Computers and Structures 83 (2005) 804820

www.elsevier.com/locate/compstruc

Anisotropic and piezoelectric materials fracture analysis

by BEM
F. Garca-Sanchez a, Andres Saez

b,*
,

J. Domnguez

a
b

Departamento de Ingeniera Civil, Universidad de Malaga, Plaza de El Ejido s/n, 29013-Malaga, Spain
Departamento de Mecanica de Medios Continuos, Escuela Superior de Ingenieros, Universidad de Sevilla,
Camino de los Descubrimientos s/n, 41092-Sevilla, Spain
Accepted 1 September 2004
Available online 11 January 2005

Abstract
A mixed boundary element approach for two-dimensional anisotropic and piezoelectric fracture mechanics problems
is presented in this paper. The numerical approach is based on displacement and traction integral equations for external
and crack boundaries, respectively. Integrals with strongly singular and hypersingular kernels are analytically transformed into weakly singular and regular integrals prior to any numerical evaluation. This is achieved by the simple election of an integration variable consistent with the material characteristic parameters. The generality of the method
allows for the use of curved and quarter-point elements and the evaluation of stress and electric displacement intensity
factors from nodal values next to the crack tip. Several crack problems in anisotropic and piezoelectric materials are
solved. The obtained results are in good agreement with previous solutions for cases where these solutions exist. The
present BE approach is more general and simple than previous procedures and has allowed for solution of a variety
of crack problems including curved cracks in piezoelectric materials which had never been studied before.
2004 Elsevier Ltd. All rights reserved.
Keywords: Boundary element method; Piezoelectric solids; Anisotropic solids; Fracture mechanics; Curved crack; Quarter-point
element

1. Introduction
Piezoelectric behavior was identied by Pierre and
Jacques Curie in 1880. They observed that some materials produce an electric eld when deformed and undergo
deformation when subjected to an electrical eld. Since
the mid of the twentieth century, there have been piezo-

*
Corresponding author. Tel.: +34 954487293; fax: +34
954487295.
E-mail address: andres@us.es (A. Saez).

electric ceramics with a piezoelectric ratio between electric eld and mechanical stress (and between mechanical
strain and electric displacement), two orders of magnitude higher than that of natural piezoelectric materials.
Lead zirconate titanate (PZT) is the most widely used
piezoceramic and polyvinylidene uoride (PVDF) the
most extended piezopolymer. They were rst produced
in 1946 and 1969, respectively. Piezoelectric materials
have become the base for construction of sensors, transducers, actuators and adaptive structures. Modeling
of piezoelectrics is complicated by the fact that they
exhibit not only electro-elastic coupling but anisotropic

0045-7949/$ - see front matter 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compstruc.2004.09.010

F. Garca-Sanchez et al. / Computers and Structures 83 (2005) 804820

behavior as well. Piezoelectric eect can only appear in

crystals that lack a center of symmetry; therefore, they
are always anisotropic.
Piezoelectric materials are brittle and due to manufacturing process and complex electromechanical loads
they are likely to develop cracks. The understanding
and evaluation of the fracture process in piezoelectric
materials is crucial to the advancement of modern intelligent material systems. Among the most signicant publications on the eld of piezoelectric materials fracture
mechanics, one can cite the works of Barnett and Lothe
[1], Deeg [2], Pak [3], Suo et al. [4], Sosa [5], Park and
Sun [6] and Wu et al. [7].
It is well known that the boundary element method
(BEM) presents signicant advantages over other
numerical techniques for the analysis of fracture
mechanics problems. This fact has led to the publication
of several BE approaches for the analysis of cracks in
piezoelectric solids in the last few years. The main
diculties in the eld are related to derivation and
integration of fundamental solutions for two- and
three-dimensional static and dynamic problems. Pan
[8] presented a single domain BE formulation for 2-D
static crack problems. He derived the fundamental solution using the complex variable function method [911]
and computed the hypersingular integrals using a
numerical quadrature. Liu and Fan [12] established
some of the basic equations in a rigorous way and
addressed the question of degeneration for cracks and
thin shell-like problems. Rajapakse and Xu [13] and
Xu and Rajapakse [14], used Lekhnitskiis formalism
and distributed dislocation modeling to derive special
Greens functions for an innite medium containing a
crack. They studied dierent crack geometries including
branched cracks. Denda and Lua [15] developed a BEM
formulation using Strohs formalism to derive the fundamental solution but they did not show any numerical
result. Davi and Milazo [16] presented a multidomain
approach based on the conventional BE formulation.
The fundamental solution was obtained by a variant
of Lekhnitskiis functions method similar to that used
by Rajapakse and Xu [13,14]. All these papers deal with
2-D domains.
Very few papers on BE formulations for three-dimensional piezoelectric solids have been published. Particular attention should be paid to the paper by Hill and
Farris [17] where the fundamental solution is obtained
using the Radon transform as in Deegs work [2] and
some simple numerical examples are analyzed using quadratic elements. A 3-D fundamental solution for displacement and electric potential discontinuity was
presented by Zhao et al. [18] and was applied to the case
of a circular crack in innite domain [19].
Boundary elements for time harmonic 2-D piezoelectric problems have been developed and applied to natural frequencies computation by Denda and Araki [20].

805

Kogl and Gaul [21] studied 3-D dynamic problems using

the static fundamental solution as derived by Deeg [2]
and a Dual Reciprocity formulation. Daros and Antes
[22] did a pioneering work obtaining a fundamental
solution for 3-D dynamic piezoelectric problems.
Sollero and Aliabadi [23] and Pan and coworkers
[2426] studied anisotropic 2-D crack problems using
the mixed formulation (dual). Pan [8] extended his
previous formulation to piezoelectric solids. Those
two-dimensional approaches for anisotropic and piezoelectric crack problems contain geometrical limitations
(on the rst case) or are based on numerical integration
procedures which are less accurate and less robust than
the simple general analytical approach presented in this
paper.
In the rst part of this paper, a general mixed BE formulation based on displacement and traction integral
equations for 2-D cracked anisotropic media is presented. Next, the formulation is extended to piezoelectric solids. In both cases, a new regularization
procedure based on a simple change of variables is presented. All the strongly singular or hypersingular integrals are transformed into regular integrals and simple
singular integrals with known analytical solution. The
traction (and normal electric displacement) integral
representation is written for the crack surface and the
displacement (and electric potential) integral representation for the external boundaries. The basic variables are
the opening displacement (and the electric potential discontinuity) along the crack, and the displacement (and
electric potential) and traction (and normal electric
displacement), on the external boundaries. Quadratic
quarter-point elements are used to represent the crack
opening displacement (COD) and potential discontinuity near the crack tip. Standard curved or straight
quadratic elements are used for the rest of the crack
and the external boundaries. Stress Intensity Factors
(SIF) are computed in a direct way from the COD and
the electric potential discontinuity at a point extremely
close to the tip. Numerical examples regarding full
plane, half plane and bounded domain crack problems
are analyzed. The obtained results are compared with
those existing in the literature. The present BE procedure is shown to be accurate and robust.

2. Basic equations
2.1. Elastic anisotropic materials
The mixed formulation for the BE solution of crack
problems is based on both displacement and traction
integral representations. In the case of zero body forces,
the 2-D displacement integral representation for a point
with coordinates n1 and n2 in an elastic anisotropic solid
can be written as

F. Garca-Sanchez et al. / Computers and Structures 83 (2005) 804820

806

Z
cij znm uj znm pij zxm ,znm uj zxm dCzxm
C
Z
 x n

uij zm ,zm pj zxm dCzxm

where i, j = 1, 2; uij and pij are the fundamental solution

displacements and tractions, respectively; cij znm dij
when n is an internal point and cij znm 12 dij for a smooth
boundary point; and the source point zn and the observation point zx are dened on the complex plane as
znm n1 lm n2 ;

zxm x1 lm x2 ;

m 1,2

lm being the roots of the following characteristic

equation:
jC 1ij1 C 1ij2 C 2ij1 lm C 2ij2 l2m j 0

where Cijkl are the elastic moduli of the material. The

roots of (3) are either complex or purely imaginary
and always occur in conjugate pairs [9].
Greens functions in 2-D anisotropic elastic and piezoelectric media may be obtained using the complex variable solution approach [911] for an unbounded plane,
a half plane and bimaterial problems among others.
Thus, for both an unbounded plane and a half plane
y 6 0, y = 0 being a traction free surface, the fundamental solution displacements and tractions are given by (see
[27])
"
(
2
X
1
 x n
uij zm ,zm  Re
Ajr H ri lnzxr  znr
p
r1
#)
2
X
x
n
Ers H si lnzr  zs
4
a

The traction integral representation for a source point n

with unit normal N is obtained by dierentiation of Eq.
(1) with respect to nk and the subsequent application of
Hookes law, to yield
Z
n
n
cij zm pj zm N r srij zxm ,znm uj zxm dCzxm
C
Z
N r d rij zxm ,znm pj zxm dCzxm
9
C

where sijk and d ijk are therefore obtained by dierentiation of pij and uij , respectively, with the following
expressions:
d ijk C ijml umk,l

10

sijk C ijml pmk,l

11

The derivatives of uij and pij at the collocation point are

evaluated from
(

2
X
1
dk1 lr dk2

x n
uij,k zm ,zm Re
Ajr H ri
p
zxr  znr
r1
#)
2
X
dk1 ls dk2
Ers H si
a
12
zxr  zns
s1
pij,k zxm ,znm

(
"
2
X
1
l n1  n2
Re
Ljr H ri dk1 lr dk2 r
x  zn 2
p
z
r1
r
r
#)
2
X
l n1  n2
Ers H si dk1 ls dk2 r
a
x  zn 2
z
s1
r
s
13

s1

(
"
2
X
1
l n1  n2
Ljr H ri r x
pij zxm ,znm Re
zr  znr
p
r1
#)
2
X
lr n1  n2
Ers H si
a
zxr  zns
s1

2.2. Piezoelectric materials

where a = 0 for the full plane case and a = 1 for the half
plane problem; n = (n1, n2) is the outward unit normal at
the observation point x. For each characteristic root lk,
the columns of the A matrix are obtained from
C 1ij1 C 1ij2 C 2ij1 lm C 2ij2 l2m Ajm
0 no sum on m

and the components of the L matrix are given by

Lim

2
X
C 2ij1 C 2ij2 lm Ajm

j1

and matrices H and E are obtained from

1

H A1 B1 B 1 ;

E L1 L

with B iAL1
8

For piezoelectric materials new eld variables need to

be considered, namely, the electric potential (/) and the
electric displacements (Di). Following Barnett and Lothe
[1] the piezoelectric problem may be formulated in an
elastic-like fashion by using a displacement vector extended with the electric potential

ui , I 1,2
uI
14
/, I 3
an extended stress tensor (extended with the electric
displacement components) and its associated extended
tractions vector (extended with the normal electric
displacement). Thus,
8
2
P
>
>
rij nj , I 1,2
> Pi

<
rij , J 1,2
j1
riJ
) PI
2
>
P
DJ , J 3
>
>D
Dj nj , I 3
:
j1

15

F. Garca-Sanchez et al. / Computers and Structures 83 (2005) 804820

The constitutive relations may be written as

rij C ijlm clm  ekji Ek ;

Di eijk cjk eij Ej

16

where cij is the strain tensor and Ei is the electric eld

(Ei = /,i); Cijkl, eijk and eij are the elastic moduli, the
piezoelectric coecients and the dielectric constants,
respectively. The elastic and electric moduli can be
grouped together as
8
C ijkl , J ,K 1,2
>
>
>
<e ,
J 1,2; K 3
lij
17
C iJKl
>
e
,
J
3; K 1,2
ikl
>
>
:
eil , J ,K 3
where the lowercase (elastic) and uppercase (extended)
subscripts take values 1, 2 and 1, 2, 3, respectively.
Based on the extended notation the integral equations for the mixed formulation of the BEM for piezoelectric problems can be written as
Z
n
n
cIJ zM uJ zM pIJ zxM ,znM uJ zxM dCzxM
C
Z

x
n
uIJ zM ,zM pJ zxM dCzxM
18

where the columns of the matrix A are obtained for each

characteristic root lM from
C 1IJ 1 C 1IJ 2 C 2IJ 1 lM C 2IJ 2 l2M AJM
0

no sum on M

Nr

Nr

LIK

3
X
C 2IR1 C 2IR2 lK ARK

and the matrix H is obtained as in (8) for the extended

variables
1

H A1 B1 B 1

with B iAL1

26

The roots lM and the matrices A and L may be computed alternatively by solving the following eigenvalue
problem [28]:
Nfa la fa

27

where
N

srIJ zxM ,znM uJ zxM dCzxM

fa

d rIJ zxM ,znM pJ zxM dCzxM

19

uIJ

pIJ

and
are the fundamental soluwhere r = 1, 2 and
tion extended displacements and tractions, respectively,
associated to a line force (I = 1, 2) or to a line charge
(I = 3) at the source point. The collocation point zn
and the observation point zx are dened on the complex
plane as
zxM x1 lM x2 ;

M 1,2,3

20

where lM are the three distinct complex roots with positive imaginary part of the following characteristic
equation:
j C 1IJ 1 C 1IJ 2 C 2IJ 1 lM C 2IJ 2 l2M j 0

C 1
2IJ2 C 1JI2

C 1
2IJ 2

C 1IJ 2 C 1
2IJ2 C 1JI2  C 1IJ 1

C 1IJ 2 C 1
2IJ 2

(
)
3
X
1

x n
x
n
uIJ zK ,zK  Re
AJR H RI lnzR  zR
p
R1
(
)
3
X
1
lR n1  n2
Re
LJR H RI x
p
zR  znR
R1

Aa
Ba

!
28

22

23

29

The terms siJK and d iJK that appear in the tractions integral equation (19) are obtained by dierentiation of pIJ
and uIJ as
d iJK C iJMl uMK ,l

30

siJK C iJMl pMK ,l

31

The derivatives at the collocation point are evaluated

from
(
)
3
X
1
dk1 lR dk2

x
n
32
AJR H RI
uIJ ,k zM ,zM Re
p
zxR  znR
R1

21

As for the anisotropic case, the complex variable solution approach provides the fundamental solution extended displacements and tractions for the unbounded
plane, the half plane and the bimaterial problems among
others, see [8]. For instance, for the full plane case the
extended fundamental solution may be written in a similar manner to (4) and (5)

pIJ zxK ,znK

25

R1

and

Z
C

znM n1 lM n2 ,

24

and L is given by

cIJ znM pJ znM

807

pIJ ,k zxM ,znM

(
)
3
X
1
lR n1  n2
Re
LJR H RI dk1 lR dk2 x
p
zR  znR 2
R1
33

3. Mixed BEM and discretization approach

In what follows we shall refer to the extended formulation only, since the elastic anisotropic problem may be
treated as a simplication of the piezoelectric case.
In some mixed BEM approaches the extended traction BIE (19) is applied at nodes on one side of the crack
surface C+ and the extended displacement BIE (18) at
nodes on the other side of the crack C and the rest of

F. Garca-Sanchez et al. / Computers and Structures 83 (2005) 804820

808

Fig. 2. Discontinuous quadratic element and natural coordinate (1 6 f 6 1).

Fig. 1. Cracked 2-D anisotropic domain and BE discretization.

the boundary CC (Fig. 1). However, if the sum of the

tractions on both sides of the crack surface is zero

(Dpk p
k p k 0) and the crack is not permeable
(no external charge resides on either crack face and the
electric induction in the environment is negligible:
D+ = D = 0), it is no longer necessary to apply both
equations on the crack surface. In such a case the extended traction BIE is enough to obtain the extended
crack opening displacement (ECOD), which is the relevant quantity in fracture problems
0
1
u1  u
1
B
 C
DuJ @ u
34
2  u2 A
/  /

Therefore, the extended displacement BIE applied on CC

Z
Z

cIJ uJ --- pIJ uJ dC
pIJ DuJ dC

CC

uIJ pJ

dC

35

CC

and the extended traction BIE applied on one side of the

crack surfaces, say C+
Z
pI N r
srIJ uJ dC N r srIJ DuJ dC
CC

Nr

d rIJ pJ dC
CC

36

Fig. 3. Discontinuous quarter-point element.

provide a complete set of equations to determine the extended tractions and extended displacements on the
boundary
and the ECOD along the crack. Symbols
R
 and in Eqs. (35) and (36) stand for the Cauchy Principal Value (CPV) and the Hadamard Finite Part (HFP)
of the integral, respectively. Note that the dierence between Eq. (36) and the general extended traction BIE
(19) is the free term, which is set to 1 to account for
the additional singularity arising from the coincidence
of the two crack surfaces.
The discretization approach follows Saez et al. [29]
and it is summarized in Fig. 1. Therefore, in order to fulll the C1 continuity of the displacements at collocation
points that the limit to the boundary requires, non-conforming (discontinuous) quadratic elements are considered to mesh the crack (Fig. 2), i.e., quadratic elements
where the two extreme collocation nodes are shifted towards the interior of the element. To reproduce the extended displacements behavior near the crack tip,
simple straight line discontinuous quadratic elements
with the mid-node located at one quarter of the element
length (Fig. 3) are used. Standard continuous quadratic
elements are used for the rest of the boundary, except for

Fig. 4. Quadratic shape functions for discontinuous elements: (a) shape functions for the geometry; (b) shape functions for the
boundary variables.

