Research
Accelerating high-throughput catalyst screening by direct generation of equilibrium adsorption structures
Nature · By Songze Huo ·
The Brief
- The authors present DBCata, a generative model trained without energy or force labels.
- DBCata directly generates near-equilibrium adsorption structures.
- The goal is faster high-throughput screening of candidate catalysts.
Why it matters
Generating adsorption structures directly could speed up the screening of candidate catalysts, which is a costly step in catalysis research.
Who’s affected
- computational catalysis researchers
- materials scientists who screen candidate catalysts
- #catalysis
- #generative models
- #adsorption
- #materials screening
- #machine learning
Nature Communications, Published online: 21 November 2026; doi:10.1038/s41467-026-79243-9 The authors develop DBCata, a generative model trained without energy or force labels to produce near-equilibrium adsorption structures, enabling faster screening of candidate catalysts.