A modular autonomous flow reactor combining monitoring technologies with a feedback algorithm is presented for the synthesis of the natural product carpanone. The autonomous self-optimizing system, controlled via MATLAB, was designed as a flexible platform enabling an adaptation of the experimental setup to the specificity of the chemical transformation to be optimized. The reaction monitoring uses either online high pressure liquid chromatography (HPLC) or in-line benchtop nuclear magnetic resonance (NMR) spectroscopy. The custom-made optimization algorithm derived from the Nelder-Mead and golden section search methods performs constrained optimizations of black-box functions in a multidimensional search domain, thereby assuming no a priori knowledge of the chemical reactions. This autonomous self-optimizing system allowed fast and efficient optimizations of the chemical steps leading to carpanone. This contribution is the first example of a multistep synthesis where all discrete steps were optimized with an autonomous flow reactor.

No abstract