N₂ activation on a molybdenum–titanium–sulfur cluster

Abstract

The FeMo-cofactor of nitrogenase, a metal–sulfur cluster that contains eight transition metals, promotes the conversion of dinitrogen into ammonia when stored in the protein. Although various metal–sulfur clusters have been synthesized over the past decades, their use in the activation of N₂ has remained challenging, and even the FeMo-cofactor extracted from nitrogenase is not able to reduce N₂. Herein, we report the activation of N₂ by a metal–sulfur cluster that contains molybdenum and titanium. An N₂ moiety bridging two [Mo₃S₄Ti] cubes is converted into NH₃ and N₂H₄ upon treatment with Brønsted acids in the presence of a reducing agent.

Nitrogenase—whose cofactor consists of a metal–sulfur cluster—catalyzes the production of NH₃ from N₂, but designing metal–sulfur complexes capable of promoting this conversion remains challenging. Here, the authors report on the activation of N₂ by a metal–sulfur cluster containing [Mo₃S₄Ti] cubes, demonstrating NH₃ and N₂H₄ production.

Introduction

Metal–sulfur clusters are a ubiquitous class of inorganic compounds, and span a wide range of structural and electronic features^1,2. One unique function of protein-supported clusters is the conversion of N₂ into NH₃, which is mediated by the FeMo-cofactor of nitrogenase (Fig. 1a)^3–5. The FeMo-cofactor can be extracted from the protein into organic solvents without significant degradation of its metal–sulfur core^6,7. The extracted form catalyzes the reduction of non-native C₁ substrates, such as CO and CO₂ in the presence of reducing agents and protons to furnish CH₄ and short-chain hydrocarbons^8,9, and similar catalytic reductions can also be accomplished with the synthetic analog [Cp*MoFe₅S₉(SH)]^3– (Cp* = η⁵-C₅Me₅)¹⁰. However, N₂ cannot be reduced with the extracted FeMo-cofactor⁷ or its synthetic analogs, highlighting the difference in N₂-reducing activity between the protein-supported FeMo-cofactor and discrete metal–sulfur clusters in solution. Thus, a molecular basis for the reduction of N₂ on metal–sulfur clusters remains elusive, while some aqueous suspensions or emulsions of Fe–S or Mo–Fe–S compounds have been found to generate NH₃^11–14 and some thiolate-supported Fe complexes have been found to activate N₂^15–17. The cuboidal cluster Cp*₃Ir₃S₄Ru(tmeda)(N₂) (tmeda = tetramethylethylenediamine) is the only structurally characterized example for the binding of N₂ to a metal–sulfur cluster, although the conversion of the Ru-bound N₂ has not yet been achieved¹⁸.

Fig. 1.

Structure of the nitrogenase FeMo-cofactor. a The resting-state structure. The peptide chains of cysteine (Cys) and histidine (His), as well as a part of (R)-homocitrate are omitted for clarity. Color legend: C, gray; Fe, black; Mo, dark blue; S, yellow; O, red; N, light blue. b Wire-frame drawing of the inorganic part of the FeMo-cofactor, which highlights the fused form of the two cubes

It should be noted that the structure of the FeMo-cofactor can be seen as a fused form of two cubes (Fig. 1b), implying the utility of cubes in the activation of N₂. However, synthetic examples are limited for the fused form of such metal–sulfur cubes^19–23, and a carbon-centered derivative still remains unprecedented. In order to develop a strategy for the activation of N₂ on synthetic metal–sulfur clusters, we examined independent cubes in this study. Our design (Fig. 2) is based on the trinuclear Mo–sulfur cluster platform Cp*₃Mo₃(μ-S)₃(μ₃-S) (1)²⁴ for the accommodation of a metal atom via three doubly bridging sulfur atoms to furnish a cubic structure. Ti was chosen as the additional metal to demonstrate the proof-of-concept, on the basis of successful examples of Ti complexes in N₂ chemistry²⁵, e.g. some recent studies on the activation of N₂ mediated by Ti–hydride complexes^26,27. The resulting [Mo₃S₄Ti] cluster (Fig. 2, top right) is reduced in the presence of N₂, to generate a Ti-based reaction site for the activation of N₂. Herein, we report the activation of N₂, as well as its conversion into NH₃ and N₂H₄, on a Mo–Ti–S cluster.

Fig. 2.

Reactions in this study. Synthesis of the cubic [Mo₃S₄Ti] cluster 2, the N₂-binding cluster 3, and its 1e reduced form [3]^–; Cp* = η⁵-pentamethylcyclopentadienyl; THF tetrahydrofuran

Results and Discussion

Synthesis of a cubic [Mo₃S₄Ti] cluster and an N₂-cluster

The cubic [Mo₃S₄Ti] precursor cluster Cp*₃Mo₃S₄TiCl₂ (2) was synthesized in 70% yield in tetrahydrofuran (THF) from the reaction of in situ generated [1]^– with TiCl₃(THF)₃ (Fig. 2), in a manner similar to the synthesis of cubic clusters from trinuclear Mo–sulfur clusters^28,29. The molecular structure of 2 (Supplementary Figure 1 and Supplementary Table 3) reveals a nearly square-pyramidal geometry of titanium with an apical sulfur atom, where the Ti–S_apical distance (2.2799(7) Å) is shorter than the Ti–S_basal distances (2.3888(6) and 2.4272(6) Å). The Ti–Cl distances (2.3429(7) and 2.3812(7) Å) are close to those in a Ti(III) complex (average 2.353(4) Å) supported by three chlorides and a tridentate phosphine ligand³⁰. The reduction of 2 with potassium graphite (KC₈) in THF under an atmosphere of N₂ resulted in the coordination and reduction of N₂ to form the brown cluster [K(THF)₂]₂[Cp*₃Mo₃S₄Ti]₂(μ-N₂) (3) in 60% yield, in which an N₂ moiety bridges two [Mo₃S₄Ti] cubes via binding to the Ti atoms. Product(s) of analogous reactions under atmospheres of H₂ or Ar have not been characterized so far. Cluster 3 is diamagnetic, and the ¹⁵N NMR spectrum of ¹⁵N-labeled 3 exhibited a signal at –75.4 ppm relative to CH₃NO₂ in THF-d₈ (Supplementary Figure 5). A single-crystal X-ray diffraction analysis of 3 established that N₂ is linearly aligned with the two Ti atoms to form a Ti–N=N–Ti moiety (Fig. 3 and Supplementary Figure 4). Its N–N distance (1.294(7) Å) is comparable to the longest bonds reported for Ti–N=N–Ti complexes^31–33, suggesting a high degree of N₂ reduction. The N–N distance in 3 is slightly longer than that in H₃CN=NCH₃ (1.25 Å)³⁴, but shorter than that in H₂N–NH₂ (1.46 Å)²⁵, indicating a character between N=N double and N–N single bonds for the bridging N₂ in 3. This notion was supported by resonance-Raman spectroscopy (Fig. 4 and Supplementary Figure 8), which showed a ν_NN band of 3 at 1240 cm⁻¹ based on the difference spectrum of 3 and ¹⁵N-labeled 3. This ν_NN frequency falls between those of H₃CN=NCH₃ (1575 cm⁻¹)³⁵ and H₂N–NH₂ (1111 cm⁻¹)³⁶. The observed shift (40 cm⁻¹) between 3 (1240 cm⁻¹) and ¹⁵N-labeled 3 (1200 cm⁻¹) is consistent with the calculated shift (42 cm⁻¹) for the replacement of ¹⁴N₂ by ¹⁵N₂ assuming  a diatomic harmonic oscillator.

Fig. 3.

Molecular structure of the N₂ cluster 3. For clarity, carbon and oxygen atoms are drawn as capped sticks, while other atoms are drawn with thermal ellipsoids set at 50% probability. Color legend: C, gray; Mo, dark blue; Ti, purple; S, yellow; O, red; N, light blue; K, green. Selected bond distances (Å): N1–N2 1.294(7), Mo1–Mo1* 2.6766(8), Mo1–Mo2 2.8853(6), Mo3–Mo3* 2.7392(8), Mo3–Mo4 2.8518(6), Mo1–Ti1 3.0315(10), Mo2–Ti1 3.0908(12), Mo3–Ti2 3.0301(10), Mo4–Ti2 3.0778(12), Ti1–N1 1.802(5), Ti2–N2 1.798(5), K–N1 2.797(3), K–N2 2.782(3)

Fig. 4.

Spectroscopic characterization of the Ti-N=N-Ti moiety. Resonance-Raman spectra for 3 (a, green line) and ¹⁵N-labeled 3 (b, orange line), as well as their difference spectrum (c, dark blue line) in THF at –30 °C

Owing to the robust Cp*–Mo and Mo–S bonds in the present system, we speculate that the three Mo atoms remain intact during the activation of N₂, leaving Ti as the reaction site in each cube. These robust bonds would also help to avoid the degradation of the cubic structure by aggregation or fragmentation.

