Search Documents

By E. P. Dahlberg, R. E. Reed-Hill

On the assumption that stress or strain relaxation occurs as the result of a thermally activated process, equations are derived relating to tensile experiments that give the strain as a function of the time under the condition of constant stress, and the stress as a function of the time for constant strain. It is demonstrated that if the strain-rate equation i = previously proPosed by Kuhlmann., is used as a starting point, then the relaxation of strain at constant stress may be expressed by the equation c = (-RT/(Y) 1tz tanh (t + is the strain capable of being relaxed at any given instant. Similarly, it is shown that the relaxation of stress at constant strain may be given by a = (-RT/B) In tanh (t + t0)/27, where a is the instantaneois value of the relaxable stress. The fact that these relationships reduce to well-known empirical equations at both large and small values of the stress Or strain is also shozcn. The present theory is shown to agree well with experimental data obtained from tensile elastic aftereffect experiments on a zirconium specimen prestrained at 77 k as to make it strongly anelastic. It is also demonstrated that elastic aftereffect data obtained using torsional specimens ?,Lay agree reasonably well with the equation derived for the case of tension. RELAXATION experiments are often employed as a means of studying metallic deformation mechanisms.' The simplest and most commonly employed techniques involve stress relaxation at constant strain and strain relaxation at constant stress. In general, however, investigations of this nature have been seriously handicapped in the past by a lack of suitable equations giving the time dependence of the relaxing variable over an interval that extends from small strains up into the region where internal-friction experiments become strain-amplitude dependent. This paper presents a derivation of such a set of equations for the case where the time-dependent part of the strain is anelastic or recoverable and the specimens are loaded in simple tension. The relaxation of strain under the condition of constant stress will be considered first. Let us assume that strain relaxation occurs as the result of a reversible thermally activated process that occurs at a number of relaxation centers lying in an elastic matrix. Then, following Kuhlmann,2 we may express the rate of strain relaxation as follows: where C is the strain rate, AFx the free energy of activation of the process controlling strain relaxation, a, the effective or average resolved stress at the relaxation centers, u an activation volume, R the universal gas constant, T the absolute temperature, > a factor with dimensions of a volume that accounts for the strain contribution of a successful operation of a unit process, N the number of relaxation centers per unit volume, and v the Debye frequency. The first term on the right of Eq. [I] represents a strain rate in the direction favored by the stress, while the second term represents the rate in the opposite direction. It is implied in Eq. [I.] that both F and v are symmetrical with respect to the two basic directions of operation of a relaxation process. Eq. [I] may also be written where and S and Q are the activation entropy and activation energy, respectively, of the relaxation process. In the following, A will be considered a constant. This is compatible with a set of experimental conditions where the relaxation rate is controlled by a single basic reversible process in which it may be assumed that the temperature dependence of the product ?Nv is negligible in comparison with the temperature variation of the exponential term. It is also implied that v, 7, and N do not depend strongly on a, . In deriving a relationship for the strain as a function of the time from her equation, equivalent to Eq. [2], Kuhlmann2 chose to consider only the limiting cases where the time was either very small or very large. It will now be shown that it is possible to integrate Eq. [2] to obtain a single equation valid over a wide range of strains if the concept of relaxable strain is introduced. The use of this quantity, which is the difference between the instantaneous value of the strain and the value of the strain at complete relaxation, represents the primary point of departure of the present theory from that of earlier workers. Let us express the effective stress at the relaxation centers in terms of strain. For this purpose we may use the following equation derived by zener3 for the case of strain relaxation at slip bands: where M? and M are the relaxed and unrelaxed moduli respectively, go the applied constant tensile stress, and m an average orientation factor that takes account of the fact that a,, the effective stress at the relaxation center, may not be a tensile stress (i.e., a

Jan 1, 1967

By Charles L. Mantell, George Ferment

This investigation determined the jeasibility of current-potential curves as an analytical tool for monitoving manganese electroicinning solutions for metallic impurities. Nine metallic impurities were studied: trickel, cobalt, silver, copper, zinc, molybdenum, catlmiun magnesium, and sodium. The individual effect of each metallic irzplrl,iLy on the standard point-izatiorr was determined for a range of 'concentration. Not all of Ihe impurities affected the polarization curve; correlation between transition current and cotrcetztration of impurity. Molybdenum had an uncorrelated effect on the polarization curie. The effects of binary mixtures of impurities were studied to see if the transition current of the mixture could be predicted. Interactions occurred between impurities which prei'etzted the pvecliction the transition current of tile rrlisLitr.e by additive laws. The polarization curve was shown to be sensitive to metallic impurities which affect the current efficiency of a manganese cell. THE application of current-potential curves in manganese electrowinning is not new but played an important part in the conversion of an inoperative large-scale pilot plant into a continuous comnercial operation.1"" The effects of trace metallic impurities in manganese electrowinning solutions on cathode current-potential curves,5-19 specifically nine impurities (nickel, cobalt, silver, copper, zinc, molybdenum, cadmium, magnesium, and sodium) were studied for a range of concentrations. Cadmium, magnesium, and sodium had no effect. Nickel, cobalt, silver, copper, and zinc yielded a correlation between transition current and concentration of impurity: molybdenum had an uncorrelated effect. Binary mixtures of impurities were studied to de-termine if the transition current could be predicted from the values for the individual constituents. Interactions between impurities prevented the prediction of the transition current of the mixtures by an additive law. The polarization curve was shown to be sensitive to metallic impurities which affected the rurrent efficiency of a manganese cell. The electrowinning of manganese relies upon maintaining a high-purity electrolyte. Purification schemes were developed to remove all of the heavy metals, the magnesium and calcium. Concentration limits have been established for single impurities. With high-purity electrolytes, favorable cell efficiencies are consistently obtained; but with impure solutions there is a reduction in efficiency. With the rapid rise to commercial prominence of electrochemical processes, there has been a lag in developing rapid methods for determining whether an electrolyte is sufficiently pure for efficient electrolysis. The most sensitive test for impurities is the behavior of the electrochemical cell. Impure electrolytes will cause a serious reduction in cell efficiency, and in many cases metal deposition will be prevented or reversed. The effects of impurities are not usually evident until after long periods of electrolysis, resulting in a disruption of the process. A method which could predict the behavior of an electrolyte in terms of cell efficiency prior to reaching the commercial cell is needed. Van Arsdale and Maier," Allmand and Campbell,a1'22 and others attempted to develop an electrochemical process for manganese,"3"29 but all met with some degree of difficulty. Jacobs et a studied the effects of metallic impurities in manganese electrowinning. Plant methods were based on a 2-hr plating in a small electrolytic cell. The final method was based on a 24-hr plating run, performed in a relatively large test cell. Here, current efficiencies were calculated with and without impurities in the electrolyte, and the critical concentration was established as the maximum amount of impurity that the electrolyte could tolerate before there was a serious loss in efficiency. The amount of impurity that could be tolerated decreased as the plating time increased, indicating deposition of the impurity. This test duplicated the conditions of commercial practire, and was sensitive to impurities. Small amounts of impurities significantly affect the hydrogen overvoltage at a metal surface, increasing or decreasing it depending upon the nature of the impurity. ockris established the limit for the onset of poisoning at 10-l moles per liter. The effects of impurities were studied by potential-time curves at a constant current density. The technique did not lend itself to rapid analysis in that several hours were needed to establish stable potential-time curves. Serfass and coworkers3' developed calorimetric and spectrophotometric techniques for analyzing for trace impurities. Our method4'" is based on current-cathode potential curves produced by continuously changing the cell voltage at a programmed rate. Starting with a clean stainless-steel cathode, the cell voltage was initially adjusted to yield zero current: the program was then started and the current was plotted as a function of the cathode potential. Fig. 1 is a typical current-potential curve. In initial region A the hydrogen-evolution reaction predominates and the polarization curve

Jan 1, 1967

By G. S. Baker, S. H. Carpenter

Langenecker has reported significant reduction in strength of high-strength metals during ultrasonic t'ibration. Decreases in static yield stress an order of magnitude greater than the calculated stress amplitude of vibrations were reported by him for a stainless steel and for tungsten. in the present work, attempts were nzadc to verify such large reductions in yield and flats stresses using resonant librations of measured amplitudes. The results show that both the soft metals (copper and aluminum) and the high-strength metals (sleels) hare reductions in yield and .flow stresses no larger than the stress amplitude of libration. BLAHA and Langenecker,' investigating the effect of vibratory stresses on the deformation of zinc and aluminum, found a decrease in the static yield and flow stresses approximately equal to the vibratory stress imposed. Langenecker later extending this work to high-strength metals such as stainless steels and tungsten2 found a much larger effect. In these metals the static yield and flow stresses decreased by an amount more than an order of magnitude greater than the calculated vibratory stresses. It was suggested that this phenomenon could be a serious hazard to high-strength structures in an acoustic environment, such as a large rocket at launch, or could be used advantageously in the working of hard-to-work materials such as the refractory metals. Langenecker and others3 have investigated the effect of high-amplitude vibrations on many metal-forming operations. The results have been quite favorable with usually a large decrease in the energy necessary for the forming operation. The processes, however, involve large frictional forces and it is not clear whether the decreases in forming forces are to be ascribed to decreased friction or to decreased flow stresses. The present paper reports an attempt to determine the effect of accurately known vibratory stresses on the static yield and flow stresses. I) EXPERIMENTAL TECHNIQUE The standard multicomponent resonator technique for measuring internal friction has been adapted to allow simultaneous high-amplitude vibration, internal-friction measurement, and tensile deformation. This has been described in detail elsewhere.4 Longitudinal oscillations of approximately 50 kc frequency are excited in a multicomponent resonator by a quartz transducer. A second transducer measures the amplitude of vibrations. Fig. 1 shows a schematic drawing of the resonator and grip arrangement and the displacement pattern produced. Since the centers of the dummy bars are at displacement nodes, they represent points where attachments will not materially affect the vibrations. The force is applied by a standard Instron tester. The sample may be deformed while simultaneously vibrating at up to 104 strain amplitude. 11) EXPERIMENTAL RESULTS The effect of high-amplitude vibrations on the yield and flow stresses of copper. aluminum, tantalum. a soft iron, and a stainless steel has been measured. In addition samples of tungsten vibrating at stress levels up to 10' dynes per sq cm have been subjected to static stresses near the static yield stress with no indication of macroscopic deformation. Fig. 2 gives typical results for 99.999 pct polycrys-talline copper. A vibratory stress of 7 x 107 dynes per sq cm is applied at points "a" and removed at points "h". The behavior of aluminum is identical. In agree-

Jan 1, 1967

By H. R. Ogden, E. S. Bartlett, A. G. Imgram

Unmodified disilicide coatirigs were applied to sheet-tensile specimens ofCb-Dg3 and Mo-TZM veJractovy- metal alloys. Coating thickness, degree of coating-substrate interdiffusion, and specimen geonzetry (notched and plain were included in the variables studied. Tensile tests were made to determine the ductile-lo-brittle transition temperature. The disilicide coating modestly increased the transition temperatlre of TZM, but had no effect on 043. Neither material condition (recrystallized or stress-velieved) nor specimen geometry (notched or unnotched) significantly altered the effects of coatings on the transilion temperatures of. the alloys. Cracks in the brittle coatings did not propagate into the substrate, and fracture modes appeared to be the same for both un-coated and coated specimens. MOST potential structural applications for refractory metals and alloys involve exposures to oxidizing environments at elevated temperatures. The general lack of oxidation resistance of these metals will require protective coatings to allow fulfillment of their potential. Currently preferred coatings for the oxidation protection of refractory metals are brittle intermetallic aluminides or silicides. These are typically formed on the surface of the refractory-metal substrate by a diffusion reaction between the substrate and a gaseous or liquid medium that is rich in aluminum or silicon. Because of the brittleness of these coatings, they will sustain no plastic deformation at low temperatures. They are frequently cracked by cooling from the coating temperature because of the thermal-expansion mismatch with the substrate alloy. Even if they survive cooling intact, they crack rather than sustain deformation under load at low temperatures. Thus, when a coated refractory metal is strained beyond the elastic limit of the coating at low temperatures, the mechanical environment of the substrate would include both static and dynamic cracks. These might be expected to influence the flow and fracture behavior of the substrate. This could be manifested in an altered fracture mode and/or an increase in the normal ductile-to-brittle transition temperature of the refractory-metal substrate. This paper presents the results of a research program that was conducted to determine the influence of the presence of a brittle surface coating on the low-strain-rate tensile behavior of typical refractory metals at low temperatures. EXPERIMENTAL PROCEDURES Material Preparation. Thirty-mil-thick sheets of molybdenum TZM alloy (Mo-0.5Ti-O.1Zr) and colum-bium D43 alloy (Cb-IOW-1Zr-O.1C) were obtained commercially. These alloys were selected as substrate materials representing two classes of materials important in current refractory-metal technology. The TZM was in the stress-relieved condition, and exhibited a heavily fibered grain structure. The D43 had been processed by the duPont "optimum" fabrication schedule,' and exhibited slightly elongated grains typical of this process. Tensile specimens of two geometries were prepared from these materials: 1) plain specimens with 0.2-in.-wide 1.0-in.-long gage sections; 2) specimens similar to above, but with a 0.06-in.-diam hole drilled in the center of the gage section, providing a stress concentration factor, Kt, of 2.5. The "notch" geometry was selected to represent a typical condition of a rivet hole or other geometric discontinuities as might be encountered in various applications. Machined specimens were degreased, with a final rinse in acetone, prior to the application of coatings. Specimens of each substrate and configuration were pack-siliconizedin a particulate mixture of 80 pct A1203, 17 pct Si, and 3 pct NaF. Specimens were embedded in this mix (contained in graphite retorts) and coated in an electrically heated argon-atmosphere furnace under time-temperature conditions to effect nominal 1- and 3-mil-thick silicide coatings: Coating Thickness, mils Thermal Treatment 0.6 to 1.4 24 hr at 982°C 2.4 to 3.2 48 hr at 1093°C Coating kinetics were similar for both the TZM and D43 substrates. These treatments had little or no visible effect on the substrate microstructure as determined by optical metallography. The coatings on TZM were essentially single-phase unmodified disilicides, while those on D43 showed substantial evidence of modification by proportionate reaction with the respective substrate elements or phases, as shown in Fig. 1. It was recognized that these coatings might not be particularly desirable regarding protective capability. However, it was desired to circumvent possible inter -ferring chemical interaction with the substrate by pack additives such as chromium, titanium, boron, aluminum, and other elements that typify the better protective coatings for these materials.' Thus, the results presented apply specifically to the simple silicide coatings investigated. They may not be rep-

Jan 1, 1967

By O. R. Bergmann

Two approaches were taken to obtain direct experimental evidence jar the existence oj a jet in the explosion-cladding pvocess: 1) direct observation of the claddirg jwocess by rtzeans ojf kig-h-speed photography using a fvartziiy catnera and 2) use of a plated-surface experiment in contribution with clzetrzical analysis. The framing-camera records showed clearly the presence of a jet it1 the form of what appears to be a line spray. The jet velocity is considerably highev than the detonation velocity arid in vacuum it is almost twice as high as the detonatiou velocily. The results jrorrz the plated-surface experiment also corzfivtned the preseizce of- a jet. The artzoutlt o,f etzergy dissipated in the jetting process was foundd lo be consislent with the basic theovetical model o/ the cladding mechanism.. MaNY reports on metal working with explosives have appeared in the literature during the last few yearsThis paper describes experiments concern- ing the mechanism of the explosive metal-cladding process. It has become increasingly important as a unique method for bonding metals and a typical cladding arrangement is shown in Fig. 1. A theoretical analysis of the mechanism of explosive bonding and a description of the nature and types of bond zones obtained was given by Cowan and oltzman." A further discussion of the basic principles of the process, a detailed description of the different bond-zone types, and extensive information on the properties and the applications of explosively bonded metals was recently published by Holtzman and cowan. " The process is now in commercial use by E. I. du Pont de Nemours & Co. The concept of the existence of a jet during metal bonding with explosives forms a central part in the theoretical analysis of the cladding mechanism described by Cowan and Holtzman When a layer of explosive is detonated on a metal plate that is separated from a second metal plate by a small distance the metal plate in contact with the detonating explosive is accelerated by the detonation pressure. An angle is established between the undeflected and deflected portions of the plate. Subsequently the deflected portion of the first plate collides with the second metal plate.

Jan 1, 1967

By K. G. Davis

Dilute alloys of silzer in tilz haz been solidzjzed a1 controlled rules, in lzorizontal tubes 2 mrz and 0.5 rn in diam. Radioactive; tracer techniques we?-e used to study solute distvihutions along llze specimens and irmpurity substructures, as functions of the conditions growth. The results have heen mzalyzed using Burton, Prim, and Slichter's tnodel for the solute disty-ihution in the liquid ahead of an advancing solid-liquid interface. DECANTING techniques have been used to establish criteria for the formation of impurity substructure during solidification in several alloy systems An analysis equating the appearance of impurity substructure with the incidence of constitutional supercooling was developed,= in which diffusion was assumed to be the predominant mode of solute mixing in the liquid. Substitution of experimental data into the equations derived from the diffusion analysis gave reasonable values for solute diffusivity in the liquid. However, in these investigations solidification rates of 10 cm per hr and over were used, to minimize the effect of natural conection, whereas for crystals of high perfection lower solidification rates are needed, usually between 1 and 5 cm per hr. A knowledge of the conditions under which impurity substructures will be formed at low growth rates, and the degree of segregation to be expected, is therefore of interest. In the present investigation, radioactive-tracer techniques have been used to study the effect of growth rate, initial solute concentration, and size of specimen on the formation of substructure in uni- directionally solidified rods of dilute silver in tin alloys, over a range of growth rate between 0.5 and 10 cm per hr. Criteria for the formation of substructure were established by auto radiography. This was believed preferable to the decanting technique, for the relationship of the structure on a decanted interface with that of the freezing interface is not clear, since a layer of liquid is left on the surface following decanting.' The effect of substructure on the distribution of solute along the rods was also investigated. EXPERIMENTAL The procedure for sample preparation has been described in detail by einber. Alloys were prepared from tin of 99.999 pct purity and silver of 99.99 pct purity. A"', an emitter of both y and 0 radiation, was used as tracer. Samples of the alloys were drawn up into silica tubes with inside diameters 0.5 or 2 mm, remelted in a horizontal furnace, and solidified at controlled rates. A few centimeters of the samples were left unmelted for a measure of the initial concentration. Total sample length was approximately 20 cm. The temperature gradient in the furnace was kept constant at 36°C per cm in the region corresponding to the melting point for tin. To find how well this corresponded to the actual temperature gradient in the melt, a thermocouple of 38-gage wire was electrically insulated with a thin layer of Sauereisen cement and placed within one of the 2-mm-diam tubes, which was then filled with tin. The maximum diameter of the coated wire was 0.5 mm, the bead of the thermocouple junction being no larger than the diameter of the wire. A plot of temperature vs distance traveled by the furnace is given in Fig. 1 for a rate of furnace travel of 1 cm per hr. The temperature gradient in the liquid

Jan 1, 1967

By T. F. Bower, M. C. Flemings, H. D. Brody

A series of experiments are reported which show thai reasonable assumptions for analysis of solute redistribution in solidification of castings and ingols illclr(de: negligible undercooling before nuclcation of solid phases, negligible increase of solute in advance of the tips of growing dendrites, complete diffusion within the Liquid oer distances the order, of dentrite arm spacings, platelike dentrite t/o?$holog?.. Qzmli/ali,e and periment is obtained JOY Al-Cu alloys front measurements of amount of nonequilibrium second phase, minimum solute content at the centers dentrites arms (after solidification), and 3) change in minimum solute content in dentrite arms during solidification. In a previous paper' analyses were given describing solute redistribution in dendritic solidification. The analyses were undertaken primarily to permit prediction of "severity of microsegregation" in a casting or ingot after solidification and cooling to room temperature, but can also be used to predict "local fraction solid' as a function of temperature in the solidifying alloy. Numerical results were given for Al-Cu alloys. The aims of this paper are to examine the validity of some of the assumptions of that analysis, and to compare predictions of the theory with experiments. All experiments reported are on Al-Cu alloys, the bulk on Al-4.5 pct Cu alloy. EXPERIMENTS Experiments conducted were on small ingots, 41/2 in. in diam by about 10 in. high. solidified in a controlled thermal environment (i.e., in a resistance-heated, re-circulating air furnace, whose temperature could be controlled to 1°C). In one series of experiments, the ingots were solidified very slowly (solidification time was varied from 8 to 1000 hr) and at constant rate of heat extraction. In a second series of experiments the ingots were solidified ( 'unidirectionally". Fig. 1 is a sketch of the method of making the two different types of ingots. The unidirectionally solidified ingots were fully columnar, and the slowly solidified ingots had equiaxed structures, Fig. 2. In making the slowly solidified ingots, the mold employed was carbon steel, coated with inert refractory. Typically the mold and its environment were heated to 700°C. The molten alloy was then poured into the mold and the temperature of the system equilibrated above the liquidus. The environment of the mold was next cooled with a controller responding to a differential thermocouple that measured the difference between the temperature of the melt and the temperature of the mold environment. The differential thermocouple consisted of one chromel-alumel junction in the melt connected in opposition with a second junction in the furnace. The potential across the differential thermocouple was input to the controller which acted to keep this potential at a set. constant value; hence the temperature difference between the two junctions was maintained nearly constant. Major resistance to heat flow in the above apparatus was between the metal surface and the furnace atmosphere, not in the metal itself. By maintaining a constant temperature difference between the metal and the furnace atmosphere, the controller maintained an essentially constant rate of heat extraction during solidification, regardless of amount of heat released by the metal per unit temperature change. When the metal temperature dropped just below the eutectic temperature. the ingot was removed from the furnace and water-quenched. The unidirectionally solidified ingots were cast in a similar mold to the foregoing except that provision was made to extract heat through the bottom face of the mold by passing air or water through the mold base, Fig. 1. For these experiments, metal and mold temperature were again equilibrated above the liquidus. Temperature of the mold environment was then held

Jan 1, 1967

By W. O. Philbrook, R. H. Spitzer, F. S. Manning

A generalized mathematical model has been developed to describe the kinetics of gas-solid reactions with special attention given to the hydrogen reduction oj dense spheres of hematite. This reduction process must pvoceed through a sequence of transport and chemical .reaction steps acting in series. Under typical experimental conditions, none oj the individual transport and reaction resistances may be neglected in the development of a physically significant model. A linear vate of thickening oj the product layer, usually taken as evidence of interface-controlled .reuction, may be observed even though transport resistances play a dominant role in the over-all kinetics. The temperature and pressure dependencies of all terms have been carefully evaluated in order to .make fill use oj arailable expe-inzental data in evaluation oj the model. FOR obvious technological reasons, the kinetics of reduction of iron oxides has become the subject of intensive experimental and theoretical investigation during recent years. Continuing efforts to construct mathematical models to describe the kinetics of reaction between a gas stream and a granular bed of solid particles have demonstrated the necessity of a thorough understanding of the reaction kinetics of a single particle within the system as a prerequisite to the construction of physically significant models of complex, reacting systems. This paper presents a critical examination of the transport and chemical-reaction processes involved in single-particle oxide reduction and develops the mathematical expressions necessary to evaluate their importance. Comparison of theoretical predictions with experimental results contributes important new understanding of the complex nature of this process. The reduction process has been studied under a wide range of experimental variables, and it is not surprising that apparent conflicts in observation and interpretation appear in the literature. The interested reader is referred to the reviews of Themelis and auvin"? and Manning and philbrook3 for a description of much of this work. The course of reaction commonly has been followed by measuring the oxygen weight loss from the particle as a function of time. A number of authors have suc- cessfully correlated their data in terms of a model which assumes that the reduction process is controlled by the rate of reaction at a single, sharply defined, advancing interface. According to this model, the interface will advance linearly with time. A mathematical model for gaseous reduction in a packed bed developed by the present authors in an earlier paper4 recognized that reduction must proceed through the series hematite (Fe203), magnetite (Fe30,), wiistite (Fe,O), and finally metallic iron. The reduction scheme permitted a stepwise topochemical reduction of the individual particles to metal or to the lowest oxide thermodynamically possible for the local gas composition in the bed. The limiting situations of theoretically dense and very porous oxide structures were considered. On the basis of available experimental evidence, the simplifying assumption was made in each case that all product layers would thicken linearly with time. A generalized single-particle model has now been developed to determine the significance of transport resistances in the product layers and in the gas film surrounding the particle. The analysis shows that the interplay of transport and chemical resistances is such that an apparently linearly advancing interface may be observed despite the fact that the reaction may be under significant transport control. Attention thus far has been focused on the development of a model for the topochemical behavior of dense spheres of hematite in atmospheres capable of reducing them completely to metal. The formation of successive product layers of iron, Fe,O, and Fe304 is assumed to follow the mechanism originally described by dstrijm.' All oxygen removal from the core occurs at the iron/Fe,O interface, intermediate oxide layers being propagated by sol id-state diffusion without change in the oxygen density of the core. Since fractional reduction is operationally defined in terms of oxygen weight loss, the kinetics of the over-all reduction of FeZ03 to iron is properly studied in terms of the gas-solid reaction occurring at the iron-wiistite interface. The fractional reduction of a partially reduced oxide sphere is given by: here refers to the radial position of the iron/Fe,O interface. This model is not universally applicable because, as explained in the earlier paper,4 a consideration of porous particles or of dense particles in atmospheres not capable of reducing wiistite must allow for the possibility of gaseous reduction at interfaces other than the one between iron and Fe,O. However, evidence that certain hematite specimens actually do follow the dense-pellet mechanism developed here is provided by the data of cewan' and of Bogdandy and

Jan 1, 1967

By Richard L. Ashbrook, John F. Wallace

Several high-temperature eutectics of cobalt and nickel alloys were modified by small additions of selected elements. Thes-e alloys were compared to unmodified melts for microstructural variations. A few modified compositions were selected on the basis of structure and their tensile and stress-rupture behavior determined. The results of stress-rupture tests in air at 1800 F and room-temperature tensile tests of four high-temperature eutectics showed modification capable of producing substantial improvements in properties. The microstructures were altered and mechanical properties improved by: a.) dispersing a subordinate phase more uniformly throughout the matrix; b) replacing large continuous -network or platelike constituents with small equiaxed discontinuous phases; and c) by forming new, finely divided phases. tHE term "modification" is most commonly used to describe the refinement in microstructure of the Al-Si eutectic that occurs as a result of rapid cooling from the melt or on slow cooling after treatment with small amounts of sodium.1 "Modification" has also been used to describe the change from flake to spheroidal graphite in the austenite-graphite eutectic of cast iron.2 This change of graphite shape is accomplished by small additions of cerium or magnesium.3 Other systems have also been shown to be capable of modification, for example: Pb-Sn by copper;4 Al-Cu by sodium;5 Al-Mn by sodium;5 and Pb-Sb by aluminum.5 Although modification of Al-Si alloys and cast irons is usually accompanied by improved mechanical properties, a general definition of modification has been suggested to include the effect of small quantities of impurity elements on the microstructures of all eutectic alloys, regardless of the effect on the mechanical properties.6 In this paper, the term "modification" is employed in a broad sense to include the formation of third phases or even the supplanting of one of the binary eutectic phases by a third phase. The mechanism of modification, as for any solidification phenomenon, must be considered primarily in terms of nucleation and growth. Although both nuclea-tion and growth mechanisms are required to explain all the phenomena observed in connection with the solidification of modified eutectics, various investigators have usually favored one or the other.7"16 It has been shown that modification of a binary eutectic can be effected by addition of a solute element that has greatly different distribution coefficients or k values in the two phases of the eutectic. Under these conditions, differences in the temperatures at the solid-liquid interfaces of the two phases retard the growth of one phase with respect to the other.4 It has also been shown that increased undercooling favors a transition from a rodlike eutectic to a globular eutectic.17 A mechanism proposed for the transition from lamellar to rodlike eutectics depends on the concept of constitutional super cooling.18 The modification of eutectic alloys may be of interest for high-temperature service as well as at room temperature. Although "high-temperature eutectic alloys" may appear to be a contradiction in terms, many eutectic systems, including those studied in this work, have melting points in excess of 2400°F. One of the mechanisms by which high-temperature alloys are strengthened is dispersion of a relatively brittle phase such as an oxide, carbide, or an inter-metallic compound throughout a ductile matrix. For a given volume fraction of dispersed phase, strength is increased by decreasing the size and spacing of the dispersed particles.19'20 Furthermore, room-temperature ductility may also be increased by changing the shape of a brittle dispersed phase from platelike to equiaxed or spheroidal.8 Overaging and redissolving of the secondary phase limit the maximum temperature to which precipitation-hardened alloys can be used. However, in a binary eutectic system, both phases coexist up to the melting point. Although all eutectic structures are not completely stable,21 a dispersion-hardened cast high-temperature alloy might result if a fine distribution of one of the eutectic phases could be obtained during solidification. An additional reason for interest in eutectic alloys is their excellent castability.22 Since eutectic alloys, like pure metals, freeze at a single temperature rather than over a range of temperatures, their castability is superior to solid-solution alloys with long liquidus to solidus ranges. The purpose of this research was to study the effects of small solute additions on the microstructure and mechanical properties of a number of high-melting-point eutectic compositions.

Jan 1, 1967

By M. M. Rao, P. G. Winchell, R. J. Russell

A detailed analysis has been used to correlate the avalable thermodynami c da In on iron-ric11 Fe-Ni (41loys in the body-centered a, and the face-centered y phases. The inp ut injovrt~ation required by the a am lysis consists of tile heat of the a, to y transformation at 1123°K, the heat of mixing of y at 1123°K, tile specific heats of both a, and y (decomposed into Debye, electronic, and magnelic contributions), and the positional and magnetic entropies of both a, and y. PYLrrzary oirt.bul of the analysis is the free energy of each phase as a Junction of temperature and composition. FOr Fe-Ni alloys the adjustment of twenty-one parunletevs allows the accurate reproduction of enthalpy data for both phases, the heat of mixing- for y, the a/y equilibrinm compositions in the Fe-Ni phase diagram, and the usually accepted enthalpy and entropy of a and y iron. Also obtaitzed are fuirly accurate reproductions of electronic specific heat coefficients predicted from electron- to-atom ratio interpolations and To temperatuves eslimanted from transformation studies. Also considered in the calculation are the elastic constants, Curie temperatures, saturation magnlelizalion values, and indications of short- range order. The measnred activity coefficient of iron in anstetnite could not be accurately matched. The major source of uncertainty in the model is thought to be the magnetic contribution to the entropy of austenite. Calcrclations were extended to Fe-ni-C alloys using available carbon-acliuity data. The free energy of transformation from martensile to r at fixed composition was obtained as a function of nickel and carbon contetzt, and binary thetal-carbon phase diagrams Were estimaled as a function of nickel contenl. We needed thermodynamic information for Fe-Ni and Fe-Ni-C alloys at temperatures between 400" and 800°K in order to interpret the results of several phase-transformation studies. The most recent theoretical analysis of the thermodynamics of Fe-Nialloysl was based on a lumped-parameter, regular-solution model and used the Fe-Ni phase diagram available at that time.' Because significant difficulties were encountered in attempting to extend this analysis to Fe-Ni-C alloys, and because more experimental data has become available recently, we decided to attempt a new correlation. The analysis used here allows the extrapolation of measured and calculated thermodynamic values to lower temperatures and the extension of the thermodynamic analysis to Fe-Ni-C alloys. The structure of this correlation consists of expressions containing a total of twenty-one adjustable parameters. The forms of these expressions are discussed in the next section; however, a less formal summary is presented now to provide a physical overview of the model and to indicate its connections to prior work. Primary input information for this correlation is described below. The measured values of the specific heat of iron3 and its analysis475 and theoretical estimates of the specific heat of y* iron at low tempera- loys at 1123°K has been measured by solution calori-metry.7 The enthalpies of both a, and y have been measured most recently by Scheil and Saftig,B who used a dry-ice calorimeter. These data are used to formulate expressions for the heat of mixing of y and the heat of transformation of a, to y both at 1123°K. The a and y solubility limits in the Fe-Ni phase diagram have been redetermined recently9 by methods which assure a close approach to equilibrium. The a to y transformation can occur without change in nickel content or free energy at a temperature called TO; TO temperatures for various nickel contents have been estimated from transformation studies. The computed free-energy functions are adjusted by means of the parameters to fit both the phase diagram and the TO temperatures. Primary emphasis is placed on the phase diagram fit. Iron activities have been measured in Fe-Ni austenite10 between 973" and 1173°K. These activity data, which cannot be fitted accurately with our model, indicate that iron is substantially ideal in iron-rich austenite. Secondary, but helpful, information for the present correlation is provided by neutron diffraction and saturation magnetization measurements on both body-centered and face-centered Fe-Ni alloys."-'3 These results were interpreted11 in terms of composition-insensitive, localized magnetic moments on both iron

Jan 1, 1968

By M. S. Seltzer

High temperature steady-state creep rates have been determined jor lead sulfide single crystals whose defect concentrations were fixed by equilibration under controlled sulfur pressure. The activation energy of creep, Qc , for undoped lead sulfide, is a function 01 composilion, decreasing from 2.9 to 2.2 ev as the carrier concentration increases from 9 x 1016 to 8 1018 electrons cm-3 for lead excess crystals, and increasing from 1.7 to 2.3 ev as the concentration of holes increases from 4 x 1017 to 1 x 1018 cm-3 in sulfur excess material. For five different composilions, Qc for undoped and bismuth-doped crystals has been found to be equal to Qpb + Qs, the self-diffusion activation energies for lead and sulfur in PbS. It is well-established1 that high-temperature steady-state creep of pure metals is a diffusion-controlled process, where the activation energy of creep is equal, within experimental error, to the activation energy of self-diffusion. The relationship between the activation energy for steady-state creep and those for self-diffusion in binary compounds is more complicated, however, because two different diffusing species must be considered, and becayse the activation energies may be influenced by compositional variations and by foreign atom additions.2-4 In some cases experimentally determined creep activation energies are correlated with those for self-diffusion of the slower-moving species, but such agreement is by no means universal. Interpretation of experimental results is difficult because few investigations include measurements of creep rates and self-diffusion coefficients of both diffusing species on single crystals of similar composition and foreign atom concentration. The objective of this investigation, therefore, is to determine the influence of stoichiometric defects and foreign atoms on steady-state creep and self-diffusion for a well-characterized binary compound. Lead sulfide, a compound semiconductor which crystallizes in the NaC1-type structure, and whose homogeneity range includes compositions on either side of the 1:l atom ratio, was chosen for study in this investigation. The pressure-temperature-composition relations for PbS are available,5,3 and deviations from stoichiometry may be monitored by electrical measurements.7 Single-crystal specimens are available to allow observation of effects due to bulk phenomena without interference from grain-boundary effects. These crystals have been employed to determine steady-state creep rates under compressive stress for "pure" and bismuth-doped PbS as a function of sulfur pressure and temperature as well as crystal orientation and applied stress in order to establish the relationships between defect concentrations and creep rates. EXPERIMENTAL PROCEDURES Crystal Growth. Lead sulfide crystals, prepared by the Bridgman method, are grown in evacuated quartz crucibles, filled with almost equal atomic proportions of high-purity (99.999 pct) lead and sulfur, and heated to a temperature above the melting point of PbS (about 1127°C) so that the lead and sulfur react to form molten PbS. The crucible and melt are dropped through the hot zone of the crystal-growing furnace at the rate of 1 cm per hr. The maximum melting point of PbS is slightly on the lead-rich side of the 50:50 composition; hence crystals grown from a near stoichiometric melt are usually n-type semiconductors containing excess lead atoms. Typically the lead excess concentration in solid solution with the PbS is 3 x 1018 atoms cm-3. Crystals containing either 1/20 or 1/2 mole pct Bi2S3 were prepared by addition of the foreign atoms prior to crystal synthesis. Preparation of Specimens of Controlled Stoichiom-etry. Since it is difficult to grow crystals of desired composition, the process of crystal growth and control of stoichiometry are separated into two distinct operations. Equilibration of as-grown PbS to a desired composition is most easily accomplished by first annealing the n-type crystals under high sulfur pressure (about 1 atm) at temperatures above 700°C. This process serves to put excess lead precipitates into solution, yielding 0 -type crystals upon quenching to room temperature. Inter diffusion occurs readily in crystals so treated, and the methods described by Bloem may then be employed to obtain crystals of controlled stoi-chiometry. For the creep measurements made in this study, this involves heating crystals in mixtures of H2 and H2S The equilibrium constant for H2S is given by: Thus the sulfur partial pressure is obtained from knowledge of the H2/H2S ratio and Kh2S for the temperature in question. Fig. 1 shows the p-T-x diagram for PbS. It can be seen that, within the stability limits for single-phase lead sulfide, desired compositions may be obtained by equilibrating under a specified sulfur pressure and temperature. Characterization of Crystals. The principal methods for evaluating PbS crystals include measurement of electrical properties and mass-spectrographic analysis. The measured electrical parameters include Hall coefficient, conductivity type, and electrical resistivity Bloem7 has shown that there is a 1:l relation between the defect concentration and carrier concentration obtained. Hence, if the contribution to the carrier con-

Jan 1, 1968

By A. W. H. Morris, E. A. Steigerwald

The jatzgue beizazriov oj Mietal matrix co))zpos~tes in tension -lensiom loading has been imestigated as a junction of Dolume fraclion veinfovcelnent using the model systems of a silver matrix reinforced with contlnuous or dzscontztinuous aligned tungsten jilamenls and continuous aligned steel filamenls. Appreciable impvocetnents in fatigue strength were obseved in all syslems inzvesLigated, the degree of sl renglhenng-increasing with increasing volume fraclion of reinforcement. The mode of fatigue failure was found to be a function of colutrre jp-actiotz and was also noted to be controlled by the relative fatigue and strength characlerislics of the matrix and reinforcemenl. THE advantages to be gained by suitably incorporating high-strength filaments or whiskers into metallic matrices are well-recognized.' To date emphasis has been placed on the experimental determination of tensile strengths and elastic moduli of metal matrix composites measured parallel to unidirectionally aligned filaments. Several attempts have been made to relate experimentally measured tensile properties to theoretical values calculated from the properties of the components on the basis of the relatively simple "rule of mixtures" analysis. However, relatively few data have been reported in the literature on the fatigue behavior and mechanism of fatigue failure of metal matrix composites. In a study of fatigue crack propagation in aluminum plus steel wires and aluminum plus tungsten wires, Forsyth et a1.2 found that the incorporation of small numbers of filaments suitably dispersed were capable of substantially reducing the rate of fatigue crack propagation. For this purpose, the strength of the filaments appeared to be more important than the aspect ratio and the authors reported that parallel filaments were not the most favorable orientation for increased strength. In a basic study of the fiber reinforcement of metals, Williams and 0'brien3 report preliminary results which indicate a considerable improvement in fatigue strength in reversed bending fatigue of a steel wire-reinforced aluminum alloy composite. In a study of the influence of interfacial bonding on fatigue behavior, Baker4 has also reported improved fatigue properties in stainless steel-reinforced aluminum composites. In contrast, Ham and place5 report that, although tensile properties are improved greatly by filament reinforcement, the reinforcement of copper with continuous, brittle tungsten wires up to 23 vol pct was comparatively ineffective against fatigue. The authors attribute the poor fatigue properties of the composite to fatigue hardening of the matrix at the tips of cracks which can build up stress concentrations sufficiently large to fracture proximate filaments. Baker and cratchley6 failed to find marked improvements in the reversed-bending fatigue properties of silica-reinforced aluminum. The study was complicated by the complex behavior in a composite during the compres-sive half-cycles imposed in this type of loading. This was manifested in the failure of the composites by de-lamination, a contributing factor being the presence of aluminum oxide at the matrix-filament interface. Prior to the general acceptance of filament-rein-

