Introduction

Terminal metal-nitrides, M≡N, represent a key fundamental class of metal-ligand linkage in coordination chemistry1. Although these M≡N triple bonds have been of elementary interest for over 170 years2, only in relatively recent times has there been a concerted effort to study their reactivity1,3. However, although a variety of reactivity patterns have emerged with metal-nitrides1, the vast majority are remarkably unreactive because strong, often highly covalent M≡N triple bonds that result from high oxidation state metal ions—needed to bind to the hard, charge-rich nitride, N3−—renders them inherently inert1,3. One strategy to increase the reactivity of metal-nitrides is to utilise low oxidation state electron-rich metals to destabilise the M≡N triple bond, but by definition such metals are ill-matched to nitrides and so are difficult to prepare4. Additionally, reactivity of metal-nitrides often involves ancillary ligands rather than the M≡N triple bond itself. Overcoming this challenge is difficult because there are very few metal-nitrides where the metal oxidation state or co-ligands can be varied within a homologous family to encourage M≡N triple bond reactivity1,5.

Since there is an isoelectronic relationship between the M≡N and N≡N triple bonds of metal-nitrides and dinitrogen, N2, respectively, the former are fundamentally mechanistically important with respect to Haber Bosch chemistry where they are invoked as intermediates in the cleavage of the latter and conversion to ammonia, NH3, by hydrogenolysis with dihydrogen, H26,7. There has thus been intense interest in the reactivity of metal-nitrides with H2, and indeed their use in N-atom transfer reactivity and catalysis more widely8,9,10,11,12, but there are few reports of molecular metal-nitrides reacting with H2, and indeed activating H2 in this homogeneous context remains a significant challenge in contrast to heterogeneous Haber Bosch chemistry where H2-cleavage is essentially barrier-less6. One solution to overcome this hydrogenolysis challenge may be to exploit Frustrated Lewis Pair (FLP) chemistry13,14, but so far this has been focussed on M–N2 complexes15,16. Usually with mid- or late-transition metals3, most metal-nitride hydrogenations involve sequential protonations17,18,19,20,21,22, but bridging nitrides in poly-iron/-titanium/-zirconium complexes have been reported to react with H2 to give imido-hydride and NH3 products23,24,25. Only three terminal metal-nitrides have been reported to undergo hydrogenolysis with H2. The isostructural d4 ruthenium(IV)- and osmium(IV)-nitrides [M{N(CH2CH2PBut2)2}(N)] (M = Ru, Os) react with H2 using the ancillary ligand to shuttle H-atoms to evolve NH326,27, and the 5d6 iridium(III)-nitride [Ir{NC5H3-2,2′-(C[Me]=N-2,6-Pri2C6H3)2}(N)] undergoes concerted reactivity with H2 to give [Ir{NC5H3-2,2′-(C[Me]=N-2,6-Pri2C6H3)2}(NH2)]28. Thus, direct hydrogenolysis of a M≡N triple bond with H2 remains exceedingly rare, and involves reasonably electron-rich (≥d4) metal complexes with low coordination numbers.

As part of our studies investigating actinide-ligand multiple bonding supported by triamidoamine ancillary ligands29,30,31,32,33,34,35, we have reported two closely related terminal uranium-nitrides [UV(TrenTIPS)(N)][K(B15C5)2] (1) and [UVI(TrenTIPS)(N)] (2) [TrenTIPS = N(CH2CH2NSiPri3)33−; B15C5 = benzo-15-crown-5 ether]36,37,38 that, unusually1,5,39, permit examination of the electronic structure and reactivity of the same isostructural terminal nitride linkage with more than one metal oxidation state. Both react with the small molecules CO, CO2, and CS240,41, but since only the protonolysis of 1 with H2O to give NH3 had been previously examined36 the ability of 1 and 2 to react with H2 has remained an open question. Indeed, the study of molecular uranium-nitride reactivity remains in its infancy36,37,40,41,42,43,44,45,46,47,48, and only very recently the diuranium(IV)-nitride-cesium complex [Cs{U(OSi[OBut]3)3}2(μ-N)] was reported to reversibly react with H2 to give the diuranium-imido-hydride complex [Cs{U(OSi[OBut]3)3}2(μ-NH)(μ-H)]47. Bridging nitrides tend to be more reactive than terminal ones, so whilst this nitride hydrogenolysis is enabled by the bridging nature of the nitride and polymetallic cooperativity effects47, we wondered whether H2 activation by 1 or 2 might still be accessible, given prior protonation studies36, since this would realise the first terminal f-block-nitride hydrogenolyses. Further motivation to study this fundamental reaction stems from the fact that bridging and terminal uranium-nitride reactivity with H2 is implicated in Haber Bosch NH3 synthesis when uranium is used as the catalyst49, and uranium-nitrides have been proposed as accident tolerant fuels (ATFs) for nuclear fission, but likely reactivity with H2 formed from radiolysis under extreme conditions or when stored as spent fuel remains poorly understood.

Here, we report that 2 does not react with H2 consistent with a strong U≡N triple bond that is inherently unreactive like many high oxidation state terminal metal-nitrides. However, in contrast 1 reacts with H2 under mild conditions despite the fact it can be considered to be a high oxidation state metal and not of a low coordination number nor electron-rich as a 5f1 metal ion. This hydrogenolysis reactivity is thus unprecedented in molecular metal-nitride chemistry, and further supports the emerging picture that suggests that the 5f-electron of 1 should not be considered as purely nonbonding. This study reveals two distinct H2-activation mechanisms. When the borane BMes3 (Mes = 2,4,6-trimethylphenyl) is present a FLP mechanism operates where two H2 heterolysis events and a borane reduction step sequentially combine to furnish a UIV-NH2 product, and this, to the best of our knowledge, is the first demonstration of the application of bona fide FLP reactivity to actinide chemistry. When the borane is absent, direct 1,2-addition of H2 across the U≡N triple bond to give a H−UV = N−H intermediate followed by H-atom migration produces a UIII-NH2 product that is easily oxidised to UIV-NH2. The direct addition is slower than the FLP-mediated mechanism, demonstrating the facilitating role of FLPs. We find evidence that treating the UIV-NH2 product with BPh3 and H2 produces further FLP hydrogenolysis reactivity, since H3NBPh3 has been detected in reaction mixtures, but this is reversible and produces products that react to give the starting materials. While currently of no practical use this demonstrates further potential for FLPs in this area. We demonstrate an azide to nitride to amide to ammonia reaction cycle, supported by overall hydrogenation involving hydrogenolysis and electrophilic quenching steps.

Results

Hydrogenolysis of the terminal uranium(V)-nitride bond

Since 2 was found to be unreactive or decomposed to a complex mixture of intractable products when exposed to boranes in the context of this study we examined the reactivity of 1. With or without H2, treatment of 1 in toluene with the strong Lewis acid B(C6F5)3 (BCF) results in decomposition as evidenced by 19F NMR spectra of reaction mixtures that show multiple fluorine resonances consonant with multiple C–F activation reactions, Fig. 1. Deleterious C–F bond activation reactivity is well documented for BCF50, and so we examined the reaction of 1 with the less Lewis acidic BPh3. However, when 1 is treated with BPh3 in toluene the adduct complex [UV(TrenTIPS)(NBPh3)][K(B15C5)2] (3), which when compared to 1 and 2 is perhaps best formulated as a uranium(V)-imido-borate rather than a uranium(V)-nitrido-borane, is rapidly formed quantitatively and isolated in crystalline form in 66% yield, Fig. 1.

Fig. 1: Synthesis of complexes 3–6.
figure 1

Treatment of 1 with the strong Lewis acid B(C6F5)3 results in decomposition, however the milder borane BPh3 produces the capped species 3, which is inert with respect to reaction with H2. Complex 1 does not react with the sterically encumbered BMes3, but exposure of that mixture to H2 produces the amide complex 4 with concomitant formation of the radical anion complex 5. Addition of H2 to 1 produces the amide complex 6, and subsequent treatment with BMes3 produces 4 and 5. Treating 4 with HCl produces NH3. B15C5 = benzo-15-crown-5 ether. Mes = 2,4,6-trimethylphenyl.

Full size image

The retention of uranium(V) in 3 is supported by absorptions in the 5000–12,500 cm−1 region of its UV/Vis/NIR spectrum (Supplementary Fig. 1) that are characteristic of intraconfigurational 2F5/2 to 2F7/2 transitions of uranium(V)38,51, and by variable-temperature SQUID magnetometry, Fig. 2 and Supplementary Fig. 2. A powdered sample of 3 returns a magnetic moment of 2.23 μB at 300 K (1.96 μB by solution Evans method) that changes little until 30 K where it falls quickly to a moment of 1.38 μB at 2 K and this is consistent with the magnetic doublet character of 5f1 uranium(V)52,53,54. The solid-state structure of 3, Fig. 3 and Supplementary Fig. 3, reveals a separated ion pair formulation where the nitride has been capped by the BPh3 unit. The U–Nimido bond length of 1.911(6) Å is consistent with the imido-borate formulation, for example distances of 1.916(4), 1.954(3), and 1.946(13) Å are found in [(ButArN)3UV(NBCF)][NBun4] (Ar = 3,5-dimethylphenyl)55, [UV(TrenTIPS)(NSiMe3)]37, and [UV(TrenTIPS)(NAd)] (Ad = 1-adamantyl)37, respectively, and the B-Nimido distance of 1.581(9) Å compares well to the sum of the single bond covalent radii of B and N (1.56 Å)56. The U–Namine distance of 2.737(5) Å is long, reflecting the dative nature of the amine donor and that it is trans to the strong imido donor, and the U–Namide distances (2.254(7)-2.312(6) Å) are slightly long for such distances57, reflecting the formal anionic nature of the uranium component of 3.

Fig. 2: Variable-temperature SQUID magnetic moment data for 3, 6, and 8.
figure 2

Key: 3 (black squares), 6 (red circles), and 8 (blue triangles). Data were measured in an applied magnetic field of 0.1 Tesla.

Full size image
Fig. 3: Molecular structure of the anion component of 3 at 150 K and displacement ellipsoids set to 40%.
figure 3

Hydrogen atoms, minor disorder components, lattice solvent, and the [K(B15C5)2]+ cation component are omitted for clarity. Selected bond lengths (Å): U1-N1, 2.305(5); U1-N2, 2.254(7); U1-N3, 2.312(6); U1-N4, 2.737(5); U1-N5, 1.911(6); B1-N5, 1.581(9).

Full size image

Complex 3 does not react with H2 (1 atm.), Fig. 1. Indeed, dissolving a mixture of 1 and BPh3 under H2 only generates 3, and so since BPh3 has shut all reactivity down by strongly binding to the nitride of 1, but BCF is too reactive, we examined the use of BMes3 (Mes = 2,4,6-trimethylphenyl). In principle, the ortho-methyls of the Mes groups of this borane block deactivating strong coordination of Lewis bases to the vacant p-orbital of boron whilst retaining a Lewis acidic boron centre.

To provide a reactivity control experiment, we stirred a 1:1 mixture of 1:BMes3 in toluene under an atmosphere of N2 and find no evidence for any adduct formation, Fig. 1, with only free BMes3 being observed as evidenced by a resonance at 76.8 ppm in the 11B NMR spectrum of the reaction mixture. Repeating this reaction, but under H2 (1 atm.), over two days at 298 K results in complete consumption of starting materials with deposition of a dark blue solid. The brown supernatant was removed and found by NMR spectroscopy to contain the known uranium(IV)-amide [UIV(TrenTIPS)(NH2)] (4) in 67% yield, Fig. 1, as evidenced by a resonance at 107 ppm in its 1H NMR spectrum that corresponds to the amide protons37. A control experiment, stirring 1 in toluene over two weeks, also produces 4 from trace, adventitious sources of H+, though in far lower proportions, so to prove that the source of H-atoms in 4 originates from H2, and not adventitious H+58, the reaction was repeated under D2 (1 atm., 99.8% atom D). Interestingly, whilst [UIV(TrenTIPS)(ND2)] (4″, 2H δ 107.5 ppm) is formed, confirming that hydrogenolysis by H2/D2 does indeed occur, it is always accompanied by 4 and [UIV(TrenTIPS)(NHD)] (4′, 1H δ 106 ppm, 2H δ 106.8 ppm, 2JHD not resolved). This reveals that H/D exchange occurs over time, so to determine the source of this exchange we studied the reaction of 1 and BMes3 with all combinations of H2/D2 with H6-/D6-benzene and H8-/D8-toluene (see Supplementary Methods). We find that when H2 is used only 4 is ever detected, but when D2 is used 4, 4′, and 4″ all form (av. 12, 24, and 64%, respectively) irrespective of whether the solvent is deuterated or not which rules out arene solvents as the H-source. However, when using C6H6 as solvent for the reaction of 1 with BMes3 and D2 a weak resonance is observed at −5.2 ppm in the 2H NMR spectrum (cf −5.35 and −5.87 ppm for iso-propyl methine and methyl protons, respectively, in the 1H NMR spectrum of 4). We therefore suggest that the H-source is the TrenTIPS Pri groups since they have precedent for forming cyclometallates57, a reversible amide/imido-cyclometallate + H2 equilibrium can be envisaged since it has been previously shown that uranium-Tren-cyclometallates can react reversibly with H2/D259, and this would also account for the absence of D-scrambling into 4 since TrenTIPS is void of D-atoms.

The dark blue solid was isolated and after work-up obtained as dark blue crystals, identified as the radical species [K(B15C5)2][BMes3] (5), in 69% yield. This compound has been structurally characterised by single crystal diffraction, see Supplementary Fig. 4. Compound 5 is very similar to [Li(12-crown-4)2][BMes3]60 that contains the same radical anion component, and the EPR data of 5 (g = 2.003, A(11B) = 9.44 G, A(10B) = 2.7 G, A(1H) = 1.2–1.4 G), Fig. 4a, confirm the formation of the BMes3•− radical anion formulation. The UV/Vis/NIR spectrum of 5 exhibits an intense, (ε = ~8000 M−1 cm−1) broad absorption centred at ~12,800 cm−1, which largely accounts for its dark blue colour, see Supplementary Fig. 5.

Fig. 4: EPR data for 5 and 6.
figure 4

a X-band (9.8 GHz) EPR spectrum of 5 at 298 K as a 9 mM solution in THF. The black line is the experimental spectrum and the red line the simulation with g = 2.003, A(11B) = 9.44 G, A(10B) = 2.7 G, A(1H) = 1.2–1.4 G. b X-band (9.3 GHz) EPR spectrum of a powdered sample of 6 at 20 K. The black line is the experimental spectrum and the red line the simulation with g = 4.19, 0.88, and 0.52. The very sharp signal marked with asterisk is a very small quantity of radical impurity with g = 2.

Full size image

Since 1 does not form an adduct with BMes3, but the introduction of H2 leads to hydrogenolysis to give the amide 4, we surmised that the 1/BMes3 mixture may constitute a Frustrated Lewis Pair (FLP) system that is evidently capable of activating H2, which is confirmed computationally (see below). However, since H2 can be a two-electron reducing agent and uranium(V) is normally quite oxidising, we hypothesised that the BMes3 may not actually be required. In effect, when 4 is produced it is essentially at the expense of the sacrificial one-electron reduction of BMes3 to BMes3•−, which would formally invoke a uranium(III)-amide precursor that would be nicely in-line with a H2-uranium(V) two-electron redox couple. In order to test whether the FLP aspect of this hydrogenolysis chemistry is vital to effecting dihydrogen activation a toluene solution of 1 under H2 (1 atm.) was stirred without BMes3. Over seven days 1 is consumed with concomitant precipitation of a gray solid identified as the uranium(III)-amide [UIII(TrenTIPS)(NH2)][K(B15C5)2] (6) (45% yield), Fig. 1. The hydrogenolysis reaction is now slower than when BMes3 is present, but the reaction is best conducted at 288 and not 298 K, which may also retard the rate of reactivity. When the reaction is alternatively conducted under D2 (1 atm.), a mix of 6, [UIII(TrenTIPS)(NHD)][K(B15C5)2] (6′) and [UIII(TrenTIPS)(ND2)][K(B15C5)2] (6″) are isolated (66% yield by uranium content) analogously to 4/4′/4″, again indicating H/D exchange but confirming the H-atoms of the amide unit in 6 originate from gaseous H2. Consistent with these observations, we find that 1 also reacts with 9,10-dihydroanthracene (pKa 31 in DMSO, cf 34±4 for H2 in DMSO)61 to produce an insoluble precipitate and 4 in solution. From this solution we isolated a small crop of red crystals formulated by 1H NMR spectroscopy and X-ray diffraction (see Supplementary Fig. 6) as [K(B15C5)2][C14H11]. We suggest that 1 is converted to 6, and this oxidises to 4 with concomitant reduction of anthracene, ultimately producing [K(B15C5)2][C14H11] via proton abstraction from solvent.

Unfortunately, 6/6′/6″ are highly insoluble in non-polar solvents and decompose in polar media so NMR and UV/Vis/NIR data could not be obtained. Complexes 6/6′/6″, as their trivalent formulations suggest, are easily oxidised, and the mother liquor from these reactions always contains variable quantities of 4/4′/4″, respectively, and heating suspensions of 6/6′/6″ in C6D6 in an attempt to obtain 1H NMR spectra results in extraction of 4/4′/4″, respectively. However, the 5f3 uranium(III) formulation of 6 is confirmed by variable-temperature SQUID magnetometry, Fig. 2 and Supplementary Fig. 7, where the magnetic moment of 6 is 3.25 μB at 300 K and this slowly decreases to 2.0 μB at ~20 K and then falls to 1.59 μB at 2 K52,53,54. Furthermore, the X-band EPR spectrum of 6 at 20 K, Fig. 4b, exhibits g values of 4.19, 0.88, and 0.52, from which a magnetic moment of 2.16 μB would be predicted that is in good agreement with the observed magnetic moment of 6 at 20 K. The solid-state structure of 6 has been determined, Fig. 5 and Supplementary Fig. 8, revealing a separated ion pair formulation. The salient feature of 6 is the presence of a UIII-NH2 linkage within the uranium component, which has no precedent in uranium(III) chemistry, as evidenced by a U–Namide distance of 2.335(3) Å, which is longer than analogous UIV-NH2 distances of 2.228(4) Å in 437, 2.217(4) Å in [UIV{η8-C8H6-1,4-(SiPri3)2}(η5-C5Me5)(NH2)]62, and 2.183(6) and 2.204(6) Å in [UIV{η5-C5H2-1,2,4-(But)3}(NH2)2]63. The Tren U–Namine and U–Namide distances of 2.721(2) and 2.385(2)-2.393(2) Å, respectively, reflect the anionic uranium(III) formulation of 6, since, for example, the latter, which are usually quite sensitive to the oxidation state of uranium, are usually ~2.27 Å for uranium(IV) congeners57.

Fig. 5: Molecular structure of the anion component of 6 at 150 K and displacement ellipsoids set to 40%.
figure 5

Nonamide hydrogen atoms and the [K(B15C5)2]+ cation component are omitted for clarity. Selected bond lengths (Å): U1-N1, 2.393(2); U1-N2, 2.404(2); U1-N3, 2.385(2); U1-N4, 2.721(2); U1-N5, 2.335(3).

Full size image

In order to experimentally link 6–4 we treated 6 with one equivalent of BMes3 resulting in immediate reduction of BMes3 to give a 1:1 mixture of 4 and 5, Fig. 1, which is in-line with the reducing nature of 6 as evidenced by ready formation of 4 in supernatant reaction mixtures. These reactions show that although the FLP aspect of the reaction of H2 with 1 certainly facilitates and accelerates the hydrogenolysis of the nitride linkage, it is not essential, and the terminal uranium(V)-nitride linkage is reactive enough in its own right to be hydrogenated with H2 to give a uranium(III)-amide, and this is confirmed computationally (see below).

Ammonia synthesis via strong acid

After the hydrogenolysis reactions that produce 4 and 6 we vacuum transferred volatile materials onto hydrochloric acid, but in each case no more than a 5% yield of NH3, as its conjugate acid NH4+, was detected by standard methods. This suggests that although the UV≡N-nitride linkage reacts with one equivalent of H2 to give UIII/IV-NH2, further reaction of the latter linkages with H2 does not occur. Direct treatment of 4′/4″ with 0.01 M HCl in THF/Et2O, to differentiate the D+ as from D2 and not D+ acid, vacuum transfer onto a 2 M HCl in Et2O acid trap, then assay, revealed a mixture of NH3D+ (2D δ 7.12 ppm) and NH4+ (1H δ 7.28 ppm, 1:1:1 triplet, JNH = 51 Hz) by 1H NMR spectroscopy. Addition of H2O results in full D/H exchange to give NH4+ as the sole ammonium species in 52% yield. Analogously, 6′/6″ produces NH4+ in 46% yield, and if the HCl acid steps are replaced with analogous DCl reagents then NHD3+ is first obtained and when this is converted to NH4+ a similar yield of 48% is obtained showing the internal consistency of this approach, Supplementary Figs. 9, 10. Under the action of strong acid the main by-product is TrenTIPSH3 from over-protonation, but up to 31% [UIV(TrenTIPS)(Cl)] (7)36 could be observed by 1H NMR spectroscopy as would be expected from the reaction of 4 with HCl.

Reversible ammonia-borane formation

Since H2 does not react with 4 or 6 on their own, we examined whether addition of a borane would facilitate a second hydrogenolysis step; utilising BCF or BMes3 with H2 results in no reaction and/or formation of unknown, intractable products. We find, however, that 4 reacts with BPh3 to form the uranium(IV)-amide-borane adduct [UIV(TrenTIPS)(NH2BPh3)] (8), Fig. 6, as evidence by its solid-state structure, Fig. 7. The salient feature of the structure of 8 is that although the Tren U–Namine and U–Namide distances of 2.645(5) and 2.221(5)-2.246(5) Å are unexceptional for Tren-uranium(IV) distances57, the U–NH2 U–Namide distance of 2.578(5) Å is very long64, suggesting that coordination of BPh3 has severely weakened the U–NH2 linkage. However, there is clearly a balance of steric clashing in this region of the molecule since the B-Namide distance of 1.637(9) Å is ~0.06 Å longer than the analogous distance in 3 and ~0.08 Å longer than the sum of the covalent single bond radii of B and N (1.56 Å)56. Variable-temperature SQUID magnetometry on a powdered sample of 8, Fig. 2 and Supplementary Fig. 11, confirms the uranium(IV) formulation of this complex. Specifically, the magnetic moment of 8 at 300 K is 3.17 μB and this decreases smoothly to a value of 0.89 μB at 2 K and is tending to zero52,53,54. This is characteristic of uranium(IV) which is a magnetic singlet at low temperature but that exhibits a small contribution from temperature independent paramagnetism to give a nonzero magnetic moment.

Fig. 6: Reactivity of 4.
figure 6

Treatment of 4 with BPh3 results in formation of the adduct 8. Addition of H2 to 8 produces the hydride 10 with concomitant elimination of H3NBPh3. Complex 10 eliminates H2 to produce the cyclometallate complex 9, which in turn can react with H3NBPh3 to reform 4.

Full size image
Fig. 7: Molecular structure of 8 at 150 K and displacement ellipsoids set to 40%.
figure 7

Nonamide hydrogen atoms and lattice solvent are omitted for clarity. Selected bond lengths (Å): U1-N1, 2.236(5); U1-N2, 2.221(5); U1-N3, 2.246(4); U1-N4, 2.645(5); U1-N5, 2.578(5); B1-N5, 1.637(9).

