Académique Documents
Professionnel Documents
Culture Documents
SUMMARY
THETWO-DIMENSIONAL
problem of an arc shaped crack lying along the interface of a circular elastic
inclusion embedded in an infmite matrix with different elastic constants is considered. Based on the
complex variable method of Muskhelishvili, closed-form solutions for the stresses and the displacements around the crack are obtained when general biaxial loads are applied at infinity. These solutions
are then combined with A. A. Griffiths virtual work argument to give a criterion of crack extension,
namely the de-bonding of the interface. The critical applied loads are expressed explicitly in terms of a
function of the inclusion radius and the central angle subtended by the crack arc. In the case of
simple tension the critical load is inversely proportional to the square-root of the inclusion radius.
By analyzing the variation of the cleavage stress near the crack tip, the deviation of the crack into the
matrix is discussed. The case of uniaxial tension is worked out in detail.
1. INTRODUCTION
THE ELASTIC problem
the interface
of two bonded
dissimilar
the stress
Later ERDOCAN
distribution
around a semi-infinite crack under plane strain conditions.
(1963, 1965a), ENGLAND (1965) and RICE and SIH (1965), using a complex variable
325
326
M. TOYA
uniform internal pressure along the interface of a rigid circular inclusion and obtained
the condition of the de-bonding of the interface (TOYA, 1973).
In this paper we consider a two-dimensional crack along the interface of a circular
elastic inclusion embedded in an infinite matrix and thus generalize the earlier results
obtained by the writer. It is assumed that each of the two materials is homogeneous
and isotropic and that the stress state at infinity is a general biaxial tension, the crack
faces being free from tractions. This elastic system provides a simplified model for the
consideration of brittle fracture of a real material containing an inclusion.
The mixed boundary-value problem is formulated on the basis of the complex
variable approach, and is readily reduced to an inhomogeneous Hilbert problem
(Section 2). The exact solution is obtained by a simple algebraic procedure, and the
stresses and the displacements at the interface are completely determined in compact
forms (Section 3). Our method of solution is similar to those of ENGLAND(1966) and
PERLMANand SIH (1967) but ours seems to be easier and more straightforward. Moreover, they did not find the displacements. Actually, the elastic field thus obtained
becomes invalid for certain combinations of stress states at infinity and the central
angle subtended by the crack arc, because the crack faces may come into contact with one
another and then the original boundary conditions change. The presence of a closedform expression of the displacements enables us to find by means of simple numerical
calculations the critical crack angle for which crack faces come into contact with one
another. The results for the case of uniaxial tension at infinity are shown graphically.
The condition for the crack to grow along the interface, namely the condition of
de-bonding of the interface, is derived by use of Griffiths virtual work argument
(Section 4). The critical applied stress is expressed in terms of a function of the
inclusion radius and the central angle subtended by the half-length of the crack. In the
case of the uniaxial tension the critical load is inversely proportional to the squareroot of the inclusion radius.
Finally, by investigating the stress state near a crack tip, the deviation of the crack
into the matrix, i.e. the fracture of the elastic system considered, is discussed (Section 5).
The application of the theory is to metallurgy and also to composite materials.
The case of epoxy-glass composite under uniaxial tensile stress at infinity is worked
out, as an example.
2.
FORMULATION
OF THEPROBLEM
327
MATRIX
(wl,)
FIG. 1. Model of a crack along the interface of a circular elastic inclusion embedded in an infinite
matrix with different elastic constants.
If the two faces of the crack do not make contact with one another, then the boundary
conditions of the interfacial crack problem are
(2.1)
and
(2.2)
where Gi and pi are components of stress on the interface in polar coordinates
uI and ui are components of displacement in Cartesian coordinates, and subscripts
1 and 2 denote the region 1 or S1 (the matrix) and the region 2 or S, (the circular
inclusion), respectively.
In polar coordinates the basic equations for two-dimensional classical linear
elasticity are
hi + Zi = W,(Z)+ ~i(~i(z>,
2(Gi+iZi)
4p,(au,/ae+ia0,/ae)
(2.3)
wi(Z)+~*(Z)-~~(Z)-(~/Z)5i(B),
(2.4)
= iz{rci ~(z)-~~(Z)+Z~~(Z)+(Z/Z)W~(Z))
(2.5)
zES1 for i = 1,
z E S, for i = 2,
(2.6)
M. TOYA
328
Wi(Z)
(a2/.Z2){
wi(a/Z)
W,(Z)
zES, for i = 1,
e S, for i = 2. >
-ZB$/(Z)},
(2.7)
(2.8)
Since k&(z)and wi(z) are holomorphic in the region Si, it follows from (2.7) that W,(z) is
holomorphic in S2 except at the point z = 0, where it may have a pole of order two
and therefore its principal part has the form
c,+cl/z+-c2/z2,
(2.9)
where ci (i = 0, 1, 2) are constants. On the other hand, W,(z) is holomorphic at
z = co and its principal part there is a constant.
Combining (2.4) and (2.6) and letting z tend to a point t (= a exp (i0)) of the
boundary from Sr for i = 1 and S, for i = 2, we obtain
W,+(t)- W;(t) = 2(%, + G,),
(2.10)
(2.11)
W:(t)- W;(t) = -2@,+iZ,),
where pi:: and Wi-(t) are the limiting values of W,(z) as z --) t from S, and S,,
respectively.
Similarly, combining (2.5) with (2.6) and using the relation
&A,/&?+iLYvi/~~= it(du,/dt + ido,/dt) = itvf(t),
which holds on the boundary IzJ = a, we obtain
(2.12)
KI w:(t)+ w;(r) = 4Plv;(r),
(2.13)
K2 w;(t)+
W2+(t)
=
@2V;(t),
where vi(t) = ui(t) + iui(t).
Now from (2.10) to (2.13) the boundary conditions (2.1) and (2.2) may be written
as
q+(t)-
P&-(t) = 0
w;(t)-w:(t)
on -4,
= 0
on As,I(4P,)-(K,W:(r)+
w;(t)) = (4Pz)-1(@%-(r)+
%+(r))on A,,
on A,. 1
w:(t)w;(t) = q-(t)w-z(t)
From (2.14) and (2.152) we can write
(2.14)
(2.15)
(2.16)
at every point of the circular boundary. Since W,(z) and W,(z) are holomorphic at
infinity we may conclude that W,(z)+ W2(z) is holomorphic over the entire plane
excluding the origin z = 0, where W,(z) may have a pole and having the principal
part given by (2.9). Therefore, we have
(2.17)
W,(z) + W,(z) = G-J+ Cl/Z + c2/zz
over the entire plane and the unknown constants cj (i = 0, 1,2) have to be determined
by the additional conditions. From (2.17), equations (2.14) and (2.15) may be written
as
(2.18)
q+(t)-w;(t)
= 0
on 4,
(2.19)
y+(t)+vFq-(t)
= B(co+c,/t+c2/t2)
on AD,
w:(t)
+ q+(t)
= q-(t)
+ w;(t)
329
(2.21)
P = &(l +%)/(Pl +KI Pz),
and W,(z) is holomorphic over the whole plane cut along the arc A, except at the
point z = 0. Thus, the interfacial crack problem is reduced to the solution of an
inhomogeneous Hilbert problem as given by (2.18) and (2.19). Once W,(z) is known,
W,(z) may also be determined from (2.17) and the components of stress and displacement may be determined from (2.3) to (2.5).
3.
STRESSES
ANDDISPLACEMENTS
AROUNDAN INTERFACIAL
CRACK
The solution of the Hilbert problem given by (2.18) and (2.19) can be written as
(MUSKHELISHVILI,
1953, Section 110)
exp (ia)]-*[z-a
exp
z- a
exp (LX)
(--WI-+ z_-aexpc_iccj ,
[
(3.2)
Iz = -(In v)/2n,
(3.3)
P(z) = Llz+do+d,/z+d,/z2.
(3.4)
It is seen that the function x(z) given by (3.2) is holomorphic over the entire plane
cut along the arc A,, and on A, x(z) satisfies the relation
x(t) = -q--(t).
(3.5)
and
(3.6)
z-u
(3.7)
W,(z) = (1-k)(co+cl/z+c2/z2)+k((co-d_~/k)z+A+B/z+C/z2}~(z),
where
(3.8)
k=-
P
1+v
A = c,-c,u(coszf2AsinI)-$!,
B = uexp[21(c(-n)]
C = a exp [21(c( - z)]
(3.10)
c1 -~(cos~-2~sincr)-_
exp [2% -
c2 - 2 exp [2&n-or)]
.
>
41 ,
1
(3.11)
(3.12)
330
M. TOYA
(3.13)
for large values of lz[, where E, denotes the rotationt
at infinity, and
(3.14)
n-a)] };+
k ( c2 -iexp[22(n-ol)]
>
-$, (3.16)
sin a)/~.
