Modelling a High Frequency/Low Amplitude Vibrating Plate

LAMMPS LAMMPS Mailing List Mirror
1

Is it possible for Lammps to model a very high frequency and low amplitude vibrating plate?

I have set up a simulation using fix wall/gran with the wiggle option. I have the frequency set at 20kHz, period = 5e-5 s. The amplitude I am using is very small though: 1.59e-5 cm.

Using this setting, Lammps shows a sphere (granular package) interacts with the plate as if it were a stationary wall. However increasing the amplitude to around 3e-3 (200 times larger) introduces extra kinetic energy to a bouncing sphere as I would expect.

Is it possible that Lammps cannot resolve this very small resolution?

I thought it could be an issue with the timestep, but decreasing it to 1e-10 did not solve the problem.

-> In my simulation a sphere is impacts the plate at around 2m/s, so it would cover the amplitude distance of the plate (1.59e-5cm) in just under 8e-8 seconds. So the smaller timestep should resolve any impact taking place?

Could it be an issue with very small values that are just too small to be computed by Lammps?

Thanks,

Tim

2

Is it possible for Lammps to model a very high frequency and low amplitude
vibrating plate?

I have set up a simulation using fix wall/gran with the wiggle option. I
have the frequency set at 20kHz, period = 5e-5 s. The amplitude I am using
is very small though: 1.59e-5 cm.

Using this setting, Lammps shows a sphere (granular package) interacts with
the plate as if it were a stationary wall. However increasing the amplitude
to around 3e-3 (200 times larger) introduces extra kinetic energy to a
bouncing sphere as I would expect.

Is it possible that Lammps cannot resolve this very small resolution?

lammps uses double precision floating point math throughout (which
provides about 14 digits of precision), so i doubt that this is the
case. now whether an effect is visible on small amplitudes depends on
what other energy fluctuations you have in your system (depends on a
lot of things). as for high frequencies, this has to be correlated
with the time step. the integrator cannot "see" frequencies that are
significantly faster than the time step you define. think of the
spokes of a fast rotating wheel seemingly turning backwards in a movie
or on tv.

if you have doubts beyond these fundamental considerations, please
provide a simple test input confirming your concerns.

thanks,
    axel.