[gmx-users] Different system volumes from gromacs versions 3.2.1 and 3.3.1

[Perttu Niemela]

Hi again,

I used the "Martini" version of the force-field. The difference in volume 
is only about 1 percent, but still much larger than the fluctuations which 
are about 0.1 percent. The effect is very systematic though, and it can be 
disturbing if you have carelessly run simulations with different versions 
and try to compare the results :)

The reason why I am testing this is that we observed a similar effect with 
other force fields too, and could not explain it. For example, area per 
lipid values for lipid bilayers (all-atom) displayed a similar dependency 
on gromacs version, but now with 5 to 10 percent differences. I am not yet 
sure if this is caused by the same effect at all, but continuing to test 
on it.

If anybody has seen similar effects and/or has other possible 
explanations, would be nice to hear them.

Regards,
Perttu

> On Tue, 8 Jan 2008 15:39:51 +0200 (EET)
>  Perttu Niemela <perttu.niemela at hut.fi> wrote:
>> Hi All,
>>
>> In December here was a short discussion about a water system, for which
>> gromacs versions 3.2.1 and 3.3.1 produced different average volumes. I have
>> experienced similar thing with Marrink's coarse grained water (about 1500
>> beads in the box), but to opposite direction.
>
> Just by curiosity: which version the ff you use?
>
>> My test runs with gromacs 3.1.4 and 3.2.1 always give a larger average
>> volume than versions 3.3.1. and 3.3.2. The difference is much larger than the
>> fluctuations. I use berendsen barostat (isotropic; p = 1 atm) and thermostat
>> (T=300K). In my simulations, the off-diagonal terms of the virial are very
>> close to zero. Additionally, I have tested using simple cutoffs instead of
>> switch/shift functions, but it does not affect the general conclusion.
>
> How much is the difference you observe and which percentage of the volume
> does this represent?
>
>> Can it be that the difference in calculating the kinetic energy (and
>> temperature) leads to the observed volume difference? I did a quick test to
>> modify the code of gromacs v. 3.3.1 such that it calculates kinetic energy
>> exactly like in v. 3.2.1, and the volume jumps back up.
>>
>> Can somebody please confirm, if this is expected? I cannot fully understand
>> how a more exact calculation of the kinetic energy would lead to a
>> significant change of the average volume (?)
>
> Well a change in temperature should affect the volume of course, so the
> change in estimation of the kinetic energy will affect the volume but
> the effect should be very small!
>
> XAvier
>