Search Documents

By O. P. Arora, M. Metzger, G. R. Ramagopal

The rates of grain boundary and general corrosion were surveyed by an approximate method. Quantitative differences between their variations with the strength or cupric ion content of the acid yielded large variations in their ratio, highly selective intergranular attack being observed only under specific chemical conditions. The boundary corrosion rate increased with the copper content of the aluminum; this was attributed to an autocatalytic effect similar to the one shown for general corrosion. High- purity aluminum in hydrochloric acid exhibits striking intergranular corrosion and other structure-dependent corrosion phenomena of fundamental interest. The behavior is highly sensitive to the metallurgical and chemical conditions, the significance of many of which has not been established. The features of the intergranular-corrosion phenomenon, first described by Rohrman,l are summarized in the remainder of this paragraph. Strong attack normally occurs only at high-angle grain boundaries.' Usually 15 to 25 pctHCl is used to obtain a high rate, but this has also been obtained with weaker (7 to 15 pct) acid either by adding cupric or ferric ion to the acid3-5 or by passing an anodic current (-0.1 ma per sq cm).2,3,5,6 The two latter corrosion conditions produce strong attack of low-angle boundaries after certain heat treatments 2-4,5,7 The susceptibility of high-angle boundaries is influenced by the grain-boundary segregation of iron atoms; this increases the slow (µ per month) attack in 10 pct HCl8 but diminishes the rapid (mm per month) attack in 20 pct Hc14 (the present and most previous investigations deal with rapid attack). That segregation of elements other than iron affects intergranular corrosion has not been demonstrated, although analysis of data for iron alloys indicated that at least one other element must be considered.9 Intergranular attack occurs over a wide range of purity although it begins to become obscured above several hundred parts per million (ppm) total im- purities by the severe general corrosion which accompanies the higher iron contents and which is associated with the presence of a second phase or an inhomogeneous distribution of the dissolved iron.4,6,10 The variation of intergranular-corrosion rate with heat treatment1'3-6'"10 can be explained in some cases in terms of the distribution of the iron impurity but is not in general adequately understood. The present work dealt with two questions not previously considered in detail. The first proceeded from the view that the extent to which a special behavior is associated with the boundaries is indicated not by the intergranular corrosion rate alone but by the ratio of the intergranular to the general corrosion rate. A comparison of the two rates was therefore made for one lot of aluminum in acids of varying strength and metallic ion content. The second question was the effect of the copper content of the aluminum. For general corrosion, it has long been known that copper ions introduced into the acid by corrosion of the solid solution are reduced on the aluminum as patches of elemental copper which enhance the corrosion rate by providing cathodes of lower hydrogen overvoltage than were originally present.''-'3 Since the copper patches remain themselves unchanged, their action is "catalytic", and since they are products of corrosion, the effect is "autocatalytic." This phenomenon is significant at copper levels much lower than previously suspected, even at those typical of high-purity aluminum, as the authors have shown in a previous paper.l4 The present work shows that copper content affects intergranular corrosion. With polycrystalline specimens, truly quantitative rate measurernents are not readily made for intergranular corrosion or for general corrosion in the presence of rapid intergranular attack, and previous work in this field has tended to be qualitative. The present study made use of a simple procedure which provided fair approximations to the quantitative corrosion rates and fitted the purpose of surveying the behavior over % wide range of conditions and displaying certain large effects. The approximations involved in this work were usually outweighed by the experimental scatter, which tends to be substantial in a system as sensitive as this one. I) EXPERIMENTAL A. Material. The material procured, Table I, represented three low copper (1 to 3 ppm) lots, two of "normal" copper content (21 and 22 ppm), and two

Jan 1, 1962

By E. C. W. Perryman

The wide grooves formed at the grain boundaries when high purity aluminum is attacked by hydrochloric acid or sodium hydroxide have been attributed by earlier workers to the high energy of the grain boundary material. The effect has been investigated for high-purity AI-Fe alloys with up to 0.055 pct Fe as a function of iron content and heat treatment. It is shown that the explanation given above is untenable, but that the results can be explained on the assumption that iron segregates to the grain boundary in solid solution. IN 1934, Rohrmann¹ showed that aluminum of 99.95 pct purity suffered intercrystalline corrosion when immersed in 10 to 20 pct hydrochloric acid, and that the susceptibility to intercrystalline corrosion depended upon the heat treatment given. The greatest susceptibility was found for specimens quenched from a high temperature (600°C) and the lowest susceptibility for specimens cooled slowly from that temperature. Lacombe and Yannaquis2 have shown that super-pure aluminum (99.9986 pct) annealed at 600°C suffers intercrystalline attack in 10 pct hydrochloric acid and that this attack is intensified by anodic dissolution in the same solution at a current density of 10 milliamperes per sq dm. No difference in extent of intercrystalline attack was found between the 99.993 and 99.986 pct Al, which led the authors to suggest that impurities played only a secondary role in the mechanism of intercrystalline corrosion. It was found, however, that the attack at the grain boundaries depended upon the relative orientation of the grains, large differences in orientation favoring rapid attack. Boundaries where the two neighboring grains were similarly orientated showed high resistance to attack as did boundaries between grains which were in twin relationship. These observations led Lacombe and Yannaquis to suggest that the intercrystalline attack was due to lattice discontinuities present at grain boundaries. Assuming that the grain boundary is a layer three to five atoms thick and has a crystal structure which is a compromise between the two neighboring grains it is clear that the discontinuities will increase with increasing difference in orientation between the neighboring grains and hence the increasing tendency to intercrystalline attack. Roald and Streicher³ investigated the effect of heat treatment of aluminum alloys ranging in purity from 99.2 to 99.998 pct on the corrosion resistance in 20 pct hydrochloric acid and 0.30N sodium hydroxide. They found that in hydrochloric acid the intercrystalline attack appeared to be determined by the type and quantity of impurities present and by the relative orientation of the grains. No difference in the susceptibility to intercrystalline attack was observed between specimens quenched and those furnace cooled, from 575°C. In 0.30N sodium hydroxide some materials exhibited intercrystalline attack, this taking the form of V-notches. Rohrmann¹ offered no explanation for the greater susceptibility to corrosion of material quenched from 600°C. It seems possible that this difference is connected in some way with a different distribution of impurity elements in the quenched and slowly cooled specimens. The fact that Roald and Streicher8 observed no difference between quenched and slowly cooled specimens may possibly be due to differences in either rate of cooling or silicon content or possibly both. Both these would be expected to have an effect on the distribution of impurity elements. Although the rate of cooling used by Rohrmann was slightly more rapid than that used by Roald and Streicher the position cannot be clarified because Rohrmann does not give the silicon content and Roald and Streicher give the silicon contents of only a few of their alloys. That Lacombe and Yannaquis2 found no difference in corrosion behavior attributable to impurities between the two materials they used may be because both were of high purity compared with the aluminum used by Rohrmann.¹ Although they found no difference in the corrosion behavior of their two materials it is possible that the results obtained by Lacombe and Yannaquis may, nevertheless, have been influenced by impurity distribution, since, on the transition lattice theory of grain boundary structure, it would be expected that sparingly soluble impurities would tend to segregate to boundaries where the orientation difference is such that there is a greater density of atomic sites of suitable size to contain them. It was considered worth while, therefore, to examine the corrosion properties of a series of materials of differing impurity content with the objects of confirming the experimental observations made

Jan 1, 1954

By Che, C. W. Spencer, C. A. Steidel, Yu Li

Grain boundary grooving as it occurs in a 5.5-deg simple-tilt nickel bicrystal immersed in a saturated Ni-S liquid has been studied. A 1/3 (t= time) dependence for the depth of the groove indicates that the rate -controlling meckanism for groove developvnent is volume diffusion in the liquid. The activation energy for the diffusive process is estimated at 15 kcal per mole. Using an extended form of W. W. Mullins&apos; analysis for grooving reactions, a computer is employed to fit the appropriate diffusion equation by a numerical method. From the analysis a possible method of measuring L.S. the solid-liquid interfacial tension is demonstrated. WHEN a polycrystalline solid is heated in equilibrium with a saturated fluid phase, grooves will form at the intersections of the grain boundaries and surface. The equilibrium root angle of these grooves is determined by the balance between the grain boundary and solid-fluid interfacial tensions. The grooves deepen and widen with time due to a difference in chemical potential between the flat interface and the curved interface of the groove. In this investigation solid-liquid interaction at a grain boundary is studied, with the object to determine the mechanism of grain boundary grooving in the presence of a saturated liquid. Extension of Mullin analysis then permits the estimation of the solid-liquid interfacial tension. Herring has demonstrated by dimensional analysis that the quasi-steady state change of the dimensions of a small particle during sintering is proportional flow, evaporation-condensation, volume-diffusion, and surface-diffusion mechanisms, respectively, if geometric similarity of consecutive configurations of the particles holds. Mullins has extended this analysis to grain boundary grooving. In the case of shallow grooves he solves the diffusion equations for evaporation-condensation, volume-diffusion, and surface-diffusion mechanisms, and obtains the same time dependence for any linear dimension of a groove as Herring found for the change in dimension of small particles. In Mullins&apos; analysis2&apos;3 grain boundary grooving of a bicrystal by volume diffusion in the presence of a saturated fluid phase was considered. Several assumptions were made prior to the analysis: 1) the solid-fluid interfacial tension is independent of crystallographic orientation; 2) the Gibbs-Thompson equation relating curvature and chemical potential is applicable at the solid-fluid interface; 3) the relaxation time of the diffusion field is small compared with the time for appreciable changes in the curvature of the groove, so that the diffusion is divergence-less; and 4) the groove has a small slope relative to the initial flat interface. Usually this last condition is fulfilled only for solid-gas systems. The solution of the diffusion equation under the above conditions produces a t1/3 dependence for any linear dimension of the groove in the case where volume diffusion is rate-controlling. The t1/3 dependence is valid because geometric similarity of the grooves exists for all time. EXPERIMENT In the experiment a 5.5-deg nickel bicrystal of <100> simple-tilt orientation was used.4 Grooving was studied in the presence of a saturated Ni-S liquid alloy as a function of time at 750°, 800°, 850°, and 900°C. A low-angle bicrystal was chosen for study to insure that 2L.S. > .b., where l.s.

Jan 1, 1964

By Leonard Bernstein, Harry Bartholomew

MANY of the properties of metals and alloys are structure dependent. Not the least of these is the grain structure. For example, in producing alloy bonds between silicon or germanium and gold-alloy foil at temperatures above their eutectics (which are, respectively, at 370" and 356°C) but below the melting points of the individual constituents, it was found that wetting was more complete with a fine-grain structure than with a coarse-grain structure. In cases where the assemblies were held together for long periods of time at elevated temperatures, wetting was not as critically dependent on the grain structure as was the case for shorter periods of time. When liquid phase forms, it first forms in the grain boundaries due to the fact that most of the impurities are concentrated at the grain boundaries and that most impurities tend to lower the melting point of the gold alloy somewhat. In addition, grain boundaries are at high-surface free-energy levels and are thus in a more favorable condition to initiate good wetting. A fine-grain structure will have more grain boundaries per unit area than coarse-grain structure thereby making wetting somewhat easier. To make the grain structure of an alloy sys- tem visible, a satisfactory method for differentiating between the different grains of which the system is composed must be employed. For the alloy bonding system previously described, the grain structure on the foil surface is of primary interest. Conventionally used etchants for the gold-alloy system will not adequately delineate grain boundaries in their structures. For this reason an alternate method is proposed for those cases where such a determination is necessary. Foil between 0.0005 and 0.005 in. thick has been used in this determination. The equipment used in this investigation was described in a previous publication.&apos; Essentially, it employs a resistance-type filament which can be heated up very rapidly and cooled very rapidly. A thin refractory material is placed on top of the heating element. This refractory material may be molybdenum, tantalum, tungsten, or even ceramic but its thickness should be less than 0.010 in. to minimize the temperature gradient across it. The gold-alloy foil is placed on top of this refractory material. The temperature of the heating filament is raised to 900°C and cooled down to below 100°C in 5 to 10 sec. During this cycle, the foil can be observed to plastically deform due to its own weight. With no further chemical etching, this assembly is then viewed under a standard metallurgical microscope and the grain size measured. Fig. 1 is a Au-1 pct Sb alloy at 400X magnification and Fig. 2 is a Au-1 pct Ga alloy at 130X. This technique has proved to be an adequate means for controlling processing conditions in producing large quantities of gold alloy foil.

Jan 1, 1963

By S. Weinig, E. S. Machlin

IN a previous study of the grain boundary stress relaxation phenomenon,&apos; the authors had arrived at the conclusion that two successive steps were involved in the complete relaxation of stress at a grain boundary. These two steps were grain boundary migration and grain boundary slip. It was further suggested that the slower of these two processes is rate controlling. In an attempt to ascertain the validity of the above, a study of grain growth kinetics on the same alloys used for the previous study was undertaken. Separate wire specimens, 0.030 in. diam, in the same as-deformed state (99 pct reduction in area) as in the previous investigation were heated to the various temperatures for a sequence of times. The compositions used and the temperatures of testing are summarized in Table I. A statistical determination of grain size was then made on each specimen, using the average length of traverse between boundaries as a measure of the grain size. Examples of typical results obtained are shown in Figs. &apos;1 and 2. The grain growth law D = ktn was found to fit most of the data. Using this relation, values of the grain growth exponent n were calculated and the inter -esting dependence of n on solute content shown in Figs. 3 and 4 was then obtained. Also, it was found that the constant K which was obtained from extrapolation is roughly independent of solute content. (K is uncertain to within ±25 pct.) An estimate of the activation energy for grain growth was made using the technique of Burke? This method utilizes the plot of the reciprocal of time necessary for the grain size to increase from a definite starting size Do, to a final size D, for each isothermal heat treatment. The slope of this curve corresponds to the heat of activation for the process. In Fig. 5 the rate of grain growth is plotted against the reciprocal of absolute temperature. The activation energies were obtained from these curves and are summarized in Fig. 6. Discussion It is interesting to note that the same saturation in the variability with concentration has been found for the grain growth exponent n in this research as was found in the previous researches on both grain boundary stress relaxationa and internal friction of dislocation origin." The significant feature is that saturation in the variation of these properties has been found to occur at about the same level of solute concentration, namely, between 0.1 and 0.5 atomic pct. The only characteristic common to these different quantities is that solute adsorption at grain boundaries and dislocation can take place and affect these values in a similar fashion. Thus, it is concluded that the grain growth exponent n depends upon solute adsorption at the grain boundaries. It appears, therefore, that theoretical considerations leading to the contrary conclusions5,6 have not treated the real process by which solute atoms exert their effect on grain growth. Prior to the analysis of the observed activation energies, certain of the aspects of the aforementioned grain boundary stress relaxation investigation will be briefly reviewed. It was found that with the addition of solute atoms the grain boundary stress relaxation peak (internal friction vs temperature) decreased in intensity. Simultaneously a second peak was observed which increased with increasing solute atoms. Both of these peaks were shown to be the result of grain boundary stress relaxation. They have been termed, respectively, poire peak and solute peak. It was suggested that the solute peak was dependent upon grain boundary migration, whereas the pure peak was a manifestation of grain boundry slip. Activation energies for both peaks were obtained as a function of Solute content. The value of the activation energy in the temperature range corresponding to the solute peak was

