2D system treatment

Hi developers,

setting up 2D slab systems we noticed something ‘unexpected’.

We prepared 2 systems, both are SPC/E water confined between graphene sheets that are frozen.

System 1 has nwalls = 2 with dummy atomtypes (charge and LJ params set to 0), wall-ewald-zfac = 3 and pbc = xy

System 2 has no walls and pbc = xyz but explicit vacuum that is two times larger than the slab (equaling the behaviour of wall-ewald-zfac = 3

Both systems have ewald-geometry = 3dc set.

In theory, the two systems should yield the same results as electrostatics and LJ interactions should be treated the same. Yet, the systems differ in nearly all observables.

We are looking into this right now, but maybe we are missing something here?


Henrik Jaeger

EDIT: Tested with 2021.4 and 2022.4

Do the systems have the same initial coordinates? And do they differ already at step 0 or after many steps?