Duo to SO interaction, elk4.3.6 produces one valence band with top just above -1 eV at Gamma point, wheras data obtained in wien2k and FPLO seem to be much more consisten (the whole band is shiftet by at least ~0.2 eV towards lower energy)
Do you know the possible reasons of the difference? I changes variuos parameters, eg rmt radii, but problem has persisted.
As I read from your plots the position of the band of interest is due to the SO splitting of the Bi-p bands at Gamma point. It is a little hard to read off from the plots, but I would get that there is a trend of increasing splittings from 1.4, 1.6 to 1.75 when going from ELK -> W2k -> FPLO.
These differences are due to a known problem of scalar relativistic treatments of the SO splittings of p-states in heavy materials, see eg PRB 63, 035103 (2001). This problem was at least partially remedied by including extra Dirac-based local orbitals to the variational basis, see PRB 64, 153102 (2001).
Now we can observe that:
FPLO uses Dirac states and therefore totally avoids this problem.
ELK uses (as most other codes do) scalar relativistic states, and therefore underestimates the splitting
Wien2k uses the above remedy which (partially) corrects this error
which then is consistent with the calculated SO splittings.
It should be mentioned that this problem only concerns the difference of p1/2 states close to the nuclei, when treated scalar or fully relativistically, and usually does not affect other states. This is consistent with your band structures. For example the states close to Fermi level are essentially unaffected.
Best wishes,
Lars
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Dear Elk Users
I calculated band structure of LaBi using both HQ and VQ parameters + SO and PBE-GGA http://www.pnas.org/content/113/25/E3475/F7.expansion.html
Duo to SO interaction, elk4.3.6 produces one valence band with top just above -1 eV at Gamma point, wheras data obtained in wien2k and FPLO seem to be much more consisten (the whole band is shiftet by at least ~0.2 eV towards lower energy)
Do you know the possible reasons of the difference? I changes variuos parameters, eg rmt radii, but problem has persisted.
FPLO:
http://picpaste.com/band-1G0hedaO.png
Elk
http://picpaste.com/elk-bands-s1wCneo9.png
Wien2k
http://www.pnas.org/content/113/25/E3475/F7.expansion.html
avec
0.5000000000 0.5000000000 0.0000000000
0.5000000000 0.0000000000 0.5000000000
0.0000000000 0.5000000000 0.5000000000
scale
12.4910895210
atoms
2 : nspecies
'Bi.in' : spfname
1
0.000000000000000 0.000000000000000 0.000000000000000 0.00000000 0.00000000 0.00000000
'La.in' : spfname
1
0.500000000000000 0.500000000000000 0.500000000000000 0.00000000 0.00000000 0.00000000
Thanks in advance.
Ana
Last edit: AnaG 2017-11-16
Hi Ana,
Could you post your entire elk.in?
Thanks,
Kay.
Dear Ana,
As I read from your plots the position of the band of interest is due to the SO splitting of the Bi-p bands at Gamma point. It is a little hard to read off from the plots, but I would get that there is a trend of increasing splittings from 1.4, 1.6 to 1.75 when going from ELK -> W2k -> FPLO.
These differences are due to a known problem of scalar relativistic treatments of the SO splittings of p-states in heavy materials, see eg PRB 63, 035103 (2001). This problem was at least partially remedied by including extra Dirac-based local orbitals to the variational basis, see PRB 64, 153102 (2001).
Now we can observe that:
FPLO uses Dirac states and therefore totally avoids this problem.
ELK uses (as most other codes do) scalar relativistic states, and therefore underestimates the splitting
Wien2k uses the above remedy which (partially) corrects this error
which then is consistent with the calculated SO splittings.
It should be mentioned that this problem only concerns the difference of p1/2 states close to the nuclei, when treated scalar or fully relativistically, and usually does not affect other states. This is consistent with your band structures. For example the states close to Fermi level are essentially unaffected.
Best wishes,
Lars