You should first build a truly periodic crystal (with fewer atoms), as there are less things that can go wrong.
ok, by periodic/non-periodic crystal I'm assuming you are referring to boundary conditions, not the periodicity of crystal geometry itself.
I have a few more questions.
Do you suggest applying PBC because the suspicion is that edge effects could give rise to problems? Energy conservation problems can crop up because of issues with time-step size, integration algorithm, discontinuities in the potential, but how does the geometry of the configuration come into play? With or without PBCs, the energy conservation problem will remain right?
with fewer atoms.... less things ....can go wrong.
Yes, with fewer atoms the rate of KE leaking into the system would be slower as it is proportional to the number of force calculations between atoms. But the Si block would eventually disintegrate after enough time steps. The performance of LAMMPS-Reax should not depend on the number of particles correct?
Nicholas J. Lee