Instability in Realtime Propagation since v2.6.0 (was working before)

[M. Brehm]

Dear Samuel,

thank you for your quick response and your fix in the code! I can confirm 
that now the input file is running fine with the current trunk version. 
Very nice for my project.

However, when I look at the temporal development of the dipole vector over 
time, there are a few quantitative differences between v2.5.0 and the 
current trunk. In my input file from above, there is a electric field pulse 
applied within the first propagation step, and the dipole moment oscillates 
afterwards. The components of the dipole vector are dumped to the log file. 
In the attachment to my post, I plotted the X component of the dipole 
vector on the Y axis and the propagation step on the X axis. The blue line 
corresponds to v2.5.0, the red line is from the trunk. The qualitative 
shape is identical, but it seems the pulse was somehow weaker in the 
current trunk version. I am not sure what causes this effect. Maybe one of 
your changes to the RTProp code can explain this difference.

Best regards,
Martin
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20150327/af2d581b/attachment.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: dip_x.png
Type: image/png
Size: 9839 bytes
Desc: not available
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20150327/af2d581b/attachment.png>