SCF calculation not converge with VASP 6.3.0

[suojiang_zhang1]

Hi，
I am using VASP6.3 to calculate a CeO2 with Sm-doped system. The INCAR file is
#Basic parameters
ISYM=0
ISTART=0
ICHARG=2
ALGO=Fast
ECUT=500
EDIFF=1E-5
EDIFFG=-0.02
PREC=Accurate
ISMEAR = 0
SIGMA = 0.1
LREAL = Auto
NELM = 100
NELMIN = 4
#For DFT+U
LDAU=T # O Ce Sm O Fe Fe Sr La
LDAUU= 0 5 5 #on site Cloulomb interaction parameter
LDAUJ= 0 1 1 #on site Exchange interaction parameter
LDAUL= -1 3 3 #-1:no term; 1:p; 2:d; 3:f .
LMAXMIX=6

#For spin polarization
ISPIN=2
AMIX=0.02 # 0.02-0.2
BMIX=0.0001
AMIX_MAG=0.8
BMIX_MAG=0.0001
MAGMOM= 64*0 7*7 25*7

LWAVE=.F.
LCHARG=.F.

IBRION = 2
ISIF=2
NSW = 500
POTIM = 0.2

and I found it is difficult to converge at NELM=100 and ALGO=Fast. But finally it is convergent to the force=0.02.
What follows that makes me hard to understand is that I copied the CONTCAR to POSCAR to single point SCF calculation at NSW=0,
I cannot finish the converging even NELM=500 and ALGO=N,
please give me some advice.

[fabien_tran1]

Hi,

Your explanations are a bit confusing. It is not clear what procedure worked or not for your case. Could you please describe more clearly the steps that you have followed? We would need also to see how the OUTCAR of a calculation that did not converge. Finally, if you want us to try, please provide also the input files KPOINT and POSCAR.

[suojiang_zhang1]

thank for your reply.
After I finished the ion relaxation, I calculated a scf single point based the optimized structure.
But it is very hard to converge even NELM=500.
Attached the OUTCAR, KPOINT,POSCAR

[fabien_tran1]

Your system is magnetic with rare-earth atoms. With such systems this may sometimes be tricky to achieve SCF convergence. Since I did not see anything wrong in your input files and there is no unique trick which works in all cases, I can only suggest you to try other algorithm schemes or to change values of parameters like AMIX or BMIX (but I guess this is what you have already done to some extent). I also suppose that you know this page: https://www.vasp.at/wiki/index.php/Cate ... ity_mixing.

What you could also try is to do first a PBE calculation (without U and supposing that it can be converged) and then start the PBE+U calculation using the WAVECAR generated with PBE. For this you need to do the PBE calculation with "LWAVE=.TRUE." and the PBE+U one with "ICHARG=0".