|Home | About | Journals | Submit | Contact Us | Français|
A comparative study of the Au(I)-catalyzed [3,3]-sigmatropic rearrangement of propargylic esters and propargyl vinyl ethers is described. Stereochemically defined cyclopropanes are employed as mechanistic probes to provide new synthetic and theoretical data concerning the reversibility of this type of rearrangement. Factors controlling the structure-reactivity relationship of Au(I)-coordinated allenes have been examined, thereby allowing for controlled access to orthogonal reactivity.
Among the large number of reactions disclosed in the past five years in the field of homogenous gold catalysis,1 the Au(I)-catalyzed rearrangement of propargylic esters has recently gained much attention. This interest is primarily due to the accessibility of the starting materials, the mild reaction conditions required for conversion, and above all the very rich chemistry that derives from the two main reaction manifolds, the Au(I)-catalyzed [2,3] and [3,3]-sigmatropic rearrangements.2 Evidence for the existence of Au-carbenoid species in the [2,3]-rearrangement has been gathered in an array of ways that include intra-3 and intermolecular trapping4 including oxidation5 of the carbenoid intermediates. In a closely related field, we have recently reported that propargyl vinyl ethers undergo irreversible Au(I)-catalyzed Claisen rearrangements6 to afford allenes 3 (Scheme 1). These allenes are proposed to result from the Grob-type fragmentation of intermediates 2.
It has been postulated that Au(I)-catalyzed [3,3]-rearrangements of propargylic esters are reversible processes7,8 that take place via cationic intermediates related to 4 and subsequent formation of Au(I)-coordinated allenes 5, which can then undergo further transformations.9 However, the reversibility of the rearrangement is generally assumed, and validation by direct experimental evidence is lacking. Also, the final products arising from 5 depend heavily on the electronic structure of the metal-bound allene. A diverse pool of reactions10,11 has been published that take advantage of substituents enforcing the dominance of specific resonance forms of reactive intermediates. Despite these successes, information about the factors dictating which resonance form dominates and relative rates of competing reaction pathways remains largely empirical. A better understanding of the various reactive intermediates in these systems would provide a more robust foundation for the design of future reaction manifolds.
In an effort to gain insight on the nature of these intermediates, we sought to capitalize on the known tendency of cyclopropyl carbinyl cations to undergo rearrangements.12,13 Phenyl-substituted cyclopropanes have proven valuable mechanistic probes in the identification of cationic mechanisms.14 To test for cationic reactivity in the Au-bound intermediates, we designed substrates incorporating a cis-disubstituted cyclopropyl group at the propargylic position. If species evolved exhibiting significant carbocationic character alpha to the cyclopropane, a fast isomerization to the thermodynamically more stable trans-cyclopropane should result (structures 7 and 8, Scheme 2). Additionally, varying the electronic properties of the cyclopropyl substituents themselves should enable modulation of the reactivity prompted by positive charge build-up in the Au-coordinated allenes or other intermediates.
Computational results are presented here in tandem with the experimental data, a combination often used in modern organic chemistry.15 The theoretical results establish consistency between our specific system and those studied by others, and lend support to many of our mechanistic claims. In this context, the present work describes a comparative study of the Au(I)-catalyzed [3,3]-rearrangement of two different cyclopropyl-substituted systems, propargylic esters and propargyl vinyl ethers.
An initial test reaction on the model substrate 9 showed that the Au(I)-catalyzed rearrangement of the pivaloate in CH2Cl2 afforded, instead of the expected allene, cyclopentene 10 (62%, 75:25 mixture of olefin isomers about the exocyclic double bond) and enyne 11 (32%) after 10 minutes (Scheme 3.1). Employing CH3NO2 as the solvent suppressed formation of 11, affording exclusively the cyclopentene 10 in 75% yield as a mixture of olefins about the exocyclic double bond (86:14 E/Z ratio), albeit with longer reaction times (8 h). In contrast, using C6H6 as the solvent favored formation of the open enyne 11 over 10. While no allene species were detected during these trials, we present strong evidence that 10 is in fact the product of a cyclopropyl ring expansion from an allene intermediate.16 In agreement with our previously reported results,6a and in contrast to the reaction of 9, treatment of vinyl ether 12 in CH2Cl2 with catalytic [(Ph3PAu)3O]BF4 resulted in the exclusive formation of the corresponding allene 13 (Scheme 3.2). The Ph3PAuSbF6-catalyzed reaction also afforded 13 as the sole reaction product, albeit in lower yield (76%).
