My Community

I have run into a strange problem. I am trying to create a PARCHG file for STM viewing and the File is large and mostly empty. The out file ends with:
NBMOD is set to -3
Selected energy range (EINT) -1.9992 -1.8992
Selected all k-points to calculate charge density.
k-point no. 1; bands inside the range 0

But a plot of the DOS clearly shows that there are many states in that energy interval. If I increase the k points to 2 2 1, about 4 bands (out of 1000) are reported.

What am I doing wrong here?

Hi,

According to the info printed at the beginning of your OUTCAR, the energy range that you selected is "EINT = -0.35 -0.25". These values are with respect to the Fermi energy (see NBMOD), which gives [-1.9992 -1.8992] if the Fermi energy (-1.649 eV) is added. Could you please upload the OUTCAR of the SCF calculation on your system to see if there are states in the interval [-1.9992 -1.8992]? This can be deduced by looking at the orbitals energies printed below

 k-point     1 :       0.0000    0.0000    0.0000
  band No.  band energies     occupation 

Note that looking at the DOS can be misleading because of the broadening that is applied to the DOS.

You are exactly right. There are no bands with that energy in that region. How can I fix this? Will increasing the number of bands work? Is the only option using a larger bias range even though I am using voltages outside the experimental region?

Thank you for your help!

KW

Would increasing the number of k points also increase the density of energies calculated near E-Fermi?

In general, using more k points leads to broader energy bands. However, in your case the system is already quite large such that using more k points may not change that much the width of the energy bands.

Probably even more important is the choice of the functional, which may influence quite a lot the position of the energy bands. However, one should also not forget that DFT does not provide orbitals energies that have a direct physical interpretation.

Dear KWH,

I have another small comment. It may indeed be necessary to adjust the bias voltage to different values compared to the experiment. I would try a couple of different ranges, keep track of how many and which bands contribute, and check how your simulated STM picture evolves with the changes. It will help you understand the system better and hopefully give you enough information to reproduce the experiment qualitatively. This is often enough to interpret the experimental STM image correctly.

For quantitative results, you might need to switch the functional or change other settings to reproduce the experimental band structure more accurately.

Cheers, Michael

Thank you Michael. Very helpful. One last question: How is the broadening determined when the DOS and PDOS are plotted?

How the broadening is determined depends a lot on how you plot your DOS
What script or program are you using?
Cheers, M

p4vasp v0.3.20r1

Hi again,

p4vasp does not add any broadening as far as I know. But I spoke to some colleagues and I think I understood the issue.

You use Fermi smearing (ISMEAR=-1) with the default SIGMA of 0.2. This alone will lead to a quite significant broadening of states in the DOS. The broadening will be reduced if you change your smearing to Gaussian (ISMEAR=0) and/or reduce your smearing width.

The charge postprocessing using LPARD ignores the smearing completely and works by manipulating the Fermi weights of the orbitals. Selected states will receive Fermi weight 1, and all others will be set to 0.
Thus, even if you see states in your DOS at the experimental bias voltage range, your settings do not guarantee that they will show up in the PARCHG file.

Let me know if this solves your problems and this topic can be locked.
Cheers, Michael

Yes. Thank you very much. I am curious about how the broadening is determined. I thought that SIGMA=0.2 was 0.2 eV, but the broadening of the DOS is clearly much more than that. How does the SIGMA value translate into eV of band width?

Dear KWH,

indeed the units of SIGMA are eV, but the effective broadening also depends on the smearing method (ISMEAR), more precisely the resulting occupation function. Some information on this is on our wiki in the section on K-point integration, to which I now introduced a link from both the ISMEAR and the SIGMA tag documentation. To learn even more about smearing, the advantages and drawbacks of different approaches, and their differences in effective broadening, I would suggest starting with a quite recent paper by Santos and Mazari: Phys. Rev. B 107, 195122 (2023). Section II B is about broadening and Fig. 1(a) will make it obvious that the same SIGMA leads to different broadening functions for different smearing methods, with Fermi-Dirac being the widest by far.

Let me know if this answers your question,
Michael

Terrific. Thanks for all your help