$C_p$ estimates for rough homogeneous singular integrals and sparse forms

TL;DR

This paper extends Coifman--Fefferman inequalities for rough homogeneous singular integrals to broader weight classes, introduces new structural insights into $C_p$ weights, and employs sparse domination techniques for sharper bounds.

Contribution

It generalizes inequalities to $C_q$ weights without extrapolation, and refines the understanding of $C_p$ classes with explicit constructions and structural analysis.

Findings

Established bounds for $T_$ on $C_q$ weights without extrapolation.

Introduced the class tp; and

Provided a new proof that $C_{p+ps}$ condition is not necessary.

Abstract

We consider Coifman--Fefferman inequalities for rough homogeneous singular integrals $T_\Omega$ and $C_p$ weights. It was recently shown by Li-P\'erez-Rivera-R\'ios-Roncal that $$ \|T_\Omega \|_{L^p(w)} \le C_{p,T,w} \|Mf\|_{L^p(w)} $$ for every $0< p < \infty$ and every $w \in A_\infty$. Our first goal is to generalize this result for every $w \in C_q$ where $q > \max\{1,p\}$ without using extrapolation theory. Although the bounds we prove are new even in a qualitative sense, we also give the quantitative bound with respect to the $C_q$ characteristic. Our techniques rely on recent advances in sparse domination theory and we actually prove most of our estimates for sparse forms. Our second goal is to continue the structural analysis of $C_p$ classes. We consider some weak self-improving properties of $C_p$ weights and weak and dyadic $C_p$ classes. We also revisit and generalize a counterexample by Kahanp\"a\"a and Mejlbro who showed that $C_p \setminus \bigcup_{q > p} C_q \neq \emptyset$. We combine their construction with techniques of Lerner to define an explicit weight class $\widetilde{C}_p$ such that $\bigcup_{q > p} C_q \subsetneq \widetilde{C}_p \subsetneq C_p$ and every $w \in \widetilde{C}_p$ satisfies Muckenhoupt's conjecture. In particular, we give a different, self-contained proof for the fact that the $C_{p+\varepsilon}$ condition is not necessary for the Coifman--Fefferman inequality and our ideas allow us to consider also dimensions higher than $1$.