Modern Organocopper Chemistry

2688 Copper-Mediated Enantioselective Substitution Reactions

Tab. 8.1. Dependence on double bond geometry in 10.

in combination with CuBr for 9 with R1 ¼ i-Pr and Y ¼ O. Although the azomethine group is crucial for the selectivity, group Y can be changed from O to S or CH2 without any large drop in obtained ees. Substrate 9 with R1 ¼ EtO2C was not suitable under the reaction conditions studied, with racemization of the heterocyclic stereocenter taking place.

As in the case of the substitution reaction of compound 7, the absolute configuration of the product depends on the double bond geometry of the starting material, as shown by the example in Tab. 8.1.

The selectivity in this process is governed by preliminary chelation of the RCu species by the azomethine group and the allylic double bond. The proposed chelates for the cases of (S)-(Z )-10 and (S)-(E )-10 are shown in Fig. 8.1.

8.2.2

Chiral Auxiliary that is Cleaved o after the Reaction

Reaction between C2 symmetric diols and a; b-unsaturated aldehydes yield chiral ethylenic acetals that undergo copper-mediated substitution reactions. With aryl or

Fig. 8.1. Proposed chelate structures.

8.2 Allylic Substitution 269

vinylcopper reagents this reaction, as studied by Alexakis et al. (Scheme 8.11), is highly anti SN20-selective. With alkyl copper reagents, however, a mixture of SN20 and SN2 substitution results [23, 24]. The copper approaches from the face of the double bond that is on the side of the equatorial substituent in the acetal, and the CaO bond nearest to the axial substituent is cleaved. The initial SN20 product is an enol ether, which is hydrolyzed to a chiral b-substituted aldehyde. The reaction sequence starting from an a; b-unsaturated aldehyde can be viewed overall as a conjugate addition of RLi.

Scheme 8.11. Reactions between an ethylenic acetal and organocopper reagents.

With the reagent PhCu in the presence of the additives BF3 and PBu3, ees of up to 95% were obtained, while values of up to 85% were achievable with a vinyl copper reagent. Chiral dienic acetals have also been studied; three regioisomeric products could be obtained in this case as the result of SN2, SN20, or SN200 attack of the organocopper reagent [25]. Mixtures were indeed obtained with alkyl copper reagents, but PhCu BF3 resulted in formation of only the SN20 and SN200 products, with selectivity for the latter (Scheme 8.12).

Scheme 8.12. Substitution of a dienyl acetal.

272 8 Copper-Mediated Enantioselective Substitution Reactions

Fig. 8.2. Transition state model for the enantioselective substitution of 17.

On the basis of this conclusion and on NMR studies of complexes of 17b with Lewis acids, a transition state model to explain the observed selectivity was proposed. This involved initial complexation of a cuprate lithium ion to the three different heteroatoms in the substrate, followed by formation of a d-p complexation product from the less hindered si face, the re face being shielded by the pyrrolidine ring (Fig. 8.2).

8.2.3

Catalytic Reactions with Chiral Ligands

Compared to the intensive and successful development of copper catalysts for asymmetric 1,4-addition reactions, discussed in Chapt. 7, catalytic asymmetric allylic substitution reactions have been the subjects of only a few studies. Di culties arise because, in the asymmetric g substitution of unsymmetrical allylic electrophiles, the catalyst has to be capable of controlling both regioselectivity and enantioselectivity.

In 1995, Ba¨ckvall and van Koten reported the first example of a catalytic, enantioselective SN20 substitution of a primary allylic acetate in the presence of a chiral copper complex [28, 29].

The copper(I) arenethiolate complexes 19 [30], first developed and studied by van Koten’s group, can be used as catalysts for a number of copper-mediated reactions such as 1,4-addition reactions to enones [31] and 1,6-addition reactions to enynes [32].

Initial studies on the application of these catalysts to allylic substitution reactions showed that the arenethiolate moiety functions as an excellent nontransferable group, and that the regioselectivity can be completely reversed by suitable changes in the reaction parameters [33]. If the reaction between geranyl acetate and n- BuMgI was carried out in THF at 30 C with fast addition of the Grignard reagent to the reaction mixture, complete a selectivity was obtained. Raising the tempera-

8.2 Allylic Substitution 273

ture to 0 C and use of Et2O as solvent, with slow addition of the Grignard reagent, gave 100% of the g product (Scheme 8.17).

Scheme 8.17. Control of regioselectivity with catalyst 19a.

These catalysts also give a remarkable reversal in leaving group ability. An allylic acetate becomes more reactive than an allylic chloride in the presence of 19a, a fact that can be explained by chelate formation with the catalyst and Grignard reagent, with the acetate group becoming activated by coordination of oxygen to magnesium [33b].

