2
$\begingroup$

The following question came to my mind while reading $\S 90.$ Minimax principle from Paul R. Halmos's Finite Dimensional Vector Spaces (second edition), and the exercises therefrom.

If $\lambda$ is an eigenvalue of a bounded linear operator $A$ on a not-necessarily finite-dimensional inner product space, does it follow that $\vert \lambda \vert \leq \Vert A \Vert$?

The norm $\Vert \cdot \Vert$ of a bounded operator on a non-trivial inner product space is defined in $\S 88.$ Expressions for the norm as follows.

$\Vert A \Vert = \sup \left\{ \frac{\Vert Ax \Vert}{\Vert x \Vert}: \Vert x \Vert \neq 0 \right\}.$

I believe that the answer to the above question is "yes". Proof: If $y$ is an eigenvector for eigenvalue $\lambda$, then $Ay = \lambda y$. Thus $\Vert Ay \Vert = \Vert \lambda y \Vert = \vert \lambda \vert \cdot \Vert y \Vert$. It follows that $\frac{\Vert Ay \Vert}{\Vert y \Vert} = \vert \lambda \vert \leq \Vert A \Vert$.

However, I am little unsure about this conclusion due to my general lack of comfort in case of infinite-dimensional spaces, and would appreciate help. Thanks.

$\endgroup$
2
  • 2
    $\begingroup$ The result holds. In general the norm on a subspace must be $\le $ the norm on the ambient space. $\endgroup$
    – copper.hat
    Oct 13 '21 at 3:47
  • 2
    $\begingroup$ Your argument is correct - and basically the way I would have shown it. $\endgroup$ Oct 13 '21 at 4:07
2
$\begingroup$

More generally, if $T: V \to V$ is a bounded operator on a normed space $V$ and if $\lambda$ is an eigenvalue of $T$ then $\|T\|\ge |\lambda|$.

Proof: Choose a unit vector $v \in V$ such that $Tv = \lambda v$ (just normalise an eigenvector associated to $\lambda$). Then $$|\lambda| = \|\lambda v\|_V = \|Tv\|_V \le \|T\|$$ and we are done!

$\endgroup$

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.