3
$\begingroup$

We know that in finite dimension spaces the closed unit ball is compact, that is if H is a finite dimension space, then there exists an $u$ in the closed unit ball in H and $T \in \mathcal{L}(H, \mathbb{R})$ such that $||T|| = |T u|$.

Why doesn't it happen in infinite dimension spaces?

For example, let $H = C([a, b])$ with the sup norm. Why doesn't exist an element $u$ in the closed unit ball in $C([a, b])$ such that $T u = ||T||$?

Thank you!

$\endgroup$
3
2
$\begingroup$

For your second question, I assume you meant to refer to the assertion:

For any linear operator $T \in \mathcal{L} (H, \mathbb R)$, there exists an element $u$ in the unit ball in $H$ (actually the unit sphere), such that $\Vert T \Vert = |Tu|$.

This is true, if and only if the unit ball is compact.

A counter-example in an infinite-dimensional space $H$ is obtained by taking an infinite sequence $(v_n)_{n\in \mathbb{N}}$ of linearly independent elements in $H$ (all of lenght $1$) and defining a linear operator $T \in \mathcal L (H, \mathbb R)$ by $$Tv_n = 1- \frac 1 n$$ expanding linearly to the span of $(v_n)_{n\in \mathbb{N}}$, i.e. $$T \left(\sum_{i=1}^k a_k v_{j_i} \right) = \sum_{i=1}^k a_k T (v_{j_i}),\quad k,j_i \in \mathbb N, a_{k} \in \mathbb R $$ and setting $T=0$ otherwise. Then $\Vert T \Vert =1$, but $|Tu|<1$ for all $u \in H$.

$\endgroup$
3
$\begingroup$

Essentially because sequences can wander off to infinite. Eg take any function with sup norm $1$ supported in the interval $[0,1].$ Call it $f.$

Define $f_n(x) = f(x-n).$

These functions have disjoint support and their distance in the sup norm is $1.$

$\endgroup$
4
  • $\begingroup$ it has to be supported in $[0,1]$ i.e. zero outside that interval. $\endgroup$ – Mark Joshi Mar 23 '15 at 23:37
  • $\begingroup$ Sorry, I misread "supported in" as "with support contained in". $\endgroup$ – Ian Mar 23 '15 at 23:38
  • $\begingroup$ Their distance in the sup norm is $1$. $\endgroup$ – Olivier Bégassat Mar 23 '15 at 23:55
  • $\begingroup$ oops ill fix this $\endgroup$ – Mark Joshi Mar 23 '15 at 23:58
1
$\begingroup$

You ask two different questions. For the second one: yes, given $u \in C([a,b])$, there is a linear functional $T \in \mathcal L(C([a,b]),\mathbb R)$ with $|T(u)| = \|u\|$ and $\|T\|=1$. But of course, since the unit ball is not compact, you cannot use compactness of the unit ball in the proof...

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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