5
$\begingroup$

Consider $f\in H^{\sigma}(S^1)=W^{\sigma, 2}$ (the usual Sobolev space on the circle) and let $S_Nf$ be its truncated Fourier series $S_Nf = \sum_{|n|\leq N} \hat{f}(n)e^{2\pi i n x}$. I am looking for the theory for inequalities of the following type: $$\|f-S_Nf\|_{H^s} \leq CN^{c(s,\sigma)} \|f\|_{H^{\sigma}} \, ,$$ What values of $s>0$ work for any given $\sigma >0$? What is known of the constant $C$ and of the rate $c(s,\sigma )$?

Motivation: I am familiar with the equivalent theory for orthogonal polynomial expansions (see Canuto and Quarteroni, Math. Comp. 1982), but figured out that the Fourier theory is older, and maybe simpler.

$\endgroup$
4
  • 2
    $\begingroup$ See Lemma A. 43 in [A. Kirsch: An Introduction to the Mathematical Theory of Inverse Problems. Springer, New York, 1996]. $\endgroup$
    – Yidong Luo
    Commented Mar 23, 2020 at 4:06
  • $\begingroup$ Thanks! Is this tight, in terms of spaces, rates, and constants? $\endgroup$
    – Amir Sagiv
    Commented Mar 23, 2020 at 14:50
  • $\begingroup$ @YidongLuo my comment above $\endgroup$
    – Amir Sagiv
    Commented Mar 23, 2020 at 15:00
  • $\begingroup$ The rates is tight for the periodic Sobolev space $ H^r, r \in \mathbb{R} $ in terms of the regular estimate proof. $\endgroup$
    – Yidong Luo
    Commented Mar 24, 2020 at 3:33

1 Answer 1

5
$\begingroup$

Let us start with pointing out that $f\in H^\sigma$ is equivalent to $$ (\langle n\rangle^\sigma\hat f(n))_{n\in \mathbb Z}\in \ell^2(\mathbb Z), \quad \text{with $\langle n\rangle=\sqrt{1+n^2}$.} $$ Then you have for $s<\sigma$ $$ \Vert f-S_N(f)\Vert_{H^s}^2=\sum_{\vert n\vert> N}\langle n\rangle^{2s}\vert\hat f(n)\vert^2= \sum_{\vert n\vert> N}\langle n\rangle^{2\sigma}\vert\hat f(n)\vert^2\langle n\rangle^{2s-2\sigma}\le\langle N\rangle^{2s-2\sigma}\Vert f\Vert^2_{H^\sigma}. $$

$\endgroup$
1
  • 1
    $\begingroup$ Yes, in my opinion, this is the most sensible answer possible here. Also, I think it is essentially sharp, apart from being the natural argument. $\endgroup$ Commented Apr 23, 2021 at 19:25

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .