My Community

Dear all,

I need to calculate a charge ordering for a insulating Fe3O4 using GGA+U method. Setting same U parameters on all Fe site (even though there are two different Fe sites, namely Fe+2 and Fe+3 in the unit cell, only one Fe type is considered in the calculations) does not give correct electronic structure, i.e., a band gap (The U is varied as well). In some literatures, two different U (J) parameters are used for two different Fe sites as mentioned before, resulting insulator. My question is how we devide Fe's sites into two charge ordered sites of +2 and +3 in input files?

Post your INCAR, you should be able to do this easily via your MAGMOM TAG. Give the appropriate number of unpaired electrons for each magnetic ion in an order corresponding to your POSCAR.

Thank you for your reply. My INCAR and POSCAR are as below,

SYSTEM = Fe3O4
NWRITE = 2
ISTART = 0
ICHARG = 2
INIWAV = 1
ISPIN = 2
MAGMOM = 2*-4.0 4*4.0 2*-4.0 4*4.0 2*-4.0 4*4.0 2*-4.0 4*4.0 32*0
LORBIT = 11
PREC = High
NELM = 100
EDIFF = 1E-05 NSW = 50
EDIFFG = 1E-03
IBRION = 2
POTIM = 0.5
ISIF = 2
ISYM = 2
ISMEAR = -5;SIGMA = 0.05
LDAU =.TRUE.
LDATYPE = 2
LDAUL = 2 -1
LDAUU = 6.8 0.0
LDAUJ = 0.9 0.0
LDAPRINT = 2
LMAXMIX = 4
ENCUT = 500.00 eV

POSCAR (only Fe):
0.0000000000000 0.0000000000000 0.0000000000000 T T T
0.5000000000000 0.5000000000000 0.0000000000000 T T T
0.6250000000000 0.1250000000000 0.1250000000000 T T T
0.8750000000000 0.3750000000000 0.1250000000000 T T T
0.1250000000000 0.6250000000000 0.1250000000000 T T T
0.3750000000000 0.8750000000000 0.1250000000000 T T T
0.2500000000000 0.2500000000000 0.2500000000000 T T T
0.7500000000000 0.7500000000000 0.2500000000000 T T T
0.8750000000000 0.1250000000000 0.3750000000000 T T T
0.6250000000000 0.3750000000000 0.3750000000000 T T T
0.3750000000000 0.6250000000000 0.3750000000000 T T T
0.1250000000000 0.8750000000000 0.3750000000000 T T T
0.5000000000000 0.0000000000000 0.5000000000000 T T T
0.0000000000000 0.5000000000000 0.5000000000000 T T T
0.1250000000000 0.1250000000000 0.6250000000000 T T T
0.3750000000000 0.3750000000000 0.6250000000000 T T T
0.6250000000000 0.6250000000000 0.6250000000000 T T T
0.8750000000000 0.8750000000000 0.6250000000000 T T T
0.7500000000000 0.2500000000000 0.7500000000000 T T T
0.2500000000000 0.7500000000000 0.7500000000000 T T T
0.3750000000000 0.1250000000000 0.8750000000000 T T T
0.1250000000000 0.3750000000000 0.8750000000000 T T T
0.8750000000000 0.6250000000000 0.8750000000000 T T T
0.6250000000000 0.8750000000000 0.8750000000000 T T T

First four Fe are Fe3+ (5 unpaired electrons), next two are Fe2+ (4 unpaired), and further keep a sequence.

I need to calculate a charge ordering for a insulating Fe3O4 using GGA+U method. Setting same U parameters on all Fe site (even though there are two different Fe sites, namely Fe+2 and Fe+3 in the unit cell, only one Fe type is considered in the calculations) does not give correct electronic structure, i.e., a band gap (The U is varied as well). In some literatures, two different U (J) parameters are used for two different Fe sites as mentioned before, resulting insulator. My question is how we devide Fe's sites into two charge ordered sites of +2 and +3 in input files?

You can have 2 copies of the Fe pseudopotentials in POTCAR. This way, VASP will treat the 2 Fe pseudopotentials as 2 different species, and then you can specify the U parameters for them separately.

Even though the calculations are not finilazed yet according to your suggestion, it would give a reasonable results in initial test. Thank you so much.

@tlchan - shouldn't he also edit his MAGMOM line to give a better initial guess closer to Fe3+ and Fe2+ , I would try to alter the MAGMOM line first:

MAGMOM = (#Fe3+)*5 (#Fe3+)*-5 (#Fe2+)*4 (#Fe2+)*-4

5 for unpaired electrons on Fe3+
4 for unpaired electrons on Fe2+

And if that didn't work I would definitely add a new Fe to my POTCAR

Thanks to tlchan and jlbettis.

The idea from tlchan is a considerable. However just setting different initial moment for different Fe sites of Fe2+ and Fe3+ is enough to describe charge ordering through the test calculations. I obtained my desired results. Thanks again, jlbettis.