Convexity properties of harmonic measures (Q927682)

Let \(U\) be an open set in \(\mathbb{R}^d,d\geq 2\) and let \(E\) be a Borel set on the boundary of \(U\). Let \(u\) be the harmonic function in \(U\) with \(u=\chi_E\) on \(\partial U\). For \(x\in U\), we set \(\mu_x^U(E)=u(x)\). As a function of \(E\), \(\mu_x^U\) is a Borel measure on \(\partial U\) called the harmonic measure of \(E\). The authors prove that any convex combination of the harmonic measures \(\mu_x^{U_1}\), \(\mu_x^{U_2}, \dots ,\mu_x^{U_k}\), where \(U_1,U_2,\dots,U_k\) are relatively compact open neighborhoods of a given point \(x\in R^d\), can be approximated by a sequence \((\mu_x^{W_n})_{n=1}^\infty\) of harmonic measures such that each \(W_n\) is an open neighborhood of \(x\) in \(U_1\cup U_2\cup\dots\cup U_k\). This answers a question raised in connection with Jensen measures by \textit{B. J. Cole} and \textit{T. J. Ransford} [J. Reine Angew. Math. 541, 29--53 (2001; Zbl 0988.31002)]. Moreover it implies that the measures obtained by balayage of the Dirac measure at \(x\) on Borel subsets are dense in the set of all representing measures \(\mu\) for \(x\), that is, of all Radon measures \(\mu\) such that \(\mu(p)\leq p(x)\) for every continuous real potential. The above results are proved in the more general context of Riesz potential theory and in fact hold in a very general potential-theoretic setting.