2
$\begingroup$

Suppose that $G$ is a finite group, and let $\# G$ denote the number of elements in $G$. Let $m$ be a point of $M$, and consider the orbit of $m$ under $G$, which contains $\#(G\cdot m)$ elements. If $n$ is an element of this orbit, then $n=am$ for some $a$ in $G$. If $n=bm$ as well, then $a^{-1}b$ must lie in $G_m$. This means that to each element $n$ there are exactly $\#G_m$ group elements which map $m$ into $n$. Therefore, since every element of $G$ carries $m$ into some orbit element, we conclude that

$$\#G=\#(G\cdot m)\#G_m$$

Above is an excerpt from Sternberg, Group Theory and Physics. To be clear, $G\cdot m \subset M$ is the orbit of $m\in M$, and $G_m \subset G$ the isotropic group of each $m$. I know that for any two elements $a,b \in G$ that map $m$ to $n$, there are two elements $a^{-1}b$ and $b^{-1}a$ in $G_m$, but I don't understand the emphasized bit, since $n$ is any element in $G\cdot m$, while $\#G_m$ is the fixed number. We can't be sure if there exists $b$ in $G$ besides $a$ such that $bm=n$, and even if with this particular $n$ there are $\#G_m$ ways to map $m$ to $n$, then nothing precludes another $n'\in G\cdot m$ from having a different number of ways to be mapped from $m$.

So why is that?

$\endgroup$
2
$\begingroup$

What the text says is that if $b$ maps $m$ to $n=a\cdot m$, then $b\in aG_m$. In fact the converse is true : if $b=ag$ for some $g\in G_m$, then $$b\cdot m=(ag)\cdot m=a\cdot (g\cdot m)=a\cdot m=n.$$ Thus the set of elements of $G$ that maps $m$ to $n$ is exactly $aG_m$, which has the same number of elements as $G_m$. This is true for any $n$ in the orbit of $G$ (although of course with a different element in the orbit you will get a different coset of $G_m$); in particular, the number of ways to map $m$ to any element of its orbit only depends on $m$.

$\endgroup$
2
$\begingroup$

For any $n$ in the orbit of $m$, there obviously exists $a \in G$, and $a \notin G_m$, such that $am = n$. Now, note that $aG_mm = am = n$. Thus, you have exactly #$G_m$ elements that map $m$ to $n$ (note that $G_m$ contains $1$, so when you say #$G_m$, it counts $a$ as well).

There cannot be another $n' \in G.m$ that has a different number of ways. The same argument as above applies. If there is $a' \in G$ such that $a'm = n'$, then $a'G_mm = a'm = n'$. Note that there cannot be $x \in G, x \notin a'G_m$ such that $xm = n'$ because in that case, $$xm = n' \implies m = x^{-1}n' \implies a'G_mm = a'G_mx^{-1}n' \implies n' = a'G_mx^{-1}n' \implies a'G_mx^{-1} = 1 \implies x = a'G_m,$$ which is a contradiction.

This observation is the departure point for any and all applications of group actions. What you have is actually a bijection between $G/G_m$ and $G.m$, $aG_m \mapsto am$. This to me is one of the most remarkable expedients of group actions. When someone says a group is acting on a set, it immediately tells us a whole bunch of things about the group and the set. (When I read about group actions for the first time, I thought that one could, without much effort, define an action between any group and any set. This is NOT TRUE.)

The orbits of a group action are equivalence classes for the relation $x \sim x'$ if $x' = gx$ for some $g \in G$. The orbits partition the set, i.e. the set can be expressed as a disjoint union of orbits.

This image is what you must keep in mind when you know a group is acting on a set.enter image description here

$\endgroup$

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.