International Journal of Solids and Structures 129 (2017) 167–176

Contents lists available at ScienceDirect

International Journal of Solids and Structures

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

A cohesive zone model and scaling analysis for mixed-mode

interfacial fracture
Shruti Jain a, Seung Ryul Na b, Kenneth M. Liechti b, Roger T. Bonnecaze a,∗
a
Department of Chemical Engineering, The University of Texas at Austin, Austin, TX 78712, USA
b
Department of Aerospace Engineering and Engineering Mechanics, Center for Mechanics of Solids, Structures and Materials, The University of Texas at
Austin, Austin, TX 78712, USA

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

Article history: A semi-analytical methodology is developed to study mixed-mode interface fracture that combines beam
Received 5 May 2017 theory and cohesive zone interactions. The method is validated with predictions from a non-linear com-
Revised 11 August 2017
mercial finite element package and results in the literature. Compared to commercial finite element pack-
Available online 5 September 2017
ages, the method is significantly faster ( > 10 0 0X) and robustly converges. Scaling equations are extracted
Keywords: that predict load, crack length, damage zone length and mode-mix. These equations can be used to ex-
Adhesion tract cohesive zone interactions from experimentally obtained load-displacement data.
Damage criteria © 2017 Elsevier Ltd. All rights reserved.
Delamination
Mixed-mode fracture
Interface
Cohesive zone model
Traction-separation relations

1. Introduction asperity shielding (Evans and Hutchinson, 1989) from interfacial

Bi-material interfaces are prevalent in multiple natural and en-
gineering applications from adhesive joints and composites to thin
film transfer and self-assembled monolayers (SAMs). To design
such layered structures which can potentially fail due to interface
cracking or to design a process requiring delamination, an under-
standing of mixed-mode interface fracture is of utmost importance.
Significant contributions to the understanding of bi-material
cracks were made through analytical approaches using linear elas-
tic fracture mechanics (Williams, 1959; Sih and Rice, 1964;
England, 1965; Comninou, 1977; Hutchinson et al., 1987;
Rice, 1988), experiments (Charalambides et al., 1989; Wang and
Suo 1990; Liechti and Chai 1992) and numerical simulations
(Lin and Mar, 1976; Smelser, 1979; Sun and Jih, 1987; Lee and
Chio, 1988). While many experimental investigations included an
adhesive interlayer and found the critical energy release rate to be
a function of the mode-mix, some experimental studies without
an interlayer (Ryoji et al., 1994; Banks-Sills et al., 20 0 0) obtained a
similar result. In the former, the toughening effect was attributed
to increased viscoplastic dissipation in the epoxy. Fiber bridging or
∗
Corresponding author at :Department of Chemical Engineering, 200 E Dean
Keeton CO400, Austin, TX 78712, USA.
E-mail address: rtb@che.utexas.edu (R.T. Bonnecaze).
roughness may have been the cause of toughening in the latter
studies.
Recently, cohesive zone models (CZMs) have gained popularity
for studying cracks to incorporate both toughness and strength pa-
rameters for the interfaces. The early approaches (Dugdale, 1960;
Barenblatt, 1962) were extended to include elastic and softening
interactions at the interface, and these models have been used ex-
tensively to study adhesion in double cantilever specimens and
predominantly mode I fracture (Kanninen, 1973; Chow et al., 1979;
Ungsuwarungsri and Knauss, 1987; Stigh, 1988; Williams and Ha-
davinia, 2002; Jain et al., 2016). In the case of adhesive interlayers,
experimental (Gowrishankar et al., 2012; Na et al., 2014; Wu et al.,
2016), analytical and numerical (Yuan et al., 2007; Mukherjee et al.,
2016) works have been presented to characterize interfaces and
develop scaling equations for pull-off force and cohesive zone size.
In this paper we consider interactions represented by mixed-
mode traction-separation relations (TSRs), which capture interac-
tions with chemical, molecular and electrostatic origins and can
also embody the effects of surface roughness and moisture. Pre-
vious mixed-mode fracture studies with linear beam theories and
CZMs (Szekrényes and Uj, 2004; Parmigiani and Thouless, 2007;
Thouless and Yang, 2008) employed finite element techniques and
developed equations for load as a function of physical parameters.
An equation for phase angle for multiple configurations was devel-
oped with tabulated coefficients (Li et al., 2004). However, closed

http://dx.doi.org/10.1016/j.ijsolstr.2017.09.002
0020-7683/© 2017 Elsevier Ltd. All rights reserved.
168 S. Jain et al. / International Journal of Solids and Structures 129 (2017) 167–176

(de Morais, 2014) and single leg bending (SLB) specimen (Lee et al.,
2010).
The fracture energy based criterion for damage initiation is
given by:
n t
+ > η, (1b)
nc tc
where η is a prescribed ratio less than unity,  n and  t are en-
ergy release rates in normal and shear directions, and  nc and  tc
are the critical energy release rates in normal and shear directions.
This condition was used for an MLDCB specimen (de Morais, 2013).
For our present work, a displacement based criterion for dam-
age initiation is used and is of the form

