next up previous
Next: The Frobenius-Cauchy lemma Up: No Title Previous: Group actions

Simplicity of the Alternating group



Denote by Sn the set of all permutations on a set with n elements, which we take without loss to be $\{1,2,\ldots ,n\}.$ Under the composition of permutations Sn is a group, called the Symmetric group of degree n.

Let Sn act on the set of polynomials with integral coefficients; specifically, $g\in S_n$ sends $Q(x_1,\cdots ,x_n)$ to $Q(x_{
g(1)},\ldots ,x_{g(n)})$. The orbit of $P=\prod_{i<j}(x_i-x_j)$ under this action is clearly $\{
P,-P\}.$ Thus the stabilizer of P in Sn is a subgroup of index 2; call it An. Note that the transposition (ij) sends P to -P, for all $i\neq j
;$ all transpositions lie therefore outside An.

Represent Sn on the left cosets of An by left translation. Denote the cosets by $A_n=\bar {1}$ and $\bar {A}_n=\bar {2}.$ The kernel of this representation is
$\{g\in S_n;gA_n=A_n,g\bar {A}_n=\bar {A}_n\}=A_
n.$
We conclude that An is normal in Sn. To summarize, this representation sends $g\in A_n$ into $(\bar {1})(\bar {2})$ and $
g\in\bar {A}_n$ into $(\bar {1}\bar {2}).$

If $g_1,\ldots ,g_m\in\bar {A}_n$, then $g_1\cdots g_m\rightarrow
(\bar {1}\bar {2})\cdots (\bar {1}\bar {2})=(\bar {1}\bar {2}
)^m.$ Clearly $(\bar {1}\bar {2})^m=(\bar {1})(\bar {2})$ if and only if m is even. Thus $g_1\cdots g_m\in A_n$ if and only if m is even. A cycle $(12\cdots k)=(1k)\cdots (12)$ decomposes into k-1 transpositions. Therefore a cycle of length k is in An if and only if k is odd. By the decomposition into disjoint cycles of any element of S_ n we conclude as follows:

* A permutation belong to An if and only if it has an even number of cycles of even length.

Elements of An are called even permutations. Elements of $
\bar {A}_n$ are called odd permutations. We call An the alternat ing group of degree n.

A group is simple if it has no normal subgroups other that itself and 1.

Theorem (Simplicity of ${\bf A}_n)$ The alternating group of degree n is simple; $n\neq 4.$

Proof: The group A3 is easily seen to be simple. When n=4 we have the normal subgroup 1, (12)(34), (13)(24), (14)(23) in A4, so A4 is not simple. Assume $n\geq 5.$

Step 1 An is generated by the 3-cycles. [In fact An is generated by the n-2 3-cycles $\{ (12k):k\geq 3\}.]$

Indeed, any element of An is a product of transpositions of the form (ab)(cd) or (ab)(ac). Since (ab)(cd)=(acb) (acd) and (ab)(ac)=(acb) we conclude that An is generated by the 3-cycles. Furthermore, (1a2)=(12a)-1, (1ab)=(12b)(12a)-1, (2ab)=(12b)- 1(12a), and (abc)=(12a)-1(12c)(12b)-1(12a), which shows that every 3-cycle is generated by a cycle of the form (12k).

Step 2 If H is a normal subgroup of An and H contains a 3-cycle, then H=An.

Without loss $(123)\in H.$ Then $(12k)=((12)(3k))(123)^{-1}((12)(3k))^{-1}=((123)^{-1})^{(12
)(3k)}\in H,$ by normality. Thus An=<(12k): all $k\geq 3>\leq H,$ and H=An.

We now assume that $H(\neq 1$) is normal in An. The idea is to show that H contains a 3-cycle, necessarily. We then invoke Step 2 to conclude the proof. Consider the exhausting possibilities examined below.

Case 1. Without loss assume that H contains $\sigma
=(12\cdots r)\tau ,$ where $r\geq 4,$ and $\tau$ is disjoint of $\{1,2,\ldots ,r\}.$ Then, by the normality of H, H contains $\sigma^{-1}\sigma^{(123)}=(12r).$ We are done by Step 2.

Case 2. Assume without loss that $\sigma =(123)(456
)\tau\in H,$ where $\tau$ is a product of disjoint transpositions. Then H contains $\sigma^{
-1}\sigma^{(124)}=(14263).$ We are done by Step 1.

Case 3. Assume without loss that $(123)\tau\in H,$ with $\tau$ a product of disjoint transpositions. Then H contains $(123)\tau (123)
\tau =(132),$ and we are done by Step 2.

Case 4. Assume that H contains elements that are products of disjoint transpositions. Without loss let $(12)(34)\tau$ be such an element of H. Then $(12)(34)\tau ((12)(34)\tau )^{(123)}=(13)(24)\in H.$ Since $
n\geq 5,$ consider $(123)\in A_n.$ By normality H contains (13)(24)((13)(24 ))(135)=(135), and we are done by Step 2. This ends the proof.



next up previous
Next: The Frobenius-Cauchy lemma Up: No Title Previous: Group actions
Gregory Constantine
1998-09-01