Search Documents

By R. W. Armstrong, D. H. Feisel

Self-diffusion in pure crystalline solids has been described through extension of the Cohen and Tum-hull volume -fluctuation model originally proposed for diffusion in simple liquids. It is shown, for a number of crystalline solids, that the self-diffusion coefficient, D, obeys the relationship where a is the lattice spacing, v the Debye frequency, Z' the lattice-volume fluctuation per atom required for a diffusion jump, and (?V/Vo) the relative increase in the lattice volume resulting from thermal expansion between O°K and the temperature under consideration. The calculated value of Z' is dependent upon the difiusion mechanism and the lattice geometry. ThE need for an increased understanding of mass-transport processes in crystalline solids has resulted in much research to determine rates of self-diffusion.' For several metals (e.g., copper, silver, and iron), the experimental values of the self-diffusion coefficient, D, have been determined over a large interval of temperature and the accuracy of D has been investigated through repeated experiments. Also, a number of empirical and theoretical diffusion studies have established a rational basis for understanding the D values for the different elements.2"6 The empirical diffusion analysis by Sherby and simnad6 supplied the main stimulus for the present study. Sherby and Simnad, reviewing the self-diffusion data for a substantial number of pure crystalline solids, suggested that sufficient data were available to establish with some certainty the apparent intrinsic material properties which influence D. Without explicit regard for any mechanism of diffusion, these investigators obtained an empirical relationship for D which correlated the self- diffusion coefficient for each element with the melting temperature, crystal structure, and valence of the particular element. Their analysis is only unsatisfactory in that the effect on D of crystal structure is not very clear and the valence to be assigned to an element is not always straightforward. The present study is an attempt, from a very different viewpoint from that of Sherby and Simnad, to establish a basis for understanding the self-diffusion coefficients for crystalline solids. This analysis was undertaken for the purpose of developing a general expression for D which would explicitly depend on the mechanism of diffusion. The main result of the study is that diffusion in crystalline solids may be described in terms of a volume-fluctuation model which was originally proposed by Cohen and Turnbull7 to describe diffusion in simple liquids. The model appears to be useful for predicting self-diffusion coefficients in crystals. The following sections describe the volume-fluctuation model for diffusion, the application of this model to crystals, and the resultant agreement with experimental data. 1) THE VOLUME-FLUCTUATION MODEL FOR DIFFUSION Cohen and Turnbull7'8 developed their model for liquid self-diffusivity for the purpose of describing certain properties of liquids, glasses, and the liquid-glass transition. A hard-sphere model for the liquid structure is used. In this model, the size of a spherical molecule or atom is fixed for all temperatures by the interatomic spacing at the absolute zero of temperature. At any temperature above O°K, the atom is viewed as being contained within a larger spherical cage defined by the interatomic spacing of the atoms at that temperature. A random distribution of varying cage sizes is assumed to exist in the liquid. An individual diffusion jump occurs when the free volume of an atom (the volume of the cage containing the atom minus the volume of the atom) is large enough for the atom to jump out of its cage. With this physical picture of diffusion in liquids, Cohen and Turnbull derive the following equation for self-diffusivity: D=Doe-yv*/vf [1] where Do is the product of three factors which are the diffusion-path geometry, the diffusion-jump distance, and the velocity of the diffusing atom, v*

Jan 1, 1964

By M. Herman, G. J. London

A series of dilute Fe-Co alloy specimens were prepared from zone-melted iron. The iron used was given ten floating-zone passes in purified hydrogen. Alloying was accomplished by plating a gradient of metallic cobalt onto a 1/4-in.-diam bar of material from a pure plating bath. The deposit was radioac-tively tagged by a small addition of co80 to the plating bath. Alloy homogenization was accomplished by passing an additional floating-zone pass through the plated section of the bar. The final concentration gradient achieved was from residual to about 1000 ppm (0.1 pct) added Co. After homogenization, the bar was swaged to 0.057 in. wire and cut to tensile specimen blanks 1-1/16 in. long. Each specimen was counted in a well-type scintillation counter. The concentration of cobalt was computed by combining the normalized specimen counting rate with the known weight (measured to the nearest 0.01 mg) of cobalt added during plating. Fig. 1 shows the cobalt gradient obtained, exclusive of residual cobalt which was estimated by neutron-activation analysis to be approximately 50 ppm. Results of recrystallization for 1 hr at 750°, 700°, 650°, and 600°C with specimens containing 40, 170, 470, and 870 ppm added Co are shown in Figs. 2, 3, 4, and 5. Recrystallization was accomplished in a dynamic vacuum of less than 1 x 10 torr. The specimens were mounted and polished parallel to the swaging direction. The recrystallization re- sults show a radical coarsening of structure with increasing cobalt concentration. The coarsest structure obtained was at the 470-ppm level (at all recrystallization temperatures) with a slight grain refinement at the 870-ppm level shown. A more normal solid-solution effect is seen in Fig. 5 (60OoC, 1 hr) where the only specimen completely recrystallized is the one containing 47 ppm Co. Since the possibility existed that some atmosphere-contamination phenomenon might be responsible for the observed results the recrystallization series was repeated using sealed quartz capsules. These were filled with purified argon which was gettered during heat treatment on metallic titanium included in the capsules. The results obtained were essentially the same as shown. Another identical alloy was made using a 5-mm-diam bar of Westinghouse Puron. Specimen preparation and recrystallization treatment were also as described. The results obtained (see Table I) were somewhat erratic. Some coarsening in structure does appear at the highest cobalt concentrations. There is no apparent relationship between the coarse, columnar, directionally solidified structure of the as-melted bars and any of the recrystallized grain structures observed, all of which are much finer than the former structure. A recrystallization process can be most simply described as one of homogeneous nucleation of recrystallized grains in a cold-worked matrix followed by more or less uniform growth of the strain free grain into the cold-worked material.' An impurity element may influence one or both of these processes. Aust and utter' have shown that impurities strongly influence grain boundary mobility in lead. Moreover, definite recrystallization textures are produced by small concentrations of tin,3 while additions of silver or gold4 produce large boundary in-

Jan 1, 1964

By D. H. Polonis, M. B. Kasen

AN unusual thermal etching phenomenon has been observed on the surface of a Cu-Si alloy. The observations were made during studies of vacancy condensation pit formation which were reported in previous papers.1,2 Figs. 1 and 2 are photomicrographs of the surface of a Cu-4.8 pct Si alloy after heating in vacuo (3 x X mm Hg) to 525°C and subsequently holding at that temperature for 16 hr in an impure (air-contaminated) argon atmosphere. Under these conditions preferential removal of silicon from the matrix by oxidation results in a surface which is entirely a phase overlying an a plus y structure. The light areas of Fig. 1 are regions in which very little thermal etching has occurred, while the majority of the surface has experienced severe etching. Similar regions in Fig. 2 are seen as plateaus above contours produced by net transfer of material from the specimen surface. Twin boundaries are also visible in Fig. 2. The difference in matrix mi-crostructure between Figs. l and 2 reflects the par- ticular orientation of low-index planes with respect to the surface of observation. Thermal etching contours of this type have been studied in detail by Hondros and Moore Both of the photomicrographs show that one or more dark spots are located approximately at the center of every protected area; where two spots are in close proximity, the protected areas have merged. The microstructures indicate that the anomalous resistance to thermal etching is related to the presence of these spots. Williams and Hayfield have shown that local regions of unusually high oxidation resistance can be caused by a surface monolayer of atoms with a work function higher than that of the base material. The result is an inhibition of the rate of electron removal from the base material and a decrease in the oxidation rate. The source of the protected layer may be either surface diffusion from contaminants on the specimen surface or pipe diffusion of solute atoms followed by migration across the surface. In the latter case the driving force is a difference in chemical potential resulting in a Gibbs adsorption phenomenon. In the present case, chemical polishing of the specimen surface precluded contaminants of a strictly surface nature. Consequently, diffusion of impurity atoms from within the lattice appears to be the most tenable explanation of the observed phenomena. The authors propose that the structures of Figs. 1 and 2 reflect the diffusion of impurity atoms along extended dislocations within the individual a phase grains prior to surface oxidation. The central spots within the protected regions are seen as sites of dislocation emergence which have become locally decorated with oxide in the manner discussed in Ref. 1. Williams and Hayfield observed protected regions adjacent to the grain boundaries of iron-contaminated copper which they attributed to the segregation of iron to the grain boundaries coupled with high dif-fusivity along the complex dislocation array at such boundaries. No evidence of such grain boundary protection was observed in the Cu-Si alloy; this

Jan 1, 1964

By E. T. Wessel

UNEXPECTED brittle failures of metals in practical applications are a serious problem to many industries and to the nation as a whole. Considerable effort has been devoted to studies of the brittle behavior in metals. However, the major portion of this effort has been directed at developing techniques and apparatus for the purposes of determining the susceptibility of metals to brittle failure and of making various types of observations of the phenomena. Relatively minor effort has been made to understand the basic mechanisms involved in initiating brittle failures. A more complete understanding of the nature and mechanism of the brittle failure problem would be very helpful in efforts directed toward predicting and avoiding brittle failure and toward the development of suitable materials and designs that are not susceptible to a brittle behavior. The intent of the present investigation is to acquire this basic understanding. The pronounced transition from a ductile-to-brittle behavior is most prevalent in metals having the body-centered-cubic lattice structure, primarily be- cause metals of this structure exhibit a strong dependence of yield strength upon temperature and the rate of straining. The generally recognized qualitative concept of the ductile-to-brittle transition is that brittle (cleavage) failure occurs when the yield strength exceeds the cleavage fracture strength. This condition can apparently be achieved by decreasing temperature or increasing strain rate. Likewise, a normally ductile metal can be made to fracture in an apparently brittle manner by the introduction of triaxial stresses (notches) where the ratio of normal to shear stresses can be increased to such an extent that the normal stress exceeds the fracture strength before the critical shear stress for plastic flow is achieved. Neither of these concepts require any prior plastic deformation or explain why the observed fracture strengths are so much less than the theoretical cohesive strength. The results of several investigations of the ductile-to-brittle transition in body-centered-cubic metals illustrate rather conclusively that measurable amounts of nonelastic deformation do precede cleavage fracture, even at extremely low temperatures. These studies suggested that the early stages of plastic flow, particularly abrupt yielding, and their relationship to brittle fracture should be examined

Jan 1, 1958

By J. H. Frye, D. L. McElroy, E. E. Stansbury

The rates of growth of pearlite in high-purity Fe-C alloys have been measured as a function of the transformation temperature. These and other data have been correlated in terms of a derived rate equation. Activation energies for the growth were found to be 24,200 cal per mol of austenite for the 0.78 pct C alloy and 27,600 cal per mol for the 0.93 pct C alloy. Theoretical implications are discussed. AN excellent review of the the theoretical and experimental studies on the mechanism and kinetics of the eutectoid reaction has been presented recently by Mehl and Dube.' They have pointed out that these investigations have followed two lines of approach: 1—experimental studies and derivations of expressions for the overall transformation characteristics, and 2—derivation of expressions for the nucleation and growth rates in terms of fundamental processes. Investigations of the latter type have been concerned for the most part with the nature of the growth at the austenite-pearlite interface in plain carbon and alloy eutectoid steels. They discuss two theories of the rate of growth of pearlite from aus-tenite—those of Brandt and of Zener. Both derive rate equations which are based on the assumption that carbon diffusion is the controlling mechanism. Mehl and Dube,' Zener,2 and Brandts. " have attempted to calculate growth rates by use of these equations together with thermodynamic and diffusion data. The results are encouraging, but agreement is not so good as to constitute verification of the theories. Furthermore, it Is not clear how these equations are to be used to study the effect of alloying elements which apparently exert little influence on carbon diffusion, but markedly decrease growth rates. It is not clear whether or not the addition to iron of an alloying element which does not diffuse during transformation can alter the rate of growth of body-centered from face-centered cubic iron. That it can greatly decrease the overall transformation rate has

