Electrowinning Magnesium Metal from Mgcl2-Ndocl Melt Using Solid-Oxide Oxygen-Ion-Conducting Membrane Technology

Electrowinning Magnesium Metal from MgCl2-NdOCl Melt Using Solid-Oxide Oxygen-Ion-Conducting Membrane Technology

David E. Woolley', Uday B. Pal2, and George B. Kenney3

1 Formerly Research Associate, Department of Manufacturing Engineering, Boston University, Boston, MA 02215 USA. Now at Saint-Gobain Ceramics and Plastics, Northboro, MA 01532 USA, email: david. e. woolley@saint-gobain. com 2 Professor, Department of Manufacturing Engineering, Boston University, Boston, MA 02215 USA, email: [email protected] 3ElMEx, LLC, Medfield, MA 02052 USA.

(Received October 31,2000)

ABSTRACT 1. INTRODUCTION

An electrolytic process for extraction of magnesium Magnesium is produced commercially by two metal from its oxide is demonstrated. The process is methods: electrolytic and metallothermic. The based on the chemical system MgCl2-NdCl3-MgO. electrolytic method uses a chloride-based melt such as

Since it has previously been shown that NdCl3 and MgO MgCl2-NaCl-KCl-CaCl2-CaF2, and the cell operating react to form NdOCl, this study has employed a MgCl2- temperature is approximately 750°C. The MgCl2 feed is NdOCI melt. Two experiments are described in detail. electrolyzed to produce magnesium metal at the steel

In the first, liquid magnesium metal is produced from a cathode and Cl2 gas at the carbon anode. If the feed

MgCl2-NdOCl melt at 700°C with a graphite cathode contains residual water, then MgO may form by the and graphite anode. In the second, liquid magnesium reaction metal is produced from melt of the same composition at 690°C with a graphite cathode and porous nickel-yttria- stabilized-zirconia cermet anode. The cermet anode is 2H20(g) + MgCl2(l) MgO(s) + 2HCl(g). (1) separated from the melt by an oxygen-ion-conducting yttria-stabilized zirconia membrane, which should Magnesium oxide has very low solubility in the chloride ensure that water vapor is produced at the anode, rather melt, and it either sinks to the bottom of the cell- than CO, C0 , or Cl . Magnesium metal globules are 2 2 forming a sludge that traps magnesium metal globules- observed in the frozen melt in the vicinity of the cathode or it reacts at the carbon anode with Cl2(g) and after completion of each experiment, and the metal consumes the anode, composition is determined using electron probe microanalysis to be greater than 99 weight percent magnesium. 2MgO(s) + 2C12(g) + C(s) - 2MgCl2(l) + C02(g). (2)

209 Vol. 20, Nos. 3-4, 2001 Electrowinning Magnesium Metal from MgCl2-NdOCl Melt Using Solid-Oxide Oxygen-Ion-Conducting Membrane

The electrolytic process is favored in the primary cathode (instead of molten tin). The second goal of the magnesium production industry: it is continuous and study was to employ solid-oxide oxygen-ion-conducting can be scaled up. However, the purification and drying membrane (SOM) technology /10/ to produce mag- of the MgCl2 feed material is expensive, accounting for nesium, using the same molten electrolyte. The cell had a large fraction of the total production cost of a graphite cathode and an anode consisting of a porous magnesium metal. There has long been interest in nickel-yttria-stabilized-zirconia cermet coating applied producing magnesium metal from impure MgCl2 on one side of an oxygen-ion-conducting yttria- feedstock, or from MgO by a continuous electrolytic stabilized zirconia membrane. process. In 1941, the U.S. Bureau of Mines initiated a program to develop a new process for magnesium production from dolomite and low-grade magnesite /1 /. Two years of development followed, and it was 2. EXPERIMENTAL concluded that magnesium metal could be produced 2.1. Preparation of batch powder from MgO in a fused MgCl2 bath. A sludge of MgO, magnesium metal, and other species collected in the In the MgCl2-NdCl3-MgO system, MgO reacts with cell, and this had to be periodically removed to prevent NdClj to form NdOCl, short-circuiting. Although these researchers were able to produce magnesium continuously, their approach was MgO(s)+NdCl3(soln)=NdOCl(s)+MgCl2(soln). (3) not adopted commercially. Several U.S. patents have recently been assigned to This reaction was shown to occur when a mixture of

