Dresden 2026 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
MM: Fachverband Metall- und Materialphysik
MM 13: Data-driven Materials Science: Big Data and Workflows I
MM 13.9: Talk
Tuesday, March 10, 2026, 12:30–12:45, SCH/A251
Score-based diffusion models for accurate crystal structure inpainting and reconstruction of hydrogen positions — •Timo Reents1, Arianna Cantarella2, Marnik Bercx1, Pietro Bonfà2,3, and Giovanni Pizzi1 — 1PSI Center for Scientific Computing, Theory and Data, CH-5232 Villigen PSI, Switzerland — 2Department of Physics and Earth Sciences, University of Parma, IT-43124 Parma, Italy — 3Dipartimento di Scienze Fisiche, Informatiche e Matematiche, University of Modena and Reggio Emilia, IT-41125 Modena, Italy
Generative AI methods are rapidly evolving to speed up and improve materials discovery. Diffusion based models can not only be adopted to generate new materials with desired properties but also to reconstruct crystal structures for which structural information is only partially available. In this work, we use Microsoft's mattergen [1], a diffusion based model originally designed to generate new stable crystal structures, and extend it to reconstruct missing hydrogen sites in crystal structures reported in experimental databases. This is particularly useful as the experimental measurement of hydrogen sites with standard XRD is typically challenging due to weak scattering of hydrogen. We show how to leverage image inpainting approaches known from computer vision, combined with universal machine learning interatomic potentials, to improve the success rate of correctly identifying the missing sites or finding lower energy configurations while significantly lowering the computational cost with respect to a direct DFT approach.
[1] Zeni, C. et al., Nature 639, 624-632 (2025)
Keywords: Generative AI for materials; Machine Learning; MLIP; Hydrogen