must be equal in both domains for the


given wall orientation. For a pair of [1 0 0] (or 
1 0 0) and
[0 1 0] (or 0 
1 0) domains, the corresponding wall plane
is (1 1 0) (or 1 
1 0), which is automatically predicted
by the simulations without a priori assumptions. For
180 domain walls, since the magnetizations are antiparallel on either side of the wall and correspond to
identical strains, there is no preferred orientation of
the 180 domain walls as shown in Fig. 1(b) and (c). It
is interesting to note that the 180 domain walls become
straight and perpendicular to the neighboring 90 domain walls as shown in Fig. 1(d). The directional alignment for the 180 walls can be easily understood as a
result of domain wall minimization associated with the
straight domain walls. Periodic segments were observed
in some 180 domain walls as shown in Fig. (1). Similar
substructures along 180 domain walls have been observed and studied experimentally [40], in which case
the subdivision of domain wall was explained as a result
of magnetostatic energy minimization. To reveal the origin of the subdivision of the 180 domain wall observed
here, we simulated the domain structure without considering both the magnetostatic energy and elastic energy.
As Fig. (3) shows, the substructures along 180 domain
walls still exist, which indicates that neither the magnetostatic energy nor elastic energy is responsible for the
formation of the substructure. They are formed due to
degeneracy of the domain wall with four dierent orientations that have equal energy. It is also noted, as expected, that since the magnetostatic energy is absent,
both the head to tail and head to head (tail to tail) domain walls exist in the domain structure.
One may also notice that in Fig. 1 the fraction of
[0 0 1] domain and 0 0 
1 domain decreases with time.
Eventually as shown in Fig. 1(c) and (d), only those
[1 0 0] (or 
1 0 0) and [0 1 0] (or 0 
1 0) domains survive.
This is easily understandable since this is a quasi 2D

2851

Fig. 3. Simulated domain structure in the absence of both magnetostatic and elastic energies. Black = [1 0 0] and 1 0 0 domains, light
gray = [0 1 0] and 0 1 0 domains, and dark gray = [0 0 1] and 0 0 1
domains.

simulation, and hence the domain walls between [0 0 1]


(or 0 0 1) domains and the other four domains with
magnetization perpendicular to the x3 axis are parallel
to the [0 0 1] direction instead of the energetically favorable 0 1 1=0 1 1 or 1 0 1=1 0 1 planes.
Experimental results [35] showed that the magnetostrictive constant of FeGa alloy is sensitive to the material composition and temperature. Therefore, it is useful
to study the domain structure of magnetostrictive materials with dierent magnetostriction constants. Fig. 4
shows the simulated domain structure with varying
magnetostrictive constants: k100 = 1.32 104, 2.64
105 (k111 = 0), which are 0.5 and 0.1 times of that for
Fe81.3Ga18.7, respectively. With the decrease of magnetostrictive constant, the domain walls become less aligned,
since the contribution of elastic energy becomes increasingly insignicant. We also note that the 0 0 1 domains
survive for the case of low magnetostrictive constants,
which conrm the responsibility of the elastic energy
for the disappearance of [0 0 1] (or 0 0 1) domains in
materials with high magnetostrictive constants in our
2D simulations as shown in Fig. 1.

Fig. 4. Simulated domain structure with varying magnetostrictive


constants: (a)k100 = 1.32 104, k111 = 0; (b) k100 = 2.64 105, k111 =
0 (in the presence of elastic energy but the absence of magnetostatic
energy). Black = [1 0 0] and 1 0 0 domains, light gray = [0 1 0] and
0 1 0 domains, and dark gray = [0 0 1] and 0 0 1 domains.

2852

J.X. Zhang, L.Q. Chen / Acta Materialia 53 (2005) 28452855

It should be noted that some other material parameters of FeGa alloy beside magnetostriction constants
were also found to depend on composition and temperature, such as saturation magnetization and cubic
anisotropy constant. They should also aect the domain
structure of FeGa alloy, which was not studied in this
work. Thus, complicated behaviors of domain structures
in FeGa are expected depending on composition and
temperature.
Fig. 5 shows the temporal evolution of domain structure when the magnetostatic energy and elastic energy
are both taken into account. In this case, the domain
wall orientation is constrained by both the elastic compatibility and the magnetic compatibility. Two adjacent
domains are magnetically compatible if the normal components of the magnetization vectors of both domains
with respect to their common domain wall are equal.
Therefore, only head to tail domain walls survive. Since
the discontinuities in the strain do not occur for 180 domain walls, the 180 domain walls are only constrained
by the magnetic compatibility. It is shown that the 180
domain walls are found to be parallel to the magnetization direction of associated domain pair (Fig. 5(d)). Finally, a twin-like domain structure was obtained as

Fig. 5. Domain evolution in the presence of both elastic energy and


magnetostatic energy: (a) 4000 steps; (b) 40,000 steps; (c) 200,000 steps;
(d) 500,000 steps. Black = [1 0 0] and 1 0 0 domains, light gray = [0 1 0]
and 0 1 0 domains, and dark gray = [0 0 1] and 0 0 1 domains.

Fig. 6. Simulated domain structure with dierent applied pre-stress (along x1 axis): (a) 5 MPa (61% domain with magnetization along x1 axis); (b) 0
MPa (49%); (c) 5 MPa (32%). Dark gray = [1 0 0] domains, black = 1 0 0 domains, light gray = 0 1 0 domains, and white = [0 1 0] domains.

300

0.6
0.4
0.2
0.0

(a)

2
4
6
Applied Field (kA/m)

(b)

300
5 MPa
0 MPa
-5 MPa

250
200
150
100
50
0

2
4
6
8
Applied Field (kA/m)

Magnetostriction (x10-6)

M/Ms

0.8

Magnetostriction (x10-6)

5 MPa
0 MPa
-5 MPa

1.0

(c)

5 MPa
0 MPa
-5 MPa

250
200
150
100
50
0
0.0

0.2

0.4
0.6
M/Ms

0.8

1.0

Fig. 7. The eld dependence of magnetization and magnetostriction for the giant magnetostrictive material with dierent pre-stress applied: (a) M/Ms vs.
applied eld curves; (b) magnetostriction vs. applied eld curves; (c) magnetostriction vs. M/Ms curves.

J.X. Zhang, L.Q. Chen / Acta Materialia 53 (2005) 28452855

shown in Fig. 5(d), which is similar to that described in a


prior theoretical prediction [17]. If the exchange stiness
constant of Fe81.3Ga18.7 has a similar magnitude as Fe
(A  1011 J/m), the system studied here is around
3 3 lm.
3.1.2. Magnetization process
The overall strain response of a magnetostrictive
material is directly related to the domain evolution in
the material under applied elds. The eld-induced magnetostriction under a certain applied eld depends on

2853

the initial demagnetized domain structure in a crystal.


The largest magnetostriction appears only when an ideal
initial domain structure is formed, i.e., all the domains
align perpendicular to the applied magnetic eld direction. Such an alignment can be obtained by an applied
pre-stress. From Eq. (21), it can easily been seen that
the elastic energy contribution can alter the anisotropy
energy. For the positive k100 we used here, a tensile
stress, that produces an elongation in the x1 axis (and
hence contraction in x2 and x3 axes), energetically favors
the domains with magnetization along the x1 axis while

Fig. 8. The domain structure evolution under applied eld along [1 0 0] (without pre-stress): (a) 0 kA/m; (b) 0.80 kA/m; (c) 4.38 kA/m; (d) 7.96 kA/m.
Dark gray = [1 0 0] domains, black = 1 0 0 domains, light gray = 0 1 0 domains, and white = [0 1 0] domains.

2854

J.X. Zhang, L.Q. Chen / Acta Materialia 53 (2005) 28452855

a compressive stress will favor domains with magnetization perpendicular to the x1 axis. Fig. 6 shows the domain structures of samples with dierent applied prestresses. As expected, while a tensile stress increases
the fraction of domains along x1 axis, a compressive
stress applied along x1 axis decreases it.
The magnetostriction and associated domain structure evolution under an applied magnetic eld for the
samples with dierent applied pre-stress were shown in
Figs. 7 and 8. In prior experiments [35], to obtain large
magnetostriction at a relatively low magnetic eld, the
samples were usually prepared in the shapes of a thin
disk or a rod to decrease the demagnetization eect.
To approximate this case, in our simulation we use the
demagnetizing factor of innite thin sheet (Nx =
Ny = 0, Nz = 1). A magnetostriction vs. M/Ms curve for
the magnetization process was given in Fig. 7(c). The
initial at stage of the curve show that the initial increase of magnetization is accompanied by a small magnetostrictive change. As Fig. (8) shows, such an initial
magnetization stage mainly consist of 180 domain wall
movement (such as region A in Fig. 8(a) and (b)), which
does not produce any magnetostrictive change in dimensions. 90 domain wall movement also occurred by
switching from 1 0 0 to [0 1 0] (or 0 
1 0) domains or
from [0 1 0] (or 0 
1 0) to [1 0 0] domains (such as region
B in Fig. 8(a) and (b)). As magnetostriction vs. M/Ms
curve measures the average behavior of the system, the
eect of such two distinct 90 domain wall motion is
equal to a 180 domain switching from 
1 0 0 to [1 0 0]
domain, which does not produce magnetostrictive
change. With further magnetization, the fraction of the

1 0 0 domain deceases, and the associated 180 domain
switching decreases too. As shown in the magnetostriction vs. M/Ms curve, the slope increases with magnetization. Finally, a single domain with saturated
magnetization and magnetostriction along the applied
eld was obtained as shown in Fig. 8(d). It is shown that
a pre-stress can indeed have a signicant eect on the
magnetostriction. A larger saturation magnetostriction
was obtained for systems with a compressive pre-stress,
while a smaller magnetostriction was obtained when a
tensile pre-stress was applied.

Fig. 9. Simulated domain structure of Terfenol-D. Black = [1 1 1] and


1 1 1 domains, dark gray = 1 1 1 and 1 1 1 domains, light gray =
1 1 1 and 1 1 1 domains, and white = 1 1 1 and 1 1 1 domains.

c44 = 4.87 1010 N/m2. Because of the extensive computational time required for 3D simulations, in this paper
we only focused on the prediction of 3D domain structures without an applied eld. An example of 3D domain structure is shown in Fig. 9. Domains with
magnetization along the 8 possible 111 easy directions
were observed. The obtained domain wall orientations
of non-180 walls, which constrained by the elastic compatibility, agree with those predicted by theory of James
and Kinderlehrer [17,18] and with prior experimental
observations [6,11]. It should be emphasized that these
domain wall orientations were predicted without a priori
assumptions. The orientations of 180 domain wall are
not exactly along those planes determined by the magnetic compatibility, which comes from the competition
between magnetostatic and exchange energy, since the
exchange energy prefers a at domain wall with minimization of wall. Extensive studies on the magnetic behavior of Terfenol-D such as the magnetic process with
applied magnetic eld or stress will be carried out using
extensive 3D simulations.

3.2. 3D Domain structures in Terfenol-D


4. Conclusion
As Terfenol-D has a negative magnetocrystalline
anisotropy with easy directions along 1 1 1, it is not
possible to approximate the domain structures in 2D.
Therefore, we chose a 3D system with 128 128 128
cells. The periodic boundary conditions are applied
along all three Cartesian axes. The material parameters
are [26]: Ms = 8.0 105 A/m, A = 9 1012 J/m,
K1 = 6 104 J/m3, K2 = 0 J/m3 and k100 = 0, k111 =
1.64 103. The bulk cubic elastic constants are
c11 = 1.41 1011 N/m2, c12 = 6.48 1010 N/m2, and

A computational model for predicting the stability of


domain structures and their temporal evolution in giant
magnetostrictive materials is developed by combining a
micromagnetic model with the phase-eld microelasticity of Khachaturyan for arbitrary domain structures.
It takes into account both the inhomogeneous stress
and magnetic eld distributions in a domain structure.
Its applications to FeGa and Terfenol-D demonstrated
that the model correctly predicts the domain wall orien-

J.X. Zhang, L.Q. Chen / Acta Materialia 53 (2005) 28452855

tations as well as the domain structure evolution and the


magnitude of magnetostriction during a magnetization
process. It is shown that the elastic energy minimization
results in an alignment of 90 domain wall orientation
while the magnetostatic energy is responsible for the orientation of 180 domain walls and the headtail magnetization conguration across a domain wall. Subdivision
of 180 domain walls was observed, which was attributed to the degeneracy of the domains with dierent
magnetic easy directions rather than the elastic interactions or the magnetostatic interactions that were believed to be responsible previously. Under an applied
magnetic eld, the domain structure evolution starts
with the 180 domain switching, which does not produce
any magnetostriction. It is also demonstrated that prestress can be eciently used to increase the magnetostriction changing the initial magnetic domain structure
in the demagnetization state.

Acknowledgment
The authors are grateful for the nancial support by
the National Science Foundation under the Grant No.
DMR 01-22638.

