My Community

Hi,

I am a new VASP-5.3.3 user and am trying to get the SCF-convence of YIG system (containing 12 Y, 20 Fe, 48 O atoms; geometry is already optimized using VASP). This is a LSDA calculation where I set default ENCUT (which I think is its max limit reading from POTCAR), 7x7x7 Monkhorst Pack, also I tuned mixing parameters as it is suggested for magnetic/insulating system. However, I an not able to get the SCF-convergence. Can anyone advise to resolve this issue? Many thanks!

Please see the last few steps of TOTEN:

free energy TOTEN = -713.87388643 eV
free energy TOTEN = -713.85440342 eV
free energy TOTEN = -713.88590381 eV
free energy TOTEN = -714.01187213 eV
free energy TOTEN = -714.00566434 eV
free energy TOTEN = -713.96289057 eV
free energy TOTEN = -713.96186526 eV
free energy TOTEN = -713.96186526 eV

And, the INCAR file:

general:
System = bcc YIG (Y3Fe5O12)

# parallization
LPLANE = .FALSE.
NPAR = 4 # sqrt(number of nodes)
NSIM = 1

ISTART = 1 # (0 = not reading WAVECAR;
# 1=read orbitals from file WAVECAR + keep energy cutoff constant;
# 2=read orbitals from file WAVECAR + keep basis constant.)
ICHARG = 1 # (2 = if ISTART=0;
# 0=Calculate charge density from initial orbitals;
# 1=read CHGCAR;
# +10=non-scf calculation)
NELM = 80 # max scf steps

ALGO=Fast # Normal # Fast
EDIFF=1.0E-4 # Energy tolerance
PREC=Normal
#ENCUT = 400 # cutoff energy in eV
ISMEAR = -5
#LORBIT = 11 # not reading DOSCAR and PROCAR file with phase factors
EMIN = -9 # min energy for DOS
EMAX = 11 # max energy for DOS
NEDOS = 401 # energy points for DOS

# Mixing parameters (this choice for for slabs, magnetic systems and insulating systems)
AMIX = 0.2 # linear mixing parameter
BMIX = 0.0001 # cutoff wave vector for Kerker mixing scheme
AMIX_MAG = 0.8 # linear mixing parameter for magnetization
BMIX_MAG = 0.0001 # cutoff wave vector for Kerker mixing scheme for mag.

# Spin
ISPIN = 2
MAGMOM = 12*0.0 20*1.0 48*0.0

LREAL = .FALSE. # .TRUE.=real-space projection scheme, helps
# for larger cell but may compromize accuracy

1. I never use ALGO=F for difficult problems.
2. In LDA+U calculations, LMAXMIX tag is important.