Suppose $G=\mathbb{Z}\ast C_n=\langle a, b; b^n\rangle$ is the free product of the infinite cyclic group with a finite cyclic group. Then $G$ has finite outer automorphism group. However, my proofs (and the proofs I have found in the literature*) of this fact use lots of Nielsen-esq automorphisms-of-free-groups theory (with a touch of Whitehead, if I recall correctly).

I was wondering if anyone knows of any short (two-liner), lower-level proofs?

*(Although I will admit I skipped the classic reference, which is a paper of Fouxe-Rabinovitch written in Russian with a MathSciNet entry in German. I speak neither language.)


Are you willing to use the fact that all subgroups of order $n$ in $G$ are conjugate? If so, then we can assume that our automorphism fixes $b$. Then it is enough to prove that it must map $a$ to $b^iab^j$ for some $i,j$, and there are only finitely many such automorphisms. So it is enough to prove that, if $G = \langle g,b \rangle$, then $g = b^iab^j$. If not, then we can assume that the normal form of $g$ has the form $awa$ or $awa^{-1}$ for some nontrivial word $w$ containing $b^{\pm 1}$, and it is not hard to see you cannot generate $a$ from such a $g$ and $b$.

  • $\begingroup$ Thanks, yes, I hadn't thought to use the normal form theorem (actually, you could translate this proof into Bass-Serre theory, because, as your observation about only having one subgroup points out, all automorphisms fix the graph of groups structure (if you view your group as an HNN-extension with trivial associated subgroups - is that allowed?)). Very nice. Although someone should say that this is only an index $\phi(n)$ subgroup of $\operatorname{Out}(G)$, where $\varphi$ is the Euler totient function. (I am sure you know this, but I am also sure that someone reading this will not.) $\endgroup$ – user1729 Sep 13 '13 at 17:31
  • $\begingroup$ Since $b^iab^j$ is conjugate to $b^ia$ by a power of $b$, representatives of the outer automorphisms seem to be $a \mapsto b^ia^{\pm 1}$, $b \mapsto b^j$ with $(j,n)=1$, which would make $|{\rm Out}(G)|= 2n\varphi(n)$ if I've got that right. $\endgroup$ – Derek Holt Sep 14 '13 at 8:46
  • $\begingroup$ Yeah, it is actually isomorphic to $D_n\rtimes\operatorname{Aut}(C_n)$. $\endgroup$ – user1729 Sep 14 '13 at 10:13
  • $\begingroup$ @DerekHolt Excuse me for asking so late. Could you provide some reference for the fact that all subgroups of order $n$ in $G$ are conjugate. Thanks in advance! $\endgroup$ – 1123581321 Oct 9 '20 at 12:49
  • $\begingroup$ @ΓιάννηςΠαπαβασιλείου Well, every finite subgroup of a free product is conjugate into one of the free factors (which is standard). As $n<\infty$ is the order of one of the free factors, the claim follows. (In fact, something stronger holds: if $\mathbb{Z\ast Z}_n$ contains a subgroup of order $m$ then all subgroups of order $m$ are conjugate. The proof is identical, except at the last step you observe that cyclic groups contain precisely one subgroup of each order.) $\endgroup$ – user1729 Oct 9 '20 at 13:12

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.