Analysis of Energy Balance When

Using Cohesive Zone Models to

C. Shet Simulate Fracture Processes
N. Chandra
Fellow ASME Cohesive Zone Models (CZMs) are being increasingly used to simulate fracture and
fragmentation processes in metallic, polymeric, and ceramic materials and their compos-
Department of Mechanical Engineering, ites. Instead of an infinitely sharp crack envisaged in fracture mechanics, CZM presup-
FAMU-FSU College of Engineering, poses the presence of a fracture process zone where the energy is transferred from exter-
Florida State University, nal work both in the forward and the wake regions of the propagating crack. In this paper,
Tallahassee, FL 32312 we examine how the external work flows as recoverable elastic strain energy, inelastic
strain energy, and cohesive energy, the latter encompassing the work of fracture and other
energy consuming mechanisms within the fracture process zone. It is clearly shown that
the plastic energy in the material surrounding the crack is not accounted in the cohesive
energy. Thus cohesive zone energy encompasses all the inelastic energy e.g., energy re-
quired for grainbridging, cavitation, internal sliding, surface energy but excludes any
form of inelastic strain energy in the bounding material. 关DOI: 10.1115/1.1494093兴

1 Introduction Table 1 shows a compilation of some of the popular CZMs

available in the literature with specific attention focused on the
Fracture studies are usually carried out under several idealized
shape and the values of the model parameters. Needleman was
conditions, as in the case of linear elastic fracture mechanics or
one of the first to use polynomial and exponential types of
the case of small scale yielding. In such cases, the details of the
traction-separation equations to simulate particle debonding in
local crack tip fields are uniquely characterized by a single mac-
metal matrices 关16,18兴. Xu and Needleman 关10,14兴 further used
roscopic parameter such as the stress intensity factors (K I,II,III ) or
the above models to study the void nucleation at the interface of
corresponding energy release rates (GI,II,III ). These global param-
particle and matrix material, fast crack growth in brittle material
eters are related to the corresponding material parameters typi-
under dynamic loading, and dynamic crack growth along the in-
cally the fracture toughness 共K IC . . . and GIC . . . 兲 that determine
terface of bimaterials. Tvergaard and Hutchinson 关19兴 used a trap-
the critical conditions of initiation of crack growth. When the
ezoidal shape of the traction-separation model to calculate the
crack tip experiences plastic yielding, the above concepts based
crack growth resistance in elasto-plastic materials. Tvergaard 关20兴
purely on the theory of elasticity are not valid and have led to the
used a quadratic traction-displacement equation to analyze inter-
introduction of a path independent J Integral 关1兴, which is strictly
faces. Camacho and Ortiz 关9兴 employed a linear cohesive fracture
valid for a nonlinear elastic material. The property of path inde-
model to propagate multiple cracks along arbitrary paths during
pendence is lost if the energy near the crack tip region is con-
impact damage in brittle materials. Geubelle et al. 关21兴 utilized a
verted into significant inelastic energy due to plasticity or when
bilinear CZM to simulate spontaneous initiation and propagation
the material locally unloads during the propagation process. The
of transverse matrix cracks and delamination fronts in thin com-
fracture mechanics analysis presupposes the existence of an infi-
posite plates subjected to low-velocity impact. It is important to
nitely sharp crack leading to the singular crack tip fields. How-
note that the magnitude of the parameters in CZMs vary widely
ever, in real materials neither the sharpness of the crack nor the
ranging from MPa to GPa for traction, Joules to Kilo Joules for
stress levels near the crack tip region can be infinite. Further, for
energy and nanometers to micrometers for the separation distance.
cracks along bimaterial interfaces, the crack tip will no longer be
In all the CZMs 共except Dugdale’s model and Camacho et al.’s
embedded in a square-root singular stress field leading to a con-
model, refer to Table 1兲, the traction-separation relations for inter-
dition that stress intensity factor may either be zero or infinity
faces are such that with increasing interfacial separation, the trac-
共Atkinson 关2兴, Wapling et al. 关3兴兲. As an alternative approach to
tion across the interface reaches a maximum, then decreases and
this singularity driven fracture approach, Barrenblatt 关4,5兴 and
eventually vanishes, permitting a complete decohesion 共see Fig. 1
Dugdale 关6兴 proposed the concept of CZM. CZM has evolved as a
for a typical variation兲. The main difference lies in the shape of
preferred method to analyze fracture problems in monolithic and
the traction-displacement response, and the constants that are used
composite material systems not only because it avoids the singu-
to describe that shape. There is a common belief that CZMs can
larity but also because it can be easily implemented in a numerical
be described by two independent parameters 关7,22,23兴. These pa-
method of analysis as in finite element or boundary element
rameters may be two of the three parameters, namely the cohesive
method.
energy ␾, and either of the cohesive strength ␴ max , or the separa-
CZMs have been used to simulate the fracture process in a
tion length ␦ se p . It should be noted that cohesive energy ␾, sur-
number of material systems including polymers 关7兴, metallic ma-
face energy and work of fracture are all expressed as energy per
terials 关8兴, ceramic materials 关9兴, bimaterial systems in polymer
unit area and hence are truly ‘‘rate quantities.’’ In general, cohe-
matrix composites 关10兴, metal matrix composites 关11兴, and fiber
reinforced plastic composites 关12兴. They have been used to simu- sive energy ␾ is obtained from experiments and is believed to be
late fracture under static 关13,14兴, dynamic 关9,10兴, and cyclic 关15兴 equivalent to the work of fracture ⌫; the latter quantity identified
loading conditions. as, for example, J IC . This relationship between ␾ and J IC is very
critical not only from the mechanics perspective but also in the
numerical solution to the fracture problem using CZMs. This pa-
Contributed by the Materials Division for publication in the JOURNAL OF ENGI-
NEERING MATERIALS AND TECHNOLOGY. Manuscript received by the Materials
per examines the origin of external work, dissipative work 共in-
Division June 5, 2001; revised manuscript received February 6, 2002. Associate cluding plastic work兲, recoverable elastic work, and surface en-
Editor: M. Zhou. ergy in a fracture process. Also, we seek to understand the

440 Õ Vol. 124, OCTOBER 2002 Copyright © 2002 by ASME Transactions of the ASME

Downloaded From: http://materialstechnology.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jemta8/27039/ on 04/23/2017 Terms of Use: http://www.asme.or

 Table 1 Various cohesive zone models and their parameters

relationship among these energy terms as a function of material, accounted for only through the traction-displacement relations.
geometry 共crack tip兲, and cohesive zone parameters. For example, plastic dissipation in the surrounding 共bounding兲
In order to achieve the above objectives of examining the en- material is not accounted for in the process zone. As discussed
ergetics during the fracture process in a ductile material, we raise already, though, there are a number of T⫺ ␦ forms available in
the following three specific questions. literature, a typical curve is shown in Fig. 1. In the figure point A
refers to a point which is yet to separate, point C corresponds to
1 Since the energy flows into the crack throughout the length
the maximum normal/shear cohesive traction, and point E where
of the fracture process zone 共FPZ兲 共the crack face traction being
the traction is once again zero where complete separation has
nonzero in this region兲, what is the rate at which the energy flows
taken place. The length of fracture process zone is the length
in the forward and wake regions of the FPZ during initiation,
along the crack face 共say in the x-direction兲 from the point where
steady-state, and stages in-between?
2 How does the macroscopic energy release rate relate to the ␦ ⫽u y ⫽0 ⫹ 共point A兲 to the point of complete separation occurs
cohesive energy and other inelastic processes in the body? ␦ ⫽ ␦ se p 共point E兲. The location of a crack tip within the cohesive
3 Is there a connection between the energy distribution ACF process zone cannot be uniquely identified. For the present pur-
and CFE of T⫺ ␦ curve in Fig. 1 and processes in the forward and pose, let us assume that to be at point C. With that definition, we
the wake regions of the FPZ? can identify a forward region as the region along the length direc-
tion, corresponding to point A from that of point C, while the
In the next section, we will introduce the practical issues in- wake region C to E, as shown in Fig. 2. The forward region is
volved in using CZM to boundary value problems. In that section identified as length l 1 and the wake as length l 2 in Fig. 1.
we will present some of the basic parameters involved in the
Where is the Crack Tip in CZM? In the classical fracture
model and illustrate the implications of using CZM in represent-
mechanics 共analysis without CZM兲, the crack growth problem is
ing fracture tip as a zone instead of a tip, and provide some mi-
identified as a moving boundary value problem in which the pri-
cromechanistic arguments. In the third section, we will outline the
mary unknown is usually the trajectory of a single point referred
geometry, material and finite element solution to the fracture prob-
to as crack tip. CZM represents a zone or a region where material
lem involving large scale plasticity. In the fourth section, we will
separates. However, in order to interpret the numerical result we
discuss the consequences of the results in terms of plastic zone
need to identify the location of a hypothetical crack tip to facili-
size and energy distribution.
tate the energy and other computations. This hypothetical crack
tip is called here as the cohesive crack tip. There seems to be no
2 Role of MicromechanicsÕProcess Zones in CZM standard way of identifying the tip of the crack from T⫺ ␦ curve.
A typical crack tip process zone is shown in Fig. 2. For our However, one can identify three specific points, one correspond-
purpose we can define process zone as the region within the sepa- ing to the case when the displacement reaches ␦ max and traction
rating surfaces where the surface traction values are nonzero. This attaining the cohesive strength 共point C on T⫺ ␦ curve, Fig. 1兲,
also implies that processes occurring within the process zone are the second point corresponding to the case when the displacement

Journal of Engineering Materials and Technology OCTOBER 2002, Vol. 124 Õ 441

Downloaded From: http://materialstechnology.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jemta8/27039/ on 04/23/2017 Terms of Use: http://www.asme.or

 Table 1 „continued…

reaches ␦ sep and traction has just vanished 共point E on T⫺ ␦ the crack has fully separated and has absorbed all the energy
curve, Fig. 1兲, and the last point when the material is just about to required for the decohesion leaving no active wake behind the tip.
separate 共point A, where ␦ ⫽0 ⫹ 兲. Selection of any one of these But it has been shown in reference 关23兴 that micromechanical
points will have different implications as discussed below. processes are active not only in the forward region of the crack
It is generally accepted that the area under the traction- but also in the wake. To be consistent with these observations,
displacement curve represents the fracture toughness, J IC 关1兴. Se- cohesive crack tip is chosen at a point where displacement reaches
lection of point A implies that the entire cohesive energy will be ␦ max and traction reaches the peak value, namely the cohesive
absorbed by CZM in the wake region. Such a definition will ex- strength, ␴ max . The cohesive crack tip coinciding with peak trac-
clude many inelastic processes, e.g., cavitation damage, occurring tion will be the best position even for extrinsic CZMs of the type
in the immediate vicinity ahead of a crack tip. On the other hand, proposed by Dugdale 关6兴, Camacho and Ortiz 关9兴, Hilderborg
if point 共E兲 corresponding to ␦ sep is chosen as the crack tip, then et al. 共see Jan et al. 关24兴兲. This selection facilitates a part of the

