slab correction in PPPM/disp and Ewald/disp

Dear lammps Users

As I concerned by reading the documentation, the slab correction is performed to constrain the Coulombic long-range interaction in one direction (as performed in Yeh and Berkowitz work).
I wonder If the slab-correction removes both Coulombic and Lenard-Jones dispersion interactions in PPPM/disp or Ewald/disp method?
I will appreciate any comments.

Best regards,
Reza

Dear lammps Users

As I concerned by reading the documentation, the slab correction is
performed to constrain the Coulombic long-range interaction in one direction
(as performed in Yeh and Berkowitz work).
I wonder If the slab-correction removes both Coulombic and Lenard-Jones
dispersion interactions in PPPM/disp or Ewald/disp method?

the slab correction is for coulomb only and only cancels inter-slab
interactions of the residual dipole moment across the slab.

axel.

Thank you so much Axel

the slab correction is for coulomb only and only cancels inter-slab
interactions of the residual dipole moment across the slab.

You mean that Coulombic interaction of Ions wont be canceled across the slab? Actually, I am confused by your expression and the Lammps documentation where the kspace_modify is introduced, it is stated that slab correction is extended to non-neutral systems.
Could you explain how the modification changes the Yeh slab-correction?

Thank you so much for your precious guides.
Best regards
Reza

Thank you so much Axel

the slab correction is for coulomb only and only cancels inter-slab
interactions of the residual dipole moment across the slab.

You mean that Coulombic interaction of Ions wont be canceled across the
slab?

that is not what i am saying. the slab correction applies to the
*residual* dipole of the entire system.

Actually, I am confused by your expression and the Lammps
documentation where the kspace_modify is introduced, it is stated that slab
correction is extended to non-neutral systems.
Could you explain how the modification changes the Yeh slab-correction?

i don't understand what you are asking here. please elaborate in more
detail and quote the particular sections to the manual you are
referring to.

any periodic system with a non-neutral cell has essentially a
diverging energy. all reciprocal space based long-range solvers work
around this by ignoring the divergent (k=0) term. this applies to slab
systems as well. if you had a residual non-zero charge, the slabs
would repel each other, but since you have the same repulsive force in
positive and negative directions, it will cancel.

axel.

that is not what i am saying. the slab correction applies to the
residual dipole of the entire system.

So, the system should be entirely neutral.

i don’t understand what you are asking here. please elaborate in more
detail and quote the particular sections to the manual you are
referring to.

In the lammps documentation on the following page:
http://lammps.sandia.gov/doc/kspace_modify.html

on the last line of the paragraph which the slab keyword is introduced, this sentence was added;" The slab option is also extended to non-neutral systems (Ballenegger)."

Actually I am looking for a way to use PPPM/disp or Ewald/disp method for a 2 dimensional system and it is important to cancel both Lenard-Jones and Coulombic long-range interactions across the non-periodic direction. Could you give a suggestion?

that is not what i am saying. the slab correction applies to the
*residual* dipole of the entire system.

So, the system should be entirely neutral.

no, but the larger the net charge the larger the risk of deviation
from the behavior of a neutral system.

i don't understand what you are asking here. please elaborate in more
detail and quote the particular sections to the manual you are
referring to.

In the lammps documentation on the following page:
http://lammps.sandia.gov/doc/kspace_modify.html
on the last line of the paragraph which the slab keyword is introduced, this
sentence was added;" The slab option is also extended to non-neutral systems
(Ballenegger)."

if you have a system with a net charge, Sum(r_i * q_i) is not
invariant to translation, so for the slab correction to be meaningful
you need to remove the translational bias, e.g. by subtracting
Sum(q_i)*(geometric center)

Actually I am looking for a way to use PPPM/disp or Ewald/disp method for a
2 dimensional system and it is important to cancel both Lenard-Jones and
Coulombic long-range interactions across the non-periodic direction. Could
you give a suggestion?

you are not making much sense here. the attractive LJ term decays with
r**6 (the r**12 term is not at all considered in the /disp styles,
since it is so extremely short ranged).
if you have a thin slab of, say, 10 \AA width, then particles will be
30 \AA apart from their images in z-direction, so that the magnitude
of their interactions will have decayed by almost 9 orders of
magnitude.
that is far smaller than the typical PPPM/Ewald accuracy threshold
used. i don't see a convincing point for having a slab correction. if
you feel this strongly about accuracy, you should rather go about and
implement a proper 2d-ewald summation for coulomb (only) with no PBC
in z, and then use an LJ cutoff long enough to span the entire width
of the slab. of course, you will also have to re-parameterize your
model, since i'd doubt that any existing force field parameters will
be tuned for such a setup at the high accuracy you are looking for.

i don't even want to argue systematic force field accuracy issues or
the fact that you are - after all - simulating an essentially
unphysical system when you have a system with a net charge.

rather than arguing about how to do what cannot be done or makes no
sense, you should explain what in your planned simulations makes this
all necessary and provide some convincing proof that the existing
features in LAMMPS are not sufficient.

axel.

I asked about non-neutral systems because I have not seen the sentence, which is added to the documentation, before and this questions comes to my mind that, might further changes is added to the slab correction.
The system I want to perform is a slab-like system(neutral-system) in order to calculate surface tension. The impact of electric field is going to be studied on the system. As far as I know, it is essential to use slab-correction in z-direction. So if I want to use PPPM/disp method I should cancel both coulombic and Lenard-Jones long-range interactions in z-direction while the slab-correction implemented in lammps just cancels the Coulombic one.

now here is the $1,000,000 question:

how are you going to model the effect of the electric field at the
same level of detail and accuracy?

axel.

So if I want to use PPPM/disp
method I should cancel both coulombic and Lenard-Jones long-range
interactions in z-direction while the slab-correction implemented in
lammps just cancels the Coulombic one.

The Yeh Berkowitz slab correction in LAMMPS (for Coulombics) is only a first order correction anyway. Like Axel said, you'd be better off using a true 2D Ewald sum if you need that much accuracy. But it's probably overkill since your potential won't be that accurate either.

Stan

how are you going to model the effect of the electric field at the
same level of detail and accuracy?

Well, by using fix Efield command.
and the field is perpendicular to the slab, that’s why I think slab-correction is essential.

Thank you Stan

How can I benefit 2d Ewald in lammps? Should I use the slab-correction together with P P F boundary condition?

Actually, I read your article " Liquid-vapor interface of Stockmayer fluid in a uniform external field (phys. rev. E 91, 022309(2015))". You were considered the long-range part of the Lenard-Jones interactions, the question is how did you implement it in the simulation? Did you apply a long cutoff radius like what Axel said?