(632d) Automated Microfluidic System for Online Optimization of a Chemical Reaction Network

Authors: 
McMullen, J. P., Massachusetts Institute of Technology
Jensen, K. F., Massachusetts Institute of Technology


Determining the optimal conditions for chemical reactions common in fine chemistry and pharmaceutical applications is a difficult task. Typically, sufficient knowledge of the reaction kinetics is not known or is difficult to evaluate accurately due to the numerous competing reactions. Therefore, black-box optimization and response surface modeling techniques are used to find the reaction conditions that maximize product yield. However, the time and resources required to perform these experiments in the traditional batch environment are significant. Alternatively, the speed and efficiency of these optimization procedures can be improved by using integrated microreactor systems with feedback algorithms.

A silicon microreactor system was developed to determine the optimal conditions of a chemical reaction by varying several process parameters. The optimization platform includes syringe pumps for fluid delivery, a silicon microreactor, thermoelectric modules for heating/cooling, and HPLC for online monitoring of the reaction. Given an objective function, such as maximizing yield, and a defined parameter space, the platform selects and performs sequential experiments according to an optimization algorithm until a specified tolerance is achieved. Several algorithms were incorporated into the platform, including the Nelder-Mead Simplex, Stable Noisy Optimization by Branch and Fit (SNOBFIT), and a gradient-based method using design of experiment methods to fit a response surface. We demonstrate rapid reaction optimization capabilities with experimental throughput rates of one completed and analyzed experiment per 10 minutes depending upon reaction residence time and HPLC analysis.

Several chemical reactions were explored in the microfluidic optimization platform. Examples include optimizing the intermediate product of a series reaction ? the oxidation of benzyl alcohol to benzaldehyde with further oxidation to benzoic acid ? and a reversible Knoevenagel condensation where the product formation can be limited to equilibrium and other competing reactions. In each of these applications, the desired objective function was maximized by varying the reaction temperature, residence time, and reagent concentrations. The potential to use the platform for kinetic modeling and production scale-up will also be discussed. orm for kinetic modeling and production scale-up will also be discussed.