On the nonequivalence of three definitions of continuous directions for vector measures (Q1907414)

Let \(\mu\) be a measure defined on the \(\sigma\)-algebra of all Borel subsets of a real linear topological space \(X\) and taking values in a Banach space \(Y\). By the shift of the measure \(\mu\) by an element \(h\in X\) we mean the measure \(\mu_h\) defined on the measurable sets \(E\) by the formula \(\mu_h(E)= \mu(E+ h)\). The semivariation of \(\mu\) with respect to all \(X\) is denoted by \(|\mu|\), while the variation by \(\text{Var } \mu\). Definitions. A vector measure \(\mu\) is said to be: 1) \(h\)-continuous (continuous in the direction \(h\)) with respect to variation if \(\text{Var } \mu< \infty\) and \(\lim_{t\to 0}\text{ Var}(\mu_{th}- \mu)= 0\), 2) \(h\)-continuous with respect to semivariation if \(\lim_{t\to 0} |\mu_{th}- \mu|= 0\), 3) \(h\)-continuous with respect to convergence on each measurable set if for any measurable set \(E\) one has \(\lim_{t\to 0} (\mu_{th}- \mu)(E)= 0\). It is obvious that 1)\(\Rightarrow\)2)\(\Rightarrow\)3). In this paper, the author establishes the fact that, in general, for vector measures the definitions of continuous directions for various forms of convergence (relative to variation, relative to semivariation, and relative to convergence on each measurable set) are not equivalent.