bug in integrated DOS for fixed moment

Message

brsahu · #1 Post by **brsahu** » Fri Jan 20, 2006 7:11 pm

Dear VASP users,

I did a fixed moment calculation ( NUPDOWN= 6 Bohr magneton/supercell; total energy self-consistent, but magnetic moment per cell is not changed from iteration to iteration) for a system having NELECT=1010 ( a supercell containing 250 atoms).

After self-consistent run, I used ICHARG=1 and copied CONTCAR to
POSCAR, NSW= 0. LORBIT=10 to get the DOS.

In the DOSCAR file the first block has 5 coloumns: energy, upDOS, dnDOS, integratedUP and integratedDN

But the integarated UP and DOWN DOS numbers upto Fermi energy are wrong. They donot add up to total number of electrons in the
supercell ie NELECT =1010 and their difference is not NUPDOWN =6.

The NON-FIXED MOMENTcalculation DOS for the SAME system do not have such behavior. It correctly adds upto NELECT=1010 and their difference is what is obtained per cell after fully self-consistent calculation.

I show here ONLY a piece of data from the DOSCAR file of FIXED MOMENT calculation near FERMI ENERGY( which is 6.0796 eV):

5.750 0.0000E+00 0.0000E+00 0.1014E+04 0.1000E+04
5.775 0.0000E+00 0.0000E+00 0.1014E+04 0.1000E+04
5.800 0.0000E+00 0.0000E+00 0.1014E+04 0.1000E+04
5.825 0.0000E+00 0.0000E+00 0.1014E+04 0.1000E+04
5.850 0.0000E+00 0.0000E+00 0.1014E+04 0.1000E+04
5.875 0.0000E+00 0.0000E+00 0.1014E+04 0.1000E+04
5.900 0.2068E+02 0.0000E+00 0.1014E+04 0.1000E+04
5.925 0.3349E+02 0.0000E+00 0.1015E+04 0.1000E+04
5.950 0.9714E+01 0.0000E+00 0.1015E+04 0.1000E+04
5.975 0.4546E+01 0.1105E+02 0.1016E+04 0.1000E+04
6.000 0.3595E+01 0.5060E+02 0.1016E+04 0.1001E+04
6.025 0.2329E+02 0.7827E+01 0.1016E+04 0.1002E+04
6.050 0.2321E+02 0.1955E+02 0.1016E+04 0.1002E+04
6.075 0.1848E+02 0.6314E+02 0.1017E+04 0.1004E+04
6.100 0.4671E+02 0.3709E+02 0.1018E+04 0.1005E+04

As one can see (from 3rd and 4th coloumn), the sum of integrated up and down is not 1010 at Fermi energy (6.075 eV) and their difference is not 6. Since I am doing a fixed moment
calculation with NUPDOWN =6 and NELECT in OUTCAR file is 1010, isn't the integrated UP and DOWN add to NELECT= 1010
upto FERMI energy and their difference upto Fermi energy should be 6?

Pl. note I don't get this behavior for NON-FIXED MOMENT calculations with the SAME SYSTEM (NELECT =1010).

The INCAR file used to calculate DOSCAR:
(CHGCAR and CONTCAR from the previous run was used here
with NSW =0 and LORBIT=10)

SYSTEM = fixedmoment

Start parameter for this Run:
ICHARG = 1
ISPIND = 2
ISPIN = 2
NUPDOWN = 6
PREC = high

Electronic Relaxation
ENCUT = 340.0000
NELMIN = 5
NELM = 90
EDIFF = 1E-05 stopping-criterion for ELM

ISYM = 2

DOS related values:
ISMEAR = -5
NEDOS = 1201
EMIN = -15.0
EMAX = 15.0
LORBIT = 10

IALGO = 48
LREAL = O

Is my understanding wrong, OR I am doing something wrong
OR is there a really bug in the code for integrated up and down
DOS fixed moment calcuation.

Which piece of the code calculate the integrated DOSUP and integreated DOSDN numbers?

Sahu

brsahu · #2 Post by **brsahu** » Fri Jan 20, 2006 7:48 pm

Dear VASP users,

the version of VASP used was VASP.4.6.28 release 25Jul05

Sahu

brsahu · #3 Post by **brsahu** » Mon Jan 23, 2006 4:18 pm

Dear VASP Admin and Vasp users,

Do you have any suggestion to my query on integrated DOS obtained from FIXED MOMENT calculation.

Sahu

#4 Post by **admin** » Wed Jan 25, 2006 2:19 pm

if you set NUPDOWN, the Fermi energy which is written in OUTCAR is the Fermi energy which is calculated from the spin down run only, it is NOT interpolated anymore to account for the spin-up states:
you should have 508 up and 502 down electrons, and, as you see, E_F is adjusted to 2*502 (1004) electrons.
--> if you set NUPDOWN, the written Fermi energy is not correct, it has to be interpolated to the energy where the sum of up and down electrons really gives the total number of electrons.