Theoretical and Applied Fracture Mechanics 43 (2005) 342359 www.elsevier.

com/locate/tafmec

Micromechanical consideration of tensile crack behavior based on virtual internal bond in contrast to cohesive stress
Z.N. Zhang *, X.R. Ge
School of Naval Architecture, Ocean and Civil Engineering, Shanghai Jiaotong University, Shanghai, 200240, PR China Available online 23 May 2005

Abstract A modied version of the virtual internal bond model (VIB) is presented. This involves the introduction of a R-bond restricting the relative rotation freedom of pairwise mass particle. Such a modication allows the VIB model to consider arbitrary values of the Poisson ratio. A linear elastic cohesive law considering both the R-bond and L-bond are assumed. The constitutive relationship is derived using the CauchyBorn rules. The derived constitutive associates the bond stiness with the Youngs modulus and Poisson ratio of materials. This gives the bond stiness in terms of the Youngs modulus and Poisson ratio of materials. The modied VIB model is then used to analyze the tensile crack behavior. In contrast to the cohesive stress method, the deformation-governed concept will be used. The local materials failure is assumed to coincide with the reduction of the bond density due to the local deformation rather than by the local cohesive stress. A phenomenological relationship between the bond density and the deformation is established. The criterion which is applied to determined crack initiation and propagation is built into the constitutive model. As an example, the method is used to study the crack initiation and propagation behavior under tensile loading. 2005 Elsevier Ltd. All rights reserved.
Keywords: Micromechanical model; Virtual internal bond; Tensile crack behaviors; Isotropic solid

1. Introduction To predict the crack behavior, many approaches have been proposed. The VIB model [1] was used to study the crack growth behavior. The method, however, is limited to a specic value of the Poissons ratio. This limitation will be overcome in this work by considering both the R- and L-bond in addition to using the CauchyBorn rule [25] that connects the continuum approach to that of the cohesive bond. Path
*

Corresponding author. Tel.: +86 21 54740526; fax: +86 21 62932194. E-mail addresses: zzn@sjtu.org, zhennanzhang@sina.com.cn (Z.N. Zhang).

0167-8442/$ - see front matter 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.tafmec.2005.03.005

Z.N. Zhang, X.R. Ge / Theoretical and Applied Fracture Mechanics 43 (2005) 342359

343

independent integral of the type in [6] has also been used. Cohesive surface methods [711] have also been advanced to represent the crack initiation and propagation process. It is dicult to associate the cohesive methods with the physics of cracking, particularly when dierent size scales are involved. The works in [12 15] have discussed multiscaling near the crack tip by using the concept of the strain energy density function. In addition, discussions were also made with reference to the use of continuum concept and that of the particle. Molecular dynamics (MD) has also been used to analyze fracture behavior at the atomistic level [16 21]. The conventional treatment of crack behavior based on fracture mechanics can be found in [22]. The approach can be regarded as a simplied version of the multiscale treatment in [1215]. It has recently been pointed out [23] that the 1/r-type of singularity for the potential function of clusters of molecules is an intrinsic property that is independent of the material constitutive relationship. This holds when the dominant lattice interaction at the crack tip is of the mutual attraction type. These fundamental researches improved the understandings of fracture behavior of materials and provided the theoretical basis for solving the practical crack problems. The development of numerical methods have involved many dierent approaches. Among them is the VIB model [1] that has been used to study crack growth in isotropic solids. The method was extended to the anisotropic materials [24]. The nite element implementation of VIB was made in [25]. The method integrates the discrete character at the microscopic level to the mechanical properties at the continuum level. By incorporating a cohesive law which governs the interactions between two discrete mass particles, the fracture criterion is directly built into the constitutive relationship. No external criterion is required when using VIB to simulate the crack growth. However, the Poissons ratio in [1] must be xed at 0.25. A R-bond is thus added to restrict the relative rotation freedom. The modied VIB model is then used to examine crack growth by fracture. The condition corresponds to a reduction of the bond density cause by the local deformation rather than the cohesive stress used in the earlier models.

2. Constitutive model for linear elastic materials 2.1. Physical consideration of material behavior Atoms within the material are known to interact and their en-mass eects can be incorporated into a continuum approach. It is important to study the macroscopic response of materials in relation to the microstructure. In the VIB model, the solid is considered to consist of numerous discrete mass particles at the microscopic scale. These mass particles are connected with a virtual internal bond, which is embedded with assumed cohesive law. These bonds only restrict the radial displacement of a two-coupled particles as shown as Fig. 1. For a two-coupled particles, besides the radial displacement, there is also relative rotation freedom as shown in Fig. 2. To restrict the rotation freedom, a R-bond is introduced into the VIB

Fig. 1. Virtual internal bond model.

344

Z.N. Zhang, X.R. Ge / Theoretical and Applied Fracture Mechanics 43 (2005) 342359

Fig. 2. Decomposition of bond deformation (b: rotation angle; I: radial displacement; I0: the original length of L-bond).

Fig. 3. The L-bond and the R-bond.

model, Fig. 3. Here, the radial bond is referred as the L-bond. Each L-bond is accompanied with three Rbonds, which restrict the L-bond rotation towards the three coordinate axes respectively. 2.2. Constitutive relationship The interaction laws between the particles presumably would determine the macroscopic properties of the material. The interaction in the VIB model is nonlinear. However, in what follows, the interaction is assumed to be linear in that both the L-bond and R-bond are linearly elastic. The energy potential stored in a single L-bond can be expressed as: UL 1 k 2 2 where k is the L-bond stiness coecient; is the stretch of the L-bond. The energy potential stored in the R-bond is: 8 2 1 > < U R1 2 rb1 U R2 1 r b2 2 2 > : 1 U R3 2 rb2 3 1

where r is the R-bond stiness coecient; URi(i = 1,2,3) is the energy potential of the R-bond corresponding to the coordinate-xi such that bi is the rotation angles of L-bond towards coordinate- xi. Hence, the energy potential within a two-coupled particles i j could be expressed as: Uij U L U R1 U R2 U R3 3

Z.N. Zhang, X.R. Ge / Theoretical and Applied Fracture Mechanics 43 (2005) 342359

345

According to Eqs. (1)(3), the pairwise particle potential is associated with the bond stiness and the particles position which must be determined. The position of the atoms may be related to that of the continuum eld through the local deformation gradient. This is the so-called CauchyBorn rule [25] as shown in Fig. 4 [5]. The theoretical basis of quasi-continuum method is to make a continuum volume element mechanically equivalent to a discrete microstructure. The equivalence principle is the strain energy density equivalence. In an arbitrary representative element with volume V, the total strain energy stored in L-bond could be expressed as: Z L2 Z 2p Z p UL U L DL 0 ; h; / sinh dh d/ d0 4
L1 0 0

where DL(0,h,/) is the spatial distribution density of 0-length bond; DL 0 ; h; / sinhdh d/ means the number of 0-length bonds per unit volume between the bond angle (h,h + dh) and (/,/ + d/) in spherical coordinate; L1, L2 is the minimum and maximum original length of L-bond respectively. The total strain energy stored in R-bond can thus be expressed as: Z L2 Z 2p Z p Z L2 Z 2p Z p UR U R1 DR1 0 ; h; / sinhdh d/ d 0 U R2 DR2 0 ; h; / sinhdh d/ d0
L1

0 L2 L1

Z
0

0 2p

Z
0

L1 p

U R3 DR3 0 ; h; / sinhdh d/ d0

where DRi(0,h,/)(i = 1,2,3) is the distribution function of R-bond related to the L-bond with original length 0. URi is the corresponding R-bond energy. The total strain energy stored in bond is: U UL UR The L-bond stretch in a certain direction n is: 0 ni eij nj 7 6

where n = (sin h cos /, sin h sin /, cos h) is the unit orientation vector of L-bond in the sphere coordinate system, shown as Fig. 5; eij is the strain tensor; 0 is the original length of L-bond. The rotation angles of L-bond with orientation n towards three coordinate axes are given by 8 0 > < b1 ni eij gj b2 ni eij g00 8 j > : b n e g000
3 i ij j

Fig. 4. The CauchyBorn rule: an undistorted lattice with vectors A1 is mapped to the deformed conguration by application of the local deformation gradient F (from [5]).

