Improving carrier separation in ZnIn2S4 to boost photocatalytic degradation of metronidazole based on machine learning prediction, experimental verification and theory calculation

[Display omitted] •A machine learning model for predicting band gap of photocatalysts was build.•N increased band gap and carrier separation efficiency of ZIS.•π* anti-bonding orbital of O2 received photogenerated electron on the catalysts.•Carrier separation process was successfully simulated by rt-TDDFT. Carrier separation efficiency significantly impacts the photocatalyst performance for wastewater purification. However, there is no established theory to accurately guide carrier separation efficiency regulation. This study used machine learning to screen dopants to change the band gap of ZnIn2S4 (ZIS), thus precisely regulating the ZIS carrier separation efficiency. Nitrogen with 0.5 % of doping amount showed the best regulation effect. Metronidazole degradation kinetic rate with nitrogen doped ZIS (N/ZIS) was 1.47 times higher than that with ZIS. Real-time time-dependent density functional theory (rt-TDDFT) was employed to assess carrier separation at the orbital level. Experiments and rt-TDDFT confirmed that N/ZIS possessed better carrier separation efficiency than that of ZIS. An “in-situ” potential research strategy was established to determine carrier acceptors, which can compare the orbital potential of molecules and catalysts in the adsorption state. This approach revealed that O2 served as the photogenerated electron acceptor, while OH– functioned as the hole acceptor. H2O was not actively involved in carrier separation. O2 anti-bonding orbital activation effect by the catalysts was studied to unveil the influence mechanism of carrier separation efficiency on the photocatalyst activity. This study provides a new insight into carrier dynamics to design efficient photocatalysts.