Search Documents

By S. T. Wlodek

The surface and subscale oxidation reactions were followed by means of continuous weight-gain and metallographic techniques over the range 1600" to 2200°F (871° to 1204 °C) for up to 400 hr. Full identification of all scale and subscale reaction products was obtained by electron and X-ray diffraction. At or below 1800°F (982°C) a linear rate of reaction (QL = 46.0 kcal per mole) governed the oxidation process, extending for up to 100 hr at 1600°F (871 "C). During linear oxidation the surface scale consisted of an amorphous SiO2 film overgrown with Cr 2O 3 and NiCr204. This initial linear process was followed, and above 1800°F completely replaced, by two successive parabolic rate laws (Qp = 60 and 57 kcal per mole). This parabolic reaction involved the formation of a complex scale consisting of Cr2 O3 and smaller amounts of NiCr2O4. Parabolic oxidation appeared to coincide with the disruplion of the silica film present during linear oxidation and was followed by subscale (internal) oxidation of crystobalite and NiCr2O4. The balance between the subscale and surface oxidation reactions controls the oxidation of this commercial alloy. The amorphous silica film appears to result in the linear rate and diffusion through Cr2O3 is the more likely rate-limiting step during parabolic oxidation. THE oxidation of a multicomponent composition is a complex phenomenon not presently amenable to a rigorous classical interpretation. Nevertheless, even a qualitative understanding of the scaling and subscale reactions that occur in a commercial composition can illuminate the reactions that limit its high-temperature stability in an oxidizing environment. This study of the oxidation of Hastelloy Alloy X presents the first of a series of studies with the above approach in mind. Hastelloy X exhibits one of the best combinations of strength and oxidation resistance available in a wrought, solution-strengthened, nickel-base alloy. Although during long time exposure some precipitation of M6C and M23C8 carbides as well as a complex Laves phase occurs, the amounts are probably small enough to have no appreciable effect on the chemistry of the matrix. Radavich has identified the oxidation products on Hastelloy X oxidized for 5 min to 10 hr at 1115°F as NiO and the NiCr2O4 spinel. Oxidation for 5 to 15 min at 1500°F produced a scale of spinel, NiO, and a rhombohedra1 phase, probably Cr2Os. Sannier et 2. have reported continuous weight-gain data for Hastelloy X at 1650" and 2010°F and internal-oxidation measurements after 150 hr at 2010°F. In addition, much of the data on binary Ni-Cr alloys recently reviewed by Kubaschewski and okins' and Ignatov and Shamgunova4 as well as studies of binary Ni-Mo alloys5 are also pertinent to the oxidation of this composition. EXPERIMENTAL Continuous weight-gain measurements and metallographic measurements of subscale reactions were the main experimental techniques used in this study. X-ray and electron diffraction backed up by a limited amount of electron-microprobe analysis served to characterize the nature of the scale- and subscale-reaction products. Two heats of commercial sheet of the composition given in Table I and identified as A and B were used in the bulk of this study. Internal-oxidation measurements were made on a third heat of material in the form of a 0.5-in.-diam bar. In order to assure homogeneity, all heats were reannealed 4 hr at 2175°F prior to sample preparation. weight-Gain Measurement. All specimens (1.5 by 0.4 by 0.03 in.) were abraded through 600 paper, electropolished, and lightly etched in an alcohol-10 pct HCl solution. An electrolyte of 150 cu cm H,O, 500 cu cm HsPO4 (85 pct conc), and 3 g CrO3 at a current density of 0.9 amp per sq cm or a solution of 10 pct HaW4 in alcohol used at 4 v and 0.3 amp per sq cm was used for electropolishing. The resultant surface exhibited a finish of 3 ± 1 p rms. Continuous weight-gain tests were made at 1600°, 1700°, 1800°, 1900°, 2000", and 2200°F on auer' type balances capable of recording a total weight change of 110 mg with an accuracy of k0.1 mg. All tests were made in air dried to a dew point of -70°F and metered into the 2-in.-diam reaction

Jan 1, 1964

By A. Szirmae, Hsun Hu

The early stages of recrystallization in a 70 pct cold-rolled Si-Fe crystal of the (110) (0011) orientation were studied with a Siemens electron microscope. Orientation studies based on electron-diffractzotz. patterns confirm the results of previous texture analysis. The driving energy for recrystallizatior and the critical radius for growth were calculated from the dislocation energy and the energy of the subgrain bourzdaries, and it was found consistent with the observed size of the recrystallized grains. The recrystallization characteristics of crystals with different initial orientations are discussed. The recrystallization of cold-rolled (110)[001] crystals of Si-Fe has been widely studied by various investigators.1-4 Their results on both deformation and annealing textures are in good agreement. The rolling texture after 70 pct reduction consists mainly of two crystallographically equivalent (111) [112] type textures and a minor component of the (100) [011] type. The latter is derived from the deformation twins, or Neumann bands, which are formed during the early stages of deformation and later rotate to the (100) [011] orientation upon further rolling reduction. Between the two main (111) [112] type textures, there is some orientation spread, because of which very low intensity areas appear in the pole figure. If these very low intensity areas are considered to be a very weak component in the texture, then a (110) [ 001 ] orientation may be assigned to them. When this rolled crystal is annealed at a sufficiently high temperature for recrystallization, the texture returns to a simple (110) [001]. The purpose of the present investigation was primarily to seek a better understanding of the recrystallization process by using the electron transmission technique. The (110) [0011 type of crystal was selected because orientation data for it are well known from previous studies with conventional techniques. Direct observations on the recrystallization of such a crystal have also been made by using a hot-stage inside the electron microscope, and the results will be reported in another paper. MATERIAL AND METHOD A single-crystal strip of the (110) [001] orientation was prepared from a commercial grade 3 pct Si-Fe alloy by the strain-anneal technique.= The strip was approximately 0.014 in. thick, and was rolled 70 pct at room temperature to a thickness of 0.004 in. Specimens were cut from the rolled strip and were annealed in a purified hydrogen or argon atmosphere. They were then electrolytically polished in a chromic-acetic acid solution to very thin foils. Best results were found by polishing first between two narrowly spaced flat cathodes with the specimen edges coated with acid-resisting paint, followed by polishing between two pointed electrodes until a hole appeared in the center as described by Bollmann.6 It was found that a thin transparent film always formed along the thin edges of the polished specimen. This film was then removed by rinsing the specimen very briefly in a solution of alcohol with a few drops of HF or HCl. RESULTS AND DISCUSSION 1) The Deformed Crystal. From the electron-diffraction patterns taken at various areas of an as-rolled specimen, the texture components as deduced - from ordinary pole-figure analysis were confirmed. Over most of the areas where orientation was examined, a (111) pattern with a [112] direction parallel to the rolling direction was obtained. This corresponds to the main deformation texture of the (111) [112] type. In a few areas the diffraction pattern was (100) [Oil], corresponding to the minor-texture component derived from the Neumann bands. The (110) [001] orientation, which corresponds to the very weak intensity area in the pole figure, was found infrequently. A typical example of the deformed matrix having the (111) type main texture is shown in Fig. 1, where (a) is the microstructure and (b) is the diffraction pattern taken from that area. It was also frequently observed that in other areas more or less continuous rings of weaker intensity were superimposed on the simple (111) diffraction pattern, suggesting the presence of a wide range of additional orientations. Other evidence indicated that the recrystallization characteristics are different in these two different types of areas. The hot-stage observations which provide this evidence will be discussed in another paper. AS shown in Fig. l(a), numerous dislocation-free areas of very small size are embedded in the "clouds" of high-dislocation density. This indicates that the deformation of a single crystal, even after a rolling reduction of 70 pct, is far from uniform on a micro-