F. Garca-Sanchez et al. / Computers and Structures 83 (2005) 804820

the case when there is an intersection between a crack

and an external boundary. In such a case a semidiscontinuous element is used on the outer boundary to avoid a
common node at the intersection. The geometry of the
discontinuous quadratic elements is represented using
the same shape functions /1G, /2G and /3G of the standard (continuous) quadratic elements (Fig. 4(a)), whilst
the extended boundary variables uK, pK and DuK are represented in terms of their values at the shifted collocation nodes NC1, NC2 and NC3 and the shape
functions /1, /2 and /3 (Fig. 4(b)).

4. Evaluation of integrals
4.1. Weakly singular integrals

809

it follows that
dvM dvM dx1 dvM dx2

n2 lM n1
dC
dx1 dC dx2 dC

40

where n1, n2 are the components of the external unit normal to the boundary at the observation point x (Fig. 5).
Eq. (40) is the key idea for all the transformations proposed below and it shows that the Jacobian dvM/dC of
the coordinate transformation that maps the geometry
of the boundary element Ce onto the complex plane
vM is included in the fundamental solution itself for both
the elastic anisotropic and piezoelectric cases and for the
full plane, the half plane and the bimaterial solutions.
Taking into account Eq. (40), (38) can be rewritten as
Z
1
IM
/q dvM
41
v
Ce M

Integrals of the kernels uIJ in Eq. (35) show a weak

singularity of the type 0 lnzxM  znM as the observation
point approaches the collocation point (x ! n), which
can be numerically integrated without diculty by using
a special quadrature including a logarithm.

which can be decomposed into the sum of a regular integral plus a singular integral with known analytical
solution
Z
Z
1
1
IM
/q  1 dvM
dvM
42
Ce vM
Ce vM

4.2. Singular integrals

Alternatively, this singular integration and the computation of the free terms cIJ(n) in (35) can be avoided by
using the extended rigid body motion condition (rigid
body motion-uniform electric potential).
The integration of the kernels d rIJ N r d IJ can be
done in a similar way as for the pIJ kernels since they
contain singularities of the same type when x ! n.
Although such singular integration is not needed when
the boundary conditions Dpk = 0 and D+ = D = 0 hold
on the crack, as in Eq. (36), it is included here for the
sake of completeness.
From Eqs. (30) and (32) (or Eqs. (10) and (12) for the
anisotropic case) it follows that the singular integrals
that arise in the piezoelectric problem are of the type
Z
1
I 0M
lM N 1  N 2 /q dC; M 1,2,3
43
v
Ce
M

The kernels pIJ and d iJK appearing in Eqs. (35) and

(36) show a strong singularity of 0 1=zxM  znM as
x ! n. The integration of such kernels over the element
Ce that contains the collocation point n can be done as
follows.
After discretization, integrals of the type
Z
pIJ /q dC
37
Ce

where /q are the boundary element shape functions,

contain basic singular integral as (see Eq. (5) for the
anisotropic case and (23) for the piezoelectric one)
Z
1
IM
lM n1  n2 x
/q dC; M 1,2,3
38
zM  znM
Ce
Calling
vM zxM  znM x1  n1 lM x2  n2 ;

M 1,2,3
39

which can be regularized as follows:



Z
Z
1
dvM
1
0
IM
/q dC
lM N 1  N 2 
/q dvM
v
v
dC
Ce M
Ce M
44
The rst integral in (44) is regular, since nj ! Nj when
x ! n and Eq. (40) holds. The second integral has been
previously studied (Eqs. (41) and (42)).
4.3. Hypersingular integrals

Fig. 5. Outward unit normal at boundary point x.

The integration of the srIJ kernels in Eq. (36) leads to

hypersingular integrals since they show a singularity of
the type 0 1=zxM  znM 2 as x ! n. From Eqs. (11)/(31)
and (13)/(33) it follows that the hypersingular integrals
are of the form

F. Garca-Sanchez et al. / Computers and Structures 83 (2005) 804820

810

I 00M

Z
Z

lM n1  n2
Ce

Ce

1
/ dv ;
v2M q M

1
/ dC
v2M q
M 1,2,3

45

where Eqs. (39) and (40) have been taken into account.
The integral I 00M can be again decomposed into the sum
of a regular integral plus singular integrals with known
analytic solution by using Eq. (40) and the rst two
terms of the series expansion of the shape function /q
at the collocation point, considered as a function of
the complex space variable vM

d/q 
/q vM /q vM 0
vM 0v2M
dvM vM 0
 /q0 /0q0 vM
Thus, I 00k
Z
I 00M
C
Z e

5. Fracture mechanics applications

46

can be written as

1
/ dv
v2M q M
1
/q  /q0  /0q0 vM dvM
2
v
Ce M
Z
Z
1
1
0
/q0
dv

/
dvM
M
q0
2
Ce vM
Ce vM

quadratures for hypersingular integrals (see, e.g., Kutt

[32] or Ladopoulos [33] among others) shows that the
numerical evaluation of such integrals is still an open issue where eciency and accuracy have to be
compromised.
The approach developed in this paper permits the
simple integration of all the kernels appearing in both
displacement and traction BIE for any curved or straight
quadratic element by transformation of all the strongly
singular and hypersingular integrals into regular integrals plus simple integrals with known analytical
solution.

