Prove that for $n\gt 2$, $| \operatorname{Aut}(D_n)|\le n\,\phi(n)$ where $D_n$ is the dihedral group with 2n elements and $\phi$ is Euler phi function.

Let $\rho$ be a rotation such that $o(\rho)=n$, that is $R_n = \langle\rho\rangle$, and $\psi$ an automorphism of $D_n$; then $\psi(\rho)$ must have order n and there are only $\phi(n)$ such elements in $D_n$ and they are all rotations, therefore $\psi(R_n)=R_n$. Now let $\iota$ be the reflection through the x-axis, we have to send it in one of the n reflection and since $D_n = \langle\rho, \iota \rangle$, $\psi$ is univocally determined. In conclusion we have at most $n\phi(n)$ choices for $\psi$.

I could have also considered all the sets with two elements that generate $D_n$ which are of the form $\{\rho, \iota\}$ with $\rho$ a rotation of order n and $\iota$ a reflection (there are $n\phi(n)$ of them); since an automorphism sends a set of generators into a set of generators we have again at most $n\phi(n)$ automorphisms (a rotation will be sent in a rotation) obtained by extending to a omomorphism the various choices (which give us bijective functions). Moreover there are exactly $n\phi(n)$ of them because every choice give us a different automorphism.

Are both solutions, and my remark, correct?

Thanks in advance

Edit: I was wrong saying that the only sets of two elements generating $D_n$ are of the form $\{\rho, \iota\}$, because there are also sets formed by two reflections, but since a rotation must be sent in a rotation my second proof should be correct.


Your explanation is correct. I have a remark that the inequality you want to prove is indeed an equality.

Let $C_n$ denote the cyclic group of order $n$. The group $G:=\text{Aut}\left(D_n\right)$ is isomorphic to the semidirect product $H:=C_n\rtimes \text{Aut}\left(C_n\right)$, where $$\left(c_1,f_1\right)\cdot \left(c_2,f_2\right):=\big(c_1\,f_1\left(c_2\right),f_1\circ f_2\big)$$ for all $c_1,c_2\in C_n$ and $f_1,f_2\in\text{Aut}\left(C_n\right)$. If $C_n$ is generated by $c$, then each element of $\text{Aut}\left(C_n\right)$ sends $c$ to $c^k$ for some $k=1,2,\ldots,n$ with $\gcd(k,n)=1$, and we write $\gamma_k$ for this element of $\text{Aut}\left(C_n\right)$.

The reason that $G$ is isomorphic to $H$ is as follows. Let $$D_n=\left\langle r,s \,|\,r^n=s^2=1\text{ and }rs=sr^{-1}\right\rangle=\left\{1,r,r^2,\ldots,r^n,s,rs,r^2s,\ldots,r^ns\right\}\,.$$ Hence, for each $\tau\in G$, it suffices to look at $r_\tau:=\tau(r)$ and $s_\tau:=\tau(s)$. We have $r_\tau=r^{k}$ and $s_\tau=r^{j}s$ for some $k=1,2,\ldots,n$ with $\gcd(k,n)=1$ and for any $j=0,1,2,\ldots,n-1$. Thus, we write $\tau_{j,k}$ for this automorphism $\tau$. Then, the map $\psi:G\to H$ sending $\tau_{j,k}$ to $\left(c^j,\gamma_k\right)$ is a group isomorphism. That is, $$\big|\text{Aut}\left(D_n\right)\big|=|G|=|H|=\left|C_n\right|\,\big|\text{Aut}\left(C_n\right)\big|=n\,\phi(n)\,.$$ You have to fill in a lot of gaps I intentionally left out, of course.

P.S. Interestingly, the condition $n>2$ is important. It turns out that $D_2\cong C_2\times C_2$ is extraordinary. That is, $$\text{Aut}\left(D_2\right)\cong\text{Aut}\left(C_2\times C_2\right)\cong \text{GL}_2\left(\mathbb{F}_2\right)\cong S_3\,,$$ where $S_3$ is the symmetric group on $3$ symbols.


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.