General Motors (Detroit, MI) for processes and MgO, MgCl2, and NdCl3 is held at 760-790°C for 1-2 h electrolyte compositions related to the use of MgO for 191. The solubility of NdOCl in MgCl2 is approximately magnesium metal production 12-11. One idea is to use a 40 mol% (58 wt% NdOCl) in the range 700-800°C /11/. molten electrolyte containing rare earth chlorides, as The solubility of NdOCl in NdC!3 is approximately 17 well as the traditional alkali metal and alkaline earth mol% (14 wt% NdOCl) in the same temperature range chlorides. When MgO is added to the electrolyte, a IUI. chemical reaction between MgO and the rare earth In this study, a batch composition of 80 wt% MgCl2- chloride occurs; this reaction produces chlorides and 20 wt% NdOCl was used. When NdOCl is used instead oxychlorides that dissolve in the electrolyte. A of MgO in the batch, there is no need during the description of this process is available in the literature experiment to wait for the reaction in Eq. (3) to take /8/. Cathro, Deutscher, and Sharma tried to produce place; one can immediately begin electrolysis once magnesium metal directly from MgO in a chloride- fusion of the batch is complete. In some preliminary based electrolyte containing NdCl3 191. They used a experiments, a batch composition of 50 wt% MgCI2-50 liquid tin cathode and a carbon anode, and they wt% NdOCl was used; this is close to the eutectic produced a Sn-Nd-Mg alloy. The magnesium content of composition in the MgCl2-NdOCl system according to the tin alloy was very low (approximately 0.5 wt% Mg), one study /ll/. Several observations indicated that the and a large amount of high-Nd-content metal fines were melt based on a 50 wt% MgCl2-50 wt% NdOCl batch is generated in the electrolyte. The anode gas was mainly not completely molten, or has a very high viscosity. For

Cl2, rather than C02 or CO. example, at the end of one experiment using 50 wt%

In this study, the investigators sought to produce MgCl2-50 wt% NdOCl, the YSZ tube was raised out of magnesium metal directly from MgO by an electrolytic the melt at 900°C, and after waiting 30 minutes at process. The first goal was to demonstrate that 900°C, the melt was cooled at 3°C min'1 to room magnesium could be extracted from MgO using an temperature. Examination of the crucible showed a deep electrolyte similar to that employed by Cathro et al. 191. pit in the frozen melt; the melt did not flow after Solid graphite was used both as the anode and the removing the tube.

210 David Ε. Woolley et al. High Temperature Materials and Processes

The MgCl2 used in the batch is prepared by heating crucible. Just above the top of the main crucible, commercial anhydrous MgCl2 under flowing HCl gas to titanium sponge was placed in three alumina crucibles eliminate residual moisture in the as-received MgCl2 that rested on a layer of silica-alumina insulation. The /13/. The NdOCl is prepared by heating commercial furnace was sealed, and 750 mL(STP) min'1 of 98 vol%

NdCl3.6H20 under flowing air /14/. Both the MgCl2 and Ar-2 vol% H2 gas was passed through the system. The NdOCl preparation procedures used in this study were gas entered at the bottom of the furnace tube, and exited validated by x-ray diffraction (XRD) tests on the at the top. The furnace was heated at 3°C min"' to products. It is likely that some residual water remained 700°C. After waiting approximately 1 hour at 700°C, in the NdOCl powder, however. In preliminary the DC power supply (Hewlett Packard model 6033A) experiments, evolution of HCl was observed using a was used to pass current through the cell in constant- mass spectrometer (AMETEK M200 Series Quadrupole current mode. After the experiment was complete, the 1 Gas Analyzer) during heating of the batch under furnace was cooled at 3°C min" to room temperature. flowing argon through the temperature range 200- 300°C. The batch for each experiment comprised 84 g NdOCl and 320 g MgCl , both prepared ahead of time 2 2.3. Electrolytic Cell with Ni-YSZ Cermet and stored before use in a glove box. The batch Anode and Graphite Cathode ingredients were weighed into a plastic milling bottle, and then mechanically mixed by rolling on a ball mill In the second type of experiment, a solid graphite without milling media for over one hour just prior to the cathode was used, and the anode was a porous cermet experiment. layer, consisting of nickel metal and Y203-stabilized

Zr02 (YSZ). The layer was deposited on the inside of a bottom-end-closed YSZ tube, and this tube was dipped 2.2. Electrolytic Cell with Graphite Anode and into the melt (Figure 2). This experiment was conducted Graphite Cathode to test the hypothesis that magnesium can be produced Two different types of experiment were conducted. at unit activity on a graphite cathode by electrolysis of a

