THE UNIVERSITY OF MICHIGAN INDUSTRY PROGRAM OF THE COLLEGE OF ENGINEERING THIRIE-ELEMENT INTERACTIONS WITH THE SYSTEM LIQUID IRON - LIQUID CALCIUM David L. Sponseller A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy in the University of Michigan 1962 June, 1962 IP-573

Doctoral Committee Professor Richard Ao Flinn, Chairman Associate Professor Wilbur Co Bigelow Professor Lee 00 Case Professor Robert Wo Parry Professor Clarence Ao Siebert Professor Lawrence Ho Van Vlack ii

PREFACE I wish to express my sincere appreciation to all those who have aided in this investigation, and particularly to the following: Professor Richard A, Flinn, Chairman of the Doctoral Committee, for his guidance and assistance. His many helpful suggestions and continuing encouragement were invaluable factors during this investigation, Professors Wilbur C, Bigelow, Lee 0, Case, Robert W, Parry, Clarence A. Siebert and Lawrence H, Van Vlack, members of the Doctoral Committee for their interest and critical analyses, Professor Paul K, Trojan, whose helpful discussion and assistance greatly facilitated this research. Professors Edward E, Hucke and Charles L, Rulfs, for their comments and suggestions. The International Nickel Company for fellowship support, My fellow students, Mr, William Mitchell, Mr. Henry Kunsmann, Mr, Jon Brokloff, Mr. Gerald Madden, Mr. William Spurgeon and Mr. James Judd, for their generous assistance. Dr, Walter Crafts and Mr. H, Vo Mosby of the Union Carbide Company, Mr, Harold Young and Mro Gordon Meeter of the Cameron Iron Works, and Mr. Robert Grinthal of Technical Research Group, for their interest and assistance, To the Industry Program of the College of Engineering for the reproduction of this thesis, iii

TABLE OF CONTENTS Page PREFACE........................................................... iii LIST OF TABLES......................................... vi LIST OF FIGURES..................... ii.......... ABSTRACT.................................................. 00 ix INTRODUCTION..............................e...... 1 A. Purpose of the Investigation........................ 2 REVIEW OF THE LITERATURE.......................................... 3 A. Experimental Studies.................................. B, Theoretical Considerations.............................. 5 1. Chemical Model...................................... 6 2. Electron Model..................................... 6 EXPERIMENTAL METHOD................... o........................... 9 A. Phase:-Rule Analysis................................... 9 1. Iron-Calcium System Under Argon Pressure............. 10 2. Iron-Calcium-Alloying Element X System Under Argon Pressure............... 0.......................... o 10 3. Effect of Argon Pressure Upon Equilibrium........... 11 B. Pressure Chamber.................... 13 C. Arrangement for Sampling in Place.................. 13 lo Coil Assembly................................... 1 2. Shutter Linkage......................o.............16 3. Sampling Device............................ o 16 D. Preparation of Crucibles......................... 20 1. Performance of Commercial Crucibles.................. 21 2. Selection of a Crucible Material.................... 21 3. The Effect of Titanium Hydride....................... 22 4. Compacting of Crucibles..................... o 22 5. Sintering.........................o 23 6. Nitriding.........................26 7. Performance of Titanium Nitride Crucibles...........o 26 iv

Page E, Temperature Control................oo..o....o...... 27 1. Method of Measuring the Temperature of the Melt.o..... 27 2. Calibrations of Thermocouple Wire and Control Equipment O............................................ 29 35 Accuracy of Temperature Control.....................o 31 F. Conduct of Experimental Heat......................... 31 1o Raw Materials...... o..o.................... o.o.,... 31 2. Preparation of Charges,.......5..... O............... 31 3. Furnace Assembly and Operation.............o....... 5533 G. Chemical Analysis................................. 35 1. Basic Procedure for Unalloyed Samples........... 37 2. Procedures for Alloyed Samples......................o 38 DISCUSSION OF RESULTS............... o *....<oQ****5. 39 A. Review of the Data.........9.................. 59 1. Solubility of Calcium in Pure Liquid Iron at 2925 ~F.. 39 2. Effects of Alloying Elements Upon the Solubility of Calcium in Liquid Iron at 2925 ~F,,.oo................ 44 B. Engineering Significance of the Solubility of Calcium in Liquid Iron.......................................... 48 1. Calcium as a Refining Agent in Open Systems.......... 48 2. Calcium as a Refining Agent in Pressurized Systems.... 50 3. Importance of the Immiscibility Between Calcium and Iron................ o.............. o.........o.. o. 52 4. Significance of the Effects of Alloying Elements Upon the Solubility of Calcium in Liquid Iron......... 52 C. Theoretical Considerations Regarding the Effects of Third Elements.........,............. o.......... 0 o.. 55 1. The Alcock and Richardson Equation...................o 55 2. Wagner's Electron-to-Atom-Ratio Concept.............. 65 3. Summary of the Present Data in Relation to Theory..... 70 CONCLUSIONS...***...e,.,,.,,,....*.O 72 CONCLUSIO NS..................................................... 72 APPENDIX 7I...................... 75 APPENDIX II.......................................... 76 BIBLIOGRAPHY............................ o ooo....o 86 V

LIST OF TABLES Table Page I Chemical Analysis of Raw Materials................ 32 II Chemical Analysis of Experimental Heats............ 40 X III Determination of Interaction Parameter E Ca 59 IV Relation of Atomistic Preference of Element X to Observed Interaction Parameter ex 64 Ca............ V Alternate Criteria for Applying Wagner's Electron Theory...................................... 68 vi

LIST OF FIGURES Figure Page 1 Experimental Pressure Chamber...............................14 2 Furnace Arrangement..................................... 15 5 Crucible in Position in Lower Section of Vapor Lock......... 17 4 Vapor Lock and Shutter Arrangement......................... 17 5 Susceptor Assembly.......................................... 17 6 Shutter Linkage.............................. 18 7 Coil Assembly Components and Shutter Linkage................ 18 8 Sampling Tube and Suction Line............................. 19 9 Sampling Tube Positioning Device............................ 19 10 Constitution of a Compacted Crucible............ 24 11 Crucible-Making Die and Accessories........................ 24 12 Induction Sintering Furnace................................. 25 13 Comparison of Crucible Before and After Sintering.......... 25 14 Location of Thermocouple.......................... 28 15 Calibration Curve for Base Thermocouple..................... 30 16 Components of a Typical Charge.............................. 34 17 Time to Reach Equilibrium.................................. 34 18 Two-Layer Melt.............................................. 36 19 Effect of Aluminum Upon the Solubility of Calcium in Liquid Iron at 2925~F.................................... 41 20 Effect of Carbon Upon the Solubility of Calcium in Liquid Iron at 2925OF...................................... 41 21 Effect of Gold Upon the Solubility of Calcium in Liquid Iron at 2925 OF.................................. 42 vii

Figure Page 22 Effect of Nickel Upon the Solubility of Calcium in Liquid Iron at 29250F.......................... 42 23 Effect of Silicon Upon the Solubility of Calcium in Liquid Iron at 2925 ~.................... o................ 43 24 Vapor Pressure of Calcium at Steelmaking Temperatures....... 49 25 Effect of Alloying Element Concentration in the IronRich Liquid Upon the Activity Coefficient of Calcium in the Iron-Rich Liquid at 2925F.............................. 60 viii

ABSTRACT The purpose of this investigation was to determine the solubility of calcium in liquid iron and the effects of various third elements upon the solubility. These relationships have not been investigated previously because of the considerable experimental problems in attaining equilibrium between liquid calcium and liquid iron at temperatures in-the range of 29250F. By using a chamber pressurized with argon and containing an induction furnace, the desired melts were made in specially developed titanium nitride crucibles and samples were taken by remote controls. The data indicate that liquid calcium and liquid iron are immiscible and, at 2925~F., the solubility of calcium in iron is 0.032 wto. Of the third elements tested, carbon is most effective in increasing the solubility of calcium, followed in order by silicon, aluminum and nickel. Gold decreases the solubility. The calciumiron-carbon-system contains an interesting inflection point; when the carbon content reaches a value of about 0.87wt%, the CaC2 phase replaces the metallic calcium and the calcium concentration in the iron decreases. The maximum solubility 0.36 wt% calcium, was obtained in an alloy containing 10.5 wt% silicon in the iron. The data have interesting applications both in process metallurgy and in providing material for testing the theories of metallic solutions of Alcock and Richardsonandof C. Wagner. In the engineering field a variety of calcium-bearing ferroalloys have been used in steelmaking without any basic knowledge of the solubility of calcium. The data of this report should lead to more effective calcium alloy design and use in deoxidizing and other refining reactions. furthermore the development of t~he processing procedure ix

for the new titanium nitride crucibles and the design of the remotely controlled melt sampler should permit the exploration of many interesting but reactive alloy systems which could not be investigated to the present time. The application of the data to theories of third-element effects in systems of restricted solubility is rewarding. Reasonably good agreement is found with current theories, and criteria are suggested which permit these theories to be applied more readily to systems of this type. x

INTRODUCTION The nature and behavior of liquid metals have eluded the complete understanding of metallurgists and other investigators even to the present day. This is largely due to the difficulty of examining structures in liquid metals, and to the relatively high melting points of most metals of engineering interest. One very important aspect of liquid-metal behavior is the relative ease with which an element can dissolve in a liquid metal. Unlike the case of solid solubility, which can be estimated rather well by use of the HumeRothery rules, relatively little is known about the factors influencing solubility in liquid metals.(1) Such information is highly desired, however, in connection with such engineering problems as the control of impurity levels in liquid metals, the use of magnesium in nodular iron production, the use of liquid metal alloys as nuclear fuels, and as coolants in nuclear power plants and other heat transfer applications. From a theoretical point of view such information should be helpful in advancing the understanding of the structure of liquid metals. One important example of restricted solubility concerns the extent to which calcium will dissolve in liquid iron. This is a matter of considerable importance in view of the strong potential of calcium for removing certain impurities in steel, as well as the ability of calcium to desulfurize, (2) inoculate, and produce spheroidal graphite in cast irons.( Since most ferrous alloys contain other elements, it is important to investigate the interactions of third elements with the calcium-iron system~ -1

-2A. Purpose of the Investigatiot In view of these considerations, the purpose of this study was to investigate the effects of third elements upon the solubility of calcium in liquid iron. For reasons of theoretical and engineering importance, the elements investigated were aluminum, carbon, gold, nickel, and silicon.

REVIEW OF THE LITERATURE For convenience, the literature pertinent to this investigation may be considered under two categories: A. Experimental studies. B. Theoretical considerations. A. Experimental Studies Several attempts to determine the solubility of calcium in liquid iron were conducted during the early part of the present century. None of these investigations, however, established the solubility of calcium in liquid iron with any degree of certainty. Quasebart attempted in 1906 to dissolve calcium in open steel melts by several techniques, with negative results.(3) In the same year, Watts reacted excess calcium with powdered iron oxide in an exothermic reaction within a simple pressure vessel.(4) A piece of the melt was found to contain 0.3% calcium, which Watts considered to be mechanically entrapped, in view of the vigorous nature of the reaction. In 1908 Hirsch and Aston reported negative results for their attempts to alloy iron with calcium in three types of open-melt experiments.(5) In an exothermic bomb-type reaction similar to that of Watts, they obtained pieces of iron analyzing from 0.29% calcium to 1.37% calcium, and considered that at least some of the calcium was alloyed with the iron. Later writers considered that any calcium reported in the investigations just described was mechanically entrapped in the liquid, with Wever considering calcium as well as the other alkaline earth and the alkali metals categorically to be insoluble in liquid and solid iron. 7) -3

-4All these investigations were unsatisfactory in that they failed to insure the pressure of a calcium liquid layer in contact with the liquid iron. In the open-melt studies the calcium would have boiled off immediately, and in the bomb-type studies, the calcium presumably would have distilled off to t+e much colder walls of the container. Furthermore, it was not possible to draw samples directly from the liquid iron in the bomb-type experiments. Any calcium that might have dissolved in the liquid iron could have been rejected by the solid iron during solidification. These early investigations are of considerable value, however, because they indicate at least that liquid calcium and liquid iron are immiscible. This follows from a consideration of the vapor pressure of calcium, which is Q \ approximately 30 psi at the melting point of iron, 2802~F(8) If iron and calcium formed ideal solutions, then the vapor pressure of calcium would not reach the boiling point, 14.7 psi, until the calcium level reached approximately 50 at%. In this case it might be expected that up to 50 at% calcium could be dissolved in liquid iron just above its melting point in an open melt. Even if strong positive deviations from Raoult's law existed in this system, a calcium solubility of at least 5 to 10 at% might still be expected. Since no calcium was detected in the open melt studies, a miscibility gap is indicated for the iron-calcium system. This observation is consistent with the criterion for immiscibility first formulated by Hildebrand and Scott, and also with the adaptation of this criterion for metals by Mott, from which an immiscibility between iron and calcium is clearly predicted.(9 ) Little work has been done to measure the solubility of calcium in liquid iron since the time of the investigations described above, presumably

-5because of the severe experimental difficulties involved. In recent years Philbrook and co-workers used radioactive calcium in a study intended to take advantage of the minute concentrations of this species which can be detected. In this study, the calcium was present as CaO in a slag layer, and it was hoped that a sufficient amount of the radioactive calcium would dissolve to permit detection. No calcium was detected in the iron melt, however, even after equilibration at the highest temperature studied, 1800~C (3272~F). Thermodynamic calculations would indicate that if metallic calcium is used instead of CaO at this temperature, then the order-of-magnitude upper limit for calcium solubility in pure iron would be 0.1 wto. Recently, Trojan and Flinn have reported magnesium solubilities as high as 3 wto in liquid iron containing carbon and silicon in the presence of a magnesium-rich layer.(12) Traditionally magnesium, like calcium, had been considered insoluble in liquid iron. Considering the Goldschmidt atomic radii, Fe(BCC) 1.28 Angstrom units Mg 1.60 Angstrom units Ca(FCC) 197 Angstrom units calcium would be expected to be considerably less soluble than magnesium in liquid iron.() Nevertheless, significant levels of calcium solubility in liquid iron and iron-base alloys might reasonably be expected on the basis of the rather high solubility of magnesium. B. Theoretical Considerations Two theoretical models are of particular interest in relation to the effects of third elements upon the activity or upon the solubility of

-6one element in another. The models, which are due to Alcock and Richardson and to Wagner, are considered in this section. 1. Chemical Model. Alcock and Richardson have considered the energies involved in U',reaking old bonds and creating new ones when solute 2 is added to a dilute solution of solute 3 in solvent 1.(14) Assuming no clustering of atoms, and that all atoms have the same coordination number, they obtain the relationship; (ln Z2(1,)) = l = in 2(3) - in Y2(1) - In Y3(1) (1) p N —> 0 2 where 7A(B) = activity coefficient of A in B, NA = mole fraction of A E = interaction parameter. This equation relates the interaction parameter to activity coefficients in the three binary alloys, and would permit the prediction of the effect which a third element would have upon the activity (and upon the solubility when the activity of solute 2 is relatively constant) of solute 2 for situations where the appropriate activity data are available. 2. Electron Model Wagner has proposed a theory of considerable interest concerning dilute ternary alloys.(15) This theory states that the activity of a solute metal 2 will be increased by a third component if metals 2 and 3 change the electron-to-atom ratio in the same direction. Conversely, the activity of a solute metal 2 will be decreased by a third component if metals 2 and 3 change the electron-to-atom ratios in opposite directions.

