Author: Shaohua Shen (Xi'an Jiaotong University) - In this study, we present a Z-scheme heterojunction photocatalyst loaded photocatalytic overall water splitting device with high scalability and encouraging performance under natural sunlight. This achievement is attributed to the machine learning (ML) discovered highly efficient Z-scheme heterojunction photocatalyst and the design of bionic photocatalytic reactor by mimicking plant chloroplasts. As for the photocatalytic material, we developed a ML prediction framework to discover Z-scheme heterojunction photocatalyst for overall water splitting. Using the ML discovered Z-scheme heterojunction photocatalyst, a STH of 1.59% can be achieved for overall water splitting in a home-made flow-cell reactor under standard AM1.5G illumination (100 mW cm-2) in the lab. As for the photocatalytic device, we designed a four-layered flow-cell reactor (light-receiving area: 4 cm2) to favor light absorption and then assembled 3 × 3 of these reactors into one module (light-receiving area: 36 cm2) with facilitated mass transfer, by mimicking the structure of chloroplasts in plant. When loaded with the ML discovered Z-scheme heterojunction, one such module achieved a STH of 1.90% under standard AM1.5G illumination (100 mW cm-2) in the lab. With 3 × 3 of these modules parallelly assembled into an upscaled photocatalytic device (light-receiving area: 324 cm2), the STH reached up to 2.11% as high and kept at ~2.0% in two independent tests under outdoor sunlight irradiation. This demonstration reveals the great feasibility and scalability of photocatalytic overall water splitting for practical solar H2 production by ML-assisted material discovery and bionic reactor design.
