Stability of a class of action functionals depending on convex functions

From MaRDI portal



Abstract: We study the stability of a class of action functionals induced by gradients of convex functions with respect to Mosco convergence, under mild assumptions on the underlying space.


The authors consider a fixed metric space \((X,d)\), a function \( f:X\rightarrow \mathbb{R}\cup \{+\infty \}\), and the functional defined on \( C([0,1],X)\) as: \(\Theta _{x_{0},x_{1}}^{f}(\gamma )=\int_{0}^{1}\left\vert \overset{.}{\gamma }\right\vert ^{2}(t)+\left\vert \partial f\right\vert (\gamma (t))dt\), if \(\gamma \in AC([0,1],X)\), \(\gamma (0)=x_{0}\), \(\gamma (1)=x_{1}\), and \(\Theta _{x_{0},x_{1}}^{f}(\gamma )=\infty \), otherwise, where \(x_{0}\) and \(x_{1}\) are given points in \(X\), \(\left\vert \partial f\right\vert \) is the descending slope of \(f\) defined as: \(\left\vert \partial f\right\vert (x)=+\infty \), if \(x\in X\setminus D(f)\), \(\left\vert \partial f\right\vert (x)=0\), if \(x\in D(f)\) is isolated and \(\left\vert \partial f\right\vert (x)=\limsup_{y\rightarrow x}\frac{(f(y)-f(x))^{-}}{ d(y,x)}\), otherwise, \(D(f)\) being the effective domain of \(f\): \(D(f)=\{x\in X:f(x)<+\infty \}\). The purpose of the paper is to describe the \(\Gamma \)-convergence of \(\Theta _{x_{0},x_{1}}^{f^{h}}(\gamma )\) when \(f^{h}\) Mosco converges to \(f\). The first main result proves that if \((X,d)\) is an Hadamard space, \(\{f_{h}\}_{h}\) and \(f\) are proper, \(\lambda \)-convex (for \( \lambda \in \mathbb{R}\)) and lower semi-continuous functions from \(X\) to \( \mathbb{R}\cup \{+\infty \}\), such that \(f^{h}\) Mosco-converges to \(f\), \( \{x_{0}^{h}\}_{h}\) and \(\{x_{1}^{h}\}_{h}\subseteq X\) are two sequences with \(x_{0}^{h}\rightarrow x_{0}\) and \(x_{0}^{h}\rightarrow x_{1}\), and \( \limsup_{h}\left\vert \partial f^{h}\right\vert (x_{0}^{h})<+\infty \) and \( \limsup_{h}\left\vert \partial f^{h}\right\vert (x_{1}^{h})<+\infty \), then \( \Theta _{x_{0},x_{1}}^{f^{h}}\) \(\Gamma \)-converges to \(\Theta _{x_{0},x_{1}}^{f}(\gamma )\) with respect to the \(C([0,1],X)\) topology. For the proof, the authors use properties of the descending slope, of the resolvent operator, and of Hadamard spaces. The \(\Gamma \)-\(\liminf\) inequality is proved applying a lower semicontinuity argument. The \(\Gamma \)-\(\limsup\) inequality is proved through the construction of an appropriate test-function and using properties of the descending slope and of the resolvent operator associated to \(f^{h}\): \(J_{\tau }^{f^{h}}x=\operatorname{argmin}\left\{ f^{h}(\cdot )+\frac{d(\cdot ,x)^{2}}{2}\right\} \). The second main result proves the same \(\Gamma \)-convergence result assuming that \((X,d)\) is a proper geodesic metric space, and \(\{f_{h}\}_{h}\) is a sequence of lower semicontinuous functions \(X\rightarrow \mathbb{R}\cup \{+\infty \}\), such that \(f^{h}\) \(\Gamma \)-converges to a proper and lower semicontinuous function \(f:X\rightarrow \mathbb{R}\cup \{+\infty \}\) and \(f^{h}\) and \(f\) satisfy further hypotheses. For the proof, the authors here replace the resolvent operators with their continuous version given by gradient flow trajectories. The paper ends with two examples.











This page was built for publication: Stability of a class of action functionals depending on convex functions

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6102523)