Search Documents

By M. S. Burton, J. E. Ritter

The sessile-drop technique was employed to study the effect of atmosphere and alloy additions of titanium and chromium on the surface tension and contact angle of nickel on sapphire substrates at 1500oC. Shear tests were conducted on the solidified sessile drops. The liquid-gas surface tension of pure nickel on sapphire was found to be 1747 ergs per sq cm, with a contact angle of 109.6 deg. The surface tension and contact angle were not affected significantly by the four atmospheres, vacuum, impure argon, pure argon, and pure hydrogen. Small additions of titanium and chromium decreased the contact angle and for the same alloy content, contact angles were lower in pure argon than in pure hydrogen. On the basis of shear-test results , the most probable cause for adhesion in the Ni-Al2O system is the formation of NiO:Al2O3 spinel at the interface; in the titanium alloy-Al2O3 system adhesion is thought to be due to a mechanical interlocking effect; in the chromium alloy-Al2O3 system adhesion is believed to be the result of solid-solution formation involving Cr2O3 and Al2O3 at the interface. In the last few years there has been considerable interest in reinforcing metals with ceramic whiskers. It has been postulated that, by reinforcing nickel with sapphire whiskers, high-temperature strengths su- perior to those of the best current nickel alloys can be achieved.' To fabricate a whisker-metal composite having optimum density and strength, understanding of wettability and adherence in the system is of prime importance. Although there has been some previous research in this a a need for further studies has been cited4-8 to understand better the effects of impurities, alloys additions, and atmosphere on the wetting and bonding of metals to metal oxides. The purpose of this research was to investigate the factors involved in the wetting and adherence of nickel/nickel alloys to sapphire (single-crystal a Al2O3). Sessile-drop experiments were conducted in vacuum, impure argon, pure argon, and pure hydrogen, to determine the effect of atmosphere on the contact angle of the metals on sapphire, and the liquid-gas surface tension of pure nickel. The solidified sessile drops were sheared from the sapphire substrate to measure the bond strength between the metal and ceramic. The alloy additions chosen were titanium and chromium since both enhance the wetting of nickel on sapphire.2, 3c Also, titanium is commonly used when brazing metal-to-ceramic seals to improve adherence. Since chromium forms an oxide that is completely soluble in sapphire, it was thought that the oxide could serve as a transition phase to promote bonding between the Ni-Cr alloy and sapphire. I) EXPERIMENTAL APPARATUS AND PROCEDURE The sessile-drop apparatus was a horizontal furnace, with an induction-heated, cylindrical molybdenum susceptor in a water-cooled Pyrex chamber. This apparatus was patterned after that of Williams and Nielsen.7 A flat Pyrex window at each end of the furnace allowed temperature measurements to be made at one

Jan 1, 1968

By R. C. Padfield

Experience over a 3-year period in Bethlehem Steel Corporation's plants has demonstrated the reliability of open-hearth roofs of bonded sprung-arch constructzon with burned basic brick. The design principles lor constructing these roofs include a minimum hot-strength requirement for the basic brick, expansion allowances that extend the full roof thickness, structural members to control arch contour, and a specified minimum roof rise. The greater stability of bonded roofs is explained in terms of the basic stress patterns of ring constrution and bonded construction. PRIOR to the development of successful sprung-arch roofs of basic brick, the majority of open-hearth furnaces in the United States were operated with sprung-arch roofs built of silica brick. Although many silica roofs used on open-hearth furnaces were ring-arch construction, Bethlehem Steel Corp. used bonded-arch construction because of its greater stability. In ring construction, each ring of brick is separately keyed and comprises an independent arch with the straight joints between rings traverse to the longitudinal axis of the furnace. In bonded construction, the bricks are laid in rows starting from the skewbacks so that the straight joints run parallel to the longitudinal axis of the furnace. Each brick in a given row is laid so that it spans the joint between two bricks in the row beneath it. Thus, the transverse joints across the arch are broken and the arch rings are thereby interlocked or bonded. When basic roofs were first being developed, the basic brick that were available had low hot strength. Such brick could not be safely used in sprung-arch construction without some means of suspending them. With the development of higher firing techniques by brick manufacturers and the recently introduced direct bonded bricks with high hot strength, the use of burned basic brick in sprung-arch roofs became feasible. The availability of high hot strength basic brick coupled with the potentially lower cost and proven stability of bonded construction prompted Bethlehem's Research Department to study the possibility of using basic brick in bonded roofs. With the full cooperation of plant ceramic engineers and open-hearth superintendents, particularly in 3 years of fur-nace trials, we developed the design criteria for bonded roofs and the corresponding property requirements for the basic brick that are discussed in this paper. DESIGN PRINCIPLES OF SPRUNG-ARCH BRICK ROOFS Stresses in Fixed Arches. A sprung-arch open-hearth furnace roof is generally built on rigidly held skewbacks. The constraint of the fixed support at each end adds a bending moment to the horizontal and vertical reactions at the ends of the arch. Fig. 11 shows the positive direction of forces acting on an arch fixed at both ends. Fixed arches can be analyzed when the members are continuous and have elastic properties. However, brick are inelastic, and arches built with individual brick segments cannot carry tensile stresses. Therefore, for practical solution of brick arches, empirical formulas have been derived from elastic theory that place design restrictions on arch dimensions to avoid development of tensile stresses. McDowell2 cites three main conditions for stability in sprung brick arches: 1) the thrust line of the arch should be maintained in the middle third of the thickness to avoid tensile stresses and resulting open joints in inner and outer curves of the arch; 2) the angle between the line of thrust at any joint and a line perpendicular to the joint must not exceed the angle of repose between brick; and 3) the maximum pressure at any point must not exceed the strength of the arch materials at furnace operating temperatures. The first and third conditions are particularly important in designing sprung-arch basic roofs because of the comparatively low hot strength of basic brick. According to McDowell's equation, if the thrust line is maintained within the middle third of the arch thickness, the unit pressure is obtained as follows: where p = unit pressure in psi, F, = resultant thrust normal to skewback in pounds per foot, t = arch thickness in inches, and z = distance in inches of thrust line from arch axis. When the resultant thrust normal to the skewback acts along the arch axis, z equals zero and unit pressure is simply the thrust divided by the cross-sectional area. If the thrust line moves to the limits of the middle third of the arch thickness, beyond which tensile forces would develop, z then equals one-sixth of the arch thickness and the unit pressure is double that when the thrust line is acting

Jan 1, 1968

By Harold M. Feder, Robert V. Schablaske, Irving Johnson, Ewald Veleckis

The stoichiometry, structure, and stability of the internzediate phases formed between zinc and some of the rare earth (RE) metals were systematically exarnined by means of a recording effusion balance and X-ray diffraction analyses. In the La-, Ce-, PY-, Nd-, and Y-Zu systems, at or below about 600 C, the following sequences of phases (REZnx) were found: La, x = 1, 2, 4.0, 5.25, 7.3, 17/2, 11, and 13.0;' Ce, x = 1, 2, 3, 11/3, 4.3, 5.25, 7.0, 17/2,* and 11; Pr,x = 1,2, 3, 11/3 ,* 4.3, 5.3(?), 7.0, 17/2,* and 11; Nd,x = 1,2, 3,* 11/3,* 4.3, 6.5, 8.5,* and 11; Y,x = 1,2,3, 11/3, 4.5, 5.0, 17/2,* and 12.* The structure types of all these phases were classified. In addition, lattice parameters were obtained for the first time for the pluses denoted by asterisks. In the absence of de tectable valency or electronegativity effects the systesnatic trends in the results have been ascribed to the effects of' the lanthanide contraction. For example, the maximum number of zinc atoms in the coordination polyhedron surrounding the RE atom decreases from twenty-four to twenty-two to twenty as the size of the RE atom decreases. THE structures and compositions of a great many intermetallic phases (e.g., the Laves phases) are known to be based primarily, but not exclusively, on the space-filling efficiency of various modes of packing together atoms of different sizes. The valencies and electronegativities of the constituent atoms are, however, also influential. In extreme cases hypothetical intermetallic phases which fulfill the efficient spacefilling requirements may not be present in the constitutional diagram because of thermodynamic instability brought about by the operation of valency or electronegativity factors. Hence, for a detailed study of the influence of atomic size on alloy structure and composition, it would be desirable to minimize variations of valency and electronegativity. The intermetallic phases formed by the rave earths (RE) with some common partner offer an excellent opportunity for isolating the effects of size from those of valency and electronegativity. The rare earths exhibit a large, but smooth, decrease in size (the lanthanide contraction) in the series from lanthanum to lutetium when inter comparison is made for a common valence state, e.g., isolated atoms or trivalent ions. The elements yttrium and scandium are frequently included as pseudo rare earths; their sizes place them in the vicinity of dysprosium and lutetium, respectively. The electronegativities of RE elements vary by less than 10 pet. The trivalent state is the most common; however, the well-known tendency of cerium, praseodymium, and terbium to achieve higher valencies, and of samarium, europium, and ytterbium to seek lower valencies, requires that caution be exercised in the assumption of equal valencies. In the present study the existence, constitution, and structure of each of the numerous intermediate phases formed by zinc with lanthanum, cerium, praseodymium, neodymium, or yttrium were examined systematically and in detail. The investigation was conducted by a recording effusion balance technique and by X-ray diffraction analysis. The results enrich our knowledge of the phase diagrams of these systems. In addition, they present some clear-cut evidences of the operation of the size factor alone. EXPERIMENTAL PROCEDURE Apparatus. The mode of operation of the recording effusion balance and its application to phase studies have been discussed in detail elsewhere.' In this work, an effusion cell containing a finely divided alloy was suspended within an evacuated tube from the beam of an analytical balance. The tube was immersed in a massive molten salt bath whose temperature was controlled to within 0.5o C during each experiment. The loss in weight of the alloy owing to effusion of zinc* was continuously recorded. Two effusion cells, 1/2 in. diam by 1 in. high, were machined from tantalum rods. Two orifices were drilled laterally into the walls of each cell. The orifice areas were determined by calibration with pure zinc: cell A had a total orifice area of 6.5 x 10-41 sq cm, and cell B an orifice area of 9.8 x 10-3 sq cm. By appropriate choices of orifice area and temperature the wide range of volatilities from pure zinc to pure rare earth metal could be investigated. X-ray diffraction powder photographs were made at room temperature with a 114.6-mm Debye-Scherrer camera with both filtered CuKa radiation and filtered CrKa radiation. Lattice parameters were refined by a computer-programmed least-squares analytical treatment which incorporated appropriate extrapolation techniques.2 Frequent use was also made of a special computer program3 designed to generate a powder pattern from an assumed structure in order to verify structural assignments. Materials. Lanthanum, neodymium, and yttrium were purchased from the Lunex Co., cerium from the Cerium Metals Corp., and praseodymium from the St.

Jan 1, 1968

By R. D. Pehlke, J. V. Gluck

A study was made of the effect of small additions of metallic solutes on the thermodynamic activity of zinc in dilute solution with molten bismuth in the range 450o to 650°C. The experimental measurements were made in a multielectrode galvanic cell using. a fused LiCl-KCl electrolyte. The activity of zinc in binary solution exhibited a strong positive deviation from Raoult's law and obeyed Henry 's law to at least 0.050 mole fraction. The ternary addition of lead, gallium, or indium increased the activity of zinc. Cadmium or tin additions slightly decreased the zinc activity. Copper, mercury, silver, or antimony caused moderate decreases, while gold strongly decreased the zinc activity. The results are expressed in terms of Wagner's interaction parameter ejZn and a second-order cross-interaction parameter. The interactions for copper, gold, silver, and antimony showed a linear dependence with reciprocal absolute temperature, while the other interactions were essentially independent of temperature over the experimental range. Since tin and antimony decreased the zinc activity, the hypothesis of a purely periodic variation with atomic number was rejected. A semiquantitative explanation was possible with a free-electron model when relative electronegativities were used to express changes in the electron/atom ratio caused by the solute addition. Random solution and quasi-chemical models correctly predicted the majority of the interactions for which binary data were available to apply them. A knowledge of the thermodynamic interactions in liquid (and solid) metallic solutions is of considerable theoretical and experimental interest in the formulation of quantitative thermodynamic relations for metallurgical processes. The approaches to defining the interactions may be either phenomenological (i.e., "What mathematical function adequately describes observed behavior and permits extrapolation to new situations?") or theoretical, such as a specific model

