[lammps-users] Inconsistent Velocity Distribution

Dear LAMMPS Support Group,

Hi there. I am fairly new to using this package and I have been having somewhat unusual results from LAMMPS.

I have been simulating a one-particle system (see code below) trying to see if it is consistent with results such as the Maxwell Boltzmann distribution for velocities.

However, when I obtain the velocity distribution and compare it to a theoretical Maxwell Boltzmann curve, I am off by a factor of 2 in the (m/kT) term. So either m should be a factor of 2 less or T should be a factor of 2 more. I'm not sure why this is happening.

Can you check my in.file and data.file to see if I have created any bugs that LAMMPS doesn't like?

Thanks for your help,
Alvin

If I understand what you are doing, you can put N particles
in a box and have them be non-interacting (no pair style)
and just use fix langevin to thermally equilibrate them
and fix nve to time integrate them. You'll get quicker
stats that way. And you want to histogram
their velocities to see if they match a Gaussian? The only
relevant operations are what fix langevin is doing which you
can look at - it's just a few lines of code in post_force(). I assume
the mean velocity (temperature) is what you expect, but
you are saying the width of the distribution is off?

The other params I would look at are the timestep and
damping constant for fix langevin. The Langevin factors
depend on both of these. You are using a very large
timestep (10 fmsec) which may mean the integrator
applies the Langevin force for a very long time. Also
the ratio of damping to timestep 18/10 is much smaller
than what I would normally use for fix Langevin - more
like a ratio of 20-100 is normal. Also your initial
velocity should be irrelevant.

Steve