International Journal of Solids and Structures 38 (2001) 4759±4776 www.elsevier.com/locate/ijsolstr The spherical-cap crack revisited P.A. Martin * Department of Mathematical and Computer Sciences, Colorado School of Mines, Golden, CO 80401-1887, USA Received 15 May 2000; in revised form 25 July 2000; accepted 11 August 2000 Abstract A crack in the shape of a spherical cap is subjected to a static loading. The exact solution for the crack-opening displacement is obtained using a method based on dual series equations and Laplace transforms. For shallow caps, the solution agrees with an asymptotic theory for perturbed penny-shaped cracks. Ó 2001 Elsevier Science Ltd. All rights reserved. 1. Introduction Let X be a loaded crack in a three-dimensional elastic solid. Assume that X is bounded and that the crack edge oX is a simple, smooth, closed curve. The form of the stresses near oX is known (Leblond and Torlai, 1992), but the stress-intensity factors themselves can only be found by solving a boundary-value problem: they cannot be obtained from a local analysis near the edge. There is only one non-planar X for which the boundary-value problem can be solved exactly, and that is a spherical cap. In fact, there are several Russian papers on this problem. One of the earliest is that of Ziuzin and Mossakovskii (1970), but their analysis was subsequently criticised by Prokhorova and Solov'ev (1976); both papers consider axisymmetric loadings, and use representations in terms of analytic functions of a complex variable. A method for non-axisymmetric problems was developed more recently by Popov (1992). He reduced the problems to some one-dimensional integral equations, whose desired solutions were shown to have non-integrable end-point singularities. A method based on dual series equations was sketched by Martynenko and Ulitko (1979). Their method is simple, in principle (we use it below), but it is, nevertheless, complicated when detailed results are required. A striking feature of these papers is that they do not contain mutual comparisons. Thus, it is dicult to know what the correct solution actually is, for any particular loading! Consequently, we decided to re-work the problem. The main purpose of the calculation is to obtain a benchmark solution, so that other techniques (such as those based on solving a boundary integral equation numerically) can be validated. We have also con®rmed the results for a shallow spherical cap by comparing with an asymptotic theory for perturbed penny-shaped cracks (Martin, 2000). * Fax: +1-303-273-3875. E-mail address: pamartin@mines.edu (P.A. Martin). 0020-7683/01/$ - see front matter Ó 2001 Elsevier Science Ltd. All rights reserved. PII: S 0 0 2 0 - 7 6 8 3 ( 0 0 ) 0 0 3 1 0 - 3 4760 P.A. Martin / International Journal of Solids and Structures 38 (2001) 4759±4776 The method used here is ®rst described in the context of a model problem for Laplace's equation. It consists of patching separated solutions together in spherical polar coordinates, leading to some dual series equations, which are then reduced to an integro-dierential equation. We solve this equation using Laplace transforms. We then generalise the method to the elasticity problem, which is much more complicated. It leads to a pair of coupled integro-dierential equations, which are essentially those obtained previously by Martynenko and Ulitko (1979). We solve these using Laplace transforms. We also derive asymptotic approximations for the stress-intensity factors, valid for shallow spherical-cap cracks. 2. A model problem: potential ¯ow past a rigid spherical cap Before embarking on the elasticity problem, it is instructive to consider a simpler problem for Laplace's equation. The techniques used to solve this problem will generalise to the much more complicated crack problem. Thus, consider a rigid spherical cap, given by r c; 0 6 h < a; 0 6 / < 2p; where r, h and / are spherical polar coordinates. We want to solve Laplace's equation, r2 u 0, outside the cap, with ou=or U cos h on both sides of the cap; 2:1 so that Uz u is the velocity potential for axisymmetric uniform ¯ow past the cap; u is required to vanish as r ! 1. Separation of variables gives the representations 1 X n An r=c Pn cos h for 0 6 r < c and u r; h Uc n0 u r; h Uc 1 X n0 Cn r=c n 1 Pn cos h for r > c; where Pn is a Legendre polynomial and the dimensionless coecients An and Cn are to be found. Continuity of ou=or across r c for 0 6 h 6 p gives nAn n 1Cn . De®ne the discontinuity in u across r c by u h lim u r; h r!c Uc lim u r; h r!c 1 X 2n 1 Un Pn cos h; n0 where Un An n 1 1 . Using Eq. (2.1) and the fact that u h 0 for a 6 h 6 p gives 1 X n n 1 Un Pn cos h cos h; 0 6 h < a; n0 1 X n0 2n 1 Un Pn cos h 0; a 6 h 6 p: 2:2 2:3 These form a pair of dual series equations for Un (Sneddon, 1966). To solve them, we use an integral representation for Un , Z a 1 1 Un u t sin n t dt; 2:4 2n 1 0 2 4761 P.A. Martin / International Journal of Solids and Structures 38 (2001) 4759±4776 where u t is to be found. This representation ensures that Eq. (2.3) is satis®ed for any u, due to the discontinuous sum (A.4). It also gives Z a u t p dt; 0 6 h < a: u h Uc 2 cos h 2 cos t h From this, we can readily show that we must have u 0 0, otherwise u 0 would be unbounded. Substitution of Eq. (2.4) in Eq. (2.2) gives Z 1 X n n 1 a 1 u t sin n t dt Pn cos h cos h; 0 6 h < a; 2n 1 0 2 n0 an equation for u t. To solve it, we would like to interchange the order of integration and summation, so as to obtain an integral equation for u; however, the resulting sum is divergent, so that we must proceed indirectly. An integration by parts in Eq. (2.4) gives Z a 2k2n Un u0 t cos kn t dt u a cos kn a; 2:5 0 where kn n 1=2 and we have used u 0 0. Then, we write Eq. (2.2) as 1 1 X 1X 2k2n Un Pn cos h Un Pn cos h 2 cos h; 0 6 h < a: 2 n0 n0 Using Eqs. (2.4), (2.5), (A.3) and (A.5) gives Z 1 a 0 u s ds 2 cos h; Tt!h u t 4 t 0 6 h < a; where we have de®ned the Abel operator T by Z h / t dt p : T / Tt!h f/ tg 2 cos t 2 cos h 0 2:6 2:7 This operator can be inverted: if T / f , we can solve for / as (Porter and Stirling, 1990, Section 9.2) Z s 2 d f h sin h 1 1 p dh: / s T f Th!s f f hg 2:8 p ds 0 2 cos h 2 cos s Applying T u0 t 1 to Eq. (2.6) gives Z 1 a 3 t ; u s ds 4=p cos 4 t 2 2:9 which is to be solved subject to u 0 0. One way to solve Eq. (2.9) is to dierentiate with respect to t, leading to a second-order ordinary differential equation for u with constant coecients. The general solution of this equation contains two arbitrary constants; these are to be determined by imposing u 0 0 and by requiring that the solution of the dierential equation actually solves Eq. (2.9). We shall use an alternative method, based on Laplace transforms. First, we write Eq. (2.9) in convolution form as u0 t 1 4 Z t 0 1 u s ds M1 4 4=p cos 3 t ; 2 2:10 4762 P.A. Martin / International Journal of Solids and Structures 38 (2001) 4759±4776 where M1 Ra u t dt is an unknown constant. Next, we de®ne Z 1 u t e pt dt; U p Lfug 0 2:11 0 where p is the transform variable. Then, taking the Laplace transform of Eq. (2.10), we obtain 4=pp2 = p2 94; p2 14 U p 14M1 whence U p M1 1 2p 4 p2 1 14 9 1 : 2 2p p 94 Inverting, using Lfsin btg b= p2 b2 , we obtain u t 12M1 p 1 sin 12t 3=p sin 32t: To determine M1 , we substitute for u in the de®nition of M1 and evaluate the integral, giving 4=p sin a sin 12a M1 and then u t 1=p sec 12a cos 32a sin 12t 3=p sin 32t: This solution agrees with the well known solution of Collins (1959). 3. Elastic ®eld representations For the crack problem, we start with representations for the elastic displacement u in terms of potentials. From Lur'e (1964, Section 6.2), we have general solutions for three-dimensional elasticity, in spherical polar coordinates r; h; /, where u ur ; uh ; u/ is independent of the azimuthal angle, /. Thus, the following representations can be obtained. Interior solution: displacements 2 4m Bnrn 1 gPn ; ur fArn1 n 1 n 4m Brn 1 gPn1 : fArn1 n 5 uh Exterior solution: displacements ur fCr n n n 3 uh fCr n n 4m 4 4m Dr n 2 Dr n 2 n 2 n 1gPn ; gPn1 : Interior solution: stresses 2l 1 srr fA n 1 n2 1 2l srh fA n2 2n 2mrn Bn n 1 2mrn B n 1rn 2 gPn ; 1rn 2 gPn1 : Exterior solution: stresses 1 fCn n2 3n 1 fC n2 2l srr 2l srh 2mr 2 2mr n 1 n 1 D n 1 n 2r D n 2r n 3 gPn1 : n 3 gPn ; P.A. Martin / International Journal of Solids and Structures 38 (2001) 4759±4776 4763 All these solutions are valid for n 0; 1; 2; . . .. In them, m is Poisson's ratio (Lur'e uses m 1=m), l is the shear modulus, Pn Pn cos h and Pn1 Pn1 cos h Pn0 cos h sin h d=dhPn cos h; note that P01 0. Expressions for the other stress components are given on p. 330 of Lur'e's book. We consider a crack in the shape of a spherical cap, given by r c, 0 6 h < a and 0 6 / < 2p. Using superscripts 1 and 2 for the regions r < c and r > c, respectively, we have the following representations for the displacements and stresses: 1 r n1 r n 1 X ur 1 A0 4m 2r c n 1 n 2 4m Bn n Pn ; 3:1 An c c n1 1 c 2 c n c n2 X 2 ur cD0 Cn c n n 3 4m Dn n 1 Pn ; 3:2 r r r n1 1 r n1 r n 1 X 1 uh c An n 5 4m Bn Pn1 ; 3:3 c c n1 1 c n c n2 X 2 uh c n 4 4m Dn 3:4 Cn Pn1 ; r r n1 1 1 2l srr 1 2 2l srr 2 1 mA0 1 r n r n 2 X 2 n 1 n n 2 2m Bn n n 1 Pn ; An c c n1 1 c 3 X c n1 c n3 2 2D0 Cn n n 3n 2m Dn n 1 n 2 Pn ; r r r n1 1 1 r n X n2 2n An c n1 2 1 X 2l 1 srh 2l 1 srh Cn n1 c n1 r n2 1 2m Bn 2 2m Dn r n c c n3 r 2 n 1 Pn1 ; 3:5 3:6 n 2 Pn1 : These formulas were obtained using A c n An , B c2 n Bn , C cn1 Cn and D cn3 Dn , where the coecients An , Bn , Cn and Dn are dimensionless. These coecients are to be determined by applying the boundary conditions on r c. The ®rst of these is that the stresses should be continuous across r c for all h, when 1 mA0 D0 ; An n 1 n2 An n2 2n n 3:7 2 2m Bn n n 1 2m Bn n Cn n n2 3n 1 1 Cn n2 2 2m 2m Dn n 1 n 2; Dn n 2; 3:8 3:9 for n 1; 2; . . .. From these, we obtain Dn Cn n 1f n 1 2n 3An 2n 1Bn g; n 2Dn Dn 2n 1fn n 2 n2 where Dn 2fn2 when n 0. n 2n 1 1 1 4 4m2 gAn n n 3:10 1 n 2 2n 1Bn ; 3:11 mg. In particular, C1 0. Note that Eq. (3.11) reduces to Eq. (3.7) 4764 P.A. Martin / International Journal of Solids and Structures 38 (2001) 4759±4776 If the material in r < c is dierent from that in r > c, one obtains an interface crack on r c. The corresponding problem has been considered by Altenbach et al. (1995), using the representations given above. 4. The crack-opening displacement We de®ne the crack-opening displacement u by ur h ur 1 c; h ur 2 c; h c 1 2 uh h uh c; h uh c; h c 1 X 2n 1 Un Pn ; 4:1 n0 1 X n1 2n 1 Vn Pn1 ; 4:2 where the dimensionless coecients Un and Vn are to be determined. Comparing with the representations (3.1)±(3.4), we ®nd that U0 4m 2A0 D0 3 1 mA0 , 2n 1Un An n 1 n 2n 1Vn An n 5 2 4m Bn n Cn n n 3 4m Bn Cn n 4 4m Dn : 4m Dn n 1; Substituting for Cn and Dn from Eqs. (3.10) and (3.11), respectively, we ®nd that n 2Dn Un mf 2n 3 n2 2 1 n 2Dn Vn 2 1 2 mf 2n 3 n 3n 2 2m n 1An 2n 2mAn 2n 1 n 2nBn g; 1 n 2Bn g: We can rewrite these, giving An and Bn in terms of Un and Vn : 2 1 2 1 m 2n 3An 2 n 2 Un nVn ; 2mUn n 1 n2 1Bn n 3n m 2n 2 2mVn : We can now state the problem to be solved: ®nd Un and Vn , so that ur h 0 and uh h 0 for a < h 6 p; 4:3 and srr h lqr h and srh h lqh h on r c for 0 6 h < a; 4:4 where qr and qh are given functions of h. In particular, for uniaxial tension at in®nity in the z-direction (so that s1 zz p0 , say), we have (Lur'e, 1964, p. 343) qr h p0 =l cos2 h qh h and p0 =l sin h cos h: 4:5 Using Eq. (4.4) in Eqs. (3.5) and (3.6) gives 1 qr 2 1 qh 2 2 1 mA0 1 X n1 An n2 2n 1 X n1 An n 1 n2 1 2m Bn n n 2 1 Pn1 : Eliminating An and Bn in favour of Un and Vn gives 2m Bn n n 1 Pn ; P.A. Martin / International Journal of Solids and Structures 38 (2001) 4759±4776 1 1 2 mqr 1 1 2 mqh 1 X n0 1 X n1 2w2n 1Un wn wn 1 m 4wn 1 m 2wn fwn 1 m 4wn 5Un wn 1 2wn 5Vn 3 3mVn g Pn ; 4wn 3 Pn1 4wn 3 4765 4:6 4:7 for 0 6 h < a, where wn n n 1. These are to be solved subject to Eq. (4.3), wherein u is de®ned by Eqs. (4.1) and (4.2). Of particular interest are the stress-intensity factors. We de®ne these by p p p p ur Kn 2a c a h and uh Ks 2a c a h as h ! a ; 4:8 where a is a length scale and Kn and Ks are dimensionless stress-intensity factors. It is convenient to take a c sin a, for we may then consider the limiting case of a penny-shaped crack of radius a, obtained by taking the limits c ! 1 and a ! 0 with a ®xed. The de®nitions of the stress-intensity factors in Eq. (4.8) are convenient, but not standard. For example, p e n = 2pq0 as q0 ! 0, where q0 is distance from the crack edge oX. Making it is usual to suppose that srr K use of the known general relations between u behind oX and the stress components ahead of oX (see, for example, Rice (1989, p. 32)), we ®nd that p e n 1l paKn = 1 m: K 4:9 2 This formula and the corresponding formula for Ks can be used to obtain expressions for the standard stress-intensity factors from the results derived below. 5. Reduction to integro-dierential equations We introduce representations (2.4) for Un and Z a 1 1 w t cos n t dt Vn 4n n 1 0 2 5:1 for Vn , where the functions u and w are to be found. Substituting Eq. (2.4) in Eq. (4.1), followed by evaluation of the sum using Eq. (A.4), shows that ur h 0 for h > a, as required, for any choice of u; we also obtain Z a u t p dt; 0 6 h < a: ur h c 5:2 2 cos h 2 cos t h ur 0 will be bounded provided that t 1 u t is integrable near t 0, so that, in particular, u 0 0: 5:3 Similarly, substituting Eq. (5.1) in Eq. (4.2), followed by use of Eq. (A.7) shows that uh h 0 for h > a, as required, provided that w satis®es Z a 1 t dt 0: 5:4 w t cos 2 0 We also obtain uh h c 2 sin h Z a h w t sin t p dt; 2 cos h 2 cos t 0 6 h < a; it turns out that uh 0 0, as expected from symmetry considerations. 5:5 4766 P.A. Martin / International Journal of Solids and Structures 38 (2001) 4759±4776 The stress-intensity factors, Kn and Ks , can be expressed directly in terms of u and w, respectively. Thus, from Eq. (5.2), we have p 1 ur h c u a sin a 2 cos h 2 cos a as h ! a with a similar approximation for uh . Then, comparison with Eq. (4.8) gives Kn u a sin a and Ks w a : 2 sin a 5:6 To obtain u and w, we substitute Eqs. (2.4) and (5.1) in Eqs. (4.6) and (4.7), and evaluate the series using results from the appendix. To do this, we note the following partial-fraction expansions: 2w2n 1 m 2wn 1 kn d1 d3 1 1 ; 5:7 2n 1 4wn 3 4 4kn 8 kn 1 kn 1 wn 1 m 4wn 5 d3 1 1 d0 ; 5:8 4wn 3 2 k n 1 kn 1 wn 1 m 4wn 5 d0 kn d2 3d3 1 1 ; 5:9 2n 1 4wn 3 2wn 8wn kn 16wn kn 1 kn 1 wn 1 2wn 3 3m k2n 3d3 1 1 : 5:10 4wn 3 kn 1 kn 1 2 4 In these, wn n n 1, kn n 1=2, d0 14 1 4m; d1 163 5 8m; d2 163 1 8m and d3 161 7 8m: However, prior to substitution, we note that the ®rst terms on the right-hand sides of Eqs. (5.7) and (5.10) will lead to divergent series. To overcome this, we ®rst integrate by parts in Eqs. (2.4) and (5.1). Thus, making use of u 0 0, we obtain Eq. (2.5) and then Z 1 X kn 1 a 0 1 u a f0 a; h 2n 1Un Pn cos h u t f0 t; h dt 4 4 4 0 n0 Z 1 h u0 t dt p ; 4 0 2 cos t 2 cos h where f0 is de®ned by Eq. (A.2) and we have used Eq. (A.3), noting that h < a. Similarly, we have Z a w0 t sin kn t dt; 4wn kn Vn w a sin kn a 0 whence 1 X n1 Vn k2n 1 1 P cos h 8 sin h 2 n Z h 0 w0 t sin t dt p ; 2 cos t 2 cos h where we have used Eqs. (A.6) and (5.4). Use of these results, together with Eqs. (2.4), (5.1) and (5.7)± (5.10), in Eqs. (4.6) and (4.7) gives Z a 1 1 0 1 mqr Tt!h fu tg 5:11 fu t S11 t; h w t S12 t; hgdt; 2 4 0 Z a 1 1 1 mqh Tt!h fw0 t sin tg 5:12 fu t S21 t; h w t S22 t; hgdt; 2 8 sin h 0 P.A. Martin / International Journal of Solids and Structures 38 (2001) 4759±4776 4767 for 0 6 h < a, where 1 X d1 d3 1 1 