Environment-adaptive machine learning potentials

The development of interatomic potentials that can effectively capture a wide range of atomic environments is a complex challenge due to several reasons. Materials can exist in numerous structural forms (e.g., crystalline, amorphous, defects, interfaces) and phases (solid, liquid, gas, plasma). Each form may require different treatment in potential modeling to reflect the real physical behavior correctly. Atoms interact through various forces such as electrostatic, van der Waals, ionic bonding, covalent bonding, and metallic bonding, which manifest differently depending on the chemical elements and their electronic structures. Furthermore, the effective interaction among atoms can change with external conditions like temperature, pressure, and chemical environment. Consequently, creating an interatomic potential that performs well across diverse conditions is difficult since optimizing the potential for one set of conditions can lead to a trade-off in the accuracy of predicted properties associated with other conditions. In this paper, we present a method to construct accurate, efficient and transferable interatomic potentials by adapting to the local atomic environment of each atom within a system. The collection of atomic environments of interest is partitioned into several clusters of atomic environments. Each cluster represents a distinctive local environment and is used to define a corresponding local potential. We introduce a many-body many-potential expansion to smoothly blend these local potentials to ensure global continuity of the potential energy surface. This is achieved by computing the probability functions that determine the likelihood of an atom belonging to each cluster. We apply the environment-adaptive machine learning potentials to predict observable properties for Ta element and InP compound, and compare them with density functional theory calculations.