[Pw_forum] Choice of pseudpotential (d orbitals for Fe)

[Paolo Giannozzi]

Cristian Degli Esposti Boschi wrote:

> initially I chose the pseudopotentials (in the pwscf site) that
> include the d orbitals contribution, namely
> 
> Fe.pbe-nd-rrkjus.UPF or
> Fe.pz-nd-rrkjus.UPF

The notation is misleading in this case: the d states are in no way
"semicore" states for Fe, they are valence states, like them or not.
Note that LDA results ("pz") are unlikely to get a warm reception
in the chemical community: don't spend too much time on it.

> Now, I wonder if the problem of excessive doming could be due
> to the use of wrong pseudopotential ... 

did you make any tests on small molecules? Sometimes 3d transition
metals require 3s and 3p "semicore" states into the valence for
good results in difficult cases like the one you are dealing with.
Those pseudopotentials are not "wrong" for sure, but they have
been tested in solid-state systems like crystalline Fe, not in
(notoriously nasty) organometallic systems. You may give a try to
Fe.pbe-sp-van.UPF, that has semicore 3s and 3p (in addition to 3d,
of course) in valence. Note that more electrons == more cpu time.

> I would like to try with BLYP's but the one for Fe
> does not seem to include explicitly d orbitals. Are there 
 > other sources of pseudopotentials?

it depends on your definition of "sources of pseudopotentials".
If it includes pseudopotential generation codes, there is plenty
of them, not necessarily easy to use. The pseudopotentials download
page and the wiki section on pseudpotentials contains a fair amount
of information.

Paolo
-- 
Paolo Giannozzi, Democritos and University of Udine, Italy