Journal of Alloys and Compounds 615 (2014) 65–74

Contents lists available at ScienceDirect

Journal of Alloys and Compounds

journal homepage: www.elsevier.com/locate/jalcom

Finite element modeling of magnetoelectric laminate composites

in considering nonlinear and load effects for energy harvesting
Hakeim Talleb ⇑, Zhuoxiang Ren
Sorbonne Universités, UPMC Univ Paris 06, UR2, L2E, F-75005 Paris, France

a r t i c l e i n f o a b s t r a c t

Article history: This paper presents the finite element analysis of a magnetoelectric energy harvester using a laminate
Received 24 April 2014 composite constituted of laminated piezoelectric and magnetostrictive layers. In this study, both the
Received in revised form 13 June 2014 nonlinear characteristics of the material and the dependency on the load impedance are considered.
Accepted 18 June 2014
The multiphysics problem involving different physics equations is solved through a strongly coupled
Available online 30 June 2014
model. The nonlinear magnetostrictive behavior is considered using the Newton–Raphson method for
various magnetic biases. The electrical circuit equation is incorporated in the finite element equations
Keywords:
for the analysis load effect. The obtained results show a good concordance with the measurements
Magnetostriction
Piezoelectricity
and with those obtained by other analytical methods.
Energy harvesting Ó 2014 Elsevier B.V. All rights reserved.
Finite elements

1. Introduction generalized in dynamic regime and enriched by Bichurin et al.

In the last decade, the multiple-phase magnetoelectric (ME)
composite materials composed of laminated piezoelectric and mag-
netostrictive layers have received intense interests from many
researchers in various scientific domains. Indeed, their property to
exhibit, under a simultaneously applied external dc bias Hdc field
and ac field Hac , a high output voltage magnetic sensitivity is very
practical for modern potential applications such as magnetic field
sensors, gyrators, variable inductances or energy transducers
[1–3]. Although early studies have started in the years 1970–1980
with notably the discovery of Terfenol-D (Tb1-xDyxFe2), it was
not until the early 2000s to see their interest arise with the emer-
gence of new magnetostrictive composite materials such as
Metglas, FeGa (Galfenol), TbFeCo (Terfecohan), NiZnFe2O4 or
FeCuNbSiB and piezoelectric materials such as PMN-PT (Pb(Mg,
Nb)O3–PbTiO3), PZT-4/5/8 (Pb(Zr, Ti)O3), BTO (BaTiO3) or NiFe2O4
(NFO). The parameter of importance to predict the performance of
a ME composite material is the magnetoelectric field or voltage
coefficient defined as aE ¼ dE=dH or aV ¼ dV=dH, where dE, dV are
dynamic electric field and dynamic voltage, respectively, when an
ac magnetic dH is applied. Employing a homogenous approach for
bilayer composite, Harshe et al. [4] proposed a 2D longitudinal (L)
aE for low-frequency depending on field orientations, boundary
conditions and material parameters (nature, number, and volume
fraction of the layer thicknesses). These expressions have been con-
firmed in non-linear case by Srinivasan et al. [5] and were
⇑ Corresponding author. Tel.: +33 144274310.
E-mail address: hakeim.talleb@upmc.fr (H. Talleb).
[6,7] in including transverse modes (T) and a coefficient factor to
take account the imperfect interface coupling. Dong et al have pro-
posed 1D expressions [8–10] of the aV for trilayer laminate compos-
ites Terfenol-D/PZT/Terfenol-D in using an equivalent circuit
approach under resonance drive according to different transversal
(T) or longitudinal (L) magnetization–polarization modes (T–T,
T–L, L–T, L–L). The works of Dong et al. are revealed that the mode
L–T (resp. L–L) is approximately five times higher than the mode
T–T (resp. T–L) for Terfenol-D/PZT laminate composites. It is inter-
esting to note that the Dong’s model aV can be established in trans-
posing the 2D models of Harsh, Bichurin in 1D. Recently Dongs
models have been enriched in adding a mechanical quality factor
Qmech [11,12] and newly they have been adjusted to integrate
the effective magnetic field in constitutive equations, the nonlinear
magnetostrictive effect due to piezomagnetic coefficient d33m , [13]
and the non-uniform magnetic field from the permeability influ-
ence [14,15]. Despite these recent adjusting, all models do not
include the global nonlinearity of each piezomagnetic coefficient
and do not take account the real mechanical impact when a resin
layer (like Epoxy) is used to stuck the magnetrostrictive and piezo-
electric layers or the mechanical damping in harmonic case. They do
not take account neither the electrical impact when the structure is
loaded by electrodes or by an electrical impedance. Moreover they
are not necessarily adapted to study specific structures incorporat-
ing different polarization and magnetization modes (L–T + T–T for
example) or thin layer structures exploiting ferromagnetic alloy
having an extremely high-permeability to achieved better ME prop-
erties [16,17]. Thus, the use of rigorous numerical modeling such as
the Finite Element Method (FEM) is essential for the design and

http://dx.doi.org/10.1016/j.jallcom.2014.06.121
0925-8388/Ó 2014 Elsevier B.V. All rights reserved.
66 H. Talleb, Z. Ren / Journal of Alloys and Compounds 615 (2014) 65–74

optimization of devices based on magnetoelectric laminate com- In considering small mechanical deformation, the mechanical
posites. Moreover, the simulation of the mechanical-magneto-elec- displacement u and the mechanical strain S is related to the
tric field distributions allows deeper understanding of the physical cross-section defined by the global coordinate directions 1 and 2:
phenomena and is helpful for design optimization. ^u
In the context of energy harvesting application, this paper pro-
S¼r ð2Þ
poses a 2D finite element model to investigate the performance of ^ is a gradient operator define as :
where r
a magnetic–electric energy transducer using conventional ME tri-
1 T
layer laminate composite. The nonlinearity of magnetic and piezo- S¼ ðgradu þ grad uÞ ð3Þ
magnetic coefficient as well as the load effect representing the 2
electronic circuit that collect and store the energy are considered. In our 2D problem, we consider the plane stress condition
The paper is organized as follows. Firstly, the general governing which satisfies the following properties:
equations and the constitutive laws as well as the finite element
T 33 ¼ T 13 ¼ T 23 ¼ 0 ð4aÞ
formulation with the applied boundary conditions are presented.
Secondly the static FEM problem is solved in considering the non-
t
S13 ¼ S23 ¼ 0; S33 ¼ ðT 11 þ T 22 Þ ð4bÞ
linear behavior of both magnetic and magnetostrictive materials E
for various magnetic biases. This allows determining the incremen- where E and t are the Young’s module and the Poisson ratio, respec-
tal magnetic and piezomagnetic properties for small signal analysis. tively. As shown in Fig. 1, the directions 1, 2, 3 represent the global
Thirdly, the harmonic problem is performed in considering a small Cartesian coordinates x, y, z.
signal ac field, Hac around a polarization point of the static case. A Magnetostatic equations:
harmonic mechanical damping is added with Rayleighs coefficients. 
curlðHÞ ¼ J
To verify the influence of volume fraction, a comparison to Dongs ð5Þ
analytical model is done. Fourthly, the load effect between the elec-
div ðBÞ ¼ 0
trodes is taking into account in adding an electrical equation in the where H the magnetic field, B the magnetic induction and J the cur-
harmonic problem. The simulation results are validated in compar- rent density.
ison with experimental results available in the literature. Finally Electrostatic equations:
the performances of ME composite are studied according to the 
curlðEÞ ¼ 0
estimation of the output power in function of the externally applied ð6Þ
dc magnetic bias field, the frequency work and the load value. div ðBÞ ¼ q
where E is the electric field, D the displacement field and q the vol-
2. Physics equations and finite element discretization ume density of the free electric charge.
The magnetic induction B and the electric field E are consid-
2.1. General constitutive laws ered in the cross-section defined by the global coordinate direc-
tions 1 and 2. The magnetic induction B can be expressed as
Fig. 1 shows the configuration of the trilayer laminate compos- the curl of the magnetic vector potential a. In 2D only the compo-
ite Terfernol-D/PZT-5A/Terfernol-D (with the local and global coor- nent a3 of a subsists and it is independent of the direction 3.
dinates) proposed to realize an energy transducer. The 2D FEM Using the potential variables, the electric field and the magnetic
formulation of this composite is based on the thermodynamical induction become:
approach presented in [18–23] that combine the mechanical 
equilibrium equation and the magnetostatic and electrostatic E ¼ gradðVÞ
ð7Þ
equations, respectively given by (1) and (5), (6) with the electro- B ¼ r  gradða3 Þ
magneto-mechanical constitutive equations given in (6).
with r  a rotation matrix in Cartesian coordinates:
@u  
div T þ f ¼ qm 2 ð1Þ 0 1
@t r ¼ ð8Þ
1 0
where T is the mechanical stress tensor, u the mechanical displace- The electro–magneto-mechanical constitutive laws for the compos-
ment, f the externally applied volume force, qm the mass density of ite material are the combination between the electro-mechanical
the medium. and the magneto-mechanical constitutive laws:

D ¼ E  eS
H ¼ vB  hS ð9Þ
T ¼ cS  et E  ht B
where  is the permittivity matrix, e is the piezoelectric coefficients
matrix, v the reluctivity matrix, h ¼ qm, with q the piezomagnetic
coefficients matrix and c the elasticity matrix.
In considering the applied magnetic field along 1-direction in
respect to global Cartesian coordinates defined in Fig. 1, the tensor
representations in L–T mode of the magnetostrictive and piezo-
electric layers in 2D are, respectively: Magnetostrictive material
in L magnetization:
2 3 2 3
q33 0 " # q33 mS33 0
t 6 7 mS33 0 t 6 7
q ¼ 4 q31 0 5; v ¼ ; h ¼ 4 q31 mS33 0 5; ð10aÞ
0 mS11
0 q15 0 q15 mS11
Fig. 1. Trilayer laminate composite in L–T mode.
 H. Talleb, Z. Ren / Journal of Alloys and Compounds 615 (2014) 65–74 67

