AI and the World's Hard Problems · Chemistry & materials · arXiv · 2026-10-19
Scalable long-range machine learning interatomic potentials with electrical response for atomistic material modeling
The paper introduces DPA4C-LR, a machine-learning interatomic potential that handles long-range electrostatics and electrical response while allowing distributed evaluation. It separates long-range electrostatic forces and Born effective charges into a global calculation and local derivatives of learned latent charges. The model scales to 5.2 million atoms on four GPUs at about twice the cost of its short-range backbone, and it reproduces the field-dependent infrared spectra of water and the polarization hysteresis of a relaxor ferroelectric. It also predicts a room-temperature ionic conductivity for an amorphous Li-based compound.