Metallophilicity-assisted Assembly of Phosphine-based Cage Molecules

Figure S1. ORTEP view of the dication 1 (counterions and hydrogen atoms are omitted for clarity; ellipsoids are shown at 50% probability).


Introduction
Remarkable progress in supramolecular coordination chemistry has been observed during the past two decades.Using a combination of discrete metal coordination geometry and rationally designed organic ligands, an impressively large number of finite supramolecular assemblies and extended inorganic polymers showing unique physical and chemical properties were successfully prepared via a self-assembly approach. 1,2By varying the nature of metal ion and tuning the stereochemistry of the organic linker, one can control the formation of fascinating polygonal 2D and polyhedral 3D aggregates of regular architectures.The latter type of compounds is particularly attractive as they possess well-defined nanosized hollows and can efficiently serve as molecular containers demonstrating rich host-guest chemistry, e.g.selective molecular recognition, 3 stabilization of the reactive or unstable species, 4,5 unconventional catalysis, 2,6 photochemistry 4 and molecular transportation including drug delivery. 7ost of the work dedicated to the metal-organic cage complexes is based on the mononuclear coordinating centers (i.e.metal ions) and polydentate hard N and/or O donor ligands of suitable geometry. 1,2The general synthetic methodology is based on so-called coordination-driven self-assembly that involves spontaneous formation of the metal-ligand bonds upon reacting the ligands with coordinatively unsaturated metal precursors or those having easy to substitute groups (e.g.weakly bound solvent molecules, counterions, labile ligands like alkenes, nitriles, etc.).
The soft P donor ligands have been extensively used in coordination chemistry of late transition metals, but their utilization for the construction of supramolecular 2D/3D molecules is quite uncommon in comparison to the N, O building blocks.Some homo-and heteroleptic diphosphine-based coinage metal rings, which mainly contain the bridging ligands with flexible backbones, were reported in the literature. 8The use of rigid multidentate phosphine ligands allowed for the preparation of certain interesting hollow aggregates of copper subgroup metals, 9 such as adamantanoid, 10 bowlshaped, 11 nanotubular 12 and tetrahedral 13 coordination clusters.Despite the well-reported ability of d 10 metal ions and of Au(I) in particular to form extensive metal-metal bonding, 14 this phenomenon is scarcely considered as a potential directing force in constructing three-dimensional cage ensembles.
We described earlier the assembly of gold(I)-diphosphine cages built of the planar tetragold Au 4 coordinating centers 15 that was one of the first examples where polynuclear metal clusters were used in the construction of 3D hollow complexes. 16Herein we report on the further development of metallophilicity-assisted cage compound synthesis using the combination of Au 3 cluster units together with di-and tridentate stereochemically rigid phosphine ligands.

Synthesis of the complexes [(Bu t
Step B. The solid obtained was dissolved in acetone (10 cm 3 ) and treated with H 2 O (4 drops) and NEt 3 (2 drops) to give a flaky white solid formulated as [(OAu 3 ) 2 PP 3 ](PF 6 ) 2 .The suspension was stirred for 20 min in the absence of light, and then it was evaporated; the solid product was washed with methanol (2 × 3 cm 3 ) and diethyl ether (2 × 3 cm 3 ) and dried to give [(OAu 3 ) 2 PP 3 ](PF 6 ) 2 as a light beige solid, which was used in the next stage without purification.

