[gmx-users] Wall potential for a membrane simulation

[Marianne Schulte]

Hello,

I would like to 
simulate a membrane system with two walls, one at the bottom of my box 
at z=0 and one at the top, using the gromos53a6 forcefield (GROMACS 
version 4.5.5).
My testing system consists of a membrane in the 
middle, water and sodium ions (40xNa+) above the membrane and water and 
chloride ions (40 Cl-) beneath it.
First, I built up the system, did 
an energy minimization and then implemented the walls for the 
npt-equilibration. I didn't change anything in the topology. (Do I have 
to define the wall-atom-type there?If yes, what exactly do I have to put
 in the topology?)
The interesting part of my nvt.mdp file looks like that:




	
	
	
	



nwall = 2


wall_type = 10-4


wall_density = 100
100		


wall_atomtype = C C


wall_r_linpot = 1 1		;
with -1 I got a fatal error, because atoms were beyond the wall


wall_ewald_zfac = 3


ewald_geometry= 3dc


pbc = xy			; Periodic
Boundary Conditions in x/y direction



What I see in VMD is the following: as expected some NA-ions are 
going through the membrane to reach the chloride/water side. But at the 
same time water molecules (and one chloride-ion) at both sides disappear
 in z-direction and never come back. In the first 10ps (with a tiny time
 step of 0.1fs) there were only a few molecules that went outside the 
box, but after 10 ps it increased dramatically and at some point the 
equilibration crashed.
I already played around with time-steps, 
atom-types and the density, but couldn't find anything that works to 
keep the molecules inside the box...

Anyone has an idea what I 
did wrong and how I can correct it? Unfortunately there is not much 
about that in the manual and I could not find anything that really 
helped me in the mailing-list archive....

Thanks in advance!!!
Cheers, 
Marianne Schulte