systems with d electrons: difference in QS total energies between 2026.2 and 2024.1 (and other older versions)

[Martin Konôpka]

Dear CP2K experts,

In my calculations using the Quickstep method I noticed in some cases slightly different total energies in  the newest released version (2026.2) compared to several older versions (e.g. 2024.1). In my more detailed tests I could see that the  differences were present when an element with d valence  electrons (e.g. gold) was among the atoms. Forces on the atoms in 2026.2 are also slightly different, also the Kohn-Sham gap. You can  see my example runs - a gold dimer in a rather arbitrary orthorombic cell.  The  same input  file was used for both  the  runs.

On  the positive side, I see that the SCF convergence  with the newest CP2K is much  easier (less iterations) in cases when the difference in the total energies occurs. I did this  kind of test also with graphene (no d electrons) and in this case  the energies were identical.

My question is if I am doing someting wrong, something which  induces the differences in  the case of d electrons or if it is a real feature of CP2K 2026.2 that it gives  slightly different (lower) total energies in such cases?

Thanks for your valued opinions.

Best wishes

Martin Konôpka.

[Jürg Hutter]

Hi

with the new version 2026.2 there was an update to the BASIS_MOLOPT_UZH and POTENTIAL_UZH files.
See the headers of the files for the elements that have been updated.
You can find the older versions in the data directory:
BASIS_MOLOPT_UZH_2026.2
POTENTIAL_UZH_2026.1

regards
JH

________________________________________
From: cp...@googlegroups.com <cp...@googlegroups.com> on behalf of Martin Konôpka <konop...@gmail.com>
Sent: Thursday, August 13, 2026 12:15 PM
To: cp2k
Subject: [CP2K:22394] systems with d electrons: difference in QS total energies between 2026.2 and 2024.1 (and other older versions)

--
You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com<mailto:cp2k+uns...@googlegroups.com>.
To view this discussion visit https://groups.google.com/d/msgid/cp2k/29f09827-8ad3-4d7f-b95d-6f1466ca3a2an%40googlegroups.com<https://groups.google.com/d/msgid/cp2k/29f09827-8ad3-4d7f-b95d-6f1466ca3a2an%40googlegroups.com?utm_medium=email&utm_source=footer>.

[Martin Konôpka]

Dear prof. Hutter,

Thanks for your very quick response and notifying me about  the updates in the files.

Best regards

Martin Konôpka.

Dátum: štvrtok 13. augusta 2026, čas: 12:23:52 UTC+2, odosielateľ: Jürg Hutter