Notes on Abstract Algebra

2.1 Definitions and Examples of Groups

2.1.2 Groups
At this point we’ve basically beat associativity to death, so let’s get on with defining
a group in a precise way. As we’ve mentioned before, the benefit of working in such
generality is the fact that we will be able to unify all of our examples under one
umbrella. We can prove results about many different examples at once, rather than
having to consider many different cases.

Definition 2.1.7. A group is a set G together with a binary operation ∗ : G×G →

G satisfying.

1. Associativity: For all a, b, c ∈ G, we have

a ∗ (b ∗ c) = (a ∗ b) ∗ c.

2. Identity: There exists an element e ∈ G with the property that

e∗a=a∗e=a

for all a ∈ G.

3. Inverses: For every a ∈ G, there is an element a−1 ∈ G with the property

that
a ∗ a−1 = a−1 ∗ a = e.

Remark 2.1.8. To be concise, we’ll often write hG, ∗i to distinguish the operation
on G. If the operation is understood, we’ll just write G for the group.

Remark 2.1.9. This seems like a good place to make some remarks about how one
should read math. Reading a math book is an active process—you need to stop and
think about the material frequently, and it helps to have a pen and paper handy to
check things on your own. In particular, when you come across a definition, it is
extremely helpful to think about two things:

• Immediately try to think of examples that satisfy that definition.

• Think about why the definition is useful. Why are we bothering to make this
definition, and why does it say what it says?

We’ll address both of these points presently.

Before we start investigating properties of groups, it will be nice to have some

examples of groups to fall back on. We’ve already seen some in our motivating
discussion, and we’ll add in some others that may also be familiar (or perhaps less
familiar).
 Group Theory

Example 2.1.10. Here are some examples of groups.

1. hZ, +i is a group, as we have already seen.

2. hMn (R), +i is a group.

3. hZn , +n i is a group.

Here are some nonexamples.

4. hZ, ·i is not a group, since inverses do not always exist. However, h{1, −1}, ·i
is a group. We do need to be careful here—the restriction of a binary oper-
ation to a smaller set need not be a binary operation, since the set may not
be closed under the operation. However, {1, −1} is definitely closed under
multiplication, so we indeed have a group.

5. hMn (R), ·i is not a group, since inverses fail. However, hGLn (R), ·i is a group.
We already saw that it is closed, and the other axioms hold as well.

6. hZn , ·n i is not a group, again because inverses fail. However, hZ×n , ·n i will be
a group. Again, we just need to verify closure: if a, b ∈ Z× n , then a and b are
both relatively prime to n. But then neither a nor b shares any prime divisors
with n, so ab is also relatively prime to n. Thus ab ∈ Z× n.

Here are some other examples that might be less familiar.

Example 2.1.11. 1. Just as Z is a group under addition, so are

Q = {rational numbers}
R = {real numbers}
C = {complex numbers}.

These are all groups under addition, but not multiplication. Since Z (and each
of these groups) is only a group under addition, we’ll usually write Z for the
additive group hZ, +i.

2. Let Q× = Q − {0} = {a/b ∈ Q : a 6= 0}, and consider hQ× , ·i. I claim that
this is a group. Multiplication is certainly associative, and the identity is 1.
Given a rational number a/b ∈ Q× , the inverse is just a/b:
a a
· = 1.
b b

Similarly, we could define R× and C× , and these would be groups under mul-
tiplication (the inverse is simply a−1 = 1/a).
2.1 Definitions and Examples of Groups

3. Let Rn = {(a1 , . . . , an ) : a1 , . . . , an ∈ R} (the addition is done coordinatewise).

Then hRn , +i is a group. Associativity follows from that for addition of real
numbers, the identity is the zero vector, and the inverse is just the negative.
More generally, any vector space is a group—we simply “forget” about scalar
multiplication.
We’ll introduce many more examples as we move through the course.
Now that we’ve done some examples, let’s address the second point that we made
above. Why does the definition of a group look the way it does? We want to come
up with a general sort of object that subsumes the objects we’re used to (such as
number systems, vector spaces, and matrices), so the “set with a binary operation”
part makes sense. We already talked about why associativity is important. But why
do we require that there be an identity and inverses? It all comes down to solving
equations (after all, that is what “algebra” originally meant). Suppose we are given
a group G, and we write down the equation

a ∗ x = b,

for a, b, x ∈ G. How could we solve for x? Multiply on the left by a−1 :

a−1 ∗ (a ∗ x) = a−1 ∗ b =⇒ (a−1 ∗ a) ∗ ba−1 ∗ b

assoc.
=⇒ e ∗ b = a−1 ∗ b
inverse
=⇒ x = a−1 ∗ b.
identity

In short, groups are algebraic structures, and it is good to be able to solve equations
within them.
We’ll often break our study of groups down into different collections of groups
that satisfy certain properties. Therefore, let’s introduce a couple of adjectives for
describing groups. We mentioned before that commutativity is “optional” when
it comes to groups. Groups that have commutative operations are special, and
therefore they have a special name.
Definition 2.1.12. A group hG, ∗i is said to be abelian if ∗ is commutative, i.e.

a∗b=b∗a

for all a, b ∈ G. If a group is not commutative, we’ll say that it is nonabelian.

Abelian groups are named in honor of Niels Henrik Abel, who is mentioned in
the historical discussion at the beginning of these notes. Except for GLn (R), all
the groups we’ve seen so far are abelian. Soon we’ll see two interesting examples of
finite nonabelian groups, the symmetric and dihedral groups.
Another useful way to describe groups is via their size. Of course we have a
special name for it.
 Group Theory

Definition 2.1.13. The order of a group G, denoted by |G|, is the number of

elements in G.

If a group G has infinitely many elements, we will write |G| = ∞.3

Example 2.1.14. Most of the examples that we have seen so far are infinite groups.
In particular, |Z| = ∞.

We know lots of examples of infinite groups, but they will actually be hard to
understand, with the exception of Z. We’ll be more interested in finite groups.

Definition 2.1.15. A group G is said to be finite if |G| < ∞.

One of the reasons that we study finite groups is that it is much easier to analyze
their structure, and we’ll be able to classify many types of finite groups. Therefore,
we’d like to have lots of examples at our disposal. Of course, we’ve already seen two
examples of finite groups, namely the groups that arise from the study of modular
arithmetic.

Example 2.1.16. 1. For any n, the additive group Zn is a finite group, with
|Zn | = n.

2. Similarly, hZ×
n , ·i is a finite group. Its order is a little harder to determine. We
know that |Zn× | will be the number of a ∈ Zn which are relatively prime to n.

We will produce two more families of finite groups very soon, namely the sym-
metric and dihedral groups to which we have already alluded.

2.1.3 Group Tables

Now we’ll make a slight detour to talk about a tool for working with finite groups.
One of the things that makes finite groups easier to handle is that we can write
down a table that completely describes the group. We list the elements out and
multiply “row by column.”

Example 2.1.17. Let’s look at Z3 , for example. We’ll write down a “multiplication
table” that tells us how the group operation works for any pair of elements. As we
mentioned, each entry is computed as “row times column”:
2.1 Definitions and Examples of Groups

+ 0 1 2
0 0 1 2
1 1 2 0
2 2 0 1

(Of course we have to remember that “times” really means “plus” in this example.)
This is called a group table (or a Cayley table).

Group tables can get pretty tedious when the order of the group is large, but
they are handy for analyzing fairly small groups. In particular, let’s look at groups
of order 1, 2, and 3. If we have a group of order one, there is really only one thing
to do. The lone element must be the identity e, and e ∗ e = e, so the group table
must be:
∗ e
e e

Suppose next that we have a group of order two, so that the group contains the
identity e and some other element a with a 6= e. Then e ∗ a = a ∗ e = a, and since a
must have an inverse (and the only option is a), we must have a ∗ a = e. Therefore,
the group table must look like:

∗ e a
e e a
a a e

Finally, suppose we have a group of order three. Then there are three distinct
elements: the identity e, and two other elements a and b. The first row and first
column are easy to fill in, since e is the identity. We are then left to determine a ∗ a,
a ∗ b, b ∗ a, and b ∗ b. We claim first that it is impossible to have a ∗ a = e. If a ∗ a = e,
then b ∗ b = e as well, since b must have an inverse. The reason for this is that if
a = b−1 , so a ∗ b = e, then

a = a ∗ e = a ∗ (a ∗ b) = (a ∗ a) ∗ b = e ∗ b = b.

If a = b, then the group no longer has three elements, so this can’t be the case.
Therefore, b ∗ b = e. But now we must have either a ∗ b = a or a ∗ b = b. In the first
case,
b = e ∗ b = (a ∗ a) ∗ b = a ∗ (a ∗ b) = a ∗ a = e,
which can’t happen either. Similarly, a ∗ b 6= b. The hypothesis that got us into all
this trouble was that a ∗ a = e, so we can’t have a = a−1 . Also, a ∗ a 6= a; if it were,
then
e = a−1 ∗ a = a−1 ∗ (a ∗ a) = (a−1 ∗ a) ∗ a = e ∗ a = a,
 Group Theory

which is not true. Therefore, we must have a ∗ a = b instead. Since a must have an
inverse (and it must be b),
a ∗ b = b ∗ a = e.
Finally, b ∗ b = a, since

b ∗ b = b ∗ (a ∗ a) = (b ∗ a) ∗ a = e ∗ a = a.

Consequently, the group table must be:

∗ e a b
e e a b
a a b e
b b e a

Note that, in particular, there is really only one group of each order (1, 2, or 3), since
there was only one way that we could fill in each group table. (In language that we’ll
develop later, there is only one group of each of these orders up to isomorphism.)
Also, any group of order 1, 2, or 3 is abelian. To see this, we can simply note that
the group table is symmetric about the diagonal. It is an exercise to show that
groups of order 4 and 5 are abelian. The result does not extend to order 6—we will
see two groups of order 6 which are not abelian. As you can see, group tables can
be quite useful, and we will look at them again once we’ve defined the symmetric
and dihedral groups.

Exercise 2.1. Show that any group of order 4 is abelian. [Hint: Compute all
possible Cayley tables. Up to a reordering of the elements, there are two possible
tables.]

Exercise 2.2. Show that any group of order 5 is abelian. [Hint: There is only
one possible Cayley table, up to relabeling.]

2.1.4 Remarks on Notation

Now we’ll make one last note on working with general groups. We will get really
sick of writing ∗ all the time, so we will oftentimes suppress the operation ∗, and
simply write
ab = a ∗ b.
when working with arbitrary groups. If we are dealing with a specific example
(such as the integers under addition), we’ll still write the operation to make things
2.2 The Symmetric and Dihedral Groups

clear. In keeping with this convention, we’ll also have a shorthand way of denoting
a product of an element a with itself. We’ll write

a2 = a ∗ a,

and in general
an = a
| ∗ a ∗{z· · · ∗ a} .
n times

We can also define negative powers by

n
a−n = a−1 .

These notational conventions will make it much easier to write down computations
when we are working with groups. The point of all this is that computations in
an arbitrary group will behave very much like usual multiplication, except that
things may not always commute. (In fact, matrix multiplication might be the best
operation to think of.) Also we should note that if we are dealing with the group
Z, an really means
an = a
| +a+ {z· · · + a} = na.
n times

This goes to show that one has to be careful when writing down computations in a
group, since we usually write things multiplicatively.

2.2 The Symmetric and Dihedral Groups

We are now ready to introduce our first really interesting examples of (finite) non-
abelian groups, which are called the symmetric and dihedral groups. (These are
actually families of groups, one for each positive integer.) Not only will they give
examples of finite nonabelian groups, but we will see that their elements are quite
different from the examples that we have been considering so far.

2.2.1 The Symmetric Group

The symmetric group is one of the most important examples of a finite group, and
we will spend quite a bit of time investigating its properties. It will arise as a
special case of the set SX of bijections from a set X back to itself, so let’s begin
there.
Before we can proceed, we need some preliminaries on functions. We’ll need
some of these facts later on when we discuss homomorphisms, so we will work a
little more generally than is absolutely necessary right now.
 Group Theory

Let X and Y be nonempty sets. Recall that a function f : X → Y is, at least

informally, a rule which assigns to each element x ∈ X a unique element f (x) ∈ T :

x 7→ f (x).

We’re eventually going to look at a particular class of functions from a set to itself.
To do this, we’ll need to know what it means for a function f : X → Y to be
one-to-one (or injective) and onto (or surjective).

Definition 2.2.1. A function f : X → Y is one-to-one or injective if whenever

x1 , x2 ∈ X with f (x1 ) = f (x2 ), then x1 = x2 . Equivalently, if x1 6= x2 , then
f (x1 ) 6= f (x2 ), or different inputs always yield different outputs.

Example 2.2.2. 1. Let X = {1, 2, 3} (or any set with three elements), and
define f : X → X by f (1) = 2, f (2) = 1, and f (3) = 1. We can represent this
pictorially as

f
1 1

2 2

3 3

This function is not injective, since 2 and 3 both map to 1.

2. Define f : R → R by f (x) = x2 . This function is not injective, since f (1) =

f (−1) = 1, for example.

3. On the other hand, define g : R → R by g(x) = x. This function is injective.

Definition 2.2.3. A function f : X → Y is onto or surjective if for any y ∈ Y ,

there exists an x ∈ X such that f (x) = y.

An intuitive way of describing surjectivity is to say that every element of Y is

“hit” by f , or lies in the image of f .

Example 2.2.4. 1. The function f : {1, 2, 3} → {1, 2, 3} described in the previ-

ous example is not onto, since no element is mapped to 3 by f .

2. The function f : R → R given by f (x) = x2 is not onto.

3. If we define g : R → R by g(x) = x, then this is a surjective function.

Any function which is both one-to-one and onto is of particular importance, so

we have a special name for such objects.
2.2 The Symmetric and Dihedral Groups

Definition 2.2.5. A function f : X → Y is bijective (or simply a bijection) if f

is both one-to-one and onto.

The proofs of the following statements would all be good exercises to try on your
own, but we will reproduce them here anyway.

Proposition 2.2.6. Let X, Y , and Z be sets, and let f : X → Y and g : Y → Z be

functions.

1. If f and g are both one-to-one, then so is g ◦ f .

2. If f and g are both onto, then so is g ◦ f .

3. If f and g are both bijections, then so is g ◦ f .

Proof. Let’s start with the first one. Suppose that f and g are one-to-one, and
suppose that x1 , x2 ∈ X with g ◦ f (x1 ) = g ◦ f (x2 ). This means that

g(f (x1 )) = g(f (x2 )),

and since g is one-to-one, f (x1 ) = f (x2 ). But f is also one-to-one, so x1 = x2 . Thus

g ◦ f is one-to-one.
Now let’s handle the question of surjectivity. Suppose f and g are both onto, and
let z ∈ Z. We need to produce an x ∈ X such that g ◦ f (x) = z, i.e., g(f (x)) = z.
Since g is onto, there exists a y ∈ Y such that g(y) = z. Also, f is onto, so there
exists an x ∈ X such that f (x) = y. Putting this together, we get

g ◦ f (x) = g(f (x)) = g(y) = z,

so g ◦ f is onto.
The third statement follows from the first two. If f and g are bijective, then
they are both one-to-one and onto. But then g ◦ f is one-to-one by (a), and g ◦ f is
onto by (b), so g ◦ f is bijective.

Now let’s formally define SX , the set of bijections from X to itself. We will then
proceed prove that SX is a group under composition.

Definition 2.2.7. Let X be a set. We define:

SX = {f : X → X : f is a bijection}.

Note that Proposition 2.2.6(c) tells us that SX is closed under composition of

functions. In other words, composition defines a binary operation on SX . But
what kind of operation is it? Why, it’s associative! We already verified this in
Example 2.1.6(5). Unfortunately, it is not a commutative operation. To see this, let
X = {1, 2, 3}, and define f and g by the following diagrams:
 Group Theory

f g
1 1 1 1

2 2 2 2

3 3 3 3

Then
g ◦ f (1) = g(f (1)) = g(2) = 3,

but
f ◦ g(1) = f (g(1)) = f (2) = 1,

so g ◦ f 6= f ◦ g. Thus SX will provide a new example of a nonabelian group.

To finish checking that SX is a group, we need to verify the existence of an
identity and inverses. For the first one, recall that any set X has a special bijection
from X to X, namely the identity function idX :

idX (x) = x

for all x ∈ X. Note that for any f ∈ SX , we have

f ◦ idX (x) = f (idX (x)) = f (x)

and
idX ◦ f (x) = idX (f (x)) = f (x)

for all x ∈ S. Thus idX ◦ f = f ◦ idX = f for all f ∈ SX , so idX serves as an

identity for SX under composition. Finally, what do we know about any bijection?
Why, it has an inverse, in the sense that there is another function that “undoes” it.
More specifically, if f ∈ SX and y ∈ X, there is an x ∈ X such that f (x) = y (since
f is onto). But f is also one-to-one, so this x is unique. Therefore, we can define
f −1 (y) = x. You can then check that

f ◦ f −1 (y) = f (f −1 (y)) = f (x) = y = idX (y)

and
f −1 ◦ f (x) = f −1 (f (x)) = f −1 (y) = x = idX (x),

so f −1 really is an inverse for f under composition. Therefore, by making all of

these observations, we have established the following result:

Proposition 2.2.8. SX forms a group under composition of functions.

2.2 The Symmetric and Dihedral Groups

If X is an infinite set, then SX is fairly hard to understand. One would have to

be either very brave or very crazy to try to work with it. Things are much more
tractable (and interesting) when X is finite. Suppose then that X is finite, say
with n elements. It doesn’t really matter what X is; it only matters that X has
n elements. That is, we can label the elements of X to be whatever we want (say,
numbers, people, flavors of ice cream, etc.), so we could simply assume that X is
the set
X = {1, 2, . . . , n}.
In this case we call SX the symmetric group on n letters, and we denote it by
Sn .
An element of Sn is a bijection from {1, 2, . . . , n} → {1, 2, . . . , n}. In other words,
it rearranges the numbers 1, . . . , n. Pictorially, we could represent the bijection σ
of {1, 2, 3} defined by σ(1) = 2, σ(2) = 1, and σ(3) = 3 with the diagram

σ
1 1

2 2

3 3
 Group Theory

for representing a particular permutation. It would be annoying if we had to specify

σ(1), σ(2), . . . , σ(n) each time, so we’ll often write an element of Sn using two-line
notation:  
1 2 ··· n
σ=
σ(1) σ(2) · · · σ(n)
For example, suppose that σ ∈ S3 is given by the picture that we considered earlier,
i.e., σ(1) = 2, σ(2) = 1, and σ(3) = 3. Then we have
 
1 2 3
σ= .
2 1 3
Of course if we are going to represent permutations in this way, it would help to
know how multiplication works in this notation. As an example, let
 
1 2 3
τ=
2 3 1
Then remember that multiplication is really just composition of functions:
    
1 2 3 1 2 3 1 2 3
στ = =
2 1 3 2 3 1 σ(τ (1)) σ(τ (2)) σ(τ (3))

   
1 2 3 1 2 3
= =
σ(2) σ(3) σ(1) 1 3 2
On the other hand, what is τ σ?
    
1 2 3 1 2 3 1 2 3
τσ = =
2 3 1 2 1 3 3 2 1
In other words, one moves right to left when computing the product of two permu-
tations. First one needs to find the number below 1 in the rightmost permutation,
then find this number in the top row of the left permutation, and write down the
number directly below it. Repeat this process for the rest of the integers 2 through
n.
In the example above, note that στ 6= τ σ, so what have we shown? We have
actually verified that S3 is nonabelian. This holds more generally:
Proposition 2.2.10. For n ≥ 3, Sn is a nonabelian group.
Proof. Let σ, τ ∈ S3 be defined as in the example, and suppose that n > 3. Define
σ̃, τ̃ ∈ Sn by (
σ(i) if 1 ≤ i ≤ 3
σ̃(i) =
i if i > 3,
and similarly for τ˜. Then the computation that we performed in S3 shows that
σ̃τ̃ 6= τ˜σ̃, so Sn is nonabelian.
2.2 The Symmetric and Dihedral Groups

It would be entirely feasible to compute the group table for Sn , at least for
relatively small n. We’ll display the group table for S3 here, though we won’t
perform any of the computations explicitly. (There are 6 · 6 = 36 different products
to compute, and these are left to the interested reader.) Label the elements of S3
as follows:
   
1 2 3 1 2 3
ι= µ1 =
1 2 3 1 3 2
   
1 2 3 1 2 3
ρ1 = µ2 =
2 3 1 3 2 1
   
1 2 3 1 2 3
ρ2 = µ3 =
3 1 2 2 1 3

Then the group table for S3 is:

∗ ι ρ1 ρ2 µ1 µ2 µ3

ι ι ρ1 ρ2 µ1 µ2 µ3

ρ1 ρ1 ρ2 ι µ3 µ1 µ2

ρ2 ρ2 ι ρ1 µ2 µ3 µ1

µ1 µ1 µ2 µ3 ι ρ1 ρ2

µ2 µ2 µ3 µ1 ρ2 ι ρ1

µ3 µ3 µ1 µ2 ρ1 ρ2 ι

The labeling of the elements may look odd at this point, but we will see a good
reason for it quite soon.

2.2.2 The Dihedral Group

Now it’s time to talk about another interesting nonabelian group, the dihedral
group. Again, we’re actually going to be dealing with a family of groups, one for
each positive integer. Suppose we have a triangle, and we label the vertices 1, 2,
and 3:
 Group Theory

1 2

We can rotate the triangle counterclockwise by 120◦ and obtain a triangle with the
same orientation, albeit with the vertices relabeled:

3 2

1 2 3 1

We could also rotate by 240◦ :

3 1

1 2 2 3

Call these two transformations r1 and r2 , respectively. We could also reflect the
triangle across any of the angle bisectors to obtain a relabeling of the vertices:

3 2

1 2 1 3

Label these reflections m1 , m2 , and m3 . Define the identity transformation i to

be the one that simply leaves the triangle unmoved. The set

{i, r1 , r2 , m1 , m2 , m3 }

is called the set of symmetries of the triangle. They are exactly the transformations
of the triangle that reorder the vertices in a way that exploits the symmetry of the
equilateral triangle. This set forms a group under composition, called D3 , or the
third dihedral group. We can define more generally the nth dihedral group
Dn to be the set of symmetries of a regular n-gon.
2.2 The Symmetric and Dihedral Groups

Definition 2.2.11. The group Dn is the set of all symmetries of the regular n-gon
under composition of transformations.

Naturally, we could ask ourselves what the order of Dn should be. In general, there
will be n rotations (including the identity transformation) and n reflections, so

|Dn | = 2n.

Of course we should talk about how to multiply two symmetries of an n-gon.

Again, we need to think of multiplication as composition of functions. For example,
suppose we wanted to compute r1 m1 in D3 . We would first have to apply m1 to the
triangle, and then apply r1 :

3 2 1

m1 r1

1 2 1 3 2 3

We see from this picture that r1 m1 = r2 . By doing this for all possible pairs of
elements, we could write down the group table for D3 :

∗ i r1 r2 m1 m2 m3

i i r1 r2 m1 m2 m3

r1 r1 r2 i m3 m1 m2

r2 r2 i r1 m2 m3 m1

m1 m1 m2 m3 i r1 r2

m2 m2 m3 m1 r2 i r1

m3 m3 m1 m2 r1 r2 i

Observe that this table is the same as that of S3 . (This is why we labeled the
elements of S3 in the way that we did.) This essentially shows that S3 and D3 are
isomorphic groups: they are the same group dressed up in different disguises. In
this case, it is not too hard to see how one could identify D3 with S3 , since the
elements of D3 can be viewed as permutations of the vertices. In general, we can
view elements of Dn as permutations of the vertices of the regular n-gon, but we
 Group Theory

cannot realize all permutations in this way. In other words, we will see that Dn and
Sn are not the same. One way to see this at this point is to notice that |Dn | = 2n
but |Sn | = n!, and these are only equal when n = 3.

2.3 Basic Properties of Groups

Now that we’ve defined groups and we have some interesting examples in our toolk-
its, it’s time to start investigating properties of groups. We’ll start of with some
simple properties before we really get into the bigger questions regarding the struc-
ture and classification of groups. At the very least, this will simplify some of the
routine calculations that we need to make when working with groups.

Proposition 2.3.1. Let G be a group. The identity element e ∈ G is unique, i.e.,

there is only one element e of G with the property that

ae = ea = a

for all a ∈ G.

Proof. For this proof, we need to use the standard mathematical trick for proving
uniqueness: we assume that there is another gadget that behaves like the one in
which we’re interested, and we prove that the two actually have to be the same.
Suppose there is another f ∈ G with the property that

af = f a = a

for all a ∈ G. Then in particular,

ef = f e = e.

But since e is an identity,

ef = f e = f.
Therefore,
e = ef = f,
so e is unique.

The next result has to do with solving equations, which was our original moti-
vation for requiring that inverses exist. It’s called cancellation.

Proposition 2.3.2 (Cancellation laws). Let G be a group, and let a, b, c ∈ G. Then:

(a) If ab = bc, then b = c.

(b) If ba = ca, then b = c.

2.3 Basic Properties of Groups

Proof. (a) Suppose that ab = ac. Multiply both sides on the left by a−1 :

a−1 (ab) = a−1 (ac).

By associativity, this is the same as

(a−1 a)b = (a−1 a)c,

and since a−1 a = e, we have

eb = ec.

Since e is the identity, b = c. The same sort of argument works for (b), except we
multiply the equation on the right by a−1 .

The cancellation laws actually give us a very useful corollary. You may have
already guessed that this result holds, but we will prove here that inverses in a
group are unique.

Corollary 2.3.3. Let G be a group. Every a ∈ G has a unique inverse, i.e. to each
a ∈ G there is exactly one element a−1 with the property that

aa−1 = a−1 a = e.

Proof. Let a ∈ G, and suppose that b ∈ G has the property that ab = ba = e. Then
in particular,
ab = e = aa−1 ,

and by cancellation, b = a−1 . Thus a−1 is unique.

While we are on the topic of inverses, we will prove two more results about them.
The first tells us what happens when we “take the inverse twice,” and the second
gives us a rule for determining the inverse of a product of two group elements.

Proposition 2.3.4. If a ∈ G, then (a−1 )−1 = a.

Proof. By definition, a−1 (a−1 )−1 = e. But a−1 a = aa−1 = e as well, so by unique-
ness of inverses, (a−1 )−1 = a.

Recall from linear algebra that if A, B ∈ GLn (R), we have a formula for (AB)−1 ;
it is simply B −1 A−1 . In general there is a rule for determining the inverse of a
product of two group elements, which looks just like the rule for matrices—it is the
product of the inverses, but in the reverse order.

Proposition 2.3.5. For any a, b ∈ G, (ab)−1 = b−1 a−1 .

 Group Theory

Proof. We’ll explicitly show that b−1 a−1 is the inverse of ab by computing:

(ab)(b−1 a−1 ) = ((ab)b−1 )a−1

= (a(bb−1 ))a−1
= (ae)a−1
= aa−1
= e.

Of course we also need to check that (b−1 a−1 )(ab) = e, which works pretty much
the same way:

(b−1 a−1 )(ab) = b−1 (a−1 (ab))

= b−1 ((a−1 a)b)
= b−1 (eb)
= b−1 b
= e.

Thus (ab)−1 = b−1 a−1 .

In order to be thorough and rigorous, we needed to check that b−1 a−1 was a
two-sided inverse in the previous proof. It would be quite annoying if we had to do
this all the time, and you may be wondering if there is a shortcut. There is one,
which allows us to check that two group elements are inverses of each other simply
by multiplying them in only one of the two possible orders. To prove it, we’ll use
the following proposition.

Proposition 2.3.6. The equations ax = b and xa = b have unique solutions in G.

Proof. The solution to ax = b is x = a−1 b, and for xa = b it is x = ba−1 . These are

unique since inverses are unique.

Proposition 2.3.7. Let G be a group, and let a, b ∈ G. If either ab = e or ba = e,

then b = a−1 .

Proof. This really amounts to solving the equation ax = e (or xa = e). We know
from Proposition 2.3.6 that there is a unique solution, namely x = a−1 e = a−1 (in
either case). Therefore, if ab = e or ba = e, then b is a solution to either ax = e or
xa = e, so b = a−1 .

Exercise 2.3. Prove that if G is a group and a, b ∈ G with ab = a, then b = e.

2.4 The Order of an Element and Cyclic Groups

2.4 The Order of an Element and Cyclic Groups

Recall that we’ve developed some shorthand notation for writing out computations
in groups. For one, we have been suppressing the binary operation ∗, and we simply
write ab in place of a ∗ b. We also developed some notation for powers of an element.
We write
an = a
| ∗ a ∗{z· · · ∗ a}
n times

for n ∈ Z+ , and
a−n = (a−1 )n .

In addition, we define a0 = e. Note that this is written for arbitrary groups—if

we’re dealing with an additive group (like Z or Zn ), then we would really write an
as
| + ·{z
a · · + a} = na.
n times

We should always keep this convention in mind when we are working with specific
examples.
Now let G be a group, and let a ∈ G. We’re going to investigate the things that
we can do with powers of a. Therefore, we’ll begin by giving a name to the set of
all powers of a.

Definition 2.4.1. Given a group G and an element a ∈ G, we define

hai = {an : n ∈ Z} = {. . . , a−2 , a−1 , e, a, a2 , a3 , . . .}.

The first observation that we will make is that the familiar “exponent rules”
hold for an arbitrary group element.

Proposition 2.4.2. Let G be a group, and let a ∈ G.

(a) an am = an+m for all n, m ∈ Z.

(b) (an )−1 = a−n for all n ∈ Z.

(c) (an )m = anm for all n, m ∈ Z.

Exercise 2.4. Prove Proposition 2.4.2.

Now we’ll investigate what happens when we keep taking powers of an element.
Let’s use a specific example to get started.
 Group Theory

Example 2.4.3. Let G = Z12 . Let’s compute the “powers” of some elements.
(Recall that in this group “power” really means “multiple.”) We’ll calculate the
powers of 2 first:

1·2=2
2 · 2 = 2 +12 2 = 4
3 · 2 = 2 +12 2 +12 2 = 6
4·2=8
5 · 2 = 10
6 · 2 = [12]12 = 0
7 · 2 = [14]12 = 2
8 · 2 = [16]12 = 4

and so on. What about powers of 3?

1·3=3
2 · 3 = 3 +12 3 = 6
3 · 3 = 3 +12 3 +12 3 = 9
4 · 3 = [12]12 = 0
5 · 3 = [15]12 = 3
6 · 3 = [18]12 = 6

and so on. Notice that the lists repeat after a while. In particular, we reach 0 (i.e.,
the identity) after a certain point. We quantify this phenomenon by saying that
these elements have finite order.

Definition 2.4.4. Let G be a group. We say that an element a ∈ G has finite

order if there exists n ∈ Z+ such that an = e. The smallest such integer is called
the order of a, denoted by o(a) (or |a|). If no such integer exists, we say that a has
infinite order.

Example 2.4.5. 1. The identity element in any group has order 1.

2. In Z12 , we have seen that o(2) = 6 and o(3) = 4.

3. In D3 , the order of the reflection m1 is 2. (This is true of all reflections in

D3 .) Also, o(r1 ) = o(r2 ) = 3.

4. In Z, 5 has infinite order, as do all other elements.

2.4 The Order of an Element and Cyclic Groups

Let’s ask ourselves a question. Do all the elements of Z12 have finite order?
Yes—if a ∈ Z12 , then 12 · a = 0, for example. What is the real reason for this? The
group is finite, so there are only so many places to put the powers of an element.
We can quantify this with the following proposition.

Proposition 2.4.6. Let G be a finite group. Then every element a ∈ G has finite
order.

Proof. Consider the set

{an : n ≥ 0} = {e, a, a2 , . . .}.

Since G is finite, this list of powers can’t be infinite. (This follows from the Pigeon-
hole principle, for instance. We have an infinite list of group elements that need to
fit into only finitely many slots.) Therefore, two different powers of a must coincide,
say ai = aj , with j 6= i. We can assume that j > i. Then

aj−i = aj a−i = ai a−i = e,

so a has finite order. (In particular, o(a) ≤ j − i.) Since a ∈ G was arbitrary, the
result follows.

Remark 2.4.7. There are two questions that we could ask here. First, you may
be wondering if the converse of Proposition 2.4.6 holds. That is, if G is a group
in which every element has finite order, is G necessarily finite? The answer to this
question is no—there are examples of infinite groups in which every element has
finite order, but we do not have the tools yet to describe them.
On the other hand, if G is an infinite group, can we have elements of finite order?
Yes—there are even examples in GLn (R):
 2  
0 1 1 0
= ,
1 0 0 1

so this matrix has order 2. It’s also easy to see this if we think about the linear
transformation that this matrix represents—it reflects the plane R2 across the line
y = x. That is, it interchanges the two standard basis vectors:

e2 Ae2
e1 Ae2
=⇒
 Group Theory

Let’s get on with proving some facts about order. First, we’ll relate the order of
an element to that of its inverse.
Proposition 2.4.8. Let G be a group and let a ∈ G. Then o(a) = o(a−1 ).
Proof. Suppose first that a has finite order, with o(a) = n. Then

(a−1 )n = a−n = (an )−1 = e−1 = e,

so o(a−1 ) ≤ n = o(a). On the other hand, if we let m = o(a−1 ), then

am = ((a−1 )−1 )m = (a−1 )−m = ((a−1 )m )−1 = e,

so n ≤ m. Thus n = m, or o(a) = o(a−1 ).

Now suppose that a has infinite order. Then for all n ∈ Z+ , we have an 6= e.
But then
(a−1 )n = a−n = (an )−1 6= e
for all n ∈ Z+ , so a−1 must have infinite order as well.

Let’s continue with our investigation of basic properties of order. The first one
says that the only integers m for which am = e are the multiples of o(a).
Proposition 2.4.9. If o(a) = n and m ∈ Z, then am = e if and only if n divides
m.
Proof. If n | m, it is easy. Write m = nd for some d ∈ Z. Then

am = and = (an )d = ed = e.

On the other hand, if m ≥ n, we can use the Division Algorithm to write

m = qn + r with 0 ≤ r < n. Then

e = am = aqn+r = aqn ar = (an )q ar = ear = ar ,

so ar = e. But r < n, and n is the smallest positive power of a which yields the
identity. Therefore r must be 0, and n divides m.

Note that this tells us something more general about powers of a: when we
proved that elements of finite groups have finite order, we saw that ai = aj implied
that aj−i = e. This means that n = o(a) divides j − i. In other words, i and j must
be congruent mod n.
Proposition 2.4.10. Let G be a group, a ∈ G an element of finite order n. Then
ai = aj if and only if i ≡ j (modn).
Along the same lines, we observed that if ai = aj with j > i, then aj−i = e, so
a has to have finite order. Taking the contrapositive of this statement, we get the
following result.
2.4 The Order of an Element and Cyclic Groups

Proposition 2.4.11. Let G be a group, a ∈ G an element of infinite order. Then

all the powers of of a are all distinct. (That is, ai = aj if and only if i = j.)
Finally, If we know that a has order n, then how can we find the orders of other
powers of a? It turns out that there is a nice formula.
Theorem 2.4.12. Let G be a group, a ∈ G an element of finite order, say o(a) = n,
and let m ∈ Z. Then
n
o(am ) = .
gcd(m, n)
Proof. Let d = gcd(m, n). Then

(am )n/d = amn/d = (an )m/d = em/d = e.

However, we need to check that n/d is the smallest positive integer which gives the
identity. Suppose that k ∈ Z+ and (am )k = e. Then

amk = e,

so n | mk by the previous proposition. It follows that n/d divides (m/d)k. Now we

observe that gcd(n/d, m/d) = 1: we use Bézout’s lemma to write

d = nx + my

for some x, y ∈ Z. Dividing by d, we get

1 = (n/d)x + (m/d)y,

so n/d and m/d have to be relatively prime. (Any common divisor would have to
divide this linear combination , hence divide 1.) This means that n/d divides k. In
particular, n/d ≤ k, so o(am ) = n/d.

