97
© 1992 Kluwer Academic Publishers. Printed in the Netherlands.
Abstract. The problem stated in the title is investigated with special emphasis on the first three
terms of the stress expansion, proportional to ri/2, t o = 1 and r 1/2 respectively, where r denotes
the distance to the crack front. The particular case of a plane crack with a straight front and of
stresses independent of the distance along the latter is studied first. It is shown that the classical
plane strain and antiplane solutions must be supplemented by a few additional particular
solutions to obtain the full stress expansion. The general case is then considered. The stress
expansion is studied by writing the field equations (equilibrium, strain compatibility and
boundary conditions) in a system of suitable curvilinear coordinates. It is shown that the number
of independent constants in the stress expansion is the same as in the particular case considered
previously but that the curvatures of the crack and its front and the nonuniformity of the stresses
along the latter induce the appearance of corrective terms in this expansion,
I. Introduction
The stress expansion near the tip of a crack is well known in the twodimen
sional case. The fundamental reference on this subject is the famous work of
Williams [1], which considered only the case of a straight crack and was
recently completed by Ting [2] who studied the general case of an arbitrarily
shaped crack. (In fact this author considered a still more general problem,
namely that of a corner or a notch with curved boundaries.)
Much less is known about the threedimensional case. One essential result
(see e.g. Bui [3]) is that the most singular term of the stress expansion can be
simply obtained by superposing the plane strain (modes I and II) and
antiplane (mode III) solutions. However higher order terms of the expansion
are involved in such questions as the path followed by a propagating crack or
the stability of such a path (see e.g. Leblond [4] and Cotterell and Rice [5]),
so that information about these terms is also needed.
~) Now with the Laboratoire de Mod~lisation en M~canique, Universit~ Paris VI, Tour 66,
4 place Jussieu, 75005 Paris, France.
98 J.B. Leblond and O. Torlai
The aim of this paper is to find the form of the stress expansion in the most
general threedimensional case: arbitrarily shaped crack, arbitrarily curved
front. Attention will be essentially focused on the first three terms of the
expansion, proportional to r 1/2, r0= 1 and rl/2 respectively (where r is the
distance to the front), but extending the analysis to all orders in r would raise
no problem of principle. The results derived will be used in a future paper to
study crack paths in three dimensions. They may also be useful to check the
validity or assess the accuracy of numerical solutions of threedimensional
crack problems.
We first consider the simpler problem of a plane crack with a straight front
and of stresses independent of the distance along the latter; in this case we
study all terms of the stress expansion. The main problem here is the
following: can all terms be simply obtained by superposing the plane strain
and antiplane solutions? As will be seen, the answer is "almost": other
solutions do exist, but there are very few of them and they do not intervene
in the most singular term.
We then introduce the curvatures of the crack and its front and the
variation of the stresses along the latter. In that case we only study the first
three terms of the stress expansion.
In a first step, the equilibrium equations, the Beltrami equations and the
boundary conditions are written in a system of "welladapted" curvilinear
coordinates; this is done by writing the equations in a system of "tangent"
Cartesian coordinates and transforming them into curvilinear coordinates.
From the equations derived, it can be seen that extra terms (with respect to
the particular case considered first) appear in the stress expansion; these terms
do not involve new, independent constants but the stress intensity factors,
their derivatives along the front and the curvatures of the crack and its front.
A second step consists in defining methods to solve the equations derived in
order to find the precise form of the new terms just mentioned. These
equations are essentially twodimensional so that techniques such as analytic
functions and Airy stress functions can be employed.
The methods defined are used in a third and final step to effectively solve
the equations and thus fully define the first three terms of the stress expansion.
II. The particular case of a plane crack with a straight front and of stresses
independent of the distance along the front
We consider a stressfree crack lying in the Oxl x3 plane, with a straight front
located along the Ox3 line (Fig. 1). The stresses are supposed to be indepen
dent of x3. The equilibrium and Beltrami equations and the boundary
conditions then split into two groups of equations involving a11, a~2, a22, 0"33
Stresses near the front of a crack 99
x2
Group 1:
0~,~ = 0; (1)
Group 2:
03~,~ = O; (4)
A03~ = 0; (5)
In these equations Greek indices take the values 1 and 2 and Latin ones the
values 1, 2 and 3. The summation convention is employed for both types
of indices. The symbol A denotes the twodimensional Laplacian, i.e.
A ~ ~2/~x21 q. 02/~x29.
