Vector Field Oriented Diffusion Model for Crystal Material Generation

Astrid Klipfel,Yaël Fregier,Adlane Sayede,Zied Bouraoui

2023-12-20

Abstract:Discovering crystal structures with specific chemical properties has become an increasingly important focus in material science. However, current models are limited in their ability to generate new crystal lattices, as they only consider atomic positions or chemical composition. To address this issue, we propose a probabilistic diffusion model that utilizes a geometrically equivariant GNN to consider atomic positions and crystal lattices jointly. To evaluate the effectiveness of our model, we introduce a new generation metric inspired by Frechet Inception Distance, but based on GNN energy prediction rather than InceptionV3 used in computer vision. In addition to commonly used metrics like validity, which assesses the plausibility of a structure, this new metric offers a more comprehensive evaluation of our model's capabilities. Our experiments on existing benchmarks show the significance of our diffusion model. We also show that our method can effectively learn meaningful representations.

Materials Science,Artificial Intelligence,Machine Learning

What problem does this paper attempt to address?

The main focus of this paper is the generation problem of crystal materials. Existing models have limitations in generating new crystal lattices, usually considering only atomic positions or chemical compositions, while ignoring crystal lattices. To address this issue, the paper proposes a probabilistic diffusion model that utilizes geometric invariant graph neural networks (GNNs) to simultaneously consider both atomic positions and crystal lattices for more comprehensive structure generation.

Specifically, the paper introduces a diffusion process on the lattice and plane to create atoms and unit cells. By using vector field-based neural networks to create lattices, the model can predict lattice parameters and maintain invariance under the E and SL3(Z) groups. In addition, the paper proposes a new evaluation metric called Frechet distance based on ALIGNN energy prediction (FAD) to more comprehensively evaluate the model's ability to generate crystals, instead of relying solely on traditional metrics such as structural rationality, density, and formation energy.

Experimental results show that the proposed diffusion model demonstrates significant performance improvements in existing benchmark tests, being able to effectively learn meaningful representations and generate crystals with specific chemical compositions, which other models cannot achieve. The paper also provides code and an independent Python toolkit for reproducibility and further research by other researchers.

In summary, the main contributions of the paper include proposing a novel method for generating crystal materials, introducing new diffusion processes and evaluation metrics, and demonstrating the advantages of the proposed model in generating crystal structures with specific chemical properties.