High-throughput screening of carbon nitride single-atom catalysts for nitrogen fixation based on machine learning

Abstract

Compared with the traditional electrocatalyst screening of the nitrogen reduction reaction (NRR), machine learning (ML) has achieved high-throughput screening with less computational cost. In this paper, 140 TM@g-C_xN_y single-atom catalysts (SACs) are constructed for the NRR. The deep neural network (DNN) classification model and the extreme gradient boosting (XGBoost) model are selected from different models. A total of 10 features are proposed based on anchoring TM atom, coordination environment and adsorption intermediates. The former model distinguishes qualified and non-qualified catalysts with an accuracy rate of 87.5%, while the latter model predicts the free energy of NRR with a fitting coefficient of 0.82 on the test set. The NN bond length and the number of outermost d electrons of TM (N_d) are found to be the most important features for both models. Moreover, the NN bond length, N_d, and adsorption energy of *N₂H (ΔE_ad[N₂H]) are proved to reflect the degree of nitrogen (N₂) activation and serve as NRR descriptors. The moderate activation and half-filled or nearly half-filled d-orbitals of the TM atom (N_d ≈ 4) favor the NRR process. Among the 20 screened catalysts, Re@g-C₄N₃ shows the best catalytic activity, with a limiting potential (U_L) of only −0.13 V under implicit solvation. The activity origin is illustrated by the electronic properties and bond changes of NRR intermediates. This research provides a new approach for the high-throughput design and screening of SACs by ML based on DFT.