Synergistic DFT and Machine Learning Screening of Z-Scheme g-SiC/TMD Heterostructures for Efficient Overall Water Splitting

Abstract

To mitigate the heavy reliance on non-renewable resources such as fossil fuels, photocatalytic water splitting has been recognized as a promising method for generating clean and renewable energy. However, achieving Z-Scheme heterostructures with suitable band edge positions for remains a significant challenge. In this work, a synergistic strategy combining density functional theory (DFT) and machine learning (ML) is proposed and implemented, which efficiently screens 1T transition metal dichalcogenide (TMD) monolayers for the oxygen evolution reaction (OER) by incorporating graphene-carbide silicon (g-SiC) monolayers with high hydrogen evolution reaction (HER) performance, thereby achieving an ideal Z-scheme band alignment. The DFT calculation results reveal that the g-SiC/SZrSe (S-C stacking) heterostructure and g-SiC/SeZrS (Se-C stacking) heterostructure are determined to be the most stable configurations for Z-Scheme heterostructures, where both the HER and OER can proceed spontaneously under light irradiation. Non-Adiabatic Molecular Dynamics (NAMD) simulations further verify that the Z-scheme pathway, with ultrafast interlayer electron-hole recombination (~1 ps). The heterostructure also achieves strong light absorption (>105 cm-1) and a remarkable solar-to-hydrogen efficiency exceeding 33%, far surpassing the conventional limit, thus demonstrating its great potential for efficient overall water splitting.