Ni(NHC) Catalyzed Rearrangement of 1-Acyl-2-vinylcyclopropanes: Tackling a Mechanistic Puzzle by Combined Experimental and Computational Studies

The Ni(NHC) catalyzed rearrangement of 1-acyl-2vinylcyclopropanes to the corresponding 4-acyl-cyclopent-1enes is highly promising for the synthesis of keto-functionalized annelated biand tricyclic subunits of natural products. Therefore, we investigated the catalytic activity of Ni(NHC) complexes in the rearrangement of 1-acyl-2-vinylcyclopropanes with different ring sizes and substitution patterns. Surprising effects regarding substrate scope and stereoselectivity of the Ni(NHC) catalyzed vinylcyclopropane-cyclopentene rearrangement were observed. Only vinylcyclopropanes with 1-methyl, 1-phenyl, 1,2-


Introduction
Cyclopropanes [1a-1f ] and particularly vinylcyclopropanes [1g-1h] are highly valuable key building blocks and intermediates for a large variety of organic target compounds due to their high reactivity and propensity for rearrangement.
Since the discovery of the [1,3]-sigmatropic rearrangement of vinylcyclopropanes to cyclopentenes more than 50 years ago, [2] the so called thermal vinylcyclopropane (VCP) rearrangement has been studied extensively both with regard to mechanistic details and applications in total synthesis of natural products. [3] Whereas the parent hydrocarbons require elevated temperatures to undergo VCP rearrangement, activated, e.g. donoracceptor-substituted vinylcyclopropanes, can react under much milder conditions, eventually promoted by Lewis acids. [4] Furthermore, transition metal-catalyzed rearrangements have been dialkyl or 2-phenyl-substitution at the vinyl moiety could be rearranged successfully. Moreover, an endo-configuration on the cyclopropane ring was required for successful rearrangement. By treatment of the vinylcyclopropanes with Rh catalysts or Lewis acids, the involvement of Lewis acid catalysis could be ruled out. In order to understand these experimental results and to rationalize the reactivity of the Ni(NHC) complexes computational studies were performed, which provided insights into mechanistic details.
developed, [5] most notably with Pd, [6] Rh, [7] Ni, [8] Mo, [9] Cr, [10] Cu, [11] Fe [12] and more recently Ir [13] complexes. However, in many cases transition metal catalysts need activated substrates, carrying at least two activating groups (EWG and/or EDG), heteroatoms, dienes, ene-ynes or allenes. In contrast, Louie and Tantillo [8d,8e] reported a Ni(NHC) catalyzed VCP rearrangement of unactivated vinylcyclopropanes with alkyl or aryl substituents (eg. 3 → 4, Scheme 1), as well as some activated alkoxysubstituted substrates. As carbonyl groups are rather useful for further functionalization of the rearrangement product, we wondered whether the substrate scope of Ni-catalyzed VCPcyclopentene rearrangements in the presence of NHC ligands could be extended to keto functionalized vinylcyclopropanes with different ring sizes and substitution pattern on the alkene moiety. The respective bicycles are important subunits of a variety of biologically active natural product families such as tetramic acid macrolactams, [14][15][16] e.g., maltophilin (10), [17] or triquinanes, e.g., (-)-hirsutene (7). [18,19] By examining previous work from Hudlicky, [3k] who described the conversion of bicyclic vinylcyclopropane 1 to the bicyclo[3.3.0]octenone 2 in the presence of stoichiometric amounts of Rh complexes, and Louie [8e] (Scheme 1), we were interested in comparing the catalytic activity of Ni(NHC) complexes with Rh complexes in the rearrangement of 1-acyl-2-vinylcyclopropanes 5 to the respective bicycles 6. Ni(NHC) complexes were recently extensively studied in catalytic C-C [20] and C-X [21] bond formations and computational studies provided mechanistic insights. [22] Furthermore, we were also interested in studying whether the steric demand of the NHC ligand is controlling the catalytic activity. Thus, a series of different NHCs was investigated. Upon application of the optimized conditions on the rearrangement of vinylcyclopropanes 8 we noticed surprising effects regarding substrate structure (reactivity) and stereoselectivity and thus examined the Ni(NHC) catalyzed rearrangement by quantum chemical methods. The results of this experimental and computational study are discussed below.

Results and Discussion
In order to obtain the vinylcyclopropane substrates for the catalytic studies, we relied on our previously reported procedure, [23] which allowed in only two steps a simplified access to 1-acyl-2-vinylcyclopropanes 5 and 8 with alkyl or aryl substituents on the alkene moiety via sulfur ylides (Scheme 2). [24] First, the tetrahydrothiophenium salts 12 were synthesized in yields up to 97 % by treating the respective allylic bromide 11 with tetrahydrothiophene. [23] With tetrahydrothiophenium salts 12 in hand the conjugate 1,4-additions to cyclopentenone and cyclohexenone were performed. Deprotonation of the allylsulfonium salts 12 with LiOtBu (2 equiv.) in DMSO at room temperature gave the corresponding ylides which reacted with the respective enones to diastereomeric mixtures of the vinylcyclopropanes 5 and 8. Yields of the bicyclo[3.1.0]hexan-2-ones 5a-e from semi-stabilized sulfur ylides and cyclopent-2-en-1-one were in the range of 19 -58 %. Reactions of cyclohex-2-en-1-one gave the bicyclo-[4.1.0]heptan-2-ones 8a-f in 17 -55 % yields (Scheme 2). Diastereomers of the bicyclo[4.1.0]heptan-2-ones 8 could be separated by flash chromatography, which was advantageous for further studies on the vinylcyclopropane rearrangement in dependence on the geometry of their alkene moiety. In contrast, for cyclopentenone-derived vinylcyclopropanes 5 chromatographic separation of the diastereomers was not possible.
As can be seen from Table 1, no conversion was observed at temperatures between room temperature and 100°C (entries 1-5). After 3 h at 120°C, only dienone 13b (13 %) was detected by GC (entry 6). When the catalysis was performed with 10 mol-% Ni(COD) 2 and 20 mol-% of ligand 14 at 150°C under microwave irradiation, the starting material 5b decomposed (entry 7). In a final attempt vinylcyclopropane 5b was treated with 20 mol-% of Ni(COD) 2 and 100 mol-% of BF 3 ·OEt 2 at room temperature. After 45 h, enone 15b was obtained in 77 % yield (entry 8), suggesting a Lewis acid catalysis mechanism for the ring opening pathway.
The failure of vinylcyclopropane-cyclopentene rearrangement of 5b to 6b agrees with results by Louie, [8d,8e] who found that non-activated substrates with vic-disubstituted C=C double bonds were in most cases poorer substrates for the Ni(NHC)catalyzed VCP rearrangement as compared to those with gem-Scheme 2. Synthesis of racemic vinylcyclopropanes from sulfur ylides and cyclopentenone or cyclohexenone.  In order to figure out the temperature dependence of the VCP rearrangement to bicyclic product 6c vs. the isomerization to dienone 13c, a series of catalytic reactions with 10 mol-% of Ni(COD) 2 and 20 mol-% of ligand precursor 14·HCl were run in toluene at various temperatures (Table 2).
Aliquots were taken after 1 h, 2 h, 3 h, and 20 h respectively and analyzed by GC using n-dodecane as internal standard. The results in Table 2 clearly indicate that reaction temperatures below 100°C favored isomerization to 13c (entries 2-5), whereas at 100°C almost equimolar amounts of bicyclo-[3.3.0]octenone 6c and dienone 13c were formed (entries 6, 7). At 120°C, VCP rearrangement was preferred resulting in a (69:31) mixture of 6c and 13c after 19 h and complete consumption of the starting material (entry 9).
It should be noted that the Ni-catalyzed reactions of nonactivated vinylcyclopropanes studied by Louie were run at room temperature or in some cases at 60°C. [8d,8e] In our case, the presence of the carbonyl group seems to increase the activation barrier of the reaction. Longer reaction times at moderate temperature changed the product ratio in favor of 13c (entries 2,3 and 4,5) indicating an induction period for the formation of 13c. Ring-opening reactions to 13c which are favored at moderate temperatures can thus be reduced by increasing the reaction temperature to above 100°C.
Based on these results we studied the influence of a series of NHC ligands 16-25 (Scheme 3) on the catalytic reactions. As described above, ligand precursors 16·HCl-25·HCl were converted in situ to the free NHC ligands 16-25. In order to sup-press the formation of the ring-opened product 13c, the reactions were performed at 120°C (Table 3). When N-heterocyclic carbene 16, the saturated counterpart of 14, was used, the desired bicycle 6c was exclusively formed and no trace of 13c could be detected (entries 1, 2). Increasing the electron donat-Scheme 3. N-heterocyclic carbene precursors used in this study. ing character, but also reducing the steric demand, by using ligand 17 resulted in decreased conversion and selectivity (entries 3, 4). After 3 h 63 % of 6c/13c (72:28) were obtained (entry 3), which changed to 70 % after 16 h (6c/13c = 60:40) (entry 4). A similar product ratio albeit at complete conversion was detected for the ligand 18 with decreased steric hindrance as compared to 16 (entries 5, 6). Substituting the 5-membered carbene in 18 by a 6-membered carbene in ligand 19 again raised the ratio in favor of 6c (100:0), but at the expense of catalytic activity (34 % conversion after 16 h, entry 8). Next aromatic NHCs with alkyl substituents of different steric bulkiness were tested. While ligands 21 and 23 gave no conversion at all (entries 11,12,15,16), ligands 22 and 20 provided only the dienone 13c (entries 9, 10, 13, 14). From the above mentioned results we would have expected that ligand 25 with a nonaromatic imidazolidinylidene unit should outperform 24 with an aromatic NHC moiety. Surprisingly, 24 showed a clear preference for 6c over 13c (90:10) at quantitative conversion after 23 h (entry 18), while 25 was completely inactive even after prolonged reaction times (entry 20).
Eur. J. Org. Chem. 2019, 6285-6295 www.eurjoc.org gous bicyclo[4.1.0]heptan-2-ones endo-8c could be rearranged successfully to the desired bicyclic compound 9c in 38 % yield (entry 4). Surprisingly, when the corresponding exo-diastereomer exo-8c was employed under the same conditions, the rearrangement to 9c failed (entry 5). It should be noted, that also the use of only 10 mol-% 16·HCl instead of 20 mol-% in the reaction of endo-8c to 9c led to a decreased yield of 8 % (Scheme S2, Supporting Information). Also, bidentate NHC ligands [26] led to decreased yields compared to the monodentate ligand 16·HCl (for details see Scheme S3, Supporting Information). Contrary to the previous parallel experiments in entries 2-5, gem-phenyl-substituted vinylcyclopropane 8d was employed as (28:72) mixture of diastereomers in the Ni-catalysis, because chromatographic separation of endo-/exo-8d could not be achieved. Fortunately, the desired rearranged product 9d could be isolated in 30 % yield (entry 6). When the endo-and exocyclopentene-substituted vinylcyclopropanes endo-8e and exo-8e were subjected to Ni catalysis, no trace of the desired tricyclic product was detected (entry 7,8). Furthermore, the vinylcyclopropanes 8f with an unsubstituted vinyl moiety did not undergo isomerization, independent of the endo or exo configuration of the vinylcyclopropanes (entries 9,10).
We surmised that the observed difficulties in the Ni-catalyzed VCP rearrangement might be due to the problems associated with the in situ formation of the NHC from the corresponding imidazolium salts and KOtBu. It should be noted that Trnka [27] has reported on a nucleophilic attack of an alcoholate to the imidazolium salt, in which especially imidazolium salts with saturated backbones were found to be sensitive. The resulting formal NHC-alcohol adduct might prevent the formation of the catalyst systems from the free carbene. Therefore, the deprotonation of the NHC precursor 16·HCl and subsequent Ni-catalyzed VCPR of gem-methyl-substituted vinylcyclopropane endo-8c was examined with various non-nucleophilic bis(trimethylsilyl)amide bases (Table 5). The catalyst was prepared in situ by equilibration of 10 mol-% Ni(COD) 2 , 20 mol-% 16·HCl and 20 mol-% LiOtBu instead of KOtBu, prior to addition of substrate endo-8c. But the change of a counterion led to a decreased yield (23 %) of the desired product 9c compared to 38 % for KOtBu (entries 1, 2). When KHMDS was employed as the base the bicyclic product 9c was only formed in 27 % yield (entry 3). The use of NaHMDS or LiHMDS gave even poorer yields (entries 4,5). Consequently, a higher nucleophilic character of the base was even beneficial for this reaction. The results revealed that the in situ formation of the catalytically active species from Ni(COD) 2 , 16·HCl and base had a pronounced influence on the VCPR.
In order to minimize potential interactions of the salts formed during the in situ generation of the carbene ligand, and to avoid decomposition of the free carbene, the generated free NHC SIPr (16) was trapped with CO 2 and converted into the corresponding imidazolium carboxylate 16·CO 2 following the method of Naumann. [28] With this CO 2 adduct in hand, subsequent thermal decarboxylation of 16·CO 2 (20 mol-%) by heating at 120°C in toluene in the presence of Ni(COD) 2 (10 mol-%) and vinylcyclopropane 8c did not give any trace of the desired product 9c. Variations of this method (Scheme S4, see Supporting Information) failed as well. Therefore, the above discussed in situ conditions seem to work best for this reaction.

Computational Investigations
In order to gain more insight into the mechanism of the VCP rearrangement and to rationalize the puzzling results discussed above regarding substrate scope, quantum chemical calculations on the B3LYP-D3-COSMO/def2-TZVP level (with single point corrections using the double-hybrid functional B2PLYP) were carried out (see Computational details).
The catalyst conformation is generally chosen, such that the rings of the aryl groups are orthogonal to the NHC five membered ring. Other conformations are considered unlikely due to the bulky isopropyl groups, however, the isopropyl groups themselves can rotate freely. To ensure consistency in the calculations the start geometries were always chosen in such a way, that the hydrogens which are not part of the methyl groups face one another, pointing towards the center of mass of the molecule. The conformational space for the substrates endo-5b, endo-5c, endo-8b and endo-8c is limited due to the two annelated rings. From the two possible conformations, only the one for which all the transition structures can be found is considered in the following. For further details on this see the Supporting Information.

Computational Investigations Regarding the Catalytically Active Species of the VCPR
One central question is which compound acts as the catalytically active species. If the Ni(COD) 2 exchanges one COD ligand for one NHC ligand, the computed free energy of the complex decreases by 1.5 kJ/mol. An exchange of the remaining COD ligand with another NHC ligand further decreases the free energy by 5.1 kJ/mol, thus Ni(NHC) 2 is the most stable complex in the regarded system. Nevertheless, at finite temperatures all three Ni complexes will be present in significant concentrations and we assume, in analogy with the theoretical results of Wang and Tantillo, [8d] that the Ni(NHC)(COD) complex (II) acts as the catalytically active species.
