HSE magnetic calculations problem

[husl]

Dear all, I am trying to do magnetic calculations for Mg doped VO2 system
with HSE functionals. I tested the way to get magnetic moments of a FM
configuration in the following procedures:
I used vasp 5.2.12 11Nov2011 version.

1, I performed a normal PBE calculations with ISPIN=1, get WAVECAR and CHGCAR

2, Then I set ISPIN=2, and MAGMOM=8*1 16*0, use the WAVECAR and CHGCAR
converged in step 1, and get total magnetic moment 0.9884, and also WAVECAR
and CHGCAR with magnetic information in it.

3, Then a ISPIN=1 HSE calculation was performed, started from the WAVECAR
and CHGCAR from step 1, to get the HSE version of WAVECAR and CHGCAR.

To get magnetic moments and energies for the FM configuration in HSE
calculations, some different procedures were tested.

Use ISPIN=2, MAGMOM=8*1,16*0, and HSE tags in INCAR. The "magnetization" is
checked for each electronic iterations in OUTCAR to see how the magetic
moments converge.

4, Use ISTART=1 to read only the WAVECAR from step 3, the charge density
will be initialized from WAVECAR, the manually set
MAGMOM in INCAR will be ignored, and the final total magnetic moment is 0.

5, Use ISTART=1 and ICHARG = 1, to read the WAVECAR and CHGCAR from step 3,
the manually set MAGMOM in INCAR is read in, but the charge density is still
initialezed from WAVECAR, and the final total magnetic moment is 0.

This behavior is different with PBE calculations (step 2), which set and kept
the initial magnetic moment=8 in the first electronic iteration step.
This is also contray to the manual of MAGMOM tag.

6, Use ICHARG=1 to read only the CHGCAR only from step 3, the manually set
MAGMOM in INCAR will be read in and kept until the 5th electronic iteration,
the final total magnetic moment is 3.

7, Use ISTART=1 and ICHARG=1, but read the WAVECAR and CHGCAR from ISPIN=2
PBE calculations from step 2, it also ignore the manually set MAGMOM, but
read in the preconverged magnetic moment 0.9884, and finally converged to
total magnetic moment to 1.

Here I got different magnetization from different ways in step 5, 6 and 7,
so I am puzzling waht is the right way to get the energy and magmom
for a special magnetic configuration?

The INCAR, POSCAR and KPOINTS files are also attached in the following.
Or what is wrong in these input files?

INCAR
SYSTEM =Mg V O !Mg VO2 2x1x1
#
ISTART = 1
ICHARG = 1
PREC = normal
ENCUT = 500
LREAL = Auto

NELM = 300
NELMIN = 4
EDIFF = 1E-5
EDIFFG = -0.01

ADDGRID=.TRUE.
NSW = 0
IBRION = -1
ISIF = 3

ISMEAR = 0
SIGMA = 0.05
ISPIN = 2
MAGMOM = 1 1 1 1 1 1 1 1 16*0
LORBIT = 11

LHFCALC = .TRUE.
HFSCREEN = 0.2
ALGO = Damped
TIME = 0.4
PRECFOCK = N

POSCAR
Mg V O
1.00000000000000
11.3843775411964607 -0.0140466356768370 -0.0761365201429826
-0.0069097870079915 4.6529669251997348 -0.0745136759386223
-2.8180460361316104 -0.0709739833527072 4.6653255575087886
Mg V O
1 7 16
Direct
0.1293595105167713 0.9974076690695391 0.0160040484114736
0.6067388774416086 0.0227099797575110 0.9566024427571977
0.3778273473985372 0.0322801060431831 0.9541733104976250
0.8700353851994246 0.0351915308546100 0.0256138233349078
0.1082606125332289 0.5196152482029240 0.5176887411643809
0.6444440413599467 0.5125092353642210 0.5042195726513017
0.3640742273462306 0.5004604044174065 0.4384384483324167
0.8853229889051479 0.5317798638367766 0.5411287264048376
0.4568267103966452 0.6767113778374636 0.3147614368418953
0.9469588282561175 0.7107732380608719 0.3055274182188427
0.0483250408926970 0.3066170525562452 0.7053051274620411
0.5518417351350884 0.2876915291650771 0.7020487753879963
0.4516115546820111 0.7775611694889581 0.7952753784388309
0.9411683831916572 0.7942588059399311 0.8077485402576979
0.0377780625350808 0.1994786040920403 0.2112721685061016
0.5510234155980256 0.1983509058736639 0.2020924948680442
0.2048448688225560 0.6573924496020176 0.3034711338139818
0.7010552100954507 0.7000563080283645 0.3047121361325543
0.3016576582076959 0.2814829356958961 0.6889380408669760
0.8020739834261391 0.2930413933513419 0.7103030618378838
0.2094723206396841 0.7840480385822954 0.7874790416870198
0.6983821539580446 0.7927708943801003 0.8021532670581202
0.3156131179073866 0.1856540482196914 0.2011513514072086
0.7953039655548099 0.2021572115798625 0.2038915136606641

KPOINTS
Monkhorst Pack
0
G
2 4 4
0 0 0

[admin]

You are forcing Mg to be magnetic. Check reasonable electronic configurations before you start expensive HSE calculations.

[husl]

Thank you for your answer.
The local magnetic moment of Mg is close to 0 after convergence in each case.
The problem is the strange behavior of VASP in HSE calculation compared to PBE.
And could you please tell me what is the right procedure to calculate magnetic moment for a specific magnetic configuration using HSE functionals, if I set MAGMOM of Mg to 0.

[admin]

Your slab is highly nonstoichiometric and therefore far from realistic structure. After the substitution V/Mg the charge of the cell is -3. Try to set up an electronically balanced configuration. E.g. two V/Mg substitutions and single O vacancy (2*-3 and +6). Then you can try to define magnetic moments of V atoms.
[ Edited Thu Jan 05 2012, 04:40PM ]

[husl]

Ah, thank you for your nice suggestions.
But I am sorry I am still not very clear what I need before I define magnetic moments of V atoms in HSE calculations. Should I use ISPIN=1 HSE calculation converged WAVECAR and CHGCAR; or CHGCAR only; or ISPIN=2 PBE calculation with initial magnetic moment converged WAVECAR and CHGCAR, or any other preconverged files?
Thank you! Sorry for bothering again.