Learning Long-Range Interactions in Equivariant Machine Learning Interatomic Potentials via Electronic Degrees of Freedom

View Full Paper

Ask AI

Bookmark

Authors

Discussion

Novel approach incorporates long-range interactions in machine learning models, improving energy predictions in materials science.

This new model demonstrates improved energy predictions, achieving higher accuracy using a 4 Å cutoff compared to a 6 Å cutoff of existing models.
Outperforming both short-range equivariant and long-range invariant MLIPs, it exhibits superior force predictions across varied data sets.
The method employs a charge equilibration scheme, explicitly treating long-range Coulomb interactions through electronic degrees of freedom.
Implications suggest a more accurate simulation capability for materials with long-range interactions and compositional heterogeneity.

Maruf et al. (2025) studied this question.

View Full Paper

Ask AI

Bookmark