Cracking and delamination of coatings

M. D. Thouless

Citation: Journal of Vacuum Science & Technology A 9, 2510 (1991); doi: 10.1116/1.577265
View online: https://doi.org/10.1116/1.577265
View Table of Contents: https://avs.scitation.org/toc/jva/9/4
Published by the American Vacuum Society

ARTICLES YOU MAY BE INTERESTED IN

Review Article: Stress in thin films and coatings: Current status, challenges, and prospects
Journal of Vacuum Science & Technology A 36, 020801 (2018); https://doi.org/10.1116/1.5011790

Thermomechanically induced voiding of Al–Cu thin films

Journal of Vacuum Science & Technology A 9, 2516 (1991); https://doi.org/10.1116/1.577266

Frost formation and ice adhesion on superhydrophobic surfaces

Applied Physics Letters 97, 234102 (2010); https://doi.org/10.1063/1.3524513

Microstructural evolution during film growth

Journal of Vacuum Science & Technology A 21, S117 (2003); https://doi.org/10.1116/1.1601610

Influence of apparatus geometry and deposition conditions on the structure and topography of thick sputtered
coatings
Journal of Vacuum Science and Technology 11, 666 (1974); https://doi.org/10.1116/1.1312732

Material selection for hard coatings

Journal of Vacuum Science & Technology A 4, 2661 (1986); https://doi.org/10.1116/1.573700
Cracking and delamination of coatings
M. D. Thouless
IBM Research Division. T. J. Watson Research Center. Yorktown Heights. New York 10598

(Received 24 September 1990; accepted 27 February 1991 )

The mechanism by which a coating will fail is governed by a number of parameters. These include
not only the properties of the coating, interface, and substrate, but also the distribution of stresses.
If the coating is under a residual tensile stress, there are at least three possible mechanisms by
which failure can occur. For example, a brittle coating may fracture by the development of cracks
through the thickness of the film. Tougher coatings may fail by delamination along the interface
or even by the propagation of a crack within the substrate. The failure mechanism associated with
a compressive stress in the coating involves simultaneous buckling and delamination. This paper
reviews these different modes of failure and discusses the conditions that determine which ones
will operate.

I. INTRODUCTION als, the advantage of fracture mechanics is that, under spe-

A number of companion papers in this symposium discuss cial conditions, the fracture resistance can be taken as a ma-
the origin and characterization of residual stresses that can terial ~onstant which is not dependent on the geometry of the
develop in thin films adhering to a substrate. This paper pre- specimen or on the loading conditions. In thin-film speci-
sents a review of some mechanisms by which these stresses mens, however, two effects conspire against this approach.
may cause cracking and delamination of films. The initial The first is that if the film is both thin and tough, the size of
part of this paper considers the characteristics of different the nonlinear region at the crack tip may be similar to the
failure modes in isolation (Fig. I). Three of these failure film thickness. The "small-scale yielding" condition will
modes, delamination of the interface, splitting of the film, then be violated and the energy absorbed in the fracture pro-
and substrate cracking, occur as a means of relieving a resid- cess will be geometry-dependent. The other complication is
ual tension in the film. The fourth mechanism, which in- one that arises in the general problem of delamination of
volves buckling of the coating and subsequent delamination interfaces: the problem of what is termed "mixed-mode frac-
ture" .2--4
of the interface, relies on the existence of a compressive stress
within the coating. The final part of this paper compares In a homogeneous material, a crack will foHow a trajec-
some of the results for these different mechanisms so as to tory in which it is subject only to an opening mode of defor-
mation at its crack tip.5.6 This is known as mode-I fracture.
illustrate the conditions under which each mechanism may
be expected to dominate.
A crack, whether it develops in the film, interface or sub-
strate, is driven by the relaxation of residual stresses and the
consequent decrease in elastic energy stored in the system.
The driving force for crack propagation is designated by [1
and is given by! a)

