Filling the gap in Extended Metal Atom Chains: Ferromagnetic Interactions in a Tetrairon(II) String Supported by Oligo- -pyridylamido Ligands

The string-like complex [Fe 4 (tpda) 3 Cl 2 ] (H 2 tpda = N 2 , N 6 -di(pyridin-2-yl)pyridine-2,6-diamine) ( 2 ) was obtained as the first homometallic extended metal atom chain based on iron(II) and oligo- -pyridylamido ligands. The synthesis was performed in strictly anaerobic and anhydrous conditions using dimesityliron, [Fe 2 (Mes) 4 ] (HMes = mesitylene) ( 1 ) as both an iron source and a deprotonating agent for H 2 tpda. The four lined-up iron(II) ions in the structure of 2 (Fe∙∙∙Fe = 2.94-2.99 Å, Fe  Fe  Fe = 171.7-168.8°) are wrapped by three doubly-deprotonated twisted ligands and the chain is capped at its termini by two chloride anions. The spectroscopic and electronic properties of 2 were investigated in dichloromethane by UV-Vis-NIR absorption spectroscopy, 1 H-NMR spectroscopy and cyclic voltammetry. The electrochemical measurements showed four fully resolved, quasi-reversible one-electron redox processes, implying that 2 can adopt five oxidation states in a potential window of only 0.8 V. Direct current magnetic measurements indicate dominant ferromagnetic coupling at room temperature, although the ground state is only weakly magnetic. Based on Density Functional Theory and Angular Overlap Model calculations, this magnetic behavior was explained as being due to two pairs of ferromagnetically-coupled iron(II) ions ( J = –21 cm –1 using 𝐽𝐒(cid:3552) (cid:3036) ∙ 𝐒(cid:3552) (cid:3037) convention) weakly antiferromagnetically coupled with each other. Alternating-current susceptibility data in the presence of a 2 kOe dc field and at frequencies up to 1.5 kHz revealed the onset of slow magnetic relaxation below 2.8 K, with an estimated energy barrier U eff / k B = 10.1(1.3) K.


INTRODUCTION
EMACs (Extended Metal Atom Chains) constitute a wide class of polynuclear metal complexes containing three to eleven metal centers lined up in a string by an array of deprotonated oligo-pyridylamine or related ligands. [1][2][3] In these materials, which may be homo-or hetero-metallic, homo-or 4 divalent metal ions helically wrapped by three tpda 2-ligands. As predicted by DFT calculations the complex shows dominant ferromagnetic interactions at room temperature, with two pairs of ferromagnetically-coupled iron(II) ions exhibiting weak antiferromagnetic interactions with each other.

EXPERIMENTAL SECTION
Materials and methods. Because iron amides are exceedingly air sensitive, all operations were carried out in an MBraun UniLAB glovebox under an inert and controlled dinitrogen atmosphere continuously purified over molecular sieves and a copper catalyst (H2O and O2 < 1 ppm). All solvents were anhydrous and of commercial origin (except for Et2O, which was distilled from its sodium benzophenone ketyl solution before use); they were deoxygenated through three freeze-pump-thaw cycles and stored over 4A molecular sieves. 28 Compounds 1 26,27 and Fe4Cl86THF 29 were prepared according to well-established literature methods. H2tpda was obtained by refluxing 2,6-diaminopyridine with 2-fluoropyridine and LiH in toluene/pyridine, following a recently-reported high-yield procedure. 30 Elemental analysis was performed using a ThermoFisher Scientific Flash 2000 analyzer. ESI-MS measurements were made on a 6310A Ion Trap LC-MS(n) instrument (Agilent Technologies) by direct infusion of dichloromethane solutions. The electronic spectra were recorded on solutions of 2 in dichloromethane using a UV-Vis-NIR Jasco V-570 spectrometer and a quartz cuvette sealed with an air-tight teflon cap (optical path length: l = 0.1 cm). The spectrum of the H2tpda ligand was also measured before and after the addition of excess t BuOK. Since solutions of H2tpda and t BuOK in dichloromethane gave time-dependent spectra, tetrahydrofuran was used as solvent. Room temperature 1 H-NMR spectrum was recorded on a CD2Cl2 solution of 2 using a valved 5 mm NMR tube and a Bruker Avance400 FT-NMR spectrometer (400.13 MHz). The chemical shifts are expressed in ppm downfield from Me4Si as external standard, by setting the residual 1 H signal of CD2Cl2 at 5.32 ppm.
Synthesis of [Fe4(tpda)3Cl2]2.6CH2Cl20.84Et2O (22.6CH2Cl20.84Et2O). A suspension of H2tpda (97.9 mg, 0.372 mmol) in toluene (3.6 mL) was added dropwise to a red solution of 1 (109.4 mg, 0.1859 mmol) in toluene (4 mL) under stirring. During the addition of the ligand the reaction mixture turned into a brown suspension. Fe4Cl86THF (22.0 mg, 0.0937 mmol of Fe) was then added to afford an orange suspension. The reaction mixture was carefully heated to the reflux temperature for 2 h 45 min and subsequently cooled down to room temperature (an ESI-MS spectrum of the reaction mixture is presented in Figure S1). The brown-orange solid was separated from the dark red liquid phase by centrifugation and repeatedly extracted with CH2Cl2 (9 × 3 mL) to give an orange solution. Slow diffusion of diethyl ether (1.4 times by volume) into the CH2Cl2 solution afforded dark-red prisms of the product ( Figure   S2). The crystals were separated from a powdery residue by flotation and stored at -80 °C in flamesealed quartz tubes containing their mother liquor. After magnetic measurements, the product was dried in vacuum to give an orange powder (27 mg, 27%). All subsequent characterizations, except for elemental analysis, were performed with strict exclusion of dioxygen and water. For elemental analysis, the sample had to be briefly exposed to the air, whereupon it immediately darkened due to reaction with dioxygen and/or water. Although incorporation of oxygen may explain the lower than calculated C,H,N content (see below), the reliability of elemental analysis data for 2 is limited. The homogeneity of bulk samples was inferred from 1 H-NMR spectra (see text) and from the X-ray analysis performed on eight single crystals taken from the same or different syntheses, which invariably gave the same unit-cell parameters. Anal. Calcd for 2: C, 50 X-ray Crystallography. The single-crystal X-ray structure determination on 22.6CH2Cl20.84Et2O was carried out at 115(2) K on a Bruker-Nonius X8APEX diffractometer equipped with Mo-K generator, area detector and Kryoflex liquid nitrogen cryostat. Crystals of the compound and a small amount of mother liquor were transferred out of the glovebox using a Schlenk flask. The Schlenk was then opened under fast dinitrogen flux and the selected crystal was covered with epoxy glue, rapidly mounted on the tip of a glass capillary and directly transferred to the cold dinitrogen flux of the cryostat.
Data collection and reduction were carried out using APEX2 31 and SAINT 31 softwares, respectively, while absorption correction was applied with SADABS. 31 The structure was solved and refined on Fo 2 by standard methods, using SIR92 32 and SHELXL-2014/7 33 programs implemented in the WINGX v2013.3 suite. 34 All non-hydrogen atoms were refined anisotropically, unless otherwise noted, while hydrogen atoms were added in idealized positions, allowed to ride on the parent carbon atoms and treated isotropically with U(H) = 1.2Ueq(C) for aromatic and methylene hydrogens and U(H) = 1.5Ueq(C) for methyl hydrogens.
A total of 2.6 and 0.84 CH2Cl2 and Et2O molecules, respectively, were located in the lattice per tetrairon(II) unit. Two interstitial CH2Cl2 molecules with full and ~0.44 occupancies, respectively, were subject to unrestrained refinement. The remaining interstitial solvent molecules were extensively disordered and overlapping, so that restraints had to be applied to their geometry and displacement parameters. The CH2Cl2 and Et2O molecules with the largest occupancy were used as structural references for geometrical restraints, with esd values of 0.01 and 0.02 Å for 1,2-and 1,3-distances, respectively. In the reference Et2O molecule, O-C and C-C bond lengths were also restrained to be equal within 0.01 Å. All atoms in the minority Et2O component and a few poorly-resolved C atoms of CH2Cl2 molecules were treated isotropically. Enhanced rigid bond restraints 35 were applied to bonded anisotropic atoms (RIGU instruction with esd values of 0.004 Å 2 for both 1,2-and 1,3-distances). Crystal data and 7 refinement parameters are gathered in Table 1 while interatomic distances and angles are presented in   Table 2 (more exhaustive listings are available in Table S1 and S2).
Electrochemistry. A Potentiostat/Galvanostat mod. PARSTAT 2273 (Princeton Applied Research, Oak Ridge, USA) was used to perform CV. Experiments were carried out at different scan rates (v = 0.02 -5 V s -1 ) using a cell for small volume samples (0.5 mL). A 1 mm-diameter GC disk (PAR), a Pt wire and an Ag wire were used as working, counter, and quasi-reference electrode, respectively. The GC electrode was cleaned following a procedure which ensures a suitable and reproducible surface for ET studies. 36 For all experiments, the potential of the quasi-reference electrode was calibrated against the ferricenium/ferrocene redox couple (in dichloromethane, E° = 0.460 V against the KCl SCE). 37 All the reported potential values are referred to ferricenium/ferrocene redox couple. Since the complex degrades quickly in the presence of O2 or H2O, all the experiments were carried out in the above-described MBraun UniLAB glovebox under dinitrogen at -10 °C. Dichloromethane/0.1 M TBACl or TBABF4 was used as electrolyte support solution. To reduce the amount of water, TBACl (TBABF4) was recrystallized twice from acetone/Et2O (EtOH/Et2O). The filtered solids were washed with Et2O and dried under vacuum. 28 The typical complex concentration was 0.2 mM. To minimize the ohmic drop between the working and the reference electrodes, careful feedback correction was applied. The formal potential value (E°') corresponding to each ET process was calculated as the semi-sum of the cathodic and anodic peak potentials, E°'=(Epc+Epa)/2. The dependence of ΔEp = Epa  Epc on v allows to obtain the standard heterogeneous ET rate constant kET, 38 which is the ET rate constant measured at the formal potential E°'.
The experiments were repeated at least five times and the kET values obtained were found to be reproducible within 6%.
Magnetic measurements. Magnetic measurements were made on an MPMS5 Quantum Design instrument. A sample of 22.6CH2Cl20.84Et2O was introduced in a flame-sealed quartz tube together with a small amount of mother liquor (a mixture of CH2Cl2 and Et2O 71:29 w/w) to preserve crystallinity and avoid field-induced torqueing at low temperature. The mass of the sample and of the mother solution were determined after magnetic measurements were completed: 14.08 mg of desolvated 2 (corresponding to 17.78 mg of 22.6CH2Cl20.84Et2O) and 30.96 mg of mother solution. The diamagnetic contributions of the sample and of the mother solution were evaluated using Pascal's constants 39 and the specific susceptibilities of the solvents, respectively, while the magnetic response of the empty quartz tube was separately measured. Magnetization (M) measurements were carried out at H = 1 kOe between 2.0 and 300 K and the magnetic susceptibility was calculated as χ = M/H. The field dependence of the magnetization was also measured up to 50 kOe at 2, 3, 4 and 5 K. AC measurements were performed using a 3.0 Oe oscillating field with frequencies in the 1−1500 Hz range. EPR experimental details are available in the SI.
Theoretical calculations. Ligand field calculations were carried out within the AOM, using the AOMX program package. 40 Experimental atomic coordinates for the terminal Fe1(N1,N2,N3,Cl1) and Fe4(N13,N14,N15,Cl2) chromophores were either used directly or averaged to C3v symmetry in order to evaluate ligand-field splittings. Ligand-field parameters that account for  and  interactions were based on previous work 41  and trans-[Fe(py)4(NCS)2], which feature a tetragonally-elongated octahedral geometry. [42][43][44] The actual ligand field parameters used in the calculations were obtained from the above-reported literature values by assuming a r -6 dependence on metal-donor distance r. Interelectronic repulsion parameters were set at B = 850 cm -1 , C = 3100 cm -1 (both ca. 20% lower than the free-ion value 45 ), the effective spin-orbit coupling constant was set at 3d = 350 cm 1 and a unitary and isotropic orbital reduction factor (k) was used 44 (see Supporting Information for details). AOM calculations on Fe2 and Fe3 ions were directly based on experimental atomic coordinates of Fe2(N4,N5,N6,N7,N8) and Fe3(N8,N9,N10,N11,N12) chromophores and on the same B, C, 3d and k values as above. The ligand-field e(N) parameters were estimated from Fe-N distances treating all donors as py-type N donors for simplicity and using the same r -6 power law described above.
