[gmx-users] Small V.S. Large simulation box.

[Mark Abraham]

Yang Ye wrote:
> As being iterated for many times in this list, use of PME shall be only
> applicable to box of certain size, not too small. The default grid size
> for PME is 0.095, with a 0.2nm box, there is only two points. And I
> don't think this can be lifted simply by further decreasing grid size.

I haven't heard of any practical minimum box size necessary for
effective application of PME. I couldn't find mention of it on a quick
search of the gmx-users list. Can you provide a pointer, please? I can't
see any reason for such a minimum from the theory.

You can actually set the grid spacing for Ewald methods to anything
within reason. The default value of the fourierspacing parameter is
0.12nm - which is a maximum grid spacing. Yinghong's solvation shell was
width around 0.2nm, not the box length, which he didn't tell us.

Mark