δ= w2 + u2 = δ0 , (1c)
Fig. 1. Schematic of a bi-linear traction-separation relation for an interface.
where w is the normal displacement and u is the shear displace-
ment at the interface. This condition has the simplicity of requiring
form solutions for crack length, damage zone length and mode- only one parameter i.e. the critical displacement δ 0 . The stress and
mix for a mixed-mode fracture configuration have not been pre- fracture energy based criteria in comparison require two or three
viously reported. Here, we present a one-dimensional (1-D) model parameters.
for mixed-mode peeling of a thin film from a rigid surface based
2.2. Failure criterion
on classical beam theory and a bi-linear TSR. The equations are
easily solved numerically and scaling analysis is used to derive al-
The most popular failure criterion is based on fracture energy
gebraic equations to correlate load, fracture length, damage zone
and is
length and mode-mix with the mixed-mode fracture properties of
the interface. The outline of the paper is as follows. Section 2 sum- n t
+ = 1. (2a)
marizes some of the most commonly used forms of bi-linear TSRs nc tc
for mixed-mode fracture. The model is developed in Section 3, fol- This condition has been successfully implemented in multiple
lowed by numerical validation in Section 4. The results of paramet- studies (Parmigiani and Thouless, 2007; Thouless and Yang, 2008;
ric analysis and scaling equations are presented in Section 5. Lee et al., 2010; de Morais, 2013, 2014).
Another popular criterion for failure and the one used in this
2. Cohesive zone models paper is the condition based on the total displacement at the in-
terface (Wu et al., 2016), namely
Cohesive zone models for interfaces have three main charac- 
teristics: the shape of the traction-separation relation (linear, bi- δ= w2 + u2 = δc , (2b)
linear, trapezoidal, exponential, polynomial or multi-linear), the where δ c is the critical value for fracture at the interface. Of course
criterion for damage initiation and the criterion for failure of the at failure, the normal and shear stress vanish at the interface. For
interfaces. In the case of mixed-mode fracture, interface strengths, total displacements between δ 0 and δ c , the stresses are assumed
interaction ranges and fracture energies may be defined separately to be linearly dependent on the displacements and the damage pa-
for both normal and tangential deformations along with a defini- rameter D (0 ≤ D ≤ 1) so that
tion of mode-mix based on energies or tractions. The review of
CZMs (Park and Paulino, 2013) provides a good summary of dif- σ = Kn (1 − D )w, τ = Ks (1 − D )u (3)
ferent models that have been used in fracture studies. This sec- and
tion briefly describes the bi-linear TSR and various criteria used
δC (δ − δ0 )
for damage initiation and failure. D= . (4)
Consider an interface with traction σ , stiffness K, opening dis-
δ (δC − δ0 )
placement δ and critical fracture energy release rate  c . In the Here it is assumed that the stiffness in the elastic region is the
case of a bi-linear TSR (Fig. 1) for Mode I, the interface first de- same for both normal and shear displacement, that is K = Kn = Ks .
forms elastically (region 1). Interface damage begins (and contin- In summary, we consider an interface with stiffness K in both
ues in region 2) once a damage initiation condition is met (δ = δ 0 normal and tangential directions, effective displacement δ 0 for
or σ = σ 0 ). The interface is fractured (region 3) on satisfying the damage initiation and effective displacement δ c for damage com-
failure criterion (δ = δ c or  =  c ). Criteria for mixed-mode fracture pletion. With these three parameters, for a bilinear TSR, the frac-
are comparatively more complex and some of them are described ture energy is given by Kδ 0 δ c /2 and the total adhesion strength σ T
below. is Kδ 0 . The stresses in each region can be expressed as
 
Kw, region 1 Ku, region 1
2.1. Damage initiation criterion
σ= K (1 − D )w, region 2 , τ = K (1 − D )u, region 2 .
0, region 3 0, region 3
The quadratic stress condition is based on the normal stress
σ and shear stress τ at the interface, and damage being initiated (5a)
when This model is equivalent to the interface behaving like a non-
 σ 2  τ 2
linear (in this case, bi-linear) spring where
+ = 1, (1a)
σ0 τ0 w u
σ = K (1 − D )δ, σ = σ, τ = σ, (5b)
where σ 0 and τ 0 are the normal and shear interface strengths. δ δ
Studies which have used this condition include analysis for where σ̄ is the magnitude of the summed normal and shear stress
a moment-loaded double cantilever beam (MLDCB) specimen vectors. The CZM formulation used in this study has the advantage
 S. Jain et al. / International Journal of Solids and Structures 129 (2017) 167–176 169

Fig. 2. (a) Schematic of peeling from a rigid substrate under fixed displacement; (b) free body diagram of an element of the top layer.