Jan 1, 1968

By J. H. Swisher

The solubility of sulfur and rate of solution of sulfur in pure Lron were measured in H2S + H2 and H2S + H2 H2O gas mixtures. The solubility and diffusivity of sulfur at 1410°Care 0.13 pet S and 1.0 x 10-5 sq cm per sec, respectively. The solubility iS the same, but the rate of sulfurization is slower in the presence of H2O in the reacting gas. Under these conditions, the over-all rate is controlled jointly by a slow surface reaction and by solid-state diffusion; the mechanism for the surface reaction has not been identified. KNOWLEDGE of the behavior of sulfur in solid iron is desirable for the metallurgy of such products as free machining steel, where a high sulfur level is required, and inclusion-free high-strength steels, where the sulfur specifications are very low. The present investigation was undertaken to check previously reported values for sulfur solubility and diffusivity in 6 iron, and to study the poisoning effect of chemisorbed oxygen on sulfurization kinetics in H2-H2S-H2O gas mixtures. All of the experiments were performed at 1410°C. The thermodynamic behavior of sulfur in 6 iron was the subject of a paper by Rosenqvist and Dunicz.' The sulfur solubility at 1400" and 1500°C was determined by equilibrating pure iron specimens with H2-H2S gas mixtures. The maximum solubility of sulfur in 6 iron was alsc determined by Barloga, Bock, and parlee2 by reacting iron wires with sulfur in sealed capsules. In another investigation, the diffusion coefficient of sulfur in 6 iron at temperatures up to 1450°C was measured by Seibel.3 The method used was to measure sulfur concentration profiles in diffusion couples containing radioactive sulfur EXPERIMENTAL Apparatus. A vertical resistance furnace wound with molybdenum wire and containing a recrystallized alumina reaction rube was used for the experiments. The hot zone in the furnace was approximately 2 in. long with a temperature variation of ±3oC. The hot zone temperature was automatically controlled to within ±2°C, and the test temperature was measured with a pt/Pt-10 pet Rh thermocouple before and after each experiment. Flow rates of the reacting gases were obtained using capillary flow meters. Materials. The source of H2S in the gas train was a premixed cylinder containing 5 pet H2S in H2. This mixture then was diluted with additional hydrogen and argon. In some experiments, water vapor was introduced by passing hydrogen and argon through a column containing 10 pet anhydrous oxalic acid and 90 pet oxalic acid dihydrate. The vapor pressure of water above this mixture is well-known.4 Argon was used as a diluent to minimize thermal segregation of H2S in the furnace5 and to reach higher H2O:H2 ratios than could be obtained in mixtures of H2 and H2S alone. Argon was purified by passage over copper chips at 350°C and subsequently over anhydrone. Hydrogen was purified by passage over platinized asbestos at 450°C and then over anhydrone. The H2-H2S mixture was purified by passage over platinized asbestos and then over P2O5. The specimen stock was made by melting and vacuum-carbon deoxidizing electrolytic "Plastiron" in a zirconia crucible. The principal impurities are listed in Table I. In some of the equilibrium experiments, six-pass zone-refined iron was used to minimize impurity side effects. This zone-refined iron had a total impurity level of about 25 ppm. Procedure. Specimens were annealed in hydrogen for a period of at least 2 hr at the beginning of each experiment. The specimens were held in the reacting gas for times varying between 10 min and 17 hr, and cooled to room temperature in a water-cooled stainless-steel block at the bottom of the furnace. The pH2S/pH2 ratios reported are those for gas equilibrium at 1410°C. Calculations based on available thermodynamic data8 showed that the only other gaseous8 species that formed in significant amounts in the furnace were S2 and S. Even when water vapor was introduced into the gas mixture, the concentrations of SO2, SO, and so forth, were negligible. The initial partial pressure of H2S was therefore corrected for its partial dissociation to S2 and S in determining the equi-

Jan 1, 1968

By A. E. Wraith, N. J. Petch, J. D. Embury, E. S. Wright

The two most important parameters controlling the fracture behavior of a solid are its intrinsic properties, e.g.,grain size, and the operative stress system. The latter may be modified in laminates by the presence of weak interfaces. This is studied in notch-impact tests on a mode1 system of mild steel laminates containing a variety of interfaces. The effect of these is evaluated in terms of the ductile/cleavage transition. Two laminate geometries are distinguished, here called "crack-arrester " and "crack-divider". In both, cleavage is inhibited. This arises because of relaxation of a state of triaxial tension. In the crack-avrester laminates, cleavage initiated at a notch is confined to the layer containing the notch. In crack-cliuider laminates, a thick specimen behaves as the Sum of a number of thinner ones. Additional benefit may derive from improved intritnsic propevties of- the lanlinate layers arising from greater deformation in their manufacture. It has long been recognized that the two most important parameters controlling the fracture behavior of solids are their intrinsic properties (e.g., grain size friction stress,3 distribution of second phase particles4) and the operative stress system under which fracture occurs. In solids that show a ductile/cleavage transition, cleavage is favored by the presence of a notch. This is because triaxial tensions, generated by the localized plastic constraint at the notch, are operative when fracture occurs.= Anything that suppresses these triaxial tensions will be unfavorable to cleavage. Such suppression may possibly occur in laminates containing weak interfaces and the purpose of the present paper is to explore this possibility. Two basic laminate geometries, here termed "crack-avrester" and "cvack-divider", are examined. They are illustrated in Fig. 1. With the crack-arrester laminates, there is the possibility that, when the fracture crack approaches the interface, this, if weak, may delaminate due to the tensile stress acting parallel to the plane of the crack.= If this happens, energy will be used in delamination, the crack will be completely blunted, and the triaxial tension associated with the crack will be relaxed. To fracture the second portion of the laminate, crack reinitiation will be necessary and, because of the relaxation of the triaxial tension, this reinitiation will occur under conditions of nearly uniaxial tension, which are unfavorable to cleavage. Thus there is the possibility of cleavage suppression in the second and subsequent subunits of a crack-arrester laminate. With the crack-divider geometry, there is again the possibility of delamination at the interfaces. This will divide the crack into a series of cracks propagating through the individual laminate subunits. If these are sufficiently thin, the triaxial tension will be relaxed towards a state of biaxial tension in each of them. Thus, with the crack-divider laminates, there is again the possibility of cleavage inhibition. In the present work, these possibilities are explored using a notched impact test on mild steel laminates bonded with soft solder, silver solder or copper. Even if delamination does not occur, it is still possible that cleavage may be inhibited in laminates. With the crack-arrester geometry, the cleavage crack in the first layer may be blunted and arrested by plastic deformation in the laminate bond, if this is ductile. Partial relaxation of the stress transmitted ahead of the crack into the second layer will then result and this will reduce the significance of this stress in the fracture of the second layer. With the crack-divider geometry, there cannot be much effect in the absence of delamination unless a large amount of energy is absorbed in rupturing the ductile material. EXPERIMENTAL DETAILS The composition of the mild steel (wt pet) was: 0.04 C, 0.29 Mn, 0.01 Si, 0.006 P, 0.008 S. "As-received" plate was annealed for 2 hr at 900°C and slowly cooled to give a grain size of 0.04 mm. The laminates were made by brazing or soldering together mild steel plates 8 by 3 in. by various thicknesses. These were obtained from the annealed plate by machining, so that the intrinsic material properties were .kept constant. Laminates containing two to six steel layers were studied using standard Charpy V-notch specimens cut from the bonded plates. Standard homogeneous specimens from the annealed plate and subsize ones from the laminate components

Jan 1, 1968

By Donald L. Ball

The equilibrium of boron nitride, gaseous nitrogen, and solute boron was investigated in Fe, Fe-C, and Fe-Si rnelts by the Sieverts technique. The free ellergy of dissociation of boron nitride, involving -the implied standard states, was evaluated in the range 1400" to 1600°C as: ?Fo = 38,000 - 23.5T, cal ber mole The solute-solute interaction coefficients eBB, ecB , and eg were determined as 0.058, 0.22, and 0.078, respectively. BORON is an important ingredient in some types of steel and cast iron. It was employed at a particularly high level, 0.01 wt pet, in the as-cast malleable iron "CentraSteel".1 In this material, the boron concentration was critical in order to assure graphite, and avoid primary cementite, in the as-cast structure. One possible function of the boron in this alloy was the nucleation of graphite through the agency of boron nitride, which, conceivably, precipitated from the melt during cooling. A research program was instituted to assess the validity of this proposed mechanism. The first undertaking, reported here, was a determination of the thermodynamic possibility of boron nitride precipitation. Data available in the literature on the activity coefficient of solute boron in iron-base melts are inadequate for the present purpose. They will be reviewed briefly, however, so that comparison can be made, when

