cu etal, Vol. 37, No 12.99. 3259-3254, 1989
rime in Grst Bria All hs xen
ON CRACK PATH SELECTION AND THE INTERFACE
FRACTURE ENERGY IN BIMATERIAL SYSTEMS
A. G. EVANS! B, J. DALGLEISH!, M. HE! and J. W, HUTCHINSON?
"Materials Department, College of Engineering, University of California, Santa Barbara, CA 93106 and
"pivision of Applied Selenees. Harvard Universi, Cambadge, MA 02138, USA,
(Received 10 May 1989)
Abstract—The intent of this article to apply resent solutions forthe mechanics of eracks at and near
‘binateril interfaces to rationale crack trajcetories observed by experiment and (0 provide basi Tor
interpcting measurements ofthe interface fracture energy. Wi Jemonstate thatthe choke of lest,
Speciman governs the tendency of eracks to cir remain at interfaces or deviate away. based on
considerations ofthe phase ane of loading, pT further eeveale that the measured nterTace fracture
“nergy may be sironelysnluenced by the erek teajctry. ay governed by J through eack shieding and
piiaticity eects Consoquenty, interfaces do not typically have unique facture energies, bu instead)
‘Sepends on yy which. in turn, Influenced by the test method.
Résumé Le but de ct article est ¢apliquer des solutions eécentes a (tude de mécanisme des fasures
sur et prés des interfaes entre deux materia afin de rtionliser les tajetores den fsures observes
txpérimentlement et de fourir une base a Fiterrstaion des mesures ce Fenerpe de rutured'ntetface
1 On démontre que le chow du type dessal gowverne le tendance dex fsuree soi Teter AUX
Inerfaces soit sn Garter suivant dophasage de charge gDe plus on monte que Téneraie de rupture
) prohibits propagation of the crack
‘out of the interfce, Then, one of two possibilities
occurs, depending upon the yield strength of the
ductile member. For a low yield strength material
plastic blunting of the interface crack is observed
[Fig 6(a)] and failure often occurs by ductile mech-
anisms (Fig, 6(6)] involving hole nucleation at the
Fig 3. Film decohosion by subatrate cracking indicating the
‘preted crack location tuned in Ky —0 and experimental
‘data ablained for {80 material vombinaionsEVANS erat
Relave Toughness,
Phase Angie of Loacng, v,
of imertace cracking
parameter Post
rows athe inset digzam,
Fig. 4. map represnting the
for the values of the casi
W is defined by the mode I
imerface [3]. Alternatively, the stress field of
the interace crack interacts with preexisting Aaws in
the prittle material, causing cracks to grow from
these flaws buck towaed the interface, resulting in a
serrated fracture with “chips” of britle material
attached to the fracture surface (Figs 1 and 5). This
‘behavior can be understood by noting that the flaws
in the brittle material are subject to subvtentiat mode
2 jouding having sign that leads to crack extension
toward the interface,
veo tk Tp ly
eee
Ce
tee yoo
7 ‘Substrate
24 tnvertace
bos rocking
Seas ale ae
Prase Angie o!Lootng
5. The region of interface fracture for a bimaterial
sm when one material F has a much higher facture
fnerty than the other, § (=F). The result potted for
the et, r= 8D. Now dha when ¥ <0, Fracture at the
Interface can inconporate segments of adjoining material
"detached at Raw.
INTERFACE PRACTURE IN BIMATERIAL SYSTEMS
mst
3. THE INTERFACE FRACTURE ENERGY
The above discussion eoncerning crack paths has
important implications for measuring and interpret-
ing information relating to the “fracture energy of
interfaces". Most importantly, the test configuration
‘determines the sign end the magnitude of ¥, which in
turn governs the crack path und thus, the mechanisms
that contribute to F,. Several examples are used 10
illustrate the salient features: “mode 1" sandwich
specimens [4,13] (Fig. 7a), peel tests [5] [Fig. 1b)]
and film decohesion tests [14] [Fig. (|. In all cases,
residual stress contributes substantially to the ob-
served hehavior and in the interpretation of the
results.
