Equivariant graph neural networks for fast electron density estimation of molecules, liquids, and solids

TL;DR

This paper introduces an equivariant graph neural network framework for rapid and accurate prediction of electron density distributions in molecules, liquids, and solids, significantly outperforming traditional DFT in speed and accuracy.

Contribution

The authors develop a novel equivariant graph neural network model that predicts electron densities at query points, achieving state-of-the-art accuracy across diverse datasets and greatly reducing computational time.

Findings

Model exceeds variability in DFT electron densities for QM9 molecules.

Achieves state-of-the-art accuracy on molecules, electrolytes, and battery cathodes.

Runs orders of magnitude faster than traditional DFT calculations.

Abstract

Electron density $\rho(\vec{r})$ is the fundamental variable in the calculation of ground state energy with density functional theory (DFT). Beyond total energy, features and changes in $\rho(\vec{r})$ distributions are often used to capture critical physicochemical phenomena in functional materials. We present a machine learning framework for the prediction of $\rho(\vec{r})$. The model is based on equivariant graph neural networks and the electron density is predicted at special query point vertices that are part of the message passing graph, but only receive messages. The model is tested across multiple data sets of molecules (QM9), liquid ethylene carbonate electrolyte (EC) and LixNiyMnzCo(1-y-z)O2 lithium ion battery cathodes (NMC). For QM9 molecules, the accuracy of the proposed model exceeds typical variability in $\rho(\vec{r})$ obtained from DFT done with different exchange-correlation functionals. The accuracy on all three datasets is beyond state of the art and the computation time is orders of magnitude faster than DFT.