C₂-Symmetric Dinickel Catalysts for Enantioselective [4 + 1]-Cycloadditions

Abstract

Dinickel naphthyridine–bis(oxazoline) catalysts promote enantioselective intermolecular [4 + 1]-cycloadditions of vinylidene equivalents and 1,3-dienes. The products of this reaction are methylenecyclopentenes, and the exocyclic alkene is generally obtained with high Z selectivity. E and Z dienes react in a stereoconvergent fashion, providing cycloadducts with the same sense of absolute stereochemistry and nearly identical ee values. This feature allows dienes that are commercially available as E/Z mixtures to be used as substrates for the cycloaddition. A DFT model for the origin of asymmetric induction is provided.

Graphical Abstract

Several of the most universal ligands in asymmetric catalysis employ substituted oxazolines as the source of chirality.¹ Prominent examples include C₂-symmetric bis(oxazoline) and pyridine–bis(oxazoline) ligands, which have been used in transformations as mechanistically diverse as cycloadditions, carbonyl additions, alkene additions, cross-couplings, and C–H functionalizations.² Being derived from α-amino acids, oxazolines can be readily configured to meet different requirements for steric discrimination. Additionally, by constraining amino acids into five-membered rings, rotational flexibility is minimized, and the stereogenic substituent is positioned close to the substrate binding site. Whereas most oxazoline-containing ligands are designed to support single metal ions, it would be attractive to incorporate such frameworks into new chiral multinucleating ligands.

Dinuclear complexes containing metal–metal bonds can exhibit catalytic properties that are challenging to replicate using mononuclear complexes.³ For example, (NDI)Ni₂ catalysts (NDI = naphthyridine–diimine) promote [4 + 1]-cycloadditions^4,5,6 of vinylidenes and 1,3-dienes.⁷ Based on experimental observations and DFT modeling studies, we hypothesized that the preference for [4 + 1]-cycloaddition over competing cyclopropanation arises due to the stepwise nature of the vinylidene addition. In our proposed mechanism, migratory insertion of the diene into a Ni₂(C=CR₂) forms a dinickel metallacycle. Reductive elimination could then proceed in either a 1,3 or a 1,5 fashion, the latter of which appears to be favored.

Seeking to render this process enantioselective, we prepared several chiral NDI ligands using imines derived from α-stereogenic primary amines.⁷ Promising levels of asymmetric induction were observed in intramolecular [4 + 1]-cycloadditions (Figure 1A). However, ee values greater than 90% could not be obtained, and these ligands proved to be ineffective for intermolecular cycloadditions. Cramer recently described alternative chiral NDI variants based on C₂-symmetric 2,6-di-(1-arylethyl)anilines.⁸ These ligands were used to carry out enantioselective methylenecyclopropanation reactions.

Figure 1.

Asymmetric [4 + 1]-cycloaddition reactions of vinylidenes and 1,3-dienes using C₂-symmetric dinickel catalysts.

We reasoned that replacing the imine donors of NDI with oxazolines might impart added rigidity and provide opportunities to sample a broader range of steric environments in the catalyst active site. To this end, we report here that (napbox)Ni₂ complexes (napbox = naphthyridine–bis(oxazoline)) catalyze enantioselective intermolecular [4 + 1]-cycloaddition reactions of terminal monosubstituted and 1,2-disusbstituted 1,3-dienes (Figure 1B).⁹ Notably, several dienes that did not undergo cycloaddition using (NDI)Ni₂ catalysts—for example, those containing Ar groups at the 1-position—are viable substrates using (napbox)Ni₂ catalysts, indicating that the two classes of catalysts have complementary substrate scopes.

(napbox)Ni₂OAc Complexes.

Synthesizing well-defined (napbox)Ni₂ complexes necessitated developing a new metalation procedure. Reacting ^Me2napbox 1 with Ni(cod)₂ (2.0 equiv) or with a combination of Ni(cod)₂ (1.0 equiv) and Ni(dme)Br₂ (1.0 equiv), routes previously used to synthesize (NDI)Ni₂ species,¹⁰ yielded monometallated products with concomitant precipitation of Ni(0). Templating the formation of the Ni–Ni bond using a carboxylate counterion proved to be a more fruitful approach. Accordingly, comproportionation of Ni(cod)₂ (1.5 equiv) and Ni(OAc)₂ (0.5 equiv) in the presence of ^Me2napbox 1 at elevated temperatures (75 °C, 24 h, THF) provided (^Me2napbox)Ni₂(OAc) (2) as a crystalline violet solid in 60% yield (Figure 2A).

Figure 2.

