which is the preference for EDIFFG flag ?

dwyuan · Post by **dwyuan** » Tue Jun 13, 2006 4:31 pm

Dear all:

When I calculate the dissociation reaction of chemical molecules on
metal surfaces, how to select the convergence flag for ionic relaxation?
Force or energy convergence, which is the preference?

thanks !

andri · Post by **andri** » Tue Jun 13, 2006 4:38 pm

Converging the force to zero is more appropriate I think. So I commonly use:

EDDIF = 1.0E-4
EDDIFG = -1.0E-3

hwang4 · Post by **hwang4** » Wed Jun 21, 2006 2:15 am

Is EDDIFG = -1.0E-3 too small? I tested a case with EDIFFG=-0.05 with another case EDIFFG=-0.001, they gave me almost same results in term of total energy and the dissociation barrier calculated by NEb. But calculation with smaller EDIFFG requires much more calculations. Maybe it is only true for my system?

andri · Post by **andri** » Wed Jun 21, 2006 2:39 am

I usually like to converge the forces to values lower than 0.03 but that's of course just my own preference and not an empirical truth. When I start a new NEB calculation I often start with setting EDDIFG = -0.01. If that converges I decrease it to -0.005 and continue. I then continue in this fashion until the forces are appropriately low or the energy is good as constant.