That's what I thought as well. Silicon is basically.... well, sand, isn't it?
Silicon has to be purified to 99.9999% at the very least to make a crappy solar cell. Add four more 9's to get to making decent microprocessors. Getting certain impurities out (that behave exactly like silicon in a chemical sense) is really difficult.
http://www.cam.ac.uk/research/features/solar-grade-silicon-a...
That's 30 seconds of Googling. I can't vouch for how accurate they are.
Silicon is capital equipment intensive. It is processed in batches. While each batch is being processed, it ties up an expensive processing station for a long time.
1 pound of polysilicon takes about 2 pounds of metallurgical grade and even more energy.
1 pound of wafer takes about 2 pounds of polysilicon and even more energy.
Aluminum probably doesn't take as much energy but the difference is Silicon's energy is cheap (coal and wood chips for the first step).
By the way, would a solar furnace be feasible to replace your first step, eliminating CO2 result?
In fact, according to the foreman at one of said plants I talked to, they use pieces of quartzite that are more like pebbles than sand; no point in crushing it further I guess.
The function of the carbon is not just to create the heat but also to give the oxygen some way to remove itself. In fact, I believe the oxygen would still rather bond to silicon than carbon but the carbon is able to pull enough physically away as gas to make the reaction work. The resultant gas is mostly CO, carbon monoxide, which somehow becomes CO2 after the plant's done with it.
"Although the newest smelters can be closer to 12,500 kWh per ton let’s say most smelters are consuming electricity at 14,500-15,000 kWh/ton of ingot produced."
So, that is 12 to 15 kWH per kilogram.
Compare that with http://www.rsi-silicon.com/media/SolarGradeSilicon_050611.pd...:
"In making MG-Si, approximately 12 kilowatt-hours of electrical energy are consumed per kilogram of silicon produced."
That already is in the same ballpark, for MG = metallurgical grade silicon. Getting from there at the purity needed for chip production is energy intensive. From the same text: "Energy consumption for the Siemens process is ~200 kilowatt hours/kilogram of silicon produced"
Even correcting for a potential bias of the author (who has his own patented process that he claims to be more efficient and, I guess, that he wants to sell), I conclude that, per kg, production of silicon-grade silicon is way more energy expensive than production of aluminum.
On the other hand, http://www.rsi-silicon.com/media/SolarGradeSilicon_050611.pd... seems to indicate that chip-quality silicon is overkill for solar cells. A dedicated factory for solar cell silicon would be more energy efficient than what exists now.
But 75% of the price of aluminium is electricity.