Substituent-Guided Cluster Nuclearity for Tetranuclear Iron(III) Compounds with Flat {Fe4(μ3-O)2} Butterfly Core

The tetranuclear iron(III) compounds [Fe4(μ3-O)2(μ-LZ)4] (1–3) were obtained by reaction of FeCl3 with the shortened salen-type N2O2 tetradentate Schiff bases N,N’-bis(salicylidene)-o-Z-phenylmethanediamine H2LZ (Z = NO2, Cl and OMe, respectively), where the one-carbon bridge between the two iminic nitrogen donor atoms guide preferentially to the formation of oligonuclear species, and the ortho position of the substituent Z on the central phenyl ring selectively drives towards Fe4 bis-oxido clusters. All compounds show a flat almost-symmetric butterfly-like conformation of the {Fe4(μ3-O)2} core, surrounded by the four Schiff base ligands, as depicted by both the X-ray molecular structures of 1 and 2 and the optimized geometries of all derivatives as obtained by UM06/6-311G(d) DFT calculations. The strength of the antiferromagnetic exchange coupling constants between the iron(III) ions varies among the three derivatives, despite their magnetic cores remain structurally almost unvaried, as well as the coordination of the metal ions, with a distorted octahedral environment for the two-body iron ions, Feb, and a pentacoordination with trigonal bipyramidal geometry for the two-wing iron ions, Few. The different magnetic behavior within the series of examined compounds may be ascribed to the influence of the electronic features of Z on the electron density distribution (EDD) of the central {Fe4(μ3-O)2} core, substantiated by a Quantum Theory of Atoms In Molecules (QTAIM) topological analysis of the EDD, as obtained by UM06 calculations 1–3.


Introduction
The need for miniaturized devices for technological applications is driving scientific research into the replacing of existing materials with molecular species of nanometer sizes [1][2][3][4]. Polynuclear compounds of transition metal ions are very promising materials for this purpose because of their tunable electronic and magnetic properties [5][6][7][8], achieved through the modulation of the magnetic exchange interactions. In the search for this variability, until a few years ago, the serendipitous assembly approach led to a huge increase in the amount of synthesized polynuclear complexes with different bridging ligands [9], but a rational approach is most desirable, with precise control of the obtained compounds and their properties by the modification of the ligands [10] and the synthetic conditions [11][12][13]. and their properties by the modification of the ligands [10] and the synthetic conditions [11][12][13].
Schiff bases are well-known ligands that have been used for decades in the synthesis of metal complexes for applications in different fields, such as catalysis [14], new materials and optics [15][16][17]. From the most famous N2O2 tetradentate Schiff base H2salen, formed by the condensation of salicylaldehyde (salH) with ethylenediamine (en) [18][19][20], new functionalized families of structurally modified molecules have been designed for the purpose of obtaining new polynuclear complexes with desired structural features and functionalities [21][22][23]. In particular, the substitution of en with the shorter methylenediamine in condensation with salH yields ligands (called H2salben's when the methylene bridge between the two nitrogen atoms carries a phenyl ring) that preferentially produce oligonuclear compounds by complexation [24][25][26][27].
Based on our previous experience in working with Schiff bases and the preferred isolation of oligonuclear metal clusters [8,[24][25][26][27][28][29][30][31], in this work, we decided to exploit the metal complexation ability toward iron(III) of three H2sal(o-Z)ben derivatives bearing a substituent Z in the ortho position of the central phenyl ring (N,N'-bis(salicylidene)-o-Zphenylmethanediamine) [24,32] (abbreviated as H2L Z in this paper) with Z going from NO2 to Cl and OMe (Scheme 1). The tetranuclear iron(III) compounds [Fe4(μ3-O)2(μ-L Z )4] (1-3, with Z = NO2, Cl, OMe, respectively, see Scheme 1) could be selectively obtained by reaction with FeCl3 in contrast with the formation of dinuclear species [Fe2((μ-OMe)2(μsal(p-Y)ben)2] when ligands with substituents Y in para position or absent were employed [27]. Here, we report the synthesis and the structural and magnetic characterization of these complexes, revealing the modulation of Z on the exchange coupling constants between the iron ions within the three derivatives. The topological analysis of the computed electron density distribution (EDD) of 1-3 will help to discuss the electronic effect of Z. Scheme 1. Synthetic scheme of 1-3 (Z = NO2, Cl, OMe, respectively) and drawing of the arrangement of the Schiff base ligands around the {Fe4(μ3-O)2} butterfly core (sal moieties are sketched as  for clarity; iron ions are divided into those of two body, Feb, and two wing, Few, ions, as applicable in butterfly-like Fe4 clusters).

Synthesis and X-ray Crystal Structures
The reaction of H2L Z with FeCl3 in non-anhydrous MeOH or MeCN in the presence of NEt3 as a base gave the tetranuclear complexes 1-3 good yields (see Scheme 1), with the best ones in MeOH. The polynucleation ability of the salben ligands due to the one-carbon bridge between the two iminic nitrogen atoms is clearly confirmed here [24][25][26][27]. There is also the control on the nature of the oligonuclear compound upon substituent shift, since the dinuclear species [Fe2((μ-OMe)2(μ-sal(p-Y)ben)2] are isolated when ligands with substituents Y in para position or absent were employed [27]. The tetranuclear nature of 1-3 is proven by the ESI + mass spectra with the [M+Na] + and [M+1] + peaks and the fragmentation of the [Fe3(O)(L Z )3] + ions (see Experimental Section). Furthermore, single crystals Scheme 1. Synthetic scheme of 1-3 (Z = NO 2 , Cl, OMe, respectively) and drawing of the arrangement of the Schiff base ligands around the {Fe 4 (µ 3 -O) 2 } butterfly core (sal moieties are sketched as ∩ for clarity; iron ions are divided into those of two body, Fe b , and two wing, Fe w , ions, as applicable in butterfly-like Fe 4 clusters).

Synthesis and X-ray Crystal Structures
The reaction of H 2 L Z with FeCl 3 in non-anhydrous MeOH or MeCN in the presence of NEt 3 as a base gave the tetranuclear complexes 1-3 good yields (see Scheme 1), with the best ones in MeOH. The polynucleation ability of the salben ligands due to the one-carbon bridge between the two iminic nitrogen atoms is clearly confirmed here [24][25][26][27]. There is also the control on the nature of the oligonuclear compound upon substituent shift, since the dinuclear species [Fe 2 ((µ-OMe) 2 (µ-sal(p-Y)ben) 2 ] are isolated when ligands with substituents Y in para position or absent were employed [27]. The tetranuclear nature of 1-3 is proven by the ESI + mass spectra with the [M+Na] + and [M+1] + peaks and the fragmentation of the [Fe 3 (O)(L Z ) 3 ] + ions (see Experimental Section). Furthermore, single crystals suitable for the X-ray structure determination of 1·1.5iPr 2 O and 2·2iPr 2 O were obtained, and the crystal structures could be refined.
The asymmetric unit of 1·1.5iPr 2 O contains two independent clusters [Fe 4 (µ 3 -O) 2 (µ-L NO2 ) 4 ], A and B, structurally very similar (see Figure S1 for molecule A and an overlap of A and B in Figure S2a in Supplementary Materials), and three iPr 2 O solvent molecules, clathrated in the lattice voids. For 2·2iPr 2 O, in contrast, only one independent molecule of [Fe 4 (µ 3 -O) 2 (µ-L Cl ) 4 ] and two iPr 2 O molecules are present in the asymmetric unit ( Figure 1). Both molecules of 1 and 2 show four iron(III) ions linked together by two triply-bridging oxido anions, giving the {Fe 4 (µ 3 -O) 2 } core. The four dianionic tetradentate ligands sal(o-Z)ben 2surround the core, bridging two iron ions each, as shown in Scheme 1 and Figure 1 and Figure S1 in Supplementary Materials. suitable for the X-ray structure determination of 1·1.5iPr2O and 2·2iPr2O were obtained, and the crystal structures could be refined. The asymmetric unit of 1·1.5iPr2O contains two independent clusters [Fe4(μ3-O)2(μ-L NO2 )4], A and B, structurally very similar (see Figure S1 for molecule A and an overlap of A and B in Figure S2a in Supplementary Materials), and three iPr2O solvent molecules, clathrated in the lattice voids. For 2·2iPr2O, in contrast, only one independent molecule of [Fe4(μ3-O)2(μ-L Cl )4] and two iPr2O molecules are present in the asymmetric unit ( Figure 1). Both molecules of 1 and 2 show four iron(III) ions linked together by two triply-bridging oxido anions, giving the {Fe4(μ3-O)2} core. The four dianionic tetradentate ligands sal(o-Z)ben 2-surround the core, bridging two iron ions each, as shown in Scheme 1 and Figures 1 and S1 in Supplementary Materials. In such clusters, the {Fe4(μ3-O)2} core adopts a so-called butterfly-like conformation where the iron(III) ions are usually divided into two 'body', Feb (Fe2, Fe3), and two 'wing', Few (Fe1, Fe4), metal ions [33]. In this butterfly-like conformation, the Fe2-Fe3 fragment features the body of the butterfly, and the Fe1Fe2Fe3 and Fe2Fe3Fe4 triangles schematize the wings, with the Fe1 and Fe4 occupying the tip positions. The Fe4O2 moieties are essentially planar, with maximum displacements from the least square (l.s.) planes of only 0.082 (Fe4A) and 0.064 Å (Fe4B) for 1, and 0.027 Å (O17) for 2 (see Figures 1 and S1 in Supplementary Materials). Another proof of the planarity of the metal cores can be detected through the dihedral angles between the l.s. planes defined by the two wings, which are 176.9 and 176.2° (molecules A and B, respectively) in 1 and 179.1° in 2, very close to the ideal value of 180°. There are few other cases that have been reported in which the two wings lie in the same plane [34][35][36][37], but they are usually associated with higher displacement values of the oxido ions from the Fe3 planes, which is not the case here.
The μ3-O bridges in 1 are asymmetric: the Few-O distances (average 1.790 Å) are shorter than the Feb-O ones (average 1.960 Å). Similar asymmetry is also present in 2, with a short Few-O average distance of 1.802 Å and a long Feb-O average distance of 1.945 Å (see Table S1 in Supplementary Materials). This feature is in line with values reported in the literature for butterfly-like Fe4O2 clusters [33], in which the Few-O distances of our compounds are among the shortest observed.  Table S1 in Supplementary Materials for all values). The opening of the Feb-O-Feb angles in 2 leads to a slightly longer Feb···Feb distance (2.9683(4) Å) compared to 1 (2.929(2) and 2.937(2) Å) and to an asymmetry in the Few position that brings Fe1 and Fe4 closer to Fe3, compared to In such clusters, the {Fe 4 (µ 3 -O) 2 } core adopts a so-called butterfly-like conformation where the iron(III) ions are usually divided into two 'body', Fe b (Fe2, Fe3), and two 'wing', Fe w (Fe1, Fe4), metal ions [33]. In this butterfly-like conformation, the Fe2-Fe3 fragment features the body of the butterfly, and the Fe1Fe2Fe3 and Fe2Fe3Fe4 triangles schematize the wings, with the Fe1 and Fe4 occupying the tip positions. The Fe 4 O 2 moieties are essentially planar, with maximum displacements from the least square (l.s.) planes of only 0.082 (Fe4A) and 0.064 Å (Fe4B) for 1, and 0.027 Å (O17) for 2 (see Figure 1 and Figure S1 in Supplementary Materials). Another proof of the planarity of the metal cores can be detected through the dihedral angles between the l.s. planes defined by the two wings, which are 176.9 and 176.2 • (molecules A and B, respectively) in 1 and 179.1 • in 2, very close to the ideal value of 180 • . There are few other cases that have been reported in which the two wings lie in the same plane [34][35][36][37], but they are usually associated with higher displacement values of the oxido ions from the Fe 3 planes, which is not the case here.
The asymmetry in the wing iron ion positions and the different rotation of the tb polyhedra can probably be explained by crystal packing and/or the steric and electronic effects of the Z substituents. In particular, the arrangement of the four sal(o-Cl)ben 2ligands around the {Fe 4 (µ 3 -O) 2 } core in 2 is noteworthy, where each of the chlorine atoms points towards the center of the adjacent phenyl ring, giving rise to a C-Cl···π halogen bond [52] (average distance of Cl from the l.s. plane through the phenyl ring = 3.66 Å, see Figure 1), due to the interaction between the positive region of the electrostatic potential on the halogen along the extension of the C-Cl bond with the π system of the phenyl ring [53][54][55]. On the other hand, the nitro group in 1 is unable to produce a stabilizing interaction with the π system of the adjacent phenyl ring, determining a more irregular arrangement of the four sal(o-NO 2 )ben 2ligands with respect to the {Fe 4 (µ 3 -O) 2 } core (see Figure S1 in Supplementary Materials). This feature also leads to slightly different conformations of the surrounding ligands, mainly affecting the central phenyl ring carrying the substituent Z, as is clearly perceivable by superimposing the two structures (see Figure S2b in Supplementary Materials for the overlap of 1A and 2).
The Fe-O bonds involving the phenoxido oxygen atoms of the Schiff base ligands are in the range 1.865(5)-1.964(5) Å for 1 and 1.8660(12)-1.9698(13) Å for 2, with usually shorter distances when the wing Fe1 and Fe4 ions are involved. Fe-N bonds are longer than Fe-O ones, ranging from 2.160(6) to 2.234(6) Å in 1 and from 2.1509(15) to 2.2730(13) Å in 2, but they show the same trend with respect to Fe w and Fe b . It is instead hard to discern a direct effect of the substituent Z on the coordination distances.

Optimized Geometries
Aimed at confirming the stability of 3 and ascertaining the influence of Z, if any, on the coordination distances of the iron ions, UM06/6-311G(d) calculations were performed on 1-3. Selected distances are reported in Table 1 in direct comparison with the corresponding experimental ones. The resulting optimized structures are essentially symmetric, unlike the experimental ones, which show a slight degree of asymmetry as described above, probably due to crystal packing effects. The comparison of the Fe w -O and Fe b -O computed distances with the corresponding experimental average values in 1 and 2 indicates an overall good agreement, with just a little overestimate for the former (1.818 and 1.825 Å vs. 1.790(5) and 1.802(11) Å, respectively) and a better correspondence for the latter (1.951 and 1.947 Å vs. 1.960(5) and 1.945(11) Å, respectively (see Figure S2c ions when going from 1 to 3. The geometry of the {Fe 4 (µ 3 -O) 2 } core is only slightly different in the three complexes, with a lengthening of the Fe w ···Fe w distance up to 0.028 Å and shortening of the Fe b ···Fe b one equal to 0.023 Å going from the electron-withdrawing NO 2 group to the electron-donating OMe group. As already evidenced in previous studies on copper(II) complexes of salen analogues with push-pull structures [15], NO 2 exerts the strongest electronic effect on the coordination geometry with respect to OMe. It is worth noting that the calculations correctly reproduce the Cl···π interaction observed in the crystal structure, providing Cl distances from the ring centroids equal to 3.668 and 3.660 Å. This agreement can be ascribed to the ability of the M06 functional to describe weak dispersion-dominated interactions, unlike the B3LYP functional, which has been shown to be unable to model the C-X···π halogen bond [53][54][55]. According to previous CCSD(T) calculations at the complete basis set limit on model systems constituted by benzene and DCl dimers (D = H, HCC, F and NC) in a T-shaped configuration [54], the energy associated with this interaction ranges from -1.43 to -3.38 kcal mol -1 at the respective equilibrium distances (from 3.45 to 3.25 Å), according to the electron-withdrawing ability of the D group. It is then expected that the C-Cl···π interaction in 2, while present, is however quite weak, owing to the observed large Cl···π distance and the non-optimal T-shaped arrangement of the interacting species.
By analyzing the conformation of the ligands in the optimized geometries of 1-3 (see Figure S2e in Supplementary Materials), it is apparent the different orientation of the central phenyl ring carries the substituent Z, similar to the experimental structures, while the cores and the coordination environments of the four iron ions are almost superimposed. Moreover, the computed reciprocal rotations of the two tb polyhedra with respect to each other (78.5, 81.1 and 83.5 • for 1, 2 and 3, respectively) and with respect to the Fe 4 O 2 plane (50.8, 49.5 and 48. 3 • for 1, 2 and 3, respectively) show a lower variability with respect to the X-ray crystal structures but follow the same experimental trend. This indicates that, besides crystal packing effects, the intrinsic electronic and steric effects induced by Z are also responsible for the observed conformational differences between 1 and 2 in the solid state.

Magnetic Properties
The magnetic susceptibility of 1-3 was measured between 5 and 300 K. The product χ M T (χ M is the molar susceptibility referred to an Fe 4 unit) of 1 (Figure 2a) is 16.5 ± 0.3 emu K mol -1 Oe -1 at 300 K, which is slightly lower than that for four uncoupled iron(III) ions (17.5 emu K mol -1 Oe -1 , g = 2.00). On cooling, χ M T of 1 decreases slowly down to 50 K and more rapidly at lower temperature up until it reaches the value 1.55 ± 0.03 emu K mol -1 Oe -1 at 5 K. Such behavior suggests that the iron(III) ions are antiferromagnetically coupled and that the coupling is weak. energy associated with this interaction ranges from -1.43 to -3.38 kcal mol -1 at the respective equilibrium distances (from 3.45 to 3.25 Å), according to the electron-withdrawing ability of the D group. It is then expected that the C-Cl···π interaction in 2, while present, is however quite weak, owing to the observed large Cl···π distance and the non-optimal T-shaped arrangement of the interacting species.