Fig. 1 A typical cohesive traction-displacement curve

442 Õ Vol. 124, OCTOBER 2002 Transactions of the ASME

Downloaded From: http://materialstechnology.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jemta8/27039/ on 04/23/2017 Terms of Use: http://www.asme.or

 zone region兲 acts predominantly in the wake and depends on
the length of the newly formed crack surface. Thus intrinsic
dissipation mechanisms are responsible for the so-called resis-
tance 共R-Curves兲 curves. A rising R-curve shows an increase in
energy dissipation beyond crack initiation until steady-state con-
dition is established, the latter condition coinciding with the full
development of cohesive zone. It is important to note that extrin-
sic dissipation mechanisms are inherent properties of material,
and do not depend on the length of the crack or geometry of the
test specimen 共except for the conditions where the local stress
state is affected e.g., plane stress/plane strain兲. In such cases, the
driving force required to initiate fracture depends only on extrin-
sic dissipation mechanisms. Thus for example, the driving force to
initiate crack growth, K IC , is affected mainly by extrinsic dissi-
pation mechanisms.
In this work, we are focusing on plastic dissipation occurring in
the adjoining material as opposed to all other forms of dissipation
processes occurring within the process zone. We are motivated by
the fact that while other forms of dissipation, 共e.g., grain bridging,
fibril bridging, cavitation, surface energy兲 can be combined to
represent the cohesive energy, plastic work in the adjoining mate-
rial cannot be done so. For this purpose we chose a commonly
used elasto-plastic material as our model material, the values of
Fig. 2 Embedded cohesive process zone K IC 共and hence J IC 兲 being well known for this material.

3 Problem Definition
cohesive energy 共domain ACF in Fig. 1兲 to be dissipated in the
forward region, while the rest of the energy 共domain FCE in Fig. Generic Formulations. Consider the two solid bodies ⍀ 1
1兲 to be dissipated in the wake region. It so happens that with this and ⍀ 2 separated by a common boundary S, an infinitesimally
interpretation, the maximum stress in the bounding material oc- thin 3-D domain with surfaces S 1 and S 2 being the part of ⍀ 1 and
curs adjacent to the crack tip since stress maximum in the cohe- ⍀ 2 , as shown in Fig. 4共a兲. Mathematically, we consider surfaces
sive zone corresponds to a stress maximum in the bulk material. It S 1 ⫽S 2 ⫽S in the initial configuration, and their corresponding
should also be noted that the crack tip location is not an input normals such that N̄ 1 ⫽N̄ 2 ⫽N̄.
parameter and hence will not affect the computational process in In any one of those cases, if S separates 共fractures兲 to Ś 1 and Ś 2
any way. Selecting a particular location as crack tip is necessary with normals n̂ 1 and n̂ 2 as shown in Figs. 4共b兲 and 共d兲, then the
for post-processing and interpreting the results. Xu and Needle- process creates a new internal/external surface violating the fun-
man 关10兴 have used points corresponding to 2 ␦ max and 5 ␦ max as damental laws of continuity. The region bounded by Ś 1 and Ś 2
crack tip locations, and the numerical predictions were unaffected belonging to a new domain ⍀* is assumed to be made of ex-
by these selections. tremely soft glue, which can be shrunk to an infinitesimally thin
Inelastic Processes in the Cohesive Zone. It is obvious that surface but can be expanded to a 3-D domain. The domains ⍀ 1
the micromechanical conditions prevailing in the FPZ within the and ⍀ 2 are governed by an elasto-plastic constitutive model such
cracking region and regions within the bounding material will that
vary widely depending on the type of material, geometric and ␴ ioj ⫽L i jkl 共 D kl ⫺D kl
In
兲. (1)
loading conditions. In this section, we explore such a relationship
based on our knowledge of what we know of the inelastic pro- The elasticity tensor L i jkl is assumed to be isotropic; where D kl
In
cesses that occur in the crack tip region. In the crack tip region we total rate of deformation tensor, D kl is inelastic part of rate of
include FPZ 共completely covering all the regions where the cohe- deformation tensor and ␴ i j is Jaumann rate of Cauchy stress. A
o
sive tractions are non zero兲 that comprises a forward region where typical constitutive relation of ⍀* is given by T⫺ ␦ relations 共see
inelastic damage processes are occurring due to current loading Fig. 4共c兲兲
conditions and a wake region in the bounding material where the
elastic unloading is taking place 共corresponding to the wake re- if 兩¯␦ 兩 ⬍ 兩¯␦ se p 兩 , ˜␴ n̂⫽T̄. (2)
gion of the T⫺ ␦ curve兲. We heavily borrow the concepts from a
recent paper by Ritchie 关24兴 in identifying the various possible Beyond a separation distance of 兩¯␦ 兩 ⭐ 兩¯␦ se p 兩 , the traction being
mechanisms dissipating inelastic energy. There are certain mecha- identically zero within ⍀*,
nisms active in the forward region within the cohesive zone and 兩¯␦ 兩 ⭓ 兩¯␦ se p 兩 , ˜␴ n̂⫽T̄⫽0. (3)
deep into the bounding material which dissipate inelastic energy
and are termed as extrinsic dissipation 共see Fig. 3兲. Ritchie calls It can also be construed that when 兩¯␦ 兩 ⭐ 兩¯␦ se p 兩 in the domain
this intrinsic toughening and we prefer to name it extrinsic dissi- ⍀*, the stiffness L i jkl ⬅0. In order to implement the vectorial
pation, to maintain our perspective from that of the crack rather inequalities given in Eqs. 共2兲 and 共3兲, typically separate identities
than from the bounding material, and also to account for the dis- are postulated for the normal and tangential components with lim-
sipated energy during the loading/unloading process. Crack its set for each of them.
growth is generally promoted by extrinsic dissipation resulting The formulation described above can be implemented in a com-
from microstructural damage mechanisms in the forward region putational scheme like FEM. The advantage of this formulation is
共both in the bounding material and in the cohesive zone兲, while that material separation is achieved without loss of continuity. By
crack advancement is impeded by intrinsic dissipation that takes creating new surfaces, the traction and the stiffness of the cohe-
place in the wake region. However, both consume energy, a part sive zone elements connecting these newly created surfaces are
of it being dissipated into the material and rest into the crack made to vanish, but the displacements across them are still con-
共within the cohesive zone region兲. tinuous. This aspect takes care of crack healing efficiently. In
It should be noted that intrinsic dissipation 共within the cohesive computational schemes like node releasing technique, new sur-

Journal of Engineering Materials and Technology OCTOBER 2002, Vol. 124 Õ 443

Downloaded From: http://materialstechnology.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jemta8/27039/ on 04/23/2017 Terms of Use: http://www.asme.or

 Fig. 3 Energy dissipating micromechanisms in the wake and forward regions

faces are created by the use of ad hoc criteria. In some other

methods cracks are modeled by altering the boundary conditions
and these techniques usually do not conserve the energy.
Selection of CZM and Material Models. Here a phenom-
enological interface constitutive relationship is defined to model
the interface or the fracture process zone. The interface is charac-
terized by the work of separation and strength in normal and tan-
gential directions. The cohesive zone interface relationship can be
expressed such that the tractions T across the interface is taken as
a function of displacement jump ⌬ ¯ across the interface. Defining a
work potential of the type ␾ (⌬ n ,⌬ t ), such that traction acting in
the interface is given by

⳵␾
T⫽⫺ ¯ . (4)
⳵⌬

In this work an exponential CZM proposed by Xu and Needleman

关14兴 is implemented in the numerical scheme.
The Problem of Doubled Edged Notched Plate. For our
purpose of computing the energy flow in a fracture process, a
double edged notched plate, as shown in Fig. 5共a兲, is analyzed.
Considering the symmetry of loading and geometry, only a quarter
portion of the plate, as shown in Fig. 5共b兲, is discretized. The
cohesive zone parameter ␦ n governs the size of cohesive elements
near the crack tip. For a convergent solution, the sizes of cohesive
elements and adjoining elements in the bounding material are Fig. 4 Conceptual frame work of cohesive zone model

444 Õ Vol. 124, OCTOBER 2002 Transactions of the ASME

Downloaded From: http://materialstechnology.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jemta8/27039/ on 04/23/2017 Terms of Use: http://www.asme.or

 The bounding material is assumed to be Al 2024 T3 alloy
with the Young’s modulus of 72 GPa, Poisson’s ratio of 0.33, and
the true stress-strain curve is given by ␧⫽ ␴ /E⫹ ␣ ( ␴ / ␴ y ) 1/n ,
where ␴ y ⫽320 MPa, ␣⫽0.01347 and n⫽0.217173, fracture
toughness K IC ⫽25 MPa m1/2. The cohesive material properties
are derived from the global material property of Al 2024 T3. In
order to relate the cohesive surface behavior to the fracture tough-
ness, the energy for interfacial normal separation is related to
critical plane strain value of J-integral, ␾ n ⫽J Ic ⫽8000 J/m2 关1兴.
By assuming an appropriate value for maximum cohesive
strength, ␴ max the characteristic normal displacement ␦ n is com-
puted using ␾ n ⫽e• ␴ max•␦n .
Finite Element Simulation. A general purpose finite element
code ABAQUS 关26兴 is used for the analysis. Four node elements
based on cohesive zone concepts have been developed and imple-
mented as user defined elements within ABAQUS. Crack growth
simulation was carried out by applying far end incremental dis-
Fig. 5 „a… Double-edge notched plate, „b… geometry and placements. The crack tip grows from an initial location of x
boundaryÕloading conditions used in FEA model ⫽0.025 m until a steady state is reached in a self-consistent man-
ner. During the application of load the crack tip 共as per our earlier
definition兲 moves past about 7000 cohesive zone elements.
A number of simulations were carried out for a range of cohe-
maintained to be of the order of ␦ n , which in the present case is sive zone properties, especially the cohesive strength. Since for all
of the order 5 ␮m. The finite element mesh is as shown in Fig. the simulations the base material was assumed to be Al 2024 T3,
6共a兲. In the simulation, crack propagates for a certain length until the cohesive energy ␾ n is fixed at 8000 J•m⫺2. The only variation
a steady-state condition is obtained. Hence for a small length near possible within the cohesive properties is a change in cohesive
the crack tip, in the direction of crack growth, the size of the strength ␴ max with corresponding changes in normal and tangen-
element is ⭐5 ␮m. The fine mesh around crack tip is shown in tial displacements ␦ n and ␦ t . In order to study the effect of cohe-
Fig. 6共b兲. After making convergence studies, the mesh was dis- sive strength on plasticity, ␴ max was varied such that the ratio of
cretized with 24340 plane strain 4 node quadrilateral elements. ␴ max /␴y was altered from 1, 1.5, 2.0 to 2.5. The spatial distribu-
Though the aspect ratio of the elements near the tip is not optimal, tion of energy and plasticity within the bounding material and the
a preliminary check using optimal ratio showed reasonable agree- cohesive zone were studied to examine the evolution of these
ments. Different mesh designs were also studied to avoid the ar- quantities as a function of loading history and cohesive strength.
tifacts from clouding the observations. The present mesh design All simulations were carried on SGI-Octane with MIPS R12000
was limited by the computing capacity 共SGI Origin with 4 pro- processor running on unix operating system IRIX 6.5, with a typi-
cessors兲. 7300 cohesive elements with each having 4 nodes are cal simulation consuming about 5 to 8 hours of CPU time.
used along the line of crack propagation. A total of 28,189 nodes
are used to model the geometry. Here the FPZ is assumed to be a 4 Discussion of the Results
line, and this line is modeled by 4 node rectangular cohesive
elements having zero thickness in the direction normal to the di- Global Energy Distribution. During fracture process, the en-
rection of crack propagation. One face of cohesive elements is ergy is supplied by the external loads. The bounding material
connected to regular elements, while the other face is given the undergoes elastic/elasto-plastic deformation involving elastic en-
symmetric displacement boundary conditions. Thus an artificial ergy and plastic dissipative energy. If other forms of dissipative
interface is created along the line of crack propagation. processes are modeled within the constitutive equation, then ad-

