Contact definitions for RVE with periodic boundary conditions - which one?

Hello,

I have a RVE made from three parts: Fibre, Interface and Matrix. The Interface is made from cohesive elements (the meshing is coarse on purpose to keep the time down while I "experiment") - first attached image. The outer surface of the interface is connected to the matrix via tie constraints as is the inner surface of the interface to the outer surface of the fibres. 

In the following a pure shear load case with periodic boundary conditions will be shown. If I don't define anything else, the matrix will eventually overlap the fibre (second attached image). This makes sense, as the matrix an fibre are not directly connected. 

To overcome this issue, I defined a surface-to-surface contact (surface of the fibres with surface of the matrix) with small sliding selected. Tangential Behavior is set as "frictionless" and normal behavior as "hard contact".

However, this causes a very strange result. Right before the two surfaces touch, everything looks fine (third image). But in the very next increment, the complete body deforms, you could even say collapses. There is no error and the simulation completes without any messages. The stress also increase in a "realistic" manner for the following increments. But the part is not recognizable and looks similar in every step to what is shown in the second to last image.

Do you have any suggestions for me?

EDIT: It does not look like the PBC are responsible. If I use the same model without PBC and calculate some simple shear cases (bottom surface fixed, top surface with a horizontal displacement), it too collapses - but the fibre now remains intact (last image)​​​​​​​


Kind regards

Mike