[lammps-users] diferent results for the same parameters with different running procedures

LAMMPS LAMMPS Mailing List Mirror
1

Hi LAMMPS users,

I was using DPD method to simulate a rod-coil copolymer system. The
simulation crashed at timestep 7150 and lammps gave me the following
error message:

ERROR on proc 1: Bond atoms 37802 37803 missing on proc 1 at step 7150
ERROR on proc 2: Angle atoms 37803 37804 37805 missing on proc 2 at step 7150

I have set

communicate single cutoff 10.0 vel yes

The cutoff for communication between different processor is 10, which
I think is big enough. I am a little surprised by this error message,
so I would like to find the distance between atoms 37802 37803, so I
set the sytem to run 7000 steps first, then continuing running another
1000 steps, where I would output the configuration at each step by the
following script:

run 6000 (note: I have run 1000 step for nve before this nvt run)
undump 1
dump 1 all atom 1 debug.nvt.T${tt}.lammpstrj
run 1000
write_restart restart.after.*

I expected the simulation would stop at the timestep 7150. However,
the simulation didn't crash. I don't understand why this happens. The
thermo outputs from two runs are as following:

1. make a long run, crashed at 7150,

Step Temp Press PotEng E_bond E_angle E_vdwl TotEng Dt

    6400 1.0967603 21.107456 3.3154345 0.70708241
0.45831164 2.1500404 4.9605358 0.01
    6500 1.1251306 21.345061 3.3126857 0.68989945
0.47751124 2.145275 5.0003414 0.01
    6600 1.0681827 21.318359 3.3146529 0.67959676
0.45927675 2.1757794 4.9168888 0.01
    6700 1.0841398 21.373696 3.334401 0.67823091
0.47134607 2.184824 4.960572 0.01
    6800 1.0663954 21.453016 3.2887075 0.64916505
0.45130369 2.1882388 4.8882625 0.01
    6900 1.1452508 21.420367 3.2686679 0.68711011
0.45067579 2.130882 4.9865032 0.01
    7000 1.1127547 21.269785 3.2826127 0.69260739
0.46066263 2.1293427 4.951705 0.01
    7100 1.5097478 22.533366 3.3207702 0.77042922
0.43797242 2.1123685 5.585338 0.01
ERROR on proc 1: Bond atoms 37802 37803 missing on proc 1 at step 7150
ERROR on proc 2: Angle atoms 37803 37804 37805 missing on proc 2 at step 7150

2. run 7000 steps,

Step Temp Press PotEng E_bond E_angle E_vdwl TotEng Dt

    6400 1.0967603 21.107456 3.3154345 0.70708241
0.45831164 2.1500404 4.9605358 0.01
    6500 1.1251306 21.345061 3.3126857 0.68989945
0.47751124 2.145275 5.0003414 0.01
    6600 1.0681827 21.318359 3.3146529 0.67959676
0.45927675 2.1757794 4.9168888 0.01
    6700 1.0841398 21.373696 3.334401 0.67823091
0.47134607 2.184824 4.960572 0.01
    6800 1.0663954 21.453016 3.2887075 0.64916505
0.45130369 2.1882388 4.8882625 0.01
    6900 1.1452508 21.420367 3.2686679 0.68711011
0.45067579 2.130882 4.9865032 0.01
    7000 1.1127547 21.269785 3.2826127 0.69260739
0.46066263 2.1293427 4.951705 0.01

the next 1000 steps

    7000 1.1127547 21.0987 3.2826127 0.69260739
0.46066263 2.1293427 4.951705 0.01
    7100 1.1676093 21.290735 3.3215576 0.74106337
0.45072515 2.1297691 5.0729299 0.01
    7200 1.1909944 21.084995 3.2404307 0.75245499
0.44470009 2.0432756 5.0268798 0.01
    7300 1.0443722 20.694148 3.2322305 0.71750029
0.44304914 2.0716811 4.7987516 0.01
    7400 1.3378401 21.61837 3.199652 0.7498716
0.43366541 2.016115 5.2063644 0.01

Comparing those two outputs, we can see the difference starts at the
time step 7100. Could anyone tell me what is going on when the
simulation continued at the timestep 7000? Thank you in advance.

Best,

Dongsheng

2

The error comes from the 2 atoms in the bond being so far apart (more
than 10.0 Angs or sigma in your case) that a processor can't compute
the bond interaction. It could be different on a restart due to when
neighbor lists are computed. The real problem is why the 2 atoms
are so far apart since that is likely unphysical. Figure that out and
the problem should go away.

Steve