Non-numeric positions - simulation unstable (src/nbin.cpp:144)

I Started to study the penetration of carbon atoms into iron oxide shell and iron core by lammps software so that the diameter of the core is equal to 20 angstroms and the inner diameter of the shell is equal to 30 angstroms.So there is an empty space between the shell and the core.

( My simulation box contains an iron-iron oxide cylinder core-shell surrounded by carbon atoms)

but I encountered a problem during simulation. During the simulation, carbon atoms diffuse into the oxide layer and accumulate at the inner border of the oxide layer and do not move towards the iron atoms in the core and suddenly Silmulation stoped by this error :

Non-numeric positions - simulation unstable (src/nbin.cpp:144)

I do not understand the reason for this error.

I think there should be a constraint so that the carbon atoms can pass through the empty space, but I don’t know what command can do this.

Guide me please…

Please see the answer of @akohlmey to this topic:

There are many possible reasons: bad choice of potential, bad geometry, bad choice of simulation parameters, badly designed model. With just a vague description like in your post, nobody can say for certain.

I doubt that this would be a scientifically sound approach. That sounds more like you want to do a computer animation rather than a serious simulation. In general, when a simulation does not do what you expect there are two possible explanations: 1) your expectation is wrong, 2) something in your simulation is wrong or rather your chosen model is not suitable. There is not likely a single command that - when added to the input - will address either of those two explanations.

I first made the core and then the shell separately and merged them together and then randomly placed carbon atoms around the shell.
When I don’t put empty space between the core and the shell, the carbon atoms penetrate into the core but by creating an empty space between the two, penetration does not occur

As far as I can tell without knowing more about the details of your simulation (like potentials and simulation settings and exact geometries), this sounds like you have problems in both of the categories I mentioned.

First off, I don’t understand the geometry of the model. What exactly is it that you want to learn from your simulations? How do you want to extract that information from the simulations? How does this geometry setup correspond to any experiments that you can relate to?

Second, I am doubtful about the potentials you are using. Representing oxides with classical models is non-trivial, but then also a lot depends on the temperature that you keep during the simulation.

Finally, I am unsure whether carbon atoms should significantly diffuse into any of the two systems (unless there is a mistake in the potentials) on the time scales accessible to regular MD simulations. Those kind of diffusion processes are dominated by their activation energies and those would be very large local fluctuation (and thus require long simulations of large enough systems or very high temperatures or both) to happen organically.

I decide to create a potential. Now I think I made a mistake in designing the potential.