Instability in NPT simulations at High Temperatures for Generating High Energy Structures

LAMMPS LAMMPS Beginners
1

I am studying periodic structures of small organic molecules (viz., amines, etc.). I would like to sample a given system’s high and low energy structures through MD simulations. I am trying to work with NPT with a range of temperatures (viz., 50K, 300K, 500K ) and pressures (1bar/atm, 1000 bars/atm, -1000 bars/atm). However, I observed that the volume blows up (or expands significantly) during the simulation when the pressure is low (viz., 1 bar/atm, -1000 bars/atm) and the temperature is high (viz., 500K). The high-temperature simulations run fine when the pressure is high (viz., 1000 bars/atm). I have been trying to understand the physical reason behind this behavior. Is it because the system is volatilizing at high temperatures? What other methods could I use to sample high-energy structures? Is doing NVT simulations a reasonable choice for such cases?
I understand that what I am observing is not Lammps’ fault; however, I would highly appreciate it if anyone could provide any insight into my problem.
Here is a sample input file I am using. I would be happy to provide more information if needed.

log             log.lammps
units           real
atom_style      charge
atom_modify     map array
boundary p p p
read_data       input.pos
pair_style      reaxff NULL
pair_coeff * * ffield.reax.FC H C N
fix 1 all qeq/reaxff 1 0.0 10.0 1.0e-6 reaxff
thermo 10
thermo_style custom step temp pe etotal fmax fnorm cella cellb cellc cellalpha cellbeta cellgamma press vol
fix 3 all box/relax aniso 0.0
minimize        0.0 1.2 200000 200001
unfix 3
dump myDump all custom 50 output.dat id type element x y z fx fy fz
dump_modify myDump element H C N
fix 2 all npt temp 500 500 $(100*dt) aniso 1.0 1.0 $(1000*dt)
timestep 0.5
run 200000
2

What is supposed to happen, if you heat a substance way beyond its boiling point?

Please note that ReaxFF parameterizations are usually not very portable to different physical conditions beyond what they were parameterized for (much less so than simpler force fields) and thus you will have a rather low accuracy for some of your simulations.

3

What is supposed to happen, if you heat a substance way beyond its boiling point?

So, this means that the system is indeed volatilizing during the high-temperature simulations. My goal for performing simulations at high temperatures was to obtain high-energy configurations (which should be structurally different from room temperature configurations) for a given system.

Please note that ReaxFF parameterizations are usually not very portable to different physical conditions beyond what they were parameterized for (much less so than simpler force fields) and thus you will have a rather low accuracy for some of your simulations.

Which other force field would you recommend to perform such simulations for simple organic molecules? Would using a potential like Chaarm be better? We also used an in-house neural network potential (trained on a different dataset), which provided similar results as ReaxFF. It seems like performing an NVT, or an NVE simulation is a better option at high temperatures.

4

All of your questions are a topic for a discussion with your adviser. This is all about your research and not about LAMMPS, which makes it off-topic for this forum category. Furthermore, I don’t know anything about your research, so even if I wanted to, I would not be able to give competent advice.

5

Okay, thank you for your help.