Organocatalyzed Birch Reduction Driven by Visible Light

Abstract

The Birch reduction is a powerful synthetic methodology that uses solvated electrons to convert inert arenes to 1,4-cyclohexadienes—valuable intermediates for building molecular complexity. Birch reductions traditionally employ alkali metals dissolved in ammonia to produce a solvated electron for the reduction of unactivated arenes such as benzene (E_red < −3.42 V vs SCE). Photoredox catalysts have been gaining popularity in highly reducing applications, but none have been reported to demonstrate reduction potentials powerful enough to reduce benzene. Here, we introduce benzo[ghi]perylene imides as new organic photoredox catalysts for Birch reductions performed at ambient temperature and driven by visible light from commercially available LEDs. Using low catalyst loadings (<1 mol percent), benzene and other functionalized arenes were selectively transformed to 1,4-cyclohexadienes in moderate to good yields in a completely metal-free reaction. Mechanistic studies support that this unprecedented visible-light-induced reactivity is enabled by the ability of the organic photoredox catalyst to harness the energy from two visible-light photons to affect a single, high-energy chemical transformation.

Graphical Abstract

INTRODUCTION

Visible-light photoredox catalysis has transformed the synthesis of small molecules and materials through the conversion of photochemical energy to chemical potentials enabling unique reactivity under mild conditions.^1–5 However, the scope of accessible chemical transformations using these catalytic platforms is fundamentally confined by the energetics of a visible photon. For example, a 400 nm photon provides 3.1 eV of energy, defining the upper limit for the thermodynamic driving force for transformations using visible light. Thus, the low electron affinity of inert substrates such as benzene render it unreactive and difficult to reduce by single electron transfer, requiring a reduction potential of −3.42 V vs SCE,⁵ while the high triplet energy of benzene (3.6 eV) prevents triplet energy sensitization.⁶ As such, the reduction of benzene requires harsher conditions than accessible by current visible-light photoredox catalyst systems. The Birch reduction—the prototypical example being the overall 2e⁻/2H⁺ reduction of benzene to 1,4-cyclohexadiene—represents one of the most demanding reductions in organic synthesis and traditionally employs solvated electrons as the reductant, generated using lithium or sodium metal under cryogenic liquid ammonia conditions (Figure 1a and 1b).^7,8 Several variations of Birch reductions have been develeloped, including ammonia free,⁹ electrochemical,^10,11 and photochemical,¹² each of which has increased the safety of performing Birch reductions. Despite these advances, the development of a mild, metal-free, visible-light-driven Birch reduction is highly desirable.

Figure 1.

Background and plausible mechanism of a visible-light-driven Birch reduction. (a) Reaction conditions and considerations for traditional Birch reduction. (b) Mechanism of 2e⁻/2H⁺ reduction of benzene to afford 1,4-cyclohexadiene. (c) Plausible consecutive photoinduced electron transfer (ConPET) catalytic cycle to merge the energetics of two photons to generate a highly reducing solvated electron. D = electron donor (OH⁻ or F⁻ in this work).

To overcome the energetic constraints of a visible-light photon, approaches to harness the energetics of two photons into a single chemical event and access more challenging reactivity have been developed.¹³ For example, under high photon flux conditions using a laser, an iridium(III) complex underwent two successive photoexcitations, allowing ionization to iridium(IV) and a solvated electron;¹⁴ while promising, the practicality and utility of this system for Birch reductions remain unknown. Efforts toward a Birch reduction driven by visible light utilized an iridium complex that served as both a triplet sensitizer of the arene as well as a photoreductant in conjunction with a sacrificial electron donor.¹⁵ However, this system was unable to reduce benzene due to the high triplet energy of benzene, and reactivity was restricted to arenes possessing a lower triplet energy.

In another approach to accessing more reducing power, the concept of consecutive photoinduced electron transfer (ConPET) was applied with a perylene diimide (PDI) system.^16,17 Here, the first photon generates an excited-state PDI that is reduced by a sacrificial electron donor to yield a radical anion. Subsequently, the radical anion is photoexcited by a second photon, generating a much stronger reducing species. This catalytic system was employed in the reduction of aryl halides to generate aryl radicals that could be coupled with an appropriate trapping agent. Similarly, reductive quenching of an acridinium PC was found to generate an acridine radical that could be photoexcited for the reduction of aryl chlorides and the reductive detosylation of amines.¹⁸ However, reduction of the aromatic ring was not observed in any of these systems.

