fix shake

Lammps community,

I am running a simulation that uses fix shake to constrain bonds and angles of extended spheres of system in 2D, using the atom_style hybrid colloid angle. I am curious if there is a known discontinuity in fix shake when the bond angle is 180 deg. Doing so triggers fix shake to become unstable and it leads to an atom moving too far:
ERROR on proc 6: Shake atoms 15973 15974 15975 missing on proc 6 at step 10
(In 3D the same thing happens but with a shake determinate of 0 error.)

I can also make this error happen when only constraining only the bond distance, but with the initial configuration of the bonds in a 180 deg arrangement (not surprisingly). To get around this (for both cases) I just offset the angle by a degree or two and the constraint is satisfied to satisfaction. Using fix rigid could be a solution, but with a downside that for each angle constraint a new body must be created. I am just curious why this happens. I will attach my script for reference. I am running lammps version 9Feb11.

Thanks,

Kevin

in_pac.txt (1.43 KB)

I am curious if there is a known discontinuity in fix shake when the
bond angle is 180 deg.

It's been a while since I looked at this, but I think there is an issue
with 180 deg. I assume it is a numerical issue of some sort
in the SHAKE constraint eqs, but it needs to be looked into to
see if there is a simple fix.

Steve

I am curious if there is a known discontinuity in fix shake when the
bond angle is 180 deg.

It's been a while since I looked at this, but I think there is an issue
with 180 deg. I assume it is a numerical issue of some sort
in the SHAKE constraint eqs, but it needs to be looked into to
see if there is a simple fix.

the 180 degree angle constraint is divergent.

i think a way to handle this would be by
turning it into a 1-3 bond constraint with the
length being the sum of the two 1-2 bonds.

cheers,
    axel.