[gmx-users] Errors during pull constrained NPT simulation for permeability calculations.

[jay patil]

Hi Experts,
I am doing simulation for permeability calculation. 
  In my simulation, water molecule is constrained (inside the Bilayer) at fixed distance from centre of the Bilayer. And another constrained in bulk water. I am doing equilibration with V-rescale(thermostat) and berendsen(for barostat) for 5ns and then production run with Nose-Hoover and parrinello-Rahman for 15ns. Simulation is crashing during production run after few nanoseconds run. 
Sometimes I get error "distance between pull group 1(Bilayer COM) and 3(water) is large than 0.49 times the box size. You might want to consider using 'pull-geometry =direction-periodic' instead "
  But I cannot use direction-periodic in NPT. 
  Sometimes I get error, "water molecule starting at atom XXX can not be settled."
   I observed, these XXX water molecules are not constrained one's  but they are from bulk water surrounding Bilayer. 
  Parameters for equilibration run.
  tcoupl=v-rescale
  tc-grps= dpp water
  tau-t = 2.0 2.0
  ref-t= 310 310
  pcoupl= Berendsen
  tau-p = 5.0 5.0
  ref-p = 1.0 1.0
  compressibility= 4.5e-5  4.5e-5
  Production:
  tcoupl=nose-Hoover 
  tc-grps= dpp water
  tau-t = 2.0 2.0
  ref-t= 310 310
  pcoupl= parrinello-Rahman 
  tau-p = 5.0 5.0
  ref-p = 1.0 1.0
  compressibility= 4.5e-5  4.5e-5
Bilayer I am using is well equilibrated, I had done 500ns NPT simulation for Bilayer at 310K and 1bar. 
  Question:
  1. What is alternative for "pull-geometry=distance" if my constrained molecule going outside the box due to volume fluctuations, since we cannot use direction-periodic in NPT.
  Molecule is constrained around 1nm inside the initial box boundary.
  2. What might be the reason for instability in bulk water which is SPC. 
  3. If there is distortions in Bilayer during pulling, leading to overlap, simulation would have crashed in initial stages, why it is crashing in later stages. 

  Any helpful tips, suggestions will be much appreciated.

Thanks

Yogesh