The use of a stereochemically defined starting material provided an opportunity to probe the reversibility of subsequent steps along the reaction pathway. A key observation emerged from 1H NMR analysis of the reaction mixture during the transformation of 9 to 10: as the reaction progressed, formation of 10 was observed along with gradual scrambling of the stereochemistry of 9 at both propargylic and cyclopropyl positions in the remaining propargyl starting material (eq 1). This scrambling occurred in two separate events. First, the relative stereochemistry at the propargylic position was completely lost after only 2 minutes, suggesting that the first event is a very fast and reversible rearrangement of the pivaloate group. It is known that Au(I) catalyzes the stereoisomerization of allenes6a,6c,17 and thus this scrambling event is likely to proceed through a reversible [3,3]-sigmatropic rearrangement. Second, a slower cis-to-trans isomerization of the cyclopropyl moiety took place over 1 h, leaving the trans-cyclopropyl isomer (as a 1:1 mixture of diastereomers of propargyl ester) as the predominant form of 9 thereafter.18 This result implies that the two scrambling events are mechanistically distinct, and allows for the proposal of an intermediate with carbocationic character which is responsible for the cis/trans cyclopropyl scrambling and which is generated along the propargylic scrambling pathway.
It has been suggested that double [2,3]-acyloxy migrations can also account for net [3,3]-rearrangements.8,9c In order to investigate this assertion, an isotopic labeling study was conducted wherein the 18O-enriched ester 18O-9 was synthesized and subjected to the reaction conditions (eq 2).19 After 8 hours, cyclopentene 18O-10 was isolated in 76% yield. Mass and IR spectra showed that the 18O label resided exclusively at the carbonyl oxygen of 18O-10. Two conclusions can be drawn from this experiment. First, 10 is exclusively generated from a [3,3]-rearrangement, provided that it is formed directly from a Au(I)-coordinated allene. A double [2,3]-acyloxy migration would place the label at the ester linkage in the allene intermediate and thus also in the ester linkage in 18O-10. In apparent contradiction with a previous computational study,8 the fact that the label does not scramble allowed us to determine that, at least in the case reported herein, the proposed double migration is not operative.20 Second, stereochemical scrambling at the propargylic position of the substrate is not caused by ionization of the pivaloate moiety. If ionization had occurred then scrambling at the labeled position should have been observed.21
A computational study (B3LYP/LACVP**, see Supporting Information for details and coordinates) was undertaken to model the reaction path from the nearly un-simplified model propargylic ester complex A1 to allene complex A3 shown in Figure 1. Our studies indicate that A1, A2 and A3 should rapidly interconvert. These results are qualitatively similar to those obtained for a more simplified system studied by Cavallo and coworkers.8 The activation barrier (ΔGSTP‡ in CH2Cl2) corresponding to the cyclization transition state Ats1 was predicted to be 7.5 kcal/mol. Heterocycle A2 and allene A3 appear to be isoenergetic and more stable than A1 by 4.3 kcal/mol.
It is well established experimentally6a,6c and theoretically17 that axially chiral allenes undergo Au-catalyzed racemization. The low-energy structure A4 was predicted to be accessible from A3, which arises from a rotation of approximately 60° about the bond connecting the central allenyl carbon and C1 (resonance forms in Scheme 2). In this configuration the four allenyl substituents are nearly coplanar and thus the allenyl stereochemistry would not be expected to persist. The calculated barrier to rotation from A3 to A4 is only 1.9 kcal/mol (5.7 kcal/mol for the reverse process). In conjunction with the reversibility of the [3,3]-rearrangement, this reactivity profile accounts for the experimentally-observed stereochemical scrambling at the propargyl position of 9, as suggested above.
