2
$\begingroup$

This is an adaptation of a Heinrich proof, but I'm missing a key ingredient.

Conjecture. Suppose $(x_n)_{n=1}^\infty$ is a Schauder basis for a Banach space $X$ whose canonical isometric copy in $X^{**}$ is complemented. Then for any free ultrafilter $\mathcal{U}$ on $\mathbb{N}$, the canonical copy of $X$ in $X^\mathcal{U}$ is complemented in $X^\mathcal{U}$.

Proof idea. Denote by $q:X\to X^{**}$ the canonical isometric embedding. Set $X_N=\text{span}(x_n)_{n=1}^N$, and note that there exists $C\in[1,\infty)$ such that for each $N\in\mathbb{N}$ there is an operator $\widehat{P}_N:X^{**}\to X$ which is a $C$-bounded linear projection onto $X_N$ and for which $\widehat{P}_Nq$ acts as the identity on $X_N$. Let's define the linear map $J:\text{span}(qx_n)_{n=1}^\infty\to X^\mathcal{U}$ by the rule $$ Ju=(\widehat{P}_Nu)_\mathcal{U}. $$ Note that if $x\in\text{span}(x_n)_{n=1}^\infty$ then there is $k\in\mathbb{N}$ and $x\in X_N$ for $N>k$, whence $$ Jqx=(\widehat{P}_1qx,\cdots,\widehat{P}_kqx,x,x,x,\cdots)_\mathcal{U}=x^\mathcal{U}. $$ It follows that $Jq$ and hence also $J$ are continuous. Now we can extend $J$ to $qX$ via continuity, so that $Jq$ is the canonical embedding $r:X\to X^\mathcal{U}$. Due to the fact that $qX$ is complemented in $X^{**}$, we can extend $J$ again to a continuous linear operator $J:X^{**}\to X^\mathcal{U}$ with range $rX$.

Next we define the linear map $V:X^\mathcal{U}\to X^{**}$ via the rule $$ V(y_n)_\mathcal{U}=\underset{\mathcal{U}}{\text{weak*-lim}}\,qy_n $$ which exists by the weak*-compactness of $B_{X^{**}}$ together with the fact that if $K$ is a compact Hausdorff space then for each $(k_n)_{n=1}^\infty\in K^\mathbb{N}$ the (unique) limit $\lim_\mathcal{U}k_n$ exists in $K$. We now have $$ \left\langle VJqx,f\right\rangle =\left\langle Vx^\mathcal{U},f\right\rangle =\langle qx,f\rangle $$ for $x\in X$ and $f\in X^*$. This means $VJ$ is the identity on $qX$ and hence that $JV$ is an idempotent with range $rX$.

GAP. We need to show that $JV$ is continuous, from which the conjecture will follow. Of course $J$ is continuous, and so if $V$ was too then that would be sufficient. But due to the clumsiness of my weak-star intuitions, I can't seem to show it. Or, alternatively, maybe there's a theorem about linear idempotents being continuous under certain conditions.

Any ideas would be much appreciated. Thanks!

$\endgroup$
6
  • 3
    $\begingroup$ From your comments, it seems like it would be sufficient to show that $V$ is norm-decreasing. Well, if I write $Q$ for the canonical quotient map from $\ell^\infty(X)$ onto $X^{\mathcal U}$, it seems to me that $VQ=T$ sends $(x_n)$ to weak-star-$\lim_{n\in\mathcal U} q(x_n)$ and $T$ is norm-decreasing by Hahn-Banach. Since $Q$ is a quotient map this shows that $V$ is norm-decreasing. Have I misunderstood something or goofed up? $\endgroup$
    – Yemon Choi
    Sep 24, 2020 at 17:42
  • $\begingroup$ Sorry, in my comment $(x_n)$ was meant to be a generic bounded sequence in $X$, I forgot that you had used that notation for your Schauder basis $\endgroup$
    – Yemon Choi
    Sep 24, 2020 at 17:55
  • $\begingroup$ @YemonChoi I must be missing something obvious. I agree that VQ is a norm-1 operator, and that Q is a norm-1 quotient map. But whence does it follow that V is norm-1 ? $\endgroup$
    – Ben W
    Sep 24, 2020 at 19:00
  • $\begingroup$ Take $\xi$ in the domain of $V$ which has norm 1 and lift it to some $z$ in the domain of $Q$ that has norm at most 1+epsilon. Then $V\xi = VQz$ has norm at most 1+epsilon $\endgroup$
    – Yemon Choi
    Sep 24, 2020 at 19:12
  • 2
    $\begingroup$ De nada :) I spent most of yesterday thinking I'd proved something only to find my argument was circular, so it happens to all of us $\endgroup$
    – Yemon Choi
    Sep 24, 2020 at 19:16

1 Answer 1

6
$\begingroup$

Theorem. Suppose $T: X \to Y^*$ is a bounded linear operator and $\mathcal{U}$ is a free ultrafilter on $\Bbb{N}$. Then $T$ extends to an operator $S:X^\mathcal{U} \to Y^*$ with $\|S\| = \|T\|$.

Proof: Define $V:\ell_\infty(X)$ to $Y^*$ by letting $V(x_n)_n$ be the weak$^*$ limit along $\mathcal{U}$ of $Tx_n$. Identify $X$ with its diagonal in $\ell_\infty(X)$. Then $V$ extends $T$ and has the same norm as $T$. If $\|x_n\| \to 0$ along $\mathcal{U}$, then $V(x_n) =0$, so $V$ induces an operator from $X^\mathcal{U}$ into $Y^*$.

Your conjecture follows. No approximation condition on $X$ is needed.

$\endgroup$
1
  • $\begingroup$ Thank you very much! (I thought I already thanked you for this long ago, but I must have forgotten to click enter.) $\endgroup$
    – Ben W
    Dec 13, 2020 at 14:06

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.