SOME GEOMETRICAL RELATIONS IN DISLOCATED CRYSTALS* J.P. NYEF When a singe erytal deforms by tide which i unevenly distributed over, the gle surfaces the Iattca bacon curved, The constant feature of distortion by glide on a viagl esto! planer i that the orthogonal trajectories of the deformed glide planes (the e-axes in hexagonal meals) are straight lines. ‘This leads t0 the conclusion that in: polygonisation experiments on single hexagonal n rvsials the palveon walls ae olanes, white the lige planes are deformed into euliders whowe sec Ste 'the involites of a stn curve, The analysts explants West's observation that the e-ancs eeitale feared ar stat ine For doe ge gn two orthogonal sto pares there -gomplete analogy between the geometrical groperties of the distorted glide planes and those ofc “lignes in He mathematical theory of pasherty. More general eases ave iscused and Forraulae arederived connecting the density of dislocations wth the lattice curvatures. Fora threesimensinal network of dislocations the state of dislocation ofa region is shown tobe spied by a second-rank, Ceier, which fas properties tke cone ofa sires Leon exvepe ta te a) nue a QUELQUES RELATIONS GEOMETRIQUES DANS DFS CRISTAUX DISLOQUES: juand un monocistal est, déformé par alissement, qui n'est pas uniformémentdistribué sur les surfaces de gliasement, le réseau devient courhe. La caractéritigue canstante de distorsion par Inccmont oars soul groupe dpe coy sus fs trajesteesorthoponaios des plans de gsoment Stor tes ape den eta esgic) unt de rie, Ces cond & i coeinon, aue lan les ensais de polyonina on suc des eomoerintan: des métas hexagonats es faces mes ‘ont planes. alors qe tes plans ce pliswmment sont aéformes.en des eines. dnt lex sechons sont ‘des dlveloppantes dune seule cousbe, L’analyseexpique observation de West, que les axere di des cristae de corindom sont des dete Poa le gisement double se dein grapes orthogo ing ya une analgie compte entre lem porte tomers des plas geet més ot cele ds lignes gssement dans la theore mathématique te fa Pater’. Des eas ps aéndraus sont cca et de formes jigrant fa denice den dnoeations ax courures dre Son déduten, Ist monte€ que porn rea de dloratons 4 tris simensone eat de does oem end pha dene sen fae woo et ‘da tenseue de tension, & exception du fait, qu'il west pas eymetrique. EINIGE GEOMETRISCHE BEZIEHUNGEN IN VERFORMTEN KRISTALLEN Wenn Einkristalle durch Gleitung verformt werden, und die Gleitung sich ungleichmassig uber die Cleitebenen vertallt, dann wird das Keistallgitter “werbogen.”” Eine der immer widerkehrenden Hegleiterscheinungen der Verformung durch Gletung im Fall einer einzigen Translationsebene is «dass die Orthogonaltrajektorien der verformten Gleitebenen (in hexogonalen Metallen die c-Acheen} Gaten sini Paraiie kann man sehlioen, dase in Palveoniitionsesperinenten an hesopanalen Einkristallen die Wande der Polygose Ebenen sind, wahrend sich die Gletebenen at Zylindeen verformt haben, deren Schnitte dle Evoluten einer einzigen Kurve sind. Diese Analyse erklirt die Beobachtuny ‘vor West, dass die ¢-Achsen von "gebogen ren” Kornd. Kristallen Graden sind, Im alle einer doppelten Giestung aul zwer aut einander senkrechten Gieitebenen besteht eme vol Re tag toes bee cities cones Gatco on ae OE Formeinabgeteltr, de dle Diche der Verseceungen mit det ese dam in cine dredimonasnaler Netoverk von Verstingen, der Vere tines, Bereiches als Tensor zweiter Ordnung dav ‘de> Spannungstensorsaalich 1, Introduction. Suppose that a single crystal deforms by sitp on a single set of parallel planes in such a way that the amount of slip is unevenly distributed over the slip planes. It may be asked: What possible geometrical forms can the slip planes tale? This is a question that arises, for example, when a single crystal wire of zinc is bent plastically to study the phenomenon of polygonisation [1]. We do not deal in this paper with the curvatures on an atomic scale but only with the average curvatures over distances large compared with d, the dislocation spacing, “Received November 7, 1952. ip fBel Teephove'Laboratoris, Murray Hil, New Jersey, ACTA METALLURGICA, VOL. 1, MARCH 1958 ‘in dee mathematischen Theorie der Plastiitat, Allgemeine Falle werden diskutiert und eqn des Gir verte Es win stant elle werden kann, dessen Eigenschaften denen nd der jedoch nfekt aymmetriach i, If the distribution of dislocations could be pre- scribed arbitrarily, there would be no limitation on the curvacures of uve glide plus. Au atbiviaty istribution of dislocations, however, gives a large scale distribution of internal strese. As an extreme example of this, if one calculates the stress at a sn point due ta a