Predicting Gas Adsorption on Dual-Atom Catalysts Embedded in Graphene via Integrated DFT and Machine Learning

The rational design of high-performance dual-atom catalysts requires a fundamental understanding of their gas adsorption properties, yet predictive frameworks for these complex systems remain underdeveloped. To address this gap, we systematically investigate the adsorption of CO, CO₂, H₂, and NO on ten bimetallic-doped graphene systems using density functional theory (DFT). DFT analysis identifies the d-band center, adsorption distance, and interfacial charge transfer as key descriptors governing adsorption strength. These descriptors were used to train ML models, with gradient boosting decision tree achieving high accuracy. On the basis of these results, we establish a descriptor-driven framework that effectively correlates atomic-scale features with adsorption energies for the ten bimetallic pairs and four gas molecules examined. The identified descriptors and the ML model offer a basis for further exploration of dual-atom catalysts, while extension to broader catalyst and adsorbate systems requires an additional validation.