Oscillations in volume related to the barostat

GROMACS version: 2020.2
GROMACS modification: No

I’m running a simulation of 1000 molecules of trichloroethylene. After an equilibration, which seems to have equilibrated well, I notice fast and large frequencies in the density (or volume) readings, which are related to the parameter ‘tau_p’ (this is the same tendency of the wave length).

I use the Parrinello-Rahman barostat. I tried to change the thermostat to the Nose-Hoover, change the compressibility, and adjust ‘tau_t’ and ‘tau_p’. However, I can’t avoid the oscillations.
Does anyone know how to solve this problem?
I’m attaching the mdp.
prod.mdp (910 Bytes)

Warning and Note:
“WARNING 1 [file topol.top, line 20]:
The following macros were defined in the ‘define’ mdp field with the -D
prefix, but were not used in the topology:
If you haven’t made a spelling error, either use the macro you defined,
or don’t define the macro.”
(I removed the POSRES.)
Cleaning up constraints and constant bonded interactions with virtual sites.
The number of degrees of freedom in the T-Coupling group System is 16,997.00.
NOTE 1 [file prod.mdp]:
There are 3000 non-linear virtual site constructions. Their contribution
to the energy error is approximated. In most cases, this does not affect
the error significantly.
(I’m not familiar with this note, as this is my first time using virtual atoms.)

Hi, historically the way to avoid these was to start the simulation with the Berendsen barostat, as the oscillations in PR are mostly present during density equilibration. In the latest Gromacs versions though (starting from 2021 - assuming you can compile a newer version somewhere), the recommended barostat is C-rescale, a recent algorithm from Giovanni Bussi’s lab that doesn’t suffer from oscillations and recovers both correct average and fluctuations.

As suggested by Milosz, the c-rescale barostat is a good choice, both for equilibration and production, if you can upgrade to a more recent GROMACS version.

Otherwise, you state that you have tried different tau_p, but not which values. How high do you go? I would suggest between 5 and 10 ps, but even 15 ps might be necessary sometimes, especially if your system is not properly equilibrated.

I appreciate your insightful responses.
It worked when I used the C-rescale (following the Berendsen barostat in the equilibrium).

Furthermore, I would like to point out that even with a greater tau (tau_p = 15), I still observe the constant ocsilation for the Nose-Hoover. Despite this, the mean and RMSD values are extremely similar to those with the C-rescale.