Help with ERROR: Lost atoms in LAMMPS?

Hi there! I’m a physics engineering student student working on a materials science project, and I’m currently running a simulation using LAMMPS. However, I’ve encountered an issue with the error message “ERROR: Lost atoms: original 182322 current 181371 (…/thermo.cpp:438) Last command: run 10000” when running the script. My goal is to simulate the collision of a Cu nanoparticle with a Ni surface under normal conditions. I have attached the input file for reference.

The problem seems to be related to the “velocity nickel create 100.0 12345 dist gaussian” command line. When I include this line in the script, the program gives me the error. However, if I remove that line, the code works perfectly, but I lost information.

I want to resolve this issue to make the simulation more realistic.

Could you please help me with this? I’m eager to improve the accuracy of the simulation. Thank you! (5.8 MB)
FeCuNi.eam.alloy (1.2 MB)

in.eam (2.0 KB)

Ive forgot to mention, the first image is the first dump file image, and the second is the last one, then the program drops me the error

No, its unlikely to be the issue here.

What is a perfectly running code that loose information? What is information here?

Most likely its a dynamics problem. I don’t know what you want to do but your code is strange looking, you are using a timestep of 8 fs, I see no thermostat.

My advise, you should follow some tutorials on LAMMPS before attempting such simulation.


Why use fixed boundaries? Why not have a periodic system in x- and z-direction? That way you don’t have to mess around with immobilizing lots of atoms. That will taint your results much more than using PBC.

Also, setting the forces of a particle to zero is only immobilizing it, if also its velocity is zero, otherwise it will just keep its velocity (remember Newton’s laws of motion from classical mechanics?) and you do assign (random) velocities to all Ni atoms. This is not what you want and in combination with the fixed boundaries that will lead to atoms leaving the box.

One more tip. If you are running in parallel, you also want to use processors * 1 * to avoid having subdomains with few or no atoms since LAMMPS creates subdomains by volume.