THE UNIVERSITY OF MICHIGAN INDUSTRY PROGRAM OF THE COLLEGE OF ENGINEERING REDUCTION OF ALUMINUM OXIDE TO ALUMINUM IN RADIO FREQUENCY GENERATED PLASMAS Roger K. Rains A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy in The University of Michigan April, 1968 IP - 816

ACKNOWLEDGMENTS The author wishes to express his gratitude to all those persons who have aided in this investigation. The interest, guidance and criticism offered by Professor Robert H. Kadlec, chairman of the doctoral committee, were especially appreciatedo The interest and assistance of the other members of the doctoral committee: Professors E. Eo Hucke (co-chairman), S. W. Churchill, Ao A. Gordus, 0. F. Kimball and R. D. Pehlke are also acknowledged. The author is appreciative of the cooperation and help given by the secretaries and shop personnel of the Department of Chemical and Metallurgical Engineering. He is also grateful for the assistance given by his friends in the plasma laboratory. The author is indebted to his wife and parents for their encouragement and help and to the Bethlehem Steel Corporation for its financial support of this project. ii

TABLE OF CONTENTS Page ACKNOWLEDGMENTS.............................................. ii LIST OF TABLES..o............................................. vi LIST OF FIGURES........................ vii ABSTRACT....................................... ix 1o INTRODUCTION........................................... 1 2, SURVEY OF PREVIOUS WORK......................a........... 2 21 Reduction of Metallic Oxides in a Plasma............... 2 2,2 Other Plasma Reductions of Interest.................... 3 2.3 Suboxide Species of Aluminum.......................... 4 204 Heating of Solids in a Plasma......................... 6 30 DISCUSSION OF PREVIOUS WORK............................... g 3.1 Plasma Reduction Reactions............................ 9 3.2 Aluminum Suboxides............................... 10 303 Heating of Solids in a Plasma.......................... 11 4. SCOPE OF THE PRESENT STUDY................................ 13 50 THEORETICAL QONSIDERATIONS......................... 15 501 Thermodynamics 0..0........................... 15 5.2 An Interpretation of the Quench Process................ 17 5.3 Heat' Mass and'Momentum Balances....................... 19 60 SPECTROGRAPHIC TEMPERATURE MEASUREMENTS..................... 23 6.1 Meaning of Temperature and Local Thermal Equilibrium. 23 6,2 Relationship Between Temperature and Atomic Line Intensity................................ 24 6,3 Determination of Argon Temperature................... 26 604 Determination of Aluminum Temperature............. 27 70 ANALYTICAL METHODS.......................................... 31 7.1 Bromate Method for Aluminum Determination............ 31 7.2 X-ray Diffraction...................................... 32 703 X-ray Fluorescence.....................................3 7~4 Optical Spectroscopy................................ 34 iii

TABLE OF CONTENTS (Continued) Page 8o EXPERIMENTAL APPARATUS.................................. 35 801 General................................................. 35 802 Plasma Reactor........................................ 35 8~3 Power Supply and Auxilliary Equipment.................. 38 804 Spectrographic Equipment o....................0 0 39 8.5 Quench Probes.......................................... 40 8.6 X-ray Sample Probe...................................... 41 8.7 Plate Development and Analysis Equipment................ 41 8.8 X-ray and Wet Chemical Analysis Equipment.............. 44 9. EXPERIMENTAL PROCEDURES.................................... 45 9.1 Start-up and Safety Procedures.......................... 45 9.2 Selection of Reactor Design............................. 49 9.3 Qualitative Analysis of Alumina......................... 50 904 Analysis of Reaction Zones.............................. 51 9.5 Identification of Product............................... 51 9.6 Determination of Aluminum............................... 52 9.7 Measurement of Argon Temperature....................... 54 908 Measurement of Aluminum Temperature..................... 55 9.9 Study of Reaction Variables............................. 55 9o10 Quench Methods.......................................... 56 10o EXPERIMENTAL RESULTS AND ANALYSIS........................... 58 10.1ol General.....o............0... o... 58 10.2 Qualitative Analysis of Alumina...................... 58 10.3 Analysis of Reaction Zones.................... 60 10.4 Identification of Product........................ 64 10o5 Accuracy of Aluminum Determinations................... 66 10.6 Effect of Alumina Flow Rate and Particle Size. 67 10.7 Effect of Power Input................................. 72 10.8 Effect of Reducing Gases in the Plasma....... 72 10.9 Quench Methods oo................... 75 10o10 Argon and Aluminum Temperatures o.. 79 10.11 Solution of Heat, Mass and Momentum Balances.......... 80 11. SUMMARY AND CONCLUSIONS..o................. 89 APPENDIX I.o DERIVATION OF MOLAR FLUX OF Al IN QUENCH ZONE.... 91 APPENDIX IIo DERIVATION OF HEAT, MASS AND MOMENTUM BALANCE EQUATIONS FOR AN ALUMINA PARTICLE.........O..OO. 93 iv

TABLE OF CONTENTS (Continued) Page APPENDIX IIIo CALIBRATION OF SA1 PLATES o...................... 97 IIIO1 Background...................................... 97 III.2 Calibration Procedure......................... 97 IIIo3 Determination of Absolute Line Intensity........ 99 APPENDIX IVo DETERMINATION OF TRANSMITTANCES OF SOLIDS AND REACTOR.................................103 APPENDIX V. CALIBRATION OF DIRECT READ-OUT SYSTEM................ 105 APPENDIX VI. CALCULATION OF THE ABSOLUTE INTENSITY OF A SPECTRAL LINE.................................106 VIo1 General...........................................106 VIo2 Integration of a Spectral Line on a Photographic Plate...............................106 VI.3 Intergration of Photomultiplier Tube Output...... 108 APPENDIX VIIo SOLUTIONS USED IN VOLUMETRIC DETERMINATIONS OF AL MINUM....................................... 110 VII l General.......... o....o................. 110 VII 2 Formation and Dissolution of Aluminum Quinolate....................... 0... 110 VII.3 Determination of 8-Quinolinol.................. 110 NOMENCLATURE..................................................... 112 BIBLIOGRAPHY... o o. o. o.....o................. o......... 116

LIST OF TABLES Table Page I Upper Energy Levels and Transition Probabilities for A11 Lines........................................ 30 II Analysis of Reactant Aluminum Oxide.................... 58 III Comparison of Composite Analysis with Carborundum's Probable Analysis........................ 59 IV Effect of Solids Build-up.............................. 68 V Effect of H2, CO and CH4............................ 74 VI Conversions Obtained with Quench Probes................ 76 VII Effect of Hydrogen-quench Flow Rate.......... 77 VIII Effect of CO and CH4 as Quench Gases............... 77 IX Argon Temperature and Enthalpy........................0 79 X Aluminum Temperatures.................................. 81 XI Percents Conversion and Vaporization for Various Particle Sizes......................................... 85 XII Percents Conversion and Vaporization at Various Argon Temperatures..................................... 85 XIII Residence Times for Alumina Particles.................. 86 XIV Calculated Heat Transfer Coefficients for Small Alumina Particles................................ 87 vi

LIST OF FIGURES Figure Page 1 Free Energy of Formation per Atom of Oxygen Versus Temperature for Various Compounds.............. 16 2 Argon Temperature from Volume Emission Coefficient..................................... 28 3 Schematic Diagram of Experimental System............... 36 4 Induction-Coupled Plasma Reactor....................... 37 5 Quench Probe........................................ 42 6 X-ray Sample Probe..................................... 43 7 Position of Quench Probe in Reactor..47 8 Powder Pick-up Arrangement in Bottom of Hopper........ 48 9 Composition of the Sections of the Plasma.............. 61 10 Typical 3082.15 and 3092.71 f A1I lines in the Core of an Argon Plasma................................ 62 11 The 4842 a A10(0,0) Band of the 4850 R., 2:_2I Band System of A10 Observed in the Tail Flame of an Argon Plasma........................................... 63 12 X-ray Diffraction Lines of Al and o(-A1203 Obtained from the Product of the Reduction of o(-A1203 in an Argon Plasma........................... 65 13 Variation of the Conversion of A1203 to Al in an Argon Plasma with the A1203 Flow Rate and Particle Size........................................... 9 14 Effect of Power Input upon the Conversion of A1203 to Al in an Argon Plasma e..A................. 73 15 Comparison of Conversions Obtained with and without the Use of Hydrogen as a Quench Gas........ 78 16 Determination of Aluminum Temperature by the "Multi-Line" Method.................. 82 17 Comparison of the Percents of Conversion and Vaporization Obtained with 26/(, 37/I, and 45u8 A1203 Particles.................. 84 vii

LIST OF FIGURES (Continued) Figure Page 18 Emulsion Calibration Curves for SA1 Plates.......... 100 19 Completion of the Missing Portion of the Intensity Profile of a 3944.03 R A1I Line by a Least Squares Fit to a Lorentzian Profile............................. o.............. 109 V 1 i

ABSTRACT The reduction of aluminum oxide to aluminum in radio frequency generated, induction-coupled plasmas was experimentally investigated. The objective was to demonstrate that appreciable conversions of stable metallic oxides, such as alumina, could be obtained in an induction plasma reactor. The reduction was effected by feeding alumina into Ar, Ar-H2, Ar-CO and Ar-CH4 plasmas. The plasmas were contained in a water-cooled quartz reactor and flowed downward. The products were collected on the reactor wall and on water-cooled quench probes placed directly in the plasma. The conversion of A1203 to Al was based on the amount of aluminum and aluminum oxide actually collected since complete recovery of the products was not attempted. In argon plasmas, the conversion was investigated as a function of A1203 particle size, A1203 mass flow rate and power input to the plasma. The effect of quenching the reaction mixture in a reducing atmosphere was determined by introducing H2, CO and CH4 into the lower section of the plasma countercurrent to the plasma flowo The products were positively identified by x-ray diffraction and wet chemical techniques. In addition, the gaseous plasma reaction zones were examined spectroscopically to give some insight as to the nature of the gas-solid reaction. Finally, mean aluminum and argon temperatures in the plasma were determined. In an atmospheric argon plasma flowing at 52 gm/min, the alumina conversion was enhanced by increasing the percent of vaporization of ix

the oxide. This was achieved by increasing the power input, and decreasing the alumina flow rate and particle size. Conversions of 3-30 percent were obtained with power levels of 5.03, 5.86 and 6.69 kw; oxide flow rates of 0~03 to 0.6 gm/min; and particle diameters of 26y, 371k and 45[i. The argon temperatures measured at these power levels are 10,9000, 11,1000 and 11,2000K respectively. The conversions qualitatively agree with the extents of vaporization predicted by computer solutions of heat, mass and momentum balances for alumina particles. Spectroscopic identification of Al and A10 in the cooler plasma regions and only Al in the hot plasma core confirmed that vaporization of A1203 to Al in an argon plasma is a two step process: the oxide dissociates to A10 and 0 and the AlO in turn dissociates to Al and 0 o With CO and CH4 in the plasma, the respective conversions were 46 and 42 percent. The corresponding conversions for pure argon plasmas at similar operating conditions were 23 and 9 percent. The use of CO and CH4 as quench gases doubled the conversions that were obtained without them. Hydrogen, in each application, had little effect. Higher conversions were obtained with the product collected on the quench probes than with that collected on the reactor wall. Examination of the reduction product from Ar and Ar-CO plasmas identified Al, o-A1203 and t-A1203. The product from-an Ar-CH4 plasma also contained A14C3. No solid aluminum suboxides or aluminum oxycarbides were found. The mean aluminum temperatures obtained for various experimental conditions ranged from 2000 to 6000~K. These results were

somewhat low and inconsistent, but it was possible to ascertain that the Al temperature increased with increasing power input and decreasing A1203 flow rate. The reduction of stable metallic oxides, such as A1203, in induction-coupled plasmas is certainly feasible. By using CO or CH4; increasing the power input; and decreasing the oxide flow rate and particle size, conversions of 50 percent or more should be realizable, depending upon the efficiency of the recovery of the product. xi

1. INTRODUCTION The area of plasma chemistry can best be described by Searcy's Laws:(67) 1. At high temperatures, everything reacts with everything. 2. The higher the temperature, the more rapidly everything reacts. 3. The products may be anything. It was with this concept in mind that this thesis was undertaken. When it was decided to investigate the plasma reduction of a metallic oxide, aluminum oxide was chosen for four reasons. First, A1203 is extremely difficult to reduce and is on a par with CeO2, Ce203, La203, Nd203, Pr203, Sc203, Sm203 and Y203 in this respect, (41) to name but a few. So if A1203 can be reduced, then a host of other oxides can also be reduced, yielding much more lucrative products. Second, since "the products may be anything," reduction of alumina itself may provide one or more products of considerable interest. Third, A1203 is inexpensive and much more readily available than most of the other difficult-to-reduce oxides. Fourth, it was felt that the two known attempts at reducing alumina had not been pursued sufficiently with the proper variables in mind. The conversion of A1203 to Al in this study was effected in radio frequency generated, induction-coupled Ar, Ar-H2, Ar-CO and Ar-CH4 plasmas. Different quench techniques and three reaction variables were investigated. The specific experimental program is outlined in Chapter 4. -1

2. SURVEY OF PREVIOUS WORK 2.1 Reduction of Metallic Oxides in a Plasma The literature on plasma reduction of metallic oxides is quite sparse, although it is known that several groups of researchers have investigated the area. The reason is that the majority of this work has been done by private industry. Grosse, et al.,(47'85) attempted to reduce 200 mesh (74yp) commercial C.P. grade alumina in an arc plasma jet. Using argon plasmas, they tried A1203 flow rates of 1.5-6.0 gm/min, power inputs of 8.55-9.6 kw and total gas flow rates of 20.4-27.5 liters/min. The oxide was fed into the plasma flame (below the arc) with H2 or CH4 as the carrier gaso Aluminum metal was definitely formed, but the yields were very poor, ranging from 0.2 to 1.25 percent. As an extension of the above work, Stokes, et al.,(86) studied the reductions of W03, Fe203, Ta205, A1203, TiO2 and ZrO2 in an arc plasma jet. Using He plasmas of 34 liters/min, they obtained maximum metal yields of 95 percent for W03, 100 percent for Fe203 and 25.6 percent for Ta205. The respective powder flow rates were 1.96, 0.3 and 1.9 gm/min at corresponding power levels of 15.2, 15.5 and 15o75 kw. The oxides were introduced into the plasma flame with H2 flowing at 13.6 liters/min. Commercial grade oxide powder of unreported size was used in each of these investigations and the products of the first two reductions were of submicron size and highly pyrophoric. Conversions of 2-5 percent were achieved with 325 mesh (44/s) A1203 in He plasmas operating at power levels of 1104-14.9 kwo The powder flow was varied -2

-3over the range 0.48 to 2.5 gm/min. At the lowest flow rate, the oxide was mixed with carbon and carried into the plasma flame in an argon stream. With oxide flows of 0.5 gm/min or greater, hydrogen was the carrier gas. No reductions were obtained with 325 mesh TiO2 at powder flow rates of 0.4-0.6 gm/min or with ZrO2 of unreported size flowing at 0.8 gm/min. Brown(12) was able to reduce ZrO2 to Zr in Ar-C and Ar-H2 arc plasmas. He found little advantage in the use of hydrogen and found no effect due to the method of carbon introduction. There was no reduction when either 60 mesh (250pu) or 100 mesh (150A) particles of zirconia were used, but with 10 zirconia, the zirconium content was increased from 66.5 percent for the injected material to 70 percent for the product. This amounts to a conversion of about 21 percent. 2.2 Other Plasma Reductions of Interest Stokes, et al.,(86) also investigated the carbothermic reduction of W03 to W2C, WC and W and the reduction of Ta205 to TaC. In each case the oxide was carried in a methane stream of 6.5 liters/min into the "flame" of a helium plasma jet flowing at 34 liters/ min. The W03 flow rates were in the range 0.51 to 4 gm/min while the operating power was varied from 10.0 to 16.3 kw. The percent conversions obtained were 9.2 - 34.6 for W2C, 4.1 - 11.0 for WC and 43.2 - 80.8 for W. A conversion to TaC of 9.7 percent was achieved for Ta205 flowing at 0.5 gm/min at an operating power of 15.75 kw. The decomposition of volatile metal halides has been studied (6) and by Brown.(12) The former found that the by Biggerstaff, et el., adb Bon

-4decompositions of boron trichloride and trifluoride in a plasma jet could be used to produce boron of very small particle size. The addition of hydrogen to the plasma did not significantly increase the conversion to the metal. Brown reduced zirconium tetrachloride in an argon plasma and tried both a hydrogen dilution quench and a straight thermal quench. With a reactant containing 38.2 percent Zr, he obtained a product with a 58.4 percent Zr content. As with Biggerstaff and his associates, he found little benefit in the use of hydrogen. The final plasma reduction of interest was performed by Huska and Clump, (58) who studied the decomposition of MoS2 in an induction-coupled argon plasma. Operating with MoS2 flow rates of o.666, 1.19 and 2.48 gm/hr and several plasma power levels between 1.9 and 5l15 kw, they obtained conversions ranging from 25.7 to 70.2 percent. The conversion increased with decreasing reactant feed rate and increased with increasing plasma energy content. 2.3 Suboxide Species of Aluminum The gaseous oxides of aluminum have been quite extensively (11) investigated. According to Brewer and Searcy, there are only two important gaseous oxides of aluminum. A120 is the primary gaseous suboxide when alumina is heated under reducing conditions and AlO is the principal gaseous product under neutral or oxidizing conditions. The bulk of the remaining work, which is discussed below, substantiates their findings. De Maria, et al., report that the main reaction upon heating A1203 from 2300 to 2600~K is the decomposition to AlO and 0.

