Conformational Analysis of Phosphinamine Ligands

Brown, John M.; Hulmes, David I.; Long, James M.; Valk, Jean Marc; Pearson, Stuart; Bayston, Daniel M.; Goeke, Andreas; Muir, Jayne M.; Alcock, Nathaniel W.
Dyson Perrins Laboratory, South Parks Rd., OXFORD OX1 3QY; Department of Chemistry, University of Warwick, Coventry CV4 7AL (NWA)
e-Mail john.brown@dpl.ox.ac.uk


Introduction

During the course of our work on asymmetric catalysis by ligands of the 1-(phosphinoaryl)-isoquinoline class[1] a number of X-ray structures of organopalladium complexes have been obtained which give insights into the conformation of the chelate ring, and the sensitivity of its geometry to changes in the backbone structure and the coordination environment. Further, comparison may be made with both the biaryl diphosphines related to BINAP[2] and other ligands such as the phosphinoaryl oxazolines,[3] which have enjoyed wide application in catalysis; the latter is discussed here. The analysis is divided into sections as follows :

1. Simple complexes of the parent ligand QUINAP and a heteroatom relative;

2. Steric discrimination in optical resolution by palladium complexes

3. Familial comparisons between P-N ligands in the series

4. Comparisons with structurally related ligands.


Discussion

1.Simple complexes

Three relevant X-ray structures have been determined, all of dichloropalladium complexes. The first of these is of the parent ligand 1,[4] shown in Figure 1 and the two independent molecules defined in the refinement are displayed as MODEL 1a and MODEL 1b. A structure of the quinoxaline analogue 2 has also been obtained[5] and is also shown in Figure 1, and as MODEL 2. From earlier work, a phosphinoarylisoquinoline was obtained as a racemate and the X-ray crystal structure of the PdCl2 complex obtained.[6] This is shown in Figure 1, and as MODEL 3.

The basic ligand structure possesses a number of surprising features common to the four distinct molecules. Basically the 6-membered chelate ring is a distorted boat conformation, hinged about the phosphorus and the 1-carbon of the heterocycle. The bite angle is quite small for a 6-ring chelating ligand, varying between 84[ring] and 86[ring] for the four structures, and some strain is obvious from the observed distortions shown in Figure 1. This is most obviously manifested as torsional strain by deviations of the N-Pd and C-P bonds from their respective ring planes, most strikingly in the first case. Angle strain is most obvious in the enhanced values of 125.5 +/- 2[ring] observed for Pd-N-C (ideally 120[ring]) and the reduced values of 101.5 +/- 2.5[ring] observed for the endocyclic Pd-P-C (ca. 112[ring] and 123[ring] for the exocyclic Pd-P-C angles). Most of the torsional strain is taken up by deviation of the N-Pd bond from the lone pair vector. The greater this distortion, which is 350 in compound 2 and 210 or 260 in the two independent structures of compound 1, the less the C-P bond is twisted out of the ring plane in compensation. The surprisingly high level of tolerance to torsional strain in the bond between an sp2-N and Pd suggests that there may be some increasing contribution from p-bonding from the arene as the palladium moves further from the ring plane; the Pd-N bond lengths are within their normal range. The variable geometry of the three complexes suggests that the chelate structures could represent points on a trajectory which maps out the dissociation path for the Pd-N bond. Only the most Pd-N distorted complex 2 shows catalytic chemistry indicative of an accessible monophosphine complex, and therefore the Pd-N linkage does not dissociate readily. The related structure 3 provides similar features to compound 1.

2.Steric discrimination in optical resolution

In the course of early synthetic work it was found that the only effective method for resolution of the racemic QUINAP ligand was through the chloropalladium complex of (R)-N,N-dimethyl-1-(1-naphthyl)ethylamine;[7] separation of the diastereomers failed completely with the chloropalladium complex of (S)-N,N-dimethyl-1-phenylethylamine. Both the resulting complexes have been crystallographically characterised, and the respective structures of the (R,R)-diastereomer 4 and the (S,R)-diastereomer 5 are shown in Figure 2, and as MODEL 4 and MODEL 5. These two complexes have very different physical properties and are easily separated by crystallisation; the (S,R)-diastereomer is formed preferentially when a deficiency of the resolving agent is employed. In contrast, attempted resolution with (S)-N,N-dimethyl-1-phenylethylamine leads to the crystallisation of a quasiracemate 6 in which the CHMe group of the resolving agent occupies a different conformation in the two otherwise mirror-related structures (ref. 4). These are also shown in Figure 2 and as MODEL 6.

Two immediate questions arise from these observations. Firstly, the origin of the dramatically different efficiencies of two closely related resolving agents needs to be explained, and secondly the distinct response of the two hands of QUINAP towards the chloropalladium complex of (R)- should be rationalised. The first of these has been discussed in our previous publications in terms of the locking effect of the naphthalene 8-H in the resolving agent on the conformation of the CHMe group in the C,N-palladocycle. Basically, the Me-group is forced into an axial conformation in order to avoid an unfavourable steric clash with the peri-hydrogen. This conformation is observed in all X-ray structures containing the N,N-dimethyl-1-(1-naphthyl)ethylamine palladocycle, but this is not the case for the corresponding N,N-dimethyl-1-phenylethylamine palladocycle (refs 1,4).

Given this explanation for the enhanced rigidity of the naphthylethylamine-derived resolution complexes, the related question of discrimination between diastereomers arises. In the resolution complexes, higher stability of the (R,S) [or (S,R) diastereomer] is observed for all the P-N ligands we have synthesised, and it is reasonable to suppose that it has a general explanation. The first efforts to rationalise it involved a search for differences in intracomplex non-bonded interactions between ligand and resolving agent. It is at first sight surprising that there are none apparent, at least as far as the "hard" H-H repulsions are concerned. The only serious clash is between H3 of the isoquinoline and one of the N-CH3 groups; in the (R,R)-diastereomer the distance of the closest pair is 2.389A, and for the (R,S)-diastereomer the closest distance is 2.377A. There are no further discernible H-H repulsions between the two fragments! This implies that the steric strain has been distributed into other channels, although analysis of the ligand chelate fails to reveal specific differences there. A number of small differences are observed which could be invoked to explain the difference in free energy between the complexes, but none consistently provides a reasonable explanation; for example the metallacyclic chelate ring is flatter in the less stable (R,R)-diastereomer 4 than in either complex 5 or 6. The single factor which appears to correlate with stability is the degree of distortion from an ideal square plane as indicated in Figure 3. The trend there is clear, and suggests that the metal-ligand bonds have a much softer potential towards angular distortion than the linkages between light atoms. At the same time the variation in bond lengths is quite small, suggesting that little strain energy is partitioned there.

Although it is not possible to define the origin of enantiomer discrimination directly, the accessibility of the coordinates of the structure 6 permit a "smoking gun" approach. Hence we have constructed models of the hypothetical undistorted precursors of complexes 4 and 5 (4-0 and 5-0) [Construct 4-0 is MODEL 7 and construct 5-0 is MODEL 8.] by converting the two forms of the quasiracemate into the two related diastereomers using Chem 3D. This was done with minimal structural change so that the models reflected an undistorted square planar complex between the palladocycle and ligand with the raw steric interactions involved. The adaption of the (R,S) diastereomer 5-0 from the X-ray structure 6 was straightforward, and showed as expected that the only significant repulsions were within the palladocyclic component and also that cross-ligand interactions were insignificant. Adaption of the (R,R) diastereomer in the structure of 6 proved less straightforward. After building the naphthalene moiety, the dihedral angles of the 5-membered ring were altered to resemble those of the authentic structure 4, shifting the C-methyl group into an axial environment. Finally, a simple energy minimisation was carried out restricting movements of all atoms save the Me2NCHMe fragment. The resulting structures are shown in Figure 4. From analysis of this the origin on the stability difference becomes clear. There is a steric clash between H3 of the isoquinoline and an N-methyl group, and at the same time another between a P-phenyl group in the ligand and the naphthalene portion of the amine reagent. This dual interaction evidently causes significant destabilisation, but in the readjustments which lead to the ground state of complex 4 (as manifested by its X-ray structure) it is no longer evident. This observation permits a more general postulate:

"Strain due to molecular crowding in an organometallic complex can be most readily accommodated by dispersal away from hard H-H and C-H repulsions into distortion of the coordination sphere; in this, angle bending strain is preferred over bond stretching strain"

This is in accord with the empirical observations in the present series and elsewhere, and presents an important factor in ligand design, equally significant in the interpretation of catalytic selectivity, especially for asymmetric synthesis. To plan on the basis of models which express the space filling properties of ligand and/or substrate disguises the true picture, since these entities will remain relatively undistorted.

