|Home | About | Journals | Submit | Contact Us | Français|
A combination of physical organic experiments and quantum chemical calculations were used to construct a detailed mechanistic model for the Ni(0)-N-heterocyclic carbene-catalyzed vinylcyclopropane—cyclopentene rearrangement that involves a muti-step oxidative addition/haptotropic shift/reductive elimination pathway. No evidence for the intermediacy of radicals or zwitterions was found. The roles of substituents on the vinylcyclopropane substrate and variations in the ligands on Ni were evaluated. It is postulated that bulky carbene ligands facilitate formation of the active catalyst species.
While transition metal-promoted cycloadditions have shown tremendous synthetic promise,1 transition metal-promoted sigmatropic shifts have not been exploited nearly as frequently.2,3,4 However, one sigmatropic shift that has been shown to be amenable to manipulation by transition metals is the [1,3] sigmatropic rearrangement of vinylcyclopropanes to cyclopentenes.3,4 In the absence of transition metals, Lewis acids, or activating substituents, vinylcyclopropane–cyclopentene rearrangements require high-temperatures and tend to produce various byproducts, most of which are thought to arise from alkyl/allyl diradical intermediates (Scheme 1).5 Various transition metals (e.g. Pd(0),6 Rh(I),7 and Ni(0)8) have been shown to facilitate vinylcyclopropane–cyclopentene rearrangements, but in many cases, activating groups on the vinylcyclopropane appear to be required (see Scheme 2 for representative examples).
Recently, Louie and co-workers reported that Ni(0)–N-heterocyclic carbene (NHC) complexes are effective catalysts for vinylcyclopropane–cyclopentene rearrangements involving unactivated vinylcyclopropane reactants (Table 1).8c 1,1-Disubstituted substrates underwent facile rearrangement at room temperature without the aid of activating groups (entry 1). Trisubstituted substrates required slightly elevated temperature (60 °C) (entry 2). 1,2-Disubstituted substrates required both elevated temperature and activation with an aromatic group (entries 3 and 4).
The most likely mechanistic pathways for the metal promoted isomerization of vinylcyclopropanes are shown in Figure 1.4 Both radical and zwitterionic mechanisms would form an η1-alkyl species through initial addition of the metal to the alkene (pathway a). Alternatively, oxidative addition to the cyclopropane would lead to a metallacyclobutane intermediate, which can occur with or without initial complexation of the substrate to the metal (pathways b and c). Which of these pathways occurs for a given experimental system, depends on the specific metal and ligands used.4
Herein, we describe quantum chemical calculations and physical organic experiments on the mechanism of the Ni(0)-promoted vinylcyclopropane–cyclopentene rearrangement. In addition to proposing a detailed mechanism based on these studies, we also predict the outcome of reactions with a variety of reactants and explore the potential of various non-NHC ligands. The transition state structures we describe in this report should provide important points of departure for the rational design of catalysts for asymmetric versions of this metal-promoted sigmatropic shift.
We began our theoretical studies (see Computational Methods section for details) by computing structures involved in the rearrangement of unsubstituted vinylcyclopropane in the presence of Ni(N,N′-dimethylimidazolylidene). All of the calculations described herein are based on the assumption that the active catalyst is a Ni(0) species. The computed intermediates (INT) and transition state structures (TS‡) for this rearrangement are shown in Figure 2, their relative energies are shown in Figure 3, and a summary of the overall computed mechanism is shown in Scheme 3 (note that we do not intend to imply that a free Ni(N,N′-dimethylimidazolylidene) is generated in solution).
The computed mechanism first involves rearrangement of an initially-formed η2-vinylcyclopropane complex (INT1; note that we also located another conformer of INT1 where the vinylcyclopropane is not oriented appropriately for oxidative addition; the energy of this conformer is marked with an INT1b label throughout the paper)9 via oxidative addition to form a vinylmetallacyclobutane (INT2). This complex then rearranges to a metallacyclohexene (INT4) by way of an η1-alkyl/η3-allyl intermediate (INT3). INT4 then undergoes a conformational change (to INT5) that prepares it for reductive elimination to form a σ-complex of cyclopentene (INT6). This intermediate then rearranges with a very small barrier to a π-complex of cyclopentene (INT7).
Several of the intermediates and transition state structures involved in this process have interesting features. First, it is somewhat unusual that the vinylmetallacyclobutane, η1-alkyl/η3-allyl, and metallacyclohexene structures are all found as minima. Even though the barrier for conversion of the vinylmetallacyclobutane (INT2) to the η1-alkyl/η3-allyl (INT3) is small enough to be of little chemical consequence (in other words, this portion of the energy surface is rather flat), the barrier for conversion of INT3 to the INT4 is substantial (approximately 13 kcal/mol; see Figure 3). Note that the Ni–CNHC distance in TS34‡ (1.85 Å) is shorter than the corresponding distance in all of the other structures (except the weakly bound σ-complex (INT6) and the transition structure (TS67‡) for its conversion to INT7), hinting that Ni–allyl bonding in this structure is weaker than in the other allyl complexes INT2-INT5. Second, a crankshaft rotation of the ethylene group in metallacyclohexene INT4 (dashed box in Figure 2) is necessary to bring the two methylenes attached directly to Ni closer together so that reductive elimination can occur readily. Third, formation of the σ-complex (INT6) is also noteworthy. Note that both of the C–H bonds near to the Ni are elongated.
While our calculations suggest that the formation of vinylmetallacyclobutane INT2 is the rate-determining step, energies of the transition state structures for formation of metallacyclohexene INT4 (TS34‡) and σ-complex INT6 (TS56‡) are only approximately 3 and 2 kcal/mol lower in energy, respectively. All of the intermediates that we located (INT1-INT6) have similar energies (spanning only approximately 3 kcal/mol), while the product π-complex INT7 is much lower in energy, reflecting the energy difference between σ- and π-complexation of cyclopentene; interestingly, throughout the reaction there appears to be a balance of strain and Ni–carbon bonding.
Several additional mechanistic possibilities were also explored. First, we considered the possibility that diradical species might be involved (e.g. Figure 1, path a). Using a system with a 2-methyl group (for additional details on this particular system, see below), the triplet diradical counterpart of INT1 was located as a minimum (using UB3LYP/LANL2DZ) and this structure is ~23 kcal/mol higher in energy than INT1. Similarly, the triplet diradical analog of INT4 was 6 kcal/mol higher than INT4 (see Supporting Information for details). Based on these results, we did not pursue further calculations on diradicals.
Second, we explored the possibility of zwitterionic intermediates. Here we assessed the energies of species such as A (below) by performing constrained optimizations with the indicated bond angle fixed to a value of 109.5°. The energies of the resulting structures (for substrates with alkyl substituents in various positions; see Supporting Information for further details) were generally 15–20 kcal/mol above that of the corresponding INT1 complexes; i.e. they were generally similar in energy to TS12‡. Thus, we think that mechanisms involving such intermediates are unlikely (note that although these are gas phase calculations, the experimental reactions were carried out in nonpolar aromatic and aliphatic solvents).8c
We also examined the effects of alkyl substitution on the rearrangement. As shown in Figure 4, the overall mechanism does not change upon alkyl substitution, although which of the three highest transition state structures is the absolute highest does vary. Interestingly, the reaction coordinates for the cis- and trans-methylvinylcyclopropane systems actually intersect at INT5 and both systems undergo reductive elimination through the same TS56‡ (actually, the two paths involve enantiomeric versions of INT5 and TS56‡). This is a result of a conformational change for the trans system that allows the Ni to avoid being in close proximity to the methyl group in the last few intermediates and transition structures.
The experimental results of Zuo and Louie indicate that systems with internal alkyl groups rearrange readily while systems with terminal alkyl groups do not (Table 1; compare entries 1 and 4).8c Our calculations are consistent with these observations. The overall barrier for rearrangement of the vinylcyclopropane with an internal methyl group is predicted to be 15.5 kcal/mol while the barriers for the systems with a terminal methyl group are higher (19.6 and 17.9 kcal/mol, from the lower energy INT1b in these cases9). Note also that the absolute energy of the transition structures is lower for the internally substituted system. Moreover, this difference in reactivity is likely to be underestimated in our calculations, given that a small NHC was used in the modeling; given the geometries of the transition structures involved (e.g., Figure 1), we expect steric repulsions to be greater for 1-substitution. The internally-substituted system also leads to a trisubstituted alkene, rather than a disubstituted alkene, providing a greater driving force for rearrangement of the internally-substituted systems. An internal methyl group is also predicted to accelerate the rearrangement as compared to the unsubstituted system (vide supra), whose overall barrier (from INT1b) is computed to be 16.8 kcal/mol.
Disubstituted vinylcyclopropanes are predicted to behave much like the monosubstituted systems described above (Figure 5). One key difference, however, is observed for the gem-dimethyl case. Unlike the trans monomethyl case described above, no simple change in conformation can lead to a reductive elimination transition structure where the Ni is not near to a methyl group, and a high energy reductive elimination transition structure is therefore observed (Figure 6, left); again, this steric problem is likely to greater in the experimental system, which involves a bulky NHC. Note also that TS67‡ for this system involves migration of the Ni “the long way around” the 5-membered ring in order to avoid the methyl group (Figure 6, right).
Although experimental results on gem-dialkyl substrates have not been reported (we would predict that the problematic reductive elimination state described above in such systems would hinder rearrangements base on our computations), rearrangements of vicinal dialkyl systems have been reported.8c These systems are observed to rearrange more sluggishly than do monoalkyl substituted systems with internal alkyl groups, but more readily than do monoalkyl substituted systems with terminal alkyl groups. Our predicted barriers for the vicinal dimethyl systems (Figure 5) are 17.3 (here, TS12‡ is the highest energy structure) and 16.0 kcal/mol (here, TS34‡ is the highest energy structure), respectively; interestingly, these barriers are between those computed for systems with internal and terminal methyl groups (vide supra).
The trimethyl substituted vinylcyclopropane shown in Figure 7 behaves very much like the gem-dimethyl substituted system described above. Again, we predict that this sort of system will not rearrange efficiently, however we do predict that the internal alkyl group will lower the overall barrier compared to that for the gem-dimethyl substituted system (23.9 vs. 28.0 kcal/mol).
Louie and coworkers showed that phenyl-substituted vinylcyclopropanes rearrange more easily than do related alkyl-substituted cases, for both internal and terminal substituted cases (Table 1).8c Our calculations do not predict a significant increase in reactivity for the phenyl-substituted systems, however (compare Figure 8 with Figure 4). This discrepancy between experiment and theory perhaps suggests that the main difference between aryl and alkyl substituents in these reactions, in terms of their effects on the overall rearrangement barriers, is likely steric rather than electronic in nature, since our calculations capture electronic effects but underestimate steric effects due to the size of our model ligand.
In the case of the cis-phenyl substituted vinylcyclopropane (Figure 8b), we observed a shorter version of our proposed mechanism (Scheme 3). We were unable to locate the vinylmetallacyclobutane complex (INT2) and instead found a transition state structure (TS13‡, Figure 9a) that connects the η2-vinylcyclopropane (INT1) and η1-alkyl/η3-allyl intermediate (INT3). Thus, in this case, the very shallow minimum associated with INT2 disappears. Additionally, it appears that no conformational changes are required before formation of the σ-complex INT6.
For the trans-phenyl substituted vinylcyclopropane (Figure 8c), an additional interesting transition state structure was also located (TS57b‡, Figure 9b). This transition structure leads to a product complex in which the nickel interacts with the π-system of the phenyl group (INT7b, Figure 9b), rather than the σ-framework of the cyclopentene (INT6). TS57b‡ is more than a kcal/mol higher in energy than TS46‡, however.
