ABINIT Discussion Forums

Hi,

I've been generating MRPP pseudos for loss function calculations using the abinit+dp combination. I need the pseudos to include semicore states for my purposes. I'd go with the HGH pseudos available on the abinit website, but unfortunately using them causes the KSS file generation in abinit to fail due
to some commutator calculation.

Currently I'm working on Cu. I use the APE code for this. When I use MRPP pseudos obtained by solving the Dirac equation, the density of states changes dramatically from what I get with by generating the pseudos using a non-relativistic atomic calculation. The fully relativistic HGH pseudos and non-relativistic MRPP pseudos give an essentially identical DOS. The pseudos themselves and all-electron KS wave functions don't change much when I change from non-rel. calculation to a relativistic one.

Can anyone enlighten me on where the big change in the DOS could arise from? Technical issue or am I missing some physics here?
Working with the non-rel. pseudos will probably not cause a problem for me, but I'd like to understand
what's going on.

Best regards,
very confused experimentalist Kari

Here's a ps file plot of the DOSs I got from the calculation I mentioned above. The HGH & non-rel MRPP DOS look pretty much the same as in Campillo et. al.
PRB 59 (1999) 12188-12191. All calculations were performed at fixed ecut and kgrid at the experimental lattice constant. The ecut is a bit of an overkill at 80 Ha and the kgrid is an unshifted 10x10x10 MP grip. I used prtdos 2 for DOS output.

I still can't figure this out, and I'd appreciate all ideas you might have on this!

Cheers,
Kari

This was mainly due to me not reading the psp info files from the abinit docs thoroughly enough. Also a small bug fix was needed for APE to print PPs in pspcod 5. So if anyone needs to generate relativistic pseudos for abinit with APE, use APE version 2.0.2 and abinit5 for the pseudopotential format.