Anion stabilised hypercloso-hexaalane Al6H6

Boron hydride clusters are an extremely diverse compound class, which are of enormous importance to many areas of chemistry. Despite this, stable aluminium hydride analogues of these species have remained staunchly elusive to synthetic chemists. Here, we report that reductions of an amidinato-aluminium(III) hydride complex with magnesium(I) dimers lead to unprecedented examples of stable aluminium(I) hydride complexes, [(ArNacnac)Mg]2[Al6H6(Fiso)2] (ArNacnac = [HC(MeCNAr)2]−, Ar = C6H2Me3-2,4,6 Mes; C6H3Et2-2,6 Dep or C6H3Me2-2,6 Xyl; Fiso = [HC(NDip)2]−, Dip = C6H3Pri2-2,6), which crystallographic and computational studies show to possess near neutral, octahedral hypercloso-hexaalane, Al6H6, cluster cores. The electronically delocalised skeletal bonding in these species is compared to that in the classical borane, [B6H6]2−. Thus, the chemistry of classical polyhedral boranes is extended to stable aluminium hydride clusters for the first time.

T he binary hydrides of boron, i.e. boranes (typically [B x H y ] z− , x ≤ y, z = 0-2), are of enormous importance to chemistry from both fundamental and applications standpoints. The vast majority of these species are low oxidation state boron cluster compounds, which exhibit an enormous array of structural types 1 . The understanding of the structures of such clusters required the early development of revolutionary theories on chemical bonding (e.g. Wade-Mingos rules for electron counting) 2,3 , which ultimately led to boranes finding applications in areas as diverse as synthesis 4 , rocket fuel technology 5 and medical science 6 .
It is remarkable that aluminium, boron's neighbour in group 13, does not form any isolable hydride cluster compounds, or indeed many binary hydride compounds at all, e.g. AlH 3 , H 2 Al (μ-H) 2 AlH 2 and [AlH 4 ] − 7 . With that said, a handful of transient, low oxidation state alane cluster compounds have been studied in the gas phase, and some, e.g. Al 4 H 6 , have been shown to have fleeting stability [8][9][10][11] . Given that numerous ligand substituted, metalloid aluminium cluster compounds, e.g. [Al 77 {N(SiMe 3 )} 20 ] 2 − , have been reported to be stable at, or close to, room temperature 12,13 , it seemed that related low-valent aluminium hydride clusters might be ultimately accessible under the right preparative conditions. As a prelude to realising this goal, we have synthesised the first stable binary low oxidation state aluminium hydride fragments, viz. [Al 2 H 6 ] 2− and Lewis base stabilised Al 2 H 4 , by reduction of aluminium(III) hydride precursors with magnesium(I) dimers 14,15 .

Results
Synthetic and spectroscopic studies.  [1][2][3][4][5], at elevated temperatures (typically 60-80°C) reproducibly afforded low yields (ca. 5-20%) of the deep red crystalline aluminium(I) hydride cluster compounds 1 (Fig. 1), upon cooling the reaction solutions to ambient temperature. On several occasions, a number of low yielding colourless crystalline by-products were isolated from the reaction mixtures, including [(Fiso)Mg  [6][7][8][9]. The nature of these by-products suggests that the reductive mechanism for the formation of 1 could involve several intermediates and/or could compete with side reactions. In order to assess these possibilities, reactions that gave 1 at 70°C were followed by 1 H NMR spectroscopy. This revealed complex mixtures of products after several minutes heating, of which [(Fiso)Mg(Nacnac)] was identified in significant quantities ( Supplementary Fig. 10). Definitive identification of products, other than those which were later isolated as crystalline solids, was not possible, and the mechanism of formation of 1 is not certain at this time.
To the best of our knowledge compounds 1 represent the first examples of isolated aluminium(I) hydride complexes, though mononuclear examples have recently been tentatively proposed as unstable intermediates in solution-based reactions 20 . The cluster compounds have negligible solubility in common deuterated solvents once crystallised, so no meaningful solution state spectroscopic data could be acquired for them. The most relevant solid state spectroscopic data (Supplementary Figs. 11,12) for the compounds come from their infrared spectra, which exhibit single bands in the characteristic region for terminal Al-H stretching modes 7 (e.g. 1a: ν = 1798 cm −1 ). In addition, stronger bands are seen at lower wavenumber (e.g. 1a: ν = 1648 cm −1 ) that possibly arise from a weakly bridging Al-H···Mg stretching mode, though these bands overlap with ligand stretching absorptions (see below). Noteworthy is the fact that the band at ν = 1798 cm −1 observed for 1a is completely absent in the infrared spectrum of its hexa-deuteride analogue, 1a-D, which was prepared by magnesium(I) reduction of [{(μ-N,N-Fiso)Al(D) (μ-D)} 2 ]. The Al-D stretching band for 1a-D should occur at ca. 1270 cm −1 , but this is likely masked by strong ligand stretching modes in that region ( Supplementary Fig. 12).
Crystallographic studies. All complexes 1 were crystallographically characterised and found to be isostructural, so only the molecular structure of 1a is depicted in Fig. 2 Supplementary  Figures 13 and 14). The hydride ligands of each cluster were located from difference maps and freely refined. A neutron diffraction study was also carried out on compound 1a, which unambiguously confirmed the presence and connectivity of the six hydride ligands, and the absence of any other terminal, bridging or interstitial hydrides within the cluster core (see Supplementary Methods and Supplementary Figs. 15, 16). The compounds can be considered as having near neutral, distorted octahedral Al 6 H 6 cores, opposing equatorial sides of which are coordinated by bridging, electronically delocalised formamidinate ligands. The remaining equatorial sides of the octahedron are bridged by [( Mes Nacnac)Mg] + cations, which have weak interactions with the two hydride ligands that project from each side. Terminal hydride ligands coordinate to the apical aluminium centres of the cluster core, though these are slightly offset from the vector passing through the two aluminium centres to which they are coordinated, presumably for steric reasons. All of the Al-Al distances within the Al 6 core lie in the known range for such bonds (mean: 2.72 (12)   Electronic structure and computational studies. The neutral distorted, octahedral Al 6 H 6 cluster cores of 1 somewhat resemble the structure of the classical polyhedral borane, closo-[B 6 H 6 ] 2− 1 , despite their aforementioned elongated equatorial Al-Al interactions. This is intriguing as Al 6 H 6 can be viewed as having 12 (i.e. 2n) valence electrons (i.e. 2 from each Al vertex) contributing to the skeletal Al-Al bonding of the cluster core. As such, it would be expected to have a more unsymmetrical, capped structure than the 14 skeletal valence electron (2n + 2) closo-[B 6 H 6 ] 2− , according to Wade-Mingos rules 2,3 . Indeed, computational studies have predicted a number of more open and unsymmetrical structures for Al 6 H 6 21,22 , which are close in energy. Of course, in 1 the distorted octahedral geometry of this fragment is likely enforced by coordination to the amidinate ligands, which computational studies suggest do not add to the skeletal electron count (see below). For sake of comparison, isoelectronic B 6 H 6 and Ga 6 H 6 , which have not been isolated experimentally, have been predicted to have capped trigonal bipyramidal hypercloso-structures, with all hydrides terminal 23,24 . Also worthy of mention are several ligand substituted analogues of Al 6 H 6 , e.g. B 6 (NMe 2 ) 6 25 and Ga 6 {SiMe(SiMe 3 ) 2 } 6 24 , which possess distorted octahedral structures with several elongated E-E bonds, not dissimilar to the situation in 1. In the case of Ga 6 {SiMe(SiMe 3 ) 2 } 6 , calculations on the model compound Ga 6 H 6 suggest that this can be attributed to a Jahn-Teller distortion arising from loss of degeneracy of the t 2g HOMOs of closo-[Ga 6 H 6 ] 2− upon removal of two electrons from that dianion 24 .
In  Table 2) optimised to be similar to those of 1, including a distorted octahedral Al 6 H 6 core with somewhat shorter Al ax -Al eq bonds (2.650-2.668 Å) than Al eq -Al eq distances (2.718-2.819 Å). Reassuringly, the calculated infrared spectrum of 1′ (Supplementary  Table 5) are suggestive of relatively strong bonding interactions, while the WBIs for the Al eq -Al eq interactions are much smaller (0.23-0.31). In line with this result is the fact that no bond critical points were found between the equatorial aluminium centres (Supplementary Fig. 8). Calculations on the [Al 6 H 6 ( H Fiso) 2 ] 2− dianion, in the absence of the [( Me Nacnac)Mg] + cations, showed this fragment to be stable, with a geometry similar to that in the full contact ion compound (Supplementary Fig. 17 and Supplementary Table 2). This, combined with the fact that the uncoordinated Al 6 H 6 octahedral unit was calculated to be an unstable entity in the electronic singlet state, confirms that the hypercloso-Al 6 H 6 moiety of 1′ is stabilised by coordination to the H Fiso anions.
The electronic structure of the [Al 6 H 6 ( H Fiso) 2 ] 2− dianion was calculated and found to be similar to that of the full contact ion compound , so only the former is displayed in Fig. 3. There are seven Al-based molecular orbitals (MOs) on the dianion, six of which are filled, in line with the view that the cluster is a 12 skeletal valence electron species. None of these MOs are degenerate, but they do closely resemble the seven filled cluster based MOs for closo-[B 6 H 6 ] 2− (triply degenerate t 2g and t 1u orbital sets, and a 1g orbital) 2 , and thus display significant electronic delocalisation over the Al 6 Table 7), as is common for polyhedral boranes 1 .

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, and Raman spectroscopies, single crystal X-ray diffraction studies, and DFT calculations. Further details are available in Supplementary Methods.  1518s, 1262m, 1215w, 1197m,1185m, 1148m, 1092m, 1021m, 996m, 857m, 848w, 775s, 758m, 636m. Note: A satisfactory reproducible microanalysis of the compound could not be obtained due to co-crystallisation of the product with small amounts (ca. 3%) of the β-diketimine, Xyl NacnacH, which could not be removed after several recrystallisations.  1542vs, 1197m, 1176m, 1146s, 1097s, 1021s, 854s, 803s, 755s; Raman (solid under N 2 , 514 nm excitation, cm −1 ): ν = 3064m, 1352s, 1318m, 522m, 386m. A similar yield of 1a was obtained when the reaction was conducted in cyclohexane (ca.  1a (and 1b-c) have negligible solubility in common deuterated solvents once crystallised, so no meaningful solution state spectroscopic data could be acquired for them. Attempts to dissolve 1a in d 8 -THF led to decomposition of the compound. (iv) Attempts were made to obtain solution state spectroscopic data on 1a from red reaction solutions before it crystallised from those solutions. NMR spectroscopic data on those solutions showed complex product mixtures ( Supplementary Fig. 10), while ESI mass spectroscopic analyses of the reaction solutions showed no ion that could be assigned to 1a or its fragmentation products. The latter compound was subsequently intentionally synthesised, spectroscopically characterised, and its X-ray crystal structure obtained (Supplementary Figure 8).