Organic photoredox catalysts for CO2 reduction: Driving discovery with genetic algorithms

Abstract

This work implements a genetic algorithm (GA) to discover organic catalysts for photoredox CO2 reduction that are both highly active and resistant to degradation. The lowest unoccupied molecular orbital energy of the ground state catalyst is chosen as the activity descriptor and the average Mulliken charge on all ring carbons is chosen as the descriptor for resistance to degradation via carboxylation (both obtained using density functional theory) to construct the fitness function of the GA. We combine the results of multiple GA runs, each based on different relative weighting of the two descriptors, and rigorously assess GA performance by calculating electron transfer barriers to CO2 reduction. A large majority of GA predictions exhibit improved performance relative to experimentally studied o-, m-, and p-terphenyl catalysts. Based on stringent cutoffs imposed on the average charge, barrier to electron transfer to CO2, and excitation energy, we recommend 25 catalysts for further experimental investigation of viability toward photoredox CO2 reduction.