PPPM method and truncated coulomb

Dear LAMMPS users

I have a question about the difference between lj/cut/coul/long and lj/cut/coul/cut pair_styles.
When I simulate water droplet on graphene surface, extracted results of contact angle is approximately similar (about 5 degree difference). But when in simulate binary fluid the difference is harsh. For water-methanol mixture is about 20 degrees in contact angle in the similar conditions(just the number of molecules are different)
water droplet with 2000 molecules (with PPPM CA is 98 and without PPPM and lj/cut/coul/cut pair_style CA is 94 )
water-methanol with 6500 molecules (witrh PPPM CA is 83, without PPPM and lj/cut/coul/cut pair_style CA is 118 )

All the Coulomb cutoff radius is assigned 40 angstrom.

I will appreciate if anyone could help me to find out the reason of this difference.

Thank you so much

Sincerely

Nikzad

Pair_style lj/cut/coul/long with PPPM includes long-range interactions, while lj/cut/coul/cut without PPPM does not and is a truncated potential.

Stan

Thank you Stan
I know this difference but I simulated all the molecules in the simulation box. I expect that the difference must not be considerable.
If I choose 5nm cut-off radius for coulomb interactions, I will cover all the simulation box. So I think results must be the same as using long range coulomb with PPPM method.

Sincerely
Reza

That’s not entirely correct. 1/r is a slow converging function so even with 40 or 50 angstrom cutoff you are cutting off considerable contributions. Such contributions are recovered with reciprocal space summation as with the PPPM method.

Even with 50 angstroms you still have interactions beyond the cutoff due to periodic boundary condition. Please read some text books on long range summations (Frenkel & Smit, Allen & Tildesley, etc).

Ray

Dear LAMMPS users

I have a question about the difference between lj/cut/coul/long and
lj/cut/coul/cut pair_styles.
When I simulate water droplet on graphene surface, extracted results of
contact angle is approximately similar (about 5 degree difference). But when
in simulate binary fluid the difference is harsh. For water-methanol mixture
is about 20 degrees in contact angle in the similar conditions(just the
number of molecules are different)
water droplet with 2000 molecules (with PPPM CA is 98 and without PPPM and
lj/cut/coul/cut pair_style CA is 94 )
water-methanol with 6500 molecules (witrh PPPM CA is 83, without PPPM and
lj/cut/coul/cut pair_style CA is 118 )

All the Coulomb cutoff radius is assigned 40 angstrom.

I will appreciate if anyone could help me to find out the reason of this
difference.

there is insufficient information here to give any meaningful
recommendation, but let me point out a few issues, that you may be
overlooking:

1) you are not comparing exactly the same system. with cutoff coulomb,
you do not "see" periodic images (provided the box is large enough);
with long-range electrostatics you do, and thus it matters a lot, how
close the periodic images of your droplets are.

2) for long-range electrostatic, the real-space cutoff should be
irrelevant, and it is often very wasteful to choose it to be too
large. however, it does matter a lot, whether your real-space +
k-space is sufficiently converged. the typical default convergence
values are ok for pure bulk systems with lots of error cancellation,
but if you want accurate results, you need to converge well.

3) for coulomb with cutoff, you can get "exact" results, if the cutoff
is sufficiently large (larger than the largest extent of the droplet)
*and* there are no periodic boundaries or the distance between
periodic images is beyond the cutoff. for coulomb with kspace, you
have to look to minimize the impact of the multipoles of the droplet
interacting with its periodic images through the kspace solver.

this is just a list of simple things that pop into my mind, and that
it seems to me you haven't considered. there is probably more.

axel.