S11 Pn cos h sin kn t; 4kn 8 kn 1 kn 1 n0 1 X d0 d3 1 1 S12 Pn cos h cos kn t; 4 8 kn 1 kn 1 n0 1 X d0 kn d2 3d3 1 1 sin kn t ; S21 Pn1 k k wn 2 8k 16 1 1 n n n n1 1 3d3 X 1 1 cos kn t : Pn1 cos h S22 wn 16 n1 kn 1 kn 1 The Abel operator T and its inverse are de®ned by Eqs. (2.7) and (2.8), respectively. The sums Sij can be evaluated explicitly using results from the appendix. We obtain 1 S11 4 S12 Z t f0 s; h fd1 0 d0 f0 t; h 4 d3 4 sin h S21 12d0 f0 t; h sin t 3d3 sin h S22 8 Z 0 d3 cos t Z t 0 f0 s; h sin t s ds d3 sin3 12h sin t 1 8 Z t f01 s; h 1 h ; 2 d3 1 h ; cos t sin 2 2 d3 sg ds sin t sin 2 sin t s ds d3 sin 2 t 0 f01 s; h fd2 3d3 cos t 1 h 2 cos 1 t 2 sin sg ds; 1 h 2 cos t ; where f01 is de®ned by Eq. (A.8). Then, changing the order of integration gives Z a d3 1 h u t sin t dt; d1 d3 cos t s u s ds sin 2 2 0 0 t Z a Z a Z a 1 d3 1 h sin w t S12 t; h dt Tt!h d0 w t d3 w s sin t s ds w t cos t dt; 4 2 2 0 t 0 Z a Z a 1b sin h d2 3d3 cos t s u s ds u t S21 t; h dt T t!h 8 t 0 Z a 1 1 3 h u t sin t dt; d0 Tt!h fu t sin tg d3 sin 2 2 0 Z a Z a Z a 3d3 b 1 3 h w t S22 t; h dt T t!h sin h w t cos t dt; w s sin t s ds d3 sin 2 8 0 0 t Z a 1 u t S11 t; h dt Tt!h 4 Z a where we have used Eq. (5.4) to obtain the last formula and the operator Tb is de®ned by Tb / Tbt!h f/ tg Z 0 h p / t 2 cos t 2 cos h dt: 4768 P.A. Martin / International Journal of Solids and Structures 38 (2001) 4759±4776 Using these results, we apply 4T 1 to Eq. (5.11), giving Z a Z a 0 u s d0 w s d3 sin s fu t sin t w t cos tg dt fd1 0 s Z a d3 w t sin s t dt Qr s d3 cos s tg u t dt s 1 1 1 for 0 < s < a, where Qr s 2 1 m Th!s sin 2 h 1=2 sin s. fqr hg and we have used Th!s 1 Next, we apply 8 sin s T 1 sin h to Eq. (5.12). To do this, we ®rst calculate that Z s 1 b T t!h f/ tg sin s Th!s / t dt; and 0 sin s sin2 12 s, whence Z s Z a Z a 0 w s 4d0 u s d2 u t t dt s u t dt 3d3 cos s fu t sin t 0 s 0 Z s Z a 3d3 w t dt 3d3 fu t sin s t w t cos s tg dt Qh s; 1 Th!s sin3 12 h 5:13 3 4 0 s w t cos tg dt 5:14 1 1 for 0 < s < a, where Qh s 4 1 m sin s Th!s fqh h sin hg. Eqs. (5.13) and (5.14) have been obtained by Martynenko and Ulitko (1979), although few details were given; see their Eq. (2.4), noting that our w is their 4w. These authors considered all-round tension at in®nity. For uniaxial tension at in®nity, qr and qh are given by Eq. (4.5); elementary calculations, using Eq. (2.8), then give Qr s 13g cos 12s 2 cos 52s with g 4 1 and Qh s 85g sin 52s; 5:15 mp0 = pl. 6. Solution of the integro-dierential equations Martynenko and Ulitko (1979) solved Eqs. (5.13) and (5.14) (for all-round tension) by repeated differentiation, leading to a pair of coupled homogeneous ordinary dierential equations with constant coecients, relating u, u00 , u iv , w0 , w000 and w v . The general solution of this system contains 17 arbitrary constants, and these were then determined by substituting back in Eqs. (5.13) and (5.14), together with the conditions (5.3) and (5.4). Here, we solve Eqs. (5.13) and (5.14) using Laplace transforms. First, we rewrite the integro-dierential equations in convolution form as Z s u0 s d0 w s 6:1 fk11 s t u t k12 s t w tgdt g1 s; Z0 s w0 s 4d0 u s 6:2 fk21 s t u t k22 s t w tgdt g2 s; 0 where k11 s d1 d3 cos s; k12 s d3 sin s; k21 s d2 s 3d3 sin s; k22 s 3d3 1 cos s; g1 s Qr s d3 M2 cos s d1 M1 ; g2 s Qh s 3d3 M2 sin s sd2 M1 ; P.A. Martin / International Journal of Solids and Structures 38 (2001) 4759±4776 M1 Z a u t dt and 0 M2 Z 0 4769 a fu t cos t w t sin tg dt: 6:3 Note that the constants M1 and M2 are unknown, as they depend on the solutions u and w. Next, we introduce U p Lfug and W p Lfwg, the Laplace transforms of u and w, respectively, de®ned by Eq. (2.11); we use upper-case letters to denote the Laplace transforms of other functions. Then, taking the Laplace transform of Eqs. (6.1) and (6.2) gives p K11 U d0 K12 W G1 ; 4d0 K21 U p K22 W G2 w0 ; where w0 w 0 and we have used u 0 0. Hence DU p K22 G1 d0 K12 G2 w0 ; DW 4d0 K21 G1 p K11 G2 w0 ; where D p K11 p K11 p d1 p K21 p K22 d0 6:4 6:5 K12 4d0 K21 . We have d3 p d3 ; ; K12 p 2 p2 1 p 1 d2 3d3 3d3 ; K22 p ; 2 2 p p 1 p p2 1 whence D p p2 94 R p2 ; p2 p2 1 6:6 where R x x2 8m2 x 169 : 1 3 2 p Note that R x 0 when x 34 2m2 2imc with c 12 3 4m2 . From Eqs. (6.4)±(6.6), we obtain U p W p p2 p2 1 4 d0 p2 d2 p2 14G1 p4 1 2d0 p2 p2 94 R p2 using d3 d0 d2 d1 G1 p g G2 p g p2 3d3 pG1 d0 p2 d2 p2 G2 w0 ; p2 94 R p2 d1 p G2 w0 2d0 . Also, for uniaxial loading, Eq. (5.15) gives p p2 94 d3 M2 p2 d1 M1 p2 1 ; 1 25 2 p p2 1 4 p 4 4 3d3 M2 p2 d2 M1 p2 1 : 25 p2 p2 1 p2 4 Thus, it is convenient to write U gU1 U2 and W gW1 W2 : We have U1 ; p2 A1 p2 p2 14 p2 94 p2 254 R p2 and U2 A2 p2 ; p2 94 R p2 4770 P.A. Martin / International Journal of Solids and Structures 38 (2001) 4759±4776 where A1 p2 p6 14p4 9 2 2 A2 p p p 2 16m 161 p2 23 2 3d0 d3 M2 p d0 p 88m 2 d2 w0 1 177 64 fp4 d1 120m; p2 d1 d0 d2 81 gM1 : 64 Next, we split into partial fractions, as p2 34D1 p2 A1 B1 C1 1 9 25 R p2 p 2 4 p2 4 p 2 4 U1 p2 34D2 p2 B2 R p2 p2 94 U2 3 E2 4 3 E1 4 ; ; where 1 ; 16 1 m A1 B1 3 16 1 m 3 5m 10m2 ; 4 1 m2 7 5m D1 m E1 25 C1 ; 4 7 5m ; 13 25m 6m2 20m3 ; 8 1 m2 7 5m 3 f16d3 M2 4w0 3 8mM1 g; 64 1 m m D2 f16d3 M2 4 3 4mw0 13 24mM1 g; 32 1 m 1 16d3 1 2mM2 4 1 6m 8m2 w0 51 E2 64 1 m B2 14m 48m2 M1 : Note that lim U1 p 1 A1 B1 C1 D1 E1 : 6:7 p!1 Inverting the Laplace transforms, we obtain u t 2gA1 sin 12t 23 gB1 B2 sin 32t 25gC1 sin 52t m 1 gD1 D2 cosh ct sin mt c 1 gE1 E2 sinh ct cos mt; using Lfcosh ct sin mtg m p2 34=R p2 and Lfsinh ct cos mtg c p2 Similarly, for W, we have p B1 p2 9 p2 254 R p2 4 and W2 B1 p2 5 4mp4 72p2 1 2m 3 29 16 4 W1 p2 p B2 p2 ; 94 R p2 p2 where 2 B2 p 1 m 4p 3 1 16 2 3d3 M2 fp 1 mf4 1 8mp Splitting into partial fractions, 2 40m; 2d0 p2 3 40mgM1 : d1 gw0 3 =R p2 . 4 6:8 4771 P.A. Martin / International Journal of Solids and Structures 38 (2001) 4759±4776 W1 W2 p B3 p C3 p p2 p2 94 p2 254 p B4 p p2 p2 94 16 B, 9 1 where B3 D3 11 3 1 C3 c2 m2 D3 2pE3 ; R p2 c2 m2 D4 2pE4 ; R p2 24 C, 25 1 13m ; m 7 5m B4 E3 16 B, 9 2 m 1 2m 15 17m ; 6 1 m 7 5m 1 f16d3 M2 4 2 3mw0 3 8mM1 g; 12 1 m m 16md3 M2 4 3 2m 6m2 w0 48 3m E4 24 1 m D4 56m2 M1 : Inverting the Laplace transforms, we obtain w t gB3 B4 cos 32t gC3 cos 52t gD3 D4 cosh ct cos mt cm 1 gE3 E4 sinh ct sin mt; 6:9 using Lfcosh ct cos mtg p p2 c2 m2 =R p2 and Lfsinh ct sin mtg 2cmp=R p2 . Note that, setting t 0 in Eq. (6.9), we obtain w 0 w0 , as expected, since B3 C3 D3 0 and B4 D4 w0 . 6.1. Determination of M1 , M2 and w0 At this stage, we have expressions for u t and w t involving the three constants w0 , M1 and M2 . To determine these, we use the de®nitions of M1 and M2 , Eq. (6.3), and constraint (5.4). We start by rewriting Eqs. (6.8) and (6.9) so as to display the three constants. Thus, we have u t gu1 t 1 w t gw1 t 1 1 m fw0 h0 t M1 h1 t M2 h2 tg; 6:10 1 m fw0 `0 t M1 `1 t M2 `2 tg; 6:11 where u1 t 2A1 sin 12t 23B1 sin 32t 25C1 sin 52t m 1 D1 cosh ct sin mt c 1 E1 sinh ct cos mt; w1 t B3 cos 32t C3 cos 52t D3 cosh ct cos mt cm 1 E3 sinh ct sin mt; h0 t h1 t h2 t `0 t `1 t `2 t 1 fsin 32t 3 4m cosh ct sin mt 12c 1 1 6m 8m2 sinh ct cos mtg; 8 1 f 3 8m sin 32t 13 24m cosh ct sin mt 12c 1 51 14m 48m2 sinh ct 32 1 d fsin 32t cosh ct sin mt 12c 1 1 2m sinh ct cos mtg; 2 2 1 fcos 32t 2 3m cosh ct cos mt 12c 1 3 2m 6m2 sinh ct sin mtg; 3 1 f 3 8m cos 32t cosh ct cos mt 12c 1 48 3m 56m2 sinh ct sin mtg; 12 4 d fcos 32t cosh ct cos mt 12 m=c sinh ct sin mtg: 3 2 cos mtg; Then, Eqs. (5.4) and (6.3) give 3 X j1 Aij xj ci ; i 1; 2; 3; where x1 M1 , x2 M2 , x3 w0 , 6:12 4772 P.A. Martin / International Journal of Solids and Structures 38 (2001) 4759±4776 c1 1 mg Z A11 1 m Z Z A21 Z A23 A31 Z u1 dt; c2 1 h1 dt; A12 mg Z h1 cos t `1 sin t dt; Z u1 cos t w1 sin t dt; A13 h2 dt; A22 1 Z m Z c3 1 mg Z w1 cos 12t dt; h0 dt; h2 cos t `2 sin t dt; h0 cos t `0 sin t dt; `1 cos 1 t dt; 2 A32 Z `2 cos 1 t dt; 2 A33 Z `0 cos 1 t dt 2 and all the integrals are over the range 0 6 t 6 a; they are all elementary. Eq. (6.12) is a system of three simultaneous algebraic equations for M1 , M2 and w0 . Then, u and w are given by Eqs. (6.10) and (6.11), respectively, the components of the crack-opening displacement u are given by Eqs. (5.2) and (5.5), and the stress-intensity factors are given by Eq. (5.6). Rather than giving the explicit solution of system (6.12) (which is straightforward but tedious), we obtain an approximate asymptotic solution for a shallow spherical cap. 7. The shallow spherical-cap crack Suppose that the cap is shallow, which means that c ! 1 and a ! 0 with a c sin a ®xed. Then, the stress-intensity factors are given by Eq. (5.6) in which a is small. We can solve the 3 3 system (6.12) in this limit. We have u1 t t O t3 ~ 2 O t4 w1 t wt and as t ! 0; making use of Eq. (6.7) and B3 C3 D3 0; the constant w~ is given by w~ 9 B 8 3 25 C 8 3 12 c2 m2 D3 E3 12 5 4m: There are similar approximations for hi t and `i t, and these lead to small-a approximations for Aij and ci . Hence, we obtain M1 M2 12ga2 O a4 as a ! 0. It follows that u a ga O a3 and and w0 and O a4 ~ 2 O a4 w a 23gwa as a ! 0, whence Kn g O a2 1 ~ 2 gwa 3 Ks 1 g 5 6 4ma O a3 as a ! 0 7:1 with g 4 1 mp0 = pl. The leading-order term corresponds to the well-known stress-intensity factor for a penny-shaped crack opened by a constant pressure; (Eq. (4.9)). The ®rst-order correction is seen to occur in the tangential component. We have derived this ®rst-order correction by an independent method, giving some credence to both approaches. In that method, we combined a perturbation expansion with an exact hypersingular boundary integral equation for u, the method being designed for cracks that are perturbations of ¯at circular cracks, namely, wrinkled penny-shaped cracks (Martin, 2000). P.A. Martin / International Journal of Solids and Structures 38 (2001) 4759±4776 4773 Appendix A. The Mehler±Dirichlet integral, sums and variants The standard Mehler±Dirichlet integral is (Whittaker and Watson, 1927, Section 15.231) Z cos n 12t 2 h p Pn cos h dt; p 0 2 cos t 2 cos h valid for n 0; 1; 2; . . . , and 0 6 h 6 p. De®ne kn n 12 and 1=2 ; 0 < t < h; f0 t; h 2 cos t 2 cos h 0; h < t 6 p; Rp so that Pn cos h 2=p 0 f0 t; h cos kn t dt. De®ne f0 t; h for p 6 t < 2p by f0 t; h f0 2p then expand the resulting extended function as a half-range Fourier cosine series. The result is f0 t; h 1 X n0 Pn cos h cos kn t; 0 < t; h 6 p; t 6 h: A:1 A:2 t; h and A:3 Such discontinuous sums are useful when solving dual series equations. Replacing t and h in Eq. (A.1) by p t and p h, respectively, we obtain Pn cos h Rp 2=p 0 f1 t; h sin kn t dt, where 0; 0 6 t < h; f1 t; h 1=2 2 cos h 2 cos t ; h < t 6 p: Extending f1 using f1 t; h f1 2p f1 t; h 1 X n0 t; h for p 6 t 6 2p and then expanding as a half-range sine series gives Pn cos h sin kn t; 0 6 t; h < p; t 6 h: From Eqs. (A.3) and (A.4), we have Z t 1 X sin kn t f0 s; h ds; Pn cos h kn 0 n0 1 X n0 Pn cos h cos kn t kn Z t p f1 s; h ds: We also have f1 t; h cos t f0 t; h sin t f0 t; h cos t f1 t; h sin t 1 X n0 1 X n0 Pn cos h sin kn 1t; Pn cos h cos kn 1t; whence 1 X n0 and Pn cos h sin kn 1t kn 1 Z t 0 ff0 s; h cos s f1 s; h sin sg ds S ; A:4 A:5 4774 P.A. Martin / International Journal of Solids and Structures 38 (2001) 4759±4776 1 X n0 Thus, 1 X n0 cos kn 1t Pn cos h kn 1 Pn cos h Z p ff1 s; h cos s f0 s; h sin sg ds C : t 1 X sin kn t cos kn t S cos t C sin t; C cos t S sin t: Pn cos h kn 1 kn 1 n0 In particular, 1 X 1 1 Pn cos h sin kn t S S cos t C C sin t kn 1 kn 1 n0 Z t 2 f0 s; h cos t s ds 4 sin t sin 12h; 0 1 X n0 1 kn 1 1 kn 1 Pn cos h cos kn t C C cos t 2 Z S S sin t t 0 f0 s; h sin t s ds 4 cos t sin 12h: We need similar formulas involving Pn1 . Use of the recurrence relation 2n 1 Pn1 cos h sin h n n 1fPn 1 cos h Pn1 cos hg gives 2n 1 1 4 Pn cos h sin h n n 1 p Z 4 p p f0 t; h sin t sin kn t dt 0 Z p 0 f1 t; h sin t cos kn t dt for n 1; 2; . . .. It follows that 1 X kn 1 1 1 1 sin t tan h sin t f0 t; h Pn cos h sin kn t 2 2 2 sin h wn n1 1 X kn 1 1 1 1 sin t h cos t f1 t; h Pn cos h cos kn t cot 2 2 2 sin h w n n1 A:6 A:7 for 0 < t; h < p, t 6 h, where wn n n 1. Integrating Eq. (A.6) once gives 1 X cos kn t 1 1 f 1 t; h h cos t 0 ; tan Pn1 cos h wn 2 2 sin h n1 where f01 t; h 2 cos t 0; 1=2 2 cos h ; 0 6 t < h; h 6 t 6 p: Integrating again gives 1 X n1 Pn1 cos h sin kn t kn w n 2 tan 1 h 2 sin Z t 1 1 t f 1 s; h ds: 2 sin h 0 0 A:8 4775 P.A. Martin / International Journal of Solids and Structures 38 (2001) 4759±4776 Similarly, Eq. (A.7) gives 1 X n1 Pn1 cos h sin kn t cot wn 1 h 2 sin 1 t 2 f11 t; h ; sin h where f11 t; h 0; 2 cos h 2 cos t 1=2 0 6 t < h; h 6 t 6 p: ; We also have sin h 1 X n1 Pn1 cos h sin kn 1t 1 cos h sin wn 1 t 2 1 t 2 cos t 1 cos h sin cos t 1 cos h cos sin t 1 t 2 f11 t; h cos t f01 t; h sin t; sin h 1 X n1 Pn1 cos h cos kn 1t wn 1 cos h cos 1 t 2 sin t f01 t; h cos t f11 t; h sin t; whence sin h 1 X n1 Pn1 cos h sin kn 1t wn kn 1 Z t 1 t cos s sin s ds 1 cos h sin 2 0 1 1 1 1 1 sin t 1 cos h sin t sin t 3 2 2 3 2 f01 s; h f11 s; h S1 ; and sin h 1 X n1 Pn1 cos h cos kn 1t wn kn 1 Z p t f11 s; h 1 cos 3 1 t 2 cos s C1 : Thus, sin h 1 X Pn1 cos h sin h 1 X sin kn t S1 cos t C1 sin t; wn kn 1 Pn1 cos h cos kn t C1 cos t S1 sin t: wn kn 1 n1 n1 In particular, 1 f01 s; h sin s ds 1 cos h cos 1 1 1 cos h cos t cos t 2 3 2 1 t 2 4776 P.A. Martin / International Journal of Solids and Structures 38 (2001) 4759±4776 sin h 1 X n1 sin h 1 X n1 1 1 1 kn 1 1 1 kn 1 kn kn 1 Pn1 cos h Pn1 cos h sin kn t S1 S1 cos t C1 C1 sin t wn Z t 2 f01 s; h cos t s ds 0 16 1 1 1 1 sin2 h sin t sin h sin t ; 3 2 2 2 2 cos kn t S1 S1 sin t C1 C1 cos t wn Z t 2 f01 s; h sin t s ds 0 16 1 1 1 sin2 h cos t sin h cos t : 3 2 2 2 References Altenbach, H., Smirnov, S.A., Kulik, V., 1995. Analysis of a spherical crack on the interface of a two-phase composite. Mechanics of Composite Materials 31, 17±26. Collins, W.D., 1959. On the solution of some axisymmetric boundary value problems by means of integral equations II. Further problems for a circular disc and a spherical cap. Mathematika 6, 120±133. Leblond, J.-B., Torlai, O., 1992. The stress ®eld near the front of an arbitrarily shaped crack in a three-dimensional elastic body. Journal of Elasticity 29, 97±131. Lur'e, A.I., 1964. Three-Dimensional Problems of the Theory of Elasticity. Interscience, New York. Martin, P.A., 2000. On wrinkled penny-shaped cracks. Journal of the Mechanics and Physics of Solids, submitted for publication. Martynenko, M.A., Ulitko, A.F., 1979. Stress state near the vertex of a spherical notch in an unbounded elastic medium. Soviet Applied Mechanics 14, 911±918. Popov, G.Ya., 1992. The non-axisymmetric problem of the stress concentration in an unbounded elastic medium near a spherical slit. Journal of Applied Mathematics and Mechanics (PMM) 56, 665±673. Porter, D., Stirling, D.S.G., 1990. Integral Equations. University Press, Cambridge. Prokhorova, N.L., Solov'ev, Iu.I., 1976. Axisymmetric problem for an elastic space with a spherical cut. Journal of Applied Mathematics and Mechanics (PMM) 40, 640±646. Rice, J.R., 1989. Weight function theory for three-dimensional elastic crack analysis. In: Wei, R.P., Ganglo, R.P. (Eds.), Fracture Mechanics: Perspectives and Directions, Twentieth Symposium, American Society for Testing and Materials, Philadelphia, pp. 29± 57. Sneddon, I.N., 1966. Mixed Boundary Value Problems in Potential Theory. North-Holland, Amsterdam. Whittaker, E.T., Watson, G.N., 1927. A Course of Modern Analysis, fourth edition. University Press, Cambridge. Ziuzin, V.A., Mossakovskii, V.I., 1970. Axisymmetric loading of a space with a spherical cut. Journal of Applied Mathematics and Mechanics (PMM) 34, 172±177.