Multiple vector-valued, mixed-norm estimates for Littlewood-Paley square functions (Q2154816)

This paper is concerned with multiple vector-valued, mixed-norm estimates for Littlewood-Paley Square Functions. The authors answer a question that came up implicitly in their recent works [Anal. PDE 9, No. 8, 1931--1988 (2016; Zbl 1361.42005); J. Funct. Anal. 273, No. 4, 1295--1353 (2017; Zbl 1392.42024); Rev. Mat. Iberoam. 37, No. 6, 2037--2118 (2021; Zbl 1472.42011)] and completes naturally classical results of the Littlewood-Paley theory. The classical Littlewood-Paley theory states that the following inequalities hold for \( 1<p<\infty\): \[ \|f\|_{L^{p}\left(\mathbb{R}^{m}\right)} \lesssim\left\|S_{m} f\right\|_{L^{p}\left(\mathbb{R}^{m}\right)} \lesssim\|f\|_{L^{p}\left(\mathbb{R}^{m}\right)}\tag{1}. \] The left-hand side of (1) can be available in the whole range \(0<p<\infty \), see for instance [\textit{C. Muscalu} and \textit{W. Schlag}, Classical and multilinear harmonic analysis. Volume II. Cambridge: Cambridge University Press (2013; Zbl 1282.42001); \textit{E. M. Stein}, Harmonic analysis: Real-variable methods, orthogonality, and oscillatory integrals. With the assistance of Timothy S. Murphy. Princeton, NJ: Princeton University Press (1993; Zbl 0821.42001)]. Using standard duality and vector-valued arguments for singular integrals, the authors extend (1) to the setting of mixed-norm spaces and \(N\)-parameter square functions. That is the following inequalities hold for \(L^{Q}\)-valued Schwartz functions defined in \(\mathbb{R}^{d}\) for every \(n\)-tuple \(Q\) and \(d\)-tuple \(P\) satisfying \(1<P\), \(Q<\infty\) componentwise: \[ \|f\|_{L^{P}\left(L^{Q}\right)} \lesssim\|S f\|_{L^{P}\left(L^{Q}\right)} \lesssim\|f\|_{L^{P}\left(L^{Q}\right)}. \] The square function to be studied in the present article is defined as follows. Recall that for any \(N \geq 1\) such Littlewood-Paley sequences \(\left(\psi_{k}^{j}\right)_{k \in \mathbb{Z}}\) defined on \(\mathbb{R}^{d_{j}}\) for \(1 \leq j \leq N\), one defines an \(N\)-parameter \(\left(\Psi_{k}\right)_{k \in \mathbb{Z}^{N}}\) on \(\mathbb{R}^{d}:=\mathbb{R}^{d_{1}} \times \cdots \times \mathbb{R}^{d_{N}}\) by \[ \Psi_{k}:=\psi_{k_{1}}^{1} \otimes \cdots \otimes \psi_{k_{N}}^{N} \] for \(k=\left(k_{1}, \ldots, k_{N}\right)\) , where \[ \psi_{k_{1}}^{1} \otimes \cdots \otimes \psi_{k_{N}}^{N}\left(x_{1}, \ldots, x_{N}\right):=\psi_{k_{1}}^{1}\left(x_{1}\right) \cdot \ldots \cdot \psi_{k_{N}}^{N}\left(x_{N}\right). \] Any Schwartz function on \(\mathbb{R}^{d} \) can be decomposed into \[ f=\sum_{k \in \mathbb{Z}^{N}} f\ast \Psi_{k}. \] Therefore, the \(N\)-parameter square function \(S f\) can be defined as follows: \[ S f(x):=\left(\sum_{k \in \mathbb{Z}^{N}}\left|f\ast \Psi_{k}(x)\right|^{2}\right)^{1 / 2} \] for \(x=\left(x_{1}, \ldots, x_{N}\right) \in \mathbb{R}^{d} \) with \(x_{j} \in \mathbb{R}^{d_{j}}\) for \(1 \leq j \leq N\). On the other hand, we also recall that given any \(n \geq 1\), \(\sigma\)-finite measurable spaces \(\left(A_{j}, \Sigma_{j}, \mu_{j}\right)\) for \(1 \leq j \leq n\) and \(R=\left(r_{1}, \ldots, r_{n}\right)\), an \(n\)-tuple of positive real numbers, one can define the iterated (or mixed-norm) Lebesgue space \(L^{R}(A, \Sigma, \mu)\) to be the space containing those functions \(g\) which are measurable on the product space \[ (A, \Sigma, \mu):=\left(\prod_{j=1}^{n} A_{j}, \prod_{j=1}^{n} \Sigma_{j}, \prod_{j=1}^{n} \mu_{j}\right) \] and for which the (quasi)-norm \(\|g\|_{R}\) defined by \[ \|g\|_{R}:=\|g\|_{L^{R}(A, \Sigma, \mu)}=\|\ldots\| g\left(a_{1}, \ldots, a_{n}\right)\left\|_{L^{r_{n}}\left(A_{n}, \Sigma_{n}, \mu_{n}\right)} \ldots\right\|_{L^{r_{1}}\left(A_{1}, \Sigma_{1}, \mu_{1}\right)} \] is finite. More specifically, this paper studies the mixed (quasi)-norm \(|h\|_{L^{P}\left(L^{Q}\right)}\) given by \[ \|h\|_{L^{P}\left(L^{Q}\right)}:=\|\| h(x, a)\|_{L^{Q}(A, \Sigma, \mu)}\|_{L^{P}\left(\mathbb{R}^{d}\right)}. \] The authors, based on the helicoidal method introduced in [\textit{C. Benea} and \textit{C. Muscalu}, Anal. PDE 9, No. 8, 1931--1988 (2016; Zbl 1361.42005); J. Funct. Anal. 273, No. 4, 1295--1353 (2017; Zbl 1392.42024); Rev. Mat. Iberoam. 37, No. 6, 2037--2118 (2021; Zbl 1472.42011)], proved that for any \(L^{Q}\) -valued Schwartz function \(f\) defined on \(\mathbb{R}^{d}\), one has the multiple vector-valued, mixed-norm estimate \[ \|f\|_{L^{P}\left(L^{Q}\right)} \lesssim\|S f\|_{L^{P}\left(L^{Q}\right)} \] valid for every \(d\)-tuple \(P\) and every \(n\)-tuple \(Q\) satisfying \(0<P\), \(Q<\infty\) componentwise. Here \(S:=S_{d_{1}} \otimes \cdots \otimes S_{d_{N}}\) is a tensor product of several Littlewood-Paley square functions \(S_{d_{j}}\) defined on arbitrary Euclidean spaces \(\mathbb{R}^{d_{j}}\) for \(1 \leq j \leq N\), with the property that \(d_{1}+\cdots+d_{N}=d\). Theorem 1.1 The estimate \[ \|f\|_{L^{P}\left(L^{Q}\right)} \lesssim\|S f\|_{L^{P}\left(L^{Q}\right)} \] is true, for every \(L^{Q}\)-valued Schwartz function \(f\) on \(\mathbb{R}^{d}\), as long as the \(n\)-tuples \(Q\) and the \(d\)-tuples \(P\) satisfy the condition \(0<P\), \(Q<\infty\) componentwise.