bands 4.6 NPAR=1 VS. 4.5.8 NPAR > 1

[giacomo giorgi]

Dear all,
there is something I am not able to understand.
I optimize my bulk system (GaAs, 8 atoms, ZB structure) with VASP 4.5.8 (PAW/GGA/91). I get my optimized structure.

I do band calculation on the optimized structure with the ICHARG=11 and I get my plot of the bands with 48 bands (I think it is the default since I do not set NBANDS explicitly, here).

From the previous optimization I take my CONTCAR. I copy it in a new POSCAR. I reoptimize with VASP 4.6 (of course same settings and same KPOINTS and POTCAR). I get the Same structure & the Same energy.

I do band calculations with the identical input of the previous one (now with NPAR=1).
What I got is the same band structure, the same Fermi energy, BUT 47 BANDS instead of 48. Why does it happen? Different cut-off settings?
Moreover, If I increase manually NBANDS, the added bands have an "orrible" zigzag aspect

This is my INCAR

SYSTEM=GaAs cell
ISPIN=2
ISTART=1
LREAL=TRUE
PREC=Med
ICHARG=11
EDIFFG=-0.05
EDIFF=1E-5
GGA=91
ENCUT=287.594
ENAUG=531.356
INIWAV = 1 electr: 0-lowe 1-rand 2-diag
NELMDL = -7
LMAXMIX = 6
IALGO = 48
NSIM = 1
LCHARG = .TRUE.
LWAVE = .TRUE.
LELF = .TRUE.
IBRION = -1
ISIF = 3
ISYM = 1
POTIM = 0.1
LORBIT = 12
ISMEAR = 1
SIGMA = 0.02 broadening in eV
NPAR = 1
LPLANE= .TRUE.

Thanks,
Giacomo

[admin]

apart from the calculated number of bands, NBANDS is an integer multiple of the number of processors unless NPAR is set explicitely.
This may cause the discrepancy in the number of bands chosen automatically by vasp.
concerning your second question: please note that there is no information about the irreducible representation of each state given by vasp, which would allow for a proper connection of the bands according to group symmetry criteria. The bands are connected according to their energies. Especially for the high conduction bands, this may lead to 'zigzag' instead of crossing bands.