Visual Abstract Algebra

Marcus Pivato
March 25, 2003
2
Contents
I Groups 1
1 Homomorphisms 3
1.1 Cosets and Coset Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Lagranges Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 Normal Subgroups & Quotient Groups . . . . . . . . . . . . . . . . . . . . . . . 7
1.4 The Fundamental Isomorphism Theorem . . . . . . . . . . . . . . . . . . . . . . 10
2 The Isomorphism Theorems 13
2.1 The Diamond Isomorphism Theorem . . . . . . . . . . . . . . . . . . . . . . . . 13
2.2 The Chain Isomorphism Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.3 The Lattice Isomorphism Theorem . . . . . . . . . . . . . . . . . . . . . . . . . 17
3 Free Abelian Groups 19
3.1 Rank and Linear Independence . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.2 Free Groups and Generators . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.3 Universal Properties of Free Abelian Groups . . . . . . . . . . . . . . . . . . . . 22
3.4 () Homological Properties of Free Abelian Groups . . . . . . . . . . . . . . . . 23
3.5 Subgroups of Free Abelian Groups . . . . . . . . . . . . . . . . . . . . . . . . . 24
4 The Structure Theory of Abelian Groups 29
4.1 Direct Products . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4.2 The Chinese Remainder Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.3 The Fundamental Theorem of Finitely Generated Abelian Groups . . . . . . . . 37
5 The Structure Theory of Nonabelian Groups 41
5.1 Three Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
5.2 Simple Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
5.3 Composition Series; The Holder program . . . . . . . . . . . . . . . . . . . . . 43
5.4 () The Classication of Finite Simple Groups . . . . . . . . . . . . . . . . . . 51
5.5 () The Extension Problem; Short Exact Sequences . . . . . . . . . . . . . . . 52
5.6 () Exact Sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
5.7 Solvability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
i
ii CONTENTS
6 Group Actions 57
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
6.2 Orbits and Stabilizers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
6.3 The Simplicity of A
N
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
II Rings 73
7 Introduction 75
7.1 Basic Denitions and Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
7.2 () Many Examples of Rings . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
7.3 () Applications of Ring Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
7.3.1 Algebraic Number Theory . . . . . . . . . . . . . . . . . . . . . . . . . . 92
7.3.2 Field Theory & Galois Theory . . . . . . . . . . . . . . . . . . . . . . . 93
7.3.3 Algebraic Geometry & Commutative Algebra . . . . . . . . . . . . . . . 95
7.3.4 Algebraic Topology & Homological Algebra . . . . . . . . . . . . . . . . 96
7.3.5 Group Representation Theory & Ring Structure Theory . . . . . . . . . 96
7.4 Subrings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
7.5 Units and Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
7.6 Zero Divisors and Integral Domains . . . . . . . . . . . . . . . . . . . . . . . . 99
8 Homomorphisms, Quotients, and Ideals 103
8.1 Homomorphisms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
8.2 () Many Examples of Ring Homomorphisms . . . . . . . . . . . . . . . . . . . . 106
8.3 Ideals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
8.4 Quotient Rings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
8.5 The Fundamental Isomorphism Theorems . . . . . . . . . . . . . . . . . . . . . 114
8.6 The Ring Isomorphism Theorems . . . . . . . . . . . . . . . . . . . . . . . . . . 115
9 Algebraic Geometry 117
9.1 Algebraic Varieties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
9.2 The Coordinate Ring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
9.3 Morphisms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
9.4 Irreducible Varieties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
10 Ideal Theory 133
10.1 Principal Ideals and PIDs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
10.2 Maximal Ideals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
10.2.1 ...in polynomial rings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
10.2.2 ...have Simple Quotients . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
10.3 The Maximal Spectrum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
10.3.1 ...of a Continuous Function Ring . . . . . . . . . . . . . . . . . . . . . . 144
CONTENTS iii
10.3.2 ...of a Polynomial Ring . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
10.3.3 ...of a Coordinate Ring . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
10.4 The Zariski Topology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
10.4.1 ...for Continuous Function Rings . . . . . . . . . . . . . . . . . . . . . . 154
10.4.2 ()...for an arbitrary Ring . . . . . . . . . . . . . . . . . . . . . . . . . . 158
10.4.3 ()...on a eld . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
10.4.4 ()...on ane n-space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
10.4.5 () More about the Zariski Topology . . . . . . . . . . . . . . . . . . . . 163
10.5 Prime Ideals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
10.6 The Prime Spectrum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
10.7 Radical, Variety, Annihilator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
10.7.1 The Jacobson Radical . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
10.7.2 Variety and Annihilator (Continuous Function Rings) . . . . . . . . . . . 169
10.7.3 The Nil Radical . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
10.7.4 Variety and Annihilator (Polynomial Rings) . . . . . . . . . . . . . . . . 173
III Fields 175
11 Field Theory 177
11.1 Compass & Straight-Edge Constructions I . . . . . . . . . . . . . . . . . . . . . 177
11.2 Field Extensions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
11.2.1 Extensions as Vector Spaces: . . . . . . . . . . . . . . . . . . . . . . . . . 186
11.2.2 Simple Extensions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
11.2.3 Finitely Generated Extensions . . . . . . . . . . . . . . . . . . . . . . . . 188
11.3 Finite and Algebraic Extensions . . . . . . . . . . . . . . . . . . . . . . . . . . 189
11.4 (multi)Quadratic Extensions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
11.5 Compass & Straight-Edge Constructions II . . . . . . . . . . . . . . . . . . . . 202
11.6 Cyclotomic Extensions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
11.7 Compass & Straight-Edge Constructions III . . . . . . . . . . . . . . . . . . . . 213
11.8 Minimal Root Extensions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
A Background: Topology 221
A.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
A.2 Abstract Topological Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226
A.3 Open sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228
A.4 Compactness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
iv CONTENTS
List of Figures
1.1 Examples (13a) and (13b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.1 Example 17a). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.2 The Chain Isomorphism Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.3 Example 19a). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.4 The Lattice Isomorphism Theorem . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.1 Z
2
= Z Z is the free group of rank 2. . . . . . . . . . . . . . . . . . . . . . . 20
3.2 Example 34a): / = ZZ, and B is the subgroup generated by x = (2, 0) and
y = (1, 2). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.3 Example 34a): / = (Za
1
) (Za
2
), where a
1
= (1, 2) and a
2
= (0, 1). Let
m
1
= 1 and m
2
= 4, so that b
1
= m
1
a
1
= (1, 2) and b
2
= m
2
a
2
= (0, 4).
Then B = (Zb
1
) (Zb
2
). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.1 The product group ( = /B, with injection and projection maps. . . . . . . . 30
4.2 Example 43a). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
5.1 Example 56a) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
5.2 An exact sequence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
6.1 The Orbit-Stabilizer Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
7.1 Functions with compact support. . . . . . . . . . . . . . . . . . . . . . . . . . . 82
7.2 Classical compass and straightedge constructions. . . . . . . . . . . . . . . . . . 94
7.3 Algebraic varieties. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
7.4 Zero divisors in the rings c(R), c
K
(R), and c

(R). . . . . . . . . . . . . . . . 100
9.1 Some algebraic varieties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
9.2 The map (x, y) = (x
2
y
2
, 2xy, x) transforms a circle into a double-loop. . 124
9.3 The map (x, y) = (x
2
, y
2
) transforms a circle into a diamond. . . . . . . . . 124
9.4 V = U W . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
10.1 A schematic representation of the lattice of ideals of the ring 1. The maximal
ideals form the top row of the lattice, just below 1. . . . . . . . . . . . . . . . 138
v
vi LIST OF FIGURES
10.2 A schematic representation of the lattice of ideals of the ring Z. The maximal
ideals are the principal ideals (2), (3), (5), etc. generated by prime numbers. . . 138
10.3 A schematic representation of the lattice of ideals of the ring R[x]. The maximal
ideals are the principal ideals (x), (x1), (x
2
+1), etc. generated by irreducible
polynomials. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
10.4 A schematic representation of the maximal spectrum of the ring 1. . . . . . . . 144
10.5 A schematic representation of the maximal spectrum of the ring c[0, 1], showing
the bijective correspondence between points in [0, 1] and maximal ideals in c[0, 1]. 144
11.1 Bisecting a line segment. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
11.2 Bisecting an angle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
11.3 Constructing a regular hexagon . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
11.4 (A) Doubling the cube. (B) Squaring the circle . . . . . . . . . . . . . . . . 179
11.5 The starting point: an origin point O, a unit circle C, and lines H and V. . . . 180
11.6 The ve basic operations with compass and straight-edge. . . . . . . . . . . . . 181
11.7 (A) Addition of lengths. (B) Subtraction of lengths. . . . . . . . . . . . . . . 182
11.8 Multiplication of lengths. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
11.9 Division of lengths. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
11.10Computing the square root of a length . . . . . . . . . . . . . . . . . . . . . . . 183
11.11The three cube roots of 3 in the complex plane. . . . . . . . . . . . . . . . . . . 192
11.12(A) Q() has degree 3 over Q, but R() only has degree 2 over R. This is
because m
;Q
(x) = x
3
2, but m
;R
(x) = x
2
+
3

2x+
3

4. (B) F() has degree

D
1
over F, while K() has degree D
2
D
1
over K. . . . . . . . . . . . . . . . . 193
11.13 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
11.14Roots of unity in the complex plane. . . . . . . . . . . . . . . . . . . . . . . . . 205
11.15E is the unique minimal root extension of F for the polynomial p(x). . . . . . . . 217
11.16E = Z
/5
(

3) has 25 elements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220

A.1 The sequence p
1
, p
2
, p
3
, . . . converges to . . . . . . . . . . . . . . . . . . . . . 222
A.2 The subsets N = (, 0] and P = (0, ) are disjoint, but there is a sequence
in P converging to a point in N. . . . . . . . . . . . . . . . . . . . . . . . . . . 225
A.3 If x
1
, x
2
, x
3
, . . . is a sequence converging to x

, and y
n
= f(x
n
) for all n, then
the sequence y
1
, y
2
, y
3
, . . . must converge to y = f(x

). . . . . . . . . . . . . 225
A.4 U is disconnected if there are two disjoint closed subsets C
1
and C
2
in X so
that U C
1
. C
2
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
A.5 tan :
_

2
,

2
_
R is a homeomorphism, with inverse homeomorphism arctan :
R :
_

2
,

2
_
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228
A.6 An open cover of (0, 1) which has no nite subcover. . . . . . . . . . . . . . . . 230
A.7 The Chinese Box property. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
A.8 The Chinese Box Property in the unit interval. . . . . . . . . . . . . . . . . . . 231
A.9 No Chinese Box Property for the real line. . . . . . . . . . . . . . . . . . . . . . 232
Part I
Groups
1
Chapter 1
Homomorphisms
1.1 Cosets and Coset Spaces
Prerequisites: Subgroups
If / < ( is a subgroup, and g (, then the corresponding left and right cosets of / are
the sets
g/ = ga ; a /, and /g = ag ; a /.
Example 1:
(a) Let ( = (Z, +), and let / = 5Z = 5z ; z Z = . . . , 5, 0, 5, 10, 15, . . ..
Then 3 + 5Z = 3 + 5z ; z Z = . . . , 2, 3, 8, 13, 18, . . .. Note that Z is
abelian, so /+ 3 = 3 +/.
(b) Let ( = S
3
=
_
e, (12), (13), (23), (123), (132)
_
, and let / =
_
e, (12)
_
. Then
(123) / =
_
(123), (123)(12)
_
=
_
(123), (13)
_
, but
/ (123) =
_
(123), (12)(123)
_
=
_
(123), (23)
_
, so (123) / ,= / (123).
(c) Again, let ( = S
3
, and let / = A
3
=
_
e, (123), (132)
_
. Then
(12) A
3
=
_
(12), (12)(123), (12)(132)
_
=
_
(12), (23), (13)
_
.
Also, (12) A
3
=
_
(12), (123)(12), (132)(12)
_
=
_
(12), (13), (23)
_
,
so that (12)A
3
= A
3
(12).
(d) Let A and be groups, and consider the product group
( = A = (x, y) ; x A and y .
3
4 CHAPTER 1. HOMOMORPHISMS
Let / = (e
X
, y) ; y = e
X
. Then for any (x, y
1
) (,
(x, y
1
)/ = (x, y
1
) (e
X
, y
2
) ; y
2
= (x, y
1
y
2
) ; y = (x, y) ; y
= x .
Lemma 2 Let ( be a group and let / < ( be a subgroup.
(a) For any g (,
_
g/ = /
_

_
g /
_

_
/g = /
_
.
(b) For any g, h (, the following are equivalent:
_
g h/
_

_
g/ = h/
_

_
h g/
_

_
h
1
g /
_

_
g
1
h /
_
Likewise, the following are equivalent:
_
g /h
_

_
/g = /h
_

_
h /g
_

_
hg
1
/
_

_
gh
1
/
_
Proof: Exercise 1 2
Example 3:
(a) Let ( = (Z, +), and let / = 5Z as in Example 1a). Then
10 + 5Z = 10 + 5z ; z Z = . . . , 5, 10, 15, 20, 25, . . . = 5Z.
(b) Let ( = S
3
and let / = A
3
=
_
e, (123), (132)
_
, as in Example 1c). Then
(123) A
3
=
_
(123), (132), e
_
= A
3
.
(c) Let A and be groups; let ( = A and let / = e
X
, as in Example 1d).
Then for any y , (e
G
, y) / = e
X
= /.
The left coset space of / is the set of all its left cosets:
(// = g/ ; g (.
Example 4:
(a) Let ( = (Z, +), and let / = 5Z as in Example 1a). Then
Z
5Z
= 5Z, 1 + 5Z, 2 + 5Z, 3 + 5Z, 4 + 5Z.
1.2. LAGRANGES THEOREM 5
(b) Let ( = S
3
, and let / =
_
e, (12)
_
, as in Example (1b). Then
S
3
/
=
_
/, (123)/, (23)/
_
=
_
e, (12), (123), (13), (23), (132)
_
.
(c) Let ( = S
3
, and let / = A
3
=
_
e, (123), (132)
_
, as in Example 1c). Then
S
3
A
3
=
_
A
3
, (12)A
3
_
=
_
e, (123), (132) , (12), (23), (13)
_
.
(d) Let A and be groups; let ( = A , and let / = e
X
, as in Example 1d).
Then
(
/
= /
x
; x A, where, for any xed x A, /
x
= (x, y) ; y .
1.2 Lagranges Theorem
Prerequisites: 1.1
Let o
1
, o
2
, . . . , o
N
( be a collection of subsets of (. We say that ( is a disjoint union
of o
1
, . . . , o
N
if:
1. ( = o
1
o
2
. . . o
N
.
2. o
n
o
m
= , whenever n ,= m.
We then write: ( = o
1
. o
2
. . . . . o
N
, or ( =
N
_
n=1
o
n
. We say that the collection
S = o
1
, o
2
, . . . , o
N
is a partition of (.
More generally, let S be any (possibly innite) collection of subsets of (. We say that ( is
a disjoint union of the elements in S if:
1. ( =
_
SS
o.
2. o o

= , whenever o and o

are distinct elements of S.

We then write: ( =
_
SS
o. We say that S is a partition of (.
Lemma 5 Let ( be a group and / a subgroup. Then (// is a partition of (.
In other words:
1. Any two cosets of / are either identical or disjoint: for any g
1
, g
2
(, either g
1
/ = g
2
/,
or (g
1
/) (g
2
/) = .
6 CHAPTER 1. HOMOMORPHISMS
2. ( =
_
(gA) G/A
(g/).
Proof: Exercise 2 Hint: Apply Lemma 2(b) 2
Example 6:
(a) Let ( = (Z, +), and let / = 5Z as in Example 4a). Then
Z = . . . , 5, 0, 5, 10, . . . . . . . , 4, 1, 6, 11, . . . . . . . , 3, 2, 7, 12, . . .
. . . . , 2, 3, 8, 13, . . . . . . . , 1, 4, 9, 14, . . .
= (5Z) . (1 + 5Z) . (2 + 5Z) . (3 + 5Z) . (4 + 5Z) =
_
(n+5Z) Z/5Z
(n + 5Z).
(b) Let ( = S
3
, and let / =
_
e, (12)
_
, as in Example (4b). Then
S
3
=
_
e, (12)
_
.
_
(123), (13)
_
.
_
(23), (132)
_
= / . (123)/ . (23)/
=
_
(gA) S
3
/A
(g/).
(c) Let A and be groups; let ( = A , and let / = e
X
, as in Example 4d). For
any xed x A, let /
x
= (x, y) ; y . Then ( =
_
xX
/
x
=
_
(x,y)A G/A
(x, y)/.
If (// is a nite set, then the index of / in ( is the cardinality of (//. It is sometimes
denoted by [( : /[.
Corollary 7 Lagranges Theorem
Let ( be a nite group and let / be any subgroup. Then:
1. [([ = [/[

(
/

.
2. In particular, [/[ divides [([.
Proof:
Claim 1: For any g (, [g/[ = [/[.
Proof: Dene a map : /g/ by: (a) = ga. It is Exercise 3 to verify that is a
bijection. Thus, [g/[ = [/[. ........................................ 2 [Claim 1]
1.3. NORMAL SUBGROUPS & QUOTIENT GROUPS 7
Now, Lemma 5 says that the cosets of / partition (. Since ( is nite, we know that there
are a nite number of distinct cosets for / lets say they are (g
1
/), (g
2
/), . . . , (g
N
/),
for some elements g
1
, g
2
, . . . , g
N
(. Thus,
( = (g
1
/) . (g
2
/) . . . . . (g
N
/),
therefore, [([ = [g
1
/[ + [g
2
/[ + . . . + [g
N
/[.
Clm.1
[/[ + [/[ + . . . + [/[
. .
N
= N [/[,
where N is the number of distinct cosetsthat is, N = [(//[. 2
1.3 Normal Subgroups & Quotient Groups
Prerequisites: 1.1; Homomorphisms
Let / < ( be a subgroup. We say that / is normal if g/ = /g for every g (. We
indicate this by writing /(.
Example 8:
(a) Let ( = Z and let / = 5Z, as in Example 4a. Then / is normal in Z, because for any
n Z, n + 5Z = n + 5z ; z Z = 5z + n ; z Z = 5Z + n.
(b) In general, if ( is any abelian group, then every subgroup is normal (Exercise 4).
(c) Let ( = S
3
, and let / =
_
e, (12)
_
, as in Example 1b). Then / is not normal in S
3
,
because (123) / ,= / (123).
(d) Let ( = S
3
, and let / = A
3
, as in Example 4c). Then A
3
is normal in S
3
, because
the only nontrivial left coset of A
3
is (12)A
3
, and the only nontrivial right coset of A
3
is
A
3
(12), and we saw in Example (1c) that (12)A
3
= A
3
(12).
(e) Let A and be groups; let ( = A , and let / = e
X
, as in Example 4d).
Then / is normal, because for any (x, y) (, (x, y) / = /
x
= / (x, y).
If /, B ( are subsets, then their product is the set /B = ab ; a / and b B. In
particular, if (g/) and (h/) are two cosets of /, then their product is the set:
(g/)(h/) = ga
1
ha
2
; a
1
, a
2
/. (1.1)
Example 9:
8 CHAPTER 1. HOMOMORPHISMS
(a) Let ( = (Z, +) and let / = 5Z, as in Example (4a). Let g = 1 and h = 2. Then
1 + 5Z = . . . , 9, 4, 1, 6, 11, 16, . . .
and 2 + 5Z = . . . , 8, 3, 2, 7, 12, 17, . . .,
so (1 + 5Z) + (2 + 5Z) = . . . , 7, 2, 3, 8, 13, 18, . . . = (3 + 5Z).
(b) Let A and be groups; let ( = A , and let / = e
X
, as in Example 4d).
Then for any (x
1
, y
1
) and (x
2
, y
2
) in (,
(x
1
, y
1
) / = (x
1
, y

) ; y

,
and (x
2
, y
2
) / = (x
2
, y

) ; y

,
so that (x
1
, y
1
) / (x
2
, y
2
) / = (x
1
, y

) (x
2
, y

) ; y

and y

=
__
(x
1
x
2
), (y

)
_
; y

and y

_
=
__
(x
1
x
2
), y
_
; y
_
=
_
x
1
x
2
, e
Y
_
/,
where e
Y
is the identity element in . Thus, the multiplication of cosets in (// mimics
the multiplication of elements in A.
Lemma 10 Let ( be a group.
(a) Subset multiplication in ( is associative. That is, for any subsets /, B, c (,
/ (B c) = (/ B) c.
(b) If / < ( is a subgroup of (, then / / = /.
Proof: Exercise 5 2
If : (H is a group homomorphism, recall that the kernel of is the subgroup:
ker() = g ( ; (g) = e
H
.
Proposition 11 Let ( be a group, and let / < ( be a subgroup. The following are
equivalent:
(a) / = ker() for some group homomorphism : (H (where H is some group).
(b) For any g (, g/g
1
= /.
(c) /(.
1.3. NORMAL SUBGROUPS & QUOTIENT GROUPS 9
(d) The coset space (//is a group under the multiplication operation (1.1). Furthermore:
1. For any g, h (, (g/)(h/) = (gh)/.
2. Dene : ((// by: (g) = g/. Then is a group epimorphism, and
ker() = /.
Proof: (a)=(b) Suppose / = ker(), and let g (. Then g/g
1
= gag
1
; a /,
and we want to show this set is just /.
Claim 1: g/g
1
/.
Proof: Let gag
1
be some element of g/g
1
. Then (gag
1
) = (g) (a) (g
1
) =
(g) e
H
(g)
1
= (g) (g)
1
= e
H
. Thus, gag
1
/. .......... 2 [Claim 1]
Claim 2: / g/g
1
.
Proof: It follows from Claim 1 that g
1
/g / (just use g
1
in place of g). Thus,
g(g
1
/g)g
1
g/g
1
. But g(g
1
/g)g
1
= (gg
1
)/(gg
1
) = e
G
/e
G
= /. In other
words, / g/g
1
. ................................................. 2 [Claim 2]
It follows from Claims 1 and 2 that g/g
1
= /.
(b)=(c) Let g (. Then g/ = g/(gg
1
) = (g/g
1
)g
(b)
/g. This holds for any
g (, so /(.
(c)=(d) We will prove this in several steps...
Claim 3: (g/)(h/) = (gh)/.
Proof: (g/)(h/)
Lem.10(a)
g(/h)/
AG
g(h/)/
Lem.10(a)
(gh)//
Lem.10(b)
(gh)/.
2 [Claim 3]
Claim 4: (// is a group.
Proof: It follows from Claim 3 that (// is closed under multiplication. By Lemma 10(a),
this multiplication is associative. By applying Claim 3, it is easy to verify:
1. The identity element of (// is just the coset e
G
/ = /.
2. For any g (, the inverse of the coset g/ is just the coset g
1
/.
Thus, (// satises all the properties of a group ....................... 2 [Claim 4]
Claim 5: : ((// is a homomorphism.
Proof: (g h) = (gh) /
Clm.3
(g/) (h/) = (g) (h). .. 2 [Claim 5]
Also, is surjective: any element of (// is a coset of the form g/ for some g (, and thus,
g/ = (g). Thus, is an epimorphism.
Claim 6: ker() = /.
10 CHAPTER 1. HOMOMORPHISMS
Proof: Let g (. Then
_
g ker()
_

_
(g) = e
G/A
_

_
g/ = /
_
Lem.2(a)
_
g /
_
. ............................................... 2 [Claim 6]
(d)=(a) This follows immediately: let H = (// and let = . 2
The group (// is called the quotient group, and the epimorphism : ((// is called
the projection map or quotient map.
1.4 The Fundamental Isomorphism Theorem
Prerequisites: 1.3
Theorem 12 Fundamental Isomorphism Theorem
Let ( and H be groups, and let : (H be a
group homomorphism, with image 1 = (() H,
and kernel /. Then:
(a) 1

= (//.
(b) For any g ( with h = (g), the -
preimage of h is the g-coset of /. That is:

1
h = g/. 2
g
1
g
2
g
1g
2
h
1
h
2 e
e
=
k
e
r
(

)
Example 13:
(a) Let ( = Z and H = Z
/3
, and let : Z n [n]
3
Z
/3
(see Figure 1.1a). Then / = 3Z,
so 1 = Z
/3
= Z/(3Z) = (//.
Observe: (5) = [5]
3
= [2]
3
, and thus,
1
[2]
3
= 5+3Z = . . . , 1, 2, 5, 8, 11, . . ..
(b) Let / and B be groups, and let ( = / B (see Figure 1.1b). Let : (/ be the
projection map; that is, (a, b) = a. Then
ker() = (e
A
, b) ; b B = e
A
B
is a subgroup isomorphic to B. Thus, the Fundamental Isomorphism Theorem says:
/ B
e
A
B
=
(
ker()

= /.
For any xed (a, b) (, observe that (a, b) = a. Thus,

1
a = (a, b) (e
A
, b

) ; b

B = (a, b

) ; b

B.
1.4. THE FUNDAMENTAL ISOMORPHISM THEOREM 11
6
3
0
3
6
0
Z
Z
/3
5
2
1
4
7
2
4
1
2
5
1
8
8
9
7
Ker() = 3 Z = {...-6, -3, 0, 3, 6, 9, ...}
:
Z Z
/3
n [n]
3
e
e

(a) (b)
Figure 1.1: Examples (13a) and (13b)
(c) Let ( = S
3
, and let / = A
3
, as in Example 8d). Then A
3
is normal in S
3
, and
S
3
/A
3
= A
3
, (12)A
3
is a two-element group, isomorphic to (Z
/2
, +), via the map
: Z
/2
S
3
/A
3
dened: (

0) = A
3
and (

1) = (12)A
3
.
Proof of Theorem 129: (a) We will build an explicit isomorphism between 1 and (//.
Claim 1: For any g
1
, g
2
(,
_
g
1
/ = g
2
/
_

_
(g
1
) = (g
2
)
_
.
Proof:
_
g
1
/ = g
2
/
_
Lem.2(b)
_
g
1
2
g
1
/
_

_
(g
1
2
g
1
) = e
H
_

_
(g
2
)
1
(g
1
) = e
H
_

_
(g
1
) = (g
2
)
_
. .... 2 [Claim 1]
Dene the map : (//1 by: (g/) = (g).
Claim 2: is well-dened and injective.
Proof: For any g
1
, g
2
(,
_
g
1
/ = g
2
/
_
Clm.1
_
(g
1
) = (g
2
)
_

_
(g
1
/) = (g
2
/)
_
. ............................................. 2 [Claim 2]
12 CHAPTER 1. HOMOMORPHISMS
Claim 3: is surjective.
Proof: By denition, any i 1 is the image of some g ( ie. (g) = i. Thus,
(g/) = i. ........................................................ 2 [Claim 3]
Claim 4: is a homomorphism.
Proof: For any cosets g
1
/ and g
2
/ in (//,

_
(g
1
/)(g
2
/)
_
()

_
(g
1
g
2
)/
_
= (g
1
g
2
) = (g
1
)(g
2
) = (g
1
/)(g
2
/),
where () follows from Proposition 11(d) part 1 ........................ 2 [Claim 4]
(b) Let g

(. Then:
_
g

g/
_
Lem2(b)
_
g

/ = g/
_
Clm.1
_
(g

) = (g) = h
_

_
g

1
h
_
. 2
Corollary 14 Let : (H be a group homomorphism. Then
_
is injective
_

_
ker = e
G

_
.
Proof: Exercise 6 2
Chapter 2
The Isomorphism Theorems
2.1 The Diamond Isomorphism Theorem
Prerequisites: 1.4
Let ( be a group and B < ( a subgroup. The normalizer of B in ( is the subgroup
^
rmlzr
G
(B) = g ( ; gB = Bg.
Thus, ^
rmlzr
G
(B) is the set of all elements of ( who think that B is normal.
Lemma 15 Let ( be a group with subgroup B < (. Then:
(a) B < ^
rmlzr
G
(B) < (.
(b)
_
B (
_

_
^
rmlzr
G
(B) = (
_
.
Proof: Exercise 7 2
Thus, the condition / ^
rmlzr
G
(B) is true automatically if B is a normal subgroup of (.
Recall that / B = a b ; a B, b /. In general, /B is not a subgroup of (. But /B
is a subgroup under the conditions of the following theorem...
13
14 CHAPTER 2. THE ISOMORPHISM THEOREMS
Theorem 16 Diamond Isomorphism Theorem
Let ( be a group, with subgroups / and B. Suppose that
/ ^
rmlzr
G
(B). Then:
(a) / B is a subgroup of (.
(b) B (/ B).
(c) (/ B) /.
(d) There is an isomorphism:
/ B
B

=
/
/ B
given by the map
:
/ B
B

/
/ B
(ab)B a (/ B)
O
Proof: (a) Let a
1
b
1
and a
2
b
2
be elements of /B. We must show that a
1
b
1
(a
2
b
2
)
1
is also
in /B. But
a
1
b
1
(a
2
b
2
)
1
= a
1
b
1
b
1
2
a
1
2
= a
1
(b
1
b
1
2
)a
1
2
()
a
1
a
1
2
b
3
= (a
1
a
1
2
)b
3
/B.
To see (), recall that / ^
rmlzr
G
(B), so Ba
1
2
= a
1
2
B. Thus (b
1
b
1
2
)a
1
2
= a
1
2
b
3
for some
b
3
B.
(b) and (c) Exercise 8 .
(d) [In progress...]
Claim 1: For any a / and b B, (ab)B = aB
2
Example 17:
(a) Let A, , and Z be groups, and let ( = A Z, as shown in Figure 2.1. Dene:
/ = A e
Y
Z =
_
(x, e
Y
, z) ; x A, z Z
_
,
and B = e
X
Z = (e
X
, y, z) ; y , z Z.
Then:
/B = A Z = (,
and / B = e
X
e
Y
Z =
_
(e
X
, e
Y
, z) ; z Z
_
.
Thus,
/B
B
=
A Z
e
X
Z

= A

=
A e
Y
Z
e
X
e
Y
Z
=
/
/ B
.
2.2. THE CHAIN ISOMORPHISM THEOREM 15
Figure 2.1: Example 17a).
2.2 The Chain Isomorphism Theorem
Prerequisites: 1.4
Let ( be a group, with normal subgroup / (. Suppose / < B < (. Then / is also a
normal subgroup of B, and the quotient group
B
/
= b/ ; b B
is a subset of the quotient group
(
/
= g/ ; g (.
Theorem 18 Chain Isomorphism Theorem
Let ( be a group, with normal subgroups /( and B (. Suppose / < B. Then:
(a)
B
/
is a normal subgroup of
(
/
.
(b) There is an isomorphism
((//)
(B//)

=
(
B
.
(c) Use bar notation to denote elements of (//. Thus, g/ = g, B// = B, (// = (,
and
(//
B//
= (/B. Then the isomorphism : (/B(/B is dened:
_
gB
_
= gB.
16 CHAPTER 2. THE ISOMORPHISM THEOREMS
mod
O O
O
O
mod
mod
m
o
d
m
o
d
m
o
d
isomorphism
isomorphism
m
o
d
Figure 2.2: The Chain Isomorphism Theorem
0
0
0
0
Figure 2.3: Example 19a).
2.3. THE LATTICE ISOMORPHISM THEOREM 17
Example 19:
(a) Let A, , and Z be groups, and let ( = A Z. Dene:
/ = A e
Y
e
Z
=
_
(x, e
Y
, e
Z
) ; x A
_
,
and B = A e
Z
= (x, y, e
Z
) ; x Z, y .
Then:
(
/
=
A Z
A e
Y
e
Z

= Z;
B
/
=
A e
Z

A e
Y
e
Z

= e
Z
;
and
(
B
=
A Z
A e
Z

= Z.
Thus,
((//)
(B//)

=
Z
e
Z

= Z

=
(
B
.
2.3 The Lattice Isomorphism Theorem
Prerequisites: 2.2
If ( is a group, recall that the subgroup lattice of ( is a directed graph L((), whose vertices
are the subgroups of (. We draw a (directed) edge from subgroup / to subgroup B if / < B
and there is no c such that / < c < B. The graph is drawn on paper so that / appears below
B on the page if and only if / < B.
This graph is called a lattice because it has two special properties:
1. For any subgroups /, B < (, there is a minimal subgroup of ( which contains both /
and B namely, the join of / and B:
/, B) = a
1
b
1
a
2
b
2
. . . a
n
b
n
; n N, a
1
, a
2
, . . . , a
n
/, and b
1
, b
2
, . . . , b
n
B.
2. For any subgroups /, B < (, there is a maximal subgroup of ( which is contained in both
/ and B namely, their intersection / B.
Let ^ (. We will adopt the bar notation for objects in the quotient group (/^. Thus,
( = (/^. If g (, then g = g^ (. If ^ < / < (, then / = //^ = a^ ; a / (.
Theorem 20 Lattice Isomorphism Theorem
Let ( be a group and let ^ ( be a normal subgroup. Let ( = (/^. Let L(() be the
subgroup lattice of (, and let L
N
(() be the fragment of L(() consisting of all subgroups which
contain ^. That is:
L
N
(() = / < ( ; ^ < /.
18 CHAPTER 2. THE ISOMORPHISM THEOREMS
N
0
G
B
A
C
D
<C,D>
C D
G
B
A
C
D
<C,D>
C D
{e}
L(G) L(G)
Figure 2.4: The Lattice Isomorphism Theorem
Then there is an order-preserving bijection from L
N
(() into L((), given:
L
N
(() / / L(().
Furthermore, for any /, B, c, T L
N
((),
(a)
_
/ < B
_

_
/ < B
_
, and in this case, [B : /[ = [B : /[.
(b) c, T) =
_
c, T
_
.
(c) c T = c T.
(d)
_
/(
_

_
/(
_
, and in this case, (//

= (//.
Proof: (d) just restates the Chain Isomorphism Theorem. The proofs of (a,b,c) are
Exercise 9 . 2
Chapter 3
Free Abelian Groups
3.1 Rank and Linear Independence
Let (/, +) be an (additive) abelian group and let a
1
, . . . , a
R
/. We say that a
1
, a
2
, . . . , a
R
are Z-linearly independent if there exist no z
1
, z
2
, . . . , z
R
Z (not all zero) such that z
1
a
1
+
z
2
a
2
+. . . +z
R
a
R
= 0. This is a natural generalization of the notion of linear independence for
vector spaces. The rank of / is the maximal cardinality of any linearly independent subset.
In other words:
_
rank (/) = R
_

_
_
_
1. There exists a linearly independent set a
1
, a
2
, . . . , a
R
/.
2. For any b
1
, b
2
, . . . , b
R
, b
R+1
/, there exist z
1
, z
2
, . . . , z
R
, z
R+1
Z
(not all zero) so that z
1
b
1
+ z
2
b
2
+ . . . + z
R
b
R
+ z
R+1
b
R+1
= 0.
_
_
_
Think of rank as a notion of dimension for abelian groups.
Example 21:
(a) rank (Z, +) = 1. First, well show that rank (Z) 1. To see this, let a Z be nonzero.
Then z a ,= 0 for all z Z, so the set a is Z-linearly independent.
Next, well show that rank (Z) 1. Suppose a
1
, a
2
Q were nonzero. Let z
1
= a
2
and
z
2
= a
1
. Then z
1
a
1
+ z
2
a
2
= a
2
a
1
a
1
a
2
= 0. So any set a
1
, a
2
with two elements
is linearly dependent.
(b) rank (Q, +) = 1.
First, well show that rank (Q) 1. To see this, let q Q be nonzero. Then z q ,= 0 for
all z Z, so the set q is Z-linearly independent.
Next, well show that rank (Q) 1. Suppose q
1
, q
2
Q were nonzero; let q
1
=
r
1
s
1
and
q
2
=
r
2
s
2
. Let z
1
= r
2
s
1
and z
2
= r
1
s
2
. Then
z
1
q
1
+ z
2
q
2
= r
2
s
1
r
1
s
1
r
1
s
2
r
2
s
2
= r
2
r
1
r
1
r
2
= 0.
19
20 CHAPTER 3. FREE ABELIAN GROUPS
(1,0)
(
0
,
1
)
(0, )
(1,0)
(
0
,
1
)
(1, )
(1,0)
(
0
,
1
)
(2, )
(1,0)
(
0
,
1
)
(3, )
(1,0)
(
0
,
1
)
(4, )
(1,0)
(
0
,
1
)
(-1, )
(1,0)
(
0
,
1
)
(-2, )
(1,0)
(
0
,
1
)
(
0
,
1
)
(
0
,
1
)
(
0
,
1
)
(
0
,
1
)
(
0
,
1
)
(
0
,
1
)
(1,0)
(
0
,
1
)
(0, )
(1,0)
(
0
,
1
)
(1, )
(1,0)
(
0
,
1
)
(2, )
(1,0)
(
0
,
1
)
(3, )
(1,0)
(
0
,
1
)
(4, )
(1,0)
(
0
,
1
)
(-1, )
(1,0)
(
0
,
1
)
(-2, )
(1,0)
(1,0)
(
0
,
1
)
(0, )
(1,0)
(
0
,
1
)
(1, )
(1,0)
(
0
,
1
)
(2, )
(1,0)
(
0
,
1
)
(3, )
(1,0)
(
0
,
1
)
(4, )
(1,0)
(
0
,
1
)
(-1, )
(1,0)
(
0
,
1
)
(-2, )
(1,0)
(1,0)
(
0
,
1
)
(0, )
(1,0)
(
0
,
1
)
(1, )
(1,0)
(
0
,
1
)
(2, )
(1,0)
(
0
,
1
)
(3, )
(1,0)
(
0
,
1
)
(4, )
(1,0)
(
0
,
1
)
(-1, )
(1,0)
(
0
,
1
)
(-2, )
(1,0)
(1,0)
(
0
,
1
)
(0, )
(1,0)
(
0
,
1
)
(1, )
(1,0)
(
0
,
1
)
(2, )
(1,0)
(
0
,
1
)
(3, )
(1,0)
(
0
,
1
)
(4, )
(1,0)
(
0
,
1
)
(-1, )
(1,0)
(
0
,
1
)
(-2, )
(1,0)
0
1
2
3
-1
0
1
2
3
-1
0
1
2
3
-1
0
1
2
3
-1
0
1
2
3
-1
0
1
2
3
-1
0
1
2
3
-1
Figure 3.1: Z
2
= Z Z is the free group of rank 2.
(c) rank
_
Z
/n
_
= 0.
To see this, note that n a = 0 for any a Z
/n
. Thus, even a set like a, containing only
one element, is not Z-linearly independent.
(d) If / is any nite abelian group, then rank (/) = 0.
To see this, suppose [/[ = n. Then n a = 0 for any a /. Thus, even a set like a,
containing only one element, is not Z-linearly independent.
(e) rank (R, +) = .
We need to construct an innite, Z-linearly independent subset r
1
, r
2
, r
3
, . . . R. This
is Exercise 10 .
3.2 Free Groups and Generators
Prerequisites: 3.1, 4.1
Let R N. The free abelian group of rank R is the group
Z
R
= (z
1
, z
2
, . . . , z
R
) ; z
1
, z
2
, . . . , z
R
Z, (with componentwise addition)
More generally, any abelian group is called free if it is isomorphic to Z
R
.
Lemma 22 rank
_
Z
R
_
= R. Furthermore, if B < Z
R
is any subgroup, then rank (B) R.
3.2. FREE GROUPS AND GENERATORS 21
Proof: I claim rank
_
Z
R
_
R. To see this, let a
1
= (1, 0, 0, . . . , 0), a
2
= (0, 1, 0, . . . , 0), . . . , a
R
=
(0, 0, . . . , 0, 1). Then a
1
, . . . , a
R
are Z-linearly independent (Exercise 11).
Claim 1: If B < Z
R
, then rank (B) R.
Proof: Suppose b
1
, b
2
, . . . , b
R
, b
R+1
B Z
R
. Think of Z
R
as a subset of R
R
. Then any
set of R+1 vectors cannot be R-linearly independent, so there are numbers q
1
, . . . , q
R+1
such
that q
1
b
1
+. . .+q
R+1
b
R+1
= 0. Since b
1
, . . . , b
R+1
have integer coecients, we can assume
that q
1
, . . . , q
R+1
are rational numbers; say q
r
=
pr
mr
for all r. Now, let M = m
1
m
2
. . . m
R
,
and let z
r
= M q
r
= m
1
m
2
. . . m
r1
p
r
m
r+1
. . . m
R
. Then z
1
, . . . , z
R+1
are integers, and
z
1
b
1
+ . . . + z
R+1
b
R+1
= M (q
1
b
1
+ . . . + q
R+1
b
R+1
) = M 0 = 0. . 2 [Claim 1]
Setting B = Z
R
in Claim 1 tells us that rank
_
Z
R
_
R. Thus, rank
_
Z
R
_
= R. 2
Lemma 23 Let / be any abelian group. Then rank
_
Z
R
/
_
= R + rank (/).
Proof: Exercise 12 2
Example 24: rank
_
Z
5
Z
/16
Z
/7
_
= rank (Z
5
) + rank
_
Z
/16
Z
/7
_
= 5 + 0 = 5.
Lemma 25 Let / and B be free abelian groups. Then:
(a) /B is also a free abelian group.
(b) rank (/B) = rank (/) + rank (B).
Proof: Suppose rank (/) = R and rank (B) = S. Thus, /

= Z
R
, and B

= Z
S
. Thus,
/B

= Z
R
Z
S
= Z
R+S
. 2
Example 26: Z
3
Z
5
= Z
8
.
If (/, +) is an (additive) abelian group, and g
1
, g
2
, . . . , g
R
/, then we say that / is
generated by g
1
, . . . , g
R
if, for any a /, we can nd integers z
1
, z
2
, . . . , z
R
Z so that
a = z
1
g
1
+ z
2
g
2
+ . . . + z
R
g
R
. (3.1)
In this case, we write / = g
1
, g
2
, . . . , g
R
), or, inspired by eqn. (3.1), we write:
/ = Zg
1
+Zg
2
+ . . . +Zg
R
If / has a nite generating set, we say that / is nitely generated.
22 CHAPTER 3. FREE ABELIAN GROUPS
Proposition 27 Let / be an abelian group, and let g
1
, g
2
, . . . , g
R
/. The following are
equivalent:
(a) / = Zg
1
+Zg
2
+ . . . +Zg
R
, and g
1
, g
2
, . . . , g
R
are Z-linearly independent
(b) / = Zg
1
Zg
2
. . . Zg
R
(where Zg
r
is the cyclic subgroup generated by g
r
).
(c) For any a /, there are unique integers z
1
, z
2
, . . . , z
R
Z so that
a = z
1
g
1
+ z
2
g
2
+ . . . + z
R
g
R
.
(d) / is isomorphic to Z
R
, via the mapping : Z
R
/ dened:
(z
1
, z
2
, . . . , z
R
) = z
1
g
1
+ z
2
g
2
+ . . . + z
R
g
R
Proof: (Exercise 13) 2
If the conditions of Proposition 27 are satised, we say that / is the free abelian group
generated by g
1
, g
2
, . . . , g
R
, and we call g
1
, g
2
, . . . , g
R
a basis for /.
3.3 Universal Properties of Free Abelian Groups
Prerequisites: 3.2
Free abelian groups are called free because they are the most structureless of all abelian
groups. It is thus very easy to construct epimorphisms from free abelian groups into any other
abelian group. This endows the free groups with certain universal properties....
Proposition 28 Universal Mapping Property of Free Abelian Groups
Let / be an abelian group, and let g
1
, g
2
, . . . , g
R
/. Dene the function : Z
R
/ by
(z
1
, z
2
, . . . , z
R
) = z
1
g
1
+ z
2
g
2
+ . . . + z
R
g
R
. Then:
(a) is always a homomorphism (regardless of the choice of g
1
, . . . , g
R
).
(b) Every homomorphism from Z
R
into / has this form.
(c) If g
1
, g
2
, . . . , g
R
generate /, then is an epimorphism.
Proof: (Exercise 14) 2
Example 29:
(a) Let / = Z
/n
, and recall that Z is the free abelian group of rank 1. Dene : ZZ
/n
by (z) = z

1 = z. Then is an epimorphism.
3.4. () HOMOLOGICAL PROPERTIES OF FREE ABELIAN GROUPS 23
(b) Let / be any abelian group and let a / be any element. Dene : Z/ by
(z) = z a. Then is a homomorphism, whose image is the cyclic subgroup a).
(c) Let / = Z
/5
Z
/7
, and dene : Z
2
/ by (z
1
, z
2
) = (z
1
[1]
5
, z
2
[1]
7
) =
([z
1
]
5
, [z
2
]
7
) (where [n]
5
is the congruence class of n, mod 5, etc.) Then is an epi-
morphism.
Corollary 30 Universal Covering Property of Free Abelian Groups
Any nitely generated abelian group is a quotient of a (nitely generated) free abelian group.
Proof: (Exercise 15) 2
This is called the covering property because any abelian group can be covered by the
projection of some free abelian group. Thus, to prove some result about all abelian groups, it
is often sucient to prove the result only for free abelian groups.
3.4 () Homological Properties of Free Abelian Groups
Prerequisites: 3.3, Homomorphism Groups
The next two properties are not important to us at present, but arise frequently in homo-
logical algebra.
Corollary 31 Projective Property of Free Abelian Groups
Suppose / and B are abelian groups, and there are homomorphisms : Z
R
/ and
: B/. Then there is a homomorphism : Z
R
B so that = .
Z
R
B

/
H
H
H
H
H
H
H
H
H
H
Hj
?
=====
Z
R
B

H
H
H
H
H
H
H
H
H
H
Hj
?
In other words, given any diagram like the one on the left, we can always complete it to get
a commuting diagram like the one on the right.
Proof: Proposition 28(c) says we can nd some elements a
1
, . . . , a
R
/ so that has the
form: (z
1
, . . . , z
R
) = z
1
a
1
+ . . . + z
R
a
R
. Now, : B/ is a surjection, so for each r,
nd some b
r
B with (b
r
) = a
r
. Now, dene the map : Z
R
B by (z
1
, . . . , z
R
) =
z
1
b
1
+ . . . + z
R
b
R
. Proposition 28(a) says is a homomorphism. To see that = ,
observe that (z
1
, . . . , z
R
) = (z
1
b
1
+. . . +z
R
b
R
) = z
1
(b
1
)+. . . +z
R
(b
R
) =
z
1
a
1
+ . . . + z
R
a
R
= (z
1
, . . . , z
R
). 2
24 CHAPTER 3. FREE ABELIAN GROUPS
Proposition 32 Adjoint Property
Let / be an abelian group, and let Hom
_
Z
R
, /
_
be the group of all homomorphisms from
Z
R
into /. Then Hom
_
Z
R
, /
_
is isomorphic to /
R
, via the map
: /
R
Hom
_
Z
R
, /
_
sending (a
1
, . . . , a
R
) to the morphism : Z
R
/ such that (z
1
, . . . , z
R
) = z
1
a
1
+. . . +z
R
a
R
.
Proof: is a homomorphism: Let a = (a
1
, . . . , a
R
) /
R
and b = (b
1
, . . . , b
R
) /
R
.
Let = (a) and = (b). Thus, , : Z
R
/ are homomomorphisms, such that
(z
1
, . . . , z
R
) = z
1
a
1
+ . . . + z
R
a
R
, and (z
1
, . . . , z
R
) = z
1
b
1
+ . . . + z
R
b
R
. We want to show
that (a+b) = +. But a+b = (a
1
+b
1
, . . . , a
R
+b
R
), so (a+b) is the homomorphism
(z
1
, . . . , z
R
) = z
1
(a
1
+b
1
)+. . .+z
R
(a
R
+b
R
) = (z
1
a
1
+. . .+z
R
a
R
)+(z
1
b
1
+. . .+z
R
b
R
) =
(z
1
, . . . , z
R
) + (z
1
, . . . , z
R
).
is surjective: This is just Proposition 28(b).
is injective: This is Exercise 16 . 2
3.5 Subgroups of Free Abelian Groups
Prerequisites: 3.2
Any subgroup of a free abelian group is also free. Furthermore, there is a common basis
for both the group and its subgroup...
Proposition 33 Let / be a free abelian group, and let B < / be any subgroup. Then:
(a) B is also a free abelian group, of rank S R.
(b) There exists a basis a
1
, a
2
, . . . , a
R
for /, and numbers m
1
, m
2
, . . . , m
S
N, so that,
if b
s
= m
s
a
s
for all s [1..S], then b
1
, b
2
, . . . , b
S
is a basis for B.
Example 34:
(a) As shown in Figure 3.2, Let / = Z
2
, and let B be the subgroup generated by x = (2, 0)
and y = (1, 2). We can see that B = (Zx) (Zy), so B is a free abelian group. However,
we want to nd a common basis for / and B.
As shown in Figure 3.3, let a
1
= (1, 2) and a
2
= (0, 1). Let m
1
= 1 and m
2
= 4, so
that b
1
= m
1
a
1
= (1, 2) and b
2
= m
2
a
2
= (0, 4). Then / = (Za
1
) (Za
2
) and
B = (Zb
1
) (Zb
2
).
3.5. SUBGROUPS OF FREE ABELIAN GROUPS 25
(
1
,
2
)
(1,0)
(
0
,
1
)
(0, )
(1,0)
(
0
,
1
)
(1, )
(1,0)
(
0
,
1
)
(2, )
(1,0)
(
0
,
1
)
(3, )
(1,0)
(
0
,
1
)
(4, )
(1,0)
(
0
,
1
)
(-1, )
(1,0)
(
0
,
1
)
(-2, )
(1,0)
(
0
,
1
)
(
0
,
1
)
(
0
,
1
)
(
0
,
1
)
(
0
,
1
)
(
0
,
1
)
(
0
,
1
)
(1,0)
(
0
,
1
)
(0, )
(1,0)
(
0
,
1
)
(1, )
(1,0)
(
0
,
1
)
(2, )
(1,0)
(
0
,
1
)
(3, )
(1,0)
(
0
,
1
)
(4, )
(1,0)
(
0
,
1
)
(-1, )
(1,0)
(
0
,
1
)
(-2, )
(1,0)
(1,0)
(
0
,
1
)
(0, )
(1,0)
(
0
,
1
)
(1, )
(1,0)
(
0
,
1
)
(2, )
(1,0)
(
0
,
1
)
(3, )
(1,0)
(
0
,
1
)
(4, )
(1,0)
(
0
,
1
)
(-1, )
(1,0)
(
0
,
1
)
(-2, )
(1,0)
(1,0)
(
0
,
1
)
(0, )
(1,0)
(
0
,
1
)
(1, )
(1,0)
(
0
,
1
)
(2, )
(1,0)
(
0
,
1
)
(3, )
(1,0)
(
0
,
1
)
(4, )
(1,0)
(
0
,
1
)
(-1, )
(1,0)
(
0
,
1
)
(-2, )
(1,0)
(1,0)
(
0
,
1
)
(0, )
(1,0)
(
0
,
1
)
(1, )
(1,0)
(
0
,
1
)
(2, )
(1,0)
(
0
,
1
)
(3, )
(1,0)
(
0
,
1
)
(4, )
(1,0)
(
0
,
1
)
(-1, )
(1,0)
(
0
,
1
)
(-2, )
(1,0)
0
1
2
3
-1
0
1
2
3
-1
0
1
2
3
-1
0
1
2
3
-1
0
1
2
3
-1
0
1
2
3
-1
0
1
2
3
-1
(1,0)
(
0
,
1
)
(0, )
(1,0)
(
0
,
1
)
(1, )
(1,0)
(
0
,
1
)
(2, )
(1,0)
(
0
,
1
)
(3, )
(1,0)
(
0
,
1
)
(4, )
(1,0)
(
0
,
1
)
(-1, )
(1,0)
(
0
,
1
)
(-2, )
(1,0)
(1,0)
(0, )
(1,0)
(1, )
(1,0)
(2, )
(1,0)
(3, )
(1,0)
(4, )
(1,0)
(-1, )
(1,0)
(-2, )
(1,0)
(
0
,
1
)
(
0
,
1
)
(
0
,
1
)
(
0
,
1
)
(
0
,
1
)
(
0
,
1
)
(
0
,
1
)
-2 -2 -2 -2 -2 -2 -2
4 4 4 4 4 4 4
(1,0)
(
0
,
1
)
(5, )
(
0
,
1
)
(1,0)
(
0
,
1
)
(5, )
(1,0)
(
0
,
1
)
(5, )
(1,0)
(
0
,
1
)
(5, )
(1,0)
(
0
,
1
)
(5, )
0
1
2
3
-1
(1,0)
(
0
,
1
)
(5, )
(1,0)
(5, )
(
0
,
1
)
-2
4
(2,0) (2,0) (2,0) (2,0)
(
1
,
2
)
(
1
,
2
)
(2,0) (2,0) (2,0) (2,0)
(2,0) (2,0) (2,0) (2,0)
(2,0) (2,0) (2,0) (2,0)
(
1
,
2
)
(
1
,
2
)
(
1
,
2
)
(
1
,
2
)
(
1
,
2
)
(
1
,
2
)
(
1
,
2
)
(
1
,
2
)
(
1
,
2
)
(
1
,
2
)
(
1
,
2
)
(1,0)
(
0
,
1
)
(0, )
(1,0)
(
0
,
1
)
(1, )
(1,0)
(
0
,
1
)
(2, )
(1,0)
(
0
,
1
)
(3, )
(1,0)
(
0
,
1
)
(4, )
(1,0)
(
0
,
1
)
(-1, )
(1,0)
(
0
,
1
)
(-2, )
(1,0)
(
0
,
1
)
(
0
,
1
)
(
0
,
1
)
(
0
,
1
)
(
0
,
1
)
(
0
,
1
)
(
0
,
1
)
(1,0)
(
0
,
1
)
(0, )
(1,0)
(
0
,
1
)
(1, )
(1,0)
(
0
,
1
)
(2, )
(1,0)
(
0
,
1
)
(3, )
(1,0)
(
0
,
1
)
(4, )
(1,0)
(
0
,
1
)
(-1, )
(1,0)
(
0
,
1
)
(-2, )
(1,0)
(1,0)
(
0
,
1
)
(0, )
(1,0)
(
0
,
1
)
(1, )
(1,0)
(
0
,
1
)
(2, )
(1,0)
(
0
,
1
)
(3, )
(1,0)
(
0
,
1
)
(4, )
(1,0)
(
0
,
1
)
(-1, )
(1,0)
(
0
,
1
)
(-2, )
(1,0)
(1,0)
(
0
,
1
)
(0, )
(1,0)
(
0
,
1
)
(1, )
(1,0)
(
0
,
1
)
(2, )
(1,0)
(
0
,
1
)
(3, )
(1,0)
(
0
,
1
)
(4, )
(1,0)
(
0
,
1
)
(-1, )
(1,0)
(
0
,
1
)
(-2, )
(1,0)
(1,0)
(
0
,
1
)
(0, )
(1,0)
(
0
,
1
)
(1, )
(1,0)
(
0
,
1
)
(2, )
(1,0)
(
0
,
1
)
(3, )
(1,0)
(
0
,
1
)
(4, )
(1,0)
(
0
,
1
)
(-1, )
(1,0)
(
0
,
1
)
(-2, )
(1,0)
0
1
2
3
-1
0
1
2
3
-1
0
1
2
3
-1
0
1
2
3
-1
0
1
2
3
-1
0
1
2
3
-1
0
1
2
3
-1
(1,0)
(
0
,
1
)
(0, )
(1,0)
(
0
,
1
)
(1, )
(1,0)
(
0
,
1
)
(2, )
(1,0)
(
0
,
1
)
(3, )
(1,0)
(
0
,
1
)
(4, )
(1,0)
(
0
,
1
)
(-1, )
(1,0)
(
0
,
1
)
(-2, )
(1,0)
(1,0)
(0, )
(1,0)
(1, )
(1,0)
(2, )
(1,0)
(3, )
(1,0)
(4, )
(1,0)
(-1, )
(1,0)
(-2, )
(1,0)
(
0
,
1
)
(
0
,
1
)
(
0
,
1
)
(
0
,
1
)
(
0
,
1
)
(
0
,
1
)
(
0
,
1
)
-2 -2 -2 -2 -2 -2 -2
4 4 4 4 4 4 4
(1,0)
(
0
,
1
)
(5, )
(
0
,
1
)
(1,0)
(
0
,
1
)
(5, )
(1,0)
(
0
,
1
)
(5, )
(1,0)
(
0
,
1
)
(5, )
(1,0)
(
0
,
1
)
(5, )
0
1
2
3
-1
(1,0)
(
0
,
1
)
(5, )
(1,0)
(5, )
(
0
,
1
)
-2
4
B = (Zx) (Zy)
A = Z Z
y=(1,2); x=(2,0);
Figure 3.2: Example 34a): / = Z Z, and B is the subgroup generated by x = (2, 0) and
y = (1, 2).
(
0
,
1
)
(0, )
(
0
,
1
)
(1, )
(
0
,
1
)
(2, )
(
0
,
1
)
(3, )
(
0
,
1
)
(4, )
(
0
,
1
)
(-1, )
(
0
,
1
)
(-2, )
(
0
,
1
)
(
0
,
1
)
(
0
,
1
)
(
0
,
1
)
(
0
,
1
)
(
0
,
1
)
(
0
,
1
)
(
0
,
1
)
(0, )
(
0
,
1
)
(1, )
(
0
,
1
)
(2, )
(
0
,
1
)
(3, )
(
0
,
1
)
(4, )
(
0
,
1
)
(-1, )
(
0
,
1
)
(-2, )
(
0
,
1
)
(0, )
(
0
,
1
)
(1, )
(
0
,
1
)
(2, )
(
0
,
1
)
(3, )
(
0
,
1
)
(4, )
(
0
,
1
)
(-1, )
(
0
,
1
)
(-2, )
(
0
,
1
)
(0, )
(
0
,
1
)
(1, )
(
0
,
1
)
(2, )
(
0
,
1
)
(3, )
(
0
,
1
)
(4, )
(
0
,
1
)
(-1, )
(
0
,
1
)
(-2, )
(
0
,
1
)
(0, )
(
0
,
1
)
(1, )
(
0
,
1
)
(2, )
(
0
,
1
)
(3, )
(
0
,
1
)
(4, )
(
0
,
1
)
(-1, )
(
0
,
1
)
(-2, )
0
1
2
3
-1
0
1
2
3
-1
0
1
2
3
-1
0
1
2
3
-1
0
1
2
3
-1
0
1
2
3
-1
0
1
2
3
-1
(
0
,
1
)
(0, )
(
0
,
1
)
(1, )
(
0
,
1
)
(2, )
(
0
,
1
)
(3, )
(
0
,
1
)
(4, )
(
0
,
1
)
(-1, )
(
0
,
1
)
(-2, )
(0, ) (1, ) (2, ) (3, ) (4, ) (-1, ) (-2, )
(
0
,
1
)
(
0
,
1
)
(
0
,
1
)
(
0
,
1
)
(
0
,
1
)
(
0
,
1
)
(
0
,
1
)
-2 -2 -2 -2 -2 -2 -2
4 4 4 4 4 4 4
(
0
,
1
)
(5, )
(
0
,
1
)
(
0
,
1
)
(5, )
(
0
,
1
)
(5, )
(
0
,
1
)
(5, )
(
0
,
1
)
(5, )
0
1
2
3
-1
(
0
,
1
)
(5, )
(5, )
(
0
,
1
)
-2
4
(
1
,
2
)
(
1
,
2
)
(
1
,
2
)
(
1
,
2
)
(
1
,
2
)
(
1
,
2
)
(
1
,
2
)
(
1
,
2
)
(
1
,
2
)
(
1
,
2
)
(
1
,
2
)
(
1
,
2
)
(
1
,
2
)
(
1
,
2
)
(
0
,
1
)
(0, )
(
0
,
1
)
(1, )
(
0
,
1
)
(2, )
(
0
,
1
)
(3, )
(
0
,
1
)
(4, )
(
0
,
1
)
(-1, )
(
0
,
1
)
(-2, )
(
0
,
1
)
(
0
,
1
)
(
0
,
1
)
(
0
,
1
)
(
0
,
1
)
(
0
,
1
)
(
0
,
1
)
(
0
,
1
)
(0, )
(
0
,
1
)
(1, )
(
0
,
1
)
(2, )
(
0
,
1
)
(3, )
(
0
,
1
)
(4, )
(
0
,
1
)
(-1, )
(
0
,
1
)
(-2, )
(
0
,
1
)
(0, )
(
0
,
1
)
(1, )
(
0
,
1
)
(2, )
(
0
,
1
)
(3, )
(
0
,
1
)
(4, )
(
0
,
1
)
(-1, )
(
0
,
1
)
(-2, )
(
0
,
1
)
(0, )
(
0
,
1
)
(1, )
(
0
,
1
)
(2, )
(
0
,
1
)
(3, )
(
0
,
1
)
(4, )
(
0
,
1
)
(-1, )
(
0
,
1
)
(-2, )
(
0
,
1
)
(0, )
(
0
,
1
)
(1, )
(
0
,
1
)
(2, )
(
0
,
1
)
(3, )
(
0
,
1
)
(4, )
(
0
,
1
)
(-1, )
(
0
,
1
)
(-2, )
0
1
2
3
-1
0
1
2
3
-1
0
1
2
3
-1
0
1
2
3
-1
0
1
2
3
-1
0
1
2
3
-1
0
1
2
3
-1
(
0
,
1
)
(0, )
(
0
,
1
)
(1, )
(
0
,
1
)
(2, )
(
0
,
1
)
(3, )
(
0
,
1
)
(4, )
(
0
,
1
)
(-1, )
(
0
,
1
)
(-2, )
(0, ) (1, ) (2, ) (3, ) (4, ) (-1, ) (-2, )
(
0
,
1
)
(
0
,
1
)
(
0
,
1
)
(
0
,
1
)
(
0
,
1
)
(
0
,
1
)
(
0
,
1
)
-2 -2 -2 -2 -2 -2 -2
4 4 4 4 4 4 4
(
0
,
1
)
(5, )
(
0
,
1
)
(
0
,
1
)
(5, )
(
0
,
1
)
(5, )
(
0
,
1
)
(5, )
(
0
,
1
)
(5, )
0
1
2
3
-1
(
0
,
1
)
(5, )
(5, )
(
0
,
1
)
-2
4
(
1
,
2
)
(
1
,
2
)
(
1
,
2
)
(
1
,
2
)
(
1
,
2
)
(
1
,
2
)
(
1
,
2
)
(
1
,
2
)
(
1
,
2
)
(
1
,
2
)
(
1
,
2
)
(
1
,
2
)
(
1
,
2
)
(
1
,
2
)
(
0
,
4
)
(
0
,
4
)
(
0
,
4
)
(
0
,
4
)
(
0
,
4
)
(
0
,
4
)
(
0
,
4
)
(
0
,
4
)
(
0
,
4
)
(
0
,
4
)
(
0
,
4
)
(
0
,
4
)
(
0
,
4
)
(
0
,
4
)
(
0
,
4
)
(
0
,
4
)
a
1
= (1,2)
a
2
= (0,1)
m
1
=1; b
1
= a
1
= (1,2)
m
2
=4; b
2
= 4a
2
= (0,1)
A = (Za
1
) (Za
2
)
B = (Zb
1
) (Zb
2
)
Figure 3.3: Example 34a): / = (Za
1
) (Za
2
), where a
1
= (1, 2) and a
2
= (0, 1). Let m
1
= 1
and m
2
= 4, so that b
1
= m
1
a
1
= (1, 2) and b
2
= m
2
a
2
= (0, 4). Then B = (Zb
1
) (Zb
2
).
26 CHAPTER 3. FREE ABELIAN GROUPS
(b) Let / = Z
3
, and let B = (5y, 7z, 0) ; y, z Z. In this case, R = 3 and S = 2. Dene:
a
1
= (1, 0, 0) and m
1
= 5, so that b
1
= (5, 0, 0);
a
2
= (0, 1, 0) and m
2
= 7, so that b
2
= (0, 7, 0);
a
3
= (0, 0, 1).
Then a
1
, a
2
, a
3
is a basis for /, and b
1
, b
2
is a basis for B.
Proof of Proposition 33: We will prove (a) by induction on R, and prove (b) by induction
on S. We will use the following fact:
Claim 8: There exists elements a
1
B and b
1
B, and a number m
1
N so that
b
1
= m
1
a
1
, such that
/ = (Za
1
)