346

Z.N. Zhang, X.R. Ge / Theoretical and Applied Fracture Mechanics 43 (2005) 342359

Fig. 5. Indication of the L-bond orientation in sphere coordinate system.

where g is the unit vector perpendicular to n, g0 n x 1 n; g00 n x 2 n; g000 n x 3 n; x i is the unit orientation vector of coordinate-xi (for details, refer to Appendix A). The concrete expression of g is: 8 0 2 2 2 2 > < g sin hsin / cos h sin h cos / sin /; sin h cos / cos h g00 sin2 h cos / sin /; cos2 h sin2 hcos2 /; sin h sin / cos h > : 000 g sin h cos / cos h; sin h sin / cos h; sin2 h Using Eqs. (1)(8) and equating the strain energy function at the continuum level to the potential energy stored in the cohesive bonds due to an imposed deformation, the stress tensor can be expressed as the relations used in the hyper-elastic theory [26,27]: Z Z Z oU=V 1 L2 2p p 2 rij k 0 nm emn nn ni nj DL 0 ; h; / sinh dh d/ d0 oeij V L1 0 0 Z Z Z 1 L2 2p p rb1 ni g0j DR1 0 ; h; / sinhdh d/ d0 V L1 0 0 Z Z Z 1 L2 2p p rb1 ni g00 j DR2 0 ; h; / sinhdh d/ d0 V L1 0 0 Z Z Z 1 L2 2p p rb1 ni g000 9 j DR3 0 ; h; / sinhdh d/ d0 V L1 0 0 where V is the representative element volume. The elastic modulus can be expressed as: Z Z Z o2 U=V 1 L2 2p p 2 C ijmn k 0 ni nj nm nn DL 0 ; h; / sinhdh d/ d0 oeij oemn V L1 0 0 Z Z Z 1 L2 2p p rni g0j nm g0n DR1 0 ; h; / sinhdh d/ d0 V L1 0 0 Z Z Z 1 L1 2p p 00 rni g00 j nm gn DR2 0 ; h; / sinhdh d/ d0 V L2 0 0 Z Z Z 1 L2 2p p 000 rni g000 j xim gn DR3 0 ; h; / sinhdh d/ d0 V L1 0 0

10

Z.N. Zhang, X.R. Ge / Theoretical and Applied Fracture Mechanics 43 (2005) 342359

347

For the simplication purpose, let the original length of L-bond distribute uniformly with original length 0. Therefore, Eqs. (9), (10), becomes the following form: Z Z Z Z 1 2p p 2 1 2p p rij k 0 nm emn nn ni nj DL h; / sinhdh d/ rb1 ni g0j DR1 h; / sinhdh d/ V 0 V 0 0 0 Z Z Z Z 1 2p p 1 2p p 00 rb1 ni gj DR2 h; / sinhdh d/ rb1 ni g000 11 j DR3 h; / sinhdh d/ V 0 V 0 0 0 C ijmn 1 V Z
0 2p

Z
0 2p

k 2 0 ni nj nm nn DL h; / sinhdh d/ Z
0 p

1 V

Z
0

1 V

Z
0

2p

Z
0

rni g0j nm g0n DR1 h; / sinhdh d/

p 000 rni g000 j nm gn DR3 h; / sinhdh d/ 0

00 rni g00 j nm gn DR2 h; / sinhdh d/

1 V

Z
0

2p

12 2.3. Relationship between the bond stiness and the material constants For the isotropic material, the bond density distributed uniformly in spatial. So, both the L-bond density and the density could be taken as constant and DL(h,/) = DR1(h,/) = DR2(h,/) = DR3 (h,/). ConR 2pR-bond Rp sider 0 0 Dh; / sin hdh d/ 1, such that the bond density can be written as: DL h; / DR1 h; / DR2 h; / DR3 h; / Using the FEM 2 C 1111 6 6 C 2211 6 6 C 3311 X6 6 C 1211 6 6 4 C 2311 C 1311 analysis, the elastic tensor Cijmn C 1122 C 2222 C 3322 C 1222 C 2322 C 1322 C 1133 C 2233 C 3333 C 1233 C 2333 C 1333
1 C 1112 2 1 C 2212 2 1 C 3312 2 1 C 1212 2 1 C 2312 2 1 C 1312 2