Full size image

The 11B NMR spectrum of 8 dissolved in C6D6 or C6D5CD3 at 293 K exhibits resonances at 67.5 and −3.2 ppm, corresponding to free BPh3 and H3NBPh3, respectively, as confirmed by comparison to authentic samples. The implication, consistent with the long U–Namide and B–Namide distances in 8, is that 8 is in equilibrium with 4 and free BPh3 in solution by B–N bond cleavage, but also that the long U–Namide bond is weakened increasing the basicity of this amide resulting in its rupture, C–H bond activation, and N–H bond formation to concomitantly form H3NBPh3 and the uranium(IV)-cyclometallate complex [UIV{N(CH2CH2NSiPri3)2(CH2CH2NSiPri2CH(Me)CH2)}] (9)65. Indeed, trace resonances that match reported data65 for 9 could be observed. A variable-temperature 1H and 11B{1H} NMR study (Supplementary Figs. 12, 13) reveals that at 293 K the dominant products are 4 and free BPh3, but as the temperature is lowered to 253 and then 233 K resonances attributable to 8 grow in as 4 diminishes such that at 253 K the ratio of 4:8 is ~2:1 and at 233 K this ratio is ~2:3. However, when 9 is treated with H3NBPh3 the formation of 4 and BPh3 are observed by 1H NMR spectroscopy. This suggests facile, unspecific reversible reactivity but also hints at FLP-type reactivity, so we dissolved a 1:1 mixture of 4 and BPh3 under H2 (1 atm.), but again find only trace quantities of H3NBPh3. If 8 reacts with H2 to form H3NBPh3 and [UIV(TrenTIPS)(H)] (10) the latter would be anticipated to eliminate H2 to give cyclometallate 959,65. Indeed, treating [UIV(TrenTIPS)(THF)][BPh4] with NaHBPh3 to nominally produce [UIV(TrenTIPS)(HBPh3)] gives H2, BPh3, and 9 in addition to the anticipated NaBPh4 by-product. The reaction cycle in Fig. 6 can thus be proposed where 4 reacts with BPh3 and H2 to give, possibly via 8, H3NBPh3 and 10, the latter of which extrudes H2 to give 9. Since it is known that 9 reacts with H3NBPh3 to give 8 and/or 4 and free BPh3 then a cycle is most likely established where reactivity is occurring but no discernable products can be isolated since the products react with one another to give the starting materials. Though of little use currently, the formation of H3NBPh3 suggests that it may be possible to extract out and trap the NH3, though so far this system has resisted attempts to do so.

Closing an ammonia synthesis reaction cycle

Having effected hydrogenolysis of 1 but found that further reaction with H2 either does not occur or seems to occur in a borane-cycle with no discernable products, we sought to close a reaction cycle utilising an electrophile. Accordingly, treatment of 4, either prepared directly from 1/H2/BMes3 or stepwise via 6, with Me3SiCl produces 7 and Me3SiNH2 that can be quantitatively converted to NH3 in the form of ammonium salts. Under nonoptimised conditions an equivalent NH3 yield of 53% was achieved. Thus, a reaction cycle for azide to nitride to amide to ammonia by hydrogenation overall is demonstrated at uranium using hydrogenolysis of H2 followed by an electrophilic elimination and acid quench, Fig. 8.

Fig. 8: Reaction cycle for the production of ammonia.
figure 8

Treatment of 7 with sodium azide, KC8, and two equivalents of B15C5 produces the terminal uranium-nitride 1, which in turn reacts with H2 and BMes3 to give 4. Treatment of 4 with Me3SiCl, followed by work-up and acidification steps, as indicated by the multiple arrows, produces ammonia. B15C5 = benzo-15-crown-5 ether. Mes = 2,4,6-trimethylphenyl.

Full size image

Computational reaction mechanism profiles

In order to understand the reactions that produce 4/5 and 6, DFT calculations (B3PW91) corrected for dispersion- and solvent-effects were carried out to determine possible reaction pathways for the reaction of complex 1 with H2 in the presence or absence of BMes3, Supplementary Tables 1–25. We also computed the reaction profile for the hypothetical reaction of 2 with H2 (Supplementary Fig. 14), which confirms the experimental situation of no observable reactivity of 2 with H2. In the absence of BMes3, Fig. 9, H2 reacts with 1 in a σ-bond metathesis fashion. The associated barrier is relatively low (14.4 kcal mol−1). At the transition state (B), the H–H bond is strongly elongated (1.02 Å) and the N–H bond is not yet formed (1.35 Å). The U–Nnitride bond is 1.84 Å and the U-H distance is long (2.20 Å). The N–H–H angle is 146.3°, which is quite acute for a metathesis reaction. The NPA charges at the transition state (TS) [U, +1.12; N, −0.84; H, +0.23; H, −0.10] indicate that the TS is better described as a proton transfer. Indeed, inspection of the spin densities of 1, the H2-adduct A, and the TS B reveal little spin-depletion at N (−0.12 for 1, −0.13 for A, −0.15 for B) and that the majority of spin density is at uranium (1.19 for 1, 1.18 for A, and 1.24 for B) so N-radical character does not appear to play a significant role in the H2-activation. Following the intrinsic reaction coordinate yields a uranium(V)-imido-hydride complex (C), whose formation is almost athermic (loosely endothermic by 2.0 kcal mol−1). Complex C can rearrange through a H-atom migration from uranium to parent imido group (transition state D), i.e. undergoing a 1,1-migratory insertion, with a reduction of uranium oxidation state at this point from V to III. The associated activation barrier is 32.1 kcal mol−1 from C (34.1 from the start point). The height of this barrier is due to the need of the hydride to be transferred as a proton to the nucleophilic imido group. However, this barrier is kinetically accessible and in-line with the slow reaction observed experimentally. This TS yields trivalent 6 that is thermodynamically stable (−21.0 kcal mol−1). However, in the presence of BMes3, complex 6 can be easily oxidised into tetravalent 4 in a process that can be considered to be an essentially athermic electron transfer process since the computed energy difference between 4 and 6 is within the error of the calculation.

Fig. 9: Computed reaction profile for the conversion of 1 to 6 in the absence of BMes3 and then conversion to 4 with the addition of BMes3.
figure 9

The iso-propyl groups of the silyl substituents, carbon-bound hydrogen atoms, and [K(B15C52)]+ cation accommodated in the calculations are omitted for clarity. Bold numbers without parentheses refer to ΔH values and numbers in parentheses are ΔG values, both quoted in kcal mol−1.

Full size image

In the presence of BMes3, Fig. 10, the computed reaction pathway is quite different. After the formation of a loosely bonded H2 adduct, the system reaches an H2 activation TS, that is reminiscent of FLP complex reactivity. Indeed, at TS 2B, the H2 molecule interacts in a bridging end-on fashion with the nitride (that is the nucleophile of the FLP) and the borane (that is the electrophile). Unlike TS 2B, the H2 molecule is very little activated at 2A (1.02 vs 0.83 Å, respectively), and neither the N–H bond (1.68 Å) nor the B-H one (1.69 Å) are yet fully formed. The U–Nnitride bond distance is similar to that found for 2A (1.81 Å). The associated barrier is relatively low (14.6 kcal mol−1 with respect to the start point) and similar to the σ-bond metathesis mechanism. Therefore, the presence of BMes3 does not impact the protonation of the strongly nucleophilic nitride that is very reactive. Again, there is essentially no spin-depletion at the nitride (−0.12 for 1, −0.13 for 2A, −0.12 for 2B) and the unpaired spin density is clearly localised at uranium (1.19 for 1, 1.20 for 2A, 1.21 for 2B), which argues against nitride radical character in this reactivity. The FLP TS 2B evolves to the formation of a fully dissociative ion pair whose formation is exothermic (−13.9 kcal mol−1 from start point). From the uranium(V)-imido complex 2C, the formation of trivalent 6 then tetravalent 4 was considered. The first, shown by the gray pathway, implies that the hydroborate (HBMes3)− delivers the hydrogen to the imido (2D1). However, this route is not favoured because, like the problem in the absence of BMes3, the hydride has to be transferred as a proton. The computed barrier of 40.4 kcal mol−1 from 2C (26.5 kcal mol−1 from the start point) is in-line with this. The second possibility, shown by the black pathway, involves a second FLP-type activation of H2 (2D2). The associated barrier is 10.5 kcal mol−1 lower than 2D1, demonstrating the beneficial role of BMes3. However, the 2D2 barrier is also higher than the first FLP barrier, indicating that the uranium(V)-imido complex 2C is a less strong nucleophile than the uranium(V)-nitride 1. This also evidenced by the geometry at the TS, where the N–H distance is far shorter than in 2B (1.44 Å vs. 1.68 Å) inducing a shorter B–H distance (1.54 Å vs. 1.69 Å). The resulting more compact geometry enhances steric repulsion that increases the activation barrier. The 2D2 TS yields ultimately tetravalent 4 (via trivalent 6) whose formation is exothermic by 21.3 kcal mol−1.

