Search Documents

By W. R. Hibbard

THE classical work on the electrical conductivity of alloys was carried out by Matthiessen and his coworkers1 in the early 1860's. He attempted to correlate the electrical conductivity of alloys with their constitution diagrams, but the information regarding the latter was too meager for success. Guertler2 reworked Matthiessen's and other conductivity data in 1906 on the basis of volume composition (an application of Le Chatelier's principle with implications as to temperature and pressure effects), and obtained the following relationships between specific conductivity and phase diagrams (plotted as volume compositions) : 1—For two-phase regions, electrical conductivity can be considered as a linear function of volume composition, following the law of mixtures. 2—For solid solutions, except intermetallic compounds, the electrical conductivity is lowered by solute additions first very extensively and later more gradually, such that a minimum occurs in systems with complete solid solubility. This minimum forms from a catenary type of curve. Intermetallic compound formation with variable compound composition results in a maximum conductivity at the stoi-chiometric composition. Landauer" has recently considered the resistivity of binary metallic two-phase mixtures on the basis of randomly distributed spherical-shaped regions of two phases having different conductivities. His derivation predicts deviations from the law of mixtures which fit measurements on alloys of 6 systems out of 13 considered. Volency (Ionic Charge) Perhaps the first comprehensive discussion of the electrical resistivity of dilute solid-solution alloys was presented by Norbury' in 1921. He collected sufficient data to show that the change in resistance caused by 1 atomic pct binary solute additions is periodic* in character. The difference between the period and/or the group of the solvent and solute elements could be correlated with the increase in resistance. Linde5-7 determined the electrical resistivity (p) of solid solutions containing up to about 4 atomic pct of various solutes in copper, silver, and gold at several temperatures. He reported that the extrapolated"" increase in resistance per atomic percent addition is a function of the square of the difference in group number of the solute and solvent as follows: ?p= a + K(N-Ng)2 where a and K are empirical constants and N and Ng are group numbers of the constituents. This empirical relation was subsequently rationalized theoretically by Mott,8 who showed that the scattering of conduction electrons is proportional to the square of the scattering charge at lattice sites. Thus, the change in resistance of dilute alloys is propor-t,ional to the square of the difference between the ionic charge (or valence) of the solvent and solute when other factors are neglected. Mott's difficulty in evaluating the volume of the lattice near each atom site where the valency electrons tend to segre-gate: limited his calculations to proportionality relations. Recently, Robinson and Dorn" reconfirmed this relationship for dilute aluminum solid-solution alloys at 20°C, using an effective charge of 2.5 for aluminum. In terms of valence, Linde's equation becomes ?P= {K2 + K1 (Z8 -Za)2} A where K1 and K2 are coefficients, A is atomic percent solute, Z, is valence of solvent, and Zß, is valence of solute. Plots of these data for copper, silver, gold, and aluminum alloys are shown in Fig. 1. The values of K1 and K2 are constant for a given chemical period (P), but vary from period to period. The value of K, increases irregularly with increasing difference between the period of the solvent and solute element (AP), being zero when AP is zero. The value of K, appears to have no obvious periodic relationship. All factors other than valence that affect resistivity are gathered in these coefficients. Because of the nature of the coefficients, Eq. 1 is of limited use in estimating the effects of solute additions on resistivity unless a large amount of experimental data are already available on the systems involved. It is the purpose of the first part of this report to investigate the factors that may be included in the coefficients of Linde's equation. On this basis, it is hoped that the relative effects of solute additions on resistivity can be better estimated from basic data, leading to a more convenient alloy design procedure. It is well 10,11 that phenomena that decrease the perfection of the periodic field in an atomic lattice, such as the introduction of a solute atom or strain due to deformation, will also increase the electrical resistivity. Thus, in an effort to relate changes in electrical resistivity to alloy composition, it appears appropriate to consider the atomic characteristics related to solution and strain hardening

Jan 1, 1955

By W. D. Robertson, E. Scala

Electrical resistivity of a number of pure liquid metals and alloys has been measured as a function of temperature and composition. The data show a close correspondence between the liquid and solid states with respect to thermal, structural, and compositional relationships. STUDY of diffuse scattering of X-rays by liquid metals1 has demonstrated that some degree of order persists above the melting point and that the atomic constitution of liquids is more analogous to the solid state than to a gas. Consideration of the change in various thermodynamic properties that accompanies transformation from solid to liquid leads to the same general conclusion.² evertheless, specific knowledge of the constitution of the liquid state is limited when compared with that of the solid and gaseous states owing, in part, to the difficulty of interpreting the X-ray data and in part to the indirect character of deductions based on thermodynamic data. Of the structure-sensitive physical properties, electrical resistivity of liquid metals is perhaps the most easily interpretable, particularly when comparison is made with the solid state. According to the available data, most of which was obtained prior to 1914, cubic metals exhibit a normal linear and positive temperature dependence in the liquid state and the data have been accepted without question. The divalent metals, zinc and cadmium, are apparently exceptional in that resistivity is not linear with temperature and it passes through a minimum. Mercury also exhibits a nonlinear temperature dependence but the slope is always positive. Important comparisons with magnesium could not be made at the time the following work was undertaken because the necessary experimental data were not available. Data on liquid solutions are very limited and are largely from a single source."-" Based on Bornemann's work, Norbury7 pointed out that the resistivity of liquid copper alloys appeared to be dependent on the valence of the solute and that the increase due to the solute was of the same order as in the solid state. However, large extrapolations were involved and no quantitative relationship was developed. Therefore, before proceeding with additional deductions regarding the resistivity of liquid metals and interpretation in terms of structure, it appeared necessary to verify and extend the previous work with particular emphasis on divalent metals and the effect of alloying elements of different valence in dilute liquid solutions. The present work includes: 1—A determination of the resistivity of high purity liquid magnesium, tin, indium, copper, zinc, and cadmium as a function of temperature. 2—The effect of 1 atom pct of copper, cadmium, aluminum, tin, antimony, and

Jan 1, 1954

By L. Delisle

This work represents the first step of an attempt to test the applicability of the electron microscope to the study of subgrain structures in copper. Observations on annealed and deformed single crystals and polycrystalline samples of copper are described. IN the course of study of the structure of fine tungsten wires and tungsten rods with the electron microscope, well defined subgrain structures were observed. The size, size distribution, and orientation uniformity of the etch figures varied widely in different samples. Figs. 1 and 2, electron micrographs of a tungsten wire and of a tungsten rod, respectively, are illustrations of the difference in size and size distribution of the etch figures in different samples of the same metal. The observed differences, as pointed out in a previous paper,' appeared to be related to the heat and mechanical treatments of the samples. They were also consistent with the results reported in the literature on the mosaic structure of metals.' For that reason a program of research was initiated in an effort to obtain more systematic evidence of the possible relation of heat and mechanical treatments to the subgrain structure of metals as observed in the electron microscope. The purpose of this paper is to present observations made on the effect of cold work on the subgrain structure of copper. Procedure Starting Materials: Copper was the metal studied because it can be obtained in a high degree of purity, much information is available in the literature on its properties and its response to cold work and heat treatment, it shows no allotropic change, and it is sufficiently hard to be handled without great difficulty. Two groups of specimens were used: 1—single crystals cast from spectroscopically pure copper and 2—polycrystalline samples of oxygen-free high conductivity copper. Single crystals were studied because it was hoped that the elimination of a number of variables, such as grain boundaries, orientation differences, degree of purity, would simplify the problem and perhaps permit a better understanding of the phenomena that would be observed. The polycrystalline samples were designed to give a general picture of the changes considered. The single crystals were made of copper which analyzed spectroscopically to better than 99.999 pct Cu. They were cast in vacuum, by the Bridgman method, in crucibles made of graphite with a maximum ash content of 0.06 pct. The mold design is shown in Fig. 3. It permitted casting crystals of the size and shape required for the experiments, so that the danger of introducing cold work in the original samples by cutting or other machining would be eliminated. The polycrystalline samples were pieces, 3/4 in. long, cut from a rod of oxygen-free high conductivity copper, % in. in diameter. A flat surface, 1/4 in. wide, was milled along the rods, polished, and etched. The samples were then annealed in vacuum at 850°C for 1 hr. Polishing and Etching: Work previously done on tungsten,' polished mechanically and etched chemically," had shown that: 1—the general appearance of the etch figures of a given sample was not altered by repeated polishings and etchings under similar conditions; 2—variations in the time of etching and the concentration of the etchant changed the definition of the etch figures, but did not alter their general size nor orientation distribution within the limits of observation. Further work confirmed the reproducibility of the subgrain structures observed in, 1—single crystals and polycrystalline samples of copper when polishing and etching were repeated under similar conditions, and 2—specimens of tungsten and polycrystalline copper when electrolytic polishing and etching were substituted for mechanical polishing and chemical etching, respectively. On the strength of these observations, it was felt that, if conditions of polishing and etching were kept constant, changes observed in the subgrain structure of a sample upon deformation and annealing would be attributable to such treatments. For that reason the conditions of polishing and etching were kept as constant as possible. The single crystals were polished electrolytically in a bath of orthophosphoric acid in water, in the ratio of 1000 g of acid of density 1.75 g per cc to 1000 cc of solution, under a potential drop of 1.6 to 1.8 V. Electrolytic polishing was selected to prevent the formation of distorted metal in polishing. The same samples were etched by immersion in a 10 pct aque-

Jan 1, 1954

By John D. Verhoeven, Edward E. Hucke

An experimental technique has been developed for the measurement of the rate of electrotransport to a precision of 1 pct in the molten Bi-Sn system. Results are presented across the phase diagram from 30°C above the liquidus to 520°C. A phenomenological description is presented and the data on Bi-Sn are thereby converted to the effective charge, Analysis of the data indicates the true charge of the migrating tin species in bismuth to be +1.8 at 500°C. The data agree well with the recent work of Belashchmko. fieldintensity. The resistivity of the molten Bi-Sn system is also presented. WHEN an electric field is applied to a metal, a relatively large current is produced due to the ease with which electrons move within the metal. The electric field also produces a force on the ions which might cause them to migrate. Such a migration of metallic ions has been detected under certain conditions. This electromigration of the metallic ions will be referred to herein as electrotransport. The study of electrotransport in metallic systems has a rather long history dating back to the initial work of Gerardin1 in 1861. General reviews of electrotransport in liquid and solid metals have been published by Schwarz,2 Jost,3 Seith,4 and Heumann.5 In reviewing the literature it is apparent that the theoretical developments are not yet complete. The difficulty arises in describing the momentum transfer between the migrating ions and the electrons. In addition, there is a lack of good quantitative electrotransport data at high temperatures, especially in liquid metal systems. This work was undertaken to establish quantitative electrotransport data in a liquid metal system which might serve as a critical test of the theoretical developments. The Bi-Sn system was chosen for analysis, and the rate of electrotransport was determined as a function of temperature and composition. THEORETICAL CONSIDERATIONS If all of the atoms of a metal are considered to be positively charged due to the loss of electrons to the

Jan 1, 1963

By E. J. Whittenberger

OVER the past two decades, the Cr-Ni stainless steels, popularly termed 18-8 steels, have been used in ever increasing amounts in the aircraft, automotive, chemical, transportation, and building industries.' The contribution of nickel to these steels is associated with its strong austenite forming tendencies which assure good hot and cold forming characteristics, and with its beneficial influence upon corrosion resistance, especially in oxidizing acids. The adjustment of the chromium and nickel contents has been used advantageously in controlling the rollability and the response of the mechanical properties of these steels to cold reduction. With the advent of the jet age and the government stockpiling of nickel, considerable emphasis has been placed upon the development of satisfactory low nickel or nickel-free substitute austenitic stainless steels, especially for nonmilitary applications.2 Fer-ritic 17 pct Cr (AISI Type 430) stainless steel has been effectively substituted for the austenitic grades in applications with mildly corrosive exposure. However, the limited formability and ferromagnetic behavior of this grade do not permit its satisfactory use for many applications. To achieve better formability in a nenmagnetic steel and conserve nickel, a number of low nickel substitute austenitic grades of stainless steel have been developed in recent years. In most of these steels, manganese and nitrogen have been substituted for a portion of the nickel content, for example, 17 pct Cr-4 pct Ni-6 pct Mn (AISI Type 201), 18 pct Cr-5 pct Ni-8 pct Mn (AISI Type 202), and 16 pct Cr-1 pct Ni-17 pct Mn, all of which contain approximately 0.15 pct N.1 The precipitation of ferrite when large ingots of these Cr-Ni-Mn-N steels are heated to temperatures high enough for conversion to slabs on conventional universal slabbing mills has seriously affected the hot workability of these steels. This has been especially true in the case of the 16 pct Cr-1 pct Ni-17 pct Mn (sometimes referred to as 15 pct Cr-1 pct Ni-15 pct Mn) composition." Because it is possible that little, if any, nickel will be available for nonmilitary stainless steel applications in the event of a national emergency, a need has existed for a nickel-free austenitic stainless steel containing hot and cold forming characteristics similar to 18-8 steels and corrosion resistance at least comparable to the 17 pct Cr ferritic steel. The aforementioned deleterious effect of ferrite upon the hot working characteristics of the 16 pct Cr-1 pct Ni-17 pct Mn, steel indicates the importance of establishing the austenite-austenite plus ferrite phase boundaries as the first step in the development of nickel-free or low nickel substitute steels. This paper presents the data collected in an elevated temperature constitutional study of the Cr-Ni-Mn-N system at the South Works of the United States Steel Corp. Experimental Procedure The 300 compositions used in the investigation were melted in a 26 Ib induction furnace and cast in 3 14 in. diam by 18 in tall hot-tcpped ingot mo1ds.

Jan 1, 1958

By C. H. Li, R. J. Stokes, T. L. Johnston

Sodium chloride crystals are brittle in the presence of surface microcracks. Water immersion leads to an enhancement of ductility which is retained indefinitely in dry air providing the crystal is dried without leaving a surface precipitate. Storage in a damp atmosphere results in a surface reaction and consequent rembrittlement. Although elimination of surface flaws in ionic solids does not necessarily enhance ductility, there is always a change in the location of the fracture origin. EVER since the original experiments of Joffel there has been controversy over their interpretation and no satisfactory generalized explanation has yet been given. It is generally admitted that when sodium

Jan 1, 1961

By Benny J. Nelson

AS a part of the fundamental research program of Aluminum Research Laboratories, some data were obtained on the ternary system Mg-Al-Mn. As very little information on the magnesium corner of this diagram has heretofore been published, it seems desirable to make available the values found for the liquidus and solidus surfaces of this system. Procedure The settling procedure was used for the determination of the liquidus compositions. Metallo-graphic examination of quenched samples, and stress-rupture upon incipient melting, were used for the solidus determinations. The settling procedure has been described in a previous paper.' Briefly, this method involved saturating the alloy with manganese at a temperature substantially above that at which the sohbility was to be determined, then cooling the melt to the latter temperature, and holding it at that temperature for a substantial period of time. Samples for analysis .were carefully ladled from the upper portion of the melt at hourly intervals during the holding period. After the ladling of each sample, the melt was stirred to redistribute some of the manganese that had already settled, because it appeared that when the latter particles of manganese again settled, they aided in carrying down more of the manganese and thus hastened the attainment of equilibrium. The melts were prepared and held in a No. 8 Tercod crucible holding approximately 4 lb of metal. The manganese was added either in the form of a prealloyed ingot (Dow M) containing about 1.5 pct Mn or by the use of a flux (Dow 250) containing manganese chloride. In calculating the flux additions, it was assumed that the manganese introduced would be equal to 22 pct of the total weight of the flux. Temperatures were measured with an iron-constantan thermocouple enclosed in a seamless steel tube, the lower end of which was welded shut. This protection tube also served as a stirring rod. The samples ladled from the upper portions of the melts at the various intervals were analyzed for aluminum, manganese, and iron. When making the alloys which were to be used for the determination of the solidus, 2½ in. diam tilt mold ingots were cast, scalped to 2.0 in. in diam, and extruded into ? in. diam wire. The principal impurities in the melts for this investigation were iron and silicon; their total not exceeding 0.03 pct. Portions of the wire, approximately 2 in. in length, were enclosed in stainless steel capsules for protection from the atmosphere. Bundles of these capsules, with a dummy capsule containing an iron-constantan thermocouple, were heated inside a large steel block (acting as a heat reservoir) in a closed circulating-air type electric furnace. At ap- propriate times, the capsules were removed and quenched in water. The wires were examined metallographically to determine the temperature of initial melting. Short times at temperature were used at the beginning for wire specimens of all alloys to obtain quickly the approximate temperatures at which melting could be first observed. When approximate solidus temperatures had thus been determined, equilibrium heating was attempted. This equilibrium heating consisted of an 8 or 16 hr period at a temperature, about 50 °F below the lowest temperature at which melting occurred when short heating cycles were used, followed by further heating for 1 hr periods at consecutive 10" higher temperatures. The theory for the method of stress-rupture at incipient melting has been well covereda and its limitations are recognized. Thus, if the interfacial tensions are such that the first minute quantity of liquid is "bunched up" at the grain boundary junctions instead of spreading out along the grain boundaries,³ temperatures higher than the solidus are required before melting will be manifested by rupture of the specimen. This point will be elaborated later. Specimens of the wires with a reduced section (approximately 1/16 in. diam) were suspended vertically in a tubular furnace. The setup used is shown in Fig. 1. The clamp holding the specimen was made from alumel thermocouple wire and the thermocouple was thus completed across the specimen by attaching a chrome1 wire to its lower end. Temperatures were read from a Speedomax recorder used in conjunction with a calibrated thermocouple. The small weight attached to the specimen and a vibrator attached to the furnace tube, to aid in distributing the molten constituent along the grain boundaries, were used to bring about rupture at a temperature closely approximating the solidus. The specimens were heated at a rate of about 5°C per min. The rupture of the specimens was indicated both by sound and by the action of the recorder. An argon atmosphere containing a small amount of SO² was used for protection of the specimen. The assembly was taken out of the furnace immediately following rupture and the specimen removed. Some of the broken specimens were examined metallographically and will be referred to later. Results and Discussion Fig. 2 shows a set of typical time-composition curves for liquid samples of the Mg-Al-Mn alloys used for the settling tests. The data as presented

Jan 1, 1952

By R. M. Hudson

Equilibrium constants have been determined for the jeaction C (in steel) + H2O = CO + H2 as a function of carbon content (0.013 to 0.74 ujt pct) and temperature (1200° to 1800°F) by using a flow system in which the water-vapor content of various gas mixtures containing hydrogen, carbon monoxide, and nitrogen was systematically controlled. The equilibrium partial-pressure ratio K1 = PcoPH2/PH2O was obtained for two low-carbon steels and four Fe-C binaries from a series of runs made at each temperature. By comparison of Ki with the analogous ratio for the reaction C (graphite) + H2O = CO + H2, the activity of carbon in steel (ac) was comDuted for each temperature and carbon content studied. From these data values of the equilibrium constants for two other carburization-decarburiza-tion reactions of importance were estimated. The chemical compositions of the two steels and four Fe-C binaries used are given in Table I. Specimens of Steel E were cut from a 0.035-in.-thick commercially box-annealed and temper-rolled lot of a low-carbon low-metalloid steel. Steel F was a commercially hot-rolled rimmed steel that was cold-reduced at the Laboratory to a 0.059-in. thickness. Fe-C binaries A, B, C, and D were vacuum-melted and poured, hot-rolled, pickled, and cold-reduced at the Laboratory to furnish specimens 0.035 in. thick. Sheet specimens were cut from all lots to approximately 3/4 by 3 in. by gage and were vapor-degreased with trichlo-roethylene and weighed before heat treatment. Heat treatments were carried out in a hinged-top Hevi-Duty electric furnace with a Vycor-glass tube. When the furnace was at the desired

