On oscillatory singular integrals and their commutators with non-convolutional Hölder class kernels (Q2034999)

One of the most fundamental results in harmonic analysis is the Calderòn-Zygmund extrapolation theorem: an \(L^{2}\) bounded singular integral operator of the form \[ Tf(x) = \int _{\mathbb{R}^{d}} K(x,y)f(y)dy \] extends to a bounded operator on \(L^{p}\) for \(p\in(1,\infty)\) when its kernel satisfies standard size (\(|K(x,y)| \lesssim |x-y|^{-d}\)) and (Hölder) regularity assumptions. In [\textit{F. Ricci} and \textit{E. M. Stein}, J. Funct. Anal. 73, 179--194 (1987; Zbl 0622.42010)] this result is extended to oscillatory singular integral operators of the form \[ Tf(x) = \int _{\mathbb{R}^{d}} \exp(iP(x,y))K(x,y)f(y)dy, \] for a real valued polynomial \(P\), under a \(C^{1}\) regularity assumption on \(K\). This is motivated by a range of applications, to analysis on Lie groups in particular (e.g. twisted convolutions). Somewhat suprisingly, this result has not yet been fully extended to the case where \(K\) only satisfies the standard Hölder regularity assumption. A result in this case is proven in [\textit{H. Al-Qassem} et al., J. Fourier Anal. Appl. 25, No. 4, 2141--2149 (2019; Zbl 1417.42012)], but only for polynomials of the form \(P(x,y) = \sum \limits_{|\alpha|=|\beta|=1} a_{\alpha,\beta}x^{\alpha}y^{\beta}\). The paper under review extends this result to polynomials of degree at most one in either the \(x\) or the \(y\) variable. For such operators, the authors also prove natural related results: weighted boundedness (for \(A_{p}\) weights), boundedness of commutators with BMO functions, and compactness of commutators with CMO functions. They also show (Theorem 1.7) that the compactness of commutators can be proven for all polynomials \(P\) under \(C^{1}\) regularity assumptions on \(K\). This is quite an interesting phenomenon which begs the question: does the ``full result'' (\(C^{1}\) regularity assumptions on \(K\) and no restriction on \(P\)) hold? If the anwer were to be no, then understanding the nature of the interactions between \(K\) and \(P\) would be highly interesting. If the answer were to be yes, it would be a very natural extension of Ricci-Stein's result. The proof of the main result is quite interesting. The operator is decomposed into its singular part (\(x\) close to \(y\)), and its off-diagonal part. For the singular part, the authors run an induction argument on the degree of the polynomial in the variable where it is above \(1\). As the degree is decreased by \(1\), one picks up a factor \(|x-y|\) which decreases the level of singularity near the diagonal. This is a fairly simple argument, but it certainly uses the structure of \(P\). For the off-diagonal part, one uses a classic dyadic decomposition and exploits the regularity of \(K\) to get off-diagonal decay. This seems less dependent on the structure of \(P\). It would be interesting to see if the off-diagonal part estimate does, indeed, hold for any polynomial \(P\). To prove the compactness of commutators theorems, the authors smoothly truncate the kernels (see 5.2) and apply a relevant (Frechet-Kolmogorov) characterisation of compact sets on (weighted) \(L^p\) spaces. Their main theorem (resp. Ricci-Stein's theorem) allows them to prove equiboundedness, leaving only the equicontinuity to be checked via subtle looking direct estimations.