You need to be very careful! You have a Dirac point with touching/crossing bands: If this specific k-point is NOT contained in your set of k-points you will for example never see a zero gap in the DOS, always a finite gap! (first "inconsistency"). So, only certain special choices of your k-sampling are allowed to avoid this problem (for a primitive cell of graphene this must be a multiple of 6x6 in the plane to hit the K-point and you need to use so-called Gamma-shifted meshes -- for your supercell you need to think about where the K-point of the primitive cell is folded into the supercell Brillouin zone but I think a 17x17x1 mesh does not contain the [folded] K-point; maybe you should even change the supercell?). And in general, be aware of the fact that band minima/maxima are not always located on (those)  high-symmetry lines (chosen to plot the band structure) and that you might observe a DOS "outside the energy range displayed in the band structure" stemming from such k-points elsewhere in the BZ. And third I have also little doubts about the use of  the (3D) tetrahedron integration scheme for a 2D k-mesh (maybe it still gives the right answer, maybe one needed to modify the scheme for 2D systems?). And to Keivan: ISMEAR=-5 vs. ISMEAR=-4 makes basically a difference in total  energies/band occupations, where ISMEAR=-5 is designed to improve k-convergence for metals (with the negative impact that forces/stresses become inconsistent with the improved total energy!), for systems with a gap one should use ISMEAR=-4 (maybe also in this case here with a zero gap? -- I'm not absolutely sure whether ISMEAR=-5 might also influence a bit the DOS near band overlaps or in general). And also ISMEAR>0 should only  be used for metals(!), not even for graphene. Conclusion: Graphene is anything else but a "simple" system and you can make a lot of mistakes in your calculations if you do not carefully think about k-meshes, BZ folding, inclusion of the Dirac point  etc. (and quite often convergence with the k-mesh size is horribly bad ...).

Please be more specific in your question. Post graphics if both the band structure and the DOS and explain in detail where you see the inconsistency.

I have got 95% results with consistency between DOS and Band structures. Please take a look on the attached figure. I used ISMEAR = -5 and Sigma 0.05 for DOS and same Sigma value for band structure with default value of ISMEAR. Moreover, I have extended band structure on more high symmetrical points, but the result is same as shown in the attached figure. I have also varied k-points in DOS calculations, but I got the same result.

can you give some more information like size of the supercell and how much kpoints you used for DOS calculations? 

Which of your results do you trust more, band structure or DOS? Do you have independent information for comparison (measurements, other calculations)?

The size of the doped graphene supercell is 5x5 and I have used 17x17x1 K-points for DOS calculations with ENCUT = 520 eV.

In my opinion, both should be consistent. However, I don't have any other calculation option to confirm either DOS is correct or band structure.

Is the point you label K where the bands cross near the Fermi level included in the kpoints used to calculate the DOS? Also try Gaussian smearing, or ISMEAR=1. ISMEAR=-5 should be used when you have a gap. In this case you do not.

Dear Prof. Keivan Esfarjani,

I have followed your advice and calculated DOS with Gaussian smearing and compared with tetrahedron method with Blöchl corrections. But in case of Gaussian smearing the DOS shows many discontinuities. 

According to vasp manual:

For the DOS and very accurate total energy calculations (no relaxation in metals) use the tetrahedron method (ISMEAR=-5).

Is it possible that vasp may have some internal error and needs a new compilation?

Thank you. 

In my questions about "independent information" (measurements, other calculations) I also had in mind to have a look at works of other people. So, what is generally known about the band structure and the DOS of graphene already?

Zero band gap semiconductor.

You need to be very careful! You have a Dirac point with touching/crossing bands: If this specific k-point is NOT contained in your set of k-points you will for example never see a zero gap in the DOS, always a finite gap! (first "inconsistency"). So, only certain special choices of your k-sampling are allowed to avoid this problem (for a primitive cell of graphene this must be a multiple of 6x6 in the plane to hit the K-point and you need to use so-called Gamma-shifted meshes -- for your supercell you need to think about where the K-point of the primitive cell is folded into the supercell Brillouin zone but I think a 17x17x1 mesh does not contain the [folded] K-point; maybe you should even change the supercell?). And in general, be aware of the fact that band minima/maxima are not always located on (those)  high-symmetry lines (chosen to plot the band structure) and that you might observe a DOS "outside the energy range displayed in the band structure" stemming from such k-points elsewhere in the BZ. And third I have also little doubts about the use of  the (3D) tetrahedron integration scheme for a 2D k-mesh (maybe it still gives the right answer, maybe one needed to modify the scheme for 2D systems?). And to Keivan: ISMEAR=-5 vs. ISMEAR=-4 makes basically a difference in total  energies/band occupations, where ISMEAR=-5 is designed to improve k-convergence for metals (with the negative impact that forces/stresses become inconsistent with the improved total energy!), for systems with a gap one should use ISMEAR=-4 (maybe also in this case here with a zero gap? -- I'm not absolutely sure whether ISMEAR=-5 might also influence a bit the DOS near band overlaps or in general). And also ISMEAR>0 should only  be used for metals(!), not even for graphene. Conclusion: Graphene is anything else but a "simple" system and you can make a lot of mistakes in your calculations if you do not carefully think about k-meshes, BZ folding, inclusion of the Dirac point  etc. (and quite often convergence with the k-mesh size is horribly bad ...).

Yes, only certain special choices of k-sampling are allowed to avoid such inconsistency between DOS and band structures.

I have one more question about Fermi level: I have calculated DOS with different k-sampling, and found that there is small change in Fermi energy with respect to different k-sampling, however, the Fermi energy is same in case of band structures calculations. It doesn't show any dependency of Fermi energy on k-sampling selected for DOS calculations. Here, I want to confirm that which value of Fermi energy Ef do we have to consider for plotting band structure with E-Ef ? Means the Fermi energy from DOS calculation or from band structure calculation?  

I have observed one more thing that the Fermi energy for DOS and band structure will be almost same in case of consistency between DOS and band structure. 

due to reference point which is zero so DOS becomes unsymmetric as i know in gnr system.

Dear All,

I am calculating PDOS of LaF3 doped with Eu and Tb (DFT+U calculation) but when I am plotting pdos, I am not getting conduction band density of states properly. I have attached here my pdos graph. My Fermi level is at zero eV and in INCAR, I gave energy range EMIN and EMAX -20 and 20 eV respectively. Still I am not getting dos upto 20 eV energy range

Where is the problem ?

Thank You All

Shilendra

Did you add enough conduction bands (sufficiently large NBANDS in INCAR) to have included all states up to +20 eV? (What does EIGENVAL tell about the energy of the highest band [after subtracting the energy of the Fermi level]?).

Juergen