1.3. Chiral sulfonium ylides

1.3.3.1. C₂ symmetric sulfide ligands

The presence of a C₂ symmetry axis within the chiral auxiliary can serve a very important function of dramatically reducing the number of possible competing, diastereomeric transition states (21). Possibly this in mind, Durst et al. decided to overcome the problems with low selectivities Furukawa and co-workers had encountered by introducing a new family of C₂-symmetric sulfide ligands to be used in asymmetric epoxidations (19, 22). The employed sulfonium salts 10, 11 and 12 are presented in Figure 1.

The sulfonium salt formation was carried out by treating the precursor thiol 13 with benzylbromide, followed by addition of AgClO₄, and the epoxidations themselves were conducted under phase transfer conditions, Scheme 8 (22).

In order to have the potential for the synthesis of optically active epoxides, the chiral sulfides employed in the reaction must be recyclable without loss of optical activity. Durst with co-workers proved this when they were able to recover thiolane 13 in 80% yield with essentially no loss of optical purity.

Despite the fact that the substituents at C-3 and C-4 in the ylide form of 13 have opposite senses of chirality compared to C-2 and C-5 in the ylide form of sulfonium salt 11, they afforded the same absolute configuration for the product trans-stilbene oxide 9a. The results of the experiments are collected in Table 2 where it can be seen that the highest enantiomeric excesses were obtained with ylides derived from thiolanes having substituents close to the ylide reaction center, i.e., at C-2 and C-5. No mechanistic rationale was given, however. The highest enantiomeric excess, 83%, was obtained when using an ylide formed from 11. The chemical yields obtained were fairly good varying from 27 to 94%.

1.3.3.2. Camphoric acid based non-C₂ symmetric sulfide ligands

Higher asymmetric induction was reported by the same group (19, 23) using non-C₂ symmetric (1R,3S)-(+)-camphoric acid-derived sulfonium ylides 14-19 (Figure 2).

The reactions were conducted under the phase transfer conditions (PTC), initially at 0 oC, then at room temperature for three hours. Both trans- and cis-stilbene oxides 9a and 20 were formed and in some cases also β -elimination took place. The asymmetric epoxidation employing ylide 14 is presented in Scheme 9. As in previous work with C-2 symmetric ligands, stoichiometric amounts of sulfides were employed in the reactions (23).

The alkylation of the sulfides had resulted in the formation of single diastereomeric sulfonium salts. From a single diastereomeric sulfonium salt only one ylide isomer having a defined conformation at sulfur would be formed and, therefore, high enantioselectivities could be expected. Indeed, benzylidene transfer gave enantioselectivities up to 96% but only low enantioselectivity for methylene transfer (16 gave up to 4% ee) with less than 50% chemical yield obtained.

Also asymmetric epoxidations of one ketone (cyclohexanone) as well as two aliphatic aldehydes were conducted. When employing cyclohexanone as the carbonyl component the corresponding epoxide was obtained with ≥ 96% enantioselectivity. The transfer of the p-chlorobenzylidene group from 15 to formaldehyde gave the (S)-enantiomer of the corresponding epoxide, and the benzylidene transfer from 14 to cyclohexanecarboxaldehyde, in turn, afforded a mixture of trans- and cis-epoxides in only fair yields. The optical purities of stilbene oxides were determined using the chiral shift reagent [Eu(hfc)] and configurations were simply based on [α]_D measurements. All results are presented in Table 3.

1.3.3.3. Asymmetric induction

Durst et al. rationalized the asymmetric induction in the following manner (23). The ylide 14 should adopt preferentially the conformation A rather than B since the aryl ring on ylide 14 has severe steric interactions with the endo hydrogens on the two-carbon bridge (Figure 3). Electrophilic attack by the carbonyl group on the ylide A will occur preferentially from the back (i.e. the si-face) of the ylide. This inducts the chirality of that center as (R) in the intermediate betaine and as (S) in the final product (24). The chirality of the second center is dependent on the approach of the carbonyl compound to the ylide. The concept that facial selectivity, with respect to the ylide rather than the carbonyl carbon, controls the stereochemical outcome predicted that all transfers from A will result in products having preferentially the (S) chirality at benzylic carbon derived from ylide, was verified by these studies.

In turn, the excess of the (R,R)-isomer of the trans-stilbene oxide obtained from ylides 17 and 18 (Figure 2) is due to the blocking of the si-face of the ylide carbon by either the benzyl or isopropyl group (structure C, Scheme 12). The lower chiral induction observed for the benzylidene transfer was proposed to be caused by the greater conformational mobility of the α-benzyl group in C vs A because the former lacks the additional β -methyl group (23).

The approach of the cyclohexanecarboxaldehyde to A was found to be almost random with respect to the facial selectivity of the aldehyde as was evidenced by an almost 1:1 trans/cis product mixture. Since the carbon of the methylene ylide 16 (D) is not prochiral, resulted approach of benzaldehyde to D in a racemic product. In contrast, the S-p-chlorobenzyl analog of A (ylide 15) reacted with formaldehyde to produce p-chlorostyrene oxide with 24% ee. Again, the sense of chirality of the product was due to the facial selectivity in the attack on the ylide and not the carbonyl carbon.

The formation of (R,R)-stilbene oxide from 19 suggested that, in this case, the ylide would preferentially adopt structure E. A small preference for attack from the rear, in this case the re-face, should generate preferentially the (R,R)-product (23).

1.3.4.1. Ligand and asymmetric induction

Durst also claimed that in order to ensure high sulfur nucleophility, the sulfide should be devoid of other electronegative heteroatoms. Solladié-Cavallo et al. have diligently reported very successful asymmetric epoxidations with chiral oxathiane derived sulfonium ylides (25,26, 27, 28, 29,30 ), however.

During the work on the asymmetric synthesis of adregenic drugs, Solladié-Cavallo et al. were confronted with the difficulty of obtaining optically pure epoxides and decided to investigate the use of chiral, optically pure sulfur ylides to convert aldehydes into chiral epoxides (25). The chiral oxathiane 23 (Eliel’s reagent) was obtained from (+)-R pulegone 22 by a literature method (31), Scheme 10.

The S-alkylation of sulfide 23 with benzylbromide in the presence of AgClO₄ afforded sulfonium salt 24 (25). NMR-analysis of 24 showed only one diastereomer and it could thus be expected that, in basic medium, 24 would yield only one ylide, a prerequisite for high asymmetric selectivity.

The reaction of sulfonium salt 24 with benzaldehyde and substituted benzaldehydes under phase transfer condititions (PTC) afforded only trans-epoxides in satisfying (60-82%) yields and enantiomeric purities up to 70-100%. Oxathiane 23 was employed in stoichiometric amounts and could be recovered and reused. The results of the epoxidations are presented in Table 4 (25).

No cis-epoxides were observed and (+) rotations obtained for trans-epoxides clearly indicated that the configuration of the major epoxides was 2R,3R. The asymmetric induction was interpreted in terms of conformation A in the ylide form of 24 together with a sterically directed approach of the aldehyde as shown in Scheme 11.

The proposed induction was based on the fact that, from a single sulfonium salt’s isomer, only one ylide isomer having the same configuration at sulfur is formed at low temperature and because in the sulfonium salts derived from six membered cyclic sulfides the third S-substituent was found equatorial (32,33), the CHPh group was, at that time, envisaged to be equatorial (conformation A). On the other hand, according to literature results (16a,16c,23,34), a 120° value was chosen for the torsional angle C4-S-C12H; therefore, as a consequence of the R absolute configuration obtained at C12, the rather hindered position of the phenyl ring had to be considered (A), instead of the less hindered position of the phenyl ring existing in conformation B. Confronted with this unecpectedly required conformation, Solladié-Cavallo decided to determine the structure of the starting sulfonium salt.

They were able to crystallize the salt, and X-ray analysis finally revealed that the benzyl group is axial in solid state. Therefore, conformation A’ (Scheme 12) should be used in their model of approach instead of conformation A, Scheme 14 (26).

This was the first example of a six-membered sulfonium salt having the third axial group at sulfur. The explanation was proposed to be the 1,3-anomeric effect: the equatorial lone pair may overlap with ó* of the axial C-O bond and so be less nucleophilic than the axial lone pair. There is also an alternative steric argument, with the consideration of gauche interaction with a neigbouring methyl group, that may also operate (14).

1.3.4.2. Exploitation of the results

The developed asymmetric ylide epoxidation method was successfully applied for the preparation of the two (R)-β -adrenergic compounds 29 and 30, Scheme 13 (27).

In the first step, (R)-monoaryl epoxides were prepared in high enantiomeric purity under monophasic aprotic conditions employing NaH as a base. The corresponding amino alcohols 29 and 30 were then obtained in one step, opening the epoxides regiospecifically by using a secondary amine developed by themselves (35).

Encouraged by the results they decided to extend this method to trans-(R,R)diaryl-epoxides to be used as intermediates for the synthesis of diaryl chiral ligands (28).

The reaction of the sulfonium salt derived from Eliel’s oxathiane 23 with various para- and ortho substituted benzaldehydes 31a-e under aprotic conditions (Scheme 14) afforded the desired epoxides 32a-e in 76% to 84% isolated yields, Table 5.

In their latest studies (29,30) phosphazene base EtP2 (EtN=P(NMe₂)₂-N=P(NMe₂)₃), instead of NaH, has been employed to generate the ylide. The conversions into desired epoxides are very high and the reaction times were significantly shortened (30 minutes instead of 1 to 2 days). The 2- and 3-pyridyl epoxides 33 and 34, as well as the only known 2-furyl epoxide 35, were synthesized for the first time in enantiomerically pure form, Scheme 15 (30).

The results are gathered in Table 6.

In agreement with previous studies with NaH as a base (28) only trans-epoxides were formed. Only in the case of 2-pyridyl epoxide 33, 12% of the cis-isomer was obtained. The R,R absolute configurations are reasonably explained by the model already proposed in the previous cases. The very high ee values observed (97-99.8%) were proposed to be caused by the participation of the salt present (Et-P₂H+TfO-) in a coordinative and/or electrostatic interaction in the transition state leading to the cyclization (Figure 4).

The participation of the salt present in this manner was proposed for the first time in the mechanistic rationale for the preparation of trans-2-acrylcyclopropane carboxylates by the use of a phosphazene base (36).

Also trans-disubstituted aryl-vinyl epoxides were prepared in the same manner, with high enantioselectivities and yields (30). The reactions yielded both epoxides as well as cyclopropanes. Both CH₂Cl₂ and THF were used as a solvent but when a p-methoxy group was present in the arylsulfonium salt, the epoxide was the sole product, whatever the solvent. It was proposed that the methoxy group would lead to a less stabilized and more reactive ylide, favoring the formation of epoxides. It was also found that higher percentages of cis-epoxides were formed in CH₂Cl₂ (3-23%) than in THF (3%).

1.3.5.1. Stoichiometric enantioselective epoxidations

Stoichiometric enantioselective epoxidations were performed employing sulfides 40, 42, 46 and 48, i.e. those with benzylthio-moiety. The sulfides were found to react smoothly with methyl iodide furnishing the corresponding sulfonium salts without the aid of silver salts. Various aldehydes were tested and reactions were conducted in CH₃CN at room temperature, Scheme 18.

For aromatic aldehydes, the reaction proceeded perfectly giving selectively the trans products, mostly with excellent yields and moderate to good ee values (19-77%). Any efforts to extend this reaction to aliphatic and heteroaromatic aldehydes and ketones were unsuccessful due to the side reactions of the aliphatic and heteroaromatic aldehydes, and the low reactivity of the ketones. The results are gathered in Table 7 (37).

The opposite asymmetric induction was achieved when employing 40 and 42 (benzylthio group at exo position) versus 46 and 48 (benzylthio group at endo position) as chiral auxiliaries.

It was also noted that the free hydroxyl group at C₂ plays an important role; when OH was converted to a methoxyl group, both the yield and the ee value of the resulting epoxide were greatly lowered. This led to postulate that there might be a nonbonded interaction between the OH and the carbonyl group of aldehydes before attack by an ylide, Scheme 19.

This interaction forced the aldehydes to approach the reactive site of the ylide, preferentially from the si-face, and the aldehyde carbonyl to be attacked on the re-face.

1.3.5.2. Catalytic enantioselective epoxidation

Generally, chiral substances, other than natural products, are difficult to obtain. Therefore, an attempt to make a stoichiometric reaction catalytic is believed to be of practical significance (40).

Sulfonium salt 50 (Scheme 23) could be prepared either by methylation of 40 or by benzylation of 41. In addition, the stoichiometric ylide epoxidation was realized through the transfer of the benzylidene group of ylide 51 instead of the methylene group. Dai reasoned that it should be possible to make this reaction catalytic when sulfides 41,43,47 and 49 were used to mediate the reaction between benzyl bromide and aldehydes. In these reactions, the sulfides were converted initially to sulfonium salts by the corresponding ylides in situ under phase-transfer conditions. The ylide was subsequently reacted with aldehydes to furnish epoxides and release the sulfide and permit it to enter a new cycle. The process employing sulfide 41 as a catalyst is illustrated in Scheme 20 (37).

The catalytic epoxidations proceeded smoothly in the presence of 0.2 equivalents of the above mentioned chiral sulfides. The results are gathered in Table 8.

As expected, the opposite asymmetric induction was again observed in the catalytic reaction when 41, which contains an exo-methylthio group, and 47 or 49, which contain an endo-methyl thio group, were used.

The opposite asymmetric induction in the catalytic epoxidation was also observed in the cases using 41, which possesses a free OH group, and 43, which contains a methoxy group instead, although the asymmetric induction of the latter was rather low. The same phenomenon was reported earlier by Furukawa et al. (20).

Also the effects of solvents and bases were investigated. In strong polar solvents, such as DMSO and DMF, the ee values of the products were decreased. Under other conditions, the ee values remained nearly the same. In the commonly used THF, the yields were low due to the low solubility of the base (KOH), leading to difficulties in producing ylides. Acetonitrile was reported to be the best solvent for this reaction, and strong bases like KOH (s), NaOH (s), and aqueous NaOH are useful. Solid KOH was mentioned to be the most suitable base when convenience was taken into account.

Increase in the amount of the sulfides did not influence the ee values, but did shorten the reaction time (15 h with 20 mol% of sulfides) and improved the yields. From the practical standpoint they came to the conclusion that 20 mol% of the sulfides are suitable for the catalytic reaction (37).

1.3.6.1. C₂ symmetric sulfide ligand

Metzner et al. (41,42,43) have reported the preparation and exploitation of a cyclic, C₂ symmetric sulfide, (2R,5R)-dimethylthiolane 56, Scheme 21.

Thiolane 56 was prepared easily in two steps starting from the commercially available (2S,5S)-hexanediol 55 (41,42). This diol can also be obtained by the enzymatic reduction of the cheap 2,5-hexanedione with baker’s yeast (44). Another yeast, Pichia farinose leads to an opposite enantiomer (45) and, thereby, to opposite asymmetric induction in the epoxidation process. The activation of the hydroxyl groups into mesylates and subsequent cyclization with sodium sulfide, by two nucleophilic substitutions with inversion, furnished the desired C₂ symmetric sulfide 56 in 95% yield.

The epoxidation procedure was the same as reported earlier by Furukawa (20) and Dai (37), i.e. a mineral base is involved and all reagents are mixed together in one pot at room temperature, Scheme 22. The reaction of benzaldehyde (1 eq) with benzyl bromide (2eq) and thiolane 56 was carried out with KOH or NaOH (2 eq) in various solvents. Experiments in nonpolar or moderately polar solvents, under heterogenous (toluene/aqueous NaOH) or homogenous (THF/ aqueous NaOH) conditions, furnished poor yields or selectivities. In more polar solvents (DMF, DMSO, CH₃CN, alcohols) with powdered KOH they observed various predominant side reactions. These problems were largely avoided by simply adding 10% of water into the solvent and, thus, stilbene oxide was obtained in excellent yields and enantioselectivities, in one or two days at ambient temperature (41), Table 9.

It is noticeable that the reaction could even be carried out in pure H₂O (Table 9, entry 5). In all cases, the trans isomer of stilbene oxide (9a) was formed preferentially, i.e. not diastereoselectivically, and the major isomer was identified as the trans-(2S,3S)-diphenyloxirane. The best enantiomeric excess, 94%, was achieved with the 9:1 EtOH:H₂O solvent system.

The reaction times are rather long varying between one and eight days. The feasibility of the catalytic process was tested by employing only 0.1 equivalents of thiolane 56 in the presence of benzyl bromide (2 eq) and benzaldehyde (1 eq) in 9:1 t-BuOH:H₂O with NaOH. Both yield and selectivity remained high but it took a whole month to complete the reaction.

1.3.6.2. Asymmetric induction

The high asymmetric induction was rationalized as shown in Scheme 23 (41).

The C₂ symmetry of 56 dictated the formation of the single sulfonium salt 57 by reaction with benzyl bromide. Deprotonation afforded the ylide 58 with a planar carbon and a tetrahedral sulfur, the sulfur doublet lying in the plane of the ylide carbon substituents to avoid repulsive interaction with the carbanion doublet. Out of two possible conformations, 58a and 58b, the former is favored as the phenyl group is away from the thiolane ring. Attack from the si face of the ylide is suggested to be preferred because the re face is hindered by the methyl group cis to the benzylidene group. A 109° approach of the aldehyde, leading to the trans epoxide 9a, avoids gauche interaction between the ylide and aldehyde phenyl groups. This model results in the formation of the trans- (S,S) enantiomer, exactly as observed experimentally (41).

1.3.7.1. Development of a catalytic cycle

To render the ylide epoxidation more synthetically useful, the catalytic mode was set as the main target for the group (53). It is plausible that the formation of the sulfonium salt in the catalytic cycle is rate-determining, and it is essential that this step is fast. Compared to previous work made in the field of asymmetric epoxidation of aldehydes, an alternative method for ylide formation, i.e., by the reaction of carbenes or carbenoids with heteroatom lone pairs, was taken into account (51,52). The reaction between sulfonium salts and carbonyl compounds giving epoxides, had usually been mediated by a base. However, carbonyl compounds containing sensitive functional groups may not be compatible with those basic conditions required. The catalytic cycle (Scheme 26) presented by Aggarwal et al. (53,54) reduced the conventional two-step sequence for ylide formation to one step.

Moreover, the reaction was now carried out under neutral conditions and is, therefore, applicable to readily enolizable and base-sensitive aldehydes (53).

The metal-catalyzed decomposition of diazo compounds in the presence of sulfides has often been used to prepare unsaturated sulfur ylides, which can then undergo [2,3]- sigmatropic rearrangements (55). In the catalytic cycle itself, ylides are generated in situ in the presence of carbonyl compounds in a reaction between a sulfide and a carbenoid, which in turn is generated from a diazo compound and a metal catalyst. At the same time ylides react with carbonyl compounds to give epoxides and are simultaneously regenerated to continue the process, as shown in Scheme 26. This method had never been used to generate sulfur ylides in the presence of carbonyl compounds and was therefore patented by Aggarwal et al. at 1995 (56).

Mixing together an aldehyde, a sulfide, a metal catalyst and a diazo compound certainly leads to competitive side reactions. For example, diazo compounds are well-known to react with carbonyl compounds directly to give homologated products (57), eq 1, Scheme 27. They are also known to dimerize in the presence of metal salts (58) (eq 2).

To minimize the extent of these potential side reactions one needs to maintain a low cocncentration of the diazo compound, and this can be achieved by slow addition (53).

Due to the convenience of the laboratory, Aggarwal et al. chose to start with diphenyl sulfide (a nonvolatile sulfide). Phenyldiazomethane 68 was chosen, in the first place due to its greater stability (59) compared to alkyldiazo compounds. The diazo compound was added, over three hours, to a solution of the above reagents, but the only product isolated was stilbene 71. This showed that the reaction of the intermediate metallocarbene with phenyl diazomethane 68 (k_Dim) was faster than the reaction of the metallocarbene with the sulfide (53). To improve the rate of the latter reaction, diphenyl sulfide was replaced with the more nucleophilic dimethyl sulfide 72, Scheme 28. This time stilbene oxide 9a was obtained in good yield. Besides benzaldehyde also other aromatic and aliphatic aldehydes worked well, giving good yields of epoxides, Table 11.

Epoxides were now formed with minimal interference from alternative side reactions. Without Rh₂(OAc)₄ only homologated products were obtained, showing that ylide formation required metal catalysis.

To give the reactions the greatest chance of success they were carried out using stoichiometric amount of sulfide in the presence of a catalytic amount of metal salt. The first attempts to use dimethyl sulfide 72 in catalytic quantities (20 mol%) gave reduced yields of epoxides together with stilbene 71 and homologated products. Aggarwal et al. reasoned that the sulfide was being “held up” in the catalytic cycle at either the sulfur ylide stage (slow k₃, eq 3) or at the betaine stage (slow k₄, eq 4) and decided to add the diazo compound over 24 hours, Scheme 29. Under these conditions good yields of epoxides were obtained for both aromatic and aliphatic aldehydes (53), Table 12.

The success of their catalytic cycle was reasoned to rely on the reactions represented in equations 1-4 in Scheme 29. They proceed at a significantly faster rate than the potential side reactions presented in Scheme 27.

1.3.7.2. Choice of aldehyde

As seen in Table 12, aromatic aldehydes gave much greater ratios of trans epoxides than aliphatic aldehydes. For comparison, benzaldehyde was chosen as the representative aldehyde (60).

1.3.7.3. Choice and generation of diazo compound

From the beginning of the project evolution Aggarwal felt that the most critical choice would be the choice of the diazo compound (60). He considered the possibility of using ethyl diazoacetate, phenyl diazomethane and diazomethane itself. Ethyl diazoacetate was, however, ruled out because the corresponding sulfur ylide that would have been generated is known not to react with aldehydes to give epoxides because it is too stabilized (61,62,63,64). Because of the greater reactivity and hazards associated with diazo methane, he primarily decided to focus on phenyl diazomethane 68, Scheme 30.

Preparation of phenyl diazomethane. The diazo compound was prepared by a literature method (59) in which sulfonamide 75 is first prepared from p-toluenesulfonyl chloride 73 and benzylamine 74 followed by nitrosation yielding the corresponding nitroso compound 76. The treatment of 76 with sodium methoxide in an ethereal solution gives phenyl diazomethane 68 as a blood red solution, Scheme 30.

Generation of diazo compound in situ. Diazo compounds are commonly known to be potentially explosive and this was regarded as a certain limitation of the original protocol (65). It was, therefore, of great interst to generate the diazo compound in situ and incorporate this reaction into an established epoxidation process. They managed to develop a pocess for the direct coupling of two different aldehydes to form epoxides with control over the relative and absolute stereochemistry (66). The process was patented (67) and extended to cyclopropanations and aziridinations, as well (68).

They focused their efforts on the use of the Bamford-Stevens reaction to generate the diazo compound (69), and found, after some experiments, that warming the suspension of the tosylhydrazone salt 77 (Scheme 31) in the presence of a phase-transfer catalyst (to aid the passage from the solid to the liquid phase), allowed the generation of the diazo compound at slightly elevated temperatures.

This protocol proved to be compatible with their established epoxidation process and was remarkably efficient. Furthermore, these new conditions proved to be general (66), Table 13. All the aromatic, heteroaromatic and unsaturated aldehydes investigated furnished the corresponding epoxide in high yields and with high trans selectivites. Aliphatic aldehydes gave lower yields as well as selectivities, though. Tetrahydrothiophene (20 mol%) was employed as a mediator (66), Scheme 31.

A more diverse range of tosylhydrazone salts was investigated in the epoxidation process (66), Scheme 32 and Table 14.

The yields of epoxides obtained were excellent, except where a sterically hindered aryl reagent was employed (entry 6). In general trans epoxides were obtained exclusively, but tosylhydrazone salts bearing electron-donating substituents led to lower diastereoselectivities (66).

The next step was to consider the possibility of generating the hydrazone itself in situ. This worked as well, providing a one-pot, atom-economical method for coupling two different aldehydes to give epoxides (66), Scheme 33 and Table 15.

Thereby, Aggarwal et al. had developed a general, user-friendly catalytic process for preparing epoxides by coupling together a carbonyl compound with either a tosylhydrazone or a second carbonyl compound.

1.3.7.4. Choice of sulfide

Employment of Durst’s sulfides in catalytic cycle. The first sulfides Aggarwal et al. tested (Scheme 24, Table 10) in asymmetric epoxidation had sulfur atom situated outside the ring system, resulting in low enantioselectivities (46) (up to 43%). Encouraged by the results achieved by Durst et al. (19,23) they decided to prepare Durst’s chiral sulfides 78 and 79 and to test those in their original catalytic cycle for epoxidation (53,70), Scheme 34.

The unhindered sulfide 78 participated well in the catalytic process, furnishing trans-stilbene oxide 9a in good yield but as expected, with low enantioselectivity (53). Under these neutral conditions the yield was significantly higher, though. Aggarwal et al. achieved 58% yield compared to 39% presented by Durst and co-workers (19,23).

To test, whether hindered sulfides could be applied in the catalytic cycle, Aggarwal et al. decided to run a test with sulfide 79 (70). Durts et al. had received remarkable results employing the very same mediator in their own studies (38%, ≥ 96% ee). No epoxides were obtained at all when employing Rh₂(OAc)₄ as a metal catalysts, only stilbenes. However, using Cu(acac)₂ in place of Rh₂(OAc)₄ and employing a stoichiometric amount of Durst’s sulfide 79, they were delighted to find that epoxidation was the dominant process again. The enantioselectivity was surprisingly moderate, only 72% (71).

To obtain higher enantiomeric excess, Aggarwal et al. assumed that the two groups attached at sulfur needed to be more dissimilar as compared, for example, to the unhindered sulfide 78, (72).

Design, preparation and testing of oxathiane ligand family. Thus, chiral sulfides 80 and 81 were chosen for study and were easily prepared in two steps (72), Scheme 35. Reduction of (+)-camphorsulfonyl chloride 80 with lithium aluminium hydride gave a mixture of the exo and endo products 5a and 5b, respectively. The thiols were separated by chromatography before cyclization.

The chiral sulfides 81 and 82 were then employed in the epoxidation cycle to prepare stilbene oxide 9a. The reactions were conducted using both stoichiometric and catalytic amounts of sulfides (72), Table 16.

The employment of sulfide 82 resulted in only moderate selectivity, but sulfide 81 gave improved ee’s of 41% and a diastereoselectivity of 10:1. The use of stoichiometric amounts of sulfides gave good yields of epoxide, but as the amounts of sulfides were reduced, a notable drop in epoxide yield was observed. The reduced amounts of sulfides result in lowering the rates of sulfur ylide formation as well as the subsequent reactions which, in turn, can be noticed as reduced yields of epoxides.

In order to maintain similar reaction rates the concentration of the sulfide was maintained the same by simply reducing the volume of the solvent used. Employing the sulfide 81 under these conditions stilbene oxide 9a could be prepared in 83% yield. The corresponding yield with sulfide 82 was 75% (72).

There is very little steric difference between the two faces of ylide 83. Therefore, Aggarwal et al. were a little surprised at the level of enantioselection shown by this sulfide. No mechanistic rationale was presented but it was believed that a single sulfonium ylide is formed as the alkylation of 81 furnishes the sulfonium salt 83, Scheme 36.

The stereochemistry of 83 was confirmed by the studies. Irradiation of the benzylic protons gave an enhancement of the axial protons flanking the sulfide moiety, thus, indicating its equatorial position and also the orientation of the phenyl group (72).

Aggarwal et al. deemed important that in the design of other chiral sulfides single sulfonium salts or single sulfonium ylides should be formed. If mixtures of diasteromeric sulfonium salts were formed then these would certainly react with different, but possibly opposite enantioselectivity and would therefore reduce the levels of enantiomeric excess observed in the product epoxides (70).

Thus, any chiral sulfide that is designed should include one lone pair that is accessible and reactive and another lone pair which is hindered and unreactive. The previously prepared oxathiane 81 seemed like an ideal templade for next generation studies. Of the two lone pairs only the equatorial one should react with the metallocarbene as the axial one is hindered by the bridging gem-dimethyl group. Moreover, it should be easy to alter the steric and electronic environment around the sulfide to maximize enantioselectivity by simply changing the aldehyde or ketone used to form the thioacetal. On this bases they designed and prepared a (+)-camphorsulfonyl chloride 80 based ligand family of general structure 84 (73), Scheme 37.

Most of these thiocetals were easily prepared via the known hydroxy thiol 6 (74,75,76), preparation of ligands 84p as well as 84q proved to be unsuccessful, though (Table 17).

The thioacetals were tested in a catalytic epoxidation process (Scheme 38), and the results are summarized in Table 18 (73). Sulfide 84a (R = Me, R’ = H) proved to be the most effective catalyst both in terms of enantioselectivity and yield. Increasing the size of the R group did not give any increase in enantioselectivity but instead lovered the yield (entries 2-4). For R = t-Bu (84c), no epoxide was obtained, stilbenes were formed instead. Evidently, the rate at which the sulfides react with the metal carbenoid is dependent upon the size of R and R’, and if either one is too hindered, epoxide formation slows down and stilbene formation dominates (73).

For R = Ph (84d), no epoxide was obtained but only limited amounts of stilbenes were formed. In this case, ring expanded thioacetals 85 were believed to have been formed via the Stevens rearrangement (73). The Stevens rearrangement of the ylide competes with carbonyl epoxidation and, in this case, is facilitated by the stability of the intermediate phenyl-substituted radical, Scheme 39.

A single heteroatom in the side chain can be tolerated (entries 7-9 and 14), and the similar enantioselectivities obtained for R = Me showed there was no substantial electronic effect. The trichloro-derivative 84l gave no epoxide, only stilbenes. It was suggested that this sulfide is either sterically too hindered or electronically too deactivated by the chlorines to react with the metal carbenoid, and therefore stilbene is obtained instead.

Aggarwal et al. expected the TMS derivative 84m to be less sterically hindered than the corresponding sulfide 84c and to be electronically activated to react with the metal carbenoid. Indeed, epoxide was obtained but in low yield and with reduced enantioselectivity. The presence of two alkyl groups adjacent to sulfur gave much reduced yields of epoxide, and it was supposed this was due to increased steric hindrance. The enantioselectivity was also low (73).

The next step was to test the optime catalyst 84a for a range of aldehydes (73), Table 19. It was found that good yields and high diastereo-and enantioselectivities were obtained for other aromatic as well as unsaturated aldehydes. However, reduced yields and diastereoselectivities were observed with aliphatic aldehydes, but the level of enantioselectivity was essentially maintained. The reaction of valeraldehyde (entry 6) gave considerably reduced enantioselectivity, Table 19.

Origin of diasteroselectivity. To account for the origin of the enantio- and diastereoselectivity it was necessary to find out whether the sulfur ylide reactions were under kinetic or thermodynamic control. From crossover experiments Aggarwal et al. found that the addition of benzylsulfonium ylide to aldehydes was remarkably fine balanced (77).

Epoxide formation is essentially under kinetic control. The trans epoxide was derived directly from the irreversible formation of the anti betaine or indirectly from the resversible formation of the syn betaine. The cis epoxide was derived from the partially reversible formation of the syn betaine. The higher trans selectivity observed in reactions with aromatic aldehydes, compared to those with aliphatic aldehydes, was proposed to be due to greater reversibility in the formation of the syn betaine (77), Scheme 40.

The reactions of simple sulfides (dimethylsulfide, tetrahydrothiophene) in the catalytic cycle with benzaldehyde gave stilbene oxide as an 86:14 ratio of trans/cis epoxides (53). However, when the camphor derived oxathiane 84a was employed, only trans epoxides were formed with a wide variety of aldehydes (70). This higher selectivity was proposed to be due to an increase in k_-3 relative to k₄. An increase in k_-3 relative to k₄ would be expected for sulfides of increasing steric hindrance or where the ylide had increased stability. In the case of the 1,3-oxathiane, Aggarwal et al. proposed that the corresponding ylide shows increased stability relative to simple benzyl sulfonium ylides as a result of the anomeric effect (73). The positive charge on sulfur can be delocalised over the oxygen, and this will lead to increased stability and therefore greated reversibility and thus greater trans selectivity. A moderate increase in the stability of the ylide is evidently not sufficient to promote reversibility in the syn betaine formation with aliphatic aldehydes, however (77).

Origin of enantioselectivity. Having established that trans epoxides are derived from the irreversible formation of anti betaines, Aggarwal et al. focused on the transition state leading to the formation of epoxides in order to understand the origin of enantioselectivity. As this was not possible, they concentrated on gaining information on the structure of the ylide (73). It was shown earlier that the alkylation of the related oxathiane 81 only gave equatorial sulfonium salt (72). Thus, a single sulfonium ylide was believed to be formed again. Ylide conformation has been studied by X-ray diffraction (16f,g), NMR (16a,c,63,78,79,80) and computation (81,82,83). All of these studies indicate that the preferred conformation of the sulfur ylides is the one in which the filled orbital of the ylide carbon is orthogonal to the lone pair of the sulfur. The barrier of rotation around the C-S bond of the semistabilized ylide, dimethyl sulfonium fluorenide, has been found to be 42 1.0 kJ mol-1 (80). This information implied that the sulfonium ylide derived from 84a will adopt conformations 86 and 87 (Scheme 41) and these will be in rapid equilibrium at room temperature (73).

Of these two, conformation 87 will be favored because 86 suffers from 1,3-diaxial interactions between the phenyl ring and the axial groups. The aldehyde can then approach either face of the ylide, but the re face is more accessible as the si face is hindered by the equatorial methyl group (70,73), Scheme 41.

The aldehyde can react in an end-on or [2 + 2] mode, but there is no evidence, experimental or theoretical, to indicate which one is preferred. While most examples invoke the end-on mode, circumstantial evidence that the [2 + 2] mode may be favored has been reported (19). The end-on and [2 + 2] transition states in the reaction of ylide 87 with benzaldehyde are shown in Scheme 42 (73).

From the analysis of the molecular models of these transition states it was clear that they all could be accommodated. Thus, the results did not provide any further evidence as to which mode is favored. All the transition states presented in Scheme 42 account for the high enantioselectivities observed.

The enantioselectivity obtained for sulfide 84a in the epoxidation of benzaldehyde was 93% (70). Therefore, Aggarwal et al. studied whether the minor enantiomer arose from the si attack (73). If so, enantioselectivity should be highly dependent upon the size of the equatorial substituent. Increasing the size of this substituent from Me to i-Pr did not result in any contominant increase in selectivity, however. The enantioselectivity was essentially the same for a range of substituents suggesting that the facial selectivity was rather complete. Lower selectivities with ligands having an aromatic ring in the side chain were proposed to be caused by possible -stacking opportunities to an incoming aromatic aldehyde.

Even in the absence of a substituent (sulfide 81), good re-face selectivity was still observed (41% ee) (70). Aggarwal et al. reasoned that the oxygen of the oxathiane would affect the facial selectivity of the ylide exerting through a combination of anomeric (84) and Cieplak effects (85,86,87).

A resonance form of the ylide derived from 84a is presented in Scheme 43. Aggarwal et al. reasoned that if there were be a contribution from this resonance form to the ground-state structure of the ylide, then the C₄-S bond should be more electronrich than the corresponding C₂-S bond which, in turn, should affect the face selectivity of the ylide. The resonance form in Scheme 43 is a result of the anomeric effect and should appear in the shortening of the C₄-O bond relative to the C₆-O bond and the lengthening of the C₄-S bond relative to the C₂-S bond (73). A significant anomeric effect was indeed proved true by preparing oxathiane 88, sulfoxides 89 and 90 and studying the structures by X-ray diffraction, Figure 5.

The assumptions came true already with oxathiane 88. The ylide mimicking sulfoxide 89 showed even greater differences in bond lengths (73), Table 20. The control compound, sulfoxide 90, showed essentially no differences in bond lengths.

In order to gain further evidence of this combined anomeric and Cieplak effect, Aggarwal et al. decided to prepare and test the sulfur and carbon analogues of 81 (73). The preparation of sulfides 93 and 99 are presented in Schemes 44 and 45.

Dithiane 93 was furnished from dithiol 92 by treating it with paraformaldehyde and tosic acid under Dean-Stark conditions. The preparation of carbon analogue was started by adding vinylcerium to 91 giving the exo alcohol 94, which upon treatment with 9-BBN gave the corresponding 4-hydroxythiane 103. The treatment of 95 with oxalylchloride resulted in elimination, giving sulfide 96. The unsaturated thiane was reduced to thiane 97, which was obtained as a mixture of diastereomers (3:1). Oxidation to sulfoxide 98 led to a substrate that they were able to crystallize, and subsequent reduction gave 99, Scheme 45.

Dithiane 93 and sulfide 99 were now tested in the epoxidation process. Sulfur analogue was expected to show enhanced selectivity (greater anomeric effect), and the carbon analogue should show reduced selectivity (no anomeric effect). Dithiane gave only slightly higher selectivity, while the carbon analogue, sulfide 99, gave much reduced selectivity (73), Table 21.

The stereochemical outcome of the ylide reaction is supposed to be determined by two factors, the face selectivity of the ylide and the conformation of the ylide (73), Scheme 46.

Dithiane 93 was expected to give enhanced selectivity due to an increase in the anomeric effect, which in turn should increase the face selectivity. It seemed, however, that the replacement of oxygen by sulfur results in a more distorted chair and, therefore, reduced 1,3-diaxial interactions, which in turn result in an increase in the population of conformer 102, Scheme 46. The ylide conformations may very well react with increased face selectivity, but a greater proportion of conformer 102 will, in turn, reduce the enantioselectivity. As expected, the carbon analogue 99 gave reduced selectivity. Since both dithiane 93 and oxathiane 81 adopt a verysimilar conformation (C-O and C-C have similar bond lengths), the ratio of conformers 87/86 and 101/103 will be similar. The difference in selectivity must result from a difference in the face selectivity of the ylide. Thus, the oxygen of the oxathiane exerts a significant stereoelectronic effect in promoting the reaction on the face opposite to this group (73).

The facial selectivity in the reaction of 84 was confirmed to be dependent on the size of the aldehyde; α-branched, aromatic and unsaturated aldehydes reacted with complete facial selectivity, while unbranched aldehydes reacted with moderate facial selectivity. Therefore, the source of the minor enantiomer was proposed to be the reaction of the minor ylide conformation 86, Scheme 45.

In order to reduce the amount of this minor conformation, Aggarwal et al. prepared thioacetals bearing axial substituents 84i (R, R’ = Me) and 84j (R = spiro-cyclobutyl) to increase 1,3-diaxial interactions. However, instead of increased selectivity, a decrease in enantioselectivity was observed. A reliable explanation was found with nOe experiments. They found that while 84a adopted the chair form only, sulfide 84i existed in both chair and boat forms. The reaction of the corresponding ylide from the boat conformer of 84i should give the opposite enantiomer to that from the chair form. The boat form could be favored because of the severe 1,3-diaxial interactions in the chair form (73).

The role of the minor conformer was further confirmed by replacing phenyl diazomethane with the bulkier mesityl diazomethane (2,4,6-MePhCHN₂). Thus, employing sulfide 84a as a catalyst increased greatly enantioselectivity. The epoxidation of benzaldehyde produced the corresponding trans-di-aryl epoxide with 98% ee (73).

[2.2.1] bicyclic sulfide. The employment of sulfide 84a in previously developed conditions, where the diazo compound is generated in situ (Scheme 31, page 51), resulted in failure, only low yields of the epoxide were obtained (66). Therefore, Aggarwal et al. embarked on the synthesis of a range of stable, mono and bicyclic chiral sulfides based on thiolanes, thianes and 1,4-oxathianes. From this study they found that the bridged bicyclic sulfides 108 and 109 performed exceptionally well (66). The bicyclic sulfides 108 and 109 were prepared in four steps from camphorsulfonyl chloride 80, Scheme 47.

The key step in the synthesis was the cycloaddition of thiolaldehyde 105 (generated photochemically from phenacyl sulfide 104) with cyclopentadiene (88,89). This reaction was reported to be both highly endo- (20:1) and diastereoselective (20:1).

The conditions for epoxidation were optimised with sulfide 108 and it was found that the sulfide loading could be reduced from 20 mol% to 5 mol% without significantly affecting the yield or ee value. The asymmetric epoxidation of benzaldehyde, employing sulfide 108 as a catalyst, produced trans-stilbene oxide with 94% ee and 82% yield. The results were essentially the same (92% ee, 82% yield) when sulfide 109 was employed. The process was even scaled up to a 50-mmol scale without any significant loss of yield (92% ee, 78% yield) (66).

The following model was proposed to account for the high enantioselectivities observed, Scheme 48. Of the two lone pairs only the exo lone pair reacts to form a single sulfonium ylide. The ylide can adopt conformations 110 or 111, but 111 should be strongly favored because of the 1,4-steric interactions present in 110 (66).

The face selectivity of the ylide is then controlled by the bulky camphor group which blocks attack from the si face, thus leading to the R,R epoxide. High enantioselectivities require both control of the conformation and high face selectivity of the ylide. The control in the ylide rigidity is proposed to arise from the conformational rigidity of sulfides 108 and 109, which would mean that the ylide conformer 110 can not be easily accommodated through small changes in the bond angles around the sulfur atom. Thus, there is a significant difference in the energy between the two ylide conformers 110 and 111, which in conjunction with the high face selectivity imposed by the bulky camphor moiety resulted in the high enantioselectivity observed (66).

1.3.7.5. Choice of metal catalyst

The first catalytic epoxidations were conducted employing Rh₂(OAc)₄ as a metal catalyst (53). As noted earlier, Rh₂(OAc)₄ did not work with the sterically hindered Durst sulfide 79 (70). Evidently, when the sulfide is slightly more hindered, the reaction of the metallocarbene with the sulfide becomes significantly lower and the reaction of the metallocarbene with phenyl diazomethane begins to dominate (70). Therefore, the relative rates of these competing reactions should be changed and this was supposed to be possible by simply changing the ligands of the rhodium. In the case of the intramolecular reactions of metal carbenes it has been shown that reaction pathways can be dramatically influenced by altering the ligands around rhodium (90,91,92,93). Therefore, a range of rhodium salts were prepared and tested in the catalytic cycle with the hindered Durst sulfide 79 (94). Ligands on rhodium, from strongly electron withdrawing to strongly electron donating groups, were tested without any success, only stilbenes were formed.

Aggarwal et al. reasoned that the wall of ligands around the rhodium carbene would present significant steric hindrance towards an incoming hindered sulfide. To make the carbene transfer to hindered sulfides more efficient, less crowded metal carbenoids were required (70). Copper salts seemed ideal as prior to the formation of the carbenoid, copper (II) is reduced to copper (I) and one of the charged ligands dissociates (95), Scheme 49.

Using Cu(acac)₂ in catalytic cycle with the hindered Durst sulfide 79 they indeed obtained trans-stilbene oxide (70). The first attempts at employing their best sulfide 84a with Cu(acac)₂ resulted in failure, only 3% yield of epoxide was obtained. This was due to the rapid decomposition of the thiocetal (within a few seconds) catalyzed by Cu(acac)₂. Under strictly anhydrous conditions the rate of hydrolysis could be reduced, but the desired improvement in yield was not obtained. There was a strong indication that the commercial material contained small quantities of impurities that showed high levels of catalytic activity in thioacetal hydrolysis. The solution to the problem was the sublimation of the metal catalyst. Thus, when the purified Cu(acac)₂ was employed in the catalytic cycle with 20 mol% of 84a, stilbene oxide was obtained in good yield and high enantioselectivity (70), Table 18.

Aggarwal et al. also tested whether the choice of metal salt would have any effect on the yield or diastereo- or enantioselectivity of the epoxidation process. The reactions were carried out in CH₂Cl₂ using 20 mol% of their optimum sulfide 84a with benzaldehyde and a range of metal catalysts (73), Scheme 50, Table 22.

All the metal salts employed gave essentially the same enantioselectivity. The results showed that the choice of metal salt used for the diazo decomposition had little effect on the enantioselectivity of the process. It also showed that metal did not participate in any way in the reaction of the sulfur ylide with the aldehyde. The optimum metal salt for sulfide 84a was copper acetylacetonate, Cu(acac)₂ (73). It is also important to note that Rh₂(OAc)₄ now worked well with the hindered sulfide 84a whilst tests with the hindered Durst sulfide 79 had resulted in complete failure.

1.3.7.6. Choice of solvent

Aggarwal et al. also investigated the effect of the solvent on the epoxidation process (73). Reactions were carried out employing copper tetramethylheptanedionate as a metal catalyst because this catalyst was soluble in all of the solvents they wished to investigate. The results are collected in Table 23.

The most successful sulfide catalyst 83a was employed. As it can be seen all the solvents gave essentially the same enantioselectivity and the highest yields were obtained using CH₂Cl₂ as a solvent (73).

1.3.7.7. Application to ketones

It is known that the dimethylbenzylsulfonium ylide reacts with ketones to give epoxides, but the reactions require elevated temperatures (60 °C) (96). Thus, Aggarwal et al. investigated reactions with cyclohexanone at various temperatures, Table 24 (94). In order to make various temperatures possible they chose the higher boiling tetrahydrothiophene instead of dimethyl sulfide as a mediator. It was found that optimum yields were obtained at 35 °C.

At lower temperatures the ylide most probably does not react rapidly enough while at elevated temperature the ylide may suffer decomposition.

1.3.7.8. Application of Simmons-Smith epoxidation

The Simmons-Smith reaction (97) and recent modifications (98) thereof are known as a powerful method for the diastereo- (99) and enantioselective cyclopropanation of alkenes (100). Some attempts to explore the reaction of Simmons-Smith reagents with sulfides had already been published (101) and Aggarwal et al. were interested to test whether the Simmons-Smith reagents could react with sulfides to generate ylides (102,103). If so, these zinc-derived sulfur ylides could then be used to generate epoxides from aldehydes. Aggarwal et al. investigated the reaction of the zinc carbenoid derived from Et₂Zn and ClCH₂I with tetrahydrothiophene, Scheme 51.

In these organozinc-mediated epoxidation reactions of aldehydes three equivalents of sulfide were employed, Scheme 52. Terminal epoxides were obtained in high yield (103). Some of the results are collected in Table 24.

Aromatic and aliphatic aldehydes worked well and the reaction even tolerated unsaturated aldehydes (entry 5) without undergoing cyclopropanation. This indicated that the reaction of the zinc carbenoid with the sulfide was much more rapid than the reaction with the alkene (103).

When diastereomers were formed, the diastereoselectivity was similar to that observed in the reactions of the trimethylsulfonium ylide with the aldehydes (104). In no case any racemization could be detected while it has been reported that in the case of the aldehyde derived from phenylalanine, substantial racemization occurs in the ylide epoxidation process (105). The epoxides derived from phenylalanine (entry 79) are reported to be key intermediates in the syntheses of many HIV protease inhibitors (105,106).