MOF cell opt with KEEP_SYMMETRY FALSE

[Corrado Bacchiocchi]

Dear Colleagues,

I ask you to kindly share your experience about the following question.

We have performed on our MOF structure the "Converge CUTOFF and REL_CUTOFF" procedure, as described in the very helpful page:

https://www.cp2k.org/howto:converging_cutoff

and found that, with  NGRIDS 4, to reach a 1.E-7 difference with respect to the limiting value of the energy, the minimum settings are:

     
      CUTOFF 1800
      REL_CUTOFF 80

We have then successfully performed CELL_OPT on a 2x2x2 supercell of our MOF with KEEP_SYMMETRY TRUE.

When we use this optimized structure for a subsequent CELL_OPT with KEEP_SYMMETRY FALSE the BFGS algorithm is not able to further decrease the energy. The structure remains essentially unmodified across hundreds of steps and the optimization does not complete.

I attach the input file and the guess structure and I kindly ask the following:

- is it essential, for any publication quality structure, that KEEP_SYMMETRY is FALSE or is it acceptable to have this constraint in place?

- more in general, for any subsequent calculation, e.g. interaction of small molecules on the surface of the MOF, Nudged Elastic Band energy path investigations, etc., is it possible/acceptable to KEEP_SYMMETRY TRUE?

Thank you very much for your kind help

Have a nice Holiday Season!

Corrado

[Krack Matthias]

Hi Corrado

 

CP2K does not take advantage of symmetry for better computational efficiency. The constraints are rather meant to preserve or enforce a certain symmetry during structure relaxations, because the symmetric structure is not the exact minimum or it distorts slightly due to numerical noise. The latter is the reason why it can make sense to use the constraint KEEP_ANGLES which will keep all cell angles exactly at 90 degree, because CP2K/Quickstep is faster for orthorhombic cells. Therefore, based on the information which you provided for your system, I would apply only the constraint KEEP_ANGLES in the runs with small molecules adsorbed.

 

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/14f06834-79c7-4ac8-922c-d1d6f008d27fn%40googlegroups.com.

[Corrado Bacchiocchi]

Dear Matthias,

thanks a lot for the information and the suggestion.

Interestingly, the CELL_OPT with KEEP_SYMMETRY FALSE does not change the angles that remain perfectly at 90 deg at all steps, so, I am afraid that by applying the constraint KEEP_ANGLES might not going to help in this case, since they are already not changing, but I will try and let you know. It might be a useful information to the community.

Thanks again!

Corrado

[Krack Matthias]

Dear Corrado

 

the CP2K output (e.g. with PRINT_LEVEL medium) informs you if the system is considered as orthorhombic. Check for the output line:

“CELL| Numerically orthorhombic:”

 

Best

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/d5f1d6ed-fabb-47ff-bb26-7564ab762a6fn%40googlegroups.com.

[Corrado Bacchiocchi]

yes, the cell remains "Numerically orthorhombic" at all steps (YES at all "CELL| Numerically orthorhombic" lines)

[Corrado Bacchiocchi]

Dear Matthias,

with "KEEP_ANGLES TRUE" the CELL_OPT got completed at the first step, i.e., the guess geometry obtained with the preliminary calculation, with KEEP_SYMMETRY TRUE, was not further modified.

Again, thank you very much for your kind help

Best

Corrado