Can we have some (new) examples of group extensions $G/N=Q$ with continuous (e.g. Lie groups) $G$ and $Q$, but a finite discrete $N$? Note that $1 \to N \to G \to Q \to 1$.

What I know already contains:



What else are the examples that you can provide?

A systematic answer to obtain new examples (a few or even a list) is of course most welcome. ; )

  • $\begingroup$ What do you mean by continuous groups more precisely? Do you just mean connected topological groups? $\endgroup$ – Tobias Kildetoft Mar 30 '17 at 19:49
  • $\begingroup$ Yes. And Lie groups. $\endgroup$ – annie heart Mar 30 '17 at 19:50
  • $\begingroup$ But being just a Lie group does not mean much if you for example allow it to be zero-dimensional. I am still not entirely sure what you mean by continuous here, as that word usually only applies to maps, not groups. $\endgroup$ – Tobias Kildetoft Mar 30 '17 at 19:59
  • $\begingroup$ OK, I guess I am just a applied math or an engineer here. Dont get confused on the term continuous, I mean something like Lie groups like SU(N). $\endgroup$ – annie heart Mar 30 '17 at 20:01

In general, there is a way to classify extensions of group $1 \to N \to G \to Q \to 1$ for given $Q$ and $N$. You need

  • a group morphism $\phi$ from $Q$ to $Out(N)$ (the outer automorphism group of $N$),

  • and the cohomological class of a $2$-cocycle in $H^2(Q,Z(N)_{\phi})$ ($Z(N)_{\phi}$ means that $Z(N)$ is considered as a $Q$-module with the action given by $\phi$).

You can read about the link between extensions of groups and second cohomology groups in Adem and Milgram's $\textit{Cohomology of Finite Groups}$.

I assume that by $G$ and $Q$ continuous groups, you mean topological groups. As far as I can tell, once you are given such extension, where $N$ is finite and $Q$ is a topological group there is only one way to extend the topology, the same is true for a Lie group. This comes from the fact that whenever we have $1 \to N \to G \to Q \to 1$ where $N$ is finite and $G$ and $Q$ are topological groups, the projection map $G\to Q$ is necessarily a topological cover.

Edit : as written in the comment below, you also need to add the condition that $\phi$ is continuous to ensure that $G$ is a topological group (here $N$ has the discrete topology).

  • 1
    $\begingroup$ Seems to me that if $Q$ is connected there are very few options for $\phi$. $\endgroup$ – Jyrki Lahtonen Mar 30 '17 at 8:26
  • $\begingroup$ Thanks! +1, could you provide us some new examples? $\endgroup$ – annie heart Mar 30 '17 at 15:26
  • $\begingroup$ Can you comment the difference between the 2nd cohomology group $H^2(Q,Z(N)_{\phi})$ and $H^2(Q,N)$? $\endgroup$ – annie heart Mar 30 '17 at 15:31
  • $\begingroup$ @Clément Guérin, Nice, what role does the Torsion play here? $\endgroup$ – wonderich Mar 30 '17 at 15:47
  • $\begingroup$ @annieheart, $H^2(Q,N)$ refers twisted cohomology with non-abelian coefficients, it is very hard to even define the second cohomology group and the higher cohomology groups. Whereas $H^2(Q,(Z(N)_{\phi})$ can be nicely define with the non-homogenous cochain definition. As for the new examples, you can take $G$ to be the Heisenberg group and $N$ to be its center. It may be interesting. Just bear in mind that $\phi$ needs to be trivial if $Q$ is connected (see Jyrki's Lahtonen's comment). $\endgroup$ – Clément Guérin Apr 2 '17 at 8:23

Let $G = Q = \{z; z \in \Bbb C, \left| z \right| = 1\}$ and $N = \{1,-1\}$.

  • $\begingroup$ That is nice +1, but I had it in $\frac{\mathbb{R}/{\mathbb{Z}}}{\mathbb{Z}_n}={\mathbb{R}}/{(n\mathbb{Z})}.$ with $n=2$ is your example. $\endgroup$ – annie heart Mar 30 '17 at 18:57
  • $\begingroup$ Let us try some more nontrivial examples. $\endgroup$ – annie heart Mar 30 '17 at 18:58
  • $\begingroup$ Same construction on the quaternions with norm $1$ and $N$ the quaternion group. $\endgroup$ – Marc Bogaerts Mar 30 '17 at 18:59
  • $\begingroup$ Thanks please update more examples on the answer so we can see explicitly. ; ) $\endgroup$ – annie heart Mar 30 '17 at 19:01
  • $\begingroup$ My previous example is wrong since $N$ is not a notmal subgroup. The good news is that $N = \{\1,-1}$ still works. Moreover for any "classic" group one can take for $N$ any finite subgtoup of the center (scalar matrices) such as the $n$ -th roots of unity. $\endgroup$ – Marc Bogaerts Mar 31 '17 at 7:42

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.