A universal descriptor for two-dimensional carbon nitride-based single-atom electrocatalysts towards the nitrogen reduction reaction

The electrocatalytic nitrogen reduction reaction (NRR) is widely regarded as one of the most promising ways for green and carbon-free ammonia (NH3) synthesis under mild conditions. Currently, strategies for designing highly efficient, selective, and stable NRR electrocatalysts are desirable. Two-dimensional (2D) carbon nitrides (CNx) with different C/N ratios (e.g., g-C2N, g-CN, g-C3N4, g-C4N3, and g-C9N4) have been widely explored as substrates for single-atom catalysts (SACs) towards the NRR. Through density functional theory (DFT) calculations, the NRR electrocatalytic activity of the single transition metal (TM) atom anchored CNx is systematically investigated and revisited. Based on evaluations of stability and catalytic activity, six promising NRR electrocatalysts are screened out, with considerably low limiting potential (UL). Most importantly, we employed a machine learning (ML)-based screening strategy aimed at predicting efficient NRR electrocatalysts by constructing a universal structure-activity descriptor, which encompasses both the intrinsic properties of the TM atom and the 2D CNx with various C/N ratios. The successful application of the newly proposed descriptor on the new system TM@C8N8 validates its universality. Our research holds promise for providing theoretical guidance in synthesizing highly active NRR electrocatalysts. A universal descriptor was constructed by combining DFT calculations and machine learning to predict highly active NRR electrocatalysts based on transition metal atom anchored 2D carbon nitrides with varied C/N ratios (TM@CNx).