Recently, replacing the first photoinduced electron transfer (PET) step to generate the radical anion, electrochemical reduction of naphthalene imides¹⁹ or dicyanoanthracene²⁰ catalysts was reported. Electrochemical reduction of the catalyst afforded a stable radical anion that can be subsequently photoexcited to a strongly reducing species capable of reducing aryl halides, including aryl chlorides, to aryl radicals for subsequent coupling reactions. Although these systems generate catalyst species with reactivity near Li⁰ and excited-state reduction potentials < −3 V vs SCE, reduction of the arene ring was not observed. Thus, Birch reduction reactivity by a visible-light photoredox catalyst (PC) system remains elusive.

Our interest in photoredox catalysis originated with the motivation to develop strongly reducing organic PCs for organocatalyzed atom transfer radical polymerization (OATRP). Using computationally accelerated discovery we identified N,N-diaryl dihydrophenazines,²¹ N-aryl phenoxazines,²² and N-aryl dimethyl-dihydroacridines²³ as classes of strongly reducing organic PCs. The most successful O-ATRP catalysts have impressively strong excited-state reduction potentials, some possessing E°(PC^•+/PC*) < −2 V vs SCE, representing some of the strongest single-photon visible-light PC reductants known. This work has motivated us to identify even more strongly reducing PC systems targeting the reduction of benzene and other arenes. Acknowledging the limitations of single-photon photoredox catalysis, we envisioned that through exploiting a ConPET process we could realize a catalyst system for the reduction of benzene (Figure 1c).

RESULTS AND DISCUSSION

Our pursuit of a visible-light photoredox-catalyzed Birch reduction led to the investigation of benzo[ghi]perylene monoimides (BPIs) as potential PCs.²⁴ This class of molecules possesses a computationally predicted high-energy lowest unoccupied molecular orbital (LUMO) [or in equivalence, relatively low electron affinity at E⁰_comp(PC/PC^•−) ≈ −1.3 V vs SCE].²⁵ This reduction potential is significantly more negative than that reported for a PDI [E_1/2(PDI/PDI^•−) = −0.43 V vs SCE] that is structurally very similar to the PC used in the ConPET system above.²⁶ Thus, we hypothesized that by utilizing the more strongly reducing PC^•− accessible with BPIs, photoexcitation to PC^•−* would access an even more reducing excited state that might be competent for the reduction of benzene. To test this hypothesis, we synthesized a family of targeted PCs in 2–4 steps from commercial reagents, resulting in a series of BPI molecules possessing electron-neutral, -withdrawing, and -donating core substituents on the 6-, 8-, and 11-core positions of the BPI (Figure 1c). All of these molecules exhibited strong visible-light absorption (wavelength of maximum absorption λ_max > 400 nm and molar absorptivity ε_max > 20 000 M⁻¹cm⁻¹), high-lying LUMOs [E_1/2,exp(PC/PC^•−) < −1.2 V vs SCE], and redox reversibility for single electron transfers as determined by cyclic voltammetry (Table S5).

To investigate the ability of these molecules to serve as PCs in a ConPET mechanism, we first examined the light-induced reduction of the BPI by an electron donor to generate the PC radical anion (PC^•−). While commonly employed trialkyl amine electron donors failed to generate the PC^•−, we found that OH⁻ and F⁻ ions could reduce the PC upon light irradiation through taking advantage of an association reaction analogous to that observed in the light-gated reduction of naphthalene diimides (NDIs) with Cl⁻ as the reductant (we note that the association reaction observed here resulted in a different type of complexation; see below for a mechanistic discussion).²⁷ The lack of reactivity of trialkylamines likely results from both their inability to associate with the PC and the lack of driving force for the electron transfer (E°*_comp[³PC*/²PC^•−] = 0.38 V vs SCE; E_p/2[Et₃N/Et₃N^•+ = 0.83 V vs SCE).²⁸ Our initial survey of the targeted photoredox-catalyzed Birch reduction implemented 2-phenylethanol as the substrate to produce the cyclohexadiene product 2-(cyclohexa-1,4-dien-1-yl)ethan-1-ol (1). Gratifyingly, using the various BPI PCs (0.25 mol %), NBu₄OH (2 equiv) as the electron source,²⁹ in mixed methanol and tert-amyl alcohol as the solvent and H⁺ source, and irradiated with a 405 nm LED resulted in conversion of the arene to the target product, albeit in low conversions (Table S2). The BPI derivative possessing p-OMePh core substituents outperformed the other PCs, resulting in 17% conversion after 16 h. Surveying potential reductants revealed that using the less sterically hindered NMe₄OH and increasing the loading to 10 equivalents resulted in an increase in conversion to 42% after 48 h.

The reaction still proceeded using a lower catalyst loading (0.1 mol %), but conversion did not improve with increased catalyst loading (e.g., 1 mol %), presumably because of quenching of the photoexcited PC (PC*) or the catalyst decomposing as an aromatic substrate in the reaction (Table S2). We found that conversion slowed dramatically after 48 h, so catalyst was added to the reaction at intervals in order to drive the reaction to higher conversions. Excitingly, it was found that 88% conversion (70% isolated yield, 1 in Figure 2b) could be achieved by adding a total of 0.75 mol % of catalyst divided over three additions during the course of the reaction (96 h). Control experiments revealed that the reaction did not proceed or resulted in minimal conversion with omission of any single component (PC, OH⁻, H⁺ source, or light). While the reaction was not oxygen tolerant (PC^•− can be quenched by O₂, E⁰(O₂/O₂^•−) ≈ −1.0 V vs SCE),³⁰ it was tolerant to water.

Figure 2.

Synthesis of 1,4-cyclohexadienes by visible-light-driven Birch reduction of arenes. (a) General reaction conditions. (b) Substrate scope; isolated yields are reported unless otherwise indicated. ^a144 h reaction, otherwise 96 h. ^bYield determined by ¹H NMR using 1,3,5-trimethoxybenzene as the internal standard. ^cSee Supporting Information for details. Scale: 0.5 mmol in arene unless otherwise indicated.

With these reaction conditions established, we explored the general applicability of this photoinduced Birch reduction (Figure 2). The reduction of structurally similar 1-phenyl-2-propanol (a secondary alcohol) also proceeded well to afford product 2 in 62% yield. Notably, additional substituents on the phenyl ring affected reaction efficiency. For example, meta- and para-methylphenylethanol exhibited reduced reactivity, and products 3 and 4 were isolated in 51% and 48% yield, respectively. Introduction of a methyl group at the ortho position significantly inhibited the reactivity (5, 24% yield). Functional groups, such as carboxylic acid (6 and 7), amide (8), carbamate (9 and 10), and strained cyclopropane (7), are well tolerated. The reduction of feedstocks benzene, toluene, and other simple mono- or disubstituted alkyl benzene derivatives were successful as well, affording 11–14 in moderate to high yields (38–80%). Remarkably, the cyclic ether motif was preserved in the high-yielding reduction of isochroman derivatives (15 and 16), while exclusive benzyl C–O bond cleavage occurs under conventional Birch conditions using lithium.³¹ Similarly, 1,3-dihydroisobenzofuran and 1,2,3,4-tetrahydroisoquinoline were transformed to 17 and 18 in 82% and 40% yield, respectively. We also demonstrated that this Birch reduction methodology was capable of generating products 1 and 17 on a larger scale. Performing the reaction at 10 mmol (1.2 g) scale, we achieved 70% conversion (52% isolated yield) to product 1 and 98% conversion (91% isolated yield) to product 17. Scale up required minimal optimization as the same reactant stoichiometry was used in a larger reactor with using four 405 nm LEDs. Under these prescribed reaction conditions this methodology proved less successful or ineffective in the reduction of electron-rich arenes as well as with substrates possessing alkenes, alkynes, alkyl halides, unprotected amines, or nitrogen-containing heterocycles (Figure S1).