We note that the energetics of the above ligand exchange reaction are difficult to compute, and strong variations have been found using different density functionals. E.g. for the B3LYP [29] functional, the Ni(NHC) 2 complex (I) is far more stable than Ni(COD) 2 . However, we take confidence in the B2PLYPcorrected [30] results, as test calculations using the high-level PNO-LCCSD(T)-F12 method [31] on a system with a simplified NHC ligand (with N,N′-dimethyl substitution) indicate that the strong energy lowering for Ni(NHC) 2 (I) predicted by B3LYP is incorrect (for details see Supporting Information). The B2PLYP computations are in much better agreement with the highlevel coupled-cluster computations, although the stability of Ni(NHC) 2 (I) might be slightly underestimated with this functional.

Calculated Mechanism of VCPR
Starting on the assumption, that complex (II) is the catalytically active species, possible reaction paths of the Ni-catalyzed VCPRs of endo-5c and endo-8c as well as endo-5b and endo-8b were determined (see Computational Details). Scheme 5 illustrates the proposed mechanism of the Ni-catalyzed VCPR of endo-8c to the desired bicyclo[4.3.0]nonenone 9c as an example. The structures of the intermediates are shown in Figure 1.
Starting from the Ni(NHC) complex II, an η 2 -coordination of the vinyl group of endo-8c to the Ni(NHC) complex II leads to intermediate INT1. Next, INT1 undergoes an oxidative addition, resulting in the vinylmetallacyclobutane species INT2 via a ring opening transition state TS2. The spatial arrangement of the substrate at the catalytic center can be seen in Figure 1, which quite nicely illustrates why only the endo isomers are found to be reactive. For INT1 the endo conformation allows the Ni center to coordinate to the carbonyl group of the substrate and keeps the substrate preoriented for the formation of the vinylmetallacycle. The resulting rearrangement of INT2 leads to the metallacyclohexene INT3, which is supposed to be additionally stabilized by coordination of the Ni center to the carbonyl group. The transition state TS4, in which the Ni-atom moves closer to the C=O bond, leads to INT4 by C-C bond formation under reduction of the Ni center.
The catalytic cycle is then closed by a ligand exchange of endo-9c and by the substrate endo-8c, i.e. going directly from INT4 to INT1. This process is strongly exergonic by 40 to 60 kJ/ mol for all investigated substrates (see below). The transition state of this bimolecular step has not been investigated in this work, but we can safely assume that this does not add any significant barrier to the overall process.
The mechanism presented here qualitatively contains the same major transition state structures as were presented by  Wang and Tantillo. [8d] They report additional intermediates and transition state structures. The absence of those structures in our study might be explained by the presence of bulky ligands leading to a more rigid system. Anyhow, these additional structures should not factor in the kinetics of the reaction since their respective barriers are very low compared to the major barriers. Because of our experimental findings it is evident, that a possible pathway not involving the keto group is not viable under the given reaction conditions. Otherwise structural isomers of the product should be detected. ing half-life estimates for the respective INT1 compounds can be determined. For substrates endo-5c and endo-8c these halflifes lie within the realm of seconds/minutes with approximately 1 s and 440 s respectively, whereas for the substrates endo-5b and endo-8b they are in the realm of days with 5600 h and 140 h. This very likely explains the unsuccessful reaction for the latter substrates in the experiment. It also has to be noted that the overall reactions are expected to be slower than these halflifes suggest, since the rate for the formation of INT1 was not taken into account.

Investigation of the Substrate Dependence
For all investigated substrates, intermediate INT4 has the lowest energy in the catalytic cycle and a subsequent ligand exchange reaction to regenerate INT1 is exergonic. In particular, for substrates endo-5c and endo-8c the exchange is exergonic by -39.0 kJ/mol and -40.4 kJ/mol, respectively, while for the substrates endo-5b and endo-8b values of -60.6 kJ/mol and -52.4 kJ/mol are computed, respectively.
In summary, the proposed mechanism is consistent with the observed reactivities and explains the preference for the geminal substituted vinyl units with the endo-configuration of the cyclopropane ring.

Conclusion
Two series of keto-functionalized vinylcyclopropanes 5, 8 with different substitution patterns at the vinyl unit were synthesized from cyclopentenone or cyclohexenone and sulfur ylides. Compounds 5, 8 were submitted to the catalytic vinylcyclopropanecyclopentene (VCP) rearrangement using Ni(NHC) complexes to obtain the bicyclo[3.3.0]octenone 6 and bicyclo[4.3.0]nonenone 9 respectively. A temperature dependent screening of different Ni(NHC) complexes revealed that the VCP rearrangement is favoured over the competing ring opening towards 3-(2methylprop-1-en-1-yl)cyclopent-2-en-1-one 13c at tempera-tures ≥ 120°C. With regard to the type of NHC ligand, combinations of the aliphatic NHC cores with peripheral bulky aryl units or aromatic NHC units with peripheral bulky alkyl units were found to be beneficial for the VCP rearrangement. Thus, the ligand 1,3-bis(2,6-diisopropylphenyl)-imidazolinium chloride (16·HCl) provided the best compromise between conversion and reactivity. Screening of the substrate scope revealed that vinylcyclopropanes with geminal-disubstituted vinyl units were better suited than vinylcyclopropanes with vicinal-disubstituted vinyl unit, which is in good agreement with the results obtained by Louie and Tantillo [8d,8e] for unfunctionalized vinylcyclopropanes. Moreover, the relative configuration of the cyclopropane unit played an important role for the reactivity, i.e. the endoconfigured vinylcyclopropanes underwent the VCP rearrangement in contrast to their exo-configured counterparts. Quantum chemical calculations of the reaction pathways showed lower energies for the relevant intermediates and transition states of the geminal-disubstituted substrates, exemplary for the endovinylcyclopropanes, supporting their high reactivity. Thus, the combined efforts of computational and experimental studies unraveled the puzzling reactivity issues, substrate scope and stereoselectivity of the Ni(NHC)-catalyzed rearrangement of 1acyl-2-vinylcyclopropanes and delivered a tool for further exploration in the synthesis of complex bicyclic target compounds.

Computational Details
The calculations were carried out with the TURBOMOLE V7.2.1 program package. [32] Resulting structures were visualized with Avogadro 1.2.0. [33] Molecular geometries were optimized using density functional theory (DFT) in conjunction with the B3LYP functional [29] including dispersion effects through Grimme′s D3 correction [34a] with Becke-Johnson damping.
[34b] Numerical in-tegration was carried out on an m3 grid and density fitting (multipole accelerated resolution of the identity) [35] was enabled to speed up the integral evaluation. For all calculations the def2-TZVP basis set [36] was used. Solvent effects were accounted for with the conductor-like screening model (COSMO) [37] using a dielectric constant of ε = 2.07 to mimic the polarity of toluene at 120°C. Finite temperature effects were accounted for using the RRHO (rigid rotor harmonic oscillator) approximation assuming a temperature of 120°C, in accordance with the usual experimental conditions. Vibrational frequencies were scaled by a factor of 1.0044. [38] The Gibbs free energies computed at the B3LYP-D3-COSMO level were corrected by additional single point calculations with the double hybrid functional B2PLYP-D3. [30] The corrected energies were obtained as G = G(B3LYP-D3/COSMO) -E(B3LYP-D3) + E(B2PLYP-D3), where E refers to purely electronic energies. Starting guess structures for transition state searches were obtained with the woelfling program [39] of TURBOMOLE and an in-house version thereof. All transition structures were verified to possess only a single mode with imaginary frequency. IRC (internal reaction coordinate) calculations starting from the transition structures yielded no relevant new intermediate structures on a low level of theory (BP86-D3/def2-SVP). [36,40] The PNO-LCCSD(T)-F12 [31] calculations were done with the quantum chemistry package MOLPRO V2019.2. [41]