The accurate prediction of excited-state properties remains a key challenge in organic photocatalyst design, where experimental screening is often time- and resource-intensive. Computational methods, particularly time-dependent density functional theory (TD-DFT), offer a promising route to estimate photophysical descriptors, such as excited-state energies, absorption, and emission spectra, relevant to photocatalytic processes prior to synthesis, thereby enabling the prioritization of candidate molecules with optimized excited-state characteristics.
In this work, a computational workflow is developed to support the efficient exploration of photocatalyst chemical space by combining quantum-chemical calculations with data-driven strategies. TD-DFT calculations using a range of hybrid and range-separated functionals in combination with multiple basis sets were evaluated against experimental values from the Joung et al. chromophore database to identify reliable approaches for predicting key photophysical observables. To ensure systematic coverage of chemical space, molecules were selected using cluster analysis, thereby identifying representative scaffolds while reducing computational cost.
This workflow is designed to be integrated into an active learning algorithm, where iterative selection of informative candidates enables data-efficient and targeted exploration of chemical space, accelerating the discovery of organic photocatalysts with desired performance. By prioritizing the most promising candidates, this approach provides a rational basis for subsequent experimental validation and synthesis, bridging computational predictions with the development of new photocatalytic systems.