Search Documents

By W. O. Philbrook, K. M. Goldman, G. Derge

T. Rosenqvist—The most interesting point in this paper is the observed transfer of iron into the slag in the initial stage of the desulphurization process, after which the iron again is reduced to the metallic state. The authors interpret this observation as showing that the sulphur enters the slag as an iron-sulphur compound which subsequently is decomposed by the slag. The present writer has previously suggested the following equation for the desulphurization process: S + O2- ? S2- + O For equilibrium in the blast furnace the oxygen potential is defined by equilibrium with graphite and CO of 1 atm pressure: C + O ? CO [2] During the desulphurization process the reactions proceed in the direction of the arrows. If one assumes eq 2 to be significantly slower than eq 1, the transfer of sulphur into the slag, in accordance with eq 1, will build up a local oxygen potential at the metal-slag interface very much higher than that corresponding to the value defined by eq 2. This is possible because the equilibrium oxygen potential in eq 1 is high as long as the sulphur content in the slag is low. This oxygen potential will again be able to oxidize some iron: Fe + O ? Fe2+ + O2- and an increase in the iron content of the slag will be observed. Adding up eqs 1 and 3 one obtains: S + Fe ? S2- + Fe2+ The net effect is thus in harmony with the experimental observation but is obtained without assuming any close ties between the sulphur and iron atoms during the process. Furthermore, it follows from eqs 1 and 2 that when the sulphur content in the slag increases, and equilibrium with C and CO is finally approached, the local oxygen potential at the metal-slag interface will decrease, and the iron in the slag will be reduced back into its metallic state. C. E. Sims-—The data and conclusions presented in this paper are thoroughly convincing in establishing the mechanism of sulphur transfer from iron to slag as in a blast furnace. The evolution of gaseous CO in step 3 of the reactions given on p. 1112 makes the process virtually irreversible. Assuming that the process is similar in slag-metal systems other than in the blast furnace, it is readily seen why free CaO and re-ducing conditions so greatly favor desulphurization. On the other hand, the very effective desulphurization obtained in oxidizing slags when strongly basic, must be attributed to the relatively high stability of CaS as compared to FeS. The ease and simplicity with which the reactions of classic chemistry agree with the experimental data and explain the mechanism is noteworthy. The concept of molecules of FeS, soluble in both phases (metallic iron is not soluble in the slag), migrating from the iron to the slag and there reacting with CaO, which is soluble only in the slag phase, is clear and uncomplicated. This is likewise true for step 3. Those who would deny the existence of molecules or molecular-type combinations in liquid iron, must strain to provide a mechanism so lucid. In the absence of molecules, the Fe and S exhibit a remarkable collusion. L. S. Darken—The investigation and interpretation of rate phenomena in the range of steelmaking temperatures is a difficult task. Most of the laboratory investigations of steelmaking reactions have been concerned with equilibrium. Having determined the equilibrium, our attention naturally focuses next on the mechanism and rate of approach to equilibrium. The authors seem to have contributed substantially to our understanding of these factors for the case of sulphur transfer. I should like to ask the authors whether they consider that the sulphur transfer reaction is diffusion controlled as many high-temperature reactions seem to be. If so, it would seem reasonable to suppose that the slow diffusion step of the process is the transfer across a pseudo-static layer or film similar to that considered in heat flow problems. As the diffusivity and fluidity are smaller for the slag than for the metal, it may tentatively be assumed that the sulphur gradient exists in a thin layer in the slag adjacent to the slag-metal interface and that the metal and the main mass of slag are each maintained uniform by convection. On this basis the amount of sulphur transferred across unit area per unit time is D p (?S%)/100 ?1, where D is the diffusivity, p the density, (?S%) the difference in percent sulphur on the two sides of the layer, and ?l is the layer thickness. At the beginning of the experiment the main body of the slag and hence one side of the layer contains no sulphur; therefore (?S%) may be replaced by (S%), the sulphur content of the slag at the slag-metal interface, which in turn is equal to L[S%] where [S%] is the sulphur content of the metal and L is the distribution coefficient. The rate of transfer thus becomes DpL[S%]/100 ?l, which the authors designate K[S%]. Equating these two quantities and setting D = 10-6 cm2 per sec, p = 3 g per cm3, L = 40, and K = lo-+ g cm-2 sec-1, it is found that ?l, the film thickness, is about 0.01 cm—a value of the order of magnitude of that found in heat transfer problems in liquids. The uncertainty of the numerical values used leaves much to be desired, but at least it can be said that this calculation tends to support the proposed model involving diffusion through a film. Although this does not seem to affect the general argument, I should like to call attention to the fact that the diffusivity3 of sulphur in hot metal is found (on conversion of units) to be about 10-4 cm2 per sec rather than 104 cm2 per sec as stated by the authors. The three equations written by the authors to express the steps in the overall process of sulphur transfer may alternatively be written ionically as only two Fe + S = Fe++ + S-- Fe++ + O-- + C (graphite or metal) = CO (gas) + Fe where the underscore is used to designate the metallic phase; ionic species are slag constituents. After the authors have so neatly demonstrated that iron and sulphur transfer together (at least initially), this fact seems almost self evident; from eq 4 it is seen that if sulphur acquires a negative charge during transfer

