Re: [AMBER] barostat with large coupling constant

From: Jason Swails <jason.swails.gmail.com>
Date: Sun, 1 Oct 2017 23:37:57 -0400

On Thu, Sep 28, 2017 at 9:14 PM, R zu <rzu512.gmail.com> wrote:

> Hi.
>
> After minimization and a short heating, I try to run NPT equilibration
> before running NVT production simulation.
>
> The volume of the last frame of the equilibration isn't close to the
> average volume.
> I worry that the NVT production would have a high pressure like 100 bar.
>

‚ÄčInstantaneous pressures are rather meaningless in MD simulations.
Pressure is a macroscopic quantity, which means you need to look at the
ensemble (average) pressure over an entire (ergodic) simulation. Given the
very high isothermal compressibility of water, tiny volume fluctuations
yield correspondingly huge pressure fluctuations.

What this means is that seeing an instantaneous pressure near 100 bar (or
-100 bar) is perfectly normal for condensed phase MD simulations. Using a
large coupling constant is actually worse than using a short one for the
Berendsen barostat, as a short coupling constant typically yields computed
compressibilities closer to experimental values (even better is to use the
Monte Carlo barostat, which samples from the correct isothermal-isobaric
ensemble). Said perhaps more clearly -- shorter coupling constants yields
more physically realistic pressure-dependent behavior.

Berendsen thermodynamic regulation (either a thermostat or barostat)
artificially reduces‚Äč fluctuations compared to what is expected in the true
ensemble. The smaller the coupling constant for the barostat, the larger
the fluctuations become (this should make sense if you think about the fact
that you need larger changes to respond to deviations from the target value
in a shorter amount of time). Therefore, shorter coupling constants will
compensate for the artificial reduction in fluctuations induced by the
Berendsen approach.

HTH,
Jason

-- 
Jason M. Swails
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Sun Oct 01 2017 - 21:00:02 PDT
Custom Search