9-9-2012

The Class Equation

If X is a G-set, X is partitioned by the G-orbits. So if X is finite,
X
|X| = |Gx|.
x∈X/G

(“x ∈ X/G” means you should take one representative x from each orbit, and sum over the set of
representatives. This is different from the notation “K ∈ X/G” which occurred in the proof of Burnside’s
Theorem; there I was actually summing over the set of orbits.)
x is a fixed point if gx = x for all g ∈ G. In this case, x is an orbit all by itself. Recall that X G
denotes the set of fixed points of X. I can pull these one-point orbits out of the sum above to obtain
X
|X| = |X G | + |Gx|.
x∈Σ

In this case, Σ will denote the set of orbits which have more than one point. (I’ll call them nontrivial
orbits for short.) Thus, “x ∈ Σ” means that you should take one representative x from each nontrivial
orbit.

Example. Here is a set X of 9 points arranged in a square:

1 2 3
4 5 6
7 8 9

Let the group D4 of symmetries of the square act on X. There is one fixed point: 5. There are two
nontrivial orbits: {1, 3, 7, 9} and {2, 4, 6, 8}. The equation above says

9 = 1 + (4 + 4).

Lemma. If G is a group, G acts on itself by conjugation:

g ⋆ x = gxg −1 for g, x ∈ G.

Proof. 1 ⋆ x = 1 · x · 1−1 = x for all x ∈ G. And if a, b, c ∈ G, then

(ab) ⋆ x = (ab)x(ab)−1 = abxb−1 a−1 = a(b ⋆ x)a−1 = a ⋆ (b ⋆ x).

Therefore, conjugation is an action.

Definition. The orbit of an element of G under the conjugation action is called a conjugacy class.
Remark. I proved earlier that if x and y are two points in the same orbit, there is a g ∈ G such that g ⋆x = y.
It follows that any two elements in a conjugacy class are conjugate.

Example. Let G = S3 . The conjugacy classes are

{id}, {(1 2), (1 3), (2 3)}, {(1 2 3), (1 3 2)}.

1
 (Note that if x, y ∈ S3 and x and y are conjugate, then x and y are either both even or both odd.)

Definition. Let G be a group, and let x ∈ G. The centralizer of x is

C(x) = {g ∈ G | gxg −1 = x}.

Lemma. Let G act on itself by conjugation, and let x ∈ G. Then Gx = C(x), the centralizer of x.
Proof. Fix x ∈ G. When is g ∈ Gx ? g ∈ Gx means g fixes x, i.e. gxg −1 = x. This is equivalent to gx = xg,
which in turn is what it means for x to be in C(x).

Corollary. Let G act on itself by conjugation, and let x ∈ G. Then:

(a) |Gx| = (G : C(x)).
(b) The number of elements in each conjugacy class divides the order of G.
Proof. For (a), recall that the order of the orbit equals the index of the isotropy group, and I just showed the
the isotropy group of x is C(x). Then (b) is immediate, since |G| = (G : C(X))|C(x)|, i.e. |G| = |Gx||C(x)|.

Corollary. Let G act on itself by conjugation. Then GG = Z(G).

(The horrible notation GG means the elements of G fixed under the conjugation action. I promise not
to do it again.)
Proof. The orbit of a fixed point consists of the point itself, so |Gx| = 1. This is equivalent to (G : C(x)) = 1,
or G = C(x). This is the same as saying that everything commutes with x, i.e. x ∈ Z(G).
I’ll apply these results to the equation
X
|X| = |X G | + |Gx|.
x∈Σ

In this case, the G-set is X = G. The set of fixed points is the center: X G = Z(G). The nontrivial
orbits Σ consist of conjugacy classes with more than one element. The order of such a conjugacy class is
(G : C(x)), where C(x) is the centralizer of x. (I must take one such x for each nontrivial orbit.) Therefore,
X
|G| = |Z(G)| + (G : C(x)).
x∈Σ

This is called the class equation.

