Engineering Fracture Mechanics 73 (2006) 2802–2828

www.elsevier.com/locate/engfracmech

Thermal fracture analysis of orthotropic functionally

graded materials using an equivalent domain integral approach
Serkan Dag *

Department of Mechanical Engineering, Middle East Technical University, Ankara 06531, Turkey

Received 2 October 2005; received in revised form 24 April 2006; accepted 24 April 2006
Available online 15 June 2006

Abstract

A new computational method based on the equivalent domain integral (EDI) is developed for mode I fracture analysis
of orthotropic functionally graded materials (FGMs) subjected to thermal stresses. By using the constitutive relations of
plane orthotropic thermoelasticity, generalized definition of the J-integral is converted to an equivalent domain integral to
calculate the thermal stress intensity factor. In the formulation of the EDI approach, all the required thermomechanical
properties are assumed to have continuous spatial variations through the functionally graded medium. Developed meth-
odology is integrated into a fracture mechanics research finite element code FRAC2D using graded finite elements that
possess cubic interpolation. Steady-state and transient temperature distribution profiles in orthotropic FGMs are com-
puted using the finite elements based heat transfer analysis software HEAT2D. EDI method is validated and domain inde-
pendence is demonstrated by comparing the numerical results obtained using EDI to those calculated by an enriched finite
element method and to those available in the literature. Single and periodic edge crack problems in orthotropic FGMs are
examined in order to study the influences of principal thermal expansion coefficient and thermal conductivity components,
relative crack length and crack periodicity on the thermal stress intensity factors. Numerical results show that among the
three principal thermal expansion coefficient components, the in-plane component perpendicular to the crack axis has the
most significant influence on the mode I stress intensity factor. Gradation profile of the thermal expansion coefficient par-
allel to the crack axis is shown to have no effect on the outcome of the fracture analysis.
Ó 2006 Elsevier Ltd. All rights reserved.

Keywords: Orthotropic functionally graded materials; Equivalent domain integral; Thermal stresses; Finite element analysis; Stress
intensity factor

1. Introduction

Functionally graded materials (FGMs) are multiphase composites that are characterized by continuous
spatial variations in the volume fractions of the constituents. These variations result in a nonhomogeneous
structure with smoothly varying thermal and mechanical properties. Among the potential application areas

*
Tel.: +90 312 2102580; fax: +90 312 2102536.
E-mail address: [email protected]

0013-7944/$ - see front matter Ó 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engfracmech.2006.04.015
 S. Dag / Engineering Fracture Mechanics 73 (2006) 2802–2828 2803

of functionally graded materials we can mention thermal barrier coatings [1], wear and impact resistant com-
ponents [2], energy conversion devices [3], biomedical materials [4] and eco-materials [5]. Some of the process-
ing methods that are used to develop FGMs are known to induce an oriented structure in addition to the
macroscopic nonhomogeneity. For example, functionally graded materials processed by the plasma spray
technique have a lamellar structure with weak cleavage planes parallel to the boundary [6]. Electron beam
physical vapor deposition technique on the other hand induces a columnar structure with weak fracture planes
perpendicular to the free surface [7]. Cracking of FGM coatings processed by these methods, under thermal
loads is one of the problems encountered in experimental studies. The cracks that are induced due to cyclic
thermal loading are generally observed to be either at the surface and parallel to the material property grada-
tion direction or at and near the interface and parallel to the free surface [8,9]. This type of cracking behavior
is a result of the oriented microstructure of the FGM layers.
In order to take the material orientation into account in the fracture analyses, graded materials are gener-
ally modeled as orthotropic with principal directions parallel and perpendicular to the boundaries. Crack
problems in inhomogeneous orthotropic media are studied by Ozturk and Erdogan [10,11] using singular inte-
gral equations. An analytical solution to compute the stress intensity factors and phase angles for an ortho-
tropic and functionally graded four point bend specimen is developed by Gu and Asaro [12]. Mode I crack
problems in a finite width graded orthotropic strip are considered by Guo et al. [13] under static loading
and by Chen et al. [14] and Guo et al. [15] under dynamic loading. Dag et al. [16] studied interface crack prob-
lems in graded orthotropic media using both the singular integral equation and enriched finite element tech-
niques. Kim and Paulino extended various finite elements based approaches for fracture mechanics analysis of
orthotropic functionally graded materials such as modified crack closure method [17], mixed-mode J-integral
[18] and interaction integral [19,20].
In all the studies mentioned in the above paragraph fracture mechanics analyses of orthotropic function-
ally graded materials are carried out under mechanical loading. Since graded coatings or interlayers could be
subjected to thermal loads in various applications, computational methods to calculate thermal stresses and
related fracture mechanics parameters in orthotropic FGMs are also required. The present study is directed
towards developing a computational technique based on the equivalent domain integral (EDI) to study crack
problems in orthotropic functionally graded materials subjected to thermal loads. In the equivalent domain
integral technique, J-integral [21] is converted to a domain integral [22] around the crack tip to compute the
energy release rate for a given loading condition. The J-integral is widely used in computational fracture
mechanics analyses in order to evaluate the stress intensity factors. In its generalized form, J is defined over
a vanishingly small curve at a crack tip. This definition is known to be equivalent to the energy release rate
for linear or nonlinear elastic materials. J can also be defined as a line integral over an arbitrary curve away
from the crack tip under certain conditions. For example, if the crack is in a homogeneous medium subjected
to mechanical loads and has traction free surfaces, J can be expressed as a path independent line integral
[21].
If the crack is located in a functionally graded medium or subjected to thermal stresses, the generalized def-
inition of J cannot be converted to a single path independent line integral. However, in these cases, it is
possible to reduce J to a form that consists of line and area integrals [23,24]. Equivalent domain integral
approach is a systematic way of converting the generalized line integral definition of J to a domain indepen-
dent form that contains only area integrals. By using this method, domain independent integrals can be
derived for nonhomogeneous materials such as FGMs and for materials subjected to thermal stresses. EDI
technique is particularly suitable to be implemented by means of the finite element method. Once the displace-
ment, strain and stress fields are computed, area integrals can be calculated using the isoparametric finite ele-
ment concept. Another advantage of the EDI method is that, even without the use of quarter point singular
elements at the crack tip, it is possible to obtain highly accurate results provided that there is sufficient degree
of mesh refinement in the integral domain. The EDI concept is previously explored to study crack problems in
both isotropic [25–27] and orthotropic [18] FGMs subjected to mechanical loads.
In the present study, the equivalent domain integral is formulated for orthotropic functionally graded mate-
rials that are subjected to mode I thermal stresses. The generalized definition of the J-integral is converted to
an area integral by carrying out the necessary modifications resulting from material property gradation and
thermal strains. The developed EDI formulation is integrated into a general purpose fracture analysis code
2804 S. Dag / Engineering Fracture Mechanics 73 (2006) 2802–2828

FRAC2D in order to calculate thermal stress intensity factors for orthotropic functionally graded materials.
FRAC2D is originally developed by the General Electric Company [28] and later the fully accessible source
code is extended with graded finite elements in order to calculate the stress and displacement fields in function-
ally graded materials under thermomechanical loading [16,29]. In the present work, temperature fields in
orthotropic FGMs are computed using the finite elements based heat transfer analysis software HEAT2D
[30]. In the finite element models, graded elements with cubic interpolation are used in order to achieve a high
degree of accuracy in the computed results. Numerical results are presented for single and periodic edge cracks
subjected to steady state and transient thermal loads. In order to examine the accuracy of the developed
method, the results computed using the EDI technique are compared to those available in the literature
and to those obtained by means of an enriched finite element method. The use of enriched finite elements
for the solution of crack problems in orthotropic FGMs is described by Dag et al. [16]. It is shown that
the developed EDI procedure yields highly accurate results. Numerical analyses also show that calculated
results are independent of the domain used to compute the stress intensity factor.
The paper is organized as follows: In Section 2, formulation and finite element implementation of EDI for
orthotropic FGMs are described. Edge crack problems in both orthotropic and isotropic FGM layers under
steady-state thermal stresses are considered in Section 3 in order to validate the developed method. Domain
independence is demonstrated in this section by comparing the stress intensity factors calculated using differ-
ent domain sizes to those obtained by enriched finite elements and to those available in the literature. Section 4
presents the analysis of elastic response of periodic cracks in orthotropic FGMs under transient thermal load-
ing. Finally, concluding remarks and a brief discussion are given in Section 5.

2. Equivalent domain integral for orthotropic FGMs under thermal stresses

2.1. Constitutive relations and the mechanical strain energy density function

In this section, we first give the constitutive relations used to model the behavior of orthotropic FGMs
under thermal loads. Assuming that x1 and x2 are the principal axes of orthotropy, the relationship between
the strain and stress components in the principal coordinate system can be written as follows for plane
stress:
2 3 2 32 3 2 3
e11 1=E1 m12 =E1 0 r11 a1 DT
6 7 6 76 7 6 7
4 e22 5 ¼ 4 m12 =E1 1=E2 0 54 r22 5 þ 4 a2 DT 5; ð1Þ
2e12 0 0 1=G12 r12 0

where eij (i, j = 1, 2) are total strain components, Ei, mij and Gij (i, j = 1, 2) are engineering constants of plane
orthotropic elasticity, a1 and a2 are thermal expansion coefficients in directions x1 and x2, respectively and
DT is the temperature change from a reference temperature. We can express DT as
DT ¼ T  T 0 : ð2Þ
In this equation, T is the temperature of the point under consideration and T0 is the temperature of the stress-
free state. For plane strain, the constitutive relationship is given by
2 3 2 32 3 2 3
e11 ð1  m31 m13 Þ=E1 ðm12 þ m13 m32 Þ=E1 0 r11 ðm31 a3 þ a1 ÞDT
6 7 6 76 7 6 7
4 e22 5 ¼ 4 ðm12 þ m13 m32 Þ=E1 ð1  m23 m32 Þ=E2 0 54 r22 5 þ 4 ðm32 a3 þ a2 ÞDT 5: ð3Þ
2e12 0 0 1=G12 r12 0
The engineering constants of orthotropic materials are related as
m21 m12 m31 m13 m32 m23
¼ ; ¼ ; ¼ : ð4Þ
E2 E1 E3 E1 E3 E2
Notice that in Eq. (3), in addition to a1 and a2, thermal expansion coefficient in the third principal direction,
i.e. a3, is also required. In general, in a functionally graded medium the thermomechanical parameters used in
Eqs. (1) and (3) could be functions of the principal coordinates x1 and x2. It will be shown in Section 2.2 that,
 S. Dag / Engineering Fracture Mechanics 73 (2006) 2802–2828 2805

one of the variables required in the formulation of the equivalent domain integral for thermal stress problems
is the mechanical strain energy density function. In the case of a general 3-D state of stress, the mechanical
strain energy density function can be written as
1
W ¼ rij em
ij ði; j ¼ 1; 2; 3Þ: ð5Þ
2
where emij is the mechanical strain. This expression reduces to the following form for plane stress and plane
strain:
(1 
2
r11 em m m m
11 þ r12 e12 þ r21 e21 þ r22 e22 ; for plane stress;
W ¼   ð6Þ
1
2
r11 em m m m m
11 þ r12 e12 þ r21 e21 þ r22 e22 þ r33 e33 ; for plane strain:

The mechanical strains in Eqs. (5) and (6) can be expressed as differences of the total strain and thermal strain
as follows:
em
11 ¼ e11  a1 DT ; em
22 ¼ e22  a2 DT ; em
33 ¼ a3 DT ; em
12 ¼ e12 ; em
21 ¼ e21 : ð7Þ

Here, we note that e33 = 0 for plane strain and r33 = 0 for plane stress. In the case of plane strain r33 is given
as
E3 E3
r33 ¼ m13 r11 þ m23 r22  E3 a3 DT : ð8Þ
E1 E2
Using Eq. (6), mechanical strain energy density function is obtained in the following form for plane stress:
E21 ðe11  a1 DT Þ þ m12 E1 E2 ðe22  a2 DT Þ  
W ¼ ðe11  a1 DT Þ þ G12 e212 þ e221
2ðE1  m212 E2 Þ
m12 E1 E2 ðe11  a1 DT Þ þ E1 E2 ðe22  a2 DT Þ
þ ðe22  a2 DT Þ: ð9Þ
2ðE1  m212 E2 Þ
Eq. (9) gives the mechanical strain energy density function in terms of total strain components, material
parameters and the temperature difference. Hence, for plane stress one can write
W ¼ W ðe11 ; e12 ; e21 ; e22 ; E1 ; E2 ; m12 ; G12 ; a1 ; a2 ; DT Þ: ð10Þ
In the case of plane strain the mechanical strain energy density function is obtained as
 
E1e11 E2 m13 m232  
W ¼ 1 E1e11 þ ðm12 þ m13 m32 ÞE2e22 þ G12 e212 þ e221
2H E1 m31
E1 E2e22 E1 m31 2
þ fðm12 þ m13 m32 Þe11 þ ð1  m31 m13 Þe22 g þ a ðDT Þ2 ; ð11Þ
2H 2m13 3
where,
 
m232 m13 2
H ¼ E1 ð1  m31 m13 Þ  E2 þ m12 þ 2m12 m13 m32 ; ð12Þ
m31
e11 ¼ e11  ðm31 a3 þ a1 ÞDT ; e22 ¼ e22  ðm32 a3 þ a2 ÞDT : ð13Þ
By using Eqs. (11)–(13), W can be represented as follows for plane strain:
W ¼ W ðe11 ; e12 ; e21 ; e22 ; E1 ; E2 ; m31 ; m13 ; m12 ; m32 ; G12 ; a1 ; a2 ; a3 ; DT Þ: ð14Þ

2.2. Formulation of the equivalent domain integral

Consider a crack in a graded orthotropic medium under mode I thermal loading as depicted in Fig. 1.
Under plane stress or plane strain condition, the generalized definition of the J-integral is given by [31]
2806 S. Dag / Engineering Fracture Mechanics 73 (2006) 2802–2828

x2

E2
Γε
n E1
x1

orthotropic FGM

Fig. 1. A crack in an orthotropic functionally graded medium.

Z
J ¼ lim fWn1  rij nj ui;1 gds ði; j ¼ 1; 2Þ; ð15Þ
Ce !0 Ce

where Ce is a curve around the crack tip starting on the lower crack surface and terminating on the upper one.
Furthermore, s is the arc length of the curve, x1, x2 are the principal directions of orthotropy, W is the
mechanical strain energy density function, ni (i = 1, 2) is the unit outward normal vector, rij and ui
(i, j = 1, 2) are the stress and displacement components, respectively and (Æ),j  o(Æ)/oxj. The J-integral given
by Eq. (15) is nonzero due to stress singularity at the crack tip and is equal to the energy release rate. In order
to convert the line integral given by Eq. (15) to a domain integral we first construct a positively oriented closed
curve C around the crack tip as shown in Fig. 2. The curve C is piecewise smooth and can be expressed as
C ¼ C0 þ Cu þ Ce þ Cl : ð16Þ
We now consider the following line integral evaluated on the closed curve C:
Z Z
I ¼ frij nj ui;1  Wn1 gq ds ¼ frij ui;1 q  W d1j qgnj ds ði; j ¼ 1; 2Þ; ð17Þ
C C

where dij (i, j = 1, 2) is the Kronecker delta function and q can be any smooth function that is equal to 0 on the
curve C0 and to unity on Ce. Using the divergence theorem in the plane, I can be converted to a domain inte-
gral as
Z Z
I¼ frij ui;1 q  W d1j qg;j dX ði; j ¼ 1; 2Þ; ð18Þ
X

x2
Ω
Γ0 E2
n E1
Γe
Γu
x1
Γl

Γ = Γ 0 + Γu + Γ e + Γl

orthotropic functionally graded medium

Fig. 2. Closed curve C around the crack tip.

 S. Dag / Engineering Fracture Mechanics 73 (2006) 2802–2828 2807

where X is the area enclosed by the curve C. Since the crack tip is a singular point with regard to the stress and
strain components, it is not included in the area X. Carrying out the differentiation of the integrand, the do-
main integral becomes
Z Z Z Z
I¼ frij;j ui;1 þ rij ui;1j  W ;1 gq dX þ frij ui;1  W d1j gq;j dX ði; j ¼ 1; 2Þ: ð19Þ
X X

Note that the first integral contains the partial derivative of W with respect to x1. Considering the fact that in a
functionally graded medium, material properties and temperature distribution could be functions of the coor-
dinates x1 and x2 and using Eqs. (10) and (14), this partial derivative can be written in the following form for
both plane stress and plane strain:
 
oW oW oeij oW
¼ þ ði; j ¼ 1; 2Þ; ð20Þ
ox1 oeij ox1 ox1 expl

where (W,1)expl is generally referred to as the explicit derivative of the mechanical strain energy density func-
tion [24]. Using Eqs. (10) and (14), this term can be expressed as
 
oW oW oE1 oW oE2 oW om12 oW oG12 oW oa1
¼ þ þ þ þ
ox1 expl oE1 ox1 oE2 ox1 om12 ox1 oG12 ox1 oa1 ox1
oW oa2 oW oðDT Þ
þ þ ðfor plane stressÞ; ð21Þ
oa2 ox1 oðDT Þ ox1
 
oW oW oE1 oW oE2 oW om31 oW om13 oW om12 oW om32
¼ þ þ þ þ þ
ox1 expl oE1 ox1 oE2 ox1 om31 ox1 om13 ox1 om12 ox1 om32 ox1
oW oG12 oW oa1 oW oa2 oW oa3 oW oðDT Þ
þ þ þ þ þ ðfor plane strainÞ: ð22Þ
oG12 ox1 oa1 ox1 oa2 ox1 oa3 ox1 oðDT Þ ox1

Notice that in Eqs. (21) and (22) we have the derivatives of W with respect to the material parameters and
temperature difference. These terms are obtained in closed form for both plane stress and plane strain using
the symbolic manipulator MAPLE. The derivatives for plane stress are obtained as
   
oW  2E2 m212  E1 em m
11 þ m12 E 2 e22 m12 E2 em m
22 þ E 1 e11
¼ 2
; ð23aÞ
oE1 2ðE1  m212 E2 Þ
 m 2

oW E2 m12 em
11 þ e22
¼ 1 2
; ð23bÞ
oE2 2ðE1  m212 E2 Þ
  
oW E1 E2 m12 em m m
11 þ e22 m12 E 2 e22 þ E1 e11
m
¼ 2
; ð23cÞ
om12 ðE1  m212 E2 Þ
oW
¼ e212 þ e221 ; ð23dÞ
oG12
 
oW E1 m12 E2 em m
22 þ E 1 e11 DT
¼ ; ð23eÞ
oa1 E1  m212 E2
 
oW E1 E2 m12 em m
11 þ e22 DT
¼ ; ð23fÞ
oa2 E1  m212 E2
 
oW E1 ðE1 a1 þ m12 E2 a2 Þem m
11 þ ðE 2 a2 þ m12 E2 a1 Þe22
¼ : ð23gÞ
oðDT Þ E1  m212 E2

The expressions for plane strain are provided in Appendix. In Eqs. (21) and (22), there are also the derivatives
of the material parameters and the temperature difference with respect to x1. These terms are calculated
numerically as described in Section 2.3.
2808 S. Dag / Engineering Fracture Mechanics 73 (2006) 2802–2828

Using the property oW/oeij = rij, (i, j = 1, 2) and the strain–displacement relationship eij = (oui/oxj + ouj/
oxi)/2 (i, j = 1, 2), the derivative of the strain energy density function given by Eq. (20) can be expressed in
the following form:
oW
¼ rij ui;1j þ ðW ;1 Þexpl ði; j ¼ 1; 2Þ: ð24Þ
ox1
Substituting Eq. (24) into Eq. (19), using the equilibrium equation rij,j = 0 (i, j = 1, 2) and Eq. (17), we re-
express the integral I as
Z Z Z Z Z
I ¼ frij ui;1 q  W d1j qgnj ds ¼ frij ui;1  W d1j gq;j dX  ðW ;1 Þexpl q dX ði; j ¼ 1; 2Þ: ð25Þ
C X X

The line integral form of I calculated on the closed curve C can also be written as follows:
Z
I ¼ frij ui;1 q  W d1j qgnj ds
ZC Z Z
¼ frij ui;1 q  W d1j qgnj ds þ frij ui;1 q  W d1j qgnj ds þ frij ui;1 q  W d1j qgnj ds
Ce Cl C0
Z
þ frij ui;1 q  W d1j qgnj ds ði; j ¼ 1; 2Þ: ð26Þ
Cu

In Eq. (26), the integral on C0 is equal to zero since q = 0 on that curve. On the curves Cu and Cl, the hori-
zontal component of the unit outward normal vector (n1) is zero. Moreover, the crack surfaces are assumed to
be traction free. As a result, the line integral evaluated on the curves Cu and Cl are also equal to zero. The only
contribution to I then comes from the line integral evaluated on the curve Ce. Changing the orientation of the
curve Ce from clockwise to counterclockwise, we obtain a new curve identical to Ce as given in Fig. 1. Thus, we
have the relationship,
Z Z
I¼ frij ui;1 q  W d1j qgnj ds ¼  frij ui;1 q  W d1j qgnj ds ði; j ¼ 1; 2Þ: ð27Þ
Ce Ce

The direction of the outward normal in each of the integrals has to be consistent with the orientation of the
path of integration. Then, using Eqs. (25)–(27), taking the limit as Ce ! 0 and taking into account the fact that
q = 1 on Ce, we recover the definition of the J-integral hence the energy release rate in the form of a domain
integral
Z Z Z Z Z
 
J ¼ lim Wn1  rij nj ui;1 ds ¼ frij ui;1  W d1j gq;j dX  ðW ;1 Þexpl q dX ði; j ¼ 1; 2Þ:
Ce !0 Ce X X

ð28Þ
X in this equation is the area enclosed by the curve C0 and the singular point, i.e. crack tip, is not included in
this region.

