Pd-η 3-C6H9 complexes of the Trost modular ligand: high nuclearity columnar aggregation controlled by concentration, solvent and counterion

Pd-η 3-C6H9 cations bearing the Trost ligand (2) undergo two-stage oligomerisation-aggregation to form high nuclearity aggregates (up to 56 Pd centres), with aggregation strongly modulated by concentration, solvent and counter-anion.


Introduction
The Trost modular ligand (TML) series 1 has been applied to an extraordinarily wide range of allylic alkylation (Tsuji-Trost) reactions. 24][5][6] These features have led to broad use of the TML in the synthesis of natural products, 7 as well as industrial application for the construction of high-enantiopurity chiral building blocks.However, reactions involving the TML can exhibit memory effects, 8,9 and a high sensitivity of the enantioselectivity to reaction temperature, catalyst concentration, solvent and nucleophile counter-ion.
Our previous mechanistic studies of this system focussed on the monomeric cationic complex [1] + , in which a Pd(h 3 -C 6 H 9 ) unit is chelated by the 1,2-diaminocyclohexane-derived TML ligand (2). 10 The monomeric cation [1] + was identied as an intermediate capable of leading to high asymmetric induction on attack of, for example, a malonate anion nucleophile, Scheme 1. Detailed NMR studies facilitated by isotopic labelling in conjunction with MM-DFT simulationsled to a model 10 in which the amide units in the catalyst facilitate enantioselective ligand-accelerated catalysis. 11or the ligand-accelerated catalysis to function efficiently, cation [1] + requires a degree of exibility.This exibility is provided by the 13-membered chelate ring, but at a cost: complex [1] + can readily undergo ring-opening oligomerisation to generate polynuclear species ([1] + ) n , Scheme 2. 12 Competing nucleophilic attack on the oligomer, rather than the monomer [1] + is, in part, responsible for a reduction in overall enantioselectivity under non-optimised conditions. 13,14o date, the structure and origin of the formation of these oligomeric species has not been studied in detail.Herein, we describe an investigation of the oligomerisation of D 0 and D 47 isotopologues of [1] + , employing NMR spectroscopy, molecular mechanics (MM), molecular dynamics (MD), and contrast variation small-angle neutron scattering (SANS).The data obtained indicate that the impact of a rst-stage of depletion of the monomeric species [1] + from the catalyst pool, via cyclic oligimerisation, is amplied by a second-stage process involving columnar aggregation of the oligomers, leading to species with very high nuclearity (up to 56 Pd centres).The effects of solvent, ligand enantiopurity and counter-ion on the degree of aggregation are explored in detail, and it is concluded that a relatively small and restricted set of conditions facilitate dissolution of the complexes in a low aggregation state, consistent with the extensive optimisation frequently required for these catalyst systems.

Results and discussion
Preliminary NMR studies and synthesis of [D 47 ]- [1] Despite extensive efforts, 15 we have been unable to crystallise any Pd(h 3 -C 6 H 9 ) complexes of 2, in either oligomeric or monomeric forms.Indeed, to date, the only the X-ray crystal structures of Pd-allyl complexes of ligand 2 16 are h 3 -C 3 H 5 complexes with triate counter anions: one a racemic tetranuclear cyclo-oligomer, 12 the other an acyclic dinuclear bis-P,O-chelate. 17he extent of solution-phase oligomerisation of cationic complexes of type [1] + can be conveniently estimated by 31 P{ 1 H} NMR spectroscopy. 10,12Analysis of [(R,R)-1][BAr 4 ] complexes in CH 2 Cl 2 , where Ar ¼ C 6 Cl 5 , 3,5-(CF 3 ) 2 C 6 H 3 , or C 6 F 5 ("BAr F "), indicates a maximum monomer concentration ([1] + ) of about 4 mM, Fig. 1.In THF, the monomer maximum is lower (approx.1.6 mM) and decreases as [Pd] tot is raised above 10 mM.With smaller, less charge-diffuse, counter-anions such as chloride or triate, the maximum monomer concentrations are lower still.We were unable to t simple analytical solutions 18 for monomer-oligomer distributions to any of the 31 P NMR data, indicative that physicochemical effects dominate over simple solution-phase equilibria, even at low [Pd] tot .
We thus elected to study the oligomeric species by SANSa technique that can be used for characterising the shape and dimensions of self-assembly structure and colloids. 19We began with [(R,R)-1][BAr F ], 20    We have previously used 31 P NMR spectroscopy to analyse the constitution of the solution-phase (i.e., lower-order) oligomers generated from various complexes of type [1][BAr 4 ] in CD 2 Cl 2 . 10,12Using PPCOSY in combination with pairs of isotopically-differentiated ligands ([D n ]-2), we were able to determine  that the oligomers are: i) non-chelated species (i.e. each of the ligands (2) in the oligomer are coordinated to two different Pd centres); ii) present in predominantly homochiral form (i.e.[(R,R)-1] + and [(S,S)-1] + oligomerise independently), and iii) contain no 'free' (i.e.not Pd-coordinated) P-centres in the ligand (2).Although a cyclic oligomer structure (Scheme 2) is fully consistent with these features, we were unable to determine the number (n) of ring-opened monomer units incorporated within the cyclo-oligomer ( Computational modelling was employed to probe the factors that control the aggregation phenomena, and to estimate the size (n) and number of cyclo-oligomers in the particle (m). 38The number of atoms in the aggregates (>2,000, vide infra) means that density functional theory (DFT) calculations of their structures would demand currently unattainable computational resources and inordinate simulation times.On the other hand, molecular mechanics (MM) can provide a good approximation in just a small fraction of the computational time required by DFT.Although MM in general calculates only a steric energy, not the full free energy, and ignores bond dissociation energies, accurate results can be obtained from MM3 calculations provided that comparisons are isodesmic and, more importantly, isoparametric. 39This requirement is fullled for all comparisons of ring oligomers and their non-covalent aggregates.We began by conrming that the structure of the 84-atom monomeric cationic P,P-chelate [1] + , optimised at MM3 level of theory with dielectric constant 3 ¼ 9.0, was almost identical to that obtained using DFT (B3LYP-D3) in a polarisable continuum model for dichloromethane. 38Further calculations involving ion-pairs, oligomers and aggregates were then performed using MM3 to provide analysis of the energies involved, albeit at a coarse-grained level of detail.
Further increase in oligomer ring size (n ¼ 5, 6, 8) yields a modest reduction in the system energy but generates species with signicantly higher radii than the 8-9 Å cylinder radius detected by SANS.In summary, the tetranuclear species (n ¼ 4),   Positively charged rod-like structures with dissociated or removed anions were estimated by MM3 to be very much higher in energy than those where the anions were closely associated with the cationic cyclo-oligomeric building blocks.Anion interactions were thus explored more deeply, and although BAr F is considered a weakly coordinating anion, 41 the charge delocalisation over its surface reduces repulsive interactions with other BAr F anions, and the presence of uorine makes it signicantly lipophilic.Indeed, the calculations indicated a favourable interleaving of the BAr F anions in sandwich layers 42 between cationic cyclo-oligomers.The estimated formation energies, (DE, Fig. 7) of such species {([1][BAr F ]) 4 } m as a function of 'm' indicated that columnar aggregates are readily attainable, with the growing entropic cost (TDS) placing limits on the aggregate length. 34,40hese conclusions were further probed by moleculardynamics (MD) simulations in which the MM3-minimised structures {([1][BAr F ]) 4 } m were computationally excited (300-700 K) over short periods (300 ps) to test the relative structural integrity of the aggregate as a function of 'm'.In the low dielectric constant medium used for the model, most systems (m ¼ 4 to 16) did not undergo any signicant changes in their tertiary structure at 300 K.As the energy input was increased the aggregate models exhibited varying degrees of structural deformation, undergoing rapid fragmentation at the highest energies.The most signicant observations were made at intermediate energies (500-550 K): aggregates with m ¼ 10-14 (e.g., Fig. 8, m ¼ 12) retained a cylinder shape, albeit mildly distorted, over the full 300 ps simulation time, whereas higher or lower order aggregates signicantly    deformed, in some cases losing one or more BAr F anions.The average dimensions of the MM3 aggregates with m ¼ 10-14 (radius 8-9 Å and length 150-200 Å) are consistent with the particle dimensions determined by SANS, Fig. 2 and 3.

The effect of counter-ion and solvent on shape and extent of aggregation
The effect of solvent type on aggregation was probed by MM, 31 P { 1 H} NMR (Fig. 9) and SANS, also comparing [1][BAr F ] with [1][OTf] to explore the impact of counter-ion.The aggregation mode for [1][BAr F ] determined by MM3, e.g., Fig. 8, involves multiple close-range electrostatic interactions that reduce the overall energy of the system.It is therefore not surprising that the solvent dielectric constant (3 r ) was found to modulate aggregation, 43 and thus also solubility, with precipitation being the ultimate consequence of strong aggregation.
As indicated in Fig. 9, both [1][OTf] and [1][BAr F ] readily oligomerise and aggregate in all of the solvents that were explored, becoming essentially insoluble at the extremes of 3 r , (e.g., in alkanes, most ethers, chloroform, aromatic hydrocarbons, and at the opposite end of the scale, in water).The lipophilicity and charge-density of the anion also affects the solubility: [1][BAr F ] (but not [1][OTf]) readily dissolves in THF, and at the opposite end of the 3 r scale, [1][OTf] (but not [1][BAr F ]) is soluble in aqueous-organic mixtures.
SANS was employed to explore how the macromolecular composition of aggregates {( [1][X]) n } m is affected by solvation.Although [1][BAr F ] is not soluble in organic-aqueous mixtures, SANS data were attainable in polar aprotic solvents (e.g., MeCN, 3 r ¼ 37.5; and DMSO, 3 r ¼ 47).This conrmed that cylindrical aggregates were still formed, but were signicantly shorter than those in THF, Fig. 10.Medium length cylinders were detected in a 50 : 50 mixture of THF and acetonitrile, consistent with the intermediate solvent polarity (3 r z 23).In all cases, the cylinders were of radius 8-10 Å, strongly suggesting the prevalence of the tetranuclear cyclo-oligomer building blocks, with the solvent modulating only the aggregation number 'm': The [1][OTf] aggregates behaved differently.Although, cylinders of radius 8-10 Å were again detected in all cases, indicative of {( [1][OTf]) 4 } m aggregates, the exibility, lengths and charge distribution in the particles were very different to those formed from [1][BAr F ].
In CD 2 Cl 2 , [1][OTf] forms cylindrical aggregates (up to 160 Å; Fig. 11) apparently with a degree of exibility, a phenomenon that can be attributed to the small and interactive triate anion being less able to rigidify the structures than the larger and more lipophilic BAr F anion.The anion effect became even more pronounced in media of higher dielectric constant.In acetonitrile-based solvent mixtures (3 r ¼ 47-58) the SANS data indicated an additional minor structure factor contribution (S(Q)), consistent with weakly charged particles, Fig. 12. Weak repulsive interactions might arise from solvation-induced ion-pair separation of triate from the cationic Pd(II) oligomeric cores.The increased cationic repulsion between the cyclo-oligomers appears to result in much shorter cylinders, just 30 Å in length, with misleadingly simple 31 P NMR spectra. 44Similar conclusions were drawn from MD simulations with the medium set at    Finally, to probe the relevance of the higher aggregates to asymmetric alkylation (Scheme 1) SANS data were acquired on reaction mixtures in which [1][BAR F ] was employed as a pre-catalyst (10 mol%) for addition of tetrabutylamonium dimethylmalonate to cyclohexenylacetate in THF.While the effects of substrate background scattering, varying acquisition times and shorter Q-range slightly affected the data quality, it remained clear that the dominant structures in solution, for the whole duration of the catalytic process, were large cylinders.This result is consistent with previous conclusions that, in THF, the catalytic turnover proceeds via a small pool of highly-active monomeric catalyst species, in competition with cyclo-oligomers and aggregates. 10,12

Conclusions
Since the initial report that [1] + readily oligomerises, 13 and that this is a largely undesirable property of an otherwise highly efficient catalyst, e.g., Scheme 1, there has been limited understanding of the oligomer structures. 10,12We have now identied, through NMR spectroscopy, SANS, and MM/MD simulations of The identity of the counter-anion has a pronounced effect on the proportion of oligomer generated from the monomer.Bulky, weakly-coordinating anions, 46 reduce the extent of oligomerisation, particularly in low polarity solvents that cannot effectively stabilise charged particles.Here the role of the bulky and relatively lipophilic anions is to solvate the monomer [1] + .Smaller harder, less lipophilic anions are less able to solvate the monomer, and have the indirect effect of shiing the equilibrium towards the oligomer; an undesirable feature for catalysis.The diminutive size of the anion also results in greater exibility of the resulting columnar aggregates, which are more ionic in nature, reducing their solubility in less polar solvents.
Overall, although the solvent polarity, counter-anion (X), and net concentration ([Pd] tot ) all affect the degree of oligomerisation and aggregation (Fig. 13) of monomer [1][X], the solvent perhaps offers the greatest degree of scope for optimisation under the conditions of catalysis.In this regard, CH 2 Cl 2 is favourable: solutions can be virtually free of oligomer at ambient temperature, provided [Pd] tot # 4 mM.Intriguingly, SANS studies of ionic and non-ionic surfactants 47 have revealed specic solvent combinations that can lead to "dead zones" where aggregation is suppressed, even for concentrated solutions.If such "dead zone" solvent combinations can be found for complexes of type [1][X], this may be highly advantageous for improving catalytic productivity whilst maintaining selectivity.

Fig. 6 ,
Fig.6, appears to be the dominant, if not exclusive, cyclo-oligomer, and aggregation of these cyclic tetramers ([1][BAr F ]) 4 was thus probed by MM3 as a process to generate the cylindrical particles, Fig.7.Positively charged rod-like structures with dissociated or removed anions were estimated by MM3 to be very much higher in energy than those where the anions were closely associated with the cationic cyclo-oligomeric building blocks.Anion interactions were thus explored more deeply, and although BAr F is considered a weakly coordinating anion,41 the charge delocalisation over its surface reduces repulsive interactions with other BAr F anions, and the presence of uorine makes it signicantly lipophilic.Indeed, the calculations indicated a favourable interleaving of the BAr F anions in sandwich layers42

Fig. 10
Fig. 10 SANS data and cylinder form factor fits for [1][BAr F ] in various solvents at 25 C.

Fig. 12
Fig. 12 SANS data and cylinder form factor fits (including an effective Hayter-Penfold structure factor S(Q) to account for charged particles) for [1][OTf] in CD 3 CN at 25 C.