Search Documents

By W. G. Lidman, H. J. Hamjian

' I HE fabrication of many materials from powders involves a sintering process. A mass of powder will sinter because of the excess free energy over the same mass in the densified state caused by the higher total surface area of the powder. An understanding of the kinetics and mechanism of sintering should assist in improving the properties of such materials. The present investigation conducted at the NACA Lewis laboratory deals with the sintering of chromium carbide. Dry sintering (sintering at a temperature below the melting point) was divided into two stages by Shaler:' the first stage, during which the particles preserve much of their original shape and the voids are interconnected, and the second stage, during which densification occurs and the pores are isolated. The mechanism of forming interfaces between particles, or welding together of particles, has been investigated by Kuczynski2 and may be described by any one or a combination of the following mechanisms: viscous flow, evaporation and condensation, volume diffusion, or surface diffusion. The mechanism by which pores are closed or eliminated (densification) during sintering, is of interest. Grain growth observed during sintering may be attributed to the variation in the surface energies of individual grains, causing some grains to grow at the expense of others. Grain boundary migration occurs presumably by a diffusion process, therefore the rate of grain growth would be expected to increase exponentially with increasing time and temperature. Thus, for practical sintering times of less than 1 hr, a certain minimum temperature may exist at which major structural and property changes will occur. Densification and kinetics of grain growth during sintering under pressure of chromium carbide were investigated to provide additional information which will aid in describing more accurately the sintering process and the mechanisms involved. This material was selected for this study because of the current interest in high strength, oxidation resistant refractory materials, such as carbides, which are sintered to produce solid, dense materials from powders. Sintering under pressure is a process where the heat and pressure are applied to the compact simultaneously, specimens for this work were prepared by sintering under pressure at different temperatures and for various time periods. Experimental Procedure Preparation of Specimens: Chemical analysis of the commercial chromium carbide used in this investigation was as follows: Cr 86.19 pct, C 12.14 pct, and Fe 0.2 pct. X-ray diffraction powder patterns gave characteristic diffraction lines of Cr3C2 crystal structure. Powder particle size was determined microscopically and the average initial particle size was 6 microns with 85 pct between 2 and 10 microns. Specimens sintered under pressure were formed in graphite dies3 heated by induction. Sintering temperatures were measured with an optical pyrometer by sighting into a 3/8-in. hole drilled 1 in. deep at the midsection of the graphite die. A load of approximately 1 ton per sq in. was applied to the powder. The die assembly was heated in 20 min to the highest temperature (2500°F') at which no increase in grain size could be observed, and less than 2.5 min were required to heat from this temperature to the maximum temperature (3000°F). Sintering temperatures and times for the specimens of this investigation are indicated in Table I. Analysis of Specimens: Specimens polished with diamond abrasives were etched to reveal the grain boundaries with a 1:1 mixture of 20 pct potassium hydroxide and 20 pct potassium ferricyanide heated to 160°F. Representative areas of each sample were photographed at 1000 diameters. The largest diameters of all well-defined grains were measured, but only the measurements of 15 of the largest grains were averaged in order to determine an index of grain size on the assumption that they were among the first to begin growth. Densities were determined from differential weighing of the samples in air and water. The reported density values are considered correct within ±0.01 g per milliliter. Results and Discussion Metal compacts have exhibited grain growth when sintered at temperatures about two-thirds of the absolute temperature of their melting point.' Grain growth also occurs during the sintering of chromium carbide and is illustrated by the micrographs shown in Fig. 1. These micrographs were prepared from specimens sintered for 90 min at temperatures ranging from 1371°C (2500°F) to 1648°C (3000°F). Average grain size and density measurements of specimens investigated are presented in Table I. The relationship between grain size and sintering tem-

Jan 1, 1954

By Harry W. Antes

The density changes that occurred on annealing cold-worked poly crystalline molybdenum, over the temperature range of 870" to 1630 SC, were determined. Increases in density were observed for recovery and the beginning of recrystallization. An initial decrease in density was observed during re-crystallization. A second decrease in density was observed after recrystallization was completed, as the microstructure changed from columnar grains to equiaxed gvains Both decreases in density were associated with an increase in gain boundary area to volume ratio. The second decrease in density was observed in the annealing range that immediately precedes secondary recrystallization as reported in the literature VARIOUS techniques have been used to identify the recovery, recrystallization, and grain growth processes that occur when cold-worked metals are annealed. As early as 1928 work was conducted that indicated measurable density changes occur when cold-worked single crystals of brass are annealed.' More recently, work by Murphy exhibited a corre- lation of density with microstructural changes observed on annealing cold-worked polycrystalline a brass.2 Other investigators have performed similar work on a brass and copper.324 In reviewing the literature, it was apparent that density work had been limited to fcc materials. Therefore, it was the purpose of this investigation to determine the density changes that accompany annealing the cold-worked bcc material, molybdenum. Two degrees of cold work, 30 and 90 pct reduction of cross-sectional area by rolling, were selected for this study. EXPERIMENTAL PROCEDURE High-purity vacuum arc-cast molybdenum was used in this study, the principal impurity being carbon in the amount of 80 ppm. The final rolling reductions were made at 1200°C in grooved rolls to provide 5/8-in.-diam rod. Density specimens 0.500 in. diam by 1.5 in. long were machined from the bar stock. The specimens were ground to the final dimensions to provide a smooth surface. Density measurements were made using a hydrostatic technique similar to that used by ~urphy.' This type of equipment permitted weighing in liquid immediately after the air weighing. During each weighing the temperature of the air and liquid was controlled to within 0.0l°C. This was accomplished by sealing the analytical balance in an air-tight brass chamber and surrounding it by an insulated wooden case. Preheated air was circulated in the air space between the brass chamber and wooden case by means of a blower. The temperature of the

