The Making of
Silicon
It's the second most abundant element on Earth — the stuff of ordinary sand. The hard part isn't finding it; it's purity. Reaching a solar cell means climbing a ladder where every rung costs more energy than the last. Reaching a computer chip means climbing further still — then adding tiny amounts of impurity back on purpose.
Every figure on this page is sourced — the notes at the end say where each one comes from.
The invisible metalloid inside almost everything electronic — and a great deal that isn't.
The whole journey at a glance — one furnace, two products
Sand and carbon go into one furnace — and the fork happens immediately. Add iron and you get ferrosilicon for steel. Leave the iron out and you get silicon metal, which can either be used as-is or sent up a long purity ladder toward solar cells and computer chips. Silicon products sit at different rungs of that ladder. Tap any stage to look closer.
↓ tap any stage to explore ↓
You start with sand and a lump of coal
Silicon is everywhere — it's the second most abundant element in the crust. The whole difficulty is that it never occurs pure. It is always locked to oxygen.
Almost all the silicon on Earth is bound up with oxygen in minerals such as quartz — the mineral form of silica, SiO2. To get the element out, you have to tear it away from the oxygen, and oxygen holds on hard. That is what the furnace is for, and it is why making even the cheapest, dirtiest silicon takes so much energy. Two raw materials go in, plus a great deal of electricity.
Quartz (silica, SiO2)
High-purity lump quartz — ordinary sand and quartzite are the same compound. It carries the silicon, chemically fused to oxygen. Roughly 2.5 tonnes of quartz go in per tonne of silicon. The related material glass is silica that has been melted rather than reduced.
Carbon (reductant)
A mix of coke, coal, charcoal and wood chips, around 1–2 tonnes per tonne of silicon. This is not fuel in the ordinary sense: the carbon is a reactant. It bonds to the oxygen and carries it away as carbon monoxide, freeing the silicon. Its raw form is coal.
Electricity (lots of it)
A submerged-arc furnace draws an enormous current: roughly 11–13 megawatt-hours per tonne of silicon — several UK homes' worth of electricity for a year, per tonne. Where that power comes from decides much of silicon's carbon footprint.
Burning the oxygen off at 2,000°C
The furnace does one job: strip the oxygen out of silica. Carbon grabs the oxygen; the silicon is left behind as a molten metal.
Quartz and carbon are fed into a submerged-arc furnace, where electrodes push the temperature past 2,000°C. At that heat the carbon reacts with the oxygen in the silica and drags it off as carbon monoxide gas. What is left is molten silicon, which pools at the base of the furnace above 1,811°C and is tapped off at around 1,600°C.
Because the reaction never runs cleanly, the raw product is refined in the ladle: air or oxygen is blown through the melt to burn off stray impurities. The furnace typically converts about 85% of the silicon in the quartz into usable metal. The finished metallurgical-grade silicon comes out at roughly 98–99% pure: good enough for alloys and chemicals, nowhere near good enough for electronics.
SiO2 + 2 C → Si + 2 CO
Read it as a tug-of-war for oxygen. Carbon wins, but only at extreme temperature, and it leaves as carbon monoxide — which is why clean electricity alone cannot make this step carbon-free. The carbon isn't heating the furnace so much as doing the chemistry. Real furnaces run several linked reactions through silicon carbide as an intermediate; this is the overall change.
Each rung costs far more than the last
98% silicon and 99.9999999% silicon are the same element. What separates them is a chemical climb so steep that most of the world's high-purity silicon is made by a single family of processes.
Metallurgical-grade silicon is about two nines pure — 99%. A solar cell needs six nines (99.9999%, under one impurity atom in a million). A computer chip needs nine to eleven nines (99.9999999%+). Those extra decimal places are not a formality: they are the whole difficulty of making silicon, and the cost climbs sharply at the high-purity end.
