M-N4-Gr/MXene Heterojunction Nanosheets as Oxygen Reduction and Evolution Reaction Catalysts: Machine Learning and Density Functional Theory Insights

MXene nanosheet materials with fast charge transfer characteristics and superior electrical conductivity are popular candidates for oxygen electrocatalysts. In this paper, the oxygen reduction and evolution reaction (ORR and OER, respectively) catalytic activities of 78 M-N4-Gr/MXene (Ti2C, Nb2C, and V2C) heterostructure nanosheets were investigated based on density functional theory (DFT) calculations and machine learning (ML) predictions. The random forest regression (RFR) algorithm proved to be the most desirable ML model for the prediction of catalytic activity with ORR/OER root mean square errors of 0.10 and 0.23 V, respectively. The obtained results show that Ni-N4-Gr/Nb2C (η*ORR = 0.31 V) and Ni-N4-Gr/V2C (η*ORR = 0.32 V) are effective ORR catalysts (prediction error of 0.03 V only), Ru-N4-Gr/Nb2C (η*OER = 0.40 V) is a promising OER catalyst, and Co-N4-Gr/V2C (η*ORR = 0.45 V/η*OER = 0.30 V) is an excellent bifunctional catalyst. Through further model analysis, the d band center of the metal active site was determined to be the most effective descriptor and a volcano curve relationship between the intrinsic descriptor and the overpotential was established based on the characteristic importance and Pearson correlation. The research approach adopted in this study may be used to screen and design other MXene heterojunction catalysts and remarkably accelerate catalyst design for many significant reactions.