Search Documents

By R. F. Mehl, R. F. Bunshah

A fast amplifier technique has been developed for the measurement of the rate of propagation of martensite in an Fe-29.5 pct Ni alloy. The time of formation of one plate of martensite is 3x10 sec and the rate of propagation is 3300 ft per sec approximately. IT has been known for some time that the plate-like structural unit of martensite forms from austenite with great rapidity. Wiester1 and Hane-mann, Hofmann, and Wiester took motion-pictures of the transformation as it occurs in a 1.65 pct C steel; they demonstrated that a single plate formed fully in the time interval between successive frames, viz., 1/20 sec, thus setting an upper limit. Forster and Scheil,3 using an Fe-Ni alloy with 29 pct Ni, recorded the sonic characteristics of the process electrically, upon an oscillograph, setting the upper time limit at 0.002 sec. Forster and Scheil,~ measuring the change in electrical resistance in the same alloy upon a cardiograph, set a limit of 0.02 sec. Forster and Scheil5 later, employing the same alloy, improved their technique, reporting an upper limit of 7.10 sec. In studying signals of such short duration, it is an important question whether the frequency response of the electrical system used is high enough compared to frequency of the pulse measured, or, put differently, whether the system is able to reproduce without distortion the signal arising, in this case, from the formation of a single martensite plate. Forster and Scheil (referring only to their last paper) obtained signals of a frequency of 30 kilocycles (hereinafter kc); this was about the frequency response of the equipment used; thus, if the signal had a frequency higher than 30 kc, it would still appear as a signal of frequency 30 kc. All of these results thus provided upper limits only. Recent developments in electronics have made available equipment with very high frequency response, very high sweep-speeds, high gain, etc. The electrical characteristics of such equipment, used in the present study, are given in Table I. Such equipment offers obvious attraction in the study of the rate of propagation of a martensite structural unit—and perhaps of other structural alterations proceeding at a very high rate. This paper reports an attempt to develop a technique employing such equipment to measure the time of propagation of a martensite structural unit and the variation of this with temperature, with the mode of formation—athermal and isothermal—in both polycrystalline and single-crystal samples; and from such measurements to obtain the rate of propagation. As will be seen, the results obtained are useful theoretically. Materials All data presented here are for an Fe-Ni alloy of the following analysis: 29.5 pct Ni, 0.027 pct C, 0.135 pct Mn, 0.094 pct Si, balance Fe. There were several reasons for choosing this alloy: 1—it is substantially the one used by previous investigators; 2—it exhibits both the athermala and the isothermal' mode of formation of martensite, both studied in detail by Machlin and Cohen; 3—the subzero temperatures of transformation in this alloy are experimentally very convenient; 4—it exhibits the "burst phenomenon";" 5—the change in electrical resistance upon the formation of martensite, a decrease, is great, approximately 50 pct.' The polycrystalline specimens were in the form of wires of 0.025 in. diameter; the single crystals were 1x1/4x1/4 in. Experimental Methods Electrical Apparatus: Fig. 1 is a schematic drawing of the electrical circuit used. The principle used in these measurements is the same as that used by Forster and Scheil." A small direct current, about 1 to 2 amp, is passed through the sample. When a martensite plate forms, the resistance of the sample changes and a high frequency signal is generated. This signal is picked up by the probes attached to the sample, fed into the bank of amplifiers and thence to the vertical deflection plates of a cathode-ray oscilloscope. The signal itself triggers the oscilloscope trace which flashes across the tube face and is photographed by means of a 35 mm movie camera at the end of a light-tight hood. The camera has no shutter. As soon as the signal flashes

Jan 1, 1954

By P. G. Shewmon, F. N. Rhines

THE determination of the self-diffusion coefficient of magnesium has been made possible recently by discovery1-1 of a radioactive isotope, Mg28 having a half-life of 21.3 hr,1 and subject to manufacture in useful quantity. In the present research this material was condensed from the vapor phase upon a surface of high purity magnesium. The progress of diffusion of the tracer atoms into polycrystalline magnesium was followed by machining layers and measuring the change in the intensity of radiation as a function of the distance of each layer from the surface. The self-diffusion coefficient was found to be 2.1 X 10-8 sq cm per sec at 627°C, 3.6 X 10-9 sq cm per sec at 551°C, and 4.4 X 10-10sq cm per sec at 468°C; the activation energy is about 32,000 cal per mol. Experimental Procedure Since there was no other published measurement of a diffusion velocity in any magnesium-base material, is was necessary to employ a number of new experimental techniques. The short half-life of Mg28 made it necessary to complete the entire experimental procedure within three or four days. This meant that the work had to be done where a cyclotron was readily accessible and that all operations, prior to the diffusion heat treatment, had to be so designed as to minimize their time requirements. Unusual problems were imposed also by the chemical reactivity of magnesium, its high vapor pressure, and the fact that no satisfactory method for elec-trodepositing magnesium on magnesium is presently available. Finally, the machining and handling of the easily air-borne radioactive-magnesium chips involved certain health hazards, resulting in the need for further experimental restrictions. Preparation of Mg28 The Mg28 was produced in the Carnegie Institute of Technology syncrocyclotron by the neutron spallation of chlorine.5 his involved bombarding a 2 gram crystal of high purity NaCl with a beam of 350 mev protons for a period of 2 hr, after which the crystal was dissolved in warm water and the Mg28 was concentrated and purified by chemical means (see Appendix). About 50 microcuries of Mg28 thus were obtained in the form of magnesium oxinate (8 hydroxyquin-olatc?), which was ignited in air to produce MgO. This in turn was reduced to magnesium metal vapor, by the method of Russell, Taylor, and Cooper," in the vacuum apparatus shown schematically in Fig. 1. Here the essential part is a tantalum ribbon, slightly dished to receive the MgO. The ribbon, pre- viously outgassed at high temperature, is heated to about 1700°C by passing an electric current through it, whereupon tantalum oxide is formed, magnesium vapor is released almost instantaneously, and condensed partly upon the diffusion sample. Diffusion-Sample Preparation: Hot-extruded magnesium rod, 21/32 in. round was used in making the diffusion specimens. The magnesium analyzed as follows: 0.004 pct Al, 0.027 pct Fe, 0.040 pct Mn, 0.0004 pct Cu, 0.0002 pct Ni, and less than 0.01 pct Ca, 0.0004 pct Pb, 0.0011 pct Si, 0.001 pct Sn, and 0.001 pct Zn. A brief study of the crystal texture of this material revealed a sharp fiber texture with the (001) plane roughly parallel to the extrusion axis. Cylindrical samples 1/2 in. long by 5/8 in. were machined from this rod, the end faces dressed on 3/0 emery, and lightly etched with 20 pct HC1 in water. These samples then were annealed for at least twice the intended time of diffusion, at the intended diffusion temperature, in order to stabilize the grain structure at about 1 mm average diameter. The annealing treatments were conducted in argon in the same apparatus and in the same manner as the subsequent diffusion treatments, which will be described presently. Thus, a strain-free plane surface was produced, but there remained a layer of MgO which had largely to be removed before the layer of Mg28 was deposited. Most of this layer was taken off by two light passes over 3/0 emery paper. The balance of the oxide and a thin layer of metal were then removed by etching 5 to 10 min in 4 pct nital (4 pct HNO3 and 96 pct ethyl alcohol) made with absolute alcohol. There followed immediately three quick rinses in: 1-49 1/2 pct methanol, 49 1/2 pct acetone, and 1 pct formic acid, 2-50 pct methanol and 50 pct acetone, and 3-pure benzene. This procedure is essentially that of Sturkey.7 The resulting surface, which was of almost elec-tropolished brightness, remained plane and was free of cold work. It could be kept clean by storing under benzene, or in a desiccator; short exposure

Jan 1, 1955

By H. S. Cannon, F. N. Rhines

The application of a static load to a copper powder compact during sintering at an elevated temperature accelerates the rate of sintering in such a way that a given load induces the same proportional increase in rate for all times of sintering. It is shown that sintering under a load is like creep under a fixed load in that the stress required to accomplish a given degree of densification is proportional to the logarithm of the sintering time. VIRTUALLY all of the theories of sintering that have been put forward within recent years have contained the assumption that the chief driving force of the process is the surface tension resident in the exposed surfaces and internal pores of the compact, or powder mass. An externally applied load might be expected to provide a somewhat equivalent driving force for sintering. It is known that the application of a compressive load, during sintering, hastens densification, but it is by no means clear whether sintering under the influence of surface tension alone and sintering under an applied load are fundamentally similar processes. The present study was undertaken in an effort to obtain an answer to this question and with the hope that the answer may contribute to an understanding of the mechanism of sintering. It has been found that there is a close relationship between sintering and creep processes. Copper powder compacts were sintered in hydrogen at 1000 °C under dead loads of 0 to 165 psi and the progress of sintering was observed by means of density measurements. Using a commercial reduced oxide copper powder, see Table I, cylindrical compacts Yz in. x x in. high were made at a pressure of 12,500 psi. For sintering, these were placed in a cylindrical graphite container into which they fitted snugly. A compressive axial load was applied to the compact through a graphite rod of similar diameter, which rested upon the upper end face of the compact; iron weights, in suitable amount, were placed upon the graphite rod to provide the load. This assembly was encased in a vertical silica tube within a resistance-type furnace capable of maintaining a substantially constant temperature within the working zone. After various predetermined times at temperature, the samples were removed from the furnace and their density was measured by conventional means. The heating-up and cooling-down times have been included in the computed sintering time, using the assumption that the sintering rate doubles for each increase of 10 °C in temperature. The experimental results are presented in Table 11. Upon metallographic examination of the specimens, it was found that there was no apparent change in the shape of the pores as a result of loading, i.e, no flattening, Fig. 1. Substantial contraction away from the container walls was observed in all samples sintered with loads of 10.5 psi and less.

Jan 1, 1952

By H. S. Cannon, F. N. Rhines

IT. H. Hausner( Sylvunia Electric Products Inc., Bay-side, N. Y.)—The results reported by the authors are interesting because they contribute some information on the principles of sintering and also because of a certain practical value for the manufacture of powder . metallurgical parts in production. Pressed powder compacts are frequently sintered in a boat in several layers so that the lower layers are actually sintered under the weight of the upper layers and, therefore, are frequently deformed. The authors gave information on this deformation process by determining the effect of the load on the density. The experimental setup of the authors, however, has a disadvantage—that the sintering occurred in a cylindrical container so that the deformation was stopped in the direction of the diameter. The authors should have rather called this a study of hot-pressing than of sintering. I want to call the attention of the authors and others interested in sintering under load to the experimental work described by Dawihl and Rix.' These authors sintered cylindrical compacts under a lengthwise ten-sional load of 0 to 25 g and found that the shrinkage in length and diameter changes considerably with the load, but that the volume shrinkage is affected very little by the load. H. S. Cannon (authors' reply)—We wish to thank Dr. Hausner for calling attention to the experimental work of Dawihl and Rix. The fact that they found volume shrinkage little affected by varying tensional loads during sintering seems to support the contention that the forces due to loading are equivalent to and additive to the forces which cause densification. This contention is the basis of our comparison of sintering and creep.

Jan 1, 1952

By R. W. Heckel, H. W. Paxton

The growth of grain boundary films and Widmanstiitten plates of proeutectoid cementite in hypereutectoid steels ulas considered quantitatively in three plain carbon steels and one "pure" iron-carbon alloy. Growth was found to take place at a slower rate than would be predicted on the basis of a model invoking rate control by long range diffusion of carbon. The films in the pure alloy grew faster than those in the plain carbon steels. A possible explanation is suggested. In 1904, Nernsst' and Brunner2 postulated that ionic crystals growing in supersaturated aqueous solutions grew at rates determined by diffusion through the solution to the growing crystal interface. The gradient down which the diffusion took place was assumed to be fixed by the supersaturated bulk solution concentration, the concentration dt the crystal surface and some arbitrary distance. The flux of material to the crystal was described as Flux where D is the diffusion coefficient, A the crystal surface area, Ci the bulk solution concentration, Co the saturated solution concentration, and 6 the thickness of the diffusion zone. It is obvious that the expression is an adaptation of F+ick's3 first law of diffusion (steady state). This was the initial formulation of what will be defined here as "diffusion controlled growth," i.e., growth limited by the long range diffusion of a component either to or from the advancing interface assuming both phases at the interface to be in equilibrium. This idea of equilibrium interface concentration implies an infinite reaction rate at the interface. Volmer, Zener,' and 11schner6 have derived expressions for the diffusion controlled growth of spherical precipitates in metallic systems. These analyses agree that the growth in a given direction should be parabolic with time. Brown and Hawkes7 observed that graphite, formed both from the decomposition of cementite and from the decomposition of eutectoid austenite in cast irons, grew according to a parabolic relationship and, therefore, concluded that the laws of diffusion controlled growth, as defined above, were obeyed. They, among others, failed to consider, however, that parabolic growth is a necessary, but not sufficient, condition for diffusion controlled growth. Mehl and Dube"' have pointed out that equilibrium may not be realized because the concentration at the interface cannot change instantaneously from that of the bulk to the saturation value. Furthermore, they propose that an equilibrium solute concentration in the parent phase adjacent to the precipitate would remove the thermodynamic activity gradient across the interface and, thus, stop the growth process. Frisch" has also emphasized the importance of interface concentrations that vary throughout the growth of a new phase. The variation in activity across an interface, as discussed by Onol' and Cahn and Hilliard,12 is of fundamental importance to this question. Ham13 considered the data of Turnbull14 on the growth of barium sulfate in aqueous solutions. He concluded that the data could be analyzed in terms of a radiation-type boundary condition. Flux across interface = K (c, - C,) [21 where C, is the concentration in the solution at the interface, Co the saturated solution concentration, and K the proportionality constant. The edgewise growth of Widmanstlitten plates, frequently referred to as the rate of lengthening, was first considered by Zener.'' Basing a mathematical analysis on steady-state (constant velocity) growth he concluded that where G is the growth rate, D the solute diffusivity, Ci the concentration in the bulk, Co the equilibrium interface concentration, C, the concentration in the precipitate, a a constant, Y the radius of curvature of the edge of the plate, and rC the critical radius, i.e., the radius which would increase the activity of solute at the interface at equilibrium under the influence of capillarity to that of the bulk. Zener1'

