Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. 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

How does one prove that $C^0([0;1],\mathbb{R})$ equipped with the sup norm is not reflexive?

I don't understand how to show that the $J$ mapping is not surjective.

share|cite|improve this question
You are missing "not" in your first sentence? Do you know what the dual of $C^0([0,1],\mathbb R)$ is? – Jonas Meyer Jun 3 '12 at 22:17
David - I am not too sure of the space that has been given in the is. I know that C^infinity([0:1],R) is the set of infinitely differentiable functions but I couldn't find the definition on the internet. Also, what is the norm that the linear functional x' is equipped with? – nada Jun 3 '12 at 22:39
Okay, yes, I get the solution now. Thank you. :) – nada Jun 3 '12 at 22:56
up vote 8 down vote accepted

As we have seen in a previous question, if $E$ is a reflexive Banach space then each linear continuous functional attains its norm. So, in order to show that $E:=C^0([0,1],\Bbb R)$ endowed with the supremum norm is not reflexive, it is enough to find a linear functional which is not norm attaining. We can define $$x'(f):=\int_0^{1/2}f(t)\mathrm dt-\int_{1/2}^1f(t)\mathrm dt;$$ $x'$ is linear and $|x'(f)|\leqslant \lVert f\rVert_{\infty}$, hence $x'$ is continuous and its norm is $\leqslant 1$.

To see that the norm is indeed $1$, for $n$ integer, consider a function $f_n$ which is $1$ on $[0,1/2-1/n)$ and $-1$ on $(1/2+1/n,1)$, and linear on $(1/2-1/n,1/2+1/n)$. We can see that $\lVert f_n\rVert=1$ and $x'(f_n)=1-2/n$.

Now, we have to show that we cannot find $f\in E$ such that $x'(f)=1$ and $\lVert f\rVert=1$. Let $f$ be continuous of norm $1$. We have to show that $x'(f)\neq 1$, for a fixed $\varepsilon>0$ we can find $\delta>0$ such that $|f(t)-f(1/2)|\leqslant \varepsilon$ whenever $|t-1/2|\leqslant \delta$. Since $$\tag{*} x'(f)=\int_0^{1/2-\delta}f(t)\mathrm dt+\int_{1/2-\delta}^{1/2+\delta}f(t)\mathrm dt -\int_{1/2+\delta}^1f(t)\mathrm dt,$$ we can assume that $f(x)=1$ on $[0,1/2-\delta]$ and $f(x)=-1$ on $[1/2+\delta,1]$, otherwise it is clear that $x'(f)\neq 1$.

By (*), it follows that $|x'(f)|\leqslant 1-2\delta+2\delta\varepsilon+\delta| f(1/2)|$. Taking $\varepsilon\lt 1-|f(1/2)|$, we get that $|x'(f)|\lt 1$.

share|cite|improve this answer
Why does $2 \delta \varepsilon$ appear? I obtained $$ \underset{1/2 - \delta}{\overset{1/2}{\int}} f(t) dt - \underset{1/2}{\overset{1/2 + \delta}{\int}} f(t) dt \leq 2 \delta f(1/2) - \frac{1}{2} f(1/2) + \varepsilon / 2 $$ Thank you! – g.pomegranate Mar 25 '15 at 19:14
@rodie You are right. Thank you for spotting an inaccuracy. I have edited. – Davide Giraudo Mar 25 '15 at 20:23
But why does $2 \delta \varepsilon$ appear? I obtained $$x'(f) \leq 2 \delta (f(\frac{1}{2}) + \varepsilon) + 1 - 2 \delta.$$ Is this because $f(\frac{1}{2}) < \infty$ and then we can consider $\varepsilon : = f(\frac{1}{2}) + \varepsilon $? Thak you! – g.pomegranate Mar 27 '15 at 11:46
You are right, I have edited again. – Davide Giraudo Mar 27 '15 at 12:11
@DavideGiraudo Is there any typing error in your definition of $(f_n)$? – Pedro Nov 6 '15 at 0:54

you could also use the following facts:

  • $BV[a,b]$ is not separable,
  • $C[a,b]$ is separable,
  • if $X$ is a Banach Space and $X^*$ is its dual and is separable, then $X$ is separable.

Since $BV[a,b]$ is the dual of $C[a,b]$, reflexivity would imply $BV[a,b]$ is separable.

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.