Statistical analysis of structural descriptors governing H2O2 selectivity on oxygen-functionalized carbons

Electrochemical on-site production of hydrogen peroxide (H 2 O 2 ) via the two-electron oxygen reduction reaction (2e - ORR) offers environmental benefits, including reduced transport hazards, demand-responsive dosing, and compatibility with renewable energy systems. Although metal-free carbon catalysts are highly promising, a reliable structural descriptor controlling their catalytic selectivity remains unclear. In this work, two series of carbon nanotubes were prepared through controlled thermal reduction and employed to decouple the catalytic contributions from oxygen functional groups and carbon hybridization state. Non-parametric screening methods (mutual information and distance correlation) and uniform quadratic visualizations consistently identify the global carbon hybridization ratio ( sp 2 / sp 3 ) as the primary determinant of catalytic selectivity. H 2 O 2 selectivity follows a volcano‑type dependence on sp 2 / sp 3 , peaking at ≈5.7–6.2. The best catalyst, OCNT‑r800 ( sp 2 / sp 3 ≈ 5.9), delivers 99.3% selectivity at 0.50 V vs RHE ( n = 2.03). Operando Fourier-transform infrared spectroscopy (FTIR) and alkaline flow‑cell tests further verify selective H₂O₂ electrosynthesis at practical current density (~330 mA cm -2 ) with FE ≥ 95% over 12 h. The results point to a balance between π‑delocalized electronic continuity and adsorption‑competent defect environments as a framework‑level guideline for designing oxygen‑functionalized carbon catalysts for H 2 O 2 production. • Dual carbon nanotube series decouple oxygen groups from carbon hybridization. • Global sp 2 / sp 3 ratio emerges as the primary descriptor of H 2 O 2 selectivity. • Selectivity shows a clear optimum at sp 2 / sp 3 around 5.7 to 6.2. • Best catalyst gives ~99.3% H 2 O 2 selectivity at 0.5 V, with n ≈ 2.03. • Data-driven screening and quadratic visualization guide carbon catalyst design.