1

Assume that $A, B\in\mathbb{R}^{n\times n}$ are both diagonalizable then $AB=BA$ implies that $A$ and $B$ are Simultaneously Diagonalizable.

I have seen a proof of this statement that goes as follows:

Let $u$ be an eigenvector of A such that $Au=\lambda u$ then consider $ABu=BAu=\lambda Bu$ which implies that $Bu$ is also an eigenvector of A with eigenvalue $\lambda$. The author proceeds to claim that this implies that $Bu=\gamma u$ that is that $Bu$ has to be a multiple of $u$ concluding that $u$ is also an eigenvector of B. The issue I am having is how does one conclude that $Bu=\gamma u$? Why can't $Bu$ be any other vector of A. The only way I see this happening is if the multiplicity of every eigenvalue of $A$ is 1 and therefore the statement follows. Can anyone offer any clarification as to how $Bu=\gamma u$ is guaranteed without the multiplicity restriction?

Darsen
  • 3,549
nvm
  • 1,296
  • 1
    It is not. This proof assumes one-dimensional eigenspaces, as you suggested. The proof in the general case is far more subtle. – Ted Shifrin Oct 09 '20 at 04:41
  • @TedShifrin Can you provide me a link of where the general case is proven? – nvm Oct 09 '20 at 04:43
  • 3
    You should try doing a careful search of the questions on this site. You might find, for example, this. – Ted Shifrin Oct 09 '20 at 04:47
  • @TedShifrin: the proof in general is not “far” more subtle. We have found that if $u$ is a $\lambda$-eigenvector of $A$, then $Bu$ is also a $\lambda$-eigenvector of $A$. From that we can conclude that $B(V_\lambda) \subseteq V_\lambda$ where $V_\lambda$ is the $\lambda$-eigenspace of $A$. Hence $B$ restricts to an operator on $V_\lambda$. Every $B$-eigenspace inside $V_\lambda$ is also an $A$-eigenspace (with eigenvalue $\lambda$). So far we have only used that $AB=BA$, finally apply the diagonalisation hypothesis to see that the whole space is a sum of simultaneous eigenspaces. – Joppy Oct 09 '20 at 11:48
  • 1
    @Joppy: Depending on what one knows, the proof that the restriction of a diagonalizable operator to any invariant subspace is likewise diagonalizable is not trivial. – Ted Shifrin Oct 09 '20 at 13:26
  • Did the link TedShifrin provided answer your question? – ViktorStein Dec 17 '22 at 18:45

0 Answers0