Very long calculations.

Message

GP · #1 Post by GP » Sat Sep 21, 2013 7:20 pm

Hello,

I have a problem running VASP: my calculations are extremely long. It can last two weeks and several thousands of iterations without reaching the "required accuracy". Usually, it runs during 50-70 hours without problem and seems to converge (visually in VMD and in the OSZICAR file). But then, I have an error like "rerun with smaller EDIFF or copy CONTCAR to POSCAR and continue". So I do that but the error appears again very fast, sometimes after only 1 hour.

Then, my calculations are very long and do not reach the final energy although the atoms barely move...

What should I do? I have set this in the INCAR file:

PREC = Normal
ENCUT = 400.
NSW = 2000
ISIF = 0
IBRION = 2
LREAL = Auto
AlGO = Fast
ISMEAR = 0
SIGMA = 0.05
SYMPREC = 1e-4
LPLANE = .True.
LSCALU = .False.
EDIFFG = -0.007
EDIFF = 0.0001

I am a beginner in VASP and some of the options were just given to me like this. I just tried to change EDIFFG to a higher value (it was 0.005 before) because I thought the forces could never reach this value but I am not sure that this is the problem... I am modeling molecules interacting with a surface of silica.

Thank you for any help.
<span class='smallblacktext'>[ Edited ]</span>

#2 Post by **admin** » Mon Sep 23, 2013 1:14 pm

Check in OUTCAR forces acting on atoms. They probably stay always larger than your stopping criterion 0.007. Use less severe stopping criterion, e.g. EDIFFG = -0.03.

GP · #3 Post by GP » Mon Sep 23, 2013 6:31 pm

Ok thank you. But it is not a problem for the coherence of my calculations? I know that if I change the ENCUT, there is no point in making a comparison or compute an adsorption energy. The precision will not be lost if I want to substract : E(final) - E(initial) where E(initial) has been calculated with a lower EDIFFG?