Chirality Transfer in Gold(I)-Catalysed Direct Allylic Etherifications of Unactivated Alcohols: Experimental and Computational Study

Gold(I)-catalysed direct allylic etherifications have been successfully carried out with chirality transfer to yield enantioenriched, γ-substituted secondary allylic ethers. Our investigations include a full substrate-scope screen to ascertain substituent effects on the regioselectivity, stereoselectivity and efficiency of chirality transfer, as well as control experiments to elucidate the mechanistic subtleties of the chirality-transfer process. Crucially, addition of molecular sieves was found to be necessary to ensure efficient and general chirality transfer. Computational studies suggest that the efficiency of chirality transfer is linked to the aggregation of the alcohol nucleophile around the reactive π-bound Au–allylic ether complex. With a single alcohol nucleophile, a high degree of chirality transfer is predicted. However, if three alcohols are present, alternative proton transfer chain mechanisms that erode the efficiency of chirality transfer become competitive.

Introduction a-Chirale thers are present in many natural products, biologically active molecules and synthetic intermediates. [1] Therefore, much effort has been directed towards efficient routes to enantiomerically enrichedchiral ethers through allylic etherification reactions. [2] Within this context, there is currently ongoing interest in utilising unactivated allylic alcohole lectrophiles in transition-metal-catalysed allylationso fv ariousn ucleophiles, [3] as the use of unactivated allylica lcohole lectrophiles reduces the number of synthetic steps required (by virtue of not requiring prior derivatisation) andm inimises byproduct formation.I nt erms of asymmetric intermolecular etherifications, ar ecent notable advance by Carreira et al. uses Ir catalysis to effect allylic etherifications on secondary allylic alcohols through formal S N 2s electivity. [4] One of the key research efforts withino ur group has been to developg old-catalysed [5] regioselective methods towards allylic ethers [6] and allylic thioethers. [7] Within this context, we recently developed am ild and air-stable gold(I)-catalysed direct allylic etherification of allylic alcohols. [8] Thisd ehydrative formal S N 2' procedure [9] requires neither the allylic alcohol electrophile nor the alcohol nucleophile to be activated (eithert oi nstall al eaving group in the former or form an alkoxide in the latter), leadingt om ild reaction conditions that are tolerant of variousf unctional groups as well as air and moisture( Scheme 1a). [3a, 10] We were keen to extend this methodology to asymmetricm ethods by investigatingv ariousc hiral, non-racemic g-substituted substrates,w hich should be amenablet o chirality transfer.I nt heory,a ne nantioenriched chirala llylic alcohol with g-substitution (e.g., 4,S cheme 1), which is easily accessible in good enantioselectivities by Sharpless kinetic resolution [11] or enzyme resolution, [12] should be able to transfer its chirality [13] to the allylic ether product 5,e specially if a6 -membered ring hydrogen-bonded intermediate I is involved (Scheme 1b). Access to chiral, non-racemic a,g-disubstituted allylic ethers such as 5 from unactivated alcohols also nicely complements recent Ir-catalysed allylation methods by Carreira et al., [4a] whicha re confined to formation of unsubstituted secondary allylic ethers (R 1 = H). It should be noted that shortly after our initial communication, [8a] Mukherjee and Widenhoefer disclosed an independent report on the same reaction. [14] Using ad ifferent set of catalysts and conditions, they carried out as ubstrate-scope study on the racemic reaction. In addi-tion, they also elegantly show one example of ac hirality-transfer reaction( see below). However,a st he substrate scope of their chirality-transfer reactionw as not reported and there was room for improvementw ith regardst ot he regioselectivity (5:1 of formal S N 2'/S N 2 5/6), we decided that it was still important to continue with our independent studies. These are reported here, andi nclude optimisation to give greatly improved regioselectivities, full substrate-scope studies and experimental and computational mechanistic investigations.

Results and Discussion
Our investigation began with the optimisation of reaction conditions to improve the regioselectivity for allylic etherifications, using secondary allylic alcohol 4a as am odel substrate (Table 1). Our previously reported conditions provided ap oor 2:1r atio of formal S N 2'/S N 2( 5aa/6aa,e ntry 1), which needed to be improved drastically before chirality transfers could be investigated. During our optimisation,w ed iscovered that addition of molecular sieves (MS) to the reaction mixture greatly improved the selectivity,e xclusivelyy ielding the formal S N 2' product 5aa (entries 2-8). 3 MS provided slightly higher yields compared to 4 MS (entry 2v s. entry 5), however,t he yields were modest when only 5mol %o fg old catalyst was employed (42 %a nd 56 %, respectively). Portion-wise addition of gold catalyst (2 5 mol %) greatlyi mproved the yields (entries 3a nd 6), but addition of another portion of molecular sieves makes little difference (entry 3v s. entry 4). Finally,a s ac ompromise between shorter reactiont imes and acceptable yield, we settled for the protocol shown in entry 8a so ur optimised conditions for investigating the chirality-transfer reaction. Note that under these newly optimised conditions, the formal S N 2' product 5 is formed exclusively for all subsequent substrate-scope screens(Ta bles 2and 3).
With theseo ptimised conditions in hand, we turned our attention to effecting chirality transfer in etherifications of ar ange of enantioenriched allylic alcohols (Table 2). For this assay,a lcohol 2b was chosen as the nucleophile fore ase of chiral stationaryp hase (CSP)-HPLC enantiomer separation in the product. Gratifyingly,o ur first attemptw ith enantioenriched n-butyl allylic alcohol( R)-4a gave the desired product (E)-5ab in 88:12 e.r.f rom > 99:1 e.r.o ft he starting material. The minor (Z)-isomer (Z)-5ab was obtained in 84:16 e.r.  (entry 1). Alternatively,s tarting with (Z)-allylic alcohol 4b,t he opposite enantiomer of the product could be obtained with good transfer of chirality (entry 2). It should be noted, however,t hat the Z-allylic alcohols tartingm aterials (e.g., 4b,e ntry 2) are more difficult to access in high e.r., resulting in poorer e.r. of product 5bb,d espite displaying ag ood degree of chirality transfer (81:19!76:24 e.r.). Reversing the substituents at the a-a nd g-positions of the allylic alcohol similarly gave high yield of product 5cb with ah igh degree of chirality transfer (entry 3). Replacing the n-butyl substituent with the sterically more demanding cyclohexyl also works well with the Cy at the a-position, but only moderatelya tt he g-position (entries 4-5).
Certain substituents on the allylic alcohols ubstrate were found to cause the chirality transfer to proceed moderatelyt o poorly (entries [10][11][12][13]. For example, dimethyl allylic alcohol 4k gave ah igh degree of racemisation (71:29 e.r.o fE-5kb from 4k of > 99:1 e.r.);l ikewise, increasing the steric bulk of the substituent at the alcohol centre to tert-butyl alcohol 4l also led to some racemisationd uring reaction (entry 11). Substrates with b-substituents performed the worst: 4m and 4n both give excellent > 20:1 E/Z ratios, but almost complete racemisation under these conditions (entries12-13) and are therefore not suitable substrates for chirality transfer.
We next turned our attention to investigating the tolerance of ar ange of different nucleophile alcohols by using model allylic alcohol substrate (R)-4a (Table 3). [15] Although para-bromobenzyla lcohol 2d (entry 3) gave ac omparable resultt ot he originaln ucleophile alcohol 2b,b enzyl alcohol 2c and paramethoxybenzyl alcohol 2e yielded products 5ac and 5ae,r espectively,w ith ag reater degree of chirality transfer (97:3 e.r., entry 2a nd > 95:5 e.r., entry 4). [16] Furfuryl alcohol 2f was also tolerated, thoughw ith ar eduction of enantioenrichmenti n product 5af (entry 5). We then turned our attention to alkyl alcohols. Extendingt he alkyl chain of benzyl alcohol by two methylene units preserved yield, formal S N 2'/S N 2a nd E/Z alkene selectivity as well as chirality-transfer efficiency( entry 2 vs. entry 6). Next, we set out to test functional group tolerance. Pleasingly, trifluoromethyl substitution of the nucleophile was tolerated (entry 7) as were haloalkanes (entry 8) and unprotected terminal alkenes (entry 9). When utilising diol 2k,r eaction occurred exclusively through the primary alcoholt og ive 5ak in high yield and selectivity (entry 10). No product from subsequent reaction through the tertiarya lcohol was observed. Acid-labile groups such as acetals 2l and 2m were also found to be compatible with the reaction( entries 11 and 12). Finally, we demonstratedt hat the reactiona lso proceeds very well (> 99:1 e.r.) using am ore hindered secondary nucleophile alcohol such as cyclohexanol 2n (entry 13). The mechanism that we originally proposed for thea llylic etherification reaction [8a] can also account for the chirality transfer and stereospecificity of the reaction( Scheme 1). As gold(I) is an excellent p-Lewis acid, [5e] it is likely to activate the alkene functionality in the allylic alcohol towards attack by an externala lcohol nucleophile (I,S cheme 2). [10f] Demetallation and elimination of water (enabled by intramolecular hydrogenbonding, II)w ill then regenerate the catalysta nd produce the desired allylic ether product 5.At ightly bound chair-like 6membered ring transition state [17] is required for efficient chirality transfer,a nd also accountsf or the stereospecificity of the E and Z isomers. As showni nS cheme 1a,t he E isomer has its substituent Ri nt he equatorial position, whereas the Z isomer has Ra xial (Scheme 2b), thus leading to the differents tereochemicalo utcomes. Havingt he substituent R' equatorial also accountsfor the E-selectivity of the reaction.
It is clear from the proposed mechanism in Scheme 1t hat the hydrogen-bonded 6-membered transition state I is crucial for the chirality transfer,a nd also the E-selectivity. Any erosion of ee could therefore be attributed to the disruption of this hydrogen-bonding pattern that would allow the reaction to occur withoutt his 6-membered transition state I.O ne such mechanism is explored in the computational section below (see Scheme 9). However,asecond possibility for erosion of ee is the racemisationo ft he product 5 through isomerisation between the formal S N 2' (5)a nd formal S N 2( 6)p roducts,c atalysed by gold(I). [6b, c, 17] Indeed, during our relateds tudies using thiols for thioetherification reactions, chirality transfer does not occur in the thioetherification reactions. [7a, 18] Experimental and computational studies showedt hat the racemisation is due to isomerisation between the formal S N 2' and S N 2t hioether products. Clearly,u sing alcohol instead of thiol as an ucleophile allows for successful chirality transfer,e xcept in certain substrates,s uch as b-substituted 4m and 4n (entries 12-13, Ta ble 2). Therefore, we carriedo ut several control experiments to ascertain the role of isomerisation of the products 5 and 6 in the erosion of ee.
Firstly,p roduct 5db(entry 4, Ta ble 2) was resubjected to the reactionc onditions and no change was observeda fter 24 h (Scheme3a). Next, 5eb was investigated, as this product was formed with only moderate chirality transfer (77:23 e.r. ,e ntry 5, Ta ble 2). Once again, no change was observed upon resubjection to the reactionc onditions (Scheme 3b). Finally,p roduct 5mb,w hich is formed as ar acemic mixture by our method (entry 12, Table 2) wasi nvestigated. This species, obtained in 98:2 e.r.b ya na lternative route, [19] was found to racemise upon resubjection to the reaction conditions (Scheme 3c). [20] From these results, it appears that slight erosion of ee is not caused by isomerisation/racemisation of the product (see later and Scheme 9f or plausible racemisation mechanism).H owever,i nstanceso fc omplete racemisation, such as the formationo f 5mb and 5nb from b-substituted 4m and 4n,r espectively, could be due to isomerisation and racemisationo ft he products under the reaction conditions.
Next, we wanted to ascertain the role of molecular sieves in the reaction. It is clear from the results in Ta ble 1t hat addition of molecular sieves is the key factor to improving the formal S N 2'/S N 2( 5/:6)r egioselectivity.O ur next control reaction (Scheme 4) shows that moleculars ieves are also crucial for chirality transfer and E/Z selectivities. Removing molecular sieves from the reaction results in ac ompletely racemic product 5db and ap oor 3:1 E/Z ratio (vs. 89:11e.r.a nd 9:1 E/Z with molecular sieves added). [21] Previously,W idenhoefer et al. had shown that chirality transfer is possible on substrate 4j,w ithoutt he need for molecular sieves (Scheme5). [14] Havingj usta scertainedt hat molecular sieves are crucial to avoid racemisation, we therefore thought it important to investigate whether the conditions in Scheme 5 allow for the omission of molecular sieves in chirality-transfer Scheme2.Proposed mechanism for successfulc hirality transfer and stereospecificity.
Scheme3.Resubjection of products 5db, 5eb and 5mb to the reaction conditions.
The effect of molecular sieves in the reactioni ss tark as well as puzzling. There are several possibilities regarding the mode of action of molecular sieves in the reaction that may lead to the observed chirality-transfer outcome. Possible reasons for this could be:i )removal of excess water from the reaction; ii)the slightly basic nature of molecular sieves, which may deactivate the gold catalyst; [22] and iii)the polar surfaceo fmolecular sieves may result in the reactiono ccurringc loser to the surface, thereby changingt he aggregation levels or transition state. However,acontrol reaction to test point (i)s hows that chirality transfer is observed regardless of whether the molecular sieves are activatedo rn ot, thusr uling out this possibility (Scheme 7). In fact, the reactiono ccurs with even better yields and e.r.w ith unactivated vs. activated sieves (67 %, 94:6 e.r.v s. 90 %, 98:2 e.r., Scheme 7).
Density functional theory (DFT) calculations were therefore employed to explore the mechanism of these direct allylic etherification reactions. In particular,w es oughtt ou nderstand why our initial expectation of chirality transfer (cf. Scheme 1 and 2) was not borne out, except in the presence of molecular sieves. In the calculations we have studied the symmetrically substituted dimethyl allylic alcohol 4k (as the R,E-isomer) reacting with ethanol (2o)t og ive 5ko.T his choice removes the potentialc omplication of any subsequentS N 2' reaction at 5ko as this would return the same 5ko product. Experimental studies indicate that the catalysis is not significantly affected by the nature of the alcohol ands oe thanol was chosen for simplicity.T he calculations (run with SDD pseudopotentialsa nd basis sets on Au and P, with d-orbital polarization on the latter, and 6-31g** basis sets on other centres) report free energies derived from aB P86-D3(toluene) protocol, that is, gas-phase free energies based on BP86 optimisations, corrected for dispersion andt oluene solvation (using Grimme'sD 3p arameter set and the PCM approach respectively,s ee Supporting Information for full details).
The Au-catalysed direct allylic etherification reaction is thought to proceed [23] via coordination of the {Au(PPh 3 )} + fragment at the C=C p-bond of the allylic alcohol. As shown by Mukherjee and Widenhoefer, [14] if the alcohol nucleophile attacks at the opposite face to Au then only two outcomesa re possible with an enantiopure substrate:w ith (R,E)-4k either (S,E)-5ko or (R,Z)-5ko will be formed (Scheme 8). The formation of both products (alongside water) is computed to be thermodynamically downhill, with the E-isomerf avouredo ver the Z-form by 1.3 kcal mol À1 .T his equatest oaE/Z ratio of approximately 9:1a t2 98 K, fairly typical of the E/Z selectivities seen with dialkyl-substituted allylic alcohols (Tables 1a nd 2). This result also suggests the reactionm ay be proceeding under thermodynamic control.
For the computed mechanism, we consider the direct etherification to start from the p-bounda dduct [(Ph 3 P)Au{(R,E)-4k}] + ·EtOH, I,i nw hich the EtOH is hydrogen-bondedt ot he OH group of the allylic alcohol. [24] Severala rrangements of this adduct were located in the course of this study and the most stable of these, Ia,h as the EtOH lying over the Au centre( i.e., syn to Au), with interactions to both the Oo ft he allyl group (1.86 ) and also to one CÀHb ond of the PPh 3 ligand (2.26 , see Figure 1, which also provides the associated labelling scheme). The most stable adduct, where the EtOH is located Scheme7.Comparing results of reactionsw ith no moleculars ieves, activated sieves and unactivated sieves.
Scheme8.Possible outcomes of the Au-mediated reaction of (R,E)-4k with ethanol(2o)tog ive either (S,E)-or (R,Z)-5ko.Computedp roductf ree energies are indicated in kcalmol À1 ,r elative to the reactant setto0 .0 kcalmol À1 . Chem.E ur.J.2015, 21,13748 -13757 www.chemeurj.org on the other side of the C=C p-bond (i.e., anti to Au, Ib), is 6.6 kcal mol À1 higher in energy,w ith the EtOH showing close contactsw ith one allylic proton, as well as the OH group. All energies in this sectionw ill be quoted relative to Ia set to 0.0 kcal mol À1 .
The key steps and associated energetics for directe therification through anti-attack of EtOH are outlined in Figure 2. Starting from Ib,C ÀOb ondf ormation proceeds throughatransition state at + 10.9 kcal mol À1 to give intermediate II b at + 9.1 kcal mol À1 .T he computed structure of this species is shown in Figure 3a nd displayst he anticipated hydrogen-bonded chairlike structure with the Au and both Me substituents all occupy-ing equatorial positions. Similar structures have been reported at a{ Au(NHC)} + fragment. [17a] From here H + transfer induces loss of water and concomitant formation of the AuÀC 3 bond to give intermediate III b in whichthe allylic ether product is bound through the C 2 =C 3 bond and water is hydrogen-bonded to the ether oxygen. Dissociation will give 5ko as the S,E-form.O f the two transition states, the higher is TS(II-III)b at + 11.1 kcal mol À1 .A na nalogous series of events accountsf or the formation of (R,Z)-5ko.S tartingf rom Ic (G =+6.8 kcal mol À1 ), the chair-like intermediate II c is formed via TS(I-II)c at + 11.1 kcal mol À1 . II c is similar to II b but now has one methyl substituent in an axial position. Loss of H 2 O via TS(II-III)c at + 13.2 kcal mol À1 leads to III c from which the allylic ether product is lost as the (R,Z)form. Overall,t hese two allylic etherification processes proceed with modest barriers (< 14 kcal mol À1 ). In addition, as these reactions are only marginally downhill thermodynamically (Scheme 8), they are likely to be reversible under the reactionc onditions. Hence, at hermodynamic distribution of products is seen that favours the (S,E)-5ko product.
The observation of the enhanced stability of precursor Ia in which EtOH is positioned syn to Au suggests the possibility of alternative syn attack mechanisms and two such processes have been characterised (see Figure 4). Hydrogen-bonded chair-like intermediates II a and II d are located,b ut now with the Au in an axial position( see Figure 3f or the structure of II a). Loss of water from II a and II d then leads to the formation of   whereas II a is formed by syn attack. Computedf ree energies (kcal mol À1 )a re quoted relative to Ia set to zero and selectedd istances are in . Phosphinep henyl substitutents are truncateda tt he ipso carbon for clarity.
(R,Z)-5ko and (S,E)-5ko,r espectively,t hat is, the same products as seen in the anti attack processes. syn attack, however,e ntails barriers of 17.7 and 18.4 kcal mol À1 ,and so these processes will not be competitive with the anti attack mechanisms described above.
