determinant error in Berry Phase calculation

[okuno]

Dear Vasp Users.

I'm now calculationg polarization by Berry Phase method.
But in some cases ( not all the case), I met the error like

Error in subroutine BERRY: did not find all determinants

Matrix CMK is not an nxn matrix for
ISTR = 4 j = 4

, and calculation stopped.

When such an error has happened, how should I do to get the results?


My INCAR file for scf is

SYSTEM=SYSTEM_NAME

ENCUT=520.0
PREC=High
ISMEAR=0
SIGMA=0.1
EDIFF=1E-05
NELMIN=8

LWAVE=.FALSE.
LORBIT=11

and for the Berry Phase calculation , my INCAR file is

Startparameter for this Run:
NWRITE = 2; LPETIM=F write-flag & timer
ICHARG = 11 initial charge 0-calc 1-from CHGCAR, 2-super potision


Electronic minimization
ALGO = NORMAL # default for IALGO=38
IALGO = 38 # default Davodason block iteration scheme
EDIFF = 1E-06 # stopping -criterion for ELM
PREC = High
ENCUT=520.0

NELM =60 # maximum electronic SC steps (default=60)


ISMEAR = 0
SIGMA = 0.1

#Berry Phase
LBERRY = .TRUE.
IGPAR = 3
NPPSTR = 15
DIPOL = 0.50 0.50 0.50

[admin]

1) please have a look (at your eigenvalue spectrum, written in OUTCAR) if your system really is insulating
2) NPPSTR may be chosen too large, please reduce it (6-8 usually are reasonable values)

[okuno]

Dear administrator.
Thank you for your suggestion.
I found the error case is not insulating case, it become metalic.

I want to calculate the absolute value of spontanious polarization,
so I calculate the centrosymmetric case ( which become metal), and ferromagnetic atomic position.

If we want to calculate spontanious polarization of the system and
the reference centrosymmetric case become the metal, how should we try to calculate polarization?

Should we introduce some adiabatic parameter like "U" (LDA+U)
to maintain insulating state , or some other method exist ?

Sincerely.