Engineering Fracture Mechanics Vol. 36, No. 3, pp.

403416, 1990 0013-7944/90

$3.00 + 0.00
Printed in Great Britain. Pergamon Press plc.

A THEORY OF DAMAGE IN BRITTLE MATERIALS

K. C. VALANIS
University of Portland, Portland, OR 97203, U.S.A.

Abstract-The present paper addresses the geometric representation of damage in brittle solids, its
equation of evolution and its incorporation in the constitutive equation. The concept of a damage
coordinate is introduced and a thermodynamic derivation of the evolution equation follows. A
continuum damage theory ensues. The theory is applied to the case of a thin plate with a central
crack under a tensile load in a direction normal to the plane of the crack. The fracture stress is
calculated numerically using a finite element code for various crack lengths and a comparison is
made with its observed value. Excellent agreement between the two is obtained with the aid of only
one damage parameter, the damage propensity constant. A congruence between the theory and
linear fracture mechanics is then shown when the damage coordinate is the crack length.

1. INTRODUCTION
DAMAGEplays an essential role in the constitutive behavior of materials. Its nature, in terms of
its measurement and mathematical representation is complex. It can vary from a well-defined crack
to randomly distributed damage that can be thought of as a “damage field”. It is important to
understand under what initial damage conditions, stress history and material properties, one type
will prevail over the other, or when the two types will co-exist.
Constitutive theories in the context of continuum damage have been proposed by Dragon
and Mroz[l], Lemaitre and Lazars[2], Chaboche[3], Krajcinovic and Fouseka[4], Fouseka and
Krajcinovic[S], and Krajcinovic and Fanella[6]. In this regard see also a recent paper by
Charboche[7].
Since, however, failure often occurs by the extension of one dominant crack, at times in the
presence of distributed damage, a comprehensive theory must be able to deal with both types, and
their mixture, simultaneously. Specifically a continuum theory should be able to deal with damage
due to well-defined sharp cracks. The theory in this study is being pursued with this goal in mind.
The state of directional damage (in the form of oriented cracks) is represented by an integrity tensor
4, and randomly distributed scalar damage by a scalar function R, both of which may co-exist.
While both types of damage were dealt with in previous work[8], in this paper we concentrate
on damage which is directed and begin with the case of a well-defined crack. The reason for this
is born out of the foregoing discussion. In addition, however, there is the question of linear elastic
fracture mechanics (LEFM) which has been successful in dealing with fracture due to sharp cracks
in brittle materials and in mode I conditions. In developing the theory we feel that we must establish
a conceptual boundary with LEFM.
To this end we use the continuum damage theory developed herein, to solve the problem of
a plate under tension with a central crack, normal to the direction of pull. Specifically we determine
the fracture stress as a function of the crack length and compare with experimental data on gray
cast iron, as well as with the solution obtained from linear fracture mechanics. The agreement
between the damage theory and experiment is excellent. The fracture mechanics solution disagrees
with the experiment for values of the crack length that are small but is otherwise in close agreement
also.
In addition to these overall results, the theoretical study has brought forth two central findings
which are the following:
(i) The evolution equation-with a firm basis in thermodynamics in the sense that it abides
by the law that the rate of change of a thermodynamic variable is proportional to the dual
thermodynamic force-gives, in the course of increasing tension, a damage field which is very
localized and is confined essentially to the vicinity of the crack root, in accord with observations
on brittle fracture and our own experiments on gray cast iron.
403
404 K. C. VALANIS

(ii) The proposed damage theory and linear fracture mechanics are brought into correspon-
dence in the context of the thermodynamic theory, when the damage coordinate is the crack length.
We have therefore a conceptual as well as a quantitative confluence among (a) the continuum
theory, (b) linear elastic fracture mechanics and (c) experimental observation in mode I conditions.

