1.2. Permutation Groups.
One of the most important and well-known types of groups are the permutation groups, which we introduce formally as follows.
Consider a finite set of elements \(X = \{1, 2, \dots, n\}\). We define a permutation to be a reordering of the elements of \(X\). More formally, a permutation is a bijective mapping \(\sigma: X \to X\), similar to one that follows.
How can we represent this information? We generally don't use sets to represent permutations, since sets don't care about order. That is, \(\{1, 2, 3\} = \{3,2,1\}\), etc.
Thus for a set \(\{1, 2, \dots, n\}\), we can represent a permutation \(\sigma\) of the set of elements as follows:
where we read this as \(1\) is assigned to \(\sigma(1)\), 2 is assigned to \(\sigma(2)\). For example, a permutation that just shifts the elements down the line is
That is, \(\sigma\) sends 1 to 2, 2 to 3 and eventually \(n\) to 1. Here we'll denote the set of all permutations of the set \(\{1, 2, \dots, n\}\), or more generally a set of \(n\) objects (since we can always enumerate objects with natural numbers) as \(S_n\).
\textcolor{NavyBlue}{What is interesting about this is that if we define "multiplication" of elements of \(S_n\) to be function composition, then the set of permutations of \(X\) form a group which we show as follows.}
Let \(X = \{1, 2, \dots, n\}\).
- Closed. For any \(\sigma_1, \sigma_2 \in S_n\), we see that \(\sigma_2 \circ \sigma_1\) is (1) a composition of bijective functions and therefore is bijective and (2) a permutation of \(X\). One way to think about the composition is that \(\sigma_1\) rearranges \(X\), and \(\sigma_2\) rearranges \(X\) again. Thus \(\sigma_2 \circ \sigma_1 \in S_n\).
- Associativity. Associativity is obvious since function composition is associative.
- Identity. Observe that the permutation \(\sigma_e: X \to X\) for which \(\sigma_e(i) = i\) is technically a permutation of \(X\). Therefore \(\sigma_e\) acts as the identity element in \(S_n\).
- Inverse. Consider a permutation \(\sigma \in S_n\). Define \(\sigma^{-1}\) to be the function where if \(\sigma(i) = j\), then \(\sigma^{-1}(j) = i\). Then by construction, (1) \(\sigma^{-1}\) is a permutation of \(X\) and (2) \(\sigma \circ \sigma^{-1} = \sigma^{-1} \circ \sigma = \sigma_e\). Thus \(S_n\) contains inverses and composition of the inverses returns \(\sigma_e\), the identity.
For all \(n \ge 1\), \(|S_n| = n!\)
This is counting the number of ways to rearrange a set of size \(n\), which we know from combinatorics to simply be \(n!\)
Now that we know that \(S_n\) is a group, we'll study the properties of this group.
Recall earlier our notation for representing a permutation \(\sigma \in S_n\):
This notation sucks, since it includes more information than we actually need to. For instance, the top row is always going to be the same.
A better way to write this is through cycle decomposition, which we will soon define.
Let \(\sigma \in S_n\) and suppose \(X = \{1, 2, \dots, n\}.\) Suppose that there exists a subset \(\{n_1, n_2, \dots, n_k\}\) of \(X\) such that
Then \(\{n_1, n_2, \dots, n_k\}\) is called a cycle, and we denote this cycle as
We then read this as "\(n_1 \to n_2\), \(n_2 \to n_3, \dots, n_k \to n_1\)".
\textcolor{Blue}{Why do we care about cycles?} \ Well, consider an arbitrary cycle $ \sigma = \begin{pmatrix} n_1 & n_2 & \cdots & n_k \end{pmatrix}.$ Then again, \(\sigma(n_1) = n_2, \sigma(n_2) = n_3, \dots, \sigma(n_k) = n_1.\) However, what this is really saying is that
However, also take a note to observe that
More generally, we see that \textcolor{blue}{the element \(\sigma \in S_n\) has order \(n_k\)}, which is why the cycle length is \(k\).
\textcolor{Blue}{We care about cycles} since, given the fact that \(S_n\) is always a finite group, each of its elements will have finite order. Thus, in some way, we can always represent the elements of \(S_n\) in this form. \ \ More definitions. \ If $ \begin{pmatrix} n_1 & n_2 & \cdots & n_k \end{pmatrix}$ and $ \begin{pmatrix} n'_1 & n'_2 & \cdots & n'_k \end{pmatrix}$ share no elements in common, i.e.,
then the cycles are defined as disjoint cycles. \
Note that if \(\sigma(i) = i\) for some \(i \in X\), then this is technically a cycle and we represent the cycle as $ \begin{pmatrix} i \end{pmatrix}.$ In this case, we say that \(\sigma\) fixes \(i\). \
For example, suppose we have a permutation \(\sigma \in S_5\) where \(\sigma(1) = 2, \sigma(2) = 4, \sigma(4) = 2\). Then we have a cycle of length 4 and we denote this as
Since \(\sigma \in S_5\), suppose further that \(\sigma(3) = 5\) and \(\sigma(5) = 3\). Then we see that we have another cycle, disjoint with the previous cycle, and we write this one as
To write the entire permutation, we then can then express \(\sigma\) as
which gives us all the information we need to know on how \(\sigma\) rearranges the elements of \(X\). Such a representation of a permutation is called a disjoint cycle decomposition. It will turn out that we can actually express every permutation \(\sigma \in S_n\) in a product of disjoint cycles. Remark. In general, 1-cycles are omitted in the representation of a disjoint cycle decomposition. Thus if we have a permutation \(\sigma \in S_3\) such that \(\sigma(1) = 2\), \(\sigma(2) = 1\) and \(\sigma(3) = 3\), then we would write this as
Such a statement leads us to conclude that \(\sigma(3) = 3\). And if \(\sigma \in S_5\), we would furthermore conclude that not only \(\sigma(3) = 3\), but also \(\sigma(4) = 4\) and \(\sigma(5) = 5\). \ \ Nonuniqueness. One thing to note is that cycles are not unique. For example, we could have written the cycle $\textcolor{ForestGreen}{\begin{pmatrix} 1 & 2 & 4 \end{pmatrix}} $ as \(\textcolor{OrangeRed}{\begin{pmatrix} 2 & 4 & 1 \end{pmatrix}}\) or \(\textcolor{Cyan}{\begin{pmatrix} 4 & 1 & 2 \end{pmatrix}}\), since the other expressions still capture the fact that 1 is sent to 2, 2 is sent to 4, and 4 is sent to 1.
Note that the colors correspond to where the cycle starts. Clearly in the diagram, there are three ways to start the cycle, and hence why there are three nonunique representations for the cycle. \ More generally, for any cycle $ \begin{pmatrix} i_1 & i_2 & \cdots & i_n \end{pmatrix}$ we have that