Jan 1, 1958

By P. K. Koh

Isothermal salt bath annealing of 0.014-in. thick 3 pct Si-Fe sheet was conducted at temperatures ranging from 927" to 1260°C in order to investigate the grain-growth behavior. Within the temperature range studied normal grain growth did not take place; instead nucleated grain growth prevailed. The nucleation frequency was found to increase with the annealing temperature, with corresponding decrease in grain diameter of ultimate secondaries after long-time annealing. Large secondaries grown by low-temperature isothermal annealing are essentially of (110)[001] orientation, and the (110)[001] texture is strong. The relatively small secondaries grown by high-temperature isothermal annealing also have orientations near (110)[001] as an average but widely scattered; thus the amount of (110)[001] texture is found to decxease with isothermal annealing temperature. Grain growth and texture development at various annealing temperatures can be described by families of sigmoidal curves with the time axis. ThE cube-on-edge (110)[001] texture and predominating large secondaries in fully annealed 3 pct Si-Fe are customarily produced by a 1100°C annealing of a primary recrystallized sheet. On annealing the primary recrystallized sheet above the temperature at which secondary recrystalliza-tion starts May and turnbull&apos; have demonstrated that secondary grains decrease in size as the temperature is increased. The phenomenon was attributed to a decrease in the (G/N) ratio where G is the growth rate of secondary grains and N is their nucleation frequency. In addition, the growth of grains in orientations rather far from (110) [0011 occurred. The present investigation is devoted to detailed rate studies of the same phenomenon. Mi-crostructure, magnetic anisotropy, and texture data were obtained on strips of 3 pct Si-Fe isothermally annealed for various lengths of time in a BaC12 salt bath at temperatures ranging from 927&apos; to 1260°C. The texture results are reported elsewhere.2 EXPERIMENTAL PROCEDURE Commercial 0.014-in. cold-rolled silicon-iron strip (3.16 pct Si), prepared by two stages of cold rolling with an intermediate short anneal, was given a decarburizing 3-min anneal at 800°C. Metallo- graphic studies indicated complete recrystallization. In order to provide a protective atmosphere and to obtain a rapid rise to an elevated temperature, a BaC12 and NaCl fused salt bath was used to heat samples isothermally at temperatures from 927" to 1260°C for various lengths of time. Torque magnetometer measurements were made on 1-in.-diam disc specimens before and after annealing, since magnetic torque curves supply information on textures and texture changes. Surface layers of the strip samples were polished and etched to reveal the grain structure. The diameters of grains were measured from micrographs at X100. The average value of diameters of matrix grains was established. Individual grains in the micrograph were cut out separately along grain boundaries. These grains were classified either as matrix grains or growth grains. The percent weight of growth grains in the cut-out patterns was reported as percent growth. Manv of the individual grains in the annealed samples were large enough to be X-rayed with a 5-mil beam for Lauegrams. Complete (110) pole figures were obtained for the primary recrystallized structure and short-time annealed structure with CoKa radiation at 30 kv in the back-reflection range, such as (220) for (110) poles, as described recently by Newkirk and ~ruce.~ Both the X-ray optics and the technique for pole figure differed slightly from those adapted previously.4

Jan 1, 1960

By Leonard P. Skolnick

PROPERTIES of materials containing a hard con-stituent dispersed in a metallic matrix are dependent on the distribution of the hard phase.&apos;-&apos; The grain growth of titanium carbide in liquid nickel was studied in order to understand microstructural development in this and similar systems. Smith" has examined the influence of surface energy on the distribution of phases in alloys. The grain growth of tungsten carbide in cobalt has been the subject of many conflicting theories. Those prior to 1950 have been summarized by Goetzel.&apos; Recently, Gurland and Norton3 and Gurland8 have indicated that growth occurs primarily by diffusion through the liquid phase in order to lower the inter-facial energy between the carbide and the liquid metal. In the development of the Ti-C-Ni ternary phase diagram, Stover and Wulff9 have shown that both eutectiferous dendritic titanium carbide, and primary titanium carbide which had developed crystals of cubic symmetry, were rounded by heating below the solidus, and that if primary carbide was already present the typical eutectic decomposition products were not formed. Experimental Procedure The infiltration technique was used to study the grain growth of titanium carbide in nickel. In contrast to sintering, infiltration more readily assures uniform dispersion of the two phases, and the problems associated with mixing by milling; i.e., mill contamination, breakdown of original carbide size in the operation, development of interfering surface films, etc., are avoided. In addition, the complicating effect of densification by shrinkage is eliminated. The titanium carbides used were commercial grades. Their chemical analyses, size distributions and shapes are summarized in Table I. Size analysis was conducted by microscopic count on a minimum

Jan 1, 1958

By R. W. Cahn, C. D. Graham

Two predictions of the oriented growth theory of recrystallization textures have been tested by measuring the orientation dependence of the rate of growth of a single grain into a strained single crystal of aluminum, and determining the orientations of artificially and spontaneously nucleated grains growing preferentially into strained aluminum single crystals. Growth rates are found to be insensitive to orientation, except that new grains with orientations similar to the matrix or to a twin of the matrix have very low mobilities. Similarly, new grains growing preferentially into a strained crystal have random orientations, except that orientations near that of the matrix or its twins are avoided. The predictions of oriented growth theory are thus not confirmed. BASICALLY, the oriented growth theory of re-crystallization textures1 rests on the assumption that the rate of growth of a recrystallizing grain depends strongly on the orientation of the growing grain relative to the strained matrix into which it grows. In particular, the theory holds that in face-centered-cubic metals the orientation which corresponds to maximum growth rate is one in which the growing grain and the matrix are related by a rotation about a common <111> direction. The amount of rotation, as derived from several kinds of experiments,2-1 has been assigned various values, generally in the range between 20" and 40". (A <111> rotation of 60" is a twin relationship.) The conclusion that boundary migration rates depend strongly on orientation is based almost entirely on indirect evidence; there have been very few direct measurements of boundary migration rates as a function of orientation.* " The present investigation was undertaken to provide such measurements, to help make possible a decision between the oriented growth and oriented nilcleation theories of recrystal-lization textures. The basic experimental program consisted of measuring the rate of growth of a single recrystal-lizing grain consuming a strained single crystal, with the orientations of both grains preselected so that the effect of orientation on growth rate could be determined. A prerequisite for such an experiment is a strained single crystal which will support the growth of a new grain but which will not spontaneously nucleate new grains on heating. That is, the crystal must support the growth of a grain nucleated artificially, but contain no recrystalliza-tion nuclei which will become active at the testing temperature. Beck was apparently the first to note that such a condition could exist," and to make use of the condition for am experiment of the type described here.&apos; The present work was actually suggested, however, by a report of Tiedema". " that an aluminum single crystal strip, oriented with a (111) plane in the plane of the strip and a < 112> direction parallel to the tensile axis, would not recrystallize after 20 pet extension even when heated almost to the melting point, provided that the strip was heavily etched before being heated. This result could not be duplicated in the present work. In fact, it was found that any crystal which deformed in multiple slip (<100>, <112> and <111> orientations) underwent spontaneous nucleation after 15 pet extension, even if etched. However, crystals which deformed in single slip, and which did not develop heavy deformation bands, could be extended 15 pet without showing spontaneous nucleation at temperatures up to 600°C. Crystals oriented within about 10o of <110> which developed heavy deformation bands could be extended 10 pet without showing spontaneous nucleation. In all cases a heavy etch was required to prevent nucleation. Etching was necessary because of the presence of an oxide layer on the crystal surface at the time of straining, which leads to preferential nucleation at the surface.&apos;&apos; Grain Growth Rates Experimental Procedure and Results—Aluminum strips of 99.6 pet purity (principal impurities 0.19 pet Fe and 0.12 pet Si), 1 mm by 1 cm in cross section, were grown into single crystals of controlled orientation by the strain-anneal method of Fuji-wara.&apos;"&apos; " A sharp temperature gradient was maintained in the strips during growth by lowering them into a salt bath controlled at 650°C. Crystal orientations were determined by the etch-pit method of Barrett and Levenson" to an accuracy of ±2O. The crystals were extended by 10 or 15 pet in a simple hand operated tensile machine. Crystal orientations were rechecked after extension, and were found to be in agreement with the orientations predicted by the formula of Schmid and Boas." After extension, the grip ends of the single crystal specimen were cut off with a jeweller&apos;s saw, and the crystal heavily etched (at least 20 pet wt loss) in hot 10 pet Na or KOH solution. A region of severe local deformation was then introduced at one corner of the strip, usually by cutting off the corner with shears. Heating this end of the strip caused a large number of new grains to nucleate at the sheared edge. One of these grains grew to occupy the full width of the strip. The appearance of the strip at

Jan 1, 1957

By W. A. Tiller

The unidirectional freezing of a semiinfinite liquid from one end has been treated by calculating the solute and temperature distributions in the liquid and solid phases with time, when the solid-liquid interface advances at a rate proportional to (time)lt2. The nucleation probability for new grain formation is calculated as a function of the constitutional and thermal properties of the alloy, the pouring conditions and the catalytic efficiency of the nucleating agents. The transition from columnar crystal formation to equiaxed crystal formation is assumed to occur at a critical nucleation frequency for a particular alloy and the quantitative dependence of the ratio (columnar zone length/equiaxed zone length) upon the freezing conditions has been predicted. The effects of (i) fluid flow in the liquid, (ii) heat transfer through a mold wall, and (iii) a finite ingot length, have been explicitly tveated as departures from ideality. These departures from ideality are found to have very important consequences on the ingot stmcture. In Part I&apos; of this series, a qualitative description of the general ingot structure was developed. The factors that determine i) the ratio of the equiaxed zone length to columnar zone length and ii) the grain sizes in these zones, were discussed. In the present paper, topic (i) will be treated quantitatively, and it is intended that topic (ii) will be treated quantitatively in Part 111. The complete treatment has been split in this manner for ease of development and coherency of presentation. The treatments are quite mathematical in nature and, although some of the conclusions to Part 11 depend upon the treatment presented in Part 111, it was felt that the intended audience would find the material more readable and more useful in its present form. To this same end, it seems worthwhile to reiterate some of the qualitative aspects of grain nucleation during ingot solidification before embarking on the mathematical description. QUALITATIVE ANALYSIS In the conventional method of casting, a liquid alloy with a certain degree of superheat is poured into a mold. The outer rim of liquid metal cools rapidly to the liquidus temperature of the alloy and begins to undercool. As the degree of undercooling increases, nuclei of solid begin to form in this chill layer, this nucleation being catalyzed by certain heterogeneous particles in the liquid. The grains in the chill layer which are generally of random orientation, begin to grow inwards because of heat conduction through the mold. The grains usually develop a columnar shape since the frequency of nucleation of new grains in the liquid ahead of the advancing interface is generally insufficient to seriously impede the growth of the original grains. Due to the slow diffusion rate of solute in the liquid, a solute distribution is formed in the liquid adjacent to the interface which is either an excess or a depletion if the partition coefficient, ko, of the solute is < 1 or > 1, respectively. Thus, the liquidus temperature of the liquid adjacent to the interface is a function of position. The actual temperature in the liquid is also a function of position and the temperature gradient at the interface decreases as the interface moves away from the mold wall. Since the rate of advance of the interface decreases with time, the equilibrium temperature distributions and the actual temperature distributions as a function of interface position, X, can be illustrated as in Fig. 1. Thus, a region of liquid may exist in which the actual temperature is below the liquidus temperature. This liquid is contitutionally supercooled and therefore unstable with respect to the nucleation of new grains. Now, since the supercooling in Fig. 1 increases as the interface moves further away from the mold wall, the nucleation rate of new grains will increase. When the nuclea-

Jan 1, 1962

By K. G. Davis, P. Fryzuk

A study has been made of the effect of undercooling on the grain structure and solute distribution in small ingots of pure bismuth and of a 100 ppm Ag in bismuth alloy. Autoradiographic evidence shows that in undercooled melts a network of dendrites is formed throughout the melt immediately after nu-cleation. At large undercoolings the dendrites are very thin and closely spaced. The origin of the observed grain morphology can be explained on the basis of information about the mode of solidification revealed by the impurity substructure. WHILE a great number of studies of nucleation in undercooled molten metals have been made, comparatively little attention has been given to the mode of solidification from melts with large undercool- ings. Colligan and Bayles1 and Walker2 have measured the rate of advance of the solid-liquid interface in undercooled nickel melts, but the morphology of the solidifying material remained a matter of speculation. Fehling and Scheil3 have described the grain structure in ingots formed from undercooled melts of nickel, palladium, and copper, and a series of Ni-Cu and Ni-Pd alloys. At small undercoolings, equiaxed structures are observed, the grain diameter tending to increase as the temperature of nucleation is lowered. At greater undercoolings, a columnar structure is developed, while with still further undercooling there is a smaller grain size incorporating rather coarse subgrains. All-bright and Colligan4 examined the grain structure shown by ingots solidified from melts of nickel and of Ni-Ag alloy at an undercooling of 105°C, and concluded that considerable grain growth had taken place after solidification. The object of the present work is to correlate the mode of solidification from undercooled melts with both the grain structure and impurity segregation, using tracers to examine the latter.

