In this question (which I think may be interesting in its own) I was asking if we can find a copy of $\ell_\infty$ between a separable subspace $Y$ contained in $\ell_\infty(\Gamma)$ and the whole space. This was in order to complete a proof, but I've realised that even if that turns out to be false, the following lemma would also complete my proof:

Lemma: Let $X$ be a Banach space and $Y$ a separable subspace. Then there exists a non-empty set $\Gamma$, an isometry $S: X \rightarrow \ell_\infty(\Gamma)$ and a subset $W \subset \ell_\infty(\Gamma)$ such that $S(Y) \subset W$ and $W$ is linearly isometric to $\ell_\infty$.

If we suppose that $Y$ is locally complemented in $X$ (which we can because of Sims-Yost theorem), then we have that $$X^* = Y^* \oplus U$$ with $U \subset Y^\perp$. Then, it seems "reasonable" to think about finding a norming set of $X$ of the form $\{y_n^*\}_{n=1}^\infty \cup \{x_i^*\}_{i \in I}$ such that the $y_n^*$'s are also norming for $Y$ and $x_i^* \in Y^\perp$ for every $i \in I$. From there it is easy to construct the isometry $S$ in the usual way.

Any ideas of how to find the norming set? If not, any ideas about the proof (if it is true) of the lemma?

EDIT: The question linked at the beggining has been solved and it is easy to deduce the lemma from that question.


1 Answer 1


In this reference (Definition 2.47) a Banach space $X$ is called separably automorphic if given a separable space $Y$ and isomorphic copies $A$ and $B$ of $Y$ in $X$, every bijective operator $T:A\to B$ can be extended to an automorphism (bijective operator) $\hat T$ of $X$.

The spaces $\ell_\infty(\Gamma)$ are separably automorphic by Proposition 2.52 in the reference.

Let $Y$ be a separable subspace of $\ell_\infty(\Gamma)$. Taking a numerable subset $\Gamma_1$ of $\Gamma$, we can identify $\ell_\infty(\Gamma_1)$ with a copy of $\ell_\infty$ in $\ell_\infty(\Gamma)$. Moreover $\ell_\infty(\Gamma_1)$ contains an isometric copy $Y_1$ of $Y$. Given a bijective isometry $T:Y_1\to Y$ and the automorphic extension $\hat T$ we mentioned, $\hat T(\ell_\infty(\Gamma_1))\supset Y$.

This result gives a positive answer to the initial question.

The Lemma can be proved with the same arguments.

ADDED on June 28, 2024: The above arguments prove the isomorphic version of the problem, but not the isometric version: in general $T$ isometry does not imply $\hat T$ isometry.

  • $\begingroup$ Thanks a lot! In fact, we are working with your notion of universally separable injective spaces :) But I'm new and I hadn't notice that section of your book yet. Thanks again! $\endgroup$ Commented Jun 27 at 18:02
  • $\begingroup$ Now that I've look at it more carefully I was wondering the same thing about $\hat T(\ell_\infty(\Gamma_1))$ $\endgroup$ Commented Jun 28 at 7:05
  • 1
    $\begingroup$ I saw it yesterday. The argument does not prove the isometric case. In fact, if every isometry between separable subspaces of $X$ extends to an isometric isomorphism of $X$ then $X$ is a space of universal disposition, but$\ell_\infty(\Gamma)$ is not (Theorem 3.34). If you need the isometric case, you have to find a different argument. $\endgroup$ Commented Jun 28 at 7:27
  • $\begingroup$ In Theorem 2.40 it is proved that every separable subspace of $\ell_\infty/c_0$ is contained in an isometric copy of $\ell_\infty$. Maybe the argument can be adapted to $\ell_\infty(\Gamma)$. $\endgroup$ Commented Jun 28 at 7:53
  • 1
    $\begingroup$ My suggestion is to use the ideas in the proof for $\ell_\infty/c_0$ to prove the case $\ell_\infty(\Gamma)$, without using the quotient $\ell_\infty(\Gamma)/c_0(\Gamma)$. $\endgroup$ Commented Jun 28 at 12:07

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.