/ and B = (Zb
1
)

B,
where

/ < / is a subgroup of rank R 1, and

B <

/ is a subgroup of rank S 1.
....we will then apply the induction hypotheses to

/ and

B.
Claim 8 will follow from Claims 1 through 7 below. Without loss of generality, suppose
/ = Z
R
(we know that / is always isomorphic to Z
R
, so this is okay). For all r [1..R], let
pr
r
: /Z be projection into the rth coordinate ie. pr
r
(z
1
, . . . , z
R
) = z
r
.
Let : /Z be any group homomorphism. Then (B) < Z is a subgroup, so there is some
m

Z such that (B) = m

Z. Dene:
M = m

; : /Z any homomorphism.
Claim 1: M,= 0.
Proof: We must show that there is some homomorphism : /Z so that (B) ,= 0.
Recall that / = Z
R
, and pr
r
: /Z is projection onto the rth coordinate. Given any
a /, there must be some r [1..R] so that pr
r
(a) ,= 0 otherwise a = (0, 0, . . . , 0). In
particular, for any b B, there is some r [1..R] so that pr
r
(b) ,= 0. Thus, pr
r
(B) ,= 0.
2 [Claim 1]
Now, let m
1
be the minimal nonzero element in M. Let
1
: /Z be a homomorphism
such that m
(
1
)
= m
1
, and let b
1
B be an element such that

1
(b
1
) = m
1
. ()
Claim 2: Let : /Z be any homomorphism. Then m
1
divides (b
1
).
Proof: Let n = (b
1
), and let d = gcd(m
1
, n). Thus d = z
1
m
1
+ zn for some integers
z
1
, z Z. Dene the homomorphism : /Z by: (a) = z
1

1
(a) + z(a). Then
(b
1
) = z
1

1
(b
1
) + z(b
1
)
by ()
z
1
m
1
+ zn = d.
Thus, d (B), so m

must divide d, so m

d. But d divides m
1
, so d m
1
. Thus,
m

m
1
. But m
1
is the minimal nonzero element in M, so m
1
m

. Therefore, m

= m
1
,
which means d = m
1
, which means m
1
divides n. ...................... 2 [Claim 2]
3.5. SUBGROUPS OF FREE ABELIAN GROUPS 27
Recall that / = Z
R
, and pr
r
: /Z is projection onto the rth coordinate. Claim 2
says m
1
divides pr
r
(b
1
) for all r [1..R]. In other words, b
1
= (b
1
, b
2
, . . . , b
R
), where
b
1
= m
1
a
1
, b
2
= m
2
a
2
, . . . , b
R
= m
R
a
R
, for some a
1
, . . . , a
R
Z. Thus, b
1
= m
1
a
1
, where
a
1
= (a
1
, a
2
, . . . , a
R
).
Claim 3:
1
(a
1
) = 1.
Proof: m
1
=
1
(b
1
) =
1
(m
1
a
1
) = m
1

1
(a
1
). Thus, (a
1
) = 1. ...... 2 [Claim 3]
Now, let

/ = ker(
1
), and let

B = K B.
Claim 4: / = (Za
1
)

/.
Proof:
Claim 4.1: / = (Za
1
) +

/.
Proof: Let a /, and let z =
1
(a). Let a = a za
1
. I claim a

/. To see this, note
that
1
(a) =
1
(a) z
1
(a
1
)
Clm 3
z z 1 = z z = 0.
Thus, a = za
1
+a (Za
1
) +

/. ................................ 2 [Claim 4.1]
Claim 4.2: (Za
1
)

/ = 0.
Proof: Any element of Za
1
has the form za
1
for some z Z. If za
1


/, this means that

1
(za
1
) = 0. But
1
(za
1
) = z
1
(a
1
)
Clm 3
z 1 = z. Thus, z = 0. 2 [Claim 4.2]
Claims 4.1 and 4.2 imply that / = (Za
1
)

/. ....................... 2 [Claim 4]
Claim 5: B = (Zb
1
)

B.
Proof:
Claim 5.1: B = (Zb
1
) +

B.
Proof: Let b B, and let z =
1
(b). Let

b = b za
1
. As in Claim 4.1,

b

/.
I claim that za
1
Zb. To see this, recall that z =
1
(b)
1
(B) = m
1
Z by
denition of m
1
. Thus, m
1
must divide z. Thus, z = ym
1
for some y Z, so that
za
1
= ym
1
a
1
= yb
1
Zb.
I also claim that

b

B. We know that

b

/. Observe that

b = b za
1
= b yb
1
is a dierence of two elements in B, so

b B also. Thus,

b

/ B =

B.
Thus, a = yb
1
+

b (Zb
1
) +

B. ............................... 2 [Claim 5.1]
Claim 4.1 implies that (Zb
1
)

B = 0. We conclude that B = (Zb

1
)

B. 2 [Claim 5]
Claim 6: rank
_

/
_
= R 1.
Proof: Claim 4 says that / = (Za
1
)

/. Thus,
R = rank (/)
Lem.23
rank (Za
1
) + rank
_

/
_
= 1 + rank
_

/
_
.
Thus rank
_

/
_
= R 1. ............................................. 2 [Claim 6]
28 CHAPTER 3. FREE ABELIAN GROUPS
Claim 7: rank
_

B
_
= S 1.
Proof: Claim 5 says B = (Zb
1
)

B. Now proceed exactly as with Claim 6. 2 [Claim 7]

Claim 8 follows from Claims 1 through 7. Now to prove the theorem....
Proof of (a): (by induction on S)
Case (S = 0) Suppose rank (B) = 0. I claim B = 0. To see this, suppose b B was
nonzero. Then for any z Z, zb ,= 0, so the set b is linearly independent, hence
rank (B) 1, a contradiction.
Induction on S: Suppose part (a) is true for all subgroups of / of rank S 1. Then in
particular, part (a) holds for all

B; hence,

B is a free abelian group. Since B = (Zb
1
)

B,
Lemma 25(a) says that B is also a free abelian group.
We conclude that all subgroups of / are free.
Proof of (b): (by induction on R)
Case (R = 1) In this case, /

= Z. Thus, B < Z, so we know there is some m
1
N so
that B = m
1
Z. So, let a
1
= 1 and b
1
= m
1
a
1
= m
1
. Then Z is generated by a
1
, and B is
generated by b
1
.
Induction on R: Suppose part (b) is true for all subgroups of all free abelian groups
of rank R 1. Consider

/. By Claim 6, we know that rank
_

/
_
= R 1. By part (a)
of the theorem (already proved), we know that

/ must be a free group. Also,

B <

/, and
rank
_

B
_
= S 1 (Claim 7) so by induction hypothesis, there is some basis a
2
, . . . , a
R
for

/, and numbers m
2
, . . . , m
S
N, so that, if b
s
= m
s
a
s
for all s [2..S], then b
2
, . . . , b
S

is a basis for

B.
Since / = (Za
1
)

/ (by Claim 4), it follows that a
1
, a
2
, . . . , a
R
is a basis for /. Since
B = (Zb
1
)

B (by Claim 5), it follows that b
1
, b
2
, . . . , b
S
is a basis for B. 2
Chapter 4
The Structure Theory of Abelian
Groups
Group structure theory concerns the decomposition of groups into elementary components.
For example, the Jordan-Holder theorem says that any group ( has a composition series
e = ^
0
^
1
. . . ^
K
= (
where the groups o
k
= ^
k
/^
k1
are simple for all k.
In this section, we will show that any nitely generated abelian group can be written as a
direct sum of cyclic groups.
4.1 Direct Products
Let (/, ), (B, ), and (c, ) be three groups. The direct product of /, B, and c is the group
/B c = (a, b, c) ; a /, b B, c c
with the multiplication operation:
(a
1
, b
1
, c
1
) (a
2
, b
2
, c
2
) = (a
1
a
2
, b
1
b
2
, c
1
c
2
)
We have dened this for three groups, but the same construction works for any number of
groups. The direct product of (
1
, (
2
, . . . , (
N
is denoted by (
1
(
2
. . . (
N
or
N

n=1
(
n
.
Example 35: Suppose / = B = c = (R, +). Then /B c = R R R = R
3
is just
three-dimensional Euclidean space, with the usual vector addition.
Proposition 36

(
1
(
2
. . . (
N

= [(
1
[ [(
2
[ [(
N
[.
29
30 CHAPTER 4. THE STRUCTURE THEORY OF ABELIAN GROUPS
inj
B
inj
A
pr
A
pr
B
B
B
A A
A
~
G
e
e
A
(a,e
B
)
(a,b)
(e
A
,b)
a a
b
B
~
e
b b
e
b
e
A
Figure 4.1: The product group ( = /B, with injection and projection maps.
Proof: (Exercise 17) 2
Example 37:

Z
/3
Z
/5
Z
/7

Z
/3

Z
/5

Z
/7

= 3 5 7 = 75.
Lemma 38 If (/, +), (B, +), and (c, +) are abelian groups, then their product /B c
is also abelian.
Proof: (Exercise 18) 2
(This theorem likewise generalizes to a product of any number of groups)
The product of nitely many (additive) abelian groups is usually called a direct sum, and
indicated with the notation /B c.
Proposition 39 (Properties of Product Groups)
Let /, B, and c be groups, and let ( = /B c be their product.
(a) The identity element of ( is just (e
A
, e
B
, e
C
).
(b) For any (a, b, c) (, (a, b, c)
1
= (a
1
, b
1
, c
1
).
(c) Dene injection maps:
inj
A
: /( by inj
A
(a) = (a, e
B
, e
C
) for all a /;
inj
B
: /( by inj
B
(b) = (e
A
, b, e
C
) for all b B;
and inj
A
: /( by inj
C
(c) = (e
A
, e
B
, c) for all c c
Then inj
A
, inj
B
, and inj
C
are monomorphisms.
4.1. DIRECT PRODUCTS 31
(d) Dene

/ = image [inj
A
] = (a, e
B
, e
C
) ; a / (;

B = image [inj
B
] = (e
A
, b, e
C
) ; b B (;
and

c = image [inj
C
] = (e
A
, e
B
, c) ; c B (
Then

/,

B, and

C are normal subgroups of (
(e)

/ is isomorphic to / (via inj
A
). Likewise,

B

= B and

C

= c.
(f ) ( =

/

B

C. In other words, for any g (, there are elements a

/,

b

B, and
c

C so that g = a

b c.
(g) The elements of

/,

B, and

C commute with each other. In other words, for any a

/,

b

B, and c

C,
a

b =

b a, a c = c a, and

b c = c

b,
(h)

/

B = e,

/

C = e, and

B

C = e.
(i) Dene projection maps:
pr
A
: (/ by pr
A
(a, b, c) = a;
pr
B
: (B by pr
B
(a, b, c) = b;
and pr
C
: (c by pr
C
(a, b, c) = c
Then pr
A
, pr
B
, and pr
C
are epimorphisms.
(j) pr
A
inj
A
= Id
A
, pr
B
inj
B
= Id
B
, and pr
C
inj
C
= Id
C
.
Proof: (Exercise 19) 2
We have stated Proposition 39 for a product of three groups for the sake of simplicity, but
the obvious generalization holds for any number of groups. The conditions of this propositon
actually characterize direct products, as follows:
Proposition 40 Let ( be a group, with subgroups /, B < (, such that ( = / B. The
following are equivalent:
(a) ( is isomorphic to /B, via the mapping:
: /B (a, b) a b (
32 CHAPTER 4. THE STRUCTURE THEORY OF ABELIAN GROUPS
(b) / B = e, and for every a / and b B, ab = ba.
(c) / B = e, and /( and B (.
Proof: The assertions (a)=(b) and (a)=(c) follow from Proposition 39(d,g,h).
(b)=(a) We must show is a homomorphism and bijective.
is a homomorphism:
_
(a
1
, b
1
) (a
2
, b
2
)
_
(DP)

_
(a
1
a
2
), (b
1
b
2
)
_
(D)
(a
1
a
2
)(b
1
b
2
)
(C)
(a
1
b
1
)
(a
2
b
2
)
(D)
(a
1
, b
1
)(a
2
, b
2
). Here, equalities (D) are by denition of ; (DP) is by denition
of the product group, and (C) is because (b) says a
2
and b
1
commute.
is surjective: By hypothesis, for any g ( we can nd a / and b B so that g = a b.
But then g = (a, b).
is injective: Well show ker() = e. Suppose (a, b) = e. Thus, a b = e. Thus,
b = a
1
, so b /. Thus, b / B, so b = e. Thus, a = e. Thus, (a, b) = (e, e).
(c)=(b) Let a / and b B. Then:
/(, so b a
1
b
1
/, B (, so a b a
1
B,
thus, a b a
1
b
1
/. thus, a b a
1
b
1
B;
Thus, aba
1
b
1
/B = e, so we conclude that aba
1
b
1
= e. Thus, ab = (a
1
b
1
)
1
=
(b
1
)
1
(a
1
)
1
= ba. 2
When the conditions of Proposition 41 are satised, we say that ( is an internal direct
product of / and B. We then abuse notation by writing, ( = / B. In terms of the
notation of Proposition 39, we are implicitly identifying / with

/ and B with

B.
If ( is abelian, then the conditions in parts (b) and (c) of Proposition 41 become trivial...
Corollary 41 (Interpretation for Abelian Groups)
Let ((, +) be an (additive) abelian group, with subgroups /, B < (. Then the following
are equivalent:
(a) / B = 0, and ( = / + B (that is, for any g ( there are a / and b B so
that g = a + b).
(b) (

= /B, via the mapping : /B (a, b) (a + b) (. 2
In the abelian cas, we say that ( is an internal direct sum of / and B. We again abuse
notation by writing, ( = /B.
4.2. THE CHINESE REMAINDER THEOREM 33
4.2 The Chinese Remainder Theorem
Prerequisites: 4.1
According to ancient legend from the War of the Three Kingdoms
1
,
The redoubtable Shu general Zhu Geliang wanted to rapidly count his troops before
entering a great battle against the Wei. He estimated that there were less than 15000
troops, and 15000 < 17017 = 7 11 13 17. So, rst he had the troops line
up in rows of 7, and found that there were 6 left over. Next, he lined them up in
rows of 11 each, and found there were 7 left over. Next, in rows of 13, there were
5 left over. Finally, in rows of 17, there were 2 left over. He concluded that there
were exactly 14384 troops.
Theorem 42 (Chinese Remainder Theorem)
(a) Let n
1
, n
2
, . . . , n
K
N be any collection of pairwise relatively prime numbers (ie.
gcd(n
j
, n
k
) = 1 whenever j ,= k), and let n = n
1
n
2
n
K
. Then
Z
/n

= Z
/n
1
Z
/n
2
. . . Z
/n
k
.
(b) To be specic, dene : Z
/n
Z
/n
1
Z
/n
2
. . . Z
/n
k
by:
([z]
n
) = ([z]
n
1
, [z]
n
2
, . . . [z]
n
K
) ,
(where [z]
n
is the congruence class of z, mod n, etc.). Then is an isomorphism.
(c) In particular, suppose n N has prime factorization n = p

1
1
p

2
2
p

K
K
. Then
Z
/n

= Z
/(p

1
1
)
Z
/(p

2
2
)
. . . Z
/(p

K
K
)
Proof: (c) is a special case of (a) which follows from (b), which is Exercise 20 . 2
Example 43:
(a) 12 = 4 3 and 4 is relatively prime to 3. Therefore, Z
/12

= Z
/4
Z
/3
. (Figure 4.2).
Part (b) provides a specic isomorphism : Z
/12
Z
/4
Z
/3
, given by:
([z]
12
) = ([z]
4
, [z]
3
) .
For example, ([5]
12
) ([5]
4
, [5]
3
) = ([1]
14
, [2]
3
).
1
Actually, I made this up. But Zhu Geliang (C.E. 181-234) was a famously brilliant Shu military leader,
inventor, and mathematician.
34 CHAPTER 4. THE STRUCTURE THEORY OF ABELIAN GROUPS
(
0
,
1
)
(
0
,
1
)
(
0
,
1
)
(
0
,
1
)
(1,0)
(
0
,
1
)
(0, )
(1,0)
(
0
,
1
)
(1, )
(1,0)
(
0
,
1
)
(2, )
(1,0)
(
0
,
1
)
(3, )
(1,0)
(1,0)
(
0
,
1
)
(0, )
(1,0)
(
0
,
1
)
(1, )
(1,0)
(
0
,
1
)
(2, )
(1,0)
(
0
,
1
)
(3, )
(1,0)
(1,0)
(
0
,
1
)
(0, )
(1,0)
(
0
,
1
)
(1, )
(1,0)
(
0
,
1
)
(2, )
(1,0)
(
0
,
1
)
(3, )
(1,0)
0
1
2
0
1
2
0
1
2
0
1
2
4 3
4 3 12
(0, ) (1, ) (2, ) (3, )
(0, ) (1, ) (2, ) (3, )
(0, ) (1, ) (2, ) (3, )
0
1
2
0
1
2
0
1
2
0
1
2
(
1
,
1
)
(
1
,
1
)
(
1
,
1
)
(
1
,
1
)
(
1
,
1
)
(
1
,
1
)
(
1
,
1
)
(
1
,
1
)
(
1
,
1
)
(
1
,
1
)
(
1
,
1
)
(
1
,
1
)
(A) (B)
Figure 4.2: Example 43a).
(b) 210 = 14 15, and 14 is relatively prime to 15. Thus, part (a) of the Chinese Remainder
Theorem says that Z
/210

= Z
/14
Z
/15
. Part (b) provides a specic isomorphism :
Z
/210
Z
/14
Z
/15
, given by: ([z]
210
) = ([z]
14
, [z]
15
).
Part (c) of the Chinese Remainder Theorem says that any nite cyclic group can be written
as a direct sum of prime power cyclic groups. This is a special case of the Fundamental Theorem
of Finitely Generated Abelian Groups, which we will see in 4.3.
Example 44:
(a) 12 = 4 3 = 2
2
3. Therefore, Z
/12

= Z
/4
Z
/3
.
(b) 210 = 2 3 5 7. Therefore, Z
/210

= Z
/2
Z
/3
Z
/5
Z
/7
.
(c) 720 = 16 9 5 = 2
4
3
2
5. Therefore, Z
/720

= Z
/16
Z
/9
Z
/5
.
Important: The Chinese Remainder Theorem does not say that Z
/nm

= Z
/n
Z
/m
when
n and m are not relatively prime. For example, although 12 = 2 6, it is not true that
Z
/12
= Z
/2
Z
/6
. Likewise, it is not true that Z
/8
= Z
/2
Z
/2
Z
/2
.
The story of Zhu Geliang comes from the following application of Theorem 42(b):
Corollary 45 Let n
1
, n
2
, . . . , n
K
N be pairwise relatively prime and let n = n
1
n
2
n
K
.
Given any numbers z
1
[0..n
1
), z
2
[0..n
2
), . . . , z
K
[0..n
K
), there is a unique z [0..n)
such that z z
1
(mod n
1
), z z
2
(mod n
2
), . . . , z z
K
(mod n
K
). 2
4.2. THE CHINESE REMAINDER THEOREM 35
Example 46: In the story of Zhu Geliang, N = 17017 = 7 11 13 17, z
1
= 6, z
2
= 7,
z
3
= 5, and z
4
= 2. Thus, Zhu knows there is a unique z [0...17016] so that
z 6 (mod 7), z 7 (mod 11), z 5 (mod 13), and z 2 (mod 17).
To be precise, z = 14384.
In Example 46), how does Zhu Geliang know that z = 14384? The Chinese Remainder
Theorem states that the solution z exists, but it does not say how to compute it. To compute
z, we need to invert the isomorphism in Theorem 42(b).
Proposition 47 Inversion Formula
Let N = n
1
n
2
n
K
as in Theorem 42. Dene:
a
1
= N/n
1
= n
2
n
3
. . . n
K
and b
1
= a
1
1
(mod n
1
)
a
2
= N/n
2
= n
1
n
3
. . . n
K
and b
2
= a
1
2
(mod n
2
)
.
.
.
a
K
= N/n
K
= n
1
n
2
. . . n
K1
and b
K
= a
1
K
(mod n
K
)
(By this, I mean that a
1
b
1
1 (mod n
1
), a
2
b
2
1 (mod n
2
), etc.)
Now, dene e
1
= a
1
b
1
, e
2
= a
2
b
2
, . . . , e
K
= a
K
b
K
. Finally, let Now dene z = z
1
e
1
+
z
2
e
2
+ . . . z
K
e
K
. Then z z
1
(mod n
1
), z z
2
(mod n
2
), . . . , z z
K
(mod n
K
).
Proof: By construction:
e
1
1 (mod n
1
); e
1
0 (mod n
2
) e
1
0 (mod n
3
) . . . e
1
0 (mod n
K
)
e
2
0 (mod n
1
); e
2
1 (mod n
2
) e
2
0 (mod n
3
) . . . e
2
0 (mod n
K
)
e
3
0 (mod n
1
); e
3
0 (mod n
2
) e
3
1 (mod n
3
) . . . e
3
0 (mod n
K
)
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
e
K
0 (mod n
1
); e
K
0 (mod n
2
) e
K
0 (mod n
3
) . . . e
K
1 (mod n
K
)
Thus,
z
1
e
1
+ z
2
e
2
+ . . . z
K
e
K
z
1
1 + z
2
0 + . . . + z
K
0 z
1
(mod n
1
)
z
1
e
1
+ z
2
e
2
+ . . . z
K
e
K
z
1
0 + z
2
1 + . . . + z
K
0 z
1
(mod n
2
)
.
.
.
z
1
e
1
+ z
2
e
2
+ . . . z
K
e
K
z
1
0 + z
2
0 + . . . + z
K
1 z
1
(mod n
K
)
2
The Chinese Remainder Theorem is really a theorem about abelian groups
2
, but it has the
following partial generalization to nonabelian groups:
2
Later we will also see versions of it for rings and for modules.
36 CHAPTER 4. THE STRUCTURE THEORY OF ABELIAN GROUPS
Proposition 48 Let ( be a (possibly nonabelian) group, with [([ = N = n
1
n
2
n
K
, where
n
1
, n
2
, . . . , n
K
are pairwise relatively prime. Suppose there are normal subgroups ^
1
, ^
2
, . . . , ^
K

( so that [^
k
[ = n
k
. Then (

= ^
1
^
2
. . . ^
K
.
Proof: (by induction)
Base Case (K = 2) We will apply Proposition 41(c).
Claim 1: ^
1
^
2
= e.
Proof: If g ^
1
, then [g[ must divide [^
1
[ = n
1
Likewise, if g ^
2
, then [g[ divides n
2
.
Thus, [g[ divides gcd(n
1
, n
2
) = 1. Thus, [g[ = 1, so g = e. .............. 2 [Claim 1]
Claim 2: ( = ^
1
^
2
.
Proof: Clearly, ^
1
^
2
is a subset (. We claim that ^
1
^
2
has the same cardinality as
(. To see this, observe that

^
1
^
2

()
[^
1
[ [^
2
[
[^
1
^
2
Claim 1
n
1
n
2
1
= n = [([.
where () follows from Prop 13, 3.2, p.94 in Dummit & Foote. ........ 2 [Claim 2]
Now combine Claims 1 and 2 with Proposition 41(c).
Induction: Observe that, since ^
2
, ^
3
, . . . , ^
K
(, the product / = ^
2
^
3
^
K
is
a subgroup of (. By induction hypothesis,
/

= ^
2
^
3
. . . ^
K
. (4.1)
Now, let m = [/[ = n
2
n
3
n
K
. Then m is relatively prime to n
1
(because each of n
2
,
n
3
, . . . , n
K
is relatively prime to n
1
). Thus, by applying the Base Case (with / playing
the role of ^
2
) we have: (

= ^
1
/
eqn.(4.1)
^
1
^
2
^
3
. . . ^
K
. 2
Corollary 49 Let ( be a (possibly nonabelian) group. Suppose [([ = n, and let n have
prime factorization: n = p

1
1
p

2
2
. . . p

K
K
. Suppose there are subgroups
3
T
1
, T
2
, . . . , T
K
< ( so
that [T
k
[ = p

k
k
.
If T
1
, . . . , T
K
are normal subgroups of (, then (

= T
1
T
2
. . . T
K
.
Proof: Apply Proposition 48, with n
1
= p

1
1
, n
2
= p

2
2
, . . . , n
K
= p

K
K
2
3
These are called Sylow subgroups.
4.3. THE FUNDAMENTAL THEOREM OF FINITELY GENERATED ABELIAN GROUPS37
Another proof of part (c) of the Chinese Remainder Theorem: Let ( = Z
/N
, and
for all k [1..K], let m
k
= n/p

k
k
. Let m
k
( be the associated congruence class.
Claim 1: [m
k
[ = p

k
k
.
Proof: (Exercise 21) ................................................ 2 [Claim 1]
Let T
k
= m
k
) be the subgroup generated by m
k
. It follows from Claim 1 that [T
k
[ = p

k
k
.
Thus, Corollary 49 says that ( = T
1
T
2
. . . T
K
. 2
4.3 The Fundamental Theorem of Finitely Generated
Abelian Groups
Prerequisites: 4.2, 3.5
Theorem 50 Let / be a nitely generated abelian group. Then there are unqiue prime
powers p

1
1
, p

2
2
, . . . , p

N
n
(where p
1
, p
2
, . . . , p
N
are prime numbers, not necessarily distinct, and

1
, . . . ,
N
are natural numbers), and a unique integer R 0 so that
/

= Z
R
Z
/(p

1
1
)
Z
/(p

2
2
)
. . . Z
/(p

N
N
)
.
Example 51: The following groups typify the Fundamental Theorem:
Z
15
; Z
/2
Z
/2
Z
/2
; Z
/2
Z
/4
Z
/8
; Z
5
Z
/2
Z
/4
Z
/3
Z
/81
Z
/25
Z
/49
Z
/17
.
On the other hand, the abelian group (Q, +) is not nitely generated, and therefore does not
admit a decomposition of this kind. Instead,
Q = Q
2
+Q
3
+Q
5
+Q
7
+Q
11
+Q
13
+ . . .
In other words, Q is a sum of the p-adic rational numbers, for all prime p. Note that this
sum is not direct, because Q
2
Q
3
Q
5
Q
7
Q
11
Q
13
. . . = Z.
To obtain something more like a direct sum, let

Q = Q/Z. If you think of the unit circle as
R/Z, then you can imagine

Q as the subgroup of rational angles on the circle. Likewise, let

Q
p
= Q
p
/Z for all prime p. Then

Q =

Q
2

Q
3

Q
5

Q
7

Q
11

Q
13
. . .
and Q =

QZ is a semidirect product of

Q with Z.
38 CHAPTER 4. THE STRUCTURE THEORY OF ABELIAN GROUPS
Proof of the Fundamental Theorem: By the Universal Covering Property (Corollary 30 on
page 23), nd a free abelian group T and an epimorphism : T/. Let / = ker(); thus,
/

= T//. (4.2)
By Proposition 33 on page 24, / is also a free abelian group, and there is a basis a
1
, a
2
, . . . , a
R

for T, and numbers m

1
, m
2
, . . . , m
S
N so that (m
1
a
1
), (m
2
a
2
), . . . , (m
S
a
S
) is a basis
for /.
In other words, T = Za
1
Za
2
. . . Za
S
Za
S+1
. . . Za
R
,
and / = Zm
1
a
1
Zm
2
a
2
. . . Zm
S
a
S
0 . . . 0.
Thus, T//

= Z/m
1
Z Z/m
2
Z . . . Z/m
S
Z Z . . . Z

= Z
/m
1
Z
/m
2
. . . Z
/m
S
Z . . . Z
In other words, if R

= R S, then
T//

= Z
/m
1
Z
/m
2
. . . Z
/m
S
Z
R

. (4.3)
Now, suppose that m
1
has prime factorization m
1
= p

1
1
p

2
2
p

i
i
. Then the part (c) of the
Chinese Remainder Theorem (Theorem 42 on page 33) says
Z
/m
1

= Z
/(p

1
1
)
Z
/(p

2
2
)
. . . Z
/(p

i
i
)
.
Likewise, if m
2
= p

i+1
i+1
p

i+2
i+2
p

j
j
. then Z
/m
1

= Z
/(p

i+1
i+1
)
Z
/(p

i+2
i+2
)
. . . Z
/(p

j
j
)
.
Proceeding this way, suppose that m
1
, . . . , m
S
collectively have prime factorizations:
m
1
= p

1
1
p

2
2
p

i
i
m
2
= p

i+1
i+1
p

i+2
i+2
p

j
j
.
.
.
m
S
= p

k
k
p

k+1
k+1
p

N
N
(where the prime number sets p
1
, . . . , p
i
,
p
i+1
, . . . , p
j
, . . . . . . , p
k
, . . . , p
N
are not
necessarily disjoint)
Then:
Z
/m
1
Z
/m
2
. . . Z
/m
S
=
_
Z
/(p

1
1
)
. . . Z
/(p

i
i
)
_

_
Z
/(p

i+1
i+1
)
. . . Z
/(p

j
j
)
_
. . .
. . .
_
Z
/(p

k
k
)
. . . Z
/(p

N
N
)
_
. (4.4)
Now, combine equations (4.2,4.3,4.4) to conclude: /

= Z
/(p

1
1
)
. . . Z
/(p

N
N
)
Z
R

. 2
4.3. THE FUNDAMENTAL THEOREM OF FINITELY GENERATED ABELIAN GROUPS39
Example: To illustrate the reasoning of the proof, suppose that T = Z
3
, so that R = 3.
Suppose S = 2, and that
a
1
= (1, 0, 0) and m
1
= 60, so that m
1
a
1
= (60, 0, 0);
a
2
= (0, 1, 0) and m
2
= 80, so that m
2
a
2
= (0, 80, 0);
a
3
= (0, 0, 1).
Thus, / = (60z
1
, 80z
2
, 0) ; z
1
, z
2
Z = (60Z) (80Z) 0. Thus,
/

= T// =
Z Z Z
(60Z) (80Z) 0

= Z
/60
Z
/80
Z.
Now, we apply the Chinese Remainder Theorem:
60 = 4 3 5 = 2
2
3 5, so Z
/60

= Z
/3
Z
/4
Z
/5
; ie. p
1
, p
2
, p
3
= 2, 3, 5.
80 = 16 5 = 2
4
5, so Z
/80

= Z
/16
Z
/5
; ie. p
4
, p
5
= 2, 5.
Observe that the set 2, 3, 5 is not disjoint from the set 2, 5, but we can still proceed to
conclude that:
/

= Z Z
/3
Z
/4
Z
/5
Z
/16
Z
/5
.
40 CHAPTER 4. THE STRUCTURE THEORY OF ABELIAN GROUPS
Chapter 5
The Structure Theory of Nonabelian
Groups
5.1 Three Examples
Prerequisites: 1.4
Example (A) The arithmetic quotient equation
6
3
= 2 (5.1)
is equivalent to the factorization:
6 = 2 3. (5.2)
Can we factor the numbers 2 and 3 any further? No, because they are prime. The equation
6 = 2 3 is a prime factorization of 6. Prime numbers are the elementary components of
number theory: any number can be written in a unique way as a product of prime numbers.
The idea of group structure theory is to factor groups into elementary components in a
similar fashion.
Example (B) Consider the cyclic group Z
/6
=

0,

1,

2,

3,

4,

5, and the subgroup / =

0,

2,

4. It is easy to check that /Z
/6
, and that
Z
/6
/

= Z
/2
. (5.3)
Observe that /

= Z
/3
. Thus, we suspect
Z
/6

= Z
/3
Z
/2
, (5.4)
41
42 CHAPTER 5. THE STRUCTURE THEORY OF NONABELIAN GROUPS
and indeed, this is true. To see this, dene : Z
/3
Z
/2
Z
/6
by ([n]
3
, [m]
2
) = 2n + 3m.
(Exercise 22 Check that is an isomorphism. Here, [n]
3
is the mod-3 congruence class of n, etc.)
Thus, the quotient equation (5.3) entails a factorization (5.4) of the group Z
/6
, just as the
quotient equation (5.1) entailed a factorization (5.2) of the integer 6.
Can we break down Z
/3
any further? No, because Z
/3
contains no normal subgroups, so it
cannot be factored in this way. Likewise, Z
/2
contains no normal subgroups. Thus, Z
/3
and
Z
/2
are analogous to prime numbers. We say that Z
/3
and Z
/2
are simple.
Example (C) Now, consider the symmetric group S
3
=
_
e, (12), (13), (23), (123), (132)
_
,
and the subgroup A
3
=
_
e, (123), (132)
_
. We saw in Example 8d) (page 7) that A
3
S
3
,
and we saw in Example 13c) (page 11) that
S
3
A
3

= Z
/2
. (5.5)
Observe that A
3
=
_
e, (123), (132)
_
is isomorphic to Z
/3
. Thus, we suspect S
3

= Z
/3
Z
/2
.
However, this is false, because Z
/3
Z
/2
is abelian, but S
3
is not abelian. So if S
3
is a product
of Z
/3
and Z
/2
, it is only a product in a metaphorical sense
1
. but we can still write:
S
3

= Z
/3
Z
/2
, (5.6)
as long as we interpret the symbol in the appropriate way. Thus, the quotient equation
(5.5) entails a factorization (5.6) of the group S
3
, just as the quotient equation (5.3) entailed a
factorization (5.4) of the group Z
/6
.
Examples (B) and (C) show an important dierence between factoring groups and factoring
numbers. Any number with prime factorization 2 3 must be equal to 6. But two dierent
groups can both factor into the same elementary components, without being the same. The
groups Z
/6
and S
3
both factor into elementary components Z
/3
and Z
/3
, but Z
/6
,

= S
3
.
5.2 Simple Groups
Prerequisites: 1.4, 1.2 Recommended: 5.1
A group ( is simple if it has no nontrivial proper normal subgroups. For example, Z
/3
is
simple (because the only proper subgroup of Z
/3
is 0, which is trivial). Simple groups are
the elementary components of group theory, and are analogous to prime numbers. Indeed, we
have the following:
Proposition 52 For any n N,
_
Z
/n
is simple
_

_
n is prime
_
.
1
In fact, S
3
is actually a semidirect product of Z
/3
and Z
/2
.
5.3. COMPOSITION SERIES; THE H

OLDER PROGRAM 43
Proof: = Let / be any subgroup of Z
/n
. Then Lagranges Theorem (Theorem 7 on
page 6) says [/[ divides [Z
/n
[ = n. But n is prime, so either [/[ = 1 (in which case / =

0)
or [/[ = n (in which case / = Z
/n
).
Thus, / has no nontrivial proper subgroups in particular, / has no nontrivial proper
normal subgroups. Hence, / is simple.
= Suppose n was not prime; then n = km for some numbers k and m, with 1 < k, m < n.
Thus, the subgroup m) =
_

0, m, 2m, . . . , (k 1)m
_
is a subgroup with k elements, so
it is a nontrivial subgroup of Z
/n
. But Z
/n
is abelian, so m) is automatically a normal
subgroup. Hence, Z
/n
cannot be simple. 2
In the proof of Proposition 52, we implicitly used the following result:
Lemma 53 Let ( be a group. Then:
(a)
_
( has no nontrivial subgroups
_
=
_
( is simple
_
.
(b) If ( is abelian, then
_
( has no nontrivial subgroups
_

_
( is simple
_
.
Proof: Exercise 23 2
Observe that the if and only if is not true if ( is not abelian. It isnt obvious, because
we havent yet seen any examples of nonabelian simple groups. We will show in 6.3 that the
alternating group A
N
is simple for all N 5. Thus, A
5
is an example of a simple group with
many nontrivial subgroups.
Lemma 54 Let ( be a simple group. If : (H is any group homomorphism, then either
is trivial or is a monomorphism.
Proof: Exercise 24 2
5.3 Composition Series; The Holder program
Prerequisites: 5.2, 2.1
Let ( be a group. A normal series is a sequence of subgroups:
^
1
^
2
^
3
. . . ^
J
(.
Note that ^
1
is normal in ^
2
, but it is not necessarily true that ^
1
is normal in (. Likewise,
for every j [1..J), ^
j
is normal in ^
j+1
, but not necessarily normal in (.
44 CHAPTER 5. THE STRUCTURE THEORY OF NONABELIAN GROUPS
N
0
={e}
N
1
=X {e} {e} N
2
=X Y {e}
N
3
=X Y Z G
S
2
=N
2
/N
1
=Y
S
1
=
N
1
/
N
0
=
X
S
3
=
N
3
/
N
2
=
Z
Figure 5.1: Example 56a)
Example 55: Let ( = Z
/120
, and let
^
3
=

2) =

0,

2,

4,

6,

8, . . . , 116, 118;
^
2
=

12
_
=

0, 12, 24, 36, 48, 60, 72, 84, 96, 108;

^
1
=

24
_
=

0, 24, 48, 72, 96.

Thus, ^
1
^
2
^
3
Z
/120
.
A composition series is a normal series
e
G
= ^
0
^
1
^
2
. . . ^
J
= (, (5.7)
such that the quotient groups
_

_
o
1
= ^
1
/^
0
= ^
1
o
2
= ^
2
/^
1
o
3
= ^
3
/^
2
.
.
.
o
J
= ^
J
/^
J1
= (/^
J1
_

_
are all simple groups.
We say this series has rank J. The groups o
1
, o
2
, . . . , o
J
are called composition factors, and
the set o
1
, o
2
, . . . , o
J
is called a composition factor set
2
.
The idea is that ( is in some way a product of its composition factors:
( = o
1
o
2
o
3
o
4
. . . o
J
where , , , etc. represent some ways of combining groups together (eg. direct product,
semidirect product, etc.). Thus, a composition series is for a group what a prime factorization
is for an integer.
Example 56:
2
Technically, this is a multiset, because we allow the same element to appear more than once.
5.3. COMPOSITION SERIES; THE H

OLDER PROGRAM 45
(a) Let A, , Z be simple groups, and let ( = A Z. As shown in Figure 5.1, dene:
^
3
= A Z = (x, y, c, e) ; x A, y and z Z
^
2
= A e
Z
= (x, y, e
Z
) ; x A and y
^
1
= A e
Y
e
Z
=
_
(x, e
Y
, e
Z
) ; x A
_
^
0
= e
X
e
Y
e
Z
= e
G

Then this is a composition series for (, with composition factors:

o
3
= ^
3
/^
2

= Z;
o
2
= ^
2
/^
1

= ;
o
1
= ^
1
/^
0

= A.
(b) Let ( = Z
/12
, and dene:
^
3
= Z
/12
=

0,

1,

2,

3,

4,

5

6,

7,

8,

9, 10, 11;
^
2
=

2) =

0,

2,

4,

6,

8, 10;
^
1
=

4) =

0,

4,

8;
^
0
=

0.
Then this is a composition series for Z
/12
, with composition factors:
o
3
= ^
3
/^
2

= Z
/2
;
o
2
= ^
2
/^
1

= Z
/2
;
o
1
= ^
1
/^
0

= Z
/3
.
(c) Again, let ( = Z
/12
, but now dene:
^
3
= Z
/12
=

0,

1,

2,

3,

4,

5

6,

7,

8,

9, 10, 11;
^
2
=

3) =

0,

3,

6,

9, ;
^
1
=

6) =

0,

6;
^
0
=

0.
Then this is another composition series for Z
/12
, with composition factors:
o
3
= ^
3
/^
2

= Z
/3
;
o
2
= ^
2
/^
1

= Z
/2
;
o
1
= ^
1
/^
0

= Z
/2
.
(d) The normal series in Example 55) is not a composition series (even if we add terms
^
4
= Z
/120
and ^
0
=

0). The reason: o

3
= ^
3
/^
2

= Z
/6
is not simple. However, we
can rene this normal chain into a composition series by inserting the normal subgroup
^
2
1
2
=

6) =

6, 12, 18, . . . , 124.

46 CHAPTER 5. THE STRUCTURE THEORY OF NONABELIAN GROUPS
Now we have a composition series:

0 = ^
0
^
1
^
2
^
2
1
2
^
3
^
4
= Z
/120
,
with composition factors:
_

_
o
4
= ^
4
/^
3

= Z
/2
;
o
3
= ^
3
/^
2
1
2

= Z
/3
;
o
2
1
2
= ^
2
1
2
/^
2

= Z
/2
;
o
2
= ^
2
/^
1

= Z
/2
;
o
1
= ^
1
/^
0

= Z
/5
.
.
(e) If ( is a simple group, then ( has a very short composition series: e
G
= ^
0
^
1
= (,
with one composition factor: o
1
= ^
1
/^
0

= (.
In Examples (56b) and (56c), we constructed two dierent composition series for Z
/12
, but
they had the same rank, and yielded the same composition factor set: Z
/2
, Z
/2
, Z
/3
(albeit
in a dierent order). This is reassuring; if a composition series is supposed to be a prime
factorization of a group, it would be unfortunate if dierent series yielded dierent factors for
the same group. The following theorem says this is true in general.
Theorem 57 (Jordan-Holder )
Let ( be a nite group. Then:
(a) ( has a composition series.
(b) Any two composition series for ( have the same rank, and yield the same composition
factor set. In other words, if
e
G
= ^
0
^
1
^
2
. . . ^
J
= (
and e
G
= /
0
/
1
/
2
. . . /
K
= (
are two composition series for (, with composition factors
o
1
= ^
1
/^
0
o
2
= ^
2
/^
1
o
3
= ^
3
/^
2
.
.
.
o
J
= ^
J
/^
J1
and
T
1
= /
1
//
0
T
2
= /
2
//
1
T
3
= /
3
//
2
.
.
.
T
K
= /
K
//
K1
then actually, J = K, and (up to reordering), o
1
, o
2
, . . . , o
J
= T
1
, T
2
, . . . , T
K
.
Proof: (a) (by complete induction on [([).
Base Case: ([([ = 2) The only group of order 2 is Z
/2
, which is simple and therefore has
composition series

0 = ^
0
^
1
= Z
/2
.
Induction: Suppose that every group of order less than n has a composition series, and let
[([ = n. We claim that ( must also have a composition series. This is Exercise 25
5.3. COMPOSITION SERIES; THE H

OLDER PROGRAM 47
Hint: there are two cases:
_
1. ( is simple.
2. ( is not simple.
If ( is simple, it has a very short composition series. If ( is not simple, let ^ be a maximal normal
subgroup of (. Then [^[ < n, so apply the induction hypothesis to ^ and proceed.
(b) (by induction on J)
Base Case: (J = 2) In this case, we have two composition series
e
G
= ^
0
^
1
^
2
= (;
and e
G
= /
0
/
1
/
2
. . . /
K
= (.
Our goal is to prove that K = 2, and that o
1
, o
2
= T
1
, T
2
. Consider the following
diamond:
Claim 1: /
K1
^
1
= (.
Claim 2: /
K1
^
1
= e
G
.
Claim 3: /
K1

= o
1
. Thus, /
K1
is simple.
Claim 4: K = 2, T
0

= o
1
, and T
1

= o
0
.
Exercise 26 Prove Claim 1 using Claim 14; prove Claim
2 using Claim 15. Prove Claims 3 and 4 using the Diamond
Isomorphism Theorem (Theorem 16 on page 14).

/
K1
^1
/
K1
^
1
/
K1
^
1
.
Induction: Suppose that part (b) is true for any group possessing a composition series
of rank J 1.
48 CHAPTER 5. THE STRUCTURE THEORY OF NONABELIAN GROUPS
Dene

( = /
K1
^
J1
.
For all k [1..K), dene

/
k
= /
k


(.
For all j [1..J), dene

^
j
=

( ^
j
.
Claim 5:

/
1
= e
G
=

^
1
.
Claim 6:
e
G
=

^
1

^
2
. . .

^
J1
=

(
and
e
G
=

/
1

/
2
. . .

/
K1
=

(
are composition series for the group

(.
Let the corresponding composition factors be:

o
J1
=

^
J1
/

^
J2

o
J2
=

^
J2
/

^
J3
.
.
.

o
3
=

^
3
/

^
2

o
2
=

^
2
/

^
1
and

T
K1
=

/
K1
/

/
K2

T
K2
=

/
K2
/

/
K3
.
.
.

T
3
=

/
3
/

/
2

T
2
=

/
2
/

/
1
Claim 7: J = K, and

o
2
,

o
3
, . . . ,

o
J1
=

T
2
,

T
3
, . . . ,

T
K1
.
Claim 8:
e
G
=

^
1

^
2

^
3
. . .

^
J1
=

( ^
J
is a composition series for the group ^
J
.

` `

` `
` `
`

`
`
`

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

`
`
`
`
`
`

` `
`
`
`
`
`
`
`'

/
K1
^
J1
/
K1
^
J1
^
j2
^
j3
/
2
^
2
/
k3

/
k3

^
j2

^
j3

^
2

/
2

/
k2
/
k2

(
/
1
^
1
/
1
^
1
Dene

o
J
= ^
J
/

(
Claim 9:

o
2
,

o
3
, . . . ,

o
J1
,

o
J
= o
1
, o
2
, . . . , o
J2
, o
J1
.
Claim 10: e
G
=

/
1

/
2

/
3
. . .

/
K1
=

( /
K
is a composition series for the group /
K
.
Dene

T
K
= /
K1
/

(.
Claim 11:

T
2
,

T
3
, . . . ,

T
K1
,

T
K
= T
1
, T
2
, . . . , T
K2
, T
K1
.
Claim 12: /
K1
^
J1
= (.
Claim 13:

o
J
= T
K
, and

T
K
= o
J
.
5.3. COMPOSITION SERIES; THE H

OLDER PROGRAM 49
The proofs of several of these claims depend upon certain basic facts about normal subgroups.
A maximal normal subgroup of ( is a normal subgroup / ( so that there exists no
normal subgroup B ( with / < B.
Claim 14: (a) ^
j1
and /
k1
are maximal normal subgroups of (.
(b) If ^ and / are distinct maximal normal subgroups of (, then ^/ = (.
A minimal normal subgroup of ( is a normal subgroup /( which is also simple ie.
there exists no nontrivial normal subgroup B /.
Claim 15: (a) ^
1
and ^
1
are minimal normal subgroups of (.
(b) If ^ and / are distinct minimal normal subgroups of (, then ^ / = e
G
.
Finally, were using the following fact about composition series:
Claim 16: Let e
G
= /
1
/
2
. . . /
R
= ( be any composition series for (, and let
H < (. For all r [1..R], let B
r
= /
r
H. Then e
G
= B
1
B
2
. . . B
R
= H is a
composition series for B.
Exercise 27 First prove Claims 14, 15, and 16. Then prove Claim 5 using Claim 14. Prove Claim
12 using Claim 15. Prove Claims 6, 8, and 10 using Claim 16. Prove Claims 7, 9, and 11 using
the induction hypothesis. Prove Claim 13 using the Diamond Isomorphism Theorem (Theorem 16 on
page 14). Now bring all this together to prove the theorem. 2
The Jordan-Holder theorem means that we can talk about the composition factors of the
group (, without making reference to any specic composition series. For instance, as we saw in
Examples (56b) and (56c), the composition factor set for Z
/12
is Z
/2
, Z
/2
, Z
/3
. Is this because
the integer 12 has prime factor set 2, 2, 3? Yes; the same pattern holds for any abelian group...
Proposition 58 Let m N have prime factorization: m = p

1
1
p

2
2
p

J
J
(where p
1
, . . . , p
J
are prime, and
1
, . . . ,
J
N). Then any abelian group of order m has composition factor set

1
..
Z
/p
1
, . . . , Z
/p
1
,

2
..
Z
/p
2
, . . . , Z
/p
2
, . . . ,

J
..
Z
/p
J
, . . . , Z
/p
J
.
Proof: We will prove this theorem by complete induction on m.
Claim 1: If n = p

1
1
p

2
2
p

J
J
, then the cyclic group Z
/n
has composition factor set

1
..
Z
/p
1
, . . . , Z
/p
1
,

2
..
Z
/p
2
, . . . , Z
/p
2
, . . . ,

J
..
Z
/p
J
, . . . , Z
/p
J
.
Proof: Exercise 28 ................................................. 2 [Claim 1]
Claim 2: Let ( be a group; let ^ ( be a normal subgroup, and let Q be a quotient
group. If ^ has composition factor set ^
1
, ^
2
, . . . , ^
L
, and Q has composition factor set
Q
1
, Q
2
, . . . , Q
K
, then ( has composition factor set ^
1
, ^
2
, . . . , ^
L
, Q
1
, Q
2
, . . . , Q
K
.
50 CHAPTER 5. THE STRUCTURE THEORY OF NONABELIAN GROUPS
Proof: Exercise 29 Hint: Suppose Q had composition series: e = 1
0
1
1
. . .1
K
= Q,
with 1
k
/1
k1
= Q
k
, for all k [1..K]. Use the Chain Isomorphism Theorem (Theorem 18 on
page 15) to transform this into a normal series
^ =

1
0


1
1


1
2
. . .

1
K
= (,
such that

1
k
/

1
k1
= 1
k
/1
k1
= Q
k
, for all k [1..K].
Now suppose ^ had composition series: e = /
0
/
1
. . . /
L
= ^, such that
/

//
1
= ^

, for all [1..L]. Conclude that ( has composition series:

e = /
0
/
1
. . . /
L
= ^ =

1
0

1
1
. . .

1
K
= (. ............ 2 [Claim 2]
Base Case: (m = 2) The only group of order 2 is Z
/2
; its only composition factor is Z
/2
.
Induction: Suppose the theorem is true for all abelian groups of order less than m,
and let / be an abelian group of order m. Let a / be any nontrivial element, and
let ^ = a). Then ^ is a cyclic subgroup of order n, for some n < m, and Lagranges
Theorem (Theorem 7 on page 6) says that n divides m. Thus, n = p

1
1
p

2
2
p

J
J
for some

1

1
,
2

1
, . . .,
J

J
. Thus, Claim 1 says that ^ has composition factor set
Z
/p
1
, . . . , Z
/p
1
. .

1
, . . . , Z
/p
J
, . . . , Z
/p
J
. .

