[Pw_forum] stress in CP and FFT grids

[Konstantin Kudin]

 Hi all,

 It appears that the stress that comes out of the CP code is rather
sensitive to the FFT grids used.

 For example, I ran a constant pressure simulation ('vc-cp') started
from a certain unit cell. Naturally, the code kept all the FFT grids
unchanged throughout (generated via the initial cell dimensions), and
optimized the cell vectors to certain values.

 Then, when I started a fresh simulation with the optimized cell
dimensions, the stress was quite a bit different, on average about 0.6 
GPa higher than before. Looking at the settings, it appears that the
FFT grids have changed because of the fresh start, and that made the
stress different.

 Any suggestions on how to deal gracefully with such situations,
especially when the system is "soft" along certain cell coordinates? It
hardly comforting to have about 5% variations in the cell parameters
due to the grids ...

 Thanks!
 Kostya
 
 

__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com