Jan 1, 1961

By Elmars Ence, Harold Margolin

The Ti-Fe and Ti-Fe-0 systems were re-examined because of the controversy regarding the existence of Ti2Fe, and to consider all available data points to the existence of Ti,Fe. The Ti-Fe-0 system contains two ternary compounds: E, corresponding to Ti2Fe; and y. SEVERAL authors have investigated the constitution of titanium-rich alloys of the Ti-Fe system. The most extensive study of titanium-rich portions of the Ti-Fe system was by Van Thyne, Kessler, and Hansen,' who investigated this system up to 50 pet Fe and used the highest purity materials available for their alloy preparation (iodide titanium was used for alloys up to 20 pet Fe). According to Van Thyne et al., phase relationships in the region of investigation are governed by phases a,ß, and TiFe. No evidence of an intermetallic compound Ti,Fe was found. although earlier. works by Laves and Wallbaum,' and Duwez and Taylor' reported its existence. Rostoker's study on the occurrence of TilX phases" confirmed the findings of Van Thyne et al. Rostoker proposed that the compound Ti Fe found by Duwez does not exist but is actually a ternary compound, Ti1Fe3O, originated by inadvertent oxygen contamination. The partial isothermal section of the ternary Ti-Fe-0 system as reported by Rostoker' seems to confirm this explanation. In the course of phase diagram work conducted at New York University, however, certain irregularities were observed to be associated with ternary systems of the type Ti-Fe-X. The ternary phase 6, observed in the systems Ti-Fe-Mo' and Ti-Fe-V." was found to be structurally identical to the com-pound Ti2Fe as reported by Duwez and Taylor, and to Rostoker's compound Ti1,Fe2O. Since the amount of oxygen possibly present in the Ti-Fe-Mo and Ti-Fe-V systems could not account for the observed amounts of the compound, it appears that the d phase in these systems is not Ti,Fe,O. On the other hand, considering the amounts of the 6 phase, particularly in the Ti-Fe-Mo system, the location of the phase was uncertain. It was felt that the introduction of the Ti2Fe phase would alleviate some of the inconsistencies of the two systems. Because of the uncertainties relating to the phase Ti2Fe, or Ti1Fe2O, it appeared worthwhile to re-examine the Ti-Fe and Ti-Fe-O systems in the vicinity of these compounds with highly sensitive metallographic techniques and with X-ray methods. Experimental Procedure Alloy Preparation—For the study of the binary system four alloys were prepared, containing 26.9 (30 wt pct), 34.0 (37.5 wt pet), 41.2 (45 wt pet), and 56.3 (60 wt pet) atomic pet Fe, and for the ternary Ti-Fe-O system 18 alloys were prepared in the composition range of 1.6 to 16.9 atomic pet 0 and 24.3 to 55.7 atomic pet Fe. The materials used for the Ti-Fe alloys were iodide titanium (99.98 pet Ti, 0.002 pet 0) and Ferro-vac-E iron (99.95 pet Fe, 0.0052 to 0.0072 pet 0). For the Ti-Fe-O study the materials used were sponge titanium containing less than 0.06 pet 0, Ferrovnc-E iron, and Baker's analyzed TiO2. The nominal weight and atomic percentages of the ternary alloys prepared are shown in Table I. Melting—Charges of 10 to 15 g were melted in a nonconsumable arc furnace in argon atmosphere according to the technique described elsewhere.'" Since there was practically no weight loss through the various stages of melting and oxygen losses have not been observed previously, it appeared justifiable to use nominal composition for interpretation of data. Iodide titanium control buttons, of the same mass as the Ti-Fe charges, and melted under the same conditions, were used to check hardness pickup during melting. A hardness increase of 2 to 3 Vhn was found. Heat Treatment—The specimens were annealed in quartz capsules under argon. To avoid contact between the specimen and capsule material all specimens were wrapped in titanium sheet. The annealing times are given in Table 11. The attain-

Jan 1, 1957

By R. Scot Clark

The reactively sputtered silicon dioxide capacitor -was developed for use in the monolithic integrated circuit, thereby requiring the capacitor to be fabricated on an oxidized silicon slice. The silicon dioxide film is formed by reactively sputtering silicon in a 50 pct Ar-50 pct 0, atmosphere and the physical properties are indistinguishable from silica. A practical upper limit of capacitance per unit area of 0.25 pf per sq mil is reached with the capacitors fabricated from the Si0, film. Methods shown to enhance this value are double stacking the thin-film capacitors and addition of aluminum to the silicon cathode during sputtering to form an aluminosilicate dielectric. A major contribution to the field of diffused integrated circuits would be the addition of auxiliary passive devices for replacement of diffused passive devices which are currently being fabricated for many integrated circuits. This accomplishment would increase the flexibility of the diffusion process in order to concentrate more on transistor parameters rather than on fabricating diffused resistors and capacitors. To pursue this technique, a thin-film SiOz capacitor was developed which has _ proven applicable to many integrated circuits. Although the SiOl capacitor has an application, the major deterrent from extending the integrated circuit into many useful areas such as low-frequency linear application is the low dielectric constant of SiO, and the low usable capacitance in an integrated circuit. This paper will cover the sputtering techniques used for forming the SiO, film, the physical and electrical properties of sputtered SiO,, and yield information concerning the application of these capacitors to integrated circuits. REACTIVE SPUTTERING Reactive sputtering is simply the sputtering of a metal or semiconductor in glow discharge in the presence of a reactive gas. The material to be sputtered is seen in Fig. 1 as the cathode of the system and is bombarded with positively charged ions of the ambient gas. In this investigation, the ambient gas is composed of 50 pct Ar and 50 pct O2 at a pressure of approximately 2.5 X l0-' Torr. The ions are accelerated toward the cathode in the cathode dark space by the electric field and upon striking the cathode effect a momentum energy transfer to the surface atoms. The transferred energy is sufficient to dislodge a surface atom or atoms and these particles diffuse through the glow discharge to the collector or anode. Several mechanisms of oxidation in the reaction chamber can be considered during reactive sputtering. 1) The surface of the cathode is oxidized and the oxide is subsequently sputtered. 2) Oxidation and sputtering occur simultaneously. 3) Sputtering of the cathode material is followed

Jan 1, 1965

By A. S. Grove, C. T. Sah, E. H. Snow, B. E. Deal

A summary of- several recent investigations in to the properties of the metal-oxide-silicon system is presented. A major portion of these studies makes use of the MOS capacitance-z)oltage method of' analysis. The particular areas of investigation which are reported include: 1) a general survey of the electvical properties of thermally oxidized silicon surjbccs; 2) a study of ion migration through silicon dioxide films ; 3) measurements of electron and hole mobilities in surface inversion layers; 4) a study of impurity redistribution due to thermal o.ridatiotz; and 5) measurements of the rates of oxidation oj-heavily doper7. silicon. THE importance of the metal-oxide-semiconductor (MOS) system in the semiconductor industry is well-known. In addition to its importance in the "planar" device technology,' the MOS structure is now also used in the fabrication of active solid-state devices. Consequently, extensive efforts have been made recently to obtain a better understanding of the characteristics of this system. A summary of some studies of the MOS system conducted in our laboratories during the past year is presented. For the most part these studies used silicon as the semiconductor, along with silicon dioxide and aluminum as the other two components of the system. Since the MOS capacitance-voltage method of analysis was used extensively in these studies, we will first briefly describe its nature and consider some of the possible causes of deviation of experimental observations from the simple theory. We will then outline the various related areas of investigation carried out in our laboratories and will briefly indicate some of the results. It should be noted that the purpose of this paper is merely to provide a brief summary of MOS studies. More detailed discussions of the various areas of investigation are given in the references cited. PRINCIPLES OF THE MOS C-V METHOD OF ANALYSIS' A sketch of the MOS structure is shown in the upper portion of Fig. 1. In this case the insulating film is Si02 and the semiconductor p-type silicon. If a large negative bias is applied to the metal field plate, holes are attracted to the silicon surface. The silicon then behaves much like a metal and the capacitance measured is that of the oxide layer alone, Co. If a small positive bias is applied to the aluminum, holes are repelled and a region depleted of majority carriers is formed at the silicon surface. This depletion I-egion adds to the width of the dielectric and the measured capacitance begins to drop. With increasing positive bias, the width of the electrical depletion region increases. At some large positive bias an inzevsion regiotr is formed at the surface and additional charges induced in the silicon appear in the form of electrons in this narrow inversion region. Thus the depletion-region width approaches a maximum value and, consequently, the capacitance reaches a minimum value and then either levels off or rises again depending on the measurement frequency and the rate of equilibration of the minority carriers in the inversion layer.3 Band diagrams, along with the corresponding charge distributions, are shown in Fig. 1 for the above bias conditions. If minority carriers cannot accumulate at the surface to form an inversion region, the depletion-region width continues to increase with increased positive bias and the capacitance drops toward zero as in a reverse biased p-n junction. The effect of a work-function difference $hs between the metal and the silicon, and of surface charges per unit area Qss located at the oxide-silicon interface, is simply to attract charges in the silicon much like the applied bias. It can be shown that this results in a parallel shift of the capacitance-voltage characteristic along the voltage axis by an amount corresponding to AV = -$bIs + Qss/Co. Theoretical curves have been calculated4 giving the capacitance of the MOS structure C normalized to the oxide capacitance Co vs the quantity VG here VG is the voltage applied to the metal field plate. In Fig. 2 such calculations are shown as points for a particular oxide thickness and bulk impurity concentration for a p-type semiconductor. (For an n-type semiconductor the curves would be mirror images of these.) All three cases, i.e., low frequency. high frequency, and depletion, are indicated. Also shown in the figure are recorder tracings of the characteristics of actual devices. These characteristics have been shifted along the voltage axis to compensate the effect of surface charges and work-function difference. It is evident that agreement between experiment and theory is good. The nature of this shift along the voltage axis is

Jan 1, 1965

By B. L. Averbach

Recovery and primary recrystalliza-tion in cold worked metals are usually considered as two competing processes. Some of the effects which usually accompany recovery are: alleviation of stress corrosion tendencies, changes in thermal emf,1 damping capacity,2 electrical resistivity,2 and magnetic properties,3 and only minor changes in hardness or the related strength properties. During primary recrystalliza-tion new unstrained grains are formed at the expense of the strained matrix. These new grains eventually become visible metallographically, and nucle-ation and growth kinetics have been indicated for this process.4,5 Frequent attempts have been made to study the cold-working phenomenon by observations on the line broadening by X ray diffraction patterns. Relatively few measurements of line intensities have been made, although Brind-ley and his collaborators, 6,7,8 by means of film techniques, compared the intensities of cold worked Cu, Ni, and Rh patterns with those from chemically precipitated powders. These precipitated powders were presumed to be strain free, and it was found that the intensities for the cold-worked materials progressively decreased as the Bragg angle increased except for the first line, where there was an increase due to reduction in extinction. This was interpreted as a randomness in atomic position induced by cold work. Such randomness is similar to that caused by thermal agitation and has been described as "frozen heat" displacement of 0.08-0.10 A from the mean atomic position. In a recent study9 on the effect of cold work in metals on their powder pattern intensities, the changes in integrated intensity for heavily cold worked alpha brass were observed as a function of the annealing temperature. These measurements were made with a manually operated Geiger-counter spectrometer using CuKa radiation monochromated with a rock salt crystal. Intensity measurements were made with a scaling meter over small intervals of angle, and the equipment was so arranged that the diffracted and incident beams made equal angles with the specimen. Intensities could be compared directly by simply interchanging specimens, and comparisons from day to day were made with a standard whose line intensities did not change on aging. It was shown that a cold worked alpha brass standard was stable for at least a year. Table 1 indicates the results of the integrated intensity measurements on a 70 Cu-30 Zn brass. In the sample preparation, a brass plate was first cold rolled 50 pct and then filed, screened to —325 mesh, compacted into briquettes at a pressure of 60,000 psi and finally annealed for one hour at various temperatures up to 400°C. The briquetting pressure did not seem to influence the integrated intensities, and most of the cold work was introduced by the filing. Although this method of cold work is not quantitative, it was used to obtain random orientation (and thus uniform diffraction lines) in order to make accurate measurements of integrated intensity. Back reflection patterns were taken in each case to check the uniformity of the lines, and from the observed line broadening it was apparent that this type of plastic deformation was quite severe. Care was taken to traverse the entire background of the pattern and to assign to each peak the total intensity above this background. The bases of the diffraction lines were quite broad and spread out over several degrees, even for the narrow peaks. The theoretical intensities were calculated to include a temperature correction, a dispersion correction, and a Lorenz- polarization factor corrected for the crystal monochromated beam. In Table 1 it was necessary to match the calculated and observed values at only one point, and the rest of the experimental values were converted directly to this arbitrary scale. The integrated intensities in Table 1 are listed in arbitrary units, and the accuracy was sufficient to reproduce any of the measured line intensities to within + 1.5 units. It is evident that the percentage error on the strongest line (111) was quite low. The calculated values and the observed intensities for the cold worked material matched reasonably well. As the annealing temperature was raised, however, the intensity of the strongest reflections, particularly the (lll), decreased measurably. Since the background intensities of all of these patterns were identical, such behavior could be interpreted as a primary

Jan 1, 1950

By J. Washburn, R. Drouard, E. R. Parker

Temperature dependence of the rate of recovery in zinc single crystals after a simple shear deformation at low temperature was investigated. Some tentative suggestions regarding the annealed and strain-hardened states of a crystal are discussed. RECOVERY may be defined as the gradual return of the mechanical and physical properties of strain-hardened metal to those characteristic of the annealed material; an increase in temperature increases the rate of recovery. The annealing process in strain-hardened polycrystalline metals is complicated by the inhomogeneity of strain which always exists in aggregates. Polygonization in bent regions of the crystals and growth of new almost strain-free grains starting at points of severe local distortion1-:' make it almost impossible to isolate and study the recovery process. Homogeneously strained single crystals, however, do not polygonize or re-crystallize and hence they can be used advantageously to study recovery. In such crystals strain hardening is completely removed by recovery alone. Since recovery is a process whereby certain lattice disturbances introduced by plastic flow are gradually reduced, a knowledge of the rate and temperature dependence of this process for various conditions of prestrain might be helpful in formulating a model of the strain-hardened state. For simplicity it seemed desirable to limit the type of prestrain to the simplest obtainable, i.e., simple shear strain. In the experiments to be described, recovery was studied by observing changes in the stress-strain curve of prestrained zinc single crystals held for various times at temperatures above that employed for straining. Single crystals were grown from the melt by a modified Bridgeman technique from Horse Head Special zinc 99.99 pct pure, and from spectrographically pure zinc 99.999 pct pure. They were grown as 1 in. diameter spheres and acid-machined' to the final specimen contour. The test section was a cylinder about 1/8 in. high and 3/4 in. in diameter. The conical sections adjacent to the test section were cemented into the grips so the load could be transmitted to the crystal as uniformly as possible. The specimens were oriented so that in testing the maximum shear stress was applied along one of the slip directions, [2110], in the (0001) plane. Details of the production and testing of such specimens have been presented.' Each test was carried out according to the following schedule: 1—The crystal was strained at — 50°C until it reached a maximum shear stress, ,,,. The strain rate was approximately 5 pct per min in all cases. 2—After straining, the crystal was unloaded before the temperature was changed. Unloading required about 3 min. 3—The temperature of the specimen was then increased from — 50°C to the temperature, T, of recovery. This change in temperature was completed in a time of less than 2 min. The specimen remained at temperature, T, for a time, t, which differed for the various specimens. 4—Thereafter the temperature was again reduced to — 50 °C in approximately 3 min. 5—While at —50°C, the stress-strain curve after recovery was obtained. 6—The specimen was then unloaded and annealed for 1 hr at 375 °C in a helium atmosphere to bring about complete recovery. Cooling to room temperature after anneal required 90 min. 7—The same crystal could be re-used for another test because the plastic properties after annealing closely duplicated those of the original crystal. The specimen was immersed during the test in a bath of methyl alcohol which, through a system of tubes, could be pumped through either of two heat exchangers to regulate the temperature; this was accomplished by circulating the liquid through coils immersed in a bath of acetone and dry ice for cooling or in a bath of warm water for heating. Test temperatures were thus maintained constant within ±1°C. The — 50°C temperature was low enough so that no measurable recovery occurred during unloading and reloading. The stress-strain curve continued after recovery along a path below, but approximately parallel to, the path of a curve obtained in an uninterrupted test. Fig. 1 shows some of the results from a specimen of 99.999 pct Zn. The amount of downward displacement of the curve due to recovery was a

Jan 1, 1954

By E. C. W. Perryman

The recovery of X-ray line broadening, hardness, and electrical resistivity from cold worked AI-Mg alloys has been investigated. The results, together with those from density measurements, suggest that magnesium atoms freeze in the excess vacancies formed by cold working, and that this affects the climb of dislocations during the recovery process. PREVIOUS work1, " has shown that while 1 pct Mg increases the activation energy for recrystalliza-tion, it decreases the activation energy for the recovery of mechanical properties and decreases the subgrain size both after cold working and annealing. A similar result has been reported by Varley2 for an A1-2 pct Mg alloy. With plastic deformation at high temperatures, McLean and Farmer' and Rachinger have found that magnesium appears to retard the motion of dislocations and, furthermore, they find little or no effect of magnesium on the size of subgrains formed during high temperature deformation. In view of the apparent contradictions as to the effect of magnesium on subgrain size, the recovery process in Al-Mg alloys has been investigated more thoroughly by studying the annealing out of X-ray line broadening, hardness, and electrical resistivity from alloys cold worked at room temperature. Material Used—The chemical analyses of the alloys are given in Table I. Fabrication—The material used was cold rolled 20 pct at room temperature, the grain size prior to the last cold rolling operation being similar to that used in previous work; i.e., 230. The experimental procedure for annealing, determination of X-rav half-veak breadth. hardness measurement, electEica1 resistivity, and density and metallographic examination was exactly the same as previously described." Variation of X-Ray Half-Peak Breadth with Annealing Time—Isothermal annealing curves at 100°, 200°, and 250°C were obtained for material cold worked 20 pct. The half-peak breadth of the (333) Ka, line directly after cold working is shown as a function of magnesium content in Fig. 1. Only with the 2.87 pct Mg alloy was overlapping of the a, and n.. reflections observed to any appreciable extent. To separate these, Rachinger's method' was used. Typical isothermal annealing curves are shown in Figs. 2 and 3 for 100" and 200°C, respectively. These curves are similar to those obtained previously for super-purity aluminum except that the

Jan 1, 1957

By Louis Raymond, John E. Dorn

The object of this investigation was to analyze the recovery that arises when the stress on a specimen undertaking creep is reduced. For this purpose annealed specimens of high-purity aluminum were precrept under a stress of 1000 bsi to a strain of 0.08 following which the stress was reduced for various periods of time to 10, 250, 500, or 700 psi. When the original stress was reapplied the subsequent creep curve lay above that for the unre-covered state and below that for the original annealed state. Analyses on the kinetics of this recovery as a function of the temperature gave a stress-sensitive activation energy that decreased as the reduced stress was increased from a value of 64,000 cal per mole at 10 psi to 37,000 cal per mole at 750 psi. Recovery was also detected and measured during creep under the reduced stress. Following a short initial period, the creep rate under the reduced stress increased monotonically until it reached the secondary-creep rate for the reduced stress. The temperature dependence of this phenomenon was also shown to be correlatable in terms of the previously deduced activation energy for recovery. The activation energies for creep of most pure metals at high temperatures have been shown to agree well with those for self-diffusion.'j2 Since the true secondary stage of creep is usually due to the steady-state balance between the rate of strain hardening and the rate of recovery, it is generally thought that the activation energy for recovery of the creep-induced substructure equals that for creep itself. A shoft time ago, however, Ludemann, Shepard, and Dorn~ found that the activation energy for recovery of the creep-induced substructure in high-purity aluminum under zero stress was almost twice that for self-diffusion, namely about 65,000 cal per mole; obviously recovery under reduced stresses differs in some significant way from the recovery that accompanies the secondary stage of creep. The major purpose of this investigation is to study the effect of stress on the re- covery of the creep-induced substructure in order to provide a better understanding of the recovery mechanism itself. EXPERIMENTAL TECHNIQUE High purity aluminum, containing 0.004 pct Cu, 0.002 pct Fe, and 0.001 pct Si, used in this investigation, was in the form of 0.100-in.-thick sheet which has been cold-rolled to the H-18 temper. Creep specimens were milled from the sheet with their tensile axes in the rolling direction. All specimens were then heated at 686°K for 1 hr followed by air cooling in order to produce an annealed structure which exhibited a uniform equiaxed grain size of about 4 grains per mm. Tests were run in creep machines fitted with Andrade-Chalmers type of lever arms so contoured as to maintain the stress constant to within 0.05 pct of the reported values. Constant temperatures to *O.l°K were obtained by complete immersion of each specimen in a temperature-controlled and agitated bath of molten KN02-KNOs mixture. Where changes in temperature were involved, the change was effected in less than 2 min by manually replacing one bath by another controlled at the second temperature. Displacements over the gage section were sensed by linear differential transformers, the output of which was autographically recorded. The calculated strain measurements were sensitive to 5x EXPERIMENTAL PROCEDURE The following analyses are based on extensions of the previously announced effect of the temperature on the creep strain,2 namely for a = constant, where e = the total true tensile creep strain for a given applied true tensile stress, t = the duration of the test, R = the gas constant, T = the absolute temperature, Q, = the activation energy per mole for creep which is independent of the stress, / = a function of 8, = and of the stress, and a = the stress. The validity of this correlation for high-purity aluminum is demonstrated in Fig. 1 for temperatures in the near vicinity of 600°K; the activation energy for creep, Q,, which is approximately that for self-diffusion, is insensitive to the applied stress