2.3. Implementation using finite elements

The EDI approach formulated in the preceding section is integrated into the fracture analysis software
FRAC2D [16]. The solutions of the crack problems considered in the present study are obtained in two steps.
First, heat transfer problem is solved using the finite elements based heat transfer analysis software HEAT2D
[30]. Then, the computed temperature distribution is provided as an input to FRAC2D to determine the stress
and displacement fields and to calculate the energy release rate and the stress intensity factor. FRAC2D uti-
lizes graded and enriched finite elements for fracture analysis of functionally graded materials. 10-node trian-
gular isoparametric finite elements are used in the creation of the finite element models in both thermal and
mechanical analyses. The triangular element shapes in the isoparametric and global coordinate systems are
shown in Fig. 3. 10-node triangles possess cubic interpolation which is particularly useful when numerical dif-
ferentiation of certain field variables or material parameters is required in the computations. The continuous
 S. Dag / Engineering Fracture Mechanics 73 (2006) 2802–2828 2809

Node (r,s)
1 (0, 0)
x2 2 (1, 0)
6 2 3 (0, 1)
7 s 4 (1/3, 0)
3 5 (2/3, 0)
3
10 5 6 (2/3, 1/3)
7 (1/3, 2/3)
8 7 8 (0, 2/3)
8
4 9 (0, 1/3)
10 (1/3, 1/3)
9 10 6
9
1
x1 r
1 4 5 2
Fig. 3. 10-node triangular element in global and isoparametric coordinate systems.

variation of the thermomechanical properties in the functionally graded medium is incorporated into the finite
element models by computing the material properties at each Gauss point of a finite element during the for-
mation of the element stiffness matrix. The elements for which the material parameters are calculated sepa-
rately at each Gauss point are generally referred to as graded finite elements [17] and widely used to model
the behavior of functionally graded materials [32–35].
In order to calculate the energy release rate and the stress intensity factor under a given thermal loading,
first the orthotropic functionally graded medium is discretized using the 10-node triangular elements. Then,
region X has to be defined around the crack tip so as to calculate the equivalent domain integral given by
Eq. (28). Although the outcome of the analysis is independent of the shape of the region used in the calcula-
tions, circular areas are used in the present study for simplicity. Fig. 4 depicts a sample crack problem and a
circular region X whose radius is taken as R. The area integrals given by Eq. (28) can be calculated over each
triangular element within the region X and the contributions can then be summed to calculate J. Assuming
that the solutions of the thermal and mechanical problems are obtained and temperature, stress and displace-
ment distributions within the functionally graded medium are computed, the energy release rate given by

W
E2
E1
h
x2
edge crack

R r
θ x1

a
Ω

orthotropic FGM

Fig. 4. A circular region X around the tip of an edge crack.

2810 S. Dag / Engineering Fracture Mechanics 73 (2006) 2802–2828

Eq. (28) can be calculated using the isoparametric finite element formulation. If the total number of triangular
elements in the circular domain X is equal to N, J can be expressed as
X
N
J¼ Ak ; ð29Þ
k¼1
Z 1 Z 1s Z 1 Z 1s
Ak ¼ frij ui;1  W d1j gq;j jJjk dr ds  ðW ;1 Þexpl qjJjk dr ds ði; j ¼ 1; 2Þ; ð30Þ
0 0 0 0

where Ak is the area integral computed over the kth element in X, jJjk is the determinant of the Jacobian ma-
trix of the kth element and r and s are the isoparametric coordinates shown in Fig. 3. The domain integrals are
calculated numerically using Gauss quadrature techniques. The material properties are computed at each
Gauss point of a finite element during the computation of the EDI given by Eqs. (29) and (30). Note that
the integrands in Eq. (30) contain the q-function and its derivatives. The q-function is an arbitrary smooth
function which is equal to unity on the inner contour, i.e. on the crack tip, and zero on the circumference
of the circular area X [25]. Pyramid and plateau functions are generally used in the EDI technique when
the integrals are evaluated over rectangular domains [27]. In the present study, since the calculations are car-
ried out over circular regions, we use a cone function whose equation is given as
pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
x21 þ x22
qðx1 ; x2 Þ ¼ 1  : ð31Þ
R
Eq. (31) represents a right circular cone centered at the origin with base radius and height being equal to R and
1, respectively.
In the evaluation of the area integrals given in Eq. (30) spatial derivatives of the displacement components,
material parameters and the temperature difference are required. The first integral in Eq. (30) contains the
derivatives of the displacement components ui,1, (i = 1, 2) and the second integral contains (W,1)expl which
itself is a function of the spatial derivatives of the material properties and temperature difference (see Eqs.
(21) and (22)). These derivatives are calculated using the isoparametric finite element concept. As an example,
consider the spatial derivative of the modulus of elasticity in x1-direction. Over a triangular element, the coor-
dinates and modulus of elasticity are interpolated as
X
10 X
10
x1 ðr; sÞ ¼ N i ðr; sÞx1i ; x2 ðr; sÞ ¼ N i ðr; sÞx2i ; ð32Þ
i¼1 i¼1
X
10
E1 ðr; sÞ ¼ N i ðr; sÞE1i ; ð33Þ
i¼1

where x1i, x2i, (i = 1, . . . , 10), are nodal coordinates, E1i (i = 1, . . . , 10), are the nodal values of the modulus of
elasticity and Ni (i = 1, . . . , 10), are the shape functions. The derivatives with respect to local coordinates are
given by
oE1 oE1 ox1 oE1 ox2
¼ þ ; ð34Þ
or ox1 or ox2 or
oE1 oE1 ox1 oE1 ox2
¼ þ : ð35Þ
os ox1 os ox2 os
Hence, derivatives of E1 with respect to global coordinates can now be calculated using
" oE1 # " ox #" oE #
ox1 1 os
2
 oxor2 or
1

¼ : ð36Þ
oE1 jJj  ox1 ox1 oE1
os
ox2 os or

The local derivatives on the right-hand-side can be computed from

oE1 X 10
oN i oE1 X 10
oN i
¼ E1i ; ¼ E1i : ð37Þ
or i¼1
or os i¼1
os
 S. Dag / Engineering Fracture Mechanics 73 (2006) 2802–2828 2811

2.4. Calculation of the mode I stress intensity factor

After calculating J, the mode I stress intensity factor KI can also be determined using the relation between
the energy release and the stress intensity factor. Consider a mode I crack in a graded orthotropic medium
under thermal loading as depicted in Fig. 4. (r, h) in this figure constitute a local polar coordinate system
whose origin is at the crack tip. Mode I stress intensity factor KI is defined using the stress component
r22(r, h) in the following form:
npffiffiffiffiffiffiffi o
K I ¼ lim 2prðr22 ðr; 0ÞÞ : ð38Þ
r!0

For orthotropic materials, relationship between J and KI is given as [18]

8  
> 1
<  2Etip I K 2I ðl1 þ l2 Þl1 l2 ; plane stress;
1
J¼ tip tip   ð39Þ
> 1m m
:  31tip 13 I K 2I ðl1 þ l2 Þl1 l2 ; plane strain;
2E 1

where the superscript (tip) refers to the material parameters calculated at the tip of the crack in the function-
ally graded medium and the overline stands for the complex conjugate. l1 and l2 can be determined by solving
the characteristic equation

atip 4 tip tip 2 tip

11 l þ 2a12 þ a66 l þ a22 ¼ 0: ð40Þ

The four roots of Eq. (40) are always complex or purely imaginary in conjugate pairs. l1 and l2 are the roots
whose imaginary parts are greater than zero. The compliance coefficients atipij are given as follows for plane
stress:
1 mtip 1 1
atip
11 ¼ ; atip
12 ¼
12
; atip
22 ¼ ; atip
66 ¼ : ð41Þ
Etip
1 Etip
1 Etip
2 Gtip
12

In the case of plane strain these constants are given by

1  mtip tip
31 m13 mtip tip tip
12 þ m13 m32 1  mtip tip
23 m32 1
atip
11 ¼ ; atip
12 ¼  ; atip
22 ¼ ; atip
66 ¼ : ð42Þ
Etip
1 Etip
1 Etip
2 Gtip
12

3. Numerical verification of the EDI technique and domain independence

3.1. An edge crack in an orthotropic FGM layer under steady-state thermal stresses

In this section, we consider calculation of the mode I stress intensity factors in an orthotropic FGM layer
under steady-state thermal loading. The geometry of the problem and thermal boundary conditions are
depicted in Fig. 5. An orthotropic functionally graded layer of thickness W contains an edge crack of length
a that is aligned parallel to the direction of the material property gradation. The surfaces x1 = 0 and x1 = W
are assumed to be kept at temperatures T1 and T2, respectively. Initially, the FGM layer is at a reference tem-
perature T0 for which all the stress components are equal to zero. Upper and lower ends x2 = ±h and surfaces
normal to x3-axis are assumed to be insulated. Layer depth in x3-direction is either very small (plane stress) or
sufficiently large (plane strain). Under these conditions, heat flow is confined to x1-direction and temperature
distribution is one-dimensional. Crack surfaces are assumed to be insulated, hence presence of the edge crack
does not disturb the one-dimensional temperature distribution. As a result, the crack undergoes mode I type of
deformation and KI gives the amplitude of the singular field around the crack tip. Research software
HEAT2D [30] is used to determine the temperature profile of the orthotropic FGM layer. This software uti-
lizes two-dimensional graded finite elements to solve the transient conduction equation which is expressed as
   
o oT o oT oT
k1 þ k2 ¼ qc ; ð43Þ
ox1 ox1 ox2 ox2 ot
2812 S. Dag / Engineering Fracture Mechanics 73 (2006) 2802–2828

x2 insulated surface

W
T = T1 T = T2
E2
E1 h

edge crack

R x1

a
Ω

Al2O3 Ni h

orthotropic FGM

reference temperature = T0

insulated surface

Fig. 5. An orthotropic functionally graded layer under steady-state thermal loading.

