We know that normal matrices are diagonalizable, but the converse is not true. For example, see here. Since a diagonalizable matrix represents a scaling operation under certain basis, so I wonder what additional geometrical meanings a normal matrix processes to be distinguished from other diagonalizable matrices. In other words, how to geometrically interpret matrices that can be diagonalized by unitary and non-unitary matrices? Thanks!

  • 4
    $\begingroup$ Diagonalizing Matrix for Normal matrices is a orthogonal matrix. Orthogonal matrices represent rotation. Thus Normal matrices first rotates your vector to their co-ordinate system (given by the columns), then scales the components of rotated vector, and then rotates it back. Note that this co-ordinate system is also orthogonal. But for other matrices, the co-ordinate system in which scaling happens is not a orthogonal co-ordinate system. $\endgroup$ – dineshdileep Jan 10 '13 at 7:16
  • $\begingroup$ @dineshdileep, I wonder if "orthogonal matrix/matrices" in your description should be relaxed to "unitary matrix/matrices". $\endgroup$ – Computist Jan 15 '15 at 1:36

So, eigenvectors are the axes where there is a magnification/contraction.

For normal matrices, these axes are orthogonal to each other, thus, a matrix is normal if and only if it represents a linear transform that scales the coordinate axes, (in a suitable chosen orthogonal basis).

A diagonalizable matrix in general is a similar linear transform, (scaling of the coordinate axes), but the basis need not to be orthogonal.

  • 1
    $\begingroup$ So is a non-diagonalizable matrix that has, for example, ones in the first super-diagonal in its Jordan form, represented by a linear transform and that added affine component? $\endgroup$ – bright-star May 30 '13 at 11:24
  • $\begingroup$ That sounds correct. $\endgroup$ – Per Alexandersson Jun 5 '13 at 18:31

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.