 Re: Calculation of accurate band gap ( No.1 ) |
- Date: 2023/02/14 20:18
- Name: Vipin Kumar <kumar.vipin118@gmail.com>
- Dear OpenMx developers,
Any kind of suggestion to calculate the accurate band gap in above mentioned systems using openmx DFT code.
Sincerely yours
Vipin
|
| Re: Calculation of accurate band gap ( No.2 ) |
- Date: 2023/02/15 20:14
- Name: Pavel Ondračka <pavel.ondracka@email.cz>
- I believe that the only option you currently have in OpenMX is using the DFT+U, however you will have to obviously select some (more or less arbitrary) value of U so at that point you unfortunately lose the predictive strength of DFT.
Also keep in mind if you are interested in the optical properties, the band gap you get from the DFT (even with hybrid like HSE06) is not corresponding to the optical band gap, there are usually also excitons at play (I have no idea about your material specifically, but halides are known for strong excitonic effects). So if this would be a starting point for optical calculations, please note that OpenMX also doesn't support calculating excitonic effects (see for example how the Si optical calculation in https://openmx-square.org/openmx_man3.9/node198.html fails to reproduce the first excitonic peak in the experimental Si spectra).
|
| Re: Calculation of accurate band gap ( No.3 ) |
- Date: 2023/02/16 14:48
- Name: Vipin Kumar
- Dear Pavel,
Thank you very much for your reply. I tried a few calculations with DFT+U, but the results are not consistence with the reported ones. I tried the U value in the hot and trial method. Is there any specific way to choose the U value for different elements or it is just the hit-and-trial method?
Sincerely,
Vipin
|
| Re: Calculation of accurate band gap ( No.4 ) |
- Date: 2023/02/16 18:00
- Name: Pavel Ondračka <pavel.ondracka@email.cz>
- Well, there are some ways how to estimate the proper U value ab initio. See for example https://doi.org/10.1103/PhysRevB.71.035105 but I'm not sure if this could be done with OpenMX and my experience in general with this method was not the good (but this specific example is quite old, so maybe you can find something better now).
Usually people try to fit the U value to match experimental measurements (lattice constants, band gap, mechanical properties, etc...). IMO it's not that good science, but if you have no better options and you can give good reasoning why you choose the specific value, than I believe it is OK (but make sure you check multiple properties, I've seen cases where people fitted U to get a better band gap, but ended with worse lattice constants than before or incorrect magnetic structure, etc. in such cases you really have to go for better method).
If you want to fit it to the band gap, it is best to compare against photoelectron + inverse photoelectron (photoemission) spectroscopy measurements if available and not the optical ones. And its also better to compare the whole DOS to the measured spectra and not just the scalar gap value, there is some significant broadening in the experiment and they usually fit the band edges somehow. In my experience best way to compare is to broaden the DOS such that the slope of the top valence band edge agrees with the experiment, that align it with the experimental one and check how the alignment of the conductive band edge looks like. See for example here: https://doi.org/10.1088/0022-3727/49/39/395301 . But its difficult to find good high-quality photoelectron + inverse photoelectron measurements :-(
 |
|
This thread is locked.Only browsing is available.
Calculation of accurate band gap