2. DAMAGE THEORY

We begin by setting out the fundamentals of a theory for incorporating directly the effects of
damage in the constitutive equation of a linear elastic material. A work which addresses the
behavior of elastic-fracturing and plastic-fracturing solids has appeared in two unpublished reports
by Valanis et aL[8], and Valanis[9]. Here we shall be concerned only with directed damage. The
question of scalar damage in the presence of directed damage was dealt with previously by Valanis
et a1.[8].
The fundamental premise of the theory is that directed damage may be represented by a tensor
valued function 4, “the integrity tensor”, which appears explicitly in the equation for the free
energy and the constitutive equation. Thermodynamically, 4 is an internal variable. In the case of
elastic-fracturing solids the free energy density is given by eq. (2.la) and the constitutive relation
by eq. (2. lc). Specifically[8],

* = (n/2)4ij$NcijEkl + WPik4j&&l~ (2.la)

Thus, since:

d = a+lat (2.lb)

it follows that:

Qij= @ij4klckl+ 2P+ik4jljltk/~ (2.lc)

Equation (2.la) is a statement to the fact that

* = 1/2Gj~,6j% (2.2)

where C, the stiffness tensor, is homogeneous and quadratic in 4.

There are two fundamental questions regarding the tensor 4:
(i) What is its geometric nature. More specifically, how is it related to the actual material
damage at some microscale-i.e. a scale smaller than the one that underlies the macroscopic
constitutive behavior of the material.
(ii) How does 4 evolve with the history of deformation.
In regard to question (i), it was previously shown, Valanis et a1.[8], that if in a material element
a principal value of 4, say 4,, is zero, and fir is the eigenvector of #,, then the traction on a surface
normal to &r is zero. That is, the element cannot support shear or direct stress on that surface.
Thus, r#~,represents damage on a plane normal to R,,, and as such it is a measure of the effective
area in the presence of microcrack in that plane. Thus when 4, = 0, an efictive plane crack has
developed across the entire element on a plane normal to 13,. Conversely if 4, = 1 no damage has
occurred on that plane.
With regard to question (ii), i.e. the evolution of r$, the following equation was proposed as
a basis for the description of damage due to mechanisms of brittle (or semi-brittle) fracture:

d#“= -(Q:)“‘d~“. (2.3)

The notation in eq. (2.3) is the following: d4” is the change in principal component 4” of 4 due
to a strain increment dc and n: are the eigenvectors of dc. Also, r” is a damage coordinate and
Q; a “damage force” in the direction ii”, i.e.

Q; = ngn;Qij (r, not summed) (2.4)

The power index m in eq. (3) is a material constant.

 Damage in brittle materials 405

The damage coordinate <” is given by eq. (2.5), i.e.

(2.5)
where C; = ~&n; (a, not summed) and k is a positive scalar. A complete discussion of these
equations when Q = $ is given by Valanis et af.[8].

Further discussion of eq. (2.2)

The appropriate “driving force” is of central importance to the evolution equation. In ref. [8],
Q was set equal to 4 because the resulting relation was of a type that governs annihilation of species
and, therefore, degradation. In brittle materials such as gray cast iron, however, and in the case
where damage is due to a well-defined crack, we found that this equation overestimates the damage
in zones of low strain intensity[8]. This result is contrary to the behavior of brittle materials and
our own observations on gray cast iron.

The physical fracturing process

The process of fracture of brittle and semi-brittle materials, such as gray cast iron which
exhibits a small amount of plasticity but fractures at very low strain (~0.5%)) shows two essential
characteristics which are always consistently and clearly observable:
(i) The damage is at first very minimal, but then increases suddenly becoming catastrophic.
(ii) In the case of a single crack, material regions away from the crack remain essentially
damage-free during the deformation process while damage, which leads to catastrophic failure,
remains confined to the tip of the crack.
The above characteristics must be depicted quantitatively by the damage evolution equation
whose suitability or otherwise, will be determined, in part, by its capability to describe the essential
physics of the damage process.