(3.18)
Next, we consider the behaviour of W,(z) and wz(z) at infinity and near z = 0.
Expanding W,(z) given by (3.8) near z = 0, we obtain
w,(z) N (l-k)c,
-I
+ $(B+CD)exp[2A(z-cz)]}:
+
f.
(3.19)
+ 2 + ...
(3.20)
(3.21)
(3.22)
where
a, = (l-k)c,
+ i (A+BD+CE)
exp [22n(rr-E)]
(3.23)
331
with
E = $
(-82
sin 21~+(1+4A.~)+(3-41~)
cos 2a),
(3.24)
and
& = CO-d-_l,
bl = c,(l-k)+kA+ka
(cO - 2)
(cos a+21sin
(3.25)
a).
To find w2(z) we substitute (3.21) and (3.22) into (2.8) and find that
w2(2) = &z,+A,)
f + 2 + ...
near z = 0.
(3.26)
+~(A+aD+CE)exp[21(7r-a)]+F,-&,
(1-k)e,+kA+ka(c,-%)(cosa+22sina)
= 0,
=O.
(327)
It is readily verified that equation (3.271) states that the moment about the center of
the stress applied to the boundary is zero while (3.272) states that the resultant of the
stress force applied to the boundary vanishes. Now, from equations (3.15), (3.17),
(3.20) and (3.27) we may determine the unknown constants. There are actually eight
conditions for the determination of the seven unknown constants. This is due to the
fact that the resultant force on the inclusion is assumed known, i.e. it is assumed a
priori to vanish, from the stresses calculated on the boundary. Hence, for example,
(3.272) is satisfied automatically.
Solving the above equations for the constants we obtain
\
Co = G+iH,
G = [+(N, + I,){ 1- (cos c(+ 21 sin a) exp [21(7c- u)]} -$(l - k)(l +4L2)(N, - T,) sin2 a cos 293 x
x 1/(2-k-k
(cos a+2A sin a) exp [2A(n-LX)]},
H=
+ z
{l+(cos
a+21sin
(3.28)
a)exp[ZA(n-LX)]~] x
(3.29)
(3.30)
(3.31)
B
C=
l-k
= k
-k
l-k
a2(N, - T,)(cos
a3(iV,-T,)
exp [2$-!-2A(a-n)],
(3.32)
(3.33)
M. TOYA
332
Hence, we have
W,(z) = k
c,+a2(N,-
W,(z) = (1 -k)
c,+a2(N,
(3.34)
- kPA4xW,
,i
>
-t kPAz)x(z),
(3.35)
where
P&z) = (CO-+)
(z-u(cos
l-k
a+21 sin a)} + 7
x a(N,-T,)
exp [2@+21(E-n)]
2(&?+ iu?) = -
(3.36)
(3.37)
From (2.12) and (2.13) the derivative of the relative displacement of the two faces of
the crack reduces to
(3.38)
v(r) = -A 1PAOX
with
1+tiz
A _ 5 l+rG + ___.
(3.39)
1- 4
( Pl
P2
>
By means of the formulae
J{t-u(cos
= y,
where
R(t) = [t-u
exp (icr)]*[t--a
exp (-icr)]+
(3.40)
t-u
t-u
exp (ia)
exp(-iia)
(3.41)
1
iA
(3.42)
which are easily verified by differentiation, (3.38) may be easily integrated and we have
(3.43)
u(N,-T,)
exp [2$+2A(a-n)]
i.
(3.44)
The function R(z) refers to the branch which is holomorphic over the whole plane cut
along the arc A, and which is such that R(z) -+ z as lzl + co.
By a similar process the displacement of the interface of the bonded part may be
obtained as
k(l+4
cg t - u2(N, -T,)exp(Zi~)~}-~(iCI-t)PI(I)H+(l),
(3.45)
4Pl
where the integration constant corresponding to a rigid body displacement has been
omitted.
We have thus obtained the exact solutions for both stresses and displacements on
the interface. These equations are rather general and some of the solutions which
VAD
___
333
have been treated separately may be obtained simply by taking suitable values for
A?,, T,, a and the material constants.
For the first example, let the matrix and the inclusion be the same material. Then,
we have the problem of the arc shaped crack opening up in a homogeneous material.
If the stresses at infinity reduce to a simple tension p in a direction making an angle C/J
with the real axis then (3.34) becomes
Wl(z) = 4
- >)}
Xl(Z),
(3.44)
where
cb = GfiH,
G
\
-cos
= g&p
p sin Msin 24
H=-------__
2(1+cosar)
(3.47)
and
Xl(Z) = (zz--2az cos a+l)_+.
J
The function xi(z) refers to the branch which is holomorphic over the plane cut along
the arc A,, and which is characterized by (3.6). W,(z) may also be obtained and the
stresses derived from these functions agree with the results obtained by ~USK~ISHvm
(1953, Section 124).
Second, consider the case where c1is zero. Then, we have the problem of a perfect
bond all round the disk (K, = (3 - qJ/( 1 + vi)) or the cylinder (Ki = 3 -4qi). We obtain
W&> = fp
1 2a2b2--id
1+
K(z) = 4P~
1
-3,
C11-tK1.4 z >
Cr20+ 75)
2112-h
+c11
k2
z f Sl,
(3.48)
z E s,,
(3,49)
which again give the stress field obtained by MUSKHELISHVILI(1953, Section 58).
For the third example, we assume that the inclusion is rigid. We let pz + co and
in this limit
W,(z) = kc,--
(kc, -d_ ,)(z - a(cos ;x+ 2Ai sin a)) + a2(N, - T,) x
ew C2Q + 2&G -
741x(z),
(3.50)
where
kc, = ikN = i f$s
{I +(cos ~+21,
sin a) exp
[2dI(7c--c41~+
x
I
a+211, sin a) exp [2&(n-(x)]j
(3.51)
with
A = (In ~,)/2n.
(3.521
334
M. TOYA
Since kc, is purely imaginary, the first term on the right-hand side of (3.50) represents
a rigid rotation of the inclusion, the angle of rotation being
(3.53)
As expected, E,,vanishes when the stress state at infinity is hydrostatic tension or when,
if the rotation E, vanishes, there is either not a crack at all or the crack is symmetrically
located with respect to one of the principal stress directions.
Reverting now to (3.37) and (3.43), we note, from the formula
([t-aexp(ior)]/[t-uexp(-ia)]}=[cos(;lln~[t-aexp(icc)]/[t-uexp(-icc)]~}+
+i sin (2 In /[t-a exp (ir)]/[t--a exp (-i!zf)](}] exp (--&I),
where w = arg {[t-u exp (icr)]/[t - a exp (- ia)] >, that the stresses and displacements
oscillate violently when t approaches the crack ends and this indicates that the upper
and lower faces of the crack will seemingly overlap one another. This peculiarity of
the solution seems to be inherent in the mixed boundary-value problem for interfaciai
cracks and it has been discussed in detail by ERDOGAN(1963, 1965a) and ENGLAND
(1965, 1966). They have shown that this unreuZistic aspect of the solution is confined
to very small regions near the crack ends. In the present case, too, the extent of the
regions of displacement and stress oscillation is very small. For example, as is discussed in Section 5, for an epoxy (matrix)-glass (inclusion) composite under simple
tension at infinity the regions of oscillation are confined to a small circle with its center
at each crack tip and whose radius is at most of the order of 10e3u. Consequently, it
may be acceptable to assume that (3.37) and (3.43) provide a good approximation to the
physical state of the body at the interface except in the immediate vicinity of the crack
en&.
Though the overlap of the displacements near the ends of the crack may thus be
permissible, from the mathematical formulation of the problem it is still required
that except in the vicinity of the crack ends the radial component of the relative
displacements of the two faces of the crack should be greater than zero. But this
condition is not always satisfied in (3.43), as may be seen at once by considering, for
example, the situation where the central angle subtended by the crack arc is close to 27~.
In practice, in the case of simple tension at infinity the relative displacement of the
two faces of the crack in polar coordinates is given by
1 2(1-k)u
exp [2i4 + 21(cr- TC)]
(3.54)
u,+iuO = -)A,pu
Go+iHo -i kt
>
where
G = 1 -(cos CI+2A sin cr) exp [2;l(n -a)] +(l - k)(l+4j1) sin2 cIcos 24,
(3.55)
0
2 - k - k(cos a + 2L sin a) exp [2i1(71- a)]
(1 -k)(l +4R2) sin2 c( sin 24
(3.56)
Ho = - k{l +(cos a+22 sin a) exp [22(7r--a>])
Or, substituting t = a exp (iQ), (3.54) may be rewritten as
u,i-iu, = -A, Pu[sin J(Lx-f?) sin *(a+@)]+ x
x
+iH
!l_2(1-k)
k
exp [ i(2+ -
k
x exp [n(n-a)
0) + 21(a - rc)] x
(3.57)
335
When a is close to z, U, takes a negative value on a certain part of the boundary for
any value of 4 and this means that there is an overlap of opposite faces of the crack.
