Search Documents

By H. D. Williams, R. A. Scriven

The distribution of elevation angle of finite plane facets distributed symmetrically about a given axis is derived in terms of the angular distribution of lines which these facets produce in a plane section parallel to the axis. It is assumed that the facets are not markedly elongated in preferential directions. The theory is intended for application to the angular distribution in polycrystalline metals of slip planes and of particular grain boundaries on which certain phenomena such as cavitation can occuv. Illustrative examples are given in closed form and simple numerical techniques are de-ueloped for practical application. The results are applied to the distribution of cavitated boundaries in a copper specimen after fatigue at 400°C and it is found that voids form on the boundaries that exper~enced maximum shear stress. When investigating the distribution of elevation angle between a given axis and planes symmetrically distributed about this axis the angular distribution observed in a plane section parallel to the axis is not necessarily the same as the required distribution. Consequently a technique for processing the results obtained by sectioning methods is required. An example where such a technique would find application is in the problem of intercrystalline void formation during uniaxial tensile creep. Two theories have been proposed to explain this cavitation in terms of the normal and the shear components of stress, respectively. Hull and Rimmer (1959) postulate that the growth of intercrystalline voids is controlled by the rate of supply of vacancies from the grain boundary, the local vacancy concentration being proportional to the normal stress on the boundary. Gifkins (1956) considers that slip arising within a grain causes a ledge to be formed at the grain boundary which is subsequently opened up into a cavity by grain boundary sliding. It might be possible to differentiate between these mechanisms by determining the angular distribution of grain boundaries containing cavities since the former theory predicts a predominance of cavitated facets normal to the stress direction whereas the latter should favor an angle of 45 deg to this direction. In a similar manner an indication may be obtained of the type of mechanism responsible for intercrystalline cavitation during high-temperature fatigue. Davies and Wilshire (1962) have analyzed the slip-plane distribution in a nickel specimen but the problem was not treated in any detail nor was any account taken of the finite extent of the planes being sectioned. Hilliard (1962) has given a very general account of how anisotropy in the microstruciure can be related to measurements taken in one or more plane sections by expressing both the measured and the true angular distributions as expansions in spherical harmonics.* The present analysis is con-

Jan 1, 1965

By T. H. Alden

In several recent studies of fatigue fracture in single-phase metals, tests have been made using the condition of essentially constant plastic strain amplitude.1-4 This procedure departs from the more common practice of holding the amplitude of stress constant. From these studies have emerged two important generalizations about the behavior of metals during cyclic straining. First, measurements of the stress amplitude, required to impose the fixed plastic strain amplitude, have shown that the rate of cyclic strain hardening in annealed metals is rapid in the early cycles but decays quickly to a very low value.1"5 This behavior is displayed particularly for a range of strain amplitudes below about 2 pct and at low and high temperature. Frequently, the hardening rate becomes so low that one can determine a "saturation" stress. This stress increases with strain amplitude. In addition, an experimental relation between fracture life and plastic strain amplitude was obtained by Coffin and Taternelli,' namely N 1/2 ep = constant. For several metal:;, this equation has been shown to hold down to a strain amplitude of 0.1 pct. Whether it also holds in the "low amplitude" range (lives of more than about 105 cycles) has not been established. The importance of the saturation stress level (a value derived from strain hardening measurements) to the question of fracture life (the usual quantity of interest of fatigue testing) lies in the observation that cracks start to form when the cyclic hardening rate is essentially zero, and not before.4"5 This point suggests immediately a means of relating tests at constant stress and constant strain amplitudes. When making comparison, the stress in a constant stress test is to be identified with the saturation stress in a constant strain test. If these ideas are correct, then it should be possible, using the relation between fracture life and strain amplitude and experimental curves of saturation stress versus strain amplitude, to predict the fracture life as a function of stress, i.e. predict the S -N curve, for low stress fatigue. Saturation stress amplitude-plastic strain amplitude data, obtained in tension-compression, exist only for a few metals and for relatively high amplitudes of strain. Comparable S -N curves (push-pull loading) were known by the author only for OFHC copper. Values for this metal taken from Ref. 2 are shown in Fig. 1. The strain amplitudes, the lowest

Jan 1, 1962

By R. E. Morgan, D. L. Douglass

The determination of phase relationships and solid-solubility limits can be performed by quantitative metallography in addition to the usual X-ray and metallographic techniques. For example, Beck and smith1 redetermined the ß/ß + ?, ß + ?/?, a/a + ß and a + ß/ß boundaries in the Cu-Zn system by measuring the volume fraction of second phase of several alloys and extrapolating the volume fraction-composition curves to 0 and 100 pct. A modification of this technique is suggested for certain alloy systems, in which it is not necessary to use several alloy compositions but merely one. A single two-phase alloy may be used to determine terminal solubilities in the following manner. The method consists of equilibrating samples of the alloy in a two-phase region adjacent to the desired solid solution, at three or more temperatures, quenching, measuring the volume fraction of second phase present, and applying an analytical treatment to calculate the unknown solid solution. However, two restrictions are inherent in this technique. They are: 1) only certain types of alloy systems are amenable to it, and 2) the general features of the system must be known. The first drawback to the new technique, i.e., that only certain types of systems may be studied, necessitates that the composition at one end of the tieline must either be constant with temperature or well established as a function of temperature. Either a pure metal or some intermetallic compounds fulfill the former. If it is assumed that the volume per gram-atom of a dilute solution is unchanged by the addition of element B to element A, the composition of the solid solution in equilibrium with the second phase may be determined by a material balance and is given by where X, = volume fraction of B in a solid solution Xc = volume fraction of B in compound c X = volume fraction of B in alloy f = volume fraction of second phase The composition by weight may then be determined by the use of tables in the Metals Handbook2 when the density ratio of the solid solution constituents is known. A possible alternative treatment involving the use of the lever rule is less precise than the above tech- nique. This may be used when the density of the solid solution is either known or may be calculated from X-ray data for several compositions. The following analysis is then made. The ratio of compound to solid solution (by weight) may be expressed as follows: x0 - x wc = xr-x = x0 - x r2i Xc -X where Wc = weight of compound w = weight of solid solution x, - alloy composition, weight percent x = unknown composition Xe = compound composition but where V, = volume of compound VA = volume of solid solution pc = density of compound p, = density of solid solution fc = volume fraction of compound fB = volume fraction of solid solution and fs = l -fc then If pc and xc are known, and f, is measured, then pB is the only unknown on the right side of Eq. [4]. The known densities of the solid solution can be plotted for various compositions and can then be expressed mathematically as a function of composition. The use of an expression of pB = f(x)reduces the equation to one unknown—the desired solubility. In the event that the densities are unknown, they may be calculated for various compositions from Vegard&apos;s law. The calculated values are then plotted and expressed analytically. The most accurate results are obtained for Eq. [4] when fc<< 1, i.c., when (x, -x,) - 0, &/l-f, - m; but as fc/l -fc - 0, (xl-x,)- (x, - x), and x - x,,. However, the accuracy with which fc can be measured decreases as f, decreases.3 Alloys for investigation must be selected by a compromise, which is based upon an error analysis of Eq. [4] and knowledge of the accuracy of volume fraction measurements. An examination of phase diagrams in the literature showed many which were amenable to the technique described here. The zirconium-copper system was selected in order to determine the solubility of copper in beta zirconium. Pieces of an alloy which was arc-melted three times were wrapped in tantalum foil and sealed under an argon atmosphere in Vycor tubes. The sealed samples were equilibrated at temperatures from 850" to 960°C for 3 weeks and quenched to room temperature by smashing the capsule in water. Several planes of polish were examined, and

Jan 1, 1960

By J. B. Clark, A. S. Yue

The zone-melting technique can be adapted for the de-termination of the eutectic composition in complex metal systerrzs. The application of this method is demonstrated in a simple eutectic system, Mg-AE, in which the eutectic composition is known and in a complex ternary system, Mg-Al-Zn, in which the literature is uncertain as to the composition of the ternary eutectic. The advantages and limitations of this unique approach for the determirution of the eutectic composition are discussed. ThE eutectic composition in phase diagrams is usually determined by thermal analysis of a series of arbitarily selected alloy compositions. The temperatures of the liquidus and eutectic arrests for each alloy are plotted and the eutectic composition is estimated by extrapolation to the entectic temperature of the curve formed by the liquidus arrest temperatures for the series of alloys. Although this approach readily yields the eutectic temperatures, several dekrminations are needed to estimate the eutectic composition, even in a simple system. In a complex ternary system, a great many determinations are often necessary to estimate, even approximately, the ternary eutectic composition. Thus, the conventional method is inherently tedious and time consuming. The application of zone melting provides a relatively efficient method of determining the eutectic composition.&apos; As an illustration, consider the zone melting of an alloy composition C, in the simple eutectic system A-B, shown in Fig. 1, under the ideal conditions of infinitely rapid diffusion in the liquid, no diffusion in the solid, and ease of nucleation of both phases. As the zone moves along the bar, the composition of the zone is enriched by solute rejected during the freezing of the a phase and the zone composition approaches the eutectic composition CE.Once the zone composition attains the lowest melting composition, i.e., the eutectic composition, it remains at this composition as the zone moves along the bar and finally freezes as such at the end of the bar. Chemical analysis of the end of such a zone-melted bar then should yield the eutectic composition CE of the system A-B. A corresponding analysis can be made for a ternary eutectic system. Under the ideal conditions noted above, only one pass of the molten zone is required to yield the eutectic composition. Experimentally, however, a solute-rich layer forms at the solid-liquid interface2 and the composition of the zone does not attain the eutectic composition as rapidly or as uniformly as in the ideal case outlined above. But with agitation of the melt, the ideal conditions can be approximated experimentally. Also, as shown below, zone melting permits a cumulative build-up of the solute at the end of the bar to the eutectic composition. The purpose of this paper is to demonstrate this unique method for determining the eutectic composition. The method is here applied to two systems: first, to the binary Mg-A1 system, in which the composition is well established, and second, to the complex Mg-Al-Zn system in which there is doubt concerning the composition of the ternary eutectic. EXPERIMENTAL PROCEDURE The zone-melting apparatus used in this study is described in detail elsewhere.3 Briefly, the binary alloy rods (5/8 in. in diam and 9 in. long) and the ternary alloy rods (5/8 in. in diam and 6 in. long), with the compositions given in Table I, were zone

Jan 1, 1962

By R. T. DeHoff

A general method for determining the geometric properties of structures composed of particles which are all the same shape and size is presented. The application of the method requires a knowledge of&apos; the qualitative shape of the particles in the structure, ad a maximum of three simple counting measurenzents, which are made on a representative plane section taken through the structure. It is also shown that the number of different kinds of measurements necessary JOY the complete description of the structure can be decreased if some independent information about individual particles is available. The geometric properties that can be determined quantitatiz~ely from counting measurements for susch structures are size, shape, number, surface area, vlolume, ad such extensive properties as volume Mction and surface area per unit volume. THE simplest kinds of measurements that may be made on a metallographic section are counting measurements. There are three such simple counting measurements, which are determined by sampling the microstructure with a point, a line, and an area. The first of these, the point count, is probably the most familiar among metallographers. The second count, the number of intersections a test line makes with particle outline, is somewhat more recent in its origin and application. The third counting measurement, the number of particle sections observed in unit area of the plane of polish, has long been used in the estimation of grain sizes. The real utility of these three counting measurements lies in the fact that they are rigorously and unambiguously related to certain extensive geometric properties of the three-dimensional structure of which the metallographic section is a sample. These relationships will be reviewed briefly below. Of somewhat secondary importance to the fundamental relationships is the observation that, if some simplifying assumptions about the geometry of the three-dimensional structure are introduced, manipulation of the counting measurements gives a more complete description of the structure. Specifically, in addition to the extensive properties which these measurements rigorously estimate, the number of particles, their size, and their shape may be determined. The purpose of the present paper is to explore the consequences of introducing the assumption of constant size and shape. A later paper will deal with the estimation of geometric properties of structures which may be characterized by a two-parameter size distribution. It should be mentioned that the introduction of this simplest of assumptions, while clearly not generally justified in metallurgical structures, is not without precedent. For example, many estimates of the number of particles per unit volume of structure, based upon a single counting measurement, are scattered throughout the literature.13 Virtually all of these developments assume a constant particle shape, e.g., spherical or polyhydral, and constant size. That such an assumption is necessary is evident from the fact that, for more general structures, more than a single parameter is required to describe the three-dimensional structure, so that the determination of a single parameter on a section would be insufficient to specify the structure. Similar developments for the estimation of particle size from a single counting measurement, e.g., routine grain-size determinations, make the same very limiting assumption,435 but have nonetheless proven of practical value. The present paper embodies a generalized approach to the development of relationships among the two-dimensional counting measurements and the three-dimensional geometry of the structure, subject to the assumption of constant particle size and shape. It is the hope of the author that by presenting the relationships and the assumptions involved in dealing with this approximation, in a single, unified treatment, the reader may be impressed with the usefulness of the counting measurements, may be guided in their application to the estimation of details of the geometry in specific structures in which he may be interested, and may be moved to apply the results, with understanding, to new structural problems. Several authors6, 7 have demonstrated independently that it is possible to obtain unbiased estimates

Jan 1, 1964

By R. T. DeHoff

Saltykov&apos;s meihod for the determination of the size distribution of spherical particles from measurements made on random plane sections is extended to particle shapes which are not equiaxed. The analysis permits the deterrrzination of the size distribution of aggregates of ellipsoidol particles provided all particles have tlze same shape. V ARIOUS solutions to the problem of the determination of the size distribution of spherical particles from measurements made on random plane sections have appeared in the literature since the original solution was first published by Scheill in 1934. Solutions have been offered utilizing measurements of the diameter1y2 or area3&apos;4 of circular sections that result when a test plane is passed through an aggregate of spherical particles, and the length of intercept5y6 that results when a test line is passed through the aggregate. To the best of the knowledge of the author, no method has been proposed which permits the determination of the size distribution of particles that are not equiaxed. The purpose of this paper is to develop a method which permits the determination of the size distribution for a specific class of particle shapes which cover awide range of axial ratios; namely, ellipsoids of revolution. In the interests of simplicity, it is necessary to limit consideration to aggregates of particles which are all similar in shape; that is, the ellipses which, by rotation about one of their axes, generate the particles, all have the same axial ratio. The size distribution, which is to be determined by the analysis, is arbitrary. It is also assumed that the section, or sections, upon which the measurements are made represent the average of all possible sections through the structure. This ideal condition can be approached as nearly as may be desired by observing a sufficient area of sections taken at random positions in a randomly oriented aggregate, and in random positions and orientations in an aggregate that has one or more axes of orientation. SIZE DISTRIBUTION FOR PROLATE ELLIPSOIDS Consider an aggregate of cigar shaped, geometrically similar particles (prolate ellipsoids), dispersed in an opaque matrix. Let the size of each particle be specified by the length of its minor axis, By Fig. 1, and the shape, q, be defined by the ratio of its minor to its major axis, B/A. Let the index, j, refer to dimensions of ellipsoids in a particular size class, and the index, i, to dimensions of an ellipse resulting from the intersection of a test plane and a particle. The size of such a section is specified by the length of its minor axis, b. Let the particles be arbitrarily divided into k size classes, and let the increment of size between each class be A, where A = &,$, B,,,being the minor axis of the ellipsoids in the largest size class. Since all sections are represented on the test plane, the minor axis of the largest of the sections observed, results from the intersection of the test plane with an ellipsoid in the largest size class at its center, and is therefore equal to Bmax Thus the increment, A = The final result will consist of the k numbers which represent the number of particles in each size class per unit volume of aggregate:

Jan 1, 1962

By R. A. Wood, R. I. Jaffee, H. R. Ogden, D. N. Williams

The tensile properties, creep resistance. and thermal stability of highly alloyed Ti-Al-Zr alloys were examined. On the basis of these studies, the Ti-7Al-1ZZr composition was selected for more complete evaluation. The alloy was found to be weldable and free from excessive directionality. In addition, it developed maximum properties without requiring heat treatment other than an annealing operation in the alpha field. The alloy was recommended for scale up and is presently being investigated on a production-level basis. One of the more attractive properties of titanium alloys is their ability to withstand stress at moderately high temperatures, and a considerable amount of effort has been devoted to increasing the maximum service temperature of titanium alloys. This work has suggested that the optimum alloys for high-temperature service will be single-phase a (close-packed hexagonal) alloys containing significant amounts of aluminum. However, the maximum amount of aluminum which can be alloyed with titanium is between 6 and 8 pct,l since at high-aluminum contents an embrittlement reaction occurs in the anticipated service temperature range, 800" to 1100°F. It has been shown that the embrittlement reaction involves decomposition of the high-aluminum a phase to one or more new phases.&apos; Since this reaction does not occur at intermediate or low-aluminum contents, it was felt that intermediate Ti-A1 alloys might be strengthened by a-soluble ternary additions without inducing the embrittlement reaction. The first alloying addition considered was tin, which shows extensive solubility in a titanium and has moderate strengthening tendencies. Unfortunately, it was soon apparent that tin also promoted the embrittlement reaction, and that to obtain a stable alloy, the aluminum content had to be reduced as the tin content was increased. The second alloying addition considered was zirconium, which is similar to tin in its effects on titanium. This element did not contribute to the embrittlement reaction and, in fact, appeared to increase the maximum amount of aluminum which could be alloyed with titanium without inducing instability. This paper describes an investigation of the Ti-A1-Zr a alloy region. Alloys containing from 4 to 12 pct A1 and from 6 to 15 pct Zr were examined. The properties of these alloys are described and the bases for selecting an optimum composition is outlined. This composition, Ti-7A1-12Zr, is presently being scaled up in tonnage quantities, and is being evaluated extensively throughout the industry. In addition to presenting the basis for its selection, this paper presents a description of the properties developed in laboratory material as determined during the alloy investigation. These properties suggest that this alloy can fill an important position in applications requiring light weight, fabrica-bility, weldability, and strength to 1000oF or higher. EXPERIMENTAL PROCEDURES Titanium alloy ingots were prepared by inert electrode arc melting under an argon atmosphere. Alloying elements used were 110 Bhn titanium sponge, high-purity aluminum, and reactor-grade zirconium. Pancake-shaped ingots were prepared weighing approximately 300 g. The composition of the ingots was checked by weight measurements before and after melting. The pancake ingots were forged at 2000°F to approximately half their original thickness to give a flat plate roughly 1/2 in. thick. This plate was then rolled at 1800&apos; to 1600°F to 0.250 in. thick. All of the alloys examined fabricated well. However, alloys containing 15 pct Zr tended to overheat due to exothermic oxidation, and scaling was excessive. As might be anticipated from its effect in decreasing the ß transus, increased zirconium appeared to improve fabricability somewhat, especially during rolling at lower temperatures. Except for a limited study of heat-treatment response, all alloys were examined in the a-annealed condition. Prior to heat treatment the a and ß tran-sus temperatures were determined by metallo-graphic examination of samples quenched after annealing at 50-deg intervals in the transformation region. These data are shown in Fig. 1. Recrystal-lization appeared to occur in about 1 hr in the range 1300º to 1500ºF. Therefore, alloys were annealed for 1 hr at 1550ºF (4 and 5 pct Al), 1600ºF (6 through 7-1/2 pct Al), or 1650°F (8 or more pct Al). This produced an equiaxed a grain structure. In most alloys, a "ghost" structure was visible after the a-annealing treatment, as shown in Fig. 2. This structure apparently resulted from the acicular

Jan 1, 1963

By J. K. Stanley

Knowledge of the diffusivity of carbon in the low temperature form of iron (alpha iron existing below 910°C) is at the moment of considerable interest in the study of the decomposition of austenite and martensite, the elastic after-effect,123 the magnetic after-effect4 and the decarburization of steel below 910°C. Information on the solubility of carbon in iron, and to a lesser extent its diffusion, is also important in consideration of such phenomena as blue-brittleness, temper-brittleness, "magnetic" aging, quench-aging, strain-aging, and possibly the yield point. In order to obtain more information on these subjects more fundamental knowledge is necessary. It is the purpose of this work to present data on the diffusion and solubility of carbon in the alpha iron. The high temperature form of iron (gamma; face-centered cubic) existing above 910°C is capable of dissolving relatively large amounts of carbon, up to 1.7 pet at 1130°C, while the low temperature form (alpha, body-centered cubic) existing below 910° dissolves only a limited maximum amount of less than 0.02 pet carbon at 725°C, according to data obtained here. Since the solubility of carbon in the face-centered or gamma iron is large, relatively speaking, no great analytical difficulties have been encountered in the determination of the solubility lines5 or of the diffusion of carbon.0 The limited solubility of carbon in alpha iron offers difficulties because experimental procedures and analytical methods for low carbon contents below say 0.01 pet have to be more refined than techniques used for work with gamma iron. Because of the difficulties of applying conventional methods to the determination of the diffusion of carbon in alpha iron, virtually no work has been done on this subject. However, by proper refinement of the analytical method for small amounts of carbon, the determination of the diffusion coefficient can be made readily using modified procedures. The solubility of carbon in alpha iron has been determined over a temperature range by various investigators, but the agreement among them is poor. The present investigation establishes the limits quite accurately. Information of this kind is useful in establishing the correctness of equilibrium diagrams but, more significantly, such information on maximum solubilities, especially when extended to alloyed ferrites, should be extremely important in the study of aging and related phenomena. Literature The literature existing on the diffusion, in particular, and on the solubility of carbon in alpha iron is not extensive. The data which exist are not of a high order of accuracy, much of them being in the realm of conjecture. THE DIFFUSION OF CARBON IN ALPHA IRON Whiteley7 made the qualitative ob- servation, using metallographic techniques, that the rate of diffusion of carbon at the A1 (725°C) point was very rapid and that its diffusion was still rapid at 550°C. Snoek,4 studying the magnetic aftereffect in high purity iron, arrived at the conclusion that the after-effect could be explained by the presence of small amounts of carbon diffusing under the influence of magnetostrictive strain (lattice distortion due to magnetic interaction). In later work, Snoek8 made an estimate of the ratio of carbon diffusion in alpha to its diffusion in gamma iron, and concluded that for a temperature of 910°C the ratio of Da/D? was 2600. Polder,9 basing his calculations of D on relaxation phenomena in the elastic after-effect, estimated that Da is about 1/3 of D? at 910°C (1183°K) and is about 1/12 of Dy at 727°C (1000°K). Polder&apos;s equation for the diffusion of carbon in alpha iron was calculated to be 18000 D = 5.2 X 10-4 e-RT cm2 per sec Ham10 obtained data for the diffusion and solid solubility of carbon in alpha iron at two temperatures by using one technique similar to that employed in this study. He found a D of 8.0 X 10-7 cm2 per sec at 702°C and of 2.7 X 10-7 at 648°C. THE SOLUBILITY OF CARBON IN ALPHA IRON Although pearlite is absent in steels containing 0.06 pet,11 0.05 pet,12 or 0.045 pet C,13 it appears that the carbon in these steels cannot be in solution in ferrite. The solubility of carbon at the A1 (725°C) point was first determined by Scott14 on the basis of cooling curves, and was found to be between 0.03 and 0.04 pet C. Tamura15 by interpolating between the solubility of carbon in delta iron at 1400°C and in alpha at room temperature (assuming zero solubility) ar-

Jan 1, 1950

By A. S. Appleton

The diffusion of carbon in Fe-Co-C austenites was studied as a function of cobalt concentration up to 12 at. pet, using an autoradiographic technique. The diffusion coefficient was found to initially decrease and then increase with increasing cobalt concentration, the trends being more marked the higher the temperature of diffusion. These effects were paralleled by variations in the activation energy and the constant for diffusion which, however, were found to be dependent also on the carbon content. The results are compared with the previously available data and their significance is discussed in terms of the kinetics of solid-state reactions in this systern. THE influence of cobalt on the diffusivity of carbon in austenite was first studied by Smoluchowski,&apos; his data being that most commonly cited in works

Jan 1, 1964

By H. Becke, D. Stolnitz, D. Flatley, W. Kern

Zinc difhsions in gallium arsenide having surface concentrations as low as 5 x 10&apos;&apos; atoms per cu cm have been attained. A multiple-difhsion sequence is employed during which zinc enters the gallium arsenide from a silicon dioxide film in a solid-to-solid diffusion step. The technique is illustrated for the case of a zinc surface concentration of -1 x 1017 atoms per cu cm and a junction depth of 1.2 p. Radiotracer measurements of -0.1-p layers of gallium arsenide are used to delineate the resulting profile. A diffusion coefficient for zinc in pyrolytically deposited silicon dioxide is reported. ZINC is frequently used as an acceptor impurity in gallium arsenide. As examples, p-n junctions are formed using zinc-diffused layers in solar cells, lasers, and varactor diodes. For these purposes surface concentrations of 10&apos;&apos; to over loz0 zinc atoms per cu cm are desired and readily obtained from a vapor source. Diffusion of zinc into gallium arsenide to yield surface concentrations of about 1017 atoms per cu cm, however, is more difficult. For instance, oldstein&apos; reports that large changes in the vapor density of zinc produce only small changes in surface concentration. Such low surface concentrations are necessary in forming the diffused base of a gallium arsenide double-diffused n-p-n transistor in order to obtain an adequate emitter efficiency. An alternative approach to zinc diffusion which can provide reduction in surface concentration of several orders of magnitude is the introduction of the diffusant from a solid source. This can be conveniently accomplished by using oxide films of the type generally employed for selective diffusion masking. Such oxides also serve to prevent the dissociation of the gallium arsenide surface at the high temperatures of diffusion. DIFFUSION PROCEDURE Fig. 1 illustrates the diffusion sequence which is employed for obtaining zinc surface concentrations in the range from 5 x lom to 1 X 10" atoms per cu cm. The description of technique will be given for the instance of -1 x 1017, which is typically utilized in double-diffused gallium arsenide transistor investigations at this laboratory. Tin- or germanium-doped gallium arsenide wafers drawn from crystals grown by horizontal Bridgman or Czochralski techniques are used, with bulk concentrations ranging from 1 to 3 X 10" carrier atoms per cu cm. Each wafer is chemically polished on the (iii) arsenic face using a mixture of five parts concentrated sul-furic acid, one part water, and one part 30 pct hydrogen peroxide at room temperature, followed by rinsing in deionized and quartz-distilled water. Pyrolytic silicon dioxide was chosen as the oxide coating because of the high quality of the films obtainable, the ease of their preparation, and the excellent control over film thickness and uniformity.

Jan 1, 1964

By Rodney P. Smith

The rate of motion of the boundary between a single-phase a-iron region and a two phase a-iron plus ?-iron, or a-iron plus cementite region has been measured on decarburization at several temperatures between 500" and 865°C. These data, together with data for the carbon content of a-iron at the boundary, yield the diffusivity of carbon in a-iron. The activation energy for carbon diffusion is 19.8 kcal per mole for the temperature range —50° to 150°C and increases to an average value of 23.0 kcaL per mole for the temperature range 727" to 865°C. AS part of an investigation of the diffusivity of carbon in iron using a steady-state method,&apos; the diffusivity of carbon in a-iron was determined at the temperatures 802" and 851°C. The diffusion constant thus obtained was larger, by nearly a factor of two, than that predicted from the combined results of stanley2 and of Wert.3 It seemed desirable, therefore, to continue the investigation. Since the temperature range in which the steady-state method can be conveniently used is limited, we have chosen a decarburization method, based on the laws of diffusion in metal accompanied by a phase change.4,5 If the initial carbon content of the Fe-C alloy is such that two phases coexist at the decar-burizing temperature* and decarburization is car- ried out such that the surface is maintained at substantially zero wt pct. C, the carbon distribution after a time, t, will be as shown in Fig. 1. The boundary between the single-phase and two-phase regions moves from y = 0 to y = x in the time, t, where x is proportional4 to tl". If the initial carbon (content of the sample is Ci, and CB is

Jan 1, 1962

By R. P. Smith

The diffusivity of carbon in iron, cobalt, and alloys of 89.7 and 79.3 wt pct Co has been determined by a decarburization method for the temperature range 850° to 1100°C. The Plots of log D us 1/T for the two Fe-Co alloys are linear through the region of the Curie temperature, showing that, unlike some cases of substitutional diffusion, carbon diffusion in these alloys is not affected by electron spin order. RECENT investigations1-&apos; have shown that for substitutional diffusion in iron the change in diffusivity with temperature does not follow an Arrhenius equation, at temperatures near the Curie temperature. Data on the diffusivity of carbon in a iron5"9 indicate that this anomaly does not exist for interstitial diffusion; however, this has not been definitely established. Since cobalt and Fe-Co alloys, 80 to 100 pct Co, have Curie temperatures in a temperature range convenient for carbon-diffusion measurements they are suitable systems for studying the relation of interstitial diffusion and the Curie temperature. EXPERIMENTAL The carbon diffusivity, D, was determined by a decarburization method. The diffusion cylinders, made from small vacuum melts of electrolytic cobalt and plastiron, were about 0.76 cm diam and 6.3 cm long. A small amount of carbon was added to each melt to keep the oxygen content as low as possible. The cylinders were first treated with a hydrogen-3 pct water vapor mixture at 1100°C to constant weight, then carburized to the desired initial carbon content with a suitable CO-C02 gas mix-

