Main

The well-defined catalytic conversion of dinitrogen (N2) to ammonia (NH3)—promoted by a molecular Mo–triamidoamine complex—was reported by Schrock in 2003 (ref. 1). The proposed mechanism involved the reduction of terminal end-on bound N2 by the stepwise, alternating addition of six protons and six electrons. Since that seminal report, a broad range of mononuclear and dinuclear Fe and Mo complexes2,3,4 capable of binding and catalytically converting N2 into NH3 have been identified5,6, with some reaching very high conversion efficiencies7. Since the discovery of the Fe–Mo cofactor, numerous mechanistic studies, mostly focused on Mo and Fe due to their biological relevance, have been carried out on metal–N2 complexes as models for the biological reduction of N2 (refs. 6,8,9,10). Although synthetic model chemistry—even based on Fe and Mo—does not provide direct mechanistic information on enzymatic species, such studies can provide relevant models and permit unforeseen reactivity patterns to be recognized. The catalytic conversion of N2 to NH3 has subsequently expanded beyond Mo and Fe, but involves a still exceedingly small number of metals (Ti, Co, V, Ru and Os)11,12,13,14.

The plethora of reported mechanistic studies on molecular catalytic N2 to NH3 conversion involve the reduction and protonation of end-on terminal or end-on/end-on bridging metal–N2 complexes5,6,8. However, the recent identification of intermediates of nitrogenase reactive sites where substrates such as CO or imide bridge—with only one atom—two neighbouring Fe centres highlights the need for more studies on dinuclear N2-bridged complexes15,16,17. Furthermore, the reduction and hydrogenation of side-on bound systems is also potentially relevant to gaining a better understanding of the mechanism of the Haber–Bosch process for converting N2 and hydrogen gas (H2) into NH3 (refs. 18,19,20). Although the catalytic N2 to NH3 conversion for a side-on bound N2 complex remains elusive, stoichiometric N2 to NH3 conversion has been reported for a few side-on bridged N2 complexes of group 4 metals21,22 and uranium23,24,25,26,27,28. Side-on coordinated N2 can undergo stepwise reduction to a variety of formal charge states including N22−, N23− and N24−, but in most cases only one of these activated N2 species is isolated for a given metal complex29. In f-element chemistry where side-on N2 complexes dominate, interconversion between N22− and the N23− radical is documented26,30,31,32,33, but further reduction could not be observed for lanthanide ions. Nonetheless, cleavage of side-on bound N24− to 2N3− following the addition of external reducing agents has also been reported by some of us for U-complexes23,26,34, which suggested to us that under suitable conditions, side-on coordinated bridging U–N2 complexes35,36,37,38 might prove to be competent for catalytic N2 to NH3 conversion. It is also germane to recall that before the use of Fe-based catalysts in the industrial Haber–Bosch process, Haber reported that uranium and uranium nitride materials (UNx) were effective heterogeneous catalysts for the production of ammonia from dinitrogen39. Accordingly, studies on molecular analogues may afford insights into heterogeneous scenarios.

Given our earlier collective work in early metal and f-block N2-activation23,25,26,27,33,40,41,42,43 and U-nitride chemistries26,34,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65, we developed our investigations using the diuranium–N2 complex [{UIV(TrenDMBS)}2(μ-η2:η2-N2)] (1, TrenDMBS = {N(CH2CH2NSiMe2tBu)3}3−), reported by Scott35 in 1998 (Fig. 1), as the formation of 1 from its trivalent precursor [UIII(TrenDMBS)] is reversible, suggesting that it could be amenable to catalysis rather than being an irreversible thermodynamic sink.

Fig. 1: Selected literature examples (1, A–C) of N2 binding and reduction by uranium complexes, as well as a summary of this work.
Fig. 1: Selected literature examples (1, A–C) of N2 binding and reduction by uranium complexes, as well as a summary of this work.
Full size image

Previously reported N22−-bridged complexes: [{UIV(TrenDMBS)}2(μ-η2:η2-N2)] (1, TrenDMBS = {N(CH2CH2NSiMe2tBu)3}3−)35; N23− radical-bridged aryloxide complex [K(2.2.2-crypt)][{(ArO)3UIV}2(μ-η2:η2-N2)] (A, OAr = 2,4,6-tri-tert-butylphenoxide)26; N24–-bridged siloxide complex [K2{[UV(OSi(OtBu)3)3]2(μ-O)(μ-η2:η2-N2)}] (B)25; bis-nitride bridged calix[4]tetrapyrrole complex [{K(DME)(Et8-calix[4]tetrapyrrole)U}2(μ-NK)2][K(DME)4] (C, DME = 1,2-dimethoxyethane)78; this work: stepwise reduction of 1 and catalytic conversion of N2 to ammonia by 1. xs., excess.

In this Article we report the stepwise reduction of N2—from free-N2 to bridging side-on bound forms to bridging nitrides—overall uniquely accessing four different states of side-on bound N2 for the same dinuclear U-complex. Experimental characterization and quantum-chemical studies suggest that the addition of reducing agents leads to the formal sequential reduction of the bound N2 rather than the U-centres. We establish molecular U-mediated catalytic N2 to NH3 conversion, moreover involving side-on bound coordination of N2 in metal-mediated N2 to NH3 catalysis. The stoichiometric reactions suggest the relevance of N2, N22−, N23−, N24− and N3− in the catalytic conversion of N2 to NH3 when involving side-on bridging N2.

Results and discussion

Synthesis and characterization

Complex 1 (Fig. 2), was prepared from hexane in 71% yield through a modified procedure, circumventing the sublimation step of the previously reported synthesis35.

Fig. 2: Synthesis of uranium–dinitrogen complexes 1–5.
Fig. 2: Synthesis of uranium–dinitrogen complexes 1–5.
Full size image

Synthesis of the previously reported pernitride-bridged complex [{UIV(TrenDMBS)}2(μ-η2:η2-N2)], 1, from [UIVTrenDMBS)Cl] via a modified procedure, followed by reduction of 1 to yield the N23− radical-bridged complexes [K(2.2.2-cryptand)][{UIV(TrenDMBS)}2(μ-η2:η2-N2)] (2-crypt), [Li{UIV(TrenDMBS)}2(μ-η2:η2-N2)] (2-Li) and [K{UIV(TrenDMBS)}2(μ-η2:η2-N2)] (2-K). Further reduction yields the hydrazido(4–)-bridged complex [Li2{UIV(TrenDMBS)}2(μ-η2:η2-N2)] (3) and ultimately leads to splitting of the N2 molecule, affording the bis-nitride complex [Li4(OEt2){UIV(TrenDMBS)}2(μ-N)2] (4). Cleavage of the hydrazido moiety is observed upon reaction of 3 with dihydrogen, and allows the isolation of the nitrido/imido-bridged complex [Li3{UIV(TrenDMBS)}2(μ-N)(μ-NH)] (5).

