How to get Distorted octahedral (D2h point group) for a metal cluster?

maximos

How to make a transition metal cluster with D2h point group (distorted octahedral) in Gaussview ?


Best regards

Uus/pMeC6H4-/キ

Try starting from the structure with octahedral symmetry, applying distortion and symmetrizing the result to D2h with Tools - Point Group. Or if there are only six atoms in the cluster, specifying coordinates in some direct form like (±x, ±y, 0) and (0, 0, ±z) would also be convenient. For bigger system, other program like ABCluster is also an available option.

maximos

Uus/pMeC6H4-/キ 发表于 2025-4-19 23:58
Try starting from the structure with octahedral symmetry, applying distortion and symmetrizing the r ...

I'm grateful for your reply. I did built an octahedral of 6 metal atoms, and distorted a little, but the "D2h" point group didnt appear (attached Image ).

Uus/pMeC6H4-/キ

maximos 发表于 2025-4-20 00:59
I'm grateful for your reply. I did built an octahedral of 6 metal atoms, and distorted a little, b ...

The distortion from Oh to D2h actually requires pulling a pair of opposite edges in or out with equal amount, without which GaussView may not be able to recognize the possibility.

For your case, specifying six atoms with Cartesian coordinates of the form (±x, ±y, 0) and (0, 0, ±z) is much better in several aspects: (1) no manual adjustment needed, just typing numbers up to arbitrary precision; (2) always guaranteed to work regardless of tolerance in symmetry perception; (3) the three C2 axes as well as the three mirror planes are perfectly aligned with three Cartesian axes, and the center is automatically at origin; (4) even if the cluster is subject to a geometric optimization, only three variables x, y, z in Cartesian coordinates are involved, instead of whatever complicated (redundant) internal coordinates as generated by program that also has the risk of not conforming to desired symmetry.

...Well, just to make sure, you know the appropriate theoretical treatments for such a transition metal cluster and any potential computational costs, right?