The climb, in orders of magnitude
Each step removes most of the impurities that remain. Getting from metallurgical silicon to solar grade means taking out roughly 9,999 impurity atoms in every 10,000. Getting from solar grade to electronic grade removes about 999 in every 1,000 of what is still left.
Purity and structure are two different problems. The first three rungs are about removing foreign atoms. The last is about arrangement: even perfectly pure silicon is useless for a chip until it is grown as a single, continuous crystal with its atoms in a flawless repeating lattice.
How you actually climb it
Two processes do almost all the work, one after the other.
Siemens process — purify by distilling a gas
You cannot filter impurities out of a solid. So metallurgical silicon is first turned into trichlorosilane, a volatile liquid that can be purified by repeated distillation — the same idea as refining spirits, run to extraordinary precision. The purified gas is then broken back down onto hot rods inside a reactor, where ultra-pure silicon grows layer by layer as polysilicon.
Alternatives exist — the fluidised-bed reactor makes polysilicon as granules with less energy, and some producers upgrade metallurgical silicon directly — but the Siemens route still dominates world supply.
Czochralski growth — pull one perfect crystal
Polysilicon is pure but it is a jumble of crystals. To make a chip you need one. The silicon is melted, a tiny seed crystal is dipped in and slowly withdrawn while rotating, and the melt freezes onto it in perfect atomic register — growing a single, continuous crystal the size of a log, called an ingot.
The ingot is sliced into thin discs — wafers — polished to a mirror. This is a physical change, not a chemical one: no new substance is made, the atoms are simply put in order.
At the end of this climb you have the purest, most perfectly ordered material humans routinely manufacture — and it still can't do anything useful. It is too pure.
Then you add the dirt back — on purpose
After climbing nine or eleven nines to make silicon perfectly pure, the final step is to contaminate it again — with exquisite control.
Perfectly pure silicon conducts electricity far worse than a metal. That is the point: it is a semiconductor, sitting between conductor and insulator. To make it useful you introduce a tiny, precisely measured amount of another element — doping — sometimes as little as a few atoms in a billion. Add boron and you get p-type silicon, hungry for electrons; add phosphorus and you get n-type, with electrons to spare.
Put a p-type region against an n-type region and you have a p–n junction — the building block from which transistors, solar cells and every microchip are constructed. The whole point of the purity ladder was to reach a material so clean that these deliberately added atoms, and nothing else, control how it behaves.
One letter, two very different things. Silicon is the element — the grey metalloid in chips and solar cells. Silicone is a family of soft, rubbery polymers built from silicon, oxygen, carbon and hydrogen — the stuff of sealants, bakeware and medical tubing. Silicones are made from silicon metal, but they never climb the purity ladder. The kitchen spatula and the microchip share an ancestor and almost nothing else.
Four products, one dominant supplier
The same furnace feeds four very different markets — and one country makes most of the world's silicon in every form, especially at the high-purity end.
Silicon leaves the process as one of four commercial forms, each sitting at a different rung of the purity ladder. They are not upgrades of one another so much as different destinations.
Ferrosilicon
Made by adding iron ore to the furnace charge. Most of it goes into steelmaking as a deoxidiser and alloying agent. By tonnage this is where most silicon-bearing furnace output goes.
Silicon metal
The furnace product without added iron. Split between aluminium alloys (silicon makes cast aluminium flow and hold its shape) and the chemicals that become silicones.
Solar-grade polysilicon
Purified silicon for photovoltaic cells. This is the fastest-growing pull on high-purity silicon by far — solar takes well over 95% of high-purity demand.
Electronic-grade polysilicon
The purest form, for computer chips. Tiny by tonnage next to the others, but the top of the ladder — and the hardest, most valuable silicon to make.
Where it's made
China dominates silicon at every grade, and the concentration tightens as purity rises. The two figures below are measured against different totals — one is a share of all silicon materials, the other a share of high-purity capacity — so they are shown side by side, not as one rising number.
All silicon materials
the purity
High-purity capacity
Two different denominators. The ~80% is China's share of all silicon materials (USGS puts it at almost 80%; the IEA at 70–80% of silicon metal). The ~95% is China's share of world high-purity silicon capacity in 2024 — out of about 1,900 kilotonnes total, with Germany (50 kt), Malaysia (23 kt) and the United States (16 kt) making up most of the rest. A large share of that Chinese capacity sits in Xinjiang and western China, which is why several countries have moved to build supply outside it.
Two waves of demand
For decades silicon metal was mostly an alloying and chemicals material. Then solar arrived. The energy sector's share of silicon-metal demand has gone from about 6% in 2010 to more than 30% today, and a second wave is forming as silicon starts to appear in battery anodes.
Projected figures are IEA scenario results, not predictions; they describe what demand would be under stated policy and technology assumptions.
The carbon is in the chemistry, not just the power
You can run the furnace on clean electricity and still emit carbon dioxide — because in silicon's case, the carbon isn't only the fuel. It's a reactant.
Most industrial decarbonisation is about swapping dirty energy for clean. Silicon has a harder problem hiding underneath that one. The carbon in the furnace is there to grab the oxygen off the silica; when it does, it leaves as carbon dioxide. Even a furnace powered entirely by renewable electricity would still produce that CO2, because the carbon is doing chemistry, not just supplying heat. In fact the chemical energy entering through the carbon is comparable to the electrical energy, so it carries roughly half the furnace's energy as well.
Why clean power isn't enough
Standard carbon accounting assigns metallurgical silicon a factor of about 5 tonnes of CO2 per tonne of silicon. Fuller life-cycle studies, which count the electricity and the raw materials as actually supplied, put the total nearer 2.7–4.5 tonnes of CO2-equivalent per tonne, depending heavily on the carbon sources and the local grid.
That grid dependence is large. China's grid is roughly 87% fossil-fuelled; Norway's is around half renewable. The same process in the two places has very different footprints — which is part of why where silicon is made matters as much as how.
Unlike iron oxide, which can be reduced with hydrogen, silica has no hydrogen escape route — so silicon can't simply switch to green hydrogen the way steelmaking hopes to.
The steel comparison — and where it breaks
Steelmaking also uses carbon to reduce its ore, so silicon is not alone in this. But steel has an exit that silicon lacks: iron oxide can be reduced by hydrogen instead of carbon, which is the basis of the emerging hydrogen-steel routes. Run that reaction and the by-product is water, not CO2.
Silica will not give up its oxygen to hydrogen. The bond is too strong; the chemistry doesn't run. So silicon cannot copy steel's cleanest planned route, and has to find its own.
Biocarbon
Replace fossil coke with charcoal and wood chips from sustainably grown biomass. The chemistry is identical, but the carbon is recently captured rather than fossil.
Molten-salt electrolysis
Extract silicon electrochemically instead of with carbon at all — in principle a genuinely carbon-free route, since no carbon reductant is used.
Recycling
Little help here. Silicon in cells and chips is hard to recover, and recovered silicon reaches only metallurgical grade — not solar or electronic.
Recycling deserves the caveat. The EU's Critical Raw Materials Act sets a 25%-by-2030 recycling benchmark for the bloc's strategic raw materials as a group; silicon metal is on the list. But that is a general target, not a figure the silicon industry is expected to hit specifically. In practice recovered silicon comes back at metallurgical grade at best, which is why almost none of it re-enters the solar or electronic supply.
Silicon's story runs opposite to most materials. For iron or copper, the hard part is finding good ore. For silicon the ore is the beach; the hard part is everything you do to it afterwards. It takes the second most common element on Earth, spends enormous energy dragging it away from oxygen, climbs nine or eleven nines of purity, and orders every atom into a single crystal. And then, at the very top of that climb, it deliberately puts a few impurity atoms back, because a material that perfect is useless until you spoil it, just so.