The trans-cyclopropyl allene complex A3trans was correctly predicted to be slightly more stable than A3. Unfortunately no satisfactory, low-energy pathway was located for the conversion of A3 to A3trans although several possible routes were investigated. Based on literature precedent12 concerning an organic system, we suggest a mechanism commencing with a 1,2-methylene shift in the cyclopropyl fragment to form a cyclobutonium intermediate. However, several species along this pathway could not be located and thus alternative mechanisms should be entertained. This issue and one other candidate mechanism are discussed in the Supporting Information.
To establish appropriate comparisons between reactivity patterns, we performed additional computations employing a model propargyl vinyl ether. As stated above (eq 2), a very clean and fast reaction affording the allenyl cis-cyclopropane 13 was observed experimentally. The computed reaction path for the vinyl ether system B is shown in Figure 2. The predicted activation barrier for the initial cyclization is considerably lower than that for model system A, and the transformation to allene B3 was found to be much more exothermic, as expected for a carbon-carbon bond forming reaction. There is a qualitative difference, however, in that the cyclic structure B2 was not located as a stationary point on the reaction path. Rather, searching along the reaction coordinate from transition state Bts1 revealed B2 as only a shoulder along a concerted pathway to B3 (Figure 2, insert, electronic energy (E) only); no nearby minima or transition states were located. As in model system A, the trans-cyclopropyl isomer B3trans was calculated to be slightly more stable than B3 (by 2 kcal/mol). Additionally, a twisted allene structure (B4) was predicted to be accessible, although in this case it is slightly less stable than B3. Therefore, scrambling of the stereochemistry in the product allene is predicted for this system, although the irreversibility of the [3,3]-rearrangement dictates that no scrambling of stereochemistry at the propargyl position of the starting substrate should be observed. This prediction is consistent with the findings of a previous study of propargyl vinyl ethers.6a
We next examined the impact that arylcyclopropyl groups with different electronic properties might exert on product distribution and stereochemistry. Replacement of the phenyl group of 9 with 4-NO2C6H4 and 4-MeOC6H4 had a dramatic effect on the reactivity of both the ester and vinyl ether systems. Specifically, reaction of the electron-deficient analog 15, bearing a 4-NO2C6H4 group, resulted exclusively in rapid scrambling of the stereochemistry at the propargylic position (1:1 ratio after 10 minutes) and cis/trans cyclopropane isomerization (a 70:30 ratio of cis/trans isomers was obtained, Scheme 4.1). The fact that a decrease in the cation-stabilizing ability of the cyclopropyl substituent results in incomplete scrambling of the cyclopropyl stereochemistry as well as failure to yield a cyclopentene product indicates that positive charge build-up near the aryl group is an important feature of the cis/trans cyclopropyl isomerization as well as the cyclopropyl ring expansion. Specifically, exposure of propargylic ester 15 (bearing a 4-MeOC6H4 group) to Ph3PAuSbF6 afforded cyclopentene 16 in 97% yield after only 10 minutes (Scheme 4.2), indicating that the electron-donating nature of the 4-MeOC6H4 group and its ability to stabilize nearby positive charge build-up greatly increases the reactivity of the system.
In analogy with propargylic esters, propargyl vinyl ethers bearing 4-NO2C6H4 and 4-MeOC6H4 groups on the cyclopropyl moiety were synthesized. In light of the results obtained for substrate 12 (Scheme 3.2), [(Ph3PAu)3O]BF4 was used as the catalyst for this part of our studies. First, reaction of 4-NO2C6H4-substituted 17 led exclusively to quantitative formation of allene 18 with complete retention of the cyclopropyl stereochemistry (Scheme 4.3). Conversely, treatment of the 4-MeOC6H4-substituted propargyl vinyl ether 19 under identical reaction conditions resulted in the very rapid formation of a mixture of species assigned as allenes 20 and 21 (by 1H NMR, Figure 5.4), indicating that cis/trans cyclopropyl isomerization was now operative. Subsequently, the peaks corresponding to 20 and 21 gradually disappeared while those belonging to cyclopentene 22 gained intensity. After 48 h, 22 was the only observed species, existing as a 75:25 mixture of olefin isomers about the exocyclic double bond. This experiment provides direct evidence that the cyclopropyl ring expansion proceeds only from an allene intermediate.22 It is reasonable to posit that this holds for the formation of 10 as well, as proposed above.
