Slow geometry optimization

[Bakary N'tji Diallo]

Dear all,

I've been trying to perform a simple geometry optimization on a deca-alanine molecule at ( revPBE-D3(BJ)/def2-TZVP). The optimization process seems rather slow (> 2 days on 64 CPUs machine), and the SCF is not converging.

I'm not sure if I am missing something in the input file (please find it attached). The same input would optimize h2 in 5 hours!

I would appreciate any help.

Best regards

--
Bakary 

[Krack Matthias]

Hi Bakary

 

All-electron basis sets like Ahlrichs-def2-TZVP are not compatible with GTH pseudopotentials. Try the MOLOPT basis sets instead. I also suggest to use another mixing method, e.g. Broyden instead of Pulay.

 

HTH

 

Matthias

--
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.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/71eeff0c-1dd5-481a-8ab1-b9fc8b8941can%40googlegroups.com.

[Bakary N'tji Diallo]

Thank you for your reply, Matthias.

Would it make sense if I need to stick to the Ahlrichs-def2-TZVP basis set to move to an all-electrons potential and use GAPW like in the attached input.

Best regards

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/ZRAP278MB082782E92EDEEEF088584185F4419%40ZRAP278MB0827.CHEP278.PROD.OUTLOOK.COM.

--

Bakary