A solution operator for \(\bar\partial\) on the Hartogs triangle and \(L^p\) estimates (Q2290807)

Gennadi Henkin constructed an explicit integral operator $H$ on the polydisk that solves the Cauchy-Riemann equations $u = \bar{\partial}f$ for a $\bar{\partial}$-closed $(0,q)$-form $f$ continuous on the closed polydisk. $Hf$ satisfies $C^0$-estimates. In her thesis from 1986 [Randregularität von Lösungen der \({\bar \partial}\)-Gleichung auf dem Polyzylinder und zweidimensionalen analytischen Polyedern. Bonn. Math. Schr. 176, 164 p. (1986; Zbl 0622.32001); Das \({\bar\partial}\)-Problem auf pseudokonvexen Polyedern nach Sergeev und Khenkin. Bonn. Math. Schr. 167, 166 p. (1985; Zbl 0597.32017)], \textit{J. Bertrams} modified $H$ to obtain operators $H^*$ that satisfy $C^k$-estimates for $k\geq 0$. She showed moreover that $H$ itself satisfies $C^k$-estimates in the case of $(0,1)$-forms; in this case (for $q=1$) the canonical (with respect to the euclidean metric) solution $Kf$ satisfies these estimates even if the form $f$ is only defined on the open polydisk. This answers a question the authors pose in their introduction. Lan Ma and Joachim Michel then used the above results to study the Hartogs triangle in dimension 2; the higher dimensional case is treated in \textit{S. Bach}'s diploma thesis [Lösungsoperatoren für den $\bar{\partial}$-Komplex. University of Bonn (Diplomarbeit) (1997)]. The case of $L^p$-estimates, $1\leq p<\infty$, had been left open -- with good reason, as the authors show in their present paper where they attack this very question. They concentrate on the case of dimension 2 and consider products of (not necessarily smoothly) bounded plane domains; the essential example is the bidisk with Henkin's solution operator $H$. Given a $\bar{\partial}$-closed $(0,1)$-form $f$ they transform $Hf$ into a sum of two Cauchy integrals applied to $f$ and one iterated Cauchy integral applied to a derivative $Df$ of $f$. The $L^p$-estimates for $Hf$ are then straightforward but they invoke $L^p$-boundedness of both $f$ and $Df$. The most interesting result shows that assumptions like the above are unavoidable: $L^p$-boundedness of $f$ alone does not imply $L^p$-boundedness of $Hf$. From here the authors follow the usual procedure of passing from the punctured bidisk to the Hartogs triangle to establish the corresponding estimates there. The authors have taken great care to describe and appreciate previous contributions by many authors; only J. Bertrams' work escaped their attention.