uniform array of positive edge locations one finds that the stress can increase to infinity as the array becomes infinite in extent. We show here from geometrical considerations that it is possible to have non-uniform distributions of dislocations that do not give accumulating stresses of this sort, Such arrays correspond to minimum strain energy configurations. In them the residual stresses average out over distances large compared to d. This property of the arrays may be statedat precisely as follows: (A) If a path is constructed through “good” crystal in the sense of Frank [2] from successive lattice translation vectors, the length 1 of any section of the path is equal to the undistorted length, provided 1>>d. The shape of the path will of course be changed by the distortion. ‘The significance of assumption (A) is easier to appreciate after the analysis that follows from it has been read. When real ctystals are distorted plastically: they do in fact contain large-scale distributions of residual strains, which contribute to the lattice curvature, However, the larger the amount of plastic deforma- tion the smaller, relatively speaking, will be this contribution to the curvature; moreover, when a distorted crystal is annealed the large-scale distri- butions of residual strains may be expected to disap- pear, while the lattice curvature will still remain. In this analysis we neglect entirely the curvature attributable to the large-scale distributions of strain, In §§2, 3, 4 we discuss single glide. This leads to the consideration of more general cases where multiple slip occurs, and then in §8 we turn to the most general problem: the curvatures of a crystal lattice produced by a three-dimensional network of dislocations. 2. Single Glide: Two-Dimensional Case It follows from assumption (A) that in single slide the distance between any two successive glide surfaces, measured along their orthogonal trajec- tories, is the same at all points. Consider first a ‘twordimensional case (Figure 1). 51 is the trace of a FIGURE 4. Single glide: rwosdlmensionat case, The glide surfaces 5; and Sy are bath involutes of cylindrical glide surface, whose centre of curvature at P is C. As P moves along Si the locus of C is 2, the evolute of Si, S: ean be generated by unwrap- ping a taut string wound round 2 and noting the lnenis of a given paint P. Any other paint Q an the string will sweep out a curve S:, and, since PQ is fixed, S: evidently satisfies the condition for being ACTA METALLURGICA, Vou. 1, 1959 another glide surface. (For the sake of rigour we should also remark that it may be shown that when PQ is a small element, St is the only surface lying ata fixed distance PQ from Sy measured along the orthogonal trajectories. It then follows that the same result is true tor PY finite, Huygens’ construc- tion in optics is somewhat analogous.) The glide urfaces are thus the involutes of 2, which might be called the generating curve for the deformation. Hf the gener to the crystal are given, the deformation is com- pletely epecified. (The sense of the generating curve rust also be given.) the construction shows that the orthogonal trajectories of the glide planes, which were straight lines before the deformation, are also straight lines afterwards. They all have the same crystallographic direction, the caxis in hexagonal crystals, although they are not parallel. Now, it a crystal is heated and, made to polygonise (1] after a distortion of this type, i has been shown theoretically that the dislocation walls should form at right angles to the glide planes, for that is the position of lowest energy. Our con- struction shows that, apart from elastic strains, the traces of the walls will theoretically be strictly straight lines, even on a macroscopic scale. Thus, for example. when a hexagonal ervstal is bent by basal glide so that the plane of bending contains its c-axis and one of the basal glide directions, the glide planes will deform into cylinders with parallel generators and the polygon walls will be planes. We shall call this case plane bending. If the shape of any one glide surface, say Sy, is given, successive glide surfaces can be constructed ‘on each side of it by drawing the normals to S: and connecting points at equal distances from Sy However, when a point is at the same distance, measured along the normal to S;, from more than cone point of S; the construction becomes ambiguous. Such points are associated with a cusp on the gene- rating curve. Physically, what has happened here is that flexural glide has become concentrated into a sharp kink. Figure 2 shows an instance of this. ABCis the generating curve with a cusp