The equilibrium equations (1) imply that the 0,a's can be expressed in terms
of an Airy stress function ql(xl, x2):
Putting then
fi~a g g ~ + f ~ a = 0; v A g ~ + (1 + v) Af = O. (9)
Summing eqns (91) over ~ and combining the result with eqn. (92), we get
AA~ = 0; Af = 0. (10)
Equations (91) reduce thus to f.~a = 0, so that the function f ( x l , x2) is of the
form
f =a+b~x~, (11)
• If ~ ~ 0 and ¢ 1, a and the b~'s are necessarily zero. The a~a's derive from
a biharmonic stress function by eqns. (7, 101), and a33 is connected to all and
a22 by the relation
which is a consequence of eqns (7, 8). Therefore the stress field corresponds to
a situation of plane strain, and the solution is the wellknown one given by
Williams [1], which depends on two arbitrary constants (the stress intensity
factors of modes I and II if y =  1 / 2 ) .
• If y = 0, the b~'s are necessarily zero but a is not. Independent solutions
can be obtained by taking • # 0, a = 0 on the one hand, • = 0, a # 0 on the
other hand. In the first case we obtain the classical Williams plane strain
solution, which depends on one arbitrary constant; this solution corresponds
in fact to a uniform all stress (and a uniform 0"33stress equal to Vail). In the
second case we get a solution which depends also on one arbitrary constant,
and corresponds to a uniform 0"33 stress.
• If ~ = 1, a is necessarily zero but the b~'s are not. For ~ # 0, bl = b2 = O,
we get the Williams plane strain solution which depends on two arbitrary
constants. For • = 0, bl and/or b2 # 0, we obtain a solution which depends
also on two arbitrary constants, and corresponds to a 0"33 stress linear with
respect to the coordinates.
Stresses near the front of a crack 101
Let us put
(h.2 + k,l).~ = 031,21  0"32,11 = 031,12 + 0"32,22 ~ (031,1 ] 032,2),2 = 0;
where c is a constant.
Now if the stresses are proportional to r 7, the derivatives of h and k are
proportional to r 7 i. Hence there are two cases:
II.3. Summary
Table 1 displays the number of independent solutions for all values of the
exponent 7. The solutions are classified as "plane strain", "antiplane" and
"particular" (i.e., corresponding to nonzero values of a, the b~'s or c). The
102 J.B. Leblond and O. Torlai
Table I. Number of independent solutions for each value of the exponent ~,.
total number of solutions is equal to 3 for all values of V except for V = 1 for
which it is equal to 6.
The stress expansion up to order r 1/2 thus takes the form
where the summation convention is employed for the index p = I, II, III. (The
upper or lower position is used for this index for typographic reasons only,
and has no particular mathematical meaning.) The constants KI, KII, Km are
the stress intensity factors. The Tp's are the "nonsingular stresses"; T~
represents a uniform tr~l stress (and a uniform a3~ stress equal to Vail: plane
strain solution), Tn a uniform o13 stress (antiplane solution) and Ti11 a
uniform 0"33 stress (particular solution). The expressions of the universal
• p, p'
functions f0' s, g;j~"s and hij s are recalled for completeness in the Appendix.
III. 1. Geometric description o f the crack  definition o f a local basis and some
curvilinear coordinates
ks,
\\
\\~...
~.~
"
~ ~
~ ~" X~
" ~'~"' ~
~x
z ~
.
1
" ~[ ~
E~
%N
"
~
E~ E~
the front which occur in the first three terms of the stress expansion). This is
done by specifying the components C~, C~3, C33 of the curvature tensor of the
crack at the point O and the curvature F of the projection of the front on the
tangent plane at the point O t2). The equation of the crack in the vicinity of the
point O is then
1
x2 = ~c~.x~x. + o[(x~ + x~)3/2]. (16)
where the summation convention is employed for the indices 2, # which take
the values 1 and 3. Furthermore the equation of the projection of the front on
the tangent plane is
F
x, = ~ x~ + o(x~).
F
x, = S x] + o(x]); x2= x~ + o(x]). (17)
Let us now define a local orthonormal basis ( e l , e2, e3) and some curvilin
ear coordinates Xl, x2, x3 "well fit" to express the stress expansion. For every
(2) It would seem more natural to use the curvature of the front itself (instead of its projection).
However the choice o f this parameter would lead to more complex calculations and final formulae.
104 J.B. Leblond and O, Torlai
point M located sufficiently near the front, there exists a unique plane ~
orthogonal to the front and containing M (see Fig. 2). Let H be the
intersection of ~ and the front, and (el, e2, e3) the analog of the basis
(E~, 22, E3) but taken at the point H instead of O. Xl and x2 are then defined
as the Cartesian coordinates of M in the plane ~ , the origin and basis chosen
being H and (ex, e2); we will also use polar coordinates r, 0 associated to xt
and x2, and the complex_variable z = x~ + ix~. Finally x3 is defined as the
curvilinear distance s = O H along the front.
