I have just implemented VASP code with Gamma K-point, POSCAR of 128-atom supercell and INCAR like this:

   ISTART = 2
   ISPIN  = 2
   ICHARG = 1
   MAGMOM = 2*3 27*4 3 2*4 64*0 29*0
   ISMEAR = 0
   SIGMA  = 0.05
   NELMIN = 3
   NELM   = 60
   EDIFF  = 1E-05
   ISIF   = 2
   NSW    = 60
   IBRION = 3
   POTIM  = 0.1
   EDIFFG =-E-02

The problem is when the electronic relaxation reaches EDIFF, the electronic step does not stop. I just wonder if there is any other convergence requirement needed to be reached so that the electronic relaxation step stops. enter image description hereenter image description here


1 Answer 1


It is not converged, it needs to hit |-1E-5| on the fourth and fifth column. It gets close on the fifth column, but never quite makes it there. The lowest I see is about |-1.7E-5|. It seems to be struggling, if this isn't an optimized structure then you could lower EDIFF to 1E-4 or even 5E-4 for some initial steps...Or simply do not worry about it and continue on regardless, over time you will converge faster with each geometric step.

However, if this is a single point job then you might need to change to ALGO=ALL for more stable convergence or change mixing settings.

  • 1
    $\begingroup$ I set EDIFF = 1E-05. I calculated another supercell using the same INCAR, POTCAR and K-point, and it stopped when dE = -0.81824E-05 (in the second figure). So I think there may be another convergence requirement. $\endgroup$
    – Binh Thien
    Jan 19, 2021 at 1:54
  • $\begingroup$ I used a unit of 0.1E-5 not 1E-5, it makes it easier to read the OSZICAR directly and think about it. This being said, I will update my answer as I realized that my original though was likely incorrect. $\endgroup$ Jan 19, 2021 at 3:43

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .