DOS bug and inaccuracies

Documentation, Web site and code modifications

Moderators: baguetl, routerov

Locked
Emile
Posts: 15
Joined: Tue Dec 04, 2012 12:54 pm

DOS bug and inaccuracies

Post by Emile » Wed Jun 17, 2015 11:05 am

Dear all,

I just wanted to signal some problems with DOS and projected DOS. I use the last version of abinit with mpi parallelization.

* When performing a projected dos in parallel mode, the number of integrated bands is something equals to nband*nprocs. See the attached picture (Tot-DOS = prtdos 2; pDOS = prtdos 3; np stands for the number of processors used for the calcualations). Calculations are correct in sequential mode. The problem was already discussed here: http://forum.abinit.org/viewtopic.php?f=9&t=2252&p=7133&hilit=projected+DOS#p7133

* Specially when semicore electrons are included, the DOS or projected DOS reach false maximum density of state of 9999.9999. This value is in disagreement with the integrated DOS information in the abinit DOS file.
If I integrate the DOS with home codes, then I do not find back the correct number of bands, and accordingly, the correct number of electrons.

* It's not a big issue, but if dosdeltae goes down to a value lower than 0.00001 Hartree, then the printed energy is rounded and the energy column becomes meaningless. I admit there is little reason to look for such a fine energy grid, but when I was dealing with the 9999.9999 DOS value I used to do it.

More infos there : http://forum.abinit.org/viewtopic.php?f=8&t=2000&p=6322#p6322

Thanks for considering these points.
Emile
Attachments
W-DOS.jpg
W-DOS.jpg (24.45 KiB) Viewed 4847 times

Locked