By analyzing the conformation of the ligands in the optimized geometries of 1-3 (see Figure S2e in Supplementary Materials), it is apparent the different orientation of the central phenyl ring carries the substituent Z, similar to the experimental structures, while the cores and the coordination environments of the four iron ions are almost superimposed. Moreover, the computed reciprocal rotations of the two tb polyhedra with respect to each other (78.5, 81.1 and 83.5° for 1, 2 and 3, respectively) and with respect to the Fe4O2 plane (50.8, 49.5 and 48. 3° for 1, 2 and 3, respectively) show a lower variability with respect to the X-ray crystal structures but follow the same experimental trend. This indicates that, besides crystal packing effects, the intrinsic electronic and steric effects induced by Z are also responsible for the observed conformational differences between 1 and 2 in the solid state.

Magnetic Properties
The magnetic susceptibility of 1-3 was measured between 5 and 300 K. The product χMT (χM is the molar susceptibility referred to an Fe4 unit) of 1 (Figure 2a) is 16.5 ± 0.3 emu K mol -1 Oe -1 at 300 K, which is slightly lower than that for four uncoupled iron(III) ions (17.5 emu K mol -1 Oe -1 , g = 2.00). On cooling, χMT of 1 decreases slowly down to 50 K and more rapidly at lower temperature up until it reaches the value 1.55 ± 0.03 emu K mol -1 Oe -1 at 5 K. Such behavior suggests that the iron(III) ions are antiferromagnetically coupled and that the coupling is weak. The product χ M T of 2 and 3 (Figure 2b,c) at 300 K is 6.04 ± 0.02 and 10.0 ± 0.2 emu K mol -1 Oe -1 , respectively, which is much lower than that of four uncoupled Fe 3+ ions. For both 2 and 3, χ M T decreases upon cooling but, differently from 1, the decrease is much less curved between 300 and 5 K. Such results suggest that the antiferromagnetic interactions between the iron ions are stronger in 2 and 3 than in 1, a conclusion seemingly contrasting with the larger χ M T values at low temperature (2.00 ± 0.01 and 2.91 ± 0.06 emu K mol -1 Oe -1 at 5 K for 2 and 3, respectively).
Several super-exchange pathways magnetically connect the wing and body iron ions of the planar {Fe 4 (µ 3 -O) 2 } core in 1-3 (see Scheme 2). Thanks to the symmetry of the cores, all wing-wing super-exchange interactions can be considered equal, and the same assumption can be made for wing-body and body-body interactions. Besides, any magnetic communication through the Schiff base ligands can be neglected because of the longer exchange pathway. Within these approximations, the spin system of 1-3 can be modelled as a butterfly spin system [56] where the exchange couplings are fully determined by the three coupling constants J bb (body-body), J wb (wing-body), and J ww (wing-wing). In complexes 1-3, the wing-wing exchange can be neglected with respect to the wing-body and body-body interactions because of the longer exchange pathway. By this further assumption, we can set J ww = 0 and obtain a simplified butterfly spin system described by the Hamiltonian: (1) exchange pathway. Within these approximations, the spin system of 1-3 can be modelled as a butterfly spin system [56] where the exchange couplings are fully determined by the three coupling constants Jbb (body-body), Jwb (wing-body), and Jww (wing-wing). In complexes 1-3, the wing-wing exchange can be neglected with respect to the wing-body and body-body interactions because of the longer exchange pathway. By this further assumption, we can set Jww = 0 and obtain a simplified butterfly spin system described by the Hamiltonian: This Hamiltonian can be treated by the Kambe vector coupling method [57], and the energy of the spin states can be expressed as where the spin state |Sbb, Sww, S is labelled as to the two intermediate (Ŝbb = Ŝb1 + Ŝb2 and Ŝww = Ŝw1 + Ŝw2) and the total (Ŝ = Ŝbb + Ŝww) spin operators. Molar susceptibility χM was calculated by inserting these spin state energies into the Van Vleck equation [56] and was then fitted to the experimental susceptibility data by varying Jbb and Jwb. In all cases, the gvalue was fixed to 2.00, as appropriate for high spin iron(III). This model could be successfully applied to the experimental data of 1-3 (Figure 2), confirming the starting assumptions.

