Question on losing atoms during minimization

Dear all,

I am new to MD and lammps. Recently one problem emerges when I try to
calculate the grain boundary energy of a fcc Ni bicrystal. At the first,
periodic boundary was set for each of the 3 dimensions. Then a bicrystal
with specific orientations was constructed. Before energy minimization,
overlapping or too close atoms were deleted judging by a cut-off distance.
However, an error occurs during the energy minimization after 34 time-steps.
In order to make a detailed description on my question, my input file and
log file are list as follows. Moreover, this error occurs only when the
code was performed a parallel run via a mpirun command on Linux, but not
happen when it runs on Windows. I will be so lucky and appreciate if you
could give me some suggestions about my question at your convenient.