[lammps-users] problem about NVT and lost atoms

Dear Axel and all others,
I have followed Axel's suggestion, but pity to find I can't say the bond breakage.
The default representation in VMD is line, when the distance between two atoms is big ,there will be a longer line , with Dynamicbond method ,there will be no bond. but from VMD, I can only see the nanotube tensiled to a little strain, no longer line, no bond breakage, so I doubt that I didn't finish the simulation .
when I set a velocity to one end of the nanotube, run 100000 setps, the error and the manual help information are:
Lost atoms: original \.15g current .15g
A thermodynamic computation has detected lost atoms.

I think after lost atoms detected by thermo dynamic computation, the simulation is end, should I make the box bigger or something ?

I want to get the simulation result like this:http://lammps.sandia.gov/images/nanowire2_small.jpg,which is a bench work in the "movie" folder in lammps website named "Au nanowire formation and extension".two end thick and the middle thin ,during the simulation, the middle part from thin to zero,