2
c33 B c31 B 0
3 XZ
6 7 ½K ii  ¼ ½Gi t ½h½Gi dX
6 7 e Xe
c ¼ 6 c31 B c11 B 0 7 ð10bÞ 8
4 5
< c if i ¼ u
> ð15Þ
0 0 c66 B h ¼ e if i ¼ p
>
:
with a if i ¼ a

ckl B ¼ ckl H þ qki mSij qjl where a=½r  t v½r  , and

XZ
Piezoelectric material in T polarization: ½K ij  ¼ ½Gi t ½U½Gj dX
e Xe
2 3 2 3  ð16Þ
0 e31 c11 E c31 E 0 e if i ¼ u; j ¼ p
" # U¼
6
6
7
7
e11 0 6
6
7
7 h if i ¼ u; j ¼ a
et ¼ 6 0 e33 7;  ¼ ; c ¼ 6 c31 E c33 E 0 7 ð11Þ
4 5 0 e33 4 5
where Xe denotes the element domain.
e15 0 0 0 c55 E The solution domain of the magnetoelectric problem is illus-
In 2D plane stress condition the elastic coefficients ckl E and ckl H trated in Fig. 2, where the magnetic, electric and elastic boundary
are calculated with the Young’s module and Poisson coefficient. conditions are, respectively, indicated. The excitation is an exter-
The finite element formulation of the combined Eqs. (1), (5), (6) nally applied magnetic field Hext . It can be noticed that the mechan-
and (9) is performed in employing the Galerkin method of ical (elastic) problem is solved only in the composite material
weighted residuals method [21]. whereas the magnetic and electric problems include the surround-
After finite element discretization, the coupling system equa- ing air domain. The magnetostatic boundary conditions involve the
tion is: external excitation In ¼ ao that equals to ao ¼ lo Hext ðymax  ymin Þ.
n o n o Electrodes on the top and bottom of the piezoelectric layer has
€ þ ½C X_ þ ½KfX g ¼ ½F 
½M X ð12Þ been introduced in the formulation in using a charge conversion
method in which the node number is modified to consider the
with top and bottom of the piezoelectric layer as explicit equipotential
2 3 2 3 2 3 connection.
K uu K up K ua M 0 0 C uu 0 0 The coupling system Eq. (12) is valid for static field analysis (by
6 7 6 7 6 7 neglecting the time dependent terms) and dynamic field analysis.
6 7 6 7 6 7
½K ¼ 6 K pu K pp 0 7; ½M ¼ 6 0 0 0 7; ½C ¼ 6 0 0 0 7; In this study, both analyses are successively solved in considering
4 5 4 5 4 5
K au 0 K aa 0 0 0 0 0 0 in first the nonlinear behaviors of v and h for static field and in sec-
ond a linear behavior (by incremental resolution) for small signal
2 3 2 3
u F dynamic analysis. In all case, the length of the trilayer composite
6 7 6 7 Terfernol-D/PZT-5A/Terfernol-D is fixed at Lx ¼ 14 mm.
6 7 6 7
fX g ¼ 6 V 7; ½F  ¼ 6 Q n 7; K pu ¼ K tup ; K a u ¼ K tua ; C uu ¼ bK uu þ aM
4 5 4 5
2.2. Nonlinear static FEM problem
a In
In this section the piezoelectric layer thickness tp is fixed at
where ½K is the mechanical stiffness matrix, ½C the damping
1mm and the magnetostrictive layer thicknesses change from
matrix, ½M the electro-magneto-mechanical mass matrix, fX g the
Ly ¼ 2 mm to Ly ¼ 5 mm. The nonlinear behaviors of both magnetic
unknown vector and fF g the excitation vector in which F; Q n ; In
and magnetostrictive properties are considered in FEM code. The
are the external excitations. In the expression of C uu ; b and a are
magnetic B–H curve of the magnetostrictive layer is presented in
Rayleighs damping coefficients respectively set to 108 and 0 in
Fig. 3.
our simulation [24]. In this study, no external body force and elec-
This curve has been reconstructed from a measurement curve of
tric charge are considered, i.e. F=0 and Q n =0.
the Terfernol-D given by [29] in using the following well-known
The unknown variables of the problem are the nodal displace-
Brauer model.
ment fug, electrical potential fVg, and magnetic vector potential
2
fag. H ¼ ðk1 eðk2 B Þ þ k3 ÞB ð17Þ
The mechanical strain, the electric field and the magnetic
induction are related to the displacement, electric potential and The nonlinearity behavior is then represented in the nonlinear
magnetic vector potential, respectively as: reductivity:
2
^ ½Nu fug ¼ ½Gu fug
S¼r mðB2 Þ ¼ ðk1 eðk2 B Þ þ k3 Þ ð18Þ
E ¼ grad½Nv fVg ¼ ½Gv fVg ð13Þ
B ¼ r  grad½Nv afag ¼ r  ½Ga fag

N u ; N v and Na are the shape functions associated with each node

which satisfy the properties:

X
m
Ni ¼ 1; Nj ðfi Þ ¼ dij ð14Þ
i¼1

m is the number of nodes by mesh element, fi the natural coordi-

nates and dij the Kronecker delta. In our study, linear triangular
element is used for the discretization of all fields. That implies
N u ; N v and N a ¼ N. In this condition, the submatrices in (1) can
be uniformly presented by Fig. 2. Magnetic, electric and elastic boundary conditions.
68 H. Talleb, Z. Ren / Journal of Alloys and Compounds 615 (2014) 65–74

Fig. 3. B–H curve measurement of the Terfenol-D.

k1 ¼ 100, k2 ¼ 14:926 et k3 ¼ 50.

The magnetostriction stress TB is expressed as a parabolic func-
tion of the magnetic induction [21,22,24]:
2 B2
3
B21  22
6 7
TB ¼ hB ¼ 2l bo 6 B2 7
4 B22  21 5 ð19Þ
3
B B
2 1 2

where l ¼ ð1þE tÞ is the Lamé coefficient and bo a magnetostrictive

parameter of the Terfernol-D equals to 2:2104 [21,25].
By identification with (10a), the piezomagnetic coefficients are
expressed by:
" #
2l bo 2 B22
q33 ¼ B  ð20aÞ
m33 B1 1 2
" #
2l bo 2 B21
q31 ¼ B  ð20bÞ
m33 B1 2 2
3l bo
q15 ¼ ½B B  ð20cÞ
m11 B1 1 2
The nonlinear problem is solved by the Newton–Raphson
method that is a robust implicit method with a good convergence
speed. The nonlinear variable is iteratively decomposed in piece- Fig. 4. Piezomagnetic coefficients in function of Hext (Oe), tp fixed at 1 mm.

wise-linear solution for unknown increment X i :


fX i g ¼ fX i1 g þ fDX i g
ð21Þ
fDX i g ¼ K1T RðX i1 Þ

where KT is the tangent stiffness matrix and RðX Þ the residual.

Considering that the excitation vector fF g is independent to X ,
@F
i.e. @X ¼ 0, the tangent stiffness matrix KT is given by:
KT ¼ ½KðX Þ þ ½@ x KðX ÞfX g ð22Þ
The KT components to be calculated are:
8 XR t
> Xe ½Gu  ½@ B TB ½Ga dX
< ½K uaT  ¼
>
e
XR ð23Þ
> t
: ½K aaT  ¼
> Xe ½Ga  ½@ B H½Ga dX
e

The derivative matrix @ B H and @ B TB are computed in respect to

the Brauer model H ¼ mðB2 ÞB and the magnetostriction stress TB
model, given respectively by (17) and (15).
Fig. 4 shows the piezomagnetic coefficients d ¼ qs (in nm/A,
s ¼ c1 ) in function of the external dc bias field Hdc in Oersted
(Oe) under different magnetostrictive thickness t m . We have a good
Fig. 5. Stress TB11 in function of Hext (Oe), tp fixed at 1 mm.
agreement quantitatively and qualitatively with the experimental
and analytical results available in different works [11–17]. Fig. 5
shows the magnetostrictive stress TB11 (in kPa) under the different
 H. Talleb, Z. Ren / Journal of Alloys and Compounds 615 (2014) 65–74 69

magnetostrictive thickness t m . The simulation results confirm that

when the magnetostrictive thickness decreases the magnetic
induction is more intense inside of these layers involving an
increase of the stress.

2.3. Harmonic FEM problem

In considering a small signal ac field Hac around a magnetization

point Hdc the system in harmonic regime with the electrodes
connected as an open circuit, can be rewritten as:

~ fX g ¼ ½F 
½K ð24Þ

with
2 3
e uu  x2 M
K e uu
e þ jx C e up
K e ua
K
6 7
~ 6
½K ¼ 4 e
K pu e pp
K 0 7 5
e au
K 0 e aa
K Fig. 7. a
~ V in function of frequency with Epoxy resin.

where x ¼ 2pf is the angular frequency.

In this section the thicknesses of the trilayer composite are set
as t m ¼ tp ¼ 1 mm. For dynamic field of small amplitude, the sys-
tem can be solved linearly around an external dc bias point. The
magnetostrictive coefficients as well as the reluctivity in Eq. (24)
are incremental values and determined according to the operation
point obtained in the static FEM analysis of the previously section.
Fig. 6 shows the obtained frequency response of the voltage coeffi-
cient a~ V defined in harmonic case in our model as j dV o =dHac j
under dc bias magnetic field Hdc excitation of 250 Oe with an ac
magnetic Hac ¼ 1 Oe.
The frequency resonance of 84.5 kHz is slightly greater than the
experiment value of 83 kHz measured by Dong et al. [8] for a sim-
ilar trilayer Terfernol-D/PZT-5A/Terfernol-D with dimensions
Lx ¼ 14 mm  Ly ¼ 3 mm  Lz ¼ 6 mm and polarized in L–T mode.
Actually, the magnetostrictive layers and piezoelectric layers are
generally bodily fixed with a resin that involves a decrease of the
(a) at 1kHz
mechanical resonance frequency. This thin layer resin can diffi-
cultly be modeled by macroscopic models such as equivalent cir-
cuits. With the coupled FEM developed in this work, the Epoxy
resin, which the material properties are given in appendix, can
be easily taken into account. In the presence of thin layer resin,
the FEM results show a decrease of mechanical resonance fre-
quency as observed in Fig. 7. In this simulation, a Epoxy resin under
a thickness varying from 0.1 mm to 0.4 mm is introduced. The
mechanical resonance frequency decreases as the resin layer

(b) at resonance

Fig. 8. Intensity of the magnetic vector potential.

increases. This same phenomenon is observed when the Young’s

module of the resin decreases.
Fig. 8a and b shows that unlike in low-frequency (1 kHz) the
total vector potential distribution at resonance frequency is stron-
ger inside of both magnetostrictive layers with an opposition
switch due to the change of phase illustrated by the real part of
a~ V . Thus, vector potential induced in each magnetostrictive layer
by the stress is higher than the applied vector potential. Figs. 9a
Fig. 6. a
~ V in function of frequency. and b and 10a and b show respectively the sign direction of
70 H. Talleb, Z. Ren / Journal of Alloys and Compounds 615 (2014) 65–74

(a) at 1kHz

(a) at 1kHz

(b) at resonance

Fig. 9. Displacement orientations (increased thousand times) inside the composite.

deformations (increased thousand times) and the stress values of

T11 and T22 in Pa unit, both cases inside the trilayer laminate com-
posite. The deformations are symmetrical but in low-frequency
there is a contraction phenomenon, and at resonance frequency,
due to switch orientation, there is a dilation phenomenon with a
stronger compression (10 Pa) of the piezoelectric. The change of
orientation and the stronger compression are confirmed in
Fig. 11a and b by the real part of the induced electric potential V
distributions symmetrically concentrated around both electrodes (b) at resonance
which the voltage U near resonance is hundred times higher than
in low-frequency. Fig. 10. Stress distribution field inside the composite.
In low-frequency with small applied Hac , the voltage coefficient
a~ V is usually studied with the following Dongs model [8]:
2 knee point of B–H curve) when the volume fraction n is equals to
nð1  nÞtlam d33H d31E
ja
~ V jL ¼ ð25aÞ 0.5 i.e. when tp ¼ 2t m . For dc bias magnetic field Hdc excitations
T
e
2 E E
33 s11 ðns11 þ ð1  nÞsH33 Þ
smaller than 200 Oe the optimal value of the volume ratio is about
where n ¼ 2t2t m
represents the volume fraction for a trilayer lami- n = 0.25 (i.e. tp ¼ 3t m ) whereas for dc bias magnetic field Hdc
m þt p
nate composite, t lam ¼ 2t m þ tp ; eS33 the permittivity under constant excitations greater than 200 Oe the optimal value of the volume
strain, sE11 and sH33 are elastic constants, and d31E a piezoelectric coef- fraction reaches n = 0.67 (i.e. t p ¼ tm ) for 400 Oe and n = 0.75 (i.e.
ficient. Actually, the formulation (25-a) must be rewritten as (25b) tp ¼ 2tm =3) for 600 Oe.
in order to introduce the dependence of the piezomagnetic coeffi-
cient d33H ðHext ; t m Þ to the external dc bias field Hdc and the magneto- 2.4. Effect of the electrical impedance load
strictive thickness t m and to introduce the elastic constant
sH33 ðd33H ; Hext ; tm Þ according to the relation sH33 ðd33H ; Hext ; t m Þ ¼ Since the piezoelectric layer is dielectric, there is no free charge
2
sB33 ðd33H Þ þ mT33 d33H ðHext ; tm Þ inside it. All the free charges are confined to the electrodes. The
current I crossings toward the electrical impedance load is then
2
nð1  nÞt lam d33H ðHext ; t m Þd31E the time derivative of the total electric charge Q ; I ¼ @ t Q .
ja
~ V jL ¼ ð25bÞ
e
2 E E
þ ð1  nÞsB33 ðd33H ; Hext ; tm ÞÞ In harmonic regime, the current can be written as:
33 s11 ðns11
T

Fig. 12a and b compare the voltage coefficient obtained by the I ¼ jxQ ð26Þ
analytical formulation (25-b) and by the simulation in employing According to Fig. 13, the terminal voltage of the electrical
the piezomagnetic coefficients illustrated in Fig. 3a. As previously impedance load Z is given by ZI and is related to the voltage U
an ac magnetic Hac ¼ 1 Oe is employed for the simulation. The total between both electrodes by:
thickness t lam has been fixed at 4mm while employing a simulta-
neous variation of t m and tp f such that volume fraction n varies U ¼ ZI ¼ jxZQ ð27Þ
from 0.25 to 0.9. It can be seen that the maximal value of the volt- We can introduce an incident matrix K pq that links the electrical
age coefficient a ~ V first increases and then decreases with the potential and the electrodes of the piezoelectric layer. The electric
increasing dc bias magnetic field Hdc excitation volume fraction. charge Q then satisfies the following condition:
The voltage coefficient a ~ V reaches a maximal value under a dc bias
magnetic field Hdc excitation of 200 Oe (which corresponds to the VK pq  jxZQ ¼ 0 ð28Þ
 H. Talleb, Z. Ren / Journal of Alloys and Compounds 615 (2014) 65–74 71

with
2 3
e uu  x2 M
K e uu
e þ jx C e up
K 0 e ua
K
6 7
6 e pu
K e pp
K e pq
K 0 7
~Z ¼ 6
½K 6
7
7
6 0 e qp
K jxZ 0 7
4 5
e au
K 0 0 e aa
K

with,
2 3 2 3
u 0
6V 7 60 7
6 7 6 7
fX g ¼ 6 7; ½F  ¼ 6 7
4Q 5 40 5
(a) at 1kHz
a In

The electric charge Q is an unknown scalar.

