Cundari Th.R. -- Computational Organometallic Chemistry-0824704789

Scheme 1

2. CYCLIZATION VERSUS ELIMINATION

We observed (1,2) that an α-diazo β-keto ester 5 (Scheme 2) would, on exposure to a catalytic amount of Rh2(OAc)4, undergo smooth cyclization to the cyclopentane derivative 6. Since the α-diazo β-keto ester 5 was readily prepared by diazo transfer (3) to the corresponding β-keto ester 4, this established a general route to highly substituted cyclopentanes. The subsequent observation (4) that use of Rh2(O2CR)4 catalysts derived from more electron-donating carboxylic acids al-

Scheme 2

lowed the efficient cyclization of simple α-diazo esters (5) such as 7 (Scheme 2) to the corresponding cyclopentanes with high diastereocontrol set the stage for the work described here.

3. DEVELOPMENT OF THE COMPUTATIONAL APPROACH

Any or all of products 11–14 (Scheme 3) could have been formed by cyclization of the α-diazo ester 10. In the event, only 12 was observed (6). In an attempt to rationalize this result, we developed a computationally based model for the transition state for C–H insertion.

An understanding of the mechanism (7) for Rh-mediated intramolecular C–H insertion begins with the recognition that these α-diazo carbonyl derivatives can also be seen as stabilized ylides, such as 15 (Scheme 4). The catalytic Rh(II) carboxylate 16 is Lewis acidic, with vacant coordination sites at the apical positions, as shown. The first step in the mechanism, carbene transfer from the diazo ester to the Rh, begins with complexation of the electron density at the diazo carbon with an open Rh coordination site, to give 17. Back-donation of electron density from the proximal Rh to the carbene carbon with concomitant loss of N2 then gives the intermediate Rh carbene complex 18.

The mechanism by which this intermediate Rh carbene complex 18 reacts can be more easily understood if it is written as the inverted ylide 19. This species would clearly be electrophilic at carbon. We hypothesized that for bond formation to proceed, a transition state (20) in which the C–Rh bond is aligned with the target C–H bond would be required (8). As the C–H insertion reaction proceeded, the electron pair in the C–H bond would drop down to form the new C–C bond,

Scheme 3

for the transition state that will allow us to discern those cyclizations that will proceed with high diastereoselectivity.

6. CHIRAL AUXILIARY CONTROL

Returning to the cyclization of a simple α-diazo ester (Scheme 6), we wanted to design (13) a chiral ester that would direct the cyclization preferentially to one absolute configuration of the product cyclopentane. In attempting to extend our computational approach to the design of such a chiral auxiliary, we found that we were missing a key piece of data, the dihedral angle between the ester carbonyl and the rhodium carbenoid at the point of commitment to cyclization (30-syn vs. 30-anti). We and others have, in the past, speculated that the ester carbonyl and the rhodium carbenoid could be syn (14), anti (15), or orthogonal (15a), but no experimental or computational evidence in favor of any of these had been put forward. Since our computational approach did not allow us to answer this question directly, we devised an indirect approach based on the cyclization of α-diazoester 29, derived from the naphthylborneol 33 (16). Our conclusion from this study is that the ester carbonyl and the rhodium carbenoid are syn in the transition state leading to the cyclization of esters such as 29.

The chiral diazo ester 29 was cyclized with four commonly used rhodium carboxylate catalysts (Table 2). It was found as before that rhodium pivalate (17) (entry 4) was most efficient for forming the cyclopentanes and that rhodium trifluoroacetate (entry 1) was best for forming the alkenes (18). For the pivalate, both the yield of the cyclization and the diastereoselectivity improved at lower temperature (entry 5).

Scheme 6

ence for the ester carbonyl and the CCRh bond to be syn at the point of commitment to cyclization, is general and is not limited to this particular case.

