(438a) Robotic Platform for Execution, Analysis, and Optimization of Multistep Reactions in Continuous Flow

Hart, T., Massachusetts Institute of Technology
Kulesza, T., Massachusetts Institute of Technology
Florit, F., Massachusetts Institute of Technology
Breen, C. P., Massachusetts Institute of Technology
Monos, T. M., Massachusetts Institute of Technology
Jamison, T., Massachusetts Institute of Technology
Jensen, K. F., Massachusetts Institute of Technology
Automated systems for chemical synthesis integrated with computer-aided synthesis planning and algorithm-guided optimization of reaction conditions have helped relieve chemists from routine tasks and accelerated process development [1,2,3]. The synthesis of organic compounds during process development often involves multiple steps with both discrete (e.g., catalyst, solvent) and continuous (e.g., temperature, time) process variables. In continuous flow synthesis where multiple steps can be telescoped, an important question is whether for a given optimization objective (e.g., yield, productivity), the combination of optimal conditions for each step is the same as the globally optimal conditions for the entire sequence. The ability to analyze the process inline between two steps in addition to at the end of a multistep synthesis can provide valuable insight into the effect of process variables on individual steps.

Here, we present a robotically reconfigurable flow chemistry platform capable of executing, analyzing, and optimizing multistep reactions. The platform contains a library of process modules that can be placed in any order onto a process stack by the robot for performing reactions, separations, and inline analysis (liquid chromatography–mass spectrometry (LC-MS) and Fourier transform infrared (FT-IR) spectroscopy). Reagents are delivered to the process stack via reconfigurable fluidic connections which the robot can place onto a reagent “switchboard”, enabling switching between multiple reagent candidates. The hardware was coupled to an optimization algorithm that employs optimal design of experiments (DoE) to intelligently navigate the design space with as few experiments as possible, and branch and bound (B&B) to handle discrete variables in addition to continuous variables [4]. The LC-MS and FT-IR modules were utilized simultaneously in multistep syntheses to analyze the process at multiple locations, providing reaction-specific information. Several case studies highlight the platform’s capabilities for chemical synthesis and explore the question of how to find globally optimal reaction conditions for multistep flow synthesis.