Search Documents

By J. E. Spruiell, F. H. Patterson, B. F. Shuler

The microstructures and textures developed by the working of high-purzly thermochemically deposited tungstem sheet have been inuesligated. Greater than 60 pct reduction was required to deslroy the columnar grain structure which existed in the as-deposited sheet. With increasing reductions the texture gradu-ally charged from the initial as-deposited orientation with (100) parallel to the sheet normal to a (100}(011) ideal ovienlation with some spread about the rolling direction. This texture was retained after recrystalli-zation a1 2000°C. THE processes for depositing tungsten by hydrogen reduction of tungsten hexafluoride have been under development for the past several years.1,2 The nature of the deposition process is such that the as-deposited material contains a highly oriented columnar grain structure perpendicular to the deposition substrate. This well-defined columnar structure may be useful in certain applications293 but a wrought or equiaxed structure may be desired for some components such as tubing. Knowledge of the effects of working ther-mochemically deposited (TCD) tungsten is therefore necessary in order to establish the feasibility of obtaining worked and recrystallized microstructures and textures. The details of a recently developed technique for thermochemically depositing flat, uniformly thick tungsten sheet are presented with emphasis on the effects of deformation by rolling on the microstructure and texture. Although no work has been previously reported on the rolling and the annealing textures of TCD tungsten, there have been some determinations of textures for cold-rolled tungsten foil prepared from powder-metal -lurgy billets.4-8 The general description is that of a predominantly lurgy{100}(011) texture, with some spread about the rolling direction as an axis. pugh5 reports data both for cold-rolled and for annealed foil having about 96 pct initial cold work. He observed no change in texture, except for a decrease in the spread, for annealing treatments of 30 min at temperatures up to 1800°C. On annealing at 1900°C, both the low-temperature components plus new components which were described alternately as {001}(320) plus {001)(230) or as {100}(011) ± 12 deg about the sheet normal were observed. Above 1900°C, the new orientation existed alone, and it was stable up to 2500°C, the highest temperature that Pugh investigated. THERMOCHEMICAL DEPOSITION OF TUNGSTEN SHEET High-purity tungsten sheet was reproducibly deposited on the inner walls of a molybdenum mandrel by the thermochemical reaction WFs(g) + 3H2(g) - W(s) + 6HF(g). The mandrel was heated by a five-zone tube furnace which afforded a controllable fixed-temperature profile. Sheets measuring 1+ by 12 in. have been deposited with uniform thicknesses of up to 0.145 m. As shown schematically in Fig. 1, the purified reaction gases were metered into a mixing chamber at a Hz-to-WF6 ratio of 15 and preheated to 350°C. The flow rate of the WF6 was 300 cu cm per min at stp. These gases were directed through a conical injector into the deposition chamber. Inside this chamber hydrogen reduction of the WF6 to pure tungsten metal occurred on the hot walls of the molybdenum mandrel at a total pressure of 5 Torr. Typical temperatures for the five zones in the chamber were, respectively, 565", 540", 550°, 570°, and 590°C. The hydrogen fluorlde generated from the reaction, the excess hydrogen, and some un-reacted WF6 were pumped on through the scavenger

Jan 1, 1968

By Pradeep K. Rohatgi, Clyde M. Adams

Structures obtained on freezing of several hypo-and hypereutectic Al-Cu alloys over a range of solidification rates have been examined. Dendrite spacing, L, increases linearly with solute concentration and with the square root of the inverse freezing rate. The relationship for hypoeutectic alloys is: where rate of change of fraction solid with time, is freezing rate, C is solute concentration, (pct Cu)=1. Mass transport in inter dendritic liquid during solidification is analyzed; the experimental observations suggest maximum concentration differences and constitutional supercooling in the inter dendritic liquid increase with an increase in the solute concentration. The dendrite morphology changes with freezing rate and alloy composition. The dendrites of the a phase are parallel, uniformly spaced plates with slow freezing and rods with rapid freezing. Nonor-thogonal side branching has been observed in phases with cubic and tetragonal structures. Side branches in a dendrites are orthogonal with slow freezing and at 60 deg with rapid freezing. Formation of second-phase envelopes around the Primary phase is also discussed. DENDRITIC structure is characteristic of many types of phase transformation. The most extensively studied so far has been solidification of liquid solutions. chalmersl and coworkers have interpreted the formation of dendrites in terms of the breakdown of a planar interface. Most of the work done concerns itself with the development of an instability at the interface. Little theoretical work has been done quantitatively to relate the parameters of dendritic structure to mass transport in the liquid phase. A few empirical relations based on the experimental2'3 observations exist in the literature. Several workers2 including Brown and Adams1 have studied dendrite spacing in A1-Cu system as a function of solidification variables. In most cases, dendrite spacing has been found to increase linearly with the square root of some parameter proportional to the freezing time. The effect of solute concentration is not clear; some workers report the dendrite spacing increases with solute concentration4 whereas others report vice versa.''' ~ohatgi' has observed an increase in the spacing between ice dendrites with an increase in solute concentration in water. Tiller has also suggested that dendrite spacing should increase with solute concentration. In the present work dendrite spacing and morphology have been examined as a function of solute concentration and freezing rate. The freezing rate is defined as the fraction of liquid solidified per unit time, dfs/dß?, where f, is the fraction solid and 8 the time. The fastest freezing rate studied was 4550 times the slowest freezing rate. THEORETICAL CONSIDERATIONS It is of interest to analyze the concentration distribution in the liquid phase between growing dendrites during solidification, Fig. 1. Since this distribution is a direct consequence of the rejection of solute by the growing solid, a diffusional process, the concentration gradients increase with the freezing rate. However, when solidification rate is the only variable in a series of experiments, the interdendritic liquid regions become smaller (i.e., the dendrites become more closely spaced) with an increase in freezing rate. The main purpose of the analytical treatment of interdendritic liquid diffusion will be to reveal a tendency for dendrite spacing to decrease with increasing solidification rate in just such a way that the maximum concentration differences developed in the liquid phase are remarkably independent of freezing rate. Two rather different analyses are set forth, one pertaining to the one-dimensional diffusion which obtains in the interdendritic liquid between parallel plate-shaped dendrites, and the other to the cylindrically symmetrical diffusion around rod-shaped dendrites during early stages of solidification. The results of the two analyses are quantitatively similar, correlating dendrite spacing, maximum concentration difference, and freezing rate. First consider the simpler one-dimensional case. Two parallel plate-shaped dendrites are separated by a distance, L, between centers, Fig. 1. Solidification takes place by the thickening of these plates, with solute being rejected into the liquid. It is assumed there is no diffusion in the solid. This thickening process is slow enough and the dendrite spacing small enough that the concentration differences which develop, although interesting and important, are very small (an important assumption which is verified ex-

Jan 1, 1968

By R. P. Agarwala, M. S. Anand

Using residual activity technique, the diffusion of palladium, silver, cadmium, indium, and tin in alunzinum has been studied in the temperature range of 400" to 630°C. The diffusivities (in units of square centimeters per second) have been expressed as: IMPURITY diffusion in aluminum,1-9 silverand lead5 for cases of low solid solubility of the impurity in the host metal has yielded frequency factors in the range of l0-6 to l0-9 sq cm per sec whereas the activation energy is practically half the self-diffusion activation energy value. From the observed values of frequency factor, activation energy, and entropy of activation, it has been suggested' that these solutes are not diffusing by vacancy or interstitial mechanisms but by a mechanism which should be consistent with such low values of the diffusion parameters (Do and Q). However in spite of extensive work on these types of systems, the mechanism of diffusion is still not well understood. The present investigation on the diffusion of palladium, silver, cadmium, indium, and tin in aluminum has been carried out to throw further light on the diffusion mechanism in systems, where the solid solubility is very low (except for the case of silver). The results are discussed on the basis of solid solubility and the structural changes involved owing to the presence of the solutes in aluminum solid solution. An attempt has also been made to apply the existing theories of charge5-8 and size8 difference between the solute and the solvent. EXPERIMENTAL PROCEDURE Specimens (1/2 in. diam by 3/8 in. high) were machined out of pure aluminum (99.995 wt pct) rod obtained from Johnson Mattheys. They were sealed under vacuum in quartz tubes and annealed at 620° C for several hours; the grains thus developed were sufficiently large to eliminate the possibility of diffusion along the grain boundaries. The flat ends were prepared carefully after polishing as described previously,10 Radioactive nitrates of cadmium, indium, and tin and chloride of palladium containing, respectively, cd115, 1n114, sn113, and pd103 were dissolved in distilled water and mixed with 30 pct acetone. By means of a micropipet a drop of this solution was placed on a smoothly polished and lightly etched surface of the specimen. Due care was taken to see that the solution spread uniformly on the surface of specimen without trickling down its sides. Radioactive silver was elec-trodeposited using a AgCN-KCN bath. The amount of sample deposited in all the cases was not more than 0.1 µ thick. The samples were then sealed in quartz tubes in vacuum. The cadmium samples were sealed in a purified argon atmosphere to avoid evaporation. The samples were then diffusion-annealed. The temperature of annealing varied between 400° and 630°C and was controlled to ±5°C. On heating to -400°C,the deposits of cadmium, indium, and tin, which were of the order of 0.1 p in thickness, were converted to their respective oxides. The contribution of oxygen present in the lattice of aluminum due to these oxides has been calculated and found to be less than 10 ppm in all cases. Oxide method has already been used by other workers11'12 in diffusion studies without any controversy on the issue. However, in some of these investigations, metallic deposition was also tried. The diffusivities calculated from these measurements were found to agree very well with the diffusivities obtained by using the oxide method. Thus it is assumed that the measured diffusivities represent true diffusion coefficients. Since palladous chloride decomposes at about 500°C, the deposited samples which were to be diffusion-annealed below 500°C were heated in vacuum for a very short time at 500°C to allow the decomposition of palladous chloride to palladium metal. Time taken in decomposition of nitrates to oxides and chloride to metal was negligibly small as compared to the period of the diffusion anneals. The residual activity technique13 was used to study the diffusion profiles where thin layers from the specimen surface were removed by grinding it parallel to a flat surface on a 600-grade carborundum paper. The specimen was washed, dried, and weighed, the differ -ence of the weight being the measure of the thickness of the layer removed. After each such abrasion and weighing, the total residual activity on the surface of the specimens was measured by counting 0.656, 0.94,

Jan 1, 1968

By E. J. Fasiska, H. Wagenblast

The dilalion of a ivon by interslilial carbon was measured by two independent techniques —dilatometric mesurements at 719 c and X-ray measurments of the urlil cell parameters a1 room temperature after quenching. The relative expansion per increase in cavbon content by both methods is (2.9 + 0.3) x 10-3 pev at. pcl C nrzd is temperature- independent within experimental error. This corresponds to (6.3 5 0.6) cu cm per g-atom for the partial gram-atomic volume of carbon in a iron-only slightly smaller than the atomic volume for iron in the same temperature vange. THE only previous quantitative study of the dilational effect of carbon dissolved in a iron was performed in 1934 by Burns1 which, at the time, generated some discussion of possible sources of experimental error.2 For a system of such widespread importance, we felt that a new investigation was merited. Both X-ray diffraction and dilation measurements were used to determine the expansion of the a iron lattice by dissolved carbon, avoiding as much as possible any previous experimental problems and deficiencies. The dilation method at solution temperature offers not only measurements which are free of residual strain but also, in conjunction with the room-temperature X-ray measurements, a method to detect any large temperature dependence of the partial gram-atomic volume of carbon. To insure that quenching strains did not affect the room-temperature X-ray measurements, wire specimens of constant carbon content but different diameter were examined for such an effect. SPECIMEN PREPARATION The material used for both experimental techniques was "Ferrovac E" iron received in the form of 19-mm-diam rod and stated as having the following impurity contents: C, Cr, Cu, Mn, P, and V, each in the range of 10 to 50 ppm; Co, O, Mo, S, and Si, 60 to 100 ppm; W, 200 ppm; Ni, 230 ppm; N, 4 ppm. The stock material was cold-swaged to 0.71 mm diam for the Debye-Scherrer X-ray camera specimens and portions were cold-rolled from 6.3 mm diam to 0.79- and 0.25-mm sheet for X-ray diffractometer and dilation measurements. The wires were annealed in wet hydrogen for 6 hr at 840°C and 15 hr at 720°C, and then quenched into 0°C water. A chemical analysis for carbon after this treatment gave 0.0046 + 0.0014 at. pct C. Three portions of these wires were subsequently held at 719°C in three different hydrogen + methane mixtures and then quenched, resulting in carbon concentrations of 0.0283, 0.0598, and 0.1067 at. pct C by chemical analysis. After carburizing, the wires were swaged to 0.48 mm and electroplated with silver to prevent carbon loss during subsequent heat treatment. The final heat treatment consisted of holding the wires at 72 1°C for 5 min in a He-2 pct H mixture followed by quenching into 0°C brine. The wires were held at room temperature for a few minutes to remove the silver plating using a phosphoric acid-hydrogen peroxide solution, and then stored in liquid nitrogen until the measurements were made. EXPERIMENTAL TECHNIQUE A) Dilation Measurement. The dilatometer consisted of a gas-atmosphere vertical tube furnace modified so that length changes of a ribbon-shaped specimen could be measured externally. This was done by installing a gas-tight mercury seal at the top of the furnace as shown schematically in Fig. 1. The specimen (21.0 cm long, 1.27 cm wide, and 0.25 mm thick) was suspended in the center of the furnace by 3-mm-diam quartz rods with the upper one passing through the cap of the mercury seal. Above the cap, the upper quartz rod was coupled to a lever exerting a load of about 25 g (-12 lb per sq in.) on the specimen and having a 10 times mechanical magnification. The vertical position of a marker at the other end of the beam was read with a traveling microscope with a precision of 0.01 mm. The temperature gradient of the furnace was meas-

Jan 1, 1968

By Pradeep K. Rohatgi, Clyde M. Adams

Dendritic aggregate of ice formed on freezing of aqueous soluitions has been studied. Chlorides of sodium, potassiutn, lithiunz , and hydroget1 were used as solutes. The spacing between ice platelets or den-drites was experimenlally found proportzonal to the square root of freezing time when the freezing rate was constant from beginrting to end of solidification; the constant freezing rates were obtained during freezing of small droplets of aqueous soliitions surrounded by relatively large heat sinks. The transverse growth velocity of the ice platelets is proportional to the square root of the freezing rate. During unidirectional freezing from a constant temperuture chill, the solution at each location is subjected to a spectrum of freezirtg rates. The freezing rate is ini- FREEZING of solutions has been a subject of investigation in many disciplines. Aqueous solutions offer the advantage of transparenty; their study is of special significance due to the increasing interest in freezing processes for desalination. In addition dendritic solidification of aqueous solutions is in many ways analogous to freezing of metallic solutions. Aqueous solutions of most ionic salts freeze with tially low, passes through a maximum, and then decreases until solidificatimz is complete. The den-drite spacing increases linearly with the distance from the chill surface; it is inversely profiortional to the square root of the maximum freezing rate. The dendrite spacings in the two systems (the constanl freezing rate and the varying freezing rate) are equal, when the maximum freezing rate of the varying freezing rate system equals the average freezing vale of the constant freezing rate system. Mass transport in interdendritic liquid duriug solidificatiotz is analyzed; the experimental observations indicate that maximum concentration difference (AC) and supercooling (AT) in the interdendritic liquid are indepexdent of freezing rate. the separation of practically pure ice crystals from the liquid phase; with very few exceptions there is no solid solubility of ionic salts in ice. As an ice crystal grows there is rejection of solute from the moving interface into the liquid. Since the rejected solute moves primarily by diffusion, concentration gradients are established in the liquid. The highest solute concentrations are found in liquid immediately adjacent to liquid-solid interfaces and lowest concentrations farthest from the interface. Two-phase equilibrium obtains only at the solid-liquid interface; at other locations the liquid becomes supercooled because temperature gradients are negligible compared to concentration gradients. This is an immediate consequence of the fact that thermal diffusivities are considerably higher than mass diffusivities.

Jan 1, 1968

By W. C. T. Yeh, T. G. Oakwood, A. A. Hendrickson, R. H. Hammar

The objective of the research is to determine whether solid-solution strengthening effects observed in dilute solutions of silver can be accounted for by the influence of the solute addition on the dislocation structure oj- the crystals. The additions of both tin and indium produced only small changes in the dislocation densities and arrangements in silver crystals. However, as found previously, small solute additions have large effects on the tensile properties; the inj-luence of the tin and indium additions on the temperature dependence of the flow stress and the easy-glide range is especially strong; It is concluded that the indirect strengthening effect of the solute due to variations in the dislocation density as proposed by Seeger is of minor importance and that solute atom-dislocation interactions are responsible for the observed strengthenirzg effects. The experimental results were combined with those of Rogausch to test the concenlvatiorz dependence of solute strengthening. Both the first and one-half power dependences of the critical resoleed shear stress on concentratiorz fail in very dilute solutions. THE objective of the research is to determine whether solid-solution strengthening effects observed in dilute solutions of silver can be accounted for, at least in part, by the influence of the solute addition on the dislocation structure of the crystals. It is recognized that the addition of solute atoms may influence the strength properties of a metal through both "direct" and "indirect" effects. The former refer to the strengthening mechanisms that result from the interaction of solute atoms with dislocations; in the latter case, the strengthening effects arise as a result of solute's influence on quantities such as dislocation density, dislocation arrangement, stacking-fault energy, diffusivities, the elastic constants, and so forth. It is clear that the correct interpretation of solid-solution strengthening phenomena cannot be given until the importance of indirect strengthening effects is properly evaluated. In the particular case of close-packed metal crystals, Seeger showed that solute strengthening effects in dilute solutions of copper and silver might be accounted for by an increase in dislocation density due to the addition of the solute. Seeger's argument was that the strengthening effects extrapolated from more concentrated solutions indicate that small concentrations of impurities raise the critical resolved shear stress much more than is predicted by a concentration-independent dislocation density. The above idea was a very reasonable one. The dislocation theories of work hardening of Taylor,2 Cot-trell, 3 Mott, 4 and seeger5 had already associated the increased flow stresses with increased dislocation densities in deformed metals; investigations of the dislocation structure of metal crystals had provided a logical basis for expecting an increased dislocation density in crystals containing impurities (see for example, Ref. 6). The numbers involved seem reasonable, too. It can be expected that the flow stress of the crystal would increase as the one-half power of the dislocation density.' Solute additions of 1 at. pct to metal crystals result in strength increases by factors in the range of three to ten. If one assumes that the strengths of the pure metal crystals are determined by their dislocation densities, then dislocation-density increases of one to two orders of magnitude as a result of solute addition would be required to account for the observed strengthening—not an unreasonable expectation. In addition to the effect of the solute addition on dislocation density, one might also anticipate important strengthening contributions to result from the solute's influence on the dislocation arrangement. Parker and washburns have reviewed a number of experimental evidences which show important strengthening effects due to the presence of subboundaries. Further, lattice strains due to impurity segregation would be expected to influence the distribution as well as the dislocation density of the as-grown crystal. As pointed out in the reviews of Chalmers,6 Elbaum,9 and winegard,lo micro segregation of impurities occurs at all interfaces of crystals in cellular growth; the impurity gradient results in lattice strains which can be reduced with the presence of dislocation arrays in the region of the impurity gradient. Hence, one would expect the presence of a solute to favor the formation of dislocation subgrain structures and that the subgrains would have an important influence on the strength of the crystal. The experimental observations that concern the possibility of an important strengthening contribution through the influence of the solute on the dislocation density or arrangement are not in agreement. Haasen has reviewed the observations of Meakin and Wils-dorf,12 Howie,13 and Bocek 36 and concluded that the solute's influence on dislocation density is not sufficient to account for strengthening effects in concentrated solutions but might, as seegerl suggested, make an important contribution in very dilute solutions. On the other hand, Hendrickson and Fine 14 concluded that changes in the dislocation density and dislocation width accounted for the solid-solution strengthening effects observed in silver-based aluminum solid solutions. Goss et a 1.I5 observed dislocation arrays in Ge-6 at. pct Si, Ge-0.2 at. pct Sn, and Ge-0.2 at. pct B crystals that were not observed in germanium crystals of

Jan 1, 1968

By G. S. Cole, G. F. Bolling

Fluid flow strongly influences ingl structure and the columnar -to-equaaxed transition. Artificial flow patterns siwzilar to the nuturul ones act to induce this transition, while dampening forces act to suppress this transition. It is shown here how simple mechanical forces can be applied to a systerrz in order to induce or slipPvess relative fluid flow and the change both the structure transition and the exact structures obtained. The example used is the effect of slow rotations and oscillations 012 a cooling- ingot (of laboratory scale). The details of rotution and oscillation for various cooling rates, lenlperal~ire gradients, and sizes of ingot are related to other solidzfication variables. THE grain structure of a casting can be changed by altering the fluid motion during solidification. Just how important these changes are depends directly upon the degree to which the fluid motion can be controlled. An early step in understanding the nature and influences of fluid motion was taken by studying natural thermal convection; when a static mechanical device was used to suppress natural flow the result was opposite to that obtained when an electromagnetic device was used to augnlent flow in a pattern similar to the natural one. ''2 Suppression allowed more columnar growth to occur, but "augmented natural convection" brought about an earlier columnar-to-equiaxed transition (CET). Since the solidification of a free ingot must have characteristics somewhere between those displayed during suppression and augmentation, we know that natural convection must influence the CET. Further, since heat transfer in the liquid is always some time-dependent combination of convection and conduction, it can easily be guessed that the temperature gradient in the liquid is the vital factor affecting the CET.1,2 It is obvious that the structure arising from the unadulterated solidification of an ingot, without external influence, is rarely if ever the best possible one. Although the pattern of the natural fluid flow may be considered efficient, the magnitude of the flow and the heat transfer decrease freely from the outset. For a

Jan 1, 1968

By R. A. Flinn, R. D. Pehlke, D. R. Glass, P. O. Hays

Knowledge of the depth of penetralion of an oxygen jet into the bath of the oxygen converter and of the correlation of penetration with driuing pressuve, lance heighl, and nozzle throat area is vital to the understanding of converter operation. If the penetration is too shallow, then severe and hazardous slopping takes place. On the other hand, if the jel penetrates entirely Lhvough lhe bath for an apprcciublc period of time, bottom damage occurs. In addilion to measurement of the penetralion of the jel, knou~ledge of the circulatory movement in the bath is also of interest in order to evaluale various theories oj-concerter operating behavior which have been published. In this investigation, experimental converters were buill of IOU- , 300-, and 4000-lb capacity. Four independent methods were used to determine penelralion: the onset of bottom marking, a nitrogen bubbler probe, observation througlt an optical syslew built into the oxygen lance, and direcl viewing of the jet issuing from the bottom of the vessel. Good correlation zuas obtained, and empirical relalions for pvedicling perletration were found. These relations were conjzrmed by bottom marking tests in 55- and 110-ton vessels. Within the operaling conditions employed in these tests, the depth to which a single oxygen jet penetrated zuas found lo vary according to the relatiorl ThE technical literature is replete with data concerning the successful use of the basic oxygen furnace or converter in steelmaking. Experimental data are lacking, however, on the vital factors of the depth of penetration of the jet into the bath and the induced circulation. Commercial operating conditions usually have been the result of cut and try experiments in lance manipulation until satisfactory results were obtained. There have been, however, two hotly argued opposing theories concerning desirable depth of penetration and these are exemplified by the Schwarz and Miles patents1,2 on one hand and the Suess patent3 on the other. The Schwarz patent teaches that the jet, issuing from the nozzle at supersonic speed, penetrates deeply "so that the reactions between the iron and the oxygen and between the oxygen and the rest of the smelting components take place in the center of the bath". Specific operating suggestions are given by Miles.2 By contrast, the Suess patent calls for surface Circulalion was investigated by lour methods: by direct observation in 200-lb open baths, by the use of graphite rudders in the 300- and 4000-lb converlers, by direct observalion through an oplical system in the lance, and by various models al room temperature. All were in excellent agreement and indicated that the motion of the bath ulas up at the center, radially outluavd at the surface, and down at the sides. Experi-ments in small and in commercial vessels indicate that it is essential to operate with a jet penetration of approximately 50 pct of the bath depth. Surface blowing results in low oxygen eficienty and in hazardous conditions which may render the process inopeuable. RejYactory dartzage al the bottom of the vessel is only encountered when the jet penetrates to the bottom, and this can be avoided by properly applying the penetration formula. The application of this en/pirical formula in commercial peraations is best when limited to combinations of lance size, pressure, and height which are typically encounteved in the use of a single-hole lance. blowing so that "...the oxygen jet does not penetrate deeply into the molten metal bath and is confined to an impingement area at the central portion of the bath surface". These references are given merely to illustrate the basic differences between the two schools of thought and to point out the need for measurement of penetration for the sake of the operator. For example, it is shown later that inefficient and even dangerous conditions can arise if improper blowing conditions are used. Differences are also evident between the two schools of thought as to the mixing, circulation, and agitation which is to be accomplished by the jet. The Schwarz patent states that "surface contact is not sufficient in most cases to bring about quick reaction, the same as the blowing of the gas over the bath surface or the mere blowing of the gas onto the bath surface". The patent goes on to call for active mixing. In contrast the claims of the Suess patent call for "discharging a stream of oxygen ... to an extent to avoid material agitation of the bath by the oxygen stream". In this patent the circulation is said to be downward in the center and up at the sides of the vessel. A number of investigators4-12 have explored penetration and circulation in transparent models. In general, it is agreed in these tests that the circulation is upward at the center (along the sides of the jet cavity),

Jan 1, 1968

By Eugene A. Mizikar

A mathetnatical model of heal transfer in continuously cast steel slabs is described. The model, consisting of a unidimensional transient conduction equation and boundary condition equations, has been pvogrammed for computer solution. Temperatuve and solidification profiles calculated for a 6-in. slab, being cast under several conditions of secondary cooling, are presented and compared. Calculated solidification profiles are in agreement with reported expevinle~ztal values. For the mold zone, the predicted slab shell thickness can be described by: Resuilts of the study indicate that multibank spray cooling followed by radiant cooling should be employed when solidifying thick slabs with minimum surface temperature of 1600°F. Under these conditions, a 6-in. slab can be expected to solidify in about 8.3 min. Computer results also indicate that radiant cooling can replace spray cooling during solidification of the final 30 pct of slab with little increase in overall solidification time. CONTINUOUS casting has come to the forefront of the steel industry in recent years because of economic advantages resulting from increased yields and elimination of several processing steps. Considerable work, however, remains to be done regarding modifications to the process and operating procedures. Both a lack of operating plants in this country and the experimental difficulties associated with direct measurements on moving castings have made determination of solidification rates and temperature distributions difficult. An alternate approach is to mathematically simulate heat transfer in a continuously cast section, and then calculate the temperature distributions as a function of the controllable variables of the process. Simulation of heat transfer during solidification requires that a nonlinear mathematical problem be solved. As pointed out by Ruddle,' there are two mathematical approaches to the problem- the analytical approach and the numerical approach. While the analytical approach is certainly the more elegant of the two, it does require a number of inexact assumptions because of the complexity of the problem. For example, noteworthy analytical treatments of heat transfer in continuous casting have been developed by Savage Hills,3 and pehlke4 only by making one or more simplifying assumptions such as invariant thermophysical properties, constant heat-transfer coefficients in the mold, and linear temperature profiles in the shell. Simplifying assumptions such as these can introduce considerable uncertainty in the validity of results calculated with analytical solutions. Numerical solutions, which are considerably more versatile, appear to be better suited for solving solidification problems. Complex variations in the boundary conditions and variable thermophysical properties can be handled readily with this technique. Whereas numerical computations can be long and tedious when done by hand, results can now be obtained quite rapidly with the use of either the digital or analog computer. Several numerical models of heat transfer in continuous casting have been published. In 1963, Adenis, Coats, and Ragones published a numerical model used to calculate temperature distributions in direct-chill-cast magnesium billets. More recently, Donaldson and Hess 6 presented results obtained with a numerical computer model of heat transfer in continuously cast steel billets. In the present study, a model of unidimensional heat transfer in continuously cast slabs is presented. The method of solution on a digital computer is also included. Calculated temperature distributions and solidification profiles for various schemes of secondary cooling along with attempts to verify the model are also discussed. MATHEMATICAL MODEL The schematic representation of the slab continuous casting process in Fig. 1 illustrates that the slab passes through three distinct zones of cooling. Accordingly, the mathematical model consists of three parts: 1) solidification in the mold; 2) solidification in the spray cooling zone ; 3) solidification in the radiant cooling zone. Heat-Transfer Equations. The model was developed bymaking a heat balance on a horizontal slice of slab over the time period required for the slice to proceed from the liquid metal meniscus in the mold to the cutoff station. As shown in Fig. 2, the imaginary slice extends from the center line to the surface of the slab. As the slice moves downward, heat is conducted from the center line to the surface of the slab at a rate governed by the surface boundary condition and thermophysical properties of the metal in the slice. The following partial differential equation describes the conduction of heat in a medium moving at velocity U in direction Z:

Jan 1, 1968

By L. C. Michels, O. G. Ricketts

The disorientations between s?nall grains, whose growth has been arrested at an early stage of recrys-tallization, and the deformed matrix in cold-rolled aluminum single crystals were determined using transmission Kikuchi line and electron diffraction patterns. The orientations of the recrystallized grains were found to be random, and the disorientations of these grains with the matrix weve found to be intermediate to large. This leads to the conclusion that the observed vecrystallization began in small areas of large disorientation present in the cold-worked structure. heavily cold-worked thin sections of aluminunz single crystals and of polycrystalline aluminum and nickel were produced directly by a mechanical technique. The specinlens thus prepared were heated with the electron beam to bring about vecrystallization during observation in the electron microscope. Motion pictures taken du.ring heating and the electvon, microg.raphs taken both before and aftev heating allowed the recrystallization process to be traced to its ovigin. Re cvystallized grains originated in very s,mall regions of the cold-worked structure and developed through rapid migration of high-angle boundaries. The boundaries either were present as such in the matrix or were formed out of dense dislocation networks. SIGNIFICANT advances have been made in recent years in the study of nucleation of recrystallization using the technique of transmission electron microscopy of thin metal foils. Bollman1 in a study of heavily rolled polycrystalline nickel found support for the Cahn-Cottrell2,3 theory of nucleation. According to this theory nuclei form by the initially slow growth of subgrains formed through polygonization. During this initial period of slow growth (the incubation period) the migrating boundary of the subgrain increases its disorientation with the cold-worked matrix and thereby increases its mobility to become a rapidly migrating high-angle boundary. Bailey4,5 investigated the annealing behavior of several metals deformed both in tension and by rolling and concluded that recrystallization took place through the migration of high-angle boundaries. With low deformations these boundaries were present in the metal before deformation. With high deformation it was not possible to tell whether the boundaries were pieces of the original grain boundaries or were produced either during deformation or by polygonization during ameal- ing. Direct observation during heating of metal foils indicated that subgrains form by polygonization and grow at an uneven rate. The grain size obtained decreased with decreasing foil thickness indicating that the foil surface resists boundary motion. Votava,6 in heating stage experiments on rolled copper, observed nuclei to appear suddenly and grow in jumps of differing magnitude. However, he found no special dislocation configurations where the nuclei appeared. Fujita,7 as a result of a study of subgrain growth in heavily worked aluminum, concluded that the boundary of a recrystallized grain initially forms from the boundary of a group of subgrains. This occurred by a process of deposition of vacancies and dislocations in the group boundary as the boundaries within the group disappear. HU8,9 directly observed a similar process in heating stage experiments on 70 pct rolled Si-Fe single crystals. The growth of subgrains appeared to proceed by a coalescence mechanism. The observed fading away of the boundary between two subgrains was explained by the moving out of dislocations from the disappearing boundary into the connecting or intersecting boundaries around the subgrains. The subgrain size and degree of disorientation with the surrounding structure were thus increased. With the increase in disorientation occurred a corresponding increase in boundary mobility, which eventually allowed the boundary to migrate rapidly. This process was observed to occur within "microbands" consisting of parallel narrow segments disoriented by a few degrees present in the as-rolled structure. The conclusion of Rzepski and Montuelle10 that growth is preceded by the coalescence of blocks through disappearance of their common boundaries supports this view. In contrast to Hu's coalescence model for nucleation were the conclusions of Walter and ~och.""~ Working with the same material as Hu, of the same orientation and rolled to the same reduction, they concluded that nucleation occurred by the Cahn-Cottrell mechanism. They observed, in agreement with Hu, that recrystallization began in the "microband" regions which they referred to as "transition" bands. Bartuska13 studied subgrain growth in heavily rolled nickel using a beam heating method in the electron microscope. He concluded that nuclei for recrystallization form from the largest most perfect subgrains present in the cold-worked structure by rapid intermittent migration of parts of subboundaries. In rare instances he observed subgrain growth by coalescence. EXPERIMENTAL PROCEDURE The materials used in this study were 99.999 pct A1 supplied by A.I.A.G. Metals, Inc., and 99.999 pct Ni supplied by Johnson and Matthey and Co., Ltd. The Hitachi HU-11 electron microscope, with uniaxial

Jan 1, 1968

By Craig R. Barrett

The influence of the stress dependence of the dislocation density on the overall stress dependence of The steady-state creep rate is discussed. Experimental measurements of dislocation densities and creep rates jor an Fe-3.1 pct Si alloy tested over a wide range of stress are presented which indicate a close correlation between the two quantities. EXPRESSING the stress dependence of high-temperature steady-state creep rates has been the subject of numerous investigations over the past decade. From these investigations have emerged a few general equations which fit most of the available data reasonably well. Garofalo 1 has proposed that for a wide range of stress and temperature the creep rate, i, can be expressed by a relationship of the form: where A, ß, and n are temperature-dependent and is the applied stress. Eq. [I] gives a stress dependence of a at low stresses and exp(nßo) at high stresses, both consistent with experimental observation where n = 5. A number of other investigatorshave proposed a similar expression where: where m and w are in general functions of temperature. Theoretical justification for Eqs. 111 and [2J comes from two different approaches, theories based on dislocation glide3-5 and theories based on the balance between work-hardening and recovery.' Dislocation glide theories predict where p is the mobile dislocation density and V is the activation volume. In order for Eq. [3] to predict a on stress dependence at low stresses with n = 5, it is necessary for p to be a strong function of stress, about proportional to u3 or u4.436 Recovery models of creep start with the general expression:' where r is the recovery rate and h is the work-hardening rate, and then employ a detailed model to determine the stress dependencies of r and h. As example, weertman2 has described a model based on pile-up hardening which results in an expression similar to Eq. [2] with m = 2 and p = 2.5. A recent, seemingly more general approach suggested by McLean is to use Friedels dislocation network growth expression to calculate the stress dependence of r. This model predicts r = p 3,2 and, as h is taken to be a relatively weak function of stress, the stress dependence of i should be about proportional to p3, 2. Once again, it is required that p be a strong function of stress to yield i This recovery model is by no means complete as the stress dependence of network growth is calculated in the absence of an applied stress meaning that r should probably be a stronger function of stress than given by p3, 2 and also some recent work by Mitra and McLean8 suggests that the stress dependence of h may be as large as Because of the importance of the stress dependence of p in determining the overall stress dependence of the strain rate it seems reasonable to examine this relationship in some detail. Measurements made on dislocation densities present during high-temperature steady-state creep generally obey the following relationship a =o0 + aGVWp [5] where uo is usually called a friction stress, a is a numerical constant about equal to unity, G is the shear modulus, and b is the Burgers vector. Depending on the magnitude of uo, data obeying Eq. [5] can give rise to a range of stress dependencies of p. For example, if we assume that the expression where 6 is some constant, adequately represents the stress dependence of p over a range of o, then from Although Eq. [6] obviously breaks down when u 5uo for uo 6 can vary over a wide range of values. In particular, if u > a, > 0 then 6 will be greater than 2. As an example the data of Ishida and McLean are plotted in Fig. 1 showing excellent agreement with dynes per sq cm) and also showing good agreement with a p cc u3 dependence when plotted on a log-log basis. As oo increases 6 tends to increase and correspondingly the stress dependence of the predicted creep rate should also increase,* as-

Jan 1, 1968

By A. F. Giamei, E. J. Freise

A discussion of the various factors affecting the accuracy of volume fraction determination by the direct comparison X-ray diffraction method is presented. To minimize errors introduced by nonrandomization effects such as preferred orientation and large grain size, as well as errors introduced by statistical effects, an iterative mathematical analysis is introduced. This analysis is applied to all observable diffractio~ spectra from a given specimen by means of computer techniques and is statistically superior to the standard direct comparison technique which utilizes only one or tm sets of intensity spectra. The computer program also presents a graphical interpretation of- the results which enables the investigator to visualize and readily assess the scatter in volume fraction estimates caused by deviations from randomness. Finally, the iterative analysis is applied to experimental data collected from a series of plain carbon steels in both the as-received and spheroidized conditions to determine the volume fraction of cemen-tite present in each case. The results show markedly good agreement with chemical analyses for all spheroidized specimens. In theory, quantitative X-ray diffraction is a power-ful analytical technique for precise determination of the volume fractions of various phases present in a multiphase poly crystalline aggregate. There are two generally used methods; both utilize the fact that the intensity of a diffraction line of a particular phase is a function of volume fraction of that phase present in the aggregate. The more common method is the so-called internal standard technique where a known amount of standard substance is mixed with the aggregate containing the unknown. The resulting ratio of the intensity of a particular diffraction line of the unknown to the intensity of a diffraction line from the standard is used to find the weight fraction of unknown material by reference to the intensity data obtained from a previously prepared set of standards containing the same amount of standard material. The details of this technique are outlined in standard works on X-ray diffraction.' The disadvantages of the technique are that it requires the use of powder specimens and that the powder mixture must be homogeneous near the surface to the maximum effective beam depth. The alternative technique is the direct comparison method in which the ratio of the intensity of a diffraction line of a phase in the aggregate to the intensity of a second-phase diffraction line yields the ratio of the volume fractions of each phase present. This method has been applied most extensively to the determination of the percentage of retained austenite in martensitic steels and to determining the amounts of hcp and fcc allotropic modifications in pure cobalt5 and the amounts of hexagonal and rhombohedra1 modifications in graphite.8 This method is applicable to bulk as well as powder aggregates and is the technique which will be considered in the following text. It is the purpose of the present communication to show that by combining the results of X-ray diffrac-tometry with the use of high-speed digital computers the results of the direct comparison quantitative X-ray analysis technique can be optimized by using all of the data available on a diffraction pattern. THEORETICAL DISCUSSION In optimizing any quantitative X-ray diffraction experiment, the problem of accuracy vs time to collect and analyze the experimental data arises. For any method to be practical, it is required that both experimental and analytical procedures are reasonably short. In most of the previous work performed by the direct comparison method, intensity data from only one or two sets of peaks were used to calculate the volume fraction of phases present in the aggregate. Consequently, rather long-time experimental measurements are required to assure that the intensity data are sufficiently accurate to justify the use of one or two sets of data. By using computer techniques to analyze the experimental data it becomes feasible to use all of the available data on a diffraction pattern. By so doing, the time spent in measuring any one diffraction maxima can be appreciably reduced. Although this results in poor counting statistics for any one maxima, the number of data points available for analysis is greatly increased. Based on the above discussion, it will be assumed that a continuous scanning counter-diffractometer technique will be used to collect intensity data from a two-phase mixture. The integrated intensity of the ith diffraction line from the a phase in the aggregate is given by:'" where is the volume fraction of the a phase, The various quantities in Eq. [la] are angular independent while those in [lb] depend on the hkl spectrum of a particular phase. The meaning of the individual quantities in Eqs. [la] and [lb] are given in Appendix 1. It is convenient at this point to outline briefly the factors which will affect the measured intensity of any one hkl spectrum. a) Extinction effects arise from two independent experimental conditions. Primary extinction arises from a high degree of crystal perfection and results in spectra having lower intensity than predicted by Eq. [I]. Any treatment, such as powdering, which reduces