J
.
Now, / is abelian, so ^ is automatically normal in /. Let Q = //^. Then Q is an abelian
group of order q = m/n = p

1
1
p

2
2
p

J
J
, where
1
=
1

1
, . . .,
J
=
J

J
.
Thus, Q is an abelian group of order less than m, so the induction hypothesis says that Q has
composition factor set Z
/p
1
, . . . , Z
/p
1
. .

1
, . . . , Z
/p
J
, . . . , Z
/p
J
. .

J
. Now apply Claim 2 to conclude
that ( has composition factor set Z
/p
1
, . . . , Z
/p
1
. .

1
+
1
, . . . , Z
/p
J
, . . . , Z
/p
J
. .

J
+
J
, as desired. 2
The H older Program: The Jordan-Holder theorem is the starting point of the Holder pro-
gram, a master strategy to build an atlas of all nite groups, in two stages:
1. Construct a complete list of all nite simple groups.
2. Characterize all possible ways that two groups ^ and Q can be combined to create a
larger group, (, such that ^ (, and (/^ = Q.
Stage 1 of the Holder program is called the Classication of Finite Simple Groups, and was
a massive, century-long eort which was nally completed in 1980 (see 5.4). If we think of
groups as molecules, then simple groups are atoms, and the Classication is analogous to
Mendeleevs construction of a Periodic Table of Elements.
Stage 2 is called the Extension Problem, and is still not completely solved (see 5.5). If we
think of groups as molecules, and simple groups as atoms, then the Extension Problem is
analogous to the entire subject of chemistry.
5.4. () THE CLASSIFICATION OF FINITE SIMPLE GROUPS 51
Renement: In Example 56d) on page 45, we saw that a normal series of Z
/120
could
be rened into a composition series, by judiciously inserting additional terms in the series.
Another example of this sort of renement is Claim 2 of Proposition 58, which we generalize
as follows:
Proposition 59 Renement Lemma
Let ( be a group, with a normal series: e = ^
0
^
1
^
2
. . . ^
J
= (.
Suppose that :
(
1
= ^
1
/^
0
has composition series: e = /
1
0
/
1
1
/
1
2
. . . /
1
K
1
= (
1
.
(
2
= ^
2
/^
1
has composition series: e = /
2
0
/
2
1
/
2
2
. . . /
2
K
2
= (
2
.
.
.
.
.
.
.
.
.
.
(
J
= ^
J
/^
J1
has composition series: e = /
J
0
/
J
1
/
J
2
. . . /
J
K
J
= (
J
.
Then ( has a composition series:
e =

/
1
0


/
1
1


/
1
2
. . .

/
1
K
1
= ^
1
=

/
2
0


/
2
1
. . .

/
2
K
2
= ^
2
=

/
3
0


/
3
1
. . .

/
3
K
3
= ^
3
=

/
4
0


/
4
1
. . .
. . . ^
J1
=

/
J
0


/
J
1


/
J
2
. . .

/
J
K
J
= ^
J
= (
such that, for all j [1..J] and k [1..K
j
],
_

/
k
j
/

/
k
j1
_

=
_
/
k
j
//
k
j1
_
. Hence,
_
_
_
(
1
has composition factor set o
1
1
, o
1
2
, . . . , o
1
K1

(
2
has composition factor set o
2
1
, o
2
2
, . . . , o
2
K2

.
.
.
.
.
.
.
.
.
(
J
has composition factor set o
J
1
, o
J
2
, . . . , o
J
KJ

_
_
_
=
_
_
_
( has composition factor set
o
1
1
, o
1
2
, . . . , o
1
K1
, o
2
1
, o
2
2
, . . . , o
2
K2
, . . . ,
. . . , o
J
1
, o
J
2
, . . . , o
J
KJ

_
_
_
Proof: Exercise 30 Hint: Generalize the proof of Claim 2 of Proposition 58. For each j [1..J],
use the Chain Isomorphism Theorem (Theorem 18 on page 15) to transform the normal series
e = /
j
0
/
j
1
/
j
2
. . . /
j
K
j
= (
j
into a normal series ^
j1
=

/
j
0


/
j
1


/
j
2
. . .

/
j
K
j
= ^
j
. 2
5.4 () The Classication of Finite Simple Groups
Prerequisites: 5.2 Recommended: 5.3
Part 1 of the Holder program called for a complete list of all nite simple groups, a project
which ultimately spanned one hundred years and generated almost ten thousand pages of
published research. At this point, the Classication is complete, but the details are beyond the
scope of these notes. Instead, we will simply summarize the main results:
52 CHAPTER 5. THE STRUCTURE THEORY OF NONABELIAN GROUPS
1. There are 18 innite families of nite simple groups. Each of these families is a collection
of simple groups built from some common template, indexed by one or more natural
numbers. For example, here are three families of simple groups:
(a) The prime cyclic groups:
_
Z
/p
; where p is any prime number
_
. (Proposition 52 on
page 42).
(b) The alternating groups: A
N
; where N 5. (see 6.3).
(c) The nite projective groups [1, Thm 8.8.3]:
_
PSL
n
(F) ; where F is any nite eld, and n 2 (with n 3 if F = Z
/2
or Z
/3
)
_
,
Here, PSL
n
(F) = SL
n
[F]/Z, where SL
n
[F] is the group of F-valued matrices with
determinant 1, and Z = Z (SL
n
[F]) = Id is its center.
2. There are also 26 sporadic simple groups, which do not t into any of these families.
3. The smallest of these sporadic simple groups is the Matthieu group, M
11
, which has
cardinality 7920.
4. The largest sporadic simple group is called the Monster, and has cardinality approximately
10
53
. The mathematical footsteps the Monster have been discovered in surprisingly
diverse areas of mathematics. The signicance of the Monster is not well understood; the
study of its footsteps is sometimes called (Monstrous) Moonshine. [?]
5. Feit-Thompson Theorem: The only simple groups of odd order are the prime cyclic
groups Z
/p
(where p is prime).
(the proof of the Feit-Thompson is 255 pages long, and occupies an entire volume of
Pacic Journal of Mathematics).
6. Likewise, the only abelian simple groups are the prime cyclic groups.
(This follows easily from Fundamental Theorem of Finitely Generated Abelian Groups; see
4.3).
7. The smallest nonabelian simple group is the icosahedral group
3
A
5
, (which is also PSL
2
(Z
/5
)),
which has order 60 (see 6.3). The next largest nonabelian simple groups have orders 168,
360, 504, 660, 1092, and 2448.
5.5 () The Extension Problem; Short Exact Sequences
Prerequisites: 1.4 Recommended: 5.1
If ( is a group, and ^ (, and Q = (/^ is the quotient group, then ( is somehow a
combination of ^ and Q. We call ( an extension of Q by ^.
3
So called because it is the group of symmetries of the regular icosahedron and regular dodecahedron.
5.5. () THE EXTENSION PROBLEM; SHORT EXACT SEQUENCES 53
Now consider the inverse problem: Given two groups ^ and Q, what are all the ways in
which ^ and Q can be combined to yield a group (, so that ^ ( and Q

= (/^? That
is, what are all the extensions of Q by ^? The most obvious extension of Q by ^ is just
the direct product, ( = ^ Q. However, this is rarely the only extension. The problem of
characterizing these extensions is the Extension Problem.
An extension is sometimes described using a short exact
sequence, a sequence of two homomorphisms and three
groups:
^

Q,
such that:
1. : ^ ( is a monomorphism.
2. : ( Q is an epimorphism.
3. (^) = ker().
e
Q
e
G
k
e
r
(

e
N

G
N
Q
In other words, if

^ = (^), then

^ is a normal subgroup of (, there is an isomorphism
^

=

^ (via ), and Q

= (/

^. The advantage of this notation is that we can apply the
following lemma.
Proposition 60 The Five Lemma
Let (^

Q) and (

^

Q) be two short exact sequences of groups, linked
together in the following commuting diagram:

(
( Q

^
^

If any two of the morphisms , , and are isomorphisms, then the third one is, as well.
In particular, suppose that ( and

( are two extensions of Q by ^, and that there is a
homomorphism : (

( so that the following diagram commutes:

-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
--
.
.
.
.
.
.
.
..
.
.
.
.
.
.
.

(
(
^ Q

54 CHAPTER 5. THE STRUCTURE THEORY OF NONABELIAN GROUPS

e
2
e
1
k
e
r
(

2 )

1
G
1
k
e
r
(

3
)
e
4
e
3
k
e
r
(

4 )

3
G
3
k
e
r
(

5
)
e
6
e
5
k
e
r
(

6 )

5
G
5
e
0
k
e
r
(

1
)
G
0
G
2
G
4
G
6
Figure 5.2: An exact sequence
Then is an isomorphism, so that (

=

(.
Proof: Exercise 31 2
The Extension Problem is an important part of the Holder program (5.3), and also arises
frequently in algebraic topology, dierential geometry, and algebraic geometry, where we often
characterize the shape of a space A using group ( (eg. a homotopy group, (co)homology group,
holonomy group, etc.). Our analysis of A often provides only indirect information about (,
in the form of two groups ^ and Q so that ( is an extension of Q by ^. Thus, to compute (,
we must solve the extension problem.
5.6 () Exact Sequences
Prerequisites: 5.5
An morphism sequence is a sequence of groups and group homomorphisms:
(
0

1
(
1

2
(
2

3
. . .

N
(
N
.
We say this sequence is exact if (as shown in Figure 5.2),
image [
1
] = ker(
2
), image [
2
] = ker(
3
), . . . . . . , image [
N1
] = ker(
N
).
For example, a short exact sequence is an exact sequence containing exactly 3 nontrivial ele-
ments:
0

^

Q

0.
In other words:
5.7. SOLVABILITY 55
1. ker() = (0) = e hence, is a monomorphism.
2. (Q) = 0 hence, ker() = Q.
3. (() = ker() = Q hence, is an epimorphism.
4. ker() = (^).
5.7 Solvability
Prerequisites: 5.3
A group ( is called solvable if it has a normal series:
e
G
= ^
0
^
1
^
2
. . . ^
J
= (. (5.8)
such that the quotient groups
_

_
/
1
= ^
1
/^
0
= ^
1
/
2
= ^
2
/^
1
/
3
= ^
3
/^
2
.
.
.
/
J
= ^
J
/^
J1
= (/^
J1
_

_
are all abelian groups.
The series (5.8) is called a solution series for (. Note that the solution series (5.8) is not
necessarily a composition series, because the groups /
1
, . . . , /
J
are not necessarily simple.
Example 61:
(a) Any abelian group is solvable. If ( is an abelian group, then the normal series
0 = ^
0
^
1
= (
is a solution series, because /
1
= ^
1
/^
0
= ( is abelian.
(b) Let ( = S
3
. Then S
3
is nonabelian, but it is still solvable. To see this, let ^
1
= A
3
.
Then the normal series 0 = ^
0
^
1
^
2
= ( is a solution series, because
/
1
= ^
1
/^
0
= A
3

= Z
/3
is abelian.
and /
2
= ^
2
/^
1
= S
3
/A
3

= Z
/2
is abelian.
(c) For any N N, the dihedral group D
2N
is nonabelian but solvable. To see this, let be
a rotation and be a reection, so that D
2N
is generated by , . Let ^
1
= )

=
Z
/N
. Then ^
1
D
2N
, and D
2N
/^
1
is a two-element group, isomorphic to Z
/2
. Thus,
/
1
= ^
1
/e = Z
/N
is abelian, and /
2
= D
2N
/^
1
= Z
/2
is abelian.
Theorem 62 Let ( be a group. The following are equivalent:
56 CHAPTER 5. THE STRUCTURE THEORY OF NONABELIAN GROUPS
(a) ( is solvable.
(b) The composition factors of ( are all cyclic groups of prime order.
(c) ( has a normal series e
G
= ^
0
^
1
^
2
. . . ^
J
= (. where each of
the factors /
j
= ^
j
/^
j1
is a cyclic group.
(d) ( has a solution series e
G
= ^
0
^
1
^
2
. . . ^
J
= ( so that ^
j
is
not only normal in ^
j+1
, but is actually normal in (.
Proof: (a)=(b) Exercise 32 Hint: combine Proposition 58 on page 49 with Proposition 59 on
page 51.
(b)=(c) follows immediately from the denition of composition factors.
(c)=(d) Exercise 33 .
(d)=(a) is immediate. 2
Chapter 6
Group Actions
6.1 Introduction
Let / be a set, and let S
A
= : // ; any bijection be the group of all permutations
of /. Thus, if / = [1..N], then S
A
= S
N
is the Nth symmetry group.
If G is a group, then an action of G upon / is a group homomorphism : GS
A
. In
other words, to every element g G, associates a permutation
g
: //. This is a way
of realizing G as a group of transformations of /.
Notation: If a / and g G, we usually write g.a to indicate the
g
(a). This notation
has a convenient associativity property: If g, h G, then
(g h).a =
gh
(a) =
g

h
(a) =
g
_

h
(a)
_
=
g
(h.a)
= g.(h.a).
Because of this associativity, we can be careless about bracketting when describing group
actions: there is no ambiguity in just writing g h.a.
Note: The product g.a is just a notational convention, and does not really represent
multiplication, since / is not, in general, a group.
Example 63:
(a) Let G = S
N
and let / = [1..N]. Then S
N
acts on [1..N] by permutations. For example,
if g = (123) and a = 2, then g.a = 3.
(b) Permutation groups: Let H < S
N
be any subgroup of S
N
. Then H also acts on [1..N].
A subgroup of S
N
is called a permutation group.
(c) Let G = D
25
be the symmetries of a pentagon, and let / = 1, 2, 3, 4, 5 represent the
ve vertices of the pentagon. Then G acts by permuting these vertices. For example, if
is counter-clockwise rotation, then .1 = 2, .2 = 3, etc.
57
58 CHAPTER 6. GROUP ACTIONS
(d) Rotation: Let G = S
1
be the unit circle group, and let / = R
2
. Then S
1
acts on R
2
by rotations about zero. To be precise, if S
1
is any angle (between 0 and 2), and
v =
_
v
1
v
2
_
R
2
, then .v =
_
c s
s c

_
v
1
v
2
_
, where s = sin() and c = cos().
(e) Conjugation: Any group acts upon itself by conjugation. Here, / = G, and for any
g G, we dene
g
: GG to be the map:
g
(a) = gag
1
.
(f) Left-multiplication: Any group acts upon itself by left-multiplication. Here, / = G,
and for any g G, we dene
g
: GG to be the map:
g
(a) = g a.
(g) Let A be a group, and let G A be a subgroup. Then G acts on A by left-multiplication.
For any g G, we dene
g
: AA to be the map:
g
(a) = g a.
6.2 Orbits and Stabilizers
Prerequisites: 6.1
If a /, then the orbit of a is the set
Orbit
G
(a) = g.a ; g G
Sometimes the orbit of a is denoted G.a.
Example 64:
(a) Let G = S
1
be the unit circle group, acting on / = R
2
by rotations about zero, as in
Example 63d). Let v = (v
1
, v
2
), and let r = |v| =
_
v
2
1
+ v
2
2
be the norm of v. Then
Orbit (v) =
_
w R
2
; |w| = r
_
is the circle of radius r about zero. In particular, note that Orbit ((0, 0)) = (0, 0).
(b) Conjugation: Let G act upon itself by conjugation, as in Example 63e). If a G,
then the orbit of a is the set:
/(a) =
_
gag
1
; g G
_
This is called the conjugacy class of a. Observe that, if a Z (G), then /(a) = a.
(c) Let A be a group, and let G A be a subgroup acting on A by left-multiplication. as
in Example 63g). For any a G, Orbit (a) is just the right coset G.a.
Exercise 34 Verify these examples.
Lemma 65 Let G act on /.
6.2. ORBITS AND STABILIZERS 59
(a) For any a /, a Orbit (a).
(b) For any a, b, c /, if a Orbit (b) and b Orbit (c), then a Orbit (c).
(c) For any a, b /, the following are equivalent:
a Orbit (b).
b Orbit (a).
Orbit (a) = Orbit (b).
(d) / is a disjoint union of orbits. For example, if / is nite, then / = O
1
. O
2
. . . . . O
M
,
where O
m
= Orbit (a
m
) for some a
m
/.
Proof: (Exercise 35) 2
We say that the action of G on / is transitive if / = Orbit (a) for some a /.
Lemma 66 The following are equivalent:
1. G acts transitively on /.
2. For any a /, / = Orbit (a).
3. For any a, b /, there is some g G so that g.a = b.
Proof: (Exercise 36) 2
Example 67:
(a) Let G = S
N
act on / = [1..N], as in Example 63a). This action is transitive. To see
this, let m, n [1..N]. If = (nm), then .n = m.
(b) Let G = D
25
act on / = 1, 2, 3, 4, 5, as in Example (63c). Then this action is
transitive.
(c) Left-multiplication: Let G act upon itself by left-multiplication, as in Example 63f).
This action is transitive. To see this, let a, b G. Let g = b a
1
. Then g.a = b.
If a /, then the stabilizer of a is the set of elements which act trivially on a:
Stab (a) = g G ; g.a = a
Sometimes the stabilizer of a is denoted G
a
.
Example 68:
60 CHAPTER 6. GROUP ACTIONS
(a) Let G = S
4
, acting by permutations on / = [1..4], as in Example 63a). Then Stab (4) is
the set of all permutations which x 4. Explicitly,
Stab (4) =
_
e, (12), (13), (15), (23), (25), (35), (12)(35), (13)(25), (15)(23), (123),
(125), (135), (132), (152), (153), (235), (253), (1235), (1532),
(1325), (1523), (1352), (1253)
_
.
(b) Let G = S
1
be the unit circle group, acting on / = R
2
by rotations about zero, as in
Example 63d). If v is any nonzero vector, then Stab (v) = e. However, Stab ((0, 0)) =
S
1
ie. all of S
1
acts trivially upon (0, 0).
(c) Conjugation: Let G act upon itself by conjugation, as in Example 63e). If a G,
then the stabilizer of a is the set:
c
G
(a) = g G ; ga = ag
This is called the centralizer of a in G. Observe that, if a Z (G), then c
G
(a) = G.
Lemma 69 If b = g.a, then Stab (b) = g Stab (a) g
1
.
Proof: (Exercise 37) 2
We say that the action of G upon / is free if Stab (a) = e for all a /. In other words,
every element of G acts nontrivially on every element of /.
Example 70:
(a) Let G act on itself by left-multiplication, as in Example (63f). This action is free: for
any g G and a G, if g.a = a, then g = e by left-cancellation.
If G acts on the set / via a homomorphism : GS
A
, then the kernel of the action is
the kernel of . In other words, the kernel is the set of all elements in G which act trivially
upon all elements of /.
Example 71: Let G act upon itself by left-multiplication, as in Example 63e). The kernel
of this action is just the center of G: Z (G) = g G ; ga = ag for all a G.
Lemma 72 Let G act on /, and let K be the kernel of the action. Then K =

aA
Stab (a).
In particular, if G acts freely, then K = e, so the map : GS
A
is a monomorphism.
Proof: (Exercise 38) 2
6.2. ORBITS AND STABILIZERS 61
Orbit(a)
G
A
a
S
=
S
t
a
b
(
a
)
g.a
g
.
S
Figure 6.1: The Orbit-Stabilizer Theorem
Corollary 73 (Cayleys Theorem)
Every group is isomorphic to a permutation group. In particular, if [G[ = N, then G is
isomorphic to a subgroup of S
N
.
Proof: Let G act upon itself by left-multiplication. Let : GS
G
be the associated
homomorphism, and let I = (G) S
G
.
As we saw in Example 71), this action is free. Thus, Lemma 72 says that ker() = e;
hence, is an isomorphism between G and I. 2
Cayleys theorem is nice because it establishes a connection between abstract (algebraic)
groups and concrete (transformational) groups. It also says that the permutation groups have
a sort of Universal Property: in a sense, every group is a subgroup of some permutation group.
Thus, if one knew everything about permutation groups, then one would know everything about
all groups. Although philosophically interesting, this result has surprisingly little application.
Partly this is because it suggests that we study a group of order N by embedding it in a (much
larger) group of order N!.
You may have noticed, in the previous examples, that the larger the stabilizer of a gets, the
smaller the orbit of a becomes. The precise formulation of this is as follows:
Theorem 74 (Orbit-Stabilizer Theorem)
Let G act on /, and let a /. Let S = Stab (a).
1. There is a bijection between the elements of Orbit (a) and the cosets of S in G, given by:
: G/S gS g.a Orbit (a) (see Figure 6.1)
62 CHAPTER 6. GROUP ACTIONS
2. Thus, [Orbit (a)[ = [G/S[. If G is nite, then [Orbit (a)[ =
|G|
|S|
.
Proof:
Claim 1: For any g, h G,
_
g.a = h.a
_

_
h g.S
_
.
Proof:
_
g.a = h.a
_

_
a = g
1
.h.a
_

_
g
1
h Stab (a) = S
_

_
h gS
_
. ......................................................... 2 [Claim 1]
(1) is well-dened and injective: Let g, h G. We want to show
_
gS = hS
_

_
(gS) = (hS)
_
. But,
_
gS = hS
_

_
h gS
_
Claim1
_
g.a = h.a
_

_
(gS) = (hS)
_
.
is surjective: Let b Orbit (a). Then b = g.a for some g. Thus, b = (gS).
(2) The rst equality follows from (1). The second is just Lagranges Theorem. 2
Example 75: Conjugacy vs. Centralizers
Let G act upon itself by conjugation, as in Example (63e). If a G, then Example (68c)
says that the stabilizer of a is its centralizer:
Stab (a) = c
G
(a) = g G ; ga = ag
Let /(a) = Orbit (a) be the conjugacy class of a, as in Example (64b). Then the Orbit-
Stabilizer Theorem tells us: [/(a)[ =
[G[
[c
G
(a)[
.
Corollary 76 (The Class Equation)
Let G be a nite group. Then
[G[ = [Z (G)[ +
[G[
[c
G
(a
1
)[
+
[G[
[c
G
(a
2
)[
+ . . . +
[G[
[c
G
(a
N
)[
where a
1
, . . . , a
N
represent the distinct nontrivial conjugacy classes of G.
Proof: Lemma 65(d) says that G is a disjoint union of conjugacy classes:
G = /
1
. /
2
. . . . . /
M
(6.1)
6.3. THE SIMPLICITY OF A
N
63
Recall that, if z Z (G), then /(z) = z. Suppose Z (G) = z
1
, z
2
, . . . , z
L
. Let N =
M L, assume without loss of generality that /
N+1
= z
1
, /
N+2
= z
2
, , . . . , /
M
= z
L
.
Then we can rewrite eqn. (6.3) as:
G = /
1
. /
2
. . . . . /
N
. z
1
. z
2
. . . . . z
L
= /
1
. /
2
. . . . . /
N
. Z (G) .
Thus,
[G[ = [/
1
[ +[/
2
[ + . . . +[/
N
[ +[Z (G)[ . (6.2)
For all n [1..N], suppose that /
n
= /(a
n
) for some a
n
G. Then Example 75 says that
[/
n
[ =
[G[
[c
G
(a
n
)[
. (6.3)
Substituting eqn. (6.3) into eqn. (6.2) yields the Class Equation. 2
Corollary 77 If [G[ = p
k
for some prime p, then Z (G) is nontrivial.
Proof: Let Z = [Z (G) [. Since e Z (G), we know that Z 1. Thus, the Class Equation
reads:
p
k
= Z +
p
k
C
1
+
p
k
C
2
+ . . . +
p
k
C
N
where C
n
= [c
G
(a
n
) [. Observe that all the terms
p
k
Cn
must be divisible by p, and of course,
p
k
is divisible by p. Thus, Z must also be divisible by p. Since Z 1, we conclude that
Z = p
m
for some m 1. 2
6.3 The Simplicity of A
N
Prerequisites: 5.2, 6.2 Recommended: 5.3
To realise the Holder program, we must identify all nite simple groups. So far, the only sim-
ple groups we know are the prime-power cyclic groups: Z
/2
, Z
/3
, Z
/5
, Z
/7
, Z
/11
, Z
/13
, Z
/17
, . . ..
All of these are abelian. Are there any nonabelian simple groups?
The answer is, Yes. The smallest nonabelian simple group is the alternating group A
5
,
which has order 60 (also called the icosahedral group, because it is the group of symmetries of
the regular icosahedron and regular dodecahedron). We will show that, for any N 5, the
group A
N
is simple.
Let G be a group. If g G, recall that the conjugacy class of g is the set /(g) =
hgh
1
; h G. If g Z (G), then /(g) = g. In particular, /(e) = e.
64 CHAPTER 6. GROUP ACTIONS
Recall from Lemma 65(d) (page 58) that G is a disjoint union of its conjugacy classes
1
. In
other words,
G = Z (G) . /
1
. /
2
. . . . . /
J
, (6.4)
where, for j [1..J], /
j
= /(g
j
) for some g
j
G.
Lemma 78 Let G be a group with conjugacy class decomposition (6.4), and let N G be
a normal subgroup.
(a) If g N, then /(g) N.
(b) Thus, N is a disjoint union of conjugacy classes. In other words, N = Y./
j
1
.. . . ./
j
I
,
where Y Z ((), and where j
1
, j
2
, . . . , j
I
[1..J] is some subcollection.
(c) Thus, [N[ = Y + K
j
1
+ K
j
2
+ . . . + K
j
I
, where Y = [Y[, and K
j
= [/
j
[ for j [1..J].
Proof: Exercise 2
Corollary 79 (Combinatorial Simplicity Criterion)
Let G be a group with conjugacy class decomposition (6.4), and let K
j
= [/
j
[ for j [1..J].
Let Z = [Z (G)[.
Suppose that there exists no number Y [1..Z], and no proper subcollection j
1
, j
2
, . . . , j
I

[1..J] so that the sum (Y + K
j
1
+ K
j
2
+ . . . + K
j
I
) divides [G[. Then G is simple.
Proof: If N was any proper normal subgroup of G, then [N[ = Y +K
j
1
+K
j
2
+. . . +K
j
I
,
as in Lemma 78(c). Here, Y = [Y[, where Y Z (G) and e Y. Thus, 1 [Y[ Z.
On the other hand, N is a subgroup of G, so by Lagranges theorem, [N[ must divide [G[.
If no sum (Y + K
j
1
+ K
j
2
+ . . . + K
j
I
) divides [G[, then no such N can exist. 2
Example 80: Suppose G was a group of order 60, with trivial center. Thus, Z = 1. Suppose
the conjugacy classes of G are /
1
, /
2
, /
3
, and /
4
, where K
1
= [/
1
[ = 20, K
2
= [/
2
[ = 12,
K
3
= [/
3
[ = 12, and K
4
= [/
4
[ = 15. Observe that none of the numbers
1 + 20 1 + 12 1 + 20 + 12 1 + 15 1 + 20 + 15
1 + 20 + 12 1 + 20 + 12 + 15 1 + 12 + 12 1 + 15 + 12 + 12 1 + 20 + 12 + 12
divides 60. Thus, G is simple by Corollary 79.
1
See also the proof of the Class Equation, Corollary 76 on page 62.
6.3. THE SIMPLICITY OF A
N
65
We will prove that A
5
is simple by applying Corollary 79. Indeed, we will show that
A
5
satises exactly the description of Example 80). First we need a way of measuring the
conjugacy classes of A
5
. To start, well look at conjugacy classes in S
5
(which are not the same,
but are related).
Lemma 81 (Conjugacy in S
N
)
Let , S
N
. If = (a
1
a
2
. . . a
n
)(b
1
b
2
. . . b
m
) . . . (c
1
c
2
. . . c

),
then
1
=
_
(a
1
)(a
2
) . . . (a
n
)
__
(b
1
)(b
2
) . . . (b
m
)
_
. . .
_
(c
1
)(c
2
) . . . (c

)
_
.
Proof: Exercise 2
Example 82: Suppose = (1 2 3) (4 5) (6 7), and = (1 5 2 3) (7 4 6). Then

1
= (5 3 1) (6 2) (7 4).
Let S
N
, and write as a product of disjoint cycles:
= (a
1
a
2
. . . a
n
)(b
1
b
2
. . . b
m
) . . . (c
1
c
2
. . . c

) (6.5)
Assume without loss of generality that n m . . . . The cycle type of is the list of
numbers (n, m, . . . , ). For example, if = (1 2 3) (4 5 6 7) (8 9 10 11) (12 13 14 15 16), then
its cycle type is (3, 4, 4, 5)
Proposition 83 Let ,
1
S
N
. Then
_
and

are conjugate in S
N
_

_
and

have the same cycle type

_
Proof: Suppose = (a
1
a
2
. . . a
n
)(b
1
b
2
. . . b
m
) . . . (c
1
c
2
. . . c

).
=: Suppose

=
1

1
. Let a

1
= (a
1
), a

2
= (a
2
), etc. Then Lemma 81 says that

= (a

1
a

2
. . . a

n
)(b

1
b

2
. . . b

m
) . . . (c

1
c

2
. . . c

), which clearly has the same cycle type as .

=: Suppose

has the same cycle type as . Thus,

= (a

1
a

2
. . . a

n
)(b

1
b

2
. . . b

m
) . . . (c

1
c

2
. . . c

),
for some elements a

1
, a

2
, . . . , a

n
, b

1
, . . . , b

m
, . . . [1..N]. Let be the permutation dened by
the property: (a
1
) = a

1
, (a
2
) = a

2
, etc. Then Lemma 81 says that
1
=

. 2
Corollary 84 Let S
N
have cycle type (m
1
, m
2
, . . . , m
k
), and let M = m
1
+m
2
+. . .+m
K
.
Then [/() [ =
N!
(m
1
m
2
. . . m
k
) (N M)!
.
66 CHAPTER 6. GROUP ACTIONS
Proof: To keep notation simple, well give the proof when has cycle type (n, m, ) the
general case is much the same. By Proposition 83, /() consists of all permutations with
this same cycle type. How many of these are there? I want to build a permutation with cycle
decomposition
(a
1
a
2
. . . a
n
)(b
1
b
2
. . . b
m
)(c
1
c
2
. . . c

)
where a
1
a
2
. . . a
n
, b
1
b
2
. . . b
m
, c
1
c
2
. . . c

are distinct elements of [1..N]. I have N choices for

a
1
. Once a
1
is chosen, I have (N 1) choices left for a
2
. Then I have (N 2) choices left for
a
3
, and so on. At the end, I have (N M + 1) choices for c

, where M = n + m + . The
total number of choices is
N(N1)(N2). . .(NM+1) =
N (N 1) . . . 2 1
(N M) (N M 1) . . . 2 1
=
N!
(N M)!
But were not done yet, because the permutation (a
1
a
2
a
3
. . . a
n
) is really the same as the per-
mutation (a
2
a
3
a
4
. . . a
n1
a
n
a
1
). Thus, in the previous argument, we have counted the same
permutation many times. To be precise, we have n times overcounted each cycle (a
1
a
2
. . . a
n
),
because there are n distinct cyclically permuted ways of writing this cycle which really rep-
resent the same permutation. Likewise we have m times overcounted each cycle (b
1
b
2
. . . b
m
)
and we have times overcounted each cycle (c
1
c
2
. . . c

). In total we have overcounted by a

factor of nm. Thus, we must divide by nm to conclude: [/() [ =
N!
n m (N M)!
.
2
Example 85: If = (1 2 3 4 5) in S
8
, then M = m
1
= 5, so that /() has cardinality
8!
5 (8 5)!
=
8 7 6 5 4
5
= 1344.
Recall that the centralizer of an element g in group Gis the set c
G
(g) = c G ; cg = gc.
Lemma 86 If G is any group, and g G, then [/(g) [ =
[G[
[c
G
(g) [
.
Proof: Let G act on itself by conjugation. Then /(g) is just the orbit of g under this group
action, and c
G
(g) is just the stabilizer of g. The result now follows from the Orbit-Stabilizer
theorem. 2
Corollary 87 If S
N
has cycle type (m
1
, m
2
, . . . , m
K
) and M = m
1
+ . . . + m
K
, then
[c
S
N
()[ = m
1
m
2
. . . m
K
(N M)!.
Proof: Combine Corollary 84 and Lemma 86. 2
6.3. THE SIMPLICITY OF A
N
67
Example 88: If = (1 2 3 4 5) in S
8
then c
S
8
() has cardinality 5(85)! = 53! = 30.
Corollary 87 provides information only about the size of c
S
N
() but in some cases this is
enough to exactly determine c
S
N
()....
Corollary 89 Suppose = (a
1
a
2
. . . a
M
) is a single M-cycle, for some M N. Let
/ = a
1
, a
2
, . . . , a
n
, and let B = [1..N] /. Then
c
S
N
() =
m
; m [0..M), a permutation of B
Proof: Let o =
m
; m [0..M), a permutation of B.
Claim 1: o c
S
N
().
Proof: Clearly, any power of commutes with , and if is a permutation of B, then
and do not touch any of the same elements of [1..N], so also commutes with . Thus,
any element
m
in o commutes with . .............................. 2 [Claim 1]
Now, Corollary 87 says that [c
S
N
()[ = M (N M)!.
Claim 2: [o[ = M (N M)! also.
Proof: Observe that [B[ = NM, so the number of permutations of B is (NM)!. The
number of distinct powers of is M. Thus, the number of products
m
is M (N M)!.
2 [Claim 2]
Combining Claims 1 and 2, we conclude that o = c
S
N
(). 2
Example 90: Consider S
5
. If = (1 2 3) then
c
S
5
() =
n
; n = 0, 1, 2, permutes 4,5 = (1 2 3)
n
(4 5)
m
; n = 0, 1, 2, m = 0, 1
=
_
e, (4 5), (1 2 3), (1 2 3)(4 5), (1 3 2), (1 3 2)(4 5)
_
.
If = (12345) then c
S
5
() =
n
; n = 0, 1, 2, 3, 4 = e, (12345), (13524), (14253), (15432).
Proposition 91 A
5
is simple.
Proof: We will apply Corollary 79. We need to measure the conjugacy classes of A
5
.
Claim 1: Let , A
N
. If and are conjugate in A
N
, then they have the same cycle
type.
Proof: If and are conjugate in A
N
, then =
1
for some A
N
. Since A
N
S
N
,
this means that If and are also conjugate in S
N
. Now apply Proposition 83. 2
68 CHAPTER 6. GROUP ACTIONS
(Note that, unlike Proposition 83, the statement of Claim 1 is not if and only if.)
Claim 2: The distinct cycle types of A
5
are:
(a b c), (a b c d e), (a b)(c d)
and the identity element e.
Proof: Exercise. ..................................................... 2 [Claim 2]
Claim 3: There are exactly 20 elements in A
5
of cycle type (a b c), and all are conjugate.
Proof: As in Corollary 87, the number of elements of type (a b c) is
543
3
= 20. To see that
all are conjugate, consider the element = (123). We want to show that [/() [ = 20.
Lemma 86 says
[/() [ =
[A
5
[
[c
A
5
()
=
60
[c
A
5
() [
,
so it suces to show that [c
A
5
() [ = 3.
Now, c
A
5
() = A
5
c
S
5
(), and in Example 90 we found that
c
S
5
() =
_
e, (45), (123), (123)(45), (132), (132)(45)
_
.
Of the six elements in this list, only three are in A
5
, namely, e, (123), and (132). Thus,
c
A
5
() = e, (123), (132) has cardinality 3, so that [/() [ has cardinality 60/3 = 20,
as desired. .......................................................... 2 [Claim 3]
Claim 4: There are exactly 24 elements in A
5
of cycle type (a b c d e), and they fall into
two conjugacy classes, having 12 elements each.
Proof: As in Corollary 87, the number of elements of type (a b c d e) is
5432
5
= 24.
Consider the element
1
= (1 2 3 4 5); well show that [/(
1
) [ = 12. As in Claim 3, we
observe that
[/(
1
) [ =
[A
5
[
[c
A
5
(
1
)
=
60
[c
A
5
() [
,
so it suces to show that [c
A
5
(
1
) [ = 5. Again, c
A
5
(
1
) = A
5
c
S
5
(), and in Example
90 we found that c
S
5
(
1
) = e, (12345), (13524), (14253), (15432). All ve of these
elements are in A
5
, so were done.
Next, let
2
= (13524); by similar reasoning, [/(
2
) [ = 12. Thus, between them, /(
1
)
and /(
2
) cover all 24 elements of type (a b c d e).
It remains to show that
1
and
2
are not conjugate in A
5
. This is Exercise 39 (Hint:
Show that, if
2

1
=
1
, then =
n
1
, where = (2453). Note that is odd thus,
must be odd. Thus,
1
and
2
cannot be conjugated with an even permutation, so they are not
conjugate in A
5
.) ..................................................... 2 [Claim 4]
6.3. THE SIMPLICITY OF A
N
69
Claim 5: There are exactly 15 elements in A
5
of cycle type (a b)(c d), and all are
conjugate.
Proof: Exercise. Hint: let = (12)(34), and show that [c
A
5
() [ = 4. .. 2 [Claim 5]
Combining Claims 2, 3, 4, and 5, we see that A
5
ts the description of Example 80). Thus,
it must be simple. 2
Proposition 92 For any N 5, A
N
is simple.
Proof: (by induction on N)
Base Case: (N = 5) This is Proposition 91.
Induction: Let N 6, and suppose A
(N1)
is simple. Let HA
N
be some normal subgroup.
To show that A
N
is simple, we must show that either H = e or H = A
N
. Suppose, by
contradiction, that e , = H ,= A
N
.
Recall that A
N
S
N
, and that S
N
acts on [1..N] by permutations.
Claim 1: If H, and ,= e, then xes no element of [1..N]. That is, (n) ,= n for
any n [1..N].
Proof: Let G
n
= Stab (n) = A
N
; (n) = n.
Claim 1.1: G
n

= A
N1
.
Proof: Let F
n
= S
N
; (n) = n. Then F
n
is really the set of all permutations of
1, 2, . . . , n1, n X, n+1, . . . , N, a set having N1 elements, so F
n

= S
N1
. But G
n
=
A
N
F
n
is just the set of even elements in F
n
hence, G
n

= A
N1
. 2 [Claim 1.1]
Claim 1.2: If e ,= H, and (n) = n, then G
n
H.
Proof: If (n) = n, then H G
n
. Thus, H G
n
,= e. But H A
N
, so that
HG
n
G
n
. But by Claim 1, G
n
is isomorphic to A
N1
, and by induction hypothesis,
A
N1
is simple. Thus, G
n
has no nontrivial proper normal subgroups. Hence, we must
have H G
n
= G
n
, which means G
n
H. ........................ 2 [Claim 1.2]
Claim 1.3: If G
n
H for some n [1..N], then G
m
H for all m [1..N].
Proof: Let A
N
be such that (n) = m. Thus, G
n

1
=
_
Stab (n)
_

1
Lemma 69
Stab ((n)) = Stab (m) = G
m
. Since G
n
H, it follows that G
m
H
1
.
However, H
1
= H, because H is normal. ....................... 2 [Claim 1.3]
Thus, G
1
, G
2
, . . . , G
N
H, so that G
1
, G
2
, . . . , G
N
) H.
Claim 1.4: G
1
, G
2
, . . . , G
N
) = A
N
.
70 CHAPTER 6. GROUP ACTIONS
Proof: We want to show that any element of A
N
can be written as a product of elements
from G
1
, G
2
, . . . , G
N
. So, let A
N
. Since is even, we can write it as a product of
an even number of 2-cycles:
=
1

1
..

2

2
..
. . .
K

K
. .
=
1

2
. . .
K
Here,
k
=
k

k
is a product of two 2-cycles.
Claim 1.4.1: For all k,
k
G
n
k
for some n
k
Proof: Suppose
k
= (a
1
a
2
) and
k
= (b
1
b
2
), for some a
1
, a
2
, b
1
, b
2
[1..N]. Thus,

k
= (a
1
a
2
)(b
1
b
2
) permutes the elements a
1
, a
2
, b
1
, b
2
, but xes all other elements in
[1..N]. But N 5 by hypothesis, so
k
must x at least one n [1..N]. Thus,

k
G
n
. .................................................... 2 [Claim 1.4.1]
Thus, =
1

2
. . .
K
is a product of elements from G
1
, G
2
, . . . , G
N
. 2 [Claim 1.4]
So, if H, and (n) = n, then Claim 1.2 says G
n
H. Then Claim 1.3 says that
G
m
H for all m. But then Claim 1.4 says A
N
H in other words, H = A
N
.
2 [Claim 1]
Claim 2: Let
1
,
2
H. If there is any n [1..N] so that
1
(n) =
2
(n), then
1
=
2
.
Proof: If
1
and
2
are elements of H, then
1
1

2
H also. But if
1
(n) =
2
(n), then

1
1

2
(n) = n. By Claim 1, this means that
1
1

2
= e. Thus,
1
=
2
. ... 2 [Claim 2]
Claim 3: All elements of H are products of disjoint 2-cycles. In other words, if H,
then = (ab)(cd)(ef) . . ., where a, b, c, d, e, f, . . . are all distinct.
Proof: Suppose contained a cycle of length 3 or longer. Thus,
= (a
1
a
2
a
3
. . .)(b
1
b
2
. . .) . . . (c
1
c
2
. . .).
for some a
1
, a
2
, a
3
, . . . , b
1
, b
2
, . . . in [1..N]. Let x, y [1..N] be any elements, such that
x ,= a
3
, and let = (a
3
xy). Then A
N
.
Let
1
=
1
. Since H is normal, and H, we know that
1
H also. By Lemma 81,

1
=
_
(a
1
)(a
2
)(a
3
) . . .
__
(b
1
)(b
2
) . . .
_
. . .
_
(c
1
)(c
2
) . . .
_
=
_
a
1
a
2
(a
3
) . . .
__
(b
1
)(b
2
) . . .
_
. . .
_
(c
1
)(c
2
) . . .
_
.
Thus,
1
(a
1
) = a
2
= (a
1
). On the other hand,
1
(a
2
) = x ,= a
3
= (a
2
), so
1
and
2
cannot be equal. This contradicts Claim 2. ........................... 2 [Claim 3]
Let a, b, c, d, e, f [1..N] be distinct elements (this is possible because N 6 by hypothesis),
and let = (ab)(cd)(ef) . . . be some element of H (by Claim 3, any element of H must have
this form). Let = (ab)(de). Then A
N
, so
1
=
1
H also. But Lemma 81 says
6.3. THE SIMPLICITY OF A
N
71
that
1
= (ab)(ce)(df). Thus,
1
,= , but (a) = b =
1
(a). Again, we have a contradiction
of Claim 2.
By contradiction, we must conclude that H = A
N
. Hence, A
N
is simple. 2
72 CHAPTER 6. GROUP ACTIONS
Part II
Rings
73
Chapter 7
Introduction
7.1 Basic Denitions and Examples
Recall the basic properties of integer arithmetic. The set of integers Z comes with two opera-
tions, addition and multiplication, which have the following properties:
1. (Z, +) is an abelian group. That is:
(a) + is associative: For all n, m, Z, n + (m + ) = (n + m) + .
(b) + is commutative: For all n, m Z, n + m = m + n.
(c) There is an additive identity element (namely, 0), such that for any n Z, 0+n =
n = n + 0.
(d) Every element n Z has an additive inverse (namely n), so that n+(n) = 0 =
(n) + n.
2. (Z, ) is not a group. For example, the element 2 has no multiplicative inverse in Z (since
1
2
is not an integer).
However, is still associative: For all n, m, Z, n (m ) = (n m) .
3. Z satises the distributive law: For all n, m, Z, n (m + ) = (n m) + (n ).
A ring is an abstract algebraic object which mimics these properties. To be precise, a ring
is a set 1, equipped with two binary operations, + and , so that:
1. (1, +) is an abelian group. That is:
(a) + is associative: For all r, s, t 1, r + (s + t) = (r + s) + t.
(b) + is commutative: For all r, s 1, r + s = s + r.
(c) There is an additive identity element 0 1 so that for any r 1, 0 + r = r =
r + 0.
75
76 CHAPTER 7. INTRODUCTION
(d) Every element r 1 has an additive inverse, denoted r, so that r +(r) = 0 =
(r) + r
2. (1, ) is not a group. However, is still associative: For all r, s, t 1, r(st) = (rs)t.
3. 1 satises the distributive law: For all r, s, t 1, r (s +t) = (r s) +(r t), and also
(s + t) r = (s r) + (t r).
Notice:
1. The operation need not be commutative (although it is for integers). If is commu-
tative, then we call 1 a commutative ring, or a domain.
2. A multiplicative identity is an element e 1so that, for any r 1, er = r = re.
If such an element exists, it is unique, and is denoted by 1 (as in the integers).
The set 1 may not have a multiplicative identity (although it almost always does). A ring
without a multiplicative identity is sometimes called a rng. A ring having a multiplicative
identity is sometimes called a ring with identity.
Example 93:
(a) Let 2Z = . . . , 4, 2, 0, 2, 4, 6, 8, . . . be the set of all even numbers. Then 2Z is a
commutative ring under the normal arithmetic operations of addition and multiplication.
However, 2Z has no identity element.
(b) More generally, for any n N, let nZ = nz ; z Z be the set of all multiples of n.
Then nZ is a commutative ring under the normal arithmetic operations of addition and
multiplication. However, unless n = 1, the ring nZ has no identity element.
(c) Rational numbers: Let Q =
_
a
b
; a, b Z
_
be the set of rational numbers. We dene
addition and multiplication on Q in the obvious way:
a
1
b
1
+
a
2
b
2
=
a
1
b
2
+ a
2
b
1
b
1
b
2
, and
a
1
b
1

a
2
b
2
=
a
1
a
2
b
1
b
2
. (7.1)
Then Q is a commutative ring, with additive identity 0 and multiplicative identity 1.
(d) Real numbers: Let R be the set of real numbers. Then R is a commutative ring under
the normal arithmetic operations of addition and multiplication. It has a multiplicative
identity, namely 1.
(e) Complex numbers: Let C be the set of complex numbers. That is, C = x + yi ; x, y R,
where i is the square root of negative one, satisfying the equation: i
2
= 1. The arith-
metic operations on C are dened:
(x
1
+ y
1
i) + (x
2
+ y
2
i) = (x
1
+ x
2
) + (y
1
+ y
2
)i
(x
1
+ y
1
i) (x
2
+ y
2
i) = (x
1
y
1
x
2
y
2
) + (x
1
y
2
+ x
2
y
1
)i
_
(7.2)
Then C is a commutative ring, with multiplicative identity 1 = 1 + 0i.
7.1. BASIC DEFINITIONS AND EXAMPLES 77
(f) Gaussian Integers: Let Z[i] = z + yi ; z, y Z, where i is the square root of nega- Num.Thr.
tive one, satisfying the equation: i
2
= 1. Thus, Z[i] is a subset of C, and the arithmetic
operations on Z[i] are again dened by the equations (7.2). It is left as Exercise 40 to
show that Z[i] is closed under these operations and forms a commutative ring.
(g) Hamiltonians: The ring of Hamiltonions is dened as follows. First, we introduce
three formal elements, i, j, and k, all of which are square roots of 1. That is:
i
2
= j
2
= k
2
= 1.
Furthermore, i, j, and k, are related by the following expressions:
ij = k; jk = i; ki = j.
ji = k; kj = i; ik = j.
_
(7.3)
(Note that this multiplication is not commutative)
The Hamiltonions are the set H of all formal linear combinations of the form
w + xi + yj + zk, where w, x, y, z are any real numbers.
We dene addition in the obvious fashion:
(w
1
+x
1
i+y
1
j+z
1
k) + (w
2
+x
2
i+y
2
j+z
2
k) = (w
1
+w
2
) + (x
1
+x
2
)i + (y
1
+y
2
)j + (z
1
+z
2
)k.
We dene addition by applying the formulae (7.3), and legislating that multiplication is
distributive. For example:
(3 + 4i 5j) (2 + 6i) = 3(2 + 6i) 4i(2 + 6i) 5j(2 + 6i)
= (6 + 12i) (8i + 24i
2
) (10j + 30ji)
= 6 + 12i 8i 24 10j 30k
= (6 24) + (12 8)i 10j 30k.
H is a noncommutative ring, with additive identity 0 = 0+0i +0j +0k and multiplicative
identity element 1 = 1 +0i +0j +0k. Note that H and is an extension of C in the same
way that C is an extension of R.
(h) Let Z
/5
=

0,

1,

2,

3,

4 be the set of congruence classes of integers, mod 5, and let +

and be addition and multiplication, mod 5. For example:

2 +

4 =

1, and

2

4 =

3.
Then Z
/5
forms a commutative ring, with additive identity element

0, and multiplicative
identity

1.
78 CHAPTER 7. INTRODUCTION
(i) More generally, for any n N, let Z
/n
be the set of congruence classes of integers, mod n,
and let + and be addition and multiplication, mod n. Then Z
/n
forms a commutative
ring, with additive identity element

0, and multiplicative identity

1.
(j) Trivial Ring: The set 0 is a ring, where 0+0 = 0 = 0 0. It is called the trivial ring.
The axioms of ring arithmetic have some basic consequences:
Proposition 94 Let 1 be a ring.
(a) If 1 has additive identity 0, then for any r 1, r 0 = 0 = 0 r.
(b) Each element in 1 has a unique additive inverse.
(c) If s 1 has additive inverse s, then for any r 1, (s) r = (s r) = s (r).
In particular, for any r 1, (1) r = r = r (1).
(d) If 1 has a multiplicative identity, it is unique. In other words, if e
1
and e
2
are two
elements of 1 such that e
1
r = r and e
2
r = r for any r 1, then e
1
= e
2
.
The unique additive inverse is denoted 1, or 1
R
.
Note: If e 1, then the fact that e r = r for some r 1, is not sucient to conclude
that e is the multiplicative identity. See Remark B of Example 95a) below.
(e) If r 1 has a multiplicative inverse, then it is unique. In other words, if i
1
and i
2
are
two elements of 1 such that i
1
r = 1 and i
2
r = 1, then i
1
= i
2
.
Proof: Exercise 41 2
7.2 () Many Examples of Rings
Prerequisites: 7.1
[This section contains many examples. It is not necessary to read and understand all these examples right
now; just read a few in order to get your intuititions working. These examples will be referred to throughout
the text.]
Example 95: (Product Rings )
a) Let 1 = R
2
, and dene addition and multiplication componentwise. That is, for any
(x
1
, y
1
) R
2
and (x
2
, y
2
) R
2
,
(x
1
, y
1
) + (x
2
, y
2
) = (x
1
+ y
1
, x
2
+ y
2
)
(x
1
, y
1
) (x
2
, y
2
) = (x
1
y
1
, x
2
y
2
)
7.2. () MANY EXAMPLES OF RINGS 79
Then R
2
is a commutative ring, with additive identity 0 = (0, 0) and multiplicative
identity 1 = (1, 1).
Remark A: The abelian group (R
2
, +) is isomorphic to the group (C, +) (see Example
93e)). However, rings (R
2
, +, ) is not isomorphic to the ring (C, +, ). Thus, we have
two rings which have identical additive structures, but which are not the same as rings.
Remark B: Let e = (1, 0). Observe that, for any element x = (x, 0) in R
2
, we have
e x = x. However, e is not the multiplicative identity of R
2
. This is an example of the
warning in Proposition (94d).
b) More generally, if 1 and o are any two rings, we dene the product ring 1o to have
componentwise addition and multiplication. That is, for any r
1
, r
2
1 and s
1
, s
2
o,
(r
1
, s
1
) + (r
2
, s
2
) = (r
1
+ s
1
, r
2
+ s
2
)
(r
1
, s
1
) (r
2
, s
2
) = (r
1
s
1
, r
2
s
2
)
Then 1o is a ring. Furthermore,
If 1 has additive identity 0
R
and o has additive identity 0
S
, then 1o has additive
identity 0 = (0
R
, 0
S
).
If 1 has multiplicative identity 1
R
and o has multiplicative identity 1
S
, then 1o
has multiplicative identity 1 = (1
R
, 1
S
).
1o is commutative if and only if 1 and o are both commutative.
Exercise 42 Verify these claims.
c) Let 1
1
, 1
2
, . . . , 1
n
be rings; by iterating this construction, we can dene the product
ring 1
1
1
2
. . . 1
n
.
d) In particular, if 1 is any ring, then we dene 1
n
= 11. . . 1
. .
n
.
Example 96: (Matrix Rings ) Grp.Repr.
a) Let /
2
(R) be the set of all 22 matrices with real coecients. We dene matrix addition
and multiplication in the familiar manner:
_
a
1
b
1
c
1
d
1
_
+
_
a
2
b
2
c
2
d
2
_
=
_
(a
1
+ a
2
) (b
1
+ b
2
)
(c
1
+ c
2
) (d
1
+ d
2
)
_
;
_
a
1
b
1
c
1
d
1
_

_
a
2
b
2
c
2
d
2
_
=
_
(a
1
a
2
+ b
1
c
2
) (a
1
b
2
+ b
1
d
2
)
(c
1
a
2
+ d
1
c
2
) (c
1
b
2
+ d
1
d
2
)
_
.
Then /
2
(R) is a noncommutative ring, with additive identity element 0 =
_
0 0
0 0

, and
multiplicative identity element Id =
_
1 0
0 1

(Exercise 43).
80 CHAPTER 7. INTRODUCTION
Remark: The abelian groups (R
4
, +), (H, +) and (/
2
(R), +) are all isomorphic as
groups (see Examples (93g) and (95)). However, the rings (R
4
, +, ), (H, +, ) and
(/
2
(R), +, ) are all dierent.
b) More generally, let /
n
(R) be the set of all n n matrices with real coecients. We
dene matrix addition and multiplication in the familiar manner. Then /
n
(R) is a
noncommutative ring (Exercise 44), with
Additive identity 0 =
_

_
0 . . . 0
.
.
.
.
.
.
.
.
.
0 . . . 0
_

_
and multiplicative identity Id =
_

_
1 0 . . . 0
0 1 . . . 0
.
.
.
.
.
.
.
.
.
.
.
.
0 0 . . . 1
_

_
.
c) Let /
n
(Z) be the set of all n n matrices with integer coecients. For example,
_
1 9
3 7

is an element of /
2
(Z), but
_
1
2
9
5
3
7
2
_
is not. Then /
n
(Z) is closed under matrix multipli-
cation, addition, and negation (Exercise 45), so it is also a (noncommutative) ring.
d) Let /
n
(Q) be the set of all n n matrices with rational coecients. For example,
_
1
2
9
5
3
7
2
_
is an element of /
2
(Q), but
_

3
3 7
_
is not. Then /
n
(Q) is closed under matrix
multiplication, addition, and negation (Exercise 46), so it is also a (noncommutative)
ring.
e) Let /
n
(C) be the set of all n n matrices with complex coecients, and let /
n
(H) be
the set of all nn matrices with Hamiltonian coecients. Then /
n
(C) and /
n
(H) are
(noncommutative) rings.
f) In general, if 1 is any ring, let /
n
(1) be the set of all n n matrices with coecients
in 1. We dene the multiplication and addition of these matrices in a manner exactly
analogous to that for real-valued matrices. Then /
n
(1) is a ring. Observe:
1. Set n = 1; then /
1
(1) is identical to 1.
2. If n 2, then /
n
(1) is never commutative (even if 1 is commutative).
3. If 1 has multiplicative identity 1, then /
n
(1) has multiplicative identity Id =

1 0 . . . 0
0 1 . . . 0
.
.
.
.
.
.
.
.
.
.
.
.
0 0 . . . 1

.
Exercise 47 Verify these claims.
Example 97: (Continuous Function Rings ) Alg.Geo.
7.2. () MANY EXAMPLES OF RINGS 81
a) Let c(R) be the set of all continuous real-valued functions f : RR. We dene addition
and multiplication of functions pointwise. That is, for any f, g : RR and any r R,
(f + g)(r) = f(r) + g(r) and (f g)(r) = f(r) g(r).
It is proved in rst-year calculus courses that the sum of two continous functions is
continuous, and the product of two continuous functions is continuous. Thus, c(R) is a
commutative ring. Observe:
1. The additive identity element is the constant zero function 0, dened by the property:
0(r) = 0 for any r R.
2. The multiplicative identity element is the constant one function 11, dened by the
property: 11(r) = 1 for any r R.
3. The additive inverse of f : RR is the function f (also continuous).
(Exercise 48)[Verify these claims.]
The algebraic structure of c(R) somehow encodes the topological structure of R. For
example, there exists no element f c(R) such that f is zero everywhere in (, 1),
but f(1) ,= 0. This reects the fact that the point 1 lies on the boundary of the domain
(, 1). This relationship between topological and algebraic structure is the starting point
of algebraic geometry.
b) Let a < b , and let c(a, b) be the set of all continuous real-valued functions
f : (a, b)R, with pointwise addition and multiplication. Then c(a, b) is a commutative
ring.
c) If U R
n
is any open subset, then let c(U) be the set of continuous real-valued functions
f : UR, with pointwise addition and multiplication. Then c(U) is a commutative
ring.
d) If X is any topological space, then let c(X) is the set of continuous real-valued functions
f : XR, with pointwise addition and multiplication. Then c(X) is a commutative
ring.
e) Similarly, if X is any topological space, let c(X; C) be the set of all continuous, complex-
valued functions f : XC, with pointwise addition and multiplication. Then c(X; C)
is a commutative ring.
f) Let a < b , and let c
1
(a, b) be the set of all continuously dierentiable real-
valued functions f : (a, b)R, with pointwise addition and multiplication. It is proved
in rst-year calculus that the sum of two dierentiable functions is dierentiable, and the
product of two dierentiable functions is dierentiable. Thus, c
1
(a, b) is a commutative
ring, with additive identity 0 and multiplicative identity 11.
82 CHAPTER 7. INTRODUCTION
supp(f)
f
supp(g)
g
supp(f+g)
f+g
supp(f)
f
supp(g)
g
supp(f.g)
supp(f)
f
f.g
(A) (B) (C)
Figure 7.1: Functions with compact support.
g) Let a < b . For any k N, let c
k
(a, b) be the set of all k times continuously
dierentiable real-valued functions f : (a, b)R, with pointwise addition and multipli-
cation. Then c
k
(a, b) is a commutative ring, with additive identity 0 and multiplicative
identity 11 (Exercise 49).
h) Let a < b , and let c

(a, b) be the set of all innitely dierentiable real-valued

functions f : (a, b)R, with pointwise addition and multiplication. Then c

(a, b) is a
commutative ring, with additive identity 0 and multiplicative identity 11 (Exercise 50).
i) If U R
n
is an open subset, then
c(U) is the set of continuous real-valued functions f : UR.
c
1
(U) is the set of continuously dierentiable real-valued functions f : UR.
c
k
(U) is the set of k times continuously dierentiable real-valued functions f :
UR.
c

(U) is the set of innitely dierentiable real-valued functions f : UR.

c

(U) is the set of analytic real-valued functions f : UR.

Then all of these are commutative rings.
j) If / is any dierentiable manifold (eg. a sphere, torus, etc), then let c

(U) be the set

of innitely dierentiable real-valued functions f : UR, with pointwise addition and
multiplication. Then c

(U) is a commutative ring.

k) Functions of Compact Support: If f : RR, then the support of f is the set of
all points where f is nonzero:
supp [f] = r R ; f(r) ,= 0 (see Figure 7.1A)
We say that f has compact support if supp [f] is a bounded subset of R. Let c
0
(R) be
the set of all continuous functions f : RR with compact support.
If f, g : RR, it is not hard to show (Exercise 51)that:
7.2. () MANY EXAMPLES OF RINGS 83
supp [f + g] supp [f] supp [g] (see Figure 7.1B).
supp [f g] = supp [f] supp [g] (see Figure 7.1C).
Thus, if f and g both have compact support, then so do the functions f +g and f g. Thus,
c
0
(R) is a commutative ring. Note, however, that c
0
(R) has no multiplicative identity,
because the function 11 does not have compact support: supp [11] = R.
Example 98: (Polynomial Rings ) Alg.Geo. ,
Galois
a) Polynomials of one variable: Let R[x] denote the set of all polynomial functions on
R that is, functions of the form
P(x) = p
n
x
n
+ p
n1
x
n1
+ . . . + p
2
x
2
+ p
1
x + p
0
,
where p
n
, p
n1
, . . . , p
1
, p
0
are real numbers.
The sum or product of two polynomials is also a polynomial, so R[x] is a ring. Polynomial
arithmetic can be described by the following formulae:
_
p
N
x
N
+ . . . + p
2
x
2
+ p
1
x + p
0
_
+
_
q
N
x
N
+ . . . + q
2
x
2
+ q
1
x + q
0
_
= (p
N
+ q
N
)x
N
+ . . . + (p
2
+ q
2
)x
2
+ (p
1
+ q
1
)x + (p
0
+ q
0
). (7.4)
_
p
N
x
N
+ . . . + p
2
x
2
+ p
1
x + p
0
_

_
q
M
x
M
+ . . . + q
2
x
2
+ q
1
x + q
0
_
= p
n
q
M
x
N+M
+ (p
N
q
M1
+ p
N1
q
M
)x
N+M1
+ . . .
. . . + (p
2
q
0
+ p
1
q
1
+ p
0
q
2
)x
2
+ (p
1
q
0
+ p
0
q
1
)x + p
0
q
0
. (7.5)
We can write formulae (7.4) and (7.5) more abstractly:
_
N

n=1
p
n
x
n
_
+
_
N

n=1
q
n
x
n
_
=
N

n=1
(p
n
+ q
n
)x
n
(7.6)
_
N

n=1
p
n
x
n
_

_
M

m=1
q
m
x
m
_
=
N+M

k=1
_

n+m=k
p
n
q
m
_
x
k
(7.7)
The degree of a polynomial is the highest exponent. For example, if p(x) = 4x
3
7x
2
+2,
then degree (p) = 3. If p(x) and q(x) are polynomials, then it is easy to verify:
degree (p q) = degree (p) + degree (q) (Exercise 52) (7.8)
b) Complex Polynomials: Let C[x] be the set of all functions f : CC which are
polynomials with complex coecients. Then C[x] is a commutative ring with identity.
c) Let Z[x] be the set of all functions f : ZZ which are polynomials with integer coe-
cients. Then Z[x] is a commutative ring with identity.
84 CHAPTER 7. INTRODUCTION
d) Let Q[x] be the set of all functions f : QQ which are polynomials with rational
coecients. Then Q[x] is a commutative ring with identity.
e) Polynomials of Many Variables: Let R[x, y] be the set of all functions f : RRR
which are polynomials in both variables. We dene addition and multiplication in the
obvious way. For example,
(3x + 4y) (x
2
2y) = 3x
3
6xy + 4x
2
y 2y
2
.
More generally, for any n N, let R[x
1
, x
2
, . . . , x
n
] be the set of all functions f : R
n
R
which are polynomials in n variables. Then R[x
1
, x
2
, . . . , x
n
] is a commutative ring with
identity.
f) Abstract Polynomial Rings: If 1 is any ring, let 1[x] be the set of all formal poly-
nomials
P(x) = p
n
x
n
+ p
n1
x
n1
+ . . . + p
2
x
2
+ p
1
x + p
0
,
where p
n
, p
n1
, . . . , p
1
, p
0
are elements of 1. Such an expression denes a function 11.
However, we do not regard P(x) as a function, but instead as an abstract algebraic
expression. We refer to the symbol x as an indeterminant, which means it is just an
abstract algebraic placeholder.
Addition and multiplication are dened using formulae formulae (7.4) and (7.5). Then:
1[x] is a ring.
1[x] is commutative if and only if 1 is commutative.
1[x] has a multiplicative identity if and only if 1 has one
Exercise 53 Verify these claims.
g) Let o = 1[x] is a ring, and consider the polynomial ring o[y] (where y is another inde-
terminant). Elements of this ring are formal polynomials in the variables x and y, with
coecients in 1. Thus, we denote this ring by 1[x, y].
More generally, for any n N, let 1[x
1
, x
2
, . . . , x
n
] be the ring of all formal polynomials
in n variables, and coecients in 1.
h) Rational functions: Let R(x) denote the set of rational functions that is, all functions
of the form f(x) =
p(x)
q(x)
, where p, q R[x] are polynomials. We dene addition and
multiplication on R(x) in the obvious way:
p
1
(x)
q
1
(x)
+
p
2
(x)
q
2
(x)
=
p
1
(x)q
2
(x) + p
2
(x)q
1
(x)
q
1
(x)q
2
(x)
, and
p
1
(x)
q
1
(x)

p
2
(x)
q
2
(x)
=
p
1
(x) p
2
(x)
q
1
(x) q
2
(x)
. (7.9)
Then R(x) is a commutative ring.
Observe that R(x) is related to R[x] in the same way that Q is related to Z; indeed,
formula (7.9) is exactly analogous to formula (7.1) on page 76.
7.2. () MANY EXAMPLES OF RINGS 85
Example 99: (Analytic Functions and Power Series ) Alg.Geo.
a) Real Analytic Functions of one variable: Let 0 < r , and let c

(r, r) be
the set of all analytic real-valued functions F : (r, r)R. A function F is analytic if
F is innitely dierentiable, and F has a Taylor series
F(x) =

n=0
f
n
x
n
(where f
0
, f
1
, f
2
, . . . are real numbers),
which converges everywhere on the interval (r, r). For example, the exponential function
exp(x) is an element of c

(, ) = c

(R), because, for any x R,

exp(x) =

n=0
1
n!
x
n
and this series converges uniformly on (, ).
Elementary calculus can be used to prove that the sum or product of two analytic functions
is also analytic. To be precise, if
F(x) =

n=0
f
n
x
n
, and G(x) =

n=0
g
n
x
n
,
then
F(x)+G(x) =

n=0
(f
n
+g
n
)x
n
, and F(x)G(x) =

n=0
_
n

k=0
f
k
g
nk
_
x
n
. (7.10)
c

(r, r) is a commutative ring, with additive identity 0 and multiplicative identity 11.
b) Holomorphic functions: If U C is an open subset, then let c

(U; C) be the set

of holomorphic functions F : UC. A function F is holomorphic if it is complex-
dierentiable; it then has a Taylor series
F(x) =

n=0
f
n
x
n
(where f
0
, f
1
, f
2
, . . . are complex numbers),
Then c

(U; C) is a commutative ring.

c) Formal Power Series: We dene R[[x]] to be the ring of all formal power series. In
other words,
R[[x]] =
_

n=0
f
n
x
n
; where f
0
, f
1
, f
2
, . . . are any real numbers
_
86 CHAPTER 7. INTRODUCTION
Observe that we place no constraints on the coecients f
0
, f
1
, f
2
, . . .. Thus, in general
the power series

n=0
f
n
may not converge anywhere. For example, the series
1 x + x
2
x
3
x
4
+ x
5
. . .
is an element of R[[x]], despite the fact that this series has zero radius of convergence.
We dene addition and multiplication in R[[x]] using formulae (7.10). Then R[[x]] is a
commutative ring, with additive identity 0 = 0+0x+0x
2
+. . . and multiplicative identity
1 = 1 + 0x + 0x
2
+ . . ..
d) Abstract Power Series: If 1 is any ring, then let 1[[x]] be the set of formal power
series with coecients in 1. In other words,
1[[x]] =
_

n=0
f
n
x
n
; where f
0
, f
1
, f
2
, . . . are any elements of 1
_
Note: in general, there is no topological structure on 1. In other words, there is no notion
of convergence on 1, so it is meaningless to talk about the convergence of this power
series. It is a purely formal construction.
We dene addition and multiplication in 1[[x]] using formulae (7.10). Then:
1[[x]] is a ring.
1[[x]] is commutative if and only if 1 is commutative.
1[[x]] has a multiplicative identity if and only if 1 has one.
Exercise 54 Verify these claims.
e) Meromorphic functions: If U C is an open subset, then let c

(U;

C) be the set of
meromorphic functions F : U

C.
Here,

C = C. is the Riemann Sphere, consisting of the complex plane and a point at
innity. A function F is meromorphic if F is complex-dierentiable everywhere except at
some discrete collection of poles, where F is innite. For example, the function F(x) =
1
xi
is meromorphic (and F(i) = ).
If U, then F has a Laurent series
F(x) =

n=N
f
n
(x )
n
_
Here, N N, and f
N
, f
1N
, . . . , f
1
, f
0
, f
1
, f
2
, . . . are
any complex numbers.
_
which converges in a neighbourhood of . If is a pole, then N > 0. If is not a pole
(ie. F() is nite) then N = 0.
7.2. () MANY EXAMPLES OF RINGS 87
c

(U;

C) is a ring. Furthermore, if U, and
F(x) =

n=N
f
n
(x )
n
, and G(x) =

n=M
g
m
(x )
m
,
then
F(x) + G(x) =

n=N
(f
n
+ g
n
)(x )
n
.
_
Assume WOLOG that M N
and dene gn = 0 if n < M
_
(7.11)
and F(x) G(x) =

n=(N+M)
_
n

k=N
f
k
g
nk
_
(x )
n
. (7.12)
(Exercise 55).
f) Formal Laurent Series: We dene C((x)) to be the ring of all formal Laurent series.
In other words,
C((x)) =
_

n=N
f
n
x
n
; where N N, and f
K
, f
1K
, . . . are complex numbers
_
Observe that we place no constraints on the coecients f
0
, f
1
, f
2
, . . .. Thus, in general,
the Laurent series may not converge anywhere.
We dene addition and multiplication in C((x)) using formulae (7.11) and (7.12). Then
C((x)) is a ring, with additive identity 0 = 0 + 0x + 0x
2
+. . . and multiplicative identity
1 = 1 + 0x + 0x
2
+ . . .. (Exercise 56)
g) Abstract Laurent Series: If 1is any ring, then let 1((x)) be the set of formal Laurent
series with coecients in 1. In other words,
1((x)) =
_

n=N
f
n
x
n
; where N N, and f
0
, f
1
, f
2
, . . . are any elements of 1
_
Note: in general, there is no topological structure on 1. In other words, there is no notion
of convergence on 1, so it is meaningless to talk about the convergence of the Laurent
series. It is a purely formal construction.
We dene addition and multiplication in 1((x)) using formulae (7.11) and (7.12). Then:
1((x)) is a ring.
1((x)) is commutative if and only if 1 is commutative.
1((x)) has a multiplicative identity if and only if 1 has one.
88 CHAPTER 7. INTRODUCTION
Exercise 57 Verify these claims.
Example 100: (Abstract Function Rings)
a) Consider a set s
1
, s
2
with two elements. Let R
{s
1
,s
2
}
be the set of all functions from
s
1
, s
2
into R. Thus, an element f R
{s
1
,s
2
}
is a function with two values: f(s
1
) and
f(s
2
). We dene addition and multiplication pointwise: (f + g)(s
1
) = f(s
1
) + g(s
1
), etc.
Then R
{s
1
,s
2
}
is a commutative ring (Exercise 58).
Observe that any element f of R
{s
1
,s
2
}
denes an element (x
1
, x
2
) of R
2
(Example 95a)),
where
x
1
= f(s
1
) and x
2
= f(s
2
). (7.13)
Conversely, any element (x
1
, x
2
) of R
2
denes an element f of R
{s
1
,s
2
}
via equation (7.13).
Thus, there is a natural bijection between R
{s
1
,s
2
}
and R
2
.
b) More generally, let ^ = s
1
, s
2
, . . . , s
N
be a set with N elements, and let R
N
be the set
of all functions from ^ into R. We dene addition and multiplication pointwise. Then
R
N
is a commutative ring (Exercise 59), and again, there is a natural correspondence
between R
N
and R
N
(Example 95d)).
c) Let R
R
be the set of all functions from R to R. Thus, R
R
includes all elements of c(R)
(Example 97a)), but also includes all discontinuous functions. Then R
R
is a commutative
ring under componentwise addition and multiplication.
d) Let X be any set, and let 1 be any ring. Let 1
X
be the set of all functions from X into
1. Then:
1
X
is a ring under componentwise addition and multiplication.
1
X
is commutative if and only if 1 is commutative.
1
X
is has a multiplicative identity if and only if 1 has one.
Exercise 60 Verify these claims.
Example 101: (Endomorphism Rings )
a) Consider the abelian group Z
2
, and let End [Z
2
] be the set of all endomorphisms of Z
2
;
that is, functions : Z
2
Z
2
such that (y +z) = (y) + (z) for any y, z Z
2
.
We dene addition on End [Z
2
] pointwise: if , End [Z
2
], and z Z
2
, then ( +
)(z) = (z) + (z).
However, we do not dene multiplication componentwise; instead, we dene it via function
composition: if , End [Z
2
], and z Z
2
, then ( )(z) =
_
(z)
_
.
It is left as Exercise 61 to verify:
7.2. () MANY EXAMPLES OF RINGS 89
If and are endomorphisms of Z
2
, then + and are also endomorphisms
of Z
2
.
The set End [Z
2
] forms a noncommutative ring under these operations.
The additive identity is the endomorphism O dened: O(z) = 0 for all z Z
2
.
The multiplicative identity is the identity endomorphism Id dened: Id(z) = z for
all z Z
2
.
Endomorphism can be understood via matrices. If End [Z
2
], then there is a 2 2
integer matrix F =
_
f
1
f
2
f
3
f
4
_
so that, for any element z = (z
1
, z
2
) in Z
2
,
(z) =
_
f
1
f
2
f
3
f
4
__
z
1
z
2
_
=
_
f
1
z
1
+ f
2
z
2
f
3
z
1
+ f
4
z
2
_
If is another endomorphism, with matrix G, then the endomorphism (+) has matrix
F+G, and the endomorphism () has matrix F G (Exercise 62). Thus, the algebra
of endomorphisms closely resembles the algebra of matrices.
b) More generally, let (/, +) be any additive abelian group, and let End [/] be the set of all
endomorphisms of /; that is, functions : // such that (a + b) = (a) + (b) for
any a, b /.
Again, we dene addition on End [/] pointwise, and we dene multiplication via function
composition. In other words, if , End [/], and a /, then
( + )(a) = (a) + (a) and ( )(a) =
_
(a)
_
.
It is left as Exercise 63 to verify:
If and are endomorphisms of /, then + and are also endomorphisms
of /.
The set End [/] forms a noncommutative ring under these operations.
The additive identity is the endomorphism O dened: O(a) = 0 for all a /.
The multiplicative identity is the identity endomorphism Id dened: Id(a) = a for
all a /.
c) A subring of End [/] is called an endomorphism ring.
d) If ( is a nonabelian group, then End [(] does not form a ring (Exercise 64).
Example 102: Rings of Subsets
Let X be some set, and let T(X) be the set of all subsets of X:
T(X) = P ; P X.
90 CHAPTER 7. INTRODUCTION
If P
1
, P
2
T(X), then we dene their sum as follows:
P
1
+P
2
= x X ; either x P
1
or x P
2
, but not in both.
(Often, this is called the symmetric dierence of P
1
and P
2
, and written as P
1
P
2
) We
dene the product of P
1
and P
2
to be their intersection:
P
1
P
2
= P
1
P
2
.
Then:
The set T(X) is a commutative ring under these operations,
The additive identity is the empty set .
The multiplicative identity is the set X.
T(X) is called the Ring of Subsets of X.
Exercise 65 Show that P
1
P
2
= P
1
+P
2
+ (P
1
P
2
).
Example 103: (Group Rings ) Grp.Repr.
a) Let R[x, x
1
] to be the set of all polynomials in x and x
1
with real coecients. An
example element of R[x, x
1
] is
3x
2
6x + +
5
2
x
1
7x
2
(7.14)
We dene multiplication and addition of these polynomials in the obvious way, similar to
in Example 98a).
b) Another way to think about R[x, x
1
] is as follows. Let A = . . . , x
2
, x
1
, x
0
, x
1
, x
2
, . . .
be the multiplicative group generated by a single element x. (Thus, A is isomorphic to
Z). We dene RA to be the set of all formal linear combinations of elements in A. In
other words, an element of RA has the form
r
1
x
n
1
+ r
2
x
n
2
+ . . . + r
k
x
n
k
where r
1
, . . . , r
k
R, and n
1
, . . . , n
k
Z. For example the polynomial (7.14) could be an
element of RA.
We dene addition and multiplication of elements in RA exactly as for polynomials. That
is:
_
r
1
x
n
1
+ r
2
x
n
2
+ . . . + r
k
x
n
k
_
+
_
s
1
x
n
1
+ s
2
x
n
2
+ . . . + s
k
x
n
k
_
= (r
1
+ s
1
)x
n
1
+ (r
2
+ s
2
)x
n
2
+ . . . + (r
k
+ s
k
)x
n
k
,
7.3. () APPLICATIONS OF RING THEORY 91
and
_
r
1
x
n
1
+ r
2
x
n
2
+ . . . + r
k
x
n
k
_

_
s
1
x
m
1
+ s
2
x
m
2
+ . . . + s
j
x
m
j
_
= r
1
s
1
x
n
1
+m
1
+ r
2
s
1
x
n
2
+m
1
+ . . . + r
k
s
1
x
n
k
+m
1
+ r
1
s
2
x
n
1
+m
2
+ r
2
s
2
x
n
2
+m
2
+ . . . + r
k
s
2
x
n
k
+m
2
+ . . .
. . . + r
1
s
j
x
n
1
+m
j
+ r
2
s
j
x
n
2
+m
j
+ . . . + r
k
s
j
x
n
k
+m
j
In other words, the algebra of RA is identical to that of R[x, x
1
].
c) Let ((, ) be any (multiplicative) group, and let 1 be any commutative ring. We dene
1A to be the set of all formal linear combinations of elements in A with coecients in
1. In other words, an element of RA has the form
r
1
g
1
+ r
2
g
2
+ . . . + r
k
g
k
where r
1
, . . . , r
k
1 and g
1
, . . . , g
k
(.
We dene addition and multiplication of elements in RA exactly as for polynomials. That
is,
_
r
1
g
1
+ r
2
g
2
+ . . . + r
k
g
k
_
+
_
s
1
g
1
+ s
2
g
2
+ . . . + s
k
g
k
_
= (r
1
+ s
1
)g
1
+ (r
2
+ s
2
)g
2
+ . . . + (r
k
+ s
k
)g
k
,
and
_
r
1
g
1
+ r
2
g
2
+ . . . + r
k
g
k
_

_
s
1
h
1
+ s
2
h
2
+ . . . + s
j
h
j
_
= r
1
s
1
g
1
h
1
+ r
2
s
1
g
2
h
1
+ . . . + r
k
s
1
g
k
h
1
+ r
1
s
2
g
1
h
2
+ r
2
s
2
g
2
h
2
+ . . . + r
k
s
2
g
k
h
2
+ . . .
. . . + r
1
s
j
g
1
h
j
+ r
2
s
j
g
2
h
j
+ . . . + r
k
s
j
g
k
h
j
Observe that this is the natural generalization of Example 103b). Then:
1( is a ring.
1( is commutative if and only if ( is abelian.
1( has multiplicative identity 0 = 0g for any g (.
If ( has identity e
G
, then 1( has multiplicative identity 1 e
G
.
(Exercise 66 Verify these claims). 1( is called the Group ring of ( with coecients in
1. It is very important in the representation theory of groups (see 7.3.5 on page 96)
7.3 () Applications of Ring Theory
Prerequisites: 7.1
Here we briey survey the major mathematical applications of ring theory.
92 CHAPTER 7. INTRODUCTION
7.3.1 Algebraic Number Theory
Number theory studies the mathematical properties of the ring of integers. Abstract ring theory
sheds light on classical number theory in two ways:
1. We can contextualize the problems and results of number theory within a larger frame-
work. Given any result in number theory, we can ask, Is this theorem true only for
integers, or does it generalize to some other rings as well?
2. Problems about integers can often be solved by relating the integers to some other ring,
such as the Gaussian Integers Z[i] of Example 93f), or the ring Z
/n
of Example 93i).
For example, a Diophantine Equation is a polynomial equation of the form
p(x
1
, x
2
, . . . , x
n
) = 0,
where p Z[x
1
, x
2
, . . . , x
n
] is a polynomial in n variables with integer coecients, and we
impose the constraint that x
1
, . . . , x
n
must be integers. For example, the equation
x
2
y = 0, (7.15)
says that y = x
2
, ie. that y is a perfect square. Equation (7.15) has solutions (x, y) = (1, 1),
(2, 4), (3, 9), etc. The Pythagorean equation
x
2
+ y
2
z
2
= 0, (7.16)
says that x
2
+ y
2
= z
2
in other words, (x, y, z) form a Pythagorean triple. One solution is
(x, y, z) = (3, 4, 5).
The obvious generalization of the Pythagorean Equation is the Fermat equation
x
n
+ y
n
z
n
= 0, (7.17)
(where n N is xed), which says that x
n
+y
n
= z
n
. Fermats Last Theorem says there are no
nontrivial solutions for any n 3. This theorem was nally proved by Andrew Wiles in 1998,
after almost 4 centuries of mathematical eort.
It is (relatively) easy to nd real or complex solutions to a polynomial equations like (7.15),
(7.16) or (7.17), but it is often very dicult to isolate the integer solutions or indeed, even
to determine whether any exist. The study of Diophantine equations is thus a major area of
number theory, and ring theory provides important tools. The simplest tool is just reduction
mod p. For example, the equation (7.17) is true only if the congruence equation
x
n
+ y
n
z
n
0 (mod p) (7.18)
is true, for any xed p N. Equation (7.18) is an equation in the nite ring Z
/p
(Example
93i)), and is probably easier to solve than (7.17).
We will indicate when a particular topic is relevant to algebraic number theory by placing
the ag Num.Thr. in the margin.
7.3. () APPLICATIONS OF RING THEORY 93
7.3.2 Field Theory & Galois Theory
Consider the following three problems:
Solution formulae for polynomial equations: If a, b, c R are constants, then the quadratic
equation
ax
2
+ bx + c = 0 (7.19)
can always be solved by recourse to the quadratic formula:
x =
b

b
2
4ac
2a
. (7.20)
Geronimo Cardano (1501-1576) developed a similar (but much more complicated) formula
for solving a cubic equation:
ax
3
+ bx
2
+ cx + d = 0.
Later, an even more complicated formula was developed solve the quartic equation:
ax
4
+ bx
3
+ cx
2
+ dx + e = 0.
This raised the question: does there exist a formula to solve an arbitrary quintic equation
ax
5
+ bx
4
+ cx
3
+ dx
2
+ ex + f = 0? (7.21)
What about higher-order polynomials?
Note that the question here is not whether a solution exists, but whether that solution can
always be found using an algebraic formula like (7.20). For example, the transcendental
equation
e
x
= 17 (7.22)
has a solution namely, x = log(17). However, there is no nite algebraic expression for
x in terms of 17 and radicals of 17, such as

17,
3

17 etc. So were asking,

Is equation (7.21) solvable via a nite expression involving radicals of a, b, c, etc? Or is
it like (7.22), possessing a solution, but not one we can easily describe?
The nature of the complex numbers: If 4ac > b
2
, then the quadratic equation (7.19) has
no real solution. In this case, the quadratic formula (7.20) produces a nonsensical result,
involving the square root of a negative number. The complex numbers were introduced to
provide an interpretation for these nonsense solutions; they extend the real numbers to
a larger eld, where the quadratic formula (7.20) always produces a meaningful answer.
Thus, in the complex numbers, every quadratic polynomial has a solution.
Does every polynomial equation have a solution in the complex numbers? Or must we
pass to an even larger eld in order to solve cubics, quartics, quintics, etc.? C.F. Gauss
answered this:
94 CHAPTER 7. INTRODUCTION
?
?
(A)
(B) (C) (D) (E)
(F) (G)
Figure 7.2: Classical compass and straightedge constructions.
Fundamental Theorem of Algebra: Let P(x) = p
n
x
n
+p
n1
x
n1
+. . .+p
2
x
2
+p
1
x+p
0
be any polynomial with complex coecients p
n
, . . . , p
1
, p
0
. Then:
1. There is a root z C so that P(z) = 0.
2. P can be completely factored over the complex numbers. In other words, there exist
z
1
, z
2
, . . . , z
n
C so that P(x) = (x z
1
) (x z
2
) (x z
n
).
This raises the question: Given an arbitrary eld F and some polynomial equation p(x) =
0 over F, can we solve this equation by extending F to a larger eld E, in the same way
that we extended R to C? Is there something analogous to the Fundamental Theorem of
Algebra for these extensions?
Classical Geometric Constructions: Since the time of Euclid, certain geometric gures
have been constructed using a compass and straightedge. A compass is a string with a pin
at one end and a pencil at the other, used to inscribe circles of any radius. A straightedge
is just a peice of wood with a perfectly straight edge, used to inscribe lines. Using these
two simple tools, one can bisect angles (Figure 7.2A) , bisect lines and construct their
perpendiculars (Figure 7.2B) , and inscribe an equilateral triangle (Figure 7.2C), regular
hexagon (Figure 7.2D), or regular pentagon (Figure 7.2E), inside a given circle.
However, many things cannot be constructed. For example, after two thousand years of
eort, there is no known method to trisect an angle (Figure 7.2F), or to construct a regular
nonagon (a nine-sided polygon; see Figure 7.2G). Is this merely a failure of ingenuity? Or
is there a fundamental obstruction which makes these things impossible?
In 1832, while in prison, a twenty-one year old mathematician named Evariste Galois found
the key to solving these three problems, by studying the symmetry groups of the roots of poly-
nomials. Shortly afterwards, Galois was killed in a duel, but his ideas became the foundation
of the vast and beautiful theory which bears his name.
7.3. () APPLICATIONS OF RING THEORY 95
3
5
Ellipse Folium of Descartes Lemniscate Conchoid of Nicomedes
(x/5)
2
+ (y/3)
2
= 1 x
3
+ y
3
- 6xy = 0 2(x
2
+ y
2
)
2
- 25 (x
2
- y
2
) = 0 x
2
y
2
- (y+1)
2
(4-y
2
) = 0
Figure 7.3: Algebraic varieties.
We will indicate when a particular topic is relevant to galois theory by placing the ag Galois
in the margin.
7.3.3 Algebraic Geometry & Commutative Algebra
An algebraic variety is the geometric gure determined by the solutions to a polynomial
equation. For example, the unit sphere is the set of solutions (x, y, z) R
3
to the equation
x
2
+ y
2
+ z
2
1 = 0.
while an ellipse is the set of solutions (x, y) R
2
to the equation
x
2
a
2
+
y
2
b
2
1
2
= 0.
where a, b R are constants. Figure 7.3 depicts several algebraic varieties and their dening
equations.
The geometric properties of algebraic varieties are closely related to the algebra the poly-
nomial ring R[x
1
, x
2
, . . . , x
n
] (Example 98e)), and classical algebraic geometry explored this
relationship. The scope of modern algebraic geometry is far greater. For example, a dieren-
tiable variety is a solution to an equation
f(x
1
, x
2
, . . . , x
n
) = 0, (7.23)
where f : R
n
R is any dierentiable function. The geometry of such varieties is related to
ring c

(R
n
) of dierentiable functions (Example 3i)). Any dierentiable manifold in R
n
can
be represented as a dierentiable variety; thus, modern algebraic geometry subsumes a large
part of dierential geometry.
At a topological level, any closed subset of R
n
can be represented as the set of solutions to
an equation like (7.23), where f : R
n
R is some continuous function. Thus, the topology of
R
n
is encoded by the structure of the ring c(R
n
) of continuous functions (Example 3c)). More
generally, if X is any regular topological space (eg. any manifold, metric space, etc.), then
the entire topological structure of X is encoded in the structure of the ring c(X) of continuous
real-valued functions on X (Example (3d)).
96 CHAPTER 7. INTRODUCTION
We can also reverse the direction of this correspondence. Given an arbitrary commutative
ring 1, we ask, Is there some space X so that 1 is the ring of continuous functions on X?
Under suitable conditions, the answer is yes; this space is called the spectrum or Zariski
topology of 1, and we can learn much about 1 by studying this space. Thus, there is a
geometric interpretation for any commutative ring.
The study of this relationship is sometimes called commutative algebra, since it concerns
commutative rings. However, recently, there has been an eort to generalize these methods
to noncommutative rings. Certain geometric structures (for example, foliations of manifolds)
can be represented with a noncommutative ring, just as the topology of a space X can be
represented using the commutative ring c(X). This burgeoning eld is called noncommutative
geometry.
We will indicate when a particular topic is relevant to algebraic geometry by placing the
ag Alg.Geo. in the margin.
7.3.4 Algebraic Topology & Homological Algebra
Algebraic topology studies the global topological properties of spaces by measuring their
structure using algebraic objects such as homotopy groups and (co)homology groups. These
objects are called homotopy invariants, because for a given space X, you will always measure
the same homotopy or homology groups for X, no matter how X is deformed.
Given two spaces X and Y, we might ask: Is Y really just a deformed version of X, or are
they fundamentally dierent? We can answer this question by computing homotopy invariants.
If X and Y yield dierent invariants, then they cannot be the same
1
.
This endeavour requires a lot of algebraic machinery, for two reasons:
First, to compute the homotopy invariants of topological spaces.
Second, to compare the invariants of two spaces, to see if they are the same or not.
This algebraic machinery involves rings, modules, and homomorphisms between modules, and
is called homological algebra.
We will indicate when a particular topic is relevant to algebraic topology or homological
algebra by placing the ag Alg.Top. in the margin.
7.3.5 Group Representation Theory & Ring Structure Theory
A representation of a group ( is a homomorphism : (GL
n
[R]. This allows us to study
the structure of ( using the well-understood algebra of matrices. Given a group (, we ask,
What representations of ( exist, and what can they tell us about (?
Representation theory has many important applications. For example:
1
On the other hand, if they yield the same invariants, X and Y might be the same, but they also might not
be.
7.4. SUBRINGS 97
Physics: Groups represent the symmetries of physical laws. Thus, group representation theory
plays central role in the Standard model of subatomic physics.
Nonabelian Harmonic Analysis: Fourier theory (aka harmonic analysis) provides a power-
ful tool for studying functions and probability distributions on the real line R, Euclidean
space R
n
, the circle T
1
, or the torus T
n
. Fourier theory is essential for solving dierential
equations, and ubiquitous in probability theory.
Fourier theory is actually the representation theory of certain abelian groups. To gener-
alize the methods of Fourier Theory to nonabelian groups (eg. linear groups, Lie groups),
we must study the representation theory of these groups.
Given a representation : (GL
n
[R], we can immediately extend to a ring homo-
morphism

: R(/
n
(R) from the group ring over ( to the ring of n n matrices (see
Example 96b) on 80, and Example 103) on page 90). Conversely, any ring homomorphism

: R(/
n
(R) determines a representation of the group (. Thus, the representation theory
of the group ( is closely related to the group ring R(. Indeed, we can completely classify the
representations of ( using the structure theory of R(.
We will indicate when a particular topic is relevant to group representation theory or ring
structure theory by placing the ag Grp.Repr. in the margin.
7.4 Subrings
Prerequisites: 7.1
If 1 is a ring with operations + and , then a subring of 1 is any subset o 1 which
is also a ring, under the same operations. We indicate this by writing: o < 1.
Example 104:
(a) 2Z < Z < Q < R < C < H (see Examples (93a-93g)).
(b) Z < Z[i] (see Example 93f)).
(c) If n N, then nZ is a subring of Z.
(d) For any n N, /
n
(Z) < /
n
(Q) < /
n
(R) < /
n
(C) < /
n
(H) (see Examples
(96b-96e)).
In general, if o is a subring of 1, then /
n
(o) is a subring of /
n
(1) (see Example
96f).
(e) R[x] < c

(R) < c

(R) < . . . < c

3
(R) < c
2
(R) < c
1
(R) < c(R) < R
R
(see
Examples (97a), (3f- 3h), (98a), (99a), and (100c).
98 CHAPTER 7. INTRODUCTION
(f) Z[x] < Q[x] < R[x] < C[x] (see Examples (98a-98d)).
In general, if o is a subring of 1, then o[x] is a subring of 1[x] (see Example 98f)).
(Exercise 67).
(g) R[x] < R[[x]]. (Examples (98a) and (99c)).
More generally, for any ring, 1[x] < 1[[x]] (Examples (98f) and (99d)) (Exercise 68).
(h) C[[x]] < C((x)). (Examples (99c) and (99f)).
More generally, for any ring, 1[[x]] < 1((x)). (Examples (99d) and (99g)) (Exercise 69).
7.5 Units and Fields
Galois
Prerequisites: 7.1
We have seen that, in a general ring, not all elements have multiplicative inverses. A unit
of 1 is an element with a multiplicative inverse. The set of all units of 1 is denoted 1

.
A eld is a commutative ring 1 such that all nonzero elements of 1 are units. In other
words: 1

= 1 0.
Example 105:
(a) Z is not a eld: the only units of Z are 1 and 1.
(b) Q and R are elds. If r ,= 0, then it is a unit, with multiplicative inverse 1/r.
(c) C is a eld; if x+yi ,= 0, then it is a unit, with multiplicative inverse
x yi
x
2
+ y
2
(Exercise 70).
(d) H is not commutative, so it cant be a eld. However, H wants to be a eld, because
every nonzero element is a unit (Exercise 71). We say that H is a division ring.
(e) Let 1 = /
n
(R) from Example 96b). The set of units of /
n
(R) is the set of all invertible
n n matrices. In other words, /

n
(R) = GL
n
[R].
(f) Let 1 = R[x] from Example 98a). The units of R[x] are just the nonzero constant
polynomials that is, polynomials of the form: p(x) = p
0
+0x +0x
2
+. . ., where p
0
,= 0.
Thus, R[x]

is identical to R

. Thus R[x] is not a eld.

(g) Let 1 be any ring, and consider 1[x] from Example 98f). The units of 1[x] are just the
unit constant polynomials that is, polynomials of the form: p(x) = p
0
+0x +0x
2
+. . .,
where p
0
1

. Thus, 1[x]

is identical to 1

, and thus, 1[x] is never itself a eld.

7.6. ZERO DIVISORS AND INTEGRAL DOMAINS 99
(h) Let 1 = c(R) from Example 97a). An element of c(R) has a multiplicative inverse if and
only if f is nonzero everywhere; in this case, the inverse of f is the function g(x) =
1
f(x)
.
The same holds for 1 = c
k
(R) for any 1 k (see Examples (3f-3h)).
(i) Let 1 = c

(U;

C), as in Example (99e). Then 1 is a eld: if f is a nonzero meromorphic
function, and g(x) =
1
f(x)
, then g is also meromorphic. The zeros of f become the poles
of g and vice versa (Exercise 72 Verify these claims.)
(j) Let T be a eld, and let 1 = T((x)) be the ring of formal Laurent series over T, as in
Example 99g). Then 1 is also a eld (Exercise 73).
(k) Let R(x) denote the set of rational functions, as in Example 98h). Then R(x) is a eld
(Exercise 74).
Lemma 106 Let 1 be any ring. Then 1

forms a group under multiplication. This is

called the group of units of 1.
Proof: Exercise 75 2
Example 107: Z
/n

Let n N. If k Z
/n
then we saw earlier that
_
k has a multiplicative inverse, mod n
_

_
k is relatively prime to n ie. gcd(n, k) = 1
_
Thus,
Z
/n

=
_
k Z
/n
; k is relatively prime to n
_
.
In particular, it follows that
_
Z
/n
is a eld
_

_
Z
/n

=
_

1,

2, . . . , n 1
_
_

_
n is prime
_
.
7.6 Zero Divisors and Integral Domains
Num.Thr.
Alg.Geo.
Alg.Top.
Prerequisites: 7.5
Let r 1 be nonzero. We say r is a zero divisor if there is some element s 1 so that
s ,= 0, but r s = 0.
Example 108:
(a) Let 1 = Z
/10
, as in Example 93i). Let r =

5, and let s =

2. Then r ,= 0 ,= s, but
r s = 10 =

0. Thus,

5 and

2 are both zero divisors in Z
/10
.
100 CHAPTER 7. INTRODUCTION
g
f
(A)
g
f
(B)
g
f
(C)
Figure 7.4: Zero divisors in the rings c(R), c
K
(R), and c

(R).
(b) Let 1 = Z
/n
, as in Example 93i). If n = p q for any integers p, q > 1, then p and q are
zero divisors, because p q =

0.
(c) Let 1 = R
2
with the product ring structure, as in Example 95a). Let r = (1, 0) and
s = (0, 1). Then r s = (0, 0), so r and s are zero divisors.
(d) Let 1 = /
2
(R), as in Example 96a). Let A =
_
1 0
0 0

and let B =
_
0 0
0 1

. Then
A B =
_
0 0
0 0

, so A and B are zero-divisors.

(e) Let 1 = c(R), as in Example 97a). As shown in Figure 7.4(A), let
f(x) =
_
0 if x 0
x if x > 0
and g(x) =
_
x if x 0
0 if x > 0
Then f and g are both nonzero continuous functions, but f g = 0. Thus, f and g are
zero-divisors.
(f) Let 1 = c
k
(R) for some k N, as in Example (3g). As shown in Figure 7.4(B), let
f(x) =
_
0 if x 0
x
k+1
if x > 0
and g(x) =
_
x
k+1
if x 0
0 if x > 0
Then f c
k+1
(R) and g c
k+1
(R) are both nonzero, but f g = 0. Thus, f and g are
zero-divisors.
(g) Let 1 = c

(R), as in Example (3h). As shown in Figure 7.4(C), let

f(x) =
_
0 if x 0
exp
_

1
x
2
_
if x > 0
and g(x) =
_
exp
_

1
x
2
_
if x 0
0 if x > 0
Then f c

(R) and g c

(R) are both nonzero, but f g = 0. Thus, f and g are

zero-divisors.
7.6. ZERO DIVISORS AND INTEGRAL DOMAINS 101
(h) Let 1 = R[x], as in Example 98a). Then 1 has no zero divisors, because the product of
any nonzero polynomials is nonzero.
(Exercise 76 Verify this. Hint: use formula (7.8) on page 83.)
(i) Let 1 = Z. Then 1 has no zero divisors, because the product of any nonzero integers is
nonzero.
An integral domain is a commutative ring with no zero divisors.
Example 109:
(a) Z is an integral domain, by Example 108i).
(b) Likewise, Q, R, and C are all integral domains.
(c) Any eld is an integral domain (Exercise 77).
(d) Z
/10
is not an integral domain, because

2 and

5 are zero divisors, from Example 108a).
(e) In general, if n = pq for any integers p, q > 1, then Z
/n
is not an integral domain, because
p and q are zero divisors, from Example 108b).
(f) Thus, combining Examples (107), (109c) and (109e), we conclude:
_
Z
/n
is an integral domain
_

_
n is prime
_

_
Z
/n
is a eld
_
.
(g) The polynomial ring R[x] is an integral domain, by Example (108h).
(h) Likewise, Z[x], Q[x], and C[x] are all integral domains.
(i) If 1 is a commutative ring, then the following are equivalent (Exercise 78):
1 is an integral domain.
1[x] is an integral domain.
1[x
1
, x
2
, . . . , x
n
] is an integral domain, for all n N.
(j) The ring c

(R) is not an integral domain, by example (108g).

(k) However, c

(R) is an integral domain. To see this, we combine the following two asser-
tions:
(i) If f, g : RR are continuous functions, and f g = 0, then there exist nonempty
open sets U R and V R so that f(u) = 0 for all u U, and g(v) = 0 for all
v V
102 CHAPTER 7. INTRODUCTION
(ii) (The Identity Theorem) If f : RR is an analytic function, and there is a
nonempty open set U R so that f(u) = 0 for all u U, then f = 0 every-
where.
Exercise 79 Use (i) and (ii) to show that c

(R) has no zero divisors.

(The proofs of assertions (i) and (ii) are beyond the scope of these notes.)
(l) Let T be any integral domain, and let 1 = T[[x]] be the ring of formal power series over
T, as in Example 99d). Then 1 is also an integral domain (Exercise 80).
Integral domains are useful because they allow us to cancel common factors out of equa-
tions. When confronted with an algebraic expression of the form
r s = r t,
we are tempted to cancel the r, and conclude that s = t. This cancellation law is valid in a
group; we simply multiply both sides by r
1
to get s = r
1
rs = r
1
rt = t. Cancellation
also works if 1 is a eld, and r is a nonzero element (therefore invertible).
However, the cancellation law is not valid in an arbitrary ring. For example, if 1 = Z
/10
,
then

3 =

6 =

2

8 even though

2 ,=

8.
The problem here is that

2 is a zero divisor in Z
/10
. To obtain a cancellation law, we must
restrict ourselves to integral domains.
Proposition 110 (Cancellation Law for Integral Domains)
Let 1 be an integral domain, and let r 1 be nonzero. Then for any s, t 1,
_
rs = rt
_

_
s = t
_
.
Proof:
_
rs = rt
_

_
rs rt = 0
_

_
r(s t) = 0
_
.
Since r is nonzero, and cannot be a zero divisor, we conclude that (s t) = 0. In other
words, s = t. 2
Chapter 8
Homomorphisms, Quotients, and Ideals
8.1 Homomorphisms
Prerequisites: 7.1
Consider the ring Z
/5
from Example 93h). For any z Z, let z Z
/5
be the congruence
class of z, mod 5. The arithmetic of Z
/5
has the following convenient property: for any z
1
, z
2

Z,
z
1
+ z
2
= z
1
+ z
2
; and z
1
z
2
= z
1
z
2
. (8.1)
Dene the function : ZZ
/5
by (z) = z. Then the equations (8.1) can be rewritten:
(z
1
+ z
2
) = z
1
+ z
2
= z
1
+ z
2
= (z
1
) + (z
2
);
and (z
1
z
2
) = z
1
z
2
= z
1
z
2
= (z
1
) (z
2
);
We say that is a homomorphism from the ring Z to the ring Z
/5
.
In general, let 1 and o be rings. A ring homomorphism is a map : 1o so that,
for any r
1
, r
2
1
(r
1
+ r
2
) = (r
1
) + (r
2
) and (r
1
r
2
) = (r
1
) (r
2
).
Example 111:
(a) Let 1 = Z and let o = Z
/5
(Example 93h)), and dene : ZZ
/5
by (z) = z. Then
is a ring homomorphism.
(b) More generally, let n N, and dene : ZZ
/n
by (z) = z, where z is the congruence
class of z, modulo n. Then is a ring homomorphism.
(c) Let 1 = R
2
be the product ring from Example 95a), and let o = R. Let
1
: R
2
R
be the projection into the rst coordinate. Then
1
is a homomorphism, because for any
103
104 CHAPTER 8. HOMOMORPHISMS, QUOTIENTS, AND IDEALS
(x
1
, y
1
) and (x
2
, y
2
) in R
2
,

1
_
(x
1
, y
1
) + (x
2
, y
2
)
_
=
1
_
(x
1
+ x
2
), (y
1
+ y
2
)
_
= x
1
+ x
2
=
1
(x
1
, y
1
) +
1
(x
2
, y
2
).
and
1
_
(x
1
, y
1
) (x
2
, y
2
)
_
=
1
_
(x
1
x
2
), (y
1
y
2
)
_
= x
1
x
2
=
1
(x
1
, y
1
)
1
(x
2
, y
2
).
(d) Let 1 = c(R), as in Example 97a), and let o = R. Dene the evaluation map

0
: c(R)R by
0
(f) = f(0). Then
0
is a ring homomorphism, because

0
(f
1
+ f
2
) = (f
1
+ f
2
)(0) = f
1
(0) + f
2
(0) =
0
(f
1
) +
0
(f
2
);
and
0
(f
1
f
2
) = (f
1
f
2
)(0) = f
1
(0) f
2
(0) =
0
(f
1
)
0
(f
2
);
(e) More generally, let r R be any xed real number, and dene the evaluation map

r
: c(R)R by
r
(f) = f(r). Then
r
is a ring homomorphism.
(f) Let 1 = R[x], as in Example 98a), and let
0
: R[x]R be the evaluation map, as
before. Again,
0
is a homomorphism.
Observe that, if P(x) = p
n
x
n
+ . . . + p
2
x
2
+ p
1
x + p
0
, then
0
(P) = p
0
.
(g) Let o be any ring, and let 1 = o[x], as in Example (98f). Now dene
0
: o[x]o
as follows: If P(x) = p
n
x
n
+ . . . + p
2
x
2
+ p
1
x + p
0
, then
0
(P) = p
0
. Then
0
is a
homomorphism (Exercise 81).
(h) Let a < b , and dene the restriction map
(a,b)
: c(R)c(a, b) where

(a,b)
(f) is just the restriction of f to a function f

(a,b)
: (a, b)R. That is, for any
r (a, b), f

(a,b)
(r) = f(r), but f

(a,b)
is not dened outside of the domain (a, b). Then

(a,b)
is a homomorphism (Exercise 82).
(i) More generally, let X be any topological space and let U X be any subspace. Dene
the restriction map
U
: c(X)c(U) by
U
(f) = f
[
U
. Then
U
is a homomorphism.
(j) Let 1 = /
2
(Z) be the ring of 2 2 integer matrices, as in Example 96c), and let o =
/
2
(Z
/5
) be the ring of 22 matrices with coecients in Z
/5
. Dene the homomorphism
: 1o by reducing each coecient mod 5. That is,

_
a b
c d
_
=
_
a b
c d
_
,
where a is the congruence class of a, mod 5, etc. Then is a ring homomorphism.
8.1. HOMOMORPHISMS 105
(k) Let 1 be any ring with multiplicative identity 1
R
. There is a natural homomomorphism
: Z1 given by (n) = 1
R
+ . . . + 1
R
. .
n
for any n N, and (n) = (n).
Indeed, this is the only homomorphism from Z into 1. (Exercise 83).
Lemma 112 Let : 1o be a ring homomorphism. Then:
1. (0
R
) = 0
S
.
2. For any r 1, (r) = (r). If r is a unit, then (r
1
) = (r)
1
.
3. (1) is a subring of o.
Proof: Exercise 84 2
Let 0
S
be the zero of the ring o. The kernel of is the preimage of 0
S
:
ker[] =
1
0 = r 1 ; (r) = 0
S
.
Example 113:
(a) Dene : ZZ
/5
by (z) = z, as in Example 111a). Then
ker() = z Z ; z = 0 (mod 5) = . . . , 5, 0, 5, 10, . . . = 5Z.
(b) More generally, let n N, and dene : ZZ
/n
by (z) = z, as in Example 111b).
Then ker() = nZ.
(c) Let
1
: R
2
R be the projection into the rst coordinate, as in Example 111c). Then
ker() =
_
(x, y) R
2
; x = 0
_
= (0, y) ; y R.
(d) Let r R, and Let
r
: c(R)R be as in Example 111e). Then ker(
r
) = f c(R) ; f(r) = 0.
(e) Let
0
: 1[x]1 be the evaluation map, as in Example 111g). If P(x) = p
n
x
n
+ . . . +
p
2
x
2
+ p
1
x + p
0
, then
_
P ker(
0
)
_

_
p
0
= 0
_

_
P(x) = p
n
x
n
+ . . . + p
2
x
2
+ p
1
x = x (p
n
x
n1
+ . . . + p
2
x
1
+ p
1
)
_
Thus, ker(
0
) = x q(x) ; q 1[x] any polynomial.
(f) Let a < b , let
(a,b)
: c(R)c(a, b) be the restriction map, as in Example
111h). Then ker(
(a,b)
) = f c(R) ; f(x) = 0 for all x (a, b).
(g) Dene the homomorphism : /
2
(Z)/
2
(Z
/5
) by reducing each coecient mod 5, as
in Example 111j). Then ker() consists of matrices whose entries are all multiples of 5.
That is, ker() = /
2
(5Z). (Exercise 85)
106 CHAPTER 8. HOMOMORPHISMS, QUOTIENTS, AND IDEALS
8.2 () Many Examples of Ring Homomorphisms
Prerequisites: 8.1
[This section contains many examples. It is not necessary to read and understand all these examples right
now; just read a few in order to get your intuititions working. These examples will be referred to throughout
the text.]
A ring monomorphism is a ring homomorphism that is injective (ie. one-to-one).
Example 114: (Ring Monomorphisms)
a) Dene the map : ZR by (z) = z for any z Z. (In other words, transforms
z considered as an element of Z to z considered as an element of R). Then is a ring
monomorphism.
b) More generally, if o is a subring of 1, we dene the inclusion map : o1by (s) = s
for any s o. Then is a monomorphism.
c) Dene : RR[x] as follows: for any r R, (r) is the constant polynomial r +0x+
0x
2
+ . . .. Then is a monomorphism.
d) Dene : Rc(R) as follows: for any r R, (r) is the constant function with value
r. Then is a monomorphism.
A ring epimorphism is a homomorphism that is surjective (ie. onto).
Example 115: (Ring Epimorphisms)
a) Let
0
: R[x]R be the evaluation map of Example 111g). Then
0
is an epimorphism
(Exercise 86).
b) Dene
i
: R[x]C by mapping any polynomial p(x) = p
n
x
n
+ . . . + p
2
x
2
+ p
1
x + p
0
to the complex number p(i) = p
n
i
n
+ . . . + p
2
i
2
+ p
1
i + p
0
For example, if p(x) =
3x
5
7x
4
+ 2x
3
3x
2
+ 4x 5, then

i
(p) = p(i) = 3(i) 7(1) + 2(i) 3(1) + 4i 4 = 8 + 5i
Then
i
is an epimorphism (Exercise 87).
c) Let Z[x] be as in Example 98c) and let Z[i] be as in Example 93f). Dene
i
: Z[x]Z[i]
by mapping any polynomial p(x) = p
n
x
n
+ . . . + p
1
x + p
0
to the complex number p(i) =
p
n
i
n
+ . . . + p
1
i + p
0
. Then
i
is an epimorphism (Exercise 88).
A ring isomorphism is a homomorphism that is bijective (ie. one-to-one and onto).
Example 116: (Ring Isomorphisms)
8.2. () MANY EXAMPLES OF RING HOMOMORPHISMS 107
a) Let /
2
(Z) be the ring of 2 2 integer matrices (Example 96c) on page 80) and let
End [Z
2
] be the ring of endomorphisms of Z
2
(Example 101a) on page 88).
If M =
_
m
1
m
2
m
3
m
4
_
is a matrix in /
2
(Z), then dene endomorphism
M
End [Z
2
] as
follows: for any element z = (z
1
, z
2
) in Z
2
,

M
(z) =
_
m
1
m
2
m
3
m
4
__
z
1
z
2
_
=
_
m
1
z
1
+ m
2
z
2
m
3
z
1
+ m
4
z
2
_
The map M
M
is an isomorphism /
2
(Z) into End [Z
2
] (Exercise 89).
b) More generally, let /
n
(Z) be the ring of n n integer matrices and End [Z
n
] be the
endomorphism ring of Z
n
, and for any M /
n
(Z), let
M
End [Z
n
] by multiplication-
by-M. Then the map M
M
is an isomorphism /
n
(Z) into End [Z
n
] (Exercise 90).
c) Fix a number N N. Let 1 be any ring, and let 1
N
be the product ring from Example
95d) on page 79. Let ^ = 1, 2, . . . , N, and let R
N
be the set of all functions from ^
into R (Example 100b) on page 88). Dene : 1
N
1
N
as follows:
Given any function F 1
N
, dene the N-tuple f = (f
1
, f
2
, . . . , f
N
) by f
n
= F(n) for all
n ^. Then the map F f is an isomorphism of 1
N
into 1
N
(Exercise 91).
d) Let X be a set, and let T(X) be the ring of subsets of X (Example 102) on page 89).
Let Z
/2
X
be the ring of all functions from X into Z
/2
(Example 100d) on page 88).
Dene : Z
/2
X
T(X) as follows: for any function f Z
/2
X
, let (f) be the set
x X ; f(x) = 1. Then is a ring isomorphism (Exercise 92).
e) Let c =
__
x y
y x
_
; x, y R
_
be the set of all 2 2 conformal matrices. Then c is a
subring of /
2
(R) (Exercise 93).
Dene the map : Cc by (x + yi) =
_
x y
y x
_
. Then is a ring isomorphism
(Exercise 94).
f) Cayleys Theorem for Rings: Any ring with identity is isomorphic to the endomor-
phism ring of some abelian group.
Proof: Let 1 be any ring with a multiplicative identity. Let

1 be the additive abelian
group obtained by forgetting the multiplicative structure on 1, and just treating 1
as a group under addition. If r is any element of 1, let r denote the corresponding
element of

1.
(For example, if 1 = C, then

1 = R
2
. If z = x + yi is a complex number, then z = (x, y) R
2
).
Let End
_

1
_
be the ring of endomorphisms of

1from Example 101) on page 88. There
is a natural homomorphism : 1End
_

1
_
, dened as follows: For any r 1, let
108 CHAPTER 8. HOMOMORPHISMS, QUOTIENTS, AND IDEALS
(r) =
r
be the map
r
:

1

1 dened:
r
( s) = r s. (Exercise 95 Verify that
this is a homomorphism.)
(For example, if 1 = C and

1 = R
2
, then for any complex number z = x +yi,
z
: R
2
R
2
is just
multiplication by the conformal matrix
_
x y
y x
_
from Example 116e).)
If c = (1), then c is a subring of End
_

1
_
, and is an isomorphism from 1 to c.
(Exercise 96 Hint: it is important that 1 has an identity element.)
(For example, if 1 = C, and

1 = R
2
, then c is just the ring of 2 2 real-valued conformal matrices,
as in Example 116e).) 2
Cayleys theorem for groups says that any group ( can be seen as a permutation group; this
is done by letting ( act upon itself. Thus, the class of permutation groups are universal
for groups.
Similarly, Cayleys theorem for rings says that any ring 1can be seen as an endomorphism
ring; again this is done by letting 1 act upon itself. Thus, the class of endomorphism
rings are universal for rings.
8.3 Ideals
Prerequisites: 7.4, 8.1
Consider the set 2Z of all even numbers, and recall:
(a) The sum of any two even numbers is an even number.
(b) The product of any two even numbers is an even number.
(c) If e is even, and n is any other number, then e n is also even.
Facts (a) and (b) together imply that 2Z is a subring. Fact (c) says that 2Z has an interesting
contagious property; multiplying any number by an even number makes it even. We say that
2Z is an ideal of Z.
Let 1 be any ring. An ideal is a subring 1 < 1 such that:
For any r 1, and any i 1, r i 1, and i r 1.
We then write: 1 1.
Example 117:
(a) 2Z is an ideal of Z.
(b) More generally, let n N, and let 1 = nZ be the set of all multiples of n. Then nZ is
an ideal. To see this, rst recall from Example 104c) that nZ is a subring of Z. Next,
observe that, if i = n z is any element of nZ, and r Z is any other integer, then
i r = n (z r) is also in nZ.
8.3. IDEALS 109
(c) Let 1 = R
2
be the product ring from Example 95a), and let 1 = (0, y) ; y R. Then
1 is an ideal. To see this, rst observe that 1 is a subring: for any (0, y
1
) and (0, y
2
) in 1,
(0, y
1
) + (0, y
2
) = (0, y
1
+ y
2
) and (0, y
1
) (0, y
2
) = (0, y
1
y
2
)
are also in 1. If (x, y) is any element of 1, and (0, i) 1, then (0, i) (x, y) = (0, i y)
is also in 1.
(d) Let 1 = c(R), as in Example 97a). Fix r 1, and dene
/
r
= f c(R) ; f(r) = 0.
Then /
r
is an ideal. To see this, suppose m
1
, m
2
are elements of /
r
, and let f c(R)
be any other function. Then
(m
1
+ m
2
)(r) = m
1
(r) + m
2
(r) = 0 + 0 = 0, so m
1
+ m
2
is in /
r
.
(m
1
m
2
)(r) = m
1
(r) m
2
(r) = 0 0 = 0, so m
1
m
2
is in /
r
.
(m
1
f)(r) = m
1
(r) f(r) = 0 f(r) = 0, so m
1
f is in /
r
.
(e) Let o be any ring, and let 1 = o[x], as in Example (98f). Let
1 = x q(x) ; q 1[x] any polynomial.
Then 1 is an ideal (Exercise 97).
(f) Let 1 = c(R), as in Example 97a). Let a < b , and dene 1 =
f c(R) ; f(x) = 0 for all x (a, b). Then 1 is an ideal (Exercise 98).
(g) Let 1 = /
2
(Z) be the ring of 2 2 integer matrices, as in Example 96c), and let
1 = /
2
(5Z) be the ring of 2 2 matrices with coecients in 5Z for example
_
5 10
20 15

.
Then /
2
(5Z) is an ideal of /
2
(Z) (Exercise 99).
(h) More generally, let 1 be any ring and let 1 1. Let /
n
(1) be the set of n n
matrices with coecients in 1, as in Example 96f). Then /
n
(1) is an ideal of /
n
(1)
(Exercise 100).
(i) Let 1 = Z[x] be the ring of polynomials with integer coecients, as in Example 98c). Let
5Z[x] be the ring of polynomials with coecients in 5Z (for example, p(x) = 15x
3
5x =
10). Then 5Z[x] is an ideal of Z[x] (Exercise 101).
(j) More generally, let 1 be any ring and let 1 1. Let 1[x] be the set of all formal
polynomials with coecients in 1, as in Example 98f). Then 1[x] is an ideal of 1[x]
(Exercise 102).
110 CHAPTER 8. HOMOMORPHISMS, QUOTIENTS, AND IDEALS
(k) Principal ideals in commutative rings: In Example 117a), we saw that 2Z =
2z ; z Z is an ideal of Z. In Example 117e), we saw that 1 = x q(x) ; q 1[x]
is an ideal of 1[x]; we could have written this ideal as 1 = x R[x].
In general, if 1 is any commutative ring, and p 1, then the principal ideal generated
by p is the set
p1 = pr ; r 1. Exercise 103 Show that this is an ideal.
We normally denote this ideal by (p). For example, if 1 = Z, then (3) = 3Z.
(l) Let 1 = c(R), as in Example 97a) on page 81, and let 1 = c
0
(R), as in Example 3k).
Then 1 1 (Exercise 104).
Ideals are to rings as normal subgroups are to groups...
Proposition 118 Let : 11 be a ring homomorphism. Then ker() is an ideal of 1.
Proof: Let k
1
, k
2
ker(), and let r 1. Then
(k
1
+ k
2
) = (k
1
) + (k
2
) = 0 + 0 = 0, so k
1
+ k
2
is in ker().
(k
1
k
2
) = (k
1
) (k
2
) = 0 0 = 0, so k
1
k
2
is in ker().
(k
1
f) = (k
1
) (f) = 0 (f) = 0, so k
1
f is in ker().
and (f k
1
) = (f) (k
1
) = (f) 0 = 0, so f k
1
is in ker(). 2
Example 119:
(a) Example 117a) is just the kernel of the homomorphism : ZZ
/2
.
(b) Example 117b) is just the kernel of the homomorphism : ZZ
/n
, from Example
(113b).
(c) Example 117c) is just the kernel of the homomorphism
1
: R
2
R, from Example
113c)
(d) Example 117d) is the kernel of the evaluation map
r
: c(R)R, from Example 113d).
(e) Example 117e) is the kernel of the evaluation map
0
: 1[x]1, from Example (113e).
(f) Example 117f) is the kernel of the restriction map
(a,b)
: c(R)c(a, b) from Example
113f).
(g) Example 117g) is the kernel of the homomorphism : /
2
(Z)/
2
(Z
/5
) from Example
113g).
Ideals are allergic to multiplicative inverses, in the following sense:
8.4. QUOTIENT RINGS 111
Lemma 120 Let 1 be a ring with multiplicative identity 1, and let 1 be an ideal. The
following are equivalent:
(a) 1 contains some unit element u.
(b) 1 1.
(c) 1 = 1.
Proof: Exercise 105 2
Corollary 121 Let F be a eld
1
. Then F contains no proper ideals.
Proof: Let 1 be any nontrivial ideal of F. Thus, 1 contains some nonzero element u F.
But F is a eld, so u is a unit, so 1 = F by the previous theorem. 2
We say that a ring 1 is simple if it contains no proper ideals. Thus, Corollary 121 can be
reformulated: Every eld is a simple ring.
8.4 Quotient Rings
Prerequisites: 8.3 Recommended: 1.4
Let 1 be a ring. If 1 < 1 is any subring, and r 1, then the corresponding coset of 1 is
the set:
r +1 = r + i ; i 1.
Example 122:
(a) Let 1 = Z, and let 1 = 5Z = 5z ; z Z = . . . , 5, 0, 5, 10, 15, . . . as in
Example 117b). Then 3 + 5Z = 3 + 5z ; z Z = . . . , 2, 3, 8, 13, 18, . . ..
(b) Let o and T be rings, and consider the product ring
1 = o T = (s, t) ; s o and t T . (Example 95) on page 78)
Let 1 = (0
S
, t) ; t T = 0
S
T . Then for any (s, t
1
) 1,
(s, t
1
)+1 = (s, t
1
) + (0
S
, t
2
) ; t
2
T = (s, t
1
+ t
2
) ; t T = (s, t) ; t T
= s T .
1
...or a division ring.
112 CHAPTER 8. HOMOMORPHISMS, QUOTIENTS, AND IDEALS
Lemma 123 Let 1 be a ring and let 1 < 1 be a subring. For any r 1:
(a)
_
r +1 = 1
_

_
r 1
_
.
(b)
_
r 1
_
=
_
r 1 1
_
. However, the converse is not generally true.
(For example, if 1 is an ideal, then r1 1 for any r 1.)
Proof: Exercise 106 2
Example 124:
(a) Let 1 = Z, and let 1 = 5Z as in Example 122a). Then 10+5Z = 10 + 5z ; z Z =
. . . , 5, 10, 15, 20, 25, . . . = 5Z.
(b) Let o and T be groups; let 1 = o T and let 1 = 0
S
T , as in Example 122b).
Then for any t T , (0
S
, t) +1 = 0
S
T = 1.
The coset space of 1 is the set of all its left cosets:
1/1 = (r +o) ; r 1.
Example 125:
(a) Let 1 = Z, and let 1 = 5Z as in Example 122a). Then
Z
5Z
= 5Z, 1 + 5Z, 2 + 5Z, 3 + 5Z, 4 + 5Z.
(b) Let o and T be rings; let 1 = o T , and let 1 = 0
S
T , as in Example 122b).
Then
1
1
= 1
s
; s o, where, for any xed s o, 1
s
= (s, t) ; t T .
If /, B 1 are subsets, then their sum is the set
/+B = a + b ; a / and b B, (8.2)
and their product is the set
/ B = a b ; a / and b B. (8.3)
Lemma 126 Let 1 be a ring.
8.4. QUOTIENT RINGS 113
(a) Subset addition in 1 is associative and commutative. That is:
for any subsets /, B, c 1, /+ (B +c) = (/+B) +c,
and, for any subsets /, B 1, /+B = B +/.
(b) Subset multiplication in 1 is associative and distributive. That is, for any subsets
/, B, c 1,
/ (B c) = (/ B) c,
/ (B +c) = (/ B) + (/ c),
and (B +c) / = (B /) + (c /),
(c) If / < 1 is a subring of 1, then /+/ = / and / / /. Furthermore,
_
/ is an ideal of 1
_

_
1 / = / = / 1
_
Proof: Exercise 107 2
Proposition 127 Let 1be a ring, and let 1 < 1be a subring. The following are equivalent:
(a) 1 = ker() for some ring homomorphism : 1o (where o is some ring).
(b) 1 1.
(c) The coset space 1/1 is a ring under the addition operation (8.2) and multiplication
operation (8.3). Furthermore:
1. If (a +1) and (b +1) are cosets of 1, then
(a +1) + (b +1) = (a + b) +1, and (a +1) (b +1) = (a b) +1. (8.4)
2. Dene : 11/1 by: (r) = r + 1. Then is a ring epimorphism, and
ker() = 1.
Proof: (a)=(b) This is just Proposition 118 on page 110.
(b)=(c) Lemma 126 says these operations are associative etc.
Exercise 108 Verify equations (8.4). Check:
The additive identity of 1/1 is the coset 1 = (0 +1). (8.5)
Show that the additive inverse of the coset (r +1) is the coset (r +1). If 1 has a multiplicative
identity 1, show that 1/1 has multiplicative identity (1 +1).
114 CHAPTER 8. HOMOMORPHISMS, QUOTIENTS, AND IDEALS
The fact that is a homomorphism follows immediately from equations (8.4):
(a + b) = (a + b) +1 = (a +1) + (b +1) = (a) + (b),
and (a b) = (a b) +1 = (a +1) (b +1) = (a) (b).
Also, for any r 1,
_
r ker()
_

_
(r) = 0
_
by (8.5)
_
(r +1) = 1
_
Lem.123(a)
_
r 1
_
.
Hence, ker() = 1.
(c)=(a) Let o = 1/1, and let = . 2
The ring 1/1 is called the quotient ring, and the epimorphism : 11/1 is called the
projection map or quotient map.
Example 128: Let 1 = Z and let 1 = 5Z. Then 1/1 is the ring Z
/5
of congruence classes,
mod 5.
8.5 The Fundamental Isomorphism Theorems
Prerequisites: 8.4 Recommended: 2.1, 2.2, 2.3
Theorem 129 Fundamental Isomorphism Theorem
Let 1 and o be rings, and let : 1o be a ring homomorphism, with image T = (1)
o, and kernel /. Then:
(a) T

= 1//.
(b) For any r 1 with t = (r), the -preimage of t is the r-coset of /. That is:

1
t = (r +/).
Proof: Exercise 109 2
Corollary 130 Let 1 and o be rings, and let : 1o be a ring homomorphism. Then
_
is injective
_

_
ker() = 0
_
Proof: Exercise 110 2
8.6. THE RING ISOMORPHISM THEOREMS 115
Corollary 131 Let F be a eld
2
, and let 1 be a ring. Let : F1 be a group homomor-
phism. Then:
either is trivial (ie. (F) = 0)
or is injective, in which case (F) is isomorphic to F.
Proof: Exercise 111 Hint: Combine Corollaries 130 and 121. 2
8.6 The Ring Isomorphism Theorems
Prerequisites: 8.4 Recommended: 2.1, 2.2, 2.3
The three isomorphism theorems for groups have analogies for rings.
Theorem 132 Diamond Isomorphism Theorem
Let 1 be a ring. Let o < 1 be a subring, and let 1 1 be an ideal. Then:
(a) o +1 is a subring of 1.
(b) 1 (o +1).
(c) (o 1) o.
(d) There is an isomorphism:
o +1
1

=
o
o 1
given by the map
:
o +1
1

o
o 1
(s + i) +1 s + (o 1)
Proof: Exercise 112 2
Let 1 be a ring, with ideal 1 1 and subring < 1. Suppose 1 < < 1. Then 1 is
also a ideal of , and the quotient ring

1
= j +1 ; j
is a subset of the quotient ring
1
1
= r +1 ; r 1.
2
...or any simple ring, for that matter.
116 CHAPTER 8. HOMOMORPHISMS, QUOTIENTS, AND IDEALS
Theorem 133 Chain Isomorphism Theorem
Let 1 be a ring, with ideals 1 1 and 1. Suppose 1 < . Then:
(a)

1
is a ideal of
1
1
.
(b) There is an isomorphism
(1/1)
(/1)

=
1

.
(c) Use bar notation to denote elements of 1/1. Thus, (r + 1) = r, /1 = ,
1/1 = 1, and
1/1
/1
= 1/. Then the isomorphism : 1/1/ is dened:

_
r +
_
= r +.
Proof: Exercise 113 2
Theorem 134 Lattice Isomorphism Theorem
Let 1 be a ring and let 1 1 be an ideal. Let 1 = 1/1. Let L(1) be the subring lattice
of 1, and let L
I
(1) be the fragment of L(1) consisting of all subrings which contain 1. That
is:
L
I
(1) = / < 1 ; 1 < /.
Then there is an order-preserving bijection from L
I
(1) into L(1), given:
L
I
(1) / / L(1).
Furthermore, for any /, B, c, T L
I
(1),
(a)
_
/ < B
_

_
/ < B
_
.
(b) c T = c T.
(c)
_
/1
_

_
/1
_
, and in this case, 1//

= 1//.
Proof: Exercise 114 2
Chapter 9
Algebraic Geometry
9.1 Algebraic Varieties
Prerequisites: 7.5 Recommended: 7.3.3
Let F be a eld for example, F = Q, R, or C (usually F = C). We refer to the set F
n
as
the ane n-space over F (we consider F
n
merely as a set of points, not as a vector space or a
ring). If p(x
1
, . . . , x
n
) is a polynomial in F[x
1
,..., x
n
], then p determines a function p : F
n
F
in the obvious way.
An (ane) algebraic variety in F
n
is a subset V F
n
which is the set of common zeros
of some collection T F[x
1
,..., x
n
] of polynomials. That is:
V = V(T) = f F
n
; p(f) = 0 for all p T.
Example 135: Real Algebraic Varieties
In each of the following examples, let F = R.
(a) Circles: Consider the singleton set T = p(x, y), where p(x, y) = x
2
+ y
2
1. Then
V(T) =
_
(x, y) R
2
; p(x, y) = 0
_
=
_
(x, y) R
2
; x
2
+ y
2
= 1
_
is the circle about the origin of radius 1, which we denote S
1
(Figure 9.1A).
(b) Ellipses: Let T = p(x, y), where now, p(x, y) = (x/5)
2
+ (y/2)
2
1. Then
V(T) =
_
(x, y) R
2
; p(x, y) = 0
_
=
_
(x, y) R
2
; (x/5)
2
+ (y/2)
2
= 1
_
is an ellipse about the origin with minor axis 2 and major axis 5 (Figure 9.1B).
(c) Spheres: Let T = p(x
1
, . . . , x
n
), where p(x
1
, . . . , x
n
) = x
2
1
+ . . . + x
2
n
1. Then
V(T) = x R
n
; p(x) = 0 =
_
x R
2
; x
2
1
+ . . . + x
n
= 1
_
is the (n 1)-dimensional sphere about the origin of radius 1, which we denote S
n1
(Figure 9.1C).
117
118 CHAPTER 9. ALGEBRAIC GEOMETRY
2
5
1
(A) (B) (D) (F)
(C) (E)
Figure 9.1: Some algebraic varieties
(d) Planes: Let T = (x, y, z), where (x, y, z) = 3x 2y z. Then
V(T) =
_
(x, y, z) R
3
; (x, y, z) = 0
_
=
_
(x, y, z) R
2
; 3x 2y z = 0
_
is the plane through the origin, orthogonal to the vector (3, 2, 1) (Figure 9.1D).
(e) Torii: Let T = p(w, x, y, z), q(w, x, y, z), where p(w, x, y, z) = w
2
+ x
2
1 and
q(w, x, y, z) = y
2
+ z
2
1. Then
V(T) =
_
(w, x, y, z) R
4
; p(w, x, y, z) = 0 = q(w, x, y, z)
_
=
_
(w, x, y, z) R
2
; w
2
+ x
2
= 1 = y
2
+ z
2
_
is the 2-dimensional torus, which we denote T
2
(Figure 9.1E).
(f ) X: Let T = p(x, y), where p(x, y) = x
2
y
2
= (x + y) (x y). Then
V(T) =
_
(x, y) R
2
; x
2
y
2
= 0
_
=
_
(x, y) R
2
; x = y
_
is two diagonal lines which cross to make an X shape (Figure 9.1F).
Example 136: Algebraic Groups
An algebraic group is an algebraic variety with a natural group structure. We will not
give a formal denition now, but instead provide some simple examples.
(a) The special linear group SL
2
[R]: Let /
2
(R) be the set of 22 real matrices. Iden-
tify /
2
(R) with R
4
in the obvious way, so that a 2 2 matrix is written
_
w x
y z
_
. and let
9.1. ALGEBRAIC VARIETIES 119
T = d(w, x, y, z), where d(w, x, y, z) = wz yx 1 = det
_
w x
y z
_
1. Then
V(d) =
__
w x
y z
_
/
2
(R) ; d(w, x, y, z) = 0
_
=
__
w x
y z
_
/
2
(R) ; det
_
w x
y z
_
= 1
_
is the special linear group of 2 2 matrices, which we denote SL
2
[R].
(b) The special linear group SL
n
[R]: Now consider /
n
(R), which we identify with R
nn
in the obvious way. It is left as an exercise to check that the determinant function
det : /
n
(R)R is a polynomial. Thus, the function d(M) = det(M) 1 is also a
polynomial, and thus, the n n special linear group
SL
n
[R] = M /
n
(R) ; d(M) = 0
is an algebraic variety.
(c) The orthogonal group O
2
(R): Again, consider /
2
(R), and now let T = p, q, r, s,
where
p(w, x, y, z) = w
2
+ x
2
1; q(w, x, y, z) = wy + xz;
r(w, x, y, z) = yw + zx; and s(w, x, y, z) = y
2
+ z
2
1.
Then
V(p, q, r, s) =
__
w x
y z
_
/
2
(R) ;
p(w, x, y, z) = q(w, x, y, z) = 0
r(w, x, y, z) = s(w, x, y, z) = 0
_
=
__
w x
y z
_
/
2
(R) ;
_
w x
y z
_

_
w y
x z
_
=
_
1 0
0 1
__
=
_
_
w x
y z
_
/
2
(R) ;
_
w x
y z
_
1
=
_
w y
x z
_
_
is the orthogonal group of 2 2 matrices, which we denote O
2
(R).
(d) The orthogonal group O
n
(R): In a similar fashion, we can formulate a set of n
2
poly-
nomials p
11
, . . . , p
nn
on /
n
(R) so that
V(T) = M /
n
(R) ; p
ij
(M) = 0 for all i, j =
_
M /
n
(R) ; M
1
= M
t
_
is the orthogonal group of n n matrices, which we denote O
n
(R). The details are
Exercise 115 .
120 CHAPTER 9. ALGEBRAIC GEOMETRY
(e) The special orthogonal group SO
2
[R] Let T = p, q, r, s, d, where p, q, r and s are
as in Example (c), and d is as in Example (a). Then
V(p, q, r, s, d) = SL
2
[R] O
2
(R) = SO
2
[R]
is the special orthogonal group of 2 2 matrices.
Example 137: The Fermat Varieties
Let F = Q, and let p
n
(x, y, z) = x
n
+ y
n
z
n
. Clearly, p
n
(0, 0, 0) = 0. Fermats famous
Last Theorem states:
For any n 3, there exist no nonzero x, y, z Z so that p
n
(x, y, z) = 0.
This is actually equivalent to the (apparently stronger) statement:
For any n 3, there exist no nonzero x, y, z Q so that p
n
(x, y, z) = 0.
To see this, suppose that x =
x
1
x
2
, y =
y
1
y
2
, and z =
z
1
z
2
are rational numbers, and let L be
the lowest common multiple of the denominators x
2
, y
2
and z
2
. Then X = Lx, Y = Ly, and
Z = Lz are integers, and
_
x
n
+ y
n
= z
n
_

_
L
n
x
n
+ L
n
y
n
= L
n
z
n
_

_
X
n
+ Y
n
= Z
n
_
.
Hence, any rational solution to Fermats equation yields an integer solution. Conversely, any
integer solution to Fermats equation yields a rational solution, because integers are rational
numbers.
Hence, if we dene the rational algebraic variety
V
n
=
_
(x, y, z) Q
3
; x
n
+ y
n
= z
n
_
,
then Fermats Last Theorem can be reformulated: V
n
= (0, 0, 0) which is a statement of
algebraic geometry.
Many other Diophantine problems can be translated into algebraic geometry in this way.
Thus, there is a close relationship between number theory and algebraic geometry.
Example 138: Riemann Surfaces
Let F = C. A Riemann surface is an algebraic variety in C
2
determined by the solutions
to a single polynomial equation p(x, y) = 0. For example:
(a) The graph of a function: Let q : CC be a polynomial function (say, q(x) =
x
2
), and dene p(x, y) = y q(x). Then
V(p) =
_
(x, y) C
2
; p(x, y) = 0
_
=
_
(x, y) C
2
; y = q(x)
_
9.2. THE COORDINATE RING 121
is the graph of the function q. For example, if q(x) = x
2
, then
V
2
=
_
(x, x
2
) ; x C
_
.
Observe that there is a natural bijection : VC given by (x, y) = x. Thus, V can
be seen as a sort of deformation of the complex plane.
(b) The graph of a branched function: The complex square root function

has two
branches, each of which constitutes a function dened almost everywhere on the com-
plex plane. To simultaneously visualize both branches, we imagine graphing them both,
to obtain a surface
V
1/2
= (y,

y) ; y C,
where, for any y C, we understand

y to represent two distinct values.
The surface V
1/2
forms a branched double covering of the complex plane, in the following
sense. Consider the projection map : V
1/2
C dened (y,

y) = y, and observe
that is two-to-one everywhere except at (0, 0), where is one-to-one.
Observe that the Riemann surface V
1/2
is actually the same as the Riemann surface V
2
from Example (a), just with the coordinates reversed. In other words, we can dene a
natural bijection
V
2
(x, x
2
) (y,

y) V
1/2
by y = x
2
.
These examples illustrate an important fact about Riemann surfaces, which we state without
proof:
Let p(x, y) C[x, y] be a polynomial of degree n. Then the Riemann surface
V =
_
(x, y) C
2
; p(x, y) = 0
_
is always nonempty, and is a branched n-fold covering of the complex plane.
9.2 The Coordinate Ring
Prerequisites: 9.1, 8.6
Let F be a eld and let V F
n
be an algebraic variety. A coordinate function on V is
the restriction to V of some polynomial function p : F
n
F. The set of all coordinate functions
on V is called the coordinate ring of V, and denoted c
oord
(V). That is:
c
oord
(V) =
_
p
[
V
; p F[x
1
,..., x
n
]
_
.
If p, q F[x
1
,..., x
n
] are two polynomials such that p(v) = q(v) for all v V, then p
[
V
and q
[
V
are really the same function. Hence, elements of c
oord
(V) can be seen as equivalence classes of
polynomials in F[x
1
,..., x
n
], where
_
p q
_

_
p(v) = q(v) for all v V
_
.
Example 139:
122 CHAPTER 9. ALGEBRAIC GEOMETRY
(a) Let V = F
n
. Then c
oord
(F
n
) = F[x
1
,..., x
n
].
(b) Consider F
2
, and let T = p(x, y), where p(x, y) = y. Then
V(p) =
_
(x, y) F
2
; y = 0
_
= (x, 0) ; x F
is just the natural embedding of F into F
2
. If q(x, y) and r(x, y) are any polynomials in
F[x, y], then
_
q
[
V
= r
[
V
_

_
q(v, w) = r(v, w) for all (v, w) V
_

_
q(x, 0) = r(x, 0) for all x F
_
.
(c) Let F = C. Let c = (c
1
, c
2
, . . . , c
n
) be a single point in C
n
, and consider the singleton set
V = c. Then V = V(p
1
, p
2
, . . . , p
n
), where, for any x = (x
1
, . . . , x
n
) in C
n
,
p
1
(x) = (x
1
c
1
); p
2
(x) = (x
2
c
2
); . . . p
n
(x) = (x
n
c
n
).
If q(x, y) and r(x, y) are any polynomials in C[x, y], then
_
q
[
V
= r
[
V
_

_
q(c) = r(c)
_

_
q(c) r(c) = 0
_

_
(q r) /
c
_
,
where /
c
= f C[x
1
,..., x
n
] ; f(c) = 0 is the maximal ideal from Hilberts
Nullstellensatz (p. 149). In other words,
_
q
[
V
= r
[
V
_

_
q and r belong to the same coset of /
c
_
.
The annihilator of V is the ideal
/
nn
(V) = f F[x
1
,..., x
n
] ; f(v) = 0 for all v V. (see 10.7.4)
For example, if V = c as in Example 139c), then /
nn
(V) = /
c
. Example 139c) then
generalizes as follows:
Proposition 140 Let V F
n
be an algebraic variety, and let /
nn
(V) be its annihilator.
There is a natural ring isomorphism c
oord
(V)

=
F[x
1
,..., x
n
]
/
nn
(V)
, given by the map:
c
oord
(V) p
[
V

_
p +/
nn
(V)
_

F[x
1
,..., x
n
]
/
nn
(V)
.
9.3. MORPHISMS 123
Proof: Dene : F[x
1
,..., x
n
]c
oord
(V) by (p) = p
[
V
. Clearly, this is a surjective ring
homomorphism. Thus, the Fundamental Isomorphism Theorem (Theorem ?? on page ??)
says that c
oord
(V)

=
F[x
1
,..., x
n
]
ker()
. It therefore suces to show that ker() = /
nn
(V). To see
this, observe that:
_
p ker()
_

_
(p) = 0
_

_
p
[
V
= 0
_

_
For all v V, p(v) = 0)
_

_
p /
nn
(V)
_
. 2
Recall that a ring 1 is perfect if 1 has no nilpotent elements ie.

0
R
= 0
R
(see
10.7.3).
Corollary 141 Let 1 be a quotient ring of C[x
1
,..., x
n
]. Then
_
1

= c
oord
(V) for some algebraic variety V C
n
_

_
1 is perfect.
_
Proof: Suppose 1 = C[x
1
,..., x
n
]/1 for some ideal 1 C[x
1
,..., x
n
]. Let V = V(1) be the
variety induced by 1. Then
_
1 is a perfect ring
_
Lem.219
_
1 is a radical ideal
_
(Nlstz)
_
1 = /
nn
(V)
_

_
1

=
C[x
1
,...,xn]
A
nn
(V)
_
Prop.140
_
1

= c
oord
(V)
_
.
Here, (Nlstz) is the Radical Nullstellensatz (Theorem 17 on page 174). 2
Corollary 142 There is a natural bijective correspondence:
_
Algebraic varieties in C
n
_

_
Perfect quotient rings of C[x
1
,..., x
n
]
_
.
9.3 Morphisms
Prerequisites: 9.2 Recommended: 10.4.1,10.7.4
If V F
n
is an algebraic variety, then a (geometric) morphism from V to F
m
is a
function : VF
m
given by
(v) =
_

1
(v),
2
(v), . . . ,
m
(v)
_
,
where
1
,
2
, . . . ,
m
c
oord
(V).
124 CHAPTER 9. ALGEBRAIC GEOMETRY

Figure 9.2: The map (x, y) = (x

2
y
2
, 2xy, x) transforms a circle into a double-loop.

Figure 9.3: The map (x, y) = (x

2
, y
2
) transforms a circle into a diamond.
Example 143: Dene : S
1
R
3
by (x, y) = (x
2
y
2
, 2xy, x). This map transforms
the circle into a double-loop in three-dimensional space; see Figure 9.2.
If W F
m
is another algebraic variety, then a morphism from V to W is a morphism
: VF
m
such that (V) W.
Example 144: Let V = S
1
= W, and dene : S
1
S
1
by (x, y) = (x
2
y
2
, 2xy).
Then is a morphism which wraps the circle twice around itself (Exercise 116 Hint: write
x = cos() and y = sin().). This is the second Legendre polynomial.
Note that the image of variety under a morphism is not necessarily a variety:
Example 145: Let S
1
= (x, y) R
2
; x
2
+ y
2
= 1 be the unit circle, and dene :
S
1
R
2
by (x, y) = (x
2
, y
2
). Then the image (S
1
) is the diamond (x, y) R
2
; [x[ +[y[ = 1,
which is not an algebraic variety. See Figure 9.3. (Exercise 117 Check this.)
If : VF
m
is a morphism, let image () be the smallest algebraic variety in F
m
containing
the image (V). In other words:
image () =

WF
m
W a variety
(V)W
W.
9.3. MORPHISMS 125
(this is called the Zariski closure of (V)). It follows from Lemma 222 on page 174 that
image () = V
_
/
nn
((V))
_
.
We say that : VW is a (geometric) epimorphism if image (V) = W. Note that this
does not necessarily mean is surjective onto W.
Example 146: Let V = R = W, and let (x) = x
2
. Then (R) = r R ; r 0 is
the positive real line. Thus, the only algebraic variety in R which contains (R) is R itself.
Hence, image () = R, so is a geometric epimorphism from R to itself, even though is
not surjective.
Note that, even if the morphism is invertible, the inverse function
1
is not necessarily
a morphism:
Example 147: Let V = R = W, and dene : RR by (x) = x
3
. Thus, is bijective,
therefore invertible. However,
1
(x) = x
1/3
is not a geometric morphism (because it is not
a polynomial).
We say that : VW is an (geometric) monomorphism if:
1. : VW is injective, and
2. If U = (V) W, then the inverse map
1
: UV is the restriction of some
polynomial function. ie.
1
(u) =
_

1
(u),
2
(u), . . . ,
m
(u)
_
for all u U, where

1
,
2
, . . . ,
n
F[x
1
, ..., x
m
] are polynomial functions on F
m
.
We say that is a (geometric) isomorphism if:
1. : VW is bijective, and
2. The inverse map
1
: WVis also a morphism ie.
1
(w) =
_

1
(w),
2
(w), . . . ,
m
(w)
_
,
where
1
,
2
, . . . ,
n
c
oord
(W).
We then say that V and W are (geometrically) isomorphic.
Example 148:
(a) Let S
1
R
2
be the unit circle, as in Example 135)(a). Let E R
2
be the ellipse
with minor axis 2 and major axis 5, as in Example 135)(b). Dene : S
1
E by
(x, y) = (5x, 2y) (in other words,
1
(x, y) = 5x and
2
(x, y) = 2y). Then is a
geometric isomorphism, with inverse
1
(x, y) = (
x
5
,
y
2
).
126 CHAPTER 9. ALGEBRAIC GEOMETRY
(b) Let SO
2
[R] /
2
(R) be the special orthogonal group from Example 136)(e), and
let S
1
R
2
be the circle. Dene : S
1
SO
2
[R] by
(x, y) =
_
x y
y x
_
.
Then is a geometric isomorphism (Exercise 118).
Proposition 149 Let V and W be algebraic varieties, and let : VW be a morphism.
(a) If f : WF is a coordinate function, then f : VF is also a coordinate function.
(b) Dene : c
oord
(W) c
oord
(V) by (f) = f . Then:
1. is a ring homomorphism, and (11
Y
) = 11
X
.
2.
_
is a geometric monomorphism
_

_
is a ring epimorphism
_
.
3.
_
is a geometric epimorphism
_

_
is a ring monomorphism
_
.
4.
_
is a geometric isomorphism
_

_
is a ring isomorphism
_
.
(c) (Functorial Property) Suppose U, V and W are algebraic varieties, and that
: UV and : VW are morphisms. Let = : UW, so that diagram
(A) below commutes.
Dene homomorphisms : c
oord
(V) c
oord
(U), : c
oord
(W) c
oord
(V), and
: c
oord
(W) c
oord
(U) as in part (b). Then = . In other words, diagram (B)
commutes:
U V

(A)
=====
c
oord
(U) c
oord
(V)

c
oord
(W)

`
(B)
Proof: (a) Suppose V F
n
and W F
m
. The construction is illustrated by the following
diagram:
V W

(
f
)
=
f

f
F

9.3. MORPHISMS 127
By hypothesis, f is a coordinate function that is, f = p
[
W
, where p F[x
1
, ..., x
m
] is some
polynomial. Likewise, (v) =
_

1
(v), . . . ,
m
(v)
_
, where
1
,
2
, . . . ,
m
F[x
1
,..., x
n
] are
polynomials. Thus, f (v) = p
_

1
(v), . . . ,
m
(v)
_
. It is Exercise 119 to check that
p
_

1
(v), . . . ,
m
(v)
_
is a polynomial.
(b1) Exercise 120 .
(b2) = Suppose is a monomorphism. Let U = (V) W. Thus, the map
1
:
UV is well-dened, and
1
= q
[
U
for some polynomial function q : F
m
F
n
.
We want to show that is surjective; hence, given any f c
oord
(V), we want some g c
oord
(W)
so that (g) = f.
Suppose f = p
[
W
, where p F[x
1
, ..., x
m
]. Let g = (pq)
[
W
. Then g is a coordinate function
on W (because p q is a polynomial), and, for any v V, we have:
(g)(v) = g (v) = p q ((v))
()
p
1
((v)) = p(v)
()
f(v).
Here, () is because (v) U and q
[
U
=
1
; () is because p
[
V
= f.
(b2) = Suppose that : c
oord
(W) c
oord
(V) is surjective.
Claim 1: is injective.
Proof: Suppose not; then there exist points v, v

V so that (v) = w = (v

). Now, let
f c
oord
(V) be any polynomial such that f(v) ,= f(v

). For example, if v = (v
1
, . . . , v
n
)
and v

= (v

1
, . . . , v

n
), then v and v

must dier in some coordinate say v

1
,= v

1
. Then
let f(x
1
, . . . , x
n
) = x
1
v
1
. Then f(v) = 0 ,= f(v

).
I claim there is no function g c
oord
(W) such that (g) = f. To see this, observe that, for
any g c
oord
(W),
(g)(v) = g
_
(v)
_
= g(w) = g
_
(v

)
_
= (g)(v

).
Hence, (g)(v) = (g)(v

), whereas f(v) ,= f(v

). Thus, (g) cannot equal f.

Thus, f is not in the image of , contradicting surjectivity. By contradiction, must be
injective. ............................................................ 2 [Claim 1]
Now, let U = (V), and let
1
: UV be the inverse function.
Claim 2:
1
= q
[
U
for some polynomial function q : F
m
F
n
.
Proof: Let
1
: F
n
F be projection into the rst coordinate; ie.
1
(x
1
, . . . , x
n
) = x
1
.
This is clearly a polynomial, so (
1
)
[
V
is a coordinate function. Thus, since is surjective,
128 CHAPTER 9. ALGEBRAIC GEOMETRY
there is some element g
1
c
oord
(W) so that (g
1
) = (
1
)
[
V
. But g
1
is the restriction of
some polynomial q
1
F[x
1
, ..., x
m
]; hence, we have:
q
1
= g
1
= (
1
)
[
V
.
Likewise, if
k
: F
n
F is projection into the kth coordinate, then there is some polynomial
q
k
F[x
1
, ..., x
m
] so that
q
k
= (
k
)
[
V
. (9.1)
I claim
1
= (q
1
, . . . , q
n
). To see this, observe that, for any v = (v
1
, . . . , v
n
) V,
(q
1
, . . . , q
n
) (v) =
_
q
1
(v), . . . , q
n
(v)
_
by(9.1)
_

1
(v), . . . ,
n
(v)
_
= (v
1
, . . . , v
n
) = v.
In other words, (q
1
, . . . , q
n
) = Id on V. ............................ 2 [Claim 2]
(b3) = Suppose is an geometric epimorphism (ie. image () = W); we claim
is a ring monomorphism. To prove this, it suces to show that ker() = 0. So, suppose
g c
oord
(W) and (g) = 0; we want to show that g 0 in other words, that g(w) = 0 for
all w W.
Let V(g) = w W ; g(w) = 0 be the algebraic variety induced by g.
Claim 3: (V) V(g).
Proof: If w (V), then there is some v V so that (v) = w. But then g(w) =
g ((v)) = (g)(v) = 0 (because (g) = 0). This holds for all w (V), so (V)
V(g). ............................................................... 2 [Claim 3]
Thus, image () V(g). But by hypothesis, image () = W; hence V(g) = W, which means
g 0.
(b3) = Suppose was not a geometric epimorphism. We claim is not a monomor-
phism in other words, ker() ,= 0.
Since is not an epimorphism, it follows that U = image () is a proper subset of W.
Let /
nn
(W) =
_
p F[x
1
, ..., x
m
] ; p
[
W
0
_
be the annihilator of W. It follows:
Claim 4: /
nn
(W) /
nn
(U).
Proof: Exercise 121 Hint: Apply Lemmas 220 and 222 on page 174 ..... 2 [Claim 4]
So, let p F[x
1
, ..., x
m
] be a polynomial so that p /
nn
(U) but p , /
nn
(W). Hence, if
g = p
[
W
, then g is a coordinate function, and g ,= 0.
Claim 5: (g) = 0.
9.4. IRREDUCIBLE VARIETIES 129
Proof: Let v V, and let u = (v) U. Then (g)(v) = g (v) = g(u) = 0,
because g /
nn
(U). ................................................. 2 [Claim 5]
Hence, g ker() is nonzero, so cannot be a monomorphism.
(b4) Combine (b2) and (b3).
(c) Exercise 122 . 2
Corollary 150 Let V and W be algebraic varieties. Then
_
V and W are geometrically isomorphic
_

_
c
oord
(V) and c
oord
(W) are isomorphic as rings
_
.
Thus, all geometric information about a variety is encoded in its coordinate ring.
Recall that a ring 1 is perfect if 1 has no nilpotent elements ie.

0
R
= 0 (see
10.7.3). Recall from Corollary 142 on page 123 that every perfect quotient of C[x
1
,..., x
n
] is
the coordinate ring of some algebraic variety V C
n
. It follows:
Corollary 151 There is a natural bijective correspondence:
_
(Geometric) Isomorphism classes
of algebraic varieties in C
n
_

_
(Algebraic) Isomorphism classes of
perfect quotient rings of C[x
1
,..., x
n
]
_
.
So, instead of dening complex algebraic geometry as the study of complex algebraic vari-
eties, we could dene it as the study of the study of perfect quotients of complex polynomial
rings, with a geometric interpretation.
9.4 Irreducible Varieties
Prerequisites: 9.1, 10.5
Let F be a eld, and let V F
n
be an algebraic variety. We say V is irreducible if V
cannot be written as a union of two smaller varieties. In other words, there are no varieties U
and W so that
V = U W.
Example 152:
(a) Let c C
n
be a point. Then the singleton variety V = c is irreducible.
130 CHAPTER 9. ALGEBRAIC GEOMETRY
V
U W
Figure 9.4: V = U W
(b) As in Example 135)(f ), let V = V(p), where p(x, y) = x
2
y
2
= (x +y) (x y). Then
V = U W, where
U = V(x + y) = (x, y) R
2
; x = y,
and W = V(x y) = (x, y) R
2
; x = y. (See Figure 9.4)
Hence, V is not irreducible.
(c) More generally, suppose that p(x
1
, . . . , x
n
) F[x
1
,..., x
n
] is a polynomial which factors:
p(x
1
, . . . , x
n
) = q(x
1
, . . . , x
n
) r(x
1
, . . . , x
n
)
Then V(p) = V(q) V(r), so if V(q) and V(r) are both nonempty, then V(p) is not
irreducible.
It appears from Examples 152b) and 152c) that irreducibility of a variety is related to
irreducibility of the polynomials which generate it. The precise formulation of this is as follows:
Proposition 153 The following are equivalent:
(a) V is irreducible.
(b) If U and W are varieties such that V U W, then either V U or V W.
(c) /
nn
(V) is a prime ideal in F[x
1
,..., x
n
].
(d) c
oord
(V) is an integral domain.
In particular, if V = V(p) for some p F[x
1
,..., x
n
], then
_
V is irreducible
_

_
p is irreducible
_
.
9.4. IRREDUCIBLE VARIETIES 131
Proof: (a)=(b) Suppose V UW. Recall that the intersection of two varieties is a
variety. Thus, U

= VU is a variety, and so is W

= VW. Not that V = U

. But
V is irreducible, so either U

= V or W

= V; in other words, either V U or V W.

(b)=(a) Suppose V = U W; then V U W, so (b) implies that either V U
or V W. Suppose that V U. But V = U W, so clearly U V. Hence, U = V.
(Likewise, if V W, then V = W.)
We conclude that V cannot be written as a union of two subvarieties; hence it is irreducible.
(b)=(c) Suppose f, g F[x
1
,..., x
n
] are polynomials such that f g /
nn
(V). I claim
that either f /
nn
(V) or g /
nn
(V). To see this observe that
_
(f g) /
nn
(V)
_

_
V V(f g) = V(f) V(g).
_
=(b)
_
Either V V(f) or V V(g).
_

_
Either f /
nn
(V) or g /
nn
(V)
_
.
(c)=(a) Suppose V is not irreducible; well show that /
nn
(V) cannot be prime.
Suppose V = U W, but V ,= U and V ,= W. Thus, U V, so Lemma 220 on page 173
implies that /
nn
(V) /
nn
(U). So, nd f /
nn
(U) with f , /
nn
(V).
Likewise W V, so /
nn
(V) /
nn
(W), so nd g /
nn
(U) with g , /
nn
(V). I claim that
(f g) /
nn
(V). To see this, let v V. Then either v U (in which case f(v) = 0) or
v W (in which case g(v) = 0). Either way, (f g)(v) = 0.
Thus, f , /
nn
(V) and g , /
nn
(V), but (f g) /
nn
(V). Hence, /
nn
(V) is not a prime ideal.
(c) (d) Exercise 123 Use Proposition 200 on page 165 and Proposition 140 on
page 122. 2
Corollary 151 on page 129 immediately implies:
Corollary 154 There is a natural bijective correspondence:
_
Isomorphism classes of irreducible
algebraic varieties in C
n
_

_
Isomorphism classes of integral
domain quotients of C[x
1
,..., x
n
]
_
.
Irreducible varieties are the basic building blocks out of which other varieties are made:
Proposition 155 Any algebraic variety is a union of irreducible varieties.
To be precise: let V F
n
be an algebraic variety with annihilator /
nn
(V). Consider the
set of all prime ideals which contain /
nn
(V):
P = T F[x
1
,..., x
n
] ; T prime, and /
nn
(V) T.
For any T P, the corresponding variety V(T) is irreducible, and V =
_
PP
V(T).
132 CHAPTER 9. ALGEBRAIC GEOMETRY
Proof: Exercise 124 2
Because of this, algebraic geometers normally restrict their attention to irreducible varieties.
Indeed, in many texts, (eg. [3]), an algebraic variety is dened to be irreducible. .
Chapter 10
Ideal Theory
10.1 Principal Ideals and PIDs
Prerequisites: ??
Let 1 be a commutative ring and let p 1. The principal ideal generated by p is the set
p1 = pr ; r 1. (see Example 117k)) (10.1)
We normally denote this ideal by (p).
Example 156:
(a) If 1 = Z, then (2) = 2Z = 2z ; z Z, the ideal of even numbers from Example 117a).
(b) If 1 = R[x], then (x) = xR[x] = x q(x) ; q 1[x], the ideal from Example 117e).
If 1 is a noncommutative ring, then in general, p1 is not an ideal. Instead, we dene the
principal ideal generated by p to be
1p1 = r
1
ps
1
+ r
2
ps
2
+ . . . + r
n
ps
n
; n N, r
1
, . . . , r
n
, s
1
, . . . , s
n
1. (10.2)
(Exercise 125 Verify this is an ideal.)
Again, we denote this ideal by (p). Observe that, if 1 is commutative, then denition
(10.2) agrees with denition (10.1) (Exercise 126).
A principal ideal domain (PID) is a commutative ring where all ideals are principal
ideals.
Proposition 157 The ring of integers is a principal ideal domain. To be precise:
If 1 Z is any ideal, then 1 = nZ, where n is the minimal positive element in 1.
133
134 CHAPTER 10. IDEAL THEORY
Proof: Let 1 Z be an ideal, and let n be the smallest positive integer in 1. That is:
n = min(1 N).
Our goal is to show that 1 = nZ.
Claim 1: nZ 1.
Proof: Since n 1 and 1 is an ideal, it follows that nz 1 for all z Z. In other words,
any element of nZ is in 1. ........................................... 2 [Claim 1]
Claim 2: 1 nZ.
Proof: Suppose not. Find some i 1 with i , nZ. Without loss of generality, we assume
i > 0 (because if i 1, then i is also in 1). Since n is the minimal positive element
of 1, we must have n < i. Apply the Division algorithm to write i = qn + r, where
0 < r < n. By hypothesis, i , nZ, so n doesnt divide i, so we must have r ,= 0. But then:
i 1;
qn nZ 1 thus, qn 1.
r = i qn; thus, r 1.
But r < n, and this is a contradiction, because n is the minimal positive element in 1.
By contradiction, no such i can exist; hence 1 (n). .................. 2 [Claim 2]
Claims 1 and 2 together imply that 1 = nZ. 2
Principal Ideals in Polynomial Rings: The proof of Proposition 157 generalizes readily
to the ring R[x] of polynomials. The key concepts used in the proof of Proposition 157 were
1. The existence of a minimal element in 1.
2. The use of the Division Algorithm.
We must generalize these ideas to polynomials. Recall that the degree of a polynomial p(x) is
the highest exponent appearing in p(x) with nonzero coecient. For example degree (() 5x
3

7x
2
+8x+6) = 3. In general, If p(x) = p
n
x
n
+. . .+p
1
x
1
+p
0
, (where p
n
,= 0), then degree (p) = n.
Recall also the the Polynomial Long Division algorithm. Given any polynomials p(x)
and P(x), where degree (P) degree (p), we can divide P(x) by p(x) to get:
P(x)
p(x)
= q(x) +
r(x)
p(x)
. (10.3)
where degree (r) < degree (p). If we multiply both sides of equation (10.3) by p(x), we get:
P(x) = q(x) p(x) + r(x).
If the remainder polynomial r(x) equals zero, then we say that p(x) divides P(x).
10.1. PRINCIPAL IDEALS AND PIDS 135
Proposition 158 The ring R[x] is a principal ideal domain. To be precise:
If 1 R[x] is any ideal, then 1 = (p), where p is a polynomial of minimal degree in 1.
Proof: Let 1 R[x] be an ideal. Let p be a polynomial of minimal degree in 1. That is:
degree (p) = min degree (q) ; q 1.
Our goal is to show that 1 = (p).
Claim 1: (p) 1.
Proof: Since p 1 and 1 is an ideal, it follows that p q 1 for all q R[x]. Hence any
element of (p) is also in 1. ........................................... 2 [Claim 1]
Claim 2: 1 (p).
Proof: Suppose not. Find some i 1 with i , (p). Since p has minimal degree in 1,
we must have degree (p) degree (i). Now apply Polynomial Long Division to write
i = qp +r, where 0 < degree (r) < degree (p). By hypothesis, i , (p), so p doesnt divide i,
so we must have r ,= 0. But then:
i 1;
qp (p) 1; thus, qp 1.
r = i qp; thus, r 1.
But degree (r) < degree (p), and this is a contradiction, because p has minimal degree in 1.
By contradiction, no such i can exist; hence 1 (p). .................. 2 [Claim 2]
Claims 1 and 2 together imply that 1 = (p). 2
There is nothing special about the real numbers or about R[x]. The notion of polynomial
division generalizes to any eld, and yields the following theorem:
Proposition 159 Let F be any eld. Then F[x] is a principal ideal domain. To be precise:
If 1 F[x] is any ideal, then 1 = (p), where p is a polynomial of minimal degree in 1.
Proof: Exercise 127 Literally just change the symbol R to F throughout the preceeding
argument 2
Do not get the idea that every polynomial ring is a PID....
Example 160:
136 CHAPTER 10. IDEAL THEORY
(a) Let 1 = Z[x] (Example 98c)), and let 1 be the polynomial generated by the elements 2
and x. That is,
1 = 2p(x) + xq(x) ; p(x) Z[x] and q(x) Z[x].
We claim that 1 is a nonprincipal ideal.
Claim 1: 1 is an ideal.
Proof: Exercise 128 ............................................ 2 [Claim 1]
Claim 2: 1 contains the polynomials 2 and x.
Proof: 2 = 2 1 + x 0, and 2 = 2 0 + x 1. ...................... 2 [Claim 2]
Claim 3: 1 , = Z[x].
Proof: We will show that any element in 1 must have an even constant coecient.
Hence, 1 cannot be all of Z[x].
Suppose i(x) 1. Then i(x) = 2p(x) + xq(x) for some p(x), q(x) Z[x].
If p(x) = p
n
x
n
+ . . . + p
1
x + p
0
, then p(x) = 2p
n
x
n
+ . . . + 2p
1
x + 2p
0
.
If q(x) = q
m
x
m
+ . . . + q
1
x + q
0
, then x q(x) = q
m
x
m+1
+ . . . + q
1
x
2
+ q
0
x.
Hence, if i(x) = i

+ . . . + i
1
x + i
0
, then i
0
= 2p
0
is even. ....... 2 [Claim 3]
Claim 4: 1 is not a principal ideal.
Proof: Suppose 1 = (g) for some polynomial g(x) Z[x].
Claim 4.1: g must be a constant polynomial ie. an integer.
Proof: Claim 2 says 2 1, so g(x) must divide 2. In other words, 2 = g(x) q(x)
for some polynomial q(x). But then
0 = degree (2) = degree (p(x) q(x)) = degree (p(x)) + degree (q(x))
Hence, we must have degree (p(x)) = 0 = degree (q(x)). Hence g(x) must be a
constant integer, say g
0
. ........................................ 2 [Claim 4]
Now, since g
0
divides 2, we must have either g
0
= 1 or g
0
= 2.
Claim 4.2: g
0
,= 1.
Proof: Claim 3 says 1 , = Z[x]. Hence, 1 is not an element of 1, so g ,= 1.
2 [Claim 4.2]
Thus, g
0
= 2. But g
0
must divide x, and 2 does not divide x in the ring Z[x]. Hence,
no such g could exist. Thus, 1 is not a principal ideal. ............. 2 [Claim 4]
As a consequence: Z[x] is not a principal ideal domain.
10.2. MAXIMAL IDEALS 137
(b) Let 1 = R[x, y] (Example 98e)) and let 1 be the polynomial generated by the elements
x and y. That is,
1 = x p(x, y) + y q(x, y) ; p(x, y) R[x, y] and q(x, y) R[x, y].
We claim that 1 is a nonprincipal ideal.
Claim 1: 1 is an ideal.
Claim 2: 1 contains the polynomials x and y.
Claim 3: 1 , = R[x, y]. In particular, every polynomial of 1 has zero constant term.
The proofs of these claims are Exercise 129 .
Claim 4: 1 is not a principal ideal.
Proof: To see this, suppose 1 = (g) for some polynomial g(x, y) R[x, y]. Then g(x)
must divide x, so it must not contain any terms in y, or any powers of x greater than
x
1
. Hence, g(x) = g
1
x + g
0
for some g
1
, g
0
R. If g(x) is to divide x, then we must
have g
0
= 0. Hence, g(x, y) = g
1
x.
But g(x, y) must divide y, and x does not divide y in the ring R[x, y]. Hence, no such
g could exist. Thus, 1 is not a principal ideal. ..................... 2 [Claim 4]
As a consequence: R[x, y] is not a principal ideal domain.
Example 160b) generalizes to the following
Proposition 161 Let F be a eld, and let n 2. If F[x
1
, x
2
, . . . , x
n
] is the ring of polynomials
over F in n variables, then F[x
1
, x
2
, . . . , x
n
] is not a principal ideal domain.
Proof: Exercise 130 2
10.2 Maximal Ideals
Prerequisites: 8.6, 10.1
Let 1 be a ring and /1 an ideal. We say / is a maximal ideal if there exists no ideal
1 1 such that / 1 1 (see Figure 10.1)
Example 162: Let 1 = Z and consider the principal ideal / = 2Z (Example 117a)). I
claim this ideal is maximal. To see this, suppose there was 1 Z such that 2Z 1 Z.
Thus, 1 contains least one odd number o (since 1 , 2Z); say o = 2n + 1. But 1 also
contains all even numbers (since 2Z 1); in particular, 1 contains 2n. Thus, 1 contains
(2n + 1) (2n) = 1. Hence 1 = Z, by Lemma 120 on page 111.
Hence, any ideal properly containing 2Z must be all of Z. Hence, 2Z is maximal.
138 CHAPTER 10. IDEAL THEORY
R
M
1
M
2
M
3
M
4
M
5 Maximal
Figure 10.1: A schematic representation of the lattice of ideals of the ring 1. The maximal
ideals form the top row of the lattice, just below 1.
(2) (3) (5) (7) (11)
(4) (6) (15) (35) (77) (121)
Maximal
Figure 10.2: A schematic representation of the lattice of ideals of the ring Z. The maximal
ideals are the principal ideals (2), (3), (5), etc. generated by prime numbers.
10.2. MAXIMAL IDEALS 139
Example 162) generalizes as follows:
Proposition 163 The maximal ideals of Z are exactly the ideals pZ, where p is prime.
(see Figure 10.2.)
Proof: We want to show:
_
/ is a maximal ideal in Z
_

_
/ = pZ for prime p
_
.
Claim 1: Let p be prime, and let / = pZ. Then / is a maximal ideal.
Proof: Recall from Example 117b) that / is an ideal. We must show it is maximal.
To see this, suppose 1Z was an ideal such that pZ 1 Z. Proposition 157 on page 133
says that 1 = (i), where i is the minimal positive element in 1. But if / 1, that means
in particular that p 1 ie. p (i), so i must divide p. But p is prime, so....
either: i = p, in which case (i) = (p) ie. 1 = /.
or: i = 1, in which case (i) = Z ie. 1 = Z.
It follows that / is maximal. ........................................ 2 [Claim 1]
Claim 2: Let /Z be a maximal ideal. Then / = pZ for some prime p.
Proof: / is an ideal of Z, so Proposition 157 on page 133 says that / = pZ, where p is
the smallest positive element in /. Our goal is to show that p is prime.
Suppose p was not prime. Let q be a divisor of p, with 1 < q < p. Let p = q d.
Claim 2.1: / qZ.
Proof: Any element of / has the form pz for some z Z. But p = q d, so pz = q (dz)
is also an element of qZ. .......................................... 2 [Claim 2.1]
Claim 2.2: qZ , /.
Proof: Recall that 1 < q < p. But p is the smallest positive element in /; hence q
cannot be in /; hence qZ , /. ................................. 2 [Claim 2.2]
Thus, / qZ Z, so / cannot be maximal. Contradiction. 2[Claim 2 & Theorem]
10.2.1 Maximal ideals in polynomial rings
The maximal ideal structure of Z is closely mirrored by the maximal ideal structure of polyno-
mial rings....
Example 164: Let 1 = R[x] and consider the principal ideal / = (x) (Example 156b)).
I claim this ideal is maximal. To see this, suppose 1 R[x] such that (x) 1 R[x]. Thus,
1 contains a polynomial p with p , (x). Hence p(x) = p
n
x
n
+. . . +p
1
x+p
0
with p
0
,= 0. Let
q(x) = p
n
x
n1
+ . . . + p
2
x + p
1
.
140 CHAPTER 10. IDEAL THEORY
(x) (x-1) (x+1)
R
[x]
(x
2
+1) (x
2
+2)
(x
2
) (x
2
-x) (x
2
-1) (x
4
+3x
2
+2) (x
3
+x
2
+x+1) (x
4
+4x
2
+4)
Maximal
Figure 10.3: A schematic representation of the lattice of ideals of the ring R[x]. The maximal
ideals are the principal ideals (x), (x 1), (x
2
+1), etc. generated by irreducible polynomials.
Then x q(x) is in (x), and thus, in 1. But
x q(x) = p
n
x
n
+ . . . + p
2
x
2
+ p
1
x = p(x) p
0
.
Thus, 1 contains p(x) x q(x) = p
0
. Thus, 1 contains p
1
0
p
0
= 1. Thus, 1 = R[x], by
Lemma 120 on page 111.
Hence, any ideal properly containing (x) must be all of R[x]. So (x) is maximal.
Example 164) generalizes as follows. Recall that a polynomial p(x) T[x] is irreducible if
there exists no polynomials q(x) T[x] which divides p(x), except for q(x) = p(x) or q(x) = 1.
Proposition 165 Let T be a eld. The maximal ideals of T[x] are exactly the ideals (p),
where p is an irreducible polynomial (see Figure 10.3).
Proof: We want to show:
_
/ is a maximal ideal in T[x]
_

_
/ = (p) for irreducible p
_
.
Claim 1: Let p be an irreducible polynomial, and let / = (p). Then / is a maximal
ideal.
Proof: Recall from Example 156b) that / is an ideal. We must show it is maximal.
To see this, suppose 1 T[x] such that (p) 1 T[x]. Proposition 159 on page 135 says
that 1 = (i), where i is a polynomial of minimal degree in 1. But if / 1, that means
in particular that p 1 ie. p (i), so i must divide p. But p is irreducible, so...
either: i = p, in which case (i) = (p) ie. 1 = /.
or: i = 1, in which case (i) = T[x] ie. 1 = T[x].
It follows that / is maximal. ........................................ 2 [Claim 1]
Claim 2: If /T[x] is a maximal ideal, then / = (p) for some irreducible polynomial p.
10.2. MAXIMAL IDEALS 141
Proof: / is an ideal of T[x], so Proposition 159 on page 135 says that / = (p), where
p is an element in / of minimal degree. Our goal is to show that p is irreducible.
Suppose p was not irreducible. Let q be a divisor of p, with 0 degree (q) < degree (p).
Let p = q d.
Claim 2.1: / (q).
Proof: Any element of / has the form pf for some f T[x]. But p = q d, so
pz = q (df) is also an element of (q). ............................. 2 [Claim 2.1]
Claim 2.2: (q) , /.
Proof: Recall that 0 degree (q) < degree (p). But p has minimal degree in /; hence
q cannot be in /; hence (q) , /. ............................... 2 [Claim 2.2]
Thus, / (q) T[x], so / cannot be maximal. Contradiction. 2[Claim 2 & Theorem]
10.2.2 Maximal Ideals have Simple Quotients
Recall that a ring 1 is simple if it contains no nontrivial ideals (ie. the only ideals in 1 are
0 and 1 itself). For example, Z
/2
is a simple ring. But Z
/2
is just the quotient ring Z/2Z
(Example 128)), and as weve seen, 2Z is a maximal ideal. This exemplies a general principle:
If / is a maximal ideal, then 1// is a simple ring. To be precise:
Proposition 166 Let 1 be a ring and let 1 1 be an ideal. Then
_
1 is maximal
_

_
1/1 is simple
_
.
Proof: Let

1 = 1/1. We apply the Lattice Isomorphism Theorem (Theorem 134 on
page 116). Let < 1 be any subring with 1 < . Let

= /1, a subring of

1. Then the
Lattice Isomorphism Theorem says
_
is a proper ideal of 1, and 1
_

_

is a proper ideal of

1, and

, = 0
_
.
Hence,
_
1 is maximal
_

_
There are no proper ideals 1 with 1
_

_
There are no proper ideals



1 with , = 0
_

_
1/1 is simple
_
. 2
Recall that the only simple commutative rings are elds. Hence, we have the following:
142 CHAPTER 10. IDEAL THEORY
Corollary 167 Let 1 be a commutative ring and let 1 1 be an ideal. Then
_
1 is maximal
_

_
1/1 is a eld
_
.
An equivalent formulation of Corollary 167 is:
Corollary 168 Let 1 be a commutative ring and let T be a eld.
(a) If : 1T is an epimorphism, then ker() is a maximal ideal.
(b) All maximal ideals of 1 arise in this fashion. 2
Example 169:
(a) Let 1 = R[x], and consider the principal ideal (x) = x q(x) ; q 1[x] any polynomial.
We already saw (Example 164)) that (x) is a maximal ideal in R[x], but now we can
provide a second proof.
Recall the evaluation homomorphism
0
: R[x]R from Example 111g), dened by

0
(p
n
x
n
+ . . . + p
1
x + p
0
) = p
0
. In Example 113e), we saw that ker(
0
) = (x). But
image [
0
] = R is a eld, so Corollary 168 implies that (x) is a maximal ideal.
(b) Let 1 = c(R), and let r R be any xed number, and consider the ideal /
r
=
f c(R) ; f(r) = 0 from Example 117d). I claim /
r
is a maximal ideal in c(R).
To see this, let
r
: c(R)R be the evalutation map from Example 111e), dened by

r
(f) = f(r). In Example 113d), we saw that ker(
r
) = /
r
. But image [
r
] = R is a
eld, so Corollary 168 implies that /
r
is a maximal ideal.
Thus, every point in R denes a maximal ideal in c(R). We will see in 10.3 that this
correspondence goes in both directions.
(c) Let 1 = C[x], and let c C be any complex number, and consider the ideal /
c
=
f C[x] ; f(c) = 0. Observe that /
c
is the kernel of the evaluation homomorphism

c
: C[x]C dened by
c
(f) = f(c). Since the image of
c
is the eld C, it follows that
/
c
is maximal in C[x].
Corollary 167 also provides a powerful mechanism for constructing elds....
Proposition 170 Let T be a eld, and let p(x) T[x] be an irreducible polynomial. Then
the quotient ring T[x]/(p) is also a eld.
Proof: Proposition 165 says that the principal ideal (p) is maximal in T[x]. Hence, Corollary
167 says that the quotient ring T[x]/(p) is a eld. 2
To illustrate, well construct the complex numbers....
10.3. THE MAXIMAL SPECTRUM 143
Example 171: The eld C is isomorphic to the quotient ring R[x]/(x
2
+ 1).
Proof: x
2
+ 1 is irreducible because you cant factor x
2
+ 1 in R[x]. Thus, Proposition 170
says that c = R[x]/(x
2
+ 1) is a eld. We want to show that c is isomorphic to C.
For any polynomial p(x) in R[x], let p be the corresponding element in c. In particular, let
j = x.
Claim 1: j is a square root of 1.
Proof: Observe that j
2
+

1 = x
2
+

1 = x
2
+ 1 = 0. Hence, j
2
= 1. 2 [Claim 1]
Claim 2: Every element of c can be written in a unique way as r
1
+ r
2
j, where r
1
and r
2
are real numbers.
Proof: Any element of c has the form p(x) for some polynomial p R[x]. To illustrate,
well use the polynomial p(x) = 8x
8
+7x
7
+6x
6
+5x
5
+4x
4
+3x
3
+2x
2
+x +
1
2
. Observe
p(x) = 8x
8
+ 7x
7
+ 6x
6
+ 5x
5
+ 4x
4
+ 3x
3
+ 2x
2
+ x + 1/2
=

8x
8
+

7x
7
+

6x
6
+

5x
5
+

4x
4
+

3x
3
+

2x
2
+ x + 1/2
=

8j
8
+

7j
7
+

6j
6
+

5j
5
+

4j
4
+

3j
3
+

2j
2
+ j + 1/2
Clm.1

8(

1)
4
+

7(

1)
3
j +

6(

1)
3
+

5(

1)
2
j +

4(

1)
2
+

3(

1)j +

2(

1) + j + 1/2
=

8

7j

6 +

5j +

4

3j

2 +j + 1/2
= (

6 +

4

2 + 1/2) + (

7 +

5

3 +

1)j = 1/2

4j.
The same reduction will clearly work for any polynomial. .............. 2 [Claim 2]
Now, dene the map : Cc by (r
1
+ r
2
i) = r
1
+ r
2
j. It is left as Exercise 131 to
verify that this is a ring isomorphism. 2
10.3 The Maximal Spectrum
Prerequisites: 10.2
Alg.Geo.
Let 1 be a ring. The maximal spectrum of 1 is the set of all maximal ideals of 1:
Spec (1) = / ; /1 is a maximal ideal. (see Figure 10.4)
Example 172: Let 1 = Z. Then Proposition 163 on page 139 says that the maximal ideals
of Z are exactly the principal ideals pZ, where p is prime. In other words,
Spec (Z) = pZ ; p N a prime number
144 CHAPTER 10. IDEAL THEORY
R
MaxSpec(R)
Figure 10.4: A schematic representation of the maximal spectrum of the ring 1.
C[0,1]
M
r
r
0 1
[0,1]
MaxSpec(C[0,1])
Figure 10.5: A schematic representation of the maximal spectrum of the ring c[0, 1], showing
the bijective correspondence between points in [0, 1] and maximal ideals in c[0, 1].
On of the fundamental concepts of algebraic geometry is the following
Correspondence Principle: If X is a space, and 1 is a ring of functions over
X, then the maximal ideals of 1 correspond to the points of X. Hence, Spec (1) is
a sort of image of X.
Thus, the maximal spectrum has a natural geometric interpretation. The basic idea of
commutative ideal theory is to extend this interpretation to arbitrary rings; given any ring 1,
we interpret Spec (1) as a kind of abstract space, and 1 as a ring of functions on this space.
10.3.1 The Maximal Spectrum of a Continuous Function Ring
Prerequisites: 10.2 Recommended: A
Our rst illustration of the Correspondence Principle is the following:
Proposition 173 Let c[0, 1] be the ring of continuous functions from [0, 1] into R. Then:
10.3. THE MAXIMAL SPECTRUM 145
(a) For every r [0, 1], there is a maximal ideal /
r
= f c[0, 1] ; f(r) = 0, which
is the kernel of the evaluation map
r
: c[0, 1] f f(r) R.
(b) Every maximal ideal of c[0, 1] arises in this fashion.
(c) Thus, there is a natural bijective correspondence [0, 1] Spec
_
c[0, 1]
_
, dened by
the map r /
r
(see Figure 10.5).
Proof: We have already seen that /
r
is a maximal ideal in Example 169b) (there we were
in the ring c(R), but the argument for c[0, 1] is identical). It remains to show that the map
[0, 1] r /
r
Spec
_
c[0, 1]
_
is bijective.
Injective: Suppose r, s [0, 1] with r ,= s. We want to show that /
r
,= /
s
. To see
this, consider the continuous function f(x) = [x r[. Clearly, f c[0, 1], and f(r) = 0, so
f /
r
. On the other hand, f(s) ,= 0, so f , /
s
. Hence, /
r
,= /
s
.
Surjective: Let / c[0, 1] be some maximal ideal; we want to show that / = /
r
for
some r [0, 1]. To do this, for any f /, dene
V(f) = v [0, 1] ; f(v) = 0.
Claim 1: V(f) ,= , and is a closed subset of [0, 1].
Proof: Nonempty: Suppose V(f) = . This means that f(x) ,= 0 for all x [0, 1]. Thus,
f is a unit element of c[0, 1]. To see this, dene g(x) =
1
f(x)
. Then g c[0, 1] also, and
f g = 11.
But if f is a unit element, then / = c[0, 1], by Lemma 120 on page 111. By contradiction,
V(f) ,=
Closed: To see that V(f) is closed, suppose that v
1
, v
2
, . . . , was a sequence of points in
V(f). Thus, f(v
n
) = 0 for all n N. Suppose v = lim
n
v
n
. Since f is continuous, we have:
f(v) = lim
n
f(v
n
) = lim
n
0 = 0.
Hence, v V(f) also. ............................................... 2 [Claim 1]
Claim 2: If f
1
, f
2
, . . . , f
N
/, then
N

n=1
V(f
n
) ,= , and is a closed subset of [0, 1].
Proof: Dene F(x) = f
2
1
(x) + f
2
(x)
2
+ . . . f
n
(x)
2
. Then F is also in / (because / is a
subring), so Claim 1 says V(F) ,= . But V(F) =
N

n=1
V(f
n
) (Exercise 132).
To see that

N
n=1
V(f
n
) is closed, recall that the intersection of any number of closed
subsets is also closed. (Lemma 267 on page 224) ....................... 2 [Claim 2]
146 CHAPTER 10. IDEAL THEORY
Now, dene
V(/) = v [0, 1] ; f(v) = 0 for all f / =

fM
V(f) .
Claim 3: V(/) ,= .
Proof: The interval [0, 1] is compact, so we apply the Finite Intersection Property
(230). Now Claim 3 follow from Claim 2. ............................. 2 [Claim 3]
Now, let x V(/).
Claim 4: / /
x
.
Proof: If f /, then x V(f); hence f(x) = 0, hence f /
x
. ..... 2 [Claim 4]
But / is a maximal ideal, so we conclude that / = /
x
. 2
There is nothing special about the interval [0, 1]; the same result holds for any compact
1
subset X R
N
:
Proposition 174 Let n N, and let X R
n
be a compact subset. Let c(X) be the ring of
continuous functions from X into R. Then:
(a) For every x X, there is a maximal ideal /
x
= f c(X) ; f(x) = 0, which is
the kernel of the evaluation map
x
: c(X) f f(x) R.
(b) Every maximal ideal of c(X) arises in this fashion.
(c) Thus, there is a natural bijective correspondence X Spec
_
c(X)
_
, dened by the
map x /
x
.
Proof: Exercise 133 Generalize the proof of Proposition 173 2
It is important that X be compact. Proposition 174 is not true if we let X = R or R
n
.
Example 175: Let 1 = c(R), and let c
0
(R) be the ring of continuous functions with
compact support from Example 3k). Then c
0
(R) is an ideal within c(R) (see Example
117l)), and indeed, a maximal ideal (Exercise 134). However, there is no point r R such
that c
0
(R) = /
r
. Hence, the Correspondence Principle fails for c(R).
To see how this is related to compactness, note that, in a sense, c
0
(R) is the set of contin-
uous functions which vanish at innity. Metaphorically we could write: c
0
(R) = /

.
The Correspondence Principle fails because is not an element of R. The solution to
this problem is to compactify R by adding a point at . This is called the Stone-

Cech
compactication.
1
X is compact if X is both closed and bounded in R
n
. See Appendix A.4.
10.3. THE MAXIMAL SPECTRUM 147
We can generalize Proposition 174 much further:
Proposition 176 Let X be a compact metric space, and let c(X) be the ring of continuous
functions from X into R. Then:
(a) For every x X, there is a maximal ideal /
x
= f c(X) ; f(x) = 0, which is
the kernel of the evaluation map
x
: c(X) f f(x) R.
(b) Every maximal ideal of c(X) arises in this fashion.
(c) Thus, there is a natural bijective correspondence X Spec
_
c(X)
_
, dened by the
map x /
x
.
Proof: Exercise 135 Hint: Use the fact that metric spaces are regular. 2
The Correspondence Principle also holds if we restrict ourselves to dierentiable func-
tions...
Proposition 177 Let X R
n
be a compact subset of R
n
(or let X be a compact dieren-
tiable manifold). Let c be one of the following rings of functions from X into R:
c = c
k
(X) (for some k N) or c = c

(X), or c = c

(X).
Then:
(a) For every x X, there is a maximal ideal /
x
= f c ; f(x) = 0, which is the
kernel of the evaluation map
x
: c f f(x) R.
(b) Every maximal ideal of c arises in this fashion.
(c) Thus, there is a natural bijective correspondence X Spec (c), dened by the map
x /
x
.
Proof: Exercise 136 2
148 CHAPTER 10. IDEAL THEORY
10.3.2 The Maximal Spectrum of a Polynomial Ring
Most important for classical algebraic geometry is a Correspondence Principle for the ring
of polynomial functions. To make this work, however, we must pass to the ring of complex
polynomials....
Proposition 178 Let C[x] be the ring of complex polynomials. Then:
(a) For every c C, there is a maximal ideal /
c
= f C[x] ; f(c) = 0, which is the
kernel of the evaluation map
c
: C[x] f f(c) C.
(b) /
c
is the principal ideal generated by the linear polynomial (x) = x c. That is,
/
c
=
_
(x c) q(x) ; q(x) C[x]
_
.
(c) Every maximal ideal of C[x] arises in this fashion.
(d) Thus, there is a natural bijective correspondence C Spec
_
C[x]
_
, dened by the
map c /
c
.
The proof of Proposition 178 requires the following result, which we state without proof:
Fundamental Theorem of Algebra: Any polynomial in C[x] can be factored into linear
polynomials. In other words, if p(x) C[x], then there are complex constants c
1
, c
2
, . . . , c
n
C
(possibly not distinct) so that p(x) = (x c
1
) (x c
2
) (x c
n
). 2
(The proof of the Fundamental Theorem does not involve the theory of maximal ideals, so
there is no circularity here)
Proof of Proposition 178: (a) This was Example 169c).
(b) Exercise 137 .
(c) Let /C[x] be a maximal ideal. The ring C[x] is a principal ideal domain (Proposi-
tion 159 on page 135) so / is the principal ideal generated by some polynomial p(x) C[x].
Our goal is to show that p(x) = x c for some c C.
The Fundamental Theorem of Algebra says that p(x) is divisible by some linear polynomial
(x) = (x c). That is, p(x) = (x c) q(x) for some polynomial q(x). Thus, p is in the
principal ideal 1 generated by (x c). But then / 1. Since / is maximal, it follows
that / = 1 that is, / is the principal ideal generated by (x c). 2
The generalization of Proposition 178 to polynomials of n variables is one of the most
important theorems in algebraic geometry:
10.3. THE MAXIMAL SPECTRUM 149
Theorem 179 Hilberts Nullstellensatz (Complex Version)
Let C[x
1
, . . . , x
n
] be the ring of complex polynomials in n variables.
(a) For every c C
n
, there is a maximal ideal /
c
= f C[x
1
, . . . , x
n
] ; f(c) = 0,
which is the kernel of the evaluation map
c
: C[x
1
, . . . , x
n
] f f(c) C.
(b) /
c
is the ideal generated by the linear polynomials

1
(x) = x
1
c
1
;
2
(x) = x
2
c
2
; . . .
n
(x) = x
n
c
n
;
That is,
/
c
=
_
(x
1
c
1
) q
1
(x) + . . . + (x
n
c
n
) q
n
(x) ; q
1
(x), . . . , q
n
(x) C[x
1
, . . . , x
n
]
_
.
(c) Every maximal ideal of C[x
1
, . . . , x
n
] arises in this fashion.
(d) Thus, there is a natural bijective correspondence C
n
Spec
_
C[x
1
, . . . , x
n
]
_
, dened
by the map c /
c
.
Proof: (a) Exercise 138 .
(b) Our strategy is to perform a change of coordinates on C
n
, so that the point c =
(c
1
, . . . , c
n
) acts as the origin. It may therefore be helpful, when rst reading this argument,
to pretend that c = (0, 0, . . . , 0).
Claim 1: Let p(x) C[x
1
, . . . , x
n
] be any polynomial. Then p(x) can be written as a
polynomial in the variables (x
1
c
1
), (x
2
c
2
), . . . , (x
n
c
n
). In other words, there are
complex coecients q

, q
1
, . . . , q
n
, q
11
, . . . , q
nn
, q
111
, . . . , q
nnn
, etc. so that
p(x) = q

+ q
1
(x
1
c
1
) + . . . + q
n
(x
n
c
n
)
+ q
11
(x
1
c
1
)
2
+ q
12
(x
1
c
1
)(x
2
c
2
) + . . . + q
1n
(x
1
c
1
)(x
n
c
n
)
+ q
22
(x
2
c
2
)
2
+ q
23
(x
2
c
2
)(x
3
c
3
) + . . . + q
2n
(x
2
c
2
)(x
n
c
n
) + . . . +
+ q
nn
(x
n
c
n
)
2
+ q
111
(x
1
c
1
)
3
+ q
112
(x
1
c
1
)
2
(x
2
c
2
) + . . .
Here, q

= p(c).
Proof: Exercise 139 Hint: There are two approaches:
Calculus Approach: Compute the Taylor series of the function p around the point c =
(c
1
, . . . , c
n
). (The advantage of this approach is that it is simple.)
Purely Algebraic Approach: Expand out the expression on the right hand side and collect
like terms, to yield a system of linear equations for the coecients q
1
, q
2
, . . . in terms of the
coecients of p. Now solve the resulting system. (The advantage of this approach is that it will
work for any eld, not just C.) ......................................... 2 [Claim 1]
150 CHAPTER 10. IDEAL THEORY
Note that every term on the right hand side except for the constant q

is divisible by at
least one of the polynomials (x
1
c
1
), . . . , (x
n
c
n
). Thus, every term except q

is in the
ideal generated by (x
1
c
1
), . . . , (x
n
c
n
). It follows:
_
p /
c
_

_
p(c) = 0
_

_
q

= 0
_

_
p is in the ideal generated by (x
1
c
1
), . . . , (x
n
c
n
)
_
.
Hence, /
c
is the ideal generated by (x
1
c
1
), . . . , (x
n
c
n
).
(c) Let / C[x
1
, . . . , x
n
] be a maximal ideal, and let T = C[x
1
, . . . , x
n
]// be the corre-
sponding quotient eld (Corollary 167). Let : C[x
1
, . . . , x
n
]T be the quotient map.
Let C[x
1
] be the ring of polynomials containing only the x
1
variable; thus, C[x
1
] is a subring
of C[x
1
, . . . , x
n
]. Let
1
: C[x
1
]T be the restriction of to C[x
1
].
Claim 2: ker(
1
) /.
Proof: By construction, / = ker(). Since
1
is just the restriction of to the smaller
domain C[x
1
], it follows that ker(
1
) ker(). ........................ 2 [Claim 2]
We want to show that / contains a linear polynomial of the form (x) = (x
1
c
1
). By
Claim 2, it will be sucient to show that ker(
1
) contains such a polynomial.
Claim 3: ker(
1
) ,= 0.
Proof: Recall the eld of rational functions:
C(x
1
) =
_
p(x
1
)
q(x
1
)
; p, q C[x
1
]
_
(Example 98h) on page 84)
Claim 3.1: If ker(
1
) = 0, then we can extend
1
to a monomorphism
1
: C(x
1
)T.
Thus, T contains an isomorphic copy of the eld C(x
1
).
Proof: Suppose p C[x
1
] has nonzero image ie.
1
(p) ,= 0 in T. Since T is a eld,
this means that
1
(p) has an inverse, which we will denote by
1

1
(p)
.
If ker(
1
) = 0, then
1
(p) is invertible for any nonzero polynomial p C[x
1
]. Thus,
then we can extend
1
to a map
1
: C(x
1
)T as follows:

1
_
p
q
_
=

1
(p)

2
(p)
, for any p, q C[x
1
].
Exercise 140 Check that
1
is well-dened, and a ring monomorphism. 2 [Claim 3.1]
We will now show that the conclusion of Claim 3.1 is impossible, because C(x
1
) is too big
to t inside of T. To do this, it will be helpful to treat C(x
1
) and T as complex vector
spaces.
10.3. THE MAXIMAL SPECTRUM 151
Recall that C[x
1
, x
2
, . . . , x
n
] is a vector space over the eld C of complex numbers. In other
words, if p and q are two polynomials, then p + q is also a polynomial, and if c C is any
scalar, then c p is also a polynomial.
Claim 3.2: C[x
1
, x
2
, . . . , x
n
] is a countable-dimensional complex vector space.
Proof: Consider the monomials x
k
1
1
x
k
2
2
. . . x
kn
n
, where k
1
, . . . , k
n
N. Clearly, any
element of C[x
1
, . . . , x
n
] can be written as C-linear combination of monomials of this
form. Thus, the (countable) set
_
x
k
1
1
x
k
2
2
. . . x
kn
n
; k
1
, . . . , k
n
N
_
spans C[x
1
, . . . , x
n
]
as a complex vector space. ........................................ 2 [Claim 3.2]
It follows that T is also a complex vector space, and : C[x
1
, . . . , x
n
]T is a surjective
C-linear map. To be precise: If p, q C[x
1
, . . . , x
n
], let p and q denote the corresponding
elements of T; then p + q = p + q, and, if if c C is any scalar, then c p = c p.
Claim 3.3: T is a countable-dimensional complex vector space.
Proof: Since T is the image of C[x
1
, . . . , x
n
], it follows that dim(T) dim
_
C[x
1
, . . . , x
n
]
_
.
To be more concrete, observe that the (countable) set
_
x
k
1
1
x
k
2
2
. . . x
kn
n
; k
1
, . . . , k
n
N
_
spans T as a complex vector space. ............................... 2 [Claim 3.3]
Observe that C(x
1
) can also be treated as a complex vector space (the sum of two rational
functions is rational, etc.).
Claim 3.4: C(x
1
) is a C-vector space of uncountable dimension.
Proof: For any c C, consider the rational function
c
(x) =
1
x c
.
Observe that
c
(x) is nite everywhere except at x = c, and
c
(c) = . It follows that,
if c
1
, . . . , c
n
are distinct complex numbers, then any nite C-linear combination
a
1

c
1
+ a
2

c
2
+ . . . + a
n

c
2
(10.4)
(where a
1
, . . . , a
n
C are nonzero complex numbers) is equal to innity on the set
c
1
, . . . , c
n
. In particular, the linear combination (10.4) can never be zero. In other
words, the set of rational functions
c
; c C is linearly independent in C(x). But C
is uncountable, so the set
c
; c C is also uncountable. Hence, C(x) has uncountable
dimension. ....................................................... 2 [Claim 3.4]
Thus, Claims 3.1, 3.2, and 3.3 together amount to an embedding of an uncountable di-
mensional C-vector space within a countable-dimensional space. This is impossible. By
contradiction, ker(
1
) cannot be 0. ................................. 2 [Claim 3]
Claim 4: ker(
1
) contains a linear polynomial
1
(x
1
) = (x
1
c
1
) for some c
1
C.
Proof: Claim 3 says ker(
1
) ,= 0, so let k(x
1
) ker(
1
) be some nonzero polynomial. The
Fundamental Theorem of Algebra (p. 148) says that there are constants a
1
, a
2
, . . . , a
m

C so that
k(x
1
) = (x
1
a
1
) (x
1
a
2
) (x
1
a
m
).
152 CHAPTER 10. IDEAL THEORY
Hence,
0 =
1
(k) =
1
(x
1
a
1
)
1
(x
1
a
2
)
1
(x
1
a
m
).
Now, T is a eld, so it has no zero divisors. So at least one of the factors
1
(x
1
a
1
),

1
(x
1
a
2
), . . . ,
1
(x
1
a
m
) must be zero. By reordering, we can assume
1
(x
1
a
1
) = 0.
In other words, the polynomial (x
1
) = (x
1
a
1
) is in ker(
1
). so, set c
1
= a
1
, and were
done. ............................................................... 2 [Claim 4]
It follows from Claims 2 and 4 that:
Claim 5: / contains a linear polynomial
1
(x) = (x
1
c
1
) for some c
1
C.
Now, we can repeat the argument of Claim 5 for x
2
, . . . , x
n
to conclude:
/ contains linear polynomials
1
(x) = (x
1
c
1
),
2
(x) = (x
2
c
2
), . . . ,
n
(x) = (x
n
c
n
),
for some c
1
, . . . , c
n
C.
Thus, / contains the ideal generated by
1
, . . . ,
n
. But from (b), we know that the ideal
generated by
1
, . . . ,
n
is the maximal ideal /
c
(where c = (c
1
, . . . , c
n
)). Thus / contains
/
c
, which means / must equal /
c
. 2
Further Reading: A good, elementary introduction to the Nullstellensatz and its relation
to algebraic geometry is [1, 10.7]. More advanced discussions of the maximal spectrum can be
found in [4, 7.5] or [6, II.5]. A thorough development of the subject is [7]. For applications
to algebraic geometry, see [8, 3]. For applications to dierential geometry, see [9]. The closely
related spectral theory of abelian Banach algebras is discussed in [2, VII.8]
10.3.3 The Maximal Spectrum of a Coordinate Ring
Prerequisites: 8.6, 9.2, 10.3.2
Recall that the Nullstellensatz
2
says there is a bijective correspondence between the
points in C
n
and the maximal ideals of C[x
1
,..., x
n
] given by
C
n
c /
c
Spec
_
C[x
1
,..., x
n
]
_
(where /
c
= p C[x
1
,..., x
n
] ; p(c) = 0.)
We will now extend this to any complex algebraic variety. If V C
n
is an algebraic variety,
then we dene
Spec
V
(C[x
1
,..., x
n
]) = /
v
; v V Spec
_
C[x
1
,..., x
n
]
_
.
2
p. 149
10.3. THE MAXIMAL SPECTRUM 153
Lemma 180 Let V C
n
, with annihilator /
nn
(V). Then Spec
V
(C[x
1
,..., x
n
]) is just the
set of maximal ideals in C[x
1
,..., x
n
] which contain /
nn
(V). Formally:
Spec
V
(C[x
1
,..., x
n
]) =
_
/ Spec
_
C[x
1
,..., x
n
]
_
; /
nn
(V) /
_
.
Proof: Exercise 141 2
Corollary 181 Nullstellensatz for Complex Algebraic Varieties
Let V C
n
be an algebraic variety, with coordinate ring c
oord
(V). Then there is a natural
bijection between the points in V and the maximal ideals c
oord
(V), given by
C
n
c /
c
Spec
_
c
oord
(V)
_
,
where /
c
= f c
oord
(V) ; f(c) = 0.
Proof: Let /
nn
(V) be the annihilator of V; Recall that Proposition 140 on page 122 says
that
c
oord
(V)

=
C[x
1
,..., x
n
]
/
nn
(V)
.
Hence, there is a natural bijection Spec
_
c
oord
(V)
_
Spec
_
C[x
1
,..., x
n
]
/
nn
(V)
_
.
The Lattice Isomorphism Theorem (p. 116) yields an order-preserving bijection:
_
ideals of C[x
1
,..., x
n
] containing /
nn
(V)
_
1 1
_
ideals of
C[x
1
,..., x
n
]
/
nn
(V)
_
.
It follows that
_
1 is maximal in C[x
1
,..., x
n
]
_

_
1 is maximal in
C[x
1
,..., x
n
]
/
nn
(V)
_
(because the map is order-preserving). Hence, we get a bijection:
_
/ Spec
_
C[x
1
,..., x
n
]
_
; /
nn
(V) /
_
/ / Spec
_
C[x
1
,..., x
n
]
/
nn
(V)
_
But Lemma 180 says
_
/ Spec
_
C[x
1
,..., x
n
]
_
; /
nn
(V) /
_
is just Spec
V
(C[x
1
,..., x
n
]).
Finally, there is a natural bijection between V and Spec
V
(C[x
1
,..., x
n
]), given by v /
v
.
2
154 CHAPTER 10. IDEAL THEORY
10.4 The Zariski Topology
Prerequisites: 10.3
Let Xbe some kind of space (eg. X = [0, 1] or X = C
n
), and let 1be some ring of functions
on X (eg. 1 = c[0, 1], or 1 = C[x
1
, . . . , x
n
]). The Correspondence Principle (page 144)
sets up a natural bijection between the maximal ideals in 1 and the points in X. We will now
see how we can turn this bijection into a homeomorphism, by endowing the maximal spectrum
of 1 with a natural topological structure called the Zariski topology.
10.4.1 The Zariski Topology (Continuous Function Rings)
Prerequisites: 10.3.1 Recommended: A
Throughout this section, let X be one of the following:
1. X = [0, 1].
2. X R
n
any compact subset.
3. X any compact metric space.
(whichever youre most comfortable with).
Let c(X) be the ring of continuous functions from X into R. Recall Propositions 173 (for
X = [0, 1]), 174 (for X R
n
) or 176 (for X a metric space), all of which say:
There is a natural bijective correspondence between the points in X and the maxi-
mal ideals in Spec
_
c(X)
_
, given by x /
x
, where /
x
= f c(X) ; f(x) = 0.
We will now see how this bijection also preserves the topological structure of X.
If C X is any closed set, then we dene a corresponding subset of Spec
_
c(X)
_
as follows:
Zar (C) = /
c
; c C.
We call Zar (C) the Zariski closed set corresponding to C.
The Zariski topology on Spec
_
c(X)
_
is the topology where a subset C Spec
_
c(X)
_
is considered closed if and only if C is a Zariski closed set. It follows:
Corollary 182 Endow Spec
_
c(X)
_
with the Zariski topology. Then the map X x
/
x
Spec
_
c(X)
_
is a homeomorphism. 2
10.4. THE ZARISKI TOPOLOGY 155
This wouldnt be very interesting if the Zariski topology on Spec
_
c(X)
_
was merely ob-
tained by importing the topology of X in the obvious way. However, well now construct the
Zariski topology using only the intrinsic algebraic structure of c(X).
If o c(X) is any subset, then the envelope of o is the set of all maximal ideals in c(X)
which contain o:
Env (o) =
_
/ Spec
_
c(X)
_
; o /
_
.
Let C Spec
_
c(X)
_
be some collection of maximal ideals. We call C an envelope if C =
Env (o) for some subset o c(X).
Proposition 183 The envelope subsets of Spec
_
c(X)
_
are exactly the Zariski closed sets.
To be specic:
(a) If o c(X) is any subset, then Env (o) = Zar
_
V(o)
_
, where
V(o) = x X ; f(x) = 0 for all f o.
is the variety of o (see 10.7.2). Hence, every envelope is a Zariski closed set.
(b) Conversely, if C X is a closed set, and Zar (C) is the corresponding Zariski closed
set, then Zar (C) = Env
_
/
nn
(C)
_
, where
/
nn
(C) = f c(X) ; f(c) = 0 for all c C.
is the annihilator of C (see 10.7.2). Hence, every Zariski closed set is an envelope.
Proof: Exercise 142 2
Thus, Zariski topology on Spec
_
c(X)
_
is a completely algebraic construction. In other
words, if you just encountered 1 = c(X) as an abstract ring, and you had no idea that it was
the ring of continuous functions on some space, you could still construct the Zariski topology,
based on purely algebraic information. Thus, the topology of X is completely encoded by the
algebraic structure of c(X).
This encoding extends to morphisms: we will see that continuous functions between spaces
X and Y correspond to ring homomorphisms between c(X) and c(Y).
Proposition 184 Suppose X and Y are compact metric spaces
3
, and let : XY be a
continuous map.
3
For example, suppose X = [0, 1] = Y.
156 CHAPTER 10. IDEAL THEORY
(a) If f : YR is continuous, then f : XR is also continuous.
Thus, if f c(Y), then f c(X).
(b) Dene : c(Y)c(X) by (f) = f . Then:
1. is a ring homomorphism, and (11
Y
) = 11
X
.
2.
_
is injective
_

_
is surjective
_
.
3.
_
is surjective
_

_
is injective
_
.
4.
_
is a homeomorphism
_

_
is a ring isomorphism
_
.
(c) Suppose X, Y and Z are compact metric spaces, and : XY and : YZ are
continuous functions. Let = : XZ, so that diagram (A) below commutes.
Dene homomorphisms : c(Y)c(X), : c(Z)c(Y), and : c(Z)c(X) as
in part (b). Then = . In other words, diagram (B) commutes:
X Y

(A)
=====
c(X) c(Y)

c(Z)

`
(B)
Proof: Exercise 143 Hint: See the proof of Proposition 149 on page 126 2
The map described in Proposition 184(b) is called the pullback of , and is usually
writted as

. Thus, in part (c), =

, =

, and =

, and part (c) could be

reformulated:
( )

.
This correspondence also goes the other way:
Proposition 185 Suppose X and Y are compact metric spaces
4
, and let : c(Y)c(X)
be a ring homomorphism, such that (11
Y
) = 11
X
.
(a) If /c(X) is any maximal ideal in c(X), then
1
(/) is a maximal ideal in c(Y).
(b) Thus, we can dene a map

: Spec
_
c(X)
_
Spec
_
c(Y)
_
by

(/) =
1
(/).
4
For example, suppose X = [0, 1] = Y.
10.4. THE ZARISKI TOPOLOGY 157
1.

is a continuous function from Spec
_
c(X)
_
to Spec
_
c(Y)
_
.
2.
_
is a ring monomorphism
_

_

is surjective
_
.
3.
_
is a ring epimorphism
_

_

is injective
_
.
4.
_
is a ring isomorphism
_

_

is a homeomorphism
_
.
(c) Recall that the map X x /
x
Spec
_
c(X)
_
is a homeomorphism. Dene
: XY by (x) = y, where y Y is the unique point such that /
y
=

(/
x
). In
other words, dene so that the following diagram commutes:

X x (x) Y


Spec
_
c(X)
_
/
x

1
(/
x
) Spec
_
c(Y)
_

Then:
1. is well-dened, and a continuous function from X to Y.
2.
_
is a ring monomorphism
_

_
is surjective
_
.
3.
_
is a ring epimorphism
_

_
is injective
_
.
4.
_
is a ring isomorphism
_

_
is a homeomorphism
_
.
5. For any f c(Y), (f) = f . In other words, =

is the pullback of
, as described in Proposition 184.
(d) Suppose X, Y and Z are compact metric spaces, and that : c(Y)c(X) and
: c(Z)c(Y) are ring homomorphisms. Let = : c(Z)c(X), so that
diagram (A) below commutes.
Dene : XY, : YZ, and : XZ as in part (c). Then = . In
other words, diagram (B) commutes:
c(X) c(Y)

c(Z)

`
(A)
=====
X Y

(B)
158 CHAPTER 10. IDEAL THEORY
Proof: Exercise 144 2
The map described in Proposition 185(c) is called the push forward of , and is usually
writted as

. Thus, in part (d), =

, =

, and =

, and part (d) could be

reformulated:
( )

.
Part (c5) of Proposition 185 says that the operations of pulling back and pushing forward
are inverse to each other. In other words:
For any continous function : XY, and any ring homomorphism : c(Y)c(X),
_
=

_

_
=

_
10.4.2 () The Zariski Topology (for an arbitrary Ring)
Prerequisites: 10.3, A Recommended: 10.4.1
Let 1 be a ring. If o 1 is any subset, then the Zariski envelope of o is the set of all
maximal ideals in 1 which contain o:
Env (o) =
_
/ Spec (1) ; o /
_
Example 186:
(a) If 1 = Z and o = 5, then Env (o) = 5Z. Also, if o = 5, 10, 35, then Env (o) =
5Z. Finally, if o = 5Z, then Env (o) = 5Z.
(b) If 1 = Z and o = 12, then Env (o) = 2Z, 3Z. If o = 12, 24, 72, then Env (o) =
2Z, 3Z. If o = 12Z, then Env (o) = 2Z, 3Z.
(c) In general, let 1 = Z, and suppose n N has prime factorization: n = p

1
1
p

2
2
p

k
k
. If
o = n, or o = nZ, then Env (o) =
_
(p
1
), (p
2
), . . . , (p
n
)
_
. (Exercise 145)
(d) Let 1 = R[x], and let p(x) = x
2
3x+2 = (x2)(x1). If o = p(x), then Env (o) =
/
1
, /
2
, where /
1
= q(x) C[x] ; q(1) = 0 and /
2
= q(x) C[x] ; q(2) = 0.
(e) Let 1 = C[x], and suppose p(x) C[x] factors into a product of linear polynomials:
p(x) = (x c
1
)
k
1
(x c
2
)
k
2
(x c
n
)
kn
,
for some c
1
, . . . , c
n
C. If o = p, then Env (o) =
_
/
c
1
, /
c
2
, . . . , /
cn
_
, where
/
c
= q(x) C[x] ; q(c) = 0. (Exercise 146)
10.4. THE ZARISKI TOPOLOGY 159
(f) If /1 is a maximal ideal, then Env (/) = /.
(g) Env (1) = (because no maximal ideal can contain the whole ring).
(h) Env (0) = Spec (1) (because every maximal ideal contains 0).
Proposition 187 Let 1 be any ring. Then
(a) The map Env () is inclusion reversing. In other words, if o T , then Env (T )
Env (o).
(b) If o 1 and T 1, then Env (o T ) = Env (o) Env (T ).
(c) If o
1
, o
2
, . . . is any collection of subsets of 1, then Env
_

_
n=1
o
n
_
=

n=1
Env (o
n
).
Proof: Exercise 147 2
Let C Spec (1) be some collection of maximal ideals. We call C a Zariski subset if
C = Env (o) for some subset o 1. It follows from Proposition 187:
Corollary 188
(a) and Spec (1) are Zariski subsets of Spec (1).
(b) If C
1
and C
2
are Zariski sets, then so is C
1
C
2
.
(c) The intersection of any number of Zariski subsets is a Zariski subset. 2
In other words, Corollary 188 says that the collection of Zariski subsets obeys all the ax-
ioms required for the closed sets in a topological space (page 226). Hence, we dene Zariski
topology on Spec (1) by the condition:
A subset C Spec (1) is considered closed if and only if C is a Zariski set.
In 10.4.1, we saw that a ring homomomorphism from c(Y) to c(X) induced a continuous
map from X to Y, and vice versa. (Propositions 184 and 185). Since the Correspondence
Principle identies X with the maximal spectrum of c(X), we could also interpret this as
a continuous map from Spec
_
c(X)
_
to Spec
_
c(Y)
_
. The same principle holds for arbitrary
rings...
Proposition 189 Let : 1o be a ring homomorphism, such that (1
R
) = 1
S
.
(a) If /o is any maximal ideal in o, then
1
(/) is a maximal ideal in 1.
160 CHAPTER 10. IDEAL THEORY
(b) Dene : Spec (o) Spec (1) by (/) =
1
(/). Then:
1. is a continuous function from Spec (o) to Spec (1).
2.
_
is a ring monomorphism
_
=
_
is surjective
_
.
3.
_
is a ring epimorphism
_
=
_
is injective
_
.
4.
_
is a ring isomorphism
_
=
_
is a homeomorphism
_
.
(c) Suppose : 1o and : oT are ring homomorphisms. Let = : 1T ,
so that diagram (A) below commutes.
Dene : Spec (o) Spec (1), : Spec (T ) Spec (o), and : Spec (T ) Spec (1)
as above. Then = . In other words, diagram (B) commutes:
1 o

(A)
=====
Spec (1) Spec (o)

Spec (T )

(B)
Proof: Exercise 148 2
The map described in Proposition 189(b) is called the push forward of , and is usually
writted as

. Thus, in part (c), =

, =

, and =

, and part (c) could be

reformulated:
( )

10.4.3 () The Zariski Topology (on a eld)

Prerequisites: 10.4.2, 10.3.2 Recommended: 10.4.1
In 10.4.1, we saw that that the Zariski topology on the ring of continuous functions c[0, 1]
completely encodes the topology of [0, 1]. What if we instead consider a ring of polynomial
functions?
Let T be a eld (eg. either T = R or T = C), and consider the ring T[x] of polynomials
over T. If T T[x] is any subset of T[x], then the (algebraic) variety dened by T is the
subset of T dened:
V(T) = x T ; p(x) = 0 for all p T.
Example 190:
10.4. THE ZARISKI TOPOLOGY 161
(a) Suppose T = R and p(x) = x
2
3x + 2 = (x 2)(x 1). If T = p(x), then
V(T) = 1, 2.
(b) More generally, suppose T = p(x) is a singleton set, where p(x) T[x] is a product of
linear factors:
p(x) = (x c
1
)
k
1
(x c
2
)
k
2
(x c
n
)
kn
,
Then V(p) = c
1
, c
2
, . . . , c
n
is just the set of roots of p(x).
The Zariski topology on T is the topology where a set is consider closed if and only if
it is an algebraic variety. That is:
For any c T,
_
c is closed
_

_
c = V(o) for some subset o T[x]
_
.
Suppose that T = C; recall that Proposition 178 says:
There is a natural bijective correspondence between the points in C and the max-
imal ideals in Spec
_
C[x]
_
, given by c /
c
, where /
c
= f C[x] ; f(c) = 0.
Well now see that this bijection is actually a homeomorphism between Zariski topologies:
Proposition 191
(a) If o C[x] is any subset, then Env (o) = /
c
; c V(o).
(b) Endow C and Spec
_
C[x]
_
with their respective Zariski topologies. Then the map
C c /
c
Spec
_
C[x]
_
is a homeomorphism.
Proof: Exercise 149 2
The Zariski topology on R or C is much dierent from the familiar Euclidean topology:
Proposition 192 The Zariski topology on T is just the conite topology. That is, for any
c T,
_
c is (Zariski) closed
_

_
c is nite
_
.
To be precise:
(a) For any o C[x], V(o) is a nite subset of C.
(b) Conversely, every nite subset of C arises in this fashion. Given any nite sub-
set c
1
, c
2
, . . . , c
n
C, let p(x) = (x c
1
) (x c
2
) (x c
n
). Then V(p) =
c
1
, c
2
, . . . , c
n
.
162 CHAPTER 10. IDEAL THEORY
Proof: Exercise 150 2
Thus, the Zariski topology has far fewer closed or open sets than the Euclidean topology.
This has two consequences:
It is easier for a sequence to converge in the Zariski topology. In other words, if the
sequence x
1
, x
2
, . . . converges to x in the Euclidean topology, then it automatically
converges in the Zariski topology. However, the converse is not true.
It is harder for a function to be continuous in the Zariski topology. If f : TT, then
_
f is continuous
_

_
either f is constant, or f is everywhere nite-to-one.
_
In particular, any polynomial function is continous.
Exercise 151 Verify these statements.
10.4.4 () The Zariski Topology (on ane n-space)
Prerequisites: 10.4.2, 10.3.2 Recommended: 10.4.1, 10.4.3
Suppose X R
n
was some compact subset. In 10.4.1, we saw that that the Zariski topology
on the ring of continuous functions c(X) completely encodes the topology of X. What if we
instead consider a ring of polynomial functions?
Let T be a eld (eg. either T = R or T = C), and consider the ring T[x
1
, . . . , x
n
] of
polynomials in n variables over T. If T T[x
1
, . . . , x
n
] be any subset of T[x
1
, . . . , x
n
], then
the (algebraic) variety dened by T is the subset of T
n
dened:
V(T) = x T
n
; p(x) = 0 for all p T
Example 193: Suppose T = R, and consider R[x, y]. Suppose T = p(x, y) is a singleton
set, where p(x, y) = x
2
+ y
2
1. Then
V(p) =
_
(x, y) R
2
; x
2
+ y
2
= 1
_
is just the circle of radius 1 around zero.
The Zariski topology on T
n
is the topology where a set is consider closed if and only if
it is an algebraic variety. That is:
For any c T
n
,
_
c is closed
_

_
c = V(o) for some subset o T[x
1
, . . . , x
n
]
_
.
Suppose that T = C; recall that Hilberts Nullstellensatz ( Proposition 179) says:
10.4. THE ZARISKI TOPOLOGY 163
There is a natural bijection between the points in C
n
and the maximal ideals in
Spec
_
C[x
1
, . . . , x
n
]
_
, given by c /
c
, where /
c
= f C[x
1
, . . . , x
n
] ; f(c) = 0.
Well now see that this bijection is actually a homeomorphism between Zariski topologies:
Proposition 194
(a) If o C[x
1
, . . . , x
n
] is any subset, then Env (o) = /
c
; c V(o).
(b) Endow C
n
and Spec
_
C[x
1
, . . . , x
n
]
_
with their respective Zariski topologies. Then
the map C
n
c /
c
Spec
_
C[x
1
, . . . , x
n
]
_
is a homeomorphism.
Proof: Exercise 152 2
10.4.5 () More about the Zariski Topology
Prerequisites: 10.4.2, 10.7.1 Recommended: 10.4.1
Recall, when constructing the Zariski topology in 10.4.1, we specically worked with com-
pact spaces. Indeed, we saw in 10.3.1 that the Correspondence Principle actually fails if X
is not compact (Example 175). One reason for this is that the Zariski topology itself is always
compact. Hence, if X was a noncompact space, we could hardly expect a homemomorphism
between X and Spec (c(X))...
Proposition 195 Let 1 be any ring. Then Spec (1) is compact in the Zariski topology. 2
If X and Y are compact metric spaces, then Propositions 184 and 185 showed that
_
X and Y are homeomorphic
_

_
c(X) and c(Y) are isomorphic as rings
_

_
Spec (c(X)) and Spec (c(Y)) are homeomorphic
_
Thus, the maximal spectrum is a complete invariant of the ring c(X). It is natural to wonder
whether a similar conclusion holds for the maximal spectrum of arbitrary rings. Unfortunately,
it does not.
Proposition 196 Let 1 be a ring, and let = 0
R
be its Jacobson radical (p. 167).
Then Spec (1) is homeomorphic to Spec (1/).
To be precise: let : 11/ be the quotient homomorphism, and let

: Spec (1/) Spec (1) be its push-forward

5
. Then

is a homeomorphism. 2
5
See Proposition 189 on page 159(b).
164 CHAPTER 10. IDEAL THEORY
Nevertheless, we can relate algebraic properties of 1to topological properties of its maximal
spectrum.
Proposition 197 Suppose 1 = 1
1
1
2
. . . 1
n
. Then Spec (1) is a disconnected union
of n components:
Spec (1) = Spec (1
1
) . Spec (1
2
) . . . . . Spec (1
n
) .
Conversely, if Spec (1) is connected, then 1 cannot be decomposed as a direct sum of two
rings. 2
Further Reading: The Zariski topology is discussed in [4, 7.5]; a thorough development is
[7]. For applications to algebraic geometry, see [8, 3].
10.5 Prime Ideals
Prerequisites: 10.2
Recall from 8.3 on page 108 that the set 2Z of even numbers is an ideal in the ring Z.
This ideal has an interesting additional property. For any integers n, m Z
_
n m is even
_

_
either n is even or m is even
_
.
In other words, the product of two odd numbers cannot be even. We say that 2Z is a prime
ideal.
Let 1 be a ring and let T 1 be an ideal. We say T is a prime ideal if, for any r, s 1,
_
r s is in T
_

_
either r T or s T
_
.
Equivalently, T is prime if its complement 1 T is multiplicatively closed. That is:
for any r, s 1,
_
r , T and s , T
_
=
_
r s , T
_
.
Prime ideals get their name from their prototypical example: principal ideals of the integers
generated by prime numbers.
Example 198: Let p N be prime. Then the principal ideal
6
pZ = pz ; z Z is a prime
ideal in Z. To see this, suppose n, m Z. Then:
_
n m pZ
_

_
p divides n m
_

_
either p divides n or p divides m
_

_
either n pZ or m pZ
_
.
6
See Example 117b) on page 108.
10.5. PRIME IDEALS 165
These are not the only prime ideals in Z. The zero ideal 0 is also prime:
Example 199: Let T be any integral domain. Then the zero ideal 0 is a prime ideal. To
see this, recall that, if d
1
, d
2
T, then
_
d
1
d
2
= 0
_

_
d
1
= 0 or d
2
= 0
_
.
Indeed, Example 199) is prototypical....
Proposition 200 Let 1 be a commutative ring and let 1 1 be an ideal. Then
_
1 is a prime ideal
_

_
1/1 is an integral domain
_
.
Proof: Let

1 = 1/1. For any elements r and s in 1, let the corresponding elements of

1
be r and s. Then
_
r 1
_

_
r =

0
_
;
_
s 1
_

_
s =

0
_
;
and
_
r s 1
_

_
r s =

0
_
.
Hence,
_
1 is a prime ideal
_

_
If r s 1, then r 1 or s 1
_

_
If r s =

0, then r =

0 or s =

0
_

_
1/1 has no zero divisors.
_

_
1/1 is an integral domain.
_
2
Corollary 201 Let 1 be a commutative ring and let 1 1 be an ideal. Then
_
1 is a maximal ideal
_
=
_
1 is a prime ideal
_
.
Proof:
_
1 is a maximal ideal
_
=(C167)
_
1/1 is a eld.
_
=(X109c)
_
1/1 is an integral domain.
_
=(C200)
_
1 is prime.
_
Here, (C167) is by Corollary 167 on page 142; (X109c) is by Example 109c) on page 101,
and (C200) is by Corollary 200 above. 2
Example 202: (Not all prime ideals are maximal.)
a) The zero ideal is prime in any integral domain (Example 199), but it is not maximal.
166 CHAPTER 10. IDEAL THEORY
b) Let 1 = Z[x] (Example 98c)). Then the principal ideal (x) is prime, because Z[x]/(x)

=
Z is an integral domain. However, (x) is not maximal, because it is contained in the ideal
(2, x) = 2 p(x) + x q(x) ; p, q Z[x]
(Exercise 153)
An equivalent formulation of Proposition 200 is:
Corollary 203 Let 1 be a commutative ring and let T be an integral domain
(a) If : 1T is an epimorphism, then ker() is a prime ideal.
(b) All prime ideals of 1 arise in this fashion. 2
Corollary 204 Let : 1o be a ring homomorphism, and let T 1 be an ideal such
that ker() T. Then
_
T is a prime ideal in 1
_

_
(T) is a prime ideal in o
_
.
Proof: Exercise 154 Hint: Combine two applications of Proposition 200 with the Chain
Isomorphism Theorem (Theorem 133 on page 116). 2
Examples 198 and 199 exhaust the prime ideals of Z:
Proposition 205 The prime ideals of Z are exactly:
1. The zero ideal 0.
2. The principal ideals pZ, where p is prime.
Proof: Weve already seen that the zero ideal and prime principal ideals are prime. Now we
show there are no others.
Suppose 1 Z was a nonzero ideal. Proposition 157 on page 133 says that 1 = (i), where i
is the minimal positive element in 1. Thus, 1/1 = Z
/i
. But from Example 109f) on page
101, we know:
_
Z
/i
is an integral domain
_

_
Z
/i
is a eld
_
.
Hence, combining this with Proposition 200, we obtain:
_
(i) is prime
_

_
(i) is is maximal
_
But Proposition 163 (p. 139) says:
_
(i) is is maximal
_

_
i is prime.
_
2
10.6. THE PRIME SPECTRUM 167
10.6 The Prime Spectrum
Prerequisites: 10.5 Recommended: 10.3
Alg.Geo.
Let 1 be a ring. The prime spectrum of 1 is the set of all prime ideals of 1:
Spec

(1) = T ; T 1 is a prime ideal.

Example 206: Let 1 = Z. Then Proposition 205 on the facing page says that the prime
ideals of Z are the zero ideal, and the principal ideals pZ, where p is prime. In other words,
Spec

(Z) = pZ ; p N a prime number, or p = 0.

10.7 Radical, Variety, Annihilator
Let X be some kind of space (eg. X = [0, 1] or X = C
n
), and let 1 be some ring of functions
on X (eg. 1 = c[0, 1], or 1 = C[x
1
, . . . , x
n
]). In this section, we will investigate an interesting
duality between subspaces of X (eg. closed subsets or algebraic varieties), and certain ideals
of 1. A key concept linking the two is that of the radical, an algebraic device for closing an
ideal.
10.7.1 The Jacobson Radical
Prerequisites: 10.2 Recommended: 10.7.3
Let 1 be a ring, and let 1 1 be any subset (usually, 1 is an ideal). The Jacobson
radical of 1 in 1 is the intersection of all maximal ideals in 1 which contain 1:
1 =

IMR
M maximal
/.
Example 207:
(a) If /1 is a maximal ideal, then / = /.
(b) Let 1 = Z, and let 1 = (12). Then 1 = (3) (2) = (6).
(c) More generally, let 1 = Z, and let i Z. Suppose i has prime factorization: i =
p

1
1
p

2
2
p

k
k
. Let 1 = (i) be the principal ideal generated by i. Then
1 = (p
1
) (p
2
) . . . (p
n
) = (P), where P = p
1
p
2
p
n
. (Exercise 155)
(d) Let 1 = R[x], and let p(x) = x
3
4x
2
+ 5x 2 = (x 1)
2
(x 2). Let 1 = (p) be
the principal ideal generated by p(x). Let /
1
be the principal ideal (x 1), and /
2
be
the principal ideal (x 2). Then
1 = /
1
/
2
= (x
2
3x + 2).
Also, observe that /
1
= q(x) R[x] ; q(1) = 0 and /
2
= q(x) R[x] ; q(2) = 0.
168 CHAPTER 10. IDEAL THEORY
(e) More generally, suppose F is a eld and 1 = F[x]. Let p(x) F[x] be a product of linear
polynomials:
p(x) = (x c
1
)
k
1
(x c
2
)
k
2
(x c
n
)
kn
,
for some c
1
, . . . , c
n
F. Let 1 = (p) be the principal ideal generated by p(x). Then
1 = /
c
1
/
c
2
. . . /
cn
, where /
c
= (x c) = q(x) F[x] ; q(c) = 0 is
the principal ideal generated by x c. (Exercise 156)
We say that the ideal 1 is a Jacobson ideal if 1 = 1.
Example 208:
(a) Any maximal ideal is a Jacobson ideal.
(b) If = 1, then is a Jacobson ideal. (Exercise 157)
The Jacobson radical of the ring 1 itself is the Jacobson radical of the zero ideal in
other words, the intersection of all maximal ideals in 1:
0
R
=

MR
maximal
/.
Example 209:
(a) Let 1 = Z. Then 0
Z
= 0.
(b) Let 1 = Z
/12
. Then 0
Z
/12
=

0,

6.
(c) More generallly, let 1 = Z
/n
, where n has prime factorization: n = p

1
1
p

2
2
p

k
k
. Then
0
Z
/n
= m ; m = p

1
1
p

k
k
, where 0 <
1
<
1
, . . . , 0 <
k
<
k
.
(Exercise 158)
Lemma 210 Let 1 1, and let Q = 1/1 be the quotient ring, with quotient map :
1Q. Then 1 =
1
_
0
Q
_
.
Proof: Exercise 159 2
Further Reading: [4, 4.2].
10.7. RADICAL, VARIETY, ANNIHILATOR 169
10.7.2 Variety and Annihilator (Continuous Function Rings)
Prerequisites: 10.7.1 Recommended: 10.7.4
Throughout the following discussion, let I be one of the following (whichever you are most
comfortable with).
1. I = [0, 1].
2. I R
n
any compact subset.
3. I any compact metric space.
Consider the ring c(I) of continuous functions from I into R. For any subset X I, we
dene the annihilator of X to be the set
/
nn
(X) = f c(I) ; f(x) = 0 for all x X.
For example, if X = r is a singleton set, then
/
nn
(r) = f c(I) ; f(r) = 0 = /
r
,
where /
r
is the maximal ideal from Example 169b).
Lemma 211
(a) For any X I, /
nn
(X) is an ideal of c(I).
(b) The operation /
nn
() is inclusion-reversing. That is: If X Y, then /
nn
(Y)
/
nn
(X).
(c) For any X I and Y I, /
nn
(X Y) = /
nn
(X) /
nn
(Y).
Proof: Exercise 160 2
Let o c(I) be any subset. Recall that the variety of o is the subset V(o) =
i I ; f(i) = 0 for all f o.
Lemma 212
(a) For any o c(I), V(o) is a closed subset of I.
(b) The operation V() is inclusion-reversing. That is: If o T , then V(T ) V(o).
Proof: Exercise 161 2
170 CHAPTER 10. IDEAL THEORY
Thus, we have dened two operations:
/ :
_
subsets of I
_

_
ideals of c(I)
_
V :
_
subsets of c(I)
_

_
closed subsets of I
_
These operations are inverses of each other, in the following sense:
Lemma 213
(a) Let X I. Then V
_
/
nn
(X)
_
= X is the topological closure of X in I.
(b) In particular,
_
X is a closed subset
_

_
V
_
/
nn
(X)
_
= X
_
.
(c) Let 1 c(I). Then /
nn
_
V(1)
_
= 1 is the Jacobson radical of 1 in c(I).
(d) In particular,
_
1 is a Jacobson ideal
_

_
/
nn
_
V(1)
_
= 1
_
.
(e) Let C = closed subsets X I and J = Jacobson ideals 1 c(I). Then the
maps
/ : C J and V : J C
are bijections, and inverse to one another.
Proof: Exercise 162 2
10.7.3 The Nil Radical
Prerequisites: 10.5 Recommended: 10.7.1
Let 1 be a ring. An element r 1 is called nilpotent if r
n
= 0
R
for some n N.
Example 214:
(a) Let 1 = Z
/72
and let r =

6. Then r is nilpotent, because (

6)
3
= 216 =

0.
(b) More generally, let 1 = Z
/m
, and suppose that m has prime factorization m = p

1
1

p

2
2
p

k
k
. Let r = p
1
p
2
. . . p
k
in 1. Then r is nilpotent in Z
/m
. To see this, let
M = max
1
, . . . ,
k
. Then r
M
= p
M
1
p
M
k
=

0, because p
M
1
p
M
k
is divisible by m.
(c) Let 1 = /
2
(R), and let M =
_
0 0
1 0

. Then M is nilpotent in /
2
(R), because M
2
= 0.
The nilradical of 1 is the set of all nilpotent elements:

0
R
= r 1 ; r
n
= 0
R
for some n N.
10.7. RADICAL, VARIETY, ANNIHILATOR 171
Proposition 215 (Krull) Let 1 be a commutative ring.
(a)

0
R
is a prime ideal in 1.
(b)

0
R
is the intersection of all prime ideals in 1:

0
R
=

PR
P prime
T.
Proof: (a)

0
R
is closed under addition: Exercise 163 .

0
R
is an ideal: Suppose z

0
R
, and let r 1 be arbitrary. We claim that r z

0
R
also. To see this, nd n N such that z
n
= 0. Then (r z)
n
(c)
r
n
z
n
= r
n
0 = 0 (where (c)
is because 1 is commutative.)

0
R
is a prime ideal: Well prove this in the case when 1 is an integral domain (the
general proof is more complicated and involves Zorns lemma).
Suppose ab

0
R
. Then (ab)
n
= 0 for some n N. But (ab)
n
= a
n
b
n
, and 1 has no zero
divisors, so either a
n
= 0 or b
n
= 0; hence, either a

0
R
or b

0
R
.
(b) Since

0
R
itself is a prime ideal, it is clear that

PR
P prime
T

0
R
.
It suces to show that

0
R
T for every prime ideal T 1. To show this, suppose
z

0
R
. Then z
n
= 0 for some n N. But then z
n
T (because 0 T). Hence, since T
is prime, it follows that z T. 2
If 1 1 is any ideal, then the nilradical of 1 in 1 is dened:

1 = r 1 ; r
n
1 for some n N.
Example 216:
(a) If 1 = 0 is the zero ideal, then the nilradical of 1 is just the nilradical of 1.
(b) If T 1 is a prime ideal, then

T = T. (Exercise 164)
(c) Let 1 = Z, and let 1 = (12). Then

1 = (6) = (3) (2).

(d) Let 1 = Z, and suppose i N has prime factorization: i = p

1
1
p

2
2
p

k
k
. Let 1 = (i)
be the principal ideal generated by i. Then:

1 =
_
all numbers of the form p
j
1
1
p
j
2
2
p
jn
n
; for some j
1
, j
2
, . . . , j
n
> 0
_
= (P) where P = p
1
p
2
p
k
= (p
1
) (p
2
) . . . (p
k
) (Exercise 165)
172 CHAPTER 10. IDEAL THEORY
(e) Let 1 = C[x], and let 1 = (x
2
1) be the principal ideal generated by p(x) = x
2
1.
Then

1 = (x + 1) (x 1) = /
(1)
/
1
,
where /
1
= (x 1) = q(x) C[x] ; q(1) = 0, and /
(1)
= (x + 1) =
q(x) C[x] ; q(1) = 0.
(f) Let 1 = C[x], and suppose p(x) C[x] factors into a product of linear polynomials:
p(x) = (x c
1
)
k
1
(x c
2
)
k
2
(x c
n
)
kn
,
for some c
1
, . . . , c
n
C. Let 1 = (p) be the principal ideal generated by p(x). Then

1 = (x c
1
) (x c
2
) (x c
n
) = /
c
1
/
c
2
. . . /
cn
,
where /
c
= (x c
1
) = q(x) C[x] ; q(c) = 0. (Exercise 166)
Proposition 217 Let 1 be a commutative ring, and let 1 1.
(a)

1 is a prime ideal in 1.
(b)

1 is the intersection of all prime ideals in 1 which contain 1:

1 =

IPR
P prime
T.
(c) Let Q = 1/1 be the quotient ring, with quotient map : 1Q. Then

1 =
1
_

0
Q
_
.
Proof: Exercise 167 Hint: Prove part (c) rst; then prove parts (a) and (b) by combining
part (c) with Proposition 215, Corollary 204 on page 166 and the Lattice Isomorphism Theorem
(Theorem 134 on page 116) 2
We say that an ideal 1 is radical if

1 = 1.
Example 218:
(a) Any prime ideal is radical.
(b) If ^ =

1, then ^ is radical. (Exercise 168)

We say that the ring 1 is perfect if 1 has no nilpotent elements ie.

0
R
= 0. In
other words, 1 is perfect if the zero ideal 0
R
is radical.
Lemma 219 Let 1 1. Then:
_
1 is radical
_

_
1/1 is perfect
_
.
Proof: Exercise 169 2
10.7. RADICAL, VARIETY, ANNIHILATOR 173
Further Reading: [4, 7.1].
10.7.4 Variety and Annihilator (Polynomial Rings)
Prerequisites: 10.7.3 Recommended: 10.7.2
Consider the ring C[x
1
, . . . , x
n
] of polynomial functions on C
n
. For any subset X C
n
, we
dene the annihilator of X to be the set
/
nn
(X) = p C[x
1
, . . . , x
n
] ; p(x) = 0 for all x X.
For example, if X = c is a singleton set, then
/
nn
(c) = p C[x
1
, . . . , x
n
] ; p(c) = 0 = /
c
,
where /
c
is the maximal ideal from Example 169b).
Lemma 220
(a) For any X C
n
, /
nn
(X) is an ideal of C[x
1
, . . . , x
n
].
(b) The operation /
nn
() is inclusion-reversing. That is: If X Y, then /
nn
(Y)
/
nn
(X).
(c) For any X C
n
and Y C
n
, /
nn
(X Y) = /
nn
(X) /
nn
(Y).
Proof: Exercise 170 2
Let o C[x
1
, . . . , x
n
] be any subset. Recall that the variety of o is the subset V(o) =
c C
n
; p(c) = 0 for all p o.
Lemma 221 The operation V() is inclusion-reversing. That is: If o T , then V(T )
V(o).
Proof: Exercise 171 2
Thus, we have dened two operations:
/ :
_
subsets of C
n
_

_
ideals of C[x
1
, . . . , x
n
]
_
V :
_
subsets of C[x
1
, . . . , x
n
]
_

_
varieties of C
n
_
In a certain sense, the operation V is the inverse of /:
174 CHAPTER 10. IDEAL THEORY
Lemma 222
(a) Let X C
n
. Then V
_
/
nn
(X)
_
is the smallest algebraic variety in C
n
which contains
X (we call this the Zariski closure of X).
(b) In particular,
_
X is an algebraic variety
_

_
V
_
/
nn
(X)
_
= X
_
.
Proof: Exercise 172 2
It would be nice if we could also say that the operation / was the inverse of V. The precise
statement of this is:
Theorem 223 Hilberts Nullstellensatz (Radical Version)
7
Let 1 C[x
1
, . . . , x
n
]. Then
(a) /
nn
_
V(1)
_
=

1 is the nil radical of 1 in C[x

1
, . . . , x
n
].
(b) In particular,
_
1 is a radical ideal
_

_
/
nn
_
V(1)
_
= 1
_
. 2
Corollary 224 Let Z = algebraic varieties X C
n
and N = radical ideals 1
C[x
1
, . . . , x
n
]. Then the maps
/ : Z N and V : N Z
are bijections, and inverse to one another. 2
Further Reading: [4, 7.12] or [6, X.2]
7
This is actually the original version of the Nullstellensatz, so it is sometimes called the classical Nullstel-
lensatz.
Part III
Fields
175
Chapter 11
Field Theory
11.1 Compass & Straight-Edge Constructions I
In Mesopotamia, India, and especially Greece, ancient mathematicians perfected the art of
constructing precise geometric gures in the plane, using only a compass and straight-edge.
A compass is a length of string with a stylus at one end; using it, you can draw a circle of
any radius around any point in the plane. A straight-edge is simply a long, perfectly straight
object (eg. a tightly stretched string); using it, you can draw a straight line passing through
any chosen pair of points. In societies lacking computers or precision-engineered tools, these
were the only devices available for drafting and engineering.
We will provide three illustrations of these techniques: bisecting a line segment, bisecting
an angle, and constructing a regular hexagon.
Example A Bisecting a Line Segment
Consider a line segment x y between two points x and y in the plane (Figure 11.1A). The
construction proceeds as follows:
1. Construct a circle of radius r about x, as in Figure 11.1B. Here, r is any value greater
than half the distance from x to y.
2. Construct another circle with the same radius r, but centred at y, as in Figure 11.1C.
(A)
y x
(B)
y x
r
(C)
y x
r
r
(D)
y x
u
v
Figure 11.1: Bisecting a line segment.
177
178 CHAPTER 11. FIELD THEORY
(A) (B) (C) (D)

/2
x
y
x
y
x
y
p p p p
Figure 11.2: Bisecting an angle
(A) (B) (C) (D)
r r
r
(E)
x
y
x
C
Figure 11.3: Constructing a regular hexagon
3. Let u and v be the points where the two circles intersect, as in Figure 11.1D. Then the
line u v is perpendicular to x y, and intersects it exactly halfway between x and y.
Example B Bisecting an Angle
Consider an angle formed by two intersecting lines in the plane, as in Figure 11.2A. The
construction proceeds as follows:
1. Let p be the place where the two lines intersect. Construct a circle of any radius about
p, as in Figure 11.2B.
2. Let x and y be the places where this circle intersects the two lines forming . Form the
line segment x y, as in Figure 11.2C.
3. Using the previous construction, bisect the line x y. Draw a line from p to the midpoint
of x y, as in Figure 11.2D. The angle formed by this line is /2
Example C Constructing a Regular Hexagon
We will construct a regular hexagon whose sides all have length r.
1. Draw a circle C of radius r as in Figure 11.3A.
2. Let x be an arbitrary point on the circle, and draw another circle of radius r around x,
as in 11.3B.
3. Let y be a point where the new circle intersects C; draw another circle of radius r around
y, as in 11.3C.
11.1. COMPASS & STRAIGHT-EDGE CONSTRUCTIONS I 179
(A) (B)
1
1
1
1
2
2
2
2
3
3
3
1
Area = Area =

Figure 11.4: (A) Doubling the cube. (B) Squaring the circle
4. Proceeding in this fashion, draw four more circles of radius r, each centered at the place
where the last circle intersects C (Figure 11.3D)
5. We now have six equally spaced circles of radius r around C. Their centres are six equally
spaced points on C. Connect these points to get a regular hexagon, as in Figure 11.3E.
Four Outstanding problems: The Greeks developed a vast and sophisticated technology
of geometric constructions. However, for more than two thousand years, four problems have
remained unsolved:
I. Trisecting an angle: We saw how to bisect an angle in Example B. Is there a similar
construction to divide an angle into three equal parts?
II. Constructing arbitrary regular polygons: We saw how to construct a regular hexagon
in Example C. By using only three of the six vertices, we also obtain a regular (ie. equi-
lateral) triangle. By bisecting the angles of the hexagon (as in Example B), we can also
construct a dodecagon (12 sides), a 24-gon, etc.
By constructing perpendicular bisectors as in Example A, we can also construct a regular
square; by bisecting its angles, we get an octagon, a 16-gon, etc.
A fairly elaborate construction yields a regular pentagon. Angle bisection yields a regular
decagon (10 sides), a 20-gon, etc.
If you could trisect angles, then you could get a regular nonagon (9 sides) from an equilat-
eral triangle. Is there another way? Can you construct a heptagon (7 sides)? An 11-gon?
In general, is there a way to construct a regular N-gon for any N?
III. Doubling the Cube: If we build a cube with sides 1 metre long, then the interior of
cube has a volume of 1 cubic metre. Is it possible to build a cube whose volume is exactly
180 CHAPTER 11. FIELD THEORY
C
H
O
V
Figure 11.5: The starting point: an origin point O, a unit circle C, and lines H and V.
2 cubic metres, as in Figure 11.4A? That is: can you construct a cube whose sides have
length
3

2?
IV. Squaring the Circle: If we draw a circle with a radius of 1 metre, then its interior has
an area of square metres. Is it possible to build a square with exactly the same area,
as in Figure 11.4B? That is: can you construct a square whose sides have length

?
We will see that the answer to these four questions is No. This does not mean we dont
yet know how to do these constructions. It is a much more sweeping statement: we can prove
that it is simply impossible to perform these constructions using only a compass and straight-
edge. While this may a depressing and negative result, it has one important consequence: we
no longer need waste mental eort trying to solve these problems.
To show that problems I to IV are unsolvable, we will use eld theory. We will show that
the set of all possible lengths constructable using compass and straight-edge forms a eld, K.
We will then show that lengths such as
3

2 or

simply cannot be elements of K. To do this
will require understanding the theory of elds and eld extensions.
To begin with, lets more precisely dene what a compass and straightedge (C&S) construc-
tion consists of. We begin with an unmarked Euclidean plane. As shown in Figure 11.5, We
set down a single point O, which we consider the origin, and we decide on a unit of length
measurement by drawing a circle C around O, and dening this to be the unit circle. We then
draw a single line H through O, which we dene to be horizontal. We can then construct the
perpendicular line V, which we dene as vertical. At this point, we can identify the Euclidean
plane with the set R
2
in the obvious way.
A C&S construction is any sequence of the following ve basic operations, illustrated in
Figure 11.6.
11.1. COMPASS & STRAIGHT-EDGE CONSTRUCTIONS I 181
(A)
(x
1
,y
1
)
(x
2
,y
2
)
(B)
(x
1
,y
1
)
(x
2
,y
2
)
(C)
L
1
L
2
(x,y)
(E)
L
C
(x
1
,y
1
)
(x
2
,y
2
)
(D)
C
1
C
2
(x
2
,y
2
)
(x
1
,y
1
)
Figure 11.6: The ve basic operations with compass and straight-edge.
(Figure 11.6A) Given two points (x
1
, y
1
) and (x
2
, y
2
) in R
2
, we can draw a unique line passing
through these points.
(Figure 11.6B) Given two points (x
1
, y
1
) and (x
2
, y
2
) in R
2
, we can draw a unique circle
centered at (x
1
, y
2
) and passing through (x
2
, y
2
).
(Figure 11.6C) Given two (nonparallel) lines L
1
and L
2
, we can nd the unique point where
they intersect.
(Figure 11.6D) Given two circles C
1
and C
2
, we can nd the unique one or two points where
they intersect.
(Figure 11.6E) Given a circle C and a line L, we can nd the unique one or two points
where they intersect.
We will say that a point, line, or circle is constructable if it can be obtained through some
sequence of these ve operations, starting from nothing more than the origin O, the unit circle
C, and the lines H and V.
A constructable length is any real number r which appears as the distance between two
constructable points x and y. Let K be the set of all constructable lengths; hence K R.
Proposition 225 K satises the following properties:
(a) K is a eld.
182 CHAPTER 11. FIELD THEORY
a b
a+b
(A)
a-b b
a
(B)
Figure 11.7: (A) Addition of lengths. (B) Subtraction of lengths.
(b) Q K.
(c) K is closed under square roots. That is: if k K, then

k is also in K.
Furthermore:
1. K is the smallest eld in R satisfying properties (A), (B), and (C). That is: all elements
of K can be obtained by starting with Q, and iteratively applying operations of addition,
subtraction, multiplication, division, and square root.
2. The set of all constructable points in the plane is exactly K
2
R
2
.
3. A line L is constructable if and only if L = (x, y) R
2
; ax + by = c, where a, b, c K.
4. A circle C is constructable if and only if C = (x, y) R
2
; (x a)
2
+ (y b)
2
= c,
where a, b, c K.
Proof: (A) We must show that K is closed under addition, subtraction, multiplication,
and division.
Closure under addition/subtraction: Let a, b K. To construct the length a + b, we
proceed as in Figure 11.7A: we construct parallel, adjoining line segments of lengths a and
b; the length of the combination is a +b. To get a b, we construct a line segment of length
a, and then remove a segment of length b, as in Figure 11.7B.
Closure under multiplication: Let a, b K. Construct a right-angle triangle whose base
has length 1 and whose height has length a as in Figure 11.8A. Now extend this to a similar
triangle whose base has length b, as in Figure 11.8B. Then the height of the new triangle is
a b.
Closure under division: Let a, b K. Construct a right-angle triangle whose base has
length b and whose height has length a as in Figure 11.9A. Now construct a similar triangle,
whose base has length 1, as in Figure 11.9B. Then the height of the new triangle is a/b.
Proof of (B) The set K contains the length 1 by denition. Since K is closed under
addition/subtraction, we then have Z K. Since K is closed under multiplication/division,
we get Q K.
11.1. COMPASS & STRAIGHT-EDGE CONSTRUCTIONS I 183
(A) (B)
a
1
a
1
b
ab
Figure 11.8: Multiplication of lengths.
(A) (B)
a/b
1
b
a
b
a
Figure 11.9: Division of lengths.
(A)
a
1
a+1
2
(B)
1
a
+
1
2
a-1
2
b
( )
b
2
=
a-1
2
a+1
2
( )
2 2
-
=
a
2
+2a+1
4
a
2
-2a+1
4
-
=
4a
4
= a
Pythagoras:
Figure 11.10: Computing the square root of a length
184 CHAPTER 11. FIELD THEORY
Proof of (C) Let a K. We want to construct the length

a. To do this, rst build a
circle of diameter a +1, as in Figure 11.10A. Thus, the radius of this circle is
a+1
2
. Now draw
a perpendicular to the diameter, as in Figure 11.10B. A simple computation shows that this
perpendicular has length b =

a.
Proof of 2: Suppose (x, y) R
2
was a constructable point. We can draw a perpendicular
line from H to (x, y); the length of this line is x, so x K. We can draw a perpendicular
line from V to (x, y); the length of this line is y, so y K.
Proof of 3: A line L is constructable if it is the unique line passing through two con-
structable points (x
1
, y
1
) and (x
2
, y
2
), as in Figure 11.6A. In this case, the coecients a, b, c
can be obtained by solving the following system of linear equations:
x
1
a + y
1
b + c = 0;
x
2
a + y
2
b + c = 0.
If (x
1
, y
1
) is directly over (x
2
, y
2
) (ie. the line L is parallel to the vertical V) then x
1
= x
2
,
and we set a = 1, b = 0, and c = x
1
= x
2
. Otherwise, set b = 1, to get the simpler system:
x
1
a c = y
1
;
x
2
a c = y
2
.
We then solve this system to obtain:
_
a
c
_
=
_
x
1
1
x
2
1
_
1
_
y
1
y
2
_
.
Since x
1
, x
2
, y
1
, y
2
and 1 are in K, and K is a eld, we conclude that a and c are also in K.
Proof of 4: A circle C is constructable if it is the unique circle centred at a constructable
point (x
1
, y
1
) and passing through another constructable point (x
2
, y
2
), as in Figure 11.6C. In
this case, set a = x
1
and b = y
1
, both of which are in K. Now let c =
_
(x
1
x
2
)
2
+ (y
1
y
2
)
2
.
Since K is closed under square roots, it follows that c K also.
Proof of 1: Let

K be the smallest eld in R satisfying (A), (B), and (C). Clearly,

K K;
well show K

K.
The arguments from the proofs of parts 2, 3, and 4 show:
If L is the line between two points in

K
2
, then L = (x, y) R
2
; ax + by = c, where
a, b, c

K. We say that L is a

K-line.
If (x, y) is the point of intersection of two

K-lines, then (x, y)

K
2
.
If C is the circle dened by two points in

K
2
, then C = (x, y) R
2
; (x a)
2
+ (y b)
2
= c,
where a, b, c

K. We say that C is a

K-circle.
It remains to show:
11.2. FIELD EXTENSIONS 185
1. If (x, y) is a point of intersection of two

K-circles then (x, y)

K
2
.
2. If (x, y) is a point of intersection of a

K-line and a

K-circle then (x, y)

K
2
.
It is Exercise 173 to show:
Claim 1: If (x, y) is a point of intersection of a

K-circle with a

K-line or another

K-circle,
then x satises a quadratic equation ax
2
+bx +c = 0, where a, b, c

K. Likewise, y satises
a quadratic equation dx
2
+ ex + f = 0, where d, e, f

K.
It follows that x =
b

b
2
4ac
2a
and y =
e

e
2
4df
2d
. Since

K is a eld closed under square
roots, it follows that x and y are also in

K. 2
To show the impossibility of constructions I to IV, well show certain lengths are not in K:
I. Trisecting an angle: Angle is trisectable i sin(/3) is in K. This is obviously the case
for some values of (eg. = /2), but well show it is not the case for most .
II. Constructing arbitrary regular polygons: A regular N-gon is constructable i sin(
2
N
)
is in K. Well show this is not true for some N.
III. Doubling the Cube: We can double a cube i
3

2 K. Well show this is false.

IV. Squaring the Circle: We can double a circle i

K. Well show this is false.
11.2 Field Extensions
Prerequisites: 7.5 Recommended: 11.1
Let F be a eld. An extension of F is another eld E such that F E.
Example 226:
(a) R is an extension of Q.
(b) C is an extension of R.
(c) Let K be the eld of constructable lengths from Proposition 225 in 11.1. Then K is an
extension of Q.
(d) Let R(x) be the eld of rational functions with real coecients (Example 98h) on page
84), and let C(x) be the eld of rational functions with complex coecients. Then C(x)
is an extension of R(x).
186 CHAPTER 11. FIELD THEORY
We consider the pair F and E as a single structure, called a eld extension. We will denote
this structure by (E F). Sometimes the notation E/F is used (read E over F). We will
also portray this eld extension diagramatically by:
E

F
11.2.1 Extensions as Vector Spaces:
It is natural to think of C as a two-dimensional real vector space. We often speak of the
complex plane (as opposed to the real line). Every element of c C can be written in a
unique way as c = r
1
+ r
2
i, so the set 1, i is a basis for C.
This idea generalizes to any eld extension. If E is a eld extension of F, then E is auto-
matically a vector space over F. That is:
1. E is an abelian group under the operation of addition:
e
1
+ e
2
= e
2
+ e
1
, for all e
1
, e
2
E
e
1
+ (e
2
+ e
3
) = (e
1
+ e
2
) + e
3
, for all e
1
, e
2
, e
3
E
e + 0 = e for all e E.
e + (e) = 0 for all e E.
2. F acts linearly on E by scalar multiplication:
For any f F and e E, the product f e is in E.
f (e
1
+ e
2
) = fe
1
+ fe
2
, for any f F and e
1
, e
2
E.
(f
1
f
2
) e = f
1
(f
2
e) for any f
1
, f
2
F and e E.
The degree of the extension (E F) is the dimension of E as an F-vector space. That is:
deg
_
E F
_
= min
_
[o[ ; where o E is an F-spanning set for E
_
= max
_
[1[ ; where 1 E is an F-linearly independent subset of E
_
= [B[, where B E is any F-basis for E.
Sometimes the notation [E : F] is used to indicate degree. We say E is a nite extension if
deg
_
E F
_
is nite, and an innite extension if deg
_
E F
_
is innite.
Example 227:
(a) C is a 2-dimensional vector space over R, with basis 1, i. Thus, deg
_
C R
_
= 2.
11.2. FIELD EXTENSIONS 187
(b) R is an innite-dimensional vector space over Q. One way to see this is to observe that
card
_
Q
N

= card [Q] =
0
for any nite N. In other words, any nite-dimensional rational
vector space is countable. But R is uncountable; hence its dimension (as a rational vector
space) must be innite (and indeed, uncountable). Thus, deg
_
R Q
_
= .
11.2.2 Simple Extensions
Let E F be any extension of F, and let E. We dene F() to be the smallest subeld of
E containing both and F. That is:
F() =

FLE
L
L.
We say F() is a simple extension of F.
Example 228:
(a) Q(

2) is the smallest subeld of R containing Q and

2. Well see later (Example 233a)

on page 191) that:
Every element of Q(

2) has the form a + b

2 for unique rational numbers a, b Q.

Addition in Q(

2) takes the obvious form:

_
a
1
+ b
1

2
_
+
_
a
2
+ b
2

2
_
= (a
1
+ a
2
) + (b
1
+ b
2
)

2.
Thus, Q(

2) is a 2-dimensional vector space over Q, with basis 1,

2. Hence Q(

2)
is a degree-2 extension of Q.
Multiplication in Q(

2) takes the obvious form:

_
a
1
+ b
1

2
_

_
a
2
+ b
2

2
_
= (a
1
a
2
+ 2b
1
b
2
) + (a
1
b
2
+ a
2
b
1
)

2.
For example:
_
1 + 3

2
_

_
5 + 7

2
_
= (1 5 + 2 3 7) + (1 7 + 3 5)

2 = 47 +
22

2.
(b) Q() is the smallest subeld of R containing Q and . Well see later that elements of
Q() take the form of rational functions in ie. they have the form
q
0
+ q
1
+ q
2

2
+ . . . + q
n

n
r
0
+ r
1
+ r
2

2
+ . . . + r
m

m
,
where n < m, and q
0
, . . . , q
n
, r
0
, . . . , r
m
Q.
Addition and multiplication proceed exactly as for rational functions. For example:
_
1 + 2
3 4
_

_
5 + 1
2 1
_
=
(1 + 2) (5 + 1)
(3 4) (2 1)
=
5 + 7 + 2
2
6 5 + 4
2
.
188 CHAPTER 11. FIELD THEORY
The elements
_
. . . ,
1

2
,
1

, 1, ,
2
,
3
, . . .
_
are all Q-linearly independent (this is because is transcendental over Q, a fact which
we will not prove). Thus, Q() is an innite-degree extension of Q.
11.2.3 Finitely Generated Extensions
Let E F be any extension of F, and let
1
,
2
, . . . ,
n
E. We dene F(
1
, . . . ,
n
) to be the
smallest subeld of E containing
1
, . . . ,
n
and F. That is:
F(
1
, . . . ,
n
) =

FLE

1
,...,nL
L.
We say F(
1
, . . . ,
n
) is a nitely generated extension of F.
Example 229:
(a) Q(

2,

3) is the smallest subeld of R containing Q and

2. Well see later (in Example

243a) on page 200) that:
Every element of Q(

2,

3) has the form a +b

2 +c

3 +d

6, for unique
rational numbers a, b, c, d Q.
(b) Q(, e) is the smallest subeld of R containing Q, , and e. Elements of Q(, e) take the
form of rational functions in and e ie. they have the form
J

j=1
K

k=0
q
jk

n
e
m
N

n=1
M

m=0
r
nm

n
e
m
where q
jk
Q and r
nm
Q.
(This is because and e are both is transcendental over Q, and further are algebraically
independent of one another. We will not prove these facts) Thus, Q(, e) is an innite-
degree extension of Q.
Lemma 230 Let E F.
(a) If , E, and A = F(), then F(, ) = A().
(b) More generally, if
1
,
2
, . . . ,
n
, E, then F(
1
, . . . ,
n
, ) = F(
1
, . . . ,
n
)().
Proof: Exercise 174 2
11.3. FINITE AND ALGEBRAIC EXTENSIONS 189
11.3 Finite and Algebraic Extensions
Prerequisites: 11.2
Let E be an extension of F, and let E. If p() = 0 for some p F[x], then we say is
algebraic over F; if not, we say is a transcendental over F.
Example 231:
(a) =
3

2 is algebraic over over Q, because p() = 0, where p(x) = x

3
2.
(b) = i is algebraic over over Q, because p() = 0, where p(x) = x
2
+ 1.
(c) = i is algebraic over over R for the same reason.
(d) and e are transcendental over Q. There is no polynomial p(x) Q[x] such that p() = 0
or p(e) = 0. (This is actually very hard to prove).
If E is an extension of F, then then we say E is an algebraic extension if is algebraic over
F for every E. Otherwise, we say E is a transcendental extension of F. For example,
well show:
Q(

3) is an algebraic extension of Q.
Q(i) is an algebraic extension of Q.
Q(

3, i) is an algebraic extension of Q.
Q() is a transcendental extension of Q.
Q(

3, ) is a transcendental extension of Q.
These claims follow from the following result:
Proposition 232 Let E be an extension of F, and let F. The following are equivalent:
(a) is algebraic over F.
(b) F() is a nite extension of F.
(c) There exists a unique monic, irreducible polynomial m(x) F[x] so that is a root
of m(x). In this case,
1. If p(x) F[x] is any polynomial such that p() = 0, then m(x) divides p(x).
2. F()

= F[x]/(m), via the isomorphism
:
F[x]
(m)
p(x) p() F().
(Note that this does not depend on the eld E).
190 CHAPTER 11. FIELD THEORY
3. deg
_
F() F
_
= degree (m(x)).
4. F() has a F-basis: 1, ,
2
, . . . ,
N
.
Proof: (a)=(c) Let / = p F[x] ; p() = 0, and let m(x) be a nonzero element of
minimal degree in /. By multiplying m(x) by some element of F, we can assume m is monic.
Claim 1: m(x) is irreducible.
Proof: Suppose m(x) = p(x) q(x). Then p() q() = m() = 0, so either p() = 0
or q() = 0. Hence, either p / or q /. Suppose p /. But if m(x) = p(x) q(x),
then degree (p) < degree (m), contradicting the minimality of m. By contradiction, m
must be irreducible. ................................................. 2 [Claim 1]
Proof of (c1) Let p(x) /; we want to show that m(x) divides p(x). Apply Polynomial
Long-Division to write p(x) = q(x)m(x) +r(x), where degree (r) < degree (q). We want to
show r = 0. To see this, observe that
0 = p() = q() m() +r(x) = q() 0 +r(x) = r(x).
Hence, r /. But q is the nonzero element of minimal degree in /, and degree (r) <
degree (q), so we must have r = 0.
Proof of (c2) Dene : F[x]F() by p(x) p(). Let I = image [] F().
It is Exercise 175 to verify that
ker() = (m).
It follows from the Fundamental Isomorphism Theorem that there is an isomorphism
I

=
F[x]
ker()
=
F[x]
(m)
.
It remains to show that I = F(). To see this, observe that
_
m(x) is irreducible.
_
=Prop.165
_
(m) is a maximal ideal.
_
=Cor.167
_
I is a eld.
_
.
Also note that I contains and F. Thus, I must contain F(). But I F(), so we conclude
that I = F().
Proof of (c4) We must show that the set 1, ,
2
, . . . ,
n1
is both linearly independent
and a spanning set for F() as a F-vector space.
Spanning Set: For any f(x) F[x], let f(x) =
_
f(x)
_
.
Claim 2: Every element of F() has the form f
0
+ f
1
+ f
2

2
+ . . . + f
m

m
for some
m N and f
0
, . . . , f
m
F.
11.3. FINITE AND ALGEBRAIC EXTENSIONS 191
Proof: Let e F(). It follows from (c2) that e = f(x), for some polynomial f(x) F[x].
Suppose f(x) = f
0
+f
1
x+f
2
x
2
+. . . +f
m
x
m
; then f(x) = f
0
+f
1
x+f
2
x
2
+. . . +f
m
x
m
=
f
0
+ f
1
+ f
2

2
+ . . . + f
m

m
. ........................................ 2 [Claim 2]
Claim 3: If m n, then
m
= r
0
+ r
1
+ r
2

2
+ . . . + r
n1

n1
for some coecients
r
0
, . . . , r
n1
F.
Proof: Apply Polynomial Long Division to write x
m
= p(x)q(x) + r(x) for some
q, r F[x], where r(x) is a polynomial of degree less than n. Then

m
= x
m
= p(x)q(x) +r(x) = r().
Thus, if r(x) = r
0
+r
1
x+r
2
x
2
+. . . +r
n1
x
n1
, then
m
= r
0
+r
1
+r
2

2
+. . . +r
n1

n1
.
2 [Claim 3]
Combine Claims 4 and 5 to conclude that 1, ,
2
, . . . ,
n1
spans F().
Linearly Independent: Suppose that f
0
+ f
1
+ f
2

2
+ . . . + f
n1

n1
= 0 for some
f
0
, . . . , f
n1
F; we want to show that f
0
= f
1
= . . . = f
n1
= 0.
Let f(x) = f
0
+ f
1
x + f
2
x
2
+ . . . + f
n1
x
n1
, a polynomial in F[x]. Then we have f(x) = 0,
which means that f(x) 1, which means that p(x) divides f(x). Thus, either f = 0, or
degree (f) degree (p) = n. But degree (f) n 1 by construction, so this is impossible.
We conclude that f = 0 in other words, f
0
= f
1
= . . . = f
n1
= 0.
Proof of (c4) It follows from (c3) that deg
_
F() F
_
= degree (m(x)).
(c)=(b) This is immediate.
(b)=(a) Suppose F() has degree n. Then the n + 1 elements 1, ,
2
, . . . ,
n
must be
F-linearly dependent. In other words, there are elements f
0
, f
1
, . . . , f
n
F (not all zero) so
that f
0
+f
1
+. . .+f
n

n
= 0. But this means that f() = 0, where f(x) = f
n
x
n
+. . .+f
1
x+f
0
.
Hence, is a root of the polynomial f, so is algebraic. 2
The polynomial m(x) in Proposition 232 is called the minimal polynomial of , and is
usually denoted m
,F
(x), to make its dependence on and F explicit. The degree of (over
F) is the degree of m
,F
(x), which we denote by degree (; F). It follows that deg
_
F()
_
=
degree (; F).
Example 233:
(a)

2 is algebraic over Q, with minimal polynomial m

2;Q
(x) = x
2
2. Thus, Proposition
232 part (c2) says Q(

2) is isomorphic to Q[x]/(x
2
2). Part (c3) says that Q(

2) is
an extension of degree 2 over Q, and Part (c4) says that Q(

2) has Q-basis 1,

2.
192 CHAPTER 11. FIELD THEORY
-1
1 2
2
3
i
-i
2
3
3
2
2
-1+ i
=
2
3
3
2
2
-1- i
=
Figure 11.11: The three cube roots of 3 in the complex plane.
(b) Let p(x) = x
3
2. Then p(x) is irreducible over Q; its three roots in C are:
3

2; =
3

2
2

_
1 +

3 i
_
; and =
3

2
2

_
1

3 i
_
(see Figure 11.11)
Thus,
3

2 is algebraic over Q, with minimal polynomial m3

2;Q
(x) = p(x). Thus, Propo-
sition 232(c2) says that Q(
3

2) is isomorphic to Q[x]/(p(x)). Part (c3) says that Q(

3

2)
is an extension of degree 3. Part (c4) says that Q(
3

2) has Q-basis 1,
3

2,
3

4.
(c) However,
3

2 R, so R(
3

2) = R in other words, R(
3

2) is a trivial extension of
3

2.
The polynomial p(x) = x
3
2 is not irreducible over R, because it factors:
x
3
2 = (x
3

2)(x
2
+
3

2 x +
3

4) (check this)
Thus, p(x) is not a minimal polynomial for
3

2 over R. Indeed, the minimal polynomial

for
3

2 over R is just m3

2;R
(x) = x
3

2.
(d) Let p(x) = x
3
2 as in Example 233b), and now let =
3

2
2

_
1 +

3 i
_
(see Figure
11.11). As in Example 233b), is algebraic over Q, with minimal polynomial p(x).
Thus, Proposition 232 part (c2) says that Q() is isomorphic to Q[x]/(p(x)); part (c3)
says Q() is an extension of degree 3, and part (c4) says Q() has basis 1, ,
2
. This
is portrayed on the left side of Figure 11.12(A).
Observe that this means Q() is isomorphic to Q(
3

2) (from Example 233b)). However,

Q() is not equal to Q(
3

2), because Q(
3

2) R, whereas Q() extends into the complex

plane.
(e) Let be as in Example 233d). Then , R, so R() is a nontrivial extension of R.
However, p(x) = x
3
2 is not the minimal polynomial for over R, because p(x) is not
irreducible, as we saw in Example 233c). Instead, the minimal polynomial for over R
is given by:
m
;R
(x) = (x
2
+
3

2 x +
3

4)
11.3. FINITE AND ALGEBRAIC EXTENSIONS 193
F
F()=
E
K() =
K
K[x]
m
;
(x)
F[x]
m
;F
(x)
Q
Q()=
C
R() =
R
3
2
D
1
D
2
D
1
D
2
Q[x]
x
3
-2
R[x]
x
2
+ x + 2
3
4
3
(A) (B)
3 2
Figure 11.12: (A) Q() has degree 3 over Q, but R() only has degree 2 over R. This is
because m
;Q
(x) = x
3
2, but m
;R
(x) = x
2
+
3

2x +
3

4. (B) F() has degree D

1
over F,
while K() has degree D
2
D
1
over K.
Thus, Proposition 232(c) says that
R()

= R[x]/(x
2
+
3

2 x +
3

4)
is an extension of degree 2, with R-basis 1, .
This is portrayed on the right side of Figure 11.12(A).
Compare Example 233b) and 233c). Compare Examples 233d) and 233e). It appears
that the minimal polynomial of an element depends on the eld we are comparing to. The
formal statement is as follows:
Proposition 234 Let E F be an extension of F, and let E be algebraic over F. Let
K F be another extension of F. Then, as shown in Figure 11.12(B):
(a) is algebraic over K.
(b) If we treat m
;F
(x) as element of K[x], then m
;F
(x) may no longer be irreducible.
However:
(c) m
;K
(x) divides m
;F
(x) (in K[x]).
(d) Thus, deg
_
K() K
_
deg
_
F() F
_
.
Proof: (a) Clearly, F[x] K[x]. Thus, m
;F
(x) is an element of K[x], so since m
;F
() = 0
by denition, it follows that is algebraic over K.
194 CHAPTER 11. FIELD THEORY
(c) Let m
;K
(x) be the minimal polynomial of over K, then Proposition 232(c1) says
m
;K
(x) divides m
;F
(x).
(b) and (d) follow immediately from (c). 2
Corollary 235 Let E F. Then
_
E is a nite extension of F
_
=
_
E is an algebraic extension of F
_
.
Proof: Let E. We want to show is algebraic.
Consider the subeld F(). Observe that deg
_
F() F
_
deg
_
E F
_
, so it is nite.
Say deg
_
F() F
_
= N. Thus, the elements 1, ,
2
, . . . ,
N
cannot be linearly independant;
hence there are some f
0
, f
1
, . . . , f
N
F (not all zero) so that
f
0
+ f
1
+ . . . + f
N

N
= 0.
But now dene polynomial f(x) F[x] by
f(x) = f
0
+ f
1
x + . . . + f
N
x
N
.
It follows that f() = 0. Hence, is algebraic.
This holds for any E, so E is nite. 2
Note: The converse of Corollary 235 is false. While it is true that every nite extension of
F is algebraic, it is not true that every algebraic extension is nite.
The next theorem can be summarized: A nite extension of a nite extension is nite; an
algebraic extension of an algebraic extension is algebraic.
Theorem 236 Let F E D be a chain of eld extensions.
(a) deg
_
D F
_
= deg
_
D E
_
deg
_
E F
_
(whether these degrees are nite or
innite).
(b) Thus, if D has nite degree over E, and E has nite degree over F, and then D also
has nite degree over F See Figure 11.13(A).
(c) In particular, suppose deg
_
E F
_
= N, and
1
, . . . ,
N
is a F-basis for E. Suppose
deg
_
D E
_
= M, and
1
, . . . ,
N
is a E-basis for D. Then
_

1

1
,
1

2
, . . . ,
1

M
,

1
,
2

2
, . . . ,
2

M
, . . .
.
.
.
.
.
.
.
.
.
.
.
.
. . .
N

1
,
N

2
, . . . ,
N

M
_
is an F-basis for D.
11.3. FINITE AND ALGEBRAIC EXTENSIONS 195
F
E
D
f
i
n
i
t
e
f
i
n
i
t
e
F
E
D
f
i
n
i
t
e
f
i
n
i
t
e f
i
n
i
t
e
implies
(A)
F
E
D
a
l
g
e
b
r
a
i
c
a
l
g
e
b
r
a
i
c
F
E
D
a
l
g
e
b
r
a
i
c
a
l
g
e
b
r
a
i
c
a
l
g
e
b
r
a
i
c
implies
(B)
If D is a finite extension of E, and E is
a finite extension of F, then D is a finite
extension of F.
If D is an algebraic extension of E, and
E is an algebraic extension of F, then D
is an algebraic extension of F.
Figure 11.13:
(d) If D is algebraic over E, and E is algebraic over F, and then D is also algebraic over F
(whether these extensions are nite or innite). See Figure 11.13(B).
Proof: Clearly (a) and (b) follows from (c), so well prove (b). We must show that the set

m=1...M
n=1...N
is F-linearly independent and an F-spanning set for D.
Spanning Set: Let d D be arbitrary. Since
1
, . . . ,
M
is a E-basis for D, we have:
d = e
1

1
+ . . . + e
M

M
(11.1)
for some e
1
, . . . , e
M
E. But
1
, . . . ,
N
is a F-basis for E. Hence, we can write:
e
1
= f
11

1
+ . . . + f
N1

N
.
.
.
.
.
.
.
.
.
e
M
= f
1M

1
+ . . . + f
NM

N
(11.2)
Combining equations (11.1) and (11.2) yields:
d =
M

m=1
e
m

m
=
M

m=1
_
N

n=1
f
nm

n
_

m
=
M

m=1
N

n=1
f
nm

m
.
Linearly Independent: Suppose we had an equation
N

n=1
M

m=1
f
nm

m
= 0 for some
coecients f
nm
F not all zero. We can rewrite this as:
0 =
M

m=1
_
N

n=1
f
nm

n
_

m
=
M

m=1
e
m

m
196 CHAPTER 11. FIELD THEORY
where e
1
= f
11

1
+. . . +f
N1

N
etc. Since
1
, . . . ,
m
are E-linearly independent, we must
conclude that e
1
= . . . = e
M
= 0. But then, since
1
, . . . ,
n
are F-linearly independent, we
must have f
nm
= 0 for all n and m.
(d) Let D; we want to show that is algebraic over F.
Since is algebraic over E, there is some polynomial p(x) E[x] so that p() = 0. Suppose
p(x) =
n
x
n
+. . . +
1
x +
0
, where
0
,
1
, . . . ,
n
E. Let P = F(
1
, . . . ,
n
) E, as in the
gure on the left.
Claim 1: P is nite over F.
Proof: Let P
0
= F(
0
). Then P
0
is a nite extension of
F by Proposition 232(b), because
0
is algebraic over
F.
Next, dene:
P
1
= P
0
(
1
) = F(
0
,
1
)
P
2
= P
1
(
2
) = F(
0
,
1
,
2
)
P
3
= P
2
(
3
) = F(
0
,
1
,
2
,
3
)
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
P
n
= P
n1
(
n
) = F(
0
,
1
,
2
,
3
, . . . ,
n
) = P
Then P
1
is a nite extension of P
0
by Proposition
232(b), because
1
is algebraic over F (and thus, over
P
1
). Likewise, P
2
is nite over P
1
; P
3
is nite over P
2
,
and so on.
Thus, we have a chain of nite extensions:
F P
0
P
1
P
2
. . . P
n
= P
Hence, by iterating part (b) of this theorem, we con-
clude that P is nite over F. ......... 2 [Claim 1]
D
E
E()
F(
1
,...,
n
)=P
P()
F
f
i
n
i
t
e
f
i
n
i
t
e
f
i
n
i
t
e
F(
0
,
1
)=P
1
f
i
n
i
t
e
f
i
n
i
t
e
F(
0
)=P
0
f
i
n
i
t
e
Claim 2: is algebraic over P.
Proof: Recall that p(x) =
n
x
n
+. . . +
1
x +
0
. Now,
1
, . . . ,
n
P by construction, so
p(x) P[x]. But p() = 0; hence is algebraic over P. ................ 2 [Claim 2]
Claim 3: P() is nite over P.
Proof: is algebraic over P, so this follows from Proposition 232. ..... 2 [Claim 3]
Claim 4: P() is nite over F.
Proof: Combine Claims 1 and 3 with part (b) of this theorem. ........ 2 [Claim 4]
11.3. FINITE AND ALGEBRAIC EXTENSIONS 197
Claim 4 and Corollary 235(b) imply that P() is algebraic over F. Hence, is algebraic over
F. This holds for any D. Hence, D is algebraic over F. 2
Corollary 237 Suppose D E F. Then:
(a) deg
_
E F
_
divides deg
_
D F
_
.
(b) Thus, deg
_
E F
_
deg
_
D F
_
, and
(c)
_
deg
_
E F
_
= deg
_
D F
_ _

_
E = D
_
.
(d) Hence, if deg
_
D F
_
is prime, then there is no subeld E such that D E F. 2
Corollary 238 Let E F. The following are equivalent:
(a) E is a nite extension of F
(b) E = F(
1
,
2
, . . . ,
n
), where
1
, . . . ,
n
are elements algebraic over F.
(c) deg
_
E F
_
d
1
d
2
d
n
, where d
k
= degree (
k
; F).
Proof: (a)=(b) (by induction on deg
_
E F
_
)
Base Case: Suppose deg
_
E F
_
= 2. Let E be any element not in F. Then F() is
an extension of F of degree 2 or greater, so Corollary 237(c) says F() = E.
Induction: Suppose that (a)=(b) is true for all elds F

and all extensions E

of
degree less than D. Suppose that deg
_
E F
_
= D.
Let
1
E. Then
1
is algebraic over F, so Corollary 232(b) says deg
_
F(
1
) F
_
= d
1
,
where d
1
= degree (
k
; F) > 1.
Now, let F

= F(
1
); then E F

F, so Theorem 236(a) says that

deg
_
E F

_
=
deg
_
E F
_
deg
_
F

F
_ =
D
d
1
< D.
Hence, by induction, E = F

(
2
, . . . ,
n
) for some elements
2
, . . . ,
n
E. But then
E = F

(
2
, . . . ,
n
) = F(
1
)(
2
, . . . ,
n
) = F(
1
,
2
, . . . ,
n
).
198 CHAPTER 11. FIELD THEORY
Since E is algebraic over F, we know that
2
, . . . ,
n
are algebraic over F.
(b)=(c) (by induction on n)
Base Case (n = 1) If E = F(), then deg
_
E F
_
= degree (; F) by denition.
Induction: Let K = F(
1
, . . . ,
n1
). By induction, we suppose
deg
_
K F
_
d
1
d
2
d
n1
. (11.3)
Note that E = K(
n
). Thus,
deg
_
E K
_
()
degree (
n
; K)
()
degree (
n
; F) = d
n
. (11.4)
Here, () is by the Base Case, and () is by Proposition 234(d). Thus,
deg
_
E F
_
()
deg
_
E K
_
deg
_
K F
_

()
d
n
(d
1
d
2
d
n1
).
Here, () is by Theorem 236(a), and () follows from equations (11.3) and (11.4).
(c)=(a) If deg
_
E F
_
d
1
d
2
d
n
, then deg
_
E F
_
is nite, so E is algebraic
over F by Corollary 235. 2
Corollary 239 Let E F, and let A = E ; is algebraic over F.
Then A is a subeld of E, and is an algebraic extension of F.
Proof: Clearly, F A E. We must show that A is a eld ie. that it is closed under
addition, subtraction, multiplication, and division.
To see this, let , A. Then , are algebraic over F, so the eld F(, ) has nite degree
over F, by Corollary 238. Hence, F(, ) is algebraic over F, by Corollary 235. Hence, all
elements of F(, ) are algebraic ie. F(, ) A.
But + , , , and / are all elements of F(, ), and thus, of A. 2
11.4 (multi)Quadratic Extensions
Prerequisites: 11.3
Let E be an extension of F, and let E. If q() = 0 for some quadratic polynomial
q(x) = x
2
+ q
1
x + q
0
in F[x], then we say is a quadratic root over F. We call F() a
quadratic extension of F.
11.4. (MULTI)QUADRATIC EXTENSIONS 199
Example 240: =

2 is a quadratic root over Q, because q() = 0, where q(x) = x

2
2.
Thus, Q(

2) is a quadratic extension of Q.
Recall that F has characteristic 2 if 1 + 1 = 0 in F. For example, Z
/2
has characteristic
2. However, Q, R, C do not have characteristic 2; nor does Z
/p
, for p > 2.
Proposition 241 Suppose F does not have characteristic 2. Let E F. The following are
equivalent:
(a) E is a quadratic extension of F.
(b) E = F(

D) for some D F which is not the square of any other element in F.

(c) deg
_
E F
_
= 2.
Proof: (a)=(b) Suppose E = F(), where is a quadratic root. Thus, q() = 0, or
some quadratic polynomial q(x) = x
2
+ bx + c.
The roots of q are given by the Quadratic Formula:
=
b

b
2
4c
2
.
(It is elementary to verify that this holds in any eld not of characteristic 2; we need char-
acteristic ,= 2 because the formula calls for division by 2).
Let D = b
2
4c; then D F. I claim that F() = F(

D). To see this, it suces to show

that F(

D) and that

D F().
To see that F(

D), observe that =

1
2
(b

D).
To see that

D F(), observe that

D = 2 b.
(b)=(a) This is immediate.
(a)=(c) Suppose E = F(), where is a quadratic root. Then Proposition 232 implies
that deg
_
F() F
_
= 2.
(c)=(a) Let E such that , F.
Claim 1: is a quadratic root over F
Proof: Note that E is a 2-dimensional F-vector space. Hence, the elements 1, ,
2

cannot be F-linearly independent. Thus, there are f

0
, f
1
, f
2
F (not all zero) such that
f
0
+ f
1
+ f
2

2
= 0. (11.5)
We know that f
2
,= 0, because
_
f
2
= 0
_
=
_
f
0
+ f
1
= 0
_
=
_
= f
0
/f
1
F
_
,
200 CHAPTER 11. FIELD THEORY
and we specied , F. Hence, we can divide equation (11.5) by f
2
, to obtain:
q
0
+ q
1
+
2
= 0,
where q
0
= f
0
/f
2
and q
1
= f
1
/f
2
. In other words, q() = 0, where q(x) = x
2
+q
1
x +x
0
is
a quadratic polynomial. .............................................. 2 [Claim 1]
It remains to show that E = F(). To see this, observe that
2 = deg
_
E F
_
deg
_
F() F
_
> 1.
Hence, we must have deg
_
F() F
_
= 2, which means F() = E. 2
We say that E is a multiquadratic extension of F if we have a sequence of extensions
F = E
0
E
1
E
2
E
N
= E,
where E
n
is a quadratic extension of E
n1
for all n 1.
Corollary 242 If E F is a multiquadratic extension, then deg
_
E F
_
= 2
n
for some n.
Proof: Exercise 176 Hint: Combine Proposition 241(c) with Theorem 236(a) on page 194. 2
Note that the converse of this is false: if deg
_
E F
_
= 2
n
, it is not necessarily true that
E is multiquadratic over F.
Example 243:
(a) Consider Q(

2,

3). Clearly, Q Q(

2) Q(

2,

3).
Claim 1:

3 , Q(

2).
Proof: If

3 Q(

2), then there are rational numbers a, b Q such that

3 = a + b

2. (11.6)
First observe that a ,= 0 ,= b. To see this, suppose b = 0; then equation (11.6) becomes

3 = a that is,

3 Q, which we know is false.

Now suppose a = 0; then equation (11.6) becomes

3 = b

2. Multiply by

2 to get:

6 = 2b, meaning ,

6 Q, which we know is false.

Thus, ab ,= 0. Square both sides of (11.6) to get:
3 = (a + b

2)
2
= a
2
+ 2b
2
2ab

2.
Since ab ,= 0, this implies:

2 =
3 a
2
2b
2
2ab
,
which means

2 Q, which we know is false. ...................... 2 [Claim 1]

11.4. (MULTI)QUADRATIC EXTENSIONS 201
Claim 2: Q(

2,

3) is a quadratic extension of Q(

2).
Proof: Claim 1 implies that Q(

2) Q(

2,

3). Also the minimal polynomial

of

3 over Q(

3) is m

3
(x) = x
2
3 (because m

3
(x) is irreducible over Q(

2), by
Claim 1). ........................................................ 2 [Claim 2]
But Q(

2) itself is a quadratic extension of Q. Thus, Q(

2,

3) is a multiquadratic
extension of Q.
Furthermore, Theorem 236(a) implies:
deg
_
Q(

2,

3) Q
_
= deg
_
Q(

2,

3) Q(

2)
_
deg
_
Q(

2) Q
_
= 2 2
= 4.
This means that Q(

2,

3) is a 4-dimensional vector space over Q. Claim 1 shows that

1,

2,

3 is a linearly independent set. Similar reasoning shows that 1,

2,

3,

6
is also linearly independent, hence, a basis. We conclude:
Every element of Q(

2,

3) has the form a +b

2 +c

3 +d

6, for unique
rational numbers a, b, c, d Q.
(b) Consider Q(
4

2). Since

2 = (
4

2)
2
, it follows that

2 Q(
4

2); hence
Q Q(

2) Q(
4

2).
Claim 3:
4

2 , Q(

2).
Proof: Exercise 177 Hint: Imitate the proof of Claim 1 above. ...... 2 [Claim 3]
Thus, Q(

2) Q(
4

2). Since
4

2 =
_

2, it follows that Q(
4

2) is a quadratic extension
of Q(

2), and thus, a multiquadratic extension of Q.

Example 243a) generalizes as follows:
Proposition 244 Let F C, and let
1
,
2
, . . . ,
n
F. Then:
(a) F(

1
,

2
, . . . ,

n
) is a multiquadratic extension of F.
(b) deg
_
F(

1
,

2
, . . . ,

n
) F
_
= 2
m
, for some m n.
Proof: Exercise 178 Hint: proceed by induction on n. 2
The following result is reminiscent of Theorem 236 on page 194:
202 CHAPTER 11. FIELD THEORY
Proposition 245 If D is a multiquadratic extension of E, and E is a multiquadratic extension
of F, then D is a multiquadratic extension of F.
Proof: Exercise 179 2
A subextension of a multiquadratic extension is also multiquadratic:
Proposition 246 If D is a multiquadratic extension of F, and F E D, then E is also a
multiquadratic extension of F.
Proof: Exercise 180 2
11.5 Compass & Straight-Edge Constructions II
Prerequisites: 11.1, 11.4
Let K R be the eld of constructable lengths, from Proposition 225 on page 181 of 11.1.
Proposition 247 Let K. Then:
(a) Q() is a multiquadratic extension of Q.
(b) Thus, deg
_
Q() Q
_
= 2
n
for some n N.
Proof: Part 1 of Proposition 225 says that K is the smallest subeld of R which contains Q
and is closed under square roots.
We will rst construct another subeld K

R so that Q K

and K

is closed under
square roots. We proceed as follows:
Let K
1
R be the smallest subeld containing Q and the square roots of all elements
in Q (thus, K
1
contains

2,

3, etc.)
Let K
2
R be the smallest subeld containing K
1
and the square roots of all elements
in K
1
(thus, K
2
contains
4

2 =
_

2,
_

2 +

3, etc.)
....inductively, let K
n+1
be the smallest subeld containing K
n
and the square roots of
all elements in K
n
.
We now have an ascending sequence of eld extensions:
Q K
1
K
2
K
3

Now, let K

_
n=1
K
n
.
Claim 1: K

is a eld, and is closed under square roots.

11.5. COMPASS & STRAIGHT-EDGE CONSTRUCTIONS II 203
Proof: Exercise 181 ................................................ 2 [Claim 1]
Clearly, Q K

R. It follows from Claim 1 and Proposition 225 (Part 1) that K K

.
(Actually, K = K

, but this is not important for our purposes.)

Now suppose that K. It follows that K

. But this means that K

n
for some
n N. It thus suces to show:
Claim 2: If K
n
for some n, then Q() is a multiquadratic extension of Q.
Proof: (by induction on n)
Base Case (n = 1): If K
1
, this means that
= a
0
+ a
1
_
b
1
+ a
2
_
b
2
+ . . . + a
m
_
b
m
,
for some a
0
, a
1
, . . . , a
m
and b
1
, . . . , b
m
Q. Thus, Q(

b
1
, . . . ,

b
n
), and thus, Q()
Q(

b
1
, . . . ,

b
n
).
But Proposition 244(a) says Q(

b
1
, . . . ,

b
n
) is a multiquadratic extension of Q. Since
Q() Q(

b
1
, . . . ,

b
n
), it follows from Proposition 246 that Q() is also a multiquadratic
extension of Q.
Induction: If K
n
, then =
0
+
1

1
+
2

2
+ . . . +
m

m
, for some

0
,
1
, . . . ,
m
and
1
, . . . ,
m
K
n1
. Let F = Q(
0
,
1
, . . . ,
m
,
1
, . . . ,
m
). Then:
By identical reasoning to the Base Case, F() is a multiquadratic extension of F.
By induction hypothesis, F is a multiquadratic extension of Q.
Thus, Proposition 245 implies that F() is a multiquadratic extension of Q.
But Q() F(), so Proposition 246 implies that Q() is a multiquadratic extension of
Q. .................................................................. 2 [Claim 2]
Part (b) of our theorem follows from part (a) and Corollary 242. 2
Corollary 248 It is impossible to do any of the following using only compass and straight-
edge:
I Trisect an arbitrary angle.
II Square the circle.
III Double the Cube.
Proof: (II) Doubling the Cube is equivalent to constructing
3

2. But Q(
3

2) is a degree-3
extension of Q, contradicting Proposition 247(b) (because 3 is not a power of 2). Thus,
3

2
cannot be in K.
204 CHAPTER 11. FIELD THEORY
(III) Squaring the circle is equivalent to constructing

, and hence constructing . But
is transcendental over Q, so Q() is an innite extension of Q, contradicting Proposition
247(b) (because is not a power of 2). Thus, cannot be in K.
(I) Let be some angle and let =
1
3
. Thus, = 3. Let = cos(). Well show that,
in general, Q() is not a multiquadratic extension of Q; hence cannot be in K.
Claim 1: cos() = 4
3
3.
Proof: We employ some standard trigonometric identities...
cos() = cos(3) = cos( + 2)
= cos() cos(2) sin() sin(2)
= cos()
_
2 cos
2
() 1
_
sin() 2 sin() cos()
= 2 cos
3
() cos() 2 sin
2
() cos()
= 2 cos
3
() cos() 2
_
1 cos
2
()
_
cos()
= 2 cos
3
() cos() 2 cos() + 2 cos
3
()
= 4 cos
3
() 3 cos() = 4
3
3
.
. . . . . . . . . . . . .......... 2 [Claim 1]
Let = 60
o
. Then is a constructable angle, and cos() =
1
2
. Thus Claim 1 says:
1
2
= 4
3
3.
In other words, p() = 0, where p(x) = 4x
3
3x
1
2
. It can be checked that p is irreducible
over Q. Thus, p(x) is the minimal polynomial of over Q, and Proposition 232(c3) says
that Q() is an extension of degree 3 over Q. But 3 is not a power of 2, so Proposition 247(b)
says that cannot be in K. 2
11.6 Cyclotomic Extensions
Prerequisites: 11.3
Let F be a eld. If n N, then an nth root of unity is a number F so that
n
= 1.
Example 249:
(a) Let F = C. The Nth roots of unity (for N = 1, 2, 3, 4, 5, 6) are shown in Figure 11.14 and
listed in the following table:
11.6. CYCLOTOMIC EXTENSIONS 205
1 -1 1 1
-1
2
3
2
i
+
-1
2
3
2
i
-
1st root
of unity
Square roots
of unity
Cube roots
of unity
-1 1
i
-i
1
e
2i/5
e
4i/5
e
6i/5
e
8i/5
4th roots
of unity
5th roots
of unity
e
2i/3
=
e
4i/3
=
1
6th roots
of unity
e
4i/6
e
8i/6
e
2i/6
e
10i/6
-1
Figure 11.14: Roots of unity in the complex plane.
N Nth roots of unity
N = 1 1.
N = 2 1 and 1.
N = 3 1, e
2i/3
=
1
2
+

3
2
i, and e
4i/3
=
1
2

3
2
i.
N = 4 1, i, 1, and i.
N = 5 1, e
2i/5
, e
4i/5
, e
6i/5
, and e
8i/5
.
N = 6 1, e
i/3
, e
2i/3
, e
3i/6
= 1, e
4i/3
, and e
5i/6
.
In general, for any n N, the nth roots of unity are: 1, e
2i/n
, e
4i/n
, . . . , e
2(n1)i
n
(see
Lemma 250(a) below).
As Figure 11.14 suggests, the nth roots of unity form a set of n equally spaced points
around the unit circle in the complex plane. There is thus a relationship between nth
roots of unity and the construction of a regular n-gon (see 11.7)
(b) Let F = Z
/5
. Then 1, 2, 3 and 4 are all 4th roots of unity:
1
4
= 1;
2
4
= 16 1 (mod 5);
3
4
= 81 1 (mod 5);
4
4
= 256 1 (mod 5).
206 CHAPTER 11. FIELD THEORY
(c) Let F = Z
/p
, for some prime p. Then all nonzero elements of Z
/p
are (p 1)th roots of
unity. In other words, we have:
Fermats Little Theorem: z
p1
1 (mod p), whenever z , 0 (mod p).
Proof: Let Z
/p

= 1, 2, . . . , (p1), and note that Z

/p

forms a group of order (p1)

under multiplication. Thus, if z Z
/p

, then the (multiplicative) order of z divides p1,

so that z
p1
= 1.
Lemma 250 Let Z
n
be the set of all nth roots of unity in C. Then
(a) Z
n
=
_
1, e
2i/n
, e
4i/n
, . . . , e
2(n1)i
n
_
. Thus, card [Z
n
] = n.
(b) Z
n
is a cyclic group under multiplication, and is generated by e
2i/n
.
(c) Z
n
is isomorphic to Z
/n
via the map : Z
/n
k e
2ki
n
Z
n
.
(d) If d 1, 2, . . . , n, then
_
d divides n
_

_
Z
d
Z
n
_
.
Proof: (a,c) Exercise 182 .
(b) First note that Z
n
is closed under multiplication: if
1
,
2
Z
n
, then
(
1

2
)
n
=
n
1

n
2
= 1 1 = 1; hence (
1

2
) Z
n
also.
Thus, Z
n
forms a group. It is clear that all elements of Z
n
are powers of e
2i/n
, so e
2i/n
generates Z
n
, so Z
n
is cyclic.
(d) Recall that = e
2i
d
generates Z
d
. Thus,
_
Z
d
Z
n
_

_
Z
n
_

_

n
= 1
_

_
e
2ni
d
= 1
_

_
2n
d
is an integer multiple of 2
_

_
n
d
is an integer
_

_
d divides n
_
. 2
The nth cyclotomic eld
1
is the eld
CF
n
= Q
_
e
2i/5
, e
4i/n
, . . . , e
2(n1)i
n
_
,
generated by all nth roots of unity over the rational numbers. Our rst task is to show that
CF
n
is in fact a simple extension of Q, generated by a single element: a primitive root of unity.
Lemma 250(a) says that Z
n
is a group under multiplication. A primitive nth root of
unit is an element of Z
n
which generates Z
n
as a group. The following table lists the primitive
Nth roots of unity for N = 2, 3, 4, 5, 6:
1
The term cyclotomic means circle-cutting (cyclo-tomic) in Greek, and refers to the fact that the nth
roots of unity cut the circle into n equal peices.
11.6. CYCLOTOMIC EXTENSIONS 207
N Primitive Nth roots of unity
N = 2 1.
N = 3 e
2i/3
and e
4i/3
.
N = 4 i, and i.
N = 5 e
2i/5
, e
4i/5
, e
6i/5
, and e
8i/5
.
N = 6 e
2i/6
and e
10i/6
.
N = 7 e
2i/7
, e
4i/7
, e
6i/7
, e
8i/7
, e
10i/7
, and e
12i/7
.
N = 8 e
i/4
, e
3i/4
, e
5i/4
, and e
7i/4
.
N = 9 e
2i/9
, e
4i/9
, e
8i/9
, e
10i/9
, e
14i/9
, and e
16i/9
.
Let Z

n
= Z
n
; a primitive root. The primitive roots allow us to express CF
n
as a
simple extension of Q:
Lemma 251 If Z

n
be any primitive nth root of unity, then CF
n
= Q().
Proof: First note that Q() CF
n
. We claim that also CF
n
Q(). To see this, note
that Q() contains all powers of ; since generates Z
n
, it follows that Q() contains all nth
roots of unity, so Q() contains CF
n
. 2
Our next task is to count Z

n
. We dene Eulers -function:
(n) = # of numbers j 1, 2, ..., n 1 which are relatively prime to n.
Example 252:
(a) (5) = 4, because all of the numbers 1, 2, 3, 4 are relatively prime to 5.
(b) (16) = 8, because the numbers 1, 3, 5, 7, 9, 11, 13, 15 are relatively prime to 16.
See the Appendix at the end of this section for more about the -function.
Lemma 253
(a) If Z
n
, then
_
Z

n
_

_
k = n 1 is the smallest k > 0 for which
k
= 1
_
.
(b) Z

n
=
_
e
2pi/n
; where p is relatively prime to n
_
.
(c) More generally, if Z

n
is any xed element, then Z

n
=
p
; p is relatively prime to n.
(d) Thus, card [Z

n
] = (n).
(e) Z
n
=
_
d divides n
Z

d
.
208 CHAPTER 11. FIELD THEORY
Proof: (a) If Z
n
, then
n1
= 1, because [Z
n
[ = (n 1). Thus, Z

n
if and only if
Z
n
= 1, ,
2
, . . . ,
n1
, and this occurs only if k = n 1 is the smallest k > 0 for which

k
= 1.
(b,c) follow from (a) (Exercise 183). (d) follows from the denition of (n).
(e) For all d n, we dene
Z

d
=
_
Z
n
; k = d 1 is the smallest k > 0 for which
k
= 1
_
.
Claim 1: Z
n
=
_
d divides n
Z

d
.
Proof: Exercise 184 . First check the following:
If d
1
,= d
2
, then Z

d
1
and Z

d
2
are disjoint.
Every element of Z
n
must belong to Z

d
for some d.
Z

d
= unless d divides n.
Now combine these to get the Claim. ................................. 2 [Claim 1]
Claim 2: For every d which divides n, Z

d
= Z

d
.
Proof: Observe that:
Z
d
Z
n
; this follows from Lemma 250(d).
Z

d
Z
d
; to see this, note that, if Z

d
, then
d
= 1 by denition.
Part (a) of the present theorem says: for any Z
d
,
_
Z

n
_

_
Z

n
_
.
It follows that Z

d
= Z

d
. ............................................. 2 [Claim 2]
(d) follows from Claims 1 and 2. 2
Remark: Note that part (d) of the theorem yields the surprising identity:
(n 1) =

d divides n
(n) .
Our next goal is to represent CF
n
as the splitting eld of a single irreducible polynomial.
We dene the nth cyclotomic polynomial:
cp
n
(x) =

n
(x ). (11.7)
11.6. CYCLOTOMIC EXTENSIONS 209
This is the smallest polynomial having all elements of Z

n
as roots. For example:
x (1) = cp
1
(x) = x 1;
x (1) = cp
2
(x) = x + 1;
_
x e
2i
3
_

_
x e
4i
3
_
= cp
3
(x) = x
2
+ x + 1;
(x +i) (x i) = cp
4
(x) = x
2
+ 1;
_
x e
2i
5
_

_
x e
4i
5
_

_
x e
6i
5
_

_
x e
8i
5
_
= cp
5
(x) = x
4
+ x
3
+ x
2
+ x + 1;
_
x e
i
3
_

_
x e
5i
3
_
= cp
6
(x) = x
2
x + 1;
_
x e
2i
7
_

_
x e
4i
7
_

_
x e
6i
7
_

_
x e
8i
7
_

_
x e
10i
7
_

_
x e
12i
7
_
= cp
7
(x) = x
6
+ x
5
+ x
4
+ x
3
+ x
2
+ x + 1;
_
x e
i
4
_

_
x e
3i
4
_

_
x e
5i
4
_

_
x e
7i
4
_
= cp
8
(x) = x
4
+ 1;
_
x e
2i
9
_

_
x e
4i
9
_

_
x e
8i
9
_

_
x e
10i
9
_

_
x e
14i
9
_

_
x e
16i
9
_
= cp
9
(x) = x
6
+ x
3
+ 1;
.
.
.
.
.
.
.
.
.
This partial list suggests that cp
n
(x) always has integer coecients, something which is not
obvious from equation (11.7).
Proposition 254
(a) cp
n
(x) is a monic, irreducible polynomial with integer coecients.
(b) degree
_
cp
n
(x)
_
= (n).
(c) For any n N, (x
n
1) =

d divides n
cp
d
(x).
Proof: (b) cp
n
(x) is a product of (n) linear factors. Thus, degree (cp
n
(x)) = (n).
(c) The roots of (x
n
1) are exactly the nth roots of unity. Hence, (x
n
1) factors
completely over CF
n
:
x
n
1 =

Zn
(x ). (11.8)
But Lemma 253(d) says Z
n
=
_
d divides n
Z

d
. Hence, we can rewrite equation (11.8) as:
x
n
1 =

d divides n

n
(x ) =

d divides n
cp
d
(x),
where the last equality is just the equation (11.7) which denes cp
d
(x). 2
210 CHAPTER 11. FIELD THEORY
The proof of part (a) will require three preliminary lemmas.
Lemma 255 Let E Q be a eld extension. Let q(x) Q[x] and e(x) E[x], and let
p(x) = q(x) e(x). If p(x) Q[x], then e(x) Q[x] also.
Proof: Apply the Division Algorithm to divide p(x) by q(x) in Q[x]. We get unique
polynomials s(x) and r(x) in Q[x] such that
p(x) = s(x) q(x) + r(x), and degree (r(x)) < degree (q(x)) . (11.9)
Now, the Division Algorithm yields a unique solution in any polynomial ring, and equation
(11.9) is also valid in the ring E[x]; hence, q(x) and r(x) are the unique elements in E[x] such
that (11.9) is true.
However, we already have a division equation in E[x], namely:
p(x) = q(x) e(x).
Hence, we conclude: e(x) = s(x) (and r(x) = 0). In other words, e(x) Q[x]. 2
(Remark: Lemma 255 holds for any eld, not just Q.)
Lemma 256 (Gauss Lemma)
Suppose that q(x) Z[x] is monic, and let e(x) Q[x]. Let p(x) = q(x) e(x).
If p(x) Z[x], then e(x) Z[x] also.
Proof: Suppose e(x) = e
0
+e
1
x+e
2
x
2
+. . . +e
n
x
n
, where e
0
, . . . , e
n
Q. Let D be the lowest
common multiple of the denominators of e
0
, . . . , e
n
. I claim D = 1 ie. all of e
0
, . . . , e
n
are
actually integers.
To see this, let
e

(x) = D e(x) = e

0
+ e

1
x + e

2
x
2
+ . . . + e

n
x
n
,
where e

0
= D e
0
, e

1
= D e
1
, etc. Thus, e

(x) has all integer coecients, and we have:

D p(x) = q(x) e

(x). (11.10)
Let p be a prime divisor of D.
Claim 1: p divides all of e

0
, . . . , e

n
.
11.6. CYCLOTOMIC EXTENSIONS 211
Proof: Reduce all coecients in p(x), q(x), and e

(x) modulo p. Then equation (11.10)

becomes:
0 = q(x) e

(x), (11.11)
an equation in the polynomial ring Z
/p
[x]. But:
_
p is prime
_
=Exmpl107
_
Z
/p
is a eld
_
=Exmpl109g
_
Z
/p
[x] is an integral domain
_
.
Thus, equation (11.11) implies that either q(x) = 0 or e

(x) = 0. In other words, either

all coecients in q are congruent to zero mod p, or all coecients in e

are zero mod p.

Now, q(x) is monic. That is, q(x) = x
n
+ q
n1
x
n1
+ . . . + q
1
x + q
0
. Thus, q(x) =
x
n
+ q
n1
x
n1
+ . . . + q
1
x + q
0
has a nonzero leading coecient, so q(x) cannot be zero.
It follows that e

0
. . . e

n
0 (mod p). Thus, all coecients e

0
, . . . , e

n
of e

must be
divisible by p. ....................................................... 2 [Claim 1]
But D is the lowest common multiple of the denominators of e
0
, . . . , e
n
, so e

0
, . . . , e

n
cannot
have any common factors. Contradiction.
We conclude that D = 1, which means that e
0
, . . . , e
n
were integers all along. 2
Lemma 257 If , Z

n
are any two primitive roots of unity, then there is a eld automor-
phism : CF
n
CF
n
such that:
(a) () = .
(b) acts as the identity on Q.
Proof: Lemma 251 says that Q() = CF
n
= Q(). Thus, every element of CF
n
has the form
q
0
+ q
1
+ q
2

2
+ . . . + q
n1

n1
for some (not necessarily unique) q
1
, . . . , q
n1
Q. Dene
the map : CF
n
CF
n
by
(q
0
+ q
1
+ q
2

2
+ . . . + q
n1

n1
) = q
0
+ q
1
+ q
2

2
+ . . . + q
n1

n1
.
Clearly, () = , and thus, (Q()) = Q() in other words, (CF
n
) = CF
n
. Also,
acts as the identity on Q (because (q
0
) = q
0
). It is Exercise 185to verify that is a eld
automorphism. 2
Proof of Proposition 254(a) cp
n
(x) is a product of monic linear polynomials, so it is
monic.
Claim 1: cp
n
(x) has integer coecients ie. cp
n
(x) Z[x].
212 CHAPTER 11. FIELD THEORY
Proof: (by induction on n).
Assume that cp
d
(x) Z[x] for all d < n.
Let p(x) = x
n
1. From (c), we know that
p(x) =

d divides n
cp
d
(x) = q(x) cp
n
(x), where q(x) =

d<n
d divides n
cp
d
(x).
By induction, cp
d
(x) Z[x] for all d < n; hence, q(x) Q[x]. Also, clearly, p(x) Q[x].
Apply Lemma 255 (with E = CF
n
and e(x) = cp
n
(x)) to conclude that cp
n
(x) Q[x]
also.
Now observe that p(x) Z[x] and q(x) Z[x], and that q(x) is monic. It follows from
Gauss Lemma (Lemma 256) that cp
n
(x) Z[x]. ...................... 2 [Claim 1]
cp
n
(x) is irreducible: Suppose not. Then cp
n
(x) = p(x) q(x) for some p(x) and q(x) in
Q[x]. But equation (11.7) says cp
n
(x) =

n
(x ), so this means that there are subsets
T, Q Z

n
so that Z

n
= T . Q, and p(x) =

P
(x ), and q(x) =

Q
(x ).
Now, let T and Q. By Lemma 255 we can dene an automorphism : CF
n
CF
n
so that () = and acts as the identity on Q.
Since, acts as the identity on Q, it follows that
_
p(x)
_
= p(x). But p(x) =

P
(x),
so this means that (T) = T. But () = Q; a contradiction.
By contradiction, the hypothetical factorization cannot exist: cp
n
(x) is irreducible. 2
Remark: In this proof, we built an automorphism of CF
n
which permuted the roots of
cp
n
(x), and used this to deduce information about the structure of cp
n
(x). This use of auto-
morphisms of eld extensions (especially those which permute the roots of some polynomial)
is the key concept of Galois Theory.
Corollary 258
(a) cp
n
(x) is the minimal polynomial (over Q) for any primitive nth root of unity.
(b) deg
_
CF
n
Q
_
= (n).
Proof: (a) Let be a primitive nth root of unity. Then cp
n
(x) is irreducible, monic, and
has as a root, so cp
n
(x) is the minimal polynomial for , according to Proposition 232(c).
(b) This follows from part (a), Proposition254(b), and Proposition 232(c3). 2
11.7. COMPASS & STRAIGHT-EDGE CONSTRUCTIONS III 213
Appendix: Eulers -function
We dene Eulers -function:
(n) = # of numbers j 1, 2, ..., n 1 which are relatively prime to n.
Example 259:
(a) (7) = 6, because all of the numbers 1, 2, 3, 4, 5, 6 are relatively prime to 7.
(b) In general, if p is prime, then (p) = p 1, because all of the numbers 1, 2, ..., p 1
are relatively prime to p.
(c) (25) = 20, because the numbers 1, 2, 3, 4, 6, 7, 8, 9, 11, 12, 13, 14, 16, 17, 18, 19, 21, 22, 23, 24
are relatively prime to 25. The only numbers not relatively prime to 25 are the multiples
of 5, namely 5, 10, 15, 20.
(d) (15) = 8, because the numbers 1, 2, 4, 7, 8, 11, 13, 14 are relatively prime to 25. Note
that 15 = 3 5, and 8 = 2 4 = (3) (5).
These examples illustrate the general rule:
Proposition 260 Properties of Eulers -function
(a) If p is prime, and n N, then (p
n
) = p
(n1)
(p 1).
(b) If n and m are relatively prime, then (n m) = (n) (m).
(c) If n has prime factorization n = p

1
1

2
2
p

k
k
, then
(n) = p
(
1
1)
1
(p
1
1) p
(
2
1)
2
(p
2
1) p
(
k
1)
k
(p
k
1).
Proof: (a) Note that the only numbers in 1, 2, . . . , p
n
1 which are not relatively prime
to p
n
are the multiples of p, namely p, 2p, . . . , p
n
p. There are exactly p
n1
of these,
leaving p
n
p
n1
= p
n1
(p 1) relatively prime elements.
(b) Exercise 186 . (c) follows by combining (a) and (b). 2
11.7 Compass & Straight-Edge Constructions III
Prerequisites: 11.1, 11.4, 11.6 Recommended: 11.5
We will now characterize the regular polygons can be constructed with compass and straight-
edge. Let K R be the eld of constructable lengths, from Proposition 225 in 11.1. Recall
Proposition 247 from 11.5:
214 CHAPTER 11. FIELD THEORY
Proposition 247 Let K. Then:
(a) Q() is a multiquadratic extension of Q.
(b) Thus, deg
_
Q() Q
_
= 2
n
for some n N. 2
A Fermat prime is a prime number of the form p = 2
(2
n
)
+ 1 for some n. For example:
3 = 2
1
+ 1 = 2
(2
0
)
+ 1 is prime.
5 = 2
2
+ 1 = 2
(2
1
)
+ 1 is prime.
17 = 2
4
+ 1 = 2
(2
2
)
+ 1 is prime.
257 = 2
8
+ 1 = 2
(2
3
)
+ 1 is prime.
65537 = 2
16
+ 1 = 2
(2
4
)
+ 1 is prime.
(However, 2
32
+ 1 is not prime, because it is divisible by 641.) It is an open problem whether
there are innitely many Fermat primes.
Proposition 261
_
The regular N-gon is constructable with compass & straight-edge
_

_
N = 2
n
p
1
p
2
p
k
, where p
1
, p
2
, . . . , p
n
are distinct Fermat primes
_
.
Thus, a regular N-gon is constructable if and only if N is one of the following values:
2 4 8 16 32 64 . . .
3 6 12 24 48 92 . . .
5 10 20 40 80 160 . . .
15 30 60 120 240 480 . . .
17 34 68 136 272 544 . . .
51 102 204 408 816 1632 . . .
85 170 340 680 1360 2720 . . .
257 514 1028 2056 4112 8224 . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
However, a regular N-gon is not constructible if N is one of:
7, 9, 11, 13, 14, 18, 19, 21, 22, 23, 25, 26, 27, 28, 29, 31, 33, 35, 36, 37, 38, 39, 41, 42, 43, 44, 45, 46, 47, 49, . . .
Proof of Proposition 261: We will only prove =. Let = e
2i/N
, a primitive Nth
root of unity. Write = z
r
+ z
i
i, where z
r
, z
i
R.
Claim 1:
_
The regular N-gon is constructable with compass & straight-edge
_

_
z
r
and z
i
are in K.
_
11.7. COMPASS & STRAIGHT-EDGE CONSTRUCTIONS III 215
Proof: Constructing the N-gon is equivalent to cutting the circle into N equal segments,
which is equivalent to constructing the angle 2/N, which is equivalent to construct-
ing the lengths cos(2/N) and sin(2/N). But z
r
= cos(2/N) and z
i
= sin(2/N).
2 [Claim 1]
Claim 2:
_
z
r
and z
i
are in K.
_
=
_
(N) is a power of 2
_
.
Proof: Let E = Q(z
r
, z
i
).
Claim 2.1: deg
_
Q() E
_
= 2.
Proof: First note that Q() E(i), because = z
r
+ z
i
i. Thus, deg
_
Q() E
_

deg
_
E(i) E
_
= 2. But deg
_
Q() E
_
cant be 1 (because E is a subset of R, whereas
Q() contains complex numbers). So deg
_
Q() E
_
must be 2. ... 2 [Claim 2.1]
. Next, observe that:
deg
_
Q() Q
_
= (N), by Corollary 258(b) (because is a primitive Nth root of
unity).
deg
_
Q() Q
_
= deg
_
Q() E
_
deg
_
E Q
_
, by Theorem 236(a).
It follows that (N) = 2 deg
_
E Q
_
. But Proposition 247(b) says:
_
z
r
and z
i
are in K.
_
=
_
deg
_
E Q
_
is a power of 2
_
.
It follows that (N) is also a power of 2. ............................. 2 [Claim 2]
Claim 3:
_
(N) is also a power of 2
_

_
N = 2
n
p
1
p
2
p
k
, where p
1
, p
2
, . . . , p
n
are distinct Fermat primes
_
.
Proof: Suppose N has prime factorization: N = 2
n
p

1
1
p

2
2
p

k
k
.
Then Proposition 260(c) says that
(N) = 2
n1
p
(
1
1)
1
(p
1
1) p
(
2
1)
2
(p
2
1) p
(
k
1)
k
(p
k
1).
Thus, (N) is a power of 2 if and only if:

1
=
2
= . . . =
k
= 1,
(p
1
1), (p
2
1), . . ., and (p
k
1) are powers of 2.
In other words, p
1
= 2
n
1
+1, . . ., p
k
= 2
n
k
+1, for some n
1
, . . . , n
k
N. It remains to show
that, p
1
, . . . , p
k
must be Fermat primes ie. that n
1
, . . . , n
k
must themselves be powers of
2. This follows from Claim 4 below. .................................. 2 [Claim 3]
216 CHAPTER 11. FIELD THEORY
Claim 4: If p = 2
n
+ 1 is a prime number, then n must be a power of 2.
Proof: Exercise 187 . .............................................. 2 [Claim 4]
The proof of the converse in the theorem requires Galois theory, and is beyond our scope. 2
11.8. MINIMAL ROOT EXTENSIONS 217
F
E
D

F
E
D

F
E

Id
(A) (B)
Figure 11.15: E is the unique minimal root extension of F for the polynomial p(x).
11.8 Minimal Root Extensions
Prerequisites: 11.2 Recommended: ??
Normally, the reason we extend the eld F is to nd solutions to polynomial equations.
For example, it is irritating that the equation
x
2
+ 1 = 0 (11.12)
has no solutions x R. Thus, we extend R to the larger eld C, which is specically designed
so that the equation (11.12) has solutions (namely x = i).
Recall that a root of the polynomial p(x) is some E so that p() = 0. Thus, the
polynomial p(x) = x
2
+1 has roots = i. We often extend elds by adjoining roots to them;
for example, we get C be adjoining the root i to R. The generalization of this procedure is
the following proposition. On rst reading, imagine F = R, p(x) = x
2
+ 1, E = C, and = i
(see Example 171 on page 143)
Proposition 262 (Minimal Root Extension) Let F be any eld and let p(x) F[x] be an
irreducible polynomial.
There is a unique eld extension E F such that:
(a) p(x) has a root in E in other words, there is some E so that p() = 0.
(b) E is the minimal extension with this property. In other words, suppose D F is
another extension of F, and D is a root of p(x) (see Figure 11.15A). Then there is a
subeld E

D (see Figure 11.15B) so that

F E

and E

;
E

= E

, via an isomorphism : EE

such that:
218 CHAPTER 11. FIELD THEORY
() = ;

[
F
= Id.
(c) deg
_
E F
_
= degree (p(x)).
(d) An F-basis for E is given by the elements 1, ,
2
, . . . ,
n1
, where n = degree (p(x)).
Proof: (a) Let 1 F[x] be the principal ideal generated by p(x), and let E = F[x]/1.
Claim 1: E is a eld.
Proof: This follows from Proposition 170, whose proof we recapitulate:
_
p(x) is irreducible.
_
= Prop.165
_
1 is a maximal ideal.
_
= Cor.167
_
E is a eld.
_
.
2 [Claim 1]
Let : F[x]E be the quotient map. For any polynomial q F[x], let q = (q) be the
corresponding element of E. Recall that F F[x] (since elements of F can be treated as
constant polynomials). Let F E be the image of F in E.
Claim 2: F

= F.
Proof: Exercise 188 ................................................ 2 [Claim 2]
Thus, we will identify F with F, and think of it F as a subeld of E.
Now, consider the polynomial x F[x]. Let = x
Claim 3: p() = 0.
Proof: p() = p(x) = p(x) = (p) = 0, by denition. .............. 2 [Claim 3]
Proof of (d) (This is very similar to the proof of Proposition 232(b4)).
We must show that the set 1, ,
2
, . . . ,
n1
is both linearly independent and a spanning set
for E as an F-vector space.
Spanning Set:
Claim 4: Every element of E has the form f
0
+ f
1
+ f
2

2
+ . . . + f
m

m
for some m N
and f
0
, . . . , f
m
F.
Proof: If e E, then e = f(x), for some polynomial f(x) F[x]. Suppose f(x) = f
0
+f
1
x+
f
2
x
2
+. . . +f
m
x
m
; then f(x) = f
0
+f
1
x+f
2
x
2
+. . . +f
m
x
m
= f
0
+f
1
+f
2

2
+. . . +f
m

m
.
2 [Claim 4]
Claim 5: If m n, then
m
= r
0
+ r
1
+ r
2

2
+ . . . + r
n1

n1
for some coecients
r
0
, . . . , r
n1
F.
11.8. MINIMAL ROOT EXTENSIONS 219
Proof: Apply Polynomial Long Division to write x
m
= p(x)q(x) + r(x) for some
q, r F[x], where r(x) is a polynomial of degree less than n. Then

m
= x
m
= p(x)q(x) +r(x) = r(). (because p(x) = 0)
Thus, if r(x) = r
0
+r
1
x +r
2
x
2
+. . . +r
n1
x
n1
, then
m
= r
0
+r
1
+r
2

2
+. . . +r
n1

n1
.
2 [Claim 5]
Combine Claims 4 and 5 to conclude that 1, ,
2
, . . . ,
n1
spans E.
Linearly Independent: Suppose that f
0
+ f
1
+ f
2

2
+ . . . + f
n1

n1
= 0 for some
f
0
, . . . , f
n1
F; we want to show that f
0
= f
1
= . . . = f
n1
= 0.
Let f(x) = f
0
+f
1
x+f
2
x
2
+. . .+f
n1
x
n1
, a polynomial in F[x]. Then we have f(x) = f() = 0,
which means that f(x) 1, which means that p(x) divides f(x). Thus, either f = 0, or
degree (f) degree (p) = n. But degree (f) n 1 by construction, so this is impossible.
We conclude that f = 0 in other words, f
0
= f
1
= . . . = f
n1
= 0.
Proof of (c) This follows immediately from (d).
Proof of (b) Let D be some other extension of F, and suppose D was a root of p(x)
ie. p() = 0. From (d), we know that every element of E has the form f
0
+ f
1
+ f
2

2
+
. . . + f
n1

n1
for some f
0
, . . . , f
n1
F. Dene : ED as follows:

_
f
0
+ f
1
+ f
2

2
+ . . . + f
n1

n1
_
= f
0
+ f
1
+ f
2

2
+ . . . + f
n1

n1
It is Exercise 189 to check that is a eld monomorphism. Thus, if E

= (E), then E

is
isomorphic to E, and is a subeld of D. Clearly, acts as the identity on F, so F E

. 2
We call E the minimal root extension of F induced by the irreducibe polynomial p(x).
It follows from the proof of Proposition 262(d) that elements of E can be treated as poly-
nomials in the symbol , of degree less than n. The multiplication of two elements of E is then
equivalent to multiplying these polynomials, subject to the reduction provided by Claim 5.
Since E is obtained by adjoining the root to F, and since elements of E are polynomials
in , we often write E = F().
Example 263:
(a) Let F = R and let p(x) = x
2
+1. Then the root extension E = R[x]/(x
2
+1) is isomorphic
to C, via the identication = i. Any element of c C is a polynomial in i of degree
1 namely, c = r
1
+ r
2
i, where r
1
, r
2
R. The multiplication of two complex numbers
then has the form:
(r
1
+ r
2
i) (s
1
+ s
2
i) = r
1
s
1
+ r
1
s
2
i + r
2
s
1
i + r
2
s
2
i
2
()
r
1
s
1
+ (r
1
s
2
+ r
2
s
1
)i r
2
s
2
= (r
1
s
1
r
2
s
2
) + (r
1
s
2
+ r
2
s
1
)i.
Here, () follows from the reduction formula: i
2
= 1, which is the form which Claim 5
takes here.
220 CHAPTER 11. FIELD THEORY
0 1 2 3 4
3
2
3
2
2
3
2
3
3
2
4
3
2
1+
3
2
1+2
3
2
1+3
3
2
1+4
3
2
2+
3
2
2+2
3
2
2+3
3
2
2+4
3
2
3+
3
2
3+2
3
2
3+3
3
2
3+4
3
2
4+
3
2
4+2
3
2
4+3
3
2
4+4
Figure 11.16: E = Z
/5
(

3) has 25 elements.
(b) Let F = Z
/5
. Observe that the polynomial p(x) = x
2
3 has no roots in Z
/5
(because
1
2
= 1 = 4
2
and 2
2
= 4 = 3
2
). Let E = Z
/5
[x]/(p). Then E is an extension of Z
/5
of
degree 2, generated by an element which satises
2
3 = 0. Thus, we could write
=

3, so that E = Z
/5
(

3). As shown in Figure 11.16, eld E contains exactly 25

elements, each of the form a + b

3, where a, b Z
/5
.
Addition in Z
/5
(

3) takes the obvious form:

_
a
1
+ b
1

3
_
+
_
a
2
+ b
2

3
_
= (a
1
+ a
2
) + (b
1
+ b
2
)

3.
Thus, Z
/5
(

3) is a 2-dimensional vector space over Z

/5
, with basis 1,

3.
Multiplication in Z
/5
(

3) takes the obvious form:

_
a
1
+ b
1

3
_

_
a
2
+ b
2

3
_
= (a
1
a
2
+ 3b
1
b
2
) + (a
1
b
2
+ a
2
b
1
)

3.
For example:
_
1 + 2

3
_

_
1 + 4

3
_
= (1 1 + 3 2 4) + (1 4 + 2 1)

3 = 25 + 6

3 =

3.
Note that E contains exactly 25 = 5
2
elements. Well see later that a similar construction
yields a eld of cardinality p
n
for any n N and any prime p.
The notation E = F() may seem ambiguous, because it suggests the simple extensions
introduced on page 187. However, Proposition 262(b) can be restated:
Corollary 264 Let F E and p(x) be as in Proposition 262. If K is any extension of F
containing a root of p(x), then there is a natural isomorphism E

= F(). 2.
Appendix A
Background: Topology
A.1 Introduction
There are several mathematical approaches to studying space and its properties. Each approach
uses a dierent mathematical structure to represent spatial structure...
Linear Algebra studies the geometry of vector spaces and their subspaces, and is concerned
primarily with at objects like lines and planes.
Dierential Geometry uses the tools of dierential calculus to study the geometry of curves,
surfaces, and other dierentiable manifolds.
Algebraic Geometry uses the algebraic structure of function rings to study the geometry of
curves, surfaces, and other algebraic varieties.
Metric Space Theory studies geometry of a space endowed with a concept of distance (a
metric).
Functional Analysis combines methods of linear algebra and metric spaces to study the
geometry of innite dimensional vector spaces, such as Banach spaces and Hilbert spaces.
Of all of these, Topology is the most abstract and fundamental. Topology studies those
properties of space which are independent of any particular metric or coordinate structure.
Topology is primarily concerned with four issues:
Convergence.
Closure.
Connectedness.
Continuity.
221
222 APPENDIX A. BACKGROUND: TOPOLOGY
3
.
1
3
9
3
.
1
4
0
3
.
1
4
1
3
.
1
4
2
p
1
p
2
p
3
3
.
1
4
1
4
3
.
1
4
1
5
3
.
1
4
1
6
3
.
1
4
1
7
p
2
p
3
p
4
p
5
3
.
1
4
1
5
9
2
6
5
0
3
.
1
4
1
5
9
2
6
5
2
3
.
1
4
1
5
9
2
6
5
4
3
.
1
4
1
5
9
2
6
5
6
p
5

3
.
1
4
1
5
9
2
6
5
1
3
.
1
4
1
5
9
2
6
5
3
3
.
1
4
1
5
9
2
6
5
5
p
6
p
7
Figure A.1: The sequence p
1
, p
2
, p
3
, . . . converges to .
Convergence: Approximation is an ancient and familiar idea. For example, consider the
number . We can never exactly express , but for practical purposes it suces to approximate
, for example, by:
3.14159265358979323854626400000 . . .
So, consider this sequence of numbers:
p
1
= 3.14000000000000000 . . .
p
2
= 3.14160000000000000 . . .
p
3
= 3.14159300000000000 . . .
p
4
= 3.14159265000000000 . . .
p
5
= 3.14159265360000000 . . .
p
6
= 3.14159265359000000 . . .
p
7
= 3.14159265358979000 . . .
.
.
.
.
.
.
.
.
.
(Figure A.1) (A.1)
Clearly, this sequence of numbers is getting closer and closer to . We say that the sequence
p
1
, p
2
, p
3
, . . . is converging to . We can justify this as follows:
p
1
agrees with up to two decimal places.
p
2
agrees with up to four decimal places.
p
3
agrees with up to six decimal places.
A.1. INTRODUCTION 223
....and in general, for any L > 0, there is some N so that p
N
agrees with up to L decimal
places. Furthermore, p
n
also agrees with up to L decimal places, for any n > N.
Observe that
_
p
n
agrees with up to L decimal places
_

_
[p
n
[ <
1
10
L
_
.
Thus, we could just say:
For any L > 0, there is some N so that [p
N
[ <
1
10
L
.
Furthermore, [p
n
[ <
1
10
L
, for any n > N.
or, more succinctly,
For any L > 0, there is some N so that, for any n N, [p
n
[ <
1
10
L
.
Of course, there is nothing special about powers of 10. Instead of talking about
1
10
L
, we could
measure convergence using any number > 0. We have arrived at the denition of convergence:
For any > 0, there is some N so that, for any n N, [p
n
[ < .
We say is the limit of the sequence p
1
, p
2
, p
3
, . . ., and write:
lim
n
p
n
= .
If we are looking at sequences of real numbers, then the concept of convergence is pretty
straightforward. But in an abstract space X, it is not so transparent. An important question
in topology is, When do sequences in X converge, and what does this mean?
Closure: We say that a subset U R is closed if no sequence of elements in U can converge
to a point outside of U.
Example 265:
(a) The open interval U = (0, ) = r R ; 0 < r is not closed. To see this, consider the
sequence
u
1
= 0.1
u
2
= 0.01
u
3
= 0.001
u
4
= 0.0001
.
.
.
.
.
.
.
.
.
(A.2)
Clearly, the points u
1
, u
2
, u
3
, . . . are all in (0, ), but they converge to 0, which is not in
(0, ). Hence, (0, ) is not closed.
224 APPENDIX A. BACKGROUND: TOPOLOGY
(b) The interval U = [0, ) = r R ; 0 r is closed. Any sequence of numbers greater
than or equal to zero must converge to a limit greater than or equal to zero. So no
sequence in [0, ) can converge to a point outside of [0, ). Hence, [0, ) is closed.
(c) The interval [0, 1] = r R ; 0 r 1 is closed.
(d) In general, if < a < b < , then (as the name suggests):
The closed interval [a, b] = r R ; a r b is closed.
The open interval (a, b) = r R ; a < r < b is not closed. (a and b are the limits
of sequences in (a, b), but a and b are not themselves in (a, b).)
The left half-open interval (a, b] = r R ; a < r b is not closed. (a is a limit of
a sequence in (a, b).)
The right half-open interval [a, b) = r R ; a r < b is not closed. (b is a limit
of a sequence in (a, b).)
(e) The set Q of rational numbers is not closed. To see this, observe that the sequence (A.1)
above can be rewritten:
p
1
=
314
100
; p
2
=
31416
10000
; p
3
=
3141593
1000000
; . . . . . .
Thus, p
1
, p
2
, p
3
, . . . a sequence of rational numbers, but converges to , which is irra-
tional. Hence Q is not closed.
The closure of a set U in R is the smallest closed subset of R which contains U.
Example 266:
(a) The closure of (0, ) is [0, ).
(b) The closure of Q is all of R.
The collection of all closed subsets of R has the following properties:
Lemma 267
(a) The empty set is closed.
(b) The set R of all real numbers is closed.
(c) If C
1
and C
2
are two closed sets, then their union C
1
C
2
is also closed.
(d) If C
1
, C
2
, C
3
, . . . is any collection of closed sets, then the intersection

n=1
C
n
is closed
(even if it is empty see (a)). 2
It is these four properties which motivate the denition of an abstract topological space (see
Page 226.)
A.1. INTRODUCTION 225
N=(-oo, 0]
P=(0, oo)
u
1
u
2
u
3
u
4
0
R
Figure A.2: The subsets N = (, 0] and P = (0, ) are disjoint, but there is a sequence in
P converging to a point in N.
x
2
x
1 x
3
x
4
x
5
x
oo
y
2
y
1
y
3
y
4
y
5
y
oo
f f f f f f
Figure A.3: If x
1
, x
2
, x
3
, . . . is a sequence converging to x

, and y
n
= f(x
n
) for all n, then
the sequence y
1
, y
2
, y
3
, . . . must converge to y = f(x

).
Connectedness: Imagine cutting the real line into two halves: N = (, 0] and P =
(0, ). Clearly, R = N.P, and the two sets are disjoint. And yet, somehow they seem glued
together. The reason is that there is a sequence in P converging to a point in N. Specically,
the sequence u
1
, u
2
, u
3
, . . . dened by (A.2) above is a sequence in P which converges to 0, a
point in N.
Because of this, it seems unnatural to cut R into N and P. In doing so, we are disrupting
the topological structure of R, by separating sequences from their limits.
The property we are encountering here is connectedness. We say that the real number line
R is connected because
It is impossible to separate R into two disjoint sets A and B, such that R =
A. B, and so that no sequence in A converges to a point in B or vice versa.
We can express this more succinctly using the concept of closed sets:
It is impossible to separate R into two disjoint closed sets A and B, such that R = A. B.
On the other hand, consider the integers Z. Let N = . . . , 3, 2, 1, 0, and let P =
1, 2, 3, . . .. Then Z = N . P, and no sequence in N can converge to a point in P or vice
versa. To see this, note that N and P are separated from one another by the interval (0, 1).
No act of sequence convergence can leap across this chasm. Thus, Z is disconnected.
Continuity A function f : RR is continuous if it preserves the topological structure of
R. For example:
226 APPENDIX A. BACKGROUND: TOPOLOGY
(i) If x
1
, x
2
, x
3
, . . . is a sequence converging to x

, and y
n
= f(x
n
) for all n, then the sequence
y
1
, y
2
, y
3
, . . . must converge to y

= f(x

) (see Figure A.3). In other words,

lim
n
f(x
n
) = f
_
lim
n
x
n
_
.
(ii) If U R is a connected subset, then f(U) is also connected.
(iii) If U R is not closed, then f(U) is not closed either.
Thus, continuous functions are the homomorphisms of topological structure.
All three properties are contained within the formal denition of continuity. If f : RR
is any function, then f is continuous if, for any subset C R,
_
C is closed
_
=
_
f
1
(C) is also closed
_
. (A.3)
(Exercise 190 Suppose that the function f satises denition (A.3). Show that f satises properties
(i), (ii), and (iii).)
A.2 Abstract Topological Spaces
So far weve looked at the topology of R. Now well develop a mathematical abstraction of
these properties. A topological space is a set X, along with a collection of subsets C of X
(called closed subsets), satisfying the following Closed Set Axioms:
(C1) The empty set is in C.
(C2) The set X is in C.
(C3) If subsets C
1
and C
2
are in C, then their union C
1
C
2
is also in C.
(C4) If K C is any collection of closed sets, then the intersection

KK
K is also in C.
Observe that these properties simply recapitulate the properties of Lemma 267 (where, in
(d), we set K = C
1
, C
2
, C
3
, . . . ...).
We extend the concepts of convergence, closure, connectedness, and continuity to this ab-
stract setting:
Closure: If U X is any subset of X, then the closure of U (denoted U) is the smallest
closed subset of X containing U. Formally:
U =

UCX
CC
C.
Exercise 191 1. Check that U is closed. (Hint: Use property (d).)
2. Show that, if C is any closed set and U C, then U C.
A.2. ABSTRACT TOPOLOGICAL SPACES 227
U
X
C
1
C
2
X
(A) (B)
Figure A.4: U is disconnected if there are two disjoint closed subsets C
1
and C
2
in X so that
U C
1
. C
2
.
Convergence: Let x
1
, x
2
, x
3
. . . be a sequence of points in X, and let x

X be some other
point. We say that the sequence x
1
, x
2
, x
3
. . . converges to x

if the closure of the set

x
1
, x
2
, x
3
. . . is exactly the set x
1
, x
2
, x
3
. . . , x

. Formally:
_
lim
n
x
n
= x

_

_
x
1
, x
2
, x
3
. . . = x
1
, x
2
, x
3
. . . , x

_
.
Connectedness: Let U X be any subset (Figure A.4A). We say U is disconnected if
there are two disjoint closed subsets C
1
and C
2
in X so that U C
1
.C
2
(Figure A.4B).
We say U is connected i it is not disconnected.
Continuity: Let X and Y be topological spaces, and let f : XY. We say that f is
continuous if, for any C Y,
_
C is closed in Y
_
=
_
f
1
(C) is closed in X
_
.
We say that f is a homeomorphism if:
1. f is continuous.
2. If f is a bijection from X to Y. (Thus, f has an inverse map f
1
: YX).
3. f
1
is also continuous.
We then say that X and Y are homeomorphic.
Continuous maps are the homomorphisms of topological spaces, and homeomorphisms are
the isomorphisms. If X and Y are homeomorphic, then, for topological purposes, they are
identical.
Example 268: Let X =
_

2
,

2
_
, and let Y = R. As shown in Figure A.5, let f : XY be
the tangent function: f(x) = tan(x). Then f is continous and bijective, and f
1
= arctan is
also continuous. (Exercise 192)
Hence, the interval
_

2
,

2
_
is homeomorphic to R.
228 APPENDIX A. BACKGROUND: TOPOLOGY
/2
/2
tan(x)
/2
/2
arctan(x)
Figure A.5: tan :
_

2
,

2
_
R is a homeomorphism, with inverse homeomorphism arctan :
R :
_

2
,

2
_
.
A.3 Open sets
An open set is the complement of a closed set. In other words, if X is a topological space,
and U X, then
_
U is open
_

_
X U is closed
_
.
Example 269:
(a) The set U = (, 0) is open in R, because R U = [0, ) is closed.
(b) The interval U = (0, 1) is open in R, because R U = (, 0] . [1, ) is closed.
An equivalent denition of openness is:
_
U is open
_

_
No point in U is a limit of a sequence in X U
_
. (Exercise 193)
Note: It is not true that U is open if U is not closed. Most subsets are neither open nor closed.
Example 270:
(a) [0, 1) is neither open nor closed in R.
(b) Q is neither open nor closed in R.
Open subsets satisfy properties which are dual to properties (C1) to (C4) of closed sets:
(O1) The empty set is open.
(O2) The set X is open.
A.4. COMPACTNESS 229
(O3) If subsets O
1
and O
2
are open, then their union O
1
O
2
is also open.
(O4) If O is any collection of closed sets, then the union
_
OO
O is also open
Exercise 194 Prove (O1) to (O4), by applying de Morgans laws to the Closed Set Axioms
(C1) to (C4).
Remark: The complement of every open set is closed, and vice versa. Thus the collection of
all open sets in a space completely determines the collection of all closed sets, and vice versa.
For this reason, we can dene a topological space just as easily by specifying which sets are
open, rather than by specifying which sets are closed (as weve done above). Indeed, in most
texts, a topological space is dened to be a set X and a collection of open sets obeying axioms
(O1) to (O4).
A.4 Compactness
Consider the subsets [0, ) and [0, 1] in R. Both sets are closed, but there is a dierence
between them: [0, 1] is bounded, and this endows it with fundamentally dierent topological
properties. We say that [0, 1] is compact.
Formally, a subset K R is compact if:
1. K is closed.
2. K is bounded (meaning that there is some M > 0 so that [k[ < M for all k K).
Example 271:
(a) If < a < b < , then the closed interval [a, b] is compact.
(b) Any nite union of nite closed intervals is compact. In other words, if < a
1
< b
1
<
a
2
< b
2
< . . . < a
n
< b
n
< , then the set [a
1
, b
1
] . [a
2
, b
2
] . . . . . [a
n
, b
n
] is compact.
The concept of compactness generalizes to abstract topological spaces. A topological space
X is compact if it has any of the following three logically equivalent properties:
1. The Finite Subcover Property.
2. The Chinese Box Property.
3. The Cluster Point Property.
230 APPENDIX A. BACKGROUND: TOPOLOGY
0 1
0
1
1/2
0
1
1/3 2/3
0
1
1/4 1/2
0 1/5 2/5
1
X=(0,1)
O
2
O
3
O
4
O
5
Figure A.6: An open cover of (0, 1) which has no nite subcover.
Finite Subcover Property: Let O be a collection of open sets which covers X, in the sense
that
_
OO
O = X. Then there is a nite subcollection O
1
, O
2
, . . . , O
N
O which also covers
X that is, O
1
O
2
. . . O
N
= X.
This is usually expressed by the slogan, Every open cover has a nite subcover.
Example 272: Let X = (0, 1). For all n 2, 3, 4, . . ., let O
n
=
_
1
n
,
2
n
_
, as in Figure A.6.
Let O = O
2
, O
3
, O
4
, . . .. Then
_
OO
O =

_
n=2
O
n
= (0, 1).
Hence, O is an open cover of (0, 1). However, no nite subcollection of O covers (0, 1). Thus,
(0, 1) is not compact.
(Weak) Chinese Box Property: Suppose C
1
C
2
C
3
is a descending sequence
of nonempty closed subsets of X, as in Figure A.7. Then

n=1
C
n
is also nonempty.
Example 273:
(a) Let X = [1, 1], and for all n N, let C
n
=
_
1
n
,
1
n

, as in Figure A.8. Let C =

C
1
, C
2
, C
3
, . . .. If C
m
1
, C
m
2
, . . . , C
m
k
C is any nite subset, then
C
m
1
C
m
2
. . . C
m
k
=
_
1
M
,
1
M
_
, where M = maxm
1
, m
2
, . . . , m
K
.
Thus, all nite intersections are nonempty, and thus, since [0, 1] is compact, we expect
that the innite intersection

n=1
C
n
is also nonempty. Indeed,

n=1
C
n
= 0.
A.4. COMPACTNESS 231
C
1
C
2 C
3
C
4
C
5
C
6
C
7
C
8
C
1
C
1
C 1 C1 C1 C1 C1C1
X
Figure A.7: The Chinese Box property.
-1
1
-1 1/2 -1/2
C
2
1
-1 1
1/3 -1/3
-1 1 1/4 -1/4
C
3
C
4
X=[-1,1]= C
1
Figure A.8: The Chinese Box Property in the unit interval.
232 APPENDIX A. BACKGROUND: TOPOLOGY
X=R
C
1
C
2
C
3
C
4
C
5
Figure A.9: No Chinese Box Property for the real line.
(b) Let X = R, and for all n N, let C
n
= [n, ), as in Figure A.9. Let C = C
1
, C
2
, C
3
, . . ..
If C
m
1
, C
m
2
, . . . , C
m
k
C is any nite subset, then
C
m
1
C
m
2
. . . C
m
k
= [M, ), where M = maxm
1
, m
2
, . . . , m
K
.
Thus, all nite intersections are nonempty. But R is not compact:

n=1
C
n
= .
The weak Chinese Box Property is actually equivalent to the
(Strong) Chinese Box Property: Let C be a collection of closed sets. Suppose that, for
any nite subset C
1
, C
2
, . . . , C
N
C, the intersection C
1
C
2
. . . C
N
is nonempty. Then
the full intersection

CC
C is also nonempty.
Lemma 274 Let X be a topological space. Then
_
X has the Weak Chinese Box Property
_

_
X has the Strong Chinese Box Property
_
.
Proof: = Exercise 195 Hint: Let C = C
1
C
2
C
3
.
= Exercise 196 Hint: Use Zorns Lemma. If you dont know Zorns Lemma, forget about it.
2
When we speak of the Chinese Box Property, we mean the strong one.
Cluster Point Property: Let C X be any innite closed set. Then C contains a cluster
point that is, a point c C so that C c is not closed.
This is usually expressed by the slogan, Every innite set has a cluster point.
Example 275:
A.4. COMPACTNESS 233
(a) Let X = [0, 1], and let C =
1
2
,
1
3
,
1
4
,
1
5
, . . . , 0. Then C X is a closed subset, and has a
cluster point namely, 0. If we remove 0, then the set C 0 =
1
2
,
1
3
,
1
4
,
1
5
, . . . is no
longer closed.
(b) Let X = R, and let C = N = 1, 2, 3, 4, . . .. Then N is a closed subset, but it has no
cluster points, because we can remove any n N to get the set N n, which is still
closed. Hence, R is not compact.
Proposition 276 Let X be a topological space. Then the following are equivalent:
(a) X has the Open Cover Property.
(b) X has the Chinese Box Property.
(c) X has the Cluster Point Property.
Proof: (a) (b) Exercise 197 Hint: use de Morgans Laws.
(b) = (c) Suppose X has the Chinese Box Property, and let D X be an innite,
closed subset. We want to show that D has a cluster point.
Suppose not. Then for every d D, let C
d
= D d. Since d is not a cluster point of D,
the set C
d
is also closed. Let C = C
d
; d D. This is an innite collection of closed sets.
Furthermore, if d
1
, d
2
, . . . , d
n
D is any nite subset, then
C
d
1
C
d
2
. . . C
dn
= D d
1
, d
2
, . . . , d
n

is nonempty (because D is innite).

Thus, the (strong) Chinese Box Property implies that

dD
C
d
is also nonempty. But

dD
C
d
= D d ; d D = D D = .
Contradiction.
(c) = (b) Exercise 198 . 2
Further Reading: A very friendly introduction to topology is [5]. An excellent and very
complete text is [10].
234 APPENDIX A. BACKGROUND: TOPOLOGY
Bibliography
[1] Michael Artin. Algebra. Prentice-Hall, Englewood Clis, NJ, 1991.
[2] John B. Conway. A Course in Functional Analysis. Springer-Verlag, New York, second edition, 1990.
[3] Robin Hartshorne. Algebraic Geometry, volume 52 of GTM. Springer-Verlag, New York, 1977.
[4] Nathan Jacobson. Basic Algebra II. W.H. Freeman and Co., New York, second edition, 1989.
[5] John G. Hocking and Gail S. Young. Topology. Dover, New York, 1961.
[6] Serge Lang. Algebra. Addison-Wesley, Reading, MA, 2nd edition, 1984.
[7] Oscar Zariski and P. Samuel. Commutative Algebra, volume I & II. van Nostrand, Princeton, NJ, 1958,1960.
[8] I. R. Shavarevich. Basic Algebraic Geometry, volume 213 of Grundlehren. Springer-Verlag, Heidelberg,
1974.
[9] Frank M. Warner. Foundations of Dierentiable Manifolds and Lie Groups. Springer-Verlag, New York,
1983.
[10] Stephen Willard. General Topology. Addison-Wesley, London, 1970.
235