Towards a General Understanding of Carbonyl‐Stabilised Ammonium Ylide‐Mediated Epoxidation Reactions

Abstract The key factors for carbonyl‐stabilised ammonium ylide‐mediated epoxidation reactions were systematically investigated by experimental and computational means and the hereby obtained energy profiles provide explanations for the observed experimental results. In addition, we were able to identify the first tertiary amine‐based chiral auxiliary that allows for high enantioselectivities and high yields for such epoxidation reactions.


Introduction
Onium ylides have foundw idespread applications for( dia)stereoselectivee poxide, aziridine and cyclopropanes yntheses. [1][2][3][4] Amongt he different classes of onium ylides that can be used for such three-ring-forming reactions, sulfur ylides have been the mostf requently used ones, and av ariety of applications by using achiral or chiral sulfur ylides (either preformed or generated in situ with catalytic quantities of ac hiral sulfide)h ave been reported since. [3] In contrast to the privileged use of sulfonium ylides, easily available ammoniumy lides have been less routinely employed in the past. [5][6][7][8][9][10] Especially,t he use of chiral amines to renders uch reactions enantioselective has so far mainlyb een limited to cyclopropanationr eactions, am ethodology that wasi mpressively developedb yG aunt et al. [5] However,t heir use in asymmetric epoxidation and aziridination reactions was found to be rather difficult, [6,7,9] which can mainly be explained by the weaker leaving-groupa bility of the amine group as compared to the use of sulfoniumy lides. [11] Our group has recently introduced ah ighly trans-selective protocolf or the synthesis of glycidic amides [7] and analogous aziridines [9b] starting from ammonium acetamides 1 (which are in situ deprotonated to give the corresponding amide-stabilised ammonium ylides) (Scheme 1). Hereby,s everal important observations were made:F irst, the reactioni ss trongly dependent on the nature of the used amine leaving group, with trimethylamine being clearly superior to othera chiral amines, [7b] at rend that was also observed when using benzylic ammonium ylides. [6d] Unfortunately,a ttempts to carry out this reaction in an enantioselective fashionb yu sing chiral Cinchona alkaloids as the amine leaving group (which is the method of choicef or cyclopropanations [5] )f ailed completely. [7b, 12] In addition, under the developed conditions, this methodology does not give any epoxides 5 when using the more stabilised esterbased ammonium salts 4. [13] Based on these resultsw ebecamei nterested in investigating this reaction in more detail, especially by computational means. Our goals were to rationalise the different reactivities of esters anda mide derivatives and to understand the high trans selectivity in this reaction. We also would like to compare the hereby obtained energy profiles with those of analogous sulfur ylide-mediated reactions to understand the different reactivities of sulfur and ammonium ylides. In addition, as Cinchonaa lkaloids failed as chiral auxiliaries for such reactions, we decidedt oi nvestigate the use of alternative (synthetic) chiral amines for their potential to facilitate this reaction in an enantioselective manner.
The mechanistics equence is similar to that of previously studied ylide-mediated epoxidation reactions:a ddition of the ylide to the aldehyde generates a cisoid betaine intermediate througha[ 2 + +2]-type approach, torsional rotation converts the later into the transoid conformer and finally ring closure, with concomitant expulsion of the amine, gives the corresponding epoxide.
As ar esult of the stabilisation of the ylide by an amide group, the initial addition step is endergonic and the cisoid betaines cis-A and trans-A lying at 17.6 and 22.7 kcal mol À1 above the reactants. [15] Noteworthy,t he cisoid cis-A betaine is more stable than the cisoid trans-A betaine. However,asignificant difference for the required bond rotationt owards the transoid betaines B and the following ring closure can be observed for the two diastereomericp athways leadingt ocis-3a or trans-3a. First, the required bond rotationt ot he transoid cis-B betaine is an endergonic step (6.8 kcal mol À1 )f or which no transition state (TS) could be located and therefore, the cis-B conformer does not represent at hermodynamic local minimum. [16] In contrast, the corresponding trans-A and trans-B betaines are of similar stabilitya nd the bond rotation between these conformations occurs with af ree energy barrier of 6.1 kcal mol À1 .F inally,t he ring-closure step is found to be the rate-limiting step in both cases, even if the elimination TS is only slightly higher than the rotation TS fort he trans-B betaine. Accordingly,t he selectivity is determined at the eliminations tep and the higher stabilityo ft he trans elimination TS (by 3.2kcal mol À1 )a sc ompared to the cis eliminationT Sa ccounts for the observed high trans selectivity.
