Gold-Catalyzed Cycloisomerization of 1,5-Allenynes via Dual Activation of an Ene Reaction

Abstract

A tris-triphenylphosphinegold oxonium tetrafluoroborate, [(Ph₃PAu)₃O]BF₄, catalyzes the rearrangement of 1,5-allenynes to produce cross-conjugated trienes. Experimental and computational evidence shows that the ene reaction proceeds through a unique nucleophilic addition of an allene double bond to a cationic phosphinegold(I) complexed phosphinegold(I) acetylide, followed by a 1,5-hydrogen shift.

Introduction

Transition metal-catalyzed allenyne cycloisomerization reactions provide an atom economical entry into polyunsaturated carbo- and heterocycles.^1,2,3 These reactions involving Rh, Pd, or Pt are generally postulated to proceed via metallacyclopentene intermediates, inherently involving an increase in the formal oxidation state of the metal.^1,4 On the other hand, gold(I)-catalyzed enyne⁵ and some related 1,6-allenyne cycloisomerization reactions⁶ are proposed to proceed without a change in the formal oxidation state of the catalyst.⁷ We report a combined experimental and computational investigation of gold(I) catalysis of an allenyne cyclization that proceeds via a unique mechanism involving cationic phosphinegold(I) activation of an in situ generated phosphinegold(I) acetylide.^2,8,9,10

Substrate Scope

Reaction of 1,5-allenyne 2a under typical conditions for the cycloisomerization of 1,5-enynes (1 mol% Ph₃PAuCl, 1% AgSbF₆, CH₂Cl₂, rt)^5c resulted in rapid formation of a complex mixture. Switching the catalyst to the less reactive tris-gold phosphine oxonium complex [(Ph₃PAu)₃O]BF₄ 1 furnished triene 3a in 31% yield. The yield was further improved to 88% when the reaction was conducted at 60 °C in chloroform (Table 1, entry 1). Under these optimal conditions, a variety of 1,5-allenynes participate in the gold-catalyzed cycloisomerization. Substitutions on the tether (entries 1–6) and on the allene moiety (entries 7, 8) are well tolerated.¹¹ Both diastereomers of bicyclo[4.3.0]nonanes 5a-b could be prepared with excellent complete retention of stereochemistry. In addition to cyclopentenes, cyclohexene 11 was produced from the gold-catalyzed cycloisomerization of 1,6-allenyne 10, albeit with diminished yield (entry 9). Of mechanistic importance, non-terminal alkyne substrates were inert under these conditions.^12,13

Table 1.

Scope of Gold(I)-Catalyzed Allenyne Cycloisomerization.

entry	substrate		mol% 1	product(s)	yielda
1		2a R1 = Bn, R2 = R3 = H	5		3a 88%
2		b R1 = R2 = R3 = H	2		b 84%
3b		c R1 = R3 = H, R2 = Ph	2		c 89%
4b		d R1 = R2 = CO2Me, R3 = H	2		d 99%
5		4a cis	1		5a 78%
6		b trans	1		b 70%
7		6	3		7 70%
8c		8	3		64% 2.4:1
9		10	5		11 40%

^aIsolated yields after flash column chromatography. For reaction times, see Supporting Information.

^bReaction run in CH₂Cl₂ at 40 °C.

^cRegio- and stereoselectivies measured by ¹H-NMR.

Deuterium Labeling Experiments

Deuterium labeling experiments reveal that the hydrogen-transfer is stereoselective; the allenic hydrogen is always incorporated syn to the newly formed C-C bond (eq 1). A double-labeled crossover experiment showed no exchange of the deuterium label, suggesting the reaction proceeds through an intramolecular hydrogen transfer (eq 2). The proton at the alkyne terminus, however, does exchange (eq 3). In deuterated methanol, nearly complete anti incorporation of deuterium is observed in the product (eq 4).

