why does the density of states look so spiky?

[riche]

Dear all,
I am doing the density of states of Al systems. The density of states looks very spiky.
Here is the INCAR file that I have:

!---general---
SYSTEM = Al-X
!
!---Startparameter_for_this_run---
PREC = High !Normal,Medium,High,Low
ISTART = 0 !WAVECAR:0-new,1-cont,2-samecut
ICHARG = 11 !charge:1-file,2-overlapping_atom,10-const
LORBIT = 11

!
!---spin_calculation---
ISPIN = 1
!
!---Electronic_Relaxation1---
ENCUT = 420
LREAL = .FALSE. !F:small_syst,T:large
NBANDS = 40
NELM = 130
NELMIN = 4 !min_nb_ELM
NELMDL = -2 !nb_ELM_before_charge_update
EDIFF = 1e-6 !stopping-criterion_for_ELM [eV]
NGX = 40
NGY = 40
NGZ = 40
EMIN = -2.3
EMAX = 44.2
NEDOS = 16000

!
!---DOS_related_values---
ISMEAR = -5
! ISMEAR:tetrahedron,-1:fermi,0:Gaussian,N:Methfessel-Paxton(Metal)
SIGMA = 0.4 !smearing_width_ISMEAR
!
!---Electronic_relaxation2----
ALGO = FAST
!---Parallel_processing
LPLANE = .TRUE.
NPAR = 4
!---Write_flags---
LWAVE = F !write_WAVECAR
LCHARG = F !write_CHGCAR
LVTOT = F !write_LOCPOT,local_potential
LELF = F !write_electronic_localiz_function(ELF)

And this is the KPOINTS file:
K-Points
0
Gamma
40 40 40
0 0 0

Thank you very much in advance!
Riche

[riche]

The answers to the following questions may clear my doubts also:
(1) Without relaxation, ISMEAR=1 makes DOS smoother than ISMEAR=-5. But can I trust the results with ISMEAR=1.
(2) what kind of KPOINTS files should I use? Explicit input or automatically generated. What is the consequence if I use automatically generated k-points?
(3) Is it absolutely necessary to set ISYM to be 0?

[panda]

First of all, you must ALWAYS relax the system first BEFORE doing a DOS calculation. Once you have obtained the ground state structure, you go back and run a DOS calculation. ISMEAR = -5 may make the DOS look more spikey, but should smooth out if you increase NEDOS. It doesn't matter which k-point generation scheme you use as long as you have enough k-points. If you use auto you can see in the OUTCAR file what your grid/mesh is. ISYM = = 0 switches off symmetry. It depends on the symmetry of your system as to whether this will matter or not, but it never hurts just to switch it off.

[riche]

Dear panda,
Thank you very much for your reply.
As you may see, I set NEDOS to be 16000 in INCAR. Previously I used NEDOS=8000. Even NEDOS=16000 does not help a bit. The energy step is like 3 meV when NEDOS=16000. But anyway, I will try it again with the smaller or much smaller energy step. And see what comes out.