Dinickel Catalyzed Vinylidene–Alkene Cyclization Reactions

Abstract

A dinickel catalyst promotes reductive cyclization reactions of 1,1-dichloroalkenes containing pendant olefins. The reactions can be conducted with a Zn reductant or electrocatalytically using a carbon working electrode. Mechanistic studies are consistent with the intermediacy of a Ni₂(vinylidene) species, which adds to the alkene and generates a metallacyclic intermediate. β-Hydride elimination followed by C–H reductive elimination forms the cyclization product. The proposed dinickel metallacycle is structurally characterized and its stoichiometric conversion to product is demonstrated. Spin polarized, unrestricted DFT calculations are used to further examine the cyclization mechanism. These computational models reveal that both nickel centers function cooperatively to mediate the key oxidative addition, migratory insertion, β-hydride elimination, and reductive elimination steps.

Graphical Abstract

INTRODUCTION

Catalytic 1,6-enyne cycloisomerizations are valuable transformations that yield cyclic unsaturated building blocks from acyclic precursors.¹ A notable feature of these reactions is the diversity of products that can be generated from a single starting material (Figure 1a). There are various mechanisms by which transition metal catalysts can engage enyne substrates: oxidative ene–yne coupling to form a metallacycle, attack of an alkene on an electrophilic M(alkyne) species, and addition of a M–H to an alkyne followed by migratory insertion of an alkene. A majority of these processes favor the formation of five-membered ring exocyclization products, with the position of substituents dictated by the specific catalyst/substrate combination being used.

Figure 1.

(a) Exo and endo cyclization modes of 1,6-enynes. (b) Reductive activation of 1,1-dichloroalkenes enable selective vinylidene–alkene cyclization reactions.

There is a less common type of enyne cycloisomerization that generates endocyclization products.² For example, Wilkinson’s catalyst converts 1,6-enynes into conjugated methylenecyclohexenes.³ On the basis of isotopic labelling experiments, Lee proposed that the reaction is initiated by the isomerization of a Rh(alkyne) complex to a Rh(vinylidene).^4,5 It stands to reason that this pathway is disfavored for most other catalysts, because alkynes are approximately 43 kcal/mol more stable than their corresponding vinylidene isomers.⁶ Consequently, only catalysts that form stable M=C multiple bonds and weak M(π-C≡C) interactions can carry out this step.⁷

If generation of the key M=C=CR₂ intermediate did not rely on an isomerization of an alkyne, it may be possible to develop a more general strategy for catalytic vinylidene transfer. In this context, it was striking to us that the hypothetical reaction of 1,1-dichloroethylene⁸ and Zn to form vinylidene⁹ and ZnCl₂ is near thermoneutral. Thus, reactive M=C=CR₂ species should be thermodynamically accessible from these starting materials. Based on this hypothesis, we discovered that dinickel catalysts could successfully promote intermolecular [2 + 1]-cycloadditions of vinylidenes generated reductively from various substituted 1,1-dichloroalkenes.¹⁰ Here, we present the intramolecular variant of this process, which generates the product of a formal vinylidene insertion into a C(sp²)–H bond (Figure 1b). This approach to vinylidene generation addresses some of the previous limitations of enyne endo-cycloisomerizations. For example, internal alkenes are tolerated, and five-membered ring formations are demonstrated. A dinickel metallacycle derived from the addition of a vinylidene to an alkene is structurally characterized, and its role in the mechanism of catalysis is discussed. DFT calculations reveal that catalysis occurs through a series of intermediates and transition states where both Ni centers cooperatively direct bonding changes.

RESULTS AND DISCUSSION

Catalytic Intramolecular Vinylidene Additions to Alkenes.

Following reaction optimization studies, we identified a set of conditions that induces the reductive cyclization of model substrate 1 to form methylenecyclohexene 2 in 92% yield (Table 1, entry 1). The active (^i-PrNDI)Ni₂Cl (3) catalyst was generated in situ by stirring ^i-PrNDI (4) (5 mol%) and Ni(dme)Cl₂ (10 mol%) over Zn powder (3.0 equiv) for 20 min prior to the addition of 1. Formation of 3 is indicated by the appearance of a deep violet colored reaction solution.¹¹ Premetallated complexes, such as (^i-PrNDI)Ni₂Cl₂ (5) and (^i-PrNDI)Ni₂(C₆H₆) (6), provided comparable yields of product 2, 91% and 90%, respectively (entries 4 and 5). The NDI variant containing ortho c-Pent substituents (7) was equally effective in the reaction (entry 6).

Table 1.

Effects of Reaction Parameters on the Reductive Vinylidene Cyclization.^a

entry	deviation from standard conditionsa	yield (2)
1	none	92%
2	no 4	0%
3	no 4 and no Ni(dme)Cl2	0%
4	(i-PrNDI)Ni2Cl2 (5) instead of 4/Ni(dme)Cl2	91%
5	(i-PrNDI)Ni2(C6H6) (6) instead of 4/Ni(dme)Cl2	90%
6	c-PentNDI (7) instead of 4	90%
7	0.05 M instead of 0.025 M	81%
8	Mn instead of Zn	88%
9	Ligands 8–10 (5 mol%) instead of 4b	<10%
10	11 or 12 (5 mol%) instead of 4/Ni(dme)Cl2	<1%

^aStandard reaction conditions: 1 (0.10 mmol, 1.0 equiv), ^i-PrNDI (5 mol%), Ni(dme)Cl₂ (10 mol%), Zn (0.3 mmol, 3.0 equiv), NMP (0.8 mL), THF (3.2 mL), 24 h, rt. All yields were determined by ¹H NMR using mesitylene (1.0 equiv) as an internal standard.

^b5 mol% of Ni(dme)Cl₂.

A low initial concentration of substrate 1 was beneficial for yield, presumably due to the suppression of competing bimolecular processes. For example, a two-fold increase in concentration to 0.05 M resulted in an 11% decrease in yield (entry 7). In our survey of reductants, Mn provided a similar yield to that obtained with Zn (entry 8).

As a point of comparison, we also tested mononickel catalysts containing various combinations of pyridine and imine donors. In all cases, only trace quantities of 2 were produced (<10% yield, entry 9). Likewise, other Ni(I) dimers 11¹² and 12¹³ also proved to be ineffective as catalysts (entry 10). Collectively, these results suggest that the constrained dinuclear active site of the (NDI)Ni₂ system is a requirement for efficient catalysis.

Substrate Scope Studies.

With optimized reaction conditions in hand, we next examined how structural changes to the substrate influence the yield of cyclization (Figure 2). A variety of synthetic routes were used to prepare the requisite 1,1-dichloroalkene substrates. For example, diastereoselective conjugate allylation using an oxazolidinone chiral auxiliary yielded 13.¹⁴ Conversion of 13 to aldehyde 14 was carried out with DIBAL-H, and one-carbon homologation using CCl₄ (2.0 equiv) and Ph₃P (4.0 equiv) yielded 1,1-dichloroalkene 15. The catalytic reductive cyclization was carried out under standard conditions and provided 16 in 81% yield.

Figure 2.

Substrate scope studies. Isolated yields were determined following purification and were averaged over two runs. Standard reaction conditions: substrate (0.2 mmol), ^i-PrNDI (4) (5–10 mol%), Ni(dme)Cl₂ (10–20 mol%), Zn (3.0 equiv), NMP (1.6 mL), THF (6.4 mL), rt or 50 °C, 24 h. ^a substrate (0.2 mmol), ^c-PentNDI (7) (10 mol%), Ni(dme)Cl₂ (20 mol%), Zn (3.0 equiv), NMP (0.8 mL), THF (7.2 mL), rt, 24 h.

Variation of the substrate tether length grants access to five-membered ring products, requiring only slight modifications to the reaction conditions: increased catalyst loading (10 mol%) and heating at 50 °C. In addition to forming methylenecyclohexenes and methylenecyclopentenes, this method is also amendable to the formation of unsaturated heterocycles (product 18). However, an attempted cyclization to form a seven-membered ring did not yield any product. Common functional groups are tolerated, including protected alcohols, acetals, and esters. Steric hindrance adjacent to the 1,1-dichloroalkene is well-tolerated (products 26, 28, and 30). However, substituting the position next to the terminal alkene results in no yield of cyclized product.

Upon further evaluation of the reaction conditions, we found that a limited scope of products containing substituted exocyclic alkenes are accessible when the c-Pent substituted NDI ligand 7 was used in the place of the i-Pr ligand 4. For example, ester substituted methylenecyclopentene 32 is formed in 63% yield (13:1 Z/E mixture) from 31-E. On the other hand, a substrate containing a Ph substituent on the alkene proved to be unreactive (see Supporting Information for a summary of the reaction limitations).

Evidence in Support of a Stepwise Ni₂(vinylidene) Addition Mechanism.

The proposed mechanism for Rh-catalyzed 1,6-enyne cycloisomerization reactions involves an initial [2 + 2]-cycloaddition of a Rh vinylidene and the pendant alkene.⁴ The metallacyclic intermediate then undergoes β-hydride elimination followed by C–H reductive elimination to form the product. There is evidence to support a similar mechanism for the Ni₂-catalyzed reductive cyclization, as opposed to a direct C–H insertion mechanism as would be typical for free vinylidene 1,5-insertions into C(sp³)–H bonds (Figure 3).¹⁵

Figure 3.

Experiments supporting a stepwise vinylidene addition mechanism.

A substrate containing a methyl substituent on the alkene was prepared and tested in the reductive cyclization. The putative metallacycle that would be generated from this substrate would have two potential pathways for β-hydride elimination. Indeed, 33 reacts under standard catalytic conditions to yield a 1:4 mixture of two isomeric products (34 and 35) in a combined yield of 84%.

Substrates 36-E and 36-Z react in a stereoconvergent fashion to provide the same product with Z stereochemistry at the exocyclic alkene. One possible explanation for this observation is that the alkene stereochemistry scrambles through a reversible β-hydride elimination/migratory insertion process. This result is inconsistent with a direct C–H insertion mechanism, which would preserve the alkene stereochemistry.

Electrocatalytic Reductive Cyclizations and the Role of Zn.

We next sought to determine whether the only role of Zn is to turn over the catalyst or whether it might be more intimately involved in the mechanism of the cyclization, for example, by forming organozinc intermediates. In the former case, it should be possible to develop an electrocatalytic variant of the cyclization, excluding Zn metal or any Zn(II) salts.¹⁶

By cyclic voltammetry (CV), (^i-PrNDI)Ni₂Cl₂ (5) possesses two reduction waves at −1.14 V and −1.34 V vs. Cp₂Fe/Cp₂Fe⁺ and one return oxidation at −1.05 V (Figure 4c). We attribute the presence of two reductive events to partial dissociation of chloride in the polar NMP solvent.¹⁷ When the CV of 5 was obtained in THF, only the more cathodic wave at −1.34 V was observed.¹⁷ There is a second reversible reduction at E_1/2 = −1.64 V. When (^i-PrNDI)Ni₂Cl₂ complex 5 is stirred over excess Zn powder, it undergoes a one-electron reduction to the violet (^i-PrNDI)Ni₂Cl complex 3, and no further reduction is observed. This result suggests that the mechanism of catalysis does not require accessing the two-electron reduced state of the Ni₂ complex. When CV experiments were conducted using solutions of complex 5 containing substrate 1, the reduction events for 5 became irreversible, indicating that activation of the 1,1-dichloroalkene by the reduced catalyst is rapid on the CV timescale.

Figure 4.

(a) Controlled potential electrolysis experiments for the catalytic conversion of 1 to 2. (b) Experimental setup (CE = counter electrode; RE = reference electrode; WE = working electrode) and a proposed one-electron catalytic cycle. (c) Cyclic voltammetry data for complex 5 in the presence and absence of 1 (0.3 M [n-BuN₄]PF₆ in NMP; 100 mV/s scan rate). All cyclic voltammograms were internally referenced to the Cp₂Fe/Cp₂Fe⁺ couple.

Based on these data, bulk electrolysis experiments were carried out at constant potential (–0.75 V vs. a Ag/Ag⁺ reference electrode; approximately −1.35 V vs. Cp₂Fe/Cp₂Fe⁺) using substrate 1 in a divided H-cell (Figure 4a and 4b). In order to ensure high mass transport, a large surface area carbon cloth working electrode was used, and the reaction was stirred vigorously at 900 rpm. After 8 h of electrolysis, an aliquot of the reaction mixture was removed. Analysis by ¹H NMR spectroscopy indicated that product 2 was formed in 94% yield, and 20.6 C of charge (2.27 F/mol of product) were passed during the 8 h reaction period, corresponding to a Faradaic efficiency of 88%.

Stoichiometric Reductive Cyclization Reactions.

