I have been using abinit to calculate absorption spectra of Aluminum. Depending on the temperature of the system, I use PAW pseudopotentials that have a frozen L shell or a valence L shell.
I have noticed that the two potentials give different results for the L-edge at low temperatures. To find out why this can be, I am trying to just plot the eigenvalues for the L-shell electrons on both systems, and compare them. However, I have not found a way to output the eigenenergies for the L-shell when it is included on the core. Is this something that is possible to obtain? I have tried the “pawfatbnd” variable but that one seems to only give the valence.
Hi, I think you might be able to get what you want in the process of making the PAW dataset. We typically use the open source code ATOMPAW for making PAW sets, and the currently distributed sets on the pseudo-dojo web site include input files to ATOMPAW. I’d suggest you get ATOMPAW running on your local machine, and try computing different Al sets with different core spaces. The output of ATOMPAW will give a lot of detail on the energies of the core and valence states it is computing. The tutorial pages include tutorials for running ATOMPAW.
No, because the core electrons are frozen and they don’t explicitly appear in the calculation. Their effect on the “valence” states is taken into account via the pseudopotential (both in PAW and in NC).
As already mentioned by Joe, you need to generate PAW pseudos including semicore states to explicitly include these states in the SCF calculation.
There’s some ongoing work to account for the relaxation of the core states within PAW but, for the time being, the easiest solution consists in using pseudos with semicore states included the “valence”
I have tried the “pawfatbnd” variable but that one seems to only give the valence.
Hi! Thanks to both of you for your detailed answers. I will have a look at the ATOMPAW manual and figure out the core energies from there. Thanks a lot!
Regards,
Pablo
Be sure to do careful comparisons between the atompaw output and abinit output, I think atompaw likes to report energies in Rydbergs, whereas abinit uses Hartrees.