A Systematic Approach to Generating Accurate Neural Network Potentials: the Case of Carbon

TL;DR

This paper presents a systematic, self-consistent method for developing accurate neural network potentials for materials, demonstrated on carbon, that reproduces first-principles results across various phases and properties.

Contribution

It introduces a novel approach combining crystal structure prediction and unsupervised data analysis to create transferable neural network potentials.

Findings

Accurately reproduces elastic and vibrational properties of carbon allotropes.

Successfully models energy ordering and structures of crystalline and amorphous phases.

Demonstrates transferability and efficiency of the proposed neural network potential.

Abstract

Availability of affordable and widely applicable interatomic potentials is the key needed to unlock the riches of modern materials modelling. Artificial neural network based approaches for generating potentials are promising; however neural network training requires large amounts of data, sampled adequately from an often unknown potential energy surface. Here we propose a self-consistent approach that is based on crystal structure prediction formalism and is guided by unsupervised data analysis, to construct an accurate, inexpensive and transferable artificial neural network potential. Using this approach, we construct an interatomic potential for Carbon and demonstrate its ability to reproduce first principles results on elastic and vibrational properties for diamond, graphite and graphene, as well as energy ordering and structural properties of a wide range of crystalline and amorphous phases.