1 4p may take the matrix form as C 1123 C 2223 C 3323 C 1223 C 2323 C 1323
1 C 1113 2 1 C 2213 2 1 C 3313 2 1 C 1213 2 1 C 2313 2 1 C 1313 2

C 1121 C 2221 C 3321 C 1221 C 2321 C 1321

1 C 1132 2 1 C 2232 2 1 C 3332 2 1 C 1232 2 1 C 2332 2 1 C 1332 2

3 C 1131 7 C 2231 7 7 C 3331 7 7 C 1231 7 7 7 C 2331 5 C 1331

13

where X can be tted into the relationship: rXe

T T

14

with r = [r11,r22,r33,r12,r23,r13] and e = [e11,e22,e33,2e12,2e23,2e13] . Note that rij and eij are the components of stress and strain tensor, respectively. According to Eq. (13), the elastic matrix corresponding to Eq. (12) becomes 3 2 k 2 k 2 0 0 0 3k 2 0 2r 0 r 0 r 7 6 3k 2 k 2 0 0 0 7 6 0 2r 0 r 7 6 2 7 3 k 2 r 0 0 0 1 6 0 7 6 15 X 7 6 2 15V 6 k 0 1.5r 0 0 7 7 6 5 4 symmetry k 2 0 0 1.5r k 2 0 1.5r In Eq. (15), the relationship between k, r and the macro material constants, i.e. the Youngs modulus E and the Poisson ratio m could be obtained:

348

Z.N. Zhang, X.R. Ge / Theoretical and Applied Fracture Mechanics 43 (2005) 342359

Fig. 6. Virtual bond in two-dimensional cases.

8 > > > <k

3EV 1 2m2 0 > 3 EV 1 4m > > :r 1 m1 2m

16

The case of r = 0 is equivalent to the case of not considering rotation in the energy potential In [1], the rotation energy potential is not considered and hence the Poisson ratio of m = 0.25 was obtained. For the two dimensional cases, the bond construction is shown in Fig. 6. Eq. (12) reduces to the form: Z 1 p 2 00 1 C ijmn dh 17 k 0 ni nj nm nn rni g0j nm g0n rni g00 j nm gn V 0 p where n = [sinh, cosh]T; g 0 = [cosh, sinh]T; g00 = [cosh, sinh]T; and 1/p the bond density. The elastic matrix corresponding to Eq. (17) is 2 3 3k 2 k 2 0 0 2r 0 2r 1 6 2 7 X 18 0 4 k 0 2r 3k 2 5 0 2r 8V 2 0 0 k 0 2r For plane-strain, the relationship between k, r and E, m could be obtained by using Eq. (18): 8 2EV > > >k 2 < 0 1 2m1 m > EV 1 4m > > :r 1 2m1 m

19

3. Extension to the nonlinear materials 3.1. Micromechanism of material weakening due to the deformation The material weakening behavior is assumed to be deformation-governed in that the local material failure is considered as the result of bond density reduction induced by the local deformation, not induced by the local cohesive stress. For the usual material, its failure process usually undergoes the linear elastic

Z.N. Zhang, X.R. Ge / Theoretical and Applied Fracture Mechanics 43 (2005) 342359

349

phase, strengthening phase and the strain softening phase, shown as Fig. 7. In the linear elastic phase, the modulus keeps constant. However, in the strengthening and the strain-softening phase, the modulus decreases with increasing deformation. This is because some aws and dislocations initiate in the material as the deformation increases. Additional increase in deformation leads to the coalescence of these aws and dislocations. This is then followed by the initiation of micro-cracks as shown in Fig. 8. Nonlinear weakening behavior of the material nally takes place. Assumed in this work is that the reduction of material modulus could be considered as the result of the bond density reduction. In the linear elastic phase, both the L-bond and the R-bond distribution density are kept constant. As the deformation exceed the linear elastic limit, the bond density reduces due to the appearance of aws, and dislocations. This is shown in Fig. 9. As the deformation further increases, the bond density reduces more seriously and tends to zero. The bond density reduction with the deformation can be qualitatively shown in Fig. 10. To quantitatively describe the density reduction with the deformation increasing, the bond density can be phenomenologically expressed. With reference to the L-bond density, it can be shown that 8 1 > < if l0 6 ec 4 p DL h; / 20 > : 1 expkl0 e 2 =e2 if l0 > e c c c 4p where l 0 = jni eijnjj. The symbol j j means the absolute value and k is a coecient used to adjust the reduction ratio of bond density while ec is the critical value of strain, the determination of which will be described in Section 3.2. The R-bond density can be expressed as:

Fig. 7. Three phases of the whole process stressstrain relationship for the uniaxial tension test (rp: the propotional strength limit; ep: the proportional strain limit; rb: the tension strength of material; eb: the strain corresponding to rb).

