Enantioselectivity of the reaction

As discussed for palladium in section 2.2.3, there can be several enantiodetermining events for the considered allylic substitution reaction (Carroll rearrangement). In the case of CpRu chemistry, there is one more element to be considered: the stereogenicity of the metal atom.

4.7.1. Stereoselectivity at the Ru-centre

If one reasons that the catalytically active species is an unsaturated 16-electron Cp-ruthenium complex (Scheme 4-2), two diastereomeric forms can be obtained with an enantiopure C₁ -symmetrical ligand upon coordination of a third ligand. Indeed, due to the pseudo tetrahedral geometry in an 18-electron piano-stool shaped complex of the Ru-atom. In the presence of four different ligands, two RRu or SRu diastereomeric complexes arise. This has been thoroughly studied by Brunner⁹ and Davies¹⁰ with (among others) pymox ligands and Cp or η⁶-arene ruthenium piano-stool complexes respectively. With a bidentate enantiopure C1 -symmetrical ligand, the two possible orientations of the ligand provide two different isomeric complexes as show in Scheme 4-9. In the first case, the proximity of the indanyl arene moiety of L18f with the Cp ring results in important steric repulsion (Scheme 4-9 (a)). Case (b) seems like the most liable complex to form a labile olefin complex that will allow the reaction

8 Lehmann, J.; Lloyd-Jones, G. C. Tetrahedron 1995, 51, 8863-8874.

9 Brunner, H.; Klankermayer, J.; Zabel, M. Organometallics 2002, 21, 5746-5756..

10 (a) Davies, D. L.; Fawcett, J.; Garratt, S. A.; Russell, D. R. Organometallics 2001, 20, 3029-3034. (b) Davies, D. L.; Fawcett, J.; Garratt, S. A.; Russell, D. R. Dalton Trans. 2004, 3629-3634..

to proceed further. In this case, the ligand delimits the environment in such a way that only one relative orientation of the cinnamyl fragment is possible (Scheme 4-9 (c)).

Scheme 4-9: 3D-drawings of the two diastereomeric forms of the alleged catalytically active species with L18f (a) and (b) (third ligand omitted for clarity) and of the postulated most stable π-allyl complex (c).

It can however not be excluded that form (a) participates in the reaction and thus accounts for a part of the imperfect enantiomeric excess (ee < 100 %). Another more plausible explanation is that form (b) solely reacts and that a perfect orientation of the cinnamyl fragment is obtained (allylic CH2 pointing towards the indanyl moiety, Scheme 4-9 (c)) but that the endo / exo isomerism is imperfectly controlled (the endo and exo isomers show a different face for the nucleophilic attack). To assess this issue, [Cp*Ru(MeCN)3][PF6]¹¹ was tried as metal source for the Carroll rearrangement of ester 1a in the standard conditions. The more bulky Cp* ligand is believed to promote a better control of the endo / exo isomerism by enhanced steric interactions with the central allylic substituent. However, using L18f as ligand, the reaction was very irreproducible in terms of enantioselectivity (ee 45 - 75 %, (–)-enantiomer still being major as with 4a) and the reaction was much slower than with [CpRu(MeCN)3][PF6] (4a) as metal source. This slower reaction rate is quite surprising knowing the traditionally accepted stronger oxidative addition efficiency of Cp* complexes.¹²

11 Mbaye, M. D.; Demerseman, B.; Renaud, J. L.; Toupet, L.; Bruneau, C. Adv. Synth. Catal. 2004, 346, 835-841.

12 (a) Hermatschweiler, R.; Fernandez, I.; Pregosin, P. S.; Breher, F. Organometallics 2006, 25, 1440-1447. (b) Trost, B. M.; Older, C. M. Organometallics 2002, 21, 2544-2546.

(a)

(b) (c)

It appeared that this reaction is sensitive to the metal-ligand stoichiometry with the Cp*Ru based catalyst. This is probably due to the fact that, contrary to the CpRu species, the Cp*Ru species are catalytically active even when the bidentate ligand is not or only partially signal for free MeCN accounting for 2 equivalents. Complex 7b was thus obtained as a single isomer or an on-the-NMR-time-scale rapidly isomerizing mixture of stereoisomers. Upon bubbling of CO into the reaction mixture, the signal for the free MeCN accounted for 3 equivalents and all ¹H signals were split into two sets. Each group of peaks, accounting for two diastereomeric species of complex [CpRu(L18f)(CO)][PF6] (7c), was obtained with a 64:36 ratio. The coordination of CO to the complex being irreversible in the reaction conditions, this result could account for the kinetic ratio of the two isomers. Unfortunately the different isomers could not be separated for catalytic activity studies. As such the stereoisomeric mixtures of complexes 7a and 7c were separately tried as catalysts for the Carroll rearrangement but none of these two catalytic combinations afforded any reaction in the standard conditions (THF, 60 °C …).

N N

Recently our group reported the synthesis, resolution and use of TRISPHAT-N (TTN), a hexacoordinated phosphate anion able to coordinate to metal centers though the Lewis basicity of the pyridine’s N-atom.¹⁴ When [CpRu(MeCN)3][PF6] was treated with 1 equivalent of L18f and 1 equivalent of enantiopure [n-Bu4N][∆-TTN], a red-orange complex was obtained after column chromatography (SiO2, CH2Cl2) in 76 % yield. Analysis of ¹H and

31P-NMR spectra showed the formation of two sets of signals and hence the presence of diastereomeric [CpRu(L18f)(∆-TTN)] complexes in a 63:37 ratio (Scheme 4-11).

Scheme 4-11: Hα signals of [CpRu(L18f)(TTN)] complexes (¹H-NMR, CDCl3, 500 MHz, with integrations); (a) Weak NOESY between HCp an H5, (b) Strong NOESY between HCp an H5 (TTN anion omitted for clarity).

Some minor signals were also noticed and compound [CpRu(L18f)(rac-TTN)] was synthesized to confirm the nature of these minor signals. Indeed, this latter complex consists of four different diastereomers ((R_Ru,∆TTN), (S_Ru,∆ TTN), (R_Ru,Λ TTN) and (S_Ru,Λ TTN) whose ¹H and ³¹P-NMR signals overlap perfectly with the major and minor signals obtained for the ∆ -TTN as starting material. Thus confirming that the minor signals account for the presence of some Λ-TTN in the analyzed sample. Furthermore, the absolute configurations of all stereogenic elements of the four different diastereomers could be deduced by NOESY experiments and the cross-peak between the signal of Cp protons (HCp) and the ones for the

14 (a) Constant, S.; Frantz, R.; Müller, J.; Bernardinelli, G.; Lacour, J. Organometallics 2007, 26, 2141-2143. (b) Constant, S.; Tortoioli, S.; Muller, J.; Linder, D.; Buron, F.; Lacour, J. Angew. Chem. Int. Ed. 2007, 46, 8979-8982.

rac-TTN

(a)

(a) S_Ru

(b) R_Ru

Weak NOESY

Strong NOESY

∆-TTN (+ Λ-TTN minor)

two diastereotopic protons α to the N atom of the oxazoline (Scheme 4-11). Unfortunately these TRISPHAT-N based CpRu-complexes did not display sufficient catalytic activity in the Carroll rearrangement and were not probed further.

The diastereomeric ratios obtained for complexes 7a to 7d are surprisingly low (~ 2:1) when compared to the enantiomeric ratios of the product from the Carroll rearrangement (~ 9:1).

However, these results do not contradict the proposed rational since the formation of the olefin complex is most probably reversible, thus providing a Curtin-Hammett situation: one of the two diastereomeric forms of the olefin Ru-complex undergoes faster oxidative addition than the other.

4.7.2. Rationalization of the enantioselectivity

The stereochemical analysis of the outcome of the reaction is also dependent of the configurational stability of the π-allyl complex. This issue was assessed by two series of experiments: one involving enantiopure secondary cinnamyl acetoacetates (Table 4-1) as substrate, and the other cinnamyl esters that cannot decarboxylate under standard reaction conditions (Table 4-2).

First, the rearrangement of the secondary acetoacetate 5c for which no trivial matched / mismatched effect could be observed using L3 as ligand was revisited.¹⁵ In the case of L18f, a clear matched / mismatched situation was observed as (–,R)-2c and (+,S)-2c were obtained with ee values of 45 and 87 % starting from enantiopure (S)-5c and (R)-5c respectively (Table 4-1). In addition, in the mismatched series, the reaction was slower and the b:l ratio was noticeably lower. When racemic 5c was submitted to the reaction conditions, a slightly enantioenriched product was obtained at 86 % conversion (ee 15 % in favor of the (–)-(R)-enantiomer) which is in good accordance with the results obtained in both enantiopure series and the fact that the reaction is globally stereospecific with branched secondary allylic substrates.^16,17 Interestingly, the b:l ratio is lower in the mismatched series; the different regioselectivity show that the reaction in the matched and the mismatched series proceed through different diastereomeric transition states which are not interconverting under the standard reaction conditions.

15 Constant, S.; Tortoioli, S.; Muller, J.; Lacour, J. Angew. Chem. Int. Ed. 2007, 46, 2082-2085.

16 Trost, B. M.; Fraisse, P. L.; Ball, Z. T. Angew. Chem. Int. Ed. 2002, 41, 1059-1061.

17 Burger, E. C.; Tunge, J. A. Chem. Commun. 2005, 2835-2837.

To account for the non-complete conservation of the enantiomeric excess, Tunge suggested, in his study with Cp*Ru(bpy) catalyst, that the branched esters 5 could isomerize into their linear regioisomer 1 in the presence of the metallic complex, thus resulting in a loss of the initial stereochemical information;¹⁷ but we were not able to verify this explanation. Globally, these results are in good accordance with the above-detailed rational but do not allow drawing any definitive conclusion concerning the conformational stability of the alleged transient π-allyl complex.

4b(10 mol%) L18f(10 mol%)

+

5c 2c 3c

O O

O O O

Entry Ester Time Conv. % ee Conf. ^b b:l ^c 1 (R)-5c 2.5 h > 97 % 87 (–)-R 95: 5

2 (S)-5c 3 h 85 % 45 (+)-S 90:10

3 rac-5c 3 h 86 % 23 (–)-R 88:12

a 4b (10 mol%), ligand L18f (10 mol%), THF, 60 °C, c(1) 2 M; the results being the average of at least two runs; ^b sign of optical rotation of 2 and absolute configuration when known; ^c ratios of branched (2) to linear (3) products were determined ¹H NMR (400 MHz).

Table 4-1: CpRu catalyzed rearrangement of allylic esters 5c.^a

To verify the validity of Tunge’s isomerization hypothesis, several branched or linear cinnamyl esters (8, rac-9 and 10) were synthesized from the corresponding cinnamyl alcohols and acyl chlorides and submitted to the reaction conditions. These esters were chosen because of their ability to get ionized and their inability to decarboxylate. Indeed, an isomerization process may occur in the presence of the Ru-catalyst upon ionization of the cinnamic ester and recombination of the two generated fragments.

In all cases, in the absence of Ru-catalyst (Table 4-2, entries 1 to 4), no reaction was observed. Linear esters 8a (R = Me) and 8b (R = t-Bu) were not transformed in any visible way when submitted to the standard Ru-catalysis conditions (Table 4-2, entries 5 and 6).