Conversion of the N₂ moiety into NH₃ and N₂H₄

The high degree of reduction in 3 prompted us to attempt a stoichiometric conversion of the coordinated N₂ into NH₃ and N₂H₄. Upon treatment of 3 with protons from either H₂O or HCl, the color changed from brown to green, and NH₃, N₂H₄, and [1]⁺ were detected. The resulting products NH₃ and N₂H₄ were quantified by indophenol and azo-dye titration methods, respectively (Supplementary Tables 1 and 2)^37,38. As summarized in Table 1 (entries 1 and 4), small to moderate amounts of NH₃ (0.12(3)–0.29(5) equiv) and N₂H₄ (0.12(10)–0.60(3) equiv) were detected by protonation of 3 with an excess H₂O or HCl. While the length of the N–N bond in 3 indicates that a reduction has occurred, additional electrons are required for the cleavage of the N–N bond to form NH₃. Therefore 3 was treated with KC₈ and H₂O or HCl. In the presence of 6 equiv of KC₈, the average amount of NH₃ generated by protonation with H₂O increased from 0.12(3) to 0.47(27) equiv (entry 5), while the fluctuations observed allow only a cautious comparison. The NH₃ yields in entries 4 and 5 (HCl) are comparable and within error. The yields of N₂H₄ in entries 2 (0.56(10) equiv from H₂O) and 5 (0.19(6) equiv from HCl) remained comparable to those of reactions in the absence of KC₈. The formation of N₂H₄ from the Ti–N=N–Ti moiety is reasonable, as this N–N bond exhibits an N=N double and N–N single bond character. In the presence of 0–6 equiv of KC₈, subsequent protonation by H₂O provided approximately three times the amount of N₂H₄ than the reactions with HCl. Such acid-dependent selectivity may arise from the affinities of the conjugate bases (OH^– or Cl^–) toward metals, where the high oxophilicity of the Ti and K atoms can facilitate the coordination of H₂O onto these atoms, which could lead to an efficient protonation of the neighboring Ti–N=N–Ti moiety.

Table 1.

Protonation of 3 in the presence/absence of KC₈^a

Entry	Proton source	KC8 (equiv)	NH3 (equiv)	N2H4 (equiv)
1	H2O	0	0.12(3)	0.60(3)
2	H2O	6	0.47(27)	0.56(10)
3	H2O	100	1.20(19)	n.d.b
4	HCl	0	0.29(5)	0.12(10)
5	HCl	6	0.43(15)	0.19(6)
6	HCl	100	0.20(8)	n.d.b

^a Yields represent the average of three runs. Standard deviations are given in parentheses

^b n.d. = not detected

Even in the presence of a slight excess (6 equiv) of KC₈, conversion of the N₂ moiety of 3 into NH₃ was not complete and notable amounts of N₂H₄ were detected (entries 2 and 5). These results indicate that some of the added KC₈ remains unreacted with the N₂ cluster and is subsequently quenched directly by H₂O or HCl. However, the formation of N₂H₄ was suppressed when a large excess of KC₈ was added prior to protonation (entries 3 and 6). While the protonation with HCl did not increase the amount of NH₃ (entry 6), the maximum yield of NH₃ (1.20(19) equiv) was achieved when 100 equiv KC₈ were added prior to the protonation with H₂O (entry 3). When conditions identical to those of entry 3 were applied for the ¹⁵N-enriched 3 under an atmosphere of ¹⁴N₂, the ¹H NMR spectrum of the resulting ammonium salt indicated the exclusive formation of ¹⁵NH₃, which originates from the Ti–¹⁵N=¹⁵N–Ti moiety (Supplementary Figure 15). After addition of an excess of protons to 3, the Ti atom probably dissociates from the [Mo₃S₄Ti] cube, which is supported by the appearance of an intense signal of [1]⁺ in the electro-spray ionization mass (ESI-MS) spectrum of the resulting mixture (Supplementary Figure 16) and the color change of the mixture from dark brown to green. Thus, at this stage, the possibility of N₂ functionalization after dissociation of the Ti atom cannot be ruled out unequivocally.