The mechanisms outlined so far are consistent with the transfer of chirality shown in Widenhoefer's example (Scheme 5) and anticipated prior to our work. However, our experimental studies have shown that in most cases such chirality transfer only occurs in the presence of molecular sieves andt hat in fact under sieve-free conditions loss of chirality dominates. To account for this am echanism, involving nucleophilic attack at one face of the allylic alcohola nd loss of water from the opposite face is required. One way to achieve this is to invoke ap roton chain transfer mechanism [25] involvings everal EtOH molecules. This is illustrated in Scheme 9f or the case of three EtOH molecules. Pathway( i) is equivalent to the anti attack in Scheme 8, where at hree EtOH molecule chain now promotes loss of water anti to Au with formation of (S,E)-5ko.I nc ontrast,p athway (ii)isa ble to accommodate a syn attack by EtOH whiles till delivering ap roton onto the allylic hydroxyl group that is in an anti position. This leads to the formation of (R,E)-5ko.I ft he barriers to pathways( i) and (ii)a re comparable, the result will be the loss of chirality transfer that is seen experimentally. Am odel incorporating three ethanolm olecules was set up to test these ideas (see Figure 5), three being the minimum number of EtOH molecules requiredt oa ccess pathway (ii), which requires both faces of the allylic alcohols ubstrate to be engaged. As in the single EtOH system, the most stable form of the hydrogen-bonded precursor has one EtOH positioned over the Au centre. This species, Ia 3(ii) ,l eads ultimately to the (R,E)-5ko product along pathway (ii), as described below.T he alternative arrangement relevant for pathway (i)i ss een in Ia 3(i) and lies 8.7 kcal mol À1 higheri ne nergy.B oth the initial attack of EtOH at this species( via TS(I-II)a 3(i) at + 12.8 kcal mol À1 )a nd the subsequentl oss of water (via TS(II-III)a 3(i) at + 14.0 kcal mol À1 )o ccur anti to the Au and hence yield the (S,E)-5ko product. Theo verall barrierf or pathway (i) is 14.0 kcal mol À1 .
In pathway (ii), the EtOH molecule lying over the Au centre in Ia 3(ii) is linked through two hydrogen-bonded EtOH molecules to the allylic hydroxyl group, which maintains ap osition anti to the Au. syn attack of EtOH proceeds to give II a 3(ii) (G = + 6.0 kcal mol À1 )v ia TS(I-II)a 3(ii) at + 15.6 kcal mol À1 .F igure 6 shows the computed structure of II a 3(ii) and highlights the anti arrangement of the EtOH nucleophile and the putative H 2 O leavingg roup.I nt his case, the water dissociationi st he easier step ands o( R,E)-5ko is formed witha no verall barrier of 15.6 kcal mol À1 .S imilar barrierh eights are therefore computed for the formation of both (S,E)-5ko (DG°c alc = 14.0 kcal mol À1 ) and (R,E)-5ko (DG°c alc = 15.6 kcal mol À1 )a nd this, coupled with reversibility of thesetransformations, means that both enantiomers will be formed over the timescale of the reaction, leading to the unexpectedl oss of chirality transfer. [26] Figure 4. Key intermediates and energetics for the Au-catalysedr eaction of (R,E)-4k with ethanol(2o)b ysyn attack to give either (S,E)-5ko or (R,Z)-5ko (L = PPh 3 ). Computedf ree energies are indicated in kcalmol À1 and quoted relative to Ia set to 0.0 kcal mol À1 .
Scheme9.Possiblem echanisms accounting for loss of chirality transfer in the Au-mediatedr eactions of (R,E)-4k with three ethanol molecules to form both (S,E)-5ko (pathway (i)) and (R,E)-5ko (pathway (ii)). Computedf ree energies are indicated in kcal mol À1 and quoted relativetoIa 3(i) set to 0.0 kcal mol À1 . Chem. Eur.J.2015, 21,13748-13757 www.chemeurj.org Ap lausible explanation for the requirement of molecular sieves for efficient chirality transfer is their role in disrupting the type of protonc hain transfer mechanism (i.e.,a ggregation levels of alcohol) shown in Scheme 9, whicha re ap otential cause of racemisation.T ot est our hypothesis, the reaction was carried out with al arge excess of alcohol nucleophile [27] (Scheme 10), which shouldoutcompete the role of the molecular sieves. Indeed, very poor enantiomeric ratios are observed in the presenceo f2 0equivalents of alcohol 2i (Scheme 10), thereby lending support to our theory.F ollowing this train of thought, we postulated that the reason substrate 4j does not requiremolecular sieves for chirality transfer (Schemes 5a nd 6) is because the ether moiety (-CH 2 OBn) within the substrate may be playingasimilar role to molecular sieves:d isrupting the protonc hain transfer mechanism, presumably by hydrogen-bonding. Indeed, when the reaction with 4j is repeated with al arge excess (20 equiv) of alcohol nucleophile, this overrides anye ffect of the ether moiety as well as molecular sieves and produces racemic product 5jb in ap oor 1:1 5jb/6jb regioselectivity (Scheme 11).
An additional explanation for erosiono fc hirality transfer is possible for allylic ether products such as 5kb.I nt his case, as econd formal S N 2' reactiono n5kb will lead to the same product, but with chirality transfer to the opposite enantiomer. Calculations indicatethat the barrierf or the second formal S N 2' is readily accessible (via at ransition state at 9.8 kcal mol À1 for 5ko). As, by definition, the energies of theset wo enantiomers of 5kb are the same, theser eversiblep rocesses will produce ar acemic mixture. This is exemplified by control experiments shown in Scheme12, where the product 5kb from Table 2, entry 10 is resubjected to the reactionc onditions without sieves to produce ar acemic mixture. In contrast, in the presence of sieves, this process is considerably slowed (Scheme12). Although the reasonsf or the remarkable impact of the molecular sieves on these transformations are currently unclear, nevertheless, their effect in providing kineticc ontrol for theseexperiments is remarkable and, moreover,synthetically useful.

Conclusion
We have successfully developed conditions for highly regioselective, gold(I)-catalysed direct allylic etherification of alcohols with chirality transfer to access enantioenriched g-substituted secondary allylic ethers. At horough substrate screen shows that very high levels of chirality transfer can be achieved (up to > 99:1 e.r.). The reaction is very functional-groupt olerant and proceeds in the presence of unprotected groups such as alkyl halides, tertiary alcohols, alkenes and acid-sensitivea cetals. Both primary and the more hindereds econdary alcohol nucleophiles are tolerated well. Furthermore, we demonstrate that the addition of molecular sieves is crucial not only for excellent formal S N 2' selectivity,b ut also to ensure efficient chirality transfer.T he molecular sieves need not be activated to achieve this effect, which implies that it is not aiding the selectivity by removal of water.D FT calculations suggest that chirality transfer should proceed under conditions that promote the reaction of as ingle alcohol as nucleophile. However,ath igheralcohol concentrationsp roton chain transfer mechanisms become accessible, which permit alternative pathways that will Figure 6. Computeds tructure of intermediate II a 3(ii) located along pathway (ii)o nr oute to the formation of (R,E)-5ko.The computed free energyis in kcal mol À1 and is relativet oIa 3(i) set to zero.S elected distances are in and phosphine phenyl substituents are truncated at the ipso carbon for clarity.
Scheme10. Comparing results of reactions with no moleculars ieves,u nactivated sievesa nd large excess of alcohol nucleophile.
Scheme11. Large excess of alcohol nucleophile results in racemisation with 4j,o verriding the effect of the ether moiety (CH 2 OBn) and molecularsieves.
Scheme12. Resubjecting 5kb to reaction conditions:racemisation without molecular sieves and much slower erosion of e.r.with molecular sieves. Chem. Eur.J.2015, 21,13748 -13757 www.chemeurj.org erode the chiralityt ransfer.Ap lausible role of the molecular sieves is to disrupt the aggregation of alcohol molecules in order to prevent loss of chirality through this pathway.W hatever the underlying reasons, the impact of molecular sieves in controlling the outcomeo ft hese allylic etherification reactions is remarkable and synthetically useful.

Experimental Section
General procedure As olution of [PPh 3 AuNTf 2 ]( 2:1t oluene adduct, 5mol %), allylic alcohol 4 (0.101 mmol), alcohol 2 (0.506 mmol) and 3 molecular sieves (8 mg) in toluene (260 mL) was stirred at 50 8Cu nder air for 8h.T hen, [PPh 3 AuNTf 2 ]( 2:1t oluene adduct, 5mol %) was added and the resulting solution was stirred at 50 8Cf or af urther 16 h. The resulting solution was filtered through as hort plug of silica, washing with 9:1h exane/Et 2 O. The filtrate was evaporated under reduced pressure to give the crude product, which was purified by flash column chromatography.F ull experimental procedures, characterisation for all new compounds and copies of 1 Ha nd 13 CNMR spectra are provided in the Supporting Information.