A Small Footprint, High Capacity Flow Reactor for UV Photochemical Synthesis on the Kilogram Scale

The development of a highly compact and powerful reactor for synthetic organic photochemistry is described enabling a 10 fold reduction in reaction times, with up to 30% more power efficiency than previous FEP tube reactors. Two reactions gave over 1kg of product in 24 hours. Two other reactions had productivities of 4 and 8 kg in 24h. The reactor consists of a succession of quartz tubes connected together in series and arranged axially around a variable power mercury lamp. This compact and relatively simple device can be safely operated in a standard fumehood.


Introduction
Over the last ten years there has been a very significant increase in the use of continuous flow reactors in organic synthesis. 1 A variety of reactors have been developed for numerous organic reactions in both academic and industrial labs. In many cases these offer an alternative to classic batch reactors and can often display significant advantages in heat and mass transfer, efficient mixing, scale-up, safety and cost. In 2005 we reported 2 the development of a practical flow reactor for photochemical synthesis on scales of up to a few hundred grams per day. This reactor consisted of a single length of UV transparent fluorinated ethylene propylene (FEP) tubing wrapped closely around a high power UV source. One to three layers of FEP could be coiled around the UV source to give a reactor of high surface area and excellent UV capture ( Figure  1as). Many labs worldwide 3,4 have adopted this reactor and variants thereof and a commercial version 5 is now available based on the original concept. Key to the success of the FEP reactor is its ability to totally encapsulate the UV emitting source and as such photon capture by the circulating substrate is near optimal.
A key brief of a recent collaboration 6 was to design a higher capacity reactor to enable photochemistry to be carried out on the kg/day scale. Initially we considered building larger FEP reactors with more powerful UV sources. However, this raised some issues that would be exacerbated by simply 'scaling-up'. For example, although FEP is a versatile material it is not completely UV transparent and any kinks or abrasions leads to weak spots that can rupture under pressure requiring the whole length of FEP tubing to be replaced. Neither is FEP immune to foul-up and blockage. In our experience a blocked FEP tube can sometimes be reversed by a short flush of solvent. On other occasions, especially if there is photochemical damage to the FEP, the whole tubing has to be replaced and the reactor rewound (see SI, With this experience behind us we set out to design a higher capacity flow reactor meeting the following specifications: • Capable of delivering ≥ 1kg per day productivity • Encapsulate UV as effectively as in FEP reactors • Small footprint to fit within a standard fumehood • Replace FEP with durable and more UV transparent tubing • Safe containment of high-power (1-5 kW) UV source

Results & Discussion
Replacing FEP with a more durable material, with at least equal performance characteristics, was non-trivial. Although a spiral quartz reactor would be ideal, and have been described 7 , obtaining the necessary compact, multi-layer coiled quartz device on the scale required for this project was not feasible. A multi-tube reactor system has been reported 8a,b but in these designs the reactor tubes were isolated from each other and acting independently and no synthetic photochemistry was described. We therefore designed a modular reactor comprised of an array of quartz tubes linked in series and arranged axially around a high power UV source so that they effectively operate as one single tube with a large surface area. We proposed that such a reactor could encapsulate light just as effectively as the previous FEP reactors. At the centre of the concentric reactor array would be placed a standard, high power medium pressure Hg-lamp, although this would be able to accommodate different types of lamps and light sources (eg LEDs) to suite a broad range of photochemistry. The reaction solution would enter through an end-cap, flow through one tube, into the opposite end-cap, then back through the adjacent tube in the opposite direction. Thus a relatively thin profile of solution would get high UV exposure by the time it had passed through the entire reactor assembly. Any UV that passes through or between the tubes would be reflected back in very close proximity to the reaction solution.
Initially we built a prototype for a 400 W Hg-lamp contained in a water-cooled jacket (Fig 2). Initial screening with a [2+2] reaction between N-methyl maleimide and hex-1-yne gave a productivity of 3.41g/h, which approached that of an optimized 3-layer FEP reactor 2d (3.62 g/h).
Similarly maleimide and propargyl alcohol gave a productivity of 2.13 g/h vs 2.35 g/h for the FEP reactor (Table 1). These were surprising initial results, especially considering that they were

Quartz Tubes Outlet
Parallel Tube Flow Reactor obtained using the reactor without the metal reflector. However, when the metal reflector was placed around the reactor significant overheating was encountered. It was clear that the stagnant hot air between the water jacket and the reactor tubing was overheating. It was conceived that this could be prevented by encapsulating the reactor tubes within a fluid-cooled annular cavity.
In addition, a fan installed at one end of the reactor would simultaneously remove the stagnant hot air. This dual mode of cooling should be effective at negating the heating effects produced by the powerful UV source (Figure 3a). The lamp would not require containment within a cooling jacket, thus simplifying the design. The inner boundary could also serve as an exchangeable UV filter. A highly polished and earthed outer metal tube would serve to reflect any transmitted UV back into the reactor tubes and to protect the user from intense radiant energy. The reactor (Fig. 3b) was constructed to these specifications, with interlocks put in place to protect against overheating, electrical and coolant failures. A sophisticated variable power supply meant that the lamp could be run at powers ranging from 1.5 -5 kW to match the photochemical reaction productivity. The footprint of this reactor was not much larger than a rotary-evaporator condenser, which is remarkable considering the power of the device. Although mains water could be used for cooling, our results were obtained using a commercial recirculating chiller (water/glycol -see SI). The final operational design 9 of the reactor was named The Firefly. After trialing a few reactions we soon found 3 kW to be a convenient power setting. At this power the reactor gave approximately 10 times greater productivity than the corresponding threelayer FEP flow reactor with a 400 W lamp. For example, the [2+2] cycloaddition of N-methyl maleimide and trichloroethene 2d (  (Table 2, entry 3) can be problematic on account of formation of insoluble by-products (see SI). When using the quartz tube reactor at 3 kW, we were able to process 2.5 L of a 0.1 M solution at 36 ml/min and still maintain almost complete conversion with a productivity of 24.6 g in 70 min (64%, 21.2 g/h).
We believe the high flow rate in this case prevents significant foul-up.
Entries 1-3 were invaluable benchmarks for testing the Firefly reactor as their performances were well understood. However, in terms of scale-up to kg quantities they were limited as the starting maleimides were relatively expensive and chromatography was required.
We therefore sought to test the Firefly with low-cost readily available starting materials that produced complex, structurally diverse molecules that could be easily purified on kg scales. Previously 10 we had shown that 3,4,5,6-tetrahydrophthalic anhydride (THPA, 1) and cis-2-butene-1,4-diol undergo an efficient batch [2+2] cycloaddition-lactonisation sequence to tricyclic lactone 2 upon direct UV irradiation. Repeating the batch irradiation at 0.4 M with just 1% of isopropythioxanthone (iTHX) as sensitizer gave an impressive 15-fold increase in productivity. Use of the parent THX 4p,q led to equally good productivity, however due to the insoluble nature of THX it was more difficult to remove from the crystallized product 2. This demonstrates how previously optimized, but scale-limited reactions can be made dramatically more productive by judicious choice of sensitizer and solvent. Transferring these newly optimized conditions to the Firefly (0.4 M @ 36 mL/min) gave a staggering 1,538 g of pure 2 in just 9.26 h of irradiation, with the product crystalizing out in the receiving flask. In a 24 hour run this productivity would enable the synthesis of almost 4 kg of 2 at this power setting.  (Fig. 2a), by simply swapping the pyrex inner tube for quartz. Ultimately a total of 6.4 moles of 7 in 32 L of MeCN/H2O was irradiated over a total of 18 h at 3kW @ 30 ml/min, giving 1,032 g of pure 8 in 85% overall yield with a productivity of 57.9 g/h ( Table 2, entry 7) Entries 4-7 have all proved to be excellent low-cost reactions with greater than 1kilogram productivities in a 24 h processing period. It is perhaps surprising that with the highpowers and concentrations involved that foul-up of the reactor was much less than expected in long runs. In general, we found that as long as the starting material was pure, and the reactor clean to start-with, then foul-up proved not to be an issue for the examples studied. For example, in entry 5 colored impurities carried over from the preparation of a particular batch of 3 had a deleterious effect on productivity and product quality as well as resulting in minor fouling of the reactor. These unidentified impurities were likely chromophoric quinone derivatives from the previous Diels-Alder step. Although in this specific case substrate 3 was ≥ 95% pure by NMR it does highlight that in photochemistry it is important that substrates are not contaminated with small quantities of compounds that may act as quenchers of the reaction or sensitizers of undesired side reactions. For all other reactions in this study substrate quality of ≥ 95% purity was tolerated without issue.
The Firefly reactor would appear to be significantly more efficient than any of the FEP reactors that we have previously designed. For example entry 2 gave 10.1 times the productivity of the 400W FEP result using only 7.5 times the UV power, making the Firefly almost 30% more power efficient. Although we can only speculate at this stage, this performance enhancement is likely due to the superior UV transparency of the quartz tubing vs. FEP combined with the efficiency of UV capture and reflection of the device.

