Supramolecular Metallacycles and Their Binding of Fullerenes

Abstract The synthesis of a new triaminoguanidinium‐based ligand with three tris‐chelating [NNO]‐binding pockets and C 3 symmetry is described. The reaction of tris‐(2‐pyridinylene‐N‐oxide)triaminoguanidinium salts with zinc(II) formate leads to the formation of cyclic supramolecular coordination compounds which in solution bind fullerenes in their spherical cavities. The rapid encapsulation of C60 can be observed by NMR spectroscopy and single‐crystal X‐ray diffraction and is verified using computation.


Introduction
The development of supramolecular coordination compounds and their correspondingc ages has attracted wide interest in recent years. [1] They have the potentialt ot ransport chemicals from one locationt oanotherinaspecific manner,for example, being used in drugd elivery or as contrast agents. [2] The structural design of these containers requires precise and complementaryb uilding blocks butt hey are not limitedt ot he incorporation of smallo rganic molecules. [3] There are only af ew exampleso ft oroid coordination compounds and larger aggregates in the literature ( Figure 1A). [4] Those compounds can be used as single molecular magnets or in the separation of fullerenes. [5] Stang introducedt he concept of using building blocks to form cage compounds, [6] whereas the groups of Saalfrank, Fujita, and Nitschke used cages to stabilizer eactive species like white phosphorous P 4 and organometallic complexes. [6b, 7] To roids can be useful in bindingg uest molecules. Covalently bondeds ystemsa re quite common, for example,c yclodextrins, cucurbiturils and cryptands which bind cations, [8] anions, [9] or hydrophobic molecules [10] by variation of their peripherald ecoration.T oroidal coordination cages are also able to bind guest molecules and separate fullerenes. [11] These systems are challenging to model using density functionalt heory (DFT) due to their large size. Herein we demonstrate not only that ag eometry-optimized structure of al arge emptym etallacycle can be obtained,b ut also the C 60 and C 70 Figure 1. (a) Previously reported supramolecular metallacycles. Hydrogen atomsa nd solvent molecules were omitted for clarity.D ifferent scales were applied and metal ions highlighted as spheres. (b) Synthesis of the ligands [H 3 (pyO) 3 L]X (5-X), (X = Cl À ,N CS À ,BF 4 À ). (c) Supramolecular metallacycle [Zn 24 Cl 24 {(pyO) 3 L} 12 ]( ZnCl 2 in green), schematic drawinga nd inclusion complex C 60 &9. encapsulated supramolecular entity.W ew erea ble to compute spectroscopic properties and assign the C 60 signals in both the free state and bound insideo ft he metallacycle ( Figure 1C). There are al arge number of studies reporting the computations of C 60 itself. However,p ublications focusingo nt he interaction of C 60 with other macrocyclesa re limited. Most studies including these interactions involve smaller, and often purely organic macrocyclic systems. [12] To the best of our knowledge, this is the first study to computationally investigate such large interacting systems (> 4000 electrons)u sing non-truncated models.

Results and Discussion
Our group demonstrated the synthesis of supramolecular structuresb ys elf-assembly of C 3 -symmetric buildingb locks with three tris-chelating binding pockets and suitable co-ligands (analogous to Figure 1B). [13] Counter ions or solvent molecules typicallys erve as templates in the synthesis, so that discrete coordination cages like M 12 L 4 ( Figure 2, left), M 18 L 6 ,o r M 24 L 8 are accessible. [13d] We did not observe any activation of small molecules with these assemblies, even though ah igh number of potentialc atalytically active metal centers are located proximal to each other.
The overall negative charge of these complexes might be the reason why substrates like lactide are not satisfyingly activated. It was thus necessary to increase the amount of positive chargeso ft he resulting coordination compounds by using stronger Lewis acids, such as Sn IV or Zr IV ,i nsteado fC d II or Pd II . [14] The resulting complexes typically form dimers or trimers and are able to oligomerize acetone in up to 15 repeating units (Figure 2, right). [13d, 15] In this work we address the chargei ssue by increasing the number of positive charges in the ligand itself, while maintaining the isoelectronic structure of the previously reported ligands.T he ligand is prepared from ac ondensation reaction between 2-formylpyridine-N-oxide and the corresponding triaminoguanidinium salts TAG-Cl (4-Cl), TAG-NCS (4-NCS), or TAG-BF 4 (4-BF 4 ), respectively ( Figure 1B).
The synthesis of N-oxide (3)r equires standard protection and deprotection procedures for the aldehyde moiety.T he oxidation of the pyridine nitrogen can be realized under mild conditions by using urea hydrogen peroxide with phthalic an-hydridei na cetonitrile. [16] Triaminoguanidiniums alts (4-X) are obtained by the amination of guanidinium salts with hydrazine hydrate. [17] The resulting compounds (5-X) serve as an excellent ligandsf or Zn II .T he reactiono f[ H 3 (pyO) 3 L]Cl  with Zn(O 2 CH) 2 in N,N-dimethylformamide resultsi nt he formation of as upramolecularc oordination compound [Zn 24 Cl 24 {(pyO) 3 L} 12 ]( 6,F igure1C) next to ac oordination polymer of unknown composition. Thist orus-shaped metallacycle exhibits an outer diameter of 31.7 and Zn II ions are octahedrally coordinated between two alternatelyo riented ligands holdingt ogether the assembly (Figure 3a nd Figure 4, left). The ZnCl 2 moieties occupy the remaining [NNO]b inding pockets. As pherical cavity of 10.7 can be found inside the complex with ap ore opening of 8.2 .E ach value is correctedb y the covalentr adii of hydrogen or carbon atoms.
Only one species is detected in addition to solvents and water. The diffusion coefficient of 7 is found to be D = (7.36 AE 0.08) 10 7 cm 2 s À1 .T he hydrodynamic diameter of 7 can be determined to be 29.9 AE 1.2 using the Stokes-Einstein equation, Figure 2. M 12 L 4 tetrahedronand M 2 L 2 dimer (M = Cd, Sn, respectively). Hydrogen atomsa nd solvent molecules wereo mitted for clarity. [14, 13a] 16 (O 2 CH) 8 {(pyO) 3 L} 12 ] 8.D isordereds olvent molecules were removed by the Squeezer outine (Platon) and hydrogen atomsw ere omitted for clarity. [18] which corresponds well with the observed diameter of the crystal structure (Supporting Information Figure S5 and S6).
Introduction of isothiocyanate leads to the formation of the analogous NCS-metallacycle [Zn 24 (NCS) 16

(O 2 CH) 8 {(pyO) 3 L} 12 ]( 8)
with a5 9% yield. The zinc(II) ions, whichw ere formally occupied by the halides Cl À or Br À ,s hare those sites with isothiocyanate and formate co-ligands (Figure4,r ight). The presence of this coordination compound, and the absence of smaller or larger aggregates in the DMSO solution, wasc onfirmed by dynamic light scattering (Supporting Information). Since the co-ligands point outwards and exhibit as lightly increased steric demand, the system crystallizes in the tetragonals pace group P4 2 1 c with solventf illed channels along the crystallographic caxis (Figure 5a). Host-guest chemistry seems feasible since the cavities of the coordinationo ligomers should be accessible in the solid state. To our surprise, it was not possible to soak crystals of [Zn 24 (NCS) 16 (O 2 CH) 8 {(pyO) 3 L} 12 ]( 8)w ith at oluene solution of C 60 ,ast here was no observed color change. [19] From thesee xperimental resultsw ed ecided to encapsulate the fullerenes into the metallacycles in solution.As olutiono f empty metallacycle (8)w as added to at oluenes olution of C 60 or C 70 .S ingle crystalso ft he respective inclusion compounds were collected after af ew days (Figure 5b-d). The C 60 and C 70 are fully incorporated into the cavity of the metallacycle [Zn 24 (NCS) 20 (O 2 CH) 4 {(pyO) 3 L} 12 ] ( 9). The lattice parameters underwentaslight change compared to 8,w hereas the space group wasm aintained. The periphery of the metallacycle was slightly perturbed, presumably due to the change in polarity of the solvent mixture.
The incorporation of C 60 in solution is validated by NMR spectroscopy (Figure 6). Crystals of 8 and C 60 &9 were removed from the crystallization solution,w ashed with THF and dissolved in [D 6 ]DMSO. The 1 HN MR spectrum shows as hift and broadening of the signals due to the molecular tumbling of C 60 .T he 13 CN MR of C 60 &9 shows only one signalf or C 60 which is in agreement with all carbon atoms being chemically and magnetically equivalent. The encapsulatedC 60 can only be detected next to free C 60 in ah ighly diluted solvent mixture,a nd therefore 13 C-enriched C 60 was used. The 2ppm signal shift from 142.18 ppm (free C 60 )t o1 40.40 ppm (C 60 &9)c learly indicates the incorporation of C 60 in the cavity of the metallacycle.
Calculationspredictedthe encapsulationofC 60 by the metallacycle (6)t ob ef avorable (Figure 7cA). However,t here is an additional local minimum (C)w here the C 60 interacts with the periphery of the metallacycle. This has also been observed crystallographically.A na nalogous binding plot is observed for the C 70 case (see Supporting Information for details). The use of dispersion corrections in these calculations is critical in order to correctly model the attractive interaction. In addition to the entropyl oss, there is an electronic energy barrier (B) that C 60 (and C 70 )m ust overcome caused by steric interactions between theC 60/70 and pyridinyl groups located att he entranceo ft he metallacycle. Due to the nature of these calculations, stepwises ingle-point calculations from the fully optimized state (A)w ere applied. The bindinge nergetics are only  The chloride metallacycle has the largest positive inner core while the outside ZnCl 2 moieties carry the negative potential. This is gradually attenuated in the series of 6 > 7 > 8 with 8 having the least positive inner core. These potentials might be useful in tuning the affinity and specificityf or guest molecules.
In agreement with experimental data, spectroscopic assignmentsu sing computations( DFT,s ee Supporting Information for details) predicta na pproximately 2.0ppm upfield shift for the encapsulatedC 60 &9 speciesv ersus the free C 60 (144.3 ppm vs. 142.2 ppm) (Figure 7a). The IR signatures of an emptyc hloride metallacycle (6)a lso match the computed IR spectrum and the fingerprint region contains several characteristic absorptions ( Figure S25 in the Supporting Information).

Summary
We report the synthesis of an ew type of tris-chelatingp yridine-N-oxide based ligands[ H 3 (pyO) 3 L]X. Coordination of zinc(II) ions leads to structurally interesting cyclic coordination oligomers which serve as hosts for fullerenesC 60 and C 70 .T he inclusion of fullerenesw as observed by single-crystal X-ray diffractometry and NMRs pectroscopy,a nd was validated using computations. These metallacycles are robust and show no sign of decomposition. Although computations on such large systemsare challenging,wewere able to model the C 60 /C 70 encapsulated metallacycles and even predict spectroscopic data, which are in strong agreement with the experimental results. Electrostatic potential maps reveal that the positive charges of the cavity cores can be tuned by the peripheralh alide co-ligands.T hese computations help to provide ad eeper understanding of host-guest interaction in these metallacycles. Similar computations will no doubt be useful for the rational design of host molecules with other cargo.I nt he future,t hese complexes coulds erve as container molecules transporting important cargo. [2f, 20] Experimental Section Experimental methods, synthesis, computational details and results can all be found in the Supporting Information. CCDC 1830922, 1830936, 1830997, 1851696, 1831512, 1831515, 1831516 (3, 5-Cl, 6, 7, 8,C 60 &9,C 70 &9,r espectively), contain the supplementary crystallographic data for this paper.T hese data are provided free of charge by The Cambridge Crystallographic Data Centre.