Let us now derive the relations connecting the bases (21,22, 23) and
(e~, e2, e3). Let Y~, Y~, Y3 be the Cartesian coordinates of H. Y~ and Y2 are
connected to Y3 by eqns (17). Therefore
dOH
= FY3Et + C33 Y3E2 + E3 + O(Y~). (18)
dY~
It follows that
dx 3 ds dOH
= 1 + o(Y]) =~ x3 = s = Y~ + o ( Y ~ )
=~ Y3 = x3 + O(x~). (19)
Equations (18, 19) imply that the unit tangent vector e 3 to the front is
d O H / d Y3
e3  []dOH/d Y3 ]1 = Fx3E~ + C33x3E 2 + E 3 + O(x~). (20)
Furthermore eqn. (16) implies that the unit normal vector to the crack is
3OM/3X3 x 0OM/3XI
Ii
II3OM/OX3 x c ~ o n / o J ( III
The vector e2 is the value of n at the point H, i.e. by eqns (17) and (19)
Finally e I is given by
Let us now relate the two sets of coordinates X~, X2, X3 and x~, x2, x 3. x 3
is easily determined as a function of X1, X2, X3 by writing that the vectors
dOH/dY3 and I t M are orthogonal, using eqns (17) (for the Yi's) and
(18, 19):
dOH
. nM= r x ~ x , + C~x~X~ + x~  x~ + O(x], x~X,, x~XO = o
dY~
~ x~ = x~ + r x ~ x ~ + c . x ~ x 3 + o(l[OMll3). (24)
Furthermore
r
Xl = H M ' e ~ = X 1  ~ X ~
3 + c,~x~x~÷o(llOMll3); (25)
The stress components will be expressed in the local basis (el, e2, e3), as
functions of the curvilinear coordinates x~, x2, x3. It is therefore necessary to
write the field equations using this basis and coordinates. This will be done by
using first the basis (El, E2, E3) and the Cartesian coordinates Xl, X2, X3 and
transforming the equations into the local basis and curvilinear coordinates.
First of all one must relate the components Yij's and 0"o.'s of the stress
tensor in the bases (E~,E2, E3) and (el,e2, e3). One gets from eqns
(20, 22, 23):
To obtain the equilibrium and Beltrami equations, one must relate the
partial derivatives O/OXi's and 02/0X~ 0X/s to the O/Oxg's and 02/0x~ Oxj's. In
fact this needs to be done only in the plane x3 = 0 orthogonal to the front at
the point O, because the field equations will be written only in that plane (3~.
Using eqns (2426), one gets
0 0
(e = 1, 2); (33)
OX~  Ox~'
(0 0)
0 X 3 ~" 0X 3 "JI C13 x 2 Ox~  x l ~ x 2 + d.t.; (34)
02 02
(~,//= 1, 2); (35)
OX, OX~ Ox, Ox~ '
02 02 (02 02 )
OX1 Ox 3 OxI Ox~ "1 C13 x2 ~  xl OxI Ox2 O~Ox2 + d.t.; (37)
o2 o2 ( 02 o2 +__o)
"OX2 O X 3  OX2 OX~
3 "J[Cl] X2 OX 1 OX2 Xl Ox~ OxlJ+ d.t. (38)
for x 3 = 0. In these equations some terms, noted "d.t.", have been discarded
because they yield terms proportional to r 1/2 and r  1/2 (which will be seen to
be negligible) in the equilibrium and Beltrami equations.
Let us now suppose that the stress expansion is given by the same
expression (15) as in the particular case of Section II, the stress components
being those in the basis ( e l , e2, e3) and r, 0 denoting the polar coordinates
associated to the coordinates x~, x2 introduced above, with the only difference
that now the Kp's, Tp's and Bp's are functions of s. Let us consider the
equilibrium and Beltrami equations in the plane x3 = 0. The lefthand sides of
these equations can be obtained through eqns (2738):
0Y~lk dKp
~  ds F~3  C3~KpF~z + C,~K~(x2F~.,  x,F~.2  F~3)
(3) This is sufficient because the point O of the front is arbitrary, and the plane orthogonal to the
front at O sweeps the whole space when O varies.
Stresses near the front of a crack 107
~ 2)'~33 ~ 2~.~kk
(1 + v )   +
~x~ ,gX,,
OX~ = (1 + v)C33KpF'~,2  (1 + v)FGF~,,+ O(r1);
(43)
~2E13 ~2Ekk OG
(1 + v )   ÷   ds F~,l  (1 + v)Ca3gvF~,a
~x~ ~x~ ~x~ ~x3
3 2~'~23 (~2~kk dG ~
(1 + v )   +   ds F~k,2  (1 + v)CaaKpF~2,2
~Xk~Xk ~a'2~X3
the commas denote differentiations with respect to the x;'s (not the X,.'s); and
A is the Laplacian with respect to x~ and x2.
