CrystalFormer-CSP: Thinking fast and slow for crystal structure prediction
Crystal structure prediction is a fundamental problem in materials science. We present CrystalFormer-CSP, an efficient framework that unifies data-driven heuristic and physics-driven optimization approaches to predict stable crystal structures for given chemical compositions. The approach combines pretrained generative models for space-group-informed structure generation and a universal machine learning force field for energy minimization. Reinforcement fine-tuning can be employed to further boost the accuracy of the framework. We demonstrate the effectiveness of CrystalFormer-CSP on benchmark problems and showcase its usage via web interface and language model integration.
Authors
- Shigang Ou
- Lei P Wang
- Zhendong Cao
Publication Details
- Journal
- Physical review. B./Physical review. B
- Published
- 2026-11-17
- DOI
- https://doi.org/10.1103/fnym-1ypn
- Citations
- 4
- Primary Topic
- Machine Learning in Materials Science
- Type
- article
- Field-Weighted Citation Impact
- 3.47
Funders
- National Natural Science Foundation of China
- Chinese Academy of Sciences