[Pw_forum] question about pseudo and lowdin analysis

[Cyrille Barreteau]

Dear all,

I have a question concerning pseudopotential and lowdin charge analysis 
(projwfc).
I have done calculation on  Ni/Au(111) system (1 layer of Nickel on 15 
layers of Au)
with a Nickel pseudopotential for which the p wavefunction are not 
included and a
Gold pseudo for which they are included.
Doing the local charge analysis with projwfc I got the following result 
(where
Atom 1 is the surface Ni atom and Atom 2 is the Au sublayer atom).
Therefore it seems that there is a charge transfer between the Nickel 
atom and
the Gold atom.


    Atom #   1: total charge =   9.5860, s =  0.6508, p =  0.0000, d =  
8.9352,
                spin up      =   5.1427, s =  0.3140, p =  0.0000, d =  
4.8288,
                spin down    =   4.4433, s =  0.3369, p =  0.0000, d =  
4.1064,
                polarization =   0.6995, s = -0.0229, p =  0.0000, d =  
0.7224,
    Atom #   2: total charge =  11.3424, s =  0.4993, p =  1.2324, d =  
9.6107,
                spin up      =   5.6558, s =  0.2460, p =  0.6055, d =  
4.8043,
                spin down    =   5.6866, s =  0.2533, p =  0.6268, d =  
4.8065,
                polarization =  -0.0308, s = -0.0073, p = -0.0213, d = 
-0.0022,
   Spilling Parameter:   0.0011

Since I suspected a kind of artefact due to the slight inconsistency 
between
Au and Ni pseudo I have redone the calculation with a pseudo of Nickel 
including
p wavefunctions and I got the following result.
The charge transfer seems to have disappeared.
However the total magnetic moment is almost identical.


    Atom #   1: total charge =   9.9425, s =  0.4115, p =  0.7846, d =  
8.7464,
                spin up      =   5.3347, s =  0.2000, p =  0.3710, d =  
4.7637,
                spin down    =   4.6079, s =  0.2115, p =  0.4136, d =  
3.9827,
                polarization =   0.7268, s = -0.0115, p = -0.0426, d =  
0.7810,
    Atom #   2: total charge =  11.0206, s =  0.4441, p =  1.0158, d =  
9.5607,
                spin up      =   5.4973, s =  0.2188, p =  0.4993, d =  
4.7792,
                spin down    =   5.5233, s =  0.2253, p =  0.5164, d =  
4.7815,
                polarization =  -0.0260, s = -0.0065, p = -0.0171, d = 
-0.0023,

    Spilling Parameter:   0.0010


I do not really understand the exact meaning of this lowdin charge 
analysis.
It seems that the electrons of p chararacter of Nickel which cannot be 
projected on p wavefunction in the first pseudo
are "evaporating" into s and p orbitals of Nickel or Gold atoms. I am 
also puzzled by the fact that the spilling
parameter is equally good in both calculation while I would expect a 
larger value for the first case.
How can that be possible?

I also have a question concerning the generation of pseudopotential.
How the inclusion or not of p wavefunction affects the quality of the 
pseudo?
It seems that both pseudo are equally good but it is just an artefact of 
the projection
procedure which can lead to different local charge analysis.

thank you very much in advance for enligthening up this point.

best wishes

 cyrille

-- 
==================================================================
Cyrille Barreteau  | phone : +33 (0)1 69 08 29 51
CEA Saclay         | fax   : +33 (0)1 69 08 84 46
IRAMIS, SPCSI      | email cyrille.barreteau at cea.fr
Batiment 462       |
91191 Gif sur Yvette Cedex  FRANCE
          ~~~~~~~~~~~~~~~~~~~~~~~~
http://iramis.cea.fr/Pisp/cyrille.barreteau/
==================================================================