WyCryst: Wyckoff Inorganic Crystal Generator Framework

Abstract

Generative design marks a significant data-driven advancement in the exploration of novel inorganic materials, which entails learning the symmetry equivalent to the crystal structure prediction (CSP) task and subsequent learning of their target properties. Generative models have been developed in the last few years that use custom Variational Autoencoders (VAEs), Generative Adversarial Networks (GANs), and diffusion models. While periodicity and global Euclidian symmetry in three dimensions through translations, rotations and reflections have recently been accounted for, symmetry constraints within allowed space groups have not. This is especially important because the final step involves energy relaxation on the generated crystal structures to find the relaxed crystal structure, typically using Density Functional Theory (DFT). To address this explicitly, we introduce a generative design framework (WyCryst), composed of three pivotal components: 1) a Wyckoff position based inorganic crystal representation, 2) a property-directed VAE model and 3) an automated DFT workflow for structure refinement. Our model selectively generates materials that follow the ground truth of unit cell space group symmetry by encoding the Wyckoff representation for each space group. We successfully reproduce a variety of existing materials: CaTiO3 (space group, SG No. 62 and 221), CsPbI3 (SG No. 221), BaTiO3 (SG No. 160), and CuInS2 (SG No.122) for both ground state as well as polymorphic structure predictions. We also generate several new ternary materials not found in the inorganic materials database (Materials Project), which are proved to be stable, retaining their symmetry, and we also check their phonon stability, using our automated DFT workflow highlighting the validity of our approach. We believe our symmetry-aware WyCryst takes a vital step towards AI-driven inorganic materials discovery.