I understand that a normal operator is an operator such that

$$ AA^\dagger = A^\dagger A $$ where $\dagger$ is the conjugate transpose.

However, what is the most intuitive way to "characterise" this? For example,

  1. $SO(3)$ is the group of rotations in $\mathbb R^3$
  2. A unitary matrix is one that represents an isometry
  3. A hermitian matrix is a generalization of symmetric, and is the "nicest" (diagonalizable, real eigenvalues, etc) of all matrices over $\mathbb C$

I was hoping for some sort of intuitive explanation of why I would care about normal matrices over $\mathbb C$ (maybe other reasons than the spectral theorem? Something more fundamental / geometric perhaps)

  • 2
    $\begingroup$ Normal operators are the ones which can be diagonalized via an unitary. In this case they behave like finite sequences of numbers (namely, the diagonal), and they have very well-behaved analytical properties. $\endgroup$ Aug 10, 2016 at 16:02
  • $\begingroup$ @LuizCordeiro - so the reason I would choose to care is because there's a unitary change of basis that makes them diagonalizable? $\endgroup$ Aug 10, 2016 at 16:07
  • $\begingroup$ Yes. Moreover, since an unitary change of basis preserves the inner product, the geometry of the vector space is also unchanged, which can be quite useful. $\endgroup$ Aug 10, 2016 at 16:10

2 Answers 2


The term "normal" here refers to orthogonality. Geometrically, a normal operator on $\mathbb C^n$ represents scaling by possibly different (complex) factors along different axes of an orthogonal coordinate frame. This, I think, is more intuitive than a Hermitian operator (which, of course, is also normal).

  • 3
    $\begingroup$ A Hermitian operator is simply a normal operator with real eigenvalues (scaling factors). $\endgroup$
    – celtschk
    Aug 10, 2016 at 17:46
  • $\begingroup$ @celtschk LOL. Of course you are right. $\endgroup$
    – user1551
    Aug 10, 2016 at 17:48

A square complex matrix $A$ is normal iff it has an orthonormal basis of eigenvectors. That's why you would care about it. Unitary and selfadjoint matrices are special cases. Normal is the most general case.


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.