2
$\begingroup$

Suppose $X$ is a strictly convex Banach space. Does there exist a bounded closed convex set $K$ in $X$ such that the set of all Chebyshev centers $C(K)$ of $K$ is a proper subset of $K$ with diameter of $C(K) > 0$ ?

$\endgroup$
2
  • 4
    $\begingroup$ As a general rule, titles more to the point are preferred (it make it easier to reach potentially interested people). Also, you can use more tags, and more precise (banach-spaces, convexity,..). I suggest to change the title accordingly. $\endgroup$ Sep 5, 2014 at 12:21
  • $\begingroup$ s $ \phantom{.} $ $\endgroup$ Sep 5, 2014 at 19:06

1 Answer 1

3
$\begingroup$

My purpose is to show that there exists $X$ for which the mentioned phenomenon occurs. Of course it does not occur for all strictly convex spaces because it does not occur for uniformly convex spaces.

I think that one can work out an example of such space $X$ and set $K$ on the following lines. We consider the space $\ell_2$ with the following equivalent norm: We pick a tending to $\infty$ increasing sequence $\{p_i\}$ of numbers $>1$. For $a=\{a_i\}_{i=0}^\infty$ we let $$||a||_s=\sup_{i\ge 1}\left(|a_0|^{p_i}+|a_i|^{p_i}\right)^{1/p_i}+\left(\sum_{i=1}^\infty|a_i|^2\right)^{1/2}.$$ This norm is strictly convex because the supremum is always attained.

Denote by $\{e_i\}_{i=0}^\infty$ the unit vector basis of this space. Let set $K$ be defined as the closed convex hull of $\left\{\pm \frac34e_0,\pm b_ie_i, i\ge 1\right\}$ where $\{b_i\}$ are chosen in such a way that $$\left\|\frac14e_0\pm b_ie_i\right\|_s=1.$$ Then $|b_i|\uparrow \frac12$. This implies that $\hbox{diam}K=2$. Therefore each element of the line segment joining $\frac14e_0$ and $-\frac14e_0$ is in the Chebyshev center of $K$. It seems straighforward to verify that actually $C(K)$ is equal to this line segment.

$\endgroup$

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.