Mechanics of Materials 36 (2004) 917–928

www.elsevier.com/locate/mechmat

Fracture criteria of piezoelectric ceramics with defects

a,*
Daining Fang , Zhen-Ke Zhang a, Ai Kah Soh b, Kwok Lun Lee b

a
Department of Engineering Mechanics, Tsinghua University, Beijing 100084, PR China
b
Department of Mechanical Engineering, The University of Hong Kong, Hong Kong, PR China
Received 20 September 2002

Abstract

In this paper, based on a solution of an elliptic cavity in piezoelectric ceramics with uniform loading at infinity, we
analyze the stress and electric fields in the limit cases of the solution when the elliptical cavity tends gradually to a crack.
The result in the zone near the crack tip is different from the result in the zone far away from the crack tip. The
calculation results of a PZT-4 material are presented. Two new fracture criteria are developed for piezoelectric fracture
in this paper. One is based on a generalized stress intensity factor related to the stress field at the tip of the crack and
another is based on the crack opening displacement (COD) related to the relative displacement of crack faces. PZT-5h
ceramics with a center crack were subjected to the combined effect of mechanical load and electrical load in this
experiment to analyze the effect of electric field on fracture. The measured fracture toughness of central crack specimens
of poled soft lead zirconate titanate ceramics (PZT-5) is presented to compare with the theoretical results. Some existing
fracture criteria are examined through the comparison of their calculation results and the experimental results.

2003 Elsevier Ltd. All rights reserved.

Keywords: Piezoelectric ceramic; Crack; Fracture criterion

1. Introduction ments suggest that the effect of the applied electric

field on fracture is complicated and to-date it still
Piezoelectric ceramics are in widespread use as remains unclear. The results obtained from differ-
sensors and actuators in smart structures. A fun- ent experiments can be drastically different, and
damental understanding of their cracking behavior there are no generally acceptable results (Park and
is still not sufficient (Ikeda, 1996; Yang, 2002). The Sun, 1995; Fu and Zhang, 2000; Singh and Wang,
natural extension of the solution of a mechanical 1995; Fang et al., 2001). This complication may
crack failed to characterize even qualitatively the be explained by the sensitivity of the fracture
behavior observed in the laboratory (Park and resistance of ferroelectric ceramics to the defect
Sun, 1995; Fu and Zhang, 2000; Singh and Wang, boundary condition (Fang et al., 2001). The crea-
1995). The discrepancies between these experi- tion of reliable design guidelines for piezoelectric
awaits an accurate fracture criterion (Yang, 2002;
*
Corresponding author. Tel.: +86-10-62772923; fax: +86-10- Fang et al., 2001; Gao et al., 1997).
62783986. Many theoretical results obtained by several
E-mail address: fangdn@mail.tsinghua.edu.cn (D. Fang). researchers (Sosa, 1991; Pak and Tobin, 1993;

0167-6636/$ - see front matter
2003 Elsevier Ltd. All rights reserved.
doi:10.1016/j.mechmat.2003.08.011
918 D. Fang et al. / Mechanics of Materials 36 (2004) 917–928

Park and Sun, 1995; Fulton and Gao, 1997) were the energy density criterion as the fracture crite-
based on the assumption of electric impermeability rion. Zhu and Yang (1997) proposed that the
at the rim of a defect and several fracture criteria main nonlinearity of ferroelectric material is the
were brought out based on them. But most of domain switching. They developed a small-scale
these criteria can only explain some aspects of domain-switching model to explain the toughening
experimental results (Tobin and Pak, 1993; Cao mechanism. The spontaneous strain in the do-
and Evans, 1994; Lynch et al., 1995). Some later main-switching zone causes the restrictive force of
work (Sosa and Khutoryansky, 1996; Chung and the zone without domain switching, which affects
Ting, 1996; Zuo and Sih, 2000) was based on the the stress intensity factor.
use of exact electric boundary conditions at the In this paper, based on a solution of an elliptic
rim of the defect. But these investigations did not cavity in piezoelectric ceramics with uniform
carefully analyze the variation of electromechani- loading at infinity, we analyze the stress and elec-
cal fields when the elliptical cavity gradually tends tric fields in the limit cases of the solution when the
to a crack and some self-contradictory conclusions elliptical cavity tends gradually to a crack. The
were even made. The experimental results show calculation results of a PZT-4 material are pre-
that fracture toughness is anisotropic and the sented. Two new fracture criteria are developed for
direction of the applied electric field can signifi- piezoelectric fracture in this paper. One is based on
cantly affect fracture behavior. Although a lot of a generalized stress intensity factor related to the
theoretical work has also been done in piezoelec- stress field at the tip of the crack and the other is
tric fracture, discrepancies between experimental based on the crack opening displacement (COD)
and analytical results indicate that the established related to the relative displacement of crack faces.
linear piezoelectric fracture theory is insufficient to The measured fracture toughness of central crack
illustrate the fracture behavior of piezoelectric or specimens of poled soft lead zirconate titanate
ferroelectric materials (Gao et al., 1997). ceramics (PZT-5) is presented to analyze the effects
As the fracture criterion, the linear energy re- of an applied electric field. Some existing fracture
lease rate is the first criterion to be used. But this criteria are examined through the comparison
criterion has the fatal disadvantage that it predicts of their calculation results and the experimental
that an electric field should always inhibit crack- results.
ing, which obviously contradicts with the experi-
mental results.
2. Fundamental solution
To avoid this defect, Park and Sun (1995)
introduced the mechanical strain energy release
A plane problem shown in Fig. 1 is considered.
rate as a fracture criterion. This criterion can
An elliptical cavity lies in an infinite and homo-
accurately predict the influence of electric field on
geneous piezoelectric matrix. The medium inside
cracking, but it has a theoretical flaw because it
the cavity is uniform and isotropic and the per-
ignores the electrical term (Gao et al., 1997). As
mittivity of the medium is e0 . The piezoelectric
argued by Gao et al. (1997), the mechanical term,
matrix is subjected to uniform fields at infinity
in fact, does not have an essential difference with
(r1 ; E1 ).
electrical term because they are both induced by
When the linear piezoelectric constitutive
the electromagnetic force.
equations are adopted, the general displacement
Fulton and Gao (1997) considered the material
and the general stress function in the piezoelectric
nonlinearity and introduced an electrical situation
body can be obtained as:
model similar to the DungdaleÕs model in plastic
metal cracking. This model assumed that the fer- u ¼ 2RefAhf1 T 1 T
 iA gg1 þ 2RefAhf iB gh1
roelectric material might electrically yield in the ð1Þ
/ ¼ 2RefBhf1 T 1 T
 iA gg1 þ 2RefBhf iB gh1
Dugdale zone. Thus, they can get the local energy
release rate Gc by means of J -integrals across where g1 , h1 are vectors determined by the load
the electric yield zone. Zuo and Sih (2000) selected and shape of the elliptical cavity, and A, B, S, H, L
 D. Fang et al. / Mechanics of Materials 36 (2004) 917–928 919

