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

Suppose we have a normed vector space $V$ and its dual $V^*$, and suppose that $X \subseteq V^*$ has the property that for every $v \in V$, there is some $\phi \in X$ with $\Vert \phi \Vert = 1$ such that $\phi(v) = \Vert v \Vert$. Is $X$ dense in $V^*$ (in the operator norm)? Note that this is a stronger property than $\Vert v \Vert = \sup_{\phi\in X} \frac{\phi(v)}{\Vert \phi \Vert}$, since we are assuming that the supremum is realized.

I think the answer is probably "no." A nice example (passed to me originally made up by Terry Tao) showing that the second condition (the supremum over $X$ gives the norm) does not imply dense is the following: consider $l^1$ and $(l^1)^* = l^\infty$. Then the space of eventually zero sequences in $l^\infty$ is sufficient for the norm: given $f\in l^1$, let $\phi_n$ be a truncation of the sign function of $f$ to the first $n$ indices. Then $\lim_{n\to \infty} \phi_n(f) = \Vert f \Vert$. However (for $f$ with infinite support), there is no finite sequence $\phi$ with $\phi(f) = \Vert f \Vert$.

share|cite|improve this question
You must assume that $X$ is a subspace of $V$, else the result is trivially wrong (take $X$ to be the unit sphere of a reflexive space $V$). – Harald Hanche-Olsen May 5 '10 at 3:52
A subset $B$ of the unit sphere of the dual of a Banach space $V$ that has the property that every vector in $V$ achieves its norm at some functional in $B$ is called a boundary for $V$. Recently Hermann Pfitzner arXiv:0807.2810 solved Godefroy's boundary problem in the affirmative. The boundary problem was whether a subset of $V$ is weakly compact if it is compact in the topology of pointwise convergence on some boundary for $V$. This was a problem of some note, with earlier related results by, including others, Grothendieck and Bourgain-Talagrand. – Bill Johnson May 6 '10 at 7:36
up vote 11 down vote accepted

The answer is negative. Since the linear span of the Dirac masses is not a dense subspace of the dual of $C[0,1]$.

share|cite|improve this answer
Excellent! One of those answers which, when you see it, makes you slap your forehead and say “why didn't I think of that?” – Harald Hanche-Olsen May 5 '10 at 14:11
Thanks! This is a great simple example. – Alden Walker May 5 '10 at 17:11

Far from a complete answer, but the answer is yes if $V=\ell^1$: For then $X$ must contain every sequence $x\in\ell^\infty$ with $|x_n|=1$ for all $n$ (consider $v\in\ell^1$ given by $v_n=\bar x_n/n^2$), and the space of linear combinations of such sequences is dense in $\ell^\infty$. To see the latter, merely note that any complex number $z$ with $|z|\le2$ can be written as $x+y$ with $|x|=|y|=1$. (Over the reals, you need to work a tiny bit harder.)

share|cite|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.