A Mixed Valence CoIICoIII2 Field-Supported Single Molecule Magnet: Solvent-Dependent Structural Variation

One-pot reaction of the Schiff base N,N’-ethylene bis(salicylaldimine) (H2L), CoCl2.6H2O, and [Ph2SnCl2] in acetone produces the mixed valence CoIICoIII2 compound [CoIICoIII2(μ-L)2(Ph)2(μ-Cl)2]·(CH3)2CO·H2O (1). Our recent study already revealed that the same reaction mixtures in methanol or ethanol produced a heterometallic SnIVCoIII (2) or monometallic CoIII complex (3), respectively. Comparison of these organometallic systems shows that the 2,1-intermetallic Ph shift occurs in any of those solvents, but their relevant structural features (mononuclear, dinuclear-heterometallic, and trinuclear mixed valence) are solvent dependent. Geometrical structural rotation is also discussed among the related organometallic CoIICoIII2 systems. The AC magnetic susceptibility measurements indicate that 1 is a single molecule magnet (SMM), exhibiting a field-induced slow magnetic relaxation with two modes. The relaxation time for the low-frequency channel is as slow as τ~0.6 s at T = 2.0 K and BDC = 1.0 T.

In the IR spectrum, the title compound 1 exhibits a strong intense absorption at 1645 cm -1 because of ν(C=N). Medium intense bands at 444 and 459 cm -1 can be assigned to ν(Co-N) and ν(Co-O), respectively. An additional broad peak at 3480 cm -1 is observed for the non-coordinated water molecule. It was also characterized by elemental analysis and single crystal X-ray diffraction.

Description of Crystal Structure
The crystal structure of the mixed valence complex [Co II Co III 2 (µ-L) 2 (Ph) 2 (µ-Cl) 2 ] ·(CH 3 ) 2 CO·H 2 O (1) was determined at 298 and 150 K (see below) and is presented in Figure 1, while some important geometrical parameters are listed in Table 1. As shown in Figure 1, two Co III ions occupy the N 2 O 2 cavities of the tetradentate Schiff base ligand (L 2− ) while the Co II ion interconnects them through bis(µ-phenoxido) and µ-chlorido bridges. While the peripheral Co(III) metal cations possess CClN 2 O 2 octahedral geometries with angle variances of ca. 30º 2 , that of the central Co(II) is constructed from two Cl and four O-atoms and is considerably more distorted in view of the measured angle variances of ca. 144-147º 2 . The octahedral volumes of 10.5-10.7 Å 3 against 13.6 Å 3 confirm the oxidation states (+3 and +2, in this order) of the metals. Bridging atoms (O and Cl) along with the metal centers (Co II and Co III ) form two corner-shared (the Co II central cation) Co 2 O 2 Cl trigonal bipyramids (Figure 1b). The Co III cations are displaced by 0.11-0.12 Å from the corresponding least-square N 2 O 2 planes and toward the phenyl ligands. Such least-squares planes of the L 2− ligands in 1 make an angle of 74.14 • .   (3) 68.43 (7) 70.21 (7) 69.17 (7) 70.70 (7) 68.91 (5) 68.86 (5)

Solvent Dependent Structural Variation and Rotation of Coordination Geometry
Structural variations that depend on solvent are not uncommon but always interesting [55]. Distinct solvents have different properties, e.g., in terms of polarity, acidity or basicity, coordinating or bridging ability, etc. Cases are also known where solvents do not affect the basic structure, forming solvatomorphs [60,61].
Heterometallic 3d-tin systems synthesized under open air atmosphere show a structural diversity which includes: i) adducts, ii) non-adduct covalent compounds, iii) cocrystals, iv) salt cocrystals, v) salts, or even vi) hetero-organometallic compounds [52]. In the present work, we have found that the mixed valence Co II Co III 2 compound (1) was isolated in acetone from a reactant mixture which was used to synthesize other compounds (2 or 3) in different solvents (methanol or ethanol) [55]. The 2,1-intermetallic Ph shift (Sn IV to Co III ) is a common phenomenon in all the cases (1, 2, and 3), occurring in all the tested solvents, while the Sn IV center of the methanol product (2), which is absent in ethanol (in 3), is formally replaced by CoCl 2 in acetone, which bridges to another Ph-Co III site generating 1. As presented in Figure 2, compound 3 is a mononuclear organocobalt(III) and 2 is a dinuclear organocobalt(III)-organotin(IV) compound, while the title compound (1) is a trinuclear mixed valence organocobalt(III)-cobalt(II) system. Therefore, not only the nuclearity changes (1→2→3), but also the metal oxidation state and/or combination (Co III →Co III Sn IV →Co II Co III 2 ) differ from one solvent to another, and such a solvent-dependent behavior, to our knowledge, was not known in tin(II/IV) systems derived from the same reactants mixture. In addition, solvents display herein three different roles: coordinated (water in 3), bridging (methanol derived methoxido in 2), and non-coordinated (acetone and water in 1, not shown in Figure 2). A search on the Crystallographic Structural Database [59] shows a considerable number (~300) of compartmental Schiff base cobalt derivatives, ca. 100 of which involve trinuclear species. Further restriction to only dichloro-diphenoxido bridges results in only four systems (CCDC ref. code: AKEHIJ, XUHSUP, XUHTAW, and XUHTEA, Figure S1, Supplementary Materials) [57,58]. AKEHIJ is a Co II 3 system but the remaining three are Co II Co III 2 systems containing Co III -C bonds (Table 1). In complex 1, as in XUHSUP and XUHTEA, the chloride anions are in cis position while in XUHTAW they are mutually trans. Our compound confirms that non-bulky ligands (phenyl groups in 1), to complete the geometry of the peripheral cobalt cations, favor their cis geometry [57,58]. Upon redrawing their metal coordination geometries by keeping one of the outer octahedra fixed, we observe that the other outer octahedron has rotated by ca. 75 • in the sequence 1, XUHSUP-1 (cis) to XUHTAW (trans) to XUHTEA, XUHSUP-2 (cis) (Figure 3). Such a structural relation in mixed valence trinuclear Co II Co III 2 systems is worthy to be mentioned [59].

Magnetic Properties
Mixed valence compounds are also very important in the field of magnetism [20,24,27,46,[62][63][64][65][66][67][68]. Their magnetic properties are usually investigated as they can show interesting magnetic exchange coupling (ferro or antiferro) or slow magnetic relaxation (characteristic of a single molecule magnet, SMM). The majority of mixed valence SMM systems is based on manganese(IV/III/II) [20,24,[64][65][66][67][68]. Although cobalt(II)-containing compounds can also display single molecule magnetism, known as mixed valence cobalt(II/III) SMM systems are only a few [45,46] and one of them shows multiple magnetic relaxation [46]. Therefore, the investigation of the magnetic behavior of such systems is promising and we have now studied both the DC and AC magnetic properties of our trinuclear mixed valence Co II Co III 2 system 1. However, the outer Co(III) ions in 1 are expected to be diamagnetic because of the low spin d 6 electronic configuration.

DC Magnetic Data
The temperature evolution of the effective magnetic moment displays features that are typical for a single hexacoordinate Co(II) complex with large zero-field splitting ( Figure 4). The room-temperature value µ eff = 5.20 µ B on cooling gradually decreases to T~15 K and then it shows a hook. The magnetization per formula unit at B = 7.0 T and T = 2.0 K saturates to M 1 = M mol /µ B = 2.39 that is far from the spin-only value owing to the zero-field splitting. There is a small remnant magnetization M r at both, T = 4.6 and 2.0 K, confirming a long-range ordering. This causes a hook visible at the low-temperature effective magnetic moment. The inverse susceptibility below 15 K turns toward zero, however on further cooling it returns back to the linear dependence. Both magnetic data-sets were fitted simultaneously by minimizing a functional of the form F = w · E(χ) + (1 − w) · E(M) that balances the relative errors of the susceptibility and magnetization, respectively. A zero-field splitting (ZFS) Hamiltonian working in the space of spin-only ketŝ has been used with the axial zero-field splitting parameter D. The Zeeman term mimics the powder-sample property: it depends upon the distribution of the magnetic field along a set of uniformly separated grids at the meridian. The eigenvalues in the basis set of spin kets enter the formulae of the statistical thermodynamics for the magnetization and susceptibility, respectively [69]. The fitting procedure was restricted to the susceptibility data above 15 K and it gave the following set of magnetic parameters: D/hc = 86.9 cm −1 , g z = 2.030, g xy = 2.778, the molecular field correction zj/hc = +0.10 cm −1 , and the temperatureindependent magnetism χ TIM = 10 × 10 −9 m 3 mol −1 . The fit is perfect as the discrepancy factors R(χ) = 0.0011 and R(M) = 0.0061. The g-factors average to g av = 2.53. A more elaborate Griffith-Figgis model working in the space of spin-orbit ketŝ involves the spin-orbit splitting parameter λ = −ξ/2S, the orbital reduction factor κ, the Figgis CI parameter A, and the axial and rhombic splitting parameters ∆ ax and ∆ rh , respectively. Owing to the T-p isomorphism the orbital angular momentum refers to L p = 1 [69]. The susceptibility and magnetization data were fitted with a reasonable set of magnetic parameters: (Aκλ)/hc = −170 cm -1 , g L = −(Aκ) = -1.44, ∆ ax /hc = −650 cm −1 , ∆ rh /hc = 25 cm -1 , zj/hc = 0.063 cm −1 and χ TIM = 11 × 10 −9 m 3 mol −1 ; R(χ) = 0.0017 and R(M) = 0.032. With negative ∆ ax the ground state refers to the E g crystal field term that is further split by ∆ rh to a pair of orbitally non-degenerate terms. The results are displayed in SI.
The results of the ab initio calculations using the ORCA package at the CASSCF+NEVP T2 level gave the data as listed in Table 2. The results need to be discussed at three levels of sophistication. First, the CASSCF calculations gave energies of the multielectron terms. In the light of the crystal field theory, the ground octahedral term 4 T 1g is split on the symmetry lowering to the tetragonally elongated system into the 4 E g (ground) and 4 A 2g (excited) terms. The orbital degeneracy of the 4 E g term is lifted on further symmetry lowering. To this end, the calculated energies of the multielectron terms are interpreted according to the following scheme: The involvement of the spin-orbit interaction causes the splitting of the multielectron terms into multiplets (Kramers doublets) with the relative energies {0, 195, 645, 952} and {1653, 1763} where the first bracket arises from the mother 4 E g term and the second one from 4 A 2g . It need be emphasized that four lowest Kramers doublets have no relationship to the spin Hamiltonian parameters such as D and E because they result from the pair of quasidegenerate terms 1 4 A and 2 4 A for which the perturbation theory diverges. The multiplet splitting δ = 195 cm −1 needs to be considered as a principal and valuable result of the calculations.

AC Susceptibility Data
In the first scan the AC susceptibility response is investigated at low temperature, depending upon the external DC field, for a set of representative frequencies f of the AC field ( Figure 5). The out-of-phase susceptibility χ" is silent at the zero field, however, it rises with the applied field, passes through a maximum, and then attenuates. Such a behavior confirms that 1 shows a field-induced slow magnetic relaxation. The frequency dependence of maxima at χ" indicates an existence of two or more relaxation channels.
The AC susceptibility data, taken for 22 frequencies ranging between f = 0.1 to 1500 Hz, are shown in Figure 6 at fixed T = 1.9 K and for an external magnetic field up to B DC = 1.0 T. Two maxima of the out-of-phase susceptibility confirm two relaxation channels: the low-frequency (LF) and the high-frequency (HF). The DC field supports the LF channel and shifts the relaxation time τ = 1/(2π f max to higher values. containing seven free parameters: the isothermal susceptibilities χ Tk , the distribution parameters α k , and the relaxation times τ k for each relaxation channel, along with the adiabatic susceptibility χ S . This equation allows separating the real (in-phase) and the imaginary (out-of-phase) components. A joint functional F = w · E(χ ) + (1 − w) · E(χ ) is minimized during the fitting process. The final relaxation parameters with their standard deviations, and the discrepancy factors of the fit, are presented in Table 3. These parameters were used in creating extrapolation/interpolation lines shown in Figure 6. Notice, at B DC = 1.0 T and T = 1.9 K the relaxation time is as long as τ LF = 0.59(3) s and the mole fraction is χ LF = 0.76. Temperature evolution of the AC susceptibility components at fixed B DC = 0.6 T is shown in Figure S2 (Supplementary Materials) for a set of frequencies f = 0.1-1500 Hz. The individual frequency components χ' merge at T > 5 K that can be assigned as a blocking temperature; simultaneously χ" vanishes.
The subsequent data acquisition was done at the external field B DC = 0.6 T for a set of temperatures (Figure 7). Based upon the profile of the χ" vs f [T, B DC ] curves, two relaxation channels are well recognized. On heating, the LF peak progressively decays which is confirmed by the fitted values of χ LF and x LF (Table 4). Above T > 4.7 K the LF-χ" component becomes too flat, and the HF tail spans outside the limit of the data taking; these features prevent a reliable data fitting.  The Argand (Cole-Cole) diagram and the Arrhenius-like plot are presented in Figure 8. The Argand diagram is formed of two overlapping arcs of different height, position, and flattening. The Arrhenius-like plot lnτ vs T −1 maps a temperature evolution of the LF and HF relaxation times. The relaxation time referring to the high-frequency relaxation channel decreases with temperature only slightly.
In order to analyze the relaxation time for the high-frequency relaxing species of 1, Figure 9 was constructed. Three high-temperature and three low-temperature points were fitted via linear regression. They refer to an individual relaxation equation τ −1 = CT n which yields the linearized form lnτ = −lnC − n · lnT = b[0]+b [1] · lnT. The hightemperature points gave n~1 which is a fingerprint of the direct process of the relaxation that is usually written as τ −1 = AB m T = A B T (For the Raman process n = 7-9 are typical values). The second straight line possess a very small slope so that probably the temperature-independent quantum tunneling process τ −1 = D(B) adopts its significance.

