[lammps-users] about fix boundary

i want use lammps for simulation of nanopore with square section and infinite length.i want that when atom in poure arrive to wall, reflect .i use this common
boundary f f p
but after afew (10000 or more or less) time step run stop or print nan.
my qustion :
is my common for this condition tru?
why atom lost in fix boundary?
thank you very much.

There are lots of ways to lose atoms or compute NAN forces
when atoms overlap - my guess is you are doing the relection
boundary incorrectly somehow ...