It has been reported that metal–sulfur clusters exhibit stability across various oxidation states, which allows the storage and transportation of electrons^39,40. It is therefore worthy of notice that the 1e-reduced cluster [3]^– was obtained in trace amounts from the reaction of 2 with 6–10 equiv of KC₈ under an atmosphere of N₂ (Fig. 2, bottom). The characterization of [3]^– was carried out only by ¹H NMR and crystallographic analysis due to the trace amounts available. Cluster [3]^– is paramagnetic and its ¹H NMR spectrum in THF-d₈ exhibited a broad Cp* signal at 5.84 ppm (Supplementary Figure 10). The molecular structure of [3]^– (Supplementary Figure 9) revealed an interaction of three potassium ions with the bridging N₂ moiety, where the N–N distance (1.293(5) Å) is comparable to that in 3 (1.294(7) Å). A structural comparison between the [Mo₃S₄Ti] cubes of 3 and [3]^– reveals the elongation of distances in [3]^–. In detail, a relatively large contribution of the metal atoms for the storage of the additional electron is indicated by the larger differences in the average distances of Mo–Ti (3.0487(12) Å for 3 and 3.0242(8) Å for [3]^–) and Mo–Mo (2.8150(8) Å for 3 and 2.8668(6) Å for [3]^–), rather than the average metal–sulfur distances, i.e. Mo–S (2.3952(16) Å for 3 and 2.3966(11) Å for [3]^–), and Ti–S (2.328(2) Å for 3 and 2.3297(13) Å for [3]^–). The detailed reaction pathways for the formation of NH₃ and N₂H₄ from 3 or [3]^– remain unclear at this point, as the dissociation of the Ti atom from the [Mo₃S₄Ti] cube renders the identification of possible cluster intermediates difficult.

In summary, we successfully achieved the activation of N₂ and its conversion into sub-stoichiometric mixtures of NH₃ and N₂H₄ on synthetic Mo–Ti–S cubes by selectively generating the Ti-based reaction sites. This work thus represents a step towards the assembly of structural and functional models for the FeMo-cofactor.

Methods

General procedure

All reactions and manipulations were carried out under an atmosphere of nitrogen using standard Schlenk line or glove box techniques. Solvents for reactions were purified by passing over columns of activated alumina and a supported copper catalyst. TiCl₃(THF)₃⁴¹, [Cp*₃Mo₃S₄][PF₆] ([1]⁺)⁴², and KC₈⁴³ were prepared according to literature procedures. All other reagents were purchased from common commercial sources and used without further purification. General considerations on the measurements and experiments as well as some experimental details are described in the Supplementary Methods.

Modified synthesis of Cp*₃Mo₃S₄ (1)

Compound 1 was originally synthesized from the reaction of Cp*Mo(S^tBu)₃ (^tBu = tert-butyl) with Na/Hg²⁴, but this reaction is not ideally suitable for a gram-scale synthesis. Instead, cluster 1 was synthesized on a multi-gram scale through the reduction of [1]⁺. For that purpose, KC₈ (1.36 g, 10.06 mmol) was added in portions to a THF (150 mL) suspension of [1]⁺ (9.78 g, 10.12 mmol) at room temperature. After stirring this mixture overnight, the solvent was removed under reduced pressure. The residue was extracted four times with toluene (350 mL). The extract was stored at –30 °C to precipitate green blocks of 1. The mother liquor was concentrated to ca. 1/5 of the original volume and stored at –30 °C to obtain a second crop. The total amount of green crystals of 1 was 6.17 g (7.51 mmol, 83%). This product was characterized by comparison with ¹H NMR data in the literature²⁴. ¹H NMR (C₆D₆): δ 8.56 (w_1/2 = 14 Hz, Cp*).

Synthesis of Cp*₃Mo₃S₄TiCl₂ (2)