Conclusions
In summary, we have developed a powerful new flow reactor that enables complex photochemical synthesis to be carried out on a variety of scales whilst reducing reaction times 10-fold compared to previous FEP reactors. The fact that such a small photo-reactor can routinely produce over a kilogram of product, and more, in a 24 h processing period is unprecedented. The small footprint enables the device to be operated in a standard (1.5 m) research lab fumehood. The versatility of the device was demonstrated by successful scale-up of reactions at UVA (Table 2, entry 4-6), UVB (entries 1-3) and UVB/UVC (entry 7) -all using the same Hg-lamp. As the UV lamp is not an integral part of the reactor, it could easily be removed and replaced with sources of different wavelengths such as visible light LEDs to match other types of photochemistry. As UV LED's continue to develop, we expect high-power, compact devices to eventually exceed the productivity and efficiency of the medium pressure lamp presented and such UV sources will be easily accommodated within the Firefly reactor. We hope these formidable capabilities can make significant contributions to synthetic photochemistry in research and process labs within universities and industry and encourage others to use this technology as an enabling synthetic tool. Current work is concerned with developing a commercial version of the prototype Firefly reactor in order to realize these ambitious goals.

Experimental Section
General For photolysis, standard reagent grade solvents were used without further purification. For all photochemical reactions, solvents were "degassed" by evacuating under vacuum and re-filling with N2. For the duration of the reaction, N2 was bubbled through the solution in the holding flask. When photochemical reactions were followed for their duration by NMR, 1,3,5trimethoxybenzene was used as an internal standard. In all cases, a stock solution of this was added to aliquots of the reaction mixture prior to NMR sample preparation.   9, 174.4, 70.3, 59.4, 48.4, 45.7, 42.7, 37.3, 26.1, 26.1, 21.5, 20.9 ppm Two additional runs yielded 577g (3.47 hr) and 192g (1.16 hr) of 2 (Total = 1,538g in 9.26 h).
Although no drop off in productivity was observed during each run, the reactor was flushed with DMSO/water as a precaution before the next run was commenced. This allowed for safe evaporation of three successive volumes of solvent in a research lab setting. All three runs were completed in a single working day. This strategy was adopted for other reactions where large volumes of flammable solvents were involved. In a certified process lab, where real-time evaporation of bulk solvent is routine, it is likely that the Firefly could be run continuously with continuous evaporation of solvent. Any drop in productivity detected by online monitoring could initiate a brief automated shutdown of the reactor, flush with DMSO/water followed by start-up of the reactor.

Representative flow procedure: Table1, Entry 5
A solution of Diels-Alder adduct 17 3 CAS: 51175-59-8 (1,307 g, 7.5 mol)  (Total = 1,032 g in 17.78 hr). Due to the relatively dilute nature of this reaction (0.2 M) it was split into 5 runs in order to make product extraction and solvent evaporation more manageable.
After one run a small residue build-up on the quartz tubes resulted in 13% drop in yield over the run. Before the subsequent run 30% water/DMSO was circulated for ca. 10 min (36 ml/min) and no significant drop in conversion was observed for the next 20 L of reaction solution (see table in   SI).

Funding Sources
Financial support from the Engineering and Physical Sciences Research Council (EPSRC; EP/L003325/1), GlaxoSmithKline and Astra Zeneca is gratefully acknowledged.

ACKNOWLEDGMENT
We thank Darren Mitchell and Paul Dinham (University of Bristol) for construction of reactor prototypes; Richard Little and Dr Jaromir Bilek (Jenton International) for technical advice and Gillian Smith Associates for project coordination

Supporting Information
The Supporting Information is available free of charge on the ACS Publications website at DOI: xxxxxxxxxxxxxxx 1 H & 13 C spectral data for all photochemically prepared compounds, experimental and optimisation procedures for photoreactions, sensitizer screens, pictures of FEP tubing (PDF).