Planar three-coordinate iron sulfide in a synthetic [4Fe-3S] cluster with biomimetic reactivity

Abstract

Iron-sulfur clusters are emerging as reactive sites for the reduction of small-molecule substrates. However, the four-coordinate iron sites of typical iron-sulfur clusters rarely react with substrates, implicating three-coordinate iron. This idea is untested because fully sulfide-coordinated three-coordinate iron is unprecedented. Here we report a new type of [4Fe-3S] cluster featuring an iron center with three bonds to sulfides. Although a high-spin electronic configuration is characteristic of other iron-sulfur clusters, the planar geometry and short Fe–S bonds lead to a surprising low-spin electronic configuration at the three-coordinate Fe center as determined by spectroscopy and ab initio calculations. In a demonstration of biomimetic reactivity, the [4Fe-3S] cluster reduces hydrazine, a natural substrate of nitrogenase. The product is the first example of NH₂ bound to an iron-sulfur cluster. Our results demonstrate that three-coordinate iron supported by sulfide donors is a plausible precursor to reactivity in iron-sulfur clusters like the FeMoco of nitrogenase.

Background

Iron-sulfur clusters are best known for electron transfer¹, and chemists’ current understanding has benefitted from synthetic models^2,3,4. Resting state structures of iron-sulfur clusters provide a natural starting point for synthetic modeling, and the most common structural elements are the [4Fe-4S] cubane and [2Fe-2S] diamond shapes. Recent advances include the isolation of all-ferrous⁵ and all-ferric⁶ [4Fe-4S] cubanes, preparation of a diferrous [2Fe-2S] cluster⁷, and isolation of several higher nuclearity [8Fe-9S] double cubanes with shapes reminiscent of the nitrogenase iron-molybdenum cofactor in its resting state^8,9. However, iron-sulfur clusters are dynamic and can undergo structural rearrangements to facilitate substrate binding and stabilize reactive intermediates¹⁰, as highlighted by reversible rearrangements of the cofactors of CODH and hydrogenase enzymes^11,12,13. In catalytically relevant examples of cluster flexibility, three-coordinate Fe in an all-sulfide environment has been postulated to enable cluster reactions, as in biotin synthase where the substrate S-adenosylmethionine coordinates to an iron atom supported by three sulfides¹⁴. Despite this emerging picture of iron-sulfur clusters as flexible assemblies, known synthetic iron-sulfur clusters have iron sites that are four-coordinate and tetrahedral like the biological resting state structures, which limits our understanding of the unsaturated active forms that may be essential for reactivity. Accordingly, few substrate reductions by iron-sulfur clusters are known^15,16,17,18. Three-coordinate species (potentially accessible from cleavage of weak iron-sulfide bonds) are expected to be quite active based on the reactivity of low-coordinate molecular Fe complexes¹⁹, making them compelling targets.

A salient example is nitrogenase²⁰, in which the crystallographically characterized FeMoco and FeVco active sites orchestrate the challenging conversion of dinitrogen into ammonia²⁰. Four-coordinate belt Fe atoms in the FeMoco (Fig. 1a) and FeVco cofactors, which are connected to three sulfide ligands (S²⁻) and an interstitial carbide ligand (C⁴⁻) in the resting state^21,22, have been shown to undergo Fe–S bond cleavage^23,24. The point in the mechanism at which bonds break, and the structure of the activated form that binds substrate, are unclear. Moreover, to our knowledge there is no experimental precedent for binding of dinitrogen reduction intermediates (N_xH_y: x = 1, 2; y = 0–5) to any synthetic iron-sulfur cluster. Three-coordinate iron in a cluster may therefore provide insight into the properties that are accessible in iron-sulfur clusters, and also address the chemistry underlying substrate binding by nitrogenase.

Figure 1. Preparation and relevance of a series of [4Fe-3S] clusters with trigonal planar iron.

a, Depiction of one hypothesis for how the FeMoco generates a substrate binding site through Fe-C bond cleavage to form three-coordinate iron sulfide; key atoms are highlighted with colored spheres. b, Synthetic route to three-coordinate iron-sulfur cluster [4Fe-3S], which features a three-coordinate iron sulfide site. c, Crystal structures of [4Fe-3S] clusters at three oxidation levels 1–3, which demonstrate that the cluster core is retained across these oxidation levels.

Results and Discussion

Synthesis and Structures of [4Fe-3S] Clusters

In this work, we report a new series of clusters supported by diketiminate ligands, which are adept at stabilizing low-coordinate metal sites in coordination complexes¹⁹. Previous work has demonstrated diketiminate-supported diiron sulfide complexes^25,26; however, in these prior examples, the iron sites each contain some biologically non-relevant N-based donors. We hypothesized that use of a smaller diketiminate framework (see ligand in Fig. 1b) might enable formation of higher nuclearity clusters while maintaining low coordination numbers. Using the previously reported iron(I) complex L^MeFe(C₆H₆)²⁷ (L^Me = 2,3,4-trimethyl-2,4-bis(2,6-dimethylphenyl)iminopent-3-yl) we found that reduction with KC₈ followed by addition of the S-atom transfer reagent SPMe₃ at low temperature resulted in formation of a single major product, as judged by ¹H NMR spectroscopy (Fig. 1b), which could be isolated in 65% yield. Single crystal X-ray diffraction revealed the structure to be a tetranuclear [4Fe-3S] iron sulfur cluster (1) with three bridging sulfides that connect three iron-diketiminate fragments in a hexagonal arrangement (Fig. 1c). A fourth Fe atom is ligated only by three sulfide ligands, and is planar (sum of angles at Fe = 360 ± 0.5°). There are also two potassium cations located above and below the plane of the cluster. The connectivity of the [4Fe-3S] core was further established by the synthesis and structural characterization of analogues of 1 with Rb and Cs countercations (described in the Supplementary Material).

The [4Fe-3S] cluster 1 is the first example of an iron-sulfur cluster that contains a three-coordinate iron site in an all-sulfide environment, providing a unique opportunity to learn about the properties of iron with one fewer sulfide ligands than found in previous clusters. We briefly compare the core of 1 to other clusters, including the FeMoco (Fig. 2). The majority of known four iron-sulfur clusters are [4Fe-4S] cubanes; to date, the most reduced biological clusters have all Fe²⁺ ions^15,28, while synthetic nitrosyl-bound cubanes have been reported with formal oxidation states of less than +2 per iron^29,30. The new [4Fe-3S] cluster 1 formally contains three Fe²⁺ sites and one Fe¹⁺ site, making it among the most reduced tetranuclear iron-sulfur clusters. The peripheral Fe–S bonds in 1 of 2.34 ± 0.01 Å are similar in length to the all-Fe²⁺ cubane of the [4Fe-4S] iron protein in nitrogenase (average Fe–S bond lengths of 2.33 ± 0.05 Ų⁸), and are within error of those reported in a synthetic cubane at the all-ferrous oxidation level (2.33 ± 0.05 Å)⁵.

Figure 2. Structural analogy between [4Fe-3S] and the FeMoco.

In blue are the Fe–S bond distances at the three-coordinate iron of 1 (a), which are within error of those at the Fe6 site of the FeMoco²² (b). In b, sulfur atoms are shown in yellow, molybdenum is in green, and carbon is in grey. The core of the [4Fe-3S] cluster—shown in red—is structurally similar to a proposed binding site in iron-sulfur clusters like the FeMoco, also highlighted in red.

The average Fe–S lengths to the central three-coordinate iron in 1 (2.17 ± 0.01 Å), on the other hand, are significantly shorter than average Fe–S bonds in [4Fe-4S] clusters (2.29 ± 0.04 Å)³¹, but within error of the Fe–S bond in a three-coordinate diketiminate-supported diiron(I) system with only one sulfide donor (2.174 ± 0.006 Å)²⁶. Similarly short Fe–S bonds have been observed at tetrahedral sulfide-supported Fe in “ladder” type [3Fe-4S] clusters (average Fe–S of 2.193 ± 0.07 Å)³² as well as in a hexanuclear [6Fe-S5] cluster (average Fe–S of 2.20 ± 0.01 Å in a Fe₄(μ₂-S)₂ core)³³. The short Fe–S bond distances to the central three-coordinate iron in 1 are also reminiscent of those to the belt Fe sites in the FeMoco (average Fe6–S bond length is 2.20 Å; Fig. 2b)²². The geometry of the central Fe site resembles the one in a three-coordinate Fe²⁺ tris(thiolate) complex reported by Power, which exhibits a planar coordination environment (sum of angles at Fe = 359.9 ± 0.2), but the Fe–S bonds at the tris(thiolate) site are 0.1 Å longer than those in 1 (average Fe–S of 2.274 ± 0.005 Å)³⁴. Although there is no precedent for planar iron-sulfide clusters with more than two Fe atoms, a planar arrangement similar to the central site in 1 has been proposed within a potential intermediate state of the FeMoco²⁰. Attainment of this planar geometry would require dissociation of the central carbide from the FeMoco resting state, in which the distorted tetrahedral geometry at Fe6 (τ = 0.88³⁵) leads to a 0.47 Å separation between Fe6 and the plane defined by three sulfide ligands.

The cyclic voltammogram (CV) of 1 in THF solution with 0.2 M tetrabutylammonium hexafluorophosphate electrolyte exhibits a reversible reduction at −3.14 V versus ferrocene (Supplementary Fig. 44), and addition of one molar equivalent of the strong reductant KC₈ led to the formation of one-electron reduced 2. X-ray crystallography shows that the [4Fe-3S] core is retained in 2, and an additional potassium cation interacts with the aryl group of a diketiminate supporting ligand (Fig. 1c). Cluster 2 formally contains two Fe¹⁺ sites and two Fe²⁺ sites, and is unstable in solution (t_1/2 = 20 min by UV-vis spectroscopy; see Supplementary Fig. 47–48), which prevented more thorough characterization. The CV of 1 also shows an irreversible oxidation at −1.89 V, and use of KBArF electrolyte (BArF = tetrakis[3,5-bis(trifluoromethyl)phenyl]borate) results in a quasi-reversible wave at −2.05 V (see Supplementary Fig. 44). Oxidation with one molar equivalent of AgPF₆ forms a new product 3 in 77% isolated yield. X-ray crystallography reveals 3 to be a more oxidized all-Fe²⁺ cluster in which one of the two potassium cations present in starting material 1 is absent (Fig. 1c). Overall, the [4Fe-3S] core is retained across three oxidation levels, with minor structural perturbations accompanying these changes in cluster oxidation level (see Supplementary Fig. 50–51 for structural comparisons of 1–3). The minor structural change across oxidation states is reminiscent of the [4Fe-4S] iron protein in nitrogenase, which also undergoes little reorganization over three oxidation levels²⁸.

Spectroscopic and Computational Studies

The physical oxidation states of the Fe atoms in the different oxidation levels of the [4Fe-3S] clusters were characterized using iron K-edge X-ray absorption spectroscopy (XAS). Analysis of three- and four-coordinate Fe²⁺ reference compounds in addition to clusters 1, 3, and 4 (a more oxidized cluster described below) revealed a clear trend in the near edge region of the XAS data (Fig. 3a and Supplementary Fig. 40). We assign a shoulder at ~7115 eV to a 1s → 4p_z transition at the central trigonal-planar iron atom (see discussion in the Supplementary Material), which shifts to higher energy by 0.41 eV going from Fe¹⁺ containing 1 to all-ferrous 3. These data are consistent with oxidation of the central iron atom.

Figure 3. Spectroscopy of three-coordinate [4Fe-3S] clusters and amide coordinated product 4 gives insight into electronic structures.

a, Fe K-edge XAS spectra of [4Fe-3S] clusters [4Fe-3S][K]₂ (1), [4Fe-3S][K] (3), and [4Fe-3S][K]-NH₂ (4) (top) and the first derivatives (bottom). N.I. represents normalized intensity. These data indicate that metal-centered oxidation occurs at the central Fe site in going from reduced cluster 1 to oxidized cluster 3. b, Mössbauer spectrum of 1 at 80 K, which shows three components in a 2:1:1 ratio for three high-spin Fe²⁺ sites and one central Fe¹⁺ site. Component 1 (fixed at 50%, green): δ = 0.80 mm/s, | ΔE_Q| = 1.57 mm/s; Component 2 (fixed at 25%, blue): δ = 0.86 mm/s, | ΔE_Q| = 0.68 mm/s; Component 3 (fixed at 25%, orange) with δ = 0.34 mm/s, |ΔE_Q| = 1.38 mm/s. c, Mössbauer spectrum of 3 at 80 K, in which the minor component corresponds to an unusual low-spin Fe²⁺ site at the center of the cluster. Component 1 (75%, green): δ = 0.81 mm/s, |ΔE_Q| = 1.87 mm/s; Component 2 (25%, blue): δ = 0.37 mm/s, |ΔE_Q| = 0 mm/s.

