Mechanism of Gold(I)-Catalyzed Rearrangements of Acetylenic Amine-N-Oxides: Computational Investigations Lead to a New Mechanism Confirmed by Experiment

Abstract

Quantum mechanical studies of the mechanism of gold-catalyzed rearrangements of acetylenic amine-N-oxides to piperidinones or azepanones have revealed a new mechanism involving a concerted hetero-retroene reaction, formally a 1, 5 hydrogen shift from the N-alkyl groups to the vinyl position of a gold-coordinated methyleneisoxazolidinium or methyleneoxazinanium. Density functional calculations (B3LYP, B3LYP-D3) on the hetero-retroene mechanism reproduce experimental regioselectivities and provide an explanation as to why the hydrogen is transferred from the smaller amine substituent. In support of the proposed mechanism, new experimental investigations show that the hydrogen shift is concerted and that gold carbenes are not involved as reaction intermediates.

Introduction

Cationic gold catalysts activate π bonds for reactions with a range of nucleophiles, leading to sigmatropic rearrangements, migrations and cycloisomerizations.¹ A particularly important class of nucleophiles is oxidant that possesses a nucleophilic oxygen and can formally deliver the oxygen atom during the reaction, and this type of gold-catalyzed alkyne oxidation has been proposed to generate a reactive α-oxo gold carbene intermediate, which would undergo an array of versatile transformations (Equation 1).² Early studies in this area use tethered oxidants such as sulfoxide, ³ nitrone, ⁴ epoxide, ⁵ nitro, ⁶ and amine N-oxide, ⁷ and recently one of us has used pyridine/quinoline N-oxides as external oxidants to achieved intermolecular alkyne oxidation.⁸ While α-oxo gold carbenes have been invoked as reactive intermediates in many of these studies, studies using external arylsulfoxides⁹ argue against such intermediacy. Here we present computational and experimental studies to suggest that gold carbene intermediates are not involved in two intramolecular cases; moreover, we disclose the first example of a gold-catalyzed reaction involving a hetero-retroene reaction.

(1)

One of our groups reported the gold-catalyzed annulation to form piperidin-4-ones or azepan-4-ones (Equation 2).⁷

(2)

These reactions were hypothesized to proceed via the mechanism in Scheme 1. The amine oxides, 1, are prepared by reactions of amines with butynyl or pentynyl tosylates followed by oxidation with mCPBA. Coordination of 1 with gold(I) catalyst is proposed to form 3. Nucleophilic attack on the gold-activated alkyne, 3, forms 4. Intermediate 4 was thought to ring-open to form a gold carbenoid intermediate, 5. This intermediate undergoes a 1, 6-hydride shift or 1, 7-hydride shift to form the iminium intermediate 6, which leads to the product, 2. While the proposed mechanism accounts for the formation of the major product, it fails to explain the selective hydride migration from one of the N-alkyl substituents.

Scheme 1.

Proposed mechanism⁷ for the rearrangements of acetylenic amine-N-oxides to piperidinones and azepanones. When n = 1 piperidinones are formed and when n = 2 azepanones are formed.

That is, when the amine oxide has two different substituents, the hydrogen shift occurs selectively from the smaller group. The transformation shown in Equation 3^7a yielded a 5:1 ratio of products (8:9). The longer chain substrate, 10, in Equation 4, ^7b yielded a single product, 11. Acidic hydrogens α to an ester group appear to migrate either less effectively (n = 2) or not at all (n = 1) (Eq. 5).^vii

(3)

(4)

(5)

We now present computational evidence for a different mechanism that accounts for the selection observed in these reactions. A novel hetero-retroene reaction is identified.

Computational Methods

Calculations were performed with Gaussian09¹⁰ with a hybrid density functional (B3LYP and B3LYP-D3¹¹) with LANL2DZ and 6-31G(d) basis sets for gold and for all other atoms, respectively. Structures were optimized with the CPCM model for dichloromethane (DCM) solvent. The gold catalyst ligand, PPh₃ or P(2-biphenyl)Cy₂, was modeled by PH₃. The coordination of the gold to either the N-oxide or alkyne as well as the syn addition to the alkyne were additionally modeled with PPh₃ and gave comparable results to that of the PH₃ ligand. Modeling PPh₃ with PH₃ is very common¹² because it gives good structures and approximates reaction barriers.^{12d, 13} B3LYP has recently been shown to give reasonable energy and geometry predictions for homogenous gold catalysis.¹⁴

Results and Discussion

Two different mechanisms for the conversion of 13 to 17 were investigated; the original proposal is shown in black and the lowest energy path described here is shown in blue in Scheme 2. The cationic gold catalyst preferentially coordinates to the N-oxide to give 13. The alkynyl-gold complex, 20, is 15.6 kcal/mol higher in energy than 13 (Scheme 3). Calculations with PPh₃ as the gold ligand predict that coordination of gold to the N-oxide is favored by 15.4 kcal/mol over coordination to the alkyne. The syn addition to the alkyne from 13 has a barrier of 24.7 kcal/mol (TS13-14). The reaction is exothermic by 24.9 kcal/mol and gives intermediate 14. Anti addition has been established for a variety of nucleophilic additions to alkynes;¹⁵ however, here the N-oxide strongly coordinates to the cationic gold, and no transition structure for the anti addition could be located (Scheme 3).¹⁶

Scheme 2.

B3LYP/6-31G(d)/CPCM(DCM) (B3LYP-D3) energetics for the conversion of the gold-coordinated N-oxide, 13, to cyclic ketones 17 and 19. Compound 19 is not observed experimentally. (t) indicates the species is a ground-state triplet. *gas phase optimization with a single point solvation correction.

Scheme 3.

Anti addition pathway.

If the N-oxide were protonated by m-chlorobenzoic acid (m-CBA), the gold cation would be less likely to coordinate to the OH group. However, calculations indicate that proton transfer from m-CBA to the N-oxide, 1, is endothermic by 7.3 kcal/mol in DCM.

Along the proposed mechanistic path, the ring-opening of 14 has a barrier of 25.6 kcal/mol and leads to a triplet carbenoid, 15. This triplet acyl carbenoid is stabilized by conjugation with the carbonyl, while the singlet (5.9 kcal/mol higher in energy) is less stabilized.¹⁷ Alpha-carbonyl carbenes are ground state triplets and adopt a planar geometry.¹⁸ Alkyl gold carbenoids have been involved in other gold(I)-catalyzed reactions and are predicted to be singlets.^{12c, 19}

Intermediate 15 can undergo a 1, 6-hydride shift with a barrier of 29.6 kcal/mol (TS15-16) to form 16. The barrier for the 1, 4-hydride shift (TS15-18) is 4.6 kcal/mol higher in energy than that for the 1, 6-hydride shift.

An alternative to the mechanism involving the sequential ring-opening and 1, 6-hydride shift was found to be energetically much more favorable. This new mechanism involves a hetero-retroene reaction and leads directly from 14 to 16, thus avoiding the gold-carbenoid intermediate, 15, altogether. This step is favored over ring-opening by 18.6 kcal/mol. Cyclization of 16 forms the product, 17, and regenerates the catalyst. This final step was calculated to have a barrier of 35.8 kcal/mol. This barrier should be lowered by the presence of an NTf₂⁻ counterion or the coordination to a molecule of the N-oxide starting material, 1.²⁰ Also, the actual aryl phosphine ligands used experimentally are more electron-rich than our model phosphine, which should facilitate the dissociation of the gold catalyst. This step is exothermic by 23.9 kcal/mol.

Scheme 4 summarizes the lowest energy path from 13 to 17. In this path, the first step is the syn addition of the gold-coordinated N-oxide, 13, to form 14. Intermediate 14 undergoes the hetero-retroene reaction, and forms 16. The final step is the cyclization of 16 to yield the piperidine product, 17, and regenerate the catalyst. This cyclization is calculated to be the rate-determining step. The overall transformation is exothermic by 96.7 kcal/mol.

Scheme 4.

B3LYP-D3 free energy diagram for the lowest energy path from 13 to 17. Carbon is grey; nitrogen is blue, oxygen is red, gold is gold and phosphorus is orange.

The role of the gold catalyst in the hetero-retroene reaction was probed by comparing the gold-catalyzed reaction to one without the AuPH₃ group. Figure 1 shows the transition structures for the reaction of 14 and of 22. In 22 AuPH₃ is replaced by H. While the reactions are equally exothermic, the barrier is 9.5 kcal/mol lower in energy with the gold catalyst than without. In 14 the cationic gold-phosphine makes the adjacent carbon more negative compared to 22. Because the carbon in 14 is more nucleophilic, the hydrogen transfer is favored, resulting in a lower energy transition structure.

Figure 1.

Comparison of the hetero-retroene reaction with and without gold including bond distances and Mulliken charges. The ΔG_rxn is −46.9 and −48.4 kcal/mol with and without gold, respectively. Free energies determined by B3LYP-D3.

As indicated earlier, experiments showed that the hydrogen is transferred from the smaller amine substituent.⁷ This observation is consistent with the newly proposed mechanism. Figure 2 shows the hetero-retroene transition structure, TS14-16. The least sterically hindered position on the amine substituent (highlighted in green) is not involved in the hetero-retroene reaction. A quantitative study of this phenomenon was undertaken as well. The acetylenic amine-N-oxide, 7, under experimental conditions afforded 8 and 9 in a 5:1 ratio (Equation 3). This product distribution corresponds to the hetero-retroene reaction preferentially involving hydrogen shift from the methyl group rather than the octyl substituent. Calculations on the N-methyl, N-ethyl substrate gave a 1.0 kcal/mol preference for TS7-8, corresponding to a 5.4:1 product ratio (Figure 3).

Figure 2.

The hetero-retroene transition structure, TS14-16. The highlighted (green) hydrogen is the at least sterically hindered position on the N substituents.

Figure 3.

Hetero-retroene reaction transition structures dictating the product distribution of 8 and 9. Highlighted hydrogens in TS7-9 are 2.28 Å apart. Free energies are determined by B3LYP-D3.

Similar mechanistic and regioselectivity studies were carried out for the annulation to azepanones. The mechanism for the conversion of 23 to 27, including competing pathways, is shown in Scheme 5. Again, the gold preferentially coordinates to the N-oxide rather than the alkyne, and no transition structure corresponding to the anti addition could be located (Scheme 6). When PPh₃ is used, calculations predict that coordination of gold to the N-oxide is favored by 10.0 kcal/mol and the barrier for the syn addition is 11.1 kcal/mol. The hetero-retroene mechanism is favored over the mechanism involving stepwise ring-opening followed by hydride shift.

Scheme 5.

B3LYP/6-31G(d)/CPCM(DCM) (B3LYP-D3) energetics for the conversion of the gold-coordinated N-oxide, 23, to piperidinone, 27 and 29. Compound 29 is not observed experimentally. (t) indicates the species is a ground-state triplet.

Scheme 6.

Anti addition pathway.

From the gold coordinated N-oxide intermediate, 23, the syn addition to the alkyne forms 24. Intermediate 24 can either ring-open or undergo a hetero-retroene reaction. Analogous to the annulation to piperidinones, the hetero-retroene transition structure, TS24-26, is favored, but only by 1.7 kcal/mol over the ring-opening. This indicates that a small amount of the gold carbenoid, 25, could be formed, although the necessity for intersystem crossing will disfavor this pathway. Intermediate 25 undergoes a 1, 7-hydride shift, rather than a 1, 5-hydride shift (ΔΔG‡ = 17.8 kcal/mol). Iminium intermediate, 26, leads to the product 27. Again, the barrier for the cyclization is overestimated by our calculations. The transformation from 23 to 27 is exothermic by 85.9 kcal/mol. The free energy diagram for the lowest energy pathway is shown in Scheme 7.

Scheme 7.

B3LYP-D3 free energy diagram for the lowest energy path from 23 to 27.

Calculations are in good agreement with the experimental regioselectivity observed in this transformation as well. The hetero-retroene transition structure, TS24-26, is in a chair conformation with the substituent involved in the rearrangement in an axial position (Figure 4). This is consistent with the experimental observation that smaller groups are involved in the hydrogen shift.

Figure 4.

The hetero-retroene transition structure, TS24-26, for the transformation from 24 to 26. The highlighted hydrogen is at the least sterically hindered position on the N substituents.

The acetylenic amine-N-oxide, 10, under experimental conditions afforded only 11 (Equation 4). As shown in Figure 5, the activation energy of the hetero-retroene reaction involving the methyl group is 2.2 kcal/mol lower in energy than that involving the ethyl substituent. The prediction is that a 43:1 ratio of products should occur.

Figure 5.

