[lammps-users] Group based cutoff?

LAMMPS LAMMPS Mailing List Mirror
1

Dear Lammps users and developers,

I want to implement a group based cutoff method in the LAMMPS code but I didn’t find how I can perform this kind of simulation. Does your PPPM method consider the “charge neutrality” inside the cutoff? Does it use an atom based cutoff?

Also, when I read articles, I encounter that they have implemented different methods such as: 1- Ewald 2- Atom based 3- Group based

My understanding is that the Ewald method is either atom based or group based. Did I miss something here? Why they separate the Ewald from group based and atom based techniques? According to my understanding they are basically subdivision of the Ewald approach. I wonder if you could give me some elaboration. Thanks!

Amir Vahid

Chem Eng Dept

The University of Akron

Akron OH 44325-3906

2

Dear Lammps users and developers,

I want to implement a group based cutoff method in the LAMMPS code but I

please explain in more detail, how a "group based" cutoff would be
different from the current scheme that LAMMPS employs and particularly
what the benefits would be. please note that LAMMPS already has a
"group" feature, but i assume this is not what you are referring to, right?

didn’t find how I can perform this kind of simulation. Does your PPPM method
consider the “charge neutrality” inside the cutoff? Does it use an atom
based cutoff?

have you read the literature about this method??
with PPPM - like any other ewald-like based method -
you apply a damping to smoothly reduce the coulomb
interactions inside the real space cutoff so that they
are vanishing at the cutoff itself. and then you compensate
for it by a computing the lattice interactions of the charges
with the opposite damping in reciprocal space.
why would there be a need to have charge neutrality
inside the real-space cutoff?

Also, when I read articles, I encounter that they have implemented different
methods such as: 1- Ewald 2- Atom based 3- Group based

My understanding is that the Ewald method is either atom based or group
based. Did I miss something here? Why they separate the Ewald from group
based and atom based techniques? According to my understanding they are
basically subdivision of the Ewald approach. I wonder if you could give me
some elaboration. Thanks!

i really don't understand what you are talking about. this is very confusing.
please elaborate on your statements or give some examples.

axel.

3

Sorry, I forgot to attach the papers. I have enclosed them to this email.

The group based cutoffs means that the entire molecule (in small organic
compounds) or some groups (with the net charge of zero) are placed inside
the cutoff. This is the basis for the group based cutoff technique in which
electrostatic forces decay faster.

If you look into the attached papers (alternative pairwise for Ewald sum),
they basically, assumed that the electrostatic interaction is short ranged
in condensed systems and there is no Fourier space (long range). Therefore,
the neutralization of the charge is important inside the cutoff radius. I
think since PPPM has the long range part would not require this criteria. I
don't know how they implemented this neutralization, did they place the
entire molecule inside the cutoff radius or they use another approach. My
understanding is that they basically apply it through force and potential
equations. Group based cutoff (placing the entire molecule inside the cutoff
radius) is another discussion. Any comments would be appreciated.

In accelrys Material Studio software there are 3 options for the
electrostatic interactions, namely: 1- Ewald 2- Atom based 3- Group based

I don't know what the difference b/w Ewald and atom based is. B/c Ewald is
essentially atom based and I don't know why they have separated it from each
other.

Regards,

Amir Vahid
Chem Eng Dept
The University of Akron
Akron OH 44325-3906

WolfEwaldChemPhys99.pdf (558 KB)

FennellEwaldJcp06.pdf (791 KB)

4

Sorry, I forgot to attach the papers. I have enclosed them to this email.

The group based cutoffs means that the entire molecule (in small organic
compounds) or some groups (with the net charge of zero) are placed inside
the cutoff. This is the basis for the group based cutoff technique in which
electrostatic forces decay faster.

so you are talking about "charge groups". they have no consequence for
ewald summation (or PPPM for that matter), since the real space
coulomb is damped smoothly to (almost) zero.

If you look into the attached papers (alternative pairwise for Ewald sum),
they basically, assumed that the electrostatic interaction is short ranged
in condensed systems and there is no Fourier space (long range). Therefore,

i know those papers, however, their assumption is only true for
homogeneous isotropic systems, which makes it useless for
most "interesting" applications.

if you have a surface, or interface that forms a surface dipole
or some higher order multipole, then those methods are wrong.

the neutralization of the charge is important inside the cutoff radius. I
think since PPPM has the long range part would not require this criteria. I
don't know how they implemented this neutralization, did they place the
entire molecule inside the cutoff radius or they use another approach. My

you have the papers, you can figure it out.

understanding is that they basically apply it through force and potential
equations. Group based cutoff (placing the entire molecule inside the cutoff
radius) is another discussion. Any comments would be appreciated.

implementing charge groups might be possible by using the "molecule"
tag in LAMMPS or you will have to write your own atomstyle that
contains the information you need. you then have to make sure that
the "skin" distance for the calculation of the neighborlists is large enough
to include you whole charge group.

In accelrys Material Studio software there are 3 options for the
electrostatic interactions, namely: 1- Ewald 2- Atom based 3- Group based

I don't know what the difference b/w Ewald and atom based is. B/c Ewald is

i assume, just truncation at cutoff. i don't care about materials studio.
doesn't it have a documentation?

essentially atom based and I don't know why they have separated it from each
other.

axel.

5

If you look into the attached papers (alternative pairwise for Ewald sum),
they basically, assumed that the electrostatic interaction is short ranged
in condensed systems and there is no Fourier space (long range). Therefore,

i know those papers, however, their assumption is only true for
homogeneous isotropic systems, which makes it useless for
most "interesting" applications.

if you have a surface, or interface that forms a surface dipole
or some higher order multipole, then those methods are wrong.

For instance, some obviously unphysical results have been reported recently:

http://www.nature.com/nnano/journal/v2/n11/full/nnano.2007.320.html

and charge groups have been later identified as the origin of this
numerical artifact...

http://www.nature.com/nnano/journal/v5/n8/full/nnano.2010.152.html

So as Axel wrote this method is limited to very boring situations and
can lead to strong artifacts if used in inhomogeneous or anisotropic
systems.

Best,
Laurent

6

I think group-based cutoffs would be tricky to
do in LAMMPS. The simple example to consider
is just wanting a Coulomb cutoff that is not based
on distance between 2 atoms but distance between
2 molecules (so that all atom pairs in the 2 molecules
are either included or excluded from the interaction
list). The problem in parallel is that one processor
may not know about all atoms in the molecule, so
it has no way to easily make that decision. I don't
think most parallel codes use group-based cutoffs.
I also don't think it matters much for PPPM or Ewald
since they are effectively long-range and don't
have a cutoff. My recollection is that group-based cutoffs
were more popular years ago.

Steve