3

Let $\{e_n\}_{n=1}^{\infty}$ be an orthonormal basis of the complex Hilbert space $\ell^2$. Fix complex numbers $\lambda_1,\lambda_2,\lambda_3,\dots$, let

$$ \mathscr{D}(T)=\{\sum_{n=1}^{\infty} x_n e_n\in \ell^2:\sum_{n=1}^{\infty}|\lambda_n x_n|^2< \infty\} $$

and define $T:\mathscr{D}(T) \rightarrow \ell^2$ as

$$ T\left( \sum_{n=1}^{\infty} x_n e_n \right) = \sum_{n=1}^{\infty} \lambda_n x_n e_n $$

for $\sum_{n=1}^{\infty} x_n e_n \in \mathscr{D}(T)$.

Determine the adjoint $T^*$ of $T$.

What I have done is the following.

Let $x,y\in\mathscr{D}(T)$. Then \begin{align*} |(Tx,y)| &=|(\sum_{n=1}^{\infty}\lambda_n x_n e_n,\sum_{n=1}^{\infty}y_n e_n)| \\ &=|(\lim_{m\rightarrow\infty}\sum_{n=1}^{m}\lambda_n x_n e_n,\lim_{m\rightarrow\infty} \sum_{n=1}^{m}y_n e_n)| \\ &=\lim_{m\rightarrow\infty} |(\sum_{n=1}^{m}\lambda_n x_n e_n,\sum_{n=1}^{m}y_n e_n)| \\ &=\lim_{m\rightarrow\infty} |\sum_{n=1}^{m}(\lambda_n x_n e_n,y_n e_n)| \\ &=\lim_{m\rightarrow\infty} |\sum_{n=1}^{m}( x_n e_n, \bar\lambda_n y_n e_n)| \\ &=\lim_{m\rightarrow\infty} |(\sum_{n=1}^{m} x_n e_n, \sum_{n=1}^{m} \bar\lambda_n y_n e_n)|\\ &=|(\sum_{n=1}^{\infty} x_n e_n, \sum_{n=1}^{\infty} \bar\lambda_n y_n e_n)|\\ &=|(x,\sum_{n=1}^{\infty} \bar\lambda_n y_n e_n)| \leq\|x\| \sqrt{\sum_{n=1}^{\infty}|\lambda_n y_n|^2} \end{align*} Where the third and forth equality follow from the continuity and linearity of the inner product respectively. Since $y\in\mathscr{D}(T)$ we have that $\sum_{n=1}^{\infty}|\lambda_n y_n|^2 < \infty$. It follows that the linear function $x\mapsto(Tx,y)$ is bounded and thus continuous. So we have $\mathscr{D}(T)\subset\mathscr{D}(T^*)$.

Now let $y\in\mathscr{D}(T^*)$ then $\Lambda x = (Tx,y)$ can be extended by the Hahn-Banach theorem 3.6 to a linear continuous function on $H$. By theorem 12.5 there exist an element $T^*y \in H$ such that $$ \Lambda x =(x,T^* y) $$ and $\|\Lambda\| = \|T^* y\|$. Since we also have that \begin{align*} \Lambda x &= (Tx,y) \\ &= (\sum_{n=1}^{\infty} \lambda_n x_n e_n, \sum_{n=1}^{\infty} y_n e_n) \\ &= \sum_{n=1}^{\infty} (\lambda_n x_n e_n,y_n e_n) \\ &= \sum_{n=1}^{\infty} (x_n e_n,\bar\lambda_n y_n e_n) \\ &= (x, \sum_{n=1}^{\infty} \bar\lambda_n y_n e_n). \\ \end{align*} It follows that $(x,T^* y) = (x, \sum_{n=1}^{\infty} \bar\lambda_n y_n e_n)$ and by theorem 12.7 that $T^* y=\sum_{n=1}^{\infty} \bar\lambda_n y_n e_n$. So $\sum_{n=1}^{\infty} |\lambda_n y_n|^2=\|\sum_{n=1}^{\infty} \bar\lambda_n y_n e_n\|^2 = \|(T^* y)\|^2 < \infty$ and $y \in \mathscr{D}(T)$. We conclude that $\mathscr{D}(T)=\mathscr{D}(T^*)$.

Furthermore since $\mathscr{D}(T)$ is dense $T^* y$ is uniquely defined.

My question now is if I'm done?

simon
  • 446
  • For $\mathscr{D}(T^*)\subset\mathscr{D}(T)$ you're missing some squares, but also note that $|(Tx,y)|^2\leq M^2||x||^2$ and $|(Tx,y)|^2\leq ||x||^2\sum_{n=1}^\infty|\lambda_n y_n|^2$ does not imply that $\sum_{n=1}^\infty|\lambda_n y_n|^2< \infty$. – ScroogeMcDuck Jan 16 '14 at 15:39
  • But if there exists a $M>0$ such that $|(Tx,y)|^2\leq M^2||x||^2$ for all $x$ then $\sum_{n=1}^\infty|\lambda_n y_n|^2$ cannot be greater than $M^2$. – simon Jan 18 '14 at 16:34
  • Why? It may be true, but it does not follow from these two statements. If $\sum_{n=1}^{\infty}| \lambda_n y_n|=\infty$, the statements are $|(Tx,y)|^2\leq M^2 ||x||^2$ and $|(Tx,y)|^2\leq \infty$, the second is always true and does not contradict the first one. – ScroogeMcDuck Jan 19 '14 at 11:58
  • You're right. I edited it and tried to prove it using riesz representation theorem. I hope it's correct now. – simon Jan 20 '14 at 12:51

1 Answers1

1

Your approach is indeed correct.