Composite Structures 151 (2016) 70–80

Contents lists available at ScienceDirect

Composite Structures
journal homepage: www.elsevier.com/locate/compstruct

Different interface models for calculating the effective properties in

piezoelectric composite materials with imperfect fiber–matrix adhesion
Humberto Brito-Santana a,⇑, Ricardo de Medeiros a, Reinaldo Rodriguez-Ramos b, Volnei Tita a
a
Department of Aeronautical Engineering, São Carlos School of Engineering, University of São Paulo, Av. João Dagnone 1100, São Carlos, SP, Brazil
b
Departamento de Matemáticas, Facultad de Matemática y Computación, Universidad de La Habana, San Lázaro y L, CP 10400, Vedado, Habana 4, Cuba

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

Article history: Piezoelectric materials are able to produce an electrical response when mechanically stressed (sensors)
Available online 11 February 2016 and inversely high precision motion can be obtained with the application of an electrical field (actuators).
The macroscopic properties of piezoelectric composites depend upon the properties and the interfacial
Keywords: bonding conditions of the constituent phases, and the microstructures of the composites. In the present
Piezoelectric composites work, a new imperfect interface model for a thin elastic interface is derived. Square unit cell model was
Effective properties used to calculate all coefficients of the material tensor. The calculation was performed via FE package
Imperfect interface
ABAQUSTM. A computational procedure, based on Python language, was developed to systematically cal-
Interface model
Finite Element Method
culate all RVE effective coefficients. Comparisons to classical Hashin’s and Nairn’s interface model show
very accurate agreement for debonding and perfect boding interface.
Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction bonds in a large variety of structures. In all cases, a basic goal of

Piezoelectric materials are able to produce an electrical
response when mechanically stressed (sensors) and inversely high
precision motion can be obtained with the application of an electri-
cal field (actuators). Sensitivity analysis [1] and optimization tech-
niques [2–4] are applied to maximize the piezoelectric actuator
efficiency.
The macroscopic properties of piezoelectric composites depend
upon the properties and the interfacial bonding conditions of the
constituent phases, and the microstructures of the composites.
Thus the effect of the interfacial bonding conditions on the
mechanical and physical properties of various composites has
attracted a lot of attention of researchers in many fields, especially,
in physics, materials science and technology, and mechanics. The
prediction of the effective moduli taking into account interface
effect is one of the fundamental problems in mechanics of compos-
ites [5–10].
Modeling interface, which are often finite-thickness inter-
phases, in composite materials is difficult. Interfaces play an
important role in determining the performance of structural mate-
rials on a wide variety of dimension scales, from grain boundaries
in metals, to inter-laminar bonds in composites and adhesive
⇑ Corresponding author. Tel.: +55 16 3373 8612; fax: +55 16 3373 9590.
E-mail address: hbritosantana@gmail.com (H. Brito-Santana).
nondestructive evaluation is the determination of the integrity of
bonds. In fact, several properties of materials, such as, just to make
a couple of examples, the mechanical behavior under stress [11] or
the ultrasonic reflection coefficient of the interface [12], are very
sensitive to boundary imperfections [13].
Regarding analytical models, different approaches have been
proposed. Broutman and Agarwal [14], Theocaris et al. [15] and
Sideridis [16] have considered the interphase as a layer between
fiber (on inclusion) and matrix, of specified thickness and of elastic
constants different from those of the matrix and the fiber. For an
alternate model, a very thin interfacial zone of unspecified thick-
ness has been considered. In that model, it is assumed that the
radial and the tangential tractions are continuous across the inter-
phase, but the displacements may be discontinuous from fiber to
matrix due to the presence of the interphase. The tractions are
assumed to be proportional to the corresponding displacement dis-
continuities. The proportionality constants then characterized the
stiffness of the interphase, which is represented by spring layer
model. Lene and Leguillar [17], Benveniste [18], Aboudi [19], Steif
and Hoysan [20], Achenbach and Zhu [21], and Hashin [22–24]
are some works related to spring layer model.
One way to model interphases is to abandon attempts for expli-
cit modeling and instead replace 3D interphases with 2D interfaces
[22]. The interphase effects are reduced for modeling the response
of 2D interfaces due to tractions normal and tangential to the

http://dx.doi.org/10.1016/j.compstruct.2016.02.003
0263-8223/Ó 2016 Elsevier Ltd. All rights reserved.
 H. Brito-Santana et al. / Composite Structures 151 (2016) 70–80 71