at B. In the ‘area to the right of DBA the glide surfaces can be constructed by imagining a taut string to be un- wrapped from the branch BA. Each point on the string sweeps out a glide surface. In the area to the left of DBC the glide surfaces can be similarly constructed by using the branch BC. There is thus an ambienity in the area within ABC ‘The amhieuity is resolved in the following way. Let EF = GH, where EF and GIf are normals to the glide surfac curve and its position in relationNYE Let the glide surface through F, constructed with branch BC, meet the glide surface through 1, constructed with the branch BA, at the point J J will then mark a sharp corner on the glide surface and the locus of J, shown by the broken line, will be a kink surface. It may be noted that the kink surface is not, in general, a plane. The fou of Uhe glide surfaces iar general cepre- sents a certain spatial distribution of edge isl: tions of like sign. (Close paira of dislocations of Ficuns 9, The affect of « step in the gooorating surve Flexural glide becomes concentrated into a Sharp kink. opposite sign produce no curvature and so are irrelevant to the present analysis.) If we look at the evolute construction in terms of dislocations we see that we have found, effectively, the distribution of least strain energy for any given base distribution along a certain part of the surface 5}, As mentioned in §1, this approach avoids any difhculties with mhmite stresses that might be caused by having an infinite array of dislocations. The method of construction automatically adjusts the stresses so that they do not accumulate, and when the density of dislocations necessary to achieve this becomes infinite (on 2) the array naturally terminates. The evolute construction focusses atten- tion on the avoidance of stresses normal to the glide planes, but it will be seen that it is the avoidance of other stresses that determines the curvature. Let n be the number of excess dislocations of one sign per unit area and b their Burgers vector. Then if we take a square Burgers circuit [2] of unit area, as shown in Figure 3, the closure failure is nb and, Ficus 3. ‘The curvature of the glide planes produced by ‘edge dislocations of the same signin unit area ts 8h by hypothesis, there is no elastic stretching of the cireuit. ‘The curvature xe of the glide planes in this region is therefore given by a ko = nb DISLOCATED CRYSTALS 189 where 6 = [b). Thus, on PC in Figure 1, m is in- versely proportional to the distance from C. We may notice that if 6+ 0 and n> = in such a way that nb is constant, the strain energy £0 but the curvature remains unchanged. Our analysis can only cover continuously varying distributions of dlistoca~ tions and for this reason it cannot predict the cerease of straint energy” kat accompanies poly- gonisation, It may be noticed, however, that the above distribution ensures that when the crystal polygonises the number of dislocations per unit length is a constant in any given palygon wall It is a simple matter to calculate the curvature of any other lattice plane in the zone parallel to the dislocation fines. In Figure 4a, ABCD is a small element bounded by glide planes, AB and DC, and their normals. Let da, da, . denote the orientation of the lattice at A, B, . . . measured anti-clockwise rom some hxed orientation. ‘Then the curvature of the lattice plane AC, which makes an angle @ with AD, i» x= Sade @) AC = ny c08 8 = nb cos 9. thus, the curcature of any lattice plane parallel fo the edge dislocation lines is equal to the component of by poratte! wv the plune. Phe cur vacure vf dhe lautive ® Fiouay 4 Curvatures produced by glide on (a) single set of planes and (b) two orthogonal sets of plane, illstra fing the relation é —3n = op — plane AC is to be distinguished from the curvature of a line inscribed on the crystal before deformation at the same angle @ to AB. Such a line changes its orientation relative to the crystal during the deformation, and after deformation different parts of the line will, in general, be in different crystallo- graphic directions. We turn now to a special case. 3. Special Case of Uniform Plane Bending If plane bending takes place uniformly about an axis O (Figure 5a), the generating curve ¥ is a circle, of, radius @ and centre O. ‘The orthogonal trajectories to the glide planes (the caxes for hexagonal metal crystals) are straight lines and the156 aide planes are the involutes of the circle. Using the notation that r is the radius vector to any point P (Figure 5b), r= re is the neutral plane, x is the angle of intersection of the glide planes with the fibre (r = constant) through P, and x» is the