Bayesian Optimized Electrosynthesis of Azobenzenes in a Self- Optimizing Flow Set-up: from DoE-guidance to Gram-Scale Preparation

The development of a self-optimizing platform for electrosynthesis in a closed-loop set-up is established, which operates on a specifically developed open-source software. With this system, the electro-reduction of nitrobenzenes to azobenzenes was optimized. In particular, a Design of Experiments (DoE) campaign was performed which was subsequently used as the initial data set for autonomous Bayesian optimization (BO) strategy. This approach allowed to illustrate both methodologies ability to thoroughly explore the parameter space of a reaction. Furthermore, BO was used to efficiently optimize inside reaction parameters bounds previously explored with DoE. The entire optimization process was performed at least on a scale of 0.5 mmol or higher and a gram scale reaction was conducted using the aforementioned automated set-up. Optimal reaction conditions are compatible with larger scale synthesis. The gram-scale purification was also performed involving only cost-efficient process compatible steps. Evaluation of reaction scope pointed out i) the ability to synthesize various azobenzenes, such as indazole precursors ii) the sensitivity toward ortho substitution, iii) a low sensitivity to electron density. Overall, this study shows how combining Bayesian optimization, as well as DoE, with self-driving labs and automation can be employed to develop robust electrosynthetic reactions on a larger scale.