(1)

(2)

(3)

(4)

Kinetic Isotope Effect

Two kinetic isotope experiments were performed (eq 5 and 6). The k_CD3/k_CH3 isotope effect (1.89±0.02) and the k_H/k_D isotope effect (1.84±0.15) were essentially identical, suggesting that the formal 1,5-sigmatropic shift is responsible for the measured isotope effects. Additionally, these measured kinetic isotope effects are much smaller than the typical values of 3–5 for 1,5-sigmatropic shifts.^14,15,16

(5)

(6)

Proposed Mechanisms

Numerous mechanistic possibilities exist for catalysis of this formal ene reaction (top, Scheme 1). In fact, no less than nine possibilities emerged during our investigation.

Scheme 1.

Potential Intermediates Involved in the Trimer (1)-Catalyzed Allenyne Cycloisomerization.

The mechanism through B involves an oxidative cyclization of the allenyne to form gold(III)-metallacycle, followed by β-hydride elimination and a reductive elimination. Mechanisms involving A and C are ene reactions of phosphinegold vinylidene and phosphinegold acetylide, respectively. Mechanisms through D and E are ene reactions catalyzed by the coordination of a phosphinegold to either the acetylene or allene¹⁷ in the substrate. Mechanisms F-I are analogous transformations to B-E involving a phosphinegold acetylide.

A mechanism involving double activation by phosphinegold(I) is plausible. This is supported by three experimental observations: i) inertness of non-terminal alkyne substrates under reaction conditions;¹² ii) deuterium exchange at the terminal alkyne position (eq 4); iii) observation of the transient formation of phosphinegold acetylide 2a′ from 2a under the reactions conditions. ¹⁸ We believe this reaction to undergo catalysis via species I.¹⁹

Computational Method

The geometry optimizations and thermodynamic corrections were performed using hybrid density functional theory (B3LYP) with the 6-31+G* and LANL2DZ+ECP basis sets. Solvation corrections for chloroform were computed using the PCM method with single points at the HF level with 6-31+G(d,p) and LANL2DZ basis sets. All calculations were performed using the Gaussian series of programs.²⁰

The computationally investigated reaction was the conversion of the unsubstituted substrate 2b to triene product 3b. Phosphine was used as a model for the triphenylphosphine. Species involving the triphenylphosphine were also optimized to compare experimental and computed KIEs. Computed KIEs between transition structures involving simple phosphines and triphenyl phosphines were indistinguishable in all computed cases.

All relative energies presented in the current manuscript are free energies in kcal/mol, with respect to separated substrate and tris-phosphinegold oxonium catalyst complex.

Possible Substrate-Gold Complexes

The tris-phosphinegold(I) oxonium catalyst 1 can transfer a phosphinegold cation to the alkyne, a process with an unfavorable enthalpy and entropy (ΔG = 37.4 and ΔG_solv = 21.8 kcal/mol, E), although this is likely to be much more favorable with sterically hindered phosphines. The formation of phosphine gold acetylide is very favorable, and subsequent transfer of a second phosphinegold cation is also favorable (Scheme 2).

Scheme 2.

Possible substrate-gold complexes.

Uncatalyzed Reaction – Ene Reaction of a 1,5-Allenyne

The parent uncatalyzed ene reaction involves a concerted C-C bond formation and asynchronous hydrogen transfer (Figure 1). The free energy barrier of this process (ΔG^‡ = 33.0 and ΔG_solv^‡ = 32.3 kcal/mol, TS-15) is typical for pericyclic reactions.¹⁷ The reaction exergonicity is computed to be large (ΔG_rxn = ~ −38 kcal/mol, 16).

Figure 1.

Uncatalyzed reaction – computed reactant, transition structure, and product.²¹

Mechanism via A – Ene Reaction of the Gold Phosphine Acetylide

The ene reaction of phosphinegold acetylide, A, was postulated based on analogy to copper catalyzed pericyclic reactions.²² The coordination of copper(I) to a terminal alkyne reportedly lowers the pK_a of the proton by ~8 pK_a units, allowing for facile formation of copper-acetylides.²² A similar pathway is proposed in the current reaction for the formation of phosphinegold acetylide (see Scheme 2).

Computationally, the ene reaction via A is found to be extremely similar to the uncatalyzed reaction; it features a concerted C-C bond formation and highly asynchronous hydrogen transfer (A-TS-1 in Figure 2). The computed free energy barrier (ΔG^‡ = 35.2 and ΔG_solv^‡ = 34.0 kcal/mol) is approximately equal to that of the uncatalyzed process (ΔG_solv^‡ = 32.3 kcal/mol, TS-15 in Figure 1). These computed barriers suggest that this mechanism is unlikely. Moreover, the computed KIEs for this process (1.25 and 1.32, respectively) did not match the experimentally observed KIEs of ~1.85 (eq 5 and 6).

Figure 2.

Computed reactant, transition structure, and product corresponding to the ene reaction of a gold acetylide A²¹

While the deuterium exchange experiments are consistent with in situ generation of phosphinegold acetylide (eq 3), independent evidence for acetylide participation in the catalytic cycle was sought. To this end, gold phosphine acetylide 2a′ was prepared by formation of the lithium acetylide of substrate 2a and subsequent transmetallation with triphenylphosphinegold(I) chloride (eq 7). ²³ However, as predicted from computations, the exposure of the acetylide 2a′ to reaction conditions without added catalyst failed to furnish the desired products (eq 8).

(7)

(8)

Mechanism via B or F –Metallacycles

Despite extensive computational searching, no metallacycle intermediates similar to B or F could be found. Any attempts to locate these metallacycle intermediates computationally lead to substrate phosphinegold complexes with a preferential mono-ligation to either the alkyne or the allene (17 and 18, Figure 3). Indeed, this is not surprising, considering that, relative to other late transition metals, oxidative additions involving cationic phosphinegold(I) complexes are known to be difficult.²⁴

Figure 3.

Computed intermediates, 17 and 18, resulting from computational efforts to locate metallacycle intermediates B and F, respectively.²¹

A mechanism involving a metallacycle intermediate necessitates a β-hydride elimination step to generate product (Scheme 3). Such a reaction should be very sensitive to the electronic bias of the hydrogen being transferred. An experimental intramolecular Hammett study was designed to probe the nature of the hydrogen transfer. Cyclization of unsymmetric substrate 19 yielded a regioisomeric product distribution of 1.3:1.0, slightly favoring transfer of the hydrogen benzylic to the para-trifluoromethyl group (eq 9). This near equal distribution of isomers is suggestive of electronic neutrality in the hydrogen transfer transition state, and is inconsistent with a mechanism involving β-hydride elimination. Finally, the measured KIEs of ~1.85 (eq 5 and 6) are inconsistent with transition metal catalyzed reactions that are proposed to proceed via metallacycle intermediates.^25,26

Scheme 3.

Hypothetical Au(III) metallacycle mechanism

(9)

Mechanism via C or G – Gold Vinylidenes

The gold vinylidene intermediate, C, was found to be extremely unstable (ΔG = 52.9, ΔG_solv = 36.7 kcal/mol, Figure 4). Since this intermediate is higher in energy than the transition structure of the uncatalyzed reaction (Figure 1), a mechanism involving C is not operative. The gold vinylidene of the gold acetylide, G, is not a minimum and calculations lead to spontaneously rearrangement to a structure corresponding to intermediate I.

Figure 4.

Computed vinylidene intermediate C, and the intermediate I arising from computational efforts to locate the gold acetylide vinylidene intermediate G.²¹

Mechanism via D or H – catalysis by gold phosphine coordination to the allene π-bond

The computed structures corresponding to the transformations arising from the phosphinegold coordination to the allene of the substrate (D, Scheme 1) are shown in Figure 5.¹⁷ This coordination mode leads to a concerted C-C bond formation and asynchronous hydrogen transfer, similar to the uncatalyzed process. The free energy of activation for this process is prohibitively high (ΔG^‡ = 63.7, ΔG_solv^‡ = 50.1 kcal/mol, D-TS-1), due to the development of an unstabilized vinyl cation in the transition state. In fact, the computed barrier is much greater (ΔG_solv^‡ > 50 kcal/mol) than the uncatalyzed reaction (ΔG_solv^‡ = 32.3 kcal/mol, TS-15 in Figure 1).

Figure 5.

Computed intermediates and transition structures corresponding to the ene reaction of Au(I) phosphine allene coordination complex D.²¹

The analogous computed structures corresponding to mechanism via H (Figure 6), which involves phosphinegold coordination to the allene of a phosphinegold acetylide, closely resemble those involving D (Figure 5); it also features a concerted C-C bond formation and asynchronous hydrogen transfer (H-TS-1). However, this process is much more facile because it does not incur the energetic penalty of transferring a single phosphinegold cation (D versus H, Scheme 1). This mechanism is still unlikely to occur because the free energy of activation (ΔG_solv^‡ = 31.4 kcal/mol, H-TS-1) is very similar to the uncatalyzed process (ΔG_solv^‡ = 32.3 kcal/mol, TS-15 in Figure 1).

Figure 6.

Computed intermediates and transition structures corresponding to the ene reaction of gold phosphine coordination to the allene of a gold acetylide, H.²¹

Mechanism via E or I – Catalysis by gold phosphine coordination to the alkyne π-bond

The cyclization involving phosphinegold alkyne complex E occurs via a stepwise process. The C-C bond formation (E-TS-1-syn and E-TS-1-anti) following the phosphinegold coordination leads to a relatively stable vinyl cation intermediate in which the gold is either syn or anti to the forming C-C bond (E-2-syn and E-2-anti, respectively). The subsequent hydrogen transfer (E-TS-3-syn and E-TS-3-anti) from an adjacent allenyl methyl group to the nascent allylgold furnishes the product phosphinegold complex E-4.²⁷

The anti coordination mode is favored over the syn for the C-C bond formation by ~7 kcal/mol. Steric repulsions will cause the anti isomer to be even more strongly favored with larger alkylphosphine ligands. While subsequent hydrogen transfer is more favored for the syn than the anti isomer, the isomerization of the vinyl gold intermediate E-2 from the anti arrangement to the syn is extremely difficult (ΔG^‡ = 44.4 kcal/mol). The anti C-C bond formation (E-TS-1-anti) and subsequent hydrogen transfer process (E-TS-3-anti) exhibit free energy barriers of ~28 kcal/mol. It is interesting to note that most of barrier for the C-C bond formation arises from the energetic penalty of transferring a single phosphinegold cation (E versus I, Scheme 1) – E-TS-1-anti is essentially barrierless from E.

The greater stability of the alkyne coordination pathway (E in Scheme 1) as compared to the allene coordination (D in Scheme 1 and D-TS-1 in Figure 5) stems primarily from the stabilization of the developing allylic substituted cation versus a vinyl cation.

A mechanism involving intermediate E appears viable, although still relatively slow. We believe it does not actually occur because; i) this mechanism fails to account for the inertness of non-terminal alkyne substrates;¹² ii) the formation of the C-C bond is predicted to result in the formation of a relatively stable vinyl gold intermediate (E-2-anti) that should be observable. Despite extensive trials to trap this intermediate via addition of methanol to the reaction mixture, products corresponding to this intermediate could not be isolated;^6b iii) the computed KIEs for this mechanism (1.02 and 4.04, for eq 5 and 6, respectively) did not fit the experimentally observed KIEs of ~1.85.

Although the acetylide 2a′ was bench top stable in both solution and solid state, addition of trimer 1 afforded triene 3a (eq 10). Use of PPh₃AuOTf also proved successful when applied toward the pre-formed acetylide 2a′. Even simple Lewis acids such as ZnCl₂ or Brønsted acid elicited product formation.²⁸

(10)

As previously stated, the use of PPh₃AuCl/AgOTf leads to the formation of a complex mixture. In light of this, the cyclization of 2a′ by Brønsted acid catalysis shown in equation 10 is curious because this should amount to an equivalent process. However, complex 1 contains a basic oxo ligand that promotes deprotonation to form the gold acetylide (Scheme 2). Additionally, in the absence of formation of the phosphinegold acetylide, the phosphinegold cation may lead to an unselective reaction, perhaps due to pathways involving activation of the allene.^29,30 It is conceivable that the phosphinegold could also catalyze the degradation of the allene in phosphinegold acetylide intermediate D, but we observe selective formation of the desired triene products. Additionally, while there is little difference in the preference for phosphinegold to coordinate to an alkyne or allene,³⁰ there is a strong preference (~22 kcal/mol) for coordination to an acetylide over an allene.³¹