On the other hand, when a is small enough and the materials of matrix and inclusion
have the same properties, u, is always positive and thus the solutions (3.43) and (3.45)
are valid for all 4. In Fig. 2, the critical values of 4 for which the crack faces come into
contact at a certain point are plotted for some typical combinations of materials under
conditions of plane strain. Each 4 vs. a curve was obtained by actually computing
(3.57) for various combinations of (p and a. The materials are taken to be epoxy
(pr = 3*46x lo5 lb/in2 or 2.39 GN mm2, q1 = 0.35) for the matrix and glass
(JL~= 6.41 x lo6 lb/in2 or 44.2 GN md2, q2 = 0.22) for the inclusion. The solutions
(3.37), (3.43) and (3.45) for uniform tension are valid only for those values of a and 4
which are in the region enclosed by each curve and the a-axis, where without loss of
generality 4 may be assumed positive.
80
EPOXY - EPOXY
-d
FIG. 2. Variation of the critical value of q+with CC,for which the crack faces come into contact at a
certain point.
When the crack faces come into contact with one another over a certain length a
much more complicated boundary-value problem than the present one will ensue.
Though it is necessary, strictly speaking, to solve this mixed crack-and-contact
problem in order to obtain the exact solution of the problem and hence to justify the
present approximate solution, we do not enter further on the subject in this paper.
4.
In Section 3 the exact solution for the components of stress and displacement on
the interface were obtained. In this section we apply Griffiths classical virtual work
argument to this solution and derive a criterion for the crack to grow along the interface, i.e. the criterion of the de-bonding of the interface. This will be done by allowing
336
M. TOYA
the end t, = a exp (ia) of the crack to extend to t; = a exp [i(c~+Q], 6 being small,
and calculating
the change in the total energy, i.e. sum of the elastic strain energy
and the potential energy of the loading mechanism.
The change of the total energy is
equal to the work needed to close up the crack to its original length (IRWIN, 1957).
The crack is stable if this work is less than the energy 2y,, ad required to produce the
new crack surface and is unstable otherwise (ylz is specific surface tension energy of
the interface).
Then, the criterion for the debonding of the interface is of the form
dU = -~Wlo16(~+i~),(u,--iu,),+,
where (z+z%),
is the traction when
is the relative displacement
(&-&J,+,
at t; = a exp [i(cc+6)], and98 denotes
As WILLIS (1971) has indicated, the
(4.1)
crack end t, = a exp (ig) is determined solely by the asymptotic forms of (&i +iz),
and (u,- iu,),+, near t, = a exp (ia), but that nevertheless (4.1) is a mathematical
consequence
of the global energy balance principle.
Hence, despite the violently
oscillatory nature of the stresses and displacements
near the end of the crack it is
justified to use equations (3.37) and (3.43) in (4.1) because they do provide a good
approximation
at most points and thus the overall energy of the cracked medium
should be very closely approximated
by the energy calculated from use of these
equations.
Taking the complex conjugate of (3.43) and substituting M = CC
+ 6 for a we obtain
(u,--&Ja,
in the vicinity
= -A,
N exp [2L(n-a)](2ia
sin c~)~+[t--a
exp (ice)]*-
(4.2)
iv+-$g+
y(N,---T,)exp[(2J,-r)i+2i.(a---rc)].
Also, near
2(a^a+iz)
= -k
1 +i
(
(4.3)
(3.37) becomes
ia(l+2ii)sincrx
>
x N(2ia sin cx)-f-i[t--a
exp (Lx)]-++.
(4.4)
Hence,
dU = -$kaA,
where
(4.6)
The functionf(t)
refers to the branch which is such thatf(t)
the integral (see Appendix II), we obtain finally
fkaA,(l
+412)7cNR
= 2y,,.
+41)nNON,,
in a direction
= 2y,,,
(4.7)
making
(4.8)
337
where
No =
G,,-l-%(I-k)vexp(%d)
+i
cm (Q-a)
-t.
HO-i(l-k)vexp(2cd)sin(2~#1-a)
>
(4.9)
M. TOYA
338
FIG. 3.
not necessarily mean symmetrical location of the crack with respect to the tension
axis. With further increase of q the crack size grows continuously until the load
necessary for the crack to deviate from the interface into the matrix or inclusion is
reached. Then, if the crack is assumed to deviate into the matrix, complete brittle
fracture may occur.
