Symmetry groups and translation invariant representations of Markov processes (Q808531): Difference between revisions

In two previous papers [author and \textit{J. Mitro}, ibid. 18, No.2, 655- 668 (1990; Zbl 0714.60060), and author, Math. Z. 204, No.1, 1-11 (1990; Zbl 0698.60065)] the author introduced and discussed a pair of ``symmetry groups'' (called G and Sym) associated to a Markov process, or more precisely, to its potential theory. These are collections of bimeasurable bijections which preserve the collection of excessive functions, or which are finely continuous and preserve the collection of excessive measures, respectively. These groups (or subgroups thereof) are used to introduce new algebraic and topological structures on the state space E of the process. In particular, if J is any transitive subgroup of G or Sym, E inherits either the group structure of J, or the structure of a certain collection of cosets of J. Even in the latter case, it is possible to define ``increments'' of the process and to decide whether a process is ``J-translation invariant''. A J-translation invariant process has ``stationary independent increments'' in a certain sense, and might be called a ``J-Lévy process''. For the case of J a subgroup of G, the author describes several sets of conditions under which a transient Borel right process with J transitive can be rendered J-translation invariant via a time change. Some of these conditions include the assumptions that E inherits the group structure of J, with the maps (x,y)\(\to xy\) and \((x,y)\to x^{-1}y\) jointly measurable, and that there exists a \(\sigma\)-finite left quasi-invariant measure on (E,\({\mathcal E})\). (In this case, there exists a topology - the Mackey-Weil topology - making E a locally compact second countable metric group.) Several conditions which guarantee the existence of a left quasi-invariant measure are discussed.