While we have had some success, we are aware of the limitations inherent in a transition-state model for rhodium-mediated C–H insertion that attempts to predict product ratios on the basis of mechanics calculations. Arbitrary decisions limiting the several degrees of freedom possible in the transition state could lead one to a model for the ‘‘point of commitment’’ to cyclization that would be far from reality. The work described here is important because it offers experimental evidence for a key rotational degree of freedom, the dihedral angle between the ester carbonyl and the rhodium carbenoid.

Our initial objective, in this investigation, had been to design a useful chiral auxiliary. We were pleased to find that naphthylborneol 33 itself, on optimization of the catalyst and the reaction temperature, served effectively. Until effective chiral catalysts are developed, naphthylborneol 33 will be of significant practical value for directing the absolute course of rhodium-mediated intramolecular C–H insertion reactions.

8. IMPLICATIONS FOR CHIRAL CATALYST DESIGN

It was striking (Table 2) how much changing the ligand on the rhodium carboxylate changed the product distribution from the cyclization of 29. It has been consistently observed by us and by others that, electronically, the ligands exert substantial control over the reactivity of the intermediate carbenoid. It is apparent (20, Scheme 4) that a strongly electron-withdrawing ligand will result in a more reactive carbenoid and that, with such a ligand, commitment to product formation will occur while the carbenoid carbon and the target C–H are still some distance apart. By changing the ligand from octanoate to pivalate, the reactivity of the carbenoid is apparently attenuated, resulting in a tighter transition state. The distance between the carbenoid carbon and the target C–H is then smaller at the point of commitment, bringing the chiral ester in closer proximity to the reaction center, where it can better influence the product distribution by the handedness of its steric bulk. An effective chiral catalyst, then, will have to direct the reaction sterically, and at the same time be electron-donating enough to have a late, tight ‘‘point of commitment.’’

9.COMPUTATIONAL DESIGN OF A RHODIUM CATALYST: BRIDGING THE TETRAKISCARBOXYLATODIRHODIUM CORE

In approaching the design of a chiral catalyst, the first question was whether or not our computational approach would allow us to predict the conformation of ligands around the Rh–Rh core. In particular, it seemed important to us, if we

were to effectively control the three-dimensional space surrounding the Rh carbene/C–H interaction, to design (19) a family of dicarboxylate ligands that could occupy four of the eight O–Rh sites on the dirhodium tetracarboxylate (35, Scheme 7).

Although several hundred tetrakiscarboxylato metal–metal dimers were known (20), there had been no report of a dicarboxylic acid that would bridge two positions on such a dimer (21). We reasoned that the best chance for success would be with a dicarboxylic acid that was specifically designed to fit across the

˚

5.4-A gap between the carboxylate ligands.

To approach the design of a dicarboxylate ligand that would effectively bridge the Rh–Rh core, we first optimized the parent Rh2(CF3CO2)4 using ZINDO (22,23). We locked this structure, bridged two of the acetate methyl groups with an increasing number of methylenes, and evaluated the strain energy of the resulting (hypothetical) complexes using mechanics (11). As expected, the initial very high strain energy for a zero-methylene bridge decreased rapidly with increasing bridge length, until the bridge reached four carbons. After that, the strain energy did not significantly decrease with increasing bridge length. Recognition that entropy considerations would favor bridging by a convergent bidentate ligand then led us to m-benzenedipropanoic acid 35 as a likely candidate (24,25).

A priori, there was cause to be concerned that exchange of a tetrakiscarboxylato metal–metal dimer with a dicarboxylic acid would lead only to oligomers. In that event (Scheme 7), heating 35 in dichloroethane with tetrakis(trifluoroacetato)dirhodium 34 led to smooth exchange to give the emerald green complex 36, which was easily purified by silica gel chromatography. Prolonged heating of the reaction mixture led to more polar materials. The structure of 36 was confirmed by X-ray diffraction of the derived bis-acetone adduct. Our calculated structure for the bis-acetone adduct was exactly superimposable (19) on the X-ray structure.

A key question was whether or not the bridged dimer would effectively catalyze the C–H insertion reaction. We were pleased to observe (Scheme 8) that

Scheme 7