1
$\begingroup$

This is a follow-up to this question of mine.

It is well-known that the Banach space $\ell_1$ does not contain any isomorphic copies of $c_0$. One can even go further and show that $\ell_1$ does not contain any bilipschitz copy of $c_0$. Is it however known whether $\ell_1$ does not contain any bilipschitz copy of the unit ball $B_{c_0}$ of $c_0$?

$\endgroup$

1 Answer 1

7
$\begingroup$

Yes, this is known. Raynaud showed that $B_{c_0}$ does not uniformly (in particular, bilipschitz) embed into any stable Banach space. $\ell_1$ is stable.

Added later. In the comments Bill pointed out an easier proof. Here is an elementary argument that doesn't use differentiation theory.

Suppose $f:B_{c_0}\to \ell_1$ with $$\frac1C\|x-y\|\le \|f(x)-f(y)\|\le C\|x-y\|$$

Let $(s_i)$ be the summing basis of $c_0$. For $\vec{n}=(n_1, \ldots, n_k)$, let $x(\vec{n})=\frac1k \sum_{i=1}^k s_{n_i}$. By standard stabilization argument you can find $M\subseteq \mathbb N$ so that for all $\vec{n}\in M^k$, $f(x(\vec{n}))\approx h_0+\sum_{i=1}^k h_{n_i}$ where $h_i$'s are successive blocks in $\ell_1$ and the support of $h_i$'s go to infinity as $n_i\to \infty$ (except $h_0$). Use Ramsey to stabilize the norms of $h_i$'s. Considering $\vec{n}<\vec{m}$, $\|x(\vec{n})-x(\vec{m})\|_{c_0}=1$ we see that $$\frac1C\le \|f(x(\vec{n}))-f(x(\vec{m}))\|\approx \|\sum_{i=1}^k h_{n_i}-\sum_{i=1}^k h_{m_i}\|=\sum_{i=1}^k (\|h_{n_i}\|+\|h_{m_i}\|).$$

Now consider $n_1<m_1<\ldots<n_k<m_k$ then $\|x(\vec{n})-x(\vec{m})\|_{c_0}=\frac1k$. The blocks $h_{n_1}<h_{m_1}<\ldots h_{n_k}<h_{m_k}$ are interlacing but the norm of the sum of the difference in $\ell_1$ is the same as above. Then

$$\sum_{i=1}^k (\|h_{n_i}\|+\|h_{m_i}\|)=\|\sum_{i=1}^k h_{n_i}-\sum_{i=1}^k h_{m_i}\|\approx\|f(x(\vec{n}))-f(x(\vec{m}))\|\le \frac{C}{k}.$$

which yields a contradiction for large $k$.

$\endgroup$
4
  • 2
    $\begingroup$ True, but this is overkill. It is easy to see that if the unit ball of $Y$ bilipschitz embeds into $X$, then $Y$ bilipschitz embeds into $X\oplus R$. Also, the same differentiation proof that $c_0$ does not bilipschitz embed gives that no open subset of $c_0$ bilipschitz embeds. $\endgroup$ Dec 4, 2022 at 21:06
  • $\begingroup$ @BillJohnson Added an elementary argument to the answer. $\endgroup$ Dec 5, 2022 at 2:55
  • $\begingroup$ @BunyaminSari, is the same true if we ask about the unit sphere $S_{c_0}$ instead of the unit ball $B_{c_0}$ (i.e. whether it embeds bilipschitz/uniformly into $\ell_1$)? $\endgroup$ Jan 27, 2023 at 12:41
  • 1
    $\begingroup$ @DamianSobota The vectors $x(\vec{n})$ and $x(\vec{m})$ are in the unit sphere. $\endgroup$ Jan 27, 2023 at 23:15

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.