Resolving Conflicting Mechanisms for Photoredox Allylic sp³-CH Arylation Using Deuterium-Labeling and Isotope Effects

Abstract

Two conflicting mechanisms have emerged for the direct arylation of allylic C−H bonds enabled by the combined use of thiol and photoredox catalysis. In the original report (Nature, 2015, 519, 74–77), a radical coupling step—between a radical anion of an arene and an allylic radical—is proposed to be the key C−C bond-forming step. A recent mechanistic study (J. Org. Chem. 2022, 87, 223–230) has suggested that the C−C bond formation occurs via radical anion capture by the olefin followed by an H atom transfer (HAT) event to deliver the allylic C−H arylation product. Utilizing cyclohexene-4,4,5,5-d₄ as a mechanistic probe to distinguish between otherwise indistinguishable regioisomeric allylic C−H arylation products in the reaction of cyclohexene and dicyanobenzene, we establish that the radical anion capture−HAT mechanism is not operative. Furthermore, experimental k_H/k_D studies and DFT calculations lend strong support to the radical coupling mechanism proceeding via irreversible HAT to form the allylic radical of cyclohexene, followed by regioselectivity-determining radical coupling (for unsymmetrical olefins) and facile decyanation.

Graphical Abstract

The direct arylation of allylic C−H bonds is an important transformation that has evolved over the years.^1–3 While strategies that utilize transition metal catalysis dominated early efforts,^4–6 recent examples have focused on the creative use of photoredox and metallophotoredox catalysis under much milder reaction conditions by exploiting the unique reactivity of radical intermediates.^7–11 In this context, the direct arylation of allylic C−H bonds using a combination of organic and photoredox catalysis was first reported by Macmillan in 2015.⁹ In this seminal report, the allylic arylation reaction of cyclohexene (1) and 1,4-dicyanobenzene (DCB, 2) is catalyzed by an iridium photocatalyst, a thiol organocatalyst, and an inorganic base with the reaction being performed in acetone under visible light irradiation (Scheme 1). According to Macmillan, the reaction is initiated by hydrogen atom transfer (HAT) from cyclohexene to a thiyl radical, resulting in the formation of allylic radical 4. Concurrently, the iridium photocatalyst reduces DCB to form the DCB radical anion (2^•−). Radical−radical coupling and elimination of the cyanide ion deliver the CH-arylated product 3 (Scheme 1, Mechanism A).

Scheme 1.

Two Distinct Mechanistic Pathways Proposed for the Allylic sp³ C−H Arylation of Cyclohexene Enabled by Organic and Photoredox Catalysis^a

^a This work evaluates the mechanism of this reaction using deuterium labeling studies, isotope effects, and DFT calculations.

An alternative mechanism was recently proposed by Swierk and Fredin¹² based on transient absorption spectroscopy (TAS), kinetic modeling, and density functional theory (DFT) analysis. In this study, the slower rate of quenching of the thiyl radical relative to 2^•− was interpreted as evidence for an alternative mechanism involving the capture of 2^•− by 1 followed by decyanation to form α-arylated cyclohexyl radical 5. The intermediate 5 then undergoes a “cooperative HAT” event to form the CH-arylated product 3 (Scheme 1, Mechanism B). DFT analysis by the authors suggests that this alternative pathway has barriers of 35.7 kcal/mol for radical anion capture and 44.3 kcal/mol for the decyanation step, while the HAT step that reforms the alkene was found to be barrierless. In the same study, the authors found that Mechanism A proceeds via the rate-limiting generation of 4, with a barrier of only 14.5 kcal/mol, followed by radical coupling and rearomatization steps. Despite the lower energetic barriers for Mechanism A, the authors favor Mechanism B for the reaction of 1 and 2; the discord between theory and the interpretation of their experiment is attributed to the underestimation of the actual barrier for the “highly orientation-dependent HAT step” (in Mechanism A) by DFT calculations.

The key difference between the two mechanisms is the involvement of allylic radical 4 in the C−C bond-forming step. When unsymmetric alkenes such as 6a or 6b are used as substrates in the reaction, Macmillan⁹ observes regioisomeric products (Scheme 2, reproduced in this work), an observation that is easily rationalized by invoking a radical coupling step between 2^•− and either end of the unsymmetric allylic radical (Scheme 2, 7a/7b). The preferred site of arylation of 6a and 6b is indicated in red (Scheme 2), and the resulting regioselectivity can be correlated to the relative stability of the allylic radical involved in the radical coupling event. It must also be noted that Mechanism B cannot account for the major product in the reaction of 6a or the minor product in the reaction of 6b. Swierk, Fredin, and co-workers¹² explain this deficiency of Mechanism B by stating that “multiple reaction mechanisms may be simultaneously accessible” for alkenes such as 6a or 6b.

Scheme 2.

Experimental Regioselectivity in the Allylic Arylation of Two Representative Unsymmetric Alkenes^a

^a Reported yields and regioselectivities are average values from reactions performed in duplicate. Mechanism A accounts for the formation of both products via the allylic radicals highlighted in boxes. Mechanism B cannot account for the formation of 8a and 8b′.

We report herein key physical organic experiments^13–18 that address these conflicting mechanistic proposals for the allylic arylation reaction of dicyanobenzene and cyclohexene. (1) Deuterium labeling studies are utilized to establish the intermediacy of an allylic radical species in this reaction. (2) Competitive isotope effects are measured to probe the first irreversible step in the catalytic cycle for cyclohexene. (3) The free energy surface is probed by implementing DFT and Marcus theory calculations. (4) Finally, the origin of regioselectivity for unsymmetrical alkenes is measured experimentally and calculated using DFT. Together, these studies provide compelling evidence for the Macmillan pathway (Scheme 1, Mechanism A) as the sole operative mechanism of this reaction.

DEUTERIUM LABELING STUDIES

In order to experimentally distinguish between competing reaction pathways, we devised a simple isotope labeling experiment (Scheme 3) using commercially available cyclohexene-4,4,5,5-d₄ (1-d₄). If the radical coupling pathway (Mechanism A) is operational, then an allylic radical will be generated adjacent to one of the deuterated carbon atoms (4-d₄). The “identical” resonance structure of this allylic radical places the radical center one carbon atom away from a deuterated carbon atom (4′-d₄). Based on a computed equilibrium isotope effect (EIE) of 1.1 for the methylene group adjacent to the allylic radical,^19–22 we estimate that 4′-d₄ is the more contributing resonance structure; that is, the allylic radical will have slightly more radical character at the carbon adjacent to a CH₂ rather than a CD₂. Therefore, radical coupling of 2^•− with the d₄-allylic radical followed by decyanation (Mechanism A) should theoretically result in a 48:52 ratio of two distinct regioisomeric products — 3-d₄ and 3′-d₄ —distinguishable based on the location of the deuterium labels relative to the olefinic carbon atoms (Scheme 3). On the other hand, if Mechanism B is operational, 2^•− will add directly to the olefin followed by decyanation and then HAT to yield a single product, 3′-d₄ where the olefin always forms adjacent to a deuterated carbon atom (Scheme 3). Finally, if both pathways are simultaneously operational, the ratio of regioisomeric products will be perturbed in favor of regioisomer 3′-d₄ which is common to both mechanisms (depending on the percent contribution of each mechanism).

Scheme 3.

Design of Isotope-Labeling Experiment to Distinguish between the Two Conflicting Mechanisms for the Photoredox Allylic sp³ C−H Arylation

We carried out the reaction of 1-d₄ with 2 in triplicate under the standard reaction conditions (Figure 1) and isolated the allylic arylation product via flash chromatography. A combination of ¹H, ²H, and ¹³C NMR analysis was used to confirm the formation of both 3-d₄ and 3′-d₄ in a 47:53 ratio in all three experiments (Figure 1). In Figure 1, the regions between δ 1.35 and δ 2.35 ppm in the ²H and ¹H NMR spectra of the purified product are stacked in order to show the formation of both regioisomeric products 3-d₄ and 3′-d₄. All peak assignments were confirmed by rigorous 1D and 2D NMR experiments. See experimental methods, Supporting Information, for a detailed explanation of NMR peak assignments. The signals for the allylic methylene of 3-d₄ (labeled c) and the methylene adjacent to the chiral center in 3′-d₄ (labeled a′) in the ¹H NMR provide an easy handle for the determination of the product ratio by simple integration of the well-separated peaks. The ²H and ¹³C NMR further confirm that both products are formed in an ~1:1 ratio. The observation of product 3-d₄ excludes the possibility that Mechanism B is the sole operative mechanism of this reaction (see Scheme 3). Moreover, the excellent agreement between the experimental product ratio and the predicted product ratio for Mechanism A (Scheme 3) lends strong support to Macmillan’s originally proposed mechanism as the sole operative pathway in this reaction.²³

Figure 1.

Sections of ²H, ¹H, and ¹³C NMR used to identify and quantify the two regioisomeric products 3-d₄ and 3′-d₄ observed in this deuterium labeling experiment performed using cyclohexene-4,4,5,5-d₄ and 1,4-dicyanobenzene.

EXPERIMENTAL KIE STUDIES

We performed a competitive k_H/k_D experiment (in triplicate) in the reaction of 1 and 2 using a 50:50 mixture of 1/1-d₁₀. In each of these experiments, aliquots from the reaction mixture were drawn at 2 min intervals, and the competitive KIEs were determined at 10 distinct time points. The % deuterated product in the reaction mixture and the % conversion of the unlabeled material were determined at each time point using GCMS (calibrated), and the KIE was calculated from the Saunders equation factoring in the percent conversion at each time point (see Supporting Information for complete details of these experiments including GCMS calibration curves and individual k_H/k_D measurements). The competitive k_H/k_D from the three independent experiments were 10.5 ± 0.5, 9.6 ± 1.2, and 9.9 ± 1.1 (Figure 2). This large normal competitive k_H/k_D value is qualitatively consistent with a hydrogen transfer as the first irreversible step in the catalytic cycle for cyclohexene.¹⁸ The next step in our study is to obtain a quantitative interpretation of this competitive k_H/k_D measurement by modeling the key steps involving 1 in both proposed mechanistic pathways and predicting the KIEs from the scaled vibrational frequencies of relevant transition structures for the first irreversible step.

Figure 2.

Competitive k_H/k_D experiments that report on the first irreversible step in the catalytic cycle for 1.

DFT EVALUATION OF COMPETING MECHANISMS

For the quantitative interpretation of the competitive k_H/k_D, we modeled the key transition states involving 1 in both mechanistic pathways discussed in Scheme 1 using ωB97XD/aug-cc-pVTZ//B97D/6-31+G*^24–28 PCM^29,30 (acetone). The reported Gibbs free energies are corrected for the experimental concentration of the relevant species in solution using Paton’s Goodvibes code.³¹ The predicted k_H/k_D values for each step in both mechanisms were obtained from the scaled vibrational frequencies²¹ of the transition structures using the program ISOEFF98,²² and a Wigner tunneling correction was applied.^20,32 If Mechanism A is operational, our experimental k_H/k_D experiments suggest that the initial HAT to generate the allylic radical has a barrier higher than those of the subsequent radical coupling and decyanation steps. Consistent with this qualitative analysis of the free energy profile, the three key TSs in this pathway—TS_HAT-A, TS_CC-A, and TS_Elim-A —were found to have barriers of 15.2, 12.1, and −0.1 kcal/mol, respectively (relative to 2^•− + 1 + thiyl radical RS^• as the arbitrary reference, Figure 3). For the quantitative interpretation of our competitive k_H/k_D experiment, we predicted k_H/k_D for each of the three TSs shown in Figure 3.^20–22 The predicted k_H/k_D for TS_HAT-A is a combination of the primary KIE for the hydrogen being abstracted and the secondary KIE for the remaining hydrogen atoms of 1. For TS_CC-A and TS_Elim-A , the predicted k_H/k_D represents the secondary KIE for the remaining nine hydrogen atoms of 1. The predicted k_H/k_D for TS_HAT-A of 8.2 is in relatively good agreement with the average experimental k_H/k_D of 10.0 ± 1.0 (average of all three measurements reported in Figure 2), providing quantitative support for Mechanism A as the likely pathway in the allylic arylation reaction of cyclohexene. The slight underprediction of k_H/k_D could be due to the choice of the DFT method or an underestimation of the contribution of tunneling to the predicted k_H/k_D. Accordingly, we explored TS_HAT-A using 14 different DFT methods and obtained a range of predicted k_H/k_D values of 8.1−9.3 (applying a Wigner tunneling correction). Additionally, implementation of a tunneling correction using an alternative model (deformed TST)³² for each of these 14 methods led to predicted k_H/k_D values ranging from 8.4 to 10.1, values that are in good to excellent agreement with the average experimental k_H/k_D of 10.0 ± 1.0.

Figure 3.

Reaction coordinate diagram representing the steps in Mechanism A that involve 1 computed using ωB97XD/aug-cc-pVTZ//B97D/6-31+G* (acetone). All energies are in kcal/mol, and distances are in angstroms. A comparison of the average competitive experimental k_H/k_D and predicted k_H/k_D for the key transition structures in this pathway. Triisopropylsilanethiol was used as the thiol (RSH).

The best possible interpretation of a competitive k_H/k_D of 10.0 ± 1.0 consistent with Mechanism B is that the final step involving HAT from radical intermediate 5 to deliver product 3 is the first irreversible step in the catalytic cycle involving 1. Our calculations of Mechanism B reveal that the radical anion capture (Figure 4, TS_CC-B) and subsequent decyanation (Figure 4, TS_Elim-B) steps have barriers of 46.4 and 50.7 kcal/mol, respectively (relative to 2^•− + 1 + thiyl radical RS^• as the arbitrary reference). Finally, in accordance with the Swierk−Fredin study, the HAT from 5 to the thiyl radical was found to be barrierless (relaxed potential energy scans confirm the absence of a saddle point corresponding to HAT from 5 to yield 3). Therefore, for Mechanism B, the decyanation step to form 5 is the first irreversible step in the catalytic cycle for 1. The predicted k_H/k_D for this step is 1.1 (a secondary k_H/k_D), which is inconsistent with the experimentally measured k_H/k_D of 10.0 ± 1.0. This mismatch of experiment and theory, along with the prohibitively high barriers (for TS_CC-B and TS_Elim-B), demonstrates that Mechanism B is unlikely to be an accessible pathway in this reaction.

Figure 4.

Reaction coordinate diagram representing the steps that involve 1 in Mechanism B computed using ωB97XD/aug-cc-pVTZ//B97D/6-31+G* (acetone). A comparison of the average competitive experimental k_H/k_D and predicted k_H/k_D for the key transition structures in this pathway. Triisopropylsilanethiol was used as the thiol (RSH).

POSSIBLE EXPLANATION OF TAS RESULTS

¹² The key observation in the Swierk−Fredin study is that the thiyl radical (RS^• that generates 4 via TS_HAT-A) was quenched at a slower rate relative to 2^•−. This led the authors to conclude that the key carbon−carbon bond-forming step occurs via capture of 2^•− by 1 (Scheme 1, TS_CC-B) and not via radical coupling with 4 (Mechanism B). Since our studies exclude Mechanism B as an operative mechanism for this reaction, we sought to offer an alternative explanation for this observation that is consistent with the radical−radical coupling mechanism (Mechanism A). Following H-atom transfer from cyclohexene to the thiyl radical to form allylic radical 4 (TS_HAT-A) and thiol (RSH), the thiyl radical (RS^•) must be regenerated for the next catalytic cycle via a deprotonation−SET sequence. Using K₂CO₃ as the base, we estimated the barrier for deprotonation to be 6.7 kcal/mol. Oxidation of the resulting thiolate anion to the thiyl radical by SET to Ir(IV) is expected to be very facile.^9,33 The low barriers for thiol deprotonation and subsequent oxidation (RSH → RS⁻ → RS^•) relative to the calculated barriers for radical coupling between 4 and 2^•− (TS_CC-A) suggest that the thiyl radical is regenerated on a much faster time scale than the consumption of 2^•−. Therefore, it is likely that the apparent slower rate of quenching of the thiyl radical (monitored at 480 nm) relative to 2^•− (concurrently monitored at 720 nm in the same experiment) can be attributed to the average concentration of the thiyl radical appearing to remain “constant” (due to rapid regeneration) during the time scale of the radical coupling event.³⁴ The compelling evidence obtained in this work supporting Mechanism A warrants a re-evaluation of the TAS results.

PREDICTION OF REGIOSELECTIVITY WITH UNSYMMETRICAL OLEFINS

As a final step in our mechanistic study, we addressed the question of regioselectivity in the allylic arylation of three unsymmetrical alkenes, such as 6a−c, that were originally reported in the Macmillan study.⁹ The reactions of 6a−c with 2 were performed in duplicate to confirm the reported yields and regioselectivities. According to the reaction coordinate for Mechanism A (Figure 3), the radical−radical coupling step should be the regioselectivity-determining step in these reactions. Accordingly, we modeled the radical coupling step (analogous to TS_CC-A) for three unsymmetric alkenes—1-hexene (6a), 2-methyl-2-hexene (6b), and 1-(tert-butyl)-cyclohex-1-ene (6c)—and found the predicted ΔΔG^‡ values for the regioisomeric radical coupling TSs to be in good agreement (and favoring the correct regioisomer) with experimental ΔΔG^‡ values that correspond to the observed regioselectivity of these reactions (Scheme 4). This agreement of experimental and theoretical regioselectivities provides additional support for Mechanism A as the operative pathway in these reactions. It must also be noted that Mechanism B cannot account for the major product in the reaction of 6a/6c or the minor product in the reaction of 6b since these regioisomers can only be formed via the intermediacy of an allylic radical.

Scheme 4.

Comparison of Experimental Regioselectivity in Reactions of Unsymmetric Alkenes to Predicted Regioselectivity Based on the Relative Energy of Regioisomeric Radical Coupling Transition States (Mechanism A)^a

^a The position undergoing arylation in the major regioisomer in each case is indicated in green, while the position undergoing arylation for the minor regioisomer is indicated in red.

In conclusion, we have evaluated the mechanism of the allylic arylation reaction of cyclohexene enabled by dual photoredox and thiol catalysis. Our results underline the inherent complexity of understanding parallel catalytic cycles that deliver a single product. In this case study, a series of physical organic tools, including isotope labeling, isotope effects, DFT and Marcus theory calculations, and product analysis, is shown to be a valuable complement to photophysical methods that are currently used to evaluate mechanisms in this important area of catalysis.