12

I had this "almost bonus" question on the final in Group Theory recently: prove that there is no such group $G$ which would satisfy ${\rm Aut}(G)\cong \mathbb{Z}_n$, where $n$ is an odd integer. I don't have much certainty if this proof is OK, and one's opinion would be appreciated.

Here's my attempt:

Suppose that there exists such a group $G$ that satisfies the above condition. Since $\mathbb{Z}_n$ is cyclic, ${\rm Aut}(G)$ is also cyclic and $G$ is abelian, which implies that ${\rm Inn}(G)\cong \{e\}$. Thus ${\rm Aut}(G)={\rm Out}(G)$. (I don't think this fact is important here though). Now, since an automorphism sends a generator to a generator, and since each automorphism is completely determined by such mapping, $|{\rm Aut}(G)|$ must be of factorial order. But an integer factorial is always even. However, $|\mathbb{Z}_n|$ is odd, which is a contradiction. Therefore, no such $G$ exists.

I'm afraid that my proof is not very nice, but at least a genuine attempt was made.

Shaun
  • 44,997
sequence
  • 9,638

3 Answers3

8

$\DeclareMathOperator{\Aut}{Aut}$ We'll show the slightly stronger result that $|\Aut(G)| \not= 2n+1$ for any $n \ge 1$ EDIT: Provided $G$ is abelian: see the comments.

If $\Aut(G)$ is trivial, then there is nothing to show. Suppose, then, that $G$ has a nontrivial automorphism group. We will show that $|\Aut(G)|$ is even by explicitly exhibiting an automorphism $\phi$ of $G$ having order $2$ for any group $G$. It will then follow by Lagrange's theorem that $\Aut(G)$ has even order.

If $G$ has an element of order at least $3$, then $\phi:x \to x^{-1}$ is the desired automorphism of even order. It only remains to consider the case where $G$ contains only elements of order less than or equal to $2$. In that case, we must have that $G$ is a vector space over $\mathbb{Z}_2$ (see Group where every element is order 2 for a proof of this result). If $G$ is one dimensional as a vector space, then $\Aut(G) = \Aut(\mathbb{Z}_2)$ is trivial, so there is nothing more to show. For $G$ having dimension greater than $1$, we have that $\Aut(G) \cong \operatorname{GL}(G)$, so if we pick an ordered basis for $G$, then the automorphism which interchanges the first and second basis vectors (and fixes all others) is an order $2$ automorphism. (This argument, which Slade gave in the comments, applies even to infinite $G$, while my previous argument tacitly assumed $G$ was finite.)

Community
  • 1
  • 2
    This assumes that $G$ is finite, but it is very close to applying to infinite groups as well. An abelian group in which every nonzero element has order $2$ is a vector space over $\mathbb{Z}_2$, so it has a basis and we can write $G = \oplus_i \mathbb{Z}_2$ and find lots of nontrivial automorphisms. – Andrew Dudzik Dec 13 '15 at 20:31
  • @Slade Good point! I've changed my answer slightly to incorporate your comment. –  Dec 13 '15 at 20:39
  • 1
    @Strants : So you are assuming $G$ is abelian ? –  Apr 30 '16 at 07:41
  • 2
    Every group has an element of order at most 2 (namely, the identity element). You mean a group whose every element has order at most 2 (i.e. "exponent" 2). As a matter of fact, there do exist finite groups $G$ whose automorphism group ${\rm Aut}(G)$ has odd order. The correct proof of OP's desired claim begins as follows: if ${\rm Aut}(G)$ is cyclic then so is ${\rm Inn}(G)\cong G/Z(G)$, and a standard exercise shows that if $G/Z(G)$ is cyclic then $G$ is abelian. – anon Apr 30 '16 at 19:14
  • @arctictern Oops, you are correct. The requirement that $G$ is abelian is necessary if $G$ has an element of order greater than $2$: otherwise, $x \to x^{-1}$ will fail to be an automorphism. I've re-edited to make it clear I'm assuming that $G$ is commutative. –  Apr 30 '16 at 19:20
  • @arctictern Would you like to expand your comment into an answer? –  Apr 30 '16 at 19:27
6

It is not true that $|Aut(G)|=m!$ for some $m$. For example, if $G=\mathbb{Z}_5$, then $Aut(G)\cong\mathbb{Z}_4$ and 4 is not a factorial.

If you know that $G$ is abelian, then it is the product of cyclic group $G=\prod \mathbb{Z}_{p_i^{k_i}}$ where the $p_i$ are prime. It follows that $Aut(\mathbb{Z}_{p_i^{k_i}})$ is a subgroup of order $(p_i-1)p_i^{k_i-1}$ of $Aut(G)$. If $p_i$ is odd, then $p_i-1$ is even so the order of $Aut(G)$ is even. Similarly, if $p_i=2$ and $k\geq 2 $ then $p^{k-1}$ is even. So you just need to check the case where $G= \mathbb{Z}_2^m$ in which unless $m=1$ (and then $Aut(G)$ is trivial), you can always find an automorphism of order 2 - for example, switch two basis elements.

Actually, it is true that $Aut(G) $ is cyclic (and not trivial) iff $G$ is cyclic of order 4, $p^k$ or $2p^k$ where $p$ is prime (see here). Using that it is easy to see that the order of $Aut(G)$ must be 2 or $(p-1)p^{k-1}$ both of which are even.

Community
  • 1
Ofir
  • 8,015
3

Suppose that $n\geq 3$ and $n$ is odd. If $Aut(G)\cong \mathbb{Z}_n$ then $Aut(G)$ is cyclic, which implies that $G$ is abelian. But if $G$ is abelian then the inversion map $x\mapsto x^{-1}$ is an automorphism of order $2$. By Lagrange's theorem, this implies that $|Aut(G)|$ is divisible by $2$. This is a contradiction since $n$ is odd.

JonHerman
  • 2,901