[gmx-users] Freeze all beyond a sphere

[Erik Lindahl]

Tivadar Orban wrote:

>Dear gromacs users,
>I want to see the differences, reliability, of a system when I simulate it
>with the following approch:
>1. Simulate (1ns) without any frozen group (this is all right)
>2. Freeze all atoms from the system that are beyond a sphere with the center
>fixed on a specific atom (in my case the center is FE from the heme), and a
>given ray.
>3. I want to take different values of the ray and perform on each of them a
>simulation.
>4. after that compare the results obtained with diff ray spheres.
>For this I wrote a short program that takes out all atom numbers (from the
>.gro file that I use before running grompp MD and after mdrun MD) where the
>distance between the atom A(i) and Fe is bigger than the chosen ray.
>The atom numbers are added in the index.ndx file after a group named
>RIsBigger.
>The mdp file that I have has the
>
> 
>freezegrps = RIsBigger
>Freezedim = Y Y Y
>
>For mdrun I specify the new index.ndx file.
>
>After running I get the following error:
>Water molecule 4756 cannot be settled.
>
>This is the first one, the second one:
>For a specific ray the simulation went smoothly, but after simulation I saw
>that the .gro file after mdrun instead of having numerical values has NaN,
>not for all rows but a lot of them.
>
>Could anybody give me a hint?
>

Hi,

The problem is almost certainly that you are only freezing part of some 
of your water molecules. The SETTLE algorithm constrains the bond 
lengths of the waters, but if e.g. the hydrogens can move while the 
oxygen is completely frozen this might not work.

Try to treat each water molecule as a group when you determine if the 
atoms in it should be frozen or not; you can either calculate the 
geometrical center, the mass center, or simply use the position of the 
oxygen.

Cheers,

Erik