NEB with charged defect using vtstcode-205

[jordanchapman]

Hi all,

I'm attempting to find an energy barrier for the migration of an anionic vacancy in LiF. I've been able to run calculations with a neutral defect (i.e., the number of electrons matches the expected number from the POSCAR + POTCAR files), but I'm running into an issue when performing a similar calc with one fewer electron. Specifically, the electronic optimization initializes until NELM=5, when I get the following message from VASP:

BRMIX: very serious problems
the old and the new charge density differ
old charge density: 1073.00000 new 1072.00000

The end points were both geometry-relaxed with the following INCAR tags:

! initialization
System = LiF
ISTART = 1
ICHARG = 2
NCORE = 8
ISPIN = 2

NELM = 200
ENCUT = 400
PREC = Accurate
ISMEAR = 0; SIGMA = 0.01
EDIFF = 1E-10

! ionic relaxation
NELECT = 1072 ! 1073 neutral
IBRION = 3
NSW = 500
POTIM = 0.1
ISIF = 0
IOPT = 1
EDIFFG = -0.010

And the NEB routine I'm trying to run uses the following INCAR tags:

! initialization
System = LiF
ISTART = 0
ICHARG = 2
NCORE = 8
ISPIN = 2

NELM = 20
ENCUT = 400
PREC = Accurate
ISMEAR = 0; SIGMA = 0.01
EDIFF = 1E-10
ALGO = Normal

! ionic relaxation
NELECT = 1072 ! 1073 neutral
IBRION = 3
NSW = 500
POTIM = 0
ISIF = 0
IOPT = 1
EDIFFG = -0.010

! NEB
ICHAIN = 0
IMAGES = 5
SPRING = -5.0
LCLIMB = .FALSE.

I'm a bit perplexed by this error message. The endpoints were run with the same number of electrons, and converged to the above specs; I verified in the INCAR that there were 1072 electrons in the final electronic configuration. I tried running the NEB calc with ICHARG=0 + ISTART=0, but ran into the same error. I also tried changing the ALGO to Fast and VeryFast. If anyone could provide any insight as to why VASP isn't reading NELECT from the NEB INCAR file (or if there's an oversight I'm not seeing), I would greatly appreciate it.

Jordan Chapman
VTNSI

[graeme]

If you post the calculation, I can take a look, but I expect that you need to remove and CHG* and WAV* files. It will read those files by default and if you have changed the number of electrons in your INCAR file, vasp will complain.

[jordanchapman]

Thanks for getting back to me quickly. I've attached a zip file containing the parent directory for the NEB calcs.

To your reply, the WAVECAR and CHGCAR files weren't generated at any point in the runs for the intermediate geometries. I included the OUTCAR files of the endpoints, too. I've linked the VTST source code with version 6.4.2 of VASP, FYI.

[graeme]

The problem is that in your initial band, you have atoms running into each other. Take a look at the initial movie in a viewer and you will see the problem. Also, there is a warning in each ??/OUTCAR that atoms are too close - you can see it in the neighborlists.

The problem is a result of atom ordering; make sure that each atom in your initial structure corresponds to the same atom in the final structure.