Scheme 2.
Magnetic exchange pathways in the simplified butterfly Fe4 spin system considered during the fitting of the experimental data (Jww = 0, see text).
Before discussing the best-fit results, we need to examine their reliability since the Jbb values obtained from the magnetic susceptibility data of butterfly-like {Fe4(μ3-O)2} complexes can be largely under-determined [37,43]. This occurs because at low temperature, only |5,5,S states are significantly populated, which differ in energy by This Hamiltonian can be treated by the Kambe vector coupling method [57], and the energy of the spin states can be expressed as E(|S bb , S ww , S ) = -J bb [S bb (S bb + 1) − 35/2] − J wb [S(S + 1) − S bb (S bb + 1) − S ww (S ww + 1)] (2) where the spin state |S bb , S ww , S is labelled as to the two intermediate (Ŝ bb =Ŝ b1 +Ŝ b2 andŜ ww =Ŝ w1 +Ŝ w2 ) and the total (Ŝ =Ŝ bb +Ŝ ww ) spin operators. Molar susceptibility χ M was calculated by inserting these spin state energies into the Van Vleck equation [56] and was then fitted to the experimental susceptibility data by varying J bb and J wb . In all cases, the g-value was fixed to 2.00, as appropriate for high spin iron(III). This model could be successfully applied to the experimental data of 1-3 (Figure 2), confirming the starting assumptions. Before discussing the best-fit results, we need to examine their reliability since the J bb values obtained from the magnetic susceptibility data of butterfly-like {Fe 4 (µ 3 -O) 2 } complexes can be largely under-determined [37,43]. This occurs because at low temperature, only |5,5,S states are significantly populated, which differ in energy by Thus, a change in J bb does not affect the energy of the low-lying states and, hence, the magnetic susceptibility χ M is scarcely sensitive to J bb [35]. To analyze the confidence level of the best-fit parameters, we computed the sum of squared weighted residuals (ssr) for a region of the (J bb , J wb ) plane about the best-fit values and found the constant-value contours of the difference ∆ssr = ssr − ssr min , i. e., the deviation from the minimum (best-fit) ssr (see Figure 2). The ∆ssr contours show a well-defined minimum in the (J bb , J wb ) plane, proving that the best fit J bb and J wb values are reliable. Realistic estimates of the parameter errors could be obtained by projecting the appropriate constant ∆ssr contours on the J bb and J wb axes [58], and they are reported in Table 2. Both J bb and J wb were statistically significant for all the three complexes. The reason for this reliability lies in the peculiar spin level structure of 1-3 that we describe below. Table 2. Best-fit parameters J bb and J wb (cm -1 ), J bb /J wb ratio, and spin ground state for 1-3 obtained by fitting the butterfly model to the experimental χ M T data. Both J wb and J bb of 1 are weaker than any tetranuclear iron butterfly complex collected in the literature [33], where J wb ranges from -91.0 to -37.2 cm -1 , and J bb ranges from -21.8 to -1.2 cm -1 . Furthermore, J wb and J bb show comparable strength (see Table 2). The J bb /J wb ratio of 1 is much larger than that of the reported butterfly {Fe 4 (µ 3 -O) 2 } complexes (J bb /J wb ranging from 0.031 to 0.265) [33], with a single exception (J bb /J wb = 1.48) [35]. The ground state of 1 is the |5,5,0 singlet state, where the body-body antiferromagnetic interaction is completely frustrated [43]. The energy separation between spin states is small, and many of them are populated even at 5 K (see Figure S3 in Supplementary Materials), including states with S bb < 5, because of the weakness of both coupling constants. Thus, χ M significantly depends on J bb through the energy difference between such states and the |5,5,0 ground state (cfr. Equation (3)). That is the reason why J bb could be reliably determined from the susceptibility data of 1 [35].