The preceding observations reveal that, although the ester and ether systems follow analogous mechanistic pathways, electronic effects play a critical role in the final outcome of the reaction. In the case of propargylic esters, the participation of resonance form 24 (Scheme 5) in the allene intermediate appears to be significant. There apparently is sufficient positive charge build-up at C1 to allow a cis/trans cyclopropyl scrambling process, even in the presence of a strongly electron-withdrawing group (Ar= 4-NO2C6H4) on the cyclopropyl ring. However, it appears that electron-deficient cyclopropyl substitution renders the central allenyl carbon insufficiently nucleophilic for the allene intermediate to undergo pentannulation. Conversely, the electron-donating 4-MeOC6H4 cyclopropyl substitution stabilizes positive charge build-up in the intermediates along the cyclopropyl scrambling pathway while also rendering the allenyl fragment more electron-rich. Thus, the rates of both cyclopropyl scrambling and pentannulation are accelerated.
The principal structural difference between the ester and ether systems is the presence of the allenyl oxygen. Non-activated systems that do not contain an allenyl oxygen, such as 12 and 17, do not experience sufficient positive charge build-up at C1 to undergo cis/trans cyclopropyl isomerization. In other words, the participation of the vinyl-Au(I) resonance form 27 is less important. However, the incorporation of an electron-rich aryl group stabilizes the system to the extent that less cationic character is needed at C1 for the cyclopropyl rearrangements to become operative.
The electronic and structural explanations posited above are supported by computational analysis of the structures of allenes A3 (Figure 3a) and B3 (Figure 3b). The most obvious difference between the geometries of the two structures is the nature of the carbon-Au bonds. While the gold center in A3 is canted strongly towards C2, B3 appears more like a η2-coordinated allene17,23 wherein C1 and C2 both participate in bonding ((C1–C2-Au) = 89°, see Figure 3 for other angles).
A Natural Bond Orbital (NBO) analysis suggested that the bonding differences are due to polarization of the C1–C2 and C2–C3 bonds by the ester oxygen in A3, which has the effect of placing more electron density at C2 (natural charges are −0.41 and −0.22 for A3 and B3, respectively) and in turn increases the ionic character of the C2-Au bond (d(C2-Au) = 2.13 and 2.21 Å for A3 and B3, respectively). The negative charge build-up on C2 of A3 makes that position more nucleophilic, promoting its attack on the cyclopropyl ring (a process that would result in the concerted formation of the pentannulation product from A3/B3). While the localized C1–C2 π bond in A3 is polarized such that 75% of its electron density resides on C2, the analogous polarization is only 63% in B3. Therefore, the donor orbital involved in the allene-Au(I) dative bond of A3 is more like an sp2-hybridized lone pair than the two-center π bond of B3. Surprisingly, C1 recovers enough electron density from the cyclopropyl ring that in both cases its partial (natural) charge is calculated to be near zero; in fact, C1 in B3 bears a slightly negative charge (−0.09). Assuming that the mechanism for cis/trans cyclopropyl isomerization is dependent on a partial positive charge on C1, the lower charge on C1 of B3 accounts for its disinclination toward this mode of reactivity. The C1 natural charges associated with the twisted allenes A4 and B4 are nearly identical at −0.02, suggesting that these species are not involved in the cyclopropyl cis/trans isomerization.
Knowledge gained from the experiments discussed above allowed us to rationally design syntheses that display orthogonal reactivity depending on the electronics of the propargyl starting materials. This was accomplished by replacing the arylcyclopropyl groups of the substrates with vinyl-substituted cyclopropanes. In this scenario, two different fates can be envisioned for a vinylcyclopropyl-allenyl intermediate: formation of the cyclopentene analogous to the products described above or a Cope rearrangement involving a η2-coordinated allene that would lead to the formation of a cycloheptadiene.24 When we examined the behavior of the diastereomeric ester series 28–31 under our standard reaction conditions, we were pleased to observe formation of the corresponding cyclopentenes 32 and 33 in excellent yields and as the sole observed reaction products25 (Scheme 6.1 and 6.2), in agreement with the system reported by Goeke and coworkers.16 The cyclopentenes were obtained regardless of the stereochemistry at the cyclopropyl and propargylic positions. It was also observed that the relative stereochemistry of the styryl moiety was retained in all cases. It is worth noting that a similar reactivity pattern can be obtained for 34, which exhibits a regiochemistry that positions the ester on the opposite side of the allene intermediate (Scheme 6.3).
Conversely, the reaction of vinyl ether 37 with [(Ph3PAu)3O]BF4 as the catalyst led to formation of cycloheptadiene 39 (Scheme 6.4). As proposed above, 37 presumably undergoes a Au(I)-catalyzed Claisen rearrangement to yield allene 38, followed by a thermal Cope rearrangement to form a seven-membered ring. Thus, the activation barrier for the pentannulation is lower than that for the Cope rearrangement from the allenyl esters, but the situation is reversed for the formyl allene which is much less prone to cyclopropyl ring expansion.
We envisioned that if a Au(I) species were involved in this step then induction could be obtained by choosing an appropriate chiral ligand.11d Indeed, treatment of 15 with (R)-DTBM-SEGPHOS(AuCl)2 and AgSbF6 afforded 16, which upon reduction with LiAlH4 in diethyl ether resulted in formation of 40 in 32% ee (eq 3). Although this enantiomeric excess is modest, it provides evidence for trapping of a chiral Au(I) intermediate.
Based on the results presented above, we propose the mechanistic hypothesis outlined in Scheme 7. Starting material A enters the catalytic cycle by coordinating to cationic Au(I), forming alkyne complex B. The [3,3]-migration then commences, passing through cyclic carbocation C. Structure C is a long-lived intermediate in the ester system but only a point along the reaction trajectory in the vinyl ether system. A Grob-type fragmentation of C results in the formation of allene complex D, a process which is irreversible in the vinyl ether case because a strong carbon-carbon bond is formed. Complex D can then fragment yielding free allene E, or if sufficient positive charge build-up is present at C1 it can ring open12 and convert to the thermodynamically-favored D′. A formyl allene with no electron-donating groups on the cyclopropyl ring apparently has insufficient cationic character at C1 for this process to take place. Alternatively, D or D′ can undergo a ring expansion to irreversibly form cyclopentene complex F by a concerted process. The same electronic factors that facilitate conversion to D′ lower the activation barrier to the pentannulation reaction, albeit to varying extents. Finally, cyclopentene G is released and the cationic Au(I) species re-enters the catalytic cycle. Therefore, the phenylcyclopropylpropargyl vinyl ethers are different from the electron-rich ethers and the esters only in that, in the former case, a larger energy barrier for the formation of F is encountered by D, and Au(I) decomplexation to form the free allene E is faster.
In this work, key mechanistic aspects on the nature of the Au(I)-catalyzed [3,3]-rearrangement of propargylic substrates have been disclosed. Specifically, our studies provide experimental evidence for the reversibility of the rearrangement in the case of propargylic esters. In contrast, the Au(I)-catalyzed [3,3]-rearrangement of propargylic vinyl ethers is irreversible and proceeds through a concerted pathway. We have found that the cationic nature of the Au(I)-coordinated allenes formed after the rearrangement and the electron-donating abilities of the allene substituents heavily influences their structure-reactivity, determining the η1- or η2-allene character of Au(I)-coordinated allenes. These findings enable controlled access to Au(I)-coordinated allenes that display orthogonal reactivity patterns. We anticipate that the mechanistic insights provided by this study will enable further developments in this area of catalysis.
We gratefully acknowledge funding from National Institute of General Medical Services support (R01GM073932), Merck Research Laboratories, Bristol-Myers Squibb and Novartis. We thank Johnson Matthey for a generous donation a gold salts. P.M. thanks the Spanish MICINN for a postdoctoral fellowship. Molecular modeling was performed at the UC Berkeley Molecular Graphics and Computation Facility, directed by Dr. Kathleen Durkin and operated with equipment funds from NSF grant CHE-0233882 and donations from Dell. We also thank Dr. David J. Gorin, Gregory L. Hamilton and Nathan D. Shapiro for helpful discussions, and Christine Musich for preliminary experiments.