[Pw_forum] total projections < 1
Lorenzo Paulatto
paulatto at sissa.it
Sat Nov 7 16:00:38 CET 2009
Dear Jiri,
first of all, welcome to the QE forum.
> am wondering about the following problem: using calculated file
> "Projections", I made a sum for each electronic state over all atoms
> (211 occupied + 43 unoccupied states, amorphous insulator).
I guess you are using projwfc.x to compute the projected densi
ty of states, if I'm wrong please correct me.
> (1) For low energy states, the sum is sufficiently close to 1.00 (>0.99).
>
> (2) However, several states below Fermi level the sum starts to
> decrease almost monotonously with increasing energy, being around 0.80
> only for the highest unoccupied state calculated. Please does it
> indicate a problem, or is it what I have to expect for some reason?
The lack of charge is a well known issue of projected density of states,
if you look at the output of the program you will notice that it reports
the "spilling factor", which is exactly what you have computed.
There are several possible explanations for the fenomenom you are
observing, I'll try to list them.
First of all, the projections only collect charge from the neighbourhood
of atoms, if some electron wish to go far away from the atom, it will not
be caught in the projection! High-energy empty states are more likely to
be highly delocalized, and they are computed to a lower accuracy during
self-consistency, which may be enough to explain what you see.
Second, computing integrals from the density of states may not be the most
sensible way to do it. The projected dos depends on the choice of smearing
and on the energy step; they are often chosen for aesthetic purpose (i.e.
to get a nice plot) more than to compute accurate integrals.
best regards
--
Lorenzo Paulatto
SISSA & DEMOCRITOS (Trieste)
phone: +39 040 3787 511
skype: paulatz
www: http://people.sissa.it/~paulatto/
----------------------------------------------------------------
SISSA Webmail https://webmail.sissa.it/
Powered by SquirrelMail http://www.squirrelmail.org/
More information about the users
mailing list