Broad-spectrum kinetic resolution of alcohols enabled by Cu–H-catalysed dehydrogenative coupling with hydrosilanes

The enantioselective silylation of racemic alcohols, where one enantiomer reacts faster than the other, is an alternative approach to established enzymatic and non-enzymatic acylation techniques. The existing art is either limited to structurally biased alcohols or requires elaborate catalysts. Simple substrates, such as benzylic and allylic alcohols, with no coordinating functionality in the proximity of the hydroxy group have been challenging in these kinetic resolutions. We report here the identification of a broadly applicable chiral catalyst for the enantioselective dehydrogenative coupling of alcohols and hydrosilanes with both the chiral ligand and the hydrosilane being commercially available. The efficiency of kinetic resolutions is characterized by the selectivity factor, that is, the ratio of the reaction rates of the fast-reacting over the slow-reacting enantiomer. The selectivity factors achieved with the new method are good for acyclic benzylic alcohols (≤170) and high for synthetically usefully cyclic benzylic (≤40.1) and allylic alcohols (≤159).

Our laboratory had approached the problem from a different angle. We had used Cu-H catalysis to couple alcohols and hydrosilanes with release of dihydrogen (Strategy 3 and Fig. 2a) 24,25 . The resulting silyl ether is usually considered waste in the catalytic generation of copper(I) hydride-reducing agents. At that time, we had chosen the dehydrogenative Si-O coupling because it allowed us to employ silicon-stereogenic hydrosilanes as resolving reagents (Fig. 2b). Unlike chlorosilanes, these react without racemization with hydroxy groups. By this, we accomplished the reagent-controlled kinetic resolution of alcohols with an achiral monodentate phosphine ligand at the copper(I) atom [26][27][28][29][30] . We were later able to turn this kinetic resolution into a catalyst-controlled process with a chiral monodentate ligand and achiral hydrosilanes 31 . However, both transformations required alcohols with pending donors (TS1 and TS2, Fig. 2c), making two-point binding of the substrate the salient feature of these coupling reactions. All subsequent attempts to extend this methodology to monools had failed for years 32,33 until we discovered that a chiral bidentate ligand together with trialkylsilanes having long aliphatic chains lead to high selectivity factors (TS3, Fig. 2c). We disclose here the Cu-H-catalysed enantioselective silylation of structurally non-biased alcohols, including synthetically useful allylic alcohols, with both a commercially available catalyst and hydrosilane.

