ABINIT Discussion Forums

Dear all,

I think there is a little problem in the DOS files printed by Abinit. Maximum printed value is 9999.9999. Such high value should be fine for most of the electronic density of states, however it appeared to be a too low value when one considers semicore states.

Here is a part of DOS file obtained on Al metal with PAW GGA PBE atomic data and following electronic configuration : 2s2-2p6-3s2-3p1 (11 valence electrons). Hereafter, energies correspond to 2s bands, but a similar behaviour also occurs for 2p bands at higher energies.

NRJ / DOS / Integrated DOS
-3.49334 4928.0850 0.1254
-3.49333 6407.1422 0.1828
-3.49332 7918.4936 0.2544
-3.49331 9999.9999 0.3435
-3.49330 9999.9999 0.4585
-3.49329 9999.9999 0.6144
-3.49328 9999.9999 0.8588
-3.49327 9999.9999 1.1714
-3.49326 9999.9999 1.5188
-3.49325 9999.9999 1.9685
-3.49324 42.0805 2.0000
-3.49323 40.7926 2.0000
-3.49322 38.7642 2.0000
-3.49321 36.1098 2.0000

DOS saturates at 9999.9999 whereas integration is still increasing strongly. This leads to an important loss of information. For instance, integrating this DOS by itself will not reproduce the total number of bands, sometimes with a strong difference. Computing total number of electrons according to this density of states leads to compute around 8 electrons instead of 11.

Same phenomenon appears for Cu metal with PAW GGA-PBE atomic data and following electronic configuration : 3s2-3p6-3d10-4s1

NRJ / DOS / Integrated DOS
-3.81726 9623.1872 1.1237
-3.81725 9983.1263 1.2219
-3.81724 9999.9999 1.3233
-3.81723 9999.9999 1.4290
-3.81722 9999.9999 1.5399
-3.81721 9999.9999 1.6580
-3.81720 9999.9999 1.7869
-3.81719 9999.9999 1.9381
-3.81718 0.6914 2.0000
-3.81717 0.6732 2.0000
-3.81716 0.6425 2.0000


An other point to be noticed is the fact that if one uses a value smaller than 1.0E-5 for dosdeltae, such as 0.5E-5 for instance, energies in the DOS file will look like this:

-3.81726
-3.81726
-3.81725
-3.81725
-3.81724
-3.81724

Instead of:
-3.817265
-3.817260
-3.817255
-3.817250
-3.817245
-3.817240

Just to let you know,
Best regards,

Emile Bévillon