big total drift

Message

lazewski · #1 Post by **lazewski** » Fri Jul 08, 2005 8:11 am

I would like optimize forces below 1E-4eV/A for phonon calculation. I can do this for a number of systems, however, for some particular ones I obtain big total drift (about 0.03eV/A). I read carefully VASP manual and VASP page FAQ and used all the tricks suggested there without sucess. Moreover, when I doubled k-mesh as well as increase FFT-mesh I obtained nearly identical total drift.

#2 Post by **admin** » Fri Jul 08, 2005 10:05 am

This is a hint that the FFT mesh is not sufficient. Usually it helps just to set
set
PREC = high
(please mind that the PREC= accurate is LOWER than PREC= high)
in INCAR.
All other cutoffs (NGX,... NGXF) are chosen quite reliably then
and need not be set explicitely, unless the drift of the forces remains
too large.
If so,
ENCUT or ENMAX should be set, but in any case do not set them to a
value which is lower than the highest ENMAX read from POTCAR (which is the default ENMAX input)

lazewski · #3 Post by **lazewski** » Fri Jul 08, 2005 12:08 pm

OK. As I mentioned before I have just done all of listed tricks: I set ENCUT 30% above ENMAX, increased NG(X,Y,Z) two times. I have also sufficient k-mesh dense (I believe) 16x16x16. Very strange for me is fact that total drift DID NOT change across my changes.

scerwin · #4 Post by **scerwin** » Wed Jul 27, 2005 3:23 pm

Also, if you are using real-space projection you might try switching it off. This is known to contribute to big drifts. See http://cst-www.nrl.navy.mil/~erwin/vasp ... tioncutoff for a more detailed explanation.