Henry Rzepa's blog

Ever since the concept of a shared two-electron bond was conjured by Gilbert N. Lewis in 1916,[1] chemists have been fascinated by the related concept of a bond order (the number of such bonds that two atoms can participate in, however a bond is defined) and pushing it ever higher for pairs of like-atoms. Lewis first showed in 1916[1] how two carbon atoms could share two, four or six electrons to achieve a bond order of up to three. It took quite a few decades for this to be extended to four for carbon (and nitrogen) and that only with some measure of controversy and dispute (for one recent brief summary, see[2]).

For the transition elements over the last forty years or so, bond orders of four, five and even six between like atom pairs have been mooted and many characterised.[3] Moving to the left of the transition elements in the periodic table, this hunt has looked at elements such as beryllium.^‡ Eleven years back, I explored here how a Be=Be double bond could be formed, strangely enough as an electronically excited state of the dispersion-bound weak Be₂ dimer.[4] This species had a calculated Be-Be distance of 1.78Å, resulting from double excitation from the 2s σ*-antibonding orbital into the degenerate π-bonding orbital above it, giving four electrons in bonding valence orbitals. In 2019, three articles appeared which showed how this bond order might be extended to the lofty heights of three as in Be≡Be[5],[6],[7] for (hypothetical) molecules in their ground electronic state. Here I discuss one example from these articles and compare it to the excited state observations made previously.

A useful starting point is the standard molecular orbital diagram for Be₂, illustrating why the ground state singlet actually has a bond order of zero.

The three 2019 suggestions[5],[6],[7] modified this to surround the Be₂ core with e.g. six Li atoms, resulting in a stable singlet species with a Be-Be distance (calculated at e.g. the CCSD/Def2-TZVP level) of 1.99Å and exhibiting C_2h symmetry.^♥ The role of the Li is to polarise and repopulate Be orbitals by delocalization of e.g. a 2c-2e bond in Be₂ dimer into a 6c-2e bond in Be₂Li₆. The reported calculations (as successfully replicated here^†, FAIR DOI: 10.14469/hpc/10106) show the resulting molecular orbitals for Be₂Li₆ comprise an (accidentally) degenerate π-pair and a higher energy weak σ-orbital, together forming the proposed triple bond. This of course inverts the normal ordering of such bonds, for which the σ-orbital is lower in energy (more stable) than π-bonds. The form of the σ-orbital also reminds to some extent of the fourth σ-bond in C⩸C.

MOs for Be2Li6
HOMO, σ orbital -0.158au	HOMO-1, π-pair, -0.175au	HOMO-2, π-pair -0.176au
Because the static 2D projections shown in the articles cited above do not always make for easy interpretation, if you click on the orbital thumbnails, you will get dynamic 3D isosurfaces to rotate and inspect. These were generated using the tool at https://www.ch.ic.ac.uk/rzepa/cub2jvxl/

The two lower energy 2s σ-orbitals, which taken together do not apparently contribute to the overall bond order in Be₂Li₆, are shown below.

Lower energy MOs for Be2Li6
σ -0.235au	σ-0.496au

ELF (electron localisation function) integrations for Be₂Li₆ show each beryllium has two basins in the Be-Be region of about 2.5e each (red arrows) typical of triple bonds and two terminal Li-Be basins of 2.3e.

One aspect arising from my earlier post on the excited state Be=Be double bond relates to the reported calculated Be-Be bond length of 1.99Å and ν 718 cm^-1 for ground state Be₂Li₆. To quote one article[5], “the Be≡Be triple bond in Li₆Be₂ may also be considered as another example of an ultraweak but ultrashort triple bond.” I had noted earlier that the electronically excited state of the Be₂ dimer has a computed bond length of 1.78Å and ν 917 cm^-1 for a double bond order, this being significantly shorter than the suggested ultrashort triple bond. We learn from this that the relationship between a bond order and a bond length may not always be linear. In other words, a longer bond may in fact have a higher bond order than a shorter bond between the same two atoms. The same was true as it happens with C⩸C; the mooted quadruple bond had a longer bond length than the triple bond in HC≡CH. That observation was controversial at the time; I suspect a similar phenomenon for Be has become less controversial.

To go back to the Be=Be dimer which started things off and that excited state with one electron in each of the degenerate π-orbitals (actually a triplet state). What would happen if two electrons were to be added, making an excited state of Be₂^2-? Yes indeed, this species (CCSD/Def2-TZVPPD) has a calculated bond length of 1.885Å and ν 766 cm^-1. If this di-anion is stabilised with a continuum water field (a milder version of surrounding the dimer with Li atoms), the Be-Be length contracts to 1.74Å, the Be-Be stretch increases to 949 cm^-1 and the σ-orbital becomes more stable than the π-orbitals. At the higher CCSD(T)/Def2-TZVPPD/SCRF=water level, the bond length still has the ultrashort value of 1.761Å, which might be assumed as the natural value for Be≡Be, a classical triple bond. From that perspective, the “ultraweak but ultrashort triple bond” predicted for Be₂Li₆ actually emerges as a relatively long triple bond!

Our final exploration is to add two lithium atoms to Be₂ to form the neutral LiBe≡BeLi. This was done in stages (see FAIR DOI 10.14469/hpc/10106), starting with a linear arrangement of atoms which revealed two negative force constants, a C_2h shape with one negative force constant and ending with a C₂ (chiral!) geometry with no negative force constants. This has a Be≡Be length of 1.705Å (ωB97XD/Def2-TZVPPD/SCRF=water), ν 1129 cm^-1, a Wiberg bond index of 2.98 and a Li-Be bond index of 0.0065, indicating an entirely ionic lithium and again a central Be₂^2- unit. As an excited state, it is 49.8 kcal/mol higher than the ground state of Be₂Li₂.

NBOs for LiBe≡BeLi
HOMO, π-pair, -0.175au	HOMO, π-pair -0.176au	HOMO-2, σ orbital -0.158au

So to conclude, we have seen two different motifs for constructing a model of a Be≡Be triple bond, one recently reported in the literature for a ground state species with six lithium atoms surrounding the Be₂ dimer and a simpler one with just two lithiums exhibiting a much shorter Be≡Be bond but which requires electronic excitation to achieve. So these two motifs are not equivalent. But hopefully this exercise shows how playing around with atoms and electrons can achieve very unusual bonding states and elevated bond orders from which one can learn a lot, although with the caveat that one does not always produce molecules capable of facile synthesis!

^‡On a slightly different theme, Cs can be shown to sustain three bonds, albeit all to different atoms. See DOI: 10.6084/m9.figshare.861030 Li≡Li^4- can also be calculated as the tetra-anion showing almost identical properties to Be≡Be^2- with a Li≡Li triple bond distance of 2.11Å. See DOI: 10.14469/hpc/10122. ^†Replication was necessary because the appropriate wavefunction files for analysis were not included in the supporting information. Only the coordinates were available for interoperation, and due to a quirk in the way Adobe Acrobat works, even those could not be easily transferred by a simple copy/paste operation to create a job input file. See e.g. here or DOI: 10.14469/hpc/10043 for more discussion. All the wavefunction files for this replication are available at the FAIR DOI noted above. ^♥The Be-Be distance in catena(dimethylberyllium), a polymer comprising two bridging Me units connecting Be atoms, is only slightly longer at 2.09Å[8] This fascinating transannular Be-Be interaction is one to be explored elsewhere.

The post has DOI: 10.14469/hpc/10125

References

1. G.N. Lewis, "THE ATOM AND THE MOLECULE.", Journal of the American Chemical Society, vol. 38, pp. 762-785, 1916. https://doi.org/10.1021/ja02261a002
2. H.S. Rzepa, "Routes involving no free C ₂ in a DFT-computed mechanistic model for the reported room-temperature chemical synthesis of C ₂", Physical Chemistry Chemical Physics, vol. 23, pp. 12630-12636, 2021. https://doi.org/10.1039/d1cp02056k
3. D. Lu, P.P. Chen, T. Kuo, and Y. Tsai, "The MoMo Quintuple Bond as a Ligand to Stabilize Transition‐Metal Complexes", Angewandte Chemie International Edition, vol. 54, pp. 9106-9110, 2015. https://doi.org/10.1002/anie.201504414
4. P.J. Bruna, and J.S. Wright, "Strongly bound doubly excited states of Be₂", Canadian Journal of Chemistry, vol. 74, pp. 998-1004, 1996. https://doi.org/10.1139/v96-111
5. S.S. Rohman, C. Kashyap, S.S. Ullah, A.K. Guha, L.J. Mazumder, and P.K. Sharma, "Ultra‐Weak Metal−Metal Bonding: Is There a Beryllium‐Beryllium Triple Bond?", ChemPhysChem, vol. 20, pp. 516-518, 2019. https://doi.org/10.1002/cphc.201900051
6. X. Liu, R. Zhong, M. Zhang, S. Wu, Y. Geng, and Z. Su, "BeBe triple bond in Be₂X₄Y₂ clusters (X = Li, Na and Y = Li, Na, K) and a perfect classical BeBe triple bond presented in Be₂Na₄K₂", Dalton Transactions, vol. 48, pp. 14590-14594, 2019. https://doi.org/10.1039/c9dt03321a
7. S.S. Rohman, C. Kashyap, S.S. Ullah, L.J. Mazumder, P.P. Sahu, A. Kalita, S. Reza, P.P. Hazarika, B. Borah, and A.K. Guha, "Revisiting ultra-weak metal-metal bonding", Chemical Physics Letters, vol. 730, pp. 411-415, 2019. https://doi.org/10.1016/j.cplett.2019.06.023
8. A.I. Snow, and R.E. Rundle, "The structure of dimethylberyllium", Acta Crystallographica, vol. 4, pp. 348-352, 1951. https://doi.org/10.1107/s0365110x51001100

I have posted often on the chemical phenomenon known as hypervalency, being careful to state that as defined it applies just to “octet excess” in main group elements. But what about the next valence shell, occurring in transition metals and known as the “18-electron rule”? You rarely hear the term hypervalency being applied to such molecules, defined presumably by the 18-electron valence shell count being exceeded. So the following molecule (drawn in three possible valence bond representations) first made in 1992 intrigues.[1]

The molecule comprises six phosphinidene ligands (RP:, R=tert-butyl), the P analogues of nitrenes and arranged around nickel to form an unusual hexagonal planar coordinate species and with three of the t-butyl groups facing up and three down. This arrangement totally obscures the two nickel diaxial positions, preventing any ligand from occupying them. One may even speculate that the dispersion attractions between the two pairs of three t-butyl groups might be unusually stabilising,^‡ maybe even on a par with those reported by Schreiner and co-workers for t-butyl substituted triphenylmethanes and noted on this blog.

Nitrenes can be represented coordinating to a metal as RN=M. If the analogy extends to P, the valence bond structure 1 above would result and the six P atoms would contribute 12 electrons to the Ni valence shell. Since the Ni shell is  [Ar].3d⁸.4s² adding another 12 electrons would make 22 electrons, thus exceeding the 18-electron rule. In fact it was never suggested as such; in the 1992 analysis of the bonding[1], the authors clearly state “The electronic structure … cannot be described in terms of the 18e rule”. To support this, they draw the structure in form 3 above, which implies a Ni valence shell of 16e, albeit also implying a Ni bond index of ~6.

Time I thought for calculations (wB97XD/Def2-TZVPP).^† The calculated NBO bond index at Ni is in fact 2.38 and the individual Ni-P bond orders are 0.37. The P bond indices are each 3.24 and the P-P bond orders are 0.88. The final electronic configuration is [Ar]3d^9.53.4s^0.22.4p^0.64.4d^0.01.5p^0.01 and the natural charge on Ni is -0.39. I show one NBO orbital which illustrates the two-electron-three-centre interaction spanning two P atoms and the Ni and giving rise to the modest Ni-P bond order.

The Ni electronic structure of [Ar].3d⁸.4s² normally corresponds to its divalency, so in this sense it is mildly hypervalent. Overall, representation 2 above is perhaps more accurate than 3. The ELF (Electron localisation function) of the electron density is shown below (t-butyl groups represented by a single carbon) with the ELF basins highlighted. Basin 2 is a P lone pair, with an integration close to 2e. Basin 1 is a P-P basin with an integration of 1.89 and finally 3 is indeed a P-Ni basin, but with an integration of only 0.06e. 

So this molecule really is a ring of six P atoms, encapsulating at its centre a lone Ni(II) atom. Rather than being hypervalent, perhaps it is most interesting as a complex in which a metal atom is contained in and perturbed by a dispersion-stabilized sphere of ligands.[2] We need more examples!

^‡The original authors[1] merely stated that electron correlation effects are decisive for the stability. Of course, dispersion attractions are indeed a form of electron correlation! ^†FAIR data doi: 10.14469/hpc/3925

References

1. R. Ahlrichs, D. Fenske, H. Oesen, and U. Schneider, "Synthesis and Structure of [Ni(P<i>t</i>Bu₆)] and [Ni₅(P<i>t</i>Bu)₆(CO)₅] and Calculations on the Electronic Structure of [Ni(P<i>t</i>Bu)₆] and (PR)₆, R = <i>t</i>Bu,Me", Angewandte Chemie International Edition in English, vol. 31, pp. 323-326, 1992. https://doi.org/10.1002/anie.199203231
2. https://doi.org/

An article with the title shown above in part recently appeared.[1] Given the apparent similarity of HF^1- to CH₃F^1- and CH₃F^2-, the latter of which I introduced on this blog previously, I thought it of interest to apply my analysis to HF^1-.

The authors[1] conclude that “the F atom of HF⁻ is negative and hypervalent and the bonding is more covalent than ionic“. So, firstly an NBO analysis. Shown below is the singly occupied NBO (ωB97XD/Def2-TZVPPD calculation, FAIR data DOI: 10.14469/hpc/3274)

The nature of this orbital is that most of it is located beyond the hydrogen and a node is apparent between the F and H (it is H-F antibonding). The Wiberg bond index for both H and F is 0.48, as is the F-H bond order. This matches with the observation of one electron in an orbital which is H-F antibonding. NBO analysis also indicates that the atomic orbital contributions to F[core]2S(1.93)2p(5.77)3S(0.05)3p(0.04)3d(0.01) and H 1S(0.96)2S(0.12)2p(0.10) show modest Rydberg character on the hydrogen, less on the fluorine. By this criterion, it is the hydrogen and not the fluorine that is hypervalent!

Next, the ELF analysis (respectively FAIR data DOI: DFT 10.14469/hpc/3377  and CASSCF(7,8) 10.14469/hpc/3394). 

DFT	CASSCF(7,8)

Both these methods reveal a monosynaptic ELF basin located away from the H, with an integration of about 1 electron. The F lone pairs form a torus around the F and the total of electrons around the fluorine is <8. So again no evidence from ELF that the fluorine is hypervalent.

In fact this analysis resembles one feature of CLi6. With nominally 12e apparently contributing to the shared C-Li shell, CLi₆ was described as hypervalent.[2] In fact ~3e of these are “expelled” from the shared C-Li regions into Li-Li regions, where they contribute only to lithium valency and not to the carbon valency. With HF^1-, the additional electron apparently responsible for the hypervalency contributes only to the H, but not to the F valencies. With CH₃F^2-, the two injected electrons do appear to contribute to the C-F bond, making it a true hyperbond.

So based on the above, I cannot entirely agree with the assertion that “the F atom of HF⁻ is negative and hypervalent”[1], but I might suggest that something more unusual is happening, the hydrogen is (mildly) hypervalent! 

References

1. M. Liu, H. Chen, C. Chin, T. Huang, Y. Chen, and Y. Wu, "Identification of a Simplest Hypervalent Hydrogen Fluoride Anion in Solid Argon", Scientific Reports, vol. 7, 2017. https://doi.org/10.1038/s41598-017-02687-z
2. H. Kudo, "Observation of hypervalent CLi6 by Knudsen-effusion mass spectrometry", Nature, vol. 355, pp. 432-434, 1992. https://doi.org/10.1038/355432a0