Jan 1, 1962

By P. Douglas, G. M. Leak, B. Mills

The sinusoidal surface relaxation technique has been used to measure the surface self-diffusion coefficient of spectroscopically pure nickel over a wide temperature range under a hydrogen atmosphere. A kink in the Arrhenius plot has been observed. In the temperature range T/T 0.98 to 0.80 (T in O K and T, is the melting temperature) the average self diffusion coefficient is given by Below the temperature T/T,- 0.80a decrease in the slope of the log Ds us 1/T plot is observed. This is associated with a diffusion process characterized by a lower activation energy (-20,000 cal mole'') and smaller preexponential term (-10- sq cm sec"). A series of experiments were carried out at T/Tm = 0.61 under a hydrogen atmosphere of higher oxygen partial pressure than for the rest of the experiments. It was found that Ds was significantly depressed due to oxygen adsorption. This evidence supports the opinion that the low temperature process (activation energy -20,000 cal mole-') is unlikely to be due to oxygen adsorption. An interesting feature of the present data is that the transition temperature (T/Tm - 0.80) is a function of orientation. For a small number of crystals of measured orientation the transition temperature was observed to be higher towards the low index (100) pole. Theories of surface diffusion are briefly reviewed and it is concluded that the present reszuts are best explained by invoking a surface roughening process. GJOSTEIN has recently analyzed available surface diffusion data for a wide range of metals. He suggested that two mechanisms were operative for fcc metals, an adatom process at high temperatures and a vacancy process at low temperatures. Results for nickel can be summarized as follows. At low temperatures (T/T, - 0.3 to 0.44) under ultra high vacuum conditions, Melmed2 measured an activation energy Q of 21 kcal mole-' using field electron emission microscopy. At higher temperatures (T/T - 0.7 to 0.9) under a vacuum of 10- ' torr, Maiya and lakel measured y as 39 kcal mole-' using the multiple scratch smoothing technique. The present work was undertaken to try to find out if two distinct processes could be observed. High temperature results give Q about 47 kcal mole-': there is evidence also for a low temperature value of about 20 kcal mole-'. These measurements were all made under a hydrogen atmosphere, in the temperature range 860" to 1412°C. Concurrent with the present study Bonze1 and jostein> have also observed a break in the Arrhenius plot for the (110) surface of nickel. These measure- ments under ultrahigh vacuum conditions using the laser diffraction technique are in excellent agreement with the work reported here under hydrogen annealing conditions. THEORY The available surface relaxation techniques include single and multiple scratch smoothing and grain boundary grooving. The processes have been compared in detail by Gjostein for conditions where surface diffusion dominates6 and Mills et al? where volume diffusion dominates. In summary the relevant points are as follows. Grain boundary grooving gives an average Ds for the two surfaces adjacent to the boundary and this can, to some extent, be simplified by using symmetrical bicrystals. This technique has been used to study the effect of environment on Ds for silver and copper.'-'' Scratch techniques yield Ds values for the small orientation range exposed by the scratches (-2 deg). The multiple scratch process is preferable because the profile rapidly becomes sinusoidal and can then be interpreted theoretically in a relatively simple way. Also corrections for mass transport processes other than surface diffusion can be introduced easily. Mullins" considered a sinusoidal profile described the wavelength of the profile. After time t the profile can be described by the equation The terms A, A', C, and B which account, respectively, for contributions due to evaporation-condensation, diffusion through the gas phase, volume diffusion through the lattice, and surface diffusion are defined as: where Ds = the surface self diffusion coefficient ys = the surface energy per unit area p = the equilibrium vapor pressure over a flat surface pa = the equilibrium vapor density over a flat surface DG= the diffusion coefficient of vapor molecules in the inert gas DM = the mass transfer diffusion coefficient which for a pure cubic metal is Dv/f where Dv is the radiotracer diffusion coefficient and f is the correlation factor H = the molecular volume V = the surface density of atoms, il2'3 M = mass of an evaporating molecule

Jan 1, 1970

By Forrest V. Williams

The dislocatiorl and stacking-fault structuve of epitaxial GaAs1-,PX lms been examined by chemical etching. The layers were groun in the (100) direction and etch Pils were developed on (111} planes which nad been lapped and polished on the epiLaxia1 layevs. Tile effecL of the jollolcing cariables on the quality of the epilaxial layers has been examined: doping leuel, grouth rate, and composition. High stacking-faullL densilies weve found in the GuAsi_xpx layers. These are not observed in heavily dolled epitaxial layers tzar in layers with low phosphorus compositions. The dislocatiorz density in GuAsi-x px was highest at the sub-stvate- epilaxia1 layer interface. Composilion changes introduced dislocations in the epitaxial layers. ManY semiconductor p-n junction lasers of Group TIT-Group V compounds and their alloys have been reported in the past several years. Laser action at visible wavelengths in GaAsl-x,Px was first reported by Holonyak and Bevacqua. GaAs, a direct transition semiconductor which lases, and Gap, an indirect transition semiconductor which does not lase, form a continuous series of solid solutions.2 Laser junctions can be fabricated in GaAsl-xPx crystals with phosphorus compositions up to about 40 mol pct. In addition to the production of coherent radiation in these crystals, the efficient recombination radiation of p-n junctions in this material has equally important potential in the development of low-power semiconductor lamps. To achieve a high conversion efficiency of electrical to optical energy in p-n junctions in this material, the relation of physical properties of the crystal to luminescence efficiency must be better understood. Although the electrical, optical, and device properties of GaAsl -xPx junction lasers are understandably of considerable interest, the work to date indicates that the more serious problems are the chemical and metallurgical difficulties encountered in the growth of this material.3 In addition to the problems of chemical purity, crystal imperfections, such as dislocations and stacking faults, can be expected to affect both the efficiency of the radiative recombination process and the perfection of the p-n junction.3 The last requirement, i.c., that of the perfection of the p-n junction, is a particularly troublesome one in the fabrication of laser diodes. To obtain good laser diodes, the p-n junction must be flat, which permits the radiation to be reflected from the resonant cavity boundaries. Junction planarity is extremely sensitive to the crystal perfection of the semiconductor material. Also, it is known that at high dislocation densities (-105 per sq cm) it has not been possible to build laser junctions in GaAsl-xPx . Few studies have been reported on the crystal defect structure of GaAsl-,P,. The first serious study seems to be that of Wolfe, Nuese, and Holonyak,3 who examined the dislocation structure of monocrystalline bulk (nonepitaxial) material grown by halogen vapor transport. In this paper are reported some observations on the dislocation and stacking-fault structure of GaAsl_,P, crystals grown by a vapor transport process on substrates of GaAs. EXPERIMENTAL Crystal Growth. The GaAsl-xPx crystals were grown in an open-tube flow system, using two sets of reagents. GaAs, Pr(red), and HC1 were employed in one method. The transport reaction is =950JC GaAs+HC1 = GaCl +1/4As4 +1/2H2 and the deposition reaction is 2GaAs1-xPx +GaCl3 Composition control is obtained by the flow rate of the HC1 and the vapor pressure of the P4, which is maintained in a separately controlled furnace. The second method has been described by Ruehr-wein4 and utilizes gallium, AsH3, PH3, and HC1. The same transport and deposition reactions as above are involved. Composition control is obtained solely by the flow rates of the three gases involved. All of the crystals were grown on chemically polished GaAs substrates oriented on the (100) plane. The thicknesses of the epitaxial layers were typically 100 to 300p. Revealing of Dislocations. Dislocations were re-vealed on both the( 111 ) and { l l l }b faces by chemical etching. The specimen to be examined was mounted at 54.7 deg, lapped on glass with 3-p alumina, polished on cloth with 3-p diamond paste, and, to remove work damage, chemically polished at room temperature for

Jan 1, 1968

By C. H. Li, R. J. Stokes

This paper compares the fracture behavior of tungsten rods in three conditions: recrystallized. recovered, and wrought. Notched specimens suhjected to a 50 in.-lb impact load showed ductile-brittle transitions at 700, 4.90°, and 440°C, respectinely. The recrystallized material had an equiaxed pain structure and jracbred by simple cleavage from a grain boundary source at all temperatures up to 700°C. The wrought and recovered material had an elongated fibrous structure and at low temperatures fractured by cleavage originating from the notch. As the transition temperature was approached cleavage was preceeded by more and more intergvanular splitting which deflected the crack front into planes parallel to the tensile axis. The enhanced toughness of wrought and recovered tungsten was attributed both to its inability to initiate cleavage because no pain boundaries were suitably oriented perpendicular to the tensile stress and to its inability to maintain cleavage because of intergranular splitting ahead of the crack. It has been appreciated for a long time in a qualitative manner that the room-temperature brittleness of fully recrystallized tungsten may be alleviated by working the material at relatively low temperatures.' More recently this difference in mechanical behavior between wrought and recrystallized tungsten has been examined quantitatively by measurement of the tensile properties as a function of temperature. In these experiments brittleness has been expressed in terms of ductility or reduction in cross-sectional area upon tensile fracture or in terms of the bend radius before fracture under bending.' This work has shown the existence of a fairly sharp transition from brittle to ductile behavior with an increase in temperature. The ductile-brittle transition temperature for recrystallized material is approximately 200°C higher than for wrought material. An increase in strain rate, small additions of impurity,' or an increase in grain size4 shift the respective transition temperatures to higher values, but the difference between them remains approximately the same at 200°C. A number of explanations for this embrittlement by recrystallization have been given. It has been blamed either on the concentration of impurity at the grain boundaries, the increase in grain size, or the change in texture which occurs upon recrystallization. The present paper examines the effect of different heat treatments on the notch-impact behavior of commercial powder-metallurgy tungsten rods. The change in the ductile-brittle transition temperature for this method of loading and the fracture mode has been related to the different mi-crostructures produced by heat treatment. EXPERIMENTAL PROCEDURE Commercial swaged powder-metallurgy tungsten rods 1-3/8 in. in length and 1/8 in. in diameter were machined to introduce a sharp V notch 0.030 in. deep. To change the microstructure from that of the as-received wrought material some of the specimens were subjected to an anneal in nitrogen either at 1300° or 1400°C for 8 hr or at 1600° or 2000°C for 1/2 hr. The notched rods were then placed in a miniature Charpy-type impact machine and struck at their midpoint (opposite the notch) with a hammer designed to deliver 50 in.-lbs of energy. The strain rate at the base of the notch was estimated to be approximately 100 sec-1 at the instant of impact. The specimens were heated in situ to the desired impact temperature. The microstructures produced by the various anneals were studied by both X-ray diffraction and metallographic techniques. Fig. 1 reproduces the microstructures observed metallographically following a 10-sec electroetch in a 10 pct KOH solution. Fig. l(a) shows the elongated fibrous grain structure of the as-received material. Following the anneal at 1300" or 1400°C the grain structure was still elongated as shown in Fig. l(b) but the etch pits delineated dense polygonized dislocation arrays within many of the grains. Occasionally a relatively dislocation-free recrystallized grain was found growing into the matrix. The anneals at 1600° and 2000°C resulted in complete recrystallization and some grain growth. The grains produced at 1600°C were still slightly elongated as shown in Fig. l(c) whereas the anneal at 2000°C produced equiaxed grains. The changes in grain size produced the expected changes in the X-ray back-reflection patterns; there was no indication either in the as-received material or the annealed material of any preferred orientation. RESULTS a) Impact Behavior. Fig. 2 reproduces the ductile-brittle transition curves measured in the manner described in the previous section. It can be seen that under these testing conditions the as-received

Jan 1, 1964

By J. Szekely, G. P. Martins

A formulation is given for describing the rate of expansion of spherical bubbles rising in liquids the freeboard of which is evacuated. The computer solution of the resultant differential equations has shown that, for low freeboard pressures (less than about 5 mm Hg), on approaching the free surface the bubbles expand much less rapidly than predictable from equilibrium considerations. In other words, in this region the pressure inside the bubbles will be significantly larger than the static pressure in the liquid corresponding to the position of the bubble. These theoretical predictions were confirmed by experimental work, using two-dimensional air bubbles rising in mercury. The important consequence of these findings is that the expansion of gas bubbles in vacuum degassing operations will be a great deal less than expected from hydrostatic considerations. This would lead to a significant reduction in the available interfacial area and may explain the apparent poor efficiency of many vacuum degassing units. VACUUM degassing as a treatment for liquid steel has gained widespread popularity in recent years; the number of known installations exceeds several hundred at the present time.' Although much information is available on both the thermodynamics of the system and the overall performance of various industrial units, much less is known about the fundamental aspects of the process kinetics.2-4 The basic physical situation common to virtually all vacuum degassing operations is the interaction between gas bubbles (swarms of bubbles) and the surrounding molten metal, held in a container, the freeborad of which is at a low absolute pressure. Once formed (or introduced from an external source) the bubbles will ascend, due to the buoyancy forces, and, during this ascent, a significant increase in their volume will occur. This progressive increase in the bubble volume is due to two factors: a) the continuous reduction in the static pressure acting on the bubble during its rise; and b) mass transfer into the bubble from the surrounding molten steel. In a recent paper Richardson and Bradshaw developed equations5 for describing mass transfer into gas bubbles from molten metals at reduced pressures. However, in deriving these expressions it was assumed that the pressure inside the bubble was identical to the static pressure in the adjacent liquid. In other words, the volume of the bubble was considered to be in equilibrium with the pressure of the fluid adjacent to it. This assumption, thus their analysis presented, was thought to be reasonably accurate for most of the bubble's ascent; however, it was unlikely to be valid in the immediate vicinity of the free surface, held at a low pressure. It was pointed out in the discussion6 that the region close to the surface may be of considerable importance as both the driving force and the interfacial area available for mass transfer are at their highest value here. The ' 'anomalous" behavior of gas bubbles when approaching a free surface at low pressures was recently confirmed in a preliminary investigation by Szekely and Martins. ?1 Here high-speed motion photography was used to study air bubbles rising in a column of n-tetradecane with a freeboard pressure of 0.1 mm Hg. It was found that significant distortion of the bubbles occurred on approaching the free surface; furthermore, the expansion observed was much less than what one could expect from hydrostatic considerations, i.e., factor a previously discussed. It follows from the foregoing that a detailed study of these phenomena would be justified both from fundamental considerations and because of their potential relevance to technology. The purpose of the paper is to present a more realistic formulation for the expansion of a gas bubble approaching a low-pressure region, together with a comparison of the theoretical prediction with experimental measurement. An inert bubble will be considered in the first instance; it is thought that the understanding of the fluid mechanics is an essential first step toward the realistic formulation of the mass transfer process. This latter problem will be the subject of a subsequent publication. FORMULATION The Physical Model. Consider a spherical bubble, of initial radius Ro, rising in a fluid, having a density pL. Initially let the bubble be at a distance H from the free surface, and at a pressure Pgo, as illustrated in Fig. 1. Pgo, the initial pressure in the bubble, is given by the following equation: pgo = Po = Ptp +pLgH [ la] where Po is the pressure in the liquid corresponding to the initial position of the bubble and Ptp is the pressure at the free surface. The fluid pressure at

Jan 1, 1970

By T. F. Kassner

The dissolution kinetics of type 304 stainless steel in the Bi-Sn eutectic alloy have been investigated under the well-defined hydrodynamic conditions produced by the rotating-disc sample geometry. In addition, the mutual solubilities of iron, chromium, nickel, and manganese from 304 stainless steel in the eutectic alloy were determined over the temperature range 450" to 985°C. The convective -diffusion model for mass transport from a rotating disc was used to interpret the experinlental dissolution data. The dissolution process was found to be liquid-diffusion-controlled under specific conditions of temperature and Reynolds number. Liquid penetration into the 304 stainless steel resulted in a reduction of the di,ffusion-controlled mass flux and thus precluded the calculation of the diffusion coeficients of the four components from 304 stainless steel in the Bi-Sn eutectic alloy. The convective-diffusion model for diffusional limitations of electrode reactions and mass transport at the tationssurface of a rotating disc set forth by Levich 1,2 has found wide applicability in the investigation of electrochemical and dissolution phenomena in aqueous systems. Riddiford 3 and Rosner have reviewed the model and also include numerous references on work of this nature. More recently the rotating-disc system has been applied to the investigation of hetereogeneous reactions in liquid-metal systems. Shurygin and Kryuk 5 have measured the dissolution rates of carbon discs in molten Fe-C, Fe-Si, Fe-P, and Fe-Ni alloys. Shurygin and shantarin6 also studied the dissolution kinetics of iron, molybdenum, chromium, and tungsten, and the carbides of chromium and tungsten in Fe-C solutions with a rotating-disc sample geometry. In these systems it was possible to distinguish between diffusion and reaction control mainly through experimental confirmation of the velocity dependence of the dissolution rate predicted by the model. However in the absence of dependable solubility data and the virtual lack of diffusion data in these systems, a quantitative check of the magnitude and the temperature dependence of the rate was not possible. In many instances, estimates of the activation energy for solute diffusion and the diffusion coefficient based upon the experimental dissolution data are not credible. A recent study by this author7 has resulted in a critical test of the model in a liquid-metal system. The solution rates of tantalum discs in liquid tin were measured over a wide range of temperature and velocity conditions. In addition, the solubility and diffusion coefficient of tantalum in liquid tin were determined as a function of temperature. The latter data were used with the model to predict both the magnitude and the temperature dependence of the dissolution flux. In that work it was also deemed necessary to reevaluate the solution to the convective diffusion equation to incorporate the effect of the lower range of Schmidt numbers encountered in liquid-metal systems. Good agreement between the model and the experimental dissolution data in the region of diffusion control was obtained in the Ta-Sn system. The Bi-Sn eutectic alloy is used as a seal between the reactor head and the reactor vessel in the Experimental Breeder Reactor-11. The alloy is fused periodically prior to fuel-handling operations. In that connection, it was necessary to investigate the compatibility of the liquid alloy with the type 304 stainless-steel containment material. The results of a rotating-disc study in this multicomponent system are presented. EXPERIMENTAL METHOD The 5.08-cm-diam discs were machined from 0.317-cm-thick plate. Chemical analysis information for the type 304 SS material is given in Table I. The discs were ground flat on metallographic paper and given a final polish on Linde B abrasive. A thin support rod was threaded into the disc and the region around the threads was fused under an inert gas. The support rod was fitted with a quartz protection tube and then was attached to a supporting shaft which passed through a rotary push-pull vacuum seal. The disc and supporting shafts were dynamically balanced prior to insertion into the furnace tube. The apparatus is shown schematically in Fig. 1. The 58 pct Bi-42 pct Sn eutectic alloy melts were prepared from 99.995 pct pure Bi and Sn by fusing the components in a 7-cm-ID Pyrex crucible. The system in which the melts were made was evacuated to a pressure of 1 x 10-6 Torr and back-filled with purified argon several times before melting the charge. The ingot was reweighed and placed in a slightly larger-diameter Vycor crucible used in the dissolution runs. A run was started by lowering the disc into the liquid

