International Journal of Fracture 81: 357-372, 1996.

357
(~) 1996 Kluwer Academic Publishers. Printed in the Netherlands.

A review of the J and I integrals and their implications for crack

growth resistance and toughness in ductile fracture

B. COTTERELL 1 and A.G. ATKINS 2

ISchool of Mechanical and Production Engineering, Nanyang Technological University, Singapore, 639798
2Department of Engineering, The University of Reading, Reading RG6 2AY, UK

Received 9 March 1996; accepted in revised form 18 July 1996

Abstract. The application of the J and the/-integrals to ductile fracture are discussed. It is shown that, because
of the finite size of the fracture process zone (FPZ), the initiation value of the J-integral is specimen dependent
even if the plastic constraint conditions are constant. The paradox that the I-integral for steady state elasto-plastic
crack growth is apparently zero is examined. It is shown that, if the FPZ at the crack tip is modelled, the/-integral
is equal to the work performed in its fracture. Thus it is essential to model the fracture process zone in ductile
fracture. The/-integral is then used to demonstrate that the breakdown in applicability of the J-integral to crack
growth in ductile fracture is as much due to the inclusion in the J-integral of progressively more work performed
in the plastic zone as it is to non-proportional deformation during unloading behind the crack tip. Thus JR-curves
combine the essential work of fracture performed in the FPZ with the plastic work performed outside of the FPZ.
These two work terms scale differently and produce size and geometry dependence. It is suggested that the future
direction of modelling in ductile fracture should be to include the FPZ. Strides have already been made in this
direction.

1. Introduction

The J-integral is a path independent integral for a non-linear elastic material given by [1]

(1)
Ox

where W is the strain energy density function, T is the traction vector on F, u is the displace-
ment vector, and F is a contour starting on the straight lower face of the crack and ending on the
upper straight face with the positive direction in an anticlockwise direction (see Figure l(a)).
During loading of an elasto-plastic material the plastic deformation is nearly proportional
before the initiation of fracture and almost identical to that of a non-linear elastic material of
the same stress-strain characteristics. Hence the initiation value, JIe, can be used as a criterion
of fracture initiation provided its dependence on the constraint of the specimen is taken into
account [2]. The fracture process cannot take place without strain-softening which because
of its unstable nature causes localization into a thin narrow fracture process zone (FPZ). At
initiation Jle can be shown to be equal to the specific essential work, Ro, necessary to cause
fracture in the FPZ [1]. Thus Ro can be obtained from the J-integral, which is very nearly path
independent and, provided the FPZ is small, Ro is also obtainable from the potential energy
release rate and is given by [1]

dH
J- dl' (2)
358 B. Cotterell and A.G. Atkins

where H is the potential energy of the system per unit thickness and l is the crack length.
In practice J is usually obtained from this energetic interpretation and not from the contour
integral.
Although up to initiation of a fracture the deformation is well modelled by a non-linear
theory, during crack growth unloading occurs behind the crack tip which is non-proportional
and cannot be modelled by non-linear elasticity. Hutchinson and Paris [3] have shown that
outside of a region of non-proportional loading defined, in terms of a radial coordinate r, by

J
r >> - - -- D, (3)
dJ/dZ

the deformation is nearly proportional. Hence if the region of non-proportional loading is well
contained within the region dominated by the J singularity there will exist an annular region
where the Hutchinson-Rice-Rosengren (HRR) equations hold [4, 5]. Provided a specimen's
uncracked ligament, b, is large compared with D, and the crack extension is dominated by the
HRR-stress field, it is argued that the crack growth will be controlled by the J-integral [6]. If
the remaining ligament, b, is fully yielded the conditions that must be fulfilled for the crack
growth resistance JR to be a function of the crack growth dl, become

b d JR
Al < 0.07b, w . . . . > 14, (4)
JR dl
for the highly constrained deep notch bend specimen and

b d JR
Al < 0.01b, w = > 100, (5)
JR dl
for the less constrained centre cracked tension specimen [6-8]. Thomason [9] has argued
that non-proportional deformation and elastic unloading effects of appreciable magnitude are
propagated completely across the stress field of the dominant singularities along certain rays
and that the concept of J-controlled crack growth is unsound. However, the J-integral does
appear to be reasonably path independent for small crack extensions and J-integral evaluations
on contours remote from the crack tip agree reasonably well with J obtained from its energetic
interpretation [7, 10, 11 ]. Despite its relative path independence during crack growth, we shall
argue that J obtained by the energy method during propagation contains increasing portions
of the non-essential work performed in the plastic zone surrounding the FPZ. In plane strain
fracture the major increase in toughness does not come from the essential work of fracture
but from the non-essential work in the plastic zone surrounding the FPZ. Different geometries
impose different constraints on plastic deformation and, in uncontained yielding, the size of
the specimen determines the amount of non-essential plastic work performed outside of the
FPZ. Hence the size and geometrical dependence of JR-Curves [11-14] is not surprising. As
Broberg [15, 16] suggested many years ago, the plastic zone screens the crack tip from the
energy flowing to it. Clearly only in well contained yielding when steady state propagation
can be established, can the energy dissipation in the plastic zone be independent of geometry
or size. In this case the crack growth resistance curve, whether KR or JR, is unique and
can be used to predict instability in any specimen or structure large enough for the yielding
to be well contained. However, the KR or JR curves do not represent a true increase in
toughness. The application of the concept of KR-curves to ceramics clearly recognizes that
 A review of the J and I integrals 359

