Actually although annealing reduces the defects, but it never creates the single sound crystal. I am going to study the effects of defects. So, I need to have an initial single sound crystal. But I cant!
May I suggest box_relax + minimize (http://lammps.sandia.gov/doc/fix_box_relax.html and http://lammps.sandia.gov/doc/minimize.html ) on the unit cell before replicating. A very small cell will suppress the nucleation of long ranged defects so you can use this to your advantage.
-Mitch