Long range coulombic cutoff

Hi dear all,
I have 3 questions about long range coulombic interaction cutoff:

I want to define the long-range coulombic interaction cutoff radius for my simulation, how can I achive the best value?

Also I have particles with different sizes in my simulation (some particles are 4 times larger than others), so is there different optimized cutoff values for interactions between these particles (large-large/small and small-small)?

I use pppm K-space, How can I different cutoffs for long-range coulombic interaction?

Thanks alot for your attention
Ly

Hi dear all,
I have 3 questions about long range coulombic interaction cutoff:

I want to define the long-range coulombic interaction cutoff radius for my
simulation, how can I achive the best value?

through experimentation. usually, there is an optimal
efficiency depending on the density of charged particles.
in my personal experience, the overall best performance
is often achieved, when the kspace part of the calculation
is taking at most about the same time and the pair part.

Also I have particles with different sizes in my simulation (some particles
are 4 times larger than others), so is there different optimized cutoff
values for interactions between these particles (large-large/small and
small-small)?

not really. if you use a kspace module, you always consider
the full coulomb interaction. what changes with the cutoff is
simply how much of it is computed in the real space and
how much is computed in reciprocal (=k) space.

I use pppm K-space, How can I different cutoffs for long-range coulombic
interaction?

you cannot and you must not. it makes no sense.

axel.

Also I have particles with different sizes in my simulation (some particles
are 4 times larger than others), so is there different optimized cutoff
values for interactions between these particles (large-large/small and
small-small)?

not really. if you use a kspace module, you always consider
the full coulomb interaction. what changes with the cutoff is
simply how much of it is computed in the real space and
how much is computed in reciprocal (=k) space.

If you mean different cutoffs for the non-Coulombic interactions,
then yes you can do that, with the pair_style command. And
you can sometimes increase the neighbor list build efficiency
with the neighbor multi option.

Steve