Learning QM/MM potential using equivariant multiscale model

• DOI: 10.1063/5.0205123
• Publication Date: 2024-06-03T09:44:23Z
• Authors: Yao-Kun Lei; Kiyoshi Yagi; Yuji Sugita
• Keywords: Operator (biology); Electrostatics

The machine learning (ML) method emerges as an efficient and precise surrogate model for high-level electronic structure theory. Its application has been limited to closed chemical systems without considering external potentials from the surrounding environment. To address this limitation incorporate influence of potentials, polarization effects, long-range interactions between a system its environment, first two terms Taylor expansion electrostatic operator have used extra input existing ML represent environments. However, high-order interaction is often essential account models based only on invariant features cannot capture significant distribution patterns potentials. Here, we propose novel that includes uses equivariant model, which can generate tensor covariant with rotations base model. Therefore, use multipole-expansion equation derive useful representation by accounting intermolecular interaction. Moreover, deal interactions, follow same strategy adopted target environment media. Our achieves higher prediction accuracy transferability among various media these modifications.