Pressure spikes/jumps with rigid/npt and shake

Hello,

I’m working on relaxing a system of rigid bodies using npt/rigid and I see sudden spikes/jumps in pressure, particularly Pxy. The system I have is basically a dislocation dipole in the y plane with x as the glide direction.

The integrators that I’m using are:

fix 1 rest rigid/npt molecule temp 0.1 0.1 100.0 aniso 1000.0 100.0 100000 pchain 20 dilate all
fix 1_s all shake 0.0001 20 10 b 2
fix 1_r rings nvt temp 0.1 0.1 100.0

I was wondering if I could do anything to reduce these spikes, because these spikes are not crashing my simulation but they’re hurting my averages.
I’ve plotted Pxy vs time in the figure below.

image.png

Sincerely,
Anirban

Hello,

I’m working on relaxing a system of rigid bodies using npt/rigid and I see sudden spikes/jumps in pressure, particularly Pxy. The system I have is basically a dislocation dipole in the y plane with x as the glide direction.

The integrators that I’m using are:

fix 1 rest rigid/npt molecule temp 0.1 0.1 100.0 aniso 1000.0 100.0 100000 pchain 20 dilate all
fix 1_s all shake 0.0001 20 10 b 2
fix 1_r rings nvt temp 0.1 0.1 100.0

I was wondering if I could do anything to reduce these spikes, because these spikes are not crashing my simulation but they’re hurting my averages.
I’ve plotted Pxy vs time in the figure below.

Impossible to say from so little information and without the ability to reproduce it. Are you sure your timestep is short enough for the rigid integrator? The rotational DOFs can be very unstable for certain conditions.

Axel

Yes, I’m aware of the small amount of information I’m providing, as I believe the problem is very specific
to the system I’m studying.

I’m using a timestep of 0.1 fs now and I see the spikes. If I use 0.01 fs, the spikes do go away. W.r.t. to the stability of the rotational DOFs, is it more of a high rotational frequency issue or some other integrator problem? Is there a direct way to estimate the rotational

frequency? Like as one does for stiff bonds?

Anirban

Yes, I’m aware of the small amount of information I’m providing, as I believe the problem is very specific
to the system I’m studying.

I’m using a timestep of 0.1 fs now and I see the spikes. If I use 0.01 fs, the spikes do go away. W.r.t. to the stability of the rotational DOFs, is it more of a high rotational frequency issue or some other integrator problem? Is there a direct way to estimate the rotational
frequency? Like as one does for stiff bonds?

Impossible to say in this generality.