Jan 1, 1968

By D. A. Woodford

In a constant-load creep test, if the density of nlobile dislocations is assumed constant for strains exceeding that corvesponding to the minimum creep rate, it is shown that the creep rate may be approximated by the equation where E is the nominal strain, o0. is the initial stress, and a' the stress fov unit dislocation velocity. The constant C contains the mean density of mobile dislocations and n is a measure of the stress sensitivity of dislocation celocity. This equation is shoum to be applicable to published data on an alloy steel and on nickel. The degree of plastic instability and hence the creep ductility is dependent on the calue of n for these materials. When instability becomes gross and local macroscopic necking occurs, the assumptions are no longer valid. On the basis of this equation, and the strain-time relation derived from it, a method of predicting the creep behavior at different stresses is suggested. MOst analyses of creep data have been concerned with time-dependent transient creep or time-independ- ent secondary creep in constant-stress tests. The far more widely used constant-load test has been neglected in terms of a mechanistic approach. In general, since the strains involved are usually small, the same laws for transient creep may be applied to both types of tests. However, there is no steady-state creep in a constant-load test since the stress increases continuously. Any apparent steady state observed on a strain-time plot is either fortuitous or illusory, and is generally revealed as a continuously varying creep rate on a carefully constructed strain rate vs time plot.* Transient creep may be understood in terms of a simple differential equation representing the change in mobile dislocation density with time:' where p is the density of mobile dislocations. The first term on the right is related to first-order kinetics of dislocation multiDlication and the second to second-order kinetics of exhaustion. On the basis of this equation, several authors2-4 have found good agreement with experimental transient creep curves for a number of materials. When k1p = k2p2, a steady-state creep

Jan 1, 1968

By D. G. Swarthout, J. W. Savage, R. W. Guard

The Nb-N constitution diagvam has been examined from 1100° to 1450°C at 1 atm N2 pressure. On the basis of- our work plus that of Brauer and coworkers the stable nitvide Phases are: a) a solid solutlion, basic niobium (bcc) with nitrogen contained interstitially; NbN,, x = <0.40; b) ß NhNx,x = 0.39 to 0.45; hexagonal (W2C type) (NbzN); a =3.058A,c = 4.961 + 0.001A c) y NbNx x = 0.75 to 0.80; tetragonal (distorted NaCl type) (Nb4N3); a = 4.386A,c = 4.332 + 0.001A; d) 6 NbN,, x = 0.88 to 0.91; cubic (NaCl type); a = 4.3885 + 0.0007A; e) E NbNx x = 0.92 to 1.00; hexagonal (Tip type); a = 2.958A, c = 11.272 * 0.0015A. In addition, a metastuble transition phase 6&apos; has been observed and occurs only during the transition 6 — E. 6&apos; NbN has these distinguishing parameters: NbNx,x = 0.96; hexagonal (anti-NiAs type); a = 2.94A, c = 5 .46A. The high-temperature phase fields in the system are governed by two reactions: a) 6 + Y + E, a entectoid reaction occurring at 1270" ± 20° C; b) o + N2 = E, a 2 012 1 reaction occurring at 1370" ± 20° C. As part of a study of the electronic low-temperature. properties of transition metal nitrides, a number of specimens of niobium nitride were prepared. Preliminary observations on these specimens showed several points of disagreement with data already in the literature. It was, therefore, decided to determine the constitution diagram for the Nb-N system between 1100" and 1450°C. The objective of this research was to define the temperature and composition regions of stability of the various phases and to explain the differences among previously proposed diagrams. Review of Previous Literature. The research on the Nb-N system to 1964 has been summarized by storms.&apos; Although he does not present a phase diagram, his review covers the research on the system thoroughly. One of the earliest works on NbN is that of Ascher-mann and coworkers2 who studied the superconductivity of the cubic 6 NbN. Subsequent work by Horn and ziegler,3 Cook, Zemansky, and Boorse,1 Rögener:5 Schroder6 and sellmaier7 has qualitatively confirmed the first results on superconductivity, although their methods of preparation differed considerably. The constitution diagrams proposed for the system by Brauer and Esselborn8 and Elliott9 are shown in Fig. l(a)and (b). The diagram in l(a) has been modified by more recent results in Brauer&apos;s laboratory (private communication). Although 6&apos; is shown on this diagram, Brauer has indicated that he now believes 6&apos; to be a metastable phase in the temperature range shown. The diagram of Elliott, Fig. l(b), is based primarily on the results of Elliott and Komjathy10 on ni-trided niobium strip or rod specimens examined by X-ray diffraction and metallography. It will be noted that these proposed diagrams are in conflict with respect to the temperature and composition ranges of

Jan 1, 1968

By E. T. Turkdogan, J. H. Swisher

bi the fivst part of the paper results ave given on the rate of decarburization of Fe-C melts ln CO2-CO atmospheres at 1580°C. The rate -controlling step is believed to he that irvlloluing dissociation of curbotz dioxide on the suvfuce of the melt. 4 genevral reaction mechanistm is poslnlated jor gels-t11eta1 veactions oc-curit~g on the surface of iron coutcotamncited with chemi-sovbed osygesL. Oxygen the present work on decavbuvization of liquid iron and previous studies on the kinetics of nitrogen absorption and desorplion are discussed in terms of the postulated mechanism, ManY of the early studies of rate of decarburization of liquid steel were of an exploratory nature and laboratory exppriments carried out pertained to open-hearth or oxygen steelmaking processes. References to previous work on this subject may be found in a literature survey made by Ward. Using more sophisticated experimental techniques, several investigators have recently studied the kinetics of decarburization of molten Fe-C alloys in oxygen-bearing gases. For example, Baker et al2.&apos; reported their findings on the rate of decarburization of liquid iron, levitated by an electromagnetic field, in carbon dioxide-carbon monoxide-helium atmospheres. In these levitation experiments the samples used were small in size, e.g., -0.6-cm-diam spheres weighing -0.7 g, and the rates were measured for decarburization from about 5 to 1 pct C at 1660°C. The rates obtained under their experimental conditions were considered to be controlled primarily by gaseous diffusion through the boundary layer at the surface of the levitated melt. Parlee and coworkers3 measured the rate of absorption of carbon monoxide in liquid iron. The rates were found to follow first-order reaction kinetics, yielding a reaction velocity or a mass transfer coefficient in the range 0.2 to 0.4 cm per min. The coefficient was found to decrease with increasing carbon content of the melt. These investigators attributed the observed rates to the transfer of carbon or oxygen through the diffusion boundary layer adjacent to the surface of the melt. In the work to be reported in this paper, an attempt has been made to study the kinetics of gas-metal surface reactions involved in the decarburization of liquid iron. EXPERIMENTAL The experiments consisted of melting 80-g samples from an Fe-1 pct C master alloy in an induction furnace and decarburizing in controlled CO2-CO mixtures at 1 atm pressure and 1580°C. The master alloy was prepared by adding graphite to electrolytic "Plastiron" melted in racuo. None of the impurities in the master alloy exceeded 0.005 pct. The reacting gases were dried by passage through columns of anhydrone; in addition, CO2 impurity in carbon monoxide was removed by passage through a column of ascarite. A schematic diagram of the apparatus is shown in Fig. 1. A 1.25-in.-diam recrys-tallized alumina crucible containing the sample was placed inside a 3-in.-diam quartz reaction tube, all of which was surrounded by an induction coil. A 450-kcps induction generator was used as the power source. Water-cooled brass flanges, which contained the gas inlet, gas exit, and sight port, were sealed to the top of the reaction tube with epoxy resin. The reacting gases were metered with capillary flowmeters and passed through a platinum wire-wound alumina preheating tube, 0.25 in. ID and 11 in. long. The gases were preheated to about 1300°C. A disappearing-filament optical pyrometer was used to measure the melt temperature. The pyrometer was initially calibrated against a Pt-6 pct Rh/Pt-30 pct Rh thermocouple. The temperature was controlled to within +10°C by manually adjusting the power input to the induction coil. In a typical experiment, an 80-g sample of the master alloy was melted in a CO2-CO atmosphere having pcO2/pco = 0.02 and flowing at 1 liter per min. A negligible amount of carbon was lost and no significant reduction of alumina from the crucible occurred during melting, e.g., 0.005 pct Al in the metal. After reaching the experimental temperature of 1580°C, the gas composition was changed to that desired for a particular series of decarburization experiments. The duration of the transient period for obtaining the desired gas composition at the surface of the melt was about 20 sec . The flow rate of the reacting gas was maintained at 1 liter per min. After a predetermined reaction time, the power to the furnace was turned off. During freezing, which took about 10 sec, the amount of gas evolution was not sufficient to result in a significant loss of carbon. The samples were analyzed for carbon by combustion and in a few cases they were analyzed for oxygen by the vacuum-fusion method. RESULTS A marked increase in the rate of decarburization of iron with increasing pcO2/pco ratio in the gas stream is demonstrated by the experimental results given in Figs. 2 and 3 for pco2/pco ratios from 0.033 to 4.0. In one series of experiments, denoted by filled triangles in Fig. 2, the reacting gas was diluted with argon (48 vol pct) resulting in a slower rate of decarburization. Samples from two series of experiments with pco2/pco = 0.033 and pco2/pco = 0.10 (with argon dilufion) were analyzed for oxygen. In these Samples the oxygen content increased with reaction time

