0% found this document useful (0 votes)

909 views448 pages

Introductory Algebraic Number Theory

Uploaded by

adam boussif

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

0% found this document useful (0 votes)

909 views448 pages

Introductory Algebraic Number Theory

Uploaded by

adam boussif

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

You are on page 1/ 448

CB609-driver CB609/Alaca & Williams August 27, 2003 17:1 Char Count= 0

This page intentionally left blank

ii
CB609-driver CB609/Alaca & Williams August 27, 2003 17:1 Char Count= 0

INTRODUCTORY ALGEBRAIC NUMBER THEORY

Algebraic number theory is a subject that came into being through the attempts of mathe-
maticians to try to prove Fermat’s last theorem and that now has a wealth of applications
to Diophantine equations, cryptography, factoring, primality testing, and public-key cryp-
tosystems.
This book provides an introduction to the subject suitable for senior under-
graduate and beginning graduate students in mathematics. The material is presented in
a straightforward, clear, and elementary fashion, and the approach is hands on, with an
explicit computational flavor. Prerequisites are kept to a minimum, and numerous examples
illustrating the material occur throughout the text. References to suggested readings and to
the biographies of mathematicians who have contributed to the development of algebraic
number theory are given at the end of each chapter. There are more than 320 exercises, an
extensive index, and helpful location guides to theorems and lemmas in the text.

Şaban Alaca is Lecturer in Mathematics at Carleton University, where he has been honored by three
teaching awards: Faculty of Science Teaching Award, Professional Achievement Award, and Students
Choice Award. His main research interest is in algebraic number theory.

Kenneth S. Williams is Professor Emeritus and Distinguished Research Professor of Mathematics at

Carleton University. Dr. Williams has published more than 240 research papers in number theory,
linear algebra, algebra, and analysis. This is his seventh book.

i
CB609-driver CB609/Alaca & Williams August 27, 2003 17:1 Char Count= 0

ii
CB609-driver CB609/Alaca & Williams August 27, 2003 17:1 Char Count= 0

INTRODUCTORY ALGEBRAIC NUMBER

THEORY

ŞABAN ALACA
Carleton University, Ottawa

KENNETH S. WILLIAMS
Carleton University, Ottawa

iii
cambridge university press
Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore, São Paulo

Cambridge University Press

The Edinburgh Building, Cambridge cb2 2ru, UK
Published in the United States of America by Cambridge University Press, New York
www.cambridge.org
Information on this title: www.cambridge.org/9780521832502

© Saban Alaca and Kenneth S. Williams 2004

This publication is in copyright. Subject to statutory exception and to the provision of

relevant collective licensing agreements, no reproduction of any part may take place
without the written permission of Cambridge University Press.

First published in print format 2004

isbn-13 978-0-511-16494-1 eBook (EBL)

isbn-10 0-511-16494-7 eBook (EBL)

isbn-13 978-0-521-83250-2 hardback

isbn-10 0-521-83250-0 hardback

isbn-13 978-0-521-54011-7 paperback

isbn-10 0-521-54011-9 paperback

Cambridge University Press has no responsibility for the persistence or accuracy of urls
for external or third-party internet websites referred to in this publication, and does not
guarantee that any content on such websites is, or will remain, accurate or appropriate.
CB609-driver CB609/Alaca & Williams August 27, 2003 17:1 Char Count= 0

To our wives
Ayşe and Carole

v
CB609-driver CB609/Alaca & Williams August 27, 2003 17:1 Char Count= 0

vi
CB609-driver CB609/Alaca & Williams August 27, 2003 17:1 Char Count= 0

Contents

List of Tables page xi

Notation xiii
Introduction xv
1 Integral Domains 1
1.1 Integral Domains 1
1.2 Irreducibles and Primes 5
1.3 Ideals 8
1.4 Principal Ideal Domains 10
1.5 Maximal Ideals and Prime Ideals 16
1.6 Sums and Products of Ideals 21
Exercises 23
Suggested Reading 25
Biographies 25
2 Euclidean Domains 27
2.1 Euclidean Domains 27
2.2 Examples of Euclidean Domains 30
2.3 Examples of Domains That are Not Euclidean 37
2.4 Almost Euclidean Domains 46
2.5 Representing Primes by Binary Quadratic Forms 47
Exercises 49
Suggested Reading 51
Biographies 53
3 Noetherian Domains 54
3.1 Noetherian Domains 54
3.2 Factorization Domains 57
3.3 Unique Factorization Domains 60
3.4 Modules 64
3.5 Noetherian Modules 67
Exercises 71

vii
CB609-driver CB609/Alaca & Williams August 27, 2003 17:1 Char Count= 0

viii Contents

Suggested Reading 72
Biographies 73
4 Elements Integral over a Domain 74
4.1 Elements Integral over a Domain 74
4.2 Integral Closure 81
Exercises 86
Suggested Reading 87
Biographies 87
5 Algebraic Extensions of a Field 88
5.1 Minimal Polynomial of an Element Algebraic over a Field 88
5.2 Conjugates of α over K 90
5.3 Conjugates of an Algebraic Integer 91
5.4 Algebraic Integers in a Quadratic Field 94
5.5 Simple Extensions 98
5.6 Multiple Extensions 102
Exercises 106
Suggested Reading 108
Biographies 108
6 Algebraic Number Fields 109
6.1 Algebraic Number Fields 109
6.2 Conjugate Fields of an Algebraic Number Field 112
6.3 The Field Polynomial of an Element of an Algebraic Number
Field 116
6.4 The Discriminant of a Set of Elements in an Algebraic Number
Field 123
6.5 Basis of an Ideal 129
6.6 Prime Ideals in Rings of Integers 137
Exercises 138
Suggested Reading 140
Biographies 140
7 Integral Bases 141
7.1 Integral Basis of an Algebraic Number Field 141
7.2 Minimal Integers 160
7.3 Some Integral Bases in Cubic Fields 170
7.4 Index and Minimal Index of an Algebraic Number Field 178
7.5 Integral Basis of a Cyclotomic Field 186
Exercises 189
Suggested Reading 191
Biographies 193
8 Dedekind Domains 194
8.1 Dedekind Domains 194
8.2 Ideals in a Dedekind Domain 195
CB609-driver CB609/Alaca & Williams August 27, 2003 17:1 Char Count= 0

Contents ix

8.3 Factorization into Prime Ideals 200

8.4 Order of an Ideal with Respect to a Prime Ideal 206
8.5 Generators of Ideals in a Dedekind Domain 215
Exercises 216
Suggested Reading 217
9 Norms of Ideals 218
9.1 Norm of an Integral Ideal 218
9.2 Norm and Trace of an Element 222
9.3 Norm of a Product of Ideals 228
9.4 Norm of a Fractional Ideal 231
Exercises 233
Suggested Reading 234
Biographies 235
10 Factoring Primes in a Number Field 236
10.1 Norm of a Prime Ideal 236
10.2 Factoring Primes in a Quadratic Field 241
10.3 Factoring Primes in a Monogenic Number Field 249
10.4 Some Factorizations in Cubic Fields 253
10.5 Factoring Primes in an Arbitrary Number Field 257
10.6 Factoring Primes in a Cyclotomic Field 260
Exercises 261
Suggested Reading 262
11 Units in Real Quadratic Fields
√ 264
11.1 The Units of Z + Z 2 264
11.2 The Equation x 2 − my 2 = 1 267
11.3 Units of Norm 1 271
11.4 Units of Norm −1 275
11.5 The Fundamental Unit 278
11.6 Calculating the Fundamental Unit 286
11.7 The Equation x 2 − my 2 = N 294
Exercises 297
Suggested Reading 298
Biographies 298
12 The Ideal Class Group 299
12.1 Ideal Class Group 299
12.2 Minkowski’s Translate Theorem 300
12.3 Minkowski’s Convex Body Theorem 305
12.4 Minkowski’s Linear Forms Theorem 306
12.5 Finiteness of the Ideal Class Group 311
12.6 Algorithm to Determine the Ideal Class Group 314
12.7 Applications to Binary Quadratic Forms 331
Exercises 341
CB609-driver CB609/Alaca & Williams August 27, 2003 17:1 Char Count= 0

x Contents

It is planned to provide solutions to selected questions, as well as corrections to

any errors, on the website
http://mathstat.carleton.ca/˜williams/books.html
or
http://www.math.carleton.ca/˜williams/books.html.

The authors would like to thank their colleagues John D. Dixon, James G. Huard,
Pierre Kaplan, Blair K. Spearman, and P. Gary Walsh for helpful suggestions in
connection with the writing of this book. The second author would like to thank the
many students who have taken the course Mathematics 70.436*/70.536 Algebraic
Number Theory with him at Carleton University over the years. Special thanks go
CB609-driver CB609/Alaca & Williams August 27, 2003 17:1 Char Count= 0

Introduction xvii

to the class of 2000–1 (Yaroslav Bezverkhnyev, Joanne Charlebois, Colette Haley,

Mathieu Lemire, Rima Rahal, Fabien Roche, Tom Wiley, and Benjamin Young) for
their suggestions for improvement to the preliminary draft of this book used in class.
Finally, the authors would like to thank Austin Behne for his help in translating van
der Waerden’s paper on Dirichlet’s unit theorem from German into English.
CB609-driver CB609/Alaca & Williams August 27, 2003 17:1 Char Count= 0

xviii
CB609-01 CB609/Alaca & Williams August 7, 2003 17:16 Char Count= 0

1
Integral Domains

1.1 Integral Domains

In this chapter we recall the definition and properties of an integral domain and
develop the concept of divisibility in such a domain. We expect the reader to be
familiar with the elementary properties of groups, rings, and fields and to have a
basic knowledge of both elementary number theory and linear algebra over a field.

Definition 1.1.1 (Integral domain) An integral domain is a commutative ring that

has a multiplicative identity but no divisors of zero.
An integral domain D is called a field if for each a ∈ D, a = 0, there exists
b ∈ D with ab = 1.

Example 1.1.1 The ring Z = {0, ±1, ±2, . . .} of all integers is an integral domain.

Example 1.1.2 Z + Zi = {a + bi | a, b ∈ Z} is an integral domain. The ele-

ments of Z + Zi are called Gaussian integers after the famous mathematician Carl
Friedrich Gauss (1777–1855), who developed their properties in his work on bi-
quadratic reciprocity. Z + Zi is called the Gaussian domain.

Example 1.1.3 Z + Zω = {a +√ bω | a, b ∈ Z}, where ω is the complex cube

root of unity given by ω = (−1 + −3)/2, is an integral domain. The elements of
Z + Zω are called Eisenstein integers after Gotthold Eisenstein (1823–1852), who
introduced them in his pioneering work on the law of cubic reciprocity. Z + Zω is
called the
√ Eisenstein domain. The other complex cube root of unity is ω2 = ω =
(−1 − −3)/2.
√
Note that Z + Zω = Z + Zω as ω2 = −ω − 1. Also Z + Zω =
2

Z + Z 1+ 2 −3 .

√ √
Example 1.1.4 Z + Z m = {a + b m | a, b ∈ Z}, where m is a positive or
√
negative integer that is not a perfect square, is an integral domain. As m is a
√
root of an irreducible quadratic polynomial (namely x 2 − m), Z + Z m is called

1
CB609-01 CB609/Alaca & Williams August 7, 2003 17:16 Char Count= 0

2 Integral Domains

a quadratic domain. If k is a nonzero integer such that k 2 divides m then

√
Z + Z m ⊆ Z + Z m/k 2
√
equality if and only if k = 1. Z + Z m is called a subdomain of Z +
2
with
Z m/k 2 . Thus Z + 2Zi ⊂ Z + Zi.
√ √
Example 1.1.5 Z + Z 1+2 m = {a + b 1+2 m | a, b ∈ Z}, where m is a non-
square integer (positive or negative), which is √congruent
to 1 modulo 4, is an
1+ m
integral domain. We emphasize that Z + Z 2
is not an integral domain if
m ≡ 1 (mod 4) since in this case it is not closed under multiplication as
√ √ √ √
1+ m 1+ m 1+ m 1− m 1−m
1− = = ∈ Z.
2 2 2 2 4
√
Again as 1+2 m is a root of an irreducible quadratic polynomial (namely x 2 − x +
1−m √
), Z + Z 1+2 m is called a quadratic domain. We note that the elements of
4
√ √
the integral domain Z + Z 1+2 m can also be written in the form 12 (x + y m),
√
where x and y are integerssuch√
that x ≡ y (mod 2). Clearly the domain Z + Z m
is a subdomain of Z + Z 1+2 m .

Example 1.1.6 F[x] = the ring of polynomials in the indeterminate x with coef-
ficients from a field F is an integral domain.

Example 1.1.7 Z[x] = the ring of polynomials in the indeterminate x with integral
coefficients is an integral domain.

Example 1.1.8 D[x] = the ring of polynomials in the indeterminate x with coef-
ficients from the integral domain D is an integral domain.

Example 1.1.9 F[x, y] = the ring of polynomials in the two indeterminates x and
y with coefficients from the field F is an integral domain.

Example 1.1.10 Z + Zθ + Zθ 2 = {a + bθ + cθ 2 | a, b, c ∈ Z}, where θ is a root

of the cubic equation θ 3 + θ + 1 = 0, is an integral domain. It is called a cubic
domain.
√ √
Example 1.1.11 D = {a + b 2 + ci + di 2 | a, c integers; b, d√both integers
√ integers} is an integral domain. Clearly Z + Z 2 ⊂ D, Z +
or both halves of odd
Zi ⊂ D, Z + Zi 2 ⊂ D.
CB609-01 CB609/Alaca & Williams August 7, 2003 17:16 Char Count= 0

1.1 Integral Domains 3

Properties of an Integral Domain

Let D be an integral domain. Then the following properties hold.
(a) The identity element of D is unique, for if 1 and 1 are two identities for D then

1 = 1 · 1 (as 1 is an identity) = 1 (as 1 is an identity).

(b) D possesses a left cancellation law, that is,

ab = ac, a = 0 =⇒ b = c (a, b, c ∈ D)

as well as a right cancellation law

ac = bc, c = 0 =⇒ a = b (a, b, c ∈ D).

(c) It is well known that if D is an integral domain then there exists a field F, called the
field of quotients of D or the quotient field of D, that contains an isomorphic copy D
of D (see, for example, Fraleigh [3]). In practice it is usual to identify D with D and so
consider D as a subdomain of F. The quotient field of Z is the field of rational numbers
Q. The quotient field of the polynomial domain F[X ] (where F is a field) is the field
F(X ) of rational functions in X .

Definition 1.1.2 (Divisor) Let a and b belong to the integral domain D. The element
a is said to be a divisor of b (or a divides b) if there exists an element c of D such
that b = ac. If a is a divisor of b, we write a | b. If a is not a divisor of b, we write
a b.

Example 1.1.12 1 + i | 2 in Z + Zi as 2 = (1 + i)(1 − i).

Example 1.1.13 x 2 + x + 1 | x 4 + x 2 + 1 in Z[x] as x 4 + x 2 + 1 = (x 2 + x + 1)

(x 2 − x + 1).

Example 1.1.14 (1 − ω)2 | 3 in Z + Zω as 3 = (1 − ω)2 (1 + ω) (see Example

1.1.3).

Example 1.1.15 1 + θ − θ 2 | − θ − 2θ 2 in Z + Zθ + Zθ 2 as −θ − 2θ 2 = (1 +
θ − θ 2 )(1 − θ) (see Example 1.1.10).
√ √ √ √ √
Example 1.1.16 2 + 2 3 in Z + Z 2 as 3/(2 + 2) = 3 − 32 2 ∈
Z + Z 2.

Properties of Divisors
Let a, b, c ∈ D, where D is an integral domain. Then the following properties hold.
(a) a | a (reflexive property).
(b) a | b and b | c implies a | c (transitive property).
CB609-01 CB609/Alaca & Williams August 7, 2003 17:16 Char Count= 0

4 Integral Domains

(c) a | b and a | c implies a | xb + yc for any x ∈ D and y ∈ D.

(d) a | b implies ac | bc.
(e) ac | bc and c = 0 implies a | b.
(f) 1 | a.
(g) a | 0.
(h) 0 | a implies a = 0.

Definition 1.1.3 (Unit) An element a of an integral domain D is called a unit if

a | 1. The set of units of D is denoted by U (D).

Properties of Units
Let D be an integral domain. Then U (D) has the following properties.
(a) ±1 ∈ U (D).
(b) If a ∈ U (D) then −a ∈ U (D).
(c) If a ∈ U (D) then a −1 ∈ U (D).
(d) If a ∈ U (D) and b ∈ U (D) then ab ∈ U (D).
(e) If a ∈ U (D) then ±a n ∈ U (D) for any n ∈ Z.

Example 1.1.17
(a) i ∈ U (Z + Zi).
(b) ω ∈ U (Z + Zω) (see Example 1.1.3).
(c) θ ∈ U (Z + Zθ + Zθ 2 ) as 1 = θ (−1 − θ 2 ) (see Example 1.1.10).

Theorem 1.1.1 If D is an integral domain then U (D) is an Abelian group with

respect to multiplication.

Proof: U (D) is closed under multiplication by property (d). Multiplication of el-

ements of U (D) is both associative and commutative as D is an integral domain.
U (D) possesses an identity element, namely 1, by property (a). Every element of
U (D) has a multiplicative inverse by property (c). Thus U (D) is an Abelian group
with respect to multiplication.

Abelian groups are named after the Norwegian mathematician Niels Henrik Abel
(1802–1829), who proved in 1824 the impossibility of solving the general quintic
equation by means of radicals.

Example 1.1.18 Let Zn denote the cyclic group of order n.

(a) U (Z) = {±1} Z2 .
(b) U (Z + Zi) = {±1, ±i} Z4 .
(c) U (F[x]) = F ∗ , where F is a field and F ∗ = F \ {0}.
CB609-01 CB609/Alaca & Williams August 7, 2003 17:16 Char Count= 0

1.2 Irreducibles and Primes 5

√= n{±1} Z2 . √
(d) U (Z[x])
(e) ±(1√ + 2)√ ∈ U (Z + Z 2), for all n ∈ Z.
(f) 12 2 + 12 i 2 ∈ U (D), where D is defined in Example 1.1.11.

We remark that in Chapter 11 we will show that

√ √
U (Z + Z 2) = {±(1 + 2)n | n ∈ Z} Z2 × Z.

Definition 1.1.4 (Associate) Two nonzero elements a and b of an integral domain

D are called associates, or said to be associated, if each divides the other. If a and
b are associates we write a ∼ b. If a and b are not associates we write a ∼ b.

Properties of Associates
Let a, b, c ∈ D ∗ = D \ {0}, where D is an integral domain. The following proper-
ties hold.
(a) a ∼ a (reflexive property).
(b) a ∼ b implies b ∼ a (symmetric property).
(c) a ∼ b and b ∼ c imply a ∼ c (transitive property).
(d) a ∼ b if and only if ab−1 ∈ U (D).
(e) a ∼ 1 if and only if a is a unit.

Properties (a), (b), and (c) show that ∼ is an equivalence relation. The equivalence
class containing a ∈ D is just the set {ua | u ∈ U (D)}.

Example 1.1.19
(a) In Z, a ∼ b if and only if a = ±b, equivalently |a| = |b|.
(b) In Z + Zi we have 1 + i ∼ 1 − i as 1+i = i ∈ U (Z + Zi).
√ √ 1−i
√ √ √ √
(c) In Z + Z 2 we have 1 + 3 2 ∼ 5 − 2 2 as 1+3 √2 = 1 +
5−2 2
2 ∈ U (Z + Z 2).

1.2 Irreducibles and Primes

In Z an integer p (≥ 2) that is divisible only by the positive integers 1 and p is
called a prime. Each prime p in Z has the following two properties:
p = ab (a, b ∈ Z) =⇒ a or b = ±1 (1.2.1)
and
p | ab (a, b ∈ Z) =⇒ p | a or p | b. (1.2.2)
Our next definition generalizes property (1.2.1) to an arbitrary integral domain D,
and an element of D with this property is called an irreducible element.
CB609-01 CB609/Alaca & Williams August 7, 2003 17:16 Char Count= 0

6 Integral Domains

Definition 1.2.1 (Irreducible) A nonzero, nonunit element a of an integral domain

D is called an irreducible, or said to be irreducible, if a = bc, where b, c ∈ D,
implies that either b or c is a unit.
A nonzero, nonunit element that is not irreducible is called reducible.

Example 1.2.1 2 is irreducible in Z, for if 2 = ab with a ∈ Z and b ∈ Z then

either a = ±1 or b = ±1.
√
Example√ 1.2.2 2 √ is irreducible in Z + Z −5. To show this, suppose that 2 =
(a + b −5)(c + d −5), where a, b, c, d ∈ Z. Taking the modulus of both sides
of this equation, we obtain 4 = (a 2 + 5b2 )(c2 + 5d 2 ). Thus a 2 + 5b2 is a positive
integral divisor of 4 and so we must have

a 2 + 5b2 = 1, 2, or 4.

Hence we see that

(a, b) = (±1, 0) or (±2, 0)

so that
√
a + b −5 = ±1 or ± 2.
√ √
In the former case a + b −5 is a unit of Z + Z −5. In the latter case
√ 2 2
c + d −5 = √ = = ±1
a + b −5 ±2
√ √
is a unit of Z + Z −5. Hence 2 is irreducible in Z + Z −5.
√ √
Example 1.2.3 7 + −5 is reducible in Z + Z −5 because
√ √ √
7 + −5 = (1 + −5)(2 − −5)
√ √ √
and neither 1 + −5 nor 2 − −5 is a unit of Z + Z −5.

Our next definition generalizes property (1.2.2) to an arbitrary integral domain, and
an element with this property is called a prime element.

Definition 1.2.2 (Prime) A nonzero, nonunit element p of an integral domain D

is called a prime if p | ab, where a, b ∈ D, implies that p | a or p | b.

Example 1.2.4 2 is a prime in Z. Suppose 2 | ab, where a, b ∈ Z, so that ab is

even. Since the product of two odd integers is odd, at least one of a and b must be
even, that is, 2 | a or 2 | b, showing that 2 is prime.
CB609-01 CB609/Alaca & Williams August 7, 2003 17:16 Char Count= 0

1.2 Irreducibles and Primes 7

√ √ √
Example√ 1.2.5 2 is not a prime in Z + Z −5 as 2 | (1 + −5)(1 − −5) yet
2 1 ± −5.

Example 1.2.6 1 + i is a prime in Z + Zi. To show this, suppose that 1 + i |

(a + bi)(c + di), where a, b, c, d ∈ Z. Then there exist integers x and y such that

(a + bi)(c + di) = (1 + i)(x + yi).

Taking the modulus of both sides of this equation, we obtain

(a 2 + b2 )(c2 + d 2 ) = 2(x 2 + y 2 ).

As 2 is a prime in Z, we have either 2 | a 2 + b2 or 2 | c2 + d 2 . Interchanging a + bi

and c + di, if necessary, we may suppose that 2 | a 2 + b2 . Thus, either a and b are
both even or they are both odd. In the former case a = 2r and b = 2s, where r and
s are integers, and

a + bi = 2(r + si) = (1 + i)((r + s) + (−r + s)i),

so that 1 + i | a + bi. In the latter case a = 2r + 1 and b = 2s + 1, where r and

s are integers, and

a + bi = 2(r + si) + (1 + i) = (1 + i)((r + s + 1) + (−r + s)i),

so that 1 + i | a + bi. Hence 1 + i is a prime in Z + Zi.

Theorem 1.2.1 In any integral domain D a prime is irreducible.

Proof: Let p ∈ D be a prime and suppose that p = ab, where a, b ∈ D. As ab =

p · 1 we have p | ab, and so, as p is prime, we deduce that p | a or p | b, that is,
a/ p ∈ D or b/ p ∈ D. Since 1 = a/ p · b or 1 = a · b/ p, either b is a unit or a is a
unit of D. This proves that p is an irreducible element of D.

The converse of Theorem 1.2.1

√ is not true. From Examples 1.2.2 and 1.2.5 we
see that the element 2 of Z + Z −5 is irreducible but not prime.
Waterhouse [6] has recently given a class of integral domains in which every
irreducible is prime.

Theorem 1.2.2 Let D be an integral domain that has the following property:

Every quadratic polynomial in D[X ] having roots in the quotient

field F of D is a product of linear polynomials in D[X ]. (1.2.3)

Then every irreducible in D is prime.

CB609-01 CB609/Alaca & Williams August 7, 2003 17:16 Char Count= 0

8 Integral Domains

Proof: Let p be an irreducible element in D, which is not prime. Then there exist
a, b ∈ D such that

p | ab, p a, p b.

Let r = ab/ p ∈ D, and consider the quadratic polynomial

f (X ) = p X 2 − (a + b)X + r.

In F[X ] we have

f (X ) = p(X − a/ p)(X − b/ p).

We show that f (X ) does not factor into linear factors in D[X ]. Indeed, suppose on
the contrary that

f (X ) = (cX + s)(d X + t)

in D[X ]. Then cd = p. As p is irreducible, one of c and d is a unit of D, say d, so

that c = d −1 p. Then the roots of f (X ) in F are −ds/ p and −d −1 t. But −d −1 t ∈ D,
while neither a/ p nor b/ p is in D. Thus no such factorization can exist. Hence
every irreducible in D is prime.

1.3 Ideals
Subsets of an integral domain D that are closed under addition and under multipli-
cation by elements of D play a special role and are called ideals.

Definition 1.3.1 (Ideal) An ideal I of an integral domain D is a nonempty subset

of D having the following two properties:

a ∈ I, b ∈ I =⇒ a + b ∈ I,

a ∈ I, r ∈ D =⇒ ra ∈ I.

It is clear that if a1 , . . . , an ∈ I then r1 a1 + · · · + rn an ∈ I for all r1 , . . . , rn ∈ D.

In particular if a ∈ I and b ∈ I then −a ∈ I and a − b ∈ I . Also 0 ∈ I , and if 1 ∈ I
then I = D.

Example 1.3.1 If {a1 , . . . , an } is a set of elements of the integral domain D then

the set of all finite linear combinations of a1 , . . . , an
n

ri ai | r1 , . . . , rn ∈ D
i=1

is an ideal of D, which we denote by a1 , . . . , an .

CB609-01 CB609/Alaca & Williams August 7, 2003 17:16 Char Count= 0

1.3 Ideals 9

Definition 1.3.2 (Principal ideal) An ideal I of an integral domain D is called a

principal ideal if there exists an element a ∈ I such that I = a. The element a is
called a generator of the ideal I .

If D is an integral domain the principal ideal a generated by a ∈ D is just the

set {ra | r ∈ D}. Clearly the principal ideal 0 is just the singleton set {0} and the
principal ideal 1 is D.

Definition 1.3.3 (Proper ideal) An ideal I of an integral domain D is called a

proper ideal of D if I = 0, 1.

Thus a proper ideal of an integral domain D is an ideal I such that {0} ⊂ I ⊂ D.

Example 1.3.2 For any positive integer k, the set

kZ = {0, ±k, ±2k, . . .}
is an ideal of Z. Indeed kZ is a principal ideal generated by k (or −k) so that
kZ = k = −k.

Example 1.3.3 Let

I = { f (x) ∈ Z[x] | f (0) = 0}.
Then I is an ideal of Z[x] and I = x.

Example 1.3.4 Let

J = { f (x) ∈ Z[x] | f (0) ≡ 0 (mod 2)}.
Then J is an ideal of Z[x] and J = 2, x. However, J is not a principal ideal.

Theorem 1.3.1 Let D be an integral domain and let a, b ∈ D ∗ = D \ {0}. Then

a = b if and only if a/b ∈ U (D).

Proof: If a/b ∈ U (D) then a = bu for some u ∈ U (D). Let x ∈ a. Then x = ac
for some c ∈ D. Hence x = buc with uc ∈ D. Thus x ∈ b. We have shown that
a ⊆ b. As a/b ∈ U (D) and U (D) is a group with respect to multiplication, we
have b/a = (a/b)−1 ∈ U (D). Then, proceeding exactly as before with the roles of
a and b interchanged, we find that b ⊆ a. Thus a = b.
Conversely, suppose that a = b. Then a = bc for some c ∈ D and b = ad for
some d ∈ D. Hence b = bcd. As b = 0 we deduce that 1 = cd so that c ∈ U (D).
Thus a/b = c ∈ U (D).
CB609-01 CB609/Alaca & Williams August 7, 2003 17:16 Char Count= 0

10 Integral Domains

1.4 Principal Ideal Domains

An important class of integral domains are those in which every ideal is principal.

Definition 1.4.1 (Principal ideal domain) An integral domain D is called a prin-

cipal ideal domain if every ideal in D is principal.

We begin by giving an example of an integral domain in which every ideal is

principal.

Theorem 1.4.1 Z is a principal ideal domain.

Proof: Let I be an ideal of Z. If I = {0} then I = 0 is a principal ideal. Thus we

may suppose that I = {0}. Hence I contains a nonzero element a. As both a and
−a belong to I , we can suppose that a > 0. Hence I contains at least one positive
integer, namely a.
We let m denote the least positive integer in I . Dividing a by m, we obtain
integers q and r such that a = mq + r and 0 ≤ r < m. As a ∈ I and m ∈ I , we
have r = a − mq ∈ I . This contradicts the minimality of m unless r = 0, in which
case a = mq; that is, I = m = mZ.

Theorems 1.3.1 and 1.4.1 show that the set of ideals of Z is {kZ | k ∈
{0, 1, 2, . . .}}. Moreover, if I is an ideal of Z then it is generated by the least
positive integer in I .
Other examples of principal ideal domains will be given in Chapter 2 where we
discuss Euclidean domains.

Theorem 1.4.2 In a principal ideal domain, an irreducible element is prime.

Proof: Let p be an irreducible element in a principal ideal domain D. Suppose

that p | ab, where a, b ∈ D. If p a we let I be the ideal p, a of D. As D is a
principal ideal domain there is an element c ∈ D such that I = c. As a ∈ I and
p ∈ I we must have c | a and c | p. If c ∼ p then p | a, contradicting p a. Hence
c ∼ p, and as p is irreducible, c must be a unit. Thus there exists d ∈ D such that
cd = 1. Now c ∈ a, p so there exist x, y ∈ D such that c = xa + yp. Hence
1 = cd = d xa + dyp,
and so
b = (d x)ab + (bdy) p.
Since p | ab this shows that p | b. Thus p | a or p | b and p is a prime element of
D.
CB609-01 CB609/Alaca & Williams August 7, 2003 17:16 Char Count= 0

1.4 Principal Ideal Domains 11

Theorem 1.4.3 In a principal ideal domain, an element is irreducible if and only

if it is prime.

Proof: This follows immediately from Theorems 1.2.1 and 1.4.2.

Example √ 1.4.1 It was noted in Section 1.2 that 2 is irreducible but not √ prime
in Z + Z −5. Hence, by Theorem 1.4.3, the integral domain √ Z + Z −5
√ is
not a principal ideal domain. Indeed the ideal 2, 1 + −5 of Z + Z −5
is not principal. This can be shown √ directly as follows. Suppose,√on the
contrary, that the ideal√2, 1 + −5 is principal, that √ is, 2, 1 + −5 =
α for some√α ∈ Z + Z −5. Hence 2 ∈ α and 1 + −5 ∈ α √ so that α | 2
and α | 1 + −5. From the first of these, as 2 is irreducible
√ in Z + Z −5, it must
be the√ case that α √∼ 1 or α ∼ 2. √If α ∼ 2 then 2 | 1 + −5, which √ is impossible
1+ −5
as 2 = 2 + 2 −5 ∈ Z + Z −5. Hence α ∼ 1, and
1 1
√ so 2, 1 + −5 = 1.
This shows
√ that 1 is a linear combination of 2 and 1 + −5 with coefficients from
Z + Z −5; that is, there exist x, y, z, w ∈ Z such that
√ √ √
1 = (x + y −5)2 + (z + w −5)(1 + −5).
√
Equating coefficients of 1 and −5, we obtain
1 = 2x + z − 5w, 0 = 2y + z + w.
The difference of these equations yields
1 = 2(x − y − 3w),
which is clearly impossible as the left-hand side is an odd
√ integer and the right-
√ is an even integer. Hence the ideal 2, 1 + −5 is not principal in
hand side
Z + Z −5.

Definition 1.4.2 (Greatest common divisor) Let D be a principal ideal domain

and let {a1 , . . . , an } be a set of elements of D. Then the ideal a1 , . . . , an is a
principal ideal. A generator of this ideal is called a greatest common divisor of
a1 , . . . , an .

Let D be a principal ideal domain. If a and b are greatest common divisors of

a1 , . . . , an ∈ D then
a = a1 , . . . , an = b,
so that, by Theorem 1.3.1, a ∼ b. We write (a1 , . . . , an ) for a greatest com-
mon divisor of a1 , . . . , an , understanding that (a1 , . . . , an ) is only defined up to
a unit. We note that (a1 , . . . , an ) = 0 if a1 = · · · = an = 0. Also (a1 , . . . , an ) =
(a1 , . . . , an−1 ) if an = 0. Furthermore,
a ∈ a = a1 , . . . , an ,
CB609-01 CB609/Alaca & Williams August 7, 2003 17:16 Char Count= 0

12 Integral Domains

so that
a = r 1 a1 + · · · + r n an
for some r1 , . . . , rn ∈ D. Thus if c ∈ D is such that
c | a j ( j = 1, 2, . . . , n)
then
c | a.
Moreover, for j = 1, 2, . . . , n, we have
a j ∈ a1 , . . . , an = a
so that
a | aj.
This justifies calling a “a greatest common divisor” of a1 , . . . , an . The elements
a1 , . . . , an are called relatively prime if (a1 , . . . , an ) is a unit, that is,
a1 , . . . , an = 1 = D.
It is easy to verify that
(a1 , . . . , an−1 , an ) = ((a1 , . . . , an−1 ), an ),
so that a greatest common divisor can be obtained by finding a succession of greatest
common divisors of pairs of elements, that is, if (a1 , a2 ) = b then (a1 , a2 , a3 ) =
(b, a3 ), etc.
In the next theorem we use our knowledge of primes and irreducibles in a principal
ideal domain to give conditions under which a prime p can be expressed as u 2 − mv 2
or mv 2 − u 2 for some integers u and v, where m is a given nonsquare integer.
√
Theorem 1.4.4 Let m be a nonsquare integer such that Z + Z m is a principal
ideal domain. Let p be an odd prime for which the Legendre symbol

m
= 1.
p
Then there exist integers u and v such that
p = u 2 − mv 2 if m < 0, or if m > 0,
and there are integers T, U such that T 2 − mU 2 = −1,
p = u 2 − mv 2 or mv 2 − u 2 , if m > 0,
and there are no integers T, U with T 2 − mU 2 = −1.
CB609-01 CB609/Alaca & Williams August 7, 2003 17:16 Char Count= 0

1.4 Principal Ideal Domains 13

Proof: As m
p
= 1, there exists an integer x such that x 2 ≡ m (mod p). Thus
√ √
p | (x + m)(x −
m)
√ √ √ √
in Z + Z m. Clearly x±p m = xp ± 1p m ∈ Z + Z m so that
√
p x ± m.
√ √
Hence p is not a prime in Z + Z m. As Z + Z m is a principal ideal domain, by
√
Theorem 1.4.3 p is not irreducible in Z + Z m. Hence
√ √
p = (u + v m)(w + t m) (1.4.1)
√ √ √ √
for some u + v m ∈ Z + Z m and w + t m ∈ Z + Z m, where neither u +
√ √ √
v m nor w + t m is a unit in Z + Z m. From (1.4.1) we deduce that
√
p − (uw + tvm) = (ut + vw) m.
√
As m is not a square, m ∈ / Q, so that
p − (uw + tvm) = ut + vm = 0.
Then
p 2 = (uw + tvm)2 = (uw + tvm)2 − m(ut + vm)2
so that
p 2 = (u 2 − mv 2 )(w 2 − mt 2 ). (1.4.2)
As m, u, v, w, t ∈ Z and m ∈ N, we see that u 2 − mv 2 ∈ Z and w 2 − mt 2 ∈ Z.
√ √
Moreover, u 2 − mv 2 = ±1 and w 2 − mt 2 = ±1, as u + v m and w + t m are
√
not units in Z + Z m. Thus, from (1.4.2), as p is a prime, we must have ± p =
u 2 − mv 2 = w 2 − mt 2 . Hence there are integers u and v such that p = u 2 − mv 2
or −(u 2 − mv 2 ).
If m < 0 then u 2 − mv 2 > 0, so we must have p = u 2 − mv 2 .
If m > 0, p = −(u 2 − mv 2 ), and there exist integers T and U such that T 2 −
mU 2 = −1 then p = u 2 − mv 2 with u = T u + mU v, v = U u + T v.

√
In Chapter 2 we give some nonsquare values of m for which Z + Z m is a
principal ideal
domain.
Then, by Theorem 1.4.4, we know that for those odd primes
p for which p = 1 there are integers u and v such that p = u 2 − mv 2 or mv 2 −
m

u 2 . For a general positive integer m it is a difficult problem to decide which primes

are expressible as u 2 − mv 2 with u, v ∈ Z. The reader interested in knowing more
about this problem should consult Cox [2].
In the next theorem we give conditions that ensure that a prime p can
be expressed in the form u 2 + uv + 14 (1 − m)v 2 or −(u 2 + uv + 14 (1 − m)v 2 )
CB609-01 CB609/Alaca & Williams August 7, 2003 17:16 Char Count= 0

14 Integral Domains

for some integers u and v, where m is a given nonsquare integer with m ≡

1 (mod 4).

√ 1.4.5 Let m ≡ 1 (mod 4) be a nonsquare integer such that Z +

Theorem
Z 1+2 m is a principal ideal domain. Let p be an odd prime for which mp = 1.
Then there exist integers u and v such that
1
p = u 2 + uv + (1 − m)v 2 ,
4
if m < 0, or if m > 0, and there are integers T, U such that
1
T 2 + T U + (1 − m)U 2 = −1,
4
and
1 1
p = u 2 + uv + (1 − m)v 2 or − (u 2 + uv + (1 − m)v 2 ) if m > 0,
4 4
and there are no integers T, U with
1
T 2 + T U + (1 − m)U 2 = −1.
4

Proof: As m
p
= 1 there exists an integer z such that z 2 ≡ −m (mod p). Set

z, if z is odd,
y=
p − z, if z is even,

so that y isan odd integer satisfying

y 2 ≡ m (mod p). Now let x = 12 (y − 1) ∈ Z.
Clearly 4 x + x + 4 (1 − m) = (2x
2 1
+ 1) − m= y − m ≡ 0 (mod p)
2 2
so that

√ √ √
p | x 2 + x + 14 (1 − m). Hence p | x + 1+2 m x + 1−2 m in Z + Z 1+ m
2
.
Clearly
√
1± m √
x+ 1+ m
2 ∈
/ Z+Z
p 2
so that
√
1± m
px+ .
2
√
Hence p is not a prime in Z + Z 1+2 m .
√
As Z + Z 1+2 m is a principal ideal domain, by Theorem 1.4.3 p is not irre-
√
ducible in Z + Z 1+2 m . Hence
√ √
1+ m 1+ m
p = u+v w+t (1.4.3)
2 2
CB609-01 CB609/Alaca & Williams August 7, 2003 17:16 Char Count= 0

1.4 Principal Ideal Domains 15

√ √ √ √
1+ m 1+ m 1+ m
for some u + v ∈ Z + Z m and w + t ∈ Z + Z ,
2 √ √ 2
√ 2
where neither u + v 1+2 m nor w + t 1+2 m is a unit in Z + Z 1+2 m . From
(1.4.3) we have

v t vt (vw + ut) √
p= u+ w+ + m+ m.
2 2 4 2
√ √
As m is not a square, m ∈
/ Q, so that 1 and m are linearly independent over Q.
Hence

v t vt
p= u+ w+ + m, vw + ut = 0.
2 2 4
Thus
√ √
1− m 1− m
p= u+v w+t . (1.4.4)
2 2
Multiplying (1.4.3) and (1.4.4) together we obtain

1 1
p = u + uv + (1 − m)v
2 2 2
w + wt + (1 − m)t
2 2
(1.4.5)
4 4
since
√ √
1+ m 1− m 1
x+y x+y = x 2 + x y + (1 − m)y 2 .
2 2 4

As m ≡ 1 (mod 4), u 2 + uv + 14 (1 − m)v 2 ∈ Z and w2 + wt + 14 (1 − m)t 2 ∈ Z.

u + uv + (1 − m)v
= ±1 and w + wt + 4 (1 − m)t = ±1 as u +
2 1 2 2 1 2
Moreover
4
√ √ √
v 1+2 m and w + t 1+ m
2
are not units in Z + Z 1+ m
2
. Thus from (1.4.5)
we deduce that
1 1
± p = u 2 + uv + (1 − m)v 2 = w 2 + wt + (1 − m)t 2
4 4
as p is a prime. Hence there are integers u and v such that
1 1
p = u 2 + uv + (1 − m)v 2 or − (u 2 + uv + (1 − m)v 2 ).
4 4
If m < 0 then u 2 + uv + 14 (1 − m)v 2 > 0 so that p = u 2 + uv + 14 (1 − m)v 2 .
If m > 0, p = − u 2 + uv + 14 (1 − m)v 2 , and there exist integers T and U such
that T 2 + T U + 14 (1 − m)U 2 = −1 then p = u 2 + u v + 14 (1 − m)v 2 with u =
uT + 14 (1 − m)vU and v = uU + vT + vU .

Examples illustrating Theorems 1.4.4 and 1.4.5 are given in Section 2.5.
CB609-01 CB609/Alaca & Williams August 7, 2003 17:16 Char Count= 0

16 Integral Domains

1.5 Maximal Ideals and Prime Ideals

In this section we give the basic properties of maximal and prime ideals. These will
be important when we discuss Dedekind domains in Chapter 8.

Definition 1.5.1 (Maximal ideal) A proper ideal M of an integral domain D is

called a maximal ideal if whenever I is an ideal of D such that M ⊆ I ⊆ D then
I = M or I = D.

Example 1.5.1 The ideal x 2 + 1 is maximal in R[x]. To show this, assume

that I is an ideal of R[x] such that x 2 + 1 ⊂ I ⊂ R[x]. As x 2 + 1 is properly
contained in I , there exists f (x) ∈ I and f (x) ∈ x 2 + 1. Dividing f (x) by
x 2 + 1, we obtain
f (x) = (x 2 + 1)q(x) + r (x),
where r (x) = 0 and deg (r (x)) < 2. Thus r (x) = ax + b, where a ∈ R and b ∈ R
are not both 0, and
ax + b = r (x) = f (x) − q(x)(x 2 + 1) ∈ I.
Thus
a 2 x 2 − b2 = (ax + b)(ax − b) ∈ I
and
a 2 (x 2 + 1) ∈ I.
Hence
a 2 + b2 = (a 2 (x 2 + 1)) − (a 2 x 2 − b2 ) ∈ I.
Thus I contains a nonzero real number, that is, I contains a unit of R[x]. This proves
that I = R[x], a contradiction. Hence no such ideal I exists, and consequently
x 2 + 1 is a maximal ideal of R[x].

Example 1.5.2 5 is not a maximal ideal of Z + Zi as

5 ⊂ 1 + 2i ⊂ Z + Zi.

Theorem 1.5.1 Let D be an integral domain. Let a ∈ D be such that a = 0 and

a ∈ U (D). Then
a is a maximal ideal of D =⇒ a is irreducible in D.

Proof: Suppose that a is not an irreducible element of D. Then, as a is neither 0 nor

a unit, it must be reducible. Hence there exist b ∈ D and c ∈ D such that a = bc
CB609-01 CB609/Alaca & Williams August 7, 2003 17:16 Char Count= 0

1.5 Maximal Ideals and Prime Ideals 17

and neither b nor c is a unit or 0. Thus

a ⊂ b ⊂ D
so that a is not a maximal ideal. Hence we have shown that
a is a maximal ideal =⇒ a is irreducible,
as asserted.

The next example shows that the converse of Theorem 1.5.1 is not true in general.

Example 1.5.3
(a) x is an irreducible element of Z[x] but x is not a maximal ideal of Z[x] as
x ⊂ 2, x ⊂ Z[x].
√ √ √
(b) 1 + −5 is an irreducible element of Z + Z −5 but 1 + −5 is not a maximal
√
ideal of Z + Z −5 as
√ √ √
1 + −5 ⊂ 2, 1 + −5 ⊂ Z + Z −5.

However, the converse of Theorem 1.5.1 is true in a principal ideal domain.

Theorem 1.5.2 Let D be a principal ideal domain. Let a ∈ D be such that a = 0

and a ∈ U (D). Then
a is a maximal ideal of D ⇐⇒ a is irreducible in D.

Proof: In view of Theorem 1.5.1 we have only to show that

a is irreducible =⇒ a is maximal. (1.5.1)
Suppose that a is irreducible but that a is not a maximal ideal. Then there exists
an ideal I such that
a ⊂ I ⊂ D.
As D is a principal ideal domain, I = b for some b ∈ D. Hence
a ⊂ b ⊂ D
and so
a = bc,
for some c ∈ D. Since b ⊂ D, b is not a unit, and since a ⊂ b, c is not a
unit. Thus a is reducible, which is a contradiction. This completes the proof of
(1.5.1).
CB609-01 CB609/Alaca & Williams August 7, 2003 17:16 Char Count= 0

18 Integral Domains

Theorem 1.5.3 Let D be an integral domain and let I be an ideal of D. Then

D/I is a field ⇐⇒ I is maximal.

Proof: Suppose that D/I is a field and that J is an ideal of D with

I ⊂ J ⊆ D.

Thus there exists b ∈ J with b ∈ I . Then b + I is a nonzero element of D/I and

therefore, as D/I is a field, there exists an element c + I ∈ D/I such that

(b + I )(c + I ) = 1 + I.

Thus

bc + I = 1 + I

and so

bc − 1 ∈ I ⊂ J.

Since b ∈ J and c ∈ D we have

bc ∈ J.

Hence

1 = bc − (bc − 1) ∈ J,

so that J = 1 = D. This proves that I is maximal.

Now suppose that I is maximal. To show that D/I is a field we have only to
show that b + I = 0 + I has a multiplicative inverse, as all the other field properties
follow trivially. As b + I = 0 + I we have b ∈ I . Consider

B = {x ∈ D | x = by + w for some y ∈ D and some w ∈ I }.

It is easy to check that B is an ideal of D such that I ⊂ B (Exercise 12). Since I is

maximal we must have B = D. Thus 1 ∈ B so that 1 = by + w for some y ∈ D
and some w ∈ I . Then

(b + I )(y + I ) = by + I = 1 − w + I = 1 + I

so that (b + I )−1 exists and is equal to y + I .

Definition 1.5.2 (Prime ideal) A proper ideal P of an integral domain D is called

a prime ideal if

a, b ∈ D and ab ∈ P implies a ∈ P or b ∈ P.
CB609-01 CB609/Alaca & Williams August 7, 2003 17:16 Char Count= 0

1.5 Maximal Ideals and Prime Ideals 19

Example 1.5.4 The principal ideal I = x 2 + 1 is not a prime ideal of C[x] as

x ± i ∈ C[x], (x + i)(x − i) = x 2 + 1 ∈ I but x ± i ∈ I.

Example 1.5.5 The ideal I = 1 + i is a prime ideal of Z + Zi. To see this,

suppose that a + bi ∈ Z + Zi and c + di ∈ Z + Zi are such that
(a + bi)(c + di) ∈ 1 + i.
Then there exists x + yi ∈ Z + Zi such that
(a + bi)(c + di) = (1 + i)(x + yi).
Equating real and imaginary parts we obtain
ac − bd = x − y, ad + bc = x + y.
Adding these two equations, we have
ac + ad + bc − bd = 2x,
so that
(a + b)(c + d) = ac + ad + bc + bd ≡ ac + ad + bc − bd = 2x ≡ 0 (mod 2).
Hence either a + b or c + d is even. Without loss of generality we may suppose
that a + b is even. Hence there exist u ∈ Z and v ∈ Z such that a + b = 2u and
a − b = 2v. Then
a + bi = (u + v) + (u − v)i = (1 + i)(u − vi)
and thus a + bi ∈ 1 + i, proving that 1 + i is a prime ideal.

We next determine which principal ideals of an integral domain are prime.

Theorem 1.5.4 Let D be an integral domain. Let a ∈ D be such that a = 0 and

a ∈ U (D). Then
a is a prime ideal of D ⇐⇒ a is prime in D.

Proof: Suppose that a is a prime ideal of D. Let b, c ∈ D be such that a | bc so

that bc ∈ a. As a is a prime ideal, we must have b ∈ a or c ∈ a; that is, a | b
or a | c, showing that a is prime.
Now suppose that a is a prime in D. Let b ∈ D and c ∈ D be such that bc ∈ a.
Hence there exists d ∈ D such that bc = ad, so that a | bc. As a is prime we
have a | b or a | c. Without loss of generality we may suppose that a | b. Hence
there exists e ∈ D such that b = ae and so b ∈ a. This proves that a is a prime
ideal.
CB609-01 CB609/Alaca & Williams August 7, 2003 17:16 Char Count= 0

20 Integral Domains

Theorem 1.5.5 Let D be an integral domain and let I be an ideal of D. Then

D/I is an integral domain ⇐⇒ I is prime.

Proof: Suppose first that D/I is an integral domain and that a, b ∈ D are such that
ab ∈ I . Then (a + I )(b + I ) = ab + I = 0 + I , the zero element of the integral
domain D/I . Because an integral domain has no divisors of zero, we have a + I =
0 + I or b + I = 0 + I ; that is, we have either a ∈ I or b ∈ I , so that I is prime.
Now suppose that I is a prime ideal of D. As I is a proper ideal of D, D/I is a
commutative ring with identity 1 + I . Thus we have only to check that when I is
prime, D/I has no divisors of zero. Suppose that a + I ∈ D/I and b + I ∈ D/I
are such that (a + I )(b + I ) = 0 + I . Then ab + I = I , so that ab ∈ I . As I is
prime, either a ∈ I or b ∈ I ; that is, a + I = 0 + I or b + I = 0 + I , so D/I has
no zero divisors.

Theorem 1.5.6 Let D be an integral domain. Let I be a maximal ideal of D. Then

I is a prime ideal of D.

Proof: Let I be a maximal ideal of D. Then, by Theorem 1.5.3, D/I is a field.

But a field is always an integral domain, so D/I is an integral domain. Then, by
Theorem 1.5.5, I is a prime ideal of D.

The next example shows that the converse of Theorem 1.5.6 is not true in general.

Example 1.5.6 x is a prime ideal of Z[x], but it is not a maximal ideal of Z[x].

The converse of Theorem 1.5.6 is true in a principal ideal domain.

Theorem 1.5.7 Let D be a principal ideal domain. Let I be a proper ideal of D.

Then
I is maximal ⇐⇒ I is prime.

Proof: In view of Theorem 1.5.6 we have only to show that if I is a prime ideal of
D then I is a maximal ideal.
Suppose that I is a prime ideal of D that is not maximal. Then there exists an
ideal J of D such that
I ⊂ J ⊂ D.
As D is a principal ideal domain, we have I = a and J = b for some a, b ∈ D.
As a ⊂ b we have a = bc for some c ∈ D. Now bc = a ∈ a = I , and I is
prime, so that either b ∈ I or c ∈ I . If b ∈ I then J = b ⊆ I ⊂ J , which is a
contradiction. Hence c ∈ I . Thus c = ad for some d ∈ D, and so a = bda. But
CB609-01 CB609/Alaca & Williams August 7, 2003 17:16 Char Count= 0

1.6 Sums and Products of Ideals 21

a = 0 so bd = 1. Thus b is a unit and J = b = D ⊃ J , a contradiction. Hence I

is maximal.

1.6 Sums and Products of Ideals

In this section we show how to add and multiply ideals to obtain further ideals.
First we define the sum of two ideals.

Definition 1.6.1 (Sum of ideals) Let I and J be ideals in an integral domain D.

The sum of I and J , written I + J , is defined by

I + J = {i + j | i ∈ I, j ∈ J }.

It is readily checked that I + J is also an ideal and that it is the minimal ideal
containing both I and J . The following properties are also easily checked: For
ideals I, J, K of the integral domain D

I + J = J + I,
(I + J ) + K = I + (J + K ),
I + 0 = I,
I + 1 = 1.

Further, if I = i and J = j are principal ideals, then I + J = i, j. It is easy
to extend Definition 1.6.1 to the sum of a finite number of ideals.
Next we define the product of two ideals.

Definition 1.6.2 (Product of ideals) Let I and J be ideals in an integral domain

D. The product of I and J , written I J , is defined by

I J = {x ∈ D | x = i 1 j1 + · · · + ir jr for some r ∈ N,
some i 1 , . . . , ir ∈ I, and some j1 , . . . , jr ∈ J }.

Clearly I J is the set of all finite sums of products of elements of I and J , and it is
easily checked that I J is an ideal. The following properties are also easily verified:
For ideals I, J, K of the integral domain D

I J = J I,
(I J )K = I (J K ),
I 0 = 0,
I 1 = I.

Further, if I = i and J = j are principal ideals, then I J = i j. We leave it to

the reader to extend Definition 1.6.2 to a product of a finite number of ideals.
CB609-01 CB609/Alaca & Williams August 7, 2003 17:16 Char Count= 0

22 Integral Domains

Addition and multiplication of ideals are related by the distributive law

(I + J )K = I K + J K .

Example 1.6.1 Let m and n be integers that are not both zero. Set d = (m, n), the
greatest common divisor of m and n. We show that

m + n = d.

Let a ∈ m + n. Then there exist integers r and s such that a = r m + sn. As
d = (m, n) there exist coprime integers m 1 and n 1 such that m = dm 1 , n = dn 1 .
Thus a = r dm 1 + sdn 1 = (r m 1 + sn 1 )d ∈ d. This shows that m + n ⊆ d.
Now let a ∈ d, so that there exists an integer e such that a = de. As d = (m, n)
there exist integers x and y such that d = xm + yn. Hence a = (xm + yn)e =
(xe)m + (ye)n ∈ m + n. This proves that d ⊆ m + n.
The two inclusions show that m + n = d.

Next we give another necessary and sufficient condition for a proper ideal to be
a prime ideal.

Theorem 1.6.1 Let P be a proper ideal of an integral domain D. Then P is a

prime ideal if and only if for any two ideals A and B of D satisfying AB ⊆ P
either A ⊆ P or B ⊆ P.

Proof: Suppose that P is a proper ideal of D with the property

AB ⊆ P =⇒ A ⊆ P or B ⊆ P (A, B ideals of D). (1.6.1)

Let a, b ∈ D be such that ab ∈ P. Set A = a, B = b so that AB = ab =

ab ⊆ P. Hence a ⊆ P or b ⊆ P. Thus a ∈ P or b ∈ P, showing that P is a
prime ideal.
Now suppose that P does not satisfy (1.6.1). Then there exist ideals A and B of
D with

A ⊆ P, B ⊆ P, AB ⊆ P.

Let a ∈ A, a ∈ P and b ∈ B, b ∈ P. Then ab ∈ AB ⊆ P but a ∈ P, b ∈ P, so P

is not a prime ideal.

Our final theorem of this chapter shows that a prime ideal P of an integral domain
D1 remains prime when restricted to a subdomain D of D1 .

Theorem 1.6.2 Let D and D1 be integral domains satisfying D ⊆ D1 . Let P be a

prime ideal of D1 such that P ∩ D = {0}, D. Then P ∩ D is a prime ideal of D.
CB609-01 CB609/Alaca & Williams August 7, 2003 17:16 Char Count= 0

Exercises 23

Proof: We show first that P ∩ D is an ideal of D. Let a, b ∈ P ∩ D. Then a, b ∈ P

and a, b ∈ D. From the first of these, as P is an ideal, we see that a + b ∈ P.
From the second, as D is an integral domain, it is closed under addition so that
a + b ∈ D. Hence a + b ∈ P ∩ D. Now suppose that a ∈ P ∩ D and d ∈ D. As
d ∈ D, a ∈ P and P is an ideal of D, we deduce that da ∈ P. As d ∈ D, a ∈ D
and D being an integral domain is closed under multiplication, we see that da ∈ D.
Thus da ∈ P ∩ D. This proves that P ∩ D is an ideal of D. Since P ∩ D = {0}, D,
by assumption, P ∩ D is a proper ideal of D.
Finally, we show that P ∩ D is a prime ideal. Let a, b ∈ D be such that ab ∈
P ∩ D. Then a, b ∈ D1 and ab ∈ P. As P is a prime ideal of D1 , we deduce that
a ∈ P or b ∈ P. This completes the proof that P ∩ D is a prime ideal of D.

Exercises
1. Prove that U (Z + Zi) = {±1, ±i}.
2. Prove that U (Z + Zω) = {±1, ±ω, ±ω2 }.
3. Let m be an integer with m < −1. Prove that
√
U (Z + Z m) = {±1}.
4. Let m be an integer with m ≡ 1 (mod 4) and m < −3. Prove that
√
1+ m
U Z+Z = {±1}.
2
5. Let

D = { f (x) ∈ Q[x] | f (0) ∈ Z}.

Prove that D is a subdomain of Q[x].

6. Determine U (D) for D as given in Exercise 5.
7. Let D be an integral domain. Let u ∈ U (D). Let I be an ideal of D that contains u.
Prove that I = D.
8. In Example 1.3.4 prove that J is not a principal ideal.
9. Let
S = {a + bi ∈ Z + Zi | b ≡ 0 (mod 2)}.
Is S an ideal of Z + Zi?
10. If A and B are ideals of an integral domain D, prove that A ∩ B is also an ideal of D.
11. Give an example to show that if A and B are ideals of an integral domain D then A ∪ B
may not be an ideal of D.
12. Prove that the set B defined in the proof of Theorem 1.5.3 is an ideal.
13. Let A and B be ideals of an integral domain D. Prove that AB ⊆ A ∩ B.
14. Let A and B be ideals of an integral domain D. Show that (A ∩ B)(A + B) ⊆ AB.
Give an example to show that equality does not always hold.
15. Give an example to show that an integral domain may not contain any irreducible
elements.
CB609-01 CB609/Alaca & Williams August 7, 2003 17:16 Char Count= 0

24 Integral Domains

16. Prove that x is a prime ideal of Z[x].

√ √
17. Let m be a positive integer that is not a perfect square. Let α = a + b m ∈ Z + Z m.
Prove that if

a 2 − mb2 = ±1
√
then α ∈ U (Z + Z m).
18. Let m be a√ positive
integer
with
√
m ≡ 1 (mod 4) that is not a perfect square. Let α =
a + b 1+2 m ∈ Z + Z 1+2 m . Prove that if

1−m
a 2 + ab + b2 = ±1
4
√
then α ∈ U Z + Z 1+2 m .
19. Prove that 1 − 3i, 3 − i is a principal ideal in Z + Zi by finding a generator for this
ideal.
√ √ √ √
20. Prove that 2, 1 + −5 = 2, 1 − −5, 3, 1 + −5 = 3, 1 − −5, 2, 1 +
√ √ √ √ √
−5 = 3, 1 + −5, and 2, 1 + −5 = 3, 1 − −5 in Z + Z −5.
√ √ √
21. Prove that 2, 1 + −5, 3, 1 + −5, and 3, 1 − −5 are prime ideals of Z +
√ √ √ √
Z −5. Determine 2, 1 + −5 ∩ Z, 3, 1 + −5 ∩ Z, and 3, 1 − −5 ∩ Z.
22. Let D be an integral domain. Let a, b, c ∈ D be such that a, c = D. Prove that
a, bc = a, b. √ √ √
23. Prove that 17 − 3 3 ∼ 83 + 47 3 in Z + Z 3.
24. Give an example of an integral domain satisfying (1.2.3).
√ √
25. Express 2 + 8 −5 as a product of irreducibles in Z + Z −5. In how many ways can
this be done?
√ √
26. Prove that −6 is not a prime in Z + Z −6.
√ √
27. Prove that −6 is an irreducible in Z + Z −6.
√
28. Prove that Z + Z −6 is not a principal ideal domain.
√
29. Give an example
√ of an ideal in Z + Z √−6 that is not principal.
30. Prove that √10 is not a prime in Z + Z 10. √
31. Prove that 10 is√an irreducible in Z + Z 10.
32. Prove that Z + Z 10 is not a principal √ideal domain.
33. Give an example of an ideal in Z + Z 10 that is not principal.
34. Let P be a prime ideal of an integral domain D. Let A1 , . . . , Ak be ideals of D such
that P ⊇ A1 · · · Ak . Prove that P ⊇ Ai for some i ∈ {1, 2, . . . , k}.
35. Let r ∈ Z \ {−2, 0}. Prove that

D = {a + bθ + cθ 2 | a, b, c ∈ Z},

where

θ3 + rθ + 1 = 0

is an integral domain. Prove that θ ∈ U (D).

36. Let p be a prime. Let m be an integer with m ≤ −( p + 1). Prove that p is irreducible
√
in Z + Z m.
CB609-01 CB609/Alaca & Williams August 7, 2003 17:16 Char Count= 0

Biographies 25

37. Let p be a prime. Let m be an integer

√
with m ≡ 1 (mod 4) and m ≤ −(4 p + 1). Prove
1+ m
that p is irreducible in Z + Z 2
.

Suggested Reading
1. P. M. Cohn, Rings of fractions, American Mathematical Monthly 78 (1971), 596–615.
The author won the Lester R. Ford award for expository writing for this paper. The paper reviews
Ore’s work on embedding certain non-commutative rings in skew fields, a generalization of the
corresponding standard result for integral domains mentioned in Section 1.1.
2. D. A. Cox, Primes of the Form x 2 + my 2 , Wiley, New York, 1989.
The main theorem of the book (Theorem 9.2, p. 180) asserts (with some details omitted) that if m
is a positive integer then there is a polynomial f m (x)∈ Z[x] (of a certain degree depending only
on m) such that if p is an odd prime satisfying −m p
= 1 then p = u 2 + mv 2 for integers u and
v if and only if the congruence f m (x) ≡ 0 (mod
p) is solvable.
For example if p is an odd prime such that −36 p
= 1, that is, p ≡ 1 (mod 4), then

p = u 2 + 36v 2 ⇐⇒ x 4 + 3 ≡ 0 (mod p) is solvable.

3. J. B. Fraleigh, A First Course in Abstract Algebra, Addison-Wesley, Reading, Mas-

sachusetts, 1968.
Chapter 26 is devoted to constructing the field of quotients of an integral domain.
4. L. Kinkade and J. Wagner, When polynomial rings are principal ideal rings, Journal of
Undergraduate Mathematics 23 (1991), 59–62.
In this paper it is shown that R[x] is a principal ideal ring if and only if R R1 ⊕ R2 ⊕ · · · ⊕ Rn ,
where R1 , . . . , Rn are fields.
5. D. E. Rowe, Gauss, Dirichlet and the law of biquadratic reciprocity, The Mathematical
Intelligencer 10 (1988), 13–26.
This paper gives a discussion of the relationship between Gauss and Dirichlet, mainly concerning
their contributions to number theory including their work on biquadratic reciprocity.
6. W. C. Waterhouse, Quadratic polynomials and unique factorization, American Mathe-
matical Monthly 109 (2002), 70–72.
Theorem 1.2.2 is taken from this paper.

Biographies
1. E. T. Bell, Men of Mathematics, Simon and Schuster, New York, 1937.
Chapters 14 and 17 are devoted to Gauss and Abel respectively.
2. W. K. Bühler, Gauss: A Biographical Study, Springer-Verlag, Berlin, Heildelberg,
New York, 1981.
This book provides a comprehensive discussion of Gauss’s life and work.
3. G. Eisenstein, Mathematische Werke, Bände I, II, Chelsea Publishing Co., New York,
1989.
The foreword to Eisenstein’s Collected Papers comprises an interesting discussion of Eisenstein’s
work by André Weil.
CB609-01 CB609/Alaca & Williams August 7, 2003 17:16 Char Count= 0

26 Integral Domains

4. O. Ore, Niels Henrik Abel, Mathematician Extraordinary, University of Minnesota Press,

1957; Chelsea, New York, 1974.
A nontechnical biography is presented here in an easy-to-read fashion.
5. A. Stubhaug, Niels Henrik Abel and His Times, Springer-Verlag, New York, 2000.
Stubhaug presents a very readable account of a remarkable mathematician.
6. The website

http://www-groups.dcs.st-and.ac.uk/˜history/

has biographies of Abel, Gauss, and Eisenstein.

CB609-02 CB609/Alaca & Williams August 27, 2003 16:51 Char Count= 0

2
Euclidean Domains

In the proof of Theorem 1.4.1 we made use of the following property of Z: Given
a, b ∈ Z with b > 0 then there exist q, r ∈ Z such that
a = qb + r, 0 ≤ r < b. (2.0.1)
In fact the integers q and r are uniquely determined by a and b. We have
q = [a/b], r = a − b[a/b], (2.0.2)
where [x] denotes the greatest integer less than or equal to the real number x.
The integer q is called the quotient and the integer r the remainder. An important
class of integral domains are those possessing a property analogous to (2.0.1). Such
domains are called Euclidean domains. In Theorem 2.1.2 we show that Euclidean
domains are principal ideal domains.

2.1 Euclidean Domains

To define a Euclidean domain we must first define a Euclidean function.

Definition 2.1.1 (Euclidean function) Let D be an integral domain. A mapping

φ : D → Z is called a Euclidean function on D if it has the following two properties:
φ(ab) ≥ φ(a), for all a, b ∈ D with b = 0, (2.1.1)

if a, b ∈ D with b = 0 then there exist q, r ∈ D (2.1.2)

such that a = qb + r and φ(r ) < φ(b).

Example 2.1.1 φ(a) = |a| (a ∈ Z) is a Euclidean function on Z.

Example 2.1.2 Let D = F[x], where F is a field. D is the domain of polynomials

in x with coefficients in F. Let p(x) ∈ D. Then

deg ( p(x)), if p(x) = 0,
φ( p(x)) =
−1, if p(x) = 0,
is a Euclidean function on D.

27
CB609-02 CB609/Alaca & Williams August 27, 2003 16:51 Char Count= 0

28 Euclidean Domains

In general the elements q and r in (2.1.2) are not uniquely determined. If D is

an integral domain that is not a field and that possesses a Euclidean function φ for
which the quotient and remainder r in (2.1.2) are always uniquely determined by
a and b then D = F[x] for some field F. This result is due to Rhai [14]; see also
Jodeit [12].

Theorem 2.1.1 (Properties of a Euclidean function) Let D be an integral

domain that possesses a Euclidean function φ. Let a, b ∈ D. Then

(a) a ∼ b =⇒ φ(a) = φ(b),

(b) a | b and φ(a) = φ(b) =⇒ a ∼ b,
(c) a ∈ U (D) ⇐⇒ φ(a) = φ(1),
(d) φ(a) > φ(0), if a = 0.

Proof: (a) As a ∼ b there exists u ∈ U (D) such that a = ub. Then by (2.1.1) we
have φ(a) = φ(ub) ≥ φ(b). As u ∈ U (D), we have u −1 ∈ U (D) and b = u −1 a, so
again by (2.1.1) we have φ(b) = φ(u −1 a) ≥ φ(a). From these two inequalities, we
deduce that φ(a) = φ(b).
(b) By (2.1.2) there exist q, r ∈ D such that a = qb + r and φ(r ) < φ(b) =
φ(a). Now a | b so that we have a | r . Suppose r = 0. Then by (2.1.1) we have
φ(r ) ≥ φ(a), which is a contradiction. Hence r = 0. Thus a = qb. But a | b so
q ∈ U (D) and thus a ∼ b.
(c) First we have
a ∈ U (D) =⇒ a ∼ 1 =⇒ φ(a) = φ(1)
by part (a). Second, we have
1 | a, φ(1) = φ(a) =⇒ 1 ∼ a =⇒ a ∈ U (D)
by part (b).
(d) By (2.1.2) there exist q, r ∈ D such that
0 = qa + r, φ(r ) < φ(a).
Suppose r = 0. Then q = 0 and by (2.1.1) we have
φ(r ) = φ((−q)a) ≥ φ(a),
which is a contradiction. Hence r = 0 and φ(0) < φ(a).

Definition 2.1.2 (Euclidean domain) Let D be an integral domain. If D possesses

a Euclidean function φ then D is called a Euclidean domain with respect to φ.

If D is a Euclidean domain with respect to some Euclidean function φ and it is not

important to specify φ, we just call D a Euclidean domain. Before giving examples
CB609-02 CB609/Alaca & Williams August 27, 2003 16:51 Char Count= 0

2.1 Euclidean Domains 29

of Euclidean domains in the next section, we prove the fundamental theorem that
every Euclidean domain is a principal ideal domain.

Theorem 2.1.2 A Euclidean domain is a principal ideal domain.

Proof: Let D be a Euclidean domain. Hence D possesses a Euclidean function, say

φ. Let I be an ideal in D. If I = {0} then I = 0 is a principal ideal. If I = {0}
we consider the set S of integers defined by

S = {φ(x) | x ∈ I, x = 0}.

As I = {0}, S is a nonempty set. By Theorem 2.1.1(d), S is bounded below. Hence

S has a least element, say φ(a), a ∈ I, a = 0. If b ∈ I then, as φ is a Euclidean
function, there exist q, r ∈ D such that

b = qa + r, φ(r ) < φ(a).

Now, as I is an ideal, r = b − qa ∈ I , and so, as φ(a) is the least element of S, we

have r = 0. Hence b = qa and so I = a. Thus every ideal in D is principal and
so D is a principal ideal domain.
√
The integral domain Z + Z 1+ 2−19 is a principal ideal domain. This will be
proved in Chapter 12 (see Example 12.6.1). However, it is not a Euclidean domain
(Theorem 2.3.8), so the converse of Theorem 2.1.2 is not true.
In a Euclidean domain D a greatest common divisor of two elements a and b
of D (see Definition 1.4.2) can be obtained by means of the Euclidean algorithm.
Since (c, 0) = (0, c) = c for all c ( = 0) ∈ D, it suffices to consider only elements
a and b that are not zero.

Theorem 2.1.3 (Euclidean algorithm) Let a and b be nonzero elements of a

Euclidean domain D with Euclidean function φ. Define elements q1 , q2 , . . . and
r−1 , r0 , r1 , r2 , . . . of D recursively by

r−1 = a, r0 = b, (2.1.3)

and

r j = q j+2r j+1 + r j+2 , φ(r j+2 ) < φ(r j+1 ), (2.1.4)

for j = −1, 0, 1, 2, . . . , k, where k is the least integer ≥ −1 such that

rk+2 = 0.

Then

(a, b) = rk+1 .
CB609-02 CB609/Alaca & Williams August 27, 2003 16:51 Char Count= 0

30 Euclidean Domains

Proof: By property (2.1.2) of the Euclidean function φ, the relations (2.1.3) and
(2.1.4) define q1 , q2 , . . . , qk+2 and r−1 , r0 , r1 , . . . , rk+2 , and since the sequence
φ(r1 ), φ(r2 ), . . . is a decreasing sequence of integers bounded below by φ(0) (The-
orem 2.1.1(d)) it must terminate after a finite number of steps (say k + 2 steps) so
that rk+2 = 0. From (2.1.4) we deduce that
r j , r j+1 = q j+2r j+1 + r j+2 , r j+1 = r j+2 , r j+1 = r j+1 , r j+2
for j = −1, 0, 1, 2, . . . , k. Hence
a, b = r−1 , r0 = r0 , r1 = · · · = rk , rk+1
= rk+1 , rk+2 = rk+1 , 0 = rk+1
so that
(a, b) = rk+1 .

2.2 Examples of Euclidean Domains

In view of Examples 2.1.1 and 2.1.2 we have

Theorem 2.2.1
(a) Z is a Euclidean domain.
(b) Let F be a field. Then F[x] is a Euclidean domain.

From Theorems 2.1.2 and 2.2.1 we see that Z and F[x] are principal ideal
domains. In the remainder of this section we
investigate
√
when the integral domains
√ 1+ m
Z + Z m (m ≡ 2, 3 (mod 4)) and Z + Z (m ≡ 1 (mod 4)) are Euclidean
√2
with respect to the function that maps r + s m to |r 2 − ms 2 |. In this section we
denote this function by φm . Later in Section 9.2 we recognize φm as the absolute
√
value of the norm of the element r + s m. Integral domains that are Euclidean
with respect to the absolute value of the norm are called norm-Euclidean.

Definition 2.2.1 (Function φm ) Let m be a squarefree integer. The function φm :

√
Q( m) → Q is defined by
√
φm (r + s m) = |r 2 − ms 2 |
for all r, s ∈ Q.

The basic properties of φm are given in the next lemma.

Lemma 2.2.1 Let m be a squarefree integer.

√
(a) φm : Z + Z m → N ∪ {0}. √
(b) If m ≡ 1 (mod 4) then φm : Z + Z 1+2 m → N ∪ {0}.
CB609-02 CB609/Alaca & Williams August 27, 2003 16:51 Char Count= 0

2.2 Examples of Euclidean Domains 31

√
(c) Let α ∈ Q( m). Then φm (α) = 0 ⇐⇒ α = 0.
√
(d) φm (αβ) = φm (α)φm (β) for all α, β ∈ Q( m).
√
(e) φm (αβ) ≥ φm (α) for all α, β ∈ Z + Z m with β = 0. √
(f) If m ≡ 1 (mod 4), then φm (αβ) ≥ φm (α) for all α, β ∈ Z + Z 1+2 m with β = 0.

√ √
Proof: (a) Let α ∈ Z + Z m so that α = x + y m for some x, y ∈ Z. Then
x 2 − my 2 ∈ Z and |x 2 − my 2 | ≥ 0 so that
√
φm (α) = φm (x + y m) = |x 2 − my 2 | ∈ N ∪ {0}.
√
(b) If m ≡ 1 (mod 4) then Z + Z 1+2 m is an integral domain (Example 1.1.5).
√ √ √
Let α ∈ Z + Z 1+2 m so that α = x + y 1+2 m = x + 2y + 2y m for some
x, y ∈ Z. Then
y y√
φm (α) = φm x + + m
2 2
y y
= |(x + )2 − m( )2 |
2 2
1 1
= |x + x y + (1 − m)y 2 | ∈ N ∪ {0}, as (1 − m) ∈ Z.
2
4 4
√ √
(c) Let α ∈ Q( m) so that α = r + s m for some r, s ∈ Q. Then, as m is
squarefree, we have
√
φm (α) = 0 ⇐⇒ φm (r + s m) = 0
⇐⇒ |r 2 − ms 2 | = 0
⇐⇒ r 2 = ms 2
⇐⇒ r = s = 0
√
⇐⇒ r + s m = 0
⇐⇒ α = 0.
√ √ √
(d) Let α, β ∈ Q( m). Then α = x + y m and β = u + v m for some
x, y, u, v ∈ Z. Thus
√ √
φm (αβ) = φm ((x + y m)(u + v m))
√
= φm ((xu + myv) + (xv + yu) m)
= |(xu + myv)2 − m(xv + yu)2 |
= |x 2 u 2 + m 2 y 2 v 2 − mx 2 v 2 − my 2 u 2 |
= |(x 2 − my 2 )(u 2 − mv 2 )|
= |x 2 − my 2 | |u 2 − mv 2 |
= φm (α)φm (β).
√
(e) Let α, β ∈ Z + Z m with β = 0. By part (c), we have φm (β) = 0. Then, by
part (a), we deduce that φm (α) ≥ 0 and φm (β) ≥ 1. Thus, by part (d), we have

φm (αβ) = φm (α)φm (β) ≥ φm (α).

CB609-02 CB609/Alaca & Williams August 27, 2003 16:51 Char Count= 0

32 Euclidean Domains

(f) This follows in exactly the same way as part (e) except that we use part (b) in
place of part (a).

Our next theorem uses the properties of φm given in Lemma 2.2.1 to give a
√
convenient necessary and sufficient condition for Z + Z m to be Euclidean with
respect to φm , that is, norm-Euclidean.

√
Theorem 2.2.2 Let m be a squarefree integer. Then the integral domain Z + Z m
is Euclidean with respect to φm if and only if for all x, y ∈ Q there exist a, b ∈ Z
such that
√ √
φm ((x + y m) − (a + b m)) < 1. (2.2.1)

√
Proof: Suppose first that Z + Z m is Euclidean with respect to φm . Let x, y ∈ Q.
√ √
Then x + y m = (r + s m)/t for integers r, s, t with t = 0. As φm is a Euclidean
√ √ √ √
function on Z + Z m there exist a + b m, c + d m ∈ Z + Z m such that
√ √ √ √
r + s m = t(a + b m) + (c + d m), φm (c + d m) < φm (t).

Hence
√
√ √ r +s m √
φm ((x + y m) − (a + b m)) = φm − (a + b m)
t
√ √
r + s m − t(a + b m)
= φm
t
√
c+d m
= φm
t
√
φm (c + d m)
= < 1,
φm (t)

by Lemma 2.2.1(d).
√
Now suppose that (2.2.1) holds. To show that Z + Z m is Euclidean with respect
to φm , we must show that (2.1.1) and (2.1.2) hold. The inequality (2.1.1) holds in
√
view of Lemma 2.2.1(e). We now show that (2.1.2) holds. Let r + s m, t +
√ √ √
u m ∈ Z + Z m with t + u m = 0. Then
√
r +s m √
√ = x + y m,
t +u m

where
r t − msu st − r u
x= ∈ Q, y = 2 ∈ Q.
t − mu
2 2 t − mu 2
CB609-02 CB609/Alaca & Williams August 27, 2003 16:51 Char Count= 0

2.2 Examples of Euclidean Domains 33

√
Note that t + u m = 0 ensures that t 2 − mu 2 = 0. By (2.2.1) there exists a +
√ √
b m ∈ Z + Z m such that
√ √
φm ((x + y m) − (a + b m)) < 1.

Set c = r − at − bum ∈ Z, d = s − au − bt ∈ Z, so that

√ √ √ √ √
c + d m = (r + s m) − (a + b m)(t + u m) ∈ Z + Z m.

Hence
√ √ √ √
r + s m = (a + b m)(t + u m) + (c + d m)

and
√ √ √ √
φm (c + d m) = φm ((r + s m) − (a + b m)(t + u m))
√ √ √ √
= φm ((x + y m)(t + u m) − (a + b m)(t + u m))
√ √ √
= φm ((t + u m)((x + y m) − (a + b m)))
√ √ √
= φm (t + u m)φm ((x + y m) − (a + b m))
√
< φm (t + u m),
by Lemma 2.2.1(d), which completes the proof of (2.1.2).

Theorem 2.2.2 enables us to determine those negative squarefree integers m for

√
which Z + Z m is Euclidean with respect to φm .

Theorem 2.2.3 Let m be a negative squarefree integer. Then the integral domain
√
Z + Z m is Euclidean with respect to φm if and only if m = −1, −2.

√
Proof: First we show that Z + Z m is Euclidean with respect to φm for m = −1
and m = −2. Let x, y ∈ Q. We can choose a, b ∈ Z such that
1 1
|x − a| ≤ , |y − b| ≤ .
2 2
Then
√ √ √
φm ((x + y m) − (a + b m)) = φm ((x − a) + (y − b) m))
= |(x − a)2 − m(y − b)2 |
≤ |x − a|2 + |m||y − b|2
1 1
≤ +2·
4 4
3
= <1
4
√
and, appealing to Theorem 2.2.2, we deduce that Z + Z m is Euclidean with
respect to φm for m = −1 and m = −2.
CB609-02 CB609/Alaca & Williams August 27, 2003 16:51 Char Count= 0

34 Euclidean Domains
√
Now suppose that Z + Z m is Euclidean with respect to φm . Then, by Theorem
2.2.2, there exist a, b ∈ Z such that

1 1√ √
φm ( + m) − (a + b m) < 1;
2 2
that is (as −m = |m|),
2 2
1 1
−a + |m| −b < 1.
2 2
But for any integer x, we have
2
1 1 1 1
| − x| ≥ , −x ≥ ,
2 2 2 4
so
1 |m|
+ < 1;
4 4
that is, |m| < 3. Hence m = −1 and m = −2 are the only possibilities.

In an exactly similar way to the proof of Theorem 2.2.2, we can prove the
following result.

Theorem 2.2.4Let√m be a squarefree integer with m ≡ 1 (mod 4). Then the integral
1+ m
domain Z + Z 2
is Euclidean with respect to φm if and only if for all x, y ∈ Q
there exist a, b ∈ Z such that
√
√ 1+ m
φm (x + y m) − a + b < 1.
2

From Theorem 2.2.4, exactly as we proved Theorem 2.2.3, we candetermine

√
those negative squarefree integers m ≡ 1 (mod 4) for which Z + Z 1+2 m is
Euclidean with respect to φm .

Theorem 2.2.5 Let m be a negative

√
squarefree integer with m ≡ 1 (mod 4). Then
1+ m
the integral domain Z + Z 2
is Euclidean with respect to φm if and only if
m = −3, −7, −11.

The determination of the positive √ squarefree

integers m for which Z +
√ 1+ m
Z m (m ≡ 2, 3 (mod 4)) and Z + Z 2
(m ≡ 1 (mod 4)) are Euclidean with
respect to φm is much more difficult and was the culmination of the efforts of numer-
ous mathematicians including E. S. Barnes (1874–1953), H. Behrbohm, E. Berg,
A. T. Brauer (1894–1985), H. Chatland, H. Davenport (1907–1969), L. E. Dickson
(1874–1954), P. Erdös (1913–1996), H. A. Heilbronn (1908–1975), N. Hofreiter,
CB609-02 CB609/Alaca & Williams August 27, 2003 16:51 Char Count= 0

2.2 Examples of Euclidean Domains 35

L. K. Hua, K. Inkeri, J. F. Keston, C. Ko, S. H. Min, A. Oppenheim, O. Perron

(1880–1975), L. Rédei, R. Remak (1888–1942), L. Schuster, W. T. Sheh, and H. P. F.
Swinnerton-Dyer.
The final step was taken in 1950 by Chatland and Davenport [4], who established
the following two theorems.

Theorem 2.2.6 Let m be a positive squarefree integer with m ≡ 2, 3 (mod 4). Then
√
the integral domain Z + Z m is Euclidean with respect to φm if and only if m =
2, 3, 6, 7, 11, 19, 57.

Theorem 2.2.7 Let m be apositive

√
squarefree integer with m ≡ 1 (mod 4). Then
1+ m
the integral domain Z + Z 2
is Euclidean with respect to φm if and only if
m = 5, 13, 17, 21, 29, 33, 37, 41, 73.

We will not prove these two theorems here. We will just prove the following
result.

√
Theorem 2.2.8 The integral domain Z + Z m is Euclidean with respect to φm for
m = 2, 3, 6.

Proof: m = 2, 3. Let x, y ∈ Q. We choose a, b ∈ Z such that

1 1
|x − a| ≤ , |y − b| ≤ .
2 2
As (x − a)2 ≥ 0 and m(y − b)2 ≥ 0, we have
3
|(x − a)2 − m(y − b)2 | ≤ max(|x − a|2 , m|y − b|2 ) ≤ .
4
Thus
√ √
φm ((x + y m) − (a + b m)) = |(x − a)2 − m(y − b)2 | < 1,

and the result follows by Theorem 2.2.2.

√
m = 6. Suppose that Z + Z 6 is not Euclidean with respect to φ6 . Then, by
Theorem 2.2.2, there exist r, s ∈ Q such that
√ √
φ6 ((r + s 6) − (x + y 6)) ≥ 1 for all x, y ∈ Z;

that is,

|(r − x)2 − 6(s − y)2 | ≥ 1 for all x, y ∈ Z.

CB609-02 CB609/Alaca & Williams August 27, 2003 16:51 Char Count= 0

36 Euclidean Domains

We can choose 1 = ±1 and u 1 ∈ Z such that

1
0 ≤ 1 r + u 1 ≤
2
and 2 = ±1 and u 2 ∈ Z such that
1
0 ≤ 2 s + u 2 ≤ .
2
Set
r1 = 1r + u 1 ∈ Q, x1 = 1 x + u 1 ∈ Z,
s1 = 2 s + u 2 ∈ Q, y1 = 2 y + u 2 ∈ Z,
so that
1 1
0 ≤ r 1 ≤ , 0 ≤ s1 ≤ , (2.2.2)
2 2
and
|(r1 − x1 )2 − 6(s1 − y1 )2 | ≥ 1 for all x1 , y1 ∈ Z. (2.2.3)
Taking (x1 , y1 ) = (0, 0), (1, 0), and (−1, 0) in (2.2.3), we obtain the inequalities
 2
 |r1 − 6s12 | ≥ 1,
|(1 − r1 )2 − 6s12 | ≥ 1, (2.2.4)

|(1 + r1 ) − 6s1 | ≥ 1.
2 2

From (2.2.2) we deduce that


 3 1

 − ≤ r12 − 6s12 ≤ ,


 2 4
5
− ≤ (1 − r1 ) − 6s12 ≤ 1,
2 (2.2.5)

 4


 1
 − ≤ (1 + r1 )2 − 6s 2 ≤ 9 .
1
2 4
From (2.2.4) and (2.2.5), we deduce that
3
− ≤ r12 − 6s12 ≤ −1, (2.2.6)
2

5
(i) (1 − r1 )2 − 6s12 = 1 or (ii) − ≤ (1 − r1 )2 − 6s12 ≤ −1, (2.2.7)
4

9
1 ≤ (1 + r1 )2 − 6s12 ≤ . (2.2.8)
4
From (2.2.6) and (2.2.8), we obtain
1 ≤ 1 + 2r1 + (r12 − 6s12 ) ≤ 2r1 ,
CB609-02 CB609/Alaca & Williams August 27, 2003 16:51 Char Count= 0

2.3 Examples of Domains That are Not Euclidean 37

so that r1 ≥ 12 . But r1 ≤ 12 so we must have r1 = 12 . Then (2.2.7)(i) gives 14 −

6s12 = 1, which is impossible, and (2.2.7)(ii) gives 14 − 6s12 ≤ −1, so that s12 ≥ 24
5
.
But from (2.2.8) we have 6s1 ≤ (1 + r1 ) − 1 = 4 ; that is, s1 ≤ 24 , so that s1 = 24
2 2 5 2 5 2 5
,
√
which is impossible. This completes the proof that Z + Z 6 is Euclidean with
respect to φ6 .

2.3 Examples of Domains That are Not Euclidean

√
We begin by giving a class of values of m for which Z + Z m is not Euclidean
with respect to φm .

Theorem 2.3.1 Let m be a positive squarefree integer. If there exist distinct odd
primes p and q such that

m m
= = −1,
p q
and positive integers t and u such that

pt + qu = m, p t, q u,

and an integer r such that

r 2 ≡ pt (mod m),
√
then Z + Z m is not Euclidean with respect to φm .
√
Proof: Suppose that Z + Z m is Euclidean with respect to φm . Then there exist
√
γ , δ ∈ Z + Z m such that
√
r m = mγ + δ, φm (δ) < φm (m).
√
Setting γ = x + y m (x, y ∈ Z) we obtain
√ √
φm (r m − m(x + y m)) < φm (m);

that is

|m 2 x 2 − m(r − my)2 | < m 2 ,

so that

|mx 2 − (my − r )2 | < m.

Since

mx 2 − (my − r )2 ≡ −r 2 ≡ − pt (mod m)
CB609-02 CB609/Alaca & Williams August 27, 2003 16:51 Char Count= 0

38 Euclidean Domains

and

0 < pt < pt + qu = m,

we must have

mx 2 − (my − r )2 = − pt or m − pt;

that is

m X 2 − Y 2 = − pt or qu

for integers X (= x) and Y (= my − r ). Suppose that m X 2 − Y 2 = − pt. As mp =
−1 we have p m. Also, as p t we have p || − pt. Hence p X and p Y . Thus
2
m m X2 Y
= = = 1,
p p p

contradicting mp = −1. Now suppose that m X 2 − Y 2 = qu. As mq = −1 we
have q m. Also, as q u we have q || qu. Hence q X and q Y . Thus
2
m m X2 Y
= = = 1,
q q q
√
contradicting mq = −1. This proves that Z + Z m is not Euclidean with respect
to φm .

We next use Theorem 2.3.1 to give some explicit, small, positive, squarefree
√
values of m for which Z + Z m is not Euclidean with respect to φm .

√
Theorem 2.3.2 Z + Z m is not Euclidean with respect to φm for m = 23, 47,
59, 83.

Proof: This follows immediately from Theorem 2.3.1 and the following table.

m p q t u r
23 3 5 1 4 7
47 3 5 4 7 23
59 3 7 15 2 24
83 3 5 1 16 13

√
1+ m
The corresponding result to Theorem 2.3.1 for Z + Z 2
(m ≡ 1 (mod 4))
is not quite so elegant.
CB609-02 CB609/Alaca & Williams August 27, 2003 16:51 Char Count= 0

2.3 Examples of Domains That are Not Euclidean 39

Theorem 2.3.3 Let m be a positive squarefree integer with m ≡ 1 (mod 4). If there
exist distinct odd primes p and q such that

m m
= = −1
p q
and an odd integer r such that

(m − 1)r 2
p || (m − 1)r − 4m
2
,
4m

(m − 1)r 2
q || (m − 1)r 2 − 4m − 4m,
4m
√
1+ m
then Z + Z 2
is not Euclidean with respect to φm .

Proof: As m and r are both odd, we have m−r

2
∈ Z. Hence
√ √ √
m +r m m −r 1+ m 1+ m
= +r ∈Z+Z .
2 2 2 2
√
Suppose that Z + Z 1+2 m is Euclidean with respect to φm . Then there exist
√
γ , δ ∈ Z + Z 1+2 m such that
√
m +r m
= mγ + δ, φm (δ) < φm (m).
2
√ √
As γ ∈ Z + Z 1+2 m there exist x, y ∈ Z such that γ = x + y 1+2 m , and
thus
√ √
m +r m 1+ m
φm −m x +y < φm (m).
2 2
Hence
2 r 2
m my my
< m2.
2 − mx − 2 −m
2
−
2
Multiplying both sides of this inequality by 4/m, we obtain

|m(1 − 2x − y)2 − (r − my)2 | < 4m.

Set X = 1 − 2x − y ∈ Z and Y = r − my ∈ Z so that

|m X 2 − Y 2 | < 4m.

As m ≡ 1 (mod 4) and (u + 2v)2 ≡ u 2 (mod 4)) for any u, v ∈ Z, we deduce that

m X 2 − Y 2 ≡ (1 − y)2 − (1 − y)2 ≡ 0 (mod 4).

CB609-02 CB609/Alaca & Williams August 27, 2003 16:51 Char Count= 0

40 Euclidean Domains

Also
m X 2 − Y 2 ≡ −Y 2 ≡ −r 2 (mod m).
Hence
m X 2 − Y 2 ≡ (m − 1)r 2 (mod 4m).
Thus

(m − 1)r 2
m X − Y = (m − 1)r − 4m
2 2 2
4m
or

(m − 1)r 2
m X − Y = (m − 1)r − 4m
2 2 2
− 4m.
4m

In the first case we have p || m X 2 − Y 2 . As mp = −1 we have p m. Thus p X
and p Y . Then
2
m m X2 Y
= = = 1,
p p p

contradicting mp = −1.
The second case can be treated similarly.
√
We now use Theorem 2.3.3 to show that Z + Z 1+2 53 is not Euclidean with
respect to φ53 .
√
Theorem 2.3.4 Z + Z 1+ 53
2
is not Euclidean with respect to φ53 .

Proof: We choose
m = 53, p = 5, q = 19, r = 29.
Clearly

m 53 3
= = = −1,
p 5 5

m 53 −4 −1
= = = = −1,
q 19 19 19

(m − 1)r 2 52 · 292
(m − 1)r − 4m
2
= 52 · 29 − 4 · 53 ·
2
4m 4 · 53
= 43732 − 212 · 206
= 43732 − 43672
= 60 = 5 · 22 · 3,
CB609-02 CB609/Alaca & Williams August 27, 2003 16:51 Char Count= 0

2.3 Examples of Domains That are Not Euclidean 41

and

(m − 1)r 2
(m − 1)r 2 − 4m − 4m = 60 − 212 = −152 = −19 · 23 ,
4m
so the result follows from Theorem 2.3.3

√
In the next theorem we show that Z + Z m (m ≡ 2, 3 (mod 4)) is not Euclidean
with
respect
√
to φm if m is sufficiently large. The same result is also true for Z +
1+ m
Z 2
(m ≡ 1 (mod 4)) but the proof is more complicated and we will not
give it here.

Theorem 2.3.5 Let m be a positive squarefree integer.

√
(a) If m ≡ 2 (mod 4) and m ≥ 42 then Z + Z m is not Euclidean with respect
to φm .
√
(b) If m ≡ 3 (mod 4) and m ≥ 91 then Z + Z m is not Euclidean with respect
to φm .

√ √ √ 2 √
Proof:
√ (a)
√ As m ≥ 42 we have m > 20 + 8 6 = 4( 3 + 2) so that m >
2( 3 + 2) and thus
√ √ √ √
3m − 1 2m − 1 3− 2 √
− = m
2 2 2
√ √
( 3 − 2) √ √
> 2( 3 + 2) = 1.
2
Hence there exists an integer u satisfying
√ √
2m − 1 3m − 1
<u< .
2 2
Set t = 2u + 1 so that t is an odd integer satisfying

2m < t 2 < 3m.

√
Now suppose that Z + Z m is Euclidean with respect to φm . Then there exist
√
γ , δ ∈ Z + Z m such that
√
t m = mγ + δ, φm (δ) < φm (m).
√ √
As γ ∈ Z + Z m there exist x, y ∈ Z such that γ = x + y m, and
√ √
φm (t m − m(x + y m)) < φm (m);

that is,

|m 2 x 2 − m(t − my)2 | < m 2 ,

CB609-02 CB609/Alaca & Williams August 27, 2003 16:51 Char Count= 0

42 Euclidean Domains

and thus

|mx 2 − (t − my)2 | < m.

Set X = my − t ∈ Z and Y = x ∈ Z so that

|X 2 − mY 2 | < m

and

X 2 − mY 2 ≡ X 2 ≡ t 2 (mod m).

Since 2m < t 2 < 3m we have

X 2 − mY 2 = t 2 − 2m

X 2 − mY 2 = t 2 − 3m.

In the first case, as t 2 ≡ 1 (mod 8) (since t is odd) and m ≡ 2 (mod 4), we have

X 2 − mY 2 ≡ 5 (mod 8).

Thus X is odd, so X 2 ≡ 1 (mod 8) and

mY 2 ≡ 4 (mod 8).

This is clearly impossible as

0 (mod 8), if Y ≡ 0 (mod 2),
mY ≡2
2 (mod 4), if Y ≡ 1 (mod 2).

In the second case, as t 2 ≡ 1 (mod 8), we have

X 2 − mY 2 ≡ 1 − 3m (mod 8).

As m is even we see that X is odd. Hence X 2 ≡ 1 (mod 8) and

m(Y 2 − 3) ≡ 0 (mod 8).

Hence, as 2 || m, we have

Y 2 ≡ 3 (mod 4),

which is impossible.
CB609-02 CB609/Alaca & Williams August 27, 2003 16:51 Char Count= 0

2.3 Examples of Domains That are Not Euclidean 43

√ √ √ 2 √ √
√ (b) As m ≥ 91 we have m > 44 + 8 30 = 4( 6 + 5) so that m > 2( 6+
5) and thus
√ √ √ √
6m − 1 5m − 1 ( 6 − 5) √
− = m
2 2 2
√ √
( 6 − 5) √ √
> 2( 6 + 5) = 1.
2
Hence there is an integer u satisfying
√ √
5m − 1 6m − 1
<u< .
2 2
Set t = 2u + 1 so that t is an odd integer satisfying
5m < t 2 < 6m.
√
Now suppose that Z + Z m is Euclidean with respect to φm . Then there exist
√
γ , δ ∈ Z + Z m such that
√
t m = mγ + δ, φm (δ) < φm (m).
√
As γ ∈ Z + Z m there exist x, y ∈ Z such that
√
γ = x + y m,
and
√ √
φm (t m − m(x + y m)) < φm (m);
that is,
|m 2 x 2 − m(t − my)2 | < m 2 ,
and thus
|mx 2 − (t − my)2 | < m.
Set X = my − t ∈ Z and Y = x ∈ Z so that
|X 2 − mY 2 | < m
and
X 2 − mY 2 ≡ X 2 ≡ t 2 (mod m).
Since 5m < t 2 < 6m we have
X 2 − mY 2 = t 2 − 5m
or
X 2 − mY 2 = t 2 − 6m.
CB609-02 CB609/Alaca & Williams August 27, 2003 16:51 Char Count= 0

44 Euclidean Domains

In the first case, as t 2 ≡ 1 (mod 8) (since t is odd) and m ≡ 3 (mod 4), we have
X 2 − mY 2 = t 2 − 5m ≡ 1 − 15 = −14 ≡ 2 (mod 4)
so that
X ≡ Y ≡ 1 (mod 2).
Thus X 2 ≡ Y 2 ≡ 1 (mod 8) so that
1 − 5m ≡ t 2 − 5m = X 2 − mY 2 ≡ 1 − m (mod 8),
giving 4m ≡ 0 (mod 8), which is clearly impossible. In the second case, as
t 2 ≡ 1 (mod 8) and m ≡ 3 (mod 4), we have
X 2 − mY 2 = t 2 − 6m ≡ 1 − 18 = −17 ≡ 7 (mod 8).
If X is odd, so X 2 ≡ 1 (mod 8), then
mY 2 ≡ 2 (mod 8),
which is impossible. If X is even, so X 2 ≡ 0 (mod 4), then −3Y 2 ≡ 3 (mod 4), so
that Y 2 ≡ 3 (mod 4), which is impossible.
√
It is a consequence of Theorem 2.2.5 that the domain Z + Z 1+ 2−19 is not
Euclidean with respect to φ−19 . But could it be Euclidean with respect to some
other function? In fact it is not. How do we see this? One way of showing that
an integral domain is not Euclidean with respect to any function is to show that
it does not possess certain distinguished elements called universal side divisors,
since a domain that has no universal side divisors
√ isnot Euclidean with respect to
any function. Indeed as we shall see Z + Z 1+ 2−19 has no universal divisors and
therefore is not Euclidean with respect to any function.
We now define a universal side divisor. For any integral domain D it is convenient
to set
D̃ = U (D) ∪ {0}
so that
D − D̃ = φ if and only if D is a field.

Definition 2.3.1 (Universal side divisor) Let D be an integral domain that is not a
field so that D − D̃ = φ. An element u ∈ D − D̃ is called a universal side divisor
if for any x ∈ D there exists some z ∈ D̃ such that u|x − z.

Theorem 2.3.6 Let D be an integral domain that is not a field. If D has no universal
side divisors then D is not Euclidean.
CB609-02 CB609/Alaca & Williams August 27, 2003 16:51 Char Count= 0

2.3 Examples of Domains That are Not Euclidean 45

Proof: Suppose that D is Euclidean with respect to the Euclidean function φ and
has no universal side divisors. Consider the set of integers defined by

S = {φ(v) | v ∈ D − D̃}.

As D is not a field, D − D̃ = φ and so S is nonempty. By Theorem 2.1.1(d),

S is bounded below. Thus S possesses a least element, say φ(u), u ∈ D − D̃. As
D is Euclidean with respect to φ, for any x ∈ D there exist y, z ∈ D such that
x = uy + z and φ(z) < φ(u). If z = 0 then x = uy and u | x. If z = 0 then by the
minimality of φ(u), z ∈ U (D). Thus in both cases u | x − z for some z ∈ D̃, and
so u is a universal side divisor, which is a contradiction.

√
If m is a negative squarefree integer with m ≡ 2, 3 (mod 4) then Z + Z m is
Euclidean with respect to φm for m = −1 and m = −2 and is not Euclidean with
respect to φm for m < −2 (Theorem 2.2.3). We now use Theorem 2.3.6 to show
√
that Z + Z m is not Euclidean with respect to any function for m < −2.

Theorem 2.3.7 Let m be a negative squarefree integer with m ≡ 2, 3 (mod 4) and

√
m < −2. Then Z + Z m is not Euclidean.

√
Proof: Let D = Z + Z m. As m = −1, U (D) = {1, −1} (see Exercise 3 of Chap-
ter 1) so that D̃ = {0, 1, −1}. Suppose that u is a universal side divisor in D. Then
u must divide one of 2 − 1, 2 + 0, or 2 + 1, that is, one of 1, 2, or 3. But u
being a universal side divisor is not a unit, so u 1. Hence u | 2 or u | 3. Since
m ≡ 2, 3 (mod 4) and m < −2 we have m ≤ −5 so that both 2 and 3 are irre-
ducible in D (Exercise 36 of Chapter 1). Hence the only possible universal side
divisors are 2, −2, 3, and −3. However, none of these divides any of the three
√
elements of Z + Z m :
√ √ √
m − 1, m, m + 1,

so that no such universal side divisor can exist. Hence, by Theorem 2.3.6, D is not
Euclidean.
√
If m is a negative squarefree integer with m ≡ 1 (mod 4) then Z + Z 1+2 m
is Euclidean with respect to φm for m = −3, −7, −11 and is not Euclidean with
respectto φ√m
for m < −11 (Theorem 2.2.5). We use Theorem 2.3.6 to show that
1+ m
Z+Z 2
is not Euclidean with respect to any function for m < −11.

Theorem 2.3.8 Let m be

a√squarefree
negative integer with m ≡ 1 (mod 4) and
1+ m
m < −11. Then Z + Z 2
is not Euclidean.
CB609-02 CB609/Alaca & Williams August 27, 2003 16:51 Char Count= 0

46 Euclidean Domains
√
1+ m
Proof: Let D = Z + Z 2
. As m = −3 we have U (D) = {1, −1} (Exercise
4 of Chapter 1) so that D̃ = {0, 1, −1}. Suppose that u is a universal side divisor in
D. Then u must divide one of 2 − 1, 2 + 0, or 2 + 1, that is, one of 1, 2, or 3. As
u is not a unit,√u must divide 2 or 3. Since m ≤ −15, both 2 and 3 are irreducible
in Z + Z 1+2 m (Exercise 37 of Chapter 1). Therefore the only possible side
divisors are 2, −2, 3, and−3. However, none of these divides any of the following
√
1+ m
three elements of Z + Z 2
,

1 √ 1 √ 1 √ 1 √ 1 √
(−1 + m) = (1 + m) − 1, (1 + m), (3 + m) = (1 + m) + 1,
2 2 2 2 2
so that no such universal side divisor can exist. Hence, by Theorem 2.3.6, D is not
Euclidean.

When m is a √
positive
squarefree integer very little is known. Clark [5] has shown
that Z + Z 1+ 69
2
is Euclidean with respect to the function
√
1+ 69 |a 2 + ab − 17b2 |, if (a, b) = (10, 3),
φ a+b =
2 26, if (a, b) = (10, 3).
√
By Theorem 2.2.7 we know that Z + Z 1+2 69 is not Euclidean with respect
to φ69 . This is the first example of a real quadratic domain that is Euclidean but
not norm-Euclidean. Since the 26√in the definition of φ can be replaced by any
integer greater than 25, Z + Z 1+2 69 is Euclidean with respect to infinitely many
√
different functions. Samuel [16] suggests that Z + Z 14 may be Euclidean with
respect to some function different from φ14 , and this has recently been proved by
Harper [9].

2.4 Almost Euclidean Domains

In this section we introduce the concept of an “almost Euclidean domain” and show
that such a domain must be a principal ideal domain. In Chapter 3 we show that a
principal ideal domain is an almost Euclidean domain (see Theorem 3.3.3). Thus
principal ideal domains are domains that are almost Euclidean in a certain sense.
We first define an “almost Euclidean function” analogously to that of a Euclidean
function (Definition 2.1.1).

Definition 2.4.1 (Almost Euclidean function) Let D be an integral domain. A

mapping φ : D → N ∪ {0} is called an almost Euclidean function on D if it has
CB609-02 CB609/Alaca & Williams August 27, 2003 16:51 Char Count= 0

2.5 Representing Primes by Binary Quadratic Forms 47

the following properties:

φ(0) = 0, (2.4.1)
φ(a) > 0, for all a ∈ D with a = 0, (2.4.2)
φ(ab) ≥ φ(a), for all a, b ∈ D with b = 0, (2.4.3)
if a, b ∈ D with b = 0 then either (2.4.4)
(i) a = bq for some q ∈ D or
(ii) 0 < φ(ax + by) < φ(b) for some x, y ∈ D.

It is clear from Definition 2.1.1 and Theorem 2.1.1(d) that if φ is a Euclidean

function satisfying φ(0) = 0 then φ is an almost Euclidean function.
The concept of an almost Euclidean domain occurs in the work of Cámpoli [2]
and Greene [8].

Definition 2.4.2 (Almost Euclidean domain) Let D be an integral domain. If D

possesses an almost Euclidean function φ then D is called an almost Euclidean
domain with respect to φ.

As for Euclidean domains, if it is not important to specify the almost Euclidean

function φ, we just call D an almost Euclidean domain.

Theorem 2.4.1 An almost Euclidean domain is a principal ideal domain.

Proof: Let D be an almost Euclidean domain. Let φ be an almost Euclidean function

defined on D. Let I be a nonzero ideal of D. Among the elements x of I , let b be an
element with a minimal positive value of φ(x). Given a ∈ I , for any x, y ∈ D, the
element ax + by is in I . By the definition of b, we cannot have 0 < φ(ax + by) <
φ(b) so that as D is almost Euclidean with respect to φ, we must have a = bq for
some q ∈ D. Thus I = b and D is a principal ideal domain.

For the converse of this theorem, see Theorem 3.3.3.

2.5 Representing Primes by Binary Quadratic Forms

Expressions of the type ax 2 + bx y + cy 2 (a, b, c ∈ Z) are called binary quadratic
forms. The integer n is said to be represented by the binary quadratic form ax 2 +
bx y + cy 2 if there are integers x and y such that n = ax 2 + bx y + cy 2 . Thus for
example 31 is represented by the form x 2 + x y + 3y 2 as 31 = 12 + 1 · 3 + 3 · 32 ,
but 2 is not represented by the form x 2 + 5y2 . √
√
As Z + Z m (m = −1, −2) and Z + Z 1+2 m (m = −3, −7, −11) are Eu-
clidean domains (Theorems 2.2.3 and 2.2.5), we can apply Theorems 1.4.4 and 1.4.5
to determine when an odd prime p is represented by each of the forms x 2 + y 2 ,
CB609-02 CB609/Alaca & Williams August 27, 2003 16:51 Char Count= 0

48 Euclidean Domains

x 2 + 2y 2 , x 2 + x y + y 2 , x 2 + x y + 2y 2 , and x 2 + x y + 3y 2 . To do this we begin

by recalling the following Legendre symbol evaluations from elementary number
theory. For an odd prime p

−1
= 1 ⇐⇒ p ≡ 1 (mod 4), (2.5.1)
p

−2
= 1 ⇐⇒ p ≡ 1, 3 (mod 8), (2.5.2)
p

−3
= 1 ⇐⇒ p ≡ 1 (mod 3), (2.5.3)
p

−7
= 1 ⇐⇒ p ≡ 1, 2, 4 (mod 7), (2.5.4)
p

−11
= 1 ⇐⇒ p ≡ 1, 3, 4, 5, 9 (mod 11). (2.5.5)
p

Theorem 2.5.1 Let p be a prime such that p ≡ 1 (mod 4). Then there exist integers
x and y such that p = x 2 + y 2 .
√
Proof: As p ≡ 1 (mod 4), by (2.5.1) we have −1 p
= 1. Since Z + Z −1 is a
Euclidean domain, by Theorem 2.1.2 it is a principal ideal domain. Thus by Theorem
1.4.4, there are integers x and y such that p = x 2 + y 2 .

Theorem 2.5.1 is called the Girard–Fermat theorem. Heath-Brown [10] gave

an interesting new proof of this theorem in 1984. Varouchas [18] and Williams
[20] have given presentations of Heath-Brown’s proof. Zagier [22] has given a
one-sentence proof.

Theorem 2.5.2 Let p be a prime such that p ≡ 1, 3 (mod 8). Then there exist
integers x and y such that p = x 2 + 2y 2 .

Proof: The proof is the same√as that of Theorem 2.5.1

√ except that (2.5.2) is used
in place of (2.5.1) and Z + Z −2 in place of Z + Z −1.

Jackson [11] has given a short proof of Theorem 2.5.2 when p ≡ 3 (mod 8).
Similarly using (2.5.3)–(2.5.5) and Theorem 1.4.5, we obtain the following three
theorems.

Theorem 2.5.3 Let p be a prime such that p ≡ 1 (mod 3). Then there exist integers
x and y such that p = x 2 + x y + y 2 .

Theorem 2.5.4 Let p be a prime such that p ≡ 1, 2, 4 (mod 7). Then there exist
integers x and y such that p = x 2 + x y + 2y 2 .
CB609-02 CB609/Alaca & Williams August 27, 2003 16:51 Char Count= 0

Exercises 49

Theorem 2.5.5 Let p be a prime such that p ≡ 1, 3, 4, 5, 9 (mod 11). Then there
exist integers x and y such that p = x 2 + x y + 3y 2 .
√
In Theorem 2.2.8 we showed that Z + Z m is Euclidean for m = 2, 3, 6. Recall
from elementary number theory that for an odd prime p

2
= 1 ⇐⇒ p ≡ 1, 7 (mod 8),
p

3
= 1 ⇐⇒ p ≡ 1, 11 (mod 12),
p

6
= 1 ⇐⇒ p ≡ 1, 5, 19, 23 (mod 24).
p

Hence, by Theorem 1.4.4, we obtain the following three theorems.

Theorem 2.5.6 Let p be a prime such that p ≡ 1, 7 (mod 8). Then there exist
integers x and y such that p = x 2 − 2y 2 .

We used the fact that T 2 − 2U 2 = −1 for T = U = 1.

Theorem 2.5.7 Let p be a prime such that p ≡ 1, 11 (mod 12). Then there exist
integers x and y such that either p = x 2 − 3y 2 or p = 3y 2 − x 2 .

In this case there are no integers T and U such that T 2 − 3U 2 = −1.

Theorem 2.5.8 Let p be a prime such that p ≡ 1, 5, 19, 23 (mod 24). Then there
exist integers x and y such that either p = x 2 − 6y 2 or p = 6y 2 − x 2 .

There are no integers T and U such that T 2 − 6U 2 = −1.

Exercises
1. Let D be an integral domain possessing a Euclidean function φ. Give an example to
show that

φ(a) = φ(b) (a, b ∈ D) =⇒

a ∼ b.

2. Prove Theorem 2.2.4.

3. Prove Theorem 2.2.5.
4. Give an example √
to show that q and r in (2.1.2) are not necessarily unique.
5. Prove that Z + Z 7 is Euclidean with respect to φ7 using the method of the proof of
Theorem 2.2.8.
CB609-02 CB609/Alaca & Williams August 27, 2003 16:51 Char Count= 0

50 Euclidean Domains
√
6. Prove that Z + Z 1+2 5 is Euclidean with respect to φ5 using the method of the proof
of Theorem 2.2.8. √
7. Use Theorem 2.3.1 to show that Z + Z 26 is not Euclidean with respect to φ26 .
8. Prove a modification of Theorem 2.3.1 that allows one of the primes p and q to be the
prime 2.
9. Prove an extension of Theorem 2.3.1 that replaces p and q in the equation pt + qu = m
by powers of p and q with odd exponents. √
10. Use Theorem 2.3.3 to prove that Z + Z 1+2 77 is not Euclidean with respect to φ77 .
11. Prove that if p is a prime with p ≡ 3 (mod 4) then there do not exist integers x and y
such that p = x 2 + y 2 .
12. Let p be a prime. Use Theorem 2.5.1 and Exercise 11 to deduce that

p = x 2 + y 2 ⇐⇒ p = 2 or p ≡ 1 (mod 4).

13. Prove that if p is a prime with p ≡ 5, 7 (mod 8) then there do not exist integers x and
y such that p = x 2 + 2y 2 .
14. Let p be a prime. Use Theorem 2.5.2 and Exercise 13 to deduce that

p = x 2 + 2y 2 ⇐⇒ p = 2 or p ≡ 1, 3 (mod 8).

15. Prove that if p is a prime with p ≡ 2 (mod 3) then there do not exist integers x and y
such that p = x 2 + x y + y 2 .
16. Let p be a prime. Use Theorem 2.5.3 and Exercise 15 to deduce that

p = x 2 + x y + y 2 ⇐⇒ p = 3 or p ≡ 1 (mod 3).

17. Prove that if p is a prime with p ≡ 3, 5, 6 (mod 7) then there do not exist integers x
and y such that p = x 2 + x y + 2y 2 .
18. Let p be a prime. Use Theorem 2.5.4 and Exercise 17 to deduce that

p = x 2 + x y + 2y 2 ⇐⇒ p = 7 or p ≡ 1, 2, 4 (mod 7).

19. Prove that if m is a positive integer possessing a prime divisor q ≡ 3 (mod 4) then there
are no integers T and U such that T 2 − mU 2 = −1.
20. Let p be a prime with p ≡ 1, 11 (mod 12). Deduce from Theorem 2.5.7 that

p = x 2 − 3y 2 , if p ≡ 1 (mod 12),
p = 3y 2 − x 2 , if p ≡ 11 (mod 12),

for some integers x and y.

21. Let p be a prime with p ≡ 1, 5, 19, 23 (mod 24). Deduce from Theorem 2.5.8 that

p = x 2 − 6y 2 , if p ≡ 1, 19 (mod 24),
p = 6y 2 − x 2 , if p ≡ 5, 23 (mod 24),

for some integers x and y.

CB609-02 CB609/Alaca & Williams August 27, 2003 16:51 Char Count= 0

Suggested Reading 51
√ √
22. Prove that the subdomain Z + 3Z −2 of the Euclidean domain Z + Z −2 is not
Euclidean. √ √
23. Prove that the subdomain Z + 7Z 2 of the Euclidean domain Z + Z 2 is not
Euclidean. √ √
24. Prove that the subdomain Z + 2Z 3 of the Euclidean domain Z + Z 3 is not
Euclidean. √ √
25. Prove that the subdomain Z + 5Z 6 of the Euclidean domain Z + Z 6 is not
Euclidean.
26. Let m be a positive integer with m ≡ 1 (mod 4). Show that the solvability of the equation
T 2 + T U + 14 (1 − m)U 2 = −1 in integers T and U (see Theorem 1.4.5) is equivalent
to the solvability of the equation X 2 − mY 2 = −4 in integers X and Y .
27. Let m be an integer with m ≡ 1 (mod 4) that possesses a prime divisor q ≡ 3 (mod 4).
Prove that there are no integers T and U such that T 2 + T U + 14 (1 − m)U 2 = −1.
28. Prove that if p is a prime with p ≡ 1, 4 (mod 5) then there are integers x and y such
that p = x 2 + x y − y 2 . [Hint: Use Theorems 1.4.5 and 2.2.7.]
29. Use Exercise 12 to show that the irreducibles in Z + Zi are 1 + i and its asso-
ciates, x ± i y, where x 2 + y 2 = p (prime) ≡ 1 (mod 4), and their associates, and
q (prime) ≡ 3 (mod 4) and its associates.
√
30. Use Exercise 14 to determine the irreducibles in Z + Z −2.

Suggested Reading
1. P. J. Arpaia, A note on quadratic Euclidean domains, American Mathematical Monthly
75 (1968), 864–865.
Examples of quadratic Euclidean domains that√possess subdomains that are not Euclidean
√ are
given. For example the Gaussian domain Z + Z −1 is Euclidean but its subdomain Z + 2Z −1
is not Euclidean.
2. O. A. Cámpoli, A principal ideal domain that is not a Euclidean domain, American
Mathematical Monthly 95 (1988), 868–871.
√
It is shown in an elementary fashion that Z + Z( 1+ 2−19 ) is a principal ideal domain but not a
Euclidean domain. The idea of a domain being almost Euclidean is introduced (p. 870).
3. H. Chatland, On the Euclidean algorithm in quadratic number fields, Bulletin of the
American Mathematical Society 55 (1949), 948–953.
This paper is a valuable source of references
√
to work on the Euclidean algorithm in quadratic
domains. It should be noted that Z + Z( 1+2 97 ) is not Euclidean, contrary to the claim by Rédei.
This was established by Barnes and Swinnerton-Dyer in 1952.
4. H. Chatland and H. Davenport, Euclid’s algorithm in real quadratic fields, Canadian
Journal of Mathematics 2 (1950), 289–296.
This is where the final steps in the proofs of Theorems 2.2.6 and 2.2.7 are given.
5. D. A. Clark, A quadratic field which is Euclidean but not norm-Euclidean, Manuscripta
Mathematica 83 (1994),√ 327–330.
It is shown that Z + Z( 1+2 69 ) is Euclidean but not norm-Euclidean.
6. D. A. Cox, Primes of the form x 2 + ny 2 , Wiley, New York, 1989.
This book presents a comprehensive treatment of the problem of deciding which primes are
represented by x 2 + ny 2 .
CB609-02 CB609/Alaca & Williams August 27, 2003 16:51 Char Count= 0

52 Euclidean Domains

7. D. W. Dubois and A. Steger, A note on division algorithms in imaginary quadratic

number fields, Canadian Journal of Mathematics 10 (1958), 285–286.
√
Let m be a negative
√
squarefree integer. The authors prove that if Z + Z m (m ≡ 2, 3 (mod 4))
and Z + Z( 2 ) (m ≡ 1 (mod 4)) are Euclidean they must be Euclidean with respect to φm .
1+ m

8. J. Greene, Principal ideal domains are almost Euclidean, American Mathematical

Monthly 104 (1997), 154–156.
The author proves that an integral domain is a principal ideal domain if and only if it is almost
Euclidean.
√
9. M. Harper, A proof that Z[ 14] is a Euclidean domain, Ph.D. thesis, McGill University,
Montréal, Canada, 2000.
√
It is shown that Z + Z 14 is Euclidean.
10. D. R. Heath-Brown, Fermat’s two-squares theorem, Invariant (1984), 3–5.
A beautifully simple proof of the Girard-Fermat theorem is given.
11. T. Jackson, A short proof that every prime p ≡ 3 (mod 8) is of the form x 2 + 2y 2 ,
American Mathematical Monthly 107 (2000), 447.
Heath-Brown’s ideas are used to prove Euler’s result that every prime p ≡ 3 (mod 8) is repre-
sented by x 2 + 2y 2 .
12. M. A. Jodeit, Uniqueness in the division algorithm, American Mathematical Monthly
74 (1967), 835–836.
Let D be an integral domain. Suppose that φ is a Euclidean function on D such that the quotient
q and remainder r in (2.1.2) are always unique. Then D is either a field or a polynomial domain
F[x], where F is a field.
13. Th. Motzkin, The Euclidean algorithm, Bulletin of the American Mathematical Society
55 (1949), 1142–1146. √
In this classic paper on Euclidean domains, it is shown that Z + Z( 1+ 2−19 ) is a principal ideal
domain but not a Euclidean domain. Universal side divisors are introduced.
14. T.-S. Rhai, A characterization of polynomial domains over a field, American Mathe-
matical Monthly 69 (1962), 984–986.
Let D be an integral domain. Suppose that φ is a Euclidean function on D such that the quotient
q and remainder r in (2.1.2) are always unique. Then D is either a field or a polynomial domain
F[x], where F is a field.
15. K. Rogers, The axioms for Euclidean domains, American Mathematical Monthly 78
(1971), 1127–1128.
The role of the condition (2.1.1) in the definition of a Euclidean function is discussed.
16. P. Samuel, About Euclidean rings, Journal of Algebra 19 (1971), 282–301.
√
In this classic paper on Euclidean rings, the author suggests (p. 294) that Z + Z 14 may be
Euclidean but not norm-Euclidean. This has recently been established by Harper [9].
17. S. Singh, Non-Euclidean domains: An example, Mathematics Magazine 49 (1976), 243.
√
It is shown that Z + Z m is not Euclidean for negative squarefree integers m with m < −2 and
m ≡ 2, 3 (mod 4).
18. Y. Varouchas, Une démonstration élémentaire du théorème des deux carrés, La Caverne,
I. R. E. M. de Lorraine, France, Bulletin No. 6, février 1984, pp. 31–39.
A presentation in French of Heath-Brown’s proof of the Girard–Fermat theorem is given.
19. K. S. Williams, Note on non-Euclidean principal ideal domains, Mathematics Magazine
48 (1975), 176–177. √
It is shown that the domains Z + Z( 1+2 m ) (m = −19, −43, −67, −163) are not Euclidean.
CB609-02 CB609/Alaca & Williams August 27, 2003 16:51 Char Count= 0

Biographies 53

20. K. S. Williams, Heath–Brown’s elementary proof of the Girard–Fermat theorem, Car-

leton Coordinates, Department of Mathematics and Statistics, Carleton University, Ot-
tawa, Ontario, Canada, January 1985, pp. 4–5.
A presentation of Heath-Brown’s proof of the Girard–Fermat theorem is given.
21. J. C. Wilson, A principal ideal ring that is not a Euclidean ring, Mathematics Magazine
46 (1973), 34–38. √
The author shows that Z + Z( 1+ 2−19 ) is a principal ideal domain that is not Euclidean.
22. D. Zagier, A one-sentence proof that every prime p ≡ 1 (mod 4) is a sum of two squares,
American Mathematical Monthly 97 (1990), 144.
A one-sentence rendition of Heath-Brown’s proof of the Girard-Fermat theorem is given.

Biographies
1. K. Barner, Pierre de Fermat (1601?–1665)—His life beside mathematics, Canadian
Mathematical Society Notes 34 (2002), 3–4, 26–30.
The author relates an interesting account of the nonmathematical life of Fermat.
2. P. Hoffman, The Man Who Loved Only Numbers, Hyperion, New York, 1998.
The story of Paul Erdös, one of the most prolific and eccentric mathematicians of the twentieth
century, is told.
3. M. S. Mahoney, The Mathematical Career of Pierre de Fermat (1601–1605), Princeton
University Press, Princeton, New Jersey, 1973.
For two completely different reviews of this book, see Isis 65 (1974), 398–400 and Bulletin of the
American Mathematical Society 79 (1973), 1138–1149.
4. C. A. Rogers, Harold Davenport, Bulletin of the London Mathematical Society 4 (1972),
66–99.
A memoir on the life and mathematics of Davenport is presented.
5. The website

http://www-groups.dcs.st-and.ac.uk/˜history/

has biographies of A. T. Brauer, L. E. Dickson, P. Erdös, H. A. Heilbronn, O. Perron,

and R. Remak.
CB609-03 CB609/Alaca & Williams August 7, 2003 17:19 Char Count= 0

3
Noetherian Domains

3.1 Noetherian Domains

Let I1 be a nonzero ideal in the domain Z. We consider ideals I such that I1 ⊆ I .
As Z is a principal ideal domain (Theorem 1.4.1), there are nonzero integers m and
n such that I1 = m, I = n, and n | m. Now m has only finitely many divisors
n so there exist only finitely many ideals I with I1 ⊆ I . Thus there cannot exist
infinitely many ideals Ik (k = 2, 3, . . .) such that

I1 ⊂ I2 ⊂ I3 ⊂ I4 ⊂ . . . . (3.1.1)

The importance of domains such as Z that do not contain infinite ascending chains
of ideals of the type (3.1.1) was first recognized by the German mathematician
Emmy Noether (1882–1935). Such domains are now called Noetherian domains in
her honor. We note that some domains do contain infinite chains of ideals of the
type (3.1.1). For example, if F is a field, the domain F[X 1 , X 2 , . . .] contains the
infinite chain of ideals

X 1 ⊂ X 1 , X 2 ⊂ X 1 , X 2 , X 3 ⊂ . . . .

Definition 3.1.1 (Ascending chain of ideals) An infinite sequence of ide-

als {In : n = 1, 2, . . .} in an integral domain is said to be an ascending
chain if
I1 ⊆ I2 ⊆ . . . ⊆ In ⊆ . . . .
The chain is said to be a strictly ascending chain if
I1 ⊂ I2 ⊂ . . . ⊂ In ⊂ . . . .

Definition 3.1.2 (Terminating ascending chain) An ascending chain of ideals

I1 ⊆ I2 ⊆ . . . ⊆ In ⊆ . . .

54
CB609-03 CB609/Alaca & Williams August 7, 2003 17:19 Char Count= 0

3.1 Noetherian Domains 55

in an integral domain is said to terminate if there exists a positive integer n 0 such

that

In = In 0 for all n ≥ n 0 .

Definition 3.1.3 (Ascending chain condition) An integral domain D is said to sat-

isfy the ascending chain condition if every ascending chain of ideals in D terminates
or, equivalently, if D does not contain a strictly ascending chain of ideals.

Definition 3.1.4 (Noetherian domain) An integral domain that satisfies the as-
cending chain condition is called a Noetherian domain.
More generally we define a Noetherian ring to be a ring R in which every
ascending chain of (two-sided) ideals in R terminates.

From the remarks preceding the definitions, we have the following two examples.

Example 3.1.1 Z is a Noetherian domain.

Example 3.1.2 If F is a field, the domain F[X 1 , X 2 , . . .] is not Noetherian.

The next theorem gives a necessary and sufficient condition for an integral domain
to be a Noetherian domain.

Theorem 3.1.1 Let D be an integral domain. Then D is Noetherian if and only if

every ideal of D is finitely generated.

Proof: Let D be a Noetherian domain. Suppose that not every ideal of D is finitely
generated. Let I be an ideal of D that is not finitely generated. Thus I = 0, and
so there exists a1 ∈ I with a1 = 0. Let A1 be the ideal given by A1 = a1 . Clearly
A1 ⊆ I . Moreover, I = A1 as A1 is finitely generated and I is not. Hence A1 ⊂ I .
Take a2 ∈ I, a2 ∈ A1 , and let A2 be the ideal given by A2 = a1 , a2 . Clearly
A1 ⊂ A2 ⊂ I . Continuing in this way, we obtain an infinite strictly increasing
sequence of ideals A1 ⊂ A2 ⊂ . . . , contradicting that D is a Noetherian domain.
Hence every ideal of a Noetherian domain must be finitely generated.
Now let D be an integral domain in which every ideal is finitely generated.
Let

I1 ⊆ I2 ⊆ I3 ⊆ . . .

be an ascending chain of ideals in D. It is easy to check that ∞ n=1 In is an ideal

of D. Hence ∞ n=1 I n is finitely generated, so there exist finitely many elements
CB609-03 CB609/Alaca & Williams August 7, 2003 17:19 Char Count= 0

56 Noetherian Domains

a1 , a2 , . . . , am of D such that
∞

In = a1 , a2 , . . . , am .
n=1

For each i = 1, 2, . . . , m, ai ∈ ∞ n=1 In , say, ai ∈ Ini . Set l = max (n 1 , n 2 , . . . , n m ).
∞
Clearly Il ⊆ n=1 In . As n i ≤ l we have Ini ⊆ Il , and thus ai ∈ Il for i =
∞
1, 2, . . . , m. Hence a1 , . . . , am ⊆ Il so that n=1 In ⊆ Il . This proves that
∞
n=1 In = Il , and thus In = Il for n ≥ l. Hence D is Noetherian.

From Theorem 3.1.1 we see that principal ideal domains are Noetherian.

Theorem 3.1.2 Let D be a principal ideal domain. Then D is a Noetherian domain.

Proof: As D is a principal ideal domain, every ideal in D is principal and therefore

finitely generated. Hence, by Theorem 3.1.1, D is Noetherian.

Example 3.1.3 By Theorems 2.1.2 and 3.1.2 a Euclidean domain is always Noethe-
rian. Thus
Z (Theorem 2.2.1(a)),
√ √
Z + Z −1, Z + Z −2 (Theorem
2.2.3),
√ √
1 + −3 1 + −7
Z+Z , Z+Z ,
2 2
√
1 + −11
Z+Z (Theorem 2.2.5),
2
√ √ √
Z + Z 2, Z + Z 3, Z + Z 6 (Theorem 2.2.8)
are all examples of Noetherian domains.

Our next objective is to give another condition (called the maximal condition)
that allows us to recognize when an integral domain is Noetherian.

Definition 3.1.5 (Maximal condition) An integral domain D is said to satisfy the

maximal condition if every nonempty set S of ideals of D contains an ideal that is
not properly contained in any other ideal of the set S; that is, S possesses an ideal
I such that if J is an ideal in S with I ⊆ J then J = I .

We show that satisfying the maximal condition is equivalent to the domain being
Noetherian.

Theorem 3.1.3 Let D be an integral domain. Then D is Noetherian if and only if

D satisfies the maximal condition.
CB609-03 CB609/Alaca & Williams August 7, 2003 17:19 Char Count= 0

3.2 Factorization Domains 57

Proof: Suppose that D is a Noetherian domain that does not satisfy the maximal
condition. Then D possesses a nonempty set S of ideals with the property that for
every ideal I of S there exists an ideal J of S with I ⊂ J . This property enables
us to construct inductively an infinite strictly ascending chain of ideals in S, which
contradicts D being a Noetherian domain. Hence every Noetherian domain must
satisfy the maximal condition.
Now let D be an integral domain that satisfies the maximal condition. Let I1 ⊆
I2 ⊆ I3 ⊆ . . . be an ascending chain of ideals of D. Set S = {In | n = 1, 2, 3, . . .}.
As D satisfies the maximal condition, S contains an ideal Im , which is not properly
contained in any other ideal of S. As Im ⊆ I j for j ≥ m we must have I j = Im
for j ≥ m. Hence the ascending chain I1 ⊆ I2 ⊆ I3 ⊆ . . . terminates and D is
Noetherian.

A famous theorem of David Hilbert (1862–1943) asserts that if D is a Noetherian

domain then the polynomial domain D[X 1 , . . . , X n ] is also Noetherian. This is the
celebrated Hilbert basis theorem. We will not prove this theorem here; a proof can
be found for example in [8, pp. 201–202].

Example 3.1.4 Z[X 1 , . . . , X n ] is a Noetherian domain. We can see this as follows.

By Theorem 2.2.1(a) Z is a Euclidean domain. Thus, by Theorem 2.1.2, Z is a
principal ideal domain. Hence, by Theorem 3.1.2, Z is a Noetherian domain. Finally,
by the Hilbert basis theorem, Z[X 1 , . . . , X n ] is a Noetherian domain.

Example 3.1.5 F[X 1 , X 2 , . . . , X n ] (n ≥ 1), where F is a field, is a Noetherian

domain. We can prove this as follows. By Theorem 2.2.1(b), F[X 1 ] is a Euclidean
domain. Hence, by Theorem 2.1.2, F[X 1 ] is a principal ideal domain, and so,
by Theorem 3.1.2, is a Noetherian domain. Then, by the Hilbert basis theorem,
(F[X 1 ])[X 2 , . . . , X n ] is a Noetherian domain; that is, F[X 1 , X 2 , . . . , X n ] is a
Noetherian domain.

3.2 Factorization Domains

Let D be an integral domain that is not a field so that D contains nonzero, nonunit
elements. It may be the case that all of these elements are reducible so that D
contains no irreducibles. The next example illustrates this.

Example 3.2.1 Let D be the domain of polynomials in positive rational powers

of x over C, that is,

D = {a1 x r1 + · · · + an x rn | n ∈ N, a1 , . . . , an ∈ C, r1 , . . . , rn ∈ Q,
0 ≤ r1 < · · · < rn }.
CB609-03 CB609/Alaca & Williams August 7, 2003 17:19 Char Count= 0

58 Noetherian Domains

Clearly U (D) = C∗ . We show that D does not possesss any irreducible elements.
Suppose that

f (x) = a1 x r1 + · · · + an x rn

is an irreducible element of D. As f (x) is a nonzero element of D, we may suppose

that an = 0. If n = 1 and r1 = 0 then f (x) = a1 = 0 is a unit of D, a contra-
diction. If n = 1 and r1 > 0 then f (x) = a1 x r1 = a1 (x r1 /2 )2 is reducible in D, a
contradiction. Hence n ≥ 2 and rn > 0. Let

t = least common multiple of the ( positive) denominators of the

rationals r1 , . . . , rn

so that t is a positive integer such that r1 t, . . . , rn t are integers with

0 ≤ r1 t < r2 t < · · · < rn t.

Then

f (x t ) = a1 x r1 t + · · · + an x rn t ∈ C[x].

Hence there exist b1 , . . . , brn t ∈ C such that

a1 x r1 t + · · · + an x rn t = an (x − b1 ) · · · (x − brn t ).

Thus

f (x) = an (x 1/t − b1 ) · · · (x 1/t − brn t ).

Since
1/2 1/2
x 1/t − b1 = (x 1/2t − b1 )(x 1/2t + b1 ),
1/2
where x 1/2t ± b1 are nonzero, nonunit elements of D, f (x) is reducible, a con-
tradiction.
Thus D does not possess any irreducible elements.

We show next that a Noetherian domain always contains irreducibles.

Theorem 3.2.1 Let D be an integral domain that is not a field. If D is Noetherian

then D contains elements that are irreducible.

Proof: Suppose that the integral domain D does not contain any irreducibles. As
we are assuming that D is not a field, D has nonzero, nonunit elements. Let a be
one of these. Then a is not an irreducible. Hence a is reducible. Thus there exists a
nonzero, nonunit element a1 of D such that a1 | a and a1 ∼ a. Clearly a ⊂ a1 .
As a1 is not an irreducible, a1 is reducible, and we can repeat the preceding argument
CB609-03 CB609/Alaca & Williams August 7, 2003 17:19 Char Count= 0

3.2 Factorization Domains 59

to obtain a nonzero, nonunit element a2 of D such that a2 | a1 and a2 ∼ a1 . Thus

a1 ⊂ a2 . Continuing in this way we obtain an ascending chain of principal ideals

a ⊂ a1 ⊂ a2 ⊂ · · · ,

contradicting that D is a Noetherian domain. Hence D contains irreducibles.

By Theorem 3.2.1 the domain D in Example 3.2.1 cannot be Noetherian. This

is easily seen directly as it contains the infinite ascending chain of ideals

x ⊂ x 1/2 ⊂ x 1/4 ⊂ x 1/8 ⊂ · · · .

Clearly for this domain it is not possible to express each nonzero, nonunit element
as a finite product of irreducibles. Domains in which this is possible are called
factorization domains. The main result of this section is that a Noetherian domain
is always a factorization domain; that is, in a Noetherian domain every nonzero,
nonunit element can be expressed as a finite product of irreducibles. The converse
of this result however is not true as we demonstrate in Example 3.2.2.

Definition 3.2.1 (Factorization domain) Let D be an integral domain. Then D is

said to be a factorization domain if every nonzero, nonunit element of D can be
expressed as a finite product of irreducible elements of D.

Our next result shows that a Noetherian domain is always a factorization domain.

Theorem 3.2.2 Let D be a Noetherian domain. Then D is a factorization domain.

Proof: Let D be a Noetherian domain and suppose that D is not a factorization

domain. Then D contains at least one nonzero, nonunit element that is not a finite
product of irreducible elements of D. Let A be the set of all such elements, so A is
not empty. Let

S = {a | a ∈ A}.

Clearly S is a nonempty set of principal ideals of D. As D is a Noetherian domain,

by the maximal condition (Theorem 3.1.3), S has a maximal element, say b. As
b ∈ S, b ∈ A so that b is a nonzero, nonunit element of D that is not a product of
irreducibles. Hence b is not irreducible. Thus we can write b in the form b = cd,
where c and d are nonzero, nonunit elements of D. Hence b = cd ⊆ c and
b = cd ⊆ d. Moreover, as d is not a unit, b and c are not associates; thus
b = c, and so b ⊂ c. Similarly, b ⊂ d. By the maximality of b, we
have c ∈ S and d ∈ S. Hence c and d are products of irreducible elements of
D. Thus b = cd is also a product of irreducible elements of D, contradicting our
assumption. Thus D is a factorization domain.
CB609-03 CB609/Alaca & Williams August 7, 2003 17:19 Char Count= 0

60 Noetherian Domains

Example 3.2.2 From Examples 3.1.1 and 3.1.3,

√
and Theorem
√ 3.2.2,
we
√ √ 1+ −3 1+ −7
see that Z, Z + Z −1, Z + Z −2, Z + Z , Z+Z , Z+
√ √ √ √
2 2

Z 1+ 2−11 , Z + Z 2, Z + Z 3, and Z + Z 6 are all examples of factorization

domains.

The next example shows that a factorization domain is not always a Notherian
domain.

Example 3.2.3 Let F be a field. We show that F[X 1 , X 2 , . . .] is a factorization

domain. Let a be a nonzero, nonunit element of F[X 1 , X 2 , . . .]. Then a is a poly-
nomial in finitely many of the indeterminates X 1 , X 2 , . . . with coefficients in F.
Thus there is a positive integer m such that a ∈ F[X 1 , X 2 , . . . , X m ]. From Ex-
ample 3.1.5 we know that F[X 1 , . . . , X m ] is a Noetherian domain. Therefore, by
Theorem 3.2.2, it is a factorization domain. Hence every nonzero, nonunit element
of F[X 1 , X 2 , . . . , X m ] can be expressed as a product of irreducible elements. Thus
a is a product of irreducible elements of F[X 1 , X 2 , . . . , X m ]. But an irreducible
element of F[X 1 , X 2 , . . . , X m ] is an irreducible element of F[X 1 , X 2 , . . .], so a
is a product of irreducible elements of F[X 1 , X 2 , . . .]. Hence F[X 1 , X 2 , . . .] is a
factorization domain. By Example 3.1.2, F[X 1 , X 2 , . . .] is not Noetherian.

Theorem 3.2.3 Let D be a principal ideal domain. Then D is a factorization

domain.

Proof: This result follows from Theorems 3.1.2 and 3.2.2.

3.3 Unique Factorization Domains

Let D be a factorization domain. Let a be a nonzero, nonunit element of D. Then
there exist irreducible elements of D such that
a = h1h2 · · · hk .
If h 1 and h 2 are associates, say h 2 = vh 1 , where v is a unit of D, then
a = vh 21 h 3 · · · h k .
Repeating this process we eventually obtain a factorization
a = wl1k1 · · · lmkm ,
where w is a unit of D, the ki are positive integers, and the li are irreducible elements
of D with no two distinct ones being associates. Suppose
k km
a = wl1 1 · · · lm
CB609-03 CB609/Alaca & Williams August 7, 2003 17:19 Char Count= 0

3.3 Unique Factorization Domains 61

is another such factorization of a into powers of nonassociated irreducible elements

of D. If m = m and after a possible rearragement of l1 , . . . , lm , we have

l1 ∼ l1 , . . . , lm ∼ lm and k1 = k1 , . . . , km = km ,

we say that a has a unique factorization as a product of irreducible elements of D.

Definition 3.3.1 (Unique factorization domain) Let D be a factorization domain.

Suppose that every nonzero, nonunit element a of D has a unique factorization as
a product of irreducible elements of D. Then D is called a unique factorization
domain.

Theorem 3.3.1 Let D be a principal ideal domain. Then D is a unique factorization

domain.

Proof: Suppose that D is a principal ideal domain. By Theorem 3.2.3 D is a fac-

torization domain. Suppose however that D is not a unique factorization domain.
Then there exists at least one nonzero, nonunit element a, which has at least two
different factorizations as a product of irreducible elements of D. Let A be the set
of all such elements a, and let

S = {a | a ∈ A}.

As A is a nonempty set, so is S. Now D is a principal ideal domain, so by Theorem

3.1.2, D is a Noetherian domain. Hence, by the maximal condition (Theorem 3.1.3),
S contains a maximal element, say b. Thus b ∈ A and b has two essentially
different factorizations as a product of irreducibles, say,
j
b = ul1k1 · · · lmkm = vh 11 · · · h njn , (3.3.1)

where u and v are units of D, l1 , . . . , lm , h 1 , . . . , h n are irreducible elements of D,

k1 , . . . , km , j1 , . . . , jn are positive integers, li ∼ l j (i = j), and h i ∼ h j (i = j). As
k1 > 0, we see that l1 | b, and thus l1 | vh d11 . . . h njn . As D is a principal ideal domain
and l1 is irreducible, by Theorem 1.4.2, l1 is prime. Thus l1 | h s for some integer s
with 1 ≤ s ≤ n. After relabeling the h’s, if necessary, we may suppose that l1 | h 1 .
Since l1 and h 1 are both irreducibles this means that l1 ∼ h 1 , say, h 1 = l1 w, where
w is a unit of D. Replacing h 1 by l1 w in (3.3.1), we obtain

b/l1 = ul1k1 −1 · · · lmkm = vwh 1d1 −1 · · · h njn .

As l1 is not a unit we have b ⊂ b/l1 . Hence, by the maximality of b, we have
after suitable rearrangement of the h’s

k1 − 1 = j1 − 1, k2 = j2 , . . . , km = jm , m = n,
l1 ∼ h 1 , l2 ∼ h 2 , . . . , lm ∼ h m .
CB609-03 CB609/Alaca & Williams August 7, 2003 17:19 Char Count= 0

62 Noetherian Domains

This contradicts the assumption that b has two essentially different factorizations.
This completes the proof that a principal ideal domain is always a unique factor-
ization domain.

Clearly from Theorems 2.1.2 and 3.3.1 we see that a Euclidean domain is always
a unique factorization domain. Thus the domains listed in Example 3.1.3 are all
unique factorization domains.

√ √ √ √
Example 3.3.1 Z, Z + Z −1, Z + Z −2, Z + Z 1+ 2 −3 , Z + Z 1+ 2 −7 ,
√ √ √ √
Z + Z 1+ 2−11 , Z + Z 2, Z + Z 3, and Z + Z 6 are unique factorization
domains.

Example 3.3.2 If F is a field then, by Theorem 2.2.1(b), F[X ] is a Euclidean

domain and thus a unique factorization domain.

It is a well-known theorem that if D is a unique factorization domain so is D[X ].

A proof is given in [2, Theorem 7, p. 305]. Appealing to this result we obtain

Example 3.3.3 Z[X ] is a unique factorization domain.

Thus (Z[X ])[Y ] = Z[X, Y ] is a unique factorization domain and generally we

have

Example 3.3.4 Z[X 1 , . . . , X n ] is a unique factorization domain.

Similarly, as F[X ] (F a field) is a unique factorization domain, we have

Example 3.3.5 F[X 1 , . . . , X n ] (F a field ) is a unique factorization domain.

The next example shows that the converse of Theorem 3.3.1 is not true.

Example 3.3.6 The unique factorization domain Z[X ] is not a principal ideal
domain as it contains the nonprincipal ideal 2, X .

Theorem 3.3.2 Let D be a unique factorization domain. Then an element of D is

irreducible if and only if it is prime.

Proof: Let p be an irreducible element of D. Suppose that p | ab, where a and b

are elements of D. Hence there exists an element c of D such that ab = pc. Since
CB609-03 CB609/Alaca & Williams August 7, 2003 17:19 Char Count= 0

3.3 Unique Factorization Domains 63

D is a factorization domain, we have

a = p1 · · · pl , b = q1 · · · qm , c = r1 · · · rn ,
where p1 , . . . , pl , q1 , . . . , qm , r1 , . . . , rn are irreducible elements of D, which are
not necessarily distinct. Then
p1 · · · pl q1 · · · qm = pr1 · · · rn .
As D is a unique factorization domain, p must be an associate of one of the pi or
q j . Hence p | a or p | b, showing that p is a prime.
This completes the proof of the theorem as a prime is always an irreducible by
Theorem 1.2.1.

Let a1 , . . . , an be nonzero elements of a unique factorization domain D. Let

{π1 , . . . , πk } be a set of irreducibles such that

(i) each πi (i = 1, 2, . . . , k) divides at least one of a1 , . . . , an ,

(ii) πi ∼ π j if i = j, and
(iii) if π is an irreducible that divides at least one of a1 , . . . , an then π ∼ πi for some
i ∈ {1, 2, . . . , k}.

We remark that if a1 , . . . , an are all units then {π1 , . . . , πk } = φ. From (i), (ii), and
(iii), we see that

k
ai = i π j ei j , i = 1, 2, . . . , n,
j=1

where i ∈ U (D) and the ei j are nonnegative integers. (ei j is positive if and only if
π j | ai .) Set
e j = min ei j , j = 1, 2, . . . , k,
1≤i≤n

and

k
e
a= π j j ∈ D.
j=1

Clearly
a | ai , i = 1, 2, . . . , n,
and if b ∈ D is such that
b | ai , i = 1, 2, . . . , n,
then
b | a.
CB609-03 CB609/Alaca & Williams August 7, 2003 17:19 Char Count= 0

64 Noetherian Domains

We call a “a greatest common divisor” of a1 , . . . , an , a quantity we had previously

defined in a principal ideal domain (Definition 1.4.2). If D is a unique factorization
domain that is also a principal ideal domain then it is easily checked that this
notion of a greatest common divisor coincides with that of Definition 1.4.2. If the
set {π1 , . . . , πk } of irreducibles is changed to any other set of irreducibles with
properties (i), (ii), and (iii) then a is changed by at most a unit. Thus a greatest
common divisor in a unique factorization domain is only defined up to a unit.
In a principal ideal domain D a greatest common divisor a of a1 , . . . , an is a
linear combination of a1 , . . . , an with coefficients from D. This is not necessarily
the case in a domain D that is a unique factorization domain but not a principal
ideal domain. For example in Z[x] a greatest common divisor of 2x and x 2 is x but
x = f (x)2x + g(x)x 2
for any f (x), g(x) ∈ Z[x].

Theorem 3.3.3 A principal ideal domain is an almost Euclidean domain.

Proof: Let D be a principal ideal domain. By Theorem 3.3.1, D is a unique factor-

ization domain. Hence we may define a function φ : D → N ∪ {0} as follows:
φ(0) = 0,
φ(a) = 1, if a ∈ U (D),
∼
φ(a) = 2n , if a ∈ D− D and a = i 1 i 2 · · · i n ,
where i 1 , . . . , i n are irreducibles.
Clearly φ satisfies (2.4.1), (2.4.2), and (2.4.3). We show that φ satisfies (2.4.4). Let
a, b ∈ D with b = 0. Let I = a, b. Since I is an ideal in D, I = r for some
r ∈ D with r = 0. If a = bq for some q ∈ D then I = bq, b = b. Otherwise
I = b. Since b ∈ I , b = xr for some x ∈ D, so φ(b) ≥ φ(r ). As I = b, x is not
a unit. Thus φ(x) > 1 so φ(r ) < φ(b). Now r = ax 0 + by0 for some x0 , y0 ∈ D
so 0 < φ(ax0 + by0 ) < φ(b) and (2.4.4) is satisfied by φ. Thus φ is an almost
Euclidean function on D and D is an almost Euclidean domain.

From Theorems 2.4.1 and 3.3.3 we deduce Greene’s theorem [3].

Theorem 3.3.4 An integral domain is a principal ideal domain if and only if it is

almost Euclidean.

3.4 Modules
Analogous to the concept of a vector space over a field is that of a module over a
ring. All rings are assumed to possess an identity.
CB609-03 CB609/Alaca & Williams August 7, 2003 17:19 Char Count= 0

3.4 Modules 65

Definition 3.4.1 (R-action) Let R be a ring with identity and M an additive Abelian
group. A function α : R × M → M is called an R-action on M if α has the following
properties:
α(r + s, m) = α(r, m) + α(s, m), (3.4.1)
α(r, m + n) = α(r, m) + α(r, n), (3.4.2)
α(r, α(s, m)) = α(r s, m), (3.4.3)
α(1, m) = m, (3.4.4)
for all r, s ∈ R and all m, n ∈ M.

Definition 3.4.2 (R-module) Let R be a ring with identity. An additive Abelian

group M together with an R-action on M is called an R-module.

It would be more accurate to call what we have just defined a left R-module.
There is a similar definition of a right R-module in which the elements of R are
written on the right. In this book we will keep to left modules throughout.
If M is an R-module with R-action α on M we write α(r, m) as r m to keep the
notation as simple as possible. With this convention (3.4.1)–(3.4.4) become
(r + s)m = r m + sm, (3.4.5)
r (m + n) = r m + r n, (3.4.6)
r (sm) = (r s)m, (3.4.7)
1m = m, (3.4.8)
valid for all r, s ∈ R, m, n ∈ M. Taking n = 0 in (3.4.6) and s = 0 in (3.4.5), we
deduce that r 0 = 0 (r ∈ R) and 0m = 0 (m ∈ M). The reader can easily check
from the axioms that
(−r )m = −(r m) = r (−m)
for all r ∈ R and all m ∈ M.

Example 3.4.1 If F is a field then an F-module is the same thing as a vector space
over F.

Example 3.4.2 Any additive Abelian group A can be thought of as a Z-module in

a natural way. The Z-action on A is just the map (n, a) → na from Z × A to A.

Example 3.4.3 Any ring R with identity can be thought of as a module over itself
in a natural way. We just take M to be the additive group R, + of R and define a
map R × R → R by (r, s) → r s (the product of r and s in R).

A submodule of an R-module M is just a subset N of M such that the operations

of M, when restricted to N , make N into an R-module. These operations are the
CB609-03 CB609/Alaca & Williams August 7, 2003 17:19 Char Count= 0

66 Noetherian Domains

Abelian group operations + and −, and the operation of “multiplying on the left”
by elements of R.

Definition 3.4.3 (Submodule) Let R be a ring with identity. Let M be an R-module.

A subgroup N of M is called a submodule of M if r n ∈ N for all r ∈ R and n ∈ N .

Example 3.4.4 Any R-module M has the submodules M and {0}.

Example 3.4.5 If A is an additive Abelian group considered as a Z-module, then

the submodules of A are precisely the subgroups of A.

Example 3.4.6 Let V be a vector space over a field F considered as an F-module.

Then the submodules of V are its subspaces.

Example 3.4.7 If R is a commutative ring with identity, then the submodules of R

considered as an R-module are the ideals of R.

Definition 3.4.4 (Submodule generated by a set) If X is a subset of an R-module

M then the submodule generated by X is the smallest submodule of M containing
X.

This definition is a valid one because the intersection of all the submodules of
M containing X is such a submodule and is thus the smallest such module. Indeed
if X is a nonempty subset of M then it is not difficult to show that the set
n

ri xi | ri ∈ R, xi ∈ X, n ≥ 1
i=1

of all finite sums of elements of the form r x with r ∈ R and x ∈ X is the smallest
submodule of M containing X , and so it is the submodule of M generated by X .

Definition 3.4.5 (Finitely generated module) An R-module M is called finitely

generated if M is generated by some finite set of elements of M.

Thus an R-module M is finitely generated if and only if there exist finitely

many elements x1 , . . . , xn ∈ M such that each x ∈ M can be expressed as a “linear
n
combination” i=1 ri xi of the xi with coefficients ri ∈ R.

Definition 3.4.6 (Factor module) Let N be a submodule of the R-module M. Then

the factor module M/N is the quotient group M/N of cosets {m + N | m ∈ M}
together with the R-action given by r (m + N ) = r m + N for each r ∈ R and each
coset m + N in M/N .
CB609-03 CB609/Alaca & Williams August 7, 2003 17:19 Char Count= 0

3.5 Noetherian Modules 67

If m + N = m + N then we have m − m ∈ N . Hence r (m − m ) ∈ N since N

is a submodule and thus r m + N = r m + N . This shows that the action of R on
M/N in Definition 3.4.6 is well defined. The axioms are easily verified, so M/N
under this R-action is an R-module. We often write m for m + N .

Definition 3.4.7 (Module homomorphism) Let M and N be R-modules. A module

homomorphism from M to N is a map θ : M → N such that

θ (m 1 + m 2 ) = θ(m 1 ) + θ (m 2 ),
θ (r m) = r θ (m),

for all m, m 1 , m 2 ∈ M and r ∈ R. A module homomorphism that is bijective is

called a module isomorphism. Two modules M and N having a module isomorphism
between them are called module isomorphic, and we write M N .

If K and L are submodules of an R-module M then K + L is a submodule

of M, K is a submodule of K + L, K ∩ L is a submodule of
L, and K + L/K L/K ∩ L since θ : K + L/K → L/K ∩ L defined by θ(k +
l + K ) = l + K ∩ L (k ∈ K , l ∈ L) is a module isomorphism.

3.5 Noetherian Modules

A Noetherian domain is an integral domain in which every ascending chain of ideals
terminates. Analogously we define a Noetherian R-module to be an R-module in
which every ascending chain of submodules terminates.

Definition 3.5.1 (Noetherian module) Let R be a ring with identity. An R-

module M is called Noetherian if every ascending chain of submodules of
M terminates.

Theorem 3.5.1 Let R be a ring with identity. Let M be an R-module and let N
be a submodule of M. Then M is Noetherian if and only if both N and M/N are
Noetherian.

Proof: Suppose that M is Noetherian. Let

N1 ⊆ N2 ⊆ . . .

be an ascending chain of submodules of N . As N is a submodule of M this chain

is also a chain of submodules of M. But M is Noetherian so this chain terminates.
Hence N is also Noetherian. Now let

M1 ⊆ M2 ⊆ . . .
CB609-03 CB609/Alaca & Williams August 7, 2003 17:19 Char Count= 0

68 Noetherian Domains

be an ascending chain of submodules of the factor module M/N . For i = 1, 2, . . .

let
Mi = {m ∈ M | m ∈ Mi }.
It is easy to check that Mi is a submodule of M and that Mi ⊆ Mi+1 . Hence
M1 ⊆ M2 ⊆ . . .
is an ascending chain of submodules of M. As M is Noetherian this chain terminates
and thus the original chain terminates too. This proves that the R-module M/N is
Noetherian.
Now suppose that both N and M/N are Noetherian. Let
M1 ⊆ M2 ⊆ . . .
be an ascending chain of submodules of M. For i = 1, 2, . . . set
M i = {m | m ∈ Mi }.
Again it is easy to check that M i is a submodule of the R-module M/N and that
M i ⊆ M i+1 . Hence
M1 ⊆ M2 ⊆ . . .
is an ascending chain of submodules of M/N . As M/N is Noetherian this chain
terminates and there is a positive integer l1 such that
M i ⊆ M l1 for i ≥ l1 .
Now Mi ∩ N is a submodule of N and Mi ∩ N ⊆ Mi+1 ∩ N so that
M1 ∩ N ⊆ M2 ∩ N ⊆ . . .
is an ascending chain of submodules of N . As N is Noetherian this chain terminates
and there exists a positive integer l2 such that
Mi ∩ N = Ml2 ∩ N for i ≥ l2 .
Set l = max(l1 , l2 ). Thus for i ≥ l we have
M i = M i+1 , Mi ∩ N = Mi+1 ∩ N .
Suppose that the original chain M1 ⊆ M2 ⊆ . . . of submodules of M does not ter-
minate. Then Mi ⊂ Mi+1 for some i ≥ l. We can choose m i+1 ∈ Mi+1 , m i+1 ∈ Mi .
Hence m i+1 ∈ M i+1 = M i and so there exist m i ∈ Mi and n ∈ N such that
m i+1 = m i + n. Thus m i+1 − m i = n ∈ N . Also as Mi ⊆ Mi+1 we have m i+1 −
m i ∈ Mi+1 . Thus m i+1 − m i ∈ Mi+1 ∩ N = Mi ∩ N ⊆ Mi so m i+1 ∈ Mi , which is
a contradiction. Hence the chain M1 ⊆ M2 ⊆ . . . must terminate and M is Noethe-
rian.
CB609-03 CB609/Alaca & Williams August 7, 2003 17:19 Char Count= 0

3.5 Noetherian Modules 69

Theorem 3.5.2 If R is a Noetherian ring, any finitely generated R-module M is

Noetherian.

Proof: Let M be a finitely generated R-module. Then there exist m 1 , m 2 , . . . , m n

∈ M such that

M = Rm 1 + Rm 2 + · · · + Rm n .

Each Rm i is an R-module.
For k = 1, 2, . . . , n we define the R-module Mk by

Mk = Rm 1 + · · · + Rm k

so that Mn = M.
We first show that each R-module Rm i (i = 1, . . . , n) is Noetherian. Let

Ni = {r ∈ R | r m i = 0}.

Clearly Ni is a submodule of R. Since the submodules of R are ideals of R and R is

a Noetherian ring, R is a Noetherian module. Hence, by Theorem 3.5.1, the factor
module R/Ni is Noetherian. But R/Ni Rm i so Rm i is Noetherian. In particular
M1 = Rm 1 is Noetherian.
Now suppose that M1 , . . . , Mk−1 (2 ≤ k ≤ n) are Noetherian. We show that Mk
is Noetherian. As Rm k is Noetherian, we see by Theorem 3.5.1 that the factor
module

Rm k /Rm k ∩ Mk−1

is Noetherian. Hence

Mk /Mk−1 = Mk−1 + Rm k /Mk−1 Rm k /Rm k ∩ Mk−1

is Noetherian. Then, by Theorem 3.5.1, Mk is Noetherian. Hence M1 , . . . , Mn are

Noetherian so M = Mn is a Noetherian module.

The consequence of Theorem 3.5.2 that we use in Chapter 6 is the following

result (see Theorem 6.5.3).

Theorem 3.5.3 Let D and E be integral domains with D ⊆ E. If D is a Noetherian

domain and E is a finitely generated D-module then E is a Noetherian domain.

Proof: Let I1 ⊆ I2 ⊆ . . . be an ascending chain of ideals in the

domain E. By Theorem 3.5.2, as D is a Noetherian domain and E is a
finitely generated D-module, E is a Noetherian D-module. But each Ii is a
D-submodule of E so the chain I1 ⊆ I2 ⊆ . . . must terminate. Hence we have
shown that E is a Noetherian domain.
CB609-03 CB609/Alaca & Williams August 7, 2003 17:19 Char Count= 0

70 Noetherian Domains
√
In Example 3.1.3 we saw that the integral domain Z + Z m is Noetherian for
m = −1, −2, 2, 3, and 6. In fact this is true for an arbitrary integer m that is not a
perfect square.
√
Theorem 3.5.4 Let m be a nonsquare integer. Then Z + Z m is a Noetherian
domain (and thus a factorization domain by Theorem 3.2.2).
√
Proof: We take D = Z and E = Z + Z m in Theorem 3.5.3. As Z is Noetherian
√
(Example 3.1.3) and Z + Z m is a finitely generated Z-module (generated by 1
√
and m) the theorem follows from Theorem 3.5.3.
√
Similarly, taking D = Z and E = Z + Z 1+2 m , where m is a nonsquare inte-
ger with m ≡ 1 (mod 4), in Theorem 3.5.3, we obtain

√ 3.5.5 Let m be a nonsquare integer with m ≡ 1 (mod 4). Then Z +

Theorem
Z 1+2 m is a Noetherian domain (and thus a factorization domain by Theorem
3.2.2).
√
Example 3.5.1 Z + Z −5 is a factorization domain by Theorem 3.5.4. However,
it is not a unique factorization
√ domain as 6 has two different factorizations into
irreducibles in Z + Z −5, namely
√ √
6 = 2 · 3 = (1 + −5)(1 − −5).
√
The fact that 2 and 3 are irreducibles
√ in Z + Z −5 follows√ from Exercise 36 of
Chapter 1. To see that 1 + −5 is an irreducible in Z + Z −5 suppose that
√ √ √
1 + −5 = (a + b −5)(c + d −5)
for some a, b, c, d ∈ Z. Then
6 = (a 2 + 5b2 )(c2 + 5d 2 )
so that (as a 2 + 5b2 is a nonnegative integer)
a 2 + 5b2 = 1, 2, 3, or 6.
√
Clearly a 2 + 5b2 = 2,√3. If a 2 + 5b2 = 1 then a = ±1, b = 0, and a + b√−5 =
±1 is a unit of Z + Z√ −5. If a 2 + 5b2 = 6 then √c + 5d = 1 and c + d √
2 2
−5 =
±1 is a unit of Z + Z −5. This√ proves that 1 + −5 is irreducible in Z + Z −5.
Similarly we can√show that 1 − −5 is also irreducible. The irreducibles 2, 3, 1 +
√
−5, and 1 − −5 are√not associates of one another because the quotient of any
two of them ∈ / U (Z + Z −5) = {−1, +1}.
√
Example 3.5.2 Z + Z 10 is a factorization domain by Theorem 3.5.4. It is not a
unique factorization domain as
CB609-03 CB609/Alaca & Williams August 7, 2003 17:19 Char Count= 0

Exercises 71
√
10 = ( 10)2 = 2 · 5,
√ √
where 2, 5, and 10 are nonassociated
√ irreducibles in Z + Z 10. We just show
that 2 is an irreducible in Z + Z 10. Suppose that
√ √
2 = (a + b 10)(c + d 10)
for some a, b, c, d ∈ Z. Then
4 = (a 2 − 10b2 )(c2 − 10d 2 ).
Hence, as a 2 − 10b2 ∈ Z, we deduce that
a 2 − 10b2 = −4, −2, −1, 1, 2, or 4.
√ √
If a 2 − 10b2 = ±1 then a +√b 10 is a unit of Z + Z√ 10. If a 2 − 10b2 = ±4 then
c2 − 10d 2 = ±1 and c + d 10 is a unit of Z + Z 10. If a 2 − 10b2 = ±2 then
a 2 ≡ ±2 (mod 5), which is a contradiction as a square is congruent to√0, 1, or
√ 2 is irreducible in Z + Z 10.√We
4 (mod 5). Thus this case cannot occur. Hence
√ to show that 5 and 10 are also irreducible in Z + Z 10
leave it to the reader
and that 2, 5, and 10 are not associates of one another (Exercise 13).

In Example 3.5.2 we showed that the equation x 2 − 10y 2 = 2 (or −2) has no
solutions in integers x and y. Here this was very easy to do: We just considered
the equation modulo 5 and got a contradiction. In general one cannot show that
an equation of the type x 2 − my 2 = N has no solutions in integers x and y by
congruence considerations alone. We show how to determine the solvability or in-
solvability of the equation x 2 − my 2 = N (m, n ∈ Z with m positive and nonsquare
√
and 0 < |N | < m) in Section 11.7.

Exercises
1. Let F be a field. If M is an F-module prove that M is a vector space over F. Conversely
show that if M is a vector space over F then M is an F-module.
2. Considering Z as a Z-module, where the Z-action on Z is just multiplication, determine
all the Z-submodules of Z.
3. Let I1 ⊆ I2 ⊆ . . . be an ascending chain of ideals in an integral domain D. Prove that
∞
n=1 In is an ideal in D.
4. Let F be a field. Is the domain F[X ] Noetherian?
5. Prove that the ideal 2, X in Z[X ] is not principal (Example 3.3.6).
6. Prove that a subset N of an R-module M is a submodule of M if and only if
(i) 0 ∈ N ,
(ii) n 1 , n 2 ∈ N =⇒ n 1 − n 2 ∈ N , and
(iii) n ∈ N , r ∈ R =⇒ r n ∈ N .
7. Prove that the intersection of any nonempty collection of submodules of an R-module
is itself a submodule of M.
CB609-03 CB609/Alaca & Williams August 7, 2003 17:19 Char Count= 0

72 Noetherian Domains

8. If M1 , . . . , Mn are n(≥ 1) nonempty subsets of an R-module M, we define

M1 + · · · + Mn = {m 1 + · · · + m n | m i ∈ Mi }.

If M1 , . . . , Mn are submodules of M prove that M1 + · · · + Mn is a submodule of M.

9. Let M and N be R-modules. Let θ : M → N be an R-homomorphism. Define

ker θ = {m ∈ M | θ (m) = 0},

im θ = {n ∈ N | n = θ (m) for some m ∈ M}.

Prove the following:

(i) ker θ is a submodule of M.
(ii) im θ is a submodule of N .
(iii) M/ker θ im θ .
10. If K and L are submodules of an R-module M with K ⊆ L, prove that

M/K / L/K M/L .

11. Suppose that D is a unique factorization domain and a(= 0) and b(= 0) are coprime
nonunits in D. Prove that if ab = cn for some c ∈ D and some n ∈ N then there is a
unit e ∈ D such that ea and e−1 b are nth powers in D.
12. Let D be a unique factorization domain. Give an example to show that the following
assertion is not true in general: If a is an irreducible element of D then a is a maximal
ideal of D. √ √ √
13. Prove that 5 and 10 are irreducible elements of Z + Z 10 and that 2, 5, and 10 are
not associates of one another, as asserted in Example 3.5.2.
14. Let
pbe a prime and m be a positive nonsquare integer such that the Legendre symbol
±p
q
= −1 for some odd prime factor q of m. Prove that the equation x 2 − my 2 = ± p
√
has no solution in integers x and y. Deduce that p is an irreducible element of Z + Z m.
√
15. Prove that Z + Z −6 is not a unique factorization domain by exhibiting an element
√
of Z + Z −6 that has two different factorizations into irreducibles.
√
16. Prove that Z + Z√−10 is not a unique factorization domain.
17. Prove that Z + Z 15 is not a unique factorization domain.

Suggested Reading
1. P. M. Cohn, Unique factorization domains, American Mathematical Monthly 80 (1973),
1–18 (correction, American Mathematical Monthly 80 (1973), 1115).
A brief survey of unique factorization domains is given in the first six sections of the article.
2. D. S. Dummit and R. M. Foote, Abstract Algebra, second edition, Prentice Hall, Upper
Saddle River, New Jersey, 1999.
In Section 9.3 it is shown that D is a unique factorization domain if and only if D[X ] is a unique
factorization domain, and that if D is a unique factorization domain so is D[X 1 , X 2 , . . .].
3. J. Greene, Principal ideal domains are almost Euclidean, American Mathematical
Monthly 104 (1997), 154–156.
This paper is where Theorem 3.3.4 was first proved.
CB609-03 CB609/Alaca & Williams August 7, 2003 17:19 Char Count= 0

Biographies 73

4. B. Hartley and T. O. Hawkes, Rings, Modules and Linear Algebra, Chapman and Hall,
London, New York, 1974.
Chapters 5 and 6 give a very nice introduction to modules.
5. N. Jacobson, Lectures in Abstract Algebra, Volume I, van Nostrand, Princeton,
New Jersey, 1955.
Chapter VI contains a proof of the Hilbert basis theorem.
6. W. Rudin, Unique factorization of Gaussian integers, American Mathematical Monthly
68 (1961), 907–908.
A very simple and short proof is given that the Gaussian domain is a unique factorization domain.
7. P. Samuel, Unique factorization, American Mathematical Monthly 75 (1968), 945–952.
A classic overview of unique factorization is given.
8. O. Zariski and P. Samuel, Commutative Algebra, Volume 1, van Nostrand Company,
Princeton, New Jersey, 1967.
Chapter 4 gives a proof of the Hilbert basis theorem.

Biographies
1. A. Dick, Emmy Noether, 1882–1935, Birkháüser, Boston, Massachusetts, 1981.
This biography of Emny Noether includes obituaries by B. L. van der Waerden, H. Weyl, and
P. S. Alexandrov as well as a list of her publications.
2. C. H. Kimberling, Emmy Noether, American Mathematical Monthly 79 (1972), 136–149
(addendum, American Mathematical Monthly 79 (1972), 755).
This personal biography has excerpts from articles on Noether by P. S. Aleksandrov and H. Weyl.
3. C. Reid, Hilbert, Springer-Verlag, Berlin, Heidelberg, New York, 1970.
This wonderful book covers the life of David Hilbert, a man deeply devoted to the world of logic
and mathematics.
4. The website

http://www-groups.dcs.st-and.ac.uk/˜history/

has biographies of both Hilbert and Noether.

CB609-04 CB609/Alaca & Williams August 7, 2003 16:37 Char Count= 0

4
Elements Integral over a Domain

4.1 Elements Integral over a Domain

Let A be an integral domain and let B be an integral domain containing A. We
are interested in those elements of B that are roots of monic polynomials with
coefficients in A.

Definition 4.1.1 (Element integral over a domain) Let A and B be integral do-
mains with A ⊆ B. The element b ∈ B is said to be integral over A if it satisfies a
polynomial equation
x n + an−1 x n−1 + · · · + a1 x + a0 = 0,
where a0 , a1 , . . . , an−1 ∈ A.

Note that every element a ∈ A is integral over A as it is a root of x − a ∈ A[x].

Definition 4.1.2 (Algebraic integer) A complex number which is integral over Z

is called an algebraic integer.
√
Example 4.1.1 2 is an algebraic integer as it satisfies the equation x 2 − 2 = 0.
√
Example 4.1.2 12 (−1 + i 3) is an algebraic integer as it satisfies the equation
x 2 + x + 1 = 0.
√ √
Example 4.1.3 3 2 − 3 4 is an algebraic integer as it satisfies the equation
x 3 + 6x + 2 = 0.
√ √ √
Example 4.1.4 14 (1 + 21 + 33 − 77) is an algebraic integer as it satisfies
the equation x 4 − x 3 − 16x 2 + 37x − 17 = 0.

Example 4.1.5 A root of unity is an algebraic integer as it is a root of x n − 1 ∈

Z[x] for some n ∈ N.

74
CB609-04 CB609/Alaca & Williams August 7, 2003 16:37 Char Count= 0

4.1 Elements Integral over a Domain 75

√ √
Example 4.1.6 2
∈ D (see Example 1.1.11) is integral over Z + Zi as it is
3 2−i 2

a root of the polynomial x 2 − (4 − 3i) ∈ (Z + Zi)[x].

√
Example
√ 4.1.7 1/ 2 is not an algebraic integer. Suppose on the contrary that
1/ 2 is an algebraic integer. Then there exists a positive integer n and integers
a0 , a1 , . . . , an−1 such that

1 n 1 n−1
√ + an−1 √ + · · · + a0 = 0.
2 2
√
Multiplying both sides of this equation by ( 2)n we obtain
√ √ √
1 + an−1 2 + an−2 ( 2)2 + · · · + a0 ( 2)n = 0.
Thus
√
(1 + 2an−2 + 4an−4 + · · · ) + 2(an−1 + 2an−3 + · · · ) = 0.
If an−1 + 2an−3 + · · · = 0 then
√ −(1 + 2an−2 + 4an−4 + · · · )
2=
(an−1 + 2an−3 + · · · )
is the quotient of two integers and thus a rational number, a contradiction. Hence,
an−1 + 2an−3 + · · · = 0 and so 1 + 2an−2 + 4an−4 + · · · = 0. This is a contradic-
tion as the integer 1 + 2an−2 + 4an−4 + · · · is clearly odd.

Definition 4.1.3 (Element algebraic over a field) Let A and B be integral domains
with A ⊆ B. Suppose that A is a field and b ∈ B is integral over A; then b is said
to be algebraic over A.

Definition 4.1.4 (Algebraic number) A complex number that is algebraic over Q

is called an algebraic number.
√
Example 4.1.8 1/ 2 is an algebraic number as it satisfies the equation x 2 −
1/2 = 0.
√ √
Example
√ 4.1.9 Let A = {a + b 2 | a, b ∈ Q} and B = {x + yi + z 2 +
wi √2 | x, y, z, w ∈ Q} so that A and B are fields with A ⊂ B. Then b = 12 (1+
i + 2) ∈ B is algebraic over A as b satisfies the equation

√ 1
x − (1 + 2)x + 1 + √
2
= 0.
2

Definition 4.1.5 (Domain integral over a subdomain) Let A and B be integral

domains with A ⊆ B. If every b ∈ B is integral over A we say that B is integral
over A.
CB609-04 CB609/Alaca & Williams August 7, 2003 16:37 Char Count= 0

76 Elements Integral over a Domain

√
Example 4.1.10 The quadratic domain Z + Z m, where m is a nonsquare
√ √
integer, is integral over Z as every element u + v m of Z + Z m is integral
over Z as it satisfies the polynomial equation

x 2 − 2ux + u 2 − mv 2 = 0,

where −2u ∈ Z and u 2 − mv 2 ∈ Z.

√
Example 4.1.11 The quadratic domain Z + Z 1+2 m , where m is a nonsquare
integer
√ with
m ≡ 1(mod
√
4), is integral over Z as every element α = u +
1+ m 1+ m
v 2
∈Z+Z 2
is integral over Z because it satisfies the polynomial
equation

1
x − (2u + v)x + u + uv + (1 − m)v = 0,
2 2 2
4

where −(2u + v) ∈ Z and u 2 + uv + 14 (1 − m)v 2 ∈ Z.

Example 4.1.12 Let A and B be integral domains with A ⊆ B. Let a ∈ A. Let

b ∈ B be integral over A. We show that ab is integral over A.
As b ∈ B is integral over A there exist a0 , a1 , . . . , an−1 ∈ A such that

bn + an−1 bn−1 + · · · + a1 b + a0 = 0.

Let a ∈ A. Then ab ∈ B and

(ab)n + an−1 a(ab)n−1 + · · · + a1 a n−1 (ab) + a0 a n = 0.

As an−1 a, . . . , a1 a n−1 , a0 a n ∈ A we deduce that ab is integral over A.

Theorem 4.1.1 Let A ⊆ B ⊆ C be a tower of integral domains. If c ∈ C is integral

over A then c is integral over B.

Proof: As c ∈ C is integral over A there exist a0 , a1 , . . . , an−1 ∈ A such that

cn + an−1 cn−1 + · · · + a1 c + a0 = 0.

As A ⊆ B, a0 , a1 , . . . , an−1 ∈ B and so c is integral over B.

Theorem 4.1.2 Let A ⊆ B ⊆ C be a tower of integral domains. If C is integral

over A then C is integral over B.

Proof: Let c ∈ C. As C is integral over A, c is integral over B. Thus c is integral

over B, so that C is integral over B.
CB609-04 CB609/Alaca & Williams August 7, 2003 16:37 Char Count= 0

4.1 Elements Integral over a Domain 77

Theorem 4.1.3 Let A and B be integral domains with A ⊆ B. Let b ∈ B. Then b

is integral over A if and only if A[b] is a finitely generated A-module.

Proof: Suppose that b is integral over A. Then there exist a0 , a1 , . . . , an−1 ∈ A

such that

bn − an−1 bn−1 − an−2 bn−2 − · · · − a1 b − a0 = 0.

Hence

bn = an−1 bn−1 + an−2 bn−2 + · · · + a1 b + a0 ∈ Abn−1 + Abn−2 + · · · + Ab + A.

Also

bn+1 = an−1 bn + an−2 bn−1 + · · · + a1 b2 + a0 b

∈ Abn + Abn−1 + · · · + Ab2 + Ab
⊆ Abn−1 + · · · + Ab + A.

By induction we see that

bk ∈ Abn−1 + · · · + Ab + A

for all nonnegative integers k. This shows that the integral domain A[b] of polyno-
mials in b with coefficients in A is a finitely generated A-module.
Conversely suppose that A[b] is a finitely generated A-module. Then there exist
u 1 , u 2 , . . . , u n ∈ A[b] such that

A[b] = Au 1 + · · · + Au n .

Clearly u 1 , . . . , u n are not all zero. Now each u i ∈ A[b] and so bu i ∈ A[b] for
i = 1, 2, . . . , n. Thus there exist ai j ∈ A (i, j = 1, 2, . . . , n) such that

 bu 1 = a11 u 1 + · · · + a1n u n ,
···

bu n = an1 u 1 + · · · + ann u n .

Thus the homogeneous system of n equations in the n unknowns x1 , . . . , xn ,



 (b − a11 )x1 − a12 x2 − · · · − a1n xn = 0,

−a21 x1 + (b − a22 )x2 − · · · − a2n xn = 0,

 ···

−an1 x1 − an2 x2 − · · · + (b − ann )xn = 0,

has a nontrivial solution (x1 , x2 , . . . , xn ) = (u 1 , u 2 , . . . , u n ) in the integral domain

A[b] and so in its quotient field. But this can only happen if the determinant of the
CB609-04 CB609/Alaca & Williams August 7, 2003 16:37 Char Count= 0

78 Elements Integral over a Domain

coefficient matrix is zero. Hence

b − a11 −a12 ··· −a1n
−a21 b − a22 ··· −a2n
.. .. .. = 0.
. . .
−an1 −an2 ··· b − ann

When this determinant is expanded, we obtain an equation

bn + an−1 bn−1 + · · · + a1 b + a0 = 0,

where a0 , a1 , . . . , an−1 ∈ A. Hence b is integral over A.

The proof of our next theorem follows closely that of the previous theorem.

Theorem 4.1.4 Let A and B be integral domains with A ⊆ B. Let b ∈ B. If there

exists an integral domain C such that

A[b] ⊆ C ⊆ B

and C is a finitely generated A-module then b is integral over A and A[b] is a

finitely generated A-module.

Proof: As C is a finitely generated A-module, there exist c1 , . . . , cn ∈ C such that

C = Ac1 + · · · + Acn .

Clearly c1 , . . . , cn are not all zero. Now b ∈ A[b] and A[b] ⊆ C so that b ∈ C. But
C is an integral domain so that bc1 , . . . , bcn ∈ C. Hence there exist ai j ∈ A (i, j =
1, 2, . . . , n) such that

 bc1 = a11 c1 + · · · + a1n cn ,
···

bcn = an1 c1 + · · · + ann cn .

Thus the homogeneous system of n equations in the n unknowns x1 , . . . , xn ,

has a nontrivial solution (x1 , . . . , xn ) = (c1 , . . . , cn ) in the integral domain C and

thus in its quotient field. Hence the determinant of its coefficient matrix is zero;
CB609-04 CB609/Alaca & Williams August 7, 2003 16:37 Char Count= 0

4.1 Elements Integral over a Domain 79

that is,
b − a11 −a12 ··· −a1n
−a21 b − a22 ··· −a2n
.. .. .. = 0.
. . .
−an1 −an2 ··· b − ann
Expanding this determinant we obtain an equation
bn + an−1 bn−1 + · · · + a1 b + a0 = 0,
where a0 , a1 , . . . , an−1 ∈ A. Hence b is integral over A and, by Theorem 4.1.3,
A[b] is a finitely generated A-module.

The special case C = B in Theorem 4.1.4 shows that if A and B are integral
domains with A ⊆ B and B is a finitely generated A-module then B is integral over
A.

Theorem 4.1.5 Let A ⊆ B ⊆ C be a tower of integral domains. If B is a finitely

generated A-module and C is a finitely generated B-module then C is a finitely
generated A-module.

Proof: As B is a finitely generated A-module there exist b1 , . . . , bm ∈ B such that

B = Ab1 + · · · + Abm .
As C is a finitely generated B-module there exist c1 , . . . , cn ∈ C such that
C = Bc1 + · · · + Bcn .
Let c ∈ C. Then

n
c= xjcj,
j=1

where x1 , . . . , xn ∈ B. Moreover, for j = 1, . . . , n we have

m
xj = ai j bi ,
i=1

where a11 , . . . , amn ∈ A. Hence

n
m
c= ai j bi c j
j=1 i=1

so that
C = Ab1 c1 + · · · + Abm cn
is a finitely generated A-module.
CB609-04 CB609/Alaca & Williams August 7, 2003 16:37 Char Count= 0

80 Elements Integral over a Domain

Theorem 4.1.6 Let A and B be integral domains with A ⊆ B. Let b1 , b2 ∈ B be

integral over A. Then b1 + b2 , b1 − b2 , and b1 b2 are integral over A.

Proof: As b1 is integral over A, by Theorem 4.1.3, A[b1 ] is a finitely generated A-

module. Moreover, b2 is integral over A and so by Theorem 4.1.1 b2 is integral over
A[b1 ]. Hence, by Theorem 4.1.3, (A[b1 ])[b2 ] = A[b1 , b2 ] is a finitely generated
A[b1 ]-module. Thus A[b1 , b2 ] is a finitely generated A-module by Theorem 4.1.5.
Let λ denote any one of b1 + b2 , b1 − b2 , b1 b2 . Then we have

A ⊆ A[λ] ⊆ A[b1 , b2 ] ⊆ B,

where the integral domain A[b1 , b2 ] is a finitely generated A-module. Hence, by

Theorem 4.1.4, λ is integral over A.

The next theorem is an immediate consequence of Theorem 4.1.6.

Theorem 4.1.7 Let A and B be integral domains with A ⊆ B. Then the set of all
elements of B that are integral over A is a subdomain of B containing A.

Taking A = Z and B = C in Theorem 4.1.7, we obtain

Theorem 4.1.8 The set of all algebraic integers is an integral domain.

The domain of all algebraic integers is denoted by .

Theorem 4.1.9 Let A and B be integral domains with A ⊆ B. Let b1 , . . . , bn ∈ B

be integral over A. Then A[b1 , . . . , bn ] is a finitely generated A-module.

Proof: We prove the theorem by induction on n. If b1 ∈ B is integral over A then

A[b1 ] is a finitely generated A-module by Theorem 4.1.3, so the theorem is true
for n = 1.
Now assume that A[b1 , . . . , bn−1 ] (n ≥ 2) is a finitely generated A-module,
where b1 , . . . , bn−1 ∈ B are integral over A. Let bn ∈ B be integral over A. Then,
by Theorem 4.1.1, bn is integral over A[b1 , . . . , bn−1 ]. Hence, by Theorem 4.1.3,
(A[b1 , . . . , bn−1 ])[bn ] = A[b1 , . . . , bn ] is a finitely generated A-module. This com-
pletes the inductive step and the theorem follows by the principle of mathematical
induction.

Theorem 4.1.10 Let A and B be integral domains with A ⊆ B. If each of

b1 , . . . , bn ∈ B is integral over A then A[b1 , . . . , bn ] is integral over A.

Proof: We prove the theorem by induction on n.

CB609-04 CB609/Alaca & Williams August 7, 2003 16:37 Char Count= 0

4.2 Integral Closure 81

Suppose first that b1 ∈ B is integral over A. Then, by Theorem 4.1.6 and Ex-
ample 4.1.12, we deduce that a0 + a1 b1 + · · · + an b1n is integral over A for all
a0 , a1 , . . . , an ∈ A. This proves that A[b1 ] is integral over A.
Next let b1 , . . . , bn−1 ∈ B be integral over A and suppose that A[b1 , . . . , bn−1 ]
is integral over A. Let bn ∈ B be integral over A. Let f be any element of
A[b1 , . . . , bn ]. Then
f = f 0 + f 1 bn + · · · + f m bnm ,
where f 0 , f 1 , . . . , f m ∈ A[b1 , . . . , bn−1 ]. By the inductive hypothesis
f 0 , f 1 , . . . , f m are all integral over A. Then, as bn is integral over A, we
deduce by Theorem 4.1.6 that f 0 + f 1 bn + · · · + f m bnm is integral over A. Hence
every element f of A[b1 , . . . , bn ] is integral over A, proving that A[b1 , . . . , bn ] is
integral over A.

Theorem 4.1.11 Let A ⊆ B ⊆ C be a tower of integral domains. If B is integral

over A and c ∈ C is integral over B then c is integral over A.

Proof: As c ∈ C is integral over B there exist b0 , b1 , . . . , bn−1 ∈ B such that

cn + bn−1 cn−1 + · · · + b1 c + b0 = 0.
This shows that c is integral over A[b0 , b1 , . . . , bn−1 ]. As each bi ∈ B and B is inte-
gral over A, each bi is integral over A. Thus, by Theorem 4.1.9, A[b0 , b1 , . . . , bn−1 ]
is a finitely generated A-module. As c is integral over A[b0 , b1 , . . . , bn−1 ], by Theo-
rem 4.1.3 we see that (A[b0 , b1 , . . . , bn−1 ])[c] = A[b0 , b1 , . . . , bn−1 , c] is a finitely
generated A-module. Hence, by Theorem 4.1.4, c is integral over A.

We can now prove that “integral over” is a transitive relation.

Theorem 4.1.12 Let A ⊆ B ⊆ C be a tower of integral domains. If C is integral

over B and B is integral over A then C is integral over A.

Proof: Let c be any element of C. Then c is integral over B. As B is integral over

A, by Theorem 4.1.11 c is integral over A. Hence C is integral over A.

4.2 Integral Closure

Let A and B be integral domains with A ⊆ B. In Theorem 4.1.7 we showed that
the set of all elements of B that are integral over A is a subdomain of B containing
A. We now give this domain a name.

Definition 4.2.1 (Integral closure) Let A and B be integral domains with A ⊆ B.

The integral closure of A in B is the subdomain of B consisting of all elements
CB609-04 CB609/Alaca & Williams August 7, 2003 16:37 Char Count= 0

82 Elements Integral over a Domain

of B that are integral over A. The integral closure of A in B is denoted

by A B .

From Theorem 4.1.7 we have

Theorem 4.2.1 Let A and B be integral domains with A ⊆ B. Then the integral
closure A B of A in B is an integral domain satisfying

A ⊆ A B ⊆ B.

Clearly A A = A for any integral domain A.

Our next theorem determines the integral closure of Z in the field Q(i) =
{x + yi | x, y ∈ Q}.

Theorem 4.2.2 The integral closure of A = Z in B = Q(i) is

A B = Z + Zi.

Proof: We first show that Z + Zi ⊆ A B . Let α ∈ Z + Zi. Then α = m + ni, where

m, n ∈ Z. Hence α is a root of the quadratic polynomial

x 2 − 2mx + (m 2 + n 2 ) ∈ Z[x].

This shows that α is an algebraic integer. Clearly α ∈ Q(i). Thus α ∈ A B so that

Z + Zi ⊆ A B .
We now show that A B ⊆ Z + Zi. Let α ∈ A B . Hence α ∈ Q(i) is algebraic over
Z. As α ∈ Q(i) we have α = r + si, where r, s ∈ Q. We just treat the case s = 0.
The case s = 0 can be treated in a similar and easier manner. Clearly α is a root
of g(x) = x 2 − 2r x + (r 2 + s 2 ) ∈ Q[x]. As α is algebraic over Z, there exists a
monic polynomial f (x) ∈ Z[x] with f (α) = 0. Since f (x), g(x) ∈ Q[x], by the
division algorithm there exist polynomials q(x), r (x) ∈ Q[x] such that

f (x) = q(x)g(x) + r (x), deg r (x) < deg g(x).

As deg g(x) = 2 we see that r (x) = r0 + r1 x, where r0 , r1 ∈ Q. Hence

f (x) = q(x)g(x) + r0 + r1 x.

Taking x = α we obtain (as f (α) = g(α) = 0)

r0 + r1 α = 0

so that

(r0 + r1r ) + ir1 s = 0.

CB609-04 CB609/Alaca & Williams August 7, 2003 16:37 Char Count= 0

4.2 Integral Closure 83

Equating real and imaginary parts, we obtain

r0 + r1r = r1 s = 0.

As s = 0 we deduce that r1 = r0 = 0. Hence

f (x) = q(x)g(x),

where q(x), g(x) ∈ Q[x]. Let a be the least common multiple of the denominators
of the coefficients of q(x) and b the least common multiple of the denominators
of g(x). Then ab f (x) = aq(x)bg(x), where aq(x) and bg(x) ∈ Z[x]. Let c be the
content of aq(x) and d the content of bg(x). (Recall that the content of a nonzero
polynomial an x n + · · · + a1 x + a0 ∈ Z[x] is the greatest common divisor of the
integers an , . . . , a1 , a0 and that a primitive polynomial is a polynomial of Z[x] with
content 1.) Then we have aq(x) = cq1 (x) and bg(x) = dg1 (x), where q1 (x) ∈ Z[x]
and g1 (x) ∈ Z[x] are both primitive polynomials. Also ab f (x) = cq1 (x)dg1 (x).
Since f (x) ∈ Z[x] is monic the content of ab f (x) is ab. By a theorem of Gauss,
the product of two primitive polynomials is primitive. Hence q1 (x)g1 (x) is primitive
and the content of cq1 (x)dg1 (x) is cd. Thus ab = cd and f (x) = q1 (x)g1 (x), where
a b
q1 (x) = q(x) ∈ Z[x], g1 (x) = g(x) ∈ Z[x].
c d
Suppose that

f (x) = x n + an−1 x n−1 + · · · + a0 ,

q1 (x) = bn−2 x n−2 + · · · + b0 ,
g1 (x) = c2 x 2 + c1 x + c0 ,

where a0 , . . . , an−1 , b0 , . . . , bn−2 , c0 , c1 , c2 ∈ Z. Equating coefficients of x n in

f (x) = q1 (x)g1 (x), we obtain bn−2 c2 = 1. As bn−2 , c2 ∈ Z we have bn−2 = c2 =
±1. Changing q1 (x) to −q1 (x) and g1 (x) to −g1 (x), if necessary, we may suppose
that c2 = 1. Then g1 (x) and g(x) are both monic so from g1 (x) = (b/d)g(x), we
deduce that b = d and

x 2 − 2r x + (r 2 + s 2 ) = g(x) = g1 (x) ∈ Z[x].

Thus 2r ∈ Z and r 2 + s 2 ∈ Z. If 2r ∈ 2Z + 1 then 2s ∈ 2Z + 1 and 4r 2 + 4s 2 ∈

4Z + 2, contradicting 4r 2 + 4s 2 ∈ 4Z. Hence 2r ∈ 2Z so that r ∈ Z and s ∈ Z,
that is, r + si ∈ Z + Zi, proving that A B ⊆ Z + Zi.

Theorem 4.2.3 Let D be a unique factorization domain. Let F be the field of

quotients of D. Then c ∈ F is integral over D if and only if c ∈ D.

Proof: If c ∈ D then c satisfies the equation x − c = 0 and so is integral over D.

CB609-04 CB609/Alaca & Williams August 7, 2003 16:37 Char Count= 0

84 Elements Integral over a Domain

Conversely, suppose that c ∈ F is integral over D. Then c satisfies a polynomial

equation
x n + an−1 x n−1 + · · · + a1 x + a0 = 0,
where a0 , a1 , . . . , an−1 ∈ D. As c ∈ F we can express c in the form c = r/s, where
r ∈ D, s( = 0) ∈ D, and gcd(r, s) = 1. Hence
r n + an−1r n−1 s + · · · + a1r s n−1 + a0 s n = 0. (4.2.1)
If s is not a unit in D then it is divisible by some prime p. From (4.2.1) we see that
p | r n , and thus, as p is prime, p | r . This contradicts that gcd(r, s) = 1. Hence s
must be a unit and c = r s −1 ∈ D.

Theorem 4.2.4 Q ∩ = Z.

Proof: By Example 3.3.1 Z is a unique factorization domain. Choose D = Z in

Theorem 4.2.3 so that F = Q. Then c ∈ Q ∩ if and only if c ∈ Z. Hence Q ∩ =
Z.

Theorem 4.2.4 tells us that a rational algebraic integer must be an ordinary integer.
We will use this result on a number of occasions.
If D is an integral domain and F its field of quotients then it may happen that
the integral closure D F of D in F is equal to D. If this happens we say that D
is integrally closed. Apparently the term “integrally closed” was first defined by
Ernst Steinitz (1871–1928) in 1912, but the importance of the concept was already
known to Richard Dedekind (1831–1916).

Definition 4.2.2 (Integrally closed domain) An integral domain D is said to be

integrally closed if the only elements of its quotient field that are integral over D
are those of D itself.

Theorem 4.2.5 Let D be a unique factorization domain. Then D is integrally

closed.

Proof: Let F be the field of quotients of D. By Theorem 4.2.3 we have D F = D

so that D is integrally closed.
√
Example
√ 4.2.1
√ Z + Z −3√is not integrally closed. The quotient
√ field of Z +
Z √−3 is Q( −3) = {x √ + y −3 | x, y ∈ Q}. Set α = 12 (1 + −3). Clearly α ∈
Q( −3) but α ∈ Z + Z −3. Moreover, α is integral over √ Z as it satisfies the
equation
√ α 2
− α + 1 = 0. Hence α is integral over Z + Z −3. This shows that
Z√+ Z −3 is not integrally closed. Further, by Theorem 4.2.5, we see that Z +
Z −3 is not a unique factorization domain. For example, 4 has two quite different
CB609-04 CB609/Alaca & Williams August 7, 2003 16:37 Char Count= 0

4.2 Integral Closure 85

√ √
factorizations
√ into irreducibles in Z + Z −3, namely 4 = 2 · 2 = (1 + −3)(1 −
−3).
√
Example 4.2.2 Z + Z 1+ 2 −3 is integrally closed. By Example 3.3.1, Z +
√
Z 1+ 2 −3 is a unique factorization domain. The assertion then follows from The-
orem 4.2.5.

Theorem 4.2.6 Every algebraic number is of the form a/b, where a is an algebraic
integer and b is a nonzero ordinary integer.

Proof: Let c be an algebraic number. Then there exist a0 , a1 , . . . , an−1 ∈ Q such

that
cn + an−1 cn−1 + · · · + a1 c + a0 = 0. (4.2.2)
Let b be the least common multiple of the denominators of a0 , a1 , . . . , an−1 . Thus
b ∈ N and bai ∈ Z for i = 0, 1, 2, . . . , n − 1. From (4.2.2) we obtain
(bc)n + (ban−1 )(bc)n−1 + · · · + (bn−1 a1 )(bc) + (bn a0 ) = 0.
This shows that bc is a root of a monic polynomial in Z. Thus bc is an algebraic
integer, say a. Then c = a/b, where a ∈ and b ∈ Z.

Example 4.2.3 Let

1 1
c= − .
34/3 32/3
Then
3
1 11 1 1 1 1 1
c =
3
−= 4 − 3 8/3 2/3 + 3 4/3 4/3 − 2
34/3 332/3 3 3 3 3 3

1 1 3 1 1 8 1
= − − 2 − = − − c,
34 32 3 34/3 32/3 81 3
so that
1 8
c3 + c + = 0.
3 81
Thus c is an algebraic number. The least common multiple of the denominators of the
coefficients 0/1, 1/3, 8/81 of c2 , c, 1 respectively is b = 81. Then a = bc = 81c
is a root of
a 3 + 2187a + 52488 = 0.
Thus
a = 38/3 − 310/3
CB609-04 CB609/Alaca & Williams August 7, 2003 16:37 Char Count= 0

86 Elements Integral over a Domain

is an algebraic integer. Hence

38/3 − 310/3
c=
81
is the quotient of an algebraic integer and an ordinary integer.

Exercises
1. Prove that
1

1 + 101/3 + 102/3
3
is an algebraic integer.
2. Prove that
102/3 − 1
√
−3
is an algebraic integer.
3. Let m and n be distinct squarefree integers such that m ≡ n ≡ 3 (mod 4). Let l = (m, n)
and set m = lm 1 , n = ln 1 so that (m 1 , n 1 ) = 1. If x0 , x1 , x2 , x3 are integers such that

x0 ≡ x3 (mod 2), x1 ≡ x2 (mod 2),

prove that
1
√ √ √ √
x0 + x1 m + x2 n + x3 m 1 n 1
2
is an algebraic integer.
4. Express the algebraic number
√ 1/3 √ 1/3
1+ 2 1− 2
+
9 9

as the quotient of an algebraic integer and an ordinary integer.

5. Express the algebraic number
√ √
3 1 + 17 3 1 − 17
+
5 5
as the quotient of an algebraic integer and an ordinary integer.
6. Let D be a principal ideal domain. Prove that D is integrally closed.
7. Let m be a nonsquare integer, which is congruent to 1 modulo 4. Prove that the domain
√
Z + Z m is not integrally closed.
8. Let θ be a root of x 3 + 6x + 34. Prove that the domain Z + Zθ + Zθ 2 is not integrally
closed. [Hint: Consider √ φ = (1 + θ )/3.]
√ √
9. Let A = Z and B = Q( 2) = {a + √ b 2 | a,
√ b ∈ Q}. Prove that A B
= Z + Z 2.
10. Prove that the domain Z + Zi + Z 2 + Zi 2 is not integrally closed.
11. If A ⊆ B ⊆ C is a tower of integral domains, prove that A B ⊆ AC ⊆ B C .
CB609-04 CB609/Alaca & Williams August 7, 2003 16:37 Char Count= 0

Biographies 87
√ √ √
12. Prove that Z + Z 2 + Z 5 + Z 10 is not integrally closed in its quotient field. [Hint:
√ 2
Consider 1+√2 5 .]
√ √ √
√ that the integral closure of Z + Z 5 in the field Q( 5, i) = {a + bi + c 5 +
13. Prove
di 5 | a, b, c, d ∈ Q} is
√
1+ 5
α + iβ | α, β ∈ Z + Z .
2
√ √ √
14. Prove
√ that the integral closure of Z + Z 5 in the field Q( 5, ω) = {a + bω + c 5+
dω 5 | a, b, c, d ∈ Q}, where ω is a primitive cube root of unity, is
√
1+ 5
α + βω | α, β ∈ Z + Z .
2
15. Let A and B be integral domains with A ⊆ B and B integral over A. If I is a nonzero
ideal of B, prove that I ∩ A is a nonzero ideal of A.

Suggested Reading
1. T. W. Atterton, A note on certain subsets of algebraic integers, Bulletin of the Australian
Mathematical Society 1 (1969), 345–352.

√ √
√ author√shows for example that the integral closure of Z + Z 5 in Q( 5, i) = {a + bi +
The
c 5 + di 5 | a, b, c, d ∈ Q} is
√
1+ 5
α + iβ | α, β ∈ Z + Z .
2

2. N. Bourbaki, Éléments d’histoire des mathématiques, second edition, Hermann, Paris,

1974.

On page 141 it is mentioned that Steinitz showed how a small number of abstract ideas, such
as an irreducible ideal, chain conditions, and an integrally closed ring, could be used to prove
general results characterizing Dedekind rings and that the last two of these ideas had already been
introduced by Dedekind.

Biographies
1. The website

http://www-groups.dcs.st-and.ac.uk/˜history/

has biographies of both Richard Dedekind (1831–1916) and Ernst Steinitz (1871–1928).
CB609-05 CB609/Alaca & Williams August 7, 2003 16:40 Char Count= 0

5
Algebraic Extensions of a Field

5.1 Minimal Polynomial of an Element Algebraic over a Field

Let K be a subfield of the field C of complex numbers. Let α ∈ C be algebraic
over K (see Definition 4.1.3). As α is algebraic over K , there exists a nonzero
polynomial g(x) ∈ K [x] such that g(α) = 0. We let I K (α) denote the set of all
polynomials in K [x] having α as a root, that is,

I K (α) = { f (x) ∈ K [x] | f (α) = 0}. (5.1.1)

Clearly the set I K (α) contains the zero polynomial. It is easy to check that I K (α) is
an ideal of K [x]. Moreover, I K (α) = 0 as g(x) ∈ I K (α).
As K is a field, by Theorem 2.2.1(b) we know that K [x] is a Euclidean domain
and thus, by Theorem 2.1.2, a principal ideal domain. Hence there exists p(x) ∈
K [x] such that

I K (α) = p(x) . (5.1.2)

Suppose p1 (x) ∈ K [x] is another polynomial that generates I K (α), that is,

I K (α) = p1 (x) .

Then

p(x) = p1 (x)

and so, by Theorem 1.3.1, we have

p1 (x) = u(x) p(x),

where u(x) is a unit in K [x]. However, from Example 1.1.18(c), we have

U (K [x]) = K ∗ ,

so that

u(x) ∈ K ∗ .

88
CB609-05 CB609/Alaca & Williams August 7, 2003 16:40 Char Count= 0

5.1 Minimal Polynomial of an Element Algebraic over a Field 89

This shows that we may take the polynomial p(x) to be monic, in which case p(x)
is uniquely determined by (5.1.2).

Definition 5.1.1 (Minimal polynomial of α over K ) Let K be a subfield of C. Let

α ∈ C be algebraic over K . Then the unique monic polynomial p(x) ∈ K [x] such
that

I K (α) = p(x)

is called the minimal polynomial of α over K and is denoted by irr K (α).

Definition 5.1.2 (Degree of α over K ) Let K be a subfield of C. Let α ∈ C be

algebraic over K . Then the degree of α over K , written deg K (α), is defined by

deg K (α) = deg(irr K (α)).

When K = Q we write deg(α) for degQ (α).

Theorem 5.1.1 Let K be a subfield of C. Let α ∈ C be algebraic over K . Then

irr K (α) is irreducible in K [x].

Proof: Suppose that irr K (α) is reducible in K [x]. Then there exist nonzero poly-
nomials r (x) ∈ K [x] and s(x) ∈ K [x] such that

irr K (α) = r (x)s(x) (5.1.3)

with r (x) ∈
/ U (K [x]) and s(x) ∈
/ U (K [x]). Hence r (x) ∈
/ K and s(x) ∈
/ K so that
deg r (x) ≥ 1 and deg s(x) ≥ 1. Thus

deg(irr K (α)) = deg r (x) + deg s(x) > max(deg r (x), deg s(x)). (5.1.4)

As α is a root of irr K (α), from (5.1.3) we have r (α)s(α) = 0, so that either

r (α) = 0 or s(α) = 0. Without loss of generality we may suppose that r (α) = 0.
Hence

r (x) ∈ I K (α) = irr K (α)

so that

irr K (α) | r (x)

and thus

deg(irr K (α)) ≤ deg r (x),

which contradicts (5.1.4). Hence irr K (α) is irreducible in K [x].

CB609-05 CB609/Alaca & Williams August 7, 2003 16:40 Char Count= 0

90 Algebraic Extensions of a Field

√
Example 5.1.1 α = (1 + i)/ 2 ∈ C is a root of x 4 + 1 ∈ Q[x]. As x 4 + 1 is
monic and irreducible in Q[x], we have

1+i 1+i
irrQ √ = x + 1, deg √
4
= 4.
2 2
√ √
Example √ 5.1.2 Let K be the field Q( 2)
√ = {a + b 2 | a, b ∈ Q}. √
Let α =
(1 + i)/ 2 ∈ C. Then α is a root of x 2 − 2x + 1 ∈ K [x]. As x 2 − 2x + 1
is monic and irreducible in K [x], we have

1+i √ 1+i
√
irrQ( 2) √ = x − 2x + 1, degQ( 2) √
2 √ = 2.
2 2

5.2 Conjugates of α over K

We define the conjugates of an element over a subfield of C.

Definition 5.2.1 (Conjugates of α over K ) Let α ∈ C be algebraic over a subfield

K of C. The conjugates of α over K are the roots in C of irr K (α).

Example 5.2.1 We have from Example 5.1.1 that

1+i
irrQ √ = x 4 + 1.
2
As

1+i 1−i 1+i 1−i
x +1=
4
x− √ x− √ x+ √ x+ √
2 2 2 2
√
the conjugates of (1 + i)/ 2 over Q are
1 + i 1 − i −1 − i −1 + i
√ , √ , √ , √ .
2 2 2 2

Example 5.2.2 We have from Example 5.1.2 that

1+i √
√
irrQ( 2) √ = x 2 − 2x + 1.
2
As

√ 1+i 1−i
x − 2x + 1 =
2
x− √ x− √
2 2
√ √
the conjugates of (1 + i)/ 2 over Q( 2) are
1+i 1−i
√ , √ .
2 2
CB609-05 CB609/Alaca & Williams August 7, 2003 16:40 Char Count= 0

5.3 Conjugates of an Algebraic Integer 91

Example √ 5.2.3 Similarly to Example 5.2.2 we find that the conjugates of

(1 + i)/ 2 over the field Q(i) = {x + yi | x, y ∈ Q} are
1 + i −1 − i
√ , √
2 2
√ √ √
and the conjugates of (1 + i)/ 2 over the field Q( −2) = {x + y −2 | x, y ∈
Q} are
1 + i −1 + i
√ , √ .
2 2

Theorem 5.2.1 Let K be a subfield of C. Let α ∈ C be algebraic over K . Then the

conjugates of α over K are distinct.

Proof: Suppose that α has two conjugates over K that are the same. Then irr K (α)
has a root of order at least 2. Let β ∈ C be such a multiple root. Then

irr K (α) = (x − β)2r (x), (5.2.1)

where r (x) ∈ C[x]. Differentiating (5.2.1) with respect to x, we obtain

irr K (α) = (x − β)2r (x) + 2(x − β)r (x).

Thus β is a root of the derivative irr K (α) of irr K (α). As irr K (α) ∈ K [x] we have

irr K (α) ∈ I K (α) = irr K (α)

so that

irr K (α) | irr K (α)

and thus

deg(irr K (α)) ≤ deg(irr K (α) ),

which is impossible. Hence the conjugates of α over K are distinct.

5.3 Conjugates of an Algebraic Integer

Theorem 5.3.1 If α is an algebraic integer then its conjugates over Q are also
algebraic integers.

Proof: As α is an algebraic integer it is a root of a polynomial

h(x) = x m + am−1 x m−1 + · · · + a1 x + a0 ∈ Z[x].

CB609-05 CB609/Alaca & Williams August 7, 2003 16:40 Char Count= 0

92 Algebraic Extensions of a Field

Since h(x) ∈ Q[x] and h