The use of the chiral catalyst 19b for asymmetric allylic substitution of allylic substrates has been studied in some detail (Scheme 8.18) and, under g-selective reaction conditions, asymmetric induction was indeed obtained [28, 34].

Scheme 8.18. Enantioselective substitution with catalyst 19b.

To optimize the enantioselectivity it was necessary to use a rather high catalyst loading (ca. 15 mol%), with reactions being carried out at fairly low substrate concentrations, with slow addition of the Grignard reagent over 2 hrs. The e ect of the leaving group was studied using substrates 20, in their reactions with n-BuMgI. Both the acetate 20a and the pivalate 20b underwent highly regioselective reactions, with 34% ee for the acetate and 25% ee for the more bulky pivalate. Trifluoroacetate (20c) or diethylphosphate (20d) as leaving groups resulted in slightly lower regioselectivities (ca. 90:10) and the ees were severely diminished to around 10%. The substituent on the allylic double bond had only a minor influence on the ee; PhOCH2 (20a) and cyclohexyl (21) gave ees of 34 and 41% respectively. A slightly lower ee of 28% was obtained with cinnamyl acetate (22). The mode of addition was important for the outcome, the best results being obtained when both the Grignard reagent and the substrate were added slowly to the reaction mixture. With this

274 8 Copper-Mediated Enantioselective Substitution Reactions

Fig. 8.3. Proposed chelate structure for the catalytically active intermediate.

technique, the ee in the case of the reaction between cyclohexyl-substituted allylic acetate 21 and n-BuMgI was 42%. This implies that a 1:1 ratio of substrate to Grignard reagent at all times is important for the selectivity. Excess substrate can disrupt the bidentate coordination necessary for the proposed chelate. The difference here, however, was very small in comparison to the situation when the Grignard reagent alone was added over 2 h. A still larger di erence was observed when the substrate was added to a mixture of catalyst and n-BuMgI, conditions favoring formation of a homocuprate, R2CuM. In that case only 18% ee was achieved. The reaction has to be performed at a rather high temperature if maximum enantioselectivity is to be achieved. Reaction temperatures of 0 C or 20 C produced similar ees, but an ee of only 7% was obtained at a lower temperature ( 20 C). This supports the hypothesis that chelate formation is important for the enantioselectivity.

The results obtained can be explained in terms of a catalytic intermediate made up of a chelate between Grignard reagent, catalyst, and substrate. The allylic substrate anchors in a bidentate fashion, through carbonyl coordination to magnesium and copper-alkene p-interaction, as represented schematically in Fig. 8.3. The chelate constitutes a rigid structure, incorporating a six-membered ring with a chiral magnesium atom. The chelate shown would produce preferential coordination from the face of the olefin indicated in Fig. 8.3, in accord with the observation that R ligands result in R products.

The coordination of the acetate in this fashion should result in enhanced leaving group reactivity, while the e ect of changes in the leaving group on enantioselectivity further supports the idea of chelate formation. The more bulky pivalate should give a less stable chelate, and a lower ee is indeed observed. The electronwithdrawing trifluoromethyl group in the trifluoroacetate moiety would weaken coordination and give a less stable chelate, which would explain the low enantioselectivity (10% ee) with the allylic trifluoroacetate. (It is also possible that the high reactivity of trifluoroacetate as a leaving group results in reaction before chelate formation takes place.) The same arguments also apply to the phosphate leaving group.

The reaction of cyclohexyl-substituted allylic acetate 21 with di erent Grignard reagents was investigated [34]. As already mentioned, a 41% ee had been obtained with n-BuMgI. Changing the counter-ion in the Grignard reagent to Br , under otherwise identical reaction conditions, gave an ee of 50%. The sterically hindered Grignard reagent Me3SiCH2MgI underwent only slow reaction, giving a moderate

8.2 Allylic Substitution 275

yield of the g product, but the observed ee, 53%, was the highest so far obtained with catalyst 19b.

To study the e ect of conformationally more rigid substrates, some cyclic allylic esters (23 and 24) were employed as substrates. Reaction of these with n-BuMgI, employing 19b as catalyst, produced very low ees, however (Scheme 8.19) [35].

Scheme 8.19. Reactions of cyclic allylic esters 23 and 24, with catalysis by 19b.

To investigate the e ect of the substituents in the arenethiolate structure, four di erently substituted copper arenethiolates, 25–28, were tested as catalysts, but very low ees were obtained in all cases [34]. The oxazolidine complex 26, developed by Pfaltz et al. [36] and used successfully in asymmetric conjugate addition reactions to cyclic enones, gave a completely racemic product with allylic substrate

20a.

To avoid the di culties in handling the highly air-sensitive copper arenethiolates, a method for their preparation and utilization in situ has been developed, the arenethiol 29 being deprotonated with n-BuLi and mixed with a copper(I) salt to yield the active catalyst [34].

Use of this technique results in an equivalent of lithium halide being present in the reaction mixture, unlike when the isolated copper arenethiolates are employed. Lithium salts can have very profound e ects on copper-mediated reactions, but in this case a similar ee (40%) and complete g selectivity were still obtained for the reaction between 21 and n-BuMgI when the catalyst was prepared from CuI. Nei-

276 8 Copper-Mediated Enantioselective Substitution Reactions

ther a change of the Cu:ligand ratio to 1:2 nor an increase in the temperature (cf. the work with the preformed catalyst) a ected the outcome of the reaction. The e ect of the arenethiolate ligand on the reactivity was confirmed by performing the reaction with only CuI as catalyst, in the absence of the ligand. In this case, the allylic acetate 21 was partly recovered, and formation of the corresponding alcohol was observed, which indicates that the reaction was much slower. The regioselectivity was also no longer complete (g=a ¼ 95:5). The source of the copper can also have a dramatic influence on the stereochemical outcome; a change from CuI to CuBr SMe2 resulted in an ee of only 7%. This can be explained in terms of coordination of the dimethyl sulfide to copper, hampering formation of the catalytic intermediate. CuCl could be employed with the same e ciency as CuI, but Cu(OTf )2 gave a lower enantioselectivity.

Investigation of di erent Grignard reagents was also carried out. In contrast to the result obtained with the isolated catalyst 19b, the in situ generation technique here gave a lower ee for BuMgBr (30% ee) than for BuMgI (40% ee). Use of CuBr instead of CuI allowed this ee to be increased somewhat, to 36%. Some bulkier Grignard reagents, such as i-PrMgI, i-PrMgBr, i-BuMgBr, and Me3SiCH2MgI, were also investigated, but no ees higher than 40% could be obtained. No allylic substitution at all was observed with PhMgI. Cinnamyl acetate (22) as the substrate gave slightly lower ees than obtained with 21, in line with the results with the preformed catalyst. Variation of the ligand structure (as in 30 and 31) produced lower ees than obtained with 29. Use of ligand 30 resulted in a very low ee of 10% for the reaction between 21 and n-BuMgI, but 31 gave a reasonable ee of 35%. Interestingly, the major enantiomers were of opposite configurations when (R)-29 and (R)-31 were used.

The moderate ees obtained with the copper arenethiolate ligands discussed above prompted a search for new chiral ligands for use in asymmetric allylic substitution reactions. The binaphthol-derived phosphoramidite ligand 32, used successfully by Feringa et al. in copper-catalyzed 1,4-addition reactions [37], was accordingly tested in the reaction between 21 and n-BuMgI.

8.2 Allylic Substitution 277

The presence of ligand 32, however, resulted in much slower allylic substitution [38], as could be seen by the formation of large amounts of the alcohol produced by carbonyl attack of the Grignard reagent on the acetate. SN20 selectivities were also lower than those obtained with copper arenethiolate catalysts. Optimization of the conditions (10% each of Cu(OTf )2 and 32, slow addition of n-BuMgI in Et2O at20 C) made it possible to obtain a 97:3 ratio of SN20 and SN2 products with less than 10% attack on the carbonyl, but the SN20 product was racemic [35]. However, it cannot be ruled out that this class of ligands might be useful for the allylic substitution reaction under reaction conditions di erent to those tested.

Chiral ferrocenes have received much attention as ligands in metal-catalyzed reactions [39], but their use in copper chemistry has been very limited [40, 41]. The ferrocene moiety o ers the possibility of utilizing both central and planar chirality in the ligand. By analogy with the copper arenethiolates described above, ferrocenyl copper complex 33 (Scheme 8.20) is extremely interesting.

Scheme 8.20. Ferrocene thiolates.

The synthesis of the corresponding ferrocene thiol 36 was therefore undertaken (Scheme 8.20) [42]. This thiol proved to be too unstable and could not be isolated, but the precursor lithium thiolate 35 could be isolated and stored under an argon atmosphere. Treatment of 35 with CuI produced a catalytically active species that gave up to 64% ee in the reaction between allylic acetates and n-BuMgI (Scheme 8.21). A rather large ratio of ligand to copper gave better results; it was concluded that this was due to the low stability of the ligand towards oxidation.

Scheme 8.21. Allylic substitution in the presence of ferrocene ligand 35.