AI and the World's Hard Problems · Physics · arXiv · 2026-10-20
NEP-kappa 2.0: A unified framework for phonon transport and nonequilibrium thermalization with machine-learning interatomic potentials
The authors present NEP-kappa 2.0, a framework that integrates DFT, material-specific machine-learning potentials and pretrained force models for phonon calculations. Applied to 3C-SiC, BAs and Cs2AgInCl6, it covers phonon transport, higher-order scattering and time-dependent dynamics. In selectively excited 3C-SiC, about half of the injected optical-phonon energy reached acoustic modes within roughly 3.75 ps, and the acoustic branches showed response times differing by a factor of five.