Section 2.13.2.
Home Tour Join! Contents Team News Catalog Search Comm

Silicon Production on the Moon

Geoffrey A. Landis

Refining of Regolith

Silicon cannot be manufactured by the industrial processes used on Earth, of course, if for no other reason than silicon production on Earth uses as a raw material high-silica sand, while lunar regolith is primarily igneous silicates such as anorthosites. Also, production of one mole of silicon on Earth uses one mole of carbon in the form of coal, not available on the moon. Any production process for the moon must use no reactants that cannot be completely recycled.

Unbeneficiated lunar soil ("regolith") is powdered igneous rock. It consists of a mixture of primarily iron, calcium, aluminum, and magnesium silicates, with a few percent titanium, plus small amounts of manganese, sodium, potassium, chromium, and rare-earth oxides. The exact composition varies from site to site, and possibly from sample to sample at a given site; regolith composition is discussed in detail in the Lunar Sourcebook [2].

Several different production processes were examined. Good reviews are found in References [3,4], and so a review will not be done here. Many of the commonly proposed processes for production of oxygen from lunar resources (e.g., ilmenite reduction) were not applicable, since they did not produce silicon. Others were discarded because silicon, although a byproduct, was not produced in pure form. The process follows (but is not quite identical to) that suggested by Seboldt and co-workers from German Aerospace [5]. It has the advantage that the silicon is produced in the form of fluorosilane, which can be easily purified of contamination of the other metals. In fact, the process can be adapted in a straightforward manner to production of other relatively pure products, including titanium, aluminum, and iron.

The basic silicon reduction process is to heat the regolith in the presence of fluorine. The fluorine will displace the oxygen, which is collected as a useful byproduct. Silicon is produced in the form of fluorosilane (silicon tetrafluoride). The silicon is recovered from the flurosilane by plasma decomposition. The metal fluorides ("fluorine salts") are then reduced with potassium to recycle the fluorine.

Existing processes to produce silicon do not use fluorosilane, but instead use silane (SiH4). Both plasma deposition processes (to produce amorphous silicon) and chemical vapor deposition processes (direct thermal decomposition, to produce either amorphous or polycrystalline Si) are commonly used in the solar cell industry to reduce silane to silicon. I am assuming here that it is possible to produce silicon of acceptable quality by plasma decomposition of tetra-fluorosilane instead of silane, but this will have to be demonstrated on the ground. If not, the process will require successive disproportionation of fluorosilane to silane. This is a multi-step chemical replacement of fluorine with hydrogen at high pressure and temperature.

Note that this process is different from that suggested in earlier work [1], which suggested a fluxed aluminothermic reduction. The process suggested here is much less sensitive to the composition of the regolith source material, and thus a better one for a lander which has to do a demonstration without extensive prospecting to find "good" rock types, and without extensive beneficiation processing to isolate the desirable type of source material. The only significant beneficiation desired is to separate the smallest particle size fraction for use as feedstock. This has no effect on the ultimate reaction endpoint, but use of small particle sizes will increase the reaction rate, since small particles have the highest surface area to mass ratio.

An additional advantage of this process sequence is that it produces oxygen from the lunar regolith with high yield, and that it produces byproduct metals such as aluminum, titanium, and iron, and byproduct magnesium and calcium oxide, which are either already purified or are relatively easy to separate in pure form. These byproducts are expected to be useful for other manufacturing processes.

Fluorine is brought to the moon in the form of potassium fluoride (KF). The KF is then electrolyzed to produce the required reactants, potassium and fluorine. Potassium fluoride has a higher affinity for fluorine than any of the other elements except magnesium, calcium, sodium, and lithium. Thus, once the lunar regolith is reacted with fluorine to produce oxygen, metallic potassium can be added to the fluoride salt. This will directly reduce the iron, aluminum, and titanium fluorides to metallic iron, aluminum, and titanium.

Potassium will also reduce fluorosilane to silicon, but this step is not employed, since it is in general undesirable to contaminate electronic silicon with alkali ions. Unless solar cell tests show that this is acceptable, potassium should not be allowed contact with the silicon. Potassium has a higher chemical exchange potential for the oxygen/fluorine substitution reaction than any of the remaining metals magnesium, calcium, sodium, and lithium. This nominal temperature of this reaction is 520 °C. Therefore, if potassium is oxidized to K2O, by adding K2O to the remaining metal fluorides, the metal fluorides are returned to the metal oxide and the potassium converted to potassium fluoride. (This can probably be done in a single step of adding oxygen to the mixture of metallic potassium with magnesium and calcium fluoride). At the end of this step, all of the fluorine is in the form of potassium fluoride, therefore, only a single salt needs to be electrolyzed to recover the initial reactants, instead of a mixture of salts. The potassium fluoride is melted, at a temperature of 858 °C, and the molten salt electrolyzed back into the reactants potassium and fluorine.

This potassium reduction is a significant improvement over the alternative of direct electrolysis of the salts to recover the fluorine, since direct electrolysis will require an electrolysis temperature as high as the calcium fluoride melt temperature of 1423 °C, which presents a difficult materials problem.

It is desirable to lower the electrolysis temperature even further, since the lower the temperature, the less corrosion of the electrodes will be expected, and it is desirable to avoid the need for frequent replacement of electrodes. The electrolysis temperature can be lowered by performing the electrolysis in a eutectic salt with a lower melt temperature. Because K is extremely low on the electromotive series, very few choices are available which will reduce potassium at the cathode. According to Alabyshev, Lantratov and Morachevskii [6], Na, Ca, Sr, and Ba are lower on the electromotive series at 700 °C, and will not be reduced from the salt. Since fluorine is the most electronegative element on the electromotive series, that there is no choice available to replace fluorine as the cation.

Na, Ca, Sr, and Ba fluorides all form eutectics with KF. The NaF/KF eutectic salt mixture at a molar ratio 60:40 NaF:KF, with a melt temperature of 710 °C, seems to be the best choice. An admixture of LiF will lower the temperature further, although a significant Li component will result in lithium instead of potassium electrolysis. It is reasonable to expect that a 5 percent (molar) admixture of LiF, which will lower the temperature to 700 °C, should be possible.

Ternary eutectics have even lower melt tempertures. A potassium/sodium/calcium fluoride eutectic mixture, with molar ratio approximately 0.6:0.3:0.1 KF:NaF:CaF2, will have a melt temperature as low as 676 °C.

The electromotive series in molten salts has a dependence on temperature and on the composition of the salt [6]. Further research may make it possible to find a temperature and a salt mixture with a lower electrolysis temperature.

The process sequence has the disadvantage that magnesium, calcium, sodium, and lithium are not reduced to the metal by this process, but returned to their original condition as oxides. Sodium and lithium are only minor components of the lunar soil. Calcium and magnesium oxides makes up about 20% of typical lunar soil, and leaving these unreduced would result in about 15% reduction in oxygen production.

Except for the lost oxygen production if the calcium and magnesium oxides are not reduced, it is not clear that these oxides are valuable in reduced form. Calcium will be required in oxide form for glass manufacture (discussed in a another document). It has also been proposed to make concrete out of lunar material; this also requires calcium oxide. Thus, it is not of great significance that the process does not reduce calcium and magnesium.

An alternative reduction strategy is to use lithium fluoride. Lithium can be separated from fluorine by electrolysis in a KF/LiF eutectic salt mixture at a temperature of only 492 °C. Lithium metal will reduce magnesium, sodium, and potassium fluoride to metal, leaving only calcium fluoride unreacted. By use of lithium fluoride, the molten salt electrolysis could be reduced from many elements to only two, lithium and calcium. If lithium oxide is then used to drive the reaction CaF2 + Li2O --> CaO + 2LiF, this can be reduced to a single electrolysis. However, this reaction exploits only a small difference in chemical potential, instead of the large difference exploited in the potassium substitution reaction. It is worth exploring to see whether the calcium/lithium oxide exchange occurs at a fast enough rate, and with a high enough recovery of lithium fluoride, for use as a production reaction.

It might be argued that fluorine is extremely reactive, and it will be difficult to find materials in which to do the electrolysis. However, electrolysis of molten fluoride salts is a high-throughput industrial process, used in the production of many materials. The most well-known fluoride electrolysis process is the Hall process for production of aluminum, in which molten sodium-aluminum silicate ("cryolite") is electrolyzed at 1000 °C. Seyboldt, et al, discuss some of the materials issues involved.

Silicon Production on the Moon

A block diagram of the basic mass flow process for the silicon production is shown in figure 2. The power requirements listed are for a demonstration-sized system to produce a few grams of silicon. The process selected for this is to use fluorine extraction to segregate silicon in the form of gaseous tetra-fluorosilane (SiF4).

