Carbon Curvature Influence on Single-Atom Catalysts for CO2 Reduction: A Study via High-Throughput Calculations and Machine Learning.

The rational design of single-atom catalysts (SACs) requires precise control over the local coordination environment. Yet the curvature of the supports, an intrinsic and tunable geometric parameter, is often overlooked. Herein, we present a high-throughput computational study combined with machine learning methods to unravel the effect of carbon support curvature on the catalytic performance of SACs for the CO2 reduction reaction (CO2RR). By investigating 28 transition metals (TMs) anchored on carbon nanotubes (CNTs) with different curvatures, we reveal that increasing curvature with higher surface energy universally enhances the thermodynamic and chemical stability of single atoms by strengthening metal-support interactions. Surprisingly, the support curvature can break the linear scaling relationship that the same metal species generally have identical adsorption trends for different species, by adjusting which the competitive hydrogen evolution reaction (HER) can be effectively suppressed. Additionally, machine learning reveals the d-band center as a key factor governing the curvature sensitivity of different metals. This work establishes support curvature as a controllable and intrinsic design dimension and proposes a general "curvature-engineering" strategy for developing efficient catalysts.