Machine learning directed multi-objective optimization of mixed variable chemical systems

The consideration of discrete variables (e.g. catalyst, ligand, solvent) in experimental self-optimization approaches remains a significant challenge. Herein we report the application new mixed variable multi-objective optimization (MVMOO) algorithm for chemical reactions. Coupling MVMOO with an automated continuous flow platform enabled identification trade-off curves different performance criteria by optimizing and concurrently. This approach utilizes Bayesian methodology to provide high efficiency, enhances process understanding considering key interactions between variables, requires no prior knowledge reaction. Nucleophilic aromatic substitution (SNAr) palladium catalyzed Sonogashira reactions were investigated, where effect solvent ligand selection on regioselectivity efficiency determined respectively whilst simultaneously determining optimum parameters each case.