Jan 1, 1951

By Lawrence H. Van Vlack, Otto K. Riegger, Gerald I. Madden

The microstructural variations of periclase (MgO) in the presence of oxide liquids are examined under the steelmaking variables of: 1) temperature, 2) liquid composition, and 3) FeO additions under different oxidation levels. Attention is given to the distribution of the phases, both liquid and solid, and to the growth of individual crystalline grains. Silicate liquids penetrate more extensively between individual periclase grains than do liquids containing high percentages of Fe2O3 Higher MgO solubilities in the liquid and lower MgO contents of the solid favor more rapid grain growth. The presence of a second solid phase reduces the periclase grain growth rate and increases the amount of the solid-to-solid contact within the oxide microstructures at high temperatures. The service suitability of a refractory depends on many factors. Two are of major importance and include 1) the thermal resistance to melting, and 2) the mechanical resistance to loads at service temperatures. Neither is a simple consequence of the service temperatures because service conditions will alter compositions, produce partial melting, and induee phase changes. Consequently, the equilibrium phase relationships have been rather thoroughly studied and give a knowledge of the thermal resistance to melting, but do not give full information about the mechanical properties because two refractories with the same types and quantities of phase may have different microstructures. Although variations of microstructures with time, temperature, and composition have been subjects for extensive investigation in metals, only a limited amount of comparable microstructural work has been performed for refractory materials.' This study was an attempt to evaluate some of the consequences of service parameters upon the microstructures of refractories so that bases may be established for the analyzing of high temperature mechanical properties. Periclase (MgO) was chosen as the refractory oxide; variables included those which are encountered under steelmaking conditions such as 1) temperature, 2) liquid composition, and 3) FeO additions under various oxidation levels. Specific attention was given to the distribution of the phases, both liquid and solid, and to the growth of individual crystalline grains. The most closely related work on microstructures of polyphase materials is that of Van Vlack and Rieg-ger2 on the microstructure of magnesiowüstite [(Mg,Fe)O] in the presence of silica. In that work which pertained to solid solutions with less than 40 pct MgO, most of the quantitative work was performed on FeO microstructures. The chief conclusions concerning these relatively low-melting oxide solids were as follows: 1) the rate of crystalline grain growth is inversely proportional to the grain diameter, 2) grain growth proceeds more rapidly at higher temperatures but is slightly retarded by additional liquid content, and 3) a Silicate-containing liquid penetrates as a film between the individual magnesiowüstite grains independent of time, temperature, amount of liquid, or the MgO/ FeO ratio. The above observations are in contrast to prior work3 on the microstructure of silica in the presence of iron oxide-containing liquids where the liquid does not penetrate as a complete film between solid grains. The phase relationships for the compositions of the present work are shown in Fig. 1 which is a summary of the work of several investigators.4 Of importance is the fact that CaO forms a more stable structure with SiO2 and Al2O3 than do either MgO or FeO. The oxygen potential has little effect on periclase unless iron oxide is also present. The iron oxide is ferrous at moderately low oxygen levels, changing to ferric as the Oxygen potential is increased so the spinels, magnetite, and magesioferrite are formed.5 These two phases are relatively stable in air at steelmaking temperatures. I) EXPERIMENTAL PROCEDURE ractory were made with reagent grade Oxides. The magnesium oxide used was 99 pct MgO after ignition, and the iron oxide raw material had a minimum content of 99 pet Fe2O3. The CaO, SiO2, and A12O3 were also reagent grade raw materials. After mixing, the required compositions were pressed into pellets at a minimum pressure of 5000 psi to insure compaction of the raw materials and prevent excess void content. A silicon carbide element tube furnace was used with thermocouple control for sin-

Jan 1, 1963

By H. C. Chao, Y. E. Smith, L. H. Van Vlack

ThE phase relationships for the MnO-MnS system have been investigated only in the eutectic region. wentrupl reported a eutectic at 1280°C (2345°F) with approximately 50 wt pct of each component, as based on the earlier work of Andrew, Maddocks, and Fowler.2 More recently silverman3 corrected these figures to 2275° ±10°F (1246°C) with 63 wt pct MnS and 37 pct MnO. Neither Wentrup nor Silverman investigated the solid-solution limits in this system. However, Wentrup did suggest with his phase diagram sketch that the maximum solubility of MnS in MnO is 10 wt pct, and the solubility of MnO in MnS is nearer 20 wt pct, Fig. 1. These figures have been utilized in subsequent publications without further verification.4 There had been evidence found in this laboratory that the above solid-solution figures were high. An investigation was therefore conducted using the following procedure. Samples containing MnO and MnS were contained in platinum foil, heated just above the eutectic temperature in a purified nitrogen at- mosphere, and quenched into distilled water. The MnO had been prepared by calcining reagent grade MnCO3 in nitrogen at 1400°C. Only MnO was detected in subsequent X-ray diffraction analyses. The MnS had been prepared by sulfur deoxidation of reagent grade MnSO4 according to the reaction: MnSO4 + S2 ? MnS + 2SO2 The details of the sulfide preparation procedure, which utilized atmospheric sulfur pressures at 950°C, are discussed elsewhere.5 The sulfur content of the sulfide was determined to be 36.8 * 0.2 pct, which matches the stoichiometric value of 36.9 pct. No evidence of either MnO or MnSO4 was detected microscopically or by X-ray diffraction. The single-phase combinations of MnO and MnS at the eutectic temperature indicate the limit of solid solution as shown in Fig. 2(a) and (c), and the revised phase diagram is shown in Fig. 2(d). The melting temperature for MnO is that provided by Glasser.6 The eutectic temperature of 1232° ± 5°C and eutectic composition of 64 wt pct MnS are in close agreement with Silverman's correction of Wentrup's work. The maximum solubility of MnO in MnS was determined to be 1.7 ± 0.2 wt pct. The maximum solubility of MnS in MnO was determined to be 1.8 ± 0.2 wt pct. These figures represent the lower limit for the presence of liquid at 1260°C and are significantly lower than the earlier estimated solid-solution figures. The authors wish to acknowledge the support of U.S. Steel's Edgar C. Bain Laboratory for Fundamental Research. The comments and critical suggestions of Drs. A. S. Keh, R. Rickett, and R. B. Snow are appreciated.

