How much did you change tau, and what happens if you use Berendsen or a different barostat instead? I have to look at my own input files, but 20 seems funny to me.

Also, how much is box size changing? PR depends on the initial box size, so if the box changes alot that messes up PR. Usually we equilibrate with Berendsen, then switch to PR for production

@B.Kelly I've tried tau from 5 to 20, with Parinello-Rahman barostat. Box size quickly becomes more or less stable.

@ShoubhikRMaiti I've started with Parinello-Rahman, probably I'll try with Berendsen. I'm using Gromacs 2021.

Are you near a phase boundary?

@B.Kelly I don't know. I have small molecules dispersed in polymer more or less uniformly.

What constraint algorithm and size of time step are you using... I assume there are hydrogen?

I use 2 fs timestep and LINCS as constraint algorithm with only h-bonds constraints, otherwise it crashes.

Hmmm... How long of a run is it? More than a nanosecond?

I tried 10-50 ns.

Did you energy minimize before starting equilibration? Part of me wonders if there is a problem with a force-field parameter(s). Gromacs uses atomic virial to calculate, first the virial, then the pressure. So, it is normal for pressure to bounce alot since the atomic virial is at the mercy of the bond(spring) forces. But it doesn't bounce like you are seeing, and, usually meets the set pressure on average. What was your starting box size? Are you sure you are in the right phase at equil? Maybe you have converged to a vapor phase?

@B.Kelly I’ve minimized the energy (Fmax ~ 50) equilibrated the system as NVT ensemble for 1-5 ns, and trying to equilibrate it as NPT ensemble. I really don’t know about phase, the density is close to the experimental. Could it happen due to my system being heterogeneous? The box size is more or less consistent, in VMD it doesn’t seem to be jumping.

Also I’ve tried Berendsen barostat, it didn’t help,

← 14 messages moved from Discussion between roma ichenko and B. Kelly

can you attach your entire .mdp file? that is easier than me one by one checking off things... like I am curious what thermostat and integrator you are using, how often neighborlist is updated, etc

@ShoubhikRMaiti @B.Kelly @roma Just a reminder: we had to move this conversation from the new room you created to here, because those rooms are extremely hard to maintain:

If you see a back-and-forth comment chain on any of the questions or answers, and you're able to step in and re-direct traffic to a specific room, I think this is much preferred over the risk of letting someone accidentally click that button which creates rooms (in the long-run this will help us keep the total number of rooms manageable, rather than the 678 pages of rooms at Math.SE!).

Here's an example where I re-directed traffic in this way: mattermodeling.stackexchange.com/questions/1733/…. I try to do that as often as possible, but can't be there all the time, so some rooms accidentally get created, but if more of us work together to stop that from happening, I think it would be good for the site overall!