We carried out DFT calculations on 2 by using NRLMOL [46][47][48] and VASP, 49,50 within PBE GGA exchange-correlation functional. 51 For NRLMOL, we considered all electrons and generally contracted Gaussian basis sets (better than triple zeta) for all elements, whereas for VASP we employed plane-wave basis sets with energy cutoff of 400 eV with PAW pseudopotentials. 52,53 The experimental geometry has been used. For the bonding analysis, we employed an on-site Coulomb repulsion term, U = 2, for the Fe d orbitals, including self-consistent spin-orbit coupling to the lowest-energy state, i.e. up-up-down-down spin configuration, in VASP.

RESULTS AND DISCUSSION
Synthesis and solution studies. The key to success in the synthesis was utilizing dimesityliron, The electronic spectra of the free H2tpda ligand in dichloromethane 57 (tetrahydrofuran) are dominated by two intense absorptions at 262 (262) and 332 (336) nm, assigned to   * transitions. 58 These bands undergo a large red shift to 320 and 390 nm upon ligand deprotonation by addition of excess t BuOK in tetrahydrofuran ( Figure S4). A qualitatively similar shift is observed in the electronic spectrum of 2 in dichloromethane, which shows two intense absorption bands at 288 and 374 nm along with a weaker shoulder at 448 nm, which conveys a light yellow-brown color to the solution ( Figure S4). Upon admission of oxygen to the cuvette, the spectrum rapidly evolves and the final profile is superimposable to the spectrum of H2tpda.
Since neither ESI-MS nor UV-Vis-NIR spectroscopy directly support the stability of 2 in dichloromethane solution, we recorded an 1 H-NMR spectrum in CD2Cl2 ( Figure 1). We found six well- is detected over NMR timescale (see below). A pentachromium(II) EMAC in dichloromethane solution has been similarly found to display its maximum possible symmetry (D4) on the timescale of NMR experiment, although the solid state structure is much less symmetric. 30 Even if a complete assignment of protons pattern is not possible at this stage of investigation, the intensity ratios suggest that signal II corresponds to Ha.  Figure S5.
X-ray structure. A single-crystal X-ray diffraction study at 115(2) K (Table 1) evidenced that 22.6CH2Cl20.84Et2O contains tetrairon string-like complexes wrapped by three doubly-deprotonated all-syn H2tpda ligands and capped at their termini by two chloride ions (selected bond distances and angles are gathered in Table 2). Consistent with this formulation, Bond-Valence Sum calculations 59 confirmed that the four metal ions are in a +2 oxidation state (Table S3).  (3) Fe (1)   Beside representing a new structural type in EMAC chemistry, complex 2 qualifies as the first ironbased EMAC supported by oligo--pyridylamido ligands, although it contains one metal less than expected from ligand's structure. Interestingly, a similar situation holds with tridentate dpa -, which has so far afforded diiron complexes only. Two of them (3 and 4) were obtained as 3[Fe4O(dpa)6]·3C7H8 and 4·C7H8 by treating 1 with excess Hdpa in hot toluene. 55 The latter compound represents the main product of the reaction, with oxide and chloride ions in 3 being adventitious. In early work, from Li(dpa) and  and the standard heterogeneous ET rate constants kET associated with the redox processes are reported in Table 3. The CV curves obtained by stopping the scan after each anodic peak strictly resemble (in terms of calculated E°', peak-to-peak separation and peak currents) the traces recorded in a single scan involving all four quasi-reversible signals. This result indicates that no appreciable chemical reactions or structural rearrangements affect the observed ET processes (at least over the time scale of the electrochemical measurement).  (Table 3) and suggests the possibility of isolating a pure solid mixed-valence compound from its solution. 71 The negative E°' values are comparable to those found for other related iron-complexes. 54,72 The heterogeneous ET rate constant kET decreases with increasing oxidation state (and charge) of the complex (Table 3). This could be related to a progressive increase in the reorganization energy λ, as already observed for other mixed valence complexes. 73 Table 3. Electrochemical data from CV for the subsequent ET processes of 2 in dichloromethane at 10 °C, using TBACl 0.1 M as base electrolyte.  To get insight into the observed magnetic behavior, we performed ligand-field calculations on the four iron(II) ions within the AOM, using the AOMX program package 40 (see Supporting Information for details). The two terminal metals (Fe1 and Fe4) possess close to trigonal pyramidal geometry and their ligand-field splitting is similar to that found in iron(II) pyrrolide complexes. [74][75][76] In exactly trigonal geometry the degenerate 3dxz and 3dyz orbitals lie lowest in energy, yielding a Jahn-Teller active 5 E ground term with unquenched orbital momentum. Spin-orbit interaction then causes a large magnetic anisotropy of the easy-axis type along the trigonal axis. Deviations from trigonal symmetry produce a splitting of the 5 E term and reduce the ground-state orbital momentum. However, in both Fe1 and Fe4 20 this splitting (2  30-35 cm -1 ) is sufficiently small (Table S4) compared with the effective spin-orbit coupling constant (3d = 350 cm -1 ) 44 that its impact on calculated χMT vs T curves is negligible (see Figure   S6). 76 Therefore, in all subsequent calculations axial symmetry (C3v) was enforced on the FeN3Cl chromophores, with average metrical parameters over Fe1 and Fe4. Next, using the PHI v3.0.6 software, 77 we determined the crystal-field Hamiltonian ( , ) to be applied to the 5 D Russell-Saunders term of the terminal iron(II) ions (i = 1, 4) in order to account for the lowest 25 zero-field levels resulting from AOM calculations (Table S5 and Figure S7). In the complete Hamiltonian used to describe Fe1 and Li and Si are the orbital (Li = 2) and spin (Si = 2) angular momenta, respectively, i = 3d/2Si = 87.5 cm -1 is the multielectronic spin-orbit coupling constant, B is the applied magnetic field and ki is the orbital reduction factor (taken as unity for consistency with ref. 44 ).
The two remaining iron(II) ions (Fe2 and Fe3) have a much more distorted coordination environment, which suggests they may exhibit a quenched first-order orbital momentum. AOM calculations (see Supporting Information for details) confirmed the occurrence of a well-isolated spin quintet (Si = 2) state (Table S6) whose Zeeman splitting is accurately described within a spin Hamiltonian formalism: where and ̿ are the ZFS and the g tensors respectively for the two ions (i = 2, 3).  (Table S7). For simplicity, in all subsequent calculations both Fe2 and Fe3 were assumed to have axial anisotropy with Di = -9.12 cm -1 , gx,i = gy,i = 2.033 and gz,i = 2.195 and their easy axes were taken orthogonal to the trigonal axes of the neighboring terminal ions.
