Raffael
Huber
,
Alessandro
Passera
and
Antonio
Mezzetti
*
Department of Chemistry and Applied Biosciences, ETH Zürich, Wolfgang-Pauli-Str. 10, CH-8093 Zürich, Switzerland. E-mail: mezzetti@inorg.chem.ethz.ch
First published on 2nd July 2019
PNP pincer ligands stabilise diamagnetic base metal catalysts, and much effort has been invested in the development of chiral analogues for asymmetric catalysis. Starting from the conformational issues that affect P-stereogenic diphosphines, we extend the analysis to our recently prepared P-stereogenic PNP pincer ligands, which we used in the iron(II)-catalysed hydrogenation of ketones. Backbone rigidity and size (and shape) of the substituents at phosphorus are pivotal in both bi- and tridentate P-stereogenic ligands, and their interplay is discussed, as well as the contribution (and shortcomings) of DFT calculations to the understanding of the conformational flexibility and enantiodiscrimination.
The goal of this article is to put the conformational issues encountered during our study in a broader context that encompasses P-stereogenic diphosphines and (the few) PNP/PCP pincer catalysts of other metals, both precious and non-noble.
The first iron(II) catalysts for the asymmetric hydrogenation of polar double bonds have been designed by analogy with their ruthenium(II) analogues.7 Thus, the first asymmetric iron(II) catalyst for the hydrogenation of ketones, reported by Morris in 2008,8 was based on the tetradentate, open-chain PNNP ligand already used with ruthenium and exploited in hydrogen transfer from 2-propanol.7b It should be noted, however, that the analogy between iron(II) and ruthenium(II) is deceiving, as iron(II) complexes are considerably less stable than their heavier counterparts. However, the combination of good σ donors, such as phosphines, with strong π-accepting ancillary ligands, e.g. CO, stabilises the low spin d6 configuration in which all M–L σ* antibonding orbitals are empty. Hence, ligand lability can be greatly reduced, and robust complexes are obtained.
For instance, Morris’ second-generation PNNP catalyst ligand is stable, but yet active and achieves turnover frequencies of up to 200 s−1.9 With the same goal, our group has developed macrocyclic N2P2 ligands that, combined with isonitriles as ancillary ligands, form robust, highly enantioselective catalysts for the transfer hydrogenation of a broad scope of ketones, including a hydride complex that catalyses the base-free reduction of benzils.10
As already observed for ruthenium(II),11 the complexes of tetradentate N2P2 ligands perform well in transfer hydrogenation, but are poor catalysts for the hydrogenation with H2,12 which is preferable for industrial application due to its irreversibility and atom economy.
As an alternative to the pyridine backbone, Beller21 and Guan22 have used a N,N-diethyleneamine bridge for PN(H)P pincer ligands whose iron borohydridocarbonylhydride complex catalyses the hydrogenation of esters (Scheme 1). The secondary amine in the backbone directs and activates the carbonyl group towards hydride attack in the bifunctional mechanism.23 Jones extended the use of this catalyst to the acceptorless dehydrogenation and hydrogenation of N-heterocycles,24 and Hazari and Schneider reported a five-coordinate iron carbonylhydride that catalyses the acceptorless dehydrogenation of alcohols as well as ketone hydrogenation.25
All PNP pincer complexes described so far were achiral. In 2014, Morris reported the first iron(II) catalyst for the asymmetric pressure hydrogenation of ketones, a PNP bromodicarbonyl complex with an enantiopure 1-methyl-2-phenylethylene chelate that hydrogenates acetophenone with 80% ee (Scheme 2).26
A second-generation chlorocarbonylhydride complex reduces acetophenone with 95% ee.27 This catalyst also reduces the heterocyclic 2-acetylfuran with full conversion and 95% ee. The more acidic 3,5-bis(trifluoromethyl)acetophenone and also the bulkier cyclohexyl phenyl ketone are not tolerated, however, and lead to low conversion (8% and 38%, respectively). Zirakzadeh introduced ferrocenyl as an additional stereogenic element, and the corresponding bromocarbonylhydride complex is an active catalyst for the hydrogenation of arylalkyl ketones (81% ee for acetophenone, Scheme 2).28
The catalysts shown in Scheme 2 all contain C-stereogenic ligands and have achieved notable enantioselectivity in selected instances, but their substrate scope is still narrow. In the quest for improvement, we have developed iron(II) catalysts with P-stereogenic pincer ligands and either pyridine (PNP) or N,N-diethyleneamine (PN(H)P) backbones. For this study, we chose dialkylphosphino units by analogy with the highly successful achiral iron(II) catalysts with pincer ligands.
An inherent difficulty of pyridine-based pincers (such as in Scheme 1 top) is that the benzylic position of the backbone is most likely to undergo deprotonation during catalysis, and a stereocentre in this position would be quickly epimerised. Hence, P-based stereocentres seemed particularly suitable. The lessons we learnt from these new PNP pincer ligands reminded us of the conformational issues that are also found in P-stereogenic diphosphines, which are recapitulated in the next section, starting with a detour to ligands with carbon based stereocentres.
In turn, the conformation of the chelate determines the orientation of the phenyl rings, which transmit the chiral information to the metal-bound substrate, leading to enantioselection. In Prophos and Chiraphos, an analysis of several crystal structures indicates that the axial phenyl is preferentially edge exposed and hence interacts with the substrate more strongly than the equatorial phenyl, which tends to be face exposed (Scheme 4 top).32 However, there is a certain degree of conformational flexibility, and three types of conformations have been observed for diphosphines that form five-membered chelates such as Chiraphos (Scheme 4).32
In the half-chair conformation (A), the substituents on phosphorus are clearly distinguishable as equatorial and axial and are arranged in an approximately C2-symmetric fashion. In structures close to an envelope conformation (B), which have a small P–M–P–Cring torsion angle α, the substituents on the in-plane P atom are best described as inclinal (Rinc in Scheme 4). In further distorted envelope conformations (C), the sign of the torsion angle α is inverted, and the axial and equatorial positions of the phosphorus substituents are switched with respect to A. A correlation between axial/equatorial and edge/face-exposed is possible only in half-chair structure A or in envelope structure C with a large torsion angle α (P–M–P–Cring).32
The mechanistic investigations of enamide hydrogenation have revealed intricacies that are still a matter of debate,34,35 but have also showed that the conformation of the chelate ring is significantly more flexible in DiPAMP than in Chiraphos.36 Also, one should keep in mind that the roles of the backbone and of the phosphorus substituents are reversed in P-stereogenic diphosphines as compared to C-stereogenic ligands with PPh2 donors such as Chiraphos. Indeed, in P-stereogenic ligands, the substituents at phosphorus play a dual role as they generate the chiral environment for the approach of the substrate and lock the conformation of the chelate ring according to their size. This can lead to some confusion. Thus, in DiPAMP, the o-anisyl group acts as the “large” substituent in the chelate and hence occupies its equatorial position, but is “small” with respect to the substrate, as it is face exposed (Scheme 6). In contrast, the smaller phenyl group is axial, but edge exposed toward the substrate, and thus acts as the bulky group toward the metal-bound substrate.35 With the so-called quadrant rule, Knowles has attempted to correlate empirically the spatial arrangement of the substituents on P in the catalyst precursor (usually a diolefin complex) with the absolute configuration of the protected amino acid obtained from enamide hydrogenation.37
However, the overall crowding in the complex affects the subtle interplay between the conformation of the chelate and the orientation of the substituents on phosphorus. Thus, the same P-stereogenic phosphine assumes opposite conformations in its norbornadiene and 1,5-cyclooctadiene complexes, but gives the same sense of induction in the hydrogenation of dehydro amino acids.38
