LUNAR MAGNESIUM PRODUCTION
Solar energy is focused into retort containing Mg ore, flux and silicon reductant.
                                                        David A. Dietzler, 2008
Producing magnesium on the Moon might be as simple as heating mare regolith that has had ilmenite and iron bearing minerals extracted magnetically and electrostatically in a solar furnace at 1500 C. and higher to volatilize silica and magnesium oxide.  In air at 1 ATM pressure MgO does not melt and volatilize until much higher temperatures; however, in the vacuum of the Moon magnesium bearing minerals will decompose and evaporate at much lower temperatures (1).

This mixture of SiO2 and MgO would then be reacted with CH4 to reduce the SiO2 leaving a mix of Si and MgO that could be mixed with flux, perhaps calcium aluminate from thermal decomposition of anorthite, and heated in a solar furnace to 1200-1500 C.  The magnesium metal will vaporize and be condensed.

From: http://solar.web.psi.ch/data/publications/pdf_books/Solar_Thermochemical_Process_Technology.pdf
we find that CH4 reduces SiO2 at 1520 K ( 1250 C) and MgO at 1770 K ( 1500 C).  However, carbothermal reduction of MgO is ineffective because the Mg vapors revert to MgO.

It should also be possible to separate SiO2 and MgO electrostatically and reduce the MgO with FeSi or silicon from magma electrolysis and some flux.  

MgO can be used as an iron and steel making flux when mixed with CaO and/or CaAl2O4 and it can be reduced with silicon to magnesium metal that evaporates and is condensed to obtain magnesium metal.  Magnesium can be used to alloy aluminum and it might be used as an explosive when made into a slurry with LOX contained in magnesium tanks detonated by a high energy electric spark. 

Silicon for MgO reduction can be obtained from FeSi obtained by molten silicate electrolysis.  Some CaO or CaO-Al2O3 flux is also required.  Iron does not participate in the reduction.

What if producing magnesium was even easier?  Aluminum can reduce silicon from anorthite in a lithium fluoride and calcium flouride flux (2).  What if  magnesium bearing olivines and pyroxenes after electrostatic sepearation of ilmenite from mare soil that also separates anorthosite, agglutinates, etc. was done (3)?  This would be followed by magnetic extraction of iron bearing olivines and pyroxene.  What if the remaining  magnesian olivine ((forsterite-Mg2SiO4) and pyroxenes ( Enstatite-Mg2Si2O6 and Diopside-MgCaSi2O6 ) were simply mixed with a CaO-Al2O3 flux and FeSi and roasted with solar energy?  More research must be done.

The reactions would be:  Mg2SiO4 + Si ==> 2SiO2(s) + Mg(g)  Mg2SiO6 + 2Si ==> 3SiO2(s) + Mg(g)
2MgCaSi2O6 +  Si  ==> 2CaO + 5 SiO2 + 2Mg(g)
1) Rudolf Keller and David B. Stofesky of EMEC Consultants
" Selective Evaporation of Lunar Oxide Components" reported in SPACE MANUFACTURING 10 PATHWAYS TO THE HIGH FRONTIER Proceedings of the Twelfth SSI-Princeton Conference May 4-7, 1995; pg. 130.
2) Christian W. Knudsen and Michael A. Gibson   Processing Lunar Soils for Oxygen and Other Materials
<http://www.belmont.k12.ca.us/ralston/programs/itech/SpaceSettlement/spaceresvol3/plsoom1.htm>
3) William N. Agosto "Lunar Beneficiation" 
<http://www.belmont.k12.ca.us/ralston/programs/itech/SpaceSettlement/spaceresvol3/lunarben1.htm>
more info about silicothermic magnesium production:

http://en.wikipedia.org/wiki/Pidgeon_process

http://members.tripod.com/Mg/mggen.htm
http://www.webelements.com/magnesium/physics.html    magnesium reflectivity 74%
http://www.webelements.com/aluminium/physics.html     aluminum reflectivity 71%

Magnesium is a slightly better reflector than aluminum. Lightweight magnesium reflectors might be the key to solar thermal power satellites.  Producing magnesium seems to be more straight forward than aluminum production. It does not require carbon, upported chlorine, H2SO4, or an upported  LiF flux as does Aluminum.  Mg production simply requires MgO, CaO or CaAl2O4, FeSi and heat; all available on the Moon. Magnesium is even easier to produce than silicon PVs.  See: Silicon and Roads

Magnesium is about 64% as dense as aluminum.  Mg 1.74 gr/cc  Al 2.7 gr/cc   Thus, we would only need to launch by mass driver 1740 tons of magnesium rather than 2700 tons of aluminum, in simple numbers, to build an equivalent mass of reflectors for a solar thermal powersat.  That's a saving of a thousand tons!  Much less energy would be required and a lot less mass driver propellant for mass catchers that haul material from L2 to L5.  If we envison an 80,000 ton solar thermal powersat like O'Neil's 5GW SPS, perhaps a fourth of it would be reflector mass.  If 20,000 tons is sheet aluminum reflectors, only 12,900 tons would be magnesium reflectors or 7100 tons less.  Conversly, if we use 20,000 tons of magnesium, the magnesium reflector will be equivalent to a 31,000 ton  aluminum reflector, therefore 1.55 times as much volume and surface area using the same thickness of sheet metal or foil for the reflector, not considering the 3% increase in reflectivity.  If we add boiler and turbine capacity to match the larger reflector the SPS generates 7.75 GW.

To really make reflectors efficient we would vapor deposit a microns thin coating of aluminum on the magnesium sheet metal reflectors.  Such thin coatings of aluminum are 98% reflective in the visible range of the light spectrum and 92% in the mid to far infrared.
MgO as well as FeO, SiO2, Na and K can be obtained by roasting regolith.  The result is cement mix and if roasting is complete enough CaAl2O4 (same as CaO-Al2O4) remains. 

FeO can be reduced to iron by magma electrolysis. SiO2 for furnace linings, furnace windows and glass. Na for lye and salt.
K for fuel cell electrolyte and farms. Na+ K make a euctectic mix for solar thermal system working fluid. CaAl2O4 for flux in steel cleaning furnaces and MgO reduction retorts. 
Alternatively, electric heat could be used to energize magnesium reduction.  With solar furnaces reflection of light and heat is a problem.  Also, solar furnaces require a very large parabolid dish, a hi temp secondary mirror that would probably be made of TiO2 on the Moon, complex structure, circular tracks to ride on, motorized aiming mechanisms and electronic controls.  Connecting electric furnaces to solar panels seems much simpler and less labor intensive and time consuming to set up on the Moon.
Vapor deposited magnesium crystals.  This is the form condensed magnesium from silicothermic reduction will take.  Other metals extracted and condensed in lunar vacuum systems will form crystals similarly.