Jan 1, 1963

By Clarence E. Sims

An effort has been made to give both a comprehensive and simplified picture of the origin, modes of formation, and characteristics of nonmetallic inclusions in steel. Exogenous inclusions, those formed by mechanical entrainment or heterogeneous reaction, have their principal source in ladle refractories. Mechanical emulsion is not considered a likely mechanism. Endogenous inclusions, resulting from homogeneous reactions, are formed mainly during cooling and freezing. They are principally oxides and sulfides but include some nitrides and carbides. Deoxidation affects both the quantity and character of such oxide inclusions and significantly affects the size and shape of sulfides. The importance of inclusions is largely in their effect on mechanical properties. It is fitting and proper that we should honor the memory of those technical giants who have left such a rich heritage to our profession. Today, according to annual custom, we pay tribute to such a man. To have been chosen to play a part in this observance prompts in me an inordinate pride while I am humble before the responsibilities of my task. The ranks are growing thin of those who had the good fortune to come under the personal influence of Professor Howe, but the number who continue to benefit from his legacy of technical wisdom con-stantly increases. I am among the latter. One has only to read his works to appreciate his unusual ability to take data and observations from many sources and give them a clear and comprehensive

Jan 1, 1960

By D. I. Cameron

A graphical method has been developed and tested for separating the effects of grain boundary and lattice diffusion in polycrystalline materials. The method is based on the assumptions that for unidirectional diffusion from a plane source of radioactive material, the penetration of activity far into the specimen from the source is due solely to grain-boundary effects, and 2) pure grain-boundary penetration is described by a linear plot of log activity vs the perpendicular distance x from the source. The former is based on the fact that at any temperature the grain-boundary diffusion coefficient is always very large relative to the lattice diffusion coefficient; and the latter is based on the derivation of Fisher.1 Using these criteria along with appropriate penetration data (in this case, data for the diffusion of Cb95 in pressed and sintered UO2) one first plots the data as log activity vs x2 and as log activity vs x, as shown by curves 1 and 2, respectively, in Fig. 1. A tangent line to curve 2 for large x (curve 3) is then transferred to the x2 plot as curve 4. The subtraction of curve 4 from curve 1 yields curve 5 which represents only lattice diffusion and may therefore be related to the diffusion coefficient by using the appropriate solution to Fick's second law. It should be noted that the method has inherent inaccuracies for small and large values of x2 and only the subtraction at intermediate values can be used. For the illustrated specimen the calculated diffusion coefficient is 3.4 X 10-11 sq cm per sec at 1445°C while data using a single crystal of UO2 at the same temperature yielded a value of 2.9 x 10-11 sq cm per sec.2 The same method of subtracting grain-boundary diffusion effects to obtain lattice diffusion coefficients has been successfully used to determine self-diffusion coefficients of Th230 in polycrystal- line thoria.

Jan 1, 1962

By G. Derge, L. D. Kirkbride

In recent years the problem of sulfur elimination in iron and steel-making has been of increasing importance. This interest has been due to the increasing amounts of sulfur coming into the system via raw materials. As a result, many investigations have been directed at obtaining information pertaining to the rate and mechanism of sulfur removal in the blast furnace as well as measuring appropriate thermodynamic quantities. Chang and Goldmanl as well as others2'3 showed that the process of sulfur transfer from molten carbon-saturated iron to blast furnace-type slags could be considered as two opposing first-order chemical reactions. Recently, Ramachandran et al.4 have shown that during desulfurization electrical neutrality is preserved in both phases and iron and silicon are transferred into the slag concurrently with the evolution of carbon monoxide gas. They concluded that the reaction s + c-+ (0)=(S) + CO(g) does not express the stoichiometry of the sulfur transfer process. As in all previous investigations, these authors studied the process of a net transfer of sulfur from metal to slag across a slag-metal interface. However, no experimental data have previously been obtained for the process of sulfurization. It was the purpose of this study to investigate the sulfurization process, i.e., net transfer of sulfur from slag to metal. EXPERIMENTAL PROCEDURE Apparatus—The equipment used in this investiga-tion was typical of that used by others.1-3The furnace was a high-frequency induction unit powered by a 35 kw Hg spark gap converter. Fig. 1 shows a schematic diagram of the apparatus. A two-compartment crucible assembly was employed to melt the slag and metal charges; being similar to that used by Ramachandran,4 in size and design. The atmosphere was that produced by reaction of hot graphite with air. Temperature measurements were made by W-Mo thermocouples which were calibrated against a Bureau of Standards Pt - Pt 10 pct Rh couple. Temperature control was always good to * 5°C and in general was much better than this figure. Raw materials used in the experiments were C. P. reagent grade where possible. The silica used in the preparation of the slags was in the form of highly purified silica flour. Silicon and titanium additions

Jan 1, 1961

By Tasuku Fuwa, John Chipman, David H. Kirkwood

Fe-C-Si alloys in silica crucibles were held at 1600°C in a controlled atmosphere of CO and Co2 and the approach to equilibrium was obsertsed. Results were not of sufficient precision to establish the activity and interaction coefficients. Observations 012 the equilibrium provided an average value for the standard free-energy change which differed by + 1.1 kcal from the earlier accepted value. The C-SL relation at other temperatures is calculated and shown graphically. IN the refining of steel in an acid-lined furnace, it is well known that, unlike in basic refining, a very significant amount of silicon may remain dissolved in the metal even when the carbon content approaches a final endpoint. Further, it is known that the residual silicon increases with increasing temperature. The quantitative relation between silicon and carbon can be calculated for equilibrium conditions from data already available on the Fe-C-O and the Fe-Si-O systems. Such equilibrium calculations could show us approximately the limitations within which the reaction would be expected to operate in practice. In the oxidation of a bath containing both carbon and silicon it is well known that at low temperatures silicon is preferentially oxidized first. It can be shown by calculation that at high temperatures carbon will be oxidized preferentially. It is the purpose of this paper to record direct observations on the Si-C relation by means of experiments carried out in an atmosphere of carbon monoxide for Fe-C-Si alloys contained in silica crucibles. When such a system is brought into equilibrium, at a fixed temperature, the composition of the liquid metal is governed by the following equations: Both K1 and K2 are known with a fair degree of accuracy, and may be calculated from the tabulated In general the percentage of CO2 is extremely small compared to that of CO and in the experiments to be described the partial pressure of the latter was generally only slightly below 1 atm. EXPERIMENTAL METHOD The induction furnace employed was the same as that described by Rist and chipman' and by Fuwa and chipman.2 It was modified slightly in view of the change from alumina to silica crucibles. The small silica crucible was surrounded by a thin susceptor of molybdenum sheet which in turn rested inside a zirconia crucible. Since the zirconia was subject to breakage, it was supported inside an alumina crucible. The charge consisted of 30 to 35 g of electrolytic iron along with the required amounts of refined silicon and graphite to produce a melt of required composition. Purified carbon monoxide and carbon dioxide were passed into the furnace through carefully calibrated flow meters. Two separate and independent series of experiments were carried out. In Series I (D.H.K.) the melt was held in the furnace at 1600°C for periods of time up to 6 hr. Samples were taken at intervals in silica tubes and were analyzed for carbon and silicon. The results of the last one or two samples taken from each experiment and the direction in which equilibrium appeared to be approached are shown in Table I. In Series II (T.F.) the silica crucibles were of less satisfactory strength and the melts were held at 1600° C for periods of only 2 to 4 hr. In heats number 29 to 45 inclusive, the metal was cooled in the furnace and the ingot itself served as a sample for analysis. In heat 46 and all subsequent heats, samples were withdrawn in silica tubes at the termination of the heat, and these samples were analyzed. The results of all of these experiments in which equilibrium appeared to be approached are listed in Table II.

Jan 1, 1965

By John F. Elliott, John R. Rawling

The rate of reduction of silica to silicon by carbon at 1550° to 1700°C in iron blast-furnace type slag-metal systems has been investigated. In the tower portion of the temperature range oxygen transport in the metal boundary layer at the slag-tal interface is probably rate limiting. A second step, which apparently involves a chemical process in the slag at the slag-metal interface, appears to be rate limiting when the oxygen-transport limitation is avoided. Above 1650°C rate control of the reduction by carbon probably is mixed between the two steps. INVESTIGATIONS1.2 into the equilibrium distribution of silicon between graphite-saturated slags and metals typical of iron blast-furnace practice show that, under the experimental conditions specified, the reduction of silica to silicon is relatively slow compared with the rates of other slag-metal reactions under similar conditions. For this reason, the reduction of silica in blast-furnace slag-metal systems was considered to be an interesting subject for further study. Investigations of the kinetics of the reaction3-5 suggest that several possible steps may be rate controlling, and that the rate-controlling step depends not only on the physical state of the systems but also on the configuration of the apparatus. The main objects of the investigation reported here were to sug- gest plausible reaction paths for the reduction of silica under the experimental conditions chosen and to determine the step or steps which control the rate of reduction of silica. EXPERIMENTAL Two types of experiments involving the reduction of silica from the slag were carried out. In the first, the only reducing agent present was carbon (graphite), for which the over-all reaction is (SiO2) +2C(gr) —Si + 2Co(g) [l] The second type was a graphite-saturated ferro-silicon-iron concentration cell. In it silica in the slag could be reduced either by carbon according to Eq. [I], or by a reaction of electrochemical nature in which silicon was oxidized at the ferrosilicon-slag interface and silica was reduced at the slag-iron interface without the formation of CO. The two types of experiment are described below. Reduction of Silica with Carbon as the Only Available Reducing Agent. The reduction of silica was accomplished by allowing a liquid lime-alumina-silica slag and graphite-saturated iron to react in an induction-heated graphite crucible. The crucible was 2.25 in. diameter and 5 in. deep; it was fitted with a graphite stirrer which penetrated through the slag and into the metal, Fig. 1. The slag and metal charges weighed 350 g each. Pure carbon monoxide at a pressure of 1 atm was maintained above the melt by passing the purified gas through the crucible at a rate of 0.5 liter per min. The gas did not stir either the slag or metal. Temperature measurements were made by sighting an optical pyrometer down the hollow stirrer rod into a 1/4-in. hole drilled in the stirrer blade. The crucible temperature was maintained within *15"C of the required value by manual control of the power supply to the

Jan 1, 1965

By W. O. Philbrook, A. E. El-Mehairy

The reduction by hydrogen of individual particles of dense hematite implanted in beds of inert spheres is controlled by single-particle kinetics. No evidence of reagent starvation was found down to low flow rates established by pressure gradients as low as 0.01 in. H2O per- in. of depth in beds of spheres as small as 3/16 in. diam. The optimum size for most rapid reduction of beds of homogeneous ores results from the effect on reducing potential of the gas of competing factors of supply and consumption. A recent publication1 presented the results of a study of the rate of reduction of iron ore in a fixed bed by hydrogen introduced at constant pressure differential. Under these conditions the reagent flow rate varied directly with the bed permeability, which was a function of particle size and shape and bed voidage, and inversely with the temperature-dependent viscosity of the gas. The present note adds data from supplementary experiments designed to reveal whether or not any partial reagent starvation arising from a flow-dependent transport resistance in the gas phase might have been operative. Kinetic studies were made of the reduction of individual spheres of iron oxide inserted into beds of inert spheres, so that the reagent flow rate and aerodynamic conditions were again determined by bed characteristics and temperature. The experimental techniques and apparatus were similar to those previously described.' The synthetic iron ore spheres were wet-rolled by hand from powdered, reagent-grade* Fe2O3 and subsequently dried and then fired in air at 1250°C. The indurated pellets has almost no porosity as indicated by their specific gravity of 5.10. The beds were made of dense alumina spheres, closely sized to the same dimension as the iron oxide spheres, in the following diameters: 3/16, 1/4, 5/16, and 3/8 in. Each bed was hand-packed of particles of a single size to a controlled voidage of 0.32 and depth of 3 in. within a capsule having an internal diameter at least 10 times the particle diameter. A few weighed iron oxide spheres were individually planted in the central plane of the bed of inert spheres at points away from the wall. Reduction was carried out at atmospheric pressure by purified hydrogen at 670°, 800°, and 985°c. The reagent was supplied at precisely controlled pressure differentials of 0.01, 0.02, and 0.03 in. of water per in. of bed depth. This gave a range of flow rates similar to that used in the study of beds of ore and comparable with blast-furnace practice in terms of mols of reducing gas per hr per sq ft of cross-section. Viscous flow prevailed in all beds under the experimental conditions of this study. The reduced particles were checked for loss in weight and then sectioned for metallographic examination. Occasional particles were found to have contained minute internal cracks that interfered with the concentricity of the reduction interfaces, and data for such particles were rejected. An experimental rate constant for the reduction of the imbedded Fe2O3 spheres was calculated on the assumption that the rate was proportional to the metal-oxide interface area by an equation that has been found by McKewan2 to describe the reduction of single spheres

Jan 1, 1962

By H. W. Hockin, D. r. Brandt, R. H. Walsh, P. L. Dietz, P. R. Girardot

New Jersey, Florida, and Canadian ilmenites were reduced with hydrogen or coke under various experimental conditions and the phase changes occurring in the ilmenite upon reduction have been studied by microscopic examination of polished sections and by X-ray diffraction. The products formed were dependent upon the type of ilmenite, temperature, time and reducing agent. Of the reducing agents, hydrogen was the more effective at lower temperatures. 1 HE possible utilization of ilmenite as a source of both iron and titanium has resulted in the development of a number of methods for the separation of the iron and titanium content. Slagging processes as currently used in Canada are typical of such methods. Somewhat less attention has been given industrially to the reduction of the iron content at less than the slagging temperature. Although references maybe found to such work, in general only one type of ilmenite, either natural or synthetic, has been studied by each author. We have attempted to draw some relationship between the results of experimental reduction and the type of deposit from which the ilmenite was derived, as evidenced by phases occurring in the ores and in the reduction products. Ilmenite ores having from 27 to 61 pct titanium dioxide were included in this study and in each case reduction was carried out at such temperatures that only limited coalescence of the particulate iron product occurred within the ilmenite grains. The history of the individual ilmenite samples and the temperature of reduction were found to determine the occurrence of various phases and the mode of distribution of the iron. REVIEW OF EARLIER WORK ON REDUCTION One of the earliest references to the reduction of ilmenite at less than the slagging temperature was made in 1917.' The metallic iron produced was_____ leached out by the action of dilute sulfuric or hydrochloric acid, to leave a product high in titanium dioxide. Subsequent to this, numerous patents2-'? and other references13"21 have appeared concerning reduction of ilmenite by carbon, hydrogen, carbon monoxide, methane, coal gas, water gas, or the like in the absence of fluxing agents. Mineragraphic examinations were not reported, and in the main, the only examinations made were chemical analyses. However, in two cases3,13, titanium suboxides were reported in the products and in one case2' rutile was reported, in addition to metallic iron. Both were identified by X-ray diffraction. While reductions in the absence of fluxing agents were generally followed by either a wet or dry chemical process for removal of the metallic iron, a Canadian source17 reported removal of the metallic iron by magnetic separation. By the addition of fluxing agents during reduction, the metallic iron has been coalesced into "pearls." Each reference to such a process23-28 has shown that the coalesced iron could be removed by magnetic or gravity means after the product was crushed. In the absence of fluxes, the coalescence did not generally occur. The major part of the present study has been limited to reductions without fluxing agents in order to determine the primary reactions of naturally occurring ilmenites. Titaniferous iron ores have also been studied,29-33 17, l9 their reduction having been used as the basis of a commercial process for beneficiation of such ores in Norway29 and more recently considered in the United States34,35. The ease of reduction of titaniferous iron ores relative to that of hematite or magnetite has been referred to,33 with the conclusion that the more

