Dear LAMMPS users,
I am using LAMMPS to perform brownian dynamics of particle clusters, using fix rigid/nve langevin, grouping my particles in rigid groups, the groups have a colloid potential between them and the internal interactions are excluded, the atoms are spheres. I am facing a recurrent problem:
The simulation runs smoothly during a large number of steps (more than 1M), and suddenly, without any visible reason (no dangerous neighbors built, no “risky” configuration nor particle overlap), the temperature starts to increase sharply until reaching NaN. This phenomenon is not sensitive to the timestep. The temperature is computed using the temp/sphere compute.
I must state that I am using the python driver with pypar and open mpi to drive my simulations and stop it before the particles overlap. The initial 1000 steps are not sufficient to cause the crash. The data77.dat is attached to this mail.
The code is the following:
Units real
atom_style sphere
atom_modify map array
dimension 3
newton off
read_data data77.dat
pair_style colloid 4000.0
pair_coeff 1 1 0.0 0.0 100.0 100.0 2000.0
pair_coeff 1 2 3.0 0.0 2000.0 2000.0 4000.0
pair_coeff 2 2 0. 0. 100.0 100.0 2000.0
group clus1 type 1
group clus2 type 2
neighbor 1000.0 bin
neigh_modify delay 0
communicate multi vel yes
neigh_modify exclude group clus1 clus1
neigh_modify exclude group clus2 clus2
velocity all create 298 3551 loop geom
compute tempreal all temp/sphere
fix 1 all rigid/nve group 2 clus1 clus2 langevin 298 298 2.44e6 85138
dump 1 all atom 10000 dump.test77
thermo_style custom step c_tempreal epair etotal evdwl fmax press vol
thermo 1000000
timestep 800000
run 1000
Do you have any idea what could be the reason of this?
Best regards,
Baptiste
data77.dat (609 Bytes)