of only requiring three parameters. Further, no prior knowledge of From this asymptotic solution, an estimate of the length over
mode-mix is needed. The local mode-mix at any location can be which the effective displacement decays L∗1 is set by the loca-
determined as part of the solution using the following equation: tion where w1 = 0.01δ 0 . Knowing the solution in region 1, we ob-
u tain boundary conditions (10a) and (10b) for the elastic-damage
tan  = . (6) boundary and solve the equations for only regions 2 and 3 with
w
a shooting method because boundary locations depend on L2 and
The following section details the differential equations, bound- L3 which are not fixed. Note that Kh/E∗ > 16/3 is required for real
ary conditions and method used for simulating mixed-mode peel- values of exponential coefficients in region 1 (Eq. 9). For Kh/E∗ >
ing. 16/3, displacement in region 1 will have imaginary exponential co-
efficients and Eqs. (10a) and (10b) will never be satisfied. Based
3. Mixed-mode peeling model on continuity of displacements, rotation, moments and forces, the
boundary conditions for the system of differential equations are:
Consider a thin film or beam of width b, length L and thickness
h adhered to a rigid substrate as shown in Fig. 2(a). There is an ( α + β )h2 d 2 w2 d 3 w2 αβ h dw2
√ + h3 + =0 : x = L2 + L3 (10a)
initial crack of length l0 10h at the interface and the top layer is 2 d x2 d x3 2 dx
displaced at one end by .
The thickness of the film or beam and its deflection are such
d 3 w2 α h2 d2 w2 β 2 h dw2 αβ 2 w2
that beam theory can be used to describe the elastic forces. The −h3 3
− √ + + √ =0 : x = L2 + L3
normal deflection of the beam w with elastic modulus E, Poisson’s
dx 2 d x2 2 dx 2 2
ratio ν and moment of inertia I is obtained using the free body (10b)
diagram in Fig. 2(b) and given by
d4 w bh dτ (x ) d w2 d w3 d 2 w2 d 2 w3
E∗I − + bσ (x ) = 0, (7) w2 = w3 , = , = ,
d x4 2 dx dx dx d x2 d x2
3 3
where the shear displacement at the bottom surface of the thin d w2 d w3
film u = 2h dw and E∗ = E/(1 − ν 2 ) in case of plane strain. The cohe-
= : x = L3 (10c − f)
dx d x3 d x3
sive zone interactions are accounted for in the terms σ (x) and τ (x)
which are the normal and shear stress acting on the beam due to d 2 w3
w 3 = , =0 :x=0 (10g − h )
the interface. We consider an elastic region of length L∗1 , damage d x2
region of length L2 and crack length L3 ≥ l0 . Substituting the equa- and boundaries are defined at steady state by the equations:
tion for tractions from Eq. (5a) into Eq. (7), we have the following 
three differential equations for each region: δ= w22 + u22 = δ0 : x = L2 + L3 ,

4
d w3 δ= w22 + u22 = δc : x = L3 (10i − j )
E∗I =0 0 ≤ x ≤ L3 (8a)
d x4 While the damage zone is growing towards its steady state
  value, Eq. (10j) is not required since the boundary is known to be
d 4 w2 bh d h d w2 at x = l0 .
E∗I + bK (1 − D )w2 − K (1 − D ) =0
d x4 2 dx 2 dx The equations are solved using a shooting method with
MATLAB® . A guess is made for dw3 /dx and d3 w3 /dx3 at x = 0 to
L3 ≤ x ≤ L2 + L3 (8b)
solve Eq. (8a) for the cracked region followed by the damaged re-
gion Eq. (8b) and guesses are updated based on Eqs. (10a) and
  (10b) for convergence.
∗ d 4 w1 bh d h d w1
E I + bK w1 − K =0
d x4 2 dx 2 dx
4. Numerical validation
L2 + L3 ≤ x ≤ L∗1 + L2 + L3 (8c)
The 1-D model is validated by comparing the solutions
Eq. (8c) for the elastic response in region 1 has the following
with those from a full 2-D finite element ABAQUS 6.14® sim-
general solution:
    ulation in the next section. As a base problem, we consider
w1 αx βx the peeling of the thin film from a rigid surface where re-
= A exp − √ + B exp − √ (9)
 2h 2h alistic values are chosen for the mixed-mode TSR based on
  the experimental results of Wu et al. (2016). The top film is
where α = 3Kh − 2
∗ ) − E ∗ and β =
( 3EKh 48Kh 2
( 3EKh
∗ ) − E∗ + E∗
48Kh 3Kh and A and B silicon with E∗ = 165,500 MPa, h = 0.6 mm, b = 5 mm, L = 45 mm,
E∗
l0 = 12 mm, K = 1,60 0,0 0 0 MPa/mm, δ c = 0.721 μm, δ 0 = 0.052 μm
are constants of integration.
and  = 0.26 mm. The displacement of the film, stresses at
170 S. Jain et al. / International Journal of Solids and Structures 129 (2017) 167–176

Fig. 3. Results for FEM simulations and beam theory model: (a) Comparison of normal displacements along the beam; (b) comparison of shear displacement along the beam.

Fig. 4. Results for FEM simulations and beam theory model: (a) Comparison of normal stresses at the interface; (b) comparison of shear stresses at the interface.

the interface and load-displacement curve are extracted from
FEM simulations in ABAQUS® with 4-node bilinear plane strain
quadrilateral elements with reduced integration for the sub-
strates with a square mesh size of 0.2 mm for the bottom sub-
strate, a finer mesh for the top layer and surface based cohesive
behavior.
Comparisons for normal and shear displacements of the thin
film are shown in Fig. 3. Excellent agreement is seen in both cases.
Normal and shear stresses at the interface are compared in Fig. 4
where the position of the crack tip matched within 0.09% and the
damage zone length matched within 7% but the peak values are
very different. We attribute this difference to our use of asymptotic
boundary conditions rather than solving the elastic region whereas
ABAQUS® solves for displacements along the entire length of the
beam. At damage initiation, this led to local mode-mix values of
−51° and −74° in the finite element and beam analyses, respec-
tively. In addition, the limited number of options for damage initi-
ation criteria in ABAQUS® did not include the one selected for the
beam analysis. The one selected for the finite element analysis was
the closest possible one, but the traction values for damage initi-
ation did differ by the amounts shown in Fig. 4, pointing to the
internal consistency of each analysis.
The comparison for load-displacement values is shown in
Fig. 5 and the results are in close agreement. These comparisons
validate our beam theory model. Note that the 1-D finite differ-
ence shooting method takes a few seconds to produce a converged
solution whereas commercial software can take hours to solve a
cohesive zone interface fracture problem and requires considerable
troubleshooting for convergence. These advantages show the im-
portance of having a simpler semi-analytical framework to model
interface fracture.
5. Results and discussion
Parametric analysis is performed to study the influence of dif-
ferent material properties, layer thickness and TSR parameters on
the resultant load P, damage zone length L2 , crack length L3 and
mode-mix among others. Scaling equations are obtained to deter-
mine the exact dependence of parameters on fracture. The base
 S. Jain et al. / International Journal of Solids and Structures 129 (2017) 167–176 171

Table 1
Material properties, geometry and TSR parameters for the base case.

E∗ (MPa) L (mm) b (mm) h (mm) l0 (mm) K (MPa/mm) δ c (μm) δ 0 (μm)  (mm)  c (J/m2 )
165,500 45 5 0.6 12 1480,0 0 0 0.721 0.052 0.3 27.74

the damage zone and the energy release rate in the damage zone
at steady state. It can be seen that the damage zone length re-
mains fairly constant during steady state crack growth and so does
the energy release rate. The mode-mix at the crack tip as a func-
tion of end displacement is shown in Fig. 8(a). It maintains a value
of approximately −50.2° which falls in the range of mode-mix
angles reported for interface cracks in a bilayer (Hutchinson and
Suo, 1991). All of these trends are consistent with characteris-
tics of crack growth and support the applicability of our effective
displacement-based cohesive zone model for mixed-mode fracture.
Interface stresses are extracted from simulation results and
compared with the input bilinear TSR as shown in Fig. 8(b), where
displacements are normalized by the maximum displacement in
respective directions. The input and output TSRs match perfectly,
and shear stresses contribute the most to interface stresses. At the
damage initiation point where the effective stress peaks at Kδ 0 ,
shear stress is at its maximum value but normal stresses continue
to increase initially during shear softening. Variations in TSR pa-
rameters and layer thickness change this behavior and it is dis-
cussed later on in Section 5.2.6.

Fig. 5. FEM and beam theory results for the load-displacement curve.
5.2. Parametric analysis

case is presented first in this section, followed by results of para-
metric analysis.
5.1. Base case
We consider a base case of a silicon layer with simulation pa-
rameters mentioned in Table 1.
Fig. 6(a) shows the load–displacement curve and Fig. 6(b)
shows the length of the crack for increasing displacement. The
load increases linearly with displacement until steady state crack
growth begins and then decreases as the crack grows. Crack length
does not grow from its initial value of l0 until the onset of steady
state crack growth. Fig. 7 (a and b) illustrates the development of
For parametric analysis of steady state crack growth, E∗ was var-
ied from 120,0 0 0–165,50 0 MPa, h was varied from 0.6–1.1 mm, K
was varied from 1480,0 0 0–2,70 0,0 0 0 MPa/mm, δ c was varied from
0.5–1.2 μm, δ 0 was varied from 0.035–0.09 μm and  was varied
from 0.22–0.3 mm. Note that the window where Kh/E∗ > 16/3 and
cracks grow in steady state sets the bounds on the parameter space
for this analysis. Next we discuss parametric analysis and scaling
arguments to predict load, crack length, damage zone length and
mode-mix for steady state crack growth.
5.2.1. Load
Results show that the load required to peel the layer increases
with increase in K, E∗ , δ c , δ 0 and h but decreases with . We use

Fig. 6. (a) Load-displacement curve for the base case; (b) crack length as a function of end displacement.
172 S. Jain et al. / International Journal of Solids and Structures 129 (2017) 167–176

Fig. 7. (a) Damage zone length as a function of displacement; (b) energy release rate in the damage zone normalized by the fracture toughness at steady state.

Fig. 8. (a) Mode-mix as a function of displacement during stead-state crack growth; (b) Comparison of input and output TSR.

the beam theory relationship for fracture in absence of cohesive which simplifies in cases of long initial cracks (or   δ c ) to:
forces (Anderson, 2005):
  P = 0.31(K δ0 δc h )3/4 bE ∗1/4 /1/2 . (13b)
 dP
c = − , (11)
2b d L3

It can be seen in Fig. 9 that simulation results from the para-
and in the cracked region of length L3 where w increases from metric analysis agree well with the scaling equation. The values
δ c to , we can approximate the third derivative to find that in the legend represent the range of variation for each parame-
d3 w −δc
E ∗ I = dx3 |x=0 ∼ 3 , which means L3 ∼ [ ( − δc )E I/P ]
P ∗ 1/3 . For a ter with units being the same as in Table 1. For example, when
L3
bi-linear TSR where  c = Kδ 0 δ c /2 or simply c ∼ K δ0 δc , which is h is varied between 0.6 mm to 1.1 mm, other parameters are kept
substituted into Eq. (11) along with scaling relationship for L3 , it is constant at the values mentioned in Table 1. It should be noted
found that that Eq. (13b) for mixed-mode peeling has the parametric depen-
dence as load for a DCB configuration obtained in our previous
E ∗ h3 ( − δc )
K δ0 δc ∼  4/3 (12) work (Jain et al., 2016) but the fracture energy in mixed-mode is
( − δc )E ∗ I the total fracture energy including contributions from both normal
P and shear stresses. This equation for the load for the DCB configu-
ration is clearly very similar to that presented here in Eq. (13b).
∗1/4 (−δ 1/4
c) In the dimensionless form, Eq. (13b) can be written as
and on rearrangement yields P ∼ O(1 )(K δ0 δc h )3/4 b E 3/4
.
After data fitting, we find that:
 3/4  1 / 2
1/4 P K δ0 δc h
E ∗1/4 ( − δc ) = 0.31 . (13c)
P = 0.31(K δ0 δc h )3/4 b (13a) E∗h E∗h 
3 / 4
 S. Jain et al. / International Journal of Solids and Structures 129 (2017) 167–176 173

Fig. 11. Comparison of simulations with empirically fitted Eq. (17) for the length of
Fig. 9. Parametric analysis for load and comparison with scaling equation. Paren-
the damage zone length L2 .
thetical values are the range of the parameters explored for the simulations.

5.2.2. Crack length The critical end displacement c at which the damage zone is
In region 3, we know that load is proportional to the third fully developed and steady -state crack growth will begin can be
derivative and we can use a scaling argument for the derivative calculated by solving Eq. (14a) for L3 = l0 . For   δ c , c is given
to obtain an equation for crack length- by
d3 w −δc
E ∗ I = dx3 |x=0 ∼
P
3 and on substituting Eq. (13a) for P, we
L3
1 / 2
Kδ δ
get: c = 1.23 ∗0 3c l02 . (15)

1 / 4 E h
∗ 3
E h
L3 = 0.9 1/4 ( − δc )1/4 (14a)
K δ0 δc
5.2.3. Rotation at loading point
which can be written in a dimensionless form as:
The shooting method in this study relies on two guesses at the

1 / 4  1 / 4  1/4
L3 E∗h   − δc loading point - dw/dx and d3 w/dx3 . A good guess for the latter can
= 0.9 . (14b) be estimated from Eq. (13a) and for the former, we use another
h K δ0 δc h h
scaling argument to estimate
Fig. 10(a) compares the parametric analysis results from simula- 
tions to results from the scaling Eq. (14a) and excellent agreement dw   − δc
∼ .
is observed. dx  L3
x=0

Fig. 10. (a) Parametric analysis for crack length L3 and comparison with scaling equation; (b) comparison of rotation at the loading point as obtained from simulations
versus scaling equation results.
174 S. Jain et al. / International Journal of Solids and Structures 129 (2017) 167–176

Fig. 12. (a) Comparison of damage zone energy release rates obtained from simulations and equation for bilinear TSR; (b) Comparison of mode-mix in simulations with
results from the empirical equation.

Using Eq. (14a), we find Our consideration of shear at the interface gives similar para-
  1/4 metric dependence as mode I (Eqs. (18a) and (18b)) but differ-
dw  K δ0 δc ( − δc )3/4 ent power law dependences on TSR parameters. Eq. (17) suggests
= 1.61
dx  E∗ 1 / 4 h 3 / 4 that materials with higher elastic moduli and thicknesses will have
x=0
 1/4  1/2 longer damage zones and so will interfaces with lower strength.
K δ0 δc 
∼ 1.61 (for δc ). (16)
E∗h h 5.2.5. Energy release rate in the damage zone
Clearly, the damage zone energy for a bilinear TSR is given
A parity plot comparing the values from simulation with the
by  d = Kδ 0 (δ c − δ 0 )/2 but it can also be estimated from simula-
scaling equation is shown in Fig. 10(b) and excellent agreement be-
tions by calculating the J-integral over the vectorial TSR σ (δ ) as
tween the two can be seen.

d = δδc σ dδ . The energies calculated from these two methods are
0
5.2.4. Damage zone length compared as another validation method for our CZM framework
Two possible scaling arguments can be made for the length of and excellent agreement is seen in Fig. 12(a).
the damage zone based on the requirement that
  5.2.6. Mode-mix at crack tip and softening at the damage initiation
d3 w  d3 w  point
= .
d x3  d x3  For same interface stiffness in both normal and tangential di-
x=0 x=L3
rections, we define mode-mix from the ratio of shear and normal
δ −δ −δ0
Thus, it is expected that −3δc ∼ c 3 0 or −3δc ∼ . displacements (Eq. 6) at the crack tip, so that
L3 L2 L3 (L3 +L2 )3 u 
However, these resulting scaling equations do not yield results tan c =
c
. (19)
in agreement with simulations. This is because the normal dis- wc
placements can vary significantly from the vectorial critical dis- In the absence of a simple differential equation in the damage
placements. Instead the empirical equation zone, we empirically obtain the following dependence of mode-
 1 / 2  1/3 mix on various parameters
δc E∗
L2 = 0.4 h2/3 . (17)  0.37  0.3
δ0 K K δ0 h
tan c = −2.7 . (20)
for the length of the damage zone is determined based on dimen- E∗ δc
sional analysis and data fitting. Fig. 11 shows the comparison be- Eq. (20) suggests a higher mode-mix for thicker layers with
tween simulations and this empirical equation. low elastic moduli and interfaces with high effective strengths and
CZM analysis with only normal stresses for the peeling process lower interaction ranges. The larger the difference between the
(Williams and Hadavinia, 2002) considered a characteristic damage separation at damage initiation and the critical separation for fail-
length l and the damage zone length was simply a function of this ure in the system, the closer it is to predominantly mode I crack
characteristic length L2 ∼ f (h, l ) where growth. Since there is a lower bound of 16/3 on the combination
    Kh/E∗ , the two separation parameters will have to differ by three
2 E c 2 E δc
l= = (18a) to four orders of magnitude to have a mode-mix under 20°. The
3 σ2 3 K δ0
parity plot comparing simulation results with empirical equation
and bi-linear TSR scaling analysis for mode I (Jain et al., 2016) is shown in Fig. 12(b). For the peeling configuration considered
yielded: here, the resulting mode-mix is always higher than −45° indicating
 1/4 significant contributions from shear at the crack tip.
E ∗ h3 (δc − δ0 ) One limitation of mixed-mode CZMs is that they can provide
L2 = 1.1 . (18b)
K δ0 positive stiffness during softening which is undesirable unless the
 S. Jain et al. / International Journal of Solids and Structures 129 (2017) 167–176
Fig. 13. (a) Comparison of normal stress at damage initiation versus peak normal stress during damage; (b) comparison of shear stress at damage initiation versus peak
shear stress during damage.
material demonstrates stiffening behavior (Park and Paulino, 2013).
In these simulations, we observe that certain parameter values re-
sult in positive stiffness in either the normal or tangential direc-
tion during damage. Fig. 13 compares the stress at damage initi-
ation to the peak stress attained during damage. While the shear
stress at initiation was equal to the peak shear stress for thin lay-
ers irrespective of TSR parameters, the normal stress at initiation
was equal to the maximum normal stress only for low values of
δ c , low values of δ 0 and thick layers. This parametric dependence
could guide the characterization of an interface based on its stiff-
ening behavior during damage.
5.2.7. Length of elastic region
It can be seen from Eqs. (8c) and (9) that scaling analysis is
not straightforward because of multiple terms of similar orders of
magnitude. An estimate of the length of the elastic region is ob-
tained from the asymptotic solution (Eq. 9) by using derivatives at
the boundary of region 1 and 2 from the shooting method solution
to find the unknown constants A and B, and then estimating the
length at which effective displacement reduces to 0.01 δ 0 . Results
indicate the following dependence:
 1/4
E ∗ δc
L∗1 ∼ f (h, δ0 ).
K the damage zone length. Another important aspect is that scaling
f(h, δ 0 ) could not be estimated by empirical power law fitting be-
cause there was a non-monotonous change in L∗1 with variation in
h and δ 0 . From Eq. (9), we can infer that the characteristic dimen-
sionless length would be a function of E∗ /Kh. In comparison, for a
triangular TSR without any damage region and considering shear
deformations in the beam but only normal damage at the inter-
face, the following equation was derived (Williams and Hadavinia,
2002):
 1/4
E δc
L∗1 ∼ h3/4
K δ0
with δ c = δ 0 in the absence of a damage zone. Eqs. (21) and
(22) have exactly the same dependence on E, K and δ c . Clearly,
the elastic zone region increases for materials with higher elastic
modulus and interfaces with lower stiffness.
In the entire analysis presented in this section, two most promi-
nent dimensional groups in the scaling equations are E∗ /σ T and
175
E∗ h/ c reinforcing the importance of effective interface strength
and fracture energy in crack growth.
6. Conclusions
A three-parameter and bilinear mixed-mode TSR is proposed
with K, δ 0 and δ c as the effective stiffness, effective separation
at initiation and separation at fracture, respectively. A shooting
method formulation is used with a 1-D classical beam model and
the asymptotic solution for the elastic region to study mixed-mode
interactions during fracture. Interface strength, mode-mix and en-
ergy release rate are extracted as results of the simulation. The
observed trends for steady-state crack growth are in agreement
with established characteristics of CZM based fracture models. The
model is also validated using the commerical finite element pack-
age ABAQUS® . Algebraic equations for steady state crack growth
are extracted from the 1-D model for load, crack length, criti-
cal displacement, rotation at load point, damage zone length and
mode-mix to clearly identify their parametric dependence on ma-
terial properties and interface properties. The equations show that
the two most important parameters in this mixed-mode TSR are
the strength σ T = Kδ 0 and effective separation at fracture δ c . Values
of K and δ 0 individually may only affect the elastic zone length and
(21)
equations for mode I and mixed-mode fracture have similar forms
when using a simple vectorial TSR analogous to a mode I bi-linear
TSR.
Acknowledgements
The authors gratefully acknowledge financial support from the
National Science Foundation Nanosystems Engineering Research
Center on Nanomanufacturing Systems for Mobile Computing
and Mobile Energy Technologies (NASCENT)- NSF EEC Grant No.
(22) 1160494. S.J. would like to thank Chenglin Wu for discussions on
his work on mixed-mode fracture.
References
Anderson, T.L., 2005. Fracture Mechanics: Fundamentals and Applications. CRC
Press, Boca Raton, Florida.
Banks-Sills, L., Travitzky, N., Ashkenazi, D., 20 0 0. Interface fracture properties of a
bimaterial ceramic composite. Mech. Mater. 32 (12), 711–722.
176 S. Jain et al. / International Journal of Solids and Structures 129 (2017) 167–176
Barenblatt, G.I., 1962. The mathematical theory of equilibrium cracks in brittle frac-
ture. Advances in Applied Mechanics vol. 7, 55–129.
Charalambides, P.G., Lund, J., Evans, A.G., McMeeking, R.M., 1989. A test specimen
for determining the fracture resistance of bimaterial interfaces. J. Appl. Mech.
56 (1), 77–82.
Chow, C.L., Woo, C.W., Sykes, J.L., 1979. On the determination and application of cod
to epoxy-bonded aluminium joints. J. Strain Anal. Eng. Des. 14 (2), 37–42.
Comninou, M., 1977. The interface crack. J. Appl. Mech. 44 (4), 631–636.
de Morais, A.B., 2013. Simplified cohesive zone analysis of mixed-mode I–II delami-
nation in composite beams. Polym. Compos. 34 (11), 1901–1911.
de Morais, A.B., 2014. Cohesive zone beam modelling of mixed-mode I–II delamina-
tion. Compos. Part A 64, 124–131.
Dugdale, D.S., 1960. Yielding of steel sheets containing slits. J. Mech. Phys. Solids 8
(2), 100–104.
England, A.H., 1965. A crack between dissimilar media. J. Appl. Mech. 32 (2),
400–402.
Evans, A.G., Hutchinson, J.W., 1989. Effects of non-planarity on the mixed mode
fracture resistance of bimaterial interfaces. Acta Metall. 37 (3), 909–916.
Gowrishankar, S., Mei, H., Liechti, K., Huang, R., 2012. A comparison of direct and
iterative methods for determining traction-separation relations. Int. J. Fract. 177
(2), 109–128.
Hutchinson, J., Suo, Z., 1991. Mixed mode cracking in layered materials. Adv. Appl.
Mech. (29) 63–191.
Hutchinson, J.W., Mear, M.E., Rice, J.R., 1987. Crack paralleling an interface between
dissimilar materials. J. Appl. Mech. 54 (4), 828–832.
Jain, S., Na, S.R., Liechti, K.M., Bonnecaze, R.T., 2016. Characteristic scaling equations
for softening interactions between beams. Int. J. Fract. 201 (1), 1–9.
Lee, M.J., Cho, T.M., Kim, W.S., Lee, B.C., Lee, J.J., 2010. Determination of cohesive
parameters for a mixed-mode cohesive zone model. Int. J. Adhes. Adhes. 30 (5),
322–328.
Lee, K.Y., Choi, H.J., 1988. Boundary element analysis of stress intensity factors for
bimaterial interface cracks. Eng. Fract. Mech. 29 (4), 461–472.
Kanninen, M.F., 1973. An augmented double cantilever beam model for studying
crack propagation and arrest. Int. J. Fract. 9 (1), 83–92.
Li, S., Wang, J., Thouless, M.D., 2004. The effects of shear on delamination in layered
materials. J. Mech. Phys. Solids 52 (1), 193–214.
Liechti, K.M., Chai, Y.S., 1992. Asymmetric shielding in interfacial fracture under in–
plane shear. J. Appl. Mech. 59 (2), 295–304.
Lin, K.Y., Mar, J.W., 1976. Finite element analysis of stress intensity factors for cracks
at a bi-material interface. Int. J. Fract. 12 (4), 521–531.
Mukherjee, B., Batra, R.C., Dillard, D.A., 2016. Edge debonding in peeling of a thin
flexible plate from an elastomer layer: a cohesive zone model analysis. J. Appl.
Mech..
Na, S.R., Suk, J.W., Ruoff, R.S., Huang, R., Liechti, K.M., 2014. Ultra long-range inter-
actions between large area graphene and silicon. ACS Nano 8 (11), 11234–11242.
Park, K., Paulino, G.H., 2013. Cohesive zone models: a critical review of trac-
tion-separation relationships across fracture surfaces. Appl. Mech. Rev. 64 (6)
060802-060802-060820.
Parmigiani, J.P., Thouless, M.D., 2007. The effects of cohesive strength and tough-
ness on mixed-mode delamination of beam-like geometries. Eng. Fract. Mech.
74 (17), 2675–2699.
Rice, J.R., 1988. Elastic fracture mechanics concepts for interfacial cracks. J. Appl.
Mech. 55 (1), 98–103.
Ryoji, Y., Jin-Qiao, L., Jin-Quan, X., Toshiaki, O., Tomoyoshi, O., 1994. Mixed mode
fracture criteria for an interface crack. Eng. Fract. Mech. 47 (3), 367–377.
Sih, G.C., Rice, J.R., 1964. The bending of plates of dissimilar materials with cracks.
J. Appl. Mech. 31 (3), 477–482.
Smelser, R.E., 1979. Evaluation of stress intensity factors for bimaterial bodies using
numerical crack flank displacement data. Int. J. Fract. 15 (2), 135–143.
Stigh, U., 1988. Damage and crack growth analysis of the double cantilever beam
specimen. Int. J. Fract. 37 (1), R13–R18.
Sun, C.T., Jih, C.J., 1987. On strain energy release rates for interfacial cracks in bi–
material media. Eng. Fract. Mech. 28 (1), 13–20.
Szekrényes, A., Uj, J., 2004. Beam and finite element analysis of quasi-unidirectional
composite SLB and ELS specimens. Compos. Sci. Technol. 64 (15), 2393–2406.
Thouless, M.D., Yang, Q.D., 2008. A parametric study of the peel test. Int. J. Adhes.
Adhes. 28 (4–5), 176–184.
Ungsuwarungsri, T., Knauss, W.G., 1987. The role of damage-softened material be-
havior in the fracture of composites and adhesives. Int. J. Fract. 35 (3), 221–241.
Wang, J.S., Suo, Z., 1990. Experimental determination of interfacial toughness curves
using Brazil-nut-sandwiches. Acta Metall. Mater. 38 (7), 1279–1290.
Williams, J.G., Hadavinia, H., 2002. Analytical solutions for cohesive zone models. J.
Mech. Phys. Solids 50 (4), 809–825.
Williams, M.L., 1959. The stresses around a fault or crack in dissimilar media. Bull.
Seismol. Soc. Am. 49 (2), 199–204.
Wu, C., Gowrishankar, S., Huang, R., Liechti, K.M., 2016. On determining mixed-mode
traction–separation relations for interfaces. Int. J. Fract. 202 (1), 1–19.
Yuan, H., Chen, J.F., Teng, J.G., Lu, X.Z., 2007. Interfacial stress analysis of a thin
plate bonded to a rigid substrate and subjected to inclined loading. Int. J. Solids
Struct. 44 (16), 5247–5271.