Let us now consider the boundary conditions in the plane x3 = 0. The polar
angle of a point M of the crack lying in that plane is
O = _ _ _ n  t a n 1 =___n ~ + O ( r : ) (47)
108 J.B. Leblond and O. Torlai
,X2
.. = c_~.o(x:)
2
Ion'1( ~_._~
,.,,~( ",? ~__~__ 0 x~1
(see Fig. 3). The normal vector to the crack is given by eqn. (21) with X3 = 0,
i.e. by eqn. (47):
Cllr
n =   ~   e r  eo + Cl3re3 + O(r2). (48)
It follows from eqns. (47, 48) that assuming again the stress expansion to be
given by eqn. (15), the lefthand sides of the boundary conditions are
CI1 r
(~.n)r(r,O)= C~r r l~Vtr + n ) + KvFP,o.o(r, ++_~)
2 vrr,, ~
CII r
(a" n)0(r, 0) = ~  KpF~o,o(r, +n) + O(r); (50)
Cllr K F¢ ~r,
(if" n)3(r, 0) = ~ p 3r\ " ~ ) +C~~KvF~oo(r,
Z ' ++_z)
It is apparent from eqns. (3945, 4951) that if the stress expansion is given
by eqn. (15), the field equations are not satisfied; the principal terms of the
lefthand sides of the equilibrium equations, the Beltrami equations and the
boundary conditions are proportional to r ~/2, r 3/2 and r 1/2 respectively,
times dK/ds, CK or F K (where K denotes a stress intensity factor and C a
component of the tensor C). To correct for this, extra terms proportional to
(dK/ds)r ~/2, CKr ~/2 and FKr ~/2 must be introduced in the stress expansion,
Stresses near thefront of a crack 109
I )
+ ~ 5 ( ° ) +~s ~5(°~ + G"g'm~"(O) + r~'n'5(O) r'/~+ O(r).
(52)
(Here again the upper or lower position of the indices has no special
mathematical meaning.) If Xij denotes one of the functions added in the
expansion, its introduction induces the appearance of the following terms in
the righthand sides of eqns (3945, 4951):
• X~.... X2~,~ and X3~,~ respectively in eqns (3941);
• ( 1 + v) AX, a + Xkg,,a, ( 1 + v) AX33, ( 1 + v) AXe3 and ( 1 + v) AXe3 respec
tively in eqns (4245);
• X~o(r, +~), Xoo(r, +~) and X3o(r, +~) respectively in eqns ( 4 9 
51).
It follows that the field equations are satisfied (i.e. the righthand sides of eqns
(3941), (4245), (4951) are equal to zero up to order r ~/~, r 3/2 and r 1/~
respectively) provided the functions
L~ =/~.(O)r'/2; Mp~U
.._ q = rn~u(O)r~/2; N~ = n~(O)r 1/2 (53)
Functions Lqp~ s:
Group 1:
Group 2:
Functions 'ij
M p11' S..
Group 1:
MP~
~t~,/~
= 0; (60)
(1 + v) A "~"~
a~'pll
ufl
±
q
M pll
kk,~tfl = 0"~ AM{~' = O; (61)
Mpll[. +n) r r
ro v,  =~(F~ro,o(r,+re)F~,r(r, +__n)); M ~ ( r , +n) =~F~o,o(r, +n).
(62)
Group 2:
Mpl~
3~t,~t = 0; (63)
AM{~ ~ = 0; (64)
r
M~o'(r, +~) = ~ (F{o,o(r, +~)  F~,(r, ~x)). (65)
Functions M~3's:
Group 1:
~#p13 1
~,~ ~,~ = ~ (  x ~ F f 3 , 1 + x~ F13,2
~ + F~3);
(66)
Mpl3 1
2u,~ = ~ ( x2F~3,1 + Xl F~3,2  F~3);
~.~3~.
2varO ~ t ~ +~)
~ ~ ~F~3(r, +z);
~ M~3(r, +_~) = O. (68)
Group 2:
1
(1 + v) AM~3~3 = ~ ( x2F~g,~l + XlF~k,12 d F~k,2);
(70)
( 1 + v) '~ i.¢p13
,~1,1 1 (  x 2F ~ , ~2 + x I F~,22  F ~ , l ) ;
23 = ~
r ~
M ~ 3 ( r , ++_~)= ~F33(r, ___~). (71)
Functions M~33's:
Group l:
l~,~ = F~2;
Mp33 ~
~1.¢p33
2~,~
= F~z  F~3; (72)
( 1 + v) ~,~A
i/~33~ + M~33k~,~= ( 1 + v)F~#,2; AM~333 = F~k,2; (73)
Group 2:
M~33
3~,~ = 2F~3; (75)
A M ~ 3 = F~z.2; (76)
M ~ z (r, ~ ~) = 0. (77)
Functions N~'s:
Group 1:
=  =
(78)
Group 2:
N%(r, ~ ) = O. (83)
112 J.B. Leblond and O. Torlai
The above equations have been divided into two groups for each tensorial
function, involving the components 11, 12, 22, 33 (or rr, tO, 00, 33) on the one
hand and 13, 23 (or r3, 03) on the other hand.
