Terminal uranium(V)-nitride hydrogenations involving direct addition or Frustrated Lewis Pair mechanisms

Despite their importance as mechanistic models for heterogeneous Haber Bosch ammonia synthesis from dinitrogen and dihydrogen, homogeneous molecular terminal metal-nitrides are notoriously unreactive towards dihydrogen, and only a few electron-rich, low-coordinate variants demonstrate any hydrogenolysis chemistry. Here, we report hydrogenolysis of a terminal uranium(V)-nitride under mild conditions even though it is electron-poor and not low-coordinate. Two divergent hydrogenolysis mechanisms are found; direct 1,2-dihydrogen addition across the uranium(V)-nitride then H-atom 1,1-migratory insertion to give a uranium(III)-amide, or with trimesitylborane a Frustrated Lewis Pair (FLP) route that produces a uranium(IV)-amide with sacrificial trimesitylborane radical anion. An isostructural uranium(VI)-nitride is inert to hydrogenolysis, suggesting the 5f1 electron of the uranium(V)-nitride is not purely non-bonding. Further FLP reactivity between the uranium(IV)-amide, dihydrogen, and triphenylborane is suggested by the formation of ammonia-triphenylborane. A reactivity cycle for ammonia synthesis is demonstrated, and this work establishes a unique marriage of actinide and FLP chemistries.

T erminal metal-nitrides, M≡N, represent a key fundamental class of metal-ligand linkage in coordination chemistry 1 . Although these M≡N triple bonds have been of elementary interest for over 170 years 2 , only in relatively recent times has there been a concerted effort to study their reactivity 1,3 . However, although a variety of reactivity patterns have emerged with metalnitrides 1 , 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, N 3− -renders them inherently inert 1,3 . 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 prepare 4 . Additionally, reactivity of metalnitrides 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 reactivity 1,5 .
Since there is an isoelectronic relationship between the M≡N and N≡N triple bonds of metal-nitrides and dinitrogen, N 2 , 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, NH 3 , by hydrogenolysis with dihydrogen, H 2 6,7 . There has thus been intense interest in the reactivity of metal-nitrides with H 2 , and indeed their use in N-atom transfer reactivity and catalysis more widely [8][9][10][11][12] , but there are few reports of molecular metal-nitrides reacting with H 2 , and indeed activating H 2 in this homogeneous context remains a significant challenge in contrast to heterogeneous Haber Bosch chemistry where H 2 -cleavage is essentially barrier-less 6 . One solution to overcome this hydrogenolysis challenge may be to exploit Frustrated Lewis Pair (FLP) chemistry 13,14 , but so far this has been focussed on M-N 2 complexes 15,16 . Usually with mid-or latetransition metals 3 , most metal-nitride hydrogenations involve sequential protonations [17][18][19][20][21][22] , but bridging nitrides in poly-iron/titanium/-zirconium complexes have been reported to react with H 2 to give imido-hydride and NH 3 products [23][24][25] . Only three terminal metal-nitrides have been reported to undergo hydrogenolysis with H 2 . The isostructural d 4 ruthenium(IV)-and osmium(IV)-nitrides [M{N(CH 2 CH 2 PBu t 2 ) 2 }(N)] (M = Ru, Os) react with H 2 using the ancillary ligand to shuttle H-atoms to evolve NH 3 26,27 , and the 5d 6 iridium(III)-nitride [Ir{NC 5 H 3 -2,2′-(C[Me]=N-2,6-Pr i 2 C 6 H 3 ) 2 }(N)] undergoes concerted reactivity with H 2 to give [Ir{NC 5 H 3 -2,2′-(C[Me]=N-2,6-Pr i 2 C 6 H 3 ) 2 } (NH 2 )] 28 . Thus, direct hydrogenolysis of a M≡N triple bond with H 2 remains exceedingly rare, and involves reasonably electronrich (≥d 4 ) metal complexes with low coordination numbers.
Here, we report that 2 does not react with H 2 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 H 2 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 5f 1 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 H 2activation mechanisms. When the borane BMes 3 (Mes = 2,4,6trimethylphenyl) is present a FLP mechanism operates where two H 2 heterolysis events and a borane reduction step sequentially combine to furnish a U IV -NH 2 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 H 2 across the U≡N triple bond to give a H−U V = N−H intermediate followed by H-atom migration produces a U III -NH 2 product that is easily oxidised to U IV -NH 2 . The direct addition is slower than the FLP-mediated mechanism, demonstrating the facilitating role of FLPs. We find evidence that treating the U IV -NH 2 product with BPh 3 and H 2 produces further FLP hydrogenolysis reactivity, since H 3 NBPh 3 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 H 2 , treatment of 1 in toluene with the strong Lewis acid B(C 6 F 5 ) 3 (BCF) results in decomposition as evidenced by 19 F 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 BCF 50 , and so we examined the reaction of 1 with the less Lewis acidic BPh 3 . However, when 1 is treated with BPh 3 in toluene the adduct complex [U V (Tren TIPS )(NBPh 3 )][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)-nitridoborane, is rapidly formed quantitatively and isolated in crystalline form in 66% yield, Fig. 1.
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 2 F 5/2 to 2 F 7/2 transitions of uranium(V) 38,51 , and by variabletemperature SQUID magnetometry, Fig. 2 and Supplementary  Fig. 2. A powdered sample of 3 returns a magnetic moment of 2.23 μ B at 300 K (1.96 μ B by solution Evans method) that changes little until 30 K where it falls quickly to a moment of 1.38 μ B at 2 K and this is consistent with the magnetic doublet character of 5f 1 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 BPh 3 unit. The U-N imido bond length of 1.911(6) Å is consistent with the imidoborate formulation, for example distances of 1.916 (4) 37 , and [U V (Tren TIPS )(NAd)] (Ad = 1-adamantyl) 37 , respectively, and the B-N imido 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-N amine 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-N amide distances (2.254(7)-2.312(6) Å) are slightly long for such distances 57 , reflecting the formal anionic nature of the uranium component of 3.
Complex 3 does not react with H 2 (1 atm.), Fig. 1. Indeed, dissolving a mixture of 1 and BPh 3 under H 2 only generates 3, and so since BPh 3 has shut all reactivity down by strongly binding to the nitride of 1, but BCF is too reactive, we examined the use of BMes 3 (Mes = 2,4,6-trimethylphenyl). In principle, the orthomethyls 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:BMes 3 in toluene under an atmosphere of N 2 and find no evidence for any adduct formation, Fig. 1, with only free BMes 3 being observed as evidenced by a resonance at 76.8 ppm in the 11 B NMR spectrum of the reaction mixture. Repeating this reaction, but under H 2 (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 [U IV (Tren TIPS )(NH 2 )] (4) in 67% yield, Fig. 1, as evidenced by a resonance at 107 ppm in its 1 H NMR spectrum that corresponds to the amide protons 37 . 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 H 2 , and not adventitious H +58 , the reaction was repeated under D 2 (1 atm., 99.8% atom D). Interestingly, whilst [U IV (Tren TIPS )(ND 2 )] (4″, 2 H δ 107.5 ppm) is formed, confirming that hydrogenolysis by H 2 /D 2 does indeed occur, it is always accompanied by 4 and [U IV (Tren TIPS )(NHD)] (4′, 1 H δ 106 ppm, 2 H δ 106.8 ppm, 2 J HD 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 BMes 3 with all combinations of H 2 /D 2 with H 6 -/D 6 -benzene and H 8 -/D 8 -toluene (see Supplementary Methods). We find that when H 2 is used only 4 is ever detected, but when D 2 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 C 6 H 6 as solvent for the reaction of 1 with BMes 3 and D 2 a weak resonance is observed at −5.2 ppm in the 2    We therefore suggest that the H-source is the Tren TIPS Pr i groups since they have precedent for forming cyclometallates 57 , a reversible amide/imido-cyclometallate + H 2 equilibrium can be envisaged since it has been previously shown that uranium-Trencyclometallates can react reversibly with H 2 /D 2 59 , and this would also account for the absence of D-scrambling into 4 since Tren TIPS 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 ] [BMes 3 ] (5), in 69% yield. This compound has been structurally characterised by single crystal diffraction, see Supplementary Since 1 does not form an adduct with BMes 3 , but the introduction of H 2 leads to hydrogenolysis to give the amide 4, we surmised that the 1/BMes 3 mixture may constitute a Frustrated Lewis Pair (FLP) system that is evidently capable of activating H 2 , which is confirmed computationally (see below). However, since H 2 can be a two-electron reducing agent and uranium(V) is normally quite oxidising, we hypothesised that the BMes 3 may not actually be required. In effect, when 4 is produced it is essentially at the expense of the sacrificial one-electron reduction of BMes 3 to BMes 3 •− , which would formally invoke a uranium(III)-amide precursor that would be nicely in-line with a H 2 -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 H 2 (1 atm.) was stirred without BMes 3 . Over seven days 1 is consumed with concomitant precipitation of a gray solid identified as the uranium(III)-amide [U III (Tren TIPS )(NH 2 )][K(B15C5) 2 ] (6) (45% yield), Fig. 1. The hydrogenolysis reaction is now slower than when BMes 3 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 D 2 (1 atm.), a mix of 6, [U III (Tren TIPS )(NHD)][K(B15C5) 2 ] (6′) and [U III (Tren TIPS )(ND 2 )][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 H 2 . Consistent with these observations, we find that 1 also reacts with 9,10-dihydroanthracene (pKa 31 in DMSO, cf 34±4 for H 2 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 1 H NMR spectroscopy and X-ray  We suggest that 1 is converted to 6, and this oxidises to 4 with concomitant reduction of anthracene, ultimately producing [K(B15C5) 2 ][C 14 H 11 ] 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 C 6 D 6 in an attempt to obtain 1 H NMR spectra results in extraction of 4/4′/4″, respectively. However, the 5f 3 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 μ B at 300 K and this slowly decreases to 2.0 μ B at~20 K and then falls to 1.59 μ B at 2 K 52-54 . 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 μ B 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 U III -NH 2 linkage within the uranium component, which has no precedent in uranium(III) chemistry, as evidenced by a U-N amide distance of 2.335(3) Å, which is longer than analogous U IV -NH 2 distances of 2.228(4) Å in 4 37 , 2.217(4) Å in [U IV {η 8 -C 8 H 6 -1,4-(SiPr i 3 ) 2 }(η 5 -C 5 Me 5 )(NH 2 )] 62 , and 2.183(6) and 2.204(6) Å in [U IV {η 5 -C 5 H 2 -1,2,4-(Bu t ) 3 }(NH 2 ) 2 ] 63 . The Tren U-N amine and U-N amide 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 usuallỹ 2.27 Å for uranium(IV) congeners 57 .
In order to experimentally link 6-4 we treated 6 with one equivalent of BMes 3 resulting in immediate reduction of BMes 3 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 H 2 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 H 2 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 NH 3 , as its conjugate acid NH 4 + , was detected by standard methods. This suggests that although the U V ≡N-nitride linkage reacts with one equivalent of H 2 to give U III/IV -NH 2 , further reaction of the latter linkages with H 2 does not occur. Direct treatment of 4′/4″ with 0.01 M HCl in THF/Et 2 O, to differentiate the D + as from D 2 and not D + acid, vacuum transfer onto a 2 M HCl in Et 2 O acid trap, then assay, revealed a mixture of NH 3 D + ( 2 D δ 7.12 ppm) and NH 4 + ( 1 H δ 7.28 ppm, 1:1:1 triplet, J NH = 51 Hz) by 1 H NMR spectroscopy. Addition of H 2 O results in full D/H exchange to give NH 4 + as the sole ammonium species in 52% yield. Analogously, 6′/6″ produces NH 4 + in 46% yield, and if the HCl acid steps are replaced with analogous DCl reagents then NHD 3 + is first obtained and when this is converted to NH 4 + 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 Tren TIPS H 3 from over-protonation, but up to 31% [U IV (Tren TIPS )(Cl)] (7) 36 could be observed by 1 H NMR spectroscopy as would be expected from the reaction of 4 with HCl.
Reversible ammonia-borane formation. Since H 2 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 BMes 3 with H 2 results in no reaction and/or formation of unknown, intractable products. We find, however, that 4 reacts with BPh 3 to form the uranium(IV)-amide-borane adduct [U IV (Tren TIPS )(NH 2 BPh 3 )] (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-N amine and U-N amide distances of 2.645(5) and 2.221(5)-2.246(5) Å are unexceptional for Tren-uranium(IV) distances 57 , the U-NH 2 U-N amide distance of 2.578(5) Å is very long 64 , suggesting that coordination of BPh 3 has severely weakened the U-NH 2 linkage. However, there is clearly a balance of steric clashing in this region of the molecule since the B-N amide 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 μ B and this decreases smoothly to a value of 0.89 μ B at 2 K and is tending to zero [52][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.
The 11 B NMR spectrum of 8 dissolved in C 6 D 6 or C 6 D 5 CD 3 at 293 K exhibits resonances at 67.5 and −3.2 ppm, corresponding to free BPh 3 and H 3 NBPh 3 , respectively, as confirmed by comparison to authentic samples. The implication, consistent with the long U-N amide and B-N amide distances in 8, is that 8 is in  gives H 2 , BPh 3 , and 9 in addition to the anticipated NaBPh 4 by-product. The reaction cycle in Fig. 6 can thus be proposed where 4 reacts with BPh 3 and H 2 to give, possibly via 8, H 3 NBPh 3 and 10, the latter of which extrudes H 2 to give 9. Since it is known that 9 reacts with H 3 NBPh 3 to give 8 and/or 4 and free BPh 3 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 H 3 NBPh 3 suggests that it may be possible to extract out and trap the NH 3 , 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 H 2 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/H 2 /BMes 3 or stepwise via 6, with Me 3 SiCl produces 7 and Me 3 SiNH 2 that can be quantitatively converted to NH 3 in the form of ammonium salts. Under nonoptimised conditions an equivalent NH 3 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 H 2 followed by an electrophilic elimination and acid quench, Fig. 8.
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 H 2 in the presence or absence of BMes 3 , Supplementary Tables 1-25. We also computed the reaction profile for the hypothetical reaction of 2 with H 2 (Supplementary Fig. 14), which confirms the experimental situation of no observable reactivity of 2 with H 2 . In the absence of BMes 3 ,  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 H 2 -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 BMes 3 , 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.
In the presence of BMes 3 , Fig. 10, the computed reaction pathway is quite different. After the formation of a loosely bonded H 2 adduct, the system reaches an H 2 activation TS, that is reminiscent of FLP complex reactivity. Indeed, at TS 2B, the H 2 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 H 2 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-N nitride 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 BMes 3 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 (HBMes 3 ) − delivers the hydrogen to the imido (2D 1 ). However, this route is not favoured because, like the problem in the absence of BMes 3 , 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 H 2 (2D 2 ). The associated barrier is 10.5 kcal mol −1 lower than 2D 1 , demonstrating the beneficial role of BMes 3 . However, the 2D 2    The resulting more compact geometry enhances steric repulsion that increases the activation barrier. The 2D 2 TS yields ultimately tetravalent 4 (via trivalent 6) whose formation is exothermic by 21.3 kcal mol −1 .

Discussion
Despite exhaustive attempts, we find no evidence for any reactivity between the uranium(VI)-nitride 2 and H 2 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 covalent 37 , 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, CO 2 , and CS 2 have been found to react with 2 40,41 , 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 H 2 that has only a σ*-orbital available for activation.
