Theoretically revealing the activity origin of the hydrogen evolution reaction on carbon-based single-atom catalysts and finding ideal catalysts for water splitting

Carbon-based single-atom catalysts (SACs) play an important role in electrochemical water splitting. Herein, using the density functional theory (DFT) and machine learning (ML) approaches, we investigated the stability and electrocatalytic activity of twenty-eight transition metal (TM) atoms on eight representative carbon-based supports (N/C-coordination graphene, C2N, C3N4, graphdiyne (GDY), phthalocyanine (Pc), covalent organic frameworks (COF), and metal-organic frameworks (MOF)) and explored the structure-property correlation and catalytic activity origin of SACs toward the hydrogen evolution reaction (HER). The DFT results show that N/C-graphene, Pc and MOF materials are desirable supports for SACs with good thermodynamic and electrochemical stability. Among them, Co-N-C, Co-Pc, V/Fe/Co/Rh/Ir-MOF, and V/Tc/Rh/Os-C SACs showed excellent HER activity with |eta(HER)| <= 0.15 V; Co/Rh-N-C, Co/Rh/Ir-MOF, Co/Rh/Ir-Pc, and Ni/Pd-C showed low overpotentials comparable to IrO2 of the oxygen evolution reaction (OER). Moreover, Co-N-C, Co-Pc, and Co/Rh/Ir-MOF presented promising electrocatalytic activity both for HER and OER, which are the ideal bifunctional catalysts. According to the ML approach, HER activities of these carbon-based SACs can be well predicted by the gradient boosting (GB) model. With the help of the feature importance analysis, we concluded that the HER activity was mainly influenced by the electronic properties of the d orbitals and the surrounding geometric structure of the TM active center in the system. This work provides a novel and efficient DFT-ML hybrid method to accelerate the screening process for high-performance catalysts and even reveals the activity origin of the HER.