As usual, $\mathbb R[x]$ denotes the vector space of polynomials in one variable with real coefficients. It is easy enough (and a good exercise for beginners) to prove that the function $$P\to\|P\|=\max_{x\in [0,1]} |P(x)|$$ of $\mathbb R[x]$ in $\mathbb R_+$ defines a norm on $\mathbb R[x]$ (then we can speak about continuity).

Prove that for all $x_0\in \mathbb R; x_0\gt 1$ the function $f_{x_0}$ of $\mathbb R[x]$ in $\mathbb R$ defined by $$f_{x_0}(P)= P(x_0)$$ is discontinuous in every point $P\in \mathbb R[x]$

  • 1
    $\begingroup$ It is false. If $x_0\in[0,1]$, then $f_{x_0}$ is continuous $\endgroup$ – sinbadh Mar 16 '16 at 14:51
  • 1
    $\begingroup$ This is not true. In fact $f_{x_0}$ is continuous if and only if $x_0\in[0,1]$. $\endgroup$ – David C. Ullrich Mar 16 '16 at 14:52
  • $\begingroup$ If you want a norm such that every $f_{x_0}$ is discontinuous you could for example define $||P||_1=\int_0^1|P(t)|\,dt$. $\endgroup$ – David C. Ullrich Mar 16 '16 at 14:53
  • $\begingroup$ I have edited the lapsus. Thank you very much. $\endgroup$ – Piquito Mar 16 '16 at 15:03
  • $\begingroup$ Pay attention please before deny. The question is true now. $\endgroup$ – Piquito Mar 16 '16 at 15:06

Lemma: every continuous linear functional on a normed vector space is bounded, i.e. it has a finite operator norm.

Now the idea is that there are continuous functions which vanish on $[0,1]$ and are large at $x_0$, and therefore there are polynomials which are close to zero on $[0,1]$ but are large at $x_0$. This conclusion follows by Weierstrass' theorem. For details, let $f(x)=0$ on $[0,1]$, $1$ at $x_0$, and linear in between (the linearity is not important). Find a polynomial $p_n$ which is uniformly within $1/n$ of $f$. Then $P(p_n)=1$, but $\| p_n \| \leq 1/n$. Thus $\| P \|$, if $P$ were continuous, would need to be at least $n$, but this is impossible.

  • $\begingroup$ Good. Whitout use of explicit theory, the sequence $\{P_n\}_{n\in \mathbb N}$ defined by $P_n(x)=(\frac xa)^n$; $1\lt a\lt x_0$ is such that $||P_n||=(\frac 1a)^n\to 0$ but $P_n(x_0)=(\frac {x_0}{a})^n\to \infty$ $\endgroup$ – Piquito Mar 16 '16 at 15:51
  • $\begingroup$ @Piquito Nice, that example does indeed work well. It also exposes the important fact that these examples require arbitrarily large degree. With a fixed maximum degree, the space is complete and homeomorphic to some $\mathbb{R}^n$, so none of these phenomena can occur. $\endgroup$ – Ian Mar 16 '16 at 16:04
  • $\begingroup$ In finite dimension there are not examples of discontinuous linear function. In other words, what you say yourself in your comment. Best regards. $\endgroup$ – Piquito Mar 16 '16 at 17:01

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.