Jan 1, 1968

By S. G. Epstein

Using the capillary-reservoir technique, electromi-gvation rates of cadmium and indium in liquid bismuth were measured at several temperatures. The electric mobility of cadmium Jrom 305° to 535°C and indium from 310° to 595°C can be expressed as a function of temperature by the equations UIn = 1.52 x 10-3 exp sq caz per v-sec Migraion of both solutes was cathode-divected at a rate rnore than four tiMes tHAt previously found for siluer in liquid bisnmth. The electric mobilities of cadmium and indiulrz in liquid bismuth at 500° C are nearly identical with their respective mobilities in mercury at room temperature. AS part of a systematic study of the variables which are considered to control electromigration in liquid metals, the electromigration rates of cadmium and indium in liquid bismuth have been measured. Mass transport properties of silver in liquid bismuth have been reported previously,&apos; and measurements of tin and antimony in liquid bismuth are forthcoming. Comparisons will be made with literature values for these same solutes in mercury.2&apos;3 This series of solutes was selected to determine the effect of the solute valence on its electromigration. Silver, cadmium, indium, tin, and antimony have nearly equal atomic masses but have chemical valences ranging from +1 to +5. They are all fairly soluble in bismuth above 300°C and all have radioactive isotopes, which are an aid in making analyses. EXPERIMENTAL TECHNIQUE Electromigration of cadmium and indium in liquid bismuth was measured by the modified capillary-reservoir technique previously described.&apos; In this method irradiated cadmium or indium is added to bismuth to form alloys containing about 1 wt pct solute (<2 at. pct solute). Several quartz or Pyrex capillaries: 1 mm ID and 5 cm long, vertically oriented, are simultaneously filled in the reservoir of the liquid alloy. A direct current is passed through two of the capillaries, which contain tungsten electrodes sealed in the upper end. The other capillaries sample the reservoir during the experiment. After a measured time interval the capillaries are removed from the reservoir and rapidly cooled. The glass is then broken away from the solidi- fied alloy, which is then weighed, dissolved in acid, and analyzed for solute content by chemical and radiochem-ical techniques. An electric mobility (velocity per unit field) can be calculated from the amount of solute entering or leaving each capillary by the simplified expression1 in which Ui is the electric mobility of the solute, ?mi the solute weight change, Ci the solute concentration of the reservoir, I the current, p the alloy resistivity, and l the duration of the experiment. This expression is valid as long as the experiment is terminated before a concentration gradient develops across the capillary orifice. Earlier experiments showed that the concentration gradient formed initially at the electrode changes with time and eventually reaches the orifice, due to back-diffusion. This condition produces a solute exchange between capillary and reservoir by diffusion or convection, opposing the electromigration, which results in a lower measured value for the electric mobility. To determine if the concentration gradient had reached the orifice, the capillaries used in some of the experiments were sectioned at 1-cm intervals and the solute content of the alloy from each section was radiochemically determined. A typical concentration profile for an experiment with indium in bismuth is shown in Fig. 1; cadmium in bismuth showed similar behavior. As illustrated in the graph, very little back-diffusion has occurred in the capillary containing the cathode, since the concentration gradient is confined to the upper 1 cm of the capillary. In the capillary containing the anode, however, the concentration gradient is much broader, extending nearly to the orifice, even though the net change in solute concentration is nearlv the same in both capillaries. Since cadmium and indium probably lower the density of bismuth when alloyed, depletion of the solute from the alloy adjacent to the anode would increase the density of the liquid in the uppermost region of the capillary. This would give rise to convective mixing within the capillary, causing the broadened concentration gradient. Conversely, the alloy adjacent to the cathode should have a reduced density as the solute concentration is increased by migration, explaining the "normal" concentration profiles found in these capillaries. This disparity was not found for electromigration of silver in bismuth. Both metals have similar densities at the operating temperatures, and nearly symmetrical concentration profiles were found in the two capillaries of each exueriment. This density effect was also apparently encountered when an attempt was made to measure diffusion coefficients for indium in liquid bismuth by the same technique which was successfully used to measure diffusion of silver in bismuth.&apos; Capillaries 1 mm ID and 2 cm

Jan 1, 1968

By R. Taggart, D. H. Polonis, J. C. McMillan

Thin foil electron microscopy techniques have been used to study the micros tructure of a Ti- 7.5 wt pct vanadium alloy in the quenched and aged cond_itions. Selected area diffraction has identified a (1011) twinning mode accompanying the martens ite t9.an.sfo?rn~ation. Precipitation of the 0 phase during the aging of quenched alloys has been obsevced to occuv in the form of plates and needles with preferential growth along the (1120) direction in the a&apos; matrix. In the present work, the micro structures of quenched and aged alloys of the 0 isomorphous Ti-V system have been investigated by means of thin-foil electron-microscopy techniques. Most of the previous studies of martensitic phases in titanium-base alloys have utilized optical microscopy and X-ray diffraction techniques. The application of transmission electron microscopy is more recent, and only a few such studies have been reported.1-5 This paper is concerned particularly with the crystallographic nature of substructures formed in the martensitic a&apos; plates. The early stages of aging or tempering have also been examined to determine the modes of precipitation from the martensitic solid solutions. Such information is essential in studying strengthening mechanisms and in determining the feasibility of high-strength rnartensitic alloys based on titanium. Liu and Margolin6 reported several types of substructures within the martensitic plates of quenched Ti-Mn alloys. It was later suggested by Liu7 that the structure of martensitic plates might be very complex and not single crystals as normally assumed. The sub-markings could not be identified with the two possible habit-plane orientations {334}ß and {344}ß, and it was not positively determined whether they are associated with mechanical twinning. The recent studies by Armitage5 on titanium martensites in a variety of commercial alloys have indicated that twinning does not generally accompany the transformation. This observation is in accordance with the Mackenzie and Bowles model for the transformation in pure titanium8 which suggests that twinning is not to be expected. However, a number of other recent investigators have reported twinlike substructures in the martensitic plates of several titanium systems including commercia1 titanium,2,3 Ti-Cu,5,9 Ti-Ta,1 and the commercial alloy Ti-8A1-1Mo-1V. Blackburn* has reported that the structure of quenched Ti-8A1-1Mo-1V depends on the prior solution temperature. He identified two types of martensitic structures, one of which is described as a heavily faulted hcp or twinned fcc structure and the other an hcp structure with a high internal dislocation density. The latter type exhibited some twinning with a (1011) twin plane. EXPERIMENTAL PROCEDURES AND RESULTS The alloy used in the present work was a 7.5 wt pct V alloy that was prepared by conventional arc-melting techniques. This alloy composition was selected

Jan 1, 1968

By R. L. Stegman, M. R. Achter

Nickel, fatigued in vacuum, shows intragranular crack initiation and growth at low temperatures. An increase in temperature initially prolongs fatigue life due to a more homogeneous distribution of slip. With a further temperature increase, the mode of failure undergoes a transitiotz to intergranular crack initiation with concurrent recrystallization which results in a shorter life. The transition temperature varies with bending strain and specimen purity. UNTIL recently, very little work has been done on the effect of temperature on the mechanism of fatigue and the resultant variation in fatigue life. Except for mild steel, where strain aging raises the strength at moderate temperature, fatigue lives had been shown to decrease with an increase in temperature. In their review article, Allen and Forrest&apos; summarized the work of McCammon and Rosenberg and of Forrest and Mann on the fatigue of copper from -269" to 325°C. The fatigue life vs stress curves remained roughly parallel over this temperature range, the fatigue strength decreasing progressively with increasing temperature. When testing was done in vacuum, the present authors&apos; found the results for nickel were different from those quoted above. Initially, the fatigue lives increased with temperature and then decreased as a grain boundary mode of failure was observed. The initial increase was attributed to a more homogeneous distribution of slip, a condition which was not as conducive to crack initiation as the bunching of slip lines observed at 20°C. Ronay, Reimann, and wood3 observed a similar change of slip line distribution in brass, tested in air, with rising temperature. Our previous observations were based on preliminary work with relatively few specimens, fatigued at comparatively low strains. To learn more about the mechanisms, the present investigation was undertaken for the purpose of studying systematically this effect over a wider range of strains and as a function of purity of the nickel. EXPERIMENTAL PROCEDURES The testing equipment, procedures, and specimen geometry have been described elsewhere.&apos; An optical technique to measure bending strains was used as de-

Jan 1, 1968

By Daniel N. Beshers, Ralph H. Condit

An expression for the diffusion coefficient of inlerstitial atoms in the bcc lattice is derived which takes account of jumps between tetrahedral and octahedral interstices. The results are applied to the case of carbon diffusion in iron to find values for jump frequencies of the types, octahedral-tetrahedral, tetra-hedral-tetrahedral, and tetrahedral-octahedral. ThERE has been renewed discussion1-5 concerning the mechanism of diffusion of interstitial atoms in the bcc lattice. This lattice has both octahedral and tetrahedral interstices, and it has been pointed out that the interstitial atoms may make jumps of the octahedral-octahedral type, hereafter called o-o, octahedral-tetrahedral, o-t, and tetrahedral-tetrahedral, t-t. Doubt has been cast on the applicability of snoek&apos;s8 model for internal friction to some interstitial migrations in bcc lattices wherein he proposed 0-0 jumps as the principal mechanism of migration. Also, the anomalous excess diffusion of carbon in iron at high temperatures4 has been ascribed to t-t jumps,2,3 even though the sites normally occupied are octahedral. In this paper an expression is derived for the diffusion coefficient, taking account of the three types of jump mentioned. The relationship of the coefficient to the corresponding internal friction relaxation time is found for the limiting case corresponding to Snoek&apos;s model. These results are applied to the case of carbon in iron. In Fig. 1, the structure is shown with the locations of the interstices. Each o site has four t site neighbors coplanar with it, and each t site lies halfway between two o sites. There are four t site first neighbors of a t site lying at the vertices of a tetrahedron. Let the average fraction of atoms occupying o sites and t sites be No and Nt, respectively. Since there are six o and twelve t interstices per unit cell having a lattice parameter, a, the concentration of interstitial atoms will It will be assumed that three types of jumps may be distinguished, t — t, t — o, o — t,for which the jump

Jan 1, 1968

By William C. Sleppy

The nmgnetic susceptibility of heat-treatable aluminuin alloys is sensitive to chanyes such as solution or dissolution of solute and the precipitation of mew phases. By measuring the change in the magnetic susceptibility of aluminum alloys caused by various heat treatments, an empirical relation was found from which atomic arrangements in dilute binary alloys of copper, zinc, and magnesiutn in aluminum have been delineated. The relation predicts the ultimate formation of C1LA12 when copper is precipitated from solid solution in aluminum. Euidexce joy silovt- range order is found for copper in solid solution in aluminum in the sense that copper atoms avoid being nearest neighbors to an extent greater than would result from a purely random arrangertzeizt. Hume-Rothery has predicted such short-range order joy solid solution of copper in aluminum The Al-Zn system agrees with evidence obtained from X-ray scattering at small angles and predicts a tendency for zinc atoms to cluster in solid solution in aluminum. In the Al-mg system, the empirical relation indicates an approach to randor distribution of magnesium in solid solution in aluminum with a tendency for magnesium segvegation which increases with incveasing temperature. ThE magnetic properties of metals are complicated by the fact that contributions are made to them both by electrons of a "metallic" type which belong to the crystal as a whole, and by electrons in states localized on particular atoms. An expression1&apos;2 for the bulk magnetic susceptibility of aluminum may be written as the sum of three contributions: where XA1 is the bulk susceptibility of aluminum per gram of material (in the cgs system, the units are those of reciprocal density); Xa1+3 is the diamagnetic contribution of the electrons localized in ion cores; Xa1 is. the paramagnetic spin contribution of conduction electrons often called Pauli paramag-netism: Xa1 is the diamagnetic contribution of the conduction electrons often called Landau diamag-netism. Ion core diamagnetism arises from the precession of the electron orbits which occurs when a magnetic field is applied to a system of electrons moving about a nucleus. Its contribution to the magnetic suscepti- bility is small, temperature-independent, and unaffected by alloying. The conduction electron diamagnetism is also temperature-independent and arises from the translatory motion of the electrons. For perfectly free electrons this contribution should be exactly one-third of the Pauli spin paramagnetism, but this relation is seldom even approximately true. Blythe2 determined the conduction electron diamagnetism in pure aluminum and found it to be extremely small. Any change in the conduction electron diamagnetism caused by alloying is neglected in this work. The Pauli paramagnetic contribution3 to the magnetic susceptibility of aluminum depends upon the number of electrons that occupy excited states and whose spins can be turned parallel to an applied magnetic field. The number of electrons free to turn in the field is proportional to the temperature and each spin contribution to the susceptibility is inversely proportional to the temperature. A slight temperature dependence of Pauli paramagnetism occurs when the number of electrons occupying excited states cannot increase sufficiently to balance the inverse dependence on temperature of each spin contribution. The decrease of the magnetic susceptibility of aluminum with increasing temperature is attributed to a temperature dependence of the Pauli paramagnetism. Estimates of the Pauli paramagnetism of aluminum have been made by several workers.2,4,5 All of the values are in reasonably good agreement with each other. In this work Xal at 17°C is taken as 0.761 X 10-8 cu cm per g. An expression similar to [I] can be written for the magnetic susceptibility of an aluminum base alloy containing a fractional weight percent x of solute:&apos; Xa = (1 -x)XAl+3 +xXsoluteion * XaPauli +Xadia) [2] where X, is the magnetic susceptibility per gram of alloy, Xal&apos;3and Xsolute ion are the ion core diamag-netic contributions, and xpauli and xdia are the Pauli and diamagnetic contributions of conduction electrons in the alloy. If the components of a mixture are not alloyed but simply mixed together in their pure states without producing a new phase, then the magnetic susceptibility of the mixture is given by the Wiedemann additivity law: Xm =x1X1 +x2x2 + ..xnxp [3] where X, is the susceptibility per gram of mixture and xnXp are the weight fractions and susceptibilities, respectively,-. for the pure components. The additivity law is not applicable to alloys because the outer electronic structures of the components are changed by alloying.&apos; Both the Pauli paramagnetism and Landau diamagnetism are affected; hence the magnetic susceptibilitv of an alloy is usually different from that calculated using the additivity law. In this work the difference, X, -X,, is taken as a measure of the change caused by alloying.

Jan 1, 1968