Fiovas 5, Uniform plane bending of a rod by single slide. The glide planes are the tavolutes of a circle, and the Bide plane noruals, the canes i Hexagonal wipstals, are Straight lines. angle of intersection of the glide plance with the neutral plane, we have the relations, pane eine patentee Vc—ae The same results may be obtained by noting that, after bending, the length J of a fibre which was originally of length fy is given by (On the other hand, the change in lattice orientation caused by this extension is expressed by [3] 1 sin x0 1 sin x Hence r sin x = ro sin x9 = constant, as before, The radial dictribution of the excess dislocation is, from equation (1), 1 n= =a er or, more simply, aot, - where b, is the component of b parallel to the neutral plane. ‘The circle r= a corresponds to a compression sufficient to turn the glide planes to a position at Tight angles to the fibres; it also corresponds to an ly high density of distovations. If dhe wd is such a way that the neutral plane remains fixed within it,-this imposes a limit to the curvature that can be applied. By using the fact that the volume is unchanged by the deformation it is easv ACTA METALLUKGICA, VOL. 1, 1993 to show that, when the innermost fibre is of radius a? = ralry ~ 28) where 4 is the distance between the innermost fibre and the neurrat plane before deformation. Usiag the relation a = r sin x» we find the minimum radii of curvature of the neutral plane and the innermost fibre respectively, @) cos xe In an actual experiment an attempt to bend a roel to a smaller radius than is given by these equations would result in an inward shift of the neutral plane relative to the rod. The bending would be accom plished, in fact, by a combination of bending with extension, Our formulae are independent of the history of the bending and of whether or not it is combined with extension or compression. rp refers always to the fbre thar ar the end of the deformation is found to be unchanged in length; on it, x = x». Formula (8) shows that, for a.given xo we can make @ as small as we like by simply choosing 6 small enough. If transverse straight Tines at right angles to the neutral axis were ruled on the erystal hefore hending, they would become curved. The calculation of their shape in, polar coordinates is straightforward but tedious. The result is given in Appendix A and 4, Experimental Verification ‘We have seen that in plane bending by basal glide in a hexagonal crystal, the crystallographic planes containing Ure casis and the axis of bending remain plane in spite of the bending. We may note that thio doce not imply that the correcponding X-ray reflexions will be free from asterism, because successive reflecting planes are not parallel. How- ever, with a transparent crystal the extinction directions between crossed nicols can be used to show up the effect. This has been done by West [4] with single crystals of corundum (trigonal). He found as an experimental fact, without explanation, that, in crystals bent round into a loop and a U-shape, “the caxis through any point remains a straight line even to the boundaries of the rod.”” He went un wy deduce dat the lide planes ia lhe uniformly bent parts were the involutes of a circle, as we saw in the last section. It may be conchided that these corundum crystals had deformed by single elide, Double lide would in general give aNYE itierent distribution of lattice rotation, as will appear in the next sections. 5, Double Orthogonal Glide: Two-Dimensional Case We now xeuealise dhe prublen of §2 by allow ing further slip to take place on the set of planes at right angles to the fret cet. Both a taken to be in the plane of the diagram (Figures 4a, b) and, as hefore, we only consider edge-type dislocations running perpendicular to the plane of the diagram. Rock salt is an example of a erystal that could deform in this way. Let Figure 4a now show a finite curvilinear quadrilateral bounded by slip-lines after slip has ‘occurred on the first set of planes, AB, DC, but before it has occurred on the second set AD, BC. Since 4 = do and $n ~ bey “ 0 ~ bn = 62 64 Now let slip take place on the second set of planes (Figure 4b), This produces additional curvatures but it is easily proved that it will not affect the curvature of AB (lor instance, by drawing a Burgers directions are Fours 6, Glide on two sets of orthogonal planes. circuit after the manner of Figure 3). Therefore, since AB is not altered in length, ¢s — 4 is unal- tered by the second st larly, @¢ — Op Is unaltered, Hence relation (4) continues to hold after Use second slip, Rearranging the terms gives to bn = 0 bn. ‘We thus find that the lattice rotation between two given slip-lines of one family where they are cut by a slip-line ‘of the other family ts constant along their length. Let AB and AD in Figure 4b now be infinitesimal with lengths ds and ds! respectively. The radii of curvature » and p’ respectively of the slip-lines through A are given by 1_ 8 1 From the geometry of the small quadrilateral, AB — DC = AD($s ~ 64)- ‘Therefore, if pp and p4 are valucs of p at D and A, DISLOCATED CRYSTALS 137 palo — 4) ~ pole ~ 40) = AD ba ~ 64) ‘Therefore, from equation (4), Pa ~ po = AD. Hence, a : Goan or dp + p'ié = along a tine ot the second set. Similarly, ae ae te eae along a line of the first set. This means that as one moves along the slip-lines of one set the radius of cureature of the slip-lines of the other set changes by the distance moved. ‘The radius decreases as one moves towards the inside of the curves. Thus, in general, the radius of curvature eventually decreases to zero and the slip-lines that are being followed run together (Figure 6). We have already had an in- stance of this in the special case of single glide. ‘The two geometrical properties of the slip-lines or glide lines expressed above are identical with Hencky’s first and second theorems in the mathemati cal theory ot plasticity [5]. On the other hand, the “slip-ines” in that theory are quite different from the slip-lines we have been discussing. They are defined as the trajectories of maximum shear stress in the strexs field, or as characteristics of the fundamen- tal hyperbolic equations of the theory. We thus have the enrions that not only are these two dissimilar families of curves called by the same name but they also obev the same equations. ‘The small arc process used in the theory of plasticity to build up the slip-line field in any finite quadrilateral ABCD, given the form of the slipsines AB and AD, is equally applicable here. ‘The answer to the question, then, as to what forms dhe stip planes cant vake after ylide on orttio= gonal planes is that they can be any families that satisfy the Hencky relations 6. Multiple Glide: Two-Dimensional Case Still confining attention to the two-dimensional case of an array of parallel edge dislocations we may now consider the effect of glide on any number of ent planes. If in unit area there are n, disloca- tions with Burgers vector b,, the curvature of any lattice plane parallel to the dislocation fines is obtained (§2) by summing the components of the vectors m:b, in the direction of the plane—which is the same as taking the component of the vector Lab, watinn, 9 cnr of mathamatical pun188 in that direction. At the point considered, therefore, the dislocations are equivalent to a set on a single plane, but of course this plane will be diferent, in general, for every point of the erystal, according to the ratios of the 1, 7. General Three-Dimensional Case for Single Glide In three dimensions it may be shown by similar reasoning to that of §2 that the curves orthogonal to the glide surfaces are necessarily straight lines, even for the most general type of distortion, pro- vided only that i takes place by glide on a single set of planes. Any number of glide directions may operate, However, in this case, unlike the two- dimensional case, the lattice orientation is not fixed along the normal, tor the lattice ean and does, in general, twist about the normal. In two dimensions the centre of curvature of Si at P (Rigure 1) was the same as the centre of curvature of S;at Q and so on, namely €, Similarly in three dimensions, all the lide surfaces intersecting a given normal share the same two centres of curvature lying on their com- mon normal. As we follow a normal, therefore, both radii of curvature change by the distance traveled. They increase or decrease respectively according to whether we are receding from or approaching the corresponding centre of curvature. Moreover, the ‘two principal curvature directions are preserved along a given normal. ‘The two surfaces described by the two centres of curvature as the normal moves represent the three-dimensional analogues of the evolute of two dimensions. For a sct of parallel like dislocations of gencial type there are simple relations (proved in Appendix B) between their direction and the curvature they produce in their glide planes. Figure 7 shows the Figune 7, Glide on a single set of plancs: three-cimen- sloal css ‘he ip plate sie lane E the agra. The principal directions of curvature beet the angle between the Alslocation ines and OW. relations between the various directions in a glide plane. The twa principal dirsctione of curvature in the glide plane are the internal and external bisec- tors of