My system blows up in Parrinello-Rahman barostat but not in Berendsen

GROMACS version: 2020.4
GROMACS modification: No

Dear all,

I have a small box of chlorobenzene molecules and I’m using OPLS-AA FF and I got the parameters and charges from the LigParGen server. When I run the production MD in NPT, if I use Parrinello-Rahman barostat and Nose-Hoover thermostat the system will blow up. Conversely, If I use the Berendsen barostat and V-rescale thermostat everything is fine.

I tried more restrict energy minimization and equilibration on NVT and NPT with more time (200ps) and also a prior NPT simulation (200ps) without constraint in Berendsen barostat followed by a production NPT in P-R and it always blows up in few steps.

If there is something wrong with my simulation why is it all good in Berendsen? It gives reasonable density and dielectric constant too.

I have attached the mdp and topol files and em.gro (as em.dat) in case you want to take a look at them.

md_relaxation_Berendsen.mdp (2.7 KB)
md_PR.mdp (2.7 KB) (6.1 KB)
em.dat (459.9 KB)

What about Parrinello-Rahman barostat and V-rescale thermostat? Or you can try also stochastic cell rescaling (as barostat). “The stochastic cell rescaling algorithm is a variant of the Berendsen algorithm that allows correct fluctuations to be sampled.”

This post may also be of interests

Best regards

1 Like

Also, 200 ps really isn’t a ton of equilibration time, if there’s something unstable about your starting configuration. It’s probably worth trying 10s of nanoseconds of Berendsen to see if your system stabilizes after that - should be very feasible with a small box.

1 Like