Pointwise selection theorems for metric space valued bivariate functions (Q2396673): Difference between revisions

Pointwise selection principles are existence theorems guaranteeing the existence of a pointwise convergent subsequence of a given sequence of functions, like the classical Helly's theorem. In the paper under review, the authors prove two theorems of such a type for functions of two variables valued in metric spaces. Analogous results can be proved for functions of several variables. Theorem 1. Let \(M\) be an arbitrary metric space and let \(I=[a_1,b_1]\times [a_2,b_2]\) be a rectangle in \(\mathbb{R}^2\). For functions \(f,g\in M^I\) let \(\mathrm{TV}(f,g,I)\) denote the total joint variation (of Vitali-Hardy-Krause-type) of \(f\) and \(g\). Let \(\{ f_j\}, \{ g_j\}\subset M^I\) be two sequences of functions such that: (a) \(\{ f_j\}\) is pointwise precompact on \(I\); (b) \(\{ g_j\}\) is pointwise convergent on \(I\) to a function \(g\in M^I\); (c) \(C=\limsup_j\mathrm{TV}(f_j,g_j,I)<\infty\). Then, there is a subsequence of \(\{ f_j\}\), which convergens pointwise on \(I\) to a function \(f\in M^I\) with \(\mathrm{TV}(f,g,I)\leq C\). Theorem 2. Let \(M\) be a finite-dimensional normed space with the norm \(\|\cdot \|\), \(I=[a_1,b_1]\times [a_2,b_2]\). For a function \(f\in M^I\) let \(\mathrm{TV}_{\varepsilon}(f,I)\) denote the total \(\varepsilon\)-variation of \(f\). Let \(\{ f_j\}\subset M^I\) be a sequence of functions such that: (a) \(A=\sup_j \|f_j(\langle a_1,a_2\rangle)\|<\infty\); (b) \(\nu_{\varepsilon}=\limsup_j\mathrm{TV}_{\varepsilon}(f_j,I)<\infty\) for all \(\varepsilon>0\). Then, there is a subsequence of \(\{ f_j\}\) which converges in \(M\) pointwise on \(I\) to a bounded function \(f\in M^I\) with \(\mathrm{TV}_{\varepsilon}(f,I)\leq\nu_{\varepsilon}\) for all \(\varepsilon>0\).