florian.s... at mat.ethz.ch
Tue Nov 13 12:19:47 UTC 2012
I did run your system with the current svn code and the optimization is
stable. As pointed out before, I recently introduced a better safeguard for
the model hessian approach which seems to solve the problem.
As far as I can see the problem was related to massive displacements
leading to bad overlaps which then leads to a problem in DFT.
Therefore I recommend to try the latest version of the code (little
reminder, the svn location has changed). If you don't have your own
version, Use the three settings described in my previous post.
In addition, maybe a little hint for your input file:
For big systems FULL_ALL becomes very expensive. For your system I would
recommend using FULL_SINGLE_INVERSE. It is slightly less efficient but much
cheaper than FULL_ALL
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the CP2K-user