Problems with momentum drift when using fix rigid

Dear LAMMPS users,

I am having problems with the fix rigid command.

I have been running initial simulations to test a force field, and have been simulating toluene in a periodic system. I am using fix rigid to hold each molecule rigid, and it seems to be working fine. However, I have analysed the centre of mass data using the variable vcom commands in an ave/time fix, and have found using VMD to visualise the data that the centre of mass is drifting over longer timescales (~1e6 fs to 2e6 fs). As can be seen from my input file 0802134.toluene.in, I have zeroed the velocity after equilibration time, but the total momentum of the system still does not appear to be zero. By running alternative simulations I believe I have identified the fix rigid as the cause of this problem, but was wondering if it is to be expected? Is the drift in momentum part of the fix rigid, or has anybody else had this problem?

I am using the version of lammps from 12th January, 2013.

Thanks,

Sophia

0802134.toluene.in (1.74 KB)

0802134.toluene.dat (394 KB)

Dear LAMMPS users,

I am having problems with the fix rigid command.

I have been running initial simulations to test a force field, and have been
simulating toluene in a periodic system. I am using fix rigid to hold each
molecule rigid, and it seems to be working fine. However, I have analysed
the centre of mass data using the variable vcom commands in an ave/time fix,
and have found using VMD to visualise the data that the centre of mass is
drifting over longer timescales (~1e6 fs to 2e6 fs). As can be seen from my
input file 0802134.toluene.in, I have zeroed the velocity after
equilibration time, but the total momentum of the system still does not
appear to be zero. By running alternative simulations I believe I have
identified the fix rigid as the cause of this problem, but was wondering if
it is to be expected? Is the drift in momentum part of the fix rigid, or
has anybody else had this problem?

does this effect depend on the length of the time step? rigid body
integrators are notoriously difficult be stable on the rotation.

axel.

I don't think fix momentum is a good way to do this.
All it does is reset the velocities of individual atoms.
But for atoms in a rigid body, the body itself stores
the velocity of the center of mass, as well as the angular
momentum. Every timestep, the velocities of individual
atoms are reset according to those 2 body values. So
I think the changes that fix momentum is imposing will be
wiped out every timestep, i.e. it has no effect.

If your initial net momentum for the system is 0.0, and you
are integrating stably (as Axel indicated), and you are
not otherwise imposing external forces on the system, then you should
not get much drift.

Steve