[CP2K:5047] Re: restarting from an already optimized structure

[Florian Schiffmann]

I see the problem. You have a farily big cell (38x38x38) and going to high 
cutoffs is likely to cause a memory issue. The problem is, BLYP 
pseudopotetnials need very high cutoffs to be smooth with the FFT's. I 
don't know what your system looks like exactly but the only way out here as 
far as I can tell is to use a smaller cell if you can't afford the settings 
as they are. I don't know what the system looks like (dimensionwise) but it 
might be possible to use MT and 2D periodic boundaries with a smaller box 
permitting you to use higher cutoffs. Do you mind sending me a snapshot or 
position file?

If you don't use the center coordinates keyword with the final geometry the 
positions should be fine (and optimal at least with the PW basis used but 
wrong compared to the basis set limit). Furthermore if you want to do 
energy comparisons (reactions or whatever) you definitely need a better 
convergence of your PW basis otherwise there will be a position depndent 
(relative to the gridpoints) error in all your energies.
About the flucuations in the energy. BFGS is very sensitive to the ripples 
in the PES and only behaves well if they are down to a reasonable magnitude 
(see above more expensive but better convergence).  

Flo
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20140308/05b96257/attachment.htm>