The magnetic response calculated from these single-ion parameters assuming four uncoupled metal centers is shown in Figure 5  To account for exchange coupling in our treatment, we modeled 2 as a linear array composed of two identical ferromagnetic Fe2 units (i.e. Fe1,Fe2 and Fe3,Fe4) weakly antiferromagnetically coupled with each other, as described by Hamiltonian: where the 's are defined in Eq. (1) and (2). Intradimer exchange coupling (J) was introduced as an interaction between true spins (Lines' approach). 77,78 Because of the expectedly much weaker 2-3 coupling, the model was halved to include only one Fe2 unit and interdimer interaction (Jeff) was treated in the mean-field approximation. Finally, for an accurate reproduction of high-temperature data a correction for TIP was introduced. 16, 60 The resulting best-fit parameters ( Figure 5) are J = -21.4(4) cm -1 , Jeff/gav 2 = 0.345(7) cm -1 and TIP = 2.1(2)·10 -3 emu mol -1 , where gav is the average g-factor of the Fe2 unit (assumed to be T-independent). The best-fit J value confirms that ferromagnetic coupling is present in 2, albeit it is considerably weaker than predicted by DFT calculations. The ferromagnetic interaction is however one order of magnitude stronger than in dimeric compound 5, which features a similar bridging motif. 60 It is also meaningful to compare 2 with diiron(II) compound 6, which resembles each Fe2 unit in 2 and indeed exhibits a similar tendency to parallel spin alignment. 16 In this case, however, a description of magnetic coupling in terms of localized spins is questionable due to the short Fe···Fe separation (2.29 Å) and delocalized metal-metal bonding.
In spite of the noncollinear anisotropies within the Fe1,Fe2 and Fe3,Fe4 pairs, the terminal iron(II) ions provide the largely dominant anisotropic contribution. Consequently each diiron(II) fragment has a huge easy-axis anisotropy, with a calculated susceptibility (at 2 K and 1 kOe) twenty-five times higher along the chain axis than orthogonal to it. Furthermore the small antiferromagnetic interdimer coupling is expected to lead to an array of low-lying excited states closely spaced in energy. Such a scenario and the prominent magnetic anisotropy, which is expected to lead to very extended spectra, may explain why the sample failed to give any detectable EPR signal (331.2 GHz and 441.6 GHz) in the temperature range from 10 to 30 K.
Given its magnetic anisotropy, 2 was tested for slow relaxation of the magnetization. AC magnetic measurements were thus performed on the same crystalline sample used for DC studies at temperatures down to 2 K and with a frequency (ν) of the 3 Oe oscillating magnetic field up to 1500 Hz. Since no outof-phase component of the molar magnetic susceptibility (χM) was detected in zero applied DC field, preliminary scans of χM vs ν were made at 2 K and different DC field values ranging from zero to 2.5 kOe. The results reveal the onset of slow relaxation of the magnetization above 500 Oe, with fields in the range 1.25-2.5 kOe leading to similar dynamic responses ( Figure S8). A static field of 2 kOe was chosen to perform isothermal χM vs ν scans between 2.0 and 5.0 K. A non-zero χM was clearly observed below 2.8 K, where χM increases with increasing frequency ( Figure S9). Since no maximum in the χM vs ν curves could be detected in the available frequency range, the temperature-dependent relaxation time could not be reliably extracted by fitting each curve to the (generalized) Debye model. 79,80 However, provided that only one characteristic relaxation process of the Debye type is present, with one energy barrier and one time constant, the relaxation parameters in the Arrhenius law τ = τ0exp(Ueff / kBT) can be roughly evaluated using equation 81,82 ln(χM/χM) = ln(ωτ0) + Ueff / kBT where ω = 2πν and χM is the in-phase component of the molar magnetic susceptibility. Plotting ln(χM/ χM) vs 1/T and performing a linear regression at each frequency above ν = 65 Hz ( Figure S10 Building up from the foregoing results, we envision that a further strengthening of ferromagnetic interactions may follow from the incorporation of an additional tpda 2ligand and a fifth iron(II) ion in the structure. 19 The resulting [Fe5(tpda)4Cl2] complex, which would most probably be isostructural to its known Cr, Co, Ni and Ru congeners 1-3,65 and would contain Fe-Fe bonds, was occasionally detected in ESI-MS spectra but never isolated. An exciting alternative is partial chemical oxidation of 2, a feasible process according to our CV studies. In fact, a mechanism of electronic interaction known as double exchange becomes operative in mixed-valent systems and generally provides a very effective source of ferromagnetic coupling between ions in different oxidation states. [84][85][86] Work in both directions is currently underway to enforce thermally-persistent high-spin states in iron-based EMACs.
ASSOCIATED CONTENT

Supporting Information
The Supporting Information is available free of charge on the ACS Publications website at DOI: XXX

Author Contributions
The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript.

Notes
The authors declare no competing financial interest.