Search Documents

By T. V. Philip, R. E. Lenhart

C. G. Dunn(General Electric Research Laboratory)— It is well recognized that understanding of the formation of the cube-on-edge texture in annealed commercial cold-rolled Si-Fe strip is important industrially as well as scientifically. I would like to make a few remarks on this subject, discuss some points in the present paper, and list two remaining problems that appear to be particularly outstanding. Attention should be drawn to the paper by P. K. Koh,23 which also considers the kinetics of secondary recrystallization in 3 pet Si-Fe, produced commercially by the Allegheny Ludlum Steel Corp. Koh covered a much larger temperature range than do either the present authors or P. N. Richards, whose work is mentioned in a footnote. Koh found an induction period (of short duration) even for isothermal anneals at 1260oC (2300°F). Koh and Dunn24 also found a change in texture during the 1260oC anneal for the same material. On the other hand, the material in coil D, according to Philip and Lenhart, underwent no sec-condary recrystallization, only continuous grain growth. The grade classifications of the four coils studied would suggest, however, that all four materials had developed the cube-on-edge or (110) [001] texture, but to varying degrees. Since no data on this are given and since watt loss depends

Jan 1, 1962

By Jean Howard

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 Jean Howard

ThE formation of the (100) [001) texture in 3-1/4 pct Si-Fe strip was first reported by Assmus ef a1.l in 1957. Since then much experimental work has been carried out with a view to establishing the mechanism involved. The papers cited above state that the (100) [001] texture was developed in strip rolled from material melted and cast in vacuum. (The impurity content of the ingot is reported as 0.005 pct.) The present note records that similar results can be obtained in material processed by powder metallurgy. A processing schedule is described.which enables the texture to be formed in strip up to 0.010 in. thick, but there seems no reason why this should not be achieved in thicker strip, provided that large grains are developed after sintering. The materials were prepared from Carbonyl Iron Powder Grade MCP (particle size 5 to 30 p) of the International Nickel Co. (Mond) Ltd. The powder contains about 0.15 pct 0, 0.01 pct C, 0.004 pct N, <0.002 pct S, $0.005 pct Mg and Si, and 0.4 pct Ni— that is, it is substantially free from metallic impurities other than nickel, which is thought to be unimportant in the present work. The silicon powder was 99.9 pct purity, or material of transistor quality (ground in pestle and mortar). The mixed powders (3-1/4 pct Si to 96-3/4 pct Fe) are heated in hydrogen at 350" and 650°C to deoxidize the iron before sintering loose at temperatures between 1350" and 1460°C (depending upon the ultimate thickness of strip required) for up to 24 hr. The object of the high-temperature sinter is to develop a large grain size at this stage. Alternatively, the loose sintering can be done at a lower temperature followed by rolling or pressing and then annealing at temperatures between 1350" and 1460°C. Both methods produce large grains, which remain large throughout the process. The compact is then hot-rolled to approximately 1/8 in. with high-temperature interstage anneals if necessary. This step is taken to avoid intercrystalline cracking which would occur if the material of such large grain size were cold-worked. The bar is then annealed at 1050°C and reduced to its final thickness by approximately 50-pct reductions and 1050°C interstage anneals. Throughout the process the dew point of the hydrogen furnace atmosphere is maintained at about -40°C. Samples were annealed in hydrogen at various temperatures and times. Secondary recrystalliza-tion to (100) [001] was developed on the thinner material (i.e., up to 0.002 in.) by annealing in hydrogen at 1050" to 1200°C with a dew point of - 40°C or in vacuum (10-5 Torr) at 1050°C. With the thicker materials (i.e., up to 0.010 in.) the best results were obtained by annealing in hydrogen at 1200°C with a dew point of - 55°C. Complete secondary recrystal-lization to (100) [001] textures was obtained. Above these temperatures secondary recrystallization to (110) [001] tended to develop. The final annealing of samples was normally carried out with the samples placed between recrystal-lized alumina plates, but some experiments were performed with the samples suspended so that their surfaces were not in contact with anything except hydrogen, and these were equally successful in developing secondary crystals. An approximate determination of the proportion of material (before secondary recrystallization took place) having crystals with the (100) or (110) planes in or near the rolling plane showed that 11 pct of the sample had (100) and 16 pct (110). The method used for the determination is described below. A sample was annealed at a temperature just below the secondary recrystallization temperature and etched to reveal the (100) planes. The approximate area covered by crystals having (100) or (110) in or very near the surface was measured on the screen of a Vickers projection microscope. This was repeated for twenty positions chosen at random and a mean of the results calculated. The main hindrance to developing the secondary crystals in the thicker materials was the difficulty of obtaining a large enough initial primary grain size before secondary recrystallization. This was overcome by increasing the particle size of the silicon powder used. During the course of the work, it had been observed that the larger the grain size after sintering the more likely it was that the material would be successful in developing secondary crystals at a later stage. An experiment was therefore carried out to determine whether the material with the larger grain was more successful in developing secondary crystals due to the large grain produced at the sintering state per se or whether it was due to the greater reduction of silica brought about when the sintering temperature was raised in order to increase the grain size. A comparison was made between two compacts, one of which was made with silicon powder of 60 to 100 mesh, the other with silicon powder which was finer than 200 mesh. F?r this experiment, use was made of a phenomenon previously observed that the larger the particle size of the silicon powder employed in making a compact, the larger is the grain size of the compact. The silicon powder was ground

Jan 1, 1964

By H. Biloni

A metallographic study of solute segregation produced during controlled solidification of Sn-Pb alloys has previously been reported.&apos; It was found that the growth conditions which produced well-developed cellular substructures led to severe segregation, and even at solute concentrations of 0.2 wt pct Pb regions of a lead-rich phase were formed at the junctions of cell boundaries (nodes). Similar sample preparation and etching techniques have now been applied to several tin alloys at more dilute concentrations. The crystals were solidified uni-directionally under less-severe growth conditions so that the predominant interface structure was composed of two-dimensional cells.* Binary alloys with lead, using 9.999 pct Sn as a solvent, and ternary alloys with lead and bismuth or lead and antimony, using zone-refined tin as a solvent, were examined. The liquid was decanted from the growing crystals in all cases to reveal the interface structure and ensure that two-dimensional cells were being formed. However, since interference may arise from a liquid layer that clings to the interface during decanting,3 an examination of the segregation pattern was always made behind the interface. In all cases, the structure within the bulk of a crystal at greater than 1 cm behind the decanted interface was substantially the same as that nearer the interface. Representative photomicrographs of the dilute Sn-Pb alloy are shown in Figs. 1, 2, and 3. Most observations are similar to those of the previous investigation. For example, the solute concentration at the cell walls is higher than in the intercellular regions; where cell boundaries meet, the segregation is enhanced; the macromosaic or stria-tion substructure and grain boundaries show enhanced segregation; there is also a depletion of solute at cell walls near a striation boundary, as Spittle et a1.4 have observed at decanted interfaces. Not all these observations can be seen in the photos shown since attention has been focused on a special feature; as evident in the photomicrographs, cer-

Jan 1, 1965

By W. G. Pfann

The simultaneous segregation of two solutes during the directional solidification of an ingot is treated mathematically on the basis of simplifying assumptions. Expressions are derived for the difference in concentration of two solutes, and for the location and concentration gradient of a pn barrier formed in a semiconductor by the segregation of a donor and an acceptor. THE problem of normal segregation of a single solute during the freezing of an alloy has been treated mathematically by a number of investigators. Gulliver&apos; showed that coring increases the quantity of eutectic above that to be expected at equilibrium, for a system having limited solid solubility and, on the basis of simplifying assumptions, calculated the fraction of eutectic to be expected. Scheuer2 expressed composition of solid solution in terms of a distribution coefficient and the fraction solidified and compared experiment with calculation for the systems A1-Cu, Al-Zn, Cu-Sn, Cu-Zn. Hayes and Chipman,3 in a detailed study of segregation in a low carbon, rimming steel ingot, calculated distribution coefficients for a number of solutes in iron, compared calculated and experimental segregation curves, and discussed the effects of process variables such as rate of solidification and stirring. In the present paper a mathematical analysis is made of the simultaneous segregation of two solutes during the orderly freezing of a solid solution system, with particular emphasis on the difference in solute concentrations. Although the analysis is quite general and can be applied to the segregation of minor elements in alloys, it is directed in particular at the solidification of a semiconductor containing a donor and an acceptor. Normal segregation has unique aspects in a semiconductor, because of the ways in which donors and acceptors affect the electrical properties. The difference between two solute concentrations becomes of importance, as does also the gradient of the difference where the difference goes through zero, this being the concentration gradient of excess carriers at a pn barrier. A primary object of this paper is to extend the mathematical treatment of segregation to include these newer aspects which are of particular significance for semiconductors. One property of interest is the electrical conduc- tivity, which arises from the presence of donors &apos;or acceptors in solid solution. The conductivity may be either n-type or p-type depending on whether donors or acceptors, respectively, are in atomic excess. For both germanium and silicon, elements of Group V of the Periodic System, such as P, As, and Sb, are donors and elements of Group 111, as B, Al, In, and Ga, are acceptors.4-6 If a donor or acceptor is present alone in solid solution, the conductivity is proportional to its concentration." If both a donor and an acceptor are present, then the conductivity is proportional to the difference in their atomic concentrations. If the donor and acceptor segregate at different rates then a pn barrier in certain circumstances may form at some point in the ingot. Accordingly, equations are derived and illustrated which express the effect of segregation on: 1—the concentration of a single solute with a discussion of the assumptions; 2—the difference between two solute concentrations and means for minimizing its variation in an ingot; and 3—the location and concentration gradient of a pn barrier. The analysis is applicable to processes in which the entire charge is melted and then progressively frozen from one end. The method in which an ingot is solidified in a crucible7 and that in which a solidifying rod is pulled from the melt8 both fall into this category. Segregation of One Solute If a cylinder of molten alloy is caused to freeze slowly from one end, a normal segregation of solutes will usually occur, producing a lengthwise concentration gradient in the ingot. Depending on whether a solute raises or lowers the melting point of the solvent, it will become concentrated in the first or last regions, respectively, to freeze. If it is assumed that freezing is such that there is no diffusion of solute in the solid, complete diffusion in the liquid, and that k, the distribution coefficient, defined as the ratio of solute concentration in the just-freezing solid to that in the liquid, is constant, then the

Jan 1, 1953

By W. Rostoker

Single isothermal sections were constructed for the titonium-rich corners of the systems Ti-Fe-O, Ti-Cr-O, and Ti-Ni-0 with a view to locating the shape and disposition of the ternary intermediate-phase fields for the compounds isomorphous with Fe3W3C. IN a previous note,&apos; it was reported that a group of ternary intermediate phases existed which occurred in the general composition range Ti,X,O-Ti&O where X was one of the following elements: copper, nickel, cobalt, iron, or manganese. In two instances, binary intermediate phases occurred (Ti²Cu and Ti²Ni) which appeared to be isomorphous. The X-ray diffraction patterns of the phases could be indexed on a cubic lattice whose cell edge was of the order of 11 kX. Karlsson2 stated that the line extinctions and line intensities indicated that these phases were isomorphous with Fe³W³C and therefore belonged to the space group Oh. This author also noted that no unique phase of this type occurred at the composition Ti,Cr,O. This latter statement was accepted as true at the time the previous technical note was written.&apos; However, work continued along these lines and that reported here has shown that a phase isomorphous with Fe³W³C does exist in the vicinity of the composition Ti³Cr³O but not at or near Ti4 Cr²O. The lattice parameter of this phase was found to be 13.01 kX. Metallographic examination of both as-cast and annealed specimens of these supposed ternary oxide phases disclosed structures which were far from single phase in many instances. It was considered a point of some interest to locate more positively the composition coordinates of the single-phase fields. To this end, single isothermal sections for the systems Ti-Fe-O (1000°C), Ti-Cr-0 (1000°C), and Ti-Ni-0 (900"C) were constructed. Phase relationships were studied by both X-ray diffraction and metallotrra~hic methods. Allovs were prepared as 10 g buttons by melting in a noncon-sumable electrode, water-cooled copper crucible arc-melting furnace using a helium atmosphere. Oxygen was introduced quantitatively by the use of weighed amounts of TiO² or TiO. Iodide titanium and the purest available alloying metals were used. All results are discussed in terms of the nominal composition of the melts. Previous experience with making oxygen-rich alloys had shown that oxygen losses were not significant if a homogeneous melt was produced. Weight-loss measurements were used to check loss in metallics during melting. Anneals were conducted in evacuated Vycor bulbs, and annealing times ranged from 24 hr at 1000 °C to 48 hr at 900°C. To achieve equilibrium in certain sluggish alloys, seven-day anneals were used. Debye-Scherrer powder patterns were taken in a 14 cm diam camera using filtered CuK radiation (K² = 1.541232 kX; K., = 1.53739 kX). Specimens were prepared by crushing and screening previously annealed buttons. In general, the X-ray diffraction results were used to identify the major phases, while the metallographic examinations were more useful in indicating the number of phases present. By the use of the combined results and general fundamental rules governing ternary equilibrium, it was possible to construct isothermal sections in good detail. In some instances where auxiliary lattice-parameter data were available, it was possible to lay down tie lines and to locate more positively corners of three-phase fields. Isothermal Section of the System Ti-Fe-O at 1000°C The binary systems Ti-O³ and Ti-Fe4 have been published. Binary phases, TiO, TiFe, and TiFe2, are likely to be encountered in the titanium-rich corner of the system. The structures are NaCl (Bl), CsCl (B2), and MgZn, (C14) types, respectively. Although preliminary X-ray diffraction studies indicated the presence of a ternary phase in the composition range Ti,Fe,O-TiJ?e,O, metallographic examination of two specimens at these compositions revealed the presence of other phases. Thirty-eight alloys were used to delineate the phase boundaries of the isothermal section illustrated in Fig. 1. The five almost single-phase ter-

Jan 1, 1956

By Ken-ichi Hirano, B. L. Averbach, Morris Cohen, N. Ujiiye

The influence of plastic deformation in compression on the self-diffisivity of a iron has been measured in the temperature range of 742º to 885°C. The diffusivity is enhanced in proportion to the strain rate, but this dependence decreases with increasing temperature. The magnitude of the strain is relatively unimportant in this connection. Strain rates from 0 to -2 x 10-3 sec-1 were investigated, with a corresponding increase in diffusivity up to 3 x 103 times. The results are analyzed in terms of vacancy diffusion and the excess vacancies introduced during dejornzntion. It is concluded that grain boundaries are the main vacancy sinks in polycrystalline iron and that the vacancy lifetime is therefore dependent on the grain size. A typical vacancy lifetime at 800°C with a nzean grain-boundary intercept of 0.3 cm is of the order of 1 sec. LATTICE imperfections are considered to be the dominant mechanism of diffusion in close-packed crystals. Vacant lattice sites are usually the most important defect, but the enhancement of diffusion along grain boundaries and other short-circuiting paths has been well documented. Inasmuch as plastic straining produces a variety of defects, it would be expected that the application of stresses sufficient to cause plastic flow during a diffusion run might result in an increase in the self-diffusivity. Enhanced self-diffusion in a iron under compressive creep has been observed by Buffington and Cohen&apos; and subsequent coworkers.2 Lee and Maddin3,4 have also found an enhancement in the self-diffusivity of silver single crystals undergoing torsional deformation. On the other hand, Darby, Tomizuka, and Balluffi5 have not observed any significant diffusion increase in single crystals of silver deformed either in tension or in compression, although the results in this case may have been affected by the occurrence of recrystallization. With poly crystalline silver, however, Forestieri and Girifalco6 have found en- hanced diffusion during compressive creep. A small decrease in the self-diffusivity of zinc single crystals under compressive loading has been reported,7 but the applied stresses in these experiments were too low to cause appreciable plastic deformation. For the cases in which enhanced diffusion has been found, a number of common effects have been noted. Usually, the ratio of the diffusivity during straining, Ds, to the diffusivity in the unstrained condition, D,, varies linearly with the strain rate, and does not seem to be a function of the plastic strain, at least up to ? = 0.2. Values of Ds/Du of the order of 100 have been reported&apos;-4 but the enhancement decreases with increasing diffusion temperature. The activation energy for diffusion decreases markedly with increasing strain rate and approaches a level of roughly one-half the unstrained value. Most of the current explanations of these observations assume that excess vacancies are generated by the deformation, but rather large concentrations of excess vacancies and/or long lifetimes are required by these theories.&apos;,&apos;,&apos;,&apos; Since the original work on iron, there has been an important improvement in the compressive-loading technique.2 The early data were also analyzed without taking into account the motion of the coordinate system in the specimens undergoing plastic deformation. This problem has now been solved by a number of investigators (see Refs. 11, 12, and the appendix of the present paper). Consequently, the strain-diffusion runs on iron for the Ds measurements were repeated and extended over wider ranges of temperature and strain rate. In addition, the original diffusivity data9 on the unstrained iron have been corrected for the divergence of the beam in the counting system and new data points have been added, resulting in somewhat different values for the self-diffusi-vities in a iron.10 The new values for Du are used in this paper. A comparison of Ds and Du confirms the large enhancement in the self-diffusivity of iron during compressive creep, as previously reported.&apos;,2

Jan 1, 1963

By R. E. Grace, H. W. Schadler

DARKEN1 has established the theoretical relation between the self-diffusion coefficients and the Boltzmann-MatanO Or interdiffusion coefficient: D is the Boltzmann-Matano or interdiffusion coef-

Jan 1, 1960

By R. F. Mehl, C. E. Birchenall

SINCE Maxwell1 first considered the self-diffusion process in 1872 its importance in the kinetic theory of matter has been recognized. Until the discovery of isotopes in 1913, a direct measurement of this quantity seemed impossible. The only information that could be gathered indirectly was for gaseous systems for which the kinetic theory was well developed. When methods of direct observation be-came available, investigations on self-diffusion were carried out in condensed phases. In metals these data are presumably of potential importance in the study of recovery, recrystallization, creep, sintering, and related phenomena. Following the pioneer work of Von Hevesy2 and his collaborators, who determined the rate of self-diffusion in lead with the naturally occurring thorium B isotope, several metals have been investigated. Pb, Ag³, Au.6,5, and Cu6,7,8,9 are examples of isotropic crystals existing in only one phase modification. Anisotropy of diffusion has been demonstrated in Bil6 and Znll. This paper is a study of a single metal in two allotropic forms and also of self-diffusion in a body-centered cubic lattice. Despite the fact that the measurements in each of the papers cited on self-diffusion in copper seem to be internally consistent to about 10 to 15 pet, the curves reported by different authors differ by factors of 2 to 4 at the same temperatures. This uncertainty may account, in part, for the failure to establish a successful correlation between the self-diffusion rates and other physical characteristics of the metals. No satisfactory theory of metallic self- diffusion has as yet been proposed to account for all the existing data and to permit estimation in other metals. Experimental Part: Two units of radioactive iron were used in these experiments, each a mixture of Fe53 and Fe59 (12). Fe53 decays by K electron capture and the emission of an X ray with a half life of about 4 years. Fe59 emits two beta spectra, one with a maximum energy of 0.26 Mev, the other 0.46 Mev, and gamma rays of 1.1 and 1.3 Mev, with a half life of about 44 days. They were present in nearly equal concentration initially. The composite half life started at about 50 days and increased as the Fe59 decayed, leaving Fe55 relatively more abundant. After aging a year, the half life was too long to measure significant decay in a month. The absorption properties also changed with time. Because of the importance of the absorption coefficient in the diffusion calculations, it was necessary to determine this quantity frequently over the period during which these experiments were carried out. This correction was unsuspected at the time of publication of notesl3 on this work and accounts for the discrepancy in alpha iron and for part of the discrepancy in gamma iron.* * The data in the present paper supersede those given in the notes entirely. In one note the scale of log D was inadvertently reversed. In addition to the correction for the drift in absorption coefficient, the D values for gamma iron at low temperatures were high owing to insuficient correction for diffusion occurring during heating and cooling through the high temperature part of the alpha range at a rate much slower than that employed in the runs reported here. The high temperature alpha rates are much higher than the low temperature gamma rates and correspond to large equivalent times at these temperatures. These experiments were discarded and new measurements taken. Independent experiments with the same activity units indicated a radioactive contaminant in the iron, but exhaustive attempts to isolate and identify it chemically failed. These experiments? indicate † These experiments will be discussed in a later publication from this laboratory. that the contaminant represented only a trace of little importance in the diffusion results. The active iron was plated from an iron chloride solution. Enough of the solution was dropped on a 1% in. square of filter paper to wet it thoroughly.

Jan 1, 1951

By R. F. Mehl, C. E. Birchenall

R. E. Hoffman and D. Turnbull—The authors have presented evidence which they have interpreted as indicating that the rate of self diffusion is not intrinsically more rapid at grain boundaries than within the grain. Grain-size effects which apparently exist are attributed rather to impurities concentrated at the grain boundaries. In view of our own experiments and the existing evidence, we believe that the support for this hypothesis is not convincing. We have in progress an investigation in which the rate of self diffusion of silver is being measured over an extended temperature range in both single-crystal and polycrystalline specimens. The results of the single-crystal experiments and some preliminary data on fine-grained polycrystalline specimens have already been reported:&apos; and it is anticipated that a complete report will be published in the near future. The self-diffusion coefficient of large-grained polycrystalline silver (1 grain per sq mm) has previously been measured by Johnson" between 730" and 940°C. The diffusion coefficients which we have measured in single crystals (210 plane normal to diffusion direction) agree within experimental error with values calculated from an extrapolation of Johnson&apos;s curve down to temperatures as low as 500 °C. However, it has been demonstrated that the overall self-diffusion rate in fine-grained polycrystalline specimens (initial grain size of 0.003 cm) becomes measurably larger than the overall rate in a single crystal at a temperature of 600°C, and the discrepancy between the two rates becomes greater as the temperature is further decreased. In fact, it has been possible to obtain satisfactory penetration curves for polycrystalline specimens using the sectioning technique at temperatures as low as 400°C. At this temperature, the penetration is 50 to 100 times greater in the polycrystalline specimens than in a single crystal. Fisher" has developed an analysis whereby the ratio of the rate of the unit diffusion process at the grain boundary to the corresponding rate within the grain can be calculated from the penetration curves and an assumption as to the width of a grain boundary. This analysis applied to our data indicates that the unit process at the grain boundary is faster by a factor of 10&apos; at 475°C when the grain boundary width is taken to be 5. The silver used in most of these experiments was obtained from the Handy and Harmon Co. and listed as 99.97 pct pure. Preliminary experiments on 99.999 pct silver from the Jarrel-Asch Co. indicate a grain-size effect of the same order of magnitude as in the less pure silver. Nominally, these purities are as good, at least, as that of the carbonyl iron used by the authors, but of course if an impurity effect does exist its magnitude might be very dependent upon the nature of both major and minor constituents. The authors have cited the work of other investigators who have found no grain-size effects. Neither Steigman, Shockley and Nix" nor Maier and Nelson8 were able to correlate self-diffusion coefficients of copper with grain size. However, all their measurements were performed at or above 750°C; and on the basis of our work with silver, no grain-size effect would be expected at temperatures above about 0.7 of the absolute melting temperature unless the grain size were exceedingly small. Likewise, in the investigation of the self diffusion of lead by Seith and Keil,14 the lowest temperature at which the diffusion coefficient was measured in polycrystalline specimens was 207°C, which is still sufficiently high so that the lack of a grain-size effect is not surprising. Finally, in those experiments on iron from which they concluded that there was no grain size effect, Drs. Birchenall and Mehl seem to have no information as to the actual grain sizes immediately prior to and following the diffusion anneal. Without this information, we believe that their own experiments offer little support for their hypothesis. F. S. Buffington, I. D. Bakalar, and M. Cohen—The results given in this paper agree in order of magnitude with those tentatively reported by us.27 However, significant differences exist in the two sets of data, and it may be well to make an explicit comparison. The diffusion studies at M.I.T. were conducted on somewhat higher purity iron (99.98 pct Fe) than the grades used by the authors, but this is undoubtedly not the answer. Fig. 4 shows the diffusion results of both laboratories for the gamma phase, omitting the authors&apos; data on the commercial steels, while fig. 5 presents a similar comparison for the alpha phase. The divergence is much more marked in the latter case than in the former. In connection with the M.I.T. determinations, all of the runs in the gamma range and those above 800 °C in the alpha range were conducted with specimens having a relatively thick (0.002 cm) electrodeposit of radioactive iron. This practice minimizes any possible error due to extraneous diffusion that may occur during the heating to and cooling from the operating temperature. An exact solution of Fick&apos;s law for these boundary conditions was used in calculating the diffusion coefficients. At a later time, three runs were made below 800°C, using very thin electrodeposits similar to those of the authors, and the points fell considerably below the values expected from the extrapolation of the results based on the specimens with the thick deposits (compare dash-dot line in fig. 5). However, in the runs with the thin deposits, deviations of 100 pct were found between the individual specimens, whereas the maximum deviation with the thick deposits was less than 25 pct. Accordingly, it is not known at the moment whether the M.I.T. points below 800 °C should be given as much weight as those above 800°C. If this were done, the frequency factor would be of the order of 400 cm2 per sec, which is quite high. In other metals, the frequency factor for self diffusion lies between about 0.1 and 10 cm2 per sec. As the points below

