A Minkowski problem for electrostatic capacity (Q1372996): Difference between revisions

If somebody, whoever, tries to classify the great deal of published mathematical papers, then the article under review should be definitely nominated as highest-level mathematics. Roughly speaking, the author applies harmonic analysis techniques in the study of the Minkowski problem. As known, this problem consists of finding a convex polyhedron from the given normals to the faces and their surface areas. The corresponding problem for convex bodies with smooth boundaries is to find the body if the Gauss curvature of the boundary is given as a function of the unit normal. In these settings, Minkowski introduced the notion of variation of a convex domain as the variation of the boundary in the direction normal to it. Indeed, the first variation of the volume of a convex body is the surface area measure on the boundary. The general goal of Jerison's paper is to develop a theory similar to that for the Minkowski problem in which the volume is replaced by the electrostatic capacity and the surface area is replaced by the first variation of the capacity. To be more precise, let \(\Omega\) be a bounded, convex and open subset of \(\mathbb{R}^N\), \(N\geq 3\). Denote by \(\Omega'\) the complement of the closure \(\overline\Omega\). The equilibrium potential of \(\Omega\) is the continuous function \(U\) defined on \(\overline\Omega'\) which satisfies \[ \Delta U=0\quad\text{ in} \Omega',\qquad U=1\quad\text{ on} \partial\Omega',\qquad \lim_{|x|\to\infty} U(x)=0. \] It is well known that \(U\) has the asymptotic expansion \[ U(x)=\gamma{1\over {(N-2)\text{ vol }(S^{n})}}|x|^{2-N}+O\left(|x|^{1-N}\right)\qquad\text{ as }x\to\infty, \] where \(S^n\) is the unit sphere in \(\mathbb{R}^N\), \(n=N-1\), and the constant \(\gamma\) is known as the electrostatic capacity of \(\Omega\), \(\gamma=\text{ cap }\Omega\). Let \(g\colon\;\partial\Omega\to S^n\) be the Gauss map, i.e., the mapping from \(x\in\partial\Omega\) to the unit outward normal at \(x\). Set \(u(\xi)\) for the Minkowski support function of \(\Omega:u(\xi)=x\cdot\xi\) for any point \(x\in P\cap\partial\Omega\), where \(P\) is a support plane of \(\Omega\) and \(\xi\) is the unit normal to \(P\) pointing from \(\Omega\) into \(\Omega'\). Clearly, the support function can be considered as a function on \(\partial \Omega:\) \(u(x)=u(g(x))\). The starting point for the Jerison paper is the Poincaré formula: \[ \text{cap }\Omega={1\over{N-2}}\int_{\partial\Omega} u|\nabla U|^{2} d\sigma,\tag{1} \] which is easily verified in the case of smooth \(\Omega\), and which implies \[ \text{vol }\Omega={1\over{N}}\int_{\partial \Omega} u d\sigma. \] Now, let \(\Omega_{1}\) be another bounded and convex domain with Minkowski support function \(u_{1}\). It is known that \[ {d\over {dt}} \text{ vol }(\Omega+t\Omega_{1})\Big|_{t=0} =\int_{\partial \Omega} u_{1}d\sigma, \tag{2} \] \((\Omega+t\Omega_{1}=\{x+ty\colon\;x\in \Omega\) and \(y\in\Omega_{1}\})\), and this leads to the measure \(d\mu=g_{*}(d\sigma)\) on \(S^{n}\) defined by \(\mu(E)=\int_{g^{-1}(E)} d\sigma\) for any Borel set \(E\subset S^{n}\). This measure plays a crucial role in the classical theory of the Minkowski problem. To realize his idea, the author uses an analog of \((2)\) known as Hadamard's variational formula: \[ {d\over {dt}} \text{ cap }(\Omega+t\Omega_{1})\Big|_{t=0} =\int_{\partial \Omega} u_{1}|\nabla U|^{2}d\sigma, \] in order to define the measure \(d\mu=g_{*}(|\nabla U|^{2}d\sigma)\) on \(S^{n}\) by \[ \mu(E)=\int_{g^{-1}(E)} |\nabla U|^{2} d\sigma\quad \text{ for\;any\;Borel\;set} E\subset S^{n}. \tag{3} \] The main result of the paper is stated as follows: Theorem: Let \(N\geq 4\) and let \(\mu\) be a positive measure on \(S^{n}\). There exists a bounded convex body \(\Omega\) such that \(g_{*}(|\nabla U|^{2}d\sigma)=d\mu\) if and only if \(\mu\) satisfies \[ \int_{S^{n}} \xi_{k} d\mu(\xi_{k})=0,\qquad k=1,\ldots,N, \] and \(\mu\) is not supported on any equator (the intersection of \(S^{n}\) with any hyperplane through the origin). Moreover, the body \(\Omega\) is unique up to translation. Further, suppose that \(k\geq0\) is an integer and \(0<\alpha<1\). If \(d\mu=Rd\xi\) for some strictly positive density \(R\in C^{k,\alpha}(S^{n})\), then \(\Omega\) is \(C^{k+2,\alpha}\)-smooth. Section 1 of the paper is devoted to the properties of the equilibrium potential \(U\). By the way, the Poincaré formula \((1)\) is deduced in the non-smooth case. In \S 2 the author proves formulas for the first and second variation of capacity in the case of smooth \(\Omega\). Continuity properties for the first variation are derived in Section 3. This property is used later in \S 5 to prove existence, and also in extending the validity of the first variation formula to non-smooth domains. Section 4 of the paper deals with a priori estimates on the inradius and diameter of \(\Omega\) which serve in the existence part of the Theorem. Existence of polyhedra is proved in \S 5 by formulating a variational problem Existence of general convex bodies is then deduced by a weak limiting procedure which essentially depends on the a priori estimates already established in \S 4 and on the fact that \(|\nabla U|\) appears in \((3)\) with a power \(2.\) The regularity part of the Theorem is proved in \S 6 by deriving new estimates for the density \(|\nabla U|^{2}\) on the boundary of a convex domain. In contrast to the interior estimates [see \textit{D. Jerison}, Invent. Math. 105, No. 2, 375-400 (1991; Zbl 0754.31007)], however, the doubling condition fails here, and the author provides a weaker estimate. In \S 7 a stronger version of Caffarelli's regularity theory of the Monge-Ampère equation is developed which makes use of this weaker estimate. Finally, in \S 8 the author gives elegant examples showing that the estimates of the paper are optimal and also explaining the reason why the exponent \(2\) of \(|\nabla U|^{2}\) in \((3)\) is the only one possible. The uniqueness part of the Theorem follows in essentially the same way as in the classical Minkowski problem, and its proof makes use of the Brunn-Minkowski isoperimetric inequality [see \textit{C. Borell}, Math. Ann., II. Ser. 263, 179-184 (1983; Zbl 0546.31001)] and of a recent result by \textit{L. A. Caffarelli, D. Jerison} and \textit{E. H. Lieb} [Adv. Math. 117, No. 2, 193-207 (1996; Zbl 0847.31005)].