Jan 1, 1964

By R. W. Heckel

A method is described whereby the relationship of both the "at-pressure" powder compact density and the "zero-pressure" compact density to the applied pressure may be obtained from continuous measurements of punch movement during a single compaction operation. The rate of density increase with applied pressure for the metal powders is found to be proportional to the volume fraction of pores for pressures exceeding a lower limit which varies from 15,000 to 30,000 psi, depending on the powder. 1 HE compaction of powdered materials is carried out primarily to increase the density of the material. If the ultimate goal of the over-all process is the attainment of minimum porosity, compaction is responsible for most of the densification. For example, loose powdered metals have porosities of about 65 to 75 pct. Axial loading in a die or hydrostatic pressure may effectively reduce the porosity of most powders to 20 pct or lower with pressures of 50,000 psi and greater. The relationship between the density of a powder compact and the pressure needed to achieve that density is most commonly obtained by pressing several compacts, each at a different pressure. Measurement of the densities of the individual compacts when they are removed from the die then provides the density-pressure relationship. However, this method suffers from several disadvantages. Because a large number of compacts and pressing operations are required, it is inherently slow. Furthermore, since density measurements are made after the specimens are removed from the die, data obtained refer only to pressures greater than those necessary to form a coherent compact. The third disadvantage is the inability of this method to provide information about the compact density at the applied pressure in the die. Because the determination of density-pressure relationships has been a slow, tedious process, investigators have sought to express the compaction behavior of powders by other means. The general terms "comparability" and L'compressibility" have been used to rank powders qualitatively according to their compaction characteristics. To restore a quantitative nature to the subject and to prevent confusion, Schwarzkopf1 has proposed that compac-tibility be defined as the minimum pressure needed to produce a given green strength, while compressibility should be used to indicate the extent to which the density of a powder is increased by a given pressure. Probably the most widely used of the compaction parameters is the "compression ratio,'' which is generally defined as the ratio of the compact density obtained by pressing at a given pressure to the apparent density of the loose powder.2 However, the description of the compaction behavior of a powder by these parameters provides only limited information about the process because of their applicability to just one specific condition such as a given green strength or a given pressure. The method of obtaining density-pressure relationships which will be described in the present paper was designed to eliminate the shortcomings of the previous techniques while maintaining equivalent accuracy and precision. EXPERIMENTAL TECHNIQUES The general principle employed makes use of the fact that the linear movement of the punch during a

Jan 1, 1962

By R. A. Vandermeer

It is usually assumed that the rate-determining step in the migration of a grain boundary involves the thermally activated transfer of single atoms across the interface. Chemical reaction rate theory then gives a migration rate which is proportional to the driving force when this is small compared to kT.1 In their recrystallization studies on aluminum, Gordon and vandermeer2 obtained data which was capable of providing the first experimental test of this proportionality. They correlated stored energy (driving force) release measurements determined calorimetrically with fraction recrystal-lized kinetic (growth rate) data acquired metallo-graphically at a temperature where a significant amount of recovery takes place concurrently with recrystallization. Within experimental error, the results of Gordon and Vandermeer confirmed experimentally, at least over a limited range, the linear dependence of boundary migration (growth) rate on driving force in accordance with theory. In obtaining these results, however, an incorrect, implicit assumption was made in relating the average instantaneous growth rate of the recrystallized grains to the measured fraction recrystallized kinetic data. The purpose of this present note is to point out and correct this assumption. An alternate approach, whereby a more proper average growth rate can be calculated from the recrystallization kinetic data, is presented. The earlier data, reevaluated in terms of this new approach, show the same general trend as before so that the original conclusion of Gordon and vandermeer2 is still valid. As was shown in a detailed study3 of the geometrical features of recrystallization in aluminum, the colonies of recrystallized grains can be regarded as uniform cylinders embedded in a three-dimensional matrix, the number of such colonies being constant. Thus during the early stages of recrystallization the volume fraction recrystallized, Xv, can be written as Xv=ynvlr2 [l] where y is a shape factor and i and r are defined as the average length and radius of a recrystallized colony and are spatial averages over the specimen. -nv, is the number of such colonies in a unit of volume. The erroneous assumption made in the earlier paper2 was that r in Eq. [I] could be replaced by

Jan 1, 1965

By T. P. Papazoglou, N. A. D. Parlee, W. C. Phelps

PRACTICAL designs for good "home made" molybdenum furnaces are hard to find in the literature. The one described briefly below left something to be desired but was good enough to operate as a rather long (30 in.) two-zone furnace for periods of over 2 months at temperature (1400° to 1600°C) without shutdown. The principal problem in our application was to establish a temperature profile with two distinct constant-temperature zones and no dip between the zones. This and other problems mentioned below would not arise, however, if both zones were operated at the same temperature and if a shorter furnace were acceptable. Since molybdenum windings oxidize very rapidly when heated in air, impervious refractory tubes were used for both the furnace core and the outer jacket tube shown in the accompanying figures. A protective atmosphere of forming gas (10 pct H2, 90 pct N2) was continuously passed through the gas-tight annular volume between the core and outer jacket. The water-cooled copper ends shown in Figs. 1 to 3 sealed the ends of the tubes. Annealed molybdenum wire (40 mil) was wound on the recrys-tallized alumina core (35.5 mm OD by 28.5 mm ID by 30 in.) at a spacing of 0.36 in. The spacing was reduced to 0.29 in. close to the ends to partially compensate for end heat losses. Fig. 2 shows two molybdenum-ground lead taps off the center of the winding to produce the two independect temperature zones. The ground lead was insulated with 1/4-in. alumina tubing and connected to the grounded copper end by passing the wires through the gas port tube and crimping each wire into notches cut at the mouth. To prevent sagging of the core at high temperatures, refractory "troughs" were cut from alumina and mullite tubing to cradle the core. The support "troughs" which were 6 in. long and located about 5 in. from each end of the annular volume also served as radiation shields. In view of the different expansion coefficients and operating temperatures of the core and outer jacket (3.5 in. OD by 3 in. ID by 30 in.), the ends of the tubes were smoothed and squared to produce good sealing surfaces, and the core was ground 1/32 in. shorter than the outer tube. The power terminals were passed through their respective conductor ports which were internally lined with alumina tubing and sealed with Armstrong adhesive A-6 leaving about 1 in. of protruding lead wire for external connections. (Adhesive A-6 is a two-component system consisting of an epoxy-type resin with a mineral fibrous filler and an amine catalyst.) Figs. 1 and 3 show the entire assembly of concentric refractory tubes, copper ends, and sealing o rings, held together with two brass flanges and 3/8-in. bolts with heavy springs for dimensional flexibility. The furnace shell (19 by 19 by 30 in.) consisted of a steel frame with sheets of transite forming a box. The copper ends protruded through two 3.75-in. circular holes cut at the center of each transite square end. Thermal insulation next to the transite exterior was effected by K-23 B&W bricks alld in a central volume of approximately 9 by 9 by 30 in. a row of K-30 bricks was lined up to support the hot assembly. The rest of the volume around the outer

Jan 1, 1965

By L. J. Buteau, R. E. Reed-Hill, W. A. Slippy

In a zirconium {1121} twins tend to have coherelzt boundaries and thus do not usually taper to a point when they intersect a grain boundary. Under the proper conditions, a (1121) twin may deform a boundary in such a manner as to clearly rcveal the twinning shear. Measurements of these grain-bounrlary rotatzons yield a value of 0.63 for the (1151) twinning shear. The magnitude of the observed shear and its shear sense are in excellent agreement with the assumption that the second undis-torted plme (K2) for this form of twinning is the basal plane (0002). The present value of the tzvinninz shear, which is approximately three limes larger than Rapperport&apos;s value? is in excellent agreement with reported results of Sokuskii and Prolsenko.5 RECENTLY westlake&apos; reported a study of the twinning modes of zirconium based on ~ap~er~ort&apos;s2 values for the twinning elements. He observed that two of the modes, (1012) and {1122), can be explained by relatively simple dislocation models. A workable model for (1121) twinning, however, proved to be very complex and questionable. Because Rapperport&apos;s second undistorted plane, (K2), for (1121) twinning has large indices, {ll2l6), it is possible that the difficulty in defining a model for (1121) twinning lies in an incorrect value for K2. A measurement of this twinning element has therefore been made, using a method based on observations of the shear displacement of grain boundaries at positions where they are intersected by (1121) twins. EXPERIMENTAL PROCEDURES The zirconium specimens were deformed in tension on an Instron machine either at the temperature of liquid nitrogen (77ºK), or prestrained at liquid nitrogen temperature and then further deformed at room temperature. Straining at 77°K was employed in order to nucleate (1 121) twins in large numbers, since simple ropm-temperature tensile deformation favors (1012) twinning. Specimens were machined from 1/2-in.-thick plate. The material was arc-melted sponge zirconium annealed to an ASTM grain size 4 and containing, in parts per million, 380 Fe, 880 0, <40 Hf, 32 N, and < 5 H. The specimen gage length and diameter were 1.25 and 0.130 in., respectively. To insure a gage section initially free of twins, the last 60 thousandths of the specimen diameter was removed by acid machining.

Jan 1, 1963

By M. E. Shank, J. Wulff

At the present time, the designer of dies for metal powder pressing is handicapped by relative ignorance of stress distribution and frictional effects at the interior surface of the die. Unckell was the first to develop a method for the study of wall friction. He used three Brinell balls on which the die rested during pressing. The total frictional wall force was determined by the size of impression these balls left on a soft metal plate. Since the method does not give radial pressures, or distribution of such pressures, coefficients of friction could not be determined. Although Unckel measured density distribution, he was not able to determine radial or shear stresses. Shaler2 has proposed theoretical expressions for the stress and density distribution within cylindrical compacts during pressing, in accordance with the experimental results of Kamm, Steinberg, and Wulff.3 By application of Siebel&apos;s method,4 Kamm et a13 plotted stress trajectories for two compacts. From the stress trajectories they calculated coefficients of friction from point to point along the die wall. As pointed out by Shaler in the discussion of Ref. 3, these values are based on progressive point-to-point calculations on finite size grid squares across the compact. In the region of the die wall such calculated values may therefore have considerable cumulative error. The purpose of the present paper is to develop an experimental method by which the nonhydrostatic pressures and shears acting on the interior wall of a cylindrical die can be measured. Such measurements can then he correlated with existing data to aid in the explanation of the pressing process. The method used is based on the elastic: properties of the thick-walled tube used as the die. The principle of super-position of force systems on an elastic body is assumed to hold. Electric strain gauges were mounted in adjacent positions on the exterior die wall in order to get an exact measurement of the variation of tangential strain over the length of the die during pressing. While in this paper, measurements in terms of only tangential strains are considered, it is well to note that similar calculations may be set up for axial strains. The latter are not preferred, since they tend to be smaller than the tangential strains and therefore permit less sensitive measurements. Discussion in this work is restricted to compacts pressed from both ends, since the elastic deformation of the die is then more amenable to analysis. Before choosing the electric strain gauge method, a more direct line of attack was considered and discarded. The discarded idea was the insertion of a pressure gauge through a hole in the die wall.* The gauge would have been in the form of a small piston. If pressure were exerted against such a gauge, it would move outward along a radius of the die. One disadvantage of the scheme is its inability to measure shears along the die wall. Another more serious disadvantage is the disturbance caused by the device itself. It would serve to change the forces it was designed to measure. No matter how small the movement of the gauge, when pressure is applied a discontinuity would exist in the wall surface at that point. Due to the stress concentration caused by the hole, abnormal deflections of the die wall would occur around the gauge. During pressing, powder would be forced into the resulting depression. The depression would then become larger with increasing compacting pressure. Powder, not being a fluid, is capable of supporting shear. The ease with which it would flow into the die wall depression to further move the piston is an indication, not of the radial pressure at that point, but of the state of shear retarding the movement. Thus the "pressure" gauge is really a criterion of flowability, and of the capability of the powder to support shear. For these reasons, it was decided that the electric strain method, herein employed, was more reliable, if more indirect. The gauges and lead wires, mounted on the external die wall do not in any way affect the behavior of the metal powder or the die during pressing. Theory of the Method THE EFFECT OF RADIAL PRESSURE ON THE DIE WALL Effect of a Single Small Band of Hydrostatic Pressure Consider a die which is a thick-walled cylinder of outer radius R. and inner radius Ri. If over a small finite length L there is a normal pressure P, a tangential strain distribution at the outer wall results. This is shown schematically in Fig 1. The exact shape of the curve may he predicted by an extension of the theory of a semi-infinite beam on an elastic foundation.6 This

Jan 1, 1950

By Dale A. Vaughan, Oliver M. Stewart, Charles M. Schwartz

Solid-solubility limits at 1500°, l000q and 500°C for carbon, nitrogen, and oxygen in high-purity tantalum were determined by X-ray lattice-parameter methods. For carbon, the solubility was found to be 0.17 at. pct at 1500°C, and less than 0.07 at. pct at 1000°C. A nitrogen solubility of 3.70 at. pct at 1500° C decreases linearly with temperature to 2.75 at. pct at 1000°C, and 1.8 at. pct at 500°C. In the case of oxygen, the solubility was found to be 3.65 at. pct at 1500°C, 2.95 at. pct at 1000°C, and 2.5 at. pct at 500°C. The phases Ta2C, the low-temperature modification of Ta205, and Ta,N of unknown composition but which has a superlattice structure based upon the original bcc tantalum lattice have been identified as the initial precipitates in the respective systems. Metallographic methods were employed to verify the X-ray analyses. The etching behavior of Ta is discussed in terms of lattice i?rzperfections and precipitate phases. The excellent fabricability, high melting point, and nuclear properties of tantalum are responsible for interest in this refractory metal. Data on the solid solubility of the interstitial elements (oxygen, nitrogen, and carbon) in tantalum and on the precipitate phases are somewhat limited. The significant contributions are discussed below. Because the purity of electron-beam melted tantalum (only recently available) is considerably higher than that used in previous studies, the present investigation was initiated. Gebhardt et all-3 have investigated the tantalum-oxygen and tantalum-nitrogen systems with particular reference to the changes in physical properties and to the rates of reaction between these gases and the metal. The solubility of oxygen in tantalum was reported2 to be 3.7 at. pet at 1500°, 2.3 at. pet at 100O°C, and 1.4 at. pct at 750°C. Schonberg4 reported that several oxide phases (Ta40, Ta,O, TaO and Ta,05) exist while X-ray studies by Gebhardtl showed only two oxides, Ta,05 and an unidentified phase which was associated with a platelet-type precipitate. La-gergren and Magneli,&apos; however, questioned the existence of compounds other than the two allotropic modifications of Ta,05 for the tantalum-oxygen system. In the case of nitrogen, the solubility was estab- lished by Gebhardt3 to be of the order of 7 at. pct at 1800°c. The solubility was reported to decrease rapidly with temperature, and, although no limits were established, a precipitate phase was observed by Gebhardt except when the high-nitrogen specimens were cooled very rapidly from the reaction temperature of 1800c. He reported the initial precipitate phase to be a tetragonal distortion of the bcc tantalum lattice while Schonberg6 reported the phase lowest in nitrogen Jo be a cubic super-lattice with a cell size of 10.11 A. Two other nitride phases, Ta,N and TaN, were reported; these appear to be isomor-phous with the carbides of tantalum. The tantalum-carbon system was investigated by Ellinger7 and by Lesser and Braurer. Two compounds, Ta,C and TaC, were reported to exist, each with a range of composition. The solubility of carbon in tantalum was found to be practically nil at all temperatures. Thus, of the interstitial elements which are present in small amounts in high-purity tantalum, carbon might be expected to form precipitates. The present investigation was initiated to obtain additional data on the solid-solubility limits of these interstitials at 1500°, 1000°, and 500&apos; with particular emphasis on the distribution and the identification of the precipitate phases. EXPERIMENTAL WORK AND RESULTS In the present investigations of the solid solubility and of the precipitate phases in the systems tantalum-nitrogen, -oxygen, and -carbon, high-purity tantalum was reacted with high-purity gases, homogenized at 1800°c, and annealed at and quenched from 1500°, 1000,

Jan 1, 1962

By A. S. Yue

It was deemed desirable to obtain an understanding of the vacuum desulfurization process. McKechnie1 has reported that the sulfur content of nickel- and cobalt-base alloys is reduced in vacuo. Keverian and Taylor2 have observed that vacuum melting causes desulfurization of iron-carbon-silicon alloys. Garnyk and samarin3 have desulfurized transformer steel by vacuum treatment. Experiments were conducted in order to investigate the desulfurization of liquid iron under reduced pressures, and to establish the effect of some alloying elements, and various crucible materials. A vacuum-induction furnace of 50-lb capacity was used. The charge was Armco iron. Silicon was added as 90 pct ferrosilicon, carbon as electrode graphite, and aluminum as high-purity aluminum metal. Magnesia and silica crucibles were used, their inside diameters were 5 1/2 and 5 3/8 in., respectively. The temperature was measured with an optical pyrometer and it was held at 1600 1 15°C. Samples were withdrawn with quartz tubes of 6 mm ID, without disturbing the furnace atmosphere. Pressures were measured with an alphatron gage. The experimental results are shown in Fig. 1. They indicate that: 1) Sulfur can be removed from liquid iron alloys by vacuum treatment; 2) The desulfurization is enhanced by the presence of silicon, carbon, and aluminum, all these elements increase the activity of sulfur dissolved in liquid iron;4 3) The pronounced effect of silicon indicates that probably a volatile silicon sulfide is formed, Alcock and Richardson5 have observed a similar phenomenon; 4) The same rate of desulfurization was obtained when the metal was contained in either magnesia or silica crucibles. Dr. J. Keverian has contributed many discussions and recommendations. M. S. Sutliff has helped in conducting the heats. The permission of General Electric Co. for this publication is gratefully acknowledged.

Jan 1, 1960

By Barbara A. Thompson

A previously reported high-resolution method for the location of boron-rich areas in metallurgical and biological specimens was been adapted for general use on a routine basis. The rnetlzod utilizes neutron activation and autoradiograpizy. Alpha-particles emitted by boron nuclei upon neutron capture are recorded on a photographic emulsion. The resulting a-particle tracks show the location of boron-rich areas. Experimental techniques, interferences, and limitations of the method are discussed in detail. The method is most useful where there is marked segregation of boron. In this type of sample, the segregation can be observed when the nominal boron concentration is as low as 0.0006 pct. THE positive identification and location of boron-rich areas in metals is frequently of great interest in metallurgical work. Unequivocal identification is often difficult to make by conventional metallo-graphic methods. Recently, a method has been described which accomplishes this objective by neutron activation and autoradiography.l-3 The method can be described briefly as follows. Upon neutron capture, a -particles are emitted by boron nuclei according to the following reaction: ,Blo + n - ,a4 + 3Li7 + 2.4 mev The energy is dissipated as kinetic energy of the products. By irradiating a boron-containing sample in contact with a photographic emulsion and subsequently developing this emulsion, a-particle tracks are obtained whose location corresponds to the location of boron-rich areas in the sample. Two factors combine to make the reaction extremely specific for boron. The first is the unusually high (755 barns) cross section of boron for thermal neutron capture. The second is the higher neutron energy required to produce (n, a ) reactions in essentially all other nuclei except lithium. These two factors make the method specific for boron by six to seven orders of magnitude when a predominantly thermal neutron source such as the Brookhaven reactor is used. The reported limit of detection of this method is of the order of 0.01 pct B., The present work was originally undertaken to determine whether this limit could be lowered by use of a thinner emulsion. However, initial experiments showed that in order to use the method at all, it was necessary to reestablish the optimum experimental conditions in terms of the available irradiation facilities. It is the purpose of this paper to describe these experimental conditions in detail, to discuss the factors influencing sensitivity, and to evaluate several techniques for increasing sensitivity. EXPERIMENTAL A) Preliminary Experiments—The first measure-ments were made using samples of crystal oriented silicon steel containing various concentrations of boron. In the later experiments, samples of various high-temperature alloys such as M-252, hcoloy 901, Nichrome V, and so forth, were used. Faraggi, et al.,2 reported that the lower limit of sensitivity in this type of sample was about 0.01 pct B using nuclear emulsions of 50- u thickness. but that it should be possible to extend this limit by the use of thinner emulsions. Accordingly, we first used Kodak Auto-radiographic Stripping Film (Permeable Base) which has an emulsion thickness of only 5 µ. This was mounted on the metallographic specimens according to the technique described by Boyd.4 The emulsion remained in contact with the metal surface throughout exposure and development. Since the emulsion is transparent after development, the autoradiograph and metal surface can be viewed simultaneously and any correlation between film blackening and structure of the metal can be made directly with no problems of realignment. Because the silicon steel is readily attacked by moisture alone, it was necessary to apply a protective coating to the metal surfaces before mounting the emulsion. The coating was made extremely thin in order to absorb as few a-particles as possible. Boyd4 and Gomberg5 have discussed various plastics used for this purpose; however, none was sufficiently impermeable to prevent chemical attack of the steel during the developing process. This attack resulted in the production of gross chemical artifacts in the emulsion. It was, therefore, necessary to use the method of Wolfsberg and John6 as follows. A very thin (approximately 1 µ) coating of Plexiglas II was applied by dipping the sample in a 2 pct solution of Plexiglas II in dichloroethylene. Then, because the emulsion will not adhere to Plexiglas 11, a thin coating of Parlodion was applied in a similar manner using 2 pct Parlodion in iso-amyl acetate. No protective coating was necessary with the high-tem-

Jan 1, 1961

By Frances H. Clark

IN wartime, the fabrication and use of metals assumes increased importance, for a modern war of sizable proportions cannot be undertaken with- out a vast supply of this material. Light alloys of aluminum and magnesium, which were comparatively unknown in the last war, will play an important role in this one. Due to research on a vast scale in the interim of years, industry is today well equipped to fabricate metals on an enormous scale. Research for the past year has been most productive and a comprehensive survey in the space of this article would be impossible. It is hoped that the stressing of a few items here and there out of a great quantity of able work will indicate in a small way the advances made within recent times. Specific references to the literature will be omitted at the Editor&apos;s re- quest but some authors of papers that I have found of special interest are mentioned.

Jan 1, 1940

By P. E. Busby, C. Wells, M. E. Warga

Fundamental data on the rate of diffusion of boron in austenite and solubility of boron in the a and y phases of iron and steel have been obtained from deboronizing experiments and provide partial explanations for some of the phenomena observed in boron steels. The rate of diffusion of boron is about the same as carbon in austenite. The solubility of boron in austenite at normal heat-treating temperatures is less than 0.001 pct. A partial tentative Fe-B phase diagram in the important low boron concentration ranges and an equation representing the diffusion of boron in austenite are presented. A LTHOUGH a search of the literature revealed ALTHOUGHthat boron diffusion had not been studied quantitatively and systematically prior to 1948, a few qualitative observations which were made by following the diffusion of boron into iron packed in ferro-boron1-3 suggest that boron obeys normal diffusion laws and that the rate of diffusion increases with increasing temperature. The claim of Cornelius and Bollenrath1 that boron does not appear to influence the rate of carbon diffusion has recently been substantiated by Wells, Batz, and Mehl.4 Calculations based on data from the paper by Campbell and Fayv indicated that the diffusion coefficient (D) for boron in iron at 900°C is approximately 3x10 sq cm per sec; a value of 2x10-7 sq cm per sec at 1038°C has been reported by Digges et al in connection with decarburizing experiments on a commercial boron steel containing 0.43 pct C. During the period 1948 to 1950 several diffusivity constant (D) values for the diffusion of boron in the y phase of iron and steel and a number of solubility values for boron in both a and y phases were reported. It was concluded tentatively as a result of Metals Research Laboratory studies that D values for boron in y and a phases are about the same as for carbon in the comparable phasesa and that carbon up to 0.4 pct did not affect the rate of diffusion of boron in y iron above 1000°C within the limits of experimental error. Whether saturation of the y or a phases with carbon would significantly affect rates of diffusion of boron through them was then and still remains in doubt. The best estimate of Q (activation energy) was reported to be about 25,000 cal, somewhat lower than that for carbon in y iron. The solubility of boron in y iron at 1000°C was thought to be higher than 0.004 pct B but not as high as a more recent analysis of available data shows it to be. The solubility of boron in a iron at 700°C was reported to be 0.0004 pct B, but it would not have surprised the authors if the true solubility value is actually lower than this. Breaks in diffusion curves which were not understood when first obtained prior to 1950 have now been recognized as indicating solubility limits, and these are included in the present paper. Experimental Procedure Two methods were employed in an effort to determine diffusion coefficients (D) for boron in austenite. Early experiments involved the use of welded couples prepared in accordance with the procedure of Wells and Mehl,&apos; but when the results of these experiments proved to be unsatisfactory for calculating D values, studies were continued by means of deboronizing experiments. In both cases, boron analyses following the diffusion anneal were carried out spectrographically and all D values were computed using the Grube solution of Fick&apos;s law. The precision of the spectrographic method used proved to be about equal to that reported by Corliss and Scribner" —average deviation from the mean about 5 pct of the amount present when the boron content is 0.003 pct. Concentration-penetration curves obtained from spectrographic analyses of welded couples generally indicated that practically no boron was transported across the weld interface during the diffusion anneal. Furthermore the concentration-penetration

Jan 1, 1954

By Frank W. Clinard, Oleg D. Sherby

The effect of multiple an transformations on diffusion in a dilute iron alloy was studied. Inter-penetration between iron and an Fe-Co alloy was evaluated, under thermal-cycling conditions chosen so as to transform only the iron-rich side of the diffusion couple. It is shown that a small but consistent diffusion enhancement by a factor of two results when data from the transforming side of thermally cycled specimens are compared with predicted inter diffusion coefficients based on isothermal data. No enhancement is found when data from the mntransforming side of such couples are compared with predicted diffusivities. It is suggested that short-circuiting by moving pain boundaries is the most likely mechanism contributing to the observed diffusion enhancement from multiple phase transJormations in iron. Transformation-induced Point defects, dislocation-pifie short-circuiting, and point defects generated by dislocation motion are less effective in contributing to enhanced diffusion under the experimental conditions utilized. SO far as is known, no studies have been made of the effect of allotropic transformation on diffusion. previous to this investigation. Studies have been made, however, of the influence of transformation on the diffusion-controlled processes of spheroidization and sintering. Cullen1 has described a method of accelerating spheroidization in an AISI 1030 steel. Partial transformation between a and ? phases was uti-

