New benchmark and model advance crystal structure determination from PXRD data

Researchers have introduced PXRD-Bench, a new benchmark designed to evaluate generative models for determining crystal structures from powder X-ray diffraction (PXRD) data. This benchmark extends previous evaluations to handle unknown elemental compositions and the recovery of structures from two-phase mixtures, utilizing approximately 4.1 million simulated training structures. Alongside the benchmark, the study presents Bragglite, a new PXRD-to-structure model that achieves a 44.2% Top-1 match rate on experimental data, significantly outperforming prior baselines. The findings indicate that while some models struggle to leverage experimental PXRD data, generalization to novel structural prototypes and recovery from complex mixtures remain significant challenges. AI

IMPACT This research could accelerate materials discovery by improving the automated analysis of crystal structures from diffraction data.

RANK_REASON The cluster contains a new academic paper introducing a benchmark and a model for a specific scientific task. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• arXiv
• Bragglite
• PXRD-Bench
• rruff

• paper
• model release
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →