[lammps-users] Error: Lost atoms

Dear lammps users:

During using Lammps, the biggest problem I have met is the loss of the atoms while simulating the structure under high temperatures.

The structure (defective carbon nanotube) is built and relaxed to be stable by other softwares. Then I run the relaxed defective nanotubes with Lammps. It is normal under lower temperatures, while atoms loss occurs under higher temperatures.The in file is displayed below, and we can see the loss from the log file.

I have looked through the past sourceforge entires(not all of them). None of decreasing timestep(0.0005-0.00008), minimizing the structure, increasing box and adjusting neighbor(0.01-10.0) is effective at all. Would you please tell me why and give some directions ? Thank you so much…

Best wishes!

Zhongqiu Li

#in file

units metal

read_data 200.lmp

pair_style airebo 3.0 1 1

pair_coeff * * ./CH.airebo C

neighbor 0.1 bin

neigh_modify delay 0 every 1 check yes

group c type 1

fix 1 all nvt 5000.0 5000.0 0.1 drag 0.2

timestep 0.0005

thermo 100

thermo_style custom step atoms temp pe ke etotal evdwl

thermo_modify lost ignore

dump 1 all atom 500 200.lammpstrj

minimize 1.0e-4 1.0e-6 1000 10000

run 200000

#part of the log file. We can see that when run to step 21500 the temperature(the 3rd) becomes out of control,then run to 52200 the number of the atoms(the 2ed) becomes less.

21300 2070 5130.3223 -14028.354 596.47569 -13431.878 -14028.354

21400 2070 5082.961 -14003.196 583.80943 -13419.387 -14003.196

21500 2070 2247426.7 -13648.744 601050.1 587401.36 -13648.744

21600 2070 1288610.7 -13710.672 344625.07 330914.39 -13710.672

52000 2070 5.4214614e+008 2024.4656 1.4499115e+008 1.4499318e+008 2024.4656

52100 2070 5.1579429e+008 4458.0106 1.3794363e+008 1.3794809e+008 4458.0106

52200 2068 4.7740628e+008 2747.7643 1.2767717e+008 1.2767992e+008 2747.7643

52300 2068 4.5225434e+008 8642.1996 1.2095056e+008 1.209592e+008 8642.1996

The first thing I would do is run NVE and not ignore lost atoms.
If you can't run an equilibrium simulation in a stable manner
with NVE, then running NVT is just putting lipstick on a pig,
i.e. it will mask something ugly that is going on.

If you do lose an atom in NVE, then you need to figure out
why it occurred. It could be that AIREBO is not parameterized
to work at extremely high temperatures. The BKS potential
for silica, has that problem, for example.


2010/1/6 Woeser <[email protected]>: