Take the 2-minute tour ×
MathOverflow is a question and answer site for professional mathematicians. It's 100% free, no registration required.

Let $(X,d)$ be a complete metric linear space whose balls are convex. Let $Y\subseteq X$ be a bounded, closed and convex subset that verifies the following property: for all $y_0\in Y$, the distance function $y\rightarrow d(y_0,y)$ attains its sup in $Y$. Does $Y$ have extreme points?

I was trying to adapt the classical Krein-Milman's proof: the family of non-empty closed faces is not empty (because, given $y_0$, the set of maxima of $d(y_0,\cdot)$ is an extremal non-empty set and one can get a face by taking a maximal and convex superset of $\bar y$, being $\bar y$ a point that maximizes the distance from $y_0$). It is inductive by completeness, boundness and convexity of the faces. The last step is the hardest one and I'm not sure it is true: given a face $F$ with two different points $x$ and $y$, can I say that $d(x,\cdot)$ attains its sup in $F$?

Or.. are there counterexample?

Thanks in advance, Valerio

share|improve this question

2 Answers 2

up vote 7 down vote accepted

There is a counter-example.

Note that in any normed space, its unit ball satisfies your supremum-attaining property. Indeed, for any $x_0\in X$ the supremum of $d(x_0,\cdot)$ on the ball is attained at the point $-x_0/\|x_0\|$ if $x_0\ne 0$, and at any point of the sphere if $x_0=0$.

It remains to construct a Banach space whose unit ball has no extreme points. Let $X$ be the space of all sequences $(x_i)_{i=1}^\infty$ such that $x_i\to 0$ as $i\to\infty$, equipped with the $\ell_\infty$ norm $\|x\|=\sup|x_i|$. It is a closed subspace of $\ell_\infty$ and hence complete. But clearly its unit ball has no extreme points.

share|improve this answer

For a counterexample, let $X=c_0$ and let $Y$ be its unit ball.

share|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.