Page 1 of 1

SCF is not converging even if dE goes below EDIFF.

Posted: Tue Aug 22, 2023 5:03 pm
by bikash.patra
Dear experts,

I am doing an SCF calculation for a large supercell using VASP 6.1.0.
My EDIFF is 1E-8. I am pasting a part of my OSZICAR file below-

DAV: 73 -0.691969227346E+03 -0.61118E-09 0.25995E-07 2016 0.215E-06 0.105E-04
DAV: 74 -0.691969227345E+03 0.49477E-09 0.25913E-07 2016 0.245E-06 0.105E-04
DAV: 75 -0.691969227346E+03 -0.77489E-09 0.25996E-07 2016 0.276E-06 0.105E-04
DAV: 76 -0.691969227346E+03 -0.87311E-10 0.25810E-07 2016 0.316E-06 0.105E-04
DAV: 77 -0.691969227346E+03 0.30923E-09 0.26024E-07 2016 0.355E-06 0.105E-04
DAV: 78 -0.691969227346E+03 0.41473E-09 0.25890E-07 2016 0.382E-06 0.105E-04

As seen from the above steps, my SCF calculation is not converging even dE goes below EDIFF.
The SCF cycle runs up to the maximum NELM steps.
Why is this happening? Is it checking anything else except EDIFF?

My INCAR is the following-

ENCUT = 330
NELM = 150
PREC = Accurate
EDIFF = 1E-8
ISMEAR = 0
SIGMA = 0.05
LREAL = .FALSE.
LCHARG = .FALSE.
ADDGRID = .TRUE.
LWAVE = .FALSE.
KPAR = 4

with regards,
Bikash Patra,
TIFR, Mumbai.

Re: SCF is not converging even if dE goes below EDIFF.

Posted: Tue Aug 22, 2023 6:57 pm
by manuel_engel1
Hi,

The break condition specified by EDIFF only triggers if the difference in both the free energy and the band-structure energy is below the cutoff. Since you are using a large supercell, I suggest raising EDIFF to a less strict value.