Jan 1, 1968

By J. C. Swartz

Measurements of the effect of precipitation stresses on the solubility of cementite (Fe3C) precipitates in a iron are reported. Solubilities were determined from measurements of the Snoek relaxation due to interstitial carbon after quenching from various equilibrium treatments. Stress-free precipitates were obtained by a spheroidizing treatment of vacuum-melted Fe 0.06 wt pet C. Subsequent equilibration treatments were designed to suppress nucleation of other precipitates. The results for stress-free cementite give a heat of solution of 14.8 i 0.2 keal per mole and an entropy change of 1.1 i 0.2 eu in the temperature range 400" to 715°C. Comparison with previous data indicates that the heat of solution increases about 6 keal per mole as temperature increases from 700" to 1000°C. Data on self-stressed cementite in quench-aged specimens of Fe 0.013 wt pet C indicate 1) an apparent heat of solution of 12.1 i 0.4 keal per mole in the range 280" to 690°C, 2) a strain energy of- 1.1 keal per mole Fe3C accompanying precipitation at 280°C, and 3') a gradual decrease in the strain energy of precipitation with increasing equilibration temperature. In com-parison with 2 and 3 the strain energy calculated for an isotropic model is 1.7 keal per mole Fe3C. The lower values obtained from the experiment indicated substantial stress velaxation by dislocation motion. WHEN a supersaturated solution of carbon in a iron is aged at temperatures in the range 200o to 700oC, the carbon clusters by interstitial diffusion and forms cementite (Fe3C) precipitate particles. Unstressed cementite has a volume per iron atom 9 pet higher than that of a iron;' hence, cementite precipitate particles are usually under pressure from the iron matrix. The pressure would tend to increase the solubility of cementite in iron.' This report describes anelastic measurements of the solubilities of self-stressed and stress-free cementite in a iron. Previous determinations of the cementite solubility are summarized in Table I. The method of cementite formation for each reference was studied to ascertain the state of stress of the cementite. The results judged to be for stress-free and stressed cementite are distinguished in the table. In the next section an estimate of the stress effect is obtained which is clearly too small to account for the large difference between the previous results in parts A and B of the table. THEORETICAL Let AGO be the standard free-energy change of the reaction Fe3C (eem.) = 3Fe (a) + C (in a) [1] in the absence of stresses. Since cementite precipitates appear to be rather pure10 and the carbon is so dilute in the iron,8 the cementite and iron of Eq. [ I] for the stress-free case have essentially unit activities and the activity of carbon equals its atom fraction xoC Then the stress-free equilibrium is expressed by RT In XoC = -?GO 121 To first approximation the change in the chemical potential of carbon due to the precipitation stresses equals the elastic strain energy W accompanying precipitation of a mole of cementite.2 Hence, in the presence of the precipitation stresses the equilibrium is described by RT In x sq = -?G° + W [3] where xsC is the mole fraction of dissolved carbon in the stressed case. An approximate value of W can be obtained from analysis of the inclusion problem in elasticity theory. The inclusion is a small portion of the matrix phase which has suffered a transformation while still imbedded in the matrix. In the absence of the constraint of the matrix the transformation would be equivalent to a uniform dilational strain eT. Both the matrix (a) phase and inclusion (0) phase are assumed to be homogeneous, isotropic materials. When the a and 0 phases have the same elastic moduli, which is approximately true for iron and cementite,11 the total strain energy per mole ß is12 2µVß(eT)2(3 +4 µx)-I [4]

Jan 1, 1968

By B. R. Hollebone

The chemical action of molten anhydrous pyridinium chloride (pyridine hydrochloride) on oxy salts and ores of some Group IV and V metals are discussed-in particular zirconium, hafnium, niobium (columbium) , and tantalum. Laboratory-scale experiments are described whose results suggest that this reaction might provide a practicable means of converting ores of these metals to the anhydrous metal chlorides. Experimental results are given which provide some insight into the nature of the reactions and some of the compounds which could be present at intermediate stages. The treatment of ores of the more refractory metals is difficult and expensive and often demands the use of gaseous chlorine, hydrogen fluoride, or other reagents which are, if nothing else, expensive and dangerous to use. This situation is, of course, due to the strong metal-oxygen affinity resulting from the high charge and the small size of the metal ion which combine to produce such high lattice energy in the metal oxide as to defeat standard reduction methods. Thus, the characteristic of these metals which leads to one of their most important properties—corrosion resistance—is the main hurdle in obtaining the metal. Many common methods' for treating ores of zirconium and niobium (columbium), for example, proceed as directly as possible to the preparation of anhydrous halides. These halides are purified by now standard processes of solvent extraction, fractional distillation, and so on. This paper reports on preliminary stages in the development of a method of treating these ores—particularly of niobium—so as to prepare the anhydrous, volatile halides by the use of chemical reagents which are much more easily handled than those used at present. CHEMICAL PRINCIPLES The solvent action of aminium halides has been referred to periodically in the literature over several decades. The chemistry of pyridinium chloride has been discussed particularly by Audreith2 and Starke.3 However, since this is not familiar chemistry in its nonaqueous setting, it is perhaps well to point out some general metallurgical principles and applications. One can use, as point of departure, the fact that HC1 dissolved in pyridine is an acid very much the same as aqueous HC1. It thus gives essentially all the reactions of the HC1 with which we are familiar. However, when the ratio of HC1:pyridine reaches 1:l (a ratio far higher than is possible with H2O) the solution becomes solid C5F5NHC1, pyridinium chloride.* Inter- estingly, the acidic properties of HC1 persist in this medium, although they are manifest only in the molten state of the compound. Some typical reactions of divalent metal compounds which we have carried out in our laboratories are the following: 2PyHCl + Zn — (ZnCl2) + H2 + 2Py 2PyHCl + MnO — (MnCl2) + H2O + 2Py 2PyHCl + CuS — (CuCl2) + H2S + 2Py 2PyHCl + CaCO3 — (CaCl2) + H2O + CO2 + 2Py In these equations, the parentheses indicate a solvated compound which in water would be predominantly a hydra ted cation but which in molten PyHCl will more likely be a complex chloro-anion of the type MC14-. Some important physical properties of pyridinium chloride (PyHCl) are given in Table I. One sees from the data in Table I that at the temperature of molten pyridinium chloride both water and pyridine will boil away. Thus the system will be kept anhydrous. For this reason the reaction 2NbCl5 + 5H2O— Nb2O5 + 10HC1 has no counterpart in molten PyHCl. Instead, the chloride is made even more stable by the formation of a hexachloro complex: Nb2O5 + 12PyHC1 — 2PyHNbC16 + 5H2O + 10Py One interesting property of these complexes is their thermal instability which for the case of niobium can be illustrated as PyHNbCl6 — PyHCl + NbCl5 With a view to possible use of this chemistry in ore treatment, we have attempted to dissolve Pyrochlore (essentially FeO - Nb2O5) and recover the NbC15 from it.

