Search Documents

By S. Ramachandran, R. A. Walsh, J. C. Fulton

A method has been developed for calculating the oxygen content of commercially melted stainless steels. Application of thermodynamics to the de-oxidation reaction in straight chromium and Cr-Ni stainless steels has resulted in a set of equations DURING the past 30 years, many rewarding developments have resulted from the application of physical chemistry to steelmaking. A need still ex- which permits the calculation of oxygen contents. The model was programmed for computer calc~ilations. The validity of the method was demonstrated on routine oxygen samples from numerous commercially melted heats of stainless steel. ists for further progress in the use of thermodynamics to meet the more stringent requirements for cleaner steel. Current electric furnace practices have been developed around minimizing residual-oxygen content, one of the most important factors controlling the inclusion content of stainless steels. The residual oxygen is controlled by the use of deoxidizers during the finishing period and a number of empirical rules have been evolved. Although this feature of electric furnace practice has long been in use, only in recent years has funda-

Jan 1, 1963

By T. C. M. Pillay, John Chipman

The Si-0 equilibrium in liquid iron was investigated in some detail by Gokcen and Chipman1 who reported equilibrium constants of the following reactions: In each case the thermodynamic equilibrium constant K was obtained by extrapolating K' to infinite dilution. Their results on reaction [I] were consistent with those of earlier observers, and there seems to be little doubt as to the approximate validity of the equilibrium product determined. On the other hand, the results on reactions [2] and [3] indicated unusual curvilinear relationships between the activity coefficients of silicon and of oxygen and the silicon concentration. Since a linear relation is observed in all other similar instances, it was concluded that some source of error had affected the gas-metal equilibrium relationships. Recently the authors undertook a restudy of these equilibria with the view especially to establishing the equilibrium constant of reaction [2]. Their experiments, however, were discontinued before a tbroughly satisfactory answer had been obtained and consequently have remained unpublished until the present time. Quite recently Matoba, Gunji, and Kuwana2 published the results of a very thorough study of these equilibria. It is the purpose of the present note to Compare these three sets of experimental data, a comparison which leads to Confirmation of the results of Matoba and coworkers. The experimental arrangements described by Dutilloy and Chipman were employed. A controlled mixture of Hz, H,0, and argon was led into the induction furnace over the surface of the Fe-Si melt contained in a silica crucible. On account of the slow approach to equilibrium, runs were continued for at least 8 hr and, in some cases, as much as 12 hr. Samples were taken by means of silica tubes at the start and during the final 4-hr period of each run. Out of twelve runs only seven appeared to have approached equilibrium with sufficient certainty to be reported here. The results of these experiments, all at 1600°c, are shown by the triangles in Fig. 1. Equilibrium was approached from both oxidizing and reducing conditions and the direction of approach is indicated by the shape of the triangle. Attempts to extend the study to higher silicon concentrations were unsuccessful on account of the transfer of silica from the metal bath to the upper portions of the crucible by means of a silicon-monoxide mechanism. Matoba, Gunji, and Kuwana used similar methods and approached equilibrium from both sides. Their results at three temperatures are shown by the solid lines of Fig. 1. A line corresponding to 1600 deg is interpolated from their equations and is shown by the dotted line. The results of Gokcen and Chipman at 1600 deg in the range 0.02 to 2.0 pct Si are also shown. At concentrations up to about 0.6 pct Si, these investigators approached equilibrium from both directions, and their results in this range appear to be trustworthy. Above 1 pct Si, however, all of their experiments involved oxidation only, as shown by a decrease in the silicon content of the bath. It now seems certain that they failed to reach equilibrium at these higher silicon concentrations and that the curvilinear relation shown by their results is incorrect. It is evident that their results in the lower silicon range are in good agreement with those of Matoba and coworkers. In fact, a straight line drawn through their results would parallel the dotted line and would differ from it by only 0.05 in log K. The new data also lie close to that line. It may thus be concluded that the three sets of data are in good agreement and that the equations proposed by Matoba and coworkers for reaction [2] are satisfactory . They take the limiting value of log K', at 0 pct Si as log K ,. The slope of each line gives the value

Jan 1, 1962

By Y. Jeannin, C. Mannerskantz, F. D. Richardson

Measurements have been made between 1040o and 1300°C of the equilibrium between chromium metal, chromic oxide, hydrogen, and water. The results are in closer agreement with those deduced from recent thermal data than with previous equilibrium values. They lead to the equation ?G°1300°-1600°k = -266,700 + 59.95T ± 300 cal for the formzation of a mole of Cr2O3 from chromium and oxygen. This same equilibrium has been studied with the chromium dissolved in solid iron, and in this way chromium activities in Fe + Cr alloys have been established. The measuremenis show that there are substantial positive deviations from ideality in the a, and b phases, and support the results obtained by Heymer and Kubaschewski against those of other investigators. There is considerable disagreement between existing data,1"4 which is referred to in the Discussion, concerning the equilibrium Cr2O3 + 3H2 = 2Cr + 3H2O [l] There is also substantial disagreement between the results obtained by various investigators5-' who have measured the activities of chromium in its alloys with iron by the Knudsen effusion vapor pressure method. In view of the importance in metallurgical processes of chromium and its oxide and of Fe-Cr alloys, an attempt has been made to study equilib- rium [I] both with pure chormium metal and with chromium alloyed with iron, at temperatures between 1040o and 1300oC. From the work with the alloys it is possible to derive the chromium activities from the relationship