Fig. 6 „a… Finite element mesh model of quarter portion of double edge notched plate „b… fine
mesh near the crack tip

Journal of Engineering Materials and Technology OCTOBER 2002, Vol. 124 Õ 445

Downloaded From: http://materialstechnology.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jemta8/27039/ on 04/23/2017 Terms of Use: http://www.asme.or

 ditional energy will be spent in such inelastic processes, for ex-
ample damage/void growth. In addition to plasticity, energy is
supplied to FPZ in the form of cohesive energy that is dissipated
within the cohesive elements. This cohesive energy is the sum
total of all dissipative processes that go within the crack tip re-
gime and surface energy. For the given system there will be a
perfect energy balance between external work and the sum of
elastic energy (E e ⫽ 兰 t0 ( 兰 V ␴ :␧˙ el dV)d ␶ ), plastic dissipative en-
ergy (E p ⫽ 兰 t0 ( 兰 V ␴ :␧˙ pl dV)d ␶ ), and cohesive energy (E c
⫽ 兰 t0 ( 兰 intS T̄:¯␦ dS)d ␶ ).

E w ⫽E e ⫹E p ⫹E c . (5)
While E e and E p are confined to the bounding material, E c is
restricted to FPZ.
Let us first consider the case of a pure elastic material. The
conventional fracture mechanics theory uses the concept of strain
energy release rate for crack growth analysis, i.e.,

⳵U
strain energy release rate⫽G⫽J⫽⫺ . (6)
⳵a
This fracture energy is dissipative in nature. Hence in an analysis Fig. 7 Variation of plastic work with cohesive energy for dif-
ferent ␴ max Õ␴y ratio
using CZM, even for an elastic material the entire fracture energy
of ␾ ⫽J IC ⫽G IC ⫽8000 J/m2 is dissipated through cohesive ele-
ments. In this case, only form of dissipation occurs in cohesive
element since,

E w ⫽E e ⫹E c . (7) Variation of Plastic Work and Cohesive Work. Figure 7

shows the relationship between cumulative plastic energy and co-
Let us consider the more general case of elasto-plastic materi- hesive energy at various stages of crack growth for different val-
als; two distinct dissipation mechanisms can be identified, one due ues of ␴ max /␴y . For the value of ␴ max /␴y⫽1.5 which represents
to plasticity within the bounding material, and another due to very small scale plasticity, the plastic energy represents about 15
micro-separation processes in the FPZ. If the fracture energy ␾ percent of the overall energy dissipated. In other words, the error
⫽J IC ⫽G IC ⫽8000 J/m2 共measured from the experiments for an incurred when plastic work is not accounted for in the dissipative
elasto-plastic material兲 is to be dissipated, the obvious question processes is of the order of 15 percent when small scale plasticity
that arises is: should the measured fracture energy go entirely into is observed. As will be discussed later, in this case, plasticity
the cohesive zone or should it be split into the two identifiable occurs in the initial stage of crack growth, i.e., crack initiation
dissipation processes? It should be realized that during testing rather than steady state growth. This implies that the deviation of
when K IC is measured, this value represents the sum total of all 15 percent occurs during crack initiation leading to the fact that
dissipative processes in the actual material for initiating and the rate of plastic energy dissipation will be much more significant
propagating fracture. If the entire fracture energy is used up in the in those stages.
cohesive zone, it will leave no energy for the plastic work in the Figure 7 shows increasing levels of plastic work as a part of
bounding material. On the other hand, if the fracture energy were total dissipation when the values of ␴ max /␴y⭓2. In these cases the
to be split into two portions, then in what ratio should that divi- amount of plastic work is significantly higher and is almost 100
sion be made? There seems to be no clear experimental method to percent to 200 percent as that of the cohesive energy. Obviously,
isolate the two components. Thus this division is nontrivial since great care needs to be exercised in isolating the plastic work from
energy consumed in plastic dissipative processes depends on the the cohesive work in these cases. From the results shown in Fig. 7
geometry, loading, and various stages of crack growth. In this it can be inferred that during crack growth with significant levels
work, we pursue two different aspects of this question. First, we of plasticity, dissipative energy of much more than the one given
would like to evaluate the contribution of plastic work toward by J IC or G IC is required. Part of this energy will be used for
total work, and then proceed to examine the various parameters plastic work and the rest in the fracture process as cohesive work.
that affect the quantity of plastic work. Whether the test result that formed the basis for the estimation of
Our numerical results indicated that during crack growth pro- 8000 J/m2 involved only fracture process or part fracture and part
cess, recoverable elastic work E e constituted most of the external plasticity is not known. If we assume that during the test, no
work, ranging almost 95–98 percent of the total work. Since elas- plasticity 共or significantly low levels兲 occurred as in the case of
tic work is used in the entire body and not just the crack tip small scale yielding then it is reasonable to assume that this 8000
region, this work depends on the geometry of the body. The plas- J/m2 represents only fracture work and not plastic work. Based on
tic work occurs primarily in the crack tip region and hence is that assumption, it is possible to assign all the 8000 J/m2 as co-
influenced by cohesive zone parameters. Though there are many hesive work. However, if the original test involved significant
parameters that can affect plasticity, cohesive strength ␴ max has plastic work as the material is highly ductile then this value will
the major influence, and occurs in the form of ␴ max /␴y ratio. In represent both plastic and fracture work.
general, the plastic work can be represented as,