Example. Let G = S3 . The center of S3 is Z(S3 ) = {id}, and the nontrivial conjugacy classes are

{(1 2), (1 3), (2 3)}, {(1 2 3), (1 3 2)}.

Take one element from each nontrivial class: say (1 2) and (1 2 3). Their centralizers are

C ((1 2)) = {id, (1 2)} and C ((1 2 3)) = {id, (1 2 3), (1 3 2)}.

Thus,
|Z(S3 )| + (S3 : C ((1 2))) + (S3 : C ((1 2 3))) = 1 + 3 + 2 = 6 = |S3 |.

2
Example. Consider the group Q = {±1, ±i, ±j, ±k} of quaternions. Here are the conjugacy classes:

central elements: {1}, {−1}

nontrivial conjugacy classes: {i, −i}, {j, −j}, {k, −k}

I found these by direct computation. For example, the conjugates of i are:

1 · i · (−1) = −i (−1) · i · 1 = −i

i · i · (−i) = i (−i) · i · i = i
j · i · (−j) = −i (−j) · i · j = −i
k · i · (−k) = −i (−k) · i · k = −i
Hence, {i, −i} is a complete conjugacy class.
In the class equation, the sum is taken over a set of representatives Σ for the nontrivial conjugacy
classes. In this example, for instance, I could take Σ = {i, j, k} or Σ = {−i, −j, k} or . . . — i.e. any three
elements, one from each of the (nontrivial) classes {i, −i}, {j, −j}, {k, −k}.
Finally, note that |Z(Q)| = 2, and the nontrivial classes have 2, 2, and 2 elements, respectively. The
class equation says (correctly) that
2 + (2 + 2 + 2) = 8.

Before I prove the next result, I’ll review some things that are apparent from the derivation of the class
equation. I had G acting on itself by conjugation. Then an orbit Gx (which does not mean “G times x” in
this case!) is the conjugacy class of x — i.e. the set of elements conjugate to x. Assuming that G is finite,
the order of the orbit equals the index of of the isotropy group, so |Gx| = (G : C(x)), where C(x) is the
centralizer of
Px — the set of elements which commute with x.
In the x∈Σ (G : C(x)) term, I’m summing over conjugacy classes with more than one element. This
implies two things:

(a) (G : C(x)) > 1, because (G : C(x)) equals the number of elements in the conjugacy class.

(b) |C(x)| < |G|, since |G| = (G : C(x))|C(x)|, and (G : C(x)) > 1.

Definition. Let p be prime. G is a p-group if for all g ∈ G, g 6= 1, the order of g is pn for some n > 0. (n
may be different for different elements.)

Finite p-groups are an important class of finite groups. Their structure is described in some detail
by Sylow theory. Sylow theory is also an important tool in determining the structure of arbitrary finite
groups. For example, it provides tools that helps answer questions like: How many different groups of order
20 are there?
As applications of the class equation, I’ll look at some results that are “preliminaries” to Sylow theory.
The first result says that if p is prime and p | |G|, then G has an element of order p. Note that if n is not
prime, n | |G| does not imply the existence of an element of order n. An easy counterexample: |Z3 × Z3 | = 9,
which is divisible by 9, but Z3 × Z3 has no elements of order 9.

Lemma. Let G be a finite abelian group, and let p be a prime number which divides |G|. Then G has an
element of order p.

Proof. Induct on |G|. Observe that Z2 has an element of order 2 and Z3 has an element of order 3 (and
those are the only groups of order 2 and 3, respectively). This gets the induction started.
Now suppose n > 3, and assume the result is true for abelian groups of order less than n. Let |G| = n,
p | n. I want to show that G has an element of order p.

3
 Let x ∈ G, where x 6= 0. If x has order pk for some k, then kx has order p. Therefore, assume that the
order of x is not divisible by p.
Now hxi 6= G, else x has order n, and n is divisible by p by assumption. Now

p | |G| = (G : hxi)|hxi|.

Since p 6 | |hxi|,  
 G 
p | (G : hxi) = 
 .
hxi 
Now |G/hxi| < |G| since |hxi| > 1, so G/hxi is an abelian group of order less than |G| whose order is
divisible by p. By the induction hypothesis, G/hxi contains an element y + hxi of order p, where y ∈ G.
Suppose y has order m. Then

G
my = 0, my + hxi = hxi, m(y + hxi) = hxi = 0 ∈ .
hxi

Since m kills y + hxi, it follows that m is divisible by the order p. Suppose m = kp. Then ky has order
p.
This shows that G has an element of order p and completes the induction.
Theorem. (Cauchy) Let G be a finite group, and let p be a prime number which divides |G|. Then G has
an element of order p.
Proof. Induct on |G|. Observe that Z2 has an element of order 2 and Z3 has an element of order 3 (and
those are the only groups of order 2 and 3, respectively). This gets the induction started.
Now suppose n > 3, and assume the result is true for groups of order less than n. Let |G| = n, p | n. I
want to show that G has an element of order p.
Consider x ∈ Σ, so the conjugacy class of x contains more than one element. By the remarks preceding
the lemma, |C(x)| < |G|. By induction, if p | |C(x)|, then C(x) has an element of order p. Since C(x) < G,
I’ll then have an element of order p in G.
Therefore, assume that p 6 | |C(x)|, for all x ∈ Σ. Now

p | |G| = (G : C(x))|C(x)|.

Hence, p | (G : C(x)) for all x ∈ Σ. Look at the class equation

X
|G| = |Z(G)| + (G : C(x)).
x∈Σ

p divides each term in the summation, and p divides |G| by assumption. Therefore, p | |Z(G)|. However,
Z(G) is an abelian group, so the Lemma shows that it has an element of order p. Since Z(G) < G, G has
an element of order p as well.

Corollary. Let p be prime. The order of a finite p-group is pn for some n > 0.
Proof. If G is a finite p-group, every element has order equal to a power of p. If |G| 6= pn for some n,
then q | |G|, where q is a prime number and q 6= p. By Cauchy’s theorem, G has an element of order q,
contradicting the fact that every element of a p-group has order equal to a power of p. Therefore, |G| = pn
for some n.
Proposition. Let p be prime. The center of a nontrivial finite p-group is nontrivial.

Proof. Examine the class equation

X
|G| = |Z(G)| + (G : C(x)).
x∈Σ

4
 Since |G| = pn for some n > 0, p | |G|. If x ∈ Σ, the remarks preceding the lemma above show that
(G : C(x)) > 1. Since pn = |G| = (G : C(x))|C(x)|, it follows that p | (G : C(x)). Now p divides |G| and
every term in the summation, so p divides |Z(G)|. Therefore, |Z(G)| 6= 1.
Proposition. Let G be a group of order pn , where p is prime. Then G has a subgroup of order pk for
0 ≤ k ≤ n.
Proof. If n = 1, then G = Zp , which plainly has subgroups of orders p0 and p1 .
Assume the result is true for groups of order pn−1 , and suppose |G| = pn . The center Z(G) of G is
nontrivial; since |Z(G)| | |G| = pn , it follows that |Z(G)| = pj for some j. By an earlier result, Z(G) contains
an element x of order p.
Now x ∈ Z(G), so hxi ⊂ Z(G). Hence, hxi is normal. Form the quotient group G/hxi. Now |G/hxi| =
pn−1 , so by induction G/hxi contains subgroups of order pk for 0 ≤ k ≤ n − 1. Each such subgroup has the
form Hk /hxi, where Hk is a subgroup of G containing hxi.
Suppose then that |Hk /hxi| = pk . I have

|Hk | = |Hk /hxi||hxi| = pk · p = pk+1 .

Thus, H0 , H1 , . . . Hn−1 are subgroups of G orders p, p2 , . . . pn . Obviously, G contains a trivial subgroup

of order p0 . This completes the induction step, and proves the result.

c 2012 by Bruce Ikenaga 5