0
$\begingroup$

Consider the Laplacian $-\Delta: C^\infty (S^1) \to C^\infty(S^1)$ where $\mathbb R/\mathbb Z=S^1$ and for a periodic function $f:\mathbb R\to\mathbb R$ we have $-\Delta f=-f''$.

For the orthonormal basis $\{\phi_n=e^{inx}\}_{n\in \mathbb Z}$ of $L^2(S^1)$ we have $-\Delta \phi_n= n^2 \phi_n $, so $\lambda_n=n^2$ are eigenvalues of $-\Delta$.

One can define the resolvent $(1-\Delta)^{-1}f= \sum_{n\in \mathbb Z} \langle f, \phi_n \rangle (1+\lambda_n)^{-1} \phi_n $ and more generally for $\delta>0$ we have \begin{align} (1-\Delta)^{-\delta}f(x)&= \sum_{n\in \mathbb Z} \langle f, e^{in(\cdot)} \rangle (1+n^2)^{-\delta} e^{inx} \\ &= \sum_{n\in \mathbb Z}\hat f(n) (1+n^2)^{-\delta} e^{inx} \end{align}

which makes sense e.g. for $L^2$ functions, or for distributions.

I want to compute an integral kernel for $(1-\Delta)^{-\delta}$, which according to a paper I am reading has a singularity of order $|x-y|^{2\delta-1}$. What is a justification for that?

EDIT: the kernel should be $K(x,y)=\sum_{n\in \mathbb Z} (1+n^2)^{-\delta} e^{inx} e^{iny}$ but how can this be converted to an expression that includes $|x-y|^{2\delta-1}?$

I suspect that we should obtain an expression of the form $$(1-\Delta)^{-\delta}f(x)= \int_0^{2\pi} \bigg(\textrm{linear expression of $f(x+y),f(x-y),f(x)...$} \bigg) \tilde K(x,y)dy $$ where $\tilde K$ has a singularity $|x-y|^{2\delta-1}$. I am also a bit worried about the periodicity of the kernel. Any ideas would be appreciated

$\endgroup$
3
  • $\begingroup$ One way to proceed — perhaps the most elementary one — is via the heat kernel and Bochner's subordination: $(I-\Delta)^{-\delta}=\frac{1}{\Gamma(\delta)} \int_0^\infty t^{-1+\delta} e^{-t(I-\Delta)} dt$, and we have good bounds for the kernel of $e^{-t(I-\Delta)}=e^{-t} e^{t\Delta}$. $\endgroup$ Commented Aug 5, 2021 at 22:23
  • $\begingroup$ First of all, since $(1-\Delta)^{-\delta}$ is multiplication on the Fourier side, it is convolution by the FT $\sum (1+n^2)^{-2\delta} e^{ixt}$ (in your formula for $K$, one of the signs of $x,y$ is wrong). $\endgroup$ Commented Aug 5, 2021 at 23:52
  • $\begingroup$ Next, in the continuous case it is well known that (up to constants) the FT of $|x|^{-\alpha}$ is $|t|^{\alpha-d}$. What you want to show looks essentially like a discrete version of this statement. $\endgroup$ Commented Aug 5, 2021 at 23:54

0

You must log in to answer this question.

Browse other questions tagged .