2
$\begingroup$

I am having trouble proving that Hermitian Matrices ($A = A^{*}$) are unitarily diagonalizable ($A = Q^{*}DQ$, where Q is a unitary matrix, $QQ^{*} = I$ and D is a diagonal matrix). I also know that given an hermitian operator T over a finite diminsional vector space V, V has an orthonormal basis which consists of eigenvectors of T.

Any suggestions?

Edit: Following Git Gud's suggestion:

We shall consider the Hermitian operator $L_A$. Let E be the standard basis of V. Also, by hypothesis, we have that there exists an orthonormal basis B for V which consists of eigenvectors of $L_A$. Since the basis consists of only eigenvectors, the basis contains $dim(V)$ linearly independent vectors, so $[L_A]_B$ is diagonal.

We know that $[L_A]_E = A$. Then, $$A = [L_A]_E = Q^{-1}DQ$$ and since Q must be unitary, then $Q^{-1} = Q^*$, by which we have that $$A = Q^*DQ$$

$\endgroup$
  • 1
    $\begingroup$ The solution follows from your last sentence. $\endgroup$ – Git Gud Jun 6 '14 at 19:32
  • $\begingroup$ I have edited the original post. Is the argument made about the Q matrix 'strong enough'? $\endgroup$ – lalaman Jun 6 '14 at 20:10
  • $\begingroup$ I think it's fine if you explain what $Q$ is. You just typed it with no explanation. $\endgroup$ – Git Gud Jun 6 '14 at 20:26
4
$\begingroup$

If you know, and therefore do not need to prove that that there exists a basis of orthonormal vectors of $A$ then this is equivalent to stating that $A$ is unitarily diagonalizable, with the matrix $Q$ having as columns the orthonormal eigenvectors of $A$...

But there is also another way - if you may make use of Schur's theorem... Every $A\in \mathbb{C}_{n \times n}$ is unitarily similar to an upper triangular matrix. So we can find unitary matrix $Q$ so that $Q^*AQ$ is upper triangular. BUT since $A$ is Hermitian, so is $Q^*AQ$ (you can prove this: $(Q^*AQ)^*=Q^*A^*Q=Q^*AQ$), and therefore $Q^*AQ$ must be diagonal.

$\endgroup$
  • $\begingroup$ Underrated answer. $\endgroup$ – Linfeng Li Dec 7 '16 at 8:38

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.