I am trying to understand what it means exactly for the spectrum of an operator to be closed. I am considering infinite dimensional operators.

Say we know, for example, that the point spectrum of some operator is made up of 4 values, $\{1,2,3,4\}$ does that mean that all numbers in $[1,4]$ are in the spectrum (in the residual or continuous spectrum) since by definition the spectrum is closed? Alternatively, is the $\{1,2,3,4\}$ set considered closed but just disjoint so that we don't necessarily have all of $[1,4]$ in the spectrum.

Can the spectrum be disjoint EDIT: not disjoint, disconnected?

I am afraid that I might be misunderstanding this idea.

I had a problem where the eigenvalues in the point spectrum were dense in $[0,1]$ and we concluded that the spectrum is made up of the closure of $[0,1]$. So I am wondering if the density was the important aspect of that, or if it works that way in general.

  • 2
    $\begingroup$ $\{1,2,3,4\}$ is already closed, so its closure is $\{1,2,3,4\}$ and not $[1, 4]$ $\endgroup$ – mercio Jul 30 '18 at 6:41
  • $\begingroup$ Thank you very much. I appreciate the clarification on that. $\endgroup$ – MathIsHard Jul 30 '18 at 6:42

Let $T(e_1)=e_1,T(e_2)=2e_2,T(e_3)=3e_3,T(e_4)=4e_4,Te_j=e_j$ for all $j>4$ where $\{e_j\}$ is the standard orthonormal basis for $l^{2}$. Then $1,2,3,4$ are eigen values of $T$. If $\lambda \notin \{1,2,3,4\}$ you can solve the equation $Tx-\lambda x=y$. You will get a unique solution for any $y \in l^{2}$. Hence $T-\lambda I$ is a bijection. Automatically the inverse is continuous (by OMT) so $\lambda $ is n at in the spectrum. Thus $\sigma (T)=\{1,2,3,4\}$. [The solution of $Tx-\lambda x=y$ is given by $x_j=\frac {y_j} {1-\lambda} $ if $j >4$ and $x_j=\frac {y_j} {j-\lambda} $ if $j \leq 4$]. Incidentally, you are using the word disjoint for disconnected. The statement $A$ is disjoint has no meaning.

  • $\begingroup$ Thank you very much for the help. If I am understanding correctly then, the set $\{1,2,3,4\}$ is closed and it is okay that the spectrum is the union closed sets? $\endgroup$ – MathIsHard Jul 30 '18 at 6:07
  • $\begingroup$ You are right. I meant disconnected... $\endgroup$ – MathIsHard Jul 30 '18 at 6:09
  • 1
    $\begingroup$ Yes, spectrum need not be a connected set. $\endgroup$ – Kavi Rama Murthy Jul 30 '18 at 6:09

The answer of Kavi Rama is not O.K.

Example: Let $T: l^2 \times \mathbb C^4 \to l^2 \times \mathbb C^4$ be defined by

$T((\xi_1, \xi_2, \xi_3,...),(z_1,z_2,z_3,z_4))=(0, \xi_1, \xi_2, \xi_3,...),(z_1,2z_2,3z_3,4z_4))$.

Then the point spectrum of $T$ is $\{1,2,3,4\}$, but the spectrum of $T$ is given by

$$\{ \lambda \in \mathbb C: |\lambda| \le 1\} \cup \{1,2,3,4\}.$$

  • $\begingroup$ Thank you for your answer. What part of Kavi Rama's answer and your's is different? $\endgroup$ – MathIsHard Jul 30 '18 at 6:47
  • $\begingroup$ Oh, you mean that you used a shift so that your Tx isn't dense in $l^2$? $\endgroup$ – MathIsHard Jul 30 '18 at 6:48

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.