[1= -au
aA ' (1)
where U is the stored elastic energy and A is the crack area. It
should be appreciated that despite the very high stresses of-
ten observed in thin films which may even exceed the yield
stress, the use of elastic solutions to determine the driving
force for fracture is appropriate. Fracture is driven by the
relaxation of the residual stresses, and unloading is always
elastic irrespective of the initial magnitude of the stresses.
This argument is not valid when discussing fracture result- c)
ing from an applied load, however, and the possibility of
gross yield must be considered.
[1 can be calculated from elasticity for any hypothesized
mode of failure. However, fracture will only occur if the
driving force exceeds the fracture resistance for the particu-
lar mechanism in question. This crack resistance consists not
only of the energy required to create newsurfaces, but also of FIG. I. Four general types of failure that may occur in coated systems: (a)
any irreversible losses, such as plasticity, associated with delamination of the interface; (b) mUltiple cracking of the film; (c) sub-
crack propagation. When considering homogeneous materi- strate cracking; and (d) film buckling and associated delamination.

2510 J. Vac. Sci. Technol. A 9 (4), JullAug 1991 0734-2101/91/042510-06$01.00 ® 1991 American Vacuum Society 2510
2511 M. D. Thouless: Cracking and delamination of coatings 2511

The amplitude of the square-root singularity in the elastic 0.8 .--------,.--,---.--0--,--,---,-...,--.,

stress field asymptotically close to the crack tip (which con-
Long Crack Asymptales
trols the fracture process) is denoted by K 1 • This quantity is
0.6
known as the mode-I stress-intensity factor and has units of
Shott Crack
Pay m. When the trajectory of a crack is determined by the -'): Asymptales I
presence of a weak plane, the crack tip will, in general, be t;,°0.4
subject to both in-plane and out-of-plane shear deforma- :;;;
tions. These are governed by the mode-II and mode-III
0.2
stress-intensity factors KII and K m , respectively. The stress-
intensity factors are related to the crack-driving force. For
example, in plane problems, 7-9 0.0 IL-_ _ _~_~_.L-....l--"--"---L--'--.J.....J
o 2 3 4 5 6 7 8 9 10
~ = {IIE/+ lIEJ{K; +K;I}/2, (2) 4/t
where E = E in plane stress, E= E 1(1 - v) in plane (a)
strain, v is Poisson's ratio, E is Young's modulus, and the
subscripts/and s refer to the film and substrate, respective-
ly Y The presence of a shear component in the crack-tip field
0.8 ,------,---,----y---,--,-r--,.-,.----r...,
appears to have a significant effect on the fracture resistance.
This effect is not completely understood, but it may result
from frictional interactions between the crack surfaces or 0.6
from changes in the size and shape of any nonlinear zone long Crack AsymptaIes
near the crack tip. Xi
Beyond the general observation that the presence of a
shear stress increases the fracture resistance, there is no rea- o;;"'~~o:
-
~
_ 0

