Sie sind auf Seite 1von 127

Chapter 3

Dislocations and Mechanical Properties



L o n c before the role of dislocatioirs in the plastic deformation of crystals was recognized, the stress-strain field around dislocations received considerable attention in the theory of elasticity. It seems t h a t ~ e i i l ~ a r t e was the first to consider dislocations il~ as simple and typical sources of internal stress in elastic solids, and Volterra2 and Timpe3 investigated their elastic field. Fig 3-1 shows the two basic types of dislocations considered by Volterra. The first arises by making a straight-edged .plane cut in a solid, displacing the two surfaces of the cut in their own plane relatively t o one another by a constant amount in a direction perpendicular to the edge, and ~velding them together; this is shown in Fig 3-la, where the outer cylindrical surface represents the boundary of the solid, and the thin spiral cylinder around' the edge of the cut has been removed in order t o avoid infinities of stress and strain. The second type of disloc'ation is sho\\rn in Fig 3-lb; here the surfaces of the cut have been displaced relatively to one another by a constant amount parallel to the edge, and then welded together. After J. R I . Burgers,"hese two types are called edge and screw'dislocations, respectively. They represent simple limiting cases; in the general case, the surfaces of the cut are displaced by a constant vector that lies in the pl$ne of the surface but has coAponents both parallel and perpendicular t o the edge of the cut. This vector is the Burgers or slip vector. The name 'Ldislocatio~l'l gi\ien by A. E. H. Love;5 it \vas bor\\.as
References for this chapter are on page 192. 68




rolved from geology, where it means a sheas displacement cutting across strata of rock and producing discontinuities between the regions separated by the shear plane. I n 1911, Praadtl6 suggested a model for explainiag mechanical hysteresis in which several characteristic elements of dislocations \\-ere present. Dehlinger7 was the first to consider dislocatiolls in connection with questions of slip; he tried to explain recrystallizatioll and strain hardening in a way that is closely related t o the theory of strain harden-

a. An edge dislocation.

0. A screw dislocation.

ing put for\vard later by G. I. Taylor.$ Simultaneously with Taylor's theory of hardening, dislocatiolls were'used to explain the striking discrepancy between the order of magnitude of the observed yield' stress and that of the shear cohesion of crystal lattices as estimated theoretically, as well as other features of plastic d e f o r m a t i ~ n . ~ ~ ' ~1934, it Since has become increasingly apparent that dislocations represent a basic element in the structure of a plastically deformed crystal; they play a role in plastic deformation similar t o that of the dipole in dielectric polarization. I n the present review, those main features of the mechailical behavior of crystalline'materials which seem to be connected essentially with the presence and properties of dislocations will be discussed; One of these, already mentioned, is the remarkable discrepancy bet\\:'& the observed yield stress of crystalline materials and the theoretical estimate of the shear stress a t which slip would be expected t o occur in a



faultless crystal lattice. Another is strain hardening, and the finite yield stress ~vhich crystals seem t o possess even before they have undergone ally plastic deformation. A11 interesting problem arises from the fact t h a t many metals call suffer vast amounts of plastic deformation, corresponding t o the passage of hundreds of dislocations over each individual atomic slip plane. Since so many dislocations cannot be present in the undeformed crystal, there must exist either a mechanism generating new dislocations during the deformation, or a possibility of large amounts of slip heing produced by one dislocation before it is used up or immobilized. T h e yield phenomenon (sharp yield point with subsecluent drop of stress) observed in carbon steel, in zinc and cadmium single crystals, and some other metals, has been the subject of interesting considerations based on the el'fect of impurities upon dislocations. Questions of deformation by slip unavoidably lead to those of recrystallization, which must also be connected more or less directly with the properties of the dislocations present in a distorted crystal before it recrystallizes. This relationship between dislocations and structural changes due to annealing is particularly clear in the case of polygonization. Finally, fracture in ductile crystalline materials is usually not the simple cleavage process coilsidered in the Griffith theory; it is, ill general, a consequence of plastic deformation, and dislocations must play a.n essential part both in ductile fracture and in mechanical fatigue. This chapter should not be regarded as a review of published literature: since this embraces many widely diverging and partly obsolete vie\vs and theolies, its impartial abstracting ~vouldconfuse the nonspecialist reader for \\:horn this introductioil has been ~vritten. Instead of giving a revie\v, the writer has attempted to present what he feels is the most likely picture of the physical basis of the phenomena discussed, and to review only those publications which contribute to this picture. For this reason, a large part of the published literature-containing many well-kno\vn and ingenious papers-has not been referred to; on the other hand, several new and previously unpublished suggestions have been discussed in detail because they seem to lead to a more realistic picture than the earlier theoretical concepts. I 1 order to 1 help the reader to recognize s\.hich of the ideas have not had the benefit of public discussion, a list of the main items should be given here (the technical terms employed will be discussed in the sectioils mentioned). The new items are as follosvs: Sectioil 3-10. T h e relatioilship bet\veen the angular disorientation and the minimum size of neighboring grains.




Section 5-14. Mechanisms for the generation of dislocation mills. Sectioil 3-15. The 2-mill mechanism. Section 5-17. The nucleatioil of twin lamellae and of martensitic transformations. Sect,ion 5-18. The initial stress as being determined by the stress required for moving a dislocation in an otherwise faultless crystal (I' driving stress "). Section 3-20 and 3-23. Strain hardening by the intersection of dislocations. Section 3-21. Disc~ussion G. I. Taylor's theory of strain hardenof ing. Sectio~i 3-24. 13. E. Warren's new work on X-ray line broadening and it,s consecluences upon the theory 'of slip and strain hardening. Section 3-25. The temperature dependence of the yield stress due to the temperature dependence of the dislocation energy and of the driving stress (see Sec. 18); impossibility of thermal activation of the dislocation bulge extrusion process and of the movement of dislocations over internal stress barriers a t the ohserved values of the yield stress. Production of stress concentrations by the arrest of dislocations a t small obstacles; 'Ldislocatio~l creep" as a possible contribution to transient or viscous creep. Section 3-26. The yield phenomenon as being due possibly to the generation of Z-mills, with unanchored connecting dislocations, in the course of a 'preparatory slip; a semiquantitative explanation of the grain size effect in iron on this basis. Section 3-27. Demonstration of the impossibility of the thermal activation of recrystallization nuclei in the classical theories. Possibility of the formation of recrystallization nuclei by a new type of nucleation process, consistiilg in a homogeneous transformation instead of spatial growth. Section 5-28. Ductile fracture and fatigue as consequences of the stress field around dislocations and the corresponding stress concentration around assemblies of dislocations.



By yield stress* is meant the stress required t o produce plastic deformation. I n the case of single crystals, it is usually understood that
* I t is advisable to use the word "strength" only to denote a stress at which fracture occurs and to avoid the expression "yield strength."



it refers to deformation taking place by slip, not by twinning or kinking; in an undistorted crystal, this starts when the shear stress in the operative slip plane, resolved in the slip direction, reaches.a critical value (E. Schmid's critical shear stress law"). Both in single crystals and in polycrystalline materials, the yield stress depends on the magnitude of the preceding plastic strain; this dependence is given by the familiar stress-strain curve. I t was believed formerly that the initial yield stress of an undeformed crystal or polycrystalline metal had no finite value and would be found to tend toward zero as the sensitivity of the strain measurements increased. Single-crystal research, however, has sho~\~n'"hat in general this is not so; single crystals usually have a fairly \\.ell-defined initial yield stress. The sharpne$s of the yield limit is somewhat blurred in polycrystalline metals, but they too possess usually a fairly distinct, although relatively low, initial yield stress. The critical shear stress (initial yield stress) of single crystals of the common ductile metals with lo\\- or medium melting point (such as Na, K, Zn, Cd, Mg, Sn, Al, Cu, Ag, Au) is usually between lo7 and lo8 dynes per cm"i.e., 150 and 1500 psi). As in the case of the'tensile strength of brittle materials, a crude theoretical estimate of the initial yield stress of a crystal can be obtained in a very simple nray. Suppose that a shear stress T acts in a possible slip plane parallel to the slip direction; it causes a certain tangential displacement x between pairs of adjacent lattice planes. In Fig 3-2, the shear stress transmitted by the interatomic forces across the slip plane is plotted as a function of the shear displacement x. In the absence of a stress (T = 0))the displacemeilt is zero; when x has reached the va1ue.b of the identity period of the lattice in the slip direction, the shear stress 7 is again zero because the lattice configuration is the same as for x = 0. Bet~veen these positions of stable equilibrium, there is a potential barrier;* .at its top, the stress is again zero, but the equilibrium is unstable, so that the stress decreases with increasing displacement x. Finally, the tangents to the curve at the points x = 0 and :c = b represent Hooke's law, and their slope is determined by the value of the shear modulus G; if c is the spacing between adjacent lattice planes, the shear strain that correspoilds to the displacement x is .. z/c, and, accordiilg to Hooke's la\v,

* T h e potential curve is not drawn in Fig 3-2; its top would be midway between the points x = 0 and x = b.


O F T A N G E N T I A L FORCE ACTING BETWEEN TWO I I A L V E S CRYSTAL WITH T H E A M O U N T X O F TANOENTIAI, DISI'1,ACENENT B E T W E E N ADJACENT LATTICE P L A N E S . a. Sinusoidal c u r v e . b. Type o f curve to be expected f o r p e r f e c t l y rigid atoms. c. C u r v e to be e x p e c t e d f o r atoms o f moderate r i g i d i t y . OF A


The most familiar curve that satisfies all those conditions is the sine with the period b (Fig 3-2a).

For small values of x/b, it call be approximated by



This equation represents Hooke's law; comparison with E q 3-1 shows that the amplitude of the sine curve is

For an order-of-magnitude estimate, the ratio of the two atomic spacings b and c can be put equal to 1; this gives

T,,,is the highest shear stress that can be withstood by the lattice in the given plane and direction. I the applied stress is less than T,, the deforf mation must be elastic (although not Hookean escept for very small strains), and the crystal must return t o its initial configuration when the applied stress is removed. When the stress reaches the magnitude T.L,ho~veve~., slightest further displacement leads into the unstable the region beyond the peak of the curve; removal of the stress causes then the lattice t o drop illto its nearest ecluilibrium position z = b, which means slip by one interatomic spacing. Consecluently, T , , is the yield stress of the lattice for the slip plane and slip direction considered. For metals and most other crystals, the sine curve cannot be a good approsimation; this can be seen in the follo\ving way.13 Let Fig 3-3a represent atoms of two adjacent atomic slip planes, and let it be assumed that the atoms are rigid spheres. When the upper plane is displaced horizontally upon the lower plane by the shear stress T (Fig 3-30), the horizontal component of the interatomic repulsio~lrepresented in Fig 3-3n by the arrows 2 disappears a t once and its full value must be replaced by the applied stress 7; as the displacement increases, ho\vever, the required stress drops because the atoms of the upper plane climb on t o less and less steep points of the lower atoms. With the assumption of rigid atoms, therefore, there could be no elasticity, and the stress vs. displacement curve would have the character indicated in Fig 3-2b. I n reality, the atoms of metals, and t h e ions or molecules in nonmetallic crystals, behave as hard but not rigid particles; consequently, the stress vs. displacement curve \\ill have the character of the full curve in Fig 3-2c. Compariso~l with the sine curve that has the same initial tangent (dotted in Fig 3-2c) shows that the shear cohesion T,,, must be l'ower than the value estimated on the basis of the sine curve (Ecl 3-4). The difference depends on the hardness of the atoms. For copper, Lomer14 has calculated the 7 vs. z relationship more accurately; he found t h a t - t h e shear cohesion \\!as somewhat lower than the value



I N METALS Eq 3-.4. I t seems reasonable to expect, therefore, that the shear cohesion of the common metals is usually between

13-51 Since the shear modulus of the commercial metals is usually between 1011 and 1012 dynes per cm2, the theoretical estimate of the

T,,, E G/10 '


T,,, E G/50




a . Before application of shearing force. b. After beginning of shear displacement.

shear cohesion of their crystals is between lo9 and loL1 dynes per cm2 (15,000 and 1,500,000 psi). As mentioned, however, the observed values of the initial yield stress are usually between lo7 and los dynes per cm2; i.e., hundreds of times lo\ver than the theoretical estimate.* * The figures given for the discrepancy in the l~tcrature usually higher by are
about one power of 10; they are obtained from the estimate Eq 3-4 based on the sinc curve as in Fig 3-2a.




Katurally, the estimate applies to faultless single crystals only. The yield stress of stroilgly cold-worked polycrystalline metals is :often quite near the theoretical shear cohesion; for instance, that of strongly cold-~vorkedcarbon steel (piano wire) can be as high as 3 pct of the shear modulus, and thus within the range of the theoretical estimate of the lattice cohesion as given by E q 3-5. Similarly, very hard, nonmetallic single crystals like diamond and mica can withstand shear stresses exceeding one per cent of the sheal. modulus: The discrepancy appears even more striking if, instead of the yield stress, the magnitude of the shear strain a t which plastic deformation starts is considered. On the assumption of a sinusoidal variation of the shear stress with the shear displacement, the maximum stress mould be reached after a shear movement between neighboring lattice planes amounting. t o one fourth of the identity period, which corresponds t o about 25 pct shear strain. I n reality, single crystals of the softermetals reach their elastic limit a t a shear strain of the order of 1/10,000; a cube of 1 meter edge length would begin t o deform by slip after an elastic shear displacement of opposite faces by as little as Momm, the thickness of a fine pencil line. I the sine curve of Fig 3-2a is replaced by a f curve of the type of Fig 3-2c, the discrepancy is reduced, but it still amounts t o a factor t h a t may be 100 or 1000.

A very similar discrepancy exists bet~veen observed and the calthe culated tensile strengths of brittle materials like glass; A. A. Griffith15 has shown that this is due t o the presence of microscopically small cracks, a t the tip of which the local tensile stress can reach the high theoretical value while the mean stress in the specimen has the low observed magnitude of the tensile strength. It would seem plausible, therefore, t o attribute the discrepancy in the case of slip also to stress concentratioil a t small cracks or similar flaws. Very soon, ho~vever, it became clear that this explanation could not be satisfactory except perhaps in a few special cases. The main reasons for this coilclusioil are as follo\vs. First, random cracks as assumed in the Griffith theory would lead t o a considerable scatter of the measured values of the strength. Such a scatter (of the order of $. 20 pct) is in fact observed in measurements of the tensile or bending strength of glasses; the yield stress of metals, however, is sometimes reproducible within $. 1 pct. Second, the stress concentration factor due t o a crack of given



length is of the same order of magnitude for a shear stress as for a tensile stress. Now, it has been k i l o ~ ~ ~ na long time that the cleavage for strength of zinc and cadmium crystals cannot be explained on the basis of the Griffith theory as anelastic stress concentration effect. The critical crack depth c required for fracture under the tensile stress u is given by the well-kno11.11 Griffith formula

in which E is Young's modulus and a the surface energy of the \\.ails of the crack. With the observed values of the cleavage strength of zinc and cadmium crystals, this equation would demand impossibly large values of crack depth (e.g., 5 to 10 mrn for a single-crystal wire of 1 mm thickness). Since the yield stress of such crystals is lo~ver than the cleavage strength, the explailation of the discrepancy as a stress concentration effect \vould demand the assumption of even deeper cracks. Consequently, the discrepancy between the shear cohesion and the observed yield stress cannot be due generally t o stress concentration st the tip of cracks.

The next attempt a t explaining the discrepancy \\;a,? made by R. Becker.16r17 Although his theory cannot be retained in-its original form, it was the starting point of important developments and its detailed discussioil is perhaps the best approach t o modern ideas on the dynamics of slip.. Becker assumed that the gap between the high theoretical estimate and the lo~v observed value of the yield stress was due t o the presence of thermal stresses. Thermal movement in a solid consists of elastic nraves running in all directions, of frequencies between the audible and the infrared ranges. The random superposition of these waves gives rise occasionally to high local stress peaks, analogous to the crests of ocean ~ v a ~ ~ e s . thought that local slip could occur a t the observed Becker lo~v values of the applied stress ~vhenever this was raised t o the value of the theoretical shear cohesion by a thermal stress fluctuation extending over a sufficiently large area in a slip plane. Once a local slip had been "llucleated," it would extend over its entire slip plane by the action of the applied stress alone; in this way, every successful thermal stress fluctuation would produce a shear displacement by one interatomic



spacilig 0. * The rate of strain is then the number of successful fluctuations per unit volume, multiplied by b. I the local accumulatioi~ free f of energy required for a successful fluctuation (i.e., the activation energy of the process) is A, the number of such fluctuations per units of time and volume is given by the Roltzmann-Einstein theorem approximately as B~-A/~T 113-7a] where the factor B is almost coilstant: that is to say, if it depends on the temperature or the stress, this dependence is slight compared with that of the exponential term. Consequently, the rate of plastic flow is dyldt = CecAIkT [3-701 \vhere C = Rb. Recker assumed t h a t the activation energy A was the isothermalreversible work required for raising the. stress from the value T of the applied stress to t h a t of the theoretical shear cohesion T,,,throughout a sufficiently large volume V; he assumed that this was given by

T h e reason why the expression (7, - T ) ~not T , , ~ ~ must be used is , - r2, as follows: When the stress rises from T to T , in the volume V, the local strain increases by ( T , ~ T)/G if Hooke's law is assumed to be applicable. The increase of the average strain in a unit cube t h a t contains the small volume V is obviously V ( T , - T)/G, and the work done by the applied stress on the unit cube as its strain increases is TV(T,~ T)/G. No\\., the activation energy is the increase of the elastic energy in the volume V less the \\-ork done simultaneously by the external forces; it is, therefore,

With this value of the activation energy, Eq 3-70 for the strain rate becomes dy /dt = Ce-(rtrc-r)2V/2C1i7' 13-91 At first sight, this equation would seem to reflect several features of the plastic behavior of crystals, such as the existence of a fairly ~velldefined yield stress and the estremely rapid rise of the flow rate with the applied stress. A closer examination, ho~vever, reveals difficulties.

* In his original paper, Ijccker assumed the amount of slip produced by one successful thermal strcss fluctuation to be proportional to the applied stress.




I the applied stress is ollly Xooo the shear cohesion T , or less, its f of entire variation ketween zero and the observed yield stress cannot have much effect on the strainrate which ~vould retain afinite and high value even if the applied st,ress hecame zero. T o avoid this difficulty, Becker added the applied stress r as a factor to the right-hand side of E q 3-9 (see footnote, p. 79); this, ho\revel., would only lead to an approsimately Xewtonian viscous behavior. The difficulty of a finite strain rate a t zero stress could be remedied if stress fluctuations opposite in direction to the applied stress would also contribute t o the strain rate; this, however, \vould agaiii lead to Ne~vtonian viscous behavior instead. of the characteristic features of plasticity.18 The deepest and most stimulating'problem raised by, the Becker theory was that of the possibility or otherwise of the thermal activation of a stable local slip process. Experience shows that, and statistical mechanics explaiii why, activatioil energies higher than a few electroil volts callnot be obtaiiled from thermal agitation a t room temperature freclueiltly enough t o give sig~lificailt rates'to processes of the type considered here. The decisive question of the Becker theory is, therefore: Can the local slip process ellvisaged in the theory be activated with an activation energy not higher than 1 or 2 electron volts? The answer depends mainly on the value of V ; i.e., on the magnitude of the volume over which the local stress fluctuation has to extend in order t o produce slip that is not ~vipedout by the elastic reaction of the surroundi~lgs as soon as the fluctuatio~lhas passed j\vay. When the theory was created, the co~lditioils that determine Vwere unknown; later, i t became clear that the calculatio~l V is closely connected with the properties of of dislocations. Before this point is discussed, however, i t is interesting to calculate the highest value of V a t which the Becker slip process could be activated a t or not much above room temperature in the absence of any elastic reaction by the surroundings. This can be done by substituting typical numerical values in Eq 3-8a. Let the highest feasible value of A be 2 electro~lvolts = 3.2 X 10-l2 ergs; T can be neglected besides T,,, the magnitude of \vhich should be assumed as G/20 (cf. Eq 3-5), with a representative value of 2 X 10" dynes per cm2for G. This leads t o V = cm3, so that the linear dimensions of V if it is a cube or a sphere amount to about 10 atomic spacings; if V is the volume of a flat interatomic spacing thick, its edges mould contain about 30 atoms. It should be repeated t h a t the energy of the elastic stresses that arise between the slipped area and its surroundings have been neglected in this estimate.



If t h e slipped region were not embedded in purely elastic surroundings, there \vould be no lower limit t o the size of a stable area of slip. As it is, however, high stresses are produced het~veen slipped region the and its surroundings, and these tend to reverse the slip when the stress fluctuation has disappeared. T o resist them, the area of slip must have a certain minimum size, which call be estimated by a method first a p plied h y \To!merlg t o problems of nucleation in phase trailsformations and subsequently used in many fields of physic^.'^^'^

FIG3-4-SECTION ACROSS LIhZITDI) AREA O F LOCAL SLIP. AB is the slip plane; t h c scctions across the disloci~tion belt are surrotlntlcd by circlcs.

Fig 3-4 shows schematically a section across an area of local slip, in a plane perpendicular to the slip plane and parallel to the slip direction. The lines of the net represent lattice planes that were orthogoilal before the stress was applied and then suffered elas1,ic shear distortion. 1,ocal slip occurred in a region of the horizontal plane A B at the center of the figure, between the two reference circles; around the center of the area of slip, the shear stress has largely disappeared and the orthogonality of t,he net has been more or less restored. The slipped area is surrounded by a characteristically distorted belt, which intersects the plane of the figure in t\vo places indicated by the reference circles. I n the circle on the left-hand side, there is a strong horizontal compression above the slip plane, and dilatation helow it; in the right-hand circle, these are reversed. Ob\~iously, distortion of the crystal caused by the local slip is of the the same type as in Fig 3 - l a ; instead of applying a cut, displacing it,s surfaces, and \velding them t,ogether, the same final result has been achieved through slip by olle interatomic spacing confined to a, part of




the slip plane. The oiily, unessential, difference between Fig 3- L and a Fig 3-4 is that in the former the boundary (edge) of the cut is straight while the latter refers to a case (small patch of slip) where the corresponding boundary of the slipped area is a closed loop. Fig 3-5 shows an example of a dislocation belt that, illstead of being a closed loop around an area of internal slip, starts and ends a t the sur-



a . ABC is the boundary of the slipped area in view upon the slip plane. b. The slip plane is seen edgewise.

face of the crystal. I n this figure, i t was assumed that the esternal shape of the crystal was cylindrical, with the slip planes parallel t o the base of the cylinder, and that slip started a t the surface point P and extended to the irregular slip front ABC. Fig 3-6 also s h o ~ a slip process ~s that has started a t the surface, but in a representation corresponding to that in Fig 3-4. The fundamental point shown by Figs 3-4, 3-5, and 3-6 is t h a t the boundaries of slip processes that have not spread over the entire slip plane represent dislocations. A slip front is a dislocation; that slip spreads over a plane can also be expressed by saying that a dislocation line moves forward or a dislocatio~l .loop expands. The question, often



heard until quite recently,~ yhether dislocations exist, coiilcides with the cluestioil ivhether slip starts a t one point and spreads from there, instead of taking place synchronously over the trillions of atoms of the \\.hole slip plane. If it is not synchronized, it must spread gradually over the plane, and then dislocatioils must exist almost by definition:* they are the boulldaries between the parts of a slip plane that have undergone slip and those t h a t have not.










Fig 3-4 shows the typical structure of edge dislocations in crystals. The dislocation in the left-hand circle, where there is compressioil above and dilatation belo~vthe slip plane, has been called by G. I. Taylor a positi\re (edge) dislocation; the right-hand circle, on the other hand, surrouilds a negati\.e dislocation with dilatatioil above and compressioil below. The present writer h8s introduced the symbols T and 1 for negative and positive edge dislocations, respectively; the horizontal line in these signs means horizontal tension and the vertical line horizo~ltalcompression, so that the symbols are self-explanatory. If a n area of local slip is circular, the surrounding dislocatioil is of the pure edge type a t the ends of the diameter that lies in the slip direction, and of the pure screw type (Fig 3-lb) at thc ends of the transverse

* Cf. section 3-7 for the reason why the transition between an area of slip and its 'sbrroundings must be fairly sharp; i.e., why it must represent a relatively narrow dislocation~belt.




diameter; every\vhere else the circular loop consists of dislocatiolls of the mixed (general) type. If the area of slip is rectangular, with the sides parallel and perpendicular t o the slip direction, the surrouildiilg dislocatioils are of the pure edge and screw type. This is shown in Fig 3-7, which represents the view of ail atomic plane just above the plane of slip; the lines of the net represent atomic rows, and slip is assumed to have taken place in the square in the middle of the plane. The slip movement in this square is vertically up~uard the plane of the figure; in

FIG3-7-LATTICE 171,ANE W I T H S Q U A R E AHE.4 O F I,OCAL S L I P . The horizontal boundaries of the square :ire edge dislocations; the vertical boundaries, screw dislocations.

in front of it (upper side of the square) there is a belt of compressionj., and behind it ,(lo\ver side of the.,square) a belt of dilatation. I n the atomic plane below t,he slip plaile-(not s h o ~ ~ lthere is naturally dilatan) tion above and compression belo\v, so t h a t the upper and lower sides of the square are edge dislocations, positive or negative according t o which side of the slip plane is oriented so as t o face upward. The vertical sides of the square are screw dislocatioils; in these, the atomic plane is not compressed or extended but sheared. I n t,he atomic plane below, the shear is in the opposite direction. Dislocations play a role not only in-elasticityand plasticity, but also in zoology: snakes, worms, and mollusks move by developing disloca-. tions, as shown in Fig 3-8. The earthworm, starts its movement by developing a tensile dislocatio~l t its neck; this is a one-sided dislocaa tion, since the compression of the earth below the extended neck of the worm is negligibly small. The dislocation moves t0war.d the tail of the



\170rmand leaves the body there; after this, or (if the worm is in a hurry) cven before, a new tensile dislocation is formed a t the neck. 111 contrast t o this, most snakes move by forming compressive (positive, 1) dislocations a t the tail and moving them toward the head. Since the axial compressibility of 3 vertebrate is small, the snake forms an arch instead of compressing its body.

FIG~ - ~ - ~ \ ' ~ o v E M O FNA N E T




BE LO^')



12arthworms move by the motion of ,z tensile dislocation from the head toward the tail; most snakes move by the motion of "compressive1' dis--: locations from the tail toward the head. -. , .




A solitary st,raight dislocation line in a crystal has a certain free energy r per unit of its length. Thiscousists partly of the elastic energy of the compressed and extended or sheared parts of the lattice around the, dislocation, and partly of a disorder (misfit) energy between the parts.of the lattice above and below the slip plane. The misfit energy is of the same kind as the energy of a boundary between adjacent grains.of substalitially different orientation in a polycrystalline metal. I a crystal contains other dislocations, or if the dislocation is not f straight, its energy is not merely the length multiplied by the quantity r ; is, in addition, a n interaction energy between the dislocations present, or bet~veenthe elements of the same dislocation line if this is curved. T h e interaction energy betIveen straight, parallel, infinitely long easily calculated (cf. sections 3-21 and 3-22); that between the elem.ents of a closed loop, however, presents a complex problem that has not received full treatment yet. Since we are interested only in a n order-of-magnitude estimate of the size of a stable area of slip, we. shall .neglect..the.interaction between the elements of the




loop and put its total energy equal to the length of its circumference multipliedhy the specific dislocation energy r. The fact that a dislocatioi~ a n energy sho\\:s a t once that a very has small area of slip may not be stable. If the area dimii~ishes.slightly, the length and thus the total energy of the surrounding-dislocation decreases. At the same time, the average plastic strain of the crystal is reduced by the decrease of the slipped area, and thus work is done against the applied stress; however, it is easily.seen that this work may be smaller than the dislocation energy released if the area of slip is very small. I n this case, the local slip \\rill disappear' spontaneously. I n order that a n area of slip may be stable, the condition must be satisfied, therefore, that the work of the applied stress for a small increase of the area must be able t o provide the.accompanying increase of the energy of the dislocation loop due t o the-increase of its length. The work of the applied stress when the area of slip increases by a n amount S is calculated, after G. I . Taylor,%n the following manner: Let the shear stress act on opposite faces of a unit cube, and let the operative slip planes be parallel t o these faces. If slip by the interatomic spacing b spreads over a n entire atomic plane (of 2rea I), i t produces a mean shear strain -y = b in the unit cuhe, and the accompanying work of the stress T is 70. If the slip extends over only a small. area S of the slip plane, the shear strain of the unit cube will be less in the proport,ion S : 1, and the work done by the stress is

Let us assume no\\-, for simplicity, that the slipped area is circular, of radius r. When r increases by dr, the area increases by Bar&, and the work done by the stress is

At the same time, the circumference increases by 2ndr and the energy of the dislocation loop by dL &Z - 2 s d r r The slipped area, therefore, will be just of the critical stable size \\:hen d W = dL, or rrb G r [3-1 1 1 T o evaluate this equation, we must now estimate the magnitude of






T h e energy of a straight solitary dislocation per unit of its length has been calculated by Nabarro" to be a few electron volts per interatomic spacing. At the same time, he found t h a t the specific energy did not depend much on the breadth of the dislocation, i.e., on the length across t,he dislocation belt over which the misfit in the dislocation extends. * I11 what f o l l o ~ ~these results are derived in a crudely approxi~s, mate hut instructive way. 'I'he disregist,er bet~veenthe atoms above and below the slip plane in a dislocation is a maximum a t its center and decreases gradually with increasing distance. For the sake of simplicity, however, nTeshall assume that the disregister is confined t o a certain belt around the center of the dislocation (e.g., the inside of the circles in Fig 3-4); ~vithin this belt, there should be a constant compressive strain in the atomic plane above, and a constant tensile strain in the atomic plane belo~vthe slip plane. If the width of the belt amounts t o n normal interatomic spacings, there will be n 5 atoms above and n - 55 ) atoms below the slip plane in the dislocation. Thus, the tensile strain immediately below the slip plane will be E = (%)In, and the compressive strain immediately above - E = - ( f 5 ) l n . The elastic energy per unit of volume a t the slip plane is then

Although the elastic strains and stresses decrease gradually with increasing distance from the slip plane, we assume for this order-ofmagnitude estimate t h a t they exist with unchanged intensity up to a certain distance, and then disappear suddenly. This effective distance, obviously, cannot be as small as one tenth of the breadth of the dislocation, nor as large as ten times the breadt,h, it must be of the same order as the breadth; i.e., it must extend over about n atomic spacings vertically. If the atomi'c spacing is 0, the effectively stressed cross section of a dislocation is then n W , and the corresponding volume per atomic spacing of its length n2b3.With Eq 3-12, the elastic energy re,
* The breadth (also called width) of a dislocation has been calculated first Although the result of such calculations depends strongly on the by P e i e r l ~ . ~ ~ approximations used in co~nplitingthe interaction between the atoms, it secms that the breadth of a dislocation is probably usually rnorc than 2 or 3 but lcss than, say, 100 identity periods.




per interatomic spacing is then of the order


ETL~~~(>$,)~/~ = %b3E


With the representative values b = 2 X cm and E = 1012 dynes per cm2, E q 3-13 gives r,l = 10-l2 ergs = 0.6 electron volt. The derivation of E q 3-13 shows why the breadth of the dislocation does not affect its elastic energy in the first approximation: if the breadth decreases, the increase of the stresses and strains is just compensated by the decrease of the effectively stressed volume. I n addition to the elastic energy, the dislocation has also a misfit energy, and this does vary with the breadth because the a?)erageatomic misfit in the disorder belt cannot depend much on its breadth, while the number of atoms in the belt is proportional to it. Consequently, the misfit energy must be roughly proportional to the breadth of the dislocation. It would seem then that the energy minimum is reached when the breadth of the dislocation is reduced t o zero; however, the inde- . pendence of the elastic energy from the breadth, as expressed by E q 3-13, is valid only as long as Hooke's law represents a reasonable approximation. At very high compressive strains, the stress rises with the strain far more rapidly than in the Hookean region, and for this reason the elastic energy of the dislocation increases sharply with decreasing breadth. The balance between the decreasing misfit energy and the increasing elastic energy then determines a finite breadth for the dislocation. As mentioned, Nabarro's calculations led to the result that the depth amounted only to a few atomic spacings. On the other hand, the soap-bubble model experiments of Sir Lawrence Bragg and NyeZ3 suggest strongly the possibility of much higher values. The misfit energy has not been calculated; however, for a dislocation in the equilibrium configuratjon of minimum total energy, it is presumably of the same order of magnitude as the elastic energy, and so the total specific energy of a dislocation per atomic spacing of its length probably amounts to a few electron volts.

Let us assume that the total specific energy of a dislocation is twice its specific elastic energy, as given by E q 3-13 per atomic spacing; in order to obtain its value per ~entimeter'len~th, must be divided by b. it This leads to r E) / i b 2 ~ [3-141

3-MECHANICALPROPERTIES Substitution in Eq 3-11 gives


Thus, the diameter of the critical stable area of slip is of the order of (E/r) interatomic spacings; even if this ratio were as low as 200, the stable area would contain a t least 10 times more atoms than could be activated thermally in a synchronous local slip process of the Becker type, even in the absknce of any dislocation energy (cf. section 3-4). In addition, ho~vever,the creation of such local slip ~vouldrequire hundreds of electron volts for the energy of the surrounding dislocation loop, and only a fraction of this could be obtained from the work of the applied stress. The conclusion is, then, that in general stable local slip cannot be activated thermally a t stresses and temperatures of the usual orders of magnitude; the process. considered in the original Becker theory, therefore, is not feasible for commercial metals a t ordinary temperatures. I t ill be seen later that similar difficulties' the majority of the theories of plasticity, including some of the latest ones. The difficulties of the Becker theory call be removed by assuming that most of the discrepancy is dui to stress concentration a t cracks, and only a small part of it to the local superposition of thermal stress f l u c t u a t i ~ i ~Is1 ~ ~ case, the local stress a t the points where ther1 . this , mally activated slip can start would be y ~ if q is the stress concentration factor of a crack, and the activation energy \vould be (7, - ~ T ) ~ V / ~ G , instead of (rm - r)?V/2G. Since the local stress is now of the same order the of magnitude as r,,~, difficulties mentioned in the preceding sedtions disappear. 'Shere is little doubt tllslt such processes of local slip generation a t stress concentrations exist; however, the arguments meiltioned ,in section 3-3 show that they cannot be the general cause of the lo\v observed value of the yield stress. This consideration has led to the dislocation theory of the discrepanc,y, \vhich is a t present generally accepted.



When the Recker theory was put forward, the tlleoretical reasons for the impossibility of its slip-activation process were unkno\vn. The




questioil whether thermal stresses could bridge the gap bet\veen the estimated and the measured values of the yield stress was examined experimentally: it was coiicluded that, if the theory were correct, crystals ought t o he unplastic a t very lo\\- temperatures. Experiments carried out a t temperatures down t o 1.21<, h o ~ ~ e v eshowed only a r, small increase of the yield stress of zinc and cadmium crystals, and little decrease of their d u c t i l i t ~ 7 ; ~ ~ Becker theory, therefore, was the rejected. Closer coilsideratioil shojvs that this conclusion from the experiments, although it happened to lead t o a substantially correct result, was in fact quite unjustified. The single crystals used in the low-temperature experiments were cut \vith pliers and gripped in the pressure jaws of the testing machine a t room temperature, so that they suffered considerable plastic deformation before they were cooled down. When the low-temperature tensile test started, therefore, they contained large quantities of stable slipped areas, which, according t o the assumptions of the theory, could have extended further a t low stress without any thermal help. I 1 order to prove esperimentally t h a t the Becker 1 activation process was impossible, single crystals would have t o be used that did not colltaiil any slip process already started a t room temperature; i.e., any dislocations. This circumstance points toward a new possibility for explaining the lo\v observed value of the yield stress by the assumption that, almost unavoidably, dislocations are present in most practically obtainable crystals and crystalliile materials, and that, as Becker has assumed, slip already started can progress (i.e., dislocatioils call mox7e) a t relatively low values of the applied stress. These suggestioils were put forward in 1934 simultaileously by Polanyig and the uriter.I0

At first sight, the explailatioil of the observed l o ~ ~ r stress by yield the assumption that dislocatiolls are present in all practically obtainable crystalline materials even before they are subjected t o plastic deformatioil seems to beg the question; if dislocations could arise only as the result of uilfinished slip processes, the difficulty would not he explained but oilly put a step further back. In 1934, the problem had a different aspect. It was thought that djslocations would arise by local slip a t stress coi~centrations,'~ and then multiply in some,nray during slip (cf. sectioil 3-12). The difficulty mentioned in sectioil3-3 that cracks of impossibly large depth would be



required for explaining the observed yield stress on the basis of the Griffith theory as a stress concentration effect was kno\\rn; it was doubted, ho~vever, whether t h e stress conceiltration factors calculated for isotropic materials were even approximately valid for crystals. I t vras only in 1947 t h a t the stress coilceiltratioil effect of cracks in the cleavage plane of crystals \\-as mathematically studied, with the result t h a t the anisotropy did not change the order of magnitude of the stress concentration factor.25

F I C - 9 - ~ O R A I A ~ ' I O NO F A X 3


A new point was introduced in 1939 \\;hen W. L. (no\\; Sir La~vrence) BraggZ6and J. R4. Burgers4 poiilted out that certain types of grain boundaries are. simply arrays of dislocations and that, vice versa, dislocations must be present ~vheneverthere are slightly disoriented regions in a crystal. This is illustrated by Fig 3-9. Suppose that two small crystals grow together a t a nlome~lt when one of theadjacent faces has a n uilcompleted atomic plane. When the crystals have united (Fig 3-90), their adjacent lattice plailes do not retain their former structure: local readjustments take place, gaps are closed and spacings equalized as far as possible, and the result is Fig 3-9c. Comparison, e.g., with



Fig 3-6, sho~vs that the union of the two crystals has produced a slightly bent.crysta1 containiilg an edge dislocation. Sjmilarly, it is easy to see that if two crystals grow together with a slight relative rotation around the normal of the plane of contact, the result is a slightly twisted crystal coi~taining one or more scren- dislocations. Fig 3-9 sho\vs that a dislocation arises by the insertion of a lattice plane~onone side of the slip plane. Fig 3-10 shows schematically the insertion of extra lattice planes 1, 2, 3,4,011 the lower sides of slip planes






sl, sz, sg, sq, which form an equidistant set. I the spacing across the f inserted lattice plane is b, and the spacing of the slip planes sl, sz, etc., is dl the angle between the lattices separated by thevertical plane AB is

This relationship gives the spacing d of the dislocations that form a grain boundary between grains differing in orientatioil by a small angle 0. Naturally, a grain boundary call be regarded'as an array of dislocations only if their spacing is large enough to allo\v at least a distinct trace of the characteristic stress-strain distribution around a dislocation to be recognizable. This condition is fulfilled if the consecutive dislocations are at least, say, or I 0 interatomic spacings apart. Taking 5 spaciilgs as an estrcme lower likit, E q 3-15 gives the correspol~diiig



angular disorieiltatioil of the two lattices as about M radian, or about 10". This is the extreme upper limit to the angle between the orientations of two grains beyond which there is no physical sense in regarding the grain bouildary as an array of dislocations. The question is now: Call the dislocations that form a lo\\--angle boundary be mobilized by the applied stress to produce slip? Calculations of the stress required to tear a dislocation out of an array indicate that such a process would demand a stress much higher than the ob-

FIG 3-11-ADJACENT CRYSTAL G R A I X S O F 1 I I N I l I U A I DISORIENTATION. Sl~owing that two adjacent'crystal grains of finite size d cannot have an b angular disorientation less than the order of magnitr~de / d (b = atomic spacing).

served yield stress of crystals \vhenever the dislocatioils of the array are close enough to form a distinct boundary between disoriented regions. In order to be mobilized by stresses of the usual magnitude, the dislocations must be very far apart. The coilsiderations attached to Fig 3-9 show an interesting point. T\\lo adjacent grains in a polycrystalline aggregate cannot have just any small amouilt of disorientation; there is a lower limit to the angle bet\\-een them, which is reached \\-hen there is only one atomic spacing more bet\veen the grains a t one edge then a t the opposite one of the commo~l face (Fig 3-11). If the atomic spacing is b and the diameter of the grains (width of the common face) (1, the smallest possible angle between the grains is of the order of b / d radians. No\\-, X-ray investi-



means of the double crystal spectrometer'and by Guinier's gations focusing method have shown t h a t the angular di~orientat~ions the in most perfect crystals investigated (diamond, synthetic NaC1, and recrystallized aluminum) must be less than 10 or 20 sec, i.e., less than shout 1/20,000 radian. It follo~vs, then, that any "mosaic" grains that may exist in these crystals must be greater than 20,000 interatomic spacings across; i.e., their diameter must be a t least ofthe order of mm. Grains of such size, of course,.\vould be easily observable microscopically a t moderate magnifications if their boundaries were sufficiently distinct. Summing up the preceding considerations, one can say t h a t dislocations may arise by slip a t stress concentrations or may be present in a n undeformed crystal as the result of accidents of growth. I n particular, grain boundaries are arrays of dislocations if the disorientation of the adjacent grains is less than about 5 t o 10 deg. However, dislocations from such arrays can hardly be mobilized by the applied stress t o produce slip, because they are too tightly bound together. Those dislocations which can be moved by stresses of the usual order of magnitude must be :qo far apart that they can hardly form microscopically recognizable boundaries between disoriented regions.


The reality of the high theoretical estimate of the tensile strength of glasses and other brittle materials has been proved abundantly by esperimeilts as well as by routine industrial processes such as the tempering of glass and the manufacture of glass fibers. I n conti.ast t o this, numerous attempts to increase the yield stress of crystalline materials by the eliminatioil of flaws have been unsuccessful, until quite recently Herring and GaltT7 enco~~ntered crystals so thin that there was practically no place for cl.ucks or dislocations on them, and found that the elastic limit of such crystals \\-as quite close t o the theoretically estimated value. Working a t the Bell Telephone Laboratories, they investigated very thin, threadlike tin crystals, which were formed by corrosion on tin-coated iron. These crystals could be bent elastically until the strain in them rose t o about one per cent. If the crystals were really pure tin, and surface layers did not influence the result, this would amount to an increase of the usual yield stress by a factor close to 100, and the measured stresses must have approached the theoretical shear \ cohesioil within a factor less than 10. This result proves the reality of the high value of the shear cohesio~l


as estimated theoretically; it does nol prove conclusively, ho~vever, that the lo\v values observed in other materials must be due to the presence of dislocations. They could concei\~ably the result of cracks be or other stress conceqtrations, as explained in the preceding section.

It is easy t o stretch a very ductile metal (e.g., copper or gold) by rolling to -100 or 1000 times its original length. This iilvolves slip by hundreds of identityspacings on each atomic slip plane; that is to say, the passage of hundreds of dislocations over each plane. So many dislocations, of course, cannot have been in the crystal before deformation; even the crudest X-ray meth'ods could show up much lower densities of dislocations. Consequently, there must exist some mechanism by which new dislocations are produced ,during the deformation process, or a possibility of one dislocation producing large amounts of slip before it disappears or becomes immobilized. I t will be see11 later that both alternatives must be present. The operation of such " multiplication mechanisms " for disloca. tions_.wasfirst ,recognized.from the existence~ofs1ipzones (slip banils, slip lines) in which a self-sustaining slip process leads to very-high local shear strains. l o The first suggestions of multiplication mechanisms were based oil the mechanical momentum of the running dislocation. If a dislocation arrives a t the free surface of a crystal with a sufficiently high velocity, it can overshoot itself by its mechanical inertia, so that slip by two spacings, instead of one, arises a t the surface. This means the formation of a new dislocation, which now travels back and may undergo the same reflexion process on the opposite face of the crystal." There are, however, several difficulties in the way of such a mechanism: one of these is that the' scattering of elastic and thermal energy by the running dislocations may. possibly be too intense, and the velocity inay not reach the high value required for the overshoot.29I n addition, the reflexion process could take place only at a free surface but not at a grain boundary. Several other multiplication mechanisms havebeen considered; in view of the uncertainty of their importance or even feasibility, however, they will not be discussed here. A new way in which large amouiits of slip'may arise without a multiplication mechanism in the narrower sense of the word has been pointed out recently by Frank and Read.30 t should be explained with reference I





to Figs 3-12a and 3-126. The first of.these shoivs a crystal of rectangular parallelepipedic shape with slip planes parallel to its faces, and slip directions parallel to the edges, the front face of which has been indented at its bottom right-hand corner by a flat, rectangular indenter; the depth of the indentation should be one identity period in the slip direction. The indentation'produces slip in a horizontal and a vertical plane; let it be assumed that the slip has propagated to the rectangle ABCD. The sides A B and AD o f t h e rectangle are then edge di~locat~ions, the in horizontal slip plane aARb a n d in the vertical' slip plane aADd, res-

FIG SIMPLE DISLOCATION MILL. a. Indentation of a rectangular punch into the corner of a crystal face producing two edge dislocatio~lsA B and A D in the crystal. b. Slip on the plane containing the dislocation i l B due to the rotary motion of this dislocation around the fixed point A . '

pectively. If now a shear stress 7 is applied in the horizontal plane, the dislocation AB starts to move on; AD, on the other hand, should be assumed to remain stationary. Since the dislocation .4B is anchored a t A, it can only rotate (in counterclork\vise direction) around A, as indicated by the curved arrow; after 90" rotation it becomes a screw dislocation, after 180' an edge dislocation of the opposite sign, and so on. When, after 360, the dislocation has returned 40 its initial position, it has swept over the whole slip plane and created in it slip by one interatomic spacing. The salient point in this process is that the dislocation does not disappear ivhen it has swept over the entire slip.plane; in fact, nothing has changed then except for the occurrence of slip by one spacing. This call be realized in the follo\\ring way. As a consequence of



the initial indentation, an additional lattice plane ABCD has been inserted in the crystal, this corresponds t o the plane that has been displaced by the indenter from the front face of the crystal. The presence of the extra lattice plane ABCD produces the edge dislocations AB and AD. Obviously, this extra plane is not removed by the rotation of the dislocation AB, i.e., by the slipping of the upper part of the crystal upon the lower one; it remains embedded in the latter and thus atomic fit between the part of the crystal above and below the level aABb cannot be reestablished by slip betrveen them. The dislocation AR, therefore, does not disappear and can continue its rotation around A any number of times, every full rotation producing slip by one interatomic spacing. After a number of rotations, the crystal will have the shape shown in Fig 3-120. The Frank-Read configuration of a fixed and a rot:tting dislocation belonging t o two different slip pl:tnes nil1 be called in \\That follows a "dislocation mill." Like several of the earlier multiplication mechanisms, but in a more plausible \yay, it can eaplaiil directly the formation of slip zones (slip bands), i.e., the occurrence of large amounts of slip concentrated in one plane (see the next section).



A particularly interesting case arises when two dislocation mills of opposite sense of rotation are present in the same slip plane. The configuration considered by Frank and Read, in which the two mills are on a line perpendicular to the slip direction, can be thought of as generated in by the indentation process sho~vn Fig 3-13a. This differs from Figs 3-12a and 3-12b only in that the rectangular indenter has now been pressed into the central part instead of a corner of the front face of the crystal. The result is the displacement of a rectangular piece of a lattice plane from the front face and the insertion of an equal piece ABCD between two parallel lattice planes in the interior of the crystal, a t the depth t o which the slip created by the indentation penetrates. This means the presence of a rectangular frame ABCD consistiilg of the horizontal edge dislocations AB and CD, and the vertical ones BC and DA. I again it is assumed that, for some reason, the vertical dislocaf tions are immobilized, a shear stress 7 (see Fig 3-13a) applied in t,he horizontal slip plane will move the dislocatioil AB forward and t.he dislocatioil CD backn-ard, in the way indicated by Fig 3-13b, which repre-




sents consecutive phases 1, 2, 3 and so on of movement of AB in its slip plane. Initially, the dislocation bulges out increasingly between its anchoring points A and B where the immobilized dislocations DA and BC end. When the bulge reaches the surface of the crystal (indicated by a dotted line), it falls apart into two separate arms, which move indepeildeiltly as the arms of the two simple dislocation mills A and B. Later, the t ~ v o arms coalesce again and form a bulge on the opposite side of the line AB, ivhich contracts until it coiilcides with AB. At this moment,, slip by one identity pericd has taken place over the entire

a. Rcctangr~lar franie A B C D of edge dislocations produced by indentation of a rcctangulnr punch. 1). Dislocation loops 1, 2, 3, 4, represcnting corlsecutivc stages of nlotion of dislocation i I3 sho\vn in a ~ ~ n d the influence of the applicd stress r . l er

slip plane; nevertheless, the dislocatioi~A B has not disappeared, and the whole process can start again for the reason explained in the preceding section. Thus far, the effect of the double mill has been the same as that of a the single mill. However, there is an important difference bet~veen two. The single mill can move as easily as a solitary dislocation if the outlines of the slip plane are not too irregular. On the other hand, it is easily seen that the double mill cannot operate unless the applied shear stress reaches a critical value that is determined by the width AB of the mill. When the dislocation arc A B bulges out, its length and thus its total dislocation energy increase; in order to keep a given bulge in equilibrium, therefore, the applied stress must be so large that, for an infinitesimal advance of the bulge, its ivork can cover the simultaneous



increase of the dislocatioil energy. The problem of the critical shear stress at which the double mill starts to work has been treated first in the theory of precipitation hardening3' (cf. section 3-19); it is an exact tivo-dimensiona,l analogue of the problem of calculatiilg the pressure required for blowing a bubble from a nozzle of given diameter submerged in a liquid. Let a be the surface tension of the liquid, 2r the





2 ~ .

diameter of the nozzle (Fig 3-14), and /3 the angle between the axis of the nozzle and the tangenl to the surface-ofthe liquid a t the mouth of the nozzle, in a plane coiltailling its axis. The force exerted by the surface tension a t the mouth of the nozzle is then a 2 a r cos /3; it must be balanced by the force pr2a of the air pressure p across the mouth of the nozzle, so that

The maximum of the pressure occurs when cos /3 bubble is hemispherical :

1 ; i.e., when the



' -


is the -pressure required for blowing bubbles. The c'oilditions for the extrusion of a dislocation bulge between two obstacles (e.g., precipitated particles) or between anchoring points such as the two fixed vertical dislocations of a double mill, are quite similar. ' Here the specific (free) length energy T of the dislocation, its "line tension," plays the role of the surface energy, and the width AB = d of the double mill that of the diameter 2r of the nozzle. The quantity that corresponds to the pressure p is not the shear stress T but its product b~ with the identity period b'in the slip direction. This can be seen in the following \tray: The air pressure.can be defined as the elementary work : . pdV done'when a. volume 'dV .o'f -air is pressed through the nozzle, divide'd by d V . The correspondingqu quantity in the case of slip is the work of the applied stress done when the area of slip increases by a small amouilt d S , divided.11~ dS.'Since the work of the applied stress on a unit cube when the slipped area increases by d S is (Eq 3-10), the "shear pressure" thnt correspondsto p is ~ b . NOW the force exerted :by. -the$shear pressure ac,ross the width AB = d of t h e double .mill-is- d ~ b for,:the same reason as in the case ;, of the nozzle, i t reaches anlaximum. wheli the dislocation bulge is semicircular, and its value is then equal t o the sum 2T' of the dislo. .cation tensions a t A and B. Consecluently, the shearstress required to extrude a dislocation- bubble between A and B, and thus to start is the double mill ~vorking, given bj7


With d

2r, E q 3-18 can be written as


1 '


I n this form, it represents a general relationship between the radius of curvature r of any part of a dislocation line and the shear stress T t h a t must be present to keep i t in equilibrium with this curvature if the stress required t o m0ve.a straight dislocation in i t s slip plane can be neglected. If the dislocation moving stress (the "driving stress") is not bulge negligible but has afinite value,rd, the applied stress T can onl~7 the dislocation line t o a radius given by



Naturally, Eqs 3-18:to 3-20 are only rough order-of-magnitude approximations when applied to the problem of the double mill. They are good approximatioils if the distance beyond jrhich tu-o elements of the dislocatioil line havc no significant interaction is small compared with the radius of curvature of the line. This condition, ho~vever,is far from being fulfilled in practical cases. It 1vas found earlier (Eq 3-14) that



where C was a large fraction of unity, probably between Since Poisson's ratio v is usually of order 55, and E = 2(1 or 3 times G, we can write approximately

+ v)G = 2



As in 3-1.4, r ' i s thevalue of the free energy of the dislocation per centi-meter; that is t o say, its line tension. Substitution of 3-21b in 3-18 gives3' b T = yG[3-2.21

where y is a numerical factor of order unity. .This equation gives the value T of the applied shear stress that is required for producing a semicircular bulge in the dislocation ailchored a t the poles A and B of the double mill and thus t o start the double mill working. It is valid only if the dislocatio~ldriving stress ~d .can be neglected; as will be seen in section 5-18, this is true only of soft metals, and so the applied stress . a t which the double mill can start working is in the general case given by b . . T=Td+qG[3-231 d The shape of the dislocation bulge during the operation of the double mill depends on how the driving stress ~d is influenced by the velocity of the dislocation. Evidently, ~d cannot be proportional to the velocity: if i t were, very high stresses could be applied for a very short time without causing considerable plastic deformation. This is contrary t o elementary experience: below the temperature range of recrystallization, the yield stress of pure metals does not vary more than 20 or 30 pct in the entire range between the lo~vest and highest practicable rates of deformation. Cdnsequently, it is plausible t o assume that the driving stress of dislocations does not depend strongly on the velocity.




