[Pw_forum] ion problem

[Giovanni Cantele]

Cyrille Barreteau wrote:
> Dear all
>
> I have a question related to this problem.
> I am interested in the study of charged clusters.
> If I understand well, due to to the periodicity of the plane wave codes
> one has to substract the correct "background" in order to keep the
> energy from going to infinity..
>   
This is automatically done by the code, as you may read in the INPUT_PW
documentation file:
nelec          REAL
               number of electron in the unit cell
               (may be noninteger if you wish)
               Default: the same as ionic charge (neutral cell)
               A compensating jellium background is inserted
               to remove divergences if the cell is not neutral

So, you set nelec in the input file, and that should be enough.

> I would like to compare the energy of clusters with different
> charges -e -2e etc... Do you have some advices to do such calculation?
> Is it necessary to take even larger supercells than for neutral clusters?
>   
The problem with charged clusters is that spurious interactions are 
introduced by the
periodic boundary conditions, so as convergence would be achieved at 
very large supercells.
Neverthelessm the total energy can be corrected, to get rid of such 
interactions, as
described by Makov-Payne, PRB (1995). Such corrections are currently 
implemented
in Quantum-ESPRESSO, with the only warning that, as far as I know, 
corrections are
applied to the total energy but are not to the forces.

If you are using espresso-3.2, you should
- run a self-consistent calculation for you charged cluster
- run post-processing code pp.x, for obtaining the total charge density 
(plot_num=0) in
a suitable format
- run dipole.x, you may read the header of PP/dipole.f90 for very short 
explanation

You finally end up with a Makov-Payne correction that must be added to 
the total energy
WITH its sign.
You may try the same charged cluster with two different supercells, and 
compare the
two uncorrected energies, as well as the corrected ones, finding the 
latter much more
converged.

Next releases will probably have an easier procedure. In the CVS version 
I understand that a new variable
(assume_isolated) has been added to the pw.x input, that, if set to 
.TRUE., should allow to get
the MP correction directly within the scf loop (so post-processing is 
not mandatory).

Giovanni


-- 



Dr. Giovanni Cantele
Coherentia CNR-INFM and Dipartimento di Scienze Fisiche
Universita' di Napoli "Federico II"
Complesso Universitario di Monte S. Angelo - Ed. 6
Via Cintia, I-80126, Napoli, Italy
Phone: +39 081 676910
Fax:   +39 081 676346
E-mail: Giovanni.Cantele at na.infn.it
Web: http://people.na.infn.it/~cantele