Commutators in the two-weight setting (Q514340)

The authors prove the following theorem: let \(\mu,\lambda\) be two weights in the Muckenhoupt \(A_p\) class and let \(T\) be a Calderón--Zygmund operator acting on functions defined on \(\mathbb{R}^d\). For a function \(b\), define \([b,T]f:=T(bf)-b(Tf)\); there holds: \[ ||{[b,T]:L^p(\mu)\to L^p(\lambda)}|| \leq C||b||_{\mathrm{BMO}(\nu)}. \] Here \(\nu=\mu^{\frac{1}{p}}\lambda^{-\frac{1}{p}}\) and \(\mathrm{BMO}(\nu)\) is a \({\mathrm{BMO}}\) space adapted to \(\nu\) defined by: \[ ||b||_{\mathrm{BMO}(\nu)} =\sup_{Q}\frac{1}{\nu(Q)} \int_{Q}\left|b(x)-\frac{1}{|Q|}\int_{Q}b(y)dy\right|dx. \] The constant \(C\) depends on \(d\), \(\nu\), and \(T\). The authors also prove that when \(T\) is a Riesz transform then we have ``\(\simeq\)'' above: \[ ||[b,R]:L^p(\mu)\to L^p(\lambda)|| \simeq ||{b}||_{\mathrm{BMO}(\nu)}. \] This result extends several important earlier results to the much more subtle two--weight setting. In particular, in 1976 Coifman, Rochberg, and Weiss prove this theorem in the setting \(\mu=\lambda=1\) and in 1985 Bloom proves this theorem in the one dimensional setting (that is, \(T\) is the Hilbert transform). This also extends earlier results of the same authors. The authors make use of the Hytönen Representation Theorem which says that every Calderón-Zygmund operator can be written as a rapidly convergent series of simpler dyadic model operators called Haar shifts. Using this, the authors write the commutator as a sum of compositions of Haar shifts with paraproducts plus a very singular term. The authors use a weighted \(H^1-{\mathrm{BMO}}\) duality to prove that the the paraproducts satisfy two weight bounds under the assumptions on \(\mu\), \(\lambda\) and \(b\). Since Haar shifts satisfy a one-weight bound, this takes care of the compositions of shifts with paraproducts. The singular term requires a more delicate analysis and it is through this term that the cancellation of the commutator expresses itself. To prove the main results, the authors must also give several equivalent characterizations of \(\mathrm{BMO}(\nu)\) and prove several other facts about this space which are interesting by themselves. As a corollary, the authors prove a weak-factorization theorem, extending another result from Coifman, Rochberg and Weiss. The methods developed by the authors have already been used by several other researchers to prove similar results in a variety of settings.