(707h) Genetic Algorithm Organic Photoredox Catalyst Evolution for Efficient CO2 Reduction and Degradation Resistance | AIChE

(707h) Genetic Algorithm Organic Photoredox Catalyst Evolution for Efficient CO2 Reduction and Degradation Resistance

Authors 

Kron, K. - Presenter, University of Southern California
Rodriguez-Katakura, A., University of Southern California
Regu, P., University of Southern California
Reed, M., University of Southern California
Elhessen, R., University of Southern California
Mallikarjun Sharada, S., University of Southern California
Though organic photoredox catalysts can perform challenging single electron reductions, their structural complexity and varied degradation reactions have limited the implementation and study of these catalysts. The structural complexity of organic photoredox catalysts makes them ideal candidates for discovery via a genetic algorithm. Genetic algorithms take ‘parents’ and their genetic footprints (molecular formulas) and generate ‘offspring’ that have fitness scores that determine which passes its genetic information to future generations. Genetic algorithms quickly evolve populations using desired features to identify novel offspring with heightened activity. Prior work in our lab studying substituted p-terphenyls for CO2 reduction enabled our identification of LUMO energies as a computationally inexpensive proxy for reduction potential. Further work studying the de-aromatization reactions that lead to lower turnover numbers has suggested average carbon charge ( as a measure of degradation likelihood. Our genetic algorithm takes a pool of substituted oligophenylenes and creates new catalysts through crossover and recombination processes. These catalysts are assessed for fitness via their LUMO energies and then propagated for 10 generations to identify 10 best catalysts per weighting. By varying the relative weight of LUMO and for several runs, we generate 103 unique catalysts that are predicted to efficiently reduce CO2 and/or resist degradation. The electronic structures of these catalysts were modelled to calculate key electron transfer parameters and verify that the predicted catalysts improve upon the parents and experimental references. While most catalysts reflect improvement, we identify 25 best viable catalysts and recommend them for future experimental study to achieve CO2 reduction. We also identify a recommended balance between weighting reduction potential and degradation resistance for future catalyst discovery.