8 9 2 3
interfacial surface, which can be modeled by interface traction
> r11 > C E11 C E12 C E13 0 0 0 0 0 e13 8 > e11 >
9
>
> >
> 6 E 7>> >
>
laws. Elimination of 3D interphases removes the resolution prob- >
> r22 >> 6 C 13 C E11 C E13 0 0 0 0 0 e13 7>> e22 >>
>
> >
> 6 7>> >
>
lem. The use of interface traction laws replaces numerous >
> >
> 6 E >
7> e >
>
>
> r >
>
> 6 C 13 C E13 C E33 0 0 0 0 0 e33 7>>
>
>
>
unknown and potentially unmeasurable interphase properties 7> >
33 33
>
> >
> 6 >
> >
>
> r >
> 6 0 0 7 > e12 >>
with a much smaller number of interface parameters. If interface < 12 = 6 0 0 C E66 0 0 0 0 >
7< > =
traction laws can be determined, one can potentially model inter- r23 ¼ 66 0 0 0 0 C E44 0 0 e15
7
e
0 7> 23 >;
7
>
> >
> 6
> r > 6 7>> >
>
> e31 >
phase effects well. That approach for interphase modeling was >
> 31 >
> > >
developed for analytical modeling of interface effects in composite >
> > 6
> 6 0 0 0 0 0 C E44 e15 0 0 7 7>> >
>
>
> D1 > > 6 7 >
> E >
>
materials [25–28], and for wave transmission in damage planes >
>
>
>
> 6 0
> 0 0 0 0 e15 j S
0 0 7> >
>
1 >
>
>
>
> D2 > > 6 11
7>> E >
>
> > jS11 > >
[29]. > ; 4 0
> 0 > 5 : >
0 0 0 e15 0 0 2
: ;
In the present work, a new imperfect interface model for a thin D3 e e13 e33 0 0 0 0 0 jS E 3
13 33
elastic interface is derived. A three dimensional (3D) representa-
ð2Þ
tive volume element (RVE) model was developed for analyzing
effective properties of piezoelectric fiber embedded in a non- where the contracted Voigt notation is used. In Eq. (2), the 3-axis is
piezoelectric matrix composite with imperfect interface using suit- aligned with the principle direction of polarization.
able interface models (including a new one), which were compared In classical lamination theory, the composite lamina is modeled
in terms of their capability to determine the effective coefficients. as a homogeneous orthotropic medium with certain effective mod-
Square unit cell model was used to calculate all coefficients of the uli that describe the ‘average’ material properties of the composite.
material tensor. The calculation was performed via FE package For micromechanical analysis, composites can be studied using
ABAQUSTM. A computational procedure, based on Python language, representative volume element (RVE) or unit cell. The RVE is the
was developed to systematically calculate all RVE effective coeffi- smallest portion of the actual composite, which has same elastic,
cients. It is important to highlight that Tita et al. [30] presented dielectric and piezoelectric constants and fiber volume fraction of
the numerical approach to evaluate the effective properties of dif- the investigated material. It is a microstructural model of a mate-
ferent volume fractions for piezoelectric fibers (with circular and rial, which can be used to obtain the response of the corresponding
square cross section) embedded in a non-piezoelectric matrix homogenized macroscopic continuum in a macroscopic material
using only a Hashin’s modified interface model. However, in the point. Thus, the proper choice of the RVE determines largely the
present work, the proposed interface model is more generalized accuracy of the modelling of a heterogeneous material. In present
than Hanshin’s and Nairn’s interface model, and it was used for work, the RVE are assumed as combinations of piezoelectric fibers
determining the effective properties of different volume fractions embedded in a polymer matrix, including an interface, obeying a
for piezoelectric fibers (with only circular cross section) embedded specified fiber volume fraction. This representative volume ele-
in a non-piezoelectric matrix. Therefore, Nairn’s interface model ment (RVE) is modeled by solid finite elements. Thus, the numeri-
can be considered as a particular case of the proposed model. Com- cal model is used to determine a homogeneous medium equivalent
parisons to classical Hashin’s and Nairn’s interface model show to the original composite and, as commented earlier, comprises the
very accurate agreement for debonding and perfect boding smallest portion of the piezoelectric composite, which keeps the
interface. most representative combination of its main materials. Thus, it is
assumed that the average mechanical and electrical properties of
2. Constitutive equations, Finite Element Method and a unit cell are equal to the average properties of the composite
representative volume element (FEM–RVE) material as follow:
Z Z
1 1
Considering that, the piezoelectric materials respond linearly to r
ij ¼ hrij i ¼ rij dV;
eij ¼ heij i ¼ eij dV;
jVj jV jVj V
changes to mechanical and electrical fields. A three-phase rein- Z Z ð3Þ
forced piezoelectric composite is studied here, in which the fiber
i ¼ hDi i ¼ 1
D Di dV; E
i ¼ hEi i ¼ 1 Ei dV;
has homogeneous and transversely isotropic properties. In addi- jVj V jVj V
tion, the matrix and the interface have homogeneous and isotropic where jVj is the unit cell volume.
properties. Three standard notation systems are commonly used to Discretizing Eq. (3) using the Finite Element Method (FEM), the
describe the constitutive modeling of linear-piezoelectric materi- average values can be calculated by:
als. Using the conventional indicial notation in which repeated
subscripts are summed over the range of i, j, k, l = 1, 2, 3, the con- 1 X nel
1 X nel
r
ij ¼ rðnÞ V ðnÞ ;
eij ¼ eðnÞ V ðnÞ ;
stitutive equations can be written as follow: jVj n¼1 ij jVj n¼1 ij
ð4Þ
Xnel Xnel
rij ¼ C Eijkl ekl  ekij Ek;
i ¼ 1
D
ðnÞ
D V ðnÞ ;
i ¼ 1
E
ðnÞ
E V ðnÞ ;
ð1Þ jVj n¼1 i jVj n¼1 i
Di ¼ eikl ekl þ jSij Ej;
where nel is the number of finite elements of the complete unit cell,
where rij and eij are respectively the stress and infinitesimal strain V ðnÞ is the volume of the nth element, and rijðnÞ , eðnÞ ðnÞ ðnÞ
ij , Di and Ei are
tensors, Di and Ei are the electric displacement and electric field the respective tensors evaluated in the nth element.
vectors. The elastic tensor C Eijkl , the piezoelectric tensor eijk and the For a complete description of a differential problem in order to
dielectric tensor jSij possess the following symmetry properties determine effective material properties, it is necessary to formu-
C Eijkl
¼ C Eijlk
¼ C Ejikl ¼ C Eklij , eijk = eikj , jSij ¼ jSji . In addition, the superscript late appropriate boundary conditions. Since periodic structures
are investigated so called periodic boundary conditions are applied
E indicates constant electric field, while the superscript S indicates
to the considered RVE. Considering of the composite as a periodical
constant strain.
array of the RVEs, the periodic boundary conditions must be
For a transversally isotropic piezoelectric material, the constitu-
applied to the RVE models. This implies that each RVE in the com-
tive equation can be written in terms of the following expanded
posite has the same deformation mode and there is no separation
matrix form:
or overlap between the neighboring RVEs. For any parallelepiped
72 H. Brito-Santana et al. / Composite Structures 151 (2016) 70–80

RVE models Dxkj ¼ xþj j

i  xi is constant, therefore the following uni-
3.1. Hashin’s interface model
fied periodic boundary conditions is obtained:
The imperfect interface conditions, which are equivalent to the
effect of a thin elastic interphase, are derived by a Taylor expansion
uþj  uj ¼ cij ði; j ¼ 1; 2; 3Þ ð5Þ
i i method in terms of interface displacement and traction jumps [24].
In this case, K n and K t can be expressed by the interphase modulus
where uij denotes the displacement along the i-direction of one and interphase thickness ti as:
node located at the boundary face whose normal vector is along
ti ti
the j-direction. The index ‘‘+j” means along the positive X j direction Kn ¼ i
; Kt ¼ ð7Þ
and ‘‘j” means along the negative X j direction.
i
k þ 2G Gi
The average stretch or contraction of the RVE model due to the A physical argument has been given in [26,35], which implies
action of the three normal traction components are represented for that Eq. (7) should be valid for very flexible interphase of any
the constants c11 , c22 and c33 , whereas the other three pairs of con- shape. Thus, it is possible to obtain the coefficients for the interface
stants, c21 ¼ c12 , c31 ¼ c13 and c32 ¼ c23 , correspond to the shear defor- by using Eq. (7) as follow:
mations due to the three shear traction components.
C i11 ¼ C i22 ¼ C i33 ¼ ki þ 2li
Thus, this RVE (square unit cell) is analyzed by FEM for different
loadings with suitable boundary conditions applied in a unique C i12 ¼ C i13 ¼ C i23 ¼ ki
way, using ABAQUSTM combined to Python language. This proce-
C i11  C i12 ð8Þ
dure has been developed to systematically calculate all RVE effec- C i44 ¼ C i55 ¼ C i66 ¼ li ¼
2
tive coefficients, thereby reducing exhaust manual work, saving
ki ¼ t i ðK n  2K t Þ
time, and diminishing the chance of numerical errors. In addition,
as commented earlier, the FEM–RVE model shows the three phases li ¼ G i
(matrix, fiber and interphase), which are modeled by solid ele-
ments, with linear interpolation, i.e. eight-node brick element with
three degrees of freedom per node (C3D8E – ABAQUS nomencla- 3.2. Nairn’s interface model
ture). It is important to highlight that ABAQUSTM calculates the
quantities for the tensors in the Gauss points, and it uses numerical For a linear-elastic interphase of thickness ti, a simple series
techniques to integrate various quantities over the volume of each analysis for either tensile or shear loading [36] such that the total
element to feed the Eq. (4). In addition, the element type C3D8E deformation of the full model agrees with the total deformation of
has 8 integration points. the interface model leads to:
The simplified set of constitutive Eq. (2) with prescribed 2t i K n E1 E2
boundary conditions allows the evaluation of the effective material Ei ¼ ð9Þ
t i K n ðE1 þ E2 Þ þ 2E1 E2
properties. Rodríguez-Ramos et al. [10] and Tita et al. [30]
presented the numerical approach to evaluate the effective proper-
2t i K t l1 l2
ties of non-piezoelectric and piezoelectric fibers embedded in a li ¼ ð10Þ
non-piezoelectric matrix. t K t ðl1 þ l2 Þ þ 2l1 l2
i

3. Imperfect condition and interphase models
Thin interphase between two media is modeled in terms of
imperfect interface in the context of potential theory for mechan-
ical, piezoelectric, and dielectric behavior. In the following, it is
shown the formulas to calculate the Young’s modulus Ei, Lame
modulus ki and shear modulus Gi of the interface by using three
different interface degradation models (including a new one).
The interface models have been widely used to simulate the
interface bonding in particle-reinforced composites [26]. Apart
from being used to simulate the imperfect interfacial bonding or
debonding [31,32], those models can also be used to simulate a
thin and compliant interphase [26,33,34]. For an isotropic inter-
face, it suffices to resolve interface traction into normal and tan-
gential tractions (Tn and Tt), and it is assumed that they are
functions of normal and tangential displacements (½un  and ½ut )
in the interface. Thus, the interface conditions for the interface
model can be written as:
where E1, E2, l1 and l2 are the tensile and shear moduli of the two
substrates 1 (matrix) and 2 (fiber), respectively. Hence, by using the
relation ki ¼ li ðEi  2li Þ=ð3li  Ei Þ [37], it is possible to obtain the
coefficients for the interface.
3.3. Proposed interface model
The proposed model is all two-phase model in the sense that
the interface region occupies a zero volume fraction in the compos-
ite, whereas, the interface model is a three-phase one, consisting of
the inhomogeneity between the interphase and the matrix. Based
on this hypothesis, it was derived new expressions for the Young’s
modulus Ei and shear modulus li, as functions of volume fraction,
matrix and fiber properties and interface parameters K n and K t in
the form:
v i ti K n E1 E2
Ei ¼ ð11Þ
t K n ½ð1  c  v 2 ÞE1 þ ðc  v 1 ÞE2  þ v i E1 E2
i

v i t i K t l1 l2
T n ¼ K n ½un ; T t ¼ K t ½ut  on C ð6Þ li ¼ ð12Þ
t K t ½ð1  c  v 2 Þl1 þ ðc  v 1 Þl2  þ v i l1 l2
i

where the symbol ½g ¼ g 1  g 2 indicates the jump in the quantity g
in the common interface denoted by C between the fiber and
matrix. K n and K t are the interface parameters, which reduce all
properties of the 3D interphase to two interface stiffness properties.
As K n and K t approach zero, tractions approach zero and the inter-
face is debonded. As K n and K t approach infinity, displacement dis-
continuities approach zero and the interface is perfect.
where c is the matrix volume fraction in the imperfect interface
two-phase model and vi, v1 and v2 are the interface, matrix and fiber
volume fraction, respectively. For the interphase model, perfect
bonding conditions are assumed. Note that when v 1 ¼ c  v i =2
and v 2 ¼ 1  c  v i =2 the formulas (9)–(10) and (11)–(12) are the
same. The coefficients for the interface are calculated of the similar
manner to the Nairn’s interface model.
 H. Brito-Santana et al. / Composite Structures 151 (2016) 70–80 73

Table 1
Material properties of fiber (PZT-5) and matrix (polymer) [38].

C 11 C 12 C 13 C 33 C 44 C 66 e13 e15 e33 e11 e33

GPa 2 nF=m
C=m
PZT-5 121.0 75.4 75.2 111.0 21.1 22.8 5.4 12.3 15.8 8.11 7.35
Polymer 3.86 2.56 2.57 3.86 0.64 0.64 – – – 0.0797 0.0797

Table 2
Effective coefficients for limit cases with fiber volume fraction equal 0.5.

Kn Kt C11 (GPa) C12 (GPa)

Hashin [24] Nairn [35] Present model Hashin [24] Nairn [35] Present model
103 103 0.704599 0.705140 0.705140 0.185862 0.186279 0.186279
1012 1012 9.302235 9.300775 9.296201 4.329506 4.329105 4.328181
C13 (GPa) C33 (GPa)
Hashin [24] Nairn [35] Present model Hashin [24] Nairn [35] Present model
3 3
10 10 0.355908 0.356282 0.356282 27.795732 27.796023 27.796023
1012 1012 5.324132 5.322686 5.320549 31.651400 31.668833 31.666963
C44 (GPa) C66 (GPa)
Hashin [24] Nairn [35] Present model Hashin [24] Nairn [35] Present model
103 103 0.195315 0.195709 0.195709 0.068303 0.068813 0.068813
1012 1012 1.777684 1.777397 1.776567 1.355313 1.355127 1.354610
e13 (C/m2) e15 (C/m2)
Hashin [24] Nairn [35] Present model Hashin [24] Nairn [35] Present model
3 3
10 10 0.021503 0.014662 0.014662 0.003683 0.003682 0.003682
1012 1012 0.204663 0.204649 0.204557 0.015426 0.015425 0.015416
e33 (C/m2)
Hashin [24] Nairn [35] Present model
103 103 9.930898 9.930886 9.930886
1012 1012 9.771260 9.771310 9.771381
e11 (nF/m) e33 (nF/m)
Hashin [24] Nairn [35] Present Model Hashin [24] Nairn [35] Present Model
103 103 0.241006 0.241006 0.241006 3.848943 3.848943 3.848943
1012 1012 0.240693 0.240694 0.240694 3.842317 3.842316 3.842319

Table 3
Elastic effective coefficient C 44 and relative difference for K n ¼ 1012 and different values of the interface parameter K t and interphase thickness t i with fiber volume fraction equal
0.6.

ti Kt Models Difference
(a) Hashin (b) Nairn (c) Present Model ½jðcÞ-ðaÞj=ðcÞx100% ½jðcÞ-ðbÞj=ðcÞx100%
4 0 0.143749 0.506830 0.506794 71.6356 0.0070
10 10
102 0.143749 2.221911 2.220563 93.5264 0.0607
104 0.143754 2.337004 2.335494 93.8448 0.0647
106 0.144195 2.338237 2.336726 93.8292 0.0647
3 0 0.143295 0.506679 0.506321 71.6988 0.0707
10 10
102 0.143295 2.224133 2.210823 93.5185 0.6020
104 0.143299 2.339483 2.324501 93.8353 0.6445
106 0.143741 2.340721 2.325717 93.8195 0.6451

102 100 0.137969 0.504930 0.501355 72.4808 0.7131

102 0.137969 2.245478 2.118862 93.4885 5.9757
104 0.137973 2.363587 2.220960 93.7877 6.4219
106 0.138420 2.364853 2.222042 93.7706 6.4270

101 100 0.085154 0.488968 0.452859 81.1965 7.9735

102 0.085154 2.461381 1.521859 94.4046 61.7351
104 0.085159 2.604401 1.567825 94.5684 66.1156
106 0.085657 2.605932 1.568301 94.5382 66.1628
 74
Table 4
Effective coefficients and relative difference for K n ¼ 1012 and different values of the interface parameter K t with fiber volume fraction equal 0.5.

Models Kt C eff
11 ðGPaÞ C eff
12 ðGPaÞ C eff
13 ðGPaÞ C eff
33 ðGPaÞ C eff
44 ðGPaÞ C eff
66 ðGPaÞ eeff
13 ðC=m Þ
2
eeff
15 ðC=m Þ
2
eeff
33 ðC=m Þ
2
eeff
11 ðnF=mÞ eeff
33 ðnF=mÞ

(a) Hashin 100 9.302146 4.329266 5.322775 31.67227 0.203306 1.348679 0.204673 1.044575  1012 9.771301 0.240763 3.842314
102 9.302146 4.329266 5.322775 31.67227 0.203306 1.348679 0.204673 3.611454  1010 9.771301 0.240763 3.842314
104 9.302146 4.329266 5.322775 31.67227 0.203309 1.348694 0.204673 3.616896  108 9.771301 0.240763 3.842314

H. Brito-Santana et al. / Composite Structures 151 (2016) 70–80

106 9.302146 4.329266 5.322775 31.67227 0.203682 1.349855 0.204673 3.616063  106 9.771301 0.24075 3.842314

(b) Nairn 100 9.300300 4.329477 5.322647 31.668803 0.504886 1.347296 0.2046448 0.002905 9.771313 0.240750 3.842316
102 9.300313 4.329465 5.322647 31.668804 1.710705 1.352469 0.2046448 0.014757 9.771313 0.240697 3.842316
104 9.300609 4.329206 5.322661 31.668814 1.776695 1.354877 0.2046462 0.015418 9.771312 0.240694 3.842316
106 9.300771 4.329106 5.322685 31.668833 1.777390 1.355123 0.2046485 0.015425 9.771310 0.240694 3.842316

(c) Present model 100 9.294291 4.329672 5.320386 31.666836 0.504858 1.340290 0.204550 0.002904 9.771387 0.240750 3.842319
102 9.294464 4.329509 5.320390 31.666839 1.709932 1.348081 0.204550 0.014749 9.771387 0.240697 3.842319
104 9.295930 4.328322 5.320499 31.666923 1.775864 1.354315 0.204555 0.015409 9.771383 0.240694 3.842319
106 9.296197 4.328183 5.320548 31.666962 1.776560 1.354607 0.204557 0.015416 9.771381 0.240694 3.842319

Difference Kt DC eff
11 DC eff
12 DC eff
13 DC eff
33 DC eff
44 DC eff
66 Deeff
13 Deeff
15 Deeff
33 Deeff
11 Deeff
33

½jðbÞ  ðaÞj=ðbÞ  100%

100 0.0198 0.0049 0.0024 0.0109 59.7324 0.1027 0.0140 99.9999 0.0001 0.0056 3.6320  105
10 2 0.0197 0.0046 0.0024 0.0109 88.1157 0.2802 0.0140 99.9999 0.0001 0.0276 3.6285  105
104 0.0165 0.0014 0.0021 0.0109 88.5569 0.4564 0.0133 99.9998 0.0001 0.0287 3.3608  105
10 6 0.0148 0.0037 0.0017 0.0108 88.5404 0.3887 0.0122 99.9766 0.0001 0.0232 2.9064  105

½jðcÞ  ðaÞj=ðcÞ  100%

100 0.0845 0.0094 0.0449 0.0172 59.7302 0.6259 0.0606 99.9999 0.0009 0.0055 0.0001
102 0.0827 0.0056 0.0448 0.0171 88.1103 0.0444 0.0604 99.9999 0.0009 0.0275 0.0001
104 0.0669 0.0218 0.0428 0.0169 88.5515 0.4151 0.0579 99.9998 0.0008 0.0287 0.0001
106 0.0640 0.0250 0.0419 0.0168 88.5350 0.3508 0.0567 99.9765 0.0008 0.0232 0.0001
 H. Brito-Santana et al. / Composite Structures 151 (2016) 70–80
It is worth to highlight that the proposed model is more gener-
alized than Nairn’s interface model. Therefore, Nairn’s interface
model is a particular case of the proposed one.
4. Analysis and discussion of results
Before investigate the coefficients, a numerical study is first
performed to evaluate the sensitivity of the predicted coefficients
by using the RVE commented earlier, considering mesh refinement.
Thus, different mesh densities for the square unit cell were inves-
tigated. Three mesh densities for the unit cell were used namely a
coarse one with approximately 2000 elements, a medium one with
approximately 4000 elements, and a fine one with approximately
8000 elements. The results showed that the differences for the
effective coefficients for the medium to fine ones are of 104 order.
Therefore, all the results are shown only for the meshed RVE
namely medium (approximately 4000 elements).
In order to illustrate the comparison of the coefficients effective
calculated using Hashin’s, Nairn’s and present interface model,
some numerical examples are presented. The material properties
used in the calculations are listed in Table 1, which are given by
[38]. For almost all calculations, the interface volume fraction is
assumed to be equal 0.0001. For example, in Table 2, it was shown
the effective coefficients for fiber volume fraction equal 0.5, con-
sidering the two limit cases, i.e. complete separation of the inter-
face and perfect bonding. And, it is observed that the results of
the three models show convergence. Only one study was per-
formed considering different values for the interphase thickness
t i with fiber volume fraction equal 0.6 (Table 3). In this study,
the effective coefficient C eff
44 was calculated using relative difference
for K n ¼ 1012 and different values of the interface parameter K t , as
well. It is verified that Hashin’s interface model introduces an error
i 2
of order ðt Þ (see [24]). This can be explained due to the increasing
of the interphase thickness. Thus, the relative difference between
Hashin’s interface model and present one increases, mainly for
the values of interface parameter K t ¼ 100 ; 102 . In addition, as
Nairn’s is a particular case of present interface model, and both
models are equal when v 1 ¼ c  v i =2 and v 2 ¼ 1  c  v i =2 (see
formulas (9)–(12)), the relative difference between both models
increase with increasing thickness, but the divergences are more
relevant for higher values of thickness.
Fig. 1. Evolution of elastic property C eff
11 versus fiber volume fraction for different values of the interface parameters K n and K t .
75
In Table 4, the effective coefficients are calculated for Kn fixed
and different values of the interface parameter Kt with fiber vol-
ume fraction equal 0.5. Calculation the percentage difference
between Hashin’s and Nairn’s models as well as between Hashin’s
and present models are shown. In general, the three interface mod-
els provide good convergence. The lower differences are obtained
for the effective coefficients eeff
33 and eeff
33 for both comparisons,
and higher differences are shown for the effective coefficients C eff
44
and eeff
15 . This can be explained due to the interface shear modulus
Gi is finite when Kt approaches infinity (see formulas (10) and (12))
for Nairn’s and present interface model. However, the opposite
behavior is obtained for the Hashin’s interface model (see formula
(7)).
Figs. 1–11 show the effective coefficients for different cases,
such as, for complete separation of the interface and perfect bond-
ing (limit cases), and for different values of Kn and Kt. For all elastic
effective coefficients, except the effective coefficient C eff
33 , in the
case of perfect interface, the coefficients increases as the fiber vol-
ume fraction increases, and decreases for complete separation of
the interface. Different behavior is shown for the effective coeffi-
cient C eff
33 , which always increases when the fiber volume fraction
increases. This can be explained by applying the loading direction
3, which corresponds to the longitudinal direction of the fibers.
Therefore, the fibers drive the composite response. On the other
hand, in the case of the other elastic effective coefficients, there
is always influence of the matrix, making the role of interface is
33 , e11 and e33 exhibit
very important. The effective coefficients eeff eff eff
similar values for both limit cases. The results for Kn fixed and dif-
ferent values of the interface parameter Kt converge to perfect
interface results, except for the effective coefficients C eff eff
44 and e15 .
For the effective coefficient C eff
44 , the results provided by Hashin’s
interface model converge to complete imperfect interface, and for
eeff
15 , the results are outside of the ‘‘lower” and ‘‘upper bound”
defined for the results of the limit cases. Notice that for combina-
tion Kn = 1012 and Kt = 1, the results provided by Nairn’s and pre-
sent models tend to debonding results. Finally, it is important to
highlight that in the previous work [30], it was assumed ki =ti is
constant, i.e. a modified Hashin’s imperfect interface model was
used. And, in the present work, it was used the classical Hashin’s
interface model.
76 H. Brito-Santana et al. / Composite Structures 151 (2016) 70–80

Fig. 2. Evolution of elastic property C eff

12 versus fiber volume fraction for different values of the interface parameters K n and K t .

Fig. 3. Evolution of elastic property C eff

13 versus fiber volume fraction for different values of the interface parameters K n and K t .

Fig. 4. Evolution of elastic property C eff

33 versus fiber volume fraction for different values of the interface parameters K n and K t .
 H. Brito-Santana et al. / Composite Structures 151 (2016) 70–80 77

Fig. 5. Evolution of elastic property C eff

44 versus fiber volume fraction for different values of the interface parameters K n and K t .

Fig. 6. Evolution of elastic property C eff

66 versus fiber volume fraction for different values of the interface parameters K n and K t .

Fig. 7. Evolution of piezoelectric property eeff

13 versus fiber volume fraction for different values of the interface parameters K n and K t .
78 H. Brito-Santana et al. / Composite Structures 151 (2016) 70–80

Fig. 8. Evolution of piezoelectric property eeff

15 versus fiber volume fraction for different values of the interface parameters K n and K t .

Fig. 9. Evolution of piezoelectric property eeff

33 versus fiber volume fraction for different values of the interface parameters K n and K t .

Fig. 10. Evolution of dielectric property eeff

11 versus fiber volume fraction for different values of the interface parameters K n and K t .
 H. Brito-Santana et al. / Composite Structures 151 (2016) 70–80
Fig. 11. Evolution of dielectric property eeff
33 versus fiber volume fraction for different values of the interface parameters K n and K t .
5. Conclusions
Regarding the limit cases, considering totally perfect and com-
pletely imperfect interface, it is concluded that there is a conver-
gence between the three models investigated. Concerning to fiber
volume fraction 0.5, fixed Kn and different values of Kt and based
on the calculation of the percentage difference between Hashin’s
and Nairn’s models, and between Hashin’s and present model’s,
it is concluded that the proposed model is more generalized than
Nairn’s interface model. Therefore, Nairn’s interface model is a par-
ticular case of the proposed model. In the case of Hashin’s interface
model, depending on the type of imperfection investigated (only
varying Kt), there is no change in the interface coefficients for cer-
tain situations. It should be noted that this type of investigated
imperfection is purely theoretical.
With respect to the graphics of the effective coefficients for the
limit cases, including the results for Kn fixed and different values of
the interface parameter Kt, it follows that, in general, the effective
coefficient values increase as the fiber volume fraction increases
for the case completely perfect. However, the coefficients values
decreases with increasing volume fraction of fiber for fully imper-
fect case. In addition, the results for Kn fixed and different values of
the interface parameter Kt in general converge to perfect interface
results. For the Hashin’s results, the effective coefficient C eff
44 con-
verge to complete imperfect interface, and the effective coefficient
eeff
15 is outside of the ‘‘lower” and ‘‘upper bound” defined for the
results of the limit cases. Finally, for future work, the authors pro-
pose to conduct a study of the coefficients varying the volume frac-
tion, as well as the values of Kt and Kn.
The results in tables can confirmed some observations and con-
clusions pointed above. For different values for the interphase
thickness t i with fiber volume fraction equal 0.6, the effective coef-
ficient C eff
44 was calculated using relative difference for K n ¼ 10
12
and different values of the interface parameter K t , as well. It is ver-
ified that the relative difference between Hashin’s results and pre-
sent ones increases, mainly for the values of interface parameter
K t ¼ 100 ; 102 . In addition, as Nairn’s is a particular case of present
interface model, the relative difference between both models
increase with increasing thickness, but the divergences are more
relevant for higher values of thickness.
79
Acknowledgements
The authors wish to dedicate this work to Prof. Carlos Mota
Soares in his 70th anniversary and, mainly, to thank for his relevant
scientific contributions. The authors are thankful to Sao Paulo State
Research Foundation (FAPESP process number: 2012/01047-8), as
well as, CNPq (process number: 502265/2014-0 and
168279/2014-2) and FAPEMIG for partially funding the present
research work through the INCT-EIE. The authors also would like
to thank Prof. Marcelo Leite Ribeiro (EESC-USP) for the ABAQUSTM
license.
References
[1] Mota Soares CA, Pereira Leal R. Mixed elements in the sensitivity analysis of
structures. J Eng Optim 1987;11(3-4):227–37.
[2] Mota Soares CA, Rodrigues HC, Oliveira Faria LM, Haug EJ. Optimization of the
geometry of shafts using boundary elements. J Mech Transm Automotion
1984;106(2):199–202.
[3] Mota Soares CM, Mota Soares CA, Franco Correira VM. Optimal design of
piezolaminated structures. Compos Struct 1999;47(1–4):625–34.
[4] Franco Correira VM, Aguilar Gomes MA, Suleman A, Mota Soares CM, Mota
Soares CA. Modelling and design of adaptive composite structures. Comput
Methods Appl Mech Eng 2000;185(2-4):325–46.
[5] Walpole LJ. A coated inclusion in an elastic medium. Math Proc Cambridge
Philos Soc 1978;83(3):495–506.
[6] Duan HL, Yi X, Huang ZP, Wang J. A unified scheme for prediction of effective
moduli of multiphase composites with interface effects. Part I: theoretical
framework. Mech Mater 2007;39(1):81–93.
[7] Duan HL, Yi X, Huang ZP, Wang J. A unified scheme for prediction of effective