In contrast, the reactivity of 1 with H 2 , the mechanisms of which sharply diverge with or without BMes 3 , is surprising and notable because the 5f 1 uranium(V) ion in 1 is high oxidation state and cannot be considered to be electron-rich nor lowcoordinate. Indeed, the only example of any molecular uraniumnitride reacting with H 2 is the diuranium(IV)-nitride-cesium complex [Cs{U(OSi[OBu t ] 3 ) 3 } 2 (μ-N)]; 47 here, the product is the bridging parent imido-hydride complex [Cs{U(OSi[OBu t ] 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 H 2 , 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 H 2 it is with 4d 4 Ru(IV) 26 or 5d 4 Os(IV) 27 to give NH 3 and 5d 6 Ir(III) 28 to give Ir-NH 2 , 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 5f 1 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 m J groundstate of uranium depending how strongly it can donate. Specifically, the nitride forms σand 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 5f 1 electron would seem to be not entirely innocent in this circumstance due to an inevitable anti-bonding interaction 38 . Nevertheless, that a metalnitride of oxidation state as high as +5 and valence electron number as low as one is capable of activating H 2 without utilising ancillary ligand reactivity and H-atom shuttling is unprecedented 1 . 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 BMes 3 , unlike BPh 3 , presumably on steric grounds presenting the potential for FLP chemistry that is intuitively invoked when considering the steric demands of Tren TIPS and BMes 3 . When 1 is reacted with H 2 in the presence of BMes 3 the [U(Tren TIPS )(N)] − and BMes 3 components constitute the FLP that can form an encounter complex with H 2 and the computed intermediate 2A and TS 2B are clear evidence for a FLP encounter complex, which facilitates the splitting of H 2 , confirming bona fide FLP reactivity and introducing actinide chemistry to the pantheon of FLP reactivity. Although the conversion of 2C and (HBMes 3 ) − to 6 and BMes 3 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 H 2 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 H 2 , and indeed the one-electron oxidation of 6 to 4 is simply a sacrificial one-electron reduction of BMes 3 to BMes 3 −• .
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 H 2 undergoes a direct 1,2-addition across the U V ≡N triple bond to give a H−U V =N−H unit that is reminiscent of the aforementioned reactivity of [Cs{U(OSi[OBu t ] 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 water 66 , and it is germane to note that concerted two-electron redox chemistry at uranium remains a relatively rare phenomena 29,36,67 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 U III -NH 2 linkage in 6 from pentavalent 1 by H-atom 1,1-migratory insertion, consistent with the two-electron reducing nature of H 2 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 R 3 CH/U VI N combination, via a R 3 C•/U V =N-H intermediate converts to U IV -N(H)CR 3 , since both involve two-electron reductions at uranium overall 37,42,48 . The reactivity of 1 with H 2 has parallels to the reactivity of the ruthenium(IV)-nitride complex [Ru{N (CH 2 CH 2 PBu t 2 ) 2 }(N)] with H 2 to give NH 3 26 , but with some important differences. The Ru-complex initially reacts with H 2 across the Ru-N amide not Ru-N nitride 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 Rucomplex does at a later stage transfer a H-atom from Ru to an imido group to form a Ru-NH 2 group like C/2C. In contrast, the iridium(III)-nitride complex [Ir{NC 5 H 3 -2,2′-(C[Me] = N-2,6-Pr i 2 C 6 H 3 ) 2 }(N)] is reported to undergo concerted reactivity with H 2 to directly afford an amide and no prior coordination of H 2 to the Ir centre 28 . Looking more widely to sulfido chemistry, the complex [Ti(η 5 -C 5 Me 5 ) 2 (S)(NC 5 H 5 )] reacts with H 2 to give the hydrosulfide-hydride [Ti(η 5 -C 5 Me 5 ) 2 (SH)(H)] 68,69 , providing a parallel to the 1,2-addition of H 2 across the U≡N triple bond of 1, but unlike 1 the titanocene reactivity halts at the hydrosulfidehydride formulation and does not undergo a subsequent H-atom 1,1-migratory insertion since that would require formation of SH 2 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 BPh 3 and H 2 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(Tren TIPS )] + and [H 2 NBPh 3 ] − components are sufficiently activated to cleave H 2 to give H 3 NBPh 3 . 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 H 2 in heterogeneous Haber Bosch and ATF scenarios.
In summary, while the uranium(VI)-nitride 2 is apparently inert with respect to reacting to H 2 , 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 H 2 , yet it is reactive. This study reveals two distinct H 2 -activation mechanisms. When the borane BMes 3 (Mes = 2,4,6-trimethylphenyl) is present a FLP mechanism operates where two H 2 heterolysis events and a borane reduction step sequentially combine to furnish a U IV -NH 2 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 H 2 across the U≡N triple bond to give a H−U V =N−H intermediate followed by H-atom migration produces a U III -NH 2 product that is easily oxidised to U IV -NH 2 .
The direct hydrogenolysis addition is slower than the FLPmediated mechanism, demonstrating the facilitating role of FLPs. We find evidence that treating the U IV -NH 2 product with BPh 3 and H 2 produces further FLP hydrogenolysis reactivity, since H 3 NBPh 3 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 H 2 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(Tren TIPS )(NBPh 3 )][K(B15C5) 2 ] (3)
. Toluene (20 ml) was added to a stirring mixture of 1 (0.54 g, 0.37 mmol) and BPh 3 (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 C 79 H 130 BKN 5 (20 ml) was degassed and exposed to H 2 (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 Tren TIPS H 3 . With D 2 : a brown solution of 1 (1.03 g, 0.71 mmol) in toluene (20 ml) was degassed and exposed D 2 (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 Tren-TIPS H 3 . X-ray quality crystals of 6 were grown from a 0.069 g/ml solution of 1 in toluene exposed to an atmosphere of H 2 for three weeks. Anal. calcd for C 61  664 (m), 654 (w), 622 (m), 603 (w), 591 (w), 560 (w), shoulder 544 (w, for D 2 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 1 H 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 C 6 D 6 resulted in the observation of resonances that correspond to 4 as evidenced by 1 H NMR spectroscopy.
Reaction between [U(Tren TIPS )(NH 2 )][K(B15C5) 2 ] (6) and BMes 3 . A colourless solution of BMes 3 (0.01 g, 0.03 mmol) in 0.5 ml of C 6 D 6 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 BMes 3 •− . Rapidly, dark blue crystals of 5 formed and 1 H NMR spectrum revealed the formation of 4 in 52% yield.
Reaction of [UN(Tren TIPS )][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). C 6 D 6 (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 1 H and 2 H 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 ][C 14 H 11 ] by a combination of X-ray diffraction and 1 H NMR spectroscopy when redissolved. 1  Volatiles were removed in vacuo and the resulting white solid was dissolved in 0.6 ml of d 6 -DMSO to quantify the amount of ammonia present using 1 H NMR spectroscopy (quantification using sealed capillary insert of 2,5-dimethylfuran in d 6 -DMSO) 59 . Integration of the NH 3 D + multiplet (7.30 ppm) revealed 40% NH 3 DCl. The 2 H NMR spectrum revealed the presence of a broad resonance at 7.12 ppm. Addition of 10 µl of H 2 O gave complete proton/deuterium exchange, as the resonance at 7.12 ppm in the 2 H NMR spectrum disappeared and a NH 4 + 1:1:1 triplet (7.28 ppm, J NH = 51 Hz) was formed, integration of the triplet revealed 52% NH 4 Cl.
