I am doing ab initio molecular dynamics in VASP. As far as I know, pseudopotentials approximate all-electron potential far from the nucleus, and deviate from it near the nucleus.
When some of my atoms are assigned large velocities, and when I take a very small time step, I noticed that atoms can approach very close to each other. Can I trust my calculations in this case? Does using pseudopotentials mean that we can use AIMD based on them only when atoms are not too close to each other? How to find the closest distance at which we still can use pseudopotential-based AIMD?