No Electronic Convergence for In_d PAW PBE dataset

[skasiraj]

Dear Vasp Users,

The system in consideration is a 4x4-Pd (111) surface slab with 2 In adatoms on the surface. The calculations were successful for the default Indium (In) PAW PBE potential with and without spin polarization turned on. The same calculation never converged electronically (multiple ionic loops with electronic iterations upto max NELM) when In_d PBE POTCAR (recommended in vasp website) was used.

Various strategies were used for the hard POTCAR potential setup to achieve convergence but none of them worked, i.e. :
1. Change ismear
2. Reduce EDIFF
3. Increase NELM to as high as 300 electronic iterations
4. Turn spin polarization on and off.
5. Changing the algorithm ialgo or algo tag.


The INCAR settings used are :
ENCUT = 400.000000
SIGMA = 0.100000
EDIFFG = -5.00e-02
PREC = Normal
GGA = RP
ALGO = Fast
ISMEAR = 0
NELM = 200
NSW = 500
IBRION = 2
NPAR = 4
IDIPOL = 3
DIPOL = 0.5000 0.5000 0.5000
LCHARG = .TRUE.
LWAVE = .FALSE.
LDIPOL = .TRUE.
LREAL = Auto

KPOINTS:
Monkhorst-Pack
3 3 1
0 0 0

PAW datasets used : PAW_PBE In_d 06Sep2000 and PAW_PBE Pd 05Jan2001

Please find the dropbox link which has the .tar file containing all the VASP files.

https://www.dropbox.com/sh/twh7wg5r4yud3s8/eKVZrtToFj

Any kind of help or support would be greatly appreciated.

Sashank Kasiraju,
PhD Student,
Dept. of Chemical Engineering,
University of Houston.

<span class='smallblacktext'>[ Edited ]</span>

[admin]

the last 2 columns in your vasp.out file indicate that the convergence of the charge density is very bad,
please try to
-- decrease the mixing parameters (AMIX, BMIX, eg to AMIX=0.1; BMIX=0.001) and do ONE ionic step to test whether it helps
(it makes no sense to continue with a predicted new ionic configuration if a step has not converged electronically)
-- set LMAXMIX = 4 (to account for proper charge density mixing of the d-electrons as well)