My Community

Hello everyone,

I'm trying to do some calculations on charged supercells and I'm having quite a bit of energy cutoff convergence trouble. The system is pure bcc Cr. I suppose that a normal metal would never be charged by any significant amount in real life, so I'm not too surprised that things don't go as smoothly as usual. I've also read various posts in the forum that report difficulties with charged systems. Electronic convergence is pretty easy though. And the system is just plain and simple, defect-free bcc Cr, the extra charge should not be concentrated around defects, requiring multipole corrections etc. The neutral system has 768 electrons. When a system with a full electron added/removed didn't converge with rspect to the planewave cutoff, I tried adding/removing 0.2 and 0.05 electrons. For a metal adding/removing non-integer numbers of electrons is no more unrealistic than adding/removing a full electron, right? But even with a small charge addition/depletion the system total energy would not converge wrt to planewave cutoff. Even when increasing the cutoff from 600 -> 800 eV (the default ENMAX is 227 eV) the energy changed 286 meV, or 2.23 eV/atom.

Does anyone know how to make it converge, or is anyone knowledgeable enough to say that charged metals are definately just too unrealistic for this ever to work?

I include info on the input files and a sample OSZICAR file below.

greets,
Peter



INCAR (ENCUT and NELECT have been tried with different values):
Cr 4x4x4 cells reference system

ISTART = 0

ENCUT = 800.0
IALGO = 38
PREC = medium
LREAL = .TRUE.
ISMEAR = 1
SIGMA = 0.2
VOSKOWN = 1

ISPIN = 2
MAGMOM = 64*1 64*-1

LWAVE = .FALSE.

NELECT = 767.95


KPOINTS was just the same MonkhP, 3x3x3 points, no special shifts

POSCAR: just a perfect 4x4x4 cells bcc system

POTCAR: PAW-PW91, 6 valence electrons

OSZICAR:
N E dE d eps ncg rms rms(c)
DAV: 1 0.850366131746E+04 0.85037E+04 -0.34802E+05 7424 0.279E+03
DAV: 2 0.244897251221E+03 -0.82588E+04 -0.81262E+04 7424 0.539E+02
DAV: 3 -0.117115851517E+04 -0.14161E+04 -0.12930E+04 10680 0.278E+02
DAV: 4 -0.139087951876E+04 -0.21972E+03 -0.21075E+03 12368 0.109E+02
DAV: 5 -0.140714067535E+04 -0.16261E+02 -0.15729E+02 11896 0.295E+01 0.895E+01
DAV: 6 -0.121718147304E+04 0.18996E+03 -0.89335E+02 9560 0.102E+02 0.155E+01
DAV: 7 -0.121371865323E+04 0.34628E+01 -0.21076E+01 11848 0.977E+00 0.591E+00
DAV: 8 -0.121262387393E+04 0.10948E+01 -0.16232E+00 10824 0.352E+00 0.445E+00
DAV: 9 -0.121207412032E+04 0.54975E+00 -0.94554E-01 12288 0.229E+00 0.125E+00
DAV: 10 -0.121213673918E+04 -0.62619E-01 -0.31847E-01 12304 0.121E+00 0.233E+00
DAV: 11 -0.121212101844E+04 0.15721E-01 -0.27654E-02 11952 0.485E-01 0.235E+00
DAV: 12 -0.121225015923E+04 -0.12914E+00 -0.19540E+00 12408 0.311E+00 0.141E+00
DAV: 13 -0.121231231910E+04 -0.62160E-01 -0.47553E-01 12240 0.144E+00 0.140E+00
DAV: 14 -0.121238135214E+04 -0.69033E-01 -0.78566E-01 12216 0.182E+00 0.103E+00
DAV: 15 -0.121244489513E+04 -0.63543E-01 -0.29788E+00 12136 0.358E+00 0.122E-01
DAV: 16 -0.121245619351E+04 -0.11298E-01 -0.90006E-02 11360 0.801E-01 0.118E-01
DAV: 17 -0.121245105996E+04 0.51335E-02 -0.14608E-01 12048 0.866E-01 0.369E-01
DAV: 18 -0.121245601194E+04 -0.49520E-02 -0.18147E-01 12280 0.967E-01 0.273E-02
DAV: 19 -0.121245632875E+04 -0.31681E-03 -0.45916E-03 11848 0.189E-01 0.721E-02
DAV: 20 -0.121245647256E+04 -0.14381E-03 -0.84424E-03 11856 0.219E-01 0.138E-02
DAV: 21 -0.121245646648E+04 0.60846E-05 -0.15974E-03 11056 0.101E-01 0.377E-02
DAV: 22 -0.121245648151E+04 -0.15031E-04 -0.24621E-03 10768 0.114E-01 0.172E-02
DAV: 23 -0.121245648031E+04 0.11908E-05 -0.12243E-04 4928 0.451E-02
1 F= -.12124565E+04 E0= -.12125254E+04 d E =0.206717E+00 mag= 0.0000

charged periodic systems diverge unless multipole corrections are added. Please read the corresponding chapters in the online handbook for this (Monopole, Dipole and Quadrupole corrections, and Dipole corrections for defects in solids) Try to add LDIPOL=True in INCAR to correct for this, but please check first if the perturbation (additional charge) is localized somehow (which I am afraid is not in a metallic system) .

Suggestion:
Did you try to converge a neutral simple bulk Cr with an atom per unit cell as a function of the cutoff energy? Try it first, it can helps to understand your problem.

Juarez