MSD calculation with ReaxFF potential

I have been trying to calculate Mean Squared Displacement for a liquid like carbon structure. To do this I have first equilibrated the structure at desired temperature for 5 ps and saved a restart file.

#equilibration run
fix 2 all temp/csvr 3500 3500 100 543248 # 0.1 ps relaxation
fix 3 all nve
run 10000 # 5 ps

Then I have carried out 5 ps data gathering run in NVE ensemble.

compute msd all msd com yes
#data gathering run
fix 3 all nve
run 10000 # 5 ps

To illustrate the problem. I am mentioning first and last three lines from the log.lammps file for the data gathering run:

Step Time Temp c_msd[4]
0 0 3645.3458
250 125 3703.4564
500 250 3773.2235

9500 4750 7299.2796
9750 4875 7520.4365
10000 5000 7739.1196

The final temperature appears to be almost double the initial temperature! Should this happen after equilibrating the temperature at 3500K?

Previously I have carried out similar simulations for AIREBO potential and observed small deviation between final and initial temperature (5-12%). The input scripts for simulation with ReaxFF are attached here. (1.4 KB) (865 Bytes)
log.lammps (5.9 KB)

It depends (as do so many things in research!).
The most common scenario, where you have this behavior, is when either your system is not yet fully equilibrated (5 ps for equilibration is rather short unless your starting geometry was already mostly equilibrated) or when your timestep is too large (ReaxFF usually requires shorter timesteps than other commonly used forces fields).

If either (or both) apply can be easily tested by running a longer equilibration or with a shorter time step or both.

1 Like