I11 this case, Eq 3-20 demands that the radius of curvature be more or less collstailt along the dislocation bulge, so that a t any moment .the shape of the bulge must be approximately circular, as indicated in Fig 3-136. Frank and Read assumed30 that the driving stress \vould be very low a t zero velocity and ~vould increase considerably with the velocity. I11 this case, the bulge-forming stress (7 - r d ) tvould be much higher near the poles of the double mill where the velocity of the dislocatioil and thus ~d \\'as lowest, and the radius of curvature of the bulge ~vould




W. T. READ).

be much smaller here than in the ceiltrnl regions; the spreading bulge \\-ould then assume the shape indicated in Fig 3-15, and its most sharply curved parts ~voulclcoalesce bcfore the central part of the bulge coulcl progress far.* I 1 this 11-ay, the double mill 11-ould emit consecutive 1 closed loops of dislocatioils as indicated in Fig 3-15, and act as a dislocatioil source; this is the term used by Frank and Read for the double mill. Since, however, thc consecutive emission of loops a t frequent spatial intervals depends on the assumptipn of a quasi-viscous behavior of thc dislocation driving stress, and since it is an ullessential feature of the dislocation mill mechanism that does not occur with a siilglc mill, the term "double mill " \\rill be used here instead of " dislocation source."
* T h e curvature of the loop near the poles of the double mill cannot be as sharp as is assumed in Fig 3-15: the smallest radius of curvature along the bulge must al\vays be equal to or greater than d / 2 , i.e., its radius in the semicircular starting position. This is obvious from Eq 3-20 r d can only be higher for a moving than for a stand~ngdislocation in a pure crystal, and so r can only be greater during movement than the value r = d/2 in the semicircular starting position.



The distinguishing feature of the double, as compared with the single, dislocation mill is the existeilce of a finite critical stress belo\\ ~vhichthe operation of the mill cannot start. This suggests the possibility of interpreting the initial yield stress of crystals as being the operating stress of double mills existing in them before deformation and having a more or less constant width d which determines the yield stress. However, there is no apparent physical reason why preexistiiig double mills should have a definite width, and there are weighty arguments pointing toward the conclusioil that the initial yield stress, in general, must be the stress required for moving dislocations in an otherwise faultless crystal, rather than the operating stress of preexisting mills (cf. sectioil 3-1 8). I t is likely, therefore, that dislocation mills are generated mainly in the course of the deformation, e.g., by the slip dellection mechailism described in the folloiiring section, or the Z-mill mechanism discussed in section 3-15.

- 3-14


The Frank-Read dislocation mill mechanism provides a plausible answer t o the question of how slip lines (concentration of slip in one slip plane) can arise \vithout large numbers of dislocations having t o be present in the slip plane before deformation. A dislocation mill, ho\v: ever, callmultiply slip only in its own plane; in order t o explain how slip can occur in many planes, either the assumption has to be made that very large numb.ers of mills exist in any undeformed crystal, or'a multiplication mechanism for mills must be found. Rilany observations indicate that slip can occur in so many plalles that the assumption of preexisting mills in all potelltially active slip planes is extremely unplausible. T o begin with, slip is by no means invariably connected with the formation of slip zones; there is little doubt that the visible zones account a t most for the major part, but not all, of the total deformatioil by slip. An extreme example for this is kinking, \\-hi'ch is a special case of slip taking place in the usual slip planes and directions of the crystals. The strain can be cluite high; nevertheless, slip links are almost al\vays absent. There are several other striking examples of zoneless slip (usually called, in a slipshod manner, L ' ~ l i p l e ~ ~ flow "). Creep in aluminum crystals under certain condit,ioi~s,~~inthe itial stages of the formation of Liiders bands in iron, the deformation of oxidation-hardened copper,33are well-known instances for this. I n such cases, therefore, slip occurs in too many planes to be attributed rea-



sonably to preexisting dislocations or dislocation mills. I t follo~vs then that the mill mechanism can prol-ide only a part of the required multiplication mechanism for slip: it can explain the multiplication of slip in a slip plane, but not its spread to so many slip planes. The necessity for a multiplication mechanism that spreads slip to other planes is clear even when there are visible slip zones: in order to attribute them to preexisting mills, an excessively large number of these JJ-ould have to be assumed. Indeed, it is well knon-11 that the spaciilgs between the zones are nearer to uniformity than to statistically random distribut i ~ nto ~ ~ ; account for the roughly equidistant slip zones, therefore, a far higher number of randomly distributed preexisting mills would have to be assumed. The simplest way in which double mills of the Frank-Read type call arise during slip is by the formation of steps, i.e., by the slip process jumping over to a parallel plane one or a few atomic spacings apart. The step formation was first mentioned by It. Becker" as one of the possible causes of strain hardening; in discussing its details, t ~ v o simple cases must be considered according as the step formation occurs at a screw or an edge part of a dislocation. Step formatioil at a screw dislocation can initiate slip in oblique planes, and it can also produce dislocation mills working in planes parallel to that of the initial slip. The same process taking place a t edge dislocations, on the othe'r hand, can give rise to dislocation mills that can operate in planes of any orientation, with the exception of that of the initial slip. These two processes mill now be discussed in some detail; their additional importance is that step formation is probably an important cause of strain hardening (cf. section 3-20). Fig 3-16 sho~vs step formation a t a screw dislocation: the slip is assumed to have been deflected from the initial horizontal slip plane into an oblique one, and, after a more or less short run in this, back into a plane parallel to the first. The deflection of slip into another plane a t a screw dislocation is particularly easy because the stress field around such a dislocation is axially symmetrical, so that the distribution of the shear stress is the same in any plane containing the dislocation (provided that- the effect of elastic anisotropy is disregarded). The asial symmetry of the stress field can be recognized a t once from Fig 3-10. Since the stress in the plane of the cut sho117n in this figure is a pure shear stress acting in the direction of the edge, symmetry and the equilibrium conditions demand that the stresses in any other plane going through the dislocation should also be pure shear stresses acting in the



same direction; moreover, if the anisotropy of the crystal is neglected, the stress distributioil must obviously he t h e same in each of these planes. The a priori chances of slip are therefore the same in every crystallographically eclui\~alentslip plane containillg a (resting) screw dislocation; the operative slip plane is preferred only if the applied shear stress is higher in it. Consequently, a small local stress irregularity or even a thermal stress fluctuat,ion may deflect the slip into an oblique plane as shol\:n in Fig 3-16. Naturally, the chances of a repeat,ed





deflection into another horizontal plane lvill be higher than those of the first if the applied shear stress is higher in the horizonbal plane. If the entire dislocation loop in Fig 3-16, consisting of one segmeilt in each of the tlvo horizontal planes and two connecting segments in the oblique plane, expands everywhere with about the same velocity, the deflection process does not lead to any slip multiplication; its only result is that slip occurs on a stepped surface instead of a plane. Such processes occur almost invariably during slip; they have been called (I cross slip" by Mathell-son, Hibbard, and RIaddin.35 The more spectacular cases of cross slip, hol\-ever, have a different origin: they arise by slip starting independently in t ~ parallel planes from opposite o directions, proceeding to\\-ard the same region, and being bridged by slip in a connecting oblique slip plane with the same slip d i r e c t i ~ n . ~ ~ ? ~ ~




The applied stress in this plane may be quite lo\\-; slip in it is the consequence of the superposed shear stresses of the disloclttions in the ti\-o parallel planes. Slip multiplication, ho\vever, call result from the process sho\vn in Fig 3-16, if the coilnecting dislocations in the oblique plane caililot move, or a t least cannot move as fast as the segments in the horizontal planes. I n this case, the loop represents a double mill in each of the horizontal planes and it can thus produce large amouilts of slip in each. Such double mills formed by the deflection of scre~v dislocations differ slightly from the particular type first considered by Frank and Read in that the poles of the mill are connected, in the absence of an applied




stress, by a screw instead of an edge dislocation (because the line connecting the poles is parallel instead of perpendicular t o the slip direction). This, ho\vever, is not an essential difference: the operation of the double mill is the same. Very different from the deflection of screw dislocations is a group of and t h a t may lead to processes that may occur \vith edge dislocatio~ls the transfer of slip to other planes. A simple example is shown in Fig 3-1 7, i11 which R is a hard inclusion representing a n unyielding obstacle to slip. A dislocation moving along the slip plane (horizontal line in the figure), therefore, is stoppecl around R and leaves it encircled with a dislocation ring, in the way indicated in the figure. With increasing a m o ~ m of slip, the ilumber of encircling rings increases; it is easy t o see t t h a t a t the same time an increasing compressive stress (in the slip direction) arises on the "slip\\-ard" side of 12, a tensile stress on its



"leeward" side, and a high shear stress in the parallel slip planes just above and below the inclusion. What happens now depends on the cohesive properties of the crystal. I the tensile cohesion 'of the lattice f is large compared with its shear cohesion, slip will start in one of the the parallel slip planes just above or helo~v inclusion; a t the same time i t is likely t o stop in the original slip plane. The generation of slip in t h e new planes may take place in the way envisaged in the Becker theory, by t h e thermal activation of a new dislocation loop; this is possible a t points of very high stress concentration. Thus, the presence of inclusions may block a slip plane after a small amount of slip and transfer the slip to another plane; in other words, it may prevent the formation of visible slip zones and produce a very even distribution of slip in the crystal. This may explain the observation t h a t the presence of small inclusions makes slip markings generally less distinct and sometimes suppresses them altogether. A striking esample of this, already mentioned, is the observation of Dewhirst and Smiths that slip markings are completely absent in osidation-hardened copper (with 0.6 pct Al), although the crystallographic deformation mechanism is manifestly identical with t h a t in the unosidized material in which very distinct slip zones arise. If t h e tensile cohesion of t h e crystal is relatively low, tensile fracture may occur on t h e leeward side of the inclusion (above or below the slip plane). Fig 3-18 shows the possible coilsequences of this. When a crack C relieves t h e tensile stress, the high shear stress a t its tip is likely to start slip in a parallel slip plane 2 that is not blocked by the inclusion; after this, slip may continue in the plane 1 to the left, and in plane 2 to the right of the inclusion. The consequence is the opening up of a cavity; naturally, this will in general have a finite length perpeudiculal. t o the plane of the figure.* If the distance between the planes I and 2 is small (one or t\vo atomic spacings), the top and bottom walls of the disk-shaped cavity may collapse o\\;ing to the forces acting between them, and then the configuration sketched in Fig 3-19 arises. There are one or two extra planes in the lattice around what was the disk-shaped cavity; in other words, the collapsed cavity is a closed loop of an edge dislocation whose possible slip vectors aye a t an angle t o the initial slip plane. Comparison with Fig 3--13 reveals t h a t such an edge dislocation frame is essentially identical with t h a t produced by a punch indentation, and therefore can represent a double mill t h a t can operate * Twinning may give rise to tllc opening up of very long a v i t i e s (the "Rosc
channels" in iron; see Sch~nid and Boas, ICristallplastieitiit, p. 101.) .



if the loop is large enough to fulfill the condition of E q 3-23. This is of interest because the Frank-Read double mill can hardly arise in reality by an indentation process such as was used t o introduce i t in section 3-13. Another interesting point is t h a t the double mill produced by cavity collapse can produce slip in planes a t an angle t o t h a t of the initial slip, but not in the initial slip plane.

FIG CAVITY FORICATION AT A H A R D I N C L U S I O N . a. Crack formcd b y t h e tensile stress a t a hard .inclusion in an activc slip plane. b. Growth o f t h e crack t o become a cavity b y furthcr slip.

It is easy t o find simple mechanisms not requiring the presence of inclusions by which cavities may open up. An obvious one is sho\\rn in Fig 3-20. When two edge dislocations of opposite sign meet, coming from opposite directions in slip planes a few atomic spacings apart, they create a high tensile or compressive stress in the space between them: At the same time, each dislocation produces shear stresses acting upon the other in oppositioil t o the applied stress, so t h a t the two dislocations immobilize each other; this effect \\rill be discussed in section 3-21 in connection with the Taylor theory of strain hardening. Suppose now that more dislocations arrive behind the two already present. They

will pile up as shown in Fig 3-20, and the normal stress between the opposite queues is reinforced. If the crystal has a relatively high tensile cohesion, the two queues of dislocations may ultimately slip past each other as the applied stress is increased; otherwise, the increasing tensile stress produces a crack as indicated in the figure. Once this has taken place, many or most of the opposite dislocations are annihilated and slip on the two planes call continue. I n the course of this, the crack opens up t o a disk-shaped cavity which, as shown in the discussion of Fig 3-19, may finally collapse and result in the formation of an edge dislocation frame which can act as a double mill in a slip plane that





is not identical with the planes that, have given rise t o the cavity formation. An easy consideration shows that a similar process call take place if the dislocations meeting in the crystal as in Fig 3-20 create a high compressive stress in the space between their slip planes. I n this case, the lattice planes between the active slip planes may finally "buckle" under the compressive stress acting in their plane, and their parts then ride up upon one another in the manner of the "overthrusts" in geology. The result is a "negative" cavity, i.e., one or several extra planes forced bet\veen the lattice planes. This is also equivalent to an edge dislocation frame and thus can give rise to a double mill. The formation of cavities on a smaller scale (of lattice vacancies consisting of missing atoms or atomic rows) has been investigated in detail by Seitz." His considerations are presented in Chapter 1 of this volume. I n the present section, particular emphasis has been given to the formation of cavities by processes taking place on a rather coarse



scale. That such processes must occur very frequently in plastic deformation is seen from the decrease of density 'that accompanies the plastic deformation of metals; The effect is far from being a small one; engineering tables of specific gravities reveal large differences between the densities of the commoil metals in the cast and in the cold-~Vorked state. It is rather surprisiilg t h a t this fact has not attracted more theoretical attention in the past.




Fig 3-21 shows a parallelepiped of a crystal whose slip planes are the faces of the parallelepiped and the slip d.irections its edges. Suppose that slip on a vertical plane parallel t o the arro\Ir T has produced the vertical edge dislocatioil A13, equivalent t o the insertion in the lattice of the extra atomic plane :4 BCD ; the correspoilding displacemeilt a t the surface ii:indicated by a step in the front face of the crystal. If it is assumed.that the dislocation A B remains firmly anchored t o the lattice when,subsequently a shear stress T causes slip in the'horizontal plane 1234, the effect of the slip \\ill be seen in Fig 3-22. The edge dislocation is cut in two parts, A E and FB; a t the same time the extra plane ABCD is also cut in two and its parts,moved t o different levels along the direction of slip. For reasons of simple geometry, a dislocatioil cannot end in the interior of a crystal but only a t a surface: this conditioil is satisfied in Fig 3-22 because the newly created edge dislocatiolls EH and FG,

represented by the cut edges of the extra plane ABCD, form continuations to the separated parts of AR beyond E and F t o the surface point If and G. E H and FG are dislocations of opposite sign; it \\rill be seen later (section 3-21) t h a t such dislocatioils attract each other; i.e., their mutual energy decreases as they come closer. Uilless the amount of slip EF is large, therefore, the new edge dislocatioils will merge as indicated b y an arc in the figure, &nd this arc colltracts in order to reduce its energy until it has become a line F E con~lectiilg severed ends of the the dislocatioil AB. This process, of course, is imaginary; in reality, the



connectiilg dislocatioil E F alone \\rill be present from the beginniilg of the slip on the plane 1234. E F is a screw dislocatioil because its direction coiilcides with t h a t of the slip; under the iilflueilce of the applied,stress T, therefore, it will tend t o move laterally (toward G and H). If E F is very small, the stress T will cause only a slight bulge in i t ; as the amount of slip EF increases, ho~vever, bulge becomes stronger. I t is easy t o the see that the configuratioil AEFB, coilsistiilg of the dislocations A E and FB assumed i s fixed to the lattice and the dislocation.EB, which is freely mobile, will behave just like a Frank-Read double mill.; the dislocation E F becomes semicircular, then espands further and sweeps over the slip plane 1234 \vhen the shear stress reaches the critical value given by Eq 3-22 or 3-23 where d = EF, and this process can repeat itself ally number of times. There is, ho~vever, great difference bettveen a the coafiguration A E F B and t h e double mill shown in Fig 3-13. In the



latter, the fixed dislocatioils (the poles of the mill) are on the same side of the slip plane, so that they form a figure U with the connecting mobile dislocation. In Fig 3-22, -on the other hand, the two poles of the mill are on opposite sides of the slip plane, so-thatthe mill resembles the letter Z. This has the consequence that the p,ole distance d, which is constant'in the U-mill,- increases with the amount of slip in the Z-mill. Eq 3-22 or 3-23 sho\\ls, thefi,.that the operating stress, which is constant for the U-mill, drops in the course~ofthe operation of the Z-mill:



PLACE IN THE HORIZONTAI, PLANE 1234. The severed ends E and F of the dislocation remain connected by a screw dislocation EF, which bulges out as shown under the applied stress r.

the Z-mill involves an instability, a self-accelerating process. A possible application of this to the theory of the yield phenomenon will be discussed in section 3-25. Naturally, the formation of operative Z-mills requires the occurrence of large amounts of slip in one plane. Once this has taken place (e.g., owing t o the initial presence of a dislocation mill), however, numerous additional mills may arise in the way just described. Fig 3-23 shows that the dislocation AB, which is cut across by the second slip process, need not have the same slip direction as the cutting slip; in the figure, its slip direction is assumed t o be perpendicular t o

that of the horizontal slip. I n this case, the extra plane represented by the dislocation A B is parallel t o the direction of the cutting slip, &nd the connecting dislocation E F is of the edge type, which has a transverse slip direction. The dislocatioil track AEFB now cannot act as a Z-mill, because all parts of it have a slip direction perpendicular to the applied stress. However, if the shear stress has a component perpendicular to EF, AEFB is capable of working as a Z-mill. The case of a screw dislocation cut by slip gives a similar picture.







1234 IN FIGURE . 21

I a number of dislocations pierce across a slip plane that subsef quently becomes active, and if the slip directions of some of the dislocations agree with that of the intersecting slip, a large number of Z-mills may become active when the amount d of the intersecting slip reaches the magnitude t h a t satisfies E q 3-22 or 3-23 with the applied stress 7 . What happens in this case is not easy to imagine; it may be that the rapid and confused slip due t o the simultaneous operation of numerous mills in the same plane quickly immobilizes the slip pla~ieby some of the processes to be discussed later in the theory of strain hardening. I this is so, an interesting coilclusion follows: the total amount of slip f in a plane in which slip is concentrated, e.g., owing to the operation of




one or a few dislocation mills, \trill be approximately equal to the value

d given by Eq 3-22 or 3-23 for a given magnitude of the applied stress.

This \vould be in remarkable agreement with observations of Heidenreich and S h o ~ k l e and~ A. F. Bra\\-n40 on aluminum. These workers y ~ of have found by investigating slip markings on deformed alumiilum electronmicroscopically with the oxide-replica method that the optically visible slip zones are really clusters of individual slip planes in which large amounts of slip have taken place. I n order to introduce a coilvenient nomenclature, the individual planes of slip conce~ltration will be called in what follows "slip tracks." The spacing of the slip tracks in the slip zone was found t o be roughly 100, and the amount of slip per track about 1000 interatomic spacings. Since, for a compressive strain of order of 10 pct as used in these experiments, G/T has the value 1000 for single crystals,* Eq 3-22 gives the magnitude of the observed slip per track correctly with q = 1.

Since the discrepancy between the theoretical estimate of the shear cohesion and the observed yield stress cannot be attributed t o therma,l agitation, it must be due, as far as call be seen a t present, substantially to one of two factors: 1. The presence of stress coilcentratio~rs such as occur a t the tip of a. crack under stress, or a t a hard iilclusion when this has been engulfed by slip; or 2. The presence of dislocatioils in the undeformed crystal. Both factors are likely to play a role; dislocations, hour eve^., are probably a fa.r more frequent cause of slip initia.tion a t relatively low stresses. Dislocatioils may arise irithout plastic deformation by accidents of crystal growth leading t o slight disorientations. Hovrever, if the disorientation bet~veen neighboring regions in a crystal is large enough t o rksult in a grain (or subgrain) bouildary in the usual sense, the dislocations between the. disoriented regions are too strongly bouild together .to be liberated b y stresses of the magnitude of the observed yield stresses of ductile metals. If a dislocation (or dislocation loop) could produce slip by only one identity period before disappearing from the crystal, vast numbers of dislocatiolls \vould have to be a s s u m e d ' t ,account for the large ~~ strains observed. So many dislocations canilot be pre'sent in an unde* s e e Fig 128, p. 163, in ~ c h m i d - ~ o a s . 4 2



formed crystal; consequently, new dislocations must he created during slip, or a dislocatioil must be able to produce large amounts of slip. Both alternatives seem to he realized in plastic deforinatioil Ily slip. Particularly iilteresting is that, whenever a dislocatioil loop does not lie \\;holly in one slip plane, any part of it that lies in a certain slip plane may sweep over this many times n~ithout disappearing; the operation of such "dislocation mills" is the likely cause of the concentration of slip in individual planes as manifested in the formatioil of slip zones or slip tracks. I a dislocatioil is anchored to a slip plane a t two points, it catlnot f suTeepover the plane hefore the applied stress reaches a critical magnitude given by Eq 3-22 or 3-23'. I t call he anchored either by hard in.elusions, as in precipitation hardening, or by its stepping out of the slip plane and running off a t a n angle to it. If it steps out of the slipplalle a t two points, it represents a double mill, urhich can give rise to a large amount of slip in the plane only if the applied stress reaches the critical value just mentioned. If the t\vo ends of the dislocation segment in the slip plane step out of it toward the same side, the configuration is that of a U-mill; if the continuations of the segment outside the slip plane lie on opposite sides of the plane, a Z-mill arises. The critical shear: stress required for operating the U-mill remains coilstant; that for a Zmill drops in the course of its operation. ~ i s l o c a t i o i lmills can produce multiple slip only in the planes in which parts of the geileratitlgloop are lying. I t urould be hard to assume in an undeformed crystal the presence of a number of mills large enough to provide for all slip planes in which significant slip is observed. Con: sequently, most of the mills must he created during slip. Some of the geometrical possibilities for this have been pointed out in the preceding sections. I n addition to the creation of mills, dislocations can be generated in other slip planes in the course of plastic deformation, probably by numerous processes iilvolving the creation of stress concentrations, cavities, and overthrusts of atomic planes; some possibilities have been mentioned in the preceding paragraphs. The present aspect of the problem of why slip takes place a t relatively low stresses, therefore,.can be summed up in the follo\ving way: Unless extraordinary care is taken, every crystal of moderate or large size contains flaws such as cracks, slight disorientations, etc., from which slip can start a t a low stress. Once it has started, there are numerous ways, some maiilly topological, others partly or mainly dynam-





ical, in which new dislocations or dislocation mills can arise. Not so long ago, the cluest for "the" multiplication mechanism of disiocations was one of the central activities in the theory of plastic-ity; a t present, the situation has become rather similar to that in the physics of the elementary particles: the former shortage has given place to an almost embarrassing abundance.



The existence of a discrepancy betnieen the observed yield stress and the estimated shear cohesion of the lattice in the case of slip a t once raises the question whether a corresponding discrepancy exists in mechanical twinning. Very little is known about the magnitude of the shear stress required for twinning: Miller4' observed twinning in zinc crystals a t resolved shear stresses between 30 and 60 bars; the initial yield stress for slip in zinc is about 9 bars, and the shear modulus for the slip plane in the slip direction 408 kilo bar^.^^ In cadmium and tin, too, twinning seems to take place at resolved shear stresses several times higher than those for slip. The twinning stress can be estimated theoretically in the same way as the yield stress f or slip (section 3-2); the only difference is that the distance between the neighboring equilibrium positions during the shear displacement of neighboring lattice planes is not the identity period b in the direction of shear but considerably less. The shear strain in tninning is sometimes extremely small; in the most familiar cases of metals, however, it is usually between 10 and 20 pct.* For zinc, i t amounts to 14.3 pct, so that the theoretical estimate of the twinning stress would be about 7 times lower than that of the yield stress for slip. Since, as just mentioned, the observed twinning stress is probably between 3 and 6 times higher than the observed initial yield stress, the discrepancy between the observed and the estimated values of the critical shear stress is between 20 and 40 times lower for twinning than it for slip. Since, ho~vever, amounts to a factor probably between 1000 and 2000 for slip (based on Eq 3-5 and the numerical data for the yield stress and shear modulus just mentioned), there is still a factor probably between 25 and 50 between the observed and the estimated twinning stress in zinc. There is no reason to expect a different result with those other crystals for which a t least the order of magnitude of the twinning stress could be recognized experimentally. * An exception is a-iron: the twin shear is here 50 pct.



Can thermal agitation alone be responsible for this discrepancy? In other words, is Becker's suggestion about the cause of the discrepancy for slip applicable to twinning? That this is very unlikely can be recognized from a consideration analogous to that carried out in the case of slip (sections 3-4 to 3-8). When a twin lamella begins to be formed, it cannot extend over the entire lattice plane at once; initially, it must be a lamella of limited width and length, and very small thickness. Fig 3-24 shows twin lamellae in polycrystalline zinc where the grain boundaries limit their length and width; the surface of the metal sheet was



polished and then twins produced by extension. The sectional views of the lamellae have a characteristic fishlike shape, with contours approximating circular arcs. This shape satisfies the condition of minimum energy (with large lamellae, this is mainly elastic energy). With increasing stress, the lamellae thicken; when they first become visible, they are extremely thin and relatively very long. As is indicated in Fig 3-25, there must be a dislocation wherever an adjacent pair of lattice planes parallel to the plane of twinning intersects the boundary of the lamella; in Fig 3-25, it is assumed that the direction of shear is parallel to the plane of the figure, and so the dislocations must be more



or less of the edge typc \ \ . l ~ e ~ , c i~itersectthis plane.* Their sign is . thcJ opposite at oppositc elicls of the section of the lamella. As an area of local slip \is surrouncled by a dislocation ring, so a localized twin lamella is surroundecl by as many dislocation rings as there are pairs of neighboring lattice planes in it. T h e twin dislocations, ho\vever, corrcspond to a much smaller shear displacemelit: the length of their shear \lectors amounts usually to some $6 or j . f o of the interatomic spacing in the shear direction. Eq 3-14 shows, then, t h a t the specific length energy of the twin dislocations is probably only of the order of a f e ~ v per cent of that of a slip dislocation, i.e., is a few times 10-"rgs per cm instead of being lo-". Another difference between the local area of the slip and the t117in lamella is that t h e boundaries-bet~veen latter and

FI(; 3-25-T\r1r I I I S L O C A T I O S S R E I ' R E S E S T I N C BOUNIIARY BET\YEIIN l,AAllBl.liA 0 1 7 I r l h l I T 1 2 D LI3SC'l':-I A N D ITS SURROUNIIINCS.


the surrounding matrix have a n interface energy. Where the boundary is parallel t o the plane of t~vinning,this energy is very low compared with the grain-boundary energies: i t seems that the twin boundary energy is of the order of 20 ergs per cm2.43 However, this low value of the boundary energy applies only in the surroundings of the points A and B in Fig 3-25; else\\-herc, the interface energy is much higher and may approach the grain-boundary energy (of order of 500 ergs per cm2) where the angle between the lamella boyndary and the plane of twinning exceeds 10 or 15 deg. The increase of the boundary energy is due t o the energy of the boundary dislocations shown in Fig 3-25 (see Chapter 2); another part of the energy of these dislocations (that due substantially t o the interaction of the dislocations on opposite sides of the lamella) represents the macroscopic elastic energy of the lamella.

* The dislocations in Fig 3-25 cannot be regarded in thc usual way as represnntirlg the insertion of a semi-infinite atomic plane into the lattice. From this point of ~ i c wthey would correspond to the insertion of a fractional atomic plane. ,



After this survey of the structure and energy of a twin lamella, it is clear how the stability consideration for the area of local slip given in sections 3-6 t o 3-8 must be modified in the case of twinning. Instead of the energy of one dislocation ring, that of n rings (including their interaction energy) must be taken into account if the thickness of the lamella, t = nb, amounts t o n interplanar spacings b; in addition, the twin boundary energy is present. If the radius of the lamella (considered a circular disk) is r, its total twin boundary energy is 2r27r . a ; as mentioned, the surface energy of the disk also contains the contribution of the boundary dislocations, which may be the major item if the thickness is not very small compared with the radius. For the present limited purpose of determining whether the lamella can arise by thermal activation, it is sufficient to estimate a lower limit of its energy if even this turns out to be too high for activation to be possible. Accordingly, the elastic energy of the lamella and the surrounding matrix mill be omitted and only the individual length energy of the dislocation rings will be taken into account; since the interaction energy of dislocations of the same sign is positive, the omitted energy terms are positive.* They are not likely to be unduly large compared with the self-energy of the dislocations if these are spread out along the boundary owing to the fish-shape of the lamella in side view. The energy of the lamella, apart from the omitted terms, can then be written as W l = 2r2aa: 12 . 2 r ~ r [3-241

As a rough approximation, the mean radius ofthe dislocation rings has been assumed to be equal t o the radius of the lamella. The volume of the lamella is V = Mr2&. nb [3-251 if its circular contours are approximated as parabolas. The rat10 of the

* A more accurate calculation in which the elastic energy is taken into account has been given by Fisher, Hollomon, a n d Turnbull [Kinetics of the Austenitchlartensite Transformation. Trans. A l M E (1949) 186, 6!)1]. I n this calculation. however, t h e dependence of t h e interfacc cnergy upon t h c orientation of t h c interfacc rclati\rcly t o t h c matrix has not been taken into account. This nlay hc a n importnot factor, since most of t h e surface of t h c lalnella is f a r from bcing p:~rnllel t o t h c planc of shear, so t h a t t h c mean interiacc crlcrgy may bc mt~cll ncarcr t o t h c ordinary grain-boundary energy (sevcr:~l 100 crgs per cn12) than t o thr: t,iirin bor~ntlnrycnergy (10 t o 30 ergs per cm2). I t is d~ffic~tlt recognize to whctlicr t h c clastic energy or t h c variation of t h c intcrfacc cncrgy is t h c rclat~vc:ly rnorc important factor.

radius r to the thickness nb is obtained by considering a lamella of given volume : dV = rsnbdr xr2sbdn = 0 [3-261 and given energy :

dW1 = 4r sadr


2 srdr

+ 2rsrdn = 0


Division of Eq 3-27 by E q 3-26 gives

With r

ergs per cm, a

20 ergs per cm2,and b

2 X 10-8cm,

substitution 'of n from Eq 3-28 in E q 3-24 gives

This is the order of magnitude of the energy of the lamella after it has sheared over into the twinning position. On the way to this, ho~vever, intermediate states of .higher lattice energy, corresponding to the partly sheared distorted lattice, have to be p r o d ~ c e d . ~ it fis assumed q that the shear stress varies in the course of the twin shear according to a curve of the type shown in Fig 3-2c, the correspondiilg potential energy will vary according to curve 1 in Fig 3-26 between the beginning and the end of the twin shear; a glance upon Fig 3-26 shows that the maximum of the potential curve is approximately X T , , ~ where y is the final shear strain (the twin shear) and T,,, theoretical twinning stress the as just discussed. At the same time, the boundary energy of the lamella varies according to a curve of type 2, which, a t the end of the twinning process, reaches the magnitude of W l as given by Eq 3-29; the curve rises roughly parabolically because the dislocation energy increases with the square of the displacement, i.e., of the Burgers vector. The curve of the total energy (lattice energy plus boundary energy) is of the type 3 in the figure; its maximum is reached toward the end of the twinning process. By means of the orders of magnitude of r', a , r, n, etc., which have partly been estimated in the preceding paragraph and

partly mill be. obtained in Eqs 3-30 through 3-31a, it is easy to verify that 1.v~ and the maximum of. the lattice distortion energy for a uniform distortion ~vithinthe volume V of the lamella are of the, same order of magnitude, as assumed in Fig 3-26. It is now plausible that the total (lattice plus boundary) energy of the lamella during the tninning


FIG3-26-TWIN NUCLEATION. Curvc 1 represents schematically the potential energy between two halves of a crystal displaced along a lattice plane. Curve 2 indicates the increase of the energy of the dislocations represcnting the boundary of a finite twin lamella with the amount of shear in the lamella. Curve 3 is the sum of the two energies plotted in curves 1 and 2; it shows how the work required for creating a finite twin lamella with fixcd I~oundary a large crystal increases with the shcar strain in the lamella. in

process will hardly exceed the value Wi,and consequently this quantity alone can be used as representing the maximum energy of the lamella. The activation energy is then obtained by subtractingfrom T z the work V W done by the applied stress during the twin shear, for a lamella of the critical size as determined by the Volmer condition lyaccordingto which, for a differential increase of the size of the lamella, the increment of the work of the applied stress should just cover the incrcnse of the boundary (plus, in the exact calculation, of the elastic) energy). The work of

the external stress is calculated in the manner used in section 3-4. If a cube of unit \~olumesuffers the twin shear y, the applied stress does the workry; if the shear is restricted t,o the volume V, the shear displacement of the opposite faces of the cube corresponds t o a mean strain Vy, and thus the work is V . r y . Consequently, the energy t o be supplied by thermal agitation during the creation of a lamella of radius r is of the order Wl - W = 6r2?ra - >5r2?rnbry - [3-301 I n this expression, n can be substituted from E q 3-28:


A lamella is of the critical stable size if,-for an increment dr of its radius, dWI = dW, i.e., if the derivative of Wl - W vanishes; this condition gives ab 12rcna - 3rc27r- ~y = 0 [3-3 11

from which

The activation energy A for the formation of the twin lamella is the value Wl - W obtained if r, is introduced for r in E q 3-30a:


ergs per cm, a:

20 ergs per cm2, b

2 X lops cm,

r = 5 X lo7 bars, and y = 0.143, y2 = 0.02 (the latter values corre-

sponding t o the case of zinc), it follo\vs that A = ergs, i.e., millions of electron volts. The extremely high value of A in spite of the relatively small discrepancy between the theoretical and the observed t~irinning stress is a consequence mainly of the low value of the twin shear; this leads t o a very large size of the critical stable lamella (Eq 3-31a gives r, 3 microns for these numerical values of I?, b, 7, and V) and thus to a very high surface energy. Obviously, thermal activatioil under such conditions is out of question; this agrees with experience which sho~vs that a lowering of the temperature promotes tninning as against slip, although the temperature dependence of the yield stress for slip is relatively small. The question for the origin of the discrepancy does not seem very

difficult to answer. Since the discrepancy is relativeljr small, its bridging does not require high stress concentrations. The internal stresses arising during slip in t h e various ways to be described in the following sections dealing with strain hardening may easily initiate twinning, other 1 types of stress conceiltration can also play a role. I 1 ferritic iron, the increase of the yield stress due t o very high rates of deformation in impact loading is the usual cause of formation of twins (Neumann bands) ; in calcite, the localized pressure of a knife edge starts t\vinning in the Baumhauer experiment. On the whole, therefore, one can say that the dynamics of twinning seems t o be relatively simple: only a small local stress concentration is required for the shear stress t o reach the theoretical value, and thermal activation seems t o be impossible until the applied (local) stress becomes nearly equal t o the theoretical twinning stress. An extremely interesting problem, closely related t o twinning, is t h a t of the formation of lamellae of the stable phase in martensitic (strain) transformations. I n these processes, the lattice of the unstable matrix is transformed into the stable phase by a simple distortion process, consisting of shears and a superposed small compression or dilatation. T h e stable phase appears in the form of lamellae similar to those in mechanical t l ~ i n stheir shape is obviously determined mainly by the ; condition of minimum elastic energy. Although martensitic transformations can be promoted by the applicatioil of stress, they take place spontaneously when the temperature is lowered to a fairly well determined value, the martensite point. The question is, how can the martensite lamellae arise when, as just demonstrated, twinning cannot be nucleated by thermal activation unless the shear stress is quite close to the theoretical twinning stress. The answer is interesting; i t can be recognized from Eq 3-32 for the activation energy in twinning. The energies involved in the formation of a martensite lamella are obviously the same as in the case of twinning; the only difference is t h a t the surface and elastic energies are covered, not by the work of the applied stress but by the difference between the free energies of t h e austenite and the martensite. The available work per unit volume of t h e twin lamella is r y ; the activation energy for the formation of the martensite lamella, therefore, is obtained simply by replacing r y in Eq 3-32 by the free energy difference 6F per unit volume between austenite and martensite a t the transformation temperature. According t o Cohen, Rfachlin, and P a r a n j ~ e this~free energy differ,~ ence is of t h e order of 300 cal per mole, i.e., about 2 X lo9ergs per cm3.



With the values of T and 7 just used, the activation energy ergs ergs; if this is replaced by is obtained and T 2 r 2 amounts to 5 X l o L 3 the square of 6F = 2 X lo9 ergs, the activation energy is reduced t o 1.25 X 10-lo ergs. This is still somewhat too high; however, if I? and a are lo~ver a factor of 3 than the values assumed for the twinning of by zinc, the activation energy is reduced to a feasible value. The important result of this consideration is that the estimate of t h e activation, energy that has led to impossibly high values for twinning a t the observed stress (as it was expected from the temperature independence of twinning) appears to give a feasible order of magnitude for the activation energy of martensite formation a t the martensite point (as it was to be expected from the fact of the spontaneous formation of martensite in the absence of significant internal stresses). At this temperature, thermal stress waves can shear over the austenite lattice to martensite within the minimum stable volume; once this is achieved, the martensite lamella can extend without further thermal help a t a rate that approaches the velocity of sound, since no diffusion process is involved. The formation of martensite, therefore, seems to take place in the following way. With decreasing temperature, the free energy difference between austenite and martensite increases; the activation energy required for the formation of a martensite lamella, hov ever, remains fartoo high until, a t the martensite point, the free energy difference reaches the magnitude of about 2 X lo9 ergs per cm3 and, consequently, the activation energy for the formation of a stable lamella is lowered to the value of 2 or 3 electron volts, which is just feasible a t the martensite point of iron-carbon alloys. It is interesting that, if the free energy difference is represented as an elastic energy density, its observed value of 2 X lo9 = T ~ / ~ G correspond to a shear stress T that is would just about equal to the lattice shear cohesion (theoretical yield stress) of iron. In fact, if T = G/20 (see E q 3-5, section 3-2), G = 1.1 X 1012 dynes per cm2 gives about I .4 X lo9 for r2/2G. I t may be asked whether the omission of the elastic energy in and around the lamella in this consideration does not introduce a serious error. I t is easy to see that this is unlikely t o be true. The elastic energy is proportional to the volume of the lamella; its value per unit of volume, therefore, is to be deducted from that of the free energy differedce between unstressed austenite and martensite. The elastic strain energy can be expressed as the product of the volume of the lamella and an elastic energy density corresponding to a certain effective stress, and



the value of the stress must be of the order of the highest stress that can exist, a t the martensite point, in the austenitic matrix. This is unlikely t o exceed 5 X lo9 dynes per cm2, and the corresponding elastic energy is about lo7 ergs per cm3. Consequently, the elastic energy of the lamella is probably only a small fraction of the free enev.gyliberated, and so the omission of the former can hardly have more effect than a moderate error in the estimate of the martensite temperature. The martensitic transformation, according to this, is a process in. which thermal activation plays only a minor part. It could be compared with the sliding of a body do\\rn a slope; strictly speaking, this is also a thermally activated process, but the activation energy becomes feasible only when the gradient of the slope has reached a value almost equal to that required for sliding \vithout any thermal agitation. It may be called, therefore, a nearly athermal process. I n most experiments, the formation of martensite is seen to progress oilly when the temperature is lowered; in a few cases, however, martensite lamellae have been observed to form a t a constant temperature. This is easily understood. The transformation is strongly influenced by the presence of stress; the martensite point (transformation tern-' perature) of a given lamella is raised by a stress acting in the direction of its martensitic shear, and depressed by a stress of the opposite direction. Owing t o the presence of internal stresses and other inhomogeneities of physical or chemical nature, different regions of the specimen will have different martensite points. At a given temperature, some regions may have extremely low activat2on energies and so they may transform almost instantaneously, while others have acti\~ationenerf gies too high t o be feasible. I the free energy differences for the various 'regions in the specimen are distributed over a wide interval, a t any given temperature only a few regions may have activation energies capable of giving a n observable reaction rate (neither too fast nor too slow) ; under such circumstances, therefore, no isothermal transformation may be noticed. If, on the other hand, a considerable fraction of the volume of the specimen has activation energies in the same narro\Ir range, a n isothermal progress of the transformation may be observed. When Polanyig and the resent writerL0 explored the possibility .of dislocations being responsible for the discrepancy between the estimated and the observed values of the yield stress, the first question was how much the resistance of a crystal to shear was lowered by the pres-




ence of a dislocation. Polanyi assumed that the factor by: which a dislocation reduced the yield stress was of the order of magnitude of t h e number of atoms across the effective breadth of the dislocation; this implied the assumption that the breadth of the dislocatioi~ amounted t o hundreds or thousands of atomic spa&ings.Since' the calculations of Peierls2' and Nabarro?' have sho\vn that a breadth amounting t o a few, or a t most a few dozen, atomic spacings is more plausible, l'olanyi's estimate of the effectiveness of a dislocation in reducing the yield stress must have been too pessimistic. Opposite to this feeling was tKat of G. I. Taylor, \\rho, in his theory of strain hardening (cf. section 3-20), assumed for the sake of simplicity that a solitary dislocation in a n othernrise perfect crystal \\:auld move under a negligibly small stress, and that any resistance to shear t h a t the crystal may possess was due to the.interaction of its dislocations and to other defects. Taylor's assumption has beell adopted in the meantime by most if not all theorists working on dislocations; accordingly, in the published literature of the theory of plasticity any initial yield stress (hardness) that an undeformed crystalline material may possess is usually attributed to various types of imperfectioils (innate dislocations, dislocatio~l mills, mosaic walls, foreign atoms, vacancies, etc.) I n particular, the observed value of the initial yield stress has heen used for estimating the number of dislocations that would be present ill the undeformed crystal on the assumption that the initial yield stress vras determined by the interaction of innate dislocations (cf. E q 3-34 iii section 3-21). I n what follo\vs, i t should be pointed out that this coilveiltionnl view of the origin of the initial yield stress now seems unrealistic. Contrary t o it, the shear stress r d required for moving a solitary dislocation in a crystal (the "dislocation driving stress") must be extremely high in many crystals, and there is no reason to espect i t to be negligible in any crystal. I n many cases, the driving stress may directly determine the initial yield stress of the crystal if this does not contain too many coarse defects or inclusions: Diamond is a p a r t i c ~ l a ~ clear examplk of a crystal whose hardness ly must be determined by the high value of the stress required for moving a solitary dislocation. If the initial yield stress were determined bj; the number of the preexisting dislocations or dislocatioil mills, its extremely high value in diamond could be explained oilly by assuming that this crystal conta;ined either no dislocation a t all, o r an extremely high number of dislocations. .The first alternative can.6e excluded a t once because the majority. of 'diamonds, .-while extremely hard, shim be-



tween crossed polarizers the typical birefringent slip bands well kno\vn from plastically deformed alkali halide crystals. 4 5 This means that such crystals must co~ltaindislocations, \vithout the presence of which ~internal stresses of this type callnot exist.-?he second alternative can be ruled out because X-ray diffraction photographs show that diamonds are usually very perfect or a t least fairly perfect crystals. A dislocatio~l dknsity such as must be present in a strongly cold-\vorked metal causes completely blurred diffraction spots and illtense line broadening, and a n even higher density \vould be needed for explaining the initial hard-. ness of diamond on the basis of an iilteraction between innate dislocations such as is represented by Taylor's equations (3-34 or 3-30). The case of silicon carbide, of the hard metal carbides and nitrides, and, indeed, of all very hard crystals can~lot very different from that. be of diamond, although the experimental evidence is not as complete as for the latter. It follo\\rs, then, that the hardness of these crystals must be due, not to the interaction of innate dislocations, but t o the high value of the dislocation driving stress rd. The reason this is so high is fairly obvious: in crystals with strong covalent binding, intense directional bonds are operative between the atoms on opposite sides of the slip plane in a dislocation, and this makes the movement of the diildcation difficult. This assumption is strikingly supported by a survey of the distribution of the haxd and soft metals in the periodic system. With the exception of the obvious marginal case of beryllium, the hard metals are those with a partially filled d - ~ h e l l and the hardest among these ,~~ are the metals \vith a n approximately half-filled d-shell: V, Cr,lR/ln, Fe, Co, RIIo, Ta, W, Re, Os, U. It is in these metals that directidnal exchange forces across the slip plane in the dislocation must be the strongest and the value of rd highest,. On the other hand, spherical symmetry of the atom, a n d a lattice strongly "blovrn up" by the electron gas pressure (i.e., by the Fermi energy playing a n important part in determining the lattice parameter) lead t o a lo\\. value. of the initial yield stress; in alkali metals, Cu, Ag, Au, the initial stress is small compared with the value t o which the yield stress can rise by strain hardening if the temperature is lo\\r enough to prevent rapid recovery or recrystallization. It can be said, therefore, that the initial yield stress, i.e., the inherent hardness, of a crystal seems to be determined mainly by the stress required for moving a dislocatio~lin an otherwise faultless crystal. I 1 1 metal crystals \\;it11 spherical dominating Fermi energy, this driving stress may be very low; such metals have, in fact, a very low




initial yield stress and the hardness they show during deformation is mainly acquired by strain hardening. This picture is much more satisfying than the old assumption of hardness being determined by defects (11 hich must be largely accidental) in spite of the manifest circumstance t h a t it is a substantially inherent individual characteristic of a crystalline material.

I n the preceding section, i t was shown that the initial yield stress of a crystalline material is far lower than the theoretical value of the shear cohesion of the crystal lattice, and t h a t i t is probably determined in a pure crystal by the driving stress r d , i.e., the stress needed for moving a solitary dislocation in the crystal. The main questions connected with the yield stress that are t o be answered now concern its value in a crystal containing inclusions of foreign atoms or particles of another phase, and in a plastically distorted and strain-hardened crys.tal. I n this section, the first of these questions will be treated, because the ansll er is simpler and more closely connected with t h e considerations of the preceding sections..Although the theory t o be given must embrace all types of material contailling inclusions (to a certain extent, even solid solutions), the case of precipitation hardening must receive particular emphasis because i t is the most general one : the solid solution and the material with microscopically visible inclusions are limiting cases, the theories of which must be contained in t h a t of precipitation hardening. The most important and first discovered precipitation hardening alloy is duralumin, ~vhich essentially aluminum with a few per cent is copper. At high temperatures, the copper goes into solid solution. Immediately after the hot solid solution has been quenched in water, its hardness is some~vhat higher than t h a t of ordinary aluminum; however, i t increases greatly and reaches a maximum after a few hours or days a t room temperature or a slightly elevated temperature, depending on the presence of other alloying elements. After the maximum has been reached, the hardness diminishes again. The first steps toward t h e understanding of the hardening process were made by RIerica, Scott, and W a l t e i ~ b e r gThey attributed i t t o .~~ precipitation from the solid solution which becomes strongly supersaturated when cooled to room temperature. However, no precipitate could be observed when the maximum hardness was reached; by the time the first trace of precipitation was seen in the microscope, the



hardness had dropped below its maximum value. Merica assumed, therefore, that the hardness depended not only on the amount ,of the precipitate, but also on the degree of dispersion, and that the maximum hardness was produced by a critical dispersioil a t which the precipitate uras still of submicroscopic size. R/Ieric,a1s first assumption (that of the presence of a precipitate) was confirmed about 20 years later by Guinier4"ind Prest0n,~9 who could demonstrate by X-ray diffraction effects the presence of copper-rich regions. A t about the same time, the use of Jacquet's electrolytic polishing technique made the direct observation of the precipitate possible in all but the initial stages of the hardening process. Thesecond assumption (that of a critical dispersion producing maximum hardness), however, presented more difficulties. As a first step, it was natural t o regard the precipitation particles as hard pegs keying to-. gether the slip planes and thus interfering with slip; this \\.as the first theory of the hardening process, put forward by Jeffries and Archer.50 However, if the hardness is determined by the resistance of the "pegs" t o shear, it ought t o be proportional to the total area occupied by the precipitated particles per unit area of the slip plane, and this is obviously a constant for a given amount of the precipitate, independent of the degree of dispersion. The first theory of the critical dispersion5' was based on the stresses developed around a precipitate when it is in atomic register with the matrix. I t is known from the X-ray work of Guinier and Preston that the hardening of duralumin starts with the segregation of copper-rich particles a few atoms thick, and perhaps 50 or 100 atoms in diameter, parallel to the cube faces of the matrix, and in atomic registlr with it. Since the natural atomic spacings of the precipitate in its unstressed state are different from those of the mattsix, the regist:r demands the development of stresses in and around the precipitate. Suppose now that the precipitated plate has a tendency t o contract; in register with the matrix, it will be then subjected t o a tensile force which is balanced mainly by shear stresses between the plate and the matrix. If the thickness of the plate increases, the tensile force rises proportionally because the strain enforced by the matrix, and thus the stress in the plate, remain approximately constant. Since the area of coiltact betweell the faces of the plate and the matrix does not change, the shear stress in these faces must rise with the thickness of the plate until it reaches the value of the shear cohesion and the plate breaks away from the matrix. When this occurs, the stresses drop t o a lower level.





FIELD. a . The shaded area around a hard inclusion shows how its effective slipstopping diameter is increased by the internal stresses it creates. b. Shows that the corresponding increase of the effective diameter of n lamellar inclusion is snlall because the internal stresses are apprcciablc only in a region around its edge the linear diincnsions of which arc comparable with the thickness of thc lamella.

The theory further assumes that the increased hardness (yield stress) is due to the circumstance that the stresses developed around the precipitated particles oppose themovements of dislocations; consequently, slip can progress only if the applied stress is large enough to overcome the adverse internal stresses due to the atomic register. Expressed quantitatively, the yield condition in this theory is


yield stress)




where T, is the adverse internal shear stress near the precipitated particle. The effect of the surrounding stress field consists in increasing the effective dislocation-stopping diameter of the particle from d to a larger value d,lr, where d,, depends on the applied stress (Fig 3-27a). In the case of duralumin, the particles are not spheres but thin lamellae (Fig 3-270). The internal stresses are then of considerable magnitude only a t the edges, and the effective depth to \vhich they extend into the matrix is of the order of the thickness of the lamella. Since the lamella is parallel t o the cube face of the matrix, whereas the slip






The inclusions left behind encircled with dislocation loops.

planes of aluminum are the octahedral planes (indicated by horizontal arrows in Fig 3-27b), the increase of the effective dislocationarresting cross section of the plate by the surrounding internal stresses (see the dotted contours in Fig 3-276) can amount to only a few per cent. For this reason, the explanation of the hardness of duralumin based on the regist5r stresses, and the interpretation of overaging as a consecluence of the breaking away of the particles from the matrix, seem t o be problematic. A different possibility for explaining precipitation hardening, the phenomenon of averaging, and the existence of a critical dispersion producing maximum hardness" is indicated in Fig 3-28, which represents a vie\\- upon a slip plane with embedded particles of a hard precipitate. Slip is assumed t o have progressed up\vard in the plane




of the figure and arrived at the wavy dislocation line, which is held up a t a number of precipitated particles and bulges for\vard bet~veen them. The salient point is nojv that the dislocation line can move over the slip plane without having to shear across the precipitated particles or overcome their stress field in the way envisaged in the former theory. It can do so by bulging for\vard between the particles in the way shown in Fig 3-29, until finally neighboring bulges fuse together to a dislocation line that moves away and leaves the particles encircled by dislocation rings. What is the shear stress required for this process? The




answer was given in section 3-13 when the bulge extrusion mechanism was discussed: it was first obtained in the theory of precipitation hardening, before the dislocation mill mechanism emerged. I t was shown in section 3-13 that the shear stress required for extruding a bulge reaches a maximum when the bulge becomes semicircular, and diminishes, when it extends beyond the semicircular position. The shear stress necessary for obtaining the semicircular shape, after which fusion with neighboring bulges can occur, is given by

if the dislocation driving stress can be neglected, and by

3-MECHANICALPROPERTIES if it cannot. Replacement of section 3-13) :


by its approximate value gives (see

for the first case, and

for the second. I n these equations, r is the free energy of the dislocation line per unit of its length, b the interatomic spacing in the slip direction, q a dimensionless factor of order unity, and a! the distance between two adjacent obstacles. For a given quantity of the precipitate, the spacing d increases with increasing coagulation (decreasing number) of the particles; the yield stress (the bulge extrusion stress) ~villthen decrease also. This is the interpretation of the fall in hardness after the maximum has been reached, But why does the hardness not continue to increase according to these equations as d diminishes? Why is the solid solution, for which d has its smallest possible value, softer than the hardened alloy? That there must be a lower limit for d a t which the applicability of these equations must break down can be seen a t once. The precipitated particles have a finite shear strength; when this is exceeded, they are sheared through or torn out of the matrix. For a given total volume of the precipitate, the average total cross section of the particles per unit area of the slip plane is independent of their number; with increasing degree of dispersion, therefore, the point must arrive beyond which the bulge extrusion stress exceeds the load-carrying capacity of the precipitated particles in the slip plane, and then the yield stress is no longer determined by E q 3-23. In fact, the strength of the inclusions is bound to be exceeded sooner or later during the plastic deformation of the crystal: every dislocation moving past an inclusion ellcircles it with a dislocation ring, inside which there is an increased shear stress that can be calculated from formulae of the type given in section 3-22.55 With increasing number of the encircling dislocation rings, the local stress in and around the inclusions rises until breakdown occurs: this may be the origin of the ~vell-kno\\rn jagged stress-strain curve of moderately aged duralumin. An ingenious argument to show \vhy the yield stress ought to increase with the spacing of the particles in the range of very small values of d has been put forward by N a b a r r ~ If~the inclusions are very . ~



close together, the dislocatioil callnot bulge between them because it has a finite breadth and a certain flexural rigidity. A segment of the dislocation, instead of threading its way het\veen the inclusions, will lie with its finite breadth over a number of them; the resulting shear stress exerted upon i t by the stress field of the inclusio~ls may be either positive or negative, and the dislocation progresses under the applied stress until its segments are stopped by accidentally high negative balances of the locally fluctuating internal stress. The yield stress is then the stress required to overcome the probable value of the accidental adverse balance of stress in the area covered by a segment. According t o a \\-ellknown theorem of probability, the probable value of this balance is ~ l inversely p r ~ ~ o r t i o i toa the square root of the number of inclusions per unit area of the slip plane; consequently, for a given amouilt of precipitate, the yield stress decreases with increasing dispersion in the range in \vhich the breadth of the dislocatioil belt is large enough compared with the spacilig of inclusions. This is the reason why, in particular, the solid solution is softer than the precipitation-hardened crystal. I t is easy to recognize the relationship between the internalstress theory and the bulge-extrusion theory of precipitation hardening. There are, in general, internal stresses around a n inclusion, particularly if i t is in atomic register with the matrix. If there are such stresses, their effect is t o increase the effective dislocation-arresting radius of the iilclusion and thus its hardening effect. Ho~vever,the presence of internal stresses is not necessary for hardening t o arise; in addition, it has been mentioned already that the effect of internal stresses cannot be very substantial if the inclusion is a thin lamella, unless i t lies in the slip plane. The principal gap in the internal-stress theory of precipitation hardening \\-as that in it the possibility of dislocations moving along the slip plane without directly overcoming the internal stresses of the inclusions, i.e., the possibility of dislocations going around the inclusions or their stress peaks and being extruded bet~veenthem, was not taken into account.

With increasing amount of slip, further deformation becomes more and more difficult, i.e., the yield stress rises; this is strain hardening. Although i t is one,of the fundamental features of plastic deformation in crystalline materials, in most cases i t is unknown which of a number of possible structural causes are mainly responsible for it in a given mate-



rial under given conditions of deformation. The main exception from. this uncertainty is the case of strain hardening in polycrystallille metals with only one plane of easy slip (Zn, Cd, RIg). This is due mainly t o the mutual interference of neighboring grains: single crystals of such metals are very much softer, and they strain-harden very much more slonrly, than the polycrystalline material. I 1 cubic metals, on the other 1 hand, the different orientation of the grains is a far less important source of hardness and hardening, and the single crystal is not much softer than the annealed polgcrystalli~le material. The seasol! for this was made clear by G. I. (no117 Sir Geoffrey) Taylor.53He has shown t h a t a crystal must have 5 slip elements (combinations of slip plane and slip direction) in order t o be able to undergo any arbitrary uniform deformation; a grain in a polycrystalline aggregate, therefore, requires a t least 5 slip elements to accommodate itself by slip to the deformation of its neighbors. Face-centered cubic metals ,have 12 slip elements (4 octahedral slip planes with 3 dodecahedra1 slip directions in each), so that they satisfy this condition; hexagonal metals, on the other hqnd, have only one plane of easy slip (the basal plane) with three slip directions. * The mutual interference of neighboring grains is the only source of strain hardening t h a t is fairly well understood; the mechanism of. intracrystalline hardening is far from being clear. I n the course of the past 20 years, several theories of strain hardening have been put for~vard, each attributing it t o a particular structural change; it is not known, ho~vever,which mechanism is the most important cause of hardening in a given material, nor is it certain that the most important mechanisms are among those so far suggested. A review of the main possibilities will be given first. Until the early 1920s, only very vague views were expressed on the cause of strain hardening; i t nras attributed in a general way to injuries of the lattice t h a t result from plastic deformation and act as obstacles

* Thc reason why polycrystallinc zinc or cadmium is ncvcrtheless fairly duct~le is t h a t the grains can undergo mechanical twinning and kinking in addition to slip; by ineans of kinking, in particular, they can in principle accommodate themselves to any deformation of their surronndings. The high initial yield stress of polycrystalline hexagonal metals, as cornpared with the softness of the corresponding single crystal, is mainly a consequencc of the fact that many grains have no favorably oriented slip plane for a glvcn dircction of tension or compression. In cubic mctals, this can never be so: there is always a slip plane available with an orientation far cnough from being parallcl or pcrpendicular to the direction of tension or compression.



to further slip. I t \\;as imagined t h a t during the deformation fragments were torn out of the lattice and rotated so that they could not conform with the slip in their surroundings; they \vould then act as "keys" locking'together the slip planes. This view no doubt contained some truth, but it was too diffuse to he, very useful. I t seemed discredited for a time when G. I. Taylor shonred, in the investigation just mentioned, that the number of slip elements in a cubic crystal was sufficient to enable any rotated crystal fragment to conform with the deformation of its surroundings without requiring a very much higher stress; in other ivords, that the fragments could i;ot have acted in the ivay just mentioned as keys hard enough to produce the observed amounts of strain hardening. This conclusion, ho\vever, implied that the yield stress of the fragments was the same as t h a t of bulk material. I n reality, a small fragment could hardly contain dislocations, and thus it could be under certain circumstances as hard as the ideally perfect lattice. The first clearly outlined suggestion of a mechanism of strain hardening mas put forward by G . I . Taylor6 in 1934. Stripped of its mathematical paraphernalia, i t amounted to the assumption that even before deformation a crystal consisted of very small domains separated by walls that slip could not penetrate. When deformation began, these walls would not interfere with it strongly a t first; with increasing distortion, however, higher and higher stress was necessary to enforce further slip against their resistance. Thus, the Taylor theory contained an element of t h e old "hard key" hypothesis, with hypothetical mosaic \valls playing the role of the obstacles to slip. Its particular distinguishing feature was that in it the obstacles \vould not arise during the deformation but \\:auld be present in the crystal from the beginning. For various reasons, this is unlikely to be the case with pure crystals, but a corresponding hardening effect does exist in mateiials that contain innate obstacles to slip, such as'precipitates of hard particles or perhaps grain boundaries with hard precipitates. A recent suggestion by Fisher, Bart, and Pry54explains the additional strain hardening due t o the presence of a hard precipitate substantially on the basis of the Taylor theory, with the precipitated particles replacing the hypothetical walls as obstacles to slip. An entirely different possibility of strain hardening \vould arise from the removal or inactivation of innate lattice defects t h a t make the crystal soft, i.e., of dislocations or dislocation mills. Without the presence of a multiplication mechanism (section 3-12), a crystal ~vould harden extremely rapidly as the initially present dislocations were used

up by slip: after a total slip by half as many interatomic spacings as the number of dislocations initially present, * the crystal ~vould cleared be of dislocations and it would assume the hardness of the ideally perfect lattice if there were no other stress concentrations in it. This process could be called strain hardening by the eshaustion of the initial supply of dislocation^.^^ A similar effect npuld arise if, instead of leaving the lattice, the dislocations or dislocation mills n-ould become t r a ~ p e d or otherwise inactivated after a certain amount of slip. I n fact, the electron-microscopic observations of Shockley and H e i d e n r e i ~ hand of ~~ A. F. B r o ~ v show that the operative slip planes in aluminum become n~~ inactive after a fairly sharply determined and fairly temperatureindependent amount of slip. Strain hardening, then, might be attributed to the gradual exhaustion by immobilization of an initial supply of dislocations or dislocation mills. The first possibility (exhaustion of an initial supply of mobile dislocations) can be disregarded: as already mentioned (section 3-12), the number of dislocations needed for producing large amounts of plastic deformation is far too high to be present initially in the crystal. The remaining alternative is then the assumption in the undeformed crystal of an initial stock of dislocation double mills with different values of the width d (section 3-13). At the begin: ning of the deformation, the double mills with large spacings and correspondingly low operating stresses (see Eqs 3-22 or 3-23) would be used up and immobilized first, and the stress n-ould rise during the deformation as mills with lower and lower values of d 111ould have t o be utilized. This possibility concerning the origin of strain hardening cannot be ruled out entirely, but it seems very unlikely t o play a considerable role. If the shape of the stress-strain curve ~ v a determined by s the number of the innate dislocation mills and the distribution function of their spacings, it would depend on the preparation of the crystal far more than it actually does. Besides, there is little doubt that dislocation mills do arise during the deformation, and that their number in the undeformed crystal may be quite small; strain hardening, then, cannot be attributed reasonably to the exhaustion of a fixed initial stock. It seems, therefore, that intracrystallii~e strain hardening cannot be due t o the factors that have so far received detailed theoretical consideration; i.e., t o innate obstacles t o slip or to the eshaustion of an

* The factor one half results if it is assumed that the dislocations are initially over the crystal, so that the mean area swept by a dislocatioii uniformly distribr~ted before it leaves the crystal is one half of the area of the slip plane.




innate stock of dislocations or dislocation mills. I t appears probable that the main source of hardening in a crystal not containing hard inclusions is the creation of lattice injuries acting as obstacles to slip in the course of the deformation; in other words, the hazy picture of 20 years ago seems to contain more truth than the cpantitatively elaborate but physically oversimplified theoretical schemes put forward since. I t must be asked, then: Can one point out mechanisms by which obstacles to slip are created as a direct consequence of the deformation process? The simplest example of the productioil of lattice imperfections other than dislocatio~ls during slip is the generation of vacancies in the way described by Seitz (see Chapter 1, by Koehler and Seitz, in the present volume) and in sectioll3-14 of the present chapter. I n a similar manner, excess (interstitial) atoms may come into the lattice; in addition, of course, twin and kink lamellae may, and in general will, arise locally, particularly i n a polycrystalline material. Another mechanism of hardening is the cutting across of dislocations by slip taking place in planes pierced by them, as was described in the discussioll of the Z-mill (section 3-15). Even before the width of the Z-mill becomes large enough t o allow its operation a t the shear stress present, the severed ends of the cut dislocation are conilected by a dislocation in the intersecting slip plane; the length of this conilecting dislocation is initially approximately equal t o the amouilt of slip in the intersecting plane. Since a dislocation has a free energy r per unit of length and thus behaves as if it had a tension r, slip in a plangpierced by fixed dislocatioils is opposed by the tension in the conilectiilg dislocatio~ls creates. it If the slip is pierced by n dislocations per unit of its area, a shear stress nr is required to overcome their oppositioil to slip. Obviously, the tension in the connecting dislocations acts as an internal stress and must give rise to a Bauschinger effect; in addition, this mechanism of hardening leads to a simple explanation of thermal softening without recrystallization (see section 3-23). A probable source of hardening that has emerged from recent Xray work by B. E. Warren is the presence of stacking faults i11.cubic face-centered crystals. I t is probably less significant a t moderate degrees of distortion, but it may be its maill cause a t very severe strains. I n the follo~\;ingsections, the Taylor theory and the production of strain hardening by the intersection of dislocations and of stacking faults mill be treated in greater detail. T h e first is discussed because Taylor's work has introduced into the theory of plasticity several con-




cepts and mathematical methods that are of great importai~ce indepeildeiltly of the particular model of strain hardening to which they were first applied, and because Taylor's model, although.not giving a realistic picture of strain hardening, does contaii~ the basic elements that are responsible for the additional hardening dbe to obst,acles such as grain boui~dariesand inclusions. The secoild and third mechai~ism (that of the intersection of dislocations-and of stacking faults by crossing slip planes) will be discussed because these possibilities have not yet received a treatment in the literature.



I n his 1934 paper, Taylor attributed strain hardening to the obstructioil, of the movement of dislocations by adverse internal stresses produced by other dislocations. On this basis, the yield stress of a strain-hardened crystal ~vould equal to the highest adverse internal be stress ri that impedes the movement of a dislocation:

The Mott-Nabarro theory of precipitation hardening (section 3-19), therefore, uras based on the same general idea as the Taylor theory of strain hardening. The quantitative execution of Taylor's theory rested on the assumption of the two-dimensional model arrangement of infinite straight dislocations shown in Fig 3-30. The dislocations are assumed to form a rectangular lattice, ~vhich supposed to remain geometrically similar is to itself during the deformation. I n what way such an arrangement could arise in reality and how its geometrical proportions could maintain themselves was not discussed. Like the electric field of an infinitely long line charge, the elastic stress field around a straight infinite dislocation is inversely proportional to the ,distance r. As a mathematical consequence of this, the shear stress 7; = Y required for moving one of the dislocations in Fig 3-30 in its plane against the resultant shear stress of all other dislocations to another equilibrium position is inversely proporti&al to the spacing-of the dislocations and thus proportional to the square root of their number n per cm2: 7 . = y = const. = C < n [3-341



if i t is assumed with Taylor t h a t a solitary dislocation can move in a faultless lattice under a negligible small shear stress (cf., however, section 3-18). Taylor assumed t h a t the undeformed crystal contailled no dislocations, but dislocations could enter a t its surfaces without any difficulty; on the other hand, they were not permitted to leave the crystal a t the other end of the slip plane. From this, the plastic strain undergone by a

1 TT

T7 --

T r

1 T

1 T -

1 1 T T -

1 T---


1-IT T -

1 T -

1 -






G. I .


crystal containing a given number n of dislocations per square centimeter arrahged as in Fig 3-30 follo~vs directly. Since all T dislocations (cf. section 3-5) must have entered the crystal (under the shear strain indicated by the arrows) a t the right-hand face, and all 1 dislocations a t the left-hand face, the mean path of a dislocation is one half of the width of the crystal. If i t were the full ~iridth, the slip per dislocation ~vouldbe one interatomic spacing b in the slip direction; but a s the mean area covered by the slip is only one half of the entire slip plane, 0 the shear strain of a crystal of width w is only wn -where, as above, 12 is

the number of dislocations per square centimeter:


Elimination of n from Eqs 3-34 and 3-35 gives

This is the equation of a parabolic stress-strain curve which represents fairly well the observed stress-strain reJationship for many metals. Ho~vever,the equation has two shortcomings: First, the stressstrain curve depends on the dimeilsions of the crystal; second, the order of magnitude of the hardening as given by E q 3-36 is wrong. The magnitude'of the constant C follows from the expressions for the elastic field around a dislocation, and with its value the rate of hardening given by E q 3-36 turns out to be negligibly small compared with the observed magnitude. Taylor circumvented both difficulties by an ad hoc assumption: he postulated that the dislocations could not move freely in the crystal because this was subdivided into mosaic cells by ~vallsthat the dislocatiolls could not penetrate, although they were supposed to be geilcrated for each cell a t its ~valls during the deformation. If the spacing of the mosaic walls is L, the mean path of a dislocation is only L/2, instead of w/2 where w is the width of the crystal.. The shear strain a t a given dislocatioll density is then only Llw times the value given by E q 3-35 and this makes 7 dependent only on ri but not on w. I 1 order to obtain agreement with observed stress-strain 1 curves, L had to be assumed to be between and cm; this magnitude fitted well into the hypotheses of mosaic structure that were in vogue in those days. Contrary to the first impression, the strain hardening in the Taylor theory did not result primarily from the interaction of dislocations; its ultimate cause was the presence of hypothetical walls impermeable to slip, the spacing of which nras assumed so as to give the required rate of hardening. The interilal stresses which are produced by the dislocations and which impede'their movement can be compared with the compressive forces in the buffer springs of a row of railroad cars when it runs against a buffer block. Although every car is stopped by the elastic force acting in its buffer springs, the ultimate cause of the process is the presence of the buffer block; similarly, the real cause of the hardening in the Taylor theory was the presence of hypothetical walls a t which slip stopped. The interaction of the dislocatiolls accumulating a t



these malls determines only the quantitative details of the process, such as the shape of the stress-strain curve. I n its original form, the theory met n ~ i t h other difficulties. The most drastic of these is that the.dislocatioil density lz cannot be increased gradually if the similarity of the dislocation lattice is to be maintained. I t is easy to see that the nearest geometrically similar lattice obtainable from that in Fig 3-30 by the addition of dislocations must contain 9 times more dislocations per square centimeter: in order t o increase the density, one has to insert a t least t ~ v o planes of dislocations (one with 1 and one with T dislocations) betn~eenpairs of esisting ones, and then the number of dislocations in each pla~ie must also be increased threefold before Eqs 3-34 and 3-36 become applicable again. I n other nlords, the plastic strain must increase ninefold before the stress-strain relationship E q 3-36 becomes valid again. While the neu7 dislocations are added, the yield stress can rise or fall steeply, according t o whether new planes are opened up, or the old ones filled up, first. Another shortcoming of the model in Fig 3-30 is that i t contains no element of the irreversibility so typical of plastic deformation. If the shear stress mere reversed, the dislocations ~vouldleave the crystal yhere they.have entered, and the yield stress \voulcl fall t o zero along the stress-strain curve E q 3-36. This does not mean t h a t the model in Fig 3-30 would lead t o a large Bauschinger effect: since the yield stress calculated on its basis does not depend on the direction of strainillg, i t leads t o full reversibility without any Rauschinger effect. Quite recently, the Taylor theory has been extended and elaborated by M ~ t with the utilization of many concepts developed in the t ~ ~ theory of dislocations since 1934. Mott suggests that the obstacles a t which slip is stopped and dislocations pile up are holes or channels in the lattice formed in the course of slip. S e i t has emphasized the possi~ ~ ~ bility of vacancies "evaporating " from moving dislocations ; a way in which sucha processmay occur on a larger scale has been pointed out in connection with Fig 3-19. Another mechanism of formation of vacuities is shown in Fig 3-31.; if slip occurs in planes 1 and 2 simultaileously in the way indicated, 'the wedge bet\veen these planes moves up~vard and a hollow channel may be formed a t its edge if the two dislocations happen to meet on their way do\vn\\:ard. This process is frequently observed in the Bragg-Nye bubble model (Lomer," Cottrell58) ; it is difficult to estimate ho\v frequently i t is likely to occur in a threedimensional crystal. More effective than vacuities are probably extra rows or lamellae of



atoms which call arise by the same process of slip-step formation as shown in Fig 3-19. The first suggestion that strain hardening may be due to step formation in the manner just described was made by R. Becker. l 7 -




As an appendix t o the foregoing sketch of the Taylor theory, some characteristic features of the elastic field around a dislocatioll should be discussed. The first solutioil of the stress field was obtained by T i m ~ e his ;~ equations were used by Taylor. Later, it was noticed by J. i\iI. Burgers4 and Koehlersgthat'an iilternal boundary conditioil in the Timpe solution did not correspond t o the conditio~is a dislocation, and they obin tained the equations for the stress field that are used a t present. The most important of their results gives the shear stress produced by a dislocatioll a t a point P in a parallel slip plane:

where v is Poisson's ratio r , the distance between the dislocatio~land the point P ; b, the interatomic spacing in the slip direction; G, the shear modulus, and 0 the angle Det~veen the slip direction and the ra-



dius vector from the dislocation to the point P (Fig 3-32). This expression can also be written as

where t is the spacing between the plane of the dis1ocat;ion and the parallel plane through the point P. The maximum shear stress occurs a t 9 = 22.5", and its value is

Ecl 3-39 sho~vs that the shear stress vanishes a t 9 = 45O, 90, and 135", and it changes its sign a t these points. Thus the ttvo maxima a t 22.5"

FIG~ - ~ ~ - R E L A T I X G TO THE




and (180" - 22.5") are of opposite sign; a t one of them, therefore, the stress exerted by the dislocation acts against the movement of another dislocation and arrests it unless the external shear stress exceeds the value given by Eq 3-39. This interaction of dislocations, upoil which the Taylor theory and the Mott-Nabarro theory of precipitation hardening were based, has a very simple physical meaning. Suppose that in Fig 3-33 the dislocation 1 has started to move from the left-hand face of the crystal, and a t the same time the dislocation 2 from the right-hand face. Unless the










The two cracks in the center have relieved the shear stress and thus prevented further propagation of the crack in the upper right-hand corner.



applied stress exceeds the value given by Eq 3-39, the two dislocations immobilize each other before the angle e reaches 22.5". No\v a slipped area boun'ded by a dislocation is closely related t o a shear crack; the difference hetween them is that the shear displacement in the slip plane is col~stant = b) escept in the neighl~orhood the center of the ( of dislocation, while in the shear crack i t increases steadily with the distance from the end of the crack. For the present problem, the difference betnreen the shear crack and the dislocation is not an essential one. If, now, Fig 3-33 is mea;nt to refer t o the propagation of shear cracks instead of slip, the reason for the mutual "attraction" of the ends of the cracks becomes directly obvious: crack I. is arrested hy crack 2 because the latter has annihilated the applied shear stress in its surroundings and thus had made the propagation of the other crack impossihle. This phenomenon is seen everywhere in natural rocks; Fig 3-34 sho~vs n a example of mutually arresting shear cracks in t,he coastal rocks a t Cape Ann, Massachusetts.





I strain hardening by the exhaustion or inimobilization of an inf nate stock of dislocations or dislocation mills is putaside as unlikely to play an important role, the only mechai~isms 11a;rdeaing so far disof cussed in the literature are those based on the presence of obstacles t o slip. I n G. I. Taylor's original theory, the obstacl6s &ere assumed to he innate mosaic it~alls; materials cont$ining inclusions grain boundin aries, these act as obstacles to slip and gs contributory causes of strain hardening. I n pure single crystals, on the other hand, the obstacles must arise during the deformation itself; they may be vacancies, cavities, interstitial atoms, intruded atomic rows or planes, or other irregularities (see section 3-14). In the present section, an entirely different mechailism of strain hardening 'will be discussed; as a n order-of-magnitude estimate of its effectiveness will show, it may possibly play a significant role in the hardenifig proc'ess. The principle of this mechanism is illustrated'in Fig 3-35. I slip occurs in the plane 1-1-1-1, all sufficiently immobile f dislocations traversing this plane are cut into two parts a-b and c-d, which are moved apart by the slip as indicated in the figure. This intersection process has already been considered in section 3-15: i t nras the mechanism by which Z-mills were generated. Now, however, attention should be given to its first stage, before the Z-mill becomes operative.'



As explained in section 3-15, the severed ends b and c of the intersected dislocation are coilnected in the operative slip plane by a dislocation b-c, which m a y be of the edge or screw type according t o the nature of the intersected dislocation and the direction of the intersecting slip. If t h e amount of the intersecting slip is small, the co~lnecting dislocation b-c is kept nearly straight by its " capillaryJ1tension, which a t the same time opposes further slip. T h e amount by which the capillary tensioil of the connecting dislocations increases the shear stress required for further slip is easily calculated. If the free energy of a connecting dislocation per unit of its length is I? (cf. section 3-G), a differential


s m N c C U T B Y s L I r I N I N T E R S E C T I N G S I J I P P1,ANE


increment dx of slip increases its total free energy by Fdz, which has t o be contributed by the applied forces. Consequently, every di:;location intersected by a slip process increases the force needed for slip by a n amouilt r ; if the number of immobile dislocatiolls tr~tversinga slip plane is 7% per unit of its area, their intersection gives rise t o all. additional shear stress n r required for slip in this plane. Since the number of dislocations traversing any slip plane increases in the course of plastic deformation, the yield stress must also rise, i.e., strain hardening must" occur. T h e possibility of this effect being a significant source of hardening \\:ill be esamined in the present section. Before discussing details and consequences of the intersection process, it should be realized that a small amount of slip can take place in the plane 1-1-1-1 (Fig 3-35) without cutting across the traversing dis-




locations. I n this case, they act as obstacles to slip, just like the precipitated particles in the \vriterls theory of precipitation hardening (section 3-19), aiid they increase the yield stress by the amount given by Eq 3-23, which represents the geileral condition for the extrusion of a dislocation through a field of ,obstacles. With every 'dislocation passing over the plane 1-1-1-1, every tra\~ersiiigdislocatioll becomes encircled by a dislocation ring. As this process continues, the shear stress a t the points of intersection of .the (raversing dislocations Irith the operative slip plane rises, and finally the traversing dislocatioils Are cut across. After this has occurred, the increase of the yield stress is given by n r instead of Eq 3-23 (in which, of course, d I/&). It is easy to obtain a rough estimate of the order of magnitude of the hardening effect due to the iiltersection of dislocations. It is usually assumed that the number a of dislocatiolls that cross a surface of unit area in a moderately deformed crystal is around 101 or 1011 per cm2; for a strongly deformed crystal, n nrould amount to 1012 or more. According to Nabarro's calculatioils (see section 9-7), the free energy of a dislocation per,interatomic spacing is of the order of a few. electron volts-say, 5 X 10-l2 ergs; its magnitude r per centimeter, therefore, is perhaps 2 or 3 times lo-! ergs. The amount of hardening n r due t o the intersection of dislocations is then a few times lo7 dynes per cm2 for .a moderately, and a few times l o Y dynes per cm2for a strongly deformed crystal. This agrees with the observed orders of magnitude of the yield stress of soft single crystals, so t h a t the intersection mechanism may in fact play a role in the generation of hardening. I the mechanism by which n increases \vith illcreasing deformatioil f were kno\vn, and a relationship bet~veeo and the plastic strain could n be established, the rate of strain hardening due to the intersectioil mechanism, and the corresponding "stress-strain" curve, could be calculated. Ho~vever, such calculatioil would a t present require a numany ber of highly arbitrary assumptiolls which could hardly be justified by any agreement of the result with observed stress-strain curves. It may be asked whether it is permissible t o assume the presence of many dislocatioils traversing the visually observed slip zones when no correspondiilg intersecting slip zones can be seen. As already mentioned (section 3-14), large amounts of slip may occur unaccompanied b y visible slip lines; there seems to be little doubt that, even if only one set .of slip zones is observed, a considerable part of the deformation may have taken place by slip in ilitersecting slip planes to which no visible slip lines correspond.



T h e intersection mechanism can explain a t o n c e an effect that has always been a difficult problem to theories of hardening: the hardening of "latent" slip planes intersecting the operative slip zones. If slip in a cubic crystal starts with the developmelit of only one set of slip lines, the corresponding slip planes may rotate into an increasingly unfavorable orientation with respect to the applied stress, and fillally the ref solved shear stress becomes higher in an alternative slip plane. I this plane had not been strain-hardened before i t started to develop slip zones, slip. in it would have started long before the resolved shear stresses became equal in the two planes. I11 reality, ho~vever, beginthe ning df double slip is delayed sometimes eve11 beyond the point of equality of the shear stress in the two planes: in other words, the "latent" slip plane is hardened, often more than the. "operative" plane. This \vould immediately follo~vfrom the intersectioil mecha-. nism: all dislocations accumulating in the operative slip zone mould give rise to intersective hardening in the alternative slip planes. The large amount of slip taking place in the operative zones, of course, does not necessarily involve the presence of a proportionate number of dislocations: if the concentration of slip in a zone is due t o a dislocation mill, any amount of slip may in principle arise in it with only one or two dislocations present. I n hexagonal crystals like zinc or cadmium, the basal plane (0001) is the only plane of easy slip, and i t is plausible to assume t h a t dislocations traversing this arise only slowly during deformation (e.g., by slip in prismatic planes). This \\:auld account for the striking difference between the.slo\v rate of hardelling of such crystals and the rapid hardening_of cubic crystals with several equivalent slip planes (SchmidGO). Andrade and HendersonG1 have recently observed t h a t the stressstrain curves of carefully produced very pure crystals of silver aild gold begin with an almost horizontal part; only after a few per cent of strain does the typical rapid hardening of cubic crystals set in. This may be the consequence of a n initial absence of traversing dislocations ill intersective hardening. Another phenomenon t h a t has presented difficulties t o earlier theories, and which is immediately uilderstood if the intersection mechanism plajls an important role, is the Bauschiiiger effect in single crjrstals (in polycrystalliile metals, the accumulation of dislocations a t grain I~oundaries a plausible source of the Bauschiuger effect). The tension is of the connecting dislocations 0-c in Fig 3-35 counteracts slip in the initial direction but a t the same time promotes it in the opposite direc-




tion. I t is iilterestiilg that this source of the Bauschinger effect is not connected 1 ith the presence of internal stresses (apart, of course, from 1 ~lg the submicroscopic stresses in the c o ~ l ~ l e c t idislocation itself). It is of the same nature as the (usually unmeasurably small) Bauschinger effect due t o surface energy, the origin of which is illustrated in Fig 3-36. This shon-s a crystal in \vhich slip has taken place along a certain f plane of which the surface energy is a . I the amount of slip increases by dx, an additional area 2b . dn: of the slip plane is laid bare where b is the width of the slip plane perpendicular t o the slip direction; consequently, an additional surface energy a . 22, . dx is created. T o produce this






through a displacement dx, a shearing force 2ab is required; since this force is determined by the interatomic forces acting in the slip direction across the slip plane, it is present irrespective of the presence or absence *of an external force. Consecluently, the surface force increases the applied force needed for slip by the amount 2ab \\-hen i t acts in the direction of increasing the surface area of the crystal, and it reduces the force needed for reverse slip by the same amount.

A very interesting mechznism of strain hardening lvhich is likely t o be of considerable importailce in intensely deformed face-centered * T h c present scc:tlon arosc o u t of a fortunate ohance mccting with Prof.
13. E. Warren, a fc\v hours aftcr t h e manuscript of t h c report \\,as delivetcd for



cubic crystals has quite recently emerged from X-ray studies. of the effect of plastic distortio~lcarried out by B. E. Warren* and M. S. Paterson f. T h e broadening of monochromatic X-ray diffraction lines (DebyeScherrer circles) due to cold ~\:orkhas beell studied este~lsively the in course of the last 30 years. Until no\\:, it was attributed either to microscopic int,er~lal strains or t,o a breaking up of the crystal leading to small size of the coherently scattering domains. I t has become iilcreasingly certain that, as a rule, the size effect cannot be of more than secondary importance, SO that inter~lalstrains have come to be coilsidered the priilcipal cause of line broadening. Itecent work of Warren, however, in striking agreement \\:it11 theoretical studies of Paterson, has led to a radical change of the situation. I t has shown 116th a high degree of certai~lty that a coilsiderable part of line broadening (perhaps one half or more) in heavily distorted 70-30 brass filings is due to the massed development of ''stacking faults." This result is of such importance to the general theory bf slip, and in particular to that of strain hardening, that a somewhat detailed accoullt seems advisable. Fig3-A shows the scheme 011 which a face-centered cubic lattice is built. Like the close-packed hexagonal lattice, it represents'a densest packing of spheres and collsists of layers of close-packed sheets of atoms, which, in the cubic lattice, are the octahedral planes (111). The cubic and the hexagonal lattices differ only in the y a y in which these sheets are stacked upon one another. The circles in Fig 3-A represent the atoms of one close-packed sheet; those of the follo\\~ing sheet lie in the hollows between atoms of the first layer. They can occupy either the sites marked with B, or those with C. Suppose they are on the sites B: The atoms of the third close-packed layer, then, can be' placed either on site A or site C. I n the first case, a hexagonal close-packed lattice results in which the atoms of the third layer lie above those of the first; the second case gives a face-centeredcubic lattice. I n this, the atoms of each three consecutive layers occupy the positions A , R, and C; those of the fourth are again on the sites A, i.e., over the atoms of the first layer. Figs 3-B, 3 4 , and 3-D show a stack of ping-pong balls in a facepublication. To avoid the risk of error in references to publications, figures, and sections, it was decided to trcat this section as an indcpcndent unit with independently numbered references and figures. * B. E. Warren: private communication; to be published. t M. S. Paterson: J. Applied Phys. (1952) 23, 805.




A, B, C,




B. Face-centered cubic arrangement of ping-pong balls; the front face, as well as the horizontal layers, are close-packed (111) planes. C. Right-hand face of the stack shown in Figure B (octahedral plane). D. Left-hand face of the stack shown in Figure B (cube plane; the cube edges are at 45 to the horizontal rows). E. The marbles show consecutive positions during slip of an individual atom in the close-packed plane lying on the atomic plane indicated by ping-pong balls.




F. Close-packed cubic stack of ping-poing balls after single-step slip between the second and third layers from below. G. The stack shown in Figure F seen from the right. H. The stack shown in Figure F seen from the left. J. The stack shown in Figure F seen almost in the direction of a closepacked row.






c'entered cubic arra~gement.The horizontal layers are close-packed (octahedral) sheets; the front face in Fig 3-B, as ~vell askhe right-hand side of the stack, are also octahedral planes, while the left-hand face is a cube plane. Fig 3-C sho\?s the right-hand face, Fig 3-D the left hand one; in the latter, the edges of the face-centered cube are indicated by dotted lines. The slip planes are the octahedral ones; in them, the three slip directims are the close-packed lzttice rows. Fig 3-E sho~vs closea packed sheet, and marbles marking on it the consecutive positions-of an individual atom of the next layer during coiltiilued slip. I t is seen that slip 'occurs along zigzag path, following the way of the lo~vest passes between neighboring atoms; every displaceineilt by one identity period in the slip direction coilsists of two consecutive steps. Comparison with Fig 3-A shonrs that one step brings an atom (and thus the atimic sheet) from, say, a B site to a C site; the sites occupied .by the coi~secutive layers after such a.step are no longer -4BCABC . . . but ARCACABC . . . . Thus, the one-step slip has caused a stacking J'au.Zt in the crystal: four consecutive sheets (CAC.4)-around the plane'of slip have nonTa hexagonal, instead of a cubic, arrangement. The next step, completing the slip to one identity period, brings the lattice hack t o the perfect cubic one. . Are the two steps of which the full identity translation consists carried out almost simultaneously, or after a significant time interval? Shockley and Heidenreich3!'have pointed out that the secoild altenlative must prevail, for the fillowing reason. If the entire identity translation were carried out a t the same time, the boundary of a spreading area of slip would be a dislocation whose slip vector mould be equal to the identity period. I n the opposite case, the area of slip must be surrounded by two dislocations loops, each representing slip by one step. Eq 3-14 in section 3-8 sho~vs that the energy of a dislocation line per unit of length is approximately proportional to the square of its slip vector b ; the energy of the one-step dislocation, therefore, is nearer to one fourth than to one half of the energy of the full dislocation with the identity period as slip vector. Consequently, the energy is lowered when the full dislocation dissociates into two L'half-dislocations," and the total energy of the two half-dislocations is the lower the farther away the second is trailing behind the first. On the other hand, the stacking fault around the partially slipped plane bet\veen the two half-dislocatioils has a hexagonal arrailgement of slightly higher energy than the undisturbed cubic one; this excess lattice energy is proportioilal to the



afea of the fault, i.e., t o the distance bet~veen two consecutive halfthe dislocations. A simple calculation sho~ved that, for a plausible value of the specific stacking-fault energy (20 ergs per cm2)', the second halfdislocations ~vould follow the first a t a distance of some 20 or 30 interatomic spacings. This meant that, a t a given time, ,only a small part of the volume of t h e crystal could have contailled stacking faults. T h e surprising result of Warren's measurement is, on the other hand, t h a t something like one half of the X-ray line broadening in brass filings must he attributed to stacking faults; every 40th or 50th atomic plane must be assumed t o contain a fault. This is in contrast to the Heidenreich-Shockley estimate; however, a.closer consideration shows t h a t the latter applies to moderately deformed material only. It assumes the primary existence of full dislocations and iilvestigates the question, how far the half-dislocations separate when the full dislocation dissociates. At high stresses, however, there may be no initial full dislocatioiis: slip may take place by half-dislocations arising and traveling individually. I n this case, every wave of single-step slip either generates or wipes out a stacking fault. Of course, the energy of the stacking fault must be provided by the work of the applied stress during limit of the stress a t which individual singleslip, and this sets a lo~ver step slip call occur. If b is the slip vector of the half-dislocation and T t h e applied shear stress resolved in the direction of the single-step slip, the work of per unit area of the slip plane is T . b ; this has to cover the specific energy of the stacking fault w per unit of area. From an assumption about the energy difference between the cubic face-centered and the hexagonal close-packed lattices, Heidenreich and Shockley derived forw the value 20 ergs per cm2. With this, the equation

gives T = lo9 dynes per cm2, a considerable fraction of the yield stress of strongly distorted brass and far higher than the initial yield stress of a brass single crystal. Of course, the value w = 20 ergs per cm2may be some~vhat high; it corresponds to the estimated free energy difference between the cubic and hexagonal lattices ill bulk, whereas the electronic energy of a hexagonal layer of 4 atoms thickness embedded in a cubic lattice may differ considerably from the bulk value. Ho\ver-er, even if the value of w is only 35 or f/4 as high, the conclusion still holds t h a t the extensive development of stacking faults can occur only a t fairly high applied stresses, in a strongly distorted state of the crystal. It is easy t o see t h a t a crystal containing numerous stacking faults





. . . .

. .

must be strongly hardened. Fig 3-F shows the stack. of cl'ose-packed balls i s already illustrated in Figs 3-B to 3-Dl after a one-step slip between the second arid third horizontal layer; Fig 3-G shows the right; and Fig 3-Hthe left-hand side of the stack. Fig 3-5 represents a front view similar to Fig 3-F, but with a close-packed. row in the viewing direction; it shows that the slip has introduced a step both in the righthand (octahedral) and in the left-hand (cubic) face. Similarly, there is a step in the front (octahedral) face. Fig 3-G indicates that, in addition to the step, there is a shear distortion in.the right-hand octahedral face(also visible on the front face, Figs 3-F and.3-5). Consequently, when slip occurs on either of the three nonhorizontal octahedral planes, either a sheet-cavity is formed, or :an overthrust occurs, as seen in Fig 3-K. This process is the same as that already disclissed in section 3-14; in the present case, however, .the step is shallower, so that the sheet-cavity f would have -a thickness of.only aboutfhalf o an atomic spacing if it did.not collapse, and the-overthrust introduces only one extra atomic plane. I t is clear from this that,.:as-a rule, disk-cavity formation will. . occur instead of overthrust. In: addition, the shear translation: seen inFigs 3-F and 3-G enforcesjt-he--fo.rmationof a weak screw dislocation . when slip occurs across the:st; .. . . . . The effect-of intersecting-slip.;~p~on.,a. stacking fault is now clear . .f&m.the cdnsiderations in secti6i.i$1.4: where the fault intersects the slip plane, two mixed edge-screw. dislocations arise, which move apartas -the intersected edges of th'e fau1t:move apart by .the slip. No slip across the fault can occur, therefore, until. the shear stress is high. enough to produce these two dislocations and move them apart. This process +requiresprobably the piling up. of two-or more dislocations a t the stacking'fault before this can be broken through. . .If, now, a strongly cold-worked face-centered cubic crystal con.-tains several intersecting sets of stacking faults, it will be strain-harclened (apart from other siinultaneous causes of hardening) 'in a way that resembles G. I . Taylor's theory of hardening. The difference is t h a t the obstacle-walls are not preexistent but are generated by slip, and that they are not impermeable to slip but can be penetrated a t a sufficiently high stress. . . I n addition to this, however, -another cause of-hardening is present - in acrystal crossed by several sets of stacking faults. As Professor Warren has pointed out to the writer, interse'cting sits.of faults ought to. produce stresses of order G / n , where n is t,he number of atomic planes between two fault planes; if the faults are 40 atoms apart, they give'




rise to stresses of order of lo9 dynes per cm2. This is both the order of magnitude of the yield stress and that of the stress corresponding to the measured X-ray line broadening in strongly deformed brass. I t is not even necessary, therefore, to assume the piling up of dislocations in the cells between the fault xvalls, as in the Taylor theory: xvithout any dislocations inside the cells, the intersecting fault xvalls could already pro- duce internal stresses high enough to explain the observed yield stress (see E q 3-33 in section 3-21). I t is to be repeated that hardening by the-intersection of.stacking faults is likely to be important only in strongly deformed face--entered cubic metals. I n this case, however, it may be the dominant factor in hardening.



Very little is knoxvn experimentally about the temperature dependence of the yield stress, because the measurements described in the literature have been made a t a time when the compound nature of the factors responsible for the temperature dependence of the stress-strain curve was not realized. I 1 general, stress-strain curves.of the crystals 1 of pure metals are lowered by factors between 10 and 30 as the temperature is raised from that of liquid hydrogen or liquid air to the neighborhood of the melting point. Oilly a part, and possibly the smaller part, of the temperature dependence ot the stress-strain curve is due directly to the temperature dependence of the yield s t r e s ~Suppose that 1 and 2 in Fig 3-37 are . ~ ~ ~ ~ ~ the stress-strain curves of a crystal a t a loxver and a higher temperature, respectively. If the crystal is strained a t the lower temperature T1 to point A on curve 1, then the temperature raised to the higher level Tz and the straining continued, the stress-strain curve obtained will be that indicated by the dotted line. I t lies above curve 1, showing that the crystal strained a t a lo\ver temperature retains a part of its higher yield stress when the temperature is raised. Correspondingly, if the straining starts a t the higher temperature Tz and the temperature is lowered during the deformation a t the point B to T,, subsequellt deformation will follow the dotted curve, sho~ving that the crystal is softer than one deformed by the same amount a t the l o ~ ~ temperature TI. er Obviously, a part, and perhaps the more important part, of the temperature dependence is due to the fact that a crystal strained a t n lower temperature acquires a more strongly strain-hardened structure



(more lattice injuries and internal stresses). On the other hand, the fact that the yield stress increases at once when the temperature is l o ~ ~ ~ e rand, decreases when it is raised during deformation, shows that ed there is also a genuine temperature dependence of the yield stress for a material in a fixed state of hardening. Naturally, the temperature dependence of the rate of hardening, and that of the yield stress for a given state of hardening, are different phenomena requiring different theoretical treatments. Before this can be done, the two effects have to be separated experimentally.


FIG ~ - ~ ~ - - I N F L U E X C E OF SUDDEN



In what follows, only the direct temperature dependence of the yield stress of a material with a given set of lattice injuries and internal stresses (dislocations, mills, cavities, etc.) will be considered and its causes discussed. Although not much is Irno\vn about it quantitatively, it is certain that such a temperature dependence exists, and, though smaller, it is of the same order of magnitude as the temperature dependence of the stress-strain curve itself. Until the possibility of the dislocation bulge extrusioil mechanism (cf. sections 3-13 and 5-19) emerged, the whole temperature dependence of the yield stress was attributed to the circumstance that the dislocaover energy barriers (e.g., regions of tions would have to move for~vard high local adverse stress). Thus, the temperature dependence was attrib-



uted in the Taylor theory t o the circumstance that the hypothetical impermeable mosaic walls would become semipermeable a t higher temperatures and dislocations could pass through them naturally by a thermal activation process; the accumulation of dislocations with increasing strain would then be slo\ver, and the stress-strain curve lo\ver. This general type of explanation is often called the "chemical reaction rate theory" of the temperature effects in plastic slip. Its central difficulty, not often realized sufficiently clearly, is that most elementary processes of slip assumed in the various theories require impossibly high activation energies in order to explain the observed order of magnitude of the temperature dependence of the yield stress, or even the very possibility of slip. T h a t the original Becker activation process is not feasible a t any practicable temperature with the common metals has already been she\\-n (cf. section 3-8). I 1 the Taylor theory, the 1 question of activation can hardly be raised; if the straight parallel dislocations of the model Fig 3-30 are long enough for their field of stress t o be considered as two-dimensional, they are too long t o be lifted over any significant potential barrier by thermal activation. Mott and Nabarro'j4 have translated the Taylor idea into three dimensions in their theory of precipitation hardening (cf. section 3-19); they assumed that, instead of moving as a \vhole, the dislocation line \vould progress by the consecutive movement of limited segments as short as was compatible with its flexural rigidity. I n spite of generous assumptioils about the maximum feasible activation energy, they had t o conclude that the yield stress a t the absolute zero could not be more than 2 or 3 pct higher than a t room temperature if the elementary step in the slip process was the thermally activated movement of the shortest independent dislocation segments. Since the publication of the theories of Taylor and of Mott and Nabarro, a new possibility has emerged in the dislocation bulge mechanism as applied t o precipitation hardening by the writer and t o the dislocation double mill by Frank and Read (cf. sections 3-13 and 3-19). It is of interest, therefore, t o investigate whether this mechanism can explain the observed order of magnitude of the temperature dependence of the yield stress if this is attributed to thermal activation. Since this interesting point has apparently not been treated in the literature, the consideration should be given here in some detail. Let it be assumed first t h a t the tension of the dislocation line (its free energy r per unit length) does not depend substantially upon the curvature and the temperature; in addition, let the dislocation driving



stress T,, be neglected for simplicity. At the absolute zero, the applied stress must have the value T,, required by E q 3-1 9 in order to make the di~locat~ion between two obstacles or between the two poles of a line double mill bplge to semicircular shape (curve 1, Fig 3-38) and thus t o enable the dislocation bulge t o escape. At a higher temperature, the yield st,ress T is lo~ver than T, and so it can bend the dislocation only t o a circular arc of a greater radius r = d/2 determined by E q 3-19:

The arc of radius r produced by the stress T is indicated by the number 2 in Fig 3-38. The stress T cannot balance the dislocatio~l arc without thermal help for all values of radius less than r, i.e., betnee11 the positions 2 and 3 in the figure; arc 3 is part of a circle that is the mirror image of the circle 2 reflected on the line AB. Thermal activation is necessary, therefore, while the arc moves from position 2 t o 3. During this movement, its length increases by L, and its free energy by L r ; a t the same time, the applied stress does the work ST^ (see E q 3-10) where S is the area of the crescent between the arcs 2 and 3. The free energy to be supplied by thermal agitation, i.e., the activation energy is then A = L r - ST^ [3-401 With the notation of Fig 3-38, \ire have and Division of E q 3-19 by the same equation written d o ~ v n absolute for zero temperature gives T / T ~ = d/2r [3-431 Further, according to Fig 3-38, sin

= d/2r


Substituting in E q 3-40 the value of L from 3-41, of S from 3-42, and of from 3-43, we obtain

dr[(a - 2p)/2 sin


- cos 9 1


This expression becomes zero, of course, for TIT, sin 9 = 1, p = n/2, = i.e., if the yield stress does not depend on the temperature. However, if



for instance T / T ~ = 0.8, which is probably the lo~vestreasonable assumption about the pure temperature variation of the yield st,ress, sin cp = 0.8, cos cp =O.G, and cp = 0.93; the square bracket has then the value 0.2. For a typical metal single crystal a t moderate strain, the shear stress is of the order T = 5 X lo7 dynes per cm2, and G



around 10" dynes per cm" Eq 3-19 gives then cl as 10H/5 X lo7 = '2000 interatomic spacings. If r amounts t o 2 electron volts per interatomic spacing, Eq 3-45 leads t o 800 electron volts for the activation energy. This value is as far from being feasible as the activation energies t o ~vhich Becker theory or the Mott-Nabarro calculation on the basis the of. the three-dimensional Taylor theory have led.



By this result, an intriguing situation arisei. It has been clear for some 20 years that the Becker slip process (creation of local slip by thermal fluctuations a t the observed values of the yield stress) cannot be activated unless the stress is quite close to the value that ~vould cause slip without any thermal help. I t is seen now that, in addition, the movement of dislocations over iilternal stress barriers or their estrusion between obstacles or from double mills cannot be activated either: if such processes occur, they must take place almost \vithout thermal help, so that the temperature dependence of the yield stress due t o thermal activation must be insignificant. The reason for this is that, in all such cases, the activation process requires an increase of the length of the dislocation by an amount about as many times greater than the interatomic spacing as is the shear modulus higher than the yield stress. This requires hundreds or thousallds of electron volts; although a large fraction of this may be obtainable from the work of the external forces during the activation process, the remainder is still far above the highest activation energy feasible a t usual temperature. The only mechanism of slip so far suggested that could avoid this difficulty mas the modification of the Becker theory obtained by the lassumption that slip always started in regions of high stress concentration.1 I n this case, the length of dislocations to be created during the activation process was small enough t o be feasible; since, however, the assumption that slip always starts a t stress concentrations due t o cracks or similar flaws cannot be a general solution of the problem (cf. sectioil3-3), the question arises whether there are alternative ways out of the difficulty. I n what follows, three possible causes of the dependence of the yield stress on the temperature should be pointed out. The first is based on the circumstance that the presence of obstacles to slip, such as hard inclusions, automatically gives rise to high local stress which may make a thermal activation possible; the second is the tem' I perature dependence of the dislocation line tension , which gives rise to a temperature dependence of the yield stress directly, without any activation process involved; and the third arises from the fact mentioned in the preceding section that the driving stress needed to move a dislocation, which is usually the main part of the initial yield stress, must depend on the temperature. Fig 3-39 shows immediately that high shear stresses arise around the hard inclusions in a precipitation-hardened crystal when a dislocatioil line is pressed against them. As mentioned in section 3-13, Eq 3-19 is a general relationship between the curvature of a dislocation line and the



shear stress acting on it. If the radius of curvature r = d/2, where d is the spacing between neighboring obstacles, correspoilds t o the yield value of the shear stress T which acts on the bulge, the local shear stress around a partly engulfed obstacle must be as many times higher than T as is d/2 greater than the radius of the dislocation arc around the obstacle. If the.obstacle is small, the shear stress around i t may be a high multiple of the yield stress, because the entire force exerted by the applied stress upon thedislocation garland is transmitted to the obstacle-pegs on which the garland is suspended. This circumstance makes possible an activation process that is quite different from the bulge-extrusion mechanism described in section 3-19 and illustrated in





T h e erlcircling clislocations left aronncl the inclusions after thc escape of t h e ~ n a i ndislocation line arc not indicated.

Fig 3-29. Instead of having to enlarge the bulge until it becomes semicircular, a thermal stress fluctuation can detach the dislocation from the obstacles in the way indicated in Fig 3-39: without any movement of the main part of the dislocation arc, two neighboring bulges can fuse together by a small encircling mol~ement around the inclusion. The activation energy required for this can be much less than that for bulge extrusion because the necessary increase of the length of the dislocation is relatively small. This activation process is easier still if the obstacle is not an inclusion of almost microscopic dimensions but, e.g., a dislocatioil traversing the slip plane or a vacant or surplus lattice row as call arise during slip in a plane pierced by a screw dislocation (cf. section 3-14). I t does not seem possible, honever, in the case of a double mill. The second possible cause of the temperature dependence of the yield stress is that the dislocatioil line tension r, like the surface energy of a liquid, must decrease n i t h increasing temperature. According to Eq 3-18, ~ b d 21', the operating stress of a double mill of pole spacing = d is proportioilal to the dislocation line tension, so that the decrease of r with increasing temperature gives rise to a proportional decrease




of that part of the yield stress that is represented by the operating stress of dislocation mills. Whether the temperature dependence of the dislocation line tension can account for much or most of the temperature dependence of the yield stress cannot be recognized before the former is calculated and the latter measured. A third cause of the temperature dependence of the yield stress is that a part of it must be represented by the stress required for moving the dislocations in the crystal (i.e., the driving stress ~ d cf. section , 3-18). I n particular, the initial yield stress of an undeformed crystal may be determined essentially, and sometimes perhaps esclusively, by the driving stress (see section 3-18). This quantity must naturally depend on the temperature: with increasing thermal agitation, the structure of the dislocation becomes more disordered and the forces acting across the slip plane and opposing the movement of the dislocation must d~crease. While much of the temperature dependence of the yield stress may be due to that of the quantity r, local slip activation must play an important role in transient creep. * An increase of the temperature a t constant stress must give rise to immediate deformation by the reduction of r , which leads to the extrusion of dislocation bulges that could not be extruded by the applied stress so long as had the higher value corresponding to the lower temperature. This circumstance, however, cannot lead t o creep, i.e., to progressive deformation a t constant stress and temperature. Of the temperature effects so far discussed, only two can produce creep: the activated movement of dislocations in an internal stress field or their activated extrusion, and the slow creeplike movement of dislocations when the stress acting on them is too small t o drive them fast. The dislocation driving stress ~ d of course, cannot,have , a rigidly fixed value a t a given temperature; it must depend on the velocity in a way that may be similar to the strain-rate dependence of the ordinary yield stress. At the lowest stresses that can still move a dislocation, the movement is then very slo~v, aiid it may be reflected in a slo~v creep of the material. The movement of the dislocation itself is a thermally activated process, but the activation energy can easily at-

* According to Andrade,'& creep in crystall~nematerials has two physically different componcnts. The first, now called "transient crcep " (or "cold creep ") consists in thermally activated slip processes; the second, "viscous creep" or "hot creep," ariscs from rearrangcnicnts within relatively srnall groups of lnore or less disordered atoms or molecnles, and is closely related t o the viscous flow of amorphor~s materials.



tain a feasible value when the lower limit of the "yield" region is reached by the stress. The preceding considerations show that the temperature effects in plastic deformation and creep must be far more complex than it ~ 7 a s imagined only a few months ago. The analysis of the differentcauses of the temperature dependence of creep'and of the yield stress will be a very interesting and difficult experimental and theoretical problem.

