I built a solution model with Materials Studio, and I am exporting the data file to be read by lammps. Running my code I am getting the Error:
“Non-numeric atom coords - simulation unstable”
Reading the answers on MATSCI for the same error I know that this error can be caused by atoms being too close together and the forces too large. For this, I did following try:
I first added the command “delete_atoms overlap 0.1 all all” before minimization. The results show that no atoms have been deleted, but I am still getting the same error.
Then I put this command after minimization, and 1 atom was deleted. When I run the simulation I got a “Bond atom missing in image check” error.
Later I changed this command to “delete_atoms overlap 0.1 all all mol yes” to delete the whole of the overlapped molecule. And I am still getting the error: “Non-numeric atom coords - simulation unstable”.
In addition, the potential function parameters are from the literature and have been cited in the same system.
I’ve also checked the units, and they seem correct. Units ‘real’ was used.
Would you please tell me what could have gone wrong for this simulation?
In units real a distance of 0.1 is 0.1 angstroms – an entirely unrealistic distance between physical atoms.
The fact that your energy minimisation leaves some atoms that close to each other indicates that, most likely, you’ve made some foundational mistake in how you specified your force field.
That’s all I would guess based on the limited information supplied.
You possibly have some very high potential energy. While you eliminated direct overlaps, that does not mean that your system is in equilibrium and may still have local configurations with very high potential energy.
Since you don’t provide any relevant details about your simulation, your guesses are as good as anybody else’s.
Thanks for the advice and your time！My LAMMPS version is LAMMPS (31 Aug 2021).
I checked the logs, and I found that the system has very high potential energy as you said. The temperature is also very high.
Based on your suggestion, I changed the initial model. I added half the bond length to each of the six boundaries to avoid overlaps. I kept the “delete_atom” command after minimization to see if any atoms were deleted. I changed the distance to 1 angstrom. I did following try and I am getting some error:
I only minimized the energy and didn’t run NPT. No errors were reported, and the temperature was normal.
I used the command “delete_atoms overlap 1 all all mol yes” after minimization and ran NPT. No atoms were deleted, but the temperature and energy became very high. I got ERROR on proc 21: Bond atoms 2497 2499 missing on proc 21 at step 9.
Reading the answers on MATSCI, I tried to run 100 steps with fix shake, and then turn off fix shake and instead enable fix NPT. Still no atoms were deleted, and I got same error as 2.
This is very old. You should upgrade, even though it will not address your immediate problem, it has other bugs fixed and better error messages.
Minimization does not consider or change velocities, so the temperature is irrelevant. What matters, however, is your extremely high pressure. That indicates an initial geometry far away from equilibrium and thus a system that “wants” to ‘explode’.
No surprise here. While you have no more overlapping atoms, you still have an unreasonable geometry. You may try running fix box/relax alongside the minimization to have the system expand a little bit.
Also, if you do have a “bad” geometry causing high pressure, running fix npt is the worst you can do. You should first run with fix nvt to see, if the system can relax without changing the box and only switch to fix npt, if the pressure has reached a minimum. Even then, it might not be sufficiently low. In that case, the problem is likely your initial geometry being too unrealistic and too far away from a reasonable equilibrated geometry.
You model either requires fix shake or does not. It is not something you can add or remove at will.
Based on your suggestion, I changed the model boundaries in MS instead of in the data file. I kept the “delete_atom” command after minimization and no atoms were deleted.
For water molecules, I’ve been using the SPC/E model in the previously attached in.file. Your suggestion about fix shake inspired me. I changed my forcefield to a flexible TIP3P model, which does not require fix shake, and it solved the problem.