The hetero-retroene reaction transition structures for the formation of 11 and 12. The highlighted hydrogens in TS10-12 are 2.10 Å apart. Free energies are determined by B3LYP-D3.

The annulation to azapanones is more regioselective than the annulation to piperidinones (Equations 3 and 4), because the axial/equatorial difference is larger in the chair-like transition structure, (TS10-12, Figure 5) than in the envelope transition structure, (TS9-11, Figures 3). The distances between the ethyl substituent and the nearest hydrogen (both highlighted in green) in TS10-12 and in TS9-11 are 2.10 and 2.28 Å, respectively.

Experimental Studies

These computational studies indicate that the hydrogen migration is intramolecular and the intermediacy of gold carbene intermediates of type 5 (Scheme 1) is unlikely. Experiments were performed to provide further support for these conclusions and thereby the calculated hetero-retroene mechanism.

(6)

To examine the hydrogen migration, the α-methylene group in the tertiary amine 28-d₂ was fully labeled by deuterium. When compound 28-d₂ was subjected to the oxidation and gold catalysis sequence, little deuterium loss was detected in the N-benzylazepanone product 29-d₂ (Equation 6), suggesting that the hydrogen/deuterium migration is intramolecular.²¹ As more than 1 equivalent of m-CBA was present in the reaction mixture during the gold catalysis, an intermolecular ionic process would most likely lead to substantial loss of deuterium. This result is consistent with the hetero-retroene mechanism.

(7)

(8)

To offer evidence against the intermediacy of gold carbene 5 or related carbenoid intermediates, the diazo ketone 31 was prepared²² and treated with Ph₃PAuNTf₂ in 1, 2-dichloroethane. The expected azepan-4-one product 33 was not formed (Equation 7); instead, the Wolf rearrangement²³ product, amino acid 34, was formed in 96% NMR yield. The addition of stoichiometric amount of m-CBA did not alter the reaction outcome. With 32, similar results were observed; amino acid 34, but not the piperidine-4-one product, was observed. When MeOH was used as solvent, the corresponding methyl ether was isolated in 70% yield (Equation 8). While there is no previous report on gold-catalyzed Wolff rearrangements, the related Ag catalysis²⁴ is known, but the mechanistic role of the metal is not well understood. It is commonly assumed, though, a carbene or carbenoid intermediate might be involved. In these clean Wolff rearrangements, either gold carbenoid 36 or gold carbene 37 is the mostly likely intermediate, and the 1, 2-alkyl migration could either be in concert with or follow the expulsion of N₂. This migration was seemingly highly facile as the tethered tertiary amine moiety did not cyclize to the electrophilic carbene/carbenoid center. Since the gold carbene 37 would be highly electrophilic (vide infra), this unexpected preference for the Wolff rearrangement suggests that concerted 1, 2-alkyl migration and nitrogen expulsion from carbenoid 36 are more likely.

(9)

An alternative approach to access gold carbene of type 37 was then pursued. One of us has recently developed a facile access to putative α-oxo gold carbenes via gold-catalyzed intermolecular oxidations of alkynes using pyridine/quinoline N-oxides as oxidants.⁸ By using this strategy, intermediate 39, with its gold carbene moiety similar to that of 37, was most likely formed from the substrate 38 in the presence of MsOH (2 equiv, Equation 9). The reaction, however, did not afford azepan-4-one 42 but instead α-chloro ketone 40 and α-methanesulfonyloxy ketone 41. The formation of 40 is a strong indication of the formation of a highly electrophilic intermediate such as gold carbene 39 as it abstracts a chloride from the solvent, dichloroethane.²⁵ This chloride abstraction is supported by the observation of 2-chloroethyl mesylate, which accounts for the remaining part of the solvent molecule. When m-CBA (1.2 equiv) replaced MsOH, no oxidation of the C-C triple bond occurred.

These studies argue against the intermediacy of gold carbene 5 (Scheme 1) and are fully consistent with the concerted hetero-retroene reaction predicted computationally.

Conclusion

A hetero-retroene mechanism for the rearrangements of acetylenic amine-N-oxides to piperidinones and azapanones has been identified. The experimental regioselectivities are reproduced by the computations of activation barriers for the hetero-retroene reaction. Experimental studies support the proposed mechanism. This is the first case in which a retroene reaction of a gold-coordinated intermediate has been identified. This type of mechanism may be important in other gold-catalyzed alkyne oxidations.