Jan 1, 1963

By Arnulf Muan, Egil Aukrust

Activities of COO in the three solid solution series COO-MgO, COO-MnO, and COO-"FeO" have been determined at 1200°C by equilibrating oxide samples with a metal phase (cobalt or a Co-Fe alloy) in atmospheres of known oxygen activities. The systems COO-MgO and COO-MnO are ideal, within limits of experimental error, whereas the system COO-"FeO" shows a slight positive deviation from ideality at the temperature of this investigation. KNOWLEDGE of activities of oxide components in solid solutions is necessary for the calculation of the cation distribution among coexisting oxide phases in many important metallurgical, ceramic, and geochemical systems. Very few activity-composition curves are available for such systems at the present time. A start on the prodigious job of obtaining such data has been made by studies of solid solutions with simple sodium chloride type structure containing NiO or "FeO" as a component.1-3 The present paper describes similar studies for solid solutions with sodium chloride type structure containing COO as a component. For the systems COO-MgO and COO-MnO, the activity solid solution is determined by the equation Here po and are the oxygen activities of the gas phase in equilibrium with the phase assemblages metallic cobalt plus oxide solid solution, and metallic cobalt plus pure COO, respectively. In the system CoO-"Fe07', the metal phase is a Co-Fe alloy rather than pure cobalt. In this case the activity of COO in the oxide solid solution is calculated from the equation where a& is the activity of cobalt in the metal phase. Because the system Co-Fe is close to ideal4 and the metal phase appearing in the present work contains less iron than cobalt even at the lowest COO concentrations of the oxide phases used may be set equal to N . Hence, Eq. [2] becomes The value of Nc, can be estimated from the relative stabilities of the oxides COO (see below) and I) EXPERIMENTAL METHOD Two different but closely related procedures were used in the present investigation to determine the activity-composition curves. In one method, the oxide solid solutions were equilibrated against

Jan 1, 1963

By J. Chipman, H. R. Larson

The data previously reported for the quantity as a function of oxygen pressure at 1550°C have been used to compute the activities of Fe, FeO, Fe2O3, and COO in slags of the ternary system. Activities of the first three have been obtained also for two quasi-ternaries involving fixed CaO:SiO2 ratios. IN a previous paper1' the authors reported the results of an investigation into the effects of oxygen pressure on the composition of various simple slags analogous to some of those which are found in steel-making practice. The ratio of ferric iron to total iron was studied at 1550°C in iron oxide melts to which lime, magnesia, lime-plus-silica, and other oxides were added. The oxygen pressures involved those of air, carbon dioxide, and carbon dioxide plus carbon monoxide in several proportions. Although very low oxygen pressures could not be used, the slag-metal equilibrium studies of Fetters and Chip-man' permitted extending the results to slags in equilibrium with iron. For the ternary system CaO-FeO-Fe,O, the oxygen pressure-composition relationship has been determined from zero percent lime to lime saturation over an oxygen pressure range from air to that represented by equilibrium with liquid iron. Lime and silica were added to iron oxide in the ratios 0.54, 1.306, and 2.235 to form three quasi-ternary systems which were also studied over the entire region of liquid melts at 1550°C. Ternary Gibbs-Duhem Equation Wagner" has developed a method by which the activities of two components of a ternary system can be calculated if' the activity of the third component is known throughout the composition range being considered. The fundamental form of the Gibbs-Duhem equation for ternary systems is N, d In a, + N2 d In a, + N3 d In a3 — 0. Wagner has developed a usable form of this equation by introducing the term y = N3/(N3 + N1) and rearranging the equation to give (? In a1/?N2)x = -y/(1-N2)2 (? In a2/?y) x2 N2/1-N2 (? In a2/?N2) y To apply this equation to the slag system CaO-FeO-Fe,O,, the activity of one of these components must be known. However, the only activity which is known from the experimental data is the activity or pressure of oxygen in the gas phase with which the slag is in equilibrium. The activity of oxygen in the slag can be defined as the square root of the oxygen pressure. In order to use oxygen as one component, the composition of the slags must be converted to the basis Fe-O-CaO. Oxygen, exclu-sive of that contained in CaO, then becomes com-ponent 2 in Eq. 1, Fe is selected as 1 and CaO as 3. Then y = Ncao/(Ncao + NFe). Eq. 1 then becomes (? In a fe/? Nx) y = -y/(1-No)2 (? In ao/?y) xp - No/ 1- No (? In ao/ O No) y. In order to evaluate Eq. 2, the boundary conditions must be known. The obvious choice of a standard state for iron is to assign slags in equilibrium with iron an activity of one. Then In a,., or log aFe, which is substituted for convenience, is determined by integrating along a line of constant y from the slag in equilibrium with iron to the composition at which log a,, is to be determined. Mathematically this can be expressed as log aFe (Nlog y) = log a'Fe + (? log a Fe/? No) dNo where the primes indicate equilibrium with liquid iron and a'Fe = 1. When Eq. 3 is integrated along a line of constant y, the following is obtained: log a Fe (No, y) = - y No ?/ No ? y ( log ao/(1- No)2) No d No - No/1-No d log ao. Lime-Iron Oxide Slags Iron Activity: The oxygen pressure of lime-iron oxide slags is shown in Fig. 1 as a function of j, de-fined as j = Fe+++/(Fe++ + Fe +++) for various constant mol percentages of lime. The j values at 0 to 60 pct lime were then determined for oxygen pressures of 1, 10-1, etc. to 10-" atm. For purposes of calculation, the line for zero percent lime was extrapolated to

Jan 1, 1955

By A. J. Jacobs, J. W. Spretnak, R. Speiser

The activities of nickel in liquid iron-nickel solutions containing from 10 to 90 at. pct Ni were measured over the temperature range .1510° to 1600°C. A special function was devised whereby activities can be computed by a graphical integration using the results of chemical analyses on the vapor phase alone. The liquid solutions studied exhibit negative deviation from Raoult's law, but they approach ideality as the temperature is raised from 1510° to 1600°C. The maximum heat of solution is less than -8.0 kcal per mole. The entropy of solution is less than —2.5 kcal per mole deg at 1510°C, and changes negligibly as the temperature is raised. The liquid iron-nickel solutions do not follow regular solution behavior. THE activities of the components of a liquid alloy system can be determined in favorable cases from the partial pressure of the vapor in equilibrium with the liquid alloy.' Assuming that the vapor over the alloy behaves as an ideal gas, the activity of component i is given by the expression, Where the superscripts 1, v, and o signify the liquid, the vapor, and the pure component, respectively; the f's are the fugacities, and the P's, the pressures. The quantities Pi and Pi can be determined by the

Jan 1, 1960

By E. T. Turkdogan

Jan 1, 1962

By John Chipman, Peter J. Koros

The distribution of copper between liquid silver and liquid iron, Fe-C, and Fe-C-Si alloys was studied at 1600°C. From the data and the activity of copper in silver obtained from the phase diagrams, the activity coefficients of copper in iron up to 3 atomic pct are log y, = 0.90 — 2.4 N, and y": = 8.0. The effect of carbon up to saturation is expressed as log y = 1.8 Ni. The effect of silicon up to 11 atomic pct is insignificant. SLOWLY rising copper content of steel being produced is focusing attention on the lack of information on the behavior of copper in liquid steel. Chou' measured the activity coefficient of copper in liquid iron through its distribution between liquid silver and iron. He obtained a value of 9.12 at infinite dilution. Chipman' calculated the activity coefficient of copper in liquid iron at 1600°C from the phase diagram and obtained a value of 12. Ameiln and Pfeiffer- investigated the possibility of removing copper from liquid steel through the use of lead, silver, and bismuth as solvents. They also calculated that the activity coefficient of copper in liquid iron would be increased five or sixfold by addition of 4 pct C. Langenberg and Lindsay' are investigating the effectiveness of lead and sodium sulfide in removing copper from liquid iron. The experimental method was similar to that used by Chou' and Chipman and Floridis." Silver, copper, and iron were melted together, the copper distributing itself between the two layers so that at equilibrium a,.,, (in Ag) - a,.,, (inFe) and N,.,, (in Ag) yc,, (inFe) (in Ag). [21 N,.,, (inFe) Thus, from the concentration of copper in each layer and the activity coefficient of copper in liquid silver, the activity coefficient of copper in liquid iron can be calculated. The effect of an alloying element such as carbon on the activity coefficient of copper can also be determined from its effect on the distribution, provided the new element is essentially insoluble in silver. The one serious limitation of this method is the increasing mutual solubility of silver and iron in the presence of a third component. The mutual solubilities of pure iron and silver are negligible, but increase with increasing copper concentration beyond the range investigated. Ag-Cu System—The activity coefficient of copper in silver at 1600°C had to be estimated, since direct experimental data are not available. Kawakami" measured the heat of mixing of Ag-Cu alloys at 1200°C by a calorimetric technique. Scheil' calculated the heat of mixing in this system by use of the phase diagram, assuming that the solution is random and that the heat of mixing is independent of temperature. The terminal solid solution may be assumed to obey Raoult's law within the requirements

Jan 1, 1957

By J. Chipman, T. Fuwa

The effects of various elements on the activity coefficient of carbon in liquid iron have been studied by two experimental methods: 1) equilibration with controlled mixtures of CO and CO2; 2) the solubility of graphite in the melt. Activity coefficient of C is increased by Al, Co, Cu, Ni, P, Si, S, and Srz. It is decreased by Cr, Cb, Mn, Mo, W, and V. THE thermodynamic properties of the iron-carbon binary system have now been fairly well established, although some uncertainty remains with respect to the exact location of some of the phase boundaries. The activity of carbon in ferrite and in austenite has been measured in the classic researches of R. P. smith' while similar measurements by Richardson and ~ennis, and by Rist and chipman3 have established the values of the activity of carbon in liquid iron up to 1760°C. On the other hand, our knowledge of the effects of alloying elements on the activity of carbon in dilute solutions is restricted to Smith's experiments on systems Fe-C-Mn and Fe-C-Si in the austenitic range and to some more recent experiments of schwarzman4 in the a range. In addition there have been a number of determinations of the effects of various elements on the solubility of graphite in liquid iron, and from these the corresponding effect in saturated solution may be obtained. The purpose of the present study was to extend the investigation of the liquid system to include the effects of alloying elements upon the activity coefficient of carbon, principally in dilute solutions. Equilibrium measurements were made on the reaction C + co, = 2 CO (g) The prepared mixture of CO and CO,, diluted with argon, flowed over the surface of the liquid metal which, after several hours' exposure to the gas, was quenched and anqlyzed. As in the earlier experiments, the principal experimental difficulty was in the deposition of carbon on the parts of the furnace at temperatures slightly below that of the metal bath. In order to minimize this difficulty, the ratio (Pco)2 /PCo2 was restricted to values not much higher than 100 atm, and correspondingly the carbon concentration in the metal seldom exceeded 0.30 pct. EXPERIMENTAL METHODS The method and apparatus were essentially the same as used by Rist and Chipman.3 The gaseous mixture consisting of highly purified CO, CO,, and argon, each controlled by a flowmeter, was led into the furnace and passed over the surface of the liquid-iron melt which was heated and stirred by high-frequency induction. One slight modification was made in that a molybdenum susceptor was placed outside the crucible for the sake of uniformity of temperature and to combat the tendency of carbon to precipitate on the crucible wall. Pure alumina crucibles approximately 25 mm ID were used. The charge consisting of about 30 g was made up of electrolytic iron, the alloying element to be added, and enough graphite to supply slightly more or less than the anticipated equilibrium carbon concentration. All metals used were of high purity. Metallic chromium, columbium, and vanadium were from special lots supplied by the Electro Metallurgical Co. Tin, copper, molybdenum, tungsten, cobalt, and nickel were of purest commercial grades. The electrolytic iron, after being cut to the proper size for charging, was prereduced by hydrogen at 850° to 1000°C to remove surface oxidation. The oxygen content of the reduced material was 0.002 pct. This treatment made it easy to control the carbon content of the initial melt. The charge was melted under the gas mixture to be used for the entire run. In some earlier melts the charge was melted under a stream of argon, but in this case some alumina was reduced from the crucible, and the aluminum thus absorbed in the melt was subsequently oxidized with the formation of a solid film of alumina on the surface of the melt. AS another safeguard against film formation, overheating of the bath was carefully avoided. All runs were made at a temperature of 1560°C. Under experimental conditions a charge of pure iron picked up 0.17 pct C in 3 hr and 0.23 pct C in 6 hr under an atmosphere for which the equilibrium concentration of carbon is 0.27. It is clear that the time required to reach equilibrium from an initially carbon-free melt would be very great. For this reason each experiment was started with a melt of known carbon concentration not far above or below the expected equilibrium value, and each melt was held at temperature for a period of at least 5 hr. Under such circumstances it was possible to chart the approach to equilibrium from both high-carbon and low-carbon materials. Temperature was controlled by frequent optical observation and adjustment and the metls were timed in such a way that the final 2 hr occurred during a time when electric power was steady; for example, 2 to 4 pm or after 11 pm. In melts containine volatile metals such as copper, tin, and mangane\e the time of holding was decreased somewhat in

Jan 1, 1960

By C. E. Lesher

Two processes for agglomerating fine sized ores with low temperature coke are described. One process (Orcarb) agglomerates ores with limited amounts of carbon; the other (ore-carbon pellets) pelletizes fine sized ores, using low temperature cokes as the binder. Data are presented on the products obtained when taconite, magnetite, and hematite concentrates and several titanium oxide ores were used. REDUCTION of finely divided oxide ores has long been a major metallurgical problem. Various methods of agglomerating, notably briquetting, sintering, nodulizing, and pelletizing, have been developed and are in industrial use. Carbon is the usual reducing agent for oxide ores and attempts have been made to agglomerate fine ores with coking coal in order to get the metallurgical advantages of intimate contact and increased particle sizes suitable for subsequent smelting or reduction.1-3 For the past three years, research has been in progress on two processes for combining fine sized ores with reactive carbon in the form of low temperature coke. By one process, designated Orcarb, the ore is coated with coke carbon in amounts limited to that required for reduction and the result is an appreciable but limited increase in product particle size over that of the original ore. The second process, called ore-carbon pellets, produces pellets of fine ore bound by low temperature coke. The first objective was to develop a process for combining an oxide ore with only enough carbon for its reduction, the amount being that derived from the simple reduction equations. The carbon (that amount theoretically required for the reduction of the oxides in an agglomerate with ore), calculated to CO, ranges from 10 pct for zinc calcines Or phosphate Ores to 21 pct for rutile and 26 pct for silica. The carbon required for reducing iron oxides falls between the extremes—10 and 26 pet' When the carbon in the agglomerate is low temperature coke, reduction by hydrogen may be expected to take place, thus modifying the calculated carbon require- ment. On the other hand, in practice, an excess of reducing agent over that required in theory is normally used. Screen sizes of three iron ores on which the greater part of the experimental work was performed are given in Table I. Most of the testing was done with Freeport (Renton Mine) coal, but Pittsburgh and Elkhorn bed coals were also used. The first objective was to agglomerate the ore with sufficient low temperature coke to provide carbon for reduction, i.e., in the range from 10 to 25 pct. This was accomplished by preheating the ore, mixing with pulverized coking coal, and completing the coal carbonization in a plain steel rotating retort. The pilot retort, as it is now constituted, is shown by diagram in Fig. 1 and photograph, Fig. 2. Ore is fed into the upper kiln by a screw; it is heated to 900° to 1100°F by direct flame and it is then discharged into an insulating hopper from which it goes

Jan 1, 1956

By W. O. Philbrook, H. W. Hosking, N. B. Melcher

Rates and patterns of gas flow in an experimental blast furnace were investigated by measurements, at three levels, of gas compo-sitions and temperatures and by direct determination of gas transit time by a new method employing mercury vapor as a tracer. Infer-ences were drawn about localized reaction zones and the mechanism of the onset of' channeling. LACK of precise knowledge of the nature of gas movement in the blast furnace has always been an impediment to better understanding of the blastfurnace process as a whole. The effects upon operation of various types of charge distribution are very largely brought about by the modifications to gas flow which they occasion. Intelligent evaluation of these effects demands a better understanding of the factors influencing gas flow and the ways in which the flow pattern is altered by conditions existing in the stock column. The same knowledge is required before a fundamental approach can be made to the problem of establishing optimum lines for new furnaces. An experimental furnace is operated at the Bureau of Mines Central Experiment Station, Pittsburgh, Pa. Advantage was taken of experimental operation to conduct an investigation to develop and evaluate methods of study which could be applied later to industrial furnaces. The gas composition and temperature determinations were made as a preliminary study so far as gas flow was concerned but yielded significant information relating to other processes occurring within the furnace. The direct measurement of gas transit times by the mercury-vapor method constituted the major object of the investigation. Development of the mercury-vapor method was carried out while the data on gas composition and temperature were being collected. A complete description of the principle and development of the mercury-vapor method is beyond the scope of this paper but is available elsewhere. The results obtained and some consideration of the method's effectiveness and its potential value in the study of industrial problems will be treated here. DESCRIPTION OF FURNACE The general lines of the experimental blast furnace are shown in Fig. 1. The hearth diameter was 3 ft at the beginning of the investigation, and this was later enlarged to 3 ft 9 in. and finally to 4 ft to increase capacity. Tuyere to stockline height is 21 ft 4 in. The bosh depth is 3 ft 4 in. over which it opens to a width of 5 ft, and this diameter is carried up for 2 ft in a cylindrical section. The inwall batter of 1/2 in. per ft. is considerably less than in larger installations This may have influenced the movement of gaseous and solid materials. The furnace is inside a laboratory building and is charged from outside by a skip and hoist. Feeding into the first of three bell chambers, the stock is distributed by rotating the upper bell, the other two being used to obtain an effective gas seal. Three water-cooled tuyeres, with a nozzle dia-

Jan 1, 1960

By E. C. Nelson

An equation is derived for calculating approximately the solubility of nitrogen in an alloy steel over a temperature range from 1200" to 1900°C using data on the effects of alloys on the activity coefficient of nitrogen in iron at 1600°C. Calculated values are compared with experimental data. STUDIES involving the solubility of nitrogen in complex iron alloys have been aided by a system devised by Lanenberg for calculating the solubility of nitrogen in ternary and higher alloys of iron at 1600C using available data on the activity coefficient of nitrogen in binary alloys. Calculated nitrogen solubilities agree well with experimental determinations on complex alloys,' and the calculation procedure is useful when experimental data on specific alloys are not required. A method for extending calculations of nitrogen solubilities in alloy steels to temperatures other than 1600°C would be useful because of the paucity of experimental data in this area. Assumptions which must be made to permit this calculation introduce an element of approximation; however, the calculation is useful in cases where a large number of alloys and temperatures are under consideration, and the precision of experimental measurements is not required. The activity coefficient can be freed from the isothermal restriction of Langenberg's calculation if it is assumed that the solution is "regular". This is a reasonably good assumption for solutions of nitrogen in iron alloys, because usually they are dilute and the heat of solution is relatively small. The presence of large amounts of certain alloying elements would make this assumption less tenable.3 One of the properties of a regular solution is: when "f" is the activity coefficient of a solute and "T" is the temperature in OK. If the activity coefficients at temperatures T and T2 are denoted by subscripts 1 and 2 respectively: The activity coefficient of nitrogen in a complex alloy then can be calculated at any temperature from the sum of the effects of the other solutes on the activity coefficient of nitrogen at 1600°C. It will be convenient to incorporate this expression into an equation which will give directly the logarithm of the solubility of nitrogen at any temperature in the presence of alloys. Two constants, the solubility of nitrogen in pure iron at 1600 "C and the partial molar heat of solution of nitrogen in iron are required. The heat of solution value of 1400 cal at 1600 °C quoted by Kubaschewski and vans will be used. The accuracy of the calculation will suffer if the heat of solution should change materially with temperature. In the absence of experimental data on the temperature dependence of this quantity, it will be expedient to restrict the application of the calcula-

Jan 1, 1963

By G. W. Healy, W. D. Forgeng, N. R. Griffing

The liquidus surface of the C-Cr-Fe system to 1900°C has been mapped from carbon solubility and freezing point measurements, metallographic observations, and published data. In the graphite field, the surface drops precipitously to meet a slope descending from the Cr3C2-C join through the mixed carbide fields to a meandering eutectic valley; this flows from the Cr-C to the Fe-C edge. On the low-cavbon side of the valley, the surface rises to the chromium corner. 1 HE purpose of this inquiry was to gain a more complete knowledge of the C-Cr-Fe liquidus surface. This required an extensive study of carbon and chromium-rich alloys which, in the past, have received the least attention. Sanbongi el al.1 measured the solubility of carbon in liquid alloys containing up to 25 wt pct Cr at temperatures of 1460" and 1545°C and found that it increased with chromium content and temperature. Lucas and wentrup2 also made solubility experiments in graphite crucibles, but at higher chromium contents in the range of 38 to 80 pct and at temperatures of 1550" to 1700°C; however, their experiments were relatively few, and the effect of temperature was not consistent. The present investigation included compsitions ranging from 0 to 94 pct Cr and 1.4 to 11.7 pct C at temperatures up to 1800°C. Saturation experiments in graphite crucibles, thermal analyses during freezing, and metallographic and X-ray diffraction studies were employed to obtain desired information which, together with published data from other investigations about the binary edges and the ternary system at lower carbon levels, made possible the construction of a comprehensive C-Cr-Fe liquidus surface. C-CR-FE LIQUIDUS SURFACE The derived ternary liquidus surface is shown in Fig. 1 where the solid curves are liquidus isotherms. The dashed curves and the binary edges are the liquidus field boundaries of the primary crystals, the arrows along these boundaries indicating their direction with decreasing temperature. Each intersection of three internal boundaries is the composition of liquid involved in four-phase isothermal equilibrium with the three primary crystals whose liquidus fields touch there. It may be observed that a total of six four-phase equilibria occur at melt temperatures. The general topography of the liquidus surface consists of a eutectic valley sloping down from the Cr-C edge at 3.2 pct C to the Fe-C edge at 4.25 pct C. The liquidus surface rises in two directions from the valley, one hillside terminating at the Cr-Fe edge while the opposing hillside rises toward the carbon corner out beyond the top edge of Fig. 1. Both hillsides are interrupted by intersections of surfaces or joints representing the liquid colmpositions of various three-phase equilibria. Seven primary crystals occur, namely a(bcc), ?(fcc), graphite, (Fe, Cr)3C capable of dissolving as much as 15 pct Cr, and three different chromium carbides in which iron is partially soluble by substitution for chromium. The composition ranges of the primary l~hases at melt temperatures are shown

Jan 1, 1962

By G. W. Healy

Activities of oxides in the ternary FeO-MnO-SiO system are calculated from data on the binaries, using the Gibbs -Schuhmann method. These activity data are used, together with thermodynamic relations, to calculate the contents of oxygen, silicon, and manganese in liquid iron, and the composition of the corresponding oxide phases in equilibrium with it. Excellent agreement was found between calculated and experimental metal compositions, indicating that thermodynamic methods, such as those presented, can supply useful data on slag-metal equilibria, and are therefore capable of wider application. When molten steel containing oxygen is treated with a deoxidizer, the product of the reaction is composed of the oxides of the deoxidizing elements together with more or less iron oxide. This assemblage of oxides is presumably very close to equilibrium with the melt since it is formed directly from elements dissolved in it. Therefore, it is likely that metal analyses corresponding to a given oxide phase composition are calculable with considerable precision where sufficient information is available on activities in both phases. The present study was made to test this hypothesis in the case of deoxidation by silicon and manganese. Here experimental data on the relationship among oxygen, silicon, and manganese in the metal phase are available, as are activity data for the MnO-SiO2, FeO-SiO2, and FeO-MnO binary systems. A further objective was to demonstrate in a specific example how gaps in the data may be filled by judicious interpolation, and how the Gibbs-Schuhmann method is applied to obtain activities in ternary systems. The rigorous derivation of the latter, requiring considerable facility in handling partial differential equations, tends to mask the real simplicity of the method, and repel the very engineers and experimenters who would benefit the most from its use. I) GENERAL PROCEDURE Deoxidation of steel with silicon and manganese results in the formation of two phases, metal and oxide. The number of components is four—Fe, Mn, Si, and 0. The phase rule then gives the number of degrees of freedom: If temperature and pressure are fixed, 2 deg of freedom remain. Consequently, it suffices to fix the concentration of two of the components of the system for all the concentrations to be determined and, at least in principle, calculable. The practical problem is then to discover a straightforward method of calculation with a minimum of trial and error solutions. The type of reactions to be dealt with are like the following: Si (in steel) + 2 O(in steel) = SiO2(in oxide phase) The standard free energy of this reaction is readily obtained from the literature, so that the equilibrium constant can be calculated by the expression: Since the metallurgist is concerned with percentages rather than activities, a real solution requires information that will permit activities to be converted to concentrations. This information is fairly adequate for the elements dissolved in molten steel, but is not available for the FeO-MnO-SiO2 oxide system. Therefore, the first step in this study was to find the relation between activity and concentration for the three oxides making up the system. The next step was the calculation of slag and metal compositions that are in equilibrium with one another. As shown by the phase rule, it is sufficient to assume the values of two composition variables in order to calculate the rest. But which two are picked makes a vast difference in how difficult the calculation turns out to be, so that it is necessary to make preliminary trial of several routes to find the easiest one. A convenient method is to begin with the slag, and to pick FeO concentration and that of one of the other oxides. Starting with the relation between oxygen dissolved in the liquid steel and FeO in the slag permits a good first estimate to be made of the oxygen content of the steel in equilibrium with FeO, because the iron content being fairly close to 100 pct, Fe has an activity close to unity. Using this first estimate of oxygen activity, approximate values of the manganese and silicon contents of the melt are readily calculated from the appropriate equilibrium constants. A minor correction for the depar-

Jan 1, 1963

By J. F. Peters

The term solution loss is discussed and defined. Examples are given showing that solution loss may either have a favorable or unfavorable effect on blast furnace performance. A theory is advanced explaining the contradictions encountered during earlier studies of the problem. MANY papers have been written and numerous theories advanced concerning the chemical reactions in the iron blast furnace. Richards, discussing the utilization of fuel in the blast furnace, said: "All the carbon burnt in the furnace should be first oxidized at the tuyeres to CO and all the reduction of oxides above the tuyeres should be caused by CO, which thus becomes CO,. This dictum is not Gruner's own words, but expresses their sense, and from the point of view of the present discussion, it is the correct principle upon which to obtain the maximum generation of heat in the furnace from a given weight of fuel." Richards also said the cubic feet of wind per pound of coke does not express how efficiently a furnace is running, but it will be shown later where Howland paid much attention to this figure. Richards pointed out a shortcoming of Grun-er's discussion in that Richards claimed there is not enough CO generated in a good working blast furnace to possibly combine with the oxygen of the ore, so some of it must be reduced by solid carbon. Mathesius2 made two important contributions. First, he wrote the direct and indirect reduction equations showing the heats of formation in each case and stressed that the indirect reduction reaction produces heat while direct reduction absorbs heat. Secondly, he substantiated Richards' conclusion that there may be a deficiency of CO gas and that some direct reduction is required, but he pointed out that "direct reduction in the hearth has often been confused with direct reduction in the stack, which more correctly should be termed 'premature combustion'." Then he said that "as this carbon is consumed partly by economical (direct reduction in the hearth) and partly by wasteful reactions in ever varying proportions, it is evident that the relation of this total carbon gasified above the tuyeres to the fuel consumption of a furnace cannot possibly have a direct bearing on the economy of furnace operation. The premature combustion of carbon (CO2+C?2CO) must therefore in all cases be considered a detrimental reaction." Johnson3 discussed this whole problem. Through experience Johnson found that when departure from Gruner's ideal working is considerable, the fuel economy is poor, and when the quantity of blast goes up, the fuel economy goes up in the same proportion. Johnson said that the wind per pound of coke is a measure of fuel economy, which will be shown to be wrong. Howland4 established the fact from operating data that there was no relationship between the coke consumption of a furnace and the percentage of coke burned at the tuyeres. But some of his calculations have led to questionable conclusions, such as "it is practically impossible to obtain low coke consumption unless we keep our wind low." Howland made an important contribution when, in his concluding paragraphs, he reached certain negative decisions as to why one coke works better than another. He said the reason one coke works better than another in a blast furnace is not because: 1—there is any difference in the percentage gasified at the tuyeres, or 2—there is any difference of wind required per pound of coke. Korevaar5 propounded a new theory on combustion which disagrees with Gruner because "as far as we can see, this is sufficient proof for the invalidity of Gruner's ideal, for if it was valid, the Low coke consumption should always be accompanied by a higher percentage of carbon burned at the tuyeres. This not being the case, we must deliberately give up our belief in Gruner's ideal." The part in italics will be proved to be wrong. Martin" contended that there was not enough reducing capacity in the blast furnace unless solution loss occurs. He came to a series of conclusions, such as: 1—The greatest efficiency of the blast furnace may not be attained when the reduction is performed entirely by carbon monoxide as demanded by Gruner's definition of the ideally perfect working of the blast furnace. 2—The so-called solution loss reactions, which should more properly be termed direct reduction reactions, promote furnace efficiency. 3—In modern blast furnace practice, the carbon consumption of the process is determined primarily by the carbon needed for reduction purposes, any thermal deficiency created by the reduction process being balanced in practice by the use of blast heat. From a practical standpoint further discussion of the theory of chemical reactions is of little moment. There is however one phase of the general theory of chemical reactions which is very important and that is the combustion of coke in the blast furnace. The purpose of this paper is to show that the reaction

