Let's first start with the example from where this question arose. I consider the metric space $\mathbb{R}$ and the Lebesgue-measure $\lambda$ on $\mathbb{R}$ as well as the transitive group action $+\colon (\mathbb{R},+)\times \mathbb{R}\to\mathbb{R}$, for which the measure and the metric are both invariant. When integrating over a set $M$ with respect to $\lambda$ I can fix a point $x\in\mathbb{R}$ (the metric space rather than the group) and, since the action is transitive, integrate over a set $M'$ with respect to some measure $\mu$ on the additive group $(\mathbb{R},+)$, such that $M'+x=M$ and $\mu(M')=\lambda(M)$. This is possible as well as completely independent of the choice of $x$ by choosing $\mu$ to be the Lebesgue-measure on $(\mathbb{R},+)$.

Now I want to generalise this statement to arbitrary metric spaces and groups. I have a metric space $X$ with a radon-measure $\mu$ as well as a transitive group action from a topological group $\cdot\colon G\times X\to X$ which leaves the metric and the measure invariant. Analogous to $\mathbb{R}$ I chose another condition that for any $\varepsilon>0$ there exists a neighbourhood $U$ of $1\in G$ such that the metric is left small when translating with elements from $U$, i.e. $\forall x\in X,g\in U\colon d(gx,x)<\varepsilon$. This so far were reasonable assumptions since they give me the required continuity of the group action. Trying to find a suitable measure on $G$ just like above I came to 3 distinct conditions, that could give me what I require.

  1. Assume that $G$ is commutative. This will make sure that $\mu(Ux)=\mu(Uy)$ for every $x,y\in X,U\subseteq G$ since $y=gx$ for some $g$ since I can simply commute the $g$ to the front and disregard it due to invariance.
  2. Assume that the action is free, i.e. $G$ and $X$ are basically the same since for any $x\in X$ the function $f\colon G\to X,g\mapsto gx$ would be a bijection, giving me a very natural choice for the measure on $G$.
  3. Assume that not only above condition, $\forall\varepsilon>0 \exists U \text{ negihbourhood of }1\in G\forall x\in X,g\in U\colon d(gx,x)<\varepsilon$ but even more: $\forall\varepsilon>0 \exists U \text{ negihbourhood of }1\in G\forall x\in X\colon g\in U \iff d(gx,x)<\varepsilon$, which would also solve some problems.

The questions now is: Which of these 3 conditions is most natural/weakest? Which would you go for in the most general case? Do I even need any of them or would this work with the above statements, without 1,2 or 3? Thanks for any help!


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.