Moreover, polysubstituted aryl substituents on phosphorus favour envelope conformations such as B and C in Scheme 4, which is beneficial for enantioselectivity, but hinders the development of simple rules-of-thumb.39 The conformational issue is acute also when the size of the P-substituents is similar, as in the case of (S,S)-tBu(Ph)PCH2CH2P(tBu)Ph,40 whose complexes show a flattened conformation of the chelate ring with no clear axial/equatorial distinction of the tert-butyl and phenyl groups. Accordingly, the tBu-substituted ligand gives the lowest enantioselectivity in the series R(Ph)PCH2CH2P(R)Ph (R = tBu, o-anisyl, 1-naphthyl).40
Therefore, the substituents on stereogenic P atoms should be significantly different in size. The P(tBu)Me unit meets this requirement, as the tert-butyl group has cylindrical symmetry with respect to P–tBu rotation and is always larger as compared to the methyl group. Imamoto's ethane-1,2-diyl-bridged, P-stereogenic diphosphines BisP* with alkylmethyl- or arylmethylphosphine donors (Scheme 7) are a privileged class of ligands and provide excellent enantioselectivity with a wide variety of catalytic reactions.41
The X-ray structures and DFT calculations of tBu–BisP* complexes exclusively show conformations with the bulky group in the equatorial position,35 but the enantioselectivity of BisP* rhodium catalysts strongly depends on the structure of the substrate.42 Therefore, more rigid backbones were introduced to improve the control of the conformation of the chelate during catalysis. In a first approach, Imamoto used a shorter chelate backbone for ligands of the general formula Me(R)P–CH2–P(Me)R (Scheme 8).43 The four-membered chelate ring formed upon complexation is conformationally rigid and almost perfectly planar.42 MiniPHOS rhodium catalysts outperform BisP* ligands in the hydrogenation of β,β-disubstituted dehydro amino acids (Scheme 8).42
However, MiniPHOS gives lower enantioselectivity than BisP* in the rhodium-catalysed hydrogenation of unsubstituted enamides,44 possibly because its smaller bite angle (ca. 72°) reduces the overall steric bulk. In the C1-symmetric modified MiniPHOS-type ligand tBu2PCH2P(CH3)tBu (Trichickenfootphos, Scheme 7), the large achiral PtBu2 donor apparently compensates for the effect of the small chelate and gives excellent enantioselectivity with dehydro amino acid derivatives.45 The NH-bridged MaxPHOS (RR′P*–NH–P*RR′) is also a MiniPHOS derivative.46
An alternative strategy to rigidify P-stereogenic diphosphines with a five-membered chelate, such as BisP*, is the use of a 1,2-phenylene linker instead of ethane-1,2-diyl. With ligands bearing pseudo-chirality47 on phosphorus such as DuPHOS and BPE, this approach has improved the enantioselectivity of the Rh-catalysed asymmetric hydrogenation of α-acetamidoacrylates (Scheme 9).48
Scheme 9 Effect of the linker rigidity on the asymmetric hydrogenation with non-P-stereogenic diphosphines. |
On this track, Imamoto prepared P-stereogenic ligands with a 1,2-phenylene backbone such as BenzP*49 and QuinoxP (Scheme 7).50 In the hydrogenation of (Z)-methyl 3-acetamidobut-2-enoate, the enantioselectivity dramatically increased from 19.7% ee with tBu–BisP*41c to 97.6% ee with BenzP* (Scheme 10).49b BenzP* and QuinoxP* have found broad application in asymmetric catalysis, such as Rh-catalysed asymmetric hydrogenation51 and hydroacylation,52 in Cu-catalysed arylation53 and borylation54 reactions, Pd-catalysed allylic alkylations,55 and Fe-catalysed cross coupling.56
In these molecules, the entropic preference for the axial conformation is larger than in cyclohexanes, and the corresponding ΔH and TΔS values are similar. Hence, the equilibrium constant K for the equatorial/axial equilibrium depends more strongly on temperature. Also, the axial isomer shows weaker 1,3-diaxial interactions because the P–C bonds are longer than C–C. Hence, for 1-methyl, 1-ethyl, and 1-phenyl phosphinanes, the axial conformer is in excess at room temperature.