Selective reduction of arenes containing multiple reactive unsaturated functional groups could be achieved through modulation of the reaction conditions (Figure 3). Interestingly, reduction of one or both phenyl rings of diphenyl ether were observed (20 and 22) as well as the over-reduction to afford vinyl ethers 21 and 23. Employing optimized conditions, benzophenone proceeded through the tandem reductive deoxygenation and Birch reduction to afford 1-benzyl-1,4-cyclohexadiene 25 in 42% yield. By manipulating the equivalents of NMe₄OH and reaction time, cinnamyl alcohol could be converted to either phenylpropanol 27 through alkene reduction or 28 via both alkene and aromatic reductions. Similarly, trans-2-phenylcyclopropane-1-carboxylic acid underwent a reductive ring-opening process to give 30 in 73% yield, while further reduction provided 31 in 40% yield. In addition, dehalogenation of the pharmaceutical loratadine was facile (33, 65% yield), although significant transesterification also occurred with the solvent (34).

Figure 3.

Selective reductions. Modulation of the reaction conditions enables selective reduction.

To investigate the mechanism underlying this reactivity, spectroscopic studies and density functional theory (DFT) calculations were performed (Figure 4). Overall, these results support a catalytic cycle involving two photon absorption steps in this photoredox-catalyzed Birch reduction (Figure 4a). Prior to irradiation, a color change from orange to yellow was observed upon OH⁻ addition to an orange solution of the PC, suggesting dark reactivity between the PC and OH⁻ (Figure 4b). To further investigate the nature of this interaction, a series of solutions was analyzed by UV–visible spectroscopy (UV–vis) in which the molar ratio of OH⁻ was increased while the total volume of each solution was held constant. In this titration experiment, absorption bands assigned to the PC (λ_abs,max = 353, 423, and 507 nm) decrease with increasing [OH⁻] in conjunction with the appearance of new absorbance features (λ_abs,max = 312 and 412 nm). Monitoring the emission of these same mixtures shows a progressive decrease in the intensity of the PC fluorescence (λ_em,max = 563 nm) with the appearance of emission from a new species (λ_em,max = 481 nm), supporting assignment of a 1:1 equilibrium binding model.³² Fitting the UV–vis data to this 1:1 model yields the equilibrium constant for OH⁻ association (K_eq = 920 M⁻¹). It is relevant to note that in formation of a charge-transfer anion–π complex between iodide and NDIs a red shift was reported, as opposed to the blue shift observed here, suggesting a fundamentally different mode of complexation in this system.³³ Interestingly, ¹³C NMR supports formation of a covalent hydroxide adduct [PC–OH]⁻ (Figure 4b, SI Figures S114–S116) rather than an anion–π type complex.²⁷ A new signal emerges after addition of 100 equiv of either OH⁻ (δ = 173.9 ppm) or F⁻ (δ = 173.0 ppm), which suggests a new covalent bond is formed in the proximity of the imide moiety. Formation of [PC–OH]⁻ is further supported by DFT calculations (Figure 4a), which predict complex formation to be exergonic by 3.0 kcal mol⁻¹ (K_eq,comp = 150 M⁻¹) along with a qualitative blue shift in the predicted lowest energy UV–vis absorption (λ_abs,max,comp from 408 to 376 nm). Thus, spectroscopic data and DFT computations support that OH⁻ attacks the PC reversibly to form [PC–OH]⁻, which comprises the majority species in the equilibrium mixture under the reaction conditions (i.e., OH⁻:PC molar ratio of 800–1333:1 in the reactions performed to determine scope (Figures 2 and 3)). Further, [PC–OH]⁻ is stable in the dark, such that thermally induced electron transfer does not occur and is predicted to be endergonic by 44.8 kcal mol⁻¹, although such ground-state reactivity was observed with the structurally related PDIs and NDIs.^34,35 Finally, control experiments reveal that in the absence of OH⁻, the PC* is not quenched by either benzene or alcohols, supporting that the catalytic cycle proceeds through [PC–OH]⁻ (Figure S18).

Figure 4.

Mechanistic studies. (a) Proposed mechanism of ConPET proceeds through a covalent [PC–OH]⁻ complex; DFT-predicted values are shown in purple and experimentally measured values in turquoise. Electrochemical reference in Volts vs SCE; Ar = p-OMePh. (b) Association of the PC and OH⁻ can be observed by absorption (left), fluorescence (middle), and ¹³C NMR (right) spectroscopy; R = 2-ethylhexyl. (c) PC^•− is stable and can be observed by absorption (left) and EPR (middle) spectroscopy. Computational characterization of PC^•− by visualization of spin density and electrostatic potential (ESP)-mapped electron density depicting electron-rich “red” and electron-poor “green” regions (right). (d) Nanosecond transient absorption spectroscopy of PC^•−; irradiation performed with a 405 nm LED (middle and right).

With the dark speciation of the PC/OH⁻ mixture driven toward [PC–OH]⁻, we next investigated its reactivity under the influence of light (Figure 4c). Upon photon absorption, we propose that PET occurs intramolecularly from OH⁻ to the imide moiety to form the radical anion PC^•− and OH^•. While the fate of OH^• (primarily O^•− under basic conditions)³⁶ is currently unknown, we hypothesize that it may react with the methanol solvent to produce a hydroxymethyl radical and OH⁻;³⁷ further studies to elucidate the ultimate fate of these reactive species are ongoing in our laboratory. This PET step is predicted by DFT to be exergonic by 11.5 kcal/mol and thermodynamically driven by the lowest singlet excited state of the [PC–OH]⁻ [E(S1)_exp = 2.57 eV; E(S1)_comp = 2.44 eV] (Figure 4a). Time-dependent DFT calculations suggest that the relevant absorption for [PC–OH]⁻ under 405 nm LED irradiation is the HOMO–LUMO transition (E_comp = 3.29 eV, 376 nm, f = 0.709), which is of π–π* nature (Figure S118).

Formation of PC^•− is further supported by multiple independent experiments. First, upon irradiation of [PC–OH]⁻ with a 405 nm LED a color change from yellow to purple is observed along with the appearance of several new absorption bands (λ_abs,max = 580 nm). In support that this new species is the persistent radical PC^•−; the same absorbance spectrum could be observed when using F⁻ as the electron source or through bulk electrolysis at an applied potential (E_app = −2.26 V vs SCE) (Figure 4c). In addition, electron paramagnetic resonance (EPR) spectroscopy was used to characterize the photogenerated PC^•−. The experimental EPR spectrum could be reasonably simulated³⁸ and demonstrated delocalization of the unpaired electron on the benzo[ghi]-perylene monoimide core as indicated by its interactions with N (a = 1.985 G), two equivalent methylene Hs adjacent to the N (a = 0.540 G), and seven nondegenerate Hs on the aromatic core (a = 0.611, 0.612, 0.639, 0.644, 0.644, 0.645, and 0.645 G); other parameters used in the simulation include g_isotropic = 2.00171, line width = 0.666 G, and line shape = Gaussian (Figure S42). PC^•− was further characterized by visualization of the DFT-predicted spin density in which the unpaired electron is localized on the imide moiety and to a lesser extent delocalized over the methylene group adjacent to nitrogen and the aromatic system on the BPI core.

Although PC^•− is a relatively strong reductant [E_1/2,exp(PC/PC^•−) = −1.24 V vs SCE; E°_comp (PC/PC^•−) = −1.30 V vs SCE], this reducing power is insufficient for a Birch reduction and reactivity is not observed in the absence of light. Thus, we hypothesized that PC^•− absorbs a second photon to generate PC^•−*, a much stronger reductant that can engage in Birch reduction (Figure 4d). Notably, PC^•− is a persistent radical with sufficient lifetime for reversible CV and EPR analysis. This long lifetime allows PC^•− to absorb a photon using a practical LED setup as opposed to requiring laser irradiation.¹⁴ The excited-state properties of the PC^•−* were predicted using time-dependent DFT calculations to evaluate the thermodynamic feasibility of generating a solvated electron or directly reducing the substrate. We determined that the first six doublet excited states of PC^•−* can be accessed with a 405 nm photon (3.06 eV) used in this study (Figure S117). Excited states 2–6 have enough energy for ionization of PC^•−* by an electron transfer to the solvent (k_ET,1) to form a solvated electron and the ground-state PC. For efficient generation of a solvated electron, k_ET,1 must be competitive with the internal conversion process (k_IC), deactivating high-lying excited states to the lowest doublet excited state of PC^•−*, which is below the energy threshold for both solvated electron formation and direct PET to benzene. Solvated electron formation has been observed to occur in as little as 11 ± 1 ps in methanol, supporting the feasibility of this competition.³⁹ Further, electron transfer from the solvated electron to an aromatic substrate (k_ET,2) for Birch reduction reactivity must also be facile relative to the unproductive back electron transfer to the lowest excited state of PC^•−* (k_ET,3).

