Tetrameric and Dimeric [N∙∙∙I + ∙∙∙N] Halogen-Bonded Supramolecular Cages

: Tripodal N -donor ligands are used to form halogen-bonded assemblies via structurally analogous Ag + -complexes. Selective formation of discrete tetrameric I 6 L 4 and dimeric I 3 L 2 halonium cages, wherein multiple [N∙∙∙I + ∙∙∙N] halogen bonds are used in concert, can be achieved by using sterically rigidified cationic tris(1-methyl-1-azonia-4-azabicyclo[2.2.2]octane)-mesitylene ligand, L1 (PF 6 ) 3 , and flexible ligand 1,3,5-tris(imidazole-1-ylmethyl)-2,4,6-trimethylbenzene, L2 , respectively. The iodonium cages, I 6 L1 4 (PF 6 ) 18 and I 3 L2 2 (PF 6 ) 3 , were obtained through the [N∙∙∙Ag + ∙∙∙N] → [N∙∙∙I + ∙∙∙N] cation exchange reaction between the corresponding Ag 6 L1 4 (PF 6 ) 18 and Ag 3 L2 2 (PF 6 ) 3 coordination cages, prepared as intermediates, and I 2 . The synthesized metallo- and halonium cages were studied in solution by NMR, in gas phase by ESI-MS and in the solid-state by single crystal X-ray diffraction.


Introduction
Halogen bonding (XB) is a non-covalent interaction between a polarized halogen atom and a Lewis base. [1] During the last decade XB has found applications in variety of research fields including crystal engineering, material sciences, medicinal chemistry and organocatalysis. [2] Halonium ions are interesting class of XB donors which are able to form three-center-fourelectron (3c4e) bonds with two XB acceptor moieties. [3,4] Behavior of halonium ions in bis-pyridine complexes have been carefully studied in solution and in the solid-state. [3,[5][6][7] Recently, we reported syntheses and characterizations of dimeric and hexameric capsules solely based on [N•••I + •••N] bonds. [8] It has been shown that [N•••I + •••N] bond is a strong interaction that exhibits highly predictable bond parameters. [4] Hence, it is particularly well-suited for the self-assembly studies of halogenbonded supramolecular species/complexes. Among different supramolecular assemblies coordination cages, as metallosupramolecular complexes, have attracted particular attention during the last two decades. The keen interest toward coordination cages is partly due to their fascinating structural features, but also because of the increase in number of potential applications that exploit their well-defined cavities. [9][10][11][12][13] These systems are prepared by coordination-driven self-assembly from spatially pre-organized organic ligands (L) and metal cations (M) with known coordination geometry. One of the commonly used group of ligands is bowl-shaped tripodal N-donor ligands which enables the construction of various types of coordination cages such as dimeric M3L2, tetrahedral M6L4 and octahedral M6L8 species. [14,15] The exact cage topology is mostly governed by the connectivity and coordination geometry of the metal ion as well as the structural and chemical properties of the ligand. Furthermore, counter anions, reaction conditions (e.g. solvent and temperature) as well as possible auxiliary ligands can all have an effect to the outcome in the selfassembly process.
We have been interested in investigating the behavior of the [N•••I + •••N] halogen bond in connection with spatially organized multitopic XB acceptors. 8 In this work we extend this approach to two different tripodal XB acceptor ligands, L1(PF6)3 and L2 (Scheme 1), both of which have previously been used in the construction of coordination cages and are easily prepared according to the reported procedures. [16,17] The ligand L2 is known to adapt different geometries due to the flexibility of the imidazole arms and hence different topologies including cages and polymeric structures with different metals and counter ions have been reported. [18][19][20][21] On the other hand, the self-assembly of L1(PF6)3 with metal cations is to a great extent affected by the intra-and intermolecular steric interactions of the bulky DABCO groups which stabilize the bowl-shaped all-cis conformation of the ligand and restrict the spatial orientation of the N-donor moieties. Recently, we reported Pd 2+ and Cu 2+ based cages of L1 3+ . [22] Their self-assembly was shown to result in tetrahedral M6L14 species instead of the expected M6L8 due to the steric interactions between the DABCO groups around the metal node leading to linear N-M-N coordination. The corresponding Ag6L14(PF6) 18

Results and Discussion
Syntheses of the [N•••I + •••N] bonded capsules follow the previously reported procedures. [8] The halonium cages are obtained first by reacting L1(PF6)3 and L2 with silver hexafluorophosphate (AgPF6) followed by the addition of molecular iodine. The self-assembly of Ag + and I + species was monitored by 1 H NMR in CD3CN ( Figure S1). Somewhat surprisingly, the addition of AgPF6 into the L1(PF6)3 in CD3CN solution did not result in noticeable shifting of the L1(PF6)3 proton signals. This may be due to solvation of Ag + as a [Ag(acetonitrile)4] + complex which seems to be energetically favored over the [Ag6L14] 18+ in acetonitrile. However, after the addition of I2, which was accompanied by an immediate precipitation of AgI and color change of the initially colorless solution to yellow (to mark the presence of slight excess of I2), the 1 H NMR spectrum reveals a new set of downfield shifted L1(PF6)3 signals together with minor peaks of unreacted L1(PF6)3 ( Figure S1c). Specifically, the merging of the two triplets that arise from the asymmetric DABCO environment into one peak at 3.6 ppm is due to the formation of the six symmetric [N•••I + •••N] halogen bonds.
To provide more evidence for the expected formation of the [I6L14] 18+ cage, diffusion-ordered NMR spectroscopy (DOSY) measurements were performed. Two set of signals were observed in the 1 H DOSY spectrum of the expected I6L14(PF6) 18 indicating the presence of two discrete species ( Figure S2). These were with diffusion coefficient (D) of 1.06 x 10 -9 m 2 •s -1 and considerable lower value of 6.09 x 10 -10 m 2 •s -1 (CD3CN, 298 K), which correspond to a mixture of unreacted L1(PF6)3 and a definitely larger species with radius of 1.05 nm (diameter of 2.1 nm) according to the Stokes-Einstein equation. The size correlates very well to our previous observations of Cu 2+ and Pd 2+ complexes with M6L14 stoichiometry (size derived from the single crystal X-ray structure of Cu6L14(PF6)24). [22] Unfortunately, our efforts to characterize either Ag6L14(PF6)18 or I6L14(PF6)18 further with ESI-MS were not successful. This is not surprising considering the high positive charge density (+18) of the assemblies. In gas phase, the stabilizing solvent and anion interactions are absent and close proximity of the silver and iodonium cations with the cationic ligand likely results in repulsion and fragmentation of the cages.
The penultimate way to investigate the nature of the formed species is to crystallize both Ag6L14(PF6)18 and I6L14(PF6)18.
Crystals emerged from a methanol solution of AgPF6 and L1(PF6)3 upon slow diffusion with ethyl acetate. Although no complexation between Ag + and L1(PF6)3 was observed in solution by NMR, the single crystal X-ray analysis revealed the formation of a Ag6L14(PF6)18 as a 2 nm cage (measured from the far-cornered H-atoms, Figure 1). The metallocage crystallized in a high-symmetry space group (tetragonal I-42m) and is similarly two-fold disordered as the Cu6L14(PF6)24 tetrameric cage. [22] The six Ag(I) ions and four ligands form a total of six [N•••Ag + •••N] coordination environments with normal Ag-N bond distances and nearly linear to slightly bent N-Ag + -N bond angles, thus being structurally analogous to the Cu6L14(PF6)24. [22]  assemblies. [8] Despite coordinative differences, the I6L14(PF6)18 binds the PF6anions similarly as its metallosupramolecular analogues Ag6L14(PF6)18 and Cu6L14(PF6)24 [22] . Four out of the 18 PF6anions reside inside the endohedral cavity of the halonium cage, each occupying one of the four cationic ligand pockets. This anion binding arises primarily from the electrostatic interactions between the cationic L1 3+ ligand and the anions. As a structurally similar tripodal supramolecular building block L2 has two significant differences to L1. Firstly, L2 is noncharged and secondly has greater structural flexibility, meaning that by using different metal cations, counteranions, reaction conditions and metal-ligand stoichiometry, cages and polymeric structures are plausible. [18][19][20][21] Of these the I3L22X3 is expected to be prominent as several Ag3L22X3 complexes (where X = N3 -, NO3 -, ClO4 -, BF4 -OTfor PF6 -) have been observed under similar conditions. [19,21,23,24] Contrary to highly-charged L1, L2 does not impose additional electrostatic interactions and can lead to less-disturbed stronger [N•••I + •••N] halogen bonds.
The consecutive reactions of AgPF6 and I2 with L2 in CD3CN showed a shifted set of signals in 1 H NMR spectrum in both cases. Comparison between the two spectra shows that I + induces a larger downfield shift to the ligand signals than Ag + (Figure 3). The 1 H DOSY measurements (Figures 3d and S3) revealed single species for both Ag + and I + complexes with D of 8.46 x 10 -10 m 2 •s -1 and 7.86 x 10 -10 m 2 •s -1 in acetonitrile at 298K, respectively. The measured D values correspond to spherical objects with diameters of 1.52 nm and 1.63 nm whereas the corresponding estimated [18,19,21,23,24] values for M3L22 3+ and M6L24 6+ are roughly 1.3 -1.4 nm and 1.65 -1.9 nm, respectively. Hence, no conclusive analysis on the compositions can be drawn from the DOSY data alone. Instead, the composition was verified by ESI-MS analysis, in which only the formation of dimeric complex was detected and ions [I3L22] 3+ at m/z 367 and [I3L22+PF6] 2+ at m/z 623 appeared in spectra ( Figure S4 and Table S1).   When comparing the two structurally analogous cages, Ag3L22(PF6)3 and I3L22(PF6)3, it becomes apparent that they show structural similarities but also differ significantly. The coordination environment of Ag + in Ag3L22(PF6)3 has clearly shorter Ag-N bonds [2.110(5) Å] compared to the I-N bonds [2.223(6) Å] in I3L22(PF6)3, while the N-Ag-N angles are very close to linear [178.2(4)°] being remarkably similar with I3L22(PF6)3. The N-Ag-N environment is known to be easily distorted away from linearity by coordinating counteranions as demonstrated, for example, by the nitrate and azide analogues of Ag3L22(PF6)3. [19,23] On the contrary, the geometry of the [N•••I + •••N] unit is expected to be mostly undisturbed by counteranions as the contact surface of the I + cation, perpendicular to the [N•••I + •••N] halogen bonds, is not electrophilic but rather charge-neutral or even nucleophilic. This is attributable to the px 2 py 2 pz 0 valence p-orbital occupancy of the iodonium cation. [26] The different charge distributions of the Ag + and I + cations is reflected also in the host-guest chemistry of the respective supramolecular M3L22 3+ cages. In the solid state Ag3L22 3+ clearly encapsulates one PF6anion within its endohedral cavity. The PF6is severely disordered over three major symmetry-related positions which were analyzed to reside at the Ag-Ag-Ag plane and being stabilized by van der Waals contacts between fluorine atoms and Ag + (d(F•••Ag) ≈ 3.0 -3.1 Å). This analysis is supported by an earlier study of a hydrate structure of Ag3L22(PF6)3 in which the equatorial fluorine atoms of an encapsulated (non-disordered) PF6anion reside at the Ag-Ag-Ag plane within 2.8 -3.1 Å away from the Ag + cations. [24] In the case of I3L22(PF6)3 the residual electron density suggests that one PF6anion occupies the endohedral cavity and is disordered over two major positions along the C3 axis of the host, perpendicular to the I-I-I plane (see residual electron density maps of non-masked refinements of Ag3L22(PF6)3 and I3L22(PF6)3 in figure S5). This is reasonable when considering the cationic nature of the halonium cage which gives the impetus for the encapsulation of the anionic guest whereas the repulsive nature of the fluorine•••I + interaction perpendicular to the N-I-N bonds pushes the anion away from the I-I-I plane. Hence, although the anion within the cavity of I3L22(PF6)3 could not be well resolved, the results suggest that the anisotropic charge distribution of the I + cation has a directing effect in the host-guest chemistry of the respective supramolecular assemblies. This phenomenon could be exploited in tuning of the interior of a supramolecular host and thus gaining target-specificity toward e.g. charge-depleted guest species.

Conclusions
This work presents the use of two structurally different N-donor tripodal ligands L1(PF6)3 and L2 in the formation of discrete cationic supramolecular halonium cages I6L14 18+ and I3L22 3+ from their parent metallosupramolecular cages Ag6L14 18+ and Ag3L22 3+ , respectively, by reaction with molecular iodine. In addition to comprehensive NMR spectroscopic and ESI-MS spectrometric evidence, the [N•••I + •••N] halogen-bonded cages were structurally characterized for the first time at an atomic resolution by single crystal X-ray crystallography. The NMR spectroscopic results illustrated the difficulty of the cationic and sterically challenging ligand L1(PF6) 3

Experimental Section
Ligands L1(PF6)3 and L2 were synthesized according to the reported procedures. [16,17] NMR spectra were recorded on a Bruker Avance III 500 or Avance 400 spectrometers. All signals are given as δ values in ppm using residual solvent signals as the internal standard. ESI-MS experiments were performed on AB Sciex QSTAR Elite ESI-Q-TOF mass spectrometer equipped with an API 200 TurboIonSpray ESI source from AB Sciex (former MDS Sciex) in Concord, Ontario (Canada). Single crystal X-ray data were collected with either Agilent SuperNova, equipped with multilayer optics monochromated dual source (Cu and Mo) and Atlas detector, using Cu Kα (1.54184 Å) radiation or Bruker-Nonius/Kappa CCD diffractometer using Mo Kα (0.71073 Å) radiation. In the latter case, data acquisitions, reductions and absorption corrections were made using programs COLLECT [27] , DENZO-SMN [28] and SADABS [29] whereas the data collected with Agilent SuperNova was treated within the CrysAlis Pro program package [30] . The structures were solved with ShelXS [31] program and refined on F 2 by full matrix least squares techniques with ShelXL [32] program in Olex 2 (v.1.2) [33] program package. Anisotropical displacement parameters were applied for all atoms except hydrogens which were calculated into their ideal positions using isotropic displacement parameters 1.2-1.5 times of the host atom. Olex 2 solvent mask routine was used to treat the residual electron density corresponding to disordered anions and solvent molecules in the crystal lattices. Details of solvent mask procedure together with crystallographic data table can be found in ESI (table S2).