[gmx-developers] PR pressure coupling

[ANick at t-online.de]

Dear Micahel,

Thanks for your detail answer.

>How bad is this in practice?  With a large coupling constant (longer
time
>scale), it won't be that bad, since it will always be very close to
>equilibrium. 
This is exactly right for my simulations the density for large coupling
constant
is closer to the MC value than that for smaller.
But unfortunately at the end I am interested in the box parameter
changes.
After the simulations I am using the MD trajectories as input for
calculations
of the elastic constants via the Parrinello-Rahman fluctuation
approach
that is based on the fluctuation of the box vectors. The convergence
of
the elastic constants is very bad for large coupling constants, since
(obviously)
the box shape does not change rapidly in this case. Therefore I
reduced
it, but as far as I now understood, I get away from the correct NpT
ensemble?

So am I correct whit my conclusion that I can use the PR trajectories
taking
into account that the ensembles are not fully correct? 
It should be as good as using Berendsen coupling?
At the moment I am rerunning some systems with Berendsen coupling
to see if there are differences in the final values for the elastic
constants. But because of the scatter between individual systems it 
needs some time to get reasonable averages.

Concerning your point of drifting box side lengths, I only see this at
high temperatures. Below the glass transition temperature of my
polymer
boxes this is no problem; they are quite stable (85° to 95°) and I do
not
see any obvious artefacts.

Btw, are you aware of any other software that allows variable shape
NpT-MD?
As far as I figured out GROMACS is the only one that has implemented
this
featrure.

Again, thanks for your answer,
Gitta