Germanium Process

The germanium process deposits or grows pure germanium — not the silicon-germanium alloy covered elsewhere, but the element that was actually the industry's first semiconductor, used in the point-contact and junction transistors of the late 1940s and early 1950s before silicon displaced it almost completely. Germanium never disappeared from the periodic table of useful fab materials; it disappeared from mainstream logic for one specific, fixable-sounding reason that turned out not to be fixable at all, and it has since come back in two roles that have nothing to do with the job it originally lost.

GeH4 precursor
CVD or epitaxial growth on Si/substrate
Low-temperature deposition
Ge melts at 938°C, well below Si
Two modern roles, same element
Ge photodetector
Strong near-IR absorption for optical links
Pure Ge channel
Highest hole mobility of any practical channel material
Silicon photonics receiver
Integrated optical interconnect
Ultra-high-mobility PMOS
Beyond what SiGe channels reach

Germanium lost the race to silicon for a reason that has nothing to do with the semiconductor itself: its native oxide is useless. GeO2, the oxide that forms on exposed germanium, is water-soluble and electrically unstable — it doesn't form the dense, self-limiting, electrically trustworthy passivating skin that SiO2 forms on silicon. Every early-era process step that depends on a reliable native or thermally grown oxide — the oxide process covered elsewhere, gate dielectric formation, masking for selective diffusion — works on silicon specifically because of that oxide's quality, and simply doesn't work the same way on germanium. By the early 1960s, the industry had moved almost entirely to silicon for exactly this reason, even though germanium's higher intrinsic carrier mobility made it, in isolation, the physically faster semiconductor.

Why Germanium Lost: the Native Oxide Problem a faster semiconductor, undone by an oxide the fab couldn't trust Silicon Si substrate SiO2 — dense, stable, insulating masks diffusion, insulates the gate, survives wet processing untouched Germanium Ge substrate GeO2 — water-soluble, unstable dissolves in wet steps, can't reliably mask diffusion or insulate a gate germanium's raw carrier mobility was never the problem — the oxide its surface forms on its own is

Germanium's modern comeback as a photodetector sidesteps the oxide problem entirely, because the job doesn't need a stable oxide at all. Germanium absorbs light strongly across the near-infrared wavelengths used in fiber-optic and silicon-photonics communication links, far more strongly than silicon does at those same wavelengths, which makes a small epitaxially grown germanium patch an efficient photodetector when integrated directly onto a silicon photonic chip. The detector only needs to absorb photons and generate carriers efficiently at the wavelength of interest — it isn't being used as a transistor channel requiring a gate oxide, so the one property that sank germanium's first run through the industry simply never comes into play.

The second comeback, as a pure-germanium transistor channel, does need to deal with the oxide problem, and the solution is to avoid growing one on germanium directly. Pure germanium offers the highest hole mobility of any semiconductor channel material practical for CMOS — higher than silicon, and higher than a SiGe channel's partial improvement — which makes it attractive for ultra-high-performance PMOS beyond what strain engineering alone can deliver. Modern germanium-channel devices solve the native-oxide problem not by fixing GeO2, but by depositing a separate high-k dielectric (hafnium-based, the same family used in HKMG gate stacks) directly on the germanium surface with careful interface passivation, bypassing the unreliable native oxide altogether rather than ever depending on it.

Germanium-on-insulator (GeOI) substrates extend the same idea to a full wafer, the way silicon-on-insulator does for silicon. A thin, high-quality germanium layer bonded or grown onto an insulating handle wafer gives device engineers a pure-germanium active layer to build transistors or photonic structures in, without needing bulk germanium crystal growth (which is far more expensive and defect-prone than silicon's mature Czochralski infrastructure) for every wafer.

Read the germanium process as a reminder that a material's fitness for a job is rarely decided by its headline property alone. Germanium's intrinsic mobility advantage over silicon was real in 1950 and is still real today — what changed was never the semiconductor, but which problems the industry needed solved: an unreliable native oxide ruled germanium out of the gate-dielectric-dependent CMOS era entirely, and only came back once two new jobs — photon absorption and a channel paired with someone else's high-k dielectric — stopped needing germanium's own oxide at all.

Take germanium process further

Ask the copilot about this term, or have our engineers assess it against your process.