-7This theory assumes that metal 2 is at least partially dissociated into positive ions and electrons, and that interactions between positive ions are unimportant. The physical basis for this theory is that the chemical potential of electrons in the solvent metal is raised by the addition of Cslute atoms having a greater number of free electrons than the solvent atoms, and is lowered by the addition of solute atoms having a smaller number of free electrons than that possessed by the solvent. Here the chemical potential is considered to increase as the Fermi energy of the electrons increases. The solubility of solute atom 2 in a solvent 1 should therefore be enhanced under conditions of relatively constant activity of 2 by the addition of solute atom 3 when 2 and 3 tend to change the electron-to-atom ratio of the alloy in opposite directions because this combination changes the chemical potential of the electrons less than if 2 were added alone. Furthermore, Wagner derives the following equation for the electronto-atom-ratio concept: 5, 2 [2~ l1/2 E = 2 = + [2 3],2 (2) 2 5 - 2 3 N2-D, N-O where 3ln Y2 23 etc =, etc. 2 and 3 = solute atoms r = activity coefficient of a species in the solvent N = mole fraction of a species in the solvent 2 5 In this equation the terms 2 and E represent self-interaction parameters for the binaries 1-2 and 1-3, respectively. These parameters are a measure

-8of the change in activity coefficient of the solute with changing solute concentration, and are therefore related to deviations from ideality in the binary system. Equation (2) permits the interactions between two solute atoms to be predicted from a knowledge of activity in the respective binaries of the solutes with the solvent. This equation agrees well with experimental data reported by Wagner for three ternary amalgams. However, use of the equation suffers from the limitations that very few values of self-interaction parameters are available, that self-interaction parameters having opposite signs result in imaginary values of interaction parameters, and that the sign of the interaction parameter must be determined by qualitative considerations. Nevertheless, Equation (2) is of sufficient interest to warrant consideration in this investigation.

EXPERIMENTAL METHOD In this investigation calcium solubility is measured in liquidiron melts which contain nominally 0, 2, 5, 10 and 20 at% alloying element, (In the carbon series, studies are at 2, 5, and 10 at% carbon.) Experimental melts are made under argon pressure, in order to prevent boiling away of the calcium, a technique used by Trojan and Flinn to study equilib(12) ria between magnesium-base liquids and iron-base liquids.( In this section equilibria between phases in such a pressurized system are analyzed by use of the Phase Rule of J. W. Gibbs. In addition, the more important aspects of the experimental technique are described. A. Phase-Rule Analysis The phase rule is a powerful tool in controlling and understanding equilibria between two or more phases. It may be written V = C + 2 - P, (3) where V = variance of the system, C = the number of components, P = the number of phases present.(16) Here the "2" represents the phase-rule variables temperature and total pressure of the system. The significance of the variance, which is sometimes referred to as the number of "degrees of freedom' is that it specifies how many of the phase-rule variables of temperature, pressure and composition must have values assigned to them in order that the remaining tbase-rule variables are uniquely specified. The phase-rule canbe applied first to m9m

-10the liquid iron-liquid calcium system under argon pressure, and then to the liquid iron-liquid calcium system to which an alloying element has been added, again under pressure of argon. 1. Iron-Calcium System Under Argon Pressure When liquid iron is equilibrated with liquid calcium in the presence of argon, the terms in the phase-rule equation (3) become: C = 3(iron, calcium, argon) P = 3(iron-rich liquid, calcium-rich liquid, gas) Hence the variance becomes V = 2. However, both temperature and pressure are held constant (29250F, 200 psig, respectively) in this investigation; therefore V = 0. This means that the equilibrium compositions of all three phases are fixed for the given temperature and pressure in question and the solubility of calcium in iron should be a unique value. 2. Iron-Calcium-Alloying Element X System Under Argon Pressure When an alloying element X is added to the above system, the number of components increases to four, and the variance at a fixed temperature and pressure in the presence of the same three phases therefore becomes: V = 1. From this it is seen that by fixing one of the composition variables in the system at equilibrium, the values of all the remaining composition variables can be uniquely specified. Thus the remaining composition variables can be plotted as unique functions of one of the composition variables. Since the

-11number of dimensions required to graphically portray an equilibrium is V + 1, a two-dimensional plot results. (16) If Ne(L) the concentration of the alloying element X in the iron-rich liquid, is plotted as the abscissa, then all the remaining compositional variables can be plotted as ordinates against (L resulting in a set of unique curves. These remaining variables are NCa(L) G = concentrations of alloying element X X x in the calcium-rich liquid and gaseous phase, respectively. NFe(L) NCa(L) NG = concentrations of calcium in the ironC a C a Ca rich liquid, calcium-rich liquid and gaseous phase, respectively. Fe(L) NCa(L) NG = concentrations of iron in the iron-rich Fe Fe Fe F e Fe Fe liquid, the calcium-rich liquid and gaseous phase, respectively. Ai^(L, N A(,= N = concentrations of argon in the ironAlAe(L)A ~ rich liquid, calcium-rich liquid and gaseous phase, respectively. Of the compositional variables presented here, the one of particular interest in this investigation is Ca L) the concentration of the calG G cium in the iron-rich liquid. The values of NX and NFe are quite small because of the low vapor pressures of the present elements and of iron, respectively. 3. Effect of Argon Pressure Upon Equilibrium The inert gas argon is provided at a pressure above the vapor pressure of the calcium, to prevent boiling away of the calcium. The manner in which the argon enters into equilibrium with the system is of considerable interest. Argon does not have measurable solubility in liquid metals.( The terms e(L) and Na(L) in the previous section become therefore essentially A A zero. The term N is not zero, and increases as the externally applied argon pressure increases, since it represents the concentration of the argon in the gaseous phase.

-12Now the concentration of argon remains unchanged (essentially zero) in the two liquid layers as the argon pressure in the system is increased. Thus, the equilibrium between the liquid layers is not affected by a change in the amount of dissolved argon. Furthermore, the effect of a change in pressure itself upon the equilibrium between two liquid phases at constant temperature is expected to be quite small, by virtue of the thermodynamic relationship T = AV, (4) P 2 T where AF = the change in free energy associated with the equilibration of the two liquid layers AV = the volume change associated with the equilibration of the two liquid layers P = pressure T = temperature.(18) Since the volume change AV involved in having calcium and iron dissolve slightly in each other is expected to be quite small, the pressure dependence of the free energy change AF of the process should also be quite small, indicating a negligible effect of pressure upon solubility. The above considerations have been borne out experimentally by Trojan and Flinn, who found the solubility of liquid magnesium in liquid ironcarbon alloys was unaffected when the argon pressure of the system was varied by a factor of four.(12) Therefore, as long as the externally applied argon pressure is sufficient to maintain a calcium-rich liquid layer, the compositions of the two liquid layers are relatively independent of the pressure of the system. Specifying the pressure serves only to p.rmit the composition of the gaseous phase to be specified. In situations where

-13this is not of interest, pressure need not be considered as a phase rule variable. The present investigation is just such a case. Pressure is held constant throughout the study for experimental purposes, but the same results are to be expected at any pressure sufficient to maintain a calcium-rich layer. B. Pressure Chamber The experimental pressure chamber used by Trojan and Flinn is employed in this investigation and is shown in Figure 1. This consists of a heavywalled pressure vessel made of austenitic stainless steel, an associated system for controlling pressure and other control devices. The interior of the pressure shell and the port for the power leads are water-cooled to avoid heating by the 3000 cycle induction heating power supply. It was necessary to redesign the interior furnace components to carry out this investigation. The new features of design are described in subsequent sections. C. Arrangement for Sampling in Place In developing the experimental phase of this investigation, it seemed highly desirable to develop a method which would permit sampling of the iron-base melt in situ. The essential components of the experimental apparatus for accomplishing this are shown schematically in Figure 2, and are considered in detail in the following sections. 1. Coil Assembly The crucible is contained within an inner graphite vapor lock, the purpose of which is to retard the loss of calcium by gaseous diffusion. A hole is provided in the upper half of this vapor lock. At the time of sampling,

-14-...... Figure 1. Experimental Pressure Chamber. The pressure chamber is shown, along with equipment for recording temperature, controlling power, and controlling pressure.

-15SHUTTER LINKAGE < \ -I -COPPER PLUG \ \ STEEL FIBER PLUG ~/ ^^^ T ^-~SAMPLE TUBE ALIGNING DEVICE TYGON SAMPLING LINE-4 //ALUMINUM i I ORIFICE SAMPLE TUBE nr |-FUSED SILICA SAMPLING TUBE -SUPPORT VALVES GRAPHITE IRON FOIL CAP SPINDLE NDLE~~.~.~.~. 4 — CO PPER COIL SAUEREIS \\\ \ \E CEMENT "bFN-B — I' INSULATING ~ t t -GORAPHITE SUSCEPTOR ~' R -- — _-GRAPHITE SHUTTER - -GRAPHITE VAPOR LOCK TWO-LAYER MELT - TITANIUM NITRIDE CRUCIBLE -Pt/Pt- I0%/o Rh T.C. - GRAPHITE RADIATION SHIELD -- ALFRAX B-I SLEEVE -I;~ -;'m^ ^X>Xa \-ALFRAX B-I BASE r ~ ",, 1,.\\\\" — \ \ FUSED SILICA SHELL e^ < _^ i^3^, s _ -THERMOCOUPLE SUPPORT TRANSITE PEDESTAL TO PRESSURE2 CONTROL SYSTEM , S.A^/^ ^-NYLON RACK a MICARTA PINION PRESSURE CHAMBER SHELL Figure 2. Furnace Arrangement. The interrelation of pressure chamber coil assembly, shutter turner and sampling device are shown in this schematic drawing.

-16a graphite shutter is remotely manipulated to uncover this hole, permitting the sampling tube to enter the melto The crucible-vapor lock assembly is contained within a cylindrical graphite susceptor, which has holes in the top to permit entry of the sampling tube and to accommodate the spindle which tu-ns the graphite shutter. The role of the susceptor is to heat the entire inner chamber and prevent calcium condensationo The inner parts of the coil assembly are shown progressively arranged in Figures 3 through 5o 2. Shutter Linkage A mechanical linkage is used to rotate the shutter at the time of sampling, by means of a handle mounted in the cover of the pressure chamber. This linkage, shown in Figure 6, features two universal joints and a sliding joint. The universal joints provide the flexibility necessary to connect the linkage at the sliding joint when the furnace cover is lowered into position. Except for the graphite spindle which projects into the hot zone of the coil assembly, the shutter linkage is made from austenitic stainless steel and aluminum. A composite view of the shutter linkage and all parts of the coil assembly is shown in Figure 7. 35 Sampling Device The sampling tube arrangement is illustrated in Figure 8. Samples of the iron-rich liquid are drawn into a fused silica tube. This tube is prepared by a meticulous cleaning, then coating the lower end with a colloidal dispersion of graphite in alcohol, wiring a piece of thin iron foil over the tip, and inserting a steel fiber plug of controlled length and weight into the upper end of the tube. The entire tube is then dried thoroughly over a Bunsen burner just before use. The iron foil serves to keep calcium vapor

