If we consider $X\neq\{0\}$ to be a complex Banach space then the product $X\times X$ is a Banach space with the norm $\|(x,y)\|=\|x\|+\|y\|$. $T(x,y)=(x + y,x - y)$ is then a bounded linear operator on $X\times X$. Find $\sigma(T)$ and the subsets $\sigma_p(T),\sigma_c(T), \sigma_r(T)$, where $\sigma(T)$ is the spectrum of operator $T$, and the subsets are the point spectrum, the continuous spectrum, and the residual spectrum, respectively.

I am stuck at this problem. I know that the spectrum is supposed to be the set of eigenvalues, which I am supposed to find with matrices I believe. So I'd have to compute $T-\lambda I$ but I am not sure how the matrix of $T$ looks in this space? Then of course I'd have to compute the subsets which seems even harder.

  • 1
    $\begingroup$ Have you studied Banach space theory/operator theory? The question has nothing to do with matrices. $\endgroup$ – Kavi Rama Murthy Feb 10 '19 at 11:50
  • $\begingroup$ Yes I have, but I thought that the spectrum is calculated through the matrix of $T$ with respect to some basis? @KaviRamaMurthy $\endgroup$ – mandella Feb 10 '19 at 11:58
  • $\begingroup$ Your line of thinking completely fails in the case of infinite dimensional spaces. I am sorry to say this, but if I post an answer to this question you may not be able to understand it. $\endgroup$ – Kavi Rama Murthy Feb 10 '19 at 12:03
  • $\begingroup$ Ah, I see. Of course I have the literature but I made a mistake. Maybe I will be able to follow the answer, so any help is appreciated. @KaviRamaMurthy $\endgroup$ – mandella Feb 10 '19 at 12:09
  • 1
    $\begingroup$ @KaviRamaMurthy Why are you saying that there are no eigenvalues? $\lambda=\pm \sqrt{2}$ and $y=(\pm\sqrt{2}-1)x$ are the solution of $T(x, y)=\lambda(x, y)$, aren't they? $\endgroup$ – Andrey Gorbunov Feb 10 '19 at 13:42

Consider the matrix $$\begin{bmatrix}1 & 1 \\ 1 & -1 \end{bmatrix}$$ This isn't the matrix of $T$ w.r.t. to any basis but it can still be useful because formally$$T\begin{bmatrix}x \\ y \end{bmatrix} = \begin{bmatrix}1 & 1 \\ 1 & -1 \end{bmatrix}\begin{bmatrix}x \\ y \end{bmatrix}, \quad \forall\begin{bmatrix}x \\ y \end{bmatrix} \in X \times X $$

Its characteristic polynomial is $\lambda^2-2$ so the eigenvalues are $\pm \sqrt{2}$.

Check that for $\lambda\ne \pm\sqrt{2}$ we have $$(T-\lambda I)^{-1}\begin{bmatrix}x \\ y \end{bmatrix} = \begin{bmatrix}1-\lambda & 1 \\ 1 & -1-\lambda \end{bmatrix}^{-1}\begin{bmatrix}x \\ y \end{bmatrix} = -\frac1{\lambda^2-2}\begin{bmatrix}\lambda+1 & 1 \\ 1 & \lambda-1 \end{bmatrix}^{-1}\begin{bmatrix}x \\ y \end{bmatrix} $$ and this is a bounded linear map so $\lambda \notin \sigma(T)$.

On the other hand, for $\lambda = \pm\sqrt{2}$, we have $$T\begin{bmatrix}(1\pm\sqrt{2})y \\ y\end{bmatrix} = \pm\sqrt{2}\begin{bmatrix}(1\pm\sqrt{2})y \\ y\end{bmatrix}$$ for any $y \in X, y \ne 0$ so $\pm \sqrt{2} \in \sigma_p(T)$.

Therefore $\sigma(T) = \sigma_p(T) = \{\pm\sqrt{2}\}$ and $\sigma_r(T) = \sigma_c(T) = \emptyset$.

We have \begin{align} \left\|(T-\lambda I)^{-1}\begin{bmatrix}x \\ y \end{bmatrix}\right\| &= \frac1{|\lambda^2-2|}\left\|\begin{bmatrix}(\lambda+1)x+y \\ x+(\lambda-1)y \end{bmatrix}\right\| \\ &= \frac1{|\lambda^2-2|}(\|(\lambda+1)x+y \| + \|x+(\lambda-1)y\|)\\ &\le \frac1{|\lambda^2-2|}(|\lambda+1|\|x\|+\|y\| + \|x\|+|\lambda-1|\|y\|)\\ &\le \frac{\max\{1+|\lambda+1|, 1+|\lambda-1|\}}{|\lambda^2-2|}(\|x\|+\|y\|)\\ &= \frac{\max\{1+|\lambda+1|, 1+|\lambda-1|\}}{|\lambda^2-2|}\left\|\begin{bmatrix} x \\ y\end{bmatrix}\right\|\\ \end{align} so $(T-\lambda I)^{-1}$ is bounded. However, this also follows from the Bounded Inverse Theorem once you verify that $(T-\lambda I)^{-1}$ is indeed the algebraic inverse of the bounded map $T-\lambda I$.

| cite | improve this answer | |
  • $\begingroup$ So the matrix can be used to express $T$ although it isn't $T$ itself? And you conclude that the residual and continuous spectrum are empty because for $\lambda \neq \pm \sqrt{2}$ we get that $\lambda\notin\sigma(T)$ right? $\endgroup$ – mandella Feb 11 '19 at 9:05
  • 1
    $\begingroup$ @mandella Yes, the matrix is merely a tool to guess what points could be in the spectrum. Yes, and $\pm\sqrt{2}$ are eigenvalues. $\endgroup$ – mechanodroid Feb 11 '19 at 9:07
  • $\begingroup$ When you calculate $(T-\lambda I)^{-1}$, should the matrix at the end when you find the inverse have the ${-1}$? and also why is it bounded? $\endgroup$ – mandella Feb 11 '19 at 9:11
  • 1
    $\begingroup$ @mandella I think the matrix is ok, notice the minus in front. I added the explanation why it's bounded. $\endgroup$ – mechanodroid Feb 11 '19 at 9:22
  • 1
    $\begingroup$ @mandella I believe $1\pm \sqrt{2}$ is correct because $$\begin{bmatrix}1 & 1 \\ 1 & -1 \end{bmatrix} \begin{bmatrix}1\pm\sqrt{2} \\ 1\end{bmatrix} = \begin{bmatrix}2\pm\sqrt{2} \\ \pm\sqrt{2}\end{bmatrix} = \pm\sqrt{2} \begin{bmatrix}1\pm\sqrt{2}\\ 1\end{bmatrix}$$ $\endgroup$ – mechanodroid Feb 11 '19 at 9:35

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.