Generating Symmetric Materials using Latent Flow Matching

TL;DR

This paper introduces SymADiT, a symmetry-aware generative model for materials that improves realism by enforcing symmetry constraints based on Wyckoff positions and space groups, outperforming prior models.

Contribution

The paper presents SymADiT, a novel symmetry-aware generative model for materials based on latent flow matching, enhancing the realism and symmetry properties of generated materials.

Findings

SymADiT generates stable, symmetric materials with realistic symmetry properties.

The model outperforms symmetry-agnostic approaches in benchmarks.

SymADiT uses a simple Transformer architecture for effective symmetry-aware generation.

Abstract

Tackling the task of materials generation, we aim to enhance the previously proposed All-atom Diffusion Transformer (ADiT) by introducing SymADiT, a symmetry-aware variant. To do so, we use a representation of materials based on Wyckoff positions. We follow ADiT and perform generative modelling in latent space, adapted to our symmetry-aware representation. By forcing the output of the generative model to adhere to the symmetry restrictions imposed by the generated crystal's space group and each atom's Wyckoff-position, the generated materials exhibit more realistic symmetry properties. We benchmark our method against both symmetry-aware and symmetry-agnostic models for materials generation and show competitive performance, generating stable, symmetric materials with a simple Transformer architecture.