Photocatalysis for organic synthesis has experienced an exponential growth in the  past 10 years. However, the variety of experimental procedures that have been reported to perform photon–based catalyst excitation has hampered the establishment of general protocols to convert visible light into chemical energy. To address this issue, we have designed an integrated photoreactor for enhanced photon capture and catalyst excitation. Moreover, the evaluation of this new reactor in eight photocatalytic transformations that are widely employed in medicinal chemistry settings has confirmed significant performance advantages of this optimized design while enabling a standardized protocol.

■ Introduction

The utilization of photon-harvesting molecules to convert visible light to chemical energy, known as photocatalysis, has long been a key technology in many important processes such as water splitting, 1 , 2 CO 2 reduction, 3 and solar energy capture. 4 Recently, however, the application of visible-light photoredox catalysis to synthetic organic chemistry has become an area of significant interest. 5 Indeed, over the past 10 years there has been an exponential growth in the number of photocatalysis studies that have been reported in the organic literature. 6 Among several important features, the ability of photoredox catalysts to generate
open-shell organic species in a controlled and selective fashion has enabled the invention of a myriad of owerful technologies for bond construction. 7 , 8 Moreover, the merger of photoredox catalysis with additional modes of catalytic activation (e.g., transition metal catalysis, organocatalysis) has led to unanticipated breakthroughs in the development of fragment- coupling reactions of value to medicinal and process chemistry. 9

The utilization of visible and UVA light and thecommoditization of low-cost high-energy LEDs has enabled
photoredox catalysis to be quickly deployed in both academic andindustrial settings. While broad focus has been placed on the goalof new reaction invention, there has also been significant interestin improving the rates of these newly discovered photon-mediated

Screenshotreactions while developing standardized operating protocols. The Beer-Lambert Law dictates that the photonic flux  decreases exponentially with depth in a given reaction medium. As such, for any given visible-light photoredox reaction, it is reasonable to assume that only the  reaction medium proximal to the vessel wall (i.e., within 2 mm) will experience irradiation. 10 Moreover, as many  studies employ the use of directional lamps with cylindrical vials or rounded flasks, a great  deal of photonic energy is lost due to reflection (Figure 1). As such, the majority of organic photoredox transformations that have been developed to date are believed to be operating  in a “photon-limited” regime as a result of (i) low photon penetration relative to vessel width  and (ii) diminished photon capture arising from poor surface area exposure. 11

 

For any photon-limited regime, it is readily appreciated that an increase in light intensity will  ead to a proportional increase in photon-capture events and thereafter the  concentration of excited state photocatalyst. As one might imagine, the formation of higher  levels of activated photocatalyst will often lead to improvements in the rates and efficiencies  of many elementary catalytic steps, a scenario that can enhance overall reaction times and  generally improve efficiencies. With this in mind, many research groups have approached  the issue of limited photon pen-

Screenshot2-etration via relatively straightforward operational changes, e.g., placing the reaction vessel closer to the  light source or increasing the effective light intensity  by “numbering-up,” wherein multiple lamps are used  in combination. Unfortunately, for these cases, the radiant energy from the light source can often result  in decreased yields arising from unproductive thermal  athways. Moreover, the introduction of  cooling systems to circumvent this problem often  results in cumbersome operational protocols (and  often without improved performance). During our own studies over the past decade, we have found that the  light source, geometry, and distance can significantly  alter the reaction efficiency and reaction profile. From  our correspondence with other research groups, a general consensus has emerged that a  standardized setup for photochemical reactions would likely be broadly adopted, not only to  enhance reproducibility but also to aid in the invention or  discovery of novel bond-forming  reactions. With this in mind, it is important to note that the use of automated flow  technologies can provide significant levels of standardization of path length, light intensity  and geometry 12 , 13 for photocatalysis; however, at the same time this technology is not  broadly deployed for small- scale applications across academic, pharmaceutical, fragrance, agrochemical, or materials laboratories. 14

photoreactor that has been engineered to optimize catalyst photon capture. In an effort to  validate both the design and applicability of this system, we have selected eight  photocatalytic reactions that are commonly employed in the realm of medicinal chemistry to function as benchmark protocols. Importantly, improved efficiencies and accelerated  reaction rates were observed across this range of reaction classes using this new integrated  photoreactor.

■ Reactor Design

To characterize the total radiant power being delivered to the reaction mixture, we chose a  simple analytical technique and a mathematical model based on Newton’s law of cooling. 15  , 16 A rudimentary calorimeter was constructed by embedding a fine thermocouple into a  piece of isotropic graphite of the same dimensions as a typical 2 mL reaction mixture in a  2-dram vial. This apparatus allows for the generation of a temperature curve during the  testing period, which entails exposing the graphite sample to irradiation for one minute,  followed by a cooling period until the temperature returns to its initial starting point. Using  the aforementioned mathematical model (see Supporting Information), the optical power  absorbed by the graphite sample was determined by curvefitting. With this model in hand,  we constructed and subsequently tested a wide range of LED setups

Screenshot3

 

 

 

 

 

 

 

 

 

 

 

 

 

 

that allow for higher photon capture in comparison to the commonly employed blue LED  lamps. In the event, optimum photon capture was observed when the reaction vial was suspended 6mm above an array of four 3.5 mm square 450 nm XTE LEDs (Cree, Inc., Durham,  NC). These LEDs were chosen for their efficiency (>35%), high output (>1.1 W per  LED) and a package size readily available in different wavelengths. 17 Notably, while a  shorter distance between the LED array and the reaction vessel resulted in higher levels of  photon capture, this system suffered from inefficient cooling and issues with reaction temperature control (vide infra). As a second critical design element, a tubular mirrored  casing was employed to ensure that surface reflected photons could be productively  redirected back to the vessel (Figure 2). More specifically, the use of this reflective chamber  ensures that 360 degrees of the vial surface area can be subjected to photon exposure (in  contrast to 180 degrees via a directed LED lamp). Indeed, calorimeter measurements  revealed a 10x increase in total incident radiant power with this XTE system relative to a  standard LED Kessil lamp apparatus. 18 With these improvements in hand, this new photon  delivery setup was quickly implemented into the design and fabrication of a prototype  photoreactor using in-house 3D printing technology.

Our next objective was to engineer an integrated photoreactor that would deliver cooling,  stirring, operational simplicity and,  most importantly, highly consistent results. With respect  to cooling, an axial fan was located underneath the LED array, providing heat extraction for  both the reaction vial and the LEDs (Figure 3). This forced convection cooling manifold  proved to be simple and highly effective in maintaining reaction temperatures across a  broad range (e.g., 25-60 °C) using variable fan speed. As outlined above, the 6 mm vertical  gap between the LED array and the reaction vial was found to be optimal for photon capture  and minimal thermal flux. For integrated stirring, a brushless motor with a rare-earth  magnet was placed immediately underneath the LED array. For operational convenience, the  tirring rate and fan speed are controlled via a Raspberry Pi controller with a touchscreen, providing simple management of reaction time and,  most significantly, LED  power. It is worth noting that this ability to have control of LED power provides an additional  reaction parameter for optimization of photoredox protocols (a parameter that is extremely  challenging to control with high accuracy using

 

Screenshot4

 

 

 

 

 

 

 

 

 

 

 

 
Screenshot5
conventional protocols). 19 Additional design features were included to further expand the  capability of the integrated photoreactor. For example, the LED array was built in a pluggable  odule, allowing users to quickly exchange the irradiation wavelength to best fit  the maximum absorbance wavelength of the photocatalyst or sensitizer (Figure 3). A modular  ial holder set was designed to accommodate different vial sizes such as 4, 8, 20,  and 40 mL vials allowing routine reaction scales from milligram to gram scale. These holders ensure consistent placement while maintaining the optimum vessel-to-LED distance. 20  Finally, the reactor features a light shield and interlock to ensure safe operation that  removes user exposure to high-energy visible and UVA light (see Supporting Information).

 

■ Reaction Comparison

We surveyed the utility of the integrated photoreactor by examining its performance in eight  hotocatalytic reactions that are commonly employed in medicinal chemistry.

Stephenson Trifluoromethylation. As a calibration point, the trifluoromethylation of  2-acetyl-N-Boc pyrrole was performed as originally described by Stephenson et al. 21 This reaction was reproduced in good yield and in an operationally concise timeframe (62% yield,  60 minutes) using a standard LED strip protocol. As shown in Figure 4, when the same  transformation was performed using our integrated photoreactor, we were able to achieve  the same level of efficiency (64% yield) in a remarkably short reaction time (~3 minutes).

Li Trifluoromethylation. The recently reported protocol for the trifluoromethylation of aryl  rings by the Li group was also investigated. 22 Again, we were able to successfully reproduce  he previously reported conditions for the trifluoromethylation of 1,3,5- trimethoxybenzene using a 26 W CFL setup, to afford the desired adduct in 60% yield after  14 hours. Using the photoreactor system, we were pleased to observe a seven-fold rate enhancement to generate the trifluoromethylated adduct in 70% yield after only 2 hours  (Figure 5).

Decarboxylative Fluorination. We next examined the capacity of this integrated  photoreactor to accelerate transformations that have already been shown to be extremely  rapid. As shown in

Screenshot6Figure 6, the decarboxylative fluorination of  secondary carboxylic acids using standard  40 W blue LEDs proceeded swiftly in  excellent yield (92% yield, 90 seconds). 23  Remarkably, this highly efficient fluorination  protocol was accelerated two-fold using our photoreactor to deliver the desired adduct  in only 45 seconds (88% yield).

We next turned our attention to  metallaphotoredox-catalyzed protocols that  enable a variety of C–C and C–heteroatom couplings. Given that these transformations  are mechanistically founded upon multiple  catalytic cycles that must function in concert, we were interested to determine the impact of  enhanced photocatalyst excitation.

Molander BF 3 K Arylation. In 2014, Molander and coworkers published a seminal  manuscript that described the metallaphotoredox-catalyzed coupling of benzyl  trifluoroborate

 

 

Screenshot7

 

 

 

 

 

 

 

 

 

 

 

 

Screenshot8

 

salts with aryl halides using nickel catalysis (Figure 7). 24 Under the standard setup (Kessil  lamp 40 W blue LEDs), we were able to successfully reproduce the reported protocol to give  the desired alkylation product in 97% yield after 24 hours. 25 Remarkably, the integrated  photoreactor was able to shorten this reaction time to only 4 hours while achieving a  comparable yield (98% yield), a net six-fold rate enhancement.

Decarboxylative Arylation. Using our previously described standard setup (Kessil lamp 40 W  lue LEDs), cyclohexane carboxylic acid was converted to the corresponding decarboxylative arylation adduct in good yield (58% yield) over the course of 3 hours. 26  When the coupling reaction was performed using the integrated photoreactor, the same transformation was accomplished in superior yield (64% yield) after only 20 minutes,  providing almost an order of magnitude of

 

 

Screenshot9

 

 

 

 

 

 

 

 

 

 

 

 

Screenshot10Decarboxylative Alkylation. Good rate acceleration was also observed when the integrated  photoreactor was deployed for the decarboxylative alkylation of N-Boc-proline. 27 As shown  in Figure 9, we observed a nearly three-fold rate acceleration over the published standard  setup, to forge the alkylated pyrrolidine adduct in 9 hours (86% yield).

Cross-Electrophile Coupling. The most commonly employed photoredox transformation in  the pharmaceutical sector at the present time appears to be the silyl-mediated cross- electrophile coupling reaction (Figure 10). 28 , 29 As such, we were disappointed to find that  our integrated photoreactor exhibited lower efficiency than the reported Kessil lamp-based  protocol in our initial comparison tests. However, we quickly recognized that the use of the  maximum LED output on the photoreactor was promoting a significant increase in reaction  rate, that in turn was causing a rapid build-up of deleterious HBr (see Supporting  Information). In the case of the reported Kessil lamp protocol, the same acid is produced,  however, at a rate at which it can be readily neutralized by sodium carbonate, the  heterogenous inorganic base employed. Indeed, when we lowered the photoreactor LED  output to 5%, we were able to completely reproduce the original report in terms of time and  efficiency. Moreover, in an effort to identify reaction conditions that would allow accelerated  reaction times, we subsequently examined a more soluble organic base with the hope that  the required neutralization step could be kinetically matched with that of HBr production.  Indeed, through the implementation of 2,6-lutidine in lieu of sodium carbonate, the  integrated photoreactor could be utilized at full power LED output without loss in efficiency  and with a six-fold enhancement in reaction time (40 minutes versus 4.5 hours). This result  serves to reinforce the utility of being able to monitor and implement variable LED intensity  with a high level of accuracy and reproducibility, as is possible using the Raspberry Pi  interface on this integrated photoreactor.

Photocatalytic C–N Coupling. In addition to C–C bond formation protocols, we also evaluated  a metalla-photocatalyzed amination of aryl halides. 30 Using the standard setup with an   inexpensive ruthenium photocatalyst (see Supporting Information)

Screenshot11we were able to reproduce the coupling of  4-bromobenzotrifluo- ride with morpholine in  excellent yield in 40 minutes (Figure 11, 95%  yield). Remarkably, by deploying the  integrated photoreactor, we observed a  four-fold reduction in reaction time to produce  he same amination product in high  yield and only 10 minutes.

Last, and perhaps most important, we have  now validated the utility of the new integrated  photoreactor across ten different medicinal  chemistry groups located at four different  Merck and Co. research sites in the USA. The  successful translation of reaction conditions from previous LED protocols to this  standardized photoreactor will be reported in the near future

■ Conclusions

In conclusion, we have designed an integrated small-scale photoreactor that can be  employed broadly in the realm of visible- light photocatalysis. The integrated photoreactor  enables enhanced light exposure and catalyst excitation across a wide range of  photocatalytic reactions and in doing so provides significant rate accelerations in all cases.  Indeed, the evaluation of this new reactor in eight photocatalytic transformations that are widely employed in medicinal chemistry settings has confirmed significant performance  advantages of this optimized design. Moreover, its successful utilization across multiple  research sites highlights its value as a standardized system that enables operational  convenience and reproducibility.

■ Author Information

Corresponding Author
*dmacmill@princeton.edu, *ian_davies1@merck.com
Author Contributions

*C.C.L., M.K.W., and Z.-C. S. contributed equally to this work.

■ Acknowledgement

Financial support was provided by the NIH NIGMS (R01 GM103558-06) and kind gifts from  Merck, Bristol-Myers Squibb, Abbvie, Janssen, and Eli Lilly. We would like to thank the following scientists and engineers at Merck & Co., Inc. and Princeton University for input on  reactor design and experimental and technical assistance during pre-production validation  of the units. From Princeton University: Patricia Zhang, Megan Shaw, Ryan Evans, Russell  Smith, Jack Twilton, Jack Terrett. From Merck & Co., Inc.: Jack Scott, Shane Krska, Haiqun Tang, Younong Yu, Steve Colletti, Ed Hudak, Kevin Dyksra, Don Henry. Catherine White, David  Candito, David Sloman, Min Lu, Sam Kattar, Charles Yeung, Michael VanHeyst, Valerie  Shurtleff, Tom Greshock, Jim Perkins, Peter Manley, Abdellatif El Marrouni, Jaume Balsells,  Dan DiRocco, Dani Schultz, Andrew Hoover, Sumei Ren, Francois Levesque, Emily Corcoran,  John Naber, Matthieu Jouffroy, Prashant Salve, Leo Joyce, Christopher Nawrat, Steven Oliver.

Supporting Information Available.  Experimental procedures and measurement data for the  ntegrated photoreactor are provided (PDF). Parts list and construction details are  available upon request.

  References.?

(1) Grätzel, M. Artificial photosynthesis: water cleavage into hydrogen and oxygen by visible  light. Acc. Chem. Res. 1981, 14, 376-384.
(2) Meyer, T. J. Chemical approaches to artificial photosynthesis. Acc. Chem. Res. 1989, 22,    63-170.
(3) Takeda, H.; Ishitani, O. Development of efficient photocatalytic systems for CO 2 reduction  sing mononuclear and multinuclear metal complexes based on mechanistic studies. Coord.  Chem. Rev. 2010, 254, 346-354.
(4) Kalyanasundaram, K.; Grätzel, M. Applications of functionalized transition metal  omplexes in photonic and optoelectronic devices. Coord. Chem. Rev. 1998, 177, 347- 414.
(5) Prier, C. K.; Rankic, D. A.; MacMillan, D. W. C. Visible Light Photoredox Catalysis with  ransition Metal Complexes: Applications in Organic Synthesis. Chem. Rev. 2013, 113,  322-5363.
(6) Shaw, M. H.; Twilton, J.; MacMillan, D. W. C. Photoredox Catalysis in Organic Chemistry. J.  rg. Chem. 2016, 81, 6898-6926.
(7) Staveness, D.; Bosque, I.; Stephenson, C. R. J. Free Radical Chemistry Enabled by Visible  ight-Induced Electron Transfer. Acc. Chem. Res. 2016, 49, 2295-2306
(8) Romero, N. A; Nicewicz, D. A. Organic Photoredox Catalysis. Chem. Rev. 2016, 116,  0075-10166.
(9) Skubi, K. L.; Blum, T. R.; Yoon, T. P. Dual Catalysis Strategies in Photochemical Synthesis.  hem. Rev. 2016, 116, 10035-10074.
(10) Shvydkiv, O.; Gallagher, S.; Nolan, K.; Oelgemöller, M. From Conventional to Microphotochemistry: Photodecarboxylation Reactions Involving Phthalimides. Org. Lett. 2010, 12, 5170-5173.
(11) Chu, L.; Lipshultz, J. M.; MacMillan, D. W. C. Merging Photoredox and Nickel Catalysis:  he  irect Synthesis of Ketones by the Decarboxylative Arylation of a-Oxo Acids. Angew. Chem.  Int. Ed. 2015, 54, 7929-7933.
(12) Andrew, R. S.; Becker, J. J.; Gagné, M. R. A Photoflow Reactor for the Continuous  Photoredox-Mediated Synthesis of C-Glycoamino Acids and C-Glycolipids. Angew. Chem. Int.  Ed. 2012, 51, 4140-4143.
(13) Tucker, J. W.; Zhang, Y.; Jamison, T. F.; Stephenson, C. R. J. Visible-Light Photoredox  atalysis in Flow. Angew. Chem. Int. Ed. 2012, 51, 4144-4147.
(14) Elliott, L. D.; Knowles, J. P.; Koovits, P. J.; Maskill, K. G.; Ralph, M. J.; Lejeune, G.; Edwards, L. J.; Robinson, R. I.; Clemens, I. R.; Cox, B.; Pascoe, D. D.; Koch, G.; Eberle, M.; Berry, M. B.;  ooker-Milburn, K. I. Batch versus Flow Photochemistry: A Revealing Comparison of Yield and Productivity. Chem. Eur. J. 2014, 20, 15226-15232.(15) Whewell, W. History of the Inductive Sciences, Nabu Press: Charleston, 2010.
(16) Burmeister, L. C. Convective Heat Transfer, 2 nd ed. Wiley-Interscience: Hoboken, 1993.
(17) Information regarding the Cree XLamp XTE LEDs can be found at www.cree.com.
(18) See Supporting Information for experimental measurement and comparison of different setups.
(19) Frazier, C. P.; Palmer, L. I.; Samoshin, A. V.; Read de Alaniz, J. Acessing nitrosocarbonyl compounds with temporal and spatial control via the photoredox oxidation of N-substituted  hydroxylamines. Tetrahedron Lett. 2015, 56, 3353-3357.
(20) Development of integrated photoreactor that can hold 6 to 12 reaction vials are underway. The design and results of this study will be reported in due course.
(21) Beatty, J. W.; Douglas, J. J.; Miller, R.; McAtee, R. C.; Stephenson, C. R. J. Photochemical Perfluoroalkylation with Pyridine N-Oxides: Mechanistic Insights and Performance on Kilogram Scale. Chem 2016, 1, 456-472. (22) Li, L.; Mu, X.; Liu, W.; Wang, Y.; Mi, Z.; Li, C.-J. Simple and Clean Photoinduced Aromatic Trifluoromethylation Reaction. J. Am. Chem. Soc. 2016, 138, 5809-5812.
(23) Ventre, S.; Petronijevic, F. R.; MacMillan, D. W. C. Decarboxylative Fluorination of Aliphatic  arboxylic Acids via Photoredox Catalysis. J. Am. Chem. Soc. 2015, 137, 5654-5657. (24) Tellis, J. C.; Primer, D. N.; Molander, G. A. Single-electron transmetalation in organoboron  cross-coupling by photoredox/nickel dual catalysis. Science 2014, 345, 433-436.
(25) Studies were carried out using a modified version of the procedure originally reported by Molander and coworkers; Luo, J.; Zhang, J. Donor-Acceptor Fluorophores for Visible- Light-Promoted Organic Synthesis: Photoredox/Ni Dual Catalytic C(sp 3 )–C(sp 2 ) Cross-Coupling.  CS Catal. 2016, 6, 873-877.
(26) Zuo, Z.; Ahneman, D. T.; Chu, L.; Terrett, J. A.; Doyle, A. G.; MacMillan, D. W. C. Merging photoredox with nickel catalysis: Coupling of a-carboxyl sp 3 -carbons with aryl halides.  cience 2014, 345, 437-440. The reactivity of cyclohexane carboxylic acid was further  ptimized in our laboratory.
(27) Johnston, C. P.; Smith, R. T.; Allmendinger, S.; MacMillan, D. W. C. Metallaphotoredox-catalysed sp 3 –sp 3 corss-coupling of carboxylic acids with alkyl halides. Nature 2016, 536, 322-325.
(28) Zhang, P.; Le, C. C.; MacMillan, D. W. C. Silyl Radical Activation of Alkyl Halides in Metallaphotoredox Catalysis: A Unique Pathway for Cross-Electrophile Coupling. J. Am.Chem. Soc. 2016, 138, 8084-8087.
(29) An informal survey of several pharmaceutical companies in the US has shown that this  ransformation is most commonly employed for C sp 3-C sp 2 coupling at the present time. (30) Corcoran, E. B.; Pirnot, M. T.; Lin, S.; Dreher, S. D.;  DiRocco, D. A.; Davies, I. W.; Buchwald, S. L.; MacMillan, D. W. C. Aryl amination using ligand-free Ni(II) salts and photoredox catalysis. Science 2016, 353, 279-283.