High-level Computational Study of the Site-, Facial- and Stereoselectivities for the Diels-Alder Reaction Between o-Benzoquinone and Norbornadiene

Abstract

Ab initio and DFT quantum chemical calculations have been applied to a study of the Diels-Alder reaction of o-benzoquinone as diene and norbornadiene as dienophile. Transition states for the different reactions are located and activation energies estimated. The prefered exo-π-facial selectivity and exo,endo-stereoselectivity exhibited in this cycloaddition are readily predicted using RHF/3-21G or higher levels of calculations. Differences between experimentally observed results and calculations may be explained by the postulation of a second, nonconcerted biradical mechanism leading to formation of hetero Diels-Alder products.

Introduction.

Diels-Alder methodology has been used by ourselves [1] and others [2] to produce rigid polyalicyclic structures. The overall shape of such systems depends on the specificity of the reactions used in each step of the construction. With a knowledge of stereospecificity, for example, it is possible to introduce a bend in the framework or to extend the structure linearly. In previous work, we have described synthetic methods for the preparation of o-benzoquinone, o-chloranil and 2,5-di-(t-butyl)- benzoquinone adducts with norbornadiene [3,4]. Here, we will present the results of the computational study of these reactions using high level quantum mechanical calculations. We have found that o-benzoquinone 1a generated in situ gives four α-dione adducts 3-6 (Scheme 1) with adduct 3 being the dominant product while adducts 4 and 5 were found in smaller amounts and adduct 6 was detected by spectroscopy, but not isolated. In contrast to this reaction, o-chloranil 1b gave six adducts 3-8, while 2,5-di-(t-bu)-benzoquinone 1c gave only adducts 9 and in minor amounts 10.

High level ab initio [5] and B3LYP/6-31G* DFT methods [6] have proven to give excellent results for energy barrier estimation of pericyclic reactions. We have previously used ab initio and DFT calculations to successfully predict stereospecificities of different Diels-Alder reactions of cyclic dienes with cyclic dienophiles [7]. In order to explain the experimentally observed stereospecificites, transition states for these reactions have been located and activation energies estimated at various levels of theory.

Computational Methodology

All ab initio calculations were conducted using the SPARTAN program [8] on a Silicon Graphics Oxygen R5000 workstations. These geometries have been used as the initial geometries for DFT calculations with Gaussian98 [9]. MP2 calculations were conducted using IBM SP2 supercomputer. Geometric optimizations were carried out without using symmetry or other structural restrictions. All calculations are performed at the restricted Hartree-Fock level [10] with 3-21G and 6-31G* basis sets [11]. Each transition structure was located using a standard routine within Gaussian 98 [12]. For all structures the vibrational analysis was performed with the same basis set used for optimization. The activation energies were also estimated from 6-31G* and MP2/6-31G* single point calculations on the RHF/3-21G and RHF/6-31G* optimized geometries. Further optimizations were carried out with density functional theory (DFT) hybrid B3LYP (Becke's 3 parameter functional [13] with the non-local correlation provided by the expression of Lee et al.) [14]. Single point energy calculations were also estimated using DFT B3LYP method: B3LYP//3-21G [15] and B3LYP/6-31G*. This study has been restricted to the parent system since the inclusion of substituents (chlorine or t-Bu) onto the o-benzoquinone considerably increased the computational effort. Such results will be reported at a later time.

Results and Discussion

The total energies of the calculated molecules and their associated transition states (Scheme 2) are collected in

Table 1. Total energies (au) of molecules under studies^a

, while activation energies and relative energies are presented in

Table 2. Calculated activation energies (kJ/mol) for the Diels-Alder reactions of 1,3-cyclohexadien-4,5-diones with norboradiene^a

and

Table 3. Relative energies (kJ/mol) for the studied Diels-Alder reactions

. B3LYP transition state structures are depicted in Figure 1, Figure 2, Figure 3 and Figure 4. All located transition state structures correspond to the concerted, synchronous mechanism. The lengths of the bonds undergoing first-order changes in these transition structures are commensurate with those expected for concerted cycloaddition reaction transition states.

FMO analysis has shown that norbornadiene HOMO and o-benzoquinone LUMO are the most important interacting orbitals, i.e. this represents an inverse electron-demand Diels-Alder reaction (norbornadiene HOMO -6.73 eV, LUMO 0.0 eV, o-benzoquinone HOMO -6.79 eV, LUMO -3.54 eV). Furthermore, quantum of charge transfer (qCT) from diene to dienophile for TS1 - 4 are estimated to be -0.124, -0.108, -0.136 and -0.120 eV, respectively, clearly indicating an inverse electron-demand mechanism.

π-facial selectivity. As mentioned in the introduction, the reaction of norbornadiene with o-benzoquinone (1a) and o-chloranil (1b) gives mixtures of the expected exo,exo-(3) and exo,endo-(4) as well as detectable amounts of the endo,exo-(5) and endo,endo-(6) adducts. The analysis of results of o-benzoquinone (1a) with norbornadiene collected in Table 1, Table 2 and Table 3 shows that all methods employed, starting with RHF/3-21G gave a good prediction of exo-π-facial selectivity in norbornene system. Regardless of the computational level employed, exo,exo-adduct 3 (which is formed via TS1) was found to be the preferred by 7.5-13.6 kJ/mol over exo,endo-adduct 4, consistent with that observed experimentally. The largest difference (13.6 kJ/mol) was estimated at the 6-31G*//3-21G level. These results reinforce our previous findings that 6-31G*//3-21G and higher theoretical levels correctly model D-facial selectivity in cycloaddition reactions with norbornenes [7]. As expected, estimated energy barriers vary with the computational level applied [17]. In contrast to these findings, there was found no significant difference between activation energies for the formation of exo,endo-4 and endo,exo-5 adducts. These predictions are less consistent, depending on the calculation level employed. While 3-21G full otimization, single point B3LYP/6-31G* energy estimations on the 3-21G and 6-31G* optimized structures and B3LYP/6-31G* calculations predict smaller activation energies for TS2 than TS3 (by 0.76, 3.5, 2.3 and 3.03 kJ/mol. respectively), the 6-31G*//3-21G, MP2/6-31G*//3-21G, 6-31G* and MP2/6-31G*//6-31G* calculations (Table 3) predict smaller activation energies for TS3 over TS2 (by 1.4, 3.1, 0.7 and 4.0 kJ/mol, respectively). However, we may conclude that, since TS2 - TS3 energy differences are relatively small, all calculations predict formation of similar amounts of products exo-face product 4 and endo-face product 5. Furthermore, the large differences in activation energies between TS1 and TS4 (within a range of 25.3 and 33.5 kJ/mol, as estimated by MP2/6-31G*//3-21G and 3-21G methods, respectively, Table 3), clearly indicate that formation of product 6 is greatly disfavoured. This prediction is in accord with our experimental results as well as computational results obtained by Hehre et al. having recently shown that ab initio calculations using 3-21G model correctly predict the relative energies of the transition states for the two modes of attack to diastereotopic faces of a diene [18].

The origin of the observed stereoselectivity may be rationalized using Mulliken population analysis, which gives a qualitative indicator of the amount of electron density shared by two atoms, and also provides some evidence for secondary orbital interaction between two reactants. Such an analysis was employed succesfully by Houk et al. to explain the stereoselectivities of several Diels - Alder reactions [19].

