Silicon: The Ordinary Element That Everything Runs On
Sand is mostly silicon dioxide. The wafer in a processor is the same element, purified until fewer than one atom in ten billion is anything else, grown into a single crystal, and sliced. That gap between the raw abundance and the required perfection is the whole story of silicon: the element is everywhere and free, and the useful form of it is one of the most demanding manufactured materials on Earth. This page explains what makes it a semiconductor at all, why it displaced its competitors, and how it functions in the graphene and silicon multilayer architecture our own research is built on.
What Makes It a Semiconductor
A silicon atom has four electrons in its outer shell, and in a crystal each one bonds to four neighbours in a tetrahedral arrangement - the same structure diamond has. Every outer electron is committed to a bond, so at absolute zero a perfect silicon crystal is an insulator with no free charge carriers at all.
Raise the temperature and some electrons acquire enough thermal energy to break free of their bonds and move through the crystal. Each departure leaves behind a vacancy that neighbouring electrons can hop into, and that vacancy behaves like a positively charged particle in its own right. This is the hole, and it is not a bookkeeping trick: holes have a measurable effective mass and mobility.
The energy needed to promote an electron out of a bond is the band gap, about 1.12 electronvolts in silicon at room temperature. That number is the single most consequential property of the material. It is large enough that thermally generated carriers are rare, so the crystal is not a conductor, and small enough that carriers can be created deliberately - by doping, by an applied field, or by absorbing a photon.
That intermediate position is what the word semiconductor means. A metal has carriers available with no threshold at all. An insulator has a gap so wide that nothing crosses it under ordinary conditions. Silicon sits between, and its conductivity can be adjusted across many orders of magnitude by design.
Why Silicon Won
The first transistors were germanium. Germanium has higher carrier mobility than silicon, a smaller band gap, and was easier to purify with 1940s technology. On the physics alone it was the better choice, and for about a decade it was the standard.
Silicon displaced it because of its oxide. Heat silicon in oxygen and it grows a layer of silicon dioxide that is chemically stable, mechanically robust, and one of the best electrical insulators known - and it grows directly on the crystal with an interface clean enough that almost no electrons get trapped there. Germanium oxide is water-soluble and useless for the same purpose.
That single material accident made the planar process possible: grow an oxide, cut windows in it, dope through the windows, grow more oxide, repeat. Every integrated circuit ever manufactured rests on it. It is a good illustration of how technology choices are made - not by picking the best-performing option, but by picking the one whose whole system of properties can be manufactured.
The second reason is purity. Silicon can be zone-refined and grown by the Czochralski method to a level where impurities are counted in parts per billion or better. Sixty years of process refinement have compounded that advantage to the point where switching to a theoretically superior semiconductor would mean abandoning an industrial base that no alternative can match.
The Indirect Gap and What It Costs
Silicon's band gap is indirect, which means the lowest available state in the conduction band does not sit at the same momentum as the highest state in the valence band. An electron cannot simply absorb a photon and cross; it also needs a lattice vibration, a phonon, to supply the missing momentum.
Requiring two things to coincide makes the process much less likely. In practice this means silicon absorbs light weakly, which is why a silicon solar cell needs to be around 200 micrometres thick to capture most of the usable spectrum, against under a micrometre for a direct-gap material such as perovskite.
The same physics runs in reverse: silicon is a very poor light emitter, which is why LEDs and laser diodes are made from gallium arsenide and its relatives rather than from silicon, and why integrating optics with silicon electronics has been a decades-long research problem.
For absorption-based devices the indirect gap is a cost that has to be engineered around, through surface texturing, back reflectors and thicker absorber layers. For the photoelectric process itself nothing changes - a photon above the gap can still promote an electron. It is the probability, not the possibility, that the indirect gap reduces.
Purity, Crystals and Thin Films
Metallurgical-grade silicon, reduced from quartz in an arc furnace, is about 98 percent pure and useless for electronics. Refining it to electronic grade means converting it to a volatile compound, distilling that, and depositing the silicon back out - a process that costs far more energy than mining the raw material ever did.
The result is then melted and pulled slowly from the melt as a single crystal, rotating as it goes, in the Czochralski process. A modern ingot is a cylinder 300 millimetres across and metres long in which the crystal lattice is continuous from end to end. Float-zone refining produces even purer material for specialist uses.
Not every application needs that. Thin-film silicon, deposited as an amorphous or microcrystalline layer, has no long-range order and much worse carrier transport, but it can be laid down cheaply over large areas at low temperature. Which form is appropriate depends entirely on whether carriers have to travel far.
This distinction matters for layered architectures. In a multilayer only a few nanometres thick per layer, carriers never travel far within the silicon, so the demands on crystalline perfection are different from those on a wafer-based device. The engineering question moves from the bulk material to the interfaces.
Silicon in a Layered Architecture
In the neutrinovoltaic multilayer that our research concerns, silicon alternates with doped graphene. The two materials could hardly be more different electronically: silicon has a 1.12 eV gap and three-dimensional bonding, while graphene has no gap at all and conducts within a single atomic plane.
Putting them in contact creates a heterojunction, and heterojunctions are where the interesting behaviour of layered devices lives. The difference in work function between the two materials sets up a built-in electric field at each interface, and it is that field, not either material alone, that can separate charge carriers.
It is worth being precise about what this does and does not imply. A built-in field at a junction is an equilibrium property. It exists whether or not anything is happening, it does not do work by itself, and a device made only of dissimilar contacts in a closed loop produces no current at all - a point our free energy page treats in general terms. Something has to disturb the equilibrium continuously for a current to flow, and identifying that something is the actual research question.
What silicon contributes is a well-understood, manufacturable partner with a gap in a useful range and sixty years of interface engineering behind it. That is a reasonable engineering choice. It is not, by itself, a claim about performance - and the honest version of the research keeps those two statements apart.
Frequently asked questions
Why is silicon used everywhere if it is not the best semiconductor?
Because of its oxide. Silicon dioxide is an excellent insulator that grows directly on the crystal with a very clean interface, which makes the planar manufacturing process possible. Germanium has higher mobility but no usable oxide.
What does the band gap of 1.12 eV mean in practice?
It is the energy needed to free an electron from a bond. Large enough that silicon is not a conductor, small enough that carriers can be created deliberately by doping, by a field, or by absorbing a photon above that energy.
Why do silicon solar cells have to be so thick?
Because silicon's band gap is indirect, so absorbing a photon also requires a lattice vibration to supply momentum. That makes absorption weak, and a cell needs roughly 200 micrometres to capture most of the usable spectrum.
How pure does electronic-grade silicon have to be?
Around one impurity atom in 10^9 to 10^11, then grown as a single crystal. Refining it to that level costs far more energy than obtaining the raw material, which is abundant.
What does silicon do in a graphene multilayer?
It forms a heterojunction with the graphene layers. The work function difference between the two sets up a built-in field at each interface. That field is an equilibrium property - it does not generate current on its own, and something must disturb the equilibrium continuously for anything to flow.
