Flow Chemistry Highlights: Selected Topic Self-Optimization
Agnieszka Ładosz, Benjamin R. Martin · CHIMIA International Journal for Chemistry · 2022
Reaction optimization is a time-and resource-consuming endeavor and a necessary part of the development of any chemical process.Automating optimization efforts could reduce cost and improve the reliability of generated data.Flow chemistry is intrinsically predisposed to automation, as flow equipment consists of automatable pumps, sensors etc., an advantage exploited by the authors of this paper.Crandall et al. constructed an automated flow reactor for closed-loop reaction optimization.The system comprises three pumps, a heated tubular reactor and a UPLC for reaction analysis.Two optimization methods were tested: SQSnobFit global optimization algorithm and Deep Reaction Optimizer, a deep reinforcement learning algorithm.Authors optimized a reduction of imine to amine in 68 experiments, achieving 70% yield and 95% selectivity.The whole run took 62 hours.Major focus of the paper lies on the software the authors developed to automate the lab hardware.Rxn Rover is written in LabView and enables users to flexibly add hardware by implementing relevant plugins.The entire Rxn Rover, with the plugins as well as documentation on the software are openly available on the team's GitHub page to aid other researchers in lab automation efforts.