The considerations of this section have been based on the assumption that the crack
will always grow along the interface of the inclusion. This may be considered to be true
when the bonding strength between the inclusion and the matrix is su@ciently small.
But when the bonding strength is large it is possible that deviation into the matrix or
inclusion may occur without any accompanying de-bonding of the interface. In this
case, we cannot accept the foregoing discussion. In Section 5 we give a discussion of
this point.
5.
In this section the cleavage stress distribution near the crack tips is analyzed.
Then, based on this stress analysis, the direction of crack growth (not necessarily
parallel to the direction of the tip) is discussed. For simplicity of discussion we confine
our attention to the case where the strength of the matrix is lower than that of the
339
so that the crack growth may be identical with fracture of our elastic system.
Also, the stress state at infinity is assumed to be one of simple tension.
We calculate the variation of the magnitude of the cleavage stress at a small
distance pr from the crack tip t1 = a exp (LX)with angular location as given by the
angle 9, where
z = a exp (iol)+p, exp [+$~+a-9)]
(5.1)
(see Fig. 4).
inclusion,
iy
5 (p,/a)+(j+-x^x+2ix^y)
X
-(I-2iA)M,
exp [-n(x-9)]
(2 sin a)+
exp
9-a
-++lnxI
(5.3)
+iM,exp
{-n(a-9)+i(-~++39+~111~~))+
+ ?exp
(--n(a+a)+i(-*+@-3,lnx,)}
x1)}+
1,
(5.4)
into (5.2) the cleavage stress 9^9near the crack tip t, = u exp (ia) may be obtained.
It may be readily verified that when 9 = 0, $ agrees with the contact stress 2
calculated from (3.37) and that when 9 = 180, % vanishes identically. It may also be
seen that for very small values of x1 the sign of % will be alternately positive and
negative. This fact makes it difficult to obtain a reliable stress distribution around the
340
M. TOYA
crack tip, while in the case of a homogeneous material it can be obtained simply from
the coefficients of the p,%ingularity.
We first evaluate the length of the regions of displacement and stress oscillation
and choose a sufficiently large value for p1 exceeding this length and then calculate the
stress distribution at this distance from the crack tip. From (3.57) it may readily be
seen that the crack faces first come into contact at the point 0 = c1- y, where, to within
the first order of y,
y [1-($$-l)sinUexp(&_(2arctan(J,/J,)+a))]
= 2,I sin a exp & (2 arctan (Jl/J2)+a)),
(
(5.5)
where
J, = kGo - I- 2( 1 - k)v exp (21.~)cos (24 - cr),
J, = kH, -2(1J, = 2(1-k)v
(5.6)
sin (~c#J-a)}/J,.
Practically, the second term within the brackets on the left-hand side of (5.5) can be
neglected. Since y increases with increase in A, the maximum size of the region of
cross-over may be evaluated when v = JJ (for which /z has the maximum value O-175),
namely when the matrix material is incompressible (qI = 0) and the inclusion is rigid
(pZ = co) under the conditions of plane strain. For example, we obtain y = 0,011 for
CI= &C and 4 = 0 and hence the cross-over region extends approximately to I.4
per cent of the crack length. The size of the region is considerably smaller for real
materials. For the crack with CI= +rt and C$= O, lying along the interface of the
epoxy-glasscompositeunderplanestrainconditions,
weobtainmerelyy = 0.284 x 10m5.
To obtain estimates of the size of the region of stress oscillation for the epoxy-glass
composite, we plot the stress 9Z&(= z(O)) , i.e. the contact stress a^a against the crack
angle CC.Figure 5 shows the variation of the stress 9^9, at varying distances p1 ahead
0.6
FIG.5. Variation of the contact stress ahead of the crack tip f 1 = a exp (ior) with the crack angle a
for the epoxy-glass composite.
341
of the crack tip for the two cases of 4 = 0, -30. Now we note that from Fig. 2,
from a macroscopic standpoint (i.e. when we neglect the oscillatory nature of the displacement near the crack tips) the crack faces do not come into contact with each
other until a reaches 65 when 4 = 0. This means that the contact stress G0 ahead
of the crack tip must be tensile or compressive according as a c 65 or tl > 65,
respectively, indicating that %,, must vanish when a = 65. Hence, turning to Fig. 5,
we see that, for example, for the crack with a = 56 the stress oscillation starts at a
distance of p1 = 10P4a. It may be concluded that (5.2) is reliable for all CI(< 65) if p1
is larger than 10e3a. By a similar reasoning we see that the size of the region of the
stress oscillation for 4 = -30 is at most p1 = O-5 x lo-a.
