BRMIX: very serious problems during ionic relaxation of a 216-atom ATAT-generated SQS supercell with DFT+U
Dear VASP developers,
I am trying to perform an ionic relaxation on a 216-atom SQS supercell generated using ATAT. The system is a transition-metal oxide, where one transition-metal species is treated using DFT+U. The supercell is required to model a substitutional dopant concentration of 8.33%, which cannot be represented in the primitive cell. Therefore, I have not attempted relaxation of the primitive cell.
The calculation fails during the electronic SCF of the first ionic step and repeatedly produces the following error:
DAV: 66 -0.162254636541E+08 -0.13384E+08 -0.30348E+07 45312 0.127E+03
BRMIX: very serious problems
the old and the new charge density differ
old charge density: 1016.10899 new 1082.85917
0.297E+03
DAV: 67 -0.618439635378E+07 0.10041E+08 -0.36107E+07 59024 0.149E+03
BRMIX: very serious problems
the old and the new charge density differ
old charge density: 1082.85917 new 1054.74188
0.353E+03
DAV: 68 -0.472028896200E+07 0.14641E+07 -0.10767E+06 47648 0.136E+03
BRMIX: very serious problems
the old and the new charge density differ
old charge density: 1054.74188 new 1041.23457
0.331E+03
My INCAR is:
Global Parameters
ISTART = 0
ISPIN = 2
ICHARG = 2
LREAL = .FALSE.
ENCUT = 520
PREC = Accurate
LWAVE = .TRUE.
LCHARG = .TRUE.
ADDGRID= .TRUE.
LASPH = .TRUE.
MAGMOM = 12*0 24*5 12*0 24*0 144*0
GGA = PE
ISYM = 0
Static Calculation
ISMEAR = 0
SIGMA = 0.05
NELM = 300
EDIFF = 1E-05
AMIX = 0.2
BMIX = 0.0001
AMIX_MAG = 0.8
BMIX_MAG = 0.0001
MAXMIX = 80
NSW = 300
IBRION = 2
ISIF = 2
EDIFFG = -2E-02
POTIM = 0.3
NCORE = 6
DFT+U Calculation
LDAU = .TRUE.
LDAUTYPE= 2
LDAUL = -1 2 -1 -1 -1
LDAUU = 0 4.3 0 0 0
LDAUJ = 0 0 0 0 0
LMAXMIX = 4
I have already tried the following without success:
ISYM = 0 and ISYM = -1
ALGO = Fast, Normal, and VeryFast
different AMIX, BMIX, AMIX_MAG, and BMIX_MAG values
loser EDIFF and EDIFFG
lower ENCUT
The electronic SCF never converges, so the ionic relaxation does not proceed. Could anyone suggest possible causes for this behavior or recommend additional settings that may improve convergence?