Fig. 8. IIlustration of materials weakening with deformation increasing.

350

Z.N. Zhang, X.R. Ge / Theoretical and Applied Fracture Mechanics 43 (2005) 342359

Fig. 9. Bond density variation with the deformation increasing.

Fig. 10. Bond density varying tendency with the deformation increasing (D0: the initial bond density).

8 1 > < if b0i 6 bc 4 p DRi bi > : 1 expkb0 b 2 =b2 if b0 > b c c i c i 4p

21

where b0i is the absolute value of the L-bond rotation angle towards the coordinate-xi; i.e. b0i jbi j (Eq. (8)); bc is the critical value of L-bond rotation angle. Refer to Section 3.2. For the two-dimensional cases, the Lbond density could be expressed as: 8 1 > < if l0 6 ec p DL h; / 22 > : 1 expkl0 e 2 =e2 if l0 > e c c c p

Z.N. Zhang, X.R. Ge / Theoretical and Applied Fracture Mechanics 43 (2005) 342359

351

and the R-bond density can be expressed as: 8 1 > < if b0i 6 bc p DRi bi > : 1 expkb0 b 2 =b2 if b0 > b c c c i p

23

where l 0 = jnieijnjj; n = (sinh, cosh) in the plane coordinate, shown as Fig. 6; b01 jni eij g0j j; b02 jni eij g00 j j; g0 cos h; sin h; g00 cos h; sin h. 3.2. Parameter determination In Eqs. (20)(23), there are totally three parameters, namely ec, bc and k , as discussed in Section 3.1. When the deformation is within the linear elastic limit, i.e. e 6 ep, both the L-bond and the R-bond density are kept constant. As the deformation exceeds the linear elastic limit ep, i.e. e > ep, shown in Fig. 7, the bond density begins to reduce. Therefore, the ec should approximately take the value of ep. At the onset of the linear elastic limit, shown in Fig. 11, the strain is: e22 = ep, e11 = me22 = mep,  ni eij g0 ep sin h cos h e12 = e21 = 0. The rotation angle of L-bond with orientation n = (sinh, cosh) is b j

Fig. 11. IIlustration of uniaxial tension test and bond model in two-dimension.

Fig. 12. Eect of k on the uniaxial stressstrain relationship, ec = ep bc = bp.

352

Z.N. Zhang, X.R. Ge / Theoretical and Applied Fracture Mechanics 43 (2005) 342359

 is: b 1 ep mep . As discussed in Section 3.1, mep cos h sin h, where g 0 = (cosh, sinh). The maximum of b p 2 when the L-bond rotation angle is smaller than bp, the R-bond density does not reduce with increasing deformation. Therefore, the value of bc should be no less than bp, i.e. bc P bp. Here, k is a coecient used to adjust the bond reduction ratio with deformation. For dierent materials, their actual strength rb may be dierent though their proportional strength limit rp and initial Youngs modulus are identical. Figs. 1214 show the eects of the three parameters on the uniaxial tension complete stressstrain relationship. The model parameters for these gures are: E = 40 GPa, m = 0.2, ep = 0.0006, bp 1 e mep 2 p 0.00036, V = 1, l0 = 1. Fig. 12 shows the eect of parameters k on the complete stressstrain relationship. From Fig. 12, it can be found that the peak-stress, i.e. rb, decreases with k increasing. This is because that the bigger the k is, the faster the density decreases with deformation increasing according to Eqs. (20)(23). The ratio of the bond density reduction dominates the material strength. Fig. 13 shows the eect of bc on the complete stressstrain relationship. From Fig. 13, it can be found that the bc dominate the post-peak behavior of material. As bc varies from bp to 5bp, the post-peak behavior changes considerably while the peak stress changes very little. The bigger the bc is, the slower the post-peak stress declines. Fig. 14 shows

Fig. 13. Eect of bc on the uniaxial stressstrain relationship, k = 0.1 ec = ep.