Jan 1, 1954

By J. E. Pavlick, A. G. Guy

The total creep of tin alloys containing antimony in solid solution was observed to decrease with increase in antimony content. However, near the solubility limit an anomalous maximum in deformation was. found, in agreement with similiar observations by Bochvar. IT has long been recognized that alloys may deform more rapidly under stress if there is a tendency for a phase change or other structural change to occur simultaneously. However, the phenomena involved are complex and in some instances the opposite effect, strengthening, may result.' This area of research has not attracted much interest in the western world, and although Russian investigators have studied it for the past 15 years their work has not received due attention here. Recently diametrically opposing views on one aspect of this problem have been espoused by two Russian investigators, Kornilov and Bochvar. The question is this: in certain binary alloys tested in creep at the temperature where the solute is at the limit of solid solubility, does the incipient formation of the second phase strengthen or weaken the alloy? Korni-lov's arguments in favor of strengthening were presented at the 1954 Symposium at the National Physical Laboratory, =Creep and Fracture of Metals at High Temperatures.The most recent statement of Boch-var's theory of weakening is in Oding's new book on creep and rupture.3 Although Bochvar agrees that strengthening occurs in some alloy systems, he argues that in certain alloys an anomalous weakening is observed. However, the extensive work of Korni-lov's group on creep properties of alloy systems has not revealed any anomaly of this type, but instead has tended to refute Bochvar's theory. This clear-cut controversy offered a good opportunity for a simple experiment to resolve the problem. The experiment and its results supporting Bochvar are the subject of this paper. However, it will be helpful first to sketch the present status of the larger question of the effect of a phase change or other structural change on plastic deformation. PREVIOUS WORK One of the first researches in this area was by Pfei1,37 who investigated the creep resistance of an 80 Ni-20 Cr alloy containing varying amounts of titanium. He found that the resistance to creep increased rapidly as the solid solubility limit in the alloy at 775'c was approached and slightly exceeded. With further addition of titanium the creep resistance decreased, presumably because of precipitation effects. Bochvar's first work in this field was a study in 1945 of "superplasticity" in zinc alloys containing 15 to 25 pct Al.4 He found that alloys quenched from about 400 had anomalously low hardness when tested below this temperature. The following year, in a paper on the dependence of mechanical properties on the composition and structure of alloys,5 he extended some observations of Kurnakov to predict that at high temperatures of testing the hardness values for solid-solution alloys might be lower than those for the component metals. He also pointed out the influence of changes in solidus temperature in affecting curves of hardness vs composition, and he

Jan 1, 1962

By P. D. Southgate

VIBRATION, both ultrasonic and sonic, can affect the course of a variety of metallurgical processes. Reports of work on this subject have appeared at intervals over the last twenty years and a thorough review has been made by Hiedemann,' who gives a full list of references. It has been established that both refinement of structure and outgassing of the melt can be produced in most metals with sufficiently intense vibration. Many of the experiments in this field have been carried out using alloys of direct commercial interest in production- type casting equipment. Here a different approach is adopted, and simple binary alloys which exhibit different types of structure are used to facilitate examination of the effects produced under well-controlled conditions. Results from this kind of experiment should enable the basis of larger scale experiments with similar types of materials to be better substantiated. The alloys selected for this work are: 1) aluminum of 99.2 pct purity, the principal impurity being iron, near 0.4 pct; 2) A1-1.8 pct Fe, which is near eutectic composition; 3) A1-4 pct Fe—the second phase, of composition FeAl3, forms distinct acicular crystals; and 4) A1-12 pct Si—although this is of eutectic composition, the silicon normally appears in the form of large plates. Experimental Method There are three practicable methods by which acoustic vibration can be transferred to liquid metals: a) vibration of the whole of the container; b) electromagnetic induction; and c) insertion of a vibrating probe into the liquid. Method c is chosen here, and a frequency of 8 kc per sec is used. The end of the probe is electrically heated so that the metal near it will be the last to solidify. This ensures that vibration is transmitted directly to the liquid metal without passing through any solidifying and pasty intermediate region, which would attenuate acoustic waves severely. In addition, the heating prevents solid metal adhering to the probe, with consequent detuning. The transducer assembly is shown in Fig. 1. A stack of Permendur laminations 30 cm long acts as a half-wave vibrator and a winding along its length carries polarizing and energizing currents. It is joined by a half-wave coupling bar to a mild steel exponential horn, which tapers from 3.8 to 1.1 cm diam. A stainless steel stub 3 cm in length is welded on the free end. The assembly is supported by a plate located at the node of the coupling bar. Acoustic radiation losses are minimized by allowing the free end of the transducer to protrude above the surface of the cooling water. The exponential horn serves the function of matching the acoustic output impedance of the transducer to the radiation resistance of the probe into the melt. Electrical drive is provided by a 1 kw generator. Measurements of probe vibration amplitude during melt treatment are obtained from a gramophone pick-up with the needle attached to the wide end of the horn. This device is calibrated with a microscope before use. Liquid aluminum is very actively erosive and solution of materials with which the molten metal comes into contactcauses difficulties. Erosion of the probes in particular can be severe since it is greatly accelerated by the vibration. Tests of various acous-

Jan 1, 1958

By Y. A. Rocher, J. E. Dorn, L. A. Shepard

Creep activation energies for single aluminum crystals favorably oriented for shear by (010) [101] glide were detemined over the temperature range from 78" to 900°K. Observations of slip bands on the specimen surface were made in conjunction with the investigation. From 78" to 780°K, the activation energies obtained in this imestigation agreed closely with those previously found for creep by (111) [101] slip. Between 78" and 140°K, the activation energy was identified with the Peierls process, while between 260°and 780°K the activation energy was close to that for cross-slip. The coarse wavy slip bands nominally parallel to the (010) plane observed above 260°K were attributed to fine cross-slip. From 800" to 900°K, unusually high apparent activation energies ranging from 28,000 to 54,000 cal per mole were obtained. These apparent activation energies were attributed to re crystallization. AS illustrated in Fig. 1, a recent investigation1 has shown that creep of aluminum single crystals by the (111) [i01] mechanism is controlled by three unique processes, each of which is characterized by a single activation energy which is independent of the applied stress and the creep strain. A comparison of the observed activation energies with theoretically calculated values permits a fairly clear identification of the three operative creep processes. Below 450°K, where the activation energy for creep is 3,400 cal per mole, the deformation is controlled by the Peierls process, the activation energy for creep agreeing well with that calculated by seeger2 for the energy required to nucleate the motion of a dislocation loop against the atomic forces of the lattice. Between 590° and 750°K, the observed activation energy for creep of about 28,000 cal per mole agrees well with the energy necessary to induce cross-slip. Seeger and schoeck3 estimate that the activation energy is about 24,000 cal per mole whereas Friedel4 recently calculated this activation energy to be 28,000 cal per mole. Above 800°K the activation energy of 35,500 cal per mole that was observed for creep agrees well with that estimated for self-diffusion in aluminum.= In this range the operative rate-controlling slip process has been clearly identified as that arising from the climb of edge dislocations. The objective of this investigation is to ascertain whether a single crystal of aluminum favorably oriented for simple shear in the [loll direction on the (010) plane might exhibit uniquely different activation energies for creep from those obtained previously for (111) [101] slip. Whereas the exis- tence of such unique activation energies would constitute incontrover table evidence for new mechanisms of slip, the absence of any new activation energies might suggest that slip of aluminum is confined to the (111) [loll mechanism. Several factors prompted the selection of the (010) [101] orientation for study. First, there are more reported observations of (010) [loll slip than of any other nonoctahe-dral mechanism.8-10Secondly, Chalmers and Martius1l have concluded from considerations of the energies of dislocations that (010) slip is the second most favored mechanism in face-centered-cubic metals. Finally, favorable orientations for simple shear by the (010) [loll mechanism provide the least favored orientation for slip by the (111) [101] mechanism. EXPE-RIMENTAL PRO-CEDURE The high-purity aluminum stock, specimen preparation, shear fixture, extensometry, and experimental technique used in this investigation were the same as those previously reported.' Single-crystal spheres grown from the melt of 99.995 pct pure Al* were _ *The high-purity aluminum used in this investigation was graciously given by the Aluminum Company of America. oriented, carefully machined into dumbbell-shaped shear specimens, annealed, and chemically polished. The finished specimen had a central reduced section 0.190 in. wide and 0.590 in. in diam and 1/4-in. grip sections at both sides, 0.690 in. in diameter. The specimen was oriented in the stainless steel grips of the shear fixture with the (010) plane perpendicular to the dumbbell axis and the [loll direction parallel to the stress axis within 2 deg. Creep activation energies were calculated in the previously described manner1 from determinations of the instantaneous change in shear strain rate produced by an abrupt 15 to 20 deg increase or decrease in test temperature. If is the instantaneous strain rate at strain y and temperature T1, and ?2 the instantaneous rate at y and T2,

Jan 1, 1960

By Louis E. Toth, Alan W. Searcy

A modification of Le Claire's microscopic model for self-diffusion is developed in a form suitable for prediction of activation energies for diffusion in disordered substitutional solutions as well as in pure metals Bonding is considered as a localized interaction, and the energy of bonding between atoms of different types is taken as the arithmetic mean of the energies in the pure elements. The activation energies for vacancy formation and migration in substitutional alloys are shown to depend on the empirical constants developed for self-diffusion when the equations are adjusted for the mole fractions of the two elements. The differences in activation energies between alloys and pure metals are calculated and compared with the experimental differences. MaNY different correlations have been proposed for use in predicting activation energies for self-diffusion or diffusion in dilute alloys. Usually theories for self-diffusion cannot be applied to alloys nor can theories for dilute alloys be satisfactorily adapted to self-diffusion. We present here a correlation scheme which can approximate activation energies for diffusion, vacancy formation, and atom migration in substitutional-alloy systems of any concentration. The activation energy for self-diffusion QD has been correlated to the melting temperature Tm,1, 2 the sublimation energy L,,l the heat of fusion Lf,3 crystal structure, and valence.4 Working with mechanical parameters Le claire5 proposed that the activation energy for vacancy migration Qm could be related to appropriate shear moduli. He also proposed that the activation energy for vacancy formation 62, is proportional to Ls. THEORY We consider that Le Claire's expression for 4, is satisfactory. But an alternate expression for Qm in terms of the bulk modulus is just as satisfactory from a theoretical point of view for the prediction of the energy of vacancy migration in pure metals and can more easily be used for alloys. In order for Qm to be a minimum, not only the moving atom but also adjacent and nearby atoms must undergo distortions. These distortions are neither isotropic as is implied by a correlation with the bulk modulus nor unidirectional as implied by a correlation with a shear modulus, but are complex and multidirectional. A complete analysis in terms of the bulk modulus and all the shear moduli is not yet available. In a first approximation in which only one parameter is used, there is no obvious theoretical reason to prefer the shear modulus over the bulk modulus. The bulk modulus is a much more convenient parameter for use in predicting Qm. Shear moduli are known for less than half of the metallic elements while bulk moduli are usually available. Furthermore, because of the directional dependence of shear moduli, their estimation for alloys appears inherently more difficult than is the estimation of bulk moduli. We assume that the bulk modulus Bs reflects an average value of the shear moduli such that the strain energy at the point of maximum distortion of the lattice during vacancy migration can be expressed as

Jan 1, 1964

By H. I-Lieh Huang, O. D. Sherby, J. E. Dorn

RECENT investigations1-4 have suggested that the total plastic strain, C, for high temperature creep under a given stress can be correlated by means of a temperature-compensated time, te- where t is the duration of test, e is the base of natural logarithms, aH is the activation energy for creep, R is the gas constant, and T is the absolute temperature. A typical set of data on which this deduction was based is given in Fig. la. Inasmuch as the e-in t curves for the various temperatures are identical except for their displacement along the In t axis where f and are appropriate functions. If the increase in creep rate with an increase in temperature is exclusively ascribable to a single process involv- ing thermal activation, (T) should equal RT and therefore it = constant [2] where ?= te Assuming the validity of this relationship, ? = ?2 at the same strain, or where the subscripts refer to two alternate test conditions.

