1
$\begingroup$

I have a question. The automorphism group of the linear groups $GL(n,q)$, the group of linear transformations of $V = \mathbb{F}_q^n$, and $SL(n,q)$, the subgroup of $GL(n,q)$ consisting of elements of determinant $1$, are well understood. Here $q$ is a prime power $p^r$ and $n$ is any positive integer.

My question is: what is known about the automorphism group of $AGL(n,q)$ and $ASL(n,q)$? Here $AGL(n,q)$ is the affine linear group $V \rtimes GL(n,q)$, semidirect product of $V = \mathbb{F}_q^n$ and $GL(n,q)$ with the natural action of $GL(n,q)$ on $V$. These groups have matrix representations, as you can see here https://en.wikipedia.org/wiki/Affine_group

What I know for sure is that there are the inner automorphisms, and the Frobenius automorphism (raise the coefficients to the power $p$) that comes from the matrix representation described in the link. But is there something else?

Is there a nice complete description of the automorphism groups of $AGL(n,q)$ and $ASL(n,q)$?

Thanks

$\endgroup$
1
  • $\begingroup$ I'm pretty sure that you mean to raise the entries to the power $p$, not $r = \log_p q$. $\endgroup$
    – LSpice
    Aug 29, 2015 at 3:12

2 Answers 2

5
$\begingroup$

This is just a brief answer. The only autmorphisms other than the ones that you know about already arise from elements of $H^1(H,V)$ (with $H = {\rm SL}(n,q)$ or ${\rm GL}(n,q)$), where the corresponding automorphisms induce the identity on $V$ and on $G/V \cong H$, but map a complement $H$ of $V$ to a complement that is not conjugate to $H$ in $G$.

These cohomology groups are all known. The only cases in which they are non-zero are when $H = {\rm SL}(2,2^k)$ with $k \ge 2$, and $H = {\rm SL}(3,2) = {\rm GL}(3,2)$, and in each of these cases, the dimension of $H^1(H,V)$ is $1$-dimensional over ${\mathbb F}_q$.

So, for example, for $G = {\rm ASL}(2,8)$, the order of the automorphism group is $24|{\rm AGL}(2,8)| = 168|G|$, with a factor $3$ coming from the field automorphisms of ${\mathbb F}_8$, and the factor $8$ coming from $H^1(H,V)$.

As a quick justification, note that $V$ is clearly characteristic in $G$, so ${\rm Aut}(G)$ induces a subgroup of the automorphism group of $G/V$. The full automorphism group of ${\rm SL}(n,q)$ is $\Gamma L(n,q)(.2)$, where the $.2$ is the duality automorphism when $n \ge 3$. The duality automorphism of $G/V$ does not lift to an automorphism of $G$, but we do have $A \Gamma L(n,q) \le {\rm Aut}(G)$.

Since $G$ acts absolutely irreducibly on $V$, automorphisms of $G$ inducing the idenity on $G/V$ must act as scalars on $V$, and we already have the full group of scalars present in $A \Gamma L(n,q)$.

So it remains to consider automorphisms of $G$ that induce the identity on both $G/V$ and on $V$ and, modulo inner automorphisms induced by conjugation by elements of $V$, these correspond exactly to $H^1(H,V)$.

A reference for the results on $H^1(H,V)$ is: G.W. Bell, On the Cohomology of the Finite Special Linear Groups, I, Journal of Algebra 54, 216-238 (1978), but I am not sure whether that was the first proof.

$\endgroup$
2
$\begingroup$

Note that $V = F^{\ast}({\rm AGL}(n,q))$, where $F^{\ast}$ denotes the generalized Fitting subgroup.

For any finite group $G$, restriction gives a homomorphism from ${\rm Aut}(G)$ into ${\rm Aut}(F^{\ast}(G))$ whose kernel $A$ is an Abelian normal subgroup of ${\rm Aut}(G)$ and all prime divisors of $|A|$ are divisors of $|G|$. This is a standard application of the three subgroups lemma, and properties of coprime automorphisms, but I outline the argument. Let $A$ denote the kernel.

Now we have $[G,F^{\ast}(G),A] = [F^{\ast}(G),A,G] = 1$, so that $[A,G,F^{\ast}(G)] =1$. Hence $[G,A] \leq Z(F(G))$, so that $[A,G,A] = [G,A,A] = 1$ Hence $[A,A,G] = 1$, so that $[A,A]$ is trivial. If $|A|$ has a prime divisor $p$ which does not divide $|G|$, and $P$ is a Sylow $p$-subgroup of $A$ then $G = [G,P]C_{G}(P)$ and $[G,P,P] = [G,P]$ while we saw above that $[G,P,P] = 1$. Hence $P$ is trivial.

$\endgroup$

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.