4
$\begingroup$

Let $A$ be a Banach algebra. Is there a Banach algebra $B$ which contains $A$ but the spectrum of each elements of $B$ has empty interior(as a subset of $\mathbb{C}$)?

The motivation comes from the fact that the spectrum of elements in a smaller algebra possibly loses its interior when we compute its spectrum in a larger algebra.(Rudin, Functional analysis)

$\endgroup$
1
  • 1
    $\begingroup$ In general, if you look for positive results (about shrinking the spectrum of certain elements by passing to a bigger algebra) then one place to start looking up some of the old results is this paper of C. J. Read: ams.org/journals/tran/1984-286-02/S0002-9947-1984-0760982-0 (Warning: I think that in this paper Charles uses the terminology "residual spectrum" in a way that is non-standard and conflicts with the accepted current terminology) $\endgroup$
    – Yemon Choi
    Dec 16, 2019 at 2:49

1 Answer 1

9
$\begingroup$

The answer is no, in general. Here is a counterexample:

Let $A$ be the algebra of bounded linear operators on $\ell^2(\mathbb{N})$, and let $a \in A$ be the left shift on $\ell^2(\mathbb{N})$. Then the spectrum of $a$ is the closed unit disk $\overline{D}$, and the point spectrum of the operator $a$ is the open unit disk $D$.

Now we note that the notion "point spectrum" - which is defined for operators - can be translated into a notion that makes sense in Banach algebras: For each $\lambda$ in the point spectrum $D$ of $a$ there exists an element $a_\lambda \in A \setminus \{0\}$ such that $(\lambda - a) a_\lambda = 0$. Indeed, let $x_\lambda \in \ell^2(\mathbb{N})$ be an eigenvector of the operator $a$ and choose an arbitrary non-zero functional $\varphi_\lambda$ on $\ell^2(\mathbb{N})$. Then the operator $a_\lambda := \varphi_\lambda \otimes x_\lambda \in A$, given by $$ a_\lambda x = \langle \varphi_\lambda, x \rangle x_\lambda \qquad \text{for } x \in \ell^2(\mathbb{N}), $$ satisfies $(\lambda - a)a_\lambda = 0$.

Hence, if $B$ is a unital Banach algebra that contains $A$ as a subalgebra, we still have $(\lambda - a)a_\lambda = 0$ in $B$, so $\lambda - a$ cannot be invertible in $B$ for any $\lambda \in D$.

$\endgroup$
3
  • $\begingroup$ Thank you very much for your very interesting answer. $\endgroup$ Dec 15, 2019 at 20:45
  • $\begingroup$ @YemonChoi: I'm not sure I follow: the right shift has empty point spectrum, so I indeed mean the left shift (as specified in the post). $\endgroup$ Dec 16, 2019 at 7:28
  • $\begingroup$ @JochenGlueck Sorry, I was reading the question while ill+tired and made a stupid misreading. Of course what you have written is fine $\endgroup$
    – Yemon Choi
    Dec 16, 2019 at 13:41

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.