Extensions of the vector-valued Hausdorff-Young inequalities (Q2231138)

The authors study several vector-valued analogues of classical inequalities for the Fourier transform. On the one hand they analyse \textit{H.~R. Pitt}'s inequality [Duke Math. J. 3, 747--755 (1937; Zbl 0018.01703)] which establishes that \[\|(\hat f(n))\|_{\ell^q(\mathbb Z^d), (|n|+1)^{-\gamma q})}\le C \|f\|_{L^p(\mathbb T^d, |\cdot|^{\beta p})}, \quad 1<p\le q<\infty\] holds if and only if \(\max\{0, d(\frac{1}{p}+\frac{1}{q}-1)\}\le \gamma<\frac{d}{q}\) and \(\beta-\gamma=d(1- \frac{1}{p}-\frac{1}{q})\) or the similar formulation for \(\mathbb R^d\). Some particular cases that are included in these inequalities, for instance \(\gamma=\beta=0\) and \(q=p'\ge 2\), have already been considered in the vector-valued setting. A Banach space is said to be of Fourier type \(p\in [1,2]\) when the Hausdorff-Young inequality holds for functions with values in \(X\). Many results can be extended when \(X\) is assumed to have nontrivial Fourier type, which turns out to be equivalent to Rademacher type \(>1\), or \(B\)-convexity due to results by \textit{J.~Bourgain} [Lect. Notes Math. 1317, 239--249 (1988; Zbl 0653.46016)] and \textit{G.~Pisier} [Ann. Math. (2) 115, 375--392 (1982; Zbl 0487.46008)]. In the present paper, the authors show the following extension of Pitt's inequality to the vector-valued setting. If \(X\) is a Banach space with Fourier-type \(p_0\in (1,2]\), \(1<p\le q<\infty\), \(\beta,\gamma\ge 0\) and \(\max\{0, d(\frac{1}{\min\{p,p_0 \}}+\frac{1}{q}-1)\}< \gamma<\frac{d}{q}\) and \(\beta-\gamma=d(1- \frac{1}{p}-\frac{1}{q})\), then \[\|\hat f\|_{L^q(\mathbb R^d), |\cdot|^{-\gamma q}, X)}\le C \|f\|_{L^p(\mathbb R^d, |\cdot|^{\beta p}), X}.\] Also, it is shown that the above vector-valued inequality holds in the case \[ \gamma=\max\Bigl\{0, d\Bigl (\frac{1}{\min\{p,p_0 \}}+\frac{1}{q}-1 \Bigr)\Bigr\} \] for different values of the parameters \(p\) and \(q\). Similar formulations for functions defined in \(\mathbb T^d\) are also shown using transference. This theorem allows to improve some previously known results by \textit{J. Parcet} et al. [Indiana Univ. Math. J. 62, No.~6, 1765--1784 (2013; Zbl 1298.42006)] which dealt with vector-valued Pitt's inequalities for \(p=q\in(1,2]\). On the other hand, they also deal with Zygmund's inequality which establishes that \[\sum_{n=1}^\infty \frac{(1+ \log n)^b}{n} \hat f^*(n)\le C \int_{\mathbb T} |f(t)|(\log^+|f(t)|)^{b+1}\,dt, \quad b+1>0.\] Their result in this direction shows that the validity of such an inequality in the vector-valued setting also characterizes the nontrivial Fourier type of the space. They actually provide a more general result which works in the setting of vector-valued Lorentz-Zygmund spaces as well as their discrete counterparts. The authors also give the vector-valued analogue of \textit{S. V. Bochkarev}'s inequality [in: Approximation theory. Harmonic analysis. Collected papers dedicated to the memory of Professor Sergei Borisovich Stechkin. Moscow: MAIK Nauka/Interperiodica Publishing (1997; Zbl 0922.46024); translation from Tr. Mat. Inst. Steklova 219, 96--107 (1997)].