Flow chemistry, where chemical reactions are run in millimeter diameter tubing instead of flasks, offers efficient mixing and heat transfer, scalability, reproducibility, facile optimization, and the ability to generate reactive species in a controlled manner. Recently these techniques have become more popular in the small molecule community for the synthesis of active pharmaceutical ingredients. Using these methods for polymer and nanomaterial synthesis or functionalization however, remains significantly underexplored. In this work, we discuss an active learning AI-planner driven experimental campaign to optimize the nucleophilic substitution reaction of pyridine with an alkyl halide. Our approach relies on an open loop, iterative workflow built on a Bayesian optimization algorithm with exploration and exploitation policies to maximize (or minimize) objective responses. This machine learning guided experimentation protocol allows objective trade-offs to be more efficiently mapped when working in large parameter spaces. Using this approach, we implemented a multi-objective optimization algorithm to maximize the yield and production rate of our target reaction. This allowed us to identify a set of non-dominated solutions, or Pareto front, where each objective is considered equally good depending on the desired goals. Furthermore, we will discuss the extension of these methods and experimental conditions to the modification of polymers.

Synthetic reaction optimization using flow chemistry and an active learning algorithm

Where thousands of chemistry professionals meet to share ideas and advance scientific and technical knowledge. Spring 2023 abstracts, posters, and presentations will launch at the end of March.

ACS Spring 2023

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.