Search Documents

By J. T. Williams

The equilibria in the V-Zr alloy system were investigated by solidus temperature determinations, thermal analysis, dilatometry, electrical resistance measurements, microscopic examination, and X-ray diffraction analysis. There is a eutectic reaction at 1230°C between a compound, V2Zr, and a solid solution containing 10 pct V in ß zirconium. V2Zr decomposes at 1300°C into liquid and a solid solution containing about 10 pct Zr in vanadium. The eutectic composition is probably about 30 pct V. A eutectoid reaction between V2 Zr and a zirconium takes place at 777°C at a very high rate. The eutectoid composition is 5 wt pct V. The limit of solubility of zirconium in vanadium was estimated to be 5 pct at 600°C. No attempt was made to determine the liquidus for the system. THE recent availability of large quantities of high purity zirconium has stimulated the study of zirconium binary systems. The equilibrium diagram for the V-Zr system has received little attention, however. Wallbauml appears to have made the first report concerning the equilibria in these alloys. He reported the existence of a compound, V2Zr, having the MgZn2 ((214) type of structure with a, = 5.277 kX and c° = 8.647 kX. Anderson, Hayes, Roberson, and Kroll2 made a survey of some potentially useful zirconium binary alloys and found that zirconium probably dissolves a small amount of vanadium. They reported the probable existence of a compound between the two elements and suggested that the zirconium-rich solid solution undergoes a eutectic reaction with this compound. Pfeil," in a critical review of the existing information, estimated that the solubility of vanadium in zirconium is less than 4.7 pct and probably less than 1.8 pct. Rostoker and Yamamoto' proposed a partial diagram for the V-Zr system in a survey paper on vanadium binary alloys. Their diagram indicates the compound, V,Zr, a eutectic reaction at 1360°C, a peritectic reaction at 1740°C, and a limit of solubility of zirconium in vanadium of about 3 pct. They obtained no information on the equilibria in the zirconium-rich alloys. In view of the potential utility of the V-Zr alloys and the incomplete knowledge concerning the equilibria in the system, an attempt was made to establish the constitutional diagram. Preparation of the Alloys Raw Materials: The vanadium for making up these alloys came from the Electro Metallurgical Corp. Zirconium came from two sources. In the beginning of the investigation, sponge zirconium from the Bureau of Mines was used in making some of the alloys. Later, iodide metal made at the Westinghouse Atomic Power Development Laboratories became available. This material was used in the preparation of all the dilatometric and resistance specimens and about two-thirds of the solidus temperature specimens. A typical manufacturer's analysis of the vanadium is shown in Table I. No other analysis of the vanadium was made. The metal contained a dispersed second phase and did not have a sharp melting point. Typical results of spectrographic analysis of the Westinghouse zirconium are shown in Table 11. These data indicate a very high purity. The Bureau of Mines sponge metal was probably less pure but had good ductility. Melting: All of the alloys used in the investigation were made by melting pieces of vanadium and zirconium together in a dc electric arc furnace similar to those of Geach and Summers-Smith, craighead, Simmons, and Eastwood," and others. Melting was done in an atmosphere of helium scavenged of residual air by the preliminary melting of a separate charge of zirconium. Each ingot was turned over and melted at least three more times before removal from the furnace to aid in the attainment of homogeneity. Alloys prepared for use in the investigation are listed with the results of solidus determinations in Table III with the exception of the following compositions upon which no solidus determinations were made: 0.29, 0.54, 4.57, and 5.55 pct V. Analysis: The weight of each ingot made from iodide zirconium was within 0.1 g of the total weight of the initial charge, about 90 g. Since each component of each charge was weighed to the nearest 20 mg for amounts less than 10 g and to the nearest 0.1 g otherwise, the gross composition of an ingot could be calculated accurately. Chemical analysis for the vanadium content of several alloys agreed

Jan 1, 1956

By P. Chiotti, G. R. Kilp

IN the study of zinc-zirconium1 and zinc-uranium2 systems, vapor-pressure measurements were made in order to establish the equilibrium phase diagrams for a pressure of 1 atm. In each of these systems, only zinc has an appreciable vapor pressure over the temperature range investigated. The zinc pressure was measured over a sufficiently wide temperature and composition range to permit calculation of the standard free energy, enthalpy, and entropy of formation for the compounds ZrZn, ZrZn2, ZrZn3, ZrZn6, ZrZn14 , and U2Zn17. In these calculations the compounds were treated as line compounds. Although the exact composition range was not determined, metallographic and thermal data indicate that the solubility range for these compounds is small. It is to be noted that the compound initially reported to correspond to the formula UZn9, 2 has been shown by the crystal structure work of Makarov and vinogradov3 to correspond to the formula U2Zn17,. The crystal structure is similar to that for the compounds Th2M17, investigated by Florio,4 et al; here M represents iron, cobalt, or nickel. Therefore, thermodynamic calculations are based on the formpla U2Zn17 EXPERIMENTAL The dew-point method employed to measure the vapor pressure of zinc over uranium-zinc alloys has been described previously.2 The same procedure and apparatus were used in measuring the zinc pressure over zinc-zirconium alloys. In the case of alloys in the two-phase region consisting of a zirconium and the compound ZrZn, the Knudsen effusion method was employed to obtain a few values in the temperature range 600" to 560°C. In most cases, tantalum crucibles were used to contain the zirconium alloys. THERMODYNAMIC PROPERTIES OFU2Zn17 A least-squares treatment of the vapor-pressure data obtained for 35 and 65 wt pct U alloys in the temperature range 650° to 940°C gave lOg10PAtm = 7,713/T + 6.33 Here, as in following equations, T represents degrees kelvin and P represents the zinc vapor pressure in atmospheres. The probable errors for the two constants were calculated to be 60.68 and 0.0563, respectively. This relation represents the equilibrium dissociation pressure for the reaction U2Zn17 (s) - 2U(s) + 17Zn(v) The solubility of zinc in solid uranium is negligible and the only additional information required to calculate the free energy of formation of U2Zn17 is the free energy of vaporization of pure zinc. K. K. Kelley5 gives the relations, AF°= 30,902 + 6.03Tlog T + 0.275 X 10-3T2 - 45.03T

Jan 1, 1961

By Glenn R. Zellars, J. P. Morris

The carrier gas method was used to measure the vapor pressure of copper over liquid copper and Fe-Cu alloys. From these results, the activity of copper in the alloys was determined at 146°, 1550°, and 1600°C. Calculations of the activity of iron were made from the results for copper. The activities of copper and iron show strong positive deviations from Raoult's law. KNOWLEDGE of vapor pressures of liquid metals has both practical and theoretical value. Fuming of metal baths is directly related to vapor pressure, and the data should bc particularly important to the technology of vacuum metallurgy. Theoretical applications include determinations of heats of vaporization, heats of solution, and activity coefficients in alloys. Methods for measuring vapor pressures of metals have been reviewed by Speiser and Spretnak.' High-vacuum procedures, such as the Langmuir and Knudsen methods, have been used most widely in recent years. However, these methods are not applicable to vapor pressures greater than lo-' mm Hg and often cannot be used in the temperature range of greatest interest. The purpose of this investigation was to develop a procedure for measuring vapor pressures of pure metals and alloys that would be applicable to pressures greater than lo-' mm Hg. The transportation or carrier gas method has been used. Copper was chosen for the initial work because its vapor pressure has been established fairly well and the reliability of the technique could be ascertained. Applicability to alloys was studied by determining the vapor pressure of copper over liquid Fe-Cu alloys. Experimental Procedure The tests were carried out in a graphite-spiral resistance furnace. The charge consisted of 50 g of metal contajned in a covered sintered-alumina crucible 1/2 in. deep and 13/8 in. ID. Two sintered-alumina tubes led from outside the furnace to within the crucible, passing through close-fitting holes in the crucible cover and radiation shield. The arrangement of the crucible and the two alumina tubes is shown in Fig. 1. The small-bore tube (1/16 in. ID by 5/16 in. OD) served as the inlet for the carrier gas. The tip of this tube touched the metal surface and the flow of gas imparted a rippling motion to the metal. In this manner the gas was brought into close contact with the metal and became saturated with metal vapor. In most of the tests the flow rate of the carrier gas was 560 ml per min, measured at room temperature. Approximately half of this saturated gas was allowed to escape from the crucible through the second alumina tube (3/16 in. ID by 5/16 in. OD) and was discharged to the atmosphere through a

Jan 1, 1957

By C. C. Herrick

The vapor pressure of indium has been measured by the torque-effusion technique, as a function of temperature between 1102o and 1422oK. For liquid indium, the vapor pressure (in atmospheres) can be represented by the equation -R In P = 5.782 x104/T - 25.29from 1000o to 1422oK. Extrapolation yields a calculated normal boiling point of 2329°K and a heat of vaporization at 298 oK of 58.09 kcal per mole by the third-law method and of 58.39 kcal per mole by the second-law method. Calculations indicate that the free-energy functions of Hultgren are to be preferred to those of Guruich. BECAUSE of the potential uses of liquid metals, it is important that convenient experimental methods be developed which will lead to a sufficient body of thermodynamic data so that reasonable estimates of properties of liquid metals can be made. Indium is particularly well-suited to investigation, since it remains liquid over a long temperature interval, it alloys readily, suitable containers for it are easily procured, and relatively few investigations of its thermodynamic properties in the region above 1000°K have been made. PREVIOUS RESEARCH The first determination of the vapor pressure of indium was made by Anderson,' using the Knudsen method and silica-lines stainless-steel cells. Kohlmeyer and spandau2 determined the normal boiling point, and a mass-spectroscopic investigation has been made by Lyubimov and Lyubitov.3 (Unfortunately, only an equation representing the data of the latter investigators has been available to us.) Hultgren et al.4 and Gurvich5 have published discordant tables of the free-energy functions for indium. EXPERIMENTAL METHOD The torsion-effusion principle has been described amply in the literature.8,7 With this method, one observes the angular deflection induced by effusion of vapor from a multihole effusion cell which is suspended from a filament producing a small rotational restoring force. The pressure within the effusion cell can be computed from the relation ziaifiai where k is the torsion constant of the filament, 6 is the observed angular deflection, and ai, fi, and qi are the area, force factor, and moment arm of the ith effusion orifice. The force factor f is analogous to the Clausing factor,' and corrects for the reduction in effusive force attributable to the finite geometry of the effusion orifice. Schultz and Searcy' have tabulated the force factors for various tube geometries. EXPERIMENTAL APPARATUS The physical arrangement employed in this study is shown in Fig. 1. The furnace elements are six Globar heat rods with 15-in. hot zones. Temperature is maintained to * 1°K by a Beck mechanism driving two 26-amp Powerstats in parallel. A Pt (Pt, Rh 10 pct) thermocouple serves as the sensing element. Four inner tubes act as vibration isolators; the air pressure required for maximum isolation was determined empirically, and is maintained by a solenoid valve. The tungsten fiber, 9 by 0.002 in., is joined to an 0.008-in. tungsten wire or graphite rod via a machined lightweight chuck.

Jan 1, 1964

By C. E. Birchenall, C L. McCabe

IN attempting to extend vapor pressure measurements of the type previously reported by Schadel and Birchenall1 for silver and by Schadel, Derge, and Birchenall' for silver-silicon to other systems, it was observed that the materials melted at indicated temperatures 10" to 15" below their accepted melting points. Further investigation revealed that the thermocouple readings were in error due to appreciable conduction losses along the reference thermocouple wires. If the wire diameter of the reference couple inserted into the Knudsen cell was reduced, the correction for the indicating couple changed in a manner tending to explain the melting behavior. When extrapolated to zero wire diameter from measurements with several reference thermocouples of different wire thickness, the melting point of silver then agreed with the indicated temperature at which silver chips were observed to coalesce into a sphere. Approximately the same calibration was given by observing the melting of small wires of silver or gold in the Knudsen cell connected in series with an ammeter, where the leads into the cell were very fine in order to minimize heat conduction. Unfortunately neither of these methods seemed to yield a sufficiently precise temperature calibration to match the apparent precision of the other aspects of the vapor pressure measurement. It was decided. therefore, to redetermine the vapor pressure of silver in another setup under conditions permitting precise temperature measurement. The vapor pressure of pure silver could then be used as an internal calibration of temperature in the older unit in making runs on alloys. This has been done; the present report is a correction to ref. 1. Experimental Procedure The apparatus, shown in Fig. 1, was very similar to that employed by Harteck,3 except that the orifice sizes were smaller and the residual pressure in the vacuum system was probably much lower. A small, sharp-edged hole, nearly circular in shape, was ground into the rounded end of a quartz tube. The orifice area was then measured by tracing the image at known magnification on graph paper and counting the squares enclosed. The silver specimen was sealed into the tube to make a Knudsen cell. A tantalum jacket surrounding the cell served to increase the uniformity of temperature. This assembly was placed in the bottom of a long quartz tube with an inside diameter of about 1 in., which was connected to the vacuum system through a ground joint sealed with picein wax well removed from the furnace. A thermocouple tube inserted through the top of the vacuum line reached into the tantalum jacket so that the thermocouple junction was immediately adjacent to the Knudsen cell except for the protection tube wall. A resistance furnace could be raised to cover the end of the quartz tube containing the cell in such a way that the cell was in the uniform temperature zone 13 in. from the end of the furnace. An ionization gage was included in the vacuum system in the cold lines of wide diameter, immediately beyond the ground joint. The vacuum system consisted of a mercury one-stage diffusion pump, backed by a Welch duo-seal mechanical pump. The pumps were separated from the reactor chamber by a dry ice trap. The ionization gage always read less than 10-5 mm Hg after initial outgassing and before each run was started. Each newly filled Knudsen cell was evacuated at high temperature overnight before the first weighing was made. The cell was returned to the system, heated for a measured time at constant temperature, cooled, and reweighed. The heating and cooling times were quite short since the hot furnace was raised to receive the reactor at the beginning of the run and removed again at the end. The tube heated or cooled quickly. The total mass loss was attributed entirely to effusion of silver vapor from the quartz cell, since empty quartz cells maintained constant mass through similar heating cycles. The vaporized silver condensed on the cold walls of the quartz tube extending above the furnace. Earlier studies in the induction heated unit had shown that the same vapor pressure was found for silver, whether the silver was in contact with the tantalum metal cell or with porcelain or quartz liners. The Pt-Pt-10 pct Rh thermocouple was calibrated against a secondary standard of the same material and found to agree with the published tables. Always operating in air at temperatures below 100O°C,

Jan 1, 1954

By C. E. Birchenall, H. M. Schadel

Vapor pressure of silver over silver-gold alloys has been measured over a range of temperatures for four compositions. Orifice effusion has been compared with electromotive force measurement as a means for determining thermodynamic activities in these solid metallic alloys. Correlated activity coefficients are given for silver-gold alloys for the temperature range 200' to 1000°C. SOUND development of a technique for the measurement of some property of matter requires a comparison with other recognized procedures which have been used to measure the same property. Such a check is especially useful if the assumptions on which the applications stand are basically different for the two methods. Agreement in the results tends to confirm both sets of assumptions. The electromotive force method for determining thermodynamic activities depends upon the reversibility of the electrode processes and a knowledge of the valence states involved. The orifice effusion technique of determining vapor pressures requires that certain mean free path conditions be met, that the state of association of the vapor be known, and that efficient collection of vapors be maintained. Thus the requirements are quite different in nature. Although procedures often exist by which the fulfillment of these conditions may be tested internally, agreement of thermodynamic activities for a single system obtained by both methods constitutes a more severe test of the assumptions. The binary alloy system silver-gold is excellent for this purpose for several series of emf measure-

Jan 1, 1954

By E. A. Gulbransen, K. F. Andrew

Weight loss measurements were made using a sensitive microbalance operating in a high vacuum system. The Langmuir equation was used to Calculate the vapor pressures of the several metals and alloys. Aluminum lowered the vapor pressure of i9,on in a 5.4 Al-94.6 Fe alloy, and of iron and chromium in a 4.8 Al-21.5 Cr-73.7 Fe alloy. The influence of oxide films formed on the alloys on the effective vapor pressures of the alloys was also studied. These studies were used to aid in the inte9,pretation of the high temperature oxidation of alloys based on iron, chromium, and aluminum. RECENT studies on the oxidation of chromium1 and the heat resistant alloy 5 Al-22 Cr-73 Fe showed transitions in the rate of oxidation at 900o and 1050oC respectively. The transition for chromium occurred at a temperature where the rate of evaporation of chromium from oxide-free surfaces equalled the rate of chromium atoms reacting with oxygen. This paper presents new vapor pressure studies on iron, chromium, and the specially prepared alloys 5.4 Al-94.6 Fe, 21.9 Cr-78.1 Fe, and 4.8 Al-21.5 Cr-73.7 Fe. The ternary alloy is similar in basic composition to the patented heat-resistant commercial alloy known as Kanthal. The binary alloys were studied to show the individual effects of aluminum on the vapor pressure of iron and of iron on the vapor pressure of chromium. The purpose of the proposed studies was to test the influence of metal volatility on the oxidation behavior of heat-resistant alloys based on iron, chromium, and aluminum. Since oxide films formed on the metal may limit metal transfer to the surface, the influence of oxide films on metal volatility was also studied. LITERATURE A) Vapor Pressure. The vapor pressure of iron has been studied by Jones, Langmuir, and Mackay,' Dornte and Nrton. Edwards. Johnston. and Ditmar. and McCabe, Hudson, and Paton. Stull and SinkeT have calculated a heat of sublimation at 298° K of 99.83 kcalper g atom. The vapor pressure of chromium has been reported by Bauer and Brunner,' Speiser, Johnston, and Blakburn, Gulbransen and Andrew,'' Vintaikin,o McCabe, Hudson, and Paton, and Kubaschewski and Heymer.12 Good agreement was obtained except for the early work of Bauer and Brunner.' Stull and Sinke7 calculate a heat of sublimation at 298°K of 95.0 kcal per g atom. Vapor pressure measurements on aluminum have been made by Brewer and Searcy,13 Bauer and Brunner,aand Farkas.I4 Stull and Sinke,7 giving Brewer and Seary's' data the most weight, derive a heat of sublimation of 77.5 kcal per g atom at 298 oK. B) Method. Two methods15 are used for measuring the vapor pressure of metals: 1) The Langhuir free evaporation method, and 2) the Knudsen effusion method, In the Langmuir method, used in the present work, the vapor pressure of the metal P is given by the equation: Here M is the molecular weight of the vapor species. T is the absolute temperature. (dw/dt) is the rate of sublimation in g per sq cm per sec and a is the condensation coefficient. a is a measure of the efficiency of condensation of molecules striking the metal surface from the vapor. If condensation results from each collision a is unity. This coefficient has been found to be unity,a, le for metals. These results establish the general validity of the Langmuir free evaporation method as applied to metals. EXPERIMENTAL A) Vacuum Microbalance Method. A modification of the Langmuir free evaporation method was used. Strip specimens were suspended from a sensitive microbalance operating inside of a high vacuum system. For alloys, microbalance methods17 must be used on large area samples at relatively low temperatures to minimize composition changes. The

Jan 1, 1962

By B. L. Averbach, E. E. Underwood

The vapor pressure of zinc in equilibrium with solid ß AgZn alloys has been measured by the dew-point method. The thermodynamic activities of silver, and the molar heats of formation have been calculated. Similar measurements have been made for the ternary alloys formed by adding gold to AgZn. An apparent particle size effect has been observed for AgZn filings. The vapor pressure of zinc over filings is 15 to 20 pct higher than over the solid alloys. STATISTICAL methods have been used to inter pret the thermodynamic activities of the components of a solid solution in terms of average equilibrium atomic configurations. In some cases, the atomic arrangements can be measured by X-ray methods, and the results are usually expressed in terms of short range order parameters. Only in the cases of Cu³Au and AgAu have there been sufficient X-ray and thermodynamic data to verify the assumed correlation between the X-ray and the thermodynamic measurements. In Cu³Au the agreement is good; negative deviations from Raoult's law are observed and the number of Cu-Au nearest neighbor pairs predicted on the basis of the thermodynamic treatments agrees quite well with the number of such pairs observed by the X-ray methods. The ß brass alloys present a different set of conditions. The ordering is so rapid in these alloys that the disordered state cannot be retained on quenching. In contrast, the disordered state in CuAu is easily retained on quenching. On heating, the ordered ß brass loses its long range order continuously as the critical temperature is approached; in Cuau the order decreases slightly on heating and then drops discontinuously to zero at the critical temperature. These differences indicate that the short range order configurations in ß brass systems may be of a different character than in Cudu systems. However, X-ray measurements of local order cannot be made in the Cu-Zn alloys because of the similarity in scattering factors. The alloy AgZn is isomorphous with CuZn, however, and recent experiments have indicated that there is a good possibility of observing short range order in this system. Another complication arises in the case of ß AgZn. The disordered body-centered cubic structure exists above 280°C. On quenching from above this temperature an ordered structure of the cesium-chloride type is formed. On slow cooling, however, a complex hexagonal structure, the 3 phase, is produced; and if an ordered quenched alloy is reheated the phase begins to form above 80°C. It is thus difficult to study local order by X-ray methods in single crystals of AgZn since the C phase formed on heating does not necessarily return to a single crystal of the original orientation. This problem has been met, however, by adding a few percent of gold to the AgZn. The gold suppresses the formation of the c

Jan 1, 1952

By Ralph Hultgren, Prodyot Roy

Manganese vapor pressures from 1250° to 1500°K were measured by conventional Knudsen and torsion-effusion methods in twelve Fe-Mn alloys with compositions from 9 to 80 at. pct Mn. The Knudsen re-sults agveed approximately with previous measurements found in the literature which indicated the solutions were nearly ideal. However, except for the higher manganese compositions, the initial torsion readings indicated much (up to 50 pct) higher vapor presszcres than the Knudsen method. These high readings decreased steadily with time. The results are interpreted as due to depletion of surface concentration of manganese during evaporation. Thus the initial torsion readings are the most nearly correct and Knudsen methods on alloys are to be regarded with suspicion unless it can be demonstrated that diffusion rates are rapid enough to replenish surface concentrations depleted by evaporation. For the lowest manganese contents and highest temperatures, depletion causes initial torsion readings to be too low. FEW thermodynamic data are available for alloys of the high-melting transition metals due primarily to experimental difficulties at the high temperatures involved. Measurement of vapor pressure is one of the most promising techniques to be applied to this problem. From the vapor pressure of a component of an alloy phase, its activity or its partial molar Gibbs energy can be directly calculated. From measurements over a sufficient range of temperatures and compositions the partial and integral Gibbs energies, enthalpies, and entropies may be determined through Gibbs-Duhem integration. For solid phases of variable composition, which commonly are found in alloys, measured vapor pressures are low because of depletion of the surface concentration of the more volatile component as it is selectively vaporized. Diffusion from the interior tends to restore the depleted concentration. The seriousness of the effect depends on the relative rates of diffusion and vaporization. Although depletion has been recognized as a factor,' direct measurements of its rate are lacking and quantitative estimates of its effect on measurements are uncertain. In the present work the vapor pressures above a series of Fe-Mn alloys have been measured by the torsion-effusion method. This method permits a measurement of the vapor pressure of manganese as soon as the sample comes to temperature and continuously throughout the run. The rate of decline of the apparent vapor pressure measures the rate of depletion and the torsion reading at zero time should be correct. At high manganese concentrations (70 and 80 at. pct Mn), the torsion reading remained nearly constant; surface depletion was negligible. There was a slow drop in measured pressure due to bulk loss of manganese from the sample. However, for lower manganese contents, especially at high temperatures, the depletion effect was considerable and the apparent vapor pressure decreased steadily with time. For these alloys the Knudsen method should give pressures which are too low. The measurement of pressure in the Knudsen method cannot be made until enough manganese has been vaporized to be weighable. The Knudsen result is therefore an integrated average between the initial and final vapor pressures. To verify this, conventional Knudsen measurements were also made with results in approximate agreement with previously published Knudsen measurements for these alloys. As expected, for the lower manganese contents the initial torsion readings of pressure were as much as 50 pct higher than the Knudsen. THE APPARATUS The apparatus, shown in Fig. 1, is capable of operating to temperatures up to 2000°K. A vacuum of 5 x 10"6 mm Hg can be maintained. Temperatures were measured to +3° by a W-Re thermocouple placed in the dummy cell, N. The torsion cell consisted of two alumina crucibles with holes in their covers. They were placed in the molybdenum holder shown suspended from the torsion wire. Rectangular torsion wires (1 by 4 mm) were found to have superior sensitivity and less residual distortion than circular (2-mm-diameter) wires. The precision of angular measurement was found to be *0.025 deg; the variation is no doubt due to temperature fluctuation and vibration. For the Knudsen experiments the alumina cell was set on the support, N. SAMPLE PREPARATION 800-g samples were prepared by melting together electrolytic iron and electrolytic manganese and pouring the melt into a water-cooled copper mold. Melting and pouring were done under a helium at-

Jan 1, 1965

By P. K. Koh, H. A. Lewis, H. F. Graff

New data on the distribution of silicon between slag and carbon-saturated iron at 1600Oand 1700OC are presented which, in combination with previously published data, permit the determination of silica activities over a broad range of compositions in the CaO-Al2O3-SiO2 system. The distribution of silicon between graphite-saturated Fe-Si-C alloys and blast furnace-type slags in equilibrium with CO has been described in previous publications.1"3 In this past work the silica-silicon relation was established at temperatures of 1425" to 1'700°C for slags containing up to 20 pct A12O3. This paper presents the results of additional studies at 1600" and 1700° C which extend the silicon distribution data at these temperatures for CaO-A12O3-SiO, slags over a range from zero pct Al2O3 to saturation with Al2O3, or CaO.2Al2O3. The upper limit of SiO2 is set by the occurrence of Sic as a stable phase when the metal contains 23.0 or 23.7 pct Si at 1600" or 1700°C, respectively. The activity of silica over the expanded range is determined directly from the distribution data.3 Recently4-7 other investigators have studied the activities of SiO, and CaO, principally in the binary system, using different methods and obtaining somewhat different results. EXPERIMENTAL STUDY The experimental apparatus and procedure have been fully described in previous publications.1, 3 Six new series of experimental heats have been made, four at 1600° and two at 1700°C. Master slags of several fixed CaO/Al203 ratios were pre-melted in graphite crucibles, and these were used with additions of silica to prepare the initial slag for each experiment. Slag and metal were stirred at 100 rpm and CO was passed through the furnace at 150 cc per min. The initial sample was taken 1 hr after addition of slag at 1600°C or 1/2 hr after addition at 1700°C. The run was normally continued for 8 hr at 1600°C or 7 hr at 1700°C, and the final sample was taken at the end of this period. Changes in Si and SiO2 content indicate the direction of approach to equilibrium, and in a series of runs where the approach is from both sides this permits approximate location of the equilibrium line. Fig. 1 shows the results of such a series of 15 runs at 1600°C for slags of CaO/Al,O3 = 1.50 by weight. Figs. 2 and 3 record other series at 1600°C and Fig. 5 a series at 1700°C with fixed CaO/Al0 ratios. The results of the experiments at 162003°C have been reported in part in a preliminary note.3 In the experiments recorded in Figs. 4 and 6, the slags were saturated with A12O3 (or with CaO.2A12O3 within its field of stability) by suspending a pure alumina tube in the melt during the course of the run. The final slag analyses were used to establish the liquidus boundaries8 in the stability fields of CaO.2Al2O3 and of Al20,. ACTIVITY OF SILICA The free-energy change in the reaction has been calculated by Fulton and chipman2 from recent and trustworthy data including heats of formation, entropies, and heat capacities. The more recent determination by Olette of the high-temperature enthalpy of liquid silicon is in satisfactory agreement with the values used and therefore requires no revision of the result which is expressed in the equation: SiO2 (crist) + 2C (graph) = Si + 2CO(g.) [1] &F° = + 161,500 - 87.4T The standard state for silica is taken as pure cristobalite and that of Si as the pure liquid metal. Since the melts were made under 1 atm of CO and were graphite-saturated, the equilibrium constant for Eq. [I] reduces to K1 = asi /asio2. The value of this constant is 1.77 at 1600°C and 16.2 at 1700°C. Through K1, the activity of silica in the slag is directly related to the activity of silicon in the equilibrium metal.

Jan 1, 1960

By C. H. Li, J. Washburn, E. R. Parker

Yield stress in single crystals of zinc was shown to be dependent on prior annealing temperature and rate of cooling after annealing. Rate of strain hardening beyond the yield was not sensitive to annealing procedure. A tentative mechanism for the effect was discussed. COMPARATIVE mechanical tests of metal single crystals generally show a large scatter of results. Many factors may contribute to this variation. The present investigation was undertaken to evaluate the influence of two factors not ordinarily considered of major importance—annealing temperature and cooling rate after annealing. Since plastic properties of crystals are thought to depend on the presence of lattice imperfections, it follows that a variation in properties should result from differences in the distribution and density of these imperfections. Furthermore, in the light of current theories of interaction between edge dislocations and lattice vacancies,' it seems possible that annealing temperature and subsequent rate of cooling might play a part in this distribution. Analysis of the effect of these variables on plastic properties may lead to a better understanding of the annealed state. Single crystals of zinc which have been strained in simple shear undergo complete recovery of mechanical properties upon subsequent annealing. This behavior is useful because a single specimen can be used repeatedly, thus eliminating the uncontrollable variables introduced by the use of many crystals. The test section of the crystals used for these experiments was a cylinder 1/8 in. thick with a cross-sectional area of about 1/3 sq in. The specimens were acid-machined from spherical single crystals 1 in. in diameter in such a way that the slip plane (0001) was at right angles to the axis of the cylinder. In testing, a shear stress was applied along one of the slip directions, [2110] in the (0001) plane.' All crystals, unless otherwise noted, were grown in a helium atmosphere from molten 99.99 pct pure Horse Head Special Zinc. Fig. 1 is a family of curves obtained from one crystal. Each curve was obtained by deforming the crystal to 10 pct strain at a rate of 3 pct per min, following an anneal of 1 hr at 260°C. The strain direction was reversed every other time so that no progressive change in shape of the specimen took place during the series of tests. Since the stress-strain curve was reproducible within a fairly narrow range under the same conditions of annealing, it was then possible to determine the effect of changes in annealing conditions. A crystal was tested repeatedly as previously described but the annealing temperature and rate of cooling preceding each test were varied (Fig. 2). Two cooling rates were used, approximately 30°C per min and 3°C per min. Annealing was conducted at three temperatures, 200°, 300°, and 400°C. For a given annealing time and rate of cooling the room temperature flow stress was higher the higher the annealing temperature. The increase in flow stress was small as the annealing temperature was increased from 200° to 300°C but rose rapidly when the annealing temperature was raised from 300" to 400°C. Faster cooling from the high temperature produced an even greater strengthening of the crystal. The curves were obtained in the order indicated by the numbers. The stress level associated with a certain annealing treatment was, within the limits of scatter, always the same regardless of the