Jan 1, 1955

By A. B. Kinzel

IT is with sincere appreciation and a deep sense of responsibility that I accept the honor of delivering the Howe Memorial Lecture. In our time metallurgical research has delved into phenomena ever more complex in character, so that much of this work must be done by organized teams. We, of the Union Carbide and Carbon Research Laboratories and the Electro Metallurgical Co., are truly a team and the lecture which is to follow is due largely to a few of the outstanding members of that team here cited and well known to you in their own right: John Lamont, whose work deserves special recognition, Walter Crafts, William Forgeng, William Binder, Robert Fowler, and David Swan. Thus it is a fitting pleasure to accept this honor for the research institution which I represent. Henry Marion Howe was truly inspirational and his breadth of view was such that he could use either the method of rigid proof or implied mechanism to further the broad understanding. But perhaps his outstanding characteristic was his sound judgment and his ability to weigh observations. The following is replete with observations and judgment as to their implication. We would like to feel that it is the type of dissertation which Howe, himself, would have enjoyed. Chromium carbides in stainless steel have, in a sense, been a key to the progress of the stainless steel industry. Brearley's first stainless steel, containing approximately 0.30 pct C, was successful because it differed from previous Fe-Cr alloys in that the carbides were present in such amount and form as to provide the stainless and cutting properties. Again, the 12 pct Cr die steels with a carbon content of 1.5 pct or more achieved their properties by virtue of the form and distribution of the chromium carbides. In the straight chromium steels with from 12 to 25 pct Cr, chromium carbide amount and distribution are once more the determinants with respect to physical properties and corrosion resistance. In each of the cases just cited the role of the carbide is in essence similar to that of iron carbide in carbon steel and this story is so well known that we need not dwell on it. In the austenitic stainless steels, the chromium carbide plays only a minor positive role with respect to mechanical properties, and for corrosion resistance the objective is to achieve maximum homogeneity. Thus the tendency has been to ever lower the carbon and correlatively the chromium carbide content of the austenitic steels. Today stainless steels with carbon less than 0.03 pct are beginning to find their true place, and this is being aided in major degree by our Company's introduction of a new type of ferrochromium having negligible carbon. The role of carbon in these austenitic steels has long been recognized, and a truly enormous amount of research has been devoted to the subject. Much has been learned about corrosion resistance, but the attack on the problem from the point of view of carbide precipitation and the nature thereof has left much to be understood. Because of its fundamental importance to the future of stainless steels, further study of the mechanism of carbide formation in austenitic steels is of interest. In the 18-8 Cr-Ni steels, which are typical of the austenitic class of steel, carbon is soluble in the aus-tenite to the extent of approximately 0.15 pct at 1832°F (1000°C). The solubility becomes less with decreasing temperature, so that at 1652°F (900°C) the solubility limit is approximately 0.06 pct, at 1472°F (800°C) it is about 0.03 pct, and at 932°F (500°C) it is approximately 0.01 pct. Obviously this is the ideal situation for the classical precipitation phenomenon. It is true that, in a sense, 18-8 austenite resulting from annealing is in a metastable state, and the phenomenon of ferritic precipitation from this austenite should be considered. However, in a so-called fully austenitic 18-8 steel this is secondary with respect to the carbides and can be neglected. A wide variety of chromium carbides is known. Only one chromium carbide, Cr,,C,,, frequently referred to as Cr,C, has been found in unmodified low carbon 18-8 steel. This, fortunately, simplifies the study of the precipitation phenomena. For this study an A.I.S.I. type 304 stainless steel of commercial quality was selected. Analysis showed C 0.07 pct, Mn 0.49 pct, Si 0.33 pct, Ni 9.34 pct, Cr 18.91 pct, and N 0.033 pct. As expected with this composition the steel was homogeneous after annealing except for a few pools of ferrite, which were disregarded in that their location was not coincident with subsequent carbide precipitation areas. On corrosion testing the steel was normal and typical of good commercial quality. Specimens 1/2xlx4 in. were annealed and subjected to precipitation heat treatment for 100 hr at 100" temperature intervals from 1000° to 1500°F and for ten selected intervals ranging from 5 min to 64 hr at 1300°F. These blocks were then cut into small pieces for metallographic studies. The first problem

Jan 1, 1953

By R. F. Mehl

Jan 1, 1961

By J. H. Swisher

An experimental study has been made of the possible mechanisms for foam stability in the system CaO-SiO2-Cr2O3, where Cr2O3is the foaming agent. The degree of lowering of surface tension by Cr2O3 was determined at 1600 "C for high -silica melts. Measurements were also made of foam stability under standardized conditions. The results of the two sets of measurements then were correlated by applying the Gibbs and Marangoni theories of film elasticity. It was demonstrated that the Marangoni elasticity effect is probably the largest single contributor to foam stability, although the high uiscosity of' silicate melts makes some contribution in controlling the rate of drainage of liquid from bubble lamellae. ThE tendency of metallurgical slags to foam has frequently been observed during steel-refining operations. In the open-hearth furnace,this tendency is considered undesirable because it slows down the rate of heat transfer from the burners to the molten-metal bath. On the other hand, it is sometimes desirable to have a "foamy" slag in the basic oxygen converter.' A more effective blanket over the steel bath is made by a "foamy" slag, and loss in yield of metal due to splashing is minimized. In addition, a flush slag can be removed more easily in this condition. Since slag-foaming problems in the field have been attacked largely on an empirical basis, it appeared desirable to obtain a better basic understanding of the mechanism of foam stability. The experimental work was directed almost entirely toward the behavior of Cr2O3 as the foaming agent in CaO-SiO,-Cr2O3 melts. In the first stage of the investigation, the degree of lowering of surface tension by Cr2O3 was determined at 1600°C. The second stage consisted of measuring foam stabilities under standardized conditions. The two sets of experimental results then were correlated by applying various theories of foam stability. For convenience in interpreting the surface-tension and foam-stability results reported here, a portion of the CaO-SiO2-Cr2O3 phase diagram2 is reproduced in Fig. 1. EXPERIMENTAL 1) Surface-Tension Measurements. Very little information is available in the literature on the lowering of surface tension of slags by foaming agents. Kozakevitch 3 mentioned that the addition of 1.5 pct Cr2O3 lowered the surface tension of FeO slightly. In another system, some data on the lowering of surface tension of foaming slags by P2O5 were reported by Cooper and Kitchener.4 The technique used for the surface-tension measurements was the maximum bubble-pressure method,' using a single tube immersed to various depths in the melt. This method has been used extensively in high-temperature systems, such as in liquid metals, slags, and glasses.6,7 It has the advantage that the contact angle between the liquid and the tube material need not be known. The equation used for calculating surface-tension values from maximum bubble-pressure measurements is the Schroedinger equatioq5 which can be written as follows:

Jan 1, 1964

By R. G. Ward, P. L. Sachdev

Using an improved maximum bubble pressure technique, the densities of iron silicates at 1400°C have been measured under nitrogen. At the wiistite composition the density has been measured by bubble-blowing experments with both argon and nitrogen and calculated from the measured volume of a known mass of melt under nitrogen. The densities under nitrogen are significantly different from those measured under argon for melts containing less than 28 wt pct SiO2. THE constitution of liquid slags is not well understood, largely because no single property or parameter can be measured which gives conclusive evi- dence of the types and packing of the ions present. Of the many property measurements which can provide information leading to a picture of the constitution of slags, density gives the most direct evidence and has been studied extensively. The work reported here is part of a program of study of densities in multicomponent melts in which the iron oxide-silica system is the solvent. These melts are important as they are fundamental to basic steelmaking, and they allow the study of basic silicate compositions which have hitherto been largely neglected. As considerable disparity exists

Jan 1, 1965

By S. D. Baumer, P. M. Hulme

IN the late thirties, the National Carbide Co. cooperated with C. E. Wood, of the U. S. Bureau of Mines, in his investigation of the relative merits of various desulphurizers, including soda ash, caustic soda, and calcium carbide. Laboratory tests showed that carbide, when it could be made to react, is an excellent desulphurizing agent for molten iron. Sulphur content can be driven to lower levels and higher extractions obtained with carbide than with actionsany of the more common reagents. Wood's results1 are shown in Table I. Unfortunately, as the Handbook of Cupola Operation puts it, the chemical fact that carbide is a good desulphurizer was of only academic interest because it was found to be extremely difficult to devise a practical means to make it react with molten iron. Calcium carbide is formed in the electric furnace at 4000°F and above, and its softening point is probably at least 500 °F above the usual working temperatures encountered in iron and steel practice. Consequently, carbide does not form a true slag but floats as a dry powder on top of the metal and only a very small portion of it ever comes in actual contact with the iron. Stirring with a rabble, or pouring the metal over the carbide, increases the efficiency only slightly. Extractions of 20 to 30 pct can be obtained in this manner, but conventional soda slag treatment can do better than this and do it more cheaply. All attempts to lower the melting point of carbide in order to obtain a reactive, liquid slag have so far proved fruitless. Directly under the arc in a metallurgical electric furnace, carbide becomes highly reactive. Excellent sulphur removal can be obtained without any slag other than a thin layer of carbide." imilarly, good results are obtained by adding small amounts of carbide to the finishing slag in double-slag arc furnace practice. To react a liquid with a solid, it is axiomatic that the liquid has to wet the solid before anything can happen. If the solid is heavier than the liquid, the problem is easy, but it becomes more difficult when the solid is much lighter than the liquid, as in the case of carbide and liquid iron. Wood recognized this problem and solved it in a unique fashion. The results shown in Table I were obtained by spinning the carbide beneath the surface of the molten iron by means of a refractory centrifuge. This technique allowed each particle of the finely divided carbide to come into intimate contact with the metal and to be wetted thereby. Wood's centrifuge technique was successful in the laboratory where it achieved excellent and consistent results. Some attempts were made to expand this method to commercial practice, but serious difficulty was encountered in obtaining a refractory centrifuge head that would be economically feasible. About this time the war intervened and the project lay dormant for several years. In 1944, it was revived. It was suggested that the carbide could be blown into the metal with a carrier gas in an attempt to eliminate the necessity for the expensive and brittle centrifuge. The idea was first tried out in a fairly large ladle of iron using natural gas as the carrier. Considerable sulphur was removed, but it was quite obvious that the use of natural gas was not practical. Attempts then were made to blow carbide into molten iron using, in turn, nitrogen, argon, carbon dioxide, air, and oxygen. The latter two gases proved unsatisfactory. Calcium evidently prefers oxygen to sulphur because in the tests calcium oxide and carbon dioxide were produced, the sulphur still being untouched in the iron. Nitrogen, argon, and carbon dioxide gave much better results, although the efficiencies and extractions were erratic, and only a few isolated tests approached the results obtained by Wood. Table II shows typical results obtained with these gases. The sulphur removals were interesting, sometimes even encouraging, but it is evident that such erratic behavior could not be tolerated in commercial practice. A number of different types of equipment, such as sand blasting machines, refractory guns, and the like can used to blow the solid into the metal. All types required relatively large quantities of gas in order to maintain the flow of solid carbide through the system and into the metal. It was observed that the bubbles of gas breaking through the surface of the metal contained quantities of unreacted carbide. The liquid metal never came in contact with these particles and if it cannot wet them it cannot react with them. The initial work had shown that carbide had great possibilities as a desulphurizer. In practice

Jan 1, 1952