Endpoint Sobolev boundedness and continuity of multilinear fractional maximal functions (Q2166175)

The centered maximal function is denoted \(\mathcal{M}\) and the uncentered maximal function as \(\widetilde{\mathcal{M}}\); I will give the results for the uncentered maximal functions and let the interested reader look at the paper for results related to the centered maximal function. Those results may be slightly different. There is much that is understood about derivatives of the Hardy-Littlewood maximal function, for example \(\widetilde{\mathcal{M}}:W^{1,p}(\mathbb{R}^n) \mapsto W^{1,p}(\mathbb{R}^n), 1< p \leq \infty\). Because of non-reflexivity, the situation at \(p = 1\) is more delicate, with the basic question being whether \(f \mapsto \nabla\widetilde{ \mathcal{M}}f\) maps \(W^{1,1}(\mathbb{R}^n)\) to \(L^1(\mathbb{R}^n)\). \textit{J. M. Aldaz} and \textit{J. Pérez Lázaro} [Trans. Am. Math. Soc. 359, No. 5, 2443--2461 (2007; Zbl 1143.42021)] proved that if \(f\) is of bounded variation, \(\widetilde{\mathcal{M}}f\) is absolutely continuous and \[ \mbox{Var} ( \widetilde{\mathcal{M}f}) \leq \mbox{Var}(f), \] and thus that \[ || (\widetilde{\mathcal{M}}f)^{\prime} ||_{L^1(\mathbb{R})} \leq || f||_{L^1(\mathbb{R})} , f \in W^{1,1}(\mathbb{R}), \] and the constant is sharp. \textit{E. Carneiro} and \textit{J. Madrid} [Trans. Am. Math. Soc. 369, No. 6, 4063--4092 (2017; Zbl 1370.26022)] proved one-dimensional results for the fractional maximal operator \[ \widetilde{\mathcal{M}}_{\alpha} f(x) = \sup_{r, s \geq 0, r + s>0} \frac{1}{(r +s)^{1 - \alpha}} \int_{x - r}^{x +s} |f(y)| \, dy \] (Note that \( \widetilde{\mathcal{M}}_{0} = \widetilde{\mathcal{M}}\)). The ultimate result was in [\textit{J. Madrid}, Rev. Mat. Iberoam. 35, No. 7, 2151--2168 (2019; Zbl 1429.42021)] who proved that if \(0 < \alpha <1, q = \frac{1}{1 - \alpha}, f \mapsto (\widetilde{\mathcal{M}}_{\alpha} f)^{\prime} \) is continuous from \(W^{1,1}(\mathbb{R})\) to \(L^q(\mathbb{R})\). One of the authors' results is to extend this to the multilinear case. If \(\vec{f} = (f_1, \ldots, f_m)\), where each \(f_j \in L^1_{\mbox{loc}}(\mathbb{R})\), the uncentered fractional maximal operator is \[ \widetilde{\mathfrak{M}}_{\alpha} \vec{f}(x) = \sup_{r, s \geq 0, r + s>0} \frac{1}{(r +s)^{m - \alpha}} \prod_{j = 1}^m \int_{x - r}^{x +s} |f_j(y)| \, dy, \] and it is known that it sends \(W^{1,p_1}(\mathbb{R}) \times \cdots \times W^{1,p_m}(\mathbb{R})\) to \(W^{1,q}(\mathbb{R})\) provided that \(1 < p_1,\dots, p_m, 0 < \alpha <m, 1/q = 1/p_1 + \cdots + 1/p_m - \alpha\). The question is about the behavior of the endpoint case \(p_1 = p_2 = \cdots = p_m = 1\). Question A is whether the mapping \(\vec{f} \mapsto (\widetilde{\mathcal{\mathcal{M}}}_{\alpha} f)^{\prime} \) bounded and continuous from \(W^{1,1}(\mathbb{R}) \times \cdots \times W^{1,1}(\mathbb{R})\) to \(L^q(\mathbb{R})\) if \(0 < \alpha <m , q = \frac{1}{m - \alpha}\). Results of \textit{F. Liu} and \textit{H. Wu} [Can. Math. Bull. 60, No. 3, 586--603 (2017; Zbl 1372.42015)] and others can be extended to handle \(m \geq 2, 1 \leq \alpha < m\) and in this paper the authors prove the remaining case; if \(0 < \alpha <1\), \(q = \frac{1}{1 - \alpha}\), then \(\vec{f} \mapsto (\widetilde{\mathfrak{M}}_{\alpha} f)^{\prime} \) is bounded and continuous from \(W^{1,1}(\mathbb{R}) \times \cdots \times W^{1,1}(\mathbb{R})\) to \(L^q(\mathbb{R})\).