Jan 1, 1954

By Gilbert Santoro

In this study tantalum carbide filaments of various compositions in the fcc region were prepared by heating a tantalum wire in a measured amount of hydrocarbon vapor. Such properties as tensile strength, transverse-rupture strength, microhard-ness, lattice parameter, electrical resistivity, and magnetic susceptibility were measured as a function of carbon content. Some of these properties showed a maximum or a minimum between a carbon-to-tantalum molar ratio of 0.80 and 0.85, the composition range at which the color changes from metallic gray to gold. The recent literature has shown that in the large homogeneity range for the fcc phase of the Ta-C system there is a marked variation for such properties as microhardness,' electrical resistivity, ' thermodynamic properties,z'3 lattice parameter and superconductivity. In addition to the technological interest of these variations within a single phase, important theoretical benefits may ensue from a continuous investigation of properties vs stoichiometry in tantalum carbide and other refractory defect compounds, for example the effect of the metal-metal distance and crystal structure upon bond formation and electronic structure. Whether there is a similar variation of the tensile strength is of practical importance. Of theoretical interest would be the knowledge of how the magnetic susceptibility varies with carbon composition. Both of these properties, as well as the transverse-rupture strength, were determined for tantalum carbide filaments in the fcc phase region. The lattice parameter, the electrical resistivity, and the microhardness were also investigated and compared with values appearing in the literature. EXPERIMENTAL PROCEDURE Material Preparation. Tantalum carbide specimens of various carbon content were prepared by statically carburizing high-purity tantalum filaments in measured quantities of pure hydrocarbon gas. The filaments, which were 10 mil in diameter and 20 or 40 cm long, were heated by their own resistance at temperatures from 1800" to 2200°C using alternating current in some cases and direct current in others. Tantalum wire from two sources was used. One had a reported purity of 99.96 wt pct Ta. The ingot from which this wire was drawn had been electron beam-melted as the final step in its purification. The reported purity of the other wire, which was prepared by a powder metallurgical process, was 99.88 wt pct Ta. The impurities of both tantalum sources are listed in Table I. Spec-trographic-grade toluene or research-grade propane was used as the source of carbon. The apparatus for the carburization of the filaments is shown schematically in Fig. 1. The system was evacuated to 1 x l0-' mm of Hg when using the electron beam-melted tantalum and to less than 1 p when using the less-pure tantalum. Neither the difference in the base vacuum, nor in the hydrocarbon gas, nor in the tantalum filaments resulted in any detectable difference in the properties studied in this work. However, only the carbides prepared from the higher-purity tantalum were used in the magnetic-susceptibility determinations. Once the base vacuum had been reached, the system was flushed with the hydrocarbon and re-evacuated. For any desired composition of the carbide, hydrocarbon gas was allowed to leak into a constant-volume flask to a predetermined pressure. These pressures were measured by a Cartesian diver-type mercury gage with a range from 0 to 20 mm and with a scale expanded to nine times. The hydrocarbon gas was allowed into the reaction tube and the filament was quickly raised to the reaction temperature. The slack in the filament due to thermal expansion was taken up by drawing out the movable electrode with a screw mechanism. The filaments were kept at temperature from a few minutes to 200 hr. In the majority of cases the heating time was 6 hr. The reaction tube was cooled by cascading water on its outer surface. By means of an ammeter and a voltmeter the electrical resistance at the reaction

Jan 1, 1963

By P. G. Shewmon, W. M. Robertson

The derivative of the surface tension with orientation, ??/??, for copper has been measured over the entire unit triangle. This derivative or torque term was determined from the variation of the dihedral angles at the intersection of twin boundaries and free surfaces. High-temperature anneals in graphite with a hydrogen atmosphere gave torque terms of 0.097 at (111), 0.066 at (100), and 0.0257 at (110). A variation from -36oto +23oC in the dew Point of the H2 had no effect on these values, nor did a change from 99.98 to 99.999 pct Pure copper. The removal of graphite from the system decreased ??/?? slightly. Considering the surface near a low-index plane to be smooth except for occasional steps, the free-energy increment per cm of step is calculated, and it is shown that the observed variation of ??/?? is consistent with a random distribution of steps and an interaction energy between steps. Integration of ??/?? gave the relative values of y at (100), (110), and (111) as comPared to ymax: (Y111/Ymax) =0.974, (Y110/Ymax) =0.996, (y100/Ymax) = 0.983. The ratio of the surface tension of coherent twin boundaries to the surface tension of the free surface was found to be (yT/ys) = 0.027 * 0.011. In a crystalline solid the surface tension y (surface free energy per unit area) depends on the orientation of the surface relative to the crystallographic axes of the crystal. Measurements of this variation of y with surface orientation have been reported for metals only within the last few years. Consider the grain boundary groove shown in Fig. 1. Here the three surfaces are all normal to the plane of the paper, and each has a different surface tension, y1, y2, and y3. If the system is at equilibrium, the total surface free energy must remain unchanged for a slight vertical displacement of the intersection (points A, B, and C are kept fixed). As shown by Herring,' the following equation is obtained:

Jan 1, 1962

By C. A. Bruch, W. E. McHugh, R. W. Hockenbury

EXPERIMENTAL results of the last decade have shown that both accelerator particles and reactor radiations produce significant changes in the properties of metals. These changes, called radiation damage, are important from both academic and engineering viewpoints. An understanding of the radiation damage produced by reactor radiations is difficult, because many kinds of particles having varying energies are present. Therefore, many fundamental studies are being conducted using accelerators in order to obtain nearly mono-energetic beams of various charged particles. Furthermore, many of these experiments are conducted at low temperatures in order to freeze in the radiation-produced lattice defects, after which the specimens are annealed to study the movement of the defects. It is found that although much of the low temperature radiation damage disappears by annealing at room temperature, a significant amount remains. Because there is no assurance that irradiation followed by annealing at a higher temperature will produce the same effect as irradiation at the higher temperature, and because there is an immediate interest in reactor radiation effects during exposures at and above room temperature, a major program is in progress at this laboratory to investigate some of these effects. This paper, representing a part of the program, compares the radiation damage to a group of metals which were exposed under nearly identical irradiation conditions. The metals, representing the three major crystal systems, include copper, nickel, titanium, zirconium, iron, molybdenum, and type 347 stainless steel. The effects of reactor radiations have been treated from a theoretical viewpoint by Seitz' and Ozeroff.' The problem considered by them has been the atomic displacements produced in an infinite medium by mono-energetic beams of both neutrons and charged particles. For reactor experiments, this treatment is not wide enough in scope, because specimens and reactor components are finite, and there are continuous energy spectra of the various particles. Therefore, the theoretical considerations will be extended in this paper to cover reactor conditions and the findings will be compared with the observed radiation damage. Experimental Materials—The six basic metals for this investigation were of commercial or higher purity, and the type 347 stainless steel was of commercial quality. These materials were supplied in various forms ranging from ingots to finished rods, and were fabricated to rods approximately % in. in diam and annealed for specimen stock. A summary of data pertinent to these materials is shown in Table I. The values for chemical compositions include both actual analytical values and typical values supplied by the manufacturer. More information concerning typical

Jan 1, 1957

By A. Phillips, H. J. Fisher

The influence of temperature and composition on the viscosity of high-purity liquid metals and alloys of the Pb-Sn and Sb-Cd systems was investigated by the logarithmic-decrement method. The variation of viscosity with temperature above the liquidus followed the equation: n = AeE/RT No large increase in viscosity was found just before normal freezing of any of the metals or alloys investigated. Certain alloys, supercooled below their stable liquidus, however, did show a marked increase in viscosity. Experimental activation energies for viscosity were determined for the metals investigated. Liquid-metal density values were obtained by a modified Joeger's method. IK an investigation of the properties of molten metals, the study of viscosity as a suitable means for studying the physical nature of the metallic liquid state has been utilized. Although interest in the relationship of the liquid to the solid-metal state holds promise of broadening the knowledge of metals in general, much information that is available regarding the influence of temperature and composition on viscosity, especially near the freezing point, is still controversial. The purpose of this paper is to present the results of recent work on this subject. Although interest in factors influencing the fluid behavior of metals and alloys has arisen from theoretical as well as applied studies of liquid-metal properties, the development of a comprehensive theory of the mechanism of viscosity has been handicapped because so little is known about the liquid state. Several theories of viscosity have been proposed and among the more recent are those of Andrade,1 Bernal,2 Frenkel,3 Eyring,4 and Gutmann and Simmons.5 Athough these contributions have helped to establish a basis of approach for studying the nature of viscosity, the fundamental aspects of the variation of viscosity with alloy composition have, in general, been neglected. Experimental Metals and Alloys Investigated: The Pb-Sn and Sb-Cd alloy systems were selected for investigation because they form constitutional features of theo-

Jan 1, 1955

By F. H. Buttner, E. R. Funk, H. Udin

Gold wires, 5 mil in diam, are found to creep viscously up to approximately 5.5x106 dynes per sq cm around 1300°K. Beyond this point, an additional slip mechanism appears. The average coefficient in the viscous range at these temperatures is 18.9 ± 7x10 12 poises. This observation and other visual observations made during the experiments agree with Herring's predictions. EARLY creep studies by Chalmers on single crystals of tin produced some of the first experimental evidence to suggest that metals deform viscously under very light stresses. That is to say, over a small initial stress range up to 150 g per sq mm, the single crystals of tin elongated at a rate proportional to the stress. The proportionality constant between strain rate and stress is by definition the coefficient of viscosity. Chalmers named this range of proportionality the microcreep range. One of the first theoretical attempts to explain the microcreep phenomenon was given by Kauzman,' who applied Eyring'sQ eneral statistical mechanical theory of shear rates to creep data then available. From the rate theory, he deduced the general hyperbolic sine law for shear rate on the assumption that the distortion of the reaction path due to stress is negligible and that the system is an infinite continuum. In order to fit the available data to the equation, it was necessary to assume that creep occurs by shear of rather large blocks of material, rather than by an atomistic shear process as in diffusion. The size of the shear units was shown to increase with increasing temperatures and lower stress, or lower strain rates. In applying the hyperbolic sine law equation to the microcreep data of tin, it was shown by Kauzman that the unit shear blocks became quite small, which suggested to him that microcreep might be a self-diffusion process. Making this an assumption, Kauz- man reworked the rate theory and arrived at an equation where the shear rate is proportional to the stress. In the form of viscosity coefficient, the equation is: o- _ \LkT V = 2qAlA where u is the stress; S, the shear rate; X, the distance in shear direction moved by flow units relative to one another in the unit process; L, the distance between layers of units of flow; T, the absolute temperature; k, Boltzmann's constant; qAL, a constant proportional to the size of flow units; and kT A, the diffusion coefficient = ------kTe-Q/RT = K X2 h Kauzman applied the expression to molten lead with a fair degree of success, the calculated viscosity being low by a factor of four. For solid lead the result is 5000 times too high. Taking q = 1, and X = dA = L = 21 = nearest neighbor distance, Udin, Shaler and Wulff8 found the equation low by a factor of lo7 for solid copper. On the basis of the discrepancy in lead, Kauzman indicated that microcreep is probably not a simple diffusion process of molecular shear units, but one involving preferential diffusion paths, or even that diffusion in solids does not involve shear at all. Cottrell and Jaswon4 treat the microcreep phenomenon on the basis of slow moving dislocations and apply it to Chalmers' data on tin with fair success. Nabarro Suggested a microcreep model in which mass movement in the direction of pull is effected by the diffusion of holes. The tension on the crystal effectively creates a greater hole concentration at regions near the crystal surfaces perpendicular to

Jan 1, 1953

By F. H. Buttner, E. R. Funk, H. Udin

A. P. Greenough (University College, swansea, Great Britain)—I have recently made some experiments on the deformation of silver wire at high temperature in an atmosphere of oxygen-free nitrogen. The observa-tions which 1 made confirm the main conclusion of this paper, that the most satisfactory mechanism so far suggested to explain the deformation which takes place

Jan 1, 1953

By A. L. Pranatis, G. M. Pound

Changes in length of copper foils of varying thickness and grain size were measured under such conditions of low stress and high temperature that it is believed that creep was predominately the result of interboundary diffusion of the type recently discussed by Conyers Herring. The surface tension of copper was calculated and results confirmed previous work within the limits of experimental error. Under the assumption of viscous flow, viscosities were calculated as a function of temperature and grain size. Predictions of the Nabarro Herring theory of surface grain boundary flow were borne out fully and the Herring theory of diffusional viscosity is strongly supported. ONLY a relatively few techniques for obtaining the surface tension of solids are presently available. Of these, the simplest and most straight forward is the direct measurement of surface tension by the application of a balancing counterforce. Thin wires or foils are lightly loaded and strain rates (either positive due to the downward force of the applied load or negative if the contracting tendency of surface tension is sufficiently greater than the applied stress) are observed. By plotting strain rates against stress, the load which exactly balances the upward pull is found and a simple calculation yields a value for the surface tension. The technique is of comparative antiquity, and solid surface tension values were reported by Chapman and Porter,' Schottky; and Berggren" in the early part of the century. Later, the filament technique became fairly well established as a method for determining the surface tension of viscous liquids, and Tammann and coworkers,'. " Sawai and co-worker and Mackh howed good agreement between the values of surface tension for glasses and tars obtained by the filament technique and by more conventional methods. With the increased confidence in the technique gained in these experiments, the method was applied to solid metals and the first reliable values of surface tension of solid metals were reported by Sawai and coworkers10' " and by Tammann and Boehme." More recently, Udin and coworkersu-'" have reported the results of experiments with gold, silver, and copper wires. Similar experiments with gold wires were carried out by Alexander, Dawson, and Kling.'" The excellent review articles of Fisher and Dunn" and of Udinl@ should be referred to for detailed criticism of the foregoing work and for discussion of underlying theory. In all the foregoing calculations, it is assumed implicitly that the material contracts or extends uni- formly along the length of the specimen and also that it flows in a viscous fashion, i.e., that strain rates are proportional to stress. For an amorphous material, such as glass, tar, or pitch, the assumptions are quite valid and good agreement is obtained with values of surface tension measured by other techniques. The values reported for metals, however, are occasionally regarded with misgiving, since it can be argued that, because of their crystalline nature, true solids can not deform in a viscous fashion. If this is true, then the results reported for solid metals over a long period of years are of only doubtful value. Thus it is clearly necessary that a mechanism be established that would explain both the viscous flow and the uniform deformation that has been assumed. Such a mechanism has been proposed by Herring."' Briefly, he suggests that, under the conditions of the experiment, deformation takes place by means of a flow of vacancies between grain boundaries and surfaces. This is a direct but independent extension of the theory proposed by Nabarro" in an attempt to explain the microcreep observed by Chalmer~.In a condensed form the Herring viscosity equation is TRL there 7 is the viscosity, T the absolute temperature, R and L grain dimensions, and D the self-diffusion coefficient. In its complete form, all constants are calculable and it includes such factors as grain shape, specimen shape, and degree of grain boundary flow. When applied to existing data, good agreement was obtained between predicted and observed flow rates. The theory received provisional confirmation from the work of Buttner, Funk, and Udin" who observed viscosities in 5 mil Au wire much higher than those in the 1 mil wire used by Alexander, Dawson, and Kling.'" More significant were the completely negligible strain rates found by Greenough" in silver single crystals. Opposed to these observations were those of Udin, Shaler, and Wulff'" who found indications of viscosity decreasing as grain size increased. Thus, complete confirmation of the theory was lacking in that the data to which it could be applied contained only a limited number of grain sizes. Hence, it was proposed that a series of experiments be carried out with thin foils of varying grain size up to and including single crystals, where, according to the Herring theory, deformation would occur only at almost infinitely slow rates.

Jan 1, 1956

By R. D. Fisher, A. G. Metcalfe, F. A. Crossley

AN investigation of the application of magnetic stirring to the consumable arc melting of aluminum and nickel demonstrated that grain refinement could be obtained when there was sufficient stirring force. Also, it was found that grain refinement could be obtained in aluminum by alternately reversing the polarity of the current through the magnetic stirring coil. The effect of the latter was to periodically reverse the direction of rotation of the molten metal.' Because of their promise, these findings were applied to conventionally cast aluminum. Shortly after this work was begun, Kondic2 reported experiments in which aluminum cast with 40° C (72° F) of superheat exhibited a grain size of 10 to 20 grains per cm2. When interrupted rotation was applied during solidification by mechanical means, a grain size of 800 to 1000 grains per cm2 was obtained. Magnetic Stirring Applied to Conventional Casting— The applicability of magnetic stirring to producing grain refinement in conventionally cast metal was evaluated in the following way. Three pounds of aluminum were cast into a clay-bonded graphite mold. The mold was surrounded by a magnetic stirring coil of 1000 turns. A direct current of 200 amp was passed through the casting by means of a graphite electrode which was immersed from the top. A current of 5 amp was passed through the stirring coil. The aluminum (1100 or 25) was superheated to 1600° F (871° C) and fluxed by passing chlorine gas through the melt for 2 min prior to casting. Two additional castings were made for comparison: 1) a conventional casting of metal superheated to 1600° F and 2) a similar casting through which 200-amp current was passed during solidification, but without current in the magnetic coil. Macrosections of these three ingots are shown in Fig. 1. Very remarkable grain refinement resulted in the magnetically stirred casting. It may be noted that the casting through which 200-amp current was passed has a coarser structure than the conventional casting. It would appear that this is due to the additional heat input generated by the current to the former casting. Mechanical Rotation Applied to Conventional Casting—A casting temperature of 1600° F was used throughout. Some 1 1/2 -lb castings rotated at 150 rpm were predominantly columnar with small areas of coarse, equiaxed grains close to the lower mold wall. Following these, a series of 3-lb castings was made. Macrosections of these ingots are shown in Fig. 2. The imposed rotational speed was 250 rpm. Turbulence or acceleration in the solidifying liquid was the significant factor. Fine grain size resulted in the material solidified during the period of acceleration of the liquid to the imposed rotational speed. When uniform rotational motion was achieved, the grain structure reverted to columnar. The final casting shown in the lower right-

Jan 1, 1962

By Harvey Cline, Donald P. Ferriss

THE violent outgassing of commercial tungsten and other refractory metals when melted in an electron beam zone refining apparatus1"3 is dealt with experimentally by one or both of two approaches. One is to provide a suitable emission current control1,2 the other is to purify the metal in the solid state before melting.4 The first of these is usually sufficient, but the authors chose also to investigate the source of the troublesome gases and to devise means of solid-state purification. Some further insights into the use of tungsten as a structural material at temperatures near its melting point were also revealed in the course of the work. One may speculate on the causes of gas formation in tungsten at high temperature by inferences from the literature and an examination of the impurities in commercial wire. Davenport5 observed microporosity in both tungsten and molybdenum filaments heated to several hundred degrees below their melting points. Studying the function of oxygen in molybdenum welds, Perry, Spacil, and wulff8 attributed similar porosity to the disproportionation of Moo2 to molybdenum metal and Moo3. Tungsten dioxide also disproportionates7 above 2000oC to yield WO3 vapor at considerable pressure. Commercial tungsten contains some 5 to 20 wt ppm O and a similar concentration of carbon depending on the details and precautions of processing. Therefore, evolution of CO is also a strong possibility as a source of gas at high temperatures. Furthermore, even for activities in the order of 0.01, it may be calculated that the pressure of CO at 3000°C would be over 4000 psi, which is in excess of the tensile strength of tungsten at this temperature.9,10 Powder metallurgy tungsten contains some 0.02 pet Fe and 0.002 to 0.005 pet of the elements Ni, Si, Al, Ca, Mn, Mg, and Mo; at temperatures over 2800°c, therefore, these elements will exhibit appreciable vapor pressure, though small compared to that of WO3 and CO. Some of these impurities may react to form high-pressure monoxides such as SiO and A10. Finally, dissolved and occluded nitrogen is an ever-present source of high-pressure gas. Commercial Type 218, General Electric filament wires, 0.025 in. diam, were heated by resistance in 10 -4 mm of Hg and in argon or hydrogen at pressures from 50 µ to atmospheric. For temperature determinations in vacuo, the Jones - Langmuir8 current-temperature relationship was employed. For gaseous atmospheres, the burn-out temperature was assumed to be the same as in vacuo so that a small, constant correction could be made for temperatures below burn-out conditions (i.e., local heating to failure). Several dozen experiments may be summarized by pointing out some structural observations. At 3050oC in 1 atm of A, porosity is very evident after 30 min, as shown in Fig. 1. For the section in Fig. 2, failure occurred 200°C below the melting point and many voids have grown in the grain bound-

Jan 1, 1962