The Origins of Stereoselectivity in the α-Alkylation of Chiral Hydrazones

Abstract

Density functional theory calculations and experiment reveal the origin of stereoselectivity in the deprotonation–alkylation of chiral N-amino cyclic carbamate (ACC) hydrazones. When the ACC is a rigid, camphor-derived carbamate, the two conformations of the azaenolate intermediate differ in energy due to conformational effects within the oxazolidinone ring and steric interactions between the ACC and the azaenolate. An electrophile adds selectively to the less-hindered π-face of the azaenolate. Although it was earlier reported that use of ACC auxiliaries led to α-alkylated ketones with ers of 82:18 to 98:2, B3LYP calculations predict higher stereoselectivity. Direct measurement of the dr of an alkylated hydrazone prior to removal of the auxiliary confirms this prediction; the removal of the auxiliary can compromise the overall stereoselectivity of the process.

Introduction

The α-alkylation of ketones is a useful synthetic operation, most often achieved via electrophilic addition to a derived azaenolate. Compared with enolates, azaenolates provide improved reactivities, yields, and regioselectivities, and can incoporate nitrogen-based chiral auxiliaries.ⁱ Enders’ SAMP/RAMP auxiliaries–the proline-based (S)- and (R)-1-amino-2-methoxymethylpyrrolidines (Scheme 1) are widely used.if,g An attractive feature of the SAMP/RAMP methodology is the well-defined stereochemical predictability; the appropriate enantiomer of the auxiliary can be chosen in advance. There have been many studies, both experimental and theoretical, of the mechanisms and stereoselectivities of alkylations of azaenolates derived from dialkylhydrazones, imines, and oximes.ⁱⁱ

Scheme 1.

Stereoselective α-alkylation of a RAMP hydrazone^ig

We recently discovered that chiral N-amino cyclic carbamate (ACC) auxiliaries are convenient alternatives for ketone α-alkylations (Scheme 2).ⁱⁱⁱ ACCs react easily with ketones to afford hydrazones, and can be recovered quantitatively from the products after alkylation. Deprotonation of an ACC hydrazone is rapid; the stereoselectivity of alkylation is high even without the use of extreme low temperature; and yields are excellent. Among the ACCs that we have investigated to date, camphor-based auxiliary 4 (Scheme 2) has proven to give the best yields and enantioselectivities.

Scheme 2.

Asymmetric α-alkylation of ketones, mediated by ACC 4

A mechanism for the stereoselective alkylation of a SAMP/RAMP hydrazone was proposed by Enders on the basis of crystallographic, spectroscopic, and computational evidence,^ig,iv,v and is depicted in Scheme 1. Deprotonation of the hydrazone (1) under equilibrating conditions gives rise to the lithium azaenolate (2), which has an E configuration at the CC bond and a Z configuration at the CN bond.^vi The bottom face of 2 is blocked by the pyrrolidine ring, and reaction with an electrophile takes place selectively at the top (β) face.^vii

We proposedⁱⁱⁱ that a similar mechanism is followed in the alkylation of an ACC azaenolate, but that the formation of the azaenolate from the hydrazone is controlled by the orientation of the carbonyl group. As shown in Scheme 3, the generalized ACC hydrazone 9 would prefer to exist in the conformation depicted as 9a, where steric interactions between R² and the larger substituent (L) on the auxiliary are minimized. Coordination of LDA to the carbonyl group would then lead to a “syn-directed” deprotonation,^viii giving the E_CC/Z_CN-azaenolate 10a as a 5-membered chelate. The bottom face of 10a is sterically blocked, and alkylation should take place selectively at the top (β) face.

Scheme 3.

Model proposed to explain stereoselectivity in the alkylation of ACC hydrazonesⁱⁱⁱ

We present here a computational study of the deprotonation of ACC hydrazones and the alkylation of the lithium azaenolates. Density functional theory calculations provide information about the transition states that lead to the selectivities shown in Scheme 2. However, computations predict even higher stereoselectivities than were originally reported.ⁱⁱⁱ This discovery prompted us to measure directly the dr of an alkylated hydrazone, prior to its conversion to the ketone. Consistent with theoretical predictions, the dr of the hydrazone was higher than the er of the final ketone, revealing that the removal of the auxiliary indeed compromised the overall stereoselectivity of the process.

Results and Discussion

To explore the structural properties of ACC hydrazones, we first examined the parent ACC hydrazone 11 (Figure 1). At the B3LYP/6-31G(d) level, two isomers of 11 were located, which differ in the conformation about the N–N bond. The more stable isomer, 11-syn, has a synclinal arrangement of the N–C_C=O bond and the N=C bond (CNNC_C=O dihedral angle 71°). The other isomer (11-anti) has an anti arrangement of these bonds (dihedral angle 150°), and is 4.7 kcal mol⁻¹ less stable (ΔH_0K). Its lower stability is due to repulsive interactions between the C=N and C=O lone pairs. In both isomers of the hydrazone, the ring nitrogen is pyramidal with an average bond angle of 114°. In this respect, the ACC hydrazone differs from simple carbamates (cf. the species 12 and 13, Scheme 4), and instead resembles an N,N-dialkylhydrazone (14). The NR₂ lone pair in 11 and 14 lies roughly in the same plane as the C=N bond, due to steric effects.^ix

Figure 1.

Conformers of the ACC hydrazone 11 (ΔH in kcal mol⁻¹ at 0 K).

Scheme 4.

Average bond angles at nitrogen in model species

To study the directed deprotonation of 11, Li(NMe₂)(THF) was used as a model for LDA in THF solution. Both 11-syn and 11-anti can undergo an intramolecular deprotonation reaction, following coordination of Li(NMe₂)(THF) to the carbonyl oxygen. The transition states, TS-11-syn and TS-11-anti, are shown in Figure 2. The reaction involving 11-syn represents the “syn-directed” deprotonation described above (Scheme 3), whereas the reaction of 11-anti leads to the opposite result, formation of a carbanion trans to the ACC group. The “syn-directed” deprotonation pathway is favored: TS-11-syn lies 1.4 kcal mol⁻¹ lower than TS-11-anti. The absolute barriers, relative to the reactant complexes, are very low or negative in the gas phase (ΔH^‡ = 0.3 and −0.6 kcal mol⁻¹, respectively). Because the lowest-energy TS (TS-11-syn) appears to have a vacant coordination site on the lithium, we reoptimized its geometry with a second THF coordinated. The second THF was found to bind in a slightly endergonic fashion (ΔG +2.4 kcal mol⁻¹ relative to TS-11-syn + THF). For the anti geometry, the lithium is already 4-coordinate in TS-11-anti and a stable structure containing a second THF could not be located.

Figure 2.

Transition states for deprotonation of 11 by Li(NMe₂)(THF). THF ligands are fogged out for clarity. The “syn-directed” deprotonation, leading to the Z_CN azaenolate, is favored. (ΔH^‡_rel in kcal mol⁻¹ at 0 K.)

For the chiral hydrazone 15, derived from the ACC 4, there are four possible stereoisomers. These are shown in Figure 3(a). Unlike the achiral hydrazone 11, 15 can only adopt a syn conformation. Steric crowding between the rigid bicycloalkane unit and the hydrazone α-methyl group is too severe for anti conformers to be energy minima. The two syn conformers, 15-syn-front and 15-syn-back, correspond to the proposed structures 9a and 9b of Scheme 3, respectively (the terms “front” and “back” refer to the orientation of the carbonyl group). Consistent with the earlier model,ⁱⁱⁱ 15-syn-front is 3.5 kcal mol⁻¹ more stable than 15-syn-back. The destabilization of 15-syn-back can be traced to the conformation about the N–C₄ bond in the oxazolidinone ring. Although the two conformers of 15-syn are formally related by rotation about the N–N bond, their interconversion also induces a change in configuration at the ring nitrogen. This is depicted in Figure 3(b), which shows Newman projections along the N–C₄ bond after the N=CMe₂ group has been replaced by a hydrogen atom (green). The N–C₄ bond in 15-syn-back shows substantial eclipsing, with CNCC and NNCC dihedral angles of 12° and 22°, respectively. By contrast, the smallest dihedral angles about N–C₄ in 15-syn-front are 29° and 56°. The energy difference between the two structures in Figure 3(b) is 2.2 kcal mol⁻¹ (ΔE).

Figure 3.

(a) Conformers of the hydrazone 15 and (b) Newman projections showing the conformation about the N–C₄ bond in the oxazolidinone ring [N=CMe₂ unit has been removed and replaced by H (green) at a distance of 1 Å, while the remaining atoms were held fixed]. Only the syn conformation is available to 15, because the anti conformers would be subject to severe steric interactions between the auxiliary and the nearby alkyl group on the hydrazone. The rigid bicycloalkane moiety enforces destabilizing eclipsing interactions about the N–C₄ bond in 15-syn-back.

Transition states for the deprotonation of 15 by coordinated Li(NMe₂)(THF) are shown in Figure 4. The relative energies of the transition states are the same as those of the hydrazones themselves; TS-15-syn-front is 3.5 kcal mol⁻¹ lower in energy than TS-15-syn-back. The barriers relative to reactants (0.5 and 1.1 kcal mol⁻¹, respectively) are similar to those for the achiral TS-11-syn and TS-11-anti. The eclipsing interactions about N–C₄ that were present in 15-syn-back are also present in TS-15-syn-back. There is also a destabilizing steric interaction (shown by the red line) between the hydrazone α-methyl group and the nearby methyl group on the auxiliary. The chiral ACC effectively blocks anti deprotonation and makes the front (β) deprotonation considerably easier than the back (α). The front/back selectivity is calculated to be only marginally affected by solvation. When the transition structures were optimized in THF using the Conductorlike Polarizable Continuum Model (CPCM),^x,xi the preference for TS-15-syn-front was ΔΔH^‡ = 3.3 kcal mol⁻¹ (ΔΔG^‡ = 4.2 kcal mol⁻¹ at 298.15 K). A transition state related to TS-15-syn-front, but containing a second THF in the coordination sphere of Li⁺, was found to be 1.9 kcal mol⁻¹ less stable (ΔG) in the gas phase.

Figure 4.

Transition states for deprotonation of 15 by Li(NMe₂)(THF). Deprotonation from the front (β) face is favored. The transition state for deprotonation from the back is destabilized both by eclipsing interactions about N–C₄ and by steric interactions between the hydrazone and the auxiliary (red line). (Interatomic distances in Å.)

The overall stereoselectivity of the deprotonation–alkylation sequence is determined by the addition of the azaenolate to the electrophile. The geometry of lithium azaenolate 16, and transition states for its reaction with MeCl, are shown in Figure 5. Two THF ligands were included in the coordination sphere of Li⁺. During the alkylation, the ACC can be oriented with its carbonyl group lying in front of or behind the plane of the azaenolate. MeCl can add to either conformer, and can approach from either the front or back. Figure 5 shows four transition states, which correspond to these four possible arrangements of the ACC and MeCl with respect to the plane of the azaenolate. The relative enthalpies and free energies are given below each transition state, both for the gas phase and with a THF solvent model in addition to the two explicit THFs.

Figure 5.

(a) Conformers of the lithium azaenolate 16 and (b) transition states for alkylation of 16 by MeCl. Both the azaenolate and the TSs for alkylation prefer the carbonyl-front conformation. Addition of MeCl to the β face of the azaenolate (TS-16-A) is preferred by 6.5 kcal mol⁻¹ over addition to the α face (TS-16-C) in the gas phase. A similar but smaller preference is retained in solution.

The lithium azaenolate 16 is subject to the same conformational effects described above for the corresponding hydrazone 15. The ring nitrogen is less pyramidal than in the corresponding hydrazone, with an average bond angle of 117° in 16-front and 118° in 16-back. Pyramidalization (109°) at the ring nitrogen has previously been observed in the crystal structure of the SAMP hydrazone 17.^vii The lithium azaenolate 16- back, like its hydrazone precursor, is destabilized by eclipsing interactions about the N–C₄ bond. 16-back is also destabilized by steric interactions between the azaenolate and one of the methyl groups on the auxiliary [red line in Figure 5(a)], similar to those present in the TS for its formation (TS-15-syn-back, Figure 4). These two effects destabilize 16-back by 5 kcal mol⁻¹ relative to 16-front. The front-back difference increases to 7 kcal mol⁻¹ in their TSs for reaction with MeCl. Additionally, regardless of the conformation of 16, there is also a 7 kcal mol⁻¹ preference for MeCl to add to the π-face where the carbonyl group (coordinated to Li⁺) is located (TS-16-A, TS-16-D). Addition to the opposite face (TS-16-B, TS-16-C) is disfavored due to steric repulsion between MeCl and the bicycloalkane group, as indicated by the red lines in Figure 5(b). Thus, the alkylation of 16 takes place exclusively through the lower-energy conformer 16-front, and MeCl adds selectively to the front side (TS-16-A). The large stereochemical preference is the result of both steric effects and Li⁺---Cl⁻ attraction in TS-16-A.

Having established the facial selectivity of alkylation at an unsubstituted azaenolate terminus, we then investigated the alkylation of a substituted azaenolate. Deprotonation at a secondary carbon introduces the additional consideration of E_CC/Z_CC selectivity. We suggested earlierⁱⁱⁱ that E_CC azaenolates are formed preferentially, on the basis that the allylation of a 3-pentanone-derived hydrazone and a cyclohexanone-derived hydrazone both led to ketones that had the same configuration at the newly-formed stereocenter. Transition states for the deprotonation of an unsymmetrical hydrazone (18) by Li(NMe₂)(THF) are shown in Figure 6. The calculated barriers confirm the E_CC selectivity. The TS leading to azaenolate 19-E_CC is favored by 2.9 kcal mol⁻¹ over the TS leading to 19-Z_CC. Similar selectivity for formation of E_CC azaenolates has previously been established for SAMP/RAMP-hydrazones.^vi

Figure 6.

Transition states for deprotonation of the substituted hydrazone 18 by Li(NMe₂)(THF). The TS leading to the E_CC isomer of the azaenolate is 2.9 kcal mol⁻¹ lower in energy than the TS leading to the Z_CC azaenolate.

Once formed, the azaenolate 19-E_CC is unlikely to undergo conversion to the Z_CC isomer. We calculate a C=C rotational barrier of 43 kcal mol⁻¹ for the azaenolate derived from 11, and the barrier for the more-hindered azaenolate 19-E_CC is likely quite higher. The reaction of 19-E_CC with MeCl is calculated to have a similar stereoselectivity to that of 16; front-side addition of MeCl to 19-E_CC is favored by 6.9 kcal mol⁻¹ (ΔΔH^‡_0K) over back-side addition (Supporting Information).

Although the B3LYP gas-phase calculations for alkylations of 16 and 19-E_CC predict the correct major products,ⁱⁱⁱ they also predict higher stereoselectivity than that reported previously. For example, the allylation of hydrazone 20 with allyl bromide (Scheme 5) was reported to give ketone 22 with an er of 96:4, but the gas-phase activation energies for the reactions of the azaenolates 16 or 19-E_CC with MeCl predict the product 22-β would be formed exclusively. This high selectivity decreases only slightly with a bulkier electrophile; for example, the alkylation of 16 by EtCl is calculated to have a stereoselectivity of 5.7 kcal mol⁻¹ (c.f. 6.5 kcal mol⁻¹ for MeCl).

Scheme 5.

Asymmetric α-allylation of 20

The very high predicted gas-phase selectivities prompted us to reappraise the experimental selectivities. In our initial study,ⁱⁱⁱ ers were determined for the ketone products, following hydrolytic cleavage of the auxiliary. We repeated the allylation of 20 (Scheme 5), and this time measured the diastereomer ratio of 21-β to 21-α. HPLC analysis revealed that 21-β and 21-α were formed in a ratio of >99:1.^xii Subsequent hydrolysis of the auxiliary led to 22-β and 22-α in a ratio of 96:4. Thus, despite the reasonably mild conditions and short reaction time, erosion of stereochemical integrity occurs during the hydrolysis. We are now seeking improved conditions for auxiliary cleavage.

The predicted stereoselectivity in solution, however, is nevertheless not as high as in the gas phase. For example, in the reaction of the azaenolate 16 with MeCl, the TS leading to the minor product (TS-16-C) has a larger degree of charge transfer to MeCl (0.35e) than the lowest-energy TS (TS-16-A, 0.29e).^xiii This would be expected to lead to enhanced stabilization of the minor TS in solution. We calculated CPCM free energies of solvation for the gas-phase structures in THF. The calculated value of ΔΔG^‡ in THF at −78 °C is only 1.1 kcal mol⁻¹, corresponding to a dr of 95:5. The same value of ΔΔG^‡ is obtained if the transition structures are fully optimized in the solvent model. Although this value is not expected to be quantitatively accurate (an accurate treatment of solvent effects would require more sophisticated modeling, including treatment of different coordination states for Li⁺), and indeed underestimates the experimental dr, it does indicate that the predicted stereoselectivity is sensitive to solvation.

Conclusion

B3LYP calculations support the model for the stereoselectivity of ketone α-alkylation shown in Scheme 3. The crucial features are that: (i) the conformation of the intermediate azaenolate is controlled by conformational effects in the oxazolidinone ring and by steric repulsion between the chiral auxiliary and the deprotonated group, and (ii) an electrophile reacts preferentially with the lower-energy azaenolate, from the side opposite the bulky bicycloalkane group. These features resemble the mechanism of stereoinduction in the alkylation of SAMP/RAMP hydrazones.^iv,v

Theoretical Calculations

B3LYP calculations^xiv–xvi were performed with Gaussian 03^xvii and Gaussian 09.^xviii The nature of each optimized point was checked by calculation of the vibrational frequencies, and transition states were further verified by IRC calculations.^xix,xx Zero-point energy and thermal corrections were derived from the B3LYP/6-31G(d) frequencies, scaled by Radom’s factors.^xxi The effects of basis set size were investigated through calculations of the transition states TS-11-syn and TS-11-anti with the 6-311+G(2d,p) basis (leaving frequencies unscaled); this raised the selectivity in favor of TS-11-syn from 1.4 to 2.9 kcal mol⁻¹, while the bond lengths involving the transferring proton changed by only 0.01–0.02 Å. The effects of solvation were simulated by means of CPCM calculations^x,xi using UAKS radii. Free energies of solvation were calculated for the gas-phase-optimized geometries, and were added to the gas-phase free energies to obtain the solution-phase free energies. We also performed geometry optimizations for selected species in THF (Gaussian 09). Solution-phase free energies are quoted at 1 mol L⁻¹. Molecular graphics were produced with the CYLview program.^xxii For simplicity, we have only considered monomeric species where the Li⁺ is coordinated by one NMe₂⁻ and one or two THF ligands (for the deprotonation step), or by two THF ligands (for the alkylation step). A fuller treatment would involve adducts having alternative coordination numbers and aggregation states, as well as multiple conformational isomers. Collum^xxiii has shown, for example, that at high THF concentrations, the lithium azaenolate derived from cyclohexanone phenylimine exists predominantly as a monomeric species with three THF ligands coordinated to Li⁺. The structures, aggregation states, and reactivities of lithium enolates and related species have been studied computationally by Pratt.^xxiv In our simple model complexes, sampling of different THF conformations showed energetic variations amounting to a few tenths of a kcal mol⁻¹.