Fig. 14 shows the frequency dependence of the voltage coeffi-
cient a~ V under dc magnetic excitation Hdc of 250 Oe with an ac
magnetic field Hac ¼ 1 Oe and under various resistive electrical
load (the 1 MOhm can be considered as the open-circuit condi-
tion). The results that show deterioration under 30 kOhm are in
good concordance with the analytical and experimental results in
[27–29].

2.5. Performances

This section is devoted to demonstrate the performance of an

(b) at resonance energy transducer employing a trilayer laminate composite Terfer-
nol-D/PZT/Terfernol-D under different external dc fields Hdc with
Fig. 11. Real part of the electrical potential distribution.
an ac field Hac ¼ 1 Oe and under different electrical loads. Figs. 15
and 16 show, respectively, the voltage coefficient a
~ V at 1 kHz under
different magnetostrictive thickness tm while maintaining
tp ¼ 1 mm and Lx ¼ 14 mm and the stress (in kPa) inside the com-
posite at 1 kHz and at the resonance. In the case when
tm ¼ t p ¼ 1 mm (so that Ly ¼ 3 mm), the optimal external dc field
Hdc value is approximately included between 150 Oe and 200 Oe.
In this configuration, Fig. 17 shows the frequency resonance
dependence of the voltage coefficient a ~ V change in function of
external dc fields Hdc . The maximal value is observed for a

(a) from the formulation 25b

Fig. 13. Trilayer loaded by an electrical impedance Z.

(b) from FEM simulation

Fig. 12. Voltage coefficient a

~ V in function of the volume fraction n.

where the elements of K pq are respectively equal to 1 or 1 if a node

is associated of the top electrode or the bottom electrode, otherwise
its elements equal to 0.
Finally, the system can be rewritten as [26]:
~ Z fX g ¼ ½F 
½K ð29Þ
Fig. 14. a
~ V in function of frequency under different load resistances.
72 H. Talleb, Z. Ren / Journal of Alloys and Compounds 615 (2014) 65–74
Fig. 15. a
~ V in function of frequency under different external dc fields Hdc .
Fig. 16. Stress at 1 kHz and at resonance.
Fig. 17. a
~ V in function of frequency under different external dc fields Hdc .
frequency resonance of 84.5 kHz under 200 Oe. Fig. 18a and b
shows at this resonance frequency the voltage coefficient a ~ V and
output deliverable power (P ¼ 12 VI ) as function of the load resis-
tance under different external dc fields Hdc . Finally for external
dc field Hdc of 200 Oe with an ac field Hac ¼ 1 Oe and
tm ¼ t p ¼ 1 mm, a deliverable maximal power of 1.1 mW is
obtained under an electrical load of 10 kOhm, that is quantitatively
and qualitatively in good concordance with the measured results
presented by Chen et al. [17] and Wang et al. [29] and more partic-
ularly by Xianzhi et al. [30] since they used a similar trilayers lam-
inate composite Terfernol-D/PZT-5A/Terfernol-D. Knowing the
Fig. 18. (a) Voltage coefficient a
~ V and (b) output deliverable power as function of
the load resistance under different external dc fields Hdc at 84.5 kHz.
optimal load, the harvesting energy can be enhanced in employing
a damping conditioning circuit such as the Synchronized Switch
Harvesting on Inductor (SSHI) nonlinear technique [31,32].
3. Conclusion
Based on the finite element method a multiphysics modeling of
magnetoelectric energy transducer has been presented. The model
has been applied to a trilayer laminate composite Terfernol-D/PZT-
5A/Terfernol-D in which the magnetic and piezomagnetic nonlin-
ear effects and the electrical load effect have been included. The
simulation results in static and dynamic cases have been validated
by comparison with different experimental and analytical results.
The present study has put in evidence the impact of the Epoxy
resin on the resonance frequency thanks to the introduction of a
thin layer in the finite element model. The mechanical-magneto-
electric field distribution results showed that the piezoelectric
polarization endures a change of sign at the resonance frequency.
The impact of volume fraction n on the voltage coefficient a~ V under
different dc bias magnetic field Hdc excitations has revealed that
the ideal configuration is reached when tp ¼ 2t m under a dc bias
magnetic field Hd c excitation around 200 Oe. The insertion of the
load effect in the trilayer composite with the dimensions
Lx ¼ 14 mm  Ly ¼ 3 mm has shown that the maximal output
deliverable power is obtained under an electrical load of 10 k Ohm
with a dc bias magnetic field Hdc excitation of 200 Oe. In conclu-
sion, the rigorous FEM modeling presented in this study provides
a useful multiphysics analysis tool that contributes to the model-
ing and the design optimization of multilayer ME transducer for
energy harvesting application with the ability of accounting for
 H. Talleb, Z. Ren / Journal of Alloys and Compounds 615 (2014) 65–74 73

