The first one:

$$a) A(x(t)) = \int_{0}^{t}x(s)ds$$

Here I proved that $\|A(x(t))\|_{L_2[0,1]} \leq \|x(t)\|_{L_2[0,1]}$. Unfortunately I cannot find a function that has the norm of one and the map having the norm of one too.

$$b) \lambda \in (0,1), A(x(t)) = \begin{cases} x(t), t \leq \lambda \\ 0, t > \lambda \end{cases} t\in [0,1]$$

$\|A(x(t))\|_{L_2[0,1]} \leq \|x(t)\|_{L_2[0,1]}$ I got this again, but I am sure the norm is less than one. Closer to $\lambda$. But I dont know how to prove this.

  • 1
    $\begingroup$ Integrate x(s) = s, see what you get then normalize it so that it has norm 1. $\endgroup$ – Paul May 8 '17 at 13:38
  • $\begingroup$ @Paul what does that hint mean? $\endgroup$ – Ben Grossmann May 8 '17 at 18:04

First of all, the proof of the "other direction" is important for motivating an example. Here's my version: $$ \|Ax\| = \int_0^1 \left|\int _0^tx(s)\,ds\right|^2 dt \leq \int_0^1 \int _0^t |x(s)|^2\,ds\, dt \\ = \int_0^1 \int_s^1 |x(s)|^2 \,dt\,ds = \int_0^1\, (1-s)|x(s)|^2\,ds\\ \leq \int_0^1 |x(s)|^2\,ds $$ Now, from that second to last step, perhaps you can intuit that for any $x(s)$, we will have $\|Ax\| < \|x\|$. However, we're ultimately interested in a supremum. Ultimately, what we really want is to find an $x_\epsilon(t)$ such that $\|A x_\epsilon\|/\|x_\epsilon\| \to 1$ as $\epsilon \to 0^+$. Intuitively, such a function should have most of its "weight" on the left side.

In particular, consider $$ x_\epsilon(t) = \begin{cases} 1/\sqrt{\epsilon} & t < \epsilon\\ 0 & t \geq \epsilon \end{cases} $$ Calculate $\|x_\epsilon\|^2 = 1$, whereas $$ Ax_\epsilon(t) = \begin{cases} t/\sqrt{\epsilon} & t < \epsilon\\ 1/\sqrt{\epsilon} & t \geq \epsilon \end{cases} \\ \|Ax_\epsilon\|^2 = (1 - \epsilon)/\epsilon + \int_0^\epsilon (t/\epsilon)^2\,dt $$

For $b$: in fact, the norm of the operator is $1$. Consider $$ x(t) = \begin{cases} 1 & t \leq \lambda\\ 0 & t \geq \lambda \end{cases} $$ Verify that $Ax = x$, so that $\|Ax\| = \|x\|$. As a linear operator, we might regard this $A$ as an orthogonal projection onto a subspace.


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.