Jan 1, 1968

R. Schuhmann, Jr. (Purdue University)— Fulton and Chipman's results on rate of silica reduction from slags are analogous in many was to the results of Parlee, Seagle, and Schuhmann10 on rate of alumina reduction from alumina crucibles. Both investigations have given comparably low rates of reduction and slow approaches to equilibrium. Accordingly, we may hypothesize that the rate-determining step is the same in both kinds of experiments; that is, oxygen diffusion across the stagnant boundary layer on the liquid-metal side of the interface between the liquid metal and the oxide phase (slag or solid oxide). I suggest that silica reduction involves the following consecutive steps: I) At the slag-metal interface: SiO2(slag) Si+ 20 II) Transport of oxygen from slag-metal to gas-metal interface: a) diffusion across liquid-metal boundary layer at slag-metal interface. b) convection within the body of liquid metal. c) diffusion across boundary layer at metal-gas interface. 111) At the metal-gas interface: C +O- CO (gas) Iv) At the graphite-metal interface: C (graphite) -C At steelmaking temperatures it is reasonable to assume that equilibrium is attained in all three chemical reactions (I, 111, and IV) right at the respective interfaces. Convection within the stirred liquid metal (step IIb) is also rapid. Transport of Si and C should be orders of magnitude easier than transport of 0, because of the relatively high concentrations of Si and C. Accordingly, we might expect the overall reaction rate to be determined by boundary-layer diffusion of oxygen, either IIa or IIc. Fulton and Chipman's demonstration that bubbling CO through the system had no major effect on reaction rate indicates that IIc is not the slowest step. Therefore, it becomes logical to estimate the maximum rate for step IIa and to compare this theoretical estimate with Fulton and Chipman's experimental data. If oxygen diffusion across the liquid metal boundary layer at the slag metal interface (step IIa) is rate-determining, In this equation, dn sio, /dt is the rate of silica reduction in moles per sec,A is the area of slag-metal interface in sq cm, Do is the diffusivity of oxygen in sq cm per sec, 6, is the boundary layer thickness in cm, c,* is the oxygen concentration right at the slag-metal interface in moles per cubic cm, and co is the oxygen concentration in the body of the liquid metal, also in moles per cubic cm. Equilibrium data" on the silicon deoxidation reaction in liquid iron and steel at 1600°C indicate that the oxygen contents of the liquid metal in Fulton and Chipman's experiments at 1600°C probably fell in the range of 0.5 x10-3 x10-3wt pct. That is, the maximum conceivable value of co*-co for the system under consideration was on the order of 10"5 mole oxygen per cubic cm. On the basis of previously published data,1O,11 it is estimated that Do/0 will fall somewhere in the range from 10-3 to 10-1 cm per sec. The surface area A in Fulton and Chipman's experiments was approximately 20 sq cm, and the weight of metal involved was about 500 grams. Combination of all these figures with the above rate equation leads to an estimate that the rate of silica reduction should fall within the range from 0.002 to 0.2 wt pct Si per hr. This estimate is consistent with the experimental data. For example, Fulton and Chipman's Fig. 2 shows a change of about 0.3 pct Si in 10 hr, corresponding to an average rate of 0.03 pct per hr. According to the proposed hypothesis, increasing the temperature will increase the reaction rate ill two ways: 1) by increasing oxygen diffusivity and 2) by increasing the oxygen concentration (oxygen solubility) in the liquid metal. The combination of these two effects accounts for the high value of the observed activation energy. Summarizing, I propose that the rate of silica reduction, like that of the carbon-oxygen reaction, is diffusion controlled and that low oxygen concentration in the liquid metal is the major factor accounting for the very low observed rates of silica reduction. John Chipman (author's reply)—The authors thank Professor Schuhmann for his interesting contribution. His proposed explanation may well prove to be the correct one. There is clearly a need for much further experimental work on this problem, and further research is in progress.

Jan 1, 1961

By Luc Delaey, Horace Pops

The structure and morphology of thermoelastic and burst type martensitic phases that form upon cooling in Cu-Zn-Si p phase alloys have been studied by transmission electron microscopy. The martensitic phases are composed of a lamellar mixture of two close-packed structures with different stacking sequence, namely ABCBCACAB (orthorhombic) and ABC (fcc). Striations within thermoelastic martensite are most likely produced during interaction with impinging burst-type martensite and not as a consequence of secondary shears. In a study of the martensitic transformation in ternary Cu-Zn based 0 phase alloys1 the dependence of the martensitic transformation temperature, M,, with composition shows variations for elements within a constant valence subgroup and between different subgroups. Such variations are not reflected in a change in habit plane, which is approximately the same for each ternary alloy, namely in the vicinity of (2, 11, 12 Ip. The fact that the habit plane remained constant, despite large differences in M, temperature and electron concentration, suggested2 that the crystal structures of the martensitic phases could be nearly the same. Crystal structures of ternary Cu-Zn based martensites have been determined recently for alloys containing the three-valent elements gallium3, 4 and aluminm. The present studies have been made to examine the structures and morphology of the martensitic phase in ternary Cu-Zn based alloys containing a four-valent element, silicon. I) PROCEDURE Two alloys were prepared by melting and casting weighed quantities of the component high-purity metals in sealed quartz tubes under half an atmosphere of argon. They were subsequently remelted by levitation under a protective atmosphere of argon. After allowing for losses of zinc as determined by the difference in weight before and after casting, the compositions in atomic percent of both alloys were established to be Cu-33.5 Zn-1.8 Si and Cu-27 Zn-5.0 Si. These alloys were homogenized in the P-phase field for 2 days at 800" C. Bulk samples consisted of a martensite phase at room temperature, the M, temperature being approximately 30' and 200" for the 1.8 and the 5 pct Si alloys, respectively. Thin disks were cut from the ingots using a spark machine, and they were heated for 5 min at 800' and quenched into water in order to obtain martensite. These slices were thinned chemically at room temperature in a solution consisting of 40 parts HN03, 50 part H3PO4, and 10 parts HC1 and thinned further electrolytically by the Window technique, using a voltage of 15 to 25 v and a mixture of 1 part HN03 and 2 parts methanol, which was kept at a temperature near -30° c. Foils were examined by transmission electron microscopy using a Philips EM 200 electron microscope. 11) RESULTS AND DISCUSSION 1) Structure and Morphology. Fig. 1 shows the martensitic phase in the alloy containing 1.8 at. pct Si. This phase is composed of contiguous platelets, each containing striations. The direction of the striations changes at the boundary between individual platelets. These internal markings resemble the striations that are usually identified as stacking faults, as for example in Cu-A1 martensites6-a or the lamellar mixture of two close-packed phases in Cu-Zn-Ga marten-sites.3p '9 lo In the present alloys, selected-area diffraction experiments have been obtained in order to determine the nature of the striations. Figs. 2(a), (61, and (c) are electron diffraction patterns of an area inside a single martensite plate. Fig. 2(a) contains diffraction spots which correspond to two close-packed structures with different stacking sequences, namely ABCBCACAB (orthorhombic) and ABC (fcc). Spots belonging only to the fcc structure are indicated by arrows. By tilting the foil either the orthorhombic structure, Fig. 2(b), or the cubic structure shown in Fig. 2(c) may be obtained. When the foil is oriented so that only the diffraction spots of the orthorhornbic structure are present, bright-field illumination shows small lamellae, as seen in Fig. 3. In this figure the lamellae that belong to the fcc structure are bright bands inside the dark extinction contours of the orthorhombic structure. The boundaries of the lamellae are parallel to the basal planes of the orthorhombic structure and to the {Ill} planes of the cubic structure, the close-packed directions of both structures being parallel. The 5 pct Si alloy shows similar features as those described for the 1.8 at. pct Si alloy.

Jan 1, 1969

By R. W. Ludt, C. C. DeWitt

The use of froth flotation for the separation of minerals has become one of the most important of ore dressing processes. Its particular adaptability to the enrichment of low grade ores has made the process an important factor in the national economy. The methods have been extended to the recovery of a great number of minerals. Among the few minerals which have resisted efforts toward industrial flotation is chrysocolla, a hydrated partly colloidal copper silicate. Chrysocolla, being a product of natural oxidation, has been found to occur in small quantities with many ores which are recovered by flotation methods. In present practice, these small quantities of copper silicate pass off with the tailings and are lost. The advantages to be gained by a satisfactory process for the recovery of chrysocolla is apparent. Any application of principles which points a way toward the satisfactory industrial flotation process for copper silicate would be of advantage. This paper presents an attack on this problem. Two methods for the recovery of chrysocolla have been developed by the United States Bureau of Mines.1,2 They have been successful on a laboratory scale but have been seriously restricted in industrial application by critical requirements in the procedure. In one of the Bureau of Mines methods,' the ore is activated with sodium or hydrogen sulphide in an aqueous solution at a pH of 4. Amy1 xanthate is then used as a collector with pine oil as a frother in the flotation process. An excess of sulphide acts as a depressant and the state of optimum conditions is difficult to control industrially. In the second Bureau of Mines method,2 soap is used as the collector at a pH of between 8 and 9. The diffi- culties with this process are that soap is not a specific collector, that heavy metal or alkaline earth ions cause the formation of insoluble soaps, and that a more acid solution causes the formation of a free acid which does not act as a collector for chrysocolla. The problem of recovering chrysocolla by flotation involves the selection of a suitable collector. The collector molecule must be composed of an active polar group that has an attraction for chrysocolla, and of a hydrocarbon chain. Certain dyes have been shown to have an attraction for certain minerals. Suida3 found that hydrated silicates are colored by basic dyes. Dittler4 showed that chrysocolla, among other colloidal minerals of acid reaction, preferentially takes up such basic dyes as fuchsin B, methylene blue, and methyl green. Endell5 gave information to show that the colloidal material in clay may be determined by its selective adsorption of fuchsin. A simple experiment, likewise, illustrates the difference in the adsorptive power of chrysocolla and of silica for the basic triphenyl methane dyes. When a mixture of chrysocolla and silica is immersed in a very dilute dye solution, less than 5 ppm, the chryso-colla is rapidly dyed and the silica is dyed more slowly. The difference is substantial but one of degree. Dean2 showed that the dyes, crystal violet and toluidine blue, are taken up by quartz in an adsorption type process. The difference in the adsorptive power, however, offers the means by which a new collector may act. To form such a collector, a hydrocarbon chain must be attached to the dye molecule. This involves a process of organic synthesis. Butyl, hexyl, and octyl hydrocarbon chains were selected for substitution in the malachite green molecule. For the purpose of identification, the alkyl-substituted dyes formed are called: butyl-malachite green; hexyl- malachite green; and octyl-malachite green. An outline of the procedure for their synthesis is given in the appendix. It is generally recognized in the preparation of this type of dye that the chemical structure of some of the dye molecules varies. However, a uniform formula is attributed to the dye. Such a procedure has been followed in specifying the structure of these alkyl-substi-tuted malachite green dyes. The structure is given on the basis of their properties as an homologous series of dyes, on their method of preparation, and on the purity of intermediates used. Structure of substituted alkyl malachite green is: C6H4 N(CH3)2 p-R C6H4 CH C6H4 N(CH)2 Procedure The flotation cell is a Bureau of Mines 100-g, batch unit provided with an air inlet at the bottom above which is a variable speed agitator. The agi-