The rema'rkable sharp "upper" yield point in annealed low-carbon steel which is followed by a drop of stress to the "Ibwer" yield point has been known for a long time. Fainter manifestations of similar effects have been observed in several alloys; yield phenomena exceeding that in steel in spectacular sharpness have been found in zinc and cadmium ~ r y s t a l s . Until~quite recently, the numerous attempts a t explaining '~~ ~ the yield phenomenon belonged to two groups. The first ( D a l b ~ , ~ ~ Nhdai,G8 and KurodaG9) was based on the assumption that the cause of the phenomenon had its seat in the grain boundary; the second assumed that it was in the crystal itself (Ludwik and Scheq70Edwards, Phillips, and Liu171and C ~ t t r e l l ~ ~ ) . situation was remarkable ;beThe cause it could be demonstrated that neither type of explanation could be adequate. According to the Dalby-N&dai-Kuroda theory, the grains \irere separated .by a very thin but very hard layer of cementite, 1vhich.resisted plastic deformation until the stress reached'the upper yield point. At the upper yield point, the hard bouildary layer was smashed and deformation could then proceed a t a much lower stress. I n this theory, i t was overlooked that the deformation could progress inside the grains almost unhindered before. the boundary layer was broken, and so this could not give rise to a yield point. The reason why this is so is seen directly from the Taylor theory of strain hardening, which deals almost with the same problem: i.e., kith the rise of stress in the course of straining in a crystal aggregate in which there are impenetrable walls between the grains. A microscopically invisible cementite layer a t the grain boundary cannot contribute to the rigidity of the metal; all it can do is t o prevent the propagation of slip from one grain to the next, as is assumed in the Taylor model. I t \\:as seen, however, in section 3-21. that' the introduction of impermeable walls produces a parabolic stress-strain curve instead of a sharp yield point, and that the strain hardening due to the presence of such walls remains very



small. until the spacings ketween the walls (to which the grain size would correspond in the case of steel) are a s small as or cm. Thus, the boundary cementite assumed in the first group ,of theories could not even make the metal significantly harder, and i t could not produce any yield phenomenon. The second group of theories is represented by those of Ludwik and Scheu70 and of Edwards, Phillips, and Liu7' \vho attribute the upper of yield point to a hard precipitate, and by the \\:ell-kno~vn-theory Cottrel172based on the assumption t h a t the dislocations existing before straining are immobilized by carbon atoms co~lgregatillgin their expanded regions.* By the adsorption of carbon, the free energy of the dislocation is reduced and the dislocation thereby locked in t o the lattice: in order t o tear it away from the carbon atoms, the applied stress has to do a ~vork equal t o the carbon adsorption energy, and this work has to be done during a very slight movem'ent of the dislocation. I n his first theory, Cottrell assumed t h a t the stress required for tearing away the dislocation from its carbon atom atmosphere was the upper yield point; when the dislocations were torn away, the stress needed for moving them mould fall to the value of the lower yield stress. It is seen t h a t the first Cottrell theory mas the exact counterpart of the DalbyNBdai-Huroda theory; in the latter, the upper yield point was attributed t o resistance to slip due t o hard carbide in the grain boundary, in the former t o lldislocation carbide" within the grains. The simple theories of iiitragranular precipitation and of dislocation lockiilg could not explain the yield phenomenon in iron: they mould have demanded the yield phenomenon to occur equally in single crystals, whereas it is kno~vn that the phenomenon disappears when the grain size increases beyond a certain limit,73and it does not exist under normal conditions 'in single crystals of iron. The second theory of C ~ t t r e l approached closely t o the boundary l~~ carbide theory: it assumed that the dislocatiolls in the boundary mere anchored by carbon atoms instead of iron carbide particles, and t h a t the hardened boundaries were broken through, as in the earlier theories, by slip starting inside the grains. T h e third theory of Cottrell,T5 on the other hand, is based on the assumption that the break-an-ay stress of carbon-locked dislocations is several times higher than the upper yield point, and t h a t it is reached only a t points of stress concentration, e.g., a t small cracks. When local slip starts a t these points by disloca* In a body-centered (but not in a face-centered) cubic mctal thcre arc slightly
expandcd regions even in screw dislocations.




tions being torn away from their carboil atoms, the stress concentration decreases o~ving the usual redistribution of stress by plastic deto formation; on the other hand, the slipped volume (shaded area in Fig 3-40) creates a region of increased stress in front of i t (in the direction of shear), because i t acts somewhat like a blunt shear crack. Cottrell assumes that the initial crack is "blunted" rapidly by the local slip i t initiates, while the stress concentration due to the slipped region inly creases more s l o \ ~ ~with its length. Cnless, therefore, the slipped region




can extend undisturbed for a certain length, it is unable to make up for the disappearance of the stress concentration of the initial crack, and the propagation of slip stops. 111 particular, if the slipped volume meets a grain boundary before its stress concentration can become high enough to replace t h a t of the initial crack, the slip process stops until the external stress has been raised to a higher value. When it is restarted by the increase of the applied stress, i t breaks through the boundary, and further slip occurs a t decreasing stress because the stress concentration in front of the gro\ving volume of slip increases. If the grain size is very large, on the other hand, the slip extends unhindered




until it is able to break through the first grain boundary (and a.fortiori through the follo\\ing ones), and no yield phenomenon is observed. , This hypothesis depends on the assumption that the stress concentration in front of a slipped region increases with its size. This cannot be taken for granted: if the shape of the yielded region does not change while its size increases, the stress distribution around it remains geometrically similar t o itself and thus the stress concentration factor is constant. The different behavior of a Griffith crack in brittle fracture is due to the fact that the radius of curvature a t its tip is invariant, being of the order of the interatomic spacing; consequently, the sharpness and the stress conceiltration factor of the crack increase as i t propagates. Since, contrary to a cleavage crack, a Liiders band gronrs not only a t its tip but also on its sides, i t is not certain whether the effective stress concentration can be assumed t o increase during its propagation. H~lden~~ has suggested that the yield may have a mechanism reminiscent of that of electrical breakdown in gases. Initially, a few dislocations \vould break away from their carbon atoms; this, however, could not represent the avalanche of slip observed in yielding. Holden assumed, therefore, that yielding in a sufficiently finegrained polycrystalliile iron would start ~vhenthe carbon-free dislocations collided with dislocations arrayed in the. grain boundaries;* by knocking them out of the boundary and setting them free to move, they would start an acalanche of slip. The difficulty in the \vay of this suggestion is that most grain boundaries cannot be considered as consisting of clislocations (see section 5-10), and even if this is possible, the dislocations of which low-angle boundaries can be thought t o consist are too strongly bound to be released by a feasible amount of energy. Since apparently essential elements to. a satisfactory theory of the yield phenomenon are still missing, two new points should be mentioned now which may play a role in it. So far, i t was assumed that easily mobile carbon-free dislocations can arise from the carbon-locked onest only if the dislocation is torn away from its anchoring atoms by the applied stress, helped by thermal fluctuations. The coilsiderations of section 3-15 show that this is by no * Collisions with dislocations within the grain cannot be assumed to lead to a
yield avalanche: otherwise this would occur in an annealed single crystal also. Holden assumed, therefore, that the mean free path of a dislocation was as large as the dimensions of the specimen, and that the yield phenomenon in annealed polycrystalline iron was due to collisions with boundary dislocations. . t In zinc or cadmium, the foreign atoms adsorbed in the dislocations are [)robably those of nitrogen."s'8



means the only possibility for the creation of carbon-free dislocations: a sufficiently large amount of slip in one plane of the crystal can produce Z-mills in which the operative dislocations are free from carbon. When the slip reaches the amount d a t which the Z-mills can operate under the shear stress present, an avalanche of slip arises because the mills work a t falling stress and produce carbon-free dislocations in other slip planes by the deflection mechanism described in section 3-14, or in some other way. Beside the Z-mill mechanism, there are no doubt other processes t h a t can give rise to the formation of carbon-free dislocations in the course of the slow movement of carbon-locked ones: the deflection mechanism offers obvious possibilities for this, and even the hulgeextrusion mechanism may perhaps be effective, since the increase of the length of the dislocation segment during the bulge formation reduces the number of carbon atoms per unit length. I n order t h a t a Z-mill may be formed, a certain amount of preparatory slip must take place before yielding can start. It seems that any explanation of the dependence of the yield phenomenon in iron on the grain size is likely to demand the assumption of a preparatory sllp which, in fact, is assumed in all recent theories of the yield phenomenon.74-76 Averbach and MuirsOcould demonstrate esperimentally the presence of preparatory slip; another esperimental indication of its reality is the yield delay pheliomenon. It was first observed in zinc crystals1 and later confirmed in iron7"hat the onset of the abrupt yielding may lag behind the application of the stress by many minutes. This, of course, may be attributed t o the circumstance t h a t the liberation of the dislocation from its carbon or nitrogen atoms by means of thermal fluctuations takes time; however, it is possible that the time lag is required for the slow preparatory slip. Once i t is assumed t h a t the development of the slip avalanche requires a certain amount of preparatory slip, there is a simple possibility for explaining why only sufficiently fine-grained iron shows the yield phenomenon but not coarse-grained or single-crystal specimens. I n a single crystal, any amouilt of preparatory slip may take place (unless strain hardening intervenes) provided that the applied stress reaches the critical value needed for the movement or the liberation of the carbon-locked dislocations. If this critical value is much higher than the stress needed for slip after carbon-free dislocations have arisen (i.e., the lower yield point), a distinct yield phenomenon is present. Ho\vever, even if the critical stress is not much higher than the lo~ver yield point, so t h a t only a faint trace of a yield pheilomenon can he observed in a




single crystal, the stress required for producing the suficient anzount of preparatory slip can be much higher in a fine-grained polycrystalline material. The reason for this is the fact that, if no significant amount of preparatory slip takes place in the surrounding grains, the grain in which it arises behaves as if it were embedded in an elastic surrounding. I n this case, any slip that occurs in it must lead t o a relaxation of the stress; consequently, the stress in the grain may drop below the value required for continuing the preparatory slip before this has reached the magnitude a t which the Z-mill can start t o operate under the stress present. I n order t o produce operative Z-mills, then, a stress has t o be applied that may be much higher than the critical stress for the preparatory slip; its value is determined by the requirement that, after the stress relaxation in the grain that accompanies the development of the necessary preparatory slip, the remaining stress should still be sufficient to operate the Z-mill. This consideration can be expressed quantitatively by calculating the rate a t which the stress in the grain T, drops in the course of the preparatory slip; the total drop until the Zmills start t o operate, plus the drop of stress during the operation of the mills until they are blocked by hardening, is the difference betn een the upper and the lower yield point, or a t least is in a close relationship n ith it. From the known difference between the two yield points, the grain size can be calculated in the following manner. Let T be the applied stress in the plane and direction of the preparatory slip before this has started. If the surroundings of the grain cannot undergo significant plastic deformation, the slip that can occur in the grain must remain below an upper limit: this is the amount of slip that would occur if there were no resistance t o slip, i.e., if the operative slip plane were replaced by a crack with frictionless walls extending t o the grain boundaries. In.Fig 3-41, the stress T, in the grain is plotted as a function of the shear displacement d represented by the preparatory slip which is assumed to be concentrated in one plane or in a few adjacent places. T = OA is the ~ a l u e T, before slip starts; OR is the of upper limit to d which ~vould occur in the case of vanishing resistance t o slip (T, = 0). For values of d between 0 and OB, the correspor~ding magnitudes of the grain stress T" are represented by the straight line A B (for a simple proof, see referencess1). The slope of this line increases with diminishing grain size; thus, AC mould represent a larger grain size, whereas the horizontal AD corresponds t o the single crystal in which no stress drop occurs. If the applied stress is increased, the line AB moves away from the origin 0 but remains parallel to itself for a



given grain size: obviously, the limiting shear displacement OB is proportional to the applied stress 0.4. A displacement d creates 2-mills of width d ; the- stress needed to operate such mills is given by Eq 3-23 (cf section 5-13) :

Here the dislocatioi~ driving stress r d is t h e stress needed for operating the mill when, in the course of its ~vorking, the width d becomes very



large; it cannot be very different in magnitude from the lower yield point, and i t should be tentatively identified with i t for the present rough calculation. Eq 3-23 is represented by the hyperbola in Fig 3-41 ; the height of its horizontal asymptote, as remarked, should be provisionally taken for the lower yield point. When the stress OA is applied, preparatory slip starts and the grain stress drops along AB. At the point of inter-





. .

section P with the hyperbola, the Z-mill starts to operate with high velocity, since the stress available (given by AB) is higher .than the stress required for the operation of the mill a t moderate speed, until the point Q is reached. Thus, after a slow beginning along AP, very rapid slip follo\\rs betveen P and Q. T h e large amount of slip represented by the abscissa of Q \vill probably initiate slip in the neighboring . grains, by the stress concen'tration i t creates a t the boundary, and the ;-avalanche.of yielding starts. 'In.'order to .introduce this process, the applied stress T = OA must be high enough for the line AB-t o intersect the hyperbola. I n addition, the point of intersection-Pmust be higher than the value of the critical stress needed for the preparatory slip; let it be assumed first t h a t 'this additional condition is fulfilled. T h e condition for the operation of the Z-mill is then a stress OA high .enoughfor the line AB t o contact the 'hyperbola, or higher. The lower: limiting stress corresponding t o AB being a tangent is obtained from the eqyation of the hyperbola 3-23'-'and:.. . $hat of the line A.B. The latter is giue.*.:when the amount of maximum .slip~O' by."tress OA is kno~vn. can be estiOB mated the follo\ving \!:ay iR2.1f:.t i%:ihethickness of the grain in the direction perpendicular t o the slipplane, complete release of its ilastic strain T/G would produce a sheai~displacementin the slip pline of amount t . T/G if its top and bottom boundaries were fixed and the surrounding matrix did not exert. ail$ other constraint (Fig 3-42). I n reality, the borders of the slip plane cannot move a t all because they are fixed a t the grain boundaries.; for this reason, the shear displacementis not the same along .the slip.plane but varies between a maximum in its middle and zero a t its edge.. In.addition, even the maximum displacement is less than it \vould. be without lateral constraint.R3 However, the order of magnitude is not affected, and so the value t . T/G should be used for the present rough estimate. If the stress in the grain drops from T = OA ollly t o T,, instead of t o 0 , the correspondjng shear displacement is [3-461 -d = t(r - T")/G This is the equation of the line AB. Since the factor q in E q 3-23 must be close t o 1 or 2 (see section 3-23), i t should be omitted for convenience. Elimination of 7, from Eqs 3-23 and 3-46 gives, then,



The plus and minus signs before the square root correspond to the two intersections P and Q of the line AB with the hyperbola (Fig 3-41). A B






becomes a tangent when P and Q coincide, i.e., when the square root vanishes; from this condition, the grain size is obtained as

(7be identified with the drop of stress a t yielding, i.e., the difference between the upper and the lower yield points. E q 3-49, therefore, gives the mii~imumgrain size required for a given value of the stress drop a t yielding. If stress inhomogeneities are carefully avoided, 7 - 7 d seems to be approximately equal to the lower yield,point ~ dthe ;

c can



ratio G/(T - ~ d is roughly 500 for annealed lo~v-carbon ) With this value, E q 3-49 gives t = 0.2 mm, a reasonable order of magnitude. This result must not be regarded as a considerable step toward the verification of the Z-mill hypothesis of the yield phenomenon. The grain size limits not only the maximum slip that can occur in an elastically constrained grain, but also the masimum stress concentration that such slip can create in the neighboring grain. The general treatment of the grain-size effect given in Eqs 3-23, 3-46, 3-48, and 3-49, therefore, could be applied with small modifications t o grain-boundary breakthrough hypotheses. I t should be regarded, therefore, as ageneral approach t o the grain-size effect that can be applied t o several hypotheses on the nature of the yield phenomenon, rather than a definite support t o the Z-mill mechanism. I n the preceding consideration, it was assumed that the stress a t which the preparatory slip ends (point Q) is higher than the critical stress for this slip. If it is not, the mathematical condition just derived is automatically fulfilled as soon as the stress is high enough t o produce the necessary preparatory slip, and the condition for yielding is simply that the applied stress must reach the value required for preparatory slip. I n this case, the yield phenomenon call be observed in single crystals also, and the simple conditioils of the first theory of Cottrell, or of the simple Z-mill mechanism primed by preparatory slip, are present. This seems to be the case with single crystals of zinc and cadmium, which show a very sharp upper yield p0intlO8~~they contain traces of if some impurity, possibly i l i t r 0 g e i 1 . ~ ~ ~ ~ intel-pretation is realistic, the If this ~ difference betnreen iron and zinc is that the preparatory slip can start in the former a t values of stress not much above the (lower) yield point, while in the latter the critical stress for preparatory slip is much higher than the lower yield point. The reason why polycrystalline zinc and cadmium do not show the yield phenomenon is fairly obvious. Many grains in these metals require a very high stress for deformation because their only plane of easy slip is nearly parallel or nearly perpendicular to the tension (or compression). For this reason, the yield stress of the polycrystalline material is a high multiple of that of favorably oriented single crystals. Since the upper yield point.of single crystals is only some 30 t o 50 pct higher than the lo~ver yield point, it is reached in grains of favorable orientation long before the specimeil as a \\,hole starts yielding; in this way, the yield phenomenon becomes imperceptible. It is remarkable that no trace of the yield point can be observed in




zinc or cadmium crjrs'tals until they have been prestrained by a few per cent and then subjected t o mild annealing or aging.10r77*78 could not This be understood on the basis of the simple dislocation locking theory, .which mould demaild the same upper yield point for crystals with few dislocations as.for those with many. However, in the 2-mill theory a necessary condition for the appearance of the yield point is the presence of anchored. dislocatioils piercing through the planes of preparatory slip; it seems that the prestraining is needed for producing such traversing dislocations, whi'ch are then anchored by foreign atoms during the subsequeilt annealing. Before leaving this.subject, a n interesting effect should be mentioned that has so far been ignored in the theoretical treatment of the yield phenomenon.: this is the jerky estension of crystals that has been observed first with NaClw and then with zinc8%ad a few other crysthat tals. I t s study in the case of zinc has sho\~~nclearly thk jerks represent the common yield phenomenon; between two jerks, the upper a yield point is restored by annealing, sometimes ~irithin small fraction of a second. The time between consecutive jerks of extension may be surprisingly constant; this indicates that the sudden yielding is not initiated by one strong thermal fluctuation but starts when some preparatory process'of creep, diffusion, or recovery is completed. By cementing closely spaced rows of tiiiy mirror-smooth spheres of till -'(sibout 100 microns diameter) by means of a soft rubbery adhesive upon the surface of zinc crystal wires, illuminating the crystal with a small light source and photographing it during its jerky extension on a rotating drum covered with bromide paper, the writer found that the consecutive jerks of estension were localized in a thin slip zone. Consequently, the jerks (sometimes in series of hundreds) do not represent indepeildeilt yield processes taking place in different parts of the crystal hut arise by Some reactivation process of the same slip zone.

The strain hardening acquired in the course of plastic deformation can be reduced or removed by annealing, i.e., by keeping the material for a certain time a t a sufficiently high temperature. For different metals, the absolute temperatures required for achieving a certain annealing effect seem t o he more or less constant fractions of the absolute melting temperatures. The structural changes in a cold-worked metal during annealing m a y take place by either of three different mechanisms, which may act



individually, successively, or simultaneously. These are recovery, polygonization, and recrystallization. T h e only effect of recovery t h a t can be observed by X-ray methods is a reduction of the internal stresses. Polygonization is a straightening of curved lattice planes, recognized by the disintegration of the asterism streaks in X-ray photographs into rows of spots. In recrystallization, nuclei of nenr grains arise in the deand formed material, grow into their surroui~dii~gs, consume the old grains. ltecrystallization is usually folloured by "grain gro\vth," i.e., the grotvth of the larger recrystallized gi.ains a t the cost of the smaller ones. This process is a simple consequence of the tendency of a polycrystalline material to reduce its total surface (interfacial) energy; it is not directly related to plastic deformation. For this reason, only the three typical effects of annealing upon a cold-worked crystalline material ~vill discussed here. Information about the considerable advances be recently made in the understanding of grain-growth phenomena can be obtained from the original papers (e.g., C. S. Smithls7 and Beck, Holzworth, and Sperryls8 and from a recent review by Burke and Turnbullsg).

Very little is kno\\rn about the mechanism of recovery. Processes that may lead to it are: 1. The retreat and dispersion of dislocations piled up a t obstacles during plastic deformation. 2. The recombination and dissolution of dislocations. T h e first process would imply the disappearance of a part of the plastic strain, an effect of which only traces have been observed.

Even \\-hen the deformation is apparently a uniform estensioi~or compression, it usually involves slight macroscopic curvatures, which give rise t o X-ray asterism in Laue photographs. Contrary to early views, this, rather than curvatures on a microscopic scale, is the most common cause of asterism streaks.go I the deformed crystal can be f annealed a t a sufficiently high temperature without recrystallization taking place, the asterism streaks break up into rows of distinct spots and (Fig 3-43); this was first observed on NaCl by I<onobeje~vski Mirer.Y1 The phenomenon \\as called "recrystallization in situ" by Crussardg2and "polygonization" hy the present writer, ~ r h o s ecollaborators Cahn93 and Honeycombelg4follo\ved by Guinier and La-



combe,95 investigated the mechanism of the process experimentally. Under the impression of X-ray photographs published by Frommer,96 the writer concluded that the process consisted in a portionwise straightening of bent lattice planes, in the course of which their initially curved sections become polygons, and that the more or less randomly distributed dislocations of the bent crystal congregates during polygonization in the planes that represented the sites of the polygon



corners. This picture was verified by experiments carried out by Cahn 93 on cadmium crystals. Fig 3-44 shows the process schematically. In Fig 3-44a, the parallel curves are traces of operative slip planes in a plastically bent crystal. Since the slipped lamellae are compressed on their concave and extended on the convex sides, atomic fit between them is no longer possible; the thickness of the boundary lines between adjacent slip lamellae would indicate the amount of the local misfit if the lattice spacings remained equal everywhere along one surface of a lamella. In reality, however, the energy of misfit is reduced considerably if the misfit is concentrated





into small regions (Fig 3-44b), outside which there is atomic registry between neighboring lamellae; in other \vords, if dislocations are formed along the operative slip planes. The reasoil why the energy of a plastically bent crystal is lowered by the concentration of the misfit into dislocatioils was given in section 3-7. Fig 3-44b represents schematically the structure of thc plasticallp bent crystal hefore polygonisation.

FIG ~ - ~ ~ - S C H E & X A C TIU R E PI T C


When polggonizatioil occurs, the dislocations congregate in radial planes (Fig 3-44c), which represent grain boundaries subdividing I\ hat was formerly a single bent crystal into a series of subgrains ("polygon facets") with straightened lattices. Each facet contributes one spot to the row of reflections into which the X-ray asterism streak breaks up after polygonization. Lacombe and Beaujardg7have observed the formation of a subgrain structure in slightly deformed aluminum after



annealing, and Guinier and Lacombeg5have recognized and verified that its mechanism mas identical with that of polygonization studied a t the time by Cahn and the writer. I n the course of the last decade, it has become clear that polygonization represeLts a very common effect of annealing upoil cold-worked metals and that a large fraction of the cases ~vherethe formatioil of apparently n e ~ v grains \\-as observed by X-ray diffraction represented polygonization, not recrystallization. A detailed discussion and references can be found in the Symposium on Polygonization in volume 2 of the Progress in Metal P h y s i ~ s . ~ ~ From the point of view of the theory of dislocations, the main questhe tion raised by polygonization is ~ v h y energy of the bent crystal is reduced by the congregatioil of the dislocations in the planes that form the boundaries of the polygoil facets. Obviously, the dislocations in Fig 3-442, are all of the same sign ( T dislocations) : in each, there is extension above the slip plane and compression belo\\. it. It was found in the Taylor theory of hardening that dislocations of equal sign repelled one another, and this would seem t o indicate that a crystal containing a given number of dislocations nrould have its minimum energy if the dislocations were as uiliformly scattered as possible. The reason why this is not so in polygonization vras demonstrated mathematically by Lom e ~ -It ~ ~ . is easy t o see directly that the l a \ of the interaction between ~ ~ dislocations used in the Taylor treatment cannot be valid for a crystal with free surfaces. If, e.g., the shear stresses produced a t a point of the surface by the dislocations present in the crystal are calculated from E q 3-37 (section 3-22), the result, in general, does not vanish; that is t o say, the Timpe-Iioehler-Burgers equations are not valid for a crystal with free surfaces. If the corresponding equatioils are obtained for a crystal 114th free surfaces, they show that the rule of the repulsion of dislocations of the same sign is no longer true; on the contrary, the dislocations assembled in a polygon boundary plane attract one another strongly, and a coilsiderable energy is required to pull one out of the array. I n other ~vords,the energy of the crystal is l o ~ ~ e r e d \\hell the initially randomly scattered dislocations congregate in the polygon boundary planes. T h a t polygonization must lower the mutual energy of the dislocations can be seen ~vithout any calculation. Fig 3-45b has been obtained from Fig 3-45a by cutting the crystal into separate lamellae along 11011operative slip planes not coiltaining dislocations. Wherever there is a dislocation in one of the resulting lamellae, there is a sharp curvature. Fig 3-45c is a similar dissected view of the polygonized crystal. Since



there is no interaction between the dislocations in the lamellae of the dissected crystals, the energy of Fig 3-453 is equal to that .of Fig 3-45c. However, the lamellae in Fig 3-45c fit upon one another, so that no work is required for welding them together, ivhereas those in Fig 3-453

FIG ~ - $ ~ - S H O W I W GS O N REA



have to be bent elastically before they can estahlish contact and welding can take place. The work of the external forces required for this elastic deformation represents the difference between the energies of the nonpolygonized and the polygonized crystal. Although polygonizition is listed in this section het\veen recovery and recry.3tallization, it is by no means certain that it has any appreciable influence upon the yield stress. It is quite possible that the en-



tire decrease of the yield stress during a heat treatment leading t o polygonization takes place by recovery before polygonization; for all we know, polygonization itself may even produce a slight hardening in some cases. Recrystallization The methods required for the evaluation of recrystallization experiments have been worked out to a considerable degree of completeness by Anderson and MehlloO and others; since, however, they do not touch directly the physical interpretation of the phenomenon, they will not be discussed here. From the physical point of view, the basic question in that of the origin and growth of the recrystallized grains. It received its first treatment by R. Becker17 who regarded the process as that of the formation of a stable phase from a metastable one (the coldworked matris), and applied t o it the formalism of the Volmerlg theory of nucleation. I n recent years, several authors have treated the problem, without modifying Becker's nucleation theory essentially. Let 6P be the difference between the bulk free energies of the coldworked and the recrystallized material per unit of volume: for simplicity, let the consideration be restricted t o that of a spherical recrystallized grain surrounded by unrecrystallized matrix. When such a grain of radius r is formed, the free energy 6F . 4ar3/3 is obtained; a t the same time, the interfacial energy a . 4ar2 has to be produced if a is the sp&ific free surface (interfacial) energy between the grain and its surrroundings. The grain is just stable if, for a differential increment of its radius, the bulk free energy obtained is equal to the surface energy needed : 61" . 4ar2dr = a . g a r . dr Thus, the critical radius of the stable nucleus i s

I n order t o produce a stable nucleus, its surface energy a . 4arC2, the less bulk energy obtained, 6F 4nrC3/3,must be provided; with the foregoing value of r,, this amounts to




If this energy is t o be produced by a thermal fluctuation, A represents the required activation energy, and the number of recrystallization nuclei formed per units of time and volume is approximately

Eqs 3-50 and 3-51 represent the core of Volmer's nucleation theory of phase transformations. The factor D is not a constant: since the fluctuations leading to the configuration of the stable nucleus take place a t a rate determined by the velocity of self-diffusion, D is proportional t o the latter quantity and thus contains another Boltzmann exponential with the activation energy of self-diffusion. I n addition to the above application of the Volmer theory t o recrystallization, Becker made a specialized assumption about the temperature dependence of A . Since the excess free energy of the cold-worked phase is due to atomic disorder, cold working can be regarded as a partial licluefaction. Since, further, the free energies of solid and melt are equal a t the melting point, Recker assumed that 6F also vanishes a t the melting point T , and is proportional to (T, - T) below it. From this he concluded that the activation energy A decreased rapidly with the temperature, and that consequently d N / d t and thus the number of recrystallized grains nucleated in the specimen was higher a t lower temperatures. This agreed with the observation that recrystallization a t lower temperatures usually produced finer grain; however, this was often a trivial consequence of grain growth following recrystall?zation a t higher temperatures. By the may, 6 is bound to decrease with inF creasing temperature for the simple reason that an increasing amount of recovery takes place before rlucleation can occur. A simpler and more fundamental conclusion drawn by Becker was that with increasing amouilt of cold work (i.e., of 6F), the rate of nucleation would increase according to Eqs 3-51 and 3-52, and the recrystallized structure would become more fine-grained. Since the interfacial energy CY depends on the relative orientation of the lattices of nucleus and matrix, the Decker theory is intrinsically capable of explaining preferred orielltations of the recrystallized grains. * A very interesting aspect of the problem emerges if it is realized that


* According to Beck and I - I ~ I a 'more important cause of preferred orientation , ~~ in recrystallized metals is the dependence of the velocity of growth upon t h e orientations of the growing grain, ancl the surrounding matrix.