A thermodynamic approach
To arrive at a suitable evolution equation we appeal directly to thermodynamics. We recall
that since 4 is an internal thermodynamic variable, the thermodynamic force Q that drives the
damage process is given by the relation

Q= - a*latp. (2.6)

Also the direction of the damage increment -dQ is dictated by the eigendirections A” of the
increment of strain dc. In view of the above observations we make the following two fundamental
stipulations:
(i) The values d4 and de are coaxial.
(ii) The operative force driving the damage process in the direction ii” is Q:,which Q projected
in direction ii” in accordance with eq. (2.3). Using then the basic linear thermodynamic law “that
the rate of change of a thermodynamic variable is proportional to the conjugate thermodynamic
force”, we obtain the relation:

(2.7)

where b is a material scalar and

(2.8)

k having the significance of a damage propensity parameter. As before, dC’ are the eigenvalues
of dc.
406 K. C. VALANIS

Previously, when eq. (2.2) was used[8], the following were the conditions that dta be nonzero:

dc”>O, 6:: =: Cij?l~nJ2 0. (2.9)

Otherwise dr” = 0. Tiowever in the present case the explicit appearance of the the~odynamic force
in the evolution equation brings the Clausius-D&em dissipation inequality, i.e.,

(2.10)

into play directly. Thus in view of eqs (2.6) and (2.7) and the fact that

d#ij = c d4*n;nq (2.11)

?I” being eigenvectors of d+, it follows that

Since eq. (2.12) must be true for each c1 individually, i.e. for each independent fracture
mechanism, then the following inequality must be statisfied for all a:

a+
-dlpaKn;n; > 0. (2.13)

Furthermore, since d@’ is always negative whenever it is nonzero, then

for all a, whenever ld4”] # 0.

We have thus obtained the full set of conditions that govern the evolution of the damage
coordinate d{ a. Thus:

d<” = k de” (2.15)

whenever:

(2.16)

but dT”= 0 otherwise.

We recall that the search for a new evolution equation was motivated by the fact eq. (2.2)
overestimated the rate of growth of damage in regions of low strain intensity. To see that eq. (2.7)
is superior to eq. (2.2) in this respect we write it in its explicit form given below:

d@” = -6 d~‘{~~ij(4~,+,) + 2CL#kt~ik~j,jn~nJ* (2.17)

Note that the right hand side of eq. (2.17) is now of the order IIf.]12-whereas previously it was of
the order ]]c]]O.Other things being equal the rate of diminution of (b” will, therefore, be expected
to be substantially higher in regions of high strain intensity such as the root of an existing crack.

Material parameters and their experimental determination

It may be seen from eqs (2.7) and (2.8) that b and k merge into a constant bk so that without
loss of generality we take b to be equal to unity. There are, therefore, three constants to be
determined: k, the fracture susceptibility, and the two elastic constants A and p. In a uniaxial test
Young’s modulus E was determined, while v, the Poisson ratio, was taken to equal 0.03. The
elastic constants A and p were then calculated and were found to be 15 x lo3 and 10 x lo3 ksi,
respectively.
 Damage in brittle materials 407

The constant k is determined from a simple tension test on an untracked specimen. Under
these conditions

6, = Eq 4: (2.18)

and using eqs (2.17) and (2.18)

d+, = kE&$ dc,. (2.19)

Equation (2.19) may be integrated to give:

4, = e-W/W:_ (2.20)

Thus, in view of eqs (2.18) and (2.20)

6, = EC, exp((2k/3)Ec:). (2.21)

Fracture will take place when cr, reaches a maximum value. The condition is

da, /dc, = 0 (2.22)

which, in conjunction with eq. (2.21) gives the relation

L, max= (1/2kE)‘” (2.23)

aI max= E(1/2keE)‘j3 (2.24)

where e in eq. (2.24) is the base of the natural logarithm.

In a series of experiments on gray case iron specimens under tension the fracture stesss a, max
was found to lie within the limits

32 x lo3 < a ,max< 38 x 103psi

while E was found to be 26.6 x lo6 psi. Thus, choosing k to be equal to 2000 in units of l/(ksi),
a, maxwas found to be 40.2 ksi which is close to the upper bound of the experimentally determined
value.