Experimental Methods

Allylation of 20

n-BuLi (2.5 M in hexanes, 100 μL, 0.250 mmol) was added dropwise over ca. 2 min to a stirred and cooled (−78 °C) solution of diisopropylamine (38.2 μL, 0.272 mmol) in THF (1.0 mL) (Ar atmosphere). The mixture was cooled for 30 min with an ice-H₂O bath, and then cooled to −40 °C. A solution of 20 (60.0 mg, 0.227 mmol) in THF (1.0 mL) was added by cannula, with additional THF (2 × 0.3 mL) as a rinse, and the mixture was stirred for 45 min. Allyl bromide (23.7 μL, 0.272 mmol) was then added and stirring was continued for 5 min. The cold bath was removed and the mixture was stirred for an additional 40 min and then partitioned between Et₂O and H₂O. The aqueous phase was extracted with Et₂O (twice) and the combined organic extracts were washed with brine, dried (MgSO₄), filtered, and evaporated under reduced pressure to give crude 21. ¹H NMR (CDCl₃, 400 MHz): δ 5.90–5.70 (m, 1H), 5.18–4.94 (m, 2H), 4.25 (dd, J = 8.1, 4.1 Hz, 1H), 3.18–3.04 (m, 1H), 2.50–2.24 (m, 4H), 2.14–1.80 (m, 4H), 1.76 (t, J = 4.4 Hz, 1H), 1.26–1.32 (m, 2H), 1.23 (s, 3H), 1.16 (s, 3H), 1.13 (t, J = 7.2 Hz, 3H), 0.94 (d, J = 7.0 Hz, 3H). HPLC analysis of this material showed a 99.3:0.7 mixture of 21-β–21-α.^xxv

The crude material was purified via flash chromatography over silica gel using 10:90 EtOAc-hexanes to give 21 as a pure, light-yellow oil (66 mg, 96%). ¹H NMR (CDCl₃, 400 MHz): δ 5.90–5.70 (m, 1H), 5.18–4.94 (m, 2H), 4.25 (dd, J = 8.1, 4.1 Hz, 1H), 3.18–3.04 (m, 1H), 2.50–2.24 (m, 4H), 2.14–1.80 (m, 4H), 1.76 (t, J = 4.4 Hz, 1H), 1.26–1.32 (m, 2H), 1.23 (s, 3H), 1.16 (s, 3H), 1.13 (t, J = 7.2 Hz, 3H), 0.94 (d, J = 7.0 Hz, 3H); ¹³C NMR (CDCl₃, 100 MHz): δ 184.4, 155.5, 136.6, 116.7, 82.9, 73.4, 47.9, 43.1, 37.6, 35.6, 35.1, 26.7, 25.8, 24.8, 21.5, 19.3, 17.3, 10.4; ESI-MS m/z [M + H]⁺ calcd for C₁₈H₂₉N₂O₂: 305.44, found 305.1. HPLC analysis of this material showed exclusively 21-β.^xxv

Hydrolysis of 21-β

p-TsOH·H₂O (83 mg; 0.436 mmol) was added to a stirred solution of 21-β (66 mg, 0.218 mmol) in acetone (2 mL). The mixture was stirred for 15 min and then partitioned between Et₂O and saturated aqueous NaHCO₃. The aqueous phase was extracted with Et₂O (twice) and the combined organic extracts were washed with brine, dried (MgSO₄), filtered, and evaporated under reduced pressure to give a colorless oil. GC analysis of this material showed a 96:4 mixture of 22-β–22-α.^xxv Flash chromatography of the remaining crude material over silica gel using 5:95 Et₂O-pentane gave 22 (25.8 mg, 94%) as a pure, colorless oil. Spectroscopic data was identical to that reported previously.^xxvi