[gmx-users] Slab gets bended in NVT

[Alex]

Dear all,
In modelling a slab and some molecules on top of it, the slab gets bended
(Fig.2) during the NVT production or equilibration simulations while I do
NpT equilibration (berendsen followed by Parrinello-Rahman) long enough
(Fig.1) so that the box parameters, pressure and energies are stable enough
prior staring the NVT one.

Fig1: After around 2 ns NpT equilibration
https://drive.google.com/open?id=1lhDruLuSx0Qf5DT4cIriTX3uoi5w_AVS

Fig2: During NVT
https://drive.google.com/open?id=1CyDBrM1Ks52ViG8fh0_pnoClqymJVjGg

The slab got equilibrated separately before I bring it to the new system
and even I restrain the slab's atoms to their equilibrated initial
positions using "7000 7000 5000" spring constants in X,Y and Z directions,
respectively.
The point is that the problem does not happen in a smaller (laterally
around 7nm*7 nm) slab, but it happens when I use a larger slab (laterally
around 20nm*20nm).
Above I have shared two pictures of the system, would you please share your
thoughts about the issue?

Thank you.
Regards,
Alex