6
$\begingroup$

I'm reading an article for my future thesis (I'm a third-year undergraduate) where the authors define the generalized Holder Spaces as a special class of Besov Spaces.

Define $\chi,\tilde{\chi}\in C_{c}^{\infty}(\mathbb{R}^d)$ such that Supp$(\chi)\subset B(0,8/3) \setminus B(0,3/4)$ and Supp$(\tilde{\chi}) \subset B(0,4/3),$ and such that $\tilde{\chi}(x)+\sum^{+ \infty}_{k=0} \chi(x/2^k)=1 \quad \forall x\in \mathbb{R}^d$.

Define $\chi_{-1}:=\tilde{\chi},$ and $\chi_k(\cdot):=\chi(\cdot/2^k).$ Now, $\forall f \in C^{\infty}(\mathbb{T}^d),$ set $\delta_k(f):=\mathscr{F}^{-1}(\hat{f} \cdot \chi_k$), where $\hat{f}(k):=\int_{\mathbb{T}^d}f(x)e^{-2\pi i k\cdot x} dx$ and $\mathscr{F}^{-1}(g)(x):=\sum_{k\in \mathbb{z}^d} g(k)e^{2 \pi i k\cdot x}.$ Heuristically, $\delta_k(f)$ is just a part of the frequencies of the smooth function $f.$

Now, for every $\alpha \in \mathbb{R},$ we can define the $\mathcal{C}^{\alpha}$ norm of a smooth function, which is $\|f\|_{\mathcal{C}^{\alpha}}:=sup_{k\geq1} 2^{\alpha k} \|\delta_k(f)\|_{L^{\infty}}.$

The space $\mathcal{C}^{\alpha}$ is defined as the completion of $C^{\infty}(\mathbb{T}^d)$ with respect to this norm.

The definition given is a bit different from the one which is mostly given in literature: the space of all tempered distributions such that the above-mentioned norm (which is well-defined because the Fourier antitransform of a compactly-supported distribution is a function) is finite.

Now, me questions are:

  1. Why the $\mathcal{C}^{\alpha}$ norm is finite for every smooth function defined on the torus?

  2. Why does it coincide (well, I don't think they precisely coincide, but they should be equivalent or at least generate the same completion) with the classical $\alpha-$Holder norm ($\|f\|_{\mathcal{C}^{\alpha}}=sup_{x\neq y} \frac{|f(x)-f(y)|}{|x-y|^{\alpha}})$?

  3. Do the $\mathcal{C}^{\alpha}$ spaces respectively defined via completion of $C^{\infty}$ and via the tempered distributions coincide?

The article can be found here.

$\endgroup$

1 Answer 1

Reset to default
5
$\begingroup$
  1. As for the first point I will just provide a hint: you can write each Littlewood-Paley block $\delta_k(f)$ as a convolution between the distribution $f$ and a smooth function $\varrho_k$ which I leave to you to think about. Crucial point is that $\varrho_k(x) = 2^{dk}\varrho(2^kx)$ for every $k$ and a fixed smooth function $\varrho.$ Now since $\varrho$ has zero Fourier coefficients near zero you can invert the laplacian, i.e. for any $\zeta \in \mathbb{N}$ find $\varphi$ such that $\Delta^\zeta \varphi = \varrho$ and now you can use integration by parts to find your bound (note that the $\mathcal{C}^{\alpha}$ norm is bounded by the supremum norm for $\alpha < 0$).
  2. This is a theorem, for $\alpha \in (0,1)$. A reference could be the book by Chemin Bahouri and Danchin (Fourier Analysis and nonlinear PDEs) or some lecture notes on the net. BUT: you do need to take the space defined through tempered distributions,NOT the completion of $C^{\infty}.$ Why is the content of the next point.
  3. $C^{\infty}$ is not dense in $\mathcal{C}^{\alpha}$ if you define the latter with tempered distributions. You should think of the difference between Lipschitz functions and $C^1$ functions. A striking example in the non-integer case is provided here. The result that makes everything work AS THOUGH $C^{\infty}$ would be dense is interpolation: if $f_n \to f$ in the sense of distributions and $$\sup_n\| f_n\|_{\mathcal{C}^{\alpha}} < {+} \infty$$ then $f_n \to f$ in $\mathcal{C}^{\beta}$ for any $\beta < \alpha.$
$\endgroup$
2
  • $\begingroup$ Thanks a lot! Now I'm going to sleep but tomorrow I'll ask you further questions if I have some doubts :) $\endgroup$ Commented Mar 21, 2018 at 23:29
  • $\begingroup$ If I understand the argument, can we proceed as follows? We fix an integer $\beta > d + \frac{\alpha}{2}$, then we can write \begin{align*} \delta_k f &= \delta_k (-\Delta)^{-\beta} g, \quad\text{with}\quad g=(-\Delta)^{\beta} f \in C^\infty(\mathbb{T}^d) \\ &=\sum_{|n|\sim 2^k} \chi(n/2^k) |n|^{-2\beta} \widehat{g}(n) e_n. \end{align*} Its $L^\infty$-norm is bounded by $$ \leq C_{\chi} 2^{-2\beta k + dk} \sum_{n\in\mathbb{Z}^d} | \widehat{g}(n)| \lesssim 2^{-2\beta k + dk} $$ This would give us $2^{\alpha k} \| \delta_k f\|_\infty$ uniformly bounded in $k$. $\endgroup$ Commented Sep 10, 2021 at 16:06

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.