Correlative Effects of Carbon Support Structures and Surface Properties on ORR Catalytic Activities of Loaded Catalysts

As a complex three-phase heterogeneous catalyst, the oxygen reduction reaction (ORR) catalyst activity is determined by the interfacial and surface structures and chemical state of the catalyst support. As a typical biomass carbon-based support, rice husk-based porous carbon (RHPC) has natural unique hierarchical porous structures, which easily regulate the microstructure and surface properties. This study explored the correlative effects of RHPC structure and surface properties on ORR catalytic activity through the typical modification methods, namely, alkali etching, high temperature, oxidation, and ball milling. The various factors for the joint effects are defined as the specific surface area, oxygen-containing functional groups, graphite edge defects, resistivity, and contact angle. The analysis of such joint influences is difficult to quantitatively evaluate due to the large number of experimental factors and small sample sizes. Partial least-squares (PLS) can better deal with such problems. Therefore, a PLS regression model was established to evaluate the relative weight of each factor on the catalytic activity for the RHPC-based support catalysts. The results reveal that the regression coefficients of four factors yield similar magnitude for the effect of the half-wave potential (E_1/2). However, graphite edge defects had a more significant impact on the limiting diffusion current density (J) and electron transfer number (n). Furthermore, an optimal support named BM-RHPC-3 was prepared with more defects and oxygen-containing functional groups, which prepared Fe-NS/BM-RHPC-3 presenting the best ORR catalytic activity (E_1/2 = 0.880 V, J of 5.15 mA cm^-2), superior to Pt/C (E_1/2 = 0.844 V, J of 4.99 mA cm^-2). The statistical regression model is validated with a relative error of less than 5% between predicted and true values for analyzing RHPC-based ORR catalysts' catalytic performance. It shows the feasibility of experiment-informed learning for data-driven material discovery and design.