Search Documents

By A. G. Guy

Zinc vapor from a reservoir of liquid zinc maintained at lower temperatures was diffused into sets of copper-zinc alloys at 759° and 870° C and into sets of silver-zinc alloys at 650° and 700° C. Each set of ,four copper-zinc alloys consisted of pure copper and 90/10, 80/20, and 70/30 brass. Similarly the set of silver-base alloys consisted of specimens containing 0, 7, 15, and 20 pct Zn. Natural markers at the initial surfaces of the specimens ware "trapped" under a thin electroplate and their depth of burying was measured after diffusion. The ratio of diffusivities of the two components determined from these data and the chemical diffusion coefficient calculated by the Matano method agreed with previously reported results. The line of markers was irregular and was often discontinuous when a grain boundary was crossed. TECHNIQUES involving the vapor phase have been widely employed in studying diffusion phenomena in metals. Many investigators have evaporated the more volatile component from an alloy, but it is generally believed that the porosity created under these conditions interferes with adequate interpretation of the experimental results. The extensive work of Gertsriken6 has been criticized on this score. Accary1 found that the diffusion coefficient for de-zincification of brass is larger by a factor of five than that for zincification, although he attributed the difference to a higher vacancy concentration under conditions of dezincification. Balluffi and siegle2 thoroughly investigated the technique of adding a more volatile component to a less volatile base metal as a means for studying diffusion. They verified that no porosity forms in a brass under these conditions and that the diffusivities of the two components can be determined from observations of the motion of markers originally placed on the surface of the specimen. The mathematical analysis that they developed for treating the experimental data was modified in a later note, but essentially it involved using the graphical Boltzmann-Matano method that is customarily applied to solid-solid diffusion couples. The experimental method used in the present research was similar to that employed by Balluffi and Siegle except for two modifications. Balluffi and Siegle used alloy chips as the source of metal vapor. This procedure has the disadvantage that the concentration of solute at the surface of the specimen changes during the diffusion experiment as the alloy content of the chips is gradually decreased. This disadvantage was overcome by employing liquid zinc as the source of zinc vapor in a manner similar to that described by Herasymenko.9 There was then no limitation on the amount of solute usefully available, and several relatively large specimens could be diffused in a given experiment. The second modification was the use of natural markers on the surface of the specimen rather than larger, artificial markers, such as the alumina powder employed by Balluffi and Siegle. It was expected that the fine, natural markers would permit more precise measurement of the buried interface as in the case of ordinary sandwich specimens.10 EXPERIMENTAL The experiments were designed so that zinc from the vapor phase would maintain a constant zinc content at the surface of pure copper (or silver) and

Jan 1, 1960

By T. S. Lundy, J. I. Federer

Chemically purified Zr95and Cb95 have been used in determining self-diffusion coefficients in the bcc phase of iodide zirconium over the temperature range of 900o to 1750°C. The temperature dependence of the diffusion coefficients could not be described by the usual Arrhenius-type equation over the entire temperature range. Values of the apparent frequency factors (D0)a and activation energies (Q)A for ZR95 varied from 4.8 x 10-6 to 2.5 x 10-2 sq cm per sec and 20,700 to 46,900 cal per mole, respectively, between 900°and 1750°C. Similarly, for Cb95, values of (Do) a and (Q)A varied from about 2.4 x 10'5 to 0.25 sq cm per sec and 26,500 to 56,800 cal per mole, respectively. The data may be described in terms of the a to ß transition temperature of 1136 °K and the absolute temperature of the diffusion anneals by the following eauations: SEVERAL studies on the diffusion of Zr95 in zirconium have been reported.1"5 The values of the constants contained in the usual Arrhenius-type equation for describing the temperature dependence of the diffusion coefficient, D = Do exp [- Q/(RT)], are summarized in Table I. The reported values of Do, the frequency factor, and Q, the activation energy, are seen to vary by factors of about 60 and 1.6, respectively. Examination of the data from any one study indicated reasonable consistency of measurements. The large discrepancies of over-all results, therefore, cannot be explained on the basis of experimental scatter. One possible explanation for the differences in results could be a varying impurity content of the zirconium. Another possibility is that the radioactive daughter, Cb95, influenced the values of some of the determined diffusion coefficients. Therefore, the first of the three purposes of this study was to reconcile the large differences in diffusion results that have been obtained. Several empirical relations have been proposed to relate the diffusion coefficients in solids with physical properties, such as the absolute melting point, the heat of fusion, and the heat of sublimation.

Jan 1, 1963

By M. T. Hepworth, W. B. Sample

HYDROGEN solubility and linear contraction measurements were made at constant temperature on an alloy of titanium of constant oxygen-to-titanium molal ratio but variable hydrogen content. A cylindrical specimen of titanium (0.25 by 1.6 in.) containing 0.75 at. pct 0 and less than 0.03 at. pct N was cast from a master button of iodide process titanium which had been alloyed with titanium dioxide. Four vacuum fusion analyses, which were made along the length of the cylinder after the hydrogen solubility studies were made, revealed a variation of oxygen content of less than 0.02 at. pct. The specimen was introduced into a quartz furnace tube, annealed and outgassedat 1050"~ for about 1 hr, and then held at 800" * 0.5" for about 30 hr, during which latter time experimental data were taken. The leak rate to the furnace chamber was not sufficient to contaminate the specimen during the run as determined by leak-rate studies and vacuum fusion analyses for oxygen and nitrogen before and after the runs. The furnace chamber was attached to a modified Sieverts apparatus used to introduce measured quantities of purified hydrogen to the specimen. A quartz pushrod transmitted changes in length of the specimen to a Chevenard dilatometer. The push-rod passed through a vacuum grease seal to the dilatometer which amplified motion of the rod by a factor of 319. During hydrogen addition, the pushrod did not contact the specimen. Measured quantities of hydrogen were added to the specimen and equilibrium hydrogen pressures recorded as a function of hydrogen content. The solubility data, which appear in Fig. 1, agree well with a separate investigation made with a different apparatus. The solubility limit of the hcp a phase appears as an abrupt discontinuity in slope at 3.4 at. pct H. The solubility plot has been extended as a dashed line at this discontinuity in order to emphasize the discontinuity. The hydrogen content at which bcc P appears is not clearly defined. An arrow indicates solubility data were taken upon hydrogen addition. Upon reaching the arbitrary maximum hydrogen content of the run, the quartz pushrod was pressed against the specimen in order to record the change in length with hydrogen removal. A pressure of about 100 g per sq cm was required to overcome the friction of the vacuum seal in order to obtain reproducible linear contraction data. The contraction ratio ~2/1", change in length divided by original length, was set equal to zero for the maximum hydrogen content. The relative contraction of the specimen was observed upon removal of hydrogen in the following manner. The system was opened to the pumps for a few seconds and an unknown quantity of hydrogen was removed. The system was then isolated from the pumps by closing a stopcock and allowed to return to equilibrium. When the temperature returned to 800°C, the pressure stabilized, and contraction ceased, measurements of contraction and pressure were made. The hydrogen content of the specimen was then deduced by measuring the equilibrium PIydrogen pressure and referring to the solubility plot. Linear contraction was then plotted against interpolated values of atomic percent hydrogen. The justification for this procedure is the established reversibility of the hydrogen equilibrium.' Uniaxial compression was assumed to have no effect on hydrogen pressure. The isothermal linear contraction data are plotted on Fig. 1. An arrow indicates data were taken upon removal of hydrogen. The solubility limits of the two phase a i /3 region can be identified by discontinuities ill slope of the dilatometric plot at 9.1 and 3.4 at. pct H, agreeing well with other studies.' Dashed extensions of the plots emphasize the discontinuities. Isothermal studies of expansion made upon hydrogen addition gave curves of much smaller slope and less pronounced discontinuities. A uniaxial compressive stress tends to accentuate volume changes in going from the more open bcc phase to the hcp phase.2 Incomplete data on ex-

Jan 1, 1962

By E. P. Abrahamson

The effect of transition element binary solid solution additions upon the re crystallization temperature of nickel has been investigated. All additions raised the re crystallization temperature. Both the rate of change of re crystallization temperature with at. pet solute and the limit of initial linearity are found to correlate with the free atom ground state outer electron configuration of the solute element. A comparison of all of the correlations for the transition elements studied has yielded a set of empirical relations, which allows the prediction of the effect of dilute binary transition elements on the recrystallization temperature of other transition elements. The work of Abrahamson, et al.1-7 on dilute binary solution alloys indicated a correlation between the free atom ground state electron configuration of the solute and both the brittle-ductile transition and re-crystallization temperature. Further, it has been shown3,4 that the limit of initial linearity may also be correlated with the solute electron configuration. Since previous studies have dealt with bee and hexagonal base elements, this study is made to note what effect, if any, the fee base structure might have. Nickel also serves to give a solvent configuration which will allow the comparison of all of the transition element solvents. Olsen8 has studied the effect of some trace elements upon the recrystallization of nickel. He included only Ti, Zr, W, Mn, and Co of the transition elements. The differences in change of recrystallization temperature were attributed to the solute atom size and to the solubility of the solute in nickel. PROCEDURE All alloys were made using 99.9 + pet Ni with 0.003 C, 0.015 Fe, 0.007 0, and 0.0007 H. The solute elements were 99.9 + pet pure. According to the published binary phase diagrams9 and metallo-graphic examinations at X750, all solute elements were in solid solution. The alloys were arc melted and remelted four times in the form of cubic 200-g buttons under an argon atmosphere. They were then hot forged at 800oC to 0.750 in. diam, machined to 0.60 in., vacuum annealed for 1 hr at 700°c, swaged to 0.40 in. and annealed at 700oC for 1 hr in vacuum. The grain size was checked and found to be 77 * 20 grains per sq mm. The specimens were then swaged to 0.187 in., yielding 76 ± 1 pet cold work. All alloys were then chemically analyzed for the solute addition. The interstitial contents remained at the values of the starting material when checked on random alloy specimens. The swaged rod was cut into 8.25 in. lengths and heat treated in a gradient furnace for 1 hr. The gradient was maintained at 250" to 850°C over an eight inch length, recorded by twelve thermocouples. Control was ± 3oC, accomplished at the hot end. The recrystallization was determined metallo-graphically over the entire specimen. The criterion chosen was that point on the specimen showing the first recrystallized grain at a constant magnification, X200. Specimens were repeated, and the agreement was found to be generally ± 3oC.

Jan 1, 1962

By Hsun Hu

Direct observations were made on the recrystalli-zntion of a cold-rolled (110) [001] crystal of 3 pct Si-Fe by annealing thin sections of the specimen inside the electron microscope during examination. Subgrain formation, presumably by polygonization, occurs rapidly, followed by a slow increase in the subgrain size before recrystallization can be detected. No migration of the subgrain boundaries was observed, and the formation of recrystallization nuclei by coalescence of subgrains is suggested. Re-c~ystallization characteristics of various texture components were individually examined. AS reported in a previous paper: orientations of the various deformation texture components in a 70 pct cold-rolled (110) [0011 crystal of 3 pct Si-Fe were completely confirmed by transmission electron diffraction patterns taken at various areas of the specimen. These orientations can be summarized as follows: Over most areas of the deformed crystal (111) planes are parallel to the surface. These areas correspond to either $- the two main defor: mation textures, the (111) [112] or the (111) [112].2-5 Within a few areas, (100) planes parallel to the surface are frequently found. These areas represent the minor deformation texture component, the (100) [Oll] type, which is derived from rotated Neumann bands.2'4'5 Only in very rare instances are (110) planes found parallel to the surface. This orientation corresponds to the very weak texture component of the (110) [001] type. In a pole figure, this (110) [0011 orientation is located within the orientation spread between the two main (111) [1121 type textures. Usually the electron diffraction patterns taken at various areas of the specimen are very simple ones, i.e., for any one area the pattern consists only of sharply defined spots of a single orientation. In some areas, however, more or less continuous rings of weaker intensities are superimposed on a simple pattern, indicating the presence of a wide range of additional orientations. In order to understand in detail the recrystallization of this cold-worked crystal, the recrywtallization characteristics of each of the various deformation texture components must be examined. This can best be done by using a high-temperature stage and transmission electron microscopy techniques, whereby small areas corresponding to the various texture components can be selected individually for examination. The present paper presents results obtained by this method. MATERIAL AND METHOD Specimens were prepared from a previously studied1 single crystal of a commercial grade 3 pct Si-Fe alloy with an initial orientation (110) [00l], cold-rolled 70 pct from 0.014 in. to 0.004 in. in thickness. Thin foils were prepared by electrolytic polishing as described in the earlier paper.' Specimens of this thin foil were then mounted in a high temperature stage similar to the one designed by Whelan.6 A direct current was used as the heating source. The temperature of the heating grid was calibrated by means of a Pt vs Pt-10 pct Rh thermocouple in a vacuum chamber (0.01 to 0.02µHg). Maximum variation within the temperature range of 250o to 000°C was about ± 25°C. After the desired areas were selected for examination, the specimen was heated to a low temperature for some time while the selected area was constantly watched. When no further change in structure and orientation could be observed, a higher temperature was applied. Photomicrographs and diffraction patterns were taken from time to time during examination, with the heater temporarily turned off* while the photographs were

Jan 1, 1962

By J. B. Clark, A. S. Yue

PfANN1 has shown that when a cylinder of molten binary alloy freezes directionally, the solute distribution can be described analytically by cg=k(l-g)k-1 [1] in which Cg is the solute concentration in the solid at the point where a fraction g of the original liquid has frozen and Co is the initial solute concentration. The term k is the effective distribution coefficient of the solute and is defined as the ratio Cx/CL where Cl is the solute concentration in the main body of the liquid. Both the slope and the intercept at -log (1 - g) = 0 of a logarithmic plot of Eq. [I.] provide a measure of k. Under conditions of freezing where diffusion in the solid is negligible and the solute concentration in the liquid is uniform, the effective distribution coefficient k, derived as above, is equal to the equilibrium distribution coefficient taken from the phase diagram. However, under ordinary freezing conditions, even though solid-state diffusion is negligible, the solute rejected from the freezing solid does not have sufficient time to diffuse uniformly throughout the main body of the liquid. A concentration gradient is formed in a boundary layer in front of the advancing solid-liquid interface as pointed out by Rutter and Chalmers.' Thus, under these conditions, the concentration of the freezing liquid in the boundary layer is higher than that predicted from the equilibrium phase diagram and consequently the effective distribution coefficient k, obtained from directional freezing experiments, is larger than the equilibrium distribution coefficient. The symbol k will designate the effective distribution coefficient in the discussion which follows. In this paper, the determination of the effective distribution coefficient is demonstrated for several dilute binary alloys and a polynary magnesium alloy. The effect of freezing rate on the value of the derived k is presented in detail. An example for the calculation of the solidus in a dilute binary solution using the experimentally determined effective distribution coefficient is given. MATERIALS STUDIED Binary magnesium alloys of copper, nickel, and zinc containing approximately 0.10 wt pct solute and a polynary alloy containing approximately 0.10 wt pct each of several elements were selected for study. The spectrographic analyses of these alloys are given in Table I. The analyses are believed accurate to within +5 pct of the stated value. EXPERIMENTAL PROCEDURE Extruded rods of the alloys, 3/8 in. in diam and 11 in. long were used for directional freezing studies. The alloy rods were melted horizontally in a high-purity graphite boat by a Nichrome-wound tube furnace which moved parallel to the length of the rod. The boat was supported inside a Vycor tube suitably connected to provide a protective atmosphere of dry SO2. For an experimental run, the following procedure was used. The alloy rod was cleaned in dilute HNO, and dried. The graphite boat containing the alloy rod was placed within the Vycor tube and the tube connected to the gas train' The 'ystem was purged for 20 min with dry argon and then a flow of dried SO, was maintained for the duration of the run. The movable furnace was positioned and either all or a desired portion of the alloy rod was melted. After attainment of steady-state temperature gradients the molten bar was solidified from one end to the other at a controlled rate by pulling the furnace along the length of the bar at the desired speed. At the conclusion of the run, the directionally frozen alloy rod was analyzed spectrographically at 1 1/4-in. intervals.

Jan 1, 1961

By A. R. Troiano, E. A. Steigerwald, F. W. Schaller

The kinetics of crack propagation in a hydrogenated high-strength steel at subzero temperatures indicated that cracking progressed in a discontinuous fashion. The delayed failure process thus involves a series of crack initiations rather than the continuous growth of a single crack. The incubation time required for the initiation of delayed failure was controlled by the diffusion of hydrogen. The activation energy for the process was 9120 cal per g-alom. As the test temperature was lowered the incubation time constituted a greater percentage of the total tinze to fracture, and al -50°F the incubation and fracture times were essentially coincident resulting in instantaneous, catastroplzic fracture. A unique feature of hydrogen embrittlement is the delayed failure process in which the material fails under the action of a sustained load even though it is capable of supporting a higher load for a finite period of time.')' Delayed failure of steel and titanium is characterized by an incubation time, preceding the initiation of a crack in a highly localized region, and the subsequent growth of the crack to ultimate failure.3-5 The kinetics of crack growth in hydro-genated high-strength steels have been studied through the use of resistance measurements and have an important bearing on the mechanism of delayed failure. Delayed failure is believed to be merely a facet of hydrogen embrittlement in which sufficient time is allowed during a test for hydrogen diffusion. The current theories employed to explain hydrogen em-brittlement can be grouped into two types. one6-8 explains the embrittlement on the basis of the pressure exerted by molecular hdrogen in lattice voids or defects while the other9-10 attributes the embrittlement to the influence of hydrogen in the lattice. The mechanism used to explain the nature of the crack kinetics in delayed failure is based on the stress-induced diffusion of hydrogen to the area of maximum triaxiality.3 When the hydrogen concentration in this region reaches a critical value, a crack is nucleated. This initial crack propagates instantaneously until stopped by the higher fracture stress of the adjacent material of lower hydrogen content. The continuation of cracking must then await the further diffusion of hydrogen to the new region of triaxiality near the base of the crack where another initiation will occur. From this mechanism it follows that crack propagation is nothing more or less than a series of initiations with limited fracture. The purpose of this investigation was to examine the kinetics of crack growth in an effort to determine whether or not the process was discontinuous. Studies of crack propagation were conducted at low temperatures to retard the diffusion of hydrogen which is presumed to be the rate controlling process in delayed failure. MATERIALS AND PROCEDURE Specimen Preparation—A commercial heat of air-craft quality SAE-AISI 4340 steel, designated as Heat G, was furnished by the Republic Steel Co. in the form of %-in. hot-rolled bars. The material had the following chemical analysis: Composition of 4340 Steel—G Heat C Mn Si Ni Cr Mo 0,395pct 0.82pct 0.29pct 1.76pct 0.85pct 0.25pct Notched tensile specimens with less than a 0.001 in. radius, Fig. 1, were used throughout the investigation. The specimens were heat treated to a 230,000 psi strength level according to the following sequence: a) Normalize bar stock at 1650°F for 1 hr, b) Stress relieve at 1200°F for 4 hr and furnace cool,

Jan 1, 1960

By I. R. Kramer

In his technical note entitled "Specimen Temperature During Electropolishing of Aluminum Crystals" Dr. Nakada 1 reported that the temperature of his aluminum specimen increased 65°C when it was polished electrolytically; he therefore concluded that the temperature of the specimens used in our work (Nakada's Refs. 4 and 5) similarly increased by this amount. From this he further concluded that the change in plastic-flow characteristics associated with the removal of the surface by electropolishing during deformation were due solely to a temperature effect. We believe that these conclusions are not justified. It is, of course, realized that the amount of the temperature increase is associated with the total heat transfer of the entire system, including the efficiency of the stirring action, heat flow through the grip on the specimen, and so forth. Of great importance is the type of electrolyte used for polishing. For example, in Dr. Nakada's experiment a methanol-perchloric acid solution was used. This solution has a relatively low conductivity and for the current densities of 0.3 amp per sq cm the electrical input appears to be 30 to 40 w. In our research a methanol-nitric acid solution was employed and for the size of specimen employed by Dr. Nakada the power never exceeded 7 w. There are major differences in his test apparatus which also appear to affect the temperature rise. In our case a steel specimen holder was used in order to provide a good heat sink and conduction path. The cathode was a coiled aluminum tube through which a coolant was passed to provide additional cooling capacity near the specimen. The stirrer was placed in a position to obtain maximum cooling efficiency. On the other hand Dr. Nakada used a teflon grip to support the specimen at one end only and a flat copper cathode, thus providing a much less efficient heat sink as compared with the metal grips used in our studies. The relationship between current density and the temperature increase for two sizes of aluminum crystals tested in our tensile apparatus is shown in Fig. 1. The temperature measurements were determined from a thermocouple spot-welded to the surface of the specimen. The thermocouple was protected from the acid by means of a thin film of stop-off lacquer which covered just the outer surface of the thermocouple junction. Whereas Dr. Nakada reported a time delay of 8 min to obtain thermal equilibrium, our temperature change occurred in less than 15 sec. This emphasizes the importance of the heat transfer in the entire system. However, readings were taken over a period in excess of 5 min to assure thermal equilibrium. The data given in Fig. 1 were also confirmed by determining the thermal expansion of the specimen in a sensitive creep apparatus. As may be seen, for the smaller specimen the increase in temperature is 8°C for a current density of 1.5 amp per sq in. The majority of our data on the effects of surface removal on the plastic behavior of aluminum were obtained at current-density values at or below this value. Above 1.5 amp per sq in. the temperature rise is uncertain; for under certain conditions, especially in the vicinity of 2 amp per sq in., the anode can become polarized and the current flow is erratic. This condition is easily recognized because wide fluctuations in the creep and stress-strain data make it virtually impossible to obtain reliable results, a condition which was always avoided in our work. The temperature increase for the larger specimen is less than that of the smaller specimens below a current density of 1.5 amp per sq in. At a current density of 2 amp per sq in. the temperature increased 13°C. The above temperature data are in accord with observations made during creep experiments. Using a transducer capable of detecting strain in the order of 10 -6, temperature changes in excess of 8°C were not detected when the current density was increased to as high a value as 1.5 amp per sq in. Finally we wish to point out that our experimental data cannot be explained solely in terms of a heat effect. For example in Fig. 1 of Ref. 2, the change in slope of Stage II as a function of the rate of re-