The computed transformations following phosphinegold coordination to the alkyne of a phosphinegold acetylide (I, Scheme 1) are shown in Figure 8. This process is stepwise, similar to mechanism via D (Figure 7). The C-C bond formation (I-TS-1) has a reasonable free energy barrier (~21 kcal/mol), leading to a gem-diauraalkene intermediate (I-2). The free energy barrier for the subsequent 1,5-sigmatropic hydrogen transfer (I-TS-3) is slightly lower (~20 kcal/mol). The reaction is exergonic by −45.5 kcal/mol.

Figure 8.

Computed intermediates and transition structures corresponding to the ene reaction of phosphinegold cation coordination to the alkyne of a phosphinegold acetylide, I²¹

Figure 7.

Computed intermediates and transition structures corresponding to the ene reaction of Au(I) phosphine alkyne coordination complex E.²¹

The structure of intermediate I-4 requires some elaboration. The computed structure of this intermediate features a three-centered two electron bond (Figure 9). The structure is best regarded as a vinyl anion coordinated by two phosphinegold cations.

Figure 9.

Side view of intermediate I-4²¹

The experimentally observed diastereoselective incorporation of hydrogen (eq 1) and anti incorporation of solvent deuterium (eq 4) are explained by the following: i) the 1,5-sigmatropic shift (ITS-3) is stereospecific; ii) subsequent isomerizations around the double bond of intermediate I-4 is hindered by the steric bulk of the gold phosphines; iii) protodemetallations are known to occur stereospecifically.^6c Additionally, the computed KIE profile of mechanism I are 1.8 and 2.0, respectively for eq 5 and 6, which are in excellent agreement with experimentally observed KIE of ~1.85 for both cases.³²

The steady state catalytic cycle of this reaction is summarized in Scheme 4. The coordination of a phosphinegold cation to the in situ generated phosphinegold acetylide lowers the alkyne LUMO toward 5-endo-dig cyclization, generating a gem-diauraalkene and tertiary allylic carbocation. A subsequent 1,5-hydrogen shift followed by protodemetallation and transfers of gold phosphines to another substrate closes the catalytic cycle (Scheme 4).

Scheme 4.

Mechanism involving I (Scheme 1)

The free energy profile for the steady state catalytic process via I is shown in Figure 10. The transfer of two phosphinegold cations from I-4 to form species I and the release of product is exergonic by 10 kcal/mol. Efforts to trap the gem-diauraalkene I-2 have failed, presumably because the subsequent hydrogen transfer is extremely fast (ΔG_solv^‡ = 12.6 kcal/mol) from this intermediate. The most difficult barrier of the catalytic cycle shown in Figure 10 is the C-C bond formation step (I-TS-1) with an activation free energy of 21.6 kcal/mol. Unfortunately, the overall free energy barrier for the catalytic cycle cannot be computed because the transition structures of the phosphinegold acetylide formation could not be located in the gas phase and the actual barrier heights are dependent on the unknown instantaneous concentrations of every species in the reaction mixture. Regardless, the observation that the deuterium labels at the alkyne terminus tend to wash out in the product, while the starting material showed no comparable decrease suggests that the formation of the phosphinegold acetylide is an irreversible process and possibly rate-limiting (eq 11).

Figure 10.

Free energy progress profile for two cycles in the mechanism via alkyne coordination to phosphinegold acetylide (I, Scheme 1). Free energies computed using gas phase standard state; Dashed lines signify transfer of phosphinegold cations from products to reactants.

(11)

Conclusions

In summary, we have presented a Au(I)-catalyzed rearrangement of allenynes to cross-conjugated trienes. Combined experimental and computational evidences reveal that a mechanism involving nucleophilic addition of an allene double bond to a phosphinegold-complexed phosphinegold acetylide is more likely than oxidative cyclization or the simple nucleophilic addition to phosphinegold-complexed substrate.