To prove that betaine formation is indeed ar eversible process, we carriedo ut crossover experiments in analogy to previous studies on sulfur ylides by the group of Aggarwal. [3f] When reactingt he racemic syn and anti b-hydroxya mmonium salts 6 with an excess of am ore reactive aldehyde, such as p-chlorobenzaldehyde, we found that only the trans epoxides 3 are formed (Scheme2). In addition, the more reactive aldehyde is incorporated predominantly,g iving 90 %o ft he corresponding epoxide startingf rom syn-6 and8 5% starting from anti-6 as precursor.A ccordingly,t hese resultsc learly show that betaine formation is indeed highly reversible for both diastereomeric pathways, althoughs lightly more for the cis one.
Aggarwal et al. have carefullyi nvestigatedt he energy profile for epoxidation reactions of amide-stabilised sulfur ylides in the past and it was found that hereby also the elimination step is the selectivity-determining step. [3f] To elucidate the re- activity difference between ammonium and sulfonium ylidemediated epoxidation reactions we thus re-addressed this transformation (considering inclusiono fe ntropic andt hermal contributions and ad ispersion correction,w hich were not included previously [3f] ). Our calculations clearly confirm the earlier conclusions by Aggarwal et al. [3f] that indeed the ring closure is the selectivity-determining step when using amide-stabilised sulfoniumy lides. As it can be seen in Figure 2, the overall energy profile for the epoxidation mediated by the sulfur ylide 7a is energetically lower than for the one mediated by the ammonium ylide 1a (green vs. blue pathway in Figure 2). This confirms the higher overall barrier for ammoniumy lides as compared to sulfonium ones. This lower reactivity of compound 1a can be accounted for by as ignificantly higherb arrier to elimination due to the poorer leaving-group ability of ammonium (7.7 kcal mol À1 ), as compared to sulfonium (2.7 kcal mol À1 ). [11] During our earlier investigations in this project, we also tried to use the propionamide-baseda mmonium ylide 1aa for the epoxidation reaction. However,u nder neither conditions we were able to obtain any product. By analysing the corresponding free energy profile for this transformation ( Figure 2, black pathway) it becomes obvious that the whole reaction is significantly higher in energy as comparedt ot he use of the acetamide-based ylides 1.N oteworthy,i st he fact that the increaseds teric demandi nt he a-positionl eads to as ignificantly higher rotationb arrier, whichb ecomest he rate-limiting step hereby.T his observation can also serve as au seful explanation to understand the lower yield when using the sterically more demanding Cinchona alkaloid leaving groups as outlined in Scheme 1.
We also used the opportunity to compare the drastic influence of an ester group on the reactivity.A sm entioned in the beginning, the so far developed reactionc onditions did not allow us to obtain any epoxides 5 when using the ester-based ammonium salts 4 (Scheme 1). [7] Analysing this observation by DFT calculations reveals that the free energy profile for the reaction of the ylide 4a involves indeed ah igh overall barrier (34.8 kcal mol À1 )( Figure 2). This lower reactivity is mainly due to ah igher stabilisation of the ylide 4a as compared to compound 1a,w hich leads to am ore endothermic betaine formation. The free energy barrier for the elimination is also slightly increased with the more electron-withdrawing ester group. [17] With these illustrative data in hand, we became interested in testingw hether it may be possible to use the ester-stabilised ammonium ylides 4 for aziridination reactions instead. It was showni nt he past by Aggarwal and co-workers that aziridinations by using ester-stabilised sulfonium ylides are possible, whereas the corresponding epoxidations are very difficult. [4b] By using the ammonium salt 4b for aziridinationsw ith the imines 9,w ef ound that indeed some aziridine 10 a can be formed under carefully optimised reaction conditions by using an excesso fs olid Cs 2 CO 3 as the base (Scheme 3). However, this reaction is accompanied by the formationo fs ignificant amountso ft he a,b-unsaturated a-amino ester 11 a and the alkyne 12 a.A lthough the formation of the alkenes 11 is as ide reactiont hat is often observedi na mmonium ylide-mediated aziridination reactions, especially with electron-poor imines, [7c, 9b] the formation of the alkyne 12 a came unexpected.B y changing the electronic properties of the imine 9,w ef ound that formation of compound 11 increases when using the bromine-substituted imine 9b,w hereas the presenceo fa ne lectron-donating group, such as am ethoxy group (i.e.,c ompound 9c), only gave the aziridine 10 c,b ut noneo ft he unsa- turateds ide products 11 c and 12 c (the trend for the formation of the alkene 11 is in analogy to our recent observations [7c, 9b] ). Control experiments showedt hat compound 12 is not formed from compound 11 but most probablye limination seems to occur from the intermediate betaines. By changing the N-protecting group to at osyl group (i.e.,c ompound 9d) we only observed the formation of the cis-aziridine 10 d, [18] but no formation of the products 11 and 12.
To elucidate if the formation of compound 12 may be rationalised by an in situ hydrolysis of the tert-butyloxycarbonyl (Boc)-protected imines 9 to aldehydes 2 andr eaction of compound 2 with the ylide, we carriedo ut the direct reactiono f compound 4b with benzaldehyde (2a) ( Scheme3,l ower part). Interestingly,w ed id not observe any formation of compound 12,b ut instead small quantitieso ft he epoxide 5b was isolated. This came as ab ig surprise as so far we have neverb een able to obtain even trace quantities of compound 5.Obviously, the crucial role in this reaction seems to be the solid carbonate base. It was reported in the past that solid inorganic bases, that is, carbonates, can have av ery speciale ffect on sulfonium ylide-mediated reactions. [19] We thus wondered, whether the liquid-solid combination of CH 2 Cl 2 /Cs 2 CO 3 may also allow us to increaset he yield for the analogouss ulfonium ylide-mediated epoxidation by using ester 13.W ew ere indeed able to isolate the epoxide 5b in 37 %y ield, which proves the positive effect of Cs 2 CO 3 as compared to other previously used bases (i.e., tBuOK, KOH or K 2 CO 3 ), but it must be admitted that this methodology could not furtherb ei mproved by using alternative sulfur leaving groups or conditions. Finally,w ea lso calculated the energy profile for the reaction of the DABCO-and quinuclidine-baseda mide-stabilised ammonium ylides 1 to rationalise the significant yield differences when using them for epoxidation reactions [7b] (compare with Scheme 1).
First, these calculations show that the barriert or ing closure is higherf or quinuclidine( 8.5 kcal mol À1 )a sc ompared to trimethylamine (7.7 kcal mol À1 )a nd DABCO (7.5 kcal mol À1 ), thus pro-viding ar easonable explanation for the lower yields obtained with this leaving groups. However,t his step alone does not explain why trimethylamine allows for significantly higher epoxidation yields than DABCO. As it can be seen in Figure 3, the whole energy profile for trimethylamine is energeticallyl ower than for DABCO, which seems to be mainly due to ah igher stabilisation of the DABCO ylide as compared to the trimethylamine derivative, [20] thus resulting in an overall higherr eactivity towards epoxidation of the trimethyl ammonium salts.

