MATH 6101: Fall 2022

Course Notes
Robert Won

These are course notes from MATH 6101 (Algebra I) at the George Washington University taught
during Fall 2022. The textbook was Abstract Algebra by Dummit and Foote.

To my students: If you find errors or typos, please let me know! I can correct them.

Contents

0. Course Logistics 3

1. Tuesday 8/30: Basic Axioms, The Dihedral Group 4

2. Thursday 9/1: More Examples, Homomorphisms 10

3. Wednesday 9/7: Group Actions, Subgroups, Normalizers, Centralizers 18

4. Monday 9/12: Cyclic Groups, Subgroup Generation 24

5. Wednesday 9/14: More on Subgroup Generation, Quotient Groups 32

6. Monday 9/19: Normal Subgroups and Lagrange’s Theorem 41

7. Wednesday 9/21: The Isomorphism Theorems, Composition Series 48

8. Monday 9/26: Composition Series and Solvable Groups 54

9. Wednesday 9/28: Alternating Groups and Group Actions 57

10. Monday 10/3: Conjugacy Classes in Sn and the Simplicity of A5 64

11. Wednesday 10/5: Sylow’s Theorem 72

12. Wednesday 10/12: Applications of Sylow and FTFGAG 76

Friday 10/14: Exam 1 84

13. Monday 10/17: Semidirect Products 84

14. Wednesday 10/19: Semidirect Product Recognition, p-Groups 91

15. Wednesday 10/26: Nilpotent Groups, Free Groups, Presentations 95

16. Monday 10/31: Free Groups, Introduction to Rings 100

17. Wednesday 11/2: Basic Properties, Homorphisms 107

18. Monday 11/7: Ideals and Isomorphism Theorems 113

1
19. Wednesday 11/9: Zorn’s Lemma 119

20. Wednesday 11/16: More on Localization 125

Friday 11/18: Exam 2 130

21. Monday 11/21: Proof of CRT, Euclidean Domains, PIDs 130

Wednesday 11/23: Thanksgiving Break 136

22. Monday 11/28: PIDs and UFDs 137

23. Wednesday 11/30: Polynomial Rings in Several Variables 142

24. Monday 12/5: Irreducibility in Polynomial Rings 148

25. Wednesday 12/7: Hilbert’s Basis Theorem 155

2
 0. Course Logistics

• Q: What is the point of this course?

This is the first semester of graduate level algebra. In algebra, we study the abstract
structure of mathematical objects. Many objects of study in mathematics have some kind of
algebraic structure, so it is useful to know algebra no matter what kind of mathematics you
are interested in: see e.g., algebraic geometry, algebraic combinatorics, algebraic number
theory, and algebraic topology.
One of the main goals of the course is to prepare you for the qualifying exam in algebra.
As I prepare lectures, problem sets, and exams, this will be the main goal in my mind.
• Logistical details: The textbook is important, and contains many many many examples
and problems. I will have several office hours each week, which you are always welcome to
attend (no specific reason or appointment needed). You can also contact me by e-mail if
there’s something you want to discuss in private.
I will be regularly updating these typed up notes (I’m writing them while teaching, so
as the semester progresses, this document will continue growing). I will upload an updated
document to Blackboard at least once a week.
You will have weekly problem sets which will be posted to Blackboard. There will be
two midterm exams, and a final exam. I will give you more information as we get closer to
the first midterm.

3
 1. Tuesday 8/30: Basic Axioms, The Dihedral Group

Definition 1. Let G be a set.

(1) A binary operation ∗ on G is a function ∗ : G × G → G. For any a, b ∈ G, we write a ∗ b for

∗(a, b).
(2) We say ∗ is associative if a ∗ (b ∗ c) = (a ∗ b) ∗ c for all a, b, c ∈ G.
(3) If, for a, b ∈ G, a ∗ b = b ∗ a, we say a and b commute. If this holds for all a, b ∈ G, we say
∗ is commutative.

Example 2. (1) Usual addition and multiplication on Z, Q, R, C are associative, commu-

tative binary operations.
(2) Is − a binary operation on Z, Q, R, C? Yes, but not commutative nor associative.
(3) Is ÷ a binary operation on C? No, not defined for all elements of C. It is on C∗ .
(4) Matrix multiplication is an associative noncommutative binary operation on M2 (C).

Definition 3. Let ∗ be a binary operation on a set G. Let H ⊆ G be a subset. We can restrict ∗

to the set H. If, for all a, b ∈ H, a ∗ b ∈ H, then we say that H is closed under ∗.

Example 4. (Viewed as subsets of Z), Z≥0 and Z>0 are closed under addition and multiplica-
tion. Z≤0 is closed under addition but not multiplication.

Definition 5. A group G is a set with a binary operation ∗ such that:

(i) ∗ is associative,
(ii) there exists an e ∈ G called the identity such that for all g ∈ G, g ∗ e = e ∗ g = g,
(iii) For all g ∈ G, there is some h ∈ G such that gh = e = hg. We usually write g −1 for this
element unless ∗ is + in which case we write −g.

If ∗ is commutative we say that G is commutative or abelian.

Note. In general, if you use the sign + for your operation, it is commutative.

4
 Example 6. (1) (Z, +), (R, +), (Q, +), (C, +) are abelian groups with e = 0 and inverse of
g given by −g.
(2) Q∗ := Q \ {0}, R∗ := R \ {0}, and C∗ := C \ {0} are groups under ×. Why isn’t Z \ {0}?
(3) For n ∈ Z≥0 , Z/nZ is an abelian group under addition of residue classes. We haven’t
shown that there is a well defined addition operation on this. But if you accept that,
then the associativity follows from associativity of addition in Z.
(4) Let F be a field (R or Q or C). Why isn’t M2 (F ) the set of 2 × 2 matrices? Not enough
inverses, but if we take only the ones with inverses, we get the general linear group
GL2 (F ).
(5) Let X be any set, then SX := {all bijective functions f : X → X} is a group under
function composition. We call this the symmetric group on X. Every bijective function
can be thought of as a permutation of the elements of X.
(6) When X = {1, . . . , n} we write Sn := SX .

Prove the following as an exercise (the proofs are in your book but you should be able to do them!)

Proposition 7. If G is a group under ∗ then:

(1) the identity of G is unique

(2) for each a ∈ G, a−1 is unique
(3) (a−1 )−1 = a for all a ∈ G
(4) (a ∗ b)−1 = b−1 ∗ a−1 ,
(5) for any a1 , . . . , an ∈ G, the value of a1 ∗ · · · ∗ an is independent of bracket. (This is the only
annoying one. You need to use induction.)

Proof. Example for (1). If f, g ∈ G are both identity elements, then f = f ∗ g = g. □

Notation. Henceforth, unless we want to be particularly precise, we write a ∗ b as just ab. The
identity is often denoted 1, unless the operation is + in which case identity is denoted 0. By part
(5) of the above proposition, xx · · · x = xn , and x−1 x−1 · · · x−1 = x−n .

Definition 8. For G a group and g ∈ G, the order of g is the smallest positive integer n such that
g n = e. This is denoted |g|. If no positive power of g is the identity, then g has infinite order.

5
Definition 9. For G a group, the order of G is the cardinality of G and is denoted |G|.

Example 10. • In Z/nZ under +, the order of an element m is given by |m| =

n/ gcd(m, n).
• |Sn | = n!

Example 11. Let G and H be groups. Then the direct product G × H of G and H is a group
whose elements are ordered pairs (g, h) where g ∈ G and h ∈ H. The operation in G × H
happens component-wise, that is

(g, h)(g ′ , h′ ) = (gg ′ , hh′ ).

Associativity follows from the fact that G and H are groups. What is the identity? What is
the inverse of (g, h)?

The Dihedral Groups.

Now that we’ve defined the structure of a group abstractly (via axioms), we would like some more
examples of groups. It will (very shortly) be very fruitful to study how groups act on other sets. An
important family of examples of groups are also best defined as symmetries of geometric objects.

Definition 12. For each n ∈ Z, n ≥ 3, let D2n be the set of symmetries of a regular n-gon (under
composition). This group is called the dihedral group of order 2n. (Be careful! Sometimes this
group is called Dn . But more often the subscript refers to the order of the group, which is 2n.)

A symmetry of the n-gon is any rigid motion which arises by taking a copy of the n-gon, moving
it in any way in 3-space and then moving it back to the original n-gon so it exactly covers it. (We
can demonstrate this with a paper triangle).

Important to note here, the elements of the group are the symmetries of the n-gon, not the vertices
of the n-gon. So one element of D2n corresponds to clockwise rotation by 2π/n radians. Is it clear
that if you compose two symmetries then you get another symmetry? What is the identity? What
is the inverse of the rotation? This operation is associative since function composition is always
associative.
6
Fact. |D2n | = 2n.

Proof. We can label the vertices of the n-gon with the integers {1, . . . , n}. Then any symmetry of
the n-gon can be described by tracking where its vertices go.

For any vertex i, there is a symmetry which sends vertex 1 to vertex i. Once you have specified
where vertex 1 goes, the symmetry is completely determined by where vertex 2 goes, and since
symmetries are rigid motions, vertex 2 can either go to vertex i − 1 or i + 1. It is easy to see that
such symmetries both exist, so the order of D2n is n · · · 2 = 2n. □

We can explicitly describe all the symmetries in D2n . There are the n symmetries of rotation
by 2πi/n radians for i = 0, . . . , n − 1. There are also n symmetries by reflection across a line of
symmetry. By the above proof, we have found all of the symmetries of the n-gon.

Since any symmetry of the n-gon can be described by tracking where its n vertices go, we can
embed D2n into Sn . A symmetry s ∈ D2n corresponds to the permutation σ ∈ Sn where if s takes
vertex i to vertex j, σ(i) = j.

There is a standard way to express the dihedral group as an abstract group. Let r be the rotation
clockwise by 2π/n radians. Let s be the reflection through the line of symmetry going through
vertex 1. The following facts are true

(1) 1, r, r2 , . . . , rn−1 are all distinct and rn = 1, so |r| = n.

(2) |s| = 2.
(3) s ̸= ri for any i.
(4) sri ̸= srj for any i ̸= j so

D2n = {e, r, r2 , . . . , rn−1 , s, sr, . . . , srn−1 },

i.e. each element can be written uniquely in the form sk ri for some k = 0 or 1 and
0 ≤ i ≤ n − 1.
(5) rs = sr−1 .
(6) ri s = sr−i .

Using these facts, we can write every element of D2n as sk ri , and we can completely describe the
multiplication in D2n by using the “relations” in (1), (2), and (6). For example, if n = 5,

(sr4 )(sr3 ) = s(sr−4 )r3 = s2 r−1 = r4 .

This brings us to the subject of presenting groups via generators and relations. We won’t do this in
total formality until chapter 6, since it requires some sophisticated technology to describe rigorously.
The idea is based on what we did with the dihedral group above. We found group elements r and
7
s such that every element of D2n could be written in terms of r and s, and then we found what
“relations” r and s satisfied (which allowed us to rewrite products when necessary).

Definition 13. A subset S ⊆ G such that every element of G can be written as a finite product
of elements of S and their inverses is called a set of generators of G. Notationally, G = ⟨S⟩.

Example 14. • The integer 1 is a generator for Z under +.

• We showed that {r, s} is a set of generators for D2n .
• 1 generates Z/nZ for any n. In general, j will generate Z/nZ if and only if gcd(n, j) = 1.
We will see this soon when we talk about cyclic groups.
• In Z, {a, b} will generate the same set as gcd(a, b). You can prove this using the
Euclidean algorithm.

Any equations that the generators satisfy are called relations. As an example, in D2n , rn = 1,
s2 = 1, and rs = sr−1 were relations. It turns out that every other equation satisfied by elements
of D2n are derived from these three.

If G is a group with a set of generators S and some relations R1 , . . . , Rm such that every equation
among the elements of S can be deduced from these relations, then we call the generators and
relations a presentation of G and write

G = ⟨S | R1 , . . . , Rm ⟩.

Example 15. D2n = ⟨r, s | rn = s2 = 1, rs = sr−1 ⟩.

Notice that this is an abstract way to work with D2n , rather than having to think about an n-
gon. Presentations of a group can be very useful when you’re doing computations (for example,
multiplication in D2n is easier with the presentation than having to think about an n-gon moving in
3-space). On the other hand, presentations can be very tricky. Just from looking at a presentation,
it is hard to tell when groups are the isomorphic, hard to tell the order of the elements in the group,
even whether the group is finite or not. This is because there can be hidden relations which are
implied by the set of relations in the presentation

8
 Example 16. Let’s define the group

X2n = ⟨x, y | xn = y 2 = 1, xy = yx2 ⟩.

Just as an example, this has the relation x = xy 2 , since y 2 = 1. Therefore

x = xy 2 = (xy)y = (yx2 )y = (yx)(xy) = (yx)(yx2 ) = y(xy)x2 = y(yx2 )x2 = x4 ,

so x4 = x and hence x3 = 1. This wasn’t obvious from the initial presentation!

There will be a problem set exercise showing that a presented group ends up being trivial.

9
 2. Thursday 9/1: More Examples, Homomorphisms

Recall the definition of the symmetric group from Example 6. Let X be any nonempty set and SX
the set of all bijections X → X, which can be thought of as permutations of X. (you can draw any
bijection with arrows, then untangle them to get a new permutation).

Then SX is a group under function composition: ◦. This is a binary operation on SX : if σ, τ ∈ SX ,

then they are bijections X → X so σ ◦ τ : X → X is a bijection. This operation is associative since
function composition is. Is it commutative? Why not?

The identity of SX is the permutation 1 defined by 1(x) = x for all x ∈ X. For every permutation
σ ∈ SX , there is a two-sided inverse σ −1 : X → X satisfying σ ◦ σ −1 = σ −1 ◦ σ = 1. This group SX
is called the symmetric group on X.

Just as with the dihedral group, where the group elements were the symmetries not the vertices,
the elements of SX are the permutations, not the elements of X.

When X = {1, . . . , n}, we use the notation Sn , and call it the symmetric group of degree n. This is
a very interesting group, and a fundamental example.

Fact. |Sn | = n!

Proof. We need to count the number of bijections {1, . . . , n} → {1, . . . , n}. Let σ ∈ Sn . Then σ is
determined by the image of each of the elements of {1, . . . , n}. Since the set is finite, every injection
is a bijection. So we have n choices for σ(1), but only n − 1 choices for σ(2) and so on. Hence,
there are n! bijections. □

We need some nice notation to write the elements of Sn . One common way (and the most common
in algebra) is called cycle notation. (There is other notation. Some combinatorialists like using
two-line notation. There is also one-line notation.)

Definition 17. A cycle is a string of integers which represents the element of Sn which cyclically
permutes these integers and fixes all others. The cycle (a1 a2 . . . am ) where ai ∈ {1, . . . , n} are
distinct is the permutation which sends ai to ai+1 for 1 ≤ i ≤ m − 1 and sends am to a1 .

Example 18. In S5 , the cycle (1 3 4) sends 1 to 3, 3 to 4, 4 to 1 and fixes 2 and 5.

A general element σ ∈ Sn can be written as a product of cycles.

10
It is easiest to do this via example so we don’t need to worry about totally general indices. So let’s
write some permutation in σ ∈ S7

σ(1) = 3, σ(2) = 5, σ(3) = 6, σ(4) = 4, σ(5) = 7, σ(6) = 1, σ(7) = 2.

Start with 1, which gets mapped to 3, so our first cycle starts (1 3. Where does 3 go? It goes to 6
which goes to 1 so we have completed the cycle (1 3 6). This of course does not describe the whole
permutation σ. So start with the next lowest number, which is 2. Repeat the above process to get
(2 5 7). Finally, 4 is fixed. So we can write this permutation

σ = (1 3 6)(2 5 7)(4).

Definition 19. The length of a particular cycle is the number of integers which appear in it, and
a cycle of length t is called a t-cycle. Two cycles are disjoint if they have no integers in common.

The example above is the product of three disjoint cycles, two 3-cycles and a 1-cycle. We do not
write 1-cycles, and understand that any integer not appearing in the cycle decomposition is fixed
by the permutation. The identity permutation is denoted by just 1.

This notation is nice and compact. Also, it gives a nice embedding of Sn in Sm for m ≥ n. For
example, (1 2) transpose 1 and 2 in Sn for all n ≥ 2.

So for any permutation in the symmetric group Sn , we can record that permutation in cycle
notation. What about products? The standard convention is to read cycles from right to left, just
like function composition (after all, f ◦ g means first do g then do f ). So you follow the elements
under successive permutations. For example:

(1 3 2)(4 5) ◦ (1 3)(2 4) = (1 2 5 4).

Now that we have products, it is also easy to write down inverses. To write σ −1 , just write each
integer in the cycle decomposition of σ in reverse order. So for the cycle σ = (1 2 5 4), we have
σ −1 = (4 5 2 1). It is easy to check that σ ◦ σ −1 = 1.

Fact. For n ≥ 3, S3 is nonabelian.

Proof. Follows since (1 2) ◦ (1 3) = (1 3 2) but (1 3) ◦ (1 2) = (1 2 3). □

Fact. You can cycle the integers appearing in a cycle to get the same cycle, i.e. (1 2 3) = (2 3 1) =
(3 1 2). By convention, we usually put the smallest number in each cycle first.
11
Fact. The cycle decomposition of an element σ ∈ Sn is a product of disjoint cycles. Up to
rearranging the cycles and cyclically permuting the numbers in each cycle, the cycle decomposition
is unique.

Recall the definition of a field from undergraduate algebra (fields are covered in much more detail
in the next semester of graduate algebra, so in this semester it is reasonable to usually take F = R).

Definition 20. A field is a set F together with two binary operations + and ∗ such that (F, +)
is an abelian group with identity 0 and (F \ {0}, ∗) is an abelian group with the distributive law
a ∗ (b + c) = a ∗ b + a ∗ c for all a, b, c ∈ F . That is, F is a ring in which every nonzero element has
a multiplicative inverse.

We use the notation F × for F \ {0}. The fields one most commonly encounters in practice are Q,
R, C, and Z/pZ for prime p. The latter we sometimes denote Fp . (It is useful to have an example
of finite vector spaces or finite groups of matrices).

In undergraduate linear algebra, you learned all about vector spaces over R, matrices, linear trans-
formations, determinants, etc. All of this works over any field F .

Definition 21. For each n ∈ Z+ , let GLn (F ) be the set of all n × n matrices whose entries come
from F and whose determinant is nonzero

GLn (F ) = {A | A is an n × n matrix with entries from F and det(A) ̸= 0}

the operation is matrix multiplication. Since det(AB) = det(A) det(B), GLn (F ) is closed under
matrix multiplication. Also, det(A) ̸= 0 if and only if A has a matrix inverse. The identity is the
n × n identity matrix I.

This is called the general linear group of degree n.

The general linear group is yet another example of a nonabelian group for you to keep in mind,
along with D2n and Sn . Our last example at the start of the course is:

The Quaternion Group.

Definition 22. The quaternion group Q8 is defined by

Q8 = {1, −1, i, −i, j, −j, k, −k}

12
with product defined by

−1 · −1 = 1 1·a=a·1 − 1 · a = a · −1 = −a for all a ∈ Q8 .

i · i = j · j = k · k = −1
and
i·j =k j · i = −k
j·k =i k · j = −i
k·i=j i · k = −j.

We can also represent Q8 with the following eight matrices in GL2 (C):
" # " # " # " #
1 0 0 1 0 i i 0
± ,± ,± ,± .
0 1 −1 0 i 0 0 −i

Group Homomorphisms.
One of the major themes in modern mathematics is that if you want to understand some sort
of structure (e.g., vector spaces, topological spaces, schemes, graphs), you gain an extraordinary
amount of insight by studying not just the structures but also the functions between the structures.
You also want your functions to play nice with whatever structures you are playing with (e.g., linear
transformations, continuous maps, regular maps, graph homomorphisms).

Indeed, much of modern mathematics is written in the language of category theory. A category
contains the data of objects and morphisms (maps) between objects. In the category of vector
spaces, the objects are vector spaces and you also record the space of linear transformations between
any two vector spaces. In the category of topological spaces, the objects are topological spaces and
the morphisms are continuous maps. In the category of groups, we study groups together with
group homomorphisms.

Definition 23. Let (G, ∗) and (H, ·) be groups. A map φ : G → H is called a (group) homomor-
phism if
φ(x ∗ y) = φ(x) · φ(y)
for all x, y ∈ G.

Another way to write this is φ(xy) = φ(x)φ(y), but it is important to remember that the multipli-
cation on the left hand side happens in G and the multiplication on the right hand side happens in
H. A homomorphism is a function that respects the group structures of the domain and codomain.
13
 Example 24. F a field, det : GLn (F ) → F × is a homomorphism because det(AB) =
det(A) det(B).

Definition 25. If φ : G → H is a homomorphism, the kernel of φ is the set

{g ∈ G | φ(g) = e}

denoted ker φ.

Later on in this semester, we will study rings and ring homomorphisms (which play nice with
ring-theoretic structure). Next semester you will also study fields and field homomorphisms.

The next definition captures the notions of two groups being “the same”

Definition 26. An isomorphism G → H is a bijective homomorphism. We say that G and H are

∼ H.
isomorphic and write G =

Intuitively, this means that G and H are the same group except that the elements and operations
may be written differently in H. Anything which depends only on the group structure of G also
holds in H.

Example 27. The exponential map φ : R → R+ defined by φ(x) = ex is an homomorphism

from (R, +) to (R+ , ·) because ex+y = ex ey .

To show that it is an isomorphism, you could show directly that exponentiation is both bijective
and surjective.

You could also show that φ has a two-sided inverse, namely ln : R+ → R, which is a group
homomorphism since ln(xy) = ln x + ln y for all x, y ∈ R+ . Since φ ◦ ln = idR+ and ln ◦φ = idR ,
this shows that both φ and ln are bijections.

In general, the most common technique you will use to show that two groups G and H are isomor-
phic, you construct a map φ : G → H and show it is a homomorphism and is bijective.
14
How can you tell if two groups are not isomorphic? It turns out that this is, in general, a very
difficult question. It is much harder to show that there cannot possibly exist a bijective homomor-
phism G → H than to construct one. Indeed, in many fields of mathematics, it is not easy to say
that two structures are not the same. For example, it is difficult to tell whether some knot you’ve
drawn is different from the unknot.

In order to tell that two groups are not isomorphic, you need to understand the groups. This is
because isomorphic groups will have all of the same group-theoretic properties. If you can identify
a property where G and H differ, then they cannot be isomorphic.1

What properties are preserved under isomorphism? Basically any of the ones that only rely on the
group-theoretic structure of the groups in question. If φ : G → H is an isomorphism then

(1) |G| = |H|,

(2) G is abelian if and only if H is abelian,
(3) for all g ∈ G, |g| = |φ(g)|,
(4) (we have not introduced the center yet but) the centers Z(G) and Z(H) have the same
cardinality,
(5) etc.

So we know that S4 ∼ ̸ D8 , since |S4 | = 24 but |D8 | = 8. How do we know D8 ∼

= ̸ Q8 ? Think about
=
this, and if you can’t figure it out, ask me next time.

Homomorphisms from presented groups.

If you have a presentation of a group G = ⟨r1 , . . . , rm | R⟩ and let s1 , . . . , sm be elements of some
group H. If every relation in the ri is satisfied by the si then you can define a unique homomorphism
φ : G → H mapping ri to si for each i. If H is generated by the si , the φ is surjective. If
|G| = |H| < ∞ then any surjective morphism is necessarily bijective so G ∼
= H.

Example 28. Recall that D2n = ⟨r, s | rn , s2 = 1, sr = rs−1 ⟩. Suppose H is a group con-
taining elements a and b such that an = 1, b2 = 1, and ba = a−1 b. Then there is
a homomorphism D2n → H mapping r to a and s to b. Let k | n and k ≥ 3. Let
D2k = ⟨r1 , s1 | r1k = s21 = 1, s1 r1 = r1 s−1
1 ⟩. Define φ : D2n → D2k by φ(r) = r1 and φ(s) = s1 .
Easy to check that the relations in D2n are satisfied by r1 and s1 . So φ extends to a unique
homomorphism. Since {r1 , s1 } generate D2k , the map is surjective.

1Sometimes you can attach an invariant to a structure which does not change under isomorphism. For example,
in topology, the most basic invariant is the fundamental group. This is a group that you can compute from a
topological space. If two topological spaces have different fundamental groups, then they cannot be homeomorphic.
If two topological spaces have the same fundamental group, they may or may not be homeomorphic. In knot theory,
Vaughn Jones won the Field’s Medal partially for introducing the Jones polynomial, an important knot invariant.
15
Note that if k < n, then φ(r) does not have order n, but this is not required (φ(r) only needs to
have order dividing n). Also, the map is not an isomorphism if k < n since |D2n | > |D2k |.

In the above example, we can also map φ(s) = 1 and get a nonsurjective, noninjective homomor-
phism.

Example 29. Let G = D6 and H = S3 . Check that in S3 , the elements a = (1 2 3) and

b = (1 2) satisfy a3 = b2 = 1 and ba = a−1 b. Hence, there is a homomorphism ψ : G → H
where r 7→ a and s 7→ b. One may check that S3 is generated by a and b so the homomorphism
is surjective. Since |D6 | = |S3 |, ψ is bijective and hence an isomorphism.

Note. Later, we will prove that up to isomorphism, there are exactly two groups of order 6, Z/6Z
is the unique abelian one and S3 is the unique nonabelian one. This is a classification theorem.

Group Actions.
One of the most important ideas in mathematics: take an algebraic structure and have it “act”
on another set in a way that preserves its algebraic structure This is the general idea behind
representation theory. We will study group actions in-depth in chapter 4 and they will be used in
Galois theory. Here, we lay out the basics of group theory before returning to it later.

Definition 30. A (left) group action of a group G on a set A is a map G × A → A, (g, a) 7→ g · a

satisfying

(1) g1 · (g2 · a) = (g1 g2 ) · a, for all g1 , g2 ∈ G, a ∈ A and

(2) 1 · a = a, for all a ∈ A.

We say that G acts on A and sometimes denote it G ↷ A. We also often write g · a as simply
ga when there is no danger of confusion. Remember, ga ∈ A. On the left-hand side of (1) above,
g2 a ∈ A so we can act on it by g1 and on the right-hand side, g1 g2 is a product in the group G.

A map G × A → A just tells you how every group element g acts on every set element a. Also
note that this map “preserves the group structure” as strongly as it can. Group multiplication is
the same thing as acting twice and identity acts like the identity permutation. We will make this
precise shortly.

Example 31. (1) G any group A any set, the trivial actions g · a = a for all g ∈ G and
a ∈ A.
16
(2) The nonzero elements of the field F × (thought of as a group under multiplication) acts
on the vector space F n by scalar multiplication.
(3) The general linear group GLn (R) acts on the vector space Rn . You learned about this
in linear algebra. Matrix multiplication gives the composition of linear transformations,
and the identity matrix acts trivially on Rn .
(4) The symmetric group Sn acts on the set {1, . . . , n}. Again, multiplication is transfor-
mation of bijections so the first axiom is satisfied.

17
 3. Wednesday 9/7: Group Actions, Subgroups, Normalizers, Centralizers

Last time we defined the action of a group G on a set A: A (left) group action of a group G on a
set A is a map G × A → A, (g, a) 7→ g · a satisfying

(1) g1 · (g2 · a) = (g1 g2 ) · a, for all g1 , g2 ∈ G, a ∈ A and

(2) 1 · a = a, for all a ∈ A.

An extremely important observation: let G act on a set A. Fix g ∈ G. Then there is a map
σg : A → A, σg (a) = g · a. It turns out that the following facts are true:

Fact. For each g ∈ G, σg is a permutation of A.

Proof of Fact 1. To say that σg is a permutation is the same thing as saying it is a bijective map
A → A. So we need only show that σg is bijective, but this follows since it has a two-sided inverse,
namely σg−1 :

(σg−1 ◦ σg )(a) = σg−1 (g · a) = g −1 · (g · a) = (g −1 g) · a = 1 · a = a

for all a ∈ A so σg−1 ◦ σg = Id. □

Fact. The map G → SA defined by g 7→ σg is a homomorphism (this is the precise meaning of a

group action preserving the group structure of G).

Proof of Fact 2. First we check we have a map φ : G → SA , φ(g) = σg but the calculation above
shows σg ∈ SA so this is true. We now need to check that this is a group homomorphism. This
means that
φ(g1 g2 ) = φ(g1 )φ(g2 )
or σg1 g2 = σg1 ◦ σg2 (since multiplication in SA is given by function composition). But this is true
by the first axiom of a group action. □

Definition 32. The homomorphism G → SA is called the permutation representation of the G

action on A.

Every group action gives a permutation representation. Also every homomorphism φ : G →

SA gives a group action by g · a = φ(g)(a). Hence, group actions are in bijection with group
homomorphisms G → SA .

Example 33. The associated permutation representation to Sn acting on {1, . . . , n} is just Sn

itself.

18
Definition 34. Let a ∈ A. The set

Orb(a) = {g · a | g ∈ G}

is called the orbit of a.

Example 35. • When R× ↷ Rn , the orbit of any nonzero vector v ∈ Rn is the line
through v and the origin, with the origin removed. The orbit of 0 is simply {0}.
• When Sn ↷ {1, . . . , n} in the usual way, there is only one orbit since for any a, b ∈
{1, . . . , n}, there is a permutation σ ∈ Sn such that σ(a) = b.
• Let S3 ↷ {1, . . . , 5} by acting on the first three integers. Then there are two orbits, the
set {1, 2, 3} the set {4} and the set {5}.

Fact. For any action G ↷ A, the set of orbits partition A.

This was true in all of the examples above, and you will prove that it is true on your next Problem
Set.

Definition 36. Let a ∈ A. The set

Stab(a) = {g ∈ G | g · a = a}

is called the stabilizer of a.

Example 37. • When R× ↷ Rn , the stabilizer of any nonzero vector v ∈ Rn is simply

the set {1}. The stabilizer of 0 is the entire group R× .
• When Sn ↷ {1, . . . , n} in the usual way, the stabilizer of 1 is all of the permutations
that fix 1, but permute the remaining n − 1 elements. Hence, the stabilizer has order
(n − 1)!, since it is a copy of Sn−1 acting on all of the integers other than 1.
• When S3 ↷ {1, . . . , 5} by acting on the first three integers, the stabilizer of 4 is all of
S3 . The stabilizer of 1 is given by all of the permutations which fix 1 and permute 2
and 3. That is, Stab(1) = {1, (2 3)}.

19
Definition 38. The set {g ∈ G | g · a = a for all a ∈ A} is called the kernel of the action of G on
A. This corresponds to the kernel of the map G → SA .

How is the kernel related to the stabilizers? With a bit of thought, we can see that the kernel is
the intersection of all of the stabilizers.

Two of the most important group actions occur when a group G acts on itself.

Example 39. • Let G be any group and let A = G. Define a map G × A → A by

g · a = ga (this makes sense because a ∈ G. This gives an action of G on itself where
each g ∈ G permutes the elements of G by left multiplication. This is called the left
regular action of G on itself.
• Another example: G acts on itself by conjugation. G = A and g · a = gag −1 .

We now move into the study of subgroups.

Definition 40. G a group. ∅ ̸= H ⊆ G is a subgroup of G if H is closed under products and

inverses (so is a group under the operation of G restricted to H). We notate this H ≤ G.

We just restrict the operation in G to H. In particular, any equation in H also happens in G. This
means eH = eG (and every subgroup of G must contain eG ) and inverses in H are the same thing
as inverses in G.

Example 41. (1) Trivially, every group G has subgroups H = G and H = {e} called the
trivial subgroup.
(2) The set of even integers in Z is a subgroup.
(3) If G = D2n then H = {e, r, r2 , . . . , rn−1 } is a subgroup isomorphic to Z/nZ.
(4) Z ≤ Q and Q ≤ R. Since being a subgroup is transitive, Z ≤ R.
(5) D6 is not a subgroup of D8 since the former is not even a subset of the latter.
(6) H = Q \ {0} under × is not a subgroup of G = Q under + even though it is a subset.
The operation in H needs to be the restriction of the operation in G.
(7) Z+ is not a subgroup of Z even though it is a subset with the same operation that is
closed under the operation. Z+ does not contain inverses so is not itself a group.

20
There is an easier criterion to check to check if H is a subgroup of G.

Proposition 42 (The Subgroup Criterion). H ⊆ G is a subgroup if and only if (1) H ̸= ∅ and (2)
for all x, y ∈ H, xy −1 ∈ H. If |H| < ∞, it suffices to check that H is nonempty and closed under
multiplication.

Proof. One direction is clear. So suppose the two conditions hold. By (1) H is nonempty so there
is some x ∈ H and by (2) xx−1 = 1 ∈ H. Now for all y ∈ H, 1y −1 ∈ H. Now if x, y ∈ H then
we just showed y −1 ∈ H so x(y −1 )−1 = xy ∈ H so H is closed under multiplication and H is a
subgroup.

For the second claim, suppose H is nonempty and closed under multiplication. Since H nonempty
then there is some x ∈ H. Now consider x, x2 , x3 , · · · ∈ H. Since H is finite, there exist some a < b
such that xa = xb in which case xb−a xa = xb so xb−a = 1. Every element of H thus has finite order
and x−1 = xn−1 ∈ H. □

Now that we’ve described subgroups, we should also define the cosets of a subgroup H, which you
can think of as “translations” of H inside the group G.

Definition 43. For any H ≤ G and any g ∈ G let

gH = {gh | h ∈ H} and Hg = {hg | h ∈ H}

called respectively a left coset and a right coset of H in G. Any element of a coset is called a
representative of the coset.

Example 44. The multiples of 3, 3Z ≤ Z. There are three cosets, namely

3Z = 0 + 3Z = 3 + 3Z = {. . . , −6, −3, 0, 3, 6, . . . }

1 + 3Z = 7 + 3Z = {. . . , −5, −2, 1, 4, 7, . . . }
2 + 3Z = −1 + 3Z = {. . . , −4, −1, 2, 5, 8, . . . }.
Note that in an additive group, instead of gH, we use the notation g + H for the coset of H
with representative g.

21
So really Z/3Z can be thought of as a group where the elements are the cosets of the subgroup 3Z.
This will be true in some generality, although it won’t work for every subgroup in a nonabelian
group. But that will come in the near future.

We now move on to introducing three very important kinds of subgroups. We will also see that
they are related to group actions.

Centralizers, Normalizers, and the Center of a Group.

Definition 45. Let ∅ =

̸ A ⊆ G. The centralizer of A in G is

CG (A) = {g ∈ G | gag −1 = a for all a ∈ A}.

Since gag −1 = a is equivalent to ga = ag, the centralizer of A is the set of group elements which
commute with every element of A.

Fact. CG (A) ≤ G.

Proof. Because 1 commutes with every element of A, CG (A) ̸= ∅. Now suppose x, y ∈ CG (A), then
xax−1 = a and yay −1 = a for all a ∈ A. Multiply on the left by y −1 and right by y to see that
y −1 ay = a for all a ∈ A so y −1 ∈ CG (A). Now

(xy)a(xy)−1 = (xy)a(y −1 x−1 ) = x(yay −1 )x−1 = xax−1 = a

so CG (A) is closed under products. □

If A = {a} we write just CG (a) for the centralizer. Since a commutes with itself (and any power),
an ∈ CG (a) for all n.

Definition 46. The center of G

Z(G) = {g ∈ G | gx = xg for all x ∈ G}.

i.e., the set of elements that commute with every element of G.

How is Z(G) related to the centralizers of different subsets? Well for one, Z(G) = CG (G) (this
shows that Z(G) is a subgroup). Further, for any A, certainly Z(G) ⊆ CG (A), since commuting
with every element in G is a stronger condition than commuting with just the elements in A.

22
 ̸ A ⊆ G. Define gAg −1 = {gag −1 | a ∈ A}. The normalizer of A in G
Definition 47. Let ∅ =

NG (A) = {g ∈ G | gAg −1 = A}.

The normalizer is the set of elements such that conjugation of a ∈ A knocks you back into A. The
centralizer is the set such that conjugation of a ∈ A gives you a exactly. In other words, to be in
the centralizer of A is a much stronger condition than to be in the normalizer of A. In general, we
will always have CG (A) ≤ NG (A). (You can prove that NG (A) is a subgroup).

Example 48. (1) If G is abelian and A is any subset of G, then, CG (A) = NG (A) = G.
Hence, these concepts are only interesting in nonabelian groups.
(2) Let G = D8 and A = {e, r, r2 , r3 } the subgroup of rotations. Then CD8 (A) = A. Since
all powers of r commute with each other, A ⊆ CD8 (A). Since sr = r−1 s ̸= rs, s does
not commute with elements of A so s ̸∈ CD8 (A). Now the other elements of D8 are sri ,
but if they were in the centralizer, then sri r−i = s would be which it isn’t.
(3) Let G and A be as above. We showed that CD8 (A) ⊆ ND8 (A) (this is true in general).
Now it is easy to see that

sAs−1 = {s1s−1 , srs−1 , sr2 s−1 , sr3 s−1 } = {e, r3 , r2 , r} = A

so s ∈ ND8 (A). And since the normalizer is a subgroup, this means D8 ⊆ ND8 (A) = D8 .
(4) For Z(D8 ), we know Z(D8 ) ⊆ CD8 (A) above. Now it is clear that r and r3 do not
commute with s, but r2 does so Z(D8 ) = {e, r2 }.

23
 4. Monday 9/12: Cyclic Groups, Subgroup Generation

Note. Next Monday (9/19) I will be traveling, so will not be on campus during our class time (I
have checked my conference/seminar/travel schedule for the rest of the semester, and I believe this
is the only time that I will miss). I will post a video on Blackboard for you to watch that day, as
well as posting the corresponding typed notes. I will remind you on Wednesday in class, and again
on Sunday evening via announcement/e-mail. Thanks for your understanding!

Recall. Last time we defined the centralizer of a subset A ⊆ G as

CG (A) = {g ∈ G | gag −1 = a for all a ∈ A}

and the normalizer of A in G as

NG (A) = {g ∈ G | gAg −1 = A}.

Example 49. (1) If G is abelian and A is any subset of G, then, CG (A) = NG (A) = G.
Hence, these concepts are only interesting in nonabelian groups.
(2) Let G = D8 and A = {e, r, r2 , r3 } the subgroup of rotations. Then CD8 (A) = A. Since
all powers of r commute with each other, A ⊆ CD8 (A). Since sr = r−1 s ̸= rs, s does
not commute with elements of A so s ̸∈ CD8 (A). Now the other elements of D8 are sri ,
but if they were in the centralizer, then sri r−i = s would be which it isn’t.
(3) Let G and A be as above. We showed that CD8 (A) ⊆ ND8 (A) (this is true in general).
Now it is easy to see that

sAs−1 = {s1s−1 , srs−1 , sr2 s−1 , sr3 s−1 } = {e, r3 , r2 , r} = A

so s ∈ ND8 (A). And since the normalizer is a subgroup, this means D8 ⊆ ND8 (A) = D8 .
(4) For Z(D8 ), we know Z(D8 ) ⊆ CD8 (A) above. Now it is clear that r and r3 do not
commute with s, but r2 does so Z(D8 ) = {e, r2 }.

How Normalizers and Centralizers are Related to Group Actions.

The fact that normalizers, centralizers, and the center are subgroups of G can be viewed as a special
case of results on group actions. Indeed, the properties of being in these subsets are all somehow
related to taking an element a or a subset A and considering gag −1 or gAg −1 . This is very much
related to the conjugation action of G on itself that we briefly discussed last time.

That action was an action G ↷ G where g · a = gag −1 . This is somewhat similar to what is
happening in the definition of the centralizer (although not quite what we want, which we’ll see in
24
a second). However, it is not quite what’s happening in the definition of the normalizer, since in
that case, we’re looking at the different sets gAg −1 for g ∈ G an A ⊆ G.

In that case, we are consider the action of G on the power set of G, denoted P(G). Recall from
your undergraduate classes that the power set of G is the set of all subsets of G. Now if A is any
subset of G, then gAg −1 is another subset of G. You can prove that this gives an action G ↷ P(G)
by conjugation.

What is NG (A) with respect to this action? Well

NG (A) = {g ∈ G | gAg −1 = A} = {g ∈ G | g · A = A}.

This was exactly the definition of the stabilizer :

Stab(x) = {g ∈ G | g · x = x}.

We sometimes use the notation Gx := Stab(x), if we want to emphasize what group is acting.

In other words, NG (A) is exactly the stabilizer of the set A under the conjugation action G ↷ P(G).

In general, for any action G ↷ X and any x ∈ X:

Fact. Gx ≤ G.

Proof. 1 ∈ Gx by axioms of a group action. Now if g ∈ Gx ,

x = 1 · x = (g −1 g) · x = g −1 (g · x) = g −1 · x,

so g −1 ∈ Gx . Finally, if g, h ∈ Gx then

(gh) · x = g · (h · x) = g · x = x. □

This gives an alternative way to see that NG (A) is a subgroup of G.

What about the centralizer? Well the centralizer is also defined in terms of some conjugation, but
the conjugation is not conjugation of subsets, rather it is the conjugation of elements of A. Can
we simply act G ↷ A by conjugation?

We cannot, since if a ∈ A, then there is no reason to think that gag −1 ∈ A, in general. If G = D8

and A = {e, r}, then conjugating by s takes us out of A, so we do not have a map G × A → A
given by conjugation. What can we act by? Well we want to make sure that for any a ∈ A, gag −1
ends up back in A. The set of elements g for which this is true is exactly the normalizer of A.

Hence, for any subset A ⊆ G, we have an action by the group NG (A) given by conjugation. Now
if g ∈ NG (A) and a ∈ A, since gAg −1 = A, we know gag −1 ∈ A, so this gives an action.

In terms of this action,

CG (A) = {g ∈ G | gag −1 = a for all a ∈ A} = {g ∈ NG (A) | gag −1 = a for all a ∈ A}

25
The latter equality is because we know that CG (A) ⊆ NG (A), so any element in G with the property
that it is in the centralizer is certainly in NG (A). Then continuing the above equality, we have

= {g ∈ NG (A) | g.a = a for all a ∈ A}.

Recall that the kernel of a group action is {g ∈ G | g · x = x for all x ∈ X}. In other words, CG (A)
is the kernel of the group action NG (A) ↷ A by conjugation.

It is true in general that the kernel of any group action is a subgroup of G (prove it!).

Example 50. • Let G = D8 act on the set A of four vertices of a square. The stabilizer
of a vertex a is {1, t} where t is the reflection through vertex a.
• Let G = D8 act on A = {{1, 3}, {2, 4}}, where 1 and 3 denote opposite vertices of the
square, and 2 and 4 denote the other pair of opposite vertices. The kernel of this action
on A is the set of elements that fix those sets, set-wise. This subgroup is {1, s, r2 , sr2 }.

Since we saw that Z(G) = CG (G), we can also recast the center as the kernel of a certain action.
In this case, A = G, so we are talking about the action G ↷ G by conjugation. Then Z(G) is the
kernel of this action.

Philosophical Aside. From a very high-level mathematical perspective, this is good evidence
that it is very fruitful to study group actions. To give some broad, hand-wavey ideas, the center
of a group seems like something that’s very internal to a group. It seems like to understand the
center of a group, you have to look inside the group and do a bunch of multiplications and see what
elements happen to commute with everything.

On the other hand, studying group actions seems somewhat external to the group. If you study
group actions, you are looking at how the group acts on other things (including sometimes itself).
However, studying actions of groups on other objects contains a wealth of information about the
group, even things that seem very internal to the group. You can recover the center as the kernel
of the conjugation action G ↷ G.

It is true in general that one of the lessons of modern mathematics is that you can learn a lot about
an algebraic structure by studying how it acts on other structures. We also saw that any group
action G ↷ A gives rise to a homomorphism G → SA . So this is also evidence that actually a great
deal of evidence about G is in fact contained in the homomorphisms G → K for other groups K.

But that’s all (extremely) big picture. For now, as graduate students in an algebra qual class, you
need to learn the definitions well and solve the kinds of problems that appear in textbooks. But if
you continue studying algebraic fields of mathematics, over a period of years you may be able to
appreciate this broad picture more and more.
26
Cyclic subgroups.
Let G be a group and x any element of G. We can look at the subgroup of all integer powers of x
(this guarantees closure under inverses and products).

Definition 51. A group H is cyclic if H can be generated by a single element. That is, there is
some x ∈ H such that H = {xn | n ∈ Z}.

We write H = ⟨x⟩ and say that H is generated by x.

Note: ⟨x⟩ = ⟨x−1 ⟩, since the powers run over Z. It is not necessarily the case that all of the xi are
distinct. And cyclic groups are automatically abelian.

Example 52. • Consider the group G = Z under +. Then G = ⟨1⟩

• Let G = D2n . Then H = ⟨r⟩ consists of the n rotations. The powers of r “cycle” with
period n.

Proposition 53. If H = ⟨x⟩ then |H| = |x| (if one side is infinite, then so is the other).

Proof. Suppose first that |x| = n is finite. The elements 1, x, . . . , xn−1 are distinct. If not, then
xa = xb for some 0 ≤ a < b ≤ n − 1 and so xb−a = 1, and so |x| < n which is a contradiction. So H
has at least n elements. We need to show that this is all of them. Let xt be an arbitrary element
of H. Use the division algorithm to write t = nq + k where 0 ≤ k ≤ n − 1. Then

xt = xnq+k = xk .

If |x| = ∞ then 1, x, x2 , . . . are all distinct, otherwise we would have xb−a = 1 and so |x| < ∞. □

Lemma 54. If xm = xn = 1, let d = gcd(m, n). Then xd = 1. In particular, if xm = 1 then |x| | m.

Proof. By the Euclidean Algorithm, there exist r and s such that d = mr + ns. Thus,

xd = xmr+ns = 1.
27
If xm = 1, let n = |x|. If m = 0, then we are done. Let d = gcd(m, n). By the preceding,
xd = 1. Since 0 < d ≤ n and n is the smallest positive power of x which is the identity, d = n so
gcd(m, n) = n and n | m. □

Theorem 55. Any two cyclic groups of the same order are isomorphic. If ⟨x⟩ and ⟨y⟩ are cyclic
groups of order n then φ : ⟨x⟩ → ⟨y⟩ given by xk 7→ y k is an isomorphism.

If ⟨x⟩ is an infinite cyclic group, then the map φ : Z → ⟨x⟩ given by k 7→ xk is an isomorphism.

Proof. Let ⟨x⟩ and ⟨y⟩ be cyclic groups of order n and φ : ⟨x⟩ → ⟨y⟩ be given by xk 7→ y k . We need
to prove that φ is well-defined, that is if xr = xs then φ(xr ) = φ(xs ). Since xr−s = 1, this implies
n | r − s. Write r = tn + s so

φ(xr ) = φ(xtn+x ) = y tn+s = (y n )t y s = y s = φ(xs )

so φ is well-defined. By laws of exponents, φ(xa xb ) = φ(xa )φ(xb ) so φ is a homomorphism. Further,

φ is surjective since every element of ⟨y⟩ can be written as y k and φ(xk ) = y k . Since both groups
have the same finite order, any surjection is a bijection so φ is an isomorphism. (Alternatively, φ
has an obvious two-sided inverse).

If ⟨x⟩ is an infinite cyclic group, let φ : Z → ⟨x⟩ be define by φ(k) = xk . This map is well-defnied
since there is no ambiguity in the representation of elements in the domain. Since xa ̸= xb for
distinct a, b ∈ Z (otherwise the group would have finite order) φ is injective. By the definition of a
cyclic group, φ is surjective. The laws of exponents guarantee that φ is a homomorphism so it is
an isomorphism. □

Notation. Your book prefers to write cyclic groups multiplicatively, so uses the notation Zn to
mean the cyclic group ⟨r⟩ where rn = 1.

Cyclic groups can have more than one generator. For example, Z = ⟨1⟩ = ⟨−1⟩ and Z/5Z = ⟨1⟩ =
⟨2⟩ = ⟨3⟩ = ⟨4⟩.

Proposition 56. Let G be a group, let x ∈ G and a ∈ Z \ {0}.

(1) If |x| = ∞, then |xa | = ∞.

n
(2) If |x| = n < ∞ then |xa | = gcd(n,a) .
(3) If |x| = n and a ∈ Z+ , a | n then |xa | = n/a.

28
Proof. (1) Assume |x| = ∞ but |xa | = m < ∞. Then 1 = (xa )m = xam . Further, x−am = 1−1 = 1.
so some positive power of x is the identity, which contradicts the hypothesis that |x| = ∞.

(2) Let y = xa , gcd(n, a) = d, and write n = db and a = dc for suitable b, c ∈ Z with b > 0. Since d
is the greatest common divisor of n and a, gcd(b, c) = 1. We must show that |y| = n/d = b. First
note that
y b = xab = xdcb = (xdb )c = 1c = 1
so |y| divides b (by the above lemma).

Let k = |y|. Then xak = y k = 1 so n | ak so db | dck so b | ck. Since gcd(b, c) = 1, b | k. Since b

and k divide each other, b = k. □

Proposition 57. Let H = ⟨x⟩.

(1) Assume |x| = ∞. Then H = ⟨xa ⟩ if and only if a = ±1.

(2) Assume |x| = n < ∞. Then H = ⟨xa ⟩ if and only if gcd(a, n) = 1. The number of generators
of H is φ(n).

Remark. Here φ means the Euler totient (or Euler phi) function. Where φ(n) is defined to be
the number of positive integers less than n which are coprime to n.

If p is prime, then φ(pn ) = pn − pn−1 . If gcd(m, n) = 1 then φ(mn) = φ(m)φ(n). For any integer,
you can compute this number by writing it at as a product or prime powers, using the second
identity to break it up, and then evaluating using the first identity.

Proof. (1) By the homework exercise, all the xi for i ∈ Z are distinct. Hence, if a ̸= ±1, then
⟨xa ⟩ = {xi | i = ad for some d ∈ Z}. And thus x ̸∈ ⟨xa ⟩ if a ̸= ±1.

(2) If |x| = n < ∞ then xa generates a subgroup of H of order |xa |. The subgroup equals all of H
if and only if |xa | = |x|. By the earlier proposition, |xa | = |x| if and only if n/ gcd(a, n) = n if and
only if gcd(a, n) = a. Since φ(n) is the number of integers a ∈ {1, . . . , n} such that gcd(a, n) = 1,
this is the number of generators of H. □

Example 58. In Z/15Z, the generators are 1, 2, 4, 7, 8, 11, 13, 14.

Given this last theorem, we are able to completely describe the subgroups of a cyclic group.
29
Theorem 59. Let H = ⟨x⟩ be a cyclic group. Then

(1) Every subgroup of H is cyclic. If K ≤ H then either K = {e} or K = ⟨xd ⟩ where d is the
smallest positive integer such that xd ∈ K.
(2) If |H| = ∞, then for any distinct nonnegative integers a and b, ⟨xa ⟩ =
̸ ⟨xb ⟩. Furthermore,
for every integer m, ⟨xm ⟩ = ⟨x|m| ⟩. So the nontrivial subgroups of H correspond bijectively
to the integers 1, 2, 3,...
(3) If |H| = n < ∞ for each positive integer a dividing n there is a unique subgroup of order
a. The subgroup is the cyclic group ⟨xd ⟩ where d = n/a. Furthermore, for every integer m,
⟨xm ⟩ = ⟨xgcd(n,m) ⟩, so the subgroups of H correspond bijectively to the positive divisors of
n.

Proof. (1) Let K ≤ H. We may assume K ̸= {e}. Thus there exists some a ̸= 0 such that xa ∈ K.
If a < 0 then since K is a group, x−a ∈ K. Hence, K contains a positive power of x. Let

P = {b | b ∈ Z+ and xb ∈ K}.

By the above, P is nonempty. By the well ordering principle, there is a minimum element, call it
d. Since K is a subgroup and xd ∈ K, ⟨xd ⟩ ≤ K. Since K is a subgroup of H, any element of K is
of the form xa for some a ∈ Z. By the division algorithm write

a = qd + r 0 ≤ r < d.

Then xr = x(a−qd) = xa (xd )−q is an element of K since both xa and xd are elements of K. By the
minimality of d it follows that r = 0, so a = qd. Hence, xa = (xd )q ∈ ⟨xd ⟩. Hence, K ≤ ⟨xd ⟩.

(3) Assume |H| = n < ∞ and a | n. Let d = n/a and apply the previous Proposition 56 (3) to
obtain ⟨xd ⟩ is a subgroup of order a. This shows the existence of a subgroup of order a.

Now for uniqueness, suppose K is a subgroup of H of order a. By part (1), we have K = ⟨xb ⟩
where b is the smallest integer such that xb ∈ K. By Proposition 56

n/d = a = |K| = |xb | = n/ gcd(n, b),

so d = gcd(n, b). In particular, d | b. Since b is a multiple of d, xb ∈ ⟨xd ⟩ hence K = ⟨xb ⟩ ≤ ⟨xd ⟩.

Since |⟨xd ⟩| = a = |K|, there fore K = ⟨xd ⟩.

Now for the last statement, it is clear that ⟨xm ⟩ ≤ ⟨xgcd(n,m) ⟩. It follows Proposition 56 that they
have the same order so are equal. Since gcd(n, m) | n, this shows that every subgroup of H arises
as a divisor of n. □

30
Example 60. The subgroups of Z/15Z are:

• Z/15Z = ⟨1⟩ = ⟨2⟩ = ⟨4⟩ = ⟨7⟩ = ⟨8⟩ = ⟨11⟩ = ⟨13⟩ = ⟨14⟩

• ⟨3⟩ = ⟨6⟩ = ⟨9⟩ = ⟨12⟩
• ⟨5⟩ = ⟨10⟩

We can picture these subgroups by drawing a lattice of subgroups, with larger subgroups at the
top, smaller subgroups toward the bottom, and lines to show containment relationships.

Z/15Z = ⟨1⟩

⟨3⟩ ⟨5⟩

⟨0⟩

Note that Z/6Z has the same lattice as Z/15Z. Here’s another example.

Z/12Z = ⟨1⟩

⟨2⟩ ⟨3⟩

⟨4⟩ ⟨6⟩

⟨0⟩

31
 5. Wednesday 9/14: More on Subgroup Generation, Quotient Groups

Here’s a common question in many areas of mathematics: given a structure G (a group, field,
vector space) and a subset A of G, is there a unique minimal subobject of G which contains A?
For example, if G = V is a vector space, and we take A = {v1 , . . . , vn }, we can ask if there is a
unique smallest subspace of V containing {v1 , . . . , vn }. I hope you remember from undergraduate
linear algebra that the answer is: yes take the span of the vectors.

Now if A is any subset of a group G, we make the notion of a subgroup of G generated by A

precise. We prove that the intersection of any set of subgroups of G is also a subgroup. The
subgroup generated by A is the unique smallest subgroup of G containing A (if there is a smaller
one, just intersect again). We show that elements of this subgroup are obtained by closing A under
the group operation and taking inverses.

Proposition 61. If A is any collection of subgroups of G, then the intersection all members of A
is also a subgroup.

T
Proof. Let K = H∈A H. Since each H ∈ A is a subgroup, 1 ∈ H for each H so 1 ∈ K. If a, b ∈ K,
then a, b ∈ H for each H ∈ A. Since each H is a subgroup, ab−1 ∈ H and since this is true for each
H, ab−1 ∈ K. By the subgroup criterion, K ≤ G. □

Definition 62. Let A be a subset of the group G. We define the subgroup of G generated by A to
be
\
⟨A⟩ = H
A⊆H,H≤G

So ⟨A⟩ is the intersection of all subgroups of G containing A. This is a subgroup by the above
proposition applied to the set A = {H ≤ G | A ⊆ H}. Since A lies in each H ∈ A, therefore
A ⊆ ⟨A⟩. Note that ⟨A⟩ is the unique minimal element of A: ⟨A⟩ is a subgroup of G containing A
so ⟨A⟩ ∈ A; and any element of A contains the intersection of all elements in A so contains ⟨A⟩.

If A = {a1 , . . . , an } then ⟨a1 , . . . , an ⟩ := ⟨A⟩. If A and B are two subsets of G, ⟨A, B⟩ := ⟨A ∪ B⟩.

Notice we just had a pretty abstract argument to show the existence of ⟨A⟩ and show that it is
the smallest subgroup containing A. It’s not very enlightening as to what ⟨A⟩ actually is. Indeed,
the definition could have worked in linear algebra: the unique smallest subspace of V containing
{v1 , . . . , vn } is the intersection of all subspaces of V which contain {v1 , . . . , vn }.
32
However, this is not the definition of span that we tell undergraduates in linear algebra, since you
are not easily able to compute with this definition. We define the span to be “the set of all possible
linear combinations of the vectors v1 , . . . , vn ” since this gives a concrete description of the vectors
in the subspace.

Similarly, for subgroups, we define the set which is the closure of A under the group operation and
inverses and prove that this set is ⟨A⟩. Let

A = ⟨aε11 aε22 . . . aεnn | n ∈ Z, n ≥ 0, ai ∈ A, εi = ±1⟩.

We think of e, the identity of G, as the empty product. Hence, if A = ∅, define A = {e}. That is,
A is the set of all finite products of elements of A and their inverses. Note that the ai ’s need not
be distinct: a2 = aa. Note also that A is not assumed to be finite or countable or anything else.

Proposition 63. A = ⟨A⟩.

Proof. First we prove that A is a subgroup. Note that A ̸= ∅ since it contains 1. If a, b ∈ A, write
a = aε11 . . . aεnn and b = bδ11 . . . bδmm . Then

ab−1 = aε11 . . . aεnn b−δ

1
1
. . . b−δ
m
m

and this is in A by definition. So A is a subgroup.

Since each a ∈ A is in A, A ⊂ A so ⟨A⟩ ⊆ A since ⟨A⟩ is the smallest subgroup containing A. But
⟨A⟩ is a group containing A and since it is closed under multiplication and inverses, A ⊆ ⟨A⟩. □

Since A = ⟨A⟩ we can take the definition of A to be the definition of ⟨A⟩. By combining exponents,
we may define
A = {aα1 1 . . . aαnn | ai ∈ A, αi ∈ Z, ai ̸= ai+1 , n ∈ Z+ }.

If G is abelian, we can commute the ai ’s and so group the like ai ’s together. If A = {a1 , . . . , an }
then
A = {aα1 1 . . . aαnn | ai ∈ A, αi ∈ Z, }.

If we further assume that each ai has finite order di , there are exactly di choices for the exponent
αi so
|⟨A⟩| ≤ d1 . . . dn .
This is only ≤ since it is possible for aα bβ = aγ bδ even though aα ̸= aγ and bβ ̸= bδ .

Just as with the span in linear algebra, when we actually work with ⟨A⟩, we usually think in terms
of A.
33
Quotient Groups.
In recent lectures, we’ve been discussing subgroups of a group. In this chapter we introduce the
notion of a quotient group which is another way of obtaining a “smaller group” from the group G.
This gives us another way to study the structure of G.

Thinking about the lattices we just discussed, if you take a subgroup H of G, then the subgroup
lattice of H is at the “bottom” of the lattice of G. For quotient groups, the lattice of subgroups
of a quotient group H will be at the “top” of the lattice of G. Before we define quotient groups,
we’ll give the canonical first example so you can think about the example when thinking about the
definitions.

Example 64. Consider the group Z. You naturally get a “smaller” group from Z in two ways:
The subgroups are nZ for various n. The other natural group associated to nZ is Z/nZ. This
is an example of a quotient group.

The study of quotient groups is essentially equivalent to the study of homomorphisms from G to
other groups.2

Definition 65. Let φ : G → H be a homomorphism. The fibers of φ are the sets of elements of G
projecting to single elements of H. That is, the fibers of φ are φ−1 (h) as h runs over H.

Example 66. Let’s write the group Z5 multiplicatively as ⟨a⟩ where |a| = 5. We have a
homomorphism φ : Z → Z5 sending k to ak . The group Z5 has five elements: e, a, a2 , a3 , a4 and
so there are five fibers of φ, e.g.,

φ−1 (e) = {. . . , −10, −5, 0, 5, 10, . . . } = 5Z

φ−1 (a) = {. . . , −9, −4, 1, 6, 11, . . . }

φ−2 (a2 ) = {. . . , −8, −3, 2, 7, 12, . . . }
etc.

The group operation in H provides a way to multiply two elements in the image of φ. This suggests
a natural multiplication on the fibers lying above these two points to make the set of fibers a group.

2This is somewhat related to the philosophy we all learned from Grothendieck: rather than study objects, study
maps. Try searching Wikipedia for “Grothendieck’s relative point of view”.
34
If Xa is the fiber above a and Xb is the fiber above b then the product of Xa and Xb is the fiber
Xab above ab.

Note that this operation is clearly well-defined, since the fiber above a is just Xa (there is no
other way to write this fiber). However, we need to check that this well-defined operation has
nice properties. This multiplication is associative since multiplication in H is associative and the
identity is the Xe . The inverse of Xa is Xa−1 . The group G is partitioned into pieces (the fibers)
and the pieces themselves have the structure of a group.

Example 67. Our homomorphism Z → Z5 gives us a way to define an operation on different

fibers. For example, if we consider Xa (the fiber over a) and Xa2 , we have an operation defined
on these elements.

The product of these two fibers Xa Xa2 is given by the product in H, so is Xa3 , the fiber over
a3 .

Recall that the kernel of a group homomorphism φ : G → H is

ker φ = {g ∈ G | φ(g) = e}.

The following proposition gives several useful facts about group homomorphisms, which we will
need for today’s lecture, but are useful more generally. In a graduate course, I think I can skip the
proofs, but you should all understand them quite easily.

Proposition 68. Let G and H be groups and let φ : G → H be a homomorphism.

(1) φ(eG ) = eH .
(2) φ(g −1 ) = φ(g)−1 for all g ∈ G.
(3) φ(g n ) = φ(g)n for all n ∈ Z.
(4) ker φ is a subgroup of G.
(5) im φ is a subgroup of H.
(6) φ is injective if and only if ker φ = {e}.

Proof. (1) Since φ(eG ) = φ(eG eG ) = φ(eG )φ(eG ) the cancellation law shows that eH = φ(eG ).

(2) Since φ(eG ) = φ(gg −1 ) = φ(g)φ(g −1 ) = eH , therefore φ(g)−1 = φ(g −1 ).

(3) This is by induction.

35
(4) By (1), eG ∈ ker φ. Now suppose a, b ∈ ker φ. Then φ(ab−1 ) = φ(a)φ(b−1 ) = φ(a)φ(b)−1 = eH
so ab−1 ∈ ker φ. By the subgroup criterion, ker φ is a subgroup.

(5) Since φ(eG ) = eH , eH is in the image. If x and y are in im φ then there exist a, b ∈ G such that
φ(a) = x and φ(b) = y. Then φ(ab−1 ) = φ(a)φ(b)−1 = xy −1 so xy −1 ∈ im φ.

(6) If ker φ ̸= {e} then there is e ̸= g ∈ ker φ and so φ(e) = φ(g) = e so φ is not injective.
Conversely, if φ is not injective then φ(g) = φ(g ′ ) for some g ̸= g ′ . Then φ(g −1 g ′ ) = e. □

Since we have seen that the fibers of a group homomorphism naturally have a group operation, we
give the following definition.

Definition 69. Let φ : G → H be a homomorphism with kernel K. The quotient group or factor
group, G/K (“G modulo K” or “G mod K”), is the group whose elements are the fibers of φ with
group operation defined above.

The notation emphasizes the fact that the kernel K is a single element in the group. We shall see
below that the other elements of G/K are “translates” of the kernel K (compare to your intuition
for Z/nZ). Hence, we can think of G/K as being obtained by collapsing K to a point or “dividing
out” by K or by “setting K to the identity”.

The definition of the quotient group G/K requires the map φ explicitly, since the multiplication is
given by mapping along φ to H, multiplying by H, and determining the fiber over the product. It
is also possible to define the product in terms of representatives from the fibers.

Proposition 70. Let φ : G → H be a homomorphism of groups with kernel K. Let X ∈ G/K be

the fiber above a. Then

(1) For any u ∈ X, X = {uk | k ∈ K} = uK

(2) For any u ∈ X, X = {ku | k ∈ K} = Ku.

Proof. We prove (1). The proof of (2) is just the same thing on the other side. Let u ∈ X so by
the definition of X, φ(u) = a. Let
uK = {uk | k ∈ K}.
We first prove uK ⊆ X. For any k ∈ K,

φ(uk) = φ(uk) = φ(u)e = a

36
so uk ∈ X. Now suppose g ∈ X and let k = u−1 g. Then

φ(k) = φ(u)−1 φ(g) = a−1 a = e

so k ∈ ker φ. Since k = u−1 g, g = uk ∈ uK. Hence, X ⊆ uK. □

Another way to say the theorem above is that you can choose any representative u of the fiber
φ−1 (a) and the whole fiber is given by uK. If you pick a different u′ then uK = u′ K.

Recall (from Definition 43) that uK is called a left coset and Ku is called a right coset of the
subgroup K. Proposition 70 shows that the fibers of a homomorphism are the left cosets of the
kernel (and also the right cosets of the kernel), i.e. the elements of the quotient G/K are the left
cosets gK, g ∈ G. In the example of Z/nZ, the multiplication in the quotient group could also be
defined in terms of representatives for cosets. This is true in general.

Theorem 71. Let G be a group and let K be the kernel of some homomorphism G → H. Then
the set whose elements are the left cosets of K in G with multiplication defined by

uK ∗ vK = (uv)K

forms a group. In particular, this operation is well-defined: if u′ ∈ uK and v ′ ∈ vK then u′ v ′ ∈ uvK

so u′ v ′ K = uvK so multiplication does not depend on the choice of representatives for the cosets.

Further, this group of left cosets of K is the same group as G/K.

Proof. We need to show this operation on left cosets is well defined. That is, if u, v ∈ G, then it
may be possible to choose different elements u′ , v ′ ∈ G such that

uK = u′ K and vK = v ′ K.

In order for our product to make sense, we need to make sure that if we had chosen to write the
cosets uK and vK as u′ K and v ′ K instead, then it would not affect the product. That is, we need
to show (uv)K = (u′ v ′ )K.

Consider h := φ(u) ∈ H. By the previous theorem, we know that uK = φ−1 (h). Similarly,
letting ℓ := φ(v) ∈ H, we know vK = φ−1 (ℓ). The product (uv)K is the fiber φ−1 (j) where
j = φ(uv) = φ(u)φ(v) = hℓ. So (uv)K is the fiber φ−1 (hℓ).

We know that any other element u′ ∈ G such that u′ K = uK is in the same fiber as u. In other
words, φ(u′ ) = h and similarly if v ′ K = vK then φ(v ′ ) = ℓ. Therefore, φ(u′ v ′ ) = hℓ so u′ v ′ is in
the fiber φ−1 (hℓ). But this is precisely (uv)K.

Summarizing, if u′ K = uK and v ′ K = vK, then (u′ v ′ )K = (uv)K, and so our claimed operation
does give a well-defined operation on the left cosets of K. To see that it is a group, it is associative
37
since the product in G is associative, the identity of the group is eK = K, and the the inverse of a
coset uK is given by u−1 K.

We now need to show that this group is the same group as G/K. The previous proposition says
that the group G/K and the group of cosets we just defined have exactly the same elements. The
fibers of φ are exactly the cosets of K. The products also coincide since by the above argument,
the product uKvK = (uv)K is the same as the product of the fibers over φ(u) and φ(v) yielding
the fiber of φ(u)φ(v) = φ(uv) since φ is a group homomorphism. □

Remark. This theorem tells us that there are two ways to view the quotient group G/K.

• We can think of group elements as fibers of the map φ : G → H. The multiplication of the
fiber φ−1 (a) and the fiber φ−1 (b) is the fiber φ−1 (ab). That is, the group operation comes
from the operation in H.
• We can think of group elements as the left cosets of the kernel K. The multiplication of
the coset uK and the coset vK is the coset (uv)K. In this case, the group operation comes
from the operation in G.

The fact that these views coincide is due to the proposition that fibers of φ are the same thing
as cosets of the kernel, and the fact that φ is a group homomorphism (so the operation in G and
operation in H are related by φ).

Example 72. Just to recap, φ : Z → Zn has fibers which are the left cosets a + nZ of the
kernel nZ. These cosets form a group under addition of representatives, namely Z/nZ which
explains the notation for the group. Z/nZ ∼
= Zn .

Example 73. If φ : G → H is an isomorphism then K = {e} and the fibers of φ are the
singleton subsets of G so G/{e} ∼
= G.

Example 74. Let G = R2 and let H = R and define φ : R2 → R to be projection (x, y) 7→ x.

It is easy to check that φ is a homomorphism. Further, ker φ = {(x, y) | x = 0} = the y-axis.

Note that ker φ is a subgroup of R2 and the fiber of φ over a ∈ R is the translate of the y-axis
by a. This is also the coset with representative (a, 0) (also it has representative (a, 1) or (a, −1)
or (a, y) for any real number y).

38
 We can view G/ ker φ as the group of vertical lines with the operation coming from adding their
x-coordinates. Since the operation on the quotient group is really coming from the operation
∼ R = H. In general, something like this will be true, but
in R, you might see that G/ ker φ =
let me point out that it is important that φ is surjective here.

By what we have developed so far, if we are given a subgroup K of a group G which we know is the
kernel of a homomorphism, we may define G/K without worrying about the homomorphism: simply
set uKvK = uvK. This raises the question: is it possible to define G/N for any subgroup N ? The
answer is no since multiplication is not well-defined in general. We will give exact conditions to
take quotient groups later. Even though we can’t take quotients, we can still talk about cosets:

Proposition 75. Let H be any subgroup of the group G. The set of left cosets of H form a
partition of G. Furthermore, for all u, v ∈ G, uH = vH if and only if v −1 u ∈ H. And in particular
uH = vH if and only if u and v are representatives of the same coset.

Proof. First note that since H is a subgroup of G, 1 ∈ H. So g ∈ gH and hence

[
G= gH.
g∈G

To show that distinct left cosets have empty intersection, suppose uH ∩ vH ̸= ∅. We will show that
uH = vH. Let x ∈ uH ∩ vH and write

x = uh = vk

for some h, k ∈ H. Multiply on the right by h−1 so

u = vkh−1 = vh′
Now for any element y ∈ uH, we can write y = uℓ for some ℓ ∈ H and so

y = uℓ = (vkh−1 )ℓ = v(kh−1 ℓ) ∈ vH

Since H is a subgroup so kh−1 ℓ ∈ H. Hence uH ⊆ vH. Reverse the roles of u and v to get the
opposite inclusion.

By the first part of the proposition, uH = vH if and only if u ∈ vH if and only if u = vh for
some h ∈ H if and only if v −1 u ∈ H, as claimed. Finally, v ∈ uH is equivalent to saying v is a
representative for uH hence uH = vH if and only if u and v are representatives for the same coset
(the coset uH = vH). □
39
Proposition 76. Let G be a group and let N be a subgroup of G

(1) The operation on the set of left cosets of N in G described by

uN · vN = (uv)N

is well-defined if and only if gng −1 ∈ N for all g ∈ G and all n ∈ N .

(2) If the above operation is well-defined then it makes the set of left cosets of N in G into a
group. In particular, the identity is 1N and the inverse of gN is g −1 N .

Proof. (1) Assume the operation is well defined. Then, for all u, v ∈ G if u′ ∈ uN and v ′ ∈ vN
then uvN = u′ v ′ N . Let g ∈ G and n ∈ N be arbitrary. Letting u = e, u′ = n, v = v ′ = g −1 and
applying the assumption above we deduce that

eg −1 N = ng −1 N i.e. g −1 N = ng −1 N.

Since e ∈ N , ng −1 e ∈ ng −1 N . So ng −1 ∈ g −1 N . Hence, ng −1 = g −1 n′ for some n′ ∈ N , Multiplying

on the left by g gives gng −1 = n′ ∈ N as claimed.

Conversely, assume gng −1 ∈ N for all g ∈ G and n ∈ N . To prove the operation is well-defined, let
u′ ∈ uN and v ′ ∈ vN . We may write

u′ = un and v ′ = vm for some n, m ∈ N .

We must prove that u′ v ′ ∈ uvN :

u′ v ′ = (un)(vm) = u(vv −1 )nvm = uv(v −1 nv)m = uvn′ m

where n′ = v −1 nv = v −1 n(v −1 )−1 is an element of N by assumption. Now N is a subgroup so

closed under products so n′ m ∈ N so u′ v ′ = (uv)n′′ for some n′′ ∈ N . Thus the left cosets uvN
and u′ v ′ N contain the common element u′ v ′ . By the preceding proposition they are equal (they
can have the same representative).

(2) If the operations on cosets is well defined, the group axioms are easy to check and induced by
their validity in G. □

These subgroups N for which you can form G/N have a name: they are called normal subgroups.
We will study them in more detail next time.

40
 6. Monday 9/19: Normal Subgroups and Lagrange’s Theorem

Note. This is our one-and-only (unless I get COVID) video lecture of the semester. Thanks for
your understanding.

Definition 77. For n ∈ N , g ∈ G, the element gng −1 is called a conjugate of n by g. The set
gN g −1 = {gng −1 | n ∈ N } is called the conjugate of N by g. The element g is said to normalize
N if gN g −1 = N . A subgroup N of a group G is called normal if every element of G normalizes N

gN g −1 = N for all g ∈ G.

If N is a normal subgroup of G, we write N ⊴ G.

We summarize the work we have done on normal subgroups in the following theorem:

Theorem 78. Let N be a subgroup of G. The following are equivalent

(1) N ⊴ G
(2) NG (N ) = G
(3) gN = N g for all g ∈ G
(4) the operation of left cosets of N in G makes the set of left cosets into a group
(5) gN g −1 ⊆ N for all g ∈ G.

When determining normality of a subgroup, one tries to minimize the computations necessary. You
don’t want to compute all of the conjugates gng −1 for n ∈ N and g ∈ G. For example, since N
is a subgroup, all elements of N automatically normalize N . If you have a set of generators of
N , it suffices to check all the conjugates of the generators (the conjugate of the products is the
product of the conjugates). Similarly, if you have a set of generators of G, it suffices to check that
the generators normalize. Sometimes you can prove that NG (N ) = G so that N is normal.

We now show that normal subgroups are the same things as kernels of homomorphisms.

Proposition 79. A subgroup N of the group G is normal if and only if it is the kernel of some
homomorphism.

41
Proof. If N is the kernel of some homomorphism φ then a previous proposition shows that the left
cosets of N are the same as the right cosets of N (both are fibers of φ) so by (3) of the previous
theorem, N is normal.

Conversely, if N ⊴ G, let H = G/N and define π : G → G/N by π(g) = gN . By definition of the

operation in G/N ,
π(g1 g2 ) = g1 g2 N = g1 N g2 N = π(g1 )π(g2 )
so π is a homomorphism. Finally,

ker π = {g ∈ G | π(g) = eN } = {g ∈ G | g ∈ N } = N. □

Definition 80. Let N ⊴ G. The homomorphism constructed above π : G → G/N defined by

π(g) = gN is called the natural projection (homomorphism) of G onto G/N . If H ≤ G/N is a
subgroup of G/N then the complete preimage of H in G is the preimage of H under the natural
projection.

The complete preimage of a subgroup of G/N is a subgroup of G containing N since those elements
all map to 1. We will see that there is a correspondence between subgroups of G/N and subgroups
of G containing N .

Based on Theorem 78, we have a criterion which determines when N is a kernel of a homomorphism,
namely NG (N ) = G. Thus, you can think of the normalizer of a subgroup as being a measure of
“how close” N is to being normal. Keep in mind that normality is a property of both N and how
N is embedded in G (depends on conjugation by elements in G). Thus, if N ≤ G ≤ G′ then it is
possible for N ⊴ G (and even G ⊴ G′ ) but N ̸⊴ G′ . Hence, normality is not transitive.

Example 81. (1) {e} and G are always normal in G and G/{e} ∼
= G and G/G ∼
= {e}, the
trivial group.
(2) If G is abelian, then any subgroup N of G is normal because for all g ∈ G and n ∈ N

gng −1 = gg −1 n = n ∈ N.

This requires G to be abelian, not just N .

(3) If G = Z, then every subgroup of G is cyclic N = ⟨n⟩ = ⟨−n⟩ = nZ and G/N = Z/nZ
is a cyclic group with generator 1 + nZ.
(4) If G = Zk is the cyclic group of order k, let x be the generator of G and N ≤ G. By
Theorem 59, N = ⟨xd ⟩ where d is the smallest power of x in N . Now

G/N = {gN | g ∈ G} = {xα N | α ∈ Z}

42
 and since xα N = (xN )α , G/N = ⟨xN ⟩ so G/N is cyclic. Given what we just saw about
Z, all subgroups and quotient groups of cyclic groups are cyclic.
(5) If N ≤ Z(G) then N ⊴ G since for all g ∈ G and n ∈ N , gng −1 = n ∈ N . In particular,
Z(G) ⊴ G.

Example 82. Last time, I vaguely talked about how subgroups can be seen “at the bottom”
of a subgroup lattice and quotient groups can be seen “at the top.” Let’s take a look again at
Z/12Z.

Z/12Z = ⟨1⟩

⟨2⟩ ⟨3⟩

⟨4⟩ ⟨6⟩

⟨0⟩

We know that G = Z/12Z is abelian, so every subgroup is normal. Hence, we can take the
quotient G/N for any of the subgroups listed. Of course when N = ⟨0⟩, then G/N ∼
= G, so
the whole diagram represents taking this quotient. On the other hand, when N = G, then
G/G = {0 + G} consists of a single element. So the subgroup diagram of G/G just looks like a
single point
G/G = {0 + G}.

Let’s take a more interesting example. The subgroup N = ⟨6⟩ = {0, 6} consists of two elements.
It therefore has 6 cosets, namely 0 + N, 1 + N, . . . , 5 + N . As you can see, these cosets partition
G. What is the group operation on this set of cosets? Well, for example

(4 + N ) + (5 + N ) = 9 + N = 3 + N.

In other words, the operation is just like addition modulo 6. Indeed, G/N ∼
= Z/6Z.
We know that Z/6Z has two proper nontrivial subgroups (usually written as ⟨2⟩ and ⟨3⟩ addi-
tively). What are those subgroups in G/N ? Well {0 + N, 2 + N, 4 + N } = ⟨2 + N ⟩ is a subgroup
of order 3 and ⟨3 + N ⟩ is a subgroup of order 2. In other words, the subgroup lattice of G/N
looks like
43
 G/N

⟨2 + N ⟩ ⟨3 + N ⟩

⟨0 + N ⟩ = 0 + N = N

The fact that ⟨6⟩ is the identity element in G/N is like setting the entire subgroup ⟨6⟩ to be 0.
The lattice of G/N looks like the part of the lattice of G sitting above N .

One of the most basic, yet important, invariants of a group is its order. We have essentially already
proved the following theorem when we proved that the cosets of any subgroup partition the group.

Theorem 83 (Lagrange’s Theorem). If G is a finite group and H ≤ G then |H| | |G| and the
number of left cosets of H in G equals |G|/|H|.

Proof. Let |H| = n and let the number of left cosets of H be k. By Proposition 75, the set of
left costs of H partition G. By definition of a left coset, the map H → gH define by h 7→ gh is a
surjection from H to the left coset gH. The left cancellation law implies this map is injective since
gh1 = gh2 implies h1 = h2 . Thus, |H| = |gH| = n. Since G is partitioned into k disjoint subsets
each of which has cardinality n, |G| = kn. Hence

k = |G|/n = |G|/|H|. □

Note that the above result does not require H to be normal. However, when N = H is normal,then
the number of left cosets is |G/N |, so the above says that (assuming G is finite), we have |G/N | =
|G|/|N |.

Definition 84. Let G be a group (not necessarily finite) and H ≤ G. The number of left cosets of
H in G is called the index of H in G and is denoted |G : H| or [G : H].

If G is finite, then |G : H| = |G|/|H| by the above. For an group G of infinite order, |G|/|H| does
not make sense. But infinite groups may have subgroups of finite index. For example, in Z, {0}
has infinite index and nZ has index n.

We have the following (fairly) immediate corollaries of Lagrange’s Theorem.

44
Corollary 85. If |G| < ∞ and x ∈ G then |x| | |G|. In particular, x|G| = 1 for all x ∈ G.

Proof. We saw that |x| = |⟨x⟩|. The first part follows from Lagrange’s Theorem. The second is
clear since |G| is a multiple of |x|. □

Corollary 86. If |G| = p a prime, then G is cyclic, hence G ∼

= Zp .

Proof. Let x ∈ G, x ̸= 1. Then |⟨x⟩| > 1 and |⟨x⟩| | |G| = p. hence, ⟨x⟩ = G so G is cyclic (with
any nonidentity element generating). We proved earlier that any two cyclic groups of the same
order are isomorphic. □

There is no general converse to Lagrange’s Theorem. That is, if k | |G| < ∞, it is possible that G
may hot have a subgroup of order k. There are partial converses, however.

If G is a finite abelian group, then G has a subgroup of order n for each divisor n of G (this will
be a consequence of the Fundamental Theorem of Finitely Generated Abelian Groups).

In general for a finite nonabelian group, there are also some partial converses to Lagrange’s Theo-
rem. We are not ready to prove either of these theorems, but they are worth stating here. We will
prove them soon enough.

Theorem 87 (Cauchy’s Theorem). If G is a finite group and p is a prime p | |G| then G has an
element of order p (and hence a subgroup of order p).

Theorem 88 (Part of Sylow’s Theorem). If G is a finite group of order pα m where p is a prime

and p ∤ m then G has a subgroup of order pα .

So if |G| = 100, then it is possible that G contains no subgroup of order 50. But since 100 = 22 · 52 ,
Cauchy’s Theorem says that G has a subgroup of order 2 and a subgroup of order 5. Sylow’s
Theorem says that G has a subgroup of order 4 and a subgroup of order 25.

The next example example exhibits a powerful use of the index of a subgroup. We will later prove
a more general result, so we won’t label this with a theorem here.
45
 Example 89. Let H ≤ G be a subgroup of index 2. We will prove that H ⊴ G. (The more
general result we will prove later is: if p is the smallest prime dividing |G| then any subgroup
of order p is normal.) Let g ∈ G \ H. Then the two left cosets of H in G are 1H = H and gH.
Since the cosets of H partition G, gH = G \ H.

Similarly, the two left cosets of H must be 1H = H and Hg = G \ H. Thus, gH = Hg so every

left coset is a right coset so by Theorem 78 (our big normality theorem), H ⊴ G.

In other words, if you have a really big subgroup (the smallest possible index), then that really
big subgroup must be normal.

We can also use this previous result to cook up an example that shows that being a normal subgroup
is not a transitive property.

Example 90. Consider

⟨s⟩ ⊴ ⟨s, r2 ⟩ ⊴ D8 .
Each subgroup is normal in the next, since each has index 2. However, ⟨s⟩ is not normal in D8
because rsr−1 = sr2 ̸∈ ⟨s⟩.

We now consider some non-normal subgroups. Recall that in an abelian group every subgroup is
normal, but in the nonabelian case this need not be the case. Indeed, there are groups G in which
the only normal subgroups are 1 and G. Such a group is called a simple group. Simple groups are
very important in the classification/study of groups and we will study them in some depth later.
However the important part is that normal subgroups may be rare in G. Finding normal subgroups
can be difficult.

Example 91. Let H = ⟨(1 2)⟩ ≤ S3 . Since H is of index 3 in S3 , and H ≤ NG (H), either
NG (H) = H or NG (H) = S3 . Direct computation shows that

(1 3)(1 2)(1 3)−1 = (1 3)(1 2)(1 3) = (2 3) ̸∈ H

so NG (H) ̸= S3 so H is not a normal subgroup of S3 . Another way to see this is

(1 3)H = {(1 3), (1 2 3)} H(1 3) = {(1 3), (1 3 2)}

Since the left coset (1 3)H is the unique left coset containing (1 3), H(1 3) cannot be a left
coset.

46
Example 92. Let G = Sn for some n ∈ Z+ and fix some i ∈ {1, 2, . . . , n}. Consider the
stabilizer
Gi = {σ ∈ G | σ(i) = i}.
Suppose τ ∈ G and τ (i) = j. It follows that for all σ ∈ Gi , τ σ(i) = j. Furthermore, if µ ∈ G
and µ(i) = j then τ −1 µ(i) = i so τ −1 µ ∈ Gi so µ ∈ τ Gi . Hence,

τ Gi = {µ ∈ G | µ(i) = j}

which is to say that the left coset τ Gi consists of the permutations in Sn which take i to j. It
is clear that distinct left cosets have empty intersection and that the number of distinct left
cosets equals the number of distinct images of the integer i under the action of G—i.e. there
are n of them. Hence, |G : Gi | = n.

Using the same notation, let k = τ −1 (i) so τ (k) = i. By similar reasoning to above, we see that

Gi τ = {µ ∈ G | µ(k) = i}

so the right coset Gi τ consists of permutations which take k to i. Hence, if τ is not the identity,

τ Gi ̸= Gi τ

since there are permutations which take i to j but do not take k to i. Hence, Gi ̸⊴ G.

47
 7. Wednesday 9/21: The Isomorphism Theorems, Composition Series

Today our goal is to learn the isomorphism theorems. These are central, fundamental results
which describe the relationship between quotients, homomorphisms, and subgroups. Since these
are all important concepts no matter what kind of structure you want to understand (groups,
rings, modules, etc.), we will also learn isomorphism theorems for rings (and in 6102, modules).
The group isomorphism theorems are the most important since: (1) they are the first ones that you
learn, so the ideas are somewhat new and (2) since rings and modules both have underlying abelian
group structures, the proofs of their isomorphism theorems actually rely on the group isomorphism
theorems.

Definition 93. Let H, K ≤ G and define

HK = {hk | h ∈ H, k ∈ K}.

Warning: HK need not be a subgroup, just a subset of G. You also might hope that HK = KH,
which is also not true in general. Even though HK is not always a subgroup, we can still always
talk about its cardinality.

Proposition 94. If H and K are finite subgroups of a group then

|H||K|
|HK| = .
|H ∩ K|

Note: Recall that H ∩ K is a subgroup (we proved this in Proposition 61).

Proof. Notice that we can write HK as a union of left cosets of K namely

[
HK = hK.
h∈H

Since each coset of K has |K| elements, we need only find the number of distinct left cosets of the
form hK where h ∈ H. But h1 K = h2 K for h1 , h2 ∈ H if and only if h−1
2 h1 ∈ K. Therefore

h1 K = h2 K ⇔ h−1
2 h1 ∈ H ∩ K ⇔ h1 (H ∩ K) = h2 (H ∩ K).

Thus, the number of distinct cosets of the form hK is the number of distinct cosets h(H ∩ K) (in
H). That number, by Lagrange’s Theorem is |H|/|H ∩ K|. Thus, HK consists of |H|/|H ∩ K|
cosets of size |K| whch yields the above formula. □

We can also fully characterize when HK is a subgroup of G.

48
Proposition 95. If H and K are subgroups of a group, HK is a subgroup if and only if HK = KH.

Proof. Assume that HK = KH and let a, b ∈ HK. We prove ab−1 ∈ HK so HK is a subgroup

by the subgroup criterion. Let a = h1 k1 and b = h2 k2 for some h1 , h2 ∈ H and k1 , k2 ∈ K. Thus
b−1 = k2−1 h−1
2 so
ab−1 = h1 k1 k2−1 h−1
2 .

Call k3 = k1 k2−1 ∈ K and h3 = h−1

2 and rewrite

ab−1 = h2 k3 h3

Since HK = KH, k3 h3 = h4 k4 for some h4 ∈ H, k4 ∈ K. Thus ab−1 = h1 h4 k4 ∈ HK.

Conversely, assume HK is a subgroup of G. Since K ≤ HK and H ≤ HK, by the closure property

of subgroups KH ⊆ HK. To show the reverse inclusion, let hk ∈ HK. Since HK is a subgroup,
write hk = a−1 for some a = h1 k1 ∈ HK. Then

hk = (h1 k1 )−1 = k1−1 h−1

1 ∈ KH. □

Warning: HK = KH does not mean every element of H commutes with every element of K.
Just that any product hk with h ∈ H, k ∈ K can be written as k ′ h′ for k ′ ∈ K, h′ ∈ H.

Corollary 96. If H, K ≤ G and H ≤ NG (K) then HK ≤ G. In particular, if K ⊴ G then

HK ≤ G for any subgroup H ≤ G.

Proof. By the above Proposition we need only show that HK = KH under the hypotheses. Let
h ∈ H and k ∈ K. By assumption, hkh−1 ∈ K and so

hk = (hkh−1 )h ∈ KH

this proves HK ⊆ KH. Similarly, kh = h(h−1 kh) ∈ HK proving the reverse containment. □

The Isomorphism Theorems

We are now ready to learn the isomorphism theorems. Note that the first of these was already
something we have already understood, even way back in Example 74.

Theorem 97 (The First Isomorphism Theorem). If φ : G → H is a homomorphism of groups,

then ker φ ⊴ G and G/ ker φ ∼
= φ(G).

49
As we said, we already proved this when we were studying quotient groups. We found that given any
homomorphism of groups φ : G → H that ker φ is a normal subgroup and G/ ker φ is isomorphic
to the image of φ.

In particular, if φ is surjective, then G/ ker φ ∼

= H.

Corollary 98. Let φ : G → H be a homomorphism of groups. Then |G : ker φ| = |φ(G)|.

Proof. |G : ker φ| is equal to the number of cosets of ker φ, but is also equal to |G|/| ker φ| = |φ(G)|,
by the first isomorphism theorem. □

The first isomorphism theorem is often how you prove the isomorphism of certain groups.

Example 99. The map φ : GLn (F ) → F × given by mapping M ∈ GLn (F ) to det(M ) is a

×
group homomorphism
  Example 24. It is also clearly surjective, since for any α ∈ F , the
by
α
 
 1 
matrix   has determinant α.
 
..

 . 

1
Now the first isomorphism theorem says that GLn (F )/ ker φ ∼= F × . But ker φ is exactly those
matrices with determinant 1, i.e., ker φ = SLn (F ). Hence, GLn (F )/ SLn (F ) ∼
= F ×.
Note that without the first isomorphism theorem, this would be a tough thing to show. You
could show that SLn (F ) is a normal subgroup by showing that conjugating by any invertible
matrix preserves determinant (since determinant is multiplicative). But then, what group do
the cosets form? Using the first isomorphism theorem by finding SLn (F ) as the kernel of an
explicit homomorphism helps to see what the quotient group is.

Theorem 100 (The Second Isomorphism Theorem/The Diamond Isomorphism Theorem). Let G
be a group and A and B be subgroups of G. Assume A ≤ NG (B) (so that AB is a subgroup of G).
Then B ⊴ AB, A ∩ B ⊴ A and AB/B ∼ = A/A ∩ B.

Proof. Since A ≤ NG (B) by assumption and B ≤ NG (B) trivially, it follows that AB ≤ NG (B).
Hence, B ⊴ AB.
50
Since B ⊴ AB, the quotient group AB/B is well defined. Define φ : A → AB/B defined by
φ(a) = aB. Since the group operation in AB/B is well defined,

φ(a1 a2 ) = (a1 a2 )B = a1 Ba2 B = φ(a1 )φ(a2 )

so φ is a homomorphism. It is also clear from the definition of AB that φ is surjective. The identity
in AB/B is the coset 1B so the kernel consists of elements a ∈ A such that aB = 1B which are
the elements a ∈ B so ker φ = A ∩ B. By the first isomorphism theorem, the result follows. □

The way I like to think about this isomorphism theorem is the group AB is like adding all of B to
A (since e ∈ A and e ∈ B, A, B ≤ AB). Then AB/B kills everything in B. You could also just
straight away kill everything in A that’s also in B. The hypothesis that A ≤ NG (B) sort of says
that B is close to normal (or at least that with respect to A, B is close to normal). So it makes
sense that B would be normal in the subgroup AB (which only sees things in A and in B).

Why is the second isomorphism theorem also called the diamond isomorphism theorem? Well we
can draw the following picture of subgroups in G related to A and B:

AB
′′

A B
′′

A∩B

Of course, in order for AB to be a subgroup, we need to assume something (in this case, A ≤
NG (B)). The theorem asserts that the two quotients marked with “ are isomorphic.

Example 101. 4Z ⊴ Z and 6Z ⊴ Z. The theorem says

4Z + 6Z ∼ 6Z
= .
4Z 4Z ∩ 6Z
We can rewrite the left group as 2Z/4Z and the right one as 6Z/12Z, and both of these are
isomorphic to Z/2Z.

The Third and Fourth Isomorphism Theorems tell us about the subgroup structure of quotient
groups.
51
Theorem 102 (The Third Isomorphism Theorem). Let G be a group and H, K ⊴ G with H ≤ K.
Then K/H ⊴ G/H and
(G/H)/(K/H) ∼
= G/K.

Proof. Define a map

φ : G/H → G/K
by gH 7→ gK. To show φ is well-defined, suppose g1 H = g2 H. Then g1 = g2 h for some h ∈ H.
Since H ≤ K, h ∈ K as well so g1 K = g2 K so φ(g1 H) = φ(g2 H). Since g may be chosen arbitrarily,
φ is a surjective homomorphism. Finally,

ker φ = {gH ∈ G/H | φ(gH) = eK = K} = {gH ∈ G/H | g ∈ K} = K/H.

Hence K/H is a normal subgroup of G/H (it is the kernel of a homomorphism) and (G/H)/(K/H) ∼
=
G/K. □

The Fourth Isomorphism Theorem establishes a bijection between subgroups of G containing N

and the subgroups of G/N . Hence, the lattice for G/N appears in the lattice for G as the collection
of subgroups of G between N and G.

Theorem 103 (The Fourth Isomorphism Theorem). Let N ⊴ G. There is a bijection between
the set of subgroups A of G containing N and the set of subgroups A/N of G/N . Hence, every
subgroup of G/N is of the form A = A/N for some subgroup A of G containing N . For all A, B ≤ G
with N ≤ A and N ≤ B

(1) A ≤ B if and only if A ≤ B

(2) If A ≤ B, then |B : A| = |B : A|
(3) ⟨A, B⟩ ∼
= ⟨A, B⟩
(4) (A ∩ B)/N ∼=A∩B
(5) A ⊴ G if and only if A ⊴ G/N .

Proof. Let φ : G → G/N be the quotient map. Let A ≤ G then prove φ(A) ≤ G/N . Now for any
subgroup H ≤ G/N , consider φ−1 (H) and prove φ−1 (H) is a subgroup of G. Use this to show that
there is a bijection between subgroups of G containing N and subgroups of G/N .

Once you have understood this bijection, actually all five parts of the theorem follow quite obviously.
For example, for any function, if A ≤ B then φ(A) ≤ φ(B) and vice versa.

You should prove the rest of the parts, as this is really just an exercise in understanding notation. □
52
 Example 104. nZ ⊴ Z. The theorem says subgroups of Z/nZ are in bijection with subgroups
of Z containing nZ which are dZ for each divisor d of n. Hence, the subgroups of Z/nZ all have
the form dZ/nZ and Z/nZ/dZ/nZ ∼ = Z/dZ.

We saw that when N is a normal subgroup, the lattice of subgroups of N is found in the lattice
of subgroups of G “above” N (and that this follows from the fourth isomorphism theorem). One
question is how much you can learn about G from N and G/N . This is the idea behind the proof
of Cauchy’s Theorem, which we deferred until now, even though we stated it as a theorem in
Theorem 87.

Theorem 105 (Cauchy’s Theorem). If G is a finite abelian group and p is a prime dividing |G|
then G contains an element of order p.

Proof. Suppose G is a finite abelian group and let p | |G|. Proceed by induction on |G|. Assume
the result is true for every group whose order is strictly smaller than |G| and then prove the result
is true for G (so this is strong induction). Since |G| > 1 there is an element x ∈ G with x ̸= e.
If |G| = p then x has order p by Lagrange’s Theorem so we are done. We may therefore assume
|G| > p.

Suppose p | |x| and write |x| = pn. Then |xn | = p so we have an element of order p. We may
therefore assume p ∤ |x|.

Let N = ⟨x⟩. Since G is abelian N ⊴ G. By Lagrange’s Theorem, |G/N | = |G|/|N | and since
N ̸= 1, |G/N | < |G|. Since p does not divide |N |, we must have p | |G/N |. Now apply the induction
hypothesis to conclude that G/N contains an element y = yN of order p. Since y ̸∈ N but y p ∈ N ,
we must have ⟨y p ⟩ =
̸ ⟨y⟩ so |y p | < |y|.

We saw that if y has order n then |y p | = n/ gcd(n, p). Hence, gcd(n, p) > 1 so p | n. But now we
are in the situation of two paragraphs ago, so we have found an element of order p, namely y m
where |y| = pm. □

53
 8. Monday 9/26: Composition Series and Solvable Groups

We started class by watching James Zhang’s talk “Open problems in noncommutative ring theory”.
Later on in this semester, we will learn ring theory, so the talk was not unrelated to this class.
As graduate students (or soon-to-be graduate students), I encourage you to go to many talks.
Most of the time, you will feel very lost, and as if you know nothing. This happens to me all the
time. But there is an art to attending math talks and getting something out of them, and it’s
something that you can (and should) actively work on. Ravi Vakil has a “three things” exercise
here: http://math.stanford.edu/~vakil/threethings.html.

Also, James had a nice quote: “it is always good to work on classification problems”. If you work
on classification problems (of algebras, say), you can discover new algebras, you can understand the
structure of algebras, and you can invent new invariants to help you classify algebras. In lecture
today, we saw an example of a success of modern mathematics in classifying finite simple groups.

Last time, we ended by proving Cauchy’s Theorem. The proof illustrated a general technique: if
we have enough information about N and G/N then we can piece the information together to learn
about G by induction because N and G/N have smaller order than G. In general it is hard to know
how much information is enough. Certainly not all of the information about G is contained in just
N and G/N , since it is possible for N and G/N to be isomorphic to N ′ and G′ /N ′ but G ∼ ̸ G′
=
(consider Z/4Z vs Z/2Z × Z/2Z).

This particular idea requires us to find a nontrivial normal subgroup of G. But there are groups
without any nontrivial normal subgroups.

Definition 106. A group G is called simple if |G| > 1 and the only normal subgroups of G are
{e} and G.

Example 107. We already know that if p is prime and |G| = p, by Lagrange’s Theorem any
subgroup has order 1 or p so G must be simple. In fact, every abelian simple group is isomorphic
to Zp for some prime p. There are nonabelian simple groups, the smallest of which has order
60 and we will learn about shortly.

Simple groups cannot be “factored” into pieces so play a role analogous to that of the primes in Z.
This analogy is supported by a “unique factorization theorem” which we describe below.

54
Definition 108. In a group G a sequence of subgroups

1 = N0 ≤ N1 ≤ N2 ≤ · · · ≤ Nk−1 ≤ Nk + G

is called a composition series if Ni ⊴ Ni+1 and Ni+1 /Ni is a simple group. In a composition series,
the quotient groups Ni+1 /Ni are called composition factors of G.

Example 109. We have two composition series for D8 :

1 ⊴ ⟨s⟩ ⊴ ⟨s, r2 ⟩ ⊴ D8

and
1 ⊴ ⟨r2 ⟩ ⊴ ⟨r⟩ ⊴ D8 .
The groups are normal in each other since they each have index 2 and the quotients are all
isomorphic to the simple group Z2 .

Theorem 110 (Jordan–Hölder3). Let G be a finite group with G ̸= {e}. Then G has a composition
series and the composition factors are unique up to order.

The Hölder Program

Because of the previous theorem, one way to try to understand finite groups is to understand (1) all
of the finite simple groups and (2) find all ways of “putting simple groups together”. Completing
(1) is one of the greatest successes of modern algebra.4

Theorem 111. Every finite simple group is isomorphic to one of the following groups:

• a member of one of three infinite classes of finite simple groups, namely:

– the cyclic groups of prime order,
– the alternating groups of degree at least 5,
– the groups of Lie type,
• one of 26 “sporadic” groups.

3Here’s a valuable lesson for you. If you are citing two different people’s last names at once, they should be separated
with an en dash (i.e., two hyphens in LATEX). If it is a single person’s hyphenated name, you use a hyphen. So
here, we have the Jordan–Hölder theorem, but in analysis you might learn Mittag-Leffler’s Theorem (one guy named
Mittag-Leffler).
4
See https://en.wikipedia.org/wiki/Classification_of_finite_simple_groups.
55
Proof. Thousands of pages of mathematics in hundreds of papers spanning over decades. □

A very powerful result (which you cannot cite on a qual) is the following:

Theorem 112 (Feit–Thompson). If G is a simple group of odd order then G ∼

= Zp for a prime p.

Proof. Feit and Thompson’s paper is 255 pages long. □

One of the classes of groups that is important in classifying groups are the solvable groups

Definition 113. A group G is solvable if there is a chain of subgroups

1 = G0 ⊴ G1 ⊴ . . . ⊴ G s = G

such that Gi+1 /Gi is abelian for i = 0, . . . , s − 1.

Lemma 114. If N and G/N are solvable then G is solvable.

Proof. Let G = G/N . Since N is solvable, let 1 = N0 ⊴ N1 ⊴ . . . ⊴ Nn = N be a chain of subgroups

of N such that Ni+1 /Ni is abelian. Let 1 = G0 ⊴ G1 ⊴ . . . ⊴ Gm = G be a chain of subgroups of
G such that Gi+1 /Gi is abelian. By the Fourth Isomorphism Theorem, there are groups Gi of G
with N ≤ Gi such that Gi /N = Gi and Gi ⊴ Gi+1 . Further, by the Third Isomorphism Theorem,

Gi+1 /Gi = (Gi+1 /N )(Gi /N ) ∼

= Gi+1 /Gi .

Thus
1 = N0 ⊴ N1 ⊴ · · · ⊴ Nn = N = G0 ⊴ G1 ⊴ · · · ⊴ Gm = G
is a chain of subgroups showing G is solvable. □

56
 9. Wednesday 9/28: Alternating Groups and Group Actions

We now turn our attention to the alternating groups appearing in Theorem 111. In a future lecture,
we will prove that A5 is simple.

When we first studied the symmetric group Sn , we saw that every element of Sn can be written
as a product of disjoint cycles in an essentially unique fashion. However, if you do not require the
cycles to be disjoint, elements of Sn can be written in many different ways. In S3 , for example, if
we let
σ = (1 2 3) = (1 3)(1 2) = (1 2)(1 3)(1 2)(1 3) = (1 2)(2 3)
then there are an infinite number of ways to write σ as a product of cycles. However, there is an
important invariant associated to permutations.

Definition 115. A 2-cycle is called a transposition.

It is not surprising that every permutation of {1, . . . , n} can be realized by a succession of transpo-
sitions (you can get any permutation of cards in a deck by switching two at a time).5 Hence, any
permutation in Sn can be written as a product of transpositions. Equivalently,

Sn = ⟨T ⟩ where T = {(i j) | 1 ≤ i < j ≤ n}.

There are of course many ways to write a permutation as a product of transpositions (we did it in
three ways in that previous example). However, there is an important invariant: the parity of the
number of cycles. It turns out for a fixed σ, any way you write σ as a product of transpositions
will have the same parity (will either be always even or always odd). We prove this now.

Let x1 , . . . , xn be independent variables and define the polynomial6

Y
∆= (xi − xj ).
1≤i<j≤n

For each σ ∈ Sn let σ act on ∆ by permuting the variables:

Y
σ(∆) = (xσ(i) − xσ(j) ).
1≤i<j≤n

5Or if your prefer, you can write (a a . . . a ) = (a a )(a a

1 2 m 1 m 1 m−1 ) . . . (a1 a2 ). Since every permutation can be
written as a product of cycles, and every cycle can be written as a product of transpositions, every permutation can
be written as a product of transpositions.
6Called the Vandermonde polynomial.
57
All of the same terms appear in ∆ and σ(∆), since for each i and j, either xi − xj or xj − xi appears
but not both. Hence, for all σ ∈ Sn , σ(∆) = ±∆. Define

+1 if σ(∆) = ∆
ϵ(σ) =
−1 if σ(∆) = −∆.

Definition 116. The sign of σ is ϵ(σ). If ϵ(σ) = 1 then σ is an even permutation while if ϵ(σ) = −1,
σ is called an odd permutation.

Proposition 117. The map ϵ : Sn → {±1} is a homomorphism (where {±1} is the multiplicative
version of the cyclic group of order 2).

Proposition 118. Transpositions are all odd permutations and ϵ is a surjective homomorphism.

Proof. A transposition changes exactly one xi − xj to xj − xi so switches the sign of ∆. □

Definition 119. The alternating group of degree n denoted An is the kernel of ϵ (i.e. the set of all
even permutations).

By the first isomorphism theorem, Sn /An ∼ = Z2 so the order of An = n!/2. Also, Sn \ An is the
coset of An which is not the identity coset. This is the set of all odd permutations of Sn . Note
that the signs of the permutations obey the usual Z2 laws: (even)(even) = (odd)(odd) = even and
(even)(odd) = (odd)(even) = odd.

Moreover, since ϵ is a homomorphism and every σ ∈ Sn can be written as a product of transpositions

σ = τ1 . . . τk then ϵ(σ) = ϵ(τ1 ) . . . ϵ(τk ) = (−1)k . Thus, the parity of the number of transpositions
is the same no matter how we write σ as a product of transpositions.

+1 if σ is a product of an even number of transpositions
ϵ(σ) =
−1 if σ is a product of an odd number of transpositions.

Finally, note that since an m-cycle can be written as a product of m − 1 transpositions, an m-cycle
is an odd permutation if and only if m is even. Now suppose σ is a permutation written in disjoint
58
cycle notation so σ = α1 . . . αk . Then ϵ(σ) = ϵ(α1 ) . . . ϵ(αk ) and you can determine ϵ(αi ) by seeing
if the length of the cycle is even or odd. Hence, you can determine the sign of σ by looking at it.

Proposition 120. The permutation of σ odd if and only if the number of cycles of even length in
its cycle decomposition is odd.

Example 121. σ = (1 2 3 4 5 6)(7 8 9)(10 11) is even, since the cycle lengths are 6, 3, 2 so are
odd, even, odd. Altogether, this is even.

Group Actions.
We now return (as promised) to studying group actions in detail. In addition to being interesting in
their own right, it turns out you can learn a lot about a group from its actions (this is basically the
whole idea behind representation theory). This study will culminate in Sylow’s Theorem, which is
a strong classification theorem for groups (that is extremely important on the qualifying exams).

Make sure to review the earlier content on group actions, as we will need to understand orbits,
stabilizers, permutation representations, kernels, etc.

Here is a quick review of key results of group actions we have already seen. They are here for
the convenience of the reader, and I did not repeat these definitions and results in lecture.

Recall that if G ↷ A then for each g ∈ G, the map σg : A → A defined by σg : a 7→ g · a

is a permutation of A. Hence, there is a map φ : G → SA defined by φ(g) = σg which is a
homomorphism called the permutation representation of the action.

Also recall the following definitions:

(1) The kernel of the action is {g ∈ G | g · a = a for all a ∈ A}.

(2) For each a ∈ A the stabilizer of a in G is Ga = {g ∈ G | g · a = a}.

Note that the kernel of an action is the same thing as the kernel of the permutation represen-
tation. Hence, the kernel is a normal subgroup of G. Two group elements induce the same
permutation if and only if they are in the same coset of the kernel.

Recall that Ga ≤ G. The kernel of the action is contained in every stabilizer. Hence, the kernel
T
of the action is a∈A Ga .

59
Definition 122. An action G ↷ A is called faithful if its kernel is the identity. [This means that
the only group element that acts as the identity permutation is the identity element.]

Example 123. Let G = D8 act on the four vertices of a square labeled 1–4 clockwise. Let r
be the rotation and s be the reflection through vertex 1. Then the permutation representation
has
φ(r) = σr = (1 2 3 4) φ(s) = σ2 = (2 4).
Since the permutation representation is a homomorphism, σrs = (1 4)(2 3). The action of D8
on the four vertices is faithful since only the identity symmetry fixes all four vertices. The
stabilizer of any vertex a is the subgroup of order 2 generated by the reflection about the line
passing through the vertex.

We recall the following results that were proved or will be proved on Problem Sets:

Proposition 124. For any group G and set A, there is a bijection between actions of G on A and
homomorphisms G → SA .

Proposition 125. A group action G ↷ A partitions A into equivalence classes called orbits.

Theorem 126 (Orbit–Stabilizer Theorem). Let G ↷ A. Then

(1) the number of elements in the orbit of a is |G : Ga |, the index of the stabilizer of a, and
(2) if g · x = y then Gy = gGx g −1 .

Proof. A future Problem Set exercise. □

Corollary 127. If |G| < ∞ then the size of any orbit is a divisor of |G|.

Proof. Apply Lagrange’s Theorem to the above. □

60
Definition 128. Let G ↷ A. If there is only one orbit then the action is called transitive. This
means that for all a, b ∈ A, there exists a g ∈ G such that g · a = b.

We can use this to prove the uniqueness of cycle decompositions we stated when we studied the
symmetric group.

Let A = {1, 2, . . . , n} and let σ ∈ Sn . We will show that σ has a unique cycle decomposition. Let
G = ⟨σ⟩. Then G ↷ A and so partitions {1, 2, . . . , n} into a set of disjoint orbits. If i ∈ {1, 2, . . . , n}
then the orbit of i under G will be {i, σ(i), σ 2 (i), . . . , σ d−1 (i)} where σ d (i) = i. Hence, on this subset
of {1, 2, . . . , n}, G acts as a d-cycle. This gives us one cycle in our cycle decomposition. Do this
for the rest of the disjoint orbits to get σ as a product of disjoint cycles.

Groups acting by left multiplication.

There are two specific actions of a group G that are extremely important and enlightening for
discovering the structure of the group. These come from left multiplication and from conjugation.

Recall that if G is any group, we can let A = G and have G act on itself by left multiplication

g·a=a

for all g, a ∈ G. We called this the left regular action. This actions is transitive and faithful, and
the stabilizer of any point is the identity. We can actually generalize this a bit.

Let H ≤ G and let A be the set of all left cosets of H in G. Although the set of cosets need not be
a group (it is iff H ⊴ G), it is still a set. We can act on the set of left cosets by

g · aH = gaH

for all g ∈ G, aH ∈ A. When H = {e}, this is the same thing as the left regular action. Hence,
any results we prove about actions on the set of left cosets also specialize to the left regular action.

Theorem 129. Let H ≤ G, let A be the set of left cosets of H and let G ↷ A by left multiplication.
Let πH be the permutation representation afforded by this action. Then

(1) G acts transitively on A

(2) the stabilizer in G of the point 1H ∈ A is the subgroup H
(3) the kernel of the action (so the kernel of πH ) is x∈G xHx−1 , and ker πH is the largest
T

normal subgroup contained in H.

61
Proof. Let aH, bH ∈ A and let g = ba−1 . then g · aH = bH so the action is transitive. For (2), the
stabilizer of 1H is defined to be {g ∈ G | gH = H} which is {g ∈ G | g ∈ H} = H.

For (3), by definition of πH we have

ker πH = {g ∈ G | gxH = xH for all x ∈ G} = {g ∈ G | (x−1 gx)H = H for all x ∈ G}

\
= {g ∈ G | x−1 gx ∈ H for all x ∈ G} = {g ∈ G | g ∈ xHx−1 for all x ∈ G} = xHx−1 .
x∈G

Now note that ker πH ⊴ G since it is a kernel and we just showed ker πH ≤ H since H is all the points
that fix 1H. Now if N is any normal subgroup contained in H, we have N = xN x−1 ≤ xHx−1 for
all x ∈ G so N ≤ x∈G xHx−1 = ker πH .
T
□

Corollary 130 (Cayley’s Theorem). Any finite group with |G| = n is isomorphic to a subgroup of
Sn .

Proof. Let G act on itself by left multiplication. This gives a homomorphism φ : G → SG . If φ is

∼ im φ ≤ Sn . But it is injective since if φ(g) = φ(h) then φ(g)(1) = φ(h)(1) so
injective then G =
g = h. □

This is a cool result but not necessarily computationally practical. To study a group of order n
you need to understand subgroups in a group of order n! which can be much bigger.

The following theorem allows you to show that large subgroups of a group must be normal. It also
uses a really important technique: acting on the set of left cosets of a subgroup by multiplication.
Note that in general, a group G need not have a subgroup of index p where p is the smallest prime
dividing |G|. However, if it does, we can say the following.

Corollary 131. If G is a finite group of order n and p is the smallest prime dividing |G|, then any
subgroup of index p is normal.

Proof. Suppose H ≤ G and |G : H| = p. Let πH be the permutation representation of the

multiplication action of the set of left cosets of H in G. Let K = ker πH and suppose |H : K| = k.
Then we know |G : K| = |G : H||H : K| = pk. Since H has p left cosets, we know that G/K is
isomorphic to a subgroup of Sp (the image of G under πH ) by the first isomorphism theorem.

Hence, by Lagrange’s Theorem, pk = |G/K| divides p!. Hence, k divides p!/p = (p − 1)!. Now note
that all prime divisors of (p − 1)! are less than p, and since we assumed that p is minimal, every
62
prime divisor of k is greater than or equal to p. This forces k = 1, so H = K ⊴ G, completing the
proof. □

Groups acting by conjugation.

We now consider G acting on itself A = G by conjugation

g · a = gag −1

for all g, a ∈ G, where gag −1 is computed in G as usual. This definition satisfies the two axioms
for a group action as

g1 · (g2 · a) = g1 · (g2 ag2−1 ) = g1 (g2 ag2−1 )g1−1 = (g1 g2 )a(g1 g2 )−1 .

Definition 132. Two elements a and b are conjugate in G if there is some g ∈ G such that
b = gag −1 . That is, if and only if they are in the same orbit of G acting on itself by conjugation.
The orbits of G acting on itself by conjugation are called the conjugacy classes of G.

Example 133. (1) If G is abelian then g · a = gag −1 = gg −1 a = a for all g, a ∈ G and

each a ∈ G is its own conjugacy class {a}.
(2) If |G| > 1 then unlike the action by left multiplication, G does not act transitively on
itself by conjugation because {e} is always a conjugacy class. More generally, {a} is a
conjugacy class if and only if gag −1 = a for all g ∈ G, i.e., if a ∈ Z(G).
(3) In GLn (F ), conjugation is the same as change of basis A 7→ P AP −1
(4) In S3 we can compute the conjugacy classes, {1}, {(1 2), (2 3), (1 3)}, {(1 2 3), (1 3 2)}.
One way to see this is to simply compute some of the conjugates. Then notice that
conjugating by an element cannot change the sign of the permutation so the 2-cycles
and 3-cycles must be in separate conjugacy classes.

63
 10. Monday 10/3: Conjugacy Classes in Sn and the Simplicity of A5

Remark. We did not quite have enough time to finish everything in this lecture. Make sure you
read the results at the end somewhat carefully.

The last thing we saw last time was the action of G on itself by conjugation, where if we let A = G,
then we define g · a = gag −1 .

We can also generalize the action of conjugation on G to an action on conjugation of subsets of G.

If S ⊆ G define
gSg −1 = {gsg −1 | s ∈ S}.
A group acts on the set P(G) of all subsets of itself by defining g · S = gSg −1 . The previous
action of G acting on itself by conjugation is the same thing as the action of conjugation on the
one-element subsets of G.

Definition 134. Two subsets S, T ⊆ G are conjugate in G if there is some g ∈ G such that
T = gSg −1 (iff they are in the same orbit of G acting on its subsets by conjugation).

Since group actions G ↷ A partitions A into orbits whose size we can count using the size of the
stabilizer, we have the following proposition.

Proposition 135. The number of conjugates of a subset S in G is the index of its normalizer,
namely, |G : NG (S)|. In particular, the number of conjugates of an element s of G is the index of
the centralizer of s, |G : CG (s)|.

Proof. This is clear since the stabilizer of a subset S is

{g ∈ G | gSg −1 = S} = NG (S)

and for a single element set {s}, NG (s) = CG (s).

The number of conjugates of S is the size of the orbit of S, so the result now follows from the
Orbit-Stabilizer Theorem. □

The action of G on itself by conjugation partitions G into the conjugacy classes of G whose orders
can be computed using the above. Since the sum of the orders of the conjugacy classes is the order
of G, we obtain the following important equation:
64
Theorem 136 (The Class Equation). Let G be a finite group and leg g1 , g2 , . . . , gr be represen-
tatives of the distinct conjugacy classes of G not contained in the center Z(G) (recall that each
element in the center is its own conjugacy class). Then
r
X
|G| = |Z(G)| + |G : CG (gi )|.
i=1

This is proved by just looking at the orbits of the elements of G under conjugation. Note that
since all of the summands on the right hand side of the equation must divide the order of the
group (since they are subgroups), their possible values are restricted. Even though (with all of the
technology we have developed so far), the class equation is nearly a trivial observation, it gives
powerful numerical tools to study groups of finite order.

Example 137. (1) The class equation gives no information for an abelian group since all
the classes have size 1 and Z(G) = G.
(2) Let G = Q8 be the quaternions. Consider CG (i). Note that ⟨i⟩ ≤ CG (i) and |⟨i⟩| = 4.
Since i ̸∈ Z(G), we must have CG (i) = ⟨i⟩. Hence, i has precisely |G : ⟨i⟩| = 2 conjugates
in Q8 , namely i and −i = kik −1 . The other conjugacy classes are determined similarly:

{1}, {−1}, {±i}, {±j}, {±k}.

The first two classes form Z(Q8 ) and the class equation says

|Q8 | = 2 + 2 + 2 + 2.

Maybe the most famous application of the class equation is to show that groups of prime power
oder must have nontrivial centers. This is a very useful fact when studying groups.

Theorem 138. If p is prime and P is a group of prime power order pα for some α ≥ 1 then P has
a nontrivial center: Z(P ) ̸= 1.

Proof. By the class equation

r
X
|P | = |Z(P )| + |P : CP (gi )|
i=1
65
where g1 , . . . , gr are representatives of the distinct non central conjugacy classes. By definition,
CP (gi ) ̸= P for each i so p divides |P : CP (gi )|. Since p also divides |P |, it follows that p | |Z(P )|
so the center is nontrivial. □

Corollary 139. If |P | = p2 for a prime p, then P is abelian. More precisely, P is isomorphic to

either Zp2 or Zp × Zp .

Proof. By the above theorem Z(P ) ̸= 1. Hence, P/Z(P ) is of order p or 1 and hence is cyclic. By
an exercise from a previous problem set, this means P is abelian. If P has an element of order
p2 then it is cyclic and isomorphic to Zp2 . So assume that every nonidentity element of P has
order p. Let x be any nonidentity element and let y ∈ P \ ⟨x⟩. Since |⟨x, y⟩| > |⟨x⟩| = p, we must
have P = ⟨x, y⟩. Both x and y have order p so ⟨x⟩ × ⟨y⟩ = Zp × Zp . It now follows that the map
(xa , y b ) 7→ xa y b is an isomorphism from ⟨x⟩ × ⟨y⟩ to P . □

Conjugacy classes in Sn .
We will now consider the conjugacy classes in Sn . Much like how in GLn (F ), conjugation was just
change of basis, we will have a similar situation in Sn 7.

Proposition 140. Let σ, τ ∈ Sn and suppose σ has cycle decomposition

(a1 a2 . . . ak1 )(b1 b2 . . . bk2 ) · · ·

then τ στ −1 has cycle decomposition

(τ (a1 ) τ (a2 ) . . . τ (ak1 ))(τ (b1 ) τ (b2 ) . . . τ (bk2 )) · · ·

i.e. τ στ −1 is obtained from σ by replacing each entry i in the cycle decomposition for σ by the
entry τ (i).

Proof. Observe that if σ(i) = j then

τ στ −1 (τ (i)) = τ (j)

Thus, if the ordered pair (i, j) appears in the cycle decomposition for σ then the ordered pair
(τ (i), τ (j)) appears in the cycle decomposition for τ στ −1 . This completes the proof. □

The above proposition says that for σ ∈ Sn every conjugate of σ has the same cycle type as σ.
7Where the “change of basis” here is just permuting the n elements around. In GL (F ), the change of basis was
n
mapping each vector in a basis to a new vector in a different basis.
66
 Example 141. Let σ = (1 2)(3 4 5) and τ = (1 3 2 4) in S5 . Then

τ στ −1 = (3 4)(2 1 5).

Definition 142. If σ ∈ Sn is the product of disjoint cycles of lengths n1 , n2 , . . . , nr with n1 ≤ n2 ≤

· · · ≤ nr (including its 1-cycles) then the integers n1 , n2 , . . . , nr are called the cycle type of σ.

Definition 143. If n ∈ Z+ , a partition of n is any nondecreasing sequence of positive integers

whose sum is n.

By what we have already proved about Sn , the cycle type of a permutation is unique. For example,
the cycle type of an m-cycle in Sn is 1, 1, 1, · · · , 1, m with n − m 1’s.

Proposition 144. Two elements of Sn are conjugate in Sn if and only if they have the same cycle
type. Hence, the number of conjugacy classes of Sn equals the number of partitions of n.

Proof. By the previous proposition, conjugate permutations have the same cycle type. Conversely,
suppose that σ1 and σ2 have the same cycle type. Order the cycles in nondecreasing length,
including 1-cycles at the beginning (if several cycles have the same length, then there are several
ways of doing this). Ignoring parentheses, each cycle decomposition is a list in which all the integers
from 1 to n appear exactly once. Define τ to be the function which maps the ith integer in the list
for σ1 to the ith integer in the list for σ2 .

Thus τ is a permutation and since the parentheses which delineate the cycle decompositions appear
at the same positions in each list, Proposition 140 ensures that τ σ1 τ −1 = σ2 so that σ1 and σ2 are
conjugate.

Since there is a bijection between the conjugacy classes of Sn and the permissible cycle types and
each cycle type for a permutation in Sn is a partition of n, the second assertion follows. □

Note that above, τ is not unique (since you could rearrange cycles of the same length). For example,
for the identity 1 ∈ Sn , you can take any τ ∈ Sn and τ 1τ −1 = 1.
67
Example 145. (1) If σ1 = (1)(3 5)(8 9)(2 4 7 6) and σ2 = (3)(4 7)(8 1)(5 2 6 9) then you
can define τ (1) = 3, τ (3) = 4, τ (5) = 7, etc. so τ = (1 3 4 2 5 7 6 9)(8) and τ σ1 τ −1 = σ2
(2) Again, this is not the only choice of τ . We could rearrange σ2 = (3)(1 8)(4 7)(5 2 6 9).
(3) If n = 5, the conjugacy classes are the partitions of 5 and we have the following
representatives of the conjugacy classes:

Partition of 5 Representative of Conjugacy Class

1, 1, 1, 1, 1 1
1, 1, 1, 2 (1 2)
1, 1, 3 (1 2 3)
1, 4 (1 2 3 4)
5 (1 2 3 4 5)
1, 2, 2 (1 2)(3 4)
2, 3 (1 2)(3 4 5)

(4) We will need the following computations in a minute so let’s do them. We can use the
above proposition to count the number of conjugates of an element, then use Proposi-
tion 135 to determine the centralizer of an m-cycle in Sn . If σ ∈ Sn is an m-cycle then
the number of conjugates of σ (number of m-cycles) is8
n · (n − 1) · · · (n − m + 1) n!
= .
m (n − m)! · m
By Proposition 135, the number of conjugates is the index of the centralizer:
|Sn |/|CSn (σ)|. Hence, we must have |CSn (σ)| = m · (n − m)!.
The element σ commutes with its m distinct powers so 1, σ, . . . , σ m−1 ∈ CSn (σ).
Further, it commutes with any permutation in Sn whose cycles are disjoint from σ and
there are (n − m)! permutations of this type. The product of elements of these types
account for m(n − m)! elements so this is the full centralizer. In other words,

CSn (σ) = {σ i τ | 0 ≤ i ≤ m − 1, τ ∈ Sn−m }.

Applying this to the above table, we can compute

Partition of 5 Representative of Conjugacy Class Size of Conjugacy Class
1, 1, 1, 1, 1 1 1
1, 1, 1, 2 (1 2) 5 · 4/2 = 10
1, 1, 3 (1 2 3) 5 · 4 · 3/3 = 20
1, 4 (1 2 3 4) 5 · 4 · 3 · 2/4 = 30
5 (1 2 3 4 5) 5!/5 = 24
1, 2, 2 (1 2)(3 4) ??
2, 3 (1 2)(3 4 5) 68 ??
 Actually, we can use the fact that |A5 | = 60 to complete the table. There are 60 even permuta-
tions, and so far we have accounted for 1 + 20 + 24 = 45 of them. This leaves 15 of type 1, 2, 2.
Finally, this leaves 20 of type 2, 3.

We now will give a proof that A5 is simple. We first observe:

Lemma 146. If H ⊴ G then for every conjugacy class K of G either K ⊆ H or K ∩ H = ∅ so every

normal subgroup is a union of conjugacy classes of G.

Proof. If x ∈ K ∩ H then gxg −1 ∈ gHg −1 for all g ∈ G. Since H ⊴ G, gHg −1 = H so H contains

all the conjugates of x so K ⊆ H. □

The previous lemma is quite powerful because it gives us some numerical conditions that could
show that a subgroup is not normal. We can compute the sizes of conjugacy classes, and so the
only normal subgroups need to be able to be represented as a sum of some of those classes. We
will show this is not possible in A5 as the main technique of the next proof.

Theorem 147. A5 is a simple group.9

Proof. We first work out the conjugacy classes of A5 .10 Proposition 144 about conjugacy classes in
Sn does not directly apply since two elements of the same cycle type (which are conjugate in S5 )
need not be conjugate in A5 (since you may need to conjugate by something not in A5 to show
they are conjugate in S5 ).

We have already seen that the representatives of the cycle types of even permutations in S5 can be
taken to be
1, (1 2 3), (1 2 3 4 5), (1 2)(3 4).
The centralizers of 3-cycles and 5-cycles were determined in the example above, and so to find the
centralizer in A5 , we have

CA5 ((1 2 3)) = ⟨(1 2 3)⟩ and CA5 ((1 2 3 4 5)) = ⟨(1 2 3 4 5)⟩.

8Since you have n choices for the first element, n − 1 for the second, and repeat this m times. We have then
overcounted by a factor of m, because any of the m elements could be written first in the list.
9The first nonabelian simple group we know about. The only other simple groups we have seen are Z for a prime p.
p
10A full determination of conjugacy classes in A will be worked through in a detailed problem set exercise.
n
69
since in the former case these are all of the ones that are even (multiplying by (4 5) gives an odd
permutation). These groups have orders 3 and 5 respectively, so the number of conjugates of (1 2 3)
is the index |A5 : CA5 ((1 2 3))| = 20 and the number of conjugates of (1 2 3 4 5) is 12. Since there
are a total of twenty 3-cycles in S5 (since (5 · 4 · 3)/3 = 20), and all lie in A5 , all twenty 3-cycles
are conjugate in A5 .

Not all 5-cycles are conjugate in A5 , as there are twenty-four 5-cycles (since 5!/5 = 24), but
(1 2 3 4 5) is only conjugate to twelve of them. However, this must mean that the 5-cycles form
two conjugacy classes of size twelve in A5 .

In A5 , the 3-cycles and 5-cycles account for all the nonidentity elements of odd order. The remaining
15 nonidentity elements must have order 2 and therefore have cycle type (2,2). It is easy to see
that (1 2)(3 4) commutes with (1 3)(2 4) and (1 4)(2 3) but does not commute with any element
of odd order in A5 . It also does not commute with any product of 2-cycles which contains a 5 in
its decomposition. Hence, |CA5 ((12)(34))| = 4. Hence, (1 2)(3 4) has 15 distinct conjugates in A5
and so all 15 elements of order two in A5 are conjugate to (1 2)(3 4).

Hence, the conjugacy classes of A5 have order 1, 15, 20, 12, and 12.

Suppose H ⊴ A5 . Then H would be a union of conjugacy classes of A5 . Then the order of H would
be a divisor of 60 and a sum of the sizes of some of the conjugacy classes. It also has to include
the conjugacy class of size 1 (the identity is always in H). The only possibilities are |H| = 1 or
|H| = 60. Hence, A5 has no proper nontrivial normal subgroups □

Definition 148. Let G be a group. An automorphism of G is an isomorphism G → G. The set of

all automorphisms of G is a group under function composition, denoted Aut(G).

Proposition 149. Let H ⊴ G. Then G acts by conjugation on H as automorphisms of H. More

specifically, for each g ∈ G, the action is given by h → ghg −1 for each h ∈ H. For each g ∈ G,
conjugation by g is an automorphism of H. The permutation reprsentation afforded by this action
is a homomorphism of G into Aut(H) with kernel CG (H).

Proof. Since H is normal, conjugation sends elements of H to elements of H so this is an action.

You can check that conjugation gives an automorphism of H. For the last claim, let ψ : G → SH
be the permutation representation of the action. Then

ker ψ = {g ∈ G | ghg −1 = h for all h ∈ H} = CG (H). □

70
Corollary 150. If K ≤ G and g ∈ G, then K ∼
= gKg −1 . Conjugate elements and conjugate
subgroups have the same order.

This was proved in the homework.

Corollary 151. For any subgroup H ≤ G, the quotient group NG (H)/CG (H) is isomorphic to a
subgroup of Aut(H). In particular, G/Z(G) is isomorphic to a subgroup of Aut(G).

Proof. H ⊴ NG (H) then apply Proposition 149. For the particular case, let H = G. □

71
 11. Wednesday 10/5: Sylow’s Theorem

Recall that last time we saw that G/Z(G) is isomorphic to a subgroup of Aut(G).

Definition 152. Let G be a group and let g ∈ G. Conjugation by g is called an inner automorphism
of G and the subgroup of Aut(G) consisting of all inner automorphisms is denoted Inn(G). Any
non-inner automorphism is called outer.

It is easy to see that Inn(G) is a subgroup. The previous corollary is an important one:

Inn(G) ∼
= G/Z(G).

Note also that if H ⊴ G, then conjugation in H by an element of G is an automorphism of H but

need not be inner.

Example 153. What is Inn(Sn )? If n = 2, then Aut(S2 ) is trivial since any automorphism
must fix the identity element. So suppose n ≥ 3. Then Z(Sn ) = 1. Hence, Inn(Sn ) = Sn .

Interestingly, although we are not yet able to prove it, for all n ̸= 6, Aut(Sn ) = Inn(Sn ) ∼
= Sn .
When n = 6, we have | Aut(Sn ) : Inn(Sn )| = 2.

Definition 154. H ≤ G is called characteristic in G, denoted H char G if every automorphism of

G maps H to itself, i.e. σ(H) = H for all σ ∈ Aut(G).

The following are true of characteristic subgroups. You will prove them in the homework exercises.

(1) characteristic subgroups are normal,

(2) if H is the unique subgroup of G of a given order, then H char G,
(3) if K char H and H ⊴ G then K ⊴ G.

Proposition 155. Aut(Zn ) ∼

= (Z/nZ)× , an abelian group of order φ(n).

Proof. Recall that (Z/nZ)× = {a | gcd(a, n) = 1}. Any automorphism in Aut(Zn ) must map the
generator x to some xa . Let ψa (x) = xa for some a ∈ Z and the integer a uniquely determines ψa .
We proved in a homework exercise that ψa is surjective if and only if gcd(a, n) = 1. Hence, there
72
is a surjective map
Ψ : Aut(Zn ) → (Z/nZ)×
where ψa 7→ a. Ψ is a homomorphism since ψa ◦ ψb (x) = ψa (xb ) = (xb )a = xab = ψab (x). Finally,
clearly Ψ is injective. □

Sylow’s Theorem.

And now onto one of the big theorems of finite group theory. Given a finite group G and d a divisor
of |G|, when does G have a subgroup of d? We’ve already seen in Cauchy’s Theorem that if d is
prime, G always does.

Definition 156. Let G be a group and p a prime.

(1) A group of order pα is called a p-group. Subgroups of G which are p-groups are called
p-subgroups.
(2) If G is a group of order pα m where p ∤ m, then a subgroup of order pα is called a Sylow
p-subgroup of G.
(3) The set of Sylow p-subgroups of G will be denoted Sylp (G). The number of Sylow p-
subgroups will be denoted np .

A previous algebra student of mine needlepointed the following motto: https://cpb-us-e1.

wpmucdn.com/blogs.gwu.edu/dist/1/4020/files/2021/07/grouptheory.jpg. The first rule of
group theory is: you do not forget Sylow’s Theorem. The second rule of group theory is: you do
not forget Sylow’s Theorem. Anyway, on to Sylow’s Theorem.

Theorem 157 (Sylow’s Theorem). Let G be a finite group, p a prime number, |G| = pα m with
p ∤ m.

(1) For all 0 ≤ i ≤ α, G has a subgroup of order pi . In particular, Sylow p-subgroups exist.
(2) If Q is any p-subgroup of G and P is any Sylow p-subgroup of G, then Q ≤ gP g −1 for some
g ∈ G. In particular, all Sylow p-subgroups are conjugate.
(3) np = |G : NG (P )| where P is any Sylow p-subgroup. In particular, np ≡ 1 (mod p) and
np | m.

73
Proof of (1). Idea: Use the class equation and induction on two cases: p | |Z(G)| or p ∤ |Z(G)|.
Induct on |G|. Recall that the class equation says
X
|G| = |Z(G)| + |G : CG (x)|.
one x from each |K|>1

Case 1: p | |Z(G)|

Cauchy’s Theorem implies Z(G) contains an element x of order p. So ⟨x⟩ ⊴ G. Let G′ = G/⟨x⟩.
This G′ has order pα−1 m. By induction, G′ satisfies (1) and has a subgroup P ′ of order pi−1 for
any 1 ≤ i ≤ α. Let π : G → G/⟨x⟩ be the quotient homomorphism, and let P = π −1 (P ′ ). Then
|P | = pi so P ≤ G has the correct order.

Case 1: p ∤ |Z(G)|

From the class equation, |G : CG (x)| is not a multiple of p for some x ̸∈ Z(G). Clearly, |G :
CG (x)| > 1 (otherwise x ∈ Z(G)) so |CG (x)| = pα m′ , gcd(p, m′ ) = 1. The induction hypothesis
applied to CG (x) implies that CG (x) has a subgroup of order pi for all 1 ≤ i ≤ α. So G has such
subgroups. □

Proof of (2). Main idea: Study conjugates of P where P is a Sylow p-subgroup. Let S denote
the set of conjugates of P , that is S = {gP g −1 | g ∈ G}.

G acts on S by conjugation: g · (hP h−1 ) = ghP h−1 g −1 . Let Q be any p-subgroup of G. In fact,
look at Q acting on S in this way.

(†) Claim: If M ∈ S, then the orbit containing M has size |Q : M ∩ Q|.

• Apply this claim to Q = P . Given M ∈ S, the size of the orbit containing M is |P : M ∩ P |.

So one orbit of size 1 (M = P ). All other orbits have size pi for some i ≥ 1 (if M ̸= P ,
M ∩ P ⪇ P ) so |S| ≡ 1 (mod p). (Since |S| = 1 + pi ).
P

• Apply claim to arbitrary Q. Then S is the disjoint union of Q orbits. The orbit containing
M has size |Q : M ∩ Q|, some power of p. In particular, this forces some orbit to have size 1
(since in the previous bullet point, we saw |S| ≡ 1 (mod p)). |Q : M ∩ Q| = 1 some M ∈ S
implies Q ⊆ M . So Q ≤ M = gP g −1 for some g (part (2) follows).
So all Sylow p-subgroups are conjugate. If Q is a Sylow p-subgroup, Q ≤ gP g −1 implies
Q = gP g −1 since both have order pα . □

Proof of (3). Now let G act on S by conjugation. By definition of S, this action is transitive. Also
we just proved that S is the set of all Sylow p-subgroups. So |S| = np implies nP ≡ 1 (mod p). Now
the Orbit-Stabilizer theorem implies |S| = |G : StabG (P )| and StabG (P ) = {g ∈ G | gP g −1 = P }
which is the definition of NG (P ). Finally, nP | m since P ≤ NG (P ). Thus, |G : NG (P )| divides
|G : P | = m. This is because we showed that |G : P | = |G : NG (P )||NG (P ) : P | on a problem
set. □
74
Proof of (†). We need to prove (†), which is: if Q acts on S (the set of Sylow p-subgroups) by
conjugation and M ∈ S, then the size of the orbit containing M has size |Q : M ∩ Q|.

Orbit-Stabilizer says |O| = |Q : StabQ (M )| = |Q : NQ (M )| where we are being a bit loose, since
M ̸≤ Q. We mean
NQ (M ) := {x ∈ Q | xM x−1 = M } = NG (M ) ∩ Q.
We need that M ∩ Q = NG (M ) ∩ Q. Note that M ⊴ NG (M ).

It is clear that M ≤ NG (M ), so M ∩ Q ≤ NG (M ) ∩ Q. We must prove the reverse inclusion. Since

certainly NG (M ) ∩ Q ≤ Q, we must show that NG (M ) ∩ Q ≤ M .

Consider M (NG (M ) ∩ Q). This is a subgroup of G since NG (M ) ∩ Q normalizes M .

|M ||NG (M ) ∩ Q|
|M (NG (M ) ∩ Q)| =
|M ∩ (NG (M ) ∩ Q)|
which is a power of p, since |M | = pα and |NG (M ) ∩ Q| | |Q| is a power of p. So. M (NG (M ) ∩ Q)
is a p-subgroup of G, but M ≤ M (NG (M ) ∩ Q) and M is a Sylow p-subgroup. Since M is as large
as possible for any p-subgroup, this implies M (NG (M ) ∩ Q) = M . Hence NG (M ) ∩ Q ≤ M . □

Corollary 158. Let P be a Sylow p-subgroup of G. Then the following are equivalent:

(1) P is the unique Sylow p-subgroup of G, i.e. np = 1.

(2) P ⊴ G
(3) P char G
(4) All subgroups generated by elements of p-power order are p-groups, i.e., if X is any subset
of G such that |x| is a power of p for all x ∈ X, then ⟨X⟩ is a p-group.

Proof. If (1) holds, then gP g −1 = P for all g ∈ G since gP g −1 ∈ Sylp (G) and hence P ⊴ G.
Conversely, if P ⊴ G and Q ∈ Sylp (G), then by Sylow’s Theorem there exists g ∈ G such that
Q = gP g −1 = P . Thus, Sylp (G) = {P }. so (1) ⇔ (2).

(3) obviously implies (2). Conversely, if P ⊴ G, we just proved that P is the unique subgroup of
G of order pα so P char G. Hence, (2) ⇔ (3).

Finally, assume (1) holds and suppose X is a subset of G such that |x| is a power of p for all
x ∈ X. By the conjugacy part of Sylow’s Theorem, for each x ∈ X, there is some g ∈ G such that
x ∈ gP g −1 = P . Thus X ⊆ P and so ⟨X⟩ ≤ P and ⟨X⟩ is a p-group. Conversely, if (4) holds, let X
be the union of all Sylow p-subgroups of G. If P is any Sylow p-subgroup, P is a subgroup of the
p-group ⟨X⟩ (if ⟨X⟩ were not a p-group, then by Cauchy’s Theorem it would contain elements of
other prime order, but it doesn’t by hypothesis). Since P is a p-subgroup of G of maximal order,
we must have P = ⟨X⟩. Hence, (1) ⇔ (4). □
75
 12. Wednesday 10/12: Applications of Sylow and FTFGAG

Two announcements. Your exam is on Friday 10/14 (in two days) in Phillips 730. Also, it
appears that the department is not offering MATH 6102 next semester.11

Last time, we stated and proved Sylow’s Theorem. Today, we’ll see some applications and also
state the Fundamental Theorem of Finitely Generated Abelian Groups (FTFGAG). We defer the
proof for a few lectures (your book proves a more general result while studying modules over rings,
but that result would be in MATH 610212).

Example 159. Let G be a finite group and let p be prime.

• If p ∤ |G| the Sylow p-subgroup of G is the trivial group. If |G| = pα , then G is the
unique Sylow p-subgroup of G.
• A consequence of the FTFGAG is that a finite abelian group has a unique Sylow p-
subgroup for each prime p. This subgroup consists of all elements x whose order is a
power of p. This is sometimes called the p-primary component of the abelian group.

Sylow’s Theorem is extremely useful in classifying groups. In particular, Sylow’s Theorem often
can be used to prove the existence of a normal subgroup in groups of a particular order, showing
that there are no simple groups of that order. The first thing to do is to factor the group order
into prime powers. The largest prime divisors of the group order tend to give the fewest possible
values for np .

Example 160. Let p < q be distinct primes and let |G| = pq. Let P ∈ Sylp (G) and let
Q ∈ Sylq (G).

• nq | p, nq ≡ 1 (mod q) implies nq = 1. So Q ⊴ G and Q ∼

= Zq .
• np | q, np ≡ 1 (mod p) implies np = 1 or np = q (happens if q ≡ 1 (mod p) ⇔ p | q − 1).
So if p ∤ q − 1 then P ⊴ G, P ∼
= Zp . (In fact, if P ⊴ G and Q ⊴ G then G ∼
= Zp × Zq ∼ =
Zpq ).
Hence, if p < q, p ∤ q − 1, all groups of order pq are cyclic.
If Q ⊴ G, P ̸⊴ G, then G = ∼ Q ⋊ P . [Shortly, we will see what this notation means.]

11Sad algebraist noises.

12Sad algebraist noises intensify.
76
Example 161. |G| = 15, G is cyclic by the above.

Example 162. |G| = 105 = 3 · 5 · 7. Then n3 = 1 or 7, n5 = 1 or 21, n7 = 1 or 15.

An important technique is to count elements. We can use this to show that G has either a
normal Sylow 5- or Sylow 7-subgroup. Suppose not, then n5 = 21 and n7 = 15.

The Sylow 5-subgroups intersect in the identity. So there are 21 · (5 − 1) = 84 elements of

order 5. Similarly, there are 15 · (7 − 1) = 90 elements of order 7. Then |G| ≥ 174 which is a
contradiction. Hence, there are no simple groups of order 105.

Example 163. |G| = 30. We will show that G has a subgroup of order 15 (which must be
normal as it has index 2).

Let P ∈ Syl5 (G) and Q ∈ Syl3 (G). If either P or Q is normal in G, then P Q is a subgroup of
G and |P Q| = |P ||Q|/|P ∩ Q| = 15.

So assume that neither P nor Q is normal. Then n5 = 6 and n3 = 10. Again, counting elements
show that either P or Q is normal.

Example 164. Omitted in lecture. Read this example.

|G| = 12. We show that either G has a normal Sylow 3-subgroup or G ∼

= A4 .
If n3 = 1 then G has a normal Sylow 3-subgroup, so suppose n3 ̸= 1 and let P ∈ Syl3 (G).
Since n3 | 4 and n3 ≡ 1 (mod 3), then n3 = 4. Since distinct Sylow 3-subgroups intersect in
the identity and each contains two elements of order 3, G contains 8 elements of order 3. Since
|G : NG (P )| = n3 = 4, NG (P ) = P .

Another important technique! G acts on its four Sylow 3-subgroups by conjugation, giving a
homomorphism φ : G → S4 . The kernel K of this action is the subgroup of G which normalizes
all Sylow 3-subgroups of G. In particular, K ≤ NG (P ) = P . Since P is not normal in G by
assumption, K = {e} so φ is injective and G ∼
= φ(G) ≤ S4 .
Since G contains 8 elements of order 3 and there are 8 elements of order 3 in S4 all contained
in A4 , it follows that φ(G) intersects A4 in a subgroup of order at least 8. Since both groups
have order 12, φ(G) = A4 so G ∼ = A4 .

77
Example below assigned on a problem set.

Example 165. Omitted in lecture. Read this example.

|G| = p2 q for distinct primes p and q.

We show that G has a normal Sylow subgroup, so G cannot be simple. Let P ∈ Sylp (G) and
Q ∈ Sylq (G).

If p > q then since np | q and np = 1 + kp, we must have np = 1 so P ⊴ G.

Now suppose p < q. If nq = 1, then Q ⊴ G. So assume nq = 1 + tq > 1 for some t > 0. Now
nq divides p2 so nq = p or p2 . Since q > p, we cannot have nq = p, so nq = p2 . Thus,

tq = p2 − 1 = (p − 1)(p + 1).

Since q is prime, either q | p − 1 or q | p + 1 but the former is impossible since q > p. So q | p + 1

but this means q = p + 1. The only cases are p = 2, q = 3 so |G| = 12. Then the result follows
by the previous example.

Punchline: The numerics of Sylow’s Theorem give you a powerful tool to understand groups. You
can also count elements. You can also have G act on the set of Sylow p-subgroups via conjugation.
This gives you a permutation representation G → Snp , which you can leverage for more information.
Section 6.2 in your textbook is an amazing section illustrating several powerful techniques. If you
are taking the qual, acquaint yourself with this section.

The Fundamental Theorem of Finitely Generated Abelian Groups.

We now move on to the Fundamental Theorem of Finitely Generated Abelian Groups, which
(amazingly) classifies all finitely generated abelian groups. In particular, it classifies all finite
abelian groups. Recall that we have already seen the direct product of two groups. We extend this
to taking direct products of arbitrarily many groups.

Definition 166. If G and H are groups, the direct product

G × H = {(g, h) | g ∈ G, h ∈ H}

is a group with (g, h)(g ′ , h′ ) = (gg ′ , hh′ ) and |(g, h)| = lcm(|g|, |h|).

Now, if {Gα }α∈A where A is any index set is a collection of groups, then
Y
Gα = {(gα )α∈A | each gα ∈ Gα }
α∈A

all A-tuples is a group, the direct product of the Gα with coordinate-wise operations.

78
We also define
M
Gα = {(gα )α∈A | each gα ∈ Gα with only finitely many not the identity}
α∈A

the direct sum of the Gα with coordinate-wise operation.

Example 167. Of course, if we have a finite number of groups, then the direct product and
L Q
the direct sum are the same. In general, one only has Gα ≤ Gα .

Q
Example 168. • G = n≥1 Z/nZ then x = (1, 1, 1, . . . ) has infinite order in G.
L
• H = n≥1 Z/nZ has no elements of infinite order. Any element will have only finitely
many of the coordinates nonzero. Hence, the order of any element will be the lcm of the
orders of the nonzero coordinates (which is again finite). The element x in the direct
product above is not an element of the direct sum.
Q
• G = n≥1 Z/2Z then |G| = ∞ but all elements have order 2 (or 1).

Proposition 169. Let G = G1 × · · · × Gn . Then

|G| = |G1 ||G2 | . . . |Gn |

where if one of the |Gi | is infinite, then so is |G|.

Theorem 170. (1) Let G = G1 × · · · × Gn , then G contains an isomorphic copy of Gi as a

normal subgroup:

Hi = {(1, 1, . . . , x, 1, . . . , 1) | x ∈ Gi } ⊴ G.

Further, G/Hi ∼= G1 × · · · × Gi−1 × Gi+1 × · · · × Gn .

(2) For each fixed i, define πi : G → Gi by

πi ((g1 , g2 , . . . gn )) = gi .

Then πi is a surjective homomorphism with

ker πi = {(g1 , . . . , gi−1 , 1, gi+1 , . . . , gn ) | gj ∈ Gj } ∼

= G1 × · · · × Gi−1 × Gi+1 × · · · × Gn .
79
 (3) Also if i ̸= j, x ∈ Hi , y ∈ Hj , then xy = yx.

Proof. (1) Gi ∼
= Hi via the map g 7→ (1, 1, . . . , g, 1, . . . , 1). To see that Hi is normal, consider
the map
G → G1 × · · · × Gi−1 × Gi+1 × · · · × Gn
where (g1 , . . . , gn ) 7→ (g1 , . . . , gi−1 , gi+1 , . . . , gn ) (the map erases the ith component). The
kernel of this homomorphism is Hi and it is clearly surjective.

The other parts are similar. □

For a direct product, there is the inclusion Gi ,→ G and the projection G ↠ Gi . Hence, Gi is both
a subgroup and a quotient group of G.13

Definition 171. A group G is finitely generated if there is a finite subset A of G such that G = ⟨A⟩.

i.e., you have a finite set of generators such that every element of G can be written as a product of
the generators.

Notice that any finite group is certainly finitely generated. Just take the entire group G for the
generators.

Definition 172. For r ∈ Z with r ≥ 0, let Zr = Z × · · · × Z be a direct product of r copies of the

group Z where Z0 = {e}. The group Zr is called the free abelian group of rank r.14

Also Zr is finitely generated by the generators e1 , . . . , er .

We now state a theorem that classifies all finitely generated (and therefore all finite) abelian groups!
As mentioned before, we defer the proof.

First just something useful to know:

Lemma 173. If gcd(m, n) = 1 then Zmn ∼

= Zm × Zn .

Theorem 174 (FTFGAG). Let G be a finitely generated abelian group. Then

(1) G ∼
= Zr × Zn1 × · · · × Zns for some integers r, n1 , . . . , ns satisfying

13Note that G ,→ H is used as notation to mean that the map is an injection. G ↠ H is used to mean that the map
is a surjection.
14In a few lectures, when we discuss free groups, we will see why this group deserves to be called a free abelian group.
It is certainly abelian.
80
 (a) r ≥ 0 and nj ≥ 2 for all j, and
(b) ni+1 | ni for 1 ≤ i ≤ s − 1
(2) and the expression in (1) is unique. If G ∼
= Zt × Zm1 × · · · × Zmu where t and the mi satisfy
(a) and (b), then t = r, u = s, and mi = ni for all i.

That is, any finitely generated abelian group is a direct product of cyclic groups. The Zr is the free
part of G while the Zn1 × · · · × Zns is the torsion subgroup of G.

Definition 175. The integer r is called the free rank or Betti number of G. The integers n1 , . . . , ns
are called the invariant factors of G. The description of G in the above theorem is called the
invariant factor of decomposition of G.

Remark. This is false if G is an infinitely generated abelian group. For example, (Q, +) is abelian,
but Q ̸∼
= H × K for any groups H, K nontrivial.

Due to the uniqueness statements in FTFGAG, two finitely generated abelian groups are isomorphic
if and only if they have the same Betti number/free rank and the same list of invariant factors.

To find (up to isomorphism) all of the finite abelian groups of a given order, is equivalent to finding
all possible lists of invariant factors: i.e., sequences of integers n1 , n2 , . . . ns such that

(1) nj ≥ 2
(2) ni+1 | ni
(3) n1 n2 · · · ns = n.

Note that n1 ≤ n2 ≤ · · · ≤ ns . If p is any prime divisor of n, then p must divide ni for some i and
also must divide nj for all j ≤ i. Hence, every prime divisor of n must divide the first invariant
factor n1 . In particular, if n is the product of distinct primes (and so squarefree), then n = n1 and
there is a unique abelian group of order n.

Corollary 176. If n is the product of distinct primes, then up to isomorphism the only abelian
group of order n is Zn .

We now do an example of finding possible invariant factors for a given order n.

81
 Example 177. Suppose n = 180 = 22 · 32 · 5. Every prime factor must divide the first invariant
factor, so 2 · 3 · 5 | n1 . Hence, the possible values of n1 are

n1 = 22 · 32 · 5, 22 · 3 · 5, 2 · 32 · 5, 2 · 3 · 5.

For each choice of n1 , we must figure out all possible n2 ’s such that n2 | n1 and n1 n2 | n. For
each possible n2 , we must figure out all possible n3 ’s, etc.

Invariant Factors Abelian Groups

22 · 32 · 5 Z180
2· 32 · 5, 2 Z90 × Z2
22 · 3 · 5, 3 Z60 × Z3
2 · 3 · 5, 2 · 3 Z30 × Z6

There is another way to describe the torsion subgroup of a finitely generated abelian group

Theorem 178 (FTFGAG, Elementary Divisors). Let G be a finitely generated abelian group.
Then
G∼
= Zr × Zn1 × · · · × Zns
for some integer r ≥ 0 and prime powers n1 , . . . , ns . The n1 , . . . , ns are called the elementary
divisors of G and they are unique up to order.

Before we do some examples computing elementary divisors and going back and forth between
elementary divisors and invariant factors, we have the following useful proposition.

Proposition 179. (1) Zm × Zn ∼ = Zmn if and only if gcd(m, n) = 1.

(2) (An easy corollary): If n = p1 . . . pαk k then Zn ∼
α1
= Z p α1 × · · · × Z p αk .
1 k

Proof. [⇒]. Let Zm = ⟨x⟩ and Zn = ⟨y⟩ and let ℓ = lcm(m, n). By contrapositive, suppose that
gcd(m, n) ̸=!. Then ℓ = mn/ gcd(m, n) ̸= mn. Let (xa , y b ) be an arbitrary element of Zm × Zn .
Then
(xa , y b )ℓ = (xℓa , xℓb ) = (e, e).
Hence, every element of Zm × Zn has order at most ℓ, hence has order strictly less that mn so
Zm × Zn ≁ Zmn .
=
82
[⇐]. Conversely, if gcd(m, n) = 1, then |xy| = lcm(|x|, |y|) = mn. Hence, xy generates Zm × Zn so
Zm × Zn is cyclic and hence isomorphic to Zmn . □

Example 180. Consider the group G = Z35 × Z105 × Z25 × Z49 . It is quite easy to find the
elementary divisors using the above theorem. If any of the factors do not have prime power
order, break it up as a product of prime powers. So

G∼
= Z5 × Z7 × Z3 × Z5 × Z7 × Z25 × Z49

so the elementary divisors are 3, 5, 5, 25, 7, 7, 49.

To find the invariant factors, for each distinct prime, list all of the elementary factors in de-
3 25 49
creasing order. So here: 5 7 The invariant factors are then 3 · 25 · 49 = 3675, 35, and
5 7
35 so G ∼
= Z3675 × Z35 × Z35 .

83
 Friday 10/14: Exam 1

You had your first midterm exam today.

13. Monday 10/17: Semidirect Products

Your syllabus says that today we will cover groups of small order (section 5.3), but this is essentially
a one-page chapter. You should read it on your own.

We have seen that you can use a direct product to construct larger groups from smaller ones. Today
we try to do the reverse: when can a larger group be decomposed as a direct product of smaller
groups? We already saw that we could do this for cyclic groups in the previous section. Recall that
if gcd(m, n) = 1 then Zmn ∼
= Zm × Zn . Today, we will develop a criterion for doing this in general.
Before we do this, we take a brief detour to discuss commutators.

Definition 181. Let G be a group, x, y ∈ G and let A, B be nonempty subsets of G.

(1) Define [x, y] = x−1 y −1 xy, called the commutator of x and y (notice that x and y commute
if and only if [x, y] = 1).
(2) Define [A, B] = ⟨[a, b] | a ∈ A, b ∈ B⟩, the group generated by commutators of elements
from A and from B.
(3) Define G′ = ⟨[x, y] | x, y ∈ G⟩, the subgroup of G generated by commutators of elements
from G, called the commutator subgroup of G.

Proposition 182. Let G be a group, x, y ∈ G, H ≤ G. Then

(1) xy = yx[x, y]
(2) H ⊴ G if and only if [H, G] ≤ H
(3) σ[x, y] = [σ(x), σ(y)] for any automorphism σ of G, also G′ char G and G/G′ is abelian.
(4) G/G′ is the largest abelian quotient of G in the sense that if H ⊴ G and G/H is abelian,
then G′ ≤ H. Conversely, if G′ ≤ H then H ⊴ G and G/H is abelian.
(5) If φ : G → A is any homomorphism of G into an abelian group A then φ factors through
G/G′ . That is, G′ ≤ ker φ and the following diagram commutes:15

G G/G′

φ
A

84
Proof. (1) is immediate from the definition.

For (2), H ⊴ G if and only if g −1 hg ∈ H for all g ∈ G and all h ∈ H. For h ∈ H, g −1 hg ∈ H if

and only if h−1 g −1 hg ∈ H so H ⊴ G if and only if [h, g] ∈ H for all h ∈ H and all g ∈ G. Hence,
H ⊴ G if and only if [H, G] ≤ H.

(3) The first part is routine. Now assuming the first part, any automorphism σ takes commutators
to commutators. By using σ −1 , it follows that any automorphism σ maps the set of commutators
bijectively to itself. Hence, σ(G′ ) = G′ so G′ char G.

To see that G/G′ is abelian, let xG′ and yG′ be elements of G/G′ . Then

(xG′ )(yG′ ) = xyG′ = yx[x, y]G′ = yxG′ = (yG′ )(xG′ )

so G/G′ is abelian.

(4) Suppose H ⊴ G and G/H is abelian. Then for all x, y ∈ G we have (xH)(yH) = (yH)(xH), so

H = (xH)−1 (yH)−1 (xH)(yH) = x−1 y −1 xyH = [x, y]H

so [x, y] ∈ H so G′ ≤ H.

Conversely, if G′ ≤ H then G/G′ is abelian and so every subgroup of G/G′ is normal. In particular,
H/G′ ⊴ G/G′ and by the Fourth Isomorphism Theorem, H ⊴ G. By the Third Isomorphism
Theorem
G/H ∼
= (G/G′ )/(H/G′ )
hence G/H is abelian.

(5) is simply (4) phrased in terms of homomorphisms. It is very useful to get comfortable with this
phrasing in terms of diagrams, as this is how many many things are phrased in algebra/category
theory/algebraic geometry. To speak precisely, if φ : G → A is any homomorphism, then we can
replace A by the image of φ (which is a subgroup, so still abelian), to get that φ : G → A is
surjective. Then its kernel is a normal subgroup. So by part (4), we have that G′ ≤ ker φ.

Now by the proof of part (4), we get a map first G → G/G′ and then a map G/G′ → (G/G′ )/(H/G′ ) ∼
=
G/H. This completes the diagram. □

Passing to the quotient by the commutator subgroup of G collapses all commutators to the identity,
so it is not surprising that all elements in the quotient group commute. The converse also holds by
(4) above: a quotient of G by H is abelian if and only if H contains the commutator subgroup.

Note that G is abelian if and only if G′ = 1.

Example 183. (1) G = Sn . Claim: [G, G] = An .

[x, y] = x−1 y −1 xy ∈ An since it has even parity. So [G, G] ≤ An .

85
 Also, (12)−1 (13)−1 (12)(13) = (123) ∈ [G, G]. And [G, G] ⊴ G so [G, G] contains all
3-cycles (all conjugates of (123)).
Thus, An ≤ [G, G] ⇒ [G, G] = An .
(2) (n ≥ 5) [An , An ] = An .
{1} ⪇ [An , An ] ⊴ An and since An is simple, [An , An ] = An .
(3) [A4 , A4 ] = V4 = {1, (12)(34), (13)(24), (14)(23)}.
(4) [A3 , A3 ] = {1} since A3 is abelian.

Proposition 184. Let H and K be subgroups of G. The number of distinct ways of writing each
element of the set HK in the form hk for some h ∈ H and k ∈ K is |H ∩ K|.

In particular, if |H ∩ K| = 1 then each element of HK can be written uniquely as a product hk.

By Proposition 94, we know that |HK| = |H||K|/|H ∩ K|. This is just all the possible pairs of
elements from h and elements from k, divided by the number of times we have overcounted (the
number of ways you can express the same element as a such a product).

Theorem 185 (Direct Product Recognition Theorem). Suppose that G is a group with subgroups
H and K such that (1) H, K ⊴ G and (2) |H ∩ K| = 1. Then HK ∼
= H × K.
In particular, if the above is satisfied and also |H||K| = |G|, then HK = G and so G ∼
= H × K.

Proof. Since H, K ⊴ G, HK is a subgroup of G. Let h ∈ H and k ∈ K. Since H ⊴ G, k −1 hk ∈ H

so h−1 k −1 hk ∈ H. Similarly, (h−1 k −1 h)k ∈ k. Since |H ∩ K| = 1, it follows that h−1 k −1 hk = 1 so
hk = kh and each element of H commutes with each element of K.

By the preceding proposition, each element of HK can be written uniquely as a product hk. So
the map
φ : HK → H × K
hk 7→ (h, k) is well-defined (many of you did not check this on the first exam). It is a homomorphism
and clearly bijective so HK ∼ = H × K. □

We can in fact extend this theorem inductively to:

86
Theorem 186. Suppose G is a group with normal subgroups H1 , . . . , Hn such that

(1) H1 H2 · · · Hn = {h1 h2 · · · hn | hi ∈ Hi } = G
(2) for all 1 ≤ i ≤ n − 1, H1 H2 . . . Hi ∩ Hi+1 = {e}

then G ∼
= H1 × · · · × Hn .

This is sometimes useful on a qualifying exam if you want to recognize a group as a direct product
of three of its subgroups.

Example 187. If |G| = 35, then by Example 160, we have P ⊴ G and Q ⊴ G where |P | = 5
and |Q| = 7. By the direct product recognition theorem, we therefore have G ∼
= P ×Q ∼
= Z5 ×Z7 .

We can say something a bit stronger.

Corollary 188. Let G be a finite group such that for each prime p | G, the Sylow p-subgroup is
normal. Say |G| = pe1 . . . pem and Pi ⊴ G with |Pi | = pei . Then G ∼
1 m = P1 × · · · × Pm .
i

Proof. First note that for all 1 ≤ i ≤ m, we have P1 · · · Pi−1 ∩ Pi = {1} since they have relatively
prime orders.

Further,
|P1 | · · · |Pi |
|P1 · · · Pi | = = pe11 · · · pei i .
|P1 ∩ · · · ∩ Pi |
Hence, P1 · · · Pm = G so by the theorem, G ∼
= P1 × · · · × Pm . □

Semidirect Products.
We now develop the semidirect product of two groups H and K. When we studied the direct
product, both H and K were (isomorphic to) normal subgroups H ×K. For the semidirect product,
we will build a larger group from H and K but only one of them will be normal in G. We will also
be able to build nonabelian groups even if H and K are abelian.

To motivate the construction of the semidirect product, remember that in the last section, we had
a group G containing normal subgroups H and K such that |H ∩ K| = 1, then HK was isomorphic
to H × K. Now suppose that we have a group G containing subgroups H and K such that H ⊴ G
but K is not necessarily normal in G, but |H ∩ K| = 1.
87
It is still true that HK is a subgroup of G (that only required one of the subgroups to be in
contained in the normalizer of the other). And we still have Proposition 184: every element of HK
can be written uniquely as a product hk for some h ∈ H and k ∈ K. Hence, there is a bijection
between HK and the collection of ordered pairs (h, k). Given two elements h1 k1 and h2 k2 of HK,
we first see how to write their product (in G) in the same form:

(h1 k1 )(h2 k2 ) = h1 k1 h2 (k1−1 k1 )k2 = h1 (k1 h2 k1−1 )k1 k2 = h3 k3

where h3 = h1 (k1 h2 k1−1 ) and k3 = k1 k2 .

We started by assuming that there exists a group G containing H and K such that H ⊴ G and
|H ∩ K| = 1. The basic idea of the semidirect product is to turn this construction around. Start
with groups H and K and construct a group containing isomorphic copies of H and K such that
the above hold.

Look back at our multiplication formula above. We had that k3 = k1 k2 so the multiplication for
the elements of K was straightforward. On the other hand, to get h3 , we had to take h1 (k1 h2 k1−1 ),
i.e. we had to first conjugate h2 by k1 . Recall now that since H ⊴ G, the group K acts on H by
conjugation.
k · h = khk −1
Hence, we can rewrite the multiplication formula as

(h1 k1 )(h2 k2 ) = (h1 [k1 · h2 ])(k1 k2 ).

The action of K on H by conjugation gives a homomorphism φ : K → Aut(H) so this multiplication

in HK depends only on the multiplication in H, in K, and the homomorphism K → Aut(H).

Theorem 189. Let H and K be groups and let φ : K → Aut(H). Let · denote the left action of
K on H determined by φ. Let G be the set of ordered pairs (h, k) with h ∈ H and k ∈ K. Define
the multiplication in G:
(h1 , k1 )(h2 , k2 ) = (h1 [k1 · h2 ], k1 k2 ).
Then

(1) The multiplication makes G into a group of order |H||K|.

e = {(h, 1) | h ∈ H} and K
(2) The sets H e = {(1, k) | k ∈ K} are subgroups of G isomorphic to
H and K.
(3) H
e ⊴ G.
e ∩K
(4) H e = 1.
(5) For all h ∈ H e khk −1 = k · h = φ(k)(h) (so in this construction, the φ : K →
e and k ∈ K,
Aut(H) becomes the action by conjugation).

88
Proof. The proof is not difficult, is in your book, and is more illuminating as an exercise than done
at the board. Check that the operation is associative with identity (1H , 1K ) and (h, k) has inverse
(φ(k −1 )(h−1 ), k −1 ).

Let’s just check that H

e ⊴ G.

(x, y)(h, e)(x, y)−1 = (x, φ(y)(h), y)(φ(y −1 )(x−1 ), y −1 ) = (xφ(y)(h)φ(y)(φ(y −1 )(x−1 )), 1) ∈ H.
e

Definition 190. Let H and K be groups and let φ : K → Aut(H) be a homomorphism. The
group described in the previous theorem is called the semidirect product of H and K with respect
to φ and is denoted H ⋊φ K. When φ is clear from context, this is denoted simply H ⋊ K. (The
direction of the triangle in ⋊ helps you remember that H ⊴ G).

The semidirect product generalizes the direct product.

Proposition 191. Let H and K be groups and let φ : K → Aut(H) be a homomorphism. Then
the following are equivalent

(1) the identity map between H ⋊ K and H × K is a group homomorphism (and hence an
isomorphism)
(2) φ is the trivial homomorphism K → Aut(H)
(3) K ⊴ H ⋊ K.

Proof. (1) ⇒ (2). By definition of the group operation in H ⋊ K

(h1 , k1 )(h2 , k2 ) = (h1 k1 · h2 , k1 k2 ).

By assumption (1), (h1 , k1 )(h2 , k2 ) = (h1 h2 , k1 k2 ) so k1 · h2 = h2 for all h2 ∈ H and all k1 ∈ K, so

K acts trivially on H. This is (2).

(2) ⇒ (3). If φ is trivial, then the action of K on H is trivial, so that the elements of H commute
with those in K. In particular, H normalizes K. Since K normalizes itself, G = HK normalizes
K so K ⊴ G.

(3) ⇒ (1). If K ⊴ H ⋊ K then for all h ∈ H and k ∈ K, [h, k] ∈ H ∩ K = {e}. Thus, hk = kh and
the action of K on H is trivial. The multiplication in the semidirect product is the same as that
in the direct product so (1) holds. □
89
Example 192. Let H be any abelian group and let K = ⟨x⟩ ∼ = Z2 be a group of order 2.
Define φ : K → Aut(H) by mapping x to the automorphism of inversion on H so x · h = h−1
for all h ∈ H. Then G contains the subgroup H of index 2 and

xh = h−1 x for all h ∈ H.

When H = Zn = ⟨a⟩, one recognizes G as D2n .

We have (a, 1)n = (1, 1), (1, x)2 = (1, 1) and (1, x)(a, 1) = (a−1 , x) = (a−1 , 1)(1, x).

Now define
φ : D2n = ⟨r, s | rn = s2 = 1, rs = sr = r−1 s⟩ → G = Zn ⋊ Z2
by r 7→ (a, 1) and s 7→ (1, x).

We checked that the relations go to (1, 1) so φ is defined. Also, it is surjective since the
generators of G are in the image of φ. Both sides have order 2n, so φ is an isomorphism.

90
 14. Wednesday 10/19: Semidirect Product Recognition, p-Groups

We didn’t quite finish semidirect products last time. One question is: how do we recognize semidi-
rect products?

Theorem 193 (Semidirect Product Recognition Theorem). Suppose G is a group with subgroups
H and K such that (1) H ⊴ G and (2) |H ∩ K| = 1.

Let φ : K → Aut(H) be the homomorphism defined by mapping k ∈ K to the automorphism of

left conjugation by k on H. Then HK ∼
= H ⋊φ K.
In particular, if G = HK with H and K satisfying (1) and (2), then G is isomorphic to some
semidirect product H ⋊ K.

Definition 194. Let H be a subgroup of the group G. A subgroup K of G is called a complement

for H in G if G = HK and |H ∩ K| = 1.

Example 195. Let |G| = pq for p, q prime, and p < q. Recall that we already know some
things about G from Example 160.

Let Q be a Sylow q-subgroup. We have already seen Q ⊴ G. Let P be a Sylow p-subgroup.

We have seen either p ∤ q − 1 and P ⊴ G or p | q − 1.

If P ⊴ G then by the direct product recognition theorem, G ∼

=P ×Q∼
= Zp × Zq ∼
= Zpq .
Otherwise, p | q − 1, P ≤ G, Q ⊴ G, P ∩ Q = {e}. Further,
|P ||Q|
|P Q| = = pq ⇒ P Q = G.
|P ∩ Q|

So G ∼
= Q ⋊φ P for some φ : P → Aut(Q).
We just need to classify the possible φ. Write P ∼
= Zp = ⟨x⟩. If φ : Zp → Aut(Zq ) = Zq−1 , then
where x 7→ a and a must have order in (Zq ) dividing p. But since (Zq )× = Zq−1 is cyclic, it
×

has a unique subgroup of order p. Say ⟨y⟩ is the subgroup of order p in (Zq )× . So the possible
a’s are 1, y, y 2 , . . . , y p−1 .

Case 1: φ(a) = 1. Then G ∼

= P × Q.
Case 2: φ(x) = y i for 1 ≤ i ≤ p − 1. Then G ∼ = Q ⋊φ P . These are all isomorphic but
nonabelian. [You will prove they are isomorphic in a homework exercise.]

91
 Example 196. We finish our example classifying groups of order 12 from Example 164. We
have already seen that either G has a normal Sylow 3-subgroup or else G ∼
= A4 .
Let P be a Sylow 2-subgroup and Q a Sylow 3-subgroup. |P | = 4 and |Q| = 3. We are now in
the case where Q ⊴ G.

Clearly, P ∩ Q = {e}. Hence P Q = G. By the semidirect product recognition theorem, we

have G ∼= Q ⋊φ P ∼
= Z3 ⋊φ Z4 or Z3 ⋊φ (Z2 × Z2 ).

• Case 1: Z3 ⋊φ Z4 . φ : Z4 → Aut(Z3 ) ∼
= (Z3 )× ∼
= Z2 . There are two possible automor-
× ×
phisms φ1 (1) = 1 ∈ Z3 or φ2 (1) = −1 ∈ Z3 .
For φ1 , this gives Z3 × Z4 . For φ2 , Z3 ⋊φ Z4 is a new group.
• Case 2: Z3 ⋊φ (Z2 × Z2 ). φ : Z2 × Z2 → Aut(Z3 ) ∼ = Z× ∼
3 = Z2 . We can have φ((1, 0)) =
±1 and φ((0, 1)) = ±1. This yields four different homomorphisms.
For φ3 (1, 0) = φ3 (0, 1) = 1, this yields Z3 × Z2 × Z2 .
For φ4,5,6 , we get Z3 ⋊φ (Z2 × Z2 ) ∼
i = Z2 × D6 ∼= Z2 × S3 .

So G ∼
= A4 or Z3 × Z4 or Z3 ⋊φ Z4 or Z3 × Z2 × Z2 or Z2 × S3 .

Remark. A remark on automorphism groups. When doing problems involving semidirect products,
you often need to understand the automorphism group Aut(H).

By Proposition 155, Aut(Zn ) ∼

= (Zn )× . Hence, the size of Aut(Zn ) is φ(n), the Euler totient
function. Indeed, the way you usually work with Aut(Zn ) is just as the invertible elements in
Zn with the operation being multiplication. The explicit correspondence here is that every auto-
morphism Zn → Zn is determined by where it maps the generator 1. To be an automorphism, 1
must be mapped to a generator, which is the same thing as an invertible element of Zn . So each
automorphism in Aut(Zn ) can be described as ψa where ψa maps 1 to a, and a ∈ (Zn )× .

Now it is a fact that for a prime p,


Z α α−1
p −p if p is odd
Aut(Zpα ) ∼
=
Z × Z α−2 if p = 2.
2 2

Hence, you can understand some properties of Aut(Zpα ) as an abstract group even without working
with the explicit automorphisms.

It is also true that if n and m are relatively prime, then Aut(Zn × Zm ) ∼

= Aut(Zn ) × Aut(Zm ). So
you can sometimes break up a cyclic group to understand its automorphism group.

However, the above remarks don’t help you understand Aut(Zn × Zn ), for example. That is, what
happens in a direct product of cyclic groups whose orders are not relatively prime. The important
92
theorem to know here is that if p is prime, then Aut(Zp × Zp ) ∼ = GL2 (Fp ), the group of invertible
2 × 2 matrices with entries in the field Fp . This is because Zp × Zp is essentially the vector space
F2p .

It is easy to see that GL2 (Fp ) = (p2 − 1)(p2 − p). This is because you can have any nonzero vector
for the first column (there are p2 total vectors and one of them is the zero vector). Then for the
second column you can take any vector that is not a scalar multiple of your first column. Knowing
the order of the group can rule out possible semidirect products. Read the example about groups
of order p3 on page 183 of the textbook to see this in action.

p-Groups.
We now move on to the penultimate topic in group theory this semester: p-groups. Recall that a
p-group is just a group of order pi for a prime P . Sylow’s Theorem says that a finite group G has
lots of p-subgroups. It turns out that we know a lot about p-groups. This is useful, since it says
that every group has lots of subgroups that we know a lot about.

Definition 197. A maximal subgroup of a group G is a proper subgroup M ⪇ G such that there
are no subgroups H of G with M ⪇ H ⪇ G.

Example 198. In a finite group, every proper subgroup is contained in some maximal sub-
group.

Example 199. In an infinite group, this may or may not be true. For example pZ is a
maximal subgroup of Z (and every maximal subgroup of Z is contained in some/potentially
many subgroups pZ). On the other hand, Q has no maximal subgroups.

We now collect a bunch of properties of p-groups into one huge theorem.

Theorem 200. Let p be a prime and let P be a group of order pα with α ≥ 1. Then

(1) The center of P is nontrivial.

(2) If H is a nontrivial normal subgroup of P , then |H ∩ Z(P )| =
̸ 1. In particular, every normal
subgroup of order p is contained in the center.

93
 (3) If H ⊴ P , then H contains a subgroup of order pb that is normal in P for each divisor pb
of |H|. In particular, P has a normal subgroup of order pb for each 0 ≤ b ≤ α.
(4) If H ⪇ P then H ⪇ NP (H) (every proper subgroup of P is a proper subgroup of its
normalizer in P ).
(5) Every maximal subgroup of P is of index p and is normal in P .

Proof. (1) This was proved already using the class equation.
(2) This is a similar argument to the above using the class equation. If H ⊴ P , we proved that
H was a union of conjugacy classes. Each conjugacy class has order [P : CP (x)] so order
pi for some i. Since H contains at least one conjugacy class with 1 element (the identity),
and H ≡ 0 (mod p), there must be other classes with only one element. These are classes
of central elements in P . Hence H ∩ Z(P ) ̸= 1.
(3) The only part of this that does not follow from Sylow’s Theorem is the normality of the
subgroups. We prove by induction on α. If α = 1 then |P | = p, so the result is trivial.
Also, if H = 1, the result is trivial.
So we assume that α > 1 and H ̸= {e}. By the above, |H ∩ Z(P )| ̸= 1 so by Cauchy’s
Theorem, H ∩ Z(P ) contains an element x of order p. Then Z = ⟨x⟩ is a normal subgroup
(since it is contained in the center). Pass to the quotient group P = P/Z. This quotient
has order pα−1 and H ⊴ P since H ⊴ P . By induction, for every 1 ≤ b ≤ α − 1, H has a
subgroup K of order pb which is normal in P . Take the complete preimage K of K to get
a normal subgroup of P of order pb+1 .
(4) We prove by induction on |P |. If P is abelian then the result is trivial, for then NP (H) = P
for any H. We may therefore assume |P | > p.
Let H ⪇ P . Since all elements of Z(P ) commute with all elements of P , therefore Z(P )
normalizes every subgroup of P . By part (1), we have Z(P ) ̸= 1. If Z(P ) is not contained
in H, then H is properly contained in ⟨H, Z(P )⟩ which is contained in NP (H) so the result
holds. We may therefore assume that Z(P ) ≤ H. Consider the quotient P = P/Z(P ).
Since P has smaller order than P , by induction H is properly contained in NP (H). Again
use the fourth isomorphism theorem.
(5) Let M be a maximal subgroup of P . By definition, M ⪇ P so M ⪇ NP (M ). Hence,
NP (M ) = P and M ⊴ P .
The fourth isomorphism theorem shows that P/M is a p-group with no proper nontrivial
subgroups (because M is maximal). But by (3), every p-group contains normal subgroups
of ever order pb . Hence, P/M = Zp which means |M | = pα−1 . □

94
 15. Wednesday 10/26: Nilpotent Groups, Free Groups, Presentations

Definition 201. A group is nilpotent if it has a series

{e} = G0 ≤ G1 ≤ G2 ≤ · · · ≤ Gn−1 ≤ Gn = G

with Gi ⊴ G and Gi+1 /Gi ≤ Z(G/Gi ) for all 0 ≤ i ≤ n − 1. Equivalently, [Gi+1 , G] ≤ Gi .

Proof that two conditions equivalent. If Gi+1 /Gi ≤ Z(G/Gi ) then for all x ∈ Gi+1 and all y ∈ G,

xGi yGi = yGi xGi ⇔ [x, y] ∈ Gi for all x ∈ Gi+1 , y ∈ G ⇔ [Gi+1 , G] ≤ Gi . □

Example 202. Note that in the above definition, Gi ⊴ Gi+1 (since it is normal in the whole
group) and each Gi+1 /Gi is abelian (since it is contained in a center). Hence, every nilpotent
group is solvable.

cyclic ⊆ abelian ⊆ nilpotent ⊆ solvable ⊆ all groups

Definition 203. (Upper central series). Let G be a group. Define Z0 = {e}, Z1 = Z(G).

Let Z2 be the group such that Z2 /Z1 = Z(G/Z1 ). This exists by subgroup correspondence, since
Z(G/Z1 ) is a subgroup of G/Z1 , there is a Z2 ≥ Z1 corresponding to it.

Similarly, Zi+1 is the group such that Zi+1 /Zi = Z(G/Zi ).

Notice that since Z(G/Z1 ) ⊴ G/Z1 , we have Z2 /Z1 ⊴ G/Z1 so Z2 ⊴ G by subgroup correspondence.

Proposition 204. A group G is nilpotent if and only if Zn = G for some n (i.e., the upper central
series reaches the top).

Proof. [⇐]. If Zn = G for some n, then

{e} = Z0 ≤ Z1 ≤ · · · ≤ Zn = G.

Claim: This series satisfies the properties from the definition of nilpotent:

Zi ⊴ G (we proved this above), and Zi+1 /Zi ≤ Z(G/Zi ) by definition.

95
[⇒]. Conversely, if G is nilpotent, then

{e} = G0 ≤ G1 ≤ · · · ≤ Gm = G.

with Gi+1 /Gi ≤ Z(G/Gi ).

Claim: Gi ≤ Zi for all i.

Prove this claim by induction on i. i = 0 is clear.

For the inductive step, assume Gi ≤ Zi . As stated in the definition, we have [Gi+1 , G] ≤ Gi . This
means that [Gi+1 , G] ≤ Zi by the inductive hypothesis.

Hence, Gi+1 Zi /Zi ≤ Z(G/Zi ) so Gi+1 Zi ≤ Zi+1 which implies Gi+1 ≤ Zi+1 .

So Gm = G implies G = Gm ≤ Zm = G. □

Example 205. • S3 is not nilpotent since Z(S3 ) = {e}. However, the series

{e} ⊴ ⟨(1 2 3)⟩ ⊴ S3

shows that S3 is solvable.

• If |G| = pi for a prime p then G is nilpotent.

Proof. We proved that a p-group has nontrivial center (using the class equation).
At each stage of forming the upper central series Z0 ≤ Z1 ≤ . . . for G, Zi+1 /Zi is
the center of G/Zi , which is a p-group.
Thus, Zi+1 /Zi is not trivial which implies Zi ⪇ Zi+1 and since |G| < ∞, Gn = G for
some n. □

• If G and H are nilpotent then G × H is nilpotent.

Proof. Zn (G × H) = Zn (G) × Zn (H). □

Corollary 206. Suppose G is a finite group such that G ∼

= P1 × · · · × Pm for some pi -groups Pi .
Then G is nilpotent.

Lemma 207. If G is nilpotent and H ⪇ G, then H ⪇ NG (H). (In nilpotent groups, normalizers
grow).

96
Proof. Note that Z(G) ≤ NG (H) since the center is in every normalizer and H ≤ NG (H) since H
is always in its own normalizer. Hence, the group HZ(G) makes sense.

If H ⪇ HZ(G) ≤ NG (H), then we are done.

So assume that H = HZ(G) which implies Z(G) ≤ H.

I claim now that Z2 ≤ NG (H). If z ∈ Z2 then for h ∈ H, we want to show that zhz −1 ∈ H.

But zhz −1 h−1 = [z, h] ∈ [Z2 , H] ≤ [Z2 , G] ≤ Z1 ≤ H.

Hence, zhz −1 ∈ H. So Z2 ≤ NG (H) and HZ2 makes sense. If H ⪇ HZ2 , then H ⪇ HZ2 ≤ NG (H)
so we are done. Otherwise, HZ2 = H so Z2 ≤ H.

Using induction, we obtain H ⪇ NG (H). Otherwise, Zm = G ≤ H, a contradiction. □

Theorem 208 (Characterization of finite nilpotent groups). Let |G| < ∞. Let p1 , . . . , ps be the
distinct primes dividing its order and let Pi ∈ Sylpi (G). The following are equivalent:

(1) G is nilpotent.
(2) H ⪇ G ⇒ H ⪇ NG (H).
(3) Pi ⊴ G for each i.
(4) G ∼
= P1 × · · · × Ps .

Proof. We have already showed (4) ⇒ (1) and (1) ⇒ (2).

In the direct product recognition section, we showed that if every Sylow subgroup is normal then
G is a direct product of Sylow subgroups so (3) ⇒ (4).

It remains to show that (2) ⇒ (3). Let P = Pi for some i. Since P ⊴ NG (P ), a corollary to
Sylow’s Theorem says P char NG (P ) ⊴ NG (NG (P )) so P ⊴ NG (NG (P )) and since NG (P ) is the
largest subgroup of G in which P is normal, we must have NG (NG (P )) ≤ NG (P ). This means
NG (P ) = NG (NG (P )). By (2), we must have NG (P ) = G so P ⊴ G. □

Remark. This partially proves FTFGAG for finite abelian groups (as opposed to finitely gen-
erated). An abelian group is automatically nilpotent, so is the direct product of its Sylow p-
subgroups. It turns out that for a finite abelian group, the Sylow p-subgroup just consists of
the terms Zpn1 × Zpn2 × · · · × Zpnk in the elementary divisor decomposition. We will fully prove
FTFGAG for finitely generated but not-necessarily-finite abelian groups later, when we learn the
Chinese Remainder Theorem in ring theory.

Remark. Also note that this shows that nilpotent groups are quite nice: they are precisely the
groups that are products of their Sylow p-subgroups. Note that the Sylow p-subgroups themselves
can be quite complicated. Even though p-groups are also nice (they are nilpotent, so “close to”
97
abelian), there are a very wide variety of them. Indeed, almost all groups have order 2n for some
n. This paper: https://www.ams.org/journals/era/2001-07-01/S1079-6762-01-00087-7/
S1079-6762-01-00087-7.pdf shows that of the groups of order ≤ 2000, over 99% of them have
order 1024.

Definition 209 (Definition of the lower central series). Let G0 = G, G1 = [G, G], G2 = [G1 , G],
. . . Gi+1 = [Gi , G].

The lower central series is G0 ≥ G1 ≥ . . . .

Theorem 210. G is nilpotent if and only if Gn = {e} for some n.

Proof. Assume G nilpotent and let {e} = H0 ≤ H1 ≤ · · · ≤ Hm = G be a central series. That is,
Hi ⊴ G and [Hi+1 , G] ≤ Hi .

Claim: Gi ≤ Hm−i for all i.

Base case: i = 0, G = G0 ≤ Hm = G.

Inductive step: Gi+1 = [G, Gi ] ≤ [G, Hm−i ] ≤ Hm−i−1 = Hm−(i+1) .

So the claim is true for all i. When i = m, Gm ≤ H0 = {e}.

Conversely, if Gm = {e} for some m, then {1} = Gm ≤ Gm−1 ≤ · · · ≤ G0 = G is a central series,

Gi ⊴ G and [Gi , G] ≤ Gi+1 . □

Lemma 211. If G is nilpotent, then all subgroups and factor groups of G are nilpotent.

Proof. Problem set. □

Example 212. If N ⊴ G and G/N and N are nilpotent, then G need not be nilpotent. Take
G = S3 , N = ⟨(1 2 3)⟩.

Free groups.
Next, we discuss free groups. This will enable us to make the idea of generators and relations that
we discussed back around Example 16 precise.
98
Let S be any set. The idea of a free group F (S) generated by S is that it is the group which
contains S and there are no relations among any of the elements in S (so S is “free” of relations).

Example 213. Let S = {x, y}. The free group F (S) on the two generators x and y consists
of elements which are words in x and y (of finite length), together with their inverses e.g.:

e, x, y, xx = x2 , xy, yx, y 2 , x3 , x2 y, xyx, xy 2 , yx2 , yxy, y 2 x, y 3 , ...

are words, but also

x−1 , y −1 , x−2 , y −1 x−1 , x−1 y −1 , y −2 , ...
The multiplication in this group is simply given by concatenation of words.

It is a technically annoying task to prove that concatenation of words is a well-defined associative

operation on F (S). This is done carefully in your book in section 6.3, and you should read it.

The most important property of the free group F (S) is (because there are no relations satisfied
by the generators) is that if G is a group, then any map S → G can be extended uniquely to a
homomorphism F (S) → G. This is because a map S → G specifies where the generators of F (S)
must go in G, and if F (S) → G is a homomorphism, this specifies where any product of generators
go. The fact that there are no relations to worry about means that we can send the generators
wherever we wish.

The important property in the previous paragraph is referred to as the universal property of the
free group. In fact, really F (S) can be defined as the group that has this property. This is the
modern algebraic way to define F (S), in terms of its universal property, and we will see examples
of more of this kind of thinking later (when we cover localizations or perhaps tensor products in
6102).

99
 16. Monday 10/31: Free Groups, Introduction to Rings

Last time we constructed (modulo showing the operation was well-defined and associative) the free
group F (S) on the set S. We also stated that F (S) has a nice property: if G is a group, then any
map S → G can be extended uniquely to a homomorphism F (S) → G. The universal property
of F (S) is often stated in the language of commutative diagrams. This is how we state it in the
theorem below.

Theorem 214. Let G be a group, S a set, and φ : S → G a set map. Then there is a unique
group homomorphism Φ : F (S) → G such that the following diagram commutes:
inclusion
S / F (S)

Φ
φ
! 
G
That is, Φ|S = φ.

Proof. Every element of F (S) is a finite product of the form

sϵ11 sϵ22 · · · sϵnn

where the si are elements of S and ϵi ∈ {±1}. We simply define

Φ(sϵ11 sϵ22 · · · sϵnn ) = φ(s1 )ϵ1 φ(s2 )ϵ2 · · · φ(sn )ϵn .

This map is clearly a homomorphism, as concatenation on the left-hand side of the equation cor-
responds to multiplication on the right-hand side. It is also unique, as any homomorphism which
restricts to φ on S must clearly map s1ϵ1 sϵ22 · · · sϵnn to the defined element. □

A typical argument in modern algebra is to show that an object is unique by using its universal
property. The corollary below shows that F (S) is unique, so it makes sense to talk about the free
group on a set of generators.

Corollary 215. F (S) is unique up to unique isomorphism which is the identity on the subset S.

Proof. Suppose F (S) and F ′ (S) are two free groups genrated by S. Since S is contained in both
F (S) and F ′ (S), we have injections S ,→ F ′ (S) and S ,→ F (S). By the universal property of the free
group, we have unique associated group homomorphisms Φ : F (S) → F ′ (S) and Φ′ : F ′ (S) → F (S),
which are both the identity on S.
100
The composition Φ′ ◦ Φ is a homomorphism F (S) → F (S) which is the identity on S, so by the
uniqueness statement in the theorem, it must be the identity map. Similarly, Φ ◦ Φ′ is the identity,
so Φ is an isomorphism. □

Definition 216. Let S be a set. The free group on S is the group F (S) defined above. A group
F is called a free group if there exists some set S such that F = F (S). In this case, we call S a set
of free generators for F . The cardinality of S is called the rank of the free group.

Presentations.
Now we are set up to talk about presentations, which are really just certain quotients of free groups.
If G is any group, then G is the homomorphic image of some free group. We can at least take
S = G and take φ : G → G to be the identity map. Then Theorem 214 produces a surjective
homomorphism F (G) → G.

Indeed, if S is any subset of G such that G = ⟨S⟩, then we have a surjective homomorphism
F (S) → G which is the identity on S (by taking φ : S → G to just be inclusion). Notice actually
that a subset T generates G if and only if the map Φ : F (T ) → G extending the inclusion T → G
is a surjection.

Definition 217. Let S be a subset of a group such that G = ⟨S⟩.

(1) A presentation for G is a pair (S, R) where R is a set of words in F (S) such that the normal
closure 16 of ⟨R⟩ in F (S) equals the kernel of the homomorphism π : F (S) → G (where
π extends the inclusion S ,→ G). The elements of S are called the generators and the
elements of R are called relations of G.
(2) We say that G is finitely generated if there is a presentation (S, R) such that S is a finite
set. We say that G is finitely presented if both S and R are finite sets.

Note that if (S, R) is a presentation, then the kernel of the map F (S) → G is not ⟨R⟩ (the subgroup
generated by R), but in general can be much larger (the subgroup generated by R and all conjugates
of elements in R, so that the result is normal).

In the above definition, the relations R contains words that are trivial in G. If S = {s1 , . . . , sn }
and R = {w1 , . . . , wk }, then we write the presentation

G = ⟨s1 , s2 , . . . , sn | w1 = w2 = · · · = wk = 1⟩.

16The normal closure of a subset T ⊆ G is the smallest normal subgroup which contains T .
101
However, we often write things like w1 = w2 , by which we really mean w1 w2−1 = 1. Since this is
unambiguous, we will continue to do so.

Example 218. • Finite groups are finitely presented (you can take S = G and write
down the relations between any pair of the elements, in what is essentially a multipli-
cation table for G).
• Z∼= F ({x}) = ⟨x⟩,
• Z×Z∼ = ⟨x, y | [x, y] = 1⟩ = ⟨x, y | xy = yx⟩.
• Zn × Zm = ∼ ⟨x, y | xn = y m = [x, y] = 1⟩.

We have of course already seen that D2n = ⟨r, s | rn = s2 = 1, sr = r−1 s⟩.

We can now see why we can use presentations of a group to find homomorphisms from the group. For
simplicity, suppose G = ⟨x, y | r1 = r2 = · · · = rk = 1⟩. Earlier in this course, we claimed that if H
is another group and x′ , y ′ are any elements satisfying the relations, then there is a homomorphism
G → H.

We have a presentation homomorphism π : F (x, y) → G. We can define π ′ : F (x, y) → H by

π ′ (x) = x′ and π ′ (y) = y ′ . Then ker π ≤ ker π ′ , so π ′ factors through ker π and we obtain a
homomorphism
G∼
= F (x, y)/ ker π → H.

Ring Theory.
We have now finished our study of groups, and move on to studying one of my favorite kinds of
mathematical structures: rings. (I am a ring theorist, kind of.)

Definition 219. A ring R is a set with binary operations + and · (called addition and multipli-
cation) such that

(1) (R, +) is an abelian group

[i.e., + is associative, there exists an identity 0 for +, for all a ∈ R, there exists an additive
inverse −a such that a + (−a) = (−a) + a = 0, and for all a, b ∈ R, a + b = b + a.].
(2) · is associative [a · (b · c) = (a · b) · c].
(3) There exists an identity 1 for multiplication [for all a ∈ R, a · 1 = 1 · a = a].
(4) The distributive laws hold [for all a, b, c ∈ R, a·(b+c) = a·b+a·c and (b+c)·a = b·a+c·a].

You often simply write ab rather than a · b. For addition, you always write +, the additive identity
is always called 0, and the additive inverse of a is always called −a.
102
Remark. Why require that the additive structure be abelian? It is actually forced by the distribu-
tive laws since
(1 + 1)(a + b) = a + b + a + b
by distributing the (1 + 1) over the (a + b) but also

(1 + 1)(a + b) = a + a + b + b.

Hence, a + b = b + a for all a, b ∈ R.

Remark. (1) If ab = ba for all a, b ∈ R, then R is a commutative ring. Otherwise the ring is
called noncommutative. (I am a noncommutative ring theorist.)
(2) Some authors leave out the axiom about the existence of a multiplicative identity 1. A ring
without 1 is called a non-unital ring or also sometimes a rng. Unlike the book, we will
always assume that rings are unital, unless I say otherwise.

Definition 220. If R is a ring such that for all a ∈ R, a ̸= 0, there exists a b ∈ R such that
ab = ba = 1, then R is a division ring or skew-field. A commutative division ring is a field.

Example 221. Z, Q, R, C with their usual operations are all commutative rings (the last three
are fields).

Example 222. The quotient group Z/nZ is not just a group (under addition) but is also a
ring under addition and multiplication. Again, Z/nZ is commutative. It is not only a quotient
group of Z but also a quotient ring of Z.

Example 223. Let R be a ring, n ≥ 1 an integer. The n×n matrices with entries in R, Mn (R)
form a ring.Let (aij ) be the matrix with entry aij in row i and column j.

(aij ) + (bij ) = (aij + bij )

n
X
(aij )(bij ) = (cij ) where cij = aik bkj
k=1
This is the usual matrix addition and multiplication. It is easy to check the axioms.
" #" # " #" #
0 1 0 0 0 0 0 1
If n ≥ 2, Mn (R) is noncommutative since ̸ = .
0 0 1 0 1 0 0 0

103
 Example 224. Note that GLn (R) is only a group and not a ring, since the group structure
was matrix multiplication. In particular, the group structure was nonabelian.

Definition 225. Let R be a ring, a ̸= 0, b ̸= 0 such that ab = 0. Then a and b are called
zero-divisors.

A ring R with no zero-divisors is called a domain.

A commutative domain is called an integral domain.

Example 226. Z, Q, R, C are integral domains.

Example 227. Z/nZ is a domain if and only if n is prime.

If n = pq where 0 < p < n, 0 < q < n, then 0 = n = pq = p · q, and p ̸= 0, q ̸= 0.

On the other hand, if p is prime, then if ab = 0, then p | ab so p | a or p | b. Hence, a = 0 or

b = 0. So Z/pZ is a domain.

Fact. Indeed, if p is a prime, then Z/pZ is a field. If a ̸= 0, then gcd(a, p) = 1, so use the Euclidean
algorithm to solve am + pn = 1. Then a · m = 1 so m = a−1 .

Example 228. If n ≥ 2, Mn (R) is not a domain since

" #" # " #
0 1 0 1 0 0
= .
0 0 0 0 0 0

You should be careful about cancelling in rings, due to the presence of zero-divisors. For example,
in the example above, if we let the first matrix be A and the second be B, we had AB = A0 but
B ̸= 0. This is in contrast to group theory, where we had left and right cancellation.

However, in integral domains, you can do this safely.

104
Proposition 229. Assume a, b, c ∈ R and with a not a zero divisor. If ab = ac then either a = 0
or b = c (i.e., if a ̸= 0, we can cancel the a’s). In particular, if R is an integral domain, then ab = ac
implies either a = 0 or b = c.

Proof. If ab = ac then a(b − c) = 0 so either a = 0 or b − c = 0. The second statement follows from

the first and noting that in an integral domain, there are no zero divisors. □

Corollary 230. A finite integral domain is a field.

Proof. Let R be a finite integral domain and choose 0 ̸= a ∈ R. By the cancellation law, the map
x 7→ ax is an injective function (since if ab = ac, then b = c). Since R is finite, any injection R → R
is a bijection. In particular, there is some b ∈ R such that ab = 1. so a is a unit. Since a was
arbitrary, R is a field. □

Definition 231. An element u ∈ R is called a unit if u has a multiplicative inverse. The set
of units in R is denoted R× . [In the past, we have sometimes called this R∗ . Both notational
conventions are used.]

Example 232. Since every unit has a multiplicative inverse, R× forms a group under mul-
tiplication. This is called the group of units and we have already seen one example in group
theory.
U (n) = {m ∈ Z/nZ | gcd(m, n) = 1} = Z/nZ× .

Example 233. (1) Q× = Q \ {0}

(2) Z× = {1, −1} ∼
= Z2
(3) Mn (R)× = GLn (R)

105
Example 234. A somewhat pathological example that we sometimes need to keep in mind is
the zero ring R = {0}. In this ring, 0 is the multiplicative identity, so 0 = 1. This is the only
ring in which the additive and multiplicative identities are the same.

106
 17. Wednesday 11/2: Basic Properties, Homorphisms

We have the following basic properties, which you should have proved in an undergraduate algebra
class. If not, you should be able to do them without much difficulty.

Proposition 235 (Basic properties in rings.). For all a, b ∈ R,

(1) 0 · a = 0 = a · 0,
(2) (−a)b = −(ab) = a(−b),
(3) (−a)(−b) = ab,
(4) a(−1) = −a = (−1)a,
(5) The multiplicative identity of R is unique.

Sample proof. 0a = (0 + 0)a = 0a + 0a ⇒ 0 = 0a. For the last claim, if e is another multiplicative
identity, then 1 = e1 = e. The remaining proofs are easy exercises. □

Remark. If R is a ring with 0 = 1, then 0 = 0a = 1a = a for all a ∈ R. Hence, R = {0}. So if you

rule out the trivial case, you may assume that 0 ̸= 1 in R.

Example 236. 2Z = {all even integers} is a non-unital ring (aka a rng) since 2b ̸= 2 for any
b ∈ 2Z.

Definition 237. A subset S of R is a subring if S is a subgroup of R under + (a, b ∈ S ⇒ a−b ∈ S)

and S is closed under multiplication (a, b ∈ S ⇒ ab ∈ S).

In other words, S is a subring of R if the operations of addition and multiplication in R when

restricted to S give S the structure of a ring.

Definition 238. S is a unital subring if 1R = 1S . Usually, we will assume that subrings are unital.

Example 239. (1) Z is a subring of Q is a subring of R is a subring of C.

(2) nZ is a non-unital subring of Z for any n ≥ 2.

107
 " # " #
∗ 0 1 0
(3) ⊆ M2 (R) is a non-unital subring since the identity element of S is but
0 0 0 0
" #
1 0
the identity element of R is .
0 1

We now discuss some fundamental examples of rings to keep in mind.

Definition 240. Let R be a any ring (not necessarily commutative). Define a ring R[x] to be the
polynomial ring over R.

• Elements of R[x] are formal polynomials a0 + a1 x + a2 x2 + · · · + an xn .

• A general element of R can be written i≥0 ai xi where ai = 0 for i ≫ 0 and
P

X X X
ai xi = bi xi = (ai + bi )xi
 
X  X  X X
ai xi bi xi =  aj bk  xi .
i≥0 j,k s.t. j+k=i

• Formally, we can identify a0 + a1 x + · · · + an xn with (a0 , a1 , · · · , an , 0, 0, · · · ) so elements of

R[x] correspond to infinite sequences of elements of R which are eventually 0.
• The degree of f ∈ R[x], f = i≥0 ai xi is the maximal n such that an ̸= 0. By convention,
P

deg 0 = −∞.
• The ring R appears as subring of R[x] as the constant polynomials.

Lemma 241. Let R be a domain and let p, q ∈ R[x]. Then

(1) deg pq = deg p + deg q,

(2) the units of R[x] are the units of R,
(3) R[x] is a domain.

Proof. If R has no zero divisors, then neither does R[x]. If p and q are polynomials with leading
terms an xn and bm xm then the leading term of pq is an bm xn+m and an bm ̸= 0 since R is a domain.
Hence, deg pq = deg p + deg q and R[x] is a domain. (Of course, if p or q is zero, then both sides of
the equality in (1) are −∞, by our convention.)

If p is a unit, then pq = 1 and deg p + deg q = 0. Hence, p and q are both elements of R and are
units in R. □
108
Conversely, if R has zero divisors, so does does R[x] since R ⊂ R[x]. Hence, in fact R is a domain
if and only if R[x] is a domain.

If S is a subring of R, then S[x] is a subring of R[x].

Example 242. Let D be any rational number that is not a perfect square in Q. Define
√ √
Q( D) = {a + b D | a, b ∈ Q}

as a subset of C. The set is closed under subtraction and

√ √ √
(a + b D)(c + d D) = ac + bdD + (ad + bc) D
√
so Q( D) is closed under multiplication and so is a subring of C.
√ √
As long as D is squarefree, every element of Q( D) can be written uniquely in the form a+b D.
√ √ √
Hence, as long as a and b are not both zero, a + b D ̸= 0. Further, (a + b D)(a − b D) =
a2 − b2 D so √
√ −1 a−b D
(a + b D) = 2
a − b2 D
√
so Q( D) is a field called a quadratic field.

√ √
Example 243. Let D be a squarefree integer and define Z[ D] = {a + b D | a, b ∈ Z}. This
√ √
is a subring of the quadratic field Q( D) called the ring of integers of Q( D).

Here is another new source of examples for rings (some of which are noncommutative). This is the
group ring construction. Just as the polynomial ring R[x] consisted of formal sums of elements
of R indexed by powers of x, so the group ring RG will consist of formal sums of elements of R
indexed by elements of the group G.

Definition 244. Let G be a group, and let R be any ring. The group ring
 
X 
RG = rg g | rg ∈ R with all but finitely many rg = 0
 
g∈G

with addition
X X X
rg g + sg g = (rg + sg )g.
g∈G g∈G g∈G

109
Distributivity forces
X  X  X X X X
rg g sg g = (rg g)(sh h) = (rg sh )gh = (rg sh )k.
g,h g,h k∈g gh∈G s.t. gh=k

With this addition and multiplication, RG forms a ring.

If G is a finite group, then we don’t need to worry about the “all but finitely many coefficients are
zero” part, so elements of RG are just formal sums indexed by group elements with coefficients in
the ring R.

Example 245. G = Z2 = {e, a}, R = R

RZ2 = R1 + Ra
√ √ √ √ √
(πe + 3a)(4e − 5 2a) = 4πe + 12a − 5 2πa − 15 2a2 = (4π − 15 2)e + (12 − 5 2)a
So the elements are indexed by group elements, and the multiplication uses both the multipli-
cation in G as well as in R.

The ring R appears in RG as the {re | r ∈ R}, the R-multiples of the identity e of G.

The group G also appears in RG as {1R g | g ∈ G}.

Example 246. You can think of the group ring construction as “almost” generalizing the
polynomial ring construction.

The polynomial ring R[x] is the “group ring” RN. The elements of R[x] are sums which are
indexed by powers of x, with coefficients in R. Multiplying powers of x just adds the coefficients.
The elements of RN are sums which are indexed by natural numbers with coefficients in R.

However N = {0, 1, 2, . . . } is not a group, since it doesn’t contain inverses. So RN is really just
a monoid ring.

Definition 247. Let R and S be rings. The direct product

R × S = {(r, s) | r ∈ R, s ∈ S}

with + and · defined coordinate-wise is again a ring. Also, can define the product of an arbitrary
collection of rings.

110
 Example 248. The multiplicative identity of R×S is (1R , 1S ) so R×{0} is a nonunital subring
of R × S.

Definition 249. Let R and S be rings. A function φ : R → S is a (ring) homomorphism if

φ(a + b) = φ(a) + φ(b)

and
φ(ab) = φ(a)φ(b)
for all a, b ∈ R.

The kernel of φ is
ker φ = {r ∈ R | φ(r) = 0}
and the image of φ is
im φ = φ(R).

If φ(1R ) = 1S then we call φ a unital homomorphism. We will assume all homomorphisms are
unital (unless stated otherwise).

Definition 250. Let R be a ring. An additive subgroup I ⊆ R is called

(1) a left ideal if rx ∈ I for all r ∈ R and all x ∈ I.

(2) a right ideal if xr ∈ I for all r ∈ R and all x ∈ I.
(3) an ideal if rx ∈ I and xr ∈ I for all r ∈ R and all x ∈ I.

If we want to be extra clear, we sometimes use the term two-sided ideal for an ideal.

In other words, a left ideal is closed under addition and subtraction, and closed under multiplication
by ring elements on the left. Of course, if R is a commutative ring, then xr = rx so a right ideal
is a left ideal is an ideal.

Example 251. In M2 (R), (" # )

α 0
I= : α, β ∈ R
β 0

111
is a left but not right ideal since
" #" # " #
a b α 0 aα + bβ 0
=
c d β 0 cα + dβ 0
and (" # )
α β
J= : α, β ∈ R
0 0
is a right but not a left ideal. In fact, M2 (R) has no ideals except {0} and M2 (R). (Q: What
should such a ring be called? A: A simple ring.)

112
 18. Monday 11/7: Ideals and Isomorphism Theorems

Definition 252. Let R be a ring, I an ideal. The quotient ring or factor ring, R/I, as a set is the
additive factor group R/I under +.

(r + I) + (s + I) = (r + s) + I

with product
(r + I)(s + I) = rs + I.

We should check to make sure the operations are well-defined in R/I. Addition is well-defined
because I is a subgroup in an abelian group (R, +), so we know that you can define addition on
cosets of I.

We just need to check multiplication. Let r +I = r′ +I and s+I = s′ +I. We need rs+I = r′ s′ +I.
This is equivalent to rs − r′ s′ ∈ I (by what we know about group cosets).

Since r + I = r′ + I, we know r − r′ ∈ I and similarly s − s′ ∈ I. But now

rs − r′ s′ = r(s − s′ ) + (r − r′ )s′ .

Since s − s′ ∈ I, therefore r(s − s′ ) ∈ I and similarly (r − r′ )s′ ∈ I so rs − r′ s′ ∈ I.

This shows that the factor ring R/I is well-defined. The additive identity is 0 + I = I and the
multiplicative identity is 1 + I.

Example 253. (1) R = Z. Then nZ ⊆ Z is an ideal and Z/nZ is a quotient ring.

(2) Let A = R[x].17 Let
I = {xf | f ∈ R} =: xA
Then I is an ideal. What is A/I?

17After R for “ring” and S since it comes next alphabetically, the next-most-popular letter to name a ring is A.
This is because “ring” in French is “anneau”. Also, many ring theorists are interested in algebras (over a field, see
https://en.wikipedia.org/wiki/Algebra_over_a_field), and so A could also stand for “algebra.” I’ll use A to not
get R confused with R.
113
Theorem 254 (First Isomorphism Theorem). If φ : R → S is a homomorphism, then ker φ is an
ideal of R, φ(R) is a subring of S, and
∼
=
R/ ker φ → φ(R)
r + ker φ 7→ φ(r).

Example 255. Let

φ : R[x] → R
where f (x) = a0 + a1 x + · · · + an xn 7→ a0 (or, stated another way f 7→ f (0)). Check that this
is a homomorphism and that ker φ = xA (since xA is exactly the set of polynomials with 0
constant term). This shows that A/I ∼
= R.

Theorem 256. If I is any ideal of R, then the map

π : R → R/I
r 7→ r + I

is a surjective ring homomorphism with ker π = I (this is called the natural projection).

Combining this result with the first part of the first isomorphism theorem, we see that every ideal
is the kernel of a ring homomorphism and conversely the kernel of any ring homomorphism is an
ideal.

Example 257. (1) R = Z. Then nZ ⊆ Z is an ideal and Z/nZ is a quotient ring.

(2) Let R = R[x]. Let
I = {xf | f ∈ R} =: xR
Then I is an ideal.
Let
φ : R[x] → R
where f (x) = a0 +a1 x+· · ·+an xn 7→ a0 (or, stated another way f 7→ f (0)). Check that
this is a homomorphism and that ker φ = xR (since xR is exactly the set of polynomials
with 0 constant term). This shows that R/I ∼
= R.

114
Theorem 258. Let R be a ring.

(1) (Second Isomorphism Theorem for Rings) Let A be a subring and let I be an ideal of
R. Then A + I = {a + b | a ∈ A, b ∈ I} is a subring of R, A ∩ I is an ideal of A and
∼ A/(A ∩ I).
(A + I)/I =
(2) (Third Isomorphism Theorem for Rings) Let I and J be ideals of R with I ⊆ J. Then J/I
is an ideal of R/I and (R/I)/(J/I) ∼
= R/J.
(3) (Fourth Isomorphism Theorem for Rings) Let I be an ideal of R. There is a bijection
between subrings of R/I and subrings A of R containing I. Furthermore, A is an ideal of
R if and only if A/I is an ideal of R/I.

Proof idea. You prove the theorems in the following way. Every ring is an additive group under
+ (with the additional operation of ring multiplication). Do exactly the same proof from group
theory using the abelian group structure of (R, +), but also check that the group homomorphism
is a multiplicative map, so defines a ring homomorphism. □

Definition 259. Let I and J be ideals of R.

(1) The sum of I and J is I + J = {a + b | a ∈ I, b ∈ J}. This is an ideal of R.18

(2) The product of I and J is

IJ = {all finite sums of elements of the form ab with a ∈ I and b ∈ J}.

This is an ideal of R.
Warning: You should be careful with this! IJ ̸= {ab | a ∈ I, b ∈ J}. In general, this is not
an ideal. If you have a1 b1 and a2 b2 where each ai ∈ I and bi ∈ J, it is not always possible
to write a1 b1 + a2 b2 as the product of something in I with something in J. We need to take
all finite sums of such products to make sure that IJ is a subgroup under addition.
(3) For any n ≥ 1, define the nth power of I denoted I n to be the set consisting of all finite
sums of elements of the form a1 a2 . . . an where ai ∈ I.
This is just the same thing as the product II · · · I, so is an ideal.

18From the perspective of R as an abelian group under addition, this is the same thing as the HK we discussed in
Definition 93. The group structure on R is additive, so we write I + J here. This is a subgroup since every subgroup
of an abelian group is normal, and when one of H or K was normal, HK was a subgroup.
115
Lemma 260. Let R be a ring and let I be an ideal of R. Then I = R if and only if 1 ∈ I if and
only if I contains a unit.

Proof. If I = R then 1 ∈ R so 1 ∈ I. Conversely, if 1 ∈ I, then since I is an ideal, r · 1 = r ∈ I for

each r ∈ R.

For the second if and only if, certainly if 1 ∈ I, then I contains the unit 1. Conversely, if u ∈ I is
a unit, then u−1 u = 1 ∈ I, so 1 ∈ I. □

Lemma 261. A commutative ring R is a field if and only if the only ideals of R are {0} and R.

Proof. If R is a field and 0 ̸= I, pick x ∈ I with x ̸= 0. Since R is a field, x has an inverse x−1 . So
xx−1 = 1 ∈ I so I = R.

Conversely, if {0} and R are the only ideals in R, pick x ̸= 0 in R. Consider the ideal

xR = {xy | y ∈ R}.

Then xR = R which implies that there exists y ∈ R such that xy = 1. Hence, x is a unit so R is a
field. □

Remark. The commutative hypothesis is important here. A noncommutative ring R might have
only the ideals {0} and R, but not be a division ring. Recall that we claimed that M2 (R) was a
simple ring, but it’s certainly not a division ring, since it contains zero divisors.

Corollary 262. If R is a field then any nonzero ring homomorphism from R into another ring is
an injection.

Proof. The kernel of a ring homomorphism is an ideal. The kernel of a nonzero homomorphism is
a proper ideal so is (0) by the above. □

Definition 263. An ideal I of R is maximal if there is no ideal J of R such that I ⊊ J ⊊ R.

116
Definition 264. Let R be a commutative ring. An ideal I is prime if xy ∈ I implies x ∈ I or
y ∈ I.

Theorem 265. Let I be an ideal of a commutative ring R.

(1) I is maximal if and only if R/I is a field.

(2) I is prime if and only if R/I is a domain.

Proof. (1) By ideal correspondence, I is maximal in R if and only if (0) and R/I are the only
ideals of R/I if and only if R/I is a field by the previous result.
(2) R/I is a domain if and only if (x + I)(y + I) = 0 implies x + I = 0 + I or y + I = 0 + I.
This is the same as saying R/I is a domain if and only if xy ∈ I implies x ∈ I or y ∈ I,
which is the definition of I being prime. □

Corollary 266. In a commutative ring R, every maximal ideal is prime.

Proof. Since every field is a domain, therefore by the above theorem, every maximal ideal is prime.
□

Definition 267. Let R be a ring and let A be any subset of R.

(1) Let (A) denote the smallest ideal of R containing A, called the ideal generated by A. (A) is
also the intersection of all ideals containing A.
(2) Let RA denote the set of all finite sums of elements of the form ra with r ∈ R and a ∈ A.
Similarly, AR is the set of all finite sums of elements of the form ar.
(3) An ideal generated by a single element is called a principal ideal. Instead of writing ({a}),
we write just (a).
(4) An ideal generated by a finite set is called a finitely generated ideal. For an ideal generated
by {a1 , . . . , an } we write (a1 , . . . , an ).

Example 268. (1) In Z, (0) is prime since Z is a domain.

117
(2) If p is prime then (p) is prime since Z/pZ is a domain. In fact, since Z/pZ is a field,
actually (p) is maximal.
(3) In Z, the ideal (8, 12) = (4). Indeed, it turns out that every ideal of Z is actually
principle.
(4) In R[x], for a ∈ R, (x − a) is maximal (and thus prime).

φ : R[x] → R

defined by f (x) 7→ f (a) is a homomorphism and ker φ = (x−a) since φ(f (x)(x−a)) = 0.
(x − a) = {f (x − a) | f ∈ R[x]}.
Thus, the first isomorphism theorem says that R[x]/(x − a) ∼
= R. Thus, (x − a) is
maximal.
(5) In R[x, y], the ideal (x) is a prime ideal which is not maximal, since it is contained in
the maximal ideal (x, y).

118
 19. Wednesday 11/9: Zorn’s Lemma

Note. You originally had an exam scheduled for Monday 11/14. This was rescheduled to Friday
11/18 at 2:20-3:35pm in Phillips 730.

Our next goal is to prove the following theorem:

Theorem 269. Let R be a commutative ring with 1. Then given any proper ideal I of R, there is
a maximal ideal M of R containing I.

In order to prove this, we need to introduce an important result called Zorn’s Lemma. This is not
done in detail in your textbook, but we will do it here in careful detail, since using Zorn’s Lemma
is actually a very important technique in mathematics.

Definition 270. A poset (partially ordered set) is a set P with a binary relation ≤ such that

(1) x ≤ x for all x ∈ P

(2) x ≤ y, y ≤ z implies x ≤ z
(3) x ≤ y and y ≤ x implies x = y.

Example 271. (1) R with ≤ is a poset.

(2) S any set and P = P(S) the power set of S (the set of all subsets of S).
P is a poset where X ≤ Y means ⊆ Y . Notice that not every pair of elements is
comparable. Let S = {1, 2, 3}. Then {1, 2} ̸⊆ {1, 3} and {1, 3} ̸⊆ {1, 2}. This is why
the set is only partially ordered.
(3) R a ring. The set of all ideals of R with ≤ meaning inclusion is a poset. That is, I ≤ J
means I ⊆ J.

Definition 272. Let P be a poset, and B ⊆ P a subset. An upper bound for B is an element
x ∈ P such that b ≤ x for all b ∈ B.

Definition 273. Let P be a poset. Then x ∈ P is a maximal element in P if for all y ∈ P , y ≥ x

implies y = x.

119
 Example 274. P = {∅, {1}, {2}, {3}, {1, 2}, {1, 3}} has 2 maximal elements (not all elements
are comparable).

There is no upper bound for all of P . However, letting B = {∅, {1}, {2}}, the element {1, 2} is
an upper bound for B.

Definition 275. P a poset, B ⊆ P is a chain 19 if any two elements of B are comparable. That is,
if x, y ∈ B then either x ≤ y or y ≤ x. Alternatively, B is called a totally ordered subset.

Lemma 276 (Zorn’s Lemma). Let P be a nonempty poset. Suppose that every chain B in P has
an upper bound in P . Then P has a maximal element.

We will not prove this, but it is equivalent to the Axiom of Choice, which is equivalent to the
Well-Ordering Theorem. So if you accept the Axiom of Choice, then Zorn’s Lemma is true.

Remark. There is a classic math joke by mathematician Jerry Bona: “The Axiom of Choice is
obviously true, the well-ordering principle obviously false, and who can tell about Zorn’s lemma?”

Jerry Bona’s Wikipedia page explains the joke:

This is a joke: although the three are all mathematically equivalent, many mathematicians find the
axiom of choice to be intuitive, the well-ordering principle to be counterintuitive, and Zorn’s lemma
to be too complex for any intuition.

Recall that our goal was to prove this statement.

Theorem 277. Let R be a commutative ring with 1. Then given any proper ideal I of R, there is
a maximal ideal M of R containing I.

19A math song that went somewhat viral when I was in high school/college was the song “Finite Simple Group (of
Order Two)” by The Klein Four, a barbershop quintet of Northwestern mathematics PhD students. Watch it here:
https://www.youtube.com/watch?v=BipvGD-LCjU. It contains lots of mathematical terms in a love song, including
“You’re the upper bound in the chains of my heart, you’re my axiom of choice you know it’s true.” As you take more
and more math, you understand more and more of the references. Every math graduate student should really be able
to sing along, so part of your homework is to listen to it. I just discovered that you can even get sheet music for it
from Matt Salomone (the soloist) http://matthematics.com/fsg.pdf.
120
Proof. Let P be the poset of proper ideals of R which contain I with ≤ being inclusion. I ∈ P so
P is nonempty. If we have the hypotheses of Zorn’s Lemma, then P has a maximal element M and
M is a maximal ideal of R containing I.

So we need to prove that every chain has an upper bound. Let {Iα }α∈A be a chain in P . That is,
I ⊆ Iα ⊊ P and α, β ∈ A implies Iα ⊆ Iβ or Iβ ⊆ Iα .
S
Take J = α∈A Iα . We claim that J is an ideal.

• Suppose x, y ∈ J. Then there exist some α, β ∈ A such that x ∈ Iα and y ∈ Iβ . But

either Iα ⊆ Iβ or Iβ ⊆ Iα so exists some γ ∈ A such that x, y ∈ Iγ . Since Iγ is an ideal,
x − y ∈ Iγ ⊆ J.
• Now suppose x ∈ J and r ∈ R. Then x ∈ Iα for some α ∈ A. Since Iα is an ideal,
rx ∈ Iα ⊆ J. So rx ∈ J.

Hence, J is an ideal. And J is proper. If not, then 1 ∈ J, but then 1 ∈ Iα for some α in which case
Iα = R. But this is a contradiction, since each Iα is proper. Thus J is the upper bound for the
chain {Iα }.

Hence, each chain in P has an upper bound in P . By Zorn’s Lemma, every proper ideal is contained
in a maximal ideal. □

This proof is so iconic, there’s even a comic about it from https://www.abstrusegoose.com

20

Localization.
Our next topic in ring theory is localization. We cover this in slightly more generality than in your
textbook, where it is done in a special case in section 7.5 as “Rings of Fractions” but then only
properly in section 15.4 as “Localization”. We take a more balanced approach, defining localizations
in full generality without going as in depth as in 15.4.

20A song, a joke, and a comic, today. What incredible cultural references.
121
Motivating Intuition: Some of the first rings we learn about as children are the integers Z and
the rationals Q. These are quite different rings, since Q is a field, while Z is just an integral domain.
Nevertheless, they are very related. In particular, Z is a subring of Q, and the normal way to define
Q is
Q = {a/b | a, b ∈ Z, b ̸= 0}/ ∼
where ∼ is the usual equivalence relation, since the same element in Q can be written in more than
one way as a quotient of two integers.

The integers Z sit inside of Q as a subring, since we can identify each integer n with the fraction
n/1. So what we’ve done is included Z in a bigger ring where every nonzero element actually has
an inverse. The idea of localization is to extend this to other rings.

Slight generalization: Let R be any integral domain. There exists a field F with R ⊆ F , the
field of fractions of R where
F = {ab−1 | a, b ∈ R, b ̸= 0}/ ∼ .
Every nonzero element of R becomes a unit in F .

The general aim: We will consider a more general situation. In a general commutative ring R
(not necessarily a domain), we will pick some subset S that we want to invert. Then we will hope
to construct a ring T which contains R where every element of S is invertible. In general, we won’t
be able to achieve all parts of this goal, but we will see why.

Definition 278. Let R be a ring. A subset S of R \ {0} is called a multiplicative system if 1 ∈ S

and if x, y ∈ S then xy ∈ S.

We want: A ring T with R ⊆ T such that for all s ∈ S, s is a unit in T . (For the field of fractions,
we took S = R \ {0}).

Remark: Suppose s ∈ S is a zero-divisor in R. If we want R ⊆ T , this is the same as wanting an

injective ring homomorphism φ : R → T . Since we want s to be a unit in T , we want φ(s) to be a
unit in T . Since s is a zero-divisor, there exists a nonzero x such that sx = 0. Hence, φ(s)φ(x) = 0
in T . But φ(s) is a unit so φ(x) = 0. Hence, φ cannot be an injection.

Thus, in general, we will give up on T containing R.

Definition 279. Let R be a commutative ring, and let S ⊆ R a multiplicative system. We define
T = RS −1 called the localization of R along S as follows:

122
Consider the set of ordered pairs {(r, s) | r ∈ R, s ∈ S} subject to the equivalence relation

(r1 , s1 ) ∼ (r2 , s2 ) if u(r1 s2 − r2 s1 ) = 0 for some u ∈ S.

Then T = RS −1 is the set of equivalence classes, where we write r/s for the equivalence class of
(r, s). The operations in T are
r1 r2 r1 s2 + r2 s1 r1 r2 r1 r2
+ = and · = .
s2 s2 s1 s2 s1 s2 s1 s2

Theorem 280. T = RS −1 is a commutative ring and there is a homomorphism

φ :R → RS −1
r 7→ r/1

with the following properties:

(1) For all s ∈ S, φ(s) is a unit in T .

(2) RS −1 has the following universal property. If ψ is any ring homomorphism ψ : R → U
such that ψ(s) is a unit in U for all s ∈ S, then there exists a unique homomorphism
θ : RS −1 → U such that the diagram commutes
φ
R / RS −1

θ
ψ " 
U
i.e. θ ◦ φ = ψ.

Remark. • The first remark is that part (2) of the theorem is called the universal property
of localizations. Universal properties are extremely important in category theory, and hence
in the language of modern mathematics. Indeed, one modern way to define the localization
would be to define it as the unique ring that has the property stated in (2). Our approach
is to construct explicitly what RS −1 is, and then show that it has property (2).
• We have ker φ = {x ∈ R | xs = 0 for some s ∈ S}.

Proof. When is x/1 = 0? If and only if there exists a s ∈ S such that s(x · 1 − 0 · 1) = 0
which happens if x · s = 0 for some s ∈ S. □

• In particular, if S consists of non-zero divisors only, then φ is injective.

123
 • We should check some things. Why is (r1 , s1 ) ∼ (r2 , s2 ) if u(r1 s2 − r2 s1 ) = 0 an equivalence
relation?
– Reflexivity and symmetry are clear.
?
– Transitivity: (r1 , s1 ) ∼ (r2 , s2 ) and (r2 , s2 ) ∼ (r3 , s3 ) ⇒ (r1 , s1 ) ∼ (r3 , s3 )

Proof. (r1 , s1 ) ∼ (r2 , s2 ) and (r2 , s2 ) ∼ (r3 , s3 ) means there exist u, v ∈ S such that
u(r1 s2 − r2 s1 ) = 0 = v(r2 s3 − r3 s2 ).
We need a w ∈ S such that w(r1 s3 − r3 s1 ) = 0. But

uvs3 (r1 s2 − r2 s1 ) = 0 = uvs1 (r2 s3 − r3 s2 ).

Hence uv(s3 r1 s2 − r2 s1 s3 + r2 s1 s3 − s1 s2 r3 ) = 0 so uvs2 (r1 s3 − s1 r3 ) = 0. So take

w = uvs2 ∈ S. □

• Since the elements of RS −1 were defined as equivalence classes, we should check that the
operations + and · that we claimed were ring operations are actually well-defined. Check
this for yourself. That is, if r1 /s1 = r1′ /s′1 and r2 /s2 = r2′ /s′2 , you must check that
r1 r2 r′ r′
+ = 1′ + 2′
s1 s2 s1 s2
in RS −1 , and similarly for multiplication.
• We should also check that these operations make RS −1 into a ring. Do this on your own,
too. 1/1 is the identity and 0/1 is the zero element.

As graduate students, I expect that you will spend time between now and our next lecture trying
to digest the definitions from this lecture and checking the above details. Next time we will spend
time talking about the universal property and look at some examples.

124
 20. Wednesday 11/16: More on Localization

Recall that we have been talking about localizations. As we are continuing our previous topic,
recall that all rings we consider are assumed to be commutative unital rings.

Theorem 281. T = RS −1 is a commutative ring and there is a homomorphism

φ :R → RS −1
r 7→ r/1

with the following properties:

(1) For all s ∈ S, φ(s) is a unit in T .

(2) RS −1 has the universal property. If ψ is any ring homomorphism ψ : R → U such that
ψ(s) is a unit in U for all s ∈ S, then there exists a unique homomorphism θ : RS −1 → U
such that the diagram commutes
φ
R / RS −1

θ
ψ " 
U
i.e. θ ◦ φ = ψ.

Proof sketch. • If s ∈ S then φ(s) = s/1 is a unit in RS −1 since 1/s is its inverse.
• Universal property: Let ψ : R → U with ψ(s) a unit for all s ∈ S. Define

θ :RS −1 → U
r/s 7→ ψ(r)ψ(s)−1 .

We need to check that θ is well-defined. (You should always be asking if the function you
wrote down is well-defined!)
If r1 /s1 = r2 /s2 then for some v ∈ S, v(r1 s2 − s1 r2 ) = 0. Then

ψ(v)[ψ(r1 )ψ(s2 ) − ψ(s1 )ψ(r2 )] = 0

and ψ(v) is a unit in U so ψ(r1 )ψ(s2 ) − ψ(s1 )ψ(r2 ) = 0 which implies ψ(r1 )ψ(s1 )−1 =
ψ(r2 )ψ(s2 )−1 so θ is well-defined.
• The diagram commutes since θ ◦ φ(r) = θ(r/1) = ψ(r)ψ(1) = ψ(r).
• For uniqueness of θ, we note that every element in RS −1 is of the form r/s = r/1·1/s. Since
θ is a ring homomorphism, θ(r/s) = θ(r/1)θ(1/s). As above, since the diagram commutes,
we must have θ(r/1) = ψ(r). We also must have θ(s/1) = ψ(s) so θ(1/s) = θ((s/1)−1 ) =
125
 θ(s/1)−1 = ψ(s)−1 . Hence, θ(r/s) = ψ(r)ψ(s)−1 is the unique homomorphism satisfying
the above properties. □

Example 282. (1) R a domain, S = R \ {0}, then RS −1 is the field of fractions of R.

(2) R a ring, 0 ̸= f ∈ R. Take S = {1, f, f 2 , . . . }. In this case, RS −1 is denoted Rf

Rf = {r/f i | r ∈ R, i ≥ 0}.

(3) R = Z, f = 2, then Rf = {a/2n | a ∈ Z, n ≥ 0}. This ring is all rational numbers with
denominators that are powers of 2 (when written in lowest terms). Notice that this is
the smallest ring containing Z such that 2 is invertible.
(4) R a ring, P a prime ideal of R.
Recall that the definition of P being prime is that if xy ∈ P then x ∈ P or y ∈ P .
Taking the contrapositive, we have x ̸∈ P and y ̸∈ P implies that xy ̸∈ P . In other
words, the complement of P is a multiplicative system!
So take S = {r ∈ R | r ∈
/ P } which is a multiplicative system since 1 ∈ S and
x ∈ S, y ∈ S implies xy ∈ S. In this case, RS −1 is denoted RP

RP = {r/s | r ∈ R, s ∈
/ P }.

(5) R = Z, P = (2), then RP = {a/b | a ∈ Z, b ∈

/ (2)}. This ring is all rational numbers
with odd denominators (when written in lowest terms).
(6) Note that if R is a domain, then (0) is a prime ideal. [Indeed, R being a domain is
equivalent to (0) being prime. Think through this logic!] So the localization at prime
ideals also covers the field of fractions case, where the field of fractions is R(0) .

The process of introducing inverses of elements doesn’t seem to be “local” in any particular way.
The origin of the word localization is really from algebraic geometry. In algebraic geometry, you
consider the geometric space Spec R whose points are given by prime ideals of R. To understand
behavior “locally” near the point P , you study the localization RP .

Example 283. We said that localizing at zero divisors can cause unexpected behavior. Intu-
itively, we have been thinking of the localization RS −1 as being bigger than then ring R, since
we have added inverses. When dealing with zero divisors though, this intuition can fail us.

R = Z/6Z, S = {1, 2, 22 , . . . } (= {1, 2, 4}).

Claim: RS −1 = Z/3Z.

126
 Use the universal property. Define ψ : Z/6Z → Z/3Z with [n]6 7→ [n]3 . Then ψ(s) is a unit for
all s ∈ S.
φ
Z/6Z / Z/6ZS −1

θ
ψ % 
Z/3Z

Claim: θ is an isomorphism.

Since ψ is surjective, θ is surjective. We want to show that θ is injective. If r/s ∈ RS −1 such

that θ(r/s) = 0 Then
0 = θ(r/s) = θ(r/1 · 1/s) = θ(r/1)θ(1/s)
and since θ(1/s) is a unit, therefore θ(r/1) = 0. Since φ(r) = r/1 therefore ψ(r) = 0 and so
r ∈ ker ψ = (3). Further, 3/1 = 0/1 in Z/6ZS −1 since 2(3 − 0) = 0 in Z/6Z.

Hence, r/1 = 0/1 in RS −1 so r/s = 0 in RS −1 . Hence ker θ is trivial so θ is an isomorphism.

The previous example shows how to work with localizations. Basically, the tool you have is to use
the universal property.

Example 284. F a field and R = F [x]. Let S = {1, x, x2 , . . . }.

RS −1 ∼
= F [x, x−1 ] = Laurent polynomials = {am xm + · · · + an xn | m ≤ n, m, n ∈ Z, ai ∈ F }.
The universal property gives
φ
F [x] / F [x]S −1

θ
ψ $ 
F [x, x−1 ]

where ψ(s) is a unit for all s ∈ S.

Claim: θ is an isomorphism.

It is clear that θ is surjective, since its image contains F [x] and x−1 = θ(1/x). We show that
θ is injective. Suppose θ(f /s) = 0. Then θ(f /1) = 0 = ψ(f ) by the same logic as the previous
example. But then f = 0 so f /s = 0 in F [x]S −1 so ker θ = {0/1}.

127
Definition 285. A local ring R is one with a unique maximal ideal.

Again the “local” in this term derives from algebraic geometry. We will see shortly that if P is a
prime ideal then the localization RP is a local ring.

Lemma 286. An ideal I in R is the unique maximal ideal of R if and only if every element of
R \ I is a unit in R.

Proof. [⇒]. Let I be the unique maximal ideal of R. Let a ∈

/ I and consider the ideal (a) = aR. If
aR ̸= R then aR is contained in a maximal ideal hence aR ⊆ I. Thus, a ∈ I, a contradiction. So
aR = R which means that a is a unit.

[⇐]. Conversely, suppose I is an ideal such that every element of R \ I is a unit in R. Then if
J ⊊ R is a proper ideal of R, we know that J does not contain any units and so J ⊆ I. Thus, I
contains every proper ideal so is the unique maximal ideal of R. □

Theorem 287. If P is a prime ideal of R, then RP is a local ring with maximal ideal

M = {r/s | r ∈ P, s ∈
/ P } ⊆ RP .

Proof. Note that M is clearly an ideal since if r1 /s1 , r2 /s2 ∈ M then r1 /s1 − r2 /s2 = (r1 s2 −
r2 s1 )/s1 s2 is also in M since r1 s2 − r2 s1 ∈ P (since r1 , r2 ∈ P ). Further, for any r/s ∈ RP , we
have r/s · r1 /s1 = rr1 /ss1 ∈ M so M is closed under multiplication by elements from RP .

Now suppose r/s ∈

/ M . Then r ∈
/ P and s ∈
/ P . Then r/s · s/r = 1 so r/s is a unit. Therefore
everything not in M is a unit so RP is a local ring with maximal ideal M . □

The Chinese Remainder Theorem.

The next section is on the Chinese Remainder Theorem. We will just state the theorem today, and
will prove it next time.

Definition 288. Let R be any ring. Two ideals I and J are called comaximal if I + J = R.

128
 Example 289. In R = Z, the ideals (2) and (3) are comaximal, since (2) + (3) contains 1 so
is equal to Z. Similarly, (2) and (9) are comaximal.

On the other hand, (2) + (4) = (2) since (4) ⊆ (2) so (2) and (4) are not comaximal. Also,
(4) + (6) = (2) so (4) and (6) are not comaximal.

Indeed, using the Euclidean algorithm it is not so hard to see that for integers m and n, the
ideals (m) and (n) are comaximal if and only if (m) + (n) = Z if and only if gcd(m, n) = 1.

Theorem 290 (Chinese Remainder Theorem). Let I1 , I2 , . . . , In be ideals in R. Suppose Ii and Ij

are comaximal for i ̸= j. Then
R ∼ R R
(1) = × ··· ×
I1 ∩ · · · ∩ In I1 In
(2) I1 ∩ · · · ∩ In = I1 I2 . . . In .

Example 291. If we take R = Z and I = (2) and J = (9) above, then the Chinese Remainder
Theorem says that
Z Z ∼ Z × Z .
= =
(18) (2) ∩ (9) (2) (9)

129
 Friday 11/18: Exam 2

You took your second midterm exam today.

21. Monday 11/21: Proof of CRT, Euclidean Domains, PIDs

Recall the Chinese Remainder Theorem:

Theorem (Chinese Remainder Theorem). Let I1 , I2 , . . . , In be ideals in R. Suppose Ii and Ij

are comaximal for i ̸= j. Then
R ∼ R R
(1) = × ··· ×
I1 ∩ · · · ∩ In I1 In
(2) I1 ∩ · · · ∩ In = I1 I2 . . . In .

Before we sketch the proof, let us look at some examples. In your problem set, you will explain
how this statement of the Chinese Remainder Theorem is related to the usual statement (about
simultaneously solving systems of congruences).

Example 292. n = pe11 · · · pemm where pi primes. Then

Z/nZ ∼
= Z/pe11 Z × · · · × Z/pemm Z

as rings.
e
The Chinese Remainder Theorem applies since (pei i ) and (pj j ) are comaximal if i ̸= j since
e e
(pei i ) + (pj j ) = (gcd(pei i , pj j ) = 1) = Z.

Notice that if we only consider the additive group structure of these rings, this proves the
elementary divisor version of FTFGAG for finite groups.

Also notice that restricting to units give

(Z/nZ)× ∼
= (Z/pe11 Z)× × · · · × (Z/pemm Z)× .

This is relevant if you are trying to understand the automorphisms of Z/nZ.

Example 293. F a field. In R = F [x], if a1 ̸= a2 ,

F [x] ∼ F [x] F [x]
= × .
(x − a1 )i1 (x − a2 )i2 (x − a1 )i1 (x − a2 )i2

130
 Since ((x − a1 )i1 ) + ((x − a2 )i2 ) = (1). We will understand the ring structure of F [x] more after
we study Euclidean domains in the next section.

Proof of CRT. Main idea: define a homomorphism

φ :R → R/I1 × · · · × R/In
r 7→ (r + I1 , . . . , r + In )

• ker φ = I1 ∩ · · · ∩ In is clear.
• Show φ is surjective. Then the first isomorphism theorem implies the Chinese Remainder
Theorem.

Why is (2) true? Think about n = 2 case. Note that I1 I2 ⊆ I1 ∩ I2 . Now pick x ∈ I1 and y ∈ I2
with x + y = 1 (can pick these x and y by comaximality).

Now if z ∈ I1 ∩ I2 , then z = z(x + y) = zx + zy ∈ I1 I2 . So I1 ∩ I2 ⊆ I1 I2 and hence I1 ∩ I2 = I1 I2 .

This proof works for general n.

Why is φ : R → R/I1 × R/I2 surjective?

Pick (r1 + I1 , r2 + I2 ) ∈ R/I1 × R/I2 . We want an r ∈ R such that (r + I1 , r + I2 ) = (r1 + I1 , r2 + I2 ).

So we need r − r1 ∈ I1 and r − r2 ∈ I2 . Since I1 + I2 = R, there exist s1 , s2 ∈ I1 and t1 , t2 ∈ I2
such that s1 + t1 = r1 and s2 + t2 = r2 .

Let r = t1 + s2 . Then
r − r1 = t1 + s2 − r1 = s2 − s1 ∈ I1
and
r − r2 = t1 + s2 − r2 = t1 − t2 ∈ I2
so φ is surjective. □

Euclidean Domains.
The next three sections of the text deals with learning about important special classes of rings,
namely Euclidean domains, principal ideal domains, and unique factorization domains. I will take
a slightly different order through these topics than the book, so you may have to look around a
bit to find the results and their proofs in the textbook. We begin by learning about Euclidean
domains.

Definition 294. An integral domain R is a Euclidean domain if there is a function

N : R → N = {0, 1, 2, . . . }

131
“the norm function” such that N (0) = 0 and for any a, b ∈ R with b ̸= 0, there exists q, r ∈ R such
that
a = qb + r
with r = 0 or N (r) < N (b).

This definition captures the fact that Euclidean domains have a division algorithm. Therefore,
they also have a Euclidean algorithm (hence the name). You should recognize from undergraduate
number theory that the above is very related to the division algorithm for the integers. Indeed,
they are the first example of a Euclidean domain.

Example 295. Let R = Z with N (a) = |a|.

Then Z is a Euclidean domain. The standard undergraduate proof goes thusly: If a, b ∈ Z with
b ̸= 0, take r to be of minimal norm among elements in S = {a − bq | q ∈ Z}. Then a = qb + r.
Check that either r = 0 or |r| < |b| (if not, add/subtract a multiple of b to r).

Another example of a Euclidean domain is the polynomial ring F [x] over a field F . In calculus, you
learn that given two polynomials in R[x], you can do long division of polynomials. For example, if
you want to divide the polynomial x3 + x2 − 1 by the polynomial x − 1, you have probably shown
your calculus students:

x2 + 2x + 2
x3 + x2


x−1 −1
− x3 + x2
2x2
− 2x2 + 2x
2x − 1
− 2x + 2
1

Therefore x3 + x2 − 1 = (x − 1)(x2 + 2x + 2) + 1. The quotient is x2 + 2x + 2 and the remainder is

1. This is also like a division algorithm, but the notion of “size” you use for a polynomial is degree.
It turns out that F [x] is always a Euclidean domain.

Example 296. Let F be a field. Then R = F [x] is a Euclidean domain with N (f ) = deg f .

132
 Proof. We can mimic the proof for Z. Let f, g ∈ F [x] with g ̸= 0. Let S = {f − hg | h ∈ F [x]}
and take r ∈ S of minimal degree (or r = 0). Then we have that r = f − qg for some fixed
q ∈ F [x].

We claim deg r < deg g. If deg r ≥ deg g, then we have

deg(r − λxdeg r−deg g g) < deg r

for some λ ∈ F . But then

r − λx∗ g = f − qg − λx∗ g = f − (q − λx∗ )g.

This element is of the form f −hg and so is in S, but it has lower degree than r. This contradicts
the choice of r. Therefore, deg r < deg g, as desired. □

For our next example of a Euclidean domain, we return to quadratic integer rings.

Example 297. Quadratic Integer Rings.

• Recall: let D be a squarefree integer. Define

√ √
Q( D) = {a + b D | a, b ∈ Q} ⊆ C

• The quadratic integer ring OQ(√D) = Z[ω] = {a + bω | a, b ∈ Z} where

√
 D if D ̸≡ 1 (mod 4)
ω= √
 1+ D if D ≡ 1 (mod 4).
2
√
• Sometimes these rings are Euclidean domains with the following norm: Let a + b D ∈
√
Q( D), then
√ √ √
N (a + b D) = |(a + b D)(a − b D)| = |a2 − Db2 | ∈ N.

√ √
Example 298. D = −1, O = Z[ −1] = {a + b −1 | a, b ∈ Z} the Gaussian integers.

N (a + bi) = (a + bi)(a − bi) = a2 + b2 = |a + bi|2 .

Z[i] is a Euclidean domain with respect to this norm N .

Proof. Given α, β ∈ Z[i], we want q, r ∈ Z[i] with α = qβ + r with r = 0 or

N (r) < N (β) ⇔ |r|2 < |β|2 ⇔ |r/β| < 1.

133
 So we want α/β = q + r/β, q ∈ Z[i], |r/β| < 1 or r = 0.

In complex plane, Z[i] are the lattice points.

Im z

α/β

Re z

√
We see that there exists q ∈ Z[i] such that |α/β − q| ≤ 2/2.

The idea that this lattice point q is very close to being the quotient of α by β. Let r = α − βq.
Then clearly α = qβ + r. Further,
√
2
|r/β| = |α/β − q| < < 1.
2
Hence, Z[i] is a Euclidean domain.

Note that in this Euclidean domain, you may have multiple choices for q that all have a
remainder r with N (r) < N (β). In the Gaussian integer picture, if α/β is in the center of
a square, there are four equally close lattice points, each of which would serve as a quotient.
This is different than in the cases of Z and F [x], where there is a unique quotient and a unique
remainder. □

Principal Ideal Domains.

Definition 299. An integral domain R is called a principal ideal domain (PID) if every ideal I of
R has the form
I = (a) = aR
for some a ∈ R.

134
Proposition 300. If R is a Euclidean domain, then R is a PID. In fact, if I is a nonzero ideal of
R, then I = (a), where a is any nonzero element of minimal norm N (a) among elements of I.

Proof. If I = {0} then I = (0). So suppose I ̸= (0).

Let a ̸= 0, a ∈ I such that N (a) = min{N (b) | b ∈ I, b ̸= 0}. Let x ∈ I. Since R is a Euclidean
domain,
x = qa + r
where r = 0 or N (r) < N (a) and r ∈ I. Thus, since a was of minimal norm, r = 0. Hence, x ∈ (a).
So I ⊆ (a).

Conversely, (a) ⊆ I since a ∈ I. Thus, I = (a). □

Example 301. The generator of an ideal in a Euclidean domain or a PID is not necessarily
unique (just generated by some element of minimal norm). In Z, (2) = (−2). In F [x], (f ) =
(λf ) for any 0 ̸= λ ∈ F .

Lemma 302. If R is an integral domain, (a) = (b) if and only if a = ub for some unit u ∈ R.

Proof. If a = 0 then (a) = (0) = {0} so b = 0 and a = b.

So assume a ̸= 0. Since a ∈ (b), a = bx for some x ∈ R. Similarly, since b ∈ (a), b = ay for some
y ∈ R.

Then a = bx = ayx so yx = 1 and hence x is a unit. Conversely, if a = ub where u is a unit then

clearly a ∈ (b) so (a) ⊆ (b). Also, b = u−1 a so (b) ⊆ (a). Thus, (a) = (b). □

Example 303. R[x]/(x2 + 1) ∼

= C.
The map φ : R[x] → C with f (x) 7→ f (i) is a homomorphism. Thus, ker φ is an ideal.

Note that x2 + 1 ∈ ker φ. We will show that ker φ = (x2 + 1) because R[x] is a PID and the
minimal degree of elements of ker φ is 2.

135
 If ax + b ∈ ker φ = I then ai + b = 0 and hence a = b = 0. Thus, x2 + 1 is of minimal
degree among nonzero elements of ker φ so ker φ = (x2 + 1). By the first isomorphism theorem,
R[x]/(x2 + 1) ∼
= C.

Definition 304. Let R be an integral domain, and let a, b ∈ R. We say that b divides a in R and
write b | a if there exists a c ∈ R such that a = bc. We also say that b is a divisor of a.

A common divisor d of a and b is called a greatest common divisor if e | a, e | b ⇒ e | d.

The gcd is defined only up to a unit. In Z, gcd(3, 6) could be ±3.

Note. Observe that if b | a then a = bc for some c ∈ R. This shows that a ∈ (b). Indeed, this is
actually an if and only if. And since the smallest ideal containing a is (a), this happens if and only
if (a) ⊆ (b).

To summarize:
b | a ⇐⇒ (a) ⊆ (b) ⇐⇒ a ∈ (b).

So there is an ideal-theoretic way to think about elements dividing other elements.

Wednesday 11/23: Thanksgiving Break

Enjoy your Thanksgiving break! Get some rest to finish the last two weeks of the semester strong.
Try not to forget everything from this semester (some forgetting is inevitable).

136
 22. Monday 11/28: PIDs and UFDs

The last thing we were talking about before break was Euclidean domains, PIDs, and divisors in
domains. We saw
b | a ⇐⇒ (a) ⊆ (b) ⇐⇒ a ∈ (b).

Lemma 305. (1) Let R be an integral domain. Then d is the gcd of a and b if and only if (d)
is the unique smallest principal ideal containing (a, b) = aR + bR.
(2) In a PID R, if (a, b) = (d) then d is a gcd of a and b.

Proof. (1) Suppose d = gcd(a, b). Then d | a and d | b so a, b ∈ (d) which implies (a, b) ⊆ (d). Also,
if (a, b) ⊆ (e) then e | a and e | b so e | d. Thus, (d) ⊆ (e).

Hence, (d) is the unique smallest principal ideal containing (a, b).

The converse is similar.

(2) follows from (1). Since if (a, b) = (d) then (d) is certainly the smallest principal ideal containing
(a, b). Hence, d is a gcd of a and b. □

Example 306. F a field, R = F [x]. Then gcd(x2 − 1, x3 − 1) = x − 1.

x2 − 1 = (x − 1)(x + 1) and x3 − 1 = (x − 1)(x2 + x + 1) and hence (x2 − 1, x3 − 1) ⊆ (x − 1).

Also, (x3 − 1) − x(x2 − 1) = x − 1 and hence x − 1 ∈ (x2 − 1, x3 − 1) so (x − 1) ⊆ (x2 − 1, x3 − 1).

Hence, (x − 1) = (x2 − 1, x3 − 1).

Note: −x + 1 is also a gcd, as is 2x − 2 (as long as 2 ̸= 0 in F ).

Definition 307. • Let R be an integral domain. Then a, b ∈ R are associates if a = ub for

some unit u ∈ R.21 [Which by Lemma 302, is equivalent to (a) = (b).]
• A nonzero, nonunit a ∈ R is irreducible if whenever a = bc for b, c ∈ R, then either b or c
is a unit.
• A nonzero, nonunit a ∈ R is prime if whenever a | bc for b, c ∈ R, then a | b or a | c.
(Equivalently, (a) is a prime ideal of R.)

21I like to think of the little ring elements wearing their business suits and carrying their briefcases, hanging out with
their associates.
137
 Example 308. In Z, irreducibles = primes = ±p (where p is a prime number).

Example 309. In F [x], F a field, finding irreducibles depends on F .

In C[x], the only irreducibles are λ(x − a), 0 ̸= λ ∈ C, a ∈ C. Also, in C[x], irreducibles =
primes.

On the other hand, in R[x], x2 + 1 is an irreducible.

Definition 310. Let R be an integral domain. Then R is a unique factorization domain (UFD) if

(1) Every nonzero, nonunit a ∈ R has an expression a = p1 p2 · · · pn with pi irreducible.

(2) Given two products of irreducibles p1 · · · pn = q1 · · · qm , then m = n and after reordering
the qj , we have pi and qi are associates for all i.

This definition says that in a UFD you can write any element as a product of irreducibles, which
is reminiscent of being able to write any integer as a product of primes. Further, this factorization
is essentially unique. Even in the integers, you can get “different” factorizations by reordering and
multiplying by −1. So part (2) says that factorizations into irreducibles are as unique as you could
hope for.

Example 311. Z is a UFD.

10 = 2 · 5 = −2 · −5 = −5 · −2.

Below, in Theorem 316, we will show that every PID is a UFD, so UFDs will fall into the following
spot in the schematic:

Euclidean domains ⊊ PIDs ⊊ UFDs ⊊ Dedekind domains ⊊ Integral domains.

We haven’t talked about Dedekind domains yet, and I’m not sure we’ll have time to this semester.
The rest of the inclusions should be clear, once we prove below that PIDs ⊆ UFDs, since we proved
in Proposition 300 that Euclidean domains ⊆ PIDs, and it is simply by definition that all of these
are examples of integral domains.
138
We still need to prove strict inclusion by giving an example of a PID that isn’t a Euclidean domain,
a UFD that isn’t a PID, and a domain that isn’t a UFD. In some sense, most of the remainder of
the semester is dedicated to understanding this whole schematic. So you should keep it in mind,
and keep updating your inventory of knowledge with respect to this chain of strict inclusions.

Lemma 312. (1) In any integral domain R, if a ∈ R is prime then a is irreducible.

(2) In a PID R, a ∈ R is prime if and only if a is irreducible if and only if (a) is a maximal
ideal.

Proof. (1) Let a be prime (and hence nonzero, nonunit). Let a = bc. Then a | bc. Since a is prime,
a | b or a | c. If a | b then b = ax so a = axc. hence, 1 = xc and c is a unit. On the other hand, if
a | c then b is a unit.

(2) Let R be a PID and a ∈ R irreducible (hence nonzero, nonunit). We claim (a) is maximal.
Take (a) ⊊ I ⊊ R. Since R is a PID, I = (x) so (a) ⊊ (x) ⊊ R.

So x | a and hence a = xy, some y ∈ R. Since a is irreducible, x or y is a unit. Since (a) ̸= (x), y
is not a unit. Since (x) ̸= R, x is not a unit. This is a contradiction, so (a) is maximal.

If (a) is maximal then (a) is prime so a is a prime element. Hence, irreducible implies prime.

By part (1), prime implies irreducible, so we are done. □

Remark. This also shows that in a PID, all nonzero prime ideals are maximal.

Definition 313. A commutative ring R is noetherian if, given ideals In in R with

I1 ⊆ I2 ⊆ I3 ⊆ . . .

there exists an N ≥ 1 such that In = IN for all n ≥ N (i.e., ascending chains of ideals eventually
stabilize). This is sometimes called the ascending chain condition (ACC). That is, R is noetherian
if it has the ACC on ideals.

Theorem 314. Let R be a commutative ring. TFAE.

(1) The ascending chain condition for ideals holds (i.e. R is noetherian).
(2) Any nonempty set of ideals of R has a maximal element.
(3) Every ideal of R is finitely generated (i.e. I = Rr1 + · · · + Rrn for ri ∈ R).

139
 S
Proof. (3) ⇒ (1). If I1 ⊆ I2 ⊆ · · · is a chain of ideals, then I = i≥0 Ii is an ideal of R. So
I = (r1 , r2 , . . . , rn ) and r1 , . . . , rn ∈ IN for some N . Hence, I = IN = IN +1 = · · · .

(1) ⇒ (2). Suppose S is a set of ideals with no maximal element. Choose I1 ∈ S. I1 is not maximal
so there exists I2 ∈ S with I1 ⊊ I2 . Then I2 is not maximal so there exists I3 ∈ S with I2 ⊊ I3 .
Continuing this process,we get
I1 ⊊ I2 ⊊ I3 ⊊ · · ·
which contradicts (1).

(2) ⇒ (3). Let I be an ideal of R. Let S be the set of finitely generated ideals contained in I.
S ̸= ∅ since (0) ∈ S. By (2), S has a maximal element, say J. If J ⊊ I, then pick any x ∈ I \ J.
If J = (y1 , . . . , ym ), then J + (x) = (y1 , . . . , ym , x) is finitely generated and J + (x) ⊆ I, which is a
contradiction. Since J ⊊ J + (x) ⊆ I. Hence, I = J so I ∈ S and so I is finitely generated. □

Lemma 315. If R is a PID, then R is noetherian.

Proof. Since every ideal is principal, every ideal is finitely generated. □

Theorem 316. Every PID is a UFD. In fact, if R is a noetherian integral domain in which every
irreducible element is prime,22 then R is a UFD.

Proof. Suppose that R is a PID. Let S be the set of principal ideals of R of the form (a) where a
is nonzero, nonunit, and a is not a finite product of irreducibles.

Assume S ̸= ∅. Since R is noetherian, therefore S has a maximal element, say (a). a is not
irreducible, so a = bc with b and c nonunits. Now (a) ⊊ (b) and (a) ⊊ (c) thus (b), (c) ∈
/ S.
So b and c are finite products of irreducibles. Hence, a = bc is also such a product. This is a
contradiction.

Hence, S = ∅. Therefore, every element of R is a finite product of irreducibles. We now need to

check that factorization is unique. Suppose p1 · · · pm = q1 · · · qn with pi , qj irreducible.

Since R is a PID, every irreducible is prime. Hence, p1 is prime and p1 | q1 · · · qn so p1 | qj for some
j. Renumber the qi ’s so that p1 | q1 .

Also q1 is irreducible, so p1 and q1 are associates. (q1 = p1 u but p1 is not a unit so u is).

Hence, p1 p2 · · · pm = up1 (q2 · · · 1n ) which implies that p2 · · · pm = uq2 · · · qn .

22Of course we already proved in Lemma 312 that in a PID every irreducible is prime.
140
By induction,we have that m = n and after rearrangement, each pi is an associate of qi . □

Lemma 317. If R is a UFD, then any two nonzero elements a, b ∈ R have a gcd.

Proof sketch. If a = upe11 pe22 · · · pemm where the pi are distinct and b = vpf11 pf22 · · · pfmm where ei , fi ≥ 0
and u, v are units. Then
min(e1 ,f1 ) min(em ,fm )
d = p1 · · · pm
is gcd(a, b). □

The above result is stated and proved more precisely in your book as Proposition 13 on Page 287,
but I just want you to realize that the gcd in a UFD is a concept that should be familiar to you.

141
 23. Wednesday 11/30: Polynomial Rings in Several Variables

We proved in Lemma 312 that in a PID every irreducible is prime. In fact, a stronger statement is
true. In any UFD, every irreducible is prime.

Lemma 318. In a UFD, if a is an irreducible element, then a is prime.

Proof. If a is irreducible and a | bc, then bc = ax. Now express b, c, and x as products of irreducibles.

(p1 · · · pm )(q1 · · · qn ) = a(r1 · · · rs )

which implies that a is an associate of a pi or a qj . Hence, a | b or a | c and so a is prime. □

We now give two examples of domains that are not UFDs.

√
Example 319. Let R = Z[ −5] = OQ(√−5) . Recall, as mentioned in Example 297, that this
√
ring comes with a norm function N defined by N (a + b −5) = a2 + 5b2 .

Fact. If r, s ∈ R then N (rs) = N (r)N (s).

Fact. An element u is a unit in R if and only if N (u) = 1. This is because if uv = 1 then

1 = N (1) = N (u)N (v). And norms only take on nonnegative integer values, so both N (u) and
N (v) must be 1.

Hence 2 and 3 are both irreducibles in R. This is because N (2) = 4. So if we write 2 = rs

as a product of irreducibles, then we must have N (r) = N (s) = 2. But R does not have any
elements of norm 2. Similarly, R does not have any elements of norm 3.
√ √
Also 1 + −5 and 1 − −5 are irreducible, since they have norm 6, so a proper factorization
would have to have factors of norm 2 and norm 3. But no such elements exist. But then
√ √
6 = 2 · 3 = (1 + −5)(1 − −5)

are two different factorizations of 6 into products of irreducibles. Hence, R is not a UFD.
However, R is an integral domain (since it is a subring of C which is an integral domain).

Example 320. Here’s another cute example.

F a field, R = {a0 + a1 x + a2 x2 + · · · + an xn | a1 = 0} ⊆ F [x].

R is a subring of F [x]. R is a domain which is not a UFD as x2 · x2 · x2 = x3 · x3 .

142
 Claim: x2 and x3 are irreducible. So this violates unique factorization.

If x2 = f g, f, g ∈ R nonunits, this forces f, g degree 1, since constants are units. But R has no
elements of degree 1.

x3 reducible also requires an element of degree 1.

Hence, x2 and x3 are irreducible.

What about a PID that is not a Euclidean domain? The example your book gives involves a little
bit more than what I want to go into, so I ask you to read it on your own.

√
Example 321. Page 277 of your textbook explains why the ring Z[(1 + −19)/2] = OQ(√−19)
is not a Euclidean domain. Page 282 explains why it is a PID. We don’t have time to do the
details here, but you should read them so that you have this in your toolkit as an example of
a ring that is a PID but not a Euclidean domain.

Once you have read the book’s example, we have now proved every part of:

Euclidean domains ⊊ PIDs ⊊ UFDs ⊊ Integral domains

other than finding a UFD that is not a PID. This takes us to our next section.

Polynomials in Several Variables.

Polynomial rings are perhaps the most important fundamental examples of rings.

Definition 322. R a commutative ring.

 
X 
R[x, y] = ai,j xi y j | ai,j ∈ R, only finitely many nonzero .
 
i,j≥0

In fact, R[x, y] ∼ j
P
= (R[x])[y]. An element of (R[x])[y] looks like j≥0 fj y with fj ∈ R[x] so
fj = i≥0 ai,j xi for ai,j ∈ R.
P
P 
Hence, j≥0 fj y j = j≥0 i y j . This looks like i j
P P P
a
i≥0 i,j x i,j≥0 ai,j x y . Construct the isomor-
phism.
143
Similarly, we have R[x1 , . . . , xn ] ∼= R[x1 ][x2 ] · · · [xn ]. Or R[xα | α ∈ S] for any index set S. For
example, F [x1 , x2 , . . . ] has elements that are finite sums of monomials xei11 · · · xeinn with F -coefficients.
(Each monomial is still a finite product and each element is a finite sum).

We now prove that being a PID is strictly stronger than being a UFD.

Lemma 323. If R is a PID that is not a field, then R[x] is not a PID.

Proof. Pick a nonzero, nonunit a ∈ R. Look at (a, x) ∈ R[x].

Suppose (a, x) = (f ) for f ∈ R[x].Then f | a so f is a constant and hence f ∈ R.

Also, f | x so x = f g. But f is a constant so f = c and g = b0 + b1 x + · · · + bn xn .

x = f g = cb0 + cb1 x + · · · + cbn xn implies 1 = cb1 which implies c is a unit in R and hence
(f ) = (c) = R[x].

But (a, x) ̸= R[x] since R[x]/(a, x) ∼

= R/(a) is not the zero ring since (a) ̸= R. Thus, (a, x) cannot
be principal. □

Example 324. Z[x] is not a PID since (2, x) is not principal.

Example 325. F a field, (F [x])[y] = F [x, y] is not a PID and (x, y) is not principal.

What is true is that both Z[x] and F [x, y] are UFDs, so that they complete our schematic. However,
we still need to prove it.

Example 326. The above examples are UFDs but not PIDs (we will see why soon).

Example 327. F [x1 , . . . , xn ] = F [x1 ] · · · [xn ] is a UFD but not a PID (we will see why soon).

We are trying to prove that if R is a UFD then R[x] is a UFD.

In fact, R is a UFD if and only if R[x] is a UFD, but the reverse direction is very straightforward.
The key thing to note is that R ⊆ R[x] is a subring, given by the constant polynomials. So suppose
144
R[x] is a UFD. Then in particular every constant polynomial c has a unique decomposition into
irreducibles. But this decomposition would necessarily have to be a decomposition into irreducible
factors of degree 0, since deg(c) = 0, and deg(f g) = deg(f ) + deg(g). But this is just a unique
decomposition into irreducibles of R.

Remark. Another thing to note from the inclusion R ⊆ R[x] is that the units in R[x] are the units
in R, since deg(1) = 0, so any units must also have degree 0.

A major idea today is to use the following. If R is a UFD, then it is a domain, so we may form the
field of fractions F . We know that F [x] is a UFD (in fact a PID). Every polynomial f ∈ R[x] can
also be viewed as a polynomial f ∈ F [x]. In F [x], it has a unique factorization into irreducibles.
So we just need to understand how factorizations in F [x] are related to those in R[x]. That is the
major work to do today.

Example 328. There are some subtleties to factorizations in R[x] vs F [x]. In this lecture, we
should usually think of R = Z and F = Q for intuition.

• 10x − 5 = 5(2x − 1) in Z[x]. So 10x − 5 is irreducible in Q[x] but not in Z[x].

• x2 − 5x + 6 = (1/3x − 2/3)(3x − 9) = (x − 2)(x − 3).

We showed in Lemma 317 that if R is a UFD, then any two elements (not both 0) have a gcd.
Similarly, any finite set of elements (not all 0) has a gcd.

Definition 329. Let R be a UFD, f ∈ R[x], f ̸= 0. Write

f = a0 + a1 x + · · · + am xm , where am ̸= 0.

The content of f is gcd(a0 , . . . , am ) in R (only defined up to unit multiple). Denote this C(f ) (this
notation is not standard, but we will use it today).

If f has content 1, then we say f is primitive.

Example 330. 12x2 + 15x − 3 ∈ Z[x] has content 3.

On the other hand, 4x2 + 5x − 1 ∈ Z[x] is primitive.

145
Lemma 331. If R is a UFD and f, g ∈ R[x] are primitive, then f g is primitive.

Proof. We need to show the gcd of the coefficients of f g is 1. Enough to show for any irreducible
p ∈ R, there is some coefficient of f g not divisible by p.

If p is irreducible, then since R is a UFD, p is prime and hence (p) ⊆ R is a prime ideal. Therefore
R/(p) is a domain. Define the map23

φp : R[x] → R/(p)[x]
X X
ai xi 7→ (ai + (p))xi .

Look at φp (f g) = φp (f )φp (g). We know that f and g have coefficients not divisible by p. So
φp (f ) ̸= 0, φp (g) ̸= 0. And R/(p)[x] is a domain so φp (f g) ̸= 0. So f g has a coefficient not
divisible by p. □

We need an important technical lemma.

Theorem 332 (Gauss’s Lemma). Let R be a UFD, and let F the field of fractions of R. Let
f ∈ R[x], f ̸= 0. If f = gh where g, h ∈ F [x], then there is a λ ∈ F such that g ′ = λg, h′ = λ−1 h
and g ′ , h′ ∈ R[x] and f = g ′ h′ .

Remark. What is Gauss’s Lemma really saying? Let’s use R = Z and F = Q, for intuition.

Take some irreducible polynomial f ∈ Z[x]. Perhaps f = 2x2 + 3. We cannot write this as a
product of two polynomials in Z[x] unless one of them is ±1 and so is a unit.

Now we can also consider this polynomial as a polynomial in Q[x]. Note that in this bigger ring,
we have more factorizations available to us! We can take 2x2 + 3 = 2 · (x2 + 3/2). So conceivably,
it might be possible to factor f in Q[x] even though it is irreducible in Z[x].

Gauss’s Lemma says that this never happens. If you can factor f in Q[x], then actually you could
factor it in Z[x].

Proof of Gauss’s Lemma. Notice that given any g ∈ F [x], there is a λ ∈ F (even λ ∈ R) so that
λg ∈ R[x]. Also, notice if h ∈ R[x], then h = rh′ where r = C(h) and h′ is primitive.

23This is a really important technique, called reduction mod p.

146
Therefore, if g ∈ F [x], we can find some λ ∈ R such that λg ∈ R[x], then we can divide by C(λg)
to get (λ/C(λg))g is a primitive element in R[x]. Call (λ/C(λg)) = λ1 ∈ F . Similarly, we can find
λ2 ∈ F such that λ2 h ∈ R[x] is primitive.

Now suppose f = gh is a factorization in F [x]. Then

λ1 λ2 f = (λ1 g)(λ2 h)

where λ1 g, λ2 h ∈ R[x] primitive. Therefore, by Lemma 331, (λ1 g)(λ2 h) is also primitive.

Now λ1 λ2 ∈ F implies that λ1 λ2 = c/d for some c, d ∈ R. Multiplying by d on both sides of the
above displayed equation yields
cf = d(λ1 g)(λ2 h).
Now both sides of the equation are actually in R[x]. Taking content of both sides, we have

cC(f ) = d

which implies C(f ) = d/c = (λ1 λ2 )−1 . Since content of a polynomial in R[x] is an element of R,
this implies that (λ1 λ2 )−1 ∈ R. Therefore,

f = (λ1 λ2 )−1 (λ1 g)(λ2 h)

is a factorization in R[x]. We can multiply the constant (λ1 λ2 )−1 ∈ R into either of the two other
factors and remain in R[x], so
f = (λ−1
2 g)(λ2 h). □

Next time, we will see how to use Gauss’s Lemma to show that F [x, y] and Z[x] are UFDs (and we
already know they are not PIDs).

147
 24. Monday 12/5: Irreducibility in Polynomial Rings

Recall that we are trying to understand why F [x, y] and Z[x] are UFDs but not PIDs. The last
thing we did was to prove Gauss’s Lemma:

Theorem (Gauss’s Lemma). Let R be a UFD, and let F the field of fractions of R. Let
f ∈ R[x], f ̸= 0. If f = gh where g, h ∈ F [x], then there is a λ ∈ F such that g ′ = λg,
h′ = λ−1 h and g ′ , h′ ∈ R[x] and f = g ′ h′ .

Corollary 333. Let R be a UFD with field of fractions F . If f ∈ R[x] and deg f ≥ 1, then f is
irreducible in R[x] if and only if f is irreducible in F [x] and f is primitive.

Proof. Suppose f is irreducible in F [x] and f is primitive (in R[x]). Suppose f = gh in F [x]. Since
f is irreducible in F [x] so g or h is constant. Say g ∈ R. But we assumed that f was primitive,
and so g is a unit in R.

Conversely, if f ∈ R[x] is irreducible in R[x] but reducible over F [x] then f = gh for some
g, h ∈ F [x]. Then f = (λg)(λ−1 h) with λg, λ−1 h ∈ R[x] by Gauss’ Lemma. So f is not irreducible
over R[x], which is a contradiction.

Also, if f does not have content 1, then f = C(f )f ′ for some f ′ ∈ R[x] so f is not irreducible over
R, which is a contradiction. □

Theorem 334. If R is a UFD then R[x] is a UFD.

Proof. We need to prove that (1) every element of R[x] is a product of irreducibles and (2) it is
unique up to reordering. We only prove (1).

Let F be the field of fractions of R. Let f ∈ R[x]. Then f = rg for some r ∈ R with g primitive.

In F [x], which is a UFD, we can factor g = q1 · · · qn where each qi irreducible in F [x]. Now we use
Gauss’s Lemma to write
g = (λ1 q1 ) · · · (λn qn )
where now each λi qi ∈ R. Since we have only multiplied by an element of F , and F is a field, each
λi qi is still irreducible.

Now we have
f = rg = r(λ1 q1 ) · · · (λn qn )
148
Since C(g) = 1, this forces C(λi qi ) = 1 for each i. Since each λi qi is a primitive in R[x] which is
irreducible in F [x], by Corollary 333, it is also irreducible in R[x].

Now since r ∈ R, and R is a UFD, we may factor r into irreducibles r = s1 · · · sm . These remain
irreducible in R[x]. Hence,
f = s1 · · · sm (λ1 q1 ) · · · (λn qn )
is a factorization of f into irreducibles in R[x].

For the proof of uniqueness, you also use uniqueness of factorizations in F [x], but it just technical
details. □

After all this work, we have finally shown that R[x] is a UFD if and only if R is a UFD.

Our next section is motivated by the following question.

Motivating Question: If F is a field, how can you tell if f ∈ F [x] is irreducible?

Recall that we saw that the irreducibles in C[x] are different from the irreducibles in R[x], and so
in polynomial ring over a field, we cannot give a totally uniform straightforward answer to what
the irreducibles are.

The following theorem should be familiar to you from calculus. A polynomial has a root a if and
only if you can factor out a linear factor of x − a from the polynomial.

Theorem 335 (Remainder/Factor Theorem). Let F be a field. If f ∈ F [x] and a ∈ F , then

f (x) = q(x)(x − a) + r(x)

where r(a) = f (a). Hence, (x − a) | f (x) iff f (a) = 0.

Proof. Since F [x] is a Euclidean domain, we may divide f by x − a to obtain

f (x) = q(x)(x − a) + r(x)

for some r(x) of degree 0. Evaluating both sides when x = a yields that r(a) = f (a).

If x − a | f (x) then f (x) = (x − a)g(x) for some g(x) ∈ F [x] and hence f (a) = (a − a)g(a) = 0.
Conversely, if f (a) = 0, then using the division algorithm as above, we have f (x) = q(x)(x−a)+r(x)
where r(x) is a constant polynomial with r(a) = 0. But this means r(x) = 0 is the constant
polynomial 0 and so f (x) = q(x)(x − a). □

As a corollary, in degree 2 or 3, we can use roots to determine irreducibility.

149
Corollary 336. If f (x) ∈ F [x] has degree 2 or 3, then f (x) is irreducible iff f has no root in F .

Proof. For any f ∈ F [x], if f has a root in F , then f = (x − a)q(x) which implies f is reducible.

Conversely, suppose that f is reducible so f = gh for two nonunits g, h. Since they are not units,
this means that deg g, deg h > 0. Since deg f = 2 or 3, then at least one of g or h must have degree
1. Suppose g = bx − c for some b, c ∈ F . Then c/b ∈ F is a root of f . □

Example 337. In Z2 [x], the polynomial x2 + x + 1 is irreducible, since it has no roots in Z2 .

Proposition 338 (Rational Roots Test). Suppose

f (x) = an xn + · · · + a1 x + a0

with ai ∈ Z. Then if f has a root in Q, say p/q in lowest terms, then p | a0 and q | an in Z.

Proof. Evaluating f at p/q yields

f (p/q) = an (p/q)n + · · · + a1 (p/q) + a0 = 0.

Clearing denominators by multiplying by q n yields

an pn + an−1 pn−1 q + · · · + a1 pq n−1 + a0 q n = 0.

Solving for a0 q n , we see that p | a0 q n . But since we have written p/q in lowest terms, gcd(p, q) = 1
so p | a0 .

Similarly, q | an pn so q | an . □

The following result is also familiar to you, although its proofs are mostly analytic in nature (since
the construction of C is fundamentally analytic). I think calling it the Fundamental Theorem of
Algebra is a misnomer.

Theorem 339 (Fundamental Theorem of Algebra). Every polynomial f ∈ C[x] of deg f ≥ 1 has
a root in C.

150
Corollary 340. So the irreducible polynomials over C are the polynomials c(x − a).

Proof. If f ∈ C[x], then f has a root a ∈ C so f = (x − a)g(x).

By induction, f = c(x − a1 ) · · · (x − an ), so every polynomial can be completely factored into linear

factors. Hence, a polynomial of degree ≥ 2 can never be irreducible.

Each linear polynomial is irreducible, since if you could write it as a product, one of the factors
would have degree 1 and the other degree 0. But the constant polynomials are all units. □

Note that actually we did not use anything in particular about C except that every polynomial in
C[x] splits completely as a product of linear factors. This makes C what is known as an algebraically
closed field.

Definition 341. A field F is called algebraically closed if every polynomial f ∈ C[x] of degree
deg f ≥ 1 has a root in F .

Really the corollary clearly holds over any algebrically closed field.

Example 342. The irreducible polynomials in R[x] are c(x − a), c ̸= 0 and c(x2 + ax + b),
c ̸= 0 where a2 − 4b < 0 (which implies that x2 + ax + b has complex roots).

The following theorem establishes a very important criterion for determining irreducibility of poly-
nomials.

Theorem 343 (Eisenstein Criterion). Let R be a UFD. Suppose f ∈ R[x] and write

f = xn + an−1 xn−1 + · · · + a1 x + a0 ,

where n ≥ 1. [Note that we are assuming the leading coefficient is 1.]

If there is an irreducible element p ∈ R such that p | an−1 , p | an−2 , ..., p | a0 and p2 ∤ a0 then f is
irreducible in R[x]. Hence, f is also irreducible in F [x] where F is the field of fractions of R.

Proof. Let R = R/(p) (which is an integral domain since p is prime).

151
Suppose f = gh with deg h < deg f and deg g < deg f . Write

g = xm + bm−1 xm−1 + · · · + b0 and h = xs + cs−1 xs−1 + · · · + c0 .

Let φ : R → R be the quotient homomorphism. This induces a homomorphism φ : R[x] → R[x]

(which reduces the coefficients mod p).

Apply φ to f = gh. Then f = xn = gh since p | ai for all i.

Suppose there is some bi such that p ∤ bi . Choose the smallest such i (so that p | bk for all k < i).
Similarly, suppose there is a cj such that p | cj but p | ck for all k < j.

g = xm + · · · + bi xi and h = xs + · · · + cj xj .

Then gh = xm+s + · · · + bi cj xi+j and bi cj ̸= 0 since R is an integral domain. But this contradicts
that f = xn . Hence, we must have p | bi and p | cj for all i and j.

In particular, p | b0 , p | c0 so p2 | b0 c0 = a0 . This is a contradiction. Hence, f is irreducible in R[x].

By Gauss’s Lemma, f is also irreducible in F [x] where F is the field of fractions of R. □

Example 344. Let R = Z so F = Q, and let f = x4 + 2x3 + 6. Use p = 2 in Eisenstein’s

Criterion to see that f is irreducible in Z[x] and Q[x].

Example 345. We can sometimes use clever change of variables to use Eisenstein’s Criterion.

Let f = xp−1 + xp−2 + · · · + x + 1 where p is prime. We claim that f is irreducible in Z[x] and
Q[x].
xp −1
Substitute x + 1 for x. Note: f = x−1 . So
p p p
xp−1 + · · · +



(x + 1)p − 1 x + x
   
p−1 1 p−1 p p−2 p
f (x + 1) = = =x + x + ··· + .
x x p−1 1
Now p | pi for 0 < i < p and p2 ∤ p



1 = p. So f (x + 1) is irreducible by Eisenstein’s Criterion.

But any factorization of f (x) = g(x)h(x) yields a factorization f (x + 1) = g(x + 1)h(x + 1).
Hence, since, f (x + 1) is irreducible, so is f (x).

Example 346. Let F be a field and consider F [x, y] = F [x][y]. Let

f = y n + y n−1 x2 + x ∈ F [x, y].

152
 Let R = F [x], which is a UFD (in fact a PID). We can then use Eisenstein’s Criterion on the
ring R[y].

Since x ∈ F [x] is irreducible in R and x divides all coefficients but x2 ∤ x, Eisenstein applies
with p = x. So f is irreducible in R[y] = F [x, y].

The final result we prove today is that Z×

p is cyclic, which is a very useful fact for group theory
qual problems. Combined with the fact that we know that Aut(Zp ) = (Zp )× has order p − 1, this
shows that Aut(Zp ) is a cyclic group of order p − 1.

Lemma 347. Let F be a field, and suppose f ∈ F [x] is of degree n. Then f has at most n distinct
roots in F .

Proof. If r1 , . . . , rm are roots of f in F , then by induction

f (x) = (x − r1 )(x − r2 ) · · · (x − rm )g(x)

which implies deg f = n ≥ m. □

Theorem 348. Let F be a field. If G is a finite subgroup of F × = F \ {0}, then G is cyclic. In

particular, Z×
p is cyclic.

Proof. Let G be a finite subgroup of F × . Let d > 1. How many elements of order dividing d can
G have? Note that if α ∈ F satisfies αd = 1 (i.e., has order dividing d in F × ), then it is a root of
the polynomial xd − 1. Since this polynomial has at most d roots in F , therefore G has at most d
elements of order dividing d.

Claim: Any finite abelian group G such that G has at most d elements of order d for all d > 1 is
cyclic.

To prove this claim, note that since multiplication in F is commutative, therefore G is an abelian
group. Therefore, by the invariant factor version of FTFGAG, we know that

G∼
= Zn1 × Zn2 × · · · × Znr

where n1 | n2 | · · · | nr .
153
If r > 1, taking any prime p | n1 implies p | n2 . By Cauchy’s Theorem, we may find x ∈ Zn1 ,
y ∈ Zn2 of order p. Then (xi , y i , e, e, . . . ) are all elements of order dividing p and there are p2 of
them. So r = 1 which implies that G is cyclic. □

154
 25. Wednesday 12/7: Hilbert’s Basis Theorem

The last result main of the semester will be Hilbert’s famous Basis Theorem.

Theorem 349 (Hilbert’s Basis Theorem). Let R be a commutative ring. If R is noetherian, then
R[x] is noetherian.

Proof. We will prove the theorem by proving that all ideals of R[x] are finitely generated. Given
f = a0 + a1 x + · · · + an xn ∈ R[x] with an ̸= 0, an is called the leading coefficient of f . Write
in(f ) = an for the leading coefficient of f (the notation in(f ) stands for for initial coefficient; note
that it is only the coefficient of the leading term). Now pick an ideal I ⊆ R[x]. For fixed d ≥ 0, the
set
I (d) = {in(f ) | f ∈ I, deg f = d} ∪ {0}
is an ideal of R.

• Closure under subtraction: f, g ∈ I, deg f = deg g = d implies f = a0 + · · · + ad xd and

g = b0 + · · · + bd xd with ad , bd ̸= 0.
Then f − g ∈ I and f − g = a0 − b0 + · · · + (ad − bd )xd either ad − bd = 0 or else f − g is
of degree d and has leading coefficient ad − bd .
• Close under multiplication: easy

We also have Ie = {in(f ) | f ∈ I} ∪ {0} is an ideal of R since

[
Ie = I (d)
d≥1

and I (d) ⊆ I (d+1) for all d. Since R is noetherian, Ie is finitely generated, so we may pick ick
f1 , f2 , . . . , fn ∈ I such that Ie = (in(f1 ), . . . , in(fn )) as an ideal of R. Let N = max(deg(fi )). For
each d ≤ N , also pick gd,1 , . . . , gd,md ∈ I with degree d such that

(in(gd,1 ), . . . , in(gd,md ) = I (d)

as ideals of R.

Claim: {f1 , . . . , fn } ∪ {gd,i | 0 ≤ d ≤ N } is a generating set for I.

Proof of Claim: Define J = (f1 , . . . , fn , gd,i )0≤d≤N,1≤i≤md ⊆ I (each generator is in I).

Suppose J ⊊ I and pick h ∈ I, h ∈

/ J, such that h has smallest degree possible degree, say D.

e Write h = a0 + · · · + aD xD . Now
• Case 1: D = deg h ≥ N . We know that in(h) ∈ I.
P
aD ∈ Ie so aD = ri in(fi ) for some ri ∈ R.
155
 ri fi xD−deg fi has smaller degree than h and is in I. By the minimality of the
P
Now h −
ri fi xD−deg fi is also clearly in J, we
P
degree of h, therefore it is also in J. But then. since
see that h ∈ J. Contradiction.
P
• Case 2: D = deg h ≤ N . in(h) = ri in(gD,i ) by definition of the gd,i .
P P
Now h − ri gd,i ∈ I has smaller degree than h. So h − ri gd,i ∈ J. But this again
implies that h ∈ J which is a contradiction.

Therefore, I = J so I is finitely generated. Hence, R[x] is noetherian. □

Corollary 350. F a field, then R = F [x1 , . . . , xn ]/I for any ideal I is noetherian.

Proof. Apply Hilbert’s Basis Theorem n times. Any quotient ring of a noetherian ring is noetherian.
Hence, we obtain the corollary. □

The rings F [x1 , . . . , xn ]/I (and their corresponding schemes) are the main objects of study in
algebraic geometry. So this is a useful corollary.

Case study: the Gaussian Integers.

At this point, we have really only carefully studied the irreducibles in Z (the prime numbers),
C[x] (the linear polynomials), and R[x] (the linear polynomials and the quadratics ax2 + bx + c
√
where b2 − 4ac < 0). We now work to understand the irreducibles in the Gaussian integers
R = Z[i] = {a + bi | a, b ∈ Z}.

• Recall that we already showed that R is a Euclidean domain back in Example 298. In
particular, R is a UFD.
• Units in R? N (a + bi) = |a2 + b2 | = ±1 if and only if a = 0, b = ±1 or a = ±1, b = 0.
Hence, the units in R are ±1 and ±i.
• What are the irreducibles in R?
Case 1: Suppose a ̸= 0, b ̸= 0 and a + bi is irreducible. Notice φ : Z[i] → Z[i],
a + bi 7→ a − bi is an automorphism. Therefore, a − bi is also irreducible.

(a + bi)(a − bi) = a2 + b2 ∈ Z
Claim: a2 + b2 = p is prime in Z. If not, p = m1 m2 for m1 , m2 ∈ Z not ±1. If m1 , m2
are irreducible in R, then p = m1 m2 = (a + bi)(a − bi). This contradicts that R is a UFD.
If m1 , m2 are not irreducible in R, then p factors into different number of irreducibles.
Contradicts that R is a UFD.
Now, it is enough to consider only integer irreducibles. Otherwise, multiply by i, a unit.
156
 Case 2: If n ∈ Z is irreducible in Z[i], it is certainly prime in Z. If p ∈ Z is prime but
not irreducible in Z[i], then p has some irreducible factor a + bi with a ̸= 0, b ̸= 0. Then
a − bi is also an irreducible factor of p (by automorphism).Thus (a + bi)(a − bi) = a2 + b2
divides p which implies a2 + b2 = p. Thus, p has precisely two irreducible factors.
(We were not that careful about when a + bi and a − bi are associates. Happens when
a = ±b. Only case to worry about is a = b = ±1. Check this case separately).

Example 351. 2 = (1 + i)(1 − i)

5 = (2 + i)(2 − i) = (1 + 2i)(1 − 2i)

7 is irreducible. 11 is irreducible.

13 = (2 + 3i)(2 − 3i)

Theorem 352. Let p ∈ Z be a prime number. Then either

(1) p is irreducible in Z[i], or

(2) p = (a + bi)(a − bi) where a + bi, a − bi are irreducible in Z[i], a ̸= 0, b ̸= 0.
Also, all irreducibles appear as a factor of some prime in Z. p is not irreducible in Z[i] if
and only if p = a2 + b2 for a, b ∈ Z.

Proof. We just did this. □

Theorem 353. Let p ∈ Z be prime in Z. Then p is irreducible in Z[i] if and only if p ≡ 3 (mod 4).

Proof. 2 is reducible. 2 = (1 + i)(1 − i). So assume p is odd.

[⇐]. If a2 + b2 = p, since a2 ≡ 0, 1 (mod 4), this implies that a2 + b2 ≡ 0, 1, or 2 (mod 4). Hence,
p ̸≡ 3 (mod 4).

[⇒]. Assume p ≡ 1 (mod 4). Now Z× ×

p is a cyclic group and |Zp | = p − 1 is a multiple of 4. So there
exists a ∈ Z× 2 2
p of order 4. So a has order 2. Since −1 is the unique element of order 2, a = −1
which implies a2 + 1 ≡ 0 (mod p) which implies p | a2 + 1 in Z.

Note a2 + 1 = (a − i)(a + i) which means p | (a + i)(a − i) in Z[i].

Suppose for contradiction that p is irreducible. Since Z[i] is a Euclidean domain, prime is equivalent
to irreducible. So p is a prime element in Z[i] so p | a + i or p | a − i.
157
But then for some c +di, a +i = p(c + di) = pc + pdi which implies pd = 1, which is a contradiction,.
Hence, p is irreducible in Z[i], p = (a + bi)(a − bi). □

Corollary 354. An integer n ∈ Z is a sum of two squares in Z if and only if n = pe11 · · · pemm where
each pi ≡ 3 (mod 4) appears to an even power.

Proof. Suppose p1 , p2 , . . . pn ≡ 3 (mod 4) and pn+1 , . . . , pm ̸≡ 3 (mod 4).

In Z[i], n = pe11 · · · penn (an+1 + bn+1 i)en+1 (an+1 − bn+1 i)en+1 · · ·

If n = a2 + b2 , n = (a + bi)(a − bi) = xx, This happens iff the irreducible factors of n occur in
conjugate pairs. □

158