NPT Ensemble: Temperature not decreasing

Dear Dr. Axel Kohlmeyer,

Hello Dr. Kolhmeyer. Once again, I deeply appreciate you many advice on LAMMPS.

I currently have a system of conjugated polymer consisting 300 chains (each consisting of 150 beads).

I am trying to simulate a quenching process where I would decrease the system temperature from 550 to 100 K using the Nose-Hoover thermostat via NPT ensemble: fix 1 all npt temp 550 100 1000 iso 1.0 1.0 2000

However, looking at the output file, the temperature is not decreasing from 550 to 100 K.

It is, in fact, increasing, and, then somewhat decreasing until 400 K where the MD run terminated with the error “Bond atoms missing on proc at step.”

Before the quenching process, I equilibrated the system where I relaxed the box size using Langevin dynamics (include solvent effect) and Nose-Hoover barostat using NPH ensemble.

The potential energy of the system (before quenching) had stabilized and reached its asymptote. (The dihedral energy showed an increasing trend at the end.) Therefore, the system should be relaxed and equilibrated when quenching begins. I do not think it is a matter of bad force field parameters or the structure of the polymer system since the potential energy of the system had stabilized and the MD run using fix Langevin and fix NPH finished without any error.

Why would the temperature not decrease using the NPT ensemble during the quenching process? I would greatly appreciate if I could request for any advice on where I should look.

Last note: When I created a small system using 10 chains with 10 beads, the NPT ensemble has successfully quenched to 100 K. When the system becomes 300 chains, then temperature did not decrease. In case, I attached my short script which works for small systems but not very much for large systems of 300 chains.

Thank you so much for your time.


Masato Koizumi


P2_P3_Bond_Table.table (69.7 KB) (74 KB) (1.63 KB)

This sounds like a case of bad dynamics. Your time step size might be too large or your initial geometry might be bad.