Weaker version of the spectral theorem for finite dimensional vector spaces

54 Views Asked by At

The spectral theorem for euclidean spaces is very well known. Let me establish the notation.

Let $V \neq \{0\}$ be a finite dimensional vector space over $\mathbb{R}$ endowed with a symmetric, positive definite, bilinear form $p:V \times V \to \mathbb{R}$.

Suppose $f:V \to V$ is an endomorphism such that for all $v,w \in V$, it is true that $p(f(v),w)=p(v,f(w))$, so $f$ is symmetric (self adjoint).

So, what the spectral theorem says is that there exists an orthonormal basis of $V$ made of eigenvectors of $f$.

Consider this claim:

"If $V$ and $f$ are as above, then $f$ is diagonalizable (equivalently, $V$ admits a basis of eigenvectors of $f$)"

This is much weaker since it does not say anything about $f$ being orthogonally diagonalizable.

Obviously, this claim is pointless if compared with the spectral theorem, because it is just an uninteresting corollary.

But the question I have is:

can the claim be proved without using the spectral theorem? In other words, is there a nice and elementary proof for this fact which does not use the same techniques for proving the spectral theorem?

The proofs I have seen for the spectral theorem do not use anything too fancy, but I just wanted to know because I read this question from someone taking a first course on linear algebra and I could not give an answer.