# Re: Erratic pressure values during equilibration of water box

From: Axel Kohlmeyer (akohlmey_at_gmail.com)
Date: Sun Jun 17 2012 - 09:08:31 CDT

On Sun, Jun 17, 2012 at 9:49 AM, Ashwin C.V. <ashwin.cv.nitt_at_gmail.com> wrote:
> Hi,
> In our learning process,we were trying to run minimization and equilibration
> on a simple water box system using NPT ensemble.We used Berendsen Pressure
> bath  coupling as follows.
> BerendsenPressure on
> BerendsenPressureTarget 1.01325
> BerendsenPressureCompressibility 4.57E-5
> BerendsenPressureRelaxationTime 20
> BerendsenPressureFreq           2
>
> Also for achieving constant temperature we used langevin dynamics as follows
>
> langevin    on
> langevinTemp  300
> langevinDamping 5
>
> After running for about 100,000 time steps of 1fs each,we observed that the
> output pressure values from namd was always fluctuating between -100 bar and
> +100 bar.Also after increasing no of time steps,we were not able to get a
> convergence zone neither a constant value.
> When we read other questions on the mailing list some of it pointed to the
> fact that these fluctuations between lower and upper bounds were acceptable.
> Is it possible in any way to achieve a convergence zone and constant
> pressure at the end of equilibration?

these fluctuations are due to the (lack of) compressibility.
their magnitude depend on the size of the sample. you
have the same kind of fluctuation on temperature, only
their magnitude is smaller. in both cases, the difference
to expectations from macroscopic experiments is due
to the ridiculously small sample size. thus to get smaller
fluctuations, you have to use a larger system. however,
to get a perfectly constant pressure, you need - consequently -
and infinitely large system.

just an ill posed question. the pressure *should* not
be constant. it only has to be "constant" (actually,
converging to a target value) when averaging over
a large enough sample of frames.

i strongly suggest to read up on these issues in
text book and understand why this has to be rather
than ask for something that cannot be.

cheers,
axel.

> Thanks
> C.V.Ashwin
```--