Answers are grounded in this paper and cite their sources.
Coordination-engineered transition-metal single-atom catalysts offer a modular platform for the hydrogen evolution reaction (HER), but their discovery is hindered by the large configurational space arising from coupled metal-ligand variations. Here we develop a symmetry-aware and uncertainty-guided workflow to accelerate the screening of TMX4@G catalysts (X = C, N, O, P or S), where a transition-metal atom is coordinated by four ligands embedded in graphene. Canonicalization of rotationally equivalent motifs reduces 17 500 nominal structures to 4900 symmetry-inequivalent candidates while preserving chemical diversity. A cascaded Gaussian process regression model encodes a physical hierarchy from atomic descriptors to electronic structure, metal-hydrogen interaction and the HER activity. Posterior uncertainty drives active learning by balancing near-thermoneutral hydrogen binding with exploration of underrepresented coordination motifs. A large language model (LLM)-assisted automation framework further integrates structure generation, VASP input preparation, job submission, output parsing and dataset updating. DFT validation shows strong agreement with model predictions (R2 = 0.898 for 24 candidates) and identifies six thermodynamically stable catalysts with ηHER < 0.15 V.
No takes yet. Share an insight, caveat, or question.
Jiang et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: