A generative model for inorganic materials design
The design of functional materials with desired properties is essential in driving technological advances in areas such as energy storage, catalysis and carbon capture<sup>1-3</sup>. Generative models accelerate materials design by directly generating new materials given desired property constraints, but current methods have a low success rate in proposing stable crystals or can satisfy only a limited set of property constraints<sup>4-11</sup>. Here we present MatterGen, a model that generates stable, diverse inorganic materials across the periodic table and can further be fine-tuned to steer the generation towards a broad range of property constraints. Compared with previous generative models<sup>4,12</sup>, structures produced by MatterGen are more than twice as likely to be new and stable, and more than ten times closer to the local energy minimum. After fine-tuning, MatterGen successfully generates stable, new materials with desired chemistry, symmetry and mechanical, electronic and magnetic properties. As a proof of concept, we synthesize one of the generated structures and measure its property value to be within 20% of our target. We believe that the quality of generated materials and the breadth of abilities of MatterGen represent an important advancement towards creating a foundational generative model for materials design.
Authors
- Chin‐Wei Huang (ORCID: https://orcid.org/0000-0003-3335-2187)
- Bichlien H. Nguyen (ORCID: https://orcid.org/0000-0001-5870-4914)
- Lixin Sun (ORCID: https://orcid.org/0000-0002-7971-5222)
- Daniel Zügner (ORCID: https://orcid.org/0000-0003-1626-5065)
- Jake A. Smith (ORCID: https://orcid.org/0000-0003-0412-1312)
- Zhongwen Wang (ORCID: https://orcid.org/0009-0004-2758-4197)
- Robert Pinsler (ORCID: https://orcid.org/0000-0003-1454-188X)
- Jielan Li (ORCID: https://orcid.org/0000-0003-4428-2452)
- Hongxia Hao (ORCID: https://orcid.org/0000-0002-4382-200X)
- Andrew T. Fowler (ORCID: https://orcid.org/0000-0002-7360-3078)
- Yichi Zhou
- Matthew K. Horton (ORCID: https://orcid.org/0000-0001-7777-8871)
- Hannes Schulz (ORCID: https://orcid.org/0000-0001-6408-9794)
- Ziheng Lu (ORCID: https://orcid.org/0000-0003-2239-8526)
- Xiang Fu (ORCID: https://orcid.org/0000-0001-7480-6312)
- Claudio Zeni
- Wenjie Li (ORCID: https://orcid.org/0000-0003-2717-9031)
- Han Yang (ORCID: https://orcid.org/0000-0002-4531-093X)
- Chunlei Yang (ORCID: https://orcid.org/0000-0002-4872-1304)
- Ryota Tomioka (ORCID: https://orcid.org/0000-0002-8092-6553)
- Tian Xie (ORCID: https://orcid.org/0000-0002-0987-4666)
- Jonathan Crabbé (ORCID: https://orcid.org/0000-0002-0341-7712)
- Sarah Lewis
- Shoko Ueda
- Roberto Sordillo
- Aliaksandra Shysheya
Institutions
- Microsoft (United States) (US)
- Chinese Academy of Sciences (CN)
- Microsoft (Germany) (DE)
- Microsoft (Netherlands) (NL)
- Microsoft Research Asia (China) (CN)
- Shenzhen Institutes of Advanced Technology (CN)
- Microsoft Research (United Kingdom) (GB)
Publication Details
- Journal
- Nature
- Published
- 2025-01-16
- DOI
- https://doi.org/10.1038/s41586-025-08628-5
- Citations
- 292
- Primary Topic
- Machine Learning in Materials Science
- Type
- article
- Field-Weighted Citation Impact
- 151.20
Funders
- Microsoft Research