[Pw_forum] Workfunction from slab calculations?

[Paolo Giannozzi]

On Monday 02 August 2004 20:27, Georg Heimel wrote:

> > not that I know. The potential at G=0 is given by the "alpha"-term
> > (the limit of (V_H+V_loc)(G) for G->0) plus V_xc(G=0).
>
> Thanks for the reply! May I ask, where in PWscf one can access
> this term (these terms)?

the  "alpha"-term is calculated in PW/init_vloc.f90, that calls 
PW/vloc_of_g.f90, and stored in real space into variable vltot(:). 
The Hartree and exchange-correlation potentials are calculated 
in PW/v_of_rho.f90 and stored in real space into variable vr(:,nspin). 
The potential that is actually used in the calculation of the Kohn-Sham 
equation is vrs(:,nspin) = vltot (:) + vr(:,nspin). The G=0 term can be
calculated as v0=SUM(vrs(:,nspin))*omega/nr1s/nr2s/nr3s. In a parallel
calculation, "call reduce(1,v0)" must follow.

> The problem is that, obviously, something is not added back on before
> the eigenvalues and the fermi energy is printed out. How else could they 
> be positive

they can be whatever they like to be: the average of the potential is
arbitrary in a periodic solid (see L. Kleinman, PRB24, 7412 (1981), 
http://prola.aps.org/pdf/PRB/v24/i12/p7412_1). This paper also 
suggests a well-defined value for the average potential in a 
semi-infinite crystal. 

> and scale with the inverse of the unit cell volume V (vacuum 
> thickness in a slab calculation)?

I guess that the average of the potential scales as 1/V if you increase 
the vacuum in a slab geometry

Anyway: I don't think that there is anything wrong or inconsistent with
the way the potential is presently calculated in PWscf. I think that in 
the past there have been calculations of band offsets, work functions,
and similar quantities, using PWscf, and nobody complained about 
the way the average of the potential is calculated, since after all it
doesn't matter, provided everything is done in a consistent way. 

If you have some specific cases yielding obviously wrong results, 
please submit them. I would like to fix the discrepancy between 
PWscf and the Car-Parrinello codes, yielding different eigenvalues 
because they use a different definition of the average potential. 
Maybe if you have a better definition, we might implement it.

Paolo

-- 
Paolo Giannozzi             e-mail:  giannozz at nest.sns.it
Scuola Normale Superiore    Phone:   +39/050-509876, Fax:-563513 
Piazza dei Cavalieri 7      I-56126 Pisa, Italy