Measure-valued continuous curves and processes in total variation norm (Q429273)

This paper treats the equivalent characterization of continuous curves in the sense of total variation norm. As a matter of fact, when \((S, {\mathcal B}(S))\) is a measurable space and \({\mathcal M}_a(S)\) denotes the space of all purely atomic measures on \(S\), then the following staments are equivalent for a curve \(\mu : [0, \infty)\to{\mathcal M}_a(S)\), namely, \((\mu_t)_{t \geq 0}\) is continuous in total variation norm if and only if (i) for each \(x \in S\), \(t \mapsto\mu_t( \{ x \})\) (and \(t \mapsto\mu_t(S)\)) is continuous; or (ii) for a measurable set \(B \subset S\), \(t \mapsto\mu_t(B)\) is continuous; or (iii) for a measurable set \(B \subset S\), the curve: \(t \mapsto\sup_{x \in B} \mu_t( \{ x \} )\) (and \(t \mapsto\mu_t(S)\)) is continuous. Based upon the above result, the author gives a sufficient condition for a purely atomic finite measure-valued process to possess a version of continuous sample paths in the variation norm. In fact, suppose that an \({\mathcal M}_a(S)\)-valued process \(( X_t)\) is continuous in weak topology, and that there exist a dense subset \(\{ x_i \} \subset S\) and a dense subset \(\{ r_j \} \subset [0, \infty)\) such that, for all \( t \geq 0\), \[ X_t \left( \bigcup_{i,j} \{ y \in S: \,\, d(x_i, y ) = r_j \}\right ) = 0, \quad \text{a.s.} \tag{1} \] Let \(( A_j^n)_{j, n \geq 1}\) be a collection of disjoint Borel subsets of \(S\). Theorem (cf. Theorem 3.3 of the paper). If the inequality \[ \sum_{j=1}^{\infty} \operatorname{E} \left| X_t ( A_j^n \cap B_{kl}) - X_s( A_j^n \cap B_{kl} ) \right|^{\alpha} \leqslant C_L | t - s |^{1 + \beta}, \quad ( \exists \alpha, \beta, C_L \geq 0) \tag{2} \] holds for \(L \geq 1\), for all \(n, k, l \in {\mathbb N}\), \(t,s \in [0, L]\), and \(B_{kl}\) \(=\) \(\{ y \in S\) : \(d(x_k, y ) < r_l \}\), then \(( X_t )\) admits a continuous version in total variation norm. Moreover, as an application of the obtained criterion, the sample path property of finite measure-valued diffusion with immigrations is studied as well. The model of measure-valued diffusion was first constructed and its explicit representation was given by \textit{T. Shiga} [J. Math. Kyoto Univ. 30, No. 2, 245--279, (1990; Zbl 0751.60044)]. For other related works, see, e.g., [\textit{J.-H. Shao}, Electron. J. Probab. 16, 271--292 (2011; Zbl 1228.60089); Bull. Sci. Math. 135, No. 4, 383--394 (2011; Zbl 1221.60118)].