Jan 1, 1950

By D. A. Dukelow, K. Li, G. C. Smith

Decarburization in the Fe-C system by oxygen top blowing has been studied in laboratory -scale experiments. It is shown that equilibrium models fail to explain or predict either the course of refining or endpoint conditions, giving results which either are incompatible with the chemistry of the system or do not satisfy material balance requirements. Also the path of decarburization was found to vary even for heats made under apparently identica1 conditions. A promising approach to analyzing the decarburization results is to relate oxygen efficiency fm carbon removal to bath carbon content. This relationship for Fe-C heats shows the same range of oxygen efficiencies as is obtained in pilot-plant and commercial heats using hot metal-scrap charges. This implies that oxygen transfer is primarily controlled by the decarburization reaction itself, independent of other refining reactions. Therefore, it should be possible to study separately decarburization and slag-metal reactions. DECARBURIZATION is probably the most important reaction in steelmaking. Not only is it a main reaction in the refining of iron to steel but it also provides the stirring action in the bath necessary for the diffusion processes to proceed at reasonable rates so as to make a steelmaking process practical. Kinetics of decarburization in the open-hearth process has been a subject of investigation for many years.&apos;-B It is generally accepted that at steelmaking temperatures the rate of homogeneous C-0 reaction is extremely high and cannot constitute a rate-controlling step. Diffusion of oxygen through a boundary film in the metal phase has been suggested by arken&apos; as rate-determining. Recently, Larsen and sordah16 concluded from experiments in a laboratory furnace that, with oxygen supplied from air or combustion gases, the rate of "steady-state" carbon boil is controlled essentially by a diffusion process of O2, Co2, or H2O through a film of nitrogen above the slag surface. Displacing this diffusion film by a stream of nearly pure oxygen produced a ten-fold increase in the rate of carbon boil with the rates of slag-metal oxygen transfer, bubble nucle-ation, and other steps all apparently able to keep pace. In the top-blown basic oxygen process, however, the transport of oxygen takes a more direct route. and the state of bath agitation is much more turbulent than in the open-hearth process. In addition, direct contact of the gas with the metal phase provides opportunity for direct oxidation of carbon. It is likely that the rate-limiting factor for the decarburization reaction will be different. However, only a few descriptive discussions of the subject have been reported in the literature.10-l2 Studies of the decarburization kinetics based on plant or pilot-plant data are necessarily complicated and are influenced by other refining reactions which occur simultaneously. In order to investigate the mechanism of decarburization, experiments have been conducted in which carbon-saturated iron melts were top-blown with pure oxygen over a range of conditions. It is hoped that this study will form a foundation on which a more basic understanding of this important reaction may be built. EXPERIMENTS One group of blowing experiments was made in a standard 200-lb induction furnace and another group in a 500-lb induction furnace. The furnaces were modified to the general shape of a basic oxygen furnace by adding a rammed refractory cone section to the regular crucible body. Crucible and cone were of high MgO (95 pct) material. A water-cooled lance, 1/2 in. in diam and threaded at one end to take a nozzle, was used for blowing oxygen. The lance with its water and oxygen lines was supported on a cantilever arrangement so that it could be moved up, down, or sideways. Oxygen of 99.5 pct purity was supplied from a cylinder and metered through a rotameter equipped with pressure and temperature gages. Another pressure gage was located at the top of the lance. A schematic diagram of the assembly is shown in Fig. 1. Before each experiment, a weighed amount of ingot iron, containing 0.02 pct C, < 0.01 pct Si, 0.10 pct Mn, 0.019 pct P, and 0.015 pct S, was charged in the furnace and melted down by induction heating. Graphite was then added to the molten charge until it became saturated. When the temperature of the charge reached the desired level, the lance was lowered to a predetermined height above the bath

Jan 1, 1964

By V. M. Karve, K. K. Majundar, K. V. Viswanathan, J. Y. Somnay

The effect of calcium, magnesium, iron (both ferrous and ferric) and aluminum ions, which are commonly encountered in a typical beryl ore, was studied in the flotation of pure beryl, soda-feldspar and quartz. The vacuumatic flotation technique was employed. With sodium oleate as collector and in the absence of any activator, beryl floated in a pH range of 3 to 7.5, whereas feldspar and quartz did not float at any pH up to 11.5. The pH range of flotation increased in the presence of the ions studied. With calcium and magnesium ions beryl floated from 3 to 11.5 pH and beyond, soda-feldspar floated beyond pH 6 and quartz floated beyond pH 8. Ferrous ion activation was found to be similar to that of calcium and magnesium. Activation by ferric and aluminium ions was found to be complex and the lower and upper critical pH for all the three minerals was around 2 and 10 respectively. These studies indicated the possibility of separation of beryl from feldspar and quartz even in the presence of calcium, magnesium and ferrous ions between pH 4 and 6. Flotation tests on a mixed feed of pure minerals in a 10 g cell revealed that beryl can be selectively floated from feldspar and quartz if ferric ion is reduced to ferrous state or if it is complexed. Beryl occurs mostly in pegmatites, and hence is associated with feldspar, quartz and micas and small amounts of other minerals such as apatite and tourmaline. The separation of beryl from these minerals is difficult because all the silicates accompanying beryl have more or less the same physical properties. Specific gravities of beryl, feldspar and quartz are 2.70, 2.56 and 2.66 respectively. Electrostatic separation has been suggested but no work has been reported. &apos; The adsorption of sodium tri-decylate tagged with Cl4 on beryl, feldspar and quartz reveal similarity in surface properties. Much work has been reported on the flotation of beryl from ores, either directly or indirectly as a by-product, but little is known about the fundamental aspects of beryl flotation. Kennedy and O&apos;Meara3 laid emphasis on prior cleaning of the mineral surfaces with HF. Mica is removed first by flotation of beryl with oleic acid, around neutral pH. Runke4 introduced calcium hypochlorite conditioning in a final separation stage for activating beryl in a mixed beryl-feldspar concentrate, and after washing to remove the hypochlorite, floated beryl with petroleum sulphonate. The Snedden and Gibbs5 procedure is somewhat similar to that of Kennedy and O&apos;Meara. Emulsified oleic acid is used as collector. Recently Fuerstenau and Bhappu6 studied the flotation of beryl, feldspar and quartz with petroleum sulfonate in the presence of activators and stressed the importance of iron in the flotation of beryl. From the studies conducted in this laboratory, it was found that feldspar and quartz as such do not float with sodium oleate, but in practice selective flotation of beryl from feldspar and quartz in an ore is found to be impossible with sodium oleate as collector. A glance at the chemical analysis of typical beryl ore indicates the presence of several ions like Ca ++, Mg++, Al + + + and Fe+++ in abundance and Ti++++ and Mn++ in traces. Hence, in an attempt to explain the behaviour of feldspar in the beryl flotation, the effect of Ca++, Mg++, Al+++ and Fe+++, which are known as gangue mineral activators7&apos;8 has been investigated. Materials and Methods: Lumps of beryl ore (hand picked) were boiled with 10% sodium hydroxide and washed with distilled water. They were further boiled many times with 10% hydrochloric acid till no positive test for iron was obtained with ammonium thio cyanate. This was followed by thorough flushing with double distilled water. The lumps were crushed in a porcelain mortar and pestle under water. The minus 65 + 100 mesh fraction was used for testing and was always stored under distilled water. Pure feldspar and quartz were similarly prepared and the minus 65 + 100 mesh fractions collected. Inorganic ions tried as activators were ca++, Mg++ , Fe++, Fe ++ and A1 +++ . Calcium nitrate, magnesium chloride, ferrous ammonium sulfate, ferric ammonium sulfate and aluminum nitrate of G.R.E. Merck grade were used. B.D.H. technical grade sodium oleate was used as a collector. The vacuumatic flotation technique developed by Schuhmann and Prakash was used for studying the effect of pH on flotability. 7 The indications given by this work were confirmed by using 10 g miniature cell.&apos;

Jan 1, 1965

By J. F. Enrietto

Internal friction measurements were used to determine the effect of manganese on the solubility and precipitation kinetics of nitrogen. Manganese, in concentrations up to 0.75 pct, has little effect on the solubility at temperatures above 250°C. On the other hand, at Concentrations as low as 0.15 pct, manganese inhibits the formation of iron nitrides, especially Fe4N, even though it may not form a precipitnte itself. The precipitation and solubility of carbides and nitrides have been extensively investigated in the pure Fe-C and Fe-N systems.1-3 In recent years, some effort has been ispent in studying the influence of substitutional alloying elements on the behavior of carbon and nitrogen in ferrite.4 -7 In particular Fast, Dijkstra, and Sladek have investigated the effect of 0.5 pct Mn on the internal friction and hardness during the quench aging of Fe-Mn-N alloys.&apos;, &apos; They found that at low temperatures (below 200°C) the presence of 0.5 pct Mn greatly retarded quench aging. For example, after 66 hr at 200°C very little precipitation had taken place in the iron alloyed with manganese, whereas precipitation was complete after a few minutes in a pure Fe-N alloy. The effect of varying the manganese content and the details of the precipitation process were not mentioned in these papers. Fast&apos; postulated that manganese causes a local lowering of the free energy of the lattice with a resulting segregation of nitrogen atoms to these low energy sites. The segregated nitrogen atoms are bound so tightly to the manganese atoms that they cannot form a precipitate. The internal friction measurements of Dijkstra and Sladek tended to confirm the concept of segregation of nitrogen around manganese atoms, and the increase in free energy on transferring a mole of nitrogen atoms from a segregated to a "normal" lattice site was computed to be - 2800 cal. Dijkstra and Sladek9 distinguished between two types of precipitates: ortho, a nitride of appreciably different manganese content than that of the matrix, and para, a nitride with a manganese content essentially that of the matrix. With each type of precipitate a solubility, again designated ortho or para, can be associated. Since the internal friction maximum in alloys which were aged several hours at 600" C dropped almost to zero, Dijkstra and Sladek9 concluded that the ortho solubility must be very low. The effect of temperature on the ortho and para solubilities has no1: been investigated. There are obviously several gaps in our knowledge concerning the influence of manganese on the behavior of nitrogen in a-iron. It was the purpose of the experiments described in this paper to determine the following: 1) The ortho and para solubilities of nitrogen as a function of temperature. 2) The details of the precipitation process at elevated temperatures. 3) The effect of varying the manganese concentration on the above phenomena. EXPERIMENTAL PROCEDURE Internal friction is conveniently employed in studying the precipitation of nitrides and/or carbides from a -iron because it is one of the few parameters, perhaps the only one, which is not affected by the presence of the precipitate itself. For this reason, internal friction techniques were heavily relied upon in the present experiment. A) Preparat of -. All specimens were prepared from electrolytic iron and electrolytic manganese. Alloys containing 0.15, 0.33, 0.65, and 0.75 wt pct Mn were vacuum melted and cast into 25 lb ingots. After being hot rolled to 3/4 in. bars, the ingots were swaged and drawn to 0.030 in. wires. The wires wen? decarburized and denitrided by annealing at 750° C for 17 hr in flowing hydrogen saturated with warer vapor. To obtain a medium grain size, - 0.1 mm, the wires were then heated to 945oC, allowed to soak for 1 hr, furnace cooled to 750°C, and water quenched. Subsequent internal friction measurements showed that this procedure reduced the nitrogen and carbon concentrations of the alloys to less than 0.001 wt pct. The wires were nitrided by sealing them in pyrex capsules containing anhydrous ammonia and annealing them for 24 hr at 580°C, the nitrogen being retained in solid solution by quenching the capsule into water. Immediately after quenching, the wires were stored in liquid nitrogen to prevent any precipitation of nitrides. By varying the pressure of ammonia in the capsule, it was possible to produce any desired nitrogen concentration. B) Internal Friction. The internal Friction measurements were made on a torsional pendulum of the Ke type,&apos;&apos; a frequency OF 1. or 2 cps being used. For

Jan 1, 1962

By J. W. Downey, M. V. Nevitt

The phase boundaries for the 950" isothermal sections in the ternary systems Zr-Co-0 and `Zr-Ni-0 have been determined for the composition range from 50 to 100 at. pct Zr. The two systems show very similar phase relations, having no extensive solid solution phase fields. Each contains a ternary phase. These phases are apparently isostructural, but their structure has not been determined. Some aspects of the phase relations are discussed in terms of the alloying behavior of transition metals. THE work described in this paper is the outgrowth of a recent study of the occurrence of phases having the Ti2Ni-type structure (structure-type E9,) in certain ternary systems involving Ti, Zr, or Hf with another transition metal and O.", In the Zr-CO-0 and Zr-Ni-0 systems no Ti2Ni-type phases were found to occur. However, there are several interesting aspects of the phase relations in the two systems which have significance from the point of view of the alloying behavior of transition metals. The results of this investigation may also have some importance in studies of the oxidation of Zr-Co and Zr-Ni alloys. In both the Zr-Co-0 and Zr-Ni-0 systems only the 950" isothermal sections were investigated and, as a further restriction, the study was limited to the composition range from 50 to 100 at. pct Zr. A tentative Zr-Co binary diagram has been published by Larsen, Williams, and Pehin. In the composition range pertinent to the present work they report a eutectic at 980°C and 75.9 at. pct Zr, the products of which are the terminal solid solution based on /3 Zr and the compound Zr2Co, and a eutectic at 1080" and 64.8 at. pct Zr whose products are Zr,Co and ZrCo. The solid solution based on /3 Zr is shown to decompose eutectoidally at 826°C into a Zr and Zr2Co. The limits of solubility of Co in a and /3 Zr have not been established. The structure of Zr2Co is not identified in the publication just cited. Dwight has reported that ZrCo has the CsC1-type strcture. The Zr-rich portion of the Zr-Ni diagram has been determined by Hayes, Roberson, and Paasche." The phase relations are very similar to those of the ZrCo system. A eutectic reaction whose products are /3 Zr containing 2.9 at. pct Ni and Zr,Ni occurs at 961°c, and a eutectic between Zr,Ni and ZrNi is found at 985". The solid solution based on /3 Zr decomposes eutectoidally at 808°C. The solubility of Ni in a Zr is not known accurately but is believed to be very small. Smith, Kirkpatrick, Bailey, and Williams7 have found that Zr2Ni has a tetragonal structure of the A1,Cu-type and that ZrNi is orthorhombic. Domagala and McPherson8 have published a constitution diagram for the system Zr-ZrO,. At 950" their diagram indicates that the solid solution of 0 in 0 Zr is stable from 0 to 0.5 at. pct while the phase field of 0 in a Zr extends from 6 to 29 at. pct. These solubility limits were adopted in the present study and no binary Zr-0 alloys were made. No previous data on the phase diagrams of the ternary systems are known to exist. EXPERIMENTAL PROCEDURE The experimental details involved in the preparation of alloys in this laboratory by arc melting have been described in several previous papers"3 and they will not be repeated here. Information concerning the purity of the metals used is given in Table I. Oxygen was added in the form of reagent grade ZrO,. All of the cast specimens in both alloy systems were annealed in air-atmosphere tube furnaces at 950 3&apos; for 72 hr and water quenched. The specimens were protected from oxidation by wrapping them in Mo foil and sealing them in quartz tubes that had been evacuated at room temperature to a pressure of 1 x 10B mm of Hg. The phase boundaries were determined by metallography, and identification of the phases was accomplished primarily by X-ray diffraction methods which employed a powder camera having a diameter of 114.6 mm. The diffraction techniques which are in use in this laboratory have been previously described.&apos; An etchant that proved satisfactory for most of the alloys consisted of 5 pct by vol of AgNO

Jan 1, 1962

By Jei Y. Choi, P. G. Shewmon

The surfrrce-diffusion coefficients (DJ for Aulg8 on (100) and (111) surfaces of copper have been determined between 1050" and 780°C using a new avuzlysis imd experimental procedure. The results are: D, has also been determined fm cua4 at 870°C, and the values found are 4.5 times larger than those measured by the grain boundary grooving technique for the same surface orientations. This difference is felt to result from the approximate nature of the mathematical solution used in the present work. Attempts to measure D, for silver on copper and silver surfaces indicated a means of matter transport different from surface diffision was dominant in moving tracer from the source out over the surface. Cnlculations and experiment both indicate that this is the flow of silver through the vapor phase which completely masks the much smaller flow due to surface diffusion. The previous self-difhsion studies of D, for silver and copper are discussed in terms of our own analysis and found to yield values of D, factors of lo5 or more greater than those found by the grain boundary grooving tech -nique. UNTIL about 5 years ago it was widely believed that the activation energy for surface diffusion, AH, , was less than that for grain boundary diffusion, AHb,, which in turn was less than that for diffusion through the lattice, AHz.&apos; This was concluded from various evidence that D,> Db>Dl, and one tracer study of D, for silver on silver from which AH, was inferred.2 In 1959 Mullins and Shewmon demonstrated that D, could be determined from the kinetics of the growth of grain-boundary grooves.3 Using this procedure, Gjostein measured D, on copper between 800" and 1050°C and found that the activation energy was roughly equal to AHl .4 Subsequent work on copper,5" silver,&apos;,&apos; and goldg between the melting temperature T, and 0.87 T, confirmed that AH, as determined using the grain boundary grooving or scratch-relaxation technique was equal to or greater than AHz. During the same period, Drew and Pye again determined AH, for silver on silver using a tracer techniquelo and a mathematical solution similar to that of Nicker son and arker.&apos; Though the values of D, Drew and Pye measured at any given temperature were about 200 times smaller than those reported by Nickerson and Parker, they again found a low activation energy of about 10 kcal, or about one fifth that found at the higher temperatures with the mass transport technique. A distinguishing characteristic of these two previous tracer studies is that they have worked at low temperatures (-1/2 T,) where they felt volume diffusion was negligible and then analyzed these data as if all tracer atoms leaving the source flowed out into and remained in a homogeneous high-diffusivity surface layer of undefined thickness. This is totally different from the model used in the mass-transport studies or the studies of grain boundary diffusion, which assume the high-diffusivity surface layer to be only a few angstroms thick. If this latter model is applied to the earlier tracer studies, it is shown that the tracer has really pe!etrated into the lattice a mean distance of 1000A. Thus the tracer distribution observed after an anneal is thought to be due to the combined effects of surface and volume diffusion. Independent of the relative validity of the two models, it seems evident to us that any comparison of the values of D, as determined in these two ways is meaningless and misleading, since the values of D, and AH, obtained in these two ways would be totally different for the same physical distributions of tracer. Once the fundamental difference in the approaches of the two techniques is established, we are faced with the question of which model better approximates physical reality. Here all the evidence seems to be on the side of the &apos;&apos;thin surface layer" analysis. In fact, the authors of Refs. 2 and 9 do not argue for the "thick-layer model" we have described; they simply invoke it through the equation they use to calculate D, . The primary evidence for the thin-film approach is: a) grain boundary grooves and scratches widen in proportion to tU4 and Mullins&apos; rigorous analysis shows that this is only valid for a surface layer which is quite thin relative to the width of the groove;11 b) all accepted or seriously discussed models of solid-vapor interfaces and high-angle grain boundaries assume that the disturbed region of the interface is at most a few a0 thick. With the above in mind, it was desirable to determine D, using a radioactive tracer and a "thin-

