[gmx-users] bd integrator problems

[Johannes Wagner]

Hello all,
trying to set up a bd system here, but observing something I cannot
explain. Perhaps someone is aware of something obvious I'm doing wrong.
If not, I'll provider further information (.mdp file, force field, and
so on).

Simulation system is the following:

- particles in a straight line
- only distance and angle bonded potentials, no dihedrals
- no Coulomb interaction (all charges=0, scheme=group)
- no LJ interactions (for the sake of isolating the problem) by setting
eps = 0, and putting LJ and rlist cuttoff below shortest particle
distance (to be absolutely sure :))
- no solvent
- starting positions of all particles is exactly at the bonded and
angular minimum. so by simulating I expect not much to happen!

Problem is the following:

If I use the bd integrator, my system sort of explodes. By decreasing dt
enough, it then smoothly expands to about 170% of the equilibrium bond
length, and my angle potentials (all 180degrees) which should just keep
the particles in a straight line, drive the system into a zig-zag
conformation and stays there stably.

NOW, if I use the md or sd integrator, the system behaves as it should:
staying at the initial bond and angle values. Any clues what the bd
integrator does differently with the bonded and angluar terms compared
to the md and sd integrator?

Any hints/helps welcome!
cheers, Johannes