In TS1 the methylene hydrogen - double bond carbon (H₇C₁₂) overlap density has a positive value of 0.0014, indicating an attractive interaction, while in TS2 the methylene hydrogen - carbonyl carbon atom (H₇C₉) overlap population is repulsive with value of -0.003. Furthermore, the methylene carbon - carbonyl carbon (C₇C₉) overlap population is also repulsive with value of -0.002, and methylene carbon - carbonyl oxygen (C₇O₁₃) overlap population is repulsive with value of -0.005, indicating larger destabilizing secondary orbital interactions in TS2. However, the similar analysis of overlap densities employed on the TS3 and TS4 does not give such conclusive differences.

Formation of hetero Diels-Alder products. We have found experimentally that o-chloranil 1b and 2,5-di-(t-bu)-benzoquinone 1c also also serve as heterodienes at the α-dione oxygens. While 1b gave a mixture of normal Diels-Alder adducts 3-6, and smaller quantities of exo- and endo- adducts 7 and 8, the quinone 1c gave only hetero Diels-Alder adducts 9 (major product) and 10 (trace). Similar behaviour of o-benzoquinione has been reported previously by Kumar et al. [20].

For each mode of attack to the π-system, we have located two transition structures: for exo-approach TS5 (where aromatic ring is facing methylene bridge) and TS6 (where aromatic ring lay in the plane of norbornene cyclohexene ring), and for the endo- addition TS7 (where aromatic moiety is facing double bond) and TS8 (where aromatic ring is outside norbornene moiety) (Scheme 2). While these structures possess some energy differences, we were unable to see discrete species using ¹H-NMR, which suggest rapid interconversion at room temperature between products derived from TS5/TS6 and TS7/TS8.

The inspection of results collected in Tables 2-4 reveals nonconsistent results for activation energies for TS5-8 at all employed theoretical levels. For most transition structures, activation energies are smaller, or very similar to the one for the TS1-4, which is opposite to experimental results. For instance, the B3LYP/6-31G* method, which gives excellent results for cycloaddition reactions, predicts TS6 to have the smallest energy followed by TS5, TS7 and TS8, while activation energies for TS1-4 are significantly greater, suggesting the exclusive formation of products 5-8. However, these adducts have not been experimentaly detected. In this case, only the RHF/6-31G* method gave the correct predictions, i.e. exclusive formation of products 1-3, while hetero Diels-Alder products have much bigger activation energies.

The lengths of the bonds undergoing first-order changes in these transtition structures are those expected for concerted cycloaddition transition states (Figure 3 and Figure 4). However, harmonic frequency calculations identified these structures as second-order saddle points. Efforts to locate concerted, synchronous transition structures for hetero Diels-Alder reaction using various spin-restricted wave functions resulted in each case, in the location of a single stationary point possessing more than one negative mode of vibration. Despite an extensive search of the singlet-state energy surface using restricted levels of ab initio and DFT theory, we were unable to locate genuine transition structures for either synchronous or a nonsynchronous concerted cycloaddition process. The same finding was experienced using various spin-restricted wave functions (either 3-21G, 6-31G* or B3LYP methods).

The discrepancies between the experimentally observed results and the theoretical analysis may be explained by the postulation of a second, nonconcerted biradical mechanism leading to formation of hetero Diels-Alder products 5-10. Given our failure to locate concerted transition structure for the o-benzoquinone hetero Diels-Alder cycloaddition, our attention is now focused on calculations of nonconcerted reactions. The calculations employing unrestricted ab initio and B3LYP calculations are currently being undertaken and these results will be reported in due course.

Conclusions

The present results demonstrate the ability of ab initio calculations to accurately predict relative reactivities and stereoselectivities for inverse electron-demand Diels - Alder reactions in alicyclic systems with cyclic 1,3-dienes. Transition states were located and activation barriers estimated at different levels of theory, by Hartree-Fock, post- Hartree-Fock and DFT methods. The high exo- π-facial selectivity exhibited in these cycloadditions are readily predicted using RHF/3-21G or higher ab initio levels. In the case of hetero Diels-Alder products, all quantum chemical levels employed failed to correctly predict energy barriers, which suggests that second, nonconcerted biradical mechanism may be operating.