Jan 1, 1968

By T. O. King, Y. P. Gupta

The self-diffusion of sodium in two sodium silicate liquids was measured in the temperature range 850" to 1500°C by the capillary-reservoir technique. Radioactive Na 22 was used as the tracer. The total count and autoradiographic methods were used for determination of the total Na 22 depletion and diffusion profiles of the dqy-used specimens. The diffusion coefficients obtained by the autoradiographic technique are slightly smaller than those obtained by a total count method. Error analysis of the two methods suggests that more confidence be placed in the results of the total count method. The experimental results were analyzed in terms of the existing activated rate process theory. The activation energy for diffusion was shown to decrease markedly as the temperature increased. This was attributed to a variation in the heat of activation with temperature, probably related to a change in the distribution of anions associated wilh the cation. In terms of a model, suggested for diffusion in liquid silicates, differences in the aclivation energies for diffusion and electrical conduction may arise from the effect of the electric field applied in conduction measurements. It is generally accepted that liquid silicates consist of cations and an equilibrium distribution of complex anions determined by temperature and composition. Transport properties in molten silicates are of interest, not merely because of their relevance to the kinetics of pyrometallurgical reactions and glassmak-ing processes, but also because they are useful to development of the theory of such ionic liquids. Self-diffusion is one such transport process that may, if studied for liquid silicates of widely varying compositions, indicate structural changes through changes in the activation energy for diffusion. The self-diffusion of calcium, silicon,' sulfur: aluminum, 4 and oxygenS in lime-alumina-silica melts and of iron6 in molten iron silicates have been measured. Unfortunately, the errors in some of the reported activation energies for diffusion are too large to allow firm conclusions concerning structure and the mechanism of the diffusion process to be drawn. The research reported here was a study of the self-diffusion of sodium in liquid sodium silicates. The soda-silica system was chosen since: i) a reasonable composition range can be covered in the binary system at moderate temperatures; ii) suitable isotopes of sodium (radioactive Na 22 and NaZ4) can be obtained; iii) data on the electrical conductivity and viscosity of sodium silicate liquids are available. However, the range of composition actually used in diffusion experiments was limited, to 20 to 35 wt pct soda, by the relatively high viscosity of silica-rich compositions and by evaporation of soda from basic melts. EXPERIMENTAL PROCEDURE The capillary-reservoir technique was used, wherein a radioactive tracer, Na 22, incorporated in the capillary melt, was allowed to diffuse out of the platinum capillary tube into a large reservoir of silicate liquid containing a chemically identical melt, but without radioactive tracer. After a specific diffusion time, both the total depletion in the tracer concentration and concentration-distance profiles of the diffused samples were measured by procedures to be described later. The diffusion cell assembly was similar to that used by Koros and King.' The temperature of the diffusion cell was measured with a calibrated Pt, Pt-10 pct Rh thermocouple located at the center of the liquid reservoir. The same thermocouple was used to obtain the temperature profile in the reservoir. The temperature at the top of the reservoir was slightly higher (2°C) than that at the bottom, to minimize convection in the capillary tubes, which were placed, open end up, in the reservoir. Convection was not expected to be a problem since the length-to-diameter ratio of the tubes was about 12:1 and the maximum capillary diameter was 1 mm. The diffusion cell was heated in a molybdenum-wire resistance furnace, previously used by Koros and King, but somewhat modified. A Pt, Pt-10 pct Rh thermocouple enclosed in an alumina tube and located near the furnace winding was used in conjunction with a preamplifier and Micromax proportional controller for temperature control, within 4°C, at temperatures near 1500°C. Before starting a diffusion run, the furnace was heated to the desired temperature and an alumina guide assembly for the capillaries was slowly lowered to within 2 cm of the liquid reservoir. The capillaries were thus heated to about the temperature of the bath. The run was started by lowering the assembly till the open ends of the capillaries were about 1 cm below the surface of the liquid. The run was ended by raising the assembly to about 15 cm above the liquid bath. Later, the sample assembly was slowly withdrawn from the furnace. Six diffusion samples were usually run at the same time. Three runs made with the open ends of the capillaries down were discarded because in the majority of these samples air bubbles trapped at the open end were observed. Some glass adhered to the outside of the capillary on withdrawal. This was carefully removed and used to check that the Na 22 activity in the sink remained at a low level. Materials. Nonradioactive sodium silicates were prepared by melting, in a platinum crucible, weighed quantities of sodium carbonate and powdered silica. To prepare radioactive sodium silicates, Na2 obtained in HC1 solution, was first checked for radioactive impurities by obtaining a y-ray energy spectrum, then the chloride ions were removed by anion-ex-

Jan 1, 1968

By A. D. Zunkel, A. H. Larson

Equilibrium arsenic contents of Pb-As alloys in contact with PbO-As2O3 slags containing less than 30 mol pct As2O3, were determined at 650°, 700: and 750 C in an inert at?rzosphere. In this temperature range, the arsenic content of the alloy increased with increasing temperature in the single-phase liquid slag region and decreased with increasing temperature in the two-phase slag region and the single-phase solid-solution slag region. The PbO-As2O3 phase diagram below 22 mol pct As2O3 was determined by thertrzal analysis and by application of a log ?As2O3/?PbO vS log XAs2o3 /x3pbO plot determined from the equilibrium actiuity data. The resulting phase diagram was not well-defined since the eutectic temperature was not detected in the thermal analysis experiments, although a region of terminal solid solubility of As2O3 was found. Results from the phase diagram determination are compared with an existing diagram in the literature. THIS experimental investigation is an extension of a study by the authors1 on the slag-metal equilibria in the systems concerned with commercial lead refining processes such as softening and dross fuming. The first part of this investigation was a study of the slag-metal equilibria in the Pb-PbO-Sb2O3 system. The only experimental work previously done on the Pb-PbO-As2O3 system was by Pelzel2,3 in which the phase diagram for the PbO-As2O3 system was determined below 50 wt pct As2O3 and the equilibrium constant for the reaction 3Pb + As2O3 + 3PbO + 2As was determined as a function of temperature. No slag-metal equilibrium data have been determined. It is due to the scarcity of information regarding the Pb-PbO-As203 system that this work was undertaken. This paper describes the determination of the slag-metal equilibria in the Pb-PbO-As203 system by equilibrating Pb-As alloys with PbO-As2O3 slags in an inert atmosphere, the effect of 1 wt pct additions of bismuth and copper on the slag-metal equilibria, and the PbO-As2O3 phase diagram both by thermal analysis and the use of the slag-metal equilibria data. EXPERIMENTAL Materials. The materials used in this investigation were analytical reagent-grade and assayed as follows: 1) 99.8 pct PbO (0.014 pct insoluble in CHJCOOH, 0.02 pct not precipitated by H2S, 0.1 pct CaO, and 0.08 pct SiO2); 2) 99.95 pct As2O3; 3) 99.99 pct Pb; 4) 99.0 pct As; 5) 99.99 pct Cu; and 6) 99.97 pct Bi. Room-temperature X-ray patterns revealed no detectable impurities in any of these materials. Apparatus for Equilibrium and Thermal-Analysis Determinations. The resistance-heated crucible furnace used in this investigation employed nichrome elements and was mounted so that it could be raised to surround the reaction tube during each experiment and, subsequently, lowered. A schematic diagram of the apparatus is shown in Fig. 1. Each charge was heated in a 3+-in.-OD by 31/2-in.-high 416 stainless-steel crucible placed in a 41/4-in.-1D by 18-in.-long fused-silica reaction tube which was closed at one end. On a shoulder around the crucible was placed a 3: -in.-OD by 12-in.-long open-end fused-silica condenser tube. The open end of the reaction tube was covered by a water-cooled brass cap with ports for 1) admitting an inert atmosphere to the system through a stopcock, 2) introducing a stainless-steel, motor-driven, paddle stirrer into the crucible, 3) evacuating the system with a mechanical vacuum pump, and 4) sampling the melt with Vycor sampling tubes. The brass cap was fitted to the open end of the reaction tube with a silicone gasket and collar clamp. The furnace temperature was controlled by a Barber-Coleman Capacitrol controller and a chromel-alumel thermocouple. Due to the corrosiveness of the melt, the controlling thermocouple also served as the measuring thermocouple. The temperature of the melt was calibrated against the controller temperature and was checked periodically during each test with a Vycor-enclosed calibrated chromel-alumel thermocouple. The temperature measurement and control can be considered accurate to ±3°C. Procedure. The charge placed in the crucible for each experiment consisted of 1000 g of Pb-As alloy and 300 g of PbO-As2O3 slag. The crucible was then placed in the reaction tube, the condenser tube was placed on the shoulder of the crucible, the silicone gasket and brass cap were fitted on the open end of the reaction tube, and the entire system was evacuated and filled with argon ten times. After the last flushing, a positive argon pressure of 1 psig was impressed on the system. The furnace was then raised to sur-

Jan 1, 1968

By Y. Baskin

Solid-solubility relationships in the Pseudobinary systems UAS-UP, UAs-US. UAS-UC, aid UAs-UN were investigated. The first two systems exhibit complete mutual solubility, whereas the component compounds in the other two systenzs are immiscible. The above information, together with solid-solubility data joy six additional pseudobinary systems , were analyzed for compliance wilh the Hurrze-Rothery rules for rnetallic systems. The relative size difference of the component nonmetal atoms was found to be the dopainant jactor determining the extent of solid solubility between the NaC1-type uranium compounds. The anionic and covalent radii of the nonmetal atoms appear to be inadequate for these systems, but compuled radii based on rare earth compounds yield consistent results for the uranium compounds. THE actinide elements, like metallic elements of the transition and rare earth series, readily form binary compounds with nonmetallic elements of groups IV, V, and VI of the periodic table. Of particular importance are the NaC1-type equiatomic compounds with carbon, nitrogen, sulfur, phosphorus, and arsenic. The uranium members of this family of compounds have high melting points, are essentially stoichiometric, and exhibit various amounts of mutual solubility. Thus, they are of interest for investigating the factors governing the extent of solid solubility. Previous investigators have determined the solid-solubility limits in the pseudobinary systems between the compounds UC, UN, US, and UP. Anselin et a1 .' reported complete miscibility in the system UC-UN. Baskin and shalek 2 and Allbutt et a1.3 reported that UP and US exhibit complete mutual solubility. Shalek and white4 reported partial miscibility in the system US-UC. At 1800°C the maximum solubility of UC in US is 40 mol pct, but that of US in UC is 4 rnol pct. shalek5 found limited solubility in the system US-UN; the maximum solubility of UN in US is 11 rnol pct at 1800°C, while that of US in UN is only 0.3 mol pct. White and askin 6 found very limited miscibility in the system UP-UC at 1800°C. Approximately 7 mol pct UC is soluble in UP, but there is no solubility of UP in the monocarbide. Phase relations in the pseudo-binary system UN-UP were investigated by askin.' Approximately 0.7 mol pct UN is soluble in UP at 1800°C, while UP is immiscible in UN. The present study was carried out to explore the extent of terminal solubility in the systems UAs-UC, UAs-UN, UAs-US, and UAs-UP. This information, combined with existing data, provided a sufficient basis on which to determine the factors governing solid solubility in pseudobinary systems containing NaC1-type uranium conpounds. I) EXPERIMENTAL 1) Materials. The compounds UC and UN were obtained from the Kerr-McGee Corp. and United Nuclear Co., respectively. The US, UP, and UAs were synthesized by reacting finely divided uranium with H2 S, pH3, or AsH3 gas at low temperature (300° to 500°C), followed by homogenization in a vacuum at moderately high temperatures (1400° to 1700°c).8-10 The materials were essentially stoichiometric, with the exception of UC, which exhibited a C/U ratio of 1.05. Oxygen was the major contaminant in these compounds, ranging from 0.05 wt pct in US to 0.30 wt pct in UC, and it was generally combined with uranium to form UO2. The UO2 content in these materials was usually of the order of 1 wt pct, and did not exceed 2 wt pct. Furthermore, no evidence was found for a high-temperature reaction between uranium dioxide and any of the compounds. Chemical analyses of equilibrated compositions in the systems UAs-UP and UAs-US showed that the non-metal atom to uranium ratios averaged about 1.01, and that the oxygen contents ranged from 0.06 to 0.22 pct. However, the small deviations from stoichiome-try or the presence of minor oxygen impurities do not invalidate the conclusions to be drawn from this study. 2) Experimental Procedures. The component compounds in powdered from were blended in the desired proportions for 5 hr in the ball mill that consisted of stainless-steel balls in a plastic container. Chemical analyses indicated very little metallic pickup from the blending operation and virtually no increase in oxygen content. The pellets were pressed in a 0.270-in.-diam steel die under 40,000 psi pressure. One wt pct of stearic acid dissolved in CCl 4 served both as a binder and as a die lubricant. Chemical analyses revealed that the stearic acid left no carbon residue in the sintered samples. The pellets were sintered in vacuum in an unsealed tantalum crucible. The temperature, measured with a calibrated optical pyrometer, was maintained at 1800" + 30°C for 3 hr. This was sufficient time for attaining equilibrium as no change occurred in either the lattice parameters or the sharpness of the X-ray patterns when samples were annealed for longer periods of time. The pellets were cooled with the furnace. Debye-Scherrer powder patterns were taken at room temperature with a 114.59-mm-diam Norelco powder camera and CuKor radiation (CuGI = 1.5405A). Unit cell dimensions were determined from a Nelson-Riley extrapolation to the high-angle reflections. The values for were precise to k 0.001A. 11) RESULTS X-ray and met allographic investigation revealed that complete mutual solid solubility exists in the pseudobinary systems UAs-UP and UAs-US. The lattice parameter vs composition plots, Fig. 1, show a

Jan 1, 1968

By D. D. Pollock

Many constitution diagrams do not indicate the limits oJ solid solubility of the silver-rich solid solutions below 200°C. Thermoelectric measurements were employed to determine this limit in the nieighborhood of room temperature. The appearance of a second phase causes a discontinuity to occur in the tkevtrroelectric power as a function of cornposition at constant temperature and pressure. This discontinuity was employed to determine the limits of the silver-rich terzinal solid solutions of the Ag-A l, Ag-As, Ag- Bi, Ag--Ge, Ag-In, Ag-Pb, Ag-Sb, and Ag-Sn systenls at 30°C. CONSIDERABLE work has been done on the constitution of binary alloys of silver. The available data have been compiled by Hansen1 and Elliott.2 However, many of the binary diagrams do not show the limits of solid solubility of the silver-rich solutions below about 200° C. This work was undertaken in an attempt to extend these limits to the neighborhood of room temperature . PREPARATION OF ALLOYS All of the alloys were made with silver containing less than 0.014 pct (by weight) of impurities. The materials were weighed out into 225-g heats and melted in a high-frequency induction furnace. The indium, tin, and antimony alloys were melted in A1203 crucibles, under a boric anhydride cover, and were poured into split graphite finger molds. The aluminum, germanium, arsenic, lead, and bismuth alloys were melted in machined graphite crucibles; they were also covered with boric anhydride and poured into split graphite finger molds. The $-in.-diam ingots were heated to 750°C and were forged and hot-rolled down to 0.1 by 0.1 in. The wires were pickled in a 25 pct HNO3-75 pct H2O solution to remove any oxygen-rich surfaces. The wires were then annealed at 550°C for 2 hr in a dissociated ammonia atmosphere. They were then furnace-cooled for 20 hr to room temperature in this atmosphere. The wires were cold-drawn to 0.0808 in. diam. The finished wires were again annealed at 550°C in the same manner as in the first anneal. Thermoelectric Power. The thermoelectric power of a single specimen of each alloy was determined against pure silver which contained a maximum impurity content of 0.014 pct by weight. The wire specimens were straightened and then cleaned with ethyl alcohol. They were then reannealed in the same way as the finished wires to assure that any residual stress would be at a minimum level. Previous work has shown that the presence of small amounts of stress has the effect of decreasing the thermoelectric power. The alloy specimen was soldered to the pure silver wire to form the measuring junction. A standard set of copper leads was then soldered to the reference junctions of the thermocouple. These junctions were maintained in separate dewar flasks which contained finely shaved melting ice during the test. The thermal electromotive forces of the thermocouples were measured at four temperatures between 15° and 45°C. A known source of electromotive force was balanced against the output of the thermocouple until a null balance was reached. The data were then expressed in terms of the thermoelectric power at 30°C. The results were then converted to absolute thermoelectric power, s303.3 These data have an average error of less than ±0.2 pct. RESULTS AND DISCUSSION All of the experimental data are given in Table 11. Since the electromotive forces of the binary alloys of silver were determined against the pure silver from which they were made, the resultant thermoelectric powers are only a function of the intentionally added binary elements at the constant test temperatures. This provides a direct measure of the effects of the alloying elements. The absolute thermoelectric power of an alloy of silver is given4 by:

Jan 1, 1968

Jan 1, 1968

By A. W. Schlechten

The proposition is advanced that edutcation specifically for exlractive metallurgy is desirable not because the course content is cmique but rather as a means of indoctrinating the student with a backgrocmd that will give him an interest and pride in the profession. Factors that have caused the decline in enrollment in metallurgical engineering courses emphasizing extractive processes are being overcome. There is a need to distinguish between the sciences of metallurgy and the engineering applications to avoid confusion both in and out of the profession. AS a first step in preparing the 1967 EMD Lecture I searched the literature to determine what my predecessors had done hoping thereby to gain some clues for my own performance. The Transactions of the Metallurgical Society show that the first EMD Lec- turer in 1959 was Dr. W. J. Kroll, known chiefly today for his pioneering work on the production of titanium and zirconium, but equally respected by an earlier generation for his contributions to lead refining. He has been followed by a series of well-known and competent metallurgists all possessing specialized knowledge that has served as a basis for their lectures. The EMD Lectures have in general been a review of the development of a certain aspect of the extractive metallurgical industry or a preview of a particular area of science or technology that has applications in extractive metallurgy. The usual format is to describe developments of the past, summarize the present state of knowledge, and make predictions for the future. There has been great variation in the relative amount of time devoted to these three subdivisions, the past, the present, and the future. For example, Professor Gaudin, the 1961 EMD Lecturer, in speaking of Separation Engineering on the Moon, did not have too many examples of past practice and so had to emphasize the future. It is cheering to observe that all of these authorities were optimistic in their predictions. Despite the fact that four of the previous EMD Lecturers have been educators, Professors Gaudin, Pidgeon, Schuhmann, and Kellogg, none of them discussed the subject of education for extractive metallurgists. I might say parenthetically that lecturers in other divisions, such as B. M. Larsen, J. B. Austin, F. D. Richardson, John Chipman, and R. F. Mehl, have commented with concern on the present state of training for extractive metallurgy. What can you say that is new about extractive metallurgy education? Certainly nothing new can be said about the educational process itself. Professional educators do have new approaches to teaching and learn-