Jan 1, 1951

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

The self-diffusion coefficients for a iron have been deternzined between 980° and 1167° K using Fe55 as the tracer. With decreasing temperature the diffusivity was found to decrease more rapidly than predicted by the Arrhenius equation in the region of the Curie temperature. The diffusion coefficients obtained well above the temperature of the magnetic transformation are more precise ard differ substantially from all the previously reported values. In the temperature region beginning at least 50" above T,, the diffusion coefficient is given by D = 118 (±2) exp 1 HE self-diffusion coefficients for a iron have been reported previously as the result of four independent investigations The general lack of agreement and the poor precision of most of the previous work prompted the redetermination by the present authors. The possible effect of the onset of ferromagnetism upon the diffusivity5 provided an additional impetus for this investigation. The earlier measurements of Birchenall and Mehl1 gave low values for the diffusion coefficient at temperatures below the Curie temperature, Tc. However, the precision of these measurements is not sufficient to establish an indisputable correlation between the low values and the magnetic transformation. The results of this investigation along with all the previous ones are shown in Fig. 1. EXPERIMENTAL Small rectangular samples, approximately 1 by 0.5 by 0.3 cm, of high-purity iron (Vacuum Metals Corp., "Ferrovac") are first heated in hydrogen at 900°C for24hr in order to remove possible trace amounts of dissolved carbon and oxygen. The small

Jan 1, 1961

By S. J. Rothman, A. L. Harkness, L. T. Lloyd

Self-diffusion in Y uranium has been measured using U235 as the tracer isotope. The diffusion coefficient fits an Arrhenius-type equation D = 2.33 x 10 -3 exp (- 28,5000/RT) cm2/sec The values of Do and Q are unusually low and are not consistent with the values expected from theories based on a simple vacancy mechanism of diffusion. CYLINDRICAL samples, 1 cm in diam and 1/2 to 1 cm long, were prepared from high-purity uranium containing 0.0343 pct u235, Table I. The samples were water quenched from 690°C (the a-ß transformation is at 668°C) to remove preferred orientation, and annealed at 450 "C to minimize residual stresses. One end of each cylinder was polished on a 1-µ diamond lap and electropolished in a phosphoric acid solution. After a final cleaning by cathodic sputtering, a 1- to 3-µ thick layer of the tracer isotope, uranium containing 93 pct u235, was sputtered onto the sample.&apos; To prevent contamination during the diffusion anneal, the diffusion couples were placed in a tantalum cup and covered with another piece of uranium, and the whole assembly was sealed off in an evacuated quartz tube. Annealing temperatures were constant within +2.0°C. The samples were heated and cooled slowly between room temperature and the ß-? transformation (774°C) in order to minimize distortion of the sample surface that might arise from aniso-tropic thermal expansion and density changes during phase transformations. After machining off 1/16 in. from the radius to

Jan 1, 1961

By P. G. Shewmon

Radioactive MgZA has been used to study the rate of self-diffusion in oriented single crystals of magnesium in the temperature range 468O to 635OC. The diffusion coefficients parallel and perpendicular to the c-axis are: Dl, = 1.0 exp (—32,200IRT) cm2 per sec and Dl = 1.5 exp (—32,500IR T) cm Qer sec. The ratio Dl/Dll was found to vary from 1.13 at 468 to 1.24 at 575 C. Assuming a vacancy mechanism, an explanation of this aniso-tropy of diffusion follows from a consideration of the difference in the saddle points for diffusion in and out of the basal plane. RECENT discovery of radioactive Mg2"&apos;-&apos; has made possible the study of self-diffusion in magnesium. The experimental procedure and the results of a study of diffusion in polycrystalline magnesium have been described in an earlier paper.V he present work is an application of the same techniques to the study of self-diffusion in oriented single crystals of magnesium. In the general diffusion problem the diffusion coefficient is a second order tensor relating the two vectors, the diffusion flux, and the concentration gradient. In a hexagonal lattice, such as magnesium, the complete determination of the diffusion coefficient D as a function of direction in the lattice requires a knowledge of D,, and D,, i.e., the diffusion coefficients parallel and perpendicular to the c-axis, respectively. It can be proven quite generally that in a hexagonal lattice D is independent of direction in the basal plane, and that out of the basal plane it varies with 0, the angle between the c-axis and the direction of diffusion, according to the equation D(B) = D,,cos2B + D,sin2B. [I] The proof of Eq. 1 depends only upon the symmetry elements of the hexagonal lattice and upon the transformation properties of a second order tensor." Therefore, Eq. 1 holds for any mechanism of diffusion and any c/a ratio. Experimental Procedure The experimental procedure used with polycrystalline specimens can be briefly outlined as follows. Radioactive Mg" was produced by bombarding a NaCl crystal with 350 mev protons and was chemically separated from the target as MgO. The Mg" was then vapor deposited on a specially cleaned magnesium specimen by heating the MgO on a tantalum ribbon in a vacuum. During the diffusion treatment, oxidation and vapor loss of the radioactive material were minimized by annealing the samples in pairs with the active faces in contact, each pair being inside a magnesium container, which was in turn surrounded by an argon atmosphere in a sealed Pyrex tube. The distance-activity profiles were obtained by measuring the activity of thin sections cut parallel to the original interface with a lathe. The only technique which was peculiar to this work was the preparation and orientation of the single crystal specimens. The single crystals used in this work were grown by E. C. Burke of the Dow Chemical Co., using a modified Bridgman method, in which the furnace and specimen were stationary while the temperature gradient moved.&apos; In growing these crystals the starting material was distilled magnesium, the crucibles were machined from Acheson electrode graphite, and the furnace atmosphere was tank argon. The crystals were grown from sublimed magnesium with the following analysis: 0.0002 pct Al, 0.0017 pct Fe, 0.0009 pct Mn, 0.0001 pct Ni, 0.0006 pct Pb, and less than 0.01 pct Ca, 0.0001 pct Cu, 0.001 pct Si, 0.001 pct Sn, and 0.02 pct Zn. The two crystals used were roughly lh in. diam and eight in. long. The c-axis made angles of about 7" and 78", respectively, with the specimen axes. If the values of D obtained by the use of these two crystals are taken equal to D,, and D,, respectively, the error introduced by this assumption is less than 1 pct. This can be shown by combining Eq. 1, the identity sin% + cos&apos;8 = 1, and the experimental fact that DJDn in magnesium was always less than 1.25.

Jan 1, 1957

By F. E. Jaumot, R. L. Smith

Self-diffusion in zinc at temperatures below 200°C has been studied using both single crystal and polycrystal samples. Anomalous results were obtained for single crystal samples, the data indicating that in some cases grain boundary diffusion predominated. These anomalous results are presumed to be due to low angle lineage boundaries in the single crystals. When volume diffusion occurred in either single or polycrystal samples, the values for the diffusion coefficient were as much as several orders of magnitude larger than would be expected from high temperature data. ONSIDERABLE work has been done on self-^-*diffusion at temperatures in the neighborhood of two thirds of the melting temperature (in OK) and higher, but very little has been done at lower temperatures. This is particularly true of self-diffusion in single crystals. This paper reports work on self-diffusion in zinc at temperatures below 200 °C where both polycrystalline and single crystal samples have been used. The measurements were made using both the sectioning and absorption techniques. The experiments on diffusion at low temperatures were initiated primarily for three reasons: to obtain data from single crystal diffusion covering a wide range of temperatures, to obtain values of the activation energy for grain boundary diffusion, and to obtain, if possible, a lower limit to the temperatures at which the absorption technique may be used. It is possible to determine whether volume or grain boundary diffusion predominates at low temperatures by analyzing the manner in which the concentration of the diffusing atoms varies with the depth of penetration. The analysis is possible only for the sectioning technique. For volume diffusion and for the boundary conditions of the present experiment, the curve of the logarithm of the concentration, C, vs the square of the penetration depth, x, is a straight line with the slope proportional to — 1/D. If grain boundary diffusion predominates, Fisher&apos; has shown that a curve of In C vs the first power of the penetration depth should be a straight line. This linear dependence of concentration on depth has been observed experimentally by several investigators. (See, for example, refs. 3 and 4.) Following Fisher&apos;s analysis, the grain boundary diffusion coefficient can be written as where 8 is the grain boundary width, D, is the volume diffusion coefficient at the temperature involved, t is the time of diffusion, C is the concentration of diffusing material, and x is the depth of penetration. A more general analysis employing the same model as Fisher&apos;s but with fewer assumptions has been given by Whipple.&apos; In contrast to Fisher&apos;s analysis the more general analysis of Whipple does not always yield a straight line relation between In C and x. Recently Turnbull and Hoffman have discussed both analyses and have applied the Fisher-Whipple model to a more refined dislocation model of the grain boundary. They have also discussed the possibility of a large effect of grain boundary direction on diffusion. In the case of the absorption technique, for which results are also given, one simply gets a value for the diffusion coefficient with no indication of the predominating mechanism. Experimental Techniques Single crystals of 99.99+ pct Zn were grown using a modified Bridgman method. Polycrystal samples were prepared in three grain sizes (small, about 4 mm2; medium, about 9 mm&apos;; and large, about 25 mmz) by quenching, air cooling, and furnace cooling. The grains of the polycrystalline samples tended to be columnar, with some preferred texture such that the c-axis was inclined at about 20" to 25" from the normal to the diffusion face, particularly in the medium and large grain size samples. The samples were cut from the rod, polished, and heavily etched, and faces cut accurately parallel using a jeweler&apos;s lathe. They were again etched to remove the lathe smear and annealed for 3 hr at 250" to 300°C. After annealing, they were etched, examined, and the orientation determined by Laue back reflection photographs. They were then electroplated with Zn. The plated surfaces were examined for uniformity; nonuniform coatings were removed by etching and the samples replated without additional preparation. For the diffusion anneals, the samples were placed in pairs, with active faces together, in evacuated Pyrex tubes. Constant temperature baths were used, and the variation in temperature did not exceed ±l°C, Diffusion runs were made at l00°,

