Space of ergodic measures is $G_\delta$ in space of invariant measures

103 Views Asked by At

I'm reading an old paper by Varadarajan (1963) titled "Groups of automorphisms of Borel spaces" and I'm trying to fulfill the details. There are many questions I thought about for quite a long time with no success. For the moment, I confine myself to the first part of the proof of Theorem 4.2 (possibly, the remaining one would be treated in another question).

Notation and Terminology: Suppose $G$ lcsc (=locally compact second countable) group. By measure we mean a Borel probability measure. A measure on a Borel space $X$ is said to be invariant if $\mu(A)=\mu(gA)$ for all borel subsets $A$ and $g\in G$ and we write $\mu\in\mathscr{I}$. It is ergodic if all $G$-invariant Borel subsets are either null or conull. The set of ergodic measures is denoted by $\mathscr{E}$.

Q: Suppose $Y$ is a compact metric $G$-space ($G$ acts continuously on $Y$), so that the space $\mathscr{I}_Y$ of $G$-invariant measures on $Y$ is compact wrt the weak-topology (the one which makes continuous the functions $\mu\mapsto \int_Y g\,d\mu$ forall bounded continuous real-valyued functions $g\in C_b(Y)$). Then, referring to a paper by Choquet titled "Existence et unicité des représentations intégrales", he claims $\mathscr{E}_Y$ is a $G_\delta$ subset. Unfortunately, I have not found such a paper, so I would like to know how to prove that $\mathscr{E}_Y$ is $G_\delta$ (any useful reference on the subject is appreciated, even better if a book).

Edit: I have made the notion of ergodicity more explicit.