Approximating the condenser equilibrium distribution

From MaRDI portal





Let \(E\), \(F\) be disjoint non-polar compact sets in \(\mathbb{R}^d\) \((d\geq 2)\), and let \(M(E,F)\) denote the collection of signed Borel measures of the form \(\sigma_E-\sigma_F\), where \(\sigma_E\) (resp. \(\sigma_F\)) is a unit measure on \(E\) (resp. on \(F\)). The equilibrium distribution of the condenser \((E,F)\) is the (unique) member \(\tau\) of \(M(E,F)\) which minimizes the energy integral \(I(\tau)\). Given \(n\geq 2\) and a discrete measure \(\sigma_n\in M(E,F)\) associating masses \(1/n\) with points \(x_1,\dots, x_n\in E\) and masses \(-1/n\) with \(y_1,\dots, y_n\in F\), there is a notion of discrete energy \(I^*(\sigma_n)\) of \(\sigma_n\) (the diagonal is excluded from the region of double integration in the energy integral). There exist (not necessarily unique) points \(x^{(n)}_1,\dots, x^{(n)}_n\in E\) and \(y^{(n)}_1,\dots, y^{(n)}_n\in F\) for which the corresponding measure \(\sigma^{(n)}\) minimizes this discrete energy. It is known that \((\sigma^{(n)})\) converges in the weak\(^*\) sense to the equilibrium distribution \(\tau\). This paper quantifies this convergence by obtaining error bounds, for the difference \(U^\tau- U^{\sigma^{(n)}}\) of the corresponding (Newtonian or logarithmic) potentials, in terms of \(n\) and the distances of \(x\) from \(\partial E\) and from \(\partial F\). It also examines the discrepancy \(\sup_B|\tau(B)- \sigma^{(n)}(B)|\), where the supremum is over a certain class of subsets \(B\) of \(\partial E\cup\partial F\).











This page was built for publication: Approximating the condenser equilibrium distribution

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q5947768)