2
$\begingroup$

It's known that every injective Banach space is of the form $C(M)$ where $M$ is a compact, Hausdorff, extremally disconnected topological space.

Let $X$, $Y$ be two injective Banach spaces such that,

  • There exists an into linear isometry $i:X\to Y$, and
  • There exists an into linear isometry $j:Y\to X$.

Q: Does there exist a surjective isometry $X\to Y$?

Dually, if $M_X$, $M_Y$ are compact Hausdorff extremally disconnected spaces such that there exist surjective continuous maps $j^{*}:M_X\to M_Y$, $i^{*}:M_Y\to M_X$, then

$\textbf{Q}^*\textbf{:}$ Are $M_X$ and $M_Y$ homeomorphic?


Edit(2024-07-05): I just realized that the answer is negative. A counterexample is the pair $\ell^{\infty}$ and $C(M_0)$, where $M_0$ is the Gleason cover of the Cantor set.

It is an old result of James (http://doi.org/10.1090/S0002-9939-1955-0076302-0) that $C(K)$ contains an isometric copy of $\ell^{\infty}$ if $K$ is extremally disconnected. So, there is an isometry $i:\ell^{\infty}\to C(M_0)$.

Conversely, there exists an isometry $\ell^2\to\ell^{\infty}$ defined by $x\to (\langle x,f_n\rangle )_{n\in\mathbb{N}}$ where $(f_n)$ is a norming subset of the dual unit ball. The injective envelope of $\ell^2$ is $C(M_0)$. Thus, there exists an isometry $j:C(M_0)\to\ell^{\infty}$.

Lastly, $C(M_0)$ and $\ell^{\infty}$ are (isomorphic as Banach spaces, but) not linearly isometric, since otherwise $M_0$ and $\beta\mathbb{N}$ would be homeomorphic, one of which contains no open point sets and the other contains a dense subset of open points.

$\endgroup$
2
  • 1
    $\begingroup$ A perhaps more standard reasoning shows directly $\ell_\infty$ and $L_\infty(0,1)$ (both of which are $1$-injective) embed isometrically into each other. In fact, this is the usual way to show that these two spaces are isomorphic as Banach spaces, and it is not hard to see that they are not isometrically isomorphic. This approach avoids a discussion of injective envelopes. That $\ell_\infty$ embeds isometrically into $L_\infty(0,1)$ is obvious, and the other embedding follows from the fact that the predual $L_1$ of $L_\infty(0,1)$, being separable, is a quotient of $\ell_1$. $\endgroup$ Commented Jul 5 at 18:04
  • $\begingroup$ Absolutely true; I was too carried away with the injective envelopes of separable Banach spaces that I totally missed $L_{\infty}$ within an arm's reach. $\endgroup$
    – Onur Oktay
    Commented Jul 5 at 19:22

0