Re: [ESPResSo-devel] User Guide on ELC and noneutralization

[Vincent Ballenegger]

Hi Stefan and Axel,

>With "-noneutralization", ELC does not add any neutralizing background, hence
>the name. Formally, that can be understood as two equally charged,
>neutralizing walls at infinite distance
That's precisely why the name "no neutralization" is not so clear. Indeed, one 
could say that there IS a neutralization when using the -noneutralization 
option, namely the neutralization by two charged walls!

I wouldn't say that these walls are at infinity, because the energy would then 
be infinite, see last term in eq. (17) of the JCP paper in the limit h--
>infinity.

Note: Rather than thinking in terms of surrounding neutralizing charged walls 
at +h and -h, one can alternatively think in terms of a regularization. The a 
priori infinite energy is regularized according to the prescription given by 
eq. (9) in the JCP paper when one uses the -noneutralization option.

>To automatically detect whether one has to neutralize or not, ELC 
>would also need to check whether charge plates are present, which is not 
>implemented.
I agree (implementing this should however not be too difficult).
If charge plates are present, one could alternatively simply issue a warning 
that the user has to adjust the total charge Q_tot in the background charge 
correction.

> And since P3M by default neutralizes, ELC also does.
OK, but the default neutralization in P3M is good for most users, while the 
default neutralization in P3M/ELC correspond to having a parabolic potential 
V(z) in the simulation box, where V(z) depends on the size Lz of the 
simulation box in the non-periodic direction. That potential is a feature that 
most users will not want to have!

Therefore, I still think that it would be better to activate by default the -
noneutralization option in P3M/ELC, which could be renamed for instance in 
"background charge correction" as in the JCP paper.

Vincent