Fig. 10: Computed reaction profile for the conversion of 1 to 6 and then 4 in the presence of BMes3.
figure 10

The iso-propyl groups of the silyl substituents, carbon-bound hydrogen atoms, and [K(B15C52)]+ cation accommodated in the calculations are omitted for clarity. Bold numbers without parentheses refer to ΔH values and numbers in parentheses are ΔG values, both quoted in kcal mol−1.

Full size image

Discussion

Despite exhaustive attempts, we find no evidence for any reactivity between the uranium(VI)-nitride 2 and H2 irrespective of whether borane promoters are present or not. However, this is not surprising since prior computational studies have suggested that the U≡N triple bond is rather covalent37, possibly even more so than group 6 congeners, and so it conforms to the general phenomenon that many metal-nitrides, and especially high oxidation state electron-poor ones, are exceedingly unreactive. To date, CO, CO2, and CS2 have been found to react with 240,41, but always much more slowly than 1, and these small, polar molecules with relatively low-lying π*-orbitals are considerably easier to activate than apolar H2 that has only a σ*-orbital available for activation.

In contrast, the reactivity of 1 with H2, the mechanisms of which sharply diverge with or without BMes3, is surprising and notable because the 5f1 uranium(V) ion in 1 is high oxidation state and cannot be considered to be electron-rich nor low-coordinate. Indeed, the only example of any molecular uranium-nitride reacting with H2 is the diuranium(IV)-nitride-cesium complex [Cs{U(OSi[OBut]3)3}2(μ-N)];47 here, the product is the bridging parent imido-hydride complex [Cs{U(OSi[OBut]3)3}2(μ-NH)(μ-H)] and this transformation is enabled by the bridging, polar nature of the nitride and polymetallic cooperativity effects. However, that chemistry stops at the imido-hydride stage, or reverts to nitride and H2, and does not proceed to the H-atom 1,1-migratory insertion stage to give an amide. When terminal M≡N triple bonds have been found to react with H2 it is with 4d4 Ru(IV)26 or 5d4 Os(IV)27 to give NH3 and 5d6 Ir(III)28 to give Ir–NH2, since these are the only nitrides that are low-coordinate and sufficiently activated and electron-rich enough to reduce the M≡N triple bond orders by populating anti-bonding interactions. This electron-rich activation is not applicable to 1 being only 5f1 and that f-electron is in principle nonbonding. However, the nitride is a very strong donor ligand that we have previously shown can modulate the mJ groundstate of uranium depending how strongly it can donate. Specifically, the nitride forms σ- and π-bonds with the l = 0 and l = 1 5f-orbitals, but also interacts with the l = 2 and l = 3 5f-orbitals where the 5f-electron must reside and thus the supposedly nonbonding 5f1 electron would seem to be not entirely innocent in this circumstance due to an inevitable anti-bonding interaction38. Nevertheless, that a metal-nitride of oxidation state as high as +5 and valence electron number as low as one is capable of activating H2 without utilising ancillary ligand reactivity and H-atom shuttling is unprecedented1. Since the reaction profile calculations do not support the notion of nitride radical character promoting the observed and unexpected reactivities, we suggest that this is due to a combination of the 5f-electron in 1 not being wholly nonbonding, and also that the uranium(V)-nitride bond is actually more polar than transition metal analogues.

The reaction profile calculations combined with experimental observations provide an internally consistent account of the reactivity reported here. It is clear from experiment that 1 does not bind BMes3, unlike BPh3, presumably on steric grounds presenting the potential for FLP chemistry that is intuitively invoked when considering the steric demands of TrenTIPS and BMes3. When 1 is reacted with H2 in the presence of BMes3 the [U(TrenTIPS)(N)]− and BMes3 components constitute the FLP that can form an encounter complex with H2 and the computed intermediate 2A and TS 2B are clear evidence for a FLP encounter complex, which facilitates the splitting of H2, confirming bona fide FLP reactivity and introducing actinide chemistry to the pantheon of FLP reactivity. Although the conversion of 2C and (HBMes3)− to 6 and BMes3 is thermodynamically favourable, it is kinetically the least feasible route to occur due to the inherent barrier of a hydride being a proton source, and instead it appears that a second FLP activation of H2 occurs along with oxidation of 6 to give 4, which is thermodynamically little different to the previous outcome but kinetically more accessible. Within the error of the calculation the oxidation of 6 to 4 is essentially athermic and likely driven by the strongly reducing nature of the uranium(III) ion in 6 coupled to its electron-rich nature. The two-electron reduction on going from 1 to 6 is entirely consistent with the two-electron redox chemistry of H2, and indeed the one-electron oxidation of 6 to 4 is simply a sacrificial one-electron reduction of BMes3 to BMes3−•.

The importance of two FLP reaction steps in the conversion of 1 to 6 and then 4 underscores the importance of the facilitating role that FLP chemistry plays in the hydrogenation of 1. However, more remarkable is that fact that the FLP component is actually not mandatory for hydrogenolysis of the U≡N triple bond to occur, though its absence does slow the reaction significantly demonstrating the facilitating role of the FLP mechanism since the main origin of this impediment is that formally a proton has to evolve from a hydride. In the absence of an FLP mechanism H2 undergoes a direct 1,2-addition across the UV≡N triple bond to give a H−UV=N−H unit that is reminiscent of the aforementioned reactivity of [Cs{U(OSi[OBut]3)3}2(μ-N)]47 when their terminal vs bridging natures, respectively, are taken into account. The reactivity of 1 is also reminiscent of aspects of recently reported mechanistic studies of the reaction of uranium(III) with water66, and it is germane to note that concerted two-electron redox chemistry at uranium remains a relatively rare phenomena29,36,67 with one-electron processes dominating. The 1,2-addition at 1 is effectively H–H heterolysis to generate H+ and H−, consistent with the polarising nature of the U≡N triple bond. Interestingly, the production of the final UIII-NH2 linkage in 6 from pentavalent 1 by H-atom 1,1-migratory insertion, consistent with the two-electron reducing nature of H2 since nucleophilic nitrides tend to react without changing metal oxidation state, is reminiscent of the reactivity of uranium(VI)-nitrides under photolytic conditions, where by a R3CH/UVIN combination, via a R3C•/UV=N–H intermediate converts to UIV-N(H)CR3, since both involve two-electron reductions at uranium overall37,42,48. The reactivity of 1 with H2 has parallels to the reactivity of the ruthenium(IV)-nitride complex [Ru{N(CH2CH2PBut2)2}(N)] with H2 to give NH326, but with some important differences. The Ru-complex initially reacts with H2 across the Ru-Namide not Ru-Nnitride bond, so like many nitrides when reactivity occurs it is with the ancillary ligand not the metal-nitride linkage itself as is the case with 1. However, the Ru-complex does at a later stage transfer a H-atom from Ru to an imido group to form a Ru-NH2 group like C/2C. In contrast, the iridium(III)-nitride complex [Ir{NC5H3-2,2′-(C[Me] = N-2,6-Pri2C6H3)2}(N)] is reported to undergo concerted reactivity with H2 to directly afford an amide and no prior coordination of H2 to the Ir centre28. Looking more widely to sulfido chemistry, the complex [Ti(η5-C5Me5)2(S)(NC5H5)] reacts with H2 to give the hydrosulfide-hydride [Ti(η5-C5Me5)2(SH)(H)]68,69, providing a parallel to the 1,2-addition of H2 across the U≡N triple bond of 1, but unlike 1 the titanocene reactivity halts at the hydrosulfide-hydride formulation and does not undergo a subsequent H-atom 1,1-migratory insertion since that would require formation of SH2 and formally the unfavourable reduction of titanium(IV) to titanium(II). So, the reactivity of 1 displays similar and divergent reactivity pathways to known transition metal-nitride reactivity, but combines 1,2-addition and 1,1-migratory insertion steps where transition metals tend to execute either 1,2-additions or 1,1-insertions at the M≡E bond, but are not capable of executing both together.

