Data-driven Strategies to Predict Ruthenium Complex Photocleavage Efficiency

Abstract

Ruthenium (Ru) complex-based photocages have shown great potential in photochemotherapy due to their activation by red- and infrared light. However, different Ru complexes exhibit dramatically varying quantum yields in the photocleavage process. To optimize the structures for a high quantum yield, in this work, we developed machine learning models and density-functional theory (DFT) protocols to predict the quantum yield for photocleavage of Ruthenium complexes. The built machine learning models were not accurate enough to classify the quantum yield for a given Ru complex structure. An alternative DFT protocol was designed to bridge the experimental quantum photorelease quantum yield and computational energy gap between the 3MLCT and 3MC states. This approach demonstrated the energy gap between the 3MLCT state and the 3MC state can serve as a ‘barrier’ in the ligand dissociation process and could help predict the quantum yield of photolabile Ru-based complexes.