5
$\begingroup$

Let $H$ and $N$ be two groups with $H$ cyclic. Let $f,g:H \rightarrow \mathrm{Aut}(N)$ be homomorphisms such that $N\rtimes _f H \cong N\rtimes _g H$. Then does that mean $f(H)$ and $g(H)$ are conjugate in $\mathrm{Aut}(N)$?

$\endgroup$
5
  • 2
    $\begingroup$ This is a question from Dummit and Foote and I do not think it's suitable for mathoverflow. $\endgroup$
    – T.B.
    Jan 22, 2011 at 4:07
  • 1
    $\begingroup$ Dear Tin Bui, I didn't see that this question in Dummit-Foote. I am asking it for following reason. If $G$ has a normal subgroup $N$, and a subgroup $H$, with $N\cap H=1$, then $G$ is a semidirect product of $N$ by $H$. Consider simple case when $H$ is cyclic. Then To determine possible dofferent semidirect products of $N$ by $H$, we can look for only those homomorphisms from $H$ to $Aut(N)$, such that the images of $H$ under homomorphisms are not conjugate. But is this sufficient to get non-isomorphic semidirect products? I worked for small groups, but I couldn't find counterexample. $\endgroup$
    – Soluble
    Jan 22, 2011 at 4:34
  • $\begingroup$ If you consider $f$ and $g$ to be functors from the one-object groupoids determined by $H$ to $Grp$ (with the object part picking out $H$), then there is a natural transformation $f\Rightarrow g$ precisely when for each $h\in H$ there is an $a\in Aut(N)$ such that $af(h)a^{-1} = g(h)$. I'm pretty sure that a natural transformation from $f$ to $g$ induces an isomorphism between the extensions determined by $f$ and $g$. You may need your isomorphism to commute with the maps to $H$. in order to reverse this argument. $\endgroup$
    – David Roberts
    Jan 22, 2011 at 6:04
  • $\begingroup$ @Rahul: I look back at the exercise and Zev was right. Sorry, although I do think you perhaps can get faster and more detailed response to this type of question at math.stackexchange.com as I did when I had tried it. $\endgroup$
    – T.B.
    Jan 22, 2011 at 6:07
  • 1
    $\begingroup$ Many books of algebra/group theory, and many articles on Semi-direct product give the following theorem: "If H is cyclic and f,g:H --> Aut(N) are homomorphisms, such that f(H) and g(H) are conjugate subgroups of Aut(N), then these two homomorphisms give isomorphic semi-direct product of N by H. But I didn't see, whether its converse is true, in any book, article, even not as an exercise, or any counterexample. I worked for small groups, but couldn't get. $\endgroup$
    – Soluble
    Jan 22, 2011 at 7:25

1 Answer 1

7
$\begingroup$

The answer is no. You can easily have a situation where $f$ is the trivial map, while $g$ makes $H$ act through an inner automorphism of $N$, so that in both cases $N\rtimes H\cong N\times H$. For concreteness, let $N=D_8$ be the dihedral group of order 8, let $\sigma$ be a non-central involution in $N$, let $H=\langle h\rangle\cong C_2$, and define $g(h)(n)=\sigma^{-1} n\sigma\;\forall n\in N$. Then $G=N\rtimes_g H\cong N\times H$, since the subgroup $\langle h\sigma\rangle$ of $G$ is of order 2, intersects $N$ trivially, and commutes with it.

$\endgroup$
5
  • 1
    $\begingroup$ You beat me to it by several seconds... What if we ask the same question after passing to Out(N)? $\endgroup$
    – Dan Ramras
    Jan 22, 2011 at 8:27
  • $\begingroup$ Oh, I guess you undeleted an old answer? $\endgroup$
    – Dan Ramras
    Jan 22, 2011 at 8:28
  • $\begingroup$ @Dan That's right, sorry. See the comment thread on the question for the explanation of why I deleted. I do agree that the question is elementary, but I have tripped over semi-direct products more than once myself. While it's something I would expect any researcher to be able to figure out rather quickly, I wouldn't expect that any mathematicians must know the answer off the top of his head. $\endgroup$
    – Alex B.
    Jan 22, 2011 at 8:31
  • $\begingroup$ Yeah, I've actually posted this sort of example somewhere else on MO before. But I've never thought about the question of conjugacy of f(H) and g(H) in Out(N). $\endgroup$
    – Dan Ramras
    Jan 22, 2011 at 8:34
  • $\begingroup$ For what it's worth "somewhere" is here: mathoverflow.net/questions/23456/… $\endgroup$
    – Dan Ramras
    Oct 11, 2016 at 4:56

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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