ABINIT Discussion Forums

At the moment I try to calculate a 2x2x2 supercell of a material with 10 electrons in the primitive cell taking SO-coupling into account. The resulting number of bands, needed to reach the fermi energy is 10x2x2x2=80 bands. Together with a nice resolution of 80 kpoints, I have to calculate at least 6400 bands to reach the fermi energy and then I still need some bands above the fermi energy.

So my question is, if there is a way to calculate only specific bands? I would be enough to calculate bands 70 to 90.

Short answer: no. What you are really asking for is to compute high lying bands in the field of fixed low energy bands, in other words, what you really want to do is use pseudopotentials with a smaller valence space and larger core space. Depending on which elements you are using, this may or may not be possible.

Since I don't want to change the physical problem and only want to calculate another part of the solution, I searched a bit further in the documentation.

I found the variables wfoptalg and eshift. If you set wfoptalg to 3, the algorithm will solve for nbands around eshift. Except for some convergence problems, this is exactly what I wanted.

For all of you, who read this: Both wfoptalg and eshift are DEVELOP variables, so be careful!

Yes, wfoptalg 3 will sort of do what you want, but note that it works only in the non-self-consistent case (iscf -2) which means that in order to work you must read in the density (_DEN file) from a completed full run. This still might help you though because you could do a full ground state run with all bands, and prtden 1, but with pretty limited convergence so it runs faster, and then refine your states in the areas of interest using a second run with wfoptalg 3 and iscf -2, using the _DEN file from the first run for the input (irdden 1).