47

The rst integral in (47) is regular and the other two can
be easily computed analytically.
The procedure presented here is a generalization of
the integration approach developed by Garca et al.
[30] for the anisotropic full plane fundamental solution
to other fundamental solutions obtained by means of
the complex variable method, namely the ones corresponding to the half plane and the bimaterials in anisotropic media and to the full plane, the half plane and the
bimaterials in piezoelectric media. Such generalization is
straightforward because the change of variable dened
in Eqs. (39) and (40) is consistent with all those fundamental solutions.
Since the series expansion is considered for the shape
functions, which have simple expressions, the procedure
presented is more general and easier to implement than
the one given in Ref. [23], where the expansion is done
for a more complicated function f(f) dened by the
product of the anisotropic fundamental solution, a
shape function, the Jacobian J(f) of the coordinate
transformation to the local parametric coordinate f
and the hypersingular term (f  f0)2, f0 dening the collocation point location. In Ref. [23] all the integrals are
solved analytically for straight elements, since for that
case the Jacobian is constant. The generalization of such
procedure to curved elements or to the other fundamental solutions considered here is rather complicated due
to the complexity of the kernel function f(f).
Pan and coworkers evaluate the hypersingular integrals numerically by using Tsamasphyros et al. quadrature rules [31]. However, the existence of dierent

The generality of the integration method developed

in previous sections allows for the use of straight line
quarter-point quadratic crack tip elements with collocation points located extremely close to the crack tip (Fig.
3), as previously presented in Saez et al. [29] for isotropic
materials and Garc
pa et al. [30] for anisotropic materials.
By doing so, the r behavior that the ECOD shows in
the vicinity of the crack front as the distance to the crack
front r tends to zero [34,4], for both the anisotropic and
piezoelectric cases, is reproduced
r
r
f2
1
48
L
f being the boundary element natural coordinate and L
the element length (Fig. 6).
These elements are used only to represent the crack
tip vicinity, i.e., only the elements containing the crack
tip are quarter-point quadratic, the rest of the elements
along the crack being discontinuous quadratic elements
with straight or curved geometry. Thus, curved cracks
are easily represented assuming only straight crack
geometry for the element containing the tip.
In the present work the collocation nodes NC1, NC2
and NC3 for the quarter-point element are located at
f1 = 3/4, f2 = 0 and f3 = +3/4, respectively (Fig. 6). In
such case, the distance r from the collocation nodes of
the quarter-point element to the crack tip follows from
Eq. (48)

Fig. 6. Discontinuous quarter-point element with collocation

nodes at f1 = 3/4; f2 = 0; f3 = 3/4.

F. Garca-Sanchez et al. / Computers and Structures 83 (2005) 804820

L
for NC1
64
L
r2
for NC2
4
49L
r3
for NC3
64

6.1. Crack in unbounded domain

r1

49

Dening the usual polar coordinate system r  h, centered at the crack tip and such that h = p are the crack
surfaces, the shear (Du1), normal (Du2) and electric potential (D/) components of the ECOD near the crack
tip may be written in terms of the Extended Stress Intensity Factors (ESIF) as [4]
1

1
0
r
K II
8r
C
C
B
B
RefBg@ K I A
@ Du2 A
p
D/
K IV
0

Du1

50

where B has been dened in Eq. (26). Only the leading

term of the asymptotic behavior of the ECOD near
the crack tip has been considered in (50). KI and KII
are the usual mode I and mode II SIF, respectively,
and KIV is the electrical displacement intensity factor.
Particularizing Eq. (50) for the collocation node NC1
(r L=64) the following one point formula for the direct
evaluation of the ESIF can be obtained
8
8 NC1 9
9
>
>
< K II >
< Du1 >
=
=
p
KI
2 2p=LRefBg1 DuNC1
2
>
>
>
:
: NC1 >
;
;
K IV
D/

811

51

Several other approaches to evaluate the ESIF directly

from the computed nodal values may be considered, as
described in Saez et al. [29] for the isotropic case. However, the procedure presented in Eq. (51) is the more
accurate and the less mesh-dependent one, since the
ESIF is calculated from the ECOD at a collocation
p
point extremely close to the crack front, where the r
behavior of the ECOD is highly dominant.
It should be pointed out that the use of this quarterpoint element does not need of any additional eort for
its implementation.

The rst example corresponds to a nite straight

crack along the x-direction in an innite PZT-4 plane
under a uniform far eld stress or electric displacement
(Fig. 7). The material constants are shown in Table 1.
The axis of symmetry of the material is the y-axis. Ten
quadratic elements with decreasing size towards the
crack tip are used to reproduce the crack. The ratio between the lengths of the largest and the shortest element
is 2 with the crack tip elements being quarter-point
elements.
The obtained crack opening displacements and electric potential increment along the crack are shown in
Fig. 8. The results are compared with the known exact
analytical solution obtained by Pak [3] and Suo et al.
[4]. The agreement between the present numerical solution and the exact solution is very good (Fig. 8). The
ESIF evaluated using Eq. (51) agree with the exact solution within a 0.1%.
6.2. Inclined crack in a rectangular plate
A rectangular PZT-4 plate with a 45 inclined crack
is considered (Fig. 9). The plate is under a uniform
traction or electric displacement applied in the y-direction. The geometry and material properties are taken
from Pan [8] for comparison purposes. The plate halflength, h = 1 m, the half-width, w = 0.5 m, and the
crack half-length, a = 0.1 m. Material properties are given in Table 1. The BE discretization consists of 24
equal quadratic elements on the external boundary
and 10 discontinuous elements of variable length along
the crack, with a length ratio between the largest elements, at the middle of the crack, and the shortest ones,
at the crack tip, of 2. Crack tip elements are quarterpoint.
Values of the ECOD along the crack are shown in
Figs. 10 and 11 for uniform traction and uniform electric displacement loading conditions, respectively. The

