Dissipativity and positive off-diagonal property of operators on ordered Banach spaces

It is known that on a Banach space $X$, the positivity and the contractivity of a semigroup $(T(t))_{t\geq 0}$ can be characterized by means of dissipativity of its generator $A$ with respect to some appropriate sublinear functions, see [2, Theorem 2.6]. This property is also proved in the case of $X$ being an ordered Banach space with normal positive cone and with respect to a canonical half-norm on $X$, see [5, Proposition 7.13]. In Banach lattices, moreover, the positivity of $(T(t))_{t\geq 0}$ can be characterized by its generator $A$ which satisfies the “positive off-diagonal” property (it is called “positive minimum principle” in [3, Theorem 1.11]). This is also studied in an ordered Banach space with a positive cone with nonempty interior, [5, Theorem 7.27]. For more positive off-diagonal properties of operators on ordered normed spaces, we refer the reader to [4, 6, 7, 10].

The goal of this paper is to investigate the positivity and contractivity of semigroups through the dissipativity of generators with respect to corresponding sublinear functions on ordered vector spaces. So the choice of sublinear functions on ordered vector spaces is crucial. In Section 3, we give two different ways of defining the sublinear functions on ordered vector spaces. One is given by a regular norm and the other one is obtained through a dual function. The latter one turns out to be more effective in studying the positivity and contractivity of semigroups on ordered Banach spaces in Section 4, these are also our main results. Moreover, in Section 5 we will show a representation for positive functions on Archimedean pre-Riesz spaces, which will be used to study the positive off-diagonal property of operators on the pre-Riesz space $C^{1}[0,1]$.

In this section, we will collect some basic terminology, which we mainly refer to [1, 2, 9]. Let $X$ be a real vector space, and $K\subseteq X$. $K$ is said to be a cone in $X$ if $x,y\in K$, $\lambda,\mu\in\mathbb{R}^{+}$ implies $\lambda x+\mu y\in K$ and $K\cap(-K)=\{0\}$. A partial order in $X$ is induced by $x\leq y$ whenever $y-x\in K$, we then say $(X,K)$ is a (partially) ordered vector space. $(X,K,\|\cdot\|)$ is called an ordered Banach space whenever it is norm complete. The space of all bounded linear operators on $X$ is denoted by $L(X)$, the positive operators in $L(X)$ are $L(X)^{+}:=\{T\in L(X):Tx\geq 0,\ \forall 0\leq x\in X\}$. Then $(L(X),L(X)^{+})$ is an ordered vector space. The domain of an operator $T$ on $X$ is denoted by $\mathcal{D}(T)$. A one-parameter semigroup of operators is written usually as $(T(t))_{t\geq 0}$, and it is called strongly continuous if the map $t\mapsto T(t)$ is continuous for the strong topology on $L(X)$.

Let $(X,K,\|\cdot\|)$ be an ordered normed vector space. A function $p$ on $X$ is called sublinear if $p(x+y)\leq p(x)+p(y)$ for all $x,y\in X$ and $p(\lambda x)=\lambda p(x)$ for all $x\in X$, $\lambda\geq 0$. It is clear that $p(0)=0$. If a sublinear function $p$ on $X$ satisfies $p(x)+p(-x)>0$ whenever $x\neq 0$ in $X$, then $p$ is called a half-norm. If there exists a constant $M>0$ such that $p(x)+p(-x)\geq M\|x\|$ for all $x\in X$, then $p$ is called a strict half-norm. Let $x\in X$, define

then $\psi$ is a half-norm in $X$. This half-norm $\psi$ is called the canonical half-norm. A seminorm $p$ on $X$ is called regular if for all $x\in X$ one has

A regular seminorm on an ordered vector space will be denoted by $\|\cdot\|_{r}$. If $(X,K)$ is an ordered vector space equipped with a seminorm $p$, then for $x\in X$, $\|x\|_{r}:=\inf\{p(y)\colon y\in X,-y\leq x\leq y\}$ is called the regularization of the original seminorm $p$. Moreover, $\|\cdot\|_{r}$ is a seminorm, and $\|\cdot\|_{r}\leq p(\cdot)$ on $K$. A seminorm $p$ on $X$ is called monotone if for every $x,y\in X$ such that $0\leq x\leq y$ one has that $p(x)\leq p(y)$.

If $(M,\|\cdot\|)$ is any normed vector space, we denote by $M^{\prime}$ the continuous linear functionals on $M$.

Let $(X,K,\|\cdot\|)$ be an ordered normed vector space, $p$ be a sublinear function on $X$. A bounded operator $T$ on $X$ is called $p$-contractive if $p(Tx)\leq p(x)$ for all $x\in X$. Similarly, a semigroup $(T(t))_{t\geq 0}$ is called $p$-contractive if $T(t)$ is $p$-contractive for all $t\geq 0$. The subdifferential of $p$ at point $x\in X$, denoted by $dp(x)$, is defined as

Let $(X,\|\cdot\|)$ be a Banach lattice, $p$ be a canonical half-norm on $X$. We have that $p(x)=\|x^{+}\|$ for all $x\in X$. In fact, let $x\in X$, $p(x)\leq\|x^{+}\|$ is obvious. Since $|x+y|\geq(x+y)^{+}\geq x^{+}$ for any $0\leq y\in X$, we have that $\|x+y\|\geq\|x^{+}\|$. So $p(x)\geq\|x^{+}\|$. To make a difference from the above, we will use distinctive notations. Let $N^{+}\colon X\rightarrow\mathbb{R}$ be the function given by $N^{+}(x)=\|x^{+}\|$. Then the subdifferential of $N^{+}$ at point $x\in X$ is

An operator $T$ on $X$ is called (strictly) dispersive if $T$ is (strictly) $N^{+}$-dissipative.

We continue by some notations of pre-Riesz theory. An ordered vector space $(X,K)$ is called directed if for every $x,y\in X$ there exists $z\in X$ such that $z\geq x,z\geq y$, and is called Archimedean if for every $x,y\in X$ with $nx\leq y$ for all $n\in\mathbb{N}$ one has $x\leq 0$. We say that $X$ is a pre-Riesz space if for every $x,y,z\in X$ the inclusion $\{x+y,x+z\}^{\text{u}}\subseteq\{y,z\}^{\text{u}}$ implies $x\in K$, where $\{x,y\}^{\text{u}}=\{z\in X\colon z\geq x,z\geq y\}$. Every directed Archimedean ordered vector space is pre-Riesz [9]. A linear subspace $D$ of $X$ is called order dense in $X$ if

Every pre-Riesz space admits a Riesz completion, and it is an order dense subspace of the Riesz completion [9]. A linear map $i\colon X\rightarrow Y$ between two ordered vector spaces is called bipositive if for every $x\in X$ one has $x\geq 0$ if and only if $i(x)\geq 0$.

In this section, we will introduce two different seminorms on ordered vector spaces. One is induced by the regular norms, and the other one involves the order structure which will be used in the following section.

We continue with a different approach of sublinear functions on ordered vector spaces, which turns out to be more useful in dealing with the contractivity and positivity of semigroups.

Let $\phi\in X^{\prime}$, define $p_{\phi}$ on $X$ by

We note that the continuous differential function space $X=C^{1}[0,1]$ and the Sobolev space $X=W^{n,p}$ are (pointwise) ordered vector spaces but not lattices. They are norm complete with respect to $\|\cdot\|_{\infty}$ and $\|\cdot\|_{W^{n,p}}$ respectively. However, these norms are not monotone, but one could still define a sublinear function $p_{\phi}$ by (3.2). Next, we will study the contractivity of $(T(t))_{t\geq 0}$ with respect to $p_{\phi}$ given by (3.2) on ordered Banach space in the following section.

In this section, we will study the relation between the $p_{\phi}$-contractivity of $(T(t))_{t\geq 0}$ and the strictly $p_{\phi}$-dissipativity of its generator $A$ for $X$ being an ordered Banach space, the sublinear function $p_{\phi}$ on $X$ is defined by (3.2). Firstly, we will give a sufficient condition under which $(T(t))_{t\geq 0}$ is contractive with respect to $p_{\phi}$ on an ordered Banach space.

We will use $X$ to denote the ordered Banach space in the following of this section, and $p_{\phi}$ on $X$ is defined by (3.2) with respect to $\phi\in X^{\prime}$.

Observe that $(I-\lambda A)^{-1}=\frac{1}{\lambda}\left(\frac{1}{\lambda}I-A\right)^{-1}=\frac{1}{\lambda}R\left(\frac{1}{\lambda},A\right)$. If a strongly continuous semigroup $(T(t))_{t\geq 0}$ is generated by the operator $A$, then for $x$ in $X$, one has

Hence, by Theorem 4.1, if $A$ is $p_{\phi}$-dissipative, then $T(t)$ is $p_{\phi}$-contractive for every $t\geq 0$. As a consequence we have the following corollary.

Next, we will study the positivity of strongly continuous semigroup $(T(t))_{t\geq 0}$ on an ordered Banach space. The generator $A$ of $(T(t))_{t\geq 0}$ is required to be $p_{\phi}$-dissipative for all $\phi$ in a total subset of an ordered Banach space.

The following example shows that the intersection of a subdifferential set and a total subset on an ordered Banach space $X=C^{1}[0,1]$ is nonempty.

In this section, we will introduce the positive off-diagonal property especially on pre-Riesz spaces, in particular $C^{1}(\Omega)$ with $\Omega\subseteq\mathbb{R}^{n}$ open. Explicitly, we investigate a representation theorem for positive linear functionals on Archimedean pre-Riesz spaces, which is also interesting independently.

The motivation of the positive off-diagonal property comes from matrix theory, where the off-diagonal elements of the matrix $A=(a_{ij})$ are positive, i.e., $a_{ij}\geq 0$ for all $i\neq j$. It is shown in [5, Lemma 7.23] that on an ordered Banach space $X$ with an order unit $u$ such that $u\in\mathcal{D}(A)$, the operator $A$ has the positive off-diagonal property and $Au\leq 0$ if and only if $A$ is $\Psi_{u}$-dissipative, where $\Psi_{u}$ is the order unit function, i.e. $\Psi_{u}(x)=\inf\{\lambda\geq 0\colon x\leq\lambda u\}$, $x\in X$. However, in general, the properties that $A$ has the positive off-diagonal property and $A$ is $p$-dissipative for a given $p$ on $X$ do not imply each other, as the following example shows.

Next, we consider a representation theorem in pre-Riesz spaces.

We give an example to illustrate that the positive off-diagonal property of $A$ can be generalized to a special kind of ordered vector space, in particular to the pre-Riesz space $C^{1}[0,1]$.