Transfer hydrogenation catalysis

Formic acid was utilised as a sacrificial source of hydride. Unlike reversible hydride donation by alcohols such as isopropanol,1, 36 formic acid provides an effectively irreversible solution due to the evolution of carbon dioxide.37 Treatment of 1 with formic acid allowed for the observation of an osmium hydride intermediate by 1H- NMR, observed as two singlets (-5.89 and -6.04 ppm) corresponding to the two hydride complex diastereomers (Figure 3.11). 187Os (I = ½, 1.6% natural abundance) satellites were also observed, with a coupling constant (1_J(187_Os,1_{H) = 44 Hz) similar}

to other reported osmium hydrides (14.4-38.1 Hz).38 Over 30 min, the ratio between the hydride resonances decreased, from 3:1 to 1.2:1. Though both hydrides form, enantioselectivity of ketone reduction likely results from either (a) greater rate of reactivity of one hydride isomer, as suggested for the analogous ruthenium hydride;39,

40 _{or (b) thermodynamically favourable interaction between a specific hydride}

enantiomer and the ketone.23

The traditionally-accepted mechanism of enantioselectivity by Noyori-type catalysts involves a combination of steric effects (involving the phenyl groups at the chiral centres of the diamine, Figure 3.11) and electrostatic interaction between the catalyst arene and the aromatic group of the substrate, occurring via a concerted outer-sphere transition state.2, 41 However, recent studies suggest that a step-wise mechanism involving the solvent may more accurately describe face-specific hydrogen transfer.

Catalytic activity of osmium sulfonamide complexes

Tosyl-DPEN complexes (2, 7, 8-10) achieved remarkably high enantioselectivity and conversions (> 99% ee).23 _{Similarly to the parent ruthenium analogue 11, (R,R)-}

configured catalysts afforded the corresponding (R)-alcohol, and (S) catalysts afforded the (S)-alcohol, respectively.19, 37 No statistically-significant difference in activity was observed between enantiomers of 2 or 7. Sulfonamide substitution (3-6) led to a reduction in enantioselectivity, however the complexes still achieved high conversion in 24 h. Reported studies have suggested that the disfavoured transition state may be

destabilised by an unfavourable interaction with the sulfonyl of the ligand.19, 42, 43

The addition of methoxy-substituents on complexes 9-10 did not have an effect on the enantioselectivity of the catalyst.

All osmium catalysts investigated displayed hydrogenation rates (TOF) greater than that of parent Ru catalyst 9 under identical conditions (23 ± 1 h-1), even after allowing the ruthenium active catalyst to form in situ. The highest turnover frequency was achieved by the biphenyl catalyst 7 (78.2 ± 0.4 h-1). Both sulfonamide substitution and methoxy substitution were found to reduce significantly the TOF relative to the tosyl compound 2 (63.9 ± 0.3). The fluorine atom of complex 5 may have led to a hindrance of reaction rate by forming a hydrogen bond with another catalyst molecule (Figure 3.12) however this is difficult to prove experimentally.

Dichlorido osmium complex 1 is also able to act as a catalyst for ATH reactions (Table 3.4). Upon treatment with a base, osmium complex 2 forms in situ.23 The high stability of osmium catalyst 2 allows for it to be prepared and used directly, and hence the catalysed reaction occurs immediately without a lag phase. This is unlike the reaction profile of Ru complex 11, which requires chloride dissociation before catalysis is initiated (Figure 3.13). Whilst the ruthenium 16-electron active catalyst has been isolated and fully characterised, it is not sufficiently stable to warrant use over the Ru(II) chlorido pre-catalyst.

Reduction of acetophenone-derived ketones using complex 1 exhibited similar behaviour for Ru and Os catalysts. In both cases, conversion is decreased for ketones bearing electron-donating substituents (e.g. MeO) while ketones containing electron withdrawing groups maintain high conversion but exhibit decreased enantiomeric excess. The resulting enantioselectivity therefore results from a combination of both steric and electronic effects in the substrate.

Figure 3.13. Ru catalyst 11 (●) must first dissociate chloride, observed as a lag-phase in the first hour

of reduction. Os catalyst 2 (▲) exhibits instantaneous rate linearity.

0 20 40 60 80 0 1 2 3 T O Nt Time / h 2 [Os(p-cymene)(TsDPEN)] 11 [RuCl(p-cymene)(TsDPEN)]

Based on experimental data, the postulated mechanism of transfer hydrogenation is summarised in Figure 3.14. Pre-catalyst 1 eliminates chloride ligands to form active catalyst 2. Hydride is then transferred to 2 from formic acid, yielding an intermediate hydride complex as a mixture of diastereomers. Favourable CH-π interactions between a prochiral ketone substrate and the catalyst orientate the transition state to favour reduction of the carbonyl on a specific face, controlled by steric interactions with the chiral diamine of the catalyst. After transfer, the 16-electron catalyst is regenerated.

3.4.4 Density Functional Theory (DFT)

To better understand and compare the catalytic performance of the osmium complexes, we looked to theoretical methods. In particular, Density Functional Theory (DFT) is a valuable tool for an N-body system, where solving the Schrödinger equation would be computationally impossible. Instead, DFT provides an approximation of the solution, using the electron density. Mixed basis sets (Lanl2DZ/6-31+G**) are frequently used to simulate organometallic complexes.44 The metal centre was approximated by using Los Alamos National Laboratory basis set with an effective core potential (ECP; Lanl2DZ)29 and all other atoms by the Pople basis set, (6-31+G**).

DFT calculations for complexes 2-10 show high similarity between all systems, in agreement with high conversions and enantioselectivities observed in catalytic studies. The more positive electrostatic potentials of p-cymene and biphenyl (compared to the phenyl groups) are due to the donation of electrons from aromatic π orbitals to form σ and π-bonds with metal-based orbitals. The metal is also capable of δ-backbonding interactions with aromatic molecular orbitals. Activating groups such as OMe in complexes 9 and 10 give rise to localised charge effects, with some donation of electron density into the corresponding aromatic rings. Little effect, however, is observed at the metal centre. This is mostly likely due to the distance of the substitution site from the inner coordination sphere. This could explain why 9-10

retain the high activity and enantioselectivity of parent compounds 2 and 7, respectively. The low TOF for acetophenone reduction using catalyst 5 may be due to the presence of intermolecular hydrogen bonding via fluorine, observed as a charge localisation at the point of substitution. However, this is difficult to prove experimentally as well as computationally and goes beyond the scope of this thesis.