Let $\{p_{n}\}$ be a sequence of polynomials and $f$ a continuous function on $[0,1]$ such that $\int\limits_{0}^{1}|p_{n}(x)-f(x)|dx\to 0$. Let $c_{n,k}$ be the coefficient of $x^{k}$ in $p_{n}(x)$. Can we conclude that $\underset{n\rightarrow \infty }{\lim }c_{n,k}$ exists for each $k$?.

What I know so far: if the degrees of $p_{n}^{\prime }s$ are bounded then this is true. In fact, we can replace $L^{1}$ convergence by convergence in any norm on $C[0,1]$; to see this we just have to note that for fixed $N$, $% \sum_{k=0}^{N}c_{i}x^{i}\rightarrow (c_{0},c_{1},...,c_{N})$ is a linear map on a finite-dimensional subspace and hence it is continuous. My guess is that the result fails when there is no restriction on the degrees. But if $p_{n}(z)$ converges uniformly in some disk around $0$ in the complex plane then the conclusion holds. To construct a counterexample we have to avoid this situation. Maybe there is a very simple example but I haven't been to find one. Thank you for investing your time on this.


1 Answer 1


Consider the sequence of polynomials

$$ p_n(x)=\left\{ \begin{array}{@{}ll@{}} (1-x)^n, & \text{if}\ n\equiv 0 \mod 2 \\ x^n, & \text{if}\ n\equiv 1 \mod 2 \end{array}\right. $$

Then $p_n(x)$ converges to $0$ but the constant term is oscillating.

  • 9
    $\begingroup$ Thanks. I just noticed that teh conclusion fails even with uniform convergence. Consider $x(1-x)p_n(x)$ with the $p_n$' you have defined and look at the coefficient of $x$. $\endgroup$ Mar 18, 2019 at 8:41
  • 4
    $\begingroup$ @KaviRamaMurthy: Or, use Ethan's sequence with the interval $[1/4, 3/4]$, on which it converges uniformly. $\endgroup$ Mar 19, 2019 at 0:36
  • 1
    $\begingroup$ What if the polynomial converges to something other than $0$ though? $\endgroup$
    – Jam
    Mar 23, 2019 at 16:07
  • 2
    $\begingroup$ @Jam You can make it work if $f$ is any function that can be approximated by polynomials. Indeed, if $g_n$ is a sequence of polynomials converging to $f$, then we can assume $g_n(0)$ converges to $f(0)$ since otherwise we're done. But then $p_n+g_n$ still converges to $f$ while $p_n(0)+g_n(0)$ has two subsequences which converge to $f(0)$ and $f(0)+1$ respectively. $\endgroup$ Mar 23, 2019 at 20:47

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.