Question: Let $\mathcal{P}$ be the space of all polynomials (with real coefficients) on the real line, endowed with sup-norm (i.e., $\|p\| = \sup_{0\le x\le 1}|p(x)|$).

For any fixed $n \in \mathbb{N}$, consider the linear functional $\ell_n \colon \mathcal{P} \to \mathbb{R}$, where $\ell_n(p)$ is equal to the coefficient of $x^n$ in $p$. Is $\ell_n$ a bounded linear functional on this normed (but incomplete) space?

Attempt: Well I can see that $\ell_0$ is a bounded linear functional with norm 1, but I don't know the answer in general. Thanks for your help!


The map $\ell_n$ is not bounded for $n\geq1$. I will do the case $n$ odd, but the even case can be done with the same idea.

Consider the function $f_m(x)=\tfrac1m\,\sin(m^2 x)$. It's easy to check that $\|f_m\|=1/m$ and that $$ f_m^{(2n-1)}(0)=m^{4n-3}. $$ Let $p_m$ be the Taylor polynomial of $f_m$, of degree big enough so that $\|f_m-p_m\|<1/m$ on $[0,1]$. Then $$ \|p_m\|\leq\|p_m-f_m\|+\|f_m\|\leq\frac2m, $$ while $$ \ell_{2n-1}(p_m)=\frac{m^{4n-3}}{(2n-1)!}. $$ As we can do this for all $m\in\mathbb N$, we obtain a sequence $\{p_m\}$ with $\|p_m\|\to0$ and $\ell_{2n-1}(p_m)\xrightarrow[m\to\infty]{}\infty$.

  • $\begingroup$ Well, I don't see why ‖p‖ is what you claimed. ‖p‖ must be at least p(1) = 1-c, so if you choose c to be small, won't ‖p‖ be very close to 1? $\endgroup$ – zzz Jan 30 '19 at 2:46
  • $\begingroup$ You want $c$ to be big, but you are right, my example doesn't work. $\endgroup$ – Martin Argerami Jan 30 '19 at 2:47
  • $\begingroup$ Please check the new version. $\endgroup$ – Martin Argerami Jan 30 '19 at 3:38
  • $\begingroup$ Thanks! This is a nice counterexample. $\endgroup$ – zzz Jan 30 '19 at 5:35

Hint : Express your functional as :

$$\ell_n(x) = a_nx^n + a_{n-1}x^{n-1} + \dots + a_1x +a_0$$

Now, take $|\ell_n(x)|$ and try to form an inequality involving the sup norm. Triangle inequality will be your friend !

  • 2
    $\begingroup$ Contrariwise, I understood the defintion to be $$ \ell_n(a_mx^m + a_{m-1}x^{m-1} + \dots + a_1x +a_0) = a_n $$ $\endgroup$ – GEdgar Jan 30 '19 at 1:50
  • $\begingroup$ @GEdgar Yes, $\ell_n$ is defined as this $\endgroup$ – zzz Jan 30 '19 at 2:48
  • $\begingroup$ -1: I fail to see how this can possibly lead to an answer. I'll be happy to be proven wrong. $\endgroup$ – Martin Argerami Jan 30 '19 at 18:00

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.