A ferromagnetically coupled pseudo-calixarene [Co 16 ] wheel that self-assembles as a tubular network of capsules

Reaction of Co(OAc) 2 ·4H 2 O and Hsal in a basic MeCN solution affords the hexadecanuclear wheel [Co 16 (sal) 16 (OAc) 16 ]·16MeCN ( 1 ·16MeCN) that displays ferromagnetic nearest neighbour exchange and has pseudo-calixarene character. Symmetry equivalent wheels self-assemble to form remarkable tubular networks of capsules in the extended structure. Johansson monochromator, 2 mm divergence slit and 2.5 degree Soller slits on the incident beam side. LynxEye detector and Bruker DIFFRAC software. Diffraction measured from 2θ = 3° - 30°; step size, 0.0101°. Samples were loaded into quartz capillaries with a 1 mm inside diameter and measured while spinning.


Introduction
Molecular wheels attract continued attention, partly due to beautiful structural aesthetics, but also because some show fascinating and potentially useful physical properties. 1 In the field of molecular magnetism, molecular wheels of 3d transition metals came into prominence in the late 1980s and early 1990s with the publication of [Cr III 8] and [Fe III 10]. 2 The former has inspired the development of a large family of homo-and heterometallic Cr wheels including the first examples of odd-numbered wheels that display topological frustration, with potential as quantum bits in information processing. 3 The latter, and other antiferromagnetically coupled even-membered wheels, are characterised by a diamagnetic spin ground state and display interesting quantum phenomena and spin dynamics, including tunneling of the Néel vector, 4 spin-multiplet mixing effects 5 and magnetic level repulsions. 6 The intervening years have witnessed the publication of wheels of all the 3d metals with nuclearities up to eighty four. 7 In Co II chemistry 8 early examples of wheels included a dodecanuclear cluster built with a substituted pyridone 9 and an heptanuclear Anderson wheel stabilised by tripodal alcohols, 10 both display ferromagnetic exchange interactions.

Results and discussion
The reaction between Co(OAc)2·4H2O and Hsal (salicylaldehyde) in a basic MeCN solution (see SI for full details) leads to the formation of pink single crystals after 3 days upon diffusion of Et2O into the mother liquor. Crystals of [Co16(sal)16(OAc)16]·16MeCN (1·16MeCN) were in a tetragonal cell and structure solution was performed in the P4/n space group. The asymmetric unit (ASU) contains one quarter of the formula, and symmetry expansion affords the wheel shown in Fig. 1. The metallic skeleton (Fig. 2) describes a single-stranded, sinusoidal [Co II 16] wheel of approximate diameter, Co1···Co1' = ~14.5 Å. The 'inside' of the wheel is stabilised by sixteen µ3-, syn, syn, anti-OAc ligands. Eight lie in the metal plane, with four above and four below the plane. The sixteen µ-sal ligands are arranged in a similar manner on the 'outside' of the wheel. Eight µ-sal ligands are in a belt around the periphery of the wheel, whilst four are above the plane, and four below (Fig. 1). These out of plane µsal ligands form an interesting arrangement that is reminiscent of calixarenes, 11 presenting a shallow hydrophobic pocket as a result. The magnetic unit between nearest neighbours contains one µ-O(alkoxide), one µ-O(carboxylate) and one syn, syn-O-C-O(carboxylate). The interaction between next nearest neighbours is mediated by the syn, anti-O-C-O(carboxylate) bridge. The Co-O-Co angles subtended by the µ-O atoms are in the range ~92.1-99.5°, with the Co II ions all being in distorted octahedral {CoO6} geometries. A search of the Cambridge Structural Database (CSD) reveals approximately twenty [Co16] structures, with more than half being squares and tetrahedra stabilised by thia-and sulfonyl-calix [4]arenes. 12 There are two other wheels, one comprising linked squares and cubes built with a bis-benzimidazolediol ligand, and one incorporating four linear {Co4} subunits constructed with polytriazolate ligands. 13 Further symmetry expansion of the new pseudo-calixarene structure presented by 1 reveals a remarkable arrangement in which symmetry equivalent (s.e.) wheels pack to form tubular networks of capsules, and channels between the tubules. Inspection of the ring structure of 1 in space filling representation (Fig. 3A, in the ab plane) shows a very small channel through the ring as a result of the syn, syn, anti-OAc ligands on the inside of the molecule. Symmetry expansion along the c axis gives rise to a dimer with the next s.e. wheel as shown in Fig. 3B (in the ac plane). These pseudo-calixarene wheels lock together in the solid state through inter-digitation of the µ-sal ligands, with some OAc ligands also forming part of the space filling belt. This gives rise to an encapsulated space with a volume of ~272 Å 3 that is occupied by disordered solvent molecules (Fig. 3C). 14 It was not possible to resolve this disorder given the high symmetry and diffuse nature of the electron density. Interdigitation continues along the c axis, forming infinite tubular stacks of capsules, and these pack as shown in Fig. 3D to form solvent filled channels that run parallel through the extended structure. Single crystals of 1 are solvent dependent, but the nature of the extended structure will be studied further with a view to forming more stable analogues that can be desolvated and explored for potential guest transport to the interior of the cages.

Magnetic properties
DC magnetic susceptibility (χ) and magnetisation (M) measurements of 1 were taken in the T = 300-2.00 K, B = 0.1 T and T = 2.0-10 K and B = 0.5-9.0 T temperature and field ranges, respectively. These are plotted as the χT product versus T and M versus B in Fig. 4. The T = 300 K value of χT = 43.2 cm 3 K mol -1 is equal to the value expected for sixteen non-interacting S = 3/2 Co II ions with g = 2.40. Upon cooling the χT value decreases slowly to ~41.7 cm 3 K mol -1 at 38 K before rising sharply to a maximum of ~52.8 cm 3 K mol -1 at T = 6 K, and then falling to ~10.9 cm 3 K mol -1 at T = 2 K. The initial drop in χT is due to the magnetic anisotropy of the octahedral Co II ions, 15 while the increase at low temperature is due to weak ferromagnetic interactions. The drop in value between 6-2 K is most likely due to intermolecular antiferromagnetic interactions. This behaviour is similar to that observed for the pyridone-stabilised [Co12] wheel. 9 The M vs B data is in agreement with this interpretation, with the magnetisation increasing rapidly with increasing field, not saturating and reaching a value of M = 33.4 µB at T = 2 K and B = 9 T. The ferromagnetic exchange in 1 is consistent with magneto-structural correlations developed for O-bridged Co II clusters where the sign and magnitude of the interaction is dictated by the Co-O-Co angle, with the exchange becoming more ferromagnetic with decreasing angle. 16 Here, the Co-O-Co angles are all ≤ 99.5° and are therefore expected to mediate weak ferromagnetic exchange.

Theoretical studies
In order to probe the nature and magnitude of the magnetic exchange and anisotropy of the Co II ions further, we now turn to theory (see the computational details in the SI for full details). 17 We have performed DFT calculations on a model of complex 1 (Model 1) based on its ASU, i.e. one [Co II 4] moiety plus one linking Co II ion (Fig. S3). Based on symmetry and structure there are four unique nearest neighbour magnetic exchange interactions with J values ranging between +1.7 cm -1 ≤ J ≤ +3.8 cm -1 (Table S2). The narrow range of ferromagnetic exchange interactions found can be attributed to the similarity of the structural parameters present, including the average Co-µO-Co angles. 16 We have also performed overlap integral calculations 17 using the singly occupied molecular orbitals (SOMOs) of the Co II ions (Fig. S4). These help to elucidate the magnitude and sign of magnetic interactions since their magnitude is directly proportional to the magnitude of the antiferromagnetic interaction, i.e. the larger the overlap the larger the antiferromagnetic interaction and vice versa. For 1, there are three intermediate and six small overlap interactions, resulting in a small ferromagnetic interaction overall. Interestingly, replacement of the phenoxide group in Model 1 with a point charge changes the sign of the magnetic interaction from ferromagnetic to antiferromagnetic (4.2 cm -1 to -3.4 cm -1 ), highlighting the importance of this moiety for obtaining ferromagnetic exchange. We have also calculated the next-nearest neighbour exchange mediated via the syn, anti-O-C-O(carboxylate). These are weak and antiferromagnetic, with J < -0.2 cm -1 (Fig. S5).
All the Co II ions in 1 are in distorted octahedral geometries (Table S3), with previous magnetostructural studies suggesting such ions would possess large easy-plane anisotropy. 18 Ab initio NEVPT2 calculations on each Co II ion in 1 confirms this, with values in the range +41.2 ≤ D ≤ +87.1 cm -1 . The dominant contribution to D arises from the dxz/yz  dxy electronic transition (Table S4, Fig. S5). Positive axial zero-field splitting can be attributed to electronic transitions between orbitals with different mL levels and the magnitude correlated to the energy separation between the orbitals involved in the electronic transition (Table S4, Fig. S6). 19