Jan 1, 1961

By Claude H. P. Lupis, John F. Elliott

IN the recent past, extensive use has been made of the interaction coefficient in treating the thermody-namic behavior of components in solutions at dilute concentrations. The development of this concept by wagner1 and chipman2,3 is sufficiently well-known so that its elucidation here is not necessary. The present discussion is focused on the relationship between E and e. The usual forms of the coefficients are:

Jan 1, 1965

By D. J. Carney, J. Chipman, N. J. Grant

An absolute calibration has been achieved for sampling and analyzing liquid steel for hydrogen based on Sieverts' values of hydrogen solubility in iron. Further checks were made in nickel, iron-nickel, and 18-8 stainless melts. The sampling method was successfully applied to a large number of commercial steel melts in various types of furnaces. The concurrent problem of sample storage was also solved. THE problem of sampling of liquid steel for hydrogen is more difficult than the problem of analysis. Hydrogen has a greater mobility than any other element; it is the only element that will diffuse out of steel at room temperature. In accordance with the known relation between temperature and diffusivity, the diffusion of hydrogen is extremely rapid at the temperature of liquid steel. Consequently, in the past, attempts to quench molten steel to retain the dissolved hydrogen have not been as successful as similar attempts with nitrogen or oxygen. The retention of hydrogen is especially difficult since it will continue to escape at room temperature even if the molten metal has been rapidly quenched. The hydrogen sampling methods may be classified broadly as (1) the rapid quenching techniques which attempt to preserve super saturation and (2) those methods which attempt to collect the gases evolved as the metal freezes and cools to room temperature. With all methods of sampling but especially with the rapid quenching methods, there is also a concurrent problem of storing the sample until analysis is undertaken. The two problems are interrelated and must be considered together when discussing liquid steel sampling techniques. The major obstacle in liquid steel sampling has been that there was no way of evaluating properly the various sampling and storage techniques which have been developed. It has not been possible to compare the accuracy of the various methods. For evaluation of sampling methods some known standard is required; this may be found in the published work on the equilibrium solubility of hydrogen in liquid iron which has been determined in three separate investigations.1,2,3 The data are shown in table I. Sieverts' original method of obtaining gas solubilities involved measurement of the volume of gas required to saturate the liquid metal contained in an evacuated bulb of known volume. This procedure, improved by introduction of high frequency induction heating, has yielded reproducible results which appear to be entirely dependable. Using Sieverts' method, it has been proved for iron and other metals and alloys that the hydrogen solubility is proportional to the square root of the pressure of the hydrogen gas above the liquid, so that one may work with various partial pressures of hydrogen to put different amounts of hydrogen in solution. From Sieverts' law it is possible to know accurately what amount of hydrogen is dissolved in the liquid metal at the time of sampling. The accuracy of a sampling method may be gauged by reference to this known solubility. It was the purpose of this investigation to develop a method for sampling liquid steel and for storing the sample so that the analytical result would represent the actual hydrogen content of the metal bath. For this purpose an induction furnace was

Jan 1, 1951

By D. J. Carney, J. Chipman, N. J. Grant

G. Derge—With the development of this last weapon, there is not much of a chance for hydrogen. It is certainly a very interesting paper, and it gives us more confidence in sampling liquid steel for hydrogen than we have ever had before. This satisfactory work gives considerable confidence to all of the chilled methods of sampling which should be more or less successful depending upon the degree of chilling obtained. J. T. MacKenzie—How do you handle a nonmagnetic specimen in that device? D. J. Carney (authors' reply)—I left that out in going over the analytical procedure. For the 18-8 steel, what was done was to get some iron wire which was degassed. One turn of wire was made around the sample; it was inserted into the analytical apparatus, and that was enough to allow the magnet to pull the sample and drop it in the furnace. Other methods could be devised for this apparatus to do the same thing. This was just a quick method. C. E. Sims—This method of sampling and analysis is very ingenious, and the results obtained attest to a satisfactory degree of accuracy. It was noted in the paper that the samples are satisfactory only when there is a complete absence of pipe. This should be empha- sized, particularly in view of the fact that they are cut off under water. Some reservations are held in regard to the adequacy of the method of sample storage prior to analysis. Cooling on dry ice undoubtedly slows down the rate of diffusion to a marked degree, and it obviously suffices when the analysis is made within a few hours. All of the samples, except some of high-nickel steels, were analyzed within 12 hr. It is still unknown how long this method would be safe. In the work at Battelle Institute, it was impracticable always to start analysis within 12 hr for various reasons. Some samples did not reach the laboratory for several days, and some were held as long as a month before analysis. During this period, they were held over mercury, and the gas evolved at room temperature was collected and analyzed. It should be mentioned that both the mercury and the container must be kept very clean.

