Assessing Zero-Point Motion Effects in MC3D via Machine Learning Interatomic Potentials

by Karl Abdelnour (PSI/EPFL)

Wednesday 4 Feb 2026, 15:45 → 16:45 Europe/Zurich

OVGA/200

Density functional theory (DFT) enables accurate phonon calculations but remains computationally demanding for large-scale applications. Machine learning interatomic potentials (MLIPs) offer a more efficient alternative, but their application to phonon zero-point motion has not yet been extensively explored [1]. This work presents a first quantification of MLIP accuracy in estimating phonon zero-point motion, considering materials from the MC3D database [2].

[1] Loew et al., npj Computational Materials 11, 178 (2025) ;

[2] Huber et al., arXiv:2508.19223 (2025);