Silicon (Si)

Electronic-grade silicon is the purest bulk material human beings routinely manufacture. The specification is usually quoted as eleven nines — 99.999999999% — which means fewer than one foreign atom in every hundred billion. That is not fastidiousness. A semiconductor works because the deliberately added impurities, at parts per billion, control the conduction; anything else present at a similar level competes with them and the device does not do what it was designed to do.

Purity is the whole industry

Turning sand into a wafer is a sequence of increasingly unreasonable purifications. Quartz is first reduced with carbon in an electric arc furnace to metallurgical-grade silicon at around 98 to 99% — good enough for alloying and nothing else. That is converted to trichlorosilane, distilled, and decomposed back onto heated rods in the Siemens process, which is where almost all the energy and cost sits.

The result is polycrystalline, and a transistor needs a single crystal with no grain boundaries. The Czochralski method supplies it: a small seed crystal is touched to the surface of molten silicon and withdrawn slowly while rotating, and the melt freezes onto it in the seed's own orientation, drawing out a cylindrical boule that can weigh hundreds of kilograms. Jan Czochralski discovered the principle in 1916, reportedly after dipping his pen into a crucible of molten tin instead of the inkwell.

Impurities largely refuse to join the growing crystal and stay behind in the melt, so the pull itself purifies. This is the same segregation effect exploited more aggressively in zone refining, where a narrow molten band is passed repeatedly along an ingot, sweeping contamination to one end to be cut off.

Most of it never goes near a chip

By mass, the semiconductor industry is a minor consumer of silicon. The bulk of production has always gone to metallurgy and to polymers.

• Ferrosilicon is the single largest outlet: an iron-silicon alloy added to steelmaking as a deoxidiser, and the basis of the electrical steels used in transformer cores.
• Aluminum alloys for casting almost all contain silicon, typically around 7 to 12%, because it improves fluidity in the mould and reduces shrinkage on solidification. An engine block is aluminum-silicon.
• Silicones — polymers with a backbone of alternating silicon and oxygen atoms — supply sealants, lubricants, medical tubing and bakeware. Frederic Stanley Kipping did the foundational chemistry in the early twentieth century and coined the name from an incorrect structural analogy he later acknowledged. The word is worth keeping distinct from the element: silicone is a family of compounds, silicon is an element, and no implant has ever been made of the latter.

Solar photovoltaics sit between the extremes, needing around six nines rather than eleven, and now consume more purified silicon than microelectronics does.

Named against Davy's wishes

Humphry Davy proposed silicium in 1808, on the assumption that the substance would turn out to be a metal, and the -ium ending survives in French, German, Spanish and most other European languages. Thomas Thomson objected in 1817: the element was clearly not metallic, and its behaviour resembled boron and carbon, whose names end in -on. English took Thomson's version.

The root itself is silex, flint — the stone whose sharp conchoidal fracture made it the first engineering material.

Three dates, one element, and none of them wrong

Attribution here is genuinely layered, which is why different sources give different years.

Joseph Louis Gay-Lussac and Louis Jacques Thénard obtained an impure amorphous form in 1811 by reacting silicon tetrafluoride with potassium, but did not purify it or identify what they had. Jöns Jacob Berzelius repeated and extended the work in 1823-24, removed the contaminants by repeated washing, and characterised the product — which is why he is usually named as the discoverer. Crystalline silicon, the form that actually matters technologically, was not produced until Henri Sainte-Claire Deville made it in 1854. A card that must carry a single credit and a single date will pair one man with whichever of those years its source preferred.

The kilogram was redefined with a sphere of silicon-28

Until 2019 the kilogram was a lump of platinum-iridium in a vault near Paris. Replacing it required tying mass to a fundamental constant, and one of the two routes to doing so was to count atoms.

The Avogadro Project built spheres of silicon isotopically enriched to better than 99.99% silicon-28, using centrifuge cascades originally developed for uranium. Enrichment matters because natural silicon is a mixture of three isotopes, and you cannot convert a volume into a number of atoms unless you know precisely what those atoms weigh. The spheres were then polished to a roundness of a few tens of nanometres — among the most spherical objects ever made — so that their volume could be computed from optical interferometry, while X-ray diffraction gave the spacing of the crystal lattice. Volume divided by the volume per atom gives the count.

The result was a value for the Avogadro constant precise enough to be fixed by definition, and it is one half of the evidence base behind the redefinition; the Kibble balance supplied the other.

Why there are no silicon lasers

Silicon has a band gap of about 1.1 electronvolts, conveniently matched to the solar spectrum, which is one reason it dominates photovoltaics. But that gap is indirect: the lowest point of the conduction band does not sit at the same crystal momentum as the top of the valence band.

An electron dropping across a direct gap can emit a photon and conserve momentum in one step. In silicon it cannot; it needs a lattice vibration to make up the difference, and the requirement for a three-body coincidence makes light emission wildly improbable compared with losing the energy as heat. Silicon is consequently an excellent light detector and a hopeless light source.

This is the entire reason that a fibre-optic transmitter contains gallium arsenide or indium phosphide while the logic driving it is silicon, and why integrating the two has been an active engineering problem for forty years.

Glass houses built by algae

Diatoms build their cell walls from hydrated amorphous silica, in structures with pore patterns regular enough to have been used as test objects for microscope resolution in the nineteenth century. They are abundant enough to matter globally: diatoms account for a large share of marine primary production, and their sinking frustules are the main route by which dissolved silicon leaves the surface ocean.

Accumulated on the seafloor over geological time those shells become diatomite, a soft, porous, extremely low-density rock that is quarried by the million tonnes for filtration — most swimming pool and brewery filters run on it — and as a mild abrasive.

Inhaled crystalline silica dust is a different matter entirely. Silicosis, an irreversible fibrotic lung disease, is among the oldest recorded occupational illnesses, and it has returned in recent years among workers cutting engineered quartz worktops, a material with a far higher crystalline silica content than natural stone. Australia banned the manufacture and installation of engineered stone outright in 2024 on that evidence.