The photoexcitation of PC^•− was investigated using nanosecond transient absorption spectroscopy (Figure 4d). Selective excitation of PC^•− could be achieved through irradiation at 532 nm due to the minimal spectral overlap with PC or [PC–OH]⁻. Laser excitation produces a ground-state bleach feature (λ_min = 580 nm) that matches the absorption of PC^•− and thus can be assigned to PC^•−*. Kinetic monitoring of this signal reveals that the baseline is not recovered on the millisecond time scale, indicating that PC^•−* decomposes. Decomposition is consistent with either direct PET to the substrate or production of a solvated electron as both processes yield the neutral, closed-shell PC which lacks an absorption at λ = 580 nm. Since the photoionization event to release a solvated electron is likely orders of magnitude faster than the time resolution of our TA setup (vide supra), the absorption signal of the solvated electron would be expected to appear in the TA spectrum immediately following the laser pulse at t = 0 ns. However, we note that in THF the solvated electron has been observed at λ_max = 2120 nm, which is outside the range of our detector.⁴⁰ Our attempts at measurement in MeOH where this absorption would fall within the visible spectrum so far have been without success, likely due to the low solubility of the PC in MeOH and the resulting challenge of generating high-enough concentrations of PC^•− for spectroscopic study.

Interestingly, the observed excited-state PC^•−* is several orders of magnitude longer lived than the doublet excited states of other aromatic imides and diimides.²⁶ This observation suggests that the signal followed by TA can be tentatively assigned to the lowest quartet excited state ⁴PC^•−* produced via intersystem crossing (ISC) from the doublet manifold, as observed to occur upon excitation of other persistent organic radicals such as the enzyme DNA photolyase.^41,42 Comparing the kinetic traces at λ_probe = 580 nm with and without the benzene substrate present as a potential quencher does not reveal a significant change in the lifetime of ⁴PC^•−* (Figure S15), suggesting that ⁴PC^•−* does not react efficiently with the substrate. Considering that the computed lowest doublet state ²PC^•−* is also unlikely to reduce benzene (insufficient by ~1.0 V vs SCE, Figure S117), direct PET to the substrate appears to be thermodynamically unfavorable. One possibility to achieve increased driving force involves the occurrence of PET from a higher lying excited state in an anti-Kasha fashion, an unlikely prospect that is typically observed only in systems in which intramolecular PET is possible.⁴³ Although we do not see evidence by UV–vis for association of [PC–OH]⁻ or PC^•− with benzene (Figure S13), we are actively investigating the alternate hypothesis that PC^•− may form ground-state π-stacking complexes or exciplexes with benzene to enable intramolecular PET, potentially through an anti-Kasha process. These considerations taken together suggest the working hypothesis that PC^•−* photoionizes with k_ET,1 competitive with k_IC, releasing a solvated electron that subsequently reduces the substrate. Overall, the initial mechanistic experiments herein support formation of [PC–OH]⁻ and subsequent photodissociation to form PC^•−. Further experiments are currently ongoing to elucidate the role of ²PC^•−*, ⁴PC^•−*, and the competition between k_ET,1 and k_IC.

CONCLUSION

In summary, a class of organic benzo[ghi]perylene imide photoredox catalysts was developed for Birch reductions under mild benchtop conditions and visible-light LED irradiation. This work represents the first visible-light photoredox catalysis system that is capable of engaging arenes such as benzene that were previously out of reach due to their high triplet energies and extremely negative reduction potentials. Despite this unprecedented reactivity, this platform requires further development to realize its full synthetic potential. Initial mechanistic experiments support formation of the hydroxide adduct [PC–OH]⁻ and its subsequent photodissociation to PC^•−, which we posit undergoes absorption of a second photon to release a solvated electron as the active reductant. Future mechanistic studies are ongoing to test this hypothesis and to identify mechanistically guided PC design principles for more robust and active PCs to improve the scope of photocatalyzed Birch reductions.