Nitrogenase-Relevant Reactivity of a Synthetic Iron-Sulfur-Carbon Site

Abstract

Simple synthetic compounds with only S and C donors offer a ligation environment similar to the active site of nitrogenase (FeMoco) and thus demonstrate reasonable mechanisms and geometries for N₂ binding and reduction in nature. We recently reported the first example of N₂ binding at a mononuclear iron site supported by only S and C donors. In this work, we report experiments that examine the mechanism of N₂ binding in this system. The reduction of an iron(II) tris(thiolate) complex with one equivalent of KC₈ leads to a new, thermally unstable intermediate and a combination of Mössbauer, EPR and X-ray absorption spectroscopies identify it as a high spin (S = 3/2) iron(I) species that maintains coordination of all three sulfur atoms. DFT calculations suggest that this iron(I) intermediate has a pseudotetrahedral geometry that resembles the S₃C iron coordination environment of the belt iron sites in the resting state of the FeMoco. Further reduction to the iron(0) oxidation level under argon causes the dissociation of one of the thiolate donors and gives an η⁶-arene species which reacts with N₂. Thus, in this system the loss of thiolate and binding of N₂ require reduction beyond the iron(I) level to the iron(0) level. Further reduction of the iron(0)-N₂ complexes gives a reactive, formally iron(-I) species. Treatment of the putative iron(-I) complex with weak acids gives low yields of ammonia and hydrazine, demonstrating that these nitrogenase products can be generated from N₂ at a synthetic Fe-S-C site. Catalytic N₂ reduction is not observed, which is attributed to protonation of the supporting ligand and degradation of the complex via ligand dissociation. Identification of the challenges in this system give insight into the design features needed for functional biomimetic complexes.

Graphical Abstract

INTRODUCTION

Biological fixation of dinitrogen to ammonia occurs at the iron- and sulfur-containing cofactors of nitrogenase enzymes.^1-2 The most thoroughly studied nitrogenase reduces N₂ at the iron-molybdenum cofactor (FeMoco), a cluster that contains seven iron atoms and a molybdenum atom (Figure 1, top).^3-5 Studies on the reactions of nitrogenase variants with substrates such as alkynes, propargyl alcohol, carbon monoxide, and selenocyanate strongly implicate the iron sites of the FeMoco as the N₂ binding sites.^6-13 These iron atoms are coordinated by three sulfides and a carbide in the resting state.

Figure 1.

Top: structure of the resting state of the FeMoco with one of the pseudotetrahedral FeS3C sites emphasized. Bottom: hypothetical structure of an N₂-bound species in which a protonated sulfide dissociates from an iron site to enable N₂ binding. Though the structure of the N₂-bound species is currently unclear, this illustrates the potential role of Fe–S cleavage and carbide coordination during N₂ binding.

Starting from the resting state, the FeMoco accumulates four reducing equivalents and four protons before N₂ binding occurs, along with reductive elimination of H₂.^{1, 14} This process presumably involves structural changes that create an open coordination site for substrate binding at an iron center. There is growing evidence that Fe–S dissociation within the cofactor is possible (Figure 1, bottom), as supported by the crystallographically observed substitution of the central sulfide bridge in the FeMoco by CO and by selenide.^12-13 More recently, Einsle and co-workers reported a structure of vanadium nitrogenase in which one of the bridging sulfides in the iron-vanadium cofactor (FeVco) is displaced from the active site, and a light atom (assigned as N₂-derived NH, though it may instead be OH^15-16) lies between these two iron centers.¹⁷ Computational studies have also suggested that sulfide lability or hemilability may be mechanistically important in nitrogenases.^18-23

The unusual structure and reactivity of the FeMoco have encouraged studies of the coordination chemistry of simple synthetic iron complexes with the ability to bind and/or reduce N₂.^24-33 Notably, catalytic ammonia and hydrazine production from N₂ using strong reductants and strong acids has been achieved using phosphine-supported complexes.^34-40 Sulfur donors, in the form of thioethers and thiolates,^41-45 and carbon donors, in the form of carbanions or N-heterocyclic carbenes,^{35-36, 46-51} have also been incorporated in ligand systems. However, none of these complexes rely on ligands that contain only the biologically relevant sulfur and carbon donors.

Recently, we reported the first example of a structurally characterized Fe-N₂ complex with a mixed sulfur/carbon ligand sphere, which utilizes a bulky bis(thiolate) ligand with a terphenyl core.⁴³ Complex 1²⁻ (Scheme 1) features N₂ coordinated to a formally iron(0) center in a pseudotetrahedral geometry (considering a C=C bond of the central arene as a single donor). This product could be formed from reduction of the iron(II) tris(thiolate) species 2-K at low temperature under an atmosphere of N₂ (Scheme 1). The transformation of 2-K to 1-K₂ provides a small-molecule model of Fe-S dissociation that leads to N₂ binding. However, the reported study does not distinguish the order of the processes that occur in this transformation (thiolate loss, N₂ binding, and two-electron reduction), which could provide insight into the behavior of iron-sulfur-carbon sites under reducing conditions.

Scheme 1.

Reduction of an Iron(II) Tris(thiolate) Complex Involving Thiolate Dissociation and N₂ Binding⁴³

Here, we report mechanistic details of the Fe-S dissociation/N₂ binding process in this synthetic system through characterization of several thermally sensitive low-valent iron species. Characterization of the iron(I) and iron(0) precursors to N₂ binding helps to elucidate the reasons for the weak N₂ binding observed in this complex. We also demonstrate the production of ammonia and hydrazine from N₂ through further reduction of 1-K₂ followed by treatment with weak acids.

RESULTS

Reduction of Iron(II) to Iron(I) Without Thiolate Dissociation.

The product of the one-electron reduction of 2-K was first examined in order to determine whether the Fe–S bond of the monothiolate remains intact at the iron(I) redox level. Treatment of a solution of 2-K in THF with 1.1 equiv of KC₈ at −70 °C under N₂ results in an immediate color change to dark red. The Mössbauer spectrum of the frozen reaction mixture indicates a single species with δ = 0.80 mm/s and ∣ΔE_Q∣ = 2.01 mm/s (Table 1 and Figure 1). In contrast, the previously isolated⁴³ iron(I) bis(thiolate) complex 3-K exhibits a broad, unresolved spectrum in frozen THF under the same conditions (Figures S5-S8). Thus, Mössbauer spectroscopy indicates that one-electron reduction of 2-K produces a species (4-K₂) that is distinct from 3-K. Furthermore, the same species is formed in an analogous reduction under an Ar atmosphere (Figure S9), indicating that 4-K₂ does not contain N₂. Based on these results, we assign 4-K₂ as an iron(I) tris(thiolate) complex, in which the iron center is reduced but retains both the monodentate thiolate and bis(thiolate) ligands (Scheme 2).

Table 1.

Mössbauer parameters for compounds 1 − 7.

Compound	Solid	Frozen THF Solution	DFT- Calculated
1-K2	δ = 0.74 mm/s a, b ∣ΔEQ∣ = 1.79 mm/s	δ = 0.70 mm/s ∣ΔEQ∣ = 1.32 mm/s	δ = 0.77 mm/s ∣ΔEQ∣ = 1.78 mm/s
2-K	δ = 0.69 mm/s a ∣ΔEQ∣ = 1.50 mm/s	δ = 0.84 mm/s c ∣ΔEQ∣ = 3.00 mm/s	δ = 0.62 mm/s ∣ΔEQ∣ = 2.76 mm/s
3-K	δ = 0.59 mm/s a ∣ΔEQ∣ = 0.53 mm/s	Broad, unresolved spectrum	δ = 0.67 mm/s ∣ΔEQ∣ = 0.75 mm/s
4-K2	Not isolable	δ = 0.80 mm/s ∣ΔEQ∣ = 2.01 mm/s	δ = 0.78 mm/s d ∣ΔEQ∣ = 1.64 mm/s
5	δ = 0.89 mm/s a ∣ΔEQ∣ = 3.77 mm/s	Poor solubility in THF	δ = 0.78 mm/s e ∣ΔEQ∣ = 3.81 mm/s
6-K2	δ = 0.97 mm/s b ∣ΔEQ∣ = 2.07 mm/s	δ = 1.00 mm/s ∣ΔEQ∣ = 2.02 mm/s	δ = 1.07 mm/s f ∣ΔEQ∣ = 1.86 mm/s
7-K3	Not isolable	δ = 0.61 mm/s ∣ΔEQ∣ = 2.48 mm/s	Not calculated

^aMössbauer parameters for solid samples were reported previously.⁴³

^bAs 18-crown-6 adduct.

^cAs previously reported⁴³, 2-K binds THF at low temperature which accounts for the difference between its solid-state and solution Mössbauer parameters (see Figure S4)

^dCalculated values for S = 3/2 model.

^eCalculated Mössbauer parameters were previously reported.⁵⁷

^fCalculated values for S = 1 model.

Scheme 2.

Iron(I) Thiolate Complexes

Further insight into the nature of 4-K₂ was obtained from EPR spectroscopy. Figure 2 shows the X-band EPR spectrum of the crude reduction product following transfer to a pre-cooled EPR tube in a −70 °C cold well and immediate flash-freezing in liquid nitrogen. Two signals with slightly different rhombicities appear at g = [5.59, 2.75, 1.76] (65%, red) and g = [5.35, 3.04, 1.85] (35%, blue), which are each indicative of an S = 3/2 spin state. The reason for the appearance of two components in the EPR spectrum is not known, but it is possible that the two components represent different conformers of 4-K₂ which have similar Mössbauer spectra that cannot be resolved at 80 K at zero field.^52-56 In these EPR spectra, the previously characterized⁴³ bis(thiolate) complex 3-K is also present as a 10% impurity (S = 1/2 signal in the g = 2 region), as determined by comparison to a sample of 3-K generated via reduction of the iron(II) bis(thiolate) complex 5 under the same conditions (Figure S20). Warming the reaction mixture to room temperature for 15 minutes results in a decrease in the S = 3/2 signals of 4-K₂ (Figure S19) and an increase in the S = 1/2 signal of 3-K (Figure S20), suggesting that the small amount of 3-K present in these samples arises from thermal decomposition of 4-K₂ during transfer to the EPR tube. This proposal is also supported by Mössbauer spectroscopy; the spectrum of 4-K₂ after stirring at room temperature for 20 minutes contains a broad feature consistent with the presence of 3-K (Figure S10). Thus, all our evidence indicates that 4-K₂ is a high-spin iron(I) complex that differs from 3-K, presumably because it retains the monodentate thiolate ligand.

Figure 2.

Mössbauer spectrum of a frozen reaction mixture containing 4-K₂ generated by treatment of 2-K with 1.1 equiv of KC₈ in THF under N₂. The black circles are the experimental data, the red line is a simulation of 4-K₂ with the fit parameters shown in Table 1, and the gray line is the residual. The slight asymmetry may be attributed to broadening from paramagnetic relaxation at the measurement temperature of 80 K.

Since attempts to crystallize 4-K₂ have not been successful, insight into its structure was obtained from X-ray absorption studies (Figure 3). In order to validate this approach, we began by collecting EXAFS data for 2-K and 3-K, which have previously been characterized crystallographically.⁴³ Fits of the k³-weighted EXAFS of 2-K indicate the presence of 3 Fe-S interactions at 2.33 Å and multiple Fe-C interactions, with the closest at 2.42 Å (Table S3). These are in reasonable agreement with the crystallographically determined bond lengths for this species (Table S4). The 2 Fe-S and 6 Fe-C interactions in 3-K determined by EXAFS (2.25 Å and 2.08 Å, respectively) are also in good agreement with the crystal structure (Tables S5 and S6). The EXAFS fits of 4-K₂ (Table S7) require 3 Fe-S interactions at a similar distance to those in 2-K (2.32 Å). Notably, these distances are significantly longer than those of the low-spin iron bis(thiolate) complex 3-K, suggesting that 4-K₂ is high-spin (in agreement with the interpretation of the EPR spectra described above). The fit also requires the inclusion of 2 short Fe-C interactions at 2.02 Å, implying η² binding to the central arene ring.

Figure 3.

X-band EPR spectrum of a frozen reaction mixture of 2-K treated with 1.1 equiv of KC₈ in 2-methyltetrahydrofuran recorded at 5 K. The experimental spectrum is shown in the top frame. A simulation for the S = 3/2 species (corresponding to 4-K₂) is shown as a solid black line in the bottom frame and the two components of the fit are shown as blue and red dashed lines. The S = 1/2 signal corresponding to 3-K in the g = 2 region is not included in the fit; this signal is power-saturated under these conditions (see Figure S20).

A more detailed structural model of 4-K₂ was obtained from DFT calculations, using a truncated model in which the triisopropylphenyl groups were removed (see Supplementary Information for computational details). This method and level of truncation gave highly accurate models of 1-K₂ and 5 in previous studies,^{43, 57} as well as of 2-K and 3-K (see Tables S10 and S11). Structures of the hypothesized iron(I) tris(thiolate) intermediate in the two possible spin states (S = 1/2 and S = 3/2) were optimized at the BP86/def2-TZVP level. The calculations indicate the S = 3/2 configuration to be more stable with ΔG_l.s.-h.s. = +14 kcal/mol, in agreement with the EPR evidence that 4-K₂ is high-spin (see above). The calculated Fe-C bond lengths of both models are longer than those determined by EXAFS. However, the S = 3/2 DFT model reproduces the experimental Fe-S distances, whereas the Fe-S bonds are too short in the S = 1/2 DFT model (see Table S8). Furthermore, the calculated Mössbauer parameters for the high spin DFT model (δ = 0.78 mm/s, ∣ΔE_Q∣ = 1.64 mm/s) are close to those observed experimentally and agree much better than the parameters for the alternative low spin DFT model (δ = 0.64 mm/s, ∣ΔE_Q∣ = 1.04 mm/s).⁵⁷ The DFT calculations and experimental results therefore both support assigning 4²⁻ as an S = 3/2 iron(I) tris(thiolate) complex. In the computationally optimized structure, three thiolate donors and an asymmetrical η² coordination of the central arene ring (Fe-C distances = 2.02, 2.21 Å) form a pseudotetrahedral ligand environment around the iron(I) site (Figure 4). The immediate coordination sphere of 4²⁻ therefore resembles the environment of the individual iron centers in the resting state of the FeMoco, although there are likely to be electronic differences (see Discussion section).

Figure 4.

Non-phase-shift-corrected FT of the Fe K-edge EXAFS spectra (black, solid) and fits (red, dashed) of a) 2-K (3S/5C model), b) 3-K (2S/6C/CC model) and c) 4-K₂ (2-3S/6C model). Parameters and details of the fits are provided in the Supplementary Information.

Reduction to Iron(0).

In order to determine whether monothiolate loss is triggered by reduction or occurs only in the presence of N₂, the two-electron reductions of the iron(II) complexes 2-K and 5 were examined under argon atmosphere (Scheme 3). Reduction of 2-K with 2.4 equiv of KC₈ under an argon atmosphere at −65 °C yields a new species (6-K₂) with Mössbauer parameters δ = 1.00 mm/s and ∣ΔE_Q∣= 2.02 mm/s) (Figure 6, top). Reduction of the iron(II) bis(thiolate) complex 5 with excess KC₈ results in formation of the same species, although samples prepared in this manner are contaminated with the iron(I) bis(thiolate) complex 3-K (Figure S11).

Scheme 3.

Iron(0) thiolate complexes

Figure 6.

Mössbauer spectra of frozen reaction mixtures of 2-K treated with 2.4 equiv of KC₈ in THF under argon (top), under argon followed by addition of N₂ (middle), and under an N₂ atmosphere (bottom). The black circles are the experimental data, and the grey lines are the residuals for the total fits. The solid lines correspond to fits with the parameters given in Table 1.

We were unable to crystallize 6-K₂ from the reduction of 2-K with excess KC₈. However, reduction of 5 with excess KC₈ at −65 °C in THF in the presence of 18-crown-6, followed by brief warming, affords 6-[K(18-crown-6)]₂, which crystallized after standing overnight at −40 °C. X-ray diffraction analysis shows that the anion in 6 is an iron(0) bis(thiolate) complex with η⁶ coordination to the central arene ring (Figure 6 and Scheme 3). Complex 6-[K(18-crown-6)]₂ is structurally similar to the iron(I) complex 3-K with longer bond lengths to ligated atoms, consistent with a more reduced iron site (Table S14).

Although it was possible to obtain clean Mössbauer spectra of solid 6-[K(18-crown-6)]₂ by careful isolation and handling at low temperature (Figure S13), further characterization of this species was hindered by its thermal instability (Figure S14) and low solubility. Insight into the nature of this species was therefore obtained from DFT calculations, using the same methods that were validated above. Geometry optimizations were performed on a truncated model of the anionic portion of 6-[K(18-crown-6)]₂ for two possible spin states (S = 0 and S = 1). The calculations predict the two spin states to be close in energy, with the calculated ground state depending on the functional used. However, the Fe-S and Fe-C bond distances for the S = 1 DFT model are closer to those observed crystallographically (see Table S13). Furthermore, the observed Mössbauer parameters of 6-[K(18-crown-6)]₂ (δ = 1.00 mm/s and ∣ΔE_Q∣= 2.02 mm/s) are closer to those predicted for the S = 1 model (δ = 1.07 mm/s, ∣ΔE_Q∣ = 1.86 mm/s) than those predicted for the S = 0 model (δ = 0.76 mm/s, ∣ΔE_Q∣ = 2.36 mm/s). Therefore, we favor assigning 6²⁻ as an iron(0) complex having a high-spin (S = 1) electronic configuration.

N₂ binding to the Iron(0) Complex.

When a reaction mixture containing 6-K₂ (generated in situ from 2-K under Ar) is treated with N₂ at −65 °C, an immediate color change from dark brown to dark red is observed. Analysis of this reaction mixture using Mössbauer spectroscopy reveals the presence of 1-K₂, along with a small amount of the over-reduced product 7-K₃ (see below) formed from unreacted KC₈ present in the crude reaction mixture. These are the same species observed when the reduction of 2-K is performed under an atmosphere of N₂ (Figure 5 middle and bottom). Since the transformation of 6-K₂ to 1-K₂ occurs rapidly upon addition of N₂, the iron(0) bis(thiolate) species 6-K₂ is a reasonable and kinetically competent intermediate for the reaction sequence that leads to N₂ binding (Scheme 3). It is also significant that reduction of 2-K to 6-K₂ was performed at −65 °C, since Mössbauer spectroscopy demonstrates that the spontaneous loss of thiolate from intermediate 4-K₂ (forming 3-K) does not occur at low temperature within 40 minutes. Therefore, direct monothiolate loss from 4-K₂ is not a kinetically competent step during the formation of 1-K₂.

Figure 5.

Computationally optimized structure of a truncated model of the Fe(I) tris(thiolate) complex 4²⁻ (S = 3/2 model). Selected bond distances: Fe–C = 2.02, 2.21 Å, Fe–S = 2.32, 2.27, 2.31 Å.

In 6-[K(18-crown-6)]₂, there is η⁶-binding of the arene which shifts to η²-binding in 1-[K(18-crown-6)]₂. Changes in hapticity to a central arene have also been observed by Agapie and co-workers in structurally similar terphenyl diphosphine systems.^58-59 This observation suggests that the binding of N₂ competes with additional strong bonding to the arene of the ligand. Correspondingly, the N₂ ligand in 1-K₂, despite engaging in strong π-backbonding based on its low N-N stretching frequency (1880 cm⁻¹), is very labile; the N-N stretching band in the infrared spectrum of 1-[K(18-crown-6)]₂ begins to disappear upon warming the compound above −40 °C.⁴³ Additionally, warming of solid 1-[K(18-crown-6)]₂ to room temperature under vacuum results in partial formation of 6-[K(18-crown-6)]₂, along with other decomposition products (Figure S15). The presence of π-backbonding interactions with both N₂ and the central arene ring is also reflected by the requirement for reduction to the iron(0) level in order to observe N₂ binding, in contrast to other systems where N₂ can bind at the iron(II) or iron(I) redox level.^{26-27, 33}

Further understanding of the bonding in 1²⁻ was obtained from DFT calculations. As reported previously, the DFT model reproduces the metrical parameters from the crystallographic structure of 1-[K(18-crown-6)]₂ well.⁴³ The DFT-predicted Mössbauer parameters are also in good agreement with experimental values (Table 1). In order to determine whether the N₂ and/or the central arene act as redox non-innocent ligands, we examined the unrestricted corresponding orbitals (UCOs)⁶⁰ of 1²⁻ at the B3LYP/def2-TZVP level of theory (Figure 7). The spatial overlaps between the α and β spin orbitals (S_αβ) are close to unity, indicating that neither the N₂ nor the arene act as redox-active ligands, although there is some spin polarization. The DFT calculation also indicates π-backbonding interactions between the iron and both the N₂ and the central arene, supporting the idea that binding of N₂ competes with binding to the central arene of the supporting ligand.

Figure 7.

X-ray crystallographic structure of the anionic portion of Fe(0) bis(thiolate) 6-[K(18-crown-6)]₂ with thermal ellipsoids shown at the 50% probability level. Hydrogen atoms are omitted and the triisopropylphenyl groups are shown in wireframe representation for clarity. Selected bond distances and angles: Fe–C = 2.083(2) − 2.194(2) Å; Fe–S = 2.4036(9), 2.4050(7) Å; S−Fe−S = 104.10(3)°.

Further Reduction and Ammonia Production.

Mössbauer spectra of samples obtained after the treatment of 2-K with 3.6 equiv of KC₈ at −65 °C for 40 min under an atmosphere of Ar are identical to those using 2.4 equiv of KC₈ (Figure S16), indicating that no reduction beyond the Fe(0) level occurs without N₂. However, under an atmosphere of N₂, reduction of 2-K with 3.6 equiv of KC₈ gives a new species (7-K₃) as observed by Mossbauer spectroscopy (Figure 8). This indicates that the iron(0)-N₂ complex 1-K₂ can be reduced by an additional electron to give a formally iron(-I) species. We were unsuccessful in isolating 7-K₃ due to its instability (see Figures S18 and S21), but the requirement for N₂ during its generation indicates that it is also an Fe-N₂ complex. Further reduction of 7-K₃ was not feasible, even with excess KC₈ (Figure S17).

Figure 8.

Unrestricted corresponding orbitals (UCOs) for 1²⁻, as calculated at the B3LYP/def2-TZVP level. S_αβ indicates the spatial overlap between the α and β spin orbitals.

Treatment of the reaction mixtures containing 1-K₂ with 20 equiv of a variety of proton sources gives 1% or less of the N₂ reduction products ammonia (NH₃) and hydrazine (N₂H₄). However, treatment of 7-K₃ with the weak acids 2,6-di-tert-butyl-4-methylphenol (BHT) and H₂O produces substoichiometric amounts of ammonia and hydrazine, respectively (Table 2 and Scheme 4). Treatment with concentrated H₂SO₄ also results in formation of a small amount of N₂H₄. These products were not detected when 7-K₃ was warmed to room temperature (resulting in its decomposition) prior to acid addition, supporting the idea that 7-K₃ itself, rather than some impurity in the system, mediates N₂ reduction.

Table 2.

Yields of NH₃ and N₂H₄ (per equivalent of iron complex) after treatment of crude reaction mixtures containing 1-K₂ or 7-K₃ with 20 equiv of acid. The yields given in bold are an average of multiple runs and the errors correspond to the range of observed values (see Table S2); all other entries are for a single run.

	1-K2	7-K3
[H(OEt2)2][BArF4]	no NH3 no N2H4	no NH3 no N2H4
2,6-di-tert-butyl-4-methylphenol	0.02 NH3 no N2H4	0.17 ± 0.06 NH3 no N2H4
H2O	0.01 NH3 no N2H4	0.02 NH3 0.26 ± 0.04 N2H4
Monothiola	no NH3 no N2H4	no NH3 no N2H4
H2SO4	no NH3 0.01 N2H4	no NH3 0.06 N2H4

^aMonothiol = 2’,4’,6’-triisopropyl-[1,1’-biphenyl]-2-thiol

Scheme 4.

Protonation of N₂ Complexes

Addition of the strong acid [H(OEt₂)₂][BAr^F₄] (Ar^F = 3,5-bis(trifluoromethyl)phenyl) to crude reaction mixtures containing 1-K₂ or 7-K₃ results in bleaching of the deep red color characteristic of these species (Figure S2), which suggests decomposition of the iron complexes via protonation of the supporting ligand. Correspondingly, the ¹H NMR spectrum of a hexanes extract of the crude reaction mixture indicates the presence of the bis(thiol) (Figure S3). No N₂ reduction products were detected from the reactions of 1-K₂ or 7-K₃ with [H(OEt₂)₂][BAr^F₄].

Attempts to perform catalytic N₂ reduction by treating 2-K with 50 equiv of KC₈ and 50 equiv of acid, under conditions similar to those utilized by Peters and others,^34-40 were unsuccessful. Using 2-K as the prospective catalyst, NH₃ is the only product detected, in a substoichiometric yield (0.13, 0.41, and 0.22 equiv of NH₃ per Fe using [H(OEt₂)₂][BAr^F₄], BHT, and H₂O, respectively, as proton sources).

DISCUSSION

Mechanism of Fe-Based N₂ Binding and Reduction.

We have identified very reactive iron(I) and iron(0) intermediates that lie on the route to N₂ binding at a sulfur- and carbon-ligated iron(0) iron center. The iron(I) tris(thiolate) intermediate has a high-spin configuration, and DFT calculations suggest a pseudotetrahedral iron-sulfur-carbon environment reminiscent of the coordination environment in the resting state of the FeMoco. Further reduction to iron(0) results in Fe-S bond cleavage and N₂ coordination. Combining the observations that (a) reduction of iron(II) tris(thiolate) 2-K to iron(I) tris(thiolate) 4-K₂ takes place without loss of thiolate or coordination of N₂ at −70 °C, (b) reduction of iron(I) tris(thiolate) 4-K₂ under argon leads to spontaneous thiolate ejection at −65 °C to give the iron(0) complex 6-K₂, and (c) 6-K₂ binds N₂ to give the iron(0)-N₂ complex 1-K₂, we propose the mechanism in Scheme 5 as the most reasonable pathway for the conversion of 2-K to 1-K₂. Our observations do not rule out an alternative associative mechanism in which N₂ binds prior to ejection of thiolate (bypassing 6-K₂), but we favor the mechanism shown in Scheme 5 because all of the individual steps have been observed.

Scheme 5.

Proposed Mechanism for Formation of 1²⁻ Through Electronation and N₂ Binding

Further reduction of 1-K₂ results in formation of the highly unstable formally iron(-I) species 7-K₃. Although rare, there have been several reports of iron-dinitrogen complexes at this redox level. Ung and Peters have described a cyclic alkyl amino carbene (CAAC) supported system in which an iron(0) can be reduced to iron(-I) under N₂.³⁶ In this case, the iron(-I)–N₂ species was isolable and thus was characterized in more detail. Their analysis indicated substantial spin delocalization onto the CAAC ligand. More recently, Schild and Peters reported a phosphine-supported system where reduction gave formally iron(-I) or iron(-II) species that were more reactive at the N₂ ligands.⁶¹ The structure of 7-K₃ is not clear based on the limited spectroscopic characterization we were able to obtain. It is possible that the central arene (rather than the Fe-N₂ unit) is reduced; it is also possible that one of the thiolate donors dissociates, as observed for highly reduced molybdenum complexes supported by a structurally similar terphenyl bis(phosphine) ligand.^{58-59, 62}

Addition of proton sources to 7-K₃ results in release of substoichiometric NH₃ or N₂H₄. The distribution of N₂-derived products depends on the acid used, with NH₃ as the major product from treatment with BHT and N₂H₄ from treatment with H₂O. However, the reasons for this reactivity difference are unclear because the yields of NH₃ and N₂H₄ in these reactions are low. We note that BHT can act as a hydrogen atom donor as well as an acid, and a related phenol can donate electrons and protons to an N₂-derived NH group.⁶³ There have been several recent reports of using hydrogen-atom donors to functionalize transition-metal coordinated N₂ ligands . These include classical H atom donors such as TEMPO-H⁶⁴, as well as H atom donors generated in situ from samarium diiodide and alcohols^65-66 or metallocenes and weak acids.^{40, 67-68} In reactions with 1-K₂ and 7-K₃, stronger acids such as [H(OEt₂)₂][BAr^F₄] give rapid decomposition via ligand protonation, even at low temperature. It is likely that achieving catalytic N₂ reduction with biomimetic sulfur-carbon supporting ligands will require the design of ligand systems that are less susceptible to acid-induced degradation.

Relevance to Nitrogenase.

As described in the Introduction, recent biochemical, spectroscopic, and crystallographic studies have implicated a S₃CFe site as the N₂ binding site on the FeMoco. The FeMoco is the only known case of a carbide being incorporated into an iron-sulfur cluster. In a very recent synthetic achievement, Rauchfuss and coworkers have isolated an iron-carbide-carbonyl cluster that incorporates a single sulfide ligand.⁶⁹ However, no clusters are yet known that have the characteristic S₃CFe or S₂CFe coordination of the resting state and putative N₂-binding state of the FeMoco. It is also important to note that replicating the exact primary coordination sphere of the FeMoco may not replicate its reactivity. For example, when extracted from nitrogenase the FeMoco can catalyze the reduction of alternative nitrogenase substrates (e.g. acetylene, CO₂, CO, and CN⁻) but not N₂.^70-71

The only examples of well-defined iron complexes containing carbon donors that catalyze N₂ reduction contain CAAC³⁶ or tris(phosphino)alkyl³⁵ ligands but do not also incorporate sulfur donors. Similarly, there are few examples of N₂ reduction by any molecular iron complexes with ligands containing only S donors. An early report from Tanaka and co-workers indicated that controlled potential electrolysis of [Fe₄S₄(SPh)₄]²⁻ or [Mo₂Fe₆S₈(SPh)₉]³⁻ under N₂ atmosphere results in the formation of NH₃ with low Faradaic yield.⁷² More recently, photochemical reduction of N₂ to NH₃ by chalcogels containing MoFeS clusters was reported.^73-74 Ohki and co-workers have also demonstrated catalytic silylation of N₂ by [Mo₂Fe₂] hydride clusters with monothiolate ligands.⁷⁵ However, the exact identities of the active species in all of these systems are unknown, and the key Fe-N₂ complexes were not isolable.

To the best of our knowledge, the results described here represent the first example of NH₃ and/or N₂H₄ production from a well-defined synthetic Fe-N₂ complex with a supporting ligand containing only sulfur and carbon donors. Though the yields are low, control experiments show that N₂ is the source of these reduced products.

The complexes described here have an S₃C ligand environment, but the sulfur donors are thiolates rather than sulfides, and the carbon donor is an arene rather than a carbide. Ligand K-edge X-ray absorption studies indicate that iron-sulfide bonds are generally more covalent than iron-thiolate bonds.^76-77 The iron-carbide bond in the FeMoco is also thought to be highly covalent.^78-80 Since neither the electronic structure of the FeMoco nor the role of the carbide in catalysis are well-understood,^81-83 it is difficult to evaluate the similarity of our model to the FeMoco. There are also other important differences: for example, the model complexes presented here have a single iron atom, while the FeMoco is a multimetallic cluster. In addition, the terphenyl supporting ligand has π-backbonding capacity that is not present in the FeMoco. Furthermore, secondary sphere effects, which are not present in the model system, could play a role in the FeMoco for N₂ binding and activation or stabilize important intermediates through hydrogen bonding. Nevertheless, our results represent the first step towards the development of synthetic iron complexes containing only S and C donors capable of N₂ binding and functionalization.

Another relevant comparison is to the oxidation states in the FeMoco. Hoffman has argued that only two oxidation levels are accessed in the FeMoco, with additional reducing equivalents stored on hydrides.¹ Thus, the intermediates (like the resting state) likely have formally iron(II) and iron(III) ions,^81-83 whereas the complexes described here bind N₂ only upon reduction to iron(0). If the iron(0) oxidation level is feasible in the FeMoco, it would only be transiently available (perhaps from reductive elimination of H₂ from an iron site, which is formally a two-electron reduction of the metal). N₂ binding is relatively weak in the iron(0) complex 1-K₂, as shown by its loss of N₂ at room temperature. Both the lability of N₂ and the requirement for reduction to the iron(0) redox level in the synthetic system may be due to the competition between backbonding to N₂ and backbonding to the internal arene of the supporting ligand (supported by DFT calculations above), rather than any deficiency in the electron-richness of the iron complex in this ligand field. Scaffolds that present a biomimetic C donor without π-backbonding capability, for example using alkyl donors,^{35, 46} may be more effective for N₂ binding and reduction.

CONCLUSIONS

We have examined the mechanism of N₂ binding to the iron(II) tris(thiolate) complex 2-K. Reduction to the iron(0) level results in monothiolate dissociation, which produces a complex that is competent for N₂ binding. These results demonstrate the feasibility of dissociation of RS⁻ leading to a low-coordinate sulfur-carbon iron site that binds N₂, which has been proposed for the FeMoco. Further reduction of the N₂ complex to the formally iron(-I) redox level gives a species that reacts with acids to give NH₃ or N₂H₄ in low yield, showing that it is possible to bind and reduce N₂ at iron without incorporating abiological donor atoms. The challenges associated with this system include the presence of a π-accepting arene in the ligand backbone, which results in weak N₂ binding, and the high acid lability of the ligand, which results in decomposition under catalytic conditions for N₂ reduction. We are currently designing new multidentate sulfur/carbon ligands in an effort to create more suitable functional models for the FeMoco.

Figure 9.

Mössbauer spectrum of a frozen reaction mixture of 2-K treated with 3.6 equiv of KC₈ in THF under N₂. The black circles are the experimental data, and the grey line is the residual for the total fit. The green and blue lines are fits corresponding to 7-K₃ and 6-K₂, respectively, with the parameters shown in Table 1.