-5Zhadanova and Sokolov(91) and Allen, et al.,(2) found only A10 and Al when aluminum burned in oxygen and air. Thus, only A10 was observed under the neutral and oxidizing conditions of the investigations. Porter, et al.,(75) Cochran(17) and Gastinger(36) each report formation of A120 when A1203 and Al were heated together. Grossman(48) heated Nb with A1203 and Grube, et al., heated Si with A12 03 to form A120. Worrell,(90) Motzfeldt(71) and Ponomarev, et al.,(74) obtained A120 by carbothermic reduction of alumina and Mal'tsev(66) formed A120 by carbothermic reduction of sodium and potassium hydroaluminates. Each of these are examples of reducing conditions and A120 is the product. Some disagreement is offered by Inghram, et al, (59) who found that A120 and Al were the major species obtained by heating A1203 under neutral conditions in a tungsten cell. Further contention is given by Hasapis, et al.,(51) who vaporized A1203 by itself in a tungsten cell and with tungsten powder. For each case, they found A10 and A120 in about equal concentrations. Ackermann and Thorn(l) also heated alumina in a tungsten cell and found a reaction between the two at high temperatures. However, they did not establish the nature of their products, which they thought were A120 and W03 The literature on the existence of solid suboxides of aluminum is quite controversial. By means of a high temperature x-ray technique, Hoch and Johnson(9) found solid A120 between 1100 and 15000C, solid A10 above 16000 and both from 1500 to 1600~Co They determined that both are cubic and that upon cooling or rapid quenching both compounds dissociate to Al and A1203 Cochran(l7) contends that the lowered

-6melting points of A1203 mixed with Al indicates subcompound formation in the condensed phase. His data were too scattered to allow a phase diagram to be constructed. Khodak and Mal'tsev(61) heated Na20oA1203-2SiO2-nH20 with carbon. Rapid chilling of the reaction mixture enabled them to confirm the formation of the solids A120 and SiO by isolation of A120 crystals. Beletskii and Rapoport(4) report the formation of coarse hexagonal crystals, believed to be A120, when they heated Al and A1203 in the presence of SiO2 and carbon. However, Brewer(10) believes that the substance may be a ternary Al-O-C compound. Brewer, in this same paper, reports that he, Searcy and McCullough were unable to find a new solid phase upon room temperature x-ray examination of fused mixtures of Al and A1203. Gitleson, et alo, could detect no indication of a solid suboxide either by microscopic examination or by x-ray powder patterns of solidified Al-A1203 melts. In studies of the carbothermic reduction of alumina, Emlin, et al, (32) Foster, et al.,(34) Ginsberg and Sparwald(39) and Gitleson, et al.,(40) could find no low melting substances except the oxycarbides A120C and A1404C 2~4 Heating of Solids in a Plasma Chludzinski determined heat transfer coefficients for axisymmetric stagnation point heat transfer at thermocouple tips in argon and argon-nitrogen plasmas. With a thermocouple 0.02 inches in diameter in an argon plasma at about 10,9000K, he obtained a maximum heat transfer coefficient of 87 BTU/hr-ft -~F. In a 90 percent Ar-10

-7percent N2 plasma at about 11,4000K, the maximum heat transfer coefficient was 162 BTU/hr-ft2-~F. He compared the convective and radiative contributions to the heat transfer from the plasma and found the former to predominate~ Heat flux measurements were made on small argon plasma jets by Stokes, et al., (87) who obtained heat fluxes as high as 4.5 kcal/cm2sec for the transient heating of a small copper slug. This corresponds to a heat transfer coefficient of about 4000 BTU/hr-ft2-~Fo Reed(77) made heat transfer intensity measurements in argon and argon with oxygen or helium induction plasmas. For the various plasmas, he obtained peak intensities of 56 to 145 watts/cm2 for heat transfer from the tail of the plasma to the surface of a water-cooled plate. The corresponding range of approximate heat transfer coefficients is 18 to 46 BTLU/hr-ft2- ~F (33) Engelke developed an idealized model for heat transfer to solid particles in an arc plasma stream. He compared his "integrated heat transfer to the particle" with observations on the melting of some carbides and found substantial agreement with his theory. In an argonhydrogen plasma at 10,3000K, he noted that 250 mesh (62,.) TiC did not begin to melt whereas the smaller 625 mesh (20o,) TiC had vaporized somewhat. In a lower energy argon plasma, even the 625 mesh TiC did not begin to melt. Titanium carbide melts at 34100K and boils at 45700K (56) Loo and Dimick, (65) Meyer(70) and Wood and Wise (89) investigated the catalytic activity of some metals and oxides on the recombination of atoms with atoms and ions with electrons. Loo and Dimick conducted experiments with an argon arc flame and ascertained that metals

-8tend to increase the Ar+-e- recombination rate while oxides tend to decrease the rate. Meyer found that metallic oxides are slightly active and metals are extremely active in catalyzing recombination processes in argon-nitrogen plasmas. Wood and Wise determined that metals catalyze the recombination of H, N and 0 atoms. High-speed cinematographic studies of metallic and metallic oxide powders (including A1203 ) in an induction-coupled argon plasma were carried out by Hedger and Hall.(52) They observed that vaporization occurred preferentially from the lower side of the descending particles and that the vaporization impeded heat transfer to the particles. It was also noted that the majority of the particles pass straight through the turbulent plasma, but a small portion is ejected from the plasma center.

3. DISCUSSION OF PREVIOUS WORK 3.1 Plasma Reduction Reactions The success obtained with W03, Fe203, and Ta205 demonstrates that the use of plasmas for reducing metallic oxides is feasible. However, these oxides normally are not difficult to reduce. The use of plasmas would be much more justifiable if it would result in significant conversions for the less reducible oxides, such as A1203, TiO2 and ZrO2. These three oxides, for which Stokes and his associates obtained negligible reduction, have generally been considered to be much too stable to allow significant successo In other words, the oxides would reform during the quenching of the reaction mixture. So unless an extremely rapid quench could be attained, most of the metal would be lost due to this recombination. The above argument is certainly meritorious, but one of equal importance is the ease of vaporization of the oxide. If the oxide is not vaporized and decomposed to its elements, then there is no need to worry about recombination during the quench. It is significant that W03, Fe203 and Ta205 are all more easily vaporized than A1203, TiO2 or ZrO2. (41) Brown's work, which occurred at about the same time as the experimental portion of this thesis, demonstrates the effect of increasing the extent of vaporization of the oxide. He found no reduction of the larger 250 and 100/ ZrO2 particles, but got a 21 percent conversion with 10lO zirconia, which was more vaporized. Thus it is reasonable to expect that A1203 and TiO2 can also be significantly reduced if small enough particles are used. -9

-10Two other variables, vizo powder flow rate and power input to the plasma, were shown to have an effect upon the reduction of metallic compounds by the work of Huska and Clump. It is also certain that the use of a reducing gas in the plasma should improve the conversion. 3.2 Aluminum Suboxides The majority of the studies of gaseous aluminum suboxides agree that A120 predominates in a reducing atmosphere and A10 in a neutral or oxidizing environment. It is further agreed that these are the only two important gaseous aluminum oxides. A closer examination of the dissenting reports will show them to be in accord with the above conclusions. Brewer and Searcy heated alumina in a tungsten cell for their neutral conditions and found only AlO. Inghram, Hasapis, et al., and Ackermann and Thorn performed similar experiments and found A120. But Ackermann and Thorn also observed a reaction between the tungsten cell and the alumina. This gave them reducing instead of neutral conditions and explains their A120 formationo It appears that Inghram and Hasapis, et al,, also had this tungsten-alumina reaction while Brewer and Searcy, due to somewhat different conditions, had no interaction. The controversy over the existence of solid aluminum suboxides can also be resolved to some extent. The two claims that stable solid suboxides were prepared at room temperature involved the reduction of A1203 by carbon. In the other four investigations of carbothermic reduction of alumina, the oxycarbides A120C and A1404C were the products. Furthermore, when systems containing A1 and A1203 without carbon were

-11examined at room temperature, no low melting compounds were observedo This suggests that the reported solid aluminum suboxides were actually oxycarbides. The formation of solid suboxides at higher temperatures is very probable. Hoch and Johnson agree with Cochran that subcompounds exist in Al-A1203 systems above 1100~C and no one has disputed their claims. Nevertheless, some more high temperature studies should be made before these results can be completely accepted. 3.3 Heating of Solids in a Plasma The heat transfer studies of Chludzinski, Stokes, et al., and Reed produced heat transfer coefficients ranging from 18 to 4000 BTU/hrft2-~FoF. Such a wide range is reasonable since three different systems were used. The coefficients estimated from Reed's results (18-46 BTU/ hr-ft2- F) should be at the low end because he investigated the tail of an induction flame, which is the coolest part. Chludzinski's results (87-162 BTU/hr-ft2-~F) should be higher than Reed's because he used thermocouples which were placed in the hot core of the plasma. The small copper slugs studied by Stokes and his co-workers were also in contact with the hotter part of the plasma. The coefficient corresponding to their maximum heat flux (4000 BTTJ/hr-ft2-~F) is reasonable for very small slugs since the coefficient should increase with a decrease in diameter as Chludzinski found for his thermocouples. Engelke's observations of the melting and vaporization of TiC particles also indicate an increase in heat transfer with a decrease in particle diameter. However, his claim that 625 mesh TiC did not

-12begin to melt in his argon plasma is dubious. If Chludzinski's results for argon plasmas are applied to Engelke's model, at least partial melting is predicted. Such partial melting would be difficult to detect by examination of particles that had passed through the plasma. It is concluded from the catalysis studies that metallic oxides decrease the Ar+-e- recombination rate and slightly increase the rate of atom-atom recombinations. Metals have the effect of increasing the rates of both types of recombination. This means that the heat transfer to oxides and metals, and hence their rate of vaporization, is affected in the same way. Two other phenomena which influence the vaporization of solids in a plasma are: its preferential occurrence at the lower side of the particle and the formation of a vapor shield around the particle. Both diminish the rate of vaporization.

4. SCOPE OF THE PRESENT STUDY The primary aim of this research was to investigate the reduction of A1203 to Al in a radio frequency generated, inductioncoupled argon plasma. It was also intended to demonstrate that an induction-coupled plasma reactor is especially suitable for such gassolid reactions. To accomplish these goals, the following program was completed~ lo Qualitative analysis of the reaction zones with an optical spectroscope. 20 Positive identification of the products. 3o Investigate various quench methods. 4. Determine mean Al and Ar temperatures in the center of the plasma. D5 Study the effects of particle size, A1203 mass flow rate and power input to the plasma. 6, Use H2, CO and CH4 as reducing agents in argon plasmas. 7. Solve heat, mass and momentum balances for an A1203 particle. The above program was designed to provide fundamental information about plasma reduction of alumina and to define the operating parameters. The purpose of the qualitative spectroscopic analysis was to acquire a better understanding of the vaporization of A1203 and quenching of the gaseous mixture. According to the literature, A1203 should decompose to AlO and this spectroscopic analysis would -13

-14be a check. The positive identification of the reduction products was necessary because the previous work on solid suboxides of aluminum suggested that it might be possible to form these compounds by a rapid quench of the gaseous suboxides. The quench methods used were: straight thermal quench on the water-cooled reactor wall; thermal quench on a water-cooled probe placed inside the plasma; counter-current introduction of H2, CO and CH4 into the bottom of the plasma through the quench probe. A basic interpretation of the quench process was obtained through an understanding of why the conversion of A1203 varied as it did with changes in the Al temperature and concentration in the plasma. The argon temperatures were used to evaluate the thermal conductivity, viscosity, density and heat capacity for the argon plasma and for the boundary layer surrounding a solid particle. These properties were used in a correlation for the heat transfer coefficient which was needed in order to solve the heat, mass and momentum balance equations. Solution of these equations allowed the time-temperature history of the particle and its extent of vaporization with axial position to be estimated. Solids flow rate and power input to the plasma were shown to be significant variables in Section 3.1. It was also thought that particle size would be an important variable (this was later substantiated, as discussed in Section 3.1) and so the effect of these three variables on. the conversion of A1203 was investigated. The reducing gases (H2, CO and CH4) were used in independent studies aimed at increasing the conversions above those attainable in the argon plasma.

5. THEORETICAL CONSIDERATIONS 5.1 Thermodynamics It was established in Sections 2.3 and 3.2 that A10 and A120 are the only two gaseous oxides of aluminum. Furthermore, when A1203 is heated, it decomposes to A10 and 0 under neutral or oxidizing conditions and to A120 and 0 in a reducing atmosphere. So A1203 should vaporize to A10 and 0 in an argon plasma and to A120 and 0 in an argon plasma containing H2, CO, or CH4 According to Brewer and Searcy,(11) the dissociation: A1203 - 2Al0 + 0 (5l1) occurs at 3800 + 2000K. This is in fair agreement with the findings of Coheur() and Coheur and Coheur, 9) who observed that the optimum temperature for formation of A10 was 4000~C. They found that the molecule is unstable above this temperature and below it the probability of formation is slight. However, Figure 1, which was prepared from information presented in the JANAF tables,(27) shows that A10 should be stable down to about 1100'K once it is formed and is unstable above 4400~Ko Since temperatures in excess of this are to be expected in an argon plasma, the A10 formed from the vaporization of A1203 should further dissociate~ A10 - Al + O (5.2) shortly after its formation. -15

FREE ENERGY OF FORMATION PER ATOM OF OXYGEN, KCAL/GMMOLE 0% N 0 00 0% to N N0 O O O O O O O O'OO O O O v~~~~~~~~~~~~~~~~~~~, C) OF~~~~~~~~~ m Al o 0 m.p0 III I I / \I I I I ~l/ r~~~~~~~~~~~~~~~~~~~~~~~~~~~ V c ~~~~~~~~~~~~~~~o H-j 0 - P- 0D +-' O~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ~ 0 g Al —'- - __- — __ O' co /-4 C I-'D () _ 0 0 0 ('0 0',./ 0,-'-A'1203 m.p. O O —' m~ - ~j Ho w +'-v Al b~p_ ~ h,~~~_ 0o n _ N, 0 %n o o - A120 3 bp.___ r~ 0 O 0 CD 01Q C) Y;0 (D2 crq o CD m~ oD 0 - 0)20 %.n~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ o C) 0i - 0),,,.. O%

-17Once A120 is formed it is stable below 43000K and unstable above it, as Figure 1 shows. So if A120 is formed in Ar-H2, Ar-CO and Ar-CH4 plasmas by: A1203 - A120 + o, (5~3) it should also undergo a decomposition: A120 4 2A1 + 0. (5.4) In either case, therefore, the plasma should contain Al and 0 atoms instead of A10 or A120 molecules. The use of reducing gases in the plasma is expected to aid in the quench process. This is done by removing oxygen during the quench; thus diminishing the chance that the gaseous Al will oxidize to A10 or A120. Figure 1 shows that carbon is much better in this respect than hydrogen as CO is considerably more stable than H20 over a wide temperature range. Oxygen also prefers carbon to aluminum, as seen by a comparison of the free energies of formation of CO, A10 and A1 20 Hydrogen, on the other hand, has less affinity for oxygen than does aluminum over fairly large ranges in temperature. Consequently, hydrogen should be of little benefit in maintaining Al during the quench and carbon should be effective. Although A1203 is more stable than even CO at low temperatures, quenching occurs rapidly enough that the formation of alumina should be slight. 5.2 An Interpretation of the Quench Process The ultrahigh temperatures of a vortex stabilized, inductioncoupled argon plasma extend in the radial direction very nearly to the

-18wall of the cylindrical confining chamber. The boundary layer between the reactor wall and the point at which the plasma temperature noticeably begins to decrease is the quench region. The concentration and temperature gradients in this quench boundary layer are extremely steep and the condensation of a species out of the plasma is a rate process controlled by these gradients. Let us consider the case of condensation of gaseous Al from an argon plasma containing the gaseois species: Al, O, Ar and various ions of these atoms. During the quench, the following reactions should occur: Al + 0 A10 (5o5) 0 + 0 0 2 (5.6) Al + 1 02 _ AlO 02 +~ (5.7) Al(g) -- Al(l) (5-8) as well as the recombinations of ions with electrons. Since recovery of aluminum is the purpose of the quench, it is desirable to increase the rate of transport of Al through the boundary layer so that less of it can be oxidizedo The radial molar flux of Al in a mixture of Al and the noncondensable gases is: C DAlm bXAl NAlr = - r (5,9) where NAlr = radial molar flux of Al, XAl = mole fraction of Al DAlm = effective binary diffusivity for Al in the mixture,

-19C = total concentration. Since only Al condenses, the bulk flow due to the other gases was neglected in the derivation of Equation (5.9), which is found in Appendix I. It is seen from Equation (5.9) that the molar flux of Al depends on both the concentration and temperature of aluminum (DAlm is temperature dependent). If the bulk concentration of Al is increased, NAlr is initially greater. But since the quench occurs with extreme rapidity, the result is that more Al is recovered. A similar argument holds for the temperature effect. The diffusion of Al atoms through a mixture depends on the temperature of the atoms; DA1m increases with temperature. Therefore NAlr also increases with Al temperature and again, more aluminum should be recovered. It is concluded that one means of improving the recovery of Al from an Al -O-Ar plasma is by increasing the concentration and/or the temperature of the aluminum in the plasmao 5.3 Heat, Mass and Momentum Balances The history of a small A1203 particle in an argon plasma is estimated from a solution of steady state heat, mass and momentum balance equations for the particle. The description is only approximate because some simplifying assumptions are used in deriving the equations (see Appendix II). The use of more exact equations is not justified since a rough comparison with the experimental results is sufficient for this worko The momentum balance derived for a small particle injected into the plasma is:

-20d2z 3./o (dz 2 -=d Ps l~dt \ $)+ g (5ol) dt2 4 d s -IS dt where dt= velocity of solid particle, dt AS = density of solid particle, ds = diameter of solid particle, 1 = density of bulk plasma, v. = velocity of bulk plasma, g = gravitational acceleration, CD = drag coefficient. The direction of the drag force on the particle depends on the velocity of the particle relative to the plasma (dz/dt - Vo = vrel). If the relative velocity, Vrel, is positive, the minus sign is used in Equation (5.10) and if vrel is negative, the plus sign is used. It is assumed that the particle is spherical and remains so after melting commences. The plasma density is obtained from the ideal gas law (which is very nearly true at plasma temperatures) and a mean particle density is used. A constant average plasma velocity and Christiansen's drag coefficient data are also usedo Magnetic drag and slip flow effects are negligible. (80) The heat balance for a particle as it is being heated to its vaporization temperature is: 1 j4 d5 dT5 4 ds CPs dT = h(T-Ts) -O-e Ts where C = heat capacity of solid particle, Ps Ts = temperature of solid, T = temperature of plasma,

-21h = heat transfer coefficient, 0-= Stefan-Boltzmann constant, es = emissivity of solid particle. The assumption is made that vaporization does not occur until the particle reaches the temperature for dissociation to A10 and 0 As a result, no mass is lost during this time and a constant particle diameter can be usedo The heat capacity of the solid varies little (27) with temperature so an average value is assumed. A constant emissivity is estimated because sufficient data were not available. At all times, the temperature throughout the particle is assumed to be equal to its surface temperature. Vaporization of the solid particle to A10 and 0 is described by the combined heat and mass balance equation: 1 Ps (d h(s) -eh(T-T Ts (5.12) - r fs dt - s s where &Hr = heat of reaction for dissociation to Al and 0 (since the A10 initially formed quickly dissociates to Al and 0). Assumptions are that the temperature of the solid remains constant during the dissociation, which proceeds uniformly. The heat transfer coefficient used in Equations (5.11) and (5.12) is' 2kf 0.5 0.3 h = + O~6 Ref Prf (5o13) where kf = thermal conductivity of gas evaluated at an average temperature in the boundary layer around a particle, Ref = Reynolds number at average boundary layer temperature, Prf Prandtl number at same temperature.

-22The relative velocity of the particle is used in evaluating the Reynolds number. The two terms on the right hand side of Equation (5.13) represent the conductive and convective contributions respectively. Engelke(33) used purely conductive heat transfer in his model while Chludzinski ) observed that the heat transfer to his thermocouples was convective. But for a small particle in relative laminar flow in a plasma, both contributions are important. Thermal radiation effects from the plasma can be neglected. The average boundary layer temperature at which the gas properties are evaluated is that which corresponds to the reference enthalpy given by: href = 0.5(h hs) + hs (5.14) where hs = gas enthalpy at particle surface temperature, h, = gas enthalpy at bulk plasma temperature. This reference enthalpy method is suggested by Eckert(30) for the case of a large temperature difference across the boundary layero The argon specific heats and enthalpies used in Equations (5.13) and (28) (5.14) are given by Drellishak, et al., and the thermal conductivities and viscosities are obtained from Sherman and Grey. (81)

6o SPECTROGRAPHIC TEMPERATURE MEASUREMENTS 6.1 Meaning of Temperature and Local Thermal Equilibrium Consider an isolated gas assembly containing a large number of identical particles which possess kinetic energy. According to statistical mechanics, an equilibrium distribution of particle energies or velocities will be established. Temperature is inherently related to this distribution. With a Maxwell-Boltzmann distribution (or M.B. distribution), the temperature can be related to the most probable velocity of the particles. In particular, the most probable kinetic energy of a particle in the gas is equal to the product of the Boltzmann constant and the temperature. By this definition, if a temperature is assigned to a gas assembly, a M.B. distribution of energies exists. The converse is also true. An ionized gas consists of a finite number of different particle assemblies corresponding to the classes of particles presento Each of these assemblies could conceivably have a different initial energy distribution (and hence temperature). The final conditions reached in an isolated set of such assemblies depends upon the nature of collisions and energy exchange between particles. If energy is exchanged only between similar particles, the attainment of "equilibrium" would find each assembly with a different temperatureo The other possibility would have all particles exchanging energy. In this case there is a unique final distribution for all the particles and the system has a single temperature. In such a system, the particles are in thermal equilibrium. -23

These isolated systems are actually nonexistent and energy exchange with the surroundings must be considered. Normally this additional exchange results in a directional flow of energy. But if a section of the assembly contains enough particles for equilibrium to exist, and the energy being transferred is much smaller than the total energy in the section, a local M.B. distribution can be obtained. In other words, collisions between particles not directly associated with energy transfer to the surroundings will predominate and maintain the distribution in the section. If the total assembly consists of such sections having local M.B. distributions and temperatures, it is in local thermal equilibrium (LTE). In practice, the temperature variation is considered to be continuous with the temperature at a point defining the local distribution. According to Chludzinski,(15) LTE exists in a radio frequency generated, induction-coupled argon plasma. 6.2 Relationship between Temperature and Atomic Line Intensity In a plasma consisting of atoms, ions and electrons, all particles have translational energy, atoms and ions can possess electronic excitation energy and ions have ionization energy. Electronic excitation occurs in an atom when energy gained by collision causes an orbital electron of the atom to make a transition to an orbit of higher energy. If this electron in the higher energy state then makes a spontaneous transition to a lower energy level, a photon of light is emitted. The energy of this photon is equal to the difference in the energies of the upper and lower levels. The rate of these downward transitions depends upon a constant known as Einstein's probability of spontaneous transition.

-25The frequency of the emitted photons is: nm -= (En-Em)/h (6.1) and the radiation intensity is: Inm 4= - Lh M Anm n(62) where Anm = Einstein's spontaneous transition probability for transition from n to m, EnEm = energies of upper level n and lower level m respectively, with respect to ground level, h = Planck's constant, Inm. = frequency of emitted photon, Inm = radiant intensity (energy emitted per unit time per unit area per steradian), L = path length through source, Nn = number density of atoms with an electron in excited state n. According to statistical mechanics, for an equilibrium system with a M.B. distribution, the concentration of excited atoms in state n is: Ngn 13 Nn = N exp(-En/kT) (6.3) 0g Q where No = concentration of atoms in ground state, gnngo = statistical weights of atoms in state n and ground state, respectively,

Q = electronic partition function, k = Boltzmann constant, T = absolute temperature. Combining Equations (6.2) and (6.3) gives the relationship between temperature and atomic line intensity: Inm = 1_Lh No Vnm Anm gn 1 exp(-En/kT) (6.4) 41~ gO Q 6.3 Determination of Argon Temperature For argon atoms, the statistical weight of the ground electronic state is 1 and the electronic partition function is essentially 1o(82) Therefore, Equation (6.4) becomes: Inm 41 Lh No inm Anm gn exp(-En/kT) (6.5) When LTE exists, the volume emission coefficient, ~, for an optically thin spectral line of argon is: h )nm = rim No gn Anm exp(-En/kT) (6.6) It follows from Equation (6o5) and (6.6) that the absolute line intensity averaged along the path length through the source is. Inm =. L * (6.7) The argon temperature can be determined from Equations (6.6) and (6.7) if the absolute intensity of a spectral line is measured and the transition probability of the line is known. This is known as the "singleline" methods

-27Since NO is temperature dependent, a graphical procedure is employed to determine the temperature. Using Drellishak's(28) values for NO and Gericke's(37) transition probabilities, the volume emission coefficient is calculated for various temperatures with Equation (6.6). Then ER is plotted against T for the spectral lines of interest. By measuring Inm and determining HE from Equation (6.7), the temperature can readily be obtained. Figure 2 is such a plot for the optically thin 4158.59~ and 4259.36A ArI lines. 6.4 Determination of Aluminum Temperature Although local thermal equilibrium exists throughout an argon induction plasma, aluminum atoms introduced into the plasma by vaporization of alumina are not necessarily at the same temperature as the argon. Aluminum is formed from A10 dissociation at about 4400~K and cannot instantaneously reach the plasma temperature. In addition, the alumina vaporization occurs continuously along the reactor. So a steady stream of low temperature aluminum atoms is being introduced into the plasma at a given point in the reactor. Aluminum is also more easily ionized than argon, so that Al atoms are less likely to reach the higher kinetic energy levels. As a result, the average temperature of the Al atoms is lower than the argon plasma temperature. The determination of the aluminum temperature is done by the "multi-line" method. Equation (6.4) is written for an unspecified line as: I = KI)gA exp(-E/kT) (6.6)

VOLUME EMISSION COEFFICIENT, erg/cm3-sec-ster ~~~~~~~~~~~~~~... 00 CD olq 0 ( ~~~~~~~~J1~~~~~~~~~~~~~~ ~~~~~~~-3~ ~ ~~~~~~ — (D Co C m rn U, ~ CD: C:) rn CD;a 0~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ rn C O 0~ CD I' C) CD c+ CD CDC

-29where K = LhNo0/4r goQ. (6.7) Taking logarithms in Equation (6.6) gives: ~n(I/i gA) - nK E/kT. (6.8) Since K does not depend on E, a plot of i.n(I/ gA) versus E for lines with different upper energy levels yields a straight line of slope -l/kT e This relative method for temperature determination is necessary because of the build-up of aluminum and unvaporized alumina on the reactor wall during a run. The argon temperature, on the other hand, could be determined in the absence of alumina. The transition probabilities used in this work are those given by Corliss and Bozman, (21) although their value for the 3082.15A Al1 line is contested by Dickerman (26) and Deuel. This is necessary because Corliss and Bozman's set is the only complete one available. In addition to uncertainties in transition probabilities, Dickerman and Deuel found that self-absorption is invariably present in spectral lines of Al. Since self-absorption intensifies the line, and a 20 percent error in the intensity causes a 29 percent error in the temperature, (82) this method is not particularly accurate. As mentioned above, however, it is the only method available. The wavelengths, upper energy levels and transition probabilities for the Al1 lines used in this work are presented in Table I.

-30TABLE I UPPER ENERGY LEVELS AND TRANSITION PROBABILITIES FOR A1I LINES Wavelength Energy Level A ev 10g/sec 3082.15 4.0214 2.7 3092 71 4.0217 5.5 3944.03 3.1427 o.66 3961.53 3.1427 1.3

7. ANALYTICAL METHODS 7.1 Bromate Method for Aluminum Determination The bromate method is the standard volumetric procedure for determination of aluminum. The method can be accurate to within one percent for milligram amounts of aluminum. Interference is encountered when iron or titanium is also present, but these impurities can be removed prior to the aluminum determination.(6263) Using samples containing A1+3 in acid solution, the procedure followed in this work involves five main steps: (1) precipitation of aluminum quinolate from a solution of A1+3 and 8-quinolinol; (2) filtration of the precipitate and its dissolution in acid, forming 8-quinolinol; (3) bromination of the 8-quinolinol with a bromate-bromide solution; (4) reduction of the excess bromate with iodide, forming iodine; and (5) determination of the iodine with thiosulfate using a starch indicator. The precipitation of aluminum quinolate by 8-quinolinol is: Al1 3+ /3 O Ar1Oj + 3H+ (7o1) and the reverse reaction is the dissolution of the quinolate. The 8-quinolinol formed by dissolving the quinolate is then brominated by: 6H+ + 5Br- + BrO3 j 3H20 + 3Br2 (7.2) OH oH + 2Br2 + 2HBr (73) Br -31

-32After the bromination reaction has reached completion, the excess bromine (obtained from the excess bromate by Equation (7.2)) is reduced: Br2 + 2I- j 2Br- + 12 (7.4) This iodine is then reduced with thiosulfate by: 12 + 2S203j 2I- + S406 (7.5) If a known volume of bromate is used, the excess is determined by the amount of thiosulfate required. Thus the amount of bromate which reacted with the 8-quinolinol is obtained. With 0.1 N BrO3 (0.1/6 molar) it can be seen from Equations (7.1), (7.2) and (7-3) that 1 ml of 0.1 N BrO_ (which reacts with 8-quinolinol) corresponds 3 to 0.22483 mg aluminum. 7.2 X-Ray Diffraction Since x-ray diffraction was used for an analysis of the reactant alumina and for the identification of the reduction products, its principles are briefly discussed. A thorough treatment is given in Cullity.(23) Diffraction occurs when electromagnetic waves encounter a set of regularly spaced objects and the wavelength of the waves are of the same order of magnitude as the distance between the scattering objects. Such is the situation when x-rays are directed onto a crystal composed of regularly spaced atoms. The diffraction of a monochromatic beam of x-rays is governed by the Bragg law:

-33) = 2d sin O (7.6) where N = wavelength of x-rays, d = spacing between crystal planes, 20 = angle between diffracted beam and transmitted beam. By using x-rays of known wavelength and continuously varying 0, it is possible to determine the spacing of the various planes in the crystal causing diffraction. This is done with an x-ray diffractometer, which can be set up to record the diffracted lines. If d is obtained for the three most intense lines of a given substance, then it can be identified with the help of the ASTM powder data file. (8) In this manner, it is possible to determine compounds and uncombined elements present in the sample and their phases. 7.3 X-Ray Fluorescence X-ray fluorescence, which was also used in the analysis of the reactant alumina, differs somewhat from x-ray diffraction. In fluorescent analysis, the sample is bombarded with x-rays of sufficient energy that each element of the substance emits a characteristic line spectrum. This spectrum is analyzed in an x-ray spectrometer by diffracting the radiation with a crystal of known d spacing. If @ is determined, then the wavelengths of the characteristic lines can be calculated from Equation (7.6) and the elements can be identified, but the state of their chemical combination is not obtained by this method.

-347.4 Optical Spectroscopy The radiation emitted by a plasma or any other form of arc or spark discharge can be analyzed by means of optical spectroscopy. As with fluorescence, the spectrum emitted by an element or molecule is characteristic of the emitter. If the radiation from the discharge is passed through a prism or diffraction grating, the total spectrum is spread out and can be detected and recorded either photographically or with phototubes. Then if the spectrum can be separated into the spectra of the various emitting elements and molecules, these emitters can be identified with the help of the proper table of wavelengths.(50'5'7'7879)

8. EXPERIMENTAL APPARATUS 8.1 General The experimental system used in this investigation consisted of a plasma reactor, a radio frequency generator, a powder feed unit, gas delivery and cooling water systems, a safety interlock system, an exhaust fan and hood, and an optical spectrograph. This experimental setup is shown schematically in Figure 3. In addition, quench probes, x-ray sample probes, and equipment for plate development and analysis, x-ray analysis and wet chemical analysis were used. 8.2 Plasma Reactor The radio frequency, induction-coupled plasmas were generated and contained in the reactor diagrammed in Figure 4. This plasma reactor consisted of a water-cooled quartz tube held in a brass distribution heado Argon was fed into the reactor through an axial entry port at the top of the head and through two 1/32 inch diameter tangential ports in the head. The use of tangential feed imparted a vortex motion to the gas, centrifuging the hotter atoms to the center and keeping the inside quartz wall relatively coolo The innermost quartz tube maintained the axial and tangential flows until just above the induction coil, where the plasma region began. Argon was also introduced axially through the powder feed tube, which was a 12 inch long, 0o125 inch O.D., 0.094 inch I.D. alumina tube. The purpose of the first-mentioned axial gas stream was to provide a sheath for the powder as it entered the plasma with the carrier gas. This helped to keep the solids on the centerline of the plasma. The -35

TO FAN CODE PROBE C.W. TO DRAIN EACTOR C.W.._... 20___ _____ ~ C )TANGENTIAL 220V COAXIAL 0 ~ ~~~D ~ARGON CARRIER POWERSTAT GENERATOR HOOD ~ GAS ~~~~~FLOW (ARGON) FLOW IO METERS CO or CH POWDER USED AS R O ~~COOLING ARGON QUENCH GAS WATER (C.W.) H2, CO or CH4 USED IN PLASMA REACTOR FRONT VIEW HOOD Figure 3. Schematic Diagram of Experimental System.

-37CARRIER GAS AND SOLIDS 0.125 IN ALUMINA COAXIAL 4 — 1ARGON TANGENTIAL ARGON 19 MM O.D, QUARTZ 30 MM O,D. QUARTZ COOLING WATER OUTLET a 38 MM O.D. PYREX J COOLING WATER INLET Figure 4. Induction-Coupled Plasma Reactor.

-38majority of the plasma gas was introduced tangentially and whenever H2, CO or CH4 was used in the plasma, they were introduced with the tangential argon. 8.3 Power Supply and Auxiliary Equipment The radio frequency power for generating and maintaining the plasma was provided by a 23.5 KVA, 3 megacycle Lepel generator, Model T-10-3-3MC. By passing the r.f. current through a five turn induction coil, a 3 me alternating electromagnetic field was produced and electrodeless discharge could be obtained. With this generator, a maximum power input to the plasma of about 10 kw was attainable. The original thyratron power control for the generator had been eliminated and replaced with a three phase powerstat on the primary side of the power transformer. This had been done to reduce the ripple of the r.f. power. The plasma column was initiated with the high voltage discharge of a tesla coil, which partially ionized the gas. Feeding of alumina into the plasma was effected with a Plasmadyne Oscillating Powder Feed System. In operation, the powder hopper was vibrated by a variable amplitude oscillator to prevent packing and the argon carrier gas was passed through the bottom of the hopper. Powder was entrained in the carrier gas by way of a venturi pickup. The top of the hopper was sealed off during operation to prevent loss of gas pressure in the container. The gases used in the plasma and for quenching were supplied through calibrated Fischer-Porter Tri-flat rotameters. The gases were delivered to the rotameters at a pressure of 30 psig and entered the

-39reactor at atmospheric pressure. Cooling of the reactor and the quench probes was effected with tap water, which was supplied at line pressure. The reactor was protected against insufficient tangential gas and cooling water flow by safety interlocks, which would shut off the power. Since H2, CO and CH4 were used in this study, the reactor was enclosed in a 2 by 4 by 7 foot plywood hood, open on one side. Air was sucked through the hood at the rate of 1210 cfm by a Buffalo Forge Company exhaust fan driven by a 5 hp motor. Spark proof construction was used for the fan shaft and wheel. 8.4 Spectrographic Equipment A 3.4 meter focal length Ebert Mark IV stigmatic plane grating spectrograph manufactured by the Jarrell-Ash Company was used to analyze the radiation from the plasma. The spectrograph was equipped with a grating ruled with 15,000 lines/in giving a first order linear dispersion of 5.1 ~/mm at the focal plane over a useful range of 2100 to 7500. The entrance slit of the spectrograph could be adjusted in width from 4 to 400^ and in height from 1 to 15 mm. The Jarrell-Ash Number 18-022 scanning and condensing system could be used to scan and focus selected areas of an extended source onto the spectrographic slit. Seven front surface mirrors and two quartz-lithium fluoride acromatic doublet lenses automatically maintained focus and alignment with the optical axis of the spectrograph. A 1.72 to 1 image reduction between the plasma and the spectrographic slit was caused by the lenses.

Depending upon the application of the spectrograph, the spectra were either recorded on two 4 by 10 inch Kbdak SAl Spectrum Analysis Plates or by an RCA 1P28 multiplier phototube located at the center of the focal plane (direct read-out). By placing the two plates end to end, a total first order range of 2400 a was recorded during a single exposure. The phototube was operated with a supply voltage of 1000 volts dc obtained from a Furst Electronics (No. 710-PR) dc power supply with a continuously adjustable voltage output from 0 to 1500 volts. The current output of the phototube was passed through a variable precision resistor and the resulting voltage amplified by a variable range Leeds and Northrup (No. 9835-A) stabilized dc, Av amplifier. The amplifier output (0-10 my) was recorded with a Leeds and Northrup Speedomax H strip chart recorder. Horizontal scanning of the plasma (perpendicular to the axis of symmetry) was done by turning a screw drive which moved the mirror assembly. Vertical scans (parallel to the plasma axis) were made by pivoting a mirroro The spectrum at a given position in the plasma could be scanned (when direct read-out was used) with a sine bar wavelength drive (Jarrell-Ash No. 70-005), which rotates the grating so that the wavelength seen by the center of the focal plane corresponds to the wavelength indicated on a counter. Twelve scan rates, ranging from 1 to 500 I/min were available8.5 Quench Probes The quench probes were used both to augment the recovery of product by collection on its outer wall and to introduce-quench gases

countercurrent to the plasma flow. To meet this dual need, a watercooled probe consisting of three concentric stainless tubes was used (see Figure 5). Three different probes were used, with outer tube outside diameters of 0.148 inch, 0250 inch and 0.375 inch. The two larger probes had stainless tips heliarced to the tubing and the smallest probe had a copper tip silver soldered to the probe. 8.6 X-Ray Sample Probe The x-ray sample probe was designed to collect product samples suitable for x-ray diffraction analysis. X-ray samples could not be prepared in the usual way (grinding to -200 mesh and mixing into a matrix of vaseline) because of the highly pyrophoric nature of some of the product~ The sample probe, shown in Figure 6, was a water-cooled, 0~250 inch O.D. copper tube. A small cup was made on the end by setting a thin copper disc inside the tubing and silver soldering it in place. After collection of a sample, the cup was cut off and a new one made for the next runs 8j7 Plate Development and Analysis Equipment The two 4 by 10 inch Spectrum Analysis Plates used for a given set of runs were developed singly in a temperature controlled tanko The developer and fixer temperatures were held to within ~ 0*2~Fo During the developing and fixing, continuous agitation was applied to the fluid in the trays~ The plates were analyzed by scanning the spectra with a Leeds and Northrup Recording Microphotometer (No. 6700-P-1) in conjunction with a Leeds and Northrup Speedomas G Logarithmic strip chart recorder.

QUENCH GAS IN (WHEN USED) / 0.148, 0.250 OR 0.375 IN O.D. OUT IN COOLING WATER Figure 5. Quench Probe.

-43COPPER DISC COPPER 0~~8 INN S COPPER COOLING WATER I N COOLING WATER OUT Figure 6. X-Ray Sample Probe.

8.8 X-Ray and Wet' Chemical Analysis Equipment The x-ray diffraction analyses were done with a Norelco Diffractometer (Type 42202) using a copper target. A Norelco Universal Vacuum X-ray Spectrograph (Type 52530) with a chromium target and either a LiF or EDDT crystal was used for the fluorescent analyses. The output from the diffractometer and spectrograp:h was recorded with a Minneapolis-Honeywell strip chart recorder. The quinolate precipitations were performed in 50 ml erlenmeyer flasks and the precipitates were filtered out with a sintered glass funnel, using vacuum. The solution was cooled to 60~C, prior to the filtration, in a constant temperature bath. The dissolution of the precipitate and all subsequent operations occurred in a 250 ml iodine determination flask. All weighings of less than 200 gm were performed with a RIGHT-A-WEIGH balance manufactured by William Ainsworth and Sons, Incorporated. This balance is accurate to + 0.2 mg. Quantities of over 200 gm were weighed on a Mettler (Tara 0-2000 gm) balance, which is accurate to ~ 0O5 gmo

9. EXPERIMENTAL PROCEDURES 9ol Start-Up and Safety Procedures After supplying cooling water to the generator and throwing a line power switch, the oscillator filament was actuated and allowed to warm up until the plate ready light began to glow. Then the argon (including carrier gas) and cooling water flows were initiated. Power was supplied to the plate, the powerstat set on 50 and the grid current adjusted to 10 percent of the plate current with the grid control. The powerstat was set on 70 and the high voltage from the tesla coil was applied to the argon above the induction coil to partially ionize the gas and create a discharge. If a stable discharge was not obtained, the powerstat reading was increased and the tesla coil treatment repeated until the discharge persisted. By slowly increasing the powerstat reading, the plasma was obtained. The plasma intensity was then maximized by readjusting the grid current to 10 percent of the plate current with the grid control. The desired operating conditions were set by proper balancing of the power and grid controls. Oxyacetylene welding goggles were worn whenever the plasma was viewed directly and warning lights, which had been installed at the laboratory entrances, were on during all plasma operations. When the reducing gases, H2, CO or CH4 were used, the exhaust fan was turned on and the feed line for the gas was flushed with helium. It was determined (by conversing with The University of Michigan Environmental Health personnel) that an exhaust rate of 1210 cfm of air was adequate protection against a maximum possible H2 usage

of 1 cfmo The plasma was initiated and optimized with argon as described above. If the reducing gas was to be used in the plasma itself, the gas was slowly bled into the plasma with the tangential argon while the power and grid controls were constantly adjustedo When the desired reducing gas flow rate was reached, the power level for the run was set and the grid current-plate current ratio balanced. The use of H2, CO or CH4 as a quench gas did not affect the generator operating conditions since the gas did not pass through the induction coil. Therefore, much higher flow rates of the reducing gases were possible. After the argon plasma was started and the operating conditions were set, the quench probe was positioned in the plasma with a vertical traversing mechanism to be about 3/4 inch below the induction coil, as shown in Figure 7. This could not be done prior to the plasma start-up since the induction coil would couple to the probe instead of the argon and a plasma could not be obtained. When the probe position was properly set, the quench gas flow was initiated and brought to the desired level. The correct aluminum oxide flow rate was attained by adjustment of the position of the venturi tube in the bottom of the hopper (a constant carrier gas flow rate was used). The powder pick-up arrangement is depicted in Figure 8. The powder flow rate was determined by weighing samples collected during a known time interval. This was done before and after each run. The oxide was not introduced into the reactor until the plasma had been established.

ARGON FLOWS WATER JACKET SOLIDS O ~~~IN QIN ~~INDUCTION COIL 0 0~ 0 0 0 O ~~~PLASMA REGION QUENCH GAS Figure 7. Position of Quench Probe in Reactor.

POWDER HOPPER BACK PRESSURE LINE ADJUSTABLE VENTURI TUBE POWDER CARRIER —,all~il U OUT GAS IN Figure 8. Powder Pick-up Arrangement in Bottom of Hopper.

9.2 Selection of Reactor Design According to Marynowski,(68) when solids are fed into a plasma, it is desirable to use an induction coil with a reverse turn at the bottom in order to achieve axial stability of the plasma. Supposedly, the introduction of solids shifts the vertical position of the plasma. This is prevented by the reverse turn, which sets up a magnetic field in opposition to the plasma flow. This approach was tried in the initial experiments, but was found to be unsuitable. With a coil of five turns in one direction and one reverse turn, an arc between the reverse coil and the cooling water was invariably obtained during the start-up of the plasma. The result was breakage of the pyrex cooling water jacket. Since a glass reactor was necessary for the spectrographic studies, the above coil had to be abandoned. The coil was replaced by a simple five turn coil and axial stability problems were not encountered. In an attempt to collect all of the reduction products, a brass product collector preceded by a water-cooled quench chamber was placed directly below the reactor. Although the chambers were supported from below, they put too much constraint on the quartz reactor. As a result, the reactor would break during a run. Since the brass chambers had to be discarded it was decided to forego total collection of the products. So the conversion was based upon the amount of material actually collected and quench probes were used to increase the recovery. The third item of importance was the choice of an alumina powder feed tube. A water-cooled stainless feed tube, similar to the quench probes (see Figure 5), was first tried. In order to achieve

-50centerline feed of the solids into the plasma, however, the tube tip had to be placed about a half inch above the induction coil. This caused arcing from the coil to the tip and water leaks developed. The alumina tube was then tried and was found to be perfectly suitable. Even though the plasma region began at the end of the tube, melting did not occur. 9.3 Qualitative Analysis of Alumina Since the aluminum oxide used in this investigation was not chemically pure, an analysis of the material was necessary. The presence of small impurities would not affect the alumina reduction but could interfere with the quantitative analysis for aluminum. A probable analysis of the alumina was supplied by the manufacturer (Carborundum Corporation) and it was checked by the methods discussed in this section. X-ray diffraction, x-ray fluorescence, and optical spectroscopy were all used for the alumina analysis. A sample of aluminum oxide in a matrix of vaseline was analyzed with the diffractometero The procedure was to scan the region from 20 = 15~ to 80~ (see Section 7.2) and identify the line spectrum as much as possible. Fluorescent analysis of alumina powder was similarly performed. The powder was placed in the spectrograph, a vacuum drawn and 20 varied over a range large enough for identification of the elements present. Both LiF and EDDT cyrstals were used for the analysis. The x-ray tube power supply was operated at 35 KV and 15 ma for the diffraction and fluorescence studies.

-51Optical spectroscopy was used in two different experimentso In the first, alumina was fed into a plasma and various regions of the spectrum were scanned using direct read-out. Identification of the spectral lines constituted an analysis of the elements present in the alumina. For the other analysis, alumina was put in a hollow cup made in the end of the vertical electrode of a carbon arc lamp (Bausch and Lamb Optical Company, Automatic Arc Lamp, 5-10 amperes dc). The lamp was mounted on the optical bench of the spectrograph, an arc was struck and the spectrum recorded on a pair of SAl plates. Analysis of the spectrum identified the elements present. 9.4 Analysis of Reaction Zones The purpose of analyzing various sections of the plasma was to identify the aluminum containing species present. This was expected to provide significant information about the nature of the vaporization and quench processes, The analysis was performed with the optical spectrograph, using direct read-out, by scanning spectral regions of interest at various points in the plasma. 9.5 Identification of Product The rigorous Group III qualitative chemical analysis given (69) by McAlpine and Soule was used to examine the product for the presence of Al, Fe and Cr. "Aluminon-reagent" (0.1 percent solution of ammonium aurin tricarboxylate) was the aluminum detector. Prior to this determination, the product was placed in half concentrated HC1 to put the metals into solution.

- 52It is conceivable that A10 or A120 might also form A1+3 in HC1 and give a positive "aluminon" test. Since A10 and A120 were not ruled out as possible products, an x-ray diffraction analysis of the product was also madeo Three product samples were obtained with the x-ray sample probe set directly in the plasma, using the vertical traversing mechanism (as in Figure 7). The samples were mounted linearly in a standard x-ray sample holder. The sample holder was rotated to give a maximum intensity for the O<- A1203 line at 20 = 37.75~. Then 20 was varied continuously to give a spectrum of diffraction lines. Samples collected in Ar-CO and Ar-CH4 plasmas were also examined by diffraction analysis. This was done to see what effect the addition of carbon to the system had on the nature of the products. 906 Determination of Aluminum The following procedure was followed to determine the amount of Al in a given product sample~ 1. Put the product in half concentrated HC1 (to dissolve the Al) and weigh the total. Use a weighed flasko 2. Allow the undissolved solids to settle and determine the density of the solution by pipetting off and weighing three 10 ml portionso 3. Pipette off most of the remaining solution and save. 4. Wash the solids, filter and weigh. 5. Take a known volume of the solution, precipitate the aluminum quinolate and cool to 60~C. Two samples were used for each run.

-536. Filter the quinolate, dissolve in HC1 and determine the Al by the bromate method. 7. Determine the volume of the total solution from steps 1, 2 and 4. The Al density of the sample in step 5, multiplied by the total volume, was the total amount of Al in the product. The method used for precipitation of aluminum quinolate(63) applies to solutions containing not more than 0.5 mg Al/ml. Tartaric acid equal to about five times the amount of Al present and 4 to 6 gm ammonium acetate were added to the solution. Bromocresol purple (8-10 drops) was added and the solution neutralized to the purple end point with half concentrated NH40H. An excess of 8-quinolinol (1 ml/3 mg Al) was added, the solution heated in a boiling water bath until a yellow precipitate was obtained and the mixture cooled to 60'C in a constant temperature bath. The 60~C solution was filtered with moderate suction through a sintered glass funnel and the quinolate washed with 100 ml of cool water. The precipitate was dissolved in half concentrated HC1 and dilute HC1 was sucked through the frit and added to the volume. The 8-quinolinol formed by dissolution of the aluminum quinolate in HC1 was brominated with the KBrO3-KBr solutiono An excess of about 2-3 ml of the solution was added; the mixture stirred and allowed to react for 60 secondso An excess of the 20 percent KI was added and the iodine titrated with thiosulfate to the starch end point. The starch was added just before the yellow iodine color disappeared to prevent

unfavorable side reactions. This method is discussed in greater detail by Kolthoff and Belcher.(63) The effect of Fe and Ti in the product was investigated by removal of the impurities and determination of the Al. Kolthoff and Belcher (63) agree with Kodama( that Fe and Ti quinolates precipitate in the presence of sodium malonate in 5 to 10 percent acetic acid. Therefore a known volume of solution (see step 3 above) was neutralized with NH40H and glacial acetic acid added. Di sodium malonate (80 times that needed to complex the Al) was added, the solution diluted to give 5 to 10 percent acetic acid and 8-quinolinol added. The solution was boiled gently to precipitate the Fe and Ti quinolates, cooled to 60~C and filtered. The filtrate was treated to form aluminum quinolate by addition of tartaric acid, ammonium acetate and dropwise addition of NH40H and 8-quinolinol. The precipitation was done in a boiling water bath. The amount of aluminum was then determined by the method outlined above. The Fe-Ti removal procedure was repeated with a pure A1+3 solution as a checko 9.7 Measurement of Argon Temperature The argon temperature was measured at the power levels studied in this work, using a clear reactor of known transmittance (see Appendix IV). The 4158.59 and 4259.36 A ArI lines were scanned at 10 A/min and the outputs of the direct read-out from the optical spectrograph were integrated to give the absolute line intensities. The argon temperature was then determined as the average of the two values obtained

-55from Figure 2. Since radial temperature profiles exist in the plasma, the argon temperature obtained above represents an average along the diameter of the plasma. 9.8 Measurement of Aluminum Temperature Aluminum temperatures were determined by the method discussed in Section 6.4 for most of the final experimental runs. The temperatures were measured at a point 1/2 inch below the coil on the centerline of the plasma and are an average for the plasma along the diameter. The spectra at this position were recorded on the SAl plates with the use of the optical spectrograph. Line profiles were obtained for the 3082.15 A, 3092.71 A, 3944.03 A, 3961.53 A Al1 lines used in this work. The lines were scanned with the recording microphotometer to give profiles of silver density of the line recorded on the plate versus wavelength. These were converted to relative intensity versus wavelength by the calibration curves (see Appendix III). Integration of these profiles gave the absolute intensities of the four lines and the temperature followed from the "multi-line" method. The integration procedure is discussed in Appendix VI. 9.9 Study of Reaction Variables The three variables that were studied in detail were the A1203 particle size, A1203 mass flow rate and power input to the plasma. This three variable investigation was preceded by an examination of the effect that build-up of solids on the reactor wall had on conversion of alumina. The purpose was to see if it would be necessary to limit the length of the runs.

-56The A1203 particle size and flow rate were varied at a constant power input level (the lowest power level of this work). Several flow rates were used for the smallest particle size to determine the shape of the flow rate versus conversion curve. This made it possible to use only three flow rates for the other two particle sizes and still cover roughly the same range of conversions. The effect of power input was determined by a single set of runs. Three power input levels were used at a constant A1203 flow rate and the smallest particle size. The use of diatomic and polyatomic gases, viz. H2, CO and CH4, in a plasma requires considerably more power than argon alone. Since the range of reducing gas flow rates was restricted by the power capacity of the r.f. generator, it was decided to use a single run for each of the three gases. An intermediate A1203 flow rate and the smallest particle size were used for the runs. The argon flow rates were maintained constant for all of the runs discussed in this section. 9.10 Quench Methods For the experiments covered in Section 9.9, the conversion of alumina to Al was based on the amount of product collected instead of the A1203 input flow rate. Water-cooled quench probes were then used to show that it would be possible to increase the recovery without lowering the conversion. The significance is that if a reactor were designed to recover nearly all of the product, the conversions obtained in this work would be maintained. This could be done, for example, with a water-cooled metallic reactor followed by a suitable

-57collection chamber. The probes were positioned directly in the plasma so that the tip was about 3/4 in below the coil (see Figure 7). This was done with a vertical traversing mechanism. Three probe sizes were tried and a single run was made with each, using the smallest A1203 particle size. The effect of H2, CO and CH4 as quench gases was determined by introducing the gases through the 0.250 in O.D. quench probe countercurrent to the plasma. The probe tip was an inch below the coilo Several H2 flow rates and one each for CO and CH4 were used.

10. EXPERIMENTAL RESULTS AND ANALYSIS 10.1 General The scope of this work was outlined in Chapter 4 and discussed in greater detail in Chapter 9. The results of the experimental program are presented and analyzed in this chapter. 10.2 Qualitative Analysis of Alumina The reactant aluminum oxide was analyzed by several qualitative techniques and the results compared with the probable composition supplied by the manufacturer (Carborundum Corporation). It was not possible to obtain a complete analysis by any one of the techniques employed, but the composite analysis was essentially complete. The results of the various analyses are presented in Table II and the composite analysis is compared with Carborundum's probable analysis in Table III. TABLE II ANALYSES OF REACTANT ALUMINUM OXIDE 1. X-ray 2. X-ray 3. Optical Spectroscopy Diffraction Fluorescence (a) Plasma (b) Carbon Arc Go-A1203 Ti Al Al Fe Ti Ti a few unidentified Zr Fe Fe impuritie s Ca Si -58

-59TABLE III COMPARISON OF COMPOSITE ANALYSIS WITH CARBORUNDUM'S PROBABLE ANALYSIS Composite Carborundum's Analysis Composition o(-A1203 o(-A1203 97.03 % Ti TiO2 2.10 Si SiO2 0.50 Fe Fe203 0.20 Zr ZrO2 0.13 Na2O 0.02 Ca MgO+CaO 0.02 100.00 % X-ray fluorescent and optical spectroscopic analyses identify only the elements present in a sample and not the state of their chemical combination. Diffraction analysis determines the compounds present but requires at least three diffraction lines for a positive identification of a substance. Thus it is not very suitable for impurities present in small amounts. None of the analytical techniques used are particularly sensitive to oxygen or small concentrations of Mg or Na. Consequently, the composite analysis is in substantial agreement with Carborundum's probable composition and the latter can be accepted as true.

