Nitrogen reduction by the Fe sites of synthetic [Mo3S4Fe] cubes

Nitrogen (N2) fixation by nature, which is a crucial process for the supply of bio-available forms of nitrogen, is performed by nitrogenase. This enzyme uses a unique transition-metal–sulfur–carbon cluster as its active-site co-factor ([(R-homocitrate)MoFe7S9C], FeMoco)1,2, and the sulfur-surrounded iron (Fe) atoms have been postulated to capture and reduce N2 (refs. 3–6). Although there are a few examples of synthetic counterparts of the FeMoco, metal–sulfur cluster, which have shown binding of N2 (refs. 7–9), the reduction of N2 by any synthetic metal–sulfur cluster or by the extracted form of FeMoco10 has remained elusive, despite nearly 50 years of research. Here we show that the Fe atoms in our synthetic [Mo3S4Fe] cubes11,12 can capture a N2 molecule and catalyse N2 silylation to form N(SiMe3)3 under treatment with excess sodium and trimethylsilyl chloride. These results exemplify the catalytic silylation of N2 by a synthetic metal–sulfur cluster and demonstrate the N2-reduction capability of Fe atoms in a sulfur-rich environment, which is reminiscent of the ability of FeMoco to bind and activate N2. Iron atoms in a synthetic metal–sulfur cluster can capture nitrogen and catalyse its silylation, demonstrating successful nitrogen reduction by iron atoms in a sulfur-rich environment.

Nitrogen (N 2 ) fixation by nature, which is a crucial process for the supply of bio-available forms of nitrogen, is performed by nitrogenase. This enzyme uses a unique transition-metal-sulfur-carbon cluster as its active-site co-factor ([(R-homocitrate)MoFe 7 S 9 C], FeMoco) 1,2 , and the sulfur-surrounded iron (Fe) atoms have been postulated to capture and reduce N 2 (refs. [3][4][5][6]. Although there are a few examples of synthetic counterparts of the FeMoco, metal-sulfur cluster, which have shown binding of N 2 (refs. [7][8][9] ), the reduction of N 2 by any synthetic metal-sulfur cluster or by the extracted form of FeMoco 10 has remained elusive, despite nearly 50 years of research. Here we show that the Fe atoms in our synthetic [Mo 3 S 4 Fe] cubes 11,12 can capture a N 2 molecule and catalyse N 2 silylation to form N(SiMe 3 ) 3 under treatment with excess sodium and trimethylsilyl chloride. These results exemplify the catalytic silylation of N 2 by a synthetic metal-sulfur cluster and demonstrate the N 2 -reduction capability of Fe atoms in a sulfur-rich environment, which is reminiscent of the ability of FeMoco to bind and activate N 2 .
Nitrogen (N 2 ) is an essential element to maintain every known form of life on Earth. Although the element is inexhaustible in the atmosphere as N 2 , only diazotrophs or lightning in thunderstorms can transform this stable molecule into bio-available forms (for example, ammonia (NH 3 ) and nitrogen oxides) in the natural world. Other organisms consequently rely on the products of the fixation processes and limited pre-existing sources to afford the necessary nitrogen. In this sense, N 2 fixation is one of the most crucial bottlenecks in Earth's ecosystem. Key players of N 2 fixation are nitrogenase enzymes that reduce N 2 to NH 3 . The most studied of these, molybdenum (Mo)-nitrogenase, has a unique metal-sulfur-carbon co-factor described as [(R-homocitrate) MoFe 7 S 9 C] (FeMoco; Fig. 1) 1,2 and performs the catalytic reduction at ambient temperature and pressure. As FeMoco is found only in nitrogenase, its chemical and physical properties have attracted significant interest regarding its desirable N 2 -reduction activity. How FeMoco reduces N 2 has long been studied by biochemical analyses of the enzyme 3,13 and by structural and functional modelling of FeMoco with small-molecule complexes 14,15 and metal-sulfur (M-S) clusters 16,17 .
Although the detailed N 2 -reduction mechanism of FeMoco remains elusive, a growing number of studies suggest that FeMoco eliminates one of the µ 2 -bridging sulfur (S) atoms and captures substrates at the produced vacant coordination sites on the iron (Fe) atoms 4-6 . Analogously, a common approach to generating small-molecule N 2 complexes has been removal of a metal-bound ligand under reducing conditions. However, applying this method to the available synthetic counterparts of FeMoco, namely, M-S clusters, has been challenging. As these clusters contain coordinative S atoms in their inorganic cores, a vacant metal site often attracts core S atoms of M-S clusters rather than N 2 , which leads to aggregation 17 . Limiting the number of vacant metal atoms in the core is thus a plausible approach and led to the isolation of N 2 -bound clusters in several previous and recent examples, including ours [7][8][9] . Nevertheless, catalytic reduction of N 2 by these clusters remains unknown despite its relevance to the natural system.
Our framework to overcome these hurdles implements a triangular [Mo 3 S 4 ] fragment bearing robust Mo-Cp R bonds (Cp R = C 5 Me 5 (Cp*), C 5 Me 4 SiMe 3 (Cp L ) and C 5 Me 4 SiEt 3 (Cp XL ), where Si is silicon, Me is methyl and Et is ethyl) 11,12 as a platform to structurally encumber and protect a fourth metal incorporated into the vertex (Fig. 1). A titanium (Ti) derivative [Cp* 3 Mo 3 S 4 Ti] captures and activates N 2 in the presence of potassium graphite (KC 8 ), indicating that the [Mo 3 S 4 Ti] cube is robust under reducing conditions and avoids undesirable aggregation 8 . In contrast, catalytic reduction of the bound N 2 was not observed, probably owing to the strong Ti-N bond that inhibits product release. We then hypothesized that a softer Fe atom as found in the biological systems might function more successfully to carry out N 2 reduction instead of the harder Ti atom. The results below demonstrate the capture and catalytic silylation of N 2 by the vertex Fe atoms of [Mo 3 S 4 Fe] cubes. Although it is likely that there are mechanistic differences between the silylation of N 2 and NH 3 production by nitrogenase, our results demonstrate that an Fe centre built into a M-S core and in a S-rich coordination environment can activate inert N 2 sufficiently for chemical conversion.
The treatment of our reported [Mo 3 S 4 Fe] clusters, [Cp R 3 Mo 3 S 4 FeCl] (Cp R = Cp* (1a), Cp R = Cp L (1b) and Cp R = Cp XL (1c)) 11,12 with strong reductants (KC 8 for 1a and sodium naphthalenide Na(C 10 H 8 ) for 1b and 1c) under N 2 in tetrahydrofuran (THF) led to the formation of the corresponding N 2 clusters (Fig. 2a) Table 5). The structure of 2a is an N 2 -bridged [Mo 3 S 4 Fe] dimer with an inversion centre at the middle of the two N atoms, whereas 2c is a monomeric cluster bearing a terminally bound N 2 . The N-N distances (1.151(4) Å for 2a and 1.136(5) Å for 2c) are in between those of free N 2 (1.098(1) Å) and N 2 H 2 (1.252 Å), suggesting a weakened N-N bond. Likewise, the N-N stretching frequencies of 2b and 2c are close to the lower end of those reported for N 2 complexes of Fe II or Fe I (ref. 18 ), and the activation levels are even comparable to an Fe 0 -N 2 complex supported by S-and carbon (C)-based ligands 19 .
Successful activation of N 2 at the Fe atoms of 2a-2c prompted us to pursue the catalytic reduction of N 2 using the [Mo 3 S 4 Fe] cubes. Reduction to NH 3 was observed but was not catalytic, giving at most 1.6 ± 0.1 equiv. NH 3 (per 2c) under typical conditions 14,15,20 (Supplementary Table 4). Protonation of S atoms possibly occurs in this reaction, which weakens the Fe-S bonds and releases a vertex Fe atom from [Mo 3 S 4 Fe], in a relevant manner to degradation of the cubes under electrochemical oxidation 11,12 . Nonetheless, more importantly, we found that the chloride (Cl) clusters 1a-1c and N 2 clusters 2b and 2c all catalytically reduce N 2 to N(SiMe 3 ) 3 in the presence of excess sodium (Na) and trimethylsilyl chloride (ClSiMe 3 ). A minimum of 122.9 ± 3.0 equiv. (per 1c) and up to 248.0 ± 15.6 equiv. (per 1b) of N(SiMe 3 ) 3 were generated after 100 h under a N 2 (1 atm) atmosphere at room temperature (Table 1 and Supplementary Tables 1 and 2). Although the cause of the differences in the activity of these complexes has not been conclusively identified at this point, we assume that the steric, not electronic, effects of the Cp R ligands play a major role (vide infra) because the Cp R substituents did not notably affect the redox behaviours of 1a-1c (half-wave potential E 1/2 ([Mo 3 S 4 Fe] 5+/4+ ) = -0.17 V (1a), -0.19 V (1b) and -0.24 V (1c) versus Ag/Ag + ) 12 or the redox features of 2b and 2c observed in cyclic voltammetry ( Supplementary Fig. 14).
The N(SiMe 3 ) 3 yields, on a per-active-metal basis produced by 1b and 2c, are about three-times higher than those of other Fe catalysts reported so far (Supplementary Table 3 [15][16][17]. These by-products should originate from reactions of the trimethylsilyl radical ·SiMe 3 with itself or the THF solvent 28 , as treatment of ClSiMe 3 with alkaline metals has been accepted to generate ·SiMe 3 (ref. 26 ). Although the THF solvent is much more abundant than the N 2 reactant, the selectivity for N(SiMe 3 ) 3 was high in the case of 1b and reached 37.2% (Table 1, (Table 1 and Supplementary Table 3). After a catalytic run using 1a, a mass spectrum of the reaction mixture revealed [Mo 3 S 4 Fe] cubes binding ring-opening products of THF ( Supplementary  Fig. 19), indicating sufficient stability of the cubic core during catalysis. Moreover Considering previous proposals for analogous reactions 26,28 , we propose a mechanism for N 2 silylation by our [Mo 3 S 4 Fe] cubes (Supplementary Fig. 18). In this pathway, we suppose that the Cl atom   (Fig. 2a,d and Supplementary Figs. 4, 9 and 13). The same cluster 3 was alternatively generated from 2c and ClSiPh 3 in C 6 D 6 (Supplementary Fig. 5). The X-ray structure of 3 shows an elongated N-N distance (1.193 (7) Table 6) 22,29,30 . Although the phenyl substituents on the Si atom of 3 differ from the methyl groups employed in the catalytic process, the isolation of 3 supports the possible generation of an Fe-NNSiMe 3 analogue of 2c in the catalytic cycle. This assumption was further reinforced when 3 was used as the precursor for the successful catalytic silylation of N 2 , yielding 258.3 ± 6.1 equiv. of N(SiMe 3 ) 3 (Table 1, entry 6).
The irreversible chemical modification of the bound N 2 molecule highlights a reactivity difference between our system and a N 2 -bridged [MoFe 3 S 4 ] dimer reported recently 9 . Treatment of [{Cp*MoFe 3 S 4 (IPr) 2 } 2 (µ-N 2 )] (IPr = 1,3-bis(2,6-diisopropylphenyl)-imidazol-2-ylidene) with [Cp L 2 Ti] has been reported to give an equilibrium mixture of the [MoFe 3 S 4 ] 2 (µ-N 2 ) dimer and a heterometallic N 2 -bridged complex [{Cp*MoFe 3 S 4 (IPr) 2 }(µ-N 2 )(Cp L 2 Ti)], but no further chemical conversions or catalytic reduction of the bound N 2 was reported. The difference in reactivity between this N 2 bridging system and the terminally bound N 2 system reported herein points to the importance of a terminal Fe-N 2 moiety for the successful reduction of N 2 . In addition, the isolation of a stable intermediate analogue by our [Mo 3 S 4 Fe] platform indicates its potential utility as a synthetic toolkit to investigate catalytic as well as stoichiometric activation of other small molecules. Structural models of the N 2 -bound clusters reveal that the Cp R ligands surround the N 2 ligand and the [Mo 3 S 4 Fe] cores (Fig. 3). The -SiMe 3 and -SiEt 3 groups of the Cp R ligands are forced into the space around the Fe sites of the [Mo 3 S 4 Fe] cubes (Fig. 3b,c) to minimize steric repulsion between the Cp R ligands. Thus, the -SiR 3 groups efficiently offer steric protection of the [Fe-N 2 ] moiety and prevent dimerization of cubes through either an Fe-N 2 -Fe bridge or an inter-cube Fe-S interaction. In contrast, the less bulky Cp* ligands lead to a more exposed Fe site (Fig. 3a) and allow the approach of the Fe site of another [Mo 3 S 4 Fe] cube to give an Fe-N 2 -Fe dimer. The bulkiness of the Cp R ligands should affect the catalytic activities as well, as we suggest that the first N-Si bond formation occurs at the distal N atom. In an Fe-N 2 -Fe dimer, both N atoms are protected until one of the Fe-N interactions breaks to generate monomers. In the catalytic reactions using Fe-Cl cubes, chloride abstraction (initiation) is expected to be slower with the bulkier Cp R ligands.
To better understand the properties of 2a-2c and 3, zero-field 57 Fe Mössbauer spectra were measured at 78 K using powdered crystals ( Supplementary Fig. 10). The spectra were fitted as single quadrupole doublets with the following values of the isomer shift (δ) and the quad- The δ values of 2a-2c are lower than those of the precursors 1a-1c featuring Fe II centres (δ = 0.555(2)-0.563(3) mm s −1 ), even though 2a-2c are supposedly more reduced. A reduced Fe centre typically shows a higher δ than an oxidized Fe in similar coordination environments, owing to shielding of electrons in s orbitals by the increased 3d electron densities. This has also been verified for [Fe 4 S 4 ] and [Fe 3 S 4 ] clusters bearing thiolate ligands 31 . The opposite trend observed here could be ascribed to π back-donation from Fe to the N 2 ligand. As suggested from theoretical calculations, such increased covalency of the Fe-ligand bond leads to high electron density at the Fe nucleus and, consequently, a low δ value 32 . Thus, although determination of the Fe oxidation state is not straightforward, we tentatively assign the Fe centres in 2a-2c as close to Fe II but only slightly reduced. Kohn-Sham frontier orbitals of optimized 2a and 2c reveal major contributions of Mo atoms in the occupied orbitals to store reducing equivalents ( Supplementary  Figs. 21-23). The density-functional-theory-calculated δ values of 2a, 2c and 3 based on the crystal and optimized structures are qualitatively in agreement with the experimental data (Supplementary Tables 7-10), supporting the utility of orbital analysis.
The covalent nature of Fe-N interaction can also rationalize the δ value of 3. As illustrated by the N-N bond distance and N-N stretching frequency of the cluster, silylation of the N 2 ligand has led to stronger back-donation from Fe and a shorter Fe-N distance (1.687(5) Å) than those of 2a and 2c. This is consistent with a highly covalent Fe-N interaction and does not contradict the decrease of the δ value in 3 compared with 2a-2c, implying a major contribution of a resonance structure of Fe=N=N-SiPh 3  2 )(C 6 H 4 ) 3 ] 3-) 33 , and regression lines from each series have nearly the same slope.
Overall, the above results represent the first catalytic silylation of N 2 by a synthetic M-S cluster and suggest some key features for successful N 2 reduction. Sufficient stability of the M-S core under reducing conditions can make an Fe centre sufficiently electron-rich to activate N 2 . These results imply that suppression of intermolecular aggregation of M-S cores is a key to stabilizing a terminal N 2 -bound species and maintaining the reactivity at the distal N atom of the bound N 2 molecule. In this sense, the steric protection given by the Cp R ligands loosely mirrors the role of a protein matrix isolating metal centres to control reactivity and avoid undesirable decomposition. Although N 2 reduction to NH 3 or to N(SiMe 3 ) 3 should have substantial mechanistic differences, our results provide compelling clues that the N 2 molecule can become susceptible to reduction by Fe centres in S-rich environments, as is the case with FeMoco.

Online content
Any methods, additional references, Nature Research reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41586-022-04848-1. Panels a and c were prepared from the crystal structures of 2a and 2c; panel b was prepared via software using the Cl-bound cluster 1b.