The various problems to be solved to find the functions Lij19' s, Mff"'s, Ni~P's
offer strong similarities. In fact they are mainly of two types; it is therefore
appropriate to define some general formalisms to solve these two kinds of
problems. (A third, more complex type of problem will also be encountered,
but this will happen only once so that developing a general method of solution
would not be useful in that case.)
This problem consists in finding some functions X,a's, X33 of xl and x2,
homogeneous of degree 1/2 and verifying the following equations:
and reasoning like in Section II. 1, one finds that • is biharmonic and f of the
form a + b,x~. But f m u s t be homogeneous of degree 1/2 like the X~'s, so it
must be zero.
The solution can thus be obtained by looking for a biharmonic function •
satisfying the boundary conditions, and calculating the unknown functions (in
Stresses near the front of a crack 113
1
Xrr = 
+_~I~ r 91V'~ Xoo : • rr "J[V;
r ~ dP'°° r " '
(88)
This problem consists in finding some functions Y3,'s of xl and xz, homoge
neous of degree 1/2 and satisfying the equations
Y3,,, = h ; (89)
AY3, = j , , (90)
where h and the j~'s are given functions of Xl and x2, homogeneous of degree
1/2 and  3 / 2 respectively and verifying the conditions
Ah = 0; j,,, = 0. (91)
f =h+ik. (92)
Let us put
F=;f=H+iK; (93)
then Y3~ = H/2, Y32 = K/2 is a particular solution of eqn. (89). Thus the
general solution of this equation is
H K
Y31~~~~,2; Y32=~Z,I (94)
114 J.B. Leblond and O. Torlai
Jl =g,2; J2 =  g a . (95)
(AX  g).~ = 0,
which implies that A:t  g is a constant. But A:t and g are homogeneous of
degree 1/2, since the Y3,'s are homogeneous of degree 1/2. Hence A:t  g
must be zero.
The solution can thus be obtained in the following way: evaluate f and F
from eqns (92, 93); evaluate g from eqns (95); determine the function ~ by
solving the equation A:t = g (together with the boundary conditions) under
the form
9
:~"(0) + ~ 2(0) = ~(0), (96)
r 3/2 g(O)
~=~(0)~; g=~; (97)
finally calculate the unknown functions (in polar coordinates) by the formulae
H* 1 K*
Y3r ~+Z,0;r Y3o =  ~    L , ; H * + iK* = F* = Fe i° (98)
(which immediately result from eqns (94)). (Note that the function F* in eqns
(98) is not analytic.)
Before solving the field equations for the various functions L,~'s, M,~.~'s, N v~''s,
it is appropriate to make the following remarks.
Firstly, if one considers two solutions of the field equations, homogeneous
of degree 1/2 (as required by eqns (52)), the difference of these two solutions
verifies the corresponding homogeneous equations and is still homogeneous of
Stresses near the front of a crack 115
where the Ap'S some constants and the functions hijp~ s are those which
a r e
appear in the third term of the stress expansion (15) for a plane crack with a
straight front. Hence the solution of the field equations is not unique; however
any two solutions, when inserted into eqns (52), yield the same form of the
stress expansion. Therefore it is sufficient to find one solution. The simplest
one will of course be selected.
Secondly, for any of the problems to be solved, either Group 1 or Group 2
of equations will have a trivial (zero) solution, because the functions which
appear in the righthand sides of the equations (the FP~p's and F~3, or the
F~'s, depending on the group of equations considered) will be zero. Only the
other, nontrivial group of equations will be considered in the sequel. (The
functions which are not given below can be taken as equal to zero.)
Mode I
Equations (5759) (Group 2) form a problem of type II; indeed the fact that
F~3 and FJ,k are harmonic (which results from the 33 component of Beltrami's
equations for the Fij~' s) implies that conditions (91) are verified. Using eqn.
(A14) of the Appendix, one gets
2v 0 2v
h x~cos~; f=x//Z~r~;
F~,k 2 0
   C O S 
l+v x/~ 2
(see eqns (A1)), i.e. ( 2 / 2 x ~ ) s i n 0/2; the general solution of eqn. (96) is
therefore
0 30 30
~ = sin ~ + A sin ~ + B cos ~,
116 J.B. Leblond and O. Torlai
where A and B are some constants. The boundary condition (59) can be
satisfied by taking A = 1  4v/3, B = 0. The solution is then
(99)
L~o:(2v~)X/~(sinO+sin3~).
