I'm attempting to reproduce the bandstructure of LaFeAsO from Ref [1], Fig 2; as well as the much simpler FeSe.
I expect (from experiments and published DFT results, i.e [1]) the bands near/crossing the Fermi surface to include: two hole-like Fermi surfaces coming from a 2-fold degeneracy at Gamma; and two electron Fermi surfaces, whose origins are two 2-fold degeneracies at k=(pi,pi).
For LaFeAsO, I managed to reproduce a hole-like degeneracy at Gamma, and additionally my Fermi energy agrees with [1]; however, the rest of the low-energy bandstructure is no good. I've tried cranking up the cutoffs sufficiently large, at a large time cost, but have yet to see convergence to the expected results. Likewise for tuning the smearing through lowering "degauss".
The simpler and computationally faster FeSe calculation doesn't even come close to what I expect, neither at Gamma or (pi,pi).
I've successfully calculated the bandstructures for Silicon and Graphene, which makes me feel like I've got a good enough grasp of the QEspresso basics. Is there something special to this family of materials that I'm missing? Any insight would be appreciated!
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
I've included my QE codes for both LaFeAsO and FeSe -- codes for the energy (scf) and bands (bands) each -- in order to save someone time in case they wanted to try to recompute it on their machine. If you're willing to look at the code, perhaps focus on the computationally simpler FeSe; but the codes are otherwise identical, up to minor details.
FYI, I'm using the SSSP pseudopotentials on MaterialsCloud [3].
[1]: K. Kuroki, et al. PRB 79, 224511 (2009) -- Note that Fig 2 has been "unfolded" into the 1-Fe BZ
[2]: The code I shared uses the experimentally determined bulk-FeSe parameters, but I have tried various others (see attached README_constants for a list of resources).
[3]: https://www.materialscloud.org/discover/sssp/table/efficiency
In principle, the steps to obtain bandstructures your structures are the same for Silicon and Graphene. I guess you may not have considered Hubbard_U parameter or starting_magnetization, because of Fe atom have d orbital and you need to use DFT+U .
You may see input files examples for atoms with d orbits and how Hubbard_U is applied.
Turns out I simply placed the second atom at the wrong location in the unit cell, which thus broke the system symmetry, and gapped the 2D irreducible representations of the space group symmetry at M. (No need for anything additional apparently, e.g Hubbard_U, if the goal is to capture the degeneracies at high symmetry points due to the symmetry. It seems vanilla QEspresso is enough.)
Since ResearchGate won't allow me to delete my silly question, I've attach a corrected QEspresso code for FeSe, so that any future researcher and DFT newbie can learn from it.
In order to run the code, the instructions are as follows (assuming Unix/Linux OS using BASH shell):
(1) Place all three files in the same folder. In that folder, create two empty folders named "out" and "pseudopot". Go to MaterialsCloud (https://www.materialscloud.org/discover/sssp/table/efficiency) and download the SSSP pseudopotentials for Fe and Se -- place them in the folder named "pseudopot".
(2) The electronic structure calculation is run with the following command:
``` pw.x FeSe.scf.out ```
which creates the file "FeSe.scf.out". One can quickly output the total and Fermi energy with the command
``` less FeSe.scf.out | grep energy ```
(3) The calculated electronic structure can be used to compute the band structure at various k-points specified in "FeSe.bands.in".
``` pw.x FeSe.bands.out ```
``` bands.x < FeSe.bands.data.in >FeSe.bands.data.out ```
(4) Finally, one can quickly plot the output with
``` plotband.x bands_FeSe.dat ```
Which will prompt you with a few plotting questions.
Cheers!
- Badges
- Science topic