Problem converge Cu bulk calculations using kpoints

[Lucas Bandeira]

Dear cp2k community, 

I am performing some calculations for Cu bulk using kpoints. However, any of the calculations that I have performed converged. It seems rather that the results diverge. In addition, the cell vectors printed at the beginning of every optimization step diverge from the experimental value that is 3.64Angs. Before the optimization, I have performed convergence tests for kpoints, and based on these, I chose the kpoint grid 8x8x8. I also tested different optimizers: LBFGS, BFGS, and conjugate gradient, but for all them I have the same problem. Could somebody help me with this? 

Attached are the input file used for the calculations with BFGS, and the output file obtained.

Yours sincerely, 

Lucas Bandeira

[Nicholas Winner]

Try removing the lines

EXTRAPOLATION USE_GUESS

EXTRAPOLATION_ORDER 1

Use guess will generally not be a good choice. The defaults should work well. Also consider trying the DZVP basis set first, which should be easier to converge, and moving to TZV2P only if necessary. 

-Nick

[Lucas Lodeiro]

Hi !

Nicholas' advice is very important. Your output shows an oscillating behavior, and one reason could be the extrapolation. Also, there could be more reasons.

One is related with the metallic character of Cu, a smearing of 300K could be so low, I would try with 2000K.

Another reason could be the REAL option on the kpoints wavefunctions... I would use the COMPLEX option.

If these advices do not work, you could explore the settings on the BFGS optimizer... for example decreasing the TRUST_RADIUS.

Regards - Lucas Lodeiro

--
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/9affdc05-6a56-4bdb-9bcd-0631c312ce10n%40googlegroups.com.

[Lucas Bandeira]

Thank you, Nicholas and Lucas. It worked now!

It worked when I only swapped the WAVEFUNCTION flag from REAL to COMPLEX. But the calculations were faster when I changed both the EXTRAPOLATION and EXTRAPOLATION_ORDER flags to the default value in cp2k, too.

Thank you both very much. 

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/CAOFT4P%2B3OmGbybFNhEY0JsT3OWJvEGcCkzY2%3DOc9-8W8okva3w%40mail.gmail.com.