Machine Learning‐Driven Band Alignment Strategy for Screening 1T‐TMDs‐Based Z‐Scheme Heterostructures toward Efficient Photocatalytic Water Splitting

Abstract To address the global energy crisis and mitigate environmental challenges stemming from fossil fuel dependence, advancing efficient photocatalytic water splitting technology has become a critical focus in renewable energy research. An innovative design strategy for high‐efficiency photocatalysts based on band edge alignment is established through the integration of machine learning (ML) and first‐principles computational methods, developing a high‐throughput screening framework specifically targeting 1T‐phase transition metal dichalcogenides (1T‐TMDs). Through optimized feature selection, ML models, and training protocols, the PdSSe monolayer is identified as exhibiting ideal band edge compatibility with the GeC monolayer. Subsequent density functional theory (DFT) verification confirmed exceptional agreement with ML predictions. The GeC/SPdSe Z‐scheme heterostructure achieves remarkable photocatalytic efficiency, driven by its optimally aligned band structure that enables spontaneous hydrogen evolution reaction (HER) and oxygen evolution reaction (OER) under visible‐light irradiation. Nonadiabatic molecular dynamics (NAMD) simulations reveal that photo‐generated carriers in heterostructures follow a Z‐scheme pathway, as supported by distinct timescales of electron‐hole migration and recombination. This heterostructure architecture exhibits broadband light absorption spanning the visible to ultraviolet spectral regions, yielding a remarkable theoretical solar‐to‐hydrogen (STH) efficiency of 29.5%.