Secondary Bracing Ligands Drive Heteroleptic Cuboctahedral PdII12 Cage Formation

The structural complexity of self-assembled metal–organic capsules can be increased by incorporating two or more different ligands into a single discrete product. Such complexity can be useful, by enabling larger, less-symmetrical, or more guests to be bound. Here we describe a rational design strategy for the use of subcomponent self-assembly to selectively prepare a heteroleptic cage with a large cavity volume (2631 Å3) from simple, commercially available starting materials. Our strategy involves the initial isolation of a tris(iminopyridyl) PdII3 complex 1, which reacts with tris(pyridyl)triazine ligand 2 to form a heteroleptic sandwich-like architecture 3. The tris(iminopyridyl) ligand within 3 serves as a “brace” to control the orientations of the labile coordination sites on the PdII centers. Self-assembly of 3 with additional 2 was thus directed to generate a large PdII12 heteroleptic cuboctahedron host. This new cuboctahedron was observed to bind multiple polycyclic aromatic hydrocarbon guests simultaneously.


General Information
Unless otherwise stated all starting materials were sourced from commercial suppliers and used without further purification. Self-assembly reactions were conducted in d 3 -MeCN.
NMR spectra were recorded using the following NMR spectrometers: Bruker 400 MHz Avance III HD Smart Probe (routine 1 H NMR, 1 H DOSY, host-guest 1 H NMR titration and 1 H NMR to monitor the conversion of 3 to 4), 500 MHz DCH Cryoprobe (High resolution 1 H, 13 C and 2D NMR experiments). Chemical shifts (δ) were reported in parts per million (ppm) for 1 H and 13 C spectra. Chemical shifts were referenced using the residual CD 3 CN solvent signal ( 1 H = 1.94 ppm, 13 C = 118.26 ppm). Signal multiplicity in 1 H NMR spectra was described using the following abbreviations: singlet (s), doublet (d), triplet (t), doublet of doublets (dd), triplet of doublets (td), doublet of doublet of doublets (ddd), multiplet (m), broad (br) and apparent (app.). 1 H DOSY NMR experiments were conducted on a Bruker 400 MHz Avance III HD Smart Probe spectrometer.
Maximum gradient strength was 5.35 G/cm A. The standard Bruker pulse program, ledbpgp2s, 3 employing a stimulated echo and longitudinal eddy-current delay (LED) using bipolar gradient pulses for diffusion using 2 spoil gradients, was utilized. A gradient ramp of 10% to 80% was used. d20 was set to 0.1 s and p30 was optimised for each species.
Low resolution electrospray ionisation mass spectra (ESI-LRMS) were recorded on a Micromass Quattro LC instrument (cone voltage 10 or 20 eV; capillary voltage 3.8 kV; desolvation temperature 313 K; source block temperature 313 K), infused from a Harvard Syringe Pump at a rate of 10 µL per minute.

Preparation and characterization of the host-guest complexes
A 0.5 mL solution of 4 (approx. 160 µM) in CD 3

Determination of the association ratio and binding constant for the pyrene ⊂4 host-guest complex
Job-Plot, stoichiometry plot and Hill function titrations were carried out to determine the association ratio and

Competitive encapsulation experiments
To a 160 µM solution of 4, phenanthrene, 9,10-dimethyl anthracene, triphenylene, pyrene and perylene (10 equiv. each) were added and the mixture stirred at 80ºC for 2 h. After removal of the excess guests by filtration, the resulting solution was concentrated using a stream of N 2 and Et 2 O (ca. 15 mL) was added. The yellow precipitate was dissolved with the minimum quantity of CD 3 CN to prevent dissolving residual free guests remaining. The solution was filtrated and the 1 H NMR spectrum was measured to reveal preferential formation of pyrene ⊂ 4 (66%) and perylene ⊂ 4 (33%).   Data were collected at Beamline I19 of Diamond Light Source employing silicon double crystal monochromated synchrotron radiation (0.6889 Å) with ω and ψ scans at 100(2) K. 1 Data integration and reduction were undertaken with Xia2. [2][3][4] Subsequent computations were carried out using the WinGX-32 graphical user interface. 5 Multi-scan empirical absorption corrections were applied to the data using the AIMLESS 6 tool in the CCP4 suite. 7 The structures were solved by direct methods using SHELXT 8 then refined and extended with SHELXL. 9 In general, non-hydrogen atoms with occupancies greater than 0.5 were refined anisotropically.
Carbon-bound hydrogen atoms were included in idealised positions and refined using a riding model. Disorder was modelled using standard crystallographic methods including constraints, restraints and rigid bodies where necessary. Crystallographic data along with specific details pertaining to the refinement follow. Crystallographic data have been deposited with the CCDC (2175722-2175723). Substantial bond length and thermal parameter restraints were applied to facilitate a reasonable refinement of the disordered anions and the lowest occupancy anions were modelled with isotropic thermal parameters. One coordinated acetonitrile was also modelled as disordered over two locations.

S32
A remaining area of diffuse electron density corresponding to highly disordered solvent could not be modelled despite many attempts using restraints or rigid bodies. Consequently the SQUEEZE 10 function of PLATON 11 was employed to remove the contribution of the electron density associated with this highly disordered solvent, which gave a potential solvent accessible void of 299 Å 3 per unit cell (a total of approximately 90 electrons). The diffuse solvent molecule could not be conclusively assigned to acetonitrile or ethyl acetate and was therefore not included in the formula. The crystals of [4]·24NTf 2 [+ solvent] were grown by diffusion of diispropyl ether into an acetonitrile solution of the complex. The crystals employed were very weakly diffracting and immediately lost solvent after removal from the mother liquor. Rapid handling prior to flash cooling in liquid nitrogen and the use of synchrotron radiation were required to collect data. Consequently few reflections at greater than 1.1 Å resolution were observed and the quality of the integration is less than ideal. Nevertheless, the quality of the data is far more than sufficient to establish the connectivity of the structure. The asymmetric unit was found to contain one quarter of a Pd 12 L 4 L 8  assembly and associated counterions.
Due to the limited resolution of the data and high level of thermal motion (or minor unresolved disorder) present throughout the structure, the GRADE program 12 was employed, using the GRADE Web Server, 13 to generate a full set of bond distance and angle restraints (DFIX, DANG, FLAT) for each of the organic ligands. Thermal parameter restraints (SIMU, RIGU) were applied to all atoms except for palladium to facilitate anisotropic refinement.
The anions within the structure also show evidence of thermal motion or disorder which could not be resolved and substantial bond length and angle restraints were required to achieve a reasonable model. The occupancies of the anions were initially refined and then fixed at the obtained values. Further reflecting the solvent loss and poor diffraction properties there is a substantial amount of void volume in the lattice containing smeared electron density from disordered solvent and ca. 21 unresolved anions per Pd 12 L 4 L 8  assembly (included as triflimide in the formula). These anions and were significantly disordered and despite numerous attempts at modelling, no satisfactory model for the electron-density associated with them could be found. Consequently the SQUEEZE 10 function of PLATON 11 was employed to remove the contribution of the electron density associated with these remaining anions and further highly disordered solvent, which gave a potential solvent accessible void of 89533 Å 3 per unit cell (a total of approximately 33403 electrons). Diffuse solvent molecules could not be assigned to acetonitrile or diisoproyl ether and were therefore not included in the formula.
Consequently, the molecular weight and density given above are underestimated.
CheckCIF gives one A level alert resulting from the limited resolution of the data and one B level alert (Large Average Ueq) resulting from unresolved disorder of a triflimide anion which is located on a special position.

Molecular Modelling
Geometry optimized host-guest structure were modelled at the PM6 level of theory using the program SCIGRESS. Molecular modelling was carried out by docking 8 pyrene guest units in the X-ray structure of 4 and minimizing using MMFF, with the following parameters: RHF (restricted HartreeFock Hamiltonian), LBFGS (low memory Broyden-Fletcher-Goldfarb-Shanno procedure), a maximum of 2000 SCF (self-consistent field) iterations and a SCF criterion at 10 -4 kcal/mol.