Machine learning-assisted research on nitrogen-sulfur co-doped carbon-loaded platinum-based oxygen reduction catalysts

In this study, a simple yet highly efficient one-pot pyrolysis method was employed to successfully synthesize nitrogen-sulfur dual-doped carbon-supported platinum-based oxygen reduction reaction (ORR) catalysts. By precisely controlling the precursor ratios and reaction conditions, a series of catalysts with varying platinum loadings and nitrogen-sulfur doping ratios were synthesized, among which an optimized high-performance sample, denoted as Pt/BP-NS₅₀₀-40%, was obtained. This catalyst demonstrated exceptional catalytic activity towards ORR, with a half-wave potential of 0.8379 V and an impressive specific activity (SA) of 2.26 A/m². After undergoing 5 000 cycles of cyclic voltammetry testing, the catalyst exhibited only a 73 mV decrease in half-wave potential, showing good stability. To further optimize catalyst design and reduce platinum metal usage, machine learning techniques were introduced in this study to construct a predictive model relating catalyst performance to its preparation conditions. This model not only provides theoretical guidance for catalyst preparation but also accelerates the development process of high-performance catalysts.