the angle, of magnitude 2a. between the ACTA MBTALLURGICA, Vou. 1, iss direction, OD. of the dislocation lines and ON. the normal to b. Thus, if the dislocations are pure edge, |, and the directions of the principal curvatures x1 and xy are parallel and perpendicular to b; while if the dislocations are pure serew, @ = x/4, and sz and ay are at angles of «/4 to b, The magnitudes of the curvatures are ky = nb cosa, a: = — mh sin* a, where m is the number of dislocation lines per unit area normal to the lines. Hence n= ap = 1b, and is independent of a. We also have eg bey = mb con Dat = 1b where 1’ is the number of lines per unit area taken normal to ON. 8. General Three-Dimensional Case Finally we come to the most general case that falls within the scope of this type of treatment, a three-dimensional network of dislocations. As before, we consider a region of the crvstal large enough for the eflects of the distocations within it to be aver- aged. Alternatively, we may make the distribution of dislocations continuous by letting & approach zero and increasing the number of dislocations of ‘each kind in proportion, so as to keep the curvatures constant (cf. §2). Our aim is to find the relation between die dislocations in the region and the curvatures of the lattice that they produce. Consider first the dislocations. They may be specified by taking Burgers circuits in different orientations and noting the closure failures (Bur gers vectors). Ifa Burgers circuit of unit area normal to the unit vector /, has a Burgers vector By (that is to say, By completes the circuit when it is traversed in the sense of a right-handed screw motion along 1) then it is shown in Appendix C that we may write 6) B= ayl, 7 =1,2,8), ‘The summation convention is to be understood ‘he components B, form an axial vector. The coefficients a1, relate the two vectors B, and [, and therefore form a second-rank tensor; it specifies the “state of dislocation’ of the region Yo calculate the «,, components for a given set of dislocations the procedure would be as follows. Suppose there are dislocations with length parallel to the unit vector r and with Burgers vector b. Lot thoro be # of theoe dislocations crossing unit area normal to t. The number crossing unit area normal to the sit vector 1 is mpl. The associatedNYE Burgers vector is b(nr-I}. Hence, in suffix notation, By = bears) and, from equation (3), aay = nbers. Jf there are other sets of dislocations present with different values of n,b and r, the total a, is obtained by summing the values of ‘nb, ry for all the sets. The form of ax, implies that, so far as curvatures are concerned, any state ot dislocation can be produced by combining nine sets of dislocations with their lengths and their Burgers vectors arran- ged parallel and perpendicular to the coordinate 2 Fiore 9 axes. The three terms of the leading diagonal represent pure screw dislocations and the six cross- terms represent the six possible types of pure edge dislocations. Alternatively, by combining the three types of dislocation lines directed along each axis the general state can be regarded as produced by three perpendicular sets of dislocations of general type. Such a representation would of course only Ue unique fur a given oct of ance We now have to specify the curvatures of the lattice. Let dé, be the small lattice rotations about the three axes, in a right-handed screw sense, associated with the displacement vector dia ‘The dg, are the components of an axial vector. If we now write ) dos the cooffisionte 1) form a vy dy wooad.ranle tensor which describes the curvature of the lattice In Append TD we show that the dislocation tensor ays is related to the curvature tensor #,, by the equation a where by = 1 for iJ, ‘The reverse relation is 8) ‘The general property of a dislocation network, that dhe Burgers vector is conserved along any dislocation line and at a node, is associated with a ‘corresponding theorem in the present formulation Consider the small rectangular parallelepiped shown in Figure 9. The faces define six Burgers circuits, ay A bise (ij R= 128) and: 8,0 for xj. fay aye ~ Bey ou DISLOCATED CRYSTALS and we specify that each is to be taken in the direction associated with the outward normal to the face. We now traverse all six in turn and show that the final result is that we have traversed a circuit of zero area. Thus (OCDBO + OBFAO + 0ABCO) + (GEAFG + GFBDG + GDCEG) CDBFAEC + EAFBDC! ‘Therefore the total Burgers vector for the six cir- cuits is zero. Equating the components to zero and proceeding to the limit we have Baus 4 0 axe Oxs and two similar equations. Thus, in general, ass ie 0 3a ax, @) = 0. The general analysis of the tensor aj, is therefore formally the sime as that for the tensor of mechani- cal otreco with the ono difference, that 1) io not symmetrical; equations (9) are the analogues of the equations of equilibrium It is interesting to notice too the formal similarity between equations (7) and the stress-strain relations in an isotropic body, ou = Ques + Murcer with 2¢ = 1, X= — 1; avy plays the part of the stress and.) that of tha strain, althengh, wnlike these tensors aj and «1, are not symmetrical. ‘These values of X and x give Young's Modulus , Rigidity Modulus = 3, Poisson's Ratio = 1 and Bulk Modulus Since 244 “1 ie wwe have ey ob. | Oe Ox, ~ GxjOx, ~ Ox, By substituting a for « from equation (8) we obtain the equations co) Sa ae Bere ae, Of the twenty-seven possible permutations of i,j # the equations given by thove nine with j trivial. OF the remaining eighteen equations, owing to tho symmetry in j and A, only nine are indepen dent. Thus there are nine independent differential equations t0 determine the nine independent a., components. By putting i= j or i= & in equations (10) a fare100) little reduetion shows that we recover the “equili- rium equations” (9). By analogy with the strain energy function in elasticity it is useful to introduce, purely formally, the function Wa dans. ‘Then, hy differentiating and using equations (7). aw. ay Bes, wars doy + Beye — 28 fous t any W is dhus a “dislocation potentials" It is a function of the components of lattice curvature at any point auch that ito dorivatives with roopect to cach component of curvature are each equal to the corresponding component of the state of dislocation. 9, Discussion Eshelby (6), using a model described by Timo- henko, has discussed a state of internal stress which Nabarro [7] has called a continuous distribution of dislocations. That this state of dislocation is different. trom the one treated here may be seen by considering the case of single glide (§2). One way of arriving at four model of Ure process would be to suppose au unstrained body cut up into small cubes of side d. “The structure is then bent by allowing sliding to tale place parallel to one set of faces only of the cubes (Boweal elie). the cubes themselves heing unstrain= ed. There will then be voids between the cubes. The strain of the cubes necessary to fill the voids is of order sd, where x is the mean curvature of the surfaces of sliding. As d—+0 with « held constant, this strain 0. In Eshelby’s model the strain does, not approach zero as the cubes are made small and he is essentially concerned with a state of internal stress, The difference between our models is that we allow stiding of the cubes or other elementary pieces, as is actually produced in crystals by the movement of dislocations, while Eshelby docs not. ‘The energy of a single dislocation at the centre of a crystal of linear dimensions 2 is af the general form oot nk, white G is an clastic constant and risa length of the order of atomic dimensions.* In the arrays of dislocations considered in this paper, where stresses do not accumulate, R may be taken as the mean spacing of the dislocations, say wm, or (0/«)!, "I am much indebted to the referee for most of the sub- ctamoe of the ramainder of thir discussion. ACTA METALLURGICA, VOL. 1, 1933 where « is the local curvature. The strain energy in a unit volume, therefore, which contains a = «/8 dislocations, is of the form «oxen (2) 2b 58 In Q) aS This is proportional to Gb la Ps = sib in, where m — 6"/r0", a constant that depends upon the details of the atomic arrangements on the dislocation axis. As b—0 with « constant, the enorgy approaches zero, as noted in §2. The analysis of this paper is closely related te Frank’s construction [8] for grain boundaries. In fact the Frank representation is essentially a surface distribution of dislocation density, or ax, and, although we do not give this development here, the Frank formulation is deducible from the present analysis. Grain boundary arrays are a special case of these more general distributions, and have the characteristic in common that they represent a state of mini is linear in the Burgers vector, or more precisely is of the general form noted above. The stability of polygonized arrays arises from the fact that the argument of the logarithmic term above really depends on the closest approach of the individual just energy and Uhat Uhe total energy lGvRE 10, Showing the form of transverse lines. after ulform plane bending of crystal barby Sle ide for dislocations rather than on their average separation, given by #7, Polygonized arrays preserve the mean dislocation density, as pointed out in §2, but they achieve a smaller minimum separation and hence a smaller total energy than uniform distributions. Appendix A The equations in polar coordinates of the trans- verse lines on the uniformly bent crystal referred to on p. 156 are, in terme of x a2 a parameter, = rosin xa cose x G— Fein 2x0 cot* x — cot x — x + const,The curves, which in general show a reversal of curvature, all have the same shape, dilferent mem bers corresponding to different values of the con stant in the expression for @. The curves for xu ==/%6, 3/4, #/8 are drawn in Figure 10. When 0 < x0 < 7/4, # has a maximum on the compression side at f= ry tan xp and a minimum at r= re When Xen a/4, 7 = re is a point of inflesion. When #4 < x0 < #2 P= ry is a maximum and 7 = rytan xo on the tension side, io a minimum. Appendix B* In Figure 7 we regard the angle NOD = 2a as fixed and we calculate the curvatures of the glide plane. Take a square Burgers circuit with sides of unit length whose normal lies in the plane of the figure at an angle 8 to UD. ‘The number ot dislocation lines threading the circuit is m cos 8. The closure fatlure is thus mb cus 8. The circuit makes an (2a — 8) with b; the component of the closure failure parallel to the circuit is nb cos 8 cos (2x 8) therefore. This component is a maximum or mini- mum when 8 = a and (a+ 2), that is, when the normal to the circuit bisects the angle NOD inter nally and externally. These bisectors then are the two directions of principal curvature. Reference to Figure 3 shows that the curvatures themselves are the two corresponding values of nb cos 8cos (2a — 8), that is ule x = mb cos’a and xs = — nb sin’ Appendix C* A proof of equations (5) is as follows, Consider the tetrahedron shown in Figure 11. Let ABC be of unit Fiore 11 area and normal to 1. Let a Burgers circuit uf unit area normal to Or: have a Burgers vector with components (aii, es; ay1)- Then the circuit OBCO has a Burgers vector (ays fy, ais Ji, asi 1). With a similar notation OCAO has a Burgers vector (ass I>. azz In, axy fa) and OABO has a Burgers vector (aus Ia, ata fa, a fh). If we now traverse these three ‘cuits one aiter the other, the arms O4, OB, OC See page TAR DISLOCATED CRYSTALS are each traversed twice, once in each directis and therefore cancel; we are left with the circuit ABCA whose closure failure must be the sum of the three for the separate circuits. Thus, for ABCA, Bes anh tals tals By = ani hy tay le + avs ls By = ax ty + ase fy + a bs or, brietty, (5) Bee aul, Appendix Dt We give here a proof of equations (7). Suppose the xq are given. Take a small square Burgers circuit of unit area normal to Oxy as shown in Figure 8. The \. jp Frovae closure failure By is C,C:. This may be regarded as compounded of separate elements corresponding to the curvatures and twist as we traverse each arm of the circuit in turn: Fieure 3 shows a special case of this. Taking A as fixed, as we move along AB, which is parallel to Osa, the lattice rotations about the three axes are us, xen ait, from equation (6). By vieualising the movements the reeulting com. ponents of displacement of C; can be seen to be ‘The displacements of C, as we move along BC; are Bers, — Fars, 0. Similarly, as we move along AD the displacements of.Cy are, Besa ~ tay = Betas an and as we move along DC: the displacements of Cx = fe, 0, He The (otal cloowe failure By io then given by the diflerence between the displacements of C, and Cs, thus: Br =~ (ent), Be = But, from equations (8) with 1,= (1, 0, 0) we have nin B= Dy ayy De = en, Dy = on {5ee pave 159,102 ‘Therefore, — (was + aa), an aud, in a sinilar way, by taking circuits normal Ors and Ors we obtain = (ean Fis) » ai = ay ~ (eur + at) « ‘Thus, in general, the relation between the a1 describing the state of dislocation and the xs describing the curvature is May en~ Fyne ABA 1,2,3) where by) = 1 for i= jy and by = 0 for i #4 j. Acknowledgments Dr. C.D. West was kind enough to show me his paper on the corundum crystals before publication ACTA METALLURGICA, vou. 1 and [ am glad to acknowledge here that it suggested to me the evalute construction of §2. 1 am also indebted to Dr. W. M. Lomer for his comments on the manuscript. References 1, Cann, RW. J. Inst. Metals, 76 (1049) 121 Faawk, F.C. Phil. Map, 42 (1951) 809, Scumtp, E. and Boas, W.. Plasticity of Crystals (Londen, FA, Hughes, 1950), p. 58, ‘Paper to be submitced to J. Appl. Phys ty (Oxford, 4. West, C.D. 5. Hit, R. ‘The Mathematical Theory of Plat University Press, 951), p. 136, 6, Esnexey, J. D. Phil. Trans., A244 (1951) 87 7. Nananno, F. RN. Adv. Phys, (1952) 269, © Prax, F.C. Plastic Deformation of Cry (Pittsburgh Conference, 1950), p. 150