6. Numerical applications
In the following, several examples with dierent
geometries and material properties are considered in order to show the generality and robustness of the present
approach. The obtained results are compared with those
presented by other authors when previous solutions exist. In the case of curved crack no comparison is carried
out because, to the best of the authors knowledge, no
previous analytical or numerical solution exists for this
problem.

Fig. 7. Crack in innite domain under far eld uniform stress/

electric displacement.

F. Garca-Sanchez et al. / Computers and Structures 83 (2005) 804820

812
Table 1
Material constants

Material constants for examples 6.1 and 6.2 [26]

C11 (GPa)
C12 (GPa)
139
74.3
e21 (C/m2)
e22 (C/m2)
5.2
15.1

C22 (GPa)
115
e16 (C/m2)
12.7

C66 (GPa)
25.6
e11 (C/(GV m))
6.461

e22(C/(GV m))
5.62

Material constants for examples 6.3 and 6.4 [35]

C12(GPa)
C11 (GPa)
139
74.3
C21 (C/m2)
e22 (C/m2)
6.98
13.84

C22 (GPa)
113
e16 (C/m2)
13.44

C66 (GPa)
25.6
e11 (C/(GV m))
6.0

e22 (C/(GV m))

5.47

Fig. 8. Crack opening displacements and electric potential increment along crack in innite domain.

by Pan [8]. The agreement between both sets of results is

excellent as well as for the ESIF evaluated from those
ECOD.
6.3. Branched crack

Fig. 9. Inclined crack in rectangular piezoelectric plate.

computed values are compared in the gures with those

obtained for the same problem and similar discretization

Crack branching is a common phenomenon in the

fracture of brittle materials. However, the coupling between electrical and mechanical elds makes crack
branching more complex for piezoelectrics than for elastic materials. One of the pioneering works in this area
was published by Xu and Rajapakse [35]. They presented a theoretical study based on a distributed dislocation model and Lekhnitskiis complex potentials to solve
branched crack problems in an unbounded piezoelectric
plane. A numerical study of one of the problems studied
by Xu and Rajapakse is presented next for validation
purposes. More general nite plate geometries can be
studied with the present approach without special
diculties.

F. Garca-Sanchez et al. / Computers and Structures 83 (2005) 804820

813

Fig. 10. Crack opening displacement and electric potential increment along the crack for uniform traction loading (ry = 1 Pa).

Fig. 11. Crack opening displacement and electric potential increment along the crack for uniform electric displacement loading
(Dy = 1 C/m2).

Fig. 12. Branched crack geometry.

The branched crack with geometry shown in Fig. 12,

in an unbounded plane, is considered. The material is

PZT-4 with the same properties as in Ref. [35]. Material

constants are given in Table 1. The axis of symmetry of
the material is the y-axis and the main crack is along the
x-axis with a branch with an angle h initiating from one
of the crack tips. Twelve quadratic elements with decreasing size to the crack tips are used for the main crack with
a length ratio between the largest and the shortest element of 2. Two equal length elements are used for the
branch when its length is b = a/10 and six elements with
increasing size to the mid-length of the branch when
b = a/2. Crack tip elements are quarter-point. The crack

814

F. Garca-Sanchez et al. / Computers and Structures 83 (2005) 804820

is under a uniform far eld stress or electric displacement along the y-axis. ESIF are evaluated form the
ECOD values, in accordance to Eq. (51).
Fig. 13 shows the computed values of KI and KII versus the branch angle h for two dierent lengths of the
branch when the load is a uniform traction along the
y-axis. A good agreement with the results obtained by
Xu and Rajapakse [35] is observed. The obtained electric
displacement intensity factor is shown in Fig. 14 for the
same two branch lengths and dierent branch angles h.
The agreement with the results in Ref. [35] is also good
in this case.
Values of KI and KII for a uniform far eld electric
displacement for the same two dierent branch lengths
are shown versus the branch angle h in Fig. 15. The ob-

tained results are in good agreement with those presented by Xu and Rajapakse [35]. Finally, the electric
displacement intensity factor, for the same far-eld electric displacement loading and branch lengths, are shown
versus the branch angle h in Fig. 16. Once again, the
agreement with the results in Ref. [35] is good.
Figs. 1316 show results for values of h between 0
and 80 or 90. Higher values of h are not possible for
the considered crack geometries without closure of the
crack branch [35].
6.4. Curved crack in unbounded domain
To show the use of the current procedure for curved
crack geometries, the problem of a crack with circular

Fig. 13. Stress intensity factors at the tip of a branched crack under far eld tension.

Fig. 14. Electric displacement intensity factor at the tip of a branched crack under far eld tension.

F. Garca-Sanchez et al. / Computers and Structures 83 (2005) 804820

815

Fig. 15. Stress intensity factors at the tip of a branched crack under far eld electric displacement.

