Pólya Enumeration Theorem(II)

Pólya Enumeration Theorem(II)

Link to Part I

Orbit and stablizer

Let us go back to our necklace with 6 beads. We can now abstract it using concepts from group theory: we let the set of orderings of the 6 beads be $X$, and our permutation group $G$ will be acting on the set $X$. For every element $x\in X$, that is one of the specific ordering of the beads, a permutation $p\in G$ can translate the ordering to another ordering, and the set that contains all such orderings is known as the orbit, denoted by G ⋅ x = { p ⋅ x ∣ p ∈ G } G\cdot x=\lbrace p\cdot x|p\in G\rbrace G⋅x={p⋅x∣p∈G}. For the same $x$, some permutations are special: they map $x$ back to itself: For example, the permutation which represents rotate the necklace by 6 beads clockwisely would be one of such permutations, since every bead would have stayed in its original position after the permutation acted on $x$. The set of all such permutations in $G$ is known as the stablizer subgroup of $G$ with respect to $x$, denoted as G x = { p ∣ p ⋅ x = x } G_x=\lbrace p| p\cdot x=x\rbrace Gx​={p∣p⋅x=x}.

Orbit-Stablizer Theorem

The Orbit-Stabilizer Theorem states that

For a group $G$ acting on the set $X$, the order of $G$ equals to the product of the order of orbit of an element $x$ in $X$ and the order of the stablizer subgroup of $G$ with respect to $x$

In mathematical terms, we have $|G|=|G_x||G\cdot x|$.

To prove this, we need to first show that the stablizer subgroup of $G$ is actually a subgroup:

Firstly, permutations are obviously commutative.

Next, consider any 2 permutations in the set $G_x$, call them $p_1$ and $p_2$, we have
$p_1\cdot p_2\cdot x=p_1\cdot (p_2\cdot x)=p_1\cdot x=x$, thus $p_1\cdot p_2\in G_x$. Hence, $G_x$ is closed.

The identity element $I$ of $G$ is in $G_x$, since $I\cdot x=x$.

Lastly, for any permutation $p$ in $G_x$,
$p^{-1}\cdot x=p^{-1}\cdot (p\cdot x)=(p^{-1}\cdot p)\cdot x=I\cdot x=x$, hence $p_{-1}\in G_x$.

Hence, the stablizer subgroup $G_x$ is a subgroup of $G$.

Recall from our discussion about Langrange Theorem that for any subgroup of $G$, we can always partition $G$ into its distinct left(or right) cosets(including itself). Thus, we have $|G|=|G_x||G:G_x|$, where $|G:G_x|$ denotes the number of distinct left cosets of $G_x$ in $G$.

Hence, to prove the Orbit-Stablizer Theorem, we only need to show that
$|G:G_x|=|G\cdot x|$. We will show that a bijection exists between the set of left cosets of $G_x$ in $G$ and the orbit of $x$. For any left coset $p\cdot G_x$, consider the specific permutation element $p$ we use to generate the exact coset, $p$ acting on $x$ will give us an element in the orbit of $x$.

If we have $p_1\cdot x=p_2\cdot x$, then $p_1^{-1}\cdot p_1\cdot x=p_1^{-1}\cdot p_2\cdot x$, $p_1^{-1}\cdot p_2\cdot x=x$, thus we have $p_1^{-1}\cdot p_2\in G_x$. We know that $p_1\cdot (p_1^{-1}\cdot p_2)=p_2$ is in the left coset generated by $p_1$, and $p_2$ is in the left coset generated by itself since $p_2\cdot I=p_2$, hence $p_1$ and $p_2$ generate the same left coset of $G_x$ in $G$, i.e. they belong to the same left coset. Notice that this process can be totally worked backward to show that if two permutation elements are in the same coset, they would also give us the same element in the orbit of $x$ when acting on $x$. Thus, the bijection exists between the set of left cosets of $G_x$ in $G$ and the orbit of $x$, and by the Bijection Principle, $|G:G_x|=|G\cdot x|$. This completes our proof for the Orbit-Stablizer Theorem.

Burnside’s Lemma

We can finally get back to the problem we are started with: how do we count the number of distinct necklaces by arranging the six beads? Burnside’s lemma aims to solve this problem of counting the number of equivalence classes in sets. It states that:

For a permutation group $G$ acting on the set of element $x$, the number of orbits(equivalent classes) $|X\backslash G|=\frac{{\sum }_{p\in G}|X^p|}{|G|}$, where $|X^p|$ denotes the number of fixed points for the permutation $p$.

To prove this, instead of summing up all the fixed points for each of the permutation $p\in G$, we can instead sum up the size of the stablizer subgroup with respect to each element $x\in X$. In the case of our necklace, it means that we are now summing up the number of permutations that do not change the positions of the beads after acting on a specific order of the beads for each possible ordering of the beads. Hence, ${\sum }_{p\in G}|X^p|={\sum }_{x\in X}|G_x|$. and according to the Orbit-Stablizer Theorem, we can write this as ${\sum }_{x\in X}\frac{|G|}{|G\cdot x|}$.

Now let us consider the nature of these “equivalent classes”. We notice that, each of these “equivalent classes” contains elements that can generate one another after permutations in the permutation group acted on them. In the following example, after the permutation “rotate the necklace by one bead clockwisely” acted on the ordering of beads on the left, it is translated into the ordering of beads on the right.
$$\begin{array}{rl}{a}_1-a_2-a_3& a_6-a_1-a_2\\ ||& =>||\\ a_6-a_5-a_4& a_5-a_4-a_3\end{array}$$
These two orderings are thus considered as equivalent and are in the same “equivalent class”. Thus, the order of the orbit for any element $x$ in an equivalent class is equal to the size of the equivalent class $E$, since the equivalent class $E$ basically contains all the elements that can be generated by some permutation $p\in G$ acting on $x$. Thus by the definition of the orbit, we have $|G\cdot x|=|E|$.

Hence, by enumerating all the different equivalence classes, we have
$$\begin{array}{rl}\frac{{\sum }_{p\in G}|X^p|}{|G|}& =\frac{{\sum }_{x\in X}|G_x|}{|G|}\\ & =\sum _{x\in X}\frac{1}{|G\cdot x|}\\ & =\sum _{E\in X/G}\sum _{x\in E}\frac{1}{|G\cdot x|}\text{(enumerating equivalence classes)}\\ & =\sum _{E\in X/G}\sum _{x\in E}\frac{1}{|E|}\\ & =\sum _{E\in X/G}1\\ & =|X/G|\end{array}$$
This completes our proof for the Burnside Lemma.

Pólya Enumeration Theorem

The Pólya enumeration theorem is a generalization of Burnside’s lemma, and it also provides a more convenient tool for finding the number of equivalence classes since it is sometimes troublesome to find all the fixed points of a specific permutation, which is also hard to express in computer language. Here we will only be introducing a simple version of the theorem, which has a rather similar form of that of the Burnside Lemma. It states that:
$$|X/G|=\frac{1}{|G|}\sum _{p\in G}|m{|}^{C(p)}$$, where $m$ denotes the number of representation an element can take(for example, the 6 different types of beads we have), and $C(p)$ denotes the number of cycles in the permutation $p$.

For example, consider the permutation $p$ acting on our six beads:$$f_p=\left(\begin{array}{c}{a}_1,a_2,a_3,a_4,a_5,a_6\\ a_3,a_1,a_2,a_5,a_4,a_6\end{array}\right)$$
It can be decomposed into three cycles, namely $\left(\begin{array}{c}{a}_1,a_2,a_3\\ a_3,a_1,a_2\end{array}\right)$, $\left(\begin{array}{c}{a}_4,a_5\\ a_5,a_4\end{array}\right)$ and$\left(\begin{array}{c}{a}_6\\ a_6\end{array}\right)$.

The theorem follows directly from the Burnside Lemma, since the sufficient and necessary condition of the $x$ being an fixed point of the permutation $p$ would be that for each cycle, all the elements concerned must have the same representation(for example, for the first cycle, $(a_1,a_2,a_3)=(a_3,a_1,a_2$), hence $a_1,a_2,a_3$ must all have the same representation). Since cycles are independent from one another, by the multiplication principle, we have $|X^p|=|m{|}^{C(p)}$.

Thus, $|X/G|=\frac{{\sum }_{p\in G}|m{|}^{C(p)}}{|G|}$.