The capacity of the trifluoromethyl (CF3) group to confer enhanced metabolic stability, bioavailability, lipophilicity, and potency to organic small molecules drives continued efforts to develop new methods for the selective incorporation of C-CF3 bonds in pharmaceuticals and agrochemicals. [1][2][3][4][5][6] Methods to prepare CF3-containing molecules are versatile and robust, and include nucleophilic, [7][8] electrophilic, [9][10] and radical 11 CF3 transfer processes. A recent emphasis on direct C-H trifluoromethylation has prompted a revisitation of radical alkylations, which can install the CF3 functional group in unactivated arenes and heteroarenes. [11][12] Minisci-type radical functionalization of heteroarenes is not new, 13 but the past decade has seen a "renaissance" in redox methods for catalytic C-H trifluoromethylation, which include the development of photoredox methods for generation of free • CF3. 12,[14][15][16][17][18][19] The • CF3 radical is a strong electrophile, so regioselectivity in these systems, or lack thereof, is determined by the stereoelectronics of the substrate.
Recent advances in selective C-H functionalization by organometallic catalysts suggest a path to selectivity in C-H fluoroalkylations. However, most state-of-the-art methods for C-C coupling cannot reliably be extended to C-CF3 bond formation due to intrinsic properties of the M-CF3 intermediates. Whereas early transition metal M-CF3 bonds readily undergo α-fluoride abstraction to generate difluoromethyl carbene complexes, [20][21][22] M-CF3 bonds to low-valent later 3d metals are often thermodynamically robust and kinetically inert. 23 Accordingly, recent successes in the development of metalmediated trifluoromethylations activate M-CF3 bonds via formation of, for instance, high formal oxidation state complexes, which are prone to C-CF3 reductive elimination. [24][25][26] In this context, photoredox activation of organometallic catalysts, termed multimetallic or metallaphotoredox catalysis, 19,27 has received considerable attention for C-C bond formation, including radical alkylations from homolysis of Co(III)-alkyl complexes. 28 With few exceptions, these processes separate the light-harvesting species from the transition metal catalyst. The role of the chromophore-most commonly a polypyridyl Ru or Ir complex-is to directly activate the metal complex via excited state oxidation or reduction or to generate a nonmetal free radical coupling partner, which is subsequently trapped at a transition metal center in a catalytic cycle for C-C or C-X coupling. 19 Although the benefits of these approaches are many, including the capacity to utilize nontraditional reaction partners in cross coupling, bimetallic excited state reactivity limits the use of base metals in photoredox catalysis. [29][30][31][32][33] Moreover, separating the chromophore from the bond-forming reaction adds a layer of complexity when pursuing metal-mediated selectivity and opens paths to side reactions from transient free radicals.
Photoinduced M-L bond activations are a staple of organometallic synthesis, and the capacity of cobaloxime organocobalt(III) complexes to generate alkyl radicals via facile Co-C homolysis has been known and exploited for synthesis applications for decades. 28 Such "visible light induced homolysis" (VLIH) has been proposed as a generalizable alternative to utilize 3d transition metals in photocatalytic applications. 34 We reported a [(OCO)Co III (CF3)(MeCN)] (I) compound supported by a pincer-type [OCO] ligand, which trifluoromethylates unactivated (hetero)arenes upon irradiation by a broad-spectrum compact fluorescent light (CFL) (Figure 1a). 35 The redox-active [OCO] is an essential feature of the observed reactivity. 36 Our search for alternative sources of CF3 led to electrophilic CF3 reagents, such as Umemoto's S-(trifluoromethyl)dibenzothiophenium triflate (1), which have been extensively studied for trifluoromethylation. 9,[37][38][39][40] Their shelf stability and ease of handling have made them oftentimes preferred over cheaper, alternative CF3 sources such as ICF3 and HCF3, 41 and over the past decade these reagents have seen extensive use in catalytic trifluoromethylations. In reactions with these reagents, the transition metal typically acts as an outer-sphere, one-electron reductant, thus generating • CF3 without M-CF3 intermediates. 4,[42][43] Addition of a + CF3 equivalent to a transition metal is a formal 2e -oxidation of the metal center, which demands a 2e -redox capacity at the metal center. This is exemplified by work of Sanford and coworkers, wherein formal + CF3 transfer generates a Ni IV -CF3 complex, which is active for trifluoromethylation of unactivated arenes. 44 By analogy, addition of + CF3 to our (OCO)Co II complex affords an (OCO)Co(CF3) species two redox levels above Co(II), and one above the (OCO)Co III (CF3) that was previously demonstrated to be photoactive for trifluoromethylation (Figure 1b). Given the propensity of the redox-active [OCO] ligand to support Co in four formal oxidation states, 36 we speculated that net + CF3 addition to an (OCO)Co II species would afford an (OCO • )Co III (CF3) product and opening avenues for catalysis based on a 2e -redox cycle (Figure 1c). But it was unclear whether (OCO • )Co III (CF3) would retain the features necessary for photoactivation of the Co-CF3 bond.
Reported herein is a Co catalyzed photoredox method for efficient trifluoromethylation of unactivated arene and heteroarene C-H bonds using 1 and visible light. Mechanistic studies and stoichiometric reactions provide strong evidence for a 2e - redox cycle involving net + CF3 addition to a Co(II) complex to generate a photoactive electronic structures include a low-spin Co(III) center with a single unpaired electron on To distinguish these possibilities, the electronic structure of II was computed with unrestricted DFT calculations (BP86, def2-TZVP) starting in the experimentally determined doublet state. The bond lengths of the optimized geometry in the doublet state were compared to the single-crystal X-ray structure and found to have a mean absolute error of 0.013 Å. In the (OCO)Co III (CF3) fragment, a maximum bond length deviation of 0.006 Å was observed, suggesting that the spin state, functional, and basis set (BP86, def2-TZVP) used for geometry optimization accurately capture the bond distances in the complex. The spin density per atom in optimized geometry was also calculated. Complex II converged as a doublet (⟨s 2 ⟩ = 0.7533), with 22.4% of the total spin density being located at cobalt (Figure 3). The balance of the spin density is delocalized over both phenoxide arms of the OCO ligand, mainly on the oxygen atoms (total of 26.4%). A small amount of spin-down density on the NHC and two aryl carbons can be attributed to spin polarization. A quartet state (⟨s 2 ⟩ = 3.7996) is computed to be +33 kcal mol -1 higher in energy than the doublet state, making its involvement highly unlikely. UCO analysis of α and β orbitals showed only one orbital with an overlap integral less than 0.999: the singly occupied molecular orbital distributed across the Co center and the [OCO] ligand. [45][46] The computed structure of II is therefore most consistent with an [(OCO•)Co III (CF3)(THF)OTf] assignment, with a low-spin Co(III) center and a monooxidized [OCO • ] -ligand radical, corroborating experimental observations and computation of very similar systems. [35][36] Partial delocalization of the unpaired spin onto the Co center reflects significant covalency in the metal-ligand bonding or minor contributors to the ground state, which would not be readily evident in common spectroscopic methods.
The UV-vis spectrum of II in CH2Cl2 (Figure 4a) exhibits four CT bands at 438, 460, 518, and 889 nm (ε = 2400-6600 M -1 cm -1 ). Excited states were examined by TDDFT calculations (BP86, def2-TZVP) from the doublet ground state. The calculated and experimental UV-vis spectra of complex II are in strong agreement (Figure 5a). The difference density CEJ plot of the transition at 447 nm (Figure 5b) shows strong ligandto-metal-charge-transfer (LMCT) character and results in 89% population of the calculated LUMO (Figure 5b), which has significant antibonding character between the Co center and the CF3 ligand, suggesting photoexcitation of II with 447 nm light should significantly destabilize the Co-CF3 bond towards homolysis. Consistent with this hypothesis, photolysis of a 0.25 mmol solution of II in CH2Cl2 using a Kessil KSPR 160L-440 LED lamp resulted in a measurable decrease in the intensity of the CT bands at 438, 460, and 518 nm within minutes, with a concomitant increase in intensity and a red shifting of the band at 890 nm to 902 nm and appearance of a new band at 337 nm (Figure 4b). Isosbestic points at 400 and 625 nm are consistent with clean conversion to a single product or mixture of products without formation of observable intermediates. The spectrum after 4 h of continuous photolysis closely matched that of isolated [(OCO)Co III (THF)2]OTf (IV). 36 Photoinduced homolysis of the Co II -CF3 bond occurs at an LMCT excited state of II to afford an excited state intravalence isomer of IV. The net conversion of II to IV is balanced by loss of • CF3. The fate of the • CF3 under these conditions was not determined, but CF3Cl is most likely from reaction with CH2Cl2 solvent. The UV-vis spectrum of II in CH2Cl2 was unchanged over 7 h in the dark at 50 °C, suggesting Co-CF3 homolysis occurs from a photoexcited state and not thermolysis from heating by the light source (Figure S2). containing byproducts. Analysis of the reaction mixture by GC-MS revealed a second product with a molecular weight of 579 m/z, consistent with CF3 addition to the [OCO] ligand and demetallation. Whereas increasing the concentration of TEMPO • to 15 equiv at the same concentration of II resulted in an increase in yield of TEMPO-CF3 to 24%, increasing both the TEMPO • and II concentration by 60% resulted in a 40% decrease in TEMPO-CF3 yield (7%). These observations are consistent with a bifurcated reaction, Photolysis of a 0.025 mM CH2Cl2 solution of II using a Kessil KSPR 160L-440 LED lamp in the presence of 5 equiv C6H6 afforded α,α,α-trifluorotoluene in 10% yield after 6 h, as determined by 19 F NMR spectroscopy. Consistent with the partitioning experiments described above, performing the reaction in neat C6H6 increased the yield of α,α,α-trifluorotoluene to 58% in 6 h. Monitoring the reaction by UV-vis spectroscopy shows the reaction occurring in two sequential phases. Consumption of II initially generates a spectrum that closely resembles IV with two quasi-isosbestic points at 637 and 405 nm (Figure 6a). 36 Continued photolysis results in bleaching of the intermediate peaks at 889, 732, and 443 nm with concomitant growth of a feature at 357 nm, which is diagnostic of [(OCO)Co II (THF)] (III) (Figure 6b). The net conversion of II + C6H6 to III + α,α,α-trifluorotoluene is balanced by loss of 1 equiv HOTf (Eq 2). HOTf formation, in the form of a triflate salt, is evident in catalytic reactions as a singlet at -79 ppm in the 19 F NMR spectrum of reactions performed in J. Young NMR tubes (vide infra). 47 Its appearance in stoichiometric reactions, however, is frequently obscured, presumably by an interaction with paramagnetic III or a rapid exchange process. Accordingly, addition of mmol solution of II in C6H6 upon exposure to ambient light. Spectra are shown from t = 55 m (green) and 5 min intervals to t = 120 m (red). A spectrum of isolated [(OCO)Co II (THF)] (III) in C6H6 (black) is shown for comparison. Eq 2 Photocatalytic arene C-H trifluoromethylation. Addition of a + CF3 fragment to the Co center in III affords II via a 2e -process analogous to oxidative addition (Figure 1b). Accordingly, reactions with Umemoto's tetrafluoro-S-(trifluoromethyl)dibenzothiophenium triflate (1) were pursued with the aim of closing a catalytic cycle for photocatalytic arene C-H trifluoromethylation.
A combination of III (5 mol%) with a 1:1 mixture of C6H6 and Umemoto's S-(trifluoromethyl)dibenzothiophenium triflate (1) in MeCN afforded α,α,α-trifluorotoluene in 18% yield after 6 h of continuous irradiation using a 440 nm blue LED lamp. (Table 1, entry 4). Major byproducts included triflouromethylated dibenzothiophenes, which result from attack of the promiscuous • CF3 radical on the heteroarene product of + CF3 removal from 1. 48 Accordingly, increasing the ratio of C6H6 to 1 to 5:1 gave a three-fold increase in yield of α,α,α-trifluorotoluene and significantly depressed the competitive dibenzothiophene trifluoromethylation (
Table 1, entry 6). A maximum yield of α,α,αtrifluorotoluene was observed with a 10-fold excess of C6H6 substrate (Table 1, entries 7 and 8). Control experiments performed without cobalt gave a maximum yield of 35% (Table 1, entries 1 and 2), but in all cases the yield of α,α,α-trifluorotoluene is significantly increased by the addition of III. Use of CoCl2 in place of III gave a statistically insignificant increase in the measured yield relative to the metal-free conditions (Table 1, entry 3).
Reactions performed in the dark with and without III gave no measurable α,α,αtrifluorotoluene (Table 1, entries 9 and 10).
a Yields were determined by integration of 19 F NMR resonances using C6F6 as an internal standard based on 1 as the limiting reagent, as described in the Supporting Information.
b Reaction performed in the dark.
The scope of the C-H trifluoromethylation was evaluated using six different (hetero)arene substrates (Figure 7). There was no significant change in yield when making the arene ring more electron rich (3b); there was a significant decrease in yield when the ring was more electron poor (3c). Pyrrole (3d) exhibited high yield and excellent selectivity with the only isomer formed being trifluoromethylation at the 2-position of the ring. Indole (3e) showed moderate reactivity, but poor selectivity with an overall 35% yield and a 1:1 ratio between the 2-and 3-positions of the indole backbone. Treatment of a dark orange solution of III in CH2Cl2 with 1 equiv 1 in the absence of light gave an immediate color change to dark olive-green. Monitoring the reaction by UV-vis spectroscopy showed the complete disappearance of diagnostic bands for III at 355 and 438 nm and the commensurate appearance of a spectrum with features at 451, 518, and 890 nm (Figure S3), which matches a 1:1 mixture of II and IV. The 50% yield of II in the stoichiometric reaction apparently results from photodegradation of II during the synthesis, suggesting the combination of III + 1 is a path to photoactive II and providing entry to functional catalysis. Accordingly, mixing III with 20 equiv 1 in a MeCN solution containing 200 equiv C6H6 in the dark gave full consumption of the diagnostic CT bands . 35 The solid-state structure of II contains a -OTf ligand in the sixth site, giving II quasi-octahedral geometry. Simple MO arguments would expect -OTf coordination to similarly raise the energy of the photoactive dz 2 LUMO, but this is insufficient to deactivate II. Computational data suggests a dz 2 -like orbital is still the primary contributor to the LUMO in II. The possibility of equilibrium -OTf dissociation generating a photochemically active five-coordinate species cannot be ruled out under the catalysis conditions, but the fact that II retains its photosensitivity in MeCN solvent argues against deactivation in octahedral geometries. This makes II more versatile by giving access to a wider range of solvents and opening avenues for trifluoromethylation of substrates that can act as strong σ donors to Co, as described below.
As a stoichiometric source of free • CF3 radical, II suffers in comparison to I. Under analogous conditions, the yield of α,α,α-trifluoromethyltoluene from C6H6 is reduced from >99% to 58%. The origin of this disparity is apparently an enhanced propensity of II to trap • CF3 via C-C coupling to the ligand backbone, which subsequently induces demetallation and formation of unidentified Co byproducts (Scheme 1). No evidence for ligand centered • CF3 radical coupling has been observed in photolysis of I. It is tempting to ascribe this partitioning to the presence of unpaired spin on the [OCO•] -ligand making II a more effective radical trap, but radical character on the arene is not a prerequisite for radical coupling and the relative kinetics of the C-C bond forming reactions in the excited states are entirely unknown at this time. This flaw is not fatal, however. Competitive trapping at the ligand can be disfavored by lowering the concentration of II and increasing the relative concentration of the organic acceptor. That is, the conditions one would typically pursue in catalysis are exactly those required for high-yielding reactions with the organic substrates.
Clean conversion of III to II using 1 establishes all the steps required for functional photocatalysis, as illustrated in Scheme 2. The use of 1 here demands a multielectron redox capacity that would typically limit the use of cobalt, as the thermodynamically preferred oxidation +2 and +3 states are incompatible with the formal 2e -redox change that occurs upon + CF3 addition to the metal center. The redox active [OCO] ligand sidesteps this issue by coupling 1e -redox at cobalt with 1e -oxidation at the ligand, as shown in the blue pathway in Scheme 3. The net 2e -reaction can occur without accumulation of 1e -intermediates because of strong electronic coupling and covalency in the Co-OCO bonding. 36
Scheme 3. Proposed photocatalysis mechanisms of arene trifluoromethylation using III.
Given its well-established propensity to participate in radical • CF3 transfer, the use of 1 as a CF3 source merits additional discussion. Although 1 was initially conceived as a source of + CF3, radical mechanisms have been frequently invoked. 49 These suggest initial 1e -transfer generates a reduced form of Umemoto's regents (1 • in Scheme 3), which is itself a source of • CF3 radical. Accordingly, single electron transfer (SET) and radical chain mechanisms have been suggested for trifluoromethylations using 1 with common photoreductants or reducing metals in their ground state. 43-44, 48, 50-52 Ground state initial ET, illustrated by the red 1e -path Scheme 3, is unlikely here based on the redox potentials of the reactants. Whereas 1 is reduced at -0.75 V vs Fc + /Fc, [51][52] oxidation of III occurs at -0.32 V vs Fc + /Fc, 36 implying initial outer-sphere 1e -transfer is uphill by over 400 mV. Moreover, reaction of III with 1 in the dark gives exclusively II; ruling out 1 • as a viable route to trifluoromethylated arene in the absence of light.
Direct of 1 has been reported and suggested to occur via in situ generation of π-π complexes with arenes, which are competent visible light chromophores. [53][54] This may account for the significant background reactivity observed herein. To our knowledge, the excited state reduction potential of 1 is unknown.
Accordingly, the red 1e -pathway cannot be rigorously excluded as a contributor to the observed catalysis under constant photolysis.
Irradiation of II with visible light induces facile Co-CF3 bond homolysis to generate a persistent • CF3 electrophile capable of attacking (hetero)arenes (Scheme 3). Rearomatization of the initially formed cyclohexadienyl radical requires net loss of H • .
Formation of III in these reactions implies a net 2e -reduction of the cobalt center, and UV-vis data are consistent with this occurring by two consecutive 1e -steps via the Co(III)
intermediate IV. The role of IV is an outer-sphere 1e -oxidant rather than an H • acceptor.
The net PCET reaction is balanced by loss of H + in the form of HOTf, which apparently generates dibenzothiophenium triflate under the catalysis conditions. 55 Throughout this manuscript, the ligand trans to the carbene in the [OCO] ligand is largely ignored because it is a spectator. Work on these and closely related ET series have revealed a capacity to bind a range of O-and N-donor substrates, including furans, pyridines, and nitriles in this position. [35][36] The photochemistry of II is seemingly unaffected by substitution at this site. Accordingly, the selectivity reported herein is determined by the persistent • CF3 radical and the arene substrates themselves. Trifluoromethylation of the [OCO] ligand demonstrates the capacity of an inner-sphere radical acceptor to function as a competitive trap, presumably via geminate recombination, suggesting an oriented substrate should be similarly susceptible to aryl trifluoromethylation within the solvent cage.
The novelty of the catalysis reported herein is not the organic products, the use of light to generate
• CF3 radical, or even the application of organocobalt(III) complexes for trifluoromethylation, but rather extrapolation of the catalyst design principles for photolytic C-H arene trifluoromethylation to functional catalysis and the avenues for selective photoredox catalysis that result. Redox noninnocence of the [OCO] ligand is a central and essential feature of every step of the catalysis cycle. First, by imparting a multielectron redox capacity at Co, it permits use of CF3 sources that demand formal 2e -oxidation of the metal center to generate the active organometallic intermediate while circumventing high energy Co(IV) or Co(I) species. Second, the ligand redox flexibility preserves the low spin Co III -CF3 core across multiple formal oxidations states, which makes the complexes thermodynamically inert but photochemically labile. Finally, the capacity of the [OCO] framework to be oxidized at modest potentials provides a reservoir of accessible ligandcentered electrons for generation of the photochemically active LMCT state with low energy light. The result is a coordination complex that functions as a combined chromophore and organometallic reaction center for a visible light photoredox catalysis cycle, obviating the requirement for long excited-state lifetimes which often limit the use of base metals in photoredox catalysis. The VLIH design principles elaborated in this system are broadly transferrable; our ongoing work is focused on extensions to other, inexpensive sources of CF3, structural modifications to tune the absorption properties and photochemistry of the organometallic chromophores, and selectivity in radical C-H trifluoroalkylations of heteroarenes that derives from spatial proximity of the C-H bonds to the incipient • CF3 than the substrate electronics.
