From: Jérôme Hénin (jhenin_at_ifr88.cnrs-mrs.fr)
Date: Mon Nov 29 2010 - 10:26:16 CST
On 26 November 2010 20:03, Sebastian Stolzenberg
> Dear Jerome,
>> Â It does sound like a problem. To distinguish whether it comes from
>> Â pressure measurement itself or from the barostat, it would be very
>> Â useful to repeat this comparative test in constant volume simulations.
> To achieve running an NVT ensemble simulation, I did the following:
> - switch off the "LangevinPiston" options
> - useFlexibleCell "yes"->"no"
> I have four runs of the same system:
> a) with piston (NPT), group pressure yes
> b) with piston (NPT), group pressure no
> c) NVT, group pressure yes
> d) NVT, group pressure no
> with the following mean GPRESSAVG values:
> a) (1.1+-1.6)bar
> b) (-14.1+-2.6)bar
> c) (8.8+-12.8)bar
> d) (-0.94+-13.01)bar
> means are over >=0.5ns trajectories. Uncertainties are simple standard
> deviations (energy fluctuations are numerically too large to compute
> Boltzmann averages).
> My Questions:
> Does this help at all? If yes, please explain: are you simply looking for
> differences in the GPRESSAVG output?
> Or are there distinct pressure values/differences in the NVT that I would
> have to expect physically?
These results (which match what I have just measured on a simple cubic
water box) indicate that the root of the issue is not a problem with
the barostat. Still, it suggests that the barostat might not be using
the same pressure value as the one in the output when useGroupPressure
is disabled, as it fails to bring that value to 1 bar.
> 2.) As a consequence, is it safer to use GroupPressure, even if the "rigidBonds"
> option is not used?
That would seem to be the case, although I would not use the word
"safe" until this surprising behavior is fully sorted out. I am not
well-versed in that part of the code (especially since the dreaded
ComputeNonBondedBase(2).h might be involved), but hopefully our
friends in Urbana will chime in.
> Qu'est-ce que c'est, je ne trouve pas "FÄ™te de la rÃ©colte" en wikipedia pour
> la France!
This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:54:48 CST