Here you have a step by step tutorial, to perform bandstructure calculations with hybrid functionals, with fake k-point method and wannier functions (http://www.wannier.org/)

http://cms.mpi.univie.ac.at/wiki/index.php/Si_bandstructure

Follow it step by step with Si first and then apply it in your system, with the specific tags that may be needed.

Hope that helps.

Thank you Oier Arcelus I will check it and let u know if there is some problem.

In general the calculation of DOS from a hybrid calculation should be exactly the same as to a non-hybrid DFT calculation with VASP. make sure to input the proper LORBIT value in the INCAR to get the DOSCAR file you want (i usually use LORBIT=10 to get the PDOS on individual atoms with s,p,d projections).

The actual calculation requires leaving some INCAR keywords to VASP defaults. I'm adding here a model INCAR file for HSE06 calculation which should get you started and you could build on:

EDIFF = 1E-4

EMIN = -10 #DOS Writting options

EMAX = 20

NEDOS = 5000

LWAVE = .TURE.

LCHARG = .TRUE.

ISPIN = 2 #spin polarized

ISYM = 0 #break symmetry

NSW = 0

LORBIT = 10 

LREAL = .FALSE.

NSIM = 2

NPAR = 112 #should be the total # of cores in HF calculation

KPAR = 7

ENCUT = 750 eV

LHFCALC = .TRUE. #set hybrid functional calculation on

AEXX = 0.25 # fraction of exact exchange

AGGAX = 0.75 # fraction of GGA exchange

AGGAC = 1.0

#NKRED = 2 #2 - reduction of q-point mesh

HFSCREEN = 0.2 # Screening in HSE06

PRECFOCK = Fast # the grid in HF calculation

In addition, if you need help plotting the actual DOS in the DOSCAR file in the SI of this paper there is an attached mathematica script to due this automatically from VASP output files and also calculate the DOS effective mass:

http://scitation.aip.org/content/aip/journal/jcp/144/16/10.1063/1.4946752

Thank you very much Dr Ofer Neufeld for your answer.

Dear Ofer Neufeld first thank you for help, I want to ask one thing here in DOS or PDOS calculation if we give default NEDOS value (i.e 300) then it shows good looking DOS but it shows bangap less than actual bandgap which we see by any script. I mean if script is showing 2.3 eV bandgap then by plotting DOS it will dhow nearby 2.1 eV. it will not show exact bandgap 2.3 eV.

When we calculate DOS with increasing NEDOS value for example NEDOS=3000 then it shows exact bandgap by script as well as from DOS plot but graph does not look good. DOS peaks becomes so intense and dense because more NEDOS.

I have seen in literature people publish so good looking graphs of DOS and PDOS. I do not know how they do ?

Shilendra

Dear Shilendra,

The keyword NEDOS just determines the amount of point in the DOS calculation. Naturally, if your resolution in the energy is pretty low you will get a bad BG in the DOS plot (if you have ~ 30eV range in the graph and you take 300 points that's a resolution of dE~0.1eV which you can't be more precise from, hence there is no point talking about 0.1eV differences in the BG). I would definitely use NEDOS and EMIN EMAX to get at least 0.01eV resolution on the plot, especially wish you wish to analyze hybridization of states.

Another point that is important is not to rely on VASP to calculate the fermi level on its own (the value in the top of the DOSCAR file) since this depends on the way VASP derived the DOS. You should however calculate the fermi level on your own from the OUTCAR file (with a script or any other way), which is more accurate, and there can be differences between the two. Also, the value of the BG can change according to the k-grid sampling if the CBM and VBM are not on the gamma point or align with the k-space grid, and thus the BG value should be converged separately from total energy with respect to k-grid density. In fact the BG you are calculating is always a type of upper limit since you can only add new k-points to the k-grid which might reduce the BG value by raising the VBM or lowering the CBM (suffice you believe the level of theory and KS eigenvalues).

In general there is no reason to expect a very smooth graph for the DOS which is defined with delta function like sampling in the energy domain. The plot will definitely be smoother by taking a smaller resolution, not to mean that its more accurate. In fact its less accurate, and the ripples, or oscillations in the DOS plot come from the discretization in energies due to quantization (i.e., a specific energy in which there is a large number of states). These states are expanded by the k-space integration scheme you are using. If you use different levels of theory, like types of tight binding you can get smoother looking DOS, and the size of the unit cell can have some influence too since this determines the size of the BZ by inverse relations, which is what you may have seen in publications (though its always a possibility it is just an under-sampled graph).

hope this helps,

ofer.

Yes I have also face this before but, using different smearing I fixed it. The parameter is SIGMA which also effect the smoothness. 

Dear Ofer Neufeld thank you for detail explanation.

For knowing fermi level I use command grep fermi OUTCAR . Do you have any script for knowing VBM and CBM values.

Thank You

Shilendra

In this paper's SI:

http://scitation.aip.org/content/aip/journal/jcp/144/16/10.1063/1.4946752

ftp://ftp.aip.org/epaps/journ_chem_phys/E-JCPSA6-144-045615

The script called "DOS eff. mass" calculates from the outcar the band gap and fermi level by subtracting the CBM and VBM (it also plots the DOS and calculates the DOS effective mass, but the output variables 'Efermi' and 'BG' contain what you want). It runs on Mathematica, and pretty self explanatory with comments, but contact me if you need help.

Sometimes, we don't have consistent picture of band structure and DOS. For example, I have calculated band structure and DOS of doped graphene but they are not consistent. I have used ISMEAR -5 with dense k-grid in vasp code. Even I have found that DOS shows band gap opening, however, there is no band gap in band structure.  Any suggestion, please.

Muhammad Ali Askari you can reduced the energy level of you axis then it'll be clear to you. You can also use Gaussian smearing with different SIGMA so, you will get clear image of band gap. 

In my experience the two should form a consistent picture, after all, the DOS is calculated by integration over the BZ. If they are not consistent it might be an indication something is wrong. If you mean the band gap calculated from the OUTCAR by subtracting the energy of the CBM and VBM and the band gap of the DOS graph, they should be identical, up to the resolution of your DOS graph, and your convergence level of the calculation in both.

Try changing NEDOS, ISMEAR, SIGMA, k-grid, which directly effect your DOS (gaussian smearing should also be good for graphene since its basically metallic). Also you can try changing numerical parameters for you calculation to try and find the culprit. This might not be an easy task, and you're better off maybe reconstructing results from a known paper where you know exactly what parameters they used. 

Which vasp parameters are most critical with respect to the symmetry of the structure? I think it might be due to the change of the symmetry of the system. 

You can try to remove symmetry with ISYM, and also increase your structure relaxation parameters. Also, since Graphene is 2d you can try increasing your z axis vacuum seoqration. But initially you should make sure you are properly calculating the band gap and DOS by reconstructing a simpler case. That is, a non-doped lattice.

Other than that any numerical parameter can be important for convergence and it's hard to say which, but even if you are not converged the two should be consistent which is the indication something maybe is off only with the dos. Not sure I can help more than that.

Good luck,

Ofer.

I also want to know how to do DOS and Band structure calculations using HSE..Zahir Muhammad, do you come to know the parameters to be set in INCAR?

Dear Neha Kapila,

You can use the following parameters in your SCF calculations

ISMEAR = 0 SIGMA = 0.01 LHFCALC = .TRUE. HFSCREEN = 0.2

TIME = 0.4

After that use the CHARGCAR file for further calculations, but better to use smaller grid size for further calculations.

Thank you so much, do we need to change the KPOINTS file too for bandstructure? copy IBZKPT to KPOINTS(as its shown on vaspwiki for Si) and DOS or shall we proceed with normal procedure. Also I want to know that I use vaspkit to generate band_gap file which need kpoints in line mode. How will it work then.

Please check the link below.

https://www.vasp.at/wiki/index.php/Bandstructure_of_SrVO3_in_GW

Hi Zahir Muhammad, won't using the scf CHGCAR file for DOS or Band calculation give incorrect result? Since the non-local exchange is not determined by the charge density but by the density matrix and/or the KS-orbitals.