Sharp weighted endpoint estimates for commutators of singular integrals (Q5954586)

In a previous paper [J. Funct. Anal. 128, No.~1, 163-185 (1995; Zbl 0831.42010)], \textit{C. Pérez} proved that if \(T\) is a Calderón-Zygmund singular integral operator and \(b\) is a BMO function, then there exists a constant \(C>0\) such that for all functions \(f\) and all \(\lambda >0\), \[ |\{ y\in \mathbb{R}^n: |[b,T]f(y)|>\lambda \} |\leq C\int_{\mathbb{R}^n}{{|f(y)|}\over {\lambda }}\left (1+\log^+\left ( {{|f(y)|}\over {\lambda }}\right)\right)dy, \] where \([b,T]\) stands for the commutator \([b,T]f=bT(f)-T(bf)\). This paper is devoted to improve the above result, obtaining a two-weight inequality for the higher order commutator \(T_b^m\) defined inductively by \(T_b^m= [b,T_b^{m-1}]\). The pairs of weights which are used are of the form \((w,M_{\Phi }w)\). By \(w\) we mean a nonnegative measurable function and \(M_{\Phi}\) is a maximal operator, associated to the Young function \(\Phi\), defined by \(M_{\Phi}f(x)=\sup_{x\in Q}\|f\|_{\Phi ,Q}\), where \(\|f\|_{\Phi ,Q}\) denotes the \(\Phi\)-average of \(f\) over \(Q\) given by means of the Luxemburg norm. More precisely, the authors show that if \(m\in \mathbb{N}\), \(w\) is a nonnegative function, \(\varepsilon >0\), \(T\) is a Calderón-Zygmund singular integral operator and \(b\in \text{BMO}\), then there exists \(C>0\) such that \[ w\left(\{ y\in \mathbb{R}^n: |T_b^mf(y)|>\lambda \} \right) \leq C\int_{\mathbb{R}^n}\Phi_m\left (\|b\|_{\text{BMO}}^m{{|f(y)|}\over {\lambda }}\right)M_{L(\log L)^{m+\varepsilon }}w(x)dx, \] where \(\Phi_m(t)=t(1+\log^+t)^m\). The proof works by induction on \(m\) and does not apply the technique of good-\(\lambda\) inequalities.