complex geometries, non-linear material properties with different Reluctivity

polarization and magnetization modes, and loading configurations. 2 3
m11 0 0
1
Appendix A m¼ 6 40 m22 7
5
m0
0 0 m33
A.1. A
Density qH ¼ 9250 kg m3
Material properties in local coordinates for the PZT-5A.
Elastic constants under electric field constant in 109 N m2 A.3. C
2 3
cE11 cE12 cE13 0 0 0 Epoxy resin.
6 E 7
6 c21 cE22 cE23 0 0 0 7 Young’s module ðEÞ ¼ 4:25  109 ; N m2 .
6 7
6 E 7 Density q ¼ 1160 kg m3 .
6 c31 cE32 cE33 0 0 0 7
C ¼6
E
60
7 Poisson’s ratio ðtÞ ¼ 0:38.
6 0 0 cE44 0 0 7 7
6 7
60 0 0 0 cE55 0 7
4 5 References
0 0 0 0 0 cE66
[1] Marian Vopsaroiu, John Blackburn, Markys G. Cain, A new magnetic technology
based on the magneto-electric effect, J. Phys. D: Appl. Phys. 40 (2007) 5027–
cE11 ¼ cE22 ¼ 138:5; cE33 ¼ 114:75; cE12 ¼ cE21 ¼ 77:37; cE13 ¼ cE31 ¼ cE23 5033.
[2] Shuxiang Dong, Junyi Zhai, J.F. Li, D. Viehland, S. Priya, Multimodal system for
¼ cE32 ¼ 73:64 harvesting magnetic and mechanical energy, Appl. Phys. Lett. 93 (2008)
103511.
[3] Robert Jahns, Henry Greve, Eric Woltermann, Eckhard Quandt, Reinhard
cE33 ¼ 114:75; cE44 ¼ cE55 ¼ 25:6; cE66 ¼ 30:6 Knchel, Sensitivity enhancement of magnetoelectric sensor through
frequency-conversion, Sensor Actuat. A: Phys., A 183 (2012) 16–21.
[4] G. Harshe, J.O. Dougherty, R.E. Newnham, Int. J. Appl. Electrom. Mater. 4 (1993)
Piezoelectric coefficients in C m2 145.
2 3 [5] G. Srinivasan, E.T. Rasmusen, J. Gallegos, R. Srinivasan, Magnetoelectric bilayer
0 0 0 0 e15 0
and multilayer structures of magnetostrictive and piezoelectric oxides, Phys.
6 7
e ¼ 40 0 0 e24 0 05 Rev. B 64 (2001) 214408.
[6] M.I. Bichurin, V.M. Petrov, S.V. Averkin, A.V. Filippov, Electromechanical
e31 e32 e33 0 0 0 resonance in magnetoelectric layered structures, Phys. Solid State 52 (10)
(2012) 2116–2122.
e31 ¼ e32 ¼ 5:2; e33 ¼ 15:08; e24 ¼ e15 ¼ 12:72 [7] M.I. Bichurin, V.M. Petrov, Theory of low frequency magnetoelectric coupling
in magnetostrictive piezoelectric bilayers, Phys. Rev. B 68 (5) (2003) 054402.
Permittivity [8] ShuXiang Dong, Zhai JunYi, Equivalent circuit method for static and dynamic
2 3 analysis of magnetoelectric laminated composites, Chin. Sci. Bull. 53 (14)
e11 0 0 (2008) 2113–2123.
1 [9] ShuXiang Dong, Jeifang Li, Dwight Viehland, Ultrahigh magnetic filed
e¼ 6 40 e22 7
5 sensitivity in aminates of TERFENOL-D and Pb(Mg1/3Nb2/3)O3–PbTiO3
e0 crystals, Appl. Phys. Lett. 83 (11) (2003).
0 0 e33
[10] ShuXiang Dong, Jeifang Li, Dwight Viehland Yang, Longitudinal and transverse
magnetoelectric voltage coefficients of magnetostrictive/piezoelectric
e11 ¼ e22 ¼ 1730; e33 ¼ 1700 laminate composite: theory, IEEE Trans. Ultrason. Ferroelectr. Freq. Controle
50 (10) (2003).
Density qE ¼ 7600 kg m3 [11] F. Yang, Y.M. Wen, P. Li, M. Zheng, L.X. Bian, Resonant magnetoelectric
response of magnetostrictive/piezoelectric laminate composite in
consideration of losses, Sensor. Actuator. A 141 (2008) 129–135.
A.2. B [12] ShuXiang Dong, Jeifang Li, Dwight Viehland, J. Cheng, L.E. Cross, A strong
magnetoelectric voltage gain effect in magnetostrictive–piezoelectric
Material properties in local coordinates for the Terfernol-D. composite, Appl. Phys. Lett. 85 (2004) 3534.
[13] Caijaing Lu, Ping Li, Yumei Wen, Aichao Yang, Wei He, Jitao Zhang, Jing Yang,
Elastic constants under magnetic field constant in 109 N m2 Jing Wen, Yong Zhu, Investigation of magnetostrictive/piezoelectric multilayer
2 3 composite with a giant zero-biased magnetoelectric effect, Appl. Phys. A 113
cH11 cH12 c 1H 0 0 0 (2013) 413–421.
3
6 H 7 [14] Zhiyi Wu, Yumei Wen, Ping Li, Jin Yang, Xianzhi Dai, Effect of permeability and
6 c2 cH22 cH23 0 0 0 7
6 1 7 piezomagnetic coefficient on magnetostrictive/piezoelectric laminate
6 H 7
6 c31 cH32 cH33 0 0 0 7 composite, J. Magnet. 16 (2) (2011) 157–160.
