Hi,
Thank you for answers.
first off, please always send a copy of your replies to the mailing list.
thanks.
Timestep is important because as you know the temperature is conducting via
phonon transfer and the vibration of the atoms should be known by very small
delta(t).
this statement is "half correct" in two ways.
a) you _always_ have to choose a suitable time step to integrate
the fastest motions in your system accurately. as such that has
nothing to do with thermal conductivity.
b) thermal conductivity through is only one component. for any compound
with delocalized electrons, e.g. metals, a significant - if not the dominant
- part of the thermal conductivity is facilitated via electrons.
since graphite
layers are effectively 2-dimensional metals, you cannot ignore this.
I would like to choose an accurate timestep in order to not make round-off
error.
you have to do an accurate MD and consider that you have good sampling.
these are just standard protocol for _any_ MD simulation (and common
sense, too).
i won't worry about rounding errors, but rather statistical errors and
systematic errors from using the wrong model or making the wrong
assumptions about your system.
I saw many Lammps scripts which are available on the user page and all has
different timestep.. why?
why not? they are different systems, use different units.
you also have some variety in your choice of time step.
you can go with a shorter time step and have more reliable
dynamics, or go with a more aggressive time step and
get better statistics. it depends a lot on the analysis of
the data which is the better approach.
axel.
p.s.: don't find it irritating when people post e-mails under
one name in their e-mail address and sign them with a different
one in the body of the e-mail?