Stereoelectronic and conformational effects on the

Pagination not final/Pagination non finale 1

Stereoelectronic and conformational effects on the stereochemical course of reduction of bicyclo[3.3.1]nonane 1,3-diketones Eugenius Butkus, Arvydas Ston…ius, Jul Malinauskien, Shuji Tomoda, and Daisuke Kaneno

Abstract: The stereoselectivity of the reduction of bicyclo[3.3.1]nonane 1,3-diketones was studied. The experimental data of π-facial stereoselection of the reduction of the carbonyl group was successfully rationalized by the application of the exterior frontier orbital extension (EFOE) model. The observed facial diastereoselectivity of the carbonyl reduction of bicyclo[3.3.1]nonane diketones could be reasonably explained by the ground-state facial anisotropy of the frontier orbital extension, steric effects, and the intrinsic reactivity of carbonyl groups. Although the EFOE density and PDAS values predicted the enhanced reactivity at the C-2 carbonyl compared to C-9 carbonyl, transition-state calculations at the B3LYP/6-31+G(d,p) level showed that the reactivity of C-9 and C-2 in a hydride addition to both carbonyl groups should be nearly the same. The EFOE analysis also strongly indicated that in the second hydride reduction step the corresponding oxobicyclononanolates are most likely to be involved in the reduction as the major species rather than the free hydroxyketones. The antiperiplanar effects in the transition states of the LiAlH4 reduction as measured by both the elongation and the natural bond population of the antiperiplanar bonds clearly indicated they should not be an essential factor of the facial diastereoselection of studied diketones. Key words: bicyclo[3.3.1]nonanedione, EFOE model, reduction, stereoselectivity, transition state. Résumé : On a étudié la stéréosélectivité de la réduction de bicyclo[3.3.1]nonane-1,3-diones. On a pu rationaliser facilement les données expérimentales de la stéréosélectivité π-faciale de la réduction du groupe carbonyle par l’application du modèle de l’extension des orbitales frontières extérieures (« EFOE »). La diastéréosélectivité faciale observée de la réduction du carbonyle des bicyclo[3.3.1]nonanediones peut s’expliquer raisonnablement par l’anisotropie faciale de l’état fondamental de l’extension de l’orbitale frontière, par des effets stériques et par la réactivité intrinsèque des groupes carbonyles. Même si la densité EFOE et les valeurs de « PDAS » ont permis de prédire la plus grande réactivité du carbonyle en C-2 par rapport à celle du carbonyle en C-9, les calculs d’état de transition au niveau B3LYP/631+G(d,p) montrent que les réactivités des deux groupes carbonyles en C-2 et en C-9 vis-à-vis de l’hydrure devraient être assez semblables. L’analyse EFO” indique aussi fortement que, dans la deuxième étape de réduction par l’hydrure, il est probable que les oxobicyclononanolates correspondants sont impliqués dans la réduction sous la forme d’espèces majeures plutôt que sous la forme d’hydroxycétones libres. Les effets antipériplanaires dans les états de transition des réductions par LiAlH4, tels que mesurés par l’élongation ainsi que la population naturelle de liaison des liaisons antipériplanaires, indiquent clairement qu’ils ne sont pas des facteurs essentiels de la diastéréosélectivité faciale des dicétones étudiées. Mots clés: bicyclo[3.3.1]nonanedione, modèle EFOE, réduction, stéréosélectivité, état de transition. [Traduit par la Rédaction]

Butkus et al.

8

Introduction The most important aspect of the conversion of organic molecules containing one or more stereogenic centers is usually that of stereochemical control. The problem of π-facial

stereoselection is one of the fundamental topics in theoretical and experimental organic chemistry. The state of art of this problem was recently reviewed in a special issue of Chem. Rev. (1). The origin of π-facial stereoselectivity is accounted for by two different transition-state (TS) models,

Received January 18, 2001. Published on the NRC Research Press Web site at http://canjchem.nrc.ca on November xx, 2001. Dedicated to Professor Victor Algirdas Snie…kus in recognition of his significant contribution to the development of organic synthesis. E. Butkus,1 A. Ston…ius, and J. Malinauskien. Department of Organic Chemistry, Vilnius University, 2734 Vilnius, Lithuania. S. Tomoda1 and D. Kaneno. Department of Life Sciences, Graduate School of Arts and Sciences, The University of Tokyo, Komaba, Meguro, Tokyo 153–8902, Japan. 1

Corresponding authors ((EB): telephone: +370-2-336517; fax: +370-2-330987; e-mail: [email protected]; (ST) e-mail: [email protected]).

Can. J. Chem. 79: 1–8 (2001)

© 2001 NRC Canada

PROOF/ÉPREUVE

Pagination not final/Pagination non finale 2

Can. J. Chem. Vol. 79, 2001

known as the Felkin–Ahn model (2) and Cieplak model (3). The former assumes that torsional and steric effects arise by interaction between the bonding orbital being formed and the antibonding orbitals of adjacent bonds can account for the experimental results (4). The Cieplak model places more importance in the interaction between the antibonding orbital of the incipient bond and the bonding orbitals of the adjacent bonds, and some interesting results in sterically unbiased systems were reviewed (5). An attempt to evaluate contributions of the above-mentioned approaches to explain the reaction selectivities by combining the stabilizing interactions was made by Li and le Noble (6) and Gung (7). Klein (8) introduced the frontier molecular orbital (FMO) distortion approach to π-facial selectivity. He pointed out that the anisotropic distribution of LUMO with respect to the carbonyl plane of the substrate in the ground state is governing the approach of nucleophile. The FMO approach was supported by ab initio calculations performed by Frenking et al. (9). The polarized π-frontier molecular orbital theory was shown to have some advantage over both the Anh and Cieplak models in the understanding of selectivity in some π-systems (10). However, these assumptions have been strongly criticized, and Felkin’s torsional strain in the transition state maintained as the main factor governing the stereochemical outcome of π-facial additions (11). In spite of this, the electronic effects in terms of remote electrostatic effect or even the through-space orbital interaction in transition state were recognized as playing the significant role in nucleophile addition to carbonyl in sterically unhindered systems (12). Recently the “exterior frontier orbital extension” (EFOE) model based on the Salem–Klopman kinetic equation has been suggested (13). The π-facial stereoselection in carbonyl group reduction and other nucleophilic addition reactions assumes the importance of the ground-state conformational and electronic properties of substrates generated over the exterior area of substrate ketones in the early stages of reaction, whereas the transition-state effects are most likely to have minor effects on facial selection (14). The calculation of the EFOE density and π-plane-divided accessible space (PDAS) provides a strong theoretical model for rediction of facial stereoselectivity of a hydride reduction and nucleophilic additions of a variety of ketones (15). Further, an interesting study on diastereoselectivity observed in gas-phase reduction of a series of bicyclic ketones by silicon hydride has been published (16). The diastereoselectivity was remarkably consistent with results obtained in solution and in accord with the predictions by molecular orbital calculations. The results are rationalized in terms of competition among steric, torsional, and electrostatic effects, and also suggest that reduction stereoselectivity might be attributed to the intrinsic nature of the reactants. However, it should be recognized that despite the power of prediction in many cases, the available models and procedures could not rationalize a number of stereochemical outcomes of π-facial additions. Consideration and comparison of the methods that are based entirely on second-order perturbation to predict diastereofacial selectivity led to a conclusion that there is no general explanation for the selectivities observed (17).

Earlier we studied the stereoselectivity of reduction of bicyclo[3.3.1]nonane ketones which seem to be suitable models to understand the stereochemical control of the reaction (18). However, an analysis of the TS structures at the 3-21G level and by the MNDO method did not give any definite explanation of stereoselectivites observed for bicyclo[3.3.1]nonane-2,9-dione. EFOE analysis is based upon a hybrid of ground-state (GS) and transition-state arguments. Therefore, a study of the applicability of the EFOE method to explain the stereoselectivity of the reduction of two diketones of this bicyclic structure was undertaken to interpret the stereoselectivity of the reaction.

Results and discussion Reduction of bicyclo[3.3.1]nonanediones The reduction of bicyclo[3.3.1]nonane-2,9-dione 1 and exo-7-methylbicyclo[3.3.1]nonane-2,9-dione 2 (Scheme 1) was performed using NaBH4, NaBH3CN, and LiAlH4 in an appropriate solvent. The examination of a reaction course of bicyclo[3.3.1]nonane-2,9-dione 1 and NaBH4 in methanol by GC and 1H NMR showed that reduction of the first carbonyl group was not regioselective but was totally diastereoselective (18a). Though solvent effect on the reduction stereoselectivity could not be excluded in NaBH4 reduction (19), there was no significant influence of it in this work. Only two of four possible hydroxyketones, i.e., the anti-9hydroxybicyclo[3.3.1]nonan-2-one 1A and the endo-2hydroxybicyclo[3.3.1]nonan-9-one 1B (Scheme 2) were formed in almost equal amounts. This ratio of hydroxyketones was found to be nearly constant during the reaction course. The stereochemical course of reduction of individual hydroxyketones was determined from the final ratio and configuration of diols, which in turn is dependent on the ratio of hydroxyketones formed in a reaction. The examination of mixtures of diastereomeric diols obtained after a complete reduction of diketones by 1H NMR, GC–MS, and circular dichroism spectroscopy of resolved enantiomeric derivatives provided us with the ratio of formed diols (Table 1). The reduction of bicyclo[3.3.1]nonane-2,9dione 1 led to two main diols 1a and 1b, and an insignificant amount of the minor diol 1c (4 ~ 8%). Essentially the same stereochemical course was observed in the reduction of exo7-methylbicyclo[3.3.1]nonane-2,9-dione 2. The distribution of the corresponding diastereomeric diols slightly varied in the reduction of 1 and 2 depending upon the hydride used and temperature of a reaction mixture. The most striking features of the investigated reduction of diones 1 and 2 was the low regioselectivity as contrasted with the high diastereoselectivity upon reduction of the first carbonyl group. At the first reduction step, hydride ion attacks the C-2 carbonyl group at the ring position from exo-face or the bridge C-9 carbonyl from syn-face, consequently syn attack with respect to both carbonyl groups occurs exclusively. Taking into account the fact that two hydroxyketones were forming in a substantial amount and the ratio of these hydroxyketones was found to be nearly constant during the reduction course, let us consider the reduction reaction as a two-step process. First, the reduction of one of the carbonyl groups of diketone leading to a formation of hydroxyketones, and second, a subsequent reduction of hydroxyketones. The © 2001 NRC Canada

PROOF/ÉPREUVE

Pagination not final/Pagination non finale Butkus et al.

3

Scheme 1. Conformers of investigated bicyclo[3.3.1]nonan-2,9dione 1 and exo-7-methylbicyclo[3.3.1]nonan-2,9-dione 2: c,c chair–chair, b,c boat–chair, c,b chair–boat, and b,b boat–boat. O

O

O 9 1

O

5

2

X

b,b

O

X

c,b

OH

O X

O

Scheme 2. Reduction scheme of bicyclo[3.3.1]nonan-2,9-diones 1 and 2.

X O

b,c

c,c

syn exo (syn) O

O

syn

[H]

endo (anti)

X

1: X = CH2 2: X = (CH)-CH3-exo

product distribution has been attempted to interpret by the theoretical evaluation of the TS though a good explanation could not be given (18a). Ground state conformational and EFOE analyses An approach of the use of the EFOE model to account for the stereochemistry of the reduction of bicyclo[3.3.1]nonane2,9-dione 1 and exo-7-methylbicyclo[3.3.1]nonane-2,9-dione 2 by complex metal hydrides is presented in this work. To explain the stereo- and regioselectivity of reduction, the theoretical calculation of properties of starting diketones and intermediate reduction products forming during the reaction was performed. The theoretical explanation of the stereochemical course of the reduction seems to be complicated due to the complexity of conformational and electronic properties of bicyclic diketones. A conformational mobility of investigated bicyclo[3.3.1]nonan-2,9-diones should be taken into account. Two six-membered rings of bicyclo[3.3.1]nonan-2,9dione may adopt four possible conformations, namely c,c, b,c, c,b, and b,b (Scheme 1). Numerous calculations performed earlier (18a, 23b) and in this work (20) showed that only c,c and b,c conformations are significantly populated and the latter appeared to be more stable. In addition, a transannular interaction between the carbonyl groups should not be neglected in bicyclic diketones as demonstrated by 13 C NMR (21), photoelectron spectroscopy and calculations (22), or CD spectroscopy (23) experiments. EFOE theory is based on a hypothesis that facial diastereoselection should not be dependent on thermodynamic stabilization effects towards the TS and one doesn’t need to focus on the latter. Namely, the information on the facial diastereoselection is contained in the initial properties (interactions) of the reactants. Since the same reagent attacks two π-faces, one can predict diastereoselectivity by looking at the properties of the substrate (ketone). Thus the GS interactions might be decisive, and EFOE analysis, which implies evaluation of these interactions, provides a method for analyzing the polarization of π-wave functions. It is commonly accepted that properties of virtual orbitals are not as precise as the occupied orbitals. However, those of LUMO are pretty accurately predicted, because LUMO most significantly reflects the properties of occupied MOs through iterative variation procedure during MO calculation. Within the framework of the same method and basis sets (in this case, HF/6-31G(d)), one can reasonably compare the energies and spatial extension of LUMOs. Two of the calculated MOs of the investigated bicyclic 1,3-diketones (1,2) could be recognized as π* MO, i.e., LUMO (lower energy) and LUMO + 1 (higher energy).

endo 1,2 c

X 1,2 A-Y = OH1,2 A Y = O syn

1, 2

1 X = CH2 2 X = (CH)-CH3-exo

exo O endo

anti

HO

Y

O

syn

exo (syn)

OH

exo

anti

HO 1,2 a

X

HO

anti Y X 1,2 B-Y = OH1,2 B Y = O

X

HO 1,2 b

X

These two MOs are located on both carbonyl groups and could be involved in π-facial stereoselection in investigated diketones. The difference in energy of LUMO and LUMO + 1, namely 0.42 ~ 1.2 eV at HF/6-31G(d) level (24), enabled us to consider that only the lower energy LUMO could be responsible for the new bond formation upon the hydride addition to the carbonyl group. The EFOE analysis (13, 14) of a complete set of conformers of 1 showed that conformers should demonstrate significantly different reactivity in nucleophilic addition reactions. LUMO of 1c,c is the lowest one (3.56 eV) and gives significantly larger values of both EFOE density and PDAS over the syn-face of both carbonyls (Table 2). The LUMO of the main conformer 1b,c and of scantily populated conformers c,b and b,b showed much reduced values of both the EFOE density and PDAS. Therefore, the most reactive conformation that mainly contributes to the reaction should be 1c,c despite the slightly higher population of 1b,c. Similar values of the EFOE density and PDAS were obtained for LUMO of the most populated conformers of the diketone 2 (Table 2). Overall, the EFOE model correctly predicted the preferred hydride approach from the syn-face to both carbonyls in agreement with experimental results, although the absence of regioselectivity at the first reduction step could not be understood by this approach. According to the EFOE analysis, the carbonyl group at C-2 is predicted to be more reactive compared to carbonyl at C-9 since both EFOE quantities of the LUMO for the syn-face are larger in the former group. However, the thermodynamic stability of the products should affect the reaction rate according to the Marcus theory (25). The regioselectivity of some reactions of investigated diketones (26) and molecular mechanics, semiempirical or ab initio estimation of the relative stability of hydroxyketones A and B gave the following results: 1Ac,c is more stable by 1.65 kcal mol–1 than 1Bc,c, and 2Ac,c is more stable by 2.90 kcal mol–1 than 2Bc,c. It should be noted that exactly the same trend for the corresponding alkoxy anions (1A– < 1B– (2.84 kcal mol–1) and 2A– < 2B– (3.00 kcal mol–1)) was observed (Table 3). Analysis of the transition states on addition of LiAlH4 Transition states possibly forming in the reduction of the first carbonyl group were calculated. Lithium bicoordinate transition complexes of LiAlH4 (Table 4) and substrate diketones were found to be uniformly more stable by about © 2001 NRC Canada

PROOF/ÉPREUVE

Pagination not final/Pagination non finale 4

Can. J. Chem. Vol. 79, 2001 Table 1. Results of hydride reduction of bicyclo[3.3.1]nonan-2,9-diones 1 and 2. Hydride/solvent

Diketone

a (anti-endo) (%)

b (syn-endo) (%)

c + d (anti-exo) + (syn-exo) (%)

NaBH4/MeOH

1 2 1 2 1 2

59 68 58 63 62 73

36 27 32 36 38 24

4