the 'toughening' that occurs with crack growth comes from the shielding of the crack tip by
a crack extension induced negative stress intensity factor [17, 18]. The same is most often
the case with ductile fracture, but has not usually been explained in this way. In fully-yielded
specimens the energy dissipated is dependent on both size and geometry for significant crack
extensions. The JR-curve obtained from a small laboratory specimen generally rises more
steeply than that obtained from a large specimen [11, 14] because a greater proportion of
the plastic work outside of the FPZ is contained in JR. Consequently laboratory tests are
likely to overestimate the toughness of large structures. The most notable case of crack growth
resistance curves being due to a real increase in toughness is the first application of the concept
of the crack growth resistance curve to the fracture of high strength thin sheet metals [ 19].
Shear lips form during crack propagation in thin sheets which consume very much more
energy than a plane strain fracture and lead to a real increase in toughness. The work involved
in forming the shear lips is an intimate part of the essential work of fracture and the Kn-curves
for thin sheet metals represent a real increase in toughness. However, in general Jn-curves
for essentially plane strain fracture do not represent a real increase in toughness and are not
unique. Hence the general usefulness of Jn-curves must be questioned. A similar situation
exists in the fracture of cementitious materials. These materials are essentially elastic outside
their FPZs so that there is no energy dissipated apart from that in the FPZ. However, the FPZs
are so large that in laboratory sized specimens the FPZ cannot develop to the size they would
in extremely large specimens. Hence crack growth resistance curves are size and geometry
dependent. Because of this specimen dependency the concept of the crack growth resistance
curve has been largely abandoned by the cementitious community and replaced by a model
that includes a FPZ [20].
The limitations of the J-integral have lead to the examination of a variety of other line
integrals [21]. The/-integral [21-24], which applies to steady state crack propagation in an
elasto-plastic material, is the most similar of these to the J-integral. The/-integral has the
same form as the J-integral and is given by Equation (1) but the strain energy density function,
W, is now the total stress work density and depends upon the history of the deformation. The
/-integral is path independent not just for non-linear elasticity but for a true elasto-plastic
material. Like the J-integral, the/-integral is zero if taken around a closed path that does not
enclose a singularity in the integrand in Equation (1) for a HRR non-linear elastic stress field
is of the order l / r and consequently the J-integral taken around a contour close to the crack
tip is non-zero. However, the singularity in the stress work function at the tip of a steadily
propagating crack in an elasto-plastic material is weaker than 1/r and therefore the/-integral
taken around the crack tip must be zero [22, 25]. Cherepanov [24] gave a physical argument
that I = R0, but then wrongly concluded that W must have a 1/r singularity. However, it is
obvious that work must be expended in the FPZ as well as in the surrounding plastic zone
during ductile fracture and the apparent conclusion that the/-integral taken around a steady
propagating crack tip is zero, leaving no energy available to fracture the FPZ, poses a paradox.
Another possible problem arises for both the J and I integrals if the material has a low
strain hardening rate. Under large deformations and low strain hardening rates near the crack
tip, Thomason [26] has shown that there may be a loss of uniqueness in the near crack tip
deformation. If there is a loss of uniqueness then the use of the J and I integrals may be
invalidated.
360 B. Cotterell and A.G. Atkins

u T ' )//,,T

(a) (b)
Figure 1. (a) The J and I integrals taken arounda crack whose FPZ is modelledas a singularityat the crack tip.
(b) The J and I integrals taken arounda crack whose FPZ is modelledby a Barenblatt/Dugdaletype zone.

2. The J-integral
The J-integral, Equation (1), can be easily seen to be non zero by closing the contour, in
Figure l(a), with an integral along Ft which follows the crack faces until it is in the vicinity
of the crack tip when it encircles the tip with a small circular path of radius r. The d-intergral
around the complete contour 1` + I-'1 is zero because there is no singularity within the contour.
The integral along the crack faces is zero, but the integrand in Equation (1) is of the order
1/r and hence the integral along the contour 1-'l tends to a finite value as r tends to zero and
therefore J has a finite value.
The J-integral ceases to be path independent if it is taken around a contour within the FPZ,
but is reasonably path-independent outside of the FPZ. Localization makes the FPZ narrow,
its width being determined largely by the microstructure of the material. In a structureless
homogeneous continuum the FPZ would collapse to become infinitesimally thin. At the
initiation of a ductile fracture, the initiation value of the J-integral, JIc, can be shown to be
the specific essential work of fracture, R0, by including an infinitesimally narrow FPZ in the
analysis (see Figure l(b)) and closing the contour 1' with a contour I'2 that goes around the
edge of the FPZ [1]. Thus

J = R0 =
fo" a d~, (6)

where cr is the stress in the FPZ, ~ is the crack opening displacement (COD) across the FPZ,
and ~f is the crack tip opening displacement (CTOD) at which a continuous crack forms in
the FPZ (see Figure 2). The initiation value, JIc, does not include any work performed outside
of the FPZ.
If the non-elastic deformation is confined to a small FPZ, it is argued that elasto-plastic
deformation is practically the same as non-linear elastic deformation because the plastic
deformation during loading is nearly proportional. However, Equation (2) is only strictly true
if the FPZ is infinitesimally small. If the fracture process zone is finite we shall show that
Equation (2) only holds if the crack propagation is steady state even for a true non-linear
elastic specimen.
The stresses acting along the surface of a narrow Barenblatt/Dugdale type FPZ, where the
stress is a function of the COD as shown in Figure 2, can be treated as 'external forces' and all
the work performed can be identified. If the stresses at the tip of the FPZ are finite no energy
 A review of the J and I integrals 361

Displaeen~nt
1 ~)f
Figure 2. The stress-displacement relationship across a FPZ.

is released from the model system. Consider the virtual work during a small crack extension,
dl, of a crack under a two-dimensional stress field in a non-linear elastic material. Since the
system considered is conservative, the total virtual work performed on the system must be
zero. Taking a plate of unit thickness, the sum of the external virtual work, dwe, performed
around a contour F (see Figure 1(b)) and the internal virtual work, dwi, performed within the
contour P is
du
[dw~ + dwi]r = [fr T-d-[ - H ~ - dx dy ]dl. (7)

The displacement, u, can be considered as a function of X, a moving coordinate measured

from the crack tip, and hence the total derivative of the displacement can be written as
du Ou Ou Ou Ou
dl -- O---X+ Ol -- Ox +-0-[' (8)
where the partial derivative with respect to 1 is taken for a constant value of the moving
coordinate, X, as it is in the rest of the paper. Hence

[dw.+dwi]r = I f f ~ x dxd9- kT-~ds] dl

-[Sl -ff[-
OW dX dy - Jr
f TOUds]
O1 dl

dX dy f
Jr dl. (9)
In the classic treatment of non-linear fracture 1, the virtual work [dwe + dwi]r]r is equal to
the change in potential energy, F, of the system (excluding the potential energy of forces at
the edge of the FPZ). The virtual external work, [dwe]FeZ performed along the boundary of
the FPZ, assuming a mode I FPZ, is
fP d~
[dwe]FPZ = - 1o a-~--~dX dl = - R d l
(lO)
= {/o'[o'
a O-X
O~
})
~-~ dX dl = -/7.0 dl - So P O~
a-~- dX d/,

I As argued by Rice [ll, there is no need to make an energy balance for the whole body because the virtual
work is of second order in dl at all distances large compared to di, and thus any contour, F, enclosing a finite area
can be used.
362 B. Cotterell and A.G. Atkins
where R is the current essential work per unit fracture area and p is the length of the FPZ.
Hence the total virtual work is given by

dw . { J . Ro . I f l O--0T
.W dX. dY frT 0Uds] fo Pa~l d X } d / = 0 . (11)

Since the J-integral is independent of the path J = Ro and, therefore,

HOW Ou P 05
- T~-ds
=/oo a--ff[dX. (12)

If the FPZ is infinitesimally small then the right hand side of Equation (12) is zero and
Equation (2) is valid. However, if the FPZ is finite then Equation (2) is only valid if the
propagation is steady state and the deformation dependent only on the moving coordinates X
and y. In general

J =R0
dII (13)
R- - - Ro + fo p a-ff[
06 dX
dl
where 17 has its usual meaning as the potential energy of the body as a system assuming that
the FPZ is infinitesimally small.
Whether the essential work of fracture, R, is greater than or less than its specific value R0
depends on whether the FPZ increases or decreases in size with crack extension. The shape of
the opening of a Barenblatt/Dugdale FPZ will not change significantly with crack propagation
unless there is a large change in the constraint. Hence if it is assumed that the COD is given
by

~f = f , (14)

the essential work of fracture is, from Equation (13), given by

dp f I df
R = Ro - 6I-d-[ J o (Xlp) d(X/p) . (15)
do
Since df/d(X/p) is negative, R < Ro if the relative size of the FPZ decreases with crack
growth, but R > Ro if the FPZ increases with crack growth. In most geometries where the
remaining ligament is limited in size, such as the compact tension specimen, the FPZ in fully
yielded specimens shortens with crack extension, because the strain gradients necessarily
become larger as the ligament decreases, and R < Ro [27]

2.1. THE ESSENTIALWORKAND THE J-INTEGRAL

The Dugdale model [28, 29], where non-elastic deformation is confined to an infinitesimally
narrow zone (see Figure 3), has an exact analytic solution and is very useful for the examination
of the relationship between R and the J-integral. Although the Dugdale model is usually
described as representing the fracture of an elasto-plastic material, the narrow yielded zone
is better interpreted as a FPZ. The material can be thought of as an ideal Barenblatt/Dugdale
 A review of the J and I integrals 363
(7

FPZ FPZ ~E_

,ol- . -I a

L 2a d

Figure 3. The Dugdale model.