Having established that Zn is only involved in catalyst turnover, we next examined a series of stoichiometric reactions using (^i-PrNDI)Ni₂(C₆H₆) (6). A reaction between 6 (1.0 equiv) and substrate 1 was carried out in C₆D₆ and monitored by ¹H NMR spectroscopy (Figure 5a). Over the course of 24 h, cyclized product 2 was generated (90% yield), concomitant with the formation of (^i-PrNDI)Ni₂Cl₂ (5). When the same experiment was repeated using substrate 25, some amount of (^i-PrNDI)Ni₂Cl₂ (5) was still generated, but the expected cyclized product 26 could not be detected, and no other diagnostic peaks were observed in the ¹H NMR spectrum, even after 24 h (Figure 5b). A frozen-solution EPR spectrum of the product mixture showed a prominent rhombic signal, indicating the presence of an S = 1/2 Ni₂ complex (Figure 5d).

Figure 5.

Experiments probing the stoichiometric activation of 1,1-dichloroalkene substrates with (^i-PrNDI)Ni₂(C₆H₆) complex 6. (a) Stoichiometric reductive cyclization of 1 using complex 6. (b) Stoichiometric reaction between complex 6 and substrate 25 to form metallacycle 38. (c) XRD structure of metallacycle 38. Selected bond metrics: Ni1–Ni2: 2.743(1) Å; Ni1–C1: 2.000(6) Å; Ni1–C3: 1.967(4) Å; Ni2–C1: 1.954(5) Å; Ni2–Cl: 2.243(2) Å. (d) Frozen solution EPR spectrum for 38 (105 K; toluene). Simulated parameters: g = [2.378, 2.226, 2.093]. * corresponds to a S = 1/2 impurity. (e) UM06-L spin density plot for 38.

After removing the (^i-PrNDI)Ni₂Cl₂ byproduct (5) from this crude mixture, the EPR active complex was successfully isolated, and single crystals were obtained from a pentane/THF solution. XRD analysis revealed that the unknown Ni₂ species (38) is a metallacycle, resulting from the addition of a vinylidene to the pendant alkene (Figure 5c). The two Ni atoms, the alkene, and the vinylidene form a five-membered ring, and Ni1 is engaged in a secondary π-interaction with the C1–C2 double bond. Only one chloride is present in the structure, consistent with 38 being in a non-integer spin system. Therefore, the balanced equation for this reaction involves 1.5 equiv of (^i-PrNDI)Ni₂(C₆H₆) (6) reacting with 1.0 equiv of substrate 25 to form 0.5 equiv of (^i-PrNDI)Ni₂Cl₂ (5) and 1.0 equiv of 38.

Complex 38 is best described as a mixed valent Ni(II)/Ni(I) species, with the NDI ligand having a neutral charge. The neutral charge of the ligand is supported by analysis of the ligand C–C and C–N bond metrics in the XRD structure. The Ni1–Ni2 distance is 2.743 Å, which is considerably longer than the Ni(I)–Ni(I) single bond in complex 6, suggesting that there is minimal covalent bonding between the two Ni atoms. The spin density plot shows that the unpaired electron is localized on Ni2, which is in a pseudo-tetrahedral geometry (τ₄ = 0.85) (Figure 5e). Ni1 is pseudo-square planar (τ₄ = 0.14) and is calculated to be in a low-spin configuration.

Metallacycle 38 is sufficiently stable at room temperature to allow for its isolation. However, upon extended heating at 50 °C in NMP/THF, the optimal solvent mixture for the catalytic cyclization, it reacts to form cyclized product 26 in 64% yield after 9 h (Figure 6). Given the relatively slow rate of this stoichiometric process and the modest yield of 26, we wondered whether the catalytic cyclization may require reduction of metallacycle 38. Indeed, metallacycle 38 possesses a reversible one-electron reduction at −1.59 V vs. Cp₂Fe/Cp₂Fe⁺, which would be accessible using Zn (see Supporting Information).

Figure 6.

Comparison of rates and yields for stoichiometric cyclizations to form 26 as a function of oxidation state.

When 38 was stirred over Zn (3.0 equiv) in NMP/THF at 50 °C, product 26 was obtained in 96% yield after only 1 h. Given the fast rate of product formation, we were unable to isolate the reduced metallacycle 39. However, when metallacycle 38 was dissolved in C₆D₆ and reduced over KC₈ (1.2 equiv), a new set of diamagnetic signals was observed in the ¹H NMR spectrum. These signals decay over the course of 3 h at 50 °C to provide product 26 in 58% yield and the (^i-PrNDI)Ni₂(C₆D₆) complex 6 in 61% yield.

DFT Calculations.

