High-frequncy oscillations in a NPT ensemble

[Ider]

Dear cp2k developers and users,

I would like to ask for some help with some issues I am having with AIMD 
simulation in the  NPT ensebmle.  I attached the input file "npt.inp" as 
well as "stress"- and "cell"-files. For convenience I also added plots of 
the cell along c and zz component of the stress tensor.  
My goal is to equilibrate a model of a Ge2Sb2Te5 (GST225) crystal phase 
(108 atoms) at T=600K and p=0 kbar. What I did was to relax the coordinates 
and the cell parameters and then run an equilibration of the system at 600K 
at constant volume ( ensemble=langevin ) which works fine. Then I switched 
the ensemble to NPT_I using the canonical velocity rescaling thermostat 
(Bussi thermostat) and the barostat with the default settings ( I tried 
both TIMECON=100 fs / TIMECON = 1000fs ). However I observe a strange 
high-frequency oscillations ( time period = 20-30 fs !! ) of the stress and 
the cell parameters, which looks very systematic. This oscillations remains 
even for longer simulation times ( see the plot "long-cell-c.png" , in 
which it was slightly different setting with flexible cell / without vdW). 

I tested the effect of following parameters on the behaviour of the system. 
Well the simulation time testing was quite short (varying 100-1000 steps). 
No significant change of the oscillation on this simulation time scale 
could be observed.

1. CUTOFF= 300 / 600 / 1200Ry.      
 
2. Larger reference cell so that the grid density is constant. In the input 
file "npt.inp" this section is uncommented, however in the restart file it 
is present. 
    &CELL_REF
       ABC 13.5 15.0 20.0
    &END CELL_REF

3. TIMECON = 100 / 1000 / 2000 / 10000 fs. 
4. TEMPERATURE of the BAROSTAT = 60K. 
5. I also tried some options in the XC_GRID section. For example RHO = 
SPLINE3 or XC_DERIV = SPLINE3 / NN50 /NN50_SMOOTH. No significant 
improvements. 

The stacking of the layers in this GST225 model is in the "inverted Petrov" 
sequence where Ge and Te are swicthed. I also tested the stacking in the 
"Petrov" sequence, but similar oscillations are present. 

I would be truly grateful for any advice!

Best regards,

Ider
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20141125/2f664d4d/attachment.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: npt.inp
Type: chemical/x-gamess-input
Size: 12316 bytes
Desc: not available
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20141125/2f664d4d/attachment.inp>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: IPCM-1.cell
Type: application/octet-stream
Size: 51983 bytes
Desc: not available
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20141125/2f664d4d/attachment.obj>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: IPCM-1.stress
Type: application/octet-stream
Size: 46258 bytes
Desc: not available
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20141125/2f664d4d/attachment-0001.obj>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: cell-c.png
Type: image/png
Size: 63318 bytes
Desc: not available
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20141125/2f664d4d/attachment.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: stress-zz.png
Type: image/png
Size: 71775 bytes
Desc: not available
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20141125/2f664d4d/attachment-0001.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: long-cell-c.png
Type: image/png
Size: 84855 bytes
Desc: not available
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20141125/2f664d4d/attachment-0002.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: IPCM-1.restart
Type: application/octet-stream
Size: 30717 bytes
Desc: not available
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20141125/2f664d4d/attachment-0002.obj>