Back to diphosphines, we have seen above that the most successful P-stereogenic ligands form small chelate rings upon coordination (five- or four-membered, Scheme 8). Six-membered chelate rings have been used with some success in combination with P-stereogenic diarylphosphine donors,58 whereas dialkylphosphine donors have been found to give low enantioselectivity.59 This is unsurprising, as six-membered chelate rings are intrinsically more flexible than smaller ones. The typical example is 2,4-bis(diphenylphosphino)pentane (bdpp), which forms, besides the most stable twist conformation with all-equatorial methyl groups, an energetically accessible achiral half-chair in which the axial substituents on phosphorus assume an achiral conformation (Scheme 12).60
Scheme 12 Conformations of (2S,4S)-[RhL2(bdpp)]+ (only one of the two equivalent half chair conformations is shown). |
To rigidify P-stereogenic diphosphines with six-membered chelate rings, gem-disubstitution on the central atom of the backbone was attempted with success with diarylphosphine donors, such as in Me2Si(CH2P(1-Np)Ph)2,61 but failed with dialkylphosphines.59 Also, as seen above for five-membered rings, the use of aromatic backbones is beneficial for enantioselectivity. Thus, P-stereogenic diphosphines featuring larger chelates take advantage of rigid xanthene structures.62 With the aim of increasing the bite angle of the P-stereogenic diphosphine, ferrocene-1,1′-diyl also has been used as the backbone, and the resulting ligands give high enantioselectivity in the rhodium-catalysed hydrogenation of acetamido cinnamates and other functionalized alkenes,63 but the rotation of the cyclopentadienyl rings of the ferrocene bridge allows for several conformations of the ligand, as observed by 31P NMR spectroscopy in solution.63b Interestingly, a recent application of a P(tBu)Me/oxazoline ligand in the iridium-catalysed hydrogenation of N-alkyl imines suggests that P-stereogenic phosphines also give stiff conformations in the presence of an external rigidifying element, in this case, a cyclometalated benzimine.64
Overall, the above discussion shows that the conformational issues of P-stereogenic diphosphines play a pivotal role in enantiodiscrimination. Before extending it to tridentate ligands, the synthetic protocols for P-stereogenic building blocks are recapitulated in the next section.
Also bulky alkyl lithium reagents can be used,39 but supermesityllithium (2,4,6-tri-tert-butylphenyl lithium) or 2-adamantyllithium fails to give the desired product.69 Organomagnesium compounds react sluggishly and require heating, which decreases the enantiospecificity.65b Usually, the aminophosphine boranes are obtained as single diastereoisomers after a simple recrystallization step. The aminophosphine is then methanolised under acidic conditions into a methyl phosphinite borane.65 Alternatively, it can be treated with anhydrous HCl in an aprotic solvent such as toluene to give the more reactive chlorophosphine borane.70 Both compound classes react as electrophiles, and after an SN2@P reaction with an organolithium reagent, tertiary phosphine boranes are obtained (Scheme 14). Alkyl- as well as aryllithium reagents react with high enantiospecificity.65
One advantage of the electrophilic chlorophosphine boranes is that they undergo umpolung to nucleophilic secondary lithium phosphide boranes feasible by transmetalation with tBuLi at −85 °C (Scheme 15).65c
Deboronation of the final product is achieved either by reaction with an excess of amine71 or by treatment with strong acids such as HBF4 etherate.72 The latter method is mandatory for trialkylphosphines due to their electron-rich nature. The Jugé–Stephan method is generally applicable when at least one of the substituents in the target (tertiary) phosphine is phenyl. A drawback of the method is the use of ephedrine, which is nowadays a restricted chemical due to its use in illicit drug manufacturing.
Imamoto exploited Evans’ enantioselective deprotonation to prepare enantiomerically pure secondary phosphine boranes.67 To this goal, the intermediate organolithium derivative was oxidised with O2 to the hydroxymethylphosphine borane with retention of configuration. Ruthenium-catalysed oxidation to the carboxylic acid and spontaneous decarboxylation under basic reaction conditions give secondary phosphine boranes with formal inversion (Scheme 17).
The secondary phosphine boranes are stable towards air and moisture, and are valuable precursors for the synthesis of a broad variety of ligands as described below.73 A drawback is the stoichiometric amount of sparteine, whose (−)-enantiomer has experienced a global shortage since the early 2010s.74 Nowadays, both enantiomers of the alkaloid are commercially available, at prices typical for fine chemicals.
However, efforts directed towards preparing enantiomerically pure secondary phosphine oxides – which, after deprotonation, offer a versatile synthon that reacts with a broad range of electrophiles – remained limited,77 possibly due to the belief that the deprotonated secondary phosphine oxide anion is inherently stereolabile.78
This prejudice was overturned by Han. In 2008, he reported that the rapid epimerization of (−)-menthyl (RP)-phenylphosphinate and its substitution products under alkaline conditions is not due to the stereolability of the secondary phosphine oxide anion, but originates from the reaction of the phosphinate ester with a metal alkoxide or hydroxide.68 Thus, by carefully ensuring that no hydroxide or alkoxide is formed under the reaction conditions (e.g. with a menthol-free starting material), he prepared a variety of secondary and tertiary phosphine oxides with excellent enantiospecificity (Scheme 19).
The method has been applied, for example, in the preparation of organocatalysts for the Morita–Baylis–Hillman reaction79 and of P-stereogenic macrocyclic N2P2 ligands.80 It is most successful when one of the substituents in the target (tertiary) phosphine is an aryl group, and is thus often an alternative to the Jugé–Stephan method. Also, menthol is cheap and readily available, and the reaction is easily scaled up. One drawback is the use of a phosphine oxide as the protecting group, though. Although a wealth of literature exists on the (stereospecific) reduction of secondary and tertiary phosphine oxides,81 the reduction conditions need to be tuned to the specific system and are less general than for the deprotection of phosphine boranes.
Imamoto prepared the P(tBu)Me analogue and used it in a variety of enantioselective reactions.85 These encompass Pd-catalysed Michael additions (1 out of 19 substrates gave >90% ee)85b and intramolecular hydroamination (up to 47% ee),85e Ir-promoted ketone hydrogenation (17% ee),85c and aza-Michael addition with a Ni catalyst (up to 46% ee).85d
Enantiomerically pure tert-butylmethylphosphine borane (3b) was prepared by Imamoto's procedure (Scheme 17). As the iron(II) complexes of the P(tBu)Me based pincer 1b turned out to be unstable (see below), we also targeted its P(Cy)Me analogue 3a. Imamoto's method gives (R)-cyclohexyl(hydroxymethyl)methylphosphine borane (5a) with 75% ee only, though.67 Conversion of 5a to the crystalline p-phenylbenzoyl ester and crystallisation gave 7a as a single enantiomer. Thus, enantiomerically pure cyclohexylmethylphosphine borane (R)-3a was obtained after saponification and Ru-catalysed oxidation (Scheme 22).88
Finally, DFT calculations (see below) prompted us to prepare the P(Me)Ph-based pincers 1c and 2c. In this case, Imamoto's approach in Scheme 17 is not suitable, as methylphenylphosphine borane cannot be prepared as in Scheme 23 because the intermediate of type 6 undergoes overoxidation to the corresponding phosphinite borane P(Me)(Ph)(OH)·BH3.67 Also, methylphenylphosphide borane has been found to be configurationally unstable.91 Therefore, we developed an alternative approach inspired by Buono's masked secondary phosphine boranes (4);92 we treated diastereomerically pure (RP)-(−)-menthyl-H-phenylphosphinate (prepared on a multi-gram scale from (−)-menthol and PhPCl2)93 with MeLi, quenched the lithium salt of the methylphenylphosphine oxide with acetaldehyde, and reduced (SP)-(1-hydroxyethyl)methylphenylphosphine oxide 8 with BH3·SMe2 to (SP)-(1-hydroxyethyl)methylphenylphosphine borane (4) with formal retention (Scheme 23).94 This route is easily scalable, and (SP)-4 was routinely prepared on a >5 g scale.89
Scheme 23 Preparation of masked secondary phosphine borane (SP)-4 from (RP)-menthyl phenylphosphinate. |
The X-ray structure of 10a (Fig. 1) reveals that the pyridine-pincer backbone of 1a is tilted to enable the envelope conformation for the two chelate rings.88
Fig. 1 ORTEP drawing of 10a (reprinted with permission from ref. 88. Copyright 2018 American Chemical Society). |
Depending on the stereocentre at phosphorus, the tilt is either λ (as for 10a with (S,S)-1a) or δ (observed for 10c with (R,R)-1c), and the large substituent on phosphorus occupies a pseudo-equatorial position.
All hydrides are extremely sensitive towards air and moisture, and the P(tBu)Me derivative 11b was not isolable in reasonable purity. As hydrides 11b–11c decompose in solution, they were always freshly prepared for catalytic experiments. A putative decomposition product (among unidentified brown precipitates) is an iron(0) dicarbonyl complex, as observed for Fe complexes of the achiral 2,6-bis(diisopropylphosphinomethyl)pyridine pincer.16 As for dibromocarbonyl complexes, bulkier phosphines (e.g. tert-butyl substituted ones) form less stable complexes.
Due to its potential role in catalysis (see below), we also prepared dihydride 12a (Scheme 27). The major species contains trans hydrides, but the other components of the reaction mixture were not characterised. The dihydride is highly sensitive towards air and moisture and decomposes in solution.
Addition of a stoichiometric amount of acetophenone to 12a did not change the major species present in a C6D6 solution as observed by NMR spectroscopy, and only trace amounts (<2%) of phenylethanol were formed (which might originate from impurities of 12a). A similar observation has been made for an achiral Fe/PNP dihydride.96 This indicates that the dihydride 12a is not involved in the catalytic cycle of ketone hydrogenation with such complexes.
Eventually, Milstein put forward the outer-sphere mechanism O, which involves the iron(0) species 15 whose chiral analogue is shown in Scheme 30.97 The base dehydrobrominates precatalyst 11a to give a dearomatised intermediate that reductively eliminates H+. Rearomatisation and EtOH coordination give the iron(0) species 15, which then transfers a benzylic H atom from the PNP ligand as hydride to the carbonyl group of acetophenone (Scheme 30, D). A hydrogen bond involving the coordinated ethanol molecule directs the incoming substrate and activates its carbonyl group towards nucleophilic attack, which is reminiscent of the well-established bifunctional mechanism.23
We reasoned that the additional stereochemical information provided by the chiral ligand may help discriminate between these mechanisms, and hence undertook a comprehensive DFT study by modelling the enantiodetermining steps of the three mechanisms in Scheme 30.88 Notably, only the outer-sphere mechanism D predicts the experimental sense of induction and the enantioselectivity, in agreement with Milstein's proposal which is mostly based on kinetic considerations.97 Beyond contributing to assess the mechanism, the DFT study revealed the conformational issues discussed below.
Fig. 2 Conformation of the five-membered chelate ring (with Newman projection and dihedral angle θ) (reprinted with permission from ref. 88. Copyright 2018 American Chemical Society). |
The conformation of the chelate ring determines the position of the P-substituents (equatorial or axial in the envelope conformation or inclinal in the planar one)32b and the tilt of the pyridine ring with respect to the Fe/PNP plane. The inversion of the ring swaps the cyclohexyl and methyl P substituents between equatorial and axial positions and tilts the pyridine ring with respect to the P(1)–Fe–N(1) and P(2)–Fe–N(1) planes. Positive torsion angles (θ1 and θ2) correspond to equatorial cyclohexyl groups and to the λ conformation of the backbone. For negative θ values (−12° to −27°), the cyclohexyl is axial, and the backbone is in the δ conformation. The calculations indicate that the five-membered chelate ring can also assume a planar conformation (θ close to 0°) with inclinal substituents.
In the enantiodetermining step of Milstein's unprecedented outer-sphere mechanism O, a benzylic H atom of the ligand is transferred as hydride from the five-coordinate iron(0) complex [Fe(CO)(EtOH)(1a)] (15) to acetophenone (Scheme 30O). Complex 15 exists in two different conformations of the chelate ring (15a and 15b). The approaching acetophenone is directed by the incipient CO⋯Fe interaction and by the CO⋯HOEt hydrogen bond to the ethanol ligand. As acetophenone approaches the iron(0) complex 15a/15b, the conformation of the chelate ring changes to allow the attack of the benzylic H atom as hydride onto the carbonyl C atom (Fig. 3). The conformation determines from which benzylic position the hydride is transferred, and acetophenone approaches with either enantioface, for a total of four transition states, O(1)–O(4). Intriguingly, the most stable TS O(1) features axial cyclohexyl groups and hence the less stable conformation.
Fig. 3 Iron(0) complexes 15a and 15b and TSs O(1)–O(4) for hydride transfer in mechanism O (product stereochemistry in brackets). Newman projections give θ1 (left) and θ2 (right). Relative Gibbs free energies are in kcal mol−1 (reprinted with permission from ref. 88. Copyright 2018 American Chemical Society). |
The intrinsic reaction coordinate (IRC) analysis shows that O(2) and O(4) connect to 15a, which bears equatorial cyclohexyl groups, whereas O(1) and O(3) are linked to 15b, which bears axial cyclohexyl groups. Also, conformer 15a, which bears equatorial cyclohexyl substituents, converts to 15b, which is less stable by 0.4 kcal mol−1.
Conformers 15a and 15b interconvert rapidly and react under Curtin–Hammett conditions to give the 1-phenylethoxide complexes 16a/16a′ and 16b/16b′ (Fig. 4). This implies that the ratio of the products depends on the energy difference between the transition states.98
Fig. 4 Reaction pathway for H− transfer in mechanism O. Relative Gibbs free energies are in kcal mol−1 (reprinted with permission from ref. 88. Copyright 2018 American Chemical Society). |
The small changes in |θ| (Fig. 3) indicate minor structural reorganization between the Fe(0) species 15a/15b and O(1)–O(4), which further supports the retention of conformation along the reaction coordinate. Conformer 15b (with axial cyclohexyl groups) is just 0.4 kcal mol−1 less stable than 15a, which bears equatorial cyclohexyls. The major contribution to the total energy of the O(1)–O(4) transition states is the distortion of the catalyst, i.e. the conformational change of the chelate ring involved in hydride transfer. However, the energetic preference for conformations in which the large P-substituents occupy equatorial positions is small, and can be reversed as the substrate approaches. Even weak intermolecular interactions, such as C–H⋯O or C–H⋯π, swap the relative stability of the conformations of different intermediates and transition states. Therefore, it is reasonable to assume that the conformational flexibility observed for 1a might be a more general problem with P-stereogenic pincer ligands with a pyridine backbone (see below).
Overall, the enantioselectivity was modest and/or strongly temperature dependent. In the ruthenium-catalysed ketone hydrogenation of ketones, the reduction of acetophenone was much more enantioselective at −40 than at 25 °C (87 and 64% ee, respectively, Scheme 31b).84 Palladium-catalysed hydrophosphinations show similar effects of the temperature (Scheme 32a, ref. 85b gives a further example).85b,87
Conformational equilibria of the 2,6-dimethylenepyridine (PNP) or 1,3-xylyl (PCP) backbones analogous to those described above for 11a may account for these temperature effects, but a conformational analysis would be needed to verify this hypothesis. Still, our observations with the PNP pincer ligand 1a prompted us to move to PN(H)P ligands with N,N-diethyleneamine as a backbone, in which the sp3 nitrogen atom should act as a conformational lock upon coordination to a metal centre.
The phenylmethylphosphine derived pincer 17c was prepared from the masked secondary phosphine borane 4 and bis(2-iodoethyl)-N-trimethylsilylamine (analogously to 9c). The borane-protected pincer ligands 17a–17c are indefinitely stable upon storage in air at room temperature.
Overall, the PN(H)P pincer ligands 2a–2c form iron complexes that are less stable than those with the pyridine backbone (1a–1c). Hence, complexes 18a–18c are best handled under an inert atmosphere, although brief exposure to ambient conditions is tolerated. In particular, the tert-butylmethylphosphine derivative 18b is unstable and slowly decomposes both in solution and in the solid state.
The conformation of the chelate rings with the N,N-diethylamine backbone is rigidified by the sp3 nitrogen (Fig. 5). The metallacycles adopt an envelope conformation with C(2) and C(3) in the endo position, and the elongated ellipsoids of the backbone indicate the presence of an isomeric form where the N–H proton points downwards, leading to a ring flip.
Fig. 5 ORTEP drawing of 18a (reprinted with permission from ref. 90). |
The hydride ligand is located trans to bromide in the major isomer of hydride 19a and trans to CO in 19b and 19c. Hydrides 19a–19c are extremely sensitive towards air and moisture. As they decompose in solution, they were freshly prepared for each catalytic run.
The diastereoisomeric transition states TS20aS and TS20aR lead to (S)- and (R)-1-phenylethanol, respectively (Fig. 6 top). The calculated energy difference (0.5 kcal mol−1, hybrid DFT model at the B3LYP level of theory) is in fair agreement with the experimental one (ca. 0.2 kcal mol−1).
Fig. 6 Transition states TS20a/cS and TS20a/cR for the hydride transfer. Relative Gibbs free energies are in kcal mol−1. Black dashed lines depict the hydride transfer and the C–H⋯π interaction. |
We speculated that the low energy difference between TS20aS and TS20aR may result from the occurrence of a similar CH/π interaction with the phenyl group of acetophenone in both transition states which involves the Cipso–H of the cyclohexyl in TS20aS and the methyl on phosphorus in TS20aR (Fig. 6 top) (in addition to a strong N–H⋯O hydrogen bond that directs and activates acetophenone). Therefore, we tried to increase the enantioselectivity by eliminating selectively one of CH/π interactions.
To this goal, the cyclohexyl group was replaced with a phenyl ring in silico, which eliminated the C–H⋯π interaction in transition state TS20cR (Fig. 6 bottom). This gave an increased energy difference of 1.6 kcal mol−1 between the diastereoisomeric TSs TS20cS and TS20cR, which corresponds to 88% ee. These computational results motivated us to tackle the synthetic challenge of introducing an enantiomerically pure phenylmethylphosphine unit onto the backbone (a primary alkyl halide) and prepare the new pincer ligand 2c as described above.
Disappointingly, however, the phenylmethylphosphine-derived bromocarbonylhydride complex [FeBrH(CO)(2c)] (19c) gave only 44% ee (Scheme 36), which implies an energy difference for the enantiodetermining transition states of 0.6 kcal mol−1. Surprised at first, we tested other DFT functionals and basis sets, but all gave energy values within ±0.4 kcal mol−1 of 1.6 kcal mol−1. Eventually, single-point energy calculations with MP2 gave an energy difference of 0.6 kcal mol−1 between TS20cS and TS20cR (Fig. 6 bottom), which perfectly corresponds to the experimental enantioselectivity of 44% ee. Most probably, DFT fails to predict the enantioselectivity accurately with the P(Me)Ph-based pincer 2c because it overestimates the energy of the C–H⋯π interactions, which is a known effect.90 When the diastereoisomeric TSs contain the same number of CH/π interactions (such as TS20aS and TS20aR, which have one each, Fig. 6), the errors cancel out, and DFT and MP2 give the same result. Instead, for 2c, only the less energetic TS20cS features a CH/π interaction whose energy is most probably overestimated by DFT, leading to a spuriously high energy difference with TS20cR (1.6 kcal mol−1, vs. 0.8 kcal mol−1 by MP2).
A further intriguing feature is that, although catalysts 19a and 19c contain pincer ligands with opposite P-stereochemistry (S vs. R, respectively) they hydrogenate acetophenone with the same sense of induction (S). Therefore, the methyl group acts as the “small” substituent with PMe(Ph) (TS20c), but as the “large” one with P(Cy)Me (TS20a) (Fig. 6), which suggests that attractive non-covalent interactions (NCI, e.g. CH/π, π/π stacking), rather than repulsive steric hindrance, dominate (or at least strongly influence) enantioselection with these ligands, which are not exceedingly bulky. The NCI analysis supports this view.90 Interestingly, similar observations have been made with P-stereogenic diphosphines in the context of the quadrant rule. Gridnev has identified attractive interactions (C–H/π or C–H/H–C) between the coordinated substrate and the ligand even in the crowded quadrants. Increasing the overall steric bulk tightens the contacts and turns attraction to repulsion.35b
The importance of non-covalent interactions poses a further challenge to the rational ligand design by calculation. In particular, hybrid DFT methods, although a reliable tool to study the geometry of transition states, do not deliver energy values that are accurate enough to predict the enantioselectivity when weak interactions (such as C–H⋯π) are involved. Post Hartree–Fock methods such as MP2 are more accurate, but much more demanding computationally than hybrid DFT methods.
Thus, in P-stereogenic diphosphines, the substituents of phosphorus must differ largely in size to lock the chelate in the conformation in which the large substituent occupies the equatorial position. As aryl substituents on phosphorus are flat, their apparent size depends on their conformation. Hence, in DiPAMP, the axial phenyl group acts as the small substituent with respect to the chelate, but as the large substituent toward the substrate (Scheme 6). With P(tBu)Me (or P(Cy)Me) as donors, instead, the large substituent occupies the equatorial position, leaving the small methyl group axial. In this conformation, both substituents act as small, notwithstanding from which direction the substrate approaches. This is illustrated by the crystal structure of 18a (Fig. 5) and in the transition state TS20aS (Fig. 6). Therefore, one might speculate that methyl as the axial group is too small to enforce strong enantiodiscrimination. In fact, the situation is more complex, as our studies with the PNP and PN(H) pincers show.
Alternatively stereogenic phosphorus can be exploited to direct the configuration of multiple chelate rings of a polydentate ligand. In this case, the multichelate structure is the element of rigidity whose necessity has been recognised in the above discussion. With our N2P2 macrocyclic ligands, we have shown that this approach generates extremely rigid structures with predictable structure and absolute configuration that deliver high enantioselectivity in the transfer hydrogenation of ketones for a broad substrate scope.10 Therefore, we believe that stereogenic phosphorus will continue to play an important role in the synthesis and stereochemical understanding of asymmetric catalysis.
This journal is © The Royal Society of Chemistry 2019 |