S-Functionalized Tripods with Monomethylene Spacers: Routes to Tetrairon(III) Single-Molecule Magnets with Ultrashort Tethering Groups

The organization of single-molecule magnets (SMMs) on surfaces is a mainstream research path in molecular magnetism. Of special importance is the control of grafting geometry in chemisorbed monolayers on metal surfaces. We herein present the synthesis, solid-state structure, and magnetic characterization of propeller-like tetrairon(III) SMMs containing the shortest-reported tethering groups for gold surfaces. Functionalization of molecular structure is attained using 2-R-2-(hydroxymethyl)propane-1,3-diol tripodal proligands (H3LR). The R substituents comprise a monomethylene spacer and three different terminations known to act as stable precursors of S-Au bonds (R = CH2SCN, CH2SAc and CH2SSnBu). These chemical groups are shown to be chemically compatible with the tetrairon(III) core and to afford fully-functional SMMs in crystalline form and in fair to excellent yields.


Introduction
The design of molecules exhibiting a directionally bistable magnetic moment, known as singlemolecule magnets (SMMs), has made tremendous progress in the last few years, with operating temperatures that now approach 80 K in some dysprosocenium derivatives [1]. A parallel research path is the investigation of SMMs processed into thin films, down to single layers or even isolated molecules deposited on surfaces. Advances in this field are also significant and demonstrate that SMMs have a real application potential in high-density information storage and molecular spintronics [2][3][4][5].
We found that the H3L CH2SAc proligand could be very conveniently synthesized from (oxetane-3,3-diyl)dimethanol (5) and thioacetic acid in a solvent-free ring-opening reaction similar to that reported for 3,3-dimethyloxetane (Scheme 1b) [31]. The third tripodal proligand, H3L CH2SSnBu , features a disulfide moiety for direct covalent grafting to a Au surface. The length of the terminal chain, a nbutyl group, was chosen to enhance the solubility of the proligand in nonpolar or moderately polar solvents while still permitting good crystal packing. The study involved a careful scrutiny of available methods for the synthesis of asymmetric disulfides. Very low yields were obtained by air oxidation of mixtures of thioacetate 7, prepared from bromide 6, and S-n-butyl thioacetate in the presence of NaOMe/MeOH, due to the extensive formation of symmetric disulfides [32]. Among methods with reportedly enhanced chemoselectivity [33,34], use of diethyl azodicarboxylate (DEAD) adducts gave the best results. As shown in Scheme 1c-e, addition product 9 was prepared in good yield by treating DEAD with butane-1-thiol in toluene at 25 °C [34], low temperatures being in this case unnecessary [35,36]. The reaction of 9 with thiol 8, produced by deprotection of 7 with NaOMe/MeOH, gave the desired compound in 73% yield after acid hydrolysis of orthoacetate 10 and final saponification of monoester 11 to liberate the triol moiety.
Tetrairon(III) complexes were synthesized from [Fe4(OMe)6(dpm)6] [37] and an excess of the appropriate proligand in diethyl ether, according to Equation (1): Although H3L CH2SCN and H3L CH2SAc are only poorly soluble in diethyl ether, the reaction proceeds smoothly with all proligands used. Slow solvent evaporation directly afforded the products as X-ray quality crystals, except for 3, which was recrystallized from 1,2-dimethoxyethane (DME). In their IR spectra, complexes 1 and 2 show C≡N and C=O stretching bands at 2153 and 1701 cm −1 , respectively, which indicate that thiocyanate and thioacetyl groups are incorporated intact in the structures. This conclusion is supported by the single-crystal X-ray diffraction data presented in the next Section.

Structural Descriptions
The structure of tetrairon(III) molecules in these compounds was determined by single-crystal X-ray diffraction at low temperature (120-140 K; see Table S1). At room temperature, the prismatic crystals of 1 are rhombohedral, with unit cell parameters similar to those of the R = Me derivative [37], suggesting a disordered SCN group; however, at 120 K the structure is monoclinic (space group C2/c) and the asymmetric unit contains half tetrairon(III) complex. The two thiocyanate groups are symmetry-related by a two-fold axis and are directed roughly perpendicular to the molecular plane ( Turning now to 2, two different crystalline phases, both belonging to triclinic space group 1 , were found to occur in erratic proportions in crystallization batches. These two phases are solvatomorphs of 2 and show a markedly different propensity to lose crystallinity when removed from their mother solution. The first phase is a mono(diethyl ether) solvate (2·Et2O) and grows as platelets which rapidly lose crystallinity upon standing in air, presumably due to loss of diethyl ether from the lattice. The asymmetric unit contains two Fe4 molecules (MOL1 and MOL2) and two diethyl ether molecules (the structure of MOL1 and MOL2 is displayed in Figure 2 and Figure S1, respectively). The second phase (2·0.375Et2O) grows as block-like crystals which exhibit much greater air-stability as compared with 2·Et2O. Two Fe4 molecules (MOL3 and MOL4) and 0.75 diethyl ether molecules are present in the asymmetric unit (see Figure S2 and Figure S3). Two crystal phases were also isolated for complex 3. Structural data were collected on welldiffracting prisms, which contain the unsolvated complex and belong to monoclinic space group C2/c. However, 3 was also obtained as extensively twinned air-stable square blocks which analyze well for unsolvated 3, but whose diffraction pattern escaped indexing. The two crystal phases likely represent polymorphs of 3. In the structurally characterized crystal phase, the asymmetric unit consists in half tetrairon(III) complex and molecules have crystallographically imposed two-fold symmetry, as in 1 ( Figure 3). All tetrairon(III) complexes reported in this study have at least one disordered tBu group in the solid state; disorder also affects one or both R groups in MOL2 and MOL4 of 2·xEt2O and in 3.  Our structural investigations directly prove that the newly synthesized S-functionalized tripods can be successfully incorporated in propeller-like tetrairon(III) cores. The analyzed crystal phases belong to monoclinic or triclinic space groups and the crystallographic molecular symmetry is thus low: C2 in 1 and 3 (two-fold axis passing through the central metal and one of the peripheral metals), and C1 in both solvatomorphs of 2. However, if the R substituents are disregarded, molecular structures approach three-fold symmetry quite closely; it is thus meaningful to use the angular parameters θ and φ, associated with the trigonally distorted coordination sphere of the central ion, and the helical pitch γ (see [38] for definitions). The resulting values are presented in Table S2, along with those of the derivative with R = CH2SMe (12) [13], and are typical for propeller-like tetrairon(III) complexes [9].
Except for thiocyanate derivative in 1, all used tripodal ligands exhibit a remarkable conformational flexibility in the solid state, which may contribute to explaining why (pseudo)polymorphs are often encountered. The top views provided in Figure 1a, Figure 2a, Figure  3a, Figure S1a, Figure S2a, and Figure S3a show that the S-C bond with the monomethylene spacer tends to adopt a staggered conformation with respect to the coordinated tripodal function. However, because of the approximately three-fold local symmetry, the relative arrangement of the two S-C bonds within the same molecule varies considerably. For instance, it is syn in MOL1 and MOL3 of 2·xEt2O ( Figure 2a and Figure S2a, respectively) but anti in 1 (Figure 1a). The four crystallographically independent molecules in 2·xEt2O further differ in the conformation of the thioacetyl groups, as resulting from rotation around the S-C bond with the monomethylene spacer. The planar SAc groups can range from being roughly perpendicular to the molecular plane to forming a dihedral angle (δ) down to about 30° with it; in this case, they approach van der Waals contact with tBu groups of dpm − ligands. In MOL1 and MOL3 of 2·xEt2O, for instance, both thioacetyl groups are ordered within experimental resolution. However, while in MOL1 one thioacetyl has δ = 85.7° and the second one has δ = 30.1° (Figure 2b), in MOL3 both thioacetyls display a perpendicular arrangement with δ = 88.7° and 89.9° ( Figure S2b). In MOL2 and MOL4 the situation is more complicated, since one or both side chains are disordered. MOL2 ( Figure S1b Finally, it is important to notice that these propeller-like molecules can exist as Δ and Λ isomers; however, they crystallize in centrosymmetric space groups and are, thus, isolated as racemic mixtures.

DC Magnetic Studies
The magnetic properties of vacuum-treated powder samples of 1, 2·Et2O and 3 were studied in DC mode by recording low-field χMT vs. T and isothermal MM vs. H data, where χM and MM are molar magnetic susceptibility and molar magnetization, respectively. Here we describe the results obtained for 1, which is representative of the whole series. Data for the remaining compounds are available in Figure S4 and Figure S5. The temperature dependence of the χMT product for 1 shows the trend typical of antiferromagnetically coupled systems with uncompensated spin moments ( Figure 4). The presence of dominant antiferromagnetic interactions is clearly revealed by the room-temperature χMT value, which is significantly lower than expected for four uncoupled si = 5/2 spins (17.5 emu K mol −1 with g = 2.00). On decreasing temperature the curve features a minimum at around 100-110 K, then rises again up to a maximum, which is close to the value of the Curie constant for an S = 5 total spin ground state (C = 15.0 emu K mol −1 with g = 2.00). χMT vs. T data at T > 20 K were then fitted to a Heisenberg-Dirac-Van Vleck plus Zeeman Hamiltonian, assuming three-fold symmetry and using two different superexchange-coupling constants to describe nearest-neighbor (J1) and next-nearestneighbor (J2) interactions within the "J" convention (Equation (2)): (s1 denotes the spin vector on the central ion). The J1, J2 and g values obtained with this procedure are gathered in Table 1 along with those of 12 [13] for comparison. J1 is invariably antiferromagnetic and typical in magnitude for this class of tetrairon(III) SMMs [9], whereas J2 is zero within the fitting error or slightly ferromagnetic. However, the best-fit value of J2 must be regarded with care since this parameter may compensate for systematic errors in data collection and reduction [14,39]. Isothermal MM vs. H data recorded at low temperature show pronounced nesting when plotted as a function of H/T (Figure 4, inset), indicating the presence of magnetic anisotropy acting on the ground spin state. We thus fitted the data using the S = 5 Giant-Spin (GS) Hamiltonian in Equation (3): retaining only second-order axial anisotropy (E = = 0) and assuming an isotropic g-factor ( = ̿ with ̿ = identity matrix). The best-fit parameters so obtained are gathered in Table 1 along with those of 12 [13] for comparison.   8 14.48 (7) 7 3.15(7)·10 −7 4.87(10)·10 −7 3.63(15)·10 −7 5.2(2)·10 −7 1 Data from [13], unless otherwise noted. 2 From low-field χMT vs. T data at T > 20 K. 3 In [13] a different model was used to extract J1, J2 and g. 4 From isothermal MM vs. H data at low T. 5 Preferential orientation correction (see Experimental Section). 6 From HF-EPR spectra. 7 From AC susceptibility data at HDC = 1 kOe. 8 Calculated as (|D|/kB)S 2 (preferably from the HF-EPR D value, when available).

HF-EPR Spectra
The powder HF-EPR spectra of 1 and 2·Et2O were recorded from 5 to 20 K at 190 and 230 GHz. Figure 5 displays the data for 1 at 230 GHz, while the remaining spectra are provided as Figures S6,  S7, and S8. At low temperature, the spectra exhibit the characteristic features for systems with a high-spin ground state associated to a quasi-axial Ising-type anisotropy: the parallel transitions are observed in the low-field part of the spectrum (i.e., below the g = 2 resonance position) whereas the perpendicular transitions appear in the high-field part. On decreasing temperature, the signals at the extremes of the spectrum get stronger and stronger as expected for transitions corresponding to MS = −5 → −4 ( Figure 5). Moving towards the center of the spectrum, the successive ΔMS = 1 transitions are observed. The unequal spacing of neighboring transitions in the parallel region is the signature of higher-order axial anisotropy terms. The simulated spectra ( Figure 5 and Figures S6-S8) were obtained with the GS Hamiltonian in Equation (3) and the resulting best-fit parameters are gathered in Table 1. Molar magnetization isotherms simulated with these parameters are presented in Figure  S9 for 1 and 2·Et2O. In spite of their low crystallographic molecular symmetry, compounds 1 and 2·Et2O have almost perfectly axial anisotropy in their S = 5 ground state. The value of |D/E| ~ 0.02 is about four times smaller than in related compound 12 [13], whose spin-Hamiltonian parameters are also reported in Table 1. A quasi-axial, Ising-type anisotropy with |D| = 0.40 ÷ 0.45 cm 1 in the ground spin state is characteristic for this class of tetrairon(III) SMMs featuring two coordinated tripodal ligands [9]. The observed anisotropy is primarily determined by single-ion terms (with the dominant contribution provided by peripheral Fe 3+ ions) and, to a lesser extent, by spin-spin interaction anisotropy [39]. Typical is also the positive fourth-order axial parameter , whereby the 11 sublevels of the S = 5 manifold define a "compressed" parabola [37,38].
An accurate reproduction of the spectra required an anisotropic linewidth (i.e., a different linewidth when the applied field is directed along x, y or z). In addition, while the linewidth of ztransitions remains essentially constant across the whole spectrum, with only minor differences between 1, 2·Et2O, and 12, x-and y-transitions behave differently. Their linewidth increases significantly when going from the center to the extremes of the spectrum. Furthermore, such a linewidth variation is largest for 1, intermediate for 2·Et2O, and smallest for 12 (Table S3). This suggests the occurrence of a compound-dependent distribution of spin-Hamiltonian parameters in the same material (strain effects), affecting primarily the E value.

AC Magnetic Studies
We investigated the dynamics of the magnetization on the same powder samples by means of AC susceptibility measurements. A small static field (HDC = 1 kOe) was applied in order to slow down under-barrier relaxation by quantum tunneling while only marginally affecting spin levels, thereby allowing to measure a larger set of relaxation times [13]. The appearance of maxima in the out-ofphase component of the molar magnetic susceptibility (χM″) is a clear signature of slow magnetic relaxation ( Figure 6). We thus fitted the χM″ vs. frequency (ν) curves to an extended Debye model [40,41] to extract the average relaxation time τ, the width of the distribution of relaxation times α, and χM,T − χM,S as functions of temperature (χM,T and χM,S are the isothermal and adiabatic molar magnetic susceptibilities, respectively). The results are gathered in Tables S4-S6. As shown in Figure 6, linear ln(τ) vs. 1/T plots are obtained indicating thermally activated relaxation mechanisms, whereby the relaxation time follows Arrhenius law τ = τ0exp[Ueff/(kBT)]. The effective energy barrier Ueff within the series of compounds 1, 2·Et2O and 3 (Table 1) is comparable to the total splitting of the S = 5 multiplet, evaluated as (|D|/kB)S 2 , showing that in the adopted experimental conditions under-barrier relaxation is largely suppressed. The fastest-relaxing species is 1, which also features the smallest |D| according to both MM vs. H data and (when available) HF-EPR spectra. In compound 12 [13] the effective barrier is instead significantly lower than (|D|/kB)S 2 . This is likely to arise from the large rhombicity of 12, whose E parameter is four times larger than in 1 and 2·Et2O. As a result, underbarrier relaxation is enhanced [38], and 12 attains a very similar spin dynamics to 1.
The α parameter in 1, 2·Et2O, and 3 is 0.16-0.21 at the lowest temperatures reached and decreases upon heating ( Figure S10), indicating that the distribution of relaxation times narrows with increasing temperature. These α values and their temperature dependence are in agreement with literature data on similar tetrairon(III) compounds [42,43]. Differences between 1 and 2·Et2O are within the experimental error, notwithstanding the narrower distribution of spin-Hamiltonian parameters in the latter (Table S3). The α value in 12 is, instead, significantly smaller at all temperatures ( Figure  S10). Rewardingly, this compound also exhibits the smallest EPR linewidth variation in the series and, by consequence, the narrowest distribution of spin-Hamiltonian parameters (Table S3).

Materials and Methods
All reactants and solvents were reagent grade and used without further purification, unless otherwise stated. Anhydrous solvents were either of commercial origin (dichloromethane, toluene) or prepared by standard methods [44]. Diethyl ether, from a freshly opened can, was pre-dried by stirring overnight with granular CaCl2, refluxed over Na/benzophenone and distilled under N2.
Methanol and DME were refluxed over Mg(OMe)2 and NaH, respectively; acetone was dried over CaSO4, while acetonitrile was sequentially dried with 3A molecular sieves and CaH2; these solvents were finally distilled with protection against moisture prior to use. Petroleum ether was the fraction with boiling point 40-60 °C. (Oxetane-3,3-diyl)dimethanol (5) was prepared in 79% yield by following a literature method [45], except that the crude product was purified by column chromatography (silica gel, CH2Cl2:MeOH 4:1). 4-(Bromomethyl)-1-methyl-2,6,7-trioxabicyclo[2.2.2]octane (6) was prepared as described elsewhere [46]. KSCN was dried in vacuum over P2O5 prior to use. NaOMe was used as a solution in methanol, prepared by careful addition of metallic Na to anhydrous methanol under N2 atmosphere. [Fe4(OMe)6(dpm)6] was prepared as described in [37]. Thin-layer chromatography (TLC) was performed on silica gel glass plates 60 F254 from Merck KGaA (Darmstadt, Germany). Compounds were visualized under an UV lamp and/or using I2 or ceric ammonium molybdate stains. Elemental analysis was carried out on an EA1110 CHNS-O automatic analyzer from CE Instruments (Milan, Italy). 1 H NMR spectra were recorded at 303 K and 200.13 MHz with an FT-DPX200 NMR spectrometer from Bruker Co. (Karlsruhe, Germany); 13 C NMR spectra were recorded at 303 K and 100.61 MHz with an FT-NMR AVANCE400 spectrometer from Bruker BioSpin S.r.l. (Milan, Italy); chemical shifts are expressed in ppm downfield from external tetramethylsilane, by setting the residual 1 H signal of acetone-d6, CD3OD, CD2Cl2 and CDCl3 at 2.05, 3.31, 5.32, and 7.26 ppm, respectively, and the methyl 13 C signal of acetone-d6 at 29.84 ppm [47]; coupling constants (J) are given in hertz. The following abbreviations are used in reporting NMR data: s = singlet, brs = broad singlet, d = doublet, t = triplet, q = quartet, m = multiplet. IR spectra were recorded as KBr pellets or on NaCl disks using an FTIR-4200 spectrophotometer with 2 cm −1 resolution from Jasco Corporation (Tokyo, Japan); ν values are given in cm −1 . The following abbreviations are used in reporting IR data: s = strong, m = medium, w = weak, br = broad.

Synthesis of [Fe4(L CH2SSnBu )2(dpm)6] (3)
H3L CH2SSnBu (27.1 mg, 0.113 mmol) dissolved in anhydrous diethyl ether (3 mL) was added to a solution of [Fe4(OMe)6(dpm)6] (56.9 mg, 0.0377 mmol) in the same solvent (23 mL). After 24 h stirring the solution was filtered and the solvent slowly evaporated in air at room temperature. The resulting partially crystalline solid was redissolved in anhydrous DME (5 mL) and the filtered solution was slowly evaporated almost to dryness over a dodecane trap. The dark-orange crystals were washed with a MeOH:DME mixture (10:1 v/v) to remove liquid residuals and dried under vacuum (40.1 mg, 59%). Crystallization batches comprise well-diffracting prisms and/or highly-twinned square blocks. All characterization data, except for single-crystal X-ray diffraction data, were collected on samples containing predominantly the latter crystal phase. Anal. calcd (%) for C84H148Fe4O18S4: C 56.12, H 8.

Single-Crystal X-ray Diffraction
All structures were analyzed with a four-circle X8APEX diffractometer from Bruker-Nonius (Delft, The Netherlands) equipped with Mo-Kα radiation (0.71073 Å) and a Kryo-Flex N2 flow cryostat for data collection at 120(2) or 140(2) K. The structures were solved by direct methods using the SIR92 program [48]. Full-or block matrix least-squares refinement on Fo 2 was performed with the SHELXL-97 or SHELXL-2014/7 programs [49], implemented in the WINGX suite [50]. Disorder effects on tBu groups, on the side chains of tripodal ligands and on lattice solvent were handled by introducing constraints/restraints on geometry and/or displacement parameters [51]. Hydrogen atoms were set in idealized positions with isotropic displacement parameters constrained to those of the attached carbon atoms. Crystal data and refinement parameters, and further refinement details are available in Table S1 and Supplementary Note 1, respectively. Graphics utilized Ortep-3 for Windows (v2.0) [50] and POV-Ray™ for Windows (v3.5) [52]. CCDC 2033840-2033843 contain the supplementary crystallographic data for this paper and can be obtained free of charge via www.ccdc.cam.ac.uk/data_request/cif, or by emailing data_request@ccdc.cam.ac.uk, or by contacting The Cambridge Crystallographic Data Centre, 12 Union Road, Cambridge CB2 1EZ, UK; fax: +44 1223 336033.

Magnetic Measurements
Magnetic data in DC and AC mode were recorded on MPMS and PPMS instruments from Quantum Design GmbH (Darmstadt, Germany). We used vacuum-treated powder samples of 1, 2·Et2O and 3 with masses of 22.19, 10.77, and 12.33 mg, respectively. Samples were either pressed in a pellet and wrapped in Teflon® tape (2·Et2O and 3) or restrained in eicosane (1). DC data were recorded at temperatures ranging from T = 1.8-2.0 to 300 K and in fields H = 1 kOe (T < 35 K) or 10 kOe (T  35 K). The MM/H ratio, where MM is the molar magnetization, was assumed to correspond to the static molar magnetic susceptibility χM. Isothermal magnetization curves were also recorded in fields up to 50-70 kOe at 1.8-2.0, 2.5 and 4.5 K. Raw data, after correction for the sample holder and addenda, were reduced using molecular weights of 1671.4 (1), 1705.5 (2) and 1797.7 (3) and diamagnetic corrections (estimated from Pascal's constants [53]) of -940 (1), -957 (2) and -1050 (3)·10 −6 emu mol −1 . AC magnetic data in a 1 kOe applied static field (HDC) were recorded on the same samples at frequencies of the oscillating field (1-10 Oe) from 10 Hz to 10 kHz and at temperatures between 1.8-1.9 and 5.5 K. The fitting of DC magnetic data was carried out using original software based on F02ABF and E04FCF NAG routines for matrix diagonalization and least-squares fitting, respectively [54]. The modelling of preferential orientation effects, i.e., a non-perfectly random distribution of crystallite orientations in the sample, was required for an accurate fitting of MM vs. H curves. To this aim, we used the correction scheme recently published by some of us, which implies the refinement of one additional least-squares parameter (a2) [16]. The estimated standard deviations of best-fit parameters, quoted in Table 1, were calculated from the variances of the regression coefficients, as provided by routine E04YCF [54].

HF-EPR Spectra
HF-EPR spectra were recorded on vacuum-treated powder samples, using a home-made spectrometer operating in single-pass configuration at T ranging from 5 to 20 K and in applied magnetic fields up to 12 T. Gunn diode sources operating at either 95 or 115 GHz associated to frequency doublers were used to produce microwaves at 190 or 230 GHz for sample excitation. To avoid field-induced orientation effects, the samples were pressed into pellets. Powder spectra were simulated using a dedicated software, after an initial fitting of the resonance positions [55,56].
Since it was not possible to introduce simultaneously an anisotropic linewidth and a strain, each spectrum was obtained as the sum of several component spectra. Each component spectrum involved only part of the transitions for which an anisotropic linewidth was considered; the sum of these component spectra allowed considering all the possible transitions (see Table S3 for the details on the linewidths used).

Conclusions
Synthetic routes were devised to 2-R-2-(hydroxymethyl)propane-1,3-diol derivatives with R = CH2SCN, CH2SAc and CH2SSnBu. The tripods were subsequently used to functionalize propellerlike tetrairon(III) SMMs by a ligand-exchange reaction. The products were obtained in crystalline form and in fair to excellent yields; according to crystallographic, spectroscopic and chemical evidence they contain intact S-based groups. Magnetic studies and HF-EPR spectroscopy showed that the new complexes retain the magnetic properties typical of SMMs of this family (S = 5, |D| = 0.41 ÷ 0.44 cm −1 , Ueff/kB = 14-16 K).
Supplementary Materials: The following are available online at www.mdpi.com/2312-7481/6/4/55/s1; Supplementary Note 1: Structure refinement details for 1, 2·Et2O, 2·0.375Et2O and 3; Figure S1. Molecular structure of MOL2 in 2·Et2O viewed along and perpendicular to the idealized three-fold axis; Figure S2. Molecular structure of MOL3 in 2·0.375Et2O viewed along and perpendicular to the idealized three-fold axis; Figure S3. Molecular structure of MOL4 in 2·0.375Et2O viewed along and perpendicular to the idealized threefold axis; Figure S4. Temperature dependence of the molar magnetic susceptibility in low field, and molar magnetization isotherms at low temperature for 2·Et2O; Figure S5. Temperature dependence of the molar magnetic susceptibility in low field, and molar magnetization isotherms at low temperature for 3; Figure S6. Temperature dependent HF-EPR spectra of 1 at 190 GHz along with best simulations; Figure S7. Temperature dependent HF-EPR spectra of 2·Et2O at 190 GHz along with best simulations; Figure S8. Temperature dependent HF-EPR spectra of 2·Et2O at 230 GHz along with best simulations; Figure S9. Molar magnetization isotherms recorded on 1 and 2·Et2O compared with calculated curves based on HF-EPR parameters in Table 1; Figure S10. Temperature dependence of the α parameter in a 0.1 T applied static field for 1, 2·Et2O, 3 and 12; Table S1. Crystal data and refinement parameters for 1, 2·Et2O, 2·0.375Et2O and 3; Table S2. Geometrical parameters for the coordination sphere of the central iron ion in 1, 2·Et2O, 2·0.375Et2O, 3 and 12, after averaging according to D3 symmetry; Table S3. Linewidths (in Gauss) used to simulate HF-EPR spectra for compounds 1, 2·Et2O, and 12, depending on the transition (M → M ') and on magnetic field orientation (x, y or z); Table S4. Fitting parameters of the isothermal χM′′ vs. ν curves (HDC = 1 kOe) based on the extended Debye model for compound 1; Table S5. Fitting parameters of the isothermal χM′′ vs. ν curves (HDC = 1 kOe) based on the extended Debye model for compound 2·Et2O; Table S6. Fitting parameters of the isothermal χM′′ vs. ν curves (HDC = 1 kOe) based on the extended Debye model for compound 3. Funding: This research was funded by European Union through Network of Excellence MAGMANET, grant number 15767, NanoSci-ERA project SMMTRANS, grant number 06NSE03, and ERC Advanced Grant MolNanoMaS, grant number 267746, to R.S.; it was also supported by Italian MIUR through a PRIN2008 project, grant number 2008FZK5AC.

Conflicts of Interest:
The authors declare no conflict of interest.