e0 =e ! 0, so the boundary conditions are electrical

impermeable. In contrast, j ! 0 corresponds to
the boundary conditions of electrical conducting.
If the piezoelectric body is transversely isotropic
and its poling axis is aligned in the positive direc-
tion of the x2 -axis (see Fig. 1), the uniform electric
field of the medium inside the cavity can be given
as follows:
D01
 ðL1 ST Þ½1;4 r1
12 ¼  E11 ¼ E10  E11
e0
ð4Þ
ðL1 Þ½2;4
 r1
22 þ D02  D1
2 ¼ jD02
Fig. 1. Schematic of an infinite piezoelectric body with an ðL1 Þ½4;4
elliptical cavity.
T
Let t2 ¼ ½r21 ; r22 ; r23 ; D2 , then we have the solu-
tion in the piezoelectric body:
tensors that depend upon material constants. hf1  i T
is a 4 · 4 diagonal matrix related to the coordi- t2 ¼ t1
2 þ 2RefBhf ðf ÞiA gg1
nates. The detailed expression and physical þ 2RefBhf ðf ÞiBT gh1 ð5Þ
meaning of the tensors, A, B, S, H, L, can be
found in Chung and TingÕs work (1996) and are where hf ðf Þi is a 4 · 4 diagonal matrix hf ðf Þi ¼
briefly given in Appendix A. diag hf ðf1 Þ; f ðf2 Þ; f ðf3 Þ; f ðf4 Þi and f ð Þ is a func-
The electric field in the medium with D0 tion
(¼ e0 E0 ) inside the cavity is given by the equations
1
ðh1 Þ½4 ¼ aðE10  E11 Þ f ðfa Þ ¼ ð6Þ
ð2Þ ca f2a þ da
_
ðh 1 Þ½4 ¼ bðE20  E21 Þ
The meanings of ca , da and fa can be found in
where the subscript ([4]) denotes the fourth Appendix A. Assume r  a where r ¼ jza  aj is
element of the vector, and a and b are defined in the distance of a point away from the crack tip.
Fig. 1. At first, we consider the zone where r  b. In
this case, Eq. (5) can be simplified to
2 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
3. Limit cases of the solutions when the cavity tends t2 ¼  pffiffiffiffiffiffiffi RefB diag ½1= cos h þ pa sin h AT gg1
to a crack 2ar
2 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 pffiffiffiffiffiffiffi RefB diag ½1= cos h þ pa sin h BT gh1
Based on the solutions given in Section 2, we 2ar
examine the limit cases of these solutions when the ð7Þ
cavity tends to a crack. To reflect the electrical
boundary conditions, a parameter j is introduced, The field solution along the x1 axis is
i.e. 1
t2 ¼  pffiffiffiffiffiffiffi g1
b 2ar
j¼ ð3Þ pffiffiffiffiffiffi
ae0 ðL1 Þ½4;4 pa 1 1 1 0 T
¼ pffiffiffiffiffiffiffi ½r12 ; r22 ; r23 ; ðD1
2  D2 Þ ð8Þ
where ðL1 Þ½4;4 is a representative variable of 2pr
permittivity of piezoelectric ceramics e. The sub- The stress and electric displacement intensity fac-
script ([4,4]) denotes the corresponding element of tors (KI ; KII ; KIII ; KD ) and the energy release rate
the matrix. For example, if j ! 1, it means G can be obtained formally:
920 D. Fang et al. / Mechanics of Materials 36 (2004) 917–928