moduli of multiphase composites with interface effects. part II-application and
scaling laws. Mech Mater 2007;39(1):94–103.
[8] Rodríguez-Ramos R, Guinovart-Díaz R, López-Realpozo JC, Bravo-Castillero J,
Sabina FJ. Influence of imperfect elastic contact condition on the antiplane
effective properties of piezoelectric fibrous composites. Arch Appl Mech
2010;80(4):377–88.
[9] López-Realpozo JC, Rodríguez-Ramos R, Guinovart-Díaz R, Bravo-Castillero J,
Sabina FJ. Transport properties in fibrous elastic rhombic composite with
imperfect contact condition. Int J Mech Sci 2011;53(2):98–107.
[10] Rodríguez-Ramos R, Medeiros R, Guinovart-Díaz R, López-Realpozo JC, Bravo-
Castillero J, Otero JA, et al. Different approaches for calculating the effective
elastic properties in composite materials under imperfect contact adherence.
Compos Struct 2013;99:264–75.
[11] Delsanto PP. Rayleigh wave propagation in deformed orthotropic materials. J
Acoust Soc Am 1987;81(4):952–60.
[12] Pilarski A, Rose JL. A transverse-wave ultrasonic oblique-incidence technique
for interfacial weakness detection in adhesive bonds. J Appl Phys 1988;63
(2):300–7.
[13] Nagy PB. Ultrasonic classification of imperfect interfaces. J Nondestr Eval
1992;11(3/4):127–39.
80 H. Brito-Santana et al. / Composite Structures 151 (2016) 70–80
[14] Broutman LJ, Agarwal BD. A theoretical study of the effect of an interfacial
layer on the properties of composites. Polym Eng Sci 1974;14(8):581–8.
[15] Theocaris PS, Sideridis EP, Papanicolaou GC. The elastic longitudinal modulus
and Poisson’s ratio of fiber composites. J Reinf Plast Compos 1985;4
(4):396–418.
[16] Sideridis E. The in-plane shear modulus of fiber reinforced composites as
defined by the concept of interphase. Compos Sci Technol 1988;31(1):35–53.
[17] Lene F, Leguillon D. Homogenized constitutive law for a partially cohesive
composite material [J]. Int J Solids Struct 1982;18(5):443–58.
[18] Benveniste Y. The effective mechanical behaviour of composite materials with
imperfect constant between the constituents. Mech Mater 1985;4
(2):197–208.
[19] Aboudi J. Damage in composites modeling of imperfect bonding. Compos Sci
Technol 1987;28(2):103–28.
[20] Steif PS, Hoysan SF. An energy method for calculating the stiffness of aligned
short-fiber composites. Mech Mater 1987;6(3):197–210.
[21] Achenbach JD, Zhu H. Effect of interfacial zone on mechanical behavior and
failure of fiber-reinforced composites. J Mech Phys Solids 1989;37(3):381–93.
[22] Hashin Z. Thermoelastic properties of fiber composites with imperfect
interface. Mech Mater 1990;8(4):333–48.
[23] Hashin Z. Thin interphase/imperfect interface in conduction. J Appl Phys
2001;89(4):2261–7.
[24] Hashin Z. Thin interphase/imperfect interface in elasticity with application to
coated fiber composite. J Mech Phys Solids 2002;50(12):2509–37.
[25] Hashin Z. Composite materials with viscoelastic interphase: creep and
relaxation. Mech Mater 1991;11(2):135–48.
[26] Hashin Z. Thermoelastic properties of particulate composites with imperfect
interface. J Mech Phys Solids 1991;39(6):745–62.
[27] Nairn JA, Liu YCH. Stress transfer into a fragmented, anisotropic fiber through
an imperfect interface. Int J Solids Struct 1997;34(10):1255–81.
[28] Nairn JA. Generalized shear-lag analysis including imperfect interfaces. Adv
Compos Lett 2004;13(6):263–74.
[29] Angel YC, Achenbach JD. Reflection and transmission of elastic waves by a
periodic array of cracks: oblique incidence. Wave Motion 1985;7(4):375–97.
[30] Tita V, Medeiros R, Marques FD, Moreno ME. Effective properties evaluation for
smart composite materials with imperfect fiber–matrix adhesion. J Compos
Mater 2015;49(29):3683–701.
[31] Tan H, Huang Y, Liu C, Geubelle PH. The Mori-Tanaka method for composite
materials with nonlinear interface debonding. Int J Plast 2005;21
(10):1890–918.
[32] Tan H, Huang Y, Liu C, Geubelle PH. The cohesive law for the particle/matrix
interfaces in high explosives. J Mech Phys Solids 2005;53(8):1892–917.
[33] Wei YJ, Anand L. Grain-boundary sliding and separation in polycrystalline
metals: application to nanocrystalline fcc metals. J Mech Phys Solids 2004;52
(11):2587–616.
[34] Wang J, Duan HL, Zhang Z, Huang ZP. An anti-interpenetration model and
connections between interphase and interface models in particle-reinforced
composites. Int J Mech Sci 2005;47(4–5):701–18.
[35] Hashin Z. The spherical inclusion with imperfect interface. J Appl Mech
1991;58(2):444–9.
[36] Nairn JA. Numerical implementation of imperfect interfaces. Comput Mater Sci
2007;40(4):525–36.
[37] Sadd MH. Elasticity: Theory, Applications, and Numerics. Oxford: Butterworth-
Heinemann; 2005.
[38] Berger H, Kari S, Gabbert U, Rodríguez-Ramos R, Medeiros R, Guinovart-Díaz R,
et al. An analytical and numerical approach for calculating effective material
coefficients of piezoelectric fiber composites. Int J Solids Struct 2005;42(21–
22):5692–714.