Physically coupled centre-environment descriptor for CO adsorption on Pt-centred PtRu(111) local environments

CO poisoning remains a central limitation in Pt-based methanol oxidation catalysts, and CO adsorption on PtRu alloys is strongly regulated by the local atomic environment around the Pt adsorption centre. Here, density functional theory calculations and transfer-oriented machine-learning validation are combined to develop a physically grounded coupled descriptor for CO adsorption on Pt-centred PtRu(111) local environments. A coherent dataset of 52 Pt-top configurations was constructed by fixing the adsorption atom as Pt while systematically varying top-layer and subsurface Pt/Ru arrangements. Descriptor benchmarking shows that a single d-band centre is too compressed to capture the adsorption-energy variation, whereas width- and filling-like centre-electronic channels combined with explicit top-layer, subsurface, and topology-sensitive environment information provide a more faithful representation. Grouped validation shows that the coupled representation transfers more reliably across unseen Pt-centred PtRu environments than the single-d-band-centre baseline, although the magnitude of the predictive gain remains model dependent. These results support an environment-regulated centre-response picture of Pt-CO bonding. The present conclusions are restricted to the fixed-Pt-centre PtRu domain and provide descriptor-level insight into vacuum CO binding relevant to CO poisoning rather than direct prediction of methanol-oxidation activity.