[gmx-users] equilibriation questions

[Erik Lindahl]

Hi,


On Sep 2, 2005, at 6:37 PM, Ramachandra Rao Gullapalli wrote:

> Dear GMX ers
> I have built a bilayer from scratch and now i am trying to  
> equilibriate the
> same. I had a few questions regarding the equilibriation process  
> itself.
>
> 1) Do i need to start by using a NVT ensemble or an NPT right away?
>

Start with NVT and monitor the pressure (XX/YY/ZZ separately in the  
energy file).

When the pressure has stabilized you can turn on NPT.

> 2) If i need a NPT ensemble, do i use isotropic, semiisotropic or  
> anisotropic
> type of pressure coupling?


Semiisotropic or anisotropic would be my recommendation. For very  
small systems (4x4 nm or so) semiisotropic avoids one dimension  
growing while the other one shrinks to less than 2*cutoff, but for  
larger systems you probably won't see those fluctuations.

>
> 3) For an equilibriation, does it matter that i use the cutoffs  
> method instead
> of PME to deal with electrostatics?

That should be fine. You can always argue that PME increases/ 
decreases ordering of the headgroups, but that's a secondary effect  
compared to the hydrocarbon tail organization.

>
> 4) I have this problem that the system which i am running seems to  
> expand in 10
> -20 picoseconds, causing a empty space to appear between the 2  
> leaflets of the
> bilayer.  Why does this problem occur? How can i rectify it?

Probably very large pressures. Lipids are large molecules, so you  
need to massage them carefully during equilibration. It might even be  
worth trying softcore initially, not for the free energy but to get  
smoother interactions.
>
> 5) How long do i need to run an equilibriation? What are the  
> criteria for a
> successful equilibriation of a lipid bilayer?

Well, ten years ago people equilibrated for 10ps, now it's probably  
closer to 10ns...

Monitor properties like potential energy, pressure volume and area of  
the system, and wait until the have stabilized enough. Translating  
"enough" to number of steps is unfortunately your call :-)


Cheers,

Erik