Re: Problem with Coulomb Energy

[Florian Weik]

Hi,

thank you for reporting this. There is a prefactor bug in the

energy calculation. I'll fix it and let you known then it is done.


For reference:

Line 1061 of src/core/electrostatics_magnetostatics/elc.cpp reads

  /* we count both i<->j and j<->i, so return just half of it */
  return 0.5 * eng;

which is incorrect, because 'eng' also contains the dipole correction

which is not a pair force.


Cheers,

Florian

On 11/13/19 2:50 PM, Fabian Glatzel wrote:
> Dear developers,
>
> using your package I encountered some problem I could not resolve by 
> myself. I'd be deeply grateful if you could offer some advice.
>
> The problem is the following:
> Using P3M and ELC with an external potential (via the "pot_diff" 
> argument), I get Coulomb energies that make no sense. The energy 
> contribution of charges in the "external" potential defined by 
> "pot_diff" seems to be off by a factor of 2.
> Let's say the system has a length of 10 and pot_diff=1V is set. 
> Further, only two particles are placed in the system at z=0 (charge 
> q=+1e) and at z=9 (charge q=-1e) and the prefactor for the P3M is set 
> such that the interactions among the particles are negligible. For 
> this case, I would expect a Coulomb energy of approx. 0.8eV. However, 
> espresso seems to have a different opinion, namely that it should be 
> 0.4eV.
>
> Please find a minimal example code attached to this email.
>
> I'm looking forward to hearing from you,
>
> Fabian Glatzel