Automated Execution and Optimization of Flow Chemistry on a Robotic Platform with Integrated Analytics
The development, optimization, and characterization of chemical processes for the synthesis of organic compounds, which play a key role in society as medicines and materials, is currently an expensive and laborious enterprise. These inefficiencies are driving efforts to develop automated, data-rich...
| 第一著者: | |
|---|---|
| その他の著者: | |
| フォーマット: | 学位論文 |
| 出版事項: |
Massachusetts Institute of Technology
2022
|
| オンライン･アクセス: | https://hdl.handle.net/1721.1/145015 https://orcid.org/ 0000-0002-5009-0361 |
| 要約: | The development, optimization, and characterization of chemical processes for the synthesis of organic compounds, which play a key role in society as medicines and materials, is currently an expensive and laborious enterprise. These inefficiencies are driving efforts to develop automated, data-rich experimentation (DRE) platforms and methods designed to maximize the amount of useful data generated per unit time and raw material expended. In this thesis, a modular robotic platform for continuous flow synthesis was utilized for machine-assisted organic reaction development. An improved version of the platform was built with new capabilities including a Cartesian robot for fast and reliable pick-and-place, integrated process analytical technology (PAT) such as LC-MS and FT-IR spectroscopy for online reaction monitoring, and closed-loop feedback optimization of reaction conditions using a Bayesian optimization algorithm. In the first case study, algorithmic reaction optimization helped partially automate the specification of critical process parameters (both continuous and categorical) for a computer-proposed and human-refined synthetic route. A representative multistep synthesis involving 3 reactions (including a heterogeneous hydrogenation) and 1 separation was chosen. In multistep flow processes where downstream residence times are physically constrained by upstream flow rates, the modular reactor volumes of the robotic platform were leveraged to introduce an independent degree of freedom. Deployment of multiple PAT tools facilitated thorough process understanding and workflow automation helped accelerate and reduce the manual burden during experimentation. In the second case study, the platform’s toolkit was further expanded with the addition of an LED array to perform photochemistry. This new capability enabled the development of two photochemical steps that lead to an important class of drugs. Bayesian optimization aided in optimizing continuous variables including residence time and stoichiometry, and characterizing the effect of critical process parameters. Finally, the design of data-rich dynamic flow experiments, where continuous reactors are operated under controlled transients in input variables, was computationally studied and experimentally validated. Mathematical modeling using transport equations and a parametric analysis helped identify a simple criterion to guide the design of dynamic trajectories. Sinusoidal dynamic experiments designed using the criterion were executed on the robotic platform with two and three simultaneously varying inputs. |
|---|