material which is elastic up to a critical stress, a0, whereupon an infinitesimally narrow FPZ
starts to form and stretch at the critical stress, a0, until a continuous crack forms at a CTOD,
5f. Surprisingly, though the Dugdale model has been employed extensively to the application
of the CTOD concept, it has been little used to investigate the work performed in a FPZ.
Goodier and Field [29] did calculate the work performed in the yielded zone, but only for
crack growth under constant remote applied stress. In that case the CTOD increases with crack
growth which is not a 'natural' condition. Here a more natural growth of a crack at a constant
CTOD, 5I, is examined 2. For equilibrium in the Dugdale model [28] (see Figure 3)

l
- = cos/3, (16)
a

where I is the half crack length, a is half the extent of the crack plus the FPZ, and
'
= ~ 7r, (17)

where aa is the remote applied stress. The essential work performed within the FPZ is given
by [29]

R = f l a a ~d5
- dz

[ a/°l
= ~o 5 / + ~
l
5 dx
J
(18)

= Ro + a o ~ 5 dx.

In the Dugdale model

8or0/
5f = 7rE---:In (sec fl), (19)

2 There is no a priori reason why the CTOD is absolutelyconstant during crack growth. If the constraint at the
crack tip changes during crack growth then ~il may vary, but usually the rate of change in 6/will not be large. That
is not to say, however,that the CTOD for geometries with different degrees of constraint may be quite different.
364 B. Cotterell and A.G. Atkins

0.8-
~k N~rnaltzed
0.7- ir~tion sbess

~ 0.6
', \

..........."-....-%..
O. 2
o ,
o.e
- -
...............
...................
......."." . . . . O. 7 ........

I
0.3
0.2
0.1
0
o 8 8 lo
Ncxmalizedcrack extension

Figure 4. Variationin applied stress (normalizedby the critical stress ao) necessaryfor equilibriumcrack growth
in the Dugdale model.

where E* = E for plane stress and E* = E / ( 1 - / ) 2 ) for plane strain. Thus for equilibrium
crack growth at constant CTOD, the remote applied stress, aa, must decrease with crack
extension. The necessary decrease in applied stress is shown in Figure 4 where the crack
extension has been normalized by the size, p ~ , of the equilibrium FPZ for a limiting semi-
infinite crack under a remote elastic K-field, and is given by
lr ( Ko ~ 2 ~rE* Ro
poo = g \ -a0
- / and (20)

For equilibrium crack growth the essential work of crack propagation, normalized by R0, can
be obtained from Equation (18) and is given by
R fl sin/3 fl 1
R---o = cos fl In (sec fl) sin 2fl 2" (21)
The normalized essential work is plotted against the normalized crack extension in Figure 5.
Only when the FPZ is very small compared with the crack length, which implies that the
initiation stress is very small compared with the critical stress, a0, does JIc = Ro ~ R. As
the crack extends the essential work of fracture tends to its plateau value, Ro, because the size
of the FPZ decreases 3 relative to the crack size until it becomes equal to the size for a semi
infinite crack, as is shown in Figure 6. Thus as the crack becomes long, s9 the propagation
tends to a steady state condition when R = R0. The initiation essential work, which is equal to
Jic, is less than the specific essential work, Ro, because the relative size of the FPZ decreases
(see Equation (15)).
The nominal linear elastic energy release rate,

G ~ E* ~ - -E- z - ' (22)

tends to the essential work in the FPZ when its size becomes small compared with the crack
length. Thus the nominal linear elastic energy release rate, G, tends to Ro as the crack becomes
long, as shown in Figure 7.
3 For applied stresses greater than 0.76ao it was found that the shortening of the FPZ became so pronounced
that healing would be required at its tip. Clearly a FPZ cannot heal and hence crack growth must be unstable for
applied stresses greater than 0.76ao for an ideal Barenblatt/Dugdalematerial. However,that does not imply that
stable crack growthis impossible in a real materialif the applied stressesis greaterthan 0.76o'0
 A review of the J and I integrals 365
1 ,

o.ss"
~ 0.96
i
0.94
"~ 0.92

0.9
0.88
ii
l
/ 0.4
0.6
0.86 0.7
0.76
i 0.84
O.82
0.8
0.78
O. 76
o 2 A 6 8 lO
Normaized crack extension

Figure 5. The essential work of fracture in the Dugdale model as a function of crack extension.

1.8
Normal~d
1.7. initiation strees
0.2
o
1.6- ~ 0.4 ..............
1.6- \ 0.6 ................
N
Q. 0.7 .......
n
1.4- ",\' , \ O. 76

1.3- '~
........ ",'-\,

1.1 ...........................~-S:S::~:.::..:=.:.:.:.::._..

0 2 4 6 8 10
Normalizedcrack extension

Figure 6. The size of the FPZ (normalized by po~) in a Dugdale model as a function of crack extension.

3. The I-integral

The equation for the I-integral is identical to Equation (1) except that for the/-integral the
stress work density, W, depends on the history and it is only path independent for steady
state crack propagation. The two terms in Equation (1) can be interpreted physically for the

2~
\
'~,,~. =/ based on a •
~14

!
...................................... m--...~, ........ :. . . . . . . . .

0.82~ ilitlation8tte~=
o.62 baaed on I
02
0'4
0.6 ....................

0.7 .......
0.76 ..........
o
o i 6 g lO
Nor~ crackextension

Figure 7. The nominal energy release rate for the Dugdale model.
366 B. Cotterell and A.G. Atkins
/-integral [22]. The first term is the rate, per unit crack extension, at which work passes
through the contour and the second is the rate of work done by the surface tractions on the
material within the boundary. The paradox that the I-integral is apparently zero for steady
state crack propagation in an elasto-plastic material is caused by the assumption that the FPZ
is infinitesimally small and has no effect on the stress distribution at the crack tip. If the FPZ
is infinitesimally small, a singularity exists in the integrand of Equation (1) that is weaker
than 1/r. The/-integral taken around the closed contour F + Fl (see Figure la) must be zero
because it does not enclose a singularity. The integral along Ft tends to zero as r tends to
zero, because the singularity for a steady propagating crack is weaker than 1/r and the crack
surfaces are stress free. Hence the/-integral taken along the contour F must be zero. Up to
fracture initiation the deformation of an elasto-plastic material is close to proportional and
the singularity in the integrand of Equation (1) is of the order 1/r hence the J-integral at
initiation has a finite value, whether the FPZ is infinitesimally small or finite, that is Ro, the
essential work of fracture. If there is no change in constraint the essential work of fracture
during steady state propagation must be the same as that at initiation. Therefore, if I = 0
the essential work of fracture must be zero and a fracture must initiate at the smallest load.
Clearly such is not the case, since work must be expended in the FPZ, and therefore in ductile
fracture of an elasto-plastic material the FPZ cannot be modelled by a singularity, but must
be modelled as a finite zone. If this finite zone is modelled as an infinitesimally narrow finite
Barenblatt/Dugdale zone, the/-integral along F2 which closes the contour 1-"(see Figure 1(b))
is given by

I2 = fr2 [W dy - T~xx dS] = -Ro. (23)

Hence the I-integral taken along any contour F that begins and ends on the crack surfaces is
give by

I = /r [W dy - T~xx dS] = Ro. (24)

Therefore in ductile fracture of an elasto-plastic material, JIc = I = Ro.

Although in practice, propagation in a large specimen with a small contained plastic zone
will reach quasi-steady state crack growth, the only geometry strictly able to attain steady
state conditions is the semi-infinite crack with a far elastic K-field. Consider steady crack
growth of such a semi-infinite crack. In an elasto-plastic material the plastic stress work (and
some locked-in elastic stress work) is not recovered and is left in the wake of the crack. If the
/-integral is taken around a circular path of radius r centred at the crack tip, the width of the
wake zone relative to r becomes infinitesimal if r is taken to infinity. Therefore the limiting
/-integral, as the radius of the circular contour tends to infinity, can be separated into an elastic
component, I~, and a residual wake term, Lo, as Hutchinson has noted [22]. Thus

i = I~ + Iw, (25)

where

g 2
~=
E* :
 A review of the d and I integrals 367

and

= 2 Wdy
f0 °
(26)
=_ _ [df~p dAr]
[-~+ d/J"
f~p is the plastic work per unit plate thickness performed in the plastic zone outside of the FPZ
and Ar is the residual elastic strain energy per unit plate thickness locked-in by non-uniform
plastic deformation which is not recovered in the wake of the crack. Therefore under steady
state crack propagation

I- E* [ dl + dl J"
If the FPZ is modelled as a singularity I = 0. However, work is expanded in the FPZ and thus
the FPZ must be modelled as a finite zone. For steady state crack propagation, with a small
plastic zone containing a finite FPZ 4, I = R0 and

dlI K2
dl E*
df~p dAr (28)
= d-T
¢I.
For an ideal elastic-brittle material, by which we mean a material that is perfectly elastic
outside of the FPZ and an example of which is the Barenblatt/Dugdale material described
in Section 2.1, a very small FPZ may be modelled as a singularity. This statement forms,
of course, the basis of linear elastic fracture mechanics. In this case the singularity in the
integrand in Equation (1) is l / r and

K 2
J = I . . . . Ro. (29)
E* E*
Thus dR = Ro is constant. If the FPZ is small and it is modelled as a finite zone then an
initiation d can be obtained from Equation (1) and die is given by

KIt
- E* - (30)

4 Note that at initiation Equation (2) is invalid for a finite FPZ hence

K 2 dlI = G.
J = ~-2 # - d--~

Only when the crack growth becomes steady state does

dfl K2
---- ~e ~ -"
dl E*

This equation (due to Irwin [30]) is identically true for an infinitesimal plastic zone for non-steady state crack
propagation, but is only true for steady state crack propagation if the plastic zone is finite.
368 B. Cotterell and A. G. Atkins

However, because

dn (31)
J=E* # at'
until the crack growth becomes steady state, JR determined from Equation (2), is not constant
and may increase or decrease slightly until it is equal to Ro when the crack growth becomes
steady state. Thus, in the absence of any real toughening action such as the formation of shear
lips which increase the essential work of fracture, an elastic-brittle material cannot possess a
significant JR crack growth resistance curve. Only elasto-plastic materials can have significant
plane strain JR-curves.
The general inconsistency of JR-CUrves, that will further be discussed in the next section,
can be seen from examination of Equations (27) and (28). At initiation JIc can be equally
described by the line integral of Equation (1) or the energetic definition given by Equation (2)
provided the FPZ is small and is approximately equal to R0. However, if the crack propagation
reaches a quasi-steady state, the J-integral given by Equation (1) is approximately5 equal to the
/-integral and equal to R0, but Equation (28) shows that the more usual energetic definition
of J, Equation (2), does not approximate to the J-integral as given by Equation (1). The
derivation of the energetic definition of d relies on the system being conservative apart from
the singularity and this is not possible for an elasto-plastic material.

3.1. CRACK GROWTH RESISTANCE AND STEADY STATE GROWTH

A problem with crack growth resistance as measured either by KR or JR is that the fraction of
the work dissipated in the plastic zone increases with crack growth. This problem is most severe
in uncontained yielding where it is now well established that JR-curves are geometry and size
dependent [11-14]. However, the problem can be appreciated theoretically by considering
crack growth of a semi-infinite crack under contained yielding. In this case there is a far
elastic K-field and, assuming the FPZ to be very small, the J-integral at initiation is given by,
Jic = R0 ~ R/c, the initiation value of the essential work of fracture. Since the far stress field
is the elastic K-field, at initiation G2c/E * = Jlc = Ro, the essential work performed within
the FPZ, and Gtc does not include any work dissipated in the plastic zone. However, when
steady state crack growth is established the plateau value of the crack growth resistance curve,
Goo = K ~ / E 2, which is equal to J ~ obtained from the energetic definition (but not from the
J-integral Equation (2)), is given by Equation (28) and includes the plastic and irrecoverable
elastic energy as well as the essential work of fracture. Thus, unless the essential work of
fracture increases because of change in constraint or a mechanism such as the growth of shear
lips at the surface of thin sheets, crack growth resistance curves, either expressed in terms
of -fiR, GR, or JR do not represent a true increase in fracture resistance, but simply show an
apparent increase in toughness because more and more of the work performed in the plastic
zone is included, until all is included if the propagation becomes steady state. That is not to
say that such curves for well contained yielding are not unique and of use in predicting stable
crack growth and instability. However, the curves represent an increase in the shielding action
of the plastic zone which filters the energy available at the crack tip; they do not represent any
real increase in crack tip toughness.

5 Exactly if the history dependence of W is taken into account.

 A review of the J and I integrals 369
Wp>0
m

Prattle zone
at ~ steady state
Pleatic zone
1
Figure 8. Schematic illustration of the initiation and propagation to steady state crack propagation in ductile
fracture.

Thus in the finite element analysis of Tvergaard and Hutchinson [31] for mode I crack
growth under small scale yielding defined by a far field stress intensity factor K , the initiation
value of the stress intensity factor Klc is given by

EJtc 1/2
-~IC ----
(1 -
(32)
ERo 1/z
(] -Z2)) ,

by virtue of the line integral definition of J. When steady state crack propagation is established,
the energy release rate is given by Equation (28) and is equal to the total energy dissipated. 6
Thus following our above argument, the KR-curves given in Figure 3 of [31] do not represent
a real increase in toughness but only the inclusion of the work in the plastic zone. The initiation
value of the stress intensity factor, Kic, is given by Equation (32), whereas the plateau value,
K ~ , comes from Equation (28).
The schematic development of a well-contained plastic zone from initiation to steady state
propagation is shown in Figure 8. The size of the plastic zone is proportional to K 2 so that
the size of the plastic zone grows as the stress intensity factor grows from its initiation value,
Kic, to the plateau value, Kin. At initiation the plastic work density, Wp, defined by

Wp ~J crij dcij,
-~ f0 e~ P (33)

where QP is the plastic strain, is zero on the plastic boundary. Thus no irrecoverable plastic
work is dumped in the wake of the crack as it begins to propagate and the only extra dissipated
work comes from the increase in the plastic work density, Wp. In contrast when steady state
crack propagation is reached there is no change in the plastic work density, Wp, but now on
the trailing edge of the plastic zone, the plastic work has accumulated and is dumped into
the wake of the growing crack along with locked-in residual strain energy. Thus the main
difference between initiation and propagation is that at initiation the material behaves like
6 Although Tvergaard and Hutchinson [31] checked that there was agreementbetween the prescribed K value
and d values found on outer contours, they do not appear to have made any J-integrals during crack growth and
simply quote the K value necessary to create the required crack extension.
370 B. Cotterell and A.G. Atkins

a non-linear elastic material that does not consume energy in its 'plastic' zone whereas in
propagation non-recoverable energy is dumped in the wake of the crack.
In the results of Tvergaard and Hutchinson [31] the increase in stress intensity factor
necessary to propagate the crack decreases as the ratio of the maximum stress in the fracture
process zone to the yield strength decreases, because in the limit when maximum stress in
the fracture process zone is the same as the yield strength there is no plastic zone only a
Barenblatt/Dugdale fracture process zone. The toughness indicated by JR in plane strain
ductile fracture does not come from an increase in the essential work of fracture with crack
growth, but comes from the work in the plastic zone surrounding the FPZ. Tvergaard and
Hutchinson [31] did not calculate the essential work of fracture. However, they reported that
the FPZ lengthens with crack growth. Hence according to Equation (15) the essential work
of fracture R is greater than R0 and decreases only to ~ as the propagation becomes steady
state. Hence, far from there being an increase in true toughness with crack extension, there is
actually a decrease in toughness!
For stable ductile crack growth over large crack extensions a large ao/ar" ratio is needed
so that large plastic zones are created. A large ao/try ratio is obtained if the volume fraction
of void nucleating inclusions is small (clean metals) and if there is a high strain hardening
rate [ 11, 31 ]. A FPZ that decreases in size will also increase that stability of fracture slightly

4. Conclusions

The early clarity of though in linear elastic fracture mechanics has become lost in post yield
fracture mechanics. The J-integral at initiation, JIc, is equal to the essential work of fracture,
R0. However, the use of the energetic interpretation depends upon the FPZ being very small
so that the deformation within it does not change significantly with crack growth. On the
other hand the/-integral for steady state propagation is zero if the FPZ is infinitesimally
small. However, work must be expended in the FPZ and the conclusion is that it is essential to
model the FPZ as a finite zone. In this case the/-integral gives the essential work of fracture
R0. Under near steady state crack propagation conditions in large specimens with contained
yielding, the J-integral, obtained from its energetic interpretation (Equation 2), contains the
plastic work and residual strain energy as well as the essential work of fracture.
While discussing crack growth resistance of advanced ceramics, it is clearly recognized
that there is no tree increase in toughness, but only a stress shielding of the crack tip. However,
the distinction is not usually clear in discussions on ductile fracture of metals and most papers
give the impression that there is a real increase in toughness. The original use of crack growth
resistance curves to describe the fracture of high strength thin sheets [19] is valid because
there is a true increase in toughness as shear lips develop, but in essentially plane strain ductile
fracture there is little or even negative true crack growth resistance. JR-curves do not give
fundamental information about the fracture process. They are size and geometry dependent,
not simply because of non-proportionate deformation during crack growth, but also because
they represent changes in the plastic work. With crack growth, JR-curves contain increasing
amounts of plastic work performed outside of the FPZ and simply reflect the increase in the
screening of the energy release rate by the plastic zone. Thus the basic problem in ductile
fracture is that the essential work of fracture in the FPZ is not separated from the plastic
work. These two work terms scale differently in fully-yielded specimens and need to be
separated before laboratory specimens can be sensibly used to predict large scale behaviour
[32, 33]. By combining terms that scale differently, the ductile fracture community has made
 A review of the J and I integrals 371

its resistance curves size and geometry dependent. The concrete community realizing, after
an early flirtation, that its crack growth resistance curves were size and geometry dependent,
albeit for a different reason, abandoned them and included a FPZ in their fracture model. The
fracture parameters for the FPZ are obtained by applying the fracture model to a small scale
fracture test. Armed with the parameters for the FPZ, the load-deflection behaviour for any
structure can be predicted [20]. In principle the same approach can be used for ductile fracture
and strides have been made in this direction by the ductile fracture community [11, 14].
Is it not time that the ductile fracture community recognized that JR-curves obtained from
specimens with large plastic zones or fully yielded ligaments are not very useful? It is realised
that much effort has been expended on obtaining JR-curve data which simply cannot be
thrown away. However, by modelling the FPZ these JR-CUrves could be reanalysed to obtain
the parameters for the FPZ, which could then be used to predict the behaviour of full-sized
structures accurately.

Acknowledgements
Prof C.F. Shih is thanked for his very useful comments on the draft of this paper. Support
for this work from the European Economic Community under Contract CI1"-CT93-0327 is
gratefully acknowledged.

References
1. J.R. Rice, Mathematical Analysis in the Mechanics of Fracture, in Fracture, H. Liebowitz (ed.), Academic
Press, New York (1968) Vol II, 192-314.
2. S.X. Wu, Y.W. Mai and B. CottereU, Prediction of the initiation of ductile fracture. Journal of the Mechanics
Physics Solids 43 (1995) 793-810.
3. J.W. Hutchinson and P.C. Pads, Stability analysis of J-controlled crack growth, in Elastic-Plastic Fracture,
ASTM STP 668, J.D. Landes, J.A. Begley, and G.A. Clarke (eds.), American Society for Testing and
Materials, Philadelphia (1979) 37--64.
4. J.W. Hutchinson, Singular behaviour at the end of a tensile crack in a hardening material. Journal of the
Mechanics Physics Solids 16 (1968) 13-31.
5. J.R. Rice and G.E Rossengren, Plane strain deformation near a crack tip in a power-law hardening material.
Journal of the Mechanics Solids 16 (1968) 1-12.
6. J.W. Hutchinson, Fundamentals of the phenomenological theory of nonlinear fracture mechanics. Journal of
Applied Mechanics 50 (1983) 1042-1051.
7. C.E Shih and M.D. German, Requirements for a one parameter characterization of crack tip fields by the
HRR singuladty. International Journal of Fractare 17 (1981) 27-43.
8. R.M. Mc Meeking and D.M. Parks, A criterionfor J-dominance of crack tip fields in large scale yielding, in
Elastic-Plastic Fracture, ASTM STP 668, J.D. Landes, J.A. Begley and G.A. Clarke (eds), American Society
for Testing and Materials, Philadelphia (1979) 175-194.
9. P.E Thomason, An assessment of the validity of J-controlled crack growth and stability of cracks in
incremental-plastic/elastic solids. International Journal of Fracture 44 (1990) 259-281.
10. J.C. Newman, B.C. Booth and K.N. Shivakumar, An elastic-plasticfinite-element analysis of the J-resistance
curve using a CTOD criterion, in Fracture Mechanics. 18th Volume, ASTM STP 945, D.T. Reed and R.E
Reed (eds), American Society for Testing and Materials, Philadelphia (1988) 665-685.
l 1. L. Xia, C.F. Shih, and J.W. Hutchinson, A computational approach to ductile crack growth under large scale
yielding conditions. Journal of the Mechanics Solids 43 (1995) 389-413.
12. J.W, Hancock, W.G. Reuter and D.M. Parks, Constraint and toughness parameterized by T, in Constraint
Effects in Fracture, ASTM STP 1171, E.M. Hackett, K.H. Schwalbe and R.H. Dobbs (eds), American Society
for Testing and Materials, Philadelphia (1993) 21-40.
13. J.A. Joyce and R.E. Link, Effects of constraint on upper shelf fracture toughness, in Fracture Mechanics:
26th Volume, ASTM STP 1256, W.G. Reuter, J.H. Underwood, and J.C. Newman (eds), American Society
for Testing and Materials, Philadelphia (1995) in Press.
14, L. Xia and C.F. Shih, Ductile crack growth- I.A. A numerical study using computational cells with microstruc-
turally based length scales. Journal of the Mechanics Physics Solids 43 (1995) 233-259.
372 B. Cotterell a n d A.G. Atkins

15. K.B. Broberg, The importance of stable crack extension in linear and non-linear fracture mechanics in
Prospects in Fracture Mechanics, G.C. Sih (ed.) Noordhoff, Leyden (1974) 125-138.
16. K.B. Broberg, On stable crack growth. Journal of the Mechanics Physics Solids 23 (1975) 215-237.
17. A.G. Evans, Perspective on development of high-toughness ceramics. Journal of American Ceramic Society
73 (1990) 187-206.
18. B. Lawn, Fracture of Brittle Solids 2nd. ed., (1993) Cambridge University Press, Cambridge.
19. J.M. Krafft, A.M. Sullivan, and R,W. Boyle, Effect of dimensions on fast fracture instability of notched sheets,
in Proc. Symp. Crack Propagation, Cranfield, College of Aeronautics, Cranfield (1961) p. 8.
20. B. Cotterell and Y.W. Mai, Fracture of Cementitious Materials (1996) Blackie Academic and Professional,
Glasgow.
21. B. Moran and C.F. Shih, Crack tip and associated domain integrals from momentum and energy balance.
Engineering Fracture Mechanics 27 (1987) 615--642.
22. J.W. Hutchinson, On steady quasi-static crack growth, Harvard University Rep. Division of Applied Sciences
DEAP S-8 1974.
23. H. Buggisch, D. Gross and K.H. Kriiger, Einige Erhaltungssfiatze der Kontinumsmechanik vom J-Integral-
typ. Ingenieur-Archiv 50 (1981) 103-I 11.
24. G.P. Cherepanov, On quasibrittle fracture. Journal of Applied Mathematics and Mechanics 31 (1968) 1050-
1058.
25. J.C. Amazigo and J.W. Hutchinson, Crack tip fields in steady crack growth with linear strain hardening.
Journal of the Mechanics Physics Solids 25 (1977) 81-97.
26. P.E Thomason, Non-singular stress and velocity fields for crack-tip plastic zones. International Journal of
Fracture 64 (1993) 27-42.
27. B. Cottereli, K.Y. Lam, Z. Chen and A.G. Atkins, The effect of geometry and size on ductile fracture, In
Proc. IUTAM Symposium on Nonlinear Analysis of Fracture (ed. J.R. Willis) Kluwer Academic Publishers,
Dordrecht (1996).
28. D.S. Dugdale, Yielding of steel sheets containing slits. Journal of the Mechanics Physics Solids 8 (1960)
100-104.
29. J.N. Goodier and F.A. Field, Plastic energy dissipation in crack propagation, in Fracture of Solids, D.C.
Drucker and J.J. Gilman (eds.) (1963) Wiley (Interscience), New York 103-118.
30. G.R. Irwin, Fracture mechanics, in Structural Mechanics: Proc. 1st Syrup. on Naval Structural Mechanics,
J.N. Goodier and N.J. Hoff (eds.), Pergamon, New York (1960) 557-591.
31. V. Tvergaard and J.W. Hutchinson, The relationship between crack growth resistance and fracture process
parameters in elastic plastic solids. Journal of the Mechanics Physics Solids 40 (1992) 377-397.
32. A.G. Atkins and Z. Chen, Scaling in elasto-plasticfracture mechanics, Proc IUTAM Conference on Scaling,
Turin Oct. 1994.
33. A.G. Atkins, Scaling in combined plastic flow and fracture. International Journal of Mechanical Sciences 30
(1988) 173-19I.