Fig. 14. Eect of ec on the uniaxial stressstrain relationship, k = 0.1 bc = bp.

Z.N. Zhang, X.R. Ge / Theoretical and Applied Fracture Mechanics 43 (2005) 342359

353

the eect of ec on the stressstrain relationship. From Fig. 14, it can be found that the value of ec has a considerable eect on the peak-stress value, but little on the varying ratio of the post-peak stress. The smaller the value of ec is, the smaller the peak-stress. This is because that the smaller the value of ec, the earlier the L-bond density reduction begins.

4. Application to the crack initiation and propagation 4.1. Mechanism of simulating crack initiation and propagation When the deformation exceeds certain value, material modulus would reduce. This simulates crack initiation and propagation as shown in Fig. 15. As the deformation of the element at the fracture tip exceed certain value, the corresponding element would have negligible contribution because its stiness tends to zero. This is equivalent to the state of fracture propagation. By the same reason, as the local deformation exceed certain value, crack would initiate as shown in Fig. 15(b). 4.2. Simulation example To simulate crack initiation and propagation, a typical example is presented. Consider a rectangular plate with an around hole at the center, whose dimensions are shown in Fig. 16(a). A uniform tension stress

Fig. 15. Mechanism of the presented model simulating crack initiation and propagation. (a) Fracture propagation (from [1]). (b) Crack initiation.

354

Z.N. Zhang, X.R. Ge / Theoretical and Applied Fracture Mechanics 43 (2005) 342359

Fig. 16. Simulation object and its boundary conditions.

Table 1 Model parameters for numerical simulation E 40 GPa m 0.2 eb 0.002 ep 0.0006 ec 0.0006 bc 0.000036 k 0.1

rp is applied at the edges. Due to symmetry, only one quarter of the geometry need to be considered. The boundary conditions are shown in Fig. 16(b). The model parameters are listed in Table 1. The FEM mesh scheme is shown as Fig. 17. The total number of elements is 870. The number of nodes along coordinate-x1 is 30, and the number of nodes along the quarter arc of the hole is 31. The element type is 4-node plane isoparameter element. In [1], the representative element volume is unit. In [25], the original bond length is taken as unity. The same will be done in this simulation. Figs. 1821 show the results of the simulation.

Fig. 17. Mesh scheme of Fig. 16(b) for FEM implementation.

Z.N. Zhang, X.R. Ge / Theoretical and Applied Fracture Mechanics 43 (2005) 342359

355

Fig. 18. Relationship between applied stress and the normal stress at point-1 (Fig. 16(b)).

Fig. 19. Normal stress along the coordinate-x1 under dierent applied stress.

Fig. 20. The nal deformed mesh (post-peak: rp = 0.005 MPa).

356

Z.N. Zhang, X.R. Ge / Theoretical and Applied Fracture Mechanics 43 (2005) 342359

Fig. 21. Eect of Poisson ratio on stress eld at crack tip (Point-1).

