[lammps-users] Equilibration


I have got a little problem that seems quite simple at the first glance but I just do not manage to sort it out.

The starting point is a small system (2100 atoms) , representing a LJ-fluid with a number density of 0.1 and temperature of 2, that is confined between two reflective walls in the x-direction (x1=0, x2=200). The other dimensions are periodic (Ly=Lz=10).
After initialisation I equilibrate the system and expect the total velocity (sum over all atoms) to be close to zero, apart from expected fluctuations.
However, what I get is a periodic oscillation of the total velocity in x-direction (www.cranfield.ac.uk/~ei2989/lammps/vx/vx_total_over_tau.jpg). The frequency is about 1/(42000 time steps).

I assumed that is a kind of sound wave which is formed because of the initial conditions. However, I am not sure about this.
Also a spectral analysis of vx measured for a cell (10x10x10) in the middle of the domain, gives a main frequency about twice as high as for the total domain.

I have tried to remove these oscillations, by
- normal standard rescaling
- normal standard rescaling + keeping the total vx = 0
- dividing the domain into blocks and rescale these smaller blocks seperately, including setting vx=0 for each block
- use a different orientation for creating the atomic positions
- initialise the positions randomly instead of using the crystal structure of the 'create_atoms' command
- repeatedly reinitialise the velocities with the 'velocity' command
- equilibate with a higher temperature
- use damped walls instead of perfect reflecting ones
- put some additional walls in between that only reflect a certain percentage of atoms (this should act as a flilter, and disperse any wave)

However, none of these worked and my simulation did always end up with almost the same oscillations. For all simulation the equilibration was over 1,000,000 time steps and another 1,000,000 time steps to check the behaviour in time.
Now, I almost think that this may be a natural behaviour, but I can not imagine that expected fluctuations should be so periodic?

I would be very grateful, if anybody can give me a hint, what I am doing wrong, or can explain why it is as it is.

Best regards,
Marco Kalweit