I am working on a MD code (http://www.acmm.nl/molsim/frenkel_smit/Exercise_10/index.html) which uses the position Verlet algorithm to integrate the equations of motion and a velocity rescaling algorithm for initialization in periodic boundary conditions.
I wondered why they use a specific step in the integration routine (integrate.f)
Mxx(I) = Mxx(I) + Xxx(I) - Rxx(I)
Myy(I) = Myy(I) + Yyy(I) - Ryy(I)
Mzz(I) = Mzz(I) + Zzz(I) - Rzz(I)
to define particles which are not put back in the box (Mxx). Xxx is the current position and Rxx is the old position.
Mxx is not used in any other way in the propagation itself.
2. The attempt at a solution
Is this step needed to calculate the diffusion coefficient in periodic boundary conditions, i.e. to have some kind of common origin? I cannot imagine what else the usage of this step could be. In the initialization Mxx is set to the current position and the differce (Xxx(I) - Rxx(I)) might give a measure for the drift.