Dátum

Előadó: Connor Williamson (Wigner FK SZFI)

Előadásának címe: Finite-Temperature Behaviour of Interacting Defects via Machine-Learned Interatomic Potentials

Dátum:  2026. november 24. kedd, 10 óra

Helyszín: 1-es épület tanácsterem

Összefoglaló:

Predicting the finite-temperature behaviour of spin defects remains a bottleneck in the computational design of solid-state qubits and, in turn, of qubit arrays. Accurately capturing the vibrational modes of these defective materials requires supercells of hundreds to thousands of atoms to avoid spurious defect-image interactions and to capture coupling to the host’s phonon continuum. Furthermore, the optical line shifts and broadening observed at T > 0 K often demand a treatment beyond the harmonic approximation, placing many systems beyond practical ab initio reach. We present an approach that uses machine-learned interatomic potentials (MLIPs) to close this gap for both single and multiple defects in diamond, silicon carbide and silicon. These hosts form a series of increasingly soft lattices in which thermal population and anharmonicity grow in importance. MLIPs are first benchmarked for each host against the Raman-active zone-centre modes and full phonon dispersions, and will then be applied to supercells large enough to converge the vibrational properties, resolving the quasi-local vibrational modes and classifying them by defect point-group symmetry. Quantum and thermal anharmonic effects are incorporated through the stochastic self-consistent harmonic approximation and its time-dependent extension, yielding temperature-dependent mode frequencies, linewidths and vibrational spectra.