Jan 1, 1968

By C. Graeme-Barber, R. L. Bell, T. G. Langdon

An analysis is made of the formulas and ,methods used to estimate the contribution of grain boundary sliding to the overall strain of a polycrystal. The errors involved in the approximations and/or ,mistakes in earlier work are assessed quantitatively so as to make possible the critical use of previously published data. SINCE it has been found that, under conditions of very slow creep, the strain due to grain boundary sliding (?gb) may contribute as much as 80 pct of the total creep strain,1,2 it is obviously important to have reliable methods for estimating this grain boundary strain. In two recent papers Stevens3,4 has drawn attention to the pitfalls encountered in assessing ?gb. He criticizes 1) the assumptions made in the derivation of some of the formulas for ?gb, and 2) the indiscriminate averaging procedures used for obtaining the quantities substituted in these formulas. These and further criticisms of earlier work are mostly valid. Unfortunately the only formulas and procedures recommended are either impossible or very difficult to apply in practice. Furthermore no attempt is made to assess the errors involved by the earlier approximations or mistakes, and so it is not possible to make critical use of any of these early dab. The present article attempts to improve on these shortcomings by reference to empirical verification of easily applied formulas whose theoretical derivations involve some otherwise unjustifiable assumptions, and by a quantitative assessment of the errors in previous work. DEFINITIONS First it is necessary to define the terms and symbols to be used. The choice of symbols preferred here is a logical combination and extension of the v introduced by McLean5 and the u by Brunner and rant.' In Fig. 1 the two grains X and Y are displaced by the sliding vector AC. u is the component of sliding resolved along the stress axis, v is that measured perpendicular both to the stress axis and to the specimen surface, and w is that measured perpendicular to the stress axis but in the plane of the surface. Two angles define the orientation of the grain boundary: 8, between the stress axis and the surface trace of the boundary, and +, the internal angle on a longitudinal section cut perpendicular to the surface. In computing the strain resulting from the boundary displacement at all the in- dividual boundaries in a polycrystal the components u, v, or w are sometimes averaged along a longitudinal line (i.e., parallel to the stress axis), sometimes along a transverse line, and sometimes at randomly chosen boundaries. Averages obtained in these three ways are given subscripts 1, t, and r, respectively. Stevens rightly calls attention to the possibility of non-equiaxed grain shapes either at the start or at the end of deformation. In this article all values of numbers of grains per unit length will refer to measurements made before deformation. The subscripts 1 and t will be used for the number of grains per unit length obtained along longitudinal and transverse lines, r for that obtained by averaging along a number of randomly directed lines. FORMULAS FOR ?gb ?gt, as a Function of u. An obvious truth is that cgb can, in principle, be obtained by summation of all the v components at the boundaries intersected by a longitudinal line of known length I, viz.: where nl is the number of grains per unit length, parallel to the stress axis, in the unstrained specimen. There are assumptions in the analytic derivation of Eq. [I] due to Brunner and rant' but ~achinger' arrived at Eq. [la] by a rigorous method. Unfortunately the experimental procedure required by Eq. [ la] proves very difficult as boundary migration tends to obscure the points at which the longitudinal line meets the grain boundary and the errors on such measurements tend to be large, see Fig. 2. For example, on a specimen of Magnox AL 80 at a creep strain of 9.7 pct, 300 measurements of u obtained from the points of intersection of a longitudinal line with the grain boundaries gave ul = 8.95 ± 3.58 pm, or an error of ±40 pct, at the 95 pct confidence limit. These large errors due to migration may be avoided by measuring u from the separation of the segments

Jan 1, 1968

By I. R. Kramer

The effect of the specimen diameter, d, on the flow stress, cra of polycrystalline aluminunz (99.997) was studied. The increase in the flow stress could be accountedfor by the increase in the surface layer stress, with decreasing specimen diameter. Both , and a, were found to be proportional to For the smaller-dianzeter specimen (< 0.033 in.) at strains less than aboul 0.1, the work hardening of the surface layer was greater than that associated with the bulk of the specimen. At higher strains the work hardening due to the bulk appears to be independent of the specimen diameter. THE increase in the strength of metals with decreasing diameter is well-known; however, an adequate explanation for the cause of the size effect is still lacking. The earliest systematic investigation of size effect appears to be that of Onol who reported that for aluminum monocrystals the resistance to slip at low strains increased as the specimen diameter decreased. A change in the stress-strain curve beyond 0.001 strain was not found. However, Suzuki et a1 .&apos; reported for monocrystals of a brass and copper having diameters in the range of 2 to 0.12 mm that the entire stress-strain curve was raised as the specimen diameter was decreased. The effect of size was most apparent when the diameter of the specimen was less than 0.5 mm. In the discussion of this paper Honey-combe reported a size effect in copper crystals as large as % in. diam. These results are in agreement with those of paterson3 and Garstone et al.4 While the majority of the investigations on size effects was conducted in terms of the variation in the diameter of the specimen, several investigators studied the influence of the specimen geometry. For example, Wu and smoluchowski 5 reported that in aluminum monocrystals the slip system was a function of the specimen dimension in the slip direction. King-man and Green 6 studied the influence of size on the compressive stress-strain relationship of aluminum monocrystals when the ratio of length to diameter was constant. Their specimen diameters ranged from to & in. For specimens oriented for single slip the critical resolved shear stress for the smaller-size specimens increased with decreasing diameter. No effect was observed in the large-size specimens. Specimens having an orientation near the corners of the stereographic triangle did not exhibit a size effect. Apparently, the increase in strength with decrease in the diameter of the specimen is a general phenomenon and has been observed in a brass |T and cadmium as well as in aluminum and copper.&apos; In a series of investigations (for example Ref. lo), it was shown that during deformation a surface layer was formed which imposes a back stress, a,, on the moving dislocations. It is reasonable to predict that this surface layer stress, as, should be a function of the specimen diameter and could possibly account for the flow stress size effect. In fact, experimental evidence will be presented to show that this is the case; i.e., the increase in flow stress with decreasing size is equal to the increase in the surface layer stress, as, with size. In addition, data will be presented on the variation with size of and a* where is the back stress associated with the generation of dislocation obstacles in the bulk of the specimen and a* is the net effective stress acting on the mobile dislocations. A limited investigation was carried out on gold specimens to determine the influence of an oxide film. EXPERIMENTAL PROCEDURE The aluminum specimens were prepared from -in. bar stock (99.997 pct purity). The 0.350- and 0.150-in.-diam specimens were machined directly from the bars while the specimens having a diameter of 0.033, 0.020, and 0.015 in. were prepared by swaging and drawing to 0.04 in. and electropolishing almost to final size. The specimens were prepared with a 2-in. gage length. The specimens were annealed in vacuum (-10-4 Torr) at 350°C for 8 hr. The grain diameter of the specimens in the various specimen diameter groups was 0.08 ± 0.02 mm. Gold specimens of two diameters, 0.14 and 0.03 in., were prepared in a similar way and annealed at 650°C for 8 hr. The grain diameter of the gold specimens was 0.2 mm. After annealing the specimens were electrochemically polished to the final size and tested in an Instron tensile machine at a strain rate, E&apos;, of 10- 3 per min. While it was possible to determine the surface layer stress, a,, in the larger-size specimens by measuring the difference, Aa, between the stress before unloading the specimens and the initial flow stress after removal of the surface layer as outlined in detail in Ref. 10, this method is not applicable for small wires because of the difficulty in obtaining a sufficiently accurate measure of the diameter. The values at the various strains were therefore determined by measuring after the specimen had been annealed at 35°C for 4 hr. It has previously been shown" that the two methods give the same results for a provided that the annealing temperature is low enough to affect only the surface layer and not the dislocation barriers in the bulk of the specimen. For the gold specimens a treatment at 150°C for 16 hr was found to be satisfactory for the determination of by the low-temperature annealing method. EXPERIMENTAL RESULTS Determination of a,, and a,. The stress-strain curves for the various diameter aluminum specimens, plotted in terms of the logarithms of the true stress, and true strain, are given in Fig. 1. These curves represent the average data taken from at least ten specimens at each size. Over the range of strains investigated the curves follow the empirical equation

Jan 1, 1968