Losing Atoms in Airebo Potential

Dear LAMMPS Users:

I plan to simulate a pyrolysis process of a “kerogen” molecule from 300K to 5000K with 5K/ps temperature increasing rate. I applied Airebo potential with langevin thermostat and NVE ensemble. However, as the temperature goes into about 1500K, LAMMPS reports an error that the atoms are lost. Has anybody encountered the same issue? I am a new user and looking forward to your help.

The input file is as follows:

units metal
dimension 3
atom_style full
boundary p p p

read_data Kerogen_all_carbon_8.data

pair_style airebo 3.0
pair_coeff * * /home/twohee/Downloads/lmp/lammps-9Dec14/potentials/CH.airebo C C C C H

velocity all create 300.0 53244 dist gaussian mom no rot no
fix initial all temp/berendsen 300.0 300.0 0.1

run 10000
minimize 1.0e-4 1.0e-6 100 1000

timestep 0.001
reset_timestep 0
neighbor 2.0 bin
neigh_modify every 10 delay 20 check yes

thermo_style multi
thermo 1000

velocity all scale 300.0
fix 1 all langevin 300.0 5000.0 0.1 48279
fix 2 all nve

dump 1 all xyz 80000 dump_kerogen_8.xyz
dump 2 all image 200000 heat.*.jpeg type type
run 800000

LAMMPS says:
ERROR: Lost atoms: original 1336 current 1335 (…/thermo.cpp:393)

Thanks.

Shuai He
Ph.D. Student
Chemical and Biomolecular Engineering, University of Houston, Houston, TX, 77057

Dear LAMMPS Users:

I plan to simulate a pyrolysis process of a "kerogen" molecule from 300K to
5000K with 5K/ps temperature increasing rate. I applied Airebo potential
with langevin thermostat and NVE ensemble. However, as the temperature goes
into about 1500K, LAMMPS reports an error that the atoms are lost. Has
anybody encountered the same issue? I am a new user and looking forward to
your help.

lots of people have reported "lost atoms". it is one of the most
mentioned topics on this mailing list. you can search the mailing list
archives and see.

what you describe reminds me of the question of how the length of time
step is related to the fastest motion in the system (and thus for
higher temperatures a shorter time step is needed). at 1fs, your time
step is already probably to large to accurately do time integration
for hydrogen atoms. thus i would recommend to first do some studies
with fix nve and withOUT fix langevin at different (initial)
temperatures to determine, which time step gives you good energy
conservation at what temperature. due to your ramping of the
temperature, it cannot be determined from the simulation as is. it is
also recommended to read up on how to monitor energy conservation and
properly choose the correct time step for a given system in your
favorite text book on MD simulations.

axel.

Why did you minimize after the first 10,000 step run?

I’ve usually seen it done the opposite way, (first relax/minimize then run).

Just wondering.

dc.