Jan 1, 1957

By B. F. Decker, D. Harker

The recrystallization reaction in OFHC and spectroscopically pure copper has been followed by X ray diffraction determinations of the amount of material with the cold-worked and recrystallized textures in specimens which had been given various heat treatments. The heats of activation for recrystallization are found to be: 29.9 kcal per rnol for OFHC copper and 22.4 kcal per rnol for spectroscopic copper. A reliable and speedy method for measuring the amount of recrystallized material in a piece of rolled metal, together with a new scheme for low temperature heat treatment, have made possible a determination of the activation energy for recrystallization in rolled copper. This method for studying recrystallization rates is different from others reported in the literature.' Oxygen free, high conductivity copper—OFHC copper—of purity 99.98 pct gave an activation energy for recrystallization of 29.9 kcal per mol with recrystallization data taken in the temperature range 208-245°C. Copper of purity 99.999 pct from the American Smelting & Refining Co. gave an activation energy for recrystallization of 22.4 kcal per mol with recrystallization data taken in the temperature range 43-135°C. The great change in activation energy for recrystallization as the small amount of impurity in the metal is decreased suggests that the motion of grain boundaries is conditioned by the impurities which must be concentrated in them. Thus the OFHC copper recrystallizes with an activation energy of the order of magnitude to be expected for the diffusion of the kind of impurities likely to be present. The high purity copper, on the other hand, recrystallizes with a much lower activation energy, in accord with the notion that the grain boundaries need not wait for the impurities to diffuse with them. Copper when strongly cold-rolled has a well-defined crystallite texture which can be described by saying that the [111] direction of the face- centered cubic crystals is aligned approximately in the cross-rolling direction and that faces of the form {110} lie approximately in the rolling plane. When such rolled copper is completely annealed the crystallite texture is quite different: the [loo] direction is in the cross-rolling direction and faces of the form (100) are in the rolling plane. At intermediate stages of annealing, the copper contains both textures. If a Geiger counter X ray diffraction spectrometer is set up so as to measure the intensity of the (200) Bragg reflection from the rolled surface of a piece of copper strip, this intensity increases from a very small value in the cold-worked state to a maximum as the specimen is annealed. The same would be true for (200) reflections from planes normal to the rolling or cross-rolling directions. In view of this fact, the amount of material in the copper which has recrystallized can be measured by comparing the intensities of one of these reflections just mentioned with the intensity from a similar specimen after complete annealing. In much the same way the intensity of (111) reflections from planes normal to the cross-rolling direction could be used to measure the amount of unrecrystallized material in the specimen. In the work to be described here the amounts of recrystallized and unrecrystallized material in partly annealed rolled copper specimens were determined in this way. Experimental Procedure Samples of oxygen free high conductivity (OFHC) copper, cold-rolled 99.7 pct to 0.002 in. thickness, were subjected to heat treatments for various chosen times at four different temperatures, and high purity copper samples, cold-rolled 98.0 pct to 0.0015 in. thickness, at six different temperatures. Each sample, after treatment, was placed in an orientation sample holder for use with a Geiger counter X ray spectrometer and a reading was taken of the intensity of the (200) reflection from planes perpendicular to

Jan 1, 1951

By R. P. Smith

THE carbon content of iron-silicon and of iron-manganese alloys at fixed activities of carbon has been reported previously.' It seemed desirable to investigate a system which allows a more extended range of alloy composition within a single phase region. The iron-nickel system is ideal in this respect, in that all alloys from 100 pct Fe to 100 pct Ni are face centered cubic ("austenitic") at 1000 °C. The experimental procedure was similar to that previously described.1,2 The alloys, in the form of pieces about 2 by 2 by 0.2 cm thick, together with a similar specimen of pure iron, were equilibrated with each of a series carbon monoxide-carbon dioxide tnixtures of fixed composition, quenched in the carburizing gas in the top of the furnace and then analyzed for carbon. The carbon content of samples containing more than 0.3 wt pct C was determined by the usual combustion method; those with less carbon

Jan 1, 1961

By R. J. Borg, C. E. Birchenall

Vapor pressures of eight compositions of Mg-Cd have been determined by the Knudsen effusion technique. Measurements are made at several temperatures for each alloy and the results interpolated to a common temperature for the calculation of the activities. It is not possible to obtain values for the partial molar enthalpies because of the narrow temperature range spanned by the measurements. The results show less precision than those obtained for pure Cd which are also reborted. These alloys are extremely susceptible to corrosion which provides a possible explanation for the lack of precision. Nevertheless, the values of the activity are in good agreement with the results of electromotive force measurements previously reported for the Mg-Cd system. MAGNESIUM and cadmium are mutually soluble in all proportions in the solid state and form a single phase which has the hexagonal close-packed structure. X-ray investigations of this system1,2 have revealed a rapid variation in the relative dimensions of the unit cell occurring at approximately 50 at. pct. The rate of change of the c/a ratio with composition increases markedly for alloys containing more than 50 at. pct Cd. The purpose of this investigation was to determine the activity of Cd in the neighborhood of 50 at. pct and to discover whether or not the thermo-dynamic properties undergo a similar rapid change. The activities are calculated from vapor pressure data obtained by the Knudsen effusion method. Unfortunately the precision of the data from which the thermodynamic values are calculated is insufficient to establish unequivocally a departure from the normal trend. The results are reported primarily because they substantiate the only previously reported activity measurements3 and because they include values for pure Cd which in contrast to those of the alloys show excellent precision. The Cd-Mg system is known to contain several ordered phases. However, this investigation was made at temperatures considerably above the transition temperatures, and the alloys may be regarded as nearly random solid solutions. EXPERIMENTAL Apparatus—The effusion cells are machined from solid tantalum rod. The lids consist of threaded, flanged rings of tantalum onto which are welded shallow cups of 4 mil molybdenum foil. The orifice, which is located at its center, is made by grinding off a small dimple produced with a sharpened punch. Such an orifice has been mounted in plastic, sectioned, and the thickness of the orifice edge measured in a microscope. The maximum thickness is 0.00016 in., and the Clausing factor is consequently equated to unity. The area of the orifice is determined by tracing its projection on the screen of the metallograph at a known magnification and measuring the area of the tracing with a planimeter. Excellent temperature control is achieved with a 4 by 18-in. alundum core, gradient-wound, furnace

