Data-Guided Exploration of Process Control in Carbon-Based Catalyst Design for Two-Electron Oxygen Reduction

Electrochemical hydrogen peroxide (H₂O₂) synthesis via the two-electron oxygen reduction reaction (2e^- ORR) is a promising alternative to the energy-intensive and high-pollution anthraquinone oxidation process. Identifying a carbon-based electrocatalyst with high selectivity and activity for 2e^- ORR is crucial to large-scale electrochemical H₂O₂ synthesis. However, optimizing catalyst composition and process parameters through experimental studies has been resource-intensive and machine learning techniques provide a solution to this problem. In this study, machine learning models were developed to enhance our comprehension of how process control and carbon-based catalyst design impact the performance of 2e^- ORR. The R² values of the optimal 2e^- ORR models for H₂O₂ selectivity and current density are 0.959 and 0.831, respectively. It revealed that nitrogen doping and oxygen content significantly enhance H₂O₂ selectivity by modifying the catalyst's electronic structure and stabilizing reaction intermediates. When it comes to current density, the I_D/I_G ratio and carbon content were found to be the key factors. Higher defect densities along with suitable carbon content can enhance catalytic activity by boosting active site density and conductivity. The practical applicability of the model, preliminary validation was conducted using catalyst compositions and process parameters different from those in the data set, confirming the good accuracy of the model in real scenarios. Our findings provide a new perspective on the influence of process control and catalyst design.