It is a classical result in harmonic analysis that

$$ \|\|P_kf\|_{\ell^2_k}\|_{L^p_x}\approx\|f\|_{L^p} $$

for $p\in(1,\infty)$, where $P_k$ is the Littlewood-Paley decomposition onto frquency $\approx 2^k$.

What if I replace the $\ell^2$ norm in $k$ by the $\ell^q$ norm? Is it possible that in addition to the trivial estimate derived from the embedding of $\ell^q$ in $\ell^2$ (or vice versa), there are some nontrivial value of $(p,q)$ for this to hold?

  • 1
    $\begingroup$ What is $P_kf$? $\endgroup$ Nov 14, 2017 at 9:36
  • 1
    $\begingroup$ It seems that you are asking whether the quasi norm $$\|f\|_{\dot{F}_{p,q}^{0}} := \| \|P_{k}f\|_{\ell_{k}^{q}}\|_{L_{x}^{p}}$$ for the homogeneous Triebel-Lizorkin space $\dot{F}_{p,q}^{0}$ is equivalent to the Lebesgue $L^{p}$ norm. I do not know have a copy of H. Triebel's Theory of Function Spaces in front of me, but I would suggest looking in there. $\endgroup$ Nov 14, 2017 at 16:00

1 Answer 1


The following counterexamples are taken from examples 6.1.10 and 6.1.11 in the textbook L. Grafakos, Classical Fourier Analysis (Third Edition).

Claim 1. Fix $1<p<\infty$ and $q<2$. Then the inequality $$\| (\sum_{j\in\mathbb{Z}} |P_{j}(f)|^{q})^{1/q}\|_{L^{p}} \lesssim_{p,q} \|f\|_{L^{p}}$$ cannot hold.

Proof. Let $\psi$ be a $C^{\infty}$ function on $\mathbb{R}$ such $\hat{\psi}\geq 0$,

$$\hat{\psi}(\xi) = \begin{cases} 1 & {\xi \in [9/8,15/8]} \\ 0 & {\xi\notin [7/8,17/8]} \end{cases},$$ and $$\sum_{j\in\mathbb{Z}} \hat{\psi}(2^{-j}\xi)^{2}=1, \qquad \xi>0.$$ Extend $\hat{\psi}$ to be an even function on $\mathbb{R}$. Let $P_{j}:=\psi(2^{-j}D)$.

Now let $\varphi$ be a Schwartz function such that $\hat{\varphi}\geq 0$ and $\mathrm{supp}(\hat{\varphi}) \subset [11/8,13/8]$. Fix $N\gg 1$, and define

$$f_{j}(x) := e^{2\pi\frac{12}{8}2^{j}x}\varphi(x), \qquad j=1,\ldots,N.$$

Since $\widehat{f}_{j}(\xi) = \hat{\varphi}(\xi-\frac{12}{8}2^{j})$ is supported on the set $$[\frac{11}{8} + \frac{12}{8}2^{j}, \frac{13}{8}+\frac{12}{8}2^{j}]\subset [\frac{9}{8}2^{j}, \frac{15}{8}2^{j}], j\gg 1,$$ $\hat{\psi}(2^{-j}\xi)=1$ on the support of $\hat{f}_{j}$.

Now define $$f := \sum_{j=3}^{N}f_{j}.$$ We have the lower bound $$\|(\sum_{j}|P_{j}(f)|^{q})^{1/q}\|_{L^{p}}\geq \|(\sum_{j=3}^{N} |f_{j}|^{q})^{1/q}\|_{L^{p}} = (N-3)^{1/q}\|\varphi\|_{L^{p}}.$$ By the Littlewood-Paley inequality, we have the upper bound $$\|f\|_{L^{p}} \lesssim_{p} \|(\sum_{j}|P_{j}f|^{2})^{1/2}\|_{L^{p}} = (N-3)^{1/2}\|\varphi\|_{L^{p}}.$$

For $N\gg 1$, we obtain a contradiction. $\Box$

Claim 2. Fix $1<p<\infty$ and $2<q<\infty$. Then the inequality $$\|g\|_{L^{p}} \lesssim_{p,q} \|(\sum_{j} |P_{j}g|^{q})^{1/q}\|_{L^{p}}$$ cannot hold.

Proof. This follows from duality and Claim 1. $\Box$.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.