Where Does Silicon Come From?

If I ever got it into my head to start believing that there had to be a god, the existence of silicon would probably be one of my main arguments in favor of.  It's just one of the several elements sitting on the edge of the metal/nonmetal divide on the periodic table, but as noted white supremacist/physicist William Shockley and friends figured out back in 1947, it's a semiconductor, meaning that whether or not it will conduct electricity is something we can exert control over pretty easily via some clever applications of quantum mechanics and materials science.

It wouldn't be an exaggeration to say that that single discovery probably transformed society more than anything has since the printing press; without it we've got no solid-state electronics, no computers, no internet, no solar cells, and basically none of the cool technology we all take for granted these days (I also wouldn't have a job).  On its own, being a semiconductor isn't that special; there are loads of materials that semiconduct, both elements and alloys.  Silicon has two other big things going for it though.  One is that it oxidizes spontaneously, meaning that if you want to make part of it into a nonconductive oxide it's almost stupidly easy to do so (most other semiconductors don't share this trait).  When you're making complex three-dimensional circuits out of silicon you're going to want to insulate lots of parts from other parts; the fact that it forms oxides so easily makes manufacturing Si-based devices vastly simpler (ergo cheaper) than other semiconductors.

The second, even more important advantage of silicon is that it's the second most common element in the earth's crust.  This miracle material that is the foundation of all modern technology is literally just sitting around all over the place in things like "sand" and "rocks"; unlike other useful things we dig out of the earth (oil, gold, helium, etc), we're never going to run out of silicon no matter how many cell phones each of us decides we need to own.  Even my shrivelled, agnostic engineer's soul is moved by the staggering serendipity of that one, although I'm sure there's a good reason for it that I just don't know (readers who know anything about planetary formation and earth science, feel free to chime in here).

Come on, it's pretty bananas right?

Unfortunately it's not as simple as just scooping up some sand and turning it into a bunch of Core i7 processors, which you'd probably already guessed from the fact that Core i7's cost significantly more than sand.  Remember when I said that silicon will oxidize spontaneously?  That's incredibly convenient in manufacturing, but it also means that basically none of the silicon found in nature is "pure;" it's all been oxidized into "silicate," or silicon dioxide (SiO₂) if you like chemistry.  The flipside of the fact that silicon oxidizes so easily is that it's extremely goddamn hard to get it to un-oxidize back into elemental form; when you're making semiconductor devices, the only way to cut through a layer of SiO₂ is with directed plasma or hydrofluoric acid (HF), a chemical whose lethality has become the stuff of legend among people in this line of work.

So that brings us to the question: how do you turn silicate from the ground into semiconductor-grade Si (which is 99.9999999% pure.  I might have missed a 9 in there) on the scale required to feed the insatiable monster that is the commercial semiconductor industry?  I assume they're not doing it with warehouse-sized plasma fields or gigantic vats of HF.  I hope not, anyway.

Did you know there's a whole field devoted to turning stuff into other stuff on industrial scales?  It's called metallurgy, and aside from that I know next to nothing about it.  Unsurprisingly, metallurgists solved this problem a long time ago (probably about when people were realizing that large quantities of very pure silicon would be a handy thing to have around) and I just never knew about it because it never occurred to me to question where the ultrapure, monocrystalline silicon wafers I work with actually come from. 

Any workable process to turn silicate into metallic Si is going to need:

1. Input energy, because the oxygen isn't just going to un-bond from the Si if you ask it nicely
   
   
2. Something else for the free oxygen to bond to and get it out of the way once you've separated it from the Si
   
   
3. A way to get the now-free Si out and separated from any other byproduct you may have created while doing the first two things

It would also be nice if the process scaled, so you could make a ton of silicon as easily as a gram just by building a bigger reactor.


The most scalable, widely-used process by which SiO₂ is turned into reasonably pure silicon is called carbothermic reduction, and for the most part it's pretty much what it sounds like.  The first thing you need is heat, a lot of heat (furnace temperatures run north of 2000C), in order to provide the necessary energy to split SiO₂ into its component atoms.  This is achieved by running the whole mess inside something called an arc furnace, which produces heat via (yup) several sustained electrical arcs inside the furnace (it's not like you're going to get to 2000C by just lighting a fire under the thing or whatever).  Into this hellish environment you'll then dump both silicate and carbon feedstock.  It helps if both are reasonably pure, both for the purity of the final product and the minimization of nasty byproducts.  Coal is generally used for the carbon, while mined crystalline quartz is used for the silicate.

There are a few steps involved in the actual reaction, but basically what happens is that the extreme heat forces the silicon and oxygen atoms in SiO₂ to separate.  Both atoms go through several intermediate reactions with the carbon we dumped in there, but eventually they'll settle into a combination of metallic silicon and carbon monoxide or carbon dioxide.  Metallic silicon is liquid at the furnace temperatures, so it'll drip down to the bottom where it can be collected.  CO/CO2, as we all know, are gases, so it'll leave via the furnace exhaust and do what it can to contribute to the global warming problem.

A Si-refining furnace, swiped from the survey paper I used as the main source for this thing.

The silicon you get out of this process is ~99% pure, or metallurgical-grade.  That's pretty good, but a far cry from the "nine 9's", or 99.9999999% purity, we need for semiconductor-grade material.  There are a variety of ways to increase the purity from here, both by adding things to the mix to precipitate out certain impurities and by good old-fashioned repeated distillation.  Even a few stray atoms of the wrong type can play absolute hell with semiconductor operation, so you really have to do a good job with this part.  That said, the details of it are extensive and boring, so I'm just going to stop here.  We've figured out the main question (how to get the Si out of SiO₂ on a large scale) anyway; the rest is just gravy.

So we've learned how to make very pure metallic silicon out of most of the earth's crust and found a brand-new way to contribute to climate change, all in one shot!  If you're interested in all the gory details of this process, I yanked a lot of them from a survey paper that goes into much more detail on every step of getting silicon from the quartz mine to the inside of your iPhone.