Failure of the $H^1\times H^1\longrightarrow L^{1/2,\infty}$ estimate for the lacunary bilinear spherical maximal function and applications

Abstract

In contrast to Christ's theorem that the lacunary spherical maximal operator is bounded from $H^1(\mathbb R^d)$ to $L^{1,\infty}(\mathbb R^d)$ ($d\geq2$), we show that the lacunary bilinear spherical maximal operator $$ \mathcal B_{\mathrm{lac}}(f,g)(x) =\sup_{\ell\in\mathbb Z} \Bigl|\int_{\mathbb S^{2d-1}}f(x-2^\ell y)\,g(x-2^\ell z)\,dσ(y,z)\Bigr| $$ is not bounded from $H^1(\mathbb R^d)\times H^1(\mathbb R^d)$ to $L^{1/2,\infty}(\mathbb R^d)$ for any $d\geq1$. In particular, it is not bounded from $L^1(\mathbb R^d)\times L^1(\mathbb R^d)$ to $L^{1/2,\infty}(\mathbb R^d)$. More precisely, for every sufficiently large integer $N$, we construct two finite linear combinations of smooth, compactly supported, mean-zero $H^1(\mathbb R^d)$-atoms for which the ratio of the weak $L^{1/2}$ quasinorm to the product of the two Hardy norms is at least $c_d\log(N+1)$. This lower bound holds already for the supremum over a suitable set of $N$ dyadic radii. For $d=1$ this answers the case $p=q=1$ of a question in Borges's survey. For every $k\ge3$ and $d\ge1$, a modification of the construction proves that the lacunary $k$-linear spherical maximal operator does not map $(H^1(\mathbb R^d))^k$ into $L^{1/k,\infty}(\mathbb R^d)$. We also indicate extensions to certain hypersurfaces and to horizontal spherical averages on the Heisenberg group.