Total drift

Message

The Evasperator · #1 Post by **The Evasperator** » Wed Oct 18, 2006 9:11 pm

The total drift for a relaxation calculation (IBRION=1, ISIF=3) that I recently performed was larger than the maximum force that I specified by EDIFFG. After reading the VASP manual, I discovered that this is a problem. My questions are:

(1) What is the meaning of drift and how is it calculated?

(2) The manual states that, among other things, drift may be reduced by increasing the FFT grid and/or by increasing ENAUG to 4*ENCUT. I am currently using ENCUT=700 eV, and I have been told that my ENCUT value alone should eliminate any need to change either ENAUG or NG(X,Y,Z). Is this true?

#2 Post by **admin** » Thu Oct 19, 2006 12:56 pm

another reason for drifts (if you have set PREC=accurate, and all ENCUT,.... parameters high enough) may show up for systems containing vacuum in the unit cell : in such a case (free clusters, molecules, surfaces,...), please keep at least one atom (or layer, if you do a slab) fixed to prevent a drift through the vacuum.: use the
Selective Dynamics
option in POTCAR and set F F F for at least one atom (layer)
(please have a look at the online handbook for a more explicit explanation of the Selective dynamics feature)

The Evasperator · #3 Post by **The Evasperator** » Fri Oct 20, 2006 2:06 pm

you mean POSCAR for Selective Dynamics, right?

The Evasperator · #4 Post by **The Evasperator** » Fri Oct 20, 2006 2:22 pm

I tried the selective dynamics. Unfortunately, the job still had drift values that were too high.

Originally, we thought that drift was the total force. However, we summed the forces and found that the total was zero. If drift is not the sum of forces, what exactly is it?