The development of zero-carbon fuel technology is essential for decreasing dependence on fossil fuels and mitigating CO₂ emissions.¹ A promising avenue within this field is the ′nitrogen economy,′ which focuses on nitrogen-based fuels for large-scale energy storage solutions.^1c-1e Among these, ammonia (NH₃) is particularly notable due to its high energy density, ease of storage, and versatility as both a hydrogen carrier and fuel.^1c-1e Advancing the ammonia economy hinges on improving the synthesis and utilization of NH₃.¹ The development of efficient catalysts is crucial for synthesizing ammonia from nitrogen (N₂) and hydrogen (H₂), potentially making the energy-intensive Haber-Bosch process more efficient by emulating the biological nitrogen fixation performed by nitrogenase enzymes.² Additionally, developing catalysts for the oxidation of ammonia is critical to effectively harnessing the energy stored in N−H bonds.^1a However, this process faces a number of challenges, including the stability of Werner-ammine coordination complexes,³ and the high dissociation energy of the N−H bond (416 kJ mol⁻¹).⁴

Recent research has focussed productively on transition metal-mediated N−H bond cleavage reactions through various strategies, including coordination-induced bond weakening, hydrogen atom abstraction, inter- and intramolecular deprotonation, oxidative addition, σ-bond metathesis, and electrochemical processes to convert NH₃ into N₂ and H₂.^{1a, 1b, 4, 5} These advances are vital for developing sustainable energy storage technologies and realizing a functional nitrogen/ammonia economy.

In contrast to d-block systems, however, the exploration of molecular main group complexes for the conversion of NH₃ to H₂ remains relatively under-explored, despite the fact that NH₃ activation by low-valent main group species has been known for nearly 20 years (Scheme 1a).^{1a, 6} Notable examples include carbenes,^{6a, 7} low-oxidation-state compounds of the heavier Group 13 and 14 elements,⁸ multiply bonded species,⁹ geometrically constrained and low-valent Group 15 compounds,¹⁰ and even frustrated Lewis pairs (FLPs).¹¹ More recently, single-electron transfer from a dithiolene zwitterion,¹² or from a Bi(II) complex has been reported in the context of NH₃ activation,¹³ although these methods are limited in their ability to subsequently transfer the primary amide (−NH₂) moiety. In a recent discovery, Breher et al. demonstrated NH₃ splitting using an ambiphilic system and subsequent catalytic transfer to a range of organic substrates, marking a significant advance in the field.¹⁴

Aluminium, the most abundant metal in Earth's crust, has shown recent promise in small molecule activation, with both neutral and anionic Al(I) complexes shown to be capable of activating species such as H₂, CO, CO₂ and N₂O.^{6d, 15} However, NH₃ activation by low-valent aluminium compounds remains rare. To date, only one example of NH₃ oxidative cleavage - effected by a cyclic(alkyl)(amino)aluminyl system - has been reported.^8k Recently, we expanded the range of ligand frameworks available for supporting Al(I) compounds: N-heterocyclic boryloxy (NHBO)-ligated aluminyl compounds could be accessed by ligand metathesis and shown to be competent in the reversible [4+1] cycloaddition of benzene.¹⁶ Here, we present NHBO-aluminyl reactivity towards ammonia, including the first example of a molecular main group compound achieving dual ammonia activation and conversion of NH₃ to H₂. This chemistry also facilitates onward mono and bis(amide) (−NH₂) transfer to Ph₂CO, CO₂ and HBpin. In the case of benzophenone, selective amide transfer leads to a rare crystallographically characterized (metal-masked) carbinolamine,¹⁷ a key class of intermediate in imine formation from an amine and an organic carbonyl compound.

Exposure of a C₆D₆ solution of K[{(HCDippN)₂BO}₂Al] (1) to NH₃ at 25 °C initially results in the formation of K[{(HCDippN)₂BO}₂Al(H)(NH₂)] (2) through the insertion of Al(I) into the N−H bond of ammonia (Scheme 2). In situ ¹H NMR spectroscopy confirms the generation of 2, with a signal at δ_H=−1.79 ppm (2H) being diagnostic of an Al−NH₂ moiety.¹⁸ Subsequently, 2 is transformed over 0.5 h by reaction with a second equivalent of NH₃, resulting in the formation of the aluminium bis(amide) complex K[{(HCDippN)₂BO}₂Al(NH₂)₂] (3) and H₂. In this case in situ ¹H NMR analysis confirms the generation of H₂ (δ_H=4.47 ppm, Figure S7); similar H₂ evolution has previously been reported in the reactions between aluminium hydrides of the type [ArAlH₂]₂ and ammonia.^18a

