308 G. Wang, S.F. Li / Theoretical and Applied Fracture Mechanics 42 (2004) 303–316micro-scale yielding due to hydrostatic stress stateis small-scale yielding: a2 1. For a 6 r 6 b, a2 1.b 2 r 2The total stress distribution within <strong>cohesive</strong> zone(a 6 r 6 b) may be approximated asr ðtÞzzr ðtÞrr¼ r ð0Þzz¼ r ð0Þrrþ r ðcÞzz¼ r 0 ð25Þþ r ðcÞrr¼ 1 2v R m þ r 02r ðtÞhh ¼ rð0Þ hh þ rðcÞ hh ¼ 1r ðtÞrz¼ r ðtÞrh ¼ rðtÞ zh ¼ 02v R m þ 2v r 02ð26Þð27Þð28ÞIt is assumed that inside the <strong>cohesive</strong> zone microplastic yielding is controlled by the Huber–vonMises criterion. There<strong>for</strong>e, one can link the <strong>cohesive</strong>strength, r 0 , with the yield stress of the virgin<strong>material</strong>, r Y ,by12r ðtÞrrr ðtÞzz 2 2þ rðtÞ 2þ rðtÞhhr ðtÞzzrrr ðtÞhh¼ r 2 Yð29ÞSubstitute Eqs. (25)–(27) into (29) and solve <strong>for</strong> r 0 .The following quadratic equation may be obtained:4 r 2 02 r 202 r Yþ 1¼ 0R m R m 1 2v R mð30Þwhich has two roots. The positive root is chosen tolink the <strong>cohesive</strong> stress r 0 with the yield stress r Yof virgin <strong>material</strong> in uniaxial tension,1 þr 0¼R msffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 24 r Y31 2v R m4ð31Þ4. Effective elastic <strong>material</strong> properties of an RVEOne of the goals of homogenization is to find theeffective <strong>material</strong> properties of an RVE to characterizethe constitutive relation of stress and strainrelation at the macro-level. Conjugate to the macrostress R, the macro strain E is defined through theoverall complementary energy density W c of an RVEE ¼ oW coRð32ÞDefine the effective compliance D of the micro<strong>crack</strong>edRVE throughE ¼ D : R ¼ D : r 1ð33Þwhere the macro stress R = r 1 , according to averagetheorem, is regarded as prescribed. It is notedthat the macro strain of an RVE may not be thevolume average strain, i.e, E 5 hei. FurthermoreEq. (33) may not be a linear relationship, becauseD may depend on R in general, which will be substantializedin later sections.A common strategy <strong>for</strong> homogenization of randomlydistributed defects is to divide the macrostrain E into two parts,E ¼ e ð0Þ þ e ðaÞð34Þwhere e (0) = D:R is known, and D is the elasticcompliance of the corresponding virgin <strong>material</strong>.The second term e (a) is so-called additional straintensor representing the effect of defects. If the relationshipbetween additional strain and macrostress can be found, say e (a) = H:R, where H isthe added compliance due to micro-<strong>crack</strong>s, subsequentlythe effective elastic compliance moduli,D ¼ D þ H, can be deduced.Energy method is applied to find an additionalstrain <strong>for</strong>mula <strong>for</strong> <strong>cohesive</strong> <strong>crack</strong>s. The essenceof energy methods is to find the energy release ina <strong>cohesive</strong> fracture process and hence to find theequivalent reduction of <strong>material</strong> properties. Forelastic <strong>crack</strong>s, the energy release rates were discussedin the classical works of Rice and hisco-worker [3,25,26]. Nevertheless, the energy dissipationprocess of <strong>cohesive</strong> fracture is much morecomplicated than a purely elastic fracture process.It includes energy dissipation from both surfaceseparation and plastic dissipation. A related discussioncan be also found in [18,22,31].Although accurate determination of energy lossduring a <strong>damage</strong> process requires an in-depthunderstanding of the physical process involved,an upper bound estimate may be made based onsimplified assumptions. It is assumed that the totalenergy release of a <strong>cohesive</strong> <strong>crack</strong> is completelyconsumed in surface separation, which may ormay not be true in <strong>cohesive</strong> fracture, because ofplastic dissipation in the <strong>cohesive</strong> zone.
G. Wang, S.F. Li / Theoretical and Applied Fracture Mechanics 42 (2004) 303–316 309Subjected to uni<strong>for</strong>m triaxial loading r 1 =R m I (2) , the total energy release rate of an RVE witha single <strong>penny</strong>-<strong>shaped</strong> <strong>cohesive</strong> micro-<strong>crack</strong> can beestimated asZZR ¼ R m ½u z ŠdS r 0 ½u z ŠdSð35ÞXX 2Carrying out the integration using <strong>crack</strong> displacementsolutions Eqs. (17), the energy release estimatecan be written as following expression:R ¼ 16ð1 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiv Þr 23l 0 a3 a1 1 ðR m =r 0 Þ 2 ð36ÞConsider that there are N <strong>penny</strong>-<strong>shaped</strong> <strong>crack</strong>s insidethe RVE. The density of energy release of theRVE is estimated as sum of each <strong>crack</strong> contribution,i.e. R ¼ P Na¼1 R a. Define the <strong>crack</strong> openingvolume fraction asf ¼ XN 4pa 3 a3V bð37Þa¼1where a a is the radius of the ath <strong>crack</strong>, and 4pa 3 a =3is the volume of a sphere with radius a a , and b isthe ratio between the volume of permanent <strong>crack</strong>opening and the volume of total <strong>crack</strong> openingof a <strong>cohesive</strong> <strong>crack</strong>. For simplicity, it is assumedthat this ratio is fixed <strong>for</strong> every <strong>crack</strong> inside anRVE. Obviously, 0 < b