where k1 and k2 are principal thermal conductivities in x1 and x2 directions, respectively, q is the density and c
is the specific heat.
All the thermomechanical properties of the FGM layer are assumed to be continuous functions of the
x1-coordinate. The material property variations are represented in the following form:
  c
E1 ðx1 Þ ¼ E01 þ Ew1  E01 ðx1 =W Þ 1 ; ð44aÞ
0
 w 0
 c2
E2 ðx1 Þ ¼ E2 þ E2  E2 ðx1 =W Þ ; ð44bÞ
  b
m12 ðx1 Þ ¼ m012 þ mw12  m012 ðx1 =W Þ 12 ; ð44cÞ
0
 w 0
 b13
m13 ðx1 Þ ¼ m13 þ m13  m13 ðx1 =W Þ ; ð44dÞ
0
 w 0
 b31
m31 ðx1 Þ ¼ m31 þ m31  m31 ðx1 =W Þ ; ð44eÞ
  b
m32 ðx1 Þ ¼ m032 þ mw32  m032 ðx1 =W Þ 32 ; ð44fÞ
0
 w 0
 c12
G12 ðx1 Þ ¼ G12 þ G12  G12 ðx1 =W Þ ; ð44gÞ
0
 w 0
 d1
a1 ðx1 Þ ¼ a1 þ a1  a1 ðx1 =W Þ ; ð44hÞ
  d
a2 ðx1 Þ ¼ a02 þ aw2  a02 ðx1 =W Þ 2 ; ð44iÞ
0
 w 0
 d3
a3 ðx1 Þ ¼ a3 þ a3  a3 ðx1 =W Þ ; ð44jÞ
0
 w 0
 x1
k 1 ðx1 Þ ¼ k 1 þ k 1  k 1 ðx1 =W Þ ; ð44kÞ
0
 w 0
 x2
k 2 ðx1 Þ ¼ k 2 þ k 2  k 2 ðx1 =W Þ ; ð44lÞ
k
qðx1 Þ ¼ q0 þ ðqw  q0 Þðx1 =W Þ ; ð44mÞ
0 w 0 v
cðx1 Þ ¼ c þ ðc  c Þðx1 =W Þ ; ð44nÞ
where the parameters with the superscripts 0 and w stand for the material properties at the surfaces x1 = 0 and
x1 = W, respectively. Remaining material properties m21, m23 and E3 are calculated by substituting the required
parameters into Eq. (4). The particular ceramic and metal components of the orthotropic FGM layer shown in
 S. Dag / Engineering Fracture Mechanics 73 (2006) 2802–2828 2813

Fig. 5 are taken as alumina (Al2O3) and nickel (Ni). The FGM layer is assumed to be 100% alumina and
orthotropic at x1 = 0 and 100% nickel and isotropic at x1 = W. Thus, there are smooth spatial variations
in all of the thermomechanical parameters from orthotropy at x1 = 0 to isotropy at x1 = W.
The power-law representations given by Eq. (44) are highly flexible and widely used to model continuous
variations in the thermomechanical properties of functionally graded materials [36,37]. The exponents in these
equations are positive constants that can be adjusted to obtain the required variation profile for the orthotro-
pic FGM layer. If an exponent is greater than one the corresponding material property has a ceramic-rich var-
iation profile whereas for an exponent that is less than unity variation profile is metal-rich. Parametric analyses
can be carried out either by specifying all the exponents for the 14 material parameters independently or by
using the same exponent for certain parameters. For example, in the studies by Lee and Erdogan [36] and Shen
[37], the exponents for all of the thermomechanical parameters are assumed to be equal. Partly for lack of
published data and partly for simplicity, same exponents are used for certain material parameters in the para-
metric analyses conducted in the present study. However, using the same exponent does not imply that vari-
ations in the corresponding parameters are identical. For instance, setting c1 = c2 merely imply that relative
differences E1  E01 and E2  E02 are proportional.
Sevostianov and Kachanov [38] developed a theoretical model to calculate the elastic and conductive prop-
erties of orthotropic plasma-sprayed alumina coatings. The study by Sevostianov and Kachanov [38] also pre-
sents experimental data on the orthotropic elastic properties of alumina. The provided experimental data is
originally obtained by Parthasarathi et al. [39] using ultrasound measurements and consists of the stiffness
coefficients of the plasma-sprayed alumina coatings. In the present study, the experimental data given in
[38] is used to calculate the elastic parameters of the orthotropic alumina surface at x1 = 0. The orthotropic
thermal conductivities of alumina on the other hand are calculated using the theoretical model of Sevostianov
and Kachanov [38]. Most of the work in the literature on the characterization of orthotropic plasma spray
deposits seems to be directed towards determining the elastic and conductive properties [38–41]. No published
data could be found in the literature on the orthotropic thermal expansion coefficients of plasma-sprayed coat-
ings. The nominal value for the thermal expansion coefficient of alumina is given as a = 7.5(10)6 (°C)1 by
Callister [42]. This value is used as the coefficient of thermal expansion in the thickness direction. Larger values
are used for the other two principal directions in the numerical simulations in order to be able examine the
influence of orthotropy on the mode I stress intensity factors. The data for density and specific heat of alumina
and for all the thermomechanical properties of the isotropic nickel surface at x1 = W are also taken from [42].
The material parameters of the 100% alumina surface are then given as follows:
E01 ¼ 116:36 GPa; E02 ¼ 90:43 GPa; G012 ¼ 38:21 GPa; ð45aÞ
m012 ¼ 0:28; m013 ¼ 0:27; m031 ¼ 0:21; m032 ¼ 0:14; ð45bÞ
k 01 ¼ 29:82 W=ðm KÞ; k 02 ¼ 21:25W=ðm KÞ; ð45cÞ
6  1 6  1 6  1
a01 ¼ 7:5ð10Þ ð CÞ ; a02 ¼ 8ð10Þ ð CÞ ; a03 ¼ 9ð10Þ ð CÞ ; ð45dÞ
0 3 0
q ¼ 3980 kg=m ; c ¼ 775 J=ðkg KÞ: ð45eÞ
Material property data for the 100% nickel surface is given by [42]
Ew1 ¼ Ew2 ¼ Ew ¼ 204 GPa; mw12 ¼ mw13 ¼ mw31 ¼ mw32 ¼ mw ¼ 0:31; ð46aÞ
Gw12 w
¼ E =ð2ð1 þ m ÞÞ ¼ 77:9 GPa; w
ð46bÞ
6  1
k w1 ¼ k w2 w
¼ k ¼ 70 W=ðm KÞ; aw1 ¼ aw2 ¼ aw3 w
¼ a ¼ 13:3ð10Þ ð CÞ ; ð46cÞ
w 3 w
q ¼ 8890 kg=m ; c ¼ 456 J=ðkg KÞ: ð46dÞ
There are certain limitations on the Poisson’s ratios in orthotropic materials. These limitations are expressed
by the following inequalities [43]:
ð1  m12 m21 Þ > 0; ð1  m13 m31 Þ > 0; ð1  m23 m32 Þ > 0; ð47aÞ
ð1  m12 m21  m13 m31  m23 m32  2m12 m23 m31 Þ > 0: ð47bÞ
These inequalities hold at every point in the alumina–nickel orthotropic functionally graded layer.
2814 S. Dag / Engineering Fracture Mechanics 73 (2006) 2802–2828

1.8

ω1 = 5

1.6
ω1 = 2 ω1 = 0.5

T / T1
1.4

ω1 = 0.2

1.2

1.0
0.0 0.2 0.4 0.6 0.8 1.0
x1 / W

Fig. 6. Steady-state temperature distribution in the orthotropic FGM layer for various values of x1.

In the numerical examples considered in the present section, the medium is initially assumed to be at a high
processing temperature of T0 = 1100 K. The temperatures of the surfaces x1 = 0 and x1 = W are taken as
T1 = 0.27(T0) and T2 = 0.5(T0), respectively. The half layer height h is set to be equal to 3W. Steady-state tem-
perature distribution in the FGM layer calculated using HEAT2D is shown in Fig. 6 for various values of the
exponent x1 (see Eq. (44k)). Giving different values to x1, it is possible to obtain different variation profiles for
the principal thermal conductivity, k1. The other principal thermal conductivity k2 has no effect on the tem-
perature distribution profile since heat flow in the layer is restricted to x1-direction. In order to verify the
developed equivalent domain integral method, we next present comparisons of the stress intensity factors
obtained by EDI to those calculated by an enriched finite element method. Tables 1 and 2 show the compar-
isons of the normalized mode I stress intensity factor KIn for cracks of relative lengths a/W = 0.1 and 0.6,
respectively. Formulation and implementation of enriched finite elements for fracture analysis of orthotropic
functionally graded materials are described in the paper by Dag et al. [16]. Domain independence of EDI is
also demonstrated in these tables by providing the results for different values of the domain radius R. The

Table 1
Normalized mode I stress intensity factors (KIn ) for a/W = 0.1 computed using equivalent domain integral (EDI) and enriched finite
elements (EFE), c1 = c2 = c12 = 2, b31 = b13 = b12 = b32 = 1.5, d1 = d2 = d3 = 2, x1 = 2
R/a Plane stress Plane strain
EDI EFE Percent difference EDI EFE Percent difference
0.1 0.1849 0.1852 0.16 0.2643 0.2649 0.23
0.2 0.1850 0.1852 0.11 0.2645 0.2649 0.15
0.3 0.1850 0.1852 0.11 0.2645 0.2649 0.15
0.4 0.1850 0.1852 0.11 0.2645 0.2649 0.15

Table 2
Normalized mode I stress intensity factors (KIn) for a/W = 0.6 computed using equivalent domain integral (EDI) and enriched finite
elements (EFE), c1 = c2 = c12 = 2, b31 = b13 = b12 = b32 = 1.5, d1 = d2 = d3 = 2, x1 = 2
R/a Plane stress Plane strain
EDI EFE Percent difference EDI EFE Percent difference
0.1 0.0233 0.0233 0 0.0369 0.0370 0.27
0.2 0.0233 0.0233 0 0.0369 0.0370 0.27
0.3 0.0233 0.0233 0 0.0369 0.0370 0.27
0.4 0.0233 0.0233 0 0.0369 0.0370 0.27
 S. Dag / Engineering Fracture Mechanics 73 (2006) 2802–2828 2815

comparisons are given for both plane stress and plane strain conditions. The normalized stress intensity factor
is defined as
KI
K In ¼ pffiffiffiffiffiffi ð48Þ
E01 a01 T 1 pa
Deformed shape of the finite element model and circular domains used in the evaluation of the domain inte-
grals for a/W = 0.6 are shown in Fig. 7. Identical finite element meshes are used in the solutions of the plane
stress and strain problems. Total number of finite elements used to compute the stress intensity factors for a/
W = 0.1 and 0.6 are 1446 and 1750, respectively. In each case, same number of elements are used in the en-
riched finite element computations. In the enriched finite element method, triangular enriched and transition
elements are used to model the crack tip region. In the EDI approach however, finite element meshes consist of
only regular 10-noded triangular cubic elements. No special crack tip elements such as quarter point or en-
riched elements are employed in EDI calculations. But, especially the mesh in the circular domain is refined
in order to evaluate the stress intensity factors within a high degree of accuracy. In both Tables 1 and 2, it can
be seen that the results calculated using the EDI method and enriched finite elements agree quite well. The
results presented for four separate values of the radius R also show that the method possesses the required
domain independence. Maximum percent difference between the results calculated using EDI and enriched
finite elements is 0.27%.
Next, we examine the effects of the variations in the principal thermal conductivity k1 and thermal expan-
sion coefficients a1, a2 and a3 on the mode I stress intensity factors. The exponents, x1, d1, d2 and d3 govern the
variations in k1, a1, a2 and a3, respectively (see Eqs. (44h)–(44k)). Normalized stress intensity factors are cal-
culated for different values of each exponent to study the effect of the variations in the corresponding thermal
material property. The influence of the variations in k1 on the normalized mode I stress intensity factor KIn is
presented in Fig. 8. This figure shows the stress intensity factors for three separate values of x1 computed
using the EDI approach and enriched finite elements under both plane strain and plane stress conditions.
It is seen that normalized mode I stress intensity factor is a decreasing function of the relative crack length
for all values of x1 considered. At a given crack length, KIn increases as x1 is increased from 0.2 to 2. An excel-
lent agreement is observed between the results evaluated by means of EDI and enriched elements. Further-
more, the stress intensity factors calculated for plane strain are found to be greater than those of plane stress.

Fig. 7. (a) Deformed shape and (b) circular domains around the crack tip, for the orthotropic FGM layer shown in Fig. 5 considering
plane strain, a/W = 0.6, c1 = c2 = c12 = 2, b12 = b13 = b31 = b32 = 1.5, d1 = d2 = d3 = 2, x1 = 2.
2816 S. Dag / Engineering Fracture Mechanics 73 (2006) 2802–2828

0.4
Equivalent domain integral
Enriched finite elements
plane strain

0.3
ω1 = 2

ω1 = 0.5
KIn
0.2
ω1 = 0.2

plane stress
0.1

ω1 = 2
ω1 = 0.5
ω1 = 0.2
0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
a/W

Fig. 8. Normalized mode I stress intensity factor versus x1 and a/W for the orthotropic FGM layer shown in Fig. 5, c1 = c2 = c12 = 2,
b12 = b13 = b31 = b32 = 1.5, d1 = d2 = d3 = 2.

In problems involving FGMs or materials subjected to thermal loads, stress distributions under identical
loading conditions are known to be different for plane stress and strain. This statement can be proven math-
ematically by using the stress function concept. In these types of problems, governing equations in terms of a
stress function are not identical for plane stress and strain [44,45]. Moreover, the governing equations contain
terms that depend on thermoelastic parameters. As a result, for FGMs and for materials that are subjected to
thermal stresses, state of stress could significantly influence the stress intensity factors. In general, it is known
that under identical loading conditions, thermal stresses and thermal stress intensity factors calculated for
plane strain are larger than those calculated for plane stress. For example, Noda et al. [44] provide an analyt-
ical solution for a homogeneous isotropic strip that is subjected to a one dimensional transient temperature
gradient. The closed form expression given for the thermal stress component indicates that plane strain solu-
tion can be obtained by multiplying the plane stress solution by a factor of 1/(1  m). Since 0 < m < 0.5, the
magnitude of the thermal stress for plane strain is larger than that for plane stress. Similarly, the results pro-
vided in Fig. 8 show that, in the case of orthotropic FGMs, thermal stress intensity factors calculated for plane
strain are larger than those calculated for plane stress.
For an isotropic functionally graded medium under a known temperature field, thermoelastic constants
required to calculate the stress intensity factors are modulus of elasticity, Poisson’s ratio and coefficient of
thermal expansion for both plane stress and strain. However, for orthotropic FGMs the number of material
parameters required for thermal fracture analysis is not equal for plane stress and strain. In the case of plane
stress six parameters are required for thermal fracture analysis whereas variation profiles of ten parameters are
needed to be known for plane strain (see for example Eqs. (1), (3) and (4)). In the present study, the additional
parameters that are specified for the plane strain case are m13, m31, m32 and a3 which are given by Eq. (44). Note
that these parameters are related to the thermoelastic behavior of the medium in the thickness direction. On
the contrary, thickness direction behavior of a material does not affect the solution of the plane stress problem.
Therefore, depending on the nature of material property distribution profiles, there could be significant differ-
ences between the results calculated for plane stress and strain states of orthotropic FGMs as shown in the
present study.
Figs. 9 and 10 present the influence of the variations in thermal expansion coefficient a2 on the normalized
mode I stress intensity factors for plane stress and plane strain, respectively. In the numerical analyses it is
found that the edge crack is generally closed, i.e. mode I stress intensity factor is negative, if d2 is less than
one. Thus, the results are provided for d2 values that are greater than unity. As shown in the figures, the stress
 S. Dag / Engineering Fracture Mechanics 73 (2006) 2802–2828 2817

0.4
Equivalent domain integral
Enriched finite elements
0.3 δ2 = 5
δ2 = 5, 2, 1.6, 1.2

KIn
0.2

0.1

δ2 = 1.2
0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
a/W

Fig. 9. Normalized mode I stress intensity factor versus d2 and a/W considering plane stress condition for the orthotropic FGM layer
shown in Fig. 5, c1 = c2 = c12 = 2, b12 = b13 = b31 = b32 = 1.5, d1 = d3 = 2, x1 = 2.

0.5
Equivalent domain integral
Enriched finite elements
0.4
δ2 = 5
δ2 = 5, 2, 1.6, 1.2
0.3
KIn

0.2

0.1 δ2 = 1.2

0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
a/W

Fig. 10. Normalized mode I stress intensity factor versus d2 and a/W considering plane strain condition for the orthotropic FGM layer
shown in Fig. 5, c1 = c2 = c12 = 2, b12 = b13 = b31 = b32 = 1.5, d1 = d3 = 2, x1 = 2.

intensity factor is found to be an increasing function of d2 and getting close to zero as d2 approaches to one.
Note that since a02 < aw2 , as d2 is increased from unity to a larger value crack tip value of the thermal expansion
coefficient a2 decreases (see Eq. (44i)). It can also be observed that the plane strain results are larger than those
of plane stress for all d2 values considered. The agreement between the results calculated using EDI and
enriched finite elements is again found to be quite good. Fig. 11 shows the set of results that illustrate the effect
of the variations in the thermal expansion coefficient a3 on the mode I stress intensity factors. It is seen that KIn
increases continuously as d3 is increased from 0.2 to 5. In this case, the results are presented only for the case of
plane strain since a3 has no effect on the stress intensity factor if the FGM layer is in a state of plane stress.
In the parametric analyses carried out to examine the influence of the thermal expansion coefficient a1, it is
found that, stress intensity factor is not sensitive to the variations in the exponent d1. Some sample results are
provided in Tables 3 and 4 for two separate data sets. The results computed by changing d1 and keeping all
other variables constant are found to be identical up to three significant digits in all cases for both plane stress
and strain. This interesting result is verified by carrying out the computations using both the EDI and enriched
finite element approaches. It is concluded that in the problem presented by Fig. 5, d1 has no effect on the mode
I stress intensity factor.
2818 S. Dag / Engineering Fracture Mechanics 73 (2006) 2802–2828

0.4
Equivalent domain integral
Enriched finite elements
0.3

δ3 = 5, 2, 1, 0.2

KIn
0.2

δ3 = 5

0.1
δ3 = 0.2

0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
a/W

Fig. 11. Normalized mode I stress intensity factor versus d3 and a/W considering plane strain condition for the orthotropic FGM layer
shown in Fig. 5, c1 = c2 = c12 = 2, b12 = b13 = b31 = b32 = 1.5, d1 = d2 = 2, x1 = 2.

Table 3
Normalized mode I stress intensity factors for various values of the exponent d1 computed using equivalent domain integral (EDI) and
enriched finite elements (EFE), a/W = 0.2, c1 = c2 = c12 = 2, b31 = b13 = b12 = b32 = 1.5, d2 = d3 = 2, x1 = 2
d1 Plane stress Plane strain
EDI EFE EDI EFE
5 0.1334 0.1335 0.1926 0.1926
2 0.1334 0.1335 0.1926 0.1926
1 0.1334 0.1335 0.1926 0.1926
0.5 0.1334 0.1335 0.1927 0.1927
0.2 0.1334 0.1335 0.1927 0.1927

Table 4
Normalized mode I stress intensity factors for various values of the exponent d1 computed using equivalent domain integral (EDI) and
enriched finite elements (EFE), a/W = 0.3, c1 = c2 = c12 = 0.5, b31 = b13 = b12 = b32 = 1.5, d2 = d3 = 1.2, x1 = 2
d1 Plane stress Plane strain
EDI EFE EDI EFE
5 0.0265 0.0265 0.0642 0.0643
2 0.0265 0.0265 0.0642 0.0643
1 0.0265 0.0265 0.0642 0.0643
0.5 0.0265 0.0265 0.0642 0.0643
0.2 0.0265 0.0265 0.0642 0.0643

In this section, we finally provide some comparisons in order to demonstrate the need for the inclusion of
material orthotropy in thermal fracture analysis of FGMs. The geometry and thermal loading conditions con-
sidered in the comparisons are shown by Fig. 5. Stress intensity factors calculated by assuming the FGM layer
shown in Fig. 5 to be orthotropic are compared to those evaluated for an isotropic FGM layer. Both ortho-
tropic and isotropic FGM layers are assumed to be 100% alumina at x1 = 0 and 100% nickel at x1 = W. Ther-
momechanical properties of the orthotropic FGM layer are given by Eqs. (44)–(46) as described above.
Material property distributions in the isotropic FGM layer are also assumed to be represented by Eq. (44).
However, for the isotropic FGM layer, 100% alumina surface at x1 = 0 is assumed to be isotropic with mod-
ulus of elasticity, coefficient of thermal expansion and thermal conductivity being equal to the corresponding
 S. Dag / Engineering Fracture Mechanics 73 (2006) 2802–2828 2819

Table 5
Normalized mode I stress intensity factors (KIn) for orthotropic and isotropic FGMs computed considering the condition of plane strain,
c1 = c2 = c12 = 2, b31 = b13 = b12 = b32 = 1.5, d1 = d2 = d3 = 2, x1 = 2, T1 = 0.27(T0 ), T2 = 0.5(T0)
a/W Orthotropic FGM Isotropic FGM Percent difference
0.1 0.2645 0.3463 30.9
0.2 0.1926 0.2484 29.0
0.3 0.1352 0.1759 30.1
0.4 0.0975 0.1200 23.1
0.5 0.0638 0.0761 19.3

x1-direction values of orthotropic alumina. An average value of 0.21 is used for the Poisson’s ratio. Therefore,
at x1 = 0, material properties of the isotropic FGM layer are expressed as follows:
E01 ¼ E02 ¼ 116:36 GPa; m012 ¼ m013 ¼ m031 ¼ m032 ¼ 0:21; ð49aÞ
6  1
a01 ¼ a02 ¼ a03 ¼ 7:5ð10Þ ð CÞ ; k 01 ¼ k 02 ¼ 29:82 W=ðm KÞ; ð49bÞ
  
G012 ¼ E01 = 2 1 þ m012 ¼ 48:1 GPa: ð49cÞ
At the surface x1 = W, isotropic FGM layer is 100% nickel whose material properties are given by Eq. (46).
Comparisons of the normalized mode I stress intensity factors evaluated for orthotropic and isotropic
FGM layers are provided in Table 5. The normalized stress intensity factor KIn is defined by Eq. (48). The
results given in the table are calculated by means of the EDI approach. For each crack length, a domain radius
of R = (0.4)a is used in the computations. Table 5 also presents percent differences calculated using the results
for the orthotropic FGM layer as reference. Examining the tabulated results, it can be seen that, there are sig-
nificant differences between the stress intensity factors calculated for orthotropic and isotropic FGM layers.
The outcome is observed to be highly sensitive to the nature of the material orientation. Therefore, inclusion
of orthotropy in fracture mechanics analyses of FGMs could lead to considerable improvements in the accu-
racy of the calculated stress intensity factors.

3.2. Comparisons to the results reported in the literature

In order to further verify the developed method, this section presents some comparisons of the thermal
stress intensity factors to those reported by Erdogan and Wu [46] and Walters et al. [35]. Erdogan and Wu
[46] examined the problem of an edge crack in an isotropic semi-infinite FGM strip that is subjected to
steady-state thermal stresses. The problem geometry and thermal boundary conditions considered by Erdogan
and Wu [46] are same as those shown in Fig. 5. Referring to this figure, material property distributions used in
Ref. [46] can be represented in the following form:
E1 ðx1 Þ ¼ E2 ðx1 Þ ¼ E3 ðx1 Þ ¼ Eðx1 Þ ¼ E0 expðbx1 Þ; ð50aÞ
a1 ðx1 Þ ¼ a2 ðx2 Þ ¼ a3 ðx1 Þ ¼ aðx1 Þ ¼ a0 expðxx1 Þ; ð50bÞ
k 1 ðx1 Þ ¼ k 2 ðx1 Þ ¼ kðx1 Þ ¼ k 0 expðdx1 Þ; ð50cÞ
m12 ¼ m13 ¼ m31 ¼ m32 ¼ m21 ¼ m23 ¼ m ¼ 0:3; ð50dÞ
Eðx1 Þ
G12 ðx1 Þ ¼ Gðx1 Þ ¼ : ð50eÞ
2ð1 þ mÞ
Note that all parameters possess exponential variations throughout the graded layer except the Poisson’s ratio
which is assumed to have a constant value of 0.3. Erdogan and Wu [46] solved this problem analytically using
Fourier transform techniques and the method of singular integral equations. Walters et al. [35] also considered
the problem described above using a three dimensional finite element method in conjunction with a domain
integral approach.
The comparisons of the normalized mode I stress intensity factors calculated in the present analysis by
means of equivalent domain integral and enriched finite element methods to those reported by Erdogan
and Wu [46] and Walters et al. [35] are given in Table 6. The table presents the comparisons for two different
2820 S. Dag / Engineering Fracture Mechanics 73 (2006) 2802–2828

Table 6
Comparisons of the normalized mode I stress intensity factors (KIn) for the case of plane strain, a/W = 0.5, bW = ln(10), xW = ln(2),
dW = ln(10)
T1 = 0.2T0, T2 = 0.5T0 T1 = 0.05T0, T2 = 0.05T0
EDI R/a = 0.1 0.0338 0.0406
R/a = 0.2 0.0340 0.0407
R/a = 0.3 0.0340 0.0408
R/a = 0.4 0.0339 0.0407
Enriched elements 0.0339 0.0407
Walters et al. [35] 0.0335 0.0409
Erdogan and Wu [46] 0.0335 0.0410

thermal loading conditions. In both EDI and enriched element calculations, the layer height is taken as
h = 4W. Normalized mode I stress intensity factor is defined as
KI
K In ¼ pffiffiffiffiffiffi : ð51Þ
ðE0 a0 T 0 paÞ=ð1  mÞ

Note that four different domain radii are used in EDI calculations. Examining the results given in the table, it
can be seen that the stress intensity factors obtained using EDI and enriched elements agree quite well with
those given in Refs. [35,46]. The results calculated for different R values are also in good agreement with each
other which implies that EDI approach possesses the required domain independence. The comparisons pre-
sented in Section 3.1 and in this Section indicate that, by using only regular 10-node triangular elements,
i.e. without any special elements at the crack tip, it is possible to generate highly accurate results via equivalent
domain integral method provided that the mesh within the integral domain is sufficiently refined.

4. Periodic cracks in orthotropic FGM coatings under transient thermal loading

Ceramic thermal barrier coatings are used in various engineering applications such as aircraft engines and
stationary gas turbines, in order to decrease the temperature of metallic surfaces and improve the thermal effi-
ciency and mechanical durability. These coatings are known to be susceptible to surface cracking due to resid-
ual thermal stresses. This type of surface failure can occur either in the form of a single dominant crack or as a
roughly regular array of periodic cracks. In this section, we consider periodic surface cracks in an orthotropic
FGM ceramic–metal coating bonded to a homogeneous metallic substrate under transient thermal stresses.
Problem geometry is shown in Fig. 12. The FGM coating is an Al2O3–Ni layer with 100% alumina at the free
surface x1 = 0 and 100% nickel at the interface x1 = W1. Material property variations in the FGM coating are
given by Eqs. (44)–(46). Note that thermal and mechanical properties of the coating are orthotropic at x1 = 0
and isotropic at x1 = W1. Hence, there is a smooth variation from orthotropy to isotropy as governed by Eqs.
(44)–(46). FGM layer is assumed to be perfectly bonded to a homogeneous and isotropic nickel substrate
whose thickness is taken as W2. Thermomechanical properties of the composite medium are continuous at
each point including the interface x1 = W1. The FGM coating contains periodic edge cracks of length a
and spacing b. Applied thermal boundary conditions are also shown in Fig. 12. Initially, both the coating
and the substrate are assumed to be at a uniform processing temperature of T0 = 1100 K. Afterwards, the tem-
perature of the surface at x1 = 0 is decreased to and kept at T1 = 0.27(T0) which is roughly the room temper-
ature. All the other outer surfaces including the surfaces normal to the x3-axis and crack faces are assumed to
be insulated. Under these conditions, temperature distribution in the coating–substrate structure is a function
of x1 only. Note that since we are particularly interested in the transient behavior of the periodic cracks, in
addition to the principal thermal conductivities k1 and k2, density q and specific heat c are also required in
the finite element computations.
The transient temperature distribution in the coating–substrate structure under the mentioned boundary
and initial conditions is calculated using the heat transfer analysis software HEAT2D. Calculated transient
temperature distribution data is then transferred to FRAC2D in order to compute the stress intensity factor
 S. Dag / Engineering Fracture Mechanics 73 (2006) 2802–2828 2821

x2
reference temperature = T0

Al2O3 – Ni FGM Ni substrate

E2
T = T1 E1
insulated surface

a
x1

W1 W2

Fig. 12. Periodic edge cracks in an orthotropic FGM coating bonded to a homogeneous substrate.

as a function of time. The inertia effects are assumed to be negligible in the calculations. In both thermal and
mechanical analyses triangular graded finite elements with cubic interpolation are utilized. The stress intensity
factor is computed by means of the equivalent domain integral and enriched finite element methods. Note that
due to the periodicity in the geometry of the problem, it is sufficient to consider a unit cell in finite element
modeling. The unit cell with thermal and mechanical boundary conditions is depicted in Fig. 13. Since, there
is no heat flux in the composite medium in x2-direction, the planes x2 = ±b/2 of the unit cell are taken as insu-
lated. Furthermore, the periodicity of the structure requires that normal displacements of both of the lines AB
and CD be tied.
Transient temperature distribution in the FGM coating–substrate structure for various values of normal-
ized time s is presented in Fig. 14. Normalized time is defined by
ðjÞðtÞ k 01
s¼ ; j¼ : ð52Þ
W 21 q0 c0
During transient thermal loading, temperature of the surface x1 = 0 is first decreased to and then kept
at T = T1 = 0.27(T0). Temperatures of the other points in the medium decrease continuously as s increases

Vertical displacements of the nodes on

both lines AB and CD are tied.
x2 ∂T
=0
∂x2
A B
a b/2
x1
b/2
C D
∂T
=0
∂x2
W1 W2

Fig. 13. Unit cell in the coating–substrate structure and periodicity conditions.
2822 S. Dag / Engineering Fracture Mechanics 73 (2006) 2802–2828

4.0 τ=0

3.5 τ = 0.1

3.0 0.3
0.6
1
2.5

T / T1
2
2.0 3

1.5 5

10
1.0
steady - state
0.5
0.0 0.5 1.0 1.5 2.0 2.5 3.0
x1 / W1

Fig. 14. Transient temperature distribution in the FGM coating–substrate structure, x1 = 2, k = v = 1.5.

from 0. The distribution is uniform at steady-state as expected (T = T1). Moreover, steady-state profile is cal-
culated to be reached at approximately s  17. The variation of the normalized mode I stress intensity factor
with normalized time and crack spacing ratio b/a is given in Fig. 15 for the relative crack length of a/W1 = 0.5.
Normalized stress intensity factor is defined by Eq. (48). The stress intensity factors are calculated under the
condition of plane strain. Fig. 16 shows the deformed shape of the finite element mesh obtained at s = 0.4 for
b/a = 1. Considering Fig. 15, it can be seen that, as the normalized time increases stress intensity factors go
through a peak point before starting to decrease. Stress intensity values approach to zero near s  2.2 for
all cases. Further increase in time will cause the stress intensity factor to be negative indicating the occurrence
of crack closure. Thus, these results show that, although at steady-state periodic cracks are closed, the stress
intensity factors go through positive peaks during the transient period of heat conduction due to tensile resid-
ual stresses that develop near the surface x1 = 0. It is also observed that KIn increases as the relative crack
spacing b/a is increased from 0.1 to 10. Results computed using equivalent domain integral and enriched finite
element methods are seen to be in very good agreement.
Next, we consider the effect of relative crack length a/W1 on the mode I stress intensity factor. The variation
of the stress intensity factor with normalized time for various values of a/W1 is illustrated in Fig. 17 for plane