We evaluated the electronic structures of 1 and 3 using Mössbauer spectroscopy, electron paramagnetic resonance (EPR) spectroscopy, and ab initio calculations. We first discuss cluster 3, which has all Fe²⁺ sites and accordingly displays a featureless perpendicular mode EPR spectrum. Its Mössbauer spectrum (Fig. 3c) is a superposition of two doublets in a 3:1 ratio that correlates to the outside and inside Fe sites. The smaller subspectrum (assigned to the central Fe²⁺ site) has an isomer shift of 0.37 mm/s that is far outside the range reported for Fe²⁺ sites in previously reported iron-sulfur clusters (0.6–0.8 mm/s)^5,7,36,37. Instead, this low isomer shift is characteristic of low-spin Fe²⁺ sites³⁸, which is surprising to find in the presence of three (normally) weak-field sulfide ligands. The low-spin d⁶ assignment is corroborated by the negligible quadrupole splitting observed for the Mössbauer signal for the central iron, and confirmed by variable-field Mössbauer spectra of 3 that reveal the central iron to be diamagnetic (Supplementary Fig. 68). Mössbauer spectra of the Rb and Cs analogues are almost identical to that of 3 (see Supplementary Fig. 35 and Supplementary Fig. 37).

Wavefunction based ab initio calculations at the complete active space self-consistent field (CASSCF) and N-electron valence-perturbation theory (NEVPT2) levels were carried out. In order to use the ab initio ligand theory (AILFT) method³⁹ for insight into the local ligand fields of each individual Fe²⁺ site in 3, we constructed four models of 3 where all but one of the Fe²⁺ ions were replaced with Zn²⁺ ions. This isolates the change in the ligand-field characteristics of each iron site. The calculations suggest two reasons for the surprising low-spin configuration at the planar Fe²⁺ site (Fe4). First, the very short Fe–S bonds to the central Fe4 (right of Fig. 4, in red) cause the near-degenerate d_xy and d_x2-y2 orbitals to be 6,000 cm⁻¹ higher in energy than the next set of orbitals (d_xz and d_yz), a splitting that is much larger than usual tetrahedral sites in iron-sulfur clusters (and the outer sites in 1–3), which have ligand-field splittings of ~3,500 cm^{−1 40}. Second, the high covalency of the Fe–S bonds gives a reduction in the Racah B parameter,⁴¹ which is calculated to be 6–8% lower at the central Fe than the outer sites. This decrease in electron-electron repulsion also contributes to a preference for a low-spin electronic configuration, in which these two high-lying orbitals are unoccupied.

Figure 4. Calculated ligand field splitting diagram of iron sites in all-ferrous [4Fe-3S][K] (3).

Orbital energies were derived from AILFT on the basis of CASSCF/NEVPT2 calculations for models that isolate each iron site by replacing the other sites with closed-shell Zn²⁺. The z axis is taken to be perpendicular to the Fe₄S₃ plane. The numbering of Fe sites corresponds to the X-ray crystal structure, and Fe4 is the central Fe site. The calculated ligand field splitting reveals that very short Fe–S bonds to the central Fe4 cause the d_xy and d_x2-y2 orbitals to be 6,000 cm⁻¹ higher in energy than the next set of orbitals (d_xz and d_yz). This splitting is much larger than in typical tetrahedral sites of iron-sulfur clusters, and helps rationalize the low-spin configuration of the central Fe4 site.

The reduced complex 1 has three Fe²⁺ ions and one Fe¹⁺ ion, and is a rare example of Fe¹⁺ bound to sulfide^26,42. At 10 K in a toluene glass, 1 exhibits a nearly axial EPR spectrum near g_eff = 2 (Supplementary Fig. 38), which indicates an S = 1/2 ground state from antiferromagnetic coupling between the four paramagnetic iron centers. We sought to determine the location of the ‘extra’ electron in this reduced mixed-valent cluster. At 80 K, zero-field Mössbauer measurements of powder samples of 1 show three quadrupole doublets in a 2:1:1 ratio (Fig. 3b). DFT calculations gave Mössbauer parameters that agree with the experimental spectrum within 0.10 mm/s only when the central iron site is assigned as Fe¹⁺ (see Supplementary Table 4 and Supplementary Fig. 59). This assignment also agrees with the XAS data described above. There are only two previous examples of Fe¹⁺ bound to sulfide: one with a phosphorus-based supporting ligand and low-spin iron(I)⁴², and one with a nitrogen-based supporting ligand and high-spin iron(I)²⁶. In 1, the isomer shift of the central iron (0.34 mm/s) is closer to the low-spin Fe¹⁺ precedent (0.22 mm/s) than the high-spin precedent (0.67 mm/s), suggesting that 1 also has low-spin Fe¹⁺. In addition, the isomer shift of this site is very similar to that in the central site of 3, which is low-spin as discussed above. However, given the empirical nature of this evidence and the few precedents, our conclusions regarding the spin state at the Fe¹⁺ site in 1 should be considered tentative. The outer three iron sites are high-spin Fe²⁺, as shown by their nearly identical isomer shifts and excellent agreement with other tetrahedral ferrous ions bound to sulfide and nitrogen atom donors (~0.8 mm/s)^7,37. One of the outer Fe²⁺ sites in 1 has a different quadrupole splitting because of a twisting of its geometry as observed in the X-ray crystal structure (Supplementary Fig. 55). Mössbauer spectra of different alkali-metal analogues show that the quadrupole splitting of the outer Fe sites has a small but systematic variation with the twist angle, decreasing in the order K>Rb>Cs (see Supplementary Fig. 56 for details).

