Synthesis of Chiral, Enantiopure Zirconocene Imido Complexes: Highly Selective Kinetic Resolution and Stereoinversion of Allenes, and Evidence for a Stepwise Cycloaddition/Retrocycloaddition Reaction Mechanism

Certain metal – carbon multiple bonds exhibit substantial reactivity toward organic compounds and are important intermediates in many catalytic processes.[1] The analogous reactivity of metal – heteroatom multiple bonds, such as those in metal – imido complexes, has been known for a much shorter time and has only relatively recently been utilized in synthetic applications.[2]

We wish to report a series of highly enantioselective reactions using isolable imido complexes with reactive metal – nitrogen double bonds.[3–9] These involve the reactions of enantiopure (ebthi)(L)Zr=NR complexes with 1,3-disubstituted allenes (L = tetrahydrofuran; ebthi = bis(tetrahydroindenyl)ethane). Our observations a) demonstrate the operation of a general and highly selective kinetic resolution, b) provide the first example of a reaction that can be used to selectively invert the absolute configuration of one (symmetrically or unsymmetrically 1,3-disubstituted) allene enantiomer within a racemic mixture, and c) provide strong mechanistic evidence for the stepwise nature of an organo-metallic cycloaddition reaction.

Previous work in our laboratories has shown that zirconocene – imido complexes undergo cycloaddition reactions with a wide range of unsaturated organic molecules including alkynes,[10] imines,[11] and certain alkenes.[12] Several of these reactions are important steps in catalytic transformations (for example, hydroamination or imine metathesis) mediated by the imido compounds. To extend this chemistry to systems capable of enantioselective reactions, we decided to examine the reactions of chiral disubstituted allenes with zirconium imido complexes bearing the C₂-symmetric ebthi ligand developed by Brintzinger.[13] The chiral imido complexes were synthesized in two steps from the known dimethyl complex 4, as shown in [Eq. (1)].[14]

(1).

In order to investigate the products and diastereoselectivity of the chiral zirconocene – allene reactions, we first examined the stoichiometric reactions of racemic ebthi complexes 1 and 2 with achiral allenes and racemic chiral 1,3-disubstituted allenes. The purple metallacycle rac-6 was formed rapidly upon condensation of allene onto a benzene solution of rac-1 [Eq. (2)]. Production of an azametallacycle with the exocyclic methylene group in the β (as opposed to the α) position was indicated by its ¹H, ¹H NOESY spectrum, which featured a strong NOE between one of the exocyclic methylene protons and the methyl groups on the phenyl ring.

(2).

The reaction of rac-1 with 1 equiv of racemic 1,3-diphenylallene, 3,4-heptadiene, 4,5-nonadiene, or 1,2-cyclononadiene at 23 °C provided a single diastereomeric product in each case, as determined by ¹H and ¹³C NMR spectra ([Eq. (3)]). A bulky imido substituent is required for the selectivity in these transformations, as an 85:15 mixture of diastereomers was formed from the reaction of the p-tolylimido complex rac-2 with diphenylallene. These results were encouraging because they suggested that the reactions are stereospecific: (S,S)-1 reacts preferentially with the (S) allene, and (R,R)-1 with the (R) allene, in order to give the diastereoselectivity observed. As the data summarized below shows, the actual situation is more complicated.

(3).

Reactions of 1 with the unsymmetrically substituted allenes 1-phenyl-1,2-butadiene and 1-phenyl-1,2-pentadiene can potentially produce a mixture of diastereo- and regioisomers. However, addition of 1 equiv of 1-phenyl-1,2-butadiene to a solution of rac-1 produced only the regioisomers rac-11 and rac-12 (1:1 ratio determined by ¹H NMR [Eq. (4)]). Therefore, even though the reaction is not regioselective, it is still highly diastereoselective. Similar results were obtained in the reaction of rac-1 with 1-phenyl-1,2-pentadiene.

Deeper insight into the actual course of these reactions was provided by examining the stereochemistry of the allene cycloaddition reactions with enantiopure 1. First, the reaction of enantiopure (S,S)-1 with the racemic allenes used in excess provided an effective kinetic resolution even at room temperature. Addition of 1.8 equiv of racemic 1,3-diphenylallene to a benzene solution of (S,S)-1 at 23 °C resulted in consumption of approximately 50% of the allene.[15–16] The unreacted 1,3-diphenylallene was highly enriched in the (R) enantiomer (>98% ee). The results of enantioselective capture experiments, in which (S,S)-l was treated with other chiral allenes, are shown in Table 1. Notably, the cycloaddition reaction is selective for both symmetrically and unsymmetrically 1,3-disubstituted allenes with high ee values in most cases.[17–18]

Table 1.

Enantioselective capture reactions of (R₁)HC=C=CH(R₂) with (S,S)-1.

Entry	R1	R2	Allene consumed [%]	ee[a] [%]	Configuration of unreacted allene
1	cyclo-(CH2)6		54	78	(R)
2	C3H5	C3H5	59	>98	(R)
3	Ph	Me	61	94	(R)
4	Ph	Et	53	>98	(R)
5	Ph	Ph	50	>98	(R)

^[a]Enantiomeric excess of the unreacted allene. See ref. [16] for the method of ee determination.

(4).

We propose that the selectivity observed in these cyclo-additions is the result of mutually reinforcing stereocontrol imposed by the combination of the ebthi ligand and the bulky imido substitutent. The ebthi ligand allows the imido complex to add preferentially onto one face of the reacting π bond,[19] while the N–Ar group favors reaction of the allene enantiomer in which one olefin substituent is oriented away from the imido ligand in the cycloaddition transition state (Scheme 1). This model is supported by the observation that the reactions of 2, which contains a smaller imido substituent, appear to be substantially less selective than the reactions of 1. The proposal is also consistent with the absolute configuration observed in the enantioselective cycloaddition reactions.

Scheme 1.

Alternative structures of the π complexes formed by interaction of (S,S)-(ebthi)Zr=NAr with the two enantiomers of a 1,3-disubstituted allene.

A single metallacycle diastereomer is, by NMR spectroscopy, observed to be the major organometallic product in the enantioselective capture reactions. In separate experiments, (S,S,R)-9 and (S,S,R)-10 could be isolated by treatment of (S,S)-1 with slightly more than 2 equiv of either 4,5-nonadiene or 1,2-cyclononadiene followed by removal of the excess allene in vacuo [Eq. (5)]. The faster reacting allene enantiomer was then regenerated from the metallacycle by treating it with the parent allene, 1,2-propadiene (C₃H₄). Thus, when (S,S,R)-9 was treated with 1,2-propadiene, retro-cycloaddition occurred slowly at 23 °C to produce (S,S)-8 and (S)-4,5-nonadiene that was highly enantioenriched (93 % ee, >95% recovery). Similarly, (S)-1,2-cyclononadiene (85% ee, >95% recovery) was produced from the reaction of 1,2-propadiene with (S,S,R)-10. By this method, it is possible to obtain good separation of allene racemates: The slower reacting allene enantiomer is recovered from the initial cycloaddition and the faster reacting allene enantiomer is recovered by displacement from the product metallacycle.

(5).

As the allene cycloaddition reactions appeared to be stereospecific, we expected that the reaction of the enantiopure imido complex with only 1 equiv of a racemic disubstituted allene (for example, 1,2-cyclononadiene) would allow reaction of the imido complex with the slower reacting allene enantiomer, once the faster reacting enantiomer had been consumed. This might then lead to a mixture of metallacycle diastereomers. We were therefore surprised to find that (S,S,R)-10 was the only detectable product of the reaction between 1 equiv of racemic 1,2-cyclononadiene and (S,S)-1 [Eq. (6)]. No intermediates were observed when this reaction was followed by ¹H NMR spectroscopy. Following treatment of the product mixture with 1,2-propadiene, (S)-1,2-cyclononadiene (84 % ee, 93 % yield) evolved.

(6).

The only way we are able to rationalize this result is to assume that when the slower reacting (R) allene reacts with (S,S)-1, its absolute configuration is inverted. To further test this conclusion, we treated (S,S)-1 with 1,2-cyclononadiene enriched in the (R) enantiomer (67% ee). This again led to (S,S,R)-10, and condensation of 1,2-propadiene into the solution regenerated (S)-1,2-cyclononadiene (85% ee). Therefore, we suggest that: a) the faster reacting (S) enantiomer of 1,2-cyclononadiene forms a metallacycle, which is then regenerated in the retrocycloaddition with retention of its absolute configuration, but b) the slower reacting (R) enantiomer reacts to produce the same metallacycle as that formed for the (S) enantiomer, which is regenerated with overall inversion of its absolute configuration. The allene cycloaddition is therefore highly stereoselective but not stereo-specific.[20] This system allows quantitative (rather than just 50%, as would be possible in a conventional kinetic resolution) conversion of racemic allene into material that is highly enriched in the (R) enantiomer. Similarly, the (R,R) imido complex can be used to convert racemic allene quantitatively into material that is highly enriched in the (S) enantiomer. Interconversion of (R) and (S) allenes is very slow under these conditions.

Other reactions of (S,S)-1 with racemic chiral allenes behave analogously. The reaction of (S,S)-1 with 1 equiv of racemic 1-phenyl-1,2-butadiene at 23°C produced (S,S,R)-11, (S,S,R)-12, and a third, unidentified metallacycle diastereomer in a 1.0:1.0:0.5 ratio, respectively (¹H NMR). After 1 h, all of the disubstituted allene was consumed in the reaction. Treatment of this product mixture with 1,2-propadiene at 75 °C quantitatively converted it to (S,S)-8 and free 1-phenyl-1,2-butadiene. Chiral GC analysis indicated that the released disubstituted allene was enantioenriched ((S) 43% ee, 90% recovery). Enantioenriched 1-phenyl-1,2-pentadiene ((S) 33% ee, >95% recovery) was also obtained by consecutive treatment of racemic 1-phenyl-1,2-pentadiene with (S,S)-1 and 1,2-propadiene.

Previous experimental work has suggested that such organometallic cycloaddition reactions are concerted and stereospecific,[21] and several theoretical studies support this postulate. [22] However, we see no economical way to explain our unusual observations by assuming that the imido – allene cycloadditions are concerted processes. We therefore offer the stepwise mechanism illustrated in Figure 1 for (S,S)-1 and 1,2-cyclononadiene, which proceeds through the diradical intermediate C,[23] as a working explanation for our results. The left-hand side of the diagram accounts for the kinetic resolution. It assumes that when the (S,S)-imido complex approaches the allene, perhaps to form the initial π complexes A and B, coordination of the (S) enantiomer of the allene gives the lower energy complex A because the two allene substituents can lie away from the N–Ar group and in the open quadrant of the metal coordination sphere. In complex B, however, the requirement that the external allene substituent lies away from the N–Ar moiety forces the internal allene substituent into the more sterically congested quadrant of the metal coordination sphere, to result in a higher energy complex. Subsequent formation of the C–N bond with rotation about one of the C–C bonds of the developing allyl radical results in the formation of intermediate C. However, the transition states A^≠ and B^≠, which lead to C and are not planar about the three carbon atom fragment, must still retain some of the energetics associated with the initial π complexes. [24] This accounts for the observation that one allene enantiomer reacts more quickly than the other. Once C has been reached, the chirality of the allene has been destroyed. Therefore a second enantioselective step is required to explain the observed stereochemistry of the metallacycle and the recovered less reactive. To account for this, we assume that ring closure in intermediate C can take place by rotation of the internal R group “up” into the open quadrant of the coordination sphere—which is favored—but not “down” into the more hindered quadrant. [25] Thus the same metallacycle (S,S,R)-10 is formed from both allene enantiomers and the (R) allene enantiomer is always regenerated from this metallacycle.[26]

Figure 1.

Reaction coordinate diagram illustrating the proposed mechanism of the reaction of imido complex (S,S)-1 with both enantiomers of 1,2-cyclononadiene.

According to this mechanism, the selectivities of both the enantioselective capture and the retrocycloaddition reactions are a function of the free energy difference (ΔG^≠_A^≠ − ΔG^≠_B^≠) for transition states A^≠ and B^≠. Therefore, if this model is correct, the selectivities of the enantioselective capture and retrocycloaddition reactions should be identical. Analysis of the cyclononadiene cycloaddition and retrocycloaddition reactions confirms that this is indeed the case. The selectivity of the enantioselective capture reaction in which (S,S)-1 is treated with excess 1,2-cyclononadiene can be estimated using [Eq. (7)], where s is the selectivity of the reaction, C is the

$$s=\frac{\text{ln}[(1-C)(1-ee)]}{\text{ln}[(1-C)(1+ee)]}$$ (7)

percentage of allene consumed in the reaction, and ee is the enantiomeric excess of the remaining allene.[27] Using the values in Table 1, a selectivity of 12.0 is calculated. This corresponds to a difference in the activation energies (ΔΔG^≠) of the reaction of the slower and faster reacting enantiomers of 1.5 kcal mol⁻¹ at 298 K. A selectivity of 11.5 (ΔΔG^≠ = 1.4 kcal mol⁻¹, 298 K) for release of the chiral allene can be calculated directly from the enantiomeric ratio (92:8) of the cyclononadiene enantiomers following the retrocycloaddition reaction. The similarity of these values strongly supports the postulate that the formation and fragmentation of these metallacycles occur via the same pathway.

In summary, 1 undergoes highly enantioselective cycloaddition reactions and, in certain cases, this system allows conversion of an allene racemate into a mixture enriched in one enantiomer. The absolute sense of stereochemistry of these reactions appears to be so predictable that we believe they can be used to assign the absolute configurations of new, optically active allenes.

Experimental Section

(rac-1): A solution of nBuLi (0.50 mL of 1.6 m solution in Et₂O) was added dropwise to a stirred solution of 2,6-dimethylaniline (98 mg, 0.80 mmol) in Et₂O (5 mL) at RT to form a yellow solution. After 10 min, the yellow solution was added dropwise to a stirred suspension of rac-5 (327 mg, 0.80 mmol) in THF (8 mL). This mixture was stirred overnight and the volatile materials were removed in vacuo to give a yellow residue. The yellow residue was dissolved in Et₂O and filtered. The volatile materials were removed in vacuo and the yellow solid was dissolved in THF (5 mL) and heated to 75 °C overnight in a sealed ampoule. The resulting orange solution was cooled to −30 °C overnight to precipitate 265 mg of rac-1 (65 %) as orange crystals.

(S,S)-1: Solid LiNH(2,6-CH₃-C₆H₃) was added to a solution of (S,S)-5 (299 mg, 0.73 mmol) in benzene. The resulting yellow solution was stirred for 1 h, the volatile materials were removed in vacuo, and the remaining solid was dissolved in pentane and filtered. The volatile materials were removed in vacuo and the yellow solid was dissolved in THF (7 mL) and heated to 75 °C overnight in a sealed ampoule. The volume of the resulting orange solution was reduced in vacuo and crystallization from THF/Et₂O at −30°C provided (S,S)-1 (210 mg, 53%). ¹H NMR (C₆D₆): (δ = 7.23 (d, J = 7.3 Hz, 2H), 6.75 (t, J = 7.3 Hz, 1H), 6.37 (d, J = 2.7 Hz, 1 H), 5.78 (d, J = 2.7 Hz, 1 H), 5.35 (d, J = 2.4 Hz, 1 H), 5.30 (d, J = 2.4 Hz, 1 H), 3.63 (m, 2H), 3.38 (m, 2H), 3.05 (m, 1H), 2.83 (m, 1H), 2.67 (m, 3 H), 2.51 (m, 11 H), 2.21 (m, 2 H), 1.72 (m, 1 H), 1.46 (m, 6 H), 1.15 (m, 5H). ¹³C{¹H} NMR ([D₈]THF; assignments made by DEPT spectrum): quaternary: δ = 157.2, 133.1, 130.1, 128.3, 127.5, 117.9, 115.7; methine: δ = 127.7, 115.0, 109.6, 108.2, 103.7, 100.5; α-methylene in coordinated THF: δ = 68.4; methyl: δ = 26.55; aliphatic CH₂: δ = 28.9, 28.2, 26.4, 25.8, 25.4, 25.3, 24.9, 24.8, 24.7, 24.0. IR (Nujol): $\overline{\nu }$ = 3061, 2881, 1582, 1315, 1005, 844, 761 cm⁻¹. Elemental analysis: calcd for C₃₂H₄₁NOZr: C 70.28, H 7.56, N 2.56; found: C 70.07, H 7.83, N 2.50. M.p. (rac)-1: 145–158°C; (S,S)-1:170–185 °C. [α]²⁵_D = − 665.5° (c = 5.27 mgmL⁻¹, C₆D₆).