Meta-LegNet framework accelerates catalyst screening with transferable adsorption environment learning

Researchers have developed Meta-LegNet, a novel graph learning framework designed to predict surface adsorption configurations in computational catalysis. This framework utilizes SE(3)-equivariant atom-level message passing and voxel-based aggregation to learn transferable representations of local adsorption environments. By providing interpretable attribution maps, Meta-LegNet can identify relevant local environments and propose likely adsorption sites on new surfaces, significantly accelerating catalyst screening. AI

Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →

IMPACT Accelerates catalyst screening by providing an interpretable and practical route for identifying adsorption sites.

RANK_REASON This is a research paper detailing a new framework for surface adsorption prediction. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• Meta-LegNet
• SE(3)
• arXiv
• Hugging Face

• paper
• infra