[gmx-users] sd simulations

[Mark Abraham]

Ran Friedman wrote:
> Hi,
> 
> I'm simulating a system of molecules, made of coarse-grained molecules.
> Each molecule has 3 uncharged CG atoms so the interactions are governed
> by LJ, bonds and angles. I use SD for the simulations. My problem is
> that the LJ energy is not converged in the simulations - it just climbs,
> and the particles diverge. I'm not talking about an initial rise in the
> energy of a minimised system. Rather, after >100,000 steps the energy
> continues to rise. If I use very high friction (small TauT) the effect
> is somewhat slower but it's still there.
> 
> Any hints about this?

You probably already knew this, but check the topology carefully! :-) 
You could eliminate SD as a source of problems by trying another integrator.

Mark