3
$\begingroup$

This is in some sense a continuation of this problem.

Given a group $G$ I would like to exhibit two actions of $G$ on a set $[n] =\{1,\ldots,n\}$ such that the two actions are isomorphic yet not equivalent.

To recall we say that two group actions $\alpha,\beta : G \mapsto S_n$ are isomorphic if there is an automorphism $f:G \mapsto G$ and bijection $\varphi:[n] \mapsto [n]$ such that

$$\varphi(o^{\alpha(g)}) = \varphi(o)^{\beta(f(g))} \quad \mbox{ for all } o \in [n], g \in G.$$

Finally, we say that they are equivalent if $f$ is the identity map.

The question is as the title says

Given a group $G$ can we find two isomorphic but not equivalent actions of $G$?

I guess not much can be done if $G$ is the trivial group, but what if not?

$\endgroup$
  • $\begingroup$ Surely this holds if and only if there is a nontrivial automorphism $f$? Given such an $f$ and a group action $\alpha$, simply define $\beta(g)=\alpha(f^{-1}(g))$ for all $g\in G$. $\endgroup$ – Jason Feb 10 '15 at 22:51
  • $\begingroup$ Where are you "recalling" your definitions of isomorphism and equivalence for group actions from? $\endgroup$ – Rob Arthan Feb 10 '15 at 22:57
  • $\begingroup$ @RobArthan Some lectures notes from a class I have. $\endgroup$ – Jernej Feb 10 '15 at 23:13
1
$\begingroup$

Your allowance of the bijection $\varphi$ is the same as allowing for an inner automorphism of $S_n$. In other words, you can rephrase your definition of isomorphic actions as:

Two actions (ie, homomorphisms) $\alpha,\beta : G\rightarrow S_n$ are isomorphic if there is an automorphism $f\in\text{Aut}(G)$ and an element $\sigma\in S_n$ corresponding to the inner automorphism $c_\sigma\in\text{Aut}(S_n)$ such that $\beta = c_\sigma\circ\alpha\circ f$. (Note that every automorphism of $S_n$ for $n\ne 2,6$ is inner, so we'll assume that $n\ne 2,6$ from now on)

Your question is then to ask: if $\alpha,\beta$ are isomorphic, then is $\beta = \alpha\circ f'$ for some $f'\in\text{Aut}(G)$.

Note that if the images of $\alpha,\beta$ are distinct, then of course it's impossible for them to be equivalent. On the other hand, they could be isomorphic. For example, if the image of $G$ is not normal in $S_n$, and $\sigma\in S_n$ an element which does not normalize $\alpha(G)$, then $\beta = c_\sigma\circ\alpha$ is isomorphic to $\alpha$ but not equivalent.

So lets assume $\alpha(G) = \beta(G)$.

If $\beta = c_\sigma\circ\alpha\circ f$ and $\sigma\in\alpha(G)$, then you can pick a preimage $g\in \alpha^{-1}(\sigma)$, in which case $\beta = \alpha\circ c_g\circ f$, where $c_g\in\text{Aut}(G)$ is just conjugation by $g$, so you can pick $f' = c_g\circ f$.

In general the answer is no. For example, you could embed $(\mathbb{Z}/5\mathbb{Z})^2$ inside $S_{25}$ via its action on itself. Let $G = \mathbb{Z}^2$, and consider the maps $\alpha,\beta : \mathbb{Z}^2\rightarrow(\mathbb{Z}/5\mathbb{Z})^2$ given by $\alpha$ sending $(x,y)\mapsto(x,y)\mod 5$, and $\beta$ sending $(x,y)\mapsto(2x,y)$.

Note that $\text{Aut}(\mathbb{Z}^2)\cong GL_2(\mathbb{Z})$ - in particular every automorphism has determinant $\pm 1$. On the other hand, the unique automorphism $\gamma$ of $(\mathbb{Z}/5\mathbb{Z})^2$ sending $(x,y)\mod 5\mapsto (2x,y)\mod 5$ has determinant $\equiv 2\mod 5$, which does not come from any automorphism of $\mathbb{Z}^2$ (since those have determinants $\pm 1$ and $2\not\equiv\pm 1\mod 5$).

Finally note that because $(\mathbb{Z}/5\mathbb{Z})^2$ is embedded in $S_{25}$ via the regular representation, the automorphism $\gamma$ extends to an (inner) automorphism of $S_{25}$ (the proof is very easy, but if you just want validation of this fact, see, for example, https://mathoverflow.net/questions/119377/automorphism-of-finite-groups-and-hurwitz-spaces). Thus, $\alpha,\beta$ are isomorphic ($\beta = \gamma\circ\alpha$) but not equivalent (there is no $f\in\text{Aut}(\mathbb{Z}^2)$ such that $\beta = \alpha\circ f$).

| cite | improve this answer | |
$\endgroup$

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.