A Ferromagnetically Coupled, Bell-Shaped [Ni4Gd5] Cage

Reaction between NiCl 2 ·6H 2 O, 2-hydroxy-4-methyl-6-phenyl-pyridine-3-amidoxime (H 2 L), benzoic acid and M(NO 3 ) 3 ·6H 2 O (M = Gd, Y) in MeCN under basic conditions, yields the complexes [Ni II4 Gd III5 (PhCOO) 10 (HL) 4 (HL zw ) 4 (OH) 2 (NO 3 ) 2 ]Cl·13.6MeCN·H 2 O ( 1 ·13.6MeCN and O). Both clusters display similar structures, consisting of a bell-shaped {Ni II4 M III5 } unit, in which a linear ‘zig-zag’ {Ni 4 } subunit bisects the central {M III5 } ‘ring’. Direct (dc) and alternating current (ac) magnetic susceptibility measurements carried out in the 2 – 300 K temperature range for complexes 1 and 2 revealed ferromagnetic intermolecular interactions, while heat-capacity measurements for the Gd analogue suggest that complex 1 lowers its temperature from T = 9.6 K down to 2.3 K by adiabatically demagnetizing from B i = 7 T to B f = 0.


Introduction
The field of molecular magnetism has become a multidisciplinary, dynamic field of science for which a breadth of potential applications can be envisioned. For example, anisotropic complexes "remembering" their magnetization upon removal of an initially applied magnetic field at low temperatures (< 80 K) can function as Single-Molecule Magnets (SMMs) with potential use in information storage devices. 1 Isotropic, high magnetic density cages that possess numerous fieldaccessible spin states have been shown to exhibit a large magnetocaloric effect that makes them excellent candidates for cryogenic refrigeration. 2 A major challenge in both areas remains the molecular design and structural control of such species, despite the fact that many, if not all, of the key ingredients are known. Indeed, synthetic control becomes increasingly more difficult as the nuclearity of the cage increases, not least when dealing with flexible organic ligands and metal ions that can display a variety of coordination geometries. Structural complexity leads to an increase in the number of intramolecular exchange pathways, and to obvious consequences for interpreting and understanding magnetic behavior. Despite these difficulties, numerous examples of such large cages exist, including an {Fe III 10Gd III 10} cyclic complex with an S = 60 ground-state, 3 a giant {Fe III 18Fe II 24} nanocluster with an S = 45 spin ground-state, 4 a {Mn III 12Mn II 7} cluster with an S = 83/2 ground-state, 5 a {Mn III 11Mn II 6} complex with an S = 37 ground-state, 6 and most recently a [Ni21Gd20] cage displaying a remarkable (field-induced) S = 91 ground state. 7 In this work, the isolation, characterization and investigation of the magnetic behavior of an enneanuclear [Ni II 4Gd III 5] cluster are reported, built by employing the amidoxime ligand 2hydroxy-4-methyl-6-phenyl-pyridine-3-amidoxime, H2L (Scheme 1), which was previously employed succesfully in heterometallic chromium chemistry yielding two Cr III Table S2, while full crystallographic details may be found in the CIF files with CCDC reference numbers 1896818-1896819, for 1 and 2, respectively.

Result and discussion
Synthesis.
In order to fully investigate the synthetic parameters that dictate the formation of 1 (and 2) we performed the reaction upon employing various bases i.e. NMe4OH and NaOH, and also increased the duration of the reaction, but in all cases complex 1 was the product as observed by IR spectroscopy and pXRD. In addition, upon increasing the metal:ligand ratio from 1:1:2:2 initially employed in the original reaction to 1:1:1:1 we again managed to isolate only complex 1, albeit in very small yield (only a handful of crystals). Finally, we performed the reaction under solvothermal conditions, but in all cases a green-white amorphous solid was obtained.
The external {Ni II 2Gd III 2} core is held in place by two η 1 : η 1 : μ benzoate and two η 2 : η 1 : η 1 : μ HL -1 ligands, with the link between the two fragments of the core occurring via two η 2 : η 1 : η 1 : μ4 HLzw -1 and two η 2 : η 1 : μ HL -1 ligands. Finally, two chelating and two monodentate benzoate ligands, as well as one chelating nitrate anion, complete the coordination environment of the metal ions. All Ni II centres are six-coordinate, adopting distorted octahedral geometries with bond lengths in the range, 2.02-2.11 Å. The Gd centers are eight-coordinate, adopting either square-antiprismatic (Gd1, Gd2 and symmetry equivalent) or cubic (Gd3) geometries, as defined by SHAPE analysis ( Figure S1, Table S1). 9 It is noteworthy that in both structures the Ni-Ni-Ni-Ni skeleton is almost identical and the geometries of the Gd5 and Y5 ions are very similar. In the crystal lattice the molecules of 1 stack in an on-set fashion on the b axis with the solvate atoms located in the void space, while no significant intermolecular interactions are present ( Figure   S2). Again, for 2 no significant intermolecular interactions are present in the extended structure.
Magnetochemistry. The magnetic properties of 1 and 2 were investigated by DC susceptibility measurements in the 300 -2 K temperature range under a 0.1 T applied magnetic field ( Figure 2).
The room-temperature χmΤ values of both complexes are very close (1, 46.13 cm 3 mol -1 K; 2, 4.82 cm 3 mol -1 K) to the Curie constants expected for four Ni II ions and five M III ions (1, 44.21 cm 3 mol -For 2, we were able to simultaneously fit the magnetic susceptibility and magnetization data, using PHI 10 and employing spin-Hamiltonian (2): where the J1 exchange represents interaction between Ni1-Ni1 mediated through one ( coupling. Remarkably, the fit reveals a change of the magnetic anisotropy, going from a stronger axial to planar symmetry, for Ni1 (Ni1΄) and Ni2 (Ni2΄), respectively consistent with bond lengths for Ni1 (Ni1΄) up to 2.11 Å, and the lack of predominant axial symmetry in Ni2 (Ni2΄). Our initial, unsuccessful, approach to fitting the magnetic data for 2 was to oversimplify the fit by imposing J  J1 = J2 and D  D1 = D2, i.e., a 1-J + 1-D model ( Figure S3). Next, we considered D  D1 = D2 and different J's, i.e., a 2-J + 1-D model, but the resulting fit, although slightly improved, remained unsatisfactory ( Figure S4). For both models, a nonzero zJ  was needed in order to account for the 11 lowering of χmΤ at the low-temperature region, in contrast with the crystal structure of 2 displaying no significant intermolecular interactions. The large ground state of Gd III (ŝGd = 7/2), combined with the L = 0 angular momentum, makes 1 a potentially attractive candidate material for magnetic refrigeration. 2 Therefore, its magnetocaloric effect (MCE) was investigated, upon studying the changes of magnetic entropy, ΔSm, and adiabatic temperature, ΔTad, following a change of the applied magnetic field, ΔB = Bi -Bf, where i and f indicate initial and final states, respectively. As heat capacity (cp) is a powerful tool for estimating the MCE, 2 we collected temperature-dependent cp measurements over the ∼(0.3 -30) K range under several applied magnetic fields (Figure 3). At the experimentally accessed high-temperature region, cp may be successfully modelled upon using the Debye heat capacity, which describes the nonmagnetic lattice contribution and simplifies to a clatt/R = aT 3 dependence at the lowest temperatures, with a = 3.5 × 10 −2 K −3 , as common for molecule-based compounds. At low temperatures, cp assumes the form of a Schottky-like anomaly that shifts to high temperature with increasing applied magnetic field. Considering that the magnetic exchange interactions between 3d-3d centres (Ni-Ni) should in principle be stronger than both the 3d-4f (Ni-Gd) and 4f-4f (Gd-Gd) interactions, the intra-molecular magnetic ordering observed at sufficiently low temperatures may be roughly described as following: four Ni II centres establish a ferromagnetic net spin of Ŝ4Ni = 4, weakly coupled to the five peripheral Gd III spins. To model the experimental data, we calculate the Schottky curves that arise from the field-split levels of the central Ŝ4Ni net spin and five independent ŝGd spins. Then, we sum these curves to the aforementioned phonon contribution and we plot the results in Figure 3, as solid lines for B = 1, 3 and 7 T, respectively. The comparison with the experimental data shows that the agreement improves with increasing field value, likely because of the presence of weak but non-negligible exchange interactions involving the Gd III spins, which are not taken into account in the calculations. Higher fields promote full decoupling of the weakly correlated Gd III spins, yielding an excellent agreement.
13 The inset of Figure 3 shows the entropy (S) that we calculate from the heat capacity data, by making use of the following equation: Assuming that the Ni-Ni correlations are sufficiently strong, as to establish a ferromagnetic net spin of Ŝ4Ni = 4, one would expect the magnetic entropy content to be Sm/R = ln(2Ŝ4Ni + 1) + 5 × ln(2ŝGd + 1) = 12.6. The experimental data corroborates this assumption. As can be seen in the 14 inset of Figure 3, the zero-field entropy curve quickly reaches S/R ≈ 12 at low temperatures (T ≈ 2.5 K). With increasing temperature, the experimental entropy gradually increases, due to the nonmagnetic lattice contribution, as expected. The maximum magnetic entropy content for 4 Ni II spins (ŝNi = 1) and 5 Gd III spins (ŝGd = 7/2) corresponds to Sm/R = 14.8. This value is for noninteracting spins and, therefore, cannot be reached experimentally at low temperatures. 2b From the entropy data, it is then straightforward to obtain the magnetocaloric effect of 1. Figure 4 shows the results of ΔSm and ΔTad as a function of temperature and for selected full From the isothermal M vs. B curves of Figure 2, the so-obtained ΔSm(T, ΔB) values obtained for ΔB = 1 and 3 T, displayed in Figure 4 (top), agree perfectly with the data obtained from cp, thus demonstrating the robustness of the procedures employed. temperature for several values of the applied magnetic field change, as labelled.

Conclusions
In conclusion, the use of the amidoxime ligand 2-hydroxy-4-methyl-6-phenyl-pyridine-3amidoxime, H2L, has led to the isolation of unusual, bell-shaped [Ni II 4M III 5] (M = Gd (1), Y(2)) cages, in which a linear, zig-zag chain of Ni II ions is sandwiched between five Gd III ions. The ligand adopts four coordination modes, bridging up to four metals. Dc magnetic susceptibility and magnetization measurements reveal the presence of dominant ferromagnetic interactions in both complexes, with analysis of the data for 2 revealing JNi-Ni = 5.26 and 2.26 cm -1 , in accordance with previously published Ni cages containing similar structural building blocks. Magnetocaloric measurements show that complex 1 lowers its temperature from T = 9.6 K down to 2.3 K by adiabatically demagnetizing from Bi = 7 T to Bf = 0, respectively. That complex 1 is a good magnetic refrigerant at liquid helium temperatures is also manifested by the value of ΔSm that reaches 19.2 J kg -1 K -1 at T = 3.3 K, for ΔB = 7 T.
Furthermore, our present and previously reported results 11a on the use of the H2L ligand strongly suggest its excellent coordination ability towards the formation of polymetallic complexes, displaying various nuclearities {Mx} (x=7-9) and metallic topologies; thus, we are confident that further exploration of its coordination chemistry will lead to new species with beautiful structures and exciting properties.

Conflicts of interest
The authors have no conflicts to declare.