ABINIT Discussion Forums

Hi,

I would like to calculate a band structure with spin orbit coupling included. I am using the GGA-PBE PAW datasets available from http://www.abinit.org/downloads/PAW2.

In addition to the standard variables defining the crystal structure, my input file includes the following to compute the band structure and specify the use of spin-orbit coupling.



However, this input gives an error:


When I run the input file without pawspnorb 1, the band structure is successfully created, so the problem seems to arise when trying to include spin-orbit coupling.

Does anyone know how I can fix this error, or has anyone successfully calculated a band structure with PAW datasets, including spin-orbit coupling?

Thank you!

When I run the input file without pawspnorb 1, the band structure is successfully created, so the problem seems to arise when trying to include spin-orbit coupling.

The problem is not related to spin-orbit coupling but to the fact that you are using prtden=3.
This option generates the _PAWDEN file with the full all-electron charge density (valence+core).
Unfortunately nspinor == 2 is not yet implemented and the code stops in the routine that produces the _PAWDEN output file.
The calculation should complete successfully if you use prtden 1

Thanks, gmatteo! When I use prtden=1, the calculation runs successfully.
However, I see no change in the band structure compared to the non-spin orbit coupling case (I expect the spin orbit coupling to split a few of the degenerate bands).

I will double-check my input file, but just to verify: with PAW band structures, if I want to include spin orbit coupling, all I have to do is add pawspnorb 1 to my input file, right? There is no need to use different PAW datasets, or change anything else in the input file?

Thanks,
Ellen

if I want to include spin orbit coupling, all I have to do is add pawspnorb 1 to my input file, right? There is no need to use different PAW datasets, or change anything else in the input file?

Yes, pawspnorb 1 and nspinor 2 are enough.
Unlike NC pseudos, there's no need for specialised PAW datasets. See also:
http://www.abinit.org/doc/helpfiles/for ... #pawspnorb