A Catalytically Generated Transition Metal Analog of the Simmons–Smith Reagent

Abstract

(PDI)CoBr₂ complexes (PDI = pyridine–diimine) serve as precatalysts for reductive alkene cyclopropanations using CH₂Br₂ and Zn. A key question regarding the mechanism of this process is whether CH₂ transfer occurs through a Co(CH₂Br) species or through an isomeric carbene complex, similar to those formed in diazoalkane-based cyclopropanation reactions. Mechanistic studies are consistent with the intermediacy of a cationic (PDI)Co(CH₂Br) species, which is proposed to function as a transition metal analog of the Simmons–Smith reagent. This model rationalizes the critical role of ZnBr₂ Lewis acid in sequestering Br^– from the reaction. Finally, the cobalt-catalyzed cyclopropanation exhibits a distinct selectivity profile from the Simmons–Smith reaction due to differences in the steric properties of (PDI)Co and Zn carbenoids.

Introduction

Cyclopropanes are highly represented in biologically active compounds due to their unique three-dimensional structure and potential to engage in strain-induced ring-opening reactions.¹ The Simmons–Smith cyclopropanation is one of the most general methods for the addition of methylene (CH₂) to an alkene.² However, despite its long history in organic synthesis, there are limitations to the reaction that remain unaddressed and stem from the intrinsic properties of Zn carbenoids.³ For example, even moderately electron-deficient alkenes react slowly or not at all due to the high electrophilicity of XZn(CH₂X) species. Additionally, there are no general approaches to achieving regioselective or stereoselective cyclopropanations in the absence of directing groups.⁴ Finally, limited opportunities are available for the development of higher order (n + 1)- or (n + m + 1)-cycloadditions due to the concerted nature of CH₂ addition.⁵

Alternative carbenoid species may be accessible using low-valent transition metal complexes capable of activating 1,1-dihalomethanes by C–X oxidative addition. Following cyclopropanation, reduction of the oxidized metal complex would complete a catalytic cycle. This approach to catalytic CH₂ transfer obviates the need for CH₂N₂, which poses significant safety concerns, particularly on scale.⁶ In this vein, we previously conducted a survey of mid-to-late first-row transition metal complexes and found that (PDI)CoBr₂ (PDI = pyridine–diimine) derivatives function as precatalysts for alkene cyclopropanation using a CH₂Br₂/Zn reagent combination.⁷ (PDI)Co complexes are most well-known for their olefin polymerization activity⁸ but have also been shown to catalyze other transformations such as alkene hydrogenation,⁹ hydrosilylation,¹⁰ and hydroboration.¹¹

There have been extensive mechanistic investigations of diazoalkane-based catalytic carbene transfer reactions.¹² However, no equivalent studies have been carried out to probe the nature of transition metal carbenoids derived from 1,1-dihaloalkanes. Here, we present experimental and computational evidence supporting the intermediacy of a cationic Co(CH₂Br) complex in (PDI)Co catalyzed cyclopropanation reactions (Figure 1). This species functions as a transition metal analog of the Simmons–Smith reagent and undergoes kinetically facile CH₂ transfer by a concerted mechanism. Key to accessing this intermediate is the presence of ZnBr₂, which is generated as a reaction byproduct and serves as a Lewis acid to sequester halide. There is a striking similarity between this proposed carbenoid and the cationic Co(II)–alkyl complexes¹³ that are the propagating intermediates in alkene polymerization catalysis. Our mechanistic framework provides a rationale for the observed selectivity properties of the cobalt-catalyzed cyclopropanation, most notably its high sensitivity toward alkene substitution.

Figure 1. Cyclopropanating agents in organic synthesis.

Investigation of a transition metal carbenoid derived from CH₂Br₂.

Results and Discussion

Examining the Role of the Reductant.

From our initial reaction optimization studies, we noted that Zn powder afforded the highest cyclopropanation yields and could not be readily replaced with other reductants, including those possessing similar reduction potentials.¹⁴ Zn is commonly used as the terminal electron source in related cross-electrophile coupling reactions.¹⁵ In some cases, alternative reductants such as tetrakis(dimethylamino)ethylene (TDAE) can be substituted, suggesting that its only role is to reduce the transition metal catalyst.¹⁶ In other cases, Zn is uniquely effective, and there is evidence that organozinc species may be generated as catalytic intermediates.¹⁷

The selectivity properties of the cobalt-catalyzed cyclopropanation exclude the possibility that Zn(CH₂X)X species are functioning as the active CH₂ transfer agent. However, these observations did not rule out other possible roles for Zn in the catalytic cycle. Table 1 summarizes our survey of different reductants in the model cyclopropanation of 4-vinylcyclohexene (2). The highest yields of 3 (86%) were obtained using Zn (entry 1). Mn and (C₆H₆)₂Cr also provided 3 but in substantially diminished yields of 52% and 26% respectively (entries 2 and 3). Cp₂Co was examined as an outer-sphere reductant and found to be ineffective at promoting catalytic cyclopropanation despite being sufficiently reducing to convert (^i-PrPDI)CoBr₂ (1) to (^i-PrPDI)CoBr (5) (entry 4). Interestingly, Cp₂Co becomes a competent reductant when used in combination with ZnBr₂ (3.0 equiv), providing up to 49% yield of 3 (entry 6). LiBr has the inverse effect of suppressing catalytic cyclopropanation (entry 8). The opposing roles of ZnBr₂ and LiBr have also been documented in Negishi cross-couplings.¹⁸ In these reactions, ZnBr₂ serves as a Lewis acid and binds halide from the reaction mixture by forming ZnBr₃^–. On the other hand, LiBr is primarily a halide source and quenches the Lewis acidity of ZnBr₂.

Table 1.

Reductant Effects in the Cobalt Catalyzed Cyclopropanation^a

entry	reductant	additive	yield 3	rr 3:4
1	Zn	–	86%	>50:1
2	Mn	–	52%	>50:1
3	(C6H6)2Cr	–	26%	>50:1
4	Cp2Co	–	0%	–
5	Cp2Co	ZnBr2 (1.0 equiv)	22%	>50:1
6	Cp2Co	ZnBr2 (3.0 equiv)	49%	>50:1
7	Cp2Co	MnBr2 (3.0 equiv)	17%	>50:1
8	Zn	LiBr (2.0 equiv)	7%	>50:1

^aStandard reaction conditions: catalyst 1 (5 mol%); reductant (2.0 equiv); CH₂Br₂ (1.1 equiv); 24 h at room temperature in THF. Yields of 3 and regioisomeric ratios (rr) of 3:4 were determined by GC-FID analysis.

Cyclic voltammetry studies were carried out to assess the reduction potential of (^i-PrPDI)CoBr₂ (1) in the presence and absence of ZnBr₂. In accordance with previous observations by Araujo, Doherty, and Batista,¹⁹ complex 1 displays two sequential one-electron reductions at E_1/2 = –1.00 and –1.93 V vs. Cp₂Fe/Cp₂Fe⁺ (Figure 2). The first reduction event exhibits a large peak-to-peak separation, which is consistent with reduction of the (^i-PrPDI)CoBr₂ (1) complex (eq 1) followed by Br^– dissociation (eq 2) then reoxidation of (^i-PrPDI)CoBr in the return scan (eq 3). The observation that a reoxidation wave for the anionic [(^i-PrPDI)CoBr₂]^– is not observed, indicates that Br^– dissociation is fast on the CV timescale.

Figure 2.

Cyclic voltammetry data for 1 in the presence and absence of ZnBr₂. Initial [1] = 0.1 mM; 0.3 M n-Bu₄PF₆ in THF; scan rate: 100 mV/s; N₂ atmosphere. Scans were started at the open circuit potential and proceed in the direction indicated by the arrow.

When ZnBr₂ is titrated into the solution, the first reduction wave at –1.38 V decreases in intensity, and a new wave grows in at a more anodic potential of –0.99 V, indicating that ZnBr₂ binds to (^i-PrPDI)CoBr₂ (1) and facilitates its reduction. The second reduction to the formally Co(0) state is unaffected by the presence of ZnBr₂, indicating that there is no significant interaction between ZnBr₂ and the reduced (^i-PrPDI)CoBr species (5).

The interaction between ZnBr₂ and the (^i-PrPDI)CoBr₂ complex (1) was further probed by UV-vis spectroscopy (Figure 3). We previously reported the solid-state structure of the (^i-PrPDI)CoBr₂(ZnBr₂(thf)) adduct (6),⁷ which reveals that ZnBr₂ interacts with the cobalt complex through one of the Br ligands. ZnBr₂ binding causes an elongation of the Co–Br distance to 2.557(1) Å and a distortion of the local geometry at Co from trigonal bipyramidal to pseudo-square pyramidal (τ₅ = 0.36). The light tan-colored (^i-PrPDI)CoBr₂ complex (1) possesses a weak absorption feature by UV–vis spectroscopy at λ_max = 702 nm, previously assigned as a d–d transition.²⁰ Upon addition of ZnBr₂, this band blue-shifts to λ_max = 661 nm, and a color change to light green is observed. As a point of comparison, (^i-PrPDI)CoBr₂ (1) was also treated with TlPF₆ (1.0 equiv) to effect quantitative Br^– abstraction. The resulting cationic complex exhibits an absorption maximum that appears at lower energy (λ_max = 647 nm).