In the first, a solid graphite cathode and solid graphite MgCl2-NdOCl melt in a cell with a Ni-YSZ cermet anode were used (Figure 1). This experiment was anode separated from the melt by a YSZ membrane. conducted to test the hypothesis that magnesium can be The construction of the cell is different than in the produced at unit activity by electrolysis of a MgCl2- previous experiment, but the experimental procedure is NdOCl melt. The anode was a graphite crucible (inner similar. diameter 64.3 mm, inner depth 121.6). The cathode was To form the anode the following procedure was a graphite rod (outer diameter 13.5 mm, length 100 used. A batch of 10 g YSZ powder (Tosoh TZ-8Y) and mm). Eight grooves were machined in this rod to trap 30 g NiO (J.T. Baker) powder was wet-milled in magnesium metal globules. To reduce the likelihood methanol with YSZ milling beads, dried at 50°C, and that magnesium metal produced at the cathode would passed through a 270-mesh screen. A slurry, formed reach the anode, an MgO crucible (outer diameter 51.4 using 4.2 g of powder and 2 mL methanol, was poured mm, inner diameter 45.0 mm, inner depth 82.9 mm) into a bottom-end-closed YSZ tube (Coors Ceramics with six pre-cut horizontal slits separated the two Company, ZDY-4, outside diameter 12.69 mm, inside electrodes. A total of 341.017 g batch powder was used diameter 9.83 mm, overall length 204 mm). The tube of the 400 g that was milled. The furnace tube was was shaken to apply the slurry evenly, and the methanol mullite, lined with graphite foil to prevent attack by the was removed using a vacuum chamber. The coating was furnace gases. The crucible temperature was monitored sintered in air for five hours at 1500°C, and then the continuously with a type Κ (chromel-alumel) NiO was reduced to Ni by holding for one hour at thermocouple placed 1 to 2 mm from the underside of 1200°C under flowing N2-5%H2. The YSZ tube was the steel crucible platform that supported the graphite removed, and the resistance of the cermet coating was

211 Vol. 20, Νos. 3-4, 2001 Electrowinning Magnesium Metal from MgCl2-NdOCl Melt Using Solid-Oxide Oxygen-Ion-Conducting Membrane

\! 74.0mm

Fig. 1: Schematic of apparatus for electrolysis (graphite cathode and graphite anode). 1-graphite crucible cover, 2-

alumina separator tube, 3-steel cathode rod, 4-graphite cathode rod, 5-MgCl2-NdOCl melt, 6-MgO separator crucible, 7-graphite crucible (anode).

212 David Ε. Woolley et al. High Temperature Materials and Processes

r— 73.00mm ~^

Fig. 2: Schematic apparatus for electrolysis by SOM method (graphite cathode and Ni-YSZ anode with YSZ membrane). 1-graphite crucible, 2-YSZ tube (bottom-end-closed), 3-Ni-YSZ cermet coating, 4-stainIess steel

anode tube, 5-steel cathode rod, 6-graphite plug, 7-MgO separator tube, 8-graphite cathode rod, 9-MgCl2- NdOCl melt, lO-MgO separator disc.

213 Vol. 20, Nos. 3-4, 2001 Electrowinning Magnesium Metal from MgCI2-NdOCl Melt Using Solid-Oxide Oxygen-Ion-Conducting Membrane measured using two copper-wire probes and a in constant-voltage mode. After the experiment was multimeter. This anode is similar to that used in solid complete, the anode assembly was raised 76 mm, and oxide fuel cells /15/. An alumina tube was joined to the the furnace was cooled at 3°C min'1 to room end of the YSZ tube using ceramic glue (Aremco, temperature. Ceramabond 552) to form a gas-tight joint. A 6.35 mm· diameter stainless steel tube with nickel metal gauze TIG-welded to its outside surface over the lower 18 cm 3. RESULTS AND DISCUSSION was inserted into the YSZ-and-alumina tube to make 3.1. Experiment #1: Graphite anode and electrical contact with the nickel metal in the cermet graphite cathode coating, while not obstructing gas flow in the tube.

During the experiment, N2-5 vol% H2 would flow The DC power supply was used in constant-current downward in the gap between the YSZ tube and (galvanostatic) mode to pass current through the cell, stainless steel tube, and upward inside the stainless steel and the voltage was measured continuously (Table 1). tube. The total charge passed through the cell is The cathode was a graphite rod (outside diameter 13 approximately 16000 C. See Table 2 for anode- and mm, length 127 mm), which was placed coaxially with cathode-melt areas, and current densities. The total mass an MgO tube (22 mm outside diameter, 17 mm inner of the graphite crucible, MgO crucible, graphite cathode diameter, 152 mm length) to reduce the likelihood that rod, and frozen electrolyte after the experiment was magnesium metal produced at the cathode would reach 757.77 g, whereas the initial mass of those items was the yttria-stabilized zirconia membrane. Two square- 765.03 g. Cross-sectioning of the crucible revealed that shaped areas were removed from the lower end of the the frozen melt depth was 68 mm. The batch weight was MgO tube to provide contact between the catholyte with initially 341.017 g, and the available inner area in the the bulk electrolyte. A graphite crucible (63.6 mm inner graphite crucible (i.e. excluding MgO crucible and diameter, 152 mm inner depth) was used in the graphite rod) is 26.2 cm2, so the density of the frozen experiment, and an MgO disc was placed in the bottom. melt is approximately 1.9 g cm'3. Metal globules were A total of 377.044 g batch powder was used of the 400 visible along the graphite cathode rod, over the bottom 2 g that was milled. At the start of the experiment, the cm of the rod (i.e. closest to the end that was resting on graphite crucible was filled with the batch powder, and the bottom of the MgO crucible). the anode assembly rested on the top of the powder bed. The frozen electrolyte itself appeared to consist of Silica-alumina insulation was placed above and below two phases, one above the other. X-ray diffraction the crucible as in the previous experiment, and titanium analysis showed that the upper phase was MgCl2-rich, sponge was used as before. The furnace tube was sealed whereas the lower phase was NdOCl-rich. It seems and Ar gas (99% pure) was passed (flow rate 2100 mL likely that some NdOCl remained in solid form during (STP) min"1). The gas entered the bottom of the furnace the experiment, and that the lower portion of the melt tube, and exited at the top. A second gas stream of N2-5 was semi-solid, consisting of NdOCl solids suspended 1 vol% H2 at 660 mL (STP) min" entered at the top of the in MgCl2-rich liquid. No magnesium metal globules anode assembly, passed downward in the space between were visible in the grooves that had been machined in the YSZ-and-alumina tube and the stainless steel tube, the cathode rod. and returned upward inside the stainless steel tube to the The cathode was cut into pieces, and two were exit at the top of the anode assembly. Stainless steel mounted in phenolic resin and carbon coated in fittings with Viton o-rings were used to seal the anode preparation for electron probe microanalysis (ΕΡΜΑ) assembly. The furnace was heated at 2.5°C min"1 to using a wavelength dispersive spectroscope (WDS). The 690°C. After waiting approximately two hours at standard used for magnesium was pure magnesium

690°C, the anode assembly was lowered and the DC metal, and the standard for neodymium was NdF3. power supply was used to pass current through the cell Several metal globules in each sample were analyzed.

214 David Ε. Wooltey et al. High Temperature Materials and Processes

Table 1 Data from experiment #1 (graphite anode and graphite cathode)

Time period Current Voltage Charge (hours: minutes) applied (A) measured (V) (C) 03:36 0.20 1.89-2.08 2592 00:20 0.20 1.10-1.79 240 09:06 0.25 1.90-1.96 8190 00:22 0.25 1.81-1.91 330 02:22 0.50 2.13-2.16 4260

Table 2 Electrode-melt areas, and current densities

Expt. # Anode-melt area Anode current Cathode-melt area Cathode current (cm2) density (mA cm'2) (cm2) density (mA cm'2) 1 140 1.5-3.6 * 29 6.9-17 * 2 27 1.5-7.0 ** 28 1.4-6.8 ** * Steady state current was in range 0.20-0.50 A during this galvanostatic experiment. ** Steady state current was 0.04 A, and highest peak current (after current interruption) was 0.19 A during this potentiostatic experiment.

