Rigid body effect, translational kinetic momentum

From: Joshua D. Moore (jdmoore_at_unity.ncsu.edu)
Date: Fri Apr 04 2008 - 19:40:35 CDT

Dear NAMD list,

I am experiencing some difficulties running NVE (microcanonical) simulations
of argon confined inside a carbon nanotube with NAMD. To start, I ran NVT
with the thermostat, and the molecules appeared to be diffusing correctly.
However, when I switched off the thermostat and ran NVE, I saw that the
particles inside the nanotube were all translating in the same direction. I
did not see this in the NVT simulations. The system is very confined, with
the argon molecules only diffusing single file (i.e., the nanotube diameter
is small).

Another issue might be the fact that I have fixed the atoms of the nanotube
using "fixedAtoms on" and the appropriate column in the *pdb file. It seems
in this case that it might be appropriate to zero the kinetic momentum (for
translation and rotation in general but only translation in this case since
Argon is spherical), and I have seen this being applied in the TINKER
package for rigid body simulations.

So the question is does NAMD zero the kinetic momentum of translation and
rotation when it initializes the velocities from a Maxwell-Boltzmann
distribution? It seems that NAMD does account for the Center of Mass to
make sure it isn't moving. If NAMD isn't zeroing the kinetic momentum of
translation, then I could try to do this manually before starting the
simulations. If this is true, then why would the simulation behave
seemingly correct with the thermostat?

I prepared some examples to show the effect and link is below. I have tried
really everything I can think of. I ran NVT simulations and didn't see this
effect. I ran NVE simulations by initializing the velocity at the
temperature I want (of course this doesn't get me the temperature I want),
minimizing the structure and then initializing at the temperature I want,
rescaling the velocities periodically to achieve the temperature, and also
starting from a set of velocities and coordinates from the NVT simulation.
All lead to the translation effect in the NVE simulations with the Center of
Mass moving (i.e. all of the particles move in one direction down the
tube.....they should just vibrate back and forth due to constant collisions
with the other molecules because of the confinement). I still can't figure
out the problem. Perhaps I am fixing the atoms wrong and should approach
the problem differently in NAMD.

By the way, I also tried decreasing the timestep from 1 fs to 0.1 fs and 0.5
fs, increasing the cutoff. All seem to have no effect, and the total energy
of the system is constant. I haven't checked for fluctuations in the
fluid-wall potential energy, this is next.

The link to my example *dcd output and input files are here (in respective
NVE and NVT directories)


Each step in the example *dcd files represents 100,000 fs (so the files are
not that large, about 5 MB each).

Thanks in advance for any help.


Joshua Moore

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:49:21 CST