MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Let $X$ be a real Banach space.

For a continuous (not necessarily linear) function $g:X \to \mathbb{R}$ and a family $\mathcal{F} \subseteq X^*$, we´ll say that $\mathcal{F}$ determines $g$ if whenever $g(x) \neq g(y)$, there is an $f \in \mathcal{F}$ such that $f(x) \neq f(y)$. Since $\mathcal{F}=X^*$ determines any function (because in fact it "determines points"), it makes sense to define: $$s(g):=\min \{|\mathcal{F}|:\mathcal{F} \mbox{ determines } g\}.$$

For example if $g$ is constant then $s(g)=0$ and if $g$ is linear then $s(g)=1$.

Now let $$s(X)=\sup_{g \in C(X)} s(g).$$

For finite-dimensional $X$ we have that $s(X) = \dim(X)$ and the supremum is attained for example by the function $g(x)=\|x\|$.

I hope someone with more background in Banach spaces than me (that probably includes most of the regulars on this site!) can easily answer some/all of the following:

1) Is there a simple way to compute $s(X)$ for infinite-dimensional $X$?

2) Is there always a $g \in C(X)$ for which $s(X)=s(g)$?

3) Is it true that if $X$ is separable then $s(X) \leq \aleph_0$?

Any comment, answer or reference will be greatly appreciated.

share|cite|improve this question
3: if $X$ is separable, then there is a countable family of bounded linear functionals that separate points in $X$. – Gerald Edgar Apr 25 '12 at 20:39
Thank you Gerald! I suppose that in general there is such a family of size at most the density character of $X$. Is that correct? – Ramiro de la Vega Apr 25 '12 at 21:16
By Hahn-Banach, for every $x \in X$ we can find $f_x \in X^*$ with $\| f_x\|=1$ and $f_x(x) = \|x\|$. Now if $E$ is dense in $X$, then $\{f_x : x \in E\}$ separates points of $X$. Choose $x_n \in E$ with $x_n \to x$, then $f_n(x_n) = \|x_n\| \to \|x\|$, and on the other hand $|f_n(x_n) - f_n(x)| \le \|x_n - x\| \to 0$. Thus $f_n(x) \to \|x\|$ and in particular $f_n(x) \ne 0$ for large enough $n$. – Nate Eldredge Apr 26 '12 at 3:37
@Nate: Thank you very much. That means $s(X) \leq dc(X)$. Michael´s answer below seems to imply that $s(X)=dc(X)=s(\|\cdot\|)$, but I still have to decrypt it. – Ramiro de la Vega Apr 26 '12 at 13:07

If the span of ${\cal F}$ is weak-* dense, it separates points. If not, it does not determine $\|x\|$.

share|cite|improve this answer
Thanks Michael. So this just means that $s(X)=d(X^*,wk-^*)=s(||\cdot||)$ which answers the three questions. But is it obvious that if $\mathcal{F}$ determines the norm then its span must be weak-$^*$ dense? – Ramiro de la Vega Sep 18 '13 at 10:18

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.