[Pw_forum] Re: Dipole correction in vacuum, polar slabs

[Guido Fratesi]

> 1) Reference to this facility seem to be completely removed from the
> documentation (distinct, of course, from the facility to add an external
> electric field, tefield =.true.). In particular, the fact that the flag
> dipfield = .true. is needed is only mentioned in the source (took me a
> while to find that out!). Is this a conscious decision? Is the dipole
> correction now "unsupported"? Was it ever "supported"? Is the routine
> known to be buggy? One of the main reasons I started using PWSCF is for
> this reason...

I don't know about "supported"/"unsupported".
However, I'm currently using dipole correction for calculations of adatom 
adsorption on a single side of a slab, and I found workfunction changes in good 
agreement (that is, slightly overestimated) with experimental findings.

To use it, I'm setting
&CONTROL
   tefield  = .true.  !!this to add some kind of electric field
   dipfield = .true.  !!this to add also the field given by the dipole
/
&SYSTEM
   edir=3       !!field along z (I'v never tried x or y)
   eamp=0       !!add only dipole field, no additional electric field
   emaxpos=0.45 !!position of the maximum of the added potential
                !!depends on how I've built the supercell
   eopreg =0.10 !!see below
/

The added potential is increasing from [emaxpos+eamp-1]*celldm(edir)*alat to 
[emaxpos]*celldm(edir)*alat with slope eamp, then decreasing to 
[emaxpos+eamp]*celldm(edir)*alat to recover periodic boundary conditions for 
the potential. The region in which the potential is rapidly decreasing to 
conserve periodicity should be in vacuum where an additional electric field 
takes no significant effect.

> 2) I see some weird behaviour when the dipole field is used, and I'd like
> to know if it's a bug, or if it arises from a real physical mechanism.
> What I've observed is that convergence depends on the relative position of
> the slab within the supercell: if the slab is roughly symmetric about
> zero, and the dipole layer located at about 0.5 of the cell, the
> correction works, and convergence is achieved; if the slab and layer are
> shifted, the convergence is never reached and the total energy shoots to
> large positive values.

Have you checked the position of the fast decrease of the potential, if it is 
in vacuum or not?
Notice: the electric field can also be plotted by using plot_num=12 in the 
postprocessing program pp.x.

> At first appearance, this to me is a bug. However, it also occurred to me
> that it might just be related to the old problematic definition of the
> dipole moment within the infinite lattice. Has anyone any idea why this
> behaviour arises? If the potential converges, can I trust the calculation?

The dipole along z is given correctly, since it is computed in 
PW/compute_dip.f90 as

1/Omega \int \rho(r) (r-r0) d^3r

where r0=(0,0,z0) and z0 is in the middle of the region in which the electric 
field is applied (which should therefore be in the slab region). As far as see 
from the code, the dipole field correction is implemented only for field 
direction along z.

Hope this helps.

Regards,
Guido Fratesi