The calculation of Indium Arsenide

[yongjunwinwin]

I used ATK to calculate the band structure of Indium Arsenide (the script and result is attached) under two different k point sampling, however, I found the band gap I got is ~0.7 which is larger than reported 0.35 eV. In addition, the split off band did not show in my calculated band structure.

I also found the the zero energy level shift with different k point sampling (3*3*3 and 6*6*6), what would be the reason for the phenomenon I mention?

[Anders Blom]

In general you should not expect DFT to give the correct band gap of semiconductors. ATK does however have tight-binding models for InAs (and other III-V binaries), which wold give a correct gap.

The energy shift is probably an artifact, may due to too few k-points; the specific position of the Fermi level in the gap is in some sense irrelevant, as long as its relatively far from the band edges (which however doesn't seem to be the case for 6x6x6). But what about 9x9x9? Or 11x11x11? You need to make sure you have enough.

The split-off band is due to spin-orbit interaction, which is coming in a future version of ATK.

[yongjunwinwin]

So DFT could not precisely predict the bandgap of semiconductor, which I think is pretty important in latter analysis?

Yeah, I have try 12*12*12, in whose result the Fermi level did not also stay in the middle, a little up I am afraid..

[Anders Blom]

Seems to be an artifact of LDA. With GGA it looks better.

[yongjunwinwin]

Yeah, my origin calculation is done using GGA ... however, what I am concerned about is that if the band gap could be precisely predicted?

[Anders Blom]

Not with LDA or GGA. Maybe with Meta-GGA/TB09. Otherwise you need to use Huckel or Slater-Koster tight-binding.

[zh]

The overestimation of band gap for InAs by the LDA/GGA in your calculations looks strange. Since InAs is a semiconductor, the use of 6*6*6 for the k-mesh may be sufficient.  You may also take care of the pseudopotential of Indium. In some cases, the treatment  of semicore states of Indium is necessary.

[Anders Blom]

Valid points. However, as a general note, one should not expect LDA/GGA to give an accurate estimate of the band gap.

[nori]

If my memory is right, this problem only occuers when GGA is used for 'In'.
So if you use LDA instead, it's expected that the band gap is under-estimated.

this problem was also seen in the calculation of InP.
I agree with zh that the pseudopotential of Indium would be the key to this issue.

[Anders Blom]

I certainly agree that you can get closer by perhaps considering different version of LDA/GGA and the pseudopotential. I just want to make sure that it's clear that one should not in general expect LDA/GGA to give correct band gaps, no matter how good your pseudopotential, k-point sampling, basis set, or whatever.

[yongjunwinwin]

Then what about other properties, I mean would it affected other properties as well?

[gM]

Hello, I have convergence issues when trying to do an SD scan of a 5nm long, 1nm wide InAs nanowire.   The simulation reaches 200 iterations and warns that it finished without convergence. I use the following:

Extended Huckel Calculator
k-point sampling 3x3x100
CerdaHuckelParameters.Arsenic_InAs_GW_Basis
CerdaHuckelParameters.Indium_InAs_GW_Basis

I have attached the script.  Could anyone tell me what I'm doing wrong with it? Should I increase the k-point sampling?

[Anders Blom]

Well, you should definitely not increase the k-point sampling. You already have too many Since this system is effectively one-dimensional, you should only have 1 k-point in A and B. That will save you a lot of time, but not solve the convergence. For that, check your boundary condition; there is a bug in Neumann for 12.2, but you don't need that anyway, just use periodic (that is, use FFT2D).

[gM]

Thanks Dr. Blom. That fixed the issue.