Figure 2

Overall Requirements:

  1. 11 kW-hr of electrical power
  2. solar furnace with capability of 676 °C operation
  3. condensor stage with capability of 178 °K refrigeration
  4. crucibles, gas chambers, plumbing for above


  1. Oxygen
  2. Silicon
  3. Aluminum/titanium/iron mixture
  4. Calcium and magnesium oxides (unseparated mixture)

The demonstration system would be designed so that only one chamber is in use at any one time. Thus, only a single solar furnace is required; the required reaction chamber is simply rotated into the hot zone when it is required. The solar dynamic ground-test demonstration, based on a power system originally designed for the Space Station Freedom, demonstrates melting a LiF/CaF2 eutectic salt as well as long-duration operation at 800 °C under simulated space conditions [7]. This design could easily be adapted to serve as the solar furnace required to produce the temperatures required for this process.

The heating process will also evolve volatiles (primarily materials implanted from the solar wind) from the regolith. These will include hydrogen, oxygen, carbon monoxide, water, etc. These may be valuable, and can be captured after the heating but before the fluorine reaction is initiated by use of condensors.

For the fluorination step, there may be some unreacted oxide if the reaction is not done in very high temperatures. As long as these do not contain fluorine, these can be considered slag and discarded.

The condensor must also separate the fluorosilane from the oxygen evolved from the sample, and must also recover any unreacted fluorine. This is probably best done by fractional liquifaction (distillation). Fluorosilane will liquify at -86 °C, oxygen at -183 °C, and fluorine at -188 °C. Since any manufacturing sequence to produce silicon for solar cells would undoubtedly also manufacture liquid oxygen for rocket fuel and other applications, liquifaction equipment will be present in any case, and it is only necessary to adapt the equipment to allow progressive liquifaction of the individual gasses. Since the boiling points of fluorine and oxygen are close, it may be necessary to repeat the distillation process to adequately purify the product gasses and recover the fluorine reactant.

TiF4, with a sublimation point of 284 °C, also sublimates at the reaction temperatures of interest. The TiF4 vapor must be condensed on a cold finger before the vapors enter the main condensor. If it is desirable to separate the titanium out, this would be easy to do. The titanium fluoride will also include trace amounts of other volatile fluorides, as discussed later; if it is necessary to produce titanium free of even trace contaminents, it may have to be purified further, probably by additional fractional distillation.


1. Landis, G., "Lunar Production of Space Photovoltaic Arrays," Proceedings of the 20th IEEE Photovoltaic Specialists Conference, Las Vegas, NV; 874-879 (1988).

2. McKay, D., et al., "The Lunar Regolith," Chapter 7, Lunar Sourcebook, Heiken, G., Vaniman, D., and French, B., eds., pp. 285-356, Cambridge University Press, Cambridge, 1991.

3. , L.A., and Carrier, W.D., "Oxygen Production on the Moon: An Overview and Evaluation," Resources of Near Earth Space, pp. 69-108, J.S. Lewis, M.S. Matthews, and M.L. Guerrieri, eds., U. Arizona Press, Tucson AZ, 1993.

4. Hepp, A., Linne, D., Landis, G., Wadel, M., and Colvin, J., "Production and Use of Metals and Oxygen for Lunar Propulsion," J. Propulsion and Power, Vol. 10, No. 6, pp. 834-840, Nov-Dec. 1994.

5. Seboldt, W., Lingner, S., Hoernes, S., Grimmeisen, W., Lekies, R., Herkelmann, R., and Burt, D., "Lunar Oxygen Extraction Using Fluorine," Resources of Near Earth Space, pp. 129-148, J.S. Lewis, M.S. Matthews, and M.L. Guerrieri, eds., U. Arizona Press, Tucson AZ, 1993.

6. Alabyshev, A.F., Lantratov M.F., and Morachevskii, A.G., Reference Electrodes for Fused Salts, Sigma Press, Washington D.C., pp. 32-39, 1965

7. Shaltens, R.K. and Boyle R.B., "Initial Results from the Solar Dynamic (SD) Ground Test Demonstration Project at NASA Lewis," 30th Intersociety Energy Conversion Engineering Conference, IECEC-95-421, Orlando, FL July 31-Aug. 4, 1995; available as NASA Technical Memorandum TM-107004 (1995).

Content by Geoffrey A. Landis <>


Home Tour Join! Contents Team News Catalog Search Comm

ASI W9700221r1.1. Copyright © 2004 Artemis Society International, for the contributors. All rights reserved.
This web site contains many trade names and copyrighted articles and images. Refer to the copyright page for terms of use.
Maintained by ASI Web Team <>.
Submit update to this page. Maintained with WebSite Director. Updated Sun, Jun 21, 1998.