Predicting charge density distribution of materials using a local-environment-based graph convolutional network

TL;DR

This paper presents a machine learning approach using graph convolutional neural networks to efficiently predict charge density distributions in materials, enabling scalable and transferable analysis based on crystal structures.

Contribution

Introduces a local-environment-based graph neural network model for charge density prediction that scales linearly and exhibits high transferability across different materials.

Findings

Prediction accuracy comparable to DFT calculations.

Scales as O(N) with system size.

High transferability across diverse structures.

Abstract

Electron charge density distribution of materials is one of the key quantities in computational materials science as theoretically it determines the ground state energy and practically it is used in many materials analyses. However, the scaling of density functional theory calculations with number of atoms limits the usage of charge-density-based calculations and analyses. Here we introduce a machine learning scheme with local-environment-based graphs and graph convolutional neural networks to predict charge density on grid-points from crystal structure. We show the accuracy of this scheme through a comparison of predicted charge densities as well as properties derived from the charge density, and the scaling is O(N). More importantly, the transferability is shown to be high with respect to different compositions and structures, which results from the explicit encoding of geometry.