Parallel multi-droplet platform for reaction kinetics and optimization

Abstract

We present an automated droplet reactor platform possessing parallel reactor channels and a scheduling algorithm that orchestrates all of the parallel hardware operations and ensures droplet integrity as well as overall efficiency. We design and incorporate all of the necessary hardware and software to enable the platform to be used to study both thermal and photochemical reactions. We incorporate a Bayesian optimization algorithm into the control software to enable reaction optimization over both categorical and continuous variables. We demonstrate the capabilities of both the preliminary single-channel and parallelized versions of the platform using a series of model thermal and photochemical reactions. We conduct a series of reaction optimization campaigns and demonstrate rapid acquisition of the data necessary to determine reaction kinetics. The platform is flexible in terms of use case: it can be used either to investigate reaction kinetics or to perform reaction optimization over a wide range of chemical domains.

Fig. 1. Parallelized platform schematic with N parallel stationary reactors.

Fig. 2. Temperature control of droplet and droplet headspace. The blue region is not explicitly temperature-controlled. The red region corresponds to the aluminum plates that sandwich the reactor, which can be both heated and cooled to match user- or algorithm-specified temperature. LH: liquid handler. LC: liquid chromatography (HPLC).

Fig. 3. SolidWorks sketches of the photochemistry hardware. (a) LED board with LED layout corresponding to the stationary reactor. (b) Stationary reactor tubing path. (c) SolidWorks sketch of the reactor, Peltier module, heat sink, and fan housing.

Scheme 1. Nucleophilic aromatic substitution reaction used to validate the single-channel version of the platform.

Scheme 2. Aerobic oxidation of 9,10-diphenylanthracene to the corresponding peroxide in the presence of Ru(bpy)₃Cl₂ used to validate irradiance uniformity.

Scheme 3. Model photoredox reaction used to verify that different LED boards behave similarly.

Fig. 4. Demonstration of the need for a scheduling operation. The reactors are bookended by bottleneck operations that only one reaction droplet can occupy at one time: the liquid handling operation (represented in green in the figure), transit form the liquid handler to the target reactor and from the target reactor to the HPLC (both of which occupy the main flowpath, shown in gray), and the HPLC itself (represented in orange in the figure). In scenario 1, the reactions – the only parallelized aspect of the platform – are long relative to the bottleneck operations. Under these circumstances, it works to run the droplet preparation steps back-to-back. In scenario 2, the reactions are relatively short; here, droplet preparation cannot be back-to-back without collisions. By pausing the droplet preparation until an appropriate time, collisions are avoided.

Scheme 4. Reaction used to demonstrate the accurate and rapid determination of kinetic parameters on the platform.

Fig. 5. Results of the kinetics investigation. (a) Temporal reaction profiles at 70 °C for two different initial concentrations of the limiting reagent, 1-fluoro-4-nitrobenzene: 0.8 M and 1.04 M. Scatter points represent experimental data; line plots represent the kinetic model. (b) Temporal reaction profiles at temperatures from 60 °C to 90 °C in ten-degree increments, with the initial concentration of the limiting reagent, 1-fluoro-4-nitrobenzene, fixed at 1.04 M. Scatter points represent experimental data; line plots represent the kinetic model. (c) Eyring plot for measured rate constants. Linear regression of the experimental data yielded eqn (3), which is plotted with a dashed line in the figure.

Scheme 5. Reaction used to showcase automated closed-loop optimization on the platform.

Fig. 6. Optimization results in DMF and DMSO. Marker shapes denote the identity of the base: circles for DBU and triangles for BTMG. Marker colors denote the identity of the catalyst and solvent: green denotes DMF, purple denotes DMSO, blue denotes tBuBrettPhos Pd G3, and orange denotes tBuXPhos Pd G3.

Scheme 6. Reaction used to showcase automated closed-loop optimization on the platform.