Hi,
As far as I’m concern, there are 3 ways to create atoms, i.e. read_data, create_atom and read_restart. From read_data command, LAMMPS will create a simulation box, not unit cell, based on the input values, and an appended replicate command will replicate the simulation box to a larger simulation box. However, from create_atom command, a unit cell will be created with lattice command and a simulation box will then be created with region command. Are the above statements true? If atoms in a system are created by read_data command and atoms in another system are created by create_atom commands, will the simulation results for both systems differ? In other words, will the results be different if the atoms were created in a unit cell (using create_atom + lattice + region command) or simulation box (using read_data + replicate command)?
Thank you.
Christopher
Hi,
As far as I'm concern, there are 3 ways to create atoms, i.e. read_data,
create_atom and read_restart. From read_data command, LAMMPS will create a
simulation box, not unit cell, based on the input values, and an appended
replicate command will replicate the simulation box to a larger simulation
box. However, from create_atom command, a unit cell will be created with
lattice command and a simulation box will then be created with region
command. Are the above statements true?
no. the second is not. the simulation cell (in MD slang often referred to
as unit cell, btw) will be created by the create_box command and while
that takes a region as input, a lattice command is not essential. it only
so happens that the region definition by default assumes using lattice units.
if you create the region with "units box" then it will use the standard length
units and not lattice command would be required. check the documentation
on this for details.
If atoms in a system are created by
read_data command and atoms in another system are created by create_atom
commands, will the simulation results for both systems differ? In other
words, will the results be different if the atoms were created in a unit
cell (using create_atom + lattice + region command) or simulation box (using
read_data + replicate command)?
they should not differ. however, reading coordinates from a file, those
numbers are subject to truncation and rounding errors and thus may
result in small differences. since MD simulations are chaotic (your
equations of motion are couple partial ordinary differential equations),
trajectories will diverge exponentially. that divergence, however, has no
consequence to the statistical relevance of the sampled conformations.
i.e. the numbers may be different, but they are equally relevant.
this topic has been discussed on this list a few times; you may want
to search the mail list archives for more details. also, having a look
at some text book(s) on statistical mechanics might provide additional
insight as well.
cheers,
axel.