C ¼6
H
60
7 [15] Hao-Miao Zhou, Xiao-Wei Ou, Ying Xiao, Shao-Xing Qu, Hua-Ping Wu, An
6 0 0 cH44 0 0 7 7 analytical nonlinear magnetoelectric coupling model of laminated composites
6 7 under combined pre-stress and magnetic bias loadings, Smart Matter. Struct.
60 0 0 0 cH55 0 7
4 5 22 (2013) 035018.
0 0 0 0 0 cH66 [16] Lei Chen, Ping Li, Yumei Wen, Dynamic magnetostriction characteristics of an
Fe-based nanocrystalline FeCuNbSiB alloy, J. Magnet. 16 (3) (2011) 211–215.
[17] Lei Chen, Ping Li, Yumei Wen, Dong Wang, Magnetoelectric transducer
cH11 ¼ cH22 ¼ 35:87; cH33 ¼ 46:11; cH12 ¼ cH21 ¼ 17:7; cH13 ¼ cH31 employing piezoelectric ceramic/ferromagnetic alloy/high-permeability
FeCuNbSiB composite, IEEE Trans. Magn. 47 (10) (2011).
¼ cH23 cH32 ¼ 23:25 [18] M. Besbes, Z. Ren, A. Razek, A generalized finite element model of
magnetostriction phenomena, IEEE Trans. Magn. 37 (5) (2001).
[19] Thu Trang Nguyen, X. Mininger, L. Daniel, F. Bouillault, A Influence of
cH33 ¼ 46:11; cH44 ¼ cH55 ¼ 4:16; cH66 ¼ 9:09 mechanical boundary conditions of magnetoelectric sensors, IEEE Trans.
Magn. 49 (5) (2013).
Piezomagnetic coefficients in C m2 [20] Thu Trang Nguyen, Xavier Mininger, Frédéric Bouillault, Laurent Daniel, Finite
element harmonic modeling of magnetoelectric effect, IEEE Trans. Magn. 47
2 3 (5) (2011).
0 0 0 0 q15 0 [21] Thu Trang Nguyen, Frédéric Bouillault, Laurent Daniel, Xavier Mininger, Finite
6 7 element modeling of magnetic field sensors based on nonlinear
q ¼ 40 0 0 q24 0 05 magnetoelectric effect, J. Appl. Phys. 109 (2011) 084904.
q31 q32 q33 0 0 0 [22] N. Galopin, X. Mininger, F. Bouillault, L. Daniel, Finite element modelling of
magneto electric sensors, IEEE Trans. Magn. 44 (6) (2008) 834–837.
74 H. Talleb, Z. Ren / Journal of Alloys and Compounds 615 (2014) 65–74
[23] X. Mininger, N. Galopin, Y. Dennemont, F. Bouillault, 3D finite element model
for magnetoelectric sensors, Eur. Phys. J. – Appl. Phys. 52 (2010).
[24] Gilder Nader, Emilio C.N. Silva, Julio C. Adamowski, Determination of
piezoelectric transducer damping by using experimental and finite element
simulations, in: Proc. SPIE 5052, Smart Structures and Materials 2003:
Damping and Isolation, vol. 116, July 31, 2003.
[25] Karim Azoum, Contribution à la modélisation numérique de phénomènes
magnéto-élastiques: étude de dispositifsà base de matériaux magnétostrictifs,
Thèse de doctorat, Université Paris 11, Orsay.
[26] Hakeim Talleb, Zhuoxiang Ren, Finite element modeling of a magnetoelectric
energy transducer including the load effect, in: IEEE Conference on
Electromagnetic Computation (CEFC), Annecy, France, 25–28 May, 2014.
[27] C.L. Zhang, J.S. Yang, W.Q. Chen, Magnetoelectric effects in multiferroic
bilayers for coupled flexure and extension, J. Intell. Mater. Syst. Struct. 21
(2010) 851.
[28] Leixiang Bian, Yumei Wen, Ping Li, Qiuling Gao, Yong Zhu, Yu. Mioa,
Magnetoelectric performances in composite of piezoelectric ceramic and
ferromagnetic constant-elasticity alloy, IEEE Sensors J. 9 (12) (2009).
[29] Yaojin Wang, Xiangyong Zhao, Jie Jiao, Linhua Liu, Wenning Di, Haosu Luo, Siu Wing,
Electrical resistance load effect on magnetoelectric coupling of magnetostrictive/
piezoelectric laminated composite, J. Alloys Comp. 500 (2010) 224–226.
[30] Xiangzhi Dai, Yumei Wen, Ping Li, Jin Yang, Gaoyong Zhang, Modeling,
characterization and fabrication of vibration energy harvester using Terfenol-
D/PZT/Terfenol-D composite transducer, Sensor. Actuator. A: Phys. 156 (2009)
350–358.
[31] Shashank Priya, Daniel J. Inman, Energy Harvesting Technologies, Springer
Edition, ISBN 978-0-387-76464-1.
[32] J. Dicken, P.D. Mitcheson, I. Stoianov, E.M. Yeatman, Power-extraction circuits
for piezoelectric energy harvesters in miniature and low-power applications,
IEEE Trans. Power. Electr. 11 (2012) 4514–4529.