Physically guided discovery of high-performance photocatalytic and photovoltaic heterostructures

Two-dimensional (2D) heterostructures provide a versatile platform for solar energy conversion owing to their highly tunable electronic structures and flexible van der Waals integration. However, the efficient discovery of high-performance photocatalytic and photovoltaic heterostructures remains challenging due to the vast combinatorial space, the need to sustain both strong oxidation and reduction reactions for photocatalysis, and inherent trade-offs among efficiency, charge transport, and stability in photovoltaics. Here, we develop a physically guided and data-driven strategy for the systematic discovery of functional 2D heterostructures for catalysis and photovoltaics. By employing machine learning-assisted predictions to complete missing hybrid functional-level electronic structure data for monolayers in the Computational 2D Materials Database, we enable exploration of a heterostructure landscape comprising over 510 000 candidates. For photocatalytic water splitting, direct high-performance Z-scheme heterostructures are identified using physically motivated descriptors based on band offsets and material electronegativity differences, enabling the realization of strong oxidation and reduction capability together with favourable light absorption. For photovoltaic applications, we indicate that evaluating band alignment and theoretical power conversion efficiency alone is insufficient to assess realistic device potential. Incorporating carrier mobility as a key descriptor is essential for capturing charge transport limitations that critically influence device performance. Remarkably, the HfBr₂/HfICl heterostructure exhibits a power conversion efficiency above 21%, along with ultrahigh and well-balanced electron and hole mobilities of 3083 and 7551 cm² V^-1 s^-1, respectively. The strategy established here offers a physically transparent route for discovering next-generation optoelectronic materials across expansive chemical spaces.