Jan 1, 1964

By E. W. Johnson, M. L. Hill

The dijfusiuity D was determined at 25° to 780°C- from hyd?-ogen evolution rates. Anomalous evolution from air-melted iron was att~zbztted to residual hydrogen, which is interpreted as a hydrogen compound that decomposes at a strongly temperature-(Ieperzdent rate above 500°C. Above 200°C, D = 0.0014 cm2sec-1 exp (-3200 cal per g atom/RT). Below, 200°C the diffusivity is anomalously low and represented by D = 0.12 cm2sec-l exp (-7820 cal per g atom/RT). It is postulated that at low temperatures the hydrogen in the metal is "trapped" with an energy 4.8 kcaL per g atom below- that of the interstitial hydrogen. 1 HE purpose of the present experimental study was to determine the diffusivity of hydrogen in a-Fe over an extended temperature range. As certain anomalies were encountered, attempts were made to understand the anomalies themselves. This research was in progress both before and after a determination of the diffusivity of hydrogen in nicke1.l The latter system was found to exhibit none of the anomalies of hydrogen in iron, and its description was far simpler. The experiments on nickel also

Jan 1, 1961

By D. Gupta, S. Weinig

This investigation comprises the study of dislocation-oxygen interactions in a! titanium and its effect on the ductile to brittle transition in titanium. Internal friction techniques using a low-frequency lorsion pendulum were used to study the oxygen atom interactions with dislocations in "superpurity" titanium. A calculation of the interaction energy between an oxygen atom and a dislocation, based on the "breakaway" strain resulted in a value of 0.015 ev as compared to theoretical values of 0.022 and 0.035 ev for edge and screw dislocations, respectively. Both the saturation of the room temperature decrement and the transition from ductile to brittle behavior occurred at 1.5 at. pct 0. It was proposed that the embrittlement was due to oxygen atoms immobilizing the dislocations. It was possible to increase the ductility of the Ti-0 binary alloys by decreasing the oxygen pinning. The increase in uniform elongation was 30 pct for an increase in the preanneal temperature of 300°C. ThE room-temperature ductility of a titanium is significantly affected by interstitial solutes. Although oxygen has a solid solubility of 15.5 pct by weight, the room-temperature ductility extends only to 0.6 pct.&apos; The ductility precipitously falls at this concentration and the fracture characteristics change from ductile to brittle.2 The small effective bulk composition of solute required to destroy the room-temperature ductility of the titanium must be indicative of an in-homogeneous distribution of solute atoms. Internal friction techniques were used in this investigation to

Jan 1, 1960

By Paul Gordon

The distribution coefficient, k, of interest in zone refining is generally defined as the ratio of the solid to the liquid solubilities of one element in another at the normal melting point of the solvent. In practice, distribution coefficients are frequently calculated from solubilities at temperatures well below the melting point of the solvent, e.g., at a eutectic temperature. This is done because of the general lack of accurate data on the liquidus and solidus curves in most phase diagrams. The practice is equivalent to assuming that the liquidus and solidus curves are straight lines in the temperature range between the melting point of the solvent and the temperature at which data for the solubilities are known. It is recognized that the k values so derived may be in serious error; we have found evidence in recent zone-refining operations that a case in point is silicon in aluminum. The phase diagram found in the literature for silicon in aluminum indicates a eutectic on the aluminum-rich side at 12.5 at. pct Si and at 577°C,1 and a solid solubility of silicon in aluminum at this temperature of 1.59 pct.2 It also shows the liquidus and solidus as nearly straight lines from 577°C to the melting point of aluminum (660°C)1,2 The k value obtained from this data is 0.13. This is almost exactly the same as the k value (about 0.14)&apos; obtained in a similar way for copper in aluminum. Yet we have found by mass-spectrograph analyses* of three bars of zone-refined aluminum that copper is efficiently removed from aluminum by zone refining but silicon is not. Two of the three bars of aluminum were zone-refined in our laboratories, the other in the laboratories of AIAG Metals, Inc. The former two bars, designated bars PG2 and PG3, were 30 in. long by 1 in. in diameter during refining; each was processed in a high-purity graphite boat* and kept under refining stage of the process consisted of thirty unidirectional zone traverses at rates of about 1 to 2 in. per hr, with a zone length of 1 to 2 in. At approximately the middle of the refining period a 6-in. piece of the impurity-laden finishing end of the bar* was cut off and replaced with new start- ing material. Following refining, the purest 20-in. section of the bar (obtained by discarding approximately 2 in. of the starting end and 8 in. of the finishing end of the bar) was leveled and then analyzed. The AIAG bar was 16 in. long and about 1.2 in. in diameter; it was purified by eight passes of a 2-in. zone at a rate of 1.6 in. per hr, the bar being contained in a sintered alumina boat lined with loose alumina powder of the highest purity. Data for the silicon and copper contents of the three refined bars are given in Table I. Derived data for the efficiencies of the zone-refining operations are listed in Table 11. It is clear from these two tables that the copper is removed much more efficiently—by a factor of 10 to 50 times—from aluminum during zone refining than is silicon. For example, in bar PG3 the thirty zone passes reduced the copper content from 10 to 0.02 ppm, i.e., by 500 times, whereas the same treatment lowered the silicon content by only a factor of about 16. It seem virtually certain, therefore, that the distribution co efficient, k, for silicon in aluminum is well above that for copper and is very likely not much below unity. This also implies that the solidus line on the aluminum-rich side of the A1-Si phase diagram is very probably appreciably curved near the aluminum melting point since the phase-diagram data indicate such curvature is unlikely in the liquidus.