Jan 1, 1960

By H. Conrad, K. Janowski, E. Stofel

In a previous paper,1 the occurrence of (0001) and (0111) twins was reported for compression tests of 60-deg-oriented sapphire rods in the temperature range from 1100° to 1300° C. Subsequent to this initial work, additional tests have been conducted, and twinning has been found to occur at temperatures as high as 1500° C. Also, the temperature and strain-rate conditions for the two types of twins and for slip have been established in more detail. The present note covers this more recent work. The compression-test data are summarized in Table I. The techniques for identifying the twins are optical examination up to X1000 and X-ray analysis.' It is evident that (0711) twins are pre-

Jan 1, 1965

By J. E. McDonald, J. G. Eberhart

A model is discussed from which the work of adhcslon .tor liquid transition metals on aluminum oxide surfaces can he calculated, A close-packed (00011 oxygen surface on A12O3 is assumed with two different types of surface sites: one type involving metal-oxygen bonds and the other man der Waals into actions. The work of adhesion is thus expressed as the sum of these two bonsding free energies. Calculated works of adhesion for nickel, titanium, chromium, and zirconium on sapphire agree well with the experimentally determined quantities. The model is extentled to the calculation of the work of adhesion and shear stress required to remove a thin metal film from a sapphire substrate and is in good agreement with experimental values. The obsewed dependence of the work of adhesion on the free energy of oxide formation of the metal is shown to also provide an interpretation of the tittle dependence of thin-film adhesion. THIS paper presents a model for the type of bonding which occurs across a metal-A12O3 interface. The model is used to explain the results of two types of experiments in which such an interface exists: 1) the adhesion of thin metal films on Al2O3 substrates and 2) the wetting of A12O3 by liquid metal drops. The adhesion of thin films to various substrates has been the subject of a variety of investigations.'-' Benjamin and weaver3 and Bowie,6 using the scratch test developed by Heavens,10 studied the adhesion of metallic films to glass substrates. Their observations for noble-metal film adhesion agree well with an adhesion model involving a van der Waals type of bonding between the film and the substrate. For films of metals whose free energy of oxide formation. ?F°f, has a negative value. Benjamin and weaver3 and Bowie6 found a time-dependent adhesion with an initial value that can be interpreted in terms of van der Waals interactions but a larger terminal value which was related to ?F°f, Karnow-sky and Estill7 deposited films on sapphire at elevated temperatures and noticed no time dependence of film adhesion but a similar correlation with ?F°f. Because of the kinetic problems associated with thin-film adhesion it is desirable to examine adhesion in an equilibrium system. The wetting behavior of liquid-metal drops on Al2O3 provides such a system. Systems of this metal-ceramic type have been studied extensively.11 Humenik and Kingeryl2 have measured the wetting of A12O3 (and other substrates) by several metals and have pointed out that the wetting ability of these metals increases with increasing values of -?F°f. It is thus seen that thin-film adhesion and metal wetting on A12O3 are both related to the tendency of the metal to react with the surface oxide ions of the Al2O3 substrate and, because of this, both phenomena should be explainable by an appropriate model for the metal-Al2O3 interfacial bonding. In the sections that follow, wetting and adhesion data on A2O3 are reviewed and a model is presented by which these phenomena can be interpreted. ANALYSIS OF WETTING EXPERIMENTS In an equilibrium system involving a liquid-metal drop on a solid Al2O3 substrate, the work of adhesion, WAD, is defined by the Dupre, equation as WAD = ?s + ?L -?sL [1] where ?s and ?L are the surface free energies of the solid substrate and the liquid drop, respectively, and ?sL is the interfacial free energy. The work of adhesion is the work required to separate a unit area of the solid-liquid interface into two surfaces. The work of adhesion is usually determined from a sessile-drop experiment in which yL and the contact angle, ?, are measured. The Young-Dupr6 equation is then used to calculate WAD Wad = ?L (1+ cos ?) [2] The literature of this subject has been examined and Table I shows work of adhesion data for various liquid metals as measured on A12O3 substrates. The standard free energy of oxide formation of the metal at the temperature of the wetting experiment, ?F°f . is also tabulated in kcal per g-atom of oxygen. The data is grouped according to the gaseous atmos-

Jan 1, 1965

By Don M. Jackson

The techniques for the growth and controlled, graded doping of silicon epitaxial overgrowth layers were established. Grading of- impurities such as arsenic or boron in arbitrarily chosen profiles oiler four orders of magnitude (from I X 10Lg to 1 X 10L5 cm) was accomplished. The graded structures were grown in a stepwise manner where in the profiles were approximated by growing many separate layers with small doping-level changes between layers. Masked-area growth of silicon through windows in a SiO2 masking film was acconiplished through the simultaneous injection of HCl gas 102th Sic14 in the epitaxial reactor. The HCl served to inhihit the growth of silicon on the Sz02 surface, and also to control the cross-sectional profile of the silicon mesas. silicon from the substrate was etched out through the windows, and then epitaxial material was used to fill the holes hack in to form planar diodes. Simullaneous HCl injection was adapted to control the cross-section profile of the hack-filled regions. Arsenic-huried regions beneath in te- In recent months, the epitaxial growth process in semiconductor materials technology has evolved into a complex system directed at both material improvement for better monolithic integrated-circuit characteristics and the development of new silicon structures for future integrated-circuit needs. The advent of high-temperature HC1 gas etching1 grated-circuit transistors reduced V(SAT) by slightly more than one third, bringing this parame-ter to that of an isolated component transistor. When arsenic was diffused into the starting sub-stvate, it was necessary to remove any excess arsenic in order to avoid excessive audodoping of the subsequent epitaxial layer. Oxide isolation of single-crystal regions in a substrate provided complete electrical isolation of integrated-circuit components, thereby eliminating parasitic capacitance which is inherent in standard integrated-circuit processing. The process involved the etching of deep mesas on silicon Wafers, the formation of an oxide over the mesa surface, and then the deposition of several mils of polycrystalline silicon over the oxide. The original substrate was lapped and polished allay until the mesas were left isolated in the polycrys-talline matrix. The Final components were fabri-cated in the single-crystal islands. Speed and power-handling capabilities of standard circuits were improved through the process. wherein work-damaged material is removed from the surface of the silicon wafer just prior to the silicon deposition step has made it possible to grow silicon films with structural quality that approaches that of the substrate. HBr used in the same volume ratio as HC1 etches silicon at the same rate and with the same results as HC1. Recent developments in epitaxial materials technology have occurred in the areas of precise doping control, buried prediffused regions between over-

Jan 1, 1965

By C. S. Barrett

The torsional after-effect in polycrystalline cadmium is interrupted by an abnormal twisting when the film is removed by etching. This is accounted for by the pile-up of dislocations beneath anodic or other oxide films during the original twisting, and the relief of the residual stresses from these by elastic and plastic processes during etching. WHEN an iron or zinc wire coated with oxide is plastically twisted and then released, it untwists according to a law of equal increments in equal intervals of log time, but if the oxide is suddenly removed by etching this normal after-effect is interrupted by a transient, and thereafter the untwisting is slower.' Etching may even reverse the strain temporarily. Related effects caused by hydrated oxide films on cadmium in creep tests have been observed by Phillips and Thompson:' definite transient strains caused by the removal of the film from single crystals of cadmium were observed; however, negligible effects were obtained with polycrystalline wires. Pronounced effects of both transient and steady nature were observed by Phillips and Thompson when anodic films 10-4 cm thick were etched off, and weak effects, transient only, occurred with hydrated oxide films 3x10 cm thick such as were formed by standing in water for 10 min. It was estimated that 10.' cm films, formed in water in 1 or 2 min, could be detected in the apparatus of Phillips and Thompson. Since abnormal after-effects were observed by the torsion method equally well with polycrystalline and monocrystalline iron and zinc wires,' it was anticipated that the torsion method would also disclose them in polycrystalline cadmium, in spite of the fact that analogous effects had not been observed in tensile creep tests on polycrystalline zinc by Pickus and Parker3 r on polycrystalline cadmium in tensile studies by Andrade and Randall." Because of its high sensitivity, particularly its sensitivity to surface conditions, the torsion method was thought to merit further study as a method of investigating thin films of various types and the action of various reagents on these films. Materials and Methods Cadmium wires were cold-drawn to a diameter of 0.039 in. from a stock that was determined spectro-scopically to contain 0.05 pct Bi, 0.005 pct Cu, 0.008 pct Pb, and 0.007 pct Si. The wires were cut to about 17/8 in. lengths and were annealed 1 hr in air at 290°C, after which the grain size was such that there were 5 to 7 grains across a diameter. The wires were then mounted as indicated in the insert in Fig. 1, one end being fastened with laboratory wax into a small diameter brass tube and the other end being waxed to a thin glass fiber which supported, on its upper end, a small galvanometer mirror. The wires were waxed with care to avoid straining or annealing, for it was found that the after-effect rate was sensitive to prior cold work in a wire. The assembly was then clamped above a beaker into which the reagent could be poured. The lower end of the wire was held in the hand while the wire was twisted 180"; it was then released immediately. Since precision timing of these manipulations was not attempted, the different wires cannot be compared as to exact rates of untwisting. The length subjected to plastic strain was 15/8 in. The surface coatings studied were those produced by the annealing in air, by anodic treatment of

Jan 1, 1954

By George Hallerman, R. J. Gray

In the customary method of studying age hardening, the process of aging is interrupted by cooling the specimen and measuring its room-temperature hardness. However, the aging process may be conveniently measured while it is occurring at an elevated temperature by a series of hardness measurements at suitable time intervals. The latter method gives data void of complications that might result from processes occurring during the quench, and the data may be more indicative of the alloy's properties at elevated temperatures. The elevated-temperature hardness tester1 used in the present study employed a dead-weight loading system with a synthetic sapphire Vickers-type in-denter mounted in a molybdenum holder. The tests were performed in a vacuum of l0 to 10 torr. Hardness impressions were made by applying a 2-kg load for 15 sec. Although the hardness impressions made in the tester at elevated temperatures were measured at room temperature, the correction for thermal contraction during cooling was insignificant and therefore neglected. The two methods of obtaining age-hardening data were compared using eighteen samples of Haynes alloy No. 25* which had been annealed for 2 hr at 2250°F (1232°C). The test surfaces of the 1 by 1 by 1/4-in. samples were ground and then polished on vibratory polishers2 prior to testing. Three groups of five samples were aged at 1400°, 1650°, and 1800°F (760°, 899", and 982°C) for 4, 16, 32, 64, and 100 hr and then water-quenched. Room-temperature hardness as a function of aging time for these samples is shown in the broken lines of Fig. 1. Only one sample for each of the three temperatures was necessary to determine the hardness of the alloy at the aging temperatures with the elevated hardness tester. The specimens were heated isothermally and hardness impressions were made after various time intervals up to 100 hr. The results of these determinations were represented by solid lines in Fig. 1. Comparison of the two sets of curves shows, as expected, the hardness values determined at the aging temperatures are lower than the room-temperature hardness of the quenched specimens and the hot-hardness values reflect a strong tempera-

