\(L^p\) estimates for non-smooth bilinear Littlewood-Paley square functions on \(\mathbb R\) (Q637588): Difference between revisions

For an interval \(I=[a,b]\), let \(\pi_{I}(f,g)(x)=\int_{\mathbb R^2}e^{ix(\xi_1+\xi_2)}\hat f(\xi_1)\hat g(\xi_2) \chi_{I}(\xi_1-\xi_2)\,d\xi_1 d\xi_2 =\int_{\mathbb R}f(x+y)g(x-y)\widehat{\chi_I}(y)\,dy\). The author's main result is: Let \(2<p,q, r'<\infty\) with \(1/r=1/p+1/q\). Assume that the intervals \(I_n\) have the same length and \(\sum_{n\in \mathbb Z}\chi_{I_n}\) is bounded. Then, there is a constant \(C=C(p,q,r)\) such that, for all \(f,g\in\mathcal S(\mathbb R)\), \(\|(\sum_{n\in\mathbb Z}|\pi_{I_n}(f,g)|^2)^{1/2}\|_{r} \leq C\|f\|_p\|g\|_q\). In particular, this holds for \(I_n=[n,\,n+1]\), and it is the bilinear version of Carleson's result on the Littlewood-Paley square function \(\|(\sum_{n\in\mathbb Z}|\mathcal F^{-1}(\chi_{[n,n+1]}\hat f)|^2)^{1/2}\|_{p}\leq C_p\|f\|_p\) for \(2\leq p<\infty\). The proof is a mixture of the well-known time-frequency analysis used for the study of operators closely related to bilinear Hilbert transforms and of some vector-valued arguments. The restriction \(p,q\geq2\) is as necessary as the restriction \(p\geq2\) in Carleson's result. A counter-example is given.