We used unrestricted DFT (UM06-L)^18,19,20 to calculate the thermodynamics, barriers, and spin crossovers (MECPro program)²¹ of reaction pathways that convert 40 to 38 and then to 44 (Figures 7 and 8). From previous benchmarking studies of DFT methods, the M06-L functional was identified as producing relative energies of different spin states for (NDI)Ni₂ complexes that matched experimental data.²² Five major pathways were explored for the conversion of 40 to 38, each having oxidative addition, migratory insertion, and reduction steps in different orders. Figure 7 displays only the lowest energy pathway.

Figure 7.

(a) DFT-calculated reaction pathways for the cyclization of 25 to give complex 38. Relative Gibbs free energies are shown in kcal/mol. Isopropyl groups were modeled as methyl groups. (b) 3D representations of key transition-state structures. Spin states are given in parenthesis. Hydrogens are removed for clarity. Distances are reported in Å.

Figure 8.

DFT-calculated reaction pathways for conversion of 38 to 44/46 as a function of oxidation state. Relative Gibbs free energies are shown in kcal/mol.

The initial π-complex 40 favors a triplet ground state, but the singlet is only 3 kcal/mol higher in energy, suggesting that it may have some multireference character. Oxidative addition of the C–Cl bond (TS40) has a low barrier on both the singlet and triplet surfaces. Both Ni centers are directly involved in the oxidative addition, such that the resulting vinyl and chloride fragments in 41 are situated in bridging positions. Vinyl complex 41 has a quintet ground state, and an energy-degenerate spin crossover structure (minimum energy crossing point; MECP) between the triplet and quintet surfaces was identified.

From 41, reduction with Zn coupled to chloride loss is endothermic by 4 kcal/mol. The thermodynamics of this first reduction was calculated as the conversion of Zn₂ to linear Zn₂Cl. Later, the second reduction was calculated as the conversion of Zn₂Cl to ZnCl₂ and Zn. This reduction step provides a low barrier pathway for activation of the second C–Cl bond through TS41. Similar to TS40, both Ni centers participate in C–Cl oxidative addition. This transition state leads to bridging vinylidene complex 42, which features a secondary interaction between the vinylidene π-bond and one of the Ni centers (Figure 7). The isolable complex 38 can be accessed through a low energy migratory insertion pathway with a barrier of <2 kcal/mol on the quartet surface (TS42). In progressing from 42 to 38, the Ni–Ni distance elongates from 2.60 to 2.69 Å to accommodate the incoming alkene.

As an alternative mechanism for the formation of 38, a non-reductive second C–Cl oxidative addition step from 41 was considered but found to be higher in energy. Furthermore, without formation of vinylidene 42, migratory insertion of the pendant alkene into the Ni(vinyl) bond of 41 was calculated to be prohibitive in energy (activation barrier: 28 kcal/mol).

Finally, we examined pathways for the conversion of metallacycle 38 to the cyclized product. The fact that complex 38 can be isolated indicates that the subsequent β-hydride elimination must have a relatively high barrier (Figure 8). Our DFT calculations corroborate this finding, and the calculated barrier for β-hydride elimination is 29 kcal/mol. Reduction of 38 and loss of chloride lowers the migratory insertion barrier to 26 kcal/mol, which is also consistent with experimental findings (Figure 8). β-Hydride elimination gives hydride 43. Reductive elimination occurs at a single Ni center and has a barrier of 3 kcal/mol, yielding the product complex 44, where the conjugated diene is η⁴-coordinated to both Ni centers.

CONCLUSIONS

In summary, dinickel catalysts promote the reductive cyclization of 1,1-dichloroalkenes containing pendant alkenes through a Ni₂-bound bridging vinylidene intermediate. Previous approaches to carrying out such cyclizations relied on a metal-induced alkyne-to-vinylidene isomerization. However, there is a limited scope of catalysts that can carry out this step, and endo cycloisomerizations are significantly less common that exo cycloisomerizations that proceed through metal(alkyne) intermediates. This reductive strategy may therefore prove useful in other transformations where alkynes are not viable as metal(vinylidene) precursors.

Mechanistic studies are consistent with a putative Ni₂(vinylidene) undergoing migratory insertion to generate a metallacyclic intermediate. Product formation would then proceed by β-hydride elimination followed by C–H reductive elimination. In the case of a five-membered ring-forming reaction, the key metallacyclic intermediate proved to be an isolable complex, and reduction with Zn resulted in high-yielding conversion to product. DFT models reveal that the canonical organometallic steps—oxidative addition, migratory insertion, β-hydride elimination, and reductive elimination—involve the direct participation of both Ni centers, with the Ni–Ni interaction responding as needed to accommodate various reactive intermediates.