(NO 2 ) 2 (Cl) 3 (OMe)
The J bb and J wb of both 2 and 3 are in line with previously reported values [33] when separately considered, but the body-body exchange interaction is unusually strong compared to the wing-body interaction. Therefore, the body-body exchange interaction is not completely frustrated, resulting in the |3,5,2 ground state and the |4,5,1 and |2,5,3 lowest excited states. Thus, the spin states with unequal S bb are populated even at low temperature, again ensuring a reliable determination of J bb . The J bb /J wb ratio (ca. 1.7 for both 2 and 3) is the highest obtained, and 2 and 3 are among the very few examples of compounds with a {Fe 4 (µ 3 -O) 2 } core possessing a non-singlet ground state (quintet S = 2), in addition to the reported complex with J bb /J wb > 1 and S = 1 [35]. Different spin ground states can be instead stabilized for Mn 4 O 2 clusters possessing a butterfly core structure [60,61].
It is not then apparent that a magneto-structural correlation [62] exists between the exchange coupling constants and spin state structure displayed by 1-3 and the structural features of their {Fe 4 (µ 3 -O) 2 } cores. The pentacoordination of Fe w for sure introduces a novelty that affects the coupling constants J bb and J wb and the spin ground state, but this cannot be the discriminating factor, since all derivatives show this feature. The different orientation of the tb polyhedra with respect to the Fe 4 O 2 plane between 1 on one side and 2 and 3 on the other side should not in principle affect the super-exchange magnetic communication path since it mainly depends on Fe-O distances and angles. The latter in any case show a certain non-negligible degree of variability among the three complexes, even if small, that alter the magnetic core. Furthermore, the presence of Cl···π interactions in 2 might further modulate its magnetic features.

Effect of the Substituents and QTAIM Analysis
The main differentiating element in 1-3 is for sure the substituent Z on the surrounding ligands, which displays a different electron demand, as shown by the Hammettlike σ ortho constants [63,64]. Small, similar J constants are found when Z is strongly electron-withdrawing (NO 2 , σ ortho = +0.8), whereas moderately electron-withdrawing (Cl, σ ortho = +0.2) and electron-donating (OMe, σ ortho = −0.39) substituents are associated with more negative coupling constants and the peculiar J bb /J wb > 1. It is thus tempting to attribute (at least in part) the different magnetic properties of 1-3 to the different electronic availability in the {Fe 4 (µ 3 -O) 2 } core, as we previously observed in electronically modulated Cu 3 (µ 3 -OH) trinuclear copper(II) complexes [8].
We have then performed a QTAIM topological analysis [65] of the computed electron density distribution (EDD) of 1-3, aimed at determining a possible effect of the substituents Z on the EDD features of the metal core [66]. In Table 3   By looking at more detailed information as given by the values of the topological parameters at the Bond Critical Points (BCP) of the central core (see Table 4), we note a rather significant variation in the ρ BCP values, and then in the bond strength, of the Fe-O bonds in the three compounds, thus emphasizing the less apparent trend as given by the bond lengths. The strength of the Fe b -O bonds increases from 1 to 3, while that of the Fe w -O bonds decreases within the series, to a greater extent with respect to the Fe b -O bonds. The ratio of the ρ BCP values for the two types of bonds is in fact 0.706 for 1 and 0.742 for 3. Owing to the large discrepancy between the experimental and theoretical ρ BCP values obtained for the Fe-O bonds in the previously reported carboxylate-bridge butterfly-like complex [34], it is quite problematic to draw some conclusion on the different density values obtained for the Fe-O bonds in 1-3 with respect to other {Fe 4 (µ 3 -O) 2 } butterfly-like compounds. It is, however, interesting to note that the ρ BCP ratios obtained for the two types of bonds in 1-3 are significantly larger than the corresponding value previously reported, 0.66 from both experiment and theory [34]. Finally, the character of the Fe-O bonds, as given by the Laplacian ∇ 2 ρ BCP and the local energy densities G BCP , V BCP and H BCP = G BCP + V BCP , is essentially unvaried along the series 1-3, all quantities being indicative of a predominantly closed-shell character for these bonds (∇ 2 ρ BCP > 0) with quite a small covalence degree, as given by negative but close to zero H BCP values.