(a) Synthesis of (^Me2napbox)Ni₂(OAc) (2). (b) Comparison of structural parameters for ^Me2napbox (1) and (^Me2napbox)Ni₂(OAc) (2) showing bond distances relevant to ligand-based redox activity. (c) Calculated SOMO for (^Me2napbox)Ni₂(OAc) (2) (BP86/6–311G(d,p) level of DFT). (d) X-band EPR spectra for (^Me2napbox)Ni₂(OAc) (2) (5.5 K, g = [2.0318, 2.0093]) and (^i-PrNDI)Ni₂(OAc) (3) (5.5 K, g = [2.0634, 2.0186]). (e) Cyclic voltammograms for (^Me2napbox)Ni₂(OAc) (2) and (^i-PrNDI)Ni₂(OAc) (3) (N₂ atmosphere, glassy carbon working electrode, 0.3 M [n-Bu₄N]PF₆ in THF, 100 mV/s scan rate).

The XRD structure of 2 reveals that acetate coordinates in a μ-O,O’ fashion coplanar with the napbox ligand (Figure 2B). The local geometry at Ni is square planar, and the Ni–Ni distance (2.372(1) Å) is within the range of other compounds containing Ni(I)–Ni(I) single bonds.¹¹ The bond metrics associated with the napbox π-system exhibit characteristic features of ligand-centered reduction, such as contracted C1–C2/C3–C4 distances and elongated N1–C1/N3–C3 and N2–C2/N4–C4 distances.¹² The EPR spectrum for 2 is indicative of an S = 1/2 ground state, with the unpaired electron predominantly residing on the napbox ligand (Figure 2D). The g_iso value at room temperature (2.0305), is near that of the free electron, and the g anisotropy in frozen solution (Me-THF, 5.5 K) is relatively small (g_max – g_min = 0.0225). Thus, the structural and spectroscopic data for 2 are consistent with it being best described as a Ni(I)–Ni(I) complex bearing a napbox radical anion.¹³ According to DFT models, 61% of the total spin density for 2 is localized on the napbox ligand (Figure 2C).

For the Ni₂ catalyzed reductive [4 + 1]-cycloaddition, it is critical for the redox potentials of the complex to be matched with the Zn/ZnCl₂ couple in order to achieve turnover. Therefore, we were interested in assessing the electronic perturbation associated with replacing imine donors with oxazolines. By cyclic voltammetry, (^Me2napbox)Ni₂(OAc) (2) possesses a quasi-reversible oxidation at E_1/2 = –0.99 V vs. Cp₂Fe/Cp₂Fe⁺ (–0.61 V vs. SCE) and a fully reversible reduction at E_1/2 = −1.81 V (–1.43 V vs. SCE) (Figure 2E).¹⁴ These redox events are nearly coincident with those observed for (^i-PrNDI)Ni₂(OAc) (3). The modest anodic shift in the potentials for the ^Me2napbox complex 2 may be a consequence of the oxygen atoms being inductively electron-withdrawing.

Enantioselective [4 + 1]-Cycloadditions.

Several chiral derivatives of napbox were then synthesized and tested in the model intermolecular [4 + 1]-cycloaddition of 1,1-dichloroalkene 4 and 1,3-diene 5 (see Supporting Information for reaction optimization studies). The (^t-Bunapbox)Ni₂OAc complex (7) provided 6 in 34% yield (18:1 Z/E ratio) and 89% ee for the major Z stereoisomer (Figure 1B). Significant improvements in yield and selectivity were obtained using the tetraarylated (^t-BuPh2napbox)Ni₂OAc catalyst 8. Under fully optimized conditions, 6 was obtained in 81% yield as a >20:1 ratio of Z/E stereoisomers. The major Z isomer was produced in 95% ee. Interestingly, (NDI)Ni₂ catalysts proved to be ineffective at promoting [4 + 1]-cycloadditions with this diene. For example, (^i-PrNDI)Ni₂(OAc) catalyst 3 generated 6 in only 4% yield and a relatively low Z/E selectivity of 5:1.

Terminal monosubstituted and 1,2-disubstituted dienes were effective substrates in the enantioselective [4 + 1]-cycloaddition reaction, providing ee values up to 98% (Figure 3). Dienes possessing other substitution patterns, such as 1,3- and 1,4-disubstitution, generally afforded low yields. A variety of common polar functional groups are tolerated, including ethers, thioethers, esters, nitriles, boronate esters, electron-rich heterocycles, and aryl chlorides. Cycloadducts 6, 22, and 32 were obtained as crystalline solids, allowing for the unambiguous assignment of absolute stereochemistry and alkene geometry by XRD analysis.

Figure 3.

