Generative structure search for efficient and diverse discovery of molecular and crystal structures

2026-04-30

AI Summary

Efficient and diverse discovery of stable and metastable structures in high-dimensional energy landscapes is hindered by the high computational cost and data dependence of conventional methods. This work proposes a Generative Structure Search (GSS) framework that unifies diffusion-based generative models with physics-guided stochastic structure search into a synergistic sampling process. By learning a score field that integrates data-driven priors and physical force–guided optimization, GSS substantially enhances sampling efficiency while preserving the ability to explore local energy minima. Applied to molecular and crystalline systems, the method discovers diverse metastable configurations at less than one-tenth the sampling cost of random search and demonstrates strong generalization to compositions outside the training distribution.

Abstract

Predicting stable and metastable structures is central to molecular and materials discovery, but remains limited by the cost of searching high-dimensional energy landscapes. Deep generative models offer efficient structure sampling, yet their outputs remain shaped by training data and can underexplore minima that are rare but physically relevant. We introduce generative structure search (GSS), a unified framework that formulates diffusion-based generation and random structure search (RSS) as limiting regimes of a common sampling process driven by learned score fields and physical forces. Coupling these drivers lets GSS use data priors to accelerate sampling while retaining energy-guided exploration of local minima. Across molecular and crystalline systems, GSS recovers diverse metastable structures with more than tenfold lower sampling cost than RSS for broad coverage and remains effective for compositions outside the training distribution. The results establish a physically grounded generative search strategy for discovering structures beyond the reach of data-driven sampling alone.

Problem

• structure prediction
• metastable structures
• energy landscape
• molecular discovery
• crystal structures

Innovation

• generative structure search
• diffusion-based generation
• random structure search
• score fields
• metastable structures