The analysis showed that the metal contained 99.0±0.2 2NdOCl(soln) + 2MgCl2(soIn) + C(s) -» wt% Mg and 0.33±0.18 wt% Nd; this is the raw data 2Mg(l) + 2NdCl3(soln) + C02(g) (7) from WDS, and the total analysis for Mg+Nd is approximately 99.3 wt%. and the anode reaction is Electrolysis of the melt should yield magnesium 2 metal, according to the following overall reaction 0 -+iC(s)-lC02(g) + 2e-. (8)

NdOCl(soln)+MgCl (soln)+C(s) -> 2 For 16000 C charge passed through the cell, 2.0 g Mg(l)+NdCl (soln)+CO(g) (4) 3 magnesium is expected (1.2 cm3), along with 2.3 g of

CO (Eq. (5)) or 1.8 g of C02 (Eq.(8)). Since the weight if the product gas is carbon monoxide. The anode loss measured from the crucible was approximately 7 g reaction is rather than 2 g, it seems likely that some volatilization

of MgCl2 occurred during the experiment, in spite of the O2" + C(s) CO(g) + 2e, (5) tight-fitting graphite crucible cover. and the cathode reaction is

3.2. Experiment #2: Ni-YSZ anode, YSZ Mg2+(soln) + 2e~ Mg(I). (6) membrane, and graphite cathode

If the product gas is carbon dioxide, then the overall A DC power supply was used to apply DC voltage reaction is to the cell (potentiostatic experiment). The applied

215 Vol. 20, Nos. 3-4, 2001 Electron inning Magnesium Metal from MgClrNdOCl Melt Using Solid-Oxide Oxygen-Ion-Conducting Membrane voltage was in the range 2.0-2.9 V, and the current was Two samples, consisting of metal globules and recorded continuously (0.05-0.20 A). See Table 2 for frozen melt, were mounted in phenolic resin and carbon anode- and cathode-melt areas, and current densities. coated for ΕΡΜΑ using WDS. The same standards were The total charge passed was approximately 3350 C. used as in the analysis for experiment # 1. Several metal After the experiment, a crater could be seen in the globules were examined, and both backscattered and center of the frozen electrolyte, indicating that the melt secondary electron images were examined. These did not flow after the anode assembly was raised at the images show that the metal is homogeneous. Small end of the experiment. The thermocouple reading was particles are visible on the metal surface, as in the 665°C when the anode assembly was raised, and fell at analysis for experiment #1. The ΕΡΜΑ showed that the approximately 3.6°C min'1 over the next twenty metal contained 101.1±0.3 wt% Mg and 0.93±0.20 wt% minutes. In order to examine the interface between the Nd; this is the raw data from the WDS analysis, and the cathode and the frozen melt, the bottom of the graphite total analysis for Mg+Nd is approximately 102 wt%. crucible was cut off, and the MgO disc was removed. Electrolysis of the melt should yield magnesium The open bottom of the graphite crucible, containing the metal, according to the following overall reaction. frozen melt, MgO tube, and graphite cathode rod, was then ground on silicon carbide paper without any NdOCl(soln) + MgCl2(soln) + H2(g) ->

lubricant. Metal droplets were observed in the bulk Mg(l) + NdClj(soln) + H20(g). (9) electrolyte and in the catholyte (Figure 3). The frozen melt depth (excluding the MgO disc thickness) was Hydrogen gas in a N2-5 vol%H2 gas mixture flowing approximately 68 mm. The batch weight was initially over the cermet coating should react with oxygen anions 377.044 g, and the available inner area in the graphite extracted through the YSZ membrane from the melt. crucible (i.e. excluding MgO tube wall, YSZ tube area, The anode reaction is and graphite rod) is 27.7 cm2, so the density of the 3 frozen melt is approximately 2.0 g cm" . 2 Ο " + H2(g) ^ H20(g) + 2e\

Fig. 3: Bottom of frozen melt, revealing magnesium metal globules, graphite cathode rod (diameter 13 mm)), and notched MgO tube (outside diameter 22 mm).

216 David Ε. Woolley et cd. High Temperature Materials and Processes

2 H2(g) + 0 (YSZ) ~> H20(g) + 2e- Anode (Ni-YSZ)

Solid-oxide oxygen-ion-conducting membrane / solid electrolyte (YSZ) Ο Ο2' (melt) --> Ο2" (YSZ) ~~ Ionic melt / molten electrolyte

(MgCl2-NdCl3-MgO) ' Mg2+ (melt) + 2e" --> Mg Cathode (graphite) Electron flow

Fig. 4: SOM process for Mg production by direction reduction of MgO.

and the cathode reaction is gas product is H20 rather than Cl2, C02, or CO. For this

demonstration, the same MgCl2-NdOCl melt was used Mg2+(soln) + 2e Mg(l). (11) as in experiment #1, although a large number of other melts may be possible. Future work on the SOM process For 3350 C charge passed through the cell, 0.42 g may be aimed at development of alternative molten magnesium is expected (0.24 cm"1). Figure 4 is a electrolyte compositions, studying the thermal and schematic of the SOM process for magnesium chemical stability of YSZ in various electrolytes, and production from a MgCl2-NdOCl melt. increasing anode current density.

4. CONCLUSIONS AND SUGGESTIONS ACKNOWLEDGMENTS FOR FURTHER WORK The authors acknowledge the help of Mr. E. Chiang Experiment #1 in the present study sought to and Mr. T. Keenan on melt preparation, cermet improve upon the results obtained in a recent study 191. processing, and set-up for electrolysis, and Mr. H. Magnesium metal with Nd content less than 1 wt% was Lingertat on NiO-YSZ slurry coating and microscopic obtained by electrolysis of a MgCl2-NdOCl melt, which examination of the YSZ tubes. Financial support from proves that liquid Mg can be obtained from a MgCl2- the Northwest Alliance for Transportation Technologies NdClj-MgO melt using a graphite anode and graphite Program at the U.S. Department of Energy's Pacific cathode. Codeposition of Nd with Mg does not appear Northwest Laboratory (Battelle) is gratefully to be a problem. Experiment #2 was a proof-of-concept acknowledged. The interest shown in this research experiment that showed that solid oxygen-ion- program by Dr. R. Jones from PNNL, Dr. A. Sherman conducting membrane (SOM) technology can be used to from USCAR, and Dr. J. Carpenter from the U.S. produce Mg from MgO. Benefits of SOM in this case Department of Energy's Office of Transportation include: the anode is non-consumable, and the anode Technologies is much appreciated.

217 1 Vol. 20, Nos. 3-4, 2001 Electron inning Magnesium Metal from MgCl2-NdOCl Melt Using Solid-Oxide Oxygen-Ion-Conducting Membrane

REFERENCES tion Process", United States Patent No. 5,665,220, General Motors Corporation, Detroit, MI (1997). 1. R.R. Lloyd, C.K. Stoddard, K.L. Mattingly, E.T. 7. R. A. Sharma: "Electrolytic Magnesium Produc- Leidigh and R.G. Knickerbocker: American tion Process", United States Patent No. 5,853,560, Institute of Mining and Metallurgical Engineers, General Motors Corporation, Detroit, MI (1998). Technical Publication No. 1848 (Class D, 8. R. A. Sharma: JOM, 39-43 (1996). NonFerrous Metallurgy, No. 89), 25 pages (1945). 9. K.J. Cathro, R.L. Deutscher, and R.A. Sharma: 2. R.A. Sharma: "Method for Producing Magnesium Journal of Applied Electrochemistry, 27, 404-413 Metal from Magnesium Oxide", United States (1997). Patent No. 5,279,716, General Motors 10. U.B. Pal and S. Britten: "Novel Electrochemical Corporation, Detroit, MI (1994). Method for Metal Extraction from Its Ores 3. R.A. Sharma: "Electrolytic Extraction of Utilizing Membranes", United States Patent No. Magnesium from a Magnesium-Containing 5,976,345, Boston University, Boston, MA (1999). Aluminum Alloy", United States Patent No. 11. P.G. Permyakov, B.G. Korshunov, and V.A. 5,395,487, General Motors Corporation, Detroit, Krokhin: Russian Journal of Inorganic Chemistry, MI (1995). 20, 1096-1098(1975). 4. R.A. Sharma: "Fused Fluoride Electrolytes for 12. F.N. T'en and I.S. Morozov: Russian Journal of Magnesium Oxide Electrolysis in the Production Inorganic Chemistry, 14, 1179-1183 (1969). of Magnesium Metal", United States Patent No. 13. P. Ehrlich: Handbook of Preparative Inorganic 5,427,657, General Motors Corporation, Detroit, Chemistry, second edition, ed. by G. Brauer, Ml (1995). Academic Press, New York, 1963; 905-909. 5. R.A. Sharma: "Electrolytic Production Process for 14. W.W. Wendlandt: Journal of Inorganic & Nuclear Magnesium and Its Alloys", United States Patent Chemistry, 5, 118-122 (1957). No. 5,593,566, General Motors Corporation, 15. O. Yamamoto: Solid State Electrochemistry, ed. by Detroit, MI (1997). P. G. Bruce, Cambridge University Press, 6. R. A. Sharma: "Electrolytic Magnesium Produc- Cambridge, UK, 1995; 292-332.

218