Ammonia formation after addition of 1 equivalent of HCl to [U(Tren TIPS ) (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/Et 2 O (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 Et 2 O (2 ml). Volatiles were removed in vacuo and the resulting white solid was dissolved in 0.6 ml of d 6 -DMSO to quantify the amount of ammonia present using 1 H NMR spectroscopy (quantification using sealed capillary insert of 2,5-dimethylfuran in d 6 -DMSO) 59 . Analysis of the brown solid residue after distillation of the volatiles revealed the presence of 7 in 1-5% yield with Tren TIPS H 3 as main product. Integration of the NH 3 D + multiplet (7.30 ppm) revealed 35% NH 3 DCl. The 2 H NMR spectrum revealed the presence of a broad resonance at 7.12 ppm. Addition of 10 µl of H 2 O gave complete proton/deuterium exchange, as the resonance at 7.12 ppm in the 2 H NMR spectrum disappeared and a NH 4 + 1:1:1 triplet (7.28 ppm, J NH = 51 Hz) was formed, integration of the triplet revealed 46% NH 4 Cl.
Ammonia formation after addition of 1 equivalent of DCl to [U(Tren TIPS ) (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/Et 2 O (0.03 mmol) and stirred for 2 h at room temperature. All volatiles were then vacuum transferred into a 2 M HCl solution in Et 2 O (2 ml). Volatiles were removed in vacuo and the resulting white solid was dissolved in 0.6 ml of d 6 -DMSO to quantify the amount of ammonia present using 1 H NMR spectroscopy (quantification using sealed capillary insert of 2,5-dimethylfuran in d 6 -DMSO) 59 . Integration of the NHD 3 + multiplet (7.37 ppm) revealed 11% NHD 3 Cl. The 2 H NMR spectrum revealed the presence of a broad triplet at 7.24 ppm. Addition of 10 µl of H 2 O gave complete proton/deuterium exchange, as the resonance at 7.24 ppm in the 2 H NMR spectrum disappeared and a NH 4 + 1:1:1 triplet (7.28 ppm, J NH = 51 Hz) was formed, integration of the triplet revealed 48% NH 4 Cl.
Synthesis of [U(Tren TIPS )(NH 2 BPh 3 )] (8). Toluene (20 ml) was added to a stirring mixture of 4 (0.20 g, 0.23 mmol) and BPh 3 (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 C 51 H 92 BN 5 Si 3 U: 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 BPh 3 and also trace H 3 NBPh 3 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 H 3 NBPh 3 could not be unequivocally confirmed in the 1 H NMR spectrum due to its low concentration level in a spectrum dominated by paragmagnetic species, but its presence is confirmed by 11 B NMR spectroscopy. Trace resonances corresponding to reported data for 9 could be observed 65  Variable-Temperature NMR study of 8. A brown solution of 4 (0.04 g, 0.05 mmol) in d 8 -toluene (0.3 ml) was added to BPh 3 (0.01 g, 0.05 mmol) in d 8 -toluene (0.2 ml). The brownish black solution was analysed by 1 H and 11 B{ 1 H} 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. 1  Reaction between [U(Tren TIPS )(NH 2 )] (4) and Me 3 SiCl. Me 3 SiCl (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 Nsilylated products were distilled under reduced pressure and stirred for 12 h into an aqueous solution of H 2 SO 4 (0.5 M, 5 ml) to convert the N-silylated products into ammonium salts 70 . After the addition of an excess amount of base (aqueous 30% KOH, 5 ml), ammonia was distilled into HCl solution in Et 2 O (2 M, 2 ml) under reduced pressure. The amount of ammonia was determined by 1 H NMR spectroscopy using sealed capillary insert of 2,5-dimethylfuran in d 6 -DMSO 59 . Yield NH 3 : 53%. To the residual solid fraction containing uranium complexes was added ferrocene as an internal standard in C 6 D 6 (0.5 ml) to quantify the amount of 7 formed. Yield of 7: 46%.

Data availability
The X-ray crystallographic coordinates for the structures of 3, 5, 6, and 8 reported in this study have been deposited at the Cambridge Crystallographic Data Centre (CCDC), under deposition numbers 1870831-1870834 and 1936479. These data can be obtained free of charge from The Cambridge Crystallographic Data Centre via www.ccdc.cam.ac. uk/data_request/cif. 1