冉 冊
LocalÕSpatial Energy Distribution. It is interesting to study
␴ max how the cohesive energy is absorbed within the fracture process
E p ⫽E p ,n,S i ,i⫽1,2, . . . (8)
␴y zone along its length. This study will provide a good physics basis
to postulate various types of micromechanisms active in the for-
where S i represents other factors arising from the shape of the ward and wake regions of the crack. There ought to be a relation-
traction-displacement relations, and this is not explored in this ship between the spatial distribution of energy flow in the FPZ,
work. dissipative mechanisms that absorb this energy, and an overall

446 Õ Vol. 124, OCTOBER 2002 Transactions of the ASME

Downloaded From: http://materialstechnology.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jemta8/27039/ on 04/23/2017 Terms of Use: http://www.asme.or

 Fig. 8 A line of patch of elements along the line of crack
propagation

manifestation of traction-separation law that embodies those de-

tails. The purpose of the next few sections is to explore this rela-
tionship in detail. Fig. 10 Variation of elastic energy in various patch of ele-
In order to study the spatial distribution of energy flow in the ments as a function of crack extension. The numbers indicate
vicinity of the crack tip, a set of patch of elements 共approximately patch numbers starting from initial crack tip.
50兲 was selected in the bounding material, adjoining the cohesive
elements as shown in the Fig. 8. The patches are approximately
squares 共with 130 ␮m side, compare this to ␦ n ⫽4.5 ␮ m兲 and are
chosen to average the effect of local variations. They are spaced Since the variations of elastic and plastic energies are nontrivial, a
equally from each other starting from the initial crack tip. As the schematic illustration of the phenomenon is presented in Fig. 12.
crack propagates these patches of elements become a part of for-
ward region first, wake region next, and then completely a free Variation of Cohesive Energy. As seen in Fig. 9 cohesive
surface 共with no traction兲 depending on the position of the crack energy dissipation takes place from the very initial stages of ex-
tip. ternal loading. The cohesive energy in the patch increases up to
point C 共corresponding to ␴ max in Fig. 1兲 when the crack tip is
Variation of Elastic, Plastic, and Cohesive Energies. In order presumed to advance. The energy consumed by the cohesive ele-
to study the spatial distribution of elastic, plastic, and cohesive ments at this stage is approximately 1/7 of the total cohesive en-
energies in the cohesive zone we have chosen the case of ergy for the present CZM. Once the point C is crossed, the patch
␴ max /␴y⫽2, which involves significant levels of plasticity. First, of elements falls into the wake region. The rate of cohesive zone
we like to study the variation of elastic energy in 10 well-spaced energy absorption depends on the slope of the T⫺ ␦ curve 共in fact
patches spanning over 7300 cohesive elements. Corresponding to ␦
the patch in the bounding material, a set of cohesive zone ele- it depends on 兰 0 currentT.d ␦ 兲, and the rates at which elastic unload-
ments adjacent to each patch has been chosen and the variation of ing and plastic dissipation take place in the adjoining material.
cohesive energy in those elements is shown in Fig. 9. Figure 10 The curves flatten out once the entire cohesive energy ␾ is dissi-
shows the variation of elastic energy in those 10 patches, while pated within a given zone. The pattern observed for the first patch
Fig. 11 plots the variation of plastic energy in the same patches. repeats for subsequent patches of elements. It is thus clear that
while the total cohesive energy absorbed in each cohesive zone
element remains the same, the rate at which the element absorbs
will depend on the length of cohesive zone 共both forward and
wake兲.

Fig. 9 The variation of cohesive energy in the wake and for-

ward region as the crack propagates. The numbers indicate the Fig. 11 Variation of dissipated plastic energy in various
cohesive element patch numbers adjoining the patches of patches as a function of crack extension. The numbers indicate
binding elements. patch numbers starting from initial crack tip.

Journal of Engineering Materials and Technology OCTOBER 2002, Vol. 124 Õ 447

Downloaded From: http://materialstechnology.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jemta8/27039/ on 04/23/2017 Terms of Use: http://www.asme.or

 Fig. 12 Schematic of crack initiation and propagation process in a ductile material

Variation of Elastic Energy. In Fig. 10 the stored elastic Fig. 12共b兲. This build up continues until the peak of T⫺ ␦ curve is
energy is shown and the behavior of patch 1 is quite different reached and the process discussed above repeats, as shown in Fig.
compared to all other patches. Just before the crack grows for the 12共c兲. Though the patterns in all the patches are identical, the
first time, considerable elastic energy is built up as shown in Fig. magnitude of elastic energy accumulated in the first patch is
12共a兲. The elastic energy built up continues until the peak of T higher than the rest, see Fig. 10. The initial crack tip is inherently
⫺␦ curve 共point C in Fig. 1兲 is reached. Once past this point, the sharp leading to high levels of stress field; once the crack tip
cohesive elements near the crack tip get separated beyond critical advances the tip tends to be blunt leading to lower levels of
displacement ␦ n resulting in advancement of the cohesive crack stresses, this in turn reduces the energy levels in other patches.
tip. With further loading, the elements in this patch become part of
The variation of elastic strain energy for other cases of ␴ max /␴y is
the wake, as shown in Fig. 12共b兲. This part of the wake will be
still active because of the fact that cohesive energy is still being similar to the one shown in Fig. 10.
dissipated in the cohesive elements adjoining this patch. At this Variation of Plastic Work. For understanding the variation
stage, the values of normal traction reduce following the down- of plastic work, we examine two different cases, one for large
ward slope of T⫺ ␦ curve 共from C toward E in Fig. 1兲. Conse- plastic yielding ( ␴ max /␴y⭓2.0) and the other for small scale
quent to the reduction in traction, the stress in the patch reduces yielding ( ␴ max /␴y⭐1.5).
accompanied by reduction in elastic strain energy. Interestingly,
this reduction in elastic strain energy is used up in dissipating Case (1) ␴ max /␴y⭓2.0. As explained earlier, plastic energy
cohesive energy to those cohesive elements adjoining this patch. accumulates considerably along with elastic energy in the first
Thus when cohesive energy is absorbed by elements in the wake patch, as shown in Fig. 12共a兲. The size of yield zone is an impor-
region, it is at the expense of elastic strain energy reduction in the tant indicator of the level of plasticity; yield zone size at various
adjoining bounding material. Elastic unloading continues to take stages of crack growth is shown in Fig. 13. As seen from this
place, until sufficient amount of strain energy gets transferred to figure, yield zone is very small during crack initiation and the size
the cohesive elements, at which point complete dissipation of co- reaches a steady state beyond patch 4. Plastic deformation occurs
hesive energy has taken place rendering this part of the wake in the bounding material when the local stresses exceed the yield
inactive. Concurrent with elastic unloading in the first patch, elas- ␴ y ; and such a state of loading occurs whenever the normal trac-
tic energy builds up in the second patch of element, as shown in tion in the adjoining cohesive elements is such that T⭓1.4 ␴ y . As

448 Õ Vol. 124, OCTOBER 2002 Transactions of the ASME

Downloaded From: http://materialstechnology.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jemta8/27039/ on 04/23/2017 Terms of Use: http://www.asme.or

 Fig. 13 Contour plot of yield locus zone around the cohesive crack tip at the various stages of crack growth

the traction in the cohesive zone continues to increase beyond forward region and the rest in the wake region. The part of the
yield, incremental plastic work occurs in the bounding material. cohesive energy spent as extrinsic dissipation in the forward re-
This process continues to occur as long as the traction continues gion is used up in advancing the crack tip. The part of energy
to increase to the peak point C on T⫺ ␦ curve. Beyond this point, spent as intrinsic dissipation in the wake region is required to
traction reduces and plastic deformation ceases in the first patch. complete the gradual separation process.
Since plastic work is dissipative in nature, the energy stored in • In the case of elastic material, the entire fracture energy given
those regions will not reduce even beyond the wake and traction by the J IC of the material is dissipated in the fracture process zone
free crack zones. The energy transfer in the wake region occurs as cohesive energy.
from elastic strain energy to the cohesive zone. • In the case of small scale yielding, a small amount of plastic
The accumulated plastic work decreases up to patch 4 from that dissipation 共of the order 15 percent兲 occurs mostly at the crack
of 1, and then increases as shown in Fig. 11. As seen while ex- initiation stage. During the crack growth stage, because of re-
plaining the elastic strain energy, the initial sharpness of the crack duced levels of stresses, plastic dissipation becomes negligible in
reduces with crack growth. Consequently, the stress intensity is the forward region. The error accrued by neglecting plasticity with
higher in patch 1 compared to others. Also, as the yield zone size respect to dissipated energy is of the order 15 percent in this case.
is smaller in the first case, the intensity in the first patch will be On the other hand, if yielding becomes substantial, there is con-
higher compared to cases where the zone is well spread out. Fur- siderable plastic dissipation together with cohesive energy; ne-
ther, it is seen that the amount of mechanical work required to glecting plasticity in such cases may lead to erroneous results.
propagate the crack increases continuously. Since the cohesive • Amount of fracture energy dissipated in the wake region de-
energy absorbed is constant and the elastic energy stored does not
increase 共see Fig. 10兲, this energy has to be stored in the form of
plastic work. That increase in plastic work causes the increase in
the stored work in patches 4 and beyond.
Case (2) ␴ max /␴y⭐1.5. In the case of ␴ max /␴y ratio equal to
1.0, there is no plastic dissipation. For ␴ max /␴y ratio equal to 1.5,
plastic work is induced only in the first patch of element as shown
in Fig. 14 and plastic dissipation does not take place anywhere in
the forward region. As explained earlier, crack tip is sharp in the
initial stages inducing high levels of stress and hence plasticity.
Once crack growth is initiated, all the cohesive elements along the
interface experiences cohesive traction and displacements, which
reduces the sharpness of the advancing crack tip, reducing the
strength of the crack tip singularity considerably. Because of this
plastic dissipation does not take place in the forward region.

5 Summary
Cohesive zone approach provides an alternate method of mod-
eling fracture process in continuous media. This approach is based
on sound physics and micromechanics, and has the added advan-
tage of easy implementation in numerical methods. This paper
addresses some of the key issues in obtaining cohesive zone pa-
rameters 共e.g., cohesive energy兲 in terms of measurable quantities.
Some of the salient observations in this work are: Fig. 14 Variation of plastic dissipation and elastic work in vari-
ous patch of elements along the interface for the case of
• The cohesive zone method allows the energy to flow into the ␴ max Õ␴yÄ1.5. The numbers indicate the energy in various patch
fracture process zone, where a part of the energy is spent in the of elements starting from the crack tip.

Journal of Engineering Materials and Technology OCTOBER 2002, Vol. 124 Õ 449

Downloaded From: http://materialstechnology.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jemta8/27039/ on 04/23/2017 Terms of Use: http://www.asme.or

 pends on the shape of the model. For example, in the exponential Damage in Brittle Materials,’’ Int. J. Solids Struct., 33, pp. 2899–2938.
关10兴 Xu, X. P., and Needleman, A., 1994, ‘‘Numerical Simulation of Fast Crack
cohesive zone model approximately 6/7th of total dissipation
Growth in Brittle Solids,’’ J. Mech. Phys. Solids, 42, pp. 1397–1434.
takes place in the wake region. 关11兴 Foulk, J. W., Allen, D. H., and Helms, K. L. E., 2000, ‘‘Formulation of a
• Plastic work depends on the shape of the crack tip in addition Three-Dimensional Cohesive Zone Model for Application to a Finite Element
to ␴ max /␴y . Before steady-state conditions are established, the Algorithm,’’ Comput. Methods Appl. Mech. Eng., 183, pp. 51– 66.
sharpness of the crack tip reduces from very sharp to blunt; the 关12兴 Espinosa, H. D., Dwivedi, S., and Lu, H. C., 2000, ‘‘Modeling Impact Induced
level of bluntness is dictated by the shape of the T⫺ ␦ curve 共l 1 /l 2 Delamination of Woven Fiber Reinforced Composites with Contact/Cohesive
Laws,’’ Comput. Methods Appl. Mech. Eng., 183共3– 4兲, pp. 259–290.
ratio兲. 关13兴 Needleman, A., 1990, ‘‘An Analysis of Tensile Decohesion Along an Inter-
face,’’ J. Mech. Phys. Solids, 38, pp. 289–324.
Acknowledgment 关14兴 Xu, X. P., and Needleman, A., 1993, ‘‘Void Nucleation by Inclusion Debond-
ing in a Crystal Matrix,’’ Modell. Simul. Mater. Sci. Eng., 1, pp. 111–132.
The authors wish to acknowledge Army Research Office and 关15兴 Yang, B., Mall, S., and Ravi-Chandar, K., 2001, ‘‘A Cohesive Zone Model for
Air Force Office of Scientific Research for providing partial finan- Fatigue Crack Growth in Quasibrittle Materials,’’ Int. J. Solids Struct., 38, pp.
cial support during the research. 3927–3944.
关16兴 Needleman, A., 1987, ‘‘A Continuum Model for Void Nucleation by Inclusion
Debonding,’’ ASME J. Appl. Mech., 54, pp. 525–531.
References 关17兴 Rice, J. R., and Jian-sheng, Wang, 1989, ‘‘Embrittlement of Interfaces by
关1兴 Rice, J. R., 1968, ‘‘A Path Independent Integral and Approximate Analysis of Solute Segregation,’’ Mater. Sci. Eng., 107, pp. 23– 40.
Strain Concentration by Notches and Cracks,’’ ASME J. Appl. Mech., 35, pp. 关18兴 Needleman, A., 1990, ‘‘An Analysis of Decohesion Along an Imperfect Inter-
379–386. face,’’ Int. J. Fract., 42, pp. 21– 40.
关2兴 Atkinson, C., 1979, ‘‘Stress Singularities and Fracture Mechanics,’’ Appl. 关19兴 Tvergaard, V., and Hutchinson, J. W., 1992, ‘‘The Relation Between Crack
Mech. Rev., 32共2兲, pp. 123–135. Growth Resistance and Fracture Process Parameters in Elastic-Plastic Solids,’’
关3兴 Wappling, D., Gunnars, J., and Stahle, 1998, ‘‘Crack Growth Across a Strength J. Mech. Phys. Solids, 40, pp. 1377–1397.
Mismatched Bimaterial Interface,’’ Int. J. Fract., 89, pp. 223–243. 关20兴 Tvergaard, V., 1990, ‘‘Effect of Fibre Debonding in a Whisker-Reinforced
关4兴 Barenblatt, G. I., 1959, ‘‘The Formation of Equilibrium Cracks During Brittle Metal,’’ Mater. Sci. Eng., A125, pp. 203–213.
Fracture. General Ideas and Hypothesis. Axisymmetrical Cracks,’’ PMM, 23,
关21兴 Geubelle, P. H., and Baylor, J., 1998, ‘‘The Impact-Induced Delamination of
pp. 434 – 444.
Laminated Composites: A 2D Simulation,’’ Composites, Part B, 29B, pp. 589–
关5兴 Barenblatt, G. I., 1962, ‘‘Mathematical Theory of Equilibrium Cracks,’’ Ad-
vances in Applied Mechanics, Academic Press, New York, pp. 55–125. 602.
关6兴 Dugdale, D. S., 1960, ‘‘Yielding of Steel Sheets Containing Slits,’’ J. Mech. 关22兴 Mohammed, I., and Liechti, K. M., 2000, ‘‘Cohesive Zone Modeling of Crack
Phys. Solids, 8, pp. 100–104. Nucleation at Bimaterial Corners,’’ J. Mech. Phys. Solids, 48, pp. 735–764.
关7兴 Rahulkumar, P., Jagota, A., Bennison, S. J., and Saigal, S., 2000, ‘‘Cohesive 关23兴 Hutchinson, J. W., and Evans, A. G., 2000, ‘‘Mechanics of Materials: Top-
Element Modeling of Viscoelastic Fracture: Application to Peel Testing of down Approaches to Fracture,’’ Acta Mater., 48, pp. 125–135.
Polymers,’’ Int. J. Solids Struct., 37, pp. 1873–1897. 关24兴 Ritchie, R. O., 1999, ‘‘Mechanisms of Fatigue-Crack Propagation in Ductile
关8兴 Siegmund, T., and Brocks, W., 2000, ‘‘A Numerical Study on the Correlation and Brittle Solids,’’ Int. J. Fract., 100, pp. 55– 83.
Between the Work of Separation and the Dissipation Rate in Ductile Fracture,’’ 关25兴 Jan G. M., and van Mier, 1996, Fracture Processes of Concrete, Assessment of
Eng. Fract. Mech., 67, pp. 139–154. Material Parameters for Fracture Models, CRC press, Boca Raton, FL, p, 291.
关9兴 Camacho, G. T., and Ortiz, M., 1996, ‘‘Computational Modeling of Impact 关26兴 ABAQUS, 1998, Manual, version 5.8, Hibbit, Karlsson & Sorensen, Inc.

450 Õ Vol. 124, OCTOBER 2002 Transactions of the ASME

Downloaded From: http://materialstechnology.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/journals/jemta8/27039/ on 04/23/2017 Terms of Use: http://www.asme.or