Theoretical insights of curvature effects on CO₂ reduction in TM₂N₆-doped carbon nanotubes

Dual-atom catalysts have emerged as enhanced platforms of single-atom catalysts for promoting the electrocatalytic carbon dioxide reduction reaction (CO₂RR). Here, we establish a curvature-resolved map of activity and selectivity for nitrogen-anchored, homonuclear dual-atom catalysts on carbon nanotubes (TM₂N₆-CNTs). We first screened ten 3d metals at the curvature ( κ = 0.25 Å⁻¹) using formation energies and dissolution potentials, and retained Fe, Co, Ni, Cu, Zn as electrochemically viable dopants. We then constructed a 100-member library spanning five curvatures, two configurations (armchair and zigzag), and two embedding orientations (parallel and vertical) of the TM₂N₆ axis relative to the tube axis, and computed full CO₂RR free-energy diagrams, HER free-energy, density of states, d-band centers, and charge-density differences. A clear metal-dependent branching emerges at the first proton-electron transfer: Cu₂N₆-CNTs funnel into the HCOOH route, while Fe/Co/Ni/Zn₂N₆-CNTs favor multi-electron hydrogenation toward CH₄/HCHO. Interpretable machine learning trained on curvature-, electronic-, and metal-interaction descriptors achieves accurate overpotential prediction (R² ≈ 0.94-0.95, RMSE≈0.08eV) and identifies curvature-electronic coupling and bimetallic interaction terms as key determinants. These results establish CNT curvature and DAC orientation as coequal design knobs for steering activity and selectivity in TM₂N₆-CNT electrocatalysts.