Abstract
Accurate band-gap prediction is essential for the computational design of photocatalysts, influencing light absorption, redox energetics, and charge-carrier dynamics. Here, we present a systematic quantitative benchmark of electronic-structure methods for predicting band-gap across diverse photocatalytic semiconductors and two-dimensional materials. In total, this review evaluates 197 unique photocatalytic materials, comprising 175 bulk semiconductors and 22 two-dimensional (2D) systems. The analysis covers GGA, GGA+SOC, OptB88vdW, hybrid functionals (HSE06, PBE0, B3LYP), meta-GGA approaches (TB-mBJ and SCAN), and many-body perturbation theory (G 0 W 0 ). Their predictive performance is evaluated against experimental data using ME, MAE, RMSE, MPE, MAPE, and standard deviation. The results demonstrate that no single approach is universally reliable across all material classes. HSE06 shows robust performance for transition-metal oxides, sulfides, and selenides, while TB-mBJ provides a favorable accuracy–cost balance for d 0 oxides, nitrides, and oxynitrides. SCAN performs particularly well for oxyhalides, OptB88vdW for layered chalcogenide-halides, and G 0 W 0 for two-dimensional materials. These findings establish a class-specific methodological framework linking functional selection to chemical composition, orbital character, dimensionality, and computational cost, providing practical guidance for reliable first-principles screening and accelerated photocatalyst discovery.