Jan 1, 1968

By W. R. Bacon

"MAGNETITE HAS BEEN found in many .parts of the coastal area of British Columbia, but the more important deposits are on Vancouver Island and Texada Island. The locations of these deposits are shown on Figure 1.Interest in the magnetite deposits dates back to the 1870s, when land was acquired on the southwest coast of Texada Island by the Puget Sound Iron Company. This ground included the Prescott, Paxton, and Lake deposits. During the years 1885 to 1903, and 1908, 28,898 tons of magnetite ore was shipped to the company's blast-furnace plant at Irondale, Washington, U.S.A. This production was almost entirely from the Prescott open-pits.In the years following the Texada Island discoveries, other magnetite deposits were found in the coastal area, and various plans were proposed to utilize the local ores in the establishment of a steel industry in the Pacific Northwest. None was carried to fruition but, in 1951, open-pit mining was begun at Iron Hill, Vancouver Island, to furnish iron concentrates for steel mills in Japan. In 1952 mining of the Texada deposits was resumed to supply the same market. Production figures for these recent operations are as follows:"

Jan 1, 1949

By Carroll C. Bailey, William F. Boericke

Mine financing, or providing adequate capital for developing and bringing a mining property into production, is an essential requirement for a successful operation. Today it presents a different picture than even twenty-five years ago. The day appears to be gone when a mining property could pay its own way from grass roots to production. The country has been quite thoroughly combed for surface showings from which sufficient initial profit could be immediately garnered to pay for further development and production facilities. The four M's for success in mining-mineral, market, management, and money-are still essential, but doubled or tripled wage costs and steep advances in cost of equipment and supplies throw increased emphasis on adequate capital funds for converting a prospect into a producer. Consequently, the old picturesque, although seldom rewarding, grub- staking procedure where a single venturesome citizen might advance funds to a prospector on a fifty-fifty basis for whatever discovery he might find, has largely disappeared. Its place, to some degree, has been taken by the small syndicate of venturesome members who band together for furnishing funds to a group of prospectors for the same general purposes. Such a syndicate has a definite organization and is usually assured of sufficient initial funds, or flow of funds, to carry on work for several years. It is usually headed by an engineer-promoter or an entrepreneur trader who has had considerable field experience as well as personal knowledge of his prospecting crews. For the owner of a potential mining property who has ample capital, the problem is to determine (1) justification for the necessary investment in exploration and development, and (2) the economics of the enterprise as to size and degree of integration of the ha1 operating entity. If the owner of such a property has little or no capital, he also faces these same basic problems but, in addition, must be able to present the findings of fact and persuade the prospective investor that the enterprise will produce

Jan 1, 1964

By Jameel Menashi

A process has been developed for recovering the refractory metal values contained in mixed secondary source materials such as contaminated grinding wastes, oxides and sludges. The wastes are calcined with sodium carbonate in air to convert the refractory metals to water soluble Na2CrO4, Na3V04, Na2Mo04 and Na2W04. The base metals, Fe, Ni, Co and Cu, are converted to water insoluble oxides. An aqueous leach of the calcine effectively separates the refractory metals from the insoluble base metals. Co, Ni and Cu can be recovered from the oxidic residue by conventional processing methods. The leach solution is purified and then treated to recover the refractory metals as Cr203, NH4VO3, (NH4)6Mo7024?4H2O and H2WO4. The purity of the products is sufficient for their use in many metallurgical applications. The Na+ values in the leach solution are recovered as Na2CO3(or NaHC03) and are recycled to the calcination step. It is shown that the process can accommodate a wide range of feed compositions and generates a minimum of aqueous effluents.

Jan 1, 1984

By R. L. Schmidt, J. D. Verhoeven

The diffusion coefficients and electrotralzsport mobilities of aluminum, gallium, and arsenic have been determined in molten germanium with the capillary reservoir technique. The diffusion coefficients are slightly higher than those determined by the solidification technique of Brirton et a1.1 It was found that aluminium and gallium migrate to the cathode in molten germanium) whereas arsenic migrates to the anode under an electric field. A possible technique for producing p-n junctions by means of electric field freezing is discussed in view of the experimental data. In the past decade or so the literature has abounded with articles concerning the solid-state properties of the semiconducting elements germanium and silicon. It is surprising by comparison how little has been published concerning the liquid-state properties of these materials. In the production of doped germanium or silicon the diffusion coefficient of the dopant in liquid germanium or silicon is an important quantity. Yet, to the best of our knowledge, no one has ever made a direct measurement of the diffusion coefficient of solutes in either molten germanium or silicon. A number of authors'-4 have determined some diffusion coefficients by the indirect technique developed by Burton et al.1 This technique appears to give numbers of the correct order of magnitude but it does depend upon a number of assumptions in the analysis and also requires prior knowledge of the viscosity of the melt. In the present work a more direct measure of the diffusion coefficients of arsenic, gallium: and aluminum in molten germanium has been made. It has been shown independently by two groups of authors5, 6 that, when an electric current is passed through the interface of a solidifying metal containing an impurity, the resulting electrotransport in the liquid boundary layer at the interface offers a mechanism for control over the redistribution of the impurity accompanying the solidification. These authors show that the degree of control is proportional to the mobility, U of the impurity atoms in the liquid boundary layer, where the mobility is defined as the drift velocity relative to the solvent per unit electric field. In principle the results apply to the solidification of a doped semiconductor and offer an interesting technique for control of the doping of a semiconductor. In the present work the mobilities of aluminum, gallium, and arsenic were determined in molten germanium in order to evaluate the potential usefulness of electro-transport as a technique for solute control when applied to the solidification of doped germanium. EXPERIMENTAL PROCEDURE Both the diffusion coefficients and the mobilities were measured using high-purity alumina capillaries and a resistivity technique for analysis. A schematic diagram of the electrotransport apparatus is shown in Fig. 1 and the details of the cell design are shown in Fig. 2. For the mobility experiments electrical contact with the bottom of the capillary was made with graphite, as shown in Fig. 2. In order to prevent electrical contact with the melt the graphite was coated with a thin layer of alumina using a plasma spray torch. The cell assembly, a graphite electrode, and another assembly containing two blank capillaries were fastened to the lavite disk shown in Fig. 1. To make a run the system was evacuated and the doped germanium was outgassed at the operating temperature of 1030°C. The cell and blank capillaries were then submerged beneath the melt and a slightly positive pressure of helium was introduced to fill them. After passing current through the cell for a given time the cell and blank capillaries were raised from the melt and the furnace was shut off. The temperature of the cells dropped below the freezing point of germanium within 1 to 2 min after the furnace was shut off.

Jan 1, 1968

By W. A. Backofen, D. L. Holt

Push-pull fatigue tests have been carried out at 4.2°K, 77oK, and room temperature on two poly crystalline materials of widely different stacking-fault energy (?): pure copper (? - 70 ergs per sq cm) and the Cu-8 wt pct A1 alloy (? - 2.8 ergs per sq cm). Constant stress-amplilude was imposed and measurement was made of the plastic-strain amplitude (ep) at saturation. Lives extended from 104 to 106 cycles. Designating lives at the various temperatures by NRT, N77, and N4.2. the ratios N77/NNT and N4.2/N77 ranged from 3.5 to 18 under the condition of common Ep . Metallo-graphic examination revealed different crack morphology in Cu-8 Al fatigued at room temperature, and at 77" and 4.2oK. At room temperature, cracks lay in or near grain and lain boundavies; at 77o and 4.2oK. cvacks were transcrystalline. Tests on single crystals of Cu-8 A1 showed that such a change in the cracking mode in polycrystallitle material accounted for a factor of- about 3.25 in N77/NRT . The longer life at lower tewperatztre (conslant cp) has heels attributed to two deuelopinents: a reduced production of the dislocation tangles and subgrain boundaries which serve as paths of rapid cracking, and suppression of oxygen chetni-sorption at the crack tip It was concluded that in both materials the luller accounted for an extension of the life at 4.2oK beyond that at room temperature by a factor of 15. XV ECENT experiments on the fatigue of Cu-A1 alloys in the so-called high-cycle range (greater than lo4 cycles) have emphasized the importance of stacking-fault energy (y) as a quantity affecting crack propagation rate and fatigue life.1,2 It was found in comparisons at essentially fixed plastic-strain amplitude that crack growth rate decreased by a factor of about 5 over the composition range from copper (? - 70 ergs per sq cm) to Cu-8 wt pct Al (? - 2.8 ergs per sq cm). The argument was made that, when stacking-fault energy is high, cross slip and climb are favored, so that dislocation tangles and/or subgrain boundaries form more readily under cyclic loading. Since the boundaries and tangles act as paths of rapid crack propagation ,3, 4 life is shortened as a result. However, when stacking-fault energy is reduced (as by alloying), cross slip and climb become more difficult, with the result that substructure formation is retarded and growth rate is also reduced. A purpose of the present work was to investigate the substructure effect in relation to temperature. As temperature is lowered, ? is varied only slightly (if at all), but decreased thermal activation can interfere with cross slip and climb. Thus substructure formation could be curtailed and life increased. Fatigue life in the high-cycle range is also known to be strongly influenced by environment. Working with copper, Wadsworth and Hutchings observed that life in a vacuum of 10-8 mm Hg exceeded life in air by a factor of 20.5 They isolated oxygen as the agent that furthered cracking. While the details are still unclear, a requirement in any mechanism of oxygen-accelerated cracking is that there be chemisorption at the crack tip. That could prevent welding on the compression half cycle,= interfere with reversal of slip,1, 6 or aid in breaking metal-metal bonds at the crack tip.5'7 In the work being reported here, temperature was lowered by immersion in liquid nitrogen and helium, which also served to reduce both the oxygen concentration and chemisorption rate. A possible effect upon life, i.e., a lengthening, had to be recognized. Several researchers have determined fatigue lives at low temperatures presenting their results in the form of stress amplitude (S) vs cycles in life (N) curves.8-11 Such curves reflect, primarily, the fact that metal is strengthened by lowering temperature; effects of substructure and changing environment tend to be masked. The difficulty can be overcome by comparisons based on identical plastic-strain amplitudes, and in the present work the dependence of life on both plastic strain and stress amplitude was established. EXPERIMENTAL Materials. The principal materials were polycrystal-line copper (? - 70 ergs per sq cm)" and the Cu-8 wt pct Al alloy (? - 2.8 ergs per sq cm),I3 the latter being near the limit of solubility of aluminum in copper and having, therefore, the lowest stacking-fault energy in the CU-Al system. Specimens were machined from 0.118-in.-diam cold-swaged rods of high-purity (99.999 pct) copper and the Cu-8 Al alloy, the latter produced initially in a graphite boat by induction vacuum melting a mixture of 99.999 pct Cu and 99.99 pct Al. The machined specimens were annealed to produce mean grain diameters of about 0.070 mm in copper and 0.190 mm in the alloy. Specimen dimensions are given in Fig. 1. Values of the tensile yield stress, ultimate strength, uniform strain (determined by the Considgre construction), and reduction of area, for both materials at 4.2oK, 77oK, and room temperature, are listed in Table I. The tensile apparatus in which these results were obtained has already been described.14 Apparatus. Specimens were fatigued in push-pull with a machine that is illustrated schematically in Fig. 2. The specimen is first soldered into the top grip (1) with Woods metal, and the grip is then screwed into the inner tube (2) which is connected to the drive rod of the Goodmans vibration genera-

Jan 1, 1968

By T. F. Bower, M. C. Flemings

The classical explanation for formation of the fine equiaxed "chill zone " in ingot solidification is that local supercooling near the chill wall causes copious nucleation in this region. This work indicates, however, that convection is an essential element in formation of the chill zone. In the absence of convection, no chill zone forms even though rate of. heat extraction is extremely vapid. Alloy studied was Al-l.5 pct Cu alloy solidified against a copper chill wall. When solidification was initiated at the chill wall in absence of. cor-vection, surface grain size was large (>I cm diam). When it was initiated in the presence of convection, it was fine (<0.05 cm diam). It is concluded that some form of "crystal multiplication" plays an important role in formation of the chill zone. INGOT macrostructures are generally classed as containing one or more of the three zones: "chill zone", "columnar zone", and central "equiaxed zone". These three zones are sketched in Fig. 1; their origin has been discussed in some detail.&apos;-&apos; The classical explanation for formation of the chill zone is that supercooling occurs in the region of liquid metal adjacent the mold wall, and copious nu-cleation then takes place in this region forming numerous small crystals. The columnar region then results by growth of favorably oriented grains (den-drites) which crowd out their less favorably oriented neighbors."5 The mechanism of the columnar-equiaxed transition has been the subject of considerable interest. The generally accepted mechanism, until recently, was that constitutional supercooling developed in front of the growing columnar grains with eventual nucleation of new, equiaxed, grains ahead of the growing solid front.&apos; Recently, it became evident this explanation was not completely satisfactory and it is now evident that "crystal multiplication" plays an important role. This crystal multiplication occurs by separation of dendrite arms from growing dendrites,8?"&apos; and has been shown to be strongly influenced by convection.6, 10-13 Reduced convection results, for a given alloy, in longer columnar zone, larger columnar grains, and larger equiaxed grains.7, 12

Jan 1, 1968

By J. L. Walter

Tungsten ingots with and without small amounts of aluminum, silicon, and potassium were prepared by conventional powder-metallurgy techniques and hot-rolled to 0.001-in.-thick sheet. The ingots were rolled three different ways: I) initially cross-rolled, then straight-rolled; 2) straight-rolled, and 3) straig-ht-rolled, then cross-rolled. Total reductions were 99.8 pct. Samples of the sheet were annealed, in vacuo, at temperatures from 1200" to 2200°C. Substructures and textures were determined by transmission electron microscopy and X-ray diffraction. The doping agents were found to form particles of mullite (Al,Si,O,,) which, on rolling., appear to form strings of particles aligned parallel to the rolling direction. These strings of purticles are associated with an increase in the recrystal-lizution temperature from 1400°C for undoped sheet ThE addition of small amounts of aluminum, silicon, and potassium to tungsten has a profound effect on the grain structure of drawn and recrystallized wire.&apos;" The grains generally do not occupy the entire cross section of the wire and have lengths many times the wire diameter. The grain boundaries are, in general, nearly parallel to the wire axis and have long &apos;&apos;fingers" projecting one into the other as may be seen in Fig. 1, a longitudinal section of recrystallized doped wire. When unsuitable additions or no additions are made or the wire is not sufficiently reduced during draw- to 2000°C for doped sheet. The particles also limit lateral grain growth while allowing growth parallel to the rolling direction. This results in the formation by primary re crystallization of large highly elongated grains of specific preferred orientation. The undoped sheet recrystallized to small equiaxed gains. The preferred orientations of the large grains were determined from axis density plots to be (113) (691). (0i3) (362), (116) (304) types for the initially cross-rolled then straight rolled material. These orientations are similar to the orientations of the large elongated grains in doped and heavily drawn tungsten wire. The nucleation of the large grains appears to occur by recovery of certain of the polygonized regions of specific orientation, followed by strain-induced boundary migration.

Jan 1, 1968

By W. J. Babyak

The habit planes for hydride precipitation in large grains of alpha zirconium and Zircaloy-4 were determined in specimens containing 161 and 136 ppm hydrogen, respectively. In zirconium, the hydride platelets precipitated on various pyramidal planes inclined to the basal plane at angles ranging from 10 to 50 deg and 75 to 90 deg. Hydrides were most frequent at inclinations of about 20 deg and 90 deg. In Zircaloy-4, hydride plates were inclined to the basal plane at angles vanging from 5 to 25 deg with the predominant inclinatioiz equal to 12 deg; no platelets were found near the prism plane position. In both zircoinium and Zir-caloy-4, hydrides were principally, but not exclusively, parallel to (1011) type planes. When the Zircaloy-4 gmins were subjected to tension stresses of 12,100 psi and 19,400 psi during hydride precipitation, the hydride habit was unchanged except that precipitation favored those habit planes which were inore nearly perpendicular to the stress direction. The stress-orientatioiz effect increased with the magnitude of the stress. THE hydride habit plane in a zirconium has been determined by several investigators. Langeron and Lehr1 in 1956 found platelets parallel to (1010) slip planes. In 1962, Westlake and Fisher2 confirmed these findings and showed that hydride precipitation on twin planes, as reported by Kunz and Bibb,3 was probably due to incomplete removal of worked metal from specimen surfaces. Secondary habit planes, ( 1011) and (1015), were also found by Westlake and Fisher under particular charging conditions where a hydride layer may have existed at the hydrogenation temperature. Most recently, walters4 and westlake5 again observed the (1010) habit exclusively in zirconium. Louthan and hgermans have reported that hydrides precipitate on various ( 101l ) and (112l) planes positioned 31.5 to 90 deg to the basal plane in Zircaloy-2. The present investigation was concerned with the hydride habit in Zircaloy-4 in relation to that in zirconium and the possible change in habit when precipitation occurs under stress. The latter consideration was prompted by the stress-orientation phenomenon originally reported by Marshall and Louthan.7, 8 I) EXPERIMENTAL Materials and Specimens. Chemical analyses of all materials are given in Table I. The zirconium specimen was a 1 by 4 in. section cut from wrought crystal bar that was hot-rolled at 790°C to 0.300 in. and cold-rolled to a final thickness of 0.095 in. All surfaces were mechanically polished and then chemically polished with a solution of H2O, HNO3, and HF (10:10:l by volume) prior to annealing. Grains as long as 0.25 in. were grown by thermally cycling the specimen three times (in 10-5 mm Hg vacuum) between 825" and 1000°C at approximately 125°C per hr. The specimen was isothermally annealed at 825°C for 24, 40, and 120 hr after each cy-

Jan 1, 1968

By K. Mukai, M. Ishihara

This report is a brief discussion of the impurity levels both in primary aluminum and super-purily alnminim in connection with raw materials and proc-essing methods. Particularly, truce amounls of im-purities were analyzed by introducing nittss-speclro-gruphy and nenlron-aclivalian analysis techniques. Vanadium and gallium content in primary aluminum metal corresponded to theor amounls in anode coke. but the nickel content in primary aluminum was hide-pendent oj its amount in the anode cake. Sulfur in the anode colic might result in an increase of the iron content in the molten metal in the reduction cell. The content of sodium and calcium increased with increasing bath ratio. In the holding furnacc sodium content decreased rapidly, and nonmetallic inclusions merit separated front the mollen metal. Three-layer elctrolysis was effectire in eliminating many impuri-ties with the exception of copper, aluminum oxide. aluminum nitride, and some rare-earth elements. THE amount of impurities and nonmetallic inclusions in aluminum metal depends on impurities in the raw materials and various production processes. This paper presents an investigation of the behavior of impurities and nonmetallic inclusions in aluminum during the electrolysis process, the period in the holding furnace, and the process of three-layer electrolysis. Chemical analysis and emission spectrometric analysis have generally been used as methods to determine small amounts of impurities in both primary aluminum and super-purity aluminum; however. mass spectrography and neutron-activation analysis have been recently introduced into this field, and trace amounts of impurities can accurately be determined by these latter methods. EXPERIMENTAL WORK Common impurities were determined by means of absorptiometric. polarographic, flame-photometric. and emission-spectrometric methods. However, in the determination of trace amounts of impurities and non-metallic inclusions, the following methods were applied. In the analyses of trace impurities by mass-spec- trographic analysis, samples were exposed in the vacuum spark of double-focus ing Mattauch-Herzog type mass spectrograph (Japan Electron Optics Laboratory, Ltd.). Ilford Q2 photographic plates were used as ion detectors. Fe was used as the internal standard, and the atomic concentration of each element was calculated by the following expression and converted into the weight concentration: where E, and Ei are the exposures (coulomb) when isotopes of the internal standard and impurity elements (on which the estimate is based) show the same density, X is the percent (atomic) concentration of the standard, IF and Ii are the percent abundances of the isotopes of the standard the impurity elements (on which the estimate is based), and Ms and hfi are the mass of the isotopes of the standard and impurity elements (on which the estimate is based). The coefficient of variation in this method is about 40 pct in the range of 0.1 to 1 ppm. In neutron-activation analysis, irradiations were done in the nuclear reactors TRIGA-I1 (neutron flux: 4.7 x 10 11 n per sq cm per sec; Rikkyo University) or JRR-2 (neutron flux: 2 x 1013 n per sq cm per sec: Japan Atomic Energy Research Institute). After the chemical separation, a NaI (3 by 3 in.) scintillator connected to a 256-channel y-ray spectrometer (R.C.L.) was used for counting and recording various activated impurities. The coefficient of variation in this method is about 20 pct. In the analyses of nonmetallic inclusions, aluminum-oxide determination was carried out according to Fischer&apos;s method, 3 as follows. A slice of sample was cut off and the surface was cleaned. It was dissolved in bromine-methanol. The solution was filtered and the residue was washed. The residue was fused with potassium bisulfate flux. Then it was dissolved in weak acid, and oxine solution was added. The aluminum oxinate was extracted with benzene. Finally, aluminum was determined absorptio-metrically. Aluminum nitride determination was carried out as follows. A slice of sample was cut off and the surface was cleaned. It was put in the distillation apparatus and decomposed with sodium hydroxide solution. The solution was distilled. Then ammonia was absorbed and