The reactivity of 4 with BPh3 and H2 is notable, though complex because it would seem products react to give reactants, because again it invokes the notion of FLP chemistry whereby weakly coordinated [U(TrenTIPS)]+ and [H2NBPh3]− components are sufficiently activated to cleave H2 to give H3NBPh3. While this is currently of no practical use it demonstrates the potential for further FLP hydrogenolysis chemistry to convert the parent amide to ammonia. However, we have demonstrated a reaction cycle, where azide is converted to nitride, which undergoes hydrogenolysis to amide, and the amide can be quenched by acid to give ammonia. Thus, overall a nitride has been hydrogenated to ammonia, and the experimentally and computationally supported proposed reactivity mechanisms contribute to our wider understanding of the reactivity of uranium-nitrides toward H2 in heterogeneous Haber Bosch and ATF scenarios.

In summary, while the uranium(VI)-nitride 2 is apparently inert with respect to reacting to H2, the uranium(V)-nitride 1 is not, suggesting that the 5f-electron of the latter is not entirely nonbonding and that the nitride imposes a strong ligand field on uranium. The absence of reactivity for 2 is entirely in-line with the lack of reactivity for high oxidation state metal-nitrides generally, but the latter is not and is notable for being neither low-coordinate nor electron-rich, which are the two requirements previously common to all terminal metal-nitrides that react with H2, yet it is reactive. This study reveals two distinct H2-activation mechanisms. When the borane BMes3 (Mes = 2,4,6-trimethylphenyl) is present a FLP mechanism operates where two H2 heterolysis events and a borane reduction step sequentially combine to furnish a UIV-NH2 product, and this, to the best of our knowledge, is the first demonstration of the application of bona fide FLP reactivity to actinide chemistry. When the borane is absent, direct 1,2-addition of H2 across the U≡N triple bond to give a H−UV=N−H intermediate followed by H-atom migration produces a UIII–NH2 product that is easily oxidised to UIV–NH2. The direct hydrogenolysis addition is slower than the FLP-mediated mechanism, demonstrating the facilitating role of FLPs. We find evidence that treating the UIV–NH2 product with BPh3 and H2 produces further FLP hydrogenolysis reactivity, since H3NBPh3 has been detected in reaction mixtures, but this is reversible and produces products that react to give the starting materials. We have demonstrated an azide to nitride to amide to ammonia reaction cycle, supported by overall hydrogenation involving hydrogenolysis and electrophilic quenching steps. Thus, overall a nitride has been converted to ammonia, and the experimentally and computationally supported proposed reactivity mechanisms inform our understanding of the reactivity of uranium-nitrides towards H2 in heterogeneous Haber Bosch and ATF scenarios.

Methods

General

Experiments were carried out under a dry, oxygen-free dinitrogen atmosphere using Schlenk-line and glove-box techniques. All solvents and reagents were rigorously dried and deoxygenated before use. Compounds were variously characterised by elemental analyses, NMR, FTIR, EPR, and UV/Vis/NIR electronic absorption spectroscopies, single crystal X-ray diffraction studies, Evans and SQUID magnetometry methods, and DFT computational methods.

Preparation of [U(TrenTIPS)(NBPh3)][K(B15C5)2] (3)

Toluene (20 ml) was added to a stirring mixture of 1 (0.54 g, 0.37 mmol) and BPh3 (0.09 g, 0.37 mmol). The resulting mixture was stirred for 16 h to afford a brown precipitate. The mixture was briefly heated to reflux and filtered. Volatiles were removed in vacuo. The resulting brown solid subsequently identified as 3 was washed with pentane (3 × 5 ml) and dried in vacuo. Yield of 3: 0.42 g, 66%. X-ray quality crystals were grown in benzene solution at room temperature. Anal. calcd for C79H130BKN5O10Si3U: C, 56.41; H, 7.79; N, 4.16%. Found: C, 56.48; H, 7.82; N, 3.95%. 1H NMR (C6D6, 298 K): δ 39.15 (s, 6 H, CH2), 23.53 (s, 6H, CH2), 10.69 (s, 6H, Ar-H), 9.27 (s, 3H, Ar-H), 7.26–6.96 (br m, 14H, Ar-H), 4.47–4.20 (m, 32H, OCH2), −7.92 (s, 54H, Pri-CH3), −9.12 (s, 9H, Pri-CH). 11B{1H} NMR (C6D6, 298 K): δ 112.8. FTIR: υ/cm−1: 2938 (w), 2914 (w), 2859 (m), 1592 (w), 1503 (m), 1454 (m), 1427 (w), 1405 (w), 1382 (w), 1360 (w), 1331 (w), 1288 (w), 1254 (m), 1217 (m), 1125 (s), 1097 (m), 1064 (m), 1050 (m), 1046 (m), 1009 (w), 993 (w), 934 (s), 881 (m), 852 (m), 836 (m), 795 (m), 781 (m), 721 (s), 703 (s), 664 (m), 640 (m), 609 (m), 566 (m), 550 (m), 511 (w), 501 (w), 465 (m), 448 (m). µeff (Evans method, C6D6, 298 K): 1.96 µB.

Attempted reaction of [U(TrenTIPS)(N)][K(B15C5)2] (1) with H2 and BPh3

A brown solution of 1 (0.040 g, 0.03 mmol) and BPh3 (0.007 g, 0.03 mmol) in C6D6 (0.5 ml) was degassed and exposed to an atmosphere of H2. The brown solution was analysed for 7 days by 1H NMR spectroscopy, after which only the formation of 3 was observed.

Attempted reaction of [U(TrenTIPS)(N)][K(B15C5)2] (1) with BMes3

A colourless solution of BMes3 (0.01 g, 0.03 mmol) in C6D6 (0.5 ml) was added to 1 (0.04 g, 0.03 mmol). The brown solution was analysed 1H NMR spectroscopy, revealing resonances of free BMes3 and 1. 1H NMR (C6D6, 298 K): δ 38.58 (s, 6H, CH2), 16.98 (s, 6H, CH2), 10.03–6.80 (br m, 20H, CH2, OCH2, Ar–H), 6.72 (s, 6H, Ar-HBMes3), 3.85 (s, 20H, OCH2), 2.16 (s, 12H, CH3BMes3), 2.14 (s, 9H, CH3BMes3), −5.61 (s, 9H, Pri-CH), −6.30 (s, 54H, Pri–CH3). 11B{1H} NMR (C6D6, 298 K): δ 76.76.

