Re: why pressure is fluctuating so much

From: Binquan luan (bqluan_at_uiuc.edu)
Date: Wed May 03 2006 - 18:56:25 CDT

Hi,
I am also very concerned about large pressure fluctuations. Since the
langevinPistonDecay is set to be 100 or 200 time steps, pressure should
converge to the targeted one in that time scale too. I assume that
langevinPistonDecay set the time scale for pressure to decay, right?
 
I also did a simple test. If I use a big dielectric constant(say 10^6),
the pressure quickly converges to 1bar and stays there with very small
fluctuations (+-0.05).I dumped results very 1000 time steps. In this
system, there are no electrostatic interactions. So, can electrostatic
interactions cause such big fluctuations?
Please correct me if I am thinking something wrong.
Thanks,
Binquan

> hi,
>
> > I am using a NPT ensemble in PBC condition for my production run.After
> > analysis what i observe is that the pressure deviates from -400 to +400
> > bar,even though i have given 1 bar in the config file using Langevin
> > dynamics.
> i think this is quite normal in NPT simulations. pressure changes are
> induced by a change in the simulation cell dimensions. due to very low
> compressibility of water (i presume you're running simulation in water),
> even very small change in volume is accompanied by large change in pressure.
> try to do a running average of the pressure curve. i think it will be
> fairly close to 1 atm.
>
> > Also the cell volume change by 2/3rd (smaller in volume) after production.
> > Prior to this i did equilibriation in NVT ensemble for a fairly long time.
> well, after switching from NVT to NPT you can expect some changes in
> simulation cell vectors. for example, density of your system in NVT
> might have been wrong and it couldn't be changed because both N and T
> vere kept constant.
> but the change is quite big... maybe you should try to calculate the
> density of your system from the nvt and npt simulation vectors and
> decide which one is closer to reality...
>
> regards,
> lubos
>

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:43:35 CST