Figure 3.

(a) UV–vis spectra of (^i-PrPDI)CoBr₂ 1 in the presence and absence of ZnBr₂. The cationic [(^i-PrPDI)CoBr]PF₆ (7) was generated by treatment of 1 with TlPF₆ (1.0 equiv).

Mechanism of CR₂Br₂ Oxidative Addition and Stoichiometric Cyclopropanation.

Four possible mechanisms for CR₂Br₂ activation by the (^i-PrPDI)CoBr complex (5) are outlined in Figure 4a: (1) three-centered, two-electron oxidative addition, (2) S_N2 substitution by a cobalt nucleophile, (3) halogen atom abstraction followed by recombination, and (4) outersphere electron transfer to form a CR₂Br₂ radical anion.²¹

Figure 4.

(a) Potential mechanisms for CR₂Br₂ oxidative addition using (^i-PrPDI)CoBr complex 5. (b) UV–vis kinetics studies for the reaction of (^i-PrPDI)CoBr 5 (0.15 mM in THF) with CH₂Br₂ (100 equiv) at room temperature. ε values at 540 nm were fit to an exponential function: k = –0.10 h^–1; t_1/2 = 5.7 h. Experiment was repeated with CHMeBr₂ and CMe₂Br₂. Relative pseudo-first order rate constants are shown. (c) Radical clock experiment: attempted cyclopropanation of cyclooctene using 8, yielding the ring-opened dimerized product 9.

The first two possibilities were ruled out by examining the rate of oxidative addition as a function of increasing dibromoalkane substitution: CH₂ vs. CHMe vs. CMe₂.²² The reaction between (^i-PrPDI)CoBr complex 5 and CH₂Br₂ was monitored by UV–vis spectroscopy (Figure 4b). Under pseudo-first order conditions (100 equiv of CH₂Br₂), 5 undergoes oxidation to (^i-PrPDI)CoBr₂ (1) with a half-life of 5.7 h at room temperature. The isosbestic conversion of 5 to 1 indicates that the putative cobalt carbenoid intermediate is unstable and does not build up under these reaction conditions. When the same experiment was repeated using CHMeBr₂ and CMe₂Br₂, the relative reaction rate increased 3.7-fold and 6.8-fold, respectively. This trend is inconsistent with either the three-centered oxidative addition or the S_N2 substitution pathways, which would be decelerated by steric hindrance. Furthermore, the calculated activation barrier state for the S_N2 oxidative addition pathway was found to be prohibitively high in energy at 34.2 kcal/mol (see Supporting Information).

The potential for an outer-sphere electron transfer mechanism was evaluated by comparing the relative reduction potentials of (^i-PrPDI)CoBr (5) and CH₂Br₂. Complex 5 possesses an E_1/2 of –1.00 V vs. Cp₂Fe/Cp₂Fe⁺ in THF solution. By cyclic voltammetry CH₂Br₂ undergoes irreversible reduction with an onset potential of approximately –1.9 V. Thus, any outer-sphere electron transfer between these two species would be highly endergonic.

In previous C–C coupling reactions promoted by (PDI)CoAr complexes, Gambarotta and Budzelaar proposed that alkyl halide activation occurs through an initial one-electron halogen atom abstraction step.²³ (PDI)CoX complexes are electronically best described as Co(II) centers anti-ferromagnetically coupled to a PDI^•–. Thus, the halogen abstraction step primarily involves ligand-centered redox. In order to examine this pathway in the cyclopropanation, radical clock substrate 8 was prepared (Figure 4c). An attempted cyclopropanation of cyclooctene using 8 resulted in only 11% conversion of cyclooctene and no detectable yield of bicycle[6.1.0]nonane. Instead, the major product of the reaction is compound 9, which presumably arises from ring-opening of the bromocyclopropylcarbinyl radical followed by dimerization. This same product (9) is also formed in comparable yield when the reaction is carried out in the absence of cyclooctene.

DFT Models for Neutral and Cationic Co(II) Carbenoids Derived from CH₂Br₂.

