Let G and H be two groups. There is a one-to-one correspondence between:

(i) an (isomorphism class of) extension of G by H, i.e. an exact sequence of group morphisms $1\to H\to E\to G\to 1$;

(ii) an (isomorphism class of) action of the group G on the category of H-sets.

Actually, this can even be made an equivalence of categories.

This is probably well-known by category-theoretists, but I found no reference on the subject. Does anybody know of an article or book where it has been treated ?

Lots of thanks, in advance.

  • $\begingroup$ I think this question is too localized for MO. It is a basic result which is contained in every introduction to semidirect products of groups. I think that the category-theory tag is wrong. $\endgroup$ – Martin Brandenburg Feb 14 '12 at 21:32
  • $\begingroup$ The category of H-sets is equivalent to the category BH with one object and H as automorphisms. Then you statement goes by the name of Schreier-Theory or Dedecker-Cocycles and dates back to the beginning of the 20th century. $\endgroup$ – Thomas Nikolaus Feb 14 '12 at 21:43
  • 4
    $\begingroup$ Martin, are you are you read well the question ? This equivalence of category, if not utterly deep, is not to be found "in any introduction to semi-direct products", and is surely about category-theory. $\endgroup$ – Joël Feb 14 '12 at 22:20
  • $\begingroup$ John Baez wrote about Schreier theory in his TWF series: math.ucr.edu/home/baez/week223.html -- he points out a good reference for the Schreier theory of groupoids too: arxiv.org/abs/math.CT/0410202 . $\endgroup$ – Finn Lawler Feb 14 '12 at 22:38
  • 1
    $\begingroup$ (Also, I think H-Set is equivalent not to BH but to [BH, Set].) $\endgroup$ – Finn Lawler Feb 14 '12 at 22:40

The beginning of SGA 7, Exposé VII, essentially proves an equivalence between extensions of $G$ by $H$ and monoidal functors from $G$ (thought of as a discrete monoidal category) to the monoidal category of $H$-bitorsors. The correspondence here takes an extension $1 \to H \to E \stackrel{\pi}{\to} G \to 1$ to the functor sending $g \in G$ to the fiber $\pi^{-1}(g)$, which naturally carries the structure of an $H$-bitorsor via the multiplication in $E$.

This latter category can be identified with the monoidal category of autoequivalences of the category of $H$-sets: an autoequivalence of the category of $H$-sets is given as tensoring by an $H$-bitorsor, and natural isomorphisms between autoequivalences correspond to isomorphisms of $H$-bitorsors. Off the top of my head, I don't know of a reference that describes this latter equivalence explicitly.

  • $\begingroup$ By the way, if $A$ is an algebra, there is a similar statement for strongly $G$-graded algebras with unit component $A$, and actions of $G$ on the category of $A$-modules. Do you know if this is also done in SGA 7 ? Thanks again. $\endgroup$ – Erwan Biland Feb 15 '12 at 1:40
  • $\begingroup$ Does this work also for $G$ and $H$ which are topological or Lie by using classifying their classifying topoi? $\endgroup$ – David Carchedi Feb 18 '12 at 14:54
  • $\begingroup$ @Erwan: I doubt it's in SGA7, but the argument should be almost exactly the same. @David: In Grothendieck's article, he works in an arbitrary (Grothendieck) topos, so I think the answer to your question is yes (modulo checking what Grothendieck actually says, as opposed to my rough translation of it). $\endgroup$ – Evan Jenkins Feb 24 '12 at 2:16
  • $\begingroup$ The equivalence between autoequivalences and bitorsors is a version of the Eilenberg-Watts theorem; it can be deduced from the universal property of the Yoneda embedding. $\endgroup$ – Qiaochu Yuan Aug 6 '16 at 0:06

Such extensions are what D. Conduché called crossed 2-modules (in French: modules croisés de longueur 2) in his 1983 paper: Modules croisés généralisés de longueurs 2, J. Pure and Appl. Alg. Vol. 34, Issues 2-3 (1984), pp. 155-178. My guess is that an action of $G$ on the category of $H$-sets amounts to what the same author called "non-abelian pre-crossed complex" (in French: complexe pré-croisé non abélien) in the same paper. The equivalence you talked about is then Théorème 2.6 of the above mentioned reference.


Another relevant paper is

R. Brown, G. Danesh-Naruie, J.P.L. Hardy ``The fundamental groupoid as a topological groupoid'', Proc. Edinburgh Math. Soc. 19 (1975) 237-244.

Section 4 explains that an extension $1 \to B \to E \to G \to 1$ is isomorphic to an exact sequence determined by a fibration of groupoids $G \ltimes A \to G$ where $A$ is a groupoid which is a $G$-module. In the case in point, $A$ is the action groupoid determined by the action of $E$ on the right of $G$.

Understanding of this really requires the notion of covering morphism of groupoids and the equivalence of categories between covering morphisms of $G$ and actions of $G$ on sets. I feel this notion of covering morphism is somewhat neglected. However in the book

Higgins, P.J. Notes on categories and groupoids, Mathematical Studies, Volume 32. van Nostrand Reinhold Co. London (1971); Reprints in Theory and Applications of Categories, No. 7 (2005) pp 1-195.

the notion is used to prove a generalisation of Grushko's theorem. And fibrations of groupoids are also useful algebraically, and topologically!

  • $\begingroup$ Thank you for this answer. It turns out that I might have to work with a notion of covering morphism of G-algebras, due to Puig. I do not know if the two notions are related. Anyway, your post gave me the opportunity to learn about groupoids. Thanks ! $\endgroup$ – Erwan Biland Feb 20 '12 at 17:02
  • $\begingroup$ I'd be grateful for a reference to the work of Puig. There is an Exercise 8 on p. 379 of "Topology and groupoids" in which I suggested a kind of research project on dealing with a subalgebra, in fact an ideal, by related methods, but I am not sure if that is the good direction! $\endgroup$ – Ronnie Brown Mar 2 '12 at 17:01

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.