Jan 1, 1964

By N. Brown, H. Green

The effect of quenching temperature and of aging temperature and time on compression stress-strain curves of ß brass was investigated. Age softening occurs at a rate which decreases with decrease of quenching temperature below the critical temperature for ordering. The activation energy for softening is 15,400 cal per mol. No matter how rapidly brass is quenched or what the temperature from which it is quenched, practically complete order is obtained at room temperature. It has been demonstrated that with respect to the specific heat,',' electrical resistivity," X-ray diffraction, and microstructure, the slowly cooled and rapidly quenched specimens are the same and that complete long range order exists in both cases. However, Smith" showed that the hardness of ß brass depended upon the cooling rate and the temperature from which brass is quenched. Smith showed that the hardness was a maximum for specimens quenched from about 445 °C and that the as-quenched structure softened at room temperature at a rate depending upon the temperature prior to quenching. This softening after quenching is in this paper referred to as age softening. The results of this investigation are qualitatively the same as the findings of Smith. However, a more detailed analysis of these phenomena has been possible since the stress-strain curve offers a more sensitive index of the state of the metal than hardness. It is the purpose of this paper to show how the mechanical properties are related to the ordering of ß brass. A 200-lb ingot made from electrolytic copper and electrolytic zinc was extruded to 1/2in. rod. Chemical analyses at frequent intervals along the rod showed that the alloy contained 51.5 pct Cu with the greatest impurity being iron, at 0.01 to 0.001 pct. To obtain a constant test material, the 1/2in. extruded rod was warm-rolled to % in. diameter at 200°C and then recrystallized to a grain size of 2.0 mm. The specimens for the compression test were % in. high and % in. in diameter. The specimen ends were polished on 600 metallographic paper and coated with Molykote for a lubricant. The compression rig consisted of an anvil resting on the machined table of the Baldwin Southwark testing machine and a plunger fixed to the movable cross

Jan 1, 1954

By G. R. Speich

The age-hardening of Fe-18 to 21 pct Ni marten-sites containing small amounts of titanium, aluminum, copper, or molybdenum has been studied by hardness measurements, transmission electron microscopy, extraction replicas, and X-ray and electron diffraction. Electyon-probe microanaly-sis of the extracted precipitates was also performed. The effects of temperature and composition on the aging kinetics have been studied in detail for Fe-20 pctNi-Ti martensites. All the martensites hardened on aging at 400" to 500 "C with a single aging peak. In the Fe-20 pct Ni-Ti martensites hardening was associated with the formation of a fine dispersion of NisTi. No precipitation was observed in an Fe-21 pct Ni-2.6 pct A1 alloy, although the hardness increased from 327 to 545 Dph during aging at 500°C. Copper was precgitated from an Fe-20 pct Ni-10.7pct Cu alloy, at least during the later stages of age-hardening. Hardening in an Fe-18 pctNi-5pct Mo alloy was associated with the precipitation of an intemzetallic compound tentatively identified as NisMo. The formation of austenite and the recovery of the defect structure of the martensite occur simultaneously with the normal precipitation processes &ring aging of these alloys. THE age-hardening of substitutional iron-base martensites is an important subject for study since age-hardening is one of the principal means of strengthening the substitutional iron-base martensites in precipitation-hardening stainless steels1 and maraging steels.a~4 In contrast to Fe-C martensites, these martensites are relatively soft and ductile in the as-quenched condition, due to the low solid-solution hardening effect of the substitutional elements as compared to carbon. Also, these martensites are hardened by aging at low temperatures in contrast to Fe-C martensites which generally soften on tempering. Titanium and aluminum, singly or in combination, as well as copper and molybdenum, have been the principal alloying additions to stainless steels to obtain an age-hardening reaction.' Ti-A1 or Co-Mo-Ti combinations have been used in the case of maraging steels.3"4 The present work has been confined to a study of the age-hardening characteristics of an Fe-18 to 21 pct Ni martensite to which small amounts of titanium, aluminum, copper, and molybdenum were added in order to study the effects of each alloying element separately. The age-hardening characteristics of these alloys were studied by hardness measurements, transmission electron microscopy, extraction replicas, and X-ray and electron diffraction. Electron-probe micro-analysis of the extracted precipitates was also performed. The types and dispersion of precipitates, the formation of austenite, and the recovery of the defect structure of the martensite all were studied. EXPERIMENTAL PROCEDURE Alloys. The compositions of the alloys studied are listed in Table I. The level of 18 to 21 pct Ni was chosen to assure complete transformation to martensite on quenching and is the same nickel level used in maraging steels. The concentrations of titanium, aluminum, copper, and molybdenum were chosen after a study of the and Fe-Ni-MO' ternary diagrams indicated how much of each alloying element could be dissolved in austenite. Alloys with 2 and 4 pct Cu were also studied, but these did not show precipitation hardening at the nickel levels studied and will not be discussed further. The alloys were prepared from electrolytic iron and high-purity* nickel, titanium, aluminum, copper, and molybdenum by vacuum melting of 20-lb heats. These were vacuum-cast into two 8-lb ingots. The ingots were hot-rolled at 1100°C to 1/2-in. plate and subsequently cut into 3/8-in. cubes. The cubes were sealed in evacuated quartz capsules and solution-treated for 16 hr at 1100°C, except for the Fe-20 pct Ni-10.7 pct Cu alloy and Fe-21 pct Ni-2.6 pct A1 alloys which were solution-treated for 16 hr at 1300" and at 1250"C, respectively. Metallographic examina-

Jan 1, 1963