However, starting from branched 9a and 9b the corresponding linear products (8a and 8b respectively) were selectively obtained (Table 4-2, entries 7 and 8). These results show that, upon treatment with the Ru-catalyst, the linear esters can, as hypothesized by Tunge,¹⁷ isomerize to the thermodynamically more stable linear isomers. In addition when 10a (Z/E >

99:1), the (Z)-isomer of 8a, was submitted to the reaction conditions: the Z/E ratio progressively diminishes to reach 80:20 after 19 h and 58:42 after 48 h. This shows that the isomerization of the Ru-π-allyl complex is possible but slow compared to the attack of the

isomerization reaction, it seems reasonable to suppose that the effect of the isomerization of the π-allyl complex during the reaction is negligible on the stereochemical outcome of the reaction. In order to more accurately compare these latter experiments with the Carroll rearrangement, two double crossover experiments were conducted starting from approximately 1:1 mixtures of linear 8a/11b and branched 9a/12b (Scheme 4-12). In both cases, GC-MS analysis of the resulting crude reaction mixture showed, after 19 hours, the formation of all possible linear products compatible with intermolecular processes. This result indicates that the rearrangement reaction clearly proceeds through a state where the in-situ generated nucleophilic and electrophilic fragments are separated in solution, resulting in a cross-over. average of at least two runs; ^b regiosisomeric purity of linear 6 (b:l < 1 : 99) and branched 7 (b:l > 99 : 1) starting materials were determined by GC-MS ^c ratios of branched (6) to linear (7) products were determined by GC-MS.

Table 4-2: CpRu catalyzed rearrangement of allylic esters 8 to 10.^a

O

Scheme 4-12: Double crossover experiments (ratios of products determined by GC-MS).

With these results (page 55 to 56) confirming the global conformational stability of the Ru-π-allyl complexes, it is interesting to reconsider the results obtained for the Z and E-cinnamyl acetoacetate reported in Table 3-5 (entries 1 and 7, 1a and 1g respectively). The lower ee obtained for the Z isomer should then not be attributed to an isomerization of the resulting anti CpRu-π-allyl complex but can be explained by a lower control of the endo/exo isomerism during the formation of the π-allyl complex in the case of an anti-allyl moiety (Scheme 4-13).

Indeed the energy difference between the endo,anti and the (exo,anti)-CpRu-π-allyl complexes is lower due to destabilizing steric interactions present in both cases (Cp-Ph for the endo,anti and Cp-CH for the exo,anti). The fact that the same (–)-(R)-2c enantiomer was predominantly obtained is both cases tends to show that the (exo,anti)-CpRu-π-allyl complex is slightly more stable. This explanation is however based on the (reasonable) hypothesis that the orientation of the ligand and the allyl fragment remains the same in both cases.

Ru

Finally, possible non-linear effects were assessed. A series of reactions with substrate 1c were performed with different mixtures of the two enantiomers of ligand L18f. Results are summarized in Table 4-3 and show that the enantioselectivity of the reaction is directly proportional to the enantiomeric purity of the ligand L18f. This indicates that, the metallic complex involved in the enantiodetermining step is probably a monomeric species composed of one single ligand.¹⁸

O L18f+ent -L18f^{4b(2 mol%)}(2.4 mol%) THF, 60 °C, 22 h

O O O

1c 2c

O

3c

Entry ee of L18f ^b b:l ^c % ee ^b,d 1 > (+)-99 % 95 : 5 (–)-76 % [(–)-76 %]

2 (+)-67 % 95 : 5 (–)-47 % [(–)-50 %]

3 (–)-44 % 95 : 5 (+)-35 % [(+)-34 %]

4 > (–)-99 % 95 : 5 (+)-76 % [(+)-76 %]

a 4b (2 mol%), ligand L18f + ent-L18f (2.4 mol%), THF, 60 °C, c(1c) 1 M; the results being the average of at least two runs; ^b sign of αD given; ^c ratios of branched (2c) to linear (3c) products were determined by GC-MS; ^d ee measured by CSP-GC, sign of αD given and calculated ee given in brackets.

Table 4-3: CpRu catalyzed rearrangement of allylic esters 1c; effect of ligand enantiopurity.^a

Overall these experimental data are in line with the proposition that the CpRu-catalyzed Carroll rearrangement is proceeding through a Ru-π-allyl complex obtained from the oxidative addition of an unsaturated Ru-complex onto an allylic ketoester; this Ru-π-allyl complex having, in the reaction conditions, a relatively high configurational stability.

However the transient allylic species could not be directly observed by spectrometric or spectroscopic methods.