5
$\begingroup$

I'm trying to prove the next inequality, like Cauchy-Schwarz standard inequality:

$$|\langle Tx,y\rangle |\leq\langle Tx,x\rangle ^{1/2}\langle Ty,y\rangle ^{1/2}\space\forall x,y\in\mathcal{H},$$ where $\mathcal{H}$ is a complex Hilbert space, $T$ bounded linear operator, $T\geq 0$ and $T=T^{*}.$

If we consider that, for $t\in\mathbb{R},$ $$0\leq\langle T(y-tx),y-tx\rangle =\langle Ty,y\rangle -2t\mathcal{Re}(\langle Ty,x\rangle )+t^{2}\langle Tx,x\rangle :=P(t),$$ then $P(t)$ is a polynomial of second grade, so its discrimante have to be $$\mathcal{Re}^{2}(\langle Ty,x\rangle )\leq\langle Tx,x\rangle \langle Ty,y\rangle ,$$ but I would like to conclude that $$|\langle Ty,x\rangle |\leq\langle Tx,x\rangle \langle Ty,y\rangle ,$$

How can we conclude the desire inequality? Is there something wrong?

$\endgroup$
1
  • $\begingroup$ (This is Schwarz without a "t"...) $\endgroup$ – paul garrett Mar 15 '18 at 12:12
5
$\begingroup$

For any $\epsilon > 0$, $$ [x,y]_{\epsilon} = \langle (T+\epsilon I)x, y\rangle $$ is an inner product. Therefore, the Cauchy-Schwarz inequality gives $$ |[x,y]_{\epsilon}| \le [x,x]_{\epsilon}^{1/2}[y,y]_{\epsilon}^{1/2}. $$ That is, the following holds for all $\epsilon > 0$ and $x,y\in H$ $$ |\langle (T+\epsilon I)x,y\rangle| \le \langle (T+\epsilon I)x,x\rangle^{1/2}\langle (T+\epsilon I)y,y\rangle^{1/2} $$ Now let $\epsilon\downarrow 0$ to obtain the desired result that $$ |\langle Tx,y\rangle| \le \langle Tx,x\rangle^{1/2}\langle Ty,y\rangle^{1/2}. $$

$\endgroup$
4
  • $\begingroup$ Thanks to answer @DisintegratingByParts, but I don't get how does work Cauchy-Schwarz inequality in this case because $[x,y]_{\epsilon}\leq||(T+\epsilon I)x||||y||$ $\endgroup$ – Squird37 Mar 15 '18 at 16:45
  • $\begingroup$ @Squird37 : The second displayed inequality is the Cauchy-Schwarz inequality for the inner product $[x,y]_{\epsilon}$. Write this out in terms of the definition above it, and then let $\epsilon\downarrow 0$ to get what you want. $\endgroup$ – Disintegrating By Parts Mar 15 '18 at 16:55
  • $\begingroup$ I get it @DisintegratingByParts! I wasn't paying attention to $[x,y]_{\epsilon}$ is a inner product. Many thanks for the help; it's a way to prove it that I didn't imagine. $\endgroup$ – Squird37 Mar 15 '18 at 17:03
  • $\begingroup$ @Squird37 : You're welcome. It's a good lazy way to avoid much work. $\endgroup$ – Disintegrating By Parts Mar 15 '18 at 18:04
2
$\begingroup$

Replace $y$ in the inequality you have obtained by $ay$ where $a$ is a complex number. You get $\{Re (a\langle Ty,x \rangle \})^{2} \leq |a|^{2}\langle Tx, x\rangle \langle Ty, y\rangle $. Now put $\bar a=\frac {\langle Ty,x \rangle} {|\langle Ty,x \rangle|}$.

$\endgroup$
1
  • $\begingroup$ Thanks @KaviRamaMurthy. This is so clear! I'm waiting for the above answer to choose one. Thanks for help. $\endgroup$ – Squird37 Mar 15 '18 at 16:47

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.