Jan 1, 1965

By Larry Kaufman

A useful method for estimating the entropy of NaCl type compounds has been developed by combining the Debye theory of specific heat with the Lindemann formula. This method pemnits calculation of the entropy from a knowledge of the atomic weight, atomic volume, and melting point of the compound. Comparison of the calculated and observed entropies of 35 NaCl type compounds over a wide temperature range yields good agreement thus establishing the usefulness of the method. Application of the method to other crystal types is also discussed. KUBASCHEWSKI~ has suggested a useful method of estimating the enthalpy of formation of inter-metallic compounds from a knowledge of the crystal structure, lattice parameter, and heats of sublimation of the component elements. This method is quite valuable in predicting the tendency for compound formation in a given system and is of particular value in considering systems for which no experimental data are available. In order to supplement the enthalpy calculation afforded by Kubschewski's method, it would be of value to be able to estimate the entropy of compounds so that the entropy and free energy of formation could be evaluated. Although a considerable body of such data determined by calorimetric and equilibration techniques exists at present, the difficulty of making such measurements (particularly at high temperatures) and the ever increasing number of new compounds which are of interest make the desirability of such a computational method clear. This paper deals with a method for calculating the entropy of a compound from a knowledge of the atomic weight, atomic volume, and melting point. Since these data are generally easier to determine and therefore more readily available than thermo-dynamic data, they should provide an additional tool in estimating the stability and reactivity of compounds at high temperatures. I) THE ENTROPY OF PURE METALS As a first approximation, the specific heat of a pure metal can be represented by a sum of terms where C,($/tI is the vibrational specific heat calculated on the basis of the Debye theory, lo4 TC, (0/T) is an approximation to the Cp - C, correction. T is the electronic specific heat, and Cp (p) is the magnetic specific heat. Although the Debye theory does not provide a complete description of the vibrational specific heat, i.e., the Debye temperature 0 is not a constant, it does provide a good approximation when used properly. Blackman6'7 has written several reviews on this subject discussing the different values of 0 determined by various methods (specific heat, X-rays, elastic constants, electrical resistivity, and so forth), as well as the observed temperature dependence of 0 with T. In the present case the 0 in Eq. [I] naturally refers to the specific heat 0. The variation of Debye 0 (as determined from specific heat measurements) with temperature is observed to exhibit its most radical behavior at very low temperatures. At higher temperatures the 8 values are observed to become more nearly constant. Blackman6'7 has presented the variation of 0 with temperature for several metals. His tabulation is reproduced in Table I. Reference to Table I shows that above a certain temperature level, 0 remains fairly constant. The value of 0 which is most useful in computing the specific heat, and hence the entropy, is the value of 0 in the range 0/6 « T « 0. In 1910, Lindemann' suggested a relation for estimating Debye 0 values. According to this formula, where V is the atomic volume, T is the melting point in 'K, M is the atomic weight, and CL is a constant. Mott and Jones9 present this formula and state that there is at present no satisfactory theoretical explanation, but that it does have an empirical basis. By using the valuesof 0 in Table I along with known1' values of M, T, and the molar volume VM, where the Lindemann constant can be evaluated. This

Jan 1, 1962

By John J. Gilman

F many years it has been suspected that a correlation existed between pits produced by etching and the density of dislocations in crystals. In 1953, the interest in this correlation was greatly stimulated when Vogel et al' demonstrated a 1:l correlation between etch pits and dislocations in low angle grain boundaries of germanium. About two years ago, the author began attempts to produce etch pits at dislocation sites in zinc mono-crystals. These attempts met with success a year or so ago,' and since that time a survey has been made of the etch pit patterns that are associated with various modes of deformation in zinc monocrystals. Unfortunately, the etching technique that will be described is too delicate for use as an accurate quantitative tool. Nevertheless, it has been of considerable assistance in clarifying the structure of plastically deformed crystals. Technique Many attempts were made to produce etch pits in high purity zinc crystals. It was found that several techniques would produce surface grooves at grain boundaries that had misorientation angles greater than about l°, but no technique was successful for boundaries with smaller angles. Among the things that were tried were: electrolytic CrO, (both etching and polishing concentrations), HNO, in various water and alcohol solutions, dilute HC1, many 0-0,-Na,SO, solutions, Cr0,-HC1 solutions, picral, electrolytic Na,S,O,, and electrolytic NaOH and KOH. A few attempts to use cathodic sputtering and thermal etching were also made.

Jan 1, 1957

By J. Bulina, J. T. Brown

SINCE the inception of vacuum cold-hearth arc melting of high-temperature alloys (about 4 years ago) it has been theoretically reasoned and qualitatively found that better mechanical properties are often obtained from alloys which have been melted by this practice. Mechanical properties are emphasized in this evaluation because of their importance in design. Vacuum-arc melting has an intrinsic advantage in that it creates an ingot grain structure which is easier to forge and thus leads to greater yields for the finished product. The greater yield naturally gives the process an economic basis when the cost per pound of alloy is rather high. However, the question stressed here is, does the process create an alloy with inherently better properties? There are several theoretical reasons for suspecting that it does. One is the vacuum refining which should permit lowering the content of harmful residuals with high vapor pressures. A second is the vacuum process which should preclude the formation of harmful distribution of oxides, nitrides, or carbo-nitrides, by keeping the amount of such constituents formed to a minimum. A third is the vacuum-arc cold-hearth method which prevents the possibility of pick-up of harmful elements from a refractory crucible since none is used. Fourth, the rapid solidification rate, induced by having a cold mold and only small amounts of molten metal at any one time, minimizes the segregation of elements like titanium, aluminum, and molybdenum from the top to the bottom of the ingot. This rapid solidification rate also decreases the tendency of the above elements to form undesirable phases by segregating locally. Another less frequently considered vacuum-arc melting advantage is the inherently finer-grained ingot structure which can be produced. Finer initial grain size leads to better forgeability, and means that after relatively small reductions to billets or final forged products a finer-grained material is obtained, which, by virtue of its finer grain size, would usually have greater stress-rupture ductility.

Jan 1, 1960

By T. S. Plaskett

A striated structure perpendicular to the growth axis was observed by the copper-decoration tech-nique in dislocation-free, .float-zoned silicon crystals. The striations, which were spaced about 100 p apart, fitted the relationship d = f/u , where d is the spacing, f is the growth rate, and u is the crystal rotation rate. Each stria was resolved into an UNDOPED silicon crystals pulled from quartz crucibles by the Czochralski technique usually exhibit a striated structure perpendicular to the growth axis.'-' This structure has been attributed to oxygen segregation, with the oxygen being introduced from the quartz crucible. If the crucible is rotated, the level of oxygen contamination has been reported as high as 10° atoms per cu cm.10 These striations are similar to solute striations commonly observed in doped Czochralski-grown crystals. The periodic nature of the striations is caused by a periodic variation in the growth rate",12 which is attributed mainly to thermal gradients in the melt.13 A finer striated structure14 attributed to constitutional supercooling is sometimes observed between the coarse striae. The oxygen striations have been observed by infrared transmission techniques,' by the copper-decoration technique,' by X-ray diffraction microscopy,6-8 and by 9 p absorption measurements3 on crystals pulled from the melt both with and without dislocations. In this investigation float-zoned dislocation-free crystals were examined by the copper-decoration technique. The level of oxygen for float-zone material is less than 1016 atoms per cu cm the lower limit of detection by 9 p absorption measurement. EXPERIMENTAL TECHNIQUE The crystals were grown by the float-zone process with the rf heating coil outside of the quartz envelope containing the silicon. All float zoning was done under an atmosphere of purified helium. The Dash technique15 was used to grow the crystal dislocation-free. This involves growing the crystal initially with a diameter between 2 and 3 mm and at array of starlike precipitates of copper. The strucLure was not .found at the surface tor a depth of about 1.5 mm, or in a region of similar width ahead of a dislocation network. The structure is postulated to consist of vacancy clusterings or dislocation loops. very rapid rates, about 20 mm per min, for a distance of about 3 cm. The diameter of the crystal is then increased to the diameter of the source of silicon, which in this case was about 19 mm. Because of the arrangement of the apparatus, the zone was passed downward rather than upward, contrary to the standard float-zoning practice. Also, the source was rotated rather than the seed. ziegler17 has made dislocation-free crystals by a similar technique but has passed the zone upwards. The starting material was zone-refined and had a p-type resistivity of 150 ohm-cm. The major impurity was boron; the total impurity excluding the boron was reported by the supplier (Dow-Corning) to be typically less than 2 x 1013 atoms per cu cm. The crystals were examined by the Dash copper-decoration technique18'19—a method in which about 10" atoms per cu cm of copper are diffused at a temperature between 900" and 1000°C into silicon which is then quenched to room temperature. On quenching, the copper precipitates on crystalline defects which are then visible when viewed by transmission infrared microscopy. The photomicrographs shown were taken either of the infrared image tube screen or directly on infrared film. All sections prior to decorating were chemically polished and, for some sections, given a sirtlZ0 dislocation etch-pit examination. After decorating, the samples were mechanically polished. RESULTS A photomicrograph, taken in transmission of a decorated cross section, is shown in Fig. 1. The portion of the section shown is near the surface of the crystal. The entire cross section showed no dislocation etch pits after being given a Sirtl etch treatment. It is seen that the copper precipitated randomly. Each precipitate, as has been reported by others, was found to have a starlike structure.

Jan 1, 1965

By Jack Washburn

The c axis indentations in zinc crystals were shown to undergo 100 pct strain recovery on heating. The mode of deformation and the details of the polygonization and collapse of indentations were found to be consistent with a number of predictions of dislocation theory. IT has been observed that indentations of the type made by Jillson in zinc crystals can, under certain conditions, be macle to shrink in size and even completely disappear on heating the crystal, Fig. 1. The details of the process are interesting in that they afford a simple experimental illustration of several predictions of dislocation theory. The indentations were made as shown in Fig. 2 by dropping a 0.2 g indenter with a hemispherical point on a cleavage face of a. zinc crystal. Such indentations can be made through relatively great thicknesses in the c-axis direction. The material close to the struck surface, where mechanical twinning and complex distortion occurred, was removed by cleavage. The remaining crystal, which was about 1 cm thick, contained a dent that was accurately round and almost the same size on both top and bottom surfaces. The indentation had the form of a flat cone of half angle approximately 89°, having a rounded apex and a rounded intersection with the unde-formed part of the crystal. The mode of deformation was primarily slip on the three equivalent systems in the basal plane. Six types of dislocation loop (plus and minus in each of the three directions) as shown in Fig. 3 are required by the geometry of the indentation. The maximum shear strain in the radial direction amounted to approximately 2.5 pct. Slip was distributed in the usual, microscopically nonuniform, manner among the possible parallel slip planes. When an indentation was made close to a (2110) surface, three separate sets of interpenetrating slip bands appeared as shown schematically in Fig. 3. A small biaxial tension strain in the plane of (0001) is also a geometrical necessity for the formation of an indentation. If the sides of the dent are tilted l° with respect to the undeformed surrounding crystal, and the surround-

Jan 1, 1957

By M. J. Spendlove, H. W. St. Clair

An automatic surface-follower mechanism was used to measure the surface temperature and the rate of evaporation of molten zinc while undergoing distillation at low pressure. At pressures of 50 to 100 microns Hg, the rate of evaporation may be 60 to 80 pct of the theoretical maximum corresponding to the measured temperature. The rate decreases rapidly at pressures above 100 microns Hg, to only 7 pct of the theoretical maximum at 2000 microns Hg. Measured temperature gradients at the surface are in agreement with theoretical gradients calculated from the heat of vaporization, rate of evaporation, and thermal conductivity of molten zinc. THIS paper covers a series of careful measurements of the rate of evaporation of molten zinc at low pressures, undertaken as a part of a general investigation on the separation of metals by distillation.' Preliminary results on the rate of evaporation of zinc were described in a previous publication.' The results of the preliminary study are as follows: 1—The observed rates of evaporation of zinc were as high as 0.60 g per sq cm per min, corresponding to 75 pct of the maximum rate as determined by the temperature of the metal. 2—The evaporation rate was decreased from 0.30 g per sq cm per min at a pressure of 25 microns to only 0.09 g per sq cm per min when the pressure was increased to 5 mm. 3—The temperature of the surface during rapid evaporation may be 35" to 100°C less than the main mass of metal. Observed temperature gradients at the surface are consistent with those calculated from the heat of vaporization and the thermal conductivity of liquid zinc. In the preliminary experiments, the rate of evaporation was determined from the loss in weight of the metal in the crucible during the test. This was compared with the weight of the condensate. Obviously, this method was subject to error because of some uncertainty as to the period of evaporation that should correspond to the loss in weight. Furthermore, this method did not permit taking into account probable variations in the rate of evaporation during the test. A more serious shortcoming was that the temperature at the evaporating surface was not known, so the maximum theoretical rate of evaporation could not be calculated accurately. To improve the precision of rate measurements and to allow accurate measurements of the surface temperature, a refined experimental technique was developed whereby the temperature and position of the surface could be measured accurately. This was done by an automatic surface-follower mechanism that gave a continuous record of the surface temperature and the amount of zinc evaporated. These data permitted an instantaneous determination of the actual rate of evaporation and the surface temperature throughout the test. Pure zinc (99.99 pct Zn) was used in all experiments. Description of Distillation Furnace The details of construction of the distillation furnace are shown in Fig. l. The furnace is enclosed in a large quartz tube open at one end. The open end of the quartz tube is sealed by means of a water-cooled head. The entire enclosure is vacuum tight. The connection between the water-cooled head and the quartz tube is sealed by means of a rubber gasket which fits tightly against the smooth end of the tube. The metal is heated by induced current set up in the molten zinc and the walls of the graphite crucible by means of a water-cooled induction coil outside the quartz tube. High frequency current is passed through the induction coil from a 6 kw mer-

Jan 1, 1952

By W. A. Krivsky, R. Schumann

Relaxation calculations were made to find the temperature distribution in the refractory wall surrounding a tuyere pipe. One set of boundary conditions approxiwated those of a standard copper converter while a second set represented a hypothetical,thermally insulated tuyere pipe. The pvonounced cooling effect of the tuyere is related to the problem of tuyere accretions and tuyere bunching. PARTICULARLY severe service conditions are encountered in the tuyere zone of a basic copper converter during the matte blow. High-temperature gradients result from the juxtaposition of the hot charge and the tuyere pipe carrying cold air. Refractory temperatures near the tuyere mouth vary rapidly with time when the blast is turned on or off. One of the most serious problems is the formation of magnetite accretions around the tuyere mouth, which occurs very rapidly under some operating conditions. As a result, continual punching of the tuyeres is necessary to maintain a satisfactory rate of converting. Thus, if some means could be devised for keeping the tuyere pipes fully open without necessity for punching, a considerable saving should result from reduction in labor cost, maintenance of full capacity operation, and reduction in refractory costs. Recent quantitative studies of the FeO-Fe,O,-SiO, system1 have shown that the solubility of magnetite in iron-silicate slags decreases rapidly with de- creasing temperature in the range of copper converting temperatures. Data taken from the previous work are shown in Fig. 1. Thus the formation of "horns" and accretions around the tuyere mouth can result simply from the freezing out of magnetite on the cooler areas of refractory immediately adjacent to the tuyere pipes. One might expect to find a correlation between the severity of magnetite formation and the area of refractory surface which is cooled below the freezing point of the slag (corresponding to magnetite precipitation). Magnetite itself has a very high melting point (1597°C) so that the accretions are for all practical purposes infusible. The accretions might be dissolved by increasing the silica content of the slag and at the same time raising the temperature, but this approach is not feasible because it results in rapid corrosion of the basic refractory lining of the converter as a whole. In order to understand better the problems involved in tuyere design and operation, it seems essential to have at least a semiquantitative idea of the thermal conditions in the tuyere zone. Accordingly, this paper presents results of thermal calculations for an idealized copper converter tuyere. These calculations follow the so-called relaxation

Jan 1, 1960

By Don M. Jackson, Robert W. Howard

Techniques for the epilaxial growth of single -crystal silicon carbide films on silicon were developed. The vapor-phase decomposition and bydrogen reduction of silicon tetrachloride (SiC14) and Propane (C3H8) resulted in clear films of silicon carbide, lip to seveval microns in thickness. The growth took place in a horizontal . silicon epilaxial reactor at 1100°C (pyrometer) at a rate of- 3000Å per minute. Electron diffraction and X-ray diffraction studies demonstrated that the films were single-cyrstal, ß -phase, or cubic silicon carbide. SiO2 film were used to mask areas of the silicon sur-lace in order that the silicon carbide might be grown in controlled geometries. Both n- and p-type films were grown on p-type silicon waters.. Heavily doped silicon films of the same conductivity type as the silicon carbide films were deposited over the silicon carbide in order to affect better probe contact to the structures. when n-type silicon carbide mesas were grown on p-type silicon substrates the de vollage-current relationships between films and substrates were that of junction diodes. These diodes showed a sensitivity to while light ill that the incident light increased forward- and reverse-satro,ation currents, P-type silicon carbide mesas grown on p-type silicon were ohmic rather than rectifying in their voltage -current relationship. No conclusions could he reached concerning heterojunc-tiou rectification in the structure. SILICON carbide is a semiconductor with many interesting properties. It decomposes at temperatures above 2200°C.1 It occurs in two general crys-tallographic forms—hexagonal (a Sic) and cubic (ß Sic)—with the cubic form having a forbidden-gap energy of 2.32 ev and the hexagonal form (specifically the 6H polytype) a gap energy of 2.86 ev.3 It behaves as an extrinsic semiconductor at temperatures approaching 5003C. It has been shown to have a high resistance to radiation damage4 and p-n junctions formed in Sic have been shown to radiate visible light under forward- or reverse-bias conditions. Epitaxial silicon carbide on silicon carbide has been successfully grown through the use of a variety of techniques, such as gaseous cracking of SiCL4 and CC4, nearly all of which require a deposition temperature above 1500°C.6 This paper will cover very recent work on the gas-phase deposition of highly ordered films of silicon carbide on high-quality silicon single-crystal substrates. The films have been shown to ex- hibit junction-rectification properties when geometrically isolated regions are electrically biased with reference to the silicon substrate. There will be no discussion of the mechanism of heterojunction rectification, but the methods of film fabrication, geometry control, and structural evaluations will be covered in detail. Electron diffraction, X-ray diffraction, and diode electrical properties were used to characterize the films and the junctions. GAS-PHASE DEPOSITION OF Sic The techniques for the deposition of silicon carbide films were a logical outgrowth of the standard silicon epitaxial process. The major premise followed was that, for any film to nucleate in an ordered fashion where there is considerable mismatch in lattice parameters (in this case 22 pct), an extremely clean, damage-free substrate surface must be presented to the gas stream. Thus a standard gas-phase HCl etching step was used to prepare the substrates for growth. A minimum of 5 µ of substrate-surface material was removed prior to the deposition of Sic overgrowth films. The techniques used for growing silicon carbide films were those of growing silicon alone, with the added injection of a hydrocarbon gas into the hydrogen and silicon tetrachloride gas stream. The hydrocarbon gases used thus far have been research-grade (99.99 pct) methane (CH4) and propane (C3H8). Propane ultimately gave the best results. The gas flows were controlled through a panel shown schematically in Fig. 1. A hydrogen main stream of 30 liters per min passed through the horizontal quartz-tube epitaxial reactor, while SiC14, C3H8, HC1, and doping gases were injected as side streams. The

Jan 1, 1965

By H. H. Klepfer, M. E. Snyder

UNTIL very recently, the commercial availability of the rare earths as metals has been very limited. Fabrication of mill products from these metals has not been studied in most cases. This note reports the results of the development of fabrication techniques for thulium. Thulium has a melting point of about 1550°C and a hexagonal close-packed crystal structure. It oxidizes in air to give a black oxide (TmzO,). The procedures for producing thin thulium foil were developed on one ingot weighing about 220 g and were subsequently applied in processing about a pound of metal to foil. Several alternates for the various fabrication steps were investigated and will be discussed. The metal fabricated was in the form of commercial chill-cast ingots 1 in. in diam and 2 in. long weighing approximately 220 g. Impurities in the ingots were reported by the vendor to be 4000 ppni tantalum, 2000 ppm calcium, 200 ppm nickel, 100 to 200 ppm iron, 100 ppm europium, and less than 100 ppm copper, lutetium, and ytterbium. In addition to these impurities, several salt-like inclusions as large as l/8 in. in diam were revealed along the center line of the one ingot sectioned. Preliminary tests indicated that small wafers cut from the as-cast ingot would not fabricate readily by rolling. Forging of copper-jacketed wafers was therefore attempted. At 1550"F forging was satisfactory but an apparent reaction of copper with thulium demanded investigation of lower temperatures. Therefore, the remainder of the test ingot was forged at 1450°F—with only minor cracking. All ingots forged were inserted into copper tubes of 1 in. 1D and 0.125-in. wall thickness. The jackets were sealed by flattening the ends of the tubes and welding under helium. Heating time at 1450°F was 30 min. Press forgings of 1/8 in. per pass were used, followed by 10 min reheats. When the ingots had been squared and reduced to 0.250 in. in thickness, the original copper jacket was stripped off and replaced by a new jacket in preparation for hot rolling. Hot rolling at 1450" F without edge cracking was readily accomplished after forging. Excellent results were obtained with 10 pct reductions of thickness followed by 5 to 10 min reheats. After reduction of thickness from 0.250 to 0.100 in. the copper jacket was removed. It was found, in fact, that hot rolling in air was possible. A tenacious black oxide similar to that seen on zirconium was formed during 3 min reheats at 1450°F. Reduction in air to 0.010 in. foil was possible taking 10 pct reductions per pass. The oxide coat formed during hot rolling in air could best be removed by sand blasting and pickling. Common pickling solutions containing polar solvents were found to attack the metal too rapidly and a concentrated nitric-hydrofluoric acid mixture attacked neither the oxide nor the metal. The most satisfactory pickling solution was 52 vol pct concentrated nitric acid-48 vol pct glacial acetic acid. After forging to 0.250 in., vacuum annealing and cold rolling was found to be another satisfactory alternate to hot rolling in copper jackets. After forging. a hardness of Rockwell B76 was found. Annealing in vacuum (2 X l0-5 mm of Hg) for 1 hr at 1200"F did not alter this value. Annealing at 1470"F for 1 hr brought the hardness down to R;]63. With cold rolling (5 pct per pass) the hardness returned to about R,1:76 after 20 pct reduction and edge cracking became noticeable. However, cold rolling to a total of 40 pct reduction in thickness (RI,,83) was possible before edge crack propagation became serious. Good surface finish was obtained, and the metal loss due to oxidation was minimized by cold rolling and vacuum annealing. Using this procedure the yield of 1.25 in. wide by 0.010 in. thick foil from a 1-in. diam ingot was about 40 pct.

Jan 1, 1961

By H. W. Weart, S. Sarian

The effect of temperature, impurities, and capillarity on the morphology of solid particles in a liquid matrix is investigated. For the NbC-liquid iron system, at least, it is found that those particles which are dissolving will assume a spherical shape; all others will take the equilibrium shape. An aniso-tropic-isotropic transition in the interfacial tension is observed at about 1725°C and is believed to be due to cooperative roughening of the interface on an atomic scale. The presence of boron as an impurity lowers the transition temperature. It is the purpose of this paper to present some observations on the shape of particles in a system consisting of solid particles of nonuniform size in a liquid matrix and to propose an explanation for these observations. Although qualitative only, these observations, which were made largely in the course of a study of particle-growth kinetics,' are presented separately to draw attention to them. They reveal the effect of particle size, temperature, and surface-active impurities on particle shape in the solid NbC-liquid iron system, and are unique in demonstrating for the first time the reversible effect of temperature and impurity concentration on the anisotropy of the energy of the solid-liquid interfaces involved. Some of the changes are readily demonstrated with the few particles whose shapes can be clearly shown in a single photomicrograph, but others are more subtle, and their recognition depends upon some experience and the study of a much larger sample. For this reason, a certain amount of judgment must be acknowledged to underlie their detection, but every effort was made to maintain proper scientific objectivity in evaluating the observations. The observations will be presented first, and then an explanation rationalizing them will be offered. Brief reminders of the relevant theories will be inserted into the discussion at the appropriate points. EXPERIMENTAL The observations were made of the shape of NbC particles dispersed in liquid iron. All specimens were prepared by pressing NbC powder and car- bony1 iron powder treated according to standard powder-metallurgical procedures. Except as noted, all specimens were heated to 1900°C for 1/2 hr. In investigating the effect of temperature, specimens were quenched from 1900°C to various lower temperatures, and held long enough (48 hr) to allow shape changes to be completed. Boron, in the form of amorphous powder, and oxygen, in the form of niobium oxide powder, were added as impurities to some specimens. Nitrogen was introduced into some specimens as an impurity by heat treating the samples under a partial pressure of nitrogen gas. Effect of Temperature. It was first observed that increasing temperature had a marked effect in rounding the corners and edges of the particles. At 1900°C the particles are all very nearly spherical and remain so even when the holding time is extended to 48 hr. As can be seen from Fig. 1, there is virtually no evidence of the presence of low-index planes. At 1300°C, however, most of the particles are prismatic with very slightly rounded edges and corners, Fig. 2. (In making these shape comparisons, attention should be focused on the larger particles, even though the smaller ones would be expected to be the first to change shape, owing to the shorter diffusion distances involved. The cause of the mixed particle shape will be explored later.) It was established that these structures are re-

