Search Documents

By J. B. Conway

RECENT studies of first-stage creep data have led to a special least-squares procedure for use in calculating the Andrade creep-equation constants. This procedure is easy to apply, uses only experimental data, and yields results which do not vary with the individual making the analysis. Taking logarithms of the Andrade equation I = lo(1 +ßl1/3)ekt leads to In l = In lo + In(1 + ßl1/3) + kt A residual can then be defined as the difference between the logarithm of an experimental length value and the logarithm of a calculated length value. Following the usual least-squares approach differentiation with respect to 10 and k of the expression for the sum of the squares of the residuals leads to two equations which cannot be solved directly for the three constants involved. However, values of ß can be assumed and corresponding values of lo and k can then be calculated. Then the sum of the squares of the residuals can be calculated. For a certain value of ß this sum will exhibit a minimum value to identify the least-squares values for ß, lo, and k. An illustration of the application of the above least-squares approach is obtained by employing the creep data for lead originally presented by Andrade' in the development of the one-third creep law. A plot of the sum of the squares of the residuals as a function of ß is shown in Fig. 1 along with the corresponding lo and k values. The minimum is seen to occur at ß = 0.0143 from which lo = 39.149 and k = 7.27 x 10-5. Based on these results the Andrade equation for these lead data is 1 = 39.149(1 + 0.0143 t1/3)e7.27 x 10-5t compared to the equation I = 39.13(1 + 0.01457/l/3)e6.95 X 10-5t developed by Andrade (calculation procedure not given but was probably a graphical solution). Calculating length values using the above equations and comparing with the experimental length values, residuals corresponding to the difference between experimental and calculated length values can be obtained. The sum of the squares of the residuals can then be calculated. For the case of the least-squares approach a value for this sum was found to be 2.37 x 10-3 compared to a value of 3.47 x 10" for the equation originally proposed by Andrade. In other words, while the constants proposed by Andrade provide an excellent fit for the lead data, a still better representation is afforded by applying the least-squares approach discussed above. A similar result was obtained for the lead data at 162oc.l The least-squares approach led to 1 = 9.866(1 + 0.0497 t1/3)el.011 x 10-2t compared to the equation I = 9.94(1 + 0.0433 l1/3)e1.06 X 10-2t given by Andrade. In this case, too, the sum of the squares of the residuals is slightly smaller for the equation based on the least-squares approach. It should be pointed out that a double trial and er-

Jan 1, 1965

By D. Kuhlmann-Wilsdorf

A new theory of work hardening is developed which rests on only a few simple principles and is applicable to a wide variety of materials and dislocation structures. It explains, qualitatively, the general characteristics of the three-stage shear stress-shear strain curve of fcc metals, as well as the most important features connected with easy glide. The value of, the work-hardening coefficient in stage 11, is derived quantitatively, from first principles, and is shown to be insensitive to many changes in detailed dislocation behavior. The low rate at which energy is stored during mechanical working in stage 11, the dependence of slip line length on stress, and the so-called Cottrell-Stokes law are explained. Although the theory is primarily developed for fcc metals and alloys, it is largely applicable also to some other materiuls, and in particular apparently to poly crystalline simple steels in theh linear range of hardening. FOR pure fcc metals and alloys, after many different pretreatments and under a great variety of testing conditions, the well-known three-stage work-hardening curve is commonly observed. Moreover, the work-hardening coefficient in stage 11, 011, bears an almost constant relationship to G, the modulus of rigidity, such that K = G/Brr" = 300, within a factor of about 2 either way. The very fact that the three-stage curve and the value of K are so very persistent, can hardly be understood except on the assumption that some quite simple phenomena are responsible. In this light all available theories of work hardening must be judged unsatisfactory: Each of them is based on some quite specific dislocation model, while it is known that metals with similar values of K may have widely different dislocation structures, for example, dislocation tangles in the case of pure fcc metals, and piled-up groups of dislocations in a-brasstype alloys. Thus, while a particular theory might conceivably represent a special case with a reasonable degree of accuracy, it cannot possibly have illuminated the underlying principle. Another severe criticism applying to the presently most widely discussed theories is that they employ empirical or even quite unexplained parameters in order to arrive at a numerical value of K. This indicates that some vital aspect of the work-hardening process has remained unexplained. In the present paper, a new theory of work-hardening is developed which does not suffer from the above shortcomings. It is derived from first principles, and is founded on the realization that the underlying causes for the three-stage work-harden- ing curve and for the persistence of the numerical value of O,, must be few and simple. As a first step it is shown herein that the experimental oklservations on "easy glide" can be explained on the assumption that in this stage dislocations multiply, beginning in a number of restricted area:;, and from there penetrate into crystal regions still substantially free of glide dislocations until a quasi-uniform dislocation distribution has been established. During "easy glide", the resistance to the movement of the foremost dislocations, advancing into still largely a dislocation-free regions, is believed to control the flow stress. This resistance is determined by various factors, among them the reaction of the dislocations against bowing-out into loops, which is due to their line tension. Without such bowing-out no dislocation multiplication would occur, but the corresponding contribution to the flow stress stays practically constant all through stage I. According to the present theory, stage TI begins as soon as there are no more areas left into which newly formed dislocations have not yet penetrated, i.e., as soon :is a quasi-uniform dislocation density has been attaned. Regardless of how, in detail, the dislocations should happen to be arranged at the end of "easy glide", whether in pileups or tangles, on one glide system or on several, the stress necessary to overcome the line tension of the segments must now increase, because the dislocation density increiises while the average free dislocation length decreases. It is argued that all other contributions to the flow stress stay at about the same level as in stage I, and that the said increase in the stress necessary to bow out dislocation segments into loops is responsible for the major part of work hardening in stage 11. Stage TII of the stress-strain curve is explained by largely follouing the ideas of Seeger and his co-workers; however, it is pointed out that the start of stage m may on occasion indicate the onset of "conservative climb" instead of cross slip. Indirect evidence indicates that linear work hardening persists even under the profuse action of cross slip, but at a lower rate. Only when, in addition to cross slip, climb operates extensively, does the work-hardening rat,. drop sharply, ultimately down to zero. BRIEF SURVEY OF EXPERIMENTAL EVIDENCE The stress-strain curve of crystals of many different substances are qualitatively similar. No significant permanent plastic deformation takes place below the so-called "critical resolved shear stress". As the applied stress is raised beyond this level. yielding occurs with little or no work hardening and

Jan 1, 1962

By David A. Thomas, John F. Peck

Sam Leber (General Electric Refractory Metals Laboratory)—I think that the authors should differentiate between the curved grains obtained by swaging and the crumpled grains produced by drawing. This can be seen most easily in the transverse sections of Fig. 2d, and is also apparent in Fig. 2c. This difference in structure indicates that the mode of deformation shou1d affect the texture, since the cylindrical texture should be maintained or even increased by the curved grains developed by swaging. Drawing, on the other hand, crumples the grains. This should convert the cylindrical texture into a normal fiber texture. This has been substantiated by a comparison of tungsten wires fabricated by similar drawing and swaging schedules. Although cylindrical textures have been detected in drawn iron wires, they were weak in comparison to tungsten wires which included swaging as part of the fabrication process. It should also be noted that large curved surfaces developed by swaging are concentric with the surface. The postulated (100) surface of these crystals would be confirmed by a (100) cylindrical texture, as was observed in tungsten. The preservalion of the (100) cylindrical texture by swaging would maintain the <100> direction in the radial position required for easy flow. This might provide a partial explanation for the decreased rate of work hardening observed in the swaged wire. J. F. Peck and D. A. Thomas (authors&apos; reply) — The authors appreciate Dr. Leber&apos;s comments on the differences in textures, microstructures, and properties of swaged and drawn wires. The evidence of the prominence and persistence of the cylindrical texture in swaged bee wires seems quite consistent with our microstructural observations and interpretations. We are presently engaged in further work to correlate the development of texture and micro-structure in drawn bee metals.

Jan 1, 1962

By E. F. Adkins, R. W. Douglas

The variation of the spectral emissivity of osmium and ruthenium with temperature can be expressed by the following relations: DURING a study of the sintering characteristics of osmium and ruthenium, discrepancies in the previously accepted melting points of the metals were noted. The melting temperatures had not been measured at the time of the work, but were estimated in the literature as 2700°C for osmium and 2450°C for ruthenium.&apos; A recent investigation by Baird2 has shown the melting temperatures of osmium and ruthenium are 3000" and 2250° C, respectively. In the present work, osmium bars sintered at temperatures as high as 2900° C showed no evidence of melting, whereas ruthenium showed definite evidence of melting at temperatures as low as 2250°C. In view of the apparent discrepancies, the melting temperatures of both metals were determined. The spectral emissivities of both metals were determined as an aid to studies of the sintering characteristics of these metals. MATERIALS AND PROCEDURES The materials used in these studies were metal powders supplied by Englehard Industries, Inc., through the courtesy of The International Nickel Co. Analytical data from three laboratories, although not in complete agreement as to the purity of the material or the major impurity elements, indicated that the maximum impurity levels were 0.5 and 1.0 wt pct in osmium and ruthenium, respectively. The metal powders were pressed to 1/4 by 1/4 by 6-in. bars at 20Tsi using carbon tetrachloride as a lubricant. The pressed bars were presintered 1 hr at 1200°C after heating 1 hr at 400°C. The presintered densities of the several bars of the metals ranged from 65 to 80 pct of ideal density. Specimens were self-resistance heated in a vacuum sintering bell. Temperatures were measured with an L&N, disappearing filament optical pyrometer. True temperatures were read at the base of an 0.30-in.-diam, 0.150-in. deep hole drilled into the center of a presintered bar. Apparent temperatures were read from the surface of the bar adjacent to the hole. MELTING POINT DETERMINATION The melting temperature of the metals was determined as the temperature at which a slight additional increase in current through the specimen resulted in a reduced temperature at the base of the hole. The apparent reduction in temperature is a result of the liquid metal filling the hole, thus altering the black-body conditions. The melting points of osmium and ruthenium, based on three deter-

Jan 1, 1962

By W. L. Phillips

Aluminum and copper single crystals were strained in shear, unloaded and rotated 60, 120, and 180 deg. The magnitude of the Bauschinger strain increased rapidly during the easy glide region and became constant as the easy glide limit was reached. An increment in stress was not required to continue slip in any close-packed plane or in the reverse direction during easy glide. A qualitative theory is proposed to explain the present results. FOLLOWING the initial work on tin by Bausch,&apos; there have been several investigatorsB7 who have employed the shear test to study the deformation of aluminum. In all cases the shear technique was employed in an attempt to obtain a more homogeneous shear than in the conventional tension test The direct shear technique has the obvious advantage, relative to the tension test, that a crystal may be stressed in any predetermined plane and direction, including the reversal of strain sense. In the present investigation the shear technique was employed with aluminum and copper single crystals and their behavior was studied with respect to deformation in all close-packed directions in the octahedral plane. In contrast to previous investigations, the shear technique was used in an attempt to obtain a stress-strain curve in shear which is identical with the corresponding tension curve and then to use the additional degree of freedom of the shear test to study the mechanical behavior of crystals in all close-packed directions in the octahedral plane. PREPARATION OF CRYSTALS Single crystals 1/2 in. in diam and 6 to 7 in. long, were prepared from aluminum containing 0.001 pct Si, 0.001 pct Fe, and 0.001 pct Zn, by spectro-graphic ana1ysis and from OFHC copper. Crystals were grown from the melt in graphite at a rate of 1 5 in. per hr. The copper crystals were grown under an argon atmosphere. No lineage structure was observed in Laue patterns, though small angle boundaries were present in the crystals. The aluminum and copper crystals were annealed at 640" and 900°C, respectively, for 24 hr and furnace-

Jan 1, 1962

By A. G. Guy

Thermodynamic activity rather than chemical composition is basic to the analysis of diffusion. This is the essential conclusion reached by Darken1-3 and by Birchenall and Mehl.4 If so, it is reasonable to expect that diffusion calculations should be carried out using activities. This paper will illustrate a method of employing activities in analyzing interstitial diffusion. For interstitial diffusion (of carbon in iron and probably generally) the diffusion of the solvent atoms, volume changes, and other such complexities not explicitly provided for in the usual diffusion equations cause a negligible error of the order of 1 pct. Therefore it is relatively easy to test the validity of an analysis based on activities in the instance of interstitial diffusion. The mathematical treatment of the more complex substitutional diffusion could be simplified once the adequacy of the activity method had been established for substitutional diffusion. Thus the present work has the purpose not only of providing a superior analysis of interstitial diffusion but also of indicating a possible approach to the more important question of an adequate treatment of substitutional diffusion. Development of the Diffusion Equation Although activities are convenient quantities to determine experimentally, they are derived from the more fundamental fugacities.9 Possible confusion in using activities can be avoided by basing equations involving their use on the relationship, dP = -D1d/ax dt [I] where, dP = the number of grams of solute crossing one cm2 in the time dt sec. D, = the diffusion constant for use with fugacities; it is assumed to be constant at a given temperature, and has units of secs. f = the fugacity of the solute in the solid solution; the units are those of pressure, dynes per cm2. x = the distance in cm. Eq 1 can be considered to be the basic form of the first Fick Law for one-dimensional diffusion. In order to convert Eq 1 into an equation involving activities, a choice of a standard or reference fugacity, f°, must be made: then. a = f [2] where a is the activity corresponding to the fugacity f. However, the value of a is also determined by the relation, a = a\c [3] where a is the activity coefficient and c is the concentration. For a given value of a (such as a = 1) it is evident that the units used for expressing concentration will affect the numerical value of a corresponding to a given amount of carbon dissolved in y-iron. Since weight per cent concentration is so widely used, this unit will be adopted as the standard in this paper. When the value off given by Eq 2 is substituted in Eq 1, the first Fick Law becomes, dP= -D,f°dl [4] Since D,f° is constant (at a given temperature), the second form of Fick&apos;s Law can be obtained in the usual manner, -dt D&apos;f° dx [5] where c&apos; is the concentration in g per cm3. In terms of weight per cent concentration, c, Eq 5 becomes, dc _ 100 d2a Tt = T /f w [6] where p is the density of the solid solution and will be assumed to be constant. In order to simplify Eq6, it is necessary to consider the standard state, f°. Standard states are chosen for convenience. There appears to be one especially convenient choice for the analysis of interstitial diffusion. If the activity coefficient, a, is set equal to unity at infinite dilution of the solute, 100 then — 100D1fl° is approximately equal P to the usual diffusion constant in dilute solutions. This is true since the activity, a, in Eq 6 can be replaced by the concentration, c, with little error in an infinitely dilute solution.* Here f1° is the standard fugacity necessary to achieve this standard state. It is proposed that a diffusion constant Da&apos; be defined, Da1=100/pD1f1° [7] where A is the metal whose diffusion is * The second derivative of a is equal to the second derivative of c in dilute solution only if -da/dr- is. zero. Although theory does not require this condition, in the systems C in Fe, and Zn in Cu, dc is in fact found to be essentially zero.

Jan 1, 1950

By P. W. Bakarian, John C. McDonald

Three magnesium alloys were processed in various ways to exhibit a wide variation in the ratios of yield strength and tensile strength in the rolling direction compared to the cross-rolling direction. Basal-plane pole figures were prepared from several specimens. A fair correlation was observed between the property ratios and the ratio of percent of poles oriented in the cross-direction quadrants compared to the percent oriented in the rolling-direction quadrants of the pole figures. THE modes by which magnesium crystals deform are such as to lead to marked preferred orientations in all forms of wrought products. The effect of the active modes at room temperature, coupled with these orientations, is often to produce a lower compressive yield stress compared to the tensile yield stress. The phenomenon is due to easy twinning in compression, and has been well-related to the preferred orientations as described by basal-plane pole figures of the hcp magnesium single crystal. The literature contains some information on the variation with test direction of the tensile properties of sheet, but no quantitative relationship to the observed pole figures has been established. These effects have been reviewed in the few com- prehensive treatises in English which have been written on magnesium technology and metallurgy.1°3 Roberts presents several pole figures, but no tensile-property data in the plane of the sheet. Raynor has no pole figures, but quotes Frommer&apos;s4 results. The information in Beck&apos;s compendium more or less parallels the results of Jones and powell.5 In both a Mg + 1.5 Mn alloy and a Mg + 4.6 A1 alloy, cold rolling was found not to change the general situation of higher properties in the cross-rolling direction,* than in the rolling direction. Thomas&apos;6 X-ray studies of Jones and Powell&apos;s specimens did not permit correlation of properties and preferred orientation. Ansel and Betterton7 report tensile and compressive properties on three other alloys as a function of degree of cold rolling, plus a recrystallizing anneal. As rolled, there was a change in anisotropy with degree of reduction, but all alloys were reasonably isotropic in tension when annealed. X-ray diffraction photograms illustrated the kinds of orientations present, but were not correlated with the properties. Bakarian8 has described a method of determining quantitatively a preferred orientation, and illustrated the method by applying it to pure magnesium and two of its alloys. This method was used in the present work as a basis for correlating the strength properties with the texture.

Jan 1, 1965

By W. F. Hosford

A series of aluminum single crystals was subjected to axially symmetric flow by drawing through dies. The orientation dependence of the drawing stress, which indicates the resistance to this type of flow, compares reasonably with predictions of Taylor and of Bishop and Hill. Measured lattice rotations likewise are generally consistent with theory. The most successful analyses of the plastic behavior of polycrystalline metal and its relationship to single crystal behavior are those of Taylor1-3 and of Bishop and Hill.4-6 The purpose of this paper is to provide a critical test of some of the basic assumptions involved in these theories. Both treatments apply to fcc metals and assume the shape change of each grain in a polycrystal to be the same as that of the polycrystal as a whole. Thus for a randomly oriented polycrystal extended under uni-axial tension each grain is assumed to undergo axially symmetric flow. dey = dez = -1/2dex, deyZ = dezx = dexy = 0 [l] where x, y, and z are a set of orthogonal axes, x being the tensile axis. Although [011.] oriented grains in bcc and fcc metals may depart markedly from this ideal behavior,7,8 the assumption is probably quite reasonable for most grains. Under a uniaxial tensile stress, ox, the work dW per unit volume expended in producing an increment of axial strain dex is, dW=oxdex [21 Since this work is expended in slip, dw =STi dyi 131 where Ti and yi are the shear stress and shear strain on the active slip systems. In both analyses Ti is taken to be the same on all slip systems, and dependent only on the total prior shear strain on all systems, so dw = Tdy [41 where t is the shear stress required for slip and y is the sum of the shear strain yi on all slip systems. Combining 121 and [4] and defining M = oX/T The ratio M is a generalized Schmid factor. It relates the tensile stress required for axially symmetric flow, ox, to the basic shear stress, 7, and it also expresses the ratio of crystallographic strain, y, to the tensile strain, Ex It depends only on the orientation of the crystal. Taylor evaluated M for various orientations by finding the least value of dy/dEx which could satisfy the shape change. Bishop and Hill, on the other hand, found M by considering the relative work (1/T)dW/dEx expended by stress states which could activate enough slip systems to produce the shape changes. The results of both methods appear to be equivalent, although certain omissions were made in Taylor&apos;s original work. The orientation dependence of M, calculated by the second method,9 is shown in Fig. 1. Taylor used the average 2= 3.06 for all possible orientations to predict the tensile stress-strain curve (oxvs Ex) for a random polycrystal from the shear stress-strain curve (t vs y) of a single crystal. Changes in M from lattice rotation were neglected so points on the ox vs ex curve were simply taken at ox = 3.06t and e = y/3.06 from corresponding points on the t vs y curve. Recent extension of the analyses of Taylor and of Bishop and Hill to crystallographically textured metals has been used to relate plastic anisotropy to textures, and to predict the potential texture hardening that could result from texture control.9 Unfortunately there have been relatively few experimental tests of the underlying theory. Taylor found that the tensile stress-strain curve of polycrystalline aluminum did agree reasonably well with that predicted from a single-crystal T-y curve. Had the single crystal undergone easy glide, how-

Jan 1, 1965

By M. G. Nicholas

The influence of deformation, relative movement and surface cleanliness on the strength of room -temperature "interparticle" bonds between copper surfaces has been studied. No bonding resulted when clean surfaces were merely pressed together, even under pressures of 50,000 psi. Bonding resulted, however, when clean surfaces were simultaneously pressed and moved relative to each other, some bonding occurring no matter how small the pressure or relative movement. Exposure of the clean surfaces to various chemicals inhibited bonding, none at all being produced by small relative movements. The size of the movements which first produce bonding between exposed surfaces depends on the nature of the chemical, being less than 0.002 in. for cleansing reagents and greater than 0.004 in. for lubricants, water and hydrogen sulfide. It is concluded from the experimental data that the prime requirement for bonding is the dispersal of surface films such as oxides and adsorbed contaminants. THE formation of interparticle bonds during the room-temperature compaction of powder metallurgy components greatly increases the ease of handling in subsequent production operations prior to sintering. The strength of unsintered compacts follows the same general trends as their density and hardness, but quantitative data are extremely Scarce. Quantitative correlation of the data on the overall strength of a compact with that of the individual interparticle bonds is virtually impossible due to the extremely complex geometry of powder compacts. Qualitatively, however, the strengths of unsintered compacts can be correlated with variations in parameters known to influence the extent of cold bonding, and hence interparticle bonding, such as surface cleanliness, relative movement, and deformation of the mating surfaces. Kuzmic and seelig4 found that contamination of powder particle surfaces with lubricants reduced the strength of unsintered iron and copper compacts. Interparticle shear has been shown to promote bonding by Train and Hersey7 while Gregory has shown that die rotation is beneficial. Eilender and schwalbe6 concluded that most bonding occurred during the early stages of pressing iron powder compacts when interparticle shear is one of the principal processes occurring. mile there can be little doubt that these and other factors, such as particle size and shape, affect the extent of interparticle bonding produced by compaction, evaluation of the influence of individual factors is not easy. Particle deformation and relative movement, for example, occur simultaneously during compaction, making any precise measurement of their relative influences difficult. Because of these and other difficulties encountered in studying interparticle bonding in actual powder compacts, the approach adopted in this study was to

Jan 1, 1963

By W. A. Wood, W. H. Reimann, K. R. Sargant

A study is made of the microstructural changes produced in armco and pure iron subjected to alternating torsion at amplitudes above and below the knee of the S/N curve. The aim was to identify the basic mechanisms of fatigue and to find why they should give rise to a sharp S/N knee in contrast to the type of S/N curve shown by fcc metals. It is found that the basic structural change above the knee in the iron is a pronounced cell formation in the grains, which determines irregular micro-cracks, and below the knee a dispersion of fine slip that determines a dispersed pore formation in the grains. Comparison of these changes with those in fcc metals similarly tested shows that they do not include an intermediate structural change, a concentrating of slip in narrow bands with associated slip-band cracking, that characterizes the fcc metals and accounts for their different type of S/N curve. RECENT conferences on basic mechanisms of fatigue emphasize that fatigue damage takes many forms: for example, cracking in slip bands (Thompson1), at grain boundaries and inclusions (Hempel&apos;), at precipitation zones in alloys (Han-stock, Forsythe), or at surface notch-peak effects, both hypothetical (Mott, cottrel16) and observed (Wood). Evidently it is not easy to see in this diversity a basic mechanism. At the same time the contributions suggest that some reconciliation may be possible if distinctions are drawn between three phases of the fatigue process: initiation of damage at nuclei, distribution of these nuclei, and the final macroscopic crack system. In initiation, the relevant features appear to be that nuclei may appear early, at least by 1/100th to 1/1000th of the life of a test specimen; that they may do so at many independent centers, appearing according to some observationsa as pores or voids; and that they are not necessarily self-propagating, damage spreading rather by their multiplication. In distribution, the relevant feature appears to be that they form preferentially in zones where the microstructure becomes abnormally strained (as defined more closely below). Now in these two phases the basic fatigue process should be systematic. But, at the same time, the "abnormal strains" in particular set the stage for considerable diversity in the final phase of macroscopic cracking, for where zones of abnormal strain develop will depend on what kind of fatigue deformation is applied and what kind of prior processing the test material has had. On this view it seems desirable to know what in

Jan 1, 1964

By A. G. Rozner

TRANSVERSE rupture tests have been commonly used in mechanical investigations of brittle materials. The specimens are simple, easy to prepare, and loading presents no difficulty. Owing to the complexity of stress distribution in bending1 one can determine stresses rigorously only for the elastic range. The bend-test results, however, are at times converted into tensile-strength data, determined as maximum fiber stress. The assumption is made that for small plastic deformations the linear stress-strain relation applies. This problem was experimentally investigated by Bach3 who tested cast iron in bending and in tension and found that the maximum fiber strength obtained in bending exceeded the experimental tensile strength by a factor depending on the geometry of the tested piece. The factor varied from 1.57 for a rectangular cross section to 2.12 for a circular cross section of the tested beam. Since this work was carried out 60 years ago, the cast iron used could be expected to have been highly brittle, thus the extent of plastic deformation prior to fracture very small. We have therefore carried out a number of tests on magnesium-modified cast iron with an average elongation of -5 pct. It was hoped that the data obtained from the tests would serve as a guideline for comparing data obtained in tension and bending for brittle materials. Mechanical properties were measured at room temperature on an Instron tensile machine at a crosshead speed of 0.05 in. per min. The three-point bend test employed cylindrical bars 0.20 in. in diameter with a 1.0-in. span. For the tensile test, cylindrical tensile specimens with a gage length of 0.85 in. and a diameter of 0.160 in. were used. Results are summarized in Table I. From Table I one can see that fracture strength

Jan 1, 1965

By B. Ramaswami, Y. Nakada, U. F. Kocks

Compression tests were performed on single crystals of aluminum, silver, gold, and a Ag-10 pet Au alloy, with height-to-width ratios from 8:1 to 1:1. As this ratio decreased, so did the amount of easy glide until, for cubic specimens of all the pure materials, it was entirely absent. Heterogeneity of deformation is proposed to he a major contribution to the explanation of easy glide in slender specimens. For comparison, one single-crystal cube of zinc was tested and showed a behavior very similar to that in tension. Beyond easy glide, tension and compression curves are very similar. It is concluded that, in this range, neither grip effects nor orientation change influence the work-hardening characteristics to any significant extent. (121) and (100) crystals of all materials had similar stress-strain curves in tension and compression, except that the stress level of the flat portion of aluminum (109 crystals at high strains mas somewhat higher in compression. These results are at variance with previously proposed mechanisms of orientation dependence by both Hosford and Kocks. The compression tests themselves are described in detail and it is shown that they are essentially frictionless and produce single slip when the crystal is suitably oriented. DEFORMATION in uniaxial compression does not necessarily give the same result as deformation in uniaxial tension, essentially because the boundary conditions chosen are practically always different. 1) The shape and size of the specimens are usually different, and any genuine effect of size on work hardening should thus influence the result. 2) The grip constraints are usually different and can vary substantially for different kinds of tests. If the compression specimen is a thin plate and the friction on the platen is large, the grip contraints are more serious than in tension and lead, in the extreme, to uniaxial strain rather than uniaxial stress. If the specimen is roughly cubical and the friction is low, the specimen is less constrained than in tension inasmuch as displacements in the contact face are allowed. 3) The orientation change of single crystals is of opposite sign in tension and in compression; it is furthermore different in detail, because the relevant orientation in a squat specimen is the normal to the contact face, whereas that in a slender specimen is the generator of the side faces.&apos; The considerable dependence of work hardening on the initial orientation in tension would suggest that the orientation change should also be very important. 4) Finally, localized deformation instabilities found in tension tests, such as necking and Lüders band propagation, are not possible in squat specimens. This is, for some investigations, a favorable consequence of the large grip influence. Even if one is not interested in any of the effects mentioned above, compression tests are often desirable just because the specimens needed are so much smaller and sturdier. For example, they have been very helpful in the investigation of latent hardening by means of alternate longitudinal and lateral compression of cubes,* and for testing

Jan 1, 1964

By Walter A. Backofen, Donald H. Avery

Intense primary and cross-slip traces were observed in easy glide on Cu: 6 pct-A1 single crystals deformed in tension. A mechanism of cooperative source operation is developed which recognizes that both the applied stress and that resolved from dislocation arrays contribute to the stress on a source. By considering the stress from arrays, the activation of secondary systems in easy glide and the difference in appearance of slip markings on pure and solid-solution crystals are explained. Furthermore the cross-slip system is predicted to be second favored during easy glide for a large range of orientations. FROM many studies of easy glide in fcc metal crystals, a pattern of observations has developed which still requires full explanation. In pure copper, aluminum, and silver, the easy-glide, slip traces are fine and uniform with steps 50 to 100A on the primary* system.1-4 Yet it has not been at all clear why slip in the pure materials should occur in this way. On the contrary, one might expect, with simple slip and a low hardening rate, that the weakest sources on the primary system would operate extensively to produce strong slip traces. In a brass, however, intense slip steps up to 0.5 are found on both primary and cross-slip system4,6-9 seeger10 has pointed out that in a material of low stacking-fault energy such as a brass, extensive easy-glide cross-slip could not be thermally activated, but rather must originate on the cross-slip plane. In further explanation, the wilsdorfs4 and seeger10 noted that only the cross-slip system would not harden differently than the primary, whatever the latent hardening tendencies of other systems; therefore, in brass, which exhibits high latent hardening, the cross-slip is the only secondary system which could be expected to operate extensively, once it became active. The explanation of how the cross-slip system, on which the applied stress is low, becomes activated has been related, in a nonspecific way, to pileups by Hirsch11 and Honeycombe.l2 In the work to be described, deformation traces on solid-solution Cu: 6 pct A1 single crvstals were studied with conventional light microscopy and a mechanism proposed to explain both the observed activation of secondary systems in easy glide and the difference in appearance of slip markings on pure and solid-solution crystals. EXPERIMENTAL In preparing the Cu-A1 single crystals, rectangular bars approximately 0.4 in. by 0.4 in. by 6 in. were first machined from ingots made by vacuum melting copper of 99.999 pct and aluminum of 99.99 pct purity. They were then packed in powdered graphite and the crystals grown in a modified Bridgman apparatus at a rate of 5 mm per hr under an atmosphere of purified nitrogen. Crystal perfection as indicated by Laue back-reflection spots was good except for the last inch to freeze, which was discarded. Orientations so determined are shown in Fig. 1. Tensile specimens with a reduced gage section were obtained by masking the ends with electrical tape and chemically polishing in a solution of 20 pct HNO3, 24 pct HAc, 52 pct H3PO4, 2 pct Hcl, and 2 pct H20 at room temperature.13 Finally, the specimens were electropolished in a 30 pct ortho-phosphoric acid bath at 12 v. Tensile deformation was performed in an Instron testing machine at a strain rate of approximately 1 pct per min. During easy glide, extensive cross-slip was observed with all crystals, in spite of the low Schmid factor on this system, and in no case was any other secondary trace encountered. A certain sequence of slip-line development was associated with the extension. At yielding, intense primary traces passing completely through the crystal developed at some point on the gage section, generally near the grips; weak cross-slip traces also appeared, extending from the primary traces into the adjacent undeformed crystal, Fig. 2. Further extension led to primary traces adjacent to the first-formed and connecting with those of the cross-

Jan 1, 1963

By G. Y. Chin, A. R. Von Neida

The relationship between crystalline texture and magnetic anisotropy in a Co-10 pct Fe alloy has been investigated by means of X-ray pole .figures, hysteresis-loop tracings, and magnetic-torque measurements. The cold-rolled texture of a sample after 95 pct thickness reduction is of the "brass-type", that is, (110)[112/. At 500°C the cold-rolled material recrystallizes fo a (311)[112] orientation, plus a minor (210) [001] component. This texture persists up to 1400°C, at which point it is replaced by a secondary recrystallization texture, (111)[112]. Magnetic-torque measurements show that the easy axis of magnetization lies along the transverse direction in the cold-rolled material and changes to the rolling direction after primary recrystallization. This change is also reflected in the hysteresis characteristic; whereas the B-H loop along the rolling direction is nonsqunre with a low value of residual induction (750 vs 19,000 gauss at saturation) for the cold-rolled material, the annealed sample exhibits a very square loop with a large value of residual induction (16,000 gauss). Theoretical torque curves calculated from the texture data on the basis of magnetocystalline anisotropy energy compare favorably with the observed torque curves. Values of the crystal anisotropy constant, K1, obbained from these comparisons are in good quantitative agreement with single-crystal data. Furthermore, both the torque amplitude of the poly crystalline material and the magnitude of K1 obtained with a single crystal show an identical temperature dependence over the range from —195° 10 +220°C. It is thus conclzided that the magnetocrystalline anisotrpopy erlergy is chiefly I-esponsible. for the nzagnetic allisotropy of textured Co-10 pct Fe alloy. THE results of a study on the relationship between crystalline texture and magnetic anisotropy of an alloy containing 90 pct Co and 10 pct Fe by weight are reported in this paper. The alloy has an fcc structure: a saturation magnetization of about 19,000 gauss,2 and a nearly zero value of magnetoatriction.3 puzey4 has reported a large crystalline anisotropy constant of -7.4 X 105 ergs per cu cm, indicating the (111) directions as the easy axes of magnetization. Hence, texture variations resulting from cold work and annealing are expected to introduce a magnetic anisotropy in the polycrystalline material through the alignment of these (111) easy directions. In the present study, magnetic anisotropy of both rolled and annealed strips was determined by magnetic-torque measurements and hysteresis-loop tracings The results were correlated quantitatively with calculations based on texture data obtained by the X-ray pole-figure technique. It is established that the magnetocrystalline anisotropy energy is indeed chiefly responsible for the magnetic anisotropy of textured Co-10 pct Fe alloy. EXPERIMENTAL Material Preparation. The alloy nominally contained 90 pct Co, 9.5 pct Fe, and 0.5 pct Mn by weight, the chief impurities being nickel, carbon, and silicon (less than 0.5 pct total). The cast ingot was rolled to strips 10 mils thick and slit to 1 in. in width. After strand annealing at 900°C in a hydrogen atmosphere, these strips were further cold-rolled to various thicknesses down to 0.25 mil (97.5 pct reduction), with samples collected for study of the development of texture and magnetic anisotropy as a function of thickness reduction. subsequently, samples cold-rolled from 100 mils to 0.25 mil were obtained, thus extending the thickness reduction range to 99.75 pct. In addition, 0.5-mil samples cold-rolled from 10 mils were selected

Jan 1, 1965

By P. K. Koh

With identical annealing heal treatment the development of major annealed texture component seems to depend primarily upon the degree of cold reduction. Cube texture was evident on annealing- 1/2-, 1,/4- , and 1/8-mil tapes at 760°C, but it dissap-)eared on 927°C annealing of 1/8-mil tape cold reduced 96.9 pct. Some cube texturee remained as tile strongest 927 °C annealed texture component in 1/4- and 1/2-mil tapes cold reduced 93.8 and 87.5 pet, respectively. The switching coefficient increases and the squareness ratio of the hysteresis loop decreases proportionally with increasing cube pole concentration. PREVIOUS studies 1,2 on the 1/8-mil ultra-thin molybdenum permalloy tape established cold rolled and annealed textures and their magnetic performance for switch-core and coincident-current memory applications, The major cold-rolled texture was found to be of {110}<335> which is quite similar to the cold-rolled texture of silver and 70-30 cartridge brass. The major annealed texture components were found to be of {120)-- <001> and {113)<785>. Although the direction for easy magnetization for this type of alloy by the single-crystal technique is not avaliable, knowledge from related studies indicated [Ill] as the easy magnetization direction, similar to that of nickel.&apos; Therefore, a cube orientation, if present in the tape core, would be most detrimental to its magnetic performance.3 In its cold-rolled state the tape had < 111 > directions in the rolling plane 14.5 deg from the rolling direction.&apos; But the coercive force of the tape in such condition will be excessive due to large amounts of internal stresses, and therefore makes its use impossible for memory applications due to long switching time. In lower nickel alloys, suchas 50-50 nickel-iron and 65 pct Ni-Fe sheets, previous investigators4"7 found the presence of cube texture on annealing at 900° to 1000°C. Although no cube texture was found in annealed 1/8-mil tape,"&apos; its presence in some annealed 1,/4-mil tape cores was firmly established a year ago as verified by significant differences in magnetic properties. The present investigation is devoted to the orientation study and related magnetic measurements on three thicknesses of the ultra-thin tape, namely, 1/8 mil, 1/4 mil, and 1/2 mil, cold rolled directly from annealed 4-mil stock in the cold-rolled state as well as after they were annealed at 760°C (1400° F) and 927°C (1700°F). The results obtained might serve to elucidate the development of annealed texture as well as to establish clearly the importance of preferred orientation on the magnetic performance of such memory tape core despite its well-known low magnetic anisotropy exchange energy. EXPERIMENTAL PROCEDURE Samples were taken from one production heat (0.010 wt pct C, 0.004 P. 0.016 S, 0.80 Mn. 0.29 Si. 79.69 Ni. 4.42 Mo. 0.054 Co, 0.002 Ti, 0.12 Al, 0.04 Mg, 14.56 Fe). Annealed 4-mil sheets were cold re-

Jan 1, 1962

By S. Feuerstein, J. M. Galligan

SURFACE effects in the deformation of metal single crystals have been noted by a variety of workers.&apos; A large majority of these experiments have used surface roughening or a second chemical constituent at the surface. We performed the following experiment to see how the presence of a fine polycrystalline layer influenced the stress-strain behavior of single crystals of zinc. Spherical single crystals oriented for single slip were acid-machined into cylindrical dumbbell-shape shear specimens having a gage length of 3.2 mm and a cross section of 20 sq mm. The crystals were then deformed varying amounts at the surface of the gage section with emery paper. Following this treatment, the crystals were heated for approximately 8 hr at 200°C. This resulted in a polycrystalline surface layer, with an observed grain size of approximately 1/10 mm in diameter. Deformation of these crystals was undertaken at 78°K in pure shear, the shear direction <1120> lying in the basal plane. The deformation apparatus has been described elsewhere.2 The presence of a polycrystalline layer markedly raises the first observed plastic deformation, as can be seen from Fig. 1. This yield stress is then followed by a region of low work hardening comparable to that observed in shear tests on similar untreated zinc single crystals. However, the extent of easy glide is quite seriously curtailed, and is followed by a more rapid increase in work hardening, reminiscent of Stage II observed in the deformation of fcc crystals. The higher yield can probably be attributed to the difficulty of penetrating the grain boundaries by dis locations from sources interior to the boundaries. If we associate this yield stress with this breakthrough, then we might expect that more dislocations would be trapped in the crystal than is normally the case. If there is an increased density of trapped dislocations, this could lead to an increase in the amount of trapping with increasing strain, giving a larger strain hardening in the crystal. Although other explanations can be given, it is quite tempting to characterize the region of increased work hardening as a manifestation of Stage II-type deformation. This seems reasonable if the polycrystalline layer is quite efficient in holding up a large number of dislocations. In zinc crystals, since it is well-known that the slip distance can be as large as the specimen dimensions, one would only expect to entrap a large number of dislocation with very large crystals or by applying surface layers. If this interpretation of increased strain hardening is correct, it might present a method of studying Stage II in single crystals of zinc. This research was supported, in part, by the Office of Naval Research (1959). The encouragement and stimulation given by Professor E. R. Parker is gratefully acknowledged.

Jan 1, 1965

By D. Y. F. Lai, R. J. Borg

Tracer diffusion of Fe59 has been measured in the a-stabilized Fe-1.8 at. pet V alloy from 700° to 1500°C. The activation energies are obtained in both the presence and absence of magnetic order. Furthermore, it is established that diffusion in the alloy is identical to that in pure iron and consequently the values of Do and Q accurately represent the temperature dependence for self-diffusion. The purpose of this investigation is to obtain an accurate estimate of the temperature dependence for self-diffusion in bee iron in both the presence and absence of magnetic order, and, in so doing, to establish the temperature range of the magnetic effect.&apos;" As the temperature interval suitable for diffusion measurements is severely limited in both bee phases of pure iron because of the intervening fee ? phase, the experiments were performed on an a-stabilized alloy containing 1.8 at. pet V. This alloy is bee over the entire range from room temperature to the melting point. Although there have been several independent investigations of self-diffusion of iron in a, iron,1, 3-6 there still exists considerable disagreement regarding the values of Do and Q for the paramagnetic region. The two systematic studies of diffusion in 6 iron6, 7 previously reported are also only in fair agreement; but in view of the extremely small temperature range available for diffusion studies, i.e., 1390o to 1535oC, this is not surprising. It is comparatively easy to obtain accurate values of Do and Q for the a-stabilized alloy inasmuch as measurements can be made over the entire temperature range -700o to 1500°C. However, in order to assume that these same values apply to pure iron requires careful comparison of the data in the a, and 6 regions in both the alloy and pure iron. We have made several measurements in the appropriate temperature ranges and are unable to establish any systematic difference between the diffusion coefficients of iron in pure iron and in the alloy. We therefore conclude that the values obtained for the alloy are truly applicable to pure iron; the complete evidence favoring this conclusion will be discussed later in this paper. EXPERIMENTAL The experimental methods will be given here only in barest detail since they have been thoroughly de- scribed elsewhere.l, 7 The alloy was prepared by induction melting and chill casting under argon. Diffusion samples were machined from the ingot and annealed in hydrogen for several days at 900°C to give an average grain diameter of 1 to 2 mm. The penetration profiles of the tracer were established by a sectioning technique, the residual activity being counted after the removal of each section. The tracer used is Fe59 which emits two high-energy ? rays of 1.098 and 1.289 Mev, respectively; these were detected by a ? scintillation counter equipped with a pulse-height analyzer. For the measurements in the temperature range -700o to 1130°C the samples were vapor-plated with Fe59, encapsulated in quartz under vacuo, and annealed in resistance-heated furnaces which are controlled to ±1°C. The specimens diffused at higher temperatures are prepared as edge-welded couples, the two halves being separated by a thin washer of the alloy to prevent sintering. The diffusion anneal is then carried out by inductive heating under a dynamic vacuum. The temperature is monitored pyrometrically. RESULTS The diffusion coefficients are obtained from the penetration profiles in the usual way using the error-function complement relationship. The results over the entire temperature range are shown in Fig. 1. In the linear region, 900o < t > 1500°C the least mean squares (lms) values of the diffusion coefficients are given by D = 1.39 exp[-(56.5 ± 1) x 103/Rt] cm2/sec [l] The average departure of the measured diffusion coefficients from the values given by Eq. [1]In order to determine whether or not the slope is truly constant over the entire range from 900" to 1500°C, the data are arbitrarily divided into two groups, the first containing values between -900" and 1133°C and the second between -1133o and 1500°C. The lms values for the two groups are given by Eqs. [2] and [3]: D = 0.519 exp[-55.7 x 103 /RT](900° to 1133°C) cm2/sec [2] D = 1.45 exp[-56.7x103/RT] (1133° to 1500=C) cm2/sec [3] Thus, there is no significant difference between the high- and low-temperature segments of the linear region. This not only assures us of the consistency of the values obtained by induction heating as compared to those obtained from the resistance-heated

Jan 1, 1965

By Kenneth A. Moon

Kenneth A. Moon (U.S. Army Materials Research Agency)—The authors are to be congratulated for a very interesting and valuable paper. Their discussion of the structural implications of the results shown in their Fig. 5, however, requires a small correction. The model of Speiser and Spretnak with 0 = 1 and z&apos; = 2 is perfectly reasonable for the Ti-O a phase, but Eq. [3] is valid only for solutions sufficiently dilute that the number of empty sites excluded by more than one filled site is negligible. The composition at which intrinsic saturation occurs therefore cannot be deduced from Eq. [3], but but must be decided by consideration of the structural details of the phase in question. In this case, intrinsic saturation is expected not at the composition Ti3O but at Ti2O, because Ti2O has the anti-Cd structurel4 for which each filled octahedral site has two nearby empty sites, one in the hexagonal plane immediately above it, and one in the similar plane below it. The other two models, with ?=1/3 and 1/2, imply a distinction between octahedral sites in pure titanium which is not easy to justify. The problem of extending Speiser&apos;s and Spretnak&apos;s treatment in a general way to more concentrated solutions will be the subject of a forthcoming publication by the writer. M. T. Hepworth and R. Schuhmann, Jr. (authors&apos; reply)—The authors wish to thank Dr. Moon for calling to their attention the paper by Holmberg which presents the results of X-ray studies of Ti-O solutions ranging from 0 to 33 at. pct 0 and quenched from various annealing temperatures between 400" and 1800°C. Within these ranges, Holmberg found three types of phases: 1) At high oxygen contents (above 25 at. pct) and low annealing temperatures, he found the anti-Cd(OH)2 structure referred to by Dr. Moon. In this structure each titanium atom is in contact with 3 oxygens and each oxygen is in contact with 6 titaniums, the "available" oxygen sites corresponding to every other layer of octahedral sites in the titanium matrix. The limiting composition is Ti2O. However, this anti-Cd(OH)2 structure was not observed for samples containing 25 at. pct 0 or less. 2) For the specimen containing 25 at. pct 0 annealed at 400°C an ordered Ti3O structure was observed. This structure is related to the anti-Cd(OH)2 structure but differs from it by providing only two oxygens adjacent to each titanium atom and thus is consistent with the model represented by Eq. [5] in our paper, with 6 = 1/3, z&apos; = 0. 3) For dilute solutions and for more concentrated solutions annealed at higher temperatures, the structure corresponds to that of the metal with random distribution of oxygens among all the octahedral sites. Thus, it is interesting to observe that the thermo-dynamic indications of ordering in the vicinity of Ti3O can now be supported qualitatively at least by Holmberg&apos;s X-ray evidence of the existence of an ordered phase of the composition Ti3O. That is, for solutions with oxygen contents close to that of Ti3O, both the thermodynamic data and the X-ray data are more consistent with models based on the ordered structure of Ti3O than they are with the model which Dr. Moon suggests corresponding to "intrinsic saturation" at Ti2O.

Jan 1, 1963

By Joachim J. Hauser

William F. Hosford, Jr. (Massachusetts Institute of Technology)—Dr. Hauser has used a very interesting method to study the interaction of dislocations on different slip systems, but it should be pointed out that the analysis of the experiments is subject to a serious criticism. He states that "the effect of the compression can be adequately described by one slip system." The effect of frictional constraint during the compression makes this quite unlikely. When a block of metal is compressed its lateral expansion requires a sliding of the metal over the compression platens. The role of friction in the interface between the deforming metal and the platens is to increase the normal force required for the compression. Analyses1213 have shown that, in general, the additional force, p, required for the compression is a function of l/h where is the coefficient of friction, I is the length of the speci- men in contact with the platens. For a constant COefficient of friction, isho&apos; shows that P should increase exponentially with pl/h for both plane strain and axially symmetric compression. For a compression specimen, Fig. 7, whose cross section perpendicular to the compression axis is a rectangle whith one dimension L, much larger than the other W, a much larger force would be required for appreciable flow parallel to L than for flow parallel to U7. Therefore lateral strain in the W direction, EW, should be much greater than the lateral strain in the L direction EL. Since L/H = 14.5 for Dr. Hauser&apos;s crystals, and since no lubrication was used, it seems questionable that appreciable lengthening of the crystals could occur. Because the theories of frictional constraint have been derived for isotropic polycrystalline materials, it was decided to make experimental measurements on a single crystal. A section 3.590 in. long was sawed from an aluminum single crystal with a square section of 0.249 in., so that L/W= 14.4. This crystal was tested in compression between dry ground steel platens. Fig. 8 shows the orientation of the crystal and the axis of compression. The face for compression was chosen so that the operation of the single-slip system which would be predicted in the absence of friction would lead to a larger strain in the length direction than in the width direction. The dimensional changes of the width and length as well as the height were measured frequently during the compression with micrometers. Width strains were calculated from the average of readings near each end and at the center. A plot of the length strain, EL, against the width strain E w is given in Fig. 9. The slope indicates that the EL amounted to only 2.5 pct of cw. Other compression tests on similar crystals with shorter lengths also showed severely limited elongation. Even for a crystal with L/W = A.0, EL was only about 26 pct of cw. Because of the frequent interruptions of the tests for dimensional measurements should have decreased the frictional constraint by allowing a reseating of the platens, even smaller elongations under continuous testing seem probable. These tests on long narrow crystals show that essentially plane strain conditions prevailed. We may assume that Dr. Hauser&apos;s crystals deformed similarly. That the compression can be described by the

Jan 1, 1962

By J. Weertman

It is shown that conditions suitable for the conversion of straight dislocations into helices are common in crystals hardened either through long-range dislocation interaction or by jog formation on dislocation lines. For dislocations other than screw dislocations or dislocations nearly screw in orientation, helices can be formed with the aid of core difhsion at temperatures low relative to the melting point of the material. It is assumed that, once formed, helices will degenerate into dislocation tangles. THE dislocation tangle is one major dislocation arrangement observed by the transmission electron microscope which had not been predicted previously from dislocation theory. Since first being observed, the tangle itself has become important in applied dislocation theory. The Cambridge school (P. Hirsch) bases part of its theory of work hardening upon this phenomenon.&apos; In spite of the fact that the tangle obviously is a structure of prime importance to a cold-worked crystal, until recently relatively little attention has been paid to the question of exactly why or how tangles arise. It was generally assumed at first that cross slip and perhaps jog formation will lead to tangles. However Wilsdorf and schmitz2 concluded that cross slip cannot markedly contribute to the tangles. The Wilsdorfs and Maddin3&apos;* have proposed that tangles arise by means of prismatic dislocation loops and superjogs which are produced by a precipitation of point defects. Dislocations with superjogs or which have joined with prismatic loops are themselves dislocation sources and they can produce new dislocations, a process they call "mushrooming", which then tangle with each other. There is no question that the Wilsdorfs et al. mechanism as well as the cross slip one can pro- duce tangles. However these may not be the only paths leading to tangle formation. In this paper we wish to consider another way tangles can form. Our work will be based on the theory of dislocation screw We wish to show that whenever a crystal is cold-worked there always is present a "driving force" leading to helix formation and thus to tangle formation. We do not expect to observe well-developed, regular helices in cold-worked metals. Rather we would expect that after a helix is formed random interactions with nearby dislocations will cause a well-developed helical dislocation to go over into an irregular form. A dislocation tangle is, of course, a grouping of dislocations with irregular forms. Therefore we feel that, in accounting for the formation of dislocation tangles, it is sufficient to show that many dislocations in a cold-worked crystal will take on helical form. The crucial step is helix formation. In order to have a straight dislocation assume a helical form, diffusion of vacancies, or interstitials, must occur. Our discussion of tangle formation therefore will be limited to crystals which have been deformed at temperatures ranging from about one fifth to half their absolute melting point. In this tem per a ture range appreciable core diffusion is possible. (There is no problem in explaining tangle formation at temperatures greater than half the melting point. At such temperatures dislocation climb through a bulk diffusion of point defects gives an extra degree of freedom to dislocation motion. This extra freedom of motion makes tangle formation easy.) At temperatures much below one fifth the absolute melting temperature diffusion processes are sufficiently slow that helix formation does not occur. Of course helix or tangle formation could occur in material deformed at very low tempera-tures which was subsequently warmed in order to make the thin specimen necessary for electron-microscope examination. Since within the temperature range of approximately one fifth to half the melting temperature the diffusion of point defects in the core of the dislocation is rapid, we can reasonably assume that within

Jan 1, 1963