A Gap Worth Trillions

Materials are usually sorted into conductors and insulators, and the distinction is about what electrons are allowed to do. In a conductor, electrons are already free to move in response to a field. In an insulator, the energy needed to liberate one is enormous: far beyond anything ordinarily available.

Between them sits a third category with a small gap. A semiconductor's electrons are bound, but only just: modest heat, a photon of visible light, or a deliberately introduced impurity can put carriers into motion. The gap is small enough to be crossed by things we can arrange. That is the entire basis of electronics, because conductivity that can be controlled from outside is a switch, and a switch is all a computer is.

The control comes from doping. Silicon has four outer electrons and bonds neatly with four neighbours. Substitute an atom with five, and the spare electron is loosely held and easily freed: the region now carries negative charge, and is called n-type. Substitute an atom with three and there is a vacancy, which behaves for all practical purposes like a mobile positive charge: p-type. The concentrations involved are around one impurity atom in ten million, which is the sense in which the semiconductor industry is really a purity industry. You cannot add one atom in ten million to something already containing a hundred times that many contaminants.

Neither doped region is useful by itself. The interesting structure is the junction where the two meet. Carriers diffuse across the boundary, leave behind fixed charges, and establish an internal electric field with a depletion region between. Current now flows readily in one direction and hardly at all in the other. That is a diode. Add a third region with a control terminal and a small voltage can govern a large current: a transistor. Shine light on the same junction so that photons liberate carriers, which the built-in field then sweeps apart, and the device generates power instead of consuming it. A solar cell and a processor are the same physics run in opposite directions.

The band gap that decides switching behaviour in one case decides absorption in the other. Photons carrying less energy than the gap pass straight through the material and contribute nothing. Photons carrying much more are absorbed with the surplus lost as heat. This trade-off sets a theoretical ceiling for a single-junction cell in the low thirties of per cent, which is why laboratory records come from stacking materials with different gaps to capture more of the spectrum.

A final point that says something general about technology. Silicon is not the best semiconductor by most physical measures: germanium has higher carrier mobility, gallium arsenide is better for high-frequency and optical work. Silicon won because it grows an excellent, stable, insulating oxide on its own surface, which makes it possible to build and isolate millions of devices on a single wafer reliably and cheaply.

Manufacturability beat raw performance. It usually does, and the history of materials is largely a list of better substances that lost to more tractable ones.