Silicon germanium is a strained, lattice-mismatched alloy of silicon and germanium grown epitaxially on silicon wafers, and it has become the load-bearing substrate for two very different things at once: stacked-channel logic beyond the 2 nm node, and the leading semiconductor spin-qubit platforms.
Summary
The parameters that decide SiGe are all about disorder and thermal budget rather than about the idea. Germanium fraction sets strain, band offsets and valley splitting. Growth temperature sets Ge-Si interdiffusion: a segmented temperature-control scheme reported in 2026 lowered the Ge diffusion coefficient to 5.6-7% of its 650 C value, giving a 4 + 4 channel stack with fully coherent strain, 1.5-2.6 nm interface transition widths and 0.08 nm RMS roughness, though the bottom of the stack still saw Ge dilution from 20% to 18.5% from cumulative thermal exposure. Metamorphic buffers relax the mismatch by generating dislocations, and those dislocations do not stay put: X-ray nano-mapping of an Intel Si/SiGe qubit chip resolved crosshatch fine structure and permanently distorted lattice planes propagating through the heterostructure at 30 nm lateral and 200 nm depth resolution, with a calculated effect on qubit energy spectra.
For the alloy in its bulk role, random substitutional disorder is the dominant phonon scatterer, which is why SiGe is a thermoelectric material as well as a logic material; a 2026 substitutional atomic distance model reproduces measured SiGe thermal conductivities and attributes the reduction chiefly to alloy scattering.
Evidence base
Open questions
- Can growth changes remove the crosshatch and dislocation-induced strain that X-ray mapping found propagating through an industrial Si/SiGe heterostructure, or will per-device tuning such as gate-biased illumination remain necessary indefinitely?
- Does the segmented-temperature growth approach hold coherent strain and 20% Ge concentration through the full downstream thermal budget of a real sub-2 nm logic flow, not just an 8-layer test stack?
- Is random alloy disorder in the SiGe barrier an intrinsic, irreducible floor on valley-splitting and g-factor variation, or can engineered composition profiles such as Wiggle Wells trade it for something more tolerable?
- Does the Ge/SiGe hole-qubit route, with isotopic 70Ge purification and simplified single-layer gating, overtake Si/SiGe electron qubits on wafer-scale uniformity, given that hole sweet spots also vary site to site?
Assessment drafted 2026-08-31 from up to 18 KB sources using the technology-scorecard framework; scores are a draft read pending review.