Singular integral operators with kernels supported in higher dimensional subvarieties (Q2157345): Difference between revisions

Classical Calderón Zygmund operators are given by \[ \mathcal{S}_{\Omega} f(x) = p. v. \int f(x - y) \frac{\Omega(y^{\prime})}{|y|^n} \, dy \] with \(y^{\prime} = y |y|^{-1}, y \ne 0\), where \(\Omega\) is homogeneous of degree 0, \(\Omega( \lambda y^{\prime}) = \Omega(y^{\prime})\). \textit{A. P. Calderón} and \textit{A. Zygmund} [Am. J. Math. 78, 289--309 (1956; Zbl 0072.11501], using their method of rotations, proved that \(S_{\Omega}\) is bounded on \(L^p\) if \(1 < p < \infty\), \(|\Omega| \in L \log L(S^{n -1})\), and that this cannot be weakened to \(|\Omega| \in L (\log L)^{1 - \epsilon}(S^{n -1})\) for any \(\epsilon > 0\). The author considers Radon transforms on \(\mathbb{R}^n\) with kernels supported on subvarieties in \(\mathbb{R}^n \times \mathbb{R}^n\) defined by a polynomial \[ \mathcal{P}(x,y) = (P_1(x,y), \ldots, P_n(x,y)): \mathbb{R}^n \times \mathbb{R}^n \mapsto \mathbb{R}^n. \] \[ \mathcal{T}_{\mathcal{P}, \Omega} f(x) = p. v. \int_{\mathbb{R}^n \times \mathbb{R}^n} f(x - \mathcal{P}(u,v) )\frac{ \Omega(u^{\prime}, v^{\prime}) }{ |u|^n |v|^n } \, du dv \] where \(u^{\prime}, v^{\prime}\) are as above, \[ \Omega \in L^1(S^{n -1} \times S^{n-1}),\ \Omega(tx, sy) = \Omega(x, y) \mbox{ for any } t, s >0; \] \[ \int_{S^{n -1}} \Omega(u^{\prime}, \cdot) d \sigma_n (u^{\prime}) = \int_{S^{n -1}} \Omega(\cdot, v^{\prime}) d\sigma_n(v^{\prime}) = 0. \] This formulation contains the composition of two homogeneous singular integral operators. If \(\mathcal{P}(u,v) = P_1(u) + P_2(v)\), \(\Omega(x,y) = \Omega_1(x) \Omega_2(y)\), where \(\Omega_1\), \(\Omega_2\) are homogeneous of degree zero and integrable on \(S^{n - 1}\), \[ \mathcal{T}_{\mathcal{P}, \Omega} f(x) = p. v. \int_{\mathbb{R}^n} S_{\Omega_1, \mathcal{P}_1} f(x - \mathcal{P}_2(y) )\frac{ \Omega_2(v^{\prime}) }{ |v|^n } \, dv, \] where \[ \mathcal{S}_{\Omega_1, \mathcal{P}_1} f(x) = p. v. \int_{\mathbb{R}^n} g(x - \mathcal{P}_1(u) )\frac{ \Omega_1(u^{\prime}) }{ |u|^n } \, du. \] If \(\Omega_1\), \(\Omega_2\) are in \(L \log L(S^{n - 1})\) it follows from [\textit{D. Fan} and \textit{Y. Pan}, Am. J. Math. 119, No. 4, 799--839 (1997; Zbl 0899.42002)] (or from Calderón and Zygmund if \(\mathcal{P}(u,v) = u + v\)), that \(\mathcal{T}\) maps \(L^p\) into \(L^p\). The author shows that this implies \(\Omega_1 \Omega_2 \in (L \log L)(S^{n - 1} \times S^{n - 1})\). His main result is that if \(\Omega \in L (\log L)^2(S^{n -1})\) and some auxiliary condition on \(\mathcal{P}\), \(\mathcal{T}_{\mathcal{P}, \Omega}\) maps \(L^p\) to \(L^p\), and the exponent cannot be replaced by \( 2 - \epsilon\), even for \(L^2\). The author proves the same result for \(\mathcal{P}(x,y) = \mathcal{R}_1(x) + \mathcal{R}_2(y)\) generalizing the previously cited result by no longer assuming that \(\Omega\) is a product. The auxiliary condition of \(\mathcal{P}\) is slightly misstated, but it is difficult to see from the proof what the correct version should be. He requires \(\mathcal{P}(x, 0) = \mathcal{P}(0,y) = (0,0)\), but since the image of \(\mathcal{P}\) lies in \(\mathbb{R}^n\), that should at least be \((0, \ldots, 0)\). In the second theorem. there is no mention of what this requires of \(\mathcal{R}_1, \mathcal{R}_2\), or why there need be no such condition in that case. One interesting consequence is that the operator \[ C_{\mathcal{P}, \Omega} f(x) = p.v. \int_{\mathbb{R}^n \times \mathbb{R}^n} e^{i\mathcal{P}(u,v)} f(x - u) \frac{ \Omega(u^{\prime}, v^{\prime}) }{ |u|^n |v|^n } \, du dv, \] is bounded on \(L^p\) for \(1 < p < \infty\) where the constants depend on the degree of the polynomial \(\mathcal{P}\), but not on its coefficients, with the same conditions on \(\Omega\). The proof involves an extension of the Ricci-Stein theorem [\textit{F. Ricci} and \textit{E. M. Stein}, J. Funct. Anal. 73, 179--194 (1987; Zbl 0622.42010)] to polynomials on \(\mathbb{R}^n \times \mathbb{R}^n\), and delicate estimates of oscillatory integrals, and some associated maximal functions.