Jan 1, 1965

By A. Yoshikawa, T. Sugeno

To detenmine the origin of the sharp fatigue limit in many ferrous metals, S-N curvces were determined in push-pull fatigue at 18.6 kc per sec at room temperature and - 67°C for various kinds of iron. These differed in carbon and nitrogen content and grain size (25 and 2500 grains per sq mm). Sharp knees appeared in the S-N curves of fine-grained specimens with high interstitial content at both temperatures. In coarse-gruined specimens, knees were not sharp .regardless of the interstitial content and the testing temperature. It is concluded from these results that the controlling factov for the appeavance of the sharp fatigue limit is not strain aging but grain size. It has already been mentioned that iron and some iron alloys show fatigue curves characterized by a more or less sharp knee and a fatigue limit.' Explanations of the origin of this characteristic behavior of fatigue in iron appear to fall into two broad classes. One is concerned with interstitial impurities through the aging or locking effect and the other is based on intrinsic plastic properties of iron. Levy and sinclair2 and Lipsit and Horne3 have suggested that the fatigue limit observed generally in iron and steel is due to interstitial solute aging during the course of fatigue. On the other hand, Thompson and wadsworthl have suggested that the fatigue limit may be analogous to the sharp yield point, in the sense that alternating stresses above the fatigue limit are necessary to break down the dislocation locking. However, there are some experimental results which cast doubt on the general validity of these suppositions.4-8 Ferro and Montalenti' have recently suggested that the typical form of the fatigue curve of iron seems to come from an intrinsic plastic characteristic of the iron structure. This suggestion is based on the experimental result that pure iron containing an insufficient amount of interstitial solute to alldw appreciable strain-aging effects still showed a fatigue limit at room temperature. Although systematic investigations on the effect of grain size are scarce, it is thought that this effect is important in relation to the sharp fatigue limit. In fact, sensible differences were observed by Oates and Wilson4 in fatigue behaviors of a low-carbon steel depending on grain size. Ideally, the fatigue limit will be defined as the upper limit of stress amplitudes at which the life is infinite. This ideal fatigue limit can never be determined because of the limited time available for experiments, and moreover scattered data points often leave the construction of S-N curves somewhat arbitrary. And, it is usually stated that a material has a fatigue limit when its statistical S-N curve becomes apparently parallel to the N axis with a more or less sharp knee. The meaning of the fatigue limit and the sharp knee will be discussed in detail in relation to the construction of S-N curves. The origin of this sharp knee was investigated by determining S-N curves of iron for various interstitial contents, testing temperatures, and grain sizes. EXPERTMENTAL PROCEDURE The material as received was a commercial-grade low-carbon steel in the form of 4-mm-diameter rods with the following composition in weight percent: C, 0.09; Si, 0.017; Mn, 0.43; P, 0.006; S, 0.012; Cu, 0.11. Five different groups of specimens were tested. The grain sizes and interstitial contents in each group are given in Table I. The specimens of group C-02 were decarburized by annealing for 100 hr at 700°C, followed by an additional 100 hr at 850°C in wet hydrogen. Specimens of groups N-2 and N-9 nitrided in a mixed atmosphere of ammonia and hydrogen after an initial treatment identical to group C-02. The specimens of group C-9 were prepared in a manner similar to those of group C-02, except heat treatment was carried out in a vacuum. After polishing their surface with emery paper, the specimens of all groups were given a final anneal for 30 min at 930°C and air-cooled in an evacuated quartz tube. For group C*-9 this

Jan 1, 1965

By E. R. Morgan

A series of manganese-rich ferromagnetic alloys has been studied. The alloys are based upon an interstitial solution of carbon in manganese. In each of the systems investigated, the ferromagnetic phase was found to be face-centered-cubic in structure. This magnetic phase is based upon the composition (Mn, X)4C; where X is a metallic element which has both a positive size factor with respect to manganese and a high positive valence. Typical elements of this type are aluminum, indium, and tin. Preliminary magnetic measurements indicate that the effective magnetic moment of manganese in the alloys is at least 1.0 Bohr magneton per atom. It is emphasized that the critical atomic separation beyond which ferromagnetism can occur will be governed by the electronic configuration of the manganese atoms. Lattice parameter measurements show that the ferromagnetic condition in the present alloys is associated with an increased atomic separation of only 3 to 6 pct compared with a hypothetical face-centered-cubic lattice of pure manganese. MANGANESE which is paramagnetic and not ferromagnetic in the pure state can become ferromagnetic if suitably alloyed. Under favorable conditions manganese can have an effective ferromagnetic moment of the order of 3.41,2 or even 4.0 compared with a moment of 2.2 Bohr magnetons for iron. The incidence of ferromagnetism is believed to be controlled by the degree of atomic separation.4,5 This is consistent with the fact that the atomic separation, when considered in terms of the ratio of the atomic diameter R to the diameter r of the 3d-shell, is larger in the ferromagnetic metals. The values of R/r calculated by Slater,' for metals of the iron group are given in Table I, and Slater has suggested that the critical ratio of R/r is 1.5. Consideration of the criterion of atomic separation alone would suggest that there are three classes of possible ferromagnetic manganese alloys: 1— Ordered structures, of either solution or compound types, such that manganese atoms are not near neighbors; 2—Structures in which the separation of near-neighbor manganese atoms is sufficiently increased by the presence of suitable solute atoms; and 3—Dilute solutions of manganese. Typical ferro- magnetic alloys of types 1 and 2 are the Heusler alloys and Mn4N, respectively. There are no recognized ferromagnetic alloys of type 3. In view of the proximity of the ratio R/r for manganese to the critical ratio of 1.5, the present work was concentrated on manganese-rich alloys rather than on ferromagnetic alloys of low manganese content, several examples of which are already known. Calculation of the distance of closest approach of manganese atoms in each of its allotropic forms, Table 11, indicates that the high temperature ? phase provides the most suitable basis for ferromagnetic alloys of high manganese content. However the high temperature forms of pure manganese cannot be retained on quenching, although a face-centered-tetragonal form apparently can be produced by electrodeposition. Since the critical separation which must be exceeded in order to obtain ferromagnetism is dependent upon the number of electrons in the d-shell, no exact critical value of R/r can be assumed. Not only will this separation vary from one pure metal to another, but it will also vary with the valency of the given type of atom when it is in an alloy. Raynor8 has shown that metals of the iron group probably

Jan 1, 1955