Let $C[a,b]$ denote the normed space of all the continuous (real or complex-valued) functions defined (and continuous!) on the closed interval $[a,b]$ on the real line, where $a, b \in \mathbb{R}$ and $a < b$, with the norm given by $$\Vert x \Vert \colon= \max_{t \in [a,b]} \left\vert x(t) \right\vert \ \mbox{ for all } \ x \in C[a,b].$$ Let $y_0$ be a given element of $C[a,b]$, and let $\alpha$ and $\beta$ be some given scalars.

Let the functionals $f_1$ and $f_2$ be defined as follows: $$f_1(x) \colon= \int_a^b \ x(t) y_0(t) \ \mathrm{d} t \ \mbox{ for all } \ x \in C[a,b],$$ $$f_2(x) \colon= \alpha x(a) + \beta x(b) \ \mbox{ for all } \ x \in C[a,b].$$ Then both $f_1$ and $f_2$ are bounded linear functionals on $C[a,b]$. Moreover, we have for all $x \in C[a,b]$ $$ \begin{align} \left\vert f_1(x) \right\vert &= \left\vert \int_a^b \ x(t) y_0(t) \ \mathrm{d} t \ \right\vert \\ &\leq \left( \int_a^b \left\vert y_0(t) \right\vert \ \mathrm{d} t \right) \Vert x \Vert, \end{align} $$ showing that $$\Vert f_1 \Vert \leq \int_a^b \left\vert y_0(t) \right\vert \ \mathrm{d} t.$$ And, $$ \Vert f_2 \Vert \leq \vert \alpha \vert + \vert \beta \vert. $$

What next?

  • $\begingroup$ The norm of the linear functional is the variation of the representing complex measure. $\|f_1\|=|\alpha|+|\beta|$. $\|f_2\|=\int_{a}^{b}|y_0(t)|dt$. $\endgroup$ – DisintegratingByParts Jun 23 '16 at 11:05

Ideally you need to find points where these suprema are attained. This is possible for $f_2$ as $\exists x \in C[a,b]$ such that $\|x\| = 1$ $$ x(a) = \text{sgn}(\alpha) \text{ and } x(b) = \text{sgn}(\beta) $$ where $\text{sgn}(z) = z/|z|$ (This follows from Urysohn's lemma, if you like, although one can just draw the graph of such a function on $[a,b]$). Now for this $x$, one has $$ f_2(x) = |\alpha| + |\beta| $$ so your inequality becomes an equality.

For $f_1$ it gets more complicated as the function $t\mapsto \text{sgn}(y_0(t))$ may not be continuous as $y_0$ may have zeroes. To fix this, for each $\epsilon > 0$, consider $x_{\epsilon} \in C[a,b]$ by $$ x_{\epsilon}(t) = \frac{\overline{y_0(t)}}{|y_0(t)| + \epsilon} $$ so that if $M := \int_a^b |y_0(t)|dt$, then $$ M - \epsilon(b-a) = \int_a^b (|y_0(t)| - \epsilon) dt = \int_a^b \frac{|y_0(t)|^2 - \epsilon^2}{|y_0(t)| + \epsilon} \leq \int_a^b y_0(t) x_{\epsilon}(t)dt $$ Hence $$ M - \epsilon(b-a) \leq f_1(x_{\epsilon}) \leq \|f_1\| $$ since $\|x_{\epsilon}\| \leq 1$. This is true for all $\epsilon > 0$, so one gets $M \leq \|f_1\|$, proving that $\|f_1\| = M$ as you have guessed.


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.