Symmetry-Aware Bayesian Flow Networks for Crystal Generation

2025-02-05

AI Summary

To address the inefficiency of conventional trial-and-error approaches in navigating the vast structural space of crystalline materials, this paper proposes Symmetry-Aware Bayesian Flow Networks (SymmBFN), the first method to explicitly incorporate space-group symmetry priors into the Bayesian flow modeling framework. SymmBFN achieves this through space-group equivariant representations, discrete-structure probability flow modeling, and property-conditioned sampling. It faithfully reproduces experimentally observed space-group distributions and enables controllable generation targeting properties such as bandgap and density. Compared to state-of-the-art alternatives, SymmBFN accelerates generation by over 50× while significantly reducing space-group distribution error. It thus overcomes two key limitations of existing generative models: simultaneous preservation of crystallographic symmetry and precise property-conditioned control.

Technology Category

• Machine Learning: Deep Generative Models & Autoencoders
• Search and Optimization: Sampling/Simulation-based Search
• Natural Language Processing: Generation

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Web Mining and Content Analysis: Web data generation and simulation

Abstract

The discovery of new crystalline materials is essential to scientific and technological progress. However, traditional trial-and-error approaches are inefficient due to the vast search space. Recent advancements in machine learning have enabled generative models to predict new stable materials by incorporating structural symmetries and to condition the generation on desired properties. In this work, we introduce SymmBFN, a novel symmetry-aware Bayesian Flow Network (BFN) for crystalline material generation that accurately reproduces the distribution of space groups found in experimentally observed crystals. SymmBFN substantially improves efficiency, generating stable structures at least 50 times faster than the next-best method. Furthermore, we demonstrate its capability for property-conditioned generation, enabling the design of materials with tailored properties. Our findings establish BFNs as an effective tool for accelerating the discovery of crystalline materials.

Problem

• Generates new crystalline materials efficiently
• Incorporates structural symmetries in predictions
• Enables design of materials with tailored properties

Innovation

• Symmetry-aware Bayesian Flow Networks
• Efficient crystalline material generation
• Property-conditioned material design