2. ech Py Sai, 96, Vo 160 Peon Pa, Pine rt Btn, SINGULAR BEHAVIOUR AT THE END OF A TENSILE CRACK IN A HARDENING MATERIAL* By J. W. Hvrcumsox ‘Harvard Univety (Recioed Mt Juy 3007) somune Durranumions of sre oomaring at the ip ofa era in tension ll represent for both ‘lane str and plane strsin. A total deformation theory of plait, in eonjunction vith two Fisdenlng stress-strain rations ated, Forapplod ta niiesty low uch hat the platiezone invery small lative tothe cack logth, the dooinant sings can be eompltly determined ‘wih the aid of «pathndependent ine itera ecanty given by ce (1087). The amptude of {he tensile srs singularity heed of the erack found tobe lrg In plane strain than in lane 1, Iemmopvertox ‘Tim nacxonouxD for the present study has been surveyed comprehensively by ‘McCumirock and Inwis (1005) and more recently by Rrcs (1967). Our investiga- tion is concerned with the plastic deformation at the tip of a crack in a twodimen- sional stress field which ie uniaxial tension in a direction perpendicular to the crack far from it. We have limited consideration to only the zone in the immediate ‘neighbouthood of the erack tip in which the plastic strains are large compared to the clastic strains, Nonlinearity is introduced in this study only through the stress~ strain relations which are chosen to model certain aspects of the elastic-plastic ‘behaviour of common metels. Equilibrium conditions and the strain-displacement ‘equations are taken to be linear. Results for plane stress and plane stzin are contrasted. 2, Dnronmariox Tnzony Equarioxs Fox & CRAck nv 4 Staane Hanpexrso Marentat ‘The fist representation of nonlinear elastic-plastic behaviour we will consider of the type suggested by Rasamxa and Oscoon (1943) to model the tensile stress-strain relations of certain metals, namely eno tat, a ‘Throughout this pape, unless otherwise stated, stress quantities if unbarré will be non-dimensionalized by a yield stress 9y and unbarred strain quantities willbe aormal- ined with respect to the corresponding yield strain é = o4/B where Eis the initial de vo at wget opty he Abn Hem Projets Any ner Cnr SDA pt ten haa ed Sp ation er st Sina byte en of Rage tk ‘ova Paps Harerd sve wu J. W, Horemoe slope ofthe stress-strain curve, Tensile curves with 2 = 0.02 eomesponding to a. ‘usual engineering definition of yield are shown in Fig. 1 for several values of the strain hardening coeflient n. “Applicability uf (1) i rotrctal Lo aouotonically increasing stress or, as more commonly stated in plasticity theory, under the condition of no unloading. Only the simplest total deformation theory will be employedin the present study. Plastic deformation is assumed to be independent of the hydrostatic component of ‘the stres, $044, and, further, is assumed to be completely determined by the fist invariant of the strest deviato ay = oy — Fone By. @ I is convenient to introduce the invariant in the form of the ‘effective stress" (e defined by 8 = Feyny @ ‘Thus, in simple tension oy = 5 and therefore the Mises yield condition auxiliary to the theory is ¢¢ = 1 oF By = ay (ofcourse a smooth stress-strain representation such at (1) admits only an effective, or approximate, yield condition). ‘The generalized stress-strain relation which reduces to (1) in simple tension is a 2 tou + ™ opp by + ny where» is Poisson's ratio. ‘Since unloading must be excluded for this relation to retain any validity, any- non-steady solution must be checked in retrospect to insure that og at every point: {snon-deereasing, Solutions must also be checked to determine the extent to which, the deformation history at every point is proportional-—that i, the extent to which the stress components remain in a fixed proportion as the deformation proceeds. eis well known that deformation theory is inadequate where there are large ‘departures from proportionality; but when this does not occur this theory can be ‘considered no more objectionable than the corresponding incremental flow theory (BvoraNsuy 1950). ‘The appliation of this theory to the emack problem will be Aiscussed at a later point.Singular behaviour atthe end of «teal rack in a hardening material 15 ‘The complementary potential energy functional, appropriate toastress boundary value problem and to be specialized to either plane stressor plane strain, is J{sa trees 1am In either generalized plane stressor plane strain, equilibrium is ensured forall stranca derived from a etrese function by arty, © og = = A8Y where cylindrical coordinates r and O[( J’ =2/2r and J =2/20] are centred at the Fight end of the erack as shown in Fig. 2. The non-dimensional stress function ¢ tnd coordinate r are given in terms of the dimensional quantities (barred) by where L is the half length of the rack. Tn plane stress, dealt with fst, the stress components acting on the plane parallel tothe plate are neglected. ‘The effective stres is 08 = of + ob — or 4 + Borat ) while the strains are given by v09 + 200% (or — bon) ea = 04 — v0, + 209" (09 — For) o $e) ons +8 002" oe ‘The partial differential equation governing the stress function (under the restric- tion of no unloading) ean be obtained by eliminating the strains from the eom- patibility equation Phra)" tre! he! art lenin) =o or, alternatively, can be obtained as the Kuler equation sociated with the fst equst6 JW, Horeaoion variation of the complementary potential energy.* ‘The result for plane stress ist Vb E(t[oe re" Ff =A EN] 4 orto tr egy] Serlfoet (— art g arts" 4s) tte t(— 9" + artg’ +artg yy} mo. ® ‘Fora stes-fre erack the boundary conditions on the ack ca be taken as bao mo. (9) 8, Sivovian Beuaviour ar si Crack Tre A. crack in afar field which is tensile in « ditection perpendicular to it is consi- dered, A solution to (8) is sought in the immediate vicinity of the erack tip where the stresees are ange, and, on the basi ofthe theory which has been laid down, are ‘unbounded ar the erick tip is approached. An asymptotic expansion of the soli ton is attempted in the form $= HHO) +H) +... (19) where, if the frst term is to be singled out at the dominant one, # <¢ ete. Our search will be restricted to only the dominant term in such an expansion : Kr gO) ay ‘where explicit display of the amplitude K will permit us to adjust the amplitude of Fin some arbitrary, yet appropriate, manner. ‘Except for its amplitude, the dominant term is determined entirely by the nonlinear terms in the governing equation (8). Formally, this follows from sub- stitution of the expansion (10) into (8), oF by a comparison of the elastic (quadratic) fand plastic (n +1" power) contributions to the complementary energy functional (), For any « < 2, one ean choote a suliciently small neighbourhood of the erack. tip such that the elastic energy is an arbitenrily small fraction ofthe plastic energy. (On physieal grounds the energy in any finite zone containing the crack tip must be finite. ‘This impotes the restriction, easly obtained using (8) and (11), that tn va (2) ‘With the biharmonie term omitted, (8) is homogeneous in both ¢ and r (and deriva tives with respect to r). For this reason, an exact separation of the nonlinear terms in (6) is accomplished by the term assumed (11). The resulting equation, whichis homogeneous in 4 and is associated with homogeneous boundary conditions, is in the form of an eigenvalue equation for # [ne 35) forte — 96-26] + {ule =) + 1}m —s)}oe-4{6(08 9 $F} $6{n (0 —2) +1} —1) (@ragy =o as) ong ee patos ens fc wh eng th ein of un etn or poe sag heh ee ig ne saat ate en fen yiSingular bebviourat he end of tn rack ns hardening mater) — Kr¥4.96(0) = Kr0¥ (02 + ont — 056 + BO and oy = Ket 80, (0) = Kot (oh +8) 04 = K-45 (0) = Kit-ta(s 8 or9 = Ket 9 (0) = Krad — 2) 8. "The stress feld surrounding the exack willbe symimetsic about @ = 0 and thus $ (0) wil be also. Stressfee boundary conditions (0) requte Slam =F len on, with the imposed symmetry, $” (0) = # (0) = 0. ‘The nonlinear eigenvalue equation (18) and its counterpart for plane strain, equation (81), were solved numerically using an iteration scheme described in the Appendix. Rather remarkably the calculated values of s for various integer values ff n for both plane stress and plane strain were found to be given to within one tenth of one percent