Prevent Cell Geometry/Vectors From Changing

[ASSIDUO Network]

Hi there. So, I've been doing a few geometry optimizations and I have seen that for some of them, the cell vectors change, sometimes a bit other times a lot. This especially happens when the SCF run is not converged multiple times in a row. The thing is, I don't want the cell vectors to change, I don't want the atoms to go outside of the box, yet it still happens. Can you give me any tips/ideas on how to do a geometry optimizations where atomic positions can only change and the cell geometry cannot, or is this not feasible?

Kind regards,

Lenard

[fabia...@gmail.com]

Hi Lenard,

in cp2k there are two types of optimizations, GEO_OPT and CELL_OPT. In GEO_OPT the lattice vectors are fixed and only the atomic positions are optimized. In a CELL_OPT both the lattice vactors and the atomic positions can be relaxed simultaneously.

So if you don't want the box to change you need

RUN_TYPE GEO_OPT

and  the

&MOTION

  &GEO_OPT

  &END

&END

section.

If the SCF does not converge the forces (and pressure) are probably very bad and thus the optimizer (e.g. BFGS) takes unreasonable steps. If the SCF does not converge you should probably stop the simulation and improve SCF stability.

Cheers,

Fabian

[ASSIDUO Network]

Thanks for feedback Fabian, I appreciate it. Does the above also apply to AIMD simulations where you have the

&MOTION

  &GEO_OPT

  &END

&END

section still included?

[fabia...@gmail.com]

No, for MD the ensemble type determines if the cell dimensions are fixed or not: https://manual.cp2k.org/cp2k-9_1-branch/CP2K_INPUT/MOTION/MD.html#ENSEMBLE

The cell only changes if you choose NPT, otherwise it is fixed.

Fabian

[ASSIDUO Network]

Great, thanks for letting me know. 

[ASSIDUO Network]

Hey Fabian, I'm messaging here since I didn't get a response on my other question. What about NEB calculations, what do I do there if I don't want the fixed atoms to move?

[fabia...@gmail.com]

I've never used &BAND together with &CONSTRAINT. From a quick glance at the source code I think that it might not be implemented.

[Lenard Carroll]

Thanks for letting me know Fabian, I appreciate it.

--
You received this message because you are subscribed to a topic in the Google Groups "cp2k" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/cp2k/5hOKf9tgi3U/unsubscribe.
To unsubscribe from this group and all its topics, send an email to cp2k+uns...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/f260cd04-0fc7-486e-b717-9ab1c01957fan%40googlegroups.com.