Jan 1, 1965

By R. E. Smallman, J. C. Terry

J. C. Terry and R. E. Smallman (Llniuersity of Birmingham, England)—In a recent publication53 Wood and Westbrook have reported the results of an investigation of the tensile behavior of fl-phase AgMg alloys. These results agree in broad outline, but differ in the details, with those obtained in the above laboratory and, consequently, some discussion of their paper would seem appropriate. In our experiments, tests were carried out on extruded wires, 2 mm diam, in a hard-beam tensile machine, and while an oil bath was used for the tests up to 150°C, above this temperature the oil bath gave rise to serious contamination of the specimens and hence tests above 150°C were made in vacuo. This procedure is different to that adopted by Wood and Westbrook. The ß-phase alloys exhibit differences in mechanical behavior when they are tested in different temperature ranges. In agreement with Wood and Westbrook, three characteristic ranges are observed which may be classified as low, intermediate, and high temperatures, respectively. Our experiments show that in the low-temperature region, the shape of the stress-strain curves is similar to that reported. However, in addition, it is observed that the temperature at which the low-temperature region ends is markedly dependent on strain rate, and increase with increasing strain rate as shown in Fig. 18. The temperature at which a transition from the low to the intermediate temperature takes place is about 0.4 to 0.45 of the absolute melting point, Tm. This transition temperature coincides with that at

Jan 1, 1963

By R. E. Smallman

R. E. Smallman (University of Birmingham, England)—Recently, Tedmon and Ferriss11 have determined the yield stress parameters oi and ky for tantalum by measuring the lower yield stress as a function of grain size 2d and fitting the results to a relationship of the form They report that although ky , which is taken to be a measure of the dislocation locking strength, is small (- 2 to 4 x 106 cgs units) a substantial yield drop is nevertheless observed in a normal tensile test. Niobium gives a similar result,12-14 as pointed out in the original work by Adams et a1.,12 and in order to check this apparent anomaly the yield-stress parameters of electron beam-melted niobium have recently been reanalyzed15 by the Luders strain technique. In this method the strain hardening part of the stress-strain curve is extrapolated to zero plastic strain; the intercept on the preyield portion of the curve is taken to give oi, whilst the difference between oi and the lower yield stress gives kyd-1/2. The results indicate that ky increases with increasing grain size and hence, a plot of vs d-112 yields an apparent ky, which is lower than the true value. A similar effect could account for the small ky found in the relatively pure tantalum used by Tedmon and Ferriss. The variation of ky with grain size shows that dislocations are more strongly locked in coarse-grained specimens than in fine-grained samples. In niobium, this may be attributed to the fact that the dislocation density in the fine-grained material is higher than that found in the coarse-grained samples which are given a sufficiently prolonged anneal to remove any residual substructure and, since the metal contains only a small amount of interstitual impurity, a variation in locking occurs. By contrast, application of both the grain size analysis and the Luders strain method to yield-stress data from commercially pure vanadium containing a large amount of interstitial impurity gives consistent values of oi and ky, with ky independent of grain size and temperature. Electron microscope observations show minor variations in dislocation density from grain size to grain size, but in any case in this material the dislocations are heavily locked with precipitate. On yielding new dislocations are generated and, as a consequence, the importance of any differences in dislocation density between the various specimens of different grain size is considerably reduced. It is perhaps significant that Adams and lannucci,16 working with a grade of tantalum containing a higher interstitial content than that used by Tedmon and Ferriss, prepared the specimens of different grain size by annealing in the temperature range 1500" to 2000° C to minimize any differences in dislocation structure, and found that ky had a value of 1.04 x 107 cgs units, independent of testing temperature. Such behavior is consistent with the dislocations being locked by carbide precipitates so that the generation of free dislocations is an athermal process. The recent work of Gilbert et al.17 also shows that in tantalum there is no significant variation of ky with grain size provided it contains 150 ppm of oxygen. In this case, however, the dislocations are not locked by precipitate and ky is temperature dependent. C. S. Tedmon and D. P. Ferriss (authors' reply)— We would like to thank Dr. Smallman for his interesting comments and discussion to our paper, "The Dependence of Yield Stress on Grain Size for Tantalum and a 10 pct W-90 pct Ta Alloy".18 It was suggested that perhaps the relatively small values obtained by us for ky of tantalum could be attributed to the same cause that accounts for the apparently small values of ky that result when it is determined by the Luders Strain technique. Since our values were obtained by plotting the lower yield stress vs the reciprocal of the square root of the grain size, it is not clear how this could be the case. The values of ky in this experiment have been calculated, using the Luders strain technique. With this method, values for ky on the order of 2 x 105 to 5 x lo6 cgs units were obtained. In spite of this rather large variation, the magnitudes are still small, and there appeared to be no good correlation between ky and the grain size or the yield stress, probably because of the difficulty in accurately extrapolating the work-hardening portion of the curve back to zero plastic strain. As was shown in the original data,18 there was little work hardening in any of the curves, at any temperature. In his discussion, Dr. Smallman also points out how ky has been observed to increase with increasing grain size, when determined by the Luders strain technique. There are at least two possible explanations for this. In the first case, if it is assumed that the bulk of the interstitial impurities are concentrated at the grain boundaries, then, of course, the available grain boundary area would decrease with increasing grain size, thus presenting less area for the interstitials, which would then presumably increase the concentration within the grains, thereby increasing the locking of the dislocations. In the second case, the increase in ky with increasing grain size would be attributed to the nature of the grain boundary itself. One of the several ways of deriving the Hall-Petch equation19 is based on the stress concentration arising from a pile-up of dislocations at the boundary. The ability of the stress concentration to unlock a source in a neighboring grain would depend on the strength of the grain boundary. As is well-known, the nature and struc-

Jan 1, 1963

By E. A. Steigerwald

E. A. Steigerwald (Thompson Ramo Wooldridge, Inc.1— he authors' results clearly indicate that cracking can be produced by the hydrogen pressure developed during a charging operation. This type of cracking or blistering has also been observed in polycrystalline materials24"28 when the charging conditions are sufficiently severe and is often referred to as irreversible embrittlement since it cannot be removed by a baking treatment. There are additional data, however, which must be considered before a pressure theory can be generally employed to account for all failures which occur as a result of hydrogen. In many cases, hydrogen embrittlement in a tensile test or in delayed failures under static loads is a reversible phenomena which cannot be simply explained on the basis of preexisting cracks generated by the charging operation. It is this aspect of the problem which has prompted many investigators to seek mechanisms other than hydrogen pressure to explain their results.24'n728 Fig. 15 indicates an example where, under specific experimental circumstances, hydrogen embrittlement can occur at liquid nitrogen temperatures. In this case, which has been previously decribed,'' specimens were charged with varying quantities of hydrogen, immediately quenched, and tested in liquid nitrogen (-320°F). A companion set of identical specimens were charged, aged at 300°F to remove the hydrogen, and also tested at -320°F. When the current density was greater than lo-' amp per sq in., embrittlekent occurred in both sets of specimens, indicating that cracking had occurred during charging and the embrittlement, as in the case of Fe-Si single crystals, was irreversible. At current densities between approximately 103 and lo-' amp per sq in, the embrittlement was present only for those specimens which contained hydrogen during the testing sequence. The embrittlement was therefore reversible with respect to the aging and charging operations. Since extensive hydrogen movement would not be expected at -320°F, it is difficult to reconcile these data with a mechanism which requires pressure and pressure dependent growth of preexisting cracks. There are other features of hydrogen embrittlement such as the reversibility of the incubation time for delayed failure with respect to applied stressg0 and the influence of prestraining31 which are also difficult to explain using a pressure model. Any general mechanism of hydrogen embrittlement will have to consider the reversible aspects of the embrittlement data as well as the irreversible portion which has been clearly presented for the Fe-Si crystals and which is consistent with a pressure model. A. S. Tetelman and W. D. Robertson (authors' reply)—he authors agree with Dr. Steigerwald that a general mechanism of hydrogen embrittlement must be applicable in all cases where embrittlement occurs. Dr. Steigerwald's experiments were performed on an iron alloy which has an extremely complex microstructure, and therefore does not lend itself to the direct and detailed observations that can be made in Fe-Si. The change in the characteristics of cracks as a consequence of annealing at 300°F is not really known in detail but it is probable that stress-relaxation (blunting) occurs, dislocations produced near the crack tips will be pinned by carbon atoms, and hydrogen will be lost to the atmosphere. Any, or all of these effects of annealing will alter the ease with which a crack subsequently propagates under applied stress. A quantitative treatment of the problem must take these effects into account. Since diffusional processes are presumably eliminated at -320°F, the general mechanism proposed by Dr. Steigerwald's colleagues cannot be operative at this temperature. A detailed discussion of the general' inapplicability of this mechanism has been presented elsewhere. However, the data presented by Dr. Steigerwald in his discussion can be explained in terms of a pressure model. Crack propagation under internal pressure P and applied stress occurs when where is the total energy expended in creating new surface area and by plastic work at the crack tip. The total crack length increases with increasing current density, since we have shown that it systematically increases with hydrogen concentration and Dr. Steigerwald has shown that hydrogen concentration increases with current density. For large L (high current density) Eq. [I] can be satisfied even if P = 0 and the embrittlement process does not require the presence of hydrogen. Specimens charged at a lower current density will con-

Jan 1, 1963

By R. C. Gifkins

R. C. Gifkins (CSIRO)—In this paper evidence is put forward to support the idea of grain boundary shearing in aluminum at 4.2°K and the phenomenon is explained in terms of a low-temperature "equicohesive temperature". It is stated that these findings show that grain boundary shear cannot be ruled out in any situation by temperature arguments alone. We believe, from our own recent study of "apparent grain boundary sliding'' at temperatures below 0.45 Tm, that it can be ruled out on temperature arguments alone.45 Our experiments were made principally with magnesium and a Ms-A1 alloy ("Magnox A12"). Careful examination of offsets of marker lines or steps at grain boundaries produced by deformation at room temperature (-0.3 Tm) showed many of these to result from curvature over narrow zones along the grain boundaries. A significant residuum did, however, appear as a sharp offset or step. It was found that the mean value of this step rose to a small value during the first 0.5 to 1 pct strain and then remained at this level with further strain. Moreover, exactly the same values were obtained (at the same stress) at 75°C instead of 21°C. That is, in this range of temperature the "apparent sliding" was not temperature-sensitive. At 125°C and above there was sliding which increased with strain and which was temperature-dependent. When this temperature-dependent sliding was used to calculate the sliding to be expected at room temperature, this underestimated the "apparent sliding" by a factor of 104. The conclusion drawn is that the "apparent sliding" consists of shear localized in a zone of a width measured perhaps in microns, whereas it seems that true sliding occurs on the boundary surface.46 A model for such localized shear has been developed in general terms, following a similar model of Urie and Wain.47 Other strong evidence for low-temper- ature sliding has been based on the offsets of marker lines consisting of a substructure seen using the electron microscope to resolve these.48 We have repeated this work and found that the effects can be accounted for in terms of cracks in an oxide film which is produced when the substructure is etched up; these cracks form over zones of shear rather than at actual interfaces. However, in aluminum at room temperature (-0.3 Tm) there is no "apparent sliding" which is optically detectable. We suggest that, when the temperature is much lower (and possibly the rate of strain higher, too), as in the experiments of Chin et al., then aluminum behaves like magnesium at room temperature, and a narrow zone of shear develops to accommodate differences in strain across the grain boundary. Such a zone could well give rise to triple-point cracks. It is hoped to investigate these suggestions further. B. Y. Chin, W. F. Hosford, Jr., and W. A. Backofen (authors' reply)—It is interesting to have Dr. Gif-kin's comments, although it is somewhat difficult to reply in detail without the reference containing the work that he describes. Apparently though, he does find a low-temperature temperature-independent sliding that would seem to be related to the observations reported here. Perhaps some of the difficulty has semantic origins, growing out of the distinction between ('true sliding" and "apparent sliding".

Jan 1, 1965

By Robin O. Williams

Robin 0. Williams (Oak Ridge National Laboratory)— The authors have questioned the degree to which the coherency strains between the iron-rich and chromium-rich phases are isotropic as proposed in Ref. 5 on the basis of the difference between the elastic properties of the two phases. The relative magnitude of the stresses is determined by the moduli as shown by Eqs. [2], [3], and [4] of Ref. 34. However, the moduli of the two phases have no direct bearing on the uniformity of either the stress or strain within either phase. The idea that the strains are isotropic within each phase (but normally of different magnitude and always of different sign) is based entirely upon the experimental observation that X-ray line broadening has not been detected even when the particles become rather large. It has not proven possible to grow the particles sufficiently large that they lose coherency. Based upon this lack of line broadening one can estimate an upper limit for the nonuniformity of the strains within each phase as follows. It is considered possible to detect line broadening if it is as great as 10 pct of the separation of the K, doublet for the (211) line using chromium radiation. The doublet separation would correspond to a total strain of 0.0017 such that the total variation of lattice parameter relative to the average lattice is now k0.05x0.0017 or something less than ± * For the present case the strain in each phase is roughly 0.002 such that the variation of strain within a phase will not exceed 5 pct. It is stated that the expression derived for strengthening for the hydrostatic straining as observed in this system would substantially overestimate the magnitude due to dislocation flexure. This is contrary to the conclusion reached in the original paper34 for the present range of particle sizes. What is the lowest temperature at which a has been observed to form in this alloy? M. J. Marcinkowski, R. M. Fisher, and A. Szirmae (nutlzors' reply)— -Williams' arguments based on X-ray findings for a chromium-rich precipitate and an iron-rich matrix strained to a common lattice parameter are certainly convincing. This being the case, there are no shear components of strain associated with the precipitate-matrix aggregate to interact with the shear components of the dislocation stress fields, contrary to the opinion expressed by the present authors. On the other hand, the present authors, in spite of this error, did not expect the shear interactions to be significant. The chief objection to Williams' model in the present case is that the various segments of the dislocation line are assumed to pass from one potential valley to the next independently of neighboring segments. This is only true for a highly flexible dislocation line, i.e., one whose radius of curvature is something less than the center to center distance between precipitate particles which amounts to about 90A in the present alloy. In order to maintain this curvature, an externally applied shear stress of at least 230,000 lb per sq in. would be required or about four times the observed stress. It is therefore concluded that the dislocation lines move rather rigidly through the lattice. This being the case, the forces on the dislocation resulting from the hydrostatic interaction between the stress fields of the edge-dislocation components and the precipitate particles should average out to zero; that is particles above the below the slip plane produce forces on the dislocation of opposite sign and therefore will cancel when averaged over the entire length of the dislocation. On the other hand, since the dislocation is not perfectly rigid, Williams' model may lead to some strengthening, but far less than that predicted. A second and equally serious objective to using Williams' strengthening model for the present alloys is that profuse wavy slip due to the motion of screw dislocations played a predominant role not only in the unaged alloys but in the fully aged ones as well. Since the screw dislocation has associated with it only shear components of stress the hydrostatic strengthening model no longer applies. In view of these arguments the present authors must reject Williams' model of strengthening as being pertinent to the present alloy system. The present authors have made no detailed study of the lowest temperature at which a forms in the quenched ferritic alloys. None was ever observed n the alloys aged at 500°C so that forma-tion must occur at temperatures higher than this and was therefore not a factor in the present study.

Jan 1, 1965

By I. R. Kramer

I. R. Kramer (Martin Co.)—In a recent paper Nakada and Chalmers24 reported some observations of effects of surface conditions on the stress-strain curves of aluminum and gold single crystals. It is of interest to compare these observations with the results published previously in this journal and to comment on their general conclusions. In brief, Nakada and Chalmers concluded that the removal of the surface layer of a prestrained specimen lowered the stress-strain curve for aluminum but not that for gold. Further, they concluded that the surface work hardening of aluminum is confined to a depth not more than l0-3 cm. In our results25 published previously, it was pointed out that when prestrained specimens of aluminum and gold were polished to reduce the thickness upon reloading the initial flow stress decreased markedly. Further if a sufficient amount of metal was removed, the yield point failed to appear. With continued application of the load the stress-strain curve became coincident with that of the virgin crystal. We have found this behavior in some 100 determinations to hold consistently for gold, aluminum, and copper in both single and poly-crystalline specimens. The amount of strain required before the curves coincided depends upon the amount of metal removed but it is usually less than 0.01. This type of behavior is the same as that reported for metarecovery by other investigators.29 For aluminum specimens which have been prestrained and then heated to temperatures above 50°C we have consistently found that the stress-strain curve was typical of the orthorecovery28 type. In this case the stress-strain curve always lies below that of the virgin specimen. With respect to Ref. 24 the curves for aluminum are always below that of the virgin curves, while those for gold become coincident. This observation indicates at least for aluminum that the specimens must have been heated to a temperature high enough to cause recovery by an alteration of the internal dislocations. In addition, a recovery would be expected because of the removal of the surface work-hardened layer. Nakada27 had reported that, with his particular apparatus in which a perchloric acid polishing solution was used, the temperature of the specimen increased 65°C. The curves presented in Ref. 24 for gold do not permit one to detect the initial flow stress upon reloading after the surface-removal treatment. In fact, contrary to the method used by Nakada and Chalmers, the change in the stress-strain curve produced by a surface-removal treatment cannot be described in terms of a decrease in stress at strains much higher than that at the initial flow stress because of the coincidence of the curves at the higher strain values. With regard to the depth of the work-hardened surface layer, our data show for aluminum single crystals (7.5 by 0.3 by 0.3 cm) that the initial flow stress remained constant after 12 x 10-3 cm had been removed from the thickness. This depth was independent of the prior strain. For gold crystals this depth is somewhere between 10 x 10-3 and 20 x 10-3 cm. Y. Nakada (author's reply)— Kramer states,28 "for aluminum specimens which have been prestrained and then heated to temperatures above 50°C, we have consistently found that the stress-strain curve was typical of the orthorecovery28 type. In this case the stress-strain curve always lies below that of the virgin specimen." However, Cherian et a1.26 discovered that aluminum polycrystals annealed at 32" and 100°C showed the metarecovery behavior. They showed that the orthorecovery behavior did not appear until the crystals were annealed at 150°C. Kramer suggests28 that the drop in the flow stress of aluminum crystals after the surface removal by electropolishing24 may be due to the recovery caused by the temperature rise which may occur during the electropolishing.27 However, as indirectly stated in Ref. 24, the current density used in these electro -polishing experiments was 0.15 amp per sq cm. According to ref. 27, this current density should cause a temperature rise of only 40°C. This may cause the metarecovery but not the orthorecovery. Furthermore, as stated explicitly in Ref. 24, the surface removal was accomplished by chemical etching as well as by electropolishing. The results were the same for both electropolishing and etching. During the etching, the specimen temperature did not rise above 35°C. Some aluminum crystals were placed in water at 35° C for 30 min. These crystals did not show the decrease in flow stress. Therefore, it is quite clear that the flow-stress drop after the surface removal is not caused by a high-temperature recovery. However, as Kramer points out,28 it is quite possible that the internal dislocation structure may have been altered because of the removal of the highly work-hardened surface layer. However, how much this rearrangement contributes toward the flow-stress decrease is not known at present.

Jan 1, 1965

By M. van Swaay, C. E. Birchenall

A. S. lBrling (Johnson, Matthey & Co. Ltd., Laboratories)— Because of its initial emphasis upon the production of membranes by powder metallurgy and by normal casting and rolling techniques, this paper is of considerable interest to those concerned with the design and application of industrial diffusion units. As the subsequent performances of the three grades of palladium were not compared in detail, it can perhaps be concluded that specific permeabilities were very similar. A short table in which the absolute diffusion rates obtained by the Authors with their three grades of palladium were compared with those of other investigators18-25 would, however, have been of particular value in view of the very variable results obtained with this membrane material. The Authors state early in the paper that the permeability of the thoriated samples decreased in much the same way as that of the nonthoriated samples. Later, they emphasize that sintered palladium without thoria was very susceptible to poisoning and did not respond as well as the other materials to the decontamination treatment. These results, which suggest that sintered membranes are less effective than those produced by normal methods, have considerable industrial implications, and amplified experimental details would be very welcome. The results of the Authors differ in one important respect from those of other investigators. Thus, no "bleeding" was required to maintain a constant rate of diffusion, although Davis, Darling, and De Rosset found this to be necessary. Davisz3 found that "bleeding" was necessary even when the hydrogen had already been purified by diffusion through a palladium tube, and it would be of value to know the purity of the gas stream emerging from the purification train shown on Fig. 1 of the Authors' paper. The Authors show that diffusion rates varied according to the square root of the pressure difference providing that the palladium membrane was not contaminated in any way. This finding is in agreement with the results of those workers2,18,19,23,24 who diffused their hydrogen through the membrane into a partly or completely evacuated system.Those workers who did not find this square-root law dependence did not have a partial vacuum on the downstream-side of the membrane.'' The difference between the two experimental con- ditions appears to be significant, as the absolute diffusion rates per sq cm of diffusion area are lower when the square-root law dependence is not observed. While the effect could be attributable to poisoning of the membrane it is conceivable that a positive pressure on the downstream side of the membrane might influence the mechanism of diffusion. In view of the capricious behavior of palladium, factors of this type, which are likely to influence its industrial significance, should be carefully considered. M. van Swaay and C. E. Birchenall (authors' 9-eply) — Our Eq. 191 may be compared with several others for hydrogen diffusion in palladium: D = 5.95 x exp (-5770/1)' D = 3.55 x 10"3 exp (-5550/)' D = 4.3 x 10"3 exp (-5620/1')' Other values are given in the literature, but in many cases surface poisoning may have affected the results. No analysis was performed on the hydrogen gas stream after purification. It should be noted that the presence of liquid nitrogen traps near the heated membranes may have set up convection currents and provided the equivalent of "bleeding." The question of whether a partial vacuum existed on the downstream side of the membrane is not clear. Does it imply that there is something unique about the permeability measured near one atmosphere pressure? We do not believe that there is. When square root pressure dependence is not observed, diffusion rates are not measured. Is "absolute diffusion rate per sq. cm. of diffusion area" equivalent to permeability? The terminology is new to us. In any case we doubt very much that pressures of the order of several atmospheres can affect the mechanism of diffusion.

Jan 1, 1962

By Elmars Ence, Harold Margolin

A. J. Goldak and J. Gordon Parr (University of Alberta) —While we appreciate the difficulties involved in any investigation of this system, and we wish to congratulate the authors on their comprehensive efforts, there are several aspects of their paper that are not clear to us. 1) If the 6 phase is, in fact, isomorphous with Ti3Sn, it should not be designated Ti,Al. An 4B structure cannot be isomorphous with an ordered structure of the DO type. Anderko et al.' suggested that the structure be called Ti3A1, and we believe this to be a more reasonable designation. The confusion is compounded when the authors designate the phase Ti3Al—yet the composition limits of this phase (according to their diagram) are approximately Ti,A1 and Ti,Al. Further, while Pietro-kowsky is said to have pointed out that the structure of the 6 phase is isomorphous with Ti3Sn, we understand that this was only a suggestion that was not confirmed. Margolin and Ence do not appear to have presented any evidence to support the suggested isomorphism. 2) In referring to the precipitation observed microscopically during the initial breakdown of high A1 a during quenching, the authors state that "such precipitation is not likely to be detected by Debye-Scherrer techniques". We suspect that if the precipitate was visible by optical microscopy, it must have at least one dimension of the order of 1 or greater, and would therefore be readily detected by diffraction techniques. 3) The authors state: "The basis for this suggestion (that 6 forms peritectically P + L—6) is the observation that the ,'3 + 6 and 6 phase fields were

Jan 1, 1962

By W. L. Worrell

R. C. Gifkins (CSIRO)—In this paper evidence is put forward to support the idea of grain boundary shearing in aluminum at 4.2°K and the phenomenon is explained in terms of a low-temperature "equicohesive temperature". It is stated that these findings show that grain boundary shear cannot be ruled out in any situation by temperature arguments alone. We believe, from our own recent study of "apparent grain boundary sliding'' at temperatures below 0.45 Tm, that it can be ruled out on temperature arguments alone.45 Our experiments were made principally with magnesium and a Ms-A1 alloy ("Magnox A12"). Careful examination of offsets of marker lines or steps at grain boundaries produced by deformation at room temperature (-0.3 Tm) showed many of these to result from curvature over narrow zones along the grain boundaries. A significant residuum did, however, appear as a sharp offset or step. It was found that the mean value of this step rose to a small value during the first 0.5 to 1 pct strain and then remained at this level with further strain. Moreover, exactly the same values were obtained (at the same stress) at 75°C instead of 21°C. That is, in this range of temperature the "apparent sliding" was not temperature-sensitive. At 125°C and above there was sliding which increased with strain and which was temperature-dependent. When this temperature-dependent sliding was used to calculate the sliding to be expected at room temperature, this underestimated the "apparent sliding" by a factor of 104. The conclusion drawn is that the "apparent sliding" consists of shear localized in a zone of a width measured perhaps in microns, whereas it seems that true sliding occurs on the boundary surface.46 A model for such localized shear has been developed in general terms, following a similar model of Urie and Wain.47 Other strong evidence for low-temper- ature sliding has been based on the offsets of marker lines consisting of a substructure seen using the electron microscope to resolve these.48 We have repeated this work and found that the effects can be accounted for in terms of cracks in an oxide film which is produced when the substructure is etched up; these cracks form over zones of shear rather than at actual interfaces. However, in aluminum at room temperature (-0.3 Tm) there is no "apparent sliding" which is optically detectable. We suggest that, when the temperature is much lower (and possibly the rate of strain higher, too), as in the experiments of Chin et al., then aluminum behaves like magnesium at room temperature, and a narrow zone of shear develops to accommodate differences in strain across the grain boundary. Such a zone could well give rise to triple-point cracks. It is hoped to investigate these suggestions further. B. Y. Chin, W. F. Hosford, Jr., and W. A. Backofen (authors' reply)—It is interesting to have Dr. Gif-kin's comments, although it is somewhat difficult to reply in detail without the reference containing the work that he describes. Apparently though, he does find a low-temperature temperature-independent sliding that would seem to be related to the observations reported here. Perhaps some of the difficulty has semantic origins, growing out of the distinction between ('true sliding" and "apparent sliding".

Jan 1, 1965

By J. Suiter, H. F. Ryan

H. F. Ryan and J. Suiter (CSIRO)—In this paper the authors have presented "a physical model which has as its central hypothesis the solution strengthening of regions along grain boundaries in the order of tens of angstroms thick", and their calculations have used 40Å as the zhalf-width of this boundary region. Selection of this value by the authors suggests that they were unaware of a number of recent field-ion microscope observations of grain boundary structures (e.g., Refs. 29 and 30). The accompanying figure, Fig. 9, shows a portion of an ion micrograph of a tungsten specimen containing a grain boundary, and the regular arrangement of atoms in various crystal planes in each grain can be seen, except in the boundary zone which is not more than two atom diameters in width. Since it is expected that solute atoms will segregate only to the region where the lattice is disturbed, it would seem that a value of 4Å for the half-width of the boundary region would be more appropriate than the value of 40Å assumed by the authors. If the former value is used in Eq. [8] of their paper to obtain curves showing solute concentration in the grain boundary as a function of temperature, the resulting curves differ significantly from those obtained by the authors. Although there are only relatively slight changes in the regions corresponding to higher temperatures (i.e., grain boundary concentrations of approximately 1 at. pet), at lower temperatures the curves are no longer realistic. For example, the curve corresponding to u = 7000 cal per mole gives a solute grain boundary concentration in excess of 100 pet at a temperature of approximately 360°K, and "flattens out" at a grain boundary concentration of about 300 pet! Similar results are obtained with each of the other curves. Ion microscopy of dilute solid-solution alloys has shown that solute atoms give rise to image points which in most cases are of higher intensity than those of the solvent atoms.31 If the same behavior occurs when the solute atoms are present in much lower concentrations then ion microscopy should provide a sensitive tool for studying the segregation of impurities to the grain boundary regions. Although many grain boundaries in tungsten have been examined after different thermal and mechanical treatments, there has not been, as yet, any systematic study of such segregation. In general the micrographs show no clear indications of marked grain boundary segregation, but occasionally areas are found where the grain boundary contains an appreciable number of image points of high intensity in a very narrow zone, see Fig. 9. If it is assumed that these correspond to impurity atoms then it is possible to estimate the impurity concentration both in the grain boundary and within the body of the grains. Application of this method (to a larger area

Jan 1, 1965

By Arthur G. Metcalfe

Arthur G. Metcalfe (SOLAR)—Three explanations to account for mechanical nucleation effects are considered in the paper, viz., theories based on fragmentation, agglomeration, and pressure-wave mechanisms. The authors show that none of these mechanisms will explain the results observed with tin and Sn-Pb alloys. A fourth mechanism of nucleation enhancement was developed during work at Armour Research Foundation on casting of beryllium and, although presented briefly on various occasions, the theory has never been discussed adequately. It is believed that this theory will explain many of the observations. Classical homogeneous nucleation theories assume that embryo nuclei are formed in the melt as a result of compositional and/or thermal fluctuations. Stable nuclei are formed when a critical size is reached such that the decrease in free energy for conversion of this volume of melt to solid offsets the increase of surface, energy. The assumption is made that the average properties of liquid and solid metal can be used. In practice this assumption is invalid and indeed is contrary to the postulation of thermal and compositional fluctuations in the melt. Actually the conditions favoring embryo formation bring the seeds of its own destruction—the thermal and compositional fluctuations leave the embryo surrounded by atoms of higher thermal energies and/or by solute atoms that depress the melting point. Any mechanical method of dispersing this destructive

Jan 1, 1964

By John L. Miles

John L. Miles (Arthur D. Little, 1nc.)—Pollack and orris" have reported measurements on electron tunneling through A1-A12O3-A1 sandwiches in which the oxide was formed by gaseous oxidation in a glow discharge. From these measurements they deduced the asymmetry of the barrier and, since this is small, conclude that the mechanism suggested by Mott19 for the growth of oxide in thin A12O3 films is inapplicable. In earlier papers20 Pollack and Morris report similar work for oxide films grown thermally. In this case they find a greater asymmetry and conclude that the Mott mechanism is valid. I would like to point out that both these conclusions are quite unjustified. Mott suggests that the growth of the oxide film on aluminum results from the passage of ions through the already present film of oxide under the action of an electric field. This field results from a constant voltage which is in effect a contact potential between metal on one side of the barrier and adsorbed oxygen ions on the other side of the barrier. The theory does not require that the oxide grown is nonuniform either in stoichiometry or structure. It does however specifically assume that the partial layer of ionized oxygen on the surface remains adsorbed on the surface of the growing oxide. In other words, the so-called "built-in field" remains in the oxide only as long as the ionized oxygen is present. When a counter electrode of aluminum is deposited on the oxide, it will react with the adsorbed oxygen on the surface of the oxide, thus forming a small additional amount of oxide. It is clear, then, that there is no requirement in the Mott theory of oxide growth which would necessitate tunneling currents through an Al-A1203-A1 sample to be different when the polarity is reversed. Neither does the theory eliminate the possibility that some additional mechanism could cause the tunneling barrier to be asymmetric and hence tunneling currents to be a function of polarity in such a sandwich. Thus these tunneling-currents measurements are not germane to the question of whether the Mott mechanism is the true method of growth of aluminum oxide films. In fact, it is not surprising that there should be a difference between the oxide properties at the two interfaces (with resulting asymmetry in the tunneling barrier) since the growth conditions and growth rates must have been quite different at these two positions. S. R. Pollack and C. E. Morris (authors' reply)— The point raised by Miles above is one has caused some confusion in the past. The following is an attempt to clarify this point. The built-in field which is responsible for the growth of the thermal oxide at low temperatures arises, according to Mott, because of the passage of electrons from the Fermi surface of the oxidizing metal to surface states introduced by the adsorbed oxygen. It is assumed that the energy of these surface states lies below the Fermi energy of the metal. Electrons therefore continue to flow from the metal to the surface until the built-in electric field raises the potential energy of the surface states to the value of the Fermi energy in the metal, at which time equilibrium is obtained between the surface states and the metal. That is in equilibrium as many excess electrons pass from the metal to the surface per unit time as vice versa. The surface of the oxide prior to deposition of a metallic counterelectrode can then be pictured as follows. The Fermi energy lies in the energy gap of the oxide and is essentially pinned at the energy of the oxygen surface states. The vacuum work function of the oxide is then given by the sum of the electron affinity of the oxide (i.e., the difference in energy between the vacuum and the conduction-band minimum) plus the energy difference between the conduction-band minimum and the Fermi energy. The deposition of a metal onto the surface of the oxide can result in a transfer of electrons across the extremely thin oxide only if there is a contact potential difference between the deposited metal and the parent metal or oxide. That is if the vacuum work function of the deposited metal differs from that of the parent metal, then charge can be redistributed across the oxide in order to equilibriate the Fermi energy across the structure. (It should be

Jan 1, 1965