11

The title basically says it all.

Is there a group with more than one element that is isomorphic to the group of automorphisms of itself?

I'm mainly interested in the case for finite groups, although the answer for infinite groups would still be somewhat interesting.

4 Answers4

29

The automorphism group of the symmetric group $S_n$ is (isomorphic to) $S_n$ when $n$ is different from $2$ or $6$. In fact, if $G$ is a complete group you can ascertain that $G \simeq \mathrm{Aut}(G)$. The reverse implication needn't hold, though.

  • 8
    As mentioned in the comments above, $D_8$ is a counterexample to the reverse implication, since it is not complete but is isomorphic to its automorphism group. – Joel David Hamkins Jul 02 '10 at 02:02
  • 3
    Could you perhaps add the definition of a complete group? – user717 Jul 02 '10 at 09:28
  • 10
    Complete = centerless + all automorphisms are inner. Such a group G is isomorphic to Aut(G) by the map taking an element g to the inner automorphism conjugation-by-g. – Joel David Hamkins Jul 02 '10 at 16:26
16

I am slightly surprised that Wielandt's automorphism tower theorem has not been mentioned: this asserts that given a finite group $G$ with trivial center, the sequence of groups defined by $G_{0} = G$ and $G_{n+1} = {\rm Aut}(G_{n})$ for $n \geq 0$ eventually stabilizes, that is at some stage ${\rm Aut}(G_{n}) = G_{n}$.

  • 3
    This result also extends to infinite centerless groups. (http://math.univ-lyon1.fr/~altinel/Master/M1Logique/simonthomasautomorphismtower.pdf) – Burak Sep 13 '15 at 22:43
  • @Burak: Thanks for pointing that out. I was not sure about the infinite case. – Geoff Robinson Sep 13 '15 at 22:46
  • 2
    Prof. Robinson: There is even a MO thread on that pearl: http://mathoverflow.net/questions/5635/does-mathrmaut-mathrmaut-mathrmautg-stabilize – José Hdz. Stgo. Sep 13 '15 at 23:12
3

For the absolute Galois group of $\mathbb Q$, the map $g\mapsto (h \mapsto ghg^{-1})$ is an isomorphism between $G_\mathbb Q$ and its automorphism group. This is a corollary of the Neukirch-Uchida theorem.

user00000
  • 376
  • 2
  • 7
3

See also Problem 2015-4C in the Nieuw Archief voor Wiskunde. A solution (of mine) can be found here: http://www.nieuwarchief.nl/home/problems/pdf/uitwerking-2015-4.pdf