-60The presence of Ti02, Fe203, ZrO2 and SiO2 in the reactant alumina indicated that Ti, Fe, Zr and Si would be possible contaminants in any reduction product. Therefore, the effect that their presence would have on a quantitative analysis for Al was determined. The result of this investigation is given in Section 10.5. 10o3 Analysis of Reaction Zones The aluminum containing species were identified at various points in the argon plasma and tail flame by using the direct read-out of the spectrograph. As is shown in Figure 9, the plasma core consisted only of atomic and ionic species. This hot core was surrounded by a thin region containing AlO and Al. Both A10 and Al were also detected throughout the cooler tail flame of the plasma. Figure 10 shows typical A1I lines in the plasma core and their intensities relative to nearby TiII lines. The presence of TiII lines throughout the spectrum instead of TiI is explained by the greater relative intensities of the Tii1 lines. The FeI lines were also much weaker than TiII lines although the former were definitely obtained. The 4850 A, 2 2 5 band system of A10 was observed in the tail flame and is partially presented in Figure 11. It was impossible to discern if A120 was also present anywhere in the plasma since A120 spectra have not been reported in the literature. However, some negative inference can be drawn from the absence of extraneous band structure, which could be attributed to A120, throughout the spectral regions scanned. Although AlO was not observed in the hot plasma core, its presence in the cooler regions of the argon plasma supports the contention

ARGON FLOWS WATER JACKET INDUCTION PLASMA COIL AWATE R B A C'O | GREEN IONS - WHITE MOSTLY AlO BLUE WATER A10.Al GREEN MOSTLY A10 ORANGE-RED Figure 9. Composition of the Sections of the Plasma.

C4 r-1 C14 r-4 r4H c)r-4 M~~~~~~~~~ cn~~~~~~~~~~c H06~ ~ H CY 0- o I~~ ~ ~ ~ ~ ~~~~~i~i,_ %0i Co oo~~~~~ H ~ I —t 0~~~C4 r'- -rH "o LO H H CH P co-H r r LLJ ~ H a Co )~~~~~~~~OE LU CY)H* O Z o E ~, _ - - - I-4 Co I I I 1 IP 0~ 0 0 a: __ — (V) *HM 3070 3080 3090H3100 O H UL) Co_ _ Fgr(.Tycl33070 3080 3090 3100 71 oE-~V cr~~~ I IO, - L1L I III1! o L I!1 I I I o WAVELENGTH,, ANGSTROM UNITS Figure 10. Ty-pical 3082.15 and 3092.71 A Al1 Lines in the Core of an Argon Plasma.

[~ —-Jl~l~ 660173 LEEDS a- -T —' P No. 66 ITS EgS......._ - - - - 4842.1'. A,"_.. —-_' -.-' _"- ___.__ - A- -.... - i. Q___ I_ _'' - - -'-. -.Co — -- ___ A1O~~~~~~~~I O0, 0)...... - -._ --— *- - BAND HEAD[- _ --- -............. - U - -_. —i_ —....-0 —----— i" o —:. 0 ~~~~~~~~~00 - _.....-.-.... ---............... J -. —--........ F-...... "............................ V............... _ _..............._...... _....._... A... -.-.-... r%................ —-............, ~........ ~........... __ __... __.........: __.........._...........!.............................. EJ) 4841 4842 48-114844 4845 WAVELENGTHe X s ANGSTROM UNITS 2 2 Figure 11.- The 418412 A P10(0,0) Band of the 41850 A 2 - 2 Band System of AMO Obs'erved in the Tail Flame of an Argon Plasma.

that A10 predominates over A120 in a neutral or oxidizing atmosphere (see Section 3.2). Thus it is quite credible that the A1203 dissociates to A10 and 0 as it is heated in the plasma. The A10 in turn quickly dissociates to Al and 0, which explains the absence of detectable A10 in the plasma core. The formation of A10 in the cooler regions also means that any quench process aimed at recovering Al from the plasma must either be rapid enough to prevent significant formation of AlO or provide a reducing agent to preferentially react with the oxygen. 10.4 Identification of Product The rigorous group III qualitative chemical analysis of product obtained from the reduction of A1203 in an argon plasma gave positive tests for Al and Fe. The test for aluminum was quite strong whereas that for Fe was weak. The presence of Al in the product was confirmed by an x-ray diffraction analysis. The only diffraction lines in addition to the Al lines were due to o(-A1203, -A1203 and the impurities that were present in the reactant alumina. The Al diffraction line at 20 = 38.50 and the adjacent o(-A1203 line are presented in Figure 12. Since these lines overlap somewhat, the missing portions of each have been sketched. It is certain from the above results that Al is the reduction product of A1203 instead of a solid aluminum suboxide. This Al was found to be highly pyrophoric and finely divided. It is also definite that Fe ispresent to interfere with the quantitative analysis for Al Since Fe203 is much more easily reduced than TiO2 or ZrO2 (see

INTENSITY,-J~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~, "3~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~1 ('D _j 4-rFO ~ ~'!.:i -- tt:... —?!-i -i..!...= T'.. ='-r'-r -r-r'., 7-1 7..i-T" r -I-"r -'...t.-H,-:i.. —"-l m 0 to'-t_ T' -T~~~~~~~~~~~~~~~~~~~~~~.i.. t.t- -t'i ~ t —'i~ — ]..-t'"'!'"!"l''.;-Tr CU t..-~-~ —i ~+; -....;;-i-i-~ — I — 7- T,-t... i'..i —r-T-'r' —r.'-:......;'.'~:/i~~~~i-]-'':'"'?t"~"?-"~-'~-T'"~'=""-~ —~'fifI i~P:~'ifit-l'=~-"-~'?-~"~'~'~-' t-.? t-' —"-q~ —iit- ~'T- ~~ I a- ( t- i 7- -T"': 4 —?-'!-''-' —i —I- -' "~t'~''!t — T c-!' 0..._ >- =7'?(D —I - —':-+.-"+....:- — i —. —-. i0. -F.":I.:.:[. ~' o ~d r ~1 -— l —; —i -;i-> 0.-.-.....t-,-?-+.-i —F.~ r ~ i-i'-t i:".-": —:.!-t:: I -:i-:-' —:-i-'-' -... —-'""," I~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~-"I'1-;,:!,,:;~ —:",,T FJ. -, —.t —-...... D-4 —-..''''1 -+'-i'' T:f —'~....' [."' ~ "T-..I. _.i-f' t -i )3~~~~~~~~~~~~~~~~~~~''")- - " " 1-'~-i-r- -:T-'t''T ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~"1 —t'rT-T ~ T"t i' x-:;'-fit~'r' r-~1- i 1.., i-T' i...L~~~~~~~~~~~~~i.!-r-~r? - -$- 1-~~~~~~~~~~~' i I i I) T'f~~~~~~~~-'r` i_~~~~'i.;-.4; 4"* ~ ~ ~ ~ ~ ~ ~ ~~ ~~~1i~ i~ i OO I I-.~-:' -I — -i i-;-f:~ f-' +t? ~i — i.,~ —" il -!i-i~ +- —';.-~ ITI vt~~~~~~..'~-.[..~:.~ —:,~ i-.; tt' -! "..'.;;t;c_.]....L "i';:-":""-e-I - -i 00~~~~~~~~~" -I i-i-i i h)W — ~: j i. i -I I~~''!''~.' ~~~~d o.;F] %~~~~~~~~~~~~~~~~~~~~~~~~~~~~; z 7~~~~~~~~-'~" "-'... —T —'""'"":'"........ ~' —I~ ~. ti.1..:.':" —' —'!: -m'"'"' i -r 1- t —.-~..-t-'I..................... _~~-~-~ ~~ l _ -.t..1.;_'._i' —;i-Jj!-i' —'":,-'~ t";"'. l.. i... ~ I'.~'',-i — f' ~~ —~t ~~-~,l-I' —'':I' ~ ~ —=-~T —-r......t.-.i...-] — r i......~'l"-.;, —?-i —-;I.~......;-......[T';[].. —'-.:'-'; Ij —t —i- I -i —t — Pi T~~~~~~~~~~~~~~;- I~~i.t. i.' i 1"1: ~~~~~~~~~~~~~~~~~~~~~~~T~~~~~~~~t i.. -i i~ ~ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~...." a) II,.... -.- -! F= tl ~ ~ ~~~~~..~- lx'~' ~I i."t]''-r'"' —i.. i-! o — i.I.~.....~' i'i'"',"' ~ —!'" —-';:..i:-i-'!', "-i-"T-i:-t i-lT-i-ri'-"."''~...........:;,....... ~,.~.....;T..~-:.4. ~'~' —"'-''~, ";-"i~t~ct 5 I i T —--- 00~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~,_...._.....,._....l...........,. ~ "' ~ ~ ~ ~.:' ~"'"?.-...-.-;!: I:.-~.?-~.. -~~,. ~;~' ~ ~5. t i...c.ID-...-. —i- t —.- t-~ —t i —. i~', i. I o.!..j'i,,., ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ i............ -..,-t~-.- I -.-:.-.........._.,:..?_..... ~~~~~~~~~~..-~,_ I.... 3~C) T -i-,~ —...-!.-,:i-t.-~:;"tt i:'i i.~ -- --, -:. —4...........iu, i- -i~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~: 1.:.i~{: _;.r ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~;~ -: -'[' " ~'~' _A ~3 0: i - -... ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~..I...~...... O PI - i -I....r:.... ri i -: i tiI i L i1 I. -: ——: i —i -~ t-j~~~~~~~~~~~~~~~~~~~~~~'...'-..... ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~.:-F ~~~Id.! —-... h _iq~~~~~~~~~~~~~~~, L'-i ii~.~ ~~~~~~~~~~~~~~~~~~~~~~~~~::-i'?':'+i''- r"'r~~~ i''~ tj C~~~~~~~~~ (Dh".......... I':!':: " ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~~~~~1i:'!:"I-;;t ":-~: -.- i:- o!... u, r 1! _i _ r.:":'~-'.-' iz -I; Ii.,i — i-. i —?-i+-ff..j.:............i' CT9 C! r~~~............................t...........- — I —~ r ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~s~~~~~~~~~~' "." I -..:''.1 _...1.1.i':, "'-r"' i''.-i..-~r, t~Zi-. I- " a vt ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~."Tl-'''""?'d~~~~~~~~~~~~~~~~~~i:i! -I~ ~ ~~ ~ ~~~~~~~:" L...i.'~- ~~ — ~~ ~-~; —~ "-'~-'"~" -"'.. i~~ ~ ~ ~ ~~~~~~~~~~~~~~.... i,........: -!: — i-i!t`i!'. ~.~ii..;... (; 3~~~~~~~1~~ i",i-.-;';...:.......... I. ii_'. m'''' i~~~~~~~~~~~~~'i ~..,..- -..~.. -~-~..'.-'..'"!i i. ~...t _. ~...i..' ~ "......~, t I -:'? i — I I~..-~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~-r O~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~'T

-66Section 2), and since the amount of Fe is much less than the Al it is probable that Ti and Zr are not present to a significant extent in the product~ Zirconium can definitely be neglected because of the very low concentration of ZrO2 in the original alumina. Samples obtained from the reduction of A12 03 in Ar-CO and Ar-CH4 plasmas were also analyzed to determine the nature of the products. Both gave positive qualitative tests for aluminum with the "faluminon-reagent" and both contained a considerable amount of carbon. Diffraction analysis of the product from the Ar- CO reduction identified Al,-A10 and Y-A1203. Similar analysis of the product from the Ar-CH reduction indicated that A14C3 was present in the sample The aluminum oxycarbides, A120C and A1404C, were not found in either case nor were solid aluminum suboxides. lO.5 Accuracy of Aluminum Determinations Extreme accuracy was not attempted for the quantitative deter-~ minations of aluminum in the product,, The alumina mass flow rate varied enough during an experimental run, sometimes by as much as~ 15 percent, that it was sufficient to know the Al content within 5 percent. The bromate method for determination of Al was checked with measured volumes of a solution containing a known amount of Al. The solution was prepared by dissolving 1K012 gm of Al powder in HrCl and diluting to 1 liter. These checks were performed prior to and during the time that experimental runs were being made. The aluminum in the known samples was determined within 1l4 to 2,,5 perc-ent of the c-orrec-t, nmount_ The valuje obtained was always

-67less than the correct value. Therefore, the aluminum determined in the reduction products should have been within 2.5 percent of the amount actually present, assuming that the contaminants Fe and Ti had no effect. Since these metallic impurities would show up as Al in the bromate method, their presence should give a high value for the amount of aluminum in a sample. To ascertain the influence of Fe and Ti, a 10 ml sample of the solution obtained by reacting the product from a run with HC1 was treated to remove Fe and Ti e The remaining solution was found to contain 2.73 mg A1 while 2.62 mg Al were found in 10 ml of solution from which Fe and Ti had not been removed. This variation of + 2 percent from the average was within the desired accuracy limit and it was decided to forego removal of Fe and Ti in the later work. Subsequent results obtained with duplicate samples also gave variations as high as + 2 percent from the average (duplicate samples were used for the final experimental runs). In view of the above results, it is felt that the aluminum determinations were easily within the desired 5 percent accuracy limits and in most cases were within 3 percent. 10.6 Effect of Alumina Flow Rate and Particle Size The alumina mass flow rate and particle size were varied while holding the power input level and argon flow rates constant. The experimental runs were of a two minute duration. The time limit was dictated by the effect of the build-up of solids on the reactor wall, which was determined from the results of runs of different duration, presented in Table IV. They show that the greater the amount of solids collected on

-68the reactor wall, the lower the percent conversion of the alumina. These results could not be explained by the variation in the alumina flow rate. So in order to determine the effect of varying this flow rate the subsequent runs were of a constant time length. Since very low flow rates were planned, at least two minute runs were necessary to collect sufficient product for analysis. TABLE IV EFFECT OF SOLIDS BUILD-UP Run Alumina Total Percent Length Flow Rate Solids Conversion (min) (gm/min) Collected (gm) 0.5 0.72 0.25 2.6 1.0 0.39 0.39 2.5 2.O o.6o o.84.1.9 4.0 0.70 1.95 1.8 Three particle sizes~ 500 mesh (26,A), 400 mesh (37/,z) and 320 mesh (45-.:; and alumina flow rates ranging from 0.03 to 0.60 gm/mm were used in the study. The power input to the plasma was 5.03 kw and.the axial (solids carrier gas), coaxial and tangential argon flow rates were 2.5, 3.7., and 45.4 gm/mmn respectively. The variation of percent conversion with alumina flow rate for the three particle sizes is shown in Figure 13. A single run was also made with 60 mesh (250At) alumina flowing at 0.19 gm/mmn. There was no aluminum in the product nor was

-69~32~-,___ ___ =IiI 30I.. SYSTEM: ARGON PLASMA POWER INPUT = 5.03 KW 26 ______ 24__ _ _ _ _ 22 a = 320 MESH (451) o = 400 MESH (37P) 0O = 500 MESH (26w) D = 325 MESH (44p) ESTIMATED FROM ~~~~18 _________ ~~OF STOKES U... z,c., 16_ _ _ _ _ _ _ 0z14 ____ Lz ~12__ _ 10 6 4 0 0 Oil 0.2 0.3 0.4 ~~~~0.5 0,6 ALUMINA FLOW RATE, GM/MIN

-70conversion of alumina for this study was based upon the amount of aluminum and alumina collected on the reactor wall. As expected, the 500 mesh oxide gave a higher conversion than the 400 mesh alumina throughout the entire range of flow rates that were used. This is logical because a smaller 500 mesh particle vaporizes more completely than a 400 mesh particle under the same conditions. The increase in conversion with a decrease in solids flow rate exhibited by both 500 and 400 mesh alumina, is also reasonable. The lower the alumina mass flow rate, the fewer particles are present in any section of the plasma at a given time. This means that energy is transferred from the plasma to each particle more efficiently as there is less interference due to nearby particles~ As the oxide flow rate increases, the hindrance to the heat transfer becomes greater and the particles are less thoroughly vaporized. When the flow rate of solids is sufficiently great a larger load is presented to the generator and the plasma begins to be affected. Eventually, the plasma is extinguished'unless the power input is increased. The anomalous behavior of the largest particles (320 mesh) can be explained. Taken by themselves, the results for 320 mesh oxide are reasonable as the increase in conversion with decrease in oxide flow rate is exhibited. In addition, the high flow rate result agrees well with the conversions obtained for 325 mesh alumina by Stokes, etl._ 86 as is shown in Figure 13. Since Stokes and his associates did not correlate conversion with alumina flow rate., the point in Figure 13 attributed to them is their highest conversion (5 percent) at their

-71which requires considerably more power than argon for comparable plasma temperatures and heat transfer, their 11.6 kw power input is not significant. This was confirmed experimentally. A helium-argon nplasma (033 gm/min He and 51.6 gm/min Ar) required 5.33 kw for stability while a pure argon plasma flowing at 51.6 gm/mi was stabilized with 5.03 kw. But the use of helium and the higher power input did not improve the conversion of alumina. The addition of H2 to the plasma flame (as the oxide carrier gas) by Stokes and his co-workers is similar lto the quench studies of this work because the H2 did not passthrough the arc. It is shown in Section 10.9 that this also does not greatly increase the conversion. The intersection of the conversion versus flow rate curve for 32,0 me sh with the curves for 4I00 and 500 mesh alumina at higher flow rates is explained by the relative extent of hindrance to the heat transfer. As the particle diameter increases, the number of particles at a given mass flow rate decreases. -So there are fewer 320 mesh particles in a given volume of plasma at any time and the heat transfer from the plasma is more efficient. As the alumina flow rate decreases., this effect is less noticeable and the curves intersect. At very low oxide flow rates', single. particle flow is approached and an increase in conversion with decrease in diameter is clearly ex-.hibited. The conversions obtained at low flow rates agree qualitatively with the extents of vaporization predicted by the computer solutions to the heat, mass and momentum balances. The comparison of the computer and experimental results is given in Section 10.11.