Results
Catalyst identification and optimization. An extensive screening of chiral ligands finally culminated in the identification of commercial (R,R)-Ph-BPE [L1, 1,2-bis((2R,5R)-2,5-diphenylphospholano)ethane] as a superior ligand in the catalytic asymmetric Si-O coupling of 1-phenylethan-1-ol and structurally related congeners (see Supplementary Table 1 for the complete ligand survey) 32,[34][35][36][37] . Systematic variation of the copper(I) source, base and solvent did not reveal any evidence of a trend, and these data are collected in Supplementary Tables 2-4. The CuCl-NaO t Bu system in toluene was subsequently used in the further optimization of the model reaction 1a-3a/1a ( Table 1). The selectivity factor was low with Ph 3 SiH (2a, s ¼ 2.96, entry 1) but at least moderate with MePh 2 SiH and Me 2 PhSiH (2b, s ¼ 5.52, entry 2 and 2c, s ¼ 6.33, entry 3). However, any steric and electronic modification of the aryl groups in these diarylmethyl-and aryldimethylsilanes had little effect (sE5 and sE6, respectively), if at all resulting in less reactive hydrosilanes (see Supplementary Table 5 for the collection of tested hydrosilanes). We then found that linear alkyl substituents instead of the methyl group(s) dramatically improved the selectivity factor, reaching promising s values of 10.0 with 2d (entry 4) and 10.6 with 2e (entry 5). This prompted a R 1 R 2 Entry Hydrosilane Ligand Conversion of alcohol (%) * ee of silyl ether (%) w ee of unreacted alcohol (%) z Selectivity factor s y 1 2a  us to (re-)investigate trialkylsilanes as coupling partners. We had initially excluded these from the present study because of their lack of reactivity in our earlier catalyst-controlled kinetic resolution of donor-functionalized alcohols 31 . To our delight, trialkylsilanes 2f-2i consistently yielded selectivity factors above 10 (entries 6-9), and the best value was obtained for 2h with n-butyl groups (s ¼ 14.3, entry 8). Branching close to the silicon atom as in 2j-2l was detrimental (entries [10][11][12], and tert-butyl substituents were generally not accepted (not shown). Bn 3 SiH performed also well (2m, s ¼ 11.6, entry 13) but was inferior to n Bu 3 SiH (2h, cf. entry 8). Reactions were routinely run for 18 h to achieve synthetically useful conversions. To boost the reactivity 28 , we tested cyclic and, hence, more Lewis-acidic hydrosilane 2n but the enormous reactivity gain was at the expense of selectivity (s ¼ 3.51, entry 14). Also, alkoxy-substituted hydrosilanes, such as (EtO) 3 SiH (2o), reacted rapidly yet without any asymmetric induction (entry 15). Cognate ligands (S,S)-Me-BPE (L2) and (R,R)-i Pr-BPE (L3) with different R groups either led to a catalytically inactive copper(I) complex (entry 16) or a clearly lower s value (entry 17) with n Bu 3 SiH (2h) (cf. entry 8).
Scope and limitations. We continued by exploring the substrate scope with this readily accessible catalytic set-up (CuCl/L1-NaO t Bu and n Bu 3 SiH , Figs 3 and 4). An analysis of the steric effects in 1-phenylethan-1-ol derivatives showed that monosubstitution of the aryl group in any of the three available positions with methyl groups is not significantly influencing the reaction outcome; selectivity factors for 1b-e were in the same range, just slightly better in the case of the ortho-substituted substrate (Fig. 3b). Additional methyl substitution as in 1f and 1g generally had little effect but selectivity factors increased substantially with CF 3 (as in 1h) or OMe (as in 1i) instead of the methyl groups. The substituent effect was dramatic for substrates with both ortho positions occupied; selectivity factors for 1j and 1k were exceedingly high (Fig. 3b). When increasing the size of the alkyl group at the carbinol carbon atom, Me (1a/1b)oEt (1l)oBn (1m)o i Pr (1n), the selectivity factor collapses (12.2/14.6411.649.93443.00) while maintaining sufficient reactivity; this effect is also seen for mesityl-substituted 1o but 81.4 is still an excellent s value (highlighted by grey ovals, Fig. 3b). Both band a-naphthylsubstituted derivatives 1p and 1q fit into the observed selectivity pattern (Fig. 3b). The selectivity factors for the acyclic benzylic alcohols 1a-q were uniformly lower than those obtained with Song's Brønsted-acid catalyst (Strategy 2, Fig. 1) 22 . Conversely, cyclic substrates 1r-w not reported by Song 22 afforded synthetically valuable selectivity factors independent of the ring size, 1r-t, and with excellent functional-group tolerance, 1u-w (Fig. 3c). Our results compare favourably with those described   by Wiskur for the same class of compounds using chlorosilanes (Strategy 1, Fig. 1) 17 .
We also investigated the resolution of isomerically pure 2-phenylcyclohexan-1-ols trans-4 and cis-4 with reasonable success (Fig. 3d). Various types of these compounds had been subject of a dedicated study by Wiskur (Strategy 1, Fig. 1) 21 , and Wiskur found that, from an isomeric mixture of 4, the isourea-based catalyst preferentially selects the trans over the cis relative configuration in the kinetic resolution of 4. Our catalytic system showed the same preference in the individual experiments. Both kinetic resolutions were slow, reaching useful conversion after several days only for trans-4. The selectivity factor was markedly higher than that achieved by Wiskur 21 (s ¼ 29.1 versus s ¼ 10).
While both acyclic 22 and cyclic 17 benzylic alcohols had been resolved by other catalytic methods before, just one example of an acyclic allylic alcohol, (E)-1,3-diphenylprop-2-en-1-ol, was described 22 . Allylic alcohols are however ubiquitous synthetic building blocks and, as such, particularly attractive substrates. Without making any changes to our catalytic set-up, allylic alcohols were equally amenable to this enantioselective alcohol silylation but underwent competing partial alkene reduction 38 . This issue was overcome by the substoichiometric addition of a sacrificial, more reactive alkene, styrene 39,40 (Fig. 4a). With this measure, representative allylic alcohols 6a-d were resolved with excellent selectivity factors (Fig. 4b). Cyclic systems with exo-and endocyclic double bonds, 6e and 6f as well as 8, participated in this kinetic resolution with superb efficiency (s455, Fig. 4c). The functional-group tolerance was expanded further by a vinylic bromide (as in 6c). For comparison, we subjected purely aliphatic alcohol 10 to the standard protocol, and the selectivity factor (s ¼ 9.62, Fig. 4d) was lower than those obtained for allylic alcohol 6d (s ¼ 14.4, Fig. 4b) and benzylic alcohol 1b (s ¼ 14.6, Fig. 3b). This indicates that the p system attached to the carbinol carbon atom is important for enantiomer discrimination by the catalyst.

Discussion
We have disclosed here that the commercially available CuCl/(R,R)-Ph-BPE-NaO t Bu catalyst system allows for the kinetic resolution of alcohols by enantioselective Si-O coupling. The choice of the hydrosilane coupling partner is crucial as high selectivity factors have only been achieved with trialkylsilanes, and commercial n Bu 3 SiH was used throughout this study. This easy-to-apply catalyst-hydrosilane combination kinetically resolves a broad range of structurally unbiased benzylic and allylic alcohols, reaching synthetically useful selectivity factors for cyclic benzylic (sr40.1, Fig. 3c) and various allylic alcohols (sr159, Fig. 4b,c). It must be emphasized here that the inaccuracy of the analytical tools to measure conversion and enantiomeric purity leads to imprecise selectivity factors, particularly for s450 (refs 27,41). Hence, the reported values are not exact but rather an approximation of the order of magnitude. While the present protocol is limited to secondary alcohols, we will now tackle the most difficult class of alcohols in this chemistry, tertiary alcohols 28 .