Let's not get hung up on the fact that it would take decades of previous large-scale development to get to that point. Why would we go at all if that wasn't the long-range plan? The farther into the future one thinks, the less need there is for people to be in space. It gets ever cheaper and safer to have robotic equipment do work instead. If we are there, it will only be because we want to be there. The argument for making the effort becomes stronger, in fact, the more people have that desire, so let's cut to the chase. Most people will not live in a cave, or under artificial lighting, or under any conditions less appealing than life on Earth. They wouldn't give up Earth unless their new home can offer an experience of equal quality. So, we'll need to make transparent domes so huge the space within offers the feel of the great outdoors of our home world. Clearly we'll need to work towards that in steps, creating ever larger windows until those windows are a large fraction of the enclosures protecting living spaces, and then make those spaces huge. The dome planned for Lalande City requires about 4 billion tons of glass.
The issue is silica - SiO2, silicon dioxide - the stuff available all over Earth as sand that in most cases can be used with little alteration to make okay glass. There is no silica sand on the Moon. Sure, there is plenty of silica - it's nearly half the mass of the lunar crust - but it isn't pure. It's mixed in complex minerals. So we have been searching for a way to purify what there is into the good stuff. I believe what follows would work, though it probably needs considerable tweaking.
1. The first step is to produce pure olivine. That is relatively straight-forward - as long as you have a sizable solar furnace, something you want on the Moon for all kinds of reasons, anyhow. By the time we contemplate major glass production, there are a variety of large ones and at least one giant one. With this you melt a quantity of regolith containing olivine - which it almost all does. If you are able to reach bedrock (which the colony has by this time and is using it for various purposes), with some rooting around you can probably find a deposit of dunite. That is over 90% olivine to start with. Melt it thoroughly, then pour it into a centrifuge.
2. In 0.16 of a gravity a centrifuge really isn't hard to make. I look forward to seeing what wonderful low-friction bearings can be made in such a place. At about 1200°C, olivine crystallizes while the rest of the melt remains liquid. On Earth, these crystals will sink through magma if nothing else interferes. On the Moon, because the gravity is so much lower, that might not happen. So the centrifuge helps by creating enough of an outward pull to cause materials to separate by density. Once you figure out the best temperature and artificial gravity to use in this process, you can get a layer of pretty pure olivine packed against the outer wall of the centrifuge vessel. Since the dissipation of heat can be reduced simply by using foil that reflects radiated heat back into the centrifuge vessel, the number of g's of artificial gravity needed could be low, as there is more time available for the crystals to collect before the the lava cools too much. A layer of pretty pure olivine gets packed against the outer wall of the centrifuge vessel.
|Spinning exerts a force of a few g's on the lava, throwing it against the vessel walls (shown in a cutaway view)|
3. Once all the olivine has solidified, pour off the remaining molten material in the middle. This will need to be done at a temperature not far below the melting point of olivine, otherwise further minerals will begin to crystallize and settle out. Set this material aside. Open the centrifuge vessel - it splits into two halves that are lined with graphite, allowing the tube of stone inside to be popped out easily.
4. Melt the olivine again. It is is still composed of two minerals: fayalite - Fe2(SiO4), and forsterite - Mg2(SiO4). Fayalite crystallizes at a temperature far below that of forsterite. It is what causes crystals to form of sufficient size to sink in magma deposits on Earth. That is what this whole concept is based on - how minerals are known to separate on Earth under certain conditions. So perhaps this is a good time to mention certain unknowns due to the fact that the Moon is not the same. Magma on Earth normally has a higher content of silica - that's why we have this problem in the first place, the Moon doesn't have enough silica for it to aggregate as a pure mineral (quartz) at the end of a magma crystallization process. Silica prefers to bond with other metal oxides, and only crystallizes as pure quartz when all of them are gone. Magma with more silica is more viscous (less runny). Earth magma has a lot more sodium and potassium, both of which make it less viscous. Raising the temperature of a melt more will make it runnier. Because the melting point of forsterite is a full 700°C above that of fayalite, at some temperature below forsterite's melting point, and at some rate of centrifugal acceleration, the fayalite will be runny enough to sink through the forsterite, which will form a crust on top of it. This for me was sort of counter-intuitive (so much so i got it wrong and am fixing it now with an edit). Iron is a much heavier element than magnesium, so even though the fayalite is liquid and the forsterite is solid, the fayalite is going to form a layer under the forsterite because it is denser. So now you can't simply stop the centrifuge and pour off the liquid contents. If you stop it, the crust of forsterite is going to break up and mix with the liquid fayalite, contaminating it. Hm. There are two options that seem plausible. Drain the liquid through taps on the centrifuge outer walls, stopping before the forsterite layer gets too close. Or, let the whole thing freeze, and then cut the fayalite layer away from the forsterite layer.
5. Now we are getting somewhere. Pour the fayalite into a fluidized bed reactor and bubble hydrogen through it. It will react to form water, iron, and silica. The water forms steam which is continuously pumped off to keep the pressure in the reactor low and the reaction going nicely in the desired direction. As more water is removed, the melt will tend to solidify unless it is heated further. Iron melts at 1538°C, and silica at 1713°C (both at 1 atmosphere). The more the mix is depleted in oxygen by the reaction, the less the iron will combine with the silica. Once no more steam is coming off the mix, let it cool to below the freezing point of iron, and the iron will sink to the bottom of the reactor. Pour off the silica. Reheat the iron so it melts, and pour it off too. (Note: there is a 2nd candidate reaction to perform this step - bubbling oxygen through the melt to produce magnetite and quartz. Possibly that would be better, and it will also be documented when all this is added to the regular contents of the website. I chose to describe the hydrogen reaction here as it is a bit less complex and also produces iron, however if it is too slow or doesn't sufficiently purify the mix, the oxygen reaction may replace it. Recycling the hydrogen by splitting the water molecules may be easier than getting more oxygen, that is another consideration.)
6. Okay. Remember how i said you could start with dunite, which is almost all olivine? Actually you might want to avoid that, unless you need extra of the products of the above reaction, which you certainly might because all the products are darn useful. Even the forsterite - you could make crucibles or refractory bricks out if it. Otherwise, it could be a lot more efficient to seek out highlands material that is a good mix of plagioclase with pyroxene and olivine. Stuff that's about 60% plagioclase with the remainder mostly olivine would do nicely. Finding something like that should not be hard. In that case, the stuff you poured off in the first step can now be added to the silica you produced from it, and the ratios will be right for alkaline earth aluminosilicate glass.
7. The ratio of silica, to aluminum oxide, to alkaline earths (calcium or magnesium oxide) needs to be close to 12:5:3. Once the melt with the extra silica has been reheated and mixed, it can be poured into molds, blown into shapes, or extruded through rollers. The roller process is the one that is the key. Taking advantage of the vacuum and low gravity once again, chilled rollers continuously fed by a vat of molten glass maintained at the right temperature can be used to produce plate glass the way they did in the old days.
A critical piece of information that is currently missing is how long you have to cool the molten glass from its working point to its annealing point before crystals start to appear in it and make it cloudy. (It will be necessary to contact a manufacturer to get that information, but i wanted to post this before then. I'll edit in the data later.) The time must be quite short, as the only thing glass with this chemistry is used for on Earth is as bulbs for halogen lamps and other applications where it can be cooled very quickly. Massive rollers chilled internally by liquid coolant can get that job done. It's just a matter of setting up the system with the right feed rate and rollers that are big enough. Because the operations can all be done in a hard vacuum, it is relatively easy to keep the vat of molten glass at the right temperature while it empties by using reflective insulation. The vacuum also ensures there are no dissolved gasses in the melt to cause bubbles. The rollers need to be cold enough that the glass is almost solid when it emerges from them, but the glass will still be about 900°C. In that case, it isn't so hard to use radiators and heat pumps to keep those rollers chilled.
If the resulting glass is more than about 2% iron, it will have a green tone visible enough to be distracting. Not nearly as distracting as living with tiny windows, but not ideal. Of course, the lunar landscape is not exactly brimming with color, so that may be a pretty minor point. If there is still a percent or two of magnesium, that is hardly a problem either. Magnesium oxide is an alkaline earth, as long as the ratio of those is right, it doesn't matter if it is made of calcium oxide, or magnesium oxide. Nothing else that is even slightly common in lunar soil presents any problems either. I'm telling you, this will totally work.
Which makes me feel a whole lot better, because now i'm going to proceed to detail our giant glass-covered city, with all the trimmings.
Edit - This post was improved to reflect better information when inaccuracies and unclear passages were pointed out to me when it was first posted yesterday. I really appreciate it when people take the time to do that. This person did so in our chat room on space.stackexchange - the address of which i'm about to add here in the sidebar, as well. Thanks to MolbOrg for that assistance. Further Edits - fixed viscosity - more SiO2 makes glass more viscous, not less. Fixed description of separation of fayalite and forsterite, to reflect that fayalite is denser.