Continuity and equicontinuity of semigroups on norming dual pairs (Q2655459)

The author studies semigroups on normal dual pairs with respect to topologies defined in terms of the duality. The focus is on the question whether the semigroup \({\mathbf T}= (T_t)_{t\geq 0}\) implies the (local/quasi) equicontinuity of the family \(\{T_t: t\geq 0\}\). The general results are applied to semigroups on the norming dual pair \((C_b(E),{\mathcal M}_0(E))\), where \(E\) is a completely regular Hausdorff space, \(C_b(E)\) denotes the space of bounded continuous functions on \(E\) and \({\mathcal M}_0(E)\) the space of bounded Radon measures on \(E\). The topology of interest is the strict topology \(\beta_0\), defined as follows: Let \({\mathcal F}_0(E)\) be the space of all bounded functions on \(E\) vanishing at infinity, i.e., given \(f\in{\mathcal F}_0(E)\) and \(\varepsilon> 0\), there is a compact set \(K\subset E\) such that \(|f(x)|<\varepsilon\) for all \(x\not\in K\). The strict topology \(\beta_0\) on \(C_b(E)\) is the locally convex topology generated by the set of seminorms \(\{p_\phi: \phi\in{\mathcal F}_0(E)\}\), where \(p_\phi(f):= \|\phi f\|_\infty\). Suppose \(E\) is a \(T'\)-space, i.e., every weakly compact set of positive Radon measure is tight. (Examples for which this assumption is satisfied are complete metric spaces and locally compact spaces.) Let \({\mathbf T}\) be an integrable semigroup on \((C_b(E),{\mathcal M}_0(E))\). If there exists a measure \(\mu\) on the Borel algebra over \(E\) which is not Radon, assume in addition that \({\mathbf T}\) is positive. Then \(\beta_0\)-continuity of \({\mathbf T}\) implies quasi-\(\beta_0\)-equicontinuity, i.e., that for some \(a\in\mathbb{R}\) the family \(\{e^{-at}T_t: t\in 0\}\) is \(\beta_0\)-equicontinuous. If \(E\) is metric or locally compact, then also the reverse implication holds. Several characterizations of \(\beta_0\)-continuous semigroups on \(C_b(E)\) are given. Of particular interest are transition sernigroups of Banach space valued Markov processes: Let \((\Omega,{\mathcal F},\mathbb{P})\) be a probability space, \((E,\rho)\) a complete metric space, and \(X: [0,\infty)\times E\to L^0(\Omega,E)\) a continuous map, where \(L^0(\Omega,E)\) is the space of all strongly measurable maps from \(\Omega\) to \(E\) (modulo equality \(\mathbb{P}\)-almost everywhere), equipped with the topology of convergence in measure. Define \(T_tf(x):= \mathbb{E} f(X(t,x))\) for \(f\in C_b(E)\). Then \((T_t)_{t\geq 0}\) is locally \(\beta_0\)-equicontinuous, i.e., for every \(t_0> 0\) the set \(\{T_t:0\leq t\leq t_0\}\) is a \(\beta_0\)-equicontinuous family of operators on \((C_b(E),{\mathcal M}_0(E))\). If \((T_t)_{t\geq0}\) is a semigroup, then it is \(\beta_0\)-ocontinuous and quasi-\(\beta_0\)-equicontinuous.