Wrong UHF assignment for `f-in-core` methods in `xtb` - Githubissues

grimme-lab / MindlessGen

Mindless molecule generator in a Python package.

Apache License 2.0

15 stars 2 forks source link

Wrong UHF assignment for `f-in-core` methods in `xtb` #50

Closed marcelmbn closed 4 days ago

marcelmbn commented 2 weeks ago

High-spin lanthanide complexes with UHF@Ln > 0 enter xtb calculation with same multiplicity as in the actual molecule
Due to f-in-core, this should not be the case (effectively, UHF is then assigned to the ligands)

marcelmbn commented 1 week ago

@jonathan-schoeps: The relevant parts of the code are here:

https://github.com/grimme-lab/MindlessGen/blob/99a4da6a319769f3089dec3c861df8467095353c/src/mindlessgen/molecules/miscellaneous.py#L19-L46

https://github.com/grimme-lab/MindlessGen/blob/main/src/mindlessgen/qm/xtb.py

Idea would be roughly the following:

When starting the singlepoint or optimize routine of xtb:

Archieve the "old" UHF assignment
Calculate the "correct" UHF value given the f-in-core approximation, while keeping the charge. We only want to get rid of the unpaired electrons for the f-electrons temporarily.
Set back either the previous UHF assignment or just use a copied Molecule object for the calculation.