[CP2K-user] [CP2K:14800] Re: Conflict of results

Leopold Talirz leopol... at gmail.com
Wed Feb 24 08:50:14 UTC 2021


Hi Monu,

thanks for following up - I did run a CP2K cell optimization starting from 
your relaxed "twisted" CP2K geometry. You find the trajectory and movie 
attached
Does this agree with your results?

It certainly seems to "untwist" the ligands (see movie) - I had to stop the 
optimization at some point though because convergence became very slow.

Some stats from the total energy:

First total energy (single-point): -2160.041714658635101
First optimization step: -2160.136731180092738
50th optimization step: -2160.2817551900
100th optimization step: -2160.2996248085

I.e. after the first "untwisting" that lowers total energy by  ~260mHa 
(quite significant), the final 50 steps lower the total energy by less than 
3 mHa.
This is with cp2k 8.1 [2] and with minimal modifications to the input file 
(see tar file). It's not clear to me why you would get a different result - 
both MULTIPLE_UNIT_CELL 1 1 2 and 2 2 2 K-points should give this result.

As to why the CASTEP gamma point calculation does not yield the twisted 
structure: it is possible that this involves breaking a symmetry that is 
for some reason preserved in the CASTEP calculation (don't know).
Can you try starting your CASTEP calculation from the twisted CP2K geometry 
and see where it ends up?
If it remains twisted, you can then compare total energies to the untwisted 
result, which should tell you which one is the actual minimum.

Best,
Leo

[1] Try  "grep i 925-pos-1.xyz" on the trajectory
[2] Our previous calculations were with cp2k v5.1 though.

On Tuesday, 23 February 2021 at 05:45:09 UTC+1 Monu Joy wrote:

> Even though the doubling of the z-axis (MULTIPLE_UNIT_CELL 1 1 2) seemed 
> fine in the initial cycles, it ended up being twisted on convergence. Then 
> I have tried the K-point scheme (SCHEME MONKHORST-PACK 2 2 2), where the 
> result was the same, twisted. However, the structure seems reasonable (no 
> twisting) in CASTEP even with the gamma point.
>
> But of course, the culprit could be the small z-axis, but I am not sure 
> how to tackle this in CP2K.
>
> -Monu
>
>
> On Friday, 19 February, 2021, 05:25:24 pm GMT-5, Leopold Talirz <
> leo... at gmail.com> wrote: 
>
>
> Since Andres Ortega from our group noticed this with a similar Al MOF, I 
> just wanted to confirm:
>
> Thanks, Marcella, for the suggestions. I just increased the Z direction by MULTIPLE_UNIT_CELL  
> 1 1 2. Albeit it's not yet converged, the intermediate geometries are 
> reasonable (no twisting so far). 
>
> This is indeed the source of the issue here - doubling the cell along z 
> results in a reversal of the twisting of the ligands, even when starting 
> from the twisted geometry (I checked).
>
> In our case, the twisted structure was the correct minimum for the 
> unit-cell only calculation, with a substantially lower energy (~250meV per 
> twisted phenyl), while for the doubled cell the twisted structure was 
> slightly higher in energy (order of 10meV per twisted phenyl).
>
> This then re-raises the original question of the thread, which is why 
> CASTEP seems to yield a different result. I believe the answer is in the 
> CASTEP output:
>
>                            -------------------------------
>                               k-Points For BZ Sampling
>                            -------------------------------
>                        MP grid size for SCF calculation is  1  1  2
>                             with an offset of   0.000  0.000  0.000
>                        Number of kpoints used =             1
>
> CASTEP chooses a 2-kpoint grid along z (which reduced to just a single 
> k-point calculation through symmetry).
>
> From the CASTEP manual [1]:
>
> > For cubic cells and for the C direction of hexagonal cells the even and 
> odd grids for the Monkhorst-Pack scheme give the same number of k-points. 
> However, the even grid provides better sampling and will always be used 
> automatically under these conditions. This ensures that a good grid with 
> k-point separation at (or less than) the specified target can be achieved 
> more economically. As a consequence, such lattices may have much finer 
> separations than requested as odd grids have been excluded - even though 
> they would have been closer to the specified separation - and the better 
> even grids have been taken in preference.
>
> I suspect that forcing CASTEP to do a Gamma-point calculation would result 
> in the same twisting as seen in CP2K (in our case, we checked with CRYSTAL).
> For our MOF, just to exclude any possibility of issues with OT, we also 
> tested this with DIAGONALIZATION and KPOINTS in cp2k and got the same 
> result as with the 1x1x2 supercell.
>
> The lesson is simply that ~6.5A can be too short along the direction of 
> these inorganic chains/rods in MOFs.
>
> Best,
> Leopold
>
> [1] 
> http://www.tcm.phy.cam.ac.uk/castep/documentation/WebHelp/CASTEP.html#modules/castep/dlgcastepelecoptkpoints.htm?Highlight=monkhorst
>  
>
> -- 
> 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 cp... at googlegroups.com.
> To view this discussion on the web visit 
>
> https://groups.google.com/d/msgid/cp2k/a018da49-5c42-4ade-b9f5-1c1b68579a91n%40googlegroups.com 
> <https://groups.google.com/d/msgid/cp2k/a018da49-5c42-4ade-b9f5-1c1b68579a91n%40googlegroups.com?utm_medium=email&utm_source=footer>
> .
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20210224/c147ef31/attachment.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: traj.tar
Type: application/x-gzip
Size: 1202998 bytes
Desc: not available
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20210224/c147ef31/attachment.bin>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: traj.mpg
Type: video/mpv
Size: 648718 bytes
Desc: not available
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20210224/c147ef31/attachment-0001.bin>


More information about the CP2K-user mailing list