Re: About linear momentum conservation

From: Tilmann Piffl <til_at_aip.de>
Date: Tue, 21 Apr 2009 18:00:19 +0200

Hello!

I'm experiencing a similar problem as Kim Do Gyun. I'm using Gadget-2
with vacuum boundary conditions and pure n-body dynamics.
A simple plummer sphere in isolation moves considerable distances (up to
100 kpc) in 10 Gyr depending on the softening and the ErrForceTolAcc
parameter. The initial net velocity was set to zero.
It is clear, that the linear momentum isn't conserved, but the force
errors should be random and add up to zero over time. This is apparently
not the case in my simulations. The sphere is accelerated in one
direction for more than one time step and also moves on smoothly curved
trajectories.
It seems like there is some self-energizing effect driving the force
error in a preferred direction.
Also the standard example , the merging galaxies, start shifting away
leaving about 2 kpc in 3 Gyr. Is this a usual behavior?
Does anybody know, how to cope with this? I heard from other people not
having this problem. What might I be doing wrong?

Til
Received on 2009-04-21 18:00:19