Enantioselective epoxidation
Controlling the absolutec onfiguration in ammonium ylidemediated epoxidation reactions has so far been av ery challenging task. Although the use of Cinchona alkaloids is the method of choice for ammonium ylide-mediated cyclopropanations, [5] their use in epoxidation reactions does not allow for any product formation (Scheme 1). [6,7] Kimachie tal. showed that brucine can be used as ac hirall eaving group for benzylic ammonium ylide-based epoxidations. [6b] We have recently reported an alternative strategy by using chiral trimethylammonium-based acetamides with ap henylglycinol-based amide auxiliary, [7c] which allowed for high selectivities in epoxidation and aziridination reactions. However,b ased on the fact that this protocol requires the cleavage of the auxiliary in as ubsequent step, as trategy by using ac hiral amine leaving group in ammonium ylide-mediated epoxidations would be much more appealing. Based on the low reactivity associated with the use of simple Cinchona alkaloids we therefore, decided to systematically screen av ariety of other chiral tertiarya mines.T able 1 gives an overview of the mosts ignificant resultso btained in ad etailed screening of different chiral tertiarya mines under different reaction conditions. Because DABCO itself was ar easonably good leaving group in our racemic approach (Scheme 1), [7a] we first focused on the known chiral DABCO derivative A. [21] Underl iquid/liquid biphasic conditions we obtained some product 3b in high enantiopurity (Table1,e ntry 1). Unfortunately,t he yield was ratherl ow and no further improvement with alternative solvents and bases was possible (e.g.,T able 1, entry 2). We next attempted the use of the prolined imer B, [21] which unfortunatelyg ave only as lightly higher yield, but with significantly lower enantioselectivity (Table 1, entries 3a nd 4). Similar observations were made by using the trans-cyclohexane diamine C (Table 1, entries 5a nd 6) or derivatives thereof as the auxiliary.F inally, we reasoned that it mayb ep ossible to increase the leavinggroup ability of the amine by using an (hemi)aminal-type structure with al ess basic nitrogen. [23] We thus, synthesised as mall collection of the proline-derived amines D. [24,25] Gratifyingly,a lready the use of the most simple derivative D1 provedo ur hypothesis right, giving the target epoxide 3b in more than 60 % yield with ap romising initial level of enantioselectivity (enantiomeric ratio (e.r.) = 76:24) under biphasic liquid/liquid conditions (Table 1, entry 7). Te sting alternative reactionc onditions showedu st hat liquid/solid conditions by using Cs 2 CO 3 as the base gave compound 3b in more than 80 %a nd comparable selectivity in solvents like iPrOH or dichloromethane (Table 1, Figure 3. Computedf ree energy profiles [kcal mol À1 ]f or the epoxidation by usingt rimethylamine, DABCO and quinuclidine-based ammonium ylides 1. [20] Chem.E ur.J. 2016, 22,1 1422 -11428 www.chemeurj.org 2016 The Authors. Published by Wiley-VCH Verlag GmbH &Co. KGaA, Weinheim entries 8a nd 9). It should be noted that changing the reaction temperature did not have any beneficial effect and we thus, kept these room-temperature conditions to further optimise the auxiliary next (Table 1, entries 9-15). Changing the aryl moiety did not allow us to improvet he outcome (see Table 1, entries 9-12 for representative results) and as we weren ot able to introduce aliphatic groups instead [25] we thus kept the phenylr esidues and started varyingt he substituent R( see Ta ble 1, entries 13-15 forr epresentatived etails). We immediately realised that the introduction of sterically demanding aliphatic groups provides reasonably high selectivities (e.r. > 90:10) and after some fine-tuningt he cyclohexane-based auxiliary D7 [26] was found to be the most promising one to obtain compound 3b in high yield and high enantiopurity (Table 1, entries [15][16][17]. Unfortunately,a ttempts to use this auxiliary in ac atalytic fashionb ys tarting from the a-bromo-acetamide (in analogy to Gaunt's cyclopropanation [5] )d id not allow us to obtain reasonable quantities of the product (< 5% when using 10 mol %o fa mine catalyst but with as imilare nantioselectivity as in the use of preformed ammonium salts). Furthermore, insitu reaction of an equimolar mixture of amine and a-bromo-acetamide with benzaldehyde also only resulted in ar ather low yield of 10-15% of epoxide (again with the same selectivity). This limited reactivity seems to be mainly because formation of the ammonium salt proceeds relatively slowlyu nder the reaction conditions and in addition, the auxiliary itself undergoes instead ap artial aminal hydrolysis under these conditions (this hydrolysis is slow when using the formaldehydebased auxiliary D1 but faster with compounds D5 and D7).
Havingi dentified the first chiral tert-aminea uxiliary that warrants both, high yield and high enantioselectivity in ammonium ylide-mediated epoxidation reactions to access the glycidic amides 3,w enext investigated the application scope of this methodology.A si tc an be seen in Scheme4,avariety of aromatic aldehydes were usually tolerated.
In most cases, we found that higher yields can be obtained when using iPrOH as solventa sc ompared to toluene, which however,i nafew cases, gave as lightly higher enantioselectivity.T his solvent effect was most striking when using the p-methoxybenzaldehyde 2h and it is still not clear why in this specific case the yield in toluene was so much lower.I nterestingly, also nitrobenzaldehyde (2g), which was found to be aproblematic substrate in the past (mainly because of competingC annizzaro disproportionation under the basic conditions), [7] could be successfully employed under these reactionc onditions giving the corresponding epoxide 3g with high selectivity and in amoderate yield. Only in the presence of the rather strongly electron-donating dimethylamino group no product 3j could Scheme4.Applicationscope of the enantioselective epoxidationb yu sing the chiral ammonium ylide-precursor 1-D7. [27] Chem. Eur.J. 2016 be obtained. Attempts to use enolisable aldehydes like cyclohexanecarbaldehydeo rd odecanal showedl ow conversion rates (< 10-20 %a fter 2-3 days) and ap ronouncedt endency towardsb yproduct formation (e.g.,a ldol reactions) what is in accordance with our earlierr acemic approach. [7] Changing the amide substituents on the other hand gave the products 3k and 3l in reasonable yields too. However,h ereby it was found that especially the dibenzylamide group gave as ignificantly lower enantioselectivity than the usually used diethylamide group. Encouraged by the successful use of the ammonium salt 1-D7 for the highly enantioselective synthesis of the glycidic amides 3 we also wanted to investigate if this methodology allows the synthesis of the analogous aziridine 14 in an enantioenriched form. Gratifyingly,d ichloromethane was found to be the solvent of choice to obtain compound 14 in amoderate yield of 42 %a nd ah ighe nantiomeric ratio of 94:6 (Scheme 5, other solvents gave almost no product formation). This methodology in general thus also holds promise for the synthesis of chiral aziridines by using ammonium ylides, albeit in lower yields.

Conclusion
Systematic experimental and computational studies allowed us to reveal the key factors fory lide-mediated epoxidation reactions when using carbonyl-stabilised ammonium ylides. The pronounced reactivityd ifferences between esters and amides as well as the high diastereoselectivity and the different leaving-group abilities of different amines were investigated and the hereby obtained energy profiles provided etailed explanations for the observed experimental results. In addition, we were ablet oi dentify the first tertiary amine-based auxiliary that allows forh igh enantioselectivities andh igh yields for such epoxidation reactions. This represents as ignificant improvement comparedt op reviousr eports which mainly focused on the use of Cinchona alkaloid-based structures and were not satisfactory for epoxidation reactions. The herein presented proline-based auxiliary approachc onstitutes an alternative strategy allowing to overcome these limitations.

Experimental Section
General details can be found in the Supporting Information. This document also contains detailed synthesis procedures for the auxiliaries and the epoxidation reactions and analytical data of the novel compounds and reaction products as well as copies of the NMR spectra and HPLC traces. The Supporting Information also includes the details of the computational investigations.
General asymmetric epoxidation procedure:T he ammonium salt 1 was dissolved in the given solvent (A:0 .1 m in iPrOH, B: 0.1 m in toluene) and Cs 2 CO 3 (20 equiv) were added followed by the addition of the corresponding carbaldehyde (2 equiv). The mixture was stirred at room temperature for 24 h( under an argon atmosphere). The reaction was quenched by addition of H 2 Oa nd extracted with dichloromethane. The combined organic phases were dried over Na 2 SO 4 and evaporated to dryness. Purification by column chromatography (gradient of heptane and EtOAc) gave the corresponding epoxides in the reported yields and enantiopurities.