Substrate Scope Studies. Isolated yields were obtained following purification by column chromatography. Reaction conditions: 1,1-dichloroalkene (0.3 mmol, 1.0 equiv), 1,3-diene (2.0 equiv), catalyst 8 (10 mol%), Zn (3.0 equiv), 5:1 MTBE:DMA (1.0 mL), 35 °C, 48 h. ^aReactions were run with 15 mol% catalyst loading. ^bReactions were run with ^tBunapbox (5 mol%) and Ni(dme)Cl₂ (15 mol%).

A double stereodifferentiation experiment was carried out using (+)-nopadiene (29),¹⁵ which is available in enantiopure form from (–)-myrtenal. Using the S,S enantiomer of catalyst 7, cycloadduct 30 was obtained with a high dr (32:1). The R,R enantiomer of 7 reverses this selectivity, favoring the other diastereomer (31) with a more moderate dr of 4:1.

The E and Z stereoisomers of 1-phenyl-1,3-butadiene were independently reacted under the [4 + 1]-cycloaddition conditions and found to converge on the same major enantiomer of cycloadduct 32 with nearly identical ee values (97%) (Figure 4A). For the reaction with (Z)-1-phenyl-1,3-butadiene, the unreacted diene was recovered at the end of the reaction with minimal isomerization to the E form, ruling out an off-cycle Z-to-E isomerization. The stereoconvergent nature of the reaction proved to be synthetically useful in cases where 1,3-dienes are most readily available as E/Z mixtures (Figure 4B). For example, a commercial sample of 1,3-hexadiene (1.3:1 E/Z ratio) provides cycloadduct 33 in 47% yield and 93% ee. β-Ocimene (2:1 E/Z ratio) was reacted with 1,1-dichloroalkene 4 to generate 34 in 64% yield and 88% ee.

Figure 4.

Stereoconvergent cycloadditions with E and Z 1,3-dienes. (a) (E)- and (Z)-1-phenyl-1,3-butadiene undergo [4 + 1]-cycloaddition to provide the same enantiomer of 32. (b) Highly enantioselective cycloadditions with dienes that are commercially available as E/Z mixtures.

Finally, we developed a computational model (BP86/6–311G(d,p) level of DFT) to rationalize the sense of stereoinduction. In our previous studies, we had identified an energetically viable stepwise pathway for the [4 + 1]-cycloaddition,⁷ involving migratory insertion of the diene into a Ni₂(C=CR₂) species followed by C–C reductive elimination. To examine the asymmetric process, we selected 1-phenyl-1,3-butadiene and 1,1-dichloroalkene 4 as model substrates. There are four possible stereoisomers of the product that can be generated in the cycloaddition: R,Z, R,E, S,Z, and S,E. Each of these products corresponds to one of four possible substrate orientations in the catalyst active site: two orientations of the diene and two orientations of the vinylidene. Additionally, the initial migratory insertion can take place on either end of the diene. Thus, eight total pathways were investigated. These results are summarized in Figure 5A, with only the lowest-energy pathway leading to each product stereoisomer shown.

Figure 5.

DFT model for asymmetric induction (BP86/6–311G(d,p) level of DFT): (^t-Bunapbox)Ni₂Cl, 4, and 1-phenyl-1,3-butadiene. (a) Calculated lowest energy [4 + 1]-cycloaddition pathways leading to the four stereoisomers of cycloadducts 32. Relative free energies for the major pathway are shown in units of kcal/mol. (b) Migratory insertion and reductive elimination transition structures highlighting key steric interactions in pathways leading to minor stereoisomers.

The migratory insertion and C–C reductive elimination transition states are sufficiently close in energy that either step in principle could be selectivity determining. Gratifyingly, the pathway leading to the R,Z cycloadduct contained the lowest activation energies, consistent with the major stereoisomer observed experimentally. The other pathways suffer from either a high-barrier migratory insertion or reductive elimination step (Figure 5B). For example, in the pro-S,Z migratory insertion transition state, prohibitive steric interactions with the diene Ph substituent causes the oxazoline N to dissociate from Ni. In the pro-S,E and pro-R,E pathways, it is the reductive elimination barriers that are high due to interactions between the oxazoline t-Bu group and either the diene Ph or vinylidene 4-OMePh substituent.

In summary, a new class of chiral (napbox)Ni₂ catalysts enabled the development of highly enantioselective intermolecular [4 + 1]-cycloadditions. Interestingly, E- and Z-dienes provide the same major enantiomer of the product and similar ee values, allowing readily available E/Z mixtures to be used as substrates. Computational models are used to rationalize the sense of enantioinduction and high preference for Z stereochemistry in the cycloadducts. The formation of the minor stereoisomers are suppressed due to unfavorable steric interactions with the catalyst t-Bu substituents in either the migratory insertion or reductive elimination transition states. Ongoing efforts are directed at developing enantioselective variants of other dinickel catalyzed processes.