Jan 1, 1965

By C. W. Beattie, R. L. Solter

ALUMINUM-KILLED, low carbon steel sheets are used extensively for severe deep drawing and other difficult forming operations. They usually, but not always, have a characteristic grain structure in which the grains are elongated both in the lengthwise and in the transverse direction. As described by Burns and McCabe,&apos; a typical grain in the plane of the sheet has its two axes in that plane from 1 Y2 to 4 times as long as the axis normal to the plane of the sheet. Rickett, Kalin, and MacKenzieZ have also reported on the recrystallization behavior of such steel. The contrast in grain structures of fully processed sheets of aluminum-killed and rimmed steel is illustrated by Figs. 1 and 2. The elongated grain structure of the aluminum-killed sheet is not developed on all heats or lots of this metal, and studies of the factors controlling and influencing its formation are reported in this paper. Jeffries and Archerb tate that unstrained grains are normally equiaxed, but exceptions are common. For example, if a metal containing a material mechanically obstructing grain growth is subjected to considerable working followed by thorough annealing, it may exhibit grains consistently elongated in the direction of working. Our experiments demonstrate that aluminum-killed, low carbon steel is such a metal, and that the substance mechanically obstructing grain growth is aluminum nitride. The effectiveness of aluminum nitride in inhibiting grain growth has been found to be influenced by the degree of cold reduction, the rate of heating in annealing, the thermal history of the sample before cold reduction, and the residual aluminum content. A correlation between grain shape and austenitic grain coarsening temperature also was indicated and additional experiments demonstrated that aluminum nitride is also the principal cause for the fine grain characteristic of aluminum-killed steels. Manufacture In conventional practice, aluminum-killed sheet steel is manufactured from a low carbon steel containing approximately 0.02 to 0.07 pct residual (HC1 soluble) Al. With the exception of certain samples containing greater or lesser amounts of aluminum, the steels used in these investigations were within the following composition range: C, 0.03 to 0.06 pct; Mn, 0.28 to 0.38; S, 0.017 to 0.032; Al, 0.03 to 0.06; P, <0.01; and Si, <0.01. Properly heated ingots are rolled to slabs about 4 in. thick. After surface conditioning, the slabs are reheated to about 2300°F and hot rolled continuouslv to strip about 1/10 in. thick. The strip rolling is completed at a temperature of 1550°F or higher, and the strip is coiled, usually at a temperature near the lower critical transformation. After cooling, the strip is pickled to remove oxide, cold reduced 40 to 70 pet to final thickness, then annealed to 1250° to 1350°F in 20 to 80 ton charges, the size of which results in slow heating and cooling rates. Effect of Cold Reduction According to Sachs and Van Horn,&apos; the deformations of the individual grains in rolling are similar to those of the total volume. Thus individual grains would elongate in rolling according to the amount of cold reduction imposed. This is true theoretically, but as cold reduction increases the individual grains tend to fragment, and measured grain elongations become less than theoretical. The amount of grain elongation may be described by a numerical rating based on grain counts made by the intercept method. Specimens are polished normal to the plane of the sheet, with the polished surface extending parallel to the rolling direction. After etching, grain intercepts are counted along a 50 mm line on a micrograph of suitable magnification. In random locations parallel to the plane of the sample 20 counts are made and 20 are made in the thickness direction of the sample the average count in the thickness direction divided by the average count parallel to the plane of the sample gives a numerical rating of the grain shape called grain elongation. For example, a grain elongation of 2.00 means that the average grain is twice as long as it is thick. The average of both counts may be converted to grains per sq mm by a nomograph relating intercept counts and grain count. By the same procedure the grain elongation in the plane of the sheet but transverse to the rolling direction may be determined, using transverse metallographic samples. A comparison of theoretical and measured grain elongation was obtained on an aluminum-killed

Jan 1, 1952

By W. W. Stephen, J. P. Walsted, W. J. Kroll

FOR the production of carbides, various furnace types are available, especially those using arc, resistance, and high-frequency heating. Selection of a specific means of heating depends primarily on the material to be treated and the physical properties of the carbide produced. In the present case, zirconium carbide had to be prepared on an industrial scale as a raw material for the production of anhydrous zirconium chloride. Considering that a rather expensive pure oxide was to be used, the arc-furnace treatment recommended for zircon sands in a previous publication&apos; was ruled out because of the considerable volatilization and dust losses caused by the blast of the arc. For this reason, either high-frequency or resistance heating seemed to offer more promise. Since there was not enough capacity of the former available, resistance heating was chosen. It was first thought that the Acheson silicon carbide furnace would be suitable for the present purpose, but the voltage in such a furnace, in which the current passes through the batch, varies from 220 to 75 v from the start to the end of a run. This variation is so great that a special tap transformer would have been required. Trouble was also expected by local melting of the carbide. Pure zirconium carbide melts at about 3527°C, but the melting point is brought down to 2427°C, according to Agte,2 when an excess of 6 pct C is present in the carbide. This we found confirmed by experiments in a high-frequency furnace. Excess carbon is needed in the batch to obtain a complete reduction. Fusion of the charge would cause great difficulties in an Acheson-type furnace because of the good electrical conductivity of the carbide as compared with that of the loose batch. Also, fused carbide is much more difficult to chlorinate than the spongy product that can be made in the radiation furnace described below. It was apparent that, to obtain a good-qual-ity zirconium carbide, the heat input would have to be well-controlled. The hairpin-resistor principle seemed to offer possibilities in this regard, and a furnace of this type was therefore developed. The advantages of the hairpin-resistor radiation principle have been discussed in previous publications, and a split-tube graphite-resistor furnace," now increasingly used in various laboratories, as well as a centrifugal quartz melting furnace4 of this type, has demonstrated the usefulness of this heating method. The hairpin-heater element has the following definite advantages over a straight resistor of the type used, for instance, by Georges:5 Its resistance is four times greater; it can expand freely; it is sturdier because of the larger diameter, and it has a larger radiation surface; there are no hot contacts that might wear out or overheat; only one clamp is used which permits assembling all electrical leads at one side of the furnace, making the other sides easily accessible to the operator. The shorter element and its larger diameter permit greater concentration of heat. The furnace developed is shown in fig. 1. The box (I), made of 2 1/2-in. graphite plates, has inside dimensions of 23x17x16 in. It contains the briquet-ted batch (2). The box is embedded in lampblack (3) up to the cover plate. The cover plate contains an opening for the gas escape (4) and for the observation hole (5), which permits measuring the temperature with an optical pyrometer. The cover plate is embedded in charcoal (6). The lampblack is contained in the insulating brick lining (7), held in the 1/4-in. sheet steel box (8). The graphite box is set on two rows of triangular graphite bars (9). The hairpin-heater element (10), the dimensions of which are given below the main drawing, extends horizontally in the graphite chamber and radiates freely on the batch. A graphite tube (11) keeps the lampblack from falling into the slot. The split electrode, which in reality is turned 90" against the drawing, is so arranged that the slot is vertical. The water-cooled packing gland (12) is insulated by an airgap from the heater element. A thin pipe (13)