Reaction of [U(TrenTIPS)(N)][K(B15C5)2] (1) with H2 and BMes3

A brown solution of 1 (0.40 g, 0.28 mmol) and BMes3 (0.10 g, 0.72 mmol) in toluene (20 ml) was degassed and exposed to H2 (1 atm.). The mixture was stirred for 2 days to ensure the complete consumption of the starting material and 5 started to precipitate as a dark blue solid after 1 day of stirring. The supernatant of 5 was removed by filtration. Dark blue 5 was washed with toluene (3 × 10 ml) and dried in vacuo. Yield of 5: 0.18 g, 69%. The volatiles of the supernatant were removed in vacuo yielding an oily brown residue containing 4 as the main uranium product. Yield of 4, based on 1H NMR spectroscopy: 67%. 1H NMR of 4 (C6D6, 298 K): δ 107 (s, 2H, NH2), 31.99 (s, 6H, CH2), 7.92 (s, 6H, CH2), −5.35 (s, 9H, Pri–CH), −5.87 (s, 54H, Pri–CH3). A similar reaction using D2 instead of H2 leads to the formation of 4/4′/4″ (4′, 1H δ 106 ppm, 2H δ 106.8 ppm, 2JHD not resolved; 4″, 1H δ no resonance in the 100–110 ppm region, 2H δ 107.5 ppm). Ammonia liberation after treatment of 4/4′/4″ with 1 equivalent of HCl led the formation of NH3DCl. [B(Mes)3][K(B15C5)2] (5): Anal. calcd for C55H73KO10: C, 69.97; H, 7.79; N, 0%. Found: C, 69.81; H, 7.88; N, 0%. FTIR ν/cm−1: 2906 (w), 2873 (w), 1582 (w), 1503 (m), 1452 (m), 1362 (w), 1331 (w), 1291 (w), 1252 (m), 1240 (m), 1219 (m), 1184 (w), 1123 (m), 1099 (m), 1074 (m), 1044 (m), 1005 (m), 936 (m), 852 (m), 840 (m), 813 (w), 777 (w), 748 (m), 734 (m), 695 (w), 675 (w), 603 (w), 573 (w), 562 (w), 542 (w), 509 (w), 467 (m), 454 (w), 411 (w).

Reaction of [U(TrenTIPS)(N)][K(B15C5)2] (1) with H2 or D2

With H2: a brown solution of 1 (1.16 g, 0.81 mmol) in toluene (20 ml) was degassed and exposed to H2 (1 atm.). The mixture was stirred for 7 days at 15 °C to ensure the complete consumption of the starting material and formation of 6 as a grey solid. The supernatant of the grey solid was removed by filtration. The solid was washed with toluene (3 × 10 ml) and dried in vacuo. Yield of 6: 0.53 g, 45%. The volatiles of the filtrate were removed in vacuo yielding an oily brown residue containing traces of 4, B15C5, and TrenTIPSH3. With D2: a brown solution of 1 (1.03 g, 0.71 mmol) in toluene (20 ml) was degassed and exposed D2 (1 atm.). The mixture was stirred for 7 days at 15 °C to ensure complete consumption of the starting material and formation of 6/6′/6″ as a grey solid. The supernatant of the grey solid was removed by filtration. The solid was washed with toluene (3 × 10 ml) and dried in vacuo. Yield of 6/6′/6″: 0.68 g, 66%. The volatiles of the filtrate were removed in vacuo yielding an oily brown residue containing traces of 4/4′/4″, B15C5, and TrenTIPSH3. X-ray quality crystals of 6 were grown from a 0.069 g/ml solution of 1 in toluene exposed to an atmosphere of H2 for three weeks. Anal. calcd for C61H117KN5O10Si3U: C, 50.81; H, 8.18; N, 4.86%. Found: C, 50.64; H, 8.38; N, 4.95%. FTIR ν/cm−1: 2940 (m), 2916 (w), 2881 (w), 2851 (m), 2814 (w), 1596 (w), 1505 (m), 1456 (m), 1411 (w), 1364 (w), 1348 (w), 1333 (w), 1295 (w), 1272 (w), 1254 (m), 1219 (m), 1125 (s), 1107 (m), 1097 (m), 1078 (m), 1046 (m), 1007 (w), 983 (w), 936 (s), 883 (m), 854 (m), 799 (w), 775 (w), 746 (s), 671 (m, for H2 only), 664 (m), 654 (w), 622 (m), 603 (w), 591 (w), 560 (w), shoulder 544 (w, for D2 only), 536 (w), 530 (w), 505 (m), 467 (w), 458 (w), 440 (m), 424 (w). The insolubility of 6 once isolated precluded the determination of its 1H NMR spectrum, the solution magnetic moment by Evans method, and acquisition of a UV/Vis/NIR electronic absorption spectrum. Heating a suspension of 6 in C6D6 resulted in the observation of resonances that correspond to 4 as evidenced by 1H NMR spectroscopy.

Reaction between [U(TrenTIPS)(NH2)][K(B15C5)2] (6) and BMes3

A colourless solution of BMes3 (0.01 g, 0.03 mmol) in 0.5 ml of C6D6 was added to 6 (0.04 g, 0.03 mmol) resulting to the formation of an intense blue solution characteristic of the formation of the radical anion BMes3•−. Rapidly, dark blue crystals of 5 formed and 1H NMR spectrum revealed the formation of 4 in 52% yield.

Reaction of [UN(TrenTIPS)][K(B15C5)2] (1) with 9,10-dihydroanthracene

A J Youngs-valve NMR tube was charged with 1 (36 mg, 25 µmol) and 9,10-dihydroanthracene (4.5 mg, 25 µmol). C6D6 (0.8 ml) was added and the resulting brown mixture was left to stand. After 10 min a turbid red mixture was observed. After standing for 24 h the resulting brown mixture was analysed by 1H and 2H NMR spectroscopy with the only observable uranium containing product being 4. During that time a small amount of red crystalline material deposited that was identified as [K(B15C5)2][C14H11] by a combination of X-ray diffraction and 1H NMR spectroscopy when redissolved. 1H NMR (C4D8O, 298 K): δ 3.59–3.64, 3.67–3.91, 3.88–3.92 (br, m, 32H, OCH2), 4.41 (s, C = CH), 5.62 (td, 2H, 1,9-Anth-CH, 3JHH = 6.85 Hz, 3JHH = 1.22 Hz), 5.89 (dd, 2H, 4,6-Anth-CH, 3JHH = 8.31 Hz, 3JHH = 1.22 Hz), 6.25 (t, 3JHH = 6.60 Hz, 2,3,7,8-Anth-CH), 6.75–6.85 (br, m, 8H, OCHCH). Resonances for the CH2 group were not observed and are likely obscured by residual d8-THF or crown ether resonances between 3.55 and 3.92 ppm.

Ammonia formation after addition of 1 equivalent of HCl to [U(TrenTIPS)(NHD)] (4′)

Complex 4′ (0.05 g, 0.06 mmol), formed from the reaction of 1 with D2 in the presence of BMes3, was treated with 1.2 ml of a 0.05 M HCl solution in THF/Et2O (0.06 mmol) and stirred for 2 h at room temperature. All volatiles were then vacuum transferred onto a 2 M HCl solution in Et2O (2 ml). Volatiles were removed in vacuo and the resulting white solid was dissolved in 0.6 ml of d6-DMSO to quantify the amount of ammonia present using 1H NMR spectroscopy (quantification using sealed capillary insert of 2,5-dimethylfuran in d6-DMSO)59. Integration of the NH3D+ multiplet (7.30 ppm) revealed 40% NH3DCl. The 2H NMR spectrum revealed the presence of a broad resonance at 7.12 ppm. Addition of 10 µl of H2O gave complete proton/deuterium exchange, as the resonance at 7.12 ppm in the 2H NMR spectrum disappeared and a NH4+ 1:1:1 triplet (7.28 ppm, JNH = 51 Hz) was formed, integration of the triplet revealed 52% NH4Cl.