3.Familial comparisons between P-N ligands

In this section we enquire about the changes incurred in ligand and complex geometry brought about by changes in the substitution pattern of the backbone. As a secondary question, the effects on the P-N chelate of changing its coordination environment is assessed. There are three structures of the stable (R,S)-diastereomer of the resolution complex, namely compounds 5 (discussed in Section 2 above), the phenanthridine analogue[8] 7 and the indole-derived ligand complex[9] 8. These latter two are respectively MODEL 9 and MODEL 10. The structures appear in Figure 5.

We first make a direct comparison between the phenanthridine-derived complex 7 and its parent; it is clear that additional distortion is apparent in the latter. The chelate angle is narrowed to 82.40, and the endocyclic Pd-P-C angle is 91.20, much more acute than the normal value which is close to tetrahedral. The aryl P-C bond is significantly longer than the norm of 0.182 nm in this series at 0.187 nm. But surprisingly there is less distortion of the N-P vector with the relevant endocyclic torsion angle being only 120 well outside the familiar 22-35[ring] seen in other complexes; for example it is 27.80 in complex 5. More surprisingly, some of the same trends are shown in the related complex 8 where the naphthalene entity has been replaced by a 2,3-disubstituted indole. The unusual feature here (most others fit the trends already discussed) is the fact that the Pd-N vector is essentially in the ring plane with an endocyclic Pd-N-C-C torsion angle of 4[ring]. Overall, however, this complex appears significantly less strained and the square plane is less distorted than in the case of the parent complex 5.

The parent in the indole series suffers from easy racemisation and its utility in catalysis is thus limited. Despite this, and in preparation for work on derivatives which should be enantiomerically stable, the allyl complex 9 [Model 11]was obtained crystalline; the central part of its X-ray structure in comparison to complex 8 is shown in Figure 6. To underscore our lack of predictive power in analysing these structures there are major structural differences between the two structures, since the allyl complex possesses a chelate ring which is much flatter than any other compound in the series. This is achieved by swinging the metal into the plane described by phosphorus and the C=N bond, with the effect of widening the bond angles in the biaryl fragment and also increasing the bite angle from 82[ring] to 940. In addition, the out-of-plane distortion of the N-Pd vector has disappeared. Other than to compensate the narrow bite angle of the allyl fragment, the origins of this profound change are unclear. But an apparently rigid chelate ring structure has access to a range of geometries, and hence subtle effects can operate in catalysis. One interesting facet is that the diastereoselectivity of formation of complex 9 - binding the 1,3-diphenylallyl group through its opposite (enantiotopic) face provides a diastereomeric structure - is much higher at 30 : 1 than in the case of the parent QUINAP at about 3 : 1 under similar conditions.

The final internal comparison is between the parent PdCl2 complex and a tridentate amide formed from the 3'-carboxylic acid, synthesised as a precursor to the corresponding side-chain oxazoline. 10 Complex 10 was crystallised in expectation of forming a dichloride and proved to have eliminated HCl in the procedure. The structure shown in Figure 7 [MODEL 12] demonstrates close proximity between a substituent at the isoquinoline 3-position and the metal, enanbling tridentate coordination. By comparison with the simple dichlorides the additional coordination is achieved with some flattening of the P-N chelate, manifested as a smaller internal P-Pd-N-C torsion angle but with little additional distortion.

4.Comparisons with structurally related ligands.

Heterotopic chelation is less common in organometallic chemistry and catalysis than homotopic P-P or N-N chelation, and a full search of the CSSR database in June 1997 revealed relatively few alternative structures where the bonding fragment is of the form P-C=C-C=N. The first such examples arise from Pignolet's work on ortho-phosphinylated Schiff's bases.[11] This led, for example to structures like complex 11 and relatives. In this case the chelate ring is less constrained and the characteristic feature of our ligands, namely the distortions in the N-Pd vector out of the aromatic ring plane, is absent. Overall, the chelate ring is flatter and the only distortion observed is in the C=N-Ir angles of 1310 and 1320 for the two chelate rings. Even this may be a consequence of the bridge between the two P-N entities rather than intrinsic to the structure.

