Hi,
i try to simulate a C-Si indentation process. I alreaded experimented
a little bit with it. I still have a problem with the distances/units.
First I created a substrate region
at the beginning the sphere and the box will shrink a little bit so
that the distance between the sphere and the substrate will be a
little bit bigger than 5 distance units, but with an amplitude of 20
distance units, the sphere should touch the substrate, anyway, but
there's still space between.
Is there something wrong with the move wiggle fix, or are box units
different units? It's not documented in manual.
Regards,
Jörg
Hi,
i try to simulate a C-Si indentation process. I alreaded experimented
a little bit with it. I still have a problem with the distances/units.
First I created a substrate region
at the beginning the sphere and the box will shrink a little bit so
that the distance between the sphere and the substrate will be a
little bit bigger than 5 distance units, but with an amplitude of 20
distance units, the sphere should touch the substrate, anyway, but
there's still space between.
Is there something wrong with the move wiggle fix, or are box units
different units? It's not documented in manual.
Most likely there is some kind of misunderstanding from your side, but
it is difficult to say anything definite with so little information.
Can you post your entire input?
fix move does use them and use them correctly.
it is also in the documentation. if you use units box,
the units for the parameters are angstrom (since you
have units real) and if you don't specify it or stick with
the default it will be in lattice spacings using the latter
of the two lattices that you define. the same units lattice/box
conventions also apply to, for example, the region commands.
a more serious issue (and there should be a big fat warning)
is the fact that you also time integrate the atoms that are
moved by fix move. you must not do this. this is likely
the reason for your unexpected trajectory.
also, whenever you define a box significantly larger than the
circumference of the atoms in the system you probably want
to use "m" boundaries instead of "s" boundaries, or else you'll
get into trouble with parallel runs (lost atoms in the first step).
final comment. for sparse systems, using a combination of
MPI parallelization and multi-threading (via multi-threaded
styles from the USER-OMP package) might give you a better
parallel efficiency. typically you would use MPI for as long as
you can partition the system to have an equal amount of atoms
(probably 4 tasks in your case) and then use threads to parallelize
over atoms instead of domains.
Thanks,
this was one point which i corrected. but comparing the
dimensions of the box with the number of created atoms, i see that
lengths of the region command are measured in lattice constants and
_not_ angstroms, as i tested the wiggle command it, it looks like it
uses - as it should - angstroms. So it's not the wiggle command but the
region command, which uses obviously the wrong units in the metal
environment.
regards,
jörg
Both the region command and fix wiggle command
have a "units" option which can be set to lattice or box,
as do many LAMMPS commands. In all cases,
the default is "lattice". "Box" means the native
distance units, Angstroms in the case of metal units.
"Lattice" means lattice spacings.