CP2K Problems with Restarting the Localization of Wannier-Centers

[tobias.l... at gmail.com]

Hi,

thank you for your quick answers.
yes the coordinates of the centers are really differing!
My proceeding was to stop the CP2K-run in a controlled manner. This works 
with the directives WALLTIME or LAST_SNAPSHOT. Then I restarted the 
calculation with the options 

&EXT_RESTART
RESTART_FILE_NAME polarizability-1.restart

&DFT
WFN_RESTART_FILE_NAME

&SCF
SCF_GUESS RESTART

&REFTRAJ
FIRST_SNAPSHOT

With the next following timestep in the FIRST_SNAPSHOT option. 
1.Run: 1-5
2.Run: FIRST_SNAPSHOT 6

Doing so the resulting Raman-Spectrum looks like a semi-circle looking like 
any systematic fault. Then I compared the coordinates and they were 
differing with a trajectory sequence from 1-5 and 7-10 for example.

The next try i did was to restart the counters in the &EXT_RESTART section. 
Then the sequence was correct from 1-10 but there were some jumping atoms 
(the trajectory was ruptured) and the coordinates of the centers after the 
restart were still differing and the Raman-Spectrum looks like a hyperbola 
with an underlying sinus oscillation.

And finally in my last try i discovered that CP2K takes everytime when 
called the first timestep to prove if the calculation converges. I changed 
the value of FIRST_SNAPSHOT to
1.Run: 1-5
2.Run: FIRST_SNAPSHOT 5

resulting in a sequence which was still correct (1-10) but without any 
jumping atoms and still differing coordinates of the wannier centers. But 
now the Raman-Spektrum calculated with TRAVIS was finally the same like in 
a single CP2K-Run.

The point of interest now is if my tries lead to any kind of systematic 
fault in TRAVIS because of the Raman-spectrums were quite regularly?

In the end I am glad to be able to do my calculations now. I did the 
imaginable simplest fault and hope to safe other people from doing the same!

Regards,
Tobias


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20150312/db27a49f/attachment.htm>