2

I have what seems to me a very simple proof that $A_{n}$ is the only subgroup of $S_{n}$ of index 2. Since I've seen other people prove it with what feel like really complicated methods (Like here.), I'm wondering if I've overlooked something.

Proof: Let $H$ be any subgroup of index 2 in $S_{n}$. Then $H\cap A_{n}$ is a normal subgroup of $A_{n}$ and since $A_{n}$ is simple, $H$ is either the trivial or the improper subgroup of $A_{n}$. If it's the trivial subgroup it doesn't have index 2 in $S_{n}$, and otherwise the theorem is proved.

Addem
  • 5,656

4 Answers4

2

As suggested, I am writing the complete answer.

First we use the not so simple result:

Theorem: Let $n = 3$ or $n\geq 5$. Then $A_n$ is simple.

Statement: Now for $n=3$ or $n \geq 5$ let's show that $\{id\}, A_n$ and $S_n$ are the only normal subgroups of $S_n$.In particular, $A_n$ is the only sugbroup of $S_n$ of index $2$.

Proof: It is clear that $\{id\}, A_n$ and $S_n$ are normal subgroups of $S_n$. Now let $H$ be a normal subgroup of $S_n$ and consider the group homomorphism $$\psi: H \to \{-1,+1\}$$

defined by $$\psi(\alpha)= \begin{cases}1&, \text {if}\ \ \alpha\ \ \text{is even}\\-1&, \text {if}\ \ \alpha\ \ \text{is odd}\end{cases} $$

Naturally, $\ker \psi = H \cap A_n$ and $(H: \ker \psi) = |\psi(H)| = 1$ or $2$. Thus, $(H: H \cap A_n) = 1$ or $2$.

  • $1^{st}$ case:

$(H: H\cap A_n) = 1$, that is, $H \subset A_n$. As $H \lhd S_n$, then a fortiori $H \lhd A_n$, then it follows from the theorem that $H=\{id\}$ or $H=A_n$.

  • $2^{nd}$ case:

$(H:H\cap A_n) = 2$. As $H \lhd S_n$, then $H \cap A_n \lhd A_n$, then from the theorem, $H \cap A_n = \{id\}$ or $H \cap A_n = A_n$. Therefore $|H| =2$ or $|H| = S_n$.

Let's suppose $|H|=2$. As $H \cap A_n = \{id\}$, then $H$ contains an odd permutation $\tau$ of order $2$. Such permutation $\tau$ is necessarily a product of disjoint transpositions, say $\tau = (12)\rho_{2}\ldots\rho_{s}$. Then

$$\tau ' = (13)\tau(13)^{-1} \in H , \text{because}\ \ \tau \in H \lhd S_n$$

As $$\tau ' (2) = [(13)\tau(13)](2) = [(13)\tau](2) = [(13)](1) =3$$

and $\tau(2) =1$, we obtain $\tau \neq \tau '$ and it follows that $|H| \geq 3$. So assuming that $|H| = 2$ leads us to a contradiction and then $H = S_n$.

Aaron Maroja
  • 17,571
  • If you want to be that precise about $;A_n;$ being simple and include (trivial simple) cyclic of order prime groups and the trivial group, then also $;A_1,,;A_2;$ are simple. – Timbuc Nov 25 '14 at 12:20
  • This is just the definition of $A_n$. It doesn't answer the question. – user1729 Nov 25 '14 at 16:11
  • 1
    @user1729 I don't understand the downvote, $\psi$ is a surjective homomorphism, and as $\beta$ can be written as a product of transpositions it's valid the associotion. Now if any subgroup $H < S_n$ has index two, then it's normal. Thus we may speak of quotients and we would have that $S_n/H$ is isomorphic to ${-1,+1}$ where $H$ is the kernel of $\psi$. Well, $H=A_n$. It's a hint by the way, it wasn't meant to give the complete answer. And it happens to be a good hint. Your lack of judgement was quite surprising given your background. I'd like to understand though what is so off. – Aaron Maroja Nov 25 '14 at 20:50
  • 1
    @Aaron I still do not understand your hint. You seem to be saying that "there is only one homomorphism from $S_n$ to $\mathbb{Z}2$ (and this is it!), hence only one subgroup of index 2". But this statement is not exactly simpler! It is true here (which is the point of the question), but not in general. For example, $\mathbb{Z}{2n}\times\mathbb{Z}_{2m}$ ($m, n\geq1$) contains precisely three subgroups of index two, as does $\mathbb{Z\times Z}$. So...perhaps you could explain more? – user1729 Nov 26 '14 at 12:48
  • @AaronMaroja That's okay, I didn't take it as rude! However, your idea doesn't work. Just think about $\mathbb{Z_2\times Z_2}$: it has 3 homomorphisms on to $\mathbb{Z}_2$, not just one. By your logic, a group can only have a single subgroup of index two, but this is not true in general (more examples are in my previous comment). Does that make sense? – user1729 Nov 26 '14 at 13:15
  • But this does not show uniqueness! I think you would be best to type out your idea in full, as I still am not getting it. – user1729 Nov 26 '14 at 16:10
  • @user1729 Anything to add? – Aaron Maroja Nov 26 '14 at 18:34
  • I still do not understand this answer, in the sense that it is (fundamentally) the same answer that the OP gave. Basically, you switch viewpoint (which I presume is the point of the original hint), but this switch does not result in a neater proof (in fact it is much more complicated than the OPs!). – user1729 Nov 27 '14 at 09:39
2

The following idea is super-simple. However, it only works if $n$ is odd (so it works precisely half the time!). But as it is so simple I thought it would be nice to record it here.

Suppose $n$ is odd. Note that $S_n$ can be generated by the elements $\alpha:=(1, 2)$ and $\beta:=(1, 2, \ldots, n)$. Now, every subgroup $H$ of index 2 is the kernel of some homomorphism $S_n\rightarrow \mathbb{Z}_2$. The key point is the following:

Homomorphisms are defined by the images of the generators. As $\beta:=(1, 2, \ldots, n)$ has odd order, it is killed by every homomorphism $\phi:S_n\rightarrow \mathbb{Z}_2$, so $\beta\in\ker\phi$ for all such $\phi$.

Therefore, if $\phi$ has non-trivial image it must be precisely the map defined by $\alpha\mapsto 1$, $\beta\mapsto0$. Hence, there is a unique homomorphism $\phi:S_n\rightarrow \mathbb{Z}_2$, and hence $S_n$ contains a unique subgroup of index 2 (for $n$ odd).

user1729
  • 31,015
1

Your proof looks simple because you assumed not so simple result that $A_n $ is simple for $n\geq 5$...

Actually something more is true...

Suppose that $H\leq S_n$ of index $m $ with $m< n$ then we have homomorphism $\eta: S_n \rightarrow S_m$.

As $Ker(\eta)$ is a normal subgroup of $S_n$ we should have $Ker(\eta)=1$ or $Ker(\eta)=A_n$.

Suppose $Ker(\eta)=(1)$ then we should have $S_n$ embedded in $S_m$, which is not possible as $m<n$.

So,$Ker(\eta)=A_n$ and we know that $Ker(\eta)\subset H$ i.e., $A_n\leq H< S_n$.

As $A_n$ is maximal subgroup of $S_n$ we have $H=A_n$.

So.. Do you see what i am concluding??

0

Assume that $A_n$ is a simple group for $n\neq 4$. If $H$ is any subgroup of index $2$ in $S_n$, then $H$ must be normal in $S_n$ (because the only non-trivial coset in $S_n/H$ must be the set-theoretic complement of $H$, and must therefore be both the right and left coset obtained by multiplying on the right, or the left by any particular $g\notin H$), and so $(H\cap A_n)$ must be a normal subgroup of $A_n$. Since $A_n$ is simple for $n\neq 4$, it follows that $(H\cap A_n)$ is either equal to $\{1\}$ (the subgroup of $S_n$ consisting only of the identity element) or $A_n$ itself, these two possibilities being mutually exclusive. If $(H\cap A_n)=\{1\}$, then ${\it every}$ element of $H$ (except for the identity permutation) must be an odd permutation. Furthermore, since $H$ has index $2$ in $S_n$, $|H|={\frac{n!}{2}}> 2$ for $n\geq 3$. Let $\sigma$ and $\tau$ be any two distinct odd permutations in $H$. Then, since ${\sigma}^2$, and ${\sigma}{\tau}$ are both even permutations in $H$, we would be forced to conlcude that both are equal to the identity permutation which is the only even permutation in $H$ by hypothesis. This in turn forces us to conclude that $\sigma = \tau$, which is a contradiction. Thus, $(H\cap A_n) = A_n$. In particular, $A_n\subset (H\cap A_n)$ and as both $A_n$ and $H$ have the same cardinality, they must be equal.

student
  • 1,324