I have the following problem. The model is a Lennard-Jones solid film on a
frozen solid substrate. Below there is the correspondiing Lammps script. This
script runs normally with the 17Jan05 Lammps version, but with 12Arp06 version
it crashes with the 'lost atoms' error after the first several hundred steps.
I should mention that initial velocity distributions generated by the
'velocity' command differ for two Lammps versions. Hence the atoms
trajectories differ as well. However if I arbitrarily change the seed number
in the 'velocity' command the results remain always the same: the 12Apr06
version crashes, the 17Jan05 version runs normally..
I made the data-file from the restart-file saved after the first 100 steps in
order to compare the runs from this identical initial condition. The result is
again the same: the 12Apr06 version crashes, the 17Jan05 version runs normally.
Presumably this problem may be connected with the free ('fm') boundary conditions.
There is no such problem while using eam potential for the same task (the same
Could you give me some advice where the bug can be? Can it be the Lammps bug?
boundary p p fm
lattice fcc 1.03383
region box block 0.0 5.0 0.0 5.0 0.0 14.0
region freezedreg block 0.0 5.0 0.0 5.0 0.0 2.01
region heatedreg block 0.0 5.0 0.0 5.0 2.02 10.0
create_box 1 box
create_atoms 1 box
mass 1 1.0
group freezed region freezedreg
group heated region heatedreg
pair_style lj/cut 2.5
pair_coeff 1 1 1.0 1.0 2.5
velocity all create 0.95 552233 mom yes rot yes dist gaussian
velocity freezed create 0.0 2745
fix 1 all nve
fix 3 freezed setforce 0.0 0.0 0.0
dump 1 all atom 100 dump_atom