Bond atoms missing in non-periodic simulation

Hi All,

I have a simulation of water + alcohol in a channel which is non-periodic in z. I have attached a VMD image of the system just before the crash. When I try to use a higher timestep of 0.5 fs, I get “Bond atoms missing error”. I don’t get this at lower timesteps. The system properties also seem stable to me (I have attached a snippet of the log file).

I searched old posts as this appears to be a common problem and saw fix wall/reflect, but I don’t want to use that since I’m already using wall atoms to bound the system in z-direction. I am out of ideas and need suggestions. I don’t have issues with this system in periodic conditions, just this non-periodic.

input-2.lammps (839 Bytes)

Hello,
I see nothing weird with the input. I suspect the problem comes from the parameters (pair coefficient, bond energy, etc.).
Simon

1 Like

This is not unexpected. You are only using fix shake on water, but you should probably also use shake on other bonds involving hydrogen atoms, specifically OH groups. Or stick to a shorter timestep.

Thank you Simon,
I’ve used the parameters (ethanol and TIp3P water) in previous simulations before. The simulations were stable and gave reasonable results that tally with published works. The only thing new is the carbon walls and the parameters are okay. But I will check again.

Thank you Axel,

I will try to constrain the OH groups as well and provide feedback.

You seem to have a water molecule slip past the wall at the top left of your screenshot. If the molecule has really gone past the wall (and it’s not just an illusion of perspective) that will definitely crash your simulation.

Yes, you are right. It’s an ethanol molecule. It’s not an illusion, its position is actually beyond the wall. Is this because of how the “kspace_modify slab 3.0” is being implemented? I tried to hold the walls in place by fix spring/self to prevent leakage from the wall, and is there a way to do this more properly? Thanks.

Yes, you are right. It’s an ethanol molecule. It’s not an illusion, its position is actually beyond the wall. Is this because of how the “kspace_modify slab 3.0” is being implemented? I tried to hold the walls in place by fix spring/self to prevent leakage from the wall, and is there a way to do this more properly? Thanks.

Well it seems that the ethanol molecule was able to slip in between your wall particles and thus out past the wall. The problem is more basic than a kspace issue, which is that since LAMMPS expects the z-axis to be non-periodic you cannot have any particle go past your stated z-axis limits.

Since you know the physics you are trying to simulate, you will have to decide how to make the wall properly impermeable – whether through adding more layers, spacing the wall particles closer together, having a tighter spring/self spring constant, or adding a wall/reflect that doesn’t affect dynamics outside the wall and only starts reflecting particles that are starting to enter the wall.

1 Like

Thank you for your response! I appreciate it.

Just feedback for future reference:
The main cause of the crash was the molecule going beyond the wall as srtee pointed out. I found that the wall did not completely span the box in the x and y periodic dimension. I fixed this, used a smaller lattice spacing, and increased the force constant slightly to still be consistent with what I want to do. Everything is okay now and I can run at 1 fs for however long I need. Thank you all.

1 Like