In a post to the relax-devel mailing list (https://mail.gna.org/public/relax-devel/2006-10/msg00136.html, Message-id: <7f080ed10610252045r3dcbba38q58aa53142af96e5a@xxxxxxxxxxxxxx>) I discussed an idea about representing the Brownian rotational diffusion tensor superimposed onto the 3D molecular structure. For a better understanding of the representation, I'll create a mini-tutorial on diffusion tensors to explain the different notations, representations, etc. This text should be useful for future reference.
The best way to think of the tensor is as a geometric object. Isotropic diffusion is diffusion as a perfectly round sphere. Prolate axial symmetry is where one axis has been stretched - i.e. the shape of a rugby/AFL/gridiron ball. Oblate axial symmetry is where one axis has been squashed - i.e. the shape of a discus. These two geometric objects are known as spheroids. Finally fully anisotropic diffusion is the shape of an ellipsoid, you can think of it as a partially flat rugby/AFL/gridiron ball that you have sat on. These geometric objects will form the representation of the tensor superimposed onto the 3D structure.
For directly comparing diffusion tensors it is the eigenvalues and eigenvectors, the geometric description, which is important. The three eigenvectors of the ellipsoid are the three axes of the tensor (two are undefined in the spheroids and all three are undefined in the sphere). The ellipsoidal eigenvalues Dx, Dy, and Dz measure the rate of rotational diffusion of a vector parallel to the corresponding axis (or eigenvector). In the spheroids, Dx = Dy = Dper and Dz = Dpar. In the sphere, Dx = Dy = Dz = Diso.
The eigenvalues tell you the geometric length of the axes of the ellipsoid and if two are equal you have a spheroid, and if all three are equal you have sphere. There are a few other ways of describing the geometric objects though. The method I prefer is the isotropic factor Diso, the anisotropic factor Da, and the rhombic or asymmetric factor Dr. You can physically picture these as follows. Diso is the starting spherical structure and is equal to 1/(6tm). Da is the amount you then stretch (Da > 0) or squash (Da < 0) this sphere. Dr is then the amount you stretch along the y dimension. This approach simplifies the parameterisation of the model. All three values are non-zero in the ellipsoid. In the spheroids, Dr = 0. In the sphere Da = 0 and Dr = 0. Hence parameters drop out of the model as it simplifies! Dratio is another way of expressing Da, but I don't like it much as it isn't a good way to describe the ellipsoid.
To see all the relationships between the geometric parameters, have a look at the documentation for the 'diffusion_tensor.init()' user function in relax, either in the current 1.2.7 version by typing 'help(diffusion_tensor.init)' or in the current relax manual at http://www.nmr-relax.com/docs.html or http://download.gna.org/relax/manual/relax.pdf.
Edward