-72107 E ctofPower Input The effect of the power input to the plasma was determined by varying the power at a constant alumina flow rate of 0.23 gm/m. Power inputs of 503, 5.86 and 6.69 kw and 500 mesh alumina were used. The argon flow rates were the same as in Section 10.6. The variation of conversion with power input is given in Figure 14. The increase in conversion with an increase in power input was due to two phenomena. First, the argon temperature and the energy content of the plasma were greater. As a result, the particle was more completely vaporized and the concentration of aluminum in the plasma was greater. Second, the aluminum temperature was higher so that the aluminum had a greater diffusivity. (The temperature results are given in Section 10.10) It was shown in Section 5.2 that both effects increase the molar flux of Al to the quench surface. Consequently., the rise in the conversion at the higher power levels was expected. As in Section 10.-6,, the conversion of alumia was based upon the amount of alumium and alumina collected on the reactor wall..10.8 Effect of Reducing Gases in the Plasma Single runs were made with each of the reducing gases (Hi2. Co and CH4) added to argon plasmas. The same argon flow rates as in Sections 10.6 and 10.7 and 500 mesh alumia were used in the study. The results are presented in Table V along with the results for argon.plasmas at comparable alumina flow rates and power levels. Only the alumium and alumina that collected on the reactor was considered.

20 18 114 14 ~ ~ ~ ~ i J3. -. < 12 U0 zB 0 10 _ _ _ _ _ _ _ LU z 0 8 ARGON PLASMA____ o PARTICLE SIZE = 500 MESH (26k) Uj ~~~~~~~ALUMINA FLOW RATE =0.23 GM/MIN 0 14.0 14.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0 POWER INPUT TO PLASMA1 KW Figure i14. Effect of Power Input Upon the Conversion of A1203 to Al

-74TABLE V EFFECT OF H2, CO AND CH) Gas Gas Alumina Power Percent Flow Flow Rate Input Conversion (gmol/min) (gm/min) (kw) of Alumina H2 0o045 0.17 5.53 22 CO 0.o16 0.15 5.63 46 Ar 0.16 5.58 23, CH4 0.025 o.o17 5.03 42 Ar --- 0.17 5.03 9 It was predicted in the discussion in Section 5.1 that hydrogen would not greatly increase the conversion of alumina and that carbon would be of considerable benefit. This is confirmed by the results obtained with H2, CO and CH4. Hydrogen had no noticeable effect while carbon monoxide nearly doubled the conversion obtained with argon alone. At a lower power level with methane the conversion was more than quadrupled. It appears that methane is more effective than carbon monoxide. When CO is dissociated, oxygen is added to the system. So some of the beneficial effect due to the carbon is negated by the additional oxygen. Since carbon condenses easily from the gas phase it is more likely to be in the quench reg ion than oxygen. As a result, the conversion of the alumia is-.increased despite the extra oxygen. Methane,, however., provides carbon without1 the_ inrdcto f xgn Frhrmrth td

-75of CO and CH4 as quench gases indicated that methane dissociated more easily than carbon monoxide. 10.9 Quench Methods The alumina conversions reported in Sections 10.6 through 10.8 were based upon the amount of aluminum and alumina collected on the reactor wall, as it was not attempted to recover all the product from the plasma. Water-cooled quench probes and different quench gases were then used in separate experiments to show that the recovery of product could be augmented while maintaining or improving the conversions obtained in the previous studies under the same operating conditions. Argon plasmas, with the same flow rate as in Sections 10.6 through 10., a power input of 5.03 kw and 500 mesh alumina were used for the quench studies. The conversions based upon the material collected on watercooled quench probes are compared with the conversions obtained with the product collected on the reactor wall in Table VI. It is clear that the quench probes can be used not only to recover additional product but also to increase the conversion. The results obtained with the two larger quench probes can be explained by the discussion in Section 5.2. The probes were placed directly in the plasma., where the aluminum concentration and temperature are the highest. Therefore., the molar flux of Al to the quench probe surface is greater than the flux to the reactor wall. In addition, the distance through which the Al must diffuse to be quenched iz less for a probe in the plasma. The result is

-76The result for the smallest probe do not contradict the above argument. With the 0o.148 in O.D. probe, the alumina flow rate was sufficiently great that the vaporization of the oxide was fairly incomplete. Therefore, the probe was in contact with a much larger number of molten alumina particles than the other probes. These molten particles collect on the probe and build up a layer of solids that inhibits the quenching of aluminum (see Section 10.6). The reactor wall, however, was not in as intimate a contact with the unvaporized alumina as the probe was and so the conversions in the two cases were about the same. TABLE VI CONVERSIONS OBTAINED WITH QUENCH PROBES Quench 500 Mesh Percent Conversion Probe Alumina Flow O.D. (gm/min) Reactor Quench (in) Wall Probe 0.148 0.47 3.1 2.9 0.250 0.22 7.1 20 00375 0.21 7 7 16 The quench gases (H2, CO, and CH4) were introduced into the lower section of the plasma core countercurrent to the plasma flow. The purpose was to see what effect these gases would have upon the conversion of the product collected on the re actor wall. The material obtained on the probes., through which the gases were fed, was not analyzed. The results obtained with four different hydrogen flow rates

-77for zero hydrogen flow in Figure 15. The amount of hydrogen used was not important and the use of any hydrogen as a quench increased the conversion only slightly. TABLE VII EFFECT OF HYDROGEN-QUENCH FLOW RATE Hydrogen 500 Mesh Percent Flow Rate Alumina Conversion (gmol/min) (gm/min) of Alumina 0.13 0.58 3.0 0.21 0.17 10 0.39 0.27 6~7 0.58 0.17 10 The conversions achieved with CO and CH4 as quench gases are compared in Table VIII. In both cases, carbon was deposited on the reactor wall., but considerably more was obtained with methane~ However, both CO and CH4 produced about the same percentage improvement in the conversion. TABLE VIII EFFECT OF. CO AND CH4 AS QUENCH GASES Quench Quench 500 Mesh Percent gas Gas Flow Alumina Conversion (gmol/min) Flow of Alumina (gm/min) CO oo0.10 0.15 26 OoO 0.15 10

11 SYSTEM: 10 \ \ ~ ARGON PLASMA POWER INPUT = 5.03 KW PARTICLE SIZE 500 MESH (26) 9 8 4: _,,..... 0 J _Aw W

-79The effect of any quench gas is to lower the temperature of the aluminum (see Section 10.10). But the diminished aluminum diffusivity is counteracted by the presence of a reducing atmosphere to react with some of the oxygen. This decreases the formation of AlO and A120 during the quench and more Al is recovered. It was shown in Section 5.1 that carbon has a much greater affinity for oxygen than hydrogen does. So the more effective quenching by CO and CH4 in comparison with H2 is reasonable. 10.10 Argon and Aluminum Temperatures Argon temperatures were determined at the three power levels of Section 10.7. The temperatures, presented in Table IX, are averages of the values obtained using the 4158.59 and 4259.36 R ArI lines. The enthalpies of atmospheric argon plasmas at these temperatures are also given. It is seen that while the argon temperatures increased only 3 percent, the enthalpy became 10 percent greater. TABLE IX ARGON TEMPERATURE AND ENTHALPY Power Argon Plasma 28 Input Temperature Enthalpy( 28) (kw) (~K) (cal/gm) 5.03 10,900 1900 5.86 11,100 2030 6.69 11,200 2100

-80Aluminum temperatures, which were obtained for most of the experimental runs of this work, are given in Table X. The temperatures are generally low and inconsistent. For example, the aluminum temperature should increase as the alumina flow rate decreases. The results for 500 mesh alumina do not exhibit this behavior throughout the range of flow rates studied. The temperatures obtained for the 00 and 320 mesh oxide, however, are consistent. The poor quality of the aluminum temperaturesis due to the nature of the silver images obtained on the spectrographic plates. In most cases, the images of the 39 03 and 396153 Al lines were so dense that complete lineprofiles were not obtained. So in order to determine the intensity of the lines, the missin portions had to be estimated (see Appendix VI). The inaccuracies inherent in the "multi-line" method (Section 6,) also add to the deviations. Figure 16 shows a typical plot of en(I/jigA) versus E for determination of the aluminum temperatures. Despite the poor results., it is possible to derive some qualitative information from the aluminum temperatures. The use of a quench gas lowers the temperature of the aluminum as does the use of He or H2 in the plasma. Higher aluminum temperatures are obtained at greater power input levels and lower alumina flow rates. Any other conclusions would be conjecture. 10.11 Solution of Heat,_Mass and Momentum Balances The computer solutions of heat, mass and momentum balance equations for aluminum oxide particles in an argon plasma are compared

-81TABLE X ALUMINUM TEMPERATURES Alumina Aluminum Flow Rate Temperature Remarks (gm/min) (OK) 0.58 2000 H2 quench, 0.1 gmol/min 0.27 3000 H2 quench, 0.27 gmol/min 0.17 2800 H2 quench, 0.58 gmol/min 0.15 3200 CO quench, 0.10 gmol/min 0.15 2600 He in plasma, 5.33 kw 0.17 3200 H2 in plasma, 5.53 kw 0.49 3000 0.37 4400 0.23 4100 0.23 5200 5.86 kw 0.23 5200 6.69 kw 0.21 3500 0.16 5900 5.58 kw, complete profiles used o.o6 3900 0.33 4200 400 mesh 0.31 5100 320 mesh, complete profiles used 0.05 6100 3~20 mes'h, comnplte

-13.5 -14.0 -14.5 PARTICLE SIZE = 500 MESH (26k) ALUMINA FLOW = 0.23 GM/MIN POWER INPUT =5.86 KW -15.0 -15.5 T 5200'K.-16.0 -16.5t 3.0 3.2 3.4 3.6 3.8 4.0 UPPER ENERGY LEVELl El ev Figure 16. Determination of Aluminum Temperature by the "Multi-Line" Method.

-83solved for various particle sizes and for 500 mesh alumina at higher power levels. The comparison is made with the results from low solids flow rates because at higher flow rates, the particles begin to interfere withheat transfer to each other. The assumption that the particle surfaces were uniformly accessible for heat and mass transfer was used in the derivation of the equations. The computer solutions estimate the amount of alumina that will vaporize. This is reported as percent vaporized, which is 100 times the volume vaporized divided by the initial volume for a spherical particle. The conversions predicted for 320, 400 and 500 mesh alumina at a flow rate of o.06 gm/min (taken from Figure 13) are compared with the percents vaporized, determined for the same plasma conditions, in Figure 17 Since the two curves have the same shape,. it appears that a constant fraction of the vaporized aluminum is recovered. It can be concluded that the conversion of alumina increases as more is vaporized. Since this is true for an argon plasma with collection of aluminum on the reactor wall, it has been proven that the recombination of aluminum and oxygen during the quench is less important than the extent of vaporization of the alumina. Apparently,, the condensation of aluminum occurs more rapidly than the diffusion of oxygen in the same direction so that once the Al is cooled sufficiently, it leaves the oxygen behind. The percents of conversion and vaporization are also presented in Table XI for the three particle sizes. The table includes the results obtained with 60 mesh (250Vu) -alumina and those obtained by Gross, et a.(7 with 200 mesh (74/A.) alumina. This comparison further demonstrates the

-84- ~ ~ ~ 60 i 55 SYSTEM: 50 ARGON PLASMA POWER INPUT = 5,.03 KW ALUMINA FLOW RATE = 0,06 GM/MIN 45........ z 0 0 N > 35........ 0 z 0 "~ -25 30...... 0 25 w Lu 0 20 15 10.......... 0 = PERCENT CONVERSION 5 25 30 35 40 145 PARTICLE DIAMETER, Figure 17. Comparison of the Percents of Conversion and Vaporization Obtained with 26k, 37~ and 4t5[ A1203 Particles.

diameter. It explains the poor results obtained by Grosse and his associates; they simply did not produce enough gaseous Al. The effect of the particle diameter also indicates that much higher conversions than those obtained in this work are possible with smaller particles. For example, the computer solution for 5/ aluminum oxide predicts complete vaporization. TABLE XI PERCENTS CONVERSION AND VAPORIZATION FOR VARIOUS PARTICLE SIZES Particle Percent Percent Size Conversion Vaporization (Mesh) ) 500 26 31 54.6 400 37 14 27.7 320 45 9.0 17.6 200 74 0.2-1.25(47) 1.86 6o 250 0.0 0.0 TABLE XII PERCENTS CONVERSION AND VAPORIZATION AT VARIOUS ARGON TEMPERATURES Argon Percent Percent Temperature Conversion Vaporization (oK) 10,900 7.1 54.6 11,100 14 59.7 11,200 16 62.1

-86The percent vaporization for 500 mesh alumina with the three power levels of Section 10.7 is compared with the experimental conversion in Table XII. (The argon temperature instead of the power input is given.) The increased vaporization due to the higher plasma temperatures and energy contents does not alone explain the size of the conversion increments. Because the aluminum temperatures are also higher in the greater energy plasmas (see Section 10.10), the combination of the two effects can account for the increases in the conversion, as explained in Section 10.7. Residence times calculated for the different particles are given in Table XIII. These times are less than those reported by (47) Grosse, et al., as they did not consider vaporization, which decreases the residence time. Since the smaller particles spend less time in the plasma and yet are more completely vaporized, the particle diameter is a more important variable than the residence time. TABLE XIII RESIDENCE TIMES FOR ALUMINA PARTICLES Particle Residence Size Time (Mesh) (Millisecond) 500 2.19 400 2.57 320 2.88 200 3.94 60 9.14

-87Previous investigations of heat transfer to solids in plasmas produced heat transfer coefficients of 87-162 and 4000 BTU/hr-ft -OF for heat transfer to thermocouple tips (0.02 in O.D.) and small copper slugs respectively (see Sections 2.4 and 3-3). The heat transfer coefficients for this study were calculated with Equation (5.13). The plasma properties used in this correlation were those corresponding to the reference temperature (see Section 5-3), which increased as the solid surface temperature:rose. So the heat transfer coefficient continuously increased until the vaporization temperature was reached. At this point the reference temperature remained constant, but the particle diameter began to shrink and the heat transfer coefficient continued to rise. So in order to compare the calculated coefficients with those given above, the values obtained with the particle at its vaporization temperature, but before vaporization has commenced, are assumed to be representative. These coefficients are presented in Table XIV for several particle diameters. TABLE XIV CALCULATED HEAT TRANSFER COEFFICIENTS FOR SMALL ALUMINA PARTICLES Particle Heat Transfer Diameter Coefficient (mesh) (y) (BTU/hr-ft -~F) 2500 5 23,300 500 26 4,48o 400 37 3,150 320 45 2, 59o 200 74 1, 570 60 250 370

-88The coefficients calculated for the 320, 400 and 500 mesh particles are on a par with those reported for the small copper slugs. The value obtained for 60 mesh (0.01 inch) agrees well with the 87-162 BTU/hr-ft 2-F range for the 0.02 inch O.D. thermocouple tips. It is concluded that the heat transfer coefficients used in this study are reasonable. Since the calculated particle residence times are also quite acceptable, the predicted vaporizations should be fairly accurate.

11. SUMMARY AND CONCLUSIONS The reduction of aluminum oxide to aluminum in an atmospheric, induction-coupled argon plasma flowing at 52 gm/min was studied experimentally. The conversion of the alumina was based upon the amount of A1203 and Al collected on the reactor wall. The variation of conversion with alumina particle size, alumina mass flow rate and the power input to the plasma was determined. Particle sizes of 500 mesh (268X), 400 mesh (37,1) and 320 mesh (45,/); flow rates of 0.03 to 0.6 gm/min; and power inputs of 5.03, 5.86 and 6.69 kw were used. These power levels correspond to argon temperatures of 10,9000, 11,100~ and 11,2000K respectively. With argon plasmas, the conversions (ranging from 3 to 30 percent) were generally found to increase with decreasing alumina flow rate and particle size and with increasing power input. The improvements in the conversion were due to increases in the percent of the oxide vaporized. The results agree qualitatively with the amount of alumina vaporization predicted by computer solutions to heat, mass and momentum balances for the oxide. The conversions obtained can be explained on the basis of a diffusion controlled quench of aluminum atoms. The use of water-cooled probes placed directly in the plasma allowed the recovery of additional aluminum at higher conversions. This indicates that with a reactor designed to collect a maximum amount of product, the conversions obtained in this work will be maintained or even improved. It was also possible to enhance the conversion by using CO and CH4 in the plasma with the argon and as quench gases introduced into the lower section of the plasma core countercurrent to the -89

-90plasma flow. Doubling and quadrupling of the conversion was obtained in this way. The use of H2 in each application was of little benefit. The relative effect of C and H2 as oxygen scavengers explains the results. Wet chemical and x-ray diffraction analyses of the reduction product from Ar and Ar-CO plasmas identified Al,,.-A1203 and t-A1203. Similar analyses of product from an Ar-CH4 plasma indicated that Al and A14C3 were present in addition to the oxides. No solid aluminum suboxides or aluminum oxycarbides were found. It was determined that A1203 vaporizes in an argon plasma to A10 and 0. The A10 in turn quickly dissociates to Al and O. Recombination of Al and 0 during the quench was not as important a consideration as the percent of the A1203 vaporized. Mean aluminum temperatures ranging from 2000 to 6000~K were determined for various operating conditions. The results were generally low and inconsistent. However, it was ascertained that the Al temperature generally increased with increasing power input and decreasing alumina flow rate. The reduction of aluminum oxide with appreciable conversion has been demonstrated to be quite feasible in an induction-coupled plasma reactor. It should be possible, on the basis of the results of this study, to achieve equal success in the reduction of other metallic oxides that are as stable as alumina.

APPENDIX I DERIVATION OF MOLAR FLUX OF Al IN QUENCH ZONE When steep concentration and temperature gradients are involved, the radial diffusion of component i in a mixture is given (8,35) by: Nit = CDimkT r + Nir - CDimk a+T r j +Xi Z Njr (I.) where Nir = radial molar flux of i, Xi = mole fraction of i, Dim = effective binary diffusivity for diffusion of i in a mixture, kT = thermal diffusion ratio (DT/Dim), where DT is the thermal diffusivity, C = total concentration, T = absolute temperature. The three terms on the right represent the ordinary diffusion, thermal diffusion and bulk flow contributions respectively. For low concentrations of species,i (as is encountered for Al in this work), kT is given by: kT = bX- (1-2) where the constant b does not exceed 0.2 through 0.3. The value of kT is also far less for gas pairs with close molecular weights. So despite the steep temperature gradients, thermal diffusion can be -91

neglected. Deleting the thermal diffusion term and rearranging Equation (I.1) gives: C Dim xi X n N = - + X Njr. (I.3) i-X r 1-X j= j.#i Consider the condensation of component i from a mixture when the other gases are noncondensable. Since these other gases do not condense, the term in Equation (1.3) due to their bulk flow is neglected: C Dim $Xi Nir = - (I.4) l-Xi $r This is the equation used in Section 5.2 to describe the condensation of Al from an argon plasma.

APPENDIX II DERIVATION OF HEAT, MASS AND MOMENTUM BALANCE EQUATIONS FOR AN ALUMINA PARTICLE The heat, mass and momentum balance equations, which were solved to give an approximate description of an alumina particle as it passed through an argon plasma, are derived in this section. The validity of the assumptions used in these derivations is discussed in Section 5.3. The velocity and axial position of the particle at a given time are governed by the momentum balance M d2z( F + Fg - Fb) g0 (II.1) dt2 where Ms = mass of a solid particle, Fg = gravitational force on the particle, Fb = buoyancy force due to the gas, Fd = drag force exerted on the particle. Assuming a spherical particle, Equation (II.1) becomes: Idf3 d2z 1 2 P (dz (11.2) d3S s2 s dt2 = + CD ds dt v + 6 d3 A g(l - He 6sPs Since A/Ps << 1, this simplifies to: dtd d Z' = + ~3 CD df (dz _ v (II.3) dt~ s/s

-94where - = velocity of solid particle, dt =s = density of solid particle, ds = diameter of solid particle, p = density of bulk plasma, vs = velocity of bulk plasma, g = gravitational acceleration, CD = drag coefficient. Thus the two forces acting on the particle are gravity and drag. Magnetic drag and slip flow effects were negligible. As a particle is heated in a plasma, energy is transferred to the particle primarily by conduction and convection and is lost from the particle to the surroundings by radiation. In this simplified model, the particles are at a uniform temperature (thermal conduction in the particle was not considered) and vaporization does not commence until the vaporization temperature is reached. The steady state heat balance for particles being heated to the vaporization temperature is: N 6ds Cps(TsTref)] - [N fls d3 Cps (TsTref)] + h( )(T-Ts) A Az - es(S) T4 A hz = O, (II.4) s, A~z -o V TsA~z=O, where Cps = heat capacity of solid, Ts = temperature of solid, T = temperature of plasma, z = axial distance along the reactor, h = conductive and convective heat transfer coefficient, d = Stefan-Boltzmann constant,

-95es = emissivity of solid, N = flow rate of particles (particles per unit time), A = plasma cross-sectional area, S/V = particle surface area per unit volume of plasma, Tref = reference temperature for particle enthalpy. In the limit, Equation (II.4) becomes: - [4N fs d Cp (Tsre )] + h(V)(T-T )A z ps S Ps ST ref V s - Tes(5)T4 A = O. (II.5) The quantity S/V is given by S NT ds V A 6) where vs = velocity of solid, (dz/dt). Combining Equations (11,5) and (II.6) and using the assumptions of constant particle flow rate and diameter as well as average heat capacity and density for the solid yields: dT h(T-T5) -'e T 4 Ps d s C s + 5 _ S_ = o (II.7) -~ PS dS ps dz vs vs And since vs = dz/dt: 1 5d5 dTs 4 6 s ds Cs dt = h(T-T) -e T (II 8) This equation gives the time-temperature history of the particle as it is heated to its vaporization temperature. When the alumina particle reaches the vaporization temperature (where it essentially dissociates to Al and 0), its heat balance is:

-96h (T-T)A - e s TA = Hr rd A AL rd A, (II.9) where /Hr = heat of reaction for dissociation to Al and 0, rd = rate of dissociation. It is assumed that the vaporization of the particle proceeds uniformly so that the particle remains spherical. The lops of mass by dissociation is described by: [N s 6 ds] [NPs N d]+ = rd A (II.10) In the limit, this becomes: -N Ps'r 2 d(ds) rd A (11.11) It follows from Equations (II.6), (II.9) and (II.11) that: 1 d(ds) 4 12 a ~s dt = h(T-T) - -e= T4 (II.12) 2 r ss dt This equation, along with Equation (II.3), gives the extent of vaporization of the particle at a given axial position in the plasma.

APPENDIX III CALIBRATION OF SAl PLATES II1.1 Background When a spectrographic plate is exposed to the radiation from a plasma, a silver image is deposited on the plate for each spectral line. The intensity of a given line can be determined from the density of its image. Since the amount of silver deposited is a non-linear function of both line intensity and wavelength, a plate (or emulsion) calibration is necessary before the intensity of a line can be determined. The transmittance or reflectance of light due to the optical components of the scanning system and the spectrograph diffraction grating are also functions of wavelength. This effect can be included in the calibration. III.2 Calibration Procedure The combined calibration was done with the use of a tungsten ribbon filament lamp, which was a source with a spectral intensity that was known as a function of wavelength. Ideally, the lamp should be positioned so that its rays would follow the same optical path as radiation from the plasma and so that the spectrograph slit would subtend the same solid angle of radiation from either source. That is, the lamp should be mounted at the reactor site. For the wavelengths used in this study, however,the tungsten continuum intensity was too weak to give' silver images dense enough for the calibration to cover spectral lines of normal intensity. Therefore, it was necessary to mount the tungsten lamp directly on the optical bench (past all of -97

-98the mirrors and one lens). Then the effect of the rest of the optical path (from the reactor site to the point on the optical bench) was determined with the direct read-out system, which could be used to greatly amplify the signals obtained at the reactor site. This was done by comparing the signals obtained at the two places. A General Electric Company, 6 volt, 108 watt lamp was used. Power was supplied to the lamp by a 120 volt constant voltage transformer (Variac) hooked up to a 117 to 6 volt step down transformer. The apparent temperature of the tungsten ribbon was measured with a Leeds and Northrup disappearing filament optical pyrometer (Cat. No. 8622) which had been calibrated by the National Bureau of Standards. The true temperature was obtained from the apparent temperature by taking into (25) account the emissivity of the tungsten ribbon, the transmittance (22) of the pyrex lamp envelope, an estimated reflectance and the N.B.S. (73) pyrometer correction tables. The temperature was uniform over the central part of the ribbon and this section was much larger than the area focused upon the spectrograph slit. The corrected temperature and the emissivity of the tungsten ribbon were then used in Planck's radiation law to calculate the intensity of the radiation emitted at a given wavelength. The actual calibration was done by exposing the plate to the tungsten continuum for several time intervals and to radiation from a plasma containing aluminum. The latter was done so that the wavelength at a point on the tungsten continuum exposure could be determined. It was assumed that all exposures having the same intensity-time product, or relative exposure (RE), would produce equal silver densities. This

-99is known as intensity-time reciprocity and holds for SA1 plates when exposure times of 5 to 300 sec are used.(9) Second order radiation dispersed by the diffraction grating did not interferein the 3000-4000 A region that was studied since the tungsten ribbon had negligible radiation below 3000 A. In fact, the tungsten radiation was weak enough at 3082.15 R and 3092.71 X that complete calibration curves could not be obtained even with the lamp mounted on the optical bench. So the calibration was repeated at these two wavelengths using the continuous spectrum of a hydrogen lamp from a Beckman DU spectrophotometer. The hydrogen lamp was also mounted on the optical bench. Since the hydrogen continuum intensity was not known, the calibration curves were matched with those obtained from the (76) tungsten lamp, as suggested by Quarderer, to give complete calibration curves. The calibration curves are plots of relative exposure versus silver density (SD), which is the negative of the common logarithm of the fractional light transmittance through the deposit. The silver densities were measured with a recording microphotometer. The emulsion calibration curves, known as Hurter-Driffield characteristic curves, are given in Figure 18 for the lines used for the aluminum temperature determinations. IIio3 Determination of Absolute Line Intensity As discussed above, the SA1 plates were calibrated with the tungsten lamp mounted on the optical bench. Then the additional transmittance due to the optical system between that point on the bench and the reactor site was determined using direct read-out. The signal

2.8 2.6 2.4 2.2 2.0 1.8 4Z 3944-& 3961A 1.6, 1.4 -- H 1.2 - - i I 00 "' 1.0 LI)/ 0,8 3o82 & 3 09 0.6 0.4 0.2 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0 20.0 30.0 40.0 50.0 600 70.080.0900)100.0 RELATIVE EXPOSURE, RE Figure 18. Emulsion Calibration Curves for SA1 Plates.

-101received by the spectrograph with the lamp on the optical bench is: Isl = I (IIIw ) and the signal from the reactor site is: Is2 Iw l1UZ22) where Iw = intensity of tungsten ribbon, r'1 = transmittance of optical system from point on bench to the plate (or phototube), =2 = transmittance of optical system from reactor site to point on beach. It follows that 52 is given by: IU2 = 1s2/Is1 (III.3) The transmittance 1 was included in the relative exposures used for the calibration curves (Figure 18), so that RE = REw/' (III.4) where RE = relative exposure used in calibration, REw= relative exposure due to tungsten lamp on optical bench. When the plate is exposed to plasma radiation, the relative exposure due to a spectral line is: RE = IA'l'2 ETZ:J r (III.5) where IQ = relative intensity of spectral line at a given wavelength, ET = exposure time,

-1027s = transmittance of solids that have collected on the reactor wall, Zr = transmittance of clear reactor, i.e. quartz tube, cooling water and pyrex jacket. If Figure 18 is entered with SD at a point on the spectral line image, then RE~/ is obtained. That is: RE, = -RE. (III.6) It follows that the relative intensity of the spectral line is given by: Ida = RE/f2' r r ET (III. 7) This intensity is a function of the wavelength near the spectral line and the absolute intensity of the line is: 00 In practice, the integration need not be carried out to infinity as it can be terminated quite near the line center. The integration is discussed in detail in Appendix VI. The transmittances Zs and Zr are functions of wavelength and are discussed in Appendix IV.

APPENDIX IV DETERMINATION OF TRANSMITTANCES OF SOLIDS AND REACTOR The transmittances ts and t, as defined in Appendix III, S r were determined with the use of the direct read-out system of the spectrograph. With the tungsten lamp mounted on the optical bench} the signals received by the spectrograph with and without the reactor in front of the lamp were compared. This was done with a clear reactor and with various amounts of solids build-up. Since the tungsten radiation passes through the entire reactor and the plasma through only half, this procedure gives 2 and Z2 As in Appendix III (for the transmittance of the optical system) we have: 2 "r = Is/I (IV. 1) r r S and = Iss/Is (IV.2) where ~s = transmittance of solids on reactor wall (in addition to that due to the reactor), ~r = transmittance due to clear reactor, Is signal received through clear reactor, Is = signal received through reactor with solids build-up. s The absolute value of Or is easy to determine, but it is impossible to know what s' is at a given time during an experimental run because the solids are continuously building up on the reactor wall. -103

-104Fortunately, for the aluminum lines used in this work, the ratios's(>A2)/Js(1)' Vs({3)/Zs(~1) and'Cs ( 4)/ s Cl) were found to be independent of solids build-up. The symbol 5s(Mi) refers to the value of Zs at the wavelength >i. This result can be used to advantage since the "multi-line" method, used for determining aluminum temperatures, needs only the relative values of the absolute intensities for the lines. That is, the ratios I(~2)/I(1l), I(>3)/I(>1), I(X4)/ I(x1) and 1 can be used just as well as I(>1), I(~12), I(/3) and I(~4) for the determination. This follows from the fact that T is related to ~n(I/QgA) (See Section 6.4). Therefore, relative values of Z s and t r were used for the aluminum temperature calculations. s ~ r

APPENDIX V CALIBRATION OF DIRECT READ-OUT SYSTEM In order to determine absolute line intensities from the direct read-out of the spectrograph, the photomultiplier tube and optical system had to be calibrated. As for the plate calibration, a 6 volt, 108 watt tungsten filament lamp served as a standard source whose radiant intensity could be calculated. This lamp was mounted at the site of the plasma reactor so that its rays followed the same path as those from the plasma and so that the spectrographic slit subtended the same solid angle of radiation from either source. Then the current output of the photomultiplier tube, at the wavelengths of interest, was compared with the radiant intensity of the tungsten filament. The ratio of the intensity to the current is the spectral response function, S(X), of the multiplier phototube, the electronic circuit and the optical system. The spectral response was a function only of wavelength in this study since the optical system was left unchanged. Calculation of the radiant intensity of the tungsten filament is discussed in Appendix III.2. -105

APPENDIX VI CALCULATION OF THE ABSOLUTE INTENSITY OF A SPECTRAL LINE VI.o General The absolute intensity of a spectral line, I, is the integral of the relative intensity, IA, over the wavelength interval that contains the line. That is, I f L= dT 1, X(VI.1) where ~o = the wavelength at the center of the spectral line, = the interval to either side of the line center that contains the particular half of the line profile, i.e. half the width of the base of the profile, IA = relative intensity of the line above the continuum intensity. The integration should ideally be carried out to infinity but in practicality it is terminated when the relative intensity of the line is indistinguishable from the continuum intensity. VI.2 Integration of a Spectral Line on a Photographic Plate When the silver image, which a spectral line caused to be deposited on a spectrographic plate, was scanned with a recording microphotometer a record of silver density versus wavelength was obtained. This was converted to relative intensity above the continuum versus wavelength by referring to the plate calibration curves. Integration of this profile gave the absolute intensity of the spectral line. -106

-107Two different situations were encountered in this work. In the first case, a complete profile of relative intensity versus wavelength was obtained. This occurred for all the 3082.15 and 3092.71 a A11 lines and the less intense 3944.03 and 3961.53 a A1 lines. These curves were divided into several intervals of equal wavelength and the integration over each segment performed by applying Simpson's Rule. The conversion from silver density to relative intensity and the subsequent integrations were done numerically by an IBM 7090 computer. In the second situation, an incomplete line profile was obtained. This occurred when the spectral line was so intense that the silver density could not be determined near the line center. Such was the case for most of the 3944.03 and 3961.53 A A,1 profiles obtained, despite the use of exposure times of 3 to 5 seconds. When shorter exposures were made, the 3082.15 and 3092.71 A Al1 lines were not recorded on the plates. Since isolated lines of heavier elements have approximate (57) profiles of the Lorentzian shape out into the far wings, it was possible to complete the missing portion of the line profile. The intensities of a known portion of the curve were fitted to a Lorentzian distribution: I(A = [ (/ /2)1 (VI.2) where L~ = the distance from line center, ~A>1/2 = the (half) half-width of the line, by the method of least squares. Parameters for the curve fitting were

the intensity of the line at the center of the profile, the continuum intensity and the (half) half-width. The data from the wings of the known curve were not used. A typical result of this analysis is given in Figure 19 for a 3944.03 A,1 line. VI.3 Integration of Photomultiplier Tube Output The current output of the photomultiplier tube was passed across a known precision resistor. The resulting voltage was amplified and integrated over the wavelength interval containing the spectral line of interest. The integration was performed directly by an operational amplifier. This integral was then converted to the absolute intensity of the spectral line by multiplying by the spectral response of the direct read-out system at the wavelength of the line.

-109180 - -. - 160 -... —.... U,, 120......... — _ _ 0) U 4 2 4-; U 1 120... I - 100......... L a) "" 80 -, z w 60 _ w cn 8 0 40 0.. < 6~0 2 0 3944. 3A -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 DISTANCE FROM LINE CENTER, AX, 0.0978A Figure 19. Completion of the Missing Portion of the Intensity Profile of a 3944.03 A AlT Line by a Least Squares Fit to a Lorentzian Profile.

APPENDIX VII SOLUTIONS USED IN VOLUMETRIC DETERMINATION OF ALUMINUM VII.1 General The solutions used in the volumetric analysis for aluminum are divided into two groups: those required for the precipitation and dissolution of aluminum quinolate; and those used to determine the resultant 8-quinolinol. The solutions were prepared by dissolving a known weight of substance in a known volume of liquid, unless otherwise stated. All weighings of less than 200 gm (including container) were done on a RIGIHT-A-WEIGH balance which is accurate to + 0.2 mg. Those of over 200 gm were done on a Mettler (Tara 0-2000 gm) balance accurate to + 0.5 gm. VII.2 Formation and Dissolution of Aluminum Quinolate 1. 8-Quinolinol (8-hydroxyquinoline; oxine): A 5 wt percent solution of 8-quinolinol in 2M acetic acid. The exact concentration is not important. 2. Tartaric Acid: 10 mg acid/mil solution. 3. Ammonium Acetate: 0.4 gm/ml solution. 4. Bromocresol Purple: 0.04 percent in ethanol. VII.3 Determination of 8-Quinol.inol 1. Bromate-Bromide Solution: 0.1000 N KBrO3 + 100 gm KBr/ liter solution. Special care was taken in the preparation of this solution because the KBrO3 was used as a primary standard. The KBrO3(M.W. = 167.01) used was 100 percent -110

-111pure, so 0.1 N KBrO3 is 16.701/6 or 2.7835 gm/l. The bromate was weighed in a crucible, dried at 180~C and 1-1/2 hours, cooled and reweighed.(63) The amount of KBrO3 used for a liter of the solution was 2.7840 gm. The exact KBr concentration is unimportant. 2. Potassium Iodide: 20 wt percent KI in water. 3. Thiosulfate: 0.1 N Na2S203 solution. Special care was also taken in the preparation of this solution. The thiosulfate was dissolved in freshly boiled (to remove C02) distilled water and 0.1 gm Na2CO3/liter was added as preservative. (63) Reagent grade (99.5 percent) Na2S203j5H20 was used and the normality determined with the primary standard, KBr03. The thiosulfate normality was checked periodically. 4. Starch Solution: 0.2 percent solution. After suspending 4 gm powdered starch in 600 ml distilled water, 20 percent KOH was added until a thick, almost clear solution was obtained. This was allowed to stand one hour for a complete reaction and then was neutralized to the litmus end point. Glacial acetic acid (2 ml) was added as a preservative and the solution was diluted to 2 liters.

NOMENCLATURE Symbol 2 A are a - cm Anm Einstein transition probability for spontaneous transition from level n to level m - sec-1 C total concentration - gmol/cm3 CD drag coefficient p Cp heat capacity - cal/gm- ~K d diameter - cm, d spacing between crystal planes Dim effective binary diffusivity for diffusion of i in a mixture - cm2/sec DT Thermal diffusivity - cm2/sec e emissivity E Energy - erg ET exposure time - sec Fb buoyant force - dyne Fd drag force - dyne Fg gravitational force - dyne g gravitation acceleration - cm/sec2 g statistical weight h heat transfer coefficient - cal/sec-cm2-~C h Planck's constant - erg-sec h specific enthalpy - cal/gm href. specific enthalpy at reference temperature in film around a solid particle - cal/gm -112

-113NOMENCLATURE (Continued) Symbol AH~r heat of reaction - cal/gm I absolute intensity of radiation - erg/cm2-sec-ster Is signal received by spectrograph k Boltzmann constant - erg/~K k thermal conductivity - cal/sec-cm- ~C kT thermal diffusion ratio L path length or source thickness - cm M mass - gm N flow rate of particles - particles/sec N number density - cm-3 2 Nir radial molar flux of i - gmol/cm2-sec Prandtl number Q electronic partition function r radial position - cm rd rate of dissociation - gm/cm3-sec Re Reynolds number RE relative exposure - erg/cm3-ster S surface area - cm s(s) spectral response of photomultiplier and optical systemerg/cm2- sec-amp- ster t time particle has been in plasma - sec T absolute temperature - ~K v velocity - cm/sec Vrel velocity of solid particle relative to the plasma - cm/sec

NOMENCLATURE (Continued) Symbol V volume - cm3 Xi mole fraction of i z axial position in plasma - cm ~A volume emission coefficient of spectral line - erg/cm3-sec-ster 20 angle between diffracted x-ray beam and transmitted beam - degrees A wavelength - cm, )XO wavelength at center of spectral line - cm, AX distance from line center - cm,, At B (half) base-width of spectral line - cm, X AVZ2 (half) half-width of spectral line - cm, frequency of emitted photon - secjP density - gm/cm3 Stefan-Boltzmann constant - cal/sec-cm2-OK -t~ transmittance Subscripts f film or boundary layer around a solid particle in plasma spectral line m upper energy state n lower energy state r clear reactor s solid particle w tungsten filament O ground level

-115NOMENCLATURE (Continued) Symbol 1 site of calibration on optical bench 2 site of plasma reactor ~0 bulk plasma

BIBLIOGRAPHY 1. Ackermann, R. J. and R. J. Thorn. "Gaseous Oxides of Aluminum, Tungsten and Tantalum," J. Am. Chem. Soc., 78, 4169 (1956). 2. Allen, William A., John S. Rinehart and W. C. White. "Phenomena Associated with the Flight of Ultra-Speed Pellets. Part I. Ballistics, " J. Appl. Phys., 23, 132-37 (1952). 3. Ayres, Gilbert H. Quantitative Chemical Analysis, Harper and Brothers, New York, 1958. 4. Beletskii, M. S. and M. B. Rapoport. Doklady Akad. Nauk S. S. R., 80, 751-4 (1951). 5. Bennett, Carl and Norman L. Franklin. Statistical Analysis in Chemistry and the Chemical Industry, John Wiley and Sons, Inc., New York, 1961. 6. Biggerstaff, G. E., W. R. Golliher, R. L. Harris and W. R. Rossmassler. "Plasma Torch Production of Elemental Boron," AEC R and D Report, KY-453, February 26, 1964. 7. Bildstein, H. "Melting Materials into Spheres with a Plasma Torch," Chem.-Ingr.-Tech., 38, No. 10, 19-25 (1966), see C. A. 64:12211 g. 8. Bird, R. Byron, Warren E. Stewart and Edwin N. Lightfoot. Transport Phenomena, John Wiley and Sons, Inc., New York, 1960. 9. Bowen, Stuart W. "A Spectroscopic Study of an Underexpanded Argon Plasma Jet," Ph.D. Thesis, The University of Michigan, 1966. 10. Brewer, Leo. "The Thermodynamic Properties of the Oxides and their Vaporization Processes," Chemical Reviews, 52, 1-75 (1953). 11-. Brewer, Leo and Alen W. Searcy. "The Gaseous Species of the Al-A1203 System," J. Am. Chem. Soc., 73, 5308-14 (1951). 12. Brown, R. A. S. "Reaction of Some Zirconium Compounds in a Plasma Jet,?" Presented at the CIMM, Conference of Metallurgists, Kingston, Ontario, August 28-30, (1967). 13. Buchanan, Robert S. "Study of a Seeded Plasma," Ph.D. Thesis, The University of Michigan, Ann Arbor, 1961. 14. Carnahan, Brice, H. A. Luther, James O. Wilkes. Preliminary Edition of Applied Numerical Methods, Vol. I, John Wiley and Sons, Inc., New York, 1964. -116

-11715. Chludzinski, George R. "Energy Transfer to Solids in R. F. Generated Plasmas," Ph.D. Thesis, The University of Michigan, Ann Arbor, 1964. 16. Christiansen, E. B. "Effect of Particle Shape on Free Settling Rates," Ph.D. Thesis, The University of Michigan, Ann Arbor, 1943. 17. Cochran, C. Norman. "Aluminum Suboxide Formed in Reaction of Aluminum with Alumina," J. Am. Chem. Soc., 77, 2190-1 (1955). 18. Coheur, F. P. and P. Coheur. Chem. Zentr., 1, 978 (1944). 19. Coheur, F. P. and P. Coheur. Rev. universelle mines, 19, 86-9 (1943). 20. Coheur, P. "Temperature Measurement by Means of Molecular Spectra, Bull. classe sci., Acad. Roy. Belg., 28, No. 7/9, 569-73 (1942). 21. Corliss, Charles H. and William R. Bozman. "Experimental Transition Probabilities for Spectral Lines of Seventy Elements," U.S. National Bureau of Standards, Monograph No. 53, 1962. 22. Corning Glass Works, "Glass Color Filter," Bulletin CF-1, Corning, New York, 1960. 23. Cullity, B. D. Elements of X-ray Diffraction, Addison-Wesley Publishing Company, Inc., Reading, Massachusetts, 1956. 24. De Maria, G., J. Drowart and M. G. Inghram. "Mass Spectrometric Study of A1203," J. Chem. Phys., 30, 318-19 (1959). 25. De Vos, J. C. "A New Determination of the Emissivity of Tungsten Ribbon," Physica, 20, 714 (1954). 26. Dickerman, P. J. and R. W. Deuel. "Measurement of Relative Transition Probabilities for Some Lines of CuI, AlI, and MoI Spectra," Journal of Quantitative Spectroscopy and Radiative Transfer, 4, No. 6, 807-17 (1964). 27. Dow Chemical Company. JANAF Thermochemical Tables, ARPA Proj. AF 04(611)-7554, Midland, Michigan (1965). 28. Drellishak, K. S., C. F. Knopp and Ali Bulent Cambel. "Partition Functions and Thermodynamic Properties of Argon Plasma," Technical Documentary Report No. AEDC-TDR-63-146, August 1963. 29. Eastman Kodak Company. "Kodak Plates and Films for Science and Industry," Kodak Publication No. P-9, Eastman Kodak Company, Rochester, N. Y., 1962. 30. Eckert, E. R. G. "Engineering Relations for Friction and Heat Transfer to Surfaces in High Velocity FlowT," Journal of the Aeronautical Sciences, 22, 585-587 (1955).

-11831. Elliott, John F. and Molly Gleiser. Thermochemistry for Steelmaking, Vol. 1, Addison-Wesley Publishing Company, Inc., Reading, Massa"chusetts, 1960. 32. Emlin, B. I., B. B. Papin and S. T. Rostovtsev. "The Mechanism of Electrocorundum Reduction by Solid Carbon," Met. i Koksokhim., Mezhvedomistv. Resp. Nauchn.-Tekhn. Sb., 3, 36-40 (1965). 33. Engelke, John L. "Heat Transfer to Particles in the Plasma Flame," Stanford Research Institute, Menlo Park, California, January 30, 1962. 34. Foster, L. M., G. Long and M. S. Hunter. "Reactions Between Aluminum Oxide and Carbon. The A1203-A14C3 Phase Diagram," J. Am. Ceram. Soc., 39, 1-11 (1956). 35. Frank-Kamenetskii, D. A. trans. N. Thon, Diffusion and Heat Exchange in Chemical Kinetics, Princeton University Press, Princeton, New Jersey, 1955. 36. Gastinger, E. "Preparation of Lower Oxides," Naturwissenschaften, 42, 95 (1955). 37. Gericke, W. E. "Messung der Ubergangswahrscheinlichkeit von AI-linien, " Z. Astrophys., 53, 68-79 (1961). 38. Gibbins, S. G. and B. Siegel. "Reduction of Aluminum Oxide by Propellant Combustion Gases at High Temperatures," Report No. TDR-269-(4210-10)-7, Contract No. AF 04(695)-269, September 1, 1964. 39. Ginsberg, H. and V. Sparwald. "Extraction of Aluminum by Carbothermic Reduction with Special Reference to the Aluminum-Carbon System, " Aluminiurn 41, No. 3, 181-93, No. 4, 219-30 (1965), 40. Gitleson, Gunnar, Oddvin Herstad and Ketil Motzfeldt. "Carbothermic Reduction of A1203," Selected Topics High Temperature Chemistry, 179-96 (1966), see C. A. 65:9819h. 41. Glassner, Alvin. "'The Thermochemical Properties of the Oxides, Fluorides, and Chlorides to 2500~K," Argonne National Laboratory Report No. ANL-5750, Contract W-31-109-eng-38 (1957). 42. Glasstone, Samuel, Keith J. Laidler and Henry Eyring. Ph.D., The Theory of Rate Processes, McGraw-Hill Book Company, Inc., New York 1941. 43. Goldsmith, Alexander, Thomas E. Waterman and Harry J. Hirschhorn. Handbook of Thermophysical Properties of Solid Materials, Vol. 3, The Macmillan Company, New York, 1961.

44. Goodlett, Vernon Wilson. "The Visible and Near Ultraviolet Spectra of the Aluminum Monoxide Molecule," Ph.D. Thesis, Vanderbilt University, August, 1959. 45. Goulden, Cyril H. Methods of Statistical Analysis, Second ed. John Wiley and Sons, Inc., New York, 1952. 46. Griem, Hans R. Plasma Spectroscopy, McGraw-Hill Book Company, New York, 1964. 47. Grosse, A. V., H. W. Leutner and C. S. Stokes. "Plasma Jet Chemistry," First Annual Report, Office of Naval Research, Contract 3085(02), Task No. NR 052-429, December 31, 1961. 48. Grossman, L. N. "Niobium-A1203 Reactions Yielding Condensed and Volatile Products," J. Chem. Phys.2 44, No. 11, 4127-31 (1966). 49. Grube, G., A. Schneider, V. Esch and M. Flad. "Zur Kenntnis des Aluminiumsuboxyds, " Z. anorg. Chem., 260, 120-6 (1949). 50. Harrison, George R. Massachusetts Institute of Technology: Wavelength Tables, John Wiley and Sons, Inc., New York, 1939. 51. Hasapis, A. A., M. B. Panish and C. Rosen. "The Vaporization and Physical Properties of Certain Refractorie, " WADD Technical Report 60-463, Part I, PB 171 413, October 1960. 52. Hedger, H. J., and A. R. Hall. "Preliminary Observations on the Use of the Induction-coupled plasma Torch for the Preparation of Spherical Powder," Powder Metallurgy, No. 8, 65-72, 1961. 53. Herzberg, Gerhard. Atomic Spectra and Atomic Structure, Dover Publications, New York, 1944. 54. Herzberg, Gerhard. Spectra of Diatomic Molecules, D. Van Nostrand Company, Inc., Princeton, New Jersey, 1950. 55. Hoch, Michael and Herrick L. Johnston. "Formation, Stability and Crystal Structure of the Solid Aluminum Suboxides: A120 and A10," J. Am. Chem. Soc., 76, 2560-1 (1954). 56. Hodgman, Charles D., Robert C. Weast, and Samuel M. Selby (editors). Handbook of Chemistry and Physics, 42nd edition, The Chemical Rubber Publishing Co., Cleveland, Ohio, 1960. 57. Huddlestone, Richard H. and Stanley L. Leonard (editors). Plasma Diagnostic Techniques, Academic Press, New York, 1965. 58. Huska, Paul A. and Curtis W. Clump. "Decomposition of Molybdenum Disulfide in an Induction-coupled Argon Plasma," I and EC Process Design and Development, 6, No. 2, 238-244 (1967).

-12059. Inghram, M. G., W. A. Chupka and J. Berkowitz. "Dissociation Energies from Thermodynamic Equilibria Studied with a Mass Spectrometer," Memoires Soc. Roy. Sci. Lie'ge, Ser. 4, 18, 513-35 (1957). 60. Kempthorne, Oscar. Design and Analysis of Experiments, John Wiley and Sons, Inc., New York, 1952. 61. Khodak, L. P. and V. S. Mal'tsev. Tr. Khim. Met. Inst., Akad. Nauk S. S. R., 1, 218-31 (1963). 62. Kodama, K. Methods of Quantitative Inorganic Analysis, Interscience Publishers, London, 1963. 63. Kolthoff, I. M. and R. Belcher. Volumetric Analysis, Vol. III, Interscience Publishers, Inc., New York, 1957. 64. Levenspiel, Octave. Chemical Reaction Engineering, John Wiley and Sons, Inc., New York, 1962. 65. Loo, S. L. and R. C. Dimick. "Interaction of Solid Particles with an Ionized Gas," Tech. Report ILL-12-p, Project Squid, Cont. Nonr - 3623(00) and NR-098-038, November, 1963. 66. Mal'tev, V. S., V. D. Ponomarev, V. T. Panyushkin and S. M. Isabaev. "The Mechanism of the Thermal Decomposition and Reduction of Sodium and Potassium Hydroaluminates," Tr. Inst. Met. i Qbogashch., Akad. Nauk Kaz. S. S. R., 12, 136-142 (1965). 67. Margrave, J. L. "Chemistry at High Temperatures," Science, 135, No. 3501, 345 (1962). 68. Marynowski, C. W. and A. G. Monroe. "R-F Generation of Thermal Plasmas, High Temperature Technology, Butttrworths, London, 1964. 69. McAlpine, Roy K. and Byron A. Soule. Fundamentals of Qualitative Chemical Analysis, 4th Edition, D. Van Nostrand Company, Inc., Princeton, New Jersey, 1956. 70. Meyer, H. "Fusion of Powders in a Plasma Jet," Ber. Dtsch. Keram. Ges., 41, No. 2, 112-119 (1964). 71. Motzfeldt, Ketil, Tek. Ukeblad., 109, 1137-42 (1962). 72. Pearse, R. W. B. and A. G. Gaydon. The Identification of Molecular Spectra, John Wiley and Sons, Inc., New York, 1963. 73. Poland, D. E., J. W. Green and J. L. Margrave. "Corrected Optical Pyrometer Readings, " NBS Monograph 30, U. S. Government Printing Office, April 21, 1961.

-12174. Ponomarev, V. D., V. T. Panyushkin and V. S. Mal'tsev. "Mechanism of Physicochemical Transformations in the Thermal Reduction with Carbon of Synthetic Nepheline," Vestn. Akad. Nauk Kaz. S. S. R., 21, No. 7, 32-5 (1965). 75. Porter, R. F., P. Schissel and M. G. Inghram. "Mass Spectrometric Study of Gaseous Species in the Al-A1203 System," J. Chem. PhysO, 23, 339-42 (1955). 76. Quarderer, George J. "Photochemical Chlorination of Sulfur Dioxide Utilizing a Plasma Light Source," Ph.D. Thesis, The University of Michigan, Ann Arbor, April 1967. 77. Reed, Thomas B. "Heat-transfer Intensity from Induction Plasma Flames and Oxyhydrogen Flames," J. Appl. Phys., 34, No. 8, 2266-9 (1963). 78. Rosen, B. Constantes Selectionnees Atlas, Hermann and C Depositaires, Paris, 1952. 79. Rosen, B. Constantes Selectionnees Donnees Spectoscopiques Concernant Les Molgcules Diatomiques, Hermann and Cle, Depositaires, Paris, 1951. 80. Seymour, Errol V. "Particle Injection Techniques for Velocity Measurement in Arc Plasmas," Technical Report No. 12, National Science Foundation Grant GP-414, Harvard University, December, 1964. 81. Sherman, Martin P. and Jerry Grey. "A Technical Report on Calculation of Transport Properties Mixtures of Helium and Partly-ionized Argon," Aeronautical Engineering Laboratory Report No. 673, Princeton University, N. J., December, 1963. 82. Smith, Dean L. "Mass and Energy Transfer Between a Confined Plasma Jet and a Gaseous Coolant," Ph.D. Thesis, The University of Michigan, February, 1965. 83. Smith, Delbert M. "Plasma Spraying of Refractory Materials," General Motors Engineering Journal, 23-28, Second Quarter 19639 84. Smith Joseph V. (editor-) "Index to the X-ray Powder Data File, " ASTM Special Technical Publication 48-L, ASTM, Philadelphia, Pa. 1962. 85. Stokes, Charles S. "Chemical Reactions with the Plasma Jet," Chemical Engineering, 191-196, April 12, 1965. 86. Stokes, C. S., J. A. Cahill, J. J. Correa and A. V. Grosse. "Plasma Jet Chemistry," Final Report, Air Force Office of Scientific Research, Contract AFOSR-62-196, December, 1964.

UNIVERSITY OF MICHIGAN II llll I lll lll 111111LII ll111111111 -1223 9015 03022 7394 87. Stokes, C. S., W. W. Knipe and L. A. Streng. "Heat Transfer Rates of an Argon Plasma," J. Electrochem. Soc., 107, 35-8 (1960). 88. Walas, Stanley M. Reaction Kinetics for Chemical Engineers, McGrawHill Book Company, Inc., New York, 1959. 89. Wood, Bernard J. and Henry Wise. "The Interaction of Atoms with Solid Surfaces, " PB Report 156, 822, U. S. Dept. Comm., Office Techo Serv. (1960). 90. Worrell, W. L. "Carbothermic Reduction of Alumina. A Thermodynamic Analysis," Can. Met.Quarterly, 4, No. 1, 87-95 (1965). 91. Zhadanova, L. V. and V. A. Sokolov. "Structure and Nature of Emission of the Spectrum of Aluminum Oxidation," Zh. Prikl. Spektroskopii, 1, No. 3, 272-5 (1964).