Convergence of defect supercells containing Yb 3+

[dsom]

Hello all,

I'm trying to minimize the energy of a Yb defect substituting on a Pb sublattice site in perovskite CsPbCl3 with different charge states. I have thus far been unable to obtain either electronic or ionic convergence. I have tried tuning several INCAR settings (including NELMDL, IMIX, AMIX, BMIX and AMIN), changing the supercell size, perturbing the initial atomic positions to break symmetry, and changing the Yb POTCAR. Most runs tend to throw the "BRMIX: very serious problems \ the old and the new charge density differ" warning within the first few self-consistent electronic steps. I am using Gamma-centered grids, typically only the Gamma k-point for larger cells, and the PBE.54 POTCARs.

My current INCAR for the neutral cell can be found below:


INCAR:

ADDGRID = True
ALGO = Normal
EDIFF = 0.0001
EDIFFG = -0.01
ENCUT = 400
GGA = PS
IBRION = 2
ICHARG = 1
ISIF = 2
ISMEAR = 0
ISPIN = 2

LASPH = True
LDAU = True
LDAUJ = 0 0.67 0.0 0
LDAUL = -1 3 2 -1
LDAUPRINT = 1
LDAUTYPE = 2
LDAUU = 0 7.0 3.6 0
LMAXMIX = 6

LORBIT = 11
LREAL = Auto
LWAVE = False
NCORE = 4
NELM = 200
NSW = 99
POTIM = 0.05
PREC = Accurate
SIGMA = 0.05

#MAGMOM = 16*0.0 1*1.0 63*0.0
#NUPDOWN = 1


The head of the current POSCAR is:

Cs16 Yb1 Pb15 Cl48
1.0
15.875458 0.000000 0.000000
0.000000 16.128863 0.000000
0.000000 0.000000 11.421408
Cs Yb Pb Cl
16 1 15 48


Any help is is appreciated. Thanks!

[dahvyd.wing]

If this is only showing up intermittently, then it could be an mpi problem. What worked for me was setting:
export I_MPI_ADJUST_REDUCE=3
in my job submission file. See https://www.nsc.liu.se/software/install ... lith/vasp/