Jan 1, 1951

By R. A., D. L. Sponseller, Flinn

Using specially del'eloped titanium nitride crucibles and a pressurized syslem, it has been possible to determine the solzibilitv of liquid calcium in liquid iron and iron-base alloys. At 2925°F the solubility is 0.032 wt pet Ca in pure iron. The solubility is increased by aluminum, carbon, nickel, and silicon additions, and lowered by additions of gold. An inflection in the calcium-solubility curve occurs as carbon is added to iron. At approximately 0.87 Wt pet Ca the CaC, phase replaces liquid calcium, and the calcium concentration de- creases as more carbon is added to the iron. The maximum solubility, 0.36 wt pet Ca, was obtained in iron containing 10.5 wt pet Si. The data have important implications regarding the use of calcium in process metallurgy, as well as providing an interesting test of theories of metallic solutions. Reasonably good agreement is found with theories of Alcock and Richardson and of Wagner, and criteria are suggested which permit these theories to be applied more readily to systems of this type. AT the present time relatively little is known about the factors influencing solubility in liquid metals. Information of this type is important for such matters as the control of impurity levels in liquid metals, the use of magnesium in ductile-iron production, and the use of liquid-metal alloys as fuels and coolants in nuclear power plants. One important example of restricted solubility concerns the extent to which liquid calcium will dissolve in liquid iron. This is a matter of considerable interest in view of the strong potential of calcium for removing certain impurities in steel,. as well as the ability of calcium to desulfurize, inoculate, and produce spheroidal graphite in cast irons.' Since ferrous materials normally contain other elements in addition to iron, the interactions of third elements with the Fe-Ca system are of interest. In view of the foregoing, this investigation was conducted to determine the solubility of liquid calcium in liquid iron and the effects of various third elements upon the solubility. For reasons of theoretical and engineering importance, the elements investigated were aluminum, carbon, gold, nickel, and silicon. Numerous attempts were made in the early part of the present century to determine the solubility of calcium in both liquid iron and steel, employing a variety of open-melt and closed bomb-type systems.2-4 From the results of these tests Wever concluded that calcium is insoluble in liquid and solid iron.5 These investigations were unsatisfactory in that they failed to insure the presence of a liquid layer

Jan 1, 1964

By T. Busch, R. A. Dodd

The solubility of hydrogen in pure iron and pure nickel, and of nitrogen in pure iron, has been determined and agrees well with earlier data. Nitrogen is insoluble in pure nickel and cobalt. The solubility of nitrogen in Fe-Ni alloys also confirms earlier- work. The solubility of hydrogen in Fe-Ni alloys increases non-linearly with increase in nickel content, but the solubility of hydrogen in Ni-Co and Fe-Co alloys reaches a minimum at 75 and 50 at. pet, respectively. The solubility of nitrogen in Fe-Co alloys decreases approximately linearly with increase in. cobalt content. Nitrogen is insoluble in Ni-Co alloys. Determined interaction parameters are in good agreement with predictions of Gokcen and Ohlani. THE solubility of hydrogen and nitrogen in liquid iron and iron alloys has been studied by various investigators.'-l' The method of solubility measuremerit due to Sievertsl is usually preferred when the vapor pressures of the liquid components are sufficiently small, but other techniques such as sampling the equilibrated melt and then analyzing by vacuum fusion have been used quite often. Previously reported data on the solubility of hydrogen and nitrogen in ostensibly pure liquid iron are summarized in Table I; certain of the quoted results were obtained by interpolation of the published data. Agreement between the various investigators is only fair, indicative of the experimental difficulties involved in obtaining gas solubility data. The only report found on the solubility of hydrogen in pure nickel is that of Sieverts.2 The solubility of nitrogen in iron-nickel alloys has received some attention, 10, 11 the results indicating that the solubility decreases with increase in nickel content, becoming zero for pure nickel. There is no published information on the solubility of these gases in either pure cobalt or alloys containing cobalt EXPERIMENTAL PROCEDURE Sieverts' method of solubility measurement was used throughout the investigation. The construction of the melting chamber is somewhat different from A. Optical pyrometer F. Alundum cover B. Vycor chamber G. Alumina crucible C. Furnace coils H. Alundum sand D. Glass water jacket I . Alundum thimble E. Chamber support J . Detachable vycor base sealed with de Khotinsky cement

Jan 1, 1961

By M. Weinstein, J. F. Elliott

IN conjunction with a study on the solubility of hydrogen in liquid pure iron and iron alloys, new and

Jan 1, 1963

By N. A. Parlee, A. E. Lord

Measurements of the solubility of lead in liquid iron were made at 1550°, 1600°, 1650°, and 1700°C using two different methods, i.e., 1) liquid iron-liquid lead equilibration and 2) liquid iron-lead vapor equilibration. LEAD was long considered insoluble in liquid iron1. Today steels containing 0.15 to 0.35 pct Pb are used. One laboratory has found a solubility of about 0.5 pet in the 1600°C range. Recently Miller and Elliott2 gave 0.60 pct as the result of one determillation at 1540°C. Because these findings resulted from the

Jan 1, 1961

By N. A. D. Parlee, N. M. El Tayeb

Fe-V alloys with small percentages of vanadium show no deviations from Sieverts' Law up to P~, = 1 atm in the 1600º to 1750ºC region. At somewhat under 8 pct V and up to at least 20pct V, at 1604" and 1661 °C, nitrogen absorption follows Sieverts' Law at low Pn2 values, but deviates from this law beyond critical values of pn2, due to formation of a nitrzde phase of approximately VN composition. The behavior below these solubility limits can be described at all temperatures by the equations: RAO and parlee1 suggested that the nitrogen absorption curve for liquid iron alloys takes the general form shown in Fig. 1 when an alloy nitride precipitate is formed. They showed that the precipitation of 6 titanium nitride from liquid Fe-Ti alloys follows this general scheme and dealt with the general theory involved. Recently, Evans and Pehlke2 reported their research on the precipitation of aluminum nitride from liquid Fe-A1 alloys. Fountain and Chipman's recent paper3 on the precipitation of boron nitride from solid solution contains a fine example of the analogous behavior which they have been studying in the solid state. Researches on the absorption of nitrogen by liquid Fe-V alloys have been described by various authors.1,4-7 Most of this work has been confined to fairly low vanadium concentrations and rather high temperatures because of difficulties variously described as, formation of surface scums, bridging, and so forth. This behavior and the rough thermo- dynamic calculations made by chipman,8 over eleven years ago, indicated that excessively high vanadium concentrations should not be required for vanadium nitride precipitation. It seemed desirable to look for and study this precipitation in order to fill out the thermodynamically interesting nitrogen absorption data on these alloys. EXPERIMENTAL METHOD The general method used was essentially the same as employed by Rao and parlee,1 in which a Sieverts type of apparatus was used and the precipitation of the nitride detected by the "break-point" in the Sieverts' Law line, and by the visual observation of the formation of the new phase on the surface of the melt. There were, however, a few differences in apparatus and technique as described below. The oil manometer used by Rao and parlee' to measure low partial pressures of nitrogen was not employed. The present authors sought to avoid as much as possible the exposure of the liquid alloy to low total pressures of gas during the 6 to 10 hr long runs because of the heavy iron evaporation-loss errors involved. This was achieved, partly by keeping evacuation times to a minimum with fast pumping but largely by the use of two burettes, one for argon and one for nitrogen, and using them to equilibrate the alloy with mixtures of argon and nitrogen rather than nitrogen alone. Each run involved a careful sequence of operations. After melting, de-

Jan 1, 1963

By J. C. Humbert, J. F. Elliott

The solubility of nitrogen in liquid pure Fe, Cr, and Ni, in liquid Fe-Ni, Fe-Cr, and Ni-Cr alloys and Fe-Cr-Ni alloys, has been measured by the Sieverts' type apparatus between 1500° and 1800°C. Tabulations and graphs of the solubility data and activity coefficients are shown. Heats of solution are reported. Interaction coefficients of the several elements in Fe. Cr, and Ni ave also reported. THE solubility of a gas in a liquid metal is usually measured by either the quenching technique or the so-called Sieverts' method. In the quenching method, 1-3 the melt is equilibrated at the desired temperature and pressure with the gas phase having a known fugacity of the gas being studied. The melt is sampled and the sample is quenched and analyzed chemically for its gas content. In Sieverts' method the melt is contained in a closed chamber. The apparent volume of this chamber, i.e. the "hot volume," is usually determined by introducing a known quantity of an inert gas having very closely the same physical characteristics as does the experimental gas. Alternatively, the experimental gas is used with an inert specimen having the same general physical characteristics as does the experimental melt. Subsequently, the experimental melt is equilibrated with the experimental gas by introducing a known quantity of the gas into the calibrated chamber. The solubility of the gas is obtained from P-V-T relationships. The quenching method suffers from' the limitation that some gas may be gained (as in the case of Cr-Fe alloys) or lost (as in the case of Fe-Ni alloys) during solidification and cooling of the sample. With Sieverts' method sampling and analysis are unnecessary. However, errors can result from the difficulty of determining accurately the "hot volume,"

Jan 1, 1961

By D. C. Hilty, W. Crafts

The solubility of oxygen in iron containing aluminum has been determined at 1550°, 1600°, and 1650°C and found to be much higher than predicted from theoretical considerations, possibly due to equilibria with an iron-aluminum spinel phase rather than pure AI,O.. The presence of manganese greatly increased the deoxidizing power of aluminum. DEOXIDATION and inclusion formation in steel have proved to be elusive phenomena that are difficult to control in spite of much effort to understand and evaluate the reactions. Study of the problem indicated that more accurate data on the solubility of oxygen in iron containing deoxidizers were essential to further progress, and an empirical determination of oxygen solubilities has been initiated at the Union Carbide and Carbon Research Laboratories, Inc. The investigation has included the effects of aluminum, silicon, manganese, and combinations of these elements at 1550°, 1600°, and 1650°C. The equipment and experimental procedure as well as the results of a study of the effects of aluminum are described below. In general, it was found that the solubility of oxygen in iron containing aluminum is much higher than predicted, possibly because of the formation of complex nonmetallic phases, and that manganese greatly increases the deoxidizing power of aluminum. The generally accepted deoxidation curves for 1600°C are summarized in fig. 1, from "Basic Open Hearth Steelmaking"'. These curves are assumed to represent the solubility of oxygen in the presence of deoxidizers and should indicate the limits of mis-cibility gaps in the phase diagrams of the systems that control the formation and character of non-metallic inclusions as described by Benedicks and Lofquist'. They were derived mainly by application of the methods of thermodynamics to the available experimental observations, and by calculations based on the thermal constants of the pure substances when experimental data were lacking. Among the more notable researches contributing to the development of knowledge of oxygen solubilities are those of Korber and Oelsen8, 4, C. H. Herty, Jr. and associates5,6, Krings and Schackmann7, Went-rup and Hieber8 Schenck and Briiggeman9, Vacher and Hamilton'" and Chipman and co-authors 1,11,12,13 Practically, the deoxidation curves are of limited value in predicting oxygen content and rationalizing deoxidation practices. The reasons for this inadequacy appear to be error from inaccurate data, oversimplifying assumptions, interacting effects of combined deoxidizers, and possibly unknown extra-equilibrium and side reactions. Although inclusion formation is a very complex reaction it must be re-

Jan 1, 1951