Universal Catalyst Design Framework for Electrochemical Hydrogen Peroxide Synthesis Facilitated by Local Atomic Environment Descriptors

A universal design framework for high-performance catalysts remains challenging due to diverse structures and active sites. We developed a framework integrating weighted atom-centered symmetry function (wACSF) descriptors with machine learning, microkinetic modeling, and high-throughput screening. The wACSF descriptors unify geometric and chemical characteristics of active sites across different catalyst families. ML models trained on wACSF accurately predicted adsorption free energies of hydroxyl (ΔG_OH _*, R² = 0.84) and oxygen (ΔG_O*, R² = 0.91) for intermetallic alloys, metal oxides, perovskites, and single-atom catalysts in the two-electron water oxidation reaction (2e^- WOR). Density functional theory and microkinetic modeling yielded a universal 2e^- WOR volcano model that agreed well with experiments. High-throughput screening with ML-predicted ΔG_OH* identified LiScO₂, which achieved 90% H₂O₂ Faradaic efficiency at 2.2 V vs. reversible hydrogen electrode (RHE) with 168-hour stability (82%-86% retention). Experimental activity (log(j) = 1.56) matched theoretical predictions (log(j) = 1.28) within 5% deviation at 2.4 V_RHE. This universal framework provides a general paradigm for rational catalyst design and is implemented in the Digital Catalysis Platform (DigCat), enabling efficient discovery across diverse material classes and electrochemical reactions.