simulating graphene nanostructures

I try to simulate a graphene nanoribbon but the simulation is stopped since some atoms are lost.

Below is the simulation input file. I would like to be sure that the use of the boundary conditions are correct. It seems that this is not a problem due to the potential since this happens with the AIREBO and the Tersoff potentials.
Moreover, 2D graphene sheet (using periodic boundary conditions) are stable. So it should mean that the problem comes from the atoms located at the boundary which only have 2 neigbour instead of 3.

As papers presenting MD simulations results on graphene nanoribbons have already been published, I figure out that this kind of problem should not occur. Maybe this has also happent to other user. If so, do they found a solution ?


units metal #mandatory unit for the AIREBO/REBO potentials
dimension 2 #just to try with the 2D lattice dynamic of graphene
atom_style atomic
boundary p f p #for 2D simulations the z direction must be periodic


lattice custom 2.53 a1 1 0 0 a2 0 1.732051 0 a3 0 0 1.361789 &
                 basis 0 0 0 &
                 basis 0.5 0.5 0 &
                 basis 0.5 0.16666 0 &
                 basis 0 0.66666 0
region boite block 0 100 0 1000 0 1
create_box 1 boite # create the simulation box
create_atoms 1 region boite # fill in the region with atoms
pair_style tersoff # define the potential
pair_coeff * * SiC.Tersoff C # define the file within the potential is parametrized. file to be included in the simulation directory
mass 1 12.0 # atomic weigth for C

velocity all create 600 5641

timestep 0.001 # time step in ps
thermo 1000 # information stored each tis many time step
thermo_style custom step temp ke pe etotal

fix 1 all nve
run 100000

Try using

boundary s s s