Fig. 16. Electric displacement intensity factor at the tip of a branched crack under far eld electric displacement.

arch shape in an unbounded domain is studied. The

crack region is under a uniform far eld stress or electric
displacement (Fig. 17) along the material and crack axis
of symmetry. The material properties are the same as in
the previous 6.3 example. Cracks with semi-angles h between 0 (straight crack) and 73.8 or 88.6 for elastic
stress or electric displacement loading, respectively, are
analysed. Larger values of h imply crack closure.
The BE mesh consists of eight curved quadratic elements and two very small (arch length/30) quarter-point
straight elements at the tips (Fig. 17). Uniform far eld
elastic stress is considered rst. Computed values of the
normalized mode I and mode II SIF are shown versus h
in Fig. 18, and values of the normalized electric displacement intensity factor KIV in Fig. 19. Notice that the ratio
v between the dielectric constant e22 and the piezoelectric

Fig. 17. Geometry, mesh and loading conditions for a circular

arch crack.

constant e22 has been used to represent a dimensionless

value of KIV. Computed values of KI and KII when the

816

F. Garca-Sanchez et al. / Computers and Structures 83 (2005) 804820

Fig. 18. Mode I and mode II SIFs for circular crack in piezoelectric solid under uniform far eld stress.

Fig. 19. Electric displacement intensity factor for circular crack in piezoelectric solid under uniform far eld stress.

load is a uniform electric displacement are shown in Fig.

20, whereas the obtained values of KIV for the same
loading conditions are shown in Fig. 21. These results
show the great inuence that crack curvature has on
ESIF.
The BE mesh for this application (Fig. 17) is very
similar to that used by the authors for the same crack
geometry in an elastic anisotropic plane in Ref. [30].
The accuracy of the results obtained in that case was
checked by comparison with known solutions. To the

authors knowledge, no solution for the curved crack

problem in piezoelectric solids has been published up
to the present.
6.5. Vertical crack in orthotropic half-plane
A crack embedded in a half-plane solid traction free
on the innite straight boundary is analysed. The crack
has a length 2a, is perpendicular to the innite boundary
(Fig. 22) and its center is located at a depth h = 1.1a

F. Garca-Sanchez et al. / Computers and Structures 83 (2005) 804820

817

Fig. 20. Mode I and mode II SIFs for circular crack in piezoelectric solid under uniform far eld electric displacement.

Fig. 21. Electric displacement intensity factor for circular crack in piezoelectric solid under uniform far eld electric displacement.

from the traction free surface. Two dierent loading

conditions are considered: uniform pressure and uniform shear on the crack faces. The material is orthotropic The material properties can be dened in terms of
two parameters j and d as done by Sung and Liou
[36]. The problem is solved using the half-plane fundamental solution and a BE discretization for the crack
consisting of twelve discontinuous quadratic elements.
For the sake of comparison with the semi-analytical
solution in Ref. [36], the analysis is restricted to materials with j = 1 and d = 1/3, 1/2, 1 (isotropic), 2 and 3. The

Fig. 22. Internal vertical crack in orthotropic half-plane under

tensile and shear stress.

818

F. Garca-Sanchez et al. / Computers and Structures 83 (2005) 804820

Fig. 23. Mode I and mode II SIFs for crack under tensile stress in orthotropic half-plane for dierent values of the c angle between the
loading direction and the orthotropic axis.

results obtained for the normalized mode I and mode II

SIF at the upper crack tip when the crack is under tensile
load are shown versus the angle c of one of the orthotropic directions in Fig. 23. It is worth to mention that
when the load axis is not parallel to one of the orthotropic axis (c 5 0, p/2), there is no symmetry in the problem
and KII 5 0 for tensile loading conditions. The obtained
results show a good agreement with those presented by

Sung and Liou [36]. Results obtained with the present

technique for a uniform shear on the crack faces are
shown in Figs. 24 and 25. Normalized mode I and mode
II SIF at the upper crack tip are shown versus the orthotropic axis angle c in Figs. 24 and 25, respectively. Once
more, the obtained solution is in good agreement with
the semi-analytical solution presented by Sung and Liou
[36] for the same problem as shown in the gures.

Fig. 24. Mode I SIF for crack under shear stress in orthotropic half-plane for dierent values of the c angle between the loading
direction and the orthotropic axis.

F. Garca-Sanchez et al. / Computers and Structures 83 (2005) 804820

819

Fig. 25. Mode II SIF for crack under shear stress in orthotropic half-plane for dierent values of the c angle between the loading
direction and the orthotropic axis.

7. Conclusions

References

A general mixed boundary element formulation for

crack problems in 2-D anisotropic and piezoelectric solids has been presented. One of the basic ideas of the
present approach is the analytical transformation of
hypersingular and strongly singular integrals into regular and weakly singular integrals by choosing a complex
integration variable consistent with the material characteristic parameters. This simple transformation approach allows for the use of dierent types of elements
along crack and external boundaries.
The approach has been applied to straight and
curved crack problems in piezoelectric bounded and unbounded domains. Half-plane crack problems have also
been studied. The obtained results show the robustness,
generality and simplicity of the procedure. To the
authors knowledge, the problem of a curved crack in
an unbounded piezoelectric solid has been studied in this
paper for the rst time. Stress and electric displacement
intensity factors have been obtained for this and other
crack problems. The present results are in good agreement with solutions presented by other authors in the
cases where those solutions exist.

[1] Barnett DM, Lothe J. Dislocations and line charges in

anisotropic piezoelectric insulators. Phys Status Solidi (b)
1975;67:10511.
[2] Deeg WF. The analysis of dislocation, crack and inclusion
problems in piezoelectric solids. PhD Thesis. Stanford
University; 1980.
[3] Pak YE. Linear electro-elastic fracture mechanics of
piezoelectric materials. Int J Fracture 1992;54:79100.
[4] Suo Z, Kuo C-M, Barnett DM, Willis JR. Fracture
mechanics for piezoelectric ceramics. J Mech Phys Solids
1992;40:73965.
[5] Sosa H. On fracture mechanics of piezoelectric solids. Int J
Solids Struct 1992;29:261322.
[6] Park SB, Sun CT. Fracture criteria for piezoelectric
ceramics. J Am Ceramic Soc 1995;78:147580. [correction
(1996)].
[7] Wu CC, Sce KY, Huang YQ. Numerical solution on
fracture of piezoelectric materials by hybrid element. Int J
Solids Struct 2001;38:431529.
[8] Pan E. A BEM analysis of fracture mechanics in 2D
anisotropic piezoelectric solids. Eng Anal Boundary Elements 1999;23:6776.
[9] Eshelby JD, Read WT, Shockley W. Anisotropic with
applications to dislocation theory. Acta Metal 1953;1:
2519.
[10] Stroh AN. Dislocations and cracks in anisotropic elasticity.
Phil Mag 1958;3:62546.
[11] Lekhnitskii SG. Theory of elasticity of an anisotropic
elastic body. San Francisco: Holden-Day; 1963.
[12] Liu Y, Fan H. On the conventional boundary integral
formulation for piezoelectric solids with defects or of thin
shapes. Eng Anal Boundary Elements 2001;25:7791.

Acknowledgment
This work was supported by the Ministerio de Ciencia y Tecnologa of Spain (DPI2004-08147-C02-02). The
nancial support is gratefully acknowledged.

820

F. Garca-Sanchez et al. / Computers and Structures 83 (2005) 804820

[13] Rajapakse RKND, Xu XL. Boundary element modelling

of cracks in piezoelectric solids. Eng Anal Boundary
Elements 2001;25:77181.
[14] Xu XL, Rajapakse RKND. On a plane crack in piezoelectric solids. Int J Solids Struct 2001;38:764358.
[15] Denda M, Lua J. Development of the boundary element
method for 2D piezoelectricity. Composites: Part B 1999;
30:699707.
[16] Davi G, Milazzo A. Multidomain boundary integral
formulation for piezoelectric materials fracture mechanics.
Int J Solids Struct 2001;38:706578.
[17] Hill LR, Farris TN. Three-dimensional piezoelectric
boundary element method. AIAA J 1998;36:1028.
[18] Zhao MH, Shen YP, Liu YJ, Liu GN. Isolated crack in
three-dimensional piezoelectric solid: part Isolution by
Hankel transform. Theor Appl Fracture Mech 1997;26:
12939.
[19] Zhao MH, Shen YP, Liu YJ, Liu GN. Isolated crack in
three-dimensional piezoelectric solid: part II: stress intensity factors for circular crack. Theor Appl Fracture Mech
1997;26:1419.
[20] Denda M, Araki Y. 2-D general anisotropic and piezoelectric time-harmonic BEM for eigenvalue analysis. In:
Proceedings of the 15th ASCE Engineering Mechanics
Conference, Columbia University, NY; 2002.
[21] Kogl M, Gaul L. A boundary element method for transient
piezoelectric analysis. Eng Anal Boundary Elements 2000;
24:5918.
[22] Daros C-H, Antes H. Dynamic fundamental solutions for
transversely isotropic piezoelectric materials. Int J Solids
Struct 2000;37:163958.
[23] Sollero P, Aliabadi MH. Anisotropic analysis of cracks in
composite laminates using the dual boundary element
method. Composite Struct 1995;31:22933.
[24] Pan E, Amadei B. Fracture mechanics analysis of cracked
2-D anisotropic media with a new formulation of the
boundary element method. Int J Fracture 1996;77:16174.

[25] Pan E, Chen C-S, Amadei B. A BEM formulation for

anisotropic half-plane problems. Eng Anal Boundary
Elements 1997;20:18595.
[26] Pan E. Boundary element analysis of fracture mechanics in
anisotropic bimaterials. Eng Anal Boundary Elements
1999;23:68391.
[27] Suo Z. Singularities, interfaces and cracks in dissimilar
anisotropic media. Proc Roy Soc Lond A 1990;420:
33158.
[28] Barnett DM, Lothe J. Synthesis of the sextic and the
integral formalism for dislocations, Green functions, and
surface waves in anisotropic elastic solids. Phys Norvegica
1973;7:139.
[29] Saez A, Gallego R, Domnguez J. Hypersingular quarterpoint boundary elements for crack problems. Int J Numer
Meth Eng 1995;38:1681701.
[30] Garca F, Saez A, Domnguez J. Traction boundary
elements for cracks in anisotropic solids. Eng Anal
Boundary Elements 2004;28:66776.
[31] Tsamasphyros G, Dimou G. Gauss quadrature rules for
nite part integrals. Int J Numer Meth Eng 1990;30:
1326.
[32] Kutt HR. Quadrature formulae for nite-part integrals.
Special Report WISK 178, National Research Institute for
Mathematical Sciences, Pretoria, South Africa, 1975.
[33] Ladopoulos EG. On the numerical evaluation of the
general type of nite-part singular integrals and integral
equations used in fracture mechanics. Eng Fracture Mech
1988;31:31537.
[34] Sih GC, Paris PC, Irwin GR. On cracks in rectilinearly
anisotropic bodies. Int J Fracture Mech 1965;1:189203.
[35] Xu X-L, Rajapakse RKND. A theoretical study of
branched cracks in piezoelectrics. Acta Mater 2000;48:
186582.
[36] Sung JC, Liou JY. Analysis of a crack embedded in a linear
elastic half-plane solid. ASME J Appl Mech 1995;62:
7886.