Machine learning assisted high-throughput screening of transition metal single atom based superb hydrogen evolution electrocatalysts

Abstract

Carbon-based transition metal (TM) single-atom catalysts (SACs) have shown great potential toward electrochemical water splitting and H₂ production. Given that two-dimensional (2D) materials are widely exploited for sustainable energy conversion and storage applications, the optimization of SACs with respect to diverse 2D materials is of importance. Herein, using density functional theory (DFT) and machine learning (ML) approaches, we highlight a new perspective for the rational design of TM-SACs. We have tuned the electronic properties of ∼364 rationally designed catalysts by embedding 3d/4d/5d TM single atoms in diverse substrates including g-C₃N₄, π-conjugated polymer, pyridinic graphene, and hexagonal boron nitride with single and double vacancy defects each with a mono- or dual-type non-metal (B, N, and P) doped configuration. In ML analysis, we use various types of electronic, geometric and thermodynamic descriptors and demonstrate that our model identifies stable and high-performance HER electrocatalysts. From the DFT results, we found 20 highly promising candidates which exhibit excellent HER activities (|ΔG_H*| ≤ 0.1 eV). Remarkably, Pd@B₄, Ru@N₂C₂, Pt@B₂N₂, Fe@N₃, Fe@P₃, Mn@P₄ and Fe@P₄ show practically near thermo-neutral binding energies (|ΔG_H*| = 0.01–0.02 eV). This work provides a fundamental understanding of the rational design of efficient TM-SACs for H₂ production through water-splitting.