Equivalent domain integral

1.6
Enriched finite elements

1.2 b/a = 10, 5, 2, 1, 0.5, 0.1

KIn

0.8
b/a = 10

0.4

b/a = 0.1
0.0
0.0 0.4 0.8 1.2 1.6 2.0 2.4
τ

Fig. 15. Normalized mode I stress intensity factor versus normalized time and crack spacing ratio b/a considering plane strain condition
for the coating–substrate structure shown in Fig. 12, c1 = c2 = c12 = 2, d1 = d2 = d3 = 2, b12 = b13 = b31 = b32 = 1.5, x1 = 2, k = v = 1.5,
a/W1 = 0.5, W2 = 2W1.
 S. Dag / Engineering Fracture Mechanics 73 (2006) 2802–2828 2823

Fig. 16. Deformed shape of the finite element model of the unit cell shown in Fig. 13 at s = 0.4, c1 = c2 = c12 = 2,
b12 = b13 = b31 = b32 = 1.5, d1 = d2 = d3 = 2, x1 = 2, k = v = 1.5, a/W1 = 0.5, W2 = 2W1, b/a = 1.

Equivalent domain integral

2.0 Enriched finite elements

1.5 a/W1 = 0.1, 0.2, 0.3, 0.4, 0.5

KIn

1.0 a/W1 = 0.1

a/W1 = 0.5
0.5

0.0
0.0 0.5 1.0 1.5 2.0 2.5
τ

Fig. 17. Normalized mode I stress intensity factor versus normalized time and relative crack length a/W1 considering plane strain
condition for the coating–substrate structure shown in Fig. 12, c1 = c2 = c12 = 2, d1 = d2 = d3 = 2, b12 = b13 = b31 = b32 = 1.5, x1 = 2,
k = v = 1.5, b = W1, W2 = 2W1.

strain. Crack spacing for this set of results is taken as b = W1. It is seen that the stress intensity factor
decreases as the relative crack length is increased from 0.1 to 0.5. In all cases, results calculated using EDI
agree well with those obtained by enriched finite elements. The effect of the variations in the thermal expansion

1.4
Equivalent domain integral
1.2
Enriched finite elements

1.0
δ2 = 0.2, 0.5, 1, 2, 5
0.8
KIn

0.6
δ2 = 0.2
0.4
δ2 = 5
0.2

0.0
0.0 0.6 1.2 1.8 2.4 3.0 3.6 4.2
τ

Fig. 18. Normalized mode I stress intensity factor versus normalized time and d2 considering plane strain condition for the coating–
substrate structure shown in Fig. 12, c1 = c2 = c12 = 2, b12 = b13 = b31 = b32 = 1.5, d1 = d3 = 2, x1 = 2, k = v = 1.5, a/W1 = 0.5, b = W1,
W2 = 2W1.
2824 S. Dag / Engineering Fracture Mechanics 73 (2006) 2802–2828

coefficient a2 on the transient stress intensity factors is given in Fig. 18. This figure shows that the exponent d2
significantly influences not only the magnitude of the stress intensity factor but also the time for KIn to

1.0

0.8
δ3 = 0.2, 0.5, 1, 2, 5
0.6
KIn

0.4
δ3 = 0.2
δ3 = 5
0.2

0.0
0.0 0.5 1.0 1.5 2.0 2.5
τ

Fig. 19. Normalized mode I stress intensity factor versus normalized time and d3 considering plane strain condition for the coating–
substrate structure shown in Fig. 12, c1 = c2 = c12 = 2, b12 = b13 = b31 = b32 = 1.5, d1 = d2 = 2, x1 = 2, k = v = 1.5, a/W1 = 0.5, b = W1,
W2 = 2W1.

1.0

0.8
ω1 = 2
0.6
KIn

ω1 = 5

0.4
ω1 = 0.2

0.2 ω1 = 0.5
ω1 = 1
0.0
0.0 0.4 0.8 1.2 1.6 2.0 2.4 2.8
τ

Fig. 20. Normalized mode I stress intensity factor versus normalized time and x1 considering plane strain condition for the coating–
substrate structure shown in Fig. 12, c1 = c2 = c12 = 2, b12 = b13 = b31 = b32 = 1.5, d1 = d2 = d3 = 2, k = v = 1.5, a/W1 = 0.5, b = W1,
W2 = 2W1.

Table 7
Transient normalized mode I stress intensity factors for various values of the exponent d1 computed using equivalent domain integral
(EDI), a/W1 = 0.5, b = W1, W2 = 2W1, c1 = c2 = c12 = 2, b31 = b13 = b12 = b32 = 1.5, d2 = d3 = 2, x1 = 2, k = v = 1.5
d1 s = 0.1 s = 0.2
Plane stress Plane strain Plane stress Plane strain
5 0.5424 0.6714 0.6935 0.8583
2 0.5424 0.6714 0.6935 0.8583
1 0.5424 0.6714 0.6935 0.8583
0.5 0.5424 0.6714 0.6935 0.8583
0.2 0.5424 0.6714 0.6935 0.8583
 S. Dag / Engineering Fracture Mechanics 73 (2006) 2802–2828 2825

approach to zero. Crack closure time and the maximum value of the stress intensity factor both increase as d2
is decreased from 5 to 0.2. Fig. 19 shows the sensitivity of the stress intensity factor to the variations in the
exponent d3. The curves are generated using the EDI method. Here, we observe that the effect of d3 is not
as pronounced as that of d2. Since the generated curves are close to each other, stress intensity factors com-
puted using enriched finite elements are not provided in order to more clearly present the trends. However,
agreement between the stress intensity factors computed using EDI and those generated by enriched elements
is again found to be quite good.
The influence of the variations in the principal thermal conductivity k1 on the mode I stress intensity factors
is presented in Fig. 20. In this figure, KIn values computed using the equivalent domain integral are plotted for
various values of the exponent x1. The effect of x1 on the time for KIn to approach to zero is seen to be rather
significant. Crack closure time is found to increase as the exponent x1 also increases. The peak value of the
stress intensity factor on the other hand goes through a minimum as x1 is increased from 0.2 to 5.
In the detailed parametric analyses conducted to examine the effect of the variations in a1 on KIn, it is
observed that d1 does not affect the stress intensity factors. Some sample results that support this statement
are provided in Table 7. This table presents the normalized stress intensity factors calculated using EDI at
two discrete points in time for five different values of the exponent d1. It can be seen that the calculated results
agree up to four significant digits for both of the cases of plane stress and strain. Therefore, it can be inferred
that variations in the thermal expansion coefficient a1 do not influence the transient stress intensity factors for
the problem shown in Fig. 12. This is a further verification of the conclusion drawn in Section 3.1 regarding
the influence of d1 on the mode I stress intensity factor.

5. Concluding remarks

In this study, a new computational method based on the generalized definition of the J-integral is developed
to study crack problems in orthotropic functionally graded materials under thermal stresses. Line integral form
of J is converted to an equivalent domain integral that can be calculated over an arbitrary area around the crack
tip. The developed technique is integrated into a fracture analysis software FRAC2D using the isoparametric
finite element concept and graded finite elements in order to calculate the thermal stress intensity factor.
Research finite element analysis code HEAT2D is used to compute temperature distribution profiles in ortho-
tropic functionally graded materials. Developed procedure is validated by comparing the stress intensity factors
computed using EDI to those calculated by an enriched finite element method and to those reported in the lit-
erature. The results obtained using the equivalent domain integral are shown to be highly accurate and indepen-
dent of the domain size used in the finite element computations. Extensive numerical results are presented
considering a single edge crack under steady-state heat flow and periodic cracks under transient thermal loading.
The results calculated for both cases point out that variations in the principal thermal expansion coefficient
parallel to the crack axis do not have any effect on the mode I thermal fracture behavior of orthotropic FGMs.
In the transient analyses conducted in Section 4, it is shown that the stress intensity factor goes through a max-
imum as the FGM coating–substrate structure is cooled down from a high processing temperature. The peak
values are found to increase with the increase in relative crack spacing. Stress intensity factors generally
become negative shortly after going through the maximum value which indicates the occurrence of crack clo-
sure at steady-state. Moreover, variations in the in-plane thermal expansion coefficient component perpendic-
ular to the crack axis and principal thermal conductivity are shown to significantly influence both the peak
values of the stress intensity factors and transient crack closure behavior. Thus, crack growth and thermal
fracture phenomena in graded orthotropic solids can be controlled by adjusting the gradation profiles of
the material properties. The computational method developed in the present study could provide useful feed-
back for the optimization of composition profiles, structural design and testing of orthotropic functionally
graded materials that are subjected to thermomechanical loads.

