I am trying to use class2 (PCFF) force filed parameters.
pair_style lj/class2/coul/cut cutoff (COMPASS with cutoff Coulomb)
My question is how can I calculate or guess value of cutoff, so that
the computational accuracy is not affected. I can make it very large,
but it add a lot to the computational cost.
If you have a forcefield parameterized for use with a cutoff,
then the original papers should state what cutoff was used.
The only caveat against using that cutoff is a cutoff should be
less than half of your box length.
Your other choice is to start with a cutoff of say 12
angstroms, do a few test runs with cutoffs bigger and smaller
than that, and see how small a cutoff you can use before the
statistics are affected.