References
[1] Clark AE. Magnetostrictive rare earth-Fe2 compounds. In:
Wohlfarth EP, editor. Ferromagnetic materials. Amsterdam: North-Holland; 1980.
[2] Jiles DC. J Phys D: Appl Phys 1994;27:1.
[3] Clark AE, Restor JB, Wun-Fogle M, Lograsso TA, Schlagel
DL. IEEE Trans Mag 2000;36:3238.
[4] Clark AE, Hathaway KB, Wun-Fogle M, Restor JB, Lograsso
TA, Keppens VM, et al. J Appl Phys 2003;93:8621.
[5] Raque S, Cullen JR, Wuttig M, Cui J. J Appl Phys 2004;95:6939.
[6] Engdahl G. Handbook of giant magnetostrictive material. New
York, NY: Academic Press; 2000.
[7] Holden AP, Lord DG, Grundy PJ. J Appl Phys 1996;79:4650.
[8] Dooley J, Graef MD, McHenry ME. J Appl Phys 1998;83:6837.

2855

[9] Holden AP, Lord DG, Grundy PJ. J Appl Phys 1996;79:6070.
[10] Lord DG, Holden AP, Grundy PJ. J Appl Phys 1997;81:5728.
[11] Schmidt J, Tickle R, Skidmore GD, Merton C, James RD,
Dahlberg ED. J Magn Magn Mater 1998;190:98.
[12] Shull RD, Quandt E, Shapiro AT, Glasmachers S, Wuttig M. J
Appl Phys 2004;95:6948.
[13] Busbridge SC, Piercy AR. J Appl Phys 1993;73:5354.
[14] Wang BW, Busbridge SC, Li YX, Wu GH, Piercy AR. J Magn
Magn Mater 2000;218:198.
[15] Wang BW, Busbridge SC, Guo ZJ, Zhang ZD. J Appl Phys
2003;93:8489.
[16] Brown WF. Magnetoelastic interactions. Berlin: Springer; 1966.
[17] James RD, Kinderlehrer D. Philos Mag B 1993;68:237.
[18] James RD, Kinderlehrer D. J Appl Phys 1994;76:7012.
[19] DeSimone A, James RD. J Mech Phys Solids 2002;50:283.
[20] Jiles DC, Thoelke JB. J Magn Magn Mater 1994;134:143.
[21] DeSimone A, James RD. J Appl Phys 1997;81:5706.
[22] Zhao XG, Lord DG. J Magn Magn Mater 1999;195:699.
[23] Armstrong WD. J Appl Phys 2002;91:2202.
[24] Park WJ, Son DR, Lee ZH. J Magn Magn Mater 2002;248:
223.
[25] Zhu B, Lo CCH, Lee SJ, Jiles DC. J Appl Phys 2001;89:7009.
[26] Shu YC, Lin MP, Wu KC. Mech Mater 2004;36:975.
[27] Fabian K, Heider F. Geophys Res Lett 1996;23:2839.
[28] Shu YC, Yen JH. Mater Res Soc Symp Proc 2004;785:D13.4.1.
[29] Hubert A, Schafer R. Magnetic domains: the analysis of magnetic
microstructure. Berlin: Springer; 1998.
[30] Khachaturyan AG, Shatalov GA. Sov Phys JETP 1969;29:557.
[31] Khachaturyan AG. Theory of structural transfomations in
solid. New York, NY: Wiley; 1983.
[32] Chen LQ, Wang YZ, Khachaturyan AG. Philos Mag Lett
1992;65:15.
[33] Wang Y, Khachaturyan AG. Acta Mater 1997;45:759.
[34] Jin YM, Artemev A, Khachaturyan AG. Acta Mater
2001;49:2309.
[35] Chen LQ, Hu SY. Phase-eld method applied to strain-dominated
microstructure evolution during solid-state phase transformations. In: Raabe D, Barlat F, Roters F, Chen LQ, editors.
Continuum scale simulation of engineering materials fundamentals microstructures process applications. New York,
NY: John Wiley; 2004.
[36] Li YL, Hu SY, Liu ZK, Chen LQ. Appl Phys Lett 2001;78:3878.
[37] Li YL, Hu SY, Liu ZK, Chen LQ. Acta Mater 2002;50:395.
[38] Seol DJ, Hu SY, Li YL, Shen J, Oh KH, Chen LQ. Acta Mater
2003;51:5173.
[39] Wang XP, Garcia-Cervera CJ, E WN. J Comput Phys 2001;171:
357.
[40] Hartmann U, Mende HH. J Appl Phys 1986;59:4123.

You might also like