Jan 1, 1964

By F. C. Bond

SIZE of grinding media is one of the principal factors affecting efficiency and capacity of tumbling-type grinding mills. It is best determined for any particular installation by lengthy plant tests with carefully kept records. However, a method of calculating the proper sizes, based on correct theoretical principles and tested by experience, can be very helpful, both for new installations and for guiding existing operations. As a general principle, the proper size of the make-up grinding balls added to an operating mill is the size that will just break the largest feed particles. If the balls are too large the number of breaking contacts will be reduced and grinding capacity will suffer. Moreover, the amount of extreme fines produced by each contact will be increased, and size distribution of the ground product may be adversely affected. If the balls added are too small, grinding efficiency is decreased by wasted contacts that are too weak to break the particles nipped; these largest particles are gradually worn down in the mill by the progressive breakage of corners and edges. Ball rationing is the regular addition of make-up balls of more than one size. The largest balls added are aimed at the largest and hardest particles. However, the contacts are governed entirely by chance, and the probability of inefficient contacts of large balls with small particles, and of small balls with large particles, is as great as the desired contact of large balls with large particles. Ball rationing should be considered an adjunct or secondary modification of the principle of selecting the make-up ball size to break the largest particle present. Empirical Equation In 19521,2 the author presented the following emerical equation for the make-up ball size: B - ball, rod, or pebble diameter in inches. F = size in microns 80 pct of new feed passes. Wi - work index at the feed size F. Cs - percentage of mill critical speed. S — specific gravity of material being ground. D == mill diameter in feet inside liners. K - 200 for balls, 300 for rods, 100 for silica pebbles. Eq. 1 was derived by selecting the factors that apparently should influence make-up ball size selection and by considering plant experience with each factor. Even though Eq. 1 is completely empirical, it has been generally successful in selecting the proper size of make-up balls for specific operations. But an equation based on theoretical considerations should be used with more confidence and have wider application. The theoretical influence of each of the governing factors listed under Eq. 1 was accordingly considered in detail, as described below, and a theoretical equation for make-up ball sizes was derived. Derivation of Theoretical Equation Ball Size and Feed Size: The basis of this analysis is that the largest ball in a mill should be just sufficient to break the largest feed particle into several pieces, excluding occasional pieces of tramp oversize. In this article the size F which 80 pct passes is considered the criterion of the effective maximum particle feed size. The smallest dimension of the largest particles present controls their breaking strength. This dimension is approximately equal to F. As a starting point for the analysis it is assumed that a 1-in. steel ball will effectively grind material with 80 pct passing 1 mm, or with F- 1000µ or about 16 mesh. The breaking force exerted by a ball varies with its weight, or as the cube of its diameter R. The force in pounds per square inch required to break a particle varies as its cross-sectional area, or as its diameter squared. It follows that when a 1-in. ball breaks a 1-mm particle, a 2-in. ball will break a 4-mm particle, and a 3-in. ball a 9-mm particle. This is in accordance with practical experience, as well as being theoretically correct. Confirmation of this reasoning is supplied by the Third Theory of Comminution," which states that the work necessary to break a particle of diameter F varies as F. Since work equals force times distance, and the distance of deformation before breaka4e varies as F it follolvs that the breaking force should vary as F½ These relationships are expressed in Table I, with a 1-in. ball representing one unit of force and breaking a 1-mm particle. This establishes theoretically the general rule used in Eq. 1 that the ball size should vary as the square root of the particle size to be broken. Ball Size and Work Index: The work input W required per ton" varies as the work index Wi, and the

Jan 1, 1959

By R. W. Regehr, R. A. Burmeister

Epitaxial growth of GaAs 1-xPx on germanium substrates was achieved using an open tube vapor transport system. The compositional range of 0.3 < x < 0.4 was examined. The best results were obtained with (311) orientation of the germanium substrate. The physical and chemical properties of the resulting layers were investigated using several techniques. Spectrographic analyses of the layers indicate substantial incorporation of germanium into the GaAs t-X Px layer. Evidence is presented which indicates that this incorporation occurs via a vapor phase transport process rather than by solid phase dijfu-sion. Electrical measurements suggest that the germanium thus incorporated behaves predominantly as a deep donor in the compositional range of 0.33 < x * 0.40 and has a deleterious effect upon the luminescent properties of GaAs1-x Px. The increasing technological importance of GaAs1-xPx for use in light-emitting devices has led to an evaluation of several aspects of existing growth processes. The method most commonly used to prepare GaAs1-xPx for electroluminescent device applications is vapor phase epitaxial growth on GaAs substrates.&apos;-4 In a typical electroluminescent diode structure the active region of the diode is entirely within the epitaxial layer and thus the electrical properties of the substrate are relatively unimportant since it is effectively a simple series resistance (assuming hetero-junction effects to be negligible). The use of germanium rather than GaAs as the substrate material is of interest for several reasons. First, GaAs of reasonable structural quality has been epitaxially grown on germanium4-2 and it is reasonable to expect that GaAs1-xPx could subsequently be deposited on the GaAs layer. Second, germanium substrates are readily available with both lower dislocation densities and larger areas than GaAs. Finally, single crystals of germanium are more economical than GaAs single crystals. The principal objective of the present investigation was to test the feasibility of growing GaAs1-xPx epi-taxially on germanium substrates, and to evaluate the properties of such layers with regard to electroluminescent device requirements. The approach used was to a) demonstrate epitaxial growth of GaAs1-xPx on germanium, and b) characterize the relevant structural, electrical, and optical properties of the GaAs1-xPx layers. The possibility of germanium incorporation into the grown layers was of special interest since there was some indication of this in previous studies of GaAs growth on germanium.5&apos;11,12 Although a study of the electrical properties of germanium in GaAs1-xPx was not an intent of this investigation, several features of the electrical properties of the layers grown in the present study which appear to be due to germanium are described. EXPERIMENTAL PROCEDURE The open-tube vapor transport system used for the epitaxial growth of GaAs1-xPx is illustrated in Fig. 1. This system utilizes the GaC1-GaC13 transport reaction and is similar in most respects to the larger system described elsewhere.&apos; The germanium substrates were n-type, with a resistivity of 40 ohm-cm (Eagle-Picher Co.). These were cut to the orientations of {100), {111), and (3111, and were mechanically polished and chemically etched in CP-4 (5 min at 0°C) prior to growth. In some cases, a GaAs substrate was employed in addition to the germanium. The orientation of the latter was {loo}, and they were also mechanically polished and chemically etched prior to growth. The initial composition of the deposited layer was pure GaAs. After approximately 10 microns of GaAs was deposited on the germanium substrate, the phosphorus content of the layer was gradually increased over a distance of approximately 15 microns to the desired concentration and maintained at this value throughout the remainder of the growth. Typical operating parameters used during growth are given in Table I. Selenium was used as a n-type dopant in several runs to facilitate comparison of the electrical properties of the layers grown on germanium with those of layers grown on GaAs substrates, which are usually doped with selenium. The concentration of H2Se in the gas phase was adjusted to a value which would normally yield a carrier density of 1 to 5 x 101 7 at room temperature in layers grown on GaAs substrates. The terminal surfaces of the epitaxial layers were examined by optical microscopy for structural characteristics. Laue back-reflection photographs (Cu radi-ation) were also made on the terminal surface to verify the epitaxial nature of the deposit. After these steps

Jan 1, 1970

By E. Miller, J. M. Eldridge, K. L. Komarek

The liquidus curve of the Mg-Pb system was accurately redetermined. The compound Mg2Pb decomposes peritectically at 538.2° ± 0.3°C to liquid and to a compound p&apos; which melts congruently at 35.0 at. pct Pb and 549.0° ± 0.3°C. The solidus curve of ß&apos; was determined. X-ray diffraction studies indicate that 4&apos; has an orthorhombic structure. Activity values of magnesium calculated from the phase diagram agree with those published in the literature. EXPERIMENTAL thermodynamic properties of binary metallic systems have to be consistent with values calculated from the phase diagram. In systems forming intermetallic compounds the shape of the liquidus curve near a compound is determined by the thermodynamic properties of the coexisting solid and liquid phases. Hauffe and Wagner&apos; neglected the temperature dependence of the chemical potentials and obtained the potential differences of the components of the liquid alloys, relative to stoichiometric liquid. Their calculations were based on the liquidus curve and on the heat of fusion of the compound, and were only valid near the congruent melting point. Steiner, Miller, and Komarek2 developed equations which account for the temperature dependence and obtained the chemical potentials of liquid Mg-Sn alloys over the entire phase diagram from the liquidus and solidus curves and from enthalpy values with the pure components as the standard states. The Mg-Pb phase diagram has been studied by several investigators whose results have been compiled and critically evaluated by Hansen.3 Although the liquidus curve was poorly defined, the general features of the diagram, i.e., one congruent melting compound, Mg2Pb, of essentially stoichiometric composition, two eutectics, and limited terminal solid solubilities, seemed to be suitable for a similar thermodynamic analysis. A careful redeter-mination of the liquidus by thermal analysis revealed, however, the existence of another compound. The liquidus curve between the two eutectics was precisely delineated and the structure and solidus curve of the new compound were investigated. The revised phase diagram was thermodynamic ally analyzed to evaluate the activity of magnesium in the liquid alloys. EXPERIMENTAL PROCEDURE The magnesium metal (Dominion Magnesium Ltd., Toronto, Canada) had a purity of 99.99+ pct; lead (American Smelting and Refining Co.) contained 99.999 pct Pb. Most experiments were carried out in graphite crucibles. Several experiments were made in high-purity alumina (Triangle R.R., Mor-ganite, Inc.) and in Armco iron crucibles to test the inertness of the graphite crucibles. Chemical analysis of magnesium and detailed description of the procedure for thermal analysis have been given previously. For the determination of the solidus curve of the compounds, specimens of initial composition Mg2Pb were equilibrated in a closed isothermal system with magnesium vapor. The source of the magnesium vapor was an alloy which had a gross composition lying in the 0&apos; + L field at the temperature of equilibration. As equilibrium was approached, the specimens lost magnesium to the two-phase reservoir thereby lowering the activity of magnesium in the specimens until activity and composition equaled that of the ß&apos;/ß&apos; + L boundary. Crucibles (1.9 cm ID by 2.2 cm OD by 4.1 cm high) and tightly fitting lids were machined from a molybdenum rod; small, shallow trays were fashioned from thin (0.005 in.) molybdenum sheet, and all the molybdenum components were degreased in hot carbon tetrachloride and then dried. The pieces were then degassed in vacuum at 950°C for about 6 hr. The two-phase alloy was placed at the bottom of the crucible and small specimens of the Mg2Pb compound, weighed on an analytical balance, were placed in two molybdenum trays above the two-phase alloy. The crucible was closed by forcing its lid on and then inserted in a titanium crucible. This crucible was evacuated, flushed twice with argon, and welded under argon. The specimens were equilibrated for about 1 week in a resistance furnace regulated by a Celectray controller, and the runs were terminated by water quenching. The specimens were again weighed and the equilibrium compositions were calculated on the basis that the weight losses were solely due to a loss of magnesium to the two-phase alloy. The structure of the B&apos; phase was investigated by the Debye-Scherrer X-ray diffraction technique. Selected ingots from thermal-analysis experiments containing about 35 at. pct Pb were re-melted, slowly cooled, and crushed in an argon-filled glovebox until the entire ingot passed through a 50-mesh sieve. The powder was thoroughly

Jan 1, 1965

By R. F. Shurtz

R. Duval (Mining Engineer, Ancien eleve de PEcole Polytechnique, Paris, France) I do not agree with the Eq. 3, reading: m =1/100- [(0.214x30.4) + (0.7B6 x0.00)] =6.5pct CaO If 0.214 and0.786 were proportions by weight, the equation would represent the well known mixtures law of the conventional arithmetics and 6. 5 pct CaO would be the correct average content. But it is not the case as the author states: "In samples consisting of single grains of mineral, those grains must, as already mentioned, be either of dolomite or magnesite. Since 78.6 pct of the deposit consists of magnesite and 21.4 pct of dolomite (excluding for present purpose the presence of other minerals), for any single grains picked at random the probability will be 0.214 that is it dolomite and0.786 that is it magnesite. In 1000 such samples the expected numbers of dolomite and mapesite grains will be 214 and 786 respectively." 0.214 and 0.786 would be proportions by weigbt under the necessary condition that all grains of dole mite and magnesite should have an identical weight. Obviously it is not the case, as the specific gravities are not the same for mapesite and dolomite and the volumes of the grains are different. Furthermore, because of these differences the conditions for a random sampling are not fulfilled and we are not authorized to state that the probabilities are, respectively, 0.214 and 0.786. The author however makes a simple application of Eq. 1: M = 1/n— ? fi x i . n Should we deduce that this relation is wrohg? Not at all, but when applying Eq. 1 you must not overlook what it actually. means. Eq. 1 gives a definition of the arithmetic mean of a total of n observed values Xi and nothing else. But the average conteht of a deposit has not the same significance. It is the ratio between the weight of concerned mineral in the deposit and the total weight of the deposit. As from 1000 particles the 214 of dolomite and the 786 of magnesite have not the same weight, the two definitions do not concur, and when applying Eq. 1 the result is an arithmetic mean of figures which has no connection with what is named average contentof a deposit. The situation is similar to the calculation of an average velocity. If a car travels a first mile over at 30 miles per hr and a second mile over at 60 miles per hr, when applying formula 1 you find as average velocity for the 2 miles: 30+60 ------- - 45 miles per hour. Many people calculate in this way and they do not realize that a mistake is involved. In fact the definition of he average velocity for the 2 miles is the quotient of the distance of 2 miles by the time (in hours) necessary for 2 miles travel, i.e.: 2 ---------- = 40 miles per hr. 1 + 1 30 60 In other words, the average volocity wanted is not the arithmetic but the harmonic average of the two velocities. The above mentioned bias in the calculation of the average contents of deposits is frequent, even in the assessments made by experienced engineers and is independant of what is named the sampling error. In order to supress the bias and to be able to use Eq. 1, you must apply a correction. An example on the subject can be found in an article by Duval et al. in the January 1955 issue of the &apos;&apos;Annales des Mines" (French), page 19. R. F. Schurtz (Author&apos;s Reply) Mr. Duval&apos;s position is quite correct. The proportions shown for dolomite and magnesite., respectively, of 0.214 and 0.786 are, in fact, proportions by weight uncorrected for specific gravity. In our day to day operation of producing magnesite from these mines at a very substantial rate, we do not normally make corrections for the difference between the specific gravity of dolomite and that of magnesite. If these corrections are made in Eq. 3 as shown in my article, then the numbers of grains turn out to be in proportions of 0.226 dolomite and 0.774 mapesite instead of the values actually shown in the equation. For the purposes of our work, and in view of the inherently low accuracy of the data, this correction was not deemed worthwhile making.

Jan 1, 1961

By D. W. Rudd, D. W. Vose, S. Johnson

The permeability of Mo-0.5 pel Ti to hydrogen was investigated over a limited range of temperature and pressuire (709° to 1100°C, 1.i and 2.0 atm). The resulting permeability, p, is found to obey the The experimental data justifies the permeation mechanism as a diffusion contl-olled pnssage of Ilvdrogen atoms through the metal barrier. 1 HE permeability of metals to hydrogen has been investigated by a number of workers and their published results have been tabulated by Barrer&apos; up to 1951. Since most of the work on the permeability has been accomplished prior to this date, the compilation is fairly complete. Mathematical discussion of the permeability process has been reported by Barrer, smithells, and more recently by zener. From these efforts several facts are observed. First, the permeability of metals to diatomic gases involves the passage through the metal of individual atoms of the permeating gas. This is evidenced by the fact that the rate of permeation is directly proportional to the square root of the gas pressure. Second, the gas permeates the lattice of the metal and not along grain boundaries. It was shown by Smithells and Ransley that the rate of permeation through single-crystal iron was the same after the iron had been recrystallized into several smaller crystals. Third, it has been observed that the rate of permeation is inversely proportional to the thickness of the metal membrane. Johnson and Larose5 verified these phenomena by measurirlg the permeation of oxygen through silver foils of various thicknesses. Similar findings were noted by Lombard6 for the system H-Ni and by Lewkonja and Baukloh7 for H-Fe. Finally, it has been determined that for a gas to permeate a metal, activated adsorption of the gas on the metal must take place. Rare gases are not adsorbed by metals, and attempts to measure permeabilities of these gases have proved futile. ~~der&apos; found negative results on the permeability of iron to argon. Also, Baukloh and Kayser found nickel impervious to helium, neon, argon, and krypton. From what was stated above concerning the dependence of the rate on the reciprocal thickness of the metal barrier, it is seen that although adsorption is a very important process, at least in determining whether permeation will or will not ensue, it is not the rate determining process for the common metals. A case in which adsorption is of sufficient inlportance to cause abnormal behavior has been noted in the case of Inconel-hydrogen and various stainless steels.&apos;&apos; APPARATUS The apparatus used in this study is shown in Fig. 1. The membrane is a thin disc (A), but is an integral part of an entire membrane assembly. The entire unit is one piece, being machined from a solid ingot of metal stock. When finished, the membrane assembly is about 5 in. long. Two membrane assemblies were made; the dimensions of the membranes are given in Table I. The wall thickness is large compared to the thickness of the membrane, being on the average in the ratio of 13 to 1. There exists in this design the possibility that some gas may diffuse around the corner section of the membrane where it joins the walls of the membrane assembly, If such an effect is present, it is of a small order of magnitude, as evidenced by the agreement of the values of permeability between the two membranes under the same temperature and pressure. A thermocouple well (B) is drilled to the vicinity of the membrane. The entire membrane assembly is then encased in an Inconel jacket and mounted in a resistance furnace. The interior of the jacket is connected to an auxiliary vacuum pump and is always kept evacuated so that the membrane assembly will suffer no oxidation at the temperatures at which measurements are taken. The advantages of this configuration are: 1) there are no welds about the membrane itself, so that the chance of welding material diffusing into the membrane at elevated temperatures is remote. 2) It is possible to maintain the membrane at a constant temperature. Since the resulting permeation rate is very dependent upon temperature, it is advisable to be as free as possible from all temperature gradients. 3) It is possible to obtain reproducible results using different specimens. The only disadvantage to this configuration is the welds (at C) in the hot zone. The welding of molybdenum to the degree of per-

Jan 1, 1962