4.3. Discussion on numerical results Fig. 18 shows the relationship between the applied stress rp and normal stress r22 at the Point-1, Fig. 16(b). From Fig. 18, it can be found that the stress at Point-1 increases with the applied stress rp increasing at rst. As the applied stress reaches a certain value, the Point-1 stress begins to decline. This decline indicates that the zone around Point-1 reaches the post-peak softening phases. With further increase of the applied stress, the stress of Point-1 becomes smaller and smaller due to the reduction of stiness. When the stiness becomes small enough, the contribution of the corresponding element to the whole structure is negligible, although these elements are still kept in the FEM iterative computation. This is equivalent to saying that a crack has initiated at Point-1. From Fig. 18, it also can be found that as the applied stress reaches a certain value, it would never reverse. It decreases though it doesnt reach the tension strength. This is because the net area along coordinate-x1 is smaller than the rectangular edge of the applied stress due to the existence of hole and the yielding area around Point-1. Although the applied stress doesnt reach the tension strength, the elements along coordinate-x1 will have reached their tension strength followed by entering the softening phases. Hence, the applied stress rp would decline. Fig. 19 shows the stress along coordinate-x1 under dierent applied stress rp. It can be found that the stress at the hole edge is much bigger than that farther away from the edge when the applied stress is small, e.g. rp = 28.8 MPa. This is the typical result of the stress intensity at the hole edge. Increasing the applied stress, the domain near the hole edge enter the strain-softening phase. As a consequence, the stress declines. This indicates crack initiation. The crack propagates further with additional increase of the applied stress. The same procedure will repeat. The peak-stress, like a wave, spread forward from the hole edge to the farther end. Moreover, it also can be noticed that the stress intensity becomes weaker and weaker with the crack growth. The nal deformation is shown in Fig. 20. It can be found that the elements at the upper part deform very little. However, the deformation of the elements along the coordinate-x1 is severe. This is equivalent to saying that the plate is intensely fractured along coordinate-x1 . It also can be noticed that a neck-narrowing phenomenon occurs. This is indicated by the point (x1 = 3) moving to the left. The eects of the Poisson ratio on the crack behaviors are shown in Fig. 21. It can be found that the Poisson ratio has considerable eect on the simulated results. The higher the Poisson ratio is, the bigger the computed stress at the tip of crack. This is because for plane strain, the transverse deformation is restricted. The Poisson ratio is a parameter which characterize the property of transverse deformation. Under the condition of the transverse deformation being restricted, the magnitude of Poisson ratio do have eect on the stress eld.

Z.N. Zhang, X.R. Ge / Theoretical and Applied Fracture Mechanics 43 (2005) 342359

357

5. Concluding remarks By adding a R-bond which is used to restrict the rotation freedom. This is in contrast to the VIB model in [1]. The present model can consider Poissons ratio other than 0.25. The derived constitutive relationship connects the discrete structure characters at the microscopic, i.e. the bond stiness, and the mechanical properties at the macroscopic, i.e. the Poisson ratio and the Youngs modulus. Based on the modied VIB model, the local materials failure and the fracture behavior are analyzed. In contrast to the cohesive stress method, the presented method attributes the local material failure to the bond density reduction induced by the local deformation rather than induced by the local cohesive stress. By incorporating a relationship between the bond density and the deformation, the criterion of fracture is directly built into the constitutive model. Through simulation, an example of crack initiation and propagation is presented and the results are discussed.

Acknowledgement During the review of this paper, Prof. G.C. Sih provided some constructive advices and help. The authors would like to thank Prof. G.C. Sih for his valuable instruction, help and encouragement.

Appendix A For the L-bond whose unit orientation vector is n, n = (sin h cos /, sin h sin /, cos h), its rotation angle towards xi-axes is equal to the angle of n towards gi, which is perpendicular to n and located in the plane * * * constructed by n and x i where x i is the unit orientation vector of xi-axes. So, gi n x i n, i = 1,2,3, shown as Fig. A.1. 8 2 2 2 0 2 > sin h cos / cos h < g sin hsin / cos h sin h cos / sin /; 2 A:1 g00 sin h cos / sin /; cos2 h sin2 hcos2 /; sin h sin / cos h > : 000 2 g sin h cos / cos h; sin h sin / cos h; sin h

Fig. A.1. Indication of gi and n.

358

Z.N. Zhang, X.R. Ge / Theoretical and Applied Fracture Mechanics 43 (2005) 342359

Fig. A.2. Indication of bi.

The rotation angle of the L-bond towards xi-axes is: bi nm emn gin The rotation angles are shown in Fig. A.2, emn is the strain tensor at the Point o. A:2

