[gmx-users] Gradual deformation of position-restrained structure

[Alex]

Hi all,

This is related to my previous question about trjcat (which Justin answered
perfectly) and now that it works, it has uncovered a problem. So, let me
describe the system and what I am trying to do with it.

I have a rectangular graphene membrane with approx. 4:1 aspect ratio,
suspended at the ends. That is, two lines of atoms along the short
direction at each end in the long direction are restrained to their initial
positions (k=50000 kj/mol/nm^2), which is part of the structure topology.

The goal of the simulation is to periodically apply a constant out-of-plane
force to a single atom at the center of the ribbon, i.e. 10 ns on, 10 ns
off, etc. I couldn't find a way to do that in a single simulation, so I
went with the cowboy version of it, i.e. I have an alternating sequence of
simulations from two mdp's, in one of which there is a constant force via
pull code. Each subsequent simulation picks up the final gro of the
previous and goes on. It works, thanks to Justin, and we are able to track
the deformation in terms of deflection at the center.

Here is the problem: the small force during each "forced" step causes the
ends shift ever so slightly toward each other, which makes the deflection
increase in every subsequent force-on simulation. This is perfectly normal,
given that the atoms at the ends are restrained and not frozen and the end
structure of the (i-1)-th simulation is the beginning of the i-th, so the
"pinned" positions keep shifting ever so slightly (but sufficiently to
cause the observed artifact).

What can we do? Is there a way to accomplish this force-on/force-off
scenario in a single simulation?
Is it possible to make the integrator omit a list of particles, i.e. freeze
them? I know GMX something to freeze atoms, but never tried (and I think it
is unphysical for our purposes)... Any other suggestions?

Thank you,

Alex