Subgradient representation of multifunctions

From MaRDI portal





Let \(X\) be a Banach space and \(f: X\to \mathbb{R}\) be a Lipschitz function. It is well known that the Clarke subdifferential mapping is weakly\(^*\) cusco, i.e., convex-compact-valued and upper semicontinuous with respect to the weak\(^*\) topology in \(X^*\). In the paper, the authors discuss the converse problem: When can a given (minimal) cusco mapping \(\Phi: X\to X^*\) be represented as the Clarke subdifferential mapping of a real-valued locally Lipschitz function, i.e., when is \(\Phi\) integrable? The first main theorem contains a sufficient condition. In detail, it is shown that for a locally bounded weak\(^*\) cusco mapping \(\Phi\) which possesses a selection \(\sigma\) such that \(\oint_C \sigma(z) dz\leq 0\) for every closed polygonal path \(C\) in \(X\) there exists a Lipschitz function \(f\) such that \(\partial f\subseteq \Phi\). Clearly, if \(\Phi\) is even minimal cusco then equality holds. As conclusions of this result the authors derive Rockafellar's cyclic monotonicity theorem for the subdifferential of convex functions and a more general version of the classical result about the integral characterization of gradient mappings. In the second main theorem, the former condition for the mapping \(\Phi\) is modified to be also necessary for a subdifferential representation.











This page was built for publication: Subgradient representation of multifunctions

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4243936)