Diffusion probabilistic models enhance variational autoencoder for crystal structure generative modeling

TL;DR

This paper introduces a diffusion probabilistic approach to enhance a variational autoencoder for crystal structure generation, resulting in more accurate and ground-state-like structures compared to previous methods.

Contribution

The study proposes a novel DP-based denoising method within CDVAE, significantly improving the accuracy of generated crystal structures and their proximity to ground states.

Findings

Generated structures are statistically comparable to original CDVAE outputs.

DP-CDVAE structures are closer to ground states than those from the original CDVAE.

Energy differences are reduced by an average of 68.1 meV/atom.

Abstract

The crystal diffusion variational autoencoder (CDVAE) is a machine learning model that leverages score matching to generate realistic crystal structures that preserve crystal symmetry. In this study, we leverage novel diffusion probabilistic (DP) models to denoise atomic coordinates rather than adopting the standard score matching approach in CDVAE. Our proposed DP-CDVAE model can reconstruct and generate crystal structures whose qualities are statistically comparable to those of the original CDVAE. Furthermore, notably, when comparing the carbon structures generated by the DP-CDVAE model with relaxed structures obtained from density functional theory calculations, we find that the DP-CDVAE generated structures are remarkably closer to their respective ground states. The energy differences between these structures and the true ground states are, on average, 68.1 meV/atom lower than those generated by the original CDVAE. This significant improvement in the energy accuracy highlights the effectiveness of the DP-CDVAE model in generating crystal structures that better represent their ground-state configurations.