Based on the observed stoichiometric reactivity, we reasoned that (^i-PrPDI)CoBr complex (5)²⁴ is the low-valent form of the catalyst that serves to activate CH₂Br₂ by a halogen-atom abstraction mechanism. Recombination of ^•CH₂Br radical with (^i-PrPDI)CoBr₂ would generate a Co(III) carbenoid, which could then be reduced by Zn to Co(II). Alternatively, if the ^•CH₂Br radical undergoes cage escape, it could recombine with another molecule of (^i-PrPDI)CoBr to directly form the Co(II) carbenoid. The optimized geometry for a putative (PDI)Co(CH₂Br)Br species (10) is shown in Figure 5a (BP86-D3BJ/6–311g(d,p) level of DFT). The two stereoisomers of 10, one with CH₂Br axial and the other with CH₂Br equatorial, were nearly isoenergetic (ΔG = 1.9 kcal/mol). The low-spin S = 1/2 state (pseudo-square pyramidal geometry at Co) was favored over the alternative S = 3/2 state by 18.6 kcal/mol.

Figure 5.

Optimized structures for (a) neutral (PDI)Co(CH₂Br)Br complex (10) and (b) cationic [(PDI)Co(CH₂Br)]⁺ species (11) (BP86-D3BJ/6–311g(d,p) level of DFT). i-Pr substituents on the ligand were truncated to Me. (c) Truncated molecular orbital diagram for 11 showing Co 3d orbitals, redox-active PDI π-orbitals, and the Co–C(σ) and C–Br(σ*) orbitals relevant to CH₂ transfer.

We next examined the cationic carbenoid (11) that would be generated by a ZnBr₂-assisted Br^– abstraction from the neutral carbenoid (10). Indeed, cationic Co(II)–alkyl complexes are the proposed propagating species in olefin polymerization reactions catalyzed by (PDI)CoBr₂ complexes.¹³ In polymerization catalysis, Lewis acidic activators such as MAO are used to abstract halide from the precatalyst.^8a

The low-spin cationic [(PDI)Co(CH₂Br)]⁺ complex (11) was calculated and found to adopt the expected square planar geometry as was observed for the analogous [(PDI)Co(CH₃)]⁺ complex described by Chirik.^13d Interestingly, the Co–C–Br angle is acute (82.9°), and the Co–Br distance is relatively short (2.59 Å). The structure of 11 may be viewed as being intermediate between a Co–CH₂Br species and a Co=CH₂ bearing an axial Br. A truncated MO diagram for 11 is shown in Figure 5c. As expected for low-spin Co(II), dz² is singly occupied, and the spin density is localized on Co. The redox-active PDI π-orbitals are unoccupied, and the ligand bond metrics are consistent with a neutral charge state. A concerted mechanism for CH₂ transfer from this species would require two orbital interactions: donation from the Co–Br σ-bond into the alkene π*, and donation from the alkene π-orbital into the C–Br σ*. These two orbitals in complex 11 could be clearly identified as HOMO–8 (125) and LUMO+6 (141), respectively.

DFT Models for Cyclopropanation Pathways from Co(II) Carbenoids.

Initial calculations suggest that cyclopropanation may occur through a Co(CH₂Br) species, akin to the Simmons–Smith reaction. Accordingly, we next sought to examine cyclopropanation pathways from 10. We excluded stepwise mechanisms involving long-lived ionic or radical intermediates from consideration because all of the cyclopropanations that we have examined proceed stereospecifically. Additionally, radical clock experiments using vinylcyclopropane did not show any evidence of cyclopropane ring-opening.⁷

A transition state for concerted CH₂ transfer from neutral carbenoid 10 to propylene was identified (Figure 6). The optimized structure resembles previously calculated butterfly transition states for the Simmons–Smith reaction, with CH₂ addition occurring simultaneously with Br migration from C1 to Co.²⁵ The activation barrier for this process was found to be sufficiently high in energy (30.6 kcal/mol) to be inconsistent with a catalytic process that occurs at room temperature. ZnBr₂ coordination to one of the axial Br ligands results in only a modest decrease in the activation barrier to 29.4 kcal/mol. An analogous transition state was located for cationic carbenoid 11, which has an activation barrier of only 6.9 kcal/mol. Thus, halide abstraction from the carbenoid species has the net result of facilitating carbene transfer. One explanation for this effect is that ligands placed along the z-axis of the Co complex are expected to be destabilizing, and Co must increase its coordination number in the cyclopropanation transition state due to Br migration. Therefore, the cyclopropanation barrier decreases when starting from the lower-coordinate square planar carbenoid.

Figure 6.

Calculated CH₂ transfer and carbometallation transition states using neutral and cationic Co(CH₂Br) species (BP86-D3BJ/6–311g(d,p) level of DFT). Energies are of the transition state relative to those for the carbenoid + propylene. All transition structures were optimized in the S = 1/2 spin state.

The two C–C bonds of the cyclopropane are formed in a concerted but highly asynchronous fashion. The C1–C2 distance is significantly shorter than the C1–C3 distance. According to NBO population analysis, C3 experiences an increase in positive charge character upon proceeding from the alkene ground state to the cyclopropanation transition state (+0.10), consistent with the small negative ρ-value observed experimentally in Hammett studies.

An alternative carbometallation pathway was also examined but found to be less favorable than the CH₂ transfer pathway by 7.7 kcal/mol. In the carbometallation transition state, the alkene adds across the Co–C1 bond with concomitant migration of Br from C1 to Co. As a point of comparison, Li(CH₂X) cyclopropanation reactions are proposed to access stepwise carbometallation/cyclization mechanisms as opposed to the concerted CH₂ transfer mechanisms favored by Zn carbenoids.^{25b, 26}

Finally, we tested whether this calculated transition state model could reproduce experimentally observed trends in alkene reactivity (Figure 7). 1,3-Dienes and monosubstituted alkenes are the most reactive substrates in the cobalt-catalyzed cyclopropanation. Accordingly, calculated barriers for 1,3-pentadiene and 1-butene were relatively low (3.9 and 6.1 kcal/mol). On the other hand, activation barriers for (E)-alkenes, which are generally unreactive in the cyclopropanation, were significantly higher in the DFT model (7.8 kcal/mol).

Figure 7.

Calculated activation barriers for the cyclopropanation of different alkenes using cationic carbenoid complex 11 (BP86-D3BJ/6–311g(d,p) level of DFT). ΔΔG^‡ values correspond to the difference in energy (kcal/mol) between the cyclopropanation transition state and the sum of the energies for 11 and the alkene.

For 4-vinylcyclohexene (2), cyclopropanation of the exocyclic alkene is calculated to be 0.16 kcal/mol more favorable than cyclopropanation of the ring alkene, consistent with the regioisomer observed experimentally (Figure 8). The lower selectivity in the DFT model as compared to that observed experimentally may be due to the truncated ligand used in the calculations. The most significant difference in the geometries of these two transition states is the orientation of the incoming alkene relative to the Co(CH₂Br) fragment. For the less-hindered terminal alkene, the alkene approaches parallel to the Co–CH₂ bond axis. The degree of twisting can be approximated by a Co–C1–C2–C3 dihedral angle of 137°, with 180° being perfectly parallel. This geometry is similar to calculated trajectories for Zn carbenoid additions.^25b By contrast, the ring alkene of 4-vinylcyclohexene is sufficiently hindered that it cannot approach in this orientation due to steric interactions with the ortho-substituents of the catalyst aryl groups. Instead, the alkene approaches nearly perpendicularly, and the Co–C1–C2–C3 dihedral angle is 110°.

Figure 8.

DFT calculations for the regioselective monocyclopropanation of 4-vinylcyclohexene (2) using a cationic carbenoid (BP86-D3BJ/6–311g(d,p) level of DFT). The catalyst was modeled using 2,6-MeC₆H₃- substituents, and the counteranion was excluded. All structures were optimized in the S = 1/2 spin state.

The preference for a parallel vs. perpendicular approach of the alkene can be rationalized on the basis of frontier orbital interactions (Figure 9). Additions of M(CH₂Br) species to alkenes involve two complementary interactions: (1) donation of the alkene π-orbital into the C–Br σ* and (2) donation from the electron-rich M–C σ-bond into the alkene π*. The former would be insensitive to the orientation of the alkene. However, the latter interaction requires parallel alignment of the Co–C σ-bond and the alkene π*-orbital.

Figure 9.

Orbital interactions in M(CH₂Br) additions to alkenes.

Conclusions

Collectively, these studies indicate that transition metals are capable of accessing carbenoid intermediates analogous to the Simmons–Smith reagent through the activation of 1,1-dihaloalkanes. An attractive feature of the (PDI)Co catalyzed process is the ability to influence the selectivity of cyclopropanation by modifying the structure of the PDI ligand. By contrast, Zn carbenoids are significantly less tunable, because their reactivity is attenuated by the presence of strongly coordinating neutral ligands.²⁷ In a broader context, these studies show that 1,1-dihaloalkanes are capable of delivering reactive carbene equivalents to transition metals, presenting a viable alternative to reactions that currently rely on diazoalkane reagents. Ongoing efforts in our lab are directed at investigating the accessibility and reactivity of M(CH₂X) and M(CH₂)X species using other transition metal catalysts.