Silicon vs Germanium

The first transistor was made of germanium, not silicon. Bardeen and Brattain's point-contact device of December 1947 used a germanium crystal, and so did the first decade of commercial transistors. Silicon took the industry anyway, and the reason it won is not a property of silicon at all — it is a property of silicon's oxide.

For anything you are designing today, the answer is silicon unless you have a specific reason that silicon cannot serve, and there are exactly three or four such reasons. This page is about what they are.

Germanium lost on an oxide

Heat silicon in oxygen and it grows silicon dioxide: a hard, dense, electrically insulating glass that adheres perfectly, tolerates high temperature, blocks dopant diffusion where you want it blocked, and forms an interface with the underlying crystal clean enough to build a transistor gate on.

That single material made the planar process possible — patterning an oxide layer, diffusing through the windows, and leaving the oxide behind as passivation — and it made the metal-oxide-semiconductor field-effect transistor possible, since the gate insulator in a MOSFET was, for fifty years, thermally grown silicon dioxide.

Germanium has no equivalent. Germanium dioxide is soluble in water, so an oxide-passivated germanium device does not survive an ordinary cleaning step, and the oxide is thermally unstable in a way that makes controlled growth difficult. Germanium could be made into individual transistors; it could not be made into the integrated process that silicon supported. The entire semiconductor industry rests on a compound rather than on the element in its name.

Two-thirds of an electronvolt is the difference between working and leaking

Silicon's forbidden gap is about 1.12 electronvolts at room temperature; germanium's is 0.66. Narrower gap means more electrons are thermally promoted across it at any given temperature, and because the relationship is exponential, germanium carries something like three orders of magnitude more intrinsic carriers than silicon at 300 K.

In a device, those carriers are reverse leakage current. Germanium junctions leak measurably at room temperature and worse as they warm, roughly doubling for every ten degrees, so germanium circuits become unreliable somewhere around 70 to 85 °C. Silicon runs comfortably past 150.

That ceiling is what actually killed germanium in the 1960s. Military and industrial specifications demanded operation over temperature ranges germanium simply could not hold, and no amount of circuit design fixes a leakage current that comes from the band structure.

The narrow gap does buy one thing worth having: a germanium diode conducts at about a fifth to a third of a volt, against roughly 0.6 to 0.7 for silicon. In a detector circuit with only microwatts of signal to work with, that lower turn-on voltage is decisive, which is why germanium point-contact diodes never quite disappeared from crystal sets and signal detectors.

Where germanium's mobility still wins

Charge carriers move faster in germanium than in silicon — electrons more than twice as fast, holes around four times. Hole mobility is germanium's standout figure, and it is the reason germanium has come back into mainstream silicon manufacturing rather than staying a museum piece.

Silicon-germanium alloys grown as strained layers inside otherwise ordinary silicon processes are used to boost p-channel performance in advanced logic, and silicon-germanium heterojunction bipolar transistors are the backbone of radio-frequency front ends, high-speed serial links and the 77 GHz radar modules in cars. Those parts are made in silicon fabs, on silicon wafers, with germanium present as an alloying element in specific layers. Germanium's second career is as an ingredient rather than a substrate.

Infrared is germanium's other career

Visible light does not get through germanium at all, while wavelengths from roughly 2 to 14 micrometres pass almost unhindered — a window that covers the long-wave infrared band thermal cameras work in. Its refractive index of about 4 is the highest of any common lens material, so germanium optics bend light strongly with shallow surfaces, at the cost of needing good antireflection coating. Essentially every uncooled thermal imaging lens is germanium.

Two more optical uses account for a large share of world consumption. Germanium dioxide is doped into the core of optical fibre to raise its refractive index above the cladding, which is what confines the light. And high-purity germanium crystals, cooled with liquid nitrogen, make gamma-ray spectrometers with energy resolution nothing else approaches — the instrument of choice when you need to identify a radionuclide rather than merely detect it.

Germanium wafers also serve as the substrate for the triple-junction solar cells that power spacecraft, where efficiency matters far more than the cost of the wafer.

Mendeleev's prediction, Winkler's mineral

Germanium is one of the periodic table's great vindications. Mendeleev left a gap below silicon in 1871 and described the missing element in detail — he called it ekasilicon and predicted an atomic weight near 72 and a density around 5.5. Clemens Winkler isolated it from the mineral argyrodite in 1886 and found an atomic weight of 72.6 and a density of 5.35. The table had described an element nobody had ever seen.

One is sand; the other comes out of zinc

Silicon is the second most abundant element in the crust and its raw material costs almost nothing. What is expensive is purity: semiconductor-grade polysilicon is refined to parts per billion and then pulled into single crystals 300 millimetres across. Worth noting too that most silicon produced is not for electronics at all — ferrosilicon for steelmaking and silicones for everything else consume far more tonnage than chips ever will.

Germanium has no ore of its own. It is recovered as a by-product of zinc refining and from coal fly ash, in quantities of a few hundred tonnes a year worldwide, and refined material sells for a thousand times the price of silicon by weight. Production is concentrated enough that germanium appears on critical-materials lists in Europe and North America, and export controls announced by China in 2023 covering germanium and gallium made that concentration a live commercial issue rather than a theoretical one.

One historical footnote connects the two: zone refining, the technique that made ultra-high-purity semiconductors possible at all, was developed by William Pfann at Bell Labs in the early 1950s on germanium, and then applied to silicon.

Which semiconductor for which device

• Anything digital, anything integrated, anything that must run warm — silicon.
• A radio-frequency front end or a millimetre-wave radar — silicon-germanium, in a silicon fab.
• A thermal imaging lens or window — germanium, with no realistic alternative.
• High-resolution gamma spectroscopy — high-purity germanium, cooled.
• A detector diode working with vanishingly small signals — germanium, for the low forward drop.
• A solar cell that has to work in orbit — a germanium substrate under III-V junctions.
• Anything where cost per wafer matters — silicon, by three orders of magnitude.