A THF (20 mL) solution of 1 (1.00 g, 1.22 mmol) was cooled to –100 °C, and a THF solution of sodium naphthalenide (19.5 mM, 62 mL, 1.21 mmol) was added. The reaction mixture was gradually warmed to room temperature to give a dark green solution. The mixture was again cooled to –100 °C, and TiCl₃(THF)₃ (451 mg, 1.22 mmol) was added. The reaction mixture was warmed to room temperature and stirred for 5 days, affording a dark brown suspension. After concentration to ca. 5 mL under reduced pressure, the mixture was cooled to –30 °C, which caused the precipitation of a dark brown powder. The supernatant was removed and the dark brown powder was washed with hexane (20 mL × 2). The brown powder was dissolved into CH₂Cl₂ (15 + 3 mL), and hexane (100 mL) was layered on top of this solution. The subsequent slow diffusion led to the formation of brown plates of Cp*₃Mo₃S₄TiCl₂·CH₂Cl₂ (2·CH₂Cl₂). These  brown crystals were washed with hexane (15 mL × 2), and then dissolved in THF (15 mL). After stirring overnight, THF and CH₂Cl₂ were removed under reduced pressure to give a brown powder of 2 (801 mg, 0.854 mmol, 70%). ¹H NMR (CD₂Cl₂): δ 9.23 (w_1/2 = 8 Hz, Cp*); solution magnetic moment (CD₂Cl₂, 297 K): μ_eff = 1.3 μ_B; Cyclic voltammogram (2 mM in CH₂Cl₂, potential vs. (C₅H₅)₂Fe/[(C₅H₅)₂Fe]⁺): E_1/2 = –0.28, –1.34 V; UV/Vis (CH₂Cl₂): λ_max 415 nm (sh, 1200 cm⁻¹ M⁻¹), 753 nm (660 cm⁻¹ M⁻¹); analysis (calcd., found for C₃₀H₄₅Cl₂Mo₃S₄Ti): C (38.31, 38.77), H (4.82, 4.70), S (13.64, 13.24).

Synthesis of [K(THF)₂]₂[Cp*₃Mo₃S₄Ti]₂(μ-N₂) (3)

Cluster 2 (200 mg, 0.213 mmol) and KC₈ (116 mg, 0.858 mmol) were charged into one side of an H-shaped glass vessel (Supplementary Fig. 3). The other arm of the vessel was charged with THF (10 mL), which was vacuum-transferred to the solid mixture at 78 K. The vessel was filled with N₂ (1 atm) and then sealed. The mixture was slowly warmed using a methanol/liquid N₂ slurry. Stirring of the mixture with a magnetic stirring bar was initiated, once part of frozen THF had melted. The mixture was gradually warmed to room temperature, giving a dark red-brown suspension. Stirring was continued overnight, before the solvent was removed under reduced pressure and the residue was extracted with THF (50 mL). The solution was filtered, concentrated to ca. 3 mL, and stored at –30 °C. Cluster 3 (136 mg, 0.064 mmol, 60%) was obtained as dark-brown crystalline blocks, which were washed with pentane. ¹H NMR (THF-d₈): δ 1.70 (Cp*); ¹³C{¹H} NMR (THF-d₈): δ 13.6 (C₅Me₅), 96.3 (C₅Me₅); UV/Vis (THF): λ_max 356 nm (35800 cm⁻¹ M⁻¹), 500 nm (15,800 cm⁻¹ M⁻¹); Resonance-Raman spectrum (0.07 mM in THF, λ_ex 355 nm, –30 °C): 1240 cm^-1 (ν_NN); analysis (calcd., found for C₇₆H₁₂₂K₂Mo₆N₂O₄S₈Ti₂): C (42.78, 42.36), H (5.76, 5.80), N (1.31, 1.37), S (12.02, 12.27).

Synthesis of ¹⁵N₂-labeled 3

This cluster was synthesized in a manner similar to that of 3, using cluster 2 (173 mg, 0.184 mmol), KC₈ (100 mg, 0.740 mmol), and ¹⁵N₂ (1 atm). Crystals of ¹⁵N₂-labeled 3 were obtained in 60% yield (117 mg, 0.055 mmol). ¹⁵N{¹H} NMR (THF-d₈, referenced to external CH₃NO₂): δ –75.4; Resonance-Raman spectrum (THF, –30 °C): 1200 cm⁻¹ (ν_NN).

Data availability

The X-ray crystallographic coordinates for structures reported in this study have been deposited at the Cambridge Crystallographic Data Centre (CCDC), under deposition numbers 1577330–1577332. These data can be obtained free of charge from The Cambridge Crystallographic Data Centre via www.ccdc.cam.ac.uk/data_request/cif. All other data are available from the authors upon request.

Author contributions

Y. O. conceived and designed the experiments; K. U. performed the experiments; T. O. and T. O. carried out the resonance Raman spectroscopic analysis; R. E. C. performed crystallographic analysis; M. T. participated in the discussion; Y. O. prepared the manuscript with input from all authors.

The authors declare no competing interests.

Electronic supplementary material

Supplementary Information accompanies this paper at 10.1038/s41467-018-05630-6.