[lammps-users] AIREBO variable atom count problem

LAMMPS LAMMPS Mailing List Mirror
1

I am running AIREBO for a problem where the atom count changes. I am heating carbon rapidly and expect atoms to be lost. Despite attempting to set the thermostat appropriately and use the ignore lost atoms options, bad energies occur after a while, presumably when atom loss occurs. The energies are large and unphysical. I presume that either (i) I have overlooked something (ii) there is something else to set for reactive calculations.

Lammps version is 7 July 2009. Runs are on 216 cores on a Cray XT. My test system has 6000 atoms.

The bad output:

LAMMPS (7 Jul 2009)
...
Step KinEng PotEng TotEng Temp
    0 0.00000000 -42409.82175532 -42409.82175532 0
  500 104.59287585 -42588.37577191 -42483.78289605 134.99595
1000 124.11260345 -42614.51987580 -42490.40727236 160.18967
1500 105.35713933 -42600.95968467 -42495.60254534 135.98237
...
128000 741.24917771 -41984.68705509 -41243.43787738 956.71558
128500 41547462.47345704 30467.07934456 41577929.55280159 73890841
129000 29481363.96341164 26365.45049593 29507729.41390757 65558424
129500 28735472.21258219 6504.13269690 28741976.34527909 76263220
130000 23144487.16392749 7932.46962369 23152419.63355118 72461977
130500 22571109.96112798 3493.05185261 22574603.01298059 83709351

The input is as follows:

units metal
atom_style atomic
read_data data.start.cnd2
pair_style airebo 2.0 1 1
pair_coeff * * CH.airebo C H
timestep 0.0002 # =0.2 fs
compute_modify thermo_temp dynamic yes
fix 1 all nvt 00 1500 .01
thermo 500 # Every 100 fs
thermo_style custom step ke pe etotal temp
thermo_modify format 2 %15.8f format 3 %15.8f format 4 %15.8f
thermo_modify lost ignore flush yes
dump traj all custom 400 dump.cnd2 id type x y z vx vy vz # Every 200 fs
restart 100000 restart.cnd2.1 restart.cnd2.2 # Every 50 ps
run 200000 # 500=100 fs 10000=2ps

Thanks for any advice.

-- Paul Kent

2

Two comments. Your system appears to be periodic, so how are
you expecting to lose atoms? They might just disappear if they
fly far out of the box (in one timestep or a few), but that is not
a very physical model. A more physical model is to have
a free surface, and a bounded box, so that when atoms leave the
box (by a small amount, then they disappear).

Also, I would not use NVT to thermostat a system that is losing atoms.
Try fix langevin. I would also print out the number of atoms in
your system, so you have a sense of what is happening. Also
I would print out thermo more often, so you see how it goes bad
near step 128500.

Also your fix NVT command didn't print out right - it doesn't show
what the damping constant is, and you are using an older version
of the code. These are minor points compared to the first one.

Steve