-17-.. ~ ~ ~ ~'.:::....... -:::.:_:-:..:_,._ 1- I,.,~~~~~~~~~~~~~~~::: i::: M - 1.. i:~~~~~~~~~~~::.~~~~!::;~~~~~~~~..~~~~~~ -.,:~~~~~~.:.,...,~~~i~i~:::,:i.:::ii::"-:':'::ii..i~::::.:i -:': j~:.:: -,''....,-...... i~i lli:iz.i.ijl:;::::::~::::/:j:\:;j: ~..11'.. -,~~~~~~~~~~~~~~~~~~~~~~, ~ ~ ~ ~ ~~.. 1.~~~~~~~~~~~...~~~~~~~::::~:::-::''''",'''',,. I ~::~F ig u r e 3. iii~ii~iiiliiIl:-I:,:_. I -.1-1-Ii:-:::..i —-:'-..'M. —- I., I I,~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~:::..'':-:- 1- _. _:.,.::,:,.::.:._.,:~~~~~~~~~~~~~~~~~~~~~~~- 1 ii;i. i;.:::: 1 - 1- 1.'.:.-:;::. ":.q_~~~~~~~~~~~~~~~~~~~~~~~~~~:::;::.':,..:........ -.... _._ ii::;:::::-::: -, —.... - -.e: -.... - I~~ ~~~ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ I..,...~~ii ~ li~ii;.-:::: 11. ~~~~~~~~~~~~~~:.::_:,:.:~~~~~~~~~~~~~~~~~~~~~~~~~~~~~:~~~~~:..%..:. ~~~i: ~~ilIi~i1:i~ii:i:I1I:..:: 11.,.- ~~~~~~:::,:::~~~~~~::,: ". ~ ~ ~ ~ ~ -,.q.,.....,:~~~~~~~~~~. - I.: I I_:..-I... I - I.:.::iii:::::: thermocouple, pedestal, r e fr a c::;::-::i:;llii':::::~~~, ~ c~,.......,::. _:_..- 1..1.-~'.1,.-"I1,-I,::~:::..:.., I.... ~~~~~~. ~~~..., -. I::.:,:::..,:::::-.: -::.:..~~~~~~~~~~~~~~~~~~~~~~~~~~~~,.I., Iiii:::: -.. -:, III: i: ~~:-~~~::::;~~~~~::;~ ~ ~~. I.I.... ~~~~~~~~...... 11..:.,.::ii::i:::::::,:: - — ~~:'~!~~j ~.~.j:~~..... ~ ~::::::~~~:~].~;~.'.~~~l...~~~:~: -.::.:: -::::'::,~::::::...: 1.,:::::: I..~~~~~ -.~~~~~:~~~~~i~~~..:~~~~~:~~~~~~:~~~:~~~~~' ~ ~ ~ ~ ~ ~;..,.: "~~~~~~~~: X:':''::~~~i,i:.:: i.:. I.:: I...::::::.. 1i: I I:':::.,.:.."::",.,.,.......~~~~~~~~~~~~~~~~~~~~~~~~~...::....::::,.....: ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~.."........-I-.1 —.1 1. -.:-ii::::~:ilii:-i 1 Ii I:::..11......, ~~~~~~~~~~ - I.., X.:::~~~~~~~~~~...'..:- ~~~~~~~~~~:::::...!,.:I:.::_:-, -.1.1. -, 11, I I~':::::'':ill i::.::.;_.-I I.,...~~~~~~~~~ ~ ~ ~ ~~~~~~~~~~~~~~~~~~~~~~::::.:.. i'iiili.. I.. i1.. -:~ ~- 1xD~li L.::::...:':::,: ~:,. -.1.:~~~~~i,.!~~~::~~~;:~~.~~:~~:i~~~~~,:.~~~:::.:~~~~.~~.~~~~: i:::i, 1 i:. -...:- -_.:I.::..: - 1:-l.:::...... -i:::i.....::..:';::i;;i;];'l:i'~~; ~ ~....i':.,.1:-:: -,.:: -1_1 _: -i..'l.I-:::::::::::I ~ii::i-:~~~~~~~~~~~~~~~~~~1..I.I..........1..:::,:,:, " I,..,. i i.-::;::.~ ~.~:"- ~ ~.::.: i.,::X:::.::... - i i.:.,, _....:-,..... I 1...I.:.. -,. -.. -. 2 1' 1, Vapor Lock and Shutter Arrangement,.::1Susce. pto r I A:::.11... I..,. 1.'...I,..,... e c. The upper half of the vapor lock tion of the graphite susceptor and~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~I.,1...1-..1I "Iand t e sh tter ave ee.... o,....l-If r t,: g h:uttI the arrangement of Figure 3. have beeIn...-FI.-. a-..dded,X t _o Vt,..-he......I II,..I..,.m InI:. I.. F...,2r.e.I4.

Figure 6. Shutter Linkage. The graphite shutter at the lower left is operated. remotely by turning the handle at the u-pper end. of the flexible linkage..7.7.477.. 7.7.7.7. opoetsad hterLnkg. hw here are the induction coil pedestal, thermocouple,~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~.7.7.7.7., rcil it aiain-hed ae.7.7.7.7.7.7.7.7.7.7.7.7. p ece an sh tte 7.7.. 7.7....7

-19 Figure 8. Samnpling Tube and Suction Line. The assembly consists of a graphite-coated fused silica tube with iron foil over the lower end, a steel fiber plug, copper plug, flexible suction line, wall connection and orifice. Figure 9. Sampling Tube Positioning Device. The sampling tube is carefully located in this device, which lowers the sampling tube into the melt at the time of sampling, and then withdraws it immediately aft:r obtaining the sample.

-20out of the tube prior to sampling, and to prevent the calcium-rich liquid from entering the tube as it passes downward through this layer. The graphite dispersion is necessary to prevent the calcium-rich liquid from attacking the fused silica tube. Metal samples are chilled when they encounter the steel fiber plug, and this prevents them from vaporizing the rubber components of the suction line. The metal samples are forced into the sampling tube by developing a pressure differential between the interior of the furnace and the sampling line. This is accomplished by bleeding argon from the samplig line to the atmosphere at the time of sampling. The pressure differential causes metal to rise rapidly in the sampling tube. The sampling rate is controlled by the use of an orifice and two needle valves, all in series, which retard the flow of the exit gas. The device for lowering the sampling tube into the melt at the time of sampling is shown in Figure 9. This unit, which permits five degrees of freedom in aligning the sampling tube, is moved by a rack and pinion drive. The sampling tube is kept in position above the upper insulating piece until the time of sampling, when the tube is lowered into the ironrich layer. Immediately after the sample is drawn, the tube is returned to its original position. D. Preparation of Crucibles In most experimental studies, crucibles which are commercially available can be used satisfactorily. Because of the extreme reactivity of calcium, however, such commercial crucibles were found unsatisfactory in this case. The performance of commercial crucible materials is described

-21in this section, as well as the method for preparing titanium nitride crucibles. The method involves the steps of pressing, sintering and nitriding. 1. Performance of Commercial Crucibles An extensive testing program was conducted to assess the suitability of commercially available crucibles for containing liquid calcium at temperatures above the melting point of iron. As a screening test, samples of the different crucible materials were tested in liquid lithium at 2500~F. Since lithium is somewhat more active chemically than calcium, such tests were used as an index of the ability of a refractory to withstand attack by liquid calcium at the higher temperature of this investigation. Tests were conducted by placing the refractory specimen and lithium within a hollow bomb-type iron crucible and inserting an iron plug, (This was done in a glove box.) After sealing the plug by arc-welding, the iron container was held for approximately 10 minutes at 2500~F in an induction furnace, The container was then sectioned and the refractory specimen was examined, The results of these screening tests are presented in Appendix I. Briefly, all the commercial crucible materials tested were found unsatisfactory. 2. Selection of a Crucible Material Of the possible classes of materials that might withstand attack by liquid calcium, the nitrides are of considerable interest because calcium is a relatively weak nitride-former. Titanium nitride is one of the most stable of the nitrides, and should therefore suffer little attack by metallic calcium. No experimental data were found concerning the

-22resistance of titanium nitride bodies to attack by molten calciumo However, pure titanium nitride crucibles were found to be unattacked by molten cerium, a metal which, like calcium, is quite active chemically (19) 3. The Effect of Titanium Hydride In order to obtain impervious pure titanium nitride bodies by sintering, temperatures relatively close to the melting point of 5300~F are required since sintering must occur without the benefit of any liquid phase. At these temperatures, vaporization of the titanium nitride becomes quite appreciable. Furthermore, the use of very high sintering temperatures intensifies the technological problems associated with sintering. The use of relatively large amounts of titanium hydride as an aid to sintering was suggested, and found to be quite effective.(20) Metallic titanium is formed at low temperatures from the decomposition of the titanium hydride, and greatly accelerates the sintering process at higher temperatures. A two stage sintering operation was developed by which the sintering process first is initiated at a temperature just below the melting point of the metallic titanium. During this time, the first stage of sintering, the metallic titanium is completely converted to TiNx, where x is considerably less than unity. In the second stage of sintering, crucibles are heated to a temperature above the melting point of titanium to accelerate the densification process. The optimum amount of titanium hydride was found to be 20 percent by weight. 4. Compacting of Crucibles The powder from which crucibles are compacted is carefully prepared from the raw materials, the proportions of which are shown in Figure

-2310. Crucibles are pressed in a double-acting die which permits the base and walls of the crucible to be compacted separately, Figure 11. By careful control of the pressing operation, compacted crucibles having high sidewalls can be obtained free of cracks. 5. Sintering Compacted crucibles are dewaxed at a final temperature of 475~F and are then placed in the induction sintering furnace shown in Figure 12, which was constructed for this purpose. A graphite susceptor is used to develop heat in the induction field, and contains the crucibles to be sintered. The susceptor is insulated by minus 100 - plus 150 mesh graphite. The water cooled cover contains a sight port for temperature measurement by an optical pyrometer. After the system has been evacuated, the crucibles are held for 30 minutes at approximately 1200~F while the hydrogen from the decomposition of the titanium hydride is drawn off. The crucibles are then heated to 2950~F and held there for one hour. (This is termed first stage of sintering.) At this point argon is bled into the sintering furnace to a pressure of 5 psi (absolute) in order to reduce subsequent vaporization of the titanium nitride, and the crucibles are then heated to 3350~F. (This is termed second stage sintering.) After one hour at this temperature the furnace atmosphere is evacuated, power is turned off, and the crucibles are cooled under vacuum. The effect of sintering is illustrated in Figure 135 Linear shrinkage and density measurements are made on all crucibles in order to keep the process in control, Also, crucibles are filled with distilled water and the drop in water level in one hour gives an index of

24 zJ~g'ur i0 Constiutioin of a Coipa e'te Cruci ble, he raw materials containecd in a comp~acted. 42-gram cruc ible aoe sho'wn here,.on.sisti. o. 8 gra...s o. titan.iu i.. hydrie, 2.. titan-i.t i nitride, Igra m of xylene andi 1.grsm of parafin, A compacted crucible is shown at the right,.............................a. e sho........ in..the c..nt. with tool fo...sembli t.he die at the le>f and ~ixtureo or disa embllng the die o~~~~~~~~~~~~~~~j:S'~:: ~j-' SSS S. SS'S.......SSSB.SSB:................................................................ lEW lE lE lE 01 1000010111101 1 1 E i' lt1 ===~~~~j~~:~"::::~~:~~::~~:::2' R B 18gg:gSS:~gS'''' S.~ _.,S,.,g:::::ji:::j~:~; sjj:~~ ~~::~~:~~:~~~~~: a-E; tnce righe

-25Figure 12. Induction Sintering Furnace. Fused-silica shell, induction coil, and water-cooled cover, with appropriate control equipment. Figure 13. Comparison of Crucible Before and After binzerlrng. Extensive strinkage is evident, especially in the center of the sidewalk.

-26the ability of the crucible to contain liquid metal. The best crucibles show practically no absorption of the water. This latter test was especially valuable in perfecting the technique for making crucibles. 6. Nitriding The addition of titanium (from titanium hydride) referred to above shifts the composition of the compound TiNx rather far in the direction of titanium. Metallographic studies show that there is no free titanium present after sintering, all having been converted to titanium nitride of low nitrogen content. Crucibles cannot be used in this nitrogen-impoverished condition because they fail by thermal shock. Presumably a phase transformation occurs at low nitrogen levels, leading to failure. Thermal-shock failures are completely eliminated by restoring the nitrogen level to a high value in a nitriding process. Crucibles are held in a purified nitrogen atmosphere at 2300~F for 48 hours. The nitrogen concentration approaches saturation in this treatment, and is accompanied by a color change of the titanium nitride from silver to gold. 7. Performance of Titanium Nitride Crucibles Crucibles prepared by the foregoing procedure give very satisfactory service. The liquid calcium does not damage the crucible, the only effect of a chemical nature being that a small amount of the crucible dissolves in the: melt. This results in a titanium concentration in the ironrich layer of less than 0.3 wt% for most of the heats. The study of a special high-titanium heat shows that these low titanium levels have a negligible effect on calcium solubility.

-27A small amount of metal normally is absorbed by the crucible. For most heats of this investigation the loss is one gram or less. The crucibles withstand the rapid heating and cooling of this investigation without any evidence of damage due to thermal shock, E. Temperature Control In any investigation of solubility accurate temperature measurements are desirbale. For this reason, a good deal of attention was given to reducing the absolute and relative errors in temperature measurements to a minimum. Since no satisfactory protective thermocouple tubes are available for immersion, the measurements are made using the crucible itself for protection as follows: 1. Method of Measuring the Temperature of the Melt Temperature measurements are made with a platinum/platinum - 10% rhodium thermocouple located in the base of the crucible. This location is used in preference to the more conventional method of placing the thermocouple inside a protection tube immersed in the melt, for two reasons, First of all, no commercial protection tube would withstand attack by the calcium, and secondly, there is insufficient space in the small crucible for a thermocouple protection tube. The location of the thermocouple is evident in the schematic drawing of Figure 2 and the photograph of Figure 14, showing the thermocouple hole in the base of a crucible and the insulated thermocouple. Crucibles are designed with a thick base to accorixnmodate the thermocouple, Prior to dewaxing, an oval-shaped hole is carefully drilled to within.1/16 in. of the melt cavity of the compacted crucible. This hole is carefully shaped

-28Figure 14. Location of Thermocouple. Bottom view of titanium nitride crucible, showing 5/16-in. deep thermocouple hole and insulated platinum/platinum - lO rhodium thermocouple which is inserted into this hole.

-29to a standard template, which corresponds to the oval shape of the Triangle RR recrystallized alumina thermocouple insulator. After the crucible is sintered, the base is ground down until the thermocouple hole is exactly 5/16 in. deep. The thermocouple insulator has a 3/64 in. notch cut in the end, permitting the thermocouple bead to be recessed so that it does not touch the crucible. The difference between the temperature at the control thermocouple and the temperature within the melt is shown for two separate calibration heats in Figure 15. These calibration heats involved a melt of Ferrovac'E' iron equal in volume to the total volume of an iron-calcium melt. A thermocouple made from the same wire as the base thermocouple was located inside a protection tube immersed in the melto Temperatures of the two thermocouples were recorded simultaneously after the temperature of the melt had stabilized at each of the three temperatures investigated. The calibration curves agree quite well, and indicate that the base thermocouple is 60~F below the temperature of the melt at 2925~F and 200 psig, the temperature and pressure respectively of this investigation, 2. Calibrations of Thermocouple Wire and Control Equipment The thermocouple wire used in this investigation was calibrated against a platinum/platinum - 10% rhodium secondary standard, which had been previously calibrated by the National Bureau of Standards. A thermocouple that had been used in an experimental heat was calibrated against an unused thermocouple made from the same wire to check for possible drift due to contamination during the course of a run, but no drift was detected. As a precaution against any slight contamination

-3080 2925 OF 60 TT =60 O/ sj 20 0 HEAT 93 - 0 HEAT 94 -T: MELT TEMP. MINUS THE TEMP | -- OF THE BASE THERMOCOUPLE 0 I _ _ _ _ _ 2800 2850 2900 2950 3000 MELT TEMPERATURE,F Figure 15. Calibration Curve for Base Thermocouple.

-31effects and possible mechanical damage, a one inch length at the bead end was cut off and discarded after each heat. Immediately before each experimental heat the Leeds and Northrup Speedomax Temperature recorder was calibrated against a Leeds and Northrup portable precision potentiometer. This potentiometer had in turn been calibrated against an Eppley standard cell. 3. Accuracy of Temperature Control The experimental melts were maintained at 2925~F during the equilibration period by adjusting the power input manually. When this small source of error is considered with other possible sources of error, the temperatures of all heats in this investigation are estimated to lie within 15~F of 2925~F. For the majority of heats, the maximum error is estimated to lie between 5~ and 10~F. F. Conduct of Experimental Heat 1. Raw Materials The calcium, iron, and alloying elements employed in making the investigation are specially selected materials of high purity, Table I. The calcium granules are remelted into an ingot and hot rolled to obtain solid bars more suitable for the preparation of charges. Carbon is added in the form of carbon-saturated iron containing approximately 4.3 wt% carbon. This material is prepared by vacuum melting the Ferrovac'E' iron in a crucible made of Graphite'G' type Graphiteo 2. Preparation of Charges It is necessary to carefully fit the melting stock into the crucible to obtain an adequately-sized melt, since it is not experimentally

-32TABLE I CHEMICAL ANALYSIS OF RAW MATERIALS MATERIAL SUPPLIER MAJOR IMPURITIES Grade 1 special purity Dominion N2 00011 percent Calcium granules Magnesium, Ltd, Mg less than. 01 percent Al 0.0028 percent Mn 0 002 percent Fe 0o006 percent Ferrovac'E' vacuum- Crucible Steel Ni o 035 percent melted iron Company Si less than 0.03 percent Cu Oo01 percent Cr less than 0.01 percent S 0 005 percent C 0,004 percent Al less than 00003 percent P 0o 002 percent Mn Oo 001 percent Mg less than o0 001 percent Ca less than 0 001 percent 99.99% aluminum Aluminum Company Cu Oo 002 percent of America Si 0,002 percent Fe 0.001 percent Carbon Graphite (Graphitite'G') Specialties, 99.5% + purity Inc. Gold Sheet J. Bishop and Co. Ag Cu Ca less than 0.02 percent total Mg Si Vacuum remelted Mond Nickel, Ltd. Co less than 0,01 percent Carbonyl nickel Mo less than Oo01 percent Mn less than Oo01 percent Zr less than Oo01 percent C 0,005 percent Si less than 0.005 percent Ti less than 0.005 percent S 0,024 percent Solar Grade E, I, DuPont No impurities present in any Silicon DeNemours and Co. significant amount.

-33feasible to add additional materials to the charge after meltdown as for open-air melts. The weight of alloying element was selected to provide for equal molar concentrations of the alloying element in the calcium-rich and iron-rich layers. In addition, a slight allowance is provided for preferential absorption of the alloying element by the titanium nitride crucible, since the alloying element melts down or is dissolved before the last iron has melted. By use of Vegard's law, the weightof the iron-base melt is calculated to correspond to a volume in the solid state of 2.85 cm3 at room temperature. For most heats the iron-rich layer weighs approximately 22 grams. The calcium-rich layer is calculated to have a final solid volume of 0.5 cm3, allowing for some loss by evaporation. The calcium and alloying element are contained within a hollow bomb-type iron crucible, Figure 16. After the charge components are carefully weighed and cleaned, the calcium and alloying element are loaded into the iron crucible and the tapered iron plug is pressed in place. This arrangement protects the calcium from oxidation, and prevents any vaporization loss of the calcium until the iron container has melted. 5. Furnace Assembly and Operation After the charge, crucible, and all components described in previous sections have been assembled in the pressure furnace, the sampling tube is carefully aligned above the coil assembly. The furnace lid is then bolted in place and the furnace is evacuated for two hours, flushed with argon, and then evacuated for one more hour. During operation, argon is admitted after the temperature has reached 15000F, and the furnace is pressurized to 200 psig. (This pressure, which is considerably higher than

-34Figure 16. Components of a Typical Charge. The components of the 5 At.% silicon charge are shown, which include from left to right the iron-bomb type container, the calcium, the silicon, and the iron plug..040.030 I'~~ —.-.020 —1 — W 1I O D10 0 0 I 2 3 4 5 6 HOLDING TIME AT 2925 F, MINUTES Figure 17. Time to Reach Equilibrium. The effect of holding time upon the solubility of calcium in pure iron at 2925~F is shown here.

-35the vapor pressure of calcium at 2925 F, is chosen to retard calcium vaporization.) The melt is held for three minutes at 2925~F, a period which is sufficient for attaining equilibrium, as seen in Figure 17. This relatively rapid rate of equilibration is quite consistent with the rate observed by Trojan and Flinn in the magnesium-iron-carbon system. ( At the conclusion of the equilibration period, the sampling tube is lowered into the iron-base melt, a sample is drawn, and the tube is returned immediately to its starting position. Samples are solid, smooth, shiny, and free from any calcium that was not in the iron-rich layer. In hevery case a calcium-rich layer is present in the crucible at the completion of a run. The appearance of a two-layer melt from a typical heat is shown in Figure 18. G, Chemical Analysis The determination of calcium in iron-base samples is not routine because relatively little analytical work has been done in this fieldo Furthermore, because a high degree of accuracy seemed desirable in this investigation, it was decided to develop and perform the analytical procedures as an integral part of the work. Two factors serve to complicate the analysis for calciumo In the first place, the chemical nature of calcium is such that it cannot be isolated directly from the other elements present, due to the high solubility of calcium compounds in aquaeous solutions. This requires that all other elements must first be removed before the concentration of calcium can be determined. The second complicating factor is the low level of

-56Figure 18. Two-Layer Melt. This photograph shows the calcium-rich layer (upper) and the iron-rich layer (lower) which are presented at the completion of a typical experimental run. No sample was drawn from this melt, and therefore the iron-rich layer is the same size as in a typical melt just before sampling.

-37calcium concentrations encountered in this investigation. This intensifies problems such as contamination, co-precipitation, interference by trace elements, and accuracy of measurements. These effects are less troublesome at concentrations two or three orders of magnitude higher. In this section, first the basic procedure for simple iron-calcium samples is outlined, then the modifications necessary for samples containing the various alloying elements, and finally the analysis for the various alloying elements. More detailed descriptions of the procedures, along with results of a reproducibility study are presented in Appendix II, 1. Basic Procedure for Unalloyed Samples In the analysis of the simple unalloyed samples, as well as in the procedures for alloyed samples, the removal of interfering elements by precipitation processes is avoided wherever possible. This is because a significant part of the trace calcium present can be lost by occlusion, absorption, or co-precipitation during such a process. Solvent extractions provide a way of avoiding certain precipitation processes, and the basic procedure involves two solvent extractions. The weight of the iron-base sample is 1.5 grams. In the first solvent extraction, iron is removed with isopropyl ether. Any remaining iron, as well as the titanium dissolved from the crucible, is removed by the use of cupferron (ammonium N-nitrosophenylhydroxylamine) and diethyl ether in a second solvent extraction. Calcium is then determined by titrating with O.001M versene (ethylenedinitrilotetracetic acid disodium salt) at a pH of 10.5, in the presence of eriochrome black T

-38indicator. A study of reproducibility for this basic procedure showed the standard deviation to be 4 percent of the amount analyzed at the 0.05 percent calcium level, an accuracy of +0.002 percent calcium at this level. 2. Procedures for Alloyed Samples. The basic procedure just described is quite compatible with the removal of the alloying elements aluminum, carbon, gold, nickel and silicon. Aluminum is removed by a second cupferron extraction at a pH of 2.5 to 4.5, carbon is removed by filtration, gold is extracted simultaneously with iron during the isopropyl ether extraction, while nickel is electrodeposited from an ammoniacal solution. Samples containing silicon are by far the most difficult to analyze for calcium. Silicon is removed as hydrated silica by two dehydration-and-filtration operations, followed by a procedure involving a cation exchange resin. Here the calcium is collected on the resin while the trace silica still in solution passes through the column and is discarded. Then the calcium is eluted from the column and analyzed. Three standards, covering the complete range of alloy concentration under test, along with a blank solution, are analyzed with each set of unknowns. Unknowns are analyzed for gold and nickel gravimetrically as a part of the procedure. Aluminum, carbon and silicon are determined by commercial laboratories previously found reliable for this type of investigation.

DISCUSSION OF RESULTS The data for the solubility of calcium in pure iron and in the various iron alloy systems are presented in Table II and Figures 19 - 23. It appears logical to divide the discussion of these data into three principal sections: lo Review of the data per se and in the light of previous investigations. 2. The engineering applications of the solubility data regarding refining reactions in actual process metallurgy. 3. The significance of the data in relation to certain theoretical concepts of liquid solutions. A. Review of the Data The data for the solubility of calcium in pure iron and in iron containing increasing amounts of the various alloying elements show regular changes in calcium concentration with increasing amounts of alloying element. Four of the elements raise the solubility of calcium sharply while gold lowers solubility strongly. 1. Solubility of Calcium in Pure Liquid Iron at 2925~F The restricted solubility of calcium in liquid iron at 2925~F, 0.032 wt% as shown by the average of duplicate heats, Figure 17, is consistent with the relatively large size of the calcium atom. Also, the immiscibility predicted by Hildebrand and Scott and by Mott for the -39

-40TABLE II CHEMICAL ANALYSIS OF EXPERIMENTAL HEATS Calcium in the Alloying Element Heat No. Iron-rich Layer in the Iron-rich Layer (Wt.%) (Wto) Pure Iron 99 0.032 102 0.031 Aluminum Series 125 o.o4o 1.07 126 0.052 2.69 127 o.o63 4.99 128 0.148 10o61 Carbon Series 122 0.044 03.7 (1) 123 0.052 0.74 (1 124 0.033 1.50 (1) Gold Series 118 0.026 0.19 118(2) 0.025 - 119(2) 0.014 0o36 120 0.004 0.90 121(2) 0.002 8.52 Nickel Series 115 0.041 2.13 112 0.051 5.14 113 0.082 10.35 114 0,207 19-71 Silicon Series 107 0.037 o 11 104 0.059 2.59 105 o.089 5.17 106 0.360 10,50 (1) Carbon concentrations obtained by subtracting average carbon pickup from the sampling tube, 0.04%, from the analyzed value, (2) Samples cut from melt.

-41-.16,.14,.14 io.10 _l..t.06 -- 3.04 0 0 1 2 3 4 5 6 7 8 9 10 II WEIGHT PERCENT ALUMINUM Figure 19. Effect of Aluminum Upon the Solubility of Calcium in Liquid Iron at 2925~F..16 --- THEORETICAL CALCIUM CONTENT OF.14 -- LIQUID IRON IN EQILIBRIUM WITH ca C2,&(co C2) = 25.6 KCAL/MOLE 12 ---------- LIMITS OF THEORETICAL CURVE O DUE TO ~ 3 KCAL ACCURACY OF <.10 - (Ca C2).08 _ 0.87 %C — C..06 -,- -- -. I I I ", H9.0 ^<| \,./Ca C2 LAYER PRESENT 3.042.02 LF, SAT'D WITH CaC, "' — --- (THEORETICAL)'- -- ~"t....t i 0 25.50.75 1.00 1.25 1.50 1.75 2.00 2.25 250 2.75 WEIGHT PERCENT CARBON Figure 20. Effect of Carbon Upon the Solubility of Calcium in Liquid Iron at 2925~F.

-42-.10 I3 v l l | | )0 RESULTS FOR CONVENTIONAL.08 - SAMPLES Q~~~~0 A ~RESULTS FOR SAMPLES i.06 CUT FROM MELT w 0 |.04 _ _-_ _-_ z.02 w A 0 2 3 4 5 6 7 8 9 10 WEIGHT PERCENT GOLD Figure 21, Effect of Gold Upon the Solubility of Calcium in Liquid Iron at 29250F..20 D /.16.14 __ z.04 w 0. H 08 ~04.02 0 2 4 6 8 10 12 14 16 18 20 WEIGHT PERCENT NICKEL Figure 22. Effect of Nickel Upon the Solubility of Calcium in Liquid Iron at 2925~F.

-43-.38.36.34.32.30.28.26.24 |.22.2.08.06.04.02 0 2 3 4 5 6 7 8 9 10 II WEIGHT PERCENT SILICON Figure 23. Effect of Silicon Upon the Solubility of Calcium in Liquid Iron at 2925~F.

-44iron-calcium system is confirmed. (9,10) Furthermore, the measured solubility of calcium corroborates the order-of-magnitude upper limit of 0.1 weight percent calcium at 3272~F, calculated from the data of Philbrook et.al. (11) The data refer of course to calcium solubility at 2925~F in the presence of a calcium-rich layer at 200 psig in a system containing argon. However, the solubility of calcium should be the same at any pressure sufficient to prevent the calcium-rich layer from boiling away, for reasons discussed previously. At pressures insufficient to maintain a calcium-rich layer, the calcium concentration should decline. Under such conditionsthe-phase rule variance at any temperature and pressure will be one, and the calcium concentration in the iron will be a function of the calcium concentration in the gaseous phase at the surface of the liquid iron. 2. Effects of Alloying Elements Upon the Solubility of Calcium in Liquid Iron at 2925~F The changes in the solubility of calcium in liquid iron with the additions of various alloying elements may now be considered, Table II, a. Aluminum Additions of aluminum increase strongly the solubility of calcium in liquid iron from the average value of 0.032 wt% in pure iron, to 0.148 wt% in iron containing 10.61 wt% aluminum, an increase of nearly five times, Figure 19. b. Carbon The solubility of calcium in liquid iron rises quite sharply initially as carbon is added to the system, Figure 20. At 0.74 wt% carbon

-45the corresponding calcium concentration in liquid iron is.052 wt%. At the highest carbon level in this series, Heat 124, 1.50 wt% carbon, the calcium concentration is less than at the two lower carbon levels. This effect in the high carbon heat is accompanied by the formation of a layer of calcium carbide, and the lower solubility may be readily explained by the presence of this new phase. At the temperature of this investigation, the theoretical carbon level at which calcium carbide should form in the presence of a calciumrich layer is approximately 0.87 wt% carbon. This is the carbon concentration at the junction in Figure 20 of the extended calcium solubility curve and the theoretical curve for the calcium concentration in an iron-rich layer saturated with calcium carbide. The theoretical curve is calculated for AFf(CaC2) = -25,600 cal/mole for calcium in the liquid state, and using the data of Rist and Chipman for the activity of carbon in iron. (21,22) The effect of carbon upon the activity coefficient of calcium in iron observed experimentally in this study, as well as the small mutual effect of calcium upon the activity coefficient of carbon in iron are taken into account, using the relationship considered by Ohtani and Gokcen for inter(23) action parameters: (2 3 = 2 2 3 This equation is valid when the concentrations of the two solute species in the ternary system approach zero, and is considered reasonably accurate for the present calculation. At this unique carbon concentration where the two curves meet, calcium carbide could exist at equilibrium with the two liquid phases and

-46the vapor phase at 2925~F. The phase rule indicates that one of the phases must be removed, however, before the carbon concentration of the iron melt becomes any higher under equilibrium conditions. In the presence of excess iron, as in these experiments, this could take place by the disappearance of the calcium-rich liquid by reaction with carbon to form calcium carbide. The equilibrium calcium concentration thus drops very sharply with increasing carbon content. The lower calcium concentration of Heat 124 is related to the shape of this theoretical curve. True equilibrium is not attained because a calcium-rich layer is still present at the conclusion of the heat. The calcium carbide may be considered as separating the melt into two systems, with the calcium carbide attempting to maintain dynamic equilibrium with the calcium-rich layer (and vapor phase) on one side and with the ironrich layer on the other. In view of the relatively close agreement between the high carbon data point and the theoretical curve, the couple between the iron-rich layer and calcium carbide is considered to be relatively close to dynamic equilibrium. c. Gold Additions of gold cause drastic reductions in the solubility of calcium in liquid iron, Figure 21. This effect is presumably related to the highly preferential distribution of gold in the calcium-rich layer, causing a very sharp reduction in calcium activity. An extreme indication of the preference of gold for the calcium rich layer is that at the highest gold level studied, nominally 20 atomic percent in the iron, the high-calcium layer sinks below the iron-rich layer, even though the density of iron is five times that of calcium.

-47The high preference of gold for the calcium-rich layer is presumably due to deviations from ideality in the gold-iron and gold-calcium systems. Gold is expected to exhibit strong positive deviations from Raoult's law when dissolved in iron, in the manner of copper. This is consistent with the similarity between the gold-iron and copper-iron phase diagrams. Furthermore, the presence of six intermetallic compounds in the gold-calcium system indicates strong negative deviations for gold in this system. Deviations of this type could easily result in a 500:1 partition of gold in calcium relative to iron. A conventional sample was not obtained from the iron-base layer of the high-gold heat, as well as from the second heat in the gold series. The analyses for these heats are taken from samples cut from the melt, as noted in the graph. To determine whether the melt sample was useful, a melt sample and a conventional sample were analyzed from the lowest gold heat. The data for these samples agree quite closely with each other, Figure 21, indicating that some confidence may be placed in the melt samples. d. Nickel Additions of nickel raise the solubility of calcium in liquid iron strongly, Figure 22. The 0.207 wt% Ca level at 19.71 wt% nickel represents a six-fold increase over the solubility in pure liquid iron. e. Silicon The highest calcium concentration in iron of this investigation is due to the addition of silicon, as seen from Figure 23. Calcium solubility is 0.360 wt% at 10.50 wt% silicon, an order-of-magnitude increase from the value for pure iron.

-48Bo Engineering Significance of the Solubility of Calcium in Liquid Iron In view of the present importance of calcium alloys and the possible use of metallic calcium as a refining agent in steelmaking, appreciable solubility of calcium in liquid iron is of considerable importance. The consequences of complete insolubility would be that when calcium vapor or metallic calcium were in contact with the liquid metal, any refining action would be confined to the contact surface layer. Because calcium can dissolve to a limited extent in liquid iron, however, the refining process should be much more effective. Calcium atoms can diffuse into the liquid steel and remove dissolved impurities at a considerable distance from the boundary of the calcium-rich phase. The potential use of calcium as a refining agent may now be considered first for open systems and then for pressurized systems. Finally, the importance of the immiscibility between liquid iron and liquid calcium is reviewed. 1. Calcium as a Refining Agent in Open Systems In an open system, the vapor pressure of calcium as a function of temperature must be considered, Figure 24. Because of the high vapor pressure of calcium, any liquid calcium would boil away rapidly in the temperature range shown, 2800 to 3000~F. If the activity of calcium in the pure liquid state is taken as unity, the maximum value of calcium activity at 1 atmosphere, a*, is the ratio of atmospheric pressure to the vapor Ca pressure, a* _ 14.7 (5) Ca V.P. of Ca The curve for a* is included in Figure 24. Boiling of the calcium can Ca be prevented, therefore, by keeping the calcium activity below a*. This Ca

-49I Jr VAPOR PRESSURE 50 - 1.0 VAPO(REFE PRENCE STATE: LIQUID C) 1 0 L ----------------.2.5 20.. -- -.4 QI3j I___ (REFERENCE STATEI I LIQUID Co) 0 i _.2 0.1 2800 2850 2900 2950 3000 TEMPERATURE,F Figure 24. Vapor Pressure of Calcium at Steelmaking Temperatures.

-50can be accomplished by employing an intermetallic compound with an element X, Ca X, for which the calcium activity is below a* It must be borne m n Ca in mind that if element X dissolves in the liquid iron before the calcium does, then the calcium would attempt to develop its equilibrium vapor pressure and boil away. Therefore if element X is highly soluble in liquid iron, the concentration of calcium in X should be kept quite low to prevent boiling. If element X is less soluble in iron, as in the case of carbon, then boiling away of the calcium is less likely. In addition, the negative slope of a* indicates that increasingly more stable compounds Ca X would Ca m n Ca be required to prevent boiling as temperature increases. Open melts of steel could also be refined by introducing calcium as a vapor instead of a compound. The vapor could be introduced directly or formed by slowly introducing metallic calcium which would vaporize rapidly. It would seem desirable to keep the vapor bubbles as fine as possible, to increase the chances of dissolving before being lost to the atmosphere over the melt. 2. Calcium as a Refining Agent in Pressurized Systems Metallic calcium can be maintained in contact with liquid steel if the pressure of the system exceeds the vapor pressure of the calcium, as in this investigation. Of course some vapor state diffusion can occur, but this is a much slower process than boiling away. If the problem of boiling is eliminated, refining by the calcium takes place for an extended length of time. The process could be enhanced by magnetic stirring of the melt, or by introducing the calcium below the surface of the melt. In order to observe the refining action of calcium in such a system, two samples of 4340 steel of the following analyses were melted

-51under a calcium layer in the pressure chamber. Melts were held 3 minutes at 2925 ~F COMPOSITION (wt %) C Mn Cr Ni Mo 02 S 2054-A o.42 0.81 0.81 1.98 0.24 0,0135 0.015 2054-B.41 0.78 0.81 1.92 0.23 0.0160 0.013 (Vacuum degassed) Both samples are from the same heat of steel, but sample 2054-B is from metal which was vacuum degassed before pouring the ingot. The effects of remelting the samples under calcium upon oxygen and sulfur concentrations are of particular interest: OXYGEN Before Calcium Treatment After Calcium Treatment 2054-A 135 parts per million 3 parts per million 2054-B 160 parts per million less than 3 parts per million SULFUR Before Calcium Treatment After Calcium Treatment 2054-A.015 wt %.013 wt % 2054-B.013 wt %.013 wt % The oxygen is almost completely removed by the calcium, being reduced from 135 ppm to the limit of analytical error. This reduction is in sharp contrast to the effect of vacuum degassing, after which the reported oxygen concentration is actually higher. The sulfur content is not affected consistently, with only a moderate reduction in one of the samples. This apparent resistance of

-52the sulfur to being removed by calcium is considered related to the fact that the initial activity of the sulfur in the samples is quite low, since sulfur has essentially unlimited solubility in liquid iron. Therefore tIe activity of the sulfur corresponding to the initial concentration of sulfur is already quite low, making a further reduction more difficult. Since oxygen on the other hand has very restricted solubility in liquid iron, its initial activity in these samples is relatively high, thus favoring a further reduction in activity by the calcium. It should also be possible to desulfurize better in the presence of a basic slag using calcium, since the calcium sulfide would have a lower activity. 3. Importance of the Immiscibility Between Calcium and Iron While the lack of extensive solubility of calcium in liquid iron is in some respects disadvantageous from a refining point of view, it does provide, however, a unique advantage over certain conventional refining agents. This advantage is that virtually no calcium should be present in the final product. In contrast to this, deoxidants such as aluminum and silicon are quite soluble in both liquid and solid iron, and therefore the alloying and graphitizing effectsof such elements in the solid state must be considered when they are used as refining agents. 4, Significance of the Effects of Alloying Elements upon the Solubility of Calcium in Liquid Iron Since the five elements of this study have strong effects upon the solubility of calcium in liquid iron, the influence of these changes upon the use of calcium as a refining agent should be considered. Also these influences upon the facility of preparing ferroalloys with large amounts of calcium are discussed.

-53a. Influence of Alloying Elements in Steel upon the Refining Action of Calcium The elements which increase the solubility of calcium would be expected to increase its effectiveness as a refining agent. This observation assumes that diffusion of calcium through liquid iron is the ratecontrolling step. Higher solubility would permit steeper concentration gradients to be established, resulting in faster diffusion of the calcium in quest of the impurity atoms. Considering the curves of Figure 19 through Figure 23, these elements can be rated according to their expected effectiveness in this regard. The percentage increase in calcium concentration at one weight percent alloying element is used here as a means of estimating the initial effectiveness of an alloying element in promoting the refining action of calcium. Alloying Element Change in Calcium Solubility 1 wt% C 90 Pet increase* 1 wt% Si 25 Pet increase 1 wt% Al 20 Pct increase 1 wt% Ni 10 Pet increase *Extrapolated value These data show that carbon should be by far the most effective of the four alloying elements on a weight percent basis in enhancing the refining action of calcium. Nickel should be the least effective, with silicon and aluminum falling between the two extremes. A corollary of the effect just considered is that the tendency to retain calcium in the solid steel would be greatestst for alloying elements which increase calcium solubility in liquid iron most strongly. While gold

-54is not commonly added to iron, its effect is quite important inasmuch as it indicates certain types of elements which might have a similar influence upon the solubility of calcium in liquid iron. According to the considerations of section A-2(c), the decrease in calcium solubility upon addition of gold is due to expected positive deviations from ideality for solutions of gold in iron, and negative deviations for solutions of gold in calcium. Copper is expected to lower the solubility of calcium in liquid iron in view of its very strong positive deviations from ideality when dissolved in iron. Copper therefore should sharply reduce the effectiveness of calcium as a refining agent. Other elements having similar deviations from ideality should behave in the same way. b. Ferroalloys Containing Large Amounts of Calcium The strong effects exerted by some of the alloying elements upon the solubility of calcium in liquid iron raise the possibility of producing alloys of iron and some other element which contain large amounts of calcium. Such ferroalloys could be used as a means of adding calcium to steel. Silicon is most interesting in this regard. Ten wt% silicon raises the calcium solubility by an order of magnitude at 2925 ~F, and if this rate of increase is projected to higher silicon levels, the miscibility gap between calcium and iron should be closed at a silicon concentration of approximately 30 to 40 wt%. Ferrosilicon alloys containing very high concentrations of calcium are therefore possible. This effect is the basis of the present calcium-bearing ferrosilicon alloys. The same possibilities exist for calcium-bearing ferroalloys of aluminum and nickel, although

-55progressively higher levels of aluminum and of nickel should be required to close the miscibility gap between iron and calcium. C, Theoretical Considerations Regarding the Effects of Third Elements While the influence of the various alloying elements upon calci~m. solubility is quite important from an engineering point of view, the theoretical implications of these relations are also of interest. Some theoretical treatments of the effects of third elements have been developed, but these have not been tested extensively, especially for systems such as the present one in which the solute of interest is a metal. The present data may therefore be helpful in the understanding and application of existing theories, The most prominent theories of interest are those of Alcock and Richardson and of Wagner and these are related to the present data. ( 45 1, The Alcock and Richardson Equation From the analysis of a chemical model which considers the energies of old bonds broken and new bonds formed when solute 2 is added to a dilute solution of solute 3 in solvent 1, Alcock and Richardson obtain an equation which may be written as follows for the present study: (~(ln 7C-a(e=Ex=lny(X)- lnFny (6) (c8 \X0 Ca Ca(x - n (Ca(Fe) X(Fe)' where YCa(Fe) activity coefficient of calcium in iron, or in ironbase solutions NX = mole fraction of element X, EX Ca = interaction parameter describing the effect of element X upon the activity coefficient of calcium in iron.

-56YCa(X) = activity coefficient of calcium in element X, Y X - activity coefficient of element X in iron, X(Fe) For clarity, the activity coefficient of calcium, rather than the solubility of calcium, is considered in this section, the two terms being uniquely related by the equation NCa(Fe) Ca(Fe) aCa (7) Here N a is inversely related to the calcium activity coefficient Ca(Fe) YCa(Fe) At calcium saturation in the special case of relatively constant activity aa an increase in calcium solubility NCa(Fe) is signified by a decrease in the calcium activity coefficient YC (Fe) For this special case negative values of the interaction parameter eX indicate that Ca additions of element X increase the solubility of calcium in iron, Because the calcium activity may change rather sharply, however, the activity coefficient y7(F) provides a much better indication of the atomic interactions which occur, While the Alcock and Richardson equation should permit a quantitative prediction of E C this is only possible when data are available for each of the three activity coefficients on the right-hand side of Equation (6), The second term, -lny Ca(Fe) for this study is - 7*7, the third term, -lny (F is 0.0 for nickel, 0,5 for cabon or aluminumer and 4O fore i (22,2) No data are availcarbon, 5,1 for aluminum No data are available for gold, Furthermore, no data are available for the first term, ln7 c(X) for any of the alloying elements, In the unlikely event that this term should have a rather high positive value, it is evident that this term should have a rather high positive value, it is evident that

-57the second term would still predominate, leading to negative interaction parameters for all of the present alloying elements, a, Determination of Interaction Parameters The interaction parameters for the present data may be determined from Equation (7), The value of NCa(Fe) is the analyzed calcium concentration in the iron, expressed as an atom fraction, The limiting reasonable values of aCa are considered here for the various alloying elements in the absence of activity data for these elements in the calcium-rich layer, Aluminum, silicon and carbon are expected to dissolve preferentially in iron in view of the negative deviations from ideality for these elements in iron. Furthermore, nickel, which forms ideal solutions with iron, can be shown to favor solution in the iron layer slightly by a mass balance, Thus the concentration of alloying element in the calcium-rich layer should not exceed that in the iron-rich layer, this limiting value representing a 1:1 partition of alloying element between the two layers, Assuming Raoult's law for the activity of calcium in the calcium-rich layer, this establishes the lower limit of calcium activity as 1-N X(Fe)* The lower limit of alloy concentration in the calcium-rich layer is zero, for which the upper limit of calcium activity is unity. Thus when the iron contains 2 at % aluminum, for example, the calcium activity should be between 0,98 and 1.00. At 5 at % aluminum, the range would be 0,95 to 1.00; at 10 at % aluminum, 0.90 to 1,00; and at 20 at % aluminum, 0,80 to 1,00, The range of uncertainty regarding calcium activity is seen to be quite small at low aluminum levels, increasing as the aluminum concentration of the iron increases,

-58The actual determination of interaction parameters may be followed in Table III for the case of an equal molar concentration of the alloying element in both liquids. The activity of the calcium, aCa, is taken as the mole fraction of calcium in the calcium-rich liquid, which for the 1:1 partition of X considered here equals (1-NX(Fe)). From this the activity coefficient of calcium in the iron-rich layer is determined by the relationship 7Ca(Fe) = aa (8) NCa(Fe) Substituting Ca(Fe) = X(Fe (9) NCa(Fe) Activity coefficients of calcium are seen to range from an average of 2270 for calcium in pure iron to 178 for calcium in iron containing 19.02 At.% (10.5 Wt.%) silicon. The natural logarithm of 7Ca(Fe) is plotted against NX(Fe) in Figure 25, and the interaction parameter for this 1:1 partition termed (ea ) is the limiting slope as NX - 0. Ca 1:1 X Values of (c )ll are listed in Table III. The dashed curve of Figure 25 represents lnYc(Fe) for a (F ) = 1, the value assuming that none of the alloying element enters the calcium-rich layer. The limiting slope for this condition is represented by (eX ) Ca 0:1' and may be shown analytically to be 1 greater than that for the condition of a 1:1 partition, Thus, (eXa) = (EXa) + 1 (10) Ca 0:1 Ca 1:1

TABLE III DETERMINATION OF INTERACTION PARAMETER cX Ca Heat Number (NCa(Fe))102 (NX(Fe))102 aCa=-(1NX(Fe)) Y Ca(Fe) aa ) lnyCa(Fe) (Ca)l:1 (CXa): Ca(Range) (from Fig. 25) [=(e Xa)1.-+l Pure Iron 99 0.045 ---- 1.0000 2220 7.706 100 0.043. --- 1.0000 2320 7.750 Aluminum Series -8. 0 -7.0 -7.0 to -8.o 125 0.056 2 19 O. 9781 1750 7.469 126 0.070 5.41 0.9459 1350 7.208 127 o.o83 9.81 0.9019 1090 6.994 128 0.185 19.72 0.8028 434 6.073 Carbon Series -16.3 -15.3 -15.3 to -16.5 122 0.061 1.70 o,9830 1610 7.386 123 0.070 3.35 0.9665 1380 7.231 Nickel Series -11.2 -10.2 -10.2 to -11.2 115 0.057 2.03 0.9797 1720 7.453 112 0.072 4.9C 0.9410 1310 7.178 113 0.115 9.90 0.9010 783 6.663 114 0.292 18.97 0.8103 278 5.627 Silicon Series -11.2 -10.2 -10.2 to -11.2 107 0.051 2.19 0.9781 1920 7.561 104 0.081 5.02 0.9498 1170 7.066 105 0.118 9.79 0.9021 764 6.639 106 o.456 19.02 0.8098 178 5.182

-6o8.0 I I — 8.) / -L SLOPE AT 0% ALUMINUM =-7.0 L 7.0 L _ —UALUMINUM SERIES; 7.0 --- SLOPE AT 0% ALUMINUMI-8.0 -_ 6.0 ----- ----- 8.0 I I I 1y S SLOPE AT 0% CARBON=-I5.3 - 7.0 7 CARBON SERIES 7.0 -5 - SLOPE AT 0% CARBON -16.3 6.0 -- -— I I I - 6.0 L -SLOPE AT 0% NICKEL=-10. 2 6.0 --—.5.0 8.0 I —-I I' I SLOPE AT 0% SILICON =-10.2 F 7.0 2o A i I SILICON SERIES n- |SLOPE AT 0% SILICON=-11.2 |^ 6.0 11 1| — BASED ON NX(CrNx(Fe)' ---- BASED ON Nx(ca=0 5.0 0 2 4 6 8 10 12 14 16 18 20 ATOMIC PERCENT ALLOYING ELEMENT Figure 25. Effect of Alloying Element Concentration in the IronRich LiJquid Upon the Activity Coefficient of Calcium in the Iron-Rich Liquid at 2925~F.

-61The value of (eX foraluminum, for example, is determined by a 11 I Al Al (e: ) = (6) +1 Ca 0-1 Ca 1.1 = -8,0+1 - -.70 Values of (eX ) are included in Table III, and the range of values Ca' 0:1 for the two terms is listed here: X EC (Range) Ca Alloying Element (X ) to (cX ) Ca 0:1 Ca 1:1 Al - 700 to - 8,0 Ni -10.2 to -11.2 Si -10.2 to -11,2 C -15.3 to -16.3 From this, the range of f" due to the uncertainty of calcium activity Ca is relatively small, and does not greatly influence the rank of the alloying elements. The heats containing gold do not lend themselves to this evaluation of the interaction parameter, This is so because the extremely high concentration of gold in the calcium layer prevents the making of a reasonable estimate of the calcium activity in this layer. b. Direct Application of the Alcock and Richardson Equation Considering the values Ea in the light of the Alcock and RichardCa son equation, the prediction of negative values of EX is verified for the Ca elements considered here, aluminum, carbon, nickel and silicon. Theoretical values of da cannot be calculated for purposes of comparison with the experimental values, however, because of lack of data for experimental values, however, because of lack of data for 7Ca(X)'

-62c, Qualitative Consideration of the Alcock and Richardson Equation If the terms on the right hand side of Equation (6) are considered individually, it is seen that this equation may be interpreted in terms of atomistic effects between the different species, In this regard, a small activity coefficient (less than 1) is considered indicative of an attraction between the two different atoms, while a large activity coefficient (greater than 1) is considered related to a repulsion between the two atoms. Considering the second term (-ln Ca(Fe)), which predominates in this equation because of the immiscibility in the iron-calcium system, this term very strongly favors negative values for - aO Atomistically, this means that iron and calcium atoms have strong repulsions for each other, making it very easy, therefore, for element X to exhibit its attraction for calcium. Qualitatively then, most elements are expected to lower the activity coefficient of calcium in iron, resulting in negative values for X, as with the four elements considered here, Since the second term of Equation (6) is the same for all alloying elements, attention may be focused on the remaining two terms, (lny ) Ca(X) and (-lny( ), in an effort to understand how one element might be more X(Fe) effective than another in changing the activity coefficient of calcium in iron, A negative interaction parameter is favored by negative values of lny,(X)' indicative of an attraction between calcium and X, and by positive values of lny (Fe), indicative of a repulsion between iron and X, Atomistically, this means that if X has a greater attraction for calcium atoms than for iron atoms, a negative value of cX is aa favored,

-63Two different alloying elements may be compared on this basis, regarding their effect on the interaction parameter E, by the following Ca criterion. The element which has the greater attraction for calcium atoms as compared to its attraction for iron atoms should result in the more negative value of e. Thus it should be possible to estimate which of Ca two elements would cause the more negative value of ~ C if information Ca concerning the attraction of these two elements for calcium and for iron is available. One reasonable index of the attraction between atoms is the standard free energy of formation of a compound formed by the two atoms at the temperature of the study, 2925~F. Of the alloying elements considered here data are available only for carbon. (21) For the other elements, one index which suggests itself is to compare the number of intermetallic compounds in the system X-Fe with that in the system X-Ca. An evaluation of the relative magnitude of the attractions X-Ca and X-Fe is presented in Table IV, with the alloying elements arranged in order of increasing negative values of eX (progressively greater reducCa Ca, (6) tions in the activity coefficient of calcium 7Ca(Fe))~ Table IV shows that there is some relation between this criterion of atom preference and the interaction parameter ~X While it would Ca be unreasonable to expect this criterion to correlate exactly with the values of C, it does seem significant that aluminum, which reduces 7 (Fe) the least, indicates a preferential attraction for iron atoms, while carbon, which reduces YCa(Fe) the most, indicates a preferential attraction for calcium atoms. In the absence of suitable activity data, the criteria of free energy of compound formation, or of the number of intermetallic compounds

TABLE IV RELATION OF ATOMISTIC PREFERENCE OF ELEMENT X TO OBSERVED INTERACTION PARAMETER eX Ca Alloying Compound formation Compound formation Indicated preference e (Range) in the- system 7A f r-Ca(6 Ca Element X in the system X-Fe(6) of X atoms Al 2 compounds 4 compounds Preference for Fe Atoms -7.0 to -8.0 N1 1 compound 0 compounds Preference for Ca Atoms -10.2 to -11.2 Si 3 compounds 3 compounds Indefinite -10.2 to -11.2 C AF0 = -12,800o* AF = -1,600 Preference for Ca Atoms -15.3 to -16.3 f f for 1/2 Ca C for Fe3C at at 2925~F (21) 2925~F.(21) * Reference of Ca taken as liquid state.

-65formed, may serve as a useful guide in assessing which of several elements will change most sharply the activity coefficient of a solute species. 2. Wagner's Electron-to-Atom-Ratio Concept The application of Wagner's theory requires a knowledge of the number of free electrons which an atom possesses in a metallic solution. As a first approach to the problem (following Wagner) the number of valence electrons may be used to represent this number. a. Correlation of the Data With Chemical Valence The interaction parameters from the previous section may be now compared with the chemical valence for each element in question: Element Chemical Valence e (Range) Ca Ca 2 Fe 2,3 Al 3 -7.0 to -8.0 Ni 2,3 -10.2 to -11.2 Si 4 -10.2 to -11.2 C 4 -15.3 to -16.3 If the valence numbers are applied directly to the Wagner model, two cases need to be considered, in view of the two possible valences for iron. If iron is considered first in the plus 2 valence state, it has the same number of free electrons as calcium, and therefore it is not possible that the number of free electrons of alloying element X and of calcium could differ in opposite directions from that for iron. One possible interpretation of the Wagner theory for a situation of this type is that if the valence of X were the same as for iron and calcium, then the electron chemical potential would not change, and the activity coefficient of the

-66X calcium would remain the same, corresponding to a value of zero for E Ca For any third element having a lower or higher valence than that of calcium and iron, the chemical potential of the electrons would change, and an increase in activity coefficient of the calcium is indicated. This would X result in a positive value of the interaction parameter,. The experiCa mental values do not support either of these predictions. If the plus 5 valence state of iron is considered, then nickel for the plus 2 state should give a positive value of e, aluminum and Ca nickel in the plus 3 state should result in a value of zero for e Ca While these values agree to some extent with the experimental data, the level of agreement is poor. The use of valence as a guide to the number of free electrons is a rather unsatisfactory index for most elements, and one which Wagner avoids when other means are available. The chemical valence can be a very misleading guide to the number of free electrons contributed to the bulk metal by an element. Silicon, with a valence of plus 4, might be expected to contribute more free electrons than calcium, with a valence of plus 2. The very low electrical conductivity of metallic silicon relative to that of calcium, however, refutes this possibility. Furthermore, the use of chemical valence would indicate that all elements with the same valence should contribute free electrons to a metallic solution with equal facility in spite of differences in electron structure between atomso Still other objections to the use of valence as a means of interpreting the Wagner theory can be raised.

-67b. Alternate Criteria for Assessing the Number of Free Electrons Various other characteristics of an element may be considered as an index of how readily an element gives up electrons when it dissolves in a metallic solution. One index of interest is electronegativity, which Pauling defines as the power of an atom in a molecule to attract electrons to itself.(25) By this definition, an atom of high electronegativity such as carbon might be expected to contribute fewer free electrons than an element such as calcium of low electronegativity. While the absolute value of electronegativity may not be meaningful here, the relative values for the various elements may provide a useful index of the number of free electrons contributed by an atom. In Table V electronegativity values are listed for calcium, iron and the alloying elements, along with the interaction parameters ~ Ca' The values of electronegativity are those for the model of Pauling based on bond energies.(26) Considering the electronegativity calculated for the plus 2 state of iron, the theory of Wagner suggests that E be positive Ca for aluminum and negative for nickel, silicon and carbon. All of the elements except aluminum conform to the theory, using Pauling's electronegativity as an index of the contribution of free electrons. Furthermore, the interaction parameters are seen to become more negative in approximate order of increasing electronegativity. For the present data, electronegativitie's permit a reasonably good qualitative and quantitative correlation with the theory of Wagner. The other possible indexes of the number of free electrons which an element contributes in a metallic solution, the ionization potential and the photoelectric work function, may also be considered.

TABLE V ALTERNATE CRITERIA FOR APPLYING WAGNER'S ELECTRON THEORY Photoelectric Element ~&C (Range) Electronegativity 1st Ionization Potential Work Function Ca Pauling's E.N. Predicted Value in Predicted Value in Predicted sign of e6a Electron sign of eX Electron sign of Ca Ca C Volts Volts** Ca Ca ---- 1.0 6.11 2.86 Fe ---- FeII 1.65 7.90 4.47 n IIII (Fe 1.80) Al -7.0 to -8.0 1.5 (+) 5.98 (+) 5.80 (+) Ni -10.2 to -11.2 1.7 (-) 7.65 (+) 4.61 (-) Si -10.2 to -11.2 1.8 -) 8.15 (-) 4.52 (-) C -15.3 to -16. 2.5 (-) 11.26 4.81 (-) II * Based on Fe11 ** Average of all listed values for each element.

-69The first ionization potential, which is the energy required to remove an electron from a neutral free atom, is listed in Table V for each element.(27) Considering the ionization potentials in the light of Wagner's theory, positive values of EX are predicted for aluminum and Ca nickel, and negative values for silicon and carbon. Silicon and carbon corroborate the Wagner theory using ionization potential as an index as to the contribution of free electrons. As with electronegativity, the values of ionization potential change rather regularly with the observed interaction parameters. Finally, photoelectric work function may be considered as a guide to the contribution of free electrons by an element. The photoelectric work function is the energy a photon must have in order to remove an electron from an element, ordinarily measured for the solid stateo Average values of the photoelectric work function are listed in Table V.(28) The signs of the interaction parameters for all the elements except aluminum corroborate the theory of Wagner on this basis, and again the values of the interaction parameters are in approximate order of increasing values of the work function. In summary, three separate guides as to the contribution of free electrons by an element dissolved in metallic solution are considered as a means of applying the qualitative theory of Wagner. In eight of twelve instances, the sign of the predicted interaction parameter agrees with that of the experimental value. Furthermore, experimental values of the interaction parameters change in an essentially regular manner with changes in each of the indexes considered. This suggests the possible use

-70of these indexes as a guide in predicting which of two elements will have the greater effect in changing the activity coefficient of an element in metallic solution. c. The Wagner Equation for the Electron-to-Atom-Ratio Concept The equation derived by Wagner for the electron model of metallic solutions may be written X Ca (Ca +X 1/(11) =e = + o Ca X \ Ca X NCa- O, NX O Ca In attempting to apply this equation to the present work, a serious problem arises. The self-interaction parameter ECa which represents the initial change in the activity coefficient of calcium in iron according to the relationship n 7 Ca(e), is expected to be negative for this Ca immiscible system. Such behavior is observed for the immiscibility between (29) x zinc and lead, for example.9) At the same time, the term X for aluminum, silicon and carbon in iron is positive, leading to imaginary values of silicon and carbon in iron is positive, leading to imaginary values of 6Ca' which have no meaning. For the alloying element nickel, which exhibits ideal behavior in liquid iron, the term ~Ni is essentially zero, leading Ni to a value of zero for E. This is quite inconsistent with the observed Ca interaction parameter for nickel. From the foregoing, the quantitative relationship of Wagner does not lend itself to an analysis of the present datao 3, Summary of the Present Data in Relation to Theory The chemical model of Alcock and Richardson and the electron model of Wagner have been considered in the light of the present data.

-71The former model correctly predicts negative values of the interaction parameters, and permits the alloying elements to be arranged in approximate order of increasing magnitude of this parameter by the use of a suggested criterion. The latter model, applied by means of several suggested criteria, is somewhat less successful in predicting the negative values of interaction parameter observed. By these criteria, however, the relative effects of various elements upon the activity of calcium are rather accurately predicted. Thus both theories provide reasonable agreement with the experimental data, in spite of differences in the theoretical models. The model of Alcock and Richardson considers interactions between the various pairs of atoms. Electron effects are not considered directly, although they presumably are reflected in the bond energies involved in this analysis. The electron model presented by Wagner takes into account the relative number of free electrons contributed by solvent and solute atoms, but does not allow for interaction effects between positive metal ions. In view of the reasonably good agreement between both theoretical models and the present experimental data, it would seem that the contributing of free electrons cannot be completely divorced from any interactions between the positive metal ions which contributed the electrons. The studies of Alcock and Richardson and Wagner are, therefore, distinct but reasonably compatible approaches to the same problem.

CONCLUSIONS On the basis of the present investigation, the following conclusions may be drawn: 1. Calcium has significant solubility, 0.032 Wt, %, in liquid iron at 2925~F, The corresponding thermodynamic activity coefficient for calcium in liquid iron is 2270. 2. The solubility of calcium in liquid iron is strongly affected by the alloying elements studied. 3. The solubility effects observed have important implications in the field of ferrous process metallurgy. 4. The effects of alloying elements agree reasonably well with current theories, Criteria are suggested to facilitate the application of these theories to studies of this type. 5. A sampling technique and crucible material have been developed which facilitate future investigations of reactive metals of high vapor pressure, Many topics for future study are suggested by the present research. Profitable areas of future work in systems of this type might include the following: the additivity of the effects of alloying elements; the effects of alloying elements at the extreme ends of the scales of electronegativity, ionization potential, or work function; the concentrations of alloying elements required to close the miscibility gap; and the solid solubility of calcium and other immiscible elements in ternary alloys of iron. -72

APPENDIX I CRUCIBLE MATERIALS 1o Evaluation of Existing Refractories Pieces from the best crucibles commercially available were exposed to molten lithium for approximately 10 minutes at 2500~F. Because lithium is more active chemically than calcium, this was considered to be an indication of the ability of a crucible to withstand attack by molten calcium at the higher temperatures of the present study. Tests were conducted by placing a small piece of the crucible to be tested, 1/16 in. to 1/8 in. in thickness, along with a piece of lithium, in a hollow iron bomb-type container, and Inserting a tapered iron plug. This was done inside a glove box, and the plug was subsequently arc-welded shut. These containers were then heated to 2500~F in an induction furnace. The materials tested were as follows: TAM ZrO2 Norton MgO, mix M 204 Magnafrax 0340 high purity MgO Triangle RR A1203 Thoria-Lined Triangle RR A1203 Leco ThO2 Synthetic sapphire single crystal Silicon nitride, experimental grade Boron nitride The level of performance of all materials was unsatisfactory, with the pieces being converted into a darkened and softened mass. Pieces of the -73

-74titanium nitride material developed in this investigation were unaffected by lithium in these tests. In addition, crucibles of hafnium carbide backed up by tungsten, produced by plasma spraying, were considered for this study. However the experimental crucible of this type failed by thermal shock when iron was melted in it. 2. Typical Data for a Set of Crucibles Compacted crucibles are sintered in groups of five, and are subsequently nitrided in larger groups. The data for the five crucibles sintered in Heat C-42 are presented in Table I, and characteristics are seen to be rather uniform. The drop in water level is reasonably low, approximately 3/32 in. in 1 hour. Linear shrinkage is seen to average 13.1 percent and sintered densities average 4.48 gm/cm3. During the nitriding treatment which follows sintering, the crucibles gained approximately 5 percent in weight, attaining an average density of 4,65 gm/cm. This corresponds to 86 percent of the theoretical density of 5.43 gm/cm3 for titanium nitride of stoichimaetric composition.(19)

APPENDIX I TABLE I CHARACTERISTICS OF CRUCIBLES FROM SINTERING HEAT C-42 Crucible Drop in Water Linear Density Weight Density No. level after Shrinkage after gain after 1 hr. (In.) (Pct.) Sintering during Nitriding (gm/cm3) Nitriding (gm/cm ) 158 3/32 13.2 4.48 5.1% 4.65 160 1/16 13.3 4.48 5.0% 4.65 161 3/32 12.7 4.48 5.1% 4.65 162 3/32 13.1 4.49 4.7% 4.66 163 3/32 13.1 4.48 4.8% 4.65 Average 3/32 13.1 4.48 4.9% 4.65 -75

APPENDIX II CHEMICAL ANALYSIS In this section the chemical analysis procedure for calcium is considered in detail. The determination of calcium in unalloyed samples is considered first, followed by sections for the analysis of alloyed samples. A. Determination of Calcium in Unalloyed Samples The unalloyed samples contain approximately 0.25% titanium picked up from the titanium nitride crucible, and trace amounts of nickel, silicon, copper, chromium and other elements, in addition to calcium. The key steps in the basic procedure are carefully considered here, as well as the reproducibility for this procedure, and the actual step-by-step procedure employed. 1. Basic Operations for Unalloyed Samples a. Dissolving of samples. The 1.5 gm iron samples are dissolved in aqua regia. This permits samples to be dissolved in less than 30 minutes as contrasted to 8 to 12 hours required to dissolve the samples in diluted hydrochloric acid. Samples are taken to dryness to expel the nitric acid. b. Removal of iron. Morrison and Freiser recommend extracting iron as the ferric ion from a 7.75-8.0 M hydrochloric acid solution with isopropyl ether.(3 Accordingly, the iron is oxidized with 30% hydrogen peroxide and then extracted with isopropyl ether in a solution containing 2 parts hydrochloric acid to 1 part water. It is possible to extract at least 99.95 percent of the iron in a double extraction by insuring that the iron is fully oxidized. -76

-77c. Removal of titanium and trace iron. The organic reagent cupferron is capable of quantitatively precipitating titanium, ferric iron, and other metals from solution.(31) Because the action is less specific at low acidity, the acidity is purposely reduced to approximately 2% hydrochloric acid for this operation. The precipitation is conducted in the presence of diethyl ether, with the precipitate dissolving entirely in the ether. This avoids the danger of losing calcium on the precipitate, and is possible because calcium chloride is insoluble in diethyl ether. No trace of iron or titanium remains after this step. d. Determination of calcium. Versene is commonly used in a titration to measure relatively high calcium concentrations. The exhaustive study of Schwarzenbach was found extremely useful in adapting the use of versene to the present determinations of trace levels of calcium.() A dilute versene solution, 0.001 M, permits accurate titrations with a 50 ml burette. Solutions are adjusted to a pH of 10.5 with a buffer solution. Approximately 0.05 Mg of magnesium is added as magnesium versenate to sharpen the endpoint. It is necessary to mask the trace nickel, chromium and any other interfering elements with sodium cyanide to permit the observation of the endpoint. Two aliquots of each solution are titrated. A concentration series was titrated and found to give exact linearity between the volume of versene and the amount of calcium taken, even at the low concentrations of this study. 2. Reproducibility for Unalloyed Samples Five synthetic unknowns containing 0.0467% calcium were prepared and analyzed for the purpose of assessing the accuracy of the method for

-78unalloyed samples. The percentage deviations of analyzed calcium from the mean for the group are listed: Solution Number Percentage deviation from mean 264 + 4.70% 265 - 3.27% 266 - 3.42% 267 - 1.85% 268 + 3.80% The standard deviation a for this set is 3.96% of the amount analyzed, or 0.00184% calcium. 3. Standard Procedure (for Analyzing Unalloyed Samples) Stock Solutions: 1. 100 ppm calcium standard. Prepared by dilution from a 952 ppm calcium standard which was analyzed by a standard calcium oxalate precipitation (triplicates analyzed). The standard calcium solutions have an HC1 concentration of 1%. 2. 2000 ppm titanium solution. Sponge titanium dissolved in 25% HC1. 3. 2/1 HC1 solution. Two parts concentrated hydrochloric acid to one part demineralized H20. 4. 1% HC1 solution. One part concentrated hydrochloric acid to 99 parts demineralized H20.

-795. H20. All H20 employed in this study is demineralized, obtained by passing distilled H20 through a "Deeminac" ion exchange resin cartridge. 6. 5% Cupferron solution. Five gm cupferron dissolved in 100 ml H20, freshly prepared. 7. pH 10.5 buffer. NH40H/NHC1 solution, adjusted to pH 10.5 on pH meter. Procedure: 1. Prepare blank by adding 2.25 ml of 2000 ppm Ti solution to 1.5 i Ferrovac'E' iron. 2. Prepare 0.0667% Ca standards by adding 10 ml of 100 ppm Ca standard and 2.25 ml of 2000 ppm Ti solution to 1.5 gm Ferrovac'E' iron. 5. Dissolve blank, standards and unknowns in 10 ml conc. HC1 and 5 ml HN03, in 150 ml beakers with watch glasses. Heat to initiate reaction. 4. Raise watch glasses on glass hooks, evaporate without boiling carefully to dryness to expel nitric acid fumes. 5. Cool, remove glass hooks, and drench with 2/1 HC1. Heat to redissolve salts. 6. Cool, add 3 ml 30% H202, simmer 30 minutes. Remove from heat, add 3 drops 30o H202. 7. When cool, dilute to 50 ml in 600 ml separatory funnel. Extract with 200 ml isopropyl ether. Rinse with 5 ml 2/1 HC1. Re-extract with 100 ml isopropyl ether. Rinse with 5 ml 2/1 HC1.

-808. Evaporate solution to 5 + 1 ml without boiling, using watch glasses and glass hooks. Add 10 ml H0, 1 ml 50% H202 evaporate solution to 5 + 1 ml. without boiling. 9. Dilute with 1% HC1 to 50 ml in separatory funnel. Add 100 ml diethyl ether. Add cupferron batchwise until yellow titanium precipitate stops forming. Shake until solution clears. Rinse with 5 ml 1% HC1. Re-extract using 1 ml 5% cupferron solution and 100 ml diethyl ether, and rinse with 5 ml 1% HC1. 10. Evaporate to 20 ml without boiling, using watch glasses and glass hooks, filter through Whatman's No. 41 paper, 11 cm, to 50 ml in volumetric flask. 11. Titrate each of two 20 ml aliquots as follows. Add 1 drop methyl red indicator. Adjust with dilute NaOH solution to give pH 10.5 on pH 10-12 paper. Add 20 ml pH 10.5 buffer. Add 10 drops.,01M magnesium versenate solution, 1 ml 2% NaCN solution, 6 drops eriochrcne black T indicator. Titrate to disappearance of last trace of red. B. Determination of Calcium in Alloyed Samples The basic procedure of the previous section can be readily adapted to samples which contain aluminum, carbon, gold, nickel and silicon. 1. Iron Samples Containing Aluminum. While aluminum is ordinarily rather difficult to remove from solution, it can be removed quantitatively from the present solutions with cupferron, according to the method of Welcher, when the solution has a pH between 2.5 and 4.5. (1) The precipitation is not complete at lower pH

-81and at a higher pH the gelatinous aluminum hydroxide precipitate forms. Aluminum is removed following the standard cupferron extraction. The aluminum precipitate is soluble in diethyl ether, permitting it to be extracted with this phase. Procedure for samples containing aluminum: 1. Steps 1 through 9 of standard procedure, adding 2, 10 and 20 at'% aluminum as metallic aluminuml to the standards of Step 2. 2. Adjust solution to pH 3 by adding concentrated NaOH dropwise. Transfer into 600 ml separatory funnel, add 200 ml diethyl ether, add 5% cupferron batchwise until white precipitate stops forming. Rinse with 5 ml H20. Readjust pH to 3, add 100 ml diethyl ether. Add 5% cupferron batchwise until white precipitate stops forming. Rinse with 5 ml H20, re-extract with 100 ml diethyl ether and 1 ml 5% cupferron. 3. Steps 10 and 11 of standard procedure. 2. Iron Samples Containing Carbon. Carbon causes little change in the standard procedure. Any carbon residue is removed by filtration prior to the isopropyl ether extraction. Procedure for samples containing carbon: 1. Steps 1 through 6 of standard procedure, adding 2 and 10 at % carbon as carbon-saturated iron to the standards of Step 2. 2. Filter through Whatman's No. 41 11-cm paper into separatory funnel, add 3 drops 30% H202, extract according to Step 7 of standard procedure.

-823. Steps 8 through 11 of standard procedure. 3. Iron Samples Containing Gold The presence of gold in iron causes no difficulty in the procedure for determining calcium. Gold is extracted by isopropyl ether simultaneously with the iron, both as trivalent chlorides.(33) Gold is subsequently precipitated with ferrous sulfate from aquaeous solution and determined gravimetrically.(33) Procedure for samples containing gold: 1. Steps 1 through 6 of standard procedure, adding 2 and 10 at % gold to the standards of Step 2 as chlorauric acid, HAuC14'3H20. 2. Step 7 of the standard procedure, saving the isopropyl ether for the subsequent determination of gold. 3. Steps 8 through 11 of standard procedure. 4. Iron Samples Containing Nickel Nickel is not removed by an isopropyl ether extraction or by cupferron, therefore requiring a special step in the procedure. The nickel is electrolytically deposited from an ammoniacal solution following the cupferron (34) extraction, according to the method described by Kolthoff and Sandell.( 4) This method involves the addition of a small amount of concentrated sulphuric acid to accelerate the deposition of nickel by increasing the conductivity of the solution. (The increase in weight of the platinum gauze electrode serves to determine the nickel in the sample.) Procedure for samples containing nickel: 1. Steps 1 through 9 of the standard procedure, adding 2, 10 and 20 at% nickel as a high-purity nickel chloride solution in Step 2. 2. Evaporate to 20 ml without boiling, using watch glasses with glass hooks.

-833. Transfer to a 250 ml electrolyzing beaker. Add 3 ml H2S04, 20 ml NH40H, dilute to 100 ml with H20. Electrolyze at 6 volts for 40 minutes, using a platinum gauze cathode and platinum wire anode. 4. Steps 10 and 11 of standard procedure adjusting pH to 10.5 with the aid of a pH meter before adding the buffer in Step 11. 5. Iron Samples Containing Silicon The presence of silicon requires very special handling of solutions. This is because the bulk of the silicon must be removed by a precipitation, requiring careful technique to minimize the loss of the calcium species with the precipitate, and because sufficient hydrated silica remains in solution even after four precipitations to obscure the endpoint during titration. Silica is removed by separate filtrations before and after the isopropyl ether extraction. Solutions are dehydrated without excessive baking, and the precipitate rinsed with hot 2/1 HC1 during filtering, all according to the best techniques described by Kolthoff and Sandell, in order to minimize any calcium loss. The calcium is freed from the hydrated silica remaining in solution by collecting the calcium on a cation exchange resin, while the silica passes on through and is discarded. The calcium is subsequently removed from the column and finally analyzed. Calcium is deposited on the column at 1% HC1 acidity, and is removed from the column with 2/1 HC1 and concentrated HC1, following the method of Campbell and Kenner.(35) The procedure employed for the ion exchange technique is the result of studies of column heights, flow rates, acidity levels and rinsing procedures, with calcium concentrations

-84being measured at all stages of the study with a Beckman flame spectrophotometer. Procedure for samples containing silicon: 1. Step 1 of standard procedure. 2. Prepare 0.0667% Ca standards by adding 10 ml of 100 ppm Ca standard and 2.25 ml of 2000 ppm Ti solution and 2, 10 and 20 at % silicon as a vacuum-melted iron alloy containing 12 wt% silicon. 3. Dissolve samples in a total of 65 ml HC1, 40 ml HN03, 20 ml H20, with watch glasses, without boiling. (Approximately 30 hours required for dissolution.) 4. Step 4 of standard procedure. 5. Cool and drench with 5 ml 2/1 HC1 for 3 minutes. Dilute to 30 ml with 1% HC1. Simmer for 5 minutes. Cool, filter through Whatman's No. 41 11-cm paper with 10 ml hot 2/1 HC1, followed by 1% HC1. 6. Evaporate solution to 10 ml without boiling, using watch glasses and glass hooks, cool and dilute to 30 ml with 2/1 HC1. 7. Steps 6 and 7 of standard procedure. 8. Evaporate without boiling or spattering just to dryness, using watch glasses and glass hooks. Hold for 30 minutes in an oven at 110 ~C. 9. Repeat Step 5. 10. Steps 8 and 9 of standard procedure substituting H20 for 1% HC1. 11. Heat 10 minutes, with watch glasses, to expel ether.

-8512. Prepare resin column by filling with Dobx 50 resin, previously washed thoroughly in 2/1 HC1, into a 50 ml burette to a height of 3.5 cm. Rinse column with 150 ml HC1 at 8 + 2 ml per minute. 13. Dilute the sample solution to 75 ml with 1% HC1, pass through column and rinse with three 10 mi-portions of 1% HC1 at 3 + 1 ml per minute. Collect this solution and test for calcium in a flame photomoter. (All solutions showed a negligible amount of calcium here.) 14. Rinse the column with 300 ml 1% HC1 at 8 + 2 ml per minute, and test this solution for calcium in a flame photometer. (All solutions showed a negligible amount of calcium here.) 15. Collect the calcium from the column with 75 ml 2/1 HC1 and 50 ml Conc. HC1 at 3 + 1 ml per minute. Prepare column for subsequent samples by rinsing with 200 ml H20 at 8 + 2 ml per minute. 16. Steps 10 and 11 of standard procedure.

BIBLIOGRAPHY 1. Hume - Rothery, W., The Structure of Metals and Alloys, 2nd Edition, Institute of Metals, London, (1950), 57. 2. deAraujo, L. A., Colorado School of Mines Quarterly, 48, (January 1953), 47. 3. Quasebart, C., Metallurgie, 3, (1906), 28. 4. Watts, 0. P., "Iron and Calcium," J. Amer. Chem. Soc., 28, (1906), 1152. 5. Hirsch, A., and Aston, J., "The Alloying of Calcium and Iron," Trans. Amer. Electrochem. Soc., 13, (1908), 143. 6. Hansen, M., Constitution of Binary Alloys, 2nd Edition, McGraw-Hill Book Co., New York, (1958). 7. Wever, F., "Ueber den Einfluss der Elemente auf den Polymorphismus des Eisens," Archiv fuer das Eisenhuettenwesen, 2, (1929), 739. 8. Kubaschewski, 0., and Evans, E. L., Metallurgical Thermochemistry, J. Wiley and Sons, New York, (1956). 9. Hildebrand, J. H., and Scott, R. L., Solubility of Non-Electrolytes, 3rd Edition, Reinhold Publ. Corp., New York, (1950). 10. Mott, B. W., "Liquid Immiscibility in Metal Systems," Philosophical Magazine, 2, (1957),259. 11. Philbrook, W. 0., Goldman, K. M., and Helzel, M. M., "The Use of Radiocalcium to Study the Distribution of Calcium Between Molten Slags and Iron Saturated with Carbon," Trans. AIME, 88, (1950), 361. 12. Trojan, P. K., and Flinn, R. A., "A New Method for Determination of Liquid Equilibria as Applied to the Fe-C-Si-Mg System," Trans. A.S.M., 54, No. 3, (September 1961), 549. 13. Smithells, C. J., Metals Reference Book, Butterworths Scientific Publ., London, (1955), 179. —-... 14. Alcock, C. B., and Richardson, F. D., "Dilute Solutions in Molten Metals and Alloys," Acta Metallurgica, 6, (1958), 385. 15. Wagner, C., "Thermodynamic Investigations of Ternary Amalgams," J. Chemical Physics, 19, (1951), 626. 16. Case, L. 0., Elements of the Phase Rule, The Edwards Letter Shop, Ann Arbor, Michigan, (1959). -86

-8717. Smith, D. P., "Fundamental Metallurgical and Thermodynamic Principles of Gas-Metal Behaviour," Gases in Metals, A.S.M., (1953), 16 18, Klotz, I. M., Chemical Thermodynamics, Prentice-Hall, New York, (1950), 114. 19. Campbell, I. E., Editor, High Temperature Technology, J. Wiley and Sons, New York, (1956), 182. 20. Frazier, L., Private Communication. 21. Elliott, J. F., and Gleiser, M., Thermochemistry for Steelmaking, 1, Addison-Wesley Publ. Co., Reading, Massachusetts, (1960), 134-138. 22. Rist, A., and Chipman, J., "L'activite du carbone dissous dans le fer liquide," Revue de Metallurgie, 53, (1956), 796. 23. Ohtani, M., and Gocken, N.A., "Thermodynamic Interaction Parameters of Elements in Liquid Iron," Trans. A.I.M.E., 218, No. 3, (June 1960), 533. 24. Philbrook, W. 0., and Bever, M. B., Editors, "Basic Open Hearth Steelmaking," 2nd Edition, A, I.M.E., (1951), 638. 25. Pauling, L., The Nature of the Chemical Bond, Cornell University Press, Ithaca, New York, (1959), 58. 26. Gordy, W., and Thomas, W. J. 0., "Electronegativities of the Elements", J. Chemical Physics, 24, (1956), 439. 27. Lange, N. A., Editor, Handbook of Chemistry, McGraw-Hill Book Co., New York, (1961), 111. 28. Hodgman, C. D., Editor, Handbook of Chemistry and Physics, Chemical Rubber Publ. Co., Cleveland, Ohio, (1960). 29. Lumsden, J., Thermodynamics of Alloys, Institute of Metals, London, (1952), 340o. 30. Morrison, G. H., and Freiser, H., Solvent Extraction in Analytical Chemistry, John Wiley and Sons, New York, (1957), 212. 31. Welcher, F. J., Organic Analytical Reagents, 3, Van Nostrand Co., New York, (1947), 371-394. 32. Schwarzenbach, G., Complexometric Titrations, Interscience Publishers, New York, (1957), 63.

-8833. Willard, H. H., and Diehl, H., Advanced Quantitative Analysis, Van Nostrand Co., New York,(1943). 34. Kolthoff, I. M., and Sandell, E. B., Textbook of Quantitative Inorganic Analysis, 3rd. Edition, MacMillan Co., (1952). 35. Campbell, D. N., and Kenner, C. T., "Separation of Magnesium from Calcium by Ion Exchange Chromatography" Anal. Chemistry, 26, No. 3, (March 1954), 560.