The Pseudomonotone Polar for Multivalued Operators
Extracted source text · arXiv 1611.02327
\documentclass[a4paper,12pt]{article}
\usepackage{amssymb,latexsym,amsthm,amsmath,stmaryrd,amsfonts}
\usepackage{mathabx}
\usepackage{graphicx}
\usepackage{color}%
\usepackage{times}
% \usepackage[mathscr]{euscript}
% \usepackage{indentfirst}
%\newtheorem{teo}{Theorem}[section]
% \usepackage{showkeys}
%\newcommand{\rn}{\mathbb{R}^n}
\newcommand{\R}{\mathbb{R}}
%\newcommand{\Q}{\mathbb{Q}}
%\newcommand{\fc}{\displaystyle\frac}
%\newcommand{\la}{\langle}
%\newcommand{\ra}{\rangle}
\newcommand{\N}{\mathbb{N}}
\newcommand{\Ce}{\mathbb{C}}
%\newcommand{\DOM}{\mbox{\rm dom\,}}
%\newcommand{\INT}{\mbox{\rm int}}
%\newcommand{\rra}{\rightrightarrows}
\DeclareMathOperator{\dom}{dom}
\DeclareMathOperator{\ran}{ran}
\DeclareMathOperator{\gra}{gra}
\DeclareMathOperator*{\argmin}{arg\,min}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\newtheorem{theorem}{Theorem}[section]
\newtheorem{lemma}[theorem]{Lemma}
\newtheorem{proposition}[theorem]{Proposition}
\newtheorem{corollary}[theorem]{Corollary}
\theoremstyle{definition}
\newtheorem{definition}[theorem]{Definition}
\newtheorem{example}[theorem]{Example}
\theoremstyle{remark}
\newtheorem{remark}[theorem]{Remark}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% \newcommand{\lam}{\lambda}
% \newcommand{\su}{\subset}
% \newcommand{\GR}{\mbox{\rm Gr}}
\newcommand{\tos}{\rightrightarrows} % point-to-set mappings
\newcommand{\ds}{\displaystyle}
% \newenvironment{dem}{\smallskip \noindent \textsl{Proof}: }
% {\hfill $\Box$\hspace{1in}\medskip \\}
\newcommand{\inner}[2]{\langle #1,#2 \rangle}
%\newcommand{\qed}{\nolinebreak \hfill \rule{2mm}{2mm} \vspace{2mm} \par}
%\newcommand{\R}{{\sf R\hspace*{-0.85ex}\rule{0.15ex}{1.4ex}\hspace*{0.9ex}}}
% \newcommand{\Rs}{{\sf R\hspace*{-0.65ex}\rule{0.11ex}{1.1ex}\hspace*{0.9ex}}}
% \newcommand{\C}{{\sf C\hspace*{-0.85ex}\rule{0.15ex}{1.4ex}\hspace*{0.9ex}}}
%\usepackage{wrapfig}
%\usepackage[all]{xy}
%\usepackage{tikz}
\usepackage{hyperref}
%\usepackage{animate}
% \usepackage{authblk}
%opening
\setlength{\topmargin}{-0.75in} \setlength{\textheight}{9.25in}
\setlength{\oddsidemargin}{0.0in} \setlength{\evensidemargin}{0.0in}
\setlength{\textwidth}{6in}
%opening
\title{The Pseudomonotone Polar for Multivalued Operators}
\author{Orestes Bueno
\thanks{Universidad del Pac\'ifico. Av. Salaverry 2020, Jes\'us Mar\'ia, Lima, Per\'u. Email:
\texttt{\{o.buenotangoa, cotrina\_je\}@up.edu.pe}}
\and John Cotrina\footnotemark[1]}
\DeclareMathOperator{\cone}{cone}
\DeclareMathOperator{\co}{co}
\DeclareMathOperator{\edom}{edom}
\newcounter{enumtemp}
\begin{document}
\maketitle
\begin{abstract}
In this work, we study the pseudomonotonicity of multivalued operators from the point of view of polarity, in an analogous way as the well-known monotone polar due to Mart\'inez-Legaz and Svaiter, and the quasimonotone polar recently
introduced by Bueno and Cotrina.
%
We show that this new polar, adapted for pseudomonotonicity, possesses analogous properties to the monotone and quasimonotone polar, among which are a characterization of pseudomonotonicity, maximality and pre-maximality. Furthermore, we characterize the notion of $D$-maximal pseudomonotonicity introduced by Hadjisavvas.
%
We conclude this work studying the connections between pseudomonotonicity and variational inequality problems.
\bigskip
\noindent{\bf Keywords:}
Pseudomonotonicity, Maximal pseudomonotonicity,
Pseudomonotone Polarity, Variational Inequality
\bigskip
\noindent{\bf MSC (2010):} 47H04, 47H05, 49J53
\end{abstract}
\section{Introduction}
The notion of pseudomonotonicity, in the sense of Karamardian~\cite{Karamardian1976}, plays a big role in certain mathematical applications, for instance, variational inequality problems and problems on consumer theory, see \cite{Crouzeix1997,Crouzeix2011,HadjSchai06,Hadjisavvas2012,John2001}, and the references therein.
The theory of pseudomonotone operators has been steadily developed in the past decades.
%
However, in contrast with the theory of maximal monotone operators, the literature about maximal generalized monotonicity is, to our best knowledge, reduced to \cite{Aussel2013-4,MR2577322,Eberhard07,Had03}.
%
In particular, in \cite{Had03}, the notion of $D$-maximal pseudomonotonicity is introduced, which is based on an equivalence related to variational inequalities.
%Between the existing characterizations of pseudomonotonicity, there is the characterization in terms of the relationship between Minty and Stampacchia variational inequalities~\cite{}.
The monotone polar was introduced by Mart\'inez-Legaz and Svaiter~\cite{BSML}.
%
Its main feature is to provide an easy characterization of monotonicity, maximal monotonicity and pre-maximal monotonicity. In a similar way, Bueno and Cotrina~\cite{BueCot16} introduced the quasimonotone polar which, besides characterizing (maximal, pre-maximal) quasimonotonicity, is related to the Minty variational inequality problem and, in particular, to the study of the adjusted normal operator defined by Aussel and Hadjisavvas~\cite{AuHad2005}.
The aim of this paper is to introduce the pseudomonotone polar and study its connection with the maximal pseudomonotonicty (including the weaker $D$-maximality) and its connection with the Minty and Stampacchia variational inequality problems. The paper is organized as follows: in section 3 we provide the definition of pseudomonotone polar, along with a characterization in terms of normal cones and the study of its set of zeros. In sections 4 and 5 we use the polar to study the pseudomonotonicity, maximal and pre-maximal pseudomonotonicity of an operator and, in addition, we characterize the $D$-maximality in terms of the pseudomonotone polar. Finally in section 6, we relate the pseudomonotonicity with the solution sets of variational inequality problems.
%%%
%
%
%El objetivo de este trabajo es definir el polar pseudo de un operador. Recordemos que el polar monotono fue definido por ML y Svaiter en 2005 []. Recientemente, Bueno y Cotrina definieron el polar quasimonotono []. In both these works, the authors related the respective polars to the (quasi-)monotonicity, maximality and pre-maximality (that is, operators with a unique maximal (quasi-)monotone extension).
%
%Respecto a la pseudomonotonia maximal, estudios al respecto fueron dados recientemente por Hadjisavvas y otros. En particular, Hadjisavvas define una nocion debil llamada D-maximalidad. En la seccion 5 de este trabajo, nosotros usamos el polar pseudo para estudiar esta nocion de maximalidad pseudo, donde conseguimos simplificar las notaciones y extender algunos resultados.
%
%Los problemas de desig. variacional fueron definidos por tal y tal y tal y tal. El nexo entre los SVIP y la pseudomonotonia, fue dado por R. John [cita] y visto nuevamente por Cotrina y su papi. En la seccion 6, nosotros estudiamos la relacion existente entre el polar pseudo y las soluciones de los VIP de minty y stampachia. En particular, probamos que blah blah y blah.
%
%Finalmente, en la seccion 7, ....
%
%In \cite{BSML}, the authors introduced the polar monotone and study the relation between maximal monotinicity
%to operator and its polar.
%In \cite{Aussel2013-4}, the authors introduced a notion of maximality, called $AE$-maximality in \cite{BueCot16},
%for quasimonotone operators and characterized
%this maximality. In \cite{BueCot16}, the authors introduced the polar quasimonotone and
%characterized the quasimonotonicity and maximal quasimonotonicity.
%
%%
%
%The aim of this work is to extend the notion of polar to the pseudomonotone case.
%We show that the \emph{pseudomonotone polar} shares the same properties with monotone and quasimonotone polar,
%and allows a characterization of pseudomonotonicity, in a parallel way than in \cite{BSML}.
%We study the maximality of this kind of operators. We understand maximality in the sense of graph inclusion.
%As expected, we show that the pseudomonotone polar is useful to characterize maximal pseudomonotone
%and also \emph{pre-maximal} pseudomonotone, that is, operators which have an unique maximal pseudomonotone extension.
%
%%
%In \cite{Had03}, introduce the $D$-maximality for pseudomonotone operators.
%We study the relation between the graph inclusion maximality and the $D$-maximality.
%Moreover, we characterize the $D$-maximality in terms of polar pseudomonotone.
%
%%
%
%The Stampacchia and Minty Variational Inequality Problems was introduced in \cite{Stampacchia1964} and
%in \cite{minty1967}, respectively. However, was John in \cite{John2001} who established the relation
%between generalized monotinicity and these variational inequalities problems.
%We study the relation between the inclusion of solutions sets of variational inequalities problems associated
%to a multivalued operator and his polar pseudomonotone. We show a new characterization of
%pseudomonotonicity and pre-maximal pseudomonotonicity terms of the relation between solution sets of variational inequalities problems
%associated to the operator and his polar.
%
%%
%
%Finally, we conclude this work, by given sufficient condition for pre-maximality under
%upper sign continuity, introduced by Hadjisavvas in \cite{Had03}.
\section{Preliminary definitions and notations}
Let $U,V$ be non-empty sets.
% For a set $S\subset U\times V$, we denote the projections over $U$ and $V$ as $\proy_U(S)=\{u\in U\::\:\exists\,v\in V,\,(u,v)\in S\}$ and $\proy_V(S)=\{v\in V\::\:\exists\,u\in U,\,(u,v)\in S\}$.
A \emph{multivalued operator} $T:U\tos V$ is an application $T:U\to \mathcal{P}(V)$, that is, for $u\in U$, $T(u)\subset V$.
The \emph{domain}, \emph{range} and \emph{graph} of $T$ are defined, respectively, as
\begin{gather*}
\dom(T)=\big\{u\in U\::\: T(u)\neq\emptyset\big\},\qquad \ran(T)=\bigcup_{u\in U}T(u),\\ %=\proy_V(\gra(T)).
\text{and }\gra(T)=\big\{(u,v)\in U\times V\::\: v\in T(u)\big\}.
\end{gather*}
Given $A\subset U$, the \emph{restriction} of $T$ to $A$ is the operator $T|_A$ defined as $T|_A(x)=T(x)$,
if $x\in A$, and $T|_A(x)=\emptyset$, otherwise.
From now on, we will identify multivalued operators with their graphs, so we will write $(u,v)\in T$ instead of $(u,v)\in \gra(T)$.
Let $V$ be a real vector space, and let $A\subset V$. The \emph{convex hull} of $A$, denoted as $\co(A)$, is the smallest convex set (in the sense of inclusion) which contains $A$. The \emph{segment} between $u$ and $v$ in $V$ is the set $[u,v]=\co\{u,v\}$.
%
The \emph{conic hull} of $A$ is the set
\[
\cone(A)=\{tv\in V\::\:t\geq 0,\, v\in A\},
\]
whereas its \emph{strict conic hull} is the set
\[
\cone_{\circ}(A)=\{tv\in V\::\:t> 0,\, v\in A\}.
\]
Furthermore, given $T:U\tos V$, we denote $\cone(T):U\tos V$ (respectively, $\cone_{\circ}(T):U\tos V$) as the operator defined by $\cone(T)(x)=\cone(T(x))$ (respectively, $\cone_{\circ}(T)(x)=\cone_{\circ}(T(x))$), for all $x\in\dom(T)$. In addition, the \emph{set of zeros} of $T$ is the set
\[
Z_T=\{x\in X\::\: 0\in T(x)\}.
\]
Note that $Z_T=Z_{\cone_{\circ}(T)}$.
%the \emph{effective domain} of $T$ is the set $\edom(T)=\{u\in U\::\:T(u)\neq \{0\}\}$. Note that $T\subset \cone(T)$ and $\edom(\cone(T))=\edom(T)=\edom(T\cup(X\times \{0\}))$.
\newcommand{\cones}{\cone_{\circ}}
Let $X$ be a Banach space and $X^*$ be its topological dual.
% We will identify $X$ with its image under the canonical injection into $X^{**}$, so in this way $X$ is called \emph{reflexive} if $X=X^{**}$ and \emph{non-reflexive}, otherwise.
The \emph{duality product} is defined as $\inner{x}{x^*}=x^*(x)$.
Let $C\subset X$ be a convex set and $x\in X$. The \emph{normal cone} of $C$ at $x$ is the set
\[
N_{C}(x)=\{x^*\in X^*\::\: \inner{y-x}{x^*}\leq 0,\,\forall\, y\in C\}.
\]
Moreover, the \emph{strict normal cone} of $C$ at $x$ is the set
\[
N^{\circ}_{C}(x)=\{x^*\in X^*\::\: \inner{y-x}{x^*}< 0,\,\forall\, y\in C\}.
\]
By a vacuity argument, we note that if $C=\emptyset$ then $N_C(x)=N^{\circ}_C(x)=X^*$.
The \emph{monotone polar}~\cite{BSML} of an operator $T:X\tos X^*$ is the operator $T^{\mu}$ defined as
\[
T^{\mu}=\{(x,x^*)\in X\times X^*\::\: \inner{x-y}{x^*-y^*}\geq 0,\,\forall\,(y,y^*)\in T\}.
\]
In the same way, the \emph{quasimonotone polar}~\cite{BueCot16} of $T$ is the operator $T^{\nu}$ defined as
\[
T^{\nu}=\{(x,x^*)\in X\times X^*\::\:\min\{\inner{y-x}{x^*},\inner{x-y}{y^*}\}\leq 0,\,\forall\,(y,y^*)\in T\}.
\]
The notions of \emph{(quasi)monotonicity}, \emph{maximal} and \emph{pre-maximal (quasi)monotonicity} can be expressed in terms of the (quasi)monotone polar. More precisely:
\begin{enumerate}
\item $T$ is (quasi)monotone if, and only if, $T\subset T^{\mu} (T^\nu)$;
\item $T$ is maximal (quasi)monotone if, and only if, $T=T^{\mu}(T^\nu)$;
\item $T$ is pre-maximal (quasi)monotone if, and only if, $T$ and $T^{\mu} (T^\nu)$ are (quasi)monotone.
\end{enumerate}
%
% \begin{proposition}
% Let $T:X\tos X^*$ be a multivalued operator. If
% $\co(\dom(T))\subset \dom(T^\mu)$ then $T$ is monotone.
% \end{proposition}
% \begin{proof}
% Let $(x,x^*),(y,y^*)\in T$ and consider $(z,z^*)\in T^\mu$ such that $z=(x+y)/2$. Then holds
% \begin{eqnarray*}
% \langle x^*-z^*,x-z\rangle&\geq&0\\Para que van a jugar si igual Alianza no va a ir a las semifinales de aca no hay fecha hasta despues del APEC ya para esa fecha sabremos quienes son los 4 y con eso no es necesario que se juegue el clasico
% \langle z^*-y^*,z-y\rangle&\geq&0
% \end{eqnarray*}
% As $x-z=z-y=(x-y)/2$ we deduce
% \[
% \langle x^*-y^*,x-y\rangle\geq0.
% \]
% Therefore, $T$ is monotone.
% \end{proof}
\section{The pseudomonotone polar}
We say that $(x,x^*)$ is in a \emph{pseudomonotone relation} with $(y,y^*)$, denoted by $(x,x^*)\sim_p(y,y^*)$, if
\[
\min\{\inner{x-y}{y^*},\inner{y-x}{x^*}\}<0\quad\text{or}\quad\inner{x-y}{y^*}=\inner{y-x}{x^*}=0.
\]
It is clear that $\sim_p$ is a reflexive and symmetric relation on $X\times X^*$. Although it is not transitive, a weaker form of transitivity can be obtained.
%%%%%%%%%%%%%55
\begin{proposition}\label{pro:pseudo-new}
Let $(x,x^*),(y,y^*),(z,z^*)\in X\times X^*$, with $z\in[x,y]$. If $(x,x^*)\sim_p(z,z^*)$ and $(z,z^*)\sim_p(y,y^*)$ then $(x,x^*)\sim_p(y,y^*)$.
\end{proposition}
\begin{proof}
Let $z=tx+(1-t)y$, with $t\in]0,1[$ (the result is trivial when $t=0,1$). Note that
\begin{gather*}
\inner{z-y}{y^*}=t\inner{x-y}{y^*},\qquad\inner{z-x}{x^*}=(1-t)\inner{y-x}{x^*}\\
(1-t)\inner{y-z}{z^*}=-t\inner{x-z}{z^*}.
\end{gather*}
Without loss of generality, assume that $\inner{x-y}{y^*}>0$ and $\inner{y-x}{x^*}\geq 0$. Then $\inner{z-x}{x^*}\geq 0$ and $\inner{z-y}{y^*}>0$. As $(y,y^*)\sim_p(z,z^*)$, the latter implies that $\inner{y-z}{z^*}<0$, which in turn implies $\inner{x-z}{z^*}>0$. This contradicts $(x,x^*)\sim_p(z,z^*)$.
\end{proof}
%%%%%%%%%%%%%%%
The proof of the following lemma is analogous to Lemma 3.4 in~\cite{BueCot16}
\begin{lemma}\label{3.6}
Let $T:X\tos X^*$ be a multivalued operator and let $(x,x^*)\in T^{\rho}$, $(y,y^*)\in T$ and $t,s> 0$. Then $(x,tx^*)\sim_p(y,sy^*)$.
\end{lemma}
\begin{definition}
The \emph{pseudomonotone polar} of $T:X\tos X^*$, $T^{\rho}$ is given by
\[
T^{\rho}=\{(x,x^*)\in X\times X^*:~(x,x^*)\sim_p(y,y^*),~\forall (y,y^*)\in T\}
\]
\end{definition}
It is clear from the definition that
\begin{equation}\label{eq:polarsubset}
T^\mu\subset T^\rho\subset T^\nu,
\end{equation}
for every operator $T:X\tos X^*$.
\begin{example}\label{exa:xtimes0}
For $x\in X$, $\{(x,0)\}^\rho=\{(x,0)\}^\mu$. Moreover, it is straightforward to verify that
$(X\times \{0\})^\rho= X\times \{0\}$.
\end{example}
The following proposition lists some polarity properties of $T^{\rho}$.
\begin{proposition}\label{pro:polarp}\quad
\begin{enumerate}
\item $\displaystyle\left(\bigcup_{i\in I}T_i\right)^{\rho}=\bigcap_{i\in I}T_i^{\rho}$.
\item $T\subset T^{\rho\rho}$.
\item $T^{\rho\rho\rho}=T^{\rho}$.
\item If $T\subset S$ then $S^{\rho}\subset T^{\rho}$.
\setcounter{enumtemp}{\theenumi}
\end{enumerate}
In addition
\begin{enumerate}
\setcounter{enumi}{\theenumtemp}
\item $\emptyset^{\rho}=X\times X^*$, $(X\times X^*)^{\rho}=\emptyset$
\item $T^{\rho\mu}\subset T^{\mu\mu}\subset T^{\mu\rho}$
\item $T^{\rho\mu}\subset T^{\rho\rho}\subset T^{\mu\rho}$
\end{enumerate}
\end{proposition}
\begin{proof}
Items {\it 1} to {\it 4} are direct consequences of the fact that the map $T\mapsto T^{\rho}$ is a polarity~\cite{BSML}. Item {\it 5} follows directly from the definition and a vacuity argument. Items {\it 6} and {\it 7} come from equation~\eqref{eq:polarsubset} and item {\it 4}.
\end{proof}
%REVISAR
% \begin{proposition}
% Let $T:X\tos X^*$ be a multivalued operator. Then .
% \end{proposition}
%% tuu subset tur
%% tu subset tr entonces tru subset tuu subset tur
%% tru subset trr
%% tu subset tr entonces tru subset trr subset tur
We now deal with the set of zeros of $T^{\rho}$.
\begin{proposition}\label{pro:zeros}
Let $T:X\tos X^*$ be a multivalued operator. Then $Z_{T^\rho}$ is a weak-closed convex set.
\end{proposition}
\begin{proof}
It is enough to observe that $(x,0)\in T^{\rho}$ if, and only if, $\inner{x-y}{y^*}\leq 0$, for all $(y,y^*)\in T$. Therefore, definining $C(y,y^*)=\{x\in X:~\langle x-y,y^*\rangle\leq0\}$, we have
\[
Z_{T^\rho}= \bigcap_{(y,y^*)\in T}C(y,y^*).
\]
The proposition follows, since each $C(y,y^*)$ is weak-closed and convex.
\end{proof}
%REVISAR
\begin{proposition}
Let $T:X\tos X^*$ be a multivalued operator. Then $Z_{T^\mu}=Z_{T^\rho}$.
\end{proposition}
\begin{proof}
Since $T^\mu\subset T^\rho$, $Z_{T^\mu}\subset Z_{T^\rho}$. On the other hand, given $x\in Z_{T^{\rho}}$, from the proof of Proposition~\ref{pro:zeros}, we have $\inner{x-y}{y^*}\leq 0$, for all $(y,y^*)\in T$. This in turn implies
\[
\inner{x-y}{0-y^*}\geq 0,\qquad\forall\,(y,y^*)\in T,
\]
that is, $(x,0)\in T^{\mu}$. This proves the proposition.
\end{proof}
Corollary~3.8 in \cite{BueCot16} shows that the images of the quasimonotone polar are conic, convex and weak$^*$-closed. In a similar way, we have the following proposition.
\begin{proposition}
Let $T:X\tos X^*$ be multivalued operator. Then the following assertions hold.
\begin{enumerate}
\item $\cones(T)^{\rho}=T^{\rho}=\cones(T^{\rho})$. In particular $T^{\rho}(x)$ is a cone, for all $x\in X$;
\item $T^{\rho}(x)$ is a convex set, for all $x\in X$;
\end{enumerate}
\end{proposition}
A natural question is if the closure of the pseudomonotone polar coincides with the quasimonotone polar of an operator.
The answer is negative, take for instance $T=\R\times\{0\}$. It is clear that $T={\rm cl(T^\rho)}\neq T^\nu=\R^2$.
Given $T:X\tos X^*$ a multivalued operator, consider the sets
\[
V_T(x)=\{y\in X\::\: \exists\,y^*\in T(y),\,\inner{x-y}{y^*}>0\}.
\]
and
\[
W_T(x)=\{y\in X\::\: \exists\,y^*\in T(y),\,\inner{x-y}{y^*}=0\}.
\]
Thus, we can state the following characterization of $T^{\rho}(x)$.
\begin{proposition}\label{pro:trhogen}
Let $T:X\tos X^*$ be a multivalued operator. Then, for $x\in X$,
\[
T^{\rho}(x)=N^{\circ}_{V_T(x)}(x)\cap N_{W_T(x)}(x).
\]
\end{proposition}
\begin{proof}
Let $x^*\in T^{\rho}(x)$. If $y\in V_T(x)$ then there exists $y^*\in T(y)$ such that $\inner{x-y}{y^*}>0$. Therefore $(x,x^*)\sim_p (y,y^*)$ so $\inner{y-x}{x^*}<0$. Thus $x^*\in N^{\circ}_{V_T(x)}(x)$, since $y\in V_T(x)$ was arbitrary . On the other hand, if $y\in W_T(x)$, then there exists $y^*\in T(y)$ such that $\inner{x-y}{y^*}=0$. Again, we have $(x,x^*)\sim_p (y,y^*)$, so $\inner{y-x}{x^*}\leq 0$ and $x^*\in N_{W_T(x)}(x)$.
Conversely, take $x^*\in N^{\circ}_{V_T(x)}(x)\cap N_{W_T(x)}(x)$ and $(y,y^*)\in T$. If $\inner{x-y}{y^*}<0$ then clearly $(x,x^*)\sim_p(y,y^*)$. Now assume that $\inner{x-y}{y^*}=0$, that is, $y\in W_T(x)$. Since $x^*\in N_{W_T(x)}(x)$, $\inner{y-x}{x^*}\leq 0$. This also implies that $(x,x^*)\sim_p(y,y^*)$. Finally, assume that $\inner{x-y}{y^*}>0$, that is $y\in V_T(x)$. In this case, $x^*\in N^{\circ}_{V_T(x)}(x)$ implies $\inner{y-x}{x^*}<0$, which in turn implies $(x,x^*)\sim_p(y,y^*)$. Hence, we conclude $(x,x^*)\in T^{\rho}$.
\end{proof}
\begin{corollary}\label{cor:trhozeros}
Let $T:X\tos X^*$ be a multivalued operator and let $x\in X$. The following are equivalent:
\begin{enumerate}
\item $V_T(x)=\emptyset$,
\item $x\in Z_{T^{\rho}}$,
\item $T^{\rho}(x)=N_{W_T(x)}(x)$,
\item $x\in\dom(T^{\rho})$ and $T^{\rho}(x)$ is weak$^*$-closed.
\end{enumerate}
\end{corollary}
\begin{proof}
If $V_T(x)=\emptyset$ then $N^{\circ}_{V_T(x)}(x)=X^*$ thus implying {\it 3}. Since a normal cone is always non-empty and weak$^*$-closed, {\it 3} implies {\it 4}. Now assume {\it 4} and take $x^*\in T^{\rho}(x)$. If $x^*=0$ then $x\in Z_{T^{\rho}}$, so assume that $x^*\neq 0$. Since $T^{\rho}(x)$ is a cone, $tx^*\in T^{\rho}(x)$, for all $t>0$. Taking the limit when $t\to 0^+$, as $T^{\rho}(x)$ is weak$^*$-closed, we conclude that $0\in T^{\rho}(x)$. Finally, assuming {\it 2}, from Proposition~\ref{pro:trhogen} we obtain $0\in N^{\circ}_{V_T(x)}(x)$, which is absurd, unless $V_T(x)=\emptyset$.
\end{proof}
%\begin{corollary}
%Let $T:X\tos X^*$ be a multivalued operator and let $x\in X$. Then $V_T(x)=\emptyset$ if, and only if, $x\in Z_{T^{\rho}}$. In addition, when these conditions hold,
%\[
%T^{\rho}(x)=N_{W_T(x)}(x).
%\]
%\end{corollary}
%
%
%\begin{corollary}
%Let $T:X\tos X^*$ be a multivalued operator and let $x\in X$. Then $x\in Z_{T^{\rho}}$ if, and only if, $x\in\dom(T^{\rho})$ and $T^{\rho}(x)$ is weak$^*$-closed.
%\end{corollary}
%\begin{proof}
%The ``only if'' part is a direct consequence of Corollary~\ref{cor:trhozeros}. For the ``if'' part, .
%\end{proof}
%%The set $L(T,x)$ is not necessarily convex, however Hadjisavvas considered the normal operator
%%\[
%%N_{L(T,x)}(x)=\{x^*\in X^*\::\:\inner{y-x}{x^*}\leq 0,\,\forall\,y\in L(T,x)\}.
%%\]
%%REVISAR
%\begin{proposition}\label{prop1}
%Let $T:X\tos X^*$ be a multivalued operator. For all $x\in Z_{T^\rho}$ hold $N_{L(T,x)}= T^\rho(x)$.
%\end{proposition}
%\begin{proof}
%Fix $x\in Z_{T^\rho}$. Take any $x^*\in N_{L(T,x)}$, and assume that $(x,x^*)\notin T^{\rho}$. Then there exists $(y_0,y_0^*)\in T$ such that
%\begin{equation}\label{eq:trho}
%\inner{x-y_0}{y_0^*}\geq 0\qquad\text{and}\qquad \inner{y_0-x}{x^*}\geq 0,
%\end{equation}
%but not both equal to zero at the same time. The first inequality in~\eqref{eq:trho} implies that $y_0\in L(T,x)$ and this in turn implies $\inner{y_0-x}{x^*}=0$, from the second inequality in~\eqref{eq:trho} and the fact that $x^*\in N_{L(T,x)}$. On the other hand, since $0\in T^{\rho}(x)$, $\inner{x-y_0}{y_0^*}\leq 0$, and this, together with the first inequality in~\eqref{eq:trho} implies $\inner{x-y_0}{y_0^*}=0$. We arrived at a contradiction, therefore, $x^*\in T^{\rho}(x)$.
%
%Conversely, assume that $x^*\notin N_{L(T,x)}(x)$. Then, there exists $y_0\in L(T,x)$ such that $\inner{y_0-x}{x^*}>0$ and, moreover, there exists $y_0^*\in T(y_0)$ such that $\inner{x-y_0}{y^*_0}\geq 0$. These inequalities, however, imply that $(x,x^*)\notin T^{\rho}$.
%%
%%Let $(y,y^*)\in T$. If $y\notin L_{T(x)}$ then for all $y^*\in T(y)$
%%\[
%% \langle x-y,y^*\rangle<0.
%%\]
%%So $(x,x^*)\sim_p(y,y^*)$. Now, if $y\in L_{T(x)}$ then $\langle y-x,x^*\rangle\leq0$ and since that $x\in Z_{T^\rho}$
%%one has that $\langle x-y,y^*\rangle\leq0$. So, $(x,x^*)\sim_p(y,y^*)$ and therefore $x^*\in T^\rho(x)$.\\
%%\newline
%%Reciprocally, let $x\in Z_{T^\rho}$ and suppose there is $x^*\in T^\rho(x)\setminus N_{L_{T(x)}}$ then there exists $y\in L_{T(x)}$ such that
%%\[
%% \langle y-x,x^*\rangle>0.
%%\]
%%Since that $y\in L_{T(x)}$ there exists $y^*\in T(y)$ such that $\langle y-x,y^*\rangle\leq0$. So,
%%\[
%% \min\{\langle y-x,x^*\rangle,~\langle x-y,y^*\rangle\}\geq0
%%\]
%%and this is a contradiction.
%\end{proof}
\section{On pseudomonotone operators}
Recall that an operator $T:X\tos X^*$ is \emph{pseudomonotone} if, for every $(x,x^*),(y,y^*)\in T$, the following implication holds
\[
\inner{y-x}{x^*}\geq0\quad\Longrightarrow\quad \inner{y-x}{y^*}\geq 0,
\]
or, equivalently, every $(x,x^*),(y,y^*)\in T$ are pseudomonotonically related, that is,
\[
\min\{\inner{y-x}{x^*},\inner{x-y}{y^*}\}< 0 \mbox{ or } \inner{y-x}{x^*}=\inner{x-y}{y^*}=0.
\]
As was in the monotone and quasimonotone case, the pseudomonotone polar provides a way of characterizing pseudomonotonicity.
\begin{proposition}\label{pro:4.1} Let $T:X\tos X^*$. The following conditions are equivalent:
\begin{enumerate}
\item $T$ is pseudomonotone,
\item $T\subset T^{\rho}$,
\item $T^{\rho\rho}\subset T^{\rho}$,
\item $T^{\rho\rho}$ is pseudomonotone,
\item $\cones(T)$ is pseudomonotone.
\end{enumerate}
\end{proposition}
%REVISAR
\begin{corollary}\label{123}
Let $T:X\tos X^*$ be a pseudomonotone operator. Then $(x,x^*)\in T^\rho$ if, and only if, $T\cup\{(x,x^*)\}$ is pseudomonotone.
\end{corollary}
\begin{example}\label{ejem-variational}
Let $T=(\R\times\{0\})\cup\{(0,1)\}$. Straightforward calculations show that
\[% \begin{align*}
T^{\nu}=\{(x,x^*)\in\R^2\::\:x\leq 0\text{ or }x^*\geq 0\}\quad\text{and}\quad
T^{\rho}=\{(x,0)\in\R^2\::\:x\leq 0\}.
\]% \end{align*}
Therefore, $T$ is quasimonotone but not pseudomonotone. Note, in addition, that $T^{\rho}$ is monotone.
\end{example}
%A natural question is: If $T$ is a quasimonotone operator and its $T^\rho$ is pseudomonotone, then is $T$ pseudomonotone?
%The following example gives a negative answer.
%REVISAR
Joining together Propositions~\ref{pro:zeros} and \ref{pro:4.1}, we obtain the following corollary.
\begin{corollary}
Let $T:X\tos X^*$ be a pseudomonotone operator. Then $\overline{\co}^{w}(Z_T)\subset Z_{T^\rho}$.
\end{corollary}
The latter inclusion is strict
in general, as shown by the following example.
\begin{example}\label{ejem1}
Let $T:\R\tos\R$ be a multivalued operator defined as $T=(\R_-\times\{-1\})\cup(\R_+\times\{1\})$.
%\[
% T(x)=\left\lbrace\begin{array}{cc}
% -1&,x<0\\
% \{-1,1\}&,x=0\\
% 1&,x>0
% \end{array}
%\right.
%\]
Clearly, $T$ is pseudomonotone and $Z_T=\emptyset$.
However,
\[
T^{\rho}=\{(x,x^*)\in\R^2\::\:xx^*>0\text{ or }x=0\}
\]
and $Z_{T^{\rho}}=\{0\}$.
\end{example}
%we give a sufficient condition for the pseudomonotonicity of $T$ by inclusion of its domain in the
%domain of its polar pseudomonotone.
\begin{proposition}\label{pseudo-dom}
Let $T:X\tos X^*$ be a multivalued operator. If
$\co(\dom(T))\subset \dom(T^\rho)$ then $T$ is pseudomonotone.
\end{proposition}
\begin{proof}
Take $(x,x^*),(y,y^*)\in T$ and let
$z=
(x+y)/2 %\dfrac{x+y}{2}
\in[x,y]\subset\co(\dom(T))$.
Then $z\in \dom(T^{\rho})$ and there exists $z^*\in T^{\rho}(z)$. In particular, $(z,z^*)$ is pseudomonotonically related to both $(x,x^*)$ and $(y,y^*)$. Therefore, by Proposition~\ref{pro:pseudo-new}, $(x,x^*)$ and $(y,y^*)$ are pseudomonotonically related. The proposition follows.
\end{proof}
%
% \begin{proof}
% Let $(x,x^*),(y,y^*)\in T$ such that
% \begin{equation}\label{1-pseudo}
% \langle y-x,x^*\rangle >0
% \end{equation}
% Consider $(z,z^*)\in T^\rho$ with $z=(x+y)/2$. Thus, the inequality (\ref{1-pseudo}) implies
% \[
% \langle z-x,x^*\rangle >0
% \]
% and as $(x,x^*)\sim_\rho(z,z^*)$ we have
% \begin{equation}\label{2-pseudo}
% \langle z-x,z^*\rangle >0
% \end{equation}
% Since $z-x=y-z$, the inequality (\ref{2-pseudo}) implies
% $\langle y-z,z^*\rangle >0$. Now, as $(z,z^*)\sim_\rho (y,y^*)$ we have
% $\langle y-z,y^*\rangle>0$. Therefore, using $y-z=(y-x)/2$
% we deduce $\langle x-y,y^*\rangle<0$.
% \end{proof}
%The following example shows that the inclusion $\dom(T)\subset\dom(T^\rho)$ dont implies the pseudomonotonicity of $T$.
The following example shows that we cannot drop the condition of taking the convex hull of the domain in the last proposition.
\begin{example}
Let $T=\{(0,1), (1,0)\}$. Then $T^{\rho}=(]-\infty,0]\times\R_-)\cup([1,+\infty[\times \R_{++})$ and $\dom(T)=\{0,1\}\subset\R\setminus]0,1[=\dom(T^{\rho})$. However $T$ is not pseudomonotone.
\end{example}
As a direct consequence of Proposition \ref{pseudo-dom} we have the following corollary.
\begin{corollary}\label{cor:pseudo}
Let $T:X\tos X^*$ be a multivalued operator. If $\dom(T^\rho)=X$ then $T$ is pseudomonotone.
\end{corollary}
\begin{remark}
Propositions~\ref{pro:pseudo-new} and \ref{pseudo-dom} and Corollary~\ref{cor:pseudo} can be restated for monotonicity instead of pseudomonotonicity, with similar proofs.
\end{remark}
We say that $T$ is \emph{maximal pseudomonotone} if $T$ is pseudomonotone and it is maximal in the sense of inclusion, that is,
if $T\subset S$ and $S$ is pseudomonotone, then $T=S$. A direct consequence of Zorn's Lemma shows that every pseudomonotone
operator must have a maximal pseudomonotone extension, that is, a maximal pseudomonotone operator which contains it.
%The following proposition is analogous to~\cite[Proposition 22]{BSML}.
\begin{proposition}\label{pro:maxpseu}
Let $T:X\tos X^*$ be a pseudomonotone operator and let $P(T)$ be the set of maximal pseudomonotone extensions of $T$. Then
\begin{enumerate}
\item $T^{\rho}=\displaystyle\bigcup_{M\in P(T)}M$;
\item $T^{\rho\rho}=\displaystyle\bigcap_{M\in P(T)}M$;
\item $T$ is maximal pseudomonotone if, and only if, $T=T^{\rho}$.
\end{enumerate}
\end{proposition}
From Example~\ref{exa:xtimes0}, the operator $T=X\times\{0\}$ satisfies $T=T^{\rho}$, so $T$ is both maximal pseudomonotone and maximal monotone. The following proposition shows that this operator is the only one with this property.
\begin{proposition}
Let $T:X\tos X^*$ be a monotone operator. Then $T$ is maximal pseudomonotone if, and only if, $T=X\times\{0\}$.
\end{proposition}
\begin{proof}
The ``if'' part follows from Example~\ref{exa:xtimes0} and Proposition~\ref{pro:maxpseu}. We now prove the ``only if'' part.
From the monotonicity and maximal pseudomonotonicity of $T$, we obtain the inclusions $T\subset T^\mu\subset T^\rho=T$, which in turn imply that $T$ is maximal monotone.
We assert that $Z_T=\dom(T)$. Indeed, let $x\in\dom(T)$ and take $x^*\in T(x)$. Note that the maximal monotonicity and pseudomonotonicity of $T$ together imply that $T^{\rho}(x)=T(x)$ is a weak$^*$-closed conic set. Using Corollary~\ref{cor:trhozeros}, we conclude that $x\in Z_T$. In addition, Proposition~\ref{pro:zeros} implies that $\dom(T)$ is weak-closed and convex.
%
Assume that there exists $x\notin \dom(T)$. Using the Hahn-Banach Theorem, there exists $x^*\in X^*$ such that $\inner{y-x}{x^*}<0$, for all $y\in\dom(T)$. This implies that $(x,x^*)\sim_p(y,y^*)$, for all $(y,y^*)\in T$, that is, $(x,x^*)\in T^{\rho}=T$, a contradiction. Therefore $\dom(T)=X$. The proposition follows by noting that $T(x)=N_X(x)=\{0\}$, for all $x\in X$.
\end{proof}
If $T$ is a pseudomonotone operator such that $T^{\rho}$ is pseudomonotone (therefore, maximal pseudomonotone),
then $T$ will be called \emph{pre-maximal pseudomonotone}. A pre-maximal pseudomonotone operator is not
necesarily maximal, but possesses a unique maximal pseudomonotone extension.
\begin{lemma}\label{pre-max}
Let $T,S:X\tos X^*$ be two pseudomonotone operators such that $T\subset S$. If $T$ is pre-maximal pseudomonotone then $S$ is pre-maximal pseudomonotone and $T^{\rho}=S^{\rho}$
\end{lemma}
\begin{proof}
Clearly $T\subset S\subset S^\rho\subset T^\rho$. Since that $T^\rho$ is pseudomonotone its follow that $S^\rho$ is pseudomonotone.
Therefore, $S$ is pre-maximal pseudomonotone.
\end{proof}
For $T:X\tos X^*$ and $\alpha^*\in X^*$, consider the \emph{linear perturbation} $T+\alpha^*:X\tos X^*$ defined as
\[
(T+\alpha^*)(x)=T(x)+\alpha^*,\qquad \forall x\in X.
\]
The relation between the quasimonotonicity of the linear perturbations $T+\alpha^*$ and the monotonicity of $T$ was originally established by Aussel, Corvellec and Lassonde~\cite{AusCorLass94},
%
and then extended to include maximality by Aussel and Eberhard~\cite[Proposition~6]{Aussel2013-4}.
Recently, Bueno and Cotrina presented an analogous version of this result considering pre-maximality~\cite{BueCot16}.
%
As pre-maximal pseudomonotonicity does not imply pre-maximal quasimonotonicity (see Example~\ref{exa:xtimes0}),
we now present a version of the previous results which relates pre-maximal pseudomonotonicity and monotonicity.
\begin{proposition}
Let $T:X\tos X^*$ be a multivalued operator such that $T+\alpha^*:X\tos X^*$ is pre-maximal pseudomonotone, for all $\alpha^*\in X^*$. Then $T$ is pre-maximal monotone
and $T^\mu$ is pre-maximal pseudomonotone.
\end{proposition}
\begin{proof}
Monotonicity of $T$ is obtained from~\cite[Proposition 2.1]{AusCorLass94}. Following the same proof
of Proposition 4.12 in \cite{BueCot16} we obtain that $T^{\mu}$ is monotone.
%Let $(x,x^*), (y,y^*)\in T^\mu$. Then, for all $\alpha^*\in X^*$,
%\[
%(x,x^*+\alpha^*), (y,y^*+\alpha^*)\in (T+\alpha^*)^\mu.
%\]
%In particular, for $\alpha_0^*=-\displaystyle\frac{x^*+y^*}{2}\in X^*$,
%\[
%\left(x,\dfrac{1}{2}(x^*-y^*)\right), \left(y,\dfrac{1}{2}(y^*-x^*)\right)\in (T+\alpha_0^*)^\mu.
%\]
%Since $(T+\alpha^*)^\mu\subset (T+\alpha^*)^\rho$, and using the fact that $(T+\alpha^*)^\rho$ is pseudomonotone, we obtain
%\[
%\min\{\inner{y-x}{x^*-y^*},\inner{x-y}{y^*-x^*}\} < 0\mbox{ or }\inner{x-y}{y^*-x^*}=0
%\]
%that is, $\langle x^*-y^*,x-y\rangle\geq0$. Therefore $T^\mu$ is monotone.
%
Finally, since $T\subset T^\mu\subset T^\rho$, Lemma \ref{pre-max} and pre-maximality of $T$ imply the
pre-maximal pseudomonotonicity of $T^\mu$. %%%revisar!
\end{proof}
\section{On $D$-maximal pseudomonotonicity}
We now address a weaker notion of maximality, defined by Hadjisavvas~\cite{Had03}, which he called \emph{$D$-maximal pseudomonotonicity}. Explicitly, $T:X\tos X^*$ is $D$-maximal pseudomonotone, if $T$ is pseudomonotone and there exists an \emph{equivalent} pseudomonotone operator $S$, that is, $\dom(T)=\dom(S)$, $Z_T=Z_S$ and $\cone(T(x))=\cone(S(x))$, for all $x\in \dom(T)\setminus Z_T$, which has no proper pseudomonotone extension with the same domain.
As expected, a maximal pseudomonotone operator is also $D$-maximal pseudomonotone.
Also in~\cite{Had03}, the author considered the maximal equivalent operator (in the sense of inclusion) of a pseudomonotone operator. Let $T:X\tos X^*$ be a pseudomonotone operator. For $x\in Z_T$, consider
\[
L(T,x)=\{y\in X\::\:\exists y^*\in T(y),\,\inner{x-y}{y^*}\geq 0\},
\]
and the operator $\widehat{T}:X\tos X^*$,
\[
\widehat{T}(x)=
\begin{cases}
N_{L(T,x)}(x),&\text{if }x\in Z_T,\\
\cone_{\circ}(T(x)),&\text{if }x\in \dom(T)\setminus Z_T,\\
\emptyset,&\text{if }x\notin\dom(T).
\end{cases}
\]
%Furthermore,
\begin{proposition}[{\cite[Proposition 2.1]{Had03}}]\label{pro:had}
Let $T:X\tos X^*$ be pseudomonotone. Then
\begin{enumerate}
\item $\widehat{T}$ is pseudomonotone;
\item $T$ is equivalent to $\widehat{T}$;
\item $\widehat{T}$ is the greatest, in the sense of inclusion, between all pseudomonotone operators equivalent to $T$;
\item if $T$ and $S$ are equivalent, then $\widehat{T}=\widehat{S}$.
\end{enumerate}
\end{proposition}
% \begin{example}
% Let us $T:\R\tos\R$ defined by $T(x)=\{x\}$ for all $x\in\R$. Clearly $T$ is a monotone operator and
% \[
% Z_T=\{0\}\mbox{ and }L(T,0)=\{0\}.
% \]
% Therefore, the operator $\widehat{T}:\R\tos\R$ is defined as:
% \[
% \widehat{T}(x)=\left\lbrace\begin{array}{ll}
% \R,&x=0\\
% ]-\infty,0[,&x<0\\
% ]0,+\infty[,&x>0
% \end{array}\right.
% \]
% Clearly $T\subset \widehat{T}=T^\rho$. %So, $\widehat{T}$ is maximal pseudomonotone and his not has
% %closed graph. Moreover, $T$ is $D$-maximal pseudomonotone.
% \end{example}
Note that $\dom(T)=\dom(\widehat{T})$. In addition, since $Z_T\subset Z_{T^{\rho}}$,
when $x\in Z_T$, $L(T,x)=W_T(x)$ and $\widehat{T}(x)=T^{\rho}(x)$. This, along with Proposition~\ref{pro:had}, Proposition \ref{pro:polarp}, item {\it 4}, and Proposition \ref{pro:4.1}, item {\it 2}, implies the following inclusions
\begin{equation}\label{eq:1}
T\subset\widehat{T}\subset(\widehat{T})^\rho\subset T^{\rho}.
\end{equation}
%Moreover, if $T$ is maximal pseudomonotone then
%$T=\widehat{T}=\widehat{T}^\rho=T^\rho$. \\
%\newline
%%
%%D-maximality
%%
The fact that the domain of $\widehat{T}$ is the same as $T$ motivates us to study the pseudomonotone polar of $T$ restricted to $T$'s domain, which we will denote as $T^{\rho}_D=T^{\rho}|_{\dom(T)}$. With this notation, we can rewrite~\eqref{eq:1} and obtain
\begin{equation}\label{eq:2}
T\subset\widehat{T}\subset(\widehat{T})^\rho_D\subset T^{\rho}_D.
\end{equation}
These four operators have the same domain, $\dom(T)$.
We now give a characterization of $D$-maximality in terms of the polar of the operator $\widehat{T}$.
\begin{theorem}\label{teo:dmax}
Let $T:X\tos X^*$ be a pseudomonotone operator. Then $T$ is $D$-maximal pseudomonotone if, and only if, $\widehat{T}=(\widehat{T})^{\rho}_D$.
\end{theorem}
\begin{proof}
Assume that $T$ is $D$-maximal pseudomonotone.
%
From Proposition~\ref{pro:had}, item~{\it 1}, we have $\widehat{T}\subset(\widehat{T})^{\rho}_D$.
On the other hand, given $(x,x^*)\in(\widehat{T})^{\rho}_D$, by Corollary~\ref{123}, $\widehat{T}\cup\{(x,x^*)\}$ is pseudomonotone. Since $T$ is $D$-maximal pseudomonotone, $(x,x^*)\in\widehat{T}$. This proves the ``only if'' part.
For the ``if'' part, note that, by Proposition~\ref{pro:had}, item {\it 3}, it is enough to prove that $\widehat T$ has no proper pseudomonotone extension with the same domain. Otherwise, there would exist $(x,x^*)\in X\times X^*$, with $x\in\dom(T)$, such that $S'=\widehat T\cup\{(x,x^*)\}$ is pseudomonotone. This implies that $(x,x^*)\in(\widehat{T})^{\rho}_D$. Therefore, $(x,x^*)\in\widehat T$ and $T$ is $D$-maximal pseudomonotone.
\end{proof}
%REVISAR
\begin{remark}
It is not difficult to show that
if $\dom(T)=X$ then $T$ is $D$-maximal pseudomonotone if, and only if, $\widehat{T}$ is maximal pseudomonotone. However,
in the case general is false, consider us the multivalued operator of Example \ref{ejem1}, the operator $\widehat{T}$ is defined by
\[
\widehat{T}(x)=\left\lbrace\begin{array}{cc}
]-\infty,0[&,x<0\\
]0,+\infty[&,x>0
\end{array}
\right.
\]
Clearly $T$ is $D$-maximal pseudomonotone but $\widehat{T}$ isnt maximal pseudomonotone.
\end{remark}
Define $T^{\rho\rho}_D=T^{\rho\rho}|_{\dom(T)}$. As a direct consequence of Theorem \ref{teo:dmax} we have the following corollary.
\begin{corollary}
Let $T:X\tos X^*$ be a multivalued operator. If $T$ is $D$-maximal pseudomonotone then $T^{\rho\rho}_D\subset \widehat{T}=(\widehat{T})_D^{\rho\rho}$.
\end{corollary}
From Theorem \ref{teo:dmax} and inclusion \eqref{eq:2} we have the following corollary.
%REVISAR
\begin{corollary}
Let $T:X\tos X^*$ be a $D$-maximal pseudomonotone operator. If $\dom(T)=\dom(T^\rho)$ then
$\widehat{T}$ is maximal pseudomonotone.
\end{corollary}
% Esto es más general y con la misma prueba.
\begin{proposition}\label{pro:domconv}
Let $T:X\tos X^*$ be a pseudomonotone operator. If $\dom(T)$ is convex then $(\widehat{T})^\rho_D=T^{\rho}_D$.
\end{proposition}
\begin{proof}
From~\eqref{eq:2}, we already have the inclusion $(\widehat{T})^\rho_D\subset T^{\rho}_D$.
%
Let $(x,x^*)\in T^{\rho}_D$, then $x\in \dom(T)$ and $(x,x^*)\in T^{\rho}$.
We now prove that $(x,x^*)\in(\widehat{T})^{\rho}$.
Indeed, take $(y,y^*)\in\widehat{T}$, then $(y,y^*)\in T^{\rho}$ and $y\in \dom(T)$.
Since $\dom(T)$ is convex, $z=\dfrac{x+y}{2}\in\dom(T)$ so there exists $z^*\in T(z)$.
Thus, $(z,z^*)$ is pseudomonotonically related to $(x,x^*)$ and $(y,y^*)$ and,
by Proposition~\ref{pro:pseudo-new}, $(x,x^*)$ and $(y,y^*)$ are pseudomonotonically related.
%
Therefore $(x,x^*)\in (\widehat{T})^{\rho}$, so $(\widehat{T})^\rho_D=T^\rho_D$, and the proposition follows.
\end{proof}
From Theorem~\ref{teo:dmax} and Proposition~\ref{pro:domconv}, we recover Lemma 3.2 in~\cite{Had03}.
\begin{corollary}\label{H-lemma}
Let $T:X\tos X^*$ be a pseudomonotone operator. If $\widehat{T}=T^{\rho}_D$ then $T$ is $D$-maximal pseudomonotone. In addition, if $\dom(T)$ is convex then the converse is true.
\end{corollary}
\begin{proof}
From the inclusions in~\eqref{eq:2}, if $\widehat{T}=T^{\rho}_D$ then $\widehat{T}=(\widehat{T})^{\rho}_D$ and, by Theorem~\ref{teo:dmax}, $T$ is $D$-maximal pseudomonotone. When $\dom(T)$ is convex, by Proposition~\ref{pro:domconv}, $T^{\rho}_D=(\widehat{T})^\rho_D$, which coincides with $\widehat{T}$ again from Theorem~\ref{teo:dmax}.
\end{proof}
%\begin{proposition}
%Let $T$ be a pseudomonotone operator. Then the following hold.
%
% If $\widehat{T}=T^{\rho}_D$ then $T$ is $D$-maximal pseudomonotone.
%
%% \item $T^{\rho}_D$ is the union of all the $D$-maximal pseudomonotone operators with the same domain as $T$.
%
%\end{proposition}
%
%\begin{proof}\quad
%\begin{enumerate}
% \item By Proposition~\ref{pro:had}, item 3, it is enough to prove that $\widehat T$ has no proper pseudomonotone extension with the same domain. Otherwise, there would exist $(x,x^*)\in X\times X^*$, with $x\in\dom(T)$, such that $S'=\widehat T\cup\{(x,x^*)\}$ is pseudomonotone. This implies that $(x,x^*)\in(\widehat{T})^{\rho}\subset T^{\rho}$. Therefore $(x,x^*)\in T^{\rho}_D=\widehat T$ and $T$ is $D$-maximal pseudomonotone.
% \item From equation~\eqref{eq:2}, we already have $\widehat{T}\subset T^{\rho}_D$.
% Let $(x,x^*)\in T^{\rho}_D$. If $x\in Z_T$ then $x^*\in T^{\rho}_D(x)=T^{\rho}(x)=\widehat{T}(x)$
% and we are done. Now assume that $x\in\dom(T)\setminus Z_T$ and further assume that $(x,x^*)\notin \widehat{T}$.
%If $\langle y-x,x^*\rangle>0$, then $x\in V_T(y)$. So, by Corollary \ref{cor:trhozeros}, $y\notin Z_T$ and this implies
%$(x,x^*)\sim_\rho(y,y^*)$ for all $y^*\in \cone_\circ(T)(y)$.
%Finally, if $\langle y-x,x^*\rangle=0$ and with out loss generality we assume that $y\in Z_T$, then $x\in L(T,y)$ and this implies
%$(x,x^*)\sim_\rho(y,y^*)$ for all $y^*\in \widehat{T}(y)$. Therefore $\widehat{T}\cup\{(x,x^*)\}$ is pseudomonotone and this is a
%contradiction due to $D-$maximality of $T$.
%
%%creo que podemos poner un lema diciendo para todo (x,x^*)\in T^\rho_D se tiene que
%%$\widehat{T}\cup\{(x,x^*)\}$ is pseudomonotone.
%
%\item It is a direct consequence of Zorn's Lemma that every pseudomonotone operator has a $D$-maximal pseudomonotone extension with the same domain. If $S$ is a $D$-maximal pseudomonotone operator containing $T$ with $\dom(S)=\dom(T)$, then $S\subset T^{\rho}_D$. This prove one inclusion. To prove the remaining inclusion, let $(x,x^*)\in T^{\rho}_D$ and consider $T'=T\cup\{(x,x^*)\}$ which is pseudomonotone. Thus, $T'$ has a $D$-maximal monotone extension $S$, with the same domain. Since $x\in\dom(T)$, $\dom(S)=\dom(T')=\dom(T)$ and $S$ is also a $D$-maximal monotone extension of $T$. The result follows by observing that $(x,x^*)\in T'\subset S$.\qedhere
%\end{enumerate}
%\end{proof}
It is a direct consequence of Zorn's Lemma that every pseudomonotone operator has a $D$-maximal pseudomonotone extension with the same domain. Moreover, it is straightforward to prove that $T^{\rho}_D$ is the union of all $D$-maximal pseudomonotone operators containing $T$.
\begin{example}[{\cite{Had03}}]
Consider $T:\R\tos\R$, $T=\{(0,-1),(1,0)\}$, which is a pseudomonotone operator.
In this case $\widehat{T}=(\{0\}\times \R_{++})\cup(\{1\}\times\R)$.
Note that $\widehat{T}$ does not have a pseudomonotone extension with the same
domain, so $T$ is $D$-maximal pseudomonotone. However, $T^{\rho}_D=(\{0\}\times \R_{+})\cup(\{1\}\times\R)$
is not pseudomonotone. Now consider $S=(\{0\}\times\R_{-})\cup(\{1\}\times\R_+)$, and observe
that $T\subset S$ and $S$ also does not have a pseudomonotone extension with the same domain.
Thus $S$ is $D$-maximal pseudomonotone aswell. Finally, observe that $T^{\rho}_D=\widehat{T}\cup S$.
\end{example}
\begin{remark}
If $T$ is pseudomonotone and $\widehat{T}=T^\rho$ then $T$ is pre-maximal pseudomonotone and $D$-maximal pseudomonotone.
\end{remark}
The condition of pre-maximal pseudomonotonicity can be used to replace convexity of the domain in Corollary~\ref{H-lemma}.
\begin{corollary}
Let $T:X\tos X^*$ be a pre-maximal pseudomonotone operator. Then $T$ is $D$-maximal pseudomonotone if, and only if, $\widehat{T}=T^\rho_D$.
\end{corollary}
\begin{proof}
First note that if $T$ is pre-maximal pseudomonotone then $T^{\rho}_D$ is pseudomonotone. This, together with~\eqref{eq:2}, implies that $T^{\rho}_D$ is a pseudomonotone extension of $\widehat{T}$. Therefore, if $T$ is $D$-maximal pseudomonotone, $\widehat{T}=T^{\rho}_D$. The converse follows again from~\eqref{eq:2} and Theorem~\ref{teo:dmax}.
\end{proof}
From Lemma \ref{pre-max} and inclusion (\ref{eq:1}) we have the following corollary.
\begin{corollary}
Let $T:X\tos X^*$ be a pre-maximal pseudomonotone operator. Then $\widehat{T}$ is pre-maximal pseudomonotone and $(\widehat{T})^\rho=T^\rho$.
\end{corollary}
\section{On variational inequalities}
We now deal with variational inequality problems, in the sense of Stampacchia and Minty. Given a multivalued operator $T:X\tos X^*$ and a set $K\subset X$, the \emph{Stampacchia Variational Inequality Problem} associated to $T$ and $K$ is
\begin{equation}
\text{to find }x\in K,\,\text{such that }\exists\,x^*\in T(x),\,\inner{y-x}{x^*}\geq 0,\,\forall\,y\in K.\tag{SVIP}
\end{equation}
In a dual way, %citar
the \emph{Minty Variational Inequality Problem} associated to $T$ and $K$ is
\begin{equation}
\text{to find }x\in K,\,\text{such that }\inner{x-y}{y^*}\leq 0,\,\forall\,(y,y^*)\in T,\, y\in K.\tag{MVIP}
\end{equation}
The solution sets of the (SVIP) and (MVIP) will be denoted as $S(T,K)$ and $M(T,K)$, respectively.
It is not difficult to prove that $M(T,K)$ is convex and (weakly-)closed, provided that $K$ is convex and (weakly-)closed. Also, when $K=X$,
\[
S(T,X)=Z_T\quad\text{and}\quad M(T,X)=Z_{T^{\rho}}.
\]
We now study the connections between the VIP solution sets and the pseudo and quasi-monotone polars. First, we need the following lemma.
\begin{lemma}\label{lemma1}
Let $T_1,T_2:X\tos X^*$ be multivalued operators and $K\subset X$. If $T_1\subset T_2$ then $S(T_1,K)\subset S(T_2,K)$ and $M(T_2,K)\subset M(T_1,K)$.
\end{lemma}
Noting that $T^\mu\subset T^\rho\subset T^\nu$, by the previous lemma, we conclude
\begin{gather*}
S(T^\mu,K)\subset S(T^\rho,K)\subset S(T^\nu,K)=K,\\
M(T^\nu,K)\subset M(T^\rho,K)\subset M(T^\mu,K).
\end{gather*}
for any $K\subset X$.
The following result says that, in general, there are relationship between solution sets of SVIP and MVIP associated to a multivalued operator and its pseudomonotone polar.
\begin{proposition}\label{S-M}
Let $T:X\tos X^*$ be a multivalued operator. Then, for every $K\subset X$,
\[
S(T^\rho,K)\subset M(T,K)\quad\text{and}\quad S(T,K)\subset M(T^\rho,K).
\]
\end{proposition}
\begin{proof}
Let $K$ be a subset of $X$, take $x\in S(T^\rho,K)$ and assume that $x\notin M(T,K)$. Then there exist $(y,y^*)\in T$, with $y\in K$, and $x^*\in T^{\rho}(x)$, such that $\inner{x-y}{y^*}>0$ and $\inner{y-x}{x^*}\geq 0$. This contradicts the fact that $(x,x^*)\sim_p(y,y^*)$, and we conclude $S(T^\rho,K)\subset M(T,K)$. The remaining inclusion can be proved analogously.
%Then there exists $x^*\in T^\rho(x)$ such that
%$ \langle y-x,x^*\rangle\geq0,\mbox{ for all }y\in K$.
%As $(x,x^*)\sim_\rho (y,y^*)$ for all $(y,y^*)\in T$, we have
%$\langle x-y,y^*\rangle\leq0\mbox{ for all }(y,y^*)\in T,~y\in K$.
%Therefore, $x\in M(T,K).$
%Let $x\in S(T,K)$. Then there exists $x^*\in T(x)$ such that
%$\langle y-x,x^*\rangle\geq0\mbox{ for all }y\in K$.
%Since $(x,x^*)\sim_\rho (y,y^*)$ for all $(y,y^*)\in T^\rho$, we have
%$ \langle x-y,y^*\rangle\leq0,\mbox{ for all }(y,y^*)\in T^\rho,~y\in K$.
%Therefore, $x\in M(T^\rho,K).$
\end{proof}
The inclusions in the previous result are strict in general. For instance, the operator defined in Example~\ref{ejem1} verifies
\[
S(T,\R)=\emptyset~\mbox{ and }~M(T^\rho,\R)=\{0\}.
\]
In addition, the operator defined in the Example \ref{ejem-variational} verifies, for $K=\{1,2\}$,
\[
S(T^\rho,K)=\emptyset~\mbox{ and }~M(T,K)=K.
\]
\newcommand{\subsetv}{\rotatebox[origin=c]{270}{$\subset$}}
Note that Proposition~\ref{S-M} does not ask for any condition on $T$ nor $K$. When $T$ is pseudomonotone (or even monotone), we can obtain a more complete array of inclusions between the VIP solution sets.
\begin{corollary}\label{coro:incl}
Let $T:X\tos X^*$ be a multivalued operator and $K\subset X$.
\begin{enumerate}
\item If $T$ is pseudomonotone,
\begin{equation}\label{eq:incpseudo}
\begin{array}{ccc}
S(T,K)&\subset & S(T^{\rho},K)\\
\subsetv&&\subsetv\\
M(T^{\rho},K)&\subset &M(T,K)\\
\end{array}
\end{equation}
\item If $T$ is monotone,
\begin{equation}\label{eq:incmono}
\begin{array}{ccccc}
S(T,K)&\subset& S(T^{\mu},K)&\subset& S(T^{\rho},K)\\
\subsetv&& &&\subsetv\\
M(T^{\rho},K)&\subset &M(T^{\mu},K)&\subset &M(T,K)\\
\end{array}
\end{equation}
\end{enumerate}
\end{corollary}
We now recall the following well known result due to John~\cite{John2001}, which characterizes the pseudomonotonicity of a multivalued operator, with certain inclusions between the VIP solution sets. We must remark that, although John originally dealt with finite dimensional spaces, his proof can be easily adapted to general Banach spaces.
\begin{theorem}[\cite{John2001},Theorem 2]\label{john-pseudomonotone}
Let $T:X\tos X^*$ be a multivalued operator. Then $T$ is pseudomonotone if, and only if, $S(T,K)\subset M(T,K)$, for all $K\subset \dom(T)$.
\end{theorem}
In a similar way to Theorem \ref{john-pseudomonotone}, we establish a characterization of pseudomonotonicity by comparing the solution sets of the VIPs associated to the operator and its pseudomonotone polar.
\begin{proposition}\label{pseudo-SVIP}
Let $T:X\tos X^*$ be a multivalued operator. Then
the following conditions are equivalent.
\begin{enumerate}
\item $T$ is pseudomonotone.
\item $S(T,K)\subset S(T^\rho,K)$, for all $K\subset X$.
\item $M(T^\rho,K)\subset M(T, K)$, for all $K\subset X$.
\end{enumerate}
\end{proposition}
\begin{proof}
From Lemma~\ref{lemma1}, we obtain that {\it 1} implies {\it 2} and {\it 3}. Moreover, from Theorem~\ref{john-pseudomonotone} and equation~\eqref{eq:incpseudo} in Corollary~\ref{coro:incl}, we conclude that items {\it 2} and {\it 3} imply {\it 1}.
%We now prove that {\it 2} implies {\it 1}. Assume that $T$ is not pseudomonotone, then there exist $(x,x^*),(y,y^*)\in T$ such that $(x,x^*)\not\sim_p (y,y^*)$ so, without loss of generality, we may assume that $\inner{y-x}{x^*}>0$ and $\inner{x-y}{y^*}\geq 0$. Now set $K=\{x,y\}$, so $y\in S(T,K)$. If $y\in S(T^{\rho},K)$ then there exists $z^*\in T(y)$ such that $\inner{x-y}{z^*}\geq 0$. This, together with $\inner{y-x}{x^*}>0$, implies $(y,z^*)\not\sim_p(x,x^*)$, a contradiction. Thus, $S(T,K)\not\subset S(T^\rho,K)$ and {\it 2} does not hold.
The theorem follows.
\end{proof}
From Theorem~\ref{john-pseudomonotone} and Proposition~\ref{pseudo-SVIP}, we can characterize pre-maximal pseudomonotonicity by the inclusions of the solution sets of the (SVIP) into the (MVIP).
\begin{corollary}
Let $T:X\tos X^*$ be a pseudomonotone operator, everywhere defined. Then $T$ is pre-maximal pseudomonotone if, and only if, $S(T^\rho,K)\subset M(T^\rho,K)$, for all $K\subset\dom(T^{\rho})$.
\end{corollary}
The following result gives a sufficient condition for the pre-maximality of a multivalued operator.
\begin{proposition}
Let $T:X\tos X^*$ be a multivalued operator. If $S(T,K)=S(T^\rho,K)$ or $M(T,K)=M(T^\rho,K)$, for all $K\subset X$ then $T$ is pre-maximal pseudomonotone.
\end{proposition}
%\begin{proof}
%The Theorem \ref{pseudo-SVIP} implies that $T$ is pseudomonotone.
%Therefore, the result follows of Proposition \ref{S-M} and again of Theorem \ref{pseudo-SVIP}..
%\end{proof}
%
%%\section{On upper sign-continuity and pseudomonotonicity}
%Recall that a multivalued operator $T:X\tos X^*$ is \emph{upper sign-continuous}~\cite{Had03} at $x\in\dom(T)$ if, for all $v\in X$, the following holds,
%\[
%\left(\forall\,t\in]0,1[,\,\inf_{y^*\in T(x+tv)}\inner{v}{y^*}\geq 0\right) \Longrightarrow \sup_{x^*\in T(x)}\inner{v}{x^*}\geq 0.
%\]
%Moreover, $T$ is \emph{upper sign-continuous} on $B\subset\dom(T)$ if $T$ is upper sign-continuous at every $x\in B$, and $T$ is \emph{upper sign-continuous} if $T$ is upper sign-continuous on its domain.
%
%%An multivalued operator $T:X\tos X^*$ with nonempty values, is called upper sign-continuous on $X$, if for every $x,y\in X$, the following implication
%%holds:
%%\[
%%\left(\forall t\in]0,1[,~\inf_{x_t^*\in T(x_t)}\langle y-x,x_t^*\rangle\geq0\right) ~\Rightarrow~\sup_{x^*\in T(x)}\langle y-x,x^*\rangle\geq0,
%%\]
%%where $x_t=tx+(1-t)y$.\\
%%\newline
%
%%Recall that the \emph{algebraic interior} of a set $A\subset X$ is
%%\[
%%A^i=\{x\in X\::\: \forall\,v\in X,\,\exists\,\delta>0,\,x+tv\in A,\,\forall\,t\in[0,\delta[\}
%%\]
%%
%%The following result establish the relationship between the upper sign-continuity and pseudomonotonicity.
%%\begin{proposition}\label{pseudo-upper-sing}
%%Let $T:X\tos X^*$ be a multivalued operator. If $-T$ is pseudomonotone then $T$ is upper sign-continuous on $\dom(T)^i$.
%%\end{proposition}
%%\begin{proof}
%%By contradiction, let us assume that $T$ is not upper sign-continuous at $x\in\dom(T)^i$, therefore there exists $v\in X$ such that
%%\begin{equation}\label{eq:uscprop}
%%\forall\,t\in]0,1[,\,\inf_{y^*\in T(x+tv)}\inner{v}{y^*}\geq 0
%%\end{equation}
%%and $\sup_{x^*\in T(x)}\inner{v}{x^*}<0$. In particular, $\inner{v}{x^*}<0$, for all $x^*\in T(x)$. Since $x$ belongs to the algebraic interior of $\dom(T)$, there exists $t\in]0,1[$ such that $x+tv\in\dom(T)$. Writing $y=x+tv$, we obtain
%%\[
%%v=\dfrac{1}{t}(y-x).
%%\]
%%so
%%\[
%%\inner{y-x}{-x^*}>0.
%%\]
%%Now, the pseudomonotonicity of $-T$ implies $\inner{x-y}{-y^*}<0$, for any $y^*\in T(y)$, which in turn implies,
%%\[
%%\inner{v}{y^*}<0
%%\]
%%which contradicts~\eqref{eq:uscprop}.
%%\end{proof}
%
%%In the previous proposition, the condition of $x$ being in at the algebraic interior of the domain is necessary. Take for instance $f:[0,1]\to\R$, $f(x)=1$, for all $x\in[0,1]$. It is easy to see that $-f$ is pseudomonotone, however, $f$ is not upper sign-continuous at $0$.
%
%%\begin{corollary}
%%Let $T:X\tos X^*$ be a multivalued operator such that $-T$ is pseudomonotone. Then $T$ is upper sign-continuous.
%%\end{corollary}
%
%\begin{proposition}[{\cite{Aussel2013-3}, Lemma 3.1, item (ii)}]\label{AC-upper}
%Let $T:X\tos X^*$ be a multivalued operator such that $T(x)$ is a $w^*$-compact and convex set, for all $x\in X$, and let $K\subset \dom(T)$ be convex. If $T$ is upper sign-continuous then $M(T,K)\subset S(T,K)$.
%\end{proposition}
%
%
%\begin{proposition}\label{upper-pseudo}
%Let $T:X\tos X^*$ be a multivalued operator such that $T(x)$ is a non-empty, $w^*$-compact and convex set, for all $x\in X$. If $T$ is upper sign-continuous
%then $T^\rho$ is pseudomonotone.
%\end{proposition}
%\begin{proof}
%The upper sign-continuity of $T$ and Proposition \ref{AC-upper} imply $M(T,K)\subset S(T,K)$, for all $K$ convex subset of $X$.
%By the Proposition \ref{S-M}, we have $S(T^\rho,K)\subset M(T^\rho,K)$.
%The result follows from Theorem \ref{john-pseudomonotone}.
%\end{proof}
%\newcommand{\equalv}{\rotatebox[origin=c]{270}{$=$}}
%
%From Propositions~\ref{AC-upper} and \ref{upper-pseudo}, together with inclusions~\eqref{eq:incpseudo} and \eqref{eq:incmono}, we obtain the following corollary.
%\begin{corollary}\label{consequence-1}
%Let $T:X\tos X^*$ be an upper sign continuous operator such that $T(x)$ is non-empty, $w^*$-compact and convex, for all $x\in X$.
%\begin{enumerate}
%\item If $T$ is pseudomonotone then $T$ is pre-maximal pseudomonotone and, in addition, for all $K\subset X$ convex,
%\[
%\begin{array}{ccc}
%S(T,K)&= & S(T^{\rho},K)\\
%\equalv&&\equalv\\
%M(T^{\rho},K)&= &M(T,K).\\
%\end{array}
%\]
%\item If $T$ is monotone then, for all $K\subset X$ convex,
%\[
%\begin{array}{ccccc}
%S(T,K)&=& S(T^{\mu},K)&=& S(T^{\rho},K)\\
%\equalv&& \equalv&&\equalv\\
%M(T^{\rho},K)&= &M(T^{\mu},K)&= &M(T,K)\\
%\end{array}
%\]
%\end{enumerate}
%\end{corollary}
%%As a direct consequence of Proposition \ref{pseudo-upper-sing} and Corollary \ref{consequence-1} we have the following corollary.
%%\begin{corollary}
%%Let $T:X\tos X^*$ be a multivalued operator such that $T(x)$ is non-empty, $w^*-$compact and convex set for all $x\in X$. If $T$
%%and $-T$ are pseudomonotone then $T$ is pre-maximal pseudomonotone
%%\end{corollary}
%%A natural question is: $-T^\rho=(-T)^\rho$? The answer of this question is given by the following example.
%%\begin{example}
%%Let $T:\R\tos\R$ be a multivalued defined as $T(0)=1$ and $T(x)=\emptyset$ for all $x\neq0$.
%%Clearly $(1,0)\notin T^\rho$, but $(1,0)\in (-T)^\rho$.
%%\end{example}
%%
%%Another consequence of Proposition \ref{upper-pseudo} is the following result.
%%
%%\begin{corollary}
%%Let $T:X\tos X^*$ be a multivalued operator such that $\dom(T)$ is open. If $T$ is maximal monotone then
%%$T$ is pre-maximal pseudomonotone.
%%\end{corollary}
\begin{thebibliography}{10}
\bibitem{AusCorLass94}
D.~Aussel, J.-N. Corvellec, and M.~Lassonde.
\newblock {Subdifferential characterization of quasiconvexity and convexity}.
\newblock {\em J. Convex Anal.}, 1(2):195--201 (1995), 1994.
\bibitem{Aussel2013-3}
D.~Aussel and J.~Cotrina.
\newblock {Stability of Quasimonotone Variational Inequality Under
Sign-Continuity}.
\newblock {\em Journal of Optimization Theory and Applications},
158(3):653--667, 2013.
\bibitem{Aussel2013-4}
D.~Aussel and A.~Eberhard.
\newblock {Maximal quasimonotonicity and dense single-directional properties of
quasimonotone operators}.
\newblock {\em Mathematical Programming}, 139(1):27--54, 2013.
\bibitem{AuHad2005}
D.~Aussel and N.~Hadjisavvas.
\newblock {Adjusted Sublevel Sets, Normal Operator, and Quasi-convex
Programming}.
\newblock {\em SIAM Journal on Optimization}, 16(2):358--367, 2005.
\bibitem{BueCot16}
O.~{Bueno} and J.~{Cotrina}.
\newblock {On maximality of quasimonotone operators}.
\newblock {\em ArXiv e-prints}, Sept. 2016.
\bibitem{Crouzeix1997}
J.-P. Crouzeix.
\newblock Pseudomonotone variational inequality problems: Existence of
solutions.
\newblock {\em Mathematical Programming}, 78(3):305--314, 1997.
\bibitem{MR2577322}
J.-P. Crouzeix, A.~Eberhard, and D.~Ralph.
\newblock {A geometrical insight on pseudoconvexity and pseudomonotonicity}.
\newblock {\em Math. Program.}, 123(1, Ser. B):61--83, 2010.
\bibitem{Crouzeix2011}
J.-P. Crouzeix, A.~Keraghel, and N.~Rahmani.
\newblock Integration of pseudomonotone maps and the revealed preference
problem.
\newblock {\em Optimization}, 60(7):783--800, 2011.
\bibitem{Eberhard07}
A.~C. Eberhard and J.-P. Crouzeix.
\newblock Existence of closed graph, maximal, cyclic pseudo-monotone relations
and revealed preference theory.
\newblock {\em Journal of Industrial and Management Optimization},
3(2):233--255, 2007.
\bibitem{Had03}
N.~Hadjisavvas.
\newblock {Continuity and Maximality Properties of Pseudomonotone Operators}.
\newblock {\em J. Convex Anal.}, 10(2):459--469, 2003.
\bibitem{HadjSchai06}
N.~Hadjisavvas and S.~Schaible.
\newblock On a generalization of paramonotone maps and its application to
solving the stampacchia variational inequality.
\newblock {\em Optimization}, 55(5-6):593--604, 2006.
\bibitem{Hadjisavvas2012}
N.~Hadjisavvas, S.~Schaible, and N.-C. Wong.
\newblock Pseudomonotone operators: A survey of the theory and its
applications.
\newblock {\em Journal of Optimization Theory and Applications}, 152(1):1--20,
2012.
\bibitem{John2001}
R.~John.
\newblock {A Note on Minty Variational Inequalities and Generalized
Monotonicity}.
\newblock In N.~Hadjisavvas, J.~E. Mart{\'i}nez-Legaz, and J.-P. Penot,
editors, {\em {Generalized Convexity and Generalized Monotonicity:
Proceedings of the 6th International Symposium on Generalized
Convexity/Monotonicity, Samos, September 1999}}, pages 240--246, Berlin,
Heidelberg, 2001. Springer Berlin Heidelberg.
\bibitem{Karamardian1976}
S.~Karamardian.
\newblock Complementarity problems over cones with monotone and pseudomonotone
maps.
\newblock {\em Journal of Optimization Theory and Applications},
18(4):445--454, 1976.
\bibitem{BSML}
J.~E. Mart{\'i}nez-Legaz and B.~F. Svaiter.
\newblock {Monotone operators representable by l.s.c.\ convex functions}.
\newblock {\em Set-Valued Anal.}, 13(1):21--46, 2005.
\end{thebibliography}
\end{document}