Geometric tuning of supports is an emerging strategy to optimize catalysts, yet its role in governing the synergy of heteronuclear p-d dual-atom catalysts (DACs) is unexplored. Using carbon nanotubes (CNTs) as tunable curvature substrates, we investigated their influence on SiFeN6 DACs via first-principles calculations. We reveal an inverted-volcano-type relationship between curvature and activity, originating from the nonlinear differential response of key intermediates. This curvature-driven trend is a general principle applicable to other 3d transition metals (TM = Mn, Co, Ni). To rationalize this complex relationship, we integrated a machine learning (SISSO) approach, which yielded a robust multidimensional descriptor (R2 = 0.92). By quantitatively revealing the dominant role of the p-block Si site, our data-driven model establishes substrate geometry as a primary and effective design strategy for optimizing these complex dual-atom catalysts.
Wang et al. (Mon,) studied this question.