pffiffiffiffiffiffi pffiffiffiffiffiffi pffiffiffiffiffiffi that a positive electric field will induce a tensile
KI ¼ r1
22 pa; KII ¼ r1 12 pa; KIII ¼ r1
23 pa
pffiffiffiffiffiffi stress and is favorable to the crack growth. This
KD ¼ ðD1 0
2  D2 Þ pa ð9Þ conclusion accords with Pak and SunÕs experi-
Z d mental result. Because U and V are matrices only
1 depending on the material constants, it has been
G ¼ lim ½ri2 ðrÞDui ðd  rÞ
d!0 2d 0 checked in the calculation that the term in the
þ D2 ðrÞDuðd  rÞ dr square brackets of 2ab ½U½2;2 þ V½2;1 ðL1 ST Þ½1;2 is
1 usually larger than 0.5 for various PZT ceramics.
¼ ½KII ; KI ; KIII ; KD ðL1 Þ½KII ; KI ; KIII ; KD T For example, U½2;2 þ V½2;1 ðL1 ST Þ½1;2 ffi 0:84 for
2
PZT-4 ceramics. Therefore, when a=b  1,
ð10Þ 2a
b
½U½2;2 þ V½2;1 ðL1 ST Þ½1;2 is much more than 1.
As we have mentioned before, however, the solu- Thus, the first term in the right side of Eq. (12) can
tion in Eq. (7) is only valid in the zone where be neglected because a=b is far larger than 1 (i.e.
r  b. If we consider the zone of r  b, the solu- a=b  1) for a crack. Therefore, the stress, r22 , is
tion is different from that in the zone of r  b. It is related to two parts. One results from the remote
more important to consider the latter case (i.e. the mechanical stress, r1 22 and the other corresponds to
zone of r  b) for the crack growth. The limit case that induced by the electric loading, ðD1 0
2  D2 Þ.
of the solution given in Eq. (5) is corresponding to
the solution at the crack tip (r ¼ jza  aj ! 0):
   4. New fracture criteria
2 T
t2 jz¼a ¼ t1
2  Re iBh1=p  iA g1
b 4.1. A generalized stress intensity factor KI
  
T
þ Re iBh1=p iB h1 ð11Þ
For a line crack with a=b  1, it can be con-
cluded from Eq. (12) that when a uniform electric
where h1=p i is a 4 · 4 diagonal matrix h1=p i ¼
field applied at infinity is nonzero, the stress in-
diag h1=p1 ; 1=p2 ; 1=p3 ; 1=p4 i. Notice that both of
  duced by electric field concentrated at the crack tip
Re iBh1=p iAT and Re iBh1=p iBT are 4 · 4 is significantly large even if there is no applied
real matrices, which can be defined as U and V. If uniform stress field at infinity. This stress can be
the piezoelectric body is transversely isotropic and likely to lead to the crack growth, which has been
its poling axis aligned in the positive x2 -direction, approved by the experimental results. However,
we can get the stress at the crack tip: the definition of stress intensity factors cannot
reflect the influence of this stress on the crack
2a
r22 ¼ r1
22 þ f½U½2;2 þ V½2;1 ðL1 ST Þ½1;2 r1
22
growth because the stress induced by electric field
b has features different from those of the stress
1 T
þ ½U½2;4 þ V½2;1 ðL S Þ½1;4 ðD12  D2 Þg
0
ð12Þ resulting from mechanical loading. For example,
from Fig. 3 presented in Section 5, one can find
where the subscript ½i; j denotes the correspond- that when r=a < 106 , the effect of electric field on
ing elements of the matrix. For the piezoelec- the stress near the crack tip is significant. How-
tric ceramics, U½2;2 > 0, U½2;4 > 0, V½2;1 > 0, ever, the conventional stress intensity factor does
ðL1 ST Þ½1;2 > 0 and ðL1 ST Þ½1;4 < 0. Therefore, not take into account this effect. The conventional
2a
b
½U½2;2 þ V½2;1 ðL1 ST Þ½1;2 , the coefficients of r1
22 in definition of stress intensity factors can only reflect
Eq. (12) are larger than zero. The coefficient of the influence of the mechanical stress field, rather
1 T
ðD1 0
2  D2 Þ, i.e. ½U½2;4 þ V½2;1 ðL S Þ½1;4 , is un- than that of the stress field induced by electric field
known and must be determined in terms of the concentrated at the crack tip. In fact, as we show
actual material properties. However, for the PZT-4 in the next section, only when r is less than 0:01b,
or PZT-5 (common piezoelectric ceramics), the is the stress induced by electric field notable. Thus,
coefficient of ðD1 0
2  D2 Þ is positive. This means the calculated conventional stress intensity factor
 D. Fang et al. / Mechanics of Materials 36 (2004) 917–928 921

is independent of electric loading. To take into " 1 T

pffiffiffiffiffiffi U½2;2 þ V½2;1 ðL S Þ½1;2 1
account the influence of electric field on the crack KI ¼ pa r22
growth, it is necessary to set up new criteria for the C
#
crack growth. To this end, we assume two U½2;4 þ V½2;1 ðL1 ST Þ½1;4
assumptions: (1) The driven force of crack growth þ ðD1 0
2  D2 Þ ð19Þ
C
is in proportion to the tensile stress at the crack
tip; (2) the critical stress of crack growth is inde- In fact, this generalized stress intensity factor, KI ,
pendent of the applied loads. Using assumption consists of two terms. One is the mechanical stress
(1), we can write a criterion with reference to intensity factor, as shown in the first term in the
Eq. (12) for a crack: right side of Eq. (19). The second term in the right
2a side of Eq. (19) represents the effect of electric
f½U½2;1 þ V½2;1 ðL1 ST Þ½1;2 r1
22 quantities on the intensity factor.
b
1 T
þ ½U½2;4 þ V½2;1 ðL S Þ½1;4 ðD1 0
2  D2 Þg > rc
4.2. A COD criterion for piezoelectric fracture
ð13Þ
If there is no electric field at infinity, Eq. (13) By another way to develop a new criterion, we
changes to adopt the concept of COD, which often is used in
elastic–plastic fracture mechanics. The common
2a 1
Cr22 > rc ð14Þ definition of COD is the distance of two points of
b intersection of two 45 lines from the crack tip as
where shown in Fig. 2 (Shih, 1981).
We have given the displacement field of the
C ¼ ½U½2;2 þ V½2;1 ðL1 ST Þ½1;2
whole piezoelectric matrix, as given in Eq. (1) of
e33 c11  e31 c13 ðL1 Þ½2;4 Section 1. Based on the COD definition, the dis-
 j tance of two points on the crack faces (see Fig. 2)
c33 c11  c213 ðL1 Þ½4;4
þ is located at a distance being l away from the crack
1þj
tip. Thus, the COD can be obtained as (assume
 ½U½2;4 þ V½2;1 ðL1 ST Þ½1;4 ð15Þ b  l  a):
pffiffiffiffiffiffiffi 0 2
is a coefficient determined by means of material d ¼ 2 2al½ðL1 Þ½2;2 r1 1 1
22 þ ðL Þ½2;4 ðD2  D2 Þ
properties. Notice that the mechanical criterion of
ð20Þ
crack growth is
pffiffiffiffiffiffi From Fig. 2 it can be seen that the COD defined
KI ¼ r122 pa > KIC ð16Þ
should meet the following requirement:
In terms of the second assumption, we can get dt ¼ 2l ð21Þ
2a KIC
C pffiffiffiffiffiffi ¼ rc ð17Þ
b pa

Substituting (17) into (13), we get a criterion:

" 1 T
pffiffiffiffiffiffi U½2;2 þ V½2;1 ðL S Þ½1;2 1
pa r22
C
#
U½2;4 V½2;1 ðL1 ST Þ½1;4
þ ðD1
2  D02 Þ > KIC ð18Þ
C

It can be seen that the left part in Eq. (18) can be

viewed as a generalized stress intensity factor KI Fig. 2. Definition of crack opening displacement (COD).
922 D. Fang et al. / Mechanics of Materials 36 (2004) 917–928

Substituting Eq. (21) into Eq. (20), the COD can In terms of the material constants and the
be expressed as: coordinates, the Barnett–Lothe tensors can be
2 calculated in the consistent units of m, N, C,
dt ¼ 4a½ðL1 Þ½2;2 r1 1 1 0
22 þ ðL Þ½2;4 ðD2  D2 Þ ð22Þ
and V, i.e.
The fracture criterion is given by 2 3
0 0:1237 0 0:2458  109
dt P dc ð23Þ 6 0:4318 0 0 0 7
S¼6 4
7
5
If we define a critical intensity factor KC as 0 0 0 0
10
pffiffiffiffiffiffiffi 0:1067  10 0 0 0
pdc
KC ¼ ð24Þ 2 3
2ðL1 Þ½2;2 0:21311010 0 0 0
6 0 0:142310 10
0 0:03027 7
H¼ 6
4
7
5
where KC is obviously identical to the critical 0 0 0:35731010 0
COD, dC , and also dependent upon the constants 0 0:03027 0 0:1077109
of materials, the criterion, Eq. (23) can be rewrit- 2 3
0:56731011 0 0 0
ten as: 6 0 0:432710 11
0 10:93 7
L¼ 6
4
7
5
0 0 0:27991011 0
ðL1 Þ½2;4 0 10:93 0 0:8646108
KI þ KD P KC ð25Þ
ðL1 Þ½2;2
and
2 3
where KI and KD are stress and electric displace- 0:06186 0 0 0
ment intensity factors, respectively. For piezo- 6 0 0:6713 0 0:4501109 7
electric ceramics, the coefficient of KD in Eq. (25), U¼6
4 0
7
5
0 0:4573 0
ðL1 Þ 10
i.e. ðL1 Þ½2;4 , is positive. This means that the positive 0 0:474810 0 0:5520
½2;2

electric field will induce a tensile stress, and hence 2 3

0 0:2163  1011 0 5:466
enhance the crack growth, which accords with Pak 6 0:2163  1011 0 0 0 7
and SunÕs experimental result. In addition, one V¼6
4
7
0 0 0 0 5
purpose of rewriting Eq. (23) in the form of an
5:466 0 0 0
intensity factor is to more conveniently compare
it with other fracture parameters (see Section 7). Thus, the generalized stress intensity factor KI is
determined by
 

pffiffiffiffiffiffi 1 j 1
5. Numerical results KI ¼ pa r22 þ 0:5111 E ð26Þ
k1 ð1 þ jÞ 2
3
In order to demonstrate the stress fields near an where k1 ¼ 1 þ 0:688610 ffi 1.
1þj
elliptical cavity or a crack in piezoelectric ceram- Next some calculation examples are given. As
ics, the field distributions of a PZT-4 material are shown in Fig. 1, the poling axis is aligned in the
calculated. The constants for PZT-4 are listed positive x2 -direction. Let b=a ¼ 0:001 so that the
below: elliptical cavity tends to a line crack. Two cases are
c11 ¼ 13:9  1010 N=m2 e31 ¼ 6:98 C=m2 considered, case I and case II. In case I, the electric
load at infinity is equal to zero but a stress r1 22
c12 ¼ 7:78  1010 N=m2 e33 ¼ 13:8 C=m2 is applied at infinity.
c13 ¼ 7:43  1010 N=m2 e15 ¼ 13:4 C=m2 In this case, the conventional fracture theory
c33 ¼ 11:3  1010 N=m2 e11 ¼ 6:00  109 C=V m gives a r1=2 singularity of the mechanical stress,
i.e. r22 / r1=2 , which is valid only in a round re-
c44 ¼ 2:56  1010 N=m2 e33 ¼ 5:47  109 C=V m gion (b  r  a). The calculated stress in terms of
1 Eq. (5) is shown in Fig. 3 with logarithmic axes.
c66 ¼ ðc11  c12 Þ ¼ 3:06  1010 N=m2
2 Fig. 3 shows clearly that the difference between
 D. Fang et al. / Mechanics of Materials 36 (2004) 917–928 923

6. Experimental results

PZT-5 ceramics with a center crack were sub-

jected to the combined effect of mechanical load
and electrical load in this investigation. The
material has a tetragonal crystal structure at room
temperature and an average grain size of 3 lm.
The measured material constants and properties of
PZT-5 are listed in Table 1. The dimensions of the
specimens were 40 · 20 · 3 mm. The top and bot-
tom surfaces (area 40 · 3 mm) of the specimens
were coated with silver paint to be used as elec-
Fig. 3. Stress field along the x1 -axis in PZT-4 with a crack trodes. Fig. 5 shows the experimental setup. The
subjected to a uniform far stress.
specimens were poled at 300 C in the direction
perpendicular to the crack prior to cutting the
crack. The specimens were poled under an electric
field of 400 V/mm in the oven. The coercive field of
PZT-5 is about 750 V/mm. Each of the specimens
tested had a central crack, which was manufac-
tured by employing the ultrasonic pulse technique
using a moving blade of 0.1 mm-thick. The basic
setup of the ultrasonic machine is shown in Fig. 5.
Note that ultrasonic lapping is a technique suitable
for machining any brittle material. By using a
profile tool in longitudinal swing mode and at high
frequency, of about 20–23 kHz, lapping grains can
Fig. 4. Stress field along the x1 -axis in PZT-4 with a crack make a crack with a width of less than 100 lm. In
under coupled electromechanical loading. our experiments, boron carbide was used as lap-
ping medium to achieve high removal rate. A high
them is significant especially for ranges at two ends frequency generator produces an alternating elec-
of the curves. For the line crack that contains a trical voltage, which is transformed by the sonic
medium with the permittivity e0 ðe0 6¼ e 6¼ 0Þ, the transformer into mechanical vibration energy
field quantities calculated from Eq. (5) remain of the same frequency. The process has been
bounded. In case II, r1 22 > 0 but the different
electric fields with zero, positive and negative val-
ues are applied. In this case, a ratio of the electric Table 1
The measured material constants and properties of PZT-5
load to the mechanical loading is defined as
E21 =r1 2 Curie temperature, Tc (C) 320
22 ¼ 1 m /C. From Eq. (26) it can be found
that a larger negative electric load can make the Relative dielectric constant, eT33 =e0 2100
generalized stress intensity factor KI equal to zero. eT11 =e0 2400
That is, under the coupled electromechanical
loading, the loading ratio that leads to KI ¼ 0 Dissipation factor (tg d) 6 0.02
should be E21 =r1 2
22 ¼ 3:43 m /C. Fig. 4 shows the
3
Density, q (10 kg/m ) 3
P 7.45
stress field along the x1 -axis in the PZT-4 ceramics c11 ¼ 12:6 c12 ¼ 5:5 c13 ¼ 5:3 c33 ¼ 11:7 c44 ¼ 3:53
with a crack under coupled electromechanical e31 ¼ 6:5 e33 ¼ 23:3 e15 ¼ 17
loading. It can be easily seen that a positive electric e11 ¼ 15:1 e33 ¼ 13:0
field promotes the crack propagation while a Note: in the table, the units of cij are 1010 N/m2 , eij are C/m2 , eij
negative electric field inhibits the crack growth. are 109 C/V m.
924 D. Fang et al. / Mechanics of Materials 36 (2004) 917–928

Fig. 5. Schematic illustration of the experimental setup for fracture test.

described in detail by Kainth et al. (1979) and, field increases it. A least-square curve fitting is
therefore, it will not be reiterated here. The ultra- performed on the measured fracture data pre-
sonic technique permits complex forms to be sented in Fig. 6 and its form is given by
produced from all engineering ceramics, regard-
KI ¼ 803505:949  18:193KE þ 2:641  104 KE2
less of their electrical conductivity.
In all the tests, the electric fields were applied to ð27Þ
the electrode surfaces. The tensile load was applied where KE is the electrical fieldpffiffiffiffiffiffiffi
intensity factor,
continuously using the crosshead displacement which is defined by KE ¼ limr!0 2prE2 jh¼0 similar
control of the mechanical testing machine (Mini- to the definition of the stress intensity factor. KI
ature Materials Tester), while the applied electric and KE are expressed in terms of Pa m1=2 and V/
field was set at a specific value till fracture oc- m1=2 . Eq. (27) gives a nonlinear function relation
curred. The electric fields were generated by a between KI and KE for an impermeable crack un-
power supply, which has a maximum rating of 50 der combined electromechanical loads. It is inter-
kV DC. The effect of applied electric field on the esting to note that the fitted fracture curve shown
critical fracture load is shown in Fig. 6. It can be in Fig. 7 is similar to that of Heyer et al. (1998) for
clearly seen that the positive electric field decreases the conducting crack, except that in our case the
the critical fracture load while the negative electric curve is concave instead of convex.

Fig. 6. The effect of applied electric field, E, on the critical Fig. 7. The relation between stress and electric-field intensity
fracture load, F . factors fitted from the experimental data.
 D. Fang et al. / Mechanics of Materials 36 (2004) 917–928 925

7. Comparison of fracture criteria where A000 , B000 are coefficients of materials, respec-
tively. The new criterion related to COD given
As the fracture criterion, the linear energy re- in Eq. (23) can be rewritten as follows:
lease rate can be expressed as (Yang, 2002; Park 0 2
dt ¼ 4a½ðL1 Þ½2;2 r1 1 1
22 þ ðL Þ½2;4 ðD2  D2 Þ
and Sun, 1995; Gao et al., 1997; Sosa, 1991; Pak
0000 0000 0000
and Tobin, 1993): ¼ A KI2 þ B KI KD þ C KD2 ð33Þ
Z d
1 0000 0000 0000
where A , B , C are coefficients of materials,
Glin ¼ lim ½ri3 ðx1 ÞDui ðd  x1 Þ
d!0 2d 0 respectively. Note that the stress intensity factor,
þ D3 ðx1 ÞD/ðd  x1 Þ dx1 ð28Þ KI , in Eqs. (29)–(33) is defined in Eq. (9), but the
where d is the distance of crack growth. For the I- electric intensity factor,
pffiffiffiffiffiffi KD , in Eqs. (29)–(31) is
type piezoelectric fracture problem under com- defined by KD ¼ D1 2 pa, which is different from
bined mechanical and electrical loading, Eq. (28) the electric intensity factor, KD in Eqs. (32),p(33)ffiffiffiffiffiffi
can be rewritten in the form of intensity factor as: that is given as in Eq. (9) as KD ¼ ðD1 0
2  D2 Þ pa.
According to the energy density criterion (Zuo
Glin ¼ AKI2 þ BKI KD  CKD2 ð29Þ and Sih, 2000), the unstable fracture is assumed to
where A, B, C are coefficients of materials, take place when
respectively.  
dW Sc
The mechanical strain energy release rate has ¼ ð34Þ
dV c rc
been given by Park and Sun (1995) as:
Z d
1 where W is the energy in the representative cell, V
Gm ¼ lim ri3 ðx1 ÞDUi ðd  x1 Þ dx1 the volume of the cell, rc the critical ligament
d!0 2d 0
ahead of crack, and Sc the critical value of the
¼ A0 KI2 þ B0 KI KD ð30Þ energy density factor. The crack is assumed grow
0 0
where A , B are coefficients of materials, respec- at the direction:
tively. Fulton and Gao (1997) gave the local en-
ergy release rate as: oS
" ¼0
oh
1 ð35Þ
Gc ¼ ðB1 Þ½2;2 KI2 þ 2ðB1 Þ½2;4 KI KD o2 S
8p >0
# oh2
ðB1 Þ½4;4 B½4;4  1 2 where S is the energy density factor, and h the
þ KD
B½4;4 angle between the real cracking direction and the
¼ A00 KI2 þ B00 KI KD þ C 00 KD2 ð31Þ direction of the original crack face.
To prove the validation of the existing fracture
where B is a 4 · 4 real, symmetric, invertible matrix criteria, a comparison of the theoretical results and
depending on the material constants and given by the experimental results is made. Fig. 8 shows the
Fulton and Gao (1997). A00 , B00 , C 00 are coefficients comparison of different criteria with our experi-
of materials, respectively. mental results. Note that the theoretical curve is
The new criterion related to the generalized obtained based on the first group of experimental
stress intensity factor KI as expressed in Eq. (19) results. The difference of the calculation results in
can be rewritten as: Fig. 8(a) and (b) is the boundary condition. That
" 1 T is, the comparison of existing fracture criteria and

pffiffiffiffiffiffi U½2;2 þ V½2;1 ðL S Þ½1;2 1
KI ¼ pa r22 experimental results includes: (a) Impermeable
C electric boundary conditions; (b) exact electric
#
U½2;4 þ V½2;1 ðL1 ST Þ½1;4 1 boundary conditions (e0 ¼ 8:85  1012 C V/m).
þ ðD2  D02 Þ From Fig. 8, it can be seen that the linear
C
energy release rate Glin cannot predict the effect
¼ A000 KI þ B000 KD ð32Þ of electric field on the critical fracture load. The
926 D. Fang et al. / Mechanics of Materials 36 (2004) 917–928

is interesting that for the impermeable electric

boundary conditions, the predicted curve based on
COD is completely overlapped with the curve
calculated from the local energy release rate Gc ,
but for the exact electric boundary conditions
there exists a difference between two criteria. It
also can be found from Fig. 8(a) and (b) that the
influence of electric field on cracking under the
exact electrical boundary conditions is smaller
than that under the impermeable electric boundary
conditions.

8. Conclusions

Based on a solution of an elliptic cavity in pie-

zoelectric ceramics with uniform loading at infin-
ity, we analyze the stress and electric fields in the
limit cases of the solution when the elliptical cavity
tends gradually to a crack. The result in the zone
far away from the crack tip is different from the
result in the zone very close to the crack tip. The
former is consistent with the results obtained by
other researchers and the latter is consistent with
the experimental results. By introducing a gen-
eralized stress intensity factor that can take into
Fig. 8. Comparison of fracture criteria and experimental re- account the effect of electric field on cracking, a
sults: (a) Impermeable electric boundary conditions; (b) exact
electric boundary conditions (e0 ¼ 8:85  1012 C/V m).
new fracture criterion is developed based on the
stress field at the tip of the crack.
The conventional fracture theory gives a r1=2
energy-density-factor criterion cannot describe the singularity of the mechanical stress, i.e. r22 / r1=2 ,
influence of electric field either. This criterion can which is valid only in a round region (b  r  a)
qualitatively predict the tendency of the electric- while the calculated stress in terms of Eq. (5) for the
field effect on the fracture load only when the line crack that contains a medium with the per-
electric field is very small (say less than ±1.2 kV/ mittivity e0 ðe0 6¼ e 6¼ 0Þ is bounded at the crack tip.
cm). When the electric field is a little bit larger, this The calculation results show clearly that the dif-
criterion may lead some abnormal conclusions. It ference between them is significant especially for
can be found that the new criterion related to the ranges at two ends of the curves. The calculation
generalized stress intensity factor KI can qualita- results indicate that the influence of electric field on
tively predict the tendency of the electric-field ef- cracking under the exact electrical boundary con-
fect on the fracture load in a quite larger range of ditions is smaller than that under the impermeable
electric field though the prediction based on this electric boundary conditions. To prove the vali-
criterion is not consistent with the experimental dation of the existing fracture criteria, a compari-
data. However, the calculated critical fracture son of the theoretical results and the experimental
loads based on the new criterion related to COD results is made. It can be found that the calculated
are in good agreement with the experimental re- critical fracture loads based on the new criterion
sults, so are the mechanical strain energy release related to COD is in good agreement with the
rate Gm , and the local energy release rate Gc . It experimental results, so are the mechanical strain
 D. Fang et al. / Mechanics of Materials 36 (2004) 917–928 927

energy release rate Gm , and the local energy release the eigenequation of the problem can be written
rate Gc . It is interesting that for the impermeable as:
electric boundary conditions, the predicted curve
based on COD is completely overlapped with the ½Q þ pðR þ RT Þ þ p2 T a ¼ 0 ðA:4Þ
curve calculated from the local energy release rate
Gc , but for the exact electric boundary conditions The eigenvalues, p, of (A.4) are all complex and
there exists a difference between two criteria. The consist of four pairs of complex conjugates. Let
total energy release rate is in total disagreement pk ; ak ðk ¼ 1; 8Þ be eigenvalues and corresponding
with the experimental data, while the energy-den- eigenvectors, then
sity-factor criterion can qualitatively explain the
influence of electric field on fracture only when the pkþ4 ¼ pk ; Im pk > 0 ðk ¼ 1–4Þ
ðA:5Þ
electric field is very small. akþ4 ¼ ak ; bkþ4 ¼ bk

where
Acknowledgements
1
b ¼ ðRT þ pTÞa ¼  ðQ þ pRÞa ðA:6Þ
Support from the National Science Foundation p
of China under grants #10025209 and #10132010,
and from the Research Grants Council of the A and B are 4 · 4 matrices given by
Hong Kong Special Administrative Region, China A ¼ ½a1 ; a2 ; a3 ; a4 ; B ¼ ½b1 ; b2 ; b3 ; b4 ðA:7Þ
(Project No. HKU 7122/99E) is acknowledged.
The three matrices S, H, L defined by
S ¼ ið2ABT  IÞ; H ¼ i2AAT ; L ¼ i2BBT
Appendix A
ðA:8Þ
The detailed expressions and physical meanings are real. The matrices H and L are symmetric and
of symbols in Eq. (1) can be found in Chung and nonsingular.
TingÕs work (1996). We will give the brief expres- The three matrices S, H, L can be obtained
sions of them in this Appendix A. directly by integral formalism. Let the tensor EiJKm
The constitutive equations for piezoelectric be defined by
materials can be given as:
rij ¼ Cijkm uk;m þ emij u;m EiJKm ¼ Cijkm ðJ; K ¼ 1; 2; 3Þ
ðA:1Þ ¼ emij ðJ ¼ 1; 2; 3; K ¼ 4Þ
Di ¼ eikm uk;m  xim u;m ði; j; k; m ¼ 1; 2; 3Þ ðA:9Þ
¼ eikm ðJ ¼ 4; K ¼ 1; 2; 3Þ
where the coefficients Cijkm , emij , xim are, respec-
¼ xim ðJ ¼ K ¼ 4Þ
tively, the elastic constants, piezoelectric con-
stants, and permittivity. By defining with vectors
" # " #
QE e11 RE e21 nT ðxÞ ¼ ½cos x; sin x; 0
Q¼ ; R¼ T ;
eT11 x11 e12 x12 ðA:10Þ
" # ðA:2Þ mT ðxÞ ¼ ½ sin x; cos x; 0
TE e22
T¼ T in which x is a real parameter ranging from 0 to
e22 x22 2p. We let
where QJK ðxÞ ¼ ni ðxÞEiJKm nm ðxÞ
E E
ðQ Þik ¼ Ci1k1 ; ðR Þik ¼ Ci1k2 ; RJK ðxÞ ¼ ni ðxÞEiJKm mm ðxÞ ðA:11Þ
E
ðA:3Þ
ðT Þik ¼ Ci2k2 ; ðeij Þm ¼ eijm TJK ðxÞ ¼ mi ðxÞEiJKm mm ðxÞ
928 D. Fang et al. / Mechanics of Materials 36 (2004) 917–928

and Chung, M.Y., Ting, T.C.T., 1996. Piezoelectric solid with an

elliptic inclusion or cavity. Int. J. Solids Struct. 33, 3343–
N1 ðxÞ ¼ T1 ðxÞRT ðxÞ 3361.
Fang, D.N., Soh, A.K., Liu, J.X., 2001. Advances in studies
N2 ðxÞ ¼ T1 ðxÞ ðA:12Þ of electromechanical deformation and fracture of piezoelec-
1 T tric/ferroelectric materials. Acta Mech. Sinica 17 (3), 193–
N3 ðxÞ ¼ RðxÞT ðxÞR ðxÞ  QðxÞ
213.
then Fu, R., Zhang, T.Y., 2000. Effects of an electric field on the
Z fracture toughness of poled lead zirconate titanate ceramics.
1 p
S¼ N1 ðxÞ dx; J. Am. Ceram. Soc. 83, 1215–1218.
p 0 Fulton, C., Gao, H.J., 1997. Electrical nonlinearity in fracture
Z p
1 of piezoelectric ceramics. Appl. Mech. Rev. 50 (2), 56–
H¼ N2 ðxÞ dx; ðA:13Þ 62.
p 0
Z Gao, H., Zhang, T.Y., Tong, P., 1997. Local and global energy
1 p release rates for an electrically yielded crack in a piezoelec-
L ¼ N3 ðxÞ dx
p 0 tric ceramic. J. Mech. Phys. Solids 45, 491–510.
Heyer, V., Schneider, G.A., Balke, H., Drescher, J., Bahr, H.A.,
Eqs. (A.13) provide an alternative to (A.8) for the
1998. A fracture criterion for conducting cracks in homo-
Barnett–Lotte tensors S, H and L. geneously poled piezoelectric PZT-PIC 151 ceramics. Acta
The fa are defined as: Mater. 46, 6615–6622.
Ikeda, T., 1996. Fundamentals of Piezoelectricity. Oxford
za ¼ ca fa þ da f1
a ða ¼ 1; 2; 3; 4Þ ðA:14Þ Science Publications, Oxford University Press, Oxford.
Kainth, G.S., Nandy, A., Singh, K., 1979. On the mechanics of
where za ¼ x1 þ pa x2 and
material removal in ultrasonic machining. Int. J. Mach.
a  ipa b a þ ipa b Tool Des. 19, 33–41.
ca ¼ ; da ¼ ðA:15Þ Lynch, C.S., Yang, W., Collier, L., Suo, Z., McMeeking, R.M.,
2 2
1995. Electric field induced cracking in ferroelectric ceram-
in which a and b are geometrical dimensions of the ics. Ferroelectrics 166, 11–30.
cavity as shown in Fig. 1, pa are the eigenvalues Pak, Y.E., Tobin, A., 1993. On electric field effects in fracture of
of Eq. (A.4). piezoelectric materials. In: Mechanics of Electromagnetic
Materials and Structures, AMD 161/MD 42. ASME.
g1 , h1 are vectors determined by the load and Park, S., Sun, C.T., 1995. Fracture criteria of piezoelectric
shape of the elliptical cavity, and can be expressed ceramics. J. Am. Ceram. Soc. 78, 1475–1480.
as: Shih, C.F., 1981. Relationship between the J -integral and crack
opening displacement for stationary and extending cracks.
g1 ¼ aðt02  t1
2 Þ J. Mech. Phys. Solids 29, 305–326.
_ ðA:16Þ
g1 ¼ bðt01  t1
1 Þ
Singh, R.N., Wang, H., 1995. Crack propagation in piezoelec-
tric materials under combined mechanical and electrical
and loadings: An experimental study. In: Carman, G.P., Lynch,
_
C.S., Sottos, N.R. (Eds.), Proceedings of AMD-206/MD-58,
h1 ¼ L1 ðST g1  g 1 Þ Adaptive Materials Systems. ASME, pp. 85–95.
_ _
ðA:17Þ Sosa, H., 1991. Plane problems in piezoelectric media with
h 1 ¼ L1 ðg1 þ ST g 1 Þ defects. Int. J. Solids Struct. 28, 491–505.
Sosa, H., Khutoryansky, N., 1996. New developments con-
where t01 ¼ ð0; 0; 0; D01 ÞT , t02 ¼ ð0; 0; 0; D02 ÞT , t1 1 ¼ cerning piezoelectric materials with defects. Int. J. Solids
1 T 1 1 T 0
ðr1
11 ; r1
12 ; r 1
13 ; D1 Þ , t 2 ¼ ðr 1
21 ; r 1
22 ; r1
23 ; D 2 Þ . D ¼ Struct. 33, 3399–3414.
0 0 T Tobin, A., Pak, Y.E., 1993. Effects of electric fields on fracture
ðD1 ; D2 Þ is the electric displacement of the med-
behavior of PZT ceramics. In: Varadan, V.K. (Ed.), Smart
ium inside the cavity.
Materials, Proceedings of SPIE. vol. 1916, pp. 78–86.
Yang, W., 2002. Mechatronic Reliability. Springer, Beijing.
Zhu, T., Yang, W., 1997. Toughness variation of ferroelectrics
References by polarization switch under nonuniform electric field.
Acta. Mater. 45 (11), 4695–4702.
Cao, H.C., Evans, A.G., 1994. Electric-field-induced fatigue Zuo, J.Z., Sih, G.C., 2000. Energy density theory formulation
crack growth in piezoelectric ceramics. J. Am. Ceram. Soc. and interpretation of cracking behavior for piezoelectric
77, 1783–1786. ceramics. Theor. Appl. Fract. Mech. 34, 17–33.