son or experimental evidence to suppose that any general 0.2

relationship exists between the degree of mixity and the ap-
parent toughness of an interface. However, for comparing
the behavior of different failure mechanisms, it is useful to
2 3 4 5 6 7 8 9 10
assume a fracture criterion. For the purposes of this paper, a./t
therefore, it will be assumed that only K[ contributes to the (b)
fracture process. This criterion, in which it is assumed that
the energy available from the other two modes are complete-
F'G. 2. Stress-intensity factors for two plane problems of delamination: (a)
ly lost in frictional or plastic dissipation, is recognized to be
delamination from an edge Haw and (b) delamination from a crack in the
an over-simplification. However, since it incorporates the film.
increase in toughness with shear, it does allow some of the
interesting features associated with interfacial fracture to be
K,/uo..Jt = 0.36(alt) 112,
illustrated.
KlI/uoft = 0.63(alt) 112 (4)
II. FAILURE MECHANISMS when alt.( 1. Conversely, if the delamination is initiated by a
A. Delamination split across the width of the coating [Fig. 2(b»), then the
The first mechanism to be considered in detail involves the stress-intensity factors decrease from an initial value of"
extension of a crack along the interface between a film and (5)
substrate [Fig. 1(a) ). General results for the asymptotic
to the values given in Eq. (3).
values of the stress-intensity factors that are pertinent when
the crack length a is much longer than the film thickness t The existence of a steady-state solution in these two cases
is a direct result of the assumption that the prOblem is a two-
have recently been obtained by Suo and Hutchinson. '0 As an
dimensional one and that there is no variation in the geome-
example of the mixed-mode nature of the delamination pro-
cess, the asymptotic values for K, and KII are given by6 try across the width of the samples. Figure 3 shows how f§
varies for an axisymmetric geometry in which a crack is initi-
K,/uo..Jt = 0.43, ated from either an external [Fig. 3(a») or an internal [Fig.
3 (b) ) boundary of the film. '2 The lack of a steady-state re-
(3)
gime in which ~ is independent of the crack length is very
when the moduli of the film and substrate are identical and evident from these plots. Similarly, if a flaw initiates from a
there is a uniform tension U o in the coating. How these finite split in the film, delamination is limited to the region
steady-state values are approached as the crack length in- around the split and scales with its length. 13
creases depends on the type of flaw that initiates delamina-
tion. If the crack grows in from the edge of the sample [Fig. B. Film splitting
2 (a) ), then the stress-intensity factors increase with crack A tensile stress in a brittle film may be relieved by the
length, having values 6 formation of a periodic array of parallel cracks growing

J. Vac. Sci. Technol. A, Vol. 9, No.4, Jul/Aug 1991

2512 M. D. Thouless: Cracking and delamination of coatings 2512

I
I
\ I
\
\ 100:
20\
, \
I

, ,
I
I
\ I

" \ I
".k:
/,,\ i
R/t t ,\
" \
OL-__________ ~ ____________ ~

o 0.5 1.0
Normalized Crack Length (1 -aiR)

>- _ 0.5 FIG. 4. Schematic illustration of the geometry for a film which is splitting
~
~ ~
W -
I W
NO
b
N"":.
..
0.4
,-. under the influence of residual tension.

c:
.~
I
i spacing and that to obtain such a spacing would require a
US
suitable density of initial flaws. In practice, it is expected that
~
'iii
CI)
II)
the cracking pattern may be dominated by the statistical dis-
E as
CI) tribution of defects. 14 The other reason is that the crack
o
z
Qi
a: spacing is determined by the critical shear stress at the inter-
oL-~~~~~=r==~~d face l8 •19 when the film is deposited on a ductile substrate, or
o 1 2 3 if film cracking is associated with delamination at the inter-
Normalized Crack Length (~ - 1) face (Sec. II A). It can be shown that under these condi-
tions, the crack spacing is not dependent on the fracture
toughness of the film and that, in contrast to the elastic anal-
FIG. 3. Strain·energy release rate for two axisymmetric problems of dela-
ysis, A increases with t: 18
mination; (a) delamination from ah edge flaw and (b) delamination from a
central cut in the film. The dashed lines in both figures illustrate asymptotic A = 2aot IT. (7)
behavior.
In this equation T is the critical shear stress at which slip
occurs and it may be controlled either by the yield stress of
the substrate or by friction at the delaminated interface. Al-
across the film (Fig. 4).14 The value of K, which acts to split though the spacing is independent of Kf in this shear-lag
the film depends on I, the spacing between the cracks. When analysis, the critical stress at which splitting can occur still
the moduli of the film and substrate are identical and the depends on it: 18
cracks are significantly longer than the the thickness of the
film, K, is given byls
10
Kdao[t "'" 1.41 (lIt) ~ 8, 9
K,/ao[t "",[O.51It-O.0316(11t)2]1I2 (//t)~8. (6) 8
These equations can be used to predict the crack spacing A '-
-<
7
expected in a film with a fracture toughness of Kf.IS.16 The rio 6
c:
results are shown in Fig. 5. There is a critical stress, 'u
0 5
a.
Vl
a* = O.71Kf l[t, below which the film .will not split. Once ~
4
v
splitting does occur, the analysis suggests that the crack e
CJ
3
spacing should be no larger than about eight times the film 2
thickness; cracks further apart than this have a negligible 1
interaction with one another. 0
Although these characteristics have been observed in at 0.0 0.5 1.0 1.5 2.0
least one study,'7 most observations suggest that the crack Normalized Film Stress, vt/K,
l1 o

spacing is much larger than the range predicted in Fig. 5. 14 •18

There are two possible reasons for this discrepancy. The first FIG. 5. A plot showing how the crack spacing is predicted to vary as a
is that this figure shows only the thermodynamic minimum function of the normalized film stress.

J. Vee. ScL Technol. A, Vol. 9, No. 4, JullAug 1991

2513 M. D. Thouless: Cracking and delamination of coatings 2513

25r-----------------------------~
a crack will propagate. 6 .2{}-24 Experiments demonstrate that
this trajectory has a remarkable stability which can be ex-
plained because any perturbation o(the crack from this path
results in a shear stress that acts to drive it back again.
Two effects of changing the moduli of the film an<,t su~­
1,5 strate can be observed from Fig. 6. The first is that as EflEs
0 decreases, the stable trajectory approaches the interface and
ti
"'
u.. the resulting spall is therefore thinner. The other concerns
~ 1,0 the magnitude of the driving force, which appears to increase
'iii
c:
C) with the compliance of the film. However, it should be re-
E D,S
membered that if the stress is a result of some residual strain,
III
III
~ 10
such as might arise from a lattice mismatch or from a differ-
Vi ence in thermal expansion, U o scales with the modulus of the
0L-----~----~----~----_7----~ film, and the dependence of KI on the film compliance will
o , 2 3 4 5
Distance Below Interface. zit be reversed.
(a)
Suo and Hutchinson have presented general results that
',0 rr---------------------, allow the effect of any stress state, elastic constants and sub-
strate thickness to be calculated. 25 One of the important re-
sults is the surprising sensitivity of the crack path to the
thickness of the substrate. 23 The effects of an axisymmetric
geometry have not been calculated in such detail, but some
preliminary finite-element studies show the presence of a
KII = 0 trajectory below the interface. 20

D.Buckling
The mechanisms by which a compressive stress in a coat-
ing may be relieved are much more limited. From an ener-
getic argument, it is conceivable that delamination along the
-, ,0 '----_ _...J-_ _ ~ _ _ _.J.__ ___'__ _ --:! interface could occur in a similar fashion to that described in
o 2 3 4 5 Sec. II A. However, this would involve a negative value of
(b) Distance Below Interface. zit KI at the crack tip. The only mechanism of delamination in
which there is a positive value of KI involves buckling of the
FIG. 6. Plots showing the stress-intensity factors as a function of the depth of film at an initial defect in the interface [Fig. l(d) J.14.26.27
a semi-infinite crack beneath a residually stressed coating: (a) mode I and The compressive stress required to buckle a portion of the
(b) mode II. coating over an initial strip of delaminated interface of width
2a is given by 14.27

"
Uc = - rrE/t la) 2112. (9)
u* = (3rK}lt) 1/3. (8)
Finally, it should be noted that when there is a biaxial As the magnitude of the compressive stress, - U o , is in-
stress in the film, the periodic array of parallel cracks will be creased beyond this value, the driving force for delamination
increases; 14.27
replaced by a pattern commonly known as "mud cracking."
The film fractures into a series of islands similar to that seen Ef f1 Ia;t = 0.5(1'. - 1)(1'. + 3), (10)
in areas of drying mud. Although the details of the mechan- where I. = uolac • I. increases as the crack begins to propa-
ics are slightly different, the general trends outlined in this gate along the interface and, consequently, a rich diversity in
section are expected to be similar. behavior is possible.1 4 Delamination may be an unstable
process with f1 increasing without limit as the crack length
C. Substrate cracking increases. It may be stable, in which case an increase in f1
The third location at which a crack may propagate to re- requires an increase in the magnitude of uo . An intermediate
lieve the strain energy stored by a film under tension is with- range of behavior is also possible in which an initial instabil-
in the substrate [Fig. 1(e) J. As in Sec. II A, there is a ity is followed by stable crack growth. The details of the
steady-state regime in which the stress-intensity factors are delamination process are controlled by the initial size of the
independent of the crack length. In this regime, KI and KII defect at the interface and by the value of the interfacial
depend on the elastic properties of the film and substrate and toughness.
on the depth of the crack beneath the interface. Values ob- It has recently been shown that when delamination occurs
tained from an approximate analytical expression for KI and by this mechanism it does so under mixed-mode conditions.
KII are shown in Fig. 6 for the case of a uniform tension in As the crack grows, the conditions at its tip change from one
the film. 2o•32 Of particular note is the existence of a depth at in which KI is slightly higher than KII to one in which the
which KII = 0; this is the depth of the trajectory along which mode-I component becomes zero and eventually negative. 27

J. Vac. Sci. Technol. A, Vol. 9, No.4, Jul/Aug 1991

2514 M. D. Thouleas: Cracking and delamination of coatings 2514

1.0 .-~~-r-~,.....--r~-r-~r---r"""""-r-.,.......,~--,

0.9
0.8 no delamination possible 2.5
N

~ 0.7 fonnaUon 0'

~ 2.0 craddng IUbatrot.
~ 0.6
~ of crock
<to:!
....0.5 ~ 1.5
g 0.4
II 0.3 1.0
><
0.2 delamination possible ----------~~~~
0.5
0.1
"'Pieto delamination
0.0 L..-o---J.................l..-..........- - ' - - - ' - - ' - -.........---'............- ' - -..........- ' - - - ' -............. 0.0 L..I....L..L..I....L..I....L..L..I....L..L..L..J....L..L..J....L..L..LJ...L..J....L..Ll..J....L...L..J..J
o 2 3 4 5 6 7 8 9 10 0.0 0.5 1.0 1.5 2.0 2.5 3.0
1: = (12uolnZE,)(ajt)2 (8) K.!uoJt
FIG. 7. Map showing the region in which delamination under a uniaxial
compression is possible by the mechanism of buckling. The dashed line
represents the condition for buckling only; buckling occurs to the right of
this line.

