LJ cuttoff

Dear all,
i’m using AIREBO potential for investigating the mechanical properties of a carbon nanotube, but it is too slow, i think it’s because of its LJ cuttoff that i used and it was 3, as mentioned in manual “In the standard AIREBO potential, sigma_CC = 3.4 Angstroms, so with a scale factor of 3.0 (the argument in pair_style), the resulting E_LJ cutoff would be 10.2 Angstroms.” can i use lower values instead of 3 ?

Yes, if you have verified that a smaller cutoff still provides you
with reasonable results.

Ray

Dear all,
i'm using AIREBO potential for investigating the mechanical properties of a
carbon nanotube, but it is too slow, i think it's because of its LJ cuttoff
that i used and it was 3, as mentioned in manual "In the standard AIREBO
potential, sigma_CC = 3.4 Angstroms, so with a scale factor of 3.0 (the
argument in pair_style), the resulting E_LJ cutoff would be 10.2 Angstroms."
can i use lower values instead of 3 ?

this kind of reasoning is almost always a bad idea.

first off, you you have proof that reducing the LJ cutoff will give
you that significant a speedup, that it is worth the loss of accuracy?
why not switch to a simpler model (tersoff, stillinger-weber, or plain
charmm or amber)? those will be even faster.

have you already exhausted all options to run more efficiently? why
not use a smaller system?

or why not running across more CPUs. LAMMPS is well parallelized and
nowadays it is rather easy to get access to parallel machines or a
cluster. many clusters in many places are underutilized all it takes
is some effort.

one final remark: what does it help you, if somebody here say "yes"?
what would you respond, if you submit your paper and the referee says
"results with a modified cutoff cannot be trusted"? if you'd respond
with "some person on the mailing list said it is ok", you are likely
to be laughed at. the only way you can be sure is, if you rigorously
test for yourself. *even* if you use the standard cutoff. *any*
simulation, *any* model needs to be validated to give reasonable
results for the kind of calculation of the properties you are
interested in. and that doesn't happen by acclamation. it is your
work, so it is your responsibility to prove that it is correct.

axel