Weighted Littlewood-Paley theory and exponential-square integrability (Q2457371)

Littlewood-Paley theory is an essential tool of Fourier analysis. The present book tries to give a gentle, well-motivated introduction to this theory for graduate students. Chapter 1 covers some elementary facts about dyadic cubes and the Calderón-Zygmund decomposition. Chapters 2, 3 and 4 discuss Littlewood-Paley theory including weighted estimates. Chapters 5,6 and 7 are devoted to the Calderón's reproducing formula. The latter half of this book gives some applications: Applications of weighted Littlewood-Paley theory to the analysis of Schrödinger and singular integral operators, Littlewood-Paley theory on Orlicz spaces, a new proof of the Hörmander-Mihlin multiplier theorem and so on. Exercises are put at the end of almost every chapter. The author emphasizes that one needs to consider weighted estimates to investigate the connection between a function and its square function. Thus the ``weighted'' in the book's title comes from this. He explains the one dimensional dyadic Littlewood-Paley theory and the Calderón's reproducing formula in detail. Let \(h_{(I)}\) be the Haar functions and \( \lambda_{I}(f) = \langle f, h_{(I)} \rangle. \) First the author proves \[ f(x) = \sum_{I} \lambda_{I}(f) h_{(I)} (x) \quad \text{and} \quad \int_{R^1} | f(x) |^2 dx = \sum_{I} | \lambda_{I}(f) |^2. \] By writing \( \sum_{I} | \lambda_{I}(f) |^2 = \int \big( \sum_{I} \frac{ | \lambda_{I}(f) |^2 }{ | I |} \chi_{I}(x) \big)dx, \) he introduces the one-dimensional dyadic square function \[ S(f)(x) = \Big( \sum_{I} \frac{ | \lambda_{I}(f) |^2 }{ | I |} \chi_{I}(x) \Big)^{1/2} \quad \text{and says} \quad \| f \|_{L^2} = \| Sf \|_{L^2}. \] He is not in a hurry to go further. He says as follows. Notice that \(| f |^2 = | \sum_{I} \lambda_{I}(f) h_{(I)} |^2\), while \( S(f)^2 = \sum_{I} | \lambda_{I}(f) h_{(I)} |^2\). Therefore the inequality \( \| f \|_{L^2} \leq \| Sf \|_{L^2} \) is remarkable and this inequality holds because \(\sum_{I}\lambda_{I}(f) h_{(I)}\) has a lot of cancellation. He also explains that this equality shows a direct relation between a signal's amplitude and its energy. In this way his explanation is instructive. He often says ``Before going further, \dots ''. Throughout this book he emphasizes that the connection between a function and its square function is tight. Next he proves the Littlewood-Paley theorem. \[ \text{If}\quad 1<p<\infty \quad \text{then} \quad \| f \|_{L^p} \approx \| S(f) \|_{L^p}. \] He does not give the ordinary short proof using a duality argument. His proof is more involved, but applicable to generalize to small \(p \;( 0 < p \leq 1)\) or to weighted settings. For this purpose he introduces the dyadic maximal function \(f^{*}(x) = \sup_{x \in I} | f_{I} |\) and the dyadic Hardy-Littlewood maximal function \(M_d f(x) = \sup_{x \in I} | f |_{I}\), and explains ``good-\(\lambda\) inequalities''. Next he considers weighted estimates. For this purpose he introduces the Rubio de Francia maximal function for a weight function \(v\); \[ {\mathbf M}_{R} (v) (x) = \sum_{k=0}^{\infty} \frac{M_d^k (v) (x)}{ R^k}, \] and \(Y\)-functional \[ Y(I,v)= \begin{cases} \frac{1}{v(I)} \int_{I} v(x) \log ( e + v(x)/v_I )dx \quad &\text{if}\quad v(I)>0, \\ 1 &\text{if}\quad v(I)=0. \end{cases} \] It is characteristic of this book to explain these two operators in detail. As corollaries of the results in this chapter the following theorems are proved. \[ \begin{aligned} &\text{If} \quad 1 < p \leq 2 \quad \text{then} \quad \int S(f)(x)^p v(x) dx \leq C_{p} \int | f(x) |^p M_d v(x) dx. \\ &\text{If} \quad 0 < p < 2 \quad \text{then} \quad \int f^{*}(x)^p v(x) dx \leq C_{p} \int Sf(x)^p M_d v(x) dx. \\ &\text{If} \quad 0 < p \leq 1 \quad \text{then} \quad \int Sf(x)^p v(x) dx \leq C_{p} \int f^{*}(x)^p M_d v(x) dx. \end{aligned} \] Next he considers Calderón's reproducing formula. Let \(\psi\) be a Littlewood-Paley function. The following formula is called the Calderón's reproducing formula. \[ f(x) = \int_{R^{d+1}_{+}} f*\psi_y (t) \cdot \psi_y (x-t) \frac{dtdy}{y}. \] By using this formula, \(f\) can be written as \[ f= \sum_{Q : \text{dyadic cubes}} \int_{T(Q)} f*\psi_y (t) \cdot \psi_y (x-t) \frac{dtdy}{y} = \sum_{Q : \text{dyadic cubes}} \lambda_Q(f) a_{(Q)}(x), \] where \(T(Q)= \{ (x,y) \in R^{d+1}_{+}; x \in Q , \ell(Q)/2 \leq y < \ell (Q) \} \) and \( \lambda_{Q}(f) = \Bigl( \int_{T(Q)} | f*\psi_y (t) |^2 \frac{dtdy}{y} \Bigr)^{1/2}. \) Another square function is defined by \[ \tilde{S}f(x) = \Bigl( \sum_{Q} \frac{ | \lambda_{Q}(f) |^2}{ | \tilde{Q} |} \chi_{\tilde{Q}}(x) \Bigr)^{1/2}. \] He considers precise estimates for this function including two weighted norm inequalities.