Reproduce “Long-range electrostatics for machine learning interatomic potentials is easier than we thought” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Dongjin Kim, Bingqing Cheng
The Latent Ewald Summation (LES) framework provides a computationally efficient method for integrating long-range electrostatics into machine learning interatomic potentials. It enhances the accuracy of predicted energies and forces while enabling the reliable inference of electrical response properties like Born Effective Charges and infrared spectra from standard energy and force training data.
Key insights
- Integrating long-range electrostatics into MLIPs can be achieved effectively and simply through physics-guided design principles, proving to be easier and more broadly applicable than previously thought.
- Accurate electrical response properties, such as polarization and Born Effective Charges, can naturally emerge from latent charges inferred purely from standard energy and force training data.
- The computational overhead introduced by the Ewald summation for long-range interactions is marginal, preserving the efficiency of MLIPs while extending their physical accuracy.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.