by the simple formula where (4) as) ‘As a consequence of this relation, the energy density varies exactly 0s 1/r as the crack tip is approached. A direct proof of the validity ofthis result ean be obtained with a path-independent integral recently given by Ricr (1007), This demonstra- tion is given in the following section Sed ra. 8 (a) and (2). Gavan of stress at crack ip for plane sees,8 J..W, Hevennetos ‘The 6-variations ofthe three stress components and the effective stress associa- ted with the dominant singularity are shown in Figa. 3a and b for n= 3 and 18, ‘Typically, then, the effective stress is given by oe = Kr #5 040) (19) where, as seen from the plots, the maximum amplitude of § is fixed so that 3, attains the maximum value of unity. Note that 2g does not vanish for any value of ‘and therefore, neglect of the elastic energy density is justified for all @ as wel. ‘The amplitude K ie undetermined and must be related to the applied tensile stress o*. In general, thie would appear to be a rather formidable task entailing a matching procedure to connect the far fld to the solution in the regime in which the elastic and plastic strains are of comparable magnitude and then in turn to the solution at the erack tip. Tt might be feasible, under eertain circumstances, to make tse of finite diference methods in much the same way as SwaLow, Wautrays land Yano (1966) have done but to incorporate direetly the form of the singularity. Te happens that itis not necessary to resort to these lengths when the plastic zone is very small in comparison with the erack length. In this ease, application of Rice's path integral yields the amplitude, K, with recourse to only e relatively simple calculation which completely by-pasies the necessity of considering the range of comparable elastic and plastic strains 4. Apuiropn or Thr Dosasaxr SiNoULAR Temi or Swazs Seats: Ynepine 'PLaN® SrRESS* [As a preliminary to the ealeulations in this section, we now demonstrate that ‘when yielding occurs on © small scale (that is, the applied stress is sufficiently low to ensure thit the plastic zone is very small compared to the erack length) the deformation is governed by a set of similar solutions. Conditions for no unloading ‘and near-proportional losing, on which the applicability of deformation theory rests, are discussed. ‘The leading singular term in the purely clastic solution for a cracked plate subject toa tension a” = ayo" is Sir (oo b0 + Sesto) an, For plane stress the effective stress easily found to be ce Src 48 + nt (as) ‘The boundary value problem for small-scale yielding is posed in the following man- ner, Ttis assumed thatthe applied stress is sufficiently low to ensure thatthe plastic ‘one isnot only very small compared to the crack length but is also embedded in an Clastie field which is dominated by the singularity of the elastic solution, For Small-scale yielding, then, ¢ rust sutisfy the fall nonlinear equation (8)and approach the funetion (17) of the elastie solution esr beeomes large, te tan igh nna Rot OST Pea pet dn sepine ocaen by cue a me ahSingular behaviour atthe end of «ten crack in @ hardening matedal = 10 ‘The following non-dimensionalization collapses the st of solutions coresponding tw different values of ow ou sel of sindlar sulutious ‘The equation governing do is again (8) with 4, vg and r replaced by 4, ow and ro, respectively. Far from the erack tip tom Jar ood + dom 8. {In terms of the similarity solution the dimensional quantities are given by, typically, m0) rato) Mo00l 4) f4- {1'no unlonding isto occur with monotonically increasing it is necessary that. 7m (fo 9) be & non-decreasing function of reas rg approaches zero from infinity for each Yalue of @. While a complete check ofthis condition must await an inspection ofthe full solution, ane might expect with some confidence that it will be satisfied. Jn much the same manner, proportionality or lack thereof at any point depends on the extent of proportionality on the corresponding radial ray of the similarity solution, We retum to this discussion at the end of the paper. ‘The boundary value problem posed for small-sale yielding is, in eect, the first term in an asymptotic expansion of the full solution about the zero spplied stress level. Tt is not clear how large the applied stress may be such that the one term in ‘the expansion remains agood approximation. Some insight into the range of validity can be obtained from an examination of the solution of the simpler problem of an lanti-plane shear erack which has been studied in detail by McCurvrocx (1958), Nevazx (1961) and Kosxnxxx (1008). Ricx (1960), in particular, hus oked closely at this solution and has suggested that the small-scale yielding assumption remains valid for overall applied stress below one third or at mast one half the yield stress Most certainly, the small-eale yielding assumption loss its validity at loads near ‘the fracture point of temi-brittle or ductile metals. For such materials dissipation of energy in plastic deformation is comparable to the elastic energy release and the small-scale yielding approximation will break down, as will be clear trem the dis- cussion which follows. Tn the remainder ofthis section use is made ofa path-independent line integral secently discovered by Rice (1967) (in cartesian coordinates) + I= | OPay — ours 0d) 09 where W is the strain energy density for a linear or nonlinear matetil (with no discontinuous stress-ttrain behaviour such as unloeding) = we fave (0) Saxons (1000) has given a related path-independent integral for thelinear elastic2 3.1, Horennon case. Rice has shown that the line integral J is zero when evaluated about any cloged cireuit which encloses no singularities on the basis of any solution to the ‘stociated equations of plane stress or plane strain. When evaluated about any closed circuit which encloses the erack, it assumes a nonzero value. ‘This value can Deidentfied with the rate of energy release per unit extension ofthe crack when the prescribed loads or displacements are held fixed. Since ths result is for « material ‘which undergoes no unloading, it lose its physical interpretation when applied to ‘common metal. Nonetheless, because ofits path-independence, it proves most ‘useful in the analysis of the stationary crack. Consider the olosed circuit I'= Ty —Ts + Ty + Ts in Fig. 2. Contributions to the integral along the erack vanish; Uhus since the total integral vanishes, the integral about I is identical to that about Ty. If T) i taken entirely inthe slastic region and if small-scale yielding is assumed then this integral can be evaluated. [ircr, (1967) has evaluated this integral for several loading eases on the basis of the purely elastic solution (con the assumption of small-sale yielding). Hehas shown that Inwin’s (1960) results [rederived by Bauzcxxen (1088) and Sawpxns (1960)] for the energy release rate are obtained. For the tensile crack under consideration (Gn dimensionless quantities) ey J ovty = eum ace) = no ‘Now, the radius of the path is chosen such that it lies within the zone dom- inated by the fully plastic singularity (11), and tis integral is evaluated in terms fof and the unknown amplinide K, This redution i fay straightforwaré. One finde we and oy ny tae = KO 8-016 O [8 (Op =) — ore (Me +] | 4: (n(¢ 2) +1) 0088 [Or + Bre Oa] ‘The radial and tangential displacements, up and up, are readily found wsirg any ‘oof the strain displacement relations: oe a 8180) = a PPE G2 +E sig men ym ety (6) = OK mte8 Loses as 2) $ — $2} — te. ‘The final form of the path integral on Ty with udu is W makes fray — ayn 2) = aot nines cn where f 1 f (aac — [an 6 0 — 8) ~ 80) “4 Pees) ramen dpa 08Siagular behaviour atthe end ofa tena crack na hardening material = 20 ‘This integral must equal the value obtained from the F integration, thus aKeii yinenieayt, [= aoe? a and this condition ean only be met (as r2-+0) if ett ati" ‘rhs demonstrates the result found approximately by the numerical ealeulation discussed enlir, ‘A re-examination ofthe above derivation shows that (26) wil also hold when the plastic zone is not necessarily very stall at Jong as the integral about 1’, does not vanish. Ite generel validity is reinforoed by the numerical solution of (18), which ‘was derived under no restrietion on the size of the plastic zone except thet no ‘unlonding occur at the crack tip. ‘One other important piece of information can also be obtained from Rice's integral—namely, the amplitude xo wr, sen from (20) canbe elated sola rom the dominant singly Silas elu of fan (71) oe preted in abl or soe ve of 9) on ‘Tanux 1 a a 5 ° 1 1 eo oa 208 ae oun 008 08 ma rd 20 an 1 oa so. +0 40 oom ones ome hardening coefficient n and an extrapolated plot of Fis given in Fig 4. Note that the strain energy density is given by west 8) nd recall that 44 takes on a maximum value of unity. ‘ew be most appropriate to discus theae results after the plane strain cleule- tions have been reported, Details of the enslogous calculations for a crac in2 5. W, Horemowson Zé Fro. 4 Value of Fax dete by equation (2, tension feld under the condition of plane strain are much the same as those given in the previous seetion and only the major points are recorded. 15, SiNovian Bunavioun 1 Piase: Seas ‘The equation governing the stres function [defined, a before, by (6)) has a more complicated form than inthe previous ease and willnot be given. On the other hand, ‘the equation governing the dominant singularity i, if anything, slightly les lengthy than its plane stress counterpart and is found to be ad 28) frag erg 9" [- 2D] forse rte 9] +4rt[roet gy] =0 (29) where now of = Nor ant 4 300 ‘hi formula or the efestive ste sony caret in the singular zone where the ‘lst strains can be neglected “Baton (2) can be acparste it i taken to be $= Kr go) (30) ‘The eigenvalue equation for «and gis [% = mle —2{n(e—2) +2) [eee 9648) + 4(e—1){ne —2) +1} 0S) = (a) og = Kr a = Krt-*[4 (8, — ae)? + B04]! ‘and dp, 0p and dye are defined in (14). ‘As discussed previously, this equation was solved numerically subject to the ‘boundary conditions F) =F" (0 where Boa an+t net ‘and @-variations of the stresses are shown in Figs. 5a and b. Similarly,Singular behaviour atthe end of «tml crack i hardening atrial 2B faa om & $F Fic. 5(a) and (), @varation of steaks at erack tp fr plane stain, For small-scale yielding, application of Rice's path-independent integral gives 2 Gent “ (the integral about 1 for plane stein is (1 — 4) is again given by (24) and for plane strain ‘our nondimensionelization). Om Fj OF +H}, (0) = (0 ~ 8) oe Values of fend (1/1)! arelisted in Table 1 and J i plotted in Fig. 4. Asin the previous case, has been normalized so that 2, assumes a maximum value of unity. Finally, the strain energy within an small radius r in the fully plastic zone is given by ey procera PPrnleyeniere and values of & sag few “ are recorded in Table 1 for both plane strain and plane stress.% 3. W. Heronnson 6, SiouLan Brxaviour ron 4 Paecewise-uncan Sruss-Sreain ReLariox ‘A somewhat less realistic representation of the tensile stras-strain behaviour ‘of common metals than the Ramberg-Osgood relation is the piecewise-linear rela- tion shown in Fig. 6, This approximation does model certain features of plastic flow. And since the singularity dominating the behaviour at the erack tip in this ‘ro. 6 Plena str-sz ration casecan be obtained in simple form withno numeral computation, tisinteresting to contrast such renulta with thse based of the Ramberg-Osgood representation. “The stress-strain relation according to deformation theory is easly obtained from the tale tres-atrain curve of Fig. 8. Continuing withthe sume notation, wwe ind y= (1 + oy ~ mp9 By +40 — 060g where =1(&-1) tama (ss) and Aso ft ch ‘The complementary energy functional appropriate to a stress boundary value problem is J{se ened +E eee +} oe mrp ad on) Plane stress calculation In a yielded region where o, > 1 the equation governing the stress function is Bowe -aa{[eS-3) te Flor rd" 4 Pg" Fort fortrAANT) <0 8) where on given by (6. Proceeding bear, shtion is sought inthe form bo kn$ and cy = Kr? o,is introduced. Ttis anticipated that « will be les than ? and thus, the dominant singularity is governed by the bihaemonic term in (88), vgno. (20) ‘A solution to this equation appropriate to our purposes (with = g° = 0 ont w together with a symmetric stress distribution) ie = Kel (cos $0 + Fo), () Arts arty — 9")‘Singular behaviour at the end of tensile ernck in hardendng ma 28 ‘Ther and @ variations of the stresses derived from this olution are the same (except: {or the amplitude K) as in the purely elastic solution, and in partioular cum Kr (cost b+ fiat @ Contrary to the assumption implicit inthe derivation of (89), is aro for the isolated values @ = 0. ‘The path-independent integral (19) is evaluated using the dominant singular solution (40). Tt is found to be P72 — au my ade E are E Bao ‘hich, for small-scale yielding, must equal ro" and thus 3.69 « Xt By = he ply clare etree ai oviuly mus Acer tng hie pnw inact, hen, the ster tthe ip re eae ciel year clay bt ec by cor Tsing fo mate th sna egy dena Ue ly plat gine ssntopenin of i dea sane peed by he pry cate sotto namely, W = trto** (cost $0 +B sino) (a) ‘and the energy within a small radius 12 is [rte | We =e ron, s) Plone strain caleulation Here, again, the dominant singularity is governed by the biharrconic, The ‘appropriate solution is given by (40) and, except for the amplitude, the in-plane stresses are the same asin the linear elastic ase, ‘The ellectve stress for plane strain and the piecewise-linear relation is OF = (1B +B) (or — ot + (1 = $8 +4) op 09 + Boye where ra26 JW Murenteon It is easily seen that o vanishes for @ = 4 and also for @=0 when = 4 (ost very neely is when E/Eyqy > 1). Thus, this solution may not be valid at these points but, as diseussed above, it should be accurate for most 8. Assuming this to ‘be 20, the amplitude for emall‘scale yielding ean be obtained as ated ee ee Ea ‘As a check, noe that when y=} and B/Byy =1, =} and K assumes the rely elastic value 1 K vi But more important, when B/Bgs > 1, B x0 ve (Fe ‘Thus, the stresses atthe tip are the linear elastic ones times the factor [using (48) notes S00) fa ea aera gi acopun meen ee es el eect etter tae aa ak un ng aie te oo rn mee i eer ai mr patil ey ed eu te ay, ea ‘and the amplitude is given by 5) 17. Iurticavioss oF te Sraess ANALYSIS Certain details of the stress distribution et the crack tip depend rather strongly ‘on the model chosen to represent the material, while others do not. For example, the Piecewise linear ealeuation indicates an variation ofthe stress; and the Ramberg “Osgood relation predicts an rl"? variation ag the erack tip is approached. (On the other hand, the predictions for the energy density at the crack tip are rela- tively insensitive to the model. One should not lose sight ofthe fact that the domi- ‘nant singularities determined in this paper are only accurate representations of the solutions ina very small portion of the plastic one. Until the solution for the entire Plastic zone and surrounding elastic region is known one ean only extrapolate the ‘character of the singularity into the region of comparable elastic and plastic strains, “Two formulas emerge from the prevent analysis which have been shown by Bare as, Brapstiaw and Rooxx (1064) to correlate reasonably well with tests on thin, ‘racked aluminium sheet, Hateman ef al, measured through-thickness thinning ‘which occurred inthe plastic zone at the erack tip. A formula suggested by Dixon ‘and Sraaxsteax (1069) for the strain component normal to the sheet, namelySingular behaviour atthe end of «teal ruck Ins hardening material = 27 (ENS: we ‘was found to give a fairly good ft for the strain ahead of the erack (Bye is the Secant modulus). Even better agreement was achieved with a formula suggested by MeCuivroce (1061) oN co) ‘tis a simple matter to show thatthe strain just aad of the crack as predicted by ‘the piecewise linear theory of plane stress for sinal-scale yielding is §-- Ale) @s ae ‘esinity boven (dale rm hc at Hg ca be Weed mi Ratrlageseuse Secondly, the plane stress predictions for the Rambeng-Osgood relation ic {apart from a constant which ean be obtained from the numerical results), sey) ” pee are eee perry eens ete ret es eerie eel gry ae per eee perro Se eee ee Fee eee ee ede (60) ‘ro. 7. Ratioof plan srsn amplitude of tena ten ingulaety ahead crc to planeatret sempltde.