5
$\begingroup$

According to some lecture notes I am reading, "it is not so difficult to find an example of a commutative unital Banach algebra which is not isomorphic to $C(X)$ for some compact Hausdorff space $X$"...

Well I was thinking about using the disk algebra $\mathcal{A}$ of continuous functions on the disk $D = \{z \mid |z| \leq 1\}$ which are holomorphic on the interior of $D$, with the sup norm.

This should be a commutative unital Banach algebra. However, since it is the beginning of the chapter about the Gelfand transformation, I would like to prove that $\mathcal{A}$ is not isomorphic to $C(X)$ without using Gelfand's result.

What other tools could I use to prove this fact?

$\endgroup$
2
  • $\begingroup$ Fourier series? }:) $\endgroup$
    – logarithm
    May 9 '19 at 17:50
  • $\begingroup$ @logarithm Could you elaborate? I would be interested to see a proof of this that uses Fourier series! $\endgroup$ May 10 '19 at 9:20
5
$\begingroup$

Here is an argument. Assume that $\pi:\mathcal A\to C(X)$ is a unital Banach algebra isomorphism. Let $f\in \mathcal A$. Put $g=\pi^{-1}(\overline{\pi(f)})$ (that is, map $f$ to $C(X)$, conjugate it, and come back). Now $$ \pi(gf)=\pi(g)\pi(f)=|\pi(f)|^2\geq0. $$ Then, for any $r>0$, $\pi(gf+r+is)=|\pi(f)|^2+r+is$ takes values at distance $r$ or more from $0$, so $gf+r+is$ is invertible. This says that $-r+is$ (we can write a plus since $s$ was arbitrary) is not in the image of $gf$. In other words, $\operatorname{Re}(gf)\geq0$. You can see proof here (applied to $-gf$, and it may require a rotation if $gf(0)$ is not real) that then $gf$ is constant. In other words, there exists $c\in\mathbb R$ such that $gf=c1$. Then $$ |\pi(f)|^2=\pi(gf)=\pi(c1)=c1. $$ Now $f$ was arbitrary and $\pi$ is onto, so every $h\in C(X)$ has $|h|^2$ constant. This can only happen if $X$ consists of a single point, and in that case $\mathcal A$ would be one-dimensional, a contradiction.

$\endgroup$
3
  • $\begingroup$ Nice proof, thanks! Just a thing in your post, you might have mixed $fg$ with $gf$ when you say "$\pi(fg + r1)$ is invertible and so $fg + r1$ is invertible". $\endgroup$ May 10 '19 at 9:22
  • $\begingroup$ Indeed, edited. Note that $\mathcal A$ is commutative, though. $\endgroup$ May 10 '19 at 14:25
  • $\begingroup$ Also, the reasoning was not complete as I had written it, so I edited it that, too. $\endgroup$ May 10 '19 at 17:25
2
$\begingroup$

Every $C^*$ algebra of dimension at least $2$ must have zero divisor but the disc algebra has no zero divisor.

Proof of existence of zero divisor is based on decomposition $x=x^+ -x^-$ for a self adjoint element $x$. This decomposition is regarded as classical decomposition $f=f^+ -f^-$ with $f^+.f^-=0$ for real valued function $f\in C(X)$.

$\endgroup$

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.