Experimental data

Lithium amides are an important class of reagents in organic synthesis and have been used extensively as strong bases for various reactions [1–3]. The deprotonation of an epoxide with a lithium amide to obtain an allylic alcohol was first reported in 1970 in a deuterium-labeling study by Thummel and Rickborn [4]. The reaction is thought to proceed via a cyclic six-membered transition state, formed by a 1:1 epoxide to base complex, where the base coordinates to the lone pair of electrons on oxygen, thereby facilitating the β-hydrogen removal. On the other hand, if a prochiral epoxide is deprotonated with a chiral lithium amide, it could give an optically active product. Such a rationale was first proposed over two decades ago, for a nonenzymatic “asymmetric deprotonation” of cyclohexene oxide to obtain 2-cyclohexen-1-ol (Scheme 1) [5].

Scheme 1
scheme 1

Asymmetric deprotonation of cyclohexene oxide

Full size image

Over the years, a number of different chiral lithium amides have been designed and used for various deprotonation reactions [6]. As a result, asymmetric synthesis using chiral lithium amides has emerged as a useful method for the preparation of nonracemic compounds [7–11]. The methodology offers the advantage that the chiral auxiliaries can be recycled easily, thereby making the process effective and cost efficient. These chiral lithium bases have been exploited by a variety of efficient enantioselective reactions. Considering the remarkable success of pinene-based reagents for various organic transformations [12, 13], we synthesized a set of chiral secondary amines derived from á-pinene (Scheme 2) (Malhotra and Brown 1998, unpublished results).

Scheme 2
scheme 2

Four kinds of lithium amides

Full size image

The lithium salts of these amines were tested in the deprotonation of meso-epoxides [14]. Diisopinocampheylamine (DIPAM) prepared from (+)-α-pinene has the advantage of C 2-symmetry. N-cyclohexyl-N-isopinocampheylamine (ChxIPAM), N-benzyl-N-isopinocampheylamine (BzIPAM) and N-isopropyl-N-isopinocampheylamine (iPIPAM) were chosen to study the steric effect of the isopinocampheyl moiety in the deprotonation of a meso-epoxide to allylic alcohol. Reaction with a stoichiometric amount of DIPAM gave the product with >99% enantiomeric excess (EE) on deprotonation of cyclohexene oxide.

To generate a catalytic cycle [15, 16] (Scheme 3), numbers of achiral lithium amides were compared with chiral amines for the opening of meso-epoxides. Lithium diisopropylamide (LDA) emerged as the achiral amide of choice because first, the chiral amine could be regenerated easily, and secondly, due to the fact that there was a significant difference in the rate of competitive reactions with the chiral amides, i.e., when two amides (achiral and chiral) were present together, the deprotonation of the epoxide was mainly due to the chiral amide. Also, a systematic investigation with varied amounts of chiral amine DIPAM, and at different temperatures revealed that the best results could be obtained at 0 °C using 20 mol% of the chiral and 125% of the achiral amines.

Scheme 3
scheme 3

Reaction flow chart

Full size image

The results of the deprotonation of meso-cyclohexene oxide with various chiral amides are shown in Table 1.

Table 1 Catalytic enantioselective deprotonation of cyclohexene oxide 1
Full size table

The experimental data clearly indicate that the isopinocampheyl moiety is the key to obtaining high yields and EEs.

Theoretical calculations

Successful rationalizations of enantioselectivity depend on comprehensive knowledge about the reaction mechanism [17, 18]. To verify the experimental results shown in Table 1, we embarked on a theoretical calculation of the deprotonation reaction of cyclohexene oxide using Gaussian 98. The possible reaction routes for deprotonation of cyclohexene oxide with various chiral amides are shown in Scheme 4.

Scheme 4
scheme 4

The deprotonation reaction equations of cyclohexene oxide

Full size image

First, all transition states for reactions with various chiral amines were located and optimized by ab initio HF/3-21G and density functional B3LYP/3-21G calculations with Gaussian 98. The deprotonation of the epoxide with lithium amide occurs via a six-membered ring transition state, where lithium approaches the lone pair of electrons on oxygen, thus facilitating the removal of the β hydrogen. As shown in Fig. 1a, in the case of deprotonation with DIPAM-Li (R1 optimized with B3LYP/3-21G), when lithium approaches the lone pair electrons on oxygen, the N–Li bond is stretched to 1.870 Å (from 1.750 Å). As a result, the C–O bond (1.840 Å) is broken. Furthermore, the β C–H bond is stretched from 1.080 to 1.240 Å and then broken. The transition states for (R) and (S) isomers obtained with different ligands are shown in Fig. 1b. When the “β hydrogen” is deprotonated from the C-3 atom of cyclohexene oxide, the transition states result in the R configuration, and the product is (R)-2-cyclohexen-1-ol. On the other hand, if the “β hydrogen” is deprotonated from the C-6 atom of cyclohexene oxide, the transition states give the S configuration, thereby (S)-2-cyclohexen-1-ol is obtained. The selected bond distances of the transition state (R and S) structures were optimized using ab initio HF/3-21G and density-functional B3LYP/3-21G calculations. The changed bond distances are shown in Table 2. The first imaginary frequency shown in the Appendix explains that ab initio HF and density-functional B3LYP methods with the 3-21G-basis set were successful in finding transition states in this study.

Fig. 1
figure 1

a Six-membered ring transtion states. b The (R) and (S) transition states for the deprotonation reactions shown in Scheme 4

Full size image
Table 2 Important bond distances in optimized transition states
Full size table

As Fig. 1 and Table 2 show, the bond distances varied with different R-groups. R isomers have absolutely the same atoms as S isomers, but the bond distances are different. Bonds a, b, c, d, e, f or a′, b′, c′, d′, e′, and f′ are most important for locating the transition states. Bond g–n or g′–n′ did not change much in various transition state structures.

The total energies in Hartree at 0 K for the reactants and transition states obtained by way of single-point calculations on the B3LYP/3-21G optimized geometries using a modern SCRF/PCM [19] implementation are shown in Table 3. The solvent in the SCRF calculations is tetrahydrofolate (THF), which is the same as used in the experiments.

Table 3 The energies for all transition states and reactants in Hartree at 0 K
Full size table

As the calculation and data in Table 3 show, the transition state energy for R isomer is lower than the S isomer in each case. Hence, the transition states for the R products are more stable and therefore easier to obtain. The energy differences between the transition states for the R and S structures were also calculated. However, all computed (S)-(R) energies in Table 3 correspond to >99% ee, which did not reflect the observed experimental selectivities. The activation energies can be calculated by subtracting the sum of the energies of reactants from the energy of the corresponding transition state. For example:

$$\begin{aligned} & {\text{Activation}}\,{\text{Energy}}\,{\text{of}}\,{\text{DIPAM}} = \begin{array}{*{20}c} { = {\text{R}}1 - {\left( {{\text{DIPAM}} - {\text{Li}} + {\text{Cyclohexene}}\,{\text{Oxide}}} \right)}} \\{ - 1152.031094 - {\left( { - 842.197713} \right)} + {\left( { - 309.889039} \right)}} \\{ = 0.055658{\left( {{\text{Hartree}}} \right)}} \\\end{array} \\& 0.055658 * 627.5095 = 34.926{\left( {{\text{kcal}}\,{\kern 1pt} {\text{mol}}^{{ - 1}} } \right)} \\\end{aligned}$$

The activation energies were calculated at the density functional B3LYP level with the 6-31+G(3df,2p) basis set. We only calculated the R configuration structures and the data for R configuration are shown in Table 4. Similarly, the calculations for the S configuration structures could use the same method. One thing that should be pointed out in this study is that the monomeric Li-amide may be a poor model for the substrates in solution because of its tendency to form higher aggregates.

Table 4 Activation energies calculated by B3LYP/6-31+G(3df,2p)//B3LYP/3-21G
Full size table

In the case of reaction with the chiral amine DIPAM, the activation energy obtained is the lowest. This reaction can occur selectively and rapidly, resulting in high EE and yield of product. Substituting one isopinocampheyl group in DIPAM with cyclohexyl leads to higher activation energy for the reaction to occur. Similarly, substitution with a benzyl or isopropyl group also gives increased reaction energy, resulting in lower product yield and EE. We also note that with the benzyl group, the relative activation energies are highest, which means that this kind of reaction is the most difficult. Therefore, lower yields and EE for this product should be obtained, as is the case in Table 1. These theoretical calculations confirm the experimental findings that the isopinocampheyl moiety plays an important role in orienting the epoxide and the lithium salt in such way that the deprotonation takes place very enantioselectively.

Conclusion

The mechanism of the enantioselective deprotonation of cyclohexene oxide with isopinocampheyl-based chiral lithium amide was examined at the B3LYP/6-31+G(3df,2p)//B3LYP/3-21G levels using Gaussian 98. We uncovered the mechanistic details of the catalytic cycle, which are supported well by the experimental results. The highest EE (95%) for the conversion of cyclohexene oxide to 2-cyclohexen-1-ol has been achieved with lithium DIPAM through the catalytic system. These theoretical and experimental results certainly promise the catalytic application of DIPAM and similar pinene-based ligands in the complex synthesis of natural products and biologically important compounds.