Crystal structure prediction with host-guided inpainting generation and foundation potentials

Peichen Zhong, Xinzhe Dai, Bowen Deng, Gerbrand Ceder, Kristin A. Persson · 2025

Read Crystal structure prediction with host-guided inpainting generation and foundation potentials by Peichen Zhong, Xinzhe Dai, Bowen Deng, Gerbrand Ceder, Kristin A. Persson (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2504-16893.txt.

Description

Predicting how atoms arrange in a new crystal is a bottleneck in materials discovery, and Peichen Zhong and co-authors offer a framework called CHGGen. Free-running generative models struggle to produce symmetric crystals as the unit cell grows. Their fix is inpainting: start from a prepared host structure with the right symmetry, and let the model repaint only a fraction of the atomic positions.

A learned potential then relaxes the result toward a stable arrangement. On two test chemistries, zinc sulfide with phosphorus sulfide and lithium-silicon, the guided method produced far more symmetric candidates than unguided generation. The scheme also suits editing known crystals, such as partly filled or layered battery materials, and can host other generators.

Catalog record

CREATORPeichen Zhong, Xinzhe Dai, Bowen Deng, Gerbrand Ceder, Kristin A. Persson
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE44 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2504-16893

Browse the full library · About the Tower of Babel