Efficient multi-objective Bayesian optimization of gas–liquid photochemical reactions using an automated flow platform

Abstract

We developed an automated platform capable of performing photochemical gas–liquid reactions. The platform was augmented with a state-of-the-art Bayesian optimization algorithm and was tested on the decatungstate-catalyzed aerobic oxidation of ethyl benzene, to optimize both yield and productivity, and identify the Pareto front of these objectives. Although photochemical gas–liquid systems are highly complex due to numerous interactions between the parameters, including effects on mass-transfer, gas solubility and light absorption, the algorithm demonstrated impressive speed to navigate the parameter space towards optimal conditions. Furthermore, this approach also proved highly flexible, allowing for modification of objectives and parameter ranges on the fly. The identified conditions were then tested on a select scope of substrates, to better understand the generality of these conditions, especially on molecules where selectivity comes into play. The results show the platform to be a useful tool for reaction optimization and process intensification.