The redox reactivity of 1 was then explored (Fig. 2). After sequential addition of 1 equiv. of 2.2.2-cryptand and then KC8 to a brown-red solution of 1 in toluene, the mixture turned dark-brown, and full consumption of the starting material occurred after 3 h of stirring at room temperature (r.t.). Brown-black crystals of the mono-reduced complex [K(2.2.2-cryptand)][{UIV(TrenDMBS)}2(μ-η2:η2-N2)] (2-crypt) were isolated in 83% yield from the filtered reaction mixture when stored at −40 °C for 2 days (Fig. 3a and Supplementary Fig. 2). The N1–N2 distance of 1.336(6) Å in 2-crypt is significantly longer compared to that found in 1 (1.109(7) Å), the latter having an N=N distance close to that of the free N2 (1.0975 Å) (Supplementary Table 1). The longer N–N distance in 2-crypt indicates reduction at the central N2 rather than at the uranium ions, and the U1–N2–U2 unit is bent (fold angle of 157.43°, Fig. 3g) compared to the essentially planar angle of 177.83° in 1. Elongation of the N1–N2 distance in 2-crypt is associated with population of its antibonding π* orbital and formation of the N23− radical. Similar values of the N–N distance were found for the only other N23− radical-bridged diuranium system [K(L)][{(ArO)3UIV}2(μ-η2:η2-N2)] (for L = 2.2.2-cryptand, 1.425(5) Å; for L = (THF)6, 1.39(1) Å, THF = tetrahydrofuran; ArO = 2,4,6-tBu-C6H2O, Fig. 1, A)26, as well as for the N23− radical-bridged lanthanide complexes, which show N=N distances in the range of 1.396(7)–1.405(3) Å (refs. 66,67). The U–Namide distances for 2-crypt (average 2.238 Å) are close to those of [UIV(TrenDMBS)Cl] (average 2.22 Å)68, and the U–Namine distance (average 2.743 Å) is longer than in [UIV(TrenDMBS)Cl] (2.656(19) Å). In contrast to 1, 1H NMR studies showed that 2-crypt is stable towards N2 loss under dynamic vacuum in the solid state and after three freeze–pump–thaw cycles in toluene-d8 solution, and no N2 displacement was evidenced by exposure to coordinating solvents such as tetrahydrofuran and diethyl ether.

Fig. 3: Single-crystal X-ray diffraction solid-state structures of the dinitrogen-bridged TrenDMBS diuranium complexes with displacement ellipsoids at 50% probability level and selective labelling.
Fig. 3: Single-crystal X-ray diffraction solid-state structures of the dinitrogen-bridged TrenDMBS diuranium complexes with displacement ellipsoids at 50% probability level and selective labelling.
Full size image

a–f, Molecular structures of the anionic complex [{UIV(TrenDMBS)}2(μ-η2:η2-N2)]− in 2-crypt (a), as well as in 2-Li (b) and 2-K (c), the hydrazido-bridged complex 3 (d) with disorder in the dinitrogen core and with one lithium position removed for clarity, the bis-nitride complex 4 (e) and the nitrido/imido complex 5 (f). For all complexes, the triamidoamine ligand framework is depicted as a wireframe, and hydrogen atoms (apart from the imido-bound H in 5) and disordered tert-butyl groups have been removed for clarity. g, Molecular U2N2 core structures and XRD-derived dinitrogen distance summaries for complexes 2–5. h, Correlation between the degree of N–N bond activation and the NH4Cl yields obtained after direct quenching with 2 M HCl in diethyl ether.

Reduction of 1 was also performed without chelating agents. Dark-brown microcrystals of [K{UIV(TrenDMBS)}2(μ-η2:η2-N2)] (2-K) were obtained in 62% yield from the filtered reaction mixture left standing at −40 °C for 2 days. When the reduction was performed under similar conditions but with adventitious Li cations present, the complex [Li{UIV(TrenDMBS)}2(μ-η2:η2-N2)] (2-Li) was isolated after 3 days of storage at −40 °C. The N1–N2 distance (1.28(1) Å) in complex 2-Li is shorter than the N–N distances in 2-K (1.364(18) Å) and in 2-crypt (1.336(6) Å), suggesting stronger N–N activation in the K- and K-crypt complexes.

Magnetic susceptibility data were collected for 1 and 2-crypt under an applied field of 0.1 T and in a temperature range of 2–300 K. The susceptibility data for 1 indicate the presence of two noninteracting 5f2 U(IV) centres, with a magnetic moment of 0.8 μB at 2.5 K, 4.2 μB at 300 K (0.6 μB at 2.5 K and 3.0 μB at 300 K per U ion), in agreement with the assignment of 1 as a pernitride-bridged U(IV)/U(IV) complex69 (Supplementary Figs. 137–140). By contrast, the magnetic data for 2-crypt show that the magnetic moment of the complex increases from 4.10 μB (2 K) to 4.90 μB (300 K) (Supplementary Fig. 143). Similar data were obtained for 2-K (per complex, 3.94 μB at 2 K and 6.84 μB at 300 K under 0.1 T applied field; 3.24 μB at 2 K and 5.12 μB at 300 K under 1 T applied field; Supplementary Fig. 149). These values are increased compared to 1 across the entire temperature range, reflecting the presence of an S = ½ radical in addition to two 5f2 U(IV) ions (Supplementary Fig. 146).

The χ versus T data for 2-crypt and 2-K do not exhibit any maxima feature that could be associated with U···U magnetic exchange coupling (Supplementary Fig. 142). However, the variable temperature behaviour is also not as expected for isolated spins. Interestingly, the magnetic data for 2-crypt and 2-K are similar to those found for the N23− radical complex [K(2.2.2-crypt)][{(ArO)3UIV}2(μ-η2:η2-N2)] (1.8 K, 3.6 μB; 300 K, 5.0 μB)26. Powdered 1 was found to be electron paramagnetic resonance (EPR)-silent (X-band) at 6 and 298 K, which agrees with its U(IV)/U(IV)-N22− formulation70. In contrast, the X-band EPR spectrum of 2-crypt, although silent at 298 K, shows a single axial feature at g = 8.6 (H0 = 780 Gs; g = Landé g-factor, H0 = magnetic field) both in solution and in frozen 2-methyltetrahydrofuran glass at 6 K, indicating that the entire molecular species possesses a Kramers ground state (Supplementary Figs. 134 and 135)71. A better-resolved EPR signal was observed for the inner-sphere N23− radical complex 2-K in the solid state at 6 K, with g1 = 7.9 (H0 = 850 Gs), g2 = 0.73 (H0 = 9,205 Gs) and g3 ~ 0.5 (H0 = 1,330 Gs) (Supplementary Fig. 136). Although EPR signals were not previously observed for the radical-bound U-complexes [{(SiMe2NPh)3-tacn}UIV(η2-N2Ph2)]72 and [{(tBuArO)3tacn}UIV(η2-NNCPh2)] (tacn = 1,4,7-triazacyclononane)73, nor the N23− radical-bridged diuranium complex [K(2.2.2-crypt)][{(ArO)3U}2(μ-η2:η2-N2)]26, the magnetic and anisotropic EPR data of 2-crypt and 2-K are similar to those of the diphosphorus radical trianion diuranium(IV) complex [K(2.2.2-crypt)][{U(TrenTIPS)}2(μ-η2:η2-P2)] (TrenTIPS = {N(CH2CH2NSiiPr3)3}3−). The latter was found to have a strong U–P radical, but very weak U···U magnetic coupling74.

Overall, the magnetic and EPR data are in agreement with the presence of two magnetically independent U(IV) ions bridged by an N22− in 1, and the presence of two N23−-bridged U(IV) ions in 2-crypt and 2-K with a strong uranium radical but weak U···U magnetic exchange coupling.

Having established that the mono-reduced N2 analogues of 1 are synthetically accessible, we probed the further reduction of the N2 unit (Fig. 2). To accommodate the increased steric and electronic loadings anticipated from further reductions, and thus facilitate the isolation of the reduced species, we carried out reductions in the presence of added Li cations. Addition at −40 °C of a solution of 1 and 2 equiv. of LiI in diethyl ether to a suspension of 2 equiv. of KC8 resulted in a colour change from brown-red to maroon over 3 days at −40 °C. After workup, dark-red crystals of [Li2{UIV(TrenDMBS)}2(μ-η2:η2-N2)] (3) were obtained after 4 days in 63% yield. The solid-state structure of 3 shows the presence of a central hydrazido(4−) moiety bridging two U ions disordered over two orientations (Fig. 3d and Supplementary Figs. 5 and 6). The N–N bond (Fig. 3g) is substantially elongated compared to 1 and 2-crypt and is within the same range for both orientations (N1–N2 = 1.483(19); N3–N4 = 1.47(2) Å), and essentially the same as free hydrazine (N–N = 1.47 Å). This value falls within the range of the N–N bond distance values found in U(IV) and U(V) hydrazido complexes, namely 1.491(5) Å in K4[UIV2(μ-N2H2)(mTPR−)2] (mTP = [{2-(OC6H2-tBu-2,Me-4)2CH}-C6H4-1,3]4−; R = Met,Bu), 1.40(1) Å (ref. 24) in [K2{[UV(OSi(OtBu)3)3]2(μ-O)(μ-η2:η2-N2)}]25 (Fig. 1, B) and 1.521(18) Å in [K3{[UV(OSi(OtBu)3)3]2(μ-N)(μ-η2:η2-N2)}23. Similar N–N distances were also found in Zr(IV) hydrazido complexes (1.457(3)–1.548(7) Å)21,75. The metrical parameters of bonding between the triamidoamine framework and the U-centres in 3 (U–Namide = 2.402 Å and U–Namine = 2.628 Å) are similar to those found in 2-Li, in agreement with the presence of U(IV) ions (Supplementary Table 1).

With the products of mono- and di-reduction of 1 in hand, we pursued further reduction to split the N2 molecule and access nitrides. When an excess of 6 equiv. of KC8 was added to the mixture of 1 and LiI (6 equiv.) and left stirring for 3 days at −40 °C, the colour of the reaction mixture gradually turned dark-brown. Subsequent workup and crystallization at −40 °C afforded the bis-nitride complex [Li4(OEt2){UIV(TrenDMBS)}2(μ-N)2] (4) as brown-black crystals in 56% yield (Fig. 3e). The solid-state structure of 4 shows two U-centres bridged by two nitrides (N6 and N3) at a non-bonding N···N distance of 2.70(1) Å in a diamond core U2N2 motif (Fig. 3g) with U–N bond distances of U2–N3 = 2.248(9) Å and U2–N6 = 2.184(9) Å (Supplementary Fig. 7 and Supplementary Table 1). Each nitride binds two Li cations, with two different binding modes: the Li cations interacting with N6 are coordinatively supported by two U-bound amide arms (Li1–N6 = 1.96(2); Li2–N6 = 1.98(2) Å), and the two cations capping the second nitride are held by one TrenDMBS arm that is dissociated from U (Li3–N3 = 2.16(3); Li3–N3 = 1.97(2) Å).

Complex 4 adds to the limited examples of U-nitrides obtained from the reductive cleavage of N2 (refs. 27,34,76,77) following the first report (Fig. 1, C)78 with only one example of a U(IV) bis-nitride obtained from cleavage of N2 having been reported77. Alternatively, brown-black crystals of 4 could be isolated through reduction of 2-K with 3 equiv. of KC8 and 4 equiv. of LiI for 1 day at −40 °C in diethyl ether, or from 3 and 2 equiv. of LiI with the addition of 2 equiv. of KC8 after 1 day under similar experimental conditions (Supplementary Figs. 24 and 25). Furthermore, it was found that 3 could in turn be accessed from 2-K by reaction with 1 equiv. of KC8 and 2 equiv. of LiI (diethyl ether at −40 °C, 1 day; Supplementary Figs. 20 and 21). These results highlight the possibility of controlled, stepwise addition of electrons into the side-on bound N2 by controlling the amount of external reducing agent. Although the stepwise reduction of end-on bound N2 in molecular complexes has been studied extensively6, here four different products of N2 reduction (N22−, N23−, N24−, N3−) have all been isolated for the same ligand system and for the same metal, and all have been shown to be accessible in a stepwise controlled manner. Recently, some of us reported the stepwise reduction of the U2–N2 complex [{(ArO)3UIV}2(μ-η2:η2-N2)]26, but in that case reduction of the N23− analogue resulted in the immediate splitting of N2, and identification of a potential hydrazido(4−) intermediate was not possible under conditions similar to those used here.

Finally, we explored the reactivity of complexes 3 and 4 with H2 (1.01 bar) at room temperature. X-ray-quality crystals of the nitrido/imido complex complex [Li3{(TrenDMBS)UIV}2(μ-N)(μ-NH)] (5) could be isolated in good yield (Supplementary Figs. 27 and 28) from the reaction mixture, indicating that H2 can cleave the hydrazido(4−) moiety at room temperature. The solid-state structure of 5 (Fig. 3f) reveals an asymmetric diamond-shape U2–N2 core (Fig. 3g) with disparate U–Nnitride distances (U2–N5 = 2.35(1); U1–N5 = 2.02(1) Å) (Supplementary Fig. 8). Loss of one Li-cation is observed following the protonation of the N6 nitrido to give an imido-group (U2–N6 = 2.21(1); U1–N6 = 2.14(1) Å).

Cleavage of metal-bound N2 by H2, a reaction directly relevant to the Haber process, has only been reported for group 4 ions at substantially higher temperatures21. To unambiguously confirm that the imido- proton originates from the H2 gas, complex 3 was reacted with D2. An immediate reaction was observed, producing 5 as a major species, with the 1H NMR spectrum of the crude reaction mixture dissolved in toluene-d8 showing the full set of resonances assigned to 5 (Supplementary Figs. 31 and 32), apart from the broad NH singlet at −170.25 ppm. The structure of 5 indicates that a second species must be formed. Other species were observed in the 1H and 7Li NMR spectra of the reaction mixture but could not be isolated. The origin of the imido- proton was further confirmed by observing the shift in the N-H/D IR stretch from 3,313 cm−1 (for 5) to 2,452 cm−1 (for 5-D2) upon isotopic substitution (Supplementary Fig. 127). The formation of 5 could be the result of H+/Li+ scrambling from a putative bis-imido complex formed upon cleavage of the N–N bond followed by rearrangement and formation of 3 and an imido–amido complex ([Li{(TrenDMBS)UIV}2(μ-NH2)(μ-NH)]).

Exposure of a frozen solution of 4 in toluene-d8 to 1.01-bar H2 and subsequent warming to room temperature resulted in an immediate colour change from dark-brown to brown. The 1H NMR and 7Li NMR spectra of the reaction mixture indicate that multiple species are formed, but only complex 5 could be isolated and crystallographically characterized. Similarly to the D2 reactivity of 3, exposure of 4 to deuterium gas gives a full set of resonances assigned to 5, with only the NH resonance absent, as the sole identifiable species (Supplementary Figs. 34 and 35). The formation of 5 is probably the result of heterolytic splitting of H2 by the bis-nitride complex, accompanied by elimination of LiH, but the latter could not be unambiguously identified. Germane to this point, heterolytic splitting of H2 by a nitride bridged U(IV) complexes to give a rare imide–hydride complex, with the hydride bridging two U(IV) and a Cs cation, has been reported53. Here, the high affinity of the hydride for the Li-cation would account for elimination of the putative LiH.

Computational analysis

To provide further insight into the U–N2 complexes reported here, we conducted density functional theory (DFT) calculations on 2-crypt, 2-Li, 2-K and 3. Additionally, although 1 has already been extensively examined by DFT79, we computed 1 at the same level of theory to provide a meaningful benchmark. The closely related analogue of 4, [{UIV(μ-NLi2)(TrenTIPS)}2], has been computationally analysed in detail previously80, so we did not compute 4 here. Given the Li-cation disorder in 3, we computed two isomers with respect to the positions of the Li cations: the side-on/end-on isomer 3A and the end-on/end-on isomer 3B. We find that 3A, in the gas phase, is more stable than 3B by 3.37 kcal mol−1, consistent with the experimental observation of crystallographic disorder.

For spin-quintet 1, there are four α-spin electrons, where HOMO (highest occupied molecular orbital) to HOMO−2 are of essentially pure 5f character, and HOMO−3 is mixed 5f/N2-π* (46/45%, the B(πg⊥) δ-symmetry orbital). HOMO−4 (and its β-spin counterpart) constitute the A(πg=) in-plane bonding combination (Supplementary Fig. 151). The addition of an extra electron to give 2-crypt, 2-Li and 2-K in each case introduces an extra α-spin 5f electron, producing spin-sextet formulations. Hence for 2-crypt, 2-Li and 2-K, in each case HOMO to HOMO−3 are 5f-character, and HOMO−4 is the 5f/N2-π* B(πg⊥) (average 33/57%) (Supplementary Figs. 152–154). HOMO−5 and its β-spin counterpart are the A(πg=) combination. On moving to spin-quintet 3A/3B, again, four α-spin electrons of 5f-character are found in HOMO to HOMO−3, and HOMO−4 and HOMO−5 and their β-spin counterparts are the B(πg⊥) and A(πg=) orbitals, confirming full occupation of the π*-manifold of the N2 unit (Supplementary Figs. 155 and 156).

The computed uranium MDCq charges (1.90–2.42; MDCq is the multipole derived charge, where q denotes quadrupolar) and MDCm (MDCm is the multipole derived charge, where m denotes multipole) net spin densities (2.09–2.53) for 1, 2-crypt, 2-Li, 2-K and 3 reflect the formal presence of U(IV) ions in all complexes, consistent with the characterization data (Supplementary Table 4). The analogous MDCq data for the N2 units progressively increase from −1.88 to −2.73, reflecting the increasing formal N2 charges across the series. This is also reflected in the MDCm data, where the net spin density of 0.25 electrons on the N22− unit in 1 increases to 0.36 (average) for the N23− radical group in 2-crypt, 2-Li and 2-K, but becomes a net spin density deficiency in 3 (−0.3), reflecting the strongly donating capacity of N24−. Notably, across the series, the N–N bond order progressively weakens, from 1.66 in 1 to 1.27 in 3A, and this is reflected by quantum theory of atoms in molecules (QTAIM) topological bond analysis and in analytical frequency calculations, where clear N–N stretches are computed at 1,207 cm−1 (1), 1,180 cm−1 (2-K), 1,139 cm−1 (2-crypt) and 990/915 cm−1 (3B/3A).

Stoichiometric and catalytic conversion of N2 to NH3

To provide a baseline for catalytic studies, we first studied the stoichiometric reactions of 1, 2-crypt, 3 and 4 with H+, because the yield of the resulting NH3, isolated as the conjugate acid NH4Cl, could indicate the degree of N2 activation, with higher yield associated with greater activation (Table 1, entries 1–4). A previous control experiment where [Ti(TrenTMS)Cl] (TrenTMS = {N(CH2CH2NSiMe3)3}3−) was treated with 1 M HCl only produced 0.04 NH3 equivalents, suggesting only minor degradation of TrenR ligands under the action of strong acids. This was confirmed by treating [UIV(TrenDMBS)Cl] with 2 M ethereal HCl, which produced 0.03 NH3 equivalents. Treatment of crystalline 1 with excess 2 M HCl affords 0.18 NH3 equivalents, indicating weak activation of the pernitride unit (Supplementary Fig. 36). Analogously, crystalline 2-crypt produced 0.43 NH3 equivalents, suggesting stronger activation than 1 (Supplementary Fig. 37). This is more NH3 equivalents than found for [K(2.2.2-crypt)][{(ArO)3UIV}2(μ-η2:η2-N2)] (0.34 equiv.), noting that no NH3 was detected from HCl treatment of [{(ArO)3UIV}2(μ-η2:η2-N2)]26. For HCl treatment of hydrazido(4−)-bridged 3, 1.57 NH3 equivalents were detected, indicating much stronger activation of the N–N bond compared to 1 and 2-crypt (Supplementary Fig. 38). For comparison, 0.54–1.08 NH3 equivalents were detected upon direct protonation of the dihydrohydrazido(2−) complex K4[UIV2(μ-N2H2)(mTPR−)2] (mTPR− = [{2-(OC6H2-tBu-2,R-4)2CH}-C6H4-1,3]4−; R = Me, tBu)24,81 with [HPy]Cl, and [K3{[UV(OSi(OtBu)3)3]2(μ-N)(μ-η2:η2-N2)}]23 gave 0.5–0.84 NH3 equivalents after addition of excess ethereal HCl, and 1.54 NH3 equivalents when the same compound was first exposed to 1.01-bar H2 gas before HCl. Finally, crystalline bis-nitride 4 quantitatively produced 2.0 equiv. of NH3 when treated with HCl, as previously found for other U(IV/V/VI)–nitride complexes (Supplementary Fig. 39)34,54,55,61,82. These reactions confirm the anticipated range of N–N activation, consistent with the experimental and computational characterization data of 1, 2-crypt, 3 and 4.

Table 1 Stoichiometric and catalytic acidification experiments for the reaction of crystalline 1, 2-crypt, 3 and 4 with 2 M ethereal HCl and solutions of 1 with acids and reductants under N2 to produce NH3 and N2H4
Full size table

We next undertook catalytic studies, and, based on the recently reported titanium-catalysed conversion of N2 to NH3 (ref. 14), we examined the reactivity of 1 with the weak acid source [Cy3PH][I] (Cy = cyclohexyl) and elemental alkali metals (Na, K, Rb, Cs) or MC8 alkali-graphite (M = K, Rb, Cs) reagents (Table 1). These combinations benefit from the low solubility of both the H+ and e− sources, minimizing their sacrifical reaction with each other rather than sequential reactivity with U2–N2 species (Supplementary Figs. 40–73).

To provide a firm basis on which to undertake catalytic studies, we established several experimental baselines. The absence of 1 is decisive, with only 0.03 equiv. of N2H4 formed when [Cy3PH][I] and KC8 are mixed together then worked up (Table 1, entries 5 and 6). To examine the potential effect of solvent (Table 1, entries 7–9), we tested the reactivity of ~1.3 mM solutions 1 with 4 equiv. of KC8 and 10 equiv. of [Cy3PH][I] and found up to 1.18 NH3 equivalents, with yields in the order diethyl ether > hexane > toluene. We thus focused on using diethyl ether for further reactions. The weakly acidic nature of [Cy3PH][I] minimizes protonation of the tren-ligand. This is underscored by the fact that attempts to obtain catalytic conversion of N2 to NH3 using 1, [Et3NH][X] (X = BPh4−, Cl− or I−) and KC8 did not result in any NH3 being detected (Table 1, entries 10–12), suggesting that the [Et3NH][BPh4] is too aggressive a H+ source, leading to protonation of the TrenDMBS ligand (Supplementary Fig. 74). The importance of a halide source to the catalysis, to produce [U(TrenDMBS)I] (after fixed-N is eliminated) that can be reduced to regenerate 1, is demonstrated by the failure to generate NH3 with [Et3NH][BPh4], and the use of [Cy3PH][BArF20] also did not sustain any catalysis (Table 1, entries 10 and 13), with only 0.4 NH3 equivalents detected (Supplementary Figs. 45–48). Finally, although the catalytic conditions in this study are quite different to those of electrochemical scenarios, to benchmark potential NOx-to-NH3 conversion false positives, we conducted control experiments83,84 with 1 equiv. of nitrite or nitrate present (Supplementary Table 3), and found that only up to 0.53 equiv. of ammonia are produced. Given that the catalysis described here is in a sealed rather than continuous-flow scenario, with scrupulously cleaned glassware, and that NOx impurities in N2 gases are at low levels, NOx-to-NH3 conversion false positives can thus be ruled out. Having ascertained the optimal solvent and H+ source, we examined varying the reductant.

Reaction of 0.65 mM solutions of 1 with 600 equiv. of Na and [Cy3PH][I] under 1.31-bar N2 produced no NH3 equivalents and only a small quantity of N2H4 (0.13 equiv.), with a maximum of 0.26 equiv. of fixed-N (Table 1, entry 14). Equivalent reactions with K and Rb under 1.31-bar N2 were similarly unproductive (Table 1, entries 15 and 20), and we attribute the poor performance of Na, K and Rb to kinetic factors. In contrast, however, the equivalent reaction with Cs (Table 1, entry 23) produced 4.94 equiv. of NH3 along with a trace amount of N2H4. We rationalize this on the basis that during the catalytic runs with Cs, a small amount of Cs would be liquid due to the low melting point of Cs (28.5 °C), and this could thus facilitate the reaction kinetics. Noting that 1 exhibited 2.5 turnovers with Cs, we turned our attention to MC8 reagents using 600 equiv. of MC8/[Cy3PH][I] (Table 1, entries 16–19, 21, 22, 24 and 25). Starting with KC8, we find that reactions conducted with 173, 300 and 600 equiv. of KC8/[Cy3PH][I] produced 1.74, 0.88 and 0.62 equiv. of NH3, respectively, along with modest amounts of N2H4 (0.12–0.46 equiv.). On first inspection these observations are counterintuitive, but they probably reflect the complexity of a number of finely balanced factors critical to successful catalytic turnover. In contrast to Cs, CsC8 produced 2.61 equiv. of NH3 with trace N2H4. However, importantly, RbC8 reproducibly produced 8.84 equiv. of NH3 and no N2H4, representing 4.4 catalytic turnovers (Supplementary Fig. 56). To confirm the source of the fixed-N over and above the baseline acidifications, we repeated the RbC8 reaction with 1-15N2 and 15N2, and found that 8.17 equiv. of NH3 are formed, which is consistent with the 14N2 data (Supplementary Fig. 57). When the same catalytic reaction (600 equiv. RbC8/600[Cy3PH][I]) was conducted with nitride complex 4, 4.4 equiv. of NH3 were formed, suggesting that a nitride species could be involved in the catalytic pathway, but the presence of strongly bound Li cations in 4 may hinder the catalytic efficiency (Supplementary Fig. 73; Table 1, entry 25). This work thus establishes catalytic N2 fixation to NH3, albeit modest, with a side-on bound complex of N2, and, when taken together with the stoichiometric sequential reductions, we propose two possible catalytic cycles (Fig. 4): (1) sequential reduction, with protonation only occurring at the nitride step, or (2) proton-coupled sequential reduction. Notably, the addition of 600 equiv. of weak acid ([Cy3PH][I]) to the N23−, N24− and N3− species in the absence of reducing agent results in ammonia formation only for the nitride complex. However, when the addition of acid is carried out in the presence of 600 equiv. RbC8 under Ar we observed the formation of ammonia for the three complexes, albeit to a different extent (1.3 equiv. for 2-crypt, 0.9 equiv. for hydrazido 3 and 2.0 equiv. for nitride 4). This suggests that protonation at earlier stages could also play a role in the catalytic cycle (Supplementary Figs. 70–72 and Supplementary Table 3, entries 33–35). Finally, given prior work on Mo–triamidoamine complexes, it may well be that the triamidoamine amide centres shuttle incoming protons to the reduced dinitrogen moiety85,86; such mechanistic detail will require further mechanistic study to elaborate.

Fig. 4: Proposed catalytic cycles for the DMBSTren uranium–N2 system.
Fig. 4: Proposed catalytic cycles for the DMBSTren uranium–N2 system.
Full size image

Two possible mechanisms are shown for the conversion of N2 to NH3 by 1: (1) sequential reduction with protonation only occurring at the nitride step; (2) proton-coupled sequential reduction (grey H atoms).

Conclusion

To conclude, we have reported the stepwise reduction of N2 from free-N2 to bridging side-on bound forms and subsequently to bridging nitrides, in the process accessing four different states of side-on bound N2 for the same dinuclear U-complex. Our combined experimental and quantum-chemical calculation characterization data provide a consistent description of the U-complexes in this study and show that although the coordinated N2 undergoes sequential reduction steps, the U ions largely remain as U(IV) throughout. Although the catalytic turnover is currently modest, we have established molecular U-mediated catalytic N2 to NH3 conversion and moreover demonstrated N2 to NH3 catalysis with side-on bound coordination of N2, adding this mode of reactivity to the paradigm of catalytic conversion of N2 to NH3 by terminal or bridging end-on metal–N2 complexes. Our stoichiometric reactions suggest the relevance of N22−, N23−, N24− and N3− in the catalytic conversion of N2 to NH3 when involving side-on bridging N2. The recognition that side-on bound N2 can be catalytically activated and converted into NH3 through polymetallic cooperativity provides conceptual homogeneous–heterogeneous links to, for example, Haber–Bosch processes, and offers new vistas for the elaboration of molecular catalytic transformations involving N2 and probably other small molecules.

Methods

General considerations

All manipulations were carried out under an inert dinitrogen atmosphere using a standard Schlenk line or glovebox techniques. Catalytic trials were carried out using a J. Young sealable two-bulb apparatus (Supplementary Fig. 1). Water and oxygen levels were kept below 0.1 ppm at all times. Glass-coated stir bars were used instead of Teflon-coated ones to prevent unwanted polytetrafluoroethylene (PTFE) reactivity during reduction reactions. All solvents and reagents were rigorously dried and deoxygenated before use. Prepared compounds were characterized by elemental analyses, NMR, infrared (IR) and EPR spectroscopy, superconducting quantum interference device (SQUID) magnetometry, single-crystal X-ray diffraction studies and DFT, natural bond orbital (NBO) and bond topology theoretical calculations. Further details on the synthetic preparations as well as additional characterization for all compounds are provided in Supplementary Information.

Synthesis of [{UIV(TrenDMBS)}2(μ-η 2:η 2-N2)] (1), modified procedure

Under an atmosphere of Ar (ref. 35), a pale-green solution of [UIV(TrenDMBS)Cl] (374.2 mg, 0.493 mmol, 1 equiv.) in n-hexane (10 ml) was added to a Schlenk tube containing a glass-coated stir bar. Solid potassium graphite (199.8 mg, 1.478 mmol, 3 equiv.) was then added in portions, which resulted in the formation of a dark-purple reaction mixture that was allowed to stir for 24 h at room temperature. Full conversion to the [UIII(TrenDMBS)] species was confirmed by 1H NMR spectroscopy. The resulting reaction mixture was filtered on a porosity ‘4’ glass frit, and the solid residue was rinsed with n-hexane (3 × 0.5 ml), yielding a dark-purple solution. The volatiles were removed in vacuo, and the resulting purple solid was brought into the dinitrogen glovebox (1.016-bar absolute N2 pressure) and dissolved in n-pentane (2 ml) to immediately form a dark-red solution. The solution was then stored at −40 °C for 2 days to yield dark-red crystals of 1 in 48% yield (178.1 mg, 0.118 mmol) that were dried under N2 flow. The mother liquor was concentrated to 1 ml, and then stored at −40 °C, yielding additional 1 (48.2 mg, 0.032 mg) after 2 days. The procedure was repeated one more time (concentration to 0.5 ml and storage at −40 °C), yielding a third and final crop of 1 after 2 days (0.024 mmol, 37.1 mg). The total yield was 71% (263.4 mg). It was found that the use of n-hexane instead of pentane for the crystallization step resulted in lower overall yield. Elemental analysis: Calcd. for C48H114N10Si6U2 (1): C, 39.06; H, 7.78; N, 9.49. Found: C, 39.22; H, 7.79; N, 9.30. 1H NMR (400 MHz, toluene-d8, 298 K): δ 10.95 (s, 12H, CH2), 7.21 (s, 12H, CH2), 3.84 (s, 54H, C(CH3)3), −20.24 ppm (s, 36H, Si(CH3)2) (Supplementary Fig. 9). The preparation of the 15N2-labelled isotopologue, 1-15N2, is described in Supplementary Information.

When the synthesis of 1 was performed under 2.21-bar N2, analytically clean 1 could be isolated in a better yield (compared to when performed under 1 atm (1.01 bar)) from a single crystallization. Once isolated, complex 1 is stable in the solid state under N2, but quickly releases N2 under dynamic vacuum. Complex 1 (~8.9 mM concentration) in toluene-d8 and cyclohexane-d12 at 1.01-bar N2 exist in equilibrium with [UIII(TrenDMBS)] in an approximate 2:1 ratio (1H NMR, Supplementary Fig. 9). No increase in N2 uptake was observed by 1H NMR spectroscopy when lowering the solution temperatures to −80 °C. However, increasing the dinitrogen pressure to 2.51 bar leads to a further shift in 1:[UIII(TrenDMBS)] equilibrium, from 2:1 to 5:1 (Supplementary Fig. 12) at room temperature. In contrast, the 1H NMR spectrum of the isolated crystalline 1 in THF-d8 showed predominantly [UIII(TrenDMBS)] resonances, indicating that N2 loss occurs in the presence of coordinating solvents (Supplementary Fig. 10). This was further corroborated by the 15N NMR spectroscopy of 1-15N2, where only the resonance for free 15N2 could be observed (Supplementary Fig. 14), with the absence of a resonance for 1-15N2 being attributed to its paramagnetism.

Synthesis of [K(2.2.2-cryptand)][{UIV(TrenDMBS)}2(μ-η 2:η 2-N2)] (2-crypt)

Under an atmosphere of dinitrogen, a solution of 1 (130.1 mg, 0.086 mmol, 1 equiv.) and 2.2.2-cryptand (32.5 mg, 0.086 mmol, 1 equiv.) in toluene (4 ml) was added to a Schlenk tube containing a glass-coated stir bar. Solid potassium graphite (11.7 mg, 0.086 mmol, 1 equiv.) was then added at room temperature, and the reaction was left stirring for 3 h to yield a dark-brown mixture. The volatiles were then removed in vacuo, and soluble residues were extracted into THF (4 ml) to facilitate graphite removal by filtration on a porosity ‘4’ glass frit. Volatiles were then removed in vacuo from the filtrate, the residue dried for 1 h, and toluene (~2 ml) added. Storage of the brown solution at −40 °C for 2 days yielded 2-crypt·2C7H8O as dark-brown crystals in 83% yield (135.6 mg, 0.072 mmol). 2-crypt is readily soluble in THF and only modestly in toluene; the complex was found to be stable towards nitrogen loss in THF and toluene (no change in the 1H NMR spectra after dissolution or a freeze–pump–thaw cycle). Elemental analysis: Calcd. for C69.5H154KN12O6Si6U2 (2-crypt · 0.5C7H8O): C, 43.08; H, 8.01; N, 8.67. Found: C, 43.04; H, 8.05; N, 7.93. 1H NMR (400 MHz, 298 K, THF-d8): δ 86.21 (s, 12H, CH2), 25.89 (s, 12H, CH2), 3.06–3.03 (m, 24H, 2.2.2-crypt), 2.04 (m, 12H, crypt), −13.02 (s, 54H, C(CH3)3), −33.64 (s, 36H, Si(CH3)2) (Supplementary Fig. 15a). 29Si{1H} NMR (79.5 MHz, 298 K, THF-d8): δ −181.81 (Supplementary Fig. 15b).

Synthesis of [K{UIV(TrenDMBS)}2(μ-η 2:η 2-N2)] (2-K)

Under an atmosphere of dinitrogen, solid potassium graphite (0.217 mmol, 29.3 mg, 3 equiv.), pre-chilled to −40 °C, was added to a cold solution of 1 (108.7 mg, 0.072 mmol, 1 equiv.) in n-hexane. The mixture was left to stir for 3 days at −40 °C, then filtered on a porosity ‘4’ glass frit, also pre-chilled to −40 °C, and concentrated to 0.5 ml while cold. Dark-brown/green crystals of 2-K were obtained from n-hexane after 2 days at −40 °C, in 62% yield (69.1 mg, 0.045 mmol). Elemental analysis: Calcd. for C51H121KN10Si6U2 (2-K·0.5C6H14): C, 39.31; H, 7.83; N, 8.99. Found: C, 39.17; H, 7.78; N, 8.63. 1H NMR (400 MHz, 233 K, toluene-d8): δ 5.88 (s, br.), −21.20 (s, very br.), 83.58 (s, very br.) (Supplementary Fig. 16a). (193 K): −12.46 (s, very br.), −27.79 (s, very br.), −36.57 (s, very br.), −132.49 (s, very br.), −150.65 (s, very br.) (Supplementary Fig. 16b).

Synthesis of [Li2{UIV(TrenDMBS)}2(μ-η 2:η 2-N2)] (3)

Under an atmosphere of dinitrogen, crystalline 1 (130.4 mg, 0.088 mmol, 1 equiv.) was combined with solid lithium iodide (23.7 mg, 0.177 mmol, 2 equiv.). At room temperature, diethyl ether (2 ml) was added, and the mixture was stirred until fully dissolved, before being stored at −40 °C for 1 h. The resulting cold mixture was added to cold (−40 °C) solid potassium graphite (23.9 mg, 0.177 mmol, 2 equiv.) in cold diethyl ether (1 ml) and stirred for 3 days at −40 °C before filtration on a porosity ‘4’ glass frit pre-chilled to −40 °C to obtain a dark-brown/red solution. Concentration in vacuo to ~1 ml and storage at −40 °C for 4 days yielded dark-red/brown crystals of 3 in 63% yield, obtained in several crops (82.9 mg, 0.056 mmol). Elemental analysis: Calcd. for C52H124Li2N10OSi6U2 (3 ·1C4H10O): C, 39.93; H, 7.99; N, 8.96. Found: C, 39.56; H, 7.79; N, 8.65. 1H NMR (400 MHz, 298 K, toluene-d8): the spectrum shows paramagnetically shifted resonances spanning a wide range from −130 to 200 ppm, indicating the presence of a highly asymmetric structure in solution (Supplementary Figs. 18 and 19).

Synthesis of [Li4(OEt2){UIV(TrenDMBS)}2(μ-N)2] (4)

Under an atmosphere of dinitrogen, crystalline 1 (107.9 mg, 0.073 mmol, 1 equiv.) was combined with solid lithium iodide (58.8 mg, 0.439 mmol, 6 equiv.). At room temperature, diethyl ether (3 ml) was added and the mixture stirred until fully dissolved, before being stored at −40 °C for 1 h. The resulting mixture was added to cold solid potassium graphite (59.3 mg, 0.439 mmol, 6 equiv.) in diethyl ether (1 ml) and allowed to stir for 3 days at −40 °C before filtration on a cold porosity ‘4’ glass frit to obtain a dark-brown solution. Concentration in vacuo to ~1 ml and storage at −40 °C for 3 days yielded dark-brown crystals of 4·1C4H10O in 56% yield (61.6 mg, 0.056 mmol). Elemental analysis: Calcd. for C52H124Li4N10OSi6U2 (4): C, 39.58; H, 7.92; N, 8.88. Found: C, 39.48; H, 7.89; N, 8.89. 1H NMR (400 MHz, 298 K, toluene-d8): the spectrum shows paramagnetically shifted resonances spanning a wide range from −135 to 150 ppm, indicating the inequivalence of many of the 114 protons present in the complex (Supplementary Figs. 22 and 23).

Synthesis of [Li3{UIV(TrenDMBS)}2(μ-N)(μ-NH)] (5)

A dark-red/brown solution of 3 (37.1 mg, 0.025 mmol, 1 equiv.) in toluene (0.5 ml) was transferred to a J. Young valve-capped tube and connected to a Schlenk line. The resulting solution was degassed by three cycles of freeze–pump–thawing, and hydrogen gas (1.01 bar) was added to the reaction mixture and the tube warmed to room temperature. A gradual colour change to brown was observed over the course of 2 days as full consumption of the starting material occurred. Brown crystals of 5 were collected from the reaction mixture after leaving it to stand at room temperature (7.9 mg) for 3 days. An additional crop (4.4 mg) of crystalline 5 was isolated after 1 day upon placing the reaction mixture at −40 °C. Elemental analysis: Calcd. for C48H115Li3N10Si6U2 (5): C, 38.49; H, 7.74; N, 9.35. Found: C, 38.97; H, 7.75; N, 9.32. 1H NMR (400 MHz, 298 K, toluene-d8): the spectrum shows paramagnetically shifted resonances spanning a wide range from −170 to 170 ppm, including the NH resonance identified at −170.25 ppm (s, 1H, NH) (Supplementary Figs. 29 and 30). IR (KBr pellet, ν/cm−1): 3,313 (w, N–H), 2,953 (s), 2,929 (s), 2,883 (s), 2,852 (s), 2,735 (w), 2,705 (w), 2,672 (w), 1,605 (m), 1,468 (m), 1,405 (w), 1,388 (w), 1,357 (w), 1,246 (m), 1,133(w), 1,113 (w), 1,092 (m), 1,072 (m), 1,057 (m), 1,023 (m), 1,011 (w), 967 (m), 925 (s), 903 (m), 888 (s), 825 (s), 777 (s), 717 (m), 655 (s), 586 (m), 560 (m), 527 (m), 505 (m) (Supplementary Fig. 127).

A second species observed in the 1H NMR spectrum (Supplementary Figs. 26 and 27) of the mother liquor measured in toluene-d8 could not be isolated despite repeated attempts.

Synthesis of [Li3{UIV(TrenDMBS)}2(μ-N)(μ-ND)] (5-D2) through the reaction of 3 with D2 gas

A dark-red/brown solution of crystalline 3 (37.1 mg, 0.025 mmol, 1 equiv.) in toluene (0.3 ml) was transferred to a J. Young valve-capped NMR tube and connected to a Schlenk line. The resulting solution was degassed by three cycles of freeze–pump–thawing and deuterium gas (0.5 bar absolute pressure of D2) was added to the reaction mixture and the tube was warmed to room temperature. An immediate colour change from dark-red/brown to brown was observed. After 1 h, all volatiles were removed in vacuo and toluene-d8 (0.3 ml) was added. Dark-brown crystals of 5-D2 (6.6 mg) were isolated from the reaction mixture after 2 days at −40 °C. 1H NMR (400 MHz, 298 K, toluene-d8): the spectrum of the reaction mixture as well as the one of isolated crystalline 5-D2 showed the full set of resonances assigned to 5 (Supplementary Figs. 31–33), except for the broadened NH singlet at −170.25 ppm (Supplementary Fig. 32), which is not present in the deuterated mixture. IR (KBr pellet, ν/cm−1): 2,953 (s), 2,928 (s), 2,883 (s), 2,851 (s), 2,704 (w), 2,452 (w, N–D), 1,468 (m), 1,444 (w), 1,387 (w), 1,357 (w), 1,288 (m), 1,133 (w), 1,111 (w), 1,091 (m), 1,056 (m), 1,024 (m), 968 (m), 924 (s), 902 (m), 889 (s), 825 (s), 779 (s), 717 (m), 654 (s), 584 (m), 526 (w), 505 (w), 484 (w) (Supplementary Fig. 127).

Reaction of 1, 2-crypt, 3 and 4 with excess HCl

To J. Young valve-capped NMR tubes containing solid complexes 1 to 4, a solution of HCl (2 M, 2.0 ml in total) in diethyl ether was added at −80 °C in a glovebox-fitted cold well. On warming each tube to −40 °C and then subsequently to room temperature, discolouration of the solids occurred, and formation of light-yellow solutions was observed. After 1 h, all volatiles were removed in vacuo and the resulting solids were dissolved in dimethylsulfoxide (DMSO)-d6 (0.5 ml). 2,5-Dimethylfuran was added to the tubes directly or in the form of an NMR insert (sealed quartz capillary) as an internal standard for the quantitative detection of NH4Cl. Formation of 0.18 equiv. of NH4Cl for 1, 0.43 equiv. for 2-crypt, 1.57 equiv. for 3 and 2.0 equiv. for 4 were detected by quantitative 1H NMR experiments (Supplementary Figs. 36–39). Under similar experimental conditions, the parent monomeric complex [UIV(TrenDMBS)Cl] produced 0.03 equiv. of NH4Cl.

Acidification of [{UIV(TrenDMBS)}2(μ-η 2:η 2-N2)] under a 14N2 atmosphere

Under an atmosphere of Ar, bulb A was charged with 1 (0.0052 mmol), the acid and reductant, along with a glass-coated stirrer bar. Bulb A was then sealed and transferred to a Schlenk line. Et2O (stored under Ar, 4–8 ml, vide infra) was added to bulb B via syringe, and then bulb B was submerged in a LN2 dewar (−196 °C) and frozen. The entire apparatus was placed under a dynamic vacuum (1 × 10−3 mbar) and sealed, leaving the apparatus under static vacuum. The LN2 dewar (−196 °C) was removed from bulb B and immediately transferred to bulb A such that the frozen solvent in bulb B could thaw and distil onto the frozen solids in bulb A. After the distillation had concluded, the apparatus was placed under a dinitrogen atmosphere, sealed, and allowed to warm to room temperature. For reactions involving MC8 (M = K, Rb, Cs) as the reductant, the reaction mixture was stirred for 17 h. In the case of the alkali metals, the reaction mixtures were subsequently sonicated until the metal was finely dispersed (1–40 min) and the reactions were stirred for 72 h. After this time, HCl (2 M in Et2O, 2 ml) was added to bulb B via syringe and both bulbs were subsequently submerged into separate LN2 (−196 °C) dewars and frozen. The entire distillation apparatus was then placed under a dynamic vacuum and sealed, leaving the vacuum under static vacuum, and the LN2 dewar removed from bulb A. During this process, the reaction mixture in bulb A thaws, and the volatiles were distilled onto the frozen HCl in bulb B. Once the acid distillation was complete (~30 min), bulb A was resubmerged in a LN2 (−196 °C) dewar and the contents frozen. Bulb B was sealed, and aqueous KOH (30%, 4 ml) was added via syringe to bulb A under a flow of dinitrogen. Once the contents had frozen, the entire apparatus was once again placed under a dynamic vacuum and sealed, leaving the apparatus under static vacuum, and the LN2 (−196 °C) dewar removed from bulb A. The base distillation was performed with vigorous stirring for 1 h, then both bulbs A and B were sealed, and bulb B was allowed to warm to room temperature and stirred for 10 min. All volatiles in bulb B were subsequently removed in vacuo and the remaining residue analysed for NH3/NH4Cl and N2H4/N2H4·2HCl via 1H NMR spectroscopy (Supplementary Figs. 40, 45–56, 58, 61–73) and pdmab methods (vide infra). As previously reported5, N2H4 is only partially transferred under these conditions, and heating the distillation mixture should be avoided because N2H4 undergoes thermal decomposition to NH3, N2 and H2. Thus, after the distillation, the solids remaining in bulb A were also analysed for N2H4/N2H4·2HCl via the pdmab method (Supplementary Figs. 78, 79, 82–105, 108–119). The acidification of 1-15N2 under a 15N2 atmosphere is described in Supplementary Information.