[gmx-users] Re: Actual box size

Justin Lemkul jalemkul at vt.edu
Mon Dec 17 21:44:49 CET 2012

On 12/17/12 3:04 PM, zugunder wrote:
> Justin Lemkul wrote
>> It is calculated correctly, the math is just a bit more complex (see the
>> manual for the equations).  The distance to the box edge is defined the
>> same way, but the two approaches don't necessarily give equally suitable
>> results.  Consider the first case, which produces a rectangular box from
>> an elongated configuration.  If your protein rotates 90 degrees about the
>> z-axis, you will likely violate the minimum image convention, as the box
>> vector along y is insufficient to accommodate the protein.  Problem!  The
>> dodecahedral box is pseudo-spherical and thus, regardless of how the
>> protein rotates, the minimum image convention is not violated.
> So, in other words, the safest way for an elongated protein (with no
> restrictions on rotation) is either a cube or dodecahedron, because only in
> these 2 cases only the longest dimension of the protein is effectively taken
> into account - do I understand it correctly? And obviously, this is true for
> any almost-spherical protein as well...

Yes, it's a reflection on the inherent rotational symmetry of the molecule.  A 
dodecahedron can give you the same periodic distance as a cube, but is much more 
efficient since there are fewer waters.

> And, therefore, any rectangular box, different from a cube will bring to a
> violation of the minimum image convention in case of unrestricted rotation
> of an elongated protein around its shorter axes assuming we set the same -d
> as in case of a cube? So do I get it right that non-cubic rectangular boxes
> are used only in such specific cases with restrictions on rotation?

Or in cases where a rectangle is suitable, i.e. for surfaces or membranes.



Justin A. Lemkul, Ph.D.
Research Scientist
Department of Biochemistry
Virginia Tech
Blacksburg, VA
jalemkul[at]vt.edu | (540) 231-9080


More information about the gromacs.org_gmx-users mailing list