The cleavage stress % at p1 = O-5 x 10m2a for 4 = 30, 0, - 30 was calculated
from (5.2) for the epoxy-glass composite under plane strain conditions (with regard
to the case of 4 = 30, the writer has checked that the size of the oscillatory region is
much smaller than 10V3a). It is plotted in Figs. 6(a), (b), (c). Also, the cleavage stress
for a homogeneous material, which is independent of the elastic constants, is shown in
Figs. 7(a), (b), (c).
Comparing Figs. 6 with Figs. 7, it may first be observed that a harder inclusion will
produce a higher stress concentration around the interfacial crack. But, apart from
the magnitude of the stress, both stress distribution curves corresponding to the same
value of 4 and their transformation with the change of c( resemble one another. For
example, in Fig. 6(b) and Fig. 7(b) the maximum value of % (= $,_.J increases with
increase in c(, but, as tl approaches the angle for which the two faces of the crack come
into contact, G maxdecreases corresponding to decrease in contact stress.
In the following we discuss mainly Fig. 6. We now take a different point of view
from that of Section 4, namely we assume that the criterion for de-bonding of the
interface is && = co (for small values of a) and that the criterion for crack deviation
into the matrix, or fracture, is sm,
De-
bonding occurs when, with the increase in the applied load, &!JOreaches co before
%,,,, reaches e1 and crack deviation occurs when 9^9, reaches cl before %, reaches
co. The magnitudes of crOand crl may be regarded as proportional to the ideal strength
of the interface (bonding strength) and of the matrix, respectively. When the strength
of the interface is greater than that of the matrix, the discussion is simple; in this case,
de-bonding of the interface never occurs, but fracture may occur for 9 = 9, for which
& = GmaX under the applied load which is lower than that given by (4.8). On the
other hand, if the bonding strength is much smaller than the strength of the matrix,
i.e. if rrOis much smaller than cl, de-bonding of the interface occurs under the applied
load given by (4.8), though Gm,, might occur at 9 # 0. But, since $&,, generally
becomes larger as c( increases, it is possible that de-bonding of the interface after all
results in fracture.
We note that Fig. 6(c) may be regarded as the stress distribution around the other
crack tip t, = a exp (- ioc)for the case 4 = 30. Comparing Figs. 6(a) and (c) we see
that the contact stress for t1 = a exp (LX)is greater than that for t, = a exp (-ia).
Hence, from the critical stress criterion, we derive the conclusion that de-bonding
occurs from the tip t I = a exp (ia), namely the crack starts to extend so as to take a
23
M. TOYA
EPOXY-GLASS
COMPOSITE
IS=303
J=o.5xlo-*)
_n
100
I 20
140
160
120
140
160
OfDEGREE)
0.6
-h
zp ( DEGREE)
lb)
FIG. 6.
180
343
EPOXY-GLASS
COMPOSITE
(9=-30,
p,=o.5xlo-2)
0.8
0.6
0.0
20
40
60
80
100
120
140
160
U(DEGREE)
(cl
FIG. 6. Variation of the cleavage stress with 9 for the epoxy-glass composite. (a) 4 = 300, pI = 0.5 x
10la; (b) Q = 0, p1 = 05 x 10-aa; (c) q = -3O, p1 = 0.5 x IO-as.
_
HOMOGENEOUS
(p-30)
0.6 -
- 0.1
I
20
I
40
-
I
60
I
80
1
100
@(DEGREE)
(a)
FIG. 7.
MATERIAL
I
Ix)
I
140
I
160
180
M. TOYA
344
C.0
20
40
60
------
80
iO0
29 i DEGREE
120
140
i60
140
160
(b)
a.+
I
0.5
c 0
Ii
3
20
40
60
---
FIQ.
7.
80
IOCP
i20
79P(DEGREEj
material.
345
symmetrical shape with respect to the tension axis, which is also derived in Section 4
from Griffiths energy criterion (4.8). On the other hand, the magnitude of 9^9,, for
the tip tz = a exp (- ia) is greater than that for the tip tl = a exp (ia) except for the
case a = 30. Hence, if crack deviation is to occur, then it must take place at the
tip t2 = a exp (--ior).
We assume that the bonding strength of the interface is large enough so that the
crack always grows in a direction 9 = 9, normal to the maximum cleavage stress. Of
interest is the angle J/ that the crack growth direction makes with the direction of the
applied stress, which is given by $ = +n + a - 4 - 9,. For example, from Fig. 6(a) we
obtain J/ = 82 for the crack with a = 30. Reading the value of 9, from Figs. 6(b)
and 7(b), the values of rj are obtained. To investigate the dependence of 4 on pr, the
values of I/I that are obtained from stress distribution curves for p1 = 10e3a and
IO-a are also plotted. The results are shown in Fig. 8. It is seen that the difference
loo-
EPOXY-GLASS
6d50
I
loa
I
30
I
x)
I
4o
I
50
I
60
I
700
-o(
FIG. 8.
between the values of J/ for a homogeneous material and for the epoxy-glass composite is at most about 10 and that as p1 increases this difference becomes even smaller.
Hence, we conclude that the direction of fracture is not so very sensitive to the elastic
properties of the inclusion and that, though we do not know yet which value of p1 is
appropriate for the determination of the crack growth direction, roughly speaking the
direction of fracture makes an angle that is somewhat lower than $I with the line of
action of the applied load. Similar conclusions apply for the case 4 = - 30.
ACKNOWLEDGMENT
The writer thanks Professor T. Tokuoka for his continuing interest and encouragement.
Thanks are also due to Mr. Y. Iwashimizu for helpful discussions on this paper and to
Mr. F. Tanabe for computing work.
REFEBENCES
CLEMJINTS,
D. L.
ENGLAND,A. H.
1971
1965
1966
M. TOYA
346
ERDOGAN,F.
GOTOH,H.
IRWIN, G. R.
KEER, L. M.
LOWENGRUB,M. and
SNEDDON,I. N.
MILNE-THOMSON,L. M.
MOSSAKOVSKY,
V. I. and
RYBKA, M. T.
MUSKHELISHVILI,
N. I.
1963
1965a
1965b
1967
1957
1967
1972
1960
1964
1953
lation by RADOK, J. R.
Groningen.
PERLMAN,A. B. and SIH, G. C.
RICE, J. R. and SIH, G. C.
TAMATE,0. and YAMADA,T.
TOYA, M.
WILLIAMS,M. L.
WILLIS, J. R.
1967
1965
1969
1973
19.59
1971
1972
M.)
Noon&off,
APPENDIX I
1)
We choose a path of integration as shown in Fig. AI.1 and then applying Cauchys
integral formula obtain
co + Cl/Z + czlz2
(AI.2)
I,+l,fl,
=
x(z)
where I,, I,, Ia denote, respectively, the integrals along the contours c, r, R, as shown
in Fig. AI.1. Expanding in power series of z, we obtain
co + Cl/Z + c2/z
x(z)
near lzl = co and
c,+c,/z+c,Iz2
x(z)
= coz+{c,-
c,a (cosa+21sina)}+O
lz [S gl(z)], (AI.3)
Z a exp [21(a - $1 x
_
--
1
c,+c,jz+c2/z2
1+v E
Y(Z)
- s1(z)-gz(z)
1*
(AU)
341
iY
t
APPENDIX II
~v~Iuff~~0~ of the integral in ~~~ti~n (4.5)
(AII.2)
on the path of integration.
(AII.3)
where c is a cIosed curve enclosing the arc aGcin clockwise direction, and then
12 = - y
l+v
then
I.
(AII.4)
(AII.5)
APPENDIX III
Derivation of equations (5.3) and (5.4)
In the vicinity of the crack tip tl = a exp (ia], we may readily obtain
W&r) 5 - 3@%i X(Z),
w;(z) % - 3pkaM&(z),
FV~(U2/Z)z -!&a%
@1x),
(AIII. 1)
(AIII.2)
(AIII.3)
(AIfi.4)
,
(Ai11.5)
(AIII.6)
Xl
P1
2a sin cc
(AIII.8)
~~bstit~~n~ from ~AIII.1) into (2.3) and making use of the ~dc~ti~ G-f= j$ S G i- 2,
we obtain equation (5.3).
On the other hand, combining
ry - x^x+ 2iX = zw;(z)+ Wl{Z)
(e.g. ~~~~~-T~u~o~,
(AIII.9)
j+-Gx;E+ixT
= (z-a21z)kY;~~)i(afz~)~~~~~2~~)~
W&).
~A~I.1~)