Data-Driven Discovery of Transition Metal Dichalcogenide-Based Z-Scheme Photocatalytic Heterostructures

The Z-scheme heterostructure is a highly promising photocatalyst for its unique electronic structure. However, a thorough examination of the heterostructure design space through experimental or computational means is prohibitively expensive. Here, we propose a highly efficient data-driven approach for fast discovering van der Waals (vdW) Z-scheme heterostructures, bypassing the need for costly calculations and experimentation. By conducting high-throughput calculations with the Heyd–Scuseria–Ernzerhof hybrid density functional (HSE06), we first generate a variety of data of electronic structures for 18 experimentally synthesized 2D transition metal dichalcogenides (TMDs) and 20 of 153 heterostructures (constructed with the 18 TMDs). Using these data, we develop an innovative and robust descriptor: Allen “material” electronegativity. Leveraging this descriptor, we identify 27 2D vdW Z-scheme heterostructures from the pool of 153 heterostructures without expensive HSE calculations. We finally refine our findings by selecting six Z-scheme heterostructures with minimal lattice mismatch, further validating them using high-fidelity ab initio calculations and studying their optical absorption. Our research not only paves the way for discovering high-performance Z-scheme photocatalysts using data-driven methods but also contributes a universal charge transfer mechanism for vdW device applications.