[gmx-users] maximum cut-off in Minimum image convention
Justin A. Lemkul
jalemkul at vt.edu
Wed May 26 13:53:53 CEST 2010
Francisco Garcia wrote:
> Dear users,
> I wish to ask a question regarding the maximum cut-off in the minimum
> image convention (MIC).
> According to my understanding, MIC ensures that the separation
> between two particles i and j along each coordinate axes, namely,
> xi- xj, yi-yj, and zi-zj, is in the interval (-0.5L, 0.5L), where L is the box
> length (assumed the box to be cubic). This implies that
> distance rij can vary between 0 and 0.5L*sqrt(3).
> In order words the maximum separation should be less than half the diagonal:
> for example if i is located at (0,0,0) and j is located at the corner
> (0.49L, 0.49L, 0.49L).
> However, I read a couple of text books and online notes that claim
> that if r > 0.5L, then MIC is violated.
> Specifically, this means that 0.5L < rij < 0.5L*sqrt(3) violates MIC.
> I find this confusing. I have been grinding hard to understand this
> point but I just
> cannot figure it out.
> In short my claim is that in MIC, rij can be at most half the diagonal
> length whereas
> the textbook says that rij can be at most half the box length.
> If I am wrong, can someone explain the flaw in my argument?
The box diagonal is the longest dimension across the unit cell, so indeed it
would seem that you can have r > 0.5L. Simplify your problem to 2 dimensions
and the answer becomes very clear. If you have a particle in the center of this
2-D lattice, and you have a cutoff > 0.5L, then you are counting interactions
with other particles within the central unit cell and duplicating at least some
of these interactions in the periodic image. See, for instance:
Extend the radius in that image ever so slightly, and particle 1 is going to
interact with *both* particles 4 and 7, the latter being the periodic image of
the same particle. Several textbooks also discuss this phenomenon in depth.
Justin A. Lemkul
ICTAS Doctoral Scholar
Department of Biochemistry
jalemkul[at]vt.edu | (540) 231-9080
More information about the gromacs.org_gmx-users