References
[1] H. Gao, P. Klein, Numerical simulation of crack growth in an isotropic solid with randomized internal cohesive bonds, J. Mech. Phys. Solids 46 (1998) 187218. [2] F. Milstein, Review: Theoretical elastic behavior at large strains, J. Mat. Sci. 15 (1980) 10711084. [3] F. Milstein, Crystal elasticity, in: H.G. Hopkins, M.J. Sewell (Eds.), Mechanics of Solids, Pergamon Press, Oxford, 1982. [4] J.L. Ericksen, in: M. Gurtin (Ed.), Phase Transformations and Material Instabilities in Solids, Academic Press, New York, 1984. [5] E.B. Tadmor, M. Ortiz, R. pillips, Quasicontinuum analysis of defects in solids, Phil. Mag. A73 (1996) 15291563. [6] J.R. Rice, A path independent integral and the approximate analysis of strain concentration by Notches and Cracks, J. Appl. Mech. 35 (1968) 379386. [7] G.I. Barenblatt, The formation of equilibrium cracks during brittle fracture: General ideas and hypotheses, axially symmetric cracks, Appl. Math. Mech. (PMM) 23 (1959) 622636. [8] D.S. Dugdale, Yielding of steel sheets containing slits, J. Mech. Phys. Solids 8 (1960) 100104. [9] J.R. Willis, A comparison of the fracture criteria of Grith and Barenblatt, J. Mech. Phys. Solids 15 (1967) 151162. [10] X.-P. Xu, A. Needleman, Numerical simulations of fast crack growth in brittle solids, J. Mech. Phys. Solids 42 (1994) 13971434. [11] G.T. Camacho, M. Ortiz, Computational modelling of impact damage in brittle materials, Int. J. Solids Struct. 33 (1996) 2899 2938. [12] G.C. Sih, X.S. Tang, Dual scaling damage model associated with weak singularity for macroscopic crack possessing a micro/ mesoscopic notch tip, Theor. Appl. Fract. Mech. 42 (2004) 124. [13] G.C. Sih, X.S. Tang, Simultaneity of multiscaling for macro-meso-micro damage model represented by strong singularities, Theor. Appl. Fract. Mech. 42 (3) (2004) 199225. [14] G.C. Sih, Implication of scaling hierarchy associated with nonequilibrium: eld and particulate, Theor. Appl. Fract. Mech. 37 (2001) 335369. [15] G.C. Sih, B. Liu, Mesofracture mechanics: A necessary link, Theor. Appl. Fract. Mech. 37 (2001) 371395. [16] S. Kohlho, P. Gumbsch, H.F. Fischmeister, Crack propagation in BCC crystals studied with a combined nite-element and atomistic model, Phil. Mag. A 64 (1991) 851878. [17] K.S. Cheung, S. Yip, A molecular dynamics simulation of crack tip extension: the brittle-to-ductile transition, Model. Simul. Mater. Sci. Eng. 2 (1994) 856892. [18] F.F. Abraham, D. Brodbeck, R.A. Rafey, W.E. Rudge, Instability dynamics of fracture: a computer simulation investigation, Phys. Rev. Lett. (1994) 272. [19] F.F. Abraham, On the transition from brittle to plastic failure in breaking a nanocrystal under tension (NUT), Europhys. Lett. 38 (1997) 103106.

Z.N. Zhang, X.R. Ge / Theoretical and Applied Fracture Mechanics 43 (2005) 342359

359

[20] S.J. Zhou, D.M. Beazley, P.S. Lomdahl, B.L. Holian, Large-scale molecular dynamics simulations of three-dimensional ductile failure, Phys. Rev. Lett. 78 (1997) 479. [21] J.Q. Broughton, N. Bernstein, E. Kaxiras, F.F. Abraham, Concurrent coupling of length scales: methodology and applications, Phys. Rev. B 60 (1998) 2391. [22] G.C. Sih, Methods of analysis and solutions of crack problemsMech. Fract., 1, Noordho, Leyden. The Netherlands, 1973. [23] D.Y. Zhou, Strain energy densityA molecular appraisal, J. Chinese Inst. Eng. 27 (2004) 919926. [24] T. Ganesh, A. Misra, Fracture simulation for anisotropic materials using a virtual internal bond model, Int. J. Solids Struct. 41 (2004) 29192938. [25] T. Ganesh, K. Jimmy Hsia, Y.G. Huang, Finite element implementation of virtual internal bond model for simulating crack behavior, Eng. Fract. Mech. 71 (2004) 401423. [26] J.E. Marsden, T.J.R. Hughes, Mathematical Foundations of Elasticity, Prentice-Hall, Englewood Clis, New Jersey, 1983. [27] R.W. Ogden, Non-linear Elastic Deformations, John Wiley and Sons, New York, 1984.