General
All used chemicals were reagent grade, and solvents were used as received (Sigma Aldrich, Europe). Elemental analyses were performed at the Microanalytical Laboratory at the Università degli Studi di Milano. ESI-MS spectra were recorded on MeOH or MeCN solutions with a LCQ Advantage Thermofluxional instrument. Infrared spectra were recorded as KBr disks using a JASCO FT-IR 410 spectrophotometer with a 2 cm -1 resolution. Schiff bases H 2 L Z (Z = NO 2 , Cl, OMe) were obtained following the synthetic method A reported in the literature [32].   Datasets were corrected for Lorentz-polarization effects and for absorption (SAD-ABS [67]). The structures were solved by direct methods (SIR-97 [68]) and completed by iterative cycles of full-matrix least squares refinement on F o 2 and ∆F synthesis using the SHELXL-97 [69] program (WinGX suite) [70]. Hydrogen atoms, located on the ∆F maps, were allowed to ride on their carbon atoms. Crystallographic data for 1·1.5iPr 2 O and 2·2iPr 2 O (excluding structure factors) were deposited into the Cambridge Crystallographic Data Centre as supplementary publication no. 883278 and 883279, respectively. These data can be obtained free of charge via www.ccdc.cam.ac.uk/conts/retrieving.html (or from CCDC, 12 Union Road, Cambridge CB2 1EZ, UK; Fax: +44 1223 336033; e-mail: deposit@ccdc.cam.ac.uk).

Theoretical Calculations
Unrestricted DFT calculations were performed on 1-3 at the highest spin multiplicity 21 (S = 10), assuming all iron(III) ions in their high spin state (S = 5/2), using Gaussian 09 [71]. All the structures were optimized in vacuo with the 6-311G(d) basis set, starting from the X-ray geometries for 1 (molecule A) and 2, and using 2 with the appropriated substitutions as a guess starting geometry for 3. The functionals B3LYP [72,73] and M06 [74] were tested to check their performances in reproducing the X-ray structures of 1 and 2. The latter functional, which was properly developed to treat organometallic complexes, provided the best agreement with the experimental geometries (see Tables S1 and S2 in Supplementary Materials for a full comparison between computed and experimental bond distances). Quantum Theory of Atoms In Molecules (QTAIM) [65], as implemented in AIMAll [75] was used for evaluating the net charges of the iron and oxido ions.

Magnetic Measurements
The magnetic moment µ of powder samples of 1-3 was measured between 5 and 300 K using a Quantum Design MPMS XL-5 SQUID magnetometer. Weighed amounts (about 15 mg) of 1-3 were sealed in polycarbonate capsules, and the magnetic moment µ was measured under an applied magnetic field of 1 kOe from 300 K down to 5 K. The molar susceptibility was obtained as χ M = (µ/H) × (MW/m), where MW is the molecular weight of the complex, m is the sample mass, and H is the applied magnetic field. Diamagnetic contributions were subtracted from µ before calculating χ M . The ligand diamagnetism was estimated using Pascal's constants [76].

Conclusions
The reaction of FeCl 3 with the Schiff bases H 2 L Z (Z = NO 2 , Cl, OMe) led selectively to the tetranuclear iron compounds 1-3, confirming the oligonucleation ability of these shortened ligands and the control of the nature of the oligonuclear compound upon substituent shift in ortho position of the central aryl ring. The analysis of the X-ray molecular structures of 1 and 2 and of the unrestricted-DFT computed geometries of all derivatives shows the similarity of the planar {Fe 4 (µ 3 -O) 2 } cores, together with intramolecular Cl···π halogen bonds in 2. In contrast, the J bb and J wb exchange coupling constants in 1-3 are different, and the unusually large J bb /J wb ratio for 2 and 3 suggests a quintet (S = 2) ground state for these compounds, which add up to another reported example of triplet (S = 1) ground state [35] being different from the singlet ground state (S = 0) usually reported for butterfly tetrairon(III) complexes [33].
The cores are slightly geometrically affected by the nature of the substituent Z, with the main differentiating factor given by the orientation of the tb polyhedra of the Fe w ions with respect to the Fe 4 O 2 plane. The different spin ground states and the great variability of the J values can hardly be attributed to this structural difference. The unequal donating or accepting power of the substituent Z, which changes the EDD at the {Fe 4 (µ 3 -O) 2 } core as confirmed by the topological QTAIM analysis, together with the different conformations of the surrounding ligands due to crystal packing and/or the steric and electronic effects of Z, may contribute to explaining the different magnetic properties of compounds 1-3.
Further studies in this respect are certainly necessary and are planned in order to shed more light on our findings. In any case, the latter represents a solid starting point for investigating new butterfly Fe 4 clusters with different substituents on the surrounding Schiff bases to fully exploit the potentiality of these shortened salen-type ligand molecules as potent oligonucleating and magnetically modulating agents.