Jan 1, 1965

By Victor V. Damiano

THOMPSON1 proposed a very convenient nomenclature in which an imaginary regular tetrahedron was used to facilitate the analysis of dislocation interactions in fcc materials. The faces of the tetrahedron were constructed parallel to four different sets of (111) planes, and the six edges of the tetrahedron then represented the six 1/2 [110] Burgers vectors. The lines which joined the midpoint of a tetrahedron face with the ends of one of its edges represented the Burgers vectors of the partial dislocations into which the six 1/2 [I101 Burgers vectors dissociate. Using Frank&apos;s2 notation, all of the dislocations and their partial dislocations were identified and numerous possible interactions were postulated. The case of hexagonal crystals is unfortunately more complex, but it appears convenient to introduce a system similar to that of Thompson1 to describe the various slip planes and Burgers vectors encountered in cph structures. The system proposed has limited application but serves to identify the following slip systems {0001} <l20>, {1010} <1120>,

Jan 1, 1963

By J. H. Jackson, R. B. Fischer

VERY little is known about the properties of relatively pure refractory metals in the cast state since these metals are customarily made by powder-metallurgy methods. Recently, the development of the vacuum arc-melting furnace has made possible the study of large, sound ingots of cast molybdenum. The ductility of molybdenum, at room temperature, may vary considerably. Often the brittle state depreciates the utility of molybdenum, even for the material prepared by powder-metallurgy techniques. At the outset of this program, it was felt that if the causes of brittleness were to be determined it might be possible to reduce the difficulties met in the working of molybdenum. Also, the problem of producing strong, fusion weldrnents of molybdenum might be solved. The conditions required to produce ductile molybdenum by the powder-metallurgy technique are reasonably well known. A fibrous or elongated-grain structure has good ductility although this property is probably directional. The equiaxed structure obtained in fully recrystallized molybdenum has been considered to be brittle. The reasons for this behavior of molybdenum have not been established. The preparation of cast molybdenum ingots by the vacuum-arc method was described by Parke and Ham.&apos; By this method, ingots weighing up to 250 Ib were produced." Parke and Ham1 found that molybdenum ingots containing more than approximately 0.0025 pct oxygen could not be hot forged. On the other hand, they found that fully deoxidized cast molybdenum, containing up to 0.06 pct carbon, could be hot forged. However, the as-cast ingots containing either carbon or oxygen were brittle at room temperature. Parke and Ham,1 Woodside,3 and Zapffe, Landgraf, and Worden4 noted the inter-granular precipitation of oxide or carbide in cast molybdenum. This study—part of an extensive program on mo-lybdenum—had the following objectives: (1) To investigate the causes for the brittleness of cast molybdenum at room temperature, (2) to investigate the means for improving the ductility of cast molybdenum at room temperature, and (3) to add to the knowledge of cast molybdenum and of cast metals in general.

Jan 1, 1951

By P. Pietrokowsky, T. E. Tietz, J. E. Dorn

The amount of solid solution hardening in aluminum alloys was found to be dictated by two factors: the lattice strain, and the change in the mean number of free electrons per atom of the solid solution. To obtain this correlation it was necessary to assume that aluminum contributes two electrons per atom to the metallic bond. WHEN the modern scientific method of analysis was first being formulated, Francis Bacon recorded in his "Essays" (circa 1600) that "an alloy . . . will make the purer but softer metal capable of longer life." During the intervening centuries voluminous data have been reported which demonstrate that the additions of alloying elements do in fact increase the hardness and strength of the pure metals. Nevertheless, the significant details of this problem on the unique effect of each element toward enhancing the mechanical properties of alloys only recently have been subjected to systematic scientific scrutiny. The major objective of this investigation is to determine how minor additions of alloying elements affect the plastic properties of polycrystalline aluminum alloys. By means of such studies it is hoped to provide not only data on the solution strengthening of aluminum alloys, but also a body of facts which will supplement the knowledge already available on the factors responsible for solution hardening in general. A review1"10 and analysis1&apos; of the existing data on the effect of solute elements on the plastic properties of solid solutions reveal that our current knowledge on solid solution hardening is somewhat meager, inconsistent, and inconclusive. Many of the inconsistencies are undoubtedly attributable to the influence of unsuspected factors, such as purity; or uncontrolled factors, such as grain size, on the plastic properties of alloys. Nevertheless the following conclusions might be tentatively accepted: 1. Addition of solute elements invariably increases the yield strength, tensile strength, and hardness of the host element. 2. The rate of strain hardening, in general, increases with the concentration of the alloying element. 3. The strengthening effect in ternary alloys is the sum of the individual strengthening effects of the two solute elements as measured in their binary alloys. 4. The lattice strain is one factor that affects the strengthening of the alloy but it is not the only factor. 5. A second factor might be the difference in valence between the solute and solvent metals. All of the available evidence is in complete agreement with the first conclusion; the remaining conclusions, however, are not in agreement with all of the published data, but, in each case, the major weight of the existing evidence favors these deductions. Additional investigations will be required before most of these tentative conclusions can be accepted without reservation. In the following report an extensive investigation of the plastic properties of binary aluminum alloys is described. This work was undertaken in an attempt to shed more light on the general problem of solid solution hardening. Materials for Test: Aluminum was selected as the solvent metal for the present investigation on the effect of solute elements on the plastic properties of alloys. This choice was made for several reasons: (1) There appears to be little fundamental data in the published literature on the effect of solute elements on the properties of high-purity aluminum alloys. In view of the ever increasing economic importance of aluminum, such data would be of basic interest to the metallurgists concerned with the development of new aluminum alloys. (2) Aluminum is thought to be only partially ionized in the metallic state1&apos; and consequently it might provide more complex relationships of the mechanical properties with the concentrations of the solute elements than more simple fully ionized solvents would reveal. (3) The data on aluminum alloys will provide a broader basis for correlations between the mechanical properties of metals in general and the concentration and atomic properties of the solute elements than is now available. Some complications, however, attend the selection of aluminum: The solubility of the various elements in the alpha aluminum phase are quite restricted, and not always well known. Consequently, only dilute solid solutions are available for study. This, however, may be somewhat advantageous because the dilute solution laws presumably are simpler than those applying to concentrated solutions. In addition, strain-hardened pure aluminum is known to recover at atmospheric temperatures. Very likely its alloys exhibit slower recovery rates. Thus, the secondary factor of effect of alloying elements on recovery might complicate the data. Such compli-

Jan 1, 1951

By John D. Verhoeven

A technique has been developed for carrying out normal freezing experiments with a current density of 2000 amp per sq cut passing through the solid-liquid interface. The equation relating the effective distribution coefficient to the equilibrium distribution coefficient in electric field-aided solidification, originally developed by Huckc et al.,1 has been modified for the case of concentrated solutions. Preliminary experiments on the Sn-Bi system give qualitative agreement with the equation. The data are analyzed by a slightly novel use of the normal freeze equation which allows one to determine the effective distribution coefficient more easily. Very extensive mixing in the liquid was found at these high current densities and it is postulated that the mixing results from a vertical component of the magnetic Lorentz force generated by the electric current. In the search for techniques of obtaining ultrahigh-purity metals the inefficient but very effective technique of electrotransport has received little attention. Electrotransport is most effective in the liquid state and a natural application, therefore, is to apply an electric field across the liquid zone of a zone-melting experiment. The present investigation was undertaken to study the effect of an electric field upon solidification of metals, so that the usefulness of electrotransport in such solidification experiments as zone melting could be determined. In zone-melting and normal-freezing experiments it is difficult to achieve complete mixing in the liquid in the immediate vicinity of the solidifying interface. Consequently a solute build-up will occur at the interface in the portion of the liquid where complete mixing does not occur (an equilibrium distribution coefficient, ko, less than one, and unidirectional atom motion will be implied throughout). This local solute build-up produces a corresponding rise in the solute concentration in the solid so that the ratio of the solute concentration between the solid and the bulk liquid is larger than the equilibrium distribution coefficient. This ratio is defined as the effective distribution coefficient, k,. The differential equation describing the solidification process may be derived by applying the continuity equation to an expression for the net solute flux at the interface. The solution to this differential equation then allows one to determine the solute distribution in the liquid and the relationship between k0 and ke. One of the most useful solutions to this equation was first derived by Burton, Prim, and Slichter,&apos; in which they assumed that a) the equilibrium distribution of solute was maintained on the plane of the interface, 11) the solute build-up ahead of the interface in the liquid disappeared at a distance 6 from the interface, and c) the solute distribution in the liquid was invariant with time. The following well-known relation between ko and ke was then obtained, where R is the rate of solidification and D the diffusion coefficient of the solute in the liquid. This equation appears to correlate the data from a number of different types of solidification experiments very well. Application of an electric field across the solid-liquid interface can produce an additional flow of solute atoms as a result of the electrotransport. When the polarity of the field is such as to direct the electrotransport flux away from the interface the solute build-up may be diminished, even to the point of producing a solute depletion and a consequent ke smaller than ko. The quantitative description of this process and the resulting form of Eq. [I] was first given by Hucke et al.1* and then inde- where E is the electric-field intensity and U is the differential mobility, i.e., the velocity of the solute atoms with respect to the solvent atoms per unit electric field. Both authors follow the method of Burton, Prim, and Slichter in their derivation, the only difference being the additional electrotransport term in the flux equation. It has been pointed out1,3 that Eq. [2] predicts the possibility of a noticeable increase in the purification of materials by solidification in an electric field. The validity of Eq. 121 has not been checked experimentally and it is possible that other factors&apos; arising from the presence of an electric field across

Jan 1, 1965

By P. W. Barton, A. A. Johnson, E. J. Hughes

THE only systematic work on the effect of carbon on the recovery and recrystallization kinetics of iron so far reported in the literature is that of Venturello et al.1 These workers found that doping high-purity iron so that it contained up to 86 ppm C caused a marked retardation of recovery but had little or no effect on the recrystallization kinetics. Such evidence for an effect as was obtained was inconclusive. In the course of a study of electron-beam zone-refined iron the authors have recently examined the effect of added carbon on the recrystallization kinetics of this material. The amount of carbon added was about 50 pct more than was used by Venturello et 01. and definite evidence was obtained that carbon at this level retards recrystallization. The starting material was taken from a batch of 0.2-in.-diameter Johnson-Matthey spectroscopically pure iron rods which had been further purified by being given three passes at 4 in. hr-1 in a standard MRC electron-beam zone-refining unit. The purity of this material. which was used in two other investigations of the properties of high-purity iron,2,3 was established by an extensive series of analyses. Four laboratories, the National Research Corp., Newton, Mass., the Naval Research Laboratory, Washington, D.C.. the Battelle Memorial Institute, Columbus, Ohio, and the National Physical Laboratory. Teddington. Middlesex, England. carried out analyses for carbon. oxygen. and nitrogen. and in addition a complete mass-spectrograph analysis was carried out at Battelle. In all, twelve analyses were carried out for carbon, seventeen for oxygen and twelve for nitrogen. The amounts of these elements found to be present were 10 * 7, 8 i 5, and 0.9 i 0.1 ppm, respectively. The limits given are rms deviations. Two results, one of 46 ppm of oxygen and another of 31 ppm of nitrogen, were rejected as anomalous. The mass-spectrograph analysis revealed the major impurity to be 25 ppm of tungsten. All of the other elements detected were present in amounts less than 0.5 ppm. Some of the rods were doped with carbon by being immersed in a suspension of carbon in oil, heated in a vacuum at 400°C, and then zone-leveled to make the distribution of carbon uniform. Carbon analyses carried out on three of the doped rods gave a mean value of 130 ± 45 ppm. Some of the doped and undoped rods were deformed 58 pct by swaging and cut into specimens 1/2 in. in length. Isothermal recrystallization experiments were then carried out on them at 575", 600°, 615", 630°, and 650°C, the volume fraction of each specimen recrystallized being measured by a point-counting method described by Hilliard and Cahn.4 The results for the two types of specimens are shown separately in Figs. 1 and 2. In Fig. 3 the pairs of recrystallization curves for each temper-

Jan 1, 1965