Completing a necessary - sufficient condition for Boundedness of Operator

57 Views Asked by At

I need to prove the following characterization for bounded diagonal linear operator:

For $\mathbb K \in \{\mathbb R, \mathbb C\}$, any $\mathbb K$-Hilbert Space $(H,\langle\cdot,\cdot\rangle,||\cdot||_{H})$, orthonormal basis $\mathbb H \subseteq H$ of $H$, any function $\lambda:\mathbb H \to \mathbb K$, and for every linear operator $A: D(A) \subseteq H \to H$ with $D(A) = \{v \in H \Big| \sum_{h \in \mathbb H}|\lambda_{h}\langle h,v\rangle_{H}|^{2} \lt \infty\}$ an $\forall$ $v \in D(A)$: $Av = \sum_{h \in \mathbb H}\lambda_{h}\langle h,v \rangle h$, it holds that $A\in L(H)$ IFF sup$(\{|\lambda_{h}| \Big |h \in \mathbb H\} \cup \{0\}) \lt \infty$

Now, starting for the proof, suppose assuming that : sup$(\{|\lambda_{h}| \Big |h \in \mathbb H\} \cup \{0\}) \lt \infty$, I need to show $A \in L(H)$.

With this in mind,

$\Big|\Big|Av\Big|\Big|_{H} = \Big|\Big|\sum\limits_{h\in \mathbb H} \lambda_{h} \langle h,v \rangle_{H} h\Big|\Big|_{H} \leq \sum\limits_{h\in \mathbb H} \Big|\lambda_{h}\Big|\Big| \langle h,v \rangle_{H}\Big|\Big|\Big|h\Big|\Big|_{H} \leq \sum\limits_{h\in \mathbb H}\Big|\lambda_{h}\Big|\Big|\Big|h\Big|\Big|_{H}\Big|\Big|v\Big|\Big|_{H}\Big|\Big|h\Big|\Big|_{H} $

[the first inequality beig the usual property for norm and the second one is due to Cauchy-Schwartz],

So now, we have $\Big|\Big|Av\Big|\Big|_{H} \leq \Bigg(\sum\limits_{h\in \mathbb H}\Big|\lambda_{h}\Big|\Big|\Big|h\Big|\Big|^{2}_{H}\Bigg)\Big|\Big|v\Big|\Big|_{H}$

Consequently, $\Big|\Big|A\Big|\Big|_{L(H)} \leq \Bigg(\sum\limits_{h\in \mathbb H}\Big|\lambda_{h}\Big|\Big|\Big|h\Big|\Big|^{2}_{H}\Bigg)$.

So, if we can show that the R.H.S. of last inequality is finite. We are done.

Now, for me, here the trouble arises. Because due to our assumption, we can pull out the $\lambda_{h}$'s outside the summation (dominating by its supremum). Then all I am left out with is : $\sum\limits_{h\in \mathbb H}\Big|\Big|h\Big|\Big|^{2}_{H}$. Which may not be finite. Isn't it ??

Please let me know how to do it from here??

And what about the converse part ??

1

There are 1 best solutions below

2
On BEST ANSWER

You are overdoing your estimates. What you have is, because $\mathbb H$ is orthonormal, \begin{align} \|Av\|_{H}^2 &= \left\|\sum\limits_{h\in \mathbb H} \lambda_{h} \langle h,v \rangle_{H} h\Big|\right\|_{H}^2 =\sum_{h\in\mathbb H}|\lambda_h\,\langle h,v\rangle|^2\\ \ \\ &=\sum_{h\in\mathbb H}|\lambda_h|^2\,|\langle h,v\rangle|^2 \leq\sup\{|\lambda_h|^2:\ h\}\,\sum_{h\in\mathbb H}|\langle h,v\rangle|^2\\ \ \\ &=\sup\{|\lambda_h|:\ h\}^2\,\|v\|^2. \end{align}

For the converse, if the supremum is not finite there is a sequence $\{h_n\}$ such that $|\lambda_{h_n}|\to\infty$. Then $$ \|Ah_n\|=|\lambda_{h_n}|\to\infty, $$ and $A$ would not be bounded.