The relevance of convex analysis for the study of monotonicity (Q1883100): Difference between revisions

Let \(X\) be a reflexive Banach space with dual space \(X^*\) and the usual coupling functional \(\langle\cdot,\cdot\rangle: X^*\times X\to\mathbb{R}\). The aim of the present paper is the characterization of set-valued mappings \(M: X\rightrightarrows X^*\) with respect to monotonicity and maximal monotonicity by means of the real-valued function \[ c_M(x, x^*):= \begin{cases} \langle x^*,x\rangle\quad &\text{if }x^*\in M(x),\\ +\infty\quad &\text{else,}\end{cases} \] and its (Fenchel) biconjugate function \(p_M(x,x^*):= c^{**}_M(x, x^*)\). It is pointed out that the mapping \(M\) is monotone if one of the following equivalent conditions is fulfilled:{\parindent=1cm\begin{itemize}\item[(i)]\(c_M(x, x^*)\geq c^*_M(x^*, x)\quad\forall(x, x^*)\in X\times X^*\), \item[(ii)]\(p_M(x, x^*)\geq p^*_M(x^*, x)\quad\forall(x, x^*)\in X\times X^*\), \item[(iii)]\(p_M(x, x^*)+ p_M(y, y^*)\geq \langle x^*, y\rangle+ \langle y^*, x\rangle\quad\forall(x, x^*),(y,y^*)\in X\times X^*\),\item[(iv)]\(p_M(x, x^*)\geq \langle x^*, x\rangle\quad\forall(x, x^*)\in X\times X^*\) with equality when \(x^*\in M(x)\),\item[(v)]there exists a (closed) convex function \(p: X\times X^*\to \mathbb{R}\) such that \newline \(p(x, x^*)\geq\langle x^*,x\rangle\quad\forall(x, x^*)\in X\times X^*\) with equality when \(x^*\in M(x)\). \end{itemize}} Moreover, if \(M\) is maximal monotone, then (vi) \(p_M(x, x^*)\geq p^*_M(x^*, x)\geq \langle x^*, x\rangle\quad\forall(x, x^*)\in X\times X^*\), (vii) there exists a (closed) convex function \(p: X\times X^*\to \mathbb{R}\) such that \(p_M(x, x^*)\geq p(x, x^*)= p^*(x^*, x)\geq p^*_M(x^*, x)\quad\forall(x, x^*)\in X\times X^*\) and in this case its graph \(M\) satisfies \[ \begin{aligned} M &= \{(x, x^*)\in X\times X^*: p_M(x, x^*)= \langle x^*, x\rangle\},\\ &= \{(x,x^*)\in X\times X^*: p^*_M(x^*, x)= \langle x^*, x\rangle\},\\ &= \{(x, x^*)\in X\times X^*: p(x, x^*)= \langle x^*, x\rangle\}.\end{aligned} \] The results derived in this paper are used for the discussion of the composition of a maximal monotone mapping with a linear map (i.e., mappings of the form \(M= A^\top NA: X\rightrightarrows X^*\), where \(N: Y\rightrightarrows Y^*\) is maximal monotone and \(A: X\to Y\) is a continuous linear map) and the sum of maximal monotone mappings.