I am quite new with LDA+U, so please forgive me if I say something of wrong.
I am trying to perform LDA+U calculation for GaAs.
After reading papers, I found a value for Ga in GaAs for U (d-states) which is ~ 20 eV (a huge value).
But I think this value is correct since VASP allows to correct only one state (d, in this case) at once.
I apply this value as U-J in the Dudarev approach and I do a single point calculation on a previously optimized (DFT/PAW/LDA) structure. What I got is a perfect agreement with experimental band gap (i.e. on Gamma Egap= 1.52 eV).
Now, for sake of comparison, I would like to optimize the structure from scratch at LDA+U level of calculation but what I got is a crash of the calculation.
I know that U-J =20 eV for d orbitals is a huge value and that leads to an exceptional shrinking of the cell. But, anyway I would like to have some explanations about the failure. Does it mean I can not directly fully optimize for GaAs?
I always obtain a message that tells me that POTIM should be increased (but now I am at POTIM=1.
This is my INCAR
SYSTEM = GaAs_PAW_LDA
ISTART=1
ISMEAR=-5
LReal=False
PREC=High
EDIFF=1E-5
EDIFFG=-0.01
ISPIN=2
NSW=100
ISIF=3
IBRION=1
ADDGRID=.True.
LDAU=.TRUE.
LDAUTYPE=2
LDAUL=2 -1
AMIX = 0.2
BMIX = 0.0001 ! almost zero, but 0 will crash some versions
AMIX_MAG = 0.8
BMIX_MAG = 0.0001 ! almost zero, but 0 will crash some versions
LDAUU=19.8 0.0
LDAUJ=0.0 0.0
ENCUT=287.594
ENAUG=531.356
#LMAXMIX=4
Thanks,
G
