Palladium hydride promoted stereoselective isomerization of unactivated di(exo)methylenes to endocyclic dienes.

The exomethylenes of 2,6-disubstituted bicyclo[3.3.1]nonan-9-ones 2 are readily isomerized over a palladium catalyst under an atmosphere of hydrogen to predominantly form the isomer 3 with C2 symmetry with very little formation of the analogous product with C(s) symmetry. A hydrogen source is essential to effect the rearrangement.

The isomerization of alkenes using catalytic amounts of transition metals and their complexes has been widely studied.[1] For example, allyl units bearing heteroatoms, e.g., allylic amines, alcohols, and ethers, can be easily converted to their propenyl isomers.[2] Allyl arenes can likewise be transformed into the alkylstyrene isomers.[3] One can also effect an equilibration between E and Z alkene isomers.[4] However, there has been little research on the isomerization of unfunctionalized alkenes. Of the few accounts reported, many use large transition-metal complexes, numerous additives, high temperatures, and long reaction times.[5] For the synthesis of rugulosone, 1 (Figure 1),[6] we required a simple protocol to prepare the bicyclo[3.3.1]nonadiene core and investigated the isomerization of a symmetric di(exo)methylene to the endocyclic diene. We tested this process on the simple analogue, 1,5-dimethyl-3,7-dimethylenebicyclo[3.3.1]nonan-9-one 2a. The required dienone 2a was prepared in one step by the tetraalkylation of 3-pentanone with 1-chloro-2-(chloromethyl)-2-propene. We report here the successful isomerization of this diene 2a to the desired C2-symmetric bicyclo[3.3.1]nonadienone 3a and the isomerization of related analogues.

Figure 1

Structure of rugulosone.

The double rearrangement of this substrate 2a to the C2 isomer 3a rather than the C isomer 4a (Scheme 1) would be required to give the natural product. We believed that 3a would be more stable than 4a due to the steric nonbonded interaction of the indicated hydrogens in 4a (Figure 2). Initial attempts to directly isomerize diene 2a with transition-metal catalysts such as Wilkinson’s or Crabtree’s catalyst failed to yield either 3a or 4a; only starting material was recovered (Scheme 2). The uniquely strained and/or hindered structure of the bicyclononane core may cause this lack of reactivity. Unable to directly isomerize the olefins, we examined a longer, more complicated process. Epoxidation of 2a (DMDO, 23 °C, 3 h) afforded the diepoxide 5 in 89% yield. All attempts at acid- or base-promoted ring-opening of these epoxides to give allylic alcohols failed, as did attempts to prepare the tertiary alcohols through various hydride reductions.

Scheme 1

Isomerization of 2a To Give 3a and/or 4a

Figure 2

Structures of the C2 and C dienes.

Scheme 2

Attempts To Isomerize 2a

The desired isomerization was, however, effected by the use of an activated palladium catalyst mixed with hydrogen gas. Thus, treatment of 2a in methanol with 4 mol % of Pd/C under a balloon of hydrogen afforded the desired C2 isomer 3a along with the monoreduced isomerized product 6a and the fully reduced material 7a in a 68:28:4 ratio (by GCMS) (Scheme 3). A number of control experiments were conducted to demonstrate that hydrogen was required for this isomerization. The use of an argon atmosphere instead of hydrogen afforded no isomerization. Pretreating the catalyst with hydrogen gas to activate the palladium and then purging the flask with argon gave no isomerization. The addition of excess cyclohexene (to remove all the H2 gas) before the addition of 2a was also unsuccessful. The use of ammonium formate (transfer hydrogenation) also produced the expected isomerization. Thus, it seems that the reaction requires a small amount of hydrogen to initiate the isomerization.[7]

Scheme 3

Isomerization of 2a To Give 3a

We postulated that alkene isomerization might be favored over reduction if the hydrogenation pathway could be slowed down, perhaps by a change in the solvent, since solvent effects on hydrogenation rates have been well studied.[8] For that reason, various solvent systems were examined to see if solvent effects could improve the yield of isomerization (Table 1). Polar, protic solvents, such as methanol and 2-propanol, gave large amounts of the monoreduced product 6a and some of the fully reduced material 7a. Reaction in ethyl acetate gave the desired C2 product 3a along with the first observation of the formation of the C isomer 4a and the monoreduced product 6a. Isomerization in other nonpolar solvents, e.g., hexane, gave similar results. Other aprotic polar solvents, e.g., acetone, dioxane and THF, also largely produced the desired C2 product 3a along with small amounts of the C isomer 4a and the monoreduced compound 6a.

Table 1

Effect of Solvents on Conversion of 2a to 3a, 4a, 6a, and 7a

entry	solvent	time (h)	3a	4a	6a	7a	2a
1	MeOH	1	68	0	28	4	0
2	iPrOH	1	68	0	26	6	0
3	EtOAc	1	87	9	4	0	0
4	hexane	1	81	7	12	0	0
5	acetone	1	79	6	14	0	0
6	dioxane	1	64	6	11	0	19
7	THF	1	77	9	13	0	1

We explored whether the isomerization to the C2 isomer in great preference to the C isomer was general. The additional di(exo)methylene compounds, 2b–d, were prepared from the corresponding substituted ketones and the bis(chloromethyl)ethylene. Treatment of both 2b and 2c under the conditions described above, namely Pd/C under an atmosphere of hydrogen gas, afforded predominately the C2 products 3b and 3c in preference to the possible C product (Scheme 4). However, the diphenyl-substituted analogue 2d gave only starting material under these conditions with no production of any isomeric or reduction products. Thus, this preference for the C2 product 3 rather than the C product 4 occurs for all alkyl substituents at the bridgehead carbons.

Scheme 4

Isomerization of Alkenes 2bc

The proposed mechanism is shown in Scheme 5. Coordination of the di(exo)methylene 2 with palladium to give A, hydride addition to form B, β-hydride elimination to afford C, and then decomplexation converts 2 to the monoisomerized product D, which has never been observed. The same type of process can convert D, via the intermediates E–H, to either the C2 isomer 3 or the C isomer 4 and can interconvert these isomers as well. All processes are reversible, and the product ratio is most likely determined by thermodynamic stabilities.[9]

Scheme 5

Proposed Mechanism for the Isomerization of 2

In order to establish the energies of each of the isomers, we calculated the structures and energies of reactants, the monoisomerized species, and the isomeric di-isomerized species with density functional theory. Using Gaussian 09,[10] optimizations were performed using B3LYP/6-31G(d),[11] followed by M06-2X/6-311+G(d,p) single-point calculations to account properly for dispersion effects.[12] The results are shown in Table 2. The parent unsubstituted system 2 (R = H) shows a 2.9 kcal/mol preference for the C2 isomer over the C diene. All the trisubstituted alkenes were significantly more stable than the disubstituted alkene starting materials. Methyl and larger alkyl substituents at the bridgehead carbons led to a greater preference for the C2 isomer. This is in good agreement with the fact that under all conditions the C2 diene 3 is the predominant product.

Table 2

Free Energy Calculations for Isomersa

R	exo (SM)	C2	Cs	difference
H	0.0	–8.0	–5.1	2.9
Me	0.0	–7.1	–3.8	3.3
Et	0.0	–9.0	–5.2	3.8
Pr	0.0	–8.7	–4.6	4.1
Ph	0.0	–9.2	–6.4	2.8

Gas-phase calculations were carried out using M06-2X/6-311+G(d,p)//B3LYP/6-31G(d) and are quoted in kcal/mol.

We set out to investigate the source of the preference for the C2 isomer. As mentioned before, the examination of molecular models revealed a possible unfavorable steric interaction involving the two allylic hydrogen atoms in 4a as shown in Figure 2. Inspection of the optimized geometries reveals that, although the decrease in H---H distance correlates well with an increase in stability across the isomers (Figure 3), a 2.28 Å distance is not sufficient to conclude that the 3–4 kcal/mol thermodynamic preference is dominated by steric repulsion. Interestingly, a twisting of the bicyclo[3.3.1]nonadienone core in 3a, which is not observed in the less stable 4a, points toward ring strain induced by nonbonding interactions as being another component of the energy difference. This slight rotation relieves some of the unfavorable eclipsing interactions and translates to an increase in the endo hydrogen distance.

Figure 3

Monoisomerized and C2 and C bis-isomerized optimized structures. Free energies calculated using M06-2X/6-311+G(d,p)//B3LYP/6-31G(d) are relative to 2a.

In summary, the facile isomerization of the di(exo)methylene bicyclo[3.3.1]nonanone systems gives rise predominantly to the C2 products rather than the possible C products. Theoretical calculations reveal that the origin of this preference stems from thermodynamics effects, involving transannular hydrogen–hydrogen interactions and ring strain induced by these interactions. Further calculations and the use of this procedure in the synthesis of rugulosone 1 are currently underway and will be reported in due course.