Conclusions
In summary, the reaction of Co(OAc)2·4H2O and Hsal affords the aesthetically pleasing [Co16(µsal)16(µ3-OAc)16] (1) wheel, which displays weak ferromagnetic nearest neighbour exchange interactions, in agreement with DFT calculations. Ab initio NEVPT2 studies suggest the presence of large single ion easy-plane anisotropy. In the extended structure the wheels form dimeric capsules via inter-digitation of the sal/acetate ions and this extends to form infinite tubular stacks of capsules upon symmetry expansion. Attempts to make analogues of 1 with different M II ions, different carboxylates and derivatised salicylaldehyde ligands are in progress, with a view to also examining guest transport to the interior of the cages.

Synthesis
All reagents and solvents were obtained from commercial sources and used without further

Physical Measurements
Elemental analyses (C, H, N) were performed by the University of Ioannina microanalysis service.

X-ray diffraction
Single crystal X-ray diffraction data were collected on a Bruker D8 VENTURE diffractometer Powder X-ray diffraction data for 1 were collected using a Bruker D8 ADVANCE with copper radiation at 40 kV, 40 mA and a Johansson monochromator, 2 mm divergence slit and 2.5 degree Soller slits on the incident beam side. LynxEye detector and Bruker DIFFRAC software. Diffraction measured from 2θ = 3° -30°; step size, 0.0101°. Samples were loaded into quartz capillaries with a 1 mm inside diameter and measured while spinning.

Magnetic Measurements
Magnetic susceptibility data were collected on a polycrystalline sample of 1 on a Quantum Design Dynacool PPMS equipped with a 9 T magnet in the temperature range 300 -2.00 K. Diamagnetic corrections were applied to the observed paramagnetic susceptibilities using Pascal's constants.

Computational Details
To estimate the intramolecular magnetic exchange interactions in 1 we have employed Density Functional Theory (DFT) in Gaussian 09 on a model complex (Model 1) created from the ASU of 1. 1 We have performed pairwise exchange interaction calculations by keeping only the two paramagnetic centres of interest in Model 1, replacing the remaining Co II ions with diamagnetic Zn II ions. This method is known to reproduce experimental magnetic exchange values for systems with weak intramolecular magnetic exchange interactions (J ≤ 10 cm -1 ). 2 The hybrid B3LYP functional 3 has been used together with the TZV basis set for Co, SVP basis set for Zn, O and SV basis set for C and H atoms. 4 We have employed Noodleman's broken symmetry methodology. 5 To calculate the zero field splitting (zfs) parameters for each Co II centre in the ASU we have used the ORCA software suite (version ORCA 4.0). 6 The zeroth-order regular approximation (ZORA) method in combination with the ZORA contracted version of basis set (ZORA-def2-TZVP for Co and ZORA-def-SVP for rest of the elements) 7 is known to be a reliable methodology to estimate zfs parameters. We have used the resolution of identity (RI) approximation. During state-average complete active space selfconsistent field (SA-CASSCF) calculations we have considered seven electrons in five d-orbitals (CAS (7 electrons / 5 3d-orbitals)) in the active space with ten triplet and fifteen singlet roots. We have used     <Co(α)dz 2 ||Co(β)dyz> = 0.003 <Co(α)dx 2 -y 2 ||Co(β)dz 2 > = 0.053 <Co(α)dz 2 ||Co(β)dx 2 -y 2 > = 0.075 Table S3. SHAPE analysis 9 performed on each Co II ion in the ASU of 1.    S6. Ab initio NEVPT2 computed d-orbital splitting for each Co II ion in the ASU of 1. Easy-plane anisotropy (+D) can be attributed to the electronic transitions between orbitals with different mL levels (dxz/yz  dxy/x 2 -y 2 ). Note that two electronic transitions dxz/yz  dxy (cyan curly arrow) and dxz/yz  dx 2 -y 2 (light green straight arrow) give positive D values with the dominant contribution arising from the dxz/yz  dxy electronic transition. The magnitude of D is correlated to the energy separation between the orbitals involved in the electronic transition (i.e. between dyz/xz and dxy/x 2 -y 2 ).