These characteristics can have a substantial influence on the

fracture behavior. If a failure criterion of KI = K j is as-
sumed, the conditions under which delamination is possible
can be calculated and are represented graphically in Fig. 7.
This figure suggests that there is a maximum defect length int.rfadal doI"";natian
for which no further growth by this mechanism of combined 0.0 I....L-~.L..J....L...L-"-I.....LJ'-'-L-..I.-.I......I.....J-,J...-L...J
buckling and delamination is possible. Once the defect has 0.0 0.5 1.0 1.5 2.0
reached this length, further failure can occur only by plastic (b) K.lai';t
deformation or by cracking of the film in buckled region.
The existence of this maximum does not depend on the pre-
FIG. 8. Two fracture maps comparing the conditions for crack propagation
cise details of the mixed-mode failure criterion; it only de-
along the interface and substrate. (al The initial defect is assumed to be a
pends on a criterion in which the fracture resistance in- crack in the film. (b) The initial defect is assumed to be an area of delamina-
creases rapidly as the ratio KII / K, _ 0() .33 tion along the interface.
A biaxial stress state appears to produce an even richer
behavior. An initial analysis suggests that buckling and dela-
mination should occur by the formation of a circular blis-
ter. 26 In fact, failure under these conditions seems to be often may result from a split in the film. Delamination along the
associated with nonaxisymmetric and sinusoidal patterns of interface is one possibility. Alternatively, the split may ex-
buckling. 28 tend as a crack into the substrate. The conditions for dela-
mination have been addressed in Sec. II A, while the split
III. FAILURE MAPS can propagate into the substrate if11 •29
It is possible to combine the results of the previous section (11 )
and generate failure maps to illustrate which mechanisms Once the crack has entered the substrate, it is possible for it
are expected to operate under particular conditions. When to turn and propagate parallel to the interface as discussed in
fracture occurs in the film or substrate it will do so under Sec. II C. For this to occur, ao has to be large enough that6
pure mode-I conditions, and it is therefore appropriate to use
the mode-I fracture toughness, Kf or K s ' for these materials. (12)
A failure mechanism that involves propagation of a crack· Figure 8(b) compares the conditions that will determine
along the interface will introduce the problem of mixed- whether a crack initially lying along the interface wiU con-
mode fracture. Ideally, the fracture resistance of the inter- tinue in that plane, or whether it will kink into the substrate.
face should be measured under the mixed-mode conditions The condition for this latter event to occur is ' I,3O
that are appropriate; however, this is not always possible and
(13)
some choice of failure criterion has to be made. For the pur-
poses of this discussion, the choice that K I = K j (the mode-I In both of these maps, a straight line drawn through the
fracture toughness of the interface) has been made. origin shows the effect of varying the stress in the film while
Two failure maps are shown in Fig. 8 for systems in which keeping the material properties constant. Two very impor-
the moduli of the film and substrate are identical and there is tant results should be noted. The first is that failure does not
a uniform tensile stress of a o in the film. Figure 8(a) com- necessarily occur along the weakest path, since the mechan-
pares the conditions for two possible modes of failure that ics of the problem will dictate different driving forces for

J. Vac. ScI. Technol. A, Vol. 9, No.4, JuliAug 1991

2515 M. D. Thouless: Cracking and delamination of coatings 2515

crack propagation at different locations. This difference may lB. R. Lawn and T. R. Wilshire, Fracture of Brittle Solids (Cambridge
University, Cambridge, 1975).
result in the fracture condition being met along a trajectory
2 A. G. Evans and J. W. Hutchinson, Acta Metall. 37, 909 (1989).
with a high fracture resistance before it is met along one with 3H. C. Caoand A. G. Evans, Mech. Mater. 7, 295 (1989).
a lower resistance. For example, it can be seen from Fig. 4M. D. Thouless, Acta Metall. Mater. 38, 1135 (1990).
8(b) that if 'F. Erdogan and G. C. Sih, J. Basic Eng. 85, 519 (1963).
6M. D. Thouless, A. G. Evans, M. F. Ashby, andJ. W. Hutchinson, Acta
Ks <2.1K;, (14)
Metall. 35, 1333 (1987).
a crack will propagate more readily in the substrate than 7J. R. Rice and G. C. Sih, J. Appl. Mech. 32, 418 (1965).
along the interface. In other words, the fracture toughness of 8J. R. Rice, J. Appl. Mech. 55, 98 (1988).
oZ. Suo and J. W. Hutchinson, Mater. Sci. Eng. A 107, 135 (1989).
the substrate can be twice that of the interface before dela-
toZ. Suo and J. W. Hutchinson, Int. J. Fract. 43, 1 (1990).
mination becomes an issue in the failure of thin films. A I1M. D. Thouless, H. C. Cao, and P. A. Mataga, J. Mater. Sci. 24, 1406
second observation to be made from these plots is that condi- (1989).
tions may be such that more than one mode of failure is 12 M. D. Thouless, Acta Metall. 36, 3131 (1988).
energetically favorable. Under these conditions, kinetic ef- 13 H. M. Jensen, 1. W. Hutchinson, and K. S. Kim, Int. J. Solids Struct. 26,
fects become important and failure will be dominated by the 1099 (1990). .
14G. Gille, in Current Topics in Materials Science, edited by E. Kaldis
mechanism that has the faster kinetics of crack growth. (North-Holland, Amsterdam, 1985), Vol. 12, Chap. 7.
I'M. D. Thouless, J. Am. Ceram. Soc. 73, 2144 (1990).
IV. SUMMARY 16M. D. Thouless, E. M. Olsson, A. P. Segmiiller, and A. Gupta (to be
A number of mechanisms by which a residually stressed published).
17E. M. Olsson, A. Gupta, M. D. Thouless, A. P. Segmiiller, and D. R.
coating may crack or delaminate have been reviewed in this
Clarke, Appl. Phys. Lett. 58, 1682 (1991).
paper. The failure of a coating is controlled by a number of laM. S. Hu and A. G. Evans, Acta Metall. 37,917 (1989).
parameters including the mechanical properties of the film, I"D. C. Agrawal and R. Raj, Acta Metall. 37,1265 (1989).
interface and substrate, the initial distributi()n of flaws with- 20M. D. Drory, M. D. Thouless, and A. G. Evans, Acta Metall. 36, 2019
in the system, and the magnitude and sign of the residual (1988).
21 M. S. Hu, M. D. Thouless, and A. G. Evans, Acta Metall. 36, 1301
stress in the film. The presence of a tensile stress can cause
(1988).
delamination along the interface, or cracking in the film or 22 R. M. Cannon, R. M. Fisher, and A. G. Evans, Mater. Res. Soc. Proc. 54,
substrate. Failure induced by a compressive stress involves a 799 (1986).
complex interaction between buckling of the film and dela- 23 Y. H. Chiao and D. R. Clarke, Acta Metall. 38, 251 (1990).
mination along the interface. 24M. D. Drory and A. G. Evans, J. Am. Ceram. Soc. 73, 634 (1990).
HZ. Suo and J. W. Hutchinson, Int. J. Solids Struct. 25, 1337 (1989).
Cracks that propagate in the film or substrate do so under
26 A. G. Evans and J. W. Hutchinson, Int. J. Solids Struct. 20, 455 (1984).
the influence of K. , and the fracture resistance is controlled 27 J. W. Hutchinson and Z. Suo in AdlJOnces in Applied Mechanics, edited by
by the mode-I fracture toughness of the appropriate materi- J. W. Hutchinson and T. Y. Wu (in press).
al. Conversely, delamination generally occurs under mixed- 28 J. W. Hutchinson, M. D. Thouless, and E. G. Liniger (to be published).
mode conditions in which both shear and normal stresses 2"H. Tada, P. C. Paris, and G. R. Irwin, The Stress Analysis of Cracks
playa role in the fracture process. Beyond the generalization Handbook (Del Research Corporation, St. Louis, 1973).
30M. Y. He and J. W. Hutchinson, J. Appl. Mech. 56,"270 (1989).
that they increase the fracture resistance, the effect of shear 31 Strictly, K J and KI\ can only be defined if a rather unique relationship
stresses on the failure ofan interface is not completely under- exists between the elastic constants on either side of the interface (Ref. 8).
stood. If a mixed-mode fracture criterion is assumed, it is In general, ;§ for an interface crack cannot be separated simply into
possible to compare different failure mechanisms and deter- normal and shear components. However, for a broad range of material
mine which ones will operate under particular conditions. It combinations the assumption implicit in Eq. (2) does not introduce large
errors (Ref. 10).
is important to appreciate that failure does not necessarily 32 If the stress in the film is not uniform, the depth at which KII = 0 is
occur along the weakest path. There may be conditions in changed in a way that is very easily calculated (Ref. 21).
which the mechanics dictate that the driving force be greater Hit is interesting to note that the existence of this maximum permits the
along a trajectory with a larger fracture resistance, and this possibility of delamination occurring upon reducing the compressive
may dominate the failure process. stress (Ref. 28).

J. Vac. Sci. Technol. A, Vol. 9, No.4, JullAug 1991