Search Documents

By C. D. King

DURING the 30-year period spanned by these annual Howe Memorial presentations, many lecturers could proudly claim a kinship either as a student or an associate of the man whose memory we honor. Although it has been my good fortune to have attended many of these annual lectures, it was not my privilege to have known Henry Marion Howe personally. However, his great repute as teacher and scientist was known to all undergraduates of my day and the later years have enhanced my appreciation of his wisdom and foresight. Those who knew him well have said he derived particular pleasure from speculations on the future world of metallurgy. For this reason, I feel that perhaps he would not be unsympathetic to a lecture in his honor which departs from the highly instructive scientific presentations made in the past by so many able Howe Memorial lecturers, and which is concerned more with the practical phases of various steelmak-ing processes and some speculations on their future form and relative importance. The word "revolutionary" is frequently applied to each seemingly important improvement in the production of steel ingots, but in retrospect these changes, impressive as they appear at the time, are merely steps of progress. In the hundred years from the inception of tonnage steelmaking, only three processes can be truly classified as revolutionary. They are the pneumatic process, known in this country as the bessemer process; the reverberatory method called the open hearth process; and, the electric furnace process. There have been many variations and combinations of the three fundamental methods, but they remain truly the only revolutionary methods in steelmaking since its early history. Everything else has been evolutionary, in effect. doing the same things that we have done in the past but doing them better, correcting our errors through experience, and slowly but inevitably reaching a higher state of accomplishment. It has often been said that coming events cast their shadows before, and the production of steel ingots is no exception. As a result of the unrelenting demands of World War I1 and the years that followed, truly impressive progress has been made in steel ingot production. The incessant pressure for immediate results during this period required the employment of initiative and daring, as in few past decades, and many developments were brought to fruition. Of equal importance is the possible effect on future steelmaking methods of the many ideas initiated but still in formative stages. Fig. 1 portrays ingot production in the United States by the three fundamental processes over a period of 75 years and is interesting because it poses some questions as to future trends. The early ascendancy of the bessemer, its replacement in importance by the open hearth process, the amazing growth of the latter, and the recent challenge of the electric furnace are evident from the chart. Management is fully aware of these changes, but is even more interested in the future trends. Our concepts of the relative importance of the more recent developments and their possible effect on future processes may perhaps be best exemplified by a specific, hypothetical problem. Let us assume management is contemplating a new ingot producing plant with an output of 100,000 net tons per month, located in an area where some purchased scrap may be obtained but where by far the largest component will be own-produced blast furnace iron. Management requires a process or combination of processes which will yield highly uniform quality characteristics in the ingot form, and represent the soundest selection in investment and operating cost. Under these conditions, the obvious selection for the past four decades has been the open hearth process but, in view of more recent developments, management may believe that it is no longer permissible to disregard other possibilities with impunity. Accordingly, to be assured of the best possible selection, they request that you review not only the possibilities of utilizing the conventional open hearth, duplex, bessemer, and electric furnace methods, but also the more recent developments, such as the turbo-hearth, the Linz-Donawitz method, the Perrin modifications, and other possibilities. With this background, one might then appraise the relative importance of these methods to meet a specific need, and concurrently speculate on the forms that future ingot processes will assume and the relative importance of these processes.

Jan 1, 1955

By G. Sachs, A. W. Dana, C. C. Chow

A great number of the investigations on the plastic flow of metals have been concerned with the establishment of a "universal" stress-strain relation. In such a relation some stress function when plotted against a strain function should yield identical curves for the various stress states. In the first investigation of this type, Ludwik and Scheu1 plotted the maximum shearing stress as a function of the maximum principal strain. Later Ros and Eichinger2 introduced two universal stress-strain relations, the one relating the maximum shearing stress to the maximum shearing strain, and the other relating a stress invariant, suggested by von Mises and Haigh, to the corresponding strain invariant. (In more recent investigations the stress and strain invariants are frequently supplemented with some factor to render their meaning more lucid.) A further suggestion which has not attracted appreciable attention is that by Baranski³ who used stress and strain deviators. The most common means of experimentation to determine the relation between stress and strain consists in subjecting thin walled tubes to combined internal pressure and axial tension.4a,4b,4c This method allows the study of plastic flow under stresses which are variable in two directions. However, the plastic flow which can be obtained in this manner is comparatively small, being limited by either tension failure or instability. For copper,'. only the relation between maximum shearing stress and maximum shearing strain yielded good agreement. On the other hand, tests on a stee14b and on an aluminum alloy4c. resulted in systematic deviations if any of the discussed universal stress-strain relations were used. It would seem, therefore, that the agreement mentioned above for copper is only incidental and explained by its high rate of strain hardening compared to that of other metals. Much larger strains than experienced in the tube tests can be obtained by subjecting a thin membrane of a ductile metal, which is restrained at its periphery, to a uniform hydraulic pressure. The thin sheet forms a deep bulge before it fails. The stresses and strains in such a bulge increase with increasing distance from the edge of the clamping "die," the maximum stresses and strains occurring at the pole (crown) of the bulge. While the stress and strain states are determined by the contour of the bulge, the absolute magnitude of the stresses and strains depends upon the hydraulic pressure. The bulge contour is in turn correlated with the geometry of the die opening. The deformation and fracture characteristics of circular bulges, that is, bulges formed with circular clamping dies, have been the subject of numerous experimental and analytical investi-gations.5,6,7 It has been shown that plastically deformed circular bulges develop large and comparatively uniform strains before failure by instability"6b,6c,6d and closely assume a spherical shape.6d Also the distribution of strains across the contour of the bulge is dependent on the metal being investigated and is correlated with, but cannot be predicted from, the metal's stress-strain characteristics. On the other hand, oblong or elliptical bulges, that is, bulges formed with elliptical clamping dies, are not as susceptible to analytical analysis and have not been investigated to the extent that circular bulges have. The few available data6c,7c indicate that stress states are obtained at the poles of the bulges, varying between plane strain and balanced biaxial tension, depending upon the geometry of the die opening. In this paper, the strain state and curvatures exhibited by three bulge shapes, a circular and two elliptical bulges, Fig 1, are analyzed experimentally using methods described in previous publications.6a,6c An attempt is made to derive the stress-strain relations for these bulges, which represent strain states in which the ratio of the two positive principal strains varied between 1.0 and 0.35. In addition, tension tests yielded data for a value of —0.5 for this strain ratio. Such an analysis should indicate the applicability of the various laws correlating stress with strain to the stress and strain states occurring in bulged shapes. Definitions and Nomenclature The definitions of the major stress and strain quantities used in this paper are as follows: s1, s2, s3 = principal normal stresses Sl > s2 > S3 t = shear stress e = conventional (unit) strain e = In (1 + e) El, E2, E3 = principal natural strains 7 = shear strain The maximum shear stress: , _ S1 — S3 lmax = 2 Frequently, the flow stress, s1 — s3 = 2lmax rather than the maximum shear stress is used.

Jan 1, 1950

By John Henderson

FOR some time now the most commonly accepted description of liquid silicate structure has been the "discrete ion" theory, proposed originally by Bockris and owe.' This theory is that when certain metal oxides and silica are melted together, the continuous three dimensional silica lattice is broken down into large anionic groups, such as sheets, chains, and rings, to form a liquid containing these complex anions and simple cations. Each composition is characterized by "an equilibrium mixture of two or more of the discrete ions",' and increasing metal oxide content causes a decrease in ion size. The implication is, and this implication has received tacit approval from subsequent workers, that these anions are rigid structures and that once formed they are quite stable. The discrete ion theory has been found to fit the results of the great majority of structural studies, but in a few areas it is not entirely satisfactory. For example it does not explain clearly the effect of temperature on melt structure,3 nor does it allow for free oxygen ions over wide composition ranges, the occurrence of which has been postulated to explain sulfur4 and water5 solubility in liquid silicates. In lime-silica-alumina melts the discrete ion theory is even less satisfactory, and in particular the apparent difference in the mechanism of transport of calcium in electrical conduction8 and self-diffusion,' and the mechanism of the self-diffusion of oxygen8 are very difficult to explain on this basis. By looking at melt structure in a slightly different way, however, a model emerges that does not pose these problems. It has been suggested5" that at each composition in a liquid silicate, there is a distribution of anion sizes; thus the dominant anionic species might be Si3,O9 but as well as these anions the melt may contain say sis0:i anions. Decreasing silica content and increasing temperature are said9 to reduce the size of the dominant species. Taking this concept further, it is now suggested that these complexes are not the rigid, stable entities originally envisaged, but rather that they exist on a time-average basis. In this way large groups are continually decaying to smaller groups and small groups reforming to larger groups. The most complete transport data 8-10 available are for a melt containing 40 wt pct CaO, 40 wt pct SiO2, and 20 wt pct Al2O3. Recalculating this composition in terms of ion fractions and bearing in mind the relative sizes of the constituent ions, Table I, it seems reasonable to regard this liquid as almost close packed oxygens, containing the other ions interstitially, in which regions of local order exist. On this basis, all oxygen positions are equivalent and, since an oxygen is always adjacent to other oxygens, its diffusion occurs by successive small movements, in a cooperative manner, in accord with modern liquid theories." Silicon diffusion is much less favorable, firstly because there are fewer positions into which it can move and secondly, because it has the rather rigid restriction that it always tends to be co-ordinated with four oxygens. Silicon self-diffusion is therefore probably best regarded as being effected by the decay and reformation of anionic groups or, in other words, by the redistribution of regions of local order. Calcium self-diffusion should occur more readily than silicon, because its co-ordination requirements are not as stringent, but not as readily as oxygen, because there are fewer positions into which it can move. There is the further restriction that electrical neutrality must be maintained, hence calcium diffusion should be regarded as the process providing for electrical neutrality in the redistribution of regions of local order. That is, silicon and calcium self-diffusion occur, basically, by the same process. Aluminum self-diffusivity should be somewhere between calcium and silicon because, for reasons discussed elsewhere,' part of the aluminum is equivalent to calcium and part equivalent to silicon. Consider now self-diffusion as a rate process. The simplest equation is: D = Do exp (-E/RT) [I] This equation can be restated in much more explicit forms but neither the accuracy of the available data, nor the present state of knowledge of rate theory as applied to liquids justifies any degree of sophistication. Nevertheless the terms of Eq. [I] do have significance;12 Do is related, however loose this relationship may be, to the frequency with which reacting species are in favorable positions to diffuse, and E is an indication of the energy barrier that must be overcome to allow diffusion to proceed. For the 40 wt pct CaO, 40 wt pct SiO2, 20 wt pct Al2O3, melt, the apparent activation energies for self-diffusion of calcium, silicon, and aluminum are not significantly different from 70 kcal per mole of diffusate,' in agreement with the postulate that these elements diffuse by the same process. For oxygen self-diffusion E is about 85 kcal per mole,' again in agreement with the idea that oxygen is transported,

Jan 1, 1963

By Gordon H. Geiger, J. Bruce Wagner

Electrical conductavity and therrnoelectvic nzeasuretnents on synthetic hematite slabs with 0.01 to 0.90 at. pet Ti and with 0.30 at. pet Ca were made. Additions of titanium and of calcium both increased the electronic conductivity with titanium yielding greater n-type behavior and calciurn p-type behavior. Reduction studies of hematite to magnetite in CO-CO2 atmospheres between 600° and 900°C initially exhibited linear kinetics. These linear rates were increased by additions of titanium and decreased by calcium. IN recent years many studies have been reported in the literature on the kinetics of the reduction of hematite and magnetite to iron. In addition, Hansen and Josephl and Schenck and Schultz2 have reported studies on the reduction of hematite to magnetite, the first step in total reduction to iron. In order to evaluate the results of the studies of over-all reduction in terms of the actual mechanism of the reaction and to define further the important factors in the rate- controlling mechanism of the magnetic roasting reaction, the following experiments were undertaken. Specimens of undoped hematite and hematite containing titanium or calcium in solution as doping agents were reduced to magnetite in CO-CO2 mixtures. In addition, electrical-conductivity measurements were correlated with the reduction kinetics using concepts of defect solid-state chemistry. EXPERIMENTAL Fisher certified Fe2O3 powder was used to make specimens for both the electrical and kinetic studies (see Table I for analysis). Doping with titanium and calcium was accomplished by adding TiO2 and CaO powders to the Fe2O3 powder. The powders were mixed in a Waring blender for minimum times of 1/2 hr. Satisfactory blending was achieved based on comparison of analyses of individual specimens af- ter reduction. The specimens were formed in a steel die using opposed punches with a rectangular opening 10.0 by 40.0 mm. The Fe2O3-TiO, platelets were pressed cold at 16,000 psi. They were sintered for approximately 16 hr at 1075°C in air, in accordance with information in the literature3, 4 on the time and temperature needed to obtain solution of the titanium in the hematite. The Fe2O3-CaO mixtures were first cold-pressed at 20,000 psi and sintered for 3 hr at 1225°C to insure solution of CaO. This material was then ground to -325 mesh in a 6-in. ball mill and the resulting powder was used to make platelets in the same manner as above. The sintered platelets were usually between 1/16 and 1/8 in. thick. They were ground on a carborundum wheel and polished on 2/0 paper. Typical dimensions of the kinetic-study specimens were 1.6 by 0.8 by 0.05 cm, and a typical weight was 0.3 g. Conductivity and thermoelectric specimens were approximately the same size, but were subjected to further sintering in order to ensure maximum densification. Finally, all specimens were measured to the nearest 0.0005 in. and weighed on a Mettler H-15 balance to the nearest 0.0001 g and the density calculated. The average density of the specimens used for the reduction studies was 4.87 g per cc, or 7.1 pet total porosity. The standard deviation ranged from 0.14 to 0.063 g per cc within a composition group. In the present study, the apparent surface area as measured with a micrometer was taken as the true surface area. The chemical analyses for calcium and for titanium within a group showed maximum variations of 50.04 pet from the average values used in Figs. 9 and 11. In order to suspend the specimen from the balance for the kinetic studies, a hole, 0.03 in. in

Jan 1, 1965

By J. M. Dahl, L. H. Van Vlack

STAFF: Editor, Gerhard Derge Associate Editor, Paul G. Shewman Carnegie lnstitute of Technology Schenley Park Pittsburgh, Pa. 15213 Editorial Assistant, M. A. Redmerski Production Editor, Otto T. Johnson THE METALLURGICAL SOCIETY: Jack H. Scoff, President; Richard C. Cole, Post-President; Harold B. Emerick, President-Elect; Ralph L. Hennebach, Treasurer; Robert W. Shearman, Secretary; Donald A. Parks, Assistant Secretary; Leslie S. Wilcoxson, Assistant Secretary. PUBLICATIONS COMMITTEE OF THE METALLURGICAL SOCIETY: R. E. Lund, Chairman; T. E. Dancy, H. B. Emerick, P. R. Gouwens, R. E. Grace, J. H. Jackson, H. H. Johnson, A. E. Lee, Jr., R. R. Nosh, R. D. Pehlke, and Paul Queneau. Review and selection of monuscripts is under the supervision of the following Publications Committees: Extractive Metollurgy Division: R. E. Grace, Chairman; G. M. Bell, D. W. Bridges, J. M. Cigon, G. W. Healy, T. A. Henrie, R. D. Holli-day, C. E. O'Neill, C. S. Sami s, A. W. Schlech-ten, and Charles Shull . Iron and Steel Division: R. D. Pehlke, Chairman; T. P. Floridis, Arnulf Muon, G. R. St. Pierre, E. T. Tuckdogon, L. H.Von Vlack, and J. M. Uys. Institute of Metals Division: H. H. Johnson, Jr., Chairman; J. B. Clark, Vice Chairman; M. R. Achter, G. F. Bolling, J. O. Brittain, G. P. Conard, II, L. A. Girifolco. G. T. Hahn, J.P. Hirth, L. Himmel, G. T. Horne, George Krauss, C. F. Leitton, Jr., J. D. Livingston, R. D. Monning, H. M. Otte, S. V. Radcliffe, R. E. Reed-Hill, H. C. Rogers, O. Sherby, M. T. Simnod, G. R. Speich, E. A. Steigerwold, R. A. Swolin, W. A. Tiller, W. R. Upthegrove, J. B. Wagner, Jr., Derek Walton, H. W. Weart, W. W. Webb, and D. WiIsdorf. Published monthly by the American lnstitute of Mining, Metollurgicol, and Petroleum Engineers, Inc., 345 East 47th Street, New York, N. Y. 10017. Telephone PLAZA 2-6800. Subscription $30 per year for non-AIME members in United States and North, South, and Central America; $32, foreign; $7.50 for AlME members. Single copies, $6.00, single copies foreign, $6.50. The AlME is not responsible for any statement mode or opinion expressed in its publications . . . Copyright 1965 by the American Institute of Mining, Metallurgical, and Petroleum Engineers, Inc. . . . Registered cable address, AlME New York . . . Indexed in Engineering Index, lndustrial Arts Index, and Chemical Abstracts . . Second class postage paid at New York, N. Y., and Ann Arbor, Mich. Volume 233, Number 1.

Jan 1, 1965

By J. M. Dahl, R. J. Warrick, O. K. Riegger, H. Van Vlack

A liquid which is rich in oxygen (and silicon) develops at steel rolling temperatures in resulfurized and plain-carbon steels. This liquid fluxes solid manganese sulfide. The composition of the liquid and the resulting microstructures vary with the manganese/silicon/oxygen ratios in the steel. The observations and conclusions zuhich are presented are used 1) to suggest a mechanism for MnS deformation in resulfurized steels, and 2) to rationulize hot-shortness anomalies between plain-carbon and resulfurized steels. 1 HIS is a summary of the dependence of sulfide inclusion microstructures upon steel composition, and an interpretation of the behavior of sulfide inclusions at steel-rolling temperatures. Sulfide inclusions are known to have several metallurgical effects upon steel. Their effects upon hot-shortness are best appreciated. Although hot-shortness is well known, it is questionable whether its mechanism is fully understood because there are several anomalies in our interpretation. For example, why should not high sulfur contents produce extensive hot-shortness in free-machining steels ? A second metallurgical consequence of sulfide inclusions appears because sulfide inclusions improve machinability. Furthermore, globular inclu- sions are to be desired over elongated inclusions. To date, we know that lower silicon contents (< 0.010 pct) favor the globular sulfides, but we do not know why this interrelationship between silicon content and inclusion shape exists. Finally, the evidence is strong that the surface quality of a hot-rolled steel is particularly sensitive to the sulfur content. SUMMARY OF PREVIOUS WORK The role of sulfur in hot-shortness was first suggested during the past century. It is uncertain who was the first to conclude that sulfur was less deleterious in the presence of manganese, and that manganese altered the sulfide phase from a liquid FeS to a solid MnS at steel-rolling temperatures. However, Urban and chipman&apos; emphasized that liquid FeS may accumulate as an interdendritic film during solidification. More recently, Keh and Van lack&apos; showed that the intergranular liquid does not completely surround the metal grains at all rolling temperatures. Rather, a small finite dihedral angle exists when geometric equilibration is attained at

Jan 1, 1962

By E. T. Turkdogan, L. S. Darken

Calcium ferrite melts were equilibrated with sulfur and oxygen-bearing gases at temperatures within the range 1290°C to 1620°C. The results show that at oxygen partial pressures below 10-4 atm the sulfide-reaction occurs and at higher oxygen pressures the sulfur dissolves in the melt principally as sulfate ion. It is also observed that when the product increases, sulfate ions are converted to pyrosulfate ions and the results permit the evaluation of the pyrosulfate/sulfate ratio. It is shown that the ratio y CaSO,/rCaO increases with decreasing concentration of calcium oxide in calcium ferrite, silicate and aluminate melts. Indications are that oxy-acids, e.g. silica, alumina and ferric oxide, increase yCaS0,. A few experiments were carried out within the sul-tide range and the results are in accord with those obtained by other investigators on aluminate and silicate melts. BECAUSE of the industrial importance and theoretical interest, many investigators have studied the reactions between molten oxides, silicates or alumi-nates, and gases containing oxygen and sulfur. The early work reviewed by Schenckl and studies made by Bardenheuer and Geller2 indicates that solution of gaseous sulfur in an oxide melt resulted in the replacement of the oxide ions by the sulfide ions. Measurements made by Grant and Chipman3 on the equilibrium partition of sulfur between molten iron and complex silicate melts also demonstrated that, for a given temperature and silicate composition, the distribution of sulfur between slag and metal increased with decreasing oxygen content of the metal. Fincham and Richardson4 studied the effects of temperature and partial pressures of sulfur and oxygen on the solubility of sulfur in simple silicate and alumino-silicate melts. They observed that at oxygen partial pressures below about 10-5 atm sulfur in the gas dissolved in the melt as sulfide ions, replacing an equal number of oxide ions of the melt. At oxygen partial pressures higher than 10-3 atm, sulfur entered the melt as sulfate ions. Similar studies were made by St. Pierre and Chipman5 using calcium ferrite and calcium silico-ferrite melts. The experimental data presented in this paper were obtained in this Laboratory about 10 years ago; this is the first formal presentation although a preliminary report was given previously.6 In each of these experiments, a pure synthetic CaO-Fe2O3 melt in a platinum crucible was equilibrated with a gas mixture (SO2-O2, SO,-air, SO2-CO, or SO2-CO2-CO) at a temperature ranging from 1296°to 1620°C. The experimental technique was essentially the same as that reported by Darken and Gurry,7 the main difference being that the atmosphere in the reaction chamber contained sulfur dioxide together with oxygen, or carbon dioxide-carbon monoxide

Jan 1, 1962

By George R. St. Pierre, John Chipman

METALLURGISTS have been studying the chem-ical behavior of sulfur in steelmaking for many years in order to have a better control of the sulfur content of finished steel. During the refining period in the open hearth, molten steel is separated from a gas phase by molten slag. Chemical reactions occur principally at the slag-metal interface and at the slag-gas interface. A need for physical-chemical investigations of reactions between sulfur-bearing gases and steelmaking slags was recognized some time ago.&apos;.&apos; From the early work, it can be qualitatively concluded that increased sulfur content and low partial pressure of oxygen in the gas phase favor the passage of sulfur from gas to slag. Using radioactive sulfur S Koch and Fink&apos; have demonstrated the rapid rate of exchange of sulfur between gas and slag in an open hearth furnace. Gurry and Darken and Darken and Shields (as reported by Derge and Marshall&apos;) have equilibrated lime-iron oxide slags with SO—O2 gas mixtures. The major conclusion from the reported results is that sulfur content increases with increasing lime content in lime-iron oxide slags equilibrated with a gas of high oxygen pressure (10-2). The data were interpreted as indicating the presence of SO, and S,O, ions. Richardson and Fincham&apos; have conducted the most extensive experimental study of sulfur equilibria between gas and slag. Their results confirmed many of the conclusions drawn from calculations of Richardson" and Withers. Richardson and Fincham&apos;s data fitted very well into the sulfur scheme represented by the reactions 1/2 s2 (g) + (0) melt - 1/202(g)+ (S) melt [II 1/2 S, (g) + 3/2 O (g) + (0) melt - (SO,) melt. [2] At oxygen pressures greater than 10 to 10 they found that reaction Eq. 2 was predominant while reaction Eq. 1 was controlling at oxygen pressures less than 10." to 10.". The slag compositions and gas mixtures used in their investigations were different from those used in the present study. Wherever possible, comparison has been made between their results and these, and agreement is excellent. It is evident that the distribution of sulfur between slag and gas depends upon the oxygen pressure of the gas. In FeO slags, a change in oxygen pressure produces a change in ferric oxide content, and a knowledge of this equilibrium is a prerequisite to studies of sulfur distribution. The data are found in the work of Darken and Gurry" on the FeO system and the more recent experiments of Gurry and Darken" and of Larson and Chipman" on FeO slags containing lime and silica. The effect of oxygen pressure on the ratio Fe /(Fe" + Fe ), the so-called j-ratio, reported by these investigators will be used in interpreting the results which follow. Experimental Method The purpose of the experimental program was to determine the equilibrium sulfur content in slags exposed to gases of known oxygen and sulfur pres- sures. The slag compositions selected for study were from the lime-silica-iron oxide system. Additions of magnesia were made to some of the slags. Apparatus—The apparatus used by Larson and Chipman was modified slightly to accommodate the addition of SO2 to the gas mixtures. As in their work, slag samples of 0.5 to 2.0 g were held at temperature (lcc) in platinum crucibles under a steady stream of gas, 100 to 200 ml per min, for several hours and then rapidly lowered to the bottom of the furnace where they were quenched in mercury. Preparation and Analysis of Gas Mixtures—Mixtures of SO,-CO were prepared from Virginia Smelting Co. Extra Dry liquid SO, and Pure Co. CO,. After removing traces of water vapor with an-hydrone, the CO, was passed over graphite at 1200°C, at which temperature the conversion to

Jan 1, 1957

By J. Chipman, G. G. Hatch

One of the important functions of the iron blast furnace is the desulphur-ization of pig iron before it enters the steelmaking furnaces. However, the increasing concentrations of sulphur in the metallurgical coke, source of approximately 90 pct of the sulphur present in the blast furnace charge, and demands for higher rates of production within recent years have increased the need for greater desulphurization within the iron blast furnace. Furnace operators are beginning to look for desulphurizing agents other than blast furnace slag to accomplish the desired degree of desulphurization. A considerable amount of work has been done on desulphurization outside the furnace with soda ash, calcium carbide and various synthetic slags. Whether the desulphurization of pig iron is accomplished wholly inside the furnace or partly inside and the remainder outside, will be determined by the economics involved. Regardless of which is the case, it is believed that it is necessary to have a better understanding of the physical chemistry of desulphurization by blast furnace slags. To this end, it is the object of the present investigation to attempt what is believed to be the first equilibrium study of the distribution of sulphur between liquid pig iron and a wide range of blast furnace slag compositions. Review of Literature There is a considerable amount of information in the literature concerning the desulphurizing power of iron blast furnace slags, the solubility of various sulphides in the slags, and the effect on desulphurization of temperature, of elements dissolved in the liquid iron, and of viscosity. However, there is nothing to indicate that the equilibrium distribution of sulphur between liquid iron saturated with carbon and iron blast furnace slags has been studied experimentally. Wentrupl has made probably the most detailed study of the desulphurization of pig iron to date. He considered that there are three distinct aspects involved, namely: 1. Desulphurization within the blast furnace (by lime and manganese). 2. Subsequent desulphurization by manganese. 3. The effect of subsidiary reactions on the desulphurization by manganese. The experimental work carried out by Wentrup was devoted mainly to obtaining a better understanding of how desulphurization by manganese was accomplished in the mixer and the ladle. Particular attention was given to the part played by carbon, silicon, and phosphorus associated with manganese in the iron, and the effect of temperature on desulphurization. The experimental results indicated that desulphurization by manganese is purely a process of crystallization of manganese sulphide. The addition of silicon to iron melts containing 3.5 pct carbon and less than 0.5 pct manganese had no noticeable effect on desulphurization, but with 1-2 pct manganese the silicon additions improved the desulphurization. Additions of phosphorus also resulted in improved desulphurizati011 by manganese, but the effect was not as marked as in the case of silicon. It was also found that desulphurization by manganese was further improved by lowering the temperature. In order to explain desulphurization inside the blast furnace, Wentrup considered the system iron, sulphur, calcium, oxygen, manganese. (silicon). The distribution of sulphur between the metal and slag was represented by the following equation: (SS) _ (S)Fe + (S)Ca + (S)Mn .... [S] = [s] [1] The parentheses and the brackets represent the equilibrium concentrations in weight per cent of the slag and metal constituents, respectively. Since FeS D (FeS) _ (FeS) (S)Fe LfeS - [FeS] [S] [2] (CaO) + S e (FeO) + (S)Ca _ (FeO)(S)Ca. (S)Ca _ (CaO) Kl = (CaO)[S] [S] ~K1(FeO) [3] Mn + S D (S)Mn (S)Mn (S)Mn K&apos; = [MnpT "1ST = *lIMnJ !4) Substitution of Eq 2, 3, and 4 into Eq 1 resulted in if = L- + *> (Sol + K^ (S) [51 Eq 5 was used to calcu1;lte -f^j and [S] at 1480°C for slags containing 30-50 pct lime, 0.1-2.5 pct iron oxide, 0-26 pct silica, 2 pct sulphur and iron analyzing 1.5 pct manganese. The value for LFaB at 1480°C was found to be equal to 4.5, based on the experimental work of Bardenheuer and Geller.2 The results of the calculations are shown in Table 1. Although the slags are hypothetical and do not represent the range of compositions found in ordinary blast furnace practice, the calculations indicate that lime is effective in controlling desulphurization only if the iron oxide and silica contents of the slag are kept low. Schenck3 did not claim K1 to be a true equilibrium constant, but an empirical value which varied with the silica content of the slag.

Jan 1, 1950

By R. E. Boni, G. Derge

Application of surface tension measurements has been made to molten silicates in order to determine the effect of sulphur upon the surface tensions of synthetic blast furnace slags. In melts with the same molar ratios of AI2O3 + SiO2 to the basic oxides BaO, CaO, and MgO, the surface tensions increase in the order of oxides listed. Sulphur lowers the surface tensions of the basic BaO, basic CaO, acid BaO, acid CaO, and acid MgO slags in that order; and the concentration of excess sulphur at the slag surface increases in this same order of slags. The rate and amount of desulphurization of iron by these slags is favored by a low excess surface sulphur in the slag. IN a previous paper,1 the authors discussed the applications of surface tension measurements of molten silicates to problems of metallurgical interest. Of major importance is the aid that such studies lend to forming a working concept of slag structure. Since the ability of a slag to desulphurize metal depends strongly upon its composition,2"4 fundamental information about the structural differences between various slags which contain sulphur should aid the development of better desulphurizing media. For that reason, the present investigation concerns the effect of sulphur upon the surface tensions of synthetic blast furnace slags. Experimental Procedure , The maximum bubble-pressure method was used to measure the surface tensions of synthetic blast furnace slags in the temperature range 1300° to 1600°C. The method is based upon the equation which was developed by Schrödinger5 from a consideration of the forces of hydrostatic pressure and surface tension of a liquid which oppose bubble formation when a gas is forced through a tube immersed in the liquid. where y is the surface tension between argon and slag in dynes per centimeter; g is the gravitational constant, centimeters per second2; H is the manometer reading in centimeters; s2 is the density of manometer fluid in grams per cubic centimeter; S1 is the density of the medium in grams per cubic

Jan 1, 1957

By T. J. Whalen, S. M. Kaufman

Surface-tension measurements for liquid Fe-C-Co and Fe-C-Ni alloys were performed by the sesszle-drop technique at 1350o and 1425°C. Cobalt was shown to he more effective than nickel in lowering the surface tension of Fe-C alloys A minimum value for surface tension of 1640 dynes per cm was found in the Fe-C-Co system at a concentration of 1.5 pet C and 12pet Co. Analysis of these measurements in terms of adsorption and desorption behavior indicated that cobalt and nickel lower surface tension in Fe-C alloys by replacing carbon at the liquid-vapor interface. Comparison of results to date for Fe-C liquid alloys has shown that most substitutional-type elements having about the same atomic diameter as iron also have relatively the same influence on surface tension when present in dilute amounts. It has been shown previously that surface-adsorption behavior in liquid Fe-C ternary alloys is related to the manner in which the ternary addition affects the activity of carbon in the bulk solution.&apos; Certain ternary additions, such as chromium, tending to decrease carbon activity cause adsorption of carbon and chromium to the liquid-vapor interface. Additions, such as silicon, increase the carbon activity, resulting in a desorption of carbon and an adsorption of silicon. Both cases were shown to cause a decrease in surface tension due to adsorption. It is the object of this paper, as part of the writers&apos; continuing interest in the effect of alloying additions on the surface tension of Fe-C alloys, to present and discuss surface-tension data on solutions in the ternary systems Fe-C-Co and Fe-C-Ni. Available activity measurements for these systems in the solid state indicate a strong increase in carbon activity when either nickel or cobalt is added to Fe-C alloys.2, 3 EXPERIMENTAL PROCEDURE The alloy compositions used are shown in Table I. These alloys were prepared by vacuum melting in stabilized zirconia crucibles and then vacuum casting into cold-steel molds. The nominal analyses for the alloys used in this investigation are given in Table I. Surface tensions and densities were measured by the sessile-drop method4,5 in conjunction with the tables of Bashforth and bdams.&apos; Specimens were prepared by slicing 1/4-in. cubes from the center of the vacuum-cast ingots and grinding the edges of these cubes until their shapes approximated spheres with flat bottoms. Proper handling techniques, including an acetone rinse immediately before use, assured minimum contamination from sources extraneous to the actual melting. For each experimental run a sample was placed on a BeO plaque and positioned and leveled inside a vacuum induction heating apparatus. The sample and plaque arrangement sat inside a molybdenum tube which acted as the susceptor. Diagrams of this apparatus may be found in the literature.= After heating to temperatures just above the liquidus at pressures of the order of 10-4 to 10- 5 mm Hg, photographs of the drop profile were made using its natural thermal radiation as a light source. Some measurements were made with partial pressures (-1 cm Hg) of a mixture of H2 and He gases. The H2-He atmospheres proved to be necessary at higher nickel contents to retard excessive vaporization losses. Chemical analysis for carbon and the ternary addition was performed for each alloy sample after completion of the experimental run.

Jan 1, 1964

By E. N. Silverman

Mixtures of&apos; MnS, ZFeO-SiO,, ZMnO.SiO, FeO, and MnO were prepared synthetically and used to study the planes ZFeO. Si0,-ZMnO.SiO,-MnS, FeO-MnO-MnS, and Fe0-MnO.Si0,-MnS in the quaternary system. Because the oxides and silicates studied flux MnS, inclusions in this system are partly liquid during rolling. Also, oxygen can modify subsurface inclusions through the reaction MnS + FeO—FeS + MnO during the soaking of ingots and the reheating of blooms and billets. NONMETALLIC inclusions in steel have long been a subject of study and conjecture because the properties of many steels are known to be affected by the the kind and amount of inclusions present. In free-machining steels, the type of inclusion present has a direct bearing on their machinability; in other steels, cleanliness is important and metallurgists direct their efforts at reducing the amount of non-metallic inclusions. Thus, the accumulation of knowledge about the formation and behavior of inclusions during all stages of steelmaking and processing should aid in improving the quality of steels and decreasing the production costs resulting from rejections. Previous studies of inclusions have contributed greatly to our understanding. The investigators have generally examined inclusions in finished steel. Boulger, Moorehead, and Garvey1 found that in Bessemer steels, low silicon contents results in globular sulfide inclusions with attendant high machin- ability ratings. Van vlack2 also found that low silicon contents favor inclusion morphology that contributes to high machinability ratings. Boquist and Stoll* found that silicon content affected sulfide inclusions and machinability of blooms, billets and bars more than did the finishing temperature in the range between 1775° and 2075°F. Carney and Rudolphy3 studied the distribution of sulfur, manganese, silicon, and carbon in a commercial ingot of resulfur-ized steel and found that silicon was concentrated near the bottom. This helped to explain why bottom cuts had poorer machinability. Klinger and Koch4 developed techniques for extracting inclusions from steel, and the technique devised by Gurry, Chris-takos, and stricker5 for extracting carbides can also be adapted to the study of inclusions. A different approach to the study of inclusions in steel consists of synthesizing the constituents that form inclusions and determining the phase eauilibria in the system that contains the constituents. The resulting phase diagrams can then be used to determine the effect of the various constituents upon each other. In this way, it becomes possible to under-

Jan 1, 1962

By S. L. Payne, J. P. Morris, R. L. Kipp, G. R. Zellars

The activities of iron and nickel in liquid Fe-Ni alloys were determined by measuring the vapor pressures of iron and nickel above the alloys and comparing with the vapor pressures of the pure metals. Nickel was found to deviate negatively from Raoult&apos;s law over the entire composition range. Iron behaves ideally down to 0.7 mole fraction but shows negative deviations at lower concentrations. At 1600°C the activity coefficients at infinite dilution are 0.40 for iron and 0.66 for nickel. 1 HE formulation of the conditions of chemical equilibrium between liquid metals and gases or slags is an important application of thermodynamics to pro- cess metallurgy. Usually the metal and slag phases are solutions of several components; and for accurate calculations, it is often necessary to use the activities of the components in place of their concentrations. Data concerning activities, however, are quite meager. The evaluation of the activities of iron and nickel, presented in this paper, is part of a project to determine activities in iron alloys at elevated temperatures.

Jan 1, 1960

By N. A. Gokcen, E. S. Tankins, G. R. Belton

Equilibrium in the reaction H2(g) + O[in liquid iron, cobalt, or nickel] = H2O(g) has been investigated over wide temperature and composition ranges. Oxygen has been found to obey Henry&apos;s Law in each of the three metals at 1550°C. The standard free-energy changes for the above reaction are given by: (1550° to 1700°C) ?Foco = -45,360(±290) + 16.16(±0.16)T (1490° to 1700°C) ?Fo Ni = -48,170(±220) + 16.12(±0.12)T (1465° to 1700°C) where the standard state for dissolved oxygen is the I wt pet solution. ThE thermodynamic behavior of dilute solutions of oxygen in molten iron has been the subject of many investigations over the last thirty years,&apos;-&apos;4 the majority of the investigators having utilized the reaction:

Jan 1, 1964

By Robert Baschwitz, John Chipman

The distribution of silicon between liquid silver and Fe-Si-C alloys has been studied at 1420oand 1530°C. The data are consistent with earlier studies. New data of Hager on the liquidus lines of the system Ag-Si and the distribution data are used to obtain the activity coefficient of silicon in both liquid phases. Data on the heat of mixing in iron permit accurate extension to 1600°C. Equilibrium data involving SiO2 and silicon in liquid iron together with revised data on the free energy of SiO2 are used to fix the activity of silicon in the infinitely dilute solution. The binary system exhibits strong negative deviation from ideality. At infinite dilution ? Si at 1600" is 1.25 x 10&apos;3, and at concentrations up to NSi = 0.4 the slope d InySi/dNSi has a constant value of r; = 13. It is found that logysi in the ternary solutzon is approximately but not exactly the same function of Nsi + NC as of NSi in the binary. The results are consistent with currently available data on the free energy of Sic and its solubility in molten iron. LIQUID solutions of the system Fe-Si-C have acquired considerable importance as the laboratory prototypes of blast furnace hot metal. Equilibrium studies involving such solutions and slags approximating those of the blast furnace have yielded useful information concerning the thermodynamic properties of blast furnace slags. In studies of this kind great importance attaches to a knowledge of the thermodynamic activity of silicon in the solution as a function of temperature and composition. An attempt was made by Chipman, Fulton, Gokcen, and askey&apos; to evaluate all of the pertinent data on this system and to deduce the desired relation between activity, composition, and temperature. These authors published data on the solubility of graphite and Sic in molten Fe-C-Si solutions and on the distribution of silicon between liquid iron and liquid silver. They showed further how the activity of silicon in very dilute solutions in liquid iron could be calculated from equilibrium data involving the molten alloy and solid SiO,. These calculations rested on the published thermodynamic properties of SiO, in- cluding its heat of formation which at that time was recorded as -209.8 kcal. This value has been under suspicion for some time and has recently been replaced by the concordant results from two independent laboratories2,3 which place the heat of formation of a-quartz at -217.6 kcal. This revision necessitates a re-evaluation not only of the activity of SiO2 in slag but also of silicon in molten iron. It is the purpose of this paper, therefore, to recalculate the activity of silicon, and in furtherance of this objective to present new data on its distribution between liquid Fe-Si-C alloys and liquid silver. HEAT OF SOLUTION OF SILICON IN IRON In order to determine the effect of temperature upon the activity coefficient it is necessary to know the heat of solution of silicon in iron as a function of composition. This is found in the data of Korber and Oelsen4 shown in Fig. 1. The curve corresponds to the following equation, which is of a form suggested by Wagner:5 Here AH is the heat absorbed in kilocalories in forming one gram atom of molten alloy from its molten elements and the N&apos;s are atom fractions. The relative partial molal enthalpies of the components, each referred to its pureliquid state and defined as zFe = aFe - PFe and zsi = HSi — -psi, are shown graphically. At low concentrations zSi = -28.5 kcal, in agreement with Kijrber and Oelsen&apos;s computation. This is in good agreement with the value of -29.3 kcal obtained by Chipman and Grant6 using an entirely different method. ACTIVITY AT INFINITE DILUTION From the known free energy of SiO, it is possible to obtain the activity of silicon in dilute solution in liquid iron from equilibrium studies. The heat of formation of a-quartz is —217.6 kcal and the heat capacity and entropy data are given by Kelley and ~ing.&apos; The free energy of formation of ß-cristo-balite at temperatures above the melting point of silicon is expressed by the following equation: Si(Z) + O2(g) = SiO2 (crist); AF" =-226,500 + 47.50T [I] The value of the deoxidation product for silicon [%Si] x [%O]2 at 1600°C according to Gokcen and chipmans is 2.8 x 10"5, in agreement with results of Hilty and Crafts.9 More recent works of Matoba,

Jan 1, 1963

By E. M. Cox, T. Rosenqvist

IN a recent investigation, carried out at the Institute for the Study of Metals, the affinity of sul-phur for iron, copper, and manganese was studied over the temperature range 700" to 1300°C.&apos; It was found that copper exhibits a rather high affinity for sulphur, especially at higher temperatures; and attention was called to the possibility that copper when present together with sulphur in liquid steel would tend to lower the chemical activity of sulphur and thus interfere with the desulphurization process. It was clear, however, that the data for the equilibrium between pure copper and pure copper sulphide (Cu2S) could not give any quantitative information about the effect of copper when dissolved in liquid steel. The present work is concerned with the activity of sulphur in the diluted range of the binary iron-sulphur system, and with the influence of small amounts of copper on this sulphur activity. This investigation is in principle based on the same method previously described for the binary metal-sulphur studies;1,2 namely, reaction with hydrogen to form hydrogen sulphide. The equilibrium ratio PH2s/Ph2 in the equilibrium gas mixture is, for a given temperature, directly proportional to the sulphur activity. The activity of sulphur in liquid iron has previously been studied by Chipman and Ta Li,3 White and Skelly,4 and recently by Sherman, Elvander, and Chipman.5 Morris and Williams6 have studied the effect of added silicon, and Kitchener, Bockris and Liberman,7 and recently Morris and Buehl,8 the effect of added carbon on the sulphur activity. In present day steelmaking processes the steel bath contains between 0.01 and 0.1 pct sulphur, and the copper content may be as high as 0.5 pct. The equi-librium gas ratio (H2S/H2) would, for such a low-sulphur content, be too low to be measured with the experimental technique we use. It has been shown by previous investigators that the H2S/H2 ratio is closely proportional to the sulphur content up to about 1 pct sulphur. The sulphur activity at very low-sulphur content can therefore be calculated by linear interpolation between 0 and 1 pct sulphur. In order to show the general trend with increasing sulphur content, the present work has been carried out to about 3.5 pct sulphur. As the effect of copper was assumed to be more a function of the Cu/S ratio than of the absolute copper content, the runs with copper were made for certain constant values for this ratio.

Jan 1, 1951

By Donald B. Evans, Robert D. Pehlke

The solubility of nitrogen in liquid Fe-A1 alloys has been measured up to the solubility limit for formation of aluminum nitride using the Sieverts method. The activity coefficient of nitrogen decreases slightly with increasing aluminum content in the range of 0 to 4 wt pct Al. Based on a nitride composition, AlN, the standard free energy of formation of aluminum nitride from fhe elements dissolved in liquid iron has been determined to be: ?F" = -59,250 + 25.55 T in the range from 1600º to 1750ºC. The solubility of nitrogen in liquid iron alloys and the interaction of nitrogen with dissolved alloying elements in liquid iron have been the subject of a number of research investigations.&apos; Most of this work, however, has been reported for concentrations well below those necessary for the formation of the alloy nitride phase. Data in the concentration region near the solubility limit of the alloy nitride, particularly for systems exhibiting stable nitrides, are important in evaluating the denitrifying power of various alloying elements. They are also useful in determining the stability of a given nitride if it is to be used as a refractory to contain liquid iron alloys. In view of the importance of aluminum as a deoxidizing agent in commercial steelmaking and the fact that its nitride, AIN, is a highly stable compound and has merited some consideration as an industrial refractory, the following investigation was undertaken. The use of the Sieverts technique provided a measurement of the equilibrium nitrogen solubility in liquid Fe-A1 alloys as a function of nitrogen gas pressure up to 3.85 wt pct A1 in the temperature range of 1600º to 1750°C. The values obtained by the Sieverts method were checked by means of a quenching method in which liquid iron was equilibrated with an A1N crucible under a known partial pressure of nitrogen gas, and the solubility of A1N in liquid iron determined by chemical analysis. EXPERIMENTAL PROCEDURE The theoretical considerations involved in determining the solubility product of a solid alloy nitride phase in liquid iron by measuring the point of departure of the nitrogen gas solubility from Sieverts law have been discussed by Rao and par lee.&apos; The principal problem is to determine the variation of nitrogen solubility in an alloy as a function of the pressure of nitrogen gas over it with sufficient precision to establish the break point in the curve at the solubility limit of the alloy nitride phase. A fairly large number of data points are required to do this. A second problem is the determination of the composition of the precipitated solid nitride phase. This is necessary in order to define completely the thermodynamic relationships. The Sieverts apparatus used to make the nitrogen solubility measurements in this investigation is of essentially the same design as that described by Pehlke and E1liott.l The charge materials were Ferrovac-E high purity iron supplied by Crucible Steel Co. and 99.99+ pct pure aluminum. Recrystal-lized alumina crucibles were used, and were not attacked by the liquid alloys. The hot volume of the system which was measured for each melt ranged from 46 to 50 standard cu cm and was found to decrease linearly with decreasing pressure and with increasing temperature. The temperature coefficient of the hot volume at 1 atm pressure of argon gas was essentially constant for all experiments at a value of -6 X 10-3 cu cm per "C. The melt temperature was measured with a Leeds and Northrup disappearing filament type optical pyrometer sighted vertically downward on the center of the melt surface. The temperature scale was calibrated against the observed melting point of pure iron taken as 1536°C. The emissivity of all melts was assumed to be that of pure iron, taken as 0.43. The charge weights ranged from 110 to 140 g and the range of aluminum contents covered was from 0 to 3.85 wt pct. Aluminum additions were made as 12 to 15 wt pct A1-Fe master alloys previously prepared in the system under purified argon. The compositions of the master alloys were checked by chemical analysis and found to be in agreement with the charge analyses. Vertical cross sections of the master-alloy ingots were used as charge material for the equilibrations in order to minimize the effect of any segregation which might have occurred during solidification of the master alloys. Determinations of the solubility product of

Jan 1, 1964

By M. L. Pearce

A comparison of the isotope-dilution, vacuum-fusion, and chemical methods of analysis for nitrogen in Si-Fe is made with particular emphasis on the effect of sample history. The superiority of the isotope-dilution method from the point of view of 1 HE analytical techniques used to determine nitrogen in Si-Fe have been subjected to close scrutiny in recent years primarily because of the marked influence of the element on the properties of the alloy. These investigations have led to a diversity of opin- both precision and accuracy is demonstrated. The solubility of nitrogen in 3.06 pct Si-Fe is investigated and the effect of silicon content on the activity coefficient of nitrogen is derived. ion regarding the reliability of the vacuum-fusion and chemical methods of nitrogen analysis when applied to si- Fe.&apos;-&apos; A majority of workers appear to favor the vacuum-fusion method over chemical methods particularly in those cases where part of the nitrogen may exist as silicon nitride or the total nitrogen content is very low. Contrary views have been expressed, however,&apos; and it has been demonstrated that for some Si-Fe alloys both methods yield results which are significantly lower than those

Jan 1, 1963

By J. M. Toguri, H. Flood

The application of the mass action law to the thermodynamic treatment of metal-slag equilibria is discussed. The consideration of the reactions essentially as ionic exchange equilibria is shown to be both an advantage and an aid for the elucidation of the theoretical treatment. This is illustrated by several examples. In general two different methods of application of the mass action law to metal-slag equilibria have been used by different groups of investigators. These could be called the oxide reaction and the ionic reaction approach respectively. At the recent discussion in the metallurgical section of the I.U.P.A.C. Congress in Montreal, 1961,&apos; divergent opinions prevailed regarding the most rational way of applying the mass action law to metal-slag equilibria. This question is of fundamental importance in the treatment and understanding of theoretical chemical metallurgy. Consequently, the following is an attempt to contribute towards its clarification. As an example, the distribution of iron and manganese between molten iron and slag will be considered. I) THE OXIDE EXCHANGE REACTION APPROACH The equilibrium between iron and manganese between molten iron and slag is frequently formulated as an oxide metal exchange reaction as follows:

Jan 1, 1963

By M. G. Benz, J. F. Elliott

Otto Schaaber (Institut für Harterei-Technik, Bremen -Schönebeck, Germany)— Langenberg, Pestel, and Honeycutt gave an interesting example of the grain refining effect nonstationary magnetic fields may execute on a solidifying metal. Probably some additional comments, based on more than a decade of experience with this method of influencing crystallization, may be useful. They refer to an attempt to explain the mechanism of grain refinement by such fields, and to recent European developments which have not been mentioned in the paper. I. During studies of the temperature distribution in a metal body near the phase boundary solid/liquid, an anomaly was observed under certain special conditions: a) direction of heat transfer as horizontal as possible, b) steady state (equilibrium, i.e,, solidification resp. melting rate exactly zero). In the liquid phase, immediately adjacent to the phase boundary, the temperature was found to be constant over an appreciable distance in spite of a high value of heat transfer (40.000 to 70.000 keal/ m2h).3 Convection, which might explain this phenomenon, could not be observed. Therefore another explanation had to be sought. The assumption that a metallic melt, especially at temperatures near the point of solidification, contains some sort of atomic arrangement with a higher degree of order than the surrounding liquid, now is generally accepted. Considering the differences in energy between such an arrangement and its environs, a transport of heat even in absence of any temperature gradient is conceivable by continued formation and redissolution of regions with a higher degree of order, i.e., with lower energy level.3 At this stage, it was not necessary to postulate any special condition as to the form or size of these regions or the distance between arrangements to be newly formed and those to be redissolved. However, a simple consideration shows that an energy transfer from one region to another will only be possible without loss of energy, if the distance between will be equal to the lattice parameters, and if, secondly, crystallographic axis or planes will be parallel.4 By the way, this assumption might give a reasonable explanation of the well known orientation effect of the heat flow during solidification. Any measures to prevent columnar crystallization should act upon this ordering mechanism: a) the regions with a lower energy (higher order) should not be allowed to arrange themselves parallel to the direction of heat flow, and their growth should be disturbed; b) all conditions leading to an appreciable heat flow of uniform direction should be avoided or removed. Condition a) may be satisfied by a physical force preferably under 90 deg to the main axis of the columnar crystals (i.e., parallel to the solidification front) acting on the ordered regions; condition b) by equalizing any differences in

Jan 1, 1962