[CP2K-user] [CP2K:11279] Re: SCF convergence criterion options

[Krack Matthias (PSI)]

Hi Matthew

You are comparing OT and TD (traditional diagonalization), but you are using Broyden with TD, which is most likely not the best choice for your case. You can converge your system also with OT, which implies that you are not using any smearing. In such cases, you should use TD with DIIS (DIRECT_P_MIXING), which is the default for good reason. TD with DIIS, if applicable, is usually by far the fastest (in the number of SCF iterations) and tightest SCF method, since once the DIIS sets in, you gain often one or even two orders of magnitude in convergence per SCF iteration step. Due to the unfavourable scaling of the diagonalization, however, OT performs much better for larger systems. TD with Broyden is employed for systems with fractional orbital occupations (smearing), for which TD/DIIS and OT are not applicable. Broyden is rather a resort for cases, which cannot be converged with other methods.

Matthias

Von: cp... at googlegroups.com <cp... at googlegroups.com> Im Auftrag von Matthew Okenyi
Gesendet: Samstag, 16. Februar 2019 14:45
An: cp2k <cp... at googlegroups.com>
Betreff: [CP2K:11279] Re: SCF convergence criterion options

Hi Martin,

Thanks for your advice. I've been comparing the convergence of the OT and diagonalization methods for the same EPS_SCF value. The change in energy at convergence is 100 times smaller for OT than for TD in my system. EPS_SCF appears to be interpreted more strictly for OT than for TD. Is this correct?

Best wishes,
Matthew

On Saturday, 16 February 2019 11:18:08 UTC, M. Brehm wrote:
To the best of my knowledge, no other critera to signal SCF convergence are implemented in CP2k.

However, it is generally a bad idea to look at the SCF energy to determine if the SCF has converged... There exist quite common cases where the energy remains almost constant in some SCF iterations, but will heavily reduce later on. Many years of practical experience have shown that the norm of the difference between two subsequent density matrices is quite a good criterion.

Best regards,
Martin
--
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... at googlegroups.com<mailto:cp2k+uns... at googlegroups.com>.
To post to this group, send email to cp... at googlegroups.com<mailto:cp... at googlegroups.com>.
Visit this group at https://groups.google.com/group/cp2k.
For more options, visit https://groups.google.com/d/optout.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20190216/99bd70fb/attachment.htm>