Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
J Am Chem Soc. Author manuscript; available in PMC 2010 April 29.
Published in final edited form as:
PMCID: PMC2675788

Mössbauer and DFT Study of the Ferromagnetically Coupled Diiron(IV) Precursor to a Complex with an FeIV2O2 Diamond Core


Recently we reported the reaction of the (μ-oxo)diiron(III) complex 1 ([FeIII2(μ-O)(μ-O2H3)(L)2]3+, L = tris(3,5-dimethyl-4-methoxypyridyl-2-methyl)amine) with 1 equivalent of H2O2 to yield a diiron(IV) intermediate, 2 (Xue, Fiedler, Martinho, Münck, and Que, Proc. Natl. Acad. Sci. U.S.A.2008, 105, 20615-20). Upon treatment with HClO4 complex 2 converted to a species with an FeIV2(μ-O)2 diamond core that serves as the only synthetic model to date for the diiron(IV) core proposed for intermediate Q of soluble methane monooxygenase. Here we report detailed Mössbauer and density functional theory (DFT) studies of 2. The Mössbauer studies reveal that 2 has distinct FeIV sites, a and b. Studies in applied magnetic fields show that the spins of sites a and b (Sa = Sb = 1) are ferromagnetically coupled to yield a ground multiplet with S = 2. Analysis of the applied field spectra of the exchange-coupled system yields for site b a set of parameters that matches those obtained for the mononuclear [LFeIV(O)(NCMe)]2+ complex, showing that site b (labeled FeO) has a terminal oxo group. Using the zero-field splitting parameters of [LFeIV(O)(NCMe)]2+ for our analysis of 2, we obtained parameters for site a that closely resemble those reported for the non-oxo FeIV complex [(β-BPMCN)FeIV(OH)(OOtBu)]2+, suggesting that a (labeled FeOH) coordinates a hydroxo group. A DFT optimization performed on 2 yielded an Fe-Fe distance of 3.39 Å and an Fe-(μ-O)-Fe angle of 131 degrees, in good agreement with the results of our previous EXAFS study. The DFT calculations reproduce the Mössbauer parameters (A-tensors, electric field gradient and isomer shift) of 2 quite well, including the observation that the largest components of the electric field gradients of FeO and FeOH are perpendicular. The ferromagnetic behavior of 2 is puzzling given that the Fe-(μ-O)-Fe angle is large but can be explained by noting that the orbital structures of FeO and FeOH are such that the unpaired electrons at the two sites delocalize into orthogonal orbitals at the bridging oxygen, rationalizing the ferromagnetic behavior of 2. Inequivalent coordinations at FeO and FeOH define magnetic orbitals favorable for ferromagnetic ineractions.


The conversion of methane to methanol in biological systems is carried out by methane monooxygenase (MMO).1 The soluble form of this enzyme, sMMO, contains a nonheme diiron center in its active site that has been characterized thoroughly by kinetic, crystallographic and spectroscopic methods.2-5 The enzyme reacts with molecular oxygen to generate a series of intermediates; two that have attracted the most interest are a (μ-peroxo)diiron(III) species called P (or Hperoxo)6-8 and a diiron(IV) species called Q.6, 7, 9 On the basis of EXAFS studies, it has been proposed that Q has an FeIV2(μ-O)2 diamond core.10 Kinetic studies show that P converts to Q in a proton-dependent core isomerization step.11 Related to this conversion is the generation of the catalytically essential intermediate X of Class 1 ribonucleotide reductases, which is formed in the reaction of O2 with the diiron(II) center found in the R2 component of this enzyme. In this case, a (μ-peroxo)diiron(III) species also forms but subsequently undergoes one-electron reduction to generate the FeIII-O-FeIV center in X.12, 13

Biomimetic complexes have been useful in providing fundamental structure-spectroscopy correlations for interpreting the properties of enzyme intermediates. While there are now a number of synthetic oxoiron(IV) complexes that serve as models for oxoiron(IV) intermediates of nonheme monoiron enzymes,14, 15 there is only one synthetic complex characterized to have the FeIV2(μ-O)2 diamond core motif proposed for sMMO intermediate Q.16 In this initial effort, the synthetic complex [FeIV2(μ-O)2(L)2]4+ (3, L = tris(3,5-dimethyl-4-methoxylpyridyl-2-methyl)amine) was obtained by one-electron oxidation of its FeIIIFeIV precursor (Scheme 1; we label the compounds as in ref. 17). More recently, however, we have discovered that 3 can also be generated from the reaction of a (μ-oxo)diiron(III) complex and H2O2, a transformation that may shed light on steps in the enzymatic P to Q conversion.17 Treatment of [FeIII2(μ-O)(OH)(H2O)(L)2](ClO4)3, 1 (the methylene groups of L were deuterated), with 1 equiv. H2O2 affords intermediate 2 in 60-80% yield. Addition of 1 equiv. HClO4 instantly converts 2 into 3, forming the [FeIV2(μ-O)2] diamond core. In the presence of H-atom donors intermediate 2 converts to the well-characterized18 green [FeIVFeIII(μ-O) (L)2]3+ intermediate, 4. The presence of 4 is easily recognized by its characteristic S = 3/2 EPR signal and Mössbauer quadrupole doublet with ΔEQ = 0.44 mm/s and isomer shift, δ = 0.11 mm/s (a doublet is observed for T > 70 K).

Scheme 1
Reaction of 1 with H2O2 to form 2 and 3.

Complex 2 is thus an interesting precursor to complexes with Fe2(μ-O)2 cores. Electrospray ionization mass spectrometry (ESI-MS) of 2 showed an ion cluster with a mass and isotope distribution pattern consistent with the formulation {[(Fe2O3(L)2H)](OTf)2}+, and 18O-labeling studies demonstrated that one oxygen atom derives from the diiron(III) precursor while the other two originate from H2O2.17 EXAFS studies revealed the presence of a 1.66-Å Fe-O bond associated with an Fe=O unit and an Fe-Fe distance of 3.32 Å.17 Mössbauer studies of 2 revealed a diiron(IV) center with inequivalent iron(IV) sites, one of which with Mössbauer parameters, reported below, strikingly similar to those of the mononuclear complex [LFeIV(O)(NCMe)]2+. These data lead us to propose that 2 is a complex with a structure as depicted in Figure 1 having an HO-FeIV-O-FeIV=O core.

Figure 1
DFT-optimized structure of 2 (ADF optimization, method 1). Hydrogen atoms have been omitted for clarity, except for the one of the hydroxo ligand on the FeOH site.

In ref. 17 we touched only briefly on the Mössbauer properties of 2, showing only a 4.2 K spectrum recorded in zero applied magnetic field. Here we report a detailed Mössbauer study of 2, together with a spin Hamiltonian analysis, that reveal 2 to be a ferromagnetically coupled diiron(IV) complex, containing one site with a terminal oxo ligand. We also report density functional theory (DFT) studies that give insight into the structure of 2 and rationalize the observed ferromagnetic coupling.

Materials and Methods

Complex 2 in MeCN was prepared by a previously described procedure.17 The Mössbauer analysis was performed on a sample prepared in 1:1 MeCN-PrCN mixture. To prepare this sample, 1 dissolved in MeCN was treated with 0.8 equiv. H2O2 in MeCN at -40 °C to generate 2; an equal volume of PrCN (pre-cooled to -40 °C) was then added prior to freezing the solution. We also prepared a sample of 2 by treating 1 with 1 equiv. D2O2 in MeCN at -40 °C. Mössbauer spectra were collected with constant acceleration spectrometers, using two cryostats that allowed studies at 4.2 K in applied fields up to 8.0 T. Isomer shifts are quoted relative to Fe metal at 298 K. Spectra were analyzed with the WMOSS software package (WEB Research Co., Edina, MN). Density functional theory (DFT) calculations were performed for complex 2 by using the Amsterdam Density Functional (ADF) 2008.01 software package19-21 on a cluster of 10 processors (Ace computers) using ADF basis set IV (triple-ζ with single polarization on the ligand atoms), an integration constant of 4.0, and the Vosko-Wilk-Nusair local density approximation22 with the nonlocal gradient corrections of Becke23 and Perdew24 (BPVWN) (Method 1) and by using Becke’s three parameter hybrid (B3LYP) functional and basis set 6-311G provided by Gaussian’03 software package25 on a cluster of 8 processors26 (Method 2). One calculation was performed in the same way of Method 2 but using the PBEVWN functional instead of the B3LYP (Method 3). For details of the DFT study, see Supporting Information.

Mössbauer analysis

We have studied with Mössbauer spectroscopy samples of 2 (14 preparations) treated with 0.2, 0.4, 0.6, 0.8, 1.0, 1.2 and 1.5 and 2.0 equivalents of H2O2 in acetonitrile and 1:1 acetonitrile/butyronitrile mixtures, with the goal of studying the formation of 2 and maximizing the content of 2 in a sample for comprehensive studies in strong applied magnetic fields. For a detailed study we have selected a sample obtained by adding 0.8 equiv. H2O2 to the diferric starting material, 1, in a 1:1 acetonitrile/butyronitrile mixture at -40°C. Figures Figures2,2, ,3,3, and S5 show Mössbauer spectra recorded for this sample.

Figure 2
Mössbauer spectrum of 2 recorded at 80 K in a parallel applied field of 8.0 T (hash marks in C). The solid line is a spectral simulation generated from eq. 1 in the strong coupling limit, using the parameters quoted in Table 1, with slight modifications ...
Figure 3
4.2 K Mössbauer spectra of 2 in a 1:1 mixture of acetonitrile/butyronitrile recorded in applied magnetic fields as indicated. The solid lines represent spectral simulations based on eq. 1 using the parameters listed in rows 2 and 3 of Table 1 ...

Our samples generally contained diferric contaminants and a small amount of the green S = 3/2 FeIIIFeIV intermediate 4; typically around 5% of the Mössbauer absorption could not be accounted for (reflecting probably residuals of imperfect subtractions). The diferric contaminants, identified by the observation of diamagnetic components with isomer shift δ ≈ 0.45 mm/s, may comprise three different species in variable proportions, namely (i) nonreacted starting complex 1, (ii) an [Fe III2(O)(H2O)2(L)2]4+ contaminant in the starting material (which does not afford 2 upon addition of peroxide), and (iii) a diferric decay product. Figure S5A shows a 4.2 K zero-field Mössbauer spectrum of the sample described below; after subtraction of the contaminants, described in Supplementary Information, we obtained the spectrum of 2 shown in Figure 3A. The spectrum exhibits two quadrupole doublets assigned to intermediate 2.

The prepared Mössbauer spectra of Figures Figures22 and and33 were obtained by subtracting the contributions of 1 (35%) and 4 (5%) from the raw data. The zero-field spectrum of Figure 3A consists of two nested quadrupole doublets, each representing ≈ 30% of the Fe in the sample. Doublet a (called site a in this section) has ΔEQ(a) = 1.96 mm/s and δ(a)= 0.00 mm/s while doublet b (site b) exhibits ΔEQ(b) = 0.92 mm/s and δ(b) = -0.03 mm/s. The isomer shifts of a and b are characteristic of FeIV TPA complexes (TPA = tris(pyridyl-2-methyl)amine),15 an assignment supported by the observation that both are transformed upon protonation into the red diiron(IV) intermediate 3 with δ = -0.05 mm/s and ΔEQ = 2.01 mm/s. Doublet a and b always appeared in a 1:1 ratio when the diferric complex 1 was titrated with H2O2. (We found that the amount of red intermediate 3 obtained by protonation correlates very well with the amount of 2 determined by the Mössbauer analysis.)

Both a and b exhibit quadrupole doublets at 4.2 K, showing that they belong to a complex with integer electron spin. In applied magnetic fields, B, the spectra of a and b exhibit paramagnetic hyperfine structure, implying that the ground state of 2 has electronic spin S > 0. The whole body of our Mössbauer analysis suggests that a and b are subsites of a dinuclear complex. Given that 2 originates from dimer 1 and transforms, respectively, into dimers 3 and 4 upon protonation and hydrogen atom transfer,17 it was expected from the outset that 2 is a dimer (Mass spectrometric data17 led to the same conclusion). The spectra recorded in applied fields are exceedingly complex and they depend on numerous unknowns. However, the salient features of the solution can be recognized, albeit with some effort, by inspection of the data.

We first consider the 8.0 T spectrum of Figure 2C recorded at 80 K, a temperature at which the electronic spin of 2 is in the fast relaxation limit. A spectral simulation (Figure 2A, dashed) assuming diamagnetism for site a yields for the low-energy feature (see comment in ref. 27) a splitting substantially larger than observed experimentally, indicating that the 57Fe nucleus of site a experiences an effective field, Beff = B + Bint, that is ca. 2-3 T smaller than the applied field. Indeed, the simulation with Bint = -2.5 T (along x and y), shown in Figure 2B, fits the data quite well. Corresponding simulations, shown in Figures 2D and E (dashed), reveal that site b has a negative Bint as well. A negative Bint for both sites indicates that the local spins Sa and Sb are parallel and that 2, therefore, is a ferromagnetically coupled diiron(IV) complex.27 All FeIV-TPA complexes studied to date have S = 1, and thus we can reasonably assume that the local sites a and b have Sa = Sb = 1.

We have analyzed the applied field spectra of 2 in the framework of the spin Hamiltonian:

H^=JS^a[center dot]S^b+i=a,b{Di(S^iz22)+Ei(S^ix2S^iy2)+geβB^[center dot]S^i+S^i[center dot]Ai[center dot]I^ignβnB^[center dot]I^i+H^Q(i)}

where J is the isotropic exchange coupling constant (J < 0). Da, Db and Ea, Eb are the axial and rhombic (local) zero-field splitting (ZFS) parameters of the Sa = 1 and Sb = 1 multiplets. The remaining terms describe the electronic Zeeman, magnetic hyperfine, nuclear Zeeman and electric quadrupole interactions. S = 1 FeIV complexes generally have g ≈ 2 and, therefore, we have fixed the g-values at ge = 2. In the limit of strong ferromagnetic coupling, |J| >> |Di|, the ground state of 2 is a multiplet with S = 2 for which the electronic terms of eq. 1 can be rewritten as:

H^e=D(S^z22)+E(S^x2S^y2)+2βB^[center dot]S^

In eq. 2 the ZFS tensor of the S = 2 multiplet is given by:28


In the strong coupling limit the nuclear part of eq. 1 can be written as (S = 2):

H^nucl=i=a,b{(1/2)S^[center dot]Ai[center dot]I^ignβnB^[center dot]I^i+H^Q(i)}

The factor 1/2 is a spin projection factor28 that has to be included when the local spin operators Ŝa and Ŝb are replaced by the cluster spin operator Ŝ. In order to discover the nature of the solution it will be useful to discuss the problem first in the strong coupling limit and use eqs. 2 and 4, H^=H^e+H^nucl. Throughout this section we will take the principal axis system of D, {x, y, z}, as the frame of reference. In order to keep the number of unknowns manageable we assume that all tensors are diagonal in {x, y, z}, but allowing 90° rotations around a principal axis.

H^Q of eqs. 1 and 4 can be written in its commonly used principal axis form (dropping the site index i) as:


where η = (Vxx-Vyy)/Vzz is the asymmetry parameter of the electric field gradient (EFG) tensor. For η = 0, -3 and +3 the EFG is axial around z, x and y, respectively.

For reasonable small values of D (say, D < 10 cm-1) the 80 K, 8.0 T spectrum of Figure 2 is (essentially) independent of D, and the thermally averaged expectation values of the site spins, <Sa>th = <Sb>th = (1/2)<S>th, are small and follow the Curie law above 50 K. Under these conditions the two sites can be treated independently (D correlates the spatial directions of the two sites at 4.2 K), and we can obtain valuable information about the EFG- and A-tensors. We temporarily use the independent frames {X’,Y’,Z’} for site a and {X“,Y”,Z“} for site b: we will shortly relate these frames with {x, y, z} and the molecular frame of 2.

Spectral simulations suggested for site a: ΔEQ = +1.96 mm/s, η ≈ 0, AX,/gnβn ≈ AY,/gnβn ≈ - (11-20) T and AZ,/gnβn ≈ +0.4 T; and for site b: ΔEQ = +0.92 mm/s, η ≈ 0, AX”/gnβn ≈ AY”/gnβn = -23 T and AZ”/gnβn ≈ -5 T. Importantly, the 80K/8.0T spectrum tolerates AX, ≠ AY, as long as their average, roughly, corresponds to the above quoted axial range. The possibility that AX, ≠ AY’ for site a enabled us to find a solution that fits both the 80 K and 4.2 K spectra.

At 4.2 K the electronic spin, S = 2, of 2 is in the slow relaxation limit on the time scale of Mössbauer spectroscopy (≈ 10-7 s). The spectra show that the magnetic hyperfine splittings of both iron sites increase with increasing applied field, reaching their saturation values near B = 2 T (In order to “see” that, one must have recognized the associations of doublets a and b with the high-field features of the spectra; the authors labored for many hours on this problem.). Such saturation behavior is approximately described by eq. 2 for D ≈ 4 cm-1 and E/D ≈ 0.4 (Figure 4, black curves). Although the 4.2 K spectra are quite complex, we can understand their principle features from the following considerations. The magnitudes of the (negative) expectation values of the electronic ground level along y, x and z are ordered such that |<Sy> | ≥ |<Sx>| > |<Sz>| (see Figure 4). For acceptable simulations for site b we must align {X”, Y”, Z”} with {x, y, z} such that Z” = z. In order to produce the correct 4.2 K magnetization for site a, i. e. the correct <Sy>, we must have, in {x, y, z}, Ay/gnβn ≈ -11 T, which would yield Ax/gnβn ≈ -(20-27) T (if X’ [equivalent] x), as the two components must average to about -(11-20) T. However, we need a solution for which the largest A-tensor component is multiplied by a small expectation value of Sa, namely <Sz,a>; without this association the calculated magnetic splittings for site a would considerably exceed the experimentally observed splittings at 4.2 K. The suggested associations are made by having the largest component of the EFG along x (which can be accomplished by replacing η’ = 0 with η = -3 and working in {x, y, z}) and associating AX’ with Az. These associations imply that the largest components of the EFGs of sites a and b are perpendicular to each other.

Figure 4
Expectation values of the components of the electronic spin for the lowest spin level of 2, with the external field B applied along x, y and z. Black: Curves generated for the zero-field splitting parameters of Table 1 using J = -1000 cm-1, the strong ...

In the course of our studies we noticed that the hyperfine parameters obtained for site b are very close to those observed for S = 1 FeIV=O complexes, an observation that prompted us to synthesize and study the FeIV-oxo complex [LFeIV(O)(NCMe)]2+ with the same TPA-related ligand; Mössbauer data, analysis and optimized DFT structure for this complex are presented in Supporting Information. Comparison of the parameters of the mononuclear complex in Table 1 with those obtained for site b forcefully suggests that site b has a terminal oxo ligand. The analysis of the Mössbauer spectra of 2 required as many as 1000 spectral simulations, and throughout this effort we had little idea what structure gave rise to this ferromagnetically coupled diiron(IV) complex. Thus, finding a parameter set for site b that closely matched those of an FeIV=O complex, prior to obtaining the EXAFS and mass spectrometric evidence,17 would suggest that the decomposition of the spectra is essentially reliable.

Table 1
Zero-field splittings and hyperfine parameters of 2 and two relevant mononuclear complexes. Rows 2 and 3 present the parameters of 2 expressed in {x, y, z}. Row 4 lists the parameters of site a in the proper coordinate system, {xp, yp, zp}, of its ZFS ...

With the strong hint that b is an FeIV site with terminal oxo group we switched to the uncoupled representation of eq. 1. Using the approximate values of D and E/D obtained in the strong exchange coupling limit (pure S = 2 ground multiplet) and inserting Db = 27 cm-1 and (E/D)b = 0 into eq. 3, we obtained a reasonable approximation for Da and (E/D)a. We then performed a global search, keeping Da, (E/D)a and the EFGs of both sites fixed. Lowering the magnitude of J from the strong exchange coupling limit decreases the spin expectation values by multiplet mixing by the Da and Db terms (Figure 4, red curves) and requires slightly larger Ay values for fitting the magnetization behavior [i. e. fitting Bint,a,b vs. B]. Our “best” solution set (listed in Table 1) was obtained for J ≈ -50 cm-1; this set has Da = -1.35 cm-1 and (E/D)a = -8.75 when quoted in {x, y, z}. By using quotation marks we remind the reader that the data of Figures Figures22 and and33 were obtained by subtracting 40% of the absorption attributable to contaminants from the raw data. It is likely that one can find solution sets of similar quality for any J value in the range -150 cm-1 < J < -30 cm-1. However, for -J < 30 cm-1 the magnetic splittings of sites a and b would saturate observably differently.

Rows 2 and 3 of Table 1 list the parameters of sites a and b in {x, y, z}. This frame is the “proper” coordinate system of the ZFS tensor of oxo site b, (frames for which 0 ≤ (E/D) ≤ 1/3 or 0 ≤ η ≤ 1 are often called “proper”). The proper coordinate frame of site a, {xp, yp, zp}, has its zp-axis along x (The use of a proper frame for Da facilitates comparison of the parameters of site a with published work). Row 4 lists the parameters of site a in {xp, yp, zp}; in this frame Da = +17.7 cm-1 and (E/D)a = 0.28. These values together with the large ΔEQ of site a pointed our attention to a Mössbauer study of the mononuclear complex [(β-BPMCN)FeIV(OH)(OOtBu)]2+.29 Guided by the observation of an {[Fe2O3L2H](OTf)2}+ cation by ESI-MS, we picture site a as an L-FeIV complex with a hydroxide ligand and linked by an oxo bridge to site b, and interestingly, the Mössbauer parameters of site a and the monomeric BPMCN complex (Table 1) are intriguingly similar (the proximal oxygen of the peroxo group representing the bridging oxygen of 2).

We have made some observations we do not yet fully understand. In order to demonstrate ferromagnetic coupling for 2 we have discussed, for developing the argument rather than out of necessity, the 80 K spectrum of Figure 2C in the limit of strong exchange coupling (Bint,b < 0 can be also inferred from inspection of the spectra of Figure 3, dashed lines). This allowed us to focus on the fact that Bint is negative for both sites of 2. After many attempts to simulate the low temperature spectra, we arrived at a set for which the magnitude of J is comparable to Da and Db. The parameter set of Table 1 describes all essential features of the 4.2 K spectra. However, for matching the simulations to the 80 K spectrum as shown in Figure 2 we had to increase Ax and Ay of site b by ca. 8%. We also increased Az of site a from -20 T to -28 T (<Sz> is small at 4.2 K, and thus the quoted value of Az,a is soft). It is possible that our J-value and ZFS-tensors are not optimal to describe the temperature dependence of the thermally averaged expectation values of the electronic spin system. However, there may be other reasons for the requirement of using slightly different A-values at 80 K. Thus, as shown in Figure S6 the recoilless fraction (Debye-Waller factor) of site a drops precipitously above 80 K, indicating perhaps, a structural change, possibly of dynamic nature (such as a rotation of the hydroxo ligand or proton hopping).30

Comparison with the DFT calculations

In this section we examine the parameters of Table 1 and compare them with the results of a DFT study of 2. To aid the reader in this discussion we will re-label site a as FeOH and site b as FeO. Using the procedures described in Supplementary Information, we have performed a series of geometry optimizations of 2 using as the starting configuration the X-ray structure of [FeIII2(O)(OH)(OH2)(5-Me3TPA)2]3+ (5-Me3-TPA = tris(5-methylpyridyl-2-methyl)amine), a complex closely related to 1, in which the water and hydroxo ligands have been replaced by hydroxo and oxo groups, respectively.18 The optimized structure of diiron(IV) complex 2, using method 1, is shown in Figure 1. The calculated structure has an Fe-Fe distance of 3.39 Å (3.32 Å), an Fe-oxo bond length of 1.67 Å (1.66 Å) and an Fe-O-Fe angle of 130.8 degrees, in good agreement with the metric data obtained from EXAFS17 (listed in italics). In the DFT structure the proton of the hydroxo group of site FeOH is hydrogen bonded to the oxo group of site FeO.

The calculations reported here (see Tables and text in SI) reveal significant differences in the structure of the OTFeOOBFeOHOH core (T = terminal, B=bridge), depending on the functional/basis set combination used for the geometry optimization. The structural discrepancies may be rooted in differences between the functionals in their ability to assess the hydrogen bonding between OH and OT (see Supporting Information). In particular, BPVWN/IV (method 1) used for the calculation of the structure in Figure 1 yields an approximately planar core, whereas B3LYP/6-311G (method 2) results in a non-planar core unit. Since the 3d-electron-carrying t2g orbitals of FeOH are engaged in π interactions with the lone-pair orbitals of OH and the bridging oxygen, OB, the torsion angles of the core unit affect the composition of the t2g orbitals (and their energies) and with it properties such as the EFG-tensor and the spin-dipolar part of the magnetic hyperfine tensor (ASD). Particularly affected are the orientations of the principal tensor axes through torsion of the terminal hydroxide at FeOH. The principal axes in the optimized structure obtained by method 2 are rotated more away from the FeOH-ligand bonds than in the optimized structure obtained by method 1, where the core exhibits less torsion.

For small torsions, as in the optimized structures obtained with method 1, the principal axes of the largest component of EFG and ASD tensors at the FeOH site are nearly parallel and roughly align with the Ntrans-Ntrans direction whereas in the case of large torsions, as obtained with method 2, these axes deviate significantly from this direction. For the following qualitative discussions we adopt the solution of method 1, unless stated otherwise, for two reasons. First, the ADF program is more convenient for broken symmetry calculations to assess the exchange interactions and, second, the consistency of the description provided by method 1 with the Mössbauer analysis presented above is easier to assess than for method 2. This does not mean that the solution obtained with method 2 can be dismissed. We wish to stress that the principal axes of the EFG- and ASD-tensors of the FeO site, where the t2g orbitals form strong π bonds with the p orbitals of the terminal oxo ligand, are insensitive to the methodological differences affecting the core structure.

The 57Fe magnetic hyperfine tensor can, in general, be written as A = AC + ASD + AL. AC is the isotropic Fermi contact term, ASD describes the traceless spin-dipolar contribution and AL accounts for the orbital contribution. AL can be decomposed into an isotropic (pseudo-contact) part, APC, and a traceless anisotropic part, AL,aniso; from our frozen solution studies we can extract (Ax + Ay + Az)/3 = Aiso = AC + APC. For the S = 1 FeIV=O complexes reported so far, AL was found to be vanishingly small, and we will assume that this holds for FeO as well.

As pointed out above, the parameters of FeO match within the uncertainties those of the mononuclear FeIV=O complex formed with the same ligand, [LFeIV(O)(NCMe)]2+. In the Mössbauer analysis section, we evaluated our data in the coordinate system {x, y, z}, the principal axis frame of the D-tensor of eq. 2, and found that its z-axis coincided with the common symmetry axis of the EFG- and A-tensor of FeO (Table 1). Three experimental observations allow us to locate {x, y, z} in the molecular frame of 2. First, the z-axis coincides with the principal axes of the EFG and ASD of FeO that are associated with the largest of the positive eigenvalues of these tensors. The DFT calculations with both methods indicate that the latter axes are collinear with the FeIV=O bond,31 in agreement with our previous studies of FeIV=O complexes.32, 33 Second, from our Mössbauer analysis we deduced for FeOH that the largest component of the (≈ axial) EFG and the smallest component of the (rhombic) A-tensor of FeOH (corresponding to the largest component of ASD) are essentially directed along x, a conclusion supported by the DFT calculation with method 1 which places these components at an angle of only ~18° relative to the N-N direction of the trans nitrogens. Third, it follows that the y-direction, the easy axis of magnetization (see Figure 4), is then nearly along the FeO-OB bond (see Figure 5). For our Mössbauer data analysis we made the assumption that the EFG- and A-tensors of both sites are collinear. Given that the Fe-O-Fe angle of the linking oxo bridge is 131°, rather than 90° or 180°, one would suspect that this assumption might not be correct. However, our assumption finds some support from the DFT calculations with method 1; the principal axes associated with the EFG components (2.15, -1.36, -0.79) mm/s at FeOH deviate by 17°, 11°, 19° from the x, y, z axes, respectively; the ASD components (14.2, - 11.0, -3.2) T are off by 18°, 25°, 20° (For FeOH the Mössbauer analysis cannot fix the tensor axes to better than ≈ 20° and is thus consistent with the results of method 1). Thus, the principal axes of the A- and EFG-tensors are roughly aligned with the axes of D,31 and neither the EFG-nor the A-tensor of FeOH has a principal axis coinciding with the FeOH-OB or FeOH-OH bond. The calculations with method 2 give much larger deviations, which is one of the reasons (see above) for focusing in the discussion on the results obtained with method 1.

Figure 5
Direction of the axes of {x, y, z} in the molecular structure of 2, according to DFT calculations. z is nearly parallel to the Fe-O bond of FeO (left) and y is within 5 degrees parallel to FeO-OB and x is roughly parallel to the trans N-N direction.

The 57Fe A-tensor of FeO has Aiso = -16.9 T, in excellent agreement with Aiso = -16.7 T obtained for the mononuclear site of [LFeIV(O)(NCMe)]2+. The spin-dipolar term, ASD = A - Aiso, of FeO, ASD;x,y,z,/gnβn = (-6.1, -5.8, +11.9) T is quite well produced by DFT, (-4.0, -5.9, +9.8) T; see Table S2. While the experimental data suggest a nearly axial EFG (η ≈ 0) for FeO, method 1 gives a rhombic tensor (η ≈ 1.28; in proper frame: η ≈ 0.75) and a quadrupole splitting, ΔEQ (DFT) = -0.80 mm/s, with reversed sign relative to experiment, ΔEQ = +0.92 mm/s. However, the negative component (-0.80 mm/s) is aligned along y and the calculated EFG has a positive component of 0.64 mm/s along x, in good agreement with the experimental value of 0.92 mm/s. Method 2 gives for FeO a less rhombic EFG (η ≈ 0.50) with a positive quadrupole splitting, ΔEQ = +1.57 mm/s, along z. We have frequently observed that our DFT calculations with method 2 yield ΔEQ values that differ by approximately 0.5 - 0.7 mm/s from the experimental data, and we are thus not surprised by the present discrepancy. However, for the [LFeIV(O)(NCMe)]2+ complex, the calculated and experimental ΔEQ and η values obtained with method 2 agree, perhaps somewhat fortuitously, within 0.02 mm/s and 0.02 respectively; the calculated and experimental isomer shift agree quite nicely within 0.05 mm/s. As observed for other FeIV=O complexes obtained with TPA or TPA-related ligands, the calculated isomer shifts for FeO and [LFeIV(O)(NCMe)]2+ with method 2 both are 0.06 mm/s larger than the experimental values, but nevertheless in good agreement with the data. (Isomer shifts have been evaluated here with method 2, for which a calibration is available.) The reader may wonder why there is such a good agreement between the Mössbauer parameters of the FeIV sites of FeO and [LFeIV(O)(NCMe)]2+, given that the MeCN ligand of the latter complex is replaced by OB in FeO. We have observed previously that a ligand substitution cis to the oxo group has little effect on the Mössbauer parameters.34 The difference in the η values predicted for FeO with methods 1 and 2 may arise from the difference in the FeO-OB distances with the two methods: 1.891 Å (1) and 1.993 Å (2) (see Supporting Information).

The FeOH site of 2 has a remarkably small Aiso value, namely Aiso/gnβn = -10.2 T, which can be compared with Aiso/gnβn values for non-oxo complexes [(β-BPMCN)FeIV(OH)(OOtBu)]2+ (-14.1 T)35 and [(Me3-cyclam-acetate)FeIV(N3)]2+ (-11.8 T).36 The small Aiso value of FeOH may reflect an orbital contribution opposing the Fermi contact term. For example, if the g-tensor of FeOH had gav = (gx + gy + gz)/3 ≈ 2.11, a pseudo-contact APC/gnβn = (gav - 2)P ≈ + 5 T would result (If we reasonably assume that the orbital term is negligible for [LFeIV(O)(NCMe)]2+, we obtain P = 2β <r-3> ≈ + 45 T for that complex). Deviation of g-values from g = 2.0 is the result of spin-orbit coupling of the (local) Sb = 1 multiplet with excited Sb =1 multiplets, and typically only one or two components of g are affected. Thus, a giso = 2.11 would suggest an anisotropic g-tensor, resulting in a substantial AL,aniso contribution to the A-tensor of FeOH. We presently do not have sufficient information to analyze the experimental A-values of FeOH in more detail. The ΔEQ of FeOH calculated with method 1 comes out about right (+2.15 mm/s vs. +1.96 mm/s experimental).

Using the Mulliken gross orbital populations for the ferromagnetic state of 2 as a guideline (Tables S3 and S4), we have identified the 3d configuration of FeO as (dxy)2 (dxz)1 (dyz)1 and the configuration of FeOH as an idealized (d’y’z’)2(d’x’y’)1(d’x’z’)1 configuration; with x’ || Ntrans-Ntrans, y’ || FeOH-OB, and z’ || FeOH-OH, see Figure 6. Accordingly, the spin density profiles of the irons in Figure S11 exhibit cylindrical symmetry along the orthogonal z and x’ ≈ x axes, respectively. The electronic structure of the FeO site is very similar to those of the well studied mononuclear S = 1, FeIV=O complexes,32-34, 37 resulting in hyperfine parameters for FeO that are similar to those found for the mononuclear [LFeIV(O)(NCMe)]2+ complex. The valence contributions to the EFGs obtained for the two configurations generate axial tensors with large, positive components along z (at FeO) and along x’ (at FeOH, x’ is nearly parallel to x), consistent with the perpendicular orientation deduced from the Mössbauer analysis. As expected for these configurations, the spin-dipolar contributions to the A-tensor from method 1, ASD = (-4.0, -5.9, +9.8) T for FeO and ASD = (+14.2, -11.0, -3.2) T for FeOH, have large positive components along z and x’ ≈ x, respectively, but with a significant degree of rhombicity in the y-z plane at FeOH, indicating that the (d’y’z’)2(d’x’y’)1(d’x’z’)1 configuration must be considered as an idealization of the DFT solution. As shown in Figure S11, the px orbitals at both the bridging and hydroxo oxygen atoms carry significant spin densities, indicating the presence of sizeable π interactions of these orbitals with the d’x’y’ and d’x’z’ orbitals of FeOH, respectively, pushing the latter up in energy and leaving the d’y’z’ orbital as the lowest, doubly occupied t2g orbital. The π interactions with pyridines, which are good σ donors, play probably a lesser role in removing the degeneracy of the t2g orbitals. Spin-orbit coupling of the ground state with the spin-allowed excitations d’y’z’ → d’x’z’ and d’y’z’ → d’x’y’ yield unquenched orbital moments along z’ and y’. The associated magnetic moments give a (positive) pseudo-contact term which may explain the low value for Aiso deduced from the spin-Hamiltonian analysis. In particular, these contributions allow us to adopt a more realistic Fermi contact term. For example, by combining AC = -14 T, the value for [(β-BPMCN)FeIV(OH)(OOtBu)]2+, with ASD = (+14.2, -11.0, -3.2) T from DFT and an AL = (+0.2, +14.0, -2.8) T we obtain the Atot = (+0.4, -11.0, -20.0) T observed. The AL components suggest that the energies of the t2g orbitals appear in the order d’y’z’ < d’x’y’ < d’x’z’. The zero-field splitting tensor is axial along x’ when the t2g excitations are degenerate and rhombic (0 < E/D ≤ ⅓) when they are non-degenerate. The latter situation is likely to apply, given the difference in the π ligand interactions of the two excited t2g orbitals. This analysis suggests that the zero-field splitting at FeOH results from spin-orbit coupling of the ground state with excited states of the same local spin (Sloc = 1) and not from spin-orbit coupling with a low-lying Sloc = 2 excited state as it is likely the case at the FeO site.32

Figure 6
Magnetically active orbitals on FeO (left) and FeOH. The px and pz orbitals reside on the bridging oxo group, OB. Coordinate system {x, y, z} coincides with the main reference frame of the Results section. {x’,y’,z’} is a local ...

Why is complex 2 ferromagnetic?

Finally, we wish to address the question why complex 2 is ferromagnetic. Our broken-symmetry calculations, using method 1, predict an exchange-coupling constant J = -320 cm-1 for 2. This value is surprisingly large and exceeds, perhaps, the magnitude of the optimal J value deduced from the Mössbauer analysis. There is an extensive literature about the relationship between exchange interactions and molecular structure.38-43 In these theoretical contributions, J is expressed as the sum of a ferromagnetic and an antiferromagnetic term, J = JF + JAF (JF < 0, JAF > 0). The magnitudes of JF and JAF are determined by the overlap properties of unpaired-electron-containing (“magnetic”) orbitals at the interacting metal centers that are defined by their respective ligand environments.44, 45 In the case that the magnetic orbitals are non-orthogonal, JAF is non-zero and often the dominant term, resulting in a net antiferromangetic exchange interaction, J > 0.

The orbitals containing the unpaired electrons are illustrated in Figure 6: dyz and dxz at FeO and d’x’y’ and d’x’z’ at FeOH. For reasons of orbital symmetry, the orbitals dxz and d’x’z’ are unfavorable for mixing with the orbitals of the bridging oxygen and contribute little to J due to vanishing overlap density between them. The orbitals dyz at FeO and d’x’y’ at FeOH match the pz and px orbitals of the bridging oxygen in symmetry, leading to the formation of magnetic orbitals with composition dπ ≈ dyz + cz pz and d’π, ≈ d’x’y’ + cx’ px, where cx’ and cz are small mixing coefficients (The spin densities suggest that |cz| << |cx’|, cf. Supporting Information). The magnetic orbitals dπ and d’π’ have non-vanishing overlap density at the bridging oxygen because the product function cx cz px(r)pz(r) is not identical to zero. The associated overlap integral, however, is zero: <dπ|d’π’> ≈ cz cx <pz|px> = cz cx ∫pz(r)px(r)dr3 = 0. Since the orbitals dπ and d’π’ have a non-vanishing overlap density for which the spatial integral is zero, the condition for ferromagnetic exchange between the unpaired electrons in these orbitals and thus for ferromagnetic exchange coupling between the spins Sa = Sb = 1 of FeO and FeOH is fulfilled. If this interpretation of the DFT results for J is valid, then we have identified a system where the inequivalent coordination of the paramagnetic sites is instrumental in defining the magnetic orbitals in a way favorable for ferromagnetic exchange.

In closing, we wish to stress that, due to lack of experimental data, we have no experience in assessing the quality of calculated J-values for diiron(IV) complexes by DFT methods. For the system at hand, we alert the reader that if the core unit of 2 is non-planar, as suggested by DFT method 2 (see Supporting Information), the magnetic orbitals may have non-orthogonal pathways, leading to antiferromagnetic contributions to the exchange-coupling constant and thus to smaller J values. The analysis of Mössbauer spectra of 2 is exceedingly complex as it may entail as many as 38 unknowns (for C1 symmetry), not counting uncertainties introduced by removing contributions of contaminants. At this time we have to be content with noting that both theory and experiment indicate substantial ferromagnetic coupling for this interesting diron(IV) complex.

Supplementary Material



This work was supported by NIH grants GM38767 (to L.Q.) and EB0001474 (to E.M.), NSF grants MCB-0424494 (to E.M.) and CHE070073 (to E.L.B.) through TeraGrid resources provided by the NCSA and by an NIH postdoctoral fellowship GM-079839 to A.T.F.


1. Dalton H. Adv. Appl. Microbiol. 1980;26:71–87.
2. Wallar BJ, Lipscomb JD. Chem. Rev. 1996;96:2625–2657. [PubMed]
3. Solomon EI, Brunold TC, Davis MI, Kemsley JN, Lee S-K, Lehnert N, Neese F, Skulan AJ, Yang Y-S, Zhou J. Chem. Rev. 2000;100:235–350. [PubMed]
4. Merkx M, Kopp DA, Sazinsky MH, Blazyk JL, Muller J, Lippard SJ. Angew. Chem. Int. Ed. 2001;40:2782–2807. [PubMed]
5. Baik MH, Newcomb M, Friesner RA, Lippard SJ. Chem. Rev. 2003;103:2385–2419. [PubMed]
6. Lee SK, Nesheim JC, Lipscomb JD. J. Biol. Chem. 1993;268:21569–21577. [PubMed]
7. Liu KE, Valentine AM, Wang D, Huynh BH, Edmondson DE, Salifoglou A, Lippard SJ. J. Am. Chem. Soc. 1995;117:10174–10175.
8. Valentine AM, Stahl SS, Lippard SJ. J. Am. Chem. Soc. 1999;121:3876–3887.
9. Lee SK, Fox BG, Froland WA, Lipscomb JD, Münck E. J. Am. Chem. Soc. 1993;115:6450–6451.
10. Shu L, Nesheim JC, Kauffmann K, Münck E, Lipscomb JD, Que L., Jr. Science. 1997;275:515–518. [PubMed]
11. Lee SK, Lipscomb JD. Biochemistry. 1999;38:4423–4432. [PubMed]
12. Wei PP, Skulan AJ, Mitic N, Yang YS, Saleh L, Bollinger JM, Solomon EI. J. Am. Chem. Soc. 2004;126:3777–3788. [PubMed]
13. Yun D, Garcia-Serres R, Chicalese BM, An YH, Huynh BH, Bollinger JM. Biochemistry. 2007;46:1925–1932. [PubMed]
14. Que L., Jr. Acc. Chem. Res. 2007;40:493–500. [PubMed]
15. Shan X, Que L., Jr. J. Inorg. Biochem. 2006;100:421–433. [PubMed]
16. Xue G, Wang D, De Hont R, Fiedler AT, Shan X, Münck E, Que L., Jr. Proc. Natl. Acad. Sci. U. S. A. 2007;104:20713–20718. [PubMed]
17. Xue G, Fiedler AT, Martinho M, Münck E, Que L., Jr. Proc. Natl. Acad. Sci. U. S. A. 2008;105:20615–20620.
18. Dong Y, Fujii H, Hendrich MP, Leising RA, Pan G, Randall CR, Wilkinson EC, Zang Y, Que L, Jr., Fox BG, Kauffmann K, Münck E. J. Am. Chem. Soc. 1995;117:2778–2792.
19. te Velde G, Baerends EJ. J. Comput. Phys. 1992;99:84–98.
20. Guerra CF, Snijders JG, Te Velde G, Baerends EJ. Theor. Chem. Acc. 1998;99:391–403.
21. ADF2008.01. SCM, Theoretical Chemistry, Vrije Universiteit; Amsterdam, The Netherlands:
22. Vosko SH, Wilk L, Nusair M. Can. J. Phys. 1980;58:1200.
23. Becke AD. J. Chem. Phys. 1986;84:4524–4529.
24. Perdew JP. Phys. Rev. B. 1986;33:8822–8824. [PubMed]
25. Frisch MJ, et al. Gaussian 03, Revision E.01. 2004.
26. Catlett C. In: “TeraGrid: Analysis of Organization, System Architecture, and Middleware Enabling New Types of Applications”, HPC and Grids in Action. Grandinetti Lucio., editor. IOS Press; Amsterdam: 2007.
27. At 80 K the internal field at site i can be written as Bint,i = -<Si>thAi/gnβn, where <Si>th is the thermally averaged spin of site i and Ai is the magnetic hyperfine tensor. The argument for parallel spin alignment draws on the observation that the major components of the A-tensors of FeIV-TPA complexes are negative, typically Ax/gnβn ≈ Ay/gnβn = -23 T, Az/gnβn = -5 T; see below. Because of the small Az values the high energy features in Fig. 2C (at +1.5 mm/s and +2.5 mm/s) exhibit very small internal fields.
28. Bencini A, Gatteschi D. EPR of Exchange Coupled Systems. Springer-Verlag; 1990.
29. Jensen MP, Costas M, Ho RYN, Kaizer J, Payeras AMI, Münck E, Que L, Rohde JU, Stubna A. J. Am. Chem. Soc. 2005;127:10512–10525. [PubMed]
30. The Debye-Waller factor is given by the expression f = exp{-Eγ2 <x2> /(ħ c)2)} where Eγ = 14.37 KeV and <x2> is the mean square vibrational amplitude along direction x.
31. We will refer to “collinear” axes when the angles between them are less than 10 degrees.
32. Chanda A, Shan X, Chakrabarti M, Ellis WC, Popescu DL, Tiago de Oliveira F, Wang D, Que L, Jr., Collins TJ, Münck E, Bominaar EL. Inorg. Chem. 2008;47:3669–3678. [PubMed]
33. Sinnecker S, Slep LD, Bill E, Neese F. Inorg. Chem. 2005;44:2245–2254. [PubMed]
34. Rohde JU, Stubna A, Bominaar EL, Münck E, Nam W, Que L. Inorg. Chem. 2006;45:6435–6445. [PubMed]
35. Az(a) comes only into play for applied fields B ≥ 6 T. It is possible to increase Az(a) from -20 T to -26 T without serious deterioration of the simulation, and thus an Aiso/gnβn = -12.2 T is possible for site a.
36. Berry JF, Bill E, Bothe E, Weyhermüller T, Wieghardt K. J. Am. Chem. Soc. 2005;127:11550–11551. [PubMed]
37. Decker A, Solomon EI. Angew. Chem. Int. Ed. 2005;44:2252–2255. [PubMed]
38. Goodenough JB. Phys. Rev. 1955;100:564–573.
39. Kanamori J. J. Phys. Chem. Solids. 1959;10:87–98.
40. Anderson PW. Phys. Rev. 1959;115:2–13.
41. Goodenough JB. Magnetism and the Chemical Bond. Wiley and Sons; New York: 1963.
42. Anderson PW. Magnetism. Vol. 1. Academic Press; New York: 1963. p. 25.
43. Hotzelmann R, Wieghardt K, Florke U, Haupt HJ, Weatherburn DC, Bonvoisin J, Blondin G, Girerd JJ. J. Am. Chem. Soc. 1992;114:1681–1696.
44. Kahn O, Briat B. J. Chem. Soc. Faraday Transactions II. 1976;72:268–281.
45. Bominaar EL, Block R. Physica B & C. 1983;121:109–120.