What is the total drift?

Message

pshea · #1 Post by **pshea** » Fri Jun 07, 2013 12:24 am

I have a basic question: what is the "total drift" printed in the OUTCAR file? That is, how is it calculated? I thought initially it was just the sum of forces on the ions (which should be 0), but when I check, this sum is always 0 to within the precision of the numbers printed in OUTCAR (10^-6).

I ask because I'm trying to calculate vibrations frequencies, and so I need accurate forces. The manual mentions that total drift is a measure of how accurate the forces are, and it should be smaller than the forces you are interested in. However, the total drift is always around 10^-2 in my calculation, and no parameters seem to change it (including PREC, LREAL, ENCUT and number of K-points). Maybe it's not possible to get the total drift smaller than this? Any insight would be appreciated.

#2 Post by **admin** » Fri Jun 07, 2013 10:02 am

Try the following:
http://cms.mpi.univie.ac.at/vasp-forum/ ... .php?4.257

pshea · #3 Post by **pshea** » Fri Jun 07, 2013 4:33 pm

Thanks for the reply. I looked at the thread you linked, and I seem to be having the same trouble as the person there was having. The largest total drift is always around 10^-2, and increasing the FFT mesh and the energy cutoff doesn't seem to have any effect. I've also tried switching off the real space projection to no effect. If it makes a difference, I'm using a fairly large unit cell (about 12x10x16 A) and 84 ions.

Also, I'm still wondering how the total drift is calculated. I can't seem to find anything in the manual. Thanks again.

tlchan · #4 Post by **tlchan** » Fri Jun 14, 2013 1:37 pm

More accurate forces can be obtained by having better self-consistency in the Kohn-Sham equations. You can try to set a much smaller EDIFF.