# Local structure of the quotient of a Lie group action

Suppose $M$ is a smooth manifold and a compact Lie group $G$ acts freely on $M$, then it is well known that $M/G$ has a manifold structure.

Are there any results for the general case? (a) If the action is not free, what can we say about the local structure of the quotient? Can we define a stratification on the space? How regular are the strata? (b) Moreover, what can we say if $M$ and $G$ are Hilbert manifolds?

Thanks a lot!

• My favorite reference for the finite-dimensional manifold case is Bredon's "Introduction to compact transformation groups". The key word is slice theorem. When $G$ or $M$ is infinite dimensional it matters very much what they are. Sometimes the slice theorem is known and sometimes it it is false. Searching on "slice theorem" "infinite dimensional" will get you started. – Igor Belegradek Jan 3 '15 at 2:17
• @IgorBelegradek Thanks a lot! This is exactly what I'm looking for. – Boyu Zhang Jan 3 '15 at 3:23

Suppose one relaxes the condition that the action is free, replacing it with the condition that every point in $M$ has a finite $G$-stabilizer. In this case, the topological quotient $M/G$ carries a natural orbifold structure. The local model is therefore the quotient of $\mathbb{R}^n$ by a finite group acting linearly.
In general, $M/G$ is a stratified space, and the stratification is indexed by conjugacy classes of $G$-stabilizers.