conversence difficulty for ismear=-5 for insulator (large system)

[okuno]

Dear Vasp Users.

I'm now compare the total energy of insulator system which are different
atom position.

My system has 40 atom and is insulator ( one impurity and we neutlizing
by back ground charge by NELECT, ant the neutlized system is
insulator.) .

I move one atom from the centro-symmetric position littile by little.
In order to get smooth energy- atom position relation,
I took ismear=-5. But the calculation has difficulty for conversence.
It took more than 50 cycle SCF for static electronic calculation, and some
case it does not reach conversence.
If I took ismear=0, I can get conversence results. But I worry about
the accuracy of total enegy for ismear=0.

For accurete total energy calculation, how we sets input parameter
by ismear=-5 (tetrahedron method) for big insulating system.

My input is as follow,

SYSTEM = SYSTEM_NAME

Startparameter for this Run:
NWRITE = 2; LPETIM=F write-flag & timer
#ISTART = 1 job : 0-new 1-cont 2-samecut
#ICHARG = 1 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 = 1.0E-05 # stopping -criterion for ELM
PREC = High

NELECT=352 # for charge of the system

NELMIN=8

LREAL=Auto

NELM =100 # maximum electronic SC steps (default=60)

GGA = 91 # GGA PB PW LM 91 PE RP

ISMEAR = -5
SIGMA = 0.1

[admin]

charged periodic systems are only conditionally convergent. Please read the manual, chapter 'Dipole corrections for defects in solids' on how to calculate such kind of systems.