Upon reduction of [4Fe-3S] cluster 2 to form 1, the ligand field at the central Fe4 site is perturbed as expected for a low-spin iron(I) ion with high-lying near-degenerate d_xy and d_x2-y2 orbitals. As shown in Supplementary Fig. 61, the energy gap between the d_xy and d_x2-y2 orbitals at the central Fe¹⁺ site in 1 is ~500 cm⁻¹, compared to only ~51 cm⁻¹ in the all-Fe²⁺ cluster (3). This splitting is consistent with the expected Jahn-Teller distortion of a low-spin iron(I) complex (see SI for details).

Thus, the new type of planar three-coordinate Fe site in clusters 1 and 3 has two distinctive spectroscopic signatures: a pronounced pre-edge feature at 7115 eV that corresponds to an unusually low energy 1s → 4p_z transition, and a Mössbauer signal with a low isomer shift of 0.36 ± 0.02 mm/s. This low isomer shift is seen in the three-coordinate all-sulfide site in both oxidation levels, and with a variety of alkali metal capping cations. We suggest that these spectroscopic signatures could help to identify unsaturated sites in intermediate states of enzymes like nitrogenases and that the low-spin configuration should be sought in unsaturated sites in natural iron-sulfur clusters.

Reactivity with N–N Bonds

The presence of an unsaturated three-coordinate iron site in 1–3 prompted us to explore the reactivity of these clusters with small molecules relevant to nitrogenase. We observed that the reaction of 3 with 0.5 molar equivalent of the nitrogenase substrate hydrazine (N₂H₄) results in rapid and near-quantitative (80% spectroscopic yield) formation of a new product with C_3v symmetry, as judged by ¹H NMR analysis (Fig. 5a). This new product (4) gave purple-red crystals for which diffraction reveals a [4Fe-3S] cluster in which a new N-based ligand is bound to the central iron (Fig. 5b). It was possible to identify this ligand as amide (NH₂) through IR and ¹H NMR spectroscopies (Fig. 5d). The IR spectrum of crystalline 4 shows two N–H stretching vibrations at 3262 and 3352 cm⁻¹, which shift to 2499 and 2372 cm⁻¹ when 4 is synthesized from N₂D₄ (Supplementary Fig. 26). The ¹H NMR spectrum of 4 can be fully assigned based on integrations of the various paramagnetically shifted signals, and the peak corresponding to the NH₂ protons is absent when deuterated 4 is measured (Fig. 5d). The identity of the NH₂ ligand was also verified by treating 4 with excess HCl, which produced 0.81 equiv of NH₃ per equiv of 4. Amide 4 can be independently synthesized using Carpino’s hydrazine⁴³ (see Supplementary Material for details) which demonstrates its utility as an amido transfer reagent.

Figure 5. [4Fe-3S] reactivity with hydrazine and characterization of Fe–NH₂ product.

a, Reaction of all-ferrous [4Fe-3S] iron-sulfur cluster 3 with hydrazine, a nitrogenase substrate. b, Crystal structure of 4, the resulting amide-coordinated [4Fe-3S] cluster. c, Mössbauer spectrum of 4 at 80 K. Component 1 (fixed at 25%, green): δ = 0.74 mm/s, |ΔE_Q| = 1.96 mm/s; Component 2 (fixed at 75%, blue): δ = 0.76 mm/s, |ΔE_Q| = 1.62 mm/s. d, Region of the IR spectrum of natural abundance 4 (blue trace) and deuterated 4 (black trace). Inset: ¹H NMR spectrum of 4 showing NH₂ protons (blue trace) overlaid with the spectrum of deuterated 4 (black trace) in which amide protons are not observed. These spectroscopic data identify the product of hydrazine reduction by all-ferrous 3 as an iron amide (Fe–NH₂). Full spectra are shown in Supplementary Figures 20, 26, and 33.

The rapid reaction of 3 with N₂H₄ shows that this all-Fe²⁺ cluster can break an N–N bond, and it is significant that we have observed a hydrazine-derived NH₂ group on an iron-sulfur cluster for the first time. Despite their implication in late-stage nitrogen reduction by the FeMoco²⁰, there are only three previously reported terminal iron amide (Fe–NH₂) complexes, none of which are iron-sulfur clusters, and none of which derive from precursors with N–N bonds^44,45,46. However, Szymczak recently reported a bridging iron amide formed from hydrazine⁴⁷. The Fe–N bond distance in 4 is 2.06 ± 0.02 Å, which lies in the range of previous parent amido iron complexes (1.918 ± 0.003 to 2.069 ± 0.007 Å)^44,45,46. The most salient difference between the molecular structure of 4 and the [4Fe-3S] clusters 1–3 is a change in geometry of the central Fe from trigonal planar to distorted tetrahedral, in which the central Fe is raised 0.9 Å out of the plane defined by the six peripheral iron and sulfur atoms (Supplementary Fig. 54).

Compound 4 formally contains three Fe²⁺ ions and one Fe³⁺ ion. The zero-field Mössbauer spectrum of 4 at 80 K has two strongly overlapping quadrupole doublets in a 3:1 ratio, with isomer shifts of 0.76 and 0.74 mm/s, respectively (Fig. 5c). While the 3:1 ratio could be taken to imply that the Fe³⁺ site is the central NH₂-bound iron, its isomer shift is inconsistent with previous four-coordinate Fe³⁺ sulfide sites (~0.3 mm/s)^6,48. Therefore, we suggest that the hole is delocalized over the three peripheral Fe sites, with averaged oxidation states of +2.33, while the central iron is high-spin Fe²⁺. Consistent with this assignment, the Fe–S bond distances at the central Fe (avg. 2.297 ± 0.002 Å) are similar to those in another four-coordinate high-spin Fe²⁺ site supported by three sulfide donors and a carbene ligand (2.33 ± 0.002 Å)⁵.

Consistent with the proposal that sulfide-supported Fe²⁺ in the key substrate binding intermediate of the FeMoco performs substrate reductions^24,49, our results demonstrate that Fe²⁺ in 3 is sufficiently reduced to cleave the N-N bond in hydrazine, and suggest that Fe²⁺ is an appropriate oxidation state to target in reactive iron-sulfur clusters. Additionally, reduction of hydrazine with 3 with to form NH₂-bound 4 is relevant to substrate reduction by the FeMoco. Rapid freeze-trapping during nitrogenase turnover with hydrazine, diazene, and methyldiazene has revealed a common intermediate with a single N environment by EPR and ENDOR^50,51. This species has been proposed to be either an NH_x species or a highly symmetric hydrazine adduct. Reaction of our [4Fe-3S] model system with N₂H₄ to form NH₂ therefore supports the feasibility of the former proposal for the FeMoco. The approach here uses a synthetic iron-sulfur cluster in which the reactive Fe site is supported exclusively by sulfide, which is an important step beyond molecular Fe complexes supported by non-biologically relevant N, P, and B coordinated supporting ligands that have previously furnished N₂ reduction intermediates^52,53,54. Cluster 4 provides the first precedent for binding of an N_xH_y intermediate to a synthetic iron-sulfur cluster. Though we have not observed reactions of 1–3 with N₂, we anticipate that future variation of the ligand, countercation, or oxidation state will enable this more difficult reduction.

Conclusion

Using diketiminate supporting ligands, it is possible to isolate a stable iron-sulfur cluster that has all-sulfide coordination at a planar three-coordinate iron site. The [4Fe-3S] clusters reported here show that lowering the coordination number of iron can cause a surprising change to a low-spin electronic configuration, which is the result of short, highly covalent in-plane Fe–S bonds that lead to strong ligand-field splitting. The unusual geometry leads to XAS and Mössbauer signatures that may help to identify trigonal-planar sites in biological clusters. In addition, we show that this new type of low-coordinate site can reduce N₂H₄, in an example of biomimetic N–N reduction. More generally, these findings demonstrate the utility of targeting intermediate structures, in addition to enzymatic resting states, as a strategy to realize biologically relevant reactivity through synthetic modeling.