Mode H
Using eqns (A2), it is easy to see that the problem is again of type II, the
functions h and g being the same as for mode I provided 0 is replaced by
0 + re. Therefore the expressions of F* and of the general solution of eqn. (96)
are also the same provided this replacement is made. Furthermore the
boundary condition (59) can be satisfied here simply by taking A = B  0.
The solution is then
Mode III
The function F~n  :~m is analytic; let us put
tr32
Then
 Fit I =  P 1; l~'IIl
•  J:32 ~" Q,I =  P,2.
V =  P = Re f (F~'I~'~'m'
*..Jr 3 2 ] ~   Re f( F~  ;L'm' ei°
I..v 3 0 ]
= Re 2i =  sm (102)
Stresses near the front o f a crack 117
(
• = r ~/~ A s i n ~ + cos~+Csin +Dcos
where ,4, B, C, D are some constants. The boundary conditions (56) can be
satisfied by taking .4 = 8/15x/~, B = C = D = 0. The corresponding solution
is
lrr
.'"= 45 ~ f ~ s i n O~ , •
 tI"=O; "l~rO
1,00
"' ~  ~/~ 0
COS ~,
•
(103)
r~"~3 3
/4+ 0
sin 2
~ "
Mode I
All righthand sides in eqns (6065) are zero, as can be checked from eqns
~ , ~ I 11 ~
(A1). Therefore all the u v can be taken as equal to zero•
Mode H
Equations (6062) (Group 1) form a problem of type I, with V = 0. The
boundary conditions (62) read
2r 5/2 0
• =~ cos ~.
Reference was made in the Introduction to Ting's study [2] of the stress
expansion near the tip of a curved crack, in the twodimensional case. Then
the dKp/ds's are necessarily zero, and so are also C13, C33, F. Therefore the
only functions which can be calculated in that case are the M,~l's. The results
/~rll I ~S and MijII11 ~_
derived by Ting for the .._;j ~ are identical to those obtained
here. (The comparison cannot be made for mode III because Ting studied
only the plane strain case.)
Mode III
Equations (6365) (Group 2) form a problem of type II, with h = 0, j, = 0.
Hence the function Z is harmonic; the boundary condition (65) reads
r 3/2 30
~ =  2  ~ sin ~.
..,,,,,
1,,3r = ~ cos~ . . . . 3o = sin~. (105)
Mode I
Equations (6971) (Group 2) form a problem of type II. Indeed
1
2 ( x2flkk'~l + xlF~kk'12 + F L , 2 ) = ~
1 (XlFIkk'l + x2F~,,2),2;
1 ~ ~ ~ l

2 ( X2Fkk'12 d XlFkk,22  F ~ , ~ ) = ~ (xlFI,~,l + x2F~k,2),~
S t r e s s e s near the f r o n t o f a c r a c k 119
since F~kk is also harmonic, which implies that condition (912) is satisfied too.
Using eqns (A1), we get
1 1 FI V 0 iv •
F =  iv ; F * =  iv
1 I 1 l+v 0
g = ~ (XlFkk,~ + x2F~k~.2) =~rF~kk.,  2~ COS~ •
~ 3O
= 1 + Vcos02+A sin + B cos~,
where A and B are some constants. The righthand side in the boundary
condition (71) is zero, so that the latter can be fulfilled by taking .4 = B = 0.
The corresponding solution is
Mode H
The problem is again of type II. Like for the functions Lon' s, the functions h
and g, and therefore the function F* and the general solution of eqn. (96), can
be simply deduced from those for mode I by replacing 0 by 0 + n. The
boundary condition (71) reads
l+v 0 1   3 v . 30
;~ = ~ sin ~ + ~ sin ~  .
120 J.B. Leblond and O. Torlai
0
3r  8 COS~+3COS ;
(107)
MII,3
30 = ~ ( V ~ 3  s i n ~ + ~09V3
sin 3~02).
Mode III
Equation (101) implies that F~I= Q,2; wul
32 =  Q , ~ , where Q is harmonic.
Therefore
This implies that the problem defined by eqns. (6668) (Group 1) is of type
I. Q is the harmonic conjugate function of P, i.e. 2x/r/2n cos 0/2 (see eqns
(101,102)). Therefore
1 1 1 ~r 0
V =  ~ (x,Q,, + x2Q.2) =  ~ rQ,r = ~ ~ / ~ cos~.
2r ~/2 0
• =  COS.
2
~ 0 ~ 0 8diI113 1~ 0
Mlrlri13= 2 COS ~ ; MIlll3
oo = 3
_ •
cos 2 . . . . rO
~
~
_o
sin 2'
(108)
MIIII3 0
33 ~~(4v  1) ~ COS   .
2
Stresses near the front o f a crack 121
Functions /~rI.~133
~t./ 'S
The function ~23K'IIIis harmonic. Furthermore the fighthand sides of eqns (76)
can be written as
i l l.
F 31.2~ ~,III ]71II
t 32,2 = ~t 31,1
for • = 1 and 2 respectively. This implies that eqns. (7577) (Group 2) form
a problem of type II. Using eqns (A3), we get
h ~''"~'IIl
a'~t 23 % /2~ c°s02", f = X /2~ ' " F = 4 ~ ; F* = 4 [ " ei°/2;
~/ 2~
1
/TIII __ _ _   s i n 0.
g = ~3  x/~ 2
1. 0 30 30
:~ =  ~ s ~ n  ~ + A sin~+ B cos~ ;
Mlli33
3r = cos~ + ~ cos ; 30 =  ~  ~ ( 0
sin ~ + sin ~ ) .
(109)
Functions  . ij~Vlll' s
Equations (8183) (Group 2) form a problem of type II, the functions h and
g, and therefore the function F* and the general solution of eqn. (96), being
the same as for the functions iyMII133"~provided 0 is replaced by 0 + n. The
boundary condition (83) can be simply met by taking A = B = 0. The solution
is then
o
~,3r =  ~ sin 2 ' ~,30 =  cos~. (110)
The systems defined by eqns (7274) and (7880) are of a more complex, yet
unencountered type. (They are not of type I because the "body forces" which
122 J.B. Leblond and O. Torlai
appear in eqns (72) and (78) are not conservative). We will solve them
simultaneously by putting
Z i j = MP.3~
v 3 +iN~.. (111)
It follows that
~f df~z Of df~z=~,;
Ox~ dzOx~  f ' ; Ox~= d~Ox~
~ ~
(117)
Of_ Of_~; Of_ 0~=_¢=,
~x~ dx~ Ox2 dx~
and inserting this expression and eqns. (119) into eqns (113), we obtain
Summing eqns (1211) over ~ and combining the result with eqn. (1212), we
obtain
This relation and eqn. (122~) imply that eqns (121~) read
( f + i~ + ice' + i¢ + 2z°),,~ = 0,
which means that the function (. • ) is of the f o ~ a + b,x,. But this function
is homogeneous of degree 1/2, since the Z~'s are homogeneous of this degree;
hence it must be zero. Using eqn. (118~), we conclude that
For each mode, the solution can therefore be obtained by the following
procedure: evaluate tO and $ from eqns (114); evaluate the Z°,a's from eqns
( 118); calculate f from eqn. ( 124); determine ¢ by solving eqn. (123) (together
with the boundary conditions) under the form
~ 25
iS"'(O) + iS"O)+ ~ iS(O) = 8~"(0), (125)
finally get the solution (in polar coordinates) through the formulae
Mode I
Using eqns (AI), we get
_ 1 0 9" 1 . ~
Ret#" 2~cos~; 2x/~ , tp= ;
1
F122 FI, + 2iFl~ = (F~oo  FI~ + 2iF~o) e 2,o = 2 ~x/2nr ( e io/: _ e  ~,o/:);
~k'= 1 . 1~
4x/~' ~=~ ;
Z°=i~f~n((12v) ~
ei°/21e~°/2  '~
1 e 3io/2 ) ;
ReZ°,=~f~nI(~2v)sinO+(~v)sin3~O2 +(1v)sin5~02];
=  cos + cos ;
furthermore
Ref = 4v  sin 0
2 + ~ sin .
The real part of ~ verifies the same differential equation (125) as ill itself, the
righthand side being Re 8itS"=  2 sin 30/2. The boundary conditions (74)
can be met by taking
~ /4v
Re • = ~  
11\ • 0 1
~~) sm ~ + ~ sin
3_~02+ 4 (lv) sin~
(This choice is not the simplest one for Re/1~, but it leads to the simplest result
for the .ijMI33's,"indeed it allows for the elimination of all sines and cosines of
the argument 50/2 in the expression of these functions.) The corresponding
126 J.B. Leblond and O. Torlai
_~ir,~~[(~~)
= 0(~ ) ~1  sin ~ +  v sin ;
(128)
~,~ ~[(~ ~) 0 (~) ~
rO =  COS~+ V COS ;
,~_~[(~ ~ ~).o ,
g~ 
~ +?g sm~+ ~ (l+v) sin~ .
ImZ~+ImZ~o=2ImZ°=~((14~)cos~cos~);
Im Z~o  I m Z ~ + 2i Im Z ~ = 2i e ~° Im Z ~
Taking the imaginary part of eqns (127), we get then the solution:
N,, = ~ /  ~ [ ( l V
i~) 0 ( ~ 6 )V COS~ ] ;
COS~+
NIoo = ( v  ]~
9)~f~n[ 3 cos ~
0 + cos 
3~1 ;
(129)
o
N~3= 8vZv~ cos~+~(l + v) cos .
