Combining the D3 dispersion correction with the neuroevolution machine-learned potential

TL;DR

This paper introduces a combined NEP-D3 model that enhances the modeling of long-range dispersion interactions in atomistic simulations, improving accuracy for materials like graphene and metal-organic frameworks.

Contribution

It proposes integrating the D3 dispersion correction with the neuroevolution potential, enabling better long-range interaction modeling in machine-learned potentials.

Findings

Improved binding and sliding energy descriptions in bilayer graphene.

Dispersion interactions reduce thermal conductivity by about 10% in metal-organic frameworks.

D3 correction implementation is compatible with multiple exchange-correlation functionals.

Abstract

Machine-learned potentials (MLPs) have become a popular approach of modelling interatomic interactions in atomistic simulations, but to keep the computational cost under control, a relatively short cutoff must be imposed, which put serious restrictions on the capability of the MLPs for modelling relatively long-ranged dispersion interactions. In this paper, we propose to combine the neuroevolution potential (NEP) with the popular D3 correction to achieve a unified NEP-D3 model that can simultaneously model relatively short-ranged bonded interactions and relatively long-ranged dispersion interactions. We show the improved descriptions of the binding and sliding energies in bilayer graphene can be obtained by the NEP-D3 approach compared to the pure NEP approach. We implement the D3 part into the GPUMD package such that it can be used out of the box for many exchange-correlation functionals. As a realistic application, we show that dispersion interactions result in approximately a 10% reduction in thermal conductivity for three typical metal-organic frameworks.