For NPT with eps_Ge=0 the Germanium pore is being stretched axially to twice
its equilibrium length. Presumably this unphysical stretching opens up holes
in the Germanium crystal for Silicon to penetrate.
The measured pressure, and therefore the amount of stretching, will depend
strongly on the value of eps_Ge.
Here are some additional concerns:
0. Check your results quantitatively against known results. Just because the
movie looks okay, this does not prove that the simulation is correct.
1. Does combining NPT and fix setforce even make sense?
2. Applying a thermostat to frozen atoms is a *bad* idea (I noticed the
temperature stayed at about half the target temperature)
3. This may explain why the pressure did not go to zero in either case.