AMBER Archive (2007)

Subject: Re: AMBER: Octahedral box orientation

From: David Cerutti (dcerutti_at_mccammon.ucsd.edu)
Date: Tue Aug 21 2007 - 19:07:32 CDT

Update: it seems I have found the solution ot my problem. If anyone is
interested, the Octahedral box in AMBER is rotated away from the
orientation I was hoping for, which would be to have square faces on the
XY/XZ/YZ planes. To get it back into this orientation, one must apply the
transformation matrix:

   cos1 = cos(PI/4.0);
   sin1 = sin(PI/4.0);
   cos2 = sqrt(2.0)/sqrt(3.0);
   sin2 = 1.0/sqrt(3.0);

   R = [
          cos1*cos2 -cos1*sin2 sin1
         -sin1*cos2 sin2*sin1 cos1
              -sin2 -cos2 0.0
       ];

to the coordinates in the AMBER coordinates file. (This is the reverse
rotation as can be found in the ptraj source code actions.c.)

Then, the dimensions given in the AMBER coordinates file apply to the
distance between two hexagonal faces. Multiply those by 2.0/sqrt(3.0) to
obtain the distance between two square faces. Then, all you need to do to
tile the box is translate the unit cell to different lattice positions:

  1 0 0
  0 1 0
  0 0 1
-1 0 0
  0 -1 0
  0 0 -1
  1/2 1/2 1/2
  1/2 -1/2 1/2
-1/2 1/2 1/2
-1/2 -1/2 1/2
  1/2 1/2 -1/2
  1/2 -1/2 -1/2
-1/2 1/2 -1/2
-1/2 -1/2 -1/2

Note that replicating the octahedron at one of the (1/2 1/2 1/2) lattice
positions, then re-imaging all particles as if there were an orthonormal
cell with the same dimensions as the distances between square faces of the
octahedron, will produce a perfectly filled, cubic simulation cell (I will
use this for FFT computations to get electrostatic energies for individual
atoms).

Dave
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu