Actually my question is about calculating the spectrum of any bounded operator. I know that if we have Banach spaces $\mathbb{E}$ and $\mathbb{F}$ over the field $\mathbb{K} = \mathbb{R}$ or $\mathbb{C}$, and a bounded operator $T: \mathbb{E} \rightarrow \mathbb{F}$, then the spectrum of $T$ is the set $$\sigma (T) = \{\lambda \in \mathbb{K}: \lambda I - T \text{ is not invertible}\}.$$ Also, I know the theorem stating that $\sigma (T) \subseteq B[0;||T||]$. I've seen similar questions here referencing things like point or residual spectrums, but I'm just thinking of the ordinary spectrum defined above.

It's clear that all eigenvalues of $T$ are in its spectrum, because if $\lambda$ is an eigenvalue of $T$, $\lambda I - T$ is going to have a kernel that is not $\{0\}$. But otherwise I've got no idea how to calculate the rest of the spectrum. My particular question is probably quite easy: finding the spectrum of the operator $R: \ell^2 (\mathbb{C}) \rightarrow \ell^2 (\mathbb{C})$ given by $$R(z_1, z_2, z_3, \dots) = (0, z_1, z_2, \dots).$$

$R$ has no eigenvalues and $-R = 0T-R$ is invertible so $0 \notin \sigma(R)$. And $||R||=1$, so if $\lambda \in \sigma(R)$ we must have $|\lambda| \leq 1$, but otherwise I don't think I can say more.

  • $\begingroup$ Be aware, that R:l^2->l^2 is not invertible, since you can't find a corresponding sequence (z_1,z_2,....) so that R(z_1,z_2,....)=(b,z_1,z_2,....) in l^2 with b not zero. The map R is not surjective. 0 is actually in the spectrum of R. But I have to admit, that calculating the whole spectrum for operators is not that fun ;) $\endgroup$ – MrMatzetoni May 9 '18 at 10:15
  • $\begingroup$ @MrMatzetoni, That's a good point about 0. But it's just occurred to me that the adjoint of R is L, the left-shift operator, and it's quite easy to show that the spectrum of L is the unit disc. Shouldn't that mean, since the spectrum of T is the conjugate set of the spectrum of its adjoint T* (also not hard to show, I think), that the spectrum of R is actually the conjugate set of the unit disc -- and therefore the whole of the unit disc, too? $\endgroup$ – SPS May 9 '18 at 10:24
  • $\begingroup$ Basically, as you know that R has no eigenvalues, thus (lambda I - R) is injective for all lambda, you would have to check the surjectivity of all operators (lambda I - R) with |lambda|<=1. Your thought about the adjoint seems legit to me, this would mean sigma_p(R)={} and sigma(R)={lambda | |lambda|<=1 } $\endgroup$ – MrMatzetoni May 9 '18 at 10:32
  • $\begingroup$ You might check this out: math.ubc.ca/~feldman/m511/spectralExamples.pdf $\endgroup$ – MrMatzetoni May 9 '18 at 10:36
  • $\begingroup$ OK, that seems to be saying that the 'point' spectrum of R is empty, but the 'residual' spectrum is the unit disc, although the open one. But then since the spectrum is closed, the 'full' spectrum must be the disc, surely. But I'm a little uncertain as to the role of these other kinds of spectrums. Thanks for the tips -- and maybe an answer will make things clearer too. $\endgroup$ – SPS May 9 '18 at 10:49

Consider $S:=R^*$, then $S(z_1, z_2, z_3, \dots) = (z_2, z_3, z_4, \dots).$

If $ | \mu| <1$ and $z=(1, \mu, \mu^2,...)$, then $Sz= \mu z$, hence

$\{\mu: |\mu|<1\} \subseteq \sigma(S)$. Since $||S||=1$ and $\sigma(S)$ is closed, we derive

$\{\mu: |\mu| \le 1\} =\sigma(S)$.

Consequence: $\{\mu: |\mu| \le 1\} =\sigma(R)$.

| cite | improve this answer | |
  • $\begingroup$ That makes sense, thanks. $\endgroup$ – SPS May 9 '18 at 11:17

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.