Transformation of an [Fe(η²-N₂H₃)]¹⁺ Species to π-Delocalized [Fe₂(μ-N₂H₂)]^2+/1+ Complexes

Abstract

A monomeric iron Fe(η²-N₂H₃) species has been prepared, and exposure to oxygen yields a diiron complex that features five-coordinate iron centers and an activated bridging diazene ligand (NH=NH). Combined structural, theoretical, and spectroscopic data for the redox pair of complexes [Fe₂(μ-N₂H₂)]^2+/1+ are consistent with 4-center, 4-electron π-delocalized bonding picture across the Fe-NH-NH-Fe core that finds analogy in butadiene and the butadiene anion.

Several mechanisms have been proposed to describe the reduction of N₂ to NH₃ at the cofactor of MoFe-nitrogenase.^[1] Although experimental evidence is consistent with initial coordination of N₂ through a single metal center of the cofactor,^[2] recent DFT studies have pointed to plausible diiron intermediates of the type Fe₂(N₂H_y) (y = 1–4) en route to NH₃ formation.^[3] In this context, it is noteworthy that diazene^[4] and hydrazine^[5] are readily reduced to NH₃ by nitrogenase under turnover conditions. Diiron model complexes that feature the Fe₂(N₂H_y) core are therefore of timely interest,^[6],[7],[8] especially as a spectral reference point to aid in the interpretation of ENDOR/ESEEM data that is being obtained with the enzymatic system during catalysis.^[1d]

Herein we describe the characterization of an [Fe(η²-N₂H₃)]¹⁺ species that gives rise to a binuclear complex with an [Fe₂(μ-N₂H₂)]²⁺ core upon exposure to O₂. The latter complex is unique in that combined structural, spectroscopic, and DFT calculations suggest that the bridging ‘diazene’ is best formulated as N₂H₂^2-. While this level of diazene activation has been observed in complexes of highly reducing early transition metals^[9] it is not well established for the later transition metals, including iron.^{[7, 10]} One-electron reduction of the [Fe₂(μ-N₂H₂)]²⁺ complex furnishes the EPR-active mixed-valent [Fe₂(μ-N₂H₂)]¹⁺ complex, whose electronic structure characterization by combined EPR/ENDOR spectroscopy is described.

Entry to this chemical manifold arises from the addition of N₂H₄ to the iron alkyl precursor [PhBP₃]FeMe, 1, ([PhBP₃]⁻ = PhB(CH₂PPH₂)3⁻) in the presence of a suitable trap. We have previously reported that the room temperature reaction between 1 and N₂H₄ quantitatively forms {[PhBP₃]Fe}₂(μ-η¹:η¹-N₂H₄)(μ-η²:η²-N₂H₂), with concomitant loss of methane.^[7b] A hydrazido complex of the type “[PhBP₃]Fe(N₂H₃)” is a plausible thermally unstable intermediate to invoke, and a strong-field trapping ligand was hence pursued. Addition of 1 equiv of N₂H₄ to 1 at −78 °C, followed by addition of 1 equiv of CO, affords orange [PhBP₃]Fe(η²-N₂H₃)(CO), 2, in ca. 70 % chemical yield (Scheme 1). Several side reactions compete with formation of 2, and the crude reaction mixtures invariably contain {[PhBP₃]Fe}₂(μ-η¹:η¹-N₂H₄)(μ-η²:η²-N₂H₂),^[7b] [PhBP₃]Fe(CO)₂H (see SI), and several other unidentified species. The similar solubilities of [PhBP₃]Fe(CO)₂H and 2 diminish the isolated yield of 2 in analytically pure form.

Scheme 1.

The solid-state structure of 2 was obtained and indicates that the N₂H₃⁻ ligand coordinates η₂ to the Fe center (Scheme 1). The Fe-N distances of 1.992(3) and 2.018(3) Å are as expected for coordination of sp³-hybridized nitrogen to Fe, and are similar to those observed in the related six-coordinate Fe(η²-N₂H₄) and Fe(η²-N₂H₂) species.^[8] The N1–N2 bond distance of 1.383(3) Å is shorter than that expected for an N-N single bond, but consistent with that of a related N₂H₃⁻ complex of tungsten.^[11]

The ¹⁵N NMR spectrum (−75 °C, [D₈]THF) of 2 shows a complicated signal centered around 32 ppm, which was fit to obtain chemical shifts and coupling constants (see SI). The NH-NH₂ and NH-NH₂ chemical shifts are noted at 31.8 ppm and 32.2 ppm, respectively, with ¹J(N,N) = 10 Hz. The ¹H NMR spectrum (−75 °C, [D₈]THF) of 2 shows three distinct protons for the hydrazido ligand that split into doublets when samples of 2 are prepared with ¹⁵N₂H₄. The NH-NH₂ chemical shift is noted at 2.85 ppm (¹J(N,H) = 56 Hz), and the inequivalent NH-NH₂ protons appear at 6.55 (¹J(N,H) = 86 Hz) and 1.88 (¹J(N,H) = 79 Hz) ppm. The NMR data collectively indicates that the N₂H₃⁻ ligand is comprised of two sp³-hybridized nitrogen atoms.

The orange hydrazido(-) complex 2 undergoes decay to the bridged blue diazene complex, {[PhBP₃]Fe(CO)}₂(μ-η¹:η¹-N₂H₂), 3, in the presence of 0.5 equiv oxygen (Scheme 1). Other oxidants (e.g., Pb(OAc)₄, Cp2Fe⁺, p-quinone), acids (e.g., pyridinium, FeCl₃, Sm(OTf)₃), and bases (e.g., N₂H₄, ⁿBuLi, ^tBuN=P(cyclo-NC₄H₈)) were canvassed but do not facilitate this transformation. The reaction is solvent dependent and proceeds in benzene but not in THF, perhaps owing to hydrogen bond stabilization of 2 by THF solvent (see SI).

The ¹⁵N NMR spectrum of 3 (prepared from ¹⁵N- 2) displays a broad doublet at 292 ppm, indicative of an sp²-hybridized nitrogen atom. The diazene protons are magnetically inequivalent, and the corresponding ¹H{³¹P} NMR spectrum of 3 shows a AA’XX’ splitting pattern centered at 9.5 ppm. The chemical shifts of both the H and N atoms of the diazene ligand differ from those observed in the related {[PhBP₃]Fe}₂(μ-η¹:η¹-N₂H₂)(μ-η²:η²-N₂H₂) (¹⁵N NMR: 407.5, 58.0; ¹H NMR: 13.20, 4.16),^[7b] and suggest that the extent of diazene activation in the two complexes may be different. Simulation of the ¹H{³¹P} spectrum of 3 gives the following coupling constants: ¹J(N,H) = −71.0 Hz, ²J(N,H) = −2.1 Hz, ³J(H,H) = 14.8 Hz, and ¹J(N,N) = 9.5 Hz. The magnitude of the three-bond HH coupling is consistent with a trans configuration, and can furthermore be used as a probe for the extent of NN activation.^[12] For example, ³J(H,H) = 28.0 Hz for [(CO)₅Cr]₂(trans-μ-N₂H₂),^[13] which has an N-N bond distance of 1.25 Å,^[14] while ³J(H,H) = 9.4 Hz for [(η⁵-C₅Me₄H)₂ZrI]₂(trans-μ-N₂H₂), which has an N-N bond distance of 1.414(3) Å.^[9b] Hence, the observed ³J(H,H) coupling in 3 is most consistent with a single bond.

The solid-state structure of 3 was obtained and its core atoms are shown in Figure 1 (see SI for complete structure). Both Fe centers have similar metrical parameters, and adopt a distorted trigonal bipyramidal geometry, with the approximate equatorial plane defined by two phosphorous and one nitrogen atom. The two Fe centers are related by a 133° rotation about the Fe-Fe vector. The trans protons on the diazene were located in the difference map, and form a planar diazene. However, the Fe-N-N-Fe linkage departs from planarity and features a 20.3° dihedral angle (Figure 1). The average Fe-N bond distance of 1.83 Å in 3 indicates the presence of π-bonding, while the elongated N-N bond distance of 1.362(4) Å establishes a significantly activated diazene unit. This distance is closer to that expected for a N(sp²)-N(sp²) single bond than that for a double bond (ca. 1.41 Å and 1.24 Å, respectively).^{[7, 15]}

Figure 1.

Displacement ellipsoid (50 %) representations of the core atoms of 3 (left, top) and 4 (right, top), and an overlay of their core atoms (bottom, left; black, 3; gray, 4), and a representation showing the twist of the Fe-N-N-Fe linkage of 4 (bottom, right).

Complex 3 is intensely colored and displays a transition at 716 nm (ε = 8500 M⁻¹ cm⁻¹) that is presumably charge transfer in nature by analogy to assignments made for similar bands observed for related dinuclear M(η¹:η¹-N₂H₂)M complexes.^{[10, 16]} The rRaman spectrum of 3 (633 nm excitation) contains an NN vibration at 1060 cm⁻¹, which shifts to 1032 cm⁻¹ in samples of ¹⁵N-enriched 3 (calculated shift for a diatomic harmonic oscillator: 1023 cm⁻¹). In addition, a second vibration is observed at 665 cm⁻¹ (¹⁵N: 651 cm⁻¹), which is tentatively assigned as the ν_s(FeN) vibration that couples with the NN vibration. Both of these vibrations are distinct from those measured by Lehnert et al. in an octahedral Fe₂(μ-η¹:η¹-N₂H₂) complex (ν(NN) = 1365 cm⁻¹; ν_as(FeN) = 496 cm⁻¹),^[17] and consistent with appreciably stronger Fe-N and weaker N-N bonds in 3. The combined structural, NMR, and vibrational data suggest that the diazene bridge in 3 might better be regarded as a dianionic hydrazido, N₂H₂²⁻, as in the lower left resonance form shown in Figure 2B.

Figure 2.

(A) HOMO and LUMO of 3 (isocontour = 0.04); see SI for computational details. (B) Plausible resonance contributors to the electronic structure of 3. (C) 35 GHz Davies ¹⁵N pulsed ENDOR spectra (black traces; ν+ manifold (ν₊ = ν_n + A/2)) from ¹⁵N-4 at the indicated g values. Simulations (red, α = ±7°) use hyperfine and g tensors given in the text (see SI); the ENDOR linewidth in the g₃ simulation is greater than that for g₁, as would be required by a distribution of α, see text.

Cyclic voltammetry of 3 shows a reversible one-electron reduction to 4 at −1.54 V (vs. Fc/Fc⁺), and chemical treatment of 3 with one equiv of Na/Hg in THF cleanly generates the purple mixed-valence [Fe₂(μ-N₂H₂)]¹⁺ complex, [{[PhBP₃]Fe}₂(μ-η¹:η¹-N₂H₂)][Na(THF)₆], 4. Crystals of 4 suitable for XRD were grown by vapor diffusion of cyclopentane into a saturated THF solution of 4.

The geometry of the [Fe₂(μ-N₂H₂)]¹⁺ core of 4 is very similar to that of the [Fe₂(μ-N₂H₂)]²⁺ core of 3, as shown by an overlay of their core atoms (Figure 1). Upon reduction the average Fe-N distance increases by ca. 0.03 Å to 1.88 Å. Consistent with this, the ν_s(FeN) stretch decreases from 665 cm⁻¹ to 643 cm⁻¹ (¹⁵N: 624 cm⁻¹) upon reduction.^[18] The N-N bond distance in 4 is found to exhibit a marginal decrease to 1.342(3) Å upon reduction. These observations are collectively consistent with π-delocalization within the Fe-N-N-Fe core, with the unpaired electron populating an orbital that is predominantly Fe-N antibonding in character.

DFT calculations (see SI) were performed to further probe the electronic structures of both 3 and 4. The frontier orbitals of 3 are isolobal to those of butadiene, and both the HOMO and LUMO are primarily composed of the Fe-N-N-Fe π-system. The HOMO displays Fe-N π-bonding and N-N π*-bonding character (Figure 2A). The LUMO features N-N π-bonding, and Fe-N π*-bonding character. Population of the LUMO should therefore result in a decrease in the N-N bond distance and an increase in the Fe-N bond distance. However, the SOMO of 4 has only minimal density on the N-N bridge, and so the actual change should be small, as observed. The observation that the reduction of 3 to 4 yields a shortened N-N distance in the N₂H₂ ligand in the present case is consistent with the DFT calculations. A similar result has been provided for a series of [Mo₂(μ-N₂)]^6+/7+/8+ species where formal overall oxidation of the complex leads to a more ‘activated’ bridging N₂ ligand.^[19]

To further probe the electronic structure of the [Fe₂(μ-N₂H₂)]¹⁺ core of 4, we turned to EPR/ENDOR spectroscopy. Complex 4 is paramagnetic, with a rhombic S = ½ EPR signal (9:1 THF:Me-THF; g = [2.125, 2.040, 2.020]) that remains essentially invariant from 77 K to 2 K. To test the model of a symmetrical, π-delocalized Fe-N-N-Fe core, 35 GHz ¹⁵N electron-nuclear double resonance (ENDOR) measurements were performed at 2 K on ¹⁵N-4.^[20] Figure 2C displays ¹⁵N ENDOR spectra selected from a 2D field-frequency pattern of ENDOR spectra (ν+ manifold) collected across the EPR envelope of ¹⁵N-4 (see SI). The 2D pattern can be simulated with a single type of ¹⁵N , having a nearly axial hyperfine coupling tensor, principal values, A(¹⁵N) = + [6.7, 5.6, 17.8] MHz, isotropic coupling, a_iso(¹⁵N) = +10 MHz, and anisotropic coupling, T(¹⁵N) = + [−3.3, −4.5, 7.8] MHz (signs have been determined by pulsed ENDOR protocols; see SI).^[21] The absence of a Mims ENDOR response associated with a second, more weakly coupled ¹⁵N nucleus (not shown),^[22] indicates that the two ¹⁵N atoms from the bridge are magnetically equivalent and contribute equally to the ENDOR response depicted in Figure 2C, as expected for a delocalized [Fe₂(μ-N₂H₂)]¹⁺ ground state.

The strong anisotropy of A requires that the spin density on the two N atoms is of π character (A₃ parallel to the π-orbital for each N).^[23] The positive sign of A for ¹⁵N indicates that the π spin density on the N-N bridge, ρ^π(N), is negative (see SI), with the anisotropic coupling corresponding to ρ^π(N) ~ −0.05 spins/nitrogen. This small negative spin density on N arises from polarization of doubly-occupied bonding core π orbitals by the large spin density on Fe. The DFT computations on 4 give ρ^π(N) ~ −0.36 spins/nitrogen, which is in satisfactory agreement with experiment given that DFT is well known to overestimate the effects of spin polarization.^[24] This finding of rather low spin delocalization onto the bridging nitrogens of 4 illustrates why it is instructive to consider 3 in terms of the butadiene-like resonance structure shown in Figure 2B. The butadiene anion is the corresponding analogue to 4, and its SOMO is minimally delocalized onto the central atoms. In 4, delocalization would be decreased further due to the greater electronegativity of N compared to that of Fe.

The orientations of the ¹⁵N hyperfine tensors also are informative. The observation of a single, very sharp ¹⁵N ENDOR feature at g₁ indicates the g₁ axis is coincident with the N-N vector and normal to the spin-bearing π orbitals on ¹⁵N_i(i = 1, 2). These are expected to be primarily defined by the Fe_i-H_i-N_j (j = 2, 1) planes (Figure 1, bottom right), and thus lie essentially normal to the g₁ axis. Indeed, the 2D ENDOR pattern is satisfactorily simulated by taking the g₃ axis to bisect the angle between the two Fe_i-H_i-N_j planes, 2α ~ 14°, and then orienting each 15N_i hyperfine tensor along the normal to its plane, which corresponds closely to simply rotating the hyperfine tensors of N1 and N2 around g₁ by equal and opposite angles, α ~ 7° (see Figure 2C, red trace). The data does not define these rotations with precision; not only is agreement with experiment at g₂ improved with α = 15° (see SI), but also the observation of broad features in the ENDOR spectrum at g₃, in contrast to the narrow peak at g₁, suggests that there is a distribution of angles in the frozen solution, likely associated with torsions about the N-N ‘single’ bond of as little as a few degrees. Overall, the ¹⁵N ENDOR results support that 4, at 2 K, contains a π-delocalized Fe-N-N-Fe core, as predicted by DFT computations, with the π-orbital ‘twist’ indicated by the X-ray structure.

In summary, we have prepared an Fe(η²-N₂H₃) species, and have shown that the coordinated hydrazido ligand is converted to diazene in the presence of oxygen. The end-on diazene ligands in the [Fe₂(μ-N₂H₂)]^2+/1+ cores of 3 and 4, are best regarded as ‘N₂H₂²⁻,’ a bonding formulation previously observed for diazene complexes of highly reducing early-transition metals. Combined structural, theoretical, and spectroscopic data for the dinuclear complex 3 indicate the presence of 4-center, 4-electron π-delocalized bonding across the Fe-N-N-Fe diiron μ-diazene core. This picture is consistent with DFT studies, as well as a combined EPR/ENDOR study of its 1-electron reduced congener 4. This electronic structure, in which the HOMO is N-N π-bonding, provides access to stable diazene complexes in both the [Fe₂(μ-N₂H₂)]^2+/1+ oxidation states. Whether such a fragment arises in the reaction pathway by which nitrogenase reduces N₂ to 2 NH₃ is being explored by detailed comparisons of the results presented here with ENDOR results for nitrogenase intermediates.^[25]