Acknowledgement

This work was partially supported by the Scientific and Technical Research Council of Turkey (TUBITAK)
through grant MISAG-TUN1-2004.
2826 S. Dag / Engineering Fracture Mechanics 73 (2006) 2802–2828

Appendix. Derivatives of the mechanical strain energy density function for plane strain

The derivatives of the mechanical strain energy density function that are used in Eq. (22) are given below
oW 1  
¼ ðA1 m31 þ 2A2 E1 Þe211 þ b1 E2e11e22 þ ðb1e11 þ b2e22 ÞE2e22
oE1 2H
E 1 b2 m31e233
 f ðA 2 E 1e 11 þ b 1 E 2 
e22 Þ
e 11 þ ðb 1e 11 þ b 2e 22 ÞE 2e 22 g þ ; ðA:1Þ
2H2 2m13
 
oW E1 m2 m13e11
¼ b1e22  32 e11 þ ðb1e11 þ b2e22 Þe22
oE2 2H m31
 
E1 2 m232 m13
þ f ðA E 
2 1 11e þ b E e
1 2 22 11 Þe þ ð b e
1 11 þ b e
2 22 ÞE 
e
2 22 g m 12 þ 2m m m
12 13 32 þ ; ðA:2Þ
2H2 m31
oW
¼ e212 þ e221 ; ðA:3Þ
oG12
oW E 1 E 2 b1 E1 E2e11e22
¼ 2
fðA2 E1e11 þ b1 E2e22 Þe11 þ ðb1e11 þ b2e22 ÞE2e22 g þ ; ðA:4Þ
om12 H H
oW E1
¼ fðA1e11  A2 E1e33 Þe11  ðA2 E1e11 þ b1 E2e22 Þe33  ðb1e33 þ m13e22 ÞE2e22 g
om31 2H
E1 ðA1  E1 m13 Þ E1e233
 fðA 2 E 1 
e 11 þ b 1 E 2e 22 Þe 11 þ ð b 1e 11 þ b 2e 22 ÞE 2e 22 g þ ; ðA:5Þ
2H2 2m13
 
oW E 1 E2 m2 e11
¼ m32e22  32 e11 þ ðm32e11  m31e22 Þe22
om13 2H m31
E1   
þ 2
fðA2 E1e11 þ b1 E2e22 Þe11 þ ðb1e11 þ b2e22 ÞE2e22 g E1 m31 þ E2 m232 =m31 þ 2m12 m32
2H
E1 m31e233
 ; ðA:6Þ
2m213
 
oW E 1 E2 2m32 m13e11
¼ m13e22  b1e33  e11 þ m13e11e22  2b2e33e22  b1e33e11
om32 2H m31
 
E1 E2 m32 m13
þ 2 fðA2 E1e11 þ b1 E2e22 Þe11 þ ðb1e11 þ b2e22 ÞE2e22 g þ m12 m13 ; ðA:7Þ
H m31
oW E1 DT
¼ fA2 E1e11 þ b1 E2e22 g; ðA:8Þ
oa1 H
oW E1 E2 DT
¼ fb1e11 þ b2e22 g; ðA:9Þ
oa2 H
oW E1 DT E1 m31e33 DT
¼ fðA2 E1 m31 þ b1 E2 m32 Þe11 þ ðb1 E2 m31 þ b2 E2 m32 Þe22 g þ ; ðA:10Þ
oa3 H m13
oW E1e11 E1 m31 a3e33
¼ fA2 E1 ðm31 a3 þ a1 Þ þ b1 E2 ðm32 a3 þ a2 Þg þ
oðDT Þ H m13
E1 E2e22
 fb1 ðm31 a3 þ a1 Þ þ b2 ðm32 a3 þ a2 Þg; ðA:11Þ
H
where,
m232 m13 m232 m13 E2
A1 ¼ E2 ; A2 ¼ 1  ; ðA:12Þ
m231 m31 E1
b1 ¼ m12 þ m13 m32 ; b2 ¼ 1  m31 m13 ; e33 ¼ a3 ðDT Þ: ðA:13Þ
The terms H, e11 and e22 in Eqs. (A.1)–(A.11) are given by Eqs. (12) and (13).
 S. Dag / Engineering Fracture Mechanics 73 (2006) 2802–2828 2827

References

[1] Kawasaki A, Watanabe R. Thermal fracture behavior of metal/ceramic functionally graded materials. Engng Fract Mech
2002;69:1713–28.
[2] Xing A, Jun Z, Chuanzhen H, Jianhua Z. Development of an advanced ceramic tool material – functionally gradient cutting ceramics.
Mater Sci Engng A 1998;248:125–31.
[3] Niino M, Kisara K, Mori M. Feasibility study of FGM technology in space solar power systems (SPSS). Mater Sci Forum
2005;492:163–8.
[4] Watari F, Yokoyama A, Omori M, Hirai T, Kondo H, Uo M, et al. Biocompatibility of materials and development to functionally
graded implant for bio-medical application. Compos Sci Technol 2004;64:893–908.
[5] Malinina M, Sammi T, Gasik M. Corrosion resistance of homogeneous and FGM coatings. Mater Sci Forum 2005;492–493:305–10.
[6] Sampath S, Herman H, Shimoda N, Saito T. Thermal spray processing of FGMs. MRS Bull 1995;20:27–31.
[7] Kaysser WA, Ilschner B. FGM research activities in Europe. MRS Bull 1995;20:22–6.
[8] Bahr HA, Balke H, Fett T, Hofinger I, Kirchhoff G, Munz D, et al. Cracks in functionally graded materials. Mater Sci Engng A
2003;362:2–16.
[9] Kokini K, Dejonge J, Rangaraj S, Beardsley B. Thermal shock of functionally graded thermal barrier coatings with similar thermal
resistance. Surf Coat Technol 2002;154:223–31.
[10] Ozturk M, Erdogan F. Mode I crack problem in an inhomogeneous orthotropic medium. Int J Engng Sci 1997;35:869–83.
[11] Ozturk M, Erdogan F. The mixed mode crack problem in an inhomogeneous orthotropic medium. Int J Fract 1999;98:243–61.
[12] Gu P, Asaro RJ. Cracks in functionally graded materials. Int J Solids Struct 1997;34:1–17.
[13] Guo LC, Wu LZ, Zeng T, Ma L. Mode I crack problem for a functionally graded orthotropic strip. Eur J Mech A/Solids
2004;23:219–34.
[14] Chen J, Liu Z, Zou Z. Transient internal crack problem for a nonhomogeneous orthotropic strip (Mode I). Int J Engng Sci
2002;40:1761–74.
[15] Guo LC, Wu LZ, Zeng T. The dynamic response of an edge crack in a functionally graded orthotropic strip. Mech Res Commun
2005;32:385–400.
[16] Dag S, Yildirim B, Erdogan F. Interface crack problems in graded orthotropic media: analytical and computational approaches. Int J
Fract 2004;130:471–96.
[17] Kim JH, Paulino GH. Mixed-mode fracture of orthotropic functionally graded materials using finite elements and the modified crack
closure method. Engng Fract Mech 2002;69:1557–86.
[18] Kim JH, Paulino GH. Mixed-mode J-integral formulation and implementation using graded finite elements for fracture analysis of
nonhomogeneous orthotropic materials. Mech Mater 2003;35:107–28.
[19] Kim JH, Paulino GH. The interaction integral for fracture of orthotropic functionally graded materials: evaluation of stress intensity
factors. Int J Solids Struct 2003;40:3967–4001.
[20] Kim JH, Paulino GH. T-stress in orthotropic functionally graded materials: Lekhnitskii and Stroh formalisms. Int J Fract
2004;126:345–84.
[21] Rice JR. A path independent integral and the approximate analysis of strain concentration by notches and cracks. ASME J Appl
Mech 1968;35:379–86.
[22] Nikishkov GP, Atluri SN. An equivalent domain integral method for computing crack-tip integral parameters in non-elastic,
thermomechanical fracture. Engng Fract Mech 1987;26:851–67.
[23] Wilson WK, Yu IW. The use of J-integral in thermal stress crack problems. Int J Fract 1979;15:377–87.
[24] Eischen JW. Fracture of nonhomogeneous materials. Int J Fract 1987;34:3–22.
[25] Gu P, Dao M, Asaro RJ. A simplified method for calculating the crack-tip field of functionally graded materials using domain
integral. ASME J Appl Mech 1999:101–8.
[26] Anlas G, Santare MH, Lambros J. Numerical calculation of stress intensity factors in functionally graded materials. Int J Fract
2000;104:131–43.
[27] Kim JH, Paulino GH. Finite element evaluation of mixed mode stress intensity factors in functionally graded materials. Int J Numer
Meth Engng 2002;53:1903–35.
[28] Kaya AC, Nied HF. Interface fracture analysis of bonded ceramic layers using enriched finite elements. In: Kokini K, editor.
Proceedings of the symposium on ceramic coatings, vol. 44. ASME MD; 1993. p. 47–71.
[29] Yildirim B, Erdogan F. Edge crack problems in homogeneous and functionally graded material thermal barrier coatings under
uniform thermal loading. J Therm Stresses 2004;27:311–29.
[30] Yildirim B, Dag S, Erdogan F. Steady state heat conduction in orthotropic functionally graded materials containing cracks: analytical
and computational techniques. In: Ziegler F, Heuer R, Adam C, editors. Proceedings of the sixth international symposium on thermal
stresses, vol. 2. Vienna, Austria, 2005. p. 517–20.
[31] Anderson TL. Fracture mechanics fundamentals and applications. Boca Raton: CRC Press; 1991.
[32] Santare MH, Lambros J. Use of graded finite elements to model the behavior of nonhomogeneous materials. ASME J Appl Mech
2000;67:819–22.
[33] Li C, Zou Z, Duan Z. Multiple isoparametric finite element method for nonhomogeneous media. Mech Res Commun
2000;27:137–42.
[34] Kim JH, Paulino GH. Isoparametric graded finite elements for nonhomogeneous isotropic and orthotropic materials. ASME J Appl
Mech 2002;69:502–14.
2828 S. Dag / Engineering Fracture Mechanics 73 (2006) 2802–2828

[35] Walters MC, Paulino GH, Dodds Jr RH. Stress intensity factors for surface cracks in functionally graded materials under mode-I
thermomechanical loading. Int J Solids Struct 2004;41:1081–118.
[36] Lee YD, Erdogan F. Interface cracking of FGM coatings under steady-state heat flow. Engng Fract Mech 1998;59:361–80.
[37] Shen HS. Postbuckling analysis of axially loaded functionally graded cylindrical panels in thermal environments. Int J Solids Struct
2002;39:5991–6010.
[38] Sevostianov I, Kachanov M. Plasma-sprayed ceramic coatings: anisotropic elastic and conductive properties in relation to
microstructure; cross-property relations. Mater Sci Engng A 2001;297:235–43.
[39] Parthasarathi S, Tittmann BR, Sampath K, Onesto EJ. Ultrasonic characterization of elastic anisotropy in plasma sprayed alumina
coatings. J Therm Spray Technol 1995;4:367–73.
[40] Leigh SH, Berndt CC. Modelling of elastic constants of plasma spray deposits with ellipsoid-shaped voids. Acta Mater
1999;47:1575–86.
[41] Sevostianov I, Kachanov M, Ruud J, Lorraine P, Dubois M. Quantitative characterization of microstructures of plasma-sprayed
coatings and their conductive and elastic properties. Mater Sci Engng A 2004;386:164–74.
[42] Callister WD. Materials science and engineering an introduction. New York: John Wiley and Sons, Inc.; 2000.
[43] Agarwal BD, Broutman LJ. Analysis and performance of fiber composites. New York: John Wiley and Sons, Inc.; 1990.
[44] Noda N, Hetnarski RB, Tanigawa Y. Thermal stresses. New York: Taylor and Francis; 2003.
[45] Shbeeb NI, Binienda WK. Analysis of an interface crack for a functionally graded strip sandwiched between two homogeneous layers
of finite thickness. Engng Fract Mech 1999;64:693–720.
[46] Erdogan F, Wu BH. Crack problems in FGM layers under thermal stresses. J Therm Stresses 1996;19:237–65.