X-ray structural determination
Single crystals of 1 and 4 were immersed in cryo-oil, mounted on a Nylon loop, and measured at the temperatures of 120 K and 210 K, respectively.The X-ray diffraction data were collected on Bruker Smart Apex II and Bruker Kappa Apex II Duo diffractometers using Mo Kα radiation.The APEX2 21 program package was used for cell refinements and data reductions.The structures were solved by direct methods using the SHELXS-2013 22 programs with the WinGX 23 graphical user interface and the SHELXS-97 program 22 incorporated into the OLEX2 program package. 24A semi-empirical absorption correction (SADABS) 25 was applied to all data.Structural refinements were carried out using SHELXS-97 and SHELXL-2013. 22One of the phenyl rings in 1 was disordered over two positions (C38-C43, C138-C143) and was refined with occupancies 0.66/0.34.The displacement parameters of the carbon atoms of both components were constrained to be equal.The aromatic ring C138-C143 was geometrically idealized.Some of the solvent molecules in the unit cells of 1 and 4 were omitted as they were disordered and could not be resolved unambiguously.The missing solvent was taken into account using a SQUEEZE routine of PLATON 26 and was not included in the cell content.The H 2 O hydrogens in 1 were positioned manually and were constrained to ride on their parent atoms O1 and O2, with U iso = 1.5U eq ( parent atom).All other hydrogen atoms were positioned geometrically and constrained to ride on their parent atoms, with C-H = 0.95-0.99Å, U iso = 1.2-1.5Ueq ( parent atom).The crystallographic details are summarized in Table 1.

Computational details
The gold-phosphine cage compounds 1-4 were studied using the hybrid PBE0 density functional method combined with Grimme's D3 dispersion correction (DFT-PBE0-D). 27The gold atoms were described by a triple-zeta-valence quality basis set with polarization functions (def2-TZVP). 28Scalar relativistic effects were taken into account by applying a 60-electron relativistic effective core potential for Au. 29 A split-valence basis set with polarization functions on non-hydrogen atoms was used for all the other atoms. 30All electronic structure calculations were carried out with the TURBOMOLE program package (version 6.4). 31
A key point in the preparation of 3D finite structures is an appropriate choice of a ligand X which can stabilize a gold(I) cluster coordinating center.Among these X moieties, reported in the literature for a number of clusters [Р n Au n (μ n -X)] m+ (n = 3, 4; P = phosphine; X = O, S, Se, As, CR, NR, PR, R 3 P:B), [32][33][34] the fragments R 3 P:B 3− , PR 2− , NR 2− , CR 3− seem to be the most promising templates as they can be introduced into the cage structures under mild conditions using the main group element nucleophiles (NH 2 R, PH 2 R, HCuCR) and simultaneously prevent the formation of intermolecular {Au n }⋯ {Au n } bonding in the solid state, often found for chalcogenido complexes. 33s we reported earlier, terminal alkynes can be easily converted into the μ 4 -methylydine group under basic conditions in the presence of diphosphine-gold cationic species, to give the stable cage complexes (Scheme 1A), which exist not only in the solid state, but retain their structures in solution. 15owever, application of this protocol to the star-shaped triphosphines, e.g.1,3,5,-tris(4-diphenylphosphinophenyl)benzene, did not bring positive results leading to insoluble material that was difficult to characterize.In order to expand this concept we investigated a possibility to use the imido μ 3 -NR 2− bridging groups for the cage assembly (exemplified by R = Bu t , Scheme 1B and C).

Structural analysis
The structures of the cages 1 and 4 in the solid state were determined by an X-ray diffraction analysis (Fig. 1 and   distances in these complexes lie in the range 2.9426(10)-3.1480(10) Å and are in good agreement with those found in the other trigold-imido clusters of the [(PR 3 Au) 3 (μ 3 -NR)] + type (2.926-3.333Å). 34,36 In the solid state the cage 4 was found to host a disordered dichloromethane crystallization molecule.According to the elemental analysis data, the CH 2 Cl 2 guest can be easily removed under vacuum from the crystalline sample.The ESI-MS (see below) did not show any appreciable signs of solvent inclusion under the conditions of mass-spectroscopic experiment.

Spectroscopic characterization
In solution the complexes 1-4 were studied by 1 H and 31 P NMR spectroscopy.The positive ion mode ESI-MS data confirm that the clusters retain their composition in solution showing the signals of the dications 1-3 with characteristic isotopic distributions at m/z 1331, 1445 and 1559, respectively (Fig. S3 †).The mass-spectrum of 4 as well displays a dominant peak of the quadruply charged ion at m/z 1520.76 that exactly matches the stoichiometry of the intact [(PPPAu 3 ) 4 (μ 3 -NBu t ) 4 ] 4+ cation (Fig. 3).
The 31 P NMR spectra of 1-4 show singlet resonances in a narrow range of δ (27.9-28.7)that is indicative of all equivalent phosphorus atoms coordinated to Au ions and is in agreement with the solid state structures.The 1 H NMR of 1-3 displays the spectroscopic patterns, which correspond to the idealized D 3h point symmetry group that is, however, higher than that found in the crystal of 1 (D 3 ).All the signals of the ortho-meta-para protons of the -PPh 2 groups represent a single set of resonances that points to the equivalence of the phenyl rings.The protons of the PP spacers -(C 6 H 4 ) n -, n = 1 (1), 2 (2), 3 (3), also give rise to the groups of signals which fit well the D 3h symmetry of the molecules.These spectroscopic patterns can be explained in terms of the fast (on the NMR timescale) rightleft twisting of the "Au 6 PP 3 " framework that results in a flip motion of the helical isomers P↔M, which eventually leads to equivalence of the phenyl rings and increases the molecular symmetry in comparison with the crystalline state (Scheme 2). 15he solution behavior of the tetrahedral cage 4 is somewhat different from that of 1-3.The idealized symmetry of the molecule in the solid state corresponds to T point group.The architecture of the complex that exhibits axial chirality makes the triphosphine phenyl rings non-equivalent in every -PPh 2 fragment.In contrast to 1-3, which display a fast P↔M equilibrium, cluster 4 retains its configuration in solution according to the 1 H NMR data, as indicated by two clearly distinguishable sets of resonances corresponding to non-equivalent Ph protons in the phosphorus atoms environment (Fig. 4).One unresolved group of signals is found in a narrow region from 7.57 to 7.80 ppm and contains all the resonances of ortho, meta, and para-H atoms.In another group of Ph ring signals there are well resolved resonances at 7.77, 7.42 and 7.24 ppm corresponding to ortho, para and meta-H, respectively.This  (10), Au(1)-Au(3) 3.1480 (10), Au(2)-Au(3) 3.0538 (10), Au(4)-Au(5) 3.0244 (10), Au(4)-Au(6) 3.0272 (10), Au(5)-Au(6) 3.1446 (11).Symmetry transformations used to generate equivalent atoms: (') 1 − x, y, 0.5 − z. observation clearly points to high robustness of the tetrahedral cage and the absence of intramolecular dynamics found in the other tetrahedral complexes built of the mononuclear metal centers. 37

Computational results
We elucidated the structural characteristics of the gold-phosphine cages 1-4 by means of quantum chemical calculations at the DFT-PBE0-D level of theory (see the Experimental section for the details).First, the geometries of the cages were fully optimized using the ideal point group symmetries derived from the solid state structures of 1 and 4 (1-3: D 3 ; 4: T ).We also carried out harmonic frequency calculations to confirm that the optimized structures are true local minima.The optimized geometries are illustrated in Fig. 5.
Even though the DFT calculations were carried out in the gas phase, the resulting geometries for 1 and 4 are in very good agreement with the respective solid state structures.For 1, the Au-Au distances in the X-ray structure are 2.96-3.07Å, while the DFT-optimized distance is 3.09 Å.The N-N distance describing the overall dimensions of the cage is 10.7 Å in the X-ray structure and 10.4 Å in the optimized structure.Various bond distances such as Au-P and Au-N are reproduced with good accuracy (differences <0.05 Å).For 4, the Au-Au distances in the X-ray structure are 2.94-3.15Å and the DFT-optimized distance of 3.06 Å is very close to the average Au-Au contact in the solid state structure.The N-N distances in the X-ray structure are 15.5-17.3Å and also here the respective value for the DFT-optimized structure (16.7 Å) is practically similar to the  average N-N separation in the X-ray structure.In the case of the cages 2 and 3, where no X-ray structure is available, the Au-Au distances are very similar to 1 (3.06 Å).The N-N distances of 2 and 3 are 14.8 and 19.3 Å, respectively.Including the D3 dispersion corrections in the structural optimizations clearly improved the agreement between the X-ray and DFToptimized structures with respect to non-dispersion corrected results.
Since the applied computational method describes the structural characteristics of the gold-phosphine cages very well, we also investigated the difference in solution behavior of the cylindrical cages 1-3 and the tetrahedral cage 4. As discussed above, in solution the cages 1-3 display NMR patterns that correspond to the ideal D 3h point group symmetry, while the cage 4 retains the T-symmetric structure, instead of showing an NMR pattern corresponding to the ideal T d point group.This difference between the two types of cages arises from the twisting-type interconversion of the gold-phosphine framework, which occurs in the case of 1-3, but not in the case of 4. We investigated the energetic barriers for the twisting-type interconversion for all four clusters.During the twisting, the cages 1-3 pass through a D 3h -symmetric transition state, where the capping gold triangles are in an eclipsed conformation instead of the staggered one in the D 3 -symmetry (see Scheme 2).Similarly, the cage 4 should pass through a T dsymmetric transition state when twisting from one T-symmetric minimum to another.We optimized the D 3h -and T dsymmetric transition states by means of a constrained optimization where the positions of the P atoms were fixed (if no atoms are fixed, the cages will revert back to D 3 or T symmetry during the optimization).The energy barriers for the D 3 → D 3h → D 3 conversion in the cages 1-3 turned out to be significantly smaller than for the T → T d → T conversion in the cage 4. The twisting barrier is the lowest for the cage 1, where it is 82 kJ mol −1 .The true twisting barriers in solution are expected to be lower than this since the transition states here are obtained from constrained optimization, but the relative twisting barriers of the cages show a very clear trend.The twisting barriers of the cages 2 and 3 relative to the cage 1 are only 24 and 3 kJ mol −1 higher (cage 3 is slightly more flexible than cage 2 due to the longer phenyl spacer).For comparison, the twisting barrier in cage 4 relative to the cage 1 is an order of magnitude higher, 230 kJ mol −1 .Hence, the difference in solution behavior of 1-3 and 4 can be attributed to the significantly higher energy barrier for the twisting-type interconversion in the cage 4.

Conclusions
In summary, we have demonstrated a so far rare possibility of using the phosphine ligands and polynuclear coordination centers for the effective construction of supramolecular cage molecules.The synthetic approach is based on aurophilicitydriven aggregation of the Au(I) ions into the small Au 3 clusters stabilized by the μ 3 -NR 2− bridging imido groups.Conducting this process in the presence of the di-or triphosphine ligands P n (n = 2, 3) of suitable stereochemistry results in self-assembly of the coordination precursors P n (AuS*) n n+ (S* = labile ligand) into the finite 3D structures.Depending on the denticity of the phosphines used the tubular-like cage clusters [(P 2 Au 2 ) 3 (μ 3 -NBu t ) 2 ] 2+ (1-3) and the tetrahedral complex [(P 3 Au 3 ) 4 (μ 3 -NBu t ) 4 ] 4+ (4) were isolated.All the compounds under study retain their composition in solution according to the NMR and ESI-MS data.The cylindrical cages 1-3 were shown to undergo fast interconversion of the helical P↔M isomers, while the architecture of the tetrahedron 4 having axial chirality found in the crystalline state remains intact in the fluid medium.The computational studies of the geometries of the complexes and of the twisting-type dynamic behavior are in good agreement with experimental observations suggesting a significantly higher energy barrier for the P↔M isomerization in the cage 4.

Fig. 4
Fig. 4 1 H-1 H COSY NMR (aromatic region) spectrum of 4, acetone-d 6 , 298 K; the inset shows schematic representation of the non-equivalent phenyl rings in the molecule; the assignment of the signals to the -PPh 2 phenyl rings is arbitrary.

Fig. 5
Fig. 5 Optimized structures of the gold-phosphine cages 1-4.The cylinders and the tetrahedron demonstrate the hollow cavities within the cages.Hydrogen atoms have been left out for clarity.

Table 1
Crystal data for 1 and 4