Jan 1, 1965

By Carl F. Lutz, Robert F. Mehl

A layer of brass was formed on 0 brass using a vapor-solid reaction technique. The variation in composition with distance within the phase layer and across the a -ß interface was determined by an electron probe. The diffusivities were calculated as a function of concentration; the diffusion coefficient in brass was found to change by a factor of twenty-five over a compositional range of 8 at. pct. An explanation is proposed to account for the large change in the diffusivity, based on possible stresses in the diffusion zone; it is suggested that the concentration dependence of the thermodynamic factor might decrease the activation energy with increasing zinc content. ALTHOUGH the body of data on diffusion coefficients (D- values) in terminal solid solutions and in isomorphous systems is quite large, there is but little quantitative information for intermediate alloy phases. This paper recounts measurements of D-values for the ? phase in the Cu-Zn system. The experimental method employed consists in the forming of alloy layers and determining and analyzing the concentration-distance (c-x) curve. The experimental method employed could have been applied to several systems; the Cu-Zn system was chosen because the D-values in the a and ß phases are well known, and can thus be used for purposes of comparison. The rates of growth, of intermediate alloy layers are known for a number of systems. In nearly all cases the thickness varies parabolically with time, as to be expected if the rate of growth is controlled by the D-values. In a few cases, nonparabolic behavior has been noted.1 Nonparabolic growth may be the result of a) a variation of the composition at the phase interface with time, b) interface-controlled kinetics, c) short diffusion times coupled with long vacancy lifetimes (in vacancy diffusion),2 and d) crystal reorientation during diffusion in anisotropic systems.3 In the Al-Ni system4 and in the Al-u5 system parabolic growth kinetics are slowly approached after an initial transient period. In general those phases stable at a given temperature in a system A-B appear as separate phase layers when A and B are put in contact and given a diffusion he at-treatment. In most cases the compositions at the interface of adjacent phase layers are those read from the phase diagrams for the termini of the respective phases. This discontinuity in concentration is taken as independent of time, and the growth of one phase at the expense of another is assumed to be independent of interface kinetics, i.e., the rate of interface movement is controlled by volume diffusion in the phases concerned. wagner6 has given many solutions for multi-phase diffusion processes, assuming that the chemical diffusion coefficient, D, is independent of concentration, as have others. Jost7 has pointed out that the familiar Boltzmann-Matano solution is as valid for a (c-x) curve exhibiting concentration discontinuities at phase boundaries as for the usual single-phase case. Heumann and associates8,9 have applied this solution to the multi-phase case, but lacking concentration data within the several layers were forced to assume linear concentration gradients, thus calculating only average D-values. The purposes of the present study were: 1) to determine the dependence of D on concentration, 2) to calculate the intrinsic diffusion coefficients D?Cu and D?Zn, 3) to establish the time-law for the movement of the interface, 4) to determine the concentration limits at the ß-? interface, and, 5) to attempt to clarify the mechanism of diffusion. EXPERIMENTAL PROCEDURE The ? phase is exceedingly brittle; conventional solid couples and conventional sectioning methods were thus inapplicable. Vapor-solid couples were accordingly employed. Unfortunately the equilibrium vapor pressures of Zn for the ? phase are unknown; to escape this awkwardness, samples were enclosed in a capsule containing powder of an alloy of known composition, sufficiently large in quantity as to be effectively an infinite source, maintaining the concentration of Zn at the sample surface constant with time. The sources of Zn-vapor which can be employed in reaction with Cu, or a brass, or ß brass, to form a layer of ? brass satisfactorily wide in concentration range, are alloys of the phases ?, or ? + E or ? + d (depending on the temperature).

Jan 1, 1962

By J. D. Braun, C. W. Powell

Diffusion between gold and indium at two temperatures below the melting point of indium (155.4oC) was investigated. The major, component of the diffusion zone was determined to be AuIn2; the other, intermediate phases in the system occur in much narrower layers In order of decreasing layer width these are: Au9In4, AuIn, Au4In. Inter diffusion coefficients of the four intermediate phases and the terminal solid solutions lie in the range of 2.4 x 10-l3 to 1.8 x 10-11 sq cm per sec. WHEN two metals which do not form a continuous series of solid solutions interdiffuse, each phase in equilibrium at the diffusion temperature should appear as a separate layer in the diffusion zone. In the Au-In system, diffusion plays a significant part in many commercial applications, particularly where indium alloys are used as solders for electrical contacts with very fine gold wire, for estab- lishing electrical contact to thin metallic films, and for mounting metal-coated piezoelectric crystals. A knowledge of the number, extent, and type of phases which form at diffusion temperatures below the melting point of indium, 155.4oC,1 and the diffusion constants for the system would be useful for these practical applications. Of more basic significance, additional information on polyphase diffusion in a system not previously investigated would be gained. The phase relationships in the Au-In system were determined primarily by Kubaschewski and Weibke1 and are based on thermal, microscopic, and X-ray analysis, Fig. 1. There are five intermediate phases in the system having approximate compositions corresponding to the formulas: Au4ln (ß)—80.0 at. pet Au; Au3In (?)— 75.0 at. pet Au; Au9In4 (d)—69.2 at. pet Au; AuIn—50.0 at. pet Au; AuIn2—33.3 at. pet Au. Solid solubility of indium in gold at 700°C is reported as 12.64 at. pet and at 406°C as 10.36 at. pet. No solid solubility of gold in indium is reported.

Jan 1, 1964

By R. E. Ogilvie, N. L. Peterson

Diffi-lsion measurements were conducted at all compositims in the bcc solid solution of the U-Nb system employing incremental couples at composition intemals of 10 at. pct. Diffusion coefficients were determined by the Matano method from concentration gradients obtained with the electron-probe microanalyzer. The activation energy for inter-diffi-lsion as a function of compositim shows three distinct regions: 1) 80 to 100 pct U.6= 30 kcal per mole; 2) 20 to 80 pct U, $ - 70 kcal per mole; 3) Oto 20 pet U, Q = i40 kcal per mole. The frequency factor, fi0 and the activation energy $ were found to be roughly related by the following equation: log Do ^9.7 X IO-5Q -6,6. The Kirkendall marker movement indicates that DU is larger than DNb between 16 and 100 pct U and DNb is larger than DU from 0 to 4 pct U. FOR practical as well as fundamental reasons, the rates of diffusion in alloys are of considerable consequence. Most solid-state reactions are largely dependent upon the diffusion of atoms through the lattice structure and along grain boundaries. The high-temperature strength and reasonable nuclear properties of niobium have prompted its use as a reactor material in contact with uranium fuel. Hence, diffusion data for the U-Nb system are of considerable importance. In the previous diffusion study1 on the U-Nb system using pure element couples, reliable data were obtained only in the range of 0 to 10 at. pct Nb due to the large variance of the diffusion coefficient with composition. Also, a large Kirkendall effect and considerable porosity in the uranium-rich areas of the specimen were reported, which suggests that the true diffusion coefficients are somewhat larger. The purpose of the present study was to obtain reliable diffusion coefficients at all compositions using incremental diffusion couples with intervals of 10 at. pct. In view of the abnormal self-diffusion be- havior of y uranium2-4 and some other bcc transition elements,&apos;-&apos; it was felt that a comparison of the interdiffusion coefficients in the bcc U-Nb system with those of Reynolds et al.9 for the fcc gold-nickel system might shed some light on the diffusion mechanism involved. Both systems have similar phase diagrams, in that complete solid solubility exists above a miscibility gap. EXPERIMENTAL PROCEDURE The uranium used in this investigation was obtained through the courtesy of Argonne National Laboratory. An analysis of this material detected only Si-30, A1-7, C-6, N < 10 and 0-18 ppm. The niobium was electron-beam melted material obtained from Stauffer-Temescal. The gaseous impurities were less than 50 ppm, and the spec troc hemical analysis showed Ta-500 and W-200 ppm. U-Nb alloys were prepared at composition intervals of 10 at. pct by melting the appropriate amounts of the pure elements in an arc furnace. The buttons were inverted and remelted 6 times to assure complete mixing. The buttons were then wrapped in molybdenum foil, canned in Zircaloy-2 or stainless steel, and hot rolled 30 pct reduction in thickness at temperatures between 850" and 1100°C. Alloys with 10, 20, 30, 40, and 90 at. pct Nb rolled quite easily under these conditions, but the 50, 60, 70, and 80 pct alloys remained brittle. After melting and rolling (when possible), the alloys were annealed for 24 hr at a temperature within 100°C of their melting point in a dynamic vacuum of better than 4 x 10-8 mm Hg. These treatments produced alloys which were homogeneous on a 1 p scale within the detectability limits of the electron probe. During fabrication, the alloys picked up as much as 100 ppm Mo and 100 ppm Zr. Other elements checked for but not found were Co, Cr, Fe, Mn, Ni, and Ti. The grain size of the annealed samples ranged from 3 mm for the uranium-rich alloys to 0.3 mm for the niobium-rich alloys. This permitted measurements of the concentration gradients in the diffusion samples without crossing more than one or two grains, thereby eliminating any grain boundary effects. The specimens were bonded by theU&apos;picture frame" technique as reported by Kittel.10 Specimens of composition b)U + (100 - x)Nb were sandwiched between two specimens of composition (x + 10)U + (90 - x)Nb after they were ground flat and parallel

Jan 1, 1963

By George Mah, Charles Wert

Internal-friction peaks caused by diffusion of oxygen and nitrogen have been observed in ytterbium. They are thought to be caused by the redistribution, under stress, of strain dipoles around an interstitial oxygen or nitrogen and a neighboring substitutional impurity. With the use of a simple geometrical model, diffusion coefficients for oxygen and nitrogen between special interstitial sites have been calculated. a) APPLICATION OF INTERNAL FRICTION TO MEASUREMENT OF DIFFUSION RATES IN DISSOLVED INTERSTITIAL ATOMS In bcc metals, the interstitial atoms occupy the centers of cube faces (0, 1/2, 1/2), or their equivalents, the center of the cube edges (0, 0, 1/2). An interstitial atom in this position causes an aniso-tropic distortion in the lattice. When stress is applied, mechanical relaxation can take place by the movements of the interstitial atoms. This relaxation has been used to obtain information concerning the movements of the interstitial atoms in the bcc metals. Data for these alloys have been compiled by Stanley and Wert. Other researchers have extended the investigation of interstitial diffusion to the close-packed metals. Gupta and weinig2 found an oxygen peak in a titanium, the height of which is linearly dependent on the oxygen concentration. They assumed, therefore, that they were dealing with movement of atomic ally dispersed oxygen in titanium. KG and wang3 determined that the diffusion coefficients of carbon in fcc steels and in nickel obtained by internal-friction methods were consistent with those obtained by macrodiffusion methods at much higher temperatures. Bisogni found oxygen and nitrogen peaks in hafnium by use of which he deduced the diffusion coefficient of these gases in hafnium. He observed further that the half-widths of these peaks corresponded almost exactly with those of a single-relaxation process; he assumed it to be caused by movement of atomically dispersed oxygen and nitrogen. From these studies, the deduction can be drawn that, in general, diffusion data can also be obtained for the interstitial diffusion in close-packed metals (or more precisely, in slightly impure close-packed metals). One group of such metals for which little data exist for the diffusion of dissolved oxygen and nitrogen is the rare earths. They have great affinity for gases, which enter solution interstitially. In this investigation results were obtained for the Yb-O and Yb-N interstitial solid solutions with traces of substitutional impurities; these solid solutions are of fcc structure. b) MECHANISM OF THE RELAXATION IN FCC CRYSTALS In the fcc structure, an interstitial gas atom is usually situated in the octahedral position (1/2, 1/2,

Jan 1, 1964

By K. G. Davis, P. Fryzuk

The diffusivity of silver in liquid tin has been determined, using the capillavy-reservoir technique, over the temperature range 250° to 500°C. The new value, D = 2.5 x 10&apos;* exp(-2480/ RT) sq cm per sec, differs from that obtained by other workers in an earlier investigation. The analysis of data from the capillary-reservoir technique is discussed. In a recent investigation of the solidification of dilute alloys, values for the diffusion constant of silver in liquid tin were required in the analysis of the formation of impurity substructures. AS a result, measurements were made of the diffusion constants in the temperature range 250° to 500°C (melting point of tin 232°C), for the alloy concentrations used in the solidification experiments and at higher concentrations, to verify previous determinations.I)2 The capillary-reservoir method was adopted, using experimental procedures similar to those followed by Ma and Swalin,1 with the main exception that radioactive silver was used in the present investigation to facilitate solute-concentration measurements. EXPERIMENTAL PROCEDURE a) 100 ppm Samples. Glass capillary tubes of 2 mm inside diameter and approximately 5 cm- long were sealed at one end, evacuated, and filled with tin of 99.999 pct purity. The region of shrinkage near the mouth was cut off, and the tubes were then placed in a graphite holder and immersed, with the open end up, in an unstirred bath of alloy containing 100 ppm Ag 110, where they remained for periods of up to 30 hr. On removal from the bath they were cooled by an air blower. The bath was kept under a small positive pressure of argon, and the temperature controlled to within +1°C. A 10-hr diffusion period was used in the majority of the tests, scatter on runs of less than 5 hr being rather large. The procedure outlined above was chosen in preference to putting alloy in the capillary and pure tin in the bath, in order to avoid segregation when the tubes filled with alloy were first solidified. To minimize segregation when the diffusion period was complete and the capillaries again solidified, the earlier samples were held in thin-walled silica tubes which could be cooled very rapidly. Later tests were made in precision-bore Pyrex tubes, to eliminate effects caused by variations in the capillary diameter. No consistent differences in diffusivity as measured in the two types of tube were detected. After removal from the glass tubing, the samples were sectioned into 2.5 mm lengths and counted for y activity, using a scintillation counter with fixed geometry. Samples were also drawn directly from the bath and counted, so that values for C/C,, the ratio of the weight of Ag 110 in the sample to that in the bath, could be obtained. b) 5000 ppm Alloy. To check for possible effects of concentration, the silver content of the bath was increased to 5000 ppm. Complete mixing was found to have taken place in the capillary after a 10-hr period at 300°C. It appears that the greater density of the alloy was sufficient for buoyancy forces to cause instability in the alloy-tin interface, leading to rapid convective mixing. For the 5000 ppm alloy, therefore, the bath was of pure tin and the capillary tube was filled with alloy. With this arrangement, values of D consistent with those for the 100 ppm alloy were obtained, Fig. 1. CALCULATIONS OF DIFFUSIVITY The terminology used applies to a capillary of pure tin immersed in a bath of alloy. 1) Error-Function Method. under the present experimental conditions, the rod of liquid tin into which silver is penetrating may be considered semi-infinite. Assuming the concentration at the mouth of the tube to remain constant at Co, the concentration C at distance x from the mouth of the tube at time / is given by3 Plots of the inverse error function of (1 -C/Co) vs .v gave straight lines passing through the origin with slope 1/2-, x being corrected for shrinkage both on solidification and while cooling to the melting point (total correction about 6 pct at 500°C). Values for log D obtained in this manner are shown in Fig. 1. A least-squares fit to the relation D = Do exp(-Q/RT)

