[lammps-users] pair_style table with ewald or pppm (kspace_style)

LAMMPS LAMMPS Mailing List Mirror
1

Hi,

I want to model a pairwise interaction having the following equation:

U = A/r^11 - C*q1*q2 exp(-b/r)/r + D exp(-c/r)/r^4 - C/r^6. The second
term (C*q1*q2 exp(-b/r)/r) accounts for the electrostatic interactions. I
am planning to model atoms having this interaction and want to add pppm or
ewald to make the electrostatics truly long range. Since lammps does not
have this kind of pair_style, aside from writing my own pair_style, can I
just tabulate the potential energy and F = -dU/dt and use pair_style table
and apply pppm or ewald?

Jan-Michael

2

Hi,

I want to model a pairwise interaction having the following equation:

U = A/r^11 - C*q1*q2 exp(-b/r)/r + D exp(-c/r)/r^4 - C/r^6. The second
term (C*q1*q2 exp(-b/r)/r) accounts for the electrostatic interactions. I
am planning to model atoms having this interaction and want to add pppm or
ewald to make the electrostatics truly long range. Since lammps does not
have this kind of pair_style, aside from writing my own pair_style, can I
just tabulate the potential energy and F = -dU/dt and use pair_style table
and apply pppm or ewald?

jan-michael,

if you want to make the electrostatics truely long range, you'd
have to _remove_ the electrostatic contribution from the potential
and then you could use pair_style hybrid/overlay with coul/long
and pppm. however, the model seems to be designed and parameterized
for a debye screening and thus i would suspect that you'd need to
re-parameterize it for your true long-range electrostatics.

please note that the point of ewald summation is that you screen
out the long range coulomb interactions in real space and the
short range coulomb in k-space and thus the screening has to have
the exact same shape in both cases (using an erfc-style screening
function equivalent to adding/subtracting a gaussian).

cheers,
   axel.

3

Axel,

Thanks, I thought of that also however reading through the manual I did
not find an equivalent of coul/long for coul/debye so I need to use
coul/long and unfortunately you are right and I need to re-parameterize.

Jan-Michael

4

Axel,

jan-michael,

Thanks, I thought of that also however reading through the manual I did
not find an equivalent of coul/long for coul/debye so I need to use

there is no long-range equivalent of coul/debye, since the debye
model is short range _by construction_. there is no need to add
k-space to it.

cheers,
    axel.