Uniform integrability and local convexity in \(\mathbb L^0\) (Q2452472)

Let \(({\mathcal W},{\mathcal F},P)\) be a probability space, and let \(L^0\) denote the space of equivalence classes of real-valued random variables over \(({\mathcal W},P)\) with the metric topology. If \(Q\) is a probability in \(({\mathcal W},{\mathcal F})\), then \(Q\ll P\) denotes that \(Q\) is absolutely continuous with respect to \(P\), and \(Q\sim P\) denotes that \(Q\) is equivalent to \(P\). The expectation of \(X\in L^{0+}=\{f\in L^0: f\geq 0\}\) under the probability \(Q\) is denoted on \(E_Q[X]\), and the metric \(\rho\) on \(L^0\) is defined by \(\rho(X,Y)= E_P[1\wedge|X-Y|]\). The set \(\zeta\subseteq L^0\) is said to be convex if \[ \{\alpha X+(1-\alpha)Y\text{ for }X,\;Y\text{ in }\zeta,\;0<\alpha< 1\}\subseteq\zeta. \] The results of the main theorems of this paper include (1) if \(\zeta\subseteq L^0\), and the non-increasing sequence \(\{O^\zeta_n,\, n= 1,2,3,\dots\}\) is defined by \(O^\zeta_n= \text{sc}\{(|X|- n)_+: X\in\zeta\}\), where \[ \text{sc}\,\zeta= \{Y\in L^{0+}: Y\leq Z\text{ for some }Z\in\text{conv\,}\zeta\}, \] then the statements (i) \(\bigcap_{n\geq 1} Q^\zeta_n= \{0\}\); (ii) \(\bigcap_{n\geq 1} \overline O^\zeta_n=\{0\}\); (iii) there is a probability \(Q\sim P\) such that \(\zeta\) is uniformly \(Q\)-integrable, are equivalent; (2) if \(\zeta\in L^{0+}\) is a convex, solid and closed set, then a number of statements including the existence of a convex neighbourhood of \(0\in\zeta\), and the identification of the \(L^0\)-topology on \(\zeta\) with the \(L^1(Q)\)-topology on \(\zeta\) are equivalent.