E q 3-50 is simply the expression of the fact that a small particle of a . phase has a higher vapor pressure than.the same phase in bulk. This means that, in. whatever way the critical stable nucleus arises, its size is always determined by E q 3-50. In particular, i t need not arise by thermal fluctuations creating locally an entirely new crystal.unrelated t o the matrix; instead, it is more likely to be the consequenceof a gradual removal, by thermal fluctuations, of the dislocations the presence of which makes the main difference between the distorted and the recrystallized material. Once this is made clear, an important difficulty to all nucleation theories of -recrystallization i s encountered. Expressed ' briefly, it amounts to this: Since the excess free e n 6 r g y . 6 ~ f t h e coldo worked a n d (in the course' of its heating to the recrystalliiation temperature) recovered m i t e r i a l i s very low, the activation energy for the formation of a stable nucleus by any process of.growth from unstable "embryos" is far too high t o be available a t the temperature of recrystallization, no matter in what way the i~uclelis grows. Of course, the local density of energy in a dislocatioi~ extremely high; however, the is quantity 6 is only the density of the mean energy of cold work a t the F most 'highly stressed points. I t is probably not an understatement t o assume t h a t only 2 or 3 pct of the cold work imparted t o a strongly deformed metal remains in it as free energy'by the time it has reached the temperature of recrystallization; on the ground of the well-known measurements of Taylor and Quinneylo2on copper, then, 6F mould be of the order of magnitude of 0.1 cal per cm3, or 4.10"rgs per cm3. The interfacial energy of high angle grain boundaries of copper is of order. 500 ergs per cm2; with these figures, E q 3-51 gives A = lo8 electron volts, a n utterly unattainable activation energy. The corresponding radius of the stable nucleus is given by E q 3-50 as 5 microns. Even if local densities of plastic strain e*ergy 10 or 100 times higher than the average density were assumed, the activation energy ~vould still' be thousands of times higher than the maximum feasible value. I t is clear, then, t h a t the classical nucleation theory cannot be applied t o the formation of nuclei in recrystallization--a fact that,apparently has escaped the attention's0 far. What isthen the way in which the nuclei arise?Fig 3-46a shows the development of a nucleus according t o the classical theories. The &cleus is assumed to possess, almost from the beginning, i t s final fixed to ~ value of specific surface energy, c ~ r r e s ~ o n d i ia~sharp boundary between the stable and the metastable phase. However, an obvious alternative which, in fact, is. far more plausible in the case of recrystalliza.



tion, is shown in Fig 3-466. Here the boundary between the two phases is initially quite diffuse, corresponding t o a n extremely lorv value of the interfacial energy; i t becomes sharper and sharper, and the surfaceenergy increases, as the bulk free .energy of the nucleus drops toward that of the undistorted metal by the gradual clearing of the dislocations in its interior. I n other words, in this case the stable nucleus is formed,

FIG 3-46-HETEROGENEOUS N D H O M O G E N E O U S NUCLEATION. A a. Scheme of the classical phase transformation process: formation and growth of a nucleus having the same structure and the same surface energy from the beginning. b. Scheme of the recrystallization process: gradual decrease of the volume energy and accompanying increase of the surface energy until a sufficiently large nucleus of the fully recrystallized material is formed.

not by the growth from smaller but structurally similar unstable "embryos," but by a homogeneous transformation of its material, extending over a volume that may be larger than that of the stable critical nucleus from the very beginning of the process. It seems very likely that the recrystallized grains are the result of such a "homogeneous nucleation," not of one of the "heterogeneous" type considered in the classical theory. I n homogeneous nucleation there is obviously no nucleation diffi-



culty that has to be overcome by thermal activation. The boundary of the nucleus may sharpen a t the rate a t which the release of free energy from the interior can provide the correspondiilg increase of the surface energy. It is interesting t o notice that, as in the case of martensitic transformations, metallurgy seems to have come across again a phase transformation m%chanism t h a t was not encountered in the narrower field of physical chemistry.

FIG ~ - ~ ~ - - S H O W HI N G ' OW A




I n the Volmer nucleation theory, the stability of the nucleus is treated only from a thermodynamic point of view. I n the case of recrystallization, the corresponding molecular consideration is simple and instructive. The increased free energy of the cold-worked matrix is due mainly to the presence of dislocations, of which a number ends on the surface of the growing nucleus, as indicated in Fig 3-47. As was seen in section 3-20, a dislocation behaves as though it were under a tension of magnitude r, eclual,$o its free energy per unit of length. Without dislocations ending a t its surface, the small grain would contract and disappear under the influence of its surface energy; this, however, is counteracted by the outward pull of the dislocations in the matrix. If,



as before, the interfacial energy is denoted by a , and the nucleus is a sphere of radius r, the surface tension across a greatest circle is 27rr . a, and the capillary pressure it produces is p = 2a/r. I the number of f dislocation lines that end on the surface is n per unit of area, they create a hydrostatic tension n . r in the sphere; the nucleus is stable if

so that the critical radius is

With n = 10" per cm?, r = ergs per cm, and a = 500 ergs per cm2, r, becomes 1 micron, in agreement with the order of magnitude of the value derived from the thermodynamical consideration That the dependence of the interfacial energy upon the relative orientation of the lattices of matrix and nucleus can lead to a preferred orientation of the recrystallized grains has already been mentioned. Fig 3-47 shows, beyond this, that the particular arrangement of the dislocations that exert the outnrard pull upon the grain call also be responsiblefor a preferred orientation both of the nuclei and of an anisotropy of their rate of growth in different directions. Becklol has pointed out that, in fact, many recrystallization textures require for their explanation the assumption that the rate of growth, not only that of nucleation, depends sharply on the relative orientation of nucleus and matrix. Specialized ideas about the mechanism of the local removal of dislocations from the volume of the recrystallization nucleus have been by and \\rho assume that the process is put for~vard BeckLo3 by Cahn,lo4 identical with polygonization, and that some of the polygon grains (sectors) act as recrystallization nuclei by starting t o grow into their surroundings. It is difficult a t present t o recognize the prospects of this interesting suggestion. There is no obvious reason why the clearing of a region from dislocations should alivays start ~irithpolygonization; ho~irever,polygonization may be a more general consequence of annealing in a cold-\vorked metal than can be recognized directly from observations. The hypothesis that recrystallization is preceded by polygonization (possibly on a very small scale) may seem to suggest that it is in reality a simple grain-grov th process that takes place in the polygonized struchave, in fact, been e~pressedrecently. However, a ture; such v i e ~ i ~ s closer consideration shows that the t~iroprocesses are fundamentally



different and the analogy between recrystallization and grain growth is superficial. The beautiful studies of C. S. Smithsi have shown that the grain boundaries in an annealed polycrystalline metal are in a state of almost stable capillary equilibrium; the shapes of the grains are practically identical with the shapes of the foam cells in a froth. I n contrast to this, the polygon sectors derived from a region of initially uniform schematically in Fig 3orientation are typically elongated, as sho~vn 48. For this reason, the polygonized structure is far from being in a




capillary equilibrium; in Fig 3-48, the capillary tension of grain 1 upon its boundary with grain 2 is much higher than that of grain 2 because the former has more polygon boundaries abutting a t the grain boundary. Consequently, the grain boundary will move into grain 1, and grain 2 will grow. This is not a typical grain-growth effect: i t belongs rather to the category of grain gro~vth the outward tension (not by balanced by the grain-boundary tension) of dislocations ending on the grain boundary as discussed above. In thermodynamic terms, the process is a growth of grain 2 which has a lower energy of cold work 6F a t the cost of the grain 1 with the higher energy. I t is t o be noted that a poly-



gon boundary has a capillary tension perpendicular t o its more or less parallel dislocatioils as well as in the direction of the dislocations; this follows from the fact that the dislocations in the boundary attract each other. The conclusion t o be drawn is, then, that in spite of its superficial similarity ~ v i t h common grain-growth process in a fine-grained anthe nealed material, recrystallization is a different phenomenon. Its driving force is the unbalanced out~vard pull of dislocations or dislocatioil walls ending on the surface of the growing grain; seen from the energetical point of view of the Becker theory, the driving force is the difference between the free energies per unit volume of the gro~ving grain and its surroundings. Whether or not the dislocations, which represent the excess free energy of the surroundings, are arranged in polygoil ~valls is hardly a point of fundamental importance, although it may influence the details of the recrystallization process.

The strength (fracture stress) of brittle materials like glass has been successfully explained by the Griffith theory ; 1 5 it is determined by the . size of the largest cracks or other stress concentrations present in the material before the load is applied. As a consequence of the statistical scatter of the size distribution, the measured strength values fluctuate , ' from specimen to specimen, sometimes within wide limits. I n contrast to this, the fracture load of ductile tensile specimens is surprisingly ' sharply reproducible. The fracture starts aln~ays the center of the I in "neck" of the specimen, so that it cannot choose the worst among the I flaws present in the specimen; a similar restriction of the locus of fZcture in a brittle material ~vouldincrease the scatter of the strength values considerably. I n this way, one is led to collclude that inherent I defects play,only a secondary role in the fracture of a ductile metal; instead of being present originally in the specimen, the crack that leads to fracture is probably produced by the deformation process itself. How can plastic deformation lead to the formation of cracks? One mechanism of crack formation, although possibly not the one of the greatest practical importance, has already been discussed in sectioil 3-14; it is the generation of cavities by the deviation of slip'into another slip plane. Another possit)ility is knonrrl to exist in t~vinuing; this process can lead t o the formation of visually observable cavities ( ( I Rose's channels " in iron). Crack formation during plastic deformation must be a very likely



occurrence for the follon;i!lg reason: It mas mentioned in section 3-22 that the shear stress a t a. distance r from the center of an edge dislocation is

its maximum value in the plane of slip (8

0) is

that the highest On the other hand, the Burgers-Koehler solution sho~vs value of the normal stress a t a point just above or below the center of the dislocation is

where h is the height or depth of the point considered relatively to the center of the dislocation. At these points, the normal stress in the plane parallel t o the slip plane is equal to that in a plane perpendicular to i t ; both stresses are tensile on the tensile side of the dislocation, and compressive on the opposite side. Eqs 3-55 and 3-.56 show that the maximum shear stress a t a distance r from the dislocation is numerically equal to the maximum tenslle stress a t the same distance. Whether slip occurs in the slip plane, or cracking in a perpendicular or nearly perpendicular plane, is therefore decided by a rather delicate balance bet\\ eel1 the tensile and the shear cohesion of the lattice. In ductile materials, of course, the shear coheut sion is relatively low and the dislocation can progress ~ v ~ t h o much danger of its being converted into a crack. However, the situation is changed if the slip is confroilted with an obstacle, e.g., a hard inclusion or one of several dislocatioi!~ traversing the slip plane. In this case, several dislocations pile up in front of the obstacle (Fig 3-49), and their elastic fields superpose. The resulting high shear stress, according to the assumption made, is resisted by the hard obstacle; however, the resulting maximum tensile stress occurs in a plane that is usually within the parent lattice. If i t reaches the tensile cohesion, cracking n7ill occur as indicated in Fig 3-49. The process illustrated In Fig 3-49 throws interesting light upon a remarkable, though infrequently realized, fact. Although precipitation hardening, or the addition of another element in solid solution, can raise the initial yield stress of a metal very considerably, the increase

of the fracture usually small; sometimes the fracture stress is even reduced by the presence of a hardening precipitate. I n other \vords, the unhardened metal strain-hardens by deformation until it approaches the fracture stress of the hardened one; the difference betnreen the t n ~ o materials is mainly in the initial yield stress. This may be



( a ) PRO(b).


due to the circumstailce that fracture, initiated by crack formation according to Fig 3-49, occurs whenever a critical number of dislocations has piled up in front of obstacles. This is likely to be the case when the applied stress reaches a critical magnitude; the nalure of the obstacle (whether it is a hard precipitate, or lattice injuries like crossing dislocations due to preceding plastic deformation) may play only a secondary role.



A similar superposition of tensile stresses occurs when edge dislocations of the same sign are present in the array shown in Fig 3-50. Obviously, all dislocations give rise to normal stress of the same sign a t the point P ; consequeritly, the array may produce a high tensile stre'ss both

FIG. 3-50-CRACK FORAIATION I N A SIIIL' PLAXE. n A row of edge dislocations in parallcl slip plancs, 'representing a limited polygon \\,all, produccs a high tensilc stress a t its end (indicated by arrows); this may lcad to crackil~galong t h e slip plane as indicated i r ~ representing the probablc mcchanis~nof f a t ~ g u e b, fracture.

in the horizontal and vertical planes through P. The way in which these stresses arise is easily visualized. The upper half of the crystal shown in Fig 3-50a represents a polygonized grain consisting of two polygon sections; if lifted off the lower half, it would assume the shape shown in Fig 3-50b. If now the polygonized half is forced upon the other part of the crystal, a high tensile stress is bound t o arise both between them a t




the end of the polygon boundary and along the polygon boundary, which is belit strongly (straightened) as the tnlo halves of the crystal are welded together. Obviously, then, there will be a tendency to cracking not only in the plane of the polygon wall, but also in the slip plane itself; in fact, the latter is likely to he more important because slip planes are atomic planes of high density, separated from their neighbors by large spacings, and so the tensile cohesion must he particularly lotv across them. This agrees well with the observation made by Ewing and Humphrey, Stanton and B a i r s t ~ r n ~ ~by M c C l i n t o ~ k ~ ~ fatigue cracks and ~ that ~ usually start in an operative slip plane. This is particularly remarkable in steel, which, as its tendency to brittle fracture shows, has a particularly easy cleavage along the (100) plane. I n spite of this, the large majority of the fatigue cracks is formed along slip planes that never coincide with a cleavage plane. It is interesting to reflect t h a t the prolonged cyclic stressing in fatigue experiments may easily have a n effect upon the arrangement of dislocations similar to that of annealing without recrystallization: i t may well "shake" the dislocations in neighboring slip planes into positions of lower energy represented by the polygon mall. Thus, fatigue may possibly arise by local polygonization due to prolonged cyclic straining, and the fatigue crack would then be the consecluence of the high tensile stress arising where a polygon wall ends in the manner indicated in Fig 3-50a. T h e writer is much indebted for valuable discussions t o many colleagues, particularly to Dr. P. A. Beck (University of Illinois), Dr. Morris Cohen (Massachusetts Institute of Technology), Dr. E. W. Hart (General Electric Co., Schenectady), and Dr. W. T. Read (Bell Telephone Laboratories). He cannot claim, of course, that all views expressed in this chapter have been endorsed hy all authorities whose help is gratefully acknowledged.
1. Weingarten, G.: Acc. Lincei Rend. (1901) [5], 10 (I),57. 2. Volterra, V.: A n n . ec. nornl., Paris (1907) [3] 24, 401. 3. Timpe, .4 : Z . math. Phys. (1905) 62, 348. 4. Burgers, J. M.: Proc. Zion. Nederland. Aknd. v. Wet., .4msterdam (1939) 42, 293, 378; Proc. Phys. Soc. (London) (1940) 62, 23. 5. Love, A. E. H.: Treatise on t h e Mathematical Theory of Elasticity, 4th ed., 221. Cambridge, 1927.

6. 7. 8. 9. .lo. 11. 12.


Reported by T h . von Iiitrrn&n; Enzykl. d. ?taalh. Ff'iss., 4, pt. 4. phys. (1929) [5] 2, 749. Dehlinger, U.: Taylor, G. I.: Proc. lioy. Soc. (London) (1934) A-146, 362. Polanyi, M.: Z. Physik. (1934) 89, 660. Oro\\ran, E.: 2. P h ~ s i k . (1934) 89, 634. Schmid, E . : Proc. First Int. Congr. Appl. Mech., Delft (1924) 342. Kochendorfer, A,: Plastisehe Eigenschaften von Ktistallen etc., 26, 32, 36, etc. Berlin, 1941. 13. Orowan, E.: Unpublished contribution to a Conference of the Anglo-French Society of Sciences on Thc Solid Statc, London, Jan. 20, 1945. [Cf. Nature . (1945) 166, 139.1 14. Lomer, \V. M . : Proc. Roy. Soc. (London) (1949) A-196, 182. '15. Griffith, A. A.: Phil. Trans. Roy. Soc. (1920) A-221, 163; Proc. First Int. Congr. Appl. Mech., Delft (1924) 55. 16. Becker, R.: Physik. %. (1925) 26, 919. 17. Becker, R.: Z. tech. Physilc. (1926) 7, 547. 18. Orowan, E . : Z. Physik. (1935) 98, 382. 19. Volmer, M., and Weber, A , : %. Physik. Chem. (1926) 119, 277; see also Volmer, M.: Kinetik der Phasenbildung. Dresden, 1939. Steinkopff. -20. Turnbull, D., and J. C. Fisher: J. Chein. Phys. (1949) 17, 71. 21. Nabnrro, F. R . N.: Proc. Phys. Sac. (London) (1947) 69, 256. 22. Peierls, R . : Proc. I'hys. Soc. (London) (1940) 62, 34. 23. Bragg, Sir Lawrence, and J . F. Nyc:Proc. IZoy. Soc. orid id on) (1947)-A-190,474. 24. Meissner, K., M. Polanyi, and E. Schmid: Z. Physik. (1930) 66, 477. 25. Elliott, H. A.: Proc. Phys. Soc. (London) (1947) 69, 208. 26. Bragg, W. L.: Proc. Phys. Soc. (London) (1940) 62, 54, 105. 27. Galt, J. K., and C. Herring: Phys. Rev. (1952) 86 (11), 656. 28. Frank, F. C.: Report of Conf. on Strength of Solids, 46. Phys. Soe. of London, 1948. The rcflexion mechanism was also discussed in the present writer's (E.0;) Diplorua Thesis, Tech. Hochsch. Berlin-Charlottenburg, 1929 (unpublished). 29. Leibfried, G.: Z. Physik (1950) 127, 344. 30. Frank, F . C., and W. T . Read: Phys. Rev. (1950) 79, 722. 31. Orowan, E.: Symposium on Internal Stresses, 451. Inst. of Metals, Lolldon, 1947. 32. Hanson, D., and M. A. Wheeler: J. Inst. Metals (1931) 46, 229. 33. Smith, G. C., and D . W. Dewhirst: Research (London) (1949) 2, 492. 34. Rosbaud, P., and E. Schmid: 2. Physilc. (1925) 32, 197. 35. Maddin, R., C. H. Mathewson and \V. R . Hibbard: Trans. AIME (1949) 186, 527. 36. Orowan, E.: Symposium on Imperfections in Nearly Perfect Crystals (Oct. 1950) 147. New York, 1952. John \Viley and Sons. 37. Mott, N. F.: Proc. Phys. Soc. (London) (1951) 64, 729. 38. Seitz, F.: Phys. Rev. (1950) 80, 239. See also Chapter .l in this volume. 39. Heidenreich, It. D., and \V. Shoekley: Report of Conf. on Strength of Solids, 57. Phys. Soc. London, 1848. 40. Brown, A. F.: Nature (1949) 163, 961; Metallurgical Applications of the Electron Microscope, 103. Inst. Metals, London, 1950.




41. Miller, R . F.: Tra,r~!s. I M E (1936) 122. A 42. Schmid, E., and \V. Boas: Kristallplastizitat, 119. Berlin,' 1935. Springer, Translation: Plasticity of Crystals, London, 1950, F: A. I-Iughes and Co: Ltd. 43. Chalmers, B.: Progress in hletal Physics, 3, 312. London, 1952. l'ergamon Press Ltd. 44. Cohen, hl., E . S. Machlin, and V. G. Paranjpe: Trans. A m . Soc. llfetals (1950) 42A,242. 45. Nye, J . F.: Nature (1948) 161, 367. 46. l'auling, L.: The Nature of the Chc~nicalBond, 2d cd., 415, Cines 5 to 8 from top. Corncll Univ. Press, 1948. 47. Mcrica, P. D., R. G. \V:iltcnl>crg, ant1 H. Scott: U. S. Bur. Stds. Science Paper 347 (1919); Trons. A I M E (1921) 64, 41. 48. Guinier, A,: Coinpt. rend. (1938) 106, 1641. Calvet, J., P. Jaquet, and A. Guinier: J. Inst. Metals (1939) 66, 121. 49. Prcston, C. D.: Proc. Roy. Soc. (Lo,ndon)(1938) A-167,526; Phil. Mag. (1038) 26, 855. 50. Jeffries, Z., and S. Archer: The Scicnce of Metals. New IYrk, 1924. 51. Mott,-N. F., and F. R. N.Nabarro: Proc. Ph?js. Soc. (London) (1940) 62, 86. 52. Nabarro, F. R. N.: Proc. Phys. Soc. (London) (1946) 68,669. Mott, N. F., and F. R. N. Nabarro: J. Inst. ildetals (1946) 72, 367. 53. Taylor, G. I.: J. Inst. Metals (1938) 62, 307. 54. Fisher, J. C., E. \V. Hart, and R. H. Pry: Acta llfetalluryica (iflay 1953) 1,336. 55. For refcrcnces, scc A. H. Cottrcll in Proyress i n Met. Phys. (1949) 1, 77. 56. hlott, N. F.: Phzl. ilfay. (1952) 43, 1151. 57. Lomer, \V. M.: Phil. May. (1951 ) 42, 1327. 58. Cottrell, A. H;: Phil. May. (1952) 43, 645. 59. Koehler, J . S.: Phvs. Rev. (1941) 60, 397. 60. Schmid, 13.: Proc. Int. Conf. Physics, London (1934) 2, 161. 61. Andradc, IS. N. da C., and C. I-Iendcrson: 'l'rans. 1Zoy. Soc. (London) (1951) A-244, 177. 62. Orowan, 13:: Trans. West o f Scotland Iron and Steel Inst. (1947) 45. 63. Ilorn, J . IS., A. Coldberg, and T . E . Tietz: l'rans. A I B d E (1949) 180, 205 (Tech. Pub. 2445, Metals Tech., Sept. 1948). 64. Mott, N. F., and F. R. N. Nabarro: Report of Conf. on Strength of Solids, 1. Phys. Soc. London, 1948. 65. Andrade, E.. N. da C.: Proc. Roy. Soc. (London) (1911) A-84, 1 ; (1914) A-90, 329. (1947) 160, 466. 66. Smith, C. L.: N a t ~ ~ r e 67. Dalby, \V. E.: Proc. IZo!j. Soc. (London) (1913) A-88,281. 68. Nhdai, 8 . : 2. Physzk. (1024) 6 , 371. 69. Kuroda, M.: Sci. Papers, Inst. Phys. Chem. ltes., Tokyo (1938) 34, 1528. 70. Ludwik, P., and R. Schcu: Ber. Fachausschusse ver. Eisenhuttenl. (1925) 6, I . 71. Edwards, C. A., D. L. Phillips, a n d Y . H. Liu: J. Iron Steel Inst. (1943) 147, 145. 72. Cottrell, A. H.: Report of Conf. on Strength of Solids, 30. Phys. Soc. London, 1948. F . R. N. Nabarro: Ibid. 38.

73. Edwards, C. A., and L. Pfeil: J. Iron Steel Znst. (1925) 112, 79. 74. Cottrell, A. H.: Symposium on Plast. Def. of Cryst. Solids, 60. Carnegie Inst. Tech. and OhTR, 1950. 75. Cottrcll, A. H.: I n L'dtat solzde (9th Solvay Congress), Brussels, 1952, 487. 7 6 . Holden, A. N.: J. o j Jfetals A Z J f E (Feb. 1952) 182. 7 7 . Cottrell, A. H., and D . F. Glbhons: hratzrre (1948) 162, 488. 7 8 . \TTain, H . L., and A. H . Cottrell: Proc. Phys. Soc. (London) (1950) 63B,339. 79. Vrecland, T . Jr., D . S. Wood, and D . S. Clark: Trans. A m . Soc. Metals (1953), 46, 620. 80. Averhach, B. L., and H . hfuir: Discussion of Vreeland, Wood, and Clark. Trans. A m . Soc. Metals (1953) 46, 632. 81. Orowan, E.: Proc. Roy. Soc. (London) (1939) A-171,7 9 ; Report on Progress (June in Physics (1949) 12, 85. Phys. Soc. London; I17eld. J.,Res. S z ~ p p l . I 952). 82. Oro\van, E.: Nature (1941) 147, 452. 83. Bragg, \V. L.: Nature (1942) 149, 511; Symposium on Intern. Strcsscs, 221. Inst. of Metals, London, 1947. 495. 84. Sylwcstro\vicz, \Ir., and E. 0. Hall: Proc. Phys. Soc. (London) (1951) 64-B, 85. Clssscn-Nekludowa, hf.: %. Physik. (1929) 66, 555. 86. Beckcr, It., and E. Orownn: 2. Physik. (1932) 79,566. Metals (1953) 87. Smith, C. S.: Trans. A Z M E (1948) 176, 15; I'rans. Aim SOC. 46,533. Harker, I)., and E. Parker: I'rans. Ain. Soc. Atetals (1945)34,156. ' 88. 13eck, 1. A., M. I. Holeworth, and P. Spcrry: Trans. A Z M E (1949) 180, 163. 89. Burkc, J . E., and D . ~ T u r n b u l l :Progress in Met. Phys., 3, 220. London and New York, 1952. 90. Orowan, E., and K. J. Pascoe: Nature (1941) 148,467. 91. I<onobejc\vski, S., and I hfirer: 2. Krist. (1932) 81, 69. 92. Crussnrd, C. : Rev. Met. (1944) 41, 111, 133. 93. Cahn, R . \V.: J. Znst. Metals (1949) 76, 121. 94. Honcycoinhe, R. \Ir. I<.:J . Znst. Metals (1951-52) 80, 45. 95. Guinier, A,, and P. Lacoinhc: Mdtaux el Corr. (1948) 24, 212. See also Rcf. 98. 96. Frommcr, F . : J . Inst. Metals (1939) 64,285, Figs 11 and 12. 97. Lacomhe, P., and L. Beaujnrd: J. Znst. Metals (1947) 74, 1. 98. Symposium on Polygonization, in Progress in Metal Physics, 2, 151, 177, 193. London and New York, 1950. 99. Lomer, \V. M.: Privatc communicat~on. 100. Anderson, W . A,, and R. F . htehl: Trans. A Z M E (1945) 161, 140. 101. Beck, P . A,, and H . I-Iu: il'rans. A Z M E (1949) 186,627; (1950) 188, 1214. 102. Taylor, G. I., and H . Quinncy: Proc. Roy. Soc. (London) (1934) A-143, 307; (1937) A-163, 157. 103. Beck, P . A,: J. Applied Phys. (1949) 20, 633. 104. Cahn, R. Iv.: Proc. Phys Soc. (London) (1950) A-164,323. 105. Ewing, J . A,, and J. C. \V. Humphrey: Trans. Roy. Soc. (London) (1903) A-200,241. Stanton, T . F , and L. Bairstow: P ~ o cInst. Czozl Eng. (London) . (1906) 166, 78. 106. McCl~ntock,F. A,: Proc. First U.S. Nat. Congr. Applied Rfcch. (1951) 653.