BRMIX: old and new charge density differ warning for doping calculations

[sophie_weber]

Hello,

I am trying to perform calculations with a slab geometry, using a material that is by default insulating. I want to explore the effect of uniform electron doping and I am doing so by changing the value of NELECT, and initially relaxing the internal positions for each value. However, for any value other than the default, uncharged value, I get a warning in the output file after about the fifth electronic iteration saying "BRMIX: very serious problems
the old and the new charge density differ
old charge density: 552.00000 new 553.00000"

However, after this warning the calculation continues and, as far as I can tell, converges perfectly fine. I have a pretty high k-mesh (17x6x1) for a 4.9 x 13.5 x 30 angstrom slab (last direction has 20 angstrom vacuum) and I'm using gaussian smearing with a fairly large sigma=0.1 (these are the things are saw in previous posts which would be carefully checked when trying to use doping with NELECT). Given this, and the following INCAR, I was wondering whether I can probably safely ignore the warning if the output seems reasonable, or if I need to change something to get rid of it? Thanks in advance.

ENCUT = 910.000000 #700*1.3 for relaxations
EDIFF = 1.00e-05
ISMEAR = 0
SIGMA=0.1
NELM=1000
ADDGRID=.TRUE.

LMAXMIX = 4

LORBIT = 11
NCORE = 18

LDAUPRINT=2
LDAUTYPE=2
LASPH = .TRUE.
LDAU = .TRUE.
LDAUL = 2 -1
LDAUU = 4.000 0.000
LDAUJ = 0.000 0.000

ISPIN=2
MAGMOM= 3 -3 -3 3 3 -3 3 -3 -3 -3 3 3 3 -3 3 -3 -3 -3 3 3 3 -3 3 -3 36*0

IDIPOL=3

NELECT=553 #adding one electron to system

ISIF=2
IBRION=1
EDIFFG=-0.01
NSW=2000

[pedro_melo]

Dear sophie_weber,

Could let me know which version of VASP you are using?

Kind regards,
Pedro Melo

[sophie_weber]

Dear Pedro,

Yes, the version is VASP 6.2.0

Best wishes,
Sophie

[sophie_weber]

Hmm, just as an update, I also noticed that when I dope with one electron, I get some weird binary section printed in the out file a few steps after the BRMIX warning, and then afterward the minimization algorthim seems to switch
LDA part: xc-table for Ceperly-Alder, standard interpolation
POSCAR, INCAR and KPOINTS ok, starting setup
FFT: planning ...
WAVECAR not read
entering main loop
N E dE d eps ncg rms rms(c)
DAV: 1 0.169253767524E+05 0.16925E+05 -0.28525E+05 80880 0.574E+03
DAV: 2 0.395901298120E+04 -0.12966E+05 -0.12369E+05 53340 0.116E+03
DAV: 3 0.335812344639E+03 -0.36232E+04 -0.34411E+04 63450 0.599E+02
DAV: 4 -0.461225967505E+03 -0.79704E+03 -0.76815E+03 89960 0.331E+02
DAV: 5 -0.538775140227E+03 -0.77549E+02 -0.72365E+02 93020 0.108E+02BRMIX: very serious problems
the old and the new charge density differ
old charge density: 552.00000 new 553.00000
0.109E+02
DAV: 6 -0.471068719772E+03 0.67706E+02 -0.13331E+03 89970 0.128E+02 0.108E+02
DAV: 7 -0.472455133986E+03 -0.13864E+01 -0.37820E+02 90510 0.926E+01 0.434E+01
DAV: 8 -0.479309876709E+03 -0.68547E+01 -0.12235E+02 96340 0.388E+01 0.243E+01
^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@
CGA: 53 0.118917352735E+06 -0.14234E+04 -0.13026E+04 6800 0.385E+04-0.102E+03
gam= 0.489 trial= 0.328 step= 0.424 mean= 0.328
gam= 0.460 g(H,U,f)= 0.160E+04 0.162E+03 0.672E-03 ort(H,U,f) = 0.798E+03-0.234E+03 0.406E-03
CGA: 54 0.118069572175E+06 -0.84778E+03 -0.80393E+03 6800 0.177E+04 0.565E+03
gam= 0.460 trial= 0.333 step= 0.714 mean= 0.333
gam= 0.910 g(H,U,f)= 0.142E+04 0.183E+03-0.742E-17 ort(H,U,f) = 0.130E+03 0.216E+02 0.000E+00
CGA: 55 0.117316536215E+06 -0.75304E+03 -0.72371E+03 6800 0.161E+04 0.152E+03
gam= 0.910 trial= 0.352 step= 0.913 mean= 0.352
gam= 1.036 g(H,U,f)= 0.148E+04 0.181E+03-0.240E-13 ort(H,U,f) = 0.445E+01-0.812E+02 0.209E-11
CGA: 56 0.116524727785E+06 -0.79181E+03 -0.79670E+03 6800 0.167E+04-0.767E+02
gam= 1.036 trial= 0.380 step= 0.946 mean= 0.380

This does not occur for larger numbers of electrons added, it sticks to the Davidson algorithm, and converges happily without any weird printing.