Introduction
In spite of the obvious deterrence for experimentalists of the high toxicity of beryllium compounds,[1] a great deal is now known about the organometallic and coordination chemistry of beryllium.[2] Beryllium compounds that have been studied experimentally and especially computationally include various systems that feature BeBe bonding and/or very short BeBe distances.[3-19] For example, it has been shown for various choices of X, such as a fluorine atom or an appropriate N‑heterocyclic carbene ligand, that certain XBeBeX species feature Be−Be bonds that are both significantly stronger and shorter than the weak and rather long bond in Be2.[3-4] We note that Liu et al. have interpreted the bonding in the octahedral Be2(µ2-Li)4 cluster, and others, in terms of a Be=Be double‑π bond[10] and the bonding in Be2X4Y2 clusters (X = Li, Na and Y = Li, Na, K) in term of Be≡Be triple bonds.[15] Calculations and analysis carried out by Rohman et al. also support the notion of Be≡Be triple bonds in various systems, including Be2X6 (X = Li, Na), but they found the BeBe bonding to be ultra-weak in spite of the very short BeBe distances.[16-17] Indeed, some systems have been studied both experimentally and computationally, such as the rhombic Be2O2 cluster, which feature very short BeBe distances in the absence of any direct BeBe chemical bonding.[8-9, 11]
There has been significant recent computational interest in predicting the existence of potentially stable beryllium complexes that feature ultra-short BeBe distances,[6, 12-13, 18-19] regardless of whether or not they actually involve any direct chemical bonds between the beryllium atoms. The present work was motivated by one such study in which three bridging hydrogen atoms were used potentially to simulate the effect of a Be≡Be triple bond, with the outcome that particularly short BeBe distances were predicted for various systems, including the D 3h[BeH3Be]+ cation.[13] Our main goal here is to investigate the nature of the bonding interactions in this type of system, especially the D 3h[MH3M]+ cations (M = Be, Mg), including those ‘capped’ by He or Ne atoms (as proxies for an inert gas matrix). To this end, we follow all-electron CCSD(T)/cc‑pVQZ geometry optimizations with spin-coupled generalized valence bond (SCGVB) calculations and the analysis of localized natural orbitals and domain-averaged Fermi holes.
It is our expectation that the various systems we study should all correspond to local minima on their respective potential energy surfaces and that their electronic structures are somewhat more likely to involve highly polar three‑centre two‑electron (3c‑2e) M−H−M bonding character rather than direct metal-metal chemical bonds. We note in this context that Kalita et al.[14] detected in the LiMH2MLi system (M =Be, Mg, Ca) with two bridging H atoms the presence of two 3c‑2e M−H−M bonds that were said to be reminiscent of the bonding situation in diborane.