The efficiency of photocatalytic reactions depends critically on the ability of a catalyst to absorb photons with energies that match the activation barrier of the desired chemical process. However, identifying metal–ligand complexes that can absorb a specific photon energy (Ea) remains a major challenge due to the vast diversity of possible combinations and the nonlinear relationship between molecular structure and electronic excitation energy. In this work, we present a machine learning framework that performs inverse design of photoactive complexes—accepting a target energy input (Ea) and predicting metal–ligand combinations capable of absorbing photons with that energy to catalyze a specified transformation.
The model leverages patterns learned from computational and experimental data to identify how structural and electronic features influence photon absorption. By reversing this relationship, the algorithm can suggest promising complexes tailored to the energy requirements of a given reaction. This approach enables a rational, data-driven route to photocatalyst discovery, reducing reliance on empirical screening and expensive quantum chemical calculations. Beyond identifying suitable complexes for specific activation energies, the framework offers a foundation for designing light-responsive materials across a range of catalytic and energy-conversion processes. It represents a step toward fully automated materials discovery, where machine learning can translate energetic requirements directly into molecular design strategies.