I would like to ask a question about automorphisms of free products of groups. More specifically, let $G = G_1 \ast ... \ G_n \ast F_r$ where $F_r$ is free group on r generators. We can define the group $Out(G)$ of outer automorphisms of $G$, as the quotient of automorphisms with the inner automorphisms.

Then we can see as a subgroup of $Out(G)$, the automorphisms that have a representative $\phi$ from $G$ to $G$, which induce the identity on $G_i$'s and restricted to $F_r$ is an automorphism from $F_r$ to $F_r$. My question is if this subgroup shares any properties with $Out(F_r)$, for example can we say that this subgroup is virtually torsion free or maybe every torsion subgroup is finite?

Thanks a lot in advance for your time.


1 Answer 1


EDITED: Thanks to @Mathieu for finding a bug.

The group you describe is actually isomorphic to $\text{Aut}(F_r)$ (which, of course, has a well known relation to $\text{Out}(F_r)$).

To see why, you have defined a subgroup of $\text{Aut}(G)$ which I will denote $A_0$, consisting of all automorhisms which take each $G_i$ to itself by the identity and which take $F_r$ to itself by some automorphism. Restricting each element of $A_0$ to $F_r$ defines an isomorphism $A_0 \mapsto \text{Aut}(F_r)$, so we just need to show that the induced map $A_0 \mapsto \text{Out}(G)$ is injective. For this it suffices to show that for each $g \in G$ with corresponding inner automorphism $i_g(h) = ghg^{-1}$, and for each $\phi \in A_0$, if $i_g \circ h \in A_0$ then $g$ is the identity.

For each nontrivial free factor $B \in \{G_1,...,G_n,F_r\}$, if $g \not\in B$ then for any non-identity element $h \in B$ we have $ghg^{-1} \not\in B$, contradicting invariance of $B$. Therefore $g$ is an element of each of the free factors $B$, and this forces $g$ to be the identity (assuming that at least one of the $G_i$'s is nontrivial and that $r \ge 1$).

  • $\begingroup$ Thanks a lot for your answer. I suspected that it was related a lot with $Out(F_r)$, but I wasn't sure for the details. It is very helpful for me, thanks again. $\endgroup$
    – user75691
    Commented Dec 16, 2014 at 14:05
  • $\begingroup$ @user75691: It is my pleasure. $\endgroup$
    – Lee Mosher
    Commented Dec 16, 2014 at 14:08
  • 2
    $\begingroup$ Even though this question is old, I thought I'd nitpick on the accepted answer: assuming some $G_i$ is non-trivial, isn't the described subgroup isomorphic to $Aut(F_r)$ instead of $Out(F_r)$? Indeed, $Aut(F_r) \subset Aut(G)$ intersects $Inn(G)$ trivially: for any non-trivial $g \in F_r$ the inner automorphism $i_g(h) = ghg^{-1}$ does not act as the identity on any non-trivial $G_i$. $\endgroup$
    – Mathieu
    Commented Jun 29, 2018 at 8:37
  • $\begingroup$ Ah, that's a good point; having checked $g \in F_r$, I should also have checked $g \in G_i$ for each $i$. I'll fix it. $\endgroup$
    – Lee Mosher
    Commented Jun 29, 2018 at 14:37

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.