A robust sub-Kelvin multiscale simulation paradigm to assess device-to-device variability in spin qubits

Integrating thousands to millions of qubits is crucial for practical quantum computers. Spin qubits (SQ) are promising due to their compact size and CMOS-compatible fabrication, but device-to-device variability (DtDV) poses significant challenges such as reduced gate fidelities and lower yields. While SQs based on quantum dots (QDs) have dimensions at the 100-nm-scale, major sources of variability—random alloying, surface roughness, and atomic impurities—occur at the atomistic scale (angstroms to nanometers). Accurate predictions of DtDV therefore require multiscale calculations.

We present a robust multiscale simulation paradigm to tackle this challenge. At the 100-nm-scale, we resolve device electrostatics by solving a nonlinear Poisson equation using the finite-element method; adaptive meshing ensures reliable convergence at sub-Kelvin temperatures. At the atomistic scale, we model the QD with an atomic structure builder that incorporates random alloying and surface roughness. We describe strain and electronic properties using the Keating and tight-binding models. We extract eigenenergies and eigenstates near the Fermi energy using a parallelized eigensolver, allowing the prediction of device figures of merit like valley splitting and g-factor. Finally, by repeating this process across random atomic configurations, we quantitatively assess DtDV, aiding in the design of more reliable SQ devices and advancing scalable quantum computing.