'Non-numeric pressure - simulation unstable' error during z-direction decompression using NPT under wall conditions

Hi all,

I was doing the equilibration on my system in the z-direction. Two walls were placed at both z faces to prevent atoms from crossing the z boundary. Spring was added to certain function groups to prevent them from moving too far away. Tip4p water was used. The simulation worked well during the compression from 100 bar to 50000 bar. It also worked well during the decompression to 500 bar. But when it finally tried to run at 1 bar condition, the error occurred. From the log file, I noticed there was a sudden drop in density from 0.7 to 0.3 and a drastic increase in total energy from -50000 to 2000. I understand this is why the simulation collapsed.

I did notice something unusual about this long, orthogonal system compared to a simple cubic mixing model without wall and spring conditions. The density and total energy were not stable in some earlier timesteps although there was no error. I would like to know if I misused any command or input any unphysical parameters that led to the collapse. My current understanding is my equilibration protocol may need to be modified to accommodate the long z-direction equilibration (Please correct me if I misunderstand). I think the force field coefficients are good since they work well in another simple system and produce a sensible equilibrated structure.

I attached the lammps input script and the last two snapshots (water is colored red) before the error for reference. Thank you!

Last snapshot

Second last snapshot

equi.in (4.8 KB)