2D PBC with non-orthogonal cell

Ondrej Krejci

unread,

Jul 5, 2021, 5:30:12 AM7/5/21

to cp2k

Dear CP2K users and developers,

I am trying to calculate systems with only 2D periodic boundary conditions, because of necessity to apply electric field. As I've found out from attempts of my predecessors, one of the way how to do it, is to use "wavelet" poisson solver with periodic "XZ" conditions. Unfortunately, I've found out, that this seem to work only for orthogonal cells. When I am using the "wavelet" with non-orthogonal cells, I get completely non-sense geometries during optimisation. I attach a difference between standard periodic and wavelet run (input files, output file and output geometries).

Please is there some way around it? - 2D periodic boundary conditions with non-orthogonal cell?

If not - what kind of setup/keywords are available for running kind big systems (2000 atoms). I am trying to run the orthogonal cell at the moment, however it seems that with the similar setup I am limited to ~ 5000 cores and also restart with usage of wfn files crushes.

Thank you very much for your answer!

Kind regards,

Ondrej

Ondrej Krejci

unread,

Jul 5, 2021, 5:33:17 AM7/5/21

to cp2k

I am trying to post the files here.

Ir_111_periodic.inp

Ir_111_wavelet.inp

Ir_111_periodic_out.txt

Ir_111_wavelet_out.xyz

Ir_111_wavelet_out.txt

Ir_111_periodic_out.xyz

DMITRII Drugov

unread,

Jul 5, 2021, 7:40:01 AM7/5/21

to cp...@googlegroups.com

Hi, I did some 2D calculations for E ads on slabs, for this 2D geometry you need to use Poison Analytic. Wavelets is only for non-periodic system. You also need to consider enough vacuum in your non periodic direction.

Regards,

Dmitrii

--
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/7d5e532b-f53c-4331-b0bc-30124f790e42n%40googlegroups.com.

Ondrej Krejci

unread,

Jul 5, 2021, 11:01:17 AM7/5/21

to cp2k

Dear Dmitrii!

Thank you a lot for your reply! I'm going to test it immediately. I normally use 5 Angstrom of Vacuum in every edge, that is not periodic. Is there some big reason to put there more vacuum?

DMITRII Drugov

unread,

Jul 5, 2021, 9:44:23 PM7/5/21

to cp...@googlegroups.com

Hi, I am not too sure about details but people suggest to use vacuum size twice longer or more than any other dimension.

Check this paper if you want to do charge surface simulation PHYSICAL REVIEW B 73, 115407 2006

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/4f73317b-42a6-4cca-841e-2f036db4dbc6n%40googlegroups.com.

Ondrej Krejci

unread,

Aug 16, 2021, 8:30:05 AM8/16/21

to cp2k

Hi,

Once more thank you for your help!

As for my system which is 30x25 Angstrom in periodic direction, to have the non-periodic axis 35 Angstrom seems to be enough and giving reasonable results. Even though it is not twice the other directions, which would be rather inefficient.