2
$\begingroup$

I'm porting this question over from MSE as it did not get any responses other than one comment on there.

Let $C^{k,\alpha}$ be a Hölder space where $0 \leq \alpha \leq 1$. I have seen various sources informally describe these spaces as functions having at least "$k + \alpha$" derivatives. How does some sense of fractional differentiability follow from the definition of the Hölder norm?

$\endgroup$
12
  • 1
    $\begingroup$ Sobolev embedding embedding theorem. (Rellich-Kondrachov theorem) $\endgroup$
    – Oxonon
    Commented Oct 24, 2023 at 8:01
  • $\begingroup$ The classic reference for this is chapter 6 of: Adams, Robert A., and John JF Fournier. Sobolev spaces. 1975. $\endgroup$
    – Oxonon
    Commented Oct 24, 2023 at 8:03
  • $\begingroup$ One thing. $\alpha$ Holder functions are indeed $\beta$ fractional differentiable for all $\beta<\alpha$. $\endgroup$
    – user479223
    Commented Oct 24, 2023 at 12:25
  • $\begingroup$ Doesn’t the definition indicate this? $\endgroup$
    – Deane Yang
    Commented Oct 24, 2023 at 12:56
  • 1
    $\begingroup$ To follow up on @PietroMajer: math.stackexchange.com/questions/2776879/… $\endgroup$ Commented Oct 24, 2023 at 15:23

2 Answers 2

2
$\begingroup$

Let $D\subset\mathbb{R}^n$ be a bounded domain. The fractional Sobolev space $W^{s,p}(D)$, $0<s<1$, $1<p<\infty$ is defined as the space of all $f\in L^p(D)$ such that $$ [f]_{s,p}=\left(\int_D\int_D\frac{|f(x)-f(y)|^p}{|x-y|^{n+sp}}\, dxdy\right)^{1/p}<\infty. $$ The space is equipped with the norm $\Vert f\Vert_{s,p}=\Vert f\Vert_p+[f]_{s,p}$.

It is easy to prove (by a straightforward estimate of the integral that $C^{0,\alpha}(D)\subset W^{1-\frac{1}{p},p}(D)$ whenever $1<p<\infty$ and $1-\frac{1}{p}<\alpha\leq 1$. In that sense H"older functions belong to some fractional Sobolev spaces.

You can find some basic facts in

Di Nezza, Eleonora; Palatucci, Giampiero; Valdinoci, Enrico, Hitchhiker’s guide to the fractional Sobolev spaces, Bull. Sci. Math. 136, No. 5, 521-573 (2012). ZBL1252.46023.

$\endgroup$
5
  • $\begingroup$ I still don’t understand why the definition of Holder itself indicates it is a fractional derivative. $\endgroup$
    – Deane Yang
    Commented Oct 24, 2023 at 19:22
  • $\begingroup$ @DeaneYang To discuss fractional derivatives in the setting of $W^{s,p}$ spaces one would have to go through the Fourier transform. While it is relatively elementary when $p=2$, it is more difficult when $p\neq 2$ as it involves singular integrals. This is quite complicated and lengthily overall and I am not going to write about it since it would require me to spend hours refreshing my memory. Sorry for that. $\endgroup$ Commented Oct 24, 2023 at 19:35
  • $\begingroup$ What I’m trying to ask is why do we have to mention Sobolev spaces at all? $\endgroup$
    – Deane Yang
    Commented Oct 24, 2023 at 19:39
  • $\begingroup$ It is not true that you need to define fractional derivatives through Fourier transform. $\endgroup$
    – user479223
    Commented Oct 24, 2023 at 19:51
  • $\begingroup$ @user479223 Yes and no. The case $n=1$ is somewhat easy. The case when $n>1$ is more delicate. You can define the fractional power of the Laplace operator through an integral formula, but if you want to prove properties of it, you need to use Fourier transform. $\endgroup$ Commented Oct 24, 2023 at 20:14
4
$\begingroup$

Theorem 14 in https://eudml.org/doc/167995 says that if $f\in C^\alpha$ and $\beta<\alpha$ then $D^\beta f$ exists and we have that $D^\beta f\in C^{\alpha-\beta}$, where $D^\beta$ is the fractional derivative.

For example $D^\beta x^\alpha=\operatorname{Const} \cdot x^{\alpha-\beta}$.

See:

https://math.stackexchange.com/questions/2685689/does-taking-a-fractional-derivative-remove-a-fractional-amount-of-holder-regular

$\endgroup$

You must log in to answer this question.

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