2
$\begingroup$

Let $Q$ be a bounded domain in $\mathbf R^N$ with smooth boundary. Let $f\in C^a(\overline{Q})$, $0<a<1$.

  • Denote $\psi_k(x)$ normalized eigenfunctions and $\lambda_k$ eigenvalues ($k=0,1,2\dotsc$, $\lambda_0=0$, $\psi_0=1/\sqrt{|Q|}$) such that $$ \begin{cases} -\Delta \psi_k= \lambda_k \psi_k,&\text{ in } Q,\\ \dfrac{\partial \psi_k}{\partial n}=0,&\text{ on }\partial \Omega. \end{cases}$$
  • Finally denote $$ S_n(x)=\sum\limits_{k=0}^{n}\int_Q f\psi_k\operatorname{d\!}x\, \psi_k(x). $$

Under what (additional) conditions we can expect $$ \lim\limits_{n\to \infty}\lVert f-S_n\rVert_{C(\overline{Q})}=0\; ? $$

Improve this question
$\endgroup$
10
  • $\begingroup$ I cerrected. Thank you. $\endgroup$
    – user500030
    Commented Feb 26, 2023 at 5:22
  • 2
    $\begingroup$ This is not true in higher dimension, basicly because the sequence of eigenfunctions need not be uniformly bounded. An example of this situation is the Laplace Beltrami on the unit sphere $S^{N-1}$ of $\mathbb R^N$. The eigenfunctions are the spherical harmonics, if $k$ is the degree, the supnorm is like $k^{N-2}$ and the expansion converges in the supnorm if $f \in C^l$ with $l>(N-1)/2$. $\endgroup$
    – Giorgio Metafune
    Commented Feb 26, 2023 at 8:36
  • 1
    $\begingroup$ Please, quote my name if you reply to me, to let me receive a notification. If $-\Delta \psi_k=\lambda_k \psi_k$ then $(-\Delta)^\ell \psi_k=\lambda_k^\ell \psi_k$ and by elliptic estimates the norm of $\psi_k$ in the Sobolev space $H^{2\ell}$ is boounded by $\lambda_k^\ell$. Since $\lambda_k \approx k^{2/N}$ choosing $2\ell >N/2$ by Sobolev embedding you get a rough estimate in the sup norm. I am counting eigenvalues and eigenfunctions with multiplicity (in the comment before all spherical harmonics of degrre $k$ belong to the same $\lambda_k$). $\endgroup$
    – Giorgio Metafune
    Commented Feb 26, 2023 at 15:31
  • 1
    $\begingroup$ @WillieWong One argument is short but uses elliptic regularity. For $\psi \in C^{2\ell}$ write $\psi=\sum_k (\psi, \psi_k)\psi_k$ in $L^2$ and $$\Delta^\ell \psi=\sum_k (\Delta^\ell \psi, \psi_k)\psi_k=\sum_k (\psi, \Delta^\ell \psi_k)\psi_k=\sum_k \lambda^\ell_k ( \psi, \psi_k)\psi_k=\Delta^\ell (\sum_k (\psi, \psi_k)\psi_k).$$ This gives that the series yielding $\psi$ converges in $H^{2\ell}$, by elliptic regularity, and by Sobolev embedding uniformly, for $\ell$ large. $\endgroup$
    – Giorgio Metafune
    Commented Feb 27, 2023 at 18:25
  • 1
    $\begingroup$ @WillieWong The other one I know uses more on spherical harmonics, zonals, sup-norm estimates. If you need, I can write down it. $\endgroup$
    – Giorgio Metafune
    Commented Feb 27, 2023 at 18:27

1 Answer 1

Reset to default
1
$\begingroup$

Hölder functions periodic with period 1 satisfy the Dini criterion $$ \int_0^{1/2}\frac{\vert u(x+t)- u(x)\vert}{t} dt<+\infty, $$ thus their Fourier series are uniformly convergent (towards $u$). On the other hand the Besov space $$ B^\alpha_{\infty, \infty} $$ is equal to the Hölder space $C^\alpha$ for $\alpha\in (0,1)$.

Improve this answer
$\endgroup$
1
  • $\begingroup$ Everything is more or less clear in periodic case.. But I can't find a suitable reference in multidimensional case $\endgroup$
    – user500030
    Commented Feb 26, 2023 at 7:25

You must log in to answer this question.

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