2.4.1 Cyclic Groups

Let’s take a few steps back now and look at the bigger picture. That is, we want to
investigate the structure of the set hai for a ∈ G. What do you notice about it?
• Closure: ai aj = ai+j ∈ hai for all i, j ∈ Z.

• Identity: e = a0 ∈ hai

• Inverses: Since (aj )−1 = a−j , we have (aj )−1 ∈ hai for all j ∈ Z.
It other words, hai is itself a group. That is, the set of all powers of a group element
is a group in its own right. We will investigate these sorts of objects further in the
next section, but let’s make the following definition now anyway.
 Group Theory

Definition 2.4.13. For a ∈ G, the set hai is called the cyclic subgroup generated
by a.

For now, let’s look at a particular situation. Is G ever a cyclic subgroup of itself?
That is, can a “generate” the whole group G? Yes, this does happen sometimes,
and such groups are quite special.

Definition 2.4.14. A group G is called cyclic if G = hai for some a ∈ G. The

element a is called a generator for G.

Cyclic groups are extremely well-understood. We’ll see that they have very nice
properties, and we can completely classify them almost immediately. First, let’s do
some examples, many of which we have already seen.

Example 2.4.15. 1. One of our first examples of a group is actually a cyclic

one: Z forms a cyclic group under addition. What is a generator for Z? Both
1 and −1 generate it, since every integer n ∈ Z can be written as a “power”
of 1 (or −1):
n = n · 1 = |1 + 1 +
{z· · · + 1} .
n times

Thus,
Z = h1i = h−1i.
These are actually the only two generators.

2. How about a finite cyclic group? For any n, Zn is cyclic, and 1 is a generator
in much the same way that 1 generates Z. There are actually plenty of other
generators, and we can characterize them by using our knowledge of greatest
common divisors. We’ll postpone this until we’ve made a couple of statements
regarding cyclic groups.

3. The group hQ, +i is not cyclic. (This is proven in Saracino.)

4. The dihedral group D3 is not cyclic. The rotations all have order 3, so

hr1 i = hr2 i = {i, r1 , r2 }.

On the other hand, all of the reflections have order 2, so

hm1 i = {i, m1 }, hm2 i = {i, m2 }, hm3 i = {i, m3 }.

Now let’s start making some observations regarding cyclic groups. First, if G =
hai is cyclic, how big is it? It turns out that our overloading of the word “order”
was fairly appropriate after all, for |G| = o(a).

Theorem 2.4.16. If G = hai is cyclic, then |G| = o(a).

2.4 The Order of an Element and Cyclic Groups

Proof. If a has infinite order, then |G| must be infinite. On the other hand, if
o(a) = n, then we know that ai = aj if and only if i ≡ j mod n, so the elements of
G are
e, a, a2 , . . . , an−1 ,
of which there are n = o(a).

If we pair this result with Theorem 2.4.12, we can characterize the generators of
any finite cyclic group.

Proposition 2.4.17. The generators of a finite cyclic group G = hai of order n are
precisely the elements ar for which gcd(r, n) = 1.

Proof. By the theorem,

n
o(ar ) = .
gcd(r, n)
Note that ar generates G if and only if o(ar ) = n. If ar has order n, then |har i| = n
and har i ⊆ hai. Since both sets have the same (finite) number of elements, they
must be the same. On the other hand, if har i = hai, then

o(ar ) = |har i| = |hai| = n.

Now note that o(ar ) = n precisely when gcd(r, n) = 1.

Remark 2.4.18. This corollary allows us to characterize the generators of Zn .

Recall that Zn = h1i. Remember that we have to be careful here—when we write
something like am in an arbitrary group, we mean “multiply a by itself m times,”
where “multiply” should be interpreted as whatever the group operation happens
to be. In particular, the operation in Zn is addition, so am really means

a {z· · · + a} = m · a.
| +a+
m times

Therefore, the generators of Zn are precisely the elements of the form r · 1, where
gcd(r, n) = 1. In other words, an element a ∈ Zn is a generator if and only if
gcd(a, n) = 1.

On a related note, let G = hai be a finite cyclic group with n = |G| = o(a). If
we take any element b ∈ G and compute bn , what do we get? Well, b = ai for some
i ∈ Z, so
bn = (ai )n = ain = (an )i = (ao(a) )i = ei = e.
Therefore:

Theorem 2.4.19. Let G be a finite cyclic group. Then for any element b ∈ G, we
have b|G| = e.
 Group Theory

If we combine this result with Proposition 2.4.9, If b|G| = e, then what can we
say about o(b) in relation to |G|? We must have that o(b) divides |G|.

Proposition 2.4.20. Let G be a finite cyclic group. Then for any b ∈ G, we have
o(b) | |G|.

These last two results are not as serendipitous as they may seem at first glance.
These phenomena for cyclic groups actually hold for any finite group. Once we have
established Lagrange’s theorem, we’ll see why this is true.
Finally, the astute reader will notice a key similarity between all of the examples
of cyclic groups that we’ve listed above: they are all abelian. It is not hard to see
that this is always the case.

Theorem 2.4.21. Every cyclic group is abelian.

Proof. Let G be a cyclic group and let a be a generator for G, i.e. G = hai. Then
given two elements x, y ∈ G, we must have x = ai and y = aj for some i, j ∈ Z.
Then
xy = ai aj = ai+j = aj+i = aj ai = yx,
and it follows that G is abelian.

Remark 2.4.22. The converse to Theorem 2.4.21 is not true. That is, there are
abelian groups that are not cyclic. Saracino gives the example of the non-cyclic
group hQ, +i. However, this is a good place to introduce a different group—the
Klein 4-group, denoted V4 .7 The Klein 4-group is an abelian group of order 4. It
has elements V4 = {e, a, b, c}, with

a2 = b2 = c2 = e

and
ab = c, bc = a, ca = b.
Note that it is abelian by a previous exercise (Exercise 2.1). However, it is not
cyclic, since every element has order 2 (except for the identity, of course). If it were
cyclic, there would necessarily be an element of order 4.

2.4.2 Classification of Cyclic Groups

As one final word on cyclic groups, we should mention that these groups are very
easy to classify. It will turn out that they have a very rigid structure, and we have
already seen some examples of this. Specifically, we will see that there are really
7
2.5 Subgroups

only two flavors of cyclic group: finite and infinite. Let’s think about what happens
in the infinite case first. If we have an infinite cyclic group G = hai, then a has
infinite order, and the group elements are simply

G = aj : j ∈ Z .


How does the group operation work? We have

aj ak = aj+k ,

so we can multiply two elements by simply adding the exponents as integers. There-
fore, we’ll see that this will allow us to set up an isomorphism between any infinite
cyclic group G and Z. That is, there is only one infinite cyclic group, namely Z.
Things are almost as easy in the finite case as well. Suppose that G = hai is a
finite cyclic group of order n. Then to add two elements, we have

aj ak = aj+k .

However, we have to remember that in doing this, j + k may have exceeded the
order of the group, and we can reduce it. That is, we can write

j + k = nq + r,

so
aj ak = aj+k = anq+r = anq ar = ear = ar .
Therefore, multiplication corresponds to addition of the exponents modulo n. We’ll
see that this allows us to identify G with the elementary cyclic group Zn . In other
words, there is only one finite cyclic group of each order n ∈ Z+ .

2.5 Subgroups
Recall that we saw in the last section that if G is a group and a ∈ G, then the set
hai is itself a group which sits inside G. That is, we have a subset of G which is
itself a group with respect to the same operation as G. A more general version of
this situation will be of great interest to us. Let’s first take a moment to say why
this is so.
We now want to begin studying groups from a much broader perspective. So far,
we have only studied groups as standalone objects, without any relations between
them. This viewpoint will begin to change now. We will start to consider more
interesting questions, like how groups are connected to each other via functions—
the concept of homomorphism and isomorphism—and how they are built out of
smaller groups. This leads naturally to the questions of classification and structure:
 Group Theory

• Classification: When are two groups the same or different? If we are dealing
with a mysterious new example of a group, perhaps it is just a more familiar
group in disguise.

• Structure: How is a group built out of smaller pieces? When working with a
big, complicated group, we might be able to break the group up into smaller
pieces, which are then easier to analyze.

We’ve already seen some primitive examples that foreshadow the classification prob-
lem. At this point we will start laying the foundation for the structure problem.
Eventually we’ll want to tear down groups into smaller bits, which will make the
larger group easier to understand. To do this, we need to know what these “bits”
really are.
Obviously we want to look at subsets of a group G, but we don’t just want any
old subset of G. We would like to consider only subsets that encode information
about the group structure on G. These subsets are exactly what are called subgroups
of G.

Definition 2.5.1. Let hG, ∗i be a group. A subgroup of G is a nonempty subset

H ⊆ G with the property that hH, ∗i is a group.

Note that in order for H to be a subgroup of G, H needs to be a group with

respect to the operation that it inherits from G. That is, H and G always carry the
same binary operation. Also, we’ll write

H≤G

to denote that H is a subgroup of G. Finally, if we want to emphasize that H ≤ G

but H 6= G, we will say that H is a proper subgroup of G.
In order to think about how to show that a subset of a group is actually a
subgroup, let’s work with an example. This will really be a specific example of
the “cyclic subgroups” that we introduced earlier, but let’s rehash the details here
anyway.

Example 2.5.2. Let’s look at the group Z (under addition, of course). Define

2Z = {even integers} = {2n : n ∈ Z}.

Is 2Z a subgroup of Z? We need to check that 2Z itself forms a group under addition:

• Closure: If a, b ∈ 2Z, then a = 2n and b = 2m for some n, m ∈ Z. Then

a + b = 2n + 2m = 2(n + m) ∈ 2Z,

so 2Z is indeed closed under addition.

2.5 Subgroups

• Associativity: Is there even anything to check here? No—the operation on

Z is already associative, so nothing changes when we pass to a subset of Z.

• Identity: The identity for addition on Z is 0, which is even: 0 = 2 · 0 ∈ 2Z.

• Inverses: If a ∈ 2Z, then a = 2n for some n ∈ Z, and −a = −2n = 2(−n) ∈

2Z.

Therefore, h2Z, +i is a group, hence a subgroup of Z.

In general, we will want to be able to check whether a subset of a group is

actually a subgroup. Fortunately, this example tells us exactly how to do it.

To check that H ≤ G, one needs to verify the following:

1. H is closed under the operation on G.

2. The identity element e ∈ G is in H.

3. For every a ∈ H, its inverse a−1 is in H.

Example 2.5.3. Naturally, it would be helpful to look at some examples of sub-

groups.

1. Every group G has two special subgroups, namely

{e} and G.

These are called the trivial subgroups of G.9

2. We saw earlier that 2Z is a subgroup of Z. There is nothing special about 2

in this example: for any n ∈ Z+ ,

nZ = {na : a ∈ Z}

is a subgroup of Z. The exact same computations that we performed for 2Z

will show that nZ ≤ Z.

3. The rational numbers Q form an additive subgroup of R.

4. Here is an example from linear algebra. Consider the n-dimensional vector

space Rn . Then Rn is, in particular, an abelian group under addition, and any
9
Many books will reserve the phrase “trivial subroup” only for the identity subgroup {e}. The
group G is sometimes referred to as the improper subgroup.
 Group Theory

vector subspace of Rn is a subgroup of Rn .10 If H is a subspace of Rn , then

it is closed under addition, and closure under scalar multiplication guarantees
that 0 ∈ H and for v ∈ H, −v = −1 · v ∈ H.

5. Let Dn be the nth dihedral group, and let

H = {i, r1 , r2 , . . . , rn−1 }

be the set of rotations in Dn . Then H ≤ Dn . It is closed, since the composition

of two rotations is another rotation, the identity i ∈ H, and for any 1 ≤ j ≤
n − 1,
rj−1 = rn−j ,
which is again in H.

6. Let GLn (R) be our usual group of invertible n × n matrices under matrix
multiplication, and define

SLn (R) = {A ∈ GLn (R) : det(A) = 1} .

Then SLn (R) ≤ GLn (R). To see that it is closed, we recall that det(AB) =
det(A) det(B), so if A, B ∈ SLn (R),

det(AB) = det(A) det(B) = 1 · 1 = 1,

and AB ∈ SLn (R). The identity matrix surely has determinant 1, and if
A ∈ SLn (R), then
1 1
det(A−1 ) = = = 1,
det(A) 1
so A−1 ∈ SLn (R). Therefore, SLn (R) is a subgroup of GLn (R), called the
special linear group.11

7. Here’s our last example, and a more interesting one at that. Exercise 2.1 leads
to toward the observation that there are two groups of order 4, and both are
abelian. One of them is of course Z4 , and the other is the Klein 4-group V4 ,
10
A word of caution about this statement: any subspace of Rn is necessarily a subgroup, but
there are plenty of subgroups that are not vector subspaces. For example, the rational vector space
Qn is an additive subgroup of Rn , but it is not a real subspace. It is a Q-subspace, however, which
goes to show that the field of scalars is critical when talking about subspaces of vector spaces. Even
worse, the set
Zn = {(a1 , a2 , . . . , an ) : ai ∈ Z}
is a subgroup of Rn , called a lattice, but it is not a vector space of any kind. (If you’re feeling
extra brave/curious, it is an example of a more general object, called a module.)
11
You might wonder what makes the special linear group “special.” If we view the matrices
in GLn (R) as invertible linear transformations from Rn to itself, then the elements of SLn (R) are
precisely the transformations which preserve volume and orientation.
2.5 Subgroups

which was introduced in Remark 2.4.22. You saw in that exercise that Z4
and V4 really are different groups, since they have completely different Cayley
tables. We will now observe this in a different way, by checking that they
have different subgroup structures. First, we claim that the only nontrivial
subgroup of Z4 is H = {0, 2}. It’s easy to check that this is a subgroup, but
why is it the only one? We will prove a result to this effect quite soon, or we
could observe that the other possible proper subgroups are
{0, 1}, {0, 3}, {0, 1, 2}, {0, 1, 3}, {0, 2, 3},
and it is easy to check that none of these are closed under addition. On the
other hand, V4 has three subgroups (in addition to {e} and V4 itself):
{e, a}, {e, b}, {e, c}.
We therefore know all the subgroups of these two groups, and we can represent
this pictorially with something called a subgroup lattice diagram:

Z4 V4

{0, 2} {e, a} {e, b} {e, c}

{0} {e}

As we have already insinuated, we will sometimes be able to tell when two

groups are different by studying these sorts of lattice diagrams. For example,
Z4 has only one subgroup of order 2, while V4 has three such subgroups. This
is one indication that these are indeed different groups. If you go on to study
Galois theory at all, you will see that subgroup diagrams are quite important
in that context.

2.5.1 Cyclic Subgroups

Some of the examples that we have mentioned are actually cases of very special kinds
of subgroups, called cyclic subgroups. We already introduced cyclic subgroups en
route to our study of cyclic groups. To recap, let G be a group and let a ∈ G. We
defined the set
hai = aj : j ∈ Z ,


and we observed that hai is itself a group which sits inside G. That is, every element
of a group G generates a whole subgroup of G, to which we attach a special name.
 Group Theory

Definition 2.5.4. The group hai is called the cyclic subgroup generated by a.
When we say that a “generates” hai, we mean that that hai is created entirely
out of the element a. In a certain sense, hai is the smallest possible subgroup of G
which contains a. Let’s try to make this more precise. If H ≤ G and a ∈ H, then
H must contain the elements
a, a2 , a3 , . . . ,
since H is closed. It also must contain e and a−1 , hence all of the elements
. . . , a−2 , a−1 , e, a, a2 , . . . ,
i.e. all powers of a. That is, hai ⊂ H, and we have proven the following fact:
Theorem 2.5.5. Let G be a group and let a ∈ G. Then hai is the smallest subgroup
of G containing a, in the sense that if H ≤ G and a ∈ H, then hai ⊆ H.
Of course we’ve already encountered several examples of cyclic subgroups in our
studies thus far.
Example 2.5.6. 1. Our first example of a subgroup, 2Z ≤ Z, is a cyclic sub-
group, namely h2i. Similarly, nZ is cyclic for any n ∈ Z.
2. The subgroup consisting of rotations on Dn ,
H = {i, r1 , r2 , . . . , rn−1 } ≤ Dn ,
is cyclic since H = hr1 i.
3. All the proper subgroups of Z4 and V4 that we listed are cyclic. In addition,
Z4 is a cyclic subgroup of itself, but V4 is not.
4. The trivial subgroup {e} is always a cyclic subgroup, namely hei.
Cyclic subgroups are useful because they will allow us to gather information
about elements of G by studying the cyclic subgroups that they generate. Properties
of the subgroup are reflected in those of the element, and vice versa. One example
is the relationship between the order of an element and the order of the associated
cyclic subgroup:
|hai| = o(a).

2.5.2 Subgroup Criteria

Now we’ll give a couple of other criteria for showing that a subset of a group is
actually a subgroup. When dealing with specific examples, it is often easiest to
simply verify the axioms that we have been using all along. However, when proving
things about subgroups it can be useful to use one of the following characterizations.
The first one shows that we can collapse the usual subgroup axioms into a single
condition.
2.5 Subgroups

Theorem 2.5.7. Let G be a group. A nonempty subset H ⊆ G is a subgroup if and

only if whenever a, b ∈ H, ab−1 ∈ H.
Proof. Suppose that H ≤ G, and let a, b ∈ H. Then b−1 ∈ H, so ab−1 ∈ H since H
is closed.
Conversely, suppose that ab−1 ∈ H for all a, b ∈ H. Then for any a ∈ H, we can
take a = b and conclude that
e = aa−1 ∈ H,
so H contains the identity. Since e ∈ H, for any a ∈ H we have

a−1 = ea−1 ∈ H,

so H is closed under taking inverses. Finally, we claim that H is closed under the
group operation. If a, b ∈ H, then b−1 ∈ H, so b−1 a−1 ∈ H, and therefore

−1
ab = (ab)−1 = (b−1 a−1 )−1 ∈ H.

Thus H is closed, hence a subgroup of G.

The next criterion is quite interesting. It obviously reduces the number of things
that one needs to check, but it only works for a finite subset of a group G.
Theorem 2.5.8. Let G be a group and H a nonempty finite subset of G. Then H
is a subgroup if and only if H is closed under the operation on G.
Proof. If H is a subgroup, then it is obviously closed by hypothesis.
On the other hand, we are assuming that H is closed, so we need to verify that
e ∈ H and that for every a ∈ H, a−1 ∈ H as well. Since {e} ≤ G, we may assume
that H is nontrivial, i.e. that H contains an element a distinct from the identity.
Since H is closed, the elements
a, a2 , a3 , . . .
are all in H, and since H is finite, this list cannot go on forever. That is, we must
eventually have duplicates on this list, so

ai = aj

for some 1 ≤ i < j ≤ |H|. Since i < j, j − i ≥ 0 and we have

ai = aj = aj−i ai ,

and using cancellation, we get

aj−i = e.
Therefore, e ∈ H. Now observe that j − i − 1 ≥ 0, so aj−i−1 ∈ H, and

aaj−i−1 = aj−i = e,

so a−1 = aj−i−1 ∈ H. Therefore, H is a subgroup of G.

 Group Theory

This theorem has an easy corollary, which is useful when the group is finite.

Corollary 2.5.9. If G is a finite group, a subset H ⊆ G is a subgroup of G if and

only if it is closed under the operation on G.

2.5.3 Subgroups of Cylic Groups

Let’s return now to the cyclic case. There is one very important thing that we can
say about cyclic groups, namely that their subgroups are always cyclic.

Theorem 2.5.10. A subgroup of a cyclic group is cyclic.

Proof. Let G = hai be a cyclic group and let H be a subgroup of G. We may

assume that H 6= {e}, since {e} is already known to be cyclic. Then H contains
an element other than e, which must have the form am for some m ∈ Z since G is
cyclic. Assume that m is the smallest positive integer for which am ∈ H. We claim
that H = ham i. To do this, we need to show that if an ∈ H, then an is a power of
am .
Suppose that an ∈ H, and use the Division Algorithm to write n = qm + r,
where 0 ≤ r < m. Then

an = aqm+r = aqm ar = (am )q ar .

Since H is a subgroup, (am )−q ∈ H, hence (am )−q an ∈ H, and it follows that

ar = (am )−q an

is in H. But r < m and we have assumed that m is the smallest positive integer such
that am ∈ H, so we must have r = 0. In other words, an = (am )q , so an ∈ ham i.
Since an was an arbitrary element of H, we have shown that H ⊆ ham i. Since
am ∈ H, we also have ham i ⊆ H, so H = ham i, and H is cyclic.

This theorem has a particularly nice corollary, which tells us a lot about the
structure of Z as an additive group.

Corollary 2.5.11. The only subgroups of Z are the cyclic subgroups nZ, where
n ∈ Z.

Proof. The cyclic subgroups of Z are simply nZ = hni for any n ∈ Z. By the
theorem, the only subgroups of Z are the cyclic ones, so we are done.

The next corollary is quite interesting. Recall that if G = hai is a finite cyclic
group of order n and aj ∈ G, we have a formula for the order of aj :
n
o(aj ) = .
gcd(j, n)
2.5 Subgroups

This tells us something in particular, which we mentioned earlier: if aj ∈ hai, then

o(aj ) divides |hai|. In particular, the order of the cyclic subgroup generated by aj
must divide the order of hai. This holds more generally—in fact, it is true for any
finite group, as we will see when we prove Lagrange’s theorem. However, a sort of
converse holds for cyclic groups: if m divides the order of the group, then there is a
subgroup of that order. Moreover, if a cyclic group has a subgroup of a given order,
then that subgroup is unique.12

Corollary 2.5.12. Let G = hai be a cyclic group of order n. If m is a positive

divisor of m, then G has exactly one subgroup of order m.

Proof. First we need to show that G even has a subgroup of order m whenever
m | n. Well, suppose that m | n, and put b = an/m . Then by Theorem 2.4.12,
n n
o(b) = = = m.
gcd(n/m, n) n/m

Therefore, hbi has m elements, so G has a subgroup of order m.

Now we need to show that G only possesses one subgroup of order m. Suppose
that aj is another element of order m, so that haj i has m elements. Then
 j n n
a  = =m= = |b|.
gcd(j, n) gcd(n/m, n)

Therefore, gcd(j, n) = gcd(n/m, n) = n/m. In particular, n/m divides j, so we can

write
j = r(n/m)
for some r ∈ Z. Then
aj = arn/m = (an/m )r = br ,
so aj ∈ hbi. This forces haj i ⊂ hbi, and since both sets have the same (finite) number
of elements, haj i = hbi.

Let’s put this to work in an example.

Example 2.5.13. Let’s write down all of the subgroups of Z12 . We know that
there will be one of each order that divides 12, and these divisors are 1, 2, 3, 4,
6, and 12. The order 1 subgroup is just {0}, and for order 12 we have the whole
group Z12 . For the others, we need to find an element of that order. We can see
that 6 has order 2, so {0, 6} is our order 2 subgroup. For order 3, we can use 4 as
the generator, so {0, 4, 8} is the subgroup. For order 4, we have 3 ∈ Z12 , and the
subgroup is {0, 3, 6, 9}. Finally, 2 has order 6, and the subgroup is {0, 2, 4, 6, 8, 10}.
Therefore, the subgroup lattice of Z12 looks like:
12
These latter two statements fail miserably for arbitrary finite groups. If m divides the order of
the group, there need not be a subgroup of order m. (It is true for abelian groups, however.) Also,
if a subgroup of order m exists, it need not be unique. (Look at the Klein 4-group, for example.)
 Group Theory

Z12

h2i h3i

h4i h6i

h0i

2.6 Lagrange’s Theorem

In our investigation of cyclic groups, we noticed one thing—that if G = hai is finite,

the order of any element divides the order of the group. This implies something
else: if H ≤ hai, then we know that H = ham i for some m ∈ Z, and

|H| = |ham i| = |am | divides |a| = |G|.

Thus the order of any subgroup divides the order of the group. This is no accident—
it holds much more generally. In fact, it is true not just for cyclic groups, but for
any finite group and any subgroup.

Theorem 2.6.1 (Lagrange). Let G be a finite group and H be a subgroup of G.

Then |H| divides |G|.

We’re not going to prove Lagrange’s theorem yet. The proof isn’t hard, but we
don’t quite have the right tools for writing it down yet. We need to develop some
new language in order to properly prove it. The concepts that we’ll talk about will
be really important for the rest of the course, but the first thing they will buy us is
a proof of Lagrange’s theorem.
The idea of the proof will be the following: given a finite group G and H ≤ G,
we want to “carve up” G into a collection of subsets, all of which are determined by
the subgroup H. We will first need to figure out what these subsets should be.
2.6 Lagrange’s Theorem

The proper way to say this is that we want to partition G. In order to do this, we
need to talk about equivalence relations.
Remark 2.6.2. Lagrange’s theorem is a statement about finite groups, but we are
going to look at arbitrary groups for this part. A lot of the techniques (specifically
the idea of cosets) will be important in a more general setting.
Before we formally introduce equivalence relations, let’s do an example for mo-
tivation. It should be eerily familiar, or it will be once we’re done with it.
Example 2.6.3. Let’s look at our favorite group—Z. We just saw that the only
subgroups of Z are those of the form nZ, for n ∈ Z+ . Let’s take 3Z, for example.
What if we start “translating” this subgroup by integers?
3Z + 0 = {. . . , −6, −3, 0, 3, 6, . . .}
3Z + 1 = {. . . , −5, −2, 1, 4, 7, . . .}
3Z + 2 = {. . . , −4, −1, 2, 5, 8, . . .}
This is depicted on the color-coded number line below:

−6 −5 −4 −3 −2 −1 0 1 2 3 4 5 6 7 8

What if we keep translating? We just get the same sets all over again:
3Z + 3 = 3Z + 0
3Z + 4 = 3Z + 1
and so on. Also, do these “translates” of 3Z have any overlap? No—if two of the
sets are distinct, then they are actually completely disjoint. Finally, they fill up all
of Z, in the sense that their union is Z. In other words, every integer belongs to one
of the sets on this list.
What do you notice about the elements of each of these sets? They are all
congruent mod 3:
3Z + 0 = {a ∈ Z : a ≡ 0 mod 3}
3Z + 1 = {a ∈ Z : a ≡ 1 mod 3}
3Z + 2 = {a ∈ Z : a ≡ 2 mod 3}
 Group Theory

In fact, we could say that these sets are determined by the relationship of “congru-
ence mod 3”—two integers a and b fall into the same “translate” if and only if they
are congruent mod 3, i.e., if and only if 3 divides a − b.

Before moving on, let’s recap the key observations that we’ve made in this ex-
ample. By shifting around the subgroup 3Z, we obtain translates of the subgroup
such that

1. They are determined by the relation a ≡ b mod 3.

2. Distinct translates are disjoint.

3. Their union is all of Z.

The latter two conditions say that we have partitioned Z. This is the situation
that we would like to generalize.

2.6.1 Equivalence Relations

In mathematics, when one wants to partition a set, the natural way to do it is with
something called an equivalence relation. These objects are important throughout
mathematics, and not just in the field of algebra. This means that we are about to
step away from algebra for a moment and talk about some even more abstract ideas
(if you can believe such a thing). The primary examples will be from algebra, of
course, but these sorts of concepts are ubiquitous throughout mathematics. They
are used in algebra, topology, analysis, combinatorics, and other fields. Therefore,
this will be a good place for you to encounter them.
Even though equivalence relations are fairly abstract, we’ll use congruence mod
n as a guiding example. In that case, we have some very nice things going on. There
are three things in particular that we can single out:

1. If a ∈ Z, then a ≡ a mod n.

2. If a, b ∈ Z and a ≡ b mod n, then b ≡ a mod n.

3. If a, b, c ∈ Z with a ≡ b mod n and b ≡ c mod n, then a ≡ c mod n.

These are the properties that we will eventually want equivalence relations to
possess. Before we can properly define them, we need to make the word “relation”
precise.

Definition 2.6.4. Let S be a set. A relation on S is a set R ⊆ S × S of ordered

pairs.

Remark 2.6.5. The definition above is the formal, precise way of describing a
relation. It is not a particularly helpful way in which to think about relations. You
should simply think of a relation as a way to “pair off” elements of S. To this end,
2.6 Lagrange’s Theorem

we will usually write x ∼ y to mean (x, y) ∈ R, and we will say that “x is related to
y.” We will usually refer to ∼ as the relation, and not worry too much about the
proper definition.

We will not concern ourselves with general relations. Our main interests will
lie in relations which behave like “congruence mod n.” These are the equivalence
relations.

Definition 2.6.6. Let S be a set. An equivalence relation on S is a relation ∼

on S satisfying the following three properties:

1. Reflexivity: a ∼ a for all a ∈ S.

2. Symmetry: If a, b ∈ S and a ∼ b, then b ∼ a.

3. Transitivity: If a, b, c ∈ S with a ∼ b and b ∼ c, then a ∼ c.

We will usually think of an equivalence relation as a way of pairing off elements

of S by equivalence. If a ∼ b, we will usually say that “a is equivalent to b.”
The whole point of introducing equivalence relations is to obtain a method of
partitioning a group. Therefore, we will of course be interested in subsets of S that
are somehow “determined” by an equivalence relation.

Definition 2.6.7. Given a ∈ S, define

[a] = {b ∈ S : b ∼ a} .

We call [a] the equivalence class of a, and a is called a representative13 of the

equivalence class.

Example 2.6.8. Here are some examples of equivalence relations.

1. This is one of the most important examples, and it’s the one we’ve already
seen. Fix n ∈ Z, and define ∼ on Z by a ∼ b if and only if a ≡ b mod n.
We have already checked that this is an equivalence relation. What are the
equivalence classes? For a ∈ Z, we have

[a] = {b ∈ Z : b ≡ a mod n} = nZ + a.

2. Define ∼ on R by a ∼ b if a = b. Then ∼ is an equivalence relation:

• Reflexive: For any a ∈ R, a = a, so a ∼ a.

• Symmetric: If a, b ∈ R and a = b, then b = a, so a ∼ b implies that
b ∼ a.
13
Note that an equivalence class may have many representatives. In fact, a representative is just a
chosen element of the equivalence class, so any element of the class can be taken as a representative.
 Group Theory

• Transitive: If a, b, c ∈ R with a = b and b = c, then a = c.

What are the equivalence classes? Each class consists of a single element:

[a] = {a}.

3. Define ∼ on GLn (R) by A ∼ B if det(A) = det(B). It’s easy to see that this
is an equivalence relation, and that

[A] = {B ∈ GLn (R) : det(B) = det(A)} .

4. This example is really the prototypical equivalence relation, in the sense that
any equivalence relation can be viewed in this way. Let SSbe any set, and let
{Si } be a partition of S. That is, Si ⊂ S for all i, S = Si , and the Si are
pairwise disjoint:
Si ∩ Sj = ∅

if i 6= j. We call the Si the cells of the partition. We can define an equivalence

relation ∼ on S by a ∼ b if and only if a and b belong to the same cell Si .
It is easy to check that ∼ is reflexive, symmetric, and transitive, hence an
equivalence relation.

Exercise 2.5. Verify that the relation ∼ defined in Example 2.6.8(4) is an equiv-
alence relation.

Aside 2.6.9. While we’re on the topic of specific examples, let’s make a couple
of comments regarding Z with the relation of congruence mod n. Note that the
equivalence classes are simply congruence classes mod n: every element of the class
[a] is congruent to a mod n. Thus all elements of [a] = nZ + a yield the same
element of Zn when reduced mod n. Therefore, the equivalence classes nZ + a can
be identified with elements of Zn . We will see soon that this is really the correct
way to view Zn . The elements of Zn are actually the equivalence classes mod n,
and we define modular addition and multiplication by

[a] + [b] = [a + b]

and
[a][b] = [ab].

Of course, we would need to know that these operations are well-defined. We’ll take
care of that when we revisit Zn later on. We will discuss these ideas in more detail
when we encounter quotient groups, of which Zn will be the foremost example.
2.6 Lagrange’s Theorem

Now let’s return to general equivalence relations. We said that we were intro-
ducing them to allow us to partition sets, so we should check that they actually do
this. We’ve already seen that a partition of a set S actually imposes an equivalence
relation on S, so we are really checking that equivalence relations and partitions go
hand in hand.
Theorem 2.6.10. Let ∼ be an equivalence relation on a set S. The equivalence
classes of ∼ partition S, in the sense that:
1. Given a, b ∈ S, either [a] ∩ [b] = ∅, or [a] = [b].

2. S is the union of all the equivalence classes of ∼. That is, every element of S
belongs to some equivalence class (and only one, by condition 1).
Proof. Let a, b ∈ S. Then either [a] ∩ [b] = ∅, or [a] ∩ [b] is nonempty. In the second
case, there is at least one element c ∈ [a] ∩ [b]. Then c ∈ [a], so a ∼ c, and c ∈ [b], so
c ∼ b. By transitivity, a ∼ b. Thus a ∈ [b], and if x ∈ [a], then x ∼ b by transitivity,
so x ∈ [b] as well. Thus [a] ⊂ [b]. By symmetry, we also have b ∈ [a], and a similar
transitivity argument shows that [b] ⊂ [a]. Thus [a] = [b].
To see that S is the union of the equivalence classes, we just need to notice that
every a ∈ S belongs to one of the equivalence classes. Specifically, a ∈ [a] since ∼ is
reflexive. Thus S is contained in the union of the equivalence classes, so in fact S
equals the union of the equivalence classes. Therefore, the equivalence classes of ∼
partition S.

2.6.2 Cosets
Let’s step back to reality now. (Or, as close to reality as abstract algebra can be.)
We are especially interested in the case where the set in question is actually a group,
and the equivalence relation has something to do with a given subgroup. That is,
we want to partition a group G into subsets, each of which is determined by some
fixed subgroup H. Once we have done this, we will able to write down a proof of
Lagrange’s theorem in a nice way. Our present goal then is to find an equivalence
relation on a group G which is somehow related to a subgroup H. There are two
very similar ones that we can define, and either one will work.
Let’s return for a moment to our key example. The relation ∼ that we defined
on Z by
a ∼ b ⇐⇒ a ≡ b mod n
is really equivalent to specifying that a ∼ b if and only if n | (a − b). This in turn is
equivalent to saying that a − b ∈ nZ. Let’s try to generalize this.
Example 2.6.11. Let G be a group and H ≤ G. Define a relation ∼H on G by:
a ∼H b if and only if ab−1 ∈ H. Is this an equivalence relation?
• Reflexive: If a ∈ G, then aa−1 = e ∈ H, so a ∼H a.
 Group Theory

• Symmetric: Suppose that a, b ∈ G and a ∼H b, so ab−1 ∈ H. To show that

b ∼H a, we need to know that ba−1 ∈ H. But

ba−1 = (ab−1 )−1 ∈ H,

since H is a subgroup of G. Thus b ∼H a, and the relation is symmetric.

• Transitive: Suppose that a ∼H b and b ∼H c. Is a ∼H c? We need to know

that ac−1 ∈ H. Well,

ac−1 = a(b−1 b)c−1 = (ab−1 )(bc−1 ),

and ab−1 ∈ H (since a ∼H b) and bc−1 ∈ H (since b ∼H c), so ac−1 ∈ H.

Thus a ∼H c.

Here’s the next logical question: what are the equivalence classes of ∼H ? Well,

[a]H = {b ∈ G : b ∼H a} = b ∈ G : ba−1 ∈ H .


If a ∼H b, then ab−1 ∈ H, i.e., there exists h ∈ H such that

ba−1 = h,

or b = ha. If we define
Ha = {ha : h ∈ H} ,
then we have just shown that [a]H ⊆ Ha. On the other hand, given b = ha ∈ Ha,

ba−1 = (ha)a−1 = h(aa−1 ) = h ∈ H,

so b ∼H a. Thus Ha ⊆ [a]H . Therefore, we have shown that the equivalence classes

are exactly
[a]H = Ha,
and we’ll generally use Ha instead of the [·]H notation when working in the group
case. These equivalence classes actually have a special name, which we will use
constantly from now on.

Definition 2.6.12. Sets of the form Ha for a ∈ G are called (right) cosets of H
in G.

Let’s summarize what we proved above.

Theorem 2.6.13. Let G be a group, H a subgroup of G, and let ∼ be the relation

on G given by
a ∼H b ⇐⇒ ab−1 ∈ H.
2.6 Lagrange’s Theorem

Then ∼ is an equivalence relation, and the equivalence classes are precisely the right
cosets of H:

[a]H = Ha.

Furthermore, if a, b ∈ G, then either Ha ∩ Hb = ∅ or Ha = Hb, and

Ha = Hb ⇐⇒ ab−1 ∈ H.

Finally, G is the union of all the right cosets of H, so the cosets partition G.

Two of the examples of equivalence relations that we mentioned last time are
actually the relation ∼H in disguise. In fact, this relation is meant to be a general-
ization of congruence mod n.

Example 2.6.14. 1. Consider the group Z and the subgroup H = nZ. Given
a, b ∈ Z, we have seen that a ∼H b if and only if a − b ∈ H, which is just the
additive way of writing ab−1 ∈ H. Thus ∼H is really just congruence mod n,
and the right cosets of H are

[a]H = nZ + a = {nc + a : c ∈ Z},

and there are n of them, namely nZ + 0, nZ + 1, . . . , nZ + (n − 1).

2. Consider H = SLn (R) ≤ GLn (R). If A, B ∈ GLn (R), A ∼H B precisely when

AB −1 ∈ SLn (R), or

det(AB −1 ) = 1.

But
det(A)
det(AB −1 ) = det(A) det(B −1 ) = ,
det(B)

so A ∼H B if and only if det(A) = det(B).

3. Recall that any vector space is an abelian group under addition, and any
vector subspace is a subgroup. In particular, let G = R2 and let H be a 1-
dimensional subspace, i.e., a line through the origin. If v ∈ R2 is any vector,
the coset H + v is just a “parallel translate” of the line H.
 Group Theory

H +v
H
v

4. Since Lagrange’s theorem deals with finite groups, it would probably be helpful
to look at one of those. Let’s look at the dihedral group D3 , and let

H = {i, r1 , r2 } ,

the rotation subgroup. What are the cosets? There are only 2: H itself, and

Hm1 = {m1 , m2 , m3 }.

(Note that we could also take m2 or m3 as a representative of the coset.)

Now that we’ve built up the appropriate machinery, let’s go ahead and use it to
formally prove Lagrange’s theorem.
Theorem 2.6.15 (Lagrange). Let G be a finite group and let H ≤ G. Then |H|
divides |G|.
Proof. Let Ha1 , . . . , Hak denote the distinct cosets of H in G. That is, a1 , . . . , ak
all represent different cosets of H, and these are all the cosets. We know that the
cosets of H partition G, so

|G| = #(Ha1 ) + · · · + #(Hak ). (2.1)

(Here # means the cardinality of the set, or simply the number of elements in that
set.) Therefore, it will be enough to show that each coset has the same number of
elements as H.
We need to exhibit a bijection between H and Hai for each i. For each i =
1, . . . , k, define a function fi : H → Hai by

f (h) = hai .

If we can prove that f is a bijection, then we will have

|H| = #(Hai )
2.6 Lagrange’s Theorem

for all i. This is fairly straightforward: if h1 , h2 ∈ H with f (h1 ) = f (h2 ), then

h1 ai = h2 ai ,

which implies that h1 = h2 , so f is one-to-one. To see that it is onto, take h ∈ H;

then f (h) = hai .
Thus all the cosets have the same number of elements, namely |H|, and (2.1)
really says that
|G| = |H| + · · · + |H| = k|H|.
| {z }
k times

Therefore, |H| does indeed divide |G|.

The number that we called k in the proof is actually quite useful, and we will
therefore give it a special name.

Definition 2.6.16. The number of distinct (right) cosets of H in G is called the

index of H in G, denoted by
[G : H].

The set of all right cosets of H in G is denoted by G/H, so

#(G/H) = [G : H].

Note that we can actually rephrase Lagrange’s theorem in terms of the index: if
G is a finite group and H ≤ G, then

|G| = |H|[G : H].

We’ll now begin to see that Lagrange’s theorem has many very useful consequences.
For one, it greatly simplifies the search for subgroups of a given finite group.

Example 2.6.17. Let’s try to find all the subgroups of D3 . Since |D3 | = 6, we
know that the only possible orders are 1, 2, 3, and 6. The subgroups are then

1 : {i}
2 : {i, m1 }, {i, m2 }, {i, m3 }
3 : {i, r1 , r2 }
6 : D3

We can even draw a lattice diagram to illustrate the subgroup structure of D3 :

 Group Theory

D3

{i, r1 , r2 }

{i, m1 } {i, m2 } {i, m3 }

{i}

Lagrange’s theorem also has a couple of easy yet powerful corollaries. One simply
states that the order of any element divides the order of the group, which we already
know for cyclic groups. The other tells us that groups of prime order are particularly
special, and they behave in a very rigid way.

Corollary 2.6.18. Let G be a finite group with |G| = n, and let a ∈ G. Then o(a)
divides n = |G|, and
an = e.

Proof. We defined o(a) to be |hai|, and hai is a subgroup of G, so its order divides
|G| by Lagrange’s theorem. Therefore, o(a) | n, and we can write n = o(a)m for
some m ∈ Z. Then

m
an = ao(a)m = ao(a) = e.

Corollary 2.6.19. If G is a finite group of prime order p, then G is cyclic.

Proof. Since p is prime, p ≥ 2, and G contains at least one element a with a 6= e. By

the previous corollary, o(a) divides p, and since a 6= e, o(a) 6= 1. Since p is prime,
we must have o(a) = p, so a generates G. Thus G is cyclic.

In this proof, we showed that any nonidentity element of a group G with prime
order is actually a generator for the group. This implies the following fact regarding
subgroups of such groups.

Corollary 2.6.20. If G is a finite group of prime order p, then G has no subgroups

other than {e} and G itself.
2.8 The Symmetric Group Redux

Note that this last part implies that if ϕ is an isomorphism, then o(ϕ(a)) = o(a)
for all a ∈ G1 .

Proposition 2.7.12. If ϕ : G1 → G2 is an isomorphism, then o(ϕ(a)) = o(a) for

all a ∈ G1 .

Proof. Suppose first that a ∈ G1 has finite order. Then we showed in Proposition
2.7.10 that o(ϕ(a)) divides o(a). But ϕ−1 is also a homomorphism, so o(a) =
o(ϕ−1 (ϕ(a))) must divide o(ϕ(a)). Since both integers divide each other, they must
be the same, and o(a) = o(ϕ(a)).
Now note that Proposition 2.7.10 also implies that a has finite order if and only
if ϕ(a) does. It follows that a has infinite order if and only if ϕ(a) does, so we are
done.

2.8 The Symmetric Group Redux

We’re now going to take a short detour into the symmetric group, which we in-
troduced quite some time ago. We’ve put off the details in order to develop some
more general tools for working with groups. Never fear; we’ll be going back into the
abstraction very soon, but right now we’ll try to have a little fun with Sn .
The reason that we are doing this is that Sn is really the most fundamental
finite nonabelian group, so it would be very useful to understand its structure. We
will see via Cayley’s theorem that every group is really a permutation group, in
a certain sense. In fact, groups were originally thought of only as permutation
groups. It was Cayley who first gave the abstract definition of a group and then
proved his eponymous theorem. Therefore, if we could somehow understand all
of the subgroups of Sn , then we would understand all finite groups. This is an
ambitious goal, and it is far too much to ask for. However, understanding the
symmetric group will give us information on it and other groups that are easily
realized as permutations (like Dn , for example).

2.8.1 Cycle Decomposition

When we originally talked about Sn , we introduced some compact notation for
writing down a permutation, called two-line notation: if σ ∈ Sn , then
 
1 2 3 ··· n
σ= .
σ(1) σ(2) σ(3) · · · σ(n)

This is obviously a nice, short way to write down an element of Sn , and multiply-
ing two permutations in two-line notation is relatively straightforward. However,
because of its simplicity it can obscure the structure of σ, and it’s not good enough
 Group Theory

from the standpoint of algebra. For example, it we look at the permutation σ ∈ S8

given by  
1 2 3 4 5 6 7 8
σ= ,
5 3 2 4 8 6 7 1
what is really going on?
• σ fixes 4, 6, and 7.

• 2 and 3 only interact with each other.

• 1, 5, and 8 only interact with each other.

Thus we’ve written down a bunch of extra information that isn’t really necessary,
and we’ve obscured some of the structure that σ possesses in the process. The
relevant data is the fact that σ “cycles through” 1, 5, and 8 and it “cycles through”
2 and 3.
1

2 3
8 5

With this in mind, perhaps there is another way to write down σ that truly captures
what σ is doing. There is, and it is called the cycle decomposition of σ.
Let’s try to be more precise about what we’ve done above: we have

σ(1) = 5
σ 2 (1) = σ(5) = 8
σ 3 (1) = σ(8) = 1
σ 4 (1) = σ(1) = 5
..
.

and

σ(2) = 3
σ 2 (2) = σ(3) = 2

and so on. This might look familiar from your take-home exam:
Theorem 2.8.1. Define a relation ∼ on the set {1, 2, . . . , n} by

i∼j ⇐⇒ σ m (i) = j

for some m ∈ Z+ . Then ∼ is an equivalence relation.

2.8 The Symmetric Group Redux

The equivalence classes would have the following form, of course:

[i] = {i, σ(i), σ 2 (i), σ 3 (i), . . .}.

Example 2.8.2. Let σ ∈ S8 be as above. What are the equivalence classes? We

have

[1] = {1, 5, 8}
[2] = {2, 3}
[4] = {4}
[6] = {6}
[7] = {7}

We can use these equivalence relations to break up σ into pieces called cycles. In
the notation that we are developing, we will write

σ = (1 5 8)(2 3)(4)(6)(7)

Note that we have written the integers in each cycle in the order in which they are
hit by σ. That is, (1 5 8) and (1 8 5) are not the same permutation. We’ll usually
suppress the numbers that are fixed by σ, since they information that they encode
is extraneous. Therefore, we will simply write

σ = (1 5 8)(2 3).

We’ve been throwing around the word “cycle” without properly defining it. You
probably have the idea already, but we should still be rigorous.

Definition 2.8.3. A permutation of the form

σ = (i1 i2 · · · ik ),

where σ(im ) = im+1 for 1 ≤ m < k and σ(ik ) = i1 is called a k-cycle.

The cycles are important in that they are the building blocks of Sn . When
finding the cycle decomposition of a permutation σ, we are really factoring σ into
cycles, in much the same way that one factors integers into primes.

Example 2.8.4. Let’s consider the permutation τ ∈ S9 given by

 
1 2 3 4 5 6 7 8 9
τ= .
4 9 6 3 7 1 8 2 5

How do we write τ as a product of cycles? We start with 1, and we find the cycle
determined by it:
 Group Theory

 
 1 2 3 4 5 6 7 8 9 
 
τ= 



 
4 9 6 3 7 1 8 2 5

Thus we have the cycle

(1 4 3 6).
Now we go to the next smallest integer, which is 2, and we find its cycle:
 
 1 2 3 4 5 6 7 8 9 
 
τ=  


 
4 9 6 3 7 1 8 2 5

so the cycle is
(2 9 5 7 8).
Thus
τ = (1 4 3 6)(2 9 5 7 8).
Example 2.8.5. Let’s find the cycle decomposition of σ ∈ S5 , where
 
1 2 3 4 5
σ= .
2 3 1 5 4
Well, the cycle determined by 1 is
(1 2 3),
and the smallest integer left is 4, whose cycle is
(4 5).
Therefore,
σ = (1 2 3)(4 5).
Remark 2.8.6. Note that we have written our permutations in such a way that the
cycles always start with the smallest possible integer. This is the standard way to
write permutations in cycle notation, but it is not necessary to do so. For example,
the cycles (1 2 3) and (3 1 2) are the same. You could think of them both as
encodings of the picture
1

2 3
Chapter 3

Ring Theory

We’re now done with our study of groups (at least for the sake of groups). The plan
now is to move on to more complicated algebraic structures, namely objects called
rings.

3.1 Rings
You’ll probably agree that groups were kind of strange at first. We had a set with
a single binary operation that satisfied certain nice properties, and we were able to
come up with many examples of such things. Some of these examples were ones we
already knew, such as Z, R, and Q, all under addition. However, these examples all
really have two binary operations, namely addition and multiplication.

Example 3.1.1. We already know that Z forms an abelian group under addition.
What nice properties are satisfied by multiplication on Z?

• associativity

• commutativity

• identity

• distributivity (if a, b ∈ Z and n ∈ Z, then n(a + b) = na + nb)

This will mean that Z is really a model example of a ring.

Definition 3.1.2. A ring is a set R equipped with two binary operations, denoted
by + and ·, such that

1. hR, +i is an abelian group:

(a) + is associative and commutative.

(b) There is an additive identity 0 ∈ R such that 0 + a = a for all a ∈ R.

111
 Ring Theory

(c) For each a ∈ R, there is an additive inverse −a ∈ R so that a + (−a) = 0.

2. The multiplication operation · is associative.

3. (Distributive law) For all a, b, r ∈ R, we have

r · (a + b) = r · a + r · b

and
(a + b) · r = a · r + b · r.

You’ll probably notice that we left two things off this list. We did not require
that multiplication be commutative, or that there is even a multiplicative identity.
Rings that have these properties are special, and thus have special names.
Definition 3.1.3. A commutative ring is a ring R for which

a·b=b·a

for all a, b ∈ R.
Definition 3.1.4. A ring with identity1 (also called a ring with unity or a
unital ring) is a ring R which contains an element 1 ∈ R (with 1 6= 0) satisfying

1·a=a·1=a

for all a ∈ R.
Before we proceed, let’s look at some familiar (and less familiar) examples of rings.
Example 3.1.5. 1. We already saw that Z is a ring, and actually a commutative
ring with identity.

2. Similarly, Q, R, and C are all commutative rings with identity with respect to
their usual operations.

3. For any n, Zn is a commutative ring with identity with respect to modular

addition and multiplication.

4. Let’s try a noncommutative example. Recall that Mn (R) is the set of all n × n
matrices with real coefficients. Then Mn (R) is a ring with respect to matrix
addition and multiplication. It is noncommutative, but it has an identity,
namely the identity matrix I.

5. Let 2Z be the set of all even integers. Then 2Z is a commutative ring with
respect to the usual arithmetic. It does not have an identity, however, since
1 6∈ 2Z.
1
Some mathematicians require that a ring contains an identity element. That is, they use “ring”
to mean “ring with identity.” Some have suggested the term rng to mean a ring without identity.
3.2 Basic Facts and Properties of Rings

6. Let C([0, 1], R) = {f : [0, 1] → R : f is continuous}. Then C([0, 1], R) is a ring

with respect to pointwise addition and multiplication:

f + g(x) = f (x) + g(x)

and
f g(x) = f (x)g(x).
It is also commutative, and its multiplicative identity is the function which is
identically 1.

7. This will be one of the main examples that we’ll study. Let Q[x] be the set
of all polynomials with rational coefficients. Then Q[x] forms a ring: we add
two polynomials by adding their coefficients, say
   
2 1 7 13 3
2x + 3x + + 5x + x + 2 = 5x3 + 2x2 + x + .
3
2 2 2 2
We multiply them by “foiling”:

(x3 + 1)(3x2 + 4x + 2) = 3x5 + 4x4 + 2x3 + 3x2 + 4x + 2.

This turns Q[x] into a commutative ring with identity. (We could also do this
with any commutative ring in place of Q.)

3.2 Basic Facts and Properties of Rings

We’ll soon start investigating particular types of rings. Before we can do this,
we should prove some relatively simple facts about rings which will be needed in
computations.
Proposition 3.2.1. Let R be a ring. Then for all a, b ∈ R, we have:
(a) 0 · a = a · 0 = 0;

(b) (−a)b = a(−b) = −(ab);

(c) (−a)(−b) = ab; and

(d) if 1 ∈ R, then (−1)a = a(−1) = −a.

Proof. (a) If a ∈ R, then we have

0 · a = (0 + 0) · a = 0 · a + 0 · a

by the right distributive law. But this means that 0 · a is an additive idempotent in
the abelian group hR, +i. There is only one such element, so 0 · a = 0. A similar
argument works to show that a · 0 = 0.
 Ring Theory

(b) Let a, b ∈ R. Then by distributivity,

ab + (−a)b = (a + (−a))b = 0 · b = 0,
so (−a)b = −(ab). The same sort of argument works to show that a(−b) = −(ab).
(c) If we apply part (b), we have
(−a)(−b) = −((−a)b) = −(−(ab)) = ab.
(d) Again by part (b), (−1)a = −(1 · a) = −a, and a(−1) = −(a · 1) = −a.

Unlike with groups, there are many different types of rings that one can con-
sider. This arises from the fact that there are some strange things that can happen
regarding the multiplication in a ring.
   
1 0 0 0
Example 3.2.2. Let R = M2 (R), and let A = and B = . Then
0 0 0 1
    
1 0 0 0 0 0
AB = = ,
0 0 0 1 0 0
but neither A nor B is the zero matrix. Such anomalies have a special name in ring
theory.
Definition 3.2.3. An element a 6= 0 of a (commutative) ring R is called a zero
divisor if there exists b ∈ R such that ab = 0.
Many of the results that we will obtain will involve commutative rings which are
free of zero divisors.
Definition 3.2.4. A commutative ring R is called an integral domain if R con-
tains no zero divisors. Equivalently, if a, b ∈ R with ab = 0, then either a = 0 or
b = 0.
Example 3.2.5. 1. The rings Z, Q, R, and C are all integral domains.
2. When n is composite, Zn is not an integral domain. For example, in Z6 ,
2 · 3 = 0.

3. The function ring C([0, 1], R) is not an integral domain. Let

(
0 x < 1/2
f (x) =
x − 1/2 x ≥ 1/2
and (
x − 1/2 x < 1/2
g(x) =
0 x ≥ 1/2.
Then f (x)g(x) = 0 for all x ∈ R, but neither function is identically zero.
3.2 Basic Facts and Properties of Rings

Another thing you may have noticed is that elements of rings (with identity)
do not necessarily possess multiplicative inverses. The ones that do have a special
name.
Definition 3.2.6. Let R be a ring with identity. An element a ∈ R is called a unit
if there is a b ∈ R such that
ab = ba = 1.
(We usually write b = a−1 .) The set of all units is denoted by R× , called the unit
group2 of R.
Example 3.2.7. Let’s compute the unit groups in some rings.
1. What are the units in Z? The only integers which have multiplicative inverses
are 1 and −1, so
Z× = {1, −1}.

2. What is the unit group in Q? Every nonzero rational number has a multi-
plicative inverse, so
Q× = Q − {0}.
Conveniently enough, this is the notation that we have already used for the
group of nonzero rational numbers under multiplication. Similarly, we have
R× = R − {0} and C× = C − {0}.

3. Let’s consider the ring Zn for n ∈ Z. We proved a long time ago that the
elements of Zn that have multiplicative inverses are precisely those which are
relatively prime to n. Therefore,

Z×
n = {a ∈ Zn : gcd(a, n) = 1}.

Note that if p is prime, then Z×

p = {1, 2, . . . , p − 1} = Zp − {0}.

Of course 0 never has a multiplicative inverse, but we’ve seen that it’s possible
that everything else might. Such rings are special.
Definition 3.2.8. A ring R with identity is called a division ring if every nonzero
a ∈ R is a unit. Equivalently, R× = R − {0}.
Definition 3.2.9. A commutative division ring is called a field.
Example 3.2.10. 1. The integral domains Q, R, and C are all fields.

2. Since only 1 and −1 are units, Z is not a field.

3. If p is prime, Zp is a field. It is an example of a finite field.

2
This name should indicate that R× forms a group under multiplication. This fact is not too
hard to check.
 Ring Theory

Exercise 3.1. Let R be a commutative ring with identity. Show that if u ∈ R is

a unit, then u cannot be a zero divisor. As a consequence, any field is an integral
domain.

3.2.1 The Quaternions

You may have noticed that we produced several examples of fields (and hence of
division rings), but no division rings that weren’t fields. There is an example of such
a thing, and it’s an interesting one mathematically and physically.

Example 3.2.11. The quaternions were invented by the Irish mathematician and
physicist William Rowan Hamilton in the middle of the 19th century. Motivated
by the complex numbers and their geometric interpretations, he had tried to define
multiplication on triples of real numbers. (That is, he tried to turn R3 into a ring.)
This didn’t quite work (and, in fact, is impossible), but he had a breakthrough one
day while walking around campus. It was so exciting to him that he actually carved
his initial identity into a stone bridge.
Hamilton’s idea was to consider three elements i, j, and k which behave like the
imaginary unit in C:
i2 = j 2 = k 2 = ijk = −1.

From this single identity, it is possible to deduce that

ij = k, jk = i, ki = j,

and that these elements anticommute:

ij = −ji, kj = −kj, ik = −ki.

The quaternions are then defined to be the set

H = {a1 + a2 i + a3 j + a4 k : a1 , a2 , a3 , a4 ∈ R} .

In other words, one can think of H as a four-dimensional real vector space with
basis {1, i, j, k}. The addition operation works just like addition of vectors in R4 :
for example,

(1 + 4i + 7j + 2k) + (−2 + 6i − 9j − 4k) = −1 + 10i − 2j − 2k.

In general, we would have

(a1 +a2 i+a3 j+a4 k)+(b1 +b2 i+b3 j+b4 k) = (a1 +b1 )+(a2 +b2 )i+(a3 +b3 )j+(a4 +b4 )k.
3.2 Basic Facts and Properties of Rings

Multiplication works like multiplication of complex numbers, but more complicated.

We simply multiply everything out, and then use the quaternion identities listed
above to simplify the result:
(3 − 5k)(1 + 4i − 5j) = 3 + 12i − 15j − 5k − 20ki + 25kj
= 3 + 12i − 15j − 5k + 20j − 25i
= 3 − 13i + 5j − 5k
There is a general formula for the coefficients of the product of two quaternions, but
it’s usually easier to simply multiply on a case-by-case basis:
(a1 + a2 i + a3 j + a4 k)(b1 + b2 i + b3 j + b4 k) = c1 + c2 i + c3 j + c4 k,
where
c1 = a1 b1 − a2 b2 − a3 b3 − a4 b4
c2 = a1 b2 + a2 b1 + a3 b4 − a4 b3
c3 = a1 b3 − a2 b4 + a3 b1 + a4 b2
c4 = a1 b4 + a2 b3 − a3 b2 + a4 b1 .
It’s possible to check that multiplication is associative and distributive, which we
won’t do here. Also, note that for example we have
(1 + 3i − 4j − 10k)(1 − 3i + 4j + 10k) = 1 − 3i + 4j + 10k + 3i − 9i2 + 12ij + 30ik
− 4j + 12ji − 16j 2 − 40jk − 10k + 30ki − 40kj − 100k 2
= 1 + 9 + 12k − 30j − 12k + 16 − 40i + 30j + 40i + 100
= 1 + 9 + 16 + 100
= 126.
In general, one can verify that
(a1 + a2 i + a3 j + a4 k)(a1 − a2 i − a3 j − a4 k) = a21 + a22 + a23 + a24 .
This allows us to define the inverse of any nonzero quaternion: if a1 +a2 i+a3 j+a4 k 6=
0, then the quantity
α = a21 + a22 + a23 + a24
is nonzero, and it can be checked that
a a2 a3 a4 
1
(a1 + a2 i + a3 j + a4 k) − i − j − k = 1.
α α α α
Thus H× = H − {0}, and H is a division ring. It is not a field, however, since we
have ij 6= ji, for example.
Remark 3.2.12. Hamilton’s quaternions may seem somewhat contrived, but they
are more than just a method for multiplying vectors in R4 together. They are
actually important in physics, since they can be used to model rotations in R3 .