Physically Motivated Recursively Embedded Atom Neural Networks: Incorporating Local Completeness and Nonlocality

TL;DR

This paper introduces a recursively embedded atom neural network that efficiently captures complete many-body correlations in interatomic potentials, addressing local completeness and nonlocality challenges.

Contribution

It proposes a novel neural network model that incorporates local completeness and nonlocality through recursive embedding, without explicitly computing high-order terms.

Findings

Successfully addresses local completeness and nonlocality issues

Provides a general framework for message-passing in atomistic models

Proves efficiency in incorporating many-body correlations

Abstract

Recent advances in machine-learned interatomic potentials largely benefit from the atomistic representation and locally invariant many-body descriptors. It was however recently argued that including three- (or even four-) body features is incomplete to distinguish specific local structures. Utilizing an embedded density descriptor made by linear combinations of neighboring atomic orbitals and realizing that each orbital coefficient physically depends on its own local environment, we propose a recursively embedded atom neural network model. We formally prove that this model can efficiently incorporate complete many-body correlations without explicitly computing high-order terms. This model not only successfully addresses challenges regarding local completeness and nonlocality in representative systems, but also provides an easy and general way to update local many-body descriptors to have a message-passing form without changing their basic structures.