JL. Sandwich tests
‘The interpretation of fracture tests performed on
sandwich specimens is. relatively. straightforward
‘when the precrack length, a, is very large compared
with the bond layer thickness, h(a 2 30h). Then, the
residual strain does not contribute to the energy
release rate G, because the bond layer remains at-
tached to one side of the specimen and thus, retains
its state of residual strain [15]. When this erack length
‘condition is satisfied, the only other significant con-
sideration is the influence of the bond layer on the
‘magnitude of G and y. When h is small compared
withthe specimen thickness, the energy release rate
‘s unaffected by the layer [13]. However, the presence
of the layer causes y to deviate from zero, For the
typical case, p = 0, y is governed solely by 2, with the
‘rend {13] plotted on Fig. 8. While the range in Wis
small, it is often sufficient to influence the crack
trajectory. Specifically, in most cass, the shear mod=
ulus of the bond layer isles than that of the honded
‘member, whereupon the sign of directs the erack
toward the interface. Consequently, when the bond
‘material is ductile, the erack tends to either remain at
the interface or Follow a serrated path near the
Interface. In either case, plasticity in the bond layer
‘can contribute to F,. causing the measured fractare
‘energy 10 be substantially greater than I, even though
the crack remains near the interface
‘When the eruck is not long compared with the layer
thickness, residual stress influences both Kj and Ky
tand must be taken into account [15]. Most signtie
cantly, when the bond layer has the larger thermal
expansion coellcient, the sign of y tends to divert the
crack away from the interface and furthermore, is
sufficiently large (= 45°) {15] that it can dominate
the overall phase angle at the fracture criticality. In
this case, cracks are often diverted into the brittle
‘material, outside the bond [2]. The nominal fracture
energy measured in such cases can havea broad range
ff values that depend on che erack fength and the
residual strain [16]. However. the actual fracture
energy can be deduced if the residual strain contri-
bution to G is properly taken into account.3282 EVANS er al
twats
|_aluminum |
INTERFACE FRACTURE IN BIMATERIAL SYSTEMS
HOLE
NUCLEATION
at
INTERFACE
Fig. 6, (a) Plastic blunting atthe imerface: a erack in the ALO,/Al system. (b) Ductile fracture surface:
Tailre adjacent to the imerfae in the ALO,/Al system,
Bond layers having a lower thermal expansion
coeflcient invariably generaie a phase angle that die
verts cracks toward the interface. Then, the measured
fracture energy can either be larger or smaller than
TT, depending upon the residual strain contribution
0 G, the influence of plasticity, roughness, et
3.2. Film decohesion
Films and coatings with interface edge eracks that
cause decohesion experience mixed mode conditions.
with g dependent largely on the sign of the residual
stress [17], Films in tension give J 45° with sign
that tends 10 deflect the erack away from the inte:
face. Consequently, when the substrate is brittle and
the interface fracture energy is relatively high. the
decohesion process proceeds in the substrate by
cracking at a characteristic depth beneath the inter-
face [9-11], The substrate fracture energy , thenEVANS er al: INTERFACE PRACTURE IN BIMATERIAL SYSTEMS
Geopert on nan inet
wena
reey
Fig. 7. A sehematic indicating the various tots used 10
reise the interface Iracture energy
becomes the relevant quantity. When the substeate
is duotile, the incidence of decohesion is governed
by T, which may be influenced by piasticty in
Fig 8. The phase angle shift fora sandwich specimen as
su Tunction of the elasie mismatch aerons the interlace,
‘expressed through the Dundurs” parameter
the substrate, ete. When decohesion does occur at
the intezace (Fig. 9). the decohesion analysis [11]
may be used to measure I,
Residual compression in the film subjects edge
cracks 10 mode Il conditions, plus compression
normal to the interfyee. As’ yet, there is no
understanding of interface fracture in such cizcum-post EVANS or ul:
33. Peel tests
‘The peel testis generally used for measuring the
fracture energy of ductile compliant films on sub-
strates [S]. The phase angle has a negative sign that
‘causes substrate cracks to deviate into the interface.
‘The test thus encourages the modes of interface
fracture depicted on the left side of Fig. 6. The
measured fracture energy can be appreciably less than
the substrate value F,and include contributions from
crack tip plasticity in the film material, One eompli-
cation with this testis that the peel load is sensitive
to non-linear deformation in the film, such that
deconvolution to obtain I) requires knowledge of the
plastic flow stress of the film [5]. However. when this
effect is taken into account, valid F, measurements
are certainly possible and indeed, can be large
compared with I,
4. CONCLUDING REMARKS
A rational mechanics basis for predicting and
interpreting crack trajectories in bimaterial systems
has been proposed, based on knowledge of the phase
angle of loading, Y. In particular. the siga and
‘magnitude of y dictate whether cracks either propa
gate along an interface or deviate into the adjoining
‘material and extend parallel to the interface. More
importantly, phase angle conditions that motivate
‘racks to remain at the interface can, in some cases,
vield interface Fracture energies Pin excess ofthat for
the brittle member even though the crack propagates
‘atthe interface. Such high fracture energies contain
contributions ffom plasticity. roughness-induced
crack shielding, ee
The phase angle of loading is strongly influenced
by the choice of test specimen and thus, Fexperimen-
tally determined on the same bimaterial interface can
‘vary appreciably between test specimens. However, it
is belicved that a unigue relationship exists between
1, and y for a given interface, It is thus emphasized
that test specimens must be characterized oth in
terms of an energy release rate and a phase angle of
loading. AS more data on Fy(p) are generated, the
INTERFACE FRACTURE IN BIMATERIAL SYSTEMS
mechanisms that contribute to 1 can be addressed,
tnd allow the development of a fundamental evalu:
ation of interface fracture.
‘An important corollary of the above conclusion is
that the prediction of such events as fiber debonding
in composites, film decohesion and bond fracture
require that both G and y be evaluated for the
problem. Then, provided that I has been measured
in the appropriate range of , prediction can be
made, This point is emphasized because is typically
much more difficult to calculate than G, and its
evaluation is essential if imerface cracking problems
are to be adequately understood. Furthermore,
schemes for calculating y using cither integral
equations or finite elements have been developed.
REFERENCES,
|. B.A Dalgleish, M.C. Luand A. G. Evans, deta metal,
36, 2029:1988),
2K Burges, D. Brenner and G. Petzow, 2 Zahn
Implant. My $47 (1987)
3, BJ. Dalgleish, K.P. Trumble and A. G, Evans, Acta
mall. 37, 1923 (1989)
4. T5.Oh, J. Rodel, RO. Ritchie and R. M. Cannon,
deta metal 36, 2083 (1985),
SKS Kim and N. Aravas, It J Solid Struct. 24,417
oss),
6. RF. Pabst and G. Elsner, J. Mater. Sch 15, 188
(1K,
7. M, He and J. W. Hutchinson, J. appl. Mech. $6, 270
(1989)
1K. M.D. Thouless, A. G. Evans, MF. Ashby and J. W.
Hluchinson, deta metall 35, 1333 (1989)
9. M.S. HU M.D. Thoules and A, G, Evans, Aca metal
6, 130, (1988),
0. M.D. Drory and A. G, Evans, J. Am. Coram. Soc
Submited
1, Z. Suo and J. W. Hutchinson, Jt J. Solids Strut. to
ress.
12. 1. R. Rice, J. appl, Mech $5, 98 (1988).
13. Z,Suo and J. W. Hutchinson, Mater. Sci Engng A1OT,
135 (1989).
14, M.S. Hu and A. G, Evans, Ata metall 37, 917 (1989),
1S HG Cao, M.'D. Thouless and’ AG. Evans, acta
metal. 36,2037 (1988).
16, H. P Kirchner J.C, Conway and A. FE, Seg
Ceram. Soe. 0, 108 (1987)
19. Z.'Suo and J. W. Hutehinson, Jn. J. Fact bn pres
dt Am.