5
$\begingroup$

Given the normed space $\ell^\infty$ of all bounded sequences of (real or complex) numbers with the norm given by $$||x||:= \sup_{j\in Z^+} |\xi_j|,$$ for each $x:=(\xi_j)_{j=1}^\infty$ in $\ell^\infty$, and given the linear operator $T \colon \ell^\infty \to \ell^\infty$ defined as $$T(\xi_j)_{j=1}^\infty := (\frac{\xi_j}{j})_{j=1}^\infty,$$ while $T$ is bounded, how to compute the norm of $T$, and how to find its range?

What is the range of $T$? And what is its norm? Is the range a closed set in $\ell^\infty$?

This operator being injective admits an inverse. (This inverse is also a linear operator from the range to the domain space.) Is the inverse a bounded operator too?

$\endgroup$
  • $\begingroup$ To find the norm is not difficult, so I'd like to give a couple of hints which will allow you to find the answer yourself. It is easy to find an estimate of $\lVert Tx\rVert_{\infty}$ in terms of $\lVert x \rVert_{\infty}$, and this will give you an upper bound $\lVert T\rVert\le C$ (I'm not revealing the exact value of $C$, you can easily find it yourself). To prove that this upper bound cannot be strict, just test the inequality $\lVert Tx\rVert_{\infty}\le C\lVert x \rVert_\infty$ with the sequence $x=(1, 1, 1,\ldots)$. $\endgroup$ – Giuseppe Negro Feb 16 '13 at 13:27
  • $\begingroup$ The value of $C$ should be at least $1$, and since in this particular case $||Tx|| = ||x||$, can we conclude that the norm is $1$? $\endgroup$ – Saaqib Mahmood Feb 16 '13 at 13:48
  • $\begingroup$ Yes, that's what I thought. $\endgroup$ – Giuseppe Negro Feb 16 '13 at 15:41
9
$\begingroup$

It is obvious that $\Vert Tx\Vert_\infty\leq\Vert x\Vert_\infty$, hence $\Vert T\Vert\leq 1$. Since $\Vert T(1,0,0,\ldots)\Vert_\infty=\Vert (1,0,\ldots)\Vert_\infty$, we conclude that $\Vert T\Vert=1$.

One can easily verify that $$T(l^\infty)=\left\{(\xi_j):\exists C>0 \text{ such that } |\xi_j|\leq C/j\text{ for all j}\right\}$$

$T(l^\infty)$ is not closed in $l^\infty$: Let $x_n=(1,\dfrac{1}{\sqrt{2}},\ldots,\dfrac{1}{\sqrt{n}},0,0,\ldots)=T(1,\sqrt{2},\ldots,\sqrt{n},0,\ldots)$. The sequence $(x_n)$ obviously converges to $x=\left(1/\sqrt{j}\right)_{j=1}^\infty\in l^\infty$, which does not lie in $T(l^\infty)$ (if it did, what would be it's pre-image?).

The inverse operator $T^{-1}$ is not bounded: Consider the sequence $(x_n)\subseteq T(l^\infty)$ as above. This sequence is bounded but the image $\left\{T^{-1}x_n\right\}$ is not, since $\Vert T^{-1}x_n\Vert_\infty=\sqrt{n}$.

You could also argue in this way: If $T^{-1}$ were bounded, then for every Cauchy sequence $(y_n)\in T(l^\infty)$, the sequence $(T^{-1} y_n)$ would also be Cauchy, and hence would converge to some $x\in l^\infty$ since $l^\infty$ is complete. But then, since $T$ is bounded, $y_n$ would converge to $Tx$. Then $T(l^\infty)$ would be complete, contradicting the fact that it is not closed in $l^\infty$ (this argument can actually be used to show that if $T:X\rightarrow Y$ is an isomorphism between a Banach space $X$ and some normed space $Y$ such that $T$ and $T^{-1}$ are bounded, then $Y$ is Banach).

$\endgroup$
  • $\begingroup$ Thank you so much! This answers my question satisfactorily enough. $\endgroup$ – Saaqib Mahmood Feb 22 '13 at 7:43

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.