This came from an old idea of mine from when I was an undergraduate for how to approach the Schreier conjecture. I (obviously) never gave it much serious thought, but wondered if much was known about it.

Let $G$ be a finite group. The group $\mathrm{Out}(G)$ has a normal subgroup $\mathrm{Out}_c(G)$ of class-preserving outer automorphisms, i.e., the group $\mathrm{Aut}_c(G)/\mathrm{Inn}(G)$, where $\mathrm{Aut}_c(G)$ is the subset of $\mathrm{Aut}(G)$ that leave invariant each conjugacy class of $G$. Note that all primes dividing $|\mathrm{Out}_c(G)|$ divide $|G|$, and if $G$ is simple then $\mathrm{Out}_c(G)=1$ (Feit-Seitz).

One approach to Schreier without CFSG is to slice $\mathrm{Out}(G)$ into two, and prove that $\mathrm{Out}_c(G)$ and $\mathrm{Out}(G)/\mathrm{Out}_c(G)$ are both soluble. Each of these looks hard, but the obvious question, even with CFSG, is:

Is $\mathrm{Out}_c(G)$ soluble for any finite group $G$?

The obvious candidates for groups with non-inner class-preserving automorphisms are $p$-groups, where this statement clearly holds. So perhaps it has a chance of being true generally. I couldn't find much progress at all on class-preserving automorphisms, so perhaps this question is too far out of reach at the moment.

  • $\begingroup$ One may ask the same question for infinite groups, but I'd be amazed if it were true. $\endgroup$ Apr 20, 2021 at 23:19

1 Answer 1



Sah, Chih-han Automorphisms of finite groups. J. Algebra 10 (1968), 47–68.

Erratum: Sah, Chih Han "Automorphisms of finite groups'' (J. Algebra 10 (1968), 47–68): addendum. J. Algebra 44 (1977), no. 2, 573–575.

Theorem 2.10 in this paper states the following (erratum has a correction to the original proof):

Theorem: Suppose that $G$ is a group which has a composition series. If $\operatorname{Out}(F)$ is solvable for every composition factor $F$ of $G$, then $\operatorname{Out}_c(G)$ is solvable.

So it follows from the Schreier conjecture that $\operatorname{Out}_c(G)$ is solvable for any finite group $G$.

For infinite groups, in the paper above Sah points out the following example. Let $S$ be the symmetric group on an infinite countable set, and let $G$ be the normal subgroup formed by the permutations with finite support. In this case $\operatorname{Aut}_c(G) = S$, you can prove this the same way you prove that a transposition-preserving automorphism of the finite symmetric group is inner. Therefore $\operatorname{Out}_c(G) \cong S/G$, which is not solvable.

  • $\begingroup$ That's great, thanks. I'd love to see a CFSG free proof just for simple groups, but I think that won't happen. $\endgroup$ Apr 21, 2021 at 7:02
  • 1
    $\begingroup$ You have misquoted the theorem. It says "for each composition factor $F$ of $G$", not "for any composition factor $F$ of $G$", which is completely ambiguous. (I was genuinely unsure what it meant when I read it for the first time.) I really wish people would stop using the word "any" in formal mathematics! $\endgroup$
    – Derek Holt
    Apr 21, 2021 at 7:50
  • $\begingroup$ @DerekHolt: Thanks and I agree, edited now. $\endgroup$ Apr 21, 2021 at 8:08

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.