A family of diastereomeric dodecanuclear coordination cages based on inversion of chirality of individual triangular cyclic helicate faces

The dodecanuclear coordination cage [Cd12(Lnaph)12(Lmes)4](BF4)24 consists of a set of four triangular, trinuclear helical panels {Cd3(μ-Lnaph)3}6+ (based on ditopic bridging ligands Lnaph), which are connected by four tritopic ligands Lmes. The result is that the four triangular helical panels and the four Lmes-capped triangular faces of the cuboctahedral core form two alternating subsets of the eight triangular faces of the cuboctahedron. Crystallographic investigations revealed that the triangular helicate faces can have ‘clockwise’ (C) or ‘anticlockwise’ (A) helicity, and that the helicity of each face can vary independently of the others as they are mechanically separated. This generates a set of three diastereoisomers in which all four cyclic helicate faces in the cuboctahedron have the same chirality (AAAA/CCCC enantiomers with T symmetry; AAAC/CCCA enantiomers with C3 symmetry; and achiral AACC with S4 symmetry). This mirrors the known behaviour of many simpler M4L6 tetrahedral cages which can likewise exist as T, C3 or S4 isomers according to the sense of tris-chelate chirality around each individual metal centre: but here it is translated onto a much larger scale by the four chiral units being entire trinuclear helicate faces rather than single metal centres. 1H NMR spectroscopy confirms the presence of the three diastereoisomers with their different molecular symmetries in a ratio slightly different from what is expected on purely statistical grounds; and 1H NMR measurements on a non-equilibrium sample (enriched by manual crystal-picking before preparing the solution) showed that the distribution does not change over several weeks in solution, indicating the kinetic inertness of the cage assemblies.


Synthesis
Ligands L mes and L naph were prepared according to our previously published methods (main text, refs, 8 and 12 respectively).
Elemental analysis calc. for Cd 12 (C 36 H 33 N 9 ) 4   Red labels indicate peaks assigned to the calculated m/z values.

Single Crystal X-Ray Diffraction Experimental Details
X-Ray quality single crystals were grown by slow diffusion of diisopropyl ether or diethyl ether into acetonitrile solution of [Cd 12 (L mes ) 4 (L naph ) 12 ](BF 4 ) 24 . Crystals of three different habits could be identified visually. Many recrystallisations were performed and multiple data sets were collected for each crystal type to try and obtain the best possible data. Single crystals were selected and mounted using Fomblin® (YR-1800 perfluoropolyether oil) on a polymer-tipped MiTeGen MicroMountTM and cooled rapidly to 120 K in a stream of cold N 2 using an Oxford Cryosystems open flow cryostat. 1 Single crystal X-ray diffraction data were collected on an Oxford Diffraction Gemini R (Ruby CCDC detector, fine focus graphite monochromated Cu-Kα radiation source; ω scans) for the AAAA/CCCC isomer or an Oxford Diffraction Supernova diffractometer (Atlas S2 detector, micro focus mirror monochromated Cu-Kα radiation source; ω scans) for AACC and AAAC/ACCC isomers. Cell parameters were refined from the observed positions of all strong reflections and absorption corrections were applied using a Gaussian numerical method with beam profile correction (CrysAlisPro). 2

AAAA-F23_sq
Refined as a 4-component inversion twin with twin law {0 -1 0 -1 0 0 0 0 -1} and three batch scale factors refining as 0.42(8), 0.42(8) and 0.07(8). Alternative twin laws were tested, all giving the same improvements in refinement and R-values. No Flack parameter is calculated for the refinement of the twinned structure. The X-ray diffraction intensity dropped off at high angle with a diffraction limit of 1.6 Å. As a result of the low-resolution diffraction limit the structure refinement has a low data to parameter ratio of 4.54:1. Extensive use of restraints on geometric and displacement parameters is used to aid the refinement.
An initial solution using direct methods (ShelXS) gave a plausible position for a cadmium atom and one pyrazolyl-pyridine moiety. The remaining ligand atoms were generated from idealised fragment libraries before being refined under the influence of extensive geometric restraints generated by Grade Web Server (DFIX, DANG, FLAT). Anti-bumping restraints have been applied to the nitrogen-cadmium distances for N31B and N21A (DFIX -2.3) and a target distance of 2.45 Å has been applied to the Cd1-N11A distance (DFIX). Geometric similarity restraints have been applied to the 1,3 and 1,4 distances around the trismethylene-mesityl moiety to maintain three-fold symmetrical geometry (SADI).
The anisotropic displacement parameters of the carbon and nitrogen atoms are restrained to be similar (SIMU) and to have a rigid bond approximation (RIGU). The anisotropic displacement parameters of the carbon atoms of the mesityl moiety are restrained to have more isotropic character (ISOR). The anisotropic displacement parameters are universally very large likely as a result of the poor crystallinity of the macromolecule species surrounded by large regions of diffuse solvent and anions. Water residue O1W is refined with an isotropic displacement parameter.
All hydrogen atoms were geometrically placed and refined with a riding model. Hydrogen atoms of water residue O1W were not observed in the electron density map; they are not included in the model, however, they are included in the unit cell contents and all values derived from it.
PLATON SQUEEZE was applied to the data to remove scattering contributions from several disordered anion and solvent residues which could not be modelled as discrete sites: 11121 electrons per unit cell were consistent with two tetrafluoroborate anions and 6.67 acetonitrile solvent molecules per asymmetric unit. These molecules have been included in the chemical formula and in all values derived from it. SQUEEZE also yielded a set of solvent-free diffraction intensities for use in the final cycles of refinement.

AACC-C2c_sq
The X-ray diffraction intensity of this large supramolecular assembly featuring large regions of diffuse anions and solvent residues dropped off rapidly at higher angles. The data used in the refinement was truncated to an upper resolution of 1.15 Å. As a result of the low-resolution diffraction limit, the refinement has a data to parameter ratio of 7.25:1. Extensive use of restraints on geometric and displacement parameters aided the refinement. Only atoms of the central metallo-cage residue were refined with isotropic displacement parameters.
The 1,2 and 1,3 distances in the pyrazolylpyridine moieties, naphthyl moieties and tetrafluoroborate anions were restrained to be similar (SAME). The 1,2 bond distances of the methylene moieties to pyrazolyl and naphthyl moieties were respectively restrained to be similar (SADI). The geometries of several pyridine, pyrazole and naphthyl ring systems were restrained to be planar (FLAT). All of the 1,2 bond distances in the pyrazolylpyridine moieties were restrained to have suitable target values taken from averages of equivalent moieties in the CSD (DFIX).
The anisotropic displacement parameters of the carbon, nitrogen, boron and fluorine atoms were restrained to be similar (SIMU) and to have a rigid bond approximation (RIGU). The tetrafluoroborate and water residues are refined with isotropic displacement parameters. The isotropic displacement parameters of all water residues were fixed at a value of 0.2. The isotropic displacement parameters of tetrafluoroborate residues B11, B31, B41, B71 and B81 were fixed at a value of 0.25. The occupancies of tetrafluoroborate and water residues were allowed to refine, using a common free variable in the case of each tetrafluoroborate residue. Tetrafluoroborate residues B11/B81 and water residues O1W/O1X are disordered pairs respectively; their refined occupancies are constrained to sum to unity.
Some small residual electron density peaks close to water residues indicate that they might be substitutionally disordered with acetonitrile solvent residues (crystallisation solvent). No sensible model could be developed for any such disorder sites. All hydrogen atoms were geometrically placed and refined with a riding model. Hydrogen atoms of water residues were not observed in the electron density map; they are not included in the model, however, they are included in the unit cell contents and all values derived from it. Some short H...H intramolecular contacts are observed between mesitylmethyl and methylene hydrogen atoms. These are caused use of idealised methyl hydrogen positions rather than positions derived from refinement of the torsion angle for which the data quality was insufficient. The clashing methyl hydrogen atoms are retained in the model rather than omitted to give the model more realistic cumulative scattering strength.
PLATON SQUEEZE was applied to the data to remove scattering contributions from several disordered anion and solvent residues which could not be modelled as discrete sites: 4335 electrons per P1 unit cell and a volume of 19877 Å 3 were equated with seven tetrafluoroborate anions, twelve acetonitrile solvent molecules and six water molecules per asymmetric unit. These molecules have been included in the chemical formula and in all values derived from it. SQUEEZE also yielded a set of solvent-free diffraction intensities for use in the final cycles of refinement.

AAAC-P21n_sq
The weakly diffracting crystals of this large supramolecular metallo-cage features large regions of poorly crystalline anion, solvent and water molecules and as a result, had a low-resolution diffraction limit. The data used for the refinement were truncated to a resolution of 1.1 Å resulting in a low data to parameter ratio of 8.3:1. Extensive use of restraints on geometric and displacement parameters aided the refinement. Only atoms of the central metallo-cage residue were refined with isotropic displacement parameters.
The 1,2 and 1,3 distances in the pyrazolylpyridine moieties, naphthyl moieties, diethyl ether solvent residues and tetrafluoroborate anions were restrained to be similar (SAME). The 1,2 bond distances of the methylene moieties to pyrazolyl and naphthyl moieties were respectively restrained to be target values (DFIX). The geometries of all pyridine, pyrazole and naphthyl ring systems were restrained to be planar (FLAT). All acetonitrile solvent residues were refined as rigid bodies with idealised coordinates taken from a fragment library (Guzei, I. A. J. Appl. Crystallogr. 2014, 47, 806-809). Hydrogen atoms were omitted for water residues and some acetonitrile residues where the hydrogen torsion angle did not converge.
The atoms of the metallo-cage complex were refined with anisotropic displacement parameters; rigid bond and similarity restraints were applied to the parameters of the carbon and nitrogen atoms (RIGU, SIMU). All other atoms in the structure (solvent residues and tetrafluorobrate anions) were refined with isotropic displacement parameters. To reduce the parameter burden on the refinement a common value for the isotropic displacement parameters was refined for the non-hydrogen atoms of each acetonitrile residue and tetrafluoroborate anion. The occupancies of tetrafluoroborate anions B31, B111, B171 and B191 were refined, whilst the occupancy of B101 was fixed at a half. The occupancies of acetonitrile residues Q, Y and X were refined. The isotropic displacement parameters of water residues 1-9 and 21-29 were fixed at a value of 0.25. The occupancies of water residues 3, 6, 7, 27, 28, and 29 were refined.
Some small residual electron density peaks close to water residues indicate that they might be substitutionally disordered with acetonitrile solvent residues (crystallisation solvent). No sensible model could be developed for any such disorder sites.
All hydrogen atoms were geometrically placed and refined with a riding model. Hydrogen atoms of water residues were not observed in the electron density map; they are not included in the model, however, they are included in the unit cell contents and all values derived from it. PLATON SQUEEZE was applied to the data to remove scattering contributions from several disordered anion and solvent residues which could not be modelled as discrete sites: 3675 electrons per P1 unit cell and a volume of 14,249 Å 3 were equated with six tetrafluoroborate anions and 30 acetonitrile solvent molecules per asymmetric unit. These molecules have been included in the chemical formula and in all values derived from it. SQUEEZE also yielded a set of solvent-free diffraction intensities for use in the final cycles of refinement.

Geometric analysis of the cages
Cavity volumes inside the crystallised cage complexes were determined using the SOLV routine in PLATON. Input files were prepared from the final refined CIFs by manually removing solvent and anion residues from inside the cage cavities and blocking the cage apertures with artificial dummyatoms (sulphur or iodine).
The geometric volume enclosed the cuboctahedral array of twelve cadmium atoms in each crystal structure was determined using the alphaShape algorithm in the software MATLAB online R2020a. The coordinates of the metal cations for a complete cage moiety were exported from Olex2 v1.3 as an XYZ file before being defined as a 12x3 matrix A in MATLAB: The coordinates were defined as an alpha shape:

CUBO = alphaShape(A)
An alpha spectrum was determined to find the critical alpha radius (largest value listed in the alpha spectrum) necessary to generate a convex alpha shape: CUBO.Alpha = 13 The cuboctahedron was plotted to ensure all vertices were appropriately connected and the volume determined:

NMR Spectroscopy
High field 1 H and 113 Cd NMR spectra were measured on a Bruker Avance III 600 MHz spectrometer. 1 H-113 Cd correlation spectra were recorded using a variant of the HMBC pulse sequence and the 113 Cd chemical shifts reported are taken from these spectra (referenced to external 0.1 M [Cd(ClO) 4 ) 2 ]). The spectra of [Cd 12 (L naph ) 12 (L mes ) 4 ](BF 4 ) 24 (mixture of isomers) were measured in CD 3 CN solution. The 113 Cd spectrum is a 1D Projection of the F2 dimension taken from the 8.51-6.36 ppm region of the 1H dimension of the 600 MHz 1 H-113 Cd HMBC spectrum. The full spectrum is dominated by cross peaks resulting from the large quantity of residual solvent. The broad peaks of overlapping 113 Cd environments preclude a full deconvolution, however, the spectrum is consistent with the expected presence of eight environments (see main text).