Jan 1, 1957

By G. C. Kuczynski

Two particles in mutual contact form a system which is not in thermo-dynamical equilibrium, because its total surface free energy is not a minimum. If such a system is left for a certain period of time, the bonding of the two particles will take place in order to decrease the total surface area, even though the temperature is lower than the melting point. This phenomenon of bonding of two or more particles with the application of heat only and at temperatures below melting point of any component of the system will be called sintering, although the powder metallurgists use this term in a broader sense, including the presence of molten phase and pressure. It is the objective of this paper to study this process and the mechanisms involved in it. This problem is of utmost importance to powder metallurgy and powder ceramics, and its technological aspects have been studied for a good many years. However, the powder metallurgical operations are too complex and include too many superimposing mechanisms, and too many variables for a direct study. It was therefore advisable for the purposes of this study to reduce the variables to a minimum. In this work the radius of the interface formed during bonding in a simple system composed of a spherical particle and a large block of the same metal was studied as a function of time and temperature. It is believed that the mechanism involved in this simple process is fundamental to any sintering operations. Previous Work J. Frenkell was the first to make a serious attempt to develop a theory of sintering. He assumed that the process consists of a slow deformation of crystalline particles under the influence of surface tension which reduces to a viscous flow where the coefficient of viscosity n is related to the self-diffusion coefficient D by the following equation 1 _ D D6kT [1] where 6 is interatomic distance, k the Boltzman constant, and T the absolute temperature. This type of viscous flow of a crystalline substance is essentially different from the ordinary plastic flow. The latter is a specific propcrty of crystals and cannot take place in amorphous bodies. According to Frenkel this viscous type of flow is due to the diffusion of the holes or vacancies arising in the lattice. He was able to derive an equation relating the growth of the interface between two spherical crystalline particles or between a particle and a semi-infinite crystal (Fig 1) to time t at constant temperature. This relationship can be written as follows: x²— 3/2 a/nt [2] where x is the radius of the interface assumed to be circular, a the original radius of the sphere and a surface tension of the material. The other assump- x tions are that x is less than 0.3 and that during the period 1 the original radius, a, of the metallic particle did not change appreciably. A. J. Shaler and J. Wulff2 followed closely the ideas of Frenkel in their theory of sintering of a mass of metallic powder. Neither Frenkel nor Shaler has validated his theoretical speculations with conclusive experimental data. Two measurements reported by

Jan 1, 1950

By W. C. Hagel

Previous inuestigators have repovted unusually low H* and Do values for self-dzf@szon in certazn bcc metals, e.g., chromium nnd y -uvanium. It has been postulated that this is nn experimental crl -tetion for the existpnce of a four -atom ring meclzanzsm, mid there could be. zmpovtant theoretrcal Implicatlons. By utilizing a precise sectioning technique with Cr51 as the tracer, self-dzffi-lsion measurements in large grains of 99.99-pct Cr were extended from 1200° fo 1600"C. A least- squares analysis of the data yields that D 0.280 exp (-73,20U/RT) cm2 sec-&apos;. These AH* and Do unlues are not anomalous; there is little reason to believe that n vacancy mechanism is not opevatiue. Apparent differences at lower temperatures may result from preferential diffusion along short-circuititing paths and/or sample contamination. WITH continued technological effort, chromium and chromium-base alloys should ultimately become useful high-temperature structural materials. Accurate measurements of self-diffusion coefficients in the pure metal are necessary for a quantitative understanding of sintering, grain growth, creep, and other related rate processes. However, two conflicting determinations have been reported.* Using an Since careful diffusion studies normally yield Do values ranging from 0.1 to 10 cm2 sec-l, the Paxton-Gondolf results require independent experimental corroboration before an undue amount of speculation arises to explain the apparent anomaly. As discussed by Nowick6 and clearly shown by Hoffman and Turn-bull7 for self-diffusion in silver, a low AH* and Do can arise from preferential diffusion along grain boundaries or other short-circuiting paths. The most reliable data come from high-temperature measurements where volume effects predominate. Following a precise sectioning technique developed for determining silver diffusion in the intermetallic compound AgMg8 and cation diffusion in the corundum oxide Cr2O3,9 the purpose of this study was to measure true volume diffusion in high-purity, large-grain samples at higher temperatures than considered previously. Another point of interest is the frequent proposal10-12 that chromium undergoes an allotropic transformation between 1375" and 1800°C where the diffusion data might then show a discontinuity. I EXPERIMENTAL 1) Sample Preparation. The highest purity (99.997 pet) Cr currently available is produced by reduction of chromium tri-iodide. Three pounds of the resulting crystal bar were shipped in a sealed container with a complete chemical analysis from the vendor.* In conformity with the procedure used for

Jan 1, 1962

By R. E. Hoffman, R. A. Ward

The self-diffusion coefficient in high purity nickel has been measured over the temperature range 870&apos; to 1248°C. The results are described by the relation D = 1.27 exp[—-66,800/RT 1cm2ec-1. The measured activation energy correlates satisfactorily with absolute melting point, heat of fusion, and heat of sublimation. KNOWING the rates of self-diffusion is important in the understanding of many metallurgical reactions. However, in spite of the rather widespread interest in nickel, both commercially and in fundamental investigations, no measurement of the self-diffusion in solid nickel has previously been re- ported." In connection with a program of measuring rates of diffusion in some nickel alloys, self-diffusion in pure nickel has been investigated over a rather wide temperature range by combining a sectioning technique and a surface counting technique.