Computational study
As discussed in the introduction, the computational study consisted of:
(a) Developing a finite element analysis program, in the presence of the damage evolution
equation, for the computation of the deformation, stress and damage fields in a flat plate in the
presence of a central crack.
(b) Computing these quantities for central cracks of various lengths.
(c) Developing a numerical algorithm for the above purpose.
The numerical experiment performed was one of the longitudinal displacement control (in the
presence of free transverse displacement) on the outer bounda
stress-free. An incremental displacement of 2.5 x 10W3u rlit&zi::$:t lzXdzi;:. Ylz
computations were carried out for a square grid of fixed size. For reasons relating to the geometry
of the specimen, 3/64” was designated as a unit of length. Half-domains were used in the
computations utilizing the longitudinal axis of symmetry.
To test the effect of specimen length on the fracture stress two such domains were used. The
first had a width of eight units and a height of eleven units of length. The second had the same
width as the first, but a height of twenty-one units. Stress, strain and damage fields were obtained
in both cases. For the domains tested the length effect was negligible. Fracture was considered to
have occurred when the displacement field was such that two halves of the plate (symmetrically
located relative to the crack line) began to move relative to each other as rigid bodies.

Brief description of the numerical algorithm

The boundary value problem for a brittle material in the presence of initiation and evolution
of damage is non-linear, and there are strong coupling effects between the elastic stiffness and the
408 K. C. VALANIS

state of damage. The coupling is manifested through the constitutive matrix [k,] which is directly
dependent on the integrity tensor 4. Specifically the constitutive response in plane stress is given
by eq. (2.25) where

(2.25)

where y12 is the engineering shear strain and

1
4:lQ. +%I Wll422 + b#d2) 4ll(d22(~ + 2cL)

M = @4l42 + %#4l) 442@ +%L) 422412@ +a1 .

(2.26)
[ 4l,dJl2(~ + 2P) A2422@ +%u) (1 +Po4:2+P#+f?52

On the other hand the evolution equation for the integrity tensor Q involves both the state of
damage (given by $), the state of strain L as well as the increment of strain dc which determines
the damage coordinate increments dt” (tl = 1,2, 3). Specifically:

d+” = Q; d<” (a not summed) (2.27)

QE = Qiin;nq (2.28)

Q, = A&,(+*c)+ 2/4k/cikcj/* (2.29)

The algorithm adopted here was one of incremental application of boundary displacements
in conjunction with the Newton-Raphson method of iteration. In the equilibrium configuration
each note is in equilibrium. Loads applied externally plus loads applied by the elements must sum
to zero. If this sum is [AR] instead of the null vector, the load imbalance [AR] will produce a
displacement difference vector [AU] which is computed by solving the equation

Kl[Aul = WI (2.30)

where [&I is the stiffness matrix of the structure in absence of damage. In iterative notation:

bl;+’ - [u]; = [K,,]- ‘[AR]: (2.31)

where k denotes the iteration number, n the incremental loading step and

[A# = [&Jnb4, (2.32)

i.e. [AR]: is calculated using the current secant stiffness matrix [KJ and the current iterated
displacement of: .
The imbalance “node force” AR and the displacement difference between two iterations will
decrease during the iteration process. Computation is stopped when a tolerance is reached, i.e. when
k+l_
IU” ~51 <euk (2.33)

where e is a small number. In this study, e was set equal to 10m6.

The plate was considered as having fractured when the difference in the displacements of any
two nodes in either the top or bottom part of the plate relative to crack line, was less than 10m6.
The numerical computations were conducted on a VAX 750.

Discussion of results
Figure l(a) shows the calculated curve of fracture stress vs crack length, based on the damage
theory. Also shown, are the experimentally determined values of fracture stress. The comparison
shows excellent agreement between calculated and experimental data. Figure l(b) shows the
calculated fracture curve when linear elastic fracture mechanics is used. It is apparent that LEFM
gives good agreement with the data for larger values of crack length, as expected.
Figures 2, 3 and 4(a) show, respectively, the stress, strain and damage distribution along the
crack line when the half-crack is one unit long. Also shown in Fig. 4(a) is the damage distribution
along the elements just below the crack line-again when the half-crack length is one unit. It may
 Damage in brittle materials

KIC
- o;=z F(a/b)

KX,
WV_ *c = x

KI,* 13.2 kri-in”2

...... Experimrntol points
[Gray cast iron)

Crack length a (in units of 3/64”) Crack length (in units of 3164”)

(4 04
Fig. I. (a) Caiculated vs observed fracture stress using the damage theory. (b) Comparison of LEFM
predictions with experimental observation.

be seen that the calculated damage away from the crack is indeed negligible. In Fig. 4(b) we show
the damage distribution along the crack line when the half-crack is one unit long and three units
long-for comparison.
Figure 5 shows the stress-strain curve for the element adjacent to the tip of the crack when
its half length is three units. Material softening is clearly visible leading to failure when the stress
reaches a maximum value. Figure 6 shows a similar type of behavior, when the stress of the element
adjacent to the crack tip is plotted vs boundary displacement-for a crack whose half-length is
again 3 units.
The attenuation effect of damage on the stress concentration is shown clearly in Fig 7, while
Fig. 8 shows the evolution of damage in the element adjacent to the crack tip vs boundary
displacement. It may be seen that damage proceeds at a slow rate in the range of small strains,
accelerating rapidly to a catastrophic rate as the strain increases. The half length of the crack was
again three units.

3. DAMAGE THEORY IN THE CONTEXT OF FRACTURE MECHANICS IN MODE I

As pointed out in the Introduction our purpose in this study is two-fold. We wish to examine
the proposed theory in the light of experimental data, as well as in the context of linear fracture
mechanics and thereby give perspective to the equations of the theory. Moreover we are also aiming
at determining how 4 depends quantitatively on the microdamage when this consists of a
wellde~ned crack of a certain extent and direction relative to a material micr~l~ent. Of course

Distoncrto
crock tiptin unitlwth) Distanceto crock tip (in unit lmpth)
Fig. 2. Stress distribution along crack line. Fig. 3. Strain distribution along crack line.
410 K. C. VALANIS

.
JO-

a Along crack ltwtth

Distance to the center of cmck (in unit length) Distance to crack tip (in unit length)

W (b)
Fig. 4. (a) Damage distribution along and below crack line. (b) Damage distribution along crack line.

a sub-part of our micromechanical aim is to determine the effective crack “extent” when the
damage is not mathematically well defined. However, this last part must await the findings of the
first phase of this work.
We begin by considering a thin sheet of “large” extent (and unit thickness) under axial
stress in the presence of a through crack normal to the direction of stress. Within the sheet, is a
circular region of radius R. The strain energy of the circular region in the absence of the crack
is YO where
1 cr2
Y 0=--nR2. (3.1)
2E

According to the Griffith calculation[9], in the limit of very large R, or vanishingly small u, the
reduction AY in strain energy due to the onset of the crack at fixed strain E is given by eq. (3.2).

(3.2)

Hence, the free energy of the circular region in the presence of the crack is:

Y=T,-dY=;;n(R’-a’) (3.3)
or

‘I’ = ;; [1 - (a/R)‘]nR2 (3.4)

or
Y = ;E~2[l - (a/R)~xR2 (3.5)
since the far field axial stress is related to the far field axial strain by the relation: e = EC

35 -

30-

Strain (x IO-9 Boutiry dir~mnl x 16% unit Length)

Fig. 5. Stress-strain curve for the tip of crack. Fig. 6. Stress vs boundary displacement.
 Damage in brittle materials 411

Boundary airpumment
x IO-4finuni-t1sngtti 8aundary diqilmemant (X IO?

Fig. 7. Stress concentration factor vs boundary displace- Fig. 8. Damage evolution vs boundary displacement.
ment.

In the Griffith fracture theory the “energy release rate” at fracture is equal to the rate of
creation of surface energy i.e.,

(3.41

since for an increment of crack length da the surface energy created is 2y da. Thus using eqs (3.5)
and (3.6) in the limit of R-+oo,

(3.7)

which is the Griffith equation for the critical stress, i.e.

