DOS does not go to zero in the gap

Message

dkanan · #1 Post by **dkanan** » Wed Jan 13, 2010 9:12 pm

Hello,
I am running HSE calculations on a small rhombohedral cell of a wide band gap insulator (Transition metal oxide to be exact) using vasp 5.2

Using the same INCAR,POSCAR,POTCAR, and converged KPOINT mesh/file I have found that the DOS does not go to zero in the case where the tetrahedron method is used for the BZ integration, whereas it does go to zero in the gap for the case where gaussian smearing is used!

Some suggestions in other threads say to make sure the gamma point is included, which I have done. I have even looked at various k-meshes including:
gamma centered 12x12x12
MP 12x12x12

in combination with both ISMEAR=0 and ISMEAR=-5, both with a very small SIGMA=0.01

Can you please help/explain why this is the case? I have seen this problem in other calculations as well.

dkanan · #2 Post by **dkanan** » Thu Jan 14, 2010 3:27 pm

By the way, I see that in other threads the suggestion is also to make sure the k-points on the BZ boundary are included. I am not sure how this should be checked. I assume that the automatically generated k-mesh in vasp already takes care of this. Is this correct?

VASP001 · #3 Post by **VASP001** » Fri Jan 15, 2010 3:11 am

Maybe ,you can see the HOMO and LOMO in the OUTCAR ,because the DOS is affected by the pepole ,because ,it's can be tunned by hand to choose some parameters ,for example sigma etc .
[quote author=0 and ISMEAR=-5, both with a very small SIGMA=0.01

Can you please help/explain why this is the case? I have seen this problem in other calculations as well.[/quote]

dkanan · #4 Post by **dkanan** » Fri Jan 15, 2010 5:39 am

VASP001, thank you for your help. however, I could not understand what you meant in your reply. Do you mean the DOS is affected by different people? as in the user. What about the HOMO and LUMO should I check?

Do you think this is a problem with VASPs implementation of the tetrahedron method or with the use of the tetrahedron method for this case??

VASP001 · #5 Post by **VASP001** » Fri Jan 15, 2010 7:40 am

[quote="dkanan"]VASP001, thank you for your help. however, I could not understand what you meant in your reply. Do you mean the DOS is affected by different people? as in the user. What about the HOMO and LUMO should I check?

Do you think this is a problem with VASPs implementation of the tetrahedron method or with the use of the tetrahedron method for this case??[/quote]
I mean some facors about the parameters when you draw the pictures of the DOS ,and it's affected by the people .Besides that ,the energy zero is be substracted by the fermi level or not ?

I'm sorry .I make a mistake ,when you see the occupancy in the OUTCAR ,the gamma point only is used and the Direct energy gap at the gamma point ,and you can see the energy gap from the OUTCAR .

dkanan · #6 Post by **dkanan** » Fri Jan 15, 2010 4:40 pm

Sorry, but I am not sure that you have understood my question. So to clarify it I am not concerned about the fermi level here, I do subtract all the energies by the fermi level in order to shift my DOS plot as you refer to.
However my problem is more serious than that because I am modeling a wide gap insulator, but the calculation results in a DOS that is not zero IN the gap. This is only true when the tetrahedron method with blochl corrections is used for BZ integration, for gaussian smearing the DOS is fine.

VASP001 · #7 Post by **VASP001** » Sat Jan 16, 2010 3:33 am

[quote="dkanan"]Sorry, but I am not sure that you have understood my question. So to clarify it I am not concerned about the fermi level here, I do subtract all the energies by the fermi level in order to shift my DOS plot as you refer to.
However my problem is more serious than that because I am modeling a wide gap insulator, but the calculation results in a DOS that is not zero IN the gap. This is only true when the tetrahedron method with blochl corrections is used for BZ integration, for gaussian smearing the DOS is fine.[/quote]
If it's like this ,I don't know the reason.

Best wishes.

dkanan · #8 Post by **dkanan** » Sat Jan 16, 2010 9:01 am

thank you for trying, I hope admin can address my issue

schani · #9 Post by **schani** » Mon Feb 01, 2010 3:04 pm

Have you tried to look at the eigenvalues themselves?

#10 Post by **admin** » Wed Feb 10, 2010 8:26 am

1) HOMO and LUMO are not valid input parameters in vasp.
2) as schani proposed: please check the eigenvalue spectrum at each k-point. for a wide gap insulator, the number of occupied bands has to be exactly the same at each k-point, all below the gap. Otherwise the tetrahedra will be interpolated between HOMO and LUMO states.