I did hybrid functional calculation on a AlN surface with VASP and met some issues.
Case 1. the electronic relaxation was performed with ALGO = All (Damped), ISMEAR = 0 and SIGMA = 0.1. In the results, The entropy in the OUTCAR is ZERO.
Case 2. the electronic relaxation was performed with ALGO = Normal and ISMEAR = -5. In the results, of course, the entropy in the OUTCAR is also ZERO.
Here comes the question. In the results of density of states (DOS), the Fermi level of case 1 is a little above VBM, However, the Fermi level of case 2 is just at the VBM.
Actually, I prefer case 2 is correct due to the manual of Vasp. but I was still confused by the zero entropy issue.
As entropy of both cases is ZERO which means there is no widening near the Fermi level, Why there is a difference between two cases?
Thanks a lot !
Well... What do you mean, saying "the Fermi level of case 1 is a little above VBM" As far, as I understand, there are two possible cases (see attached figure).
Generally speaking, there is a standard problem in DFT calculations - while we calculate band structure of semiconductor, the Fermi level appears to be slightly (~ 0.01-0.02 eV typically) below Valence band maximum making the valence band being slightly unfilled - even if we ignore the thermodynamical aspects at all. Actulally this is purely computational drawback with no physical meaning - the result often can be improved increasing the k-mesh and so on, however there is no serious reasons to take care on it, because there are no ideal insulators in nature.
Dear Dr.Bannikov,
Thank you for your kind and helpful replay! The k-mesh is set as 5x5x1 for the 2x2 slab model and I think it is fine. I just attached a figure of DOS to the question to make it clear. The difference is so obvious that can be as large as 0.25 eV!
So, do you have any further comment on this question?
Thank you very much!
Okay.... Have you tried to look the file with numerical data of DOS? Or (in principle, the same) to plot the DOS in the near-Fermi region (~ +/- 0.5 eV as related to your plot) - taking the range for DOS axis ~ 0-1 your units? Perhaps, the states of low density could appear near EF - which are not distinguishible in your scale and "smeared", so to say. The amount of electronic states they cover is about to be negligible, however, the Fermi level may shift considerably due to their presence. I would advise to check this point, at first.
And... 5x5x1 k - mesh? Frankly speaking, I would not trust to "fine details" of DOS obtained within that.
Dear Dr.Bannikov,
You are RIGHT!! I checked the original file of case 1 and find there are two values much tiny between VBM and Fermi level which are 0.0133 at -0.204104 eV and
0.0001 at -0.103604 eV.
So, as you mentioned, the coarse k-mesh might be the problem...
I have referred published results to determined this k-mesh and actually I was wondering this case always....
the length of the 2x2 surface cell is 6.26 angstrom. what is your suggestion for the k-mesh? or simply increase NEDOS in the INCAR ?
Thank you very much !!
Hi,
you should always converge the k mesh for every new system, meaning you should increase the number of k points until the resulting energies are converging, the changes in energy that still occur should be in accordance with the desired accuracy
by the way, also other settings such as encut and sigma should always be converged
hope this is some help, cheers
Dear Dr.Feldbauer,
Thank you for your suggestion! Actually, I have changed k-mesh from 5x5x1 to 9x9x1 and did the calculation. However, the results near the Fermi level make no difference which means the "smearing" is still there. The encut was chosen as 500 eV which should be fairly enough for the system. and as the question mentioned, sigma =0.1 gives zero entropy in the OUTCAR.
So, I am still confused with the question. Do you have more suggestions ?
Thank you very much!
Entropy: A concept that is not a physical quantity
https://www.researchgate.net/publication/230554936_Entropy_A_concept_that_is_not_a_physical_quantity
Jiaduo Zhu ...... I was wondering about your solution....., any valuable comments would be greatly appreciated. Thank you. having the same problem.
Yisehak Gebredingle Please check the occupation in the EIGENVAL, it can tell the VBM and CBM.
Hi~ Dr. Zhu. If I wanna align the Evac to zero, then how should i do with VASP?
Is it correct that first calculates the WF, then aligns the Fermi-level， Ef= Evac-WF?
Jiaduo Zhu
- Badges
- Science topic