[CP2K-user] [CP2K:21741] scf convergence error with large systems

[Konstantin Tokarev]

>I find that when I was calculating a large system, in this example, 744 
atoms, the scf convergence always failed, no matter how I tried with 
cutoff, minimizer ...
>However, when I use a smaller system (same elements but fewer atoms), the 
convergence is fine.

Such thing is often a sign of (near) linear dependence in the basis set, 
which is very common in large systems. Everything converges perfectly fine, 
than you extend your molecular cluster, or increase supercell, and it 
suddenly crumbles. If reducing the system is not an option, there are a few 
other ways to proceed:
1) try less diffuse basis set — probably not an option as you are using SR 
variant already;
2) try smaller basis set — for example, there is SZV-MOLOPT-SR-GTH, or try 
something from a different basis set family;
3) filter basis set by eigenvalues of overlap matrix — I'm not exactly sure 
how does it work in CP2K, but changing SCF/EPS_EIGVAL might help. Here is a 
documentation for a similar feature in ORCA: 
https://www.faccts.de/docs/orca/6.1/manual/contents/essentialelements/basisset.html#sec-essentialelements-lindep



среда, 13 августа 2025 г. в 10:43:03 UTC+3, Que Zhu: 

sorry,I can not open the geom.cell file.Can you find me the cif file


FWIW, it's just XYZ file without number of atoms. It's @INCLUDE'd in 
NiO-perovskite.inp in COORD section.



-- 
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+unsubscribe at googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/cp2k/2f65216a-7d11-4c98-9fc2-d6f5e9966408n%40googlegroups.com.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20250813/723025cd/attachment.htm>