Optimisation of flow chemistry: tools and algorithms

The coupling of flow chemistry with automated laboratory equipment has become increasingly common and used to support the efficient manufacturing of chemicals. A variety of reactors and analytical techniques have been used in such configurations for investigating and optimising the processing conditions of different reactions. However, the integrated reactors used thus far have been constrained to single phase mixing, greatly limiting the scope of reactions for such studies. This thesis presents the development and integration of a millilitre-scale CSTR, the fReactor, that is able to process multiphase flows, thus broadening the range of reactions susceptible of being investigated in this way.
Following a thorough review of the literature covering the uses of flow chemistry and lab-scale reactor technology, insights on the design of a temperature-controlled version of the fReactor with an accuracy of ±0.3 ºC capable of cutting waiting times 44% when compared to the previous reactor are given. A demonstration of its use is provided for which the product of a multiphasic reaction is analysed automatically under different reaction conditions according to a sampling plan. Metamodeling and cross-validation techniques are applied to these results, where single and multi-objective optimisations are carried out over the response surface models of different metrics to illustrate different trade-offs between them. The use of such techniques allowed reducing the error incurred by the common least squares polynomial fitting by over 12%. Additionally, a demonstration of the fReactor as a tool for synchrotron X-Ray Diffraction is also carried out by means of successfully assessing the change in polymorph caused by solvent switching, this being the first synchrotron experiment using this sort of device.
The remainder of the thesis focuses on applying the same metamodeling and cross-validation techniques used previously, in the optimisation of the design of a miniaturised continuous oscillatory baffled reactor. However, rather than using these techniques with physical experimentation, they are used in conjunction with computational fluid dynamics. This reactor shows a better residence time distribution than its CSTR counterparts. Notably, the effect of the introduction of baffle offsetting in a plate design of the reactor is identified as a key parameter in giving a narrow residence time distribution and good mixing. Under this configuration it is possible to reduce the RTD variance by 45% and increase the mixing efficiency by 60% when compared to the best performing opposing baffles geometry.