6
$\begingroup$

Let $X$ be a real Banach space and $Y\subset X$ be a (closed) subspace of $X$. Assume that a sequence $y_n^*\in S_{Y^*}$ weak*-converges to some $y^*\in S_{Y*}$. (Here $S_{Y^*}$ stands for the dual unit sphere.)

Let $x_0 \in X \setminus Y$ and define $Z:= \textrm{span} (Y\cup \{x_0\})$.

Is the following true?

There exist a sequence $z_n^*\in S_{Z^*}$ with $z_n^*\restriction_Y = y_n^*$, and $z^*\in S_{Z^*}$ with $z^*\restriction_Y = y^*$, such that $z_n^*$ converges weak* to $z^*$.

Remarks:

  1. It is easy to obtain that a SUBSEQUENCE of $z_n^*$ converges weak* to $z^*$.
  2. I'd like to consider general Banach spaces without additional assumptions (e.g. reflexivity, separability).
  3. I wouldn't mind using an additional set-theoretic axiom consistent with ZFC.
$\endgroup$
2
  • 1
    $\begingroup$ Isn't this the "inductive step" from the proof of Hahn-Banach? You prove that you can always extend a functional to a space of dimension one larger, without increasing its norm. This part of the proof can be done explicitly (you don't need AC at this stage) and it should be possible to see whether the extension can be done in a weak-* continuous fashion. $\endgroup$ Commented Dec 27, 2017 at 16:05
  • 1
    $\begingroup$ The only obstacle on the way to constructing the extensions like in, for example, Johnson's answer to mathoverflow.net/questions/245773 is that you would like to have norm-preserving extensions. This leaves some hope to construct a counterexample (extensions of different functionals can be `distorted' in different ways). $\endgroup$ Commented Dec 27, 2017 at 17:56

1 Answer 1

6
$\begingroup$

The norm preserving condition, is as Mikhail points out, essential to prove that the such an extension does not exist. The question is stated for Banach spaces, but notice that the same question for normed spaces is equivalent to the question for Banach spaces because a bounded sequence of functionals converges weak$^*$ if it converges pointwise on a dense set.

So assume that the question has an affirmative answer in the category of normed spaces. We claim that a stronger thing is true; namely, in the set up of the OP, the superspace $Z$ of $Y$ can be taken to be $X$ itself. Indeed, the union of a nested family of extensions that are weak$^*$ convergent gives extension to the union of the domains of the extending functionals, so one can Zornicate just as in the standard proof of the Hahn-Banach theorem. More explicitly, consider all $(Z, (z_n^*)_n)$ such that $Y \subset Z $, $Z$ is a subspace of $X$, and $(z_n^*)_n$ are linear functionals in the unit ball of $Z^*$ such that for all $n$, $z_n^*$ extends $y_n^*$, and $(z_n^*)_n$ converges pointwise on $Z$. Partially order by $(Z, (z_n^*)_n) \le (W, (w_n^*)_n)$ provided $Z\subset W$ and for all $n$, $z_n^* \subset w_n^*$, and observe that the hypothesis in Zorn's Lemma applies.

To see that the OP's question has a negative answer, let $X=\ell_\infty$, $Y=c$ (the space of convergent sequences under the supremum norm), and let $y_n^*$ be the unit vector basis for $\ell_1 \subset c^*$. There are not uniformly bounded weak$^*$ convergent extensions to $\ell_\infty$ because weak$^*$ converging sequences in $\ell_\infty^*$ are weakly convergent.

EDIT 12/30/17: If you want an explicit example in the separable setting to the OP's original formulation, again let $X=\ell_\infty$, $Y=c$ (the space of convergent sequences under the supremum norm), let $y_n^*$ be the unit vector basis for $\ell_1 \subset c^*$, and let $z_0 $ be the element of $\ell_\infty$ whose odd coordinates are all $1$ and whose even coordinates are all $-1$.

$\endgroup$
6
  • $\begingroup$ Thank you very much for your answer. You are right about the application of Zorn's Lemma. This is exactly how I wanted to use it later (once I have the inductive step). However, if I understand correctly, the standard basis in $\ell_1$ is not a counterexample. It weak* converges to $0$, whereas I need the limit functional $y^*$ to have norm 1. $\endgroup$ Commented Dec 30, 2017 at 9:39
  • $\begingroup$ ZORNICATE -- Ouch! I once heard that Max Zorn was quite angry that his name was only known because of a simple lemma. With this verb he is even more reduced to that. $\endgroup$ Commented Dec 30, 2017 at 16:05
  • $\begingroup$ Oh, I missed that part. Just replace $c_0$ by $c$ (the space of convergent sequences) to make the weak$^*$ limit have norm one. I'll change it when I do an edit I planned to do. $\endgroup$ Commented Dec 30, 2017 at 17:26
  • 2
    $\begingroup$ @Jochen Wengenroch: I know how Zorn feels. :) $\endgroup$ Commented Dec 30, 2017 at 17:28
  • $\begingroup$ I think Zornicate is the best thing ever happened on MO! $\endgroup$ Commented Dec 30, 2017 at 19:15

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .