Maximal monotone operators, convex functions and a special family of enlargements (Q1863700): Difference between revisions

Given a maximal monotone operator \(T:X\rightrightarrows X^{\ast}\), the authors introduce and study the family \(\mathcal{H}(T)\) of lower semi-continuous convex functions \(h:X\times X^{\ast }\rightarrow\mathbb{R}\cup\{+\infty\}\) satisfying \[ h(x,x^{\ast})\geq\langle x,x^{\ast}\rangle\text{,\quad}\forall(x,x^{\ast})\in X\times X^{\ast} \] and \[ x^{\ast}\in T(x)\Longrightarrow h(x,x^{\ast})=\langle x,x^{\ast}\rangle. \] The authors show that if there exists some \(h\in\mathcal{H}(T)\) of the form \(h(x,x^{\ast})=f(x)+g(x^{\ast})\) for some \(f:X\rightarrow\mathbb{R} \cup\{+\infty\}\) and \(g:X^{\ast}\rightarrow\mathbb{R}\cup\{+\infty\},\) then \(T\) is maximal cyclically monotone, and in particular, \(T=\partial f\) and \(g=f^{\ast}.\) They also show that for every \(h\in\mathcal{H}(T),\) the function \[ J(h)(x,x^{\ast}):=h^{\ast}(x^{\ast},x),\quad\forall(x,x^{\ast})\in X\times X^{\ast}, \] also belongs to \(\mathcal{H}(T)\) (\(h^{\ast}\) denotes the conjugate of \(h\)). In particular, for all \((x,x^{\ast})\in X\times X^{\ast}\) one has \[ \min\text{ }\{h(x,x^{\ast}),h^{\ast}(x^{\ast},x)\}\geq\langle x,x^{\ast }\rangle.\tag{3} \] (In a follow-up paper, the authors show that if \(X\) is reflexive, then every lower semi-continuous convex function \(h:X\times X^{\ast}\rightarrow \mathbb{R}\cup\{+\infty\}\) satisfying (3) determines a unique maximal monotone operator \(T\) such that \(h\in\mathcal{H}(T)\) [Proc. Am. Math. Soc. 131, 2379--2383 (2003; Zbl 1019.47038)].) A bijection between \(\mathcal{H}(T)\) and the set \(\mathbb{E}_{c}(T)\) of the non-decreasing closed enlargements of \(T\) that satisfy the transportation formula is also established. This latter set had been previously considered in [\textit{B. F. Svaiter}, Set-Valued Anal. 8, 311--328 (2000; Zbl 0977.47042)]. The least element of \(\mathcal{H}(T)\) with respect to the usual pointwise order of functions is the function \[ \Lambda(x,x^{\ast}):=\beta_{T}(x,x^{\ast})+\langle x^{\ast},x\rangle ,\text{\quad}\forall(x,x^{\ast})\in X\times X^{\ast} \] where \[ \beta_{T}(x,x^{\ast}):=\sup\left\{ \langle x^{\ast}-y^{\ast},y-x\rangle :(y,y^{\ast})\in G(T)\right\} \] is the Brézis-Haraux function for \(T\) [ see \textit{H. Brézis} and \textit{A. Haraux} [Isr. J. Math. 23, 165--186 (1976; Zbl 0323.47041)]. The function \(\Lambda(x,x^{\ast})\) has been independently introduced by \textit{S. Fitzpatrick} [Proc. Cent. Math. Anal. Aust. Natl. Univ. 20, 59--65 (1988; Zbl 0669.47029)] and is also considered in [\textit{J. E. Martínez-Legaz} and \textit{M. Théra}, J. Nonlinear Convex Anal. 2, 243--247 (2001; Zbl 0999.47037)].