Hi LAMMPS users,

I just want to make sure I have an exact/correct understanding of the calculation of long range Coulomb interaction in lammps.

About “coul/long”, the doc page says: “The Coulombic cutoff specified for this style means that pairwise interactions within this distance are computed directly; interactions outside that distance are computed in reciprocal space.”

While the textbook’s way (like A.Leach’s Molecular Modeling) handles it by adding Gaussian charge distribution to point charges, so the screened charges and screening charges both converge in real and k space, thus both terms can be cut. And finally the total Coulomb energy is the sum of those two terms plus a self term.

It seems to me the lammps’s way and textbook’s way are slightly different by replacing the real space term + self term with a direct computation within a cutoff, am I right ?

If I am wrong and the two ways are actually equivalent, could someone enlighten me how? as I don’t see it.

If I am right, then why would lammps choose such a way? what’s the trade-off?

Thanks a lot.

Liu

Hi LAMMPS users,

I just want to make sure I have an exact/correct understanding of the

calculation of long range Coulomb interaction in lammps.

About "coul/long", the doc page says: "The Coulombic cutoff specified for

this style means that pairwise interactions within this distance are

computed directly; interactions outside that distance are computed in

reciprocal space."

While the textbook's way (like A.Leach's Molecular Modeling) handles it by

adding Gaussian charge distribution to point charges, so the screened

charges and screening charges both converge in real and k space, thus both

terms can be cut. And finally the total Coulomb energy is the sum of those

two terms plus a self term.

It seems to me the lammps's way and textbook's way are slightly different by

replacing the real space term + self term with a direct computation within a

cutoff, am I right ?

no.

If I am wrong and the two ways are actually equivalent, could someone

enlighten me how? as I don't see it.

they are the same. any "coul/long" style does add the screening

and thus has to be used in combination with a kspace style

to provide the corresponding reciprocal space contributions.

axel.

Thank, so the real-space term cutoff is set by user, but where is the k-space cutoff set?

And, how is the Gaussian distribution width chosen? does lammps use a fixed universal value throughout? or variable?

Thank, so the real-space term cutoff is set by user, but where is the

k-space cutoff set?

And, how is the Gaussian distribution width chosen? does lammps use a fixed

universal value throughout? or variable?

how about reading the documentation? e.g.:

http://lammps.sandia.gov/doc/kspace_style.html

axel.

I see the kspace_style documentation says the accuracy setting determines the # of k-space vectors for ewald.

But I don’t find anywhere say about the width of the screening charge distribution, could you please give more hint or perhaps there’s another doc page that I don’t know related ? Thank you.

I see the kspace_style documentation says the accuracy setting determines

the # of k-space vectors for ewald.

But I don't find anywhere say about the width of the screening charge

distribution, could you please give more hint or perhaps there's another doc

page that I don't know related ? Thank you.

more details are in the referenced papers and the sources.

axel.