Compound 3 can be isolated in 77 % yield as a white powder and single crystals could be obtained from benzene solution. 3 was fully characterized by multinuclear NMR spectroscopy, elemental microanalysis, and X-ray diffraction. In the ¹H NMR spectrum, a 4H signal at −2.15 ppm is assigned to the aluminium-bound NH₂ group, i.e. downfield of previously reported aluminium primary amide complexes (δ_H=−0.68 to −0.22 ppm).¹⁸ The molecular structure of compound 3 in the solid state (Figure 1) reveals a dinuclear arrangement featuring two inequivalent tetra-coordinate aluminium centres. Each aluminium centre is bonded to two NHBO ligands and two NH₂ fragments. The Al−N bond lengths, ranging from 1.797(4) to 1.821(4) Å, are similar to those measured previously for Al(III) bis(amide) complexes (1.788(2)–1.790(2) Å).^{18b, 19} Notably, the two K⁺ cations in 3 exhibit distinct coordination environments: one is encapsulated between two oxygen atoms (thereby constituting a 4-membered AlO₂K core), while the other is coordinated by four NH₂ units.

The presence of two −NH₂ groups at each metal centre in 3 is consistent with the activation of two equivalents of ammonia at each aluminium. As such, it represents the first example of single-site main group element-mediated dual activation of NH₃, and only the second example involving any metal.²⁰ A superficially comparable process involving the conversion of two NH₃ molecules to −NH₂ ligands with accompanying hydrogen evolution, was reported for the molybdenum bis(imino)pyridine complex, (ⁱPrPDI)Mo(η⁶-C₆H₆) (Scheme 1b; ⁱPrPDI=2,6-(DippNCMe)₂C₅H₃N), via a process which exploits metal-ligand cooperativity.²⁰ It is noteworthy that aluminium(III) diamide compounds featuring two terminal NH₂ groups, are rare: Roesky and co-workers, for example, reported the synthesis of {HC(MeCDippN)₂}Al(NH₂)₂ via the direct reaction of the corresponding Al(III) dichloride with excess ammonia.^{18b, 19}

The role of the counter cations has been shown to be crucial in aluminyl-mediated small molecule activation, with different reaction pathways and/or regioselectivities, for example, being observed in the presence/absence of a K⁺ counterion.²¹ To probe the role of K⁺ in the current context, we treated the ‘naked’ acyclic aluminyl 4 (in which K⁺ is encapsulated within 2.2.2-cryptand) with NH₃ (Scheme 2b). This chemistry leads to the formation of a similar aluminium bis(amide) complex 4 (plus H₂) albeit with a mono-nuclear structure in the solid state (Figure 1), suggesting that the K⁺ cation does not play a decisive role in this specific activation process.

The reactivity of aluminyl compounds 1 and 4 towards ammonia differs from that determined for Kinjo's cyclic C,N-chelated aluminyl system, [{(Me₃Si)₂C(CH)₂NAd}Al]⁻.^8k 1/4 activate two ammonia molecules and liberate H₂, while the latter is reported to undergo a single ammonia activation step involving the insertion of Al(I) into the N−H bond. In a broader context, we hypothesized that the nature of the Al−H bond formed in the first activation step might influence the feasibility of the second (hydrogen evolution) process. With this in mind, and given the isoelectronic relationship between anionic aluminyl and charge-neutral silylene systems, we examined the reactivity of bis(boryloxy)silylene 6 towards NH₃, hypothesizing that a less hydridic Si−H bond would be less prone to formal protonolysis by a second equivalent of NH₃.²² Consistently, under comparable conditions to those used for 1/4, this chemistry generates the ‘simple’ oxidative addition product 7 via insertion of the Si(II) centre into the N−H bond (Scheme 3 and Figure 2). Further treatment of compound 7 with excess ammonia under forcing conditions did not result in dual ammonia activation.

The mechanistic steps implicit in the conversion of the anionic component of aluminyl 4 to bis(amide) 5 (and related systems) have been investigated by quantum chemical calculations (PCM-PBE0-GD3BJ/Def2-TZVP//PBE0-GD3BJ/Def2-SVP). These calculations show that a two-step process involving initial N−H activation at Al(I), followed by protonolysis of the resulting Al−H bond is feasible, with associated activation barriers of 115.4 and 120.4 kJ mol⁻¹, respectively (Figure 3). The overall thermodynamics associated with the two steps are strongly exergonic (−148.8 and −65.8 kJ mol⁻¹, respectively). Moreover, the relative energetic barriers associated with the formation and consumption of (amido)hydride complex 2⁻ are consistent with the observation (by NMR spectroscopy) of 2 in solution at short reaction times.

For the N−H activation step, we considered two possible mechanistic variants through which an initially-formed NH₃ adduct (Int1) could be converted into the corresponding (amido)hydride via proton transfer from N to Al. Accordingly, we find that a bimolecular reaction involving a second molecule of ammonia acting as a proton shuttle incurs a lower energetic cost (via TS2b), than the corresponding reaction involving a single NH₃ molecule and a three-centre transition state (i.e. TS2a). In this respect, the initial N−H activation step mirrors those reported for NH₃ activation by isoelectronic group 14 systems.^{8b, 8g, 8j, 8n, 23}

The observation that aluminyl complexes 1/4 activate two molecules of NH₃ – while silylene system 6 effects only the initial N−H oxidative addition process – speaks to the feasibility of the second protonolysis step. Consistently, the barrier associated with accessing the AlHHN-containing TS3 (120.4 kJ mol⁻¹), is significantly lower than that calculated for the corresponding silicon-containing system (217 kJ mol⁻¹; Figure S23). This difference can be related in turn to the less hydridic nature of the silicon-bonded hydride, which bears a significantly less negative partial charge than its aluminium-bonded counterpart (Mulliken charges: −0.05 vs. −0.23 e). Interestingly our calculations also suggest that Kinjo's amido/hydride [K(12-crown-4)₂] [{(Me₃Si)₂C(CH)₂NAd}Al(H)(NH₂)] should react in a thermodynamically favourable (ΔG^o=−25 kJ mol⁻¹) and kinetically facile manner (ΔG^≠=114 kJ mol⁻¹) to generate the corresponding bis(amide) and dihydrogen. Noting the low temperature conditions utilized in this study (and in the absence of any experimental data on the further reaction with NH₃),^8k it is conceivable that the onward reaction instead leads to protolytic loss of the C,N-donor supporting ligand scaffold. In the case of 2 the less basic nature of the O-donor supporting scaffold would be expected to render protolytic ligand loss less favourable, and we see no evidence for the formation of (HCDippN)₂BOH.

Finally, we set out to explore whether doubly activated bis(amide) 3 could be used to deliver the −NH₂ fragment to unsaturated and/or electrophilic substrates. As such, we probed the chemistry of 3 towards the carbonyl-containing systems CO₂ and Ph₂CO, and the borane HBpin. The reaction with CO₂ leads to transfer of both amide moieties (to two CO₂ molecules), yielding bis(carbamate) complex 8 (Scheme 4). Compound 8 has been characterized by multinuclear NMR spectroscopy and X-ray diffraction (Figure 4). On the other hand, the reaction of compound 3 with more sterically encumbered substrate Ph₂CO leads to transfer of a single amide unit, despite the presence of an additional molecule of ketone in the product (Figure 4). The resulting product (9) contains a (unique) metalated carbinolamine fragment – the protonated form of which is critical intermediate in imine formation from ammonia and a carbonyl compound.¹⁷

In addition, we investigated the reactivity of compound 3 towards (hydridic) B−H bonds with the aim of regenerating the Al−H bond present in intermediate 2. The reaction with pinacolborane (HBpin) does indeed generate pinBNH₂ (presumably driven thermodynamically by B−N bond formation), as confirmed by ¹¹B{¹H} NMR monitoring.²⁴ However, ¹H NMR data implies a mixture of aluminium-containing products (which could not easily be separated), potentially derived from 3 by the substitution of variable numbers of −NH₂ groups by H.

In conclusion, we report a novel advance in main-group-mediated ammonia activation and conversion, demonstrating the efficacy of aluminyl compounds in both the activation of NH₃ and the generation of H₂ (along with an aluminium(III) bis(amide)) on uptake of a second molecule of ammonia. Critically, this chemistry is facilitated by the accessibility of a (coordination/proton shuttling) N−H oxidative addition pathway, and by the inertness of the NHBO-supporting ligand set in the subsequent protonolysis step. As such, we demonstrate the capabilities of low-valent main group complexes in a topical transformation previously precedented by only a single transition metal system.