Nose-Hoover vs v-rescale, low density of the system

GROMACS version: :-) GROMACS - gmx mdrun, 2020.5 (double precision) (-:
GROMACS modification: Yes/No

Hello Everyone,

I have a small cage system (51 atoms) in chloroform solvent, total 2046 atoms. I did NPT equilibriation using T = 298 K, P = 1bar, Thermostat = V-rescale, Barostat = Berendsen. which showed the desired temperature, pressure and density is well equilibriated.

MD run 1: I ran a MD simulation using Thermostat = Nose-Hoover, Barostat = Parrinello-Rahman for 20 ns. The average values of the temperature and pressure were as expected. But the density droped to an averge value of 1217.11 kg/m^3, which is much lower than the liquid choloroform density 1477.75 kg/m^3 at 298 K. Then I have checked the temperature plot against the time of the MD which shows the values are fluctuating a lot.

MD run 2: I changed the thermostat = v-rescale ( and kept Barostat = Parrinello-Rahman as previous) for another MD production run. The average value of density is 1424.7 kg/m^3, which seem to be more acceptable than the previous run.

I found in many discussions that Nose-Hoover is good for big systems, for small systems the flucuations are high, resulting the low density.

My queries are,

  1. Is this combination [Thermostat = v-rescale and Barostat = Parrinello-Rahman] is good enough for MD run?
  2. Do we have any rule to determine which thermostat or barostat will work better for a perticular system? What things should I keep in mind while using thermostat or barostat?
  3. How much difference in the density (compared to the solvent exact density) and/or fuctuations in the temperature or pressure during the production run of equilibriaion is acceptable?
  4. How many atoms are considered as small (or big) system here? I did only QM before, and the cage system seem to be big to me.

Please write about your opinions and suggestions.
Thank you for reading.