(3.8)

The present theory

It is easily shown that in the case of a thin plate under axial stress conditions (+# = 0,
i #j, & = c#J~~ = 1) eq. (2.la) gives:

‘P = t/2E&f4. (3.9)

Thus UTwhich is the energy of a regin with radius R becomes

Y = nR2JI = l/2Er2c#+R2 (3.10)

comparison of eq. (3.10) with eq. (3.4) shows that in the limit of a vanishingly small crack
length

(t)=J_. (3.11)

This is a highly interesting result which though precisely valid only for small values of a, il
nonetheless satisfies, for circular regions, the condition (9, = 0 for a = R. Thus, when the crack haI
spread across the entire circular region Q, = 0 as expected. This suggests that in a strip of finite
width b and height h

4 =A1 - (a/b12, h/b) (3.12:

where the form of the function f is to be determined from an exact elastic solution of the
problem.
412 K. C. VALANIS

We now return to the Griffith crack criterion, i.e. eq. (3.6). Regarding the entire circular region
as a thermodynamic system, 4, is an internal variable of the system and since c$, depends on a,
then so is a. We proceed to define a generalized strain t where

6 =ZR,,& (3.13)

and a generalized stress 5 where

a=eRJ;E (3.14)

then

d =awlar=E+: (3.15)

recovering the relation

0 = Ecc+:. (3.16)
It also follows from the usual thermodynamic considerations that -aY/aa is the internal force
operating on the fracture mechanism, i.e. the “driving thermodynamic force” that is the cause of
crack growth. In view of eqs (3.10) (3.11), (3.13), (3.14) and (3.15):

ay
--_=
a2
(3.17)
aa [l -~/R)z]z~’

Now if we use linear thermodynamics of internal variables[lO], the classical equation of evolution
for a is

H/au + rj da/d[ = 0 (3.18)

where 1 is the resistance to cracking and d< is an intrinsic time measure and q is the resistance
coefficient of the internal fracture mechanism. In our case 5 is the damage coordinate. If, therefore,
we set

dc = da (3.19)

eq. (3.18) becomes

-aY/ai2 = rj. (3.20)

Thus eqs (3.6) and (3.18) are completely reconciled if q = 27. Therefore, using eq. (3.20)

a, = (3.21)

which in the limit of vanishingly small a agrees with the Griffith crack criterion, but is also
qualiratively true for finite regions of width R in the range 0 < a < 1, as will be shown.
We have thus demonstrated that the Griffith fracture theory is totally consistent with the
damage theory proposed here, in the context of irreversible thermodynamics put forth above, for
the choice of damage co-ordinate given in eq. (3.19). It also provides a means for determining the
dependence of 4 on the crack extent in mode I, for crack lengths that are small in relation to the
size of the size of the overall thermodynamic system.

4. GENERALIZATION TO OTHER DOMAINS

We wish to extend the above ideas to other domains. To this end we begin with a strip of
finite width as shown in Fig. 8. Our purpose here is to use linear fracture mechanics as a basis
from which to view the damage theory. Thus we are looking for a paradigm, an example of
convergence of the two theories, when the material is an elastic solid that fractures in an ideally
brittle manner.
 Damage in brittle materials 413

In accordance with eq, (3.9) the free energy of the plate (width 2b, height 2h) is given by the
expression

(4.1)

where
di =t#i: (4.2)
#* being the first principal component of the integrity tensor 4. Again we define a generalized stress
(5 and strain i’ such that
IS= bt?, c = hcz. (4.3a,b)

Thus the thermodynamic relation

d = ayaz (4.4)
gives rise to the

(4.5)

(4.6)

-at/i/as
= K:/E. (4.7)
Thus, in view of eq. (3.18) damage theory and fracture mechanics are brought into correspondence
if

q = K:/E t4.8)

and da = d{ in accordance with eq. (3.19).

Using eqs (4.7) and (4.1):

E+/b*)(a = K;/E (4.9)

where d = a@,@(1- az/b2) at constant b/h. At this point, using eq. (4.5) we find that

K:=o*a(&,lC~~) ; . (4.10)
0
In the notation of fracture mechanics[l l]

. (4.1 Oa)

Thus

(4. lob)

Since F is known from an elastic analysis, we have, therefore, established an exact connection
between the damage tensor and the crack length a under mode I conditions and the stipulation
that the damage co-ordinate e is a hnear function of the crack length,
EfM MC-D
414 K. C. VALANIS

More specifically if we write 1 - d/b* = x, h/b = y, eq. (4.10b) becomes

a i
--- =fP(x,y). (4.1Oc)
ax0Q
Thus
hlb

s
Qi= (4.10d)
c-n F’dx

where c is a constant of integration to be determined from the condition

@I+=, = 0. (4.10e)
Equation (4.9) provides another limiting condition on b. Fracture mechanics requires that[l 11, in
the limit of h + co
K: = a2naF2(b/a) (4.11)

and since under these conditions cr = EC, it follows from eq. (3.9) that

h&&D*(;) (4.12)

where the function @* is a function of (a/b) only.

Now if we use limiting conditions (4.6) and (4.12), and in view of eq. (4.10), we find that:
K,=aJ?m&@* (4.13)

thus providing a relation between the damage function and the fracture toughness coefficient in
mode I, for a strip of finite width. (Note if a, is the effective area representation of the damage
on a plane normal to Rr then $,(a,) must satisfy the end conditions 4(O) = 1; +(l) = 0.)

A specljic example

To illustrate the ideas involved in the section we consider a form of @ given in eq. (4.14), which
satisfies the limiting conditions Qi = 1 at a =Oand@=lath=co.Thusinthelimitofh+cowe
set:

(4.14)

and explore the extent to which K, derived from this form, agrees with the one derived from the
principles of fracture mechanics.
Use of eq. (4.10b) gives the result

(4.15)

In the specific case of the infinite strip (h = 00)

1
F= (4.16)

In Table 1 we compare the exact solution for F as given in [lo] with that of eq. (4.16). One
can see that the agreement is reasonably close for a trial solution such as the one given by eq. (4.14).

Table I. Comparison of the exact values of F with those of ea. (4.16)

alb 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
l/FW) 1 0.944 0.976 0.945 0.901 0.843 0.767 0.672 0.551 0.388 0
II - (a/bW2 1 0.955 0.980 0.954 0.916 0.866 0.800 0.714 0.599 0.435 0
 Damagein brittle materials 41.5

Equation (4.14), however, does not lead to a good representation of F for$&e values of h. Other
simple forms of # will be explored in the future.

5. MORE ON THE RATE EQUATION (3.18)

In the previous sections we showed a precise correspondence between the integrity tensor and
the stress intensity factor in the context of linear fracture mechanics. The key to this correspondence
is the evolution equation (3.18) in the specific instance when the fracture coordinate 4 is such that
d< = da. (5-l)
One disadvantage of linear fracture mechanics is that the evolution law for the crack length
degenerates into an instability statement in the sense that when the stress is below a critical stress
level the rate of growth is zero, whereas at the critical stress level the crack grows in an unstable
fashion.
Previously, we proposed an evolution law of damage given by eq. (2.2), where the damage
coordinate < is such that when d< > 0, then
d5 = k dc:, dr > 0. (5.2)
It is therefore of interest to explore the effect of eq. (5.2) on the evolution eq. (3.18). In this specific
case we find that the Griffith crack theory is used, in the limit of R+co, but in the context of
eq. (5.2):
am2E = (~/~)(da/d~). (5.3)
This equation may be integrated to give the relation
a = a0 ewn/34k3
(5.4)
where a, is the ~~~~~a~ crack length. Note that in this approach a crack will not grow unless
an initial crack length is assumed. While this result is physically unrealistic (in the sense that
eq. (5.4) does not account for crack initiation), it agrees with linear fracture mechanics precepts
whereby a specimen with a zero crack length cannot fail. Note, however, the rapid growth of a
with c3.

The question of coupling between a and e, must inevitably arise in so far as the definition of
the damage coordinate 6 is concerned. In keeping with our previous ideas on the concept of intrinsic
time we set dr equal to the right hand side of eq. (5.5)
d< = (k* a62 + da*)“* (5.5)
where k is a material constant that has the dimension of length. Thus in view of eqs (3.17), (3.18)
and (5.5) and in the limit of R -+co, one finds that
da kEnac*
‘;i; = ($ _ ,&2&4)W ’ (5.6)

Evidently, unstable crack growth occurs when da/dE = 00, i.e. where
q = Earn**= ana*/E (5.7)
or at

(5.8)

thus recovering the Griffith crack condition. However, eq. (5.6) is now a crack growth equation
Mich describes the rate of crack growth prior to instability. In view of eq. (5.6), for low values
of t
da/de = kEitac*fq (5.9)
416 K. C. VALANIS

and thus eq. (5.4) applies. The crack grows continuously until fracture takes place, when the
denominator of the right hand side of eq. (5.6) becomes zero, in accordance with the Griffith
condition. Thus the evolution eq. (3.18) in conjunction with the definition of damage coordinate
given by eq. (5.5) gives an ultimate Griffith crack condition, but also, a crack growth law prior to
instability.
These ideas will prove useful in the future for developing a comprehensive damage theory for
brittle or semi-brittle solids.

Efect of size on specimen strength

In eq. (3.8) the driving force on the fracture process is the rate of change of the entire energy
of the specimen with respect to the crack length (actually crack area of unit width). In the case
of a strip, height h and width 6, the fracture stress is given by eq. (4.10a). Thus, given two different
geometrically similar specimens so that a/b and h/b are the same in both specimens, but different
in size, the larger specimen will have a smaller fracture stress oC(K, being a material property), since

=c N 114 (5.10)
all other factors being equal. This rule is not satisfied precisely since the inverse square law is
satisfied only approximately and only so for larger values of a.
The question of size on fracture strength is basically an unresolved problem which could be
investigated in the future. The effect can be accounted for, however, in a formalistic manner by
stipulating that q (the damage resistance coefficient) is a function of the geometry in some sense
which is not quantized at this stage. Thus, we denote the geometry by G and we proceed to set

rl = V(G). (5.11)
This formalism allows one to treat the rate equation in terms of the average density $ of the free
energy of the specimen.

Acknowledgement-The author gratefully acknowledges the financial support provided by the Air Force Office of Scientific
Research, Directorate of Aerospace Sciences, Civil Engineering Program, Bolling Air Force Base, Washington, D. C.

REFERENCES
PI A. Dragon and Z. Mroz, A continuum model for plastic-brittle behavior of rock and concrete. Int. J. Engng Sci. 17,
121-137 (1979).
I21J. Lemaitre and J. Lazars, Application de la theorie de l’endommagament au comportement nonlineaire et al rupture
du beton de structure. Annales de L’ITBTP 41 (1982).
131J. L. Chaboche, The concept of effective stress applied to elasticity and to viscoelasticity in the presence of anisotropic
damage. Royal Aircraft Est., ONERA Report No. 1979-77, Farnborough (1979).
I41 D. Krajcinovic and G. U. Fouseka, The continuous damage theory of brittle materials, Part I. J. appl. Mech. 48, 809
(1981).
151 G. U. Fouseka and D. Krajcinovic, The continuous damage theory of brittle materials, Part II. J. appl. Mech. 48, 816
(1981).
[61 D. Krajcinovic and D. Fanella, A micromechanical damage model for concrete. Engng Fracture Mech. 25, 386396
(1986).
171J. L. Caboche, Continuum damage mechanics: Part I-General concepts; Part II-Damage growth, crack initiation
and crack growth. J. appl. Mech. 55, 59-72 (1988).
181 G. A. Hegemier, H. E. Read, K. C. Valanis and H. Murakami, Development of advanced constitutive models for plain
and reinforced concrete. S-CUBED report SSS-R-87-8454 (1987). _
[91 K. C. Valanis. A theorv of damage in brittle and semibrittle materials. ENDOCHRONICS Reoort Endic-OOS-AFOSR-
1988 (1988). - -
1101J. F. Knott, Fundamentals of Fracture Mechanics. Butterworth, London (1973).
1111 H. Tada, P. Paris and G. Irwin, The Stress Analysis of Cracks Handbook. Del Corporation, Hellertown, PA (1973).

(Received 10 April 1989)