Molcas Forum

andrewshyichuk

Member

Registered: 2020-02-13

Posts: 93

Odd behaviour of embedding AIMPs

Dear Users,

I currently hunt a potential bug in embedding AIMPs - or maybe that is just me not seeing something.

General problem: transition energies in a cluster depend on dpseud radius in lattgen. In an huge embedding of, say, about 150 Angstrom radius, the positions within certain radius R from the cluster are embedding AIMPs, and the rest are point charges. I can change the R as I see fit, and I get first excitation energy changing within 1-5 eV - NONSYSTEMATICALLY, it jumps around. Now, for the funny part: the R is 36-49 Angstrom, all non-electrostatic contributions should have been long gone. The total charge on Xfield is zero.

Anyway, as I do not want to share the system (also, it is huge), I came up with a test: a Ti3+ surrounded by 6 positions - a basic crystal field setup. If I use some -2 charges, or an oxygen embedding AIMP - I get the same splitting, and the dependence on the interatomic distance dies off at about 15-20 Angstrom - very reasonable result.

I then switched to +3 charges and positive embedding AIMPs - surprise! The latter behave as if their charge is much smaller than the formal 3+. So - WHY??
Did the test with all the available 3+, Y and Ga.

ChatGPT says something about cations lacking electrons and being undershielded or whatnot, but that's a nonsense - embedding AIMPs are built from core orbitals and from a distance much look like a point charge. Or should they not?

Files: https://www.dropbox.com/scl/fi/0nsvgj9o … mdm7e&dl=0

That, however, does not resolve a bigger question: why the energy dependence is so large, and why is it jumpy? AIMPs not producing enough positive charge cannot explain that. Or rather, ok, when a 3+ charge replaces a 3+ AIMP, the effective charge on the cluster kinda goes up. But, from 50 Angstrom away - it should not make a difference, not a jumpy one.

Maybe integration grid is too sparse out there?

Thank you