Experiencing NaN Values and Temperature Instability during NVT Simulation

[eshaanraval]

Note: I'm relatively new to molecular dynamics (MD) simulations and might have some gaps in my understanding of the concepts and implementation.

I’m currently trying to recreate a simulation based on Florain Müller-Plathe's paper titled "A simple non-equilibrium molecular dynamics method for calculating the thermal conductivity." Unfortunately, I am encountering NaN values during the integration step, specifically from step 0 to around step 10. Furthermore, the system's temperature appears to fluctuate dramatically.

Based on my observations from the temperature and energy plots, my hypothesis is that the instability might stem from a few atoms being placed extremely close together due to the use of rng.uniform, which could result in exceedingly high potential energy.

I attempted to use optimizers to relax the system, thereby reducing its potential energy and forces, but the nan issue persists. The temperature continues to randomly spike well beyond logical variance.

Is there any advice or guidance on what might be going wrong in my approach? What steps could I take to resolve these NaN values and stabilize the system's temperature?

NVT Simulation Code: simulation-code.py
Sample Output: output.txt
Other Trials Output: Equilibrium-Phase-Testing.odt
Simulation Paper: NEMD Simulation

[eshaanraval]

Following Thomas Loux's Suggestion

You shouldn't use nvt_nose_hoover_init AND ts.integrate. ts.integrate does it already.
If you use LJ models, use LJ units but not MetalUnits.

I was able to solve this issue.