Open Materials Generation with Stochastic Interpolants
🤖 AI Summary
This work addresses generative discovery of inorganic crystalline materials by modeling the manifold of stable crystal structures within an infinite design space. We propose OMG, a unified generative framework that extends stochastic interpolation (SI) to periodic crystal modeling for the first time. OMG integrates equivariant graph neural networks, periodic boundary handling, and joint continuous–discrete flow matching, supporting both diffusion- and flow-based paradigms. It simultaneously generates atomic species, fractional coordinates, and lattice vectors while strictly preserving translational, rotational, and permutation symmetries inherent to crystals. On crystal structure prediction (CSP) and de novo generation (DNG) benchmarks, OMG achieves new state-of-the-art performance—outperforming dedicated diffusion-only or flow-only approaches—and yields structures with significantly improved thermodynamic stability and chemical novelty.
Application Category
📝 Abstract
The discovery of new materials is essential for enabling technological advancements. Computational approaches for predicting novel materials must effectively learn the manifold of stable crystal structures within an infinite design space. We introduce Open Materials Generation (OMG), a unifying framework for the generative design and discovery of inorganic crystalline materials. OMG employs stochastic interpolants (SI) to bridge an arbitrary base distribution to the target distribution of inorganic crystals via a broad class of tunable stochastic processes, encompassing both diffusion models and flow matching as special cases. In this work, we adapt the SI framework by integrating an equivariant graph representation of crystal structures and extending it to account for periodic boundary conditions in unit cell representations. Additionally, we couple the SI flow over spatial coordinates and lattice vectors with discrete flow matching for atomic species. We benchmark OMG's performance on two tasks: Crystal Structure Prediction (CSP) for specified compositions, and 'de novo' generation (DNG) aimed at discovering stable, novel, and unique structures. In our ground-up implementation of OMG, we refine and extend both CSP and DNG metrics compared to previous works. OMG establishes a new state-of-the-art in generative modeling for materials discovery, outperforming purely flow-based and diffusion-based implementations. These results underscore the importance of designing flexible deep learning frameworks to accelerate progress in materials science.
Problem
Research questions and friction points this paper is trying to address.
Generative design of inorganic crystals
Bridging distributions with stochastic interpolants
Advancing computational materials discovery
Innovation
Methods, ideas, or system contributions that make the work stand out.
Stochastic interpolants bridge distributions
Equivariant graph representation for crystals
Periodic boundary conditions in unit cells
🔎 Similar Papers
No similar papers found.
