Dear gromacs users,
I am running a simulation for acetonitrile-water binary mixture systems. The simulation system consists of 1800 acetonitrile and 200 water molecules, box dimension: 5.5 nm (Cubic). The acetonitrile amber-ff parameters are obtained from RED server (RESP charges) and acpype program. I have run two sets of simulations:
- energy minimization followed by 100 ns of NVT equilibration and then 100 ns of NPT equilibration using velocity-rescale thermostate (0.1 ps coupling time) and Berendsen barostat (using 1 ps coupling time); However, as soon as I am starting the production MD simulation (NPT ensemble) using Parrinello-Rahman barostat (5 ps time constant), the simulation is crashing throwing many LINCS warnings.
The box length is becoming huge: from 5.6 nm it is increasing to over 150 nm or so.
- I ran the production MD with Berendsen barostat (5 ps time constant) and things worked fine without any error or unrealistic box length fluctuation.
I am using compressibility of water for NPT i.e. 4.5e-5 as the isothermal compressibility in the mdp file.
For some reason the Parrinello-Rahman barostat is not working well for this system. As far as I know, the major difference between Berendsen barostat and Parrinello-Rahman barostat is that PR scales the box vector instead of the particle coordinates as Berendsen does. Does the compressibility factor play an important role here? As PR barostat is changing the box volume to keep up with the reference pressure, is that resulting in a very large change of volume with respect to pressure and the simulation is crashing?
I am stuck with this situation and any help would be much appreciated, thank you.