Jan 1, 1965

By F. E. Jaumot, A. Sawatzky

IT as long been recognized that there is a need for accurate systematic data on diffusion in solids. Originally, such data were needed to check existing theories but, more recently, it would appear that it could be used most advantageously as a guide to, or a check for, much needed extensions of the theory of diffusion. The awareness of the conditions which are important to experimental accuracy, the availability of the materials for accurate experiments, and the techniques have improved to the extent that it would now seem to an experimentalist that, unless extensions of the theory are made which take into account the fundamental parameters of the atoms, the electrons, and the lattice in some detail, the field of diffusion will rapidly become an empirical or semi-empirical field. The present series of experiments was designed to provide accurate data on the diffusion of the elements of the IB and IIB subgroups in silver; that is, to provide data on the diffusion in silver of three face-centered-cubic elements with roughly similar chemical properties, but with different atomic or ionic sizes and, similarly, for three noncubic elements. The elements of these two subgroups have different valences, so that the data will not give any significant information on differences due to the crystal structure of the solute. The present paper reports data on the diffusion of copper and mercury in single crystals of silver; the data for the other elements of the IB and IIB subgroups were reported previously by the authors1&apos;2 and by Tomizuka and Slifkin.8 Experiment The general technique used was identical to that previously described1,2 and will not be repeated here. In all cases, the sectioning technique and high specific activity isotopes" were used. The short half-life of Cu64 made decay corrections more important than for the other isotopes, and necessitated around the clock shifts while data were being taken. Each section was counted at least twice (as was true for all longer half-life isotopes) and a monitor was counted along with each sample. Data for the diffusion of mercury in silver are presented at a greater number of temperatures than for the other solutes. This is a result of two circumstances. First, six (of 30) samples had to be discarded because of pitting or recrystallization which occurred during the diffusion runs, even though the samples had sustained high temperature prediffusion anneals. Either of these was a very rare phenomenon in the other experiments, and their occurrence in this case is considered in more detail below. Second, the curves of the logarithm of the activity of the sections vs the square of the penetration deptht were, in general, quite good, but frequently seemed to show one point to have high activity and the next point to have slightly low activity in terms of the best straight line. When this occurred, there was remarkable regularity of the high-low sequence. In any event, the results, as far as diffusion coefficients are concerned, were not affected by this amalgam effect. In fact, the mercury data had the highest internal consistency of any of the data. Results Diffusion of Copper in Silver—-Diffusion runs were made at eight temperatures In the temperature range from 700° to 950°C, for times ranging from 5 to 45 hr. A linear curve of the logarithm or the activity vs the square of the penetration depth was obtained for all samples. Table I summarizes the results for individual samples and Fig. 1 gives a plot 02 the logarithm of the diffusion coefficients as

Jan 1, 1958

By R. W. Balluffi, R. Resnick

NUMEROUS investigations of chemical diffusion in a brass have been made and the results are collected in several places.1-3 This work has been mainly concerned with the determination of the chemical diffusivity as a function of composition and temperature. In 1947 Smigelskas and Kirken-dall&apos; showed that zinc and copper diffuse at different rates in face-centered-cubic brass, and since then, a number of efforts have been made to determine the intrinsic diffusivities of zinc and copper in this alloy.1, 5-9 Horne and Mehl8 in particular have recently determined the intrinsic diffusivities as functions of temperature and composition using sandwich-type couples and inert markers. Inman et al." also have determined the intrinsic diffusivities in homogeneous alloys using tracer techniques. When the present work was started, no information of this type was available. Consequently, measurements of the intrinsic diffusivities were made as a function of temperature at a constant composition of 28 atomic pct Zn with vapor-solid diffusion couples where the zinc was diffused into the diffusion couple from the vapor phase. The application of these couples to the study of diffusion in a: brass has been described previously.0,7 The temperature dependence of the intrinsic diffusivities was found to follow the relation D, = A, exp(-Hi/RT) and the values of Hzn, and Hcu, were found to be closely the same. It is emphasized, however, that the chemical dif-fusivity (D = N1D2 + N2D1) is a composite diffusivity and does not necessarily follow this exponential form. It is usually found to do so within experimental error for substitutional alloys because the heats of activation of the intrinsic diffusivities generally are not greatly different.&apos;" Also, at the onset of this work, there was no information available concerning possible unequal diffusion rates of individual components and the existence of a Kirkendall effect in alloys with other than face-centered-cubic structures. Since then, two reports indicating a Kirkendall effect in body-centered-cubic ß brass have appeared. Landergren and Mehl" have published a note describing Kirkendall diffusion experiments with sandwich-type couples. Inman et a1.9 also find a Kirkendall effect in this alloy using the tracer technique. In the present work, several aspects of the Kirkendall effect in ß brass were further investigated using vapor-solid couples. Two different couples were used, one in which the zinc was diffused into the specimen from the vapor phase and the other in which the zinc was diffused out of the specimen into the vapor phase. Briefly, the existence of a Kirkendall effect is confirmed and it is found that Dzn/Dcu = 3 at about the 46 atomic pct composition in this alloy at 600°, 700°, and 800°C. As a result of the unequal diffusion rates of zinc and copper, volume changes occur and subgrain formation is observed in the diffusion zone. In addition, significant porosity is produced by the precipitation of supersaturated vacancies. Diffusion in this alloy is therefore outwardly similar to diffusion in a brass where these effects are also observed, a Brass Experimental Methods—The use of vapor-solid couples in studying diffusion in a brass has been described in previous articles.6,7 The method briefly consists of sealing a copper specimen with Kirkendall markers initially placed on its surface in an evacuated quartz capsule along with a large zinc source of fine a brass chips and then diffusing the zinc into the specimen through the vapor phase. The zinc concentration at the specimen surface rises rapidly enough to a value near that of the a brass source so that the surface concentration may be regarded as constant during diffusion. Under these boundary conditions, values of the chemical diffu-sivity may be obtained by applying the Boltzmann-Matano analysis to the concentration penetration curve, and the intrinsic diffusivities may be obtained from Darken&apos;s5 equations when the velocity of marker movement is known. The diffusion specimens were made from OFHC copper in the form of disks 3.2 cm diam and 0.5 cm thick with faces surface-ground parallel to within +0.001 cm. Markers in the form of fine alumina particles <0.0002 cm diam were placed on the specimen surface. These specimens were then sealed in quartz capsules along with enough a brass chips of a 30.0 atomic pct Zn composition to keep the source concentration from decreasing by more than 0.3 atomic pct Zn as a result of the loss of zinc to the specimen during diffusion. The quartz capsules which were initially evacuated to a pressure of

Jan 1, 1956