Jan 1, 1964

By H. L. Couch, J. D. Lubahn

In decarburized mild steel, the strain hardening arising from 1/2 pct strain can he partly recovered by subsequent heating, even though re crystallization or grain growth does not occur. This recovery varies from 5 pct in an hour at room temperature to 70 pct at 800°C. The variation with temperature was qualitatively similar to that reported for copper, but occurred at lower temperatures. Except for the initial portion (Bazdschinger effect), the stress-strain curve following recovery can be superimposed on the original curve by shifting it to the left. In other words, the tensile behavior is essentially the same after straining and heating as if there had been less strain and no heating. The importance of this observation in understanding simultaneous deformation and recouery is explained. AN earlier investigation1 on copper showed that strengthening due to strain hardening was augmented by heating at moderate temperatures, but was diminished by heating at higher temperatures, Fig. 1. Auxiliary experiments1 indicated that recovery and strain aging are independent processes, which may proceed simultaneously, and whose effects are additive as indicated in Fig. 1. In order to study recovery by itself, similar tests were made on decarburized mild steel, where strain-aging effects are known2 to be absent. MATERIAL, TESTING PROCEDURE The specimens were made from rimmed, hot-rolled SAE 1020 steel 0.040 in. thick, by stamping out rectangular blanks and end holes, Fig. 2, and pack milling the reduced section. These were wet-hydrogen treated2 for 16 hr at 720°C to remove carbon and nitrogen completely and then heated 1 hr at 850°C in vacuum to stabilize the grain size. Auxiliary ex- periments showed that heating up to 800°C will not cause grain growth. They were prestrained 5 to 7.5 mils per in. using the special grips shown in Fig. 2. Such small strains are less than the "critical" strain for recrystallization and, therefore, will not cause recrystallization, even at high temperature. The recovery treatment after prestraining and before the retesting consisted of heating for 1 hr in wet hydrogen at the desired temperature. In one test, Huggenberger gages were used to define more accurately the loading and unloading curves. Fig. 3 illustrates the creep during the early stages of unloading, the hysteresis due to cold stretching, and removal of hysteresis by heating.3 RESULTS The load-deflection curves had ratner widely differing shapes for different specimens, both before and after the recovery treatment. Some specimens showed perfectly smooth curves, Fig. 4, while others exhibited an inflection more or less suggestive of a yield point (Fig. 4 shows all extreme case). Some curves were much steeper than others; Fig. 5 shows two extremes. Some additional ex-

Jan 1, 1960

By C. L. Meyers, J. L. Lytton, T. E. Tietz

The recovery of tensile flow stress of 99.995-pct Al and Al-1 pct Mg alloy was investigated using a fractional recovery parameter. Tensile strainirg was conducted at room temperature, and recovery treatments were made in the range from 80° to 200°C. Equal values of fractional flow stress recovery were found to result in equivalent structural states. Identical recovery treatments were found to produce constant values of fractional flow stress re covery, independent of prestrain. The activation energy for recovery of the 99.995-pct A1 between 80" and 200°C was found to be 23,000 +2000 cal per mole, and the activation energy for the Al-1 pct Mg alloy was 27,500 cal per mole. 1 HE degree of recovery of a cold-worked metal may be studied by measuring the changes which take place in any of a number of properties. Some of the properties which have been used by past investigators include X-ray diffraction phenomena, electrical resistivity, hardness, creep strength, elastic limit, and flow stress. Despite the apparent preponderance of such studies, little is known concerning the effect of recovery on the strength of metals except as deduced from changes in various types of hardness measurements. Furthermore, little information is avail- able concerning the effects which specific solute atoms have on the recovery process. Limited studies have indicated that in some cases solid-solution additions greatly decrease the rate of recovery,' whereas in at least the case of magnesium additions to alluminum an increase in the rate has been reported.2 The purpose of this investigation was to determine the effect of a 1 pct Mg addition on the recovery of high-purity aluminum. Because of the particular interest in the effect of solute additions on the recovery behavior as related to mechanical properties, the degree of recovery was evaluated in terms of tensile flow stress. For the purposes of this study, a recovery process shall be defined as any process, short of recrystallization, whereby the properties, of work-hardened metals tend toward their values ill the unstrained state. EXPERIMENTAL PROCEDURE Test Materials. The materials used in this study were a 99.995-3 A1 and an Al-1 pct Mg alloy made from high-purity base metal. The high-purity aluminum contained the following wt pct impurities: 0.001 pct Cu, 0.002 pct Fe, 0.001 pct Si, and 0.0004 pct Mg. The Pil-1 pct Mg alloy contained 0.001 pct Cu, 0.002 pct Fe, 0.003 pct Si, and 0.99 pct Mg. Both materials were obtained in the form of as-rolled 0.125-in.-thick sheets, from the Alcoa Research Laboratories, New Kensington, Penn. Tensile specimens were cut from these sheets with the longitudinal axis in the rolling direction. Both materials were recrystallized in a molten salt bath; the high-purity aluminum at 450° C for 20 min and

Jan 1, 1962

By W. D. Ludemann, J. E. Dor, L. A. Shepard

Recovery of the creep resistance of 99.99 pct pure Al was studied at temperatures 540°, 573°, 600°, and 611°K. Poly-crystalline specimens crept under a stress of 950 psi to a strain of 5.5 pct were allowed to recover for periods of from 1 min to 16 days under a residual stress of 4.4 psi. Increased creep rates upon reapplication of the 950 psi stress evidenced softening of the material. The activation energy for the recovery process was found to be 64,000 cal-per mol. THREE major observations have clearly revealed that the creep of polycrystalline aluminum above about 510°K is controlled by the rate of climb of jogged edge components of dislocations past the barriers impeding their motion: 1) Extensive polygonization, characteristic of climb, takes place during creep.'9' 2) The activation energy for creep equals the estimated activation for self-diffusion. 3) The stress law for creep coincides with theoretical predictions based on the climb mechanism.4,5 The decreasing rate during the primary stage of creep must be ascribed to the introduction of additional barriers to the glide of dislocations. During secondary creep the density of barriers must remain constant, the rate at which new barriers are formed being equal to the rate at which they recover. For this reason the nature of the barriers, and their rates of formation and recovery, are significant to a complete understanding of creep. It is proposed here to study the kinetics of the recovery of barriers to creep of high-purity polycrystalline aluminum in the climb range. The technique will consist of creeping a series of tensile specimens under a prescribed stress to given state following which the specimens will be recovered for various times and temperatures under approximately zero stress. The amount of recovery will be determined by comparing creep curves following the recovery. EXPERIMENTAL PROCEDURE The material used for this investigation was a 99.99 pct pure Al supplied by the Aluminum Co. of America in sheets of 0.001-in. thickness having an H-18 temper. The spectroscopic analysis of impurities gave Cu-0.004 wt pct, Fe-0.002 pct, Si-0.001 pct, others-0.000 pct. Creep specimens were machined with their tensile axes in the rolling direction of the sheets. Annealing in a molten potassium nitrate-potassium nitrite bath for 1 hr at 686K, followed by air-cooling, resulted in a grain diameter of from 0.22 to 0.27 mm. Creep machines were equipped with constant stress lever arms of the type suggested by Fullman, Carreker, and Fishher,' which maintained the applied stress to within 0.04 pct of the reported value of 950 psi. Creep strains were calculated from extensions over a 6-in. gage section to the nearest 3 X 10"5. The temperatures of testing, 540°, 573", 600°, and 611°K, were obtained by immersing the specimen and extensometer assembly in a molten KNO2-KNO3 bath. Temperatures during the periods of creeping could be maintained constant to within better than ± 1°K. Correlation of creep data at different temperatures of testing demanded the use of an appropriate temperature-compensated time 8. Previous investigations have shown that the creep of pure metals at temperatures above about 0.55 of their melting temperatures can be correlated by the functional relationship3" e=f(0), s= const. [1] where e = total plastic strain during creep f = a function that depends on stress a = stress 6 = te-Q/RT = a temperature-compensated time t = time under test Q = 35,500 cal per mol = activation energy for creep R = gas constant, and T = absolute temperature In the present series of tests all of the creep was conducted under the same stress of 950 psi. To provide identical initial conditions for the recovery all specimens were first crept to a temperature-compensated time of = 47 X 10-14 min which resulted in a creep strain of 0.055 + 0.005 regardless of the creep temperature. During recovery, a small stress (4.4 psi) was permitted to remain upon the specimens to maintain tautness in the pulling assembly and to prevent mechanical damage to the soft specimen. After each recovery period, the full stress was reapplied and

Jan 1, 1961

By W. L. Phillips

The recovery properties of compression-deformed lithium fluoride single crystals were investigated as a function of prior strain, annealing time, and annealing temperature. The recovery process was studied by observation of etch pits and birefringence stress patterns. The recovery process involved first a redistribution of the dislocations and then a decrease in their number. Only by annealing at high temperatures after small strains (<0.02 in. per in.) was it possible to obtain complete recovery of mechanical properties. LITHIUM fluoride affords unusual opportunities to extend the knowledge of deformation mechanisms of nonmetallic crystalline compounds. Single crystals of high purity are readily available. These crystals deform by slip on the (110) <110> system.&apos; Unlike those of several other ionic crystals, the stress-strain curves of annealed lithium fluoride are reproducible.2 Dislocations generated in the crystals during deformation can be revealed by etch pits.3 This fact makes possible more detailed studies of mechanisms than is usually the case with metals. Also, because lithium fluoride is transparent, it is possible to follow the recovery process with birefringence stress patterns. It has been shown that with metals an increase in temperature increases the amount of recovery for a given time, and the amount of recovery increases with increasing time at the same temperature.4-6 The yield stress has also been shown to be dependent on the rate of cooling after annealing.&apos; Annealing germanium, a covalent-bond material, at high temperatures after deformation redistributes the dislocations and reduces their number. The purpose of the present paper is to extend the knowledge of recovery properties to an ionic crystal, lithium fluoride. The effect of strain, time, and temperature of recovery on the stress-strain curve was investigated. The recovery process was followed by observations of etch pits and birefringence stress patterns. EXPERIMENTAL PROCEDURE Specimens measuring approximately 0.10 by 0,10 by 0.30 in. long were cleaved from a single bulk

Jan 1, 1961

By R. M. Treco

THE purpose of this work has been to investigate recrystallization characteristics of zirconium after cold working and annealing. In order to do this, hafnium-contain ing crystal bars of Foote Mineral Co. iodide zirconium were selected for high purity. maximum diameter (about ½ in.), brightness of surface, and absence of core wire. A typical lot analysis is shown in Table I. In order to obtain uniformity of starting material, all bars were cold-swaged to 0.310 in. diam, annealed for 2 hr at 1000°C, and furnace cooled.

Jan 1, 1957

By R. L. Rickett, S. H. Kalin, J. T. Mackenzie

Aluminum killed low carbon steel, § which is now used extensively for severe deep drawing or other difficult forming operations, is unusual in that its grain structure, after cold reduction and box annealing in accordance with conventional continuous sheet or strip mill practice, often is elongated, although at times it is equiaxed. Since this unusual structure has been found superior for many, but not all, severe forming operations, recrystallization of the steel, both at constant temperature and on continuous heating, was investigated and compared with that of rimmed steel in the hope that something might be learned about the mechanism of, and the factors controlling, the formation of such elongated grains. In this structure, the grains are elongated both in the lengthwise direction of the strip and transverse to this direction, even though nearly all of the extension in both hot and cold rolling is in the lengthwise direction. The grains are thus roughly pancake-shaped, being longer and wider than they are thick, as observed also by Burns and McCabe,1 and as illustrated by the typical structures shown in Fig 1. Fig la, representing a conventional longitudinal section, shows the length and thickness of the grains, whereas Fig Ib shows their length and width as seen by examining a section parallel to the sheet surface. Both illustrate the very irregular grain boundaries usually associated with the elongated grain shape. A finer equiaxed grain structure in this same grade is shown in Fig Ic. Either the elongated or the equiaxed structure may be present in the annealed product, and in rare instances the two types may coexist in a single specimen, as shown in Fig 1 d. Isothermal Recrystalliza-tion of Rimmed and Alamimum Killed Steel An aluminum killed steel known to have an elongated grain structure after conventional processing (Steel B, Table l), was selected for the initial recrystallization studies; for comparison, a rimmed steel, A in Table 1, was used. Samples of each in the form of hot rolled strip 0.075 and 0.095 in. thick, respectively, were cold rolled on a small laboratory mill in steps of about 0.010 in. per pass to obtain total reductions of 40 and 60 pct. Small pieces of the cold reduced strip were heated in lead at selected constant temperatures for one of several periods of time, then cooled in air. Rate of heating in the lead was, of course, very fast. Hardness of the cooled specimen was measured and a longitudinal section examined metallographically. Isothermal recrystallization curves for these two steels at 1050°F, based on hardness of the air cooled specimens, are shown in Fig 2 in which the amount of recrystallization corresponding to each plotted point is indicated. The marked difference in the behavior of these two types of steel is evident. After a corresponding amount of cold reduction, the rimmed steel recrys-tallizes in a much shorter time than the killed steel and the shape of its recrystallization curve, (plotted on a logarithmic time scale), is very different. The curve for rimmed steel indicates that recrystallization is analogous to isothermal transformation of aus-i.enite in that it proceeds at a progressively faster rate up to some 50 pct recrystallization, then at an increasingly slower rate. For the aluminum killed steel, however, the start of

Jan 1, 1950

By H. P. Leighly, J. W. Marx, H. L. Walker

A relationship between recrystallized grain size and prior deformation is predicted from elementary statistical considerations, and reasonable agreement with experiment is obtained. RECRYSTALLIZATION phenomena have received much theoretical and experimental attention, and very adequate discussions of the various mechanisms suggested have been given by Anderson and Mehl,&apos; Cook and Richards,&apos; and Burgers," and others. It suffices to note here that the initial concepts fitted one of two general hypotheses: I—The new crystals are nucleated within microscopic regions at which there is a concentration of internal stress, i.e., volume elements of maximum strain energy. 2—The new crystals have their inception in volume elements of minimum strain energy. In a comprehensive early review of the subject, van Arkel&apos; considered both possibilities and came to the conclusion that hypothesis 2 was inconsistent with the experimental fact that the number of nu-cleations increased with increasing deformation. He suggested that, if allowance was made for a preliminary atomic rearrangement at the points of large internal strain, hypothesfs 1 provided the more reasonable concept. Following the initial observations of Collins and Mathewsonb nd Crussard,Y he systematic X-ray investigations of Cahn&apos; and Guinier and Tennevin" established the existence of a polygonized structure in metals that had been cold-worked, and subsequently annealed, under conditions such that re-crystallization did not occur. This new experimental evidence by no means disproved the first nucleation hypothesis, a fact recognized by both Beck" and Cahn,10 who independently suggested nucleation mechanisms based on the idea that regions of large strain might polygonize before the bulk of the less deformed material. These mechanisms, together with others recently proposed," are dependent upon the prior existence of some degree of polygonization in the deformed material. On the other hand, a critical examination of the published data leads to the supposition that polygonization may not be the necessary precursor of recrystallization, but merely its favored energetic competitor. In regions where the lattice distortion is not too severe, the limited adjustment afforded by polygonization may permit the lattice to drop to a relatively stable low energy state in which the nucleation of a new crystal becomes improbable, even at elevated temperatures. Such structures, of course, might provide a fertile field for the growth of nuclei, if these could be introduced. In regions of concentrated lattice distortion, polygonization alone may not be able to accomplish an adequate strain reduction, and in these volume elements sufficient thermal activity permits the more drastic atomic rearrangement involved in nucleation. Completely polygonized structures are known to withstand extended heating close to their melting points," with no other effect than a slow enlargement of some of the subgrains at the expense of others. Once started, macroscopic recrystallization might be expected to go to completion in a matter of minutes under similar heat treatment. Crussard12 and his coworkers have found that recrystallization may occur in single crystals of impure (about 99.7 pct) aluminum without any detectable prior polygonization. The indication here is that large Cottrell impurity atmospheres inhibit dislocation movement, thus suppressing polygonization to such an extent that the alternative process, recrystallization, dominates the lattice restoration. Experimental studies relating recrystallization directly to stored energy have not, at the present time,

Jan 1, 1954

By E. C. W. Perryman

IN spite of the large amount of work which has been carried out on the recrystallization of aluminum and its alloys, there has been no complete investigation of the recrystallization characteristics of A1-Mg alloys. The most complete investigation is that reported by Bungardt and Osswald&apos; who, working with commercial-purity materials, found that the temperature both of beginning and of complete recrystallization increased with magnesium content up to about 2 pct Mg, then decreased with a further increase in magnesium up to 5.0 pct, and thereafter increased again. Michaud and Segolz working with 99.93 pct A1 also found the temperature of beginning of recrystallization to decrease as the magnesium content increased from 2.13 pct to about 4.5 pct. Chossat3 using aluminum of 99.99 pct purity found that 0.12 pct Mg had no effect but that 0.4 pct Mg increased the recrystallization temperature by about 30°C. That no effect was found for 0.12 pct Mg does not agree with the later results reported by Demer and Beck4 who found that 0.025 pct Mg in 99.99 pct A1 increased the time for complete recrystallization and furthermore increased the activation energy for the recrystallization process. Kratz" using 99.997 pct A1 found that 0.5 pct Mg increased both the temperature of beginning and of complete recrystallization by 100°C. Chadwick and Hooper" working with commercial-purity material found that 1 pct Mg increased the temperature of complete recrystallization by about 40 °C for cold reductions in the range 17 to 72 pct. With smaller amounts of cold work produced by stretching (10 pct strain), they found no significant effect of magnesium. The latter result agrees with that obtained by Williams and Eborall&apos; who found no effect of 2 pct Mg on the recrystallization temperature of commercial-purity aluminum strained 10 pct or less by stretching. Experimental Procedure Material Used: The chemical analyses of the materials used throughout this investigation are given in Table I. Fabrication: The alloys were cast in iron molds which had been washed thoroughly with alumina. After scalping, the ingots were annealed for 16 hr at 450°C, pressed, annealed at 450°C for 16 hr and hot rolled to 0.25 in. After hot rolling, the slabs were alternately annealed and cold rolled to 0.036 in. Cold-rolling reductions between intermediate anneals were kept at about 30 pct and a high intermediate annealing temperature of 450°C was used so as to obtain, as nearly as possible, similar starting grain sizes for the different materials. All intermediate anneals were done in a salt bath and were followed by cold-water quenching so as to retain all the magnesium in solid solution. The final cold reductions investigated were 20, 30, 40, 50, 60, and 80 pct. The initial grain sizes, that is, the grain size before the final cold reduction, were all very similar. Annealing Procedure: Specimens, 1 in. sq, were cut from the 0.036 in. sheet and annealed in a salt bath, the temperature of which was controlled to k2"C. Separate specimens were annealed for different times at 350°, 375", 400°, and in one or two cases at 415°C and after annealing were quenched in cold water. The time taken for a specimen to reach temperature was approximately 5 sec. Metal-lographic examination after annealing showed that no precipitation of Mg+413 occurred during cooling. Procedures Used for Determining Recrystallization Characteristics: All specimens were electrolyti-cally polished and etched by the method described previously1° and polishing was continued until about 0.002 in. had been removed from each side of the specimen. Using this technique, the smallest recrys-tallized grain which could be observed with cer-

Jan 1, 1956