More closely relevant structures are to be found in the catalytic chemistry of isoxazoles and in particular the phosphinoaryl oxazolines simultaneously studied by Williams, Helmchen and Pfaltz(ref 3). In this series there are two deposited crystal structures, both from Helmchen's group.[12] Geometrical features similar to the imines above are found in the chelate ring of complexes 12 and 13. The chelate is much flatter here than in the QUINAP series, perhaps inevitably since the structure of the latter must accommodate the twisted biaryl moiety. There is some strain in the compounds in this series but it is manifested differently, largely through an expansion of the internal chelate angles above the ideal 1200 value. Again, the ligating angle at nitrogen is widened to ca. 1250. Key features of the structures are displayed in Figure 8. Although superficially similar, the oxazoline P-N family are distinct from the isoquinoline P-N family in their catalytic chemistry. Both work well in catalytic asymmetric allylation, but the oxazolines are relatively insensitive to the reaction conditions and have a broader range of successful reactions. In the very different sphere of catalytic hydroboration, the isoquinoline-based ligands are superior and indeed out-perform diphosphines in comparable reactions. The intrinsic conformational flexibility of the QUINAP series was not predicted, and is not seen in diphosphines beyond the 7-ring chelate series of DIOP and its relatives, where at least two conformations compete for the ground state.[13] It remains to be seen how far this flexibility can be exploited in catalysis.

Acknowledgment

We warmly acknowledge the support provided by LINK Asymmetric Synthesis, and in particular the interest currently shown by sponsoring companies Chiroscience (Dr. U. Berens) Glaxo-Wellcome (Dr. A. Payne), Robinsons (Dr. J.C.P Sly) SmithKline Beecham (Dr. P. Sheldrake) and Zeneca (Dr. A. J. Blacker).

References

[1] N.W. Alcock, J.M. Brown and D.I. Hulmes Tetrahedron: Asymmetry 1993, 4, 743

[2] S. Akutagawa Applied Catalysis A General, 1995, 128, 171, and references therein.

[3] P. Vonmatt and A. Pfaltz Angew. Chem., Int. Ed. Engl., 1993, 32, 566; J. Sprinz and G. Helmchen Tetrahedron Lett., 1993, 34, 1769. G. J. Dawson, C. G. Frost and J. Williams Tetrahedron Lett., 1993, 34, 3149.

[4] N.W. Alcock, D.I. Hulmes and J.M. Brown J. Chem. Soc., Chem. Commun. 1995, 395 - see also reference 1.

[5] N.W. Alcock, S. Pearson and J.M. Brown, unpublished work, 1996

[6] N.W. Alcock, J.M. Brown, M. Pearson and S.C. Woodward, Tetrahedron : Asymmetry, 1992, 3, 17.

[7] D. Hockless, P. A. Gugger, P. H. Leung, R. C. Mayadunne, M. Pabel and S. B. Wild Tetrahedron, 53, 1997, 4083, and references therein.

[8] J. M. Valk, T.D.W. Claridge, J. M. Brown, D. Hibbs and M. B. Hursthouse Tetrahedron: Asymmetry, 1995, 6, 2597.

[9] T. Claridge, J. M. Long, J. M. Brown, D. Hibbs and M. B. Hursthouse Tetrahedron, 1997, 53, 4035.

10 A. Goeke and N.W. Alcock, unpublished work 1997

[11] T.L.Marxen,B.J.Johnson,P.V.Nilsson,L.H.Pignolet, Inorg.Chem., 1984, 23, 4663

[12] J. Sprinz, M. Kiefer, G. Helmchen, G. Huttner, O. Walter, L. Zsolnai and M. Reggelin Tetrahedron Lett. 1994, 35, 1523.

[13] G. Balavoine, S. Brunic and H.B. Kagan, J. Organometal. Chem., 1980, 187, 125; J.M. Brown and P.A. Chaloner, J. Am. Chem. Soc., 1978, 100, 4307.