General Considerations.
Unless otherwise mentioned, all operations were carried out under anaerobic conditions using standard vacuum line techniques or in an inert-atmosphere glovebox under nitrogen. All NMR spectra were recorded on a Varian Mercury 400 spectrometer and chemical shifts are reported in parts per million (ppm) relative to TMS, with the residual solvent peak as an internal reference. 56 All 19 F chemical shifts are reported in ppm relative to CFCl3 with hexafluorobenzene as an internal standard. Solution magnetic moments were obtained by the Evan's NMR method. [57][58] UV-vis absorption spectra were acquired using a Hitachi 4150 spectrophotometer. Unless otherwise noted, all electronic absorption spectra were recorded at ambient temperature in a 1 cm quartz cell. All mass spectra were recorded in the Georgia Institute of Technology Bioanalytical Mass Spectrometry Facility. Electrospray Ionization mass spectrometry (ESI-MS) was carried out with acetonitrile solutions using a Micromass Quattro LC spectrometer. Elemental analyses were performed by Atlantic Microlab, Inc., Norcross, GA. All analyses were performed in duplicate, and the reported compositions are the average of the two runs. Full details of X-ray data collection and refinement are provided in the Supporting information.
Anhydrous acetonitrile (MeCN), dichloromethane (CH2Cl2), benzene, tetrahydrofuran (THF), pentane, and toluene solvents for air-and moisture-sensitive manipulations were purchased from Sigma-Aldrich and further dried by passage through columns of activated alumina, degassed by at least three freeze-pump-thaw cycles, and stored under N2 prior to use. Hexamethyldisiloxane (HMDSO) was degassed by at least three freeze-pumpthaw cycles and stored over activated 4 Å molecular sieves under N2 prior to use.
Methanol (Drisolv) was purchased from EMD Millapore and used as received. Deuterated solvents were purchased from Cambridge Isotope Laboratories. Acetonitrile-d3, DCM-d2, and THF-d8 were placed an in oven dried sealable flask and degassed by freeze-pumpthaw cycles then stored over activated 4 Å molecular sieves under N2 prior to use. AgCF3
was prepared according to a published procedure. 59 Silver fluoride (Strem), TMSCF3
(Oakwood), Umemoto's reagent (Sigma), hexafluorobenzene (Sigma) were all used as received. [(OCO)Co(THF)], 36 [(OCO)Co(THF)2]OTf (IV), 36 and [(OCO)Co(CF3)(MeCN)] (I) 35 were prepared by published procedures. C-H Trifluoromethylation of (Hetero)Aryls.
In a representative procedure, a borosilicate NMR tube with a J Young valve was charged with 1 (80.5 mg, 0.2 mmol), C6H6 (180 L, 2 mmol, 10 equiv.), III (5 mg, .01 mmol, 5% mol) and CD3CN (1 mL). The tube was sealed and placed ~3 inches from a Kessil KSPR 160L-440 LED lamp for 6 hours. Hexafluorobenzene (11.5 L, 0.1 mmol) was added as an internal standard and yields were measured by integration against the 19 F resonances for the CF3 containing products. The NMR spectra matched those previously reported. 14,44,50,[60][61][62] Computational Studies.
DFT calculations were performed using ORCA 4.2.1 [63][64] using the BP86 [65][66] functional and def2-TZVP [67][68] basis set (default grid) on the full model. TD-DFT calculations were performed at the same level of theory using the output coordinates from the geometry optimization as input. Spin density, difference density, and molecular orbital plots were generated using IQmol (http://www.iqmol.org/), and IboView (http://www.iboview.org).