Ammonia formation after addition of 1 equivalent of HCl to [U(TrenTIPS)(NHD)][K(B15C5)2] (6′)

Complex 6′ (0.03 g, 0.021 mmol) was treated with 2.1 ml of a 0.01 M HCl solution in THF/Et2O (0.02 mmol) and was stirred for 2 h at room temperature. All the volatiles were then vacuum transferred into a 2 M HCl solution in Et2O (2 ml). Volatiles were removed in vacuo and the resulting white solid was dissolved in 0.6 ml of d6-DMSO to quantify the amount of ammonia present using 1H NMR spectroscopy (quantification using sealed capillary insert of 2,5-dimethylfuran in d6-DMSO)59. Analysis of the brown solid residue after distillation of the volatiles revealed the presence of 7 in 1–5% yield with TrenTIPSH3 as main product. Integration of the NH3D+ multiplet (7.30 ppm) revealed 35% NH3DCl. The 2H NMR spectrum revealed the presence of a broad resonance at 7.12 ppm. Addition of 10 µl of H2O gave complete proton/deuterium exchange, as the resonance at 7.12 ppm in the 2H NMR spectrum disappeared and a NH4+ 1:1:1 triplet (7.28 ppm, JNH = 51 Hz) was formed, integration of the triplet revealed 46% NH4Cl.

Ammonia formation after addition of 1 equivalent of DCl to [U(TrenTIPS)(NHD)][K(B15C5)2] (6′)

Complex 6′ (0.04 g, 0.03 mmol) was treated with 2.8 ml of a 0.01 M DCl solution in THF/Et2O (0.03 mmol) and stirred for 2 h at room temperature. All volatiles were then vacuum transferred into a 2 M HCl solution in Et2O (2 ml). Volatiles were removed in vacuo and the resulting white solid was dissolved in 0.6 ml of d6-DMSO to quantify the amount of ammonia present using 1H NMR spectroscopy (quantification using sealed capillary insert of 2,5-dimethylfuran in d6-DMSO)59. Integration of the NHD3+ multiplet (7.37 ppm) revealed 11% NHD3Cl. The 2H NMR spectrum revealed the presence of a broad triplet at 7.24 ppm. Addition of 10 µl of H2O gave complete proton/deuterium exchange, as the resonance at 7.24 ppm in the 2H NMR spectrum disappeared and a NH4+ 1:1:1 triplet (7.28 ppm, JNH = 51 Hz) was formed, integration of the triplet revealed 48% NH4Cl.

Synthesis of [U(TrenTIPS)(NH2BPh3)] (8)

Toluene (20 ml) was added to a stirring mixture of 4 (0.20 g, 0.23 mmol) and BPh3 (0.06 g, 0.23 mmol). The resulting mixture was stirred for a further 16 h to afford a brown precipitate. The mixture was filtered and volatiles were removed in vacuo. X-ray quality crystals grew in the brown oily residue overnight. Crystals were washed with pentane (2 × 5 ml) and dried in vacuo. Yield of 8: 0.16 g, 62%. Anal. calcd for C51H92BN5Si3U: C, 55.26; H, 8.37; N, 6.32%. Found: C, 55.52; H, 8.16; N, 5.91%. NMR spectroscopy reveals that when isolated 8 is dissolved in solution it dissociates to 4 and free BPh3 and also trace H3NBPh3 and 9, but this equilibrium can be manipulated by cooling samples favouring the formation of 8 so a variable-temperature NMR study was performed, see below. The presence of H3NBPh3 could not be unequivocally confirmed in the 1H NMR spectrum due to its low concentration level in a spectrum dominated by paragmagnetic species, but its presence is confirmed by 11B NMR spectroscopy. Trace resonances corresponding to reported data for 9 could be observed65. 11B{1H} NMR (C6D6, 298 K): δ 67.5 (BPh3), −3.20 (H3NBPh3), −55.2 (U-H2NBPh3). FTIR: υ/cm−1: 3293 (w), 3228 (w), 3044 (w), 2938 (m), 2886 (m), 2861 (m), 1590 (w), 1502 (w), 1461 (m), 1428 (m), 1372 (m), 1339 (w), 1316 (w), 1270 (m), 1238 (m), 1166 (w), 1133 (w), 1116 (w), 1047 (m), 1010 (m), 988 (m), 925 (s), 880 (s), 816 (w), 731 (s), 701 (s), 670 (s), 632 (s), 596 (m), 565 (m), 554 (m), 514 (s). µeff (Evans method, C6D6, 298 K): 2.96 µB.

Variable-Temperature NMR study of 8

A brown solution of 4 (0.04 g, 0.05 mmol) in d8-toluene (0.3 ml) was added to BPh3 (0.01 g, 0.05 mmol) in d8-toluene (0.2 ml). The brownish black solution was analysed by 1H and 11B{1H} NMR spectroscopies at 293, 253, and 233 K. Integrations are listed relatively for functional units within a given species, but note at 293 K 8 is fully dissociated to 4, at 253 K the ratio of 4:8 is ~2:1, and at 233 K that ratio is then ~2:3. 1H NMR (C6D5CD3, 293 K): δ 107 (s, 2H, NH2, 4), 31.99 (s, 6H, CH2, 4), 7.92 (s, 6H, CH2, 4), 7.4–6.2 (m, br, 15H, B(C6H5)3), −5.35 (s, 9H, Pri-CH, 4), −5.87 (s, 54H, Pri-CH3, 4). 11B{1H} NMR (C6D5CD3, 298 K): δ 67.5 (BPh3), −3.2 (H3NBPh3). 1H NMR (C6D5CD3, 253 K): δ 148.1 (s, br, 2H, NH2, 4), 43.5 (s, br, 6H, CH2, 4), 11.5 (s, vbr, 63H, Pri-CH and Pri-CH3, 8), 9.2 (s, br, 6H, CH2, 4), 7.5–3.7 (s, vbr, 15H, B(C6H5)3), −7.5 (s, br, 9H, Pri-CH, 4), −8.5 (s, br, 54H, Pri-CH3, 4), −34.5 (s, vbr, 6H, CH2, 8), −44.5 (s, vbr, 6H, CH2, 8), −154.4 (s, vbr, 1H, NH2, 8), −173.6 (s, vbr, 1H, NH2, 8). 11B{1H} NMR (C6D5CD3, 253 K): δ 69.0 (BPh3), −95.8 (U-H2NBPh3, 8). 1H NMR (C6D5CD3, 233 K): δ 171.9 (s, br, 2H, NH2, 4), 50.3 (s, br, 6H, CH2, 4), 16.7 (s, br, 9H, Pri-CH, 8), 15.8 (s, br, 6H, CH2, 4), 10.04 (s, br, 54H, Pri-CH3, 8), 3.7–2.1 (m, br, 15H, H2NB(C6H5)3) −8.5 (s, br, 9H, Pri-CH, 4), −9.5 (s, br, 54H, Pri-CH3, 4), −38.4 (s, br, 6H, CH2, 8), −51.2 (s, br, 6H, CH2, 8), −159.6 (s, br, 1H, NH2, 8), −196.1 (s, br, 1H, NH2, 8). 11B{1H} NMR (C6D5CD3, 233 K): δ −107.9 (U-H2NBPh3, 8).

Reaction between [U(TrenTIPS)(NH2)] (4) and Me3SiCl

Me3SiCl (6 µl, 0.05 mmol) was added to a brown solution of 4 (0.04 g, 0.05 mmol) in benzene (0.5 ml). The mixture was stirred at room temperature for 2 h. All volatiles containing N-silylated products were distilled under reduced pressure and stirred for 12 h into an aqueous solution of H2SO4 (0.5 M, 5 ml) to convert the N-silylated products into ammonium salts70. After the addition of an excess amount of base (aqueous 30% KOH, 5 ml), ammonia was distilled into HCl solution in Et2O (2 M, 2 ml) under reduced pressure. The amount of ammonia was determined by 1H NMR spectroscopy using sealed capillary insert of 2,5-dimethylfuran in d6-DMSO59. Yield NH3: 53%. To the residual solid fraction containing uranium complexes was added ferrocene as an internal standard in C6D6 (0.5 ml) to quantify the amount of 7 formed. Yield of 7: 46%.