The reactivity of several phenyl substituted vinylcyclopropanes was investigated experimentally by determining their rearrangement rates (Table 2). The isomerization of (1-cyclopropylvinyl)benzene (1), was significantly faster than that of (E)-(2-cyclopropylvinyl)benzene (2); note that the rate for 2 (entry 2) was determined at 60°C, while the rate for 1 (entry 1) was determined at 27 °C, since the rate of rearrangement for 1 was too fast to measure at 60 °C. This reactivity difference was borne out in our calculations as well (Figure 8). (Z)-(1-Cyclopropylprop-1-2-yl)benzene (3) isomerizes to the (E)-substrate (2) before rearrangement (entry 3), so a rearrangement rate was not determined for this specific substrate. When (E)-(1-cyclopropylprop-1-2-yl)benzene (4) was subjected to the isomerization conditions, no reaction was observed, even at higher temperatures (80 °C, entry 4). These results are consistent with the unwillingness of gem-disubstituted vinylcyclopropanes to rearrange (vide supra).
The initial cis-trans isomerization of 3 could be explained by a diradical mechanism, but our calculations (vide supra) suggest that this is unlikely. In an attempt to settle this issue, the reaction of 3 was carried out in the presence of TEMPO and BHT (2,6-dit-tert-butyl-p-cresol) in an effort to trap any radical intermediates. Under these conditions, no decrease in conversion or yield was observed, consistent with the absence of radical intermediates.
The effect of substituents on the phenyl ring on the rate of rearrangement was also examined (Table 3). Substrates bearing electron-withdrawing groups (entries 1 and 2) were found to rearrange more quickly than the unsubstituted substrate (entry 4) and substrates bearing electron-donating groups (entries 5 and 6) rearranged more slowly. The Hammett plot (Figure 10)11, 12 constructed from this data had a ρ value of 0.11. A ρ value greater than 0.2 is typically considered to indicate a significant substituent effect.18 Therefore, although substituents on the aryl ring have an effect on the rate of the reaction, this effect is small, consistent with zwitterionic (or radical) mechanisms not contributing significantly. The effect of varying the electron density of the aromatic group was also explored by theoretical calculations, and as shown in Table 3 (last column), the experimentally determined order of reactivity is reflected in the computed rearrangement barriers for processes not involving zwitterionic or diradical intermediates.
The effect of the structure of the carbene ligand on the rearangement was also examined by measuring the 1st order rate constants for rearrangement of a variety of ligands (Table 4). The rearrangement of (E)-(2-cyclopropylvinyl)benzene (2) in the presence of 1,3-bis-(2,6-diisopropylphenyl)-4,5-dihydroimidazol-2-ylidene (SIPr) is 3.2 times faster than the rearrangement of 2 in the presence of 1,3-bis-(2,6-diisopropylphenyl)-imidazol-2-ylidene (IPr) (entries 1 and 2). Rearrangement is not observed with 1,3-bis-(1,3,5-trimethylphenyl)-imidazol-2-ylidene (IMes) even under more forcing conditions (80°C or 100°C) and longer reaction times (16 hours, entry 3). These results are consistent with previously reported vinylcyclopropane rearrangements in which rearrangements involving the IMes ligand resulted in a dramatic decrease in yield.8c Rearrangement in the presence of IPrCl2 is slightly faster than with the parent IPr ligand (entry 4, perhaps because IPrCl2 is more robust than IPr). The difference in reactivity between the electronically similar IPr, SIPr, and IMes ligands suggests that steric factors play a significant role in the rearrangement, in that the three NHCs with bulky iso-propyl groups near the carbene center promoted reaction, while the one with only methyl groups near the carbene center (IMes) did not.
A variety of ligand/additive combinations were also examined in order to assess the composition of the active catalyst (Table 5). Interestingly, Ni(IPr)2 without any additional additives promoted the rearrangement, although with a much lower yield and rate (entry 4, 32% yield and 14.2 ×10−3 first order rate constant) than Ni(COD)2 with IPr additive (entry 1, 81% yield and 26.2 ×10−3 first order rate constant). Ni(IPr)2 with cyclooctadiene additive gave comparable results (entry 2, 79% yield and 19.0 ×10−3 first order rate constant) to Ni(COD)2 with IPr. When excess IPr was added to the Ni(IPr)2 catalyst, the yield and rate (entry 3, 28% yield and 14.7 ×10−3 first order rate constant) were comparable to Ni(IPr)2 without an additive. Styrene and 1,5-dimethyl-cycloocta-1,5-diene were also examined as additives, and these led to slower reactions and lower yields than did Ni(IPr)2 with COD additive. In addition, the optimized ratio of IPr to COD was determined to be 1:2. These results suggest that COD plays an important role in the formation (or stabilization) of the active catalyst.
In addition, calculations were performed on the rearrangement of the parent vinylcyclopropane system with other heterocyclic carbenes. Adding two chlorines to the imidazolylidene (Figure 11) seems to have only a very small effect on the rearrangement (the predicted overall barrier only changes from 16.8 to 16.5 kcal/mol), consistent with the experimental results in Table 4 (entries 2 and 4). Replacing one NR group of the imidazolylidene with an O or S is also predicted to have only small effects (Figure 11). These two ligands are sterically smaller than the corresponding imidazolylidenes, however, which could present a problem, as described above (also, vide infra). Overall, these results indicate that the rearrangement mechanism—from INT1 to INT7—is not very sensitive to the heterocyclic carbene ligand on Ni.
Rearrangement of the parent vinylcyclopropane system was also examined with trimethylphosphine as a simple model of alkyl phosphines (Figure 11). Although tricyclohexylphosphine was shown experimentally to be an ineffective ligand for promoting the rearrangement of unactivated vinylcyclopropanes,8c the computed mechanism and energetics for the trimethylphosphine promoted rearrangement (Figure 11) are similar to those for the reaction with N,N′-dimethylimidazolylidene. Although the transition structure for conversion of the η1-alkyl/η3-allyl intermediate to the metallacyclohexene intermediate becomes the rate-determining transition state structure in the phosphine system, the overall barrier predicted for rearrangement with trimethylphosphine is comparable to that predicted for rearrangement with N,N′-dimethylimidazolylidene. This suggests that the problem with phosphine ligands observed experimentally8 may be related to initial formation of INT1 rather than the subsequent rearrangement steps.
Taken together, this experimental and theoretical data hints that the most important difference between NHCs that promote rearrangement and NHCs and phosphines that do not may be their bulkiness. Our calculations indicate that in the absence of bulky groups, all of the ligand frameworks should allow for rearrangement. Our calculations suggest, however, that two full-sized IPr ligands are unable to bind simultaneously to Ni when a vinylcyclopropane substrate is bound (i.e., optimization of such structures leads to dissociation of one IPr; see Supporting Information for details). Interestingly, a complex with two tricyclohexylphoshine ligands and a vinylcyclopropane substrate all bound to Ni could be located as a minimum. Thus, we suggest that the key role of bulky groups on IPr (and the related SIPr and IPrCl2) is to promote dissociation of one NHC, allowing INT1 to form; the smaller IMes and differently shaped phosphine ligands are less inclined to dissociate. The effectiveness of adding COD is also likely due to its ability to facilitate NHC dissociation.
We have proposed a catalytic cycle (Scheme 3) for the Ni(0)-catalyzed isomerization of vinylcyclopropanes to cyclopentenes based on physical organic experiments and theoretical calculations. Which step in this process is rate-determining appears to depend on the exact nature of the substrate used. In addition, the importance of generating a catalyst species with only one bound NHC was revealed, pointing to the importance of bulky groups on the NHC and the utility of COD as an additive. These results should aid in future rational design of catalysts for additional substrates and asymmetric versions of this useful metal-promoted sigmatropic shift.
GAUSSIAN0313 was used for all calculations. All geometries were optimized without symmetry constraints using B3LYP/LANL2DZ.14 As a test, the B3LYP/LANL2DZ level was used to optimize the geometries of several Ni-allyl and Ni-NHC complexes for which X-ray structures have been reported, and no significant deviations between the computed and experimental geometries were observed (see Supporting Information for details). The vibrational frequencies of all stationary points were analyzed to characterize structures as minima or transition state structures. Intrinsic reaction coordinate (IRC) calculations15 were used to further characterize the nature of some transition state structures by mapping out the portions of the reaction coordinates near to them (see Supporting Information for details). Geometries for the parent system (Figure 3) were also reoptimized using B3LYP/DZVP2+16 and no significant changes were observed (see Supporting Information for coordinates). Single point energies were calculated for all structures at the B3LYP/DZVP2+ level16 and these B3LYP/DZVP2+//B3LYP/LANL2DZ + ZPE(B3LYP/LANL2DZ) energies are reported throughout the text. (The corresponding energies from B3LYP/LANL2DZ + ZPE(B3LYP/LANL2DZ) calculations can be found in the Supporting Information). This theoretical approach has been used previously to characterize the chemistry of various organometallic complexes involving metals from the first transition series;17 additional tests on the validity of this theoretical approach for the Ni(0) systems explored herein are described below. In general, in order to reduce computational cost, the diisopropylphenyl groups of the IPr ligands were replaced by methyl groups (see Scheme 3). However, tests on the validity of this approximation were also performed using ONIOM(B3LYP/LANL2DZ:UFF) calculations18 with full-sized carbene ligands. In these calculations, the quantum mechanics layer (treated with B3LYP/LANL2DZ) consisted of the vinylcyclopropane system, the Ni atom, and the heterocycle, while the molecular mechanics layer (treated with the Universal Force Field [UFF]) included the diisopropylphenyl groups. For the subset of structures from Figure 2 examined using this method, no significant changes to relative energies were observed (see Supporting Information for details), although we were not able to locate transition structures for reductive elimination, where steric effects may be largest (vide infra). Structural drawings were produced using Ball & Stick.19
In the drybox, vinyl cyclopropane was weighed directly into an oven-dried screw cap vial equipped with a magnetic stir bar and dissolved in hexane (0.1 M). A solution of Ni(COD)2/IPr (2 mol%) (Ni(COD)2 (1.0 equiv.) and IPr (2.0 equiv.) were dissolved in benzene and allowed to equilibrate at room temperature for 3 h) was added and the vial was sealed with a PTFE lined cap. The vial was removed from the drybox and the dark greenish-black reaction was stirred at 60°C. After complete consumption of the substrate, the solvent was removed in vacuo, and the products were purified by silica gel chromatography. See Supporting Information for additional details.
A stock solution of vinyl cyclopropane and ferrocene (10 mol%) was prepared in deuterated benzene. A stock solution of Ni(COD)2 (1.0 equiv.) and IPr (2.0 equiv.) was prepared in deuterated benzene and stirred for 4 h. The solution of vinyl cyclopropane (0.104 mmol) and ferrocene (0.0104 mmol) were syringed into an NMR with a screw cap. Deuterated benzene was added to give an overall solution of 0.2 M. The solution was then frozen in the glovebox before the Ni(COD)2/IPr (0.00520 mmol) was added to prevent mixing. The NMR tube was frozen again after addition of the Ni(COD)2/IPr solution and removed from the drybox. The NMR tube was warmed just before being placed in the NMR. The reaction was monitored by 1H NMR, taking data points every minute. Reaction monitored for at least 3 half lives. See Supporting Information for additional details.
We gratefully acknowledge the University of California—Davis, the National Science Foundation’s CAREER and Partnership for Advanced Computational Infrastructure (PSC) programs, the American Chemical Scoiety’s Petroleum Research Fund (SCW and DJT), and the National Institutes of Health (SCW, DT, MS, GZ, DL, JL, DJT) for support.