On null quadrature domains (Q934525): Difference between revisions

An open set \(\omega\subset\mathbb R^n\) is a null quadrature domain if for all harmonic and integrable functions \(h\) on \(\Omega\) the integral \(\int_\Omega h(x)\,dx = 0\). Two papers of \textit{M. Sakai} from 1981 and 1993 [J. Anal. Math. 40, 144--154 (1981; Zbl 0483.30002) and SIAM J. Math. Anal. 24, No. 2, 341--364 (1993; Zbl 0771.30041)] give a complete characterisation of two-dimensional null quadrature domains: \(\Omega\) is a null quadrature domain if, and only if, \(\Omega^c\) is an ellipse or a strip, or \(\Omega\) is a half-plane or \(\Omega\) is the non-convex domain bounded by a parabola. \textit{A. S. Margulis} and the present author conjectured in [J. Anal. Math. 70, 1--63 (1996; Zbl 0876.35136)] that for general \(n\) a set \(\Omega\) is a null quadrature domain if, and only if (a) \(\Omega^c\) is an ellipsoid or (b) a slab, (c) \(\Omega\) is a half-space or (d) the non-convex domain bounded by an elliptic paraboloid, or \(\Omega\) is a cylinder with a base of type (a) or (d). The conjecture is true if \(\Omega^c\) is bounded. In the present paper the author shows that for any null quadrature domain \(\Omega\subset\mathbb R^n\) such that for some coordinate system the boundary \(\partial\Omega\subset\{x: | x_n| \leq R\}\) for some \(R>0\) and such that \[ \liminf_{r\to\infty} \frac{\text{Cap}_n(\Omega^c\cap B_r)}{\text{Cap}_n(B_r)}>0, \tag \(*\) \] (Cap stands for the Newtonian capacity) then \(\Omega\) is either a half-space or the complement of a slab. The proof uses an `unorthodox' Phragmen-Lindelöf-type theorem which is proved in an appendix: let \(u\) be subharmonic on an unbounded domain \(\Omega\) and continuous up to the boundary such that \(u| \partial\Omega\geq 0\). If \((*)\) holds, then either \(u\geq 0\) in \(\Omega\) or there is some \(\beta>0\) such that \[ \liminf_{r\to\infty} \frac{\sup_{\omega\cap\{| x| =r\}}| u(x)| }{r^\beta}>0. \]