Materials and Methods
All the reagents and solvents were purchased from commercial sources and used as received. The Schiff base H 2 L was prepared according to the reported procedures [56]. Elemental (C, H, and N) analyses were performed on a Perkin-Elmer 2400 II analyzer. IR spectra were recorded in the region 400-4000 cm -1 on a Perkin-Elmer RXIFT spectrophotometer with samples as KBr disks.

Synthesis
To an acetone suspension (20 mL) of H 2 L (0.134 g, 0.5 mmol) were added CoCl 2 ·6H 2 O (0.120 g, 0.5 mmol) and [SnPh 2 Cl 2 ] (0.172 g, 0.5 mmol). The resulted reaction mixture was refluxed for 30 min to obtain a dark red solution which was kept at room temperature for slow evaporation. Within 3 h, the formed dark red crystals, suitable for X-ray diffraction analysis, were collected by filtration and washed with cold acetone. Other fractions were then obtained from the reaction mixture, one of those sent for elemental analysis.

Crystal Structure Determination
An X-ray quality crystal of 1 was immersed in cryo-oil, mounted in a Nylon loop and measured at 150 K and 298 K. Intensity data were collected using a Bruker APEX II SMART CCD diffractometer with graphite monochromated Mo-Kα (λ = 0.71073 Å) radiation. Data were collected using omega scans of 0.5º per frame and full sphere of data were obtained. Cell parameters were retrieved using Bruker SMART software and refined using Bruker SAINT [70] on all the observed reflections. Raw data were corrected for absorption effects using the multi-scan method (SADABS) [70]. Structures were solved by direct methods by using the SHELXS-2014 package [71] and refined with SHELXL-2014/7 [71]. Calculations were performed using the WinGX Version 2014.01 [72]. All the hydrogen atoms attached to carbon atoms were inserted at geometrically calculated positions and included in the refinement using the riding-model approximation. U iso (H) were defined as 1.2 U eq of the parent carbon atoms for phenyl and methylene residues and 1.5 U eq of the parent carbon atom for the methyl group. Least square refinements with anisotropic thermal motion parameters for all the non-hydrogen atoms were employed. Solvent molecules in the structure of 1 obtained at room temperature were highly disordered and could not be modelled; therefore, they were removed by using PLATON SQUEEZE routine [73]. A total void of 1301 Å 3 containing 312 electrons per unit cell was found and fits well for one acetone (32 electrons) and one water molecule (10 electrons) per asymmetric unit. These results are reasonably consistent with the elemental analysis and the low temperature (150 K) X-ray data collection where such crystallization solvent molecules were located. The final refinement converged to R 1 (I > 2σ(I)) value of 0.0334 (at 150 K) and 0.0559 (at 298 K). Crystallographic data of 1 at both 150 and 298 K are summarized in Table S1 (Supplementary Materials).

Magnetic Measurements
The DC magnetic data were taken with a SQUID magnetometer (MPMS-XL7, Quantum Design) using the RSO mode of detection at B DC = 0.1 T. Freshly prepared samples were encapsulated in a diamagnetic gelatin-made sample holder. The molar magnetic susceptibility data were corrected for the underlying diamagnetism and presented in the form of the effective magnetic moment. Magnetization data were taken at T = 2.0 and 4.6 K until B DC = 7.0 T and presented in the form of magnetization per formula unit per Bohr magneton, M 1 = M mol /(N A µ B ). The AC susceptibility data were taken at the oscillating field B AC = 0.38 mT for 22 frequencies in the range of f = 0.1-1500 Hz. In these measurements an external magnetic field B DC = 0.2, 0.4, 0.6, 0.8, and 1.0 T was applied at a fixed temperature (T = 2.0 K) while at a fixed field (B DC = 0.6) AC susceptibility data were for a set of different temperatures (T = 1.9-4.7 K).

Theoretical Calculations
Ab initio calculations were performed with ORCA 4.0.0 computational package [74] using the experimental geometry of the complex under study. The relativistic effects were included in the calculations with zero-order regular approximation (ZORA) together with the scalar relativistic contracted version of def2-TZVP basis functions for Co atom and def2-SV(P) basis functions for other elements. The calculations of ZFS parameters were based on state average complete active space self-consistent field (SA-CASSCF) wave functions complemented by N-electron valence second order perturbation theory (NEVPT2) [75][76][77]. The active space of the CASSCF calculations comprised of seven electrons in five metalbased d-orbitals. The state averaged approach was used, in which all 10 quartet and 40 doublet states were equally weighted. The calculations utilized the RI approximation with appropriate decontracted auxiliary basis set and the chain-of-spheres (RIJCOSX) approximation to exact exchange. Increased integration grids (Grid4 and GridX5) and tight SCF convergence criteria were used. The ZFS parameters were calculated through quasi-degenerate perturbation theory in which an approximation to the Breit-Pauli form of the spin-orbit coupling operator (SOMF) and the effective Hamiltonian theory was utilized [78][79][80].

Conclusions
The mixed valence Co II Co III 2 compound [Co II Co III 2 (µ-L) 2 (Ph) 2 (µ-Cl) 2 ]·(CH 3 ) 2 CO·H 2 O (1) was obtained from the reaction mixture of the Schiff base N,N'-ethylenebis(salicylaldimine) (H 2 L), CoCl 2 .6H 2 O, and [Ph 2 SnCl 2 ] in acetone, while a dinuclear organocobalt(III)-organotin (IV) compound (2) and a mononuclear organocobalt(III) (3) were isolated from the same reaction mixture but using different alcohols as solvents. The strategy revealed a few interesting points: (i) The occurrence of the 2,1-intermetallic Ph shift (Sn IV →Co III ) is independent of the solvent used; (ii) the solvent-dependent nuclearity of the metal complexes [1 (in 3)→2 (in 2)→3 (in 1)]; (iii) the solvent-dependent combination of metal ions (Co III →Co III Sn IV →Co II Co III 2 ) resulting in homo, hetero, or mixed valence systems. To our knowledge, such a solvent-dependent structural diversity in association with a solventindependent occurrence of Ph transfer has never been found in tin(II/IV) systems derived from a common reactants mixture.
Further comparison of the title compound 1 with related mixed valence organometallic Co II Co III 2 systems shows that their different metal coordination geometries (involving cis or trans chloro ligands) are related by rotations of one of the outer moieties relatively to the other one.
The AC susceptibility measurements indicate that 1 is a single molecule magnet showing a field-induced slow magnetic relaxation with two relaxation modes. While the high-frequency process spans the usual range [4][5][6][7] of the relaxation time for other single molecule magnets (τ HF~1 0 −7 s), the low-frequency branch is as slow as τ LF~0 .6 s at T = 2.0 K and B = 1.0 T. While the barrier height and the blocking temperature for 1 are in the range of other reported SMMs , two relaxation modes were not investigated commonly in a Co II Co III 2 mixed valence system.

Supplementary Materials:
The following are available online, Figure S1: Trinuclear cobalt compounds containing compartmental salen type Schiff base ligands [57][58][59]. Compounds are presented with their corresponding CCDC reference codes. Figure S2: Temperature dependence of the AC susceptibility components of 1 for frequencies f = 0.1-1500 Hz at B DC = 0.6 T. Table S1: Crystallographic data of 1 at 150 and 298 K.