Jan 1, 1957

By J. D. Meakin

The self-diffusion coefficients of ß tin have been deterttlltled using a plating and sectioning technique. The principal diffusivities pavallel and perpendicu1ar to the "c" axis are given by the Arrhenius relations Dc = 8.2 exp (-25,600/RT) sq cnl per sec and Da = 1.4 exp (-23,300/RT) sq cm) per sec. The results are discussed in terms of a vacancy mecharism of diffusion. In conjunction with an investigation of diffusion in dilute indium-tin alloys we have reexamined self-diffusion in tin single crystals. The published self-diffusion coefficients of tin, eensham&apos;) have always appeared anomalous in comparison to diffusion data for other metals and do not correlate with parameters derived from creep tests.&apos; In addition to the necessity of confirming the self-diffusion values for use in analyzing the alloy work, self-diffusion in an anisotropic lattice is of considerable fundamental interest and confirmation of the published data seemed worthwhile. For a crystal with an axis of symmetry three-fold or higher the diffusivity at an angle 6 to the axis is given by D? = (Da - Do) sin2?+ Do [1] where Da and Do are the principal diffusivities perpendicular and parallel to the axis. For a particular diffusion mechanism, again in a crystal of axial symmetry, these principal diffusivities are [2] Do - ? ?i2 COS?i In these expressions ?i is the displacement undergone by the ith atom and 8, is the angle this displacement makes with the symmetry axis. In this, as

Jan 1, 1961

By H. W. Mead, C. E. Birchenall

SELF-DIFFUSION of iron in austenite is a process which may play a significant role in some of the practically important reactions which occur in solid irons and steels. It also provides a system in which the interaction of vacancies on substitutional sites and atoms in interstitial positions may be investigated. For both purposes, accurate values of the diffusivities are required. Two sets of previous measurements exist, each containing features which require checking. In the course of an investigation to determine the rate of diffusion of iron in certain steels, Garwick and Rosenqvist&apos; measured the self-diffusion coefficient of iron in a 0.10 pct* C steel and obtained a dicated that while at lower temperatures D increased with increasing carbon content, at the higher temperatures it first increased and then decreased. At the same time the activation energy decreased linearly from a value of 69,000 cal per g-atom for pure iron to 33,000 cal per g-atom at 1.06 pct C. This rapid decrease in activation energy implies that iron at higher temperatures within the austenite range would diffuse more slowly in high carbon steels than in steels of lower carbon content. Experimental Procedure The radioactive iron used in this work was obtained from Oak Ridge National Laboratory. The unit consisted of about 1 microcurie each of Fe7 and Fe It was allowed to decay for two years. At the end of this time most of the Fe with a half-life of 46.3 days, had decayed, while most of the original Fe(half-life 2.91 years) still remained. The iron as chloride was extracted into an ether layer, using the method of Ashley and M~rray. After evaporation of the ether the iron chloride was dissolved in 250 ml of aqueous solution containing 2.4 gpl Fe. The compositions of the steels used are listed in Table I. Disks of ¾ in. diam and about ¼ in. thick were machined from these steels for use in diffusion couples. The disks were surface ground on both sides to ensure parallel opposite faces. One face of each disk was polished through 000 emery paper. Half of the prepared disks were plated with radioactive iron. For this, 1 ml of the radioactive iron

Jan 1, 1957

By L. Himmel, R. F. Mehl, C. E. Birchenall

The rates of self-diffusion of iron in artifically prepared wustites of various compositions have been determined using the decrease in surface activity technique. Similar measurements are reported for artificial magnetites of nearly stoichiometric composition and for natural hematite single crystals. These data, together with appropriate thermodynamic data derived from the present study or from the literature, are used to calculate the rates of oxidation of iron and its oxides, taking as a basis the theoretical rate equations developed by Wagner. The calculations are compared with experimental rate constant data for these same reactions and are shown to provide essential confirmation for the Wagner theory. For the most part, the results are also consistent with the previously proposed transport mechanisms in the oxides of iron. WHEN a metal is exposed to an oxidizing atmosphere at elevated temperatures, a superficial oxide layer is formed which then thickens according to some characteristic rate law. Provided that the oxide layer is dense and continuous and does not present any paths for gaseous diffusion, continued growth must inevitably involve the migration of the reactants—either the metal or oxygen or both—through the lattice of the solid oxide. This is true whether the reaction rate is controlled by the diffusion process itself or by a slow step in a reaction at one of the interfaces. If the overall rate is limited by diffusion, and if the concentrations or thermodynamic activities of the reactants at the boundaries of the oxide layer are independent of time, the familiar parabolic rate law is observed. Under these conditions, Wagner1 has shown that the rate constant may be expressed in terms of the mobilities of the reacting components and the appropriate concentration or activity gradients which are established across the growing oxide layer. In presenting the phenomenological basis for the theory, Wagner has emphasized the role played by lattice defects in diffusion processes in oxides. The essential validity of this theory has been confirmed by experiment for a number of simple systems including the growth of Cu2O on copper.&apos;-&apos; One of the objectives of the present study has been to extend the Wagner approach to the more complex process of the oxidation of iron and its oxides, and particularly to wustite, which is normally the major constituent of the oxide scales formed on iron at temperatures above approximately 570°C. For this purpose the rates of self-diffusion of iron in w?stite, magnetite, and hematite have been measured by tracer techniques and the necessary thermodynamic data have been obtained from equilibrium studies involving the oxides and their atmospheres. These data have subsequently been used to test the underlying assumptions of the Wagner theory as well as the detailed mechanism of the scaling of iron previously proposed by Davies, Simnad, and Birchenall.5 Lengthy treatments of diffusion and conduction phenomena in ionic crystals have been published." * The latest, by Jost,8 also summarizes much of the existing experimental data. In general, however, the oxides of iron have not been considered specifically and hence a brief review of the available

Jan 1, 1954