Mode H
Here making the substitution 0 ~ 0 + ~r is of no help in the search for a
solution. Following the same steps as for mode I, we obtain
tp=i ; ~b=~ ,
Z ~ = 2i(v  l) ~ ei°/~;
~ ,
2
128 J.B. Leblond and O. Torlai
~o~,o~~[(~)cos~(~~)oos~ 1,~os~l~
ReZ~0= ~[(~ ) cos~+0(~)
r
~0
v  cos~+(1v)
+2v cos ~] ;
The righthand side in eqn. (125) for Re~ is 2cos 30/2. The bounda~
conditions (74) are met by the choice
~,~,~L[(,~
r
~)~o~(~_~)co~;
_13 0 9
~ ~L[(~~)~o~(~~)~o~];
M~
(~)L( sin ~0+ sin ~~) ;
33 = v 
(~30)
~,,~_~[(
3~  8v :  5 v + ~) 0,
cos~+~(l+v) cos 7] .
Considering now ima~na~ parts, we get
ImZ~o=~[(~2v) ~,~(~_~)~i~,l_~,
" 0 sinai ~
Stresses near the front of a crack 129
Imf=~2~I(4v~)sin~~sin~l"
The righthand side in eqn. (125) for Im~i is 2sin30/2. The boundary
conditions (80) are met by taking
4v 31) 0 1 . 30 4 50
Imi~= ~~ s i n ~  ~ s l n  ~  + ]  ~ ( v  1) sin~.
NI)r = ~ /~~rFf,07
/ ~ 1_\.~~ 7~) sin _02+ (v _ ~6) sin ~ ] ;
N~ =  v sin ~ + sin
9 ;
(131)
Nlrlo.~~_ r(31
L \ ~  ~ / cv,~
° s ~ + 0 (~6)
v ~os _.~0_];
9 s i n ~  ~ (1l ÷ v )
N~: [_[,~÷~~ sin
7] .
Only the first three terms of the stress expansion for an arbitrarily shaped
crack were considered in this paper. In principle, the next terms could also be
studied by the same technique. (However it is clear, from the amount of
algebra involved in the determination of the term proportional to r 1/2, that in
practice the calculations would soon become untractable). It would then be
necessary to describe the geometry of the crack and its front with a greater
degree of accuracy, by introducing higher order curvature parameters. The
corrective terms which would appear in the stress expansion would be
proportional to the stress intensity factors, the nonsingular stresses ... or
some of their successive derivatives with respect to s, times a product of
curvature parameters (possibly degenerate, i.e. equal to unity).
130 J.B. Leblond and O. Torlai
Only the nonzero components are given. For conciseness of the formulae, the
f and h functions are expressed in polar coordinates and the g functions in
Cartesian coordinates.
Functions f~'s:
1 ( 0 3 0 , 1 ( 0 ~)
fJ,~(O) = 4 x / ~ 5cos ~  cos ~),• fIoo(O) = ~ 3 cos i + cos ;
(A1)
fio(O)= ~ 1
(sin~
o + sin ; f~3(0) = =c°s~/2rt 2"
o
fI,~(0)
1(o
4v/~ cos~+3cos ; f~(0)=~sin~.
1 0 1 0
fr3(0)
~ n =l sin ~; f~I(0) = ~ cos ~. (A3)
Functions go s.
g~3(O) = 1. (A5)
g[t3x(O) = 1. (A6)
Functions h pe) s:
n 1 ( 0 ~)
h~(O) = 4x/~ 3 sin ~ + 5 sin ; hlo~o(O) 4v/~ sin ~  sin ;
(A8)
n 1 ( 0 ~_) 2v 0
h ~0(0)  4 x / ~  cos ~ + 5 cos ; h~313(0) ~ sin ~.
1 30 1 30
IIl
h~3 (0) =
~ sin ~, hlor3~(O)= ~
.
cos ~. (A9)
References
1. M.L. Williams, On the stress distribution at the base of a stationary crack. A S M E J. Appl.
Mech. 24 (1957) 109114.
2. T.C.T. Ting, Asymptotic solution near the apex of an elastic wedge with curved boundaries.
Q. Appl. Math. 42 (1985) 467476.
3. H.D. Bui, M~canique de la rupture fragile. Paris: Masson (1978).
4. J.B. Leblond, Crack paths in plane situations  I: General form of the expansion of the stress
intensity factors. Int. J. Solids Structures 25 (1989) 13111325.
5. B. Cotterell and J.R. Rice, Slightly curved or kinked cracks. Int. J. Fract. 16 (1980) 155169.
6. N.I. Muskhelishvili, Some Basic Problems in the Mathematical Theory of Elasticity. Groningen:
Noordhoff (1953).