Jan 1, 1967

By R. T. Southin

Cast ingots of 99.99 pct purity A1 and aluminum/copper alloys containing up to 2 pct Cu have been found to contain four zones rather than the three previously accepted, i.e., chill, colummar, and equiaxed zones. The fourth zone consists of a coarse dendritic layer across the top surface of the ingot which is open to the atmosphere. It is also found that the grains in the equiaxed region have a "cornet" shape, with a head which is coarsely dendritic and a tail which grew with the same structure as the columnar grains. Experiments using mechanical barriers, and others involving the elimination of the surface dendrile layer, have indicated that nucleation of the equiaxed zone is caused by the showering of dendritic particles from this surface dendrite layer. These particles are the dendritic heads found in the comet grains. A mechanism of nucleation and growth of these comet grains is put forward. Experiments with a number of other metals and alloys suggests that these conclusions may apply generally to ingots cast without the addition of nucleating agents, and solidified under normal conditions. THE macrostructure of cast ingots may contain the following three zones: i) a peripheral zone of fine equiaxed grains, commonly called the chill zone; ii) a zone of columnar grains extending inwards from the chill zone; iii) a central zone of equiaxed grains; these are normally larger than the chill grains. All three zones are not necessarily present in any particular ingot since their existence and magnitude depend upon the casting conditions and composition of the cast material. The mode of formation of zones i and ii is fairly well understood. However, the cause of the central equiaxed zone remains a matter of conjecture. Previous theories put forward to explain the mechanism of nucleation in the equiaxed zone have been examined by Chalmers1 and Winegard.2 At the time when the present work was carried out the most re -cent hypothesis was that due to Chalmers,1 who recognized more fully than previous workers that conditions in this zone during solidification were not conducive to nucleation. In his hypothesis, he suggested that nuclei were formed near the mould wall during the very early stages of solidification, and that they can drift into the central zone by convective motion. This was supported by evidence from experiments using mechanical barriers to stop the drift of nuclei. A similar hypothesis had previously been put forward by Genders.3 More recently, Jackson et al. 4 have observed that dendrite arms in organic compounds can melt off during solidification. They have postulated that this occurs in metals and alloys, and that the dendrite arms which have been freed from the main dendrite act as nuclei for the equiaxed zone. In the present work, critical metallographic examination has shown that there is a fourth zone in the macrostructure of cast ingots which is a dendritic layer over the top surface of the ingot. It is also shown that there is a relationship between this zone and nucleation of the equiaxed zone. EXPERIMENTAL DETAILS The alloys examined were prepared from super -purity aluminum of 99.99 pct min purity (spectro-graphically detected impurities of Cu, Fe, Si, and Mg) and copper of 99.98 pct min purity (spectrographically detected impurities of Si, Zn, As, Fe, Pb, Bi, Ni, Sn, Mn, Ag, Cr, and Al). Three compositions of ingots were examined, containing 0.0, 0.1, and 2 pct Cu, respectively. The alloys were melted in an alumina-lined salamander crucible using a medium-frequency electric induction furnace; when the appropriate casting temperature was reached, the alloys were cast without degassing or fluxing treatment. The mould used was made of electrode graphite and had the following dimensions: external diameter 6 in., height 9 in., with a mould cavity 8 in. deep and a diameter tapering from 5 in. at the top to 4 in. at the bottom. Casting temperatures used were generally in the order of 70" to 90°C above the liquidus and the mould was at room temperature before casting. Inserts kere placed in the mould cavity in some experiments. These consisted of either an 18-g stainless-steel tube of 2 in. diam held axially in the mould i in. off the bottom and extending 1 in. out of the top of the mould or a "trap-door" barrier of stainless-steel gauze of 24-mesh held horizontally in the mould 31/2 in. down from the top. The construction of this insert is shown in Fig. 1. All inserts were preheated prior to casting to prevent chilling. RESULTS The macrostructures of ingots of the three compositions are described, in terms of the three generally recognized zones, in Table I, and illustrated in Figs. 2(a) to (c). The most interesting observation in this table is that relating to the shape of the equiaxed grains, which were not in fact equiaxed but of almost tear-drop shape. Closer examination of these grains showed that each consisted of a head which had a coarse dendritic structure and a tail which always grew in the opposite direction to heat flow, and which had the same structure as that of the columnar material of the ingot concerned; i.e., in the pure aluminum and 0.1 pct Cu alloy it was cellular and in the 2 pct Cu alloy it was cellular-dendritic. Identification of original cast grains was greatly hampered in the super-purity aluminum ingots, and to a lesser degree in the 0.1 pct Cu alloy,

Jan 1, 1968

By D. R. Uhlmann, T. P. Seward, K. A. Jackson, J. D. Hunt

microscopic ohservations on alloys of organic trzaterials show that dendrite arms can melt off under normal conditiorzs of growth. This occurs because of the interactiorz of&apos; heat and matter fluxes during dendritic growth. The re melting occurs where the den-drite arms are joined onto the main deizclrite sterrz. Many detached crystals can be produced in this fashion. It is postulated that this remelting phenornenon occurs in metal alloys, giving rise to the equi-axed region in castings. Experiments peyformed orr castings of transparent materials support remelting as a primary source of detached crystals. Ob~ervations an equiaxed tones in metal castings are discussed in terms of this and other mechanisms. It is concluded that the partial remelting of dendrites is on important mechanisnl for producing the equiaxed regions in castings. THE freezing of a casting can be divided into three regions: a chill zone formed near the mold wall, a columnar region, and a third region, at the center of the casting, which is called the equiaxed zone. The origin of the grains that comprise the equiaxed region is the subject of this paper. We will first outline the mechanisms which have been proposed to account for the origin of the equiaxed region and discuss some recent experiments which cannot be explained by these theories. We will then present some microscopic observations of dendritic growth which have led us to conclude that the equiaxed zone in castings can arise from partial remelting of dendrites. This proposal is supported by observations on transparent castings. Finally, we will discuss several observations on metal castings which could not be explained previously, and which can be accounted for by the present proposal. Previous Mechanisms. Winegard and Chalmers3 pointed out that the molten central region of an ingot could be constitutionally supercooled, because of the solute layer at the growing interface. In this supercooled region, nucleation on particles could occur, giving rise to the equiaxed crystals. Another mechanism has also been suggested,4y5 that the nuclei originate in the chill zone then float or are carried by convection to the center of the casting, where they grow to produce equiaxed crystals. In this case, all the crystals nucleate immediately on casting, so this mechanism will be termed "big-bang" nucleation. Walker&apos;s Experiments. J. L. Walker has made observations on the solidification of nickel and Ni-Cu alloys which cannot be accounted for by either of the above mechanisms. In studying the grain size resulting from freezing of nickel at various undercoolings, walker6 found that at large undercoolings the grain size was small, due to mechanical nucleation, as discussed by walker7 and Horvay.&apos; At small undercoolings, the grain size was large, and the growth dendritic. In Ni-Cu alloys, however, for initial undercoolings starting from just below the melting point and extending for a hundred or so degrees of undercooling, fine grains were also observed. Samples were undercooled 200 Centigrade degrees without nucleating, and were then heated 130 Centigrade degrees so that the undercooling was 70 Centigrade degrees, and nucleated. The samples had a small grain size, indicating multiple nucleation. There were no heterogeneous nuclei in the liquid which could operate at the temperature of growth. Constitutional supercooling could not have been greater than the supercooling of the initial melt, since the temperature rose as the freezing continued. walker -as also shown that a Ni-Cu alloy which can be undercooled to below its solidus will produce typical columnar and equiaxed regions when cast into a cold mold. This alloy could not reach a temperature below its solidus by constitutional supercooling. It could not, therefore, have nucleated heterogeneously. Neither the big-bang mechanism nor the Winegard-Chalmers mechanism can account for the fine grain

Jan 1, 1967