High-Productivity Single-Pass Electrochemical Birch Reduction of Naphthalenes in a Continuous Flow Electrochemical Taylor Vortex Reactor

We report the development of a single-pass electrochemical Birch reduction carried out in a small footprint electrochemical Taylor vortex reactor with projected productivities of >80 g day–1 (based on 32.2 mmol h–1), using a modified version of our previously reported reactor [Org. Process Res. Dev.2021, 25, 7, 1619–1627], consisting of a static outer electrode and a rapidly rotating cylindrical inner electrode. In this study, we used an aluminum tube as the sacrificial outer electrode and stainless steel as the rotating inner electrode. We have established the viability of using a sacrificial aluminum anode for the electrochemical reduction of naphthalene, and by varying the current, we can switch between high selectivity (>90%) for either the single ring reduction or double ring reduction with >80 g day–1 projected productivity for either product. The concentration of LiBr in solution changes the fluid dynamics of the reaction mixture investigated by computational fluid dynamics, and this affects equilibration time, monitored using Fourier transform infrared spectroscopy. We show that the concentrations of electrolyte (LiBr) and proton source (dimethylurea) can be reduced while maintaining high reaction efficiency. We also report the reduction of 1-aminonaphthalene, which has been used as a precursor to the API Ropinirole. We find that our methodology produces the corresponding dihydronaphthalene with excellent selectivity and 88% isolated yield in an uninterrupted run of >8 h with a projected productivity of >100 g day–1.

. Preliminary Optimisation of conditions for the flow de-aromatisation of 1a using TPPA.
When the reaction mixture contained TPPA, in each case the reactor blocked between 20-30 min after the current was applied. Figure S5. Showing the stainless steel rotor after a blockage occurred under the conditions using TPPA. The rotor was found to be coated in a black tar like material (this material also filled the annulus of the reactor).

S10
Extended Reactor Operation Figure S6. The substrate (1b) used for the extended run. Figure S7. The crude product (2b) recovered from the extended 8hr run. Figure S8. The reaction mixture after centrifuging to separate the black residue. S12 Figure S9. 1H NMR spectra of the crude reaction mixture (following mini work-up by extraction with 1.0 M aqueous potassium sodium tartrate), showing consistent >90% selectivty between the Birch reduction product and starting material.

Examination of the sacrificial electrode following the > 8 hr run.
The bore of the tube was measured at both ends prior to the experiment. At the end of the run, the electrode was cut in half at the point indicated by the blue bands. The bore was then measured at the red, green and two blue ends, Figure S10. The fluid flowed from the red end to green. Then the green/blue tubular section was cut in half longditudinally so that their interior surfaces could be examined (see Figure S11). S13 Figure S10. Diagram of the cut Al tubular electrode. Figure S11. Cut Al tubular electrode. Note the traces of dark deposit at the green (outflow) end of the electrode. There is no evidence for any localised erosion of the surface. Visual inspection of the bore of the blue/red section did not reveal any obvious erosionso it was not cut in half. Additional FTIR Spectra Figure S12. Infrared spectra collected as the reactor feed was switched from dry THF to the reaction mixture as it was introduced into the reactor and flowed out of the reactor outlet (Blue to Red, thin lines). Pure reference spectra of the individual mixture components are overlaid for comparison and characterisation (thick lines) This process was monitored via an ATR-FTIR fibre optic probe and allowed for the reactor dynamics to be monitored. S16 Figure S13. Infrared spectra of the reactor outlet (Blue to Red, thin lines) the current was applied to the reaction mixture and reduction began. A characteristic product band of the first reduction

S14
(1,4-dihydronaphthalene) can be observed to emerge at ca. 750 cm -1 as the primary naphthalene peak at ca.790 cm -1 decreases in intensity. Other changes can be observed in the spectrum between 1150-1700 cm -1 these relate to additional 1,4-dihydronaphthalene bands as well as a shift in the DMU peaks. S18 Figure S15. These changes can be more easily seen in this difference spectrum. This difference spectrum again shows the appearance product band at ca. 750 cm -1 and the disappearance of the naphthalene peak at ca. 790 cm -1 , as well the new product peaks at ca. 1200 cm -1 and shift in the DMU peaks. This shift in the DMU manifests as an increase in intensity around 1600 cm -1 , and a decrease around 1570 and 1670 cm -1 , resembling the difference observed by Baran and co-workers when DMU is removed from the presence of LiBr. This could suggest that the lithium is no longer complexing the DMU after the reduction has taken place.

Figure S16
These data could then be analysed using multivariate curve resolution to extract he kinetic information from the process of introducing the reactants to the reactor, and then initiating the reduction. This was achieved by using the reference spectra as hard bounds in the algorithm, and then allowing for two extra "free" factors to be fit to the data in order to extract the rest of the constituents, (in this case the shifting THF-LiBr / DMU / 1-4 Dihydronaphthalene for product formation, and dry THF) Figure S17. Inline FTIR spectra showing peaks relating to 1b (~785 cm -1 ) and DMU (~775 cm -1 ) during the production of 2b, demonstrating the need for Raman spectroscopy to probe <700 cm -1 in order to monitor the production of 2b which possess characteristic vibrational modes in that range.

S21
Additional Raman Spectra Figure S18. Raman spectra of components 1a and 2a with the solvent subtracted.

A-TEEM Spectroscopy
EEM spectra consist of an excitation emission matrix with an intensity axis, usually derived from number of photons of a given wavelength reaching the detector. This intensity is therefore given in counts can then be normalized between 0 and 1. This can be illustrated by the EEM spectrum of 2b. as demonstrated in Figure S19.

Figure S19
Normalized EEM Spectrum of 2b as a contour plot (left) and 3D mesh (right). The normalized EEM spectra can be depicted as a contour plot of excitation vs. emission vs.
fluorescence intensity (Blue: low intensity, Red: high intensity) or as a 3D projection, In this paper, we have used the 2-D projections for simplicity.

Computation Fluid Dynamic (CFD) Details
The CFD simulations were conducted using the commercial software ANSYS-Fluent 2022R1.
The modelled electro-vortex reactor was two dimensional and axisymmetric with a mesh size of 146k nodes. The CFD method is validated with the experiemtnal data from literature (J. Colloid Interface Sci. 2005, 285, 167-178), where the velocity profiles are used for the comparison. A species of a mass fraction of 0.1 that has the same properties of the working solution is injected from the rotor to model the electro-reaction. At low rotation speed (e.g., 100 RPM), the flow in the reactor is laminar, but at high rotation speed (e.g., 4000 RPM), the flow is turbulent and was modelled using the k-w turbulence model. The laminar case was modelled without acceleration due to the low rotation speed, thus, the simulation was initiated from the rest. However, for the turbulent case, an acceleration of 500 RPM is used progressively starting from the laminar case The mixing efficiency is calculated using the following formula: where Yi,ini is the initial mass fraction injected from the rotor and equal 0.1, and Yi,area-ave is the area-averaged mass fraction of the injected species through the whole reactor.