Jan 1, 1951

By T. G. Digges, M. R. Achter

ALTHOUGH zone melting has found favor in recent years because of its convenience and its faster rate of production of single crystals, the older technique of strain annealing still has a number of advantages. It has been reported1 that the lattice of a crystal prepared in this manner has fewer defects than one grown from the melt and, of course, the control of the external geometry is easier. Also, the diameter of the crystal which can be grown by zone melting has a theoretical limit determined by the surface tension, thermal conductivity, and specific gravity of the material, while the size of that which can be grown in the solid state is limited only by practical considerations. In a program for the study of the structure-sensitive properties of niobium and its alloys it was desirable, for ease of handling, to have available single crystals of large diameter. This requirement dictated the choice of the strain-anneal technique. Induction heating was chosen in the present work in preference to other methods because of its deep penetration of large-diameter sections. A five-turn coil, powered by a motor generator operating at 10,000 cps, is contained in a vacuum chamber. The work is lowered through the coil for both the recrystallization and the crystal-growth steps. Starting with electron-beam melted material, the bars are cold-swaged, recrystallized, strained in a tensile machine, and then subjected to the growth step. Single crystals have been grown under the following conditions: Reduction in area during swaging— 79 to 99 pct, recrystallization temperature—1500o to 1667°C at the rate of 1.25 in. per hr, strain—1.0 to 2.5 pct, growth temperature—1600o to 2400°C at rates from 0.15 to 1.25 in. per hr. Starting grain sizes achieved after the recrystallization step varied from 0.6 to 1.55 mm. The starting grain size was found to be critical; if it is too fine and has any residual cold work, as shown by the presence of a fiberlike texture, a single crystal cannot be grown. Using this technique niobium single crystals have been grown from 0.250 in. diameter and 12 in. long to 1.00 in. diameter and 5 in. long as shown in Fig. 1. The 1-in. crystal is the largest to be reported for a refractory metal as compared to 0.500 in. for molybdenum grown by zone melting.2 As with zone melting, control of orientation is possible by the use of special procedures. Other investigators, see for example Williamson and Smallman,3 have reported that orientation control may be achieved by a bending technique. In the present work the strained rod is partially lowered through the coil to start the growth of the crystal. Then it is removed and bent at a point ahead of the interface between the single and polycrystalline portions. Finally, it is returned to the chamber and growth is continued "around the corner". The amount of bending which can be imparted to the rod is limited by coil geometry which up to now has been 10 deg. However, by repeating the bending and growing operations it should be possible to attain any desired orientation. The perfection of these single-crystal specimens as a function of process variables is being measured by X-ray and metallographic techniques and the results will be reported in a forthcoming paper. This work was partially supported by the Advanced Research Projects Agency.

Jan 1, 1964

By C. M. Wayman, M. A. Gedwill, C. J. Altstetter

Cobalt Whiskers were grown by the hydrogen reduction of CoBr,. The fcc = hcp martensitic trans-formation in these whiskers was studied using X-ray and metallographic techniques. Present theories of the fee + hcp martensitic transformation were ex-amined in the light of the present findings Experimental evidence for slip as a part of the transformation mechanism is presented. In the bulk form cobalt is known to undergo a martensitic transformation1 from an hcp to an fcc structure upon heating above about 430°C. Upon cooling, the reverse transformation starts at about 400°C, but does not go to completion except in single crystalline or coarse-grained samples, or upon plastic deformation. The transformation may be thought of as the shearing of close-packed atomic planes in such a way as to change the stacking sequence from ABCABC to ABABAB with a contraction normal to the shearing planes and an expansion in the planes,2 producing a c/a ratio slightly less than ideal. This restacking to give {0001) and (112) 11 (1010) can be accomplished by the passage of a shockfey partial dislocation over every other (111)f plane resulting in a theoretical shear of 19.47 deg. To date there has been no completely satisfactory explanation of how the partial dislocations are produced on the close-packed planes or propagated from plane to plane to give the correct stacking sequence. The object of the present research was to study some of the characteristics of the fcc to hcp transformation in geometrically simple single crystals. It was felt that whiskers of cobalt grown by halide reduction would be ideal specimens with which to study the transformation, because whiskers have the following desirable features: 1) are easily grown as fcc single crystals; 2) exhibit relatively simple geometries; 3) have comparatively low dislocation densities;

Jan 1, 1964

By K. R. Janowski, A. B. Chase, E. J. Stofel

Ruby crystals of exceptional optical quality have been grown from fluxes composed of Bi2O3 + PbF2 or Bi2O3 + BiF3. Careful control of growth conditions was required to minimize the number of growth defects consisting primarily of dislocations, occluded flux, and growth twins. Slow growth rates coupled with the addition of small amounts of La2O3 as a flux dopant were helpful in producing nearly perfect crystals. Chromium concentrations up to 1 pet were incorporated into these crystals without adversely affecting their perfection. Crystals up to 10 mm in diameter having as few as one to ten dislocations per crystal were typical of those produced under optimum growth conditions. Under these same conditions the occlusions were eliminated, and growth twins were either eliminated or reduced to one to three twin boundaries per crystal. INTEREST in the nature of lattice imperfections produced during the growth of ruby crystals (Al2O3-corundum structure) has been stimulated recently by the need for ruby lasers of good optical uniformity. As a result, ruby crystals grown by the flame-fusion (Verneuil) process have been studied in detail and have been found to contain high densities of dislocations (106 to 107 dislocations per sq cm) and mosaic structures that impair their optical quality.&apos;-3 Attempts to improve the quality of Verneuil-grown crystals by controlling the growth variables have not been entirely successful. On the other hand, rubies grown by the flux process have received little attention even though their more nearly equilibrium growth conditions suggest that they would have fewer growth defects. Nelson and Remeika have shown recently that small flux-grown rubies can indeed be made into very low-loss lasers. However, they did not report information on crystalline imperfections in these crystals.4 Stephens and Alford have reported briefly on defects found in a flux-grown corundum crystal but they neither described these defects in detail nor related them to the method of growth.5 The present investigation was undertaken to establish the relationship between growth conditions and crystalline perfection of flux-grown rubies, sapphires, and other crystals of the corundum family. EXPERIMENTAL PROCEDURE The corundum crystals used in this study were grown by slowly cooling solutions of A12O3 in molten-salt solvents of PbO + PbF2, Bi2O3 + PbF2 or Bi2O3 + BiF3. The materials employed were Linde A (a Al2O3 powder) with reagent-grade chemicals as solvents. Some of the solutions also contained small additions (1 pet or less) of one or more of the transition metal elements chromium, manganese, iron, and titanium as crystal dopants. La2O3 was added to some of the solutions to alter the growth habit of the crystals. The La2O3 promoted the occurrence of growth facets parallel to (0112) and {1011} planes, where the indexing system is the morphological system described in detail by Kron-berg6 and used throughout this paper. The melts from which the crystals were grown were prepared by mechanically mixing the various constituents and placing them in platinum crucibles. The covered 50- and 100-ml platinum crucibles were heated in muffle furnaces at 1250°C for short periods of time (1 to 4 hr) and then program-cooled at either 2° or 4°C per hr to 1000°C. At this lower

Jan 1, 1965

By S. J. Matas, R. F. Hehemann, R. H. Goodenow

The hot-stage metallographic method has been employed to determine radial growth rates of upper and lower bainite. Lower bainite appears to grow as individual plates that increase in both length and thickness while upper bainite grows as sheaves of closely spaced plates or needles. A discontinuity in the temperature dependence of the growth rate appears to be associated with the transition from upper to lower bainite and stepped quenching experiments suggest that lower bainite will not continue to grow when the temperature is raised to the upper range. Thus, a single process does not appear to control the growth of both upper and lower bainite. A tentative model based on the forces acting on a dislocation interface is proposed to account for the slow growth of bainite. It has been recognized since the early studies of Davenport and Bain that a close relationship exists between the martensite and bainite transformations in steels. However, diffusion of carbon during the transformation has lead to considerable controversy concerning the reaction mechanism.1-7 Many studies4,5,7-9 have emphasized the structura and kinetic aspects of the bainite reaction and, in particular, the difference observed between the two forms of bainite—frequently termed upper and lower bainite. A kinetic difference between the two types of bainite has been revealed in the activation energy determined from the overall reaction;4,5 however, many difficulties arise in an interpretation of these activation energies since a number of temperature dependent factors contribute to the kinetics of the overall reaction. The displacive surface relief10 accompanying the bainite reaction permits a separation of the nuclea-tion and growth contributions to the reaction and allows the kinetics of the growth process to be examined. Several investigations of this type have been conducted; however, they have been restricted to the low temperature range.3&apos;6&apos;11 Here, bainite plates are observed to grow slowly in both length and width. The present investigation extends such studies to include the upper bainite range. MATERIALS AND PROCEDURE Materials. The composition of the steels studied in this investigation are presented in Table I. The range of composition permitted a study of the effect of carbon content on the growth rate of upper bainite in the 9 pct Ni steels and consideration of both upper and lower bainite in the higher carbon steels. These steels were received as hot-rolled rounds and were forged to approximately 1/2 in. thick plates for preparation of metallographic specimens. In order to produce an essentially zero carbon steel, Steel A was decarburized in wet hydrogen at 1550°F for 15 days with the result tabulated as Steel AD. Experimental Procedure. In order to study the slow growth of the displacive transformation product, a commercially available Leitz hot stage was modified to permit heating of the specimen by its own resistance in a manner similar to that employed by other investigators.6 The furnace chamber was maintained at a pressure of 2 to 3 X 10-5 mm Hg prior to austenitizing at 2000° ± 12° F for 5 min. Samples could be cooled from the austenitizing to the transformation temperature in approximately 15 sec and the progress of transformation was recorded with a 9 cm by 12 cm single frame camera or a 16 mm movie camera. RESULTS AND DISCUSSION Growth Characteristics. The general growth characteristics observed in this investigation were analogous to those reported previously.3,6,11 Typical growth sequences for the high temperature and low temperature forms are illustrated in Figs. 1 to 5. In both temperature ranges, bainite plates increase in length at slow, temperature dependent rates until interrupted by a grain boundary, another plate, or

Jan 1, 1963

By L. S. Castleman, H. A. Froot

A study was made of the growth kinetics of B-brass layers in a-brass us y-brass diffusion couples in which both phases were saturated in the 6 phase. It was found that useful estimates can be made of the difhsion coefficients at the ß-phase interfaces from interface displacement data, which are relatively easy to measure experimentally; also, a useful estimate can be made of the mean heat of activation for diffusion from layer growth data. The effects of pressure on layer growth kinetics were explored and found to be consistent with a vacancy model for diffusion in 6 brass in both the disordered and ordered temperature range. 1 HE study of multiphase diffusion receives relatively little attention in comparison with that devoted to diffusion in homogeneous materials. Of the hundred or so papers on diffusion written each year,&apos; only a few are devoted to the subject, and it is the rare review article2,3 that devotes any space to multiphase diffusional phenomena. This is unfortunate, since the metallurgist frequently encounters practical situations involving cementation processes such as cladding, protective coating of metals, and so forth in which he needs guidance of a fundamental nature. The investigation of multiphase diffusion phenomena is complicated by the fact that one must consider reactions at phase interfaces as well as the migration of atoms within the individual phases; these reactions cause the interfaces to be displaced within the diffusion zone. The relative displacement of two adjacent interfaces corresponds to the growth of an intermetallic layer, and it is not difficult to determine experimentally the kinetics of layer growth. In the past few years it has been pointed out4,5 that valuable fundamental information may be gained about diffusion coefficients and the concentrations at interfaces through a study of their motion. Accordingly, it is the purpose of this paper to report the results of an investigation of interface motion in a simple multiphase binary system. In the work reported herein, attention was focussed on the growth of a 6 -brass layer in a diffusion couple prepared from a- and y-brass compositions saturated with respect to the 6 phase. This particular system was chosen for a number of reasons. First, the bcc ß-brass phase is disordered for all compositions at temperatures above 468°C and becomes increasingly long-range ordered for all compositions at temperatures below 454° C;6,7 thus, the effects of a variation in order on interface displacement could be determined. Second, diffusion coefficients have already been obtained as a function of temperature in the ß phase both in the ordered and disordered Re- and a theoretical picture of the mechanism of diffusion in this and other ordered solutions is emerging.11&apos;12 Third, some thermodynamic data are available for the a-, ß-, and ?-brass phases,13,4 and these data could be used to gain insight into questions of phase equilibria and concentrations at phase interfaces.

Jan 1, 1963

By S. T. Wlodek, John Wulff

Dense, adherent, and ductile coatings of chromium can be applied to Molybdenum by selectively freezing out the chromium solute from a supersaturated copper-or tin-rich liquid alloy. The successful exploitation of this technique depends on the control of variables common to any freezing process. Under the most favorable combimtion of supersaturation, nucleation, and mass-transfey, coatings as thick as 0.004 in. can be obtained. In exploratory experiments it was found that a coating of chromium may be applied to molybdenum by immersing the molybdenum body in a supercooled copper-chromium or tin-chromium melt and selectively freezing out the contained chromium onto the molybdenum body. Supercooling of the liquid metal plating bath can be achieved by cooling, or the maintenance of thermal or concentration gradients in the melt. The technique is similar to conventional dip-plating procedures where the sample is immersed in a molten bath of the pure plating medium, except that in these studies, the plating medium consisted of a dilute solution of chromium in liquid tin or copper. The success of this method requires that the base material to be plated be isomorphous with the coating and insoluble in the plating bath. In addition, the plating bath, while having an appreciable solubility for the coating component, must have negligible solid solubility in the solid plate. Such relationships exist in the molybdenum-chromium, chromium-tin, and chromium-copper systems. Liquid tin and copper are satisfactory as the plating solvent, because although chromium is appreciably soluble in both of these liquid metals, solid inter-solubility is almost negligible, and no compound formation occurs. The same relationships exist between molybdenum and copper or tin except that the solubility of molybdenum in liquid tin or copper is insignificant. It is this combination of solubility relationships, as well as the completely isomorphous nature of the molybdenum-chromium system, which allows the formation of a chromium coating on molybdenum by the selective freezing mechanism. Since the main deterrent to the widespread use of molybdenum for high-temperature structural applications is its poor oxidation resistance, the formation of a dense and ductile chromium coating on molybdenum may well be of commercial significance. In addition, the basic principle of coating formation by selective freezing should be applicable to other metallic or even nonmetallic systems. The precipitation of the solute from a liquid metallic solution, for the formation of a protective coating, has not been previously considered in the literature. Frequently, however, a similar phenomenon has been described in the corrosion of metallic heat exchangers by liquid metals1 wherein temperature gradients in the liquid metal heat-transfer medium result in the formation of layers or deposits of metallic residue2 in the cooler portions of the system. Impregnating techniques using liquid baths have been tried as a plating process. Semchyshen and Barr, for instance, have successfully diffused chromium into molybdenum by immersing the molybdenum specimen in a liquid tin-chromium or copper-chromium alloy. This simple holding technique fails to produce a layer of pure chromium and requires prolonged immersion to allow for the diffusion-activated enrichment of the molybdenum. During the course of this investigation, most of the phase relations published for the copper-chromium4 and molybdenum-chromium4 systems, were found reliable. It was necessary, however, to investigate the solubilities of chromium and molybdenum in liquid tin as well as the solubility of molybdenum in liquid copper. These determinations were made by a simpie weight loss technique on materials of the same purity as used in the plating studies. The results of these solubility studies are given in Tables I and I1 and indicate that although the solubility of molybdenum in liquid

Jan 1, 1961

By E. Teghtsoonian, K. G. Davis

THE preparation of cobalt crystals offers problems: on cooling through 400°C a phase transformation takes place whereby the structure changes from face-centered cubic to the low temperature close-packed hexagonal form. There are a few references in the literature which describe the preparation of cobalt crystals for magnetic work1,&apos; where melt-solidification methods were used. Reasons will be given why such methods were found unsuitable for our purposes. The following techniques were tried: 1) Solid state methods. Any mechanical shaping of tensile specimens would lead to uncertainty with regard to residual strains. With this in mind, following Kaa, strain-anneal and grain growth techniques were extensively investigated. In no Case were useful single crystals obtained. The transformation boundaries, which are crystallographically similar to twin boundaries, are probably extremely stable and inhibit the grain growth. 2) Growth from the melt. Normal melt-solidification methods were tried, using recrystallised alumina crucibles, and also a osoft moldo technique in which cobalt rod was packed in alumina powder. Trouble was encountered with gas evolution on solidification of the melt, giving porosity. Some adhesion of cobalt to the alumina also took place. Only a small proportion of the runs gave single crystals. 3) Zone-refining. Much the best results have been obtained using an electron-beam floating zone refiner. The refiner, based on a design by Calverly et Al .5 had a single-turn tungsten filament. It was operated at 600 v with an emission current around 45 ma. Commercial cobalt rod 1/8 in. in diam was given one pass at a rate of zone travel normally 25 cm per hr, under a vacuum approximately 10-5mm of Hg. Over 50 pct of the runs gave long sections of single crystal. Crystals up to 2o Cm long with good uniformity of cross section and straightness have been grown in this way. Increase in the length of the molten zone resulted in higher single-crystal yields, but control of zone stability was reduced. Sub-boundaries in the zoned crystals were few, and Laue X-ray spots were sharp. The success of the method is believed to lie in two factors, namely i) the absence of crucible restraints during the transformation, and ii) extreme axiality of heat flow during cooling. The axes of the zoned crystals were clustered around the [ 1010] pole, the pole of the basal plane being at least 60 deg from the specimen axis. Crystals grown at lower rates of zone travel showed the same orientation range, Fig. 1. This orientation is similar to that for other hexagonal metals grown from the melt, which is a little surprising when the mode of formation of the hexagonal phase is considered. The preferred orientation is probably caused by anisotropy of heat conduction, i.e., that crystal which can most effectively conduct heat away from the transformation interface will be favored. Fig. 2 shows the area on a standard cubic stereo-gram in which there are no poles more than 60 deg from one of the four < 111> directions. This area

Jan 1, 1962

By E. M. Porbansky

DURING the investigation of the electrical transport properties of copper, it became necessary to prepare large single crystals of the highest obtainable purity. In an effort to meet these demands, single crystals of copper have been grown by the conventional pulling technique—as has been used for the growth of germanium and silicon crystals.&apos; Low-temperature resistance measurements made on these crystals show that, as far as their electrical properties are concerned, they are generally of significantly higher purity than the original high-purity material. The use of these pure single crystals with very high resistance ratios has made possible the acquisition of detailed information regarding the electron energy band structure of copper2-&apos; and has stimulated widespread effort on Fermi surface studies of a number of other pure metals. It is the purpose of this note to describe our method of preparing very pure copper crystals by the Czochralski technique. Precautions were taken to prevent contamination of the melt from the crystal growing apparatus. A new fused silica growing chamber was used to prevent possible contamination from previous groqths of other materials such as germanium, silicon, and so forth. A new high-purity graphite crucible was used to contain the melt. This crucible was baked out in a hydrogen atmosphere at -1200°C for an hour, prior to its use in crystal growth. Commercial tank helium, containing uncontrolled traces of oxygen, was used as the protective atmosphere. A trace of oxygen in the atmosphere appears to be necessary for obtaining high-purity copper single crystals. A 3/8-in-diam polycrystalline copper rod of the same purity as the melt was used as a seed. The copper rod was allowed to come in contact with the melt while rotating at 57 rpm. When an equilibrium was observed between the melt and the seed (that is, the seed neither grew nor melted), the seed was pulled away from the melt at a rate of 0.5 mils per sec. As the seed was raised, the melt temperature was slowly increased, so that the grown material diminished in diameter with increasing length. When this portion of the grown crystal was -1 in. long and the diameter reduced to less than 1/8 in., the melt was slowly cooled and the crystal was allowed to increase to - 1-1/4 in. diam as it was grown. By reducing the diameter of the crystal in this manner, the number of crystals at the liquid-solid interface was decreased until only one crystal remained. Fig. 1 shows a typical pulled copper single crystal. The purity of the starting material and the crystals was determined by the resistance ratio method: where the ratio is taken as R273ok/R4.2ok. The starting material, obtained from American Smelting and Refining Co., was the purest copper available. Most of the pulled copper crystals had much higher resistance ratios than the starting material. The highest ratio obtained to data is 8000. Table I is an example of the data obtained from some of the copper crystals. Note that Crystal No. 126 had a lower resistance ratio than its starting material and this might be due to carbon in the melt. The melt of this crystal was heated 250" to 300°C above the melting point of copper. At this temperature it was observed that copper dissolved appreciable amounts of carbon. The possible presence of carbon at the interface between the liquid and the crystal will result in reducing conditions and negate the slight oxidizing condition required for high purity as discussed below. The possible explanations of the improvement in the copper purity compared to the starting material are: improvement in crystal perfection, segregation, and oxidation of impurities. Of these, the latter seems to be most probable. A study of the etch pits in the pulled crystals showed them to have between 107 and 108 pits per sq cm. The etch procedure used was developed by Love11 and Wernick.10 The resistivity of the purest copper crystal grown was 2 x 10-10 ohm-cm at 4.2oK; from the work of H. G. vanBuren,11 the resistivity due to the dislocations would be approximately 10-l3 ohm-cm, which indicates that. the dislocations in the copper crystals would contribute relatively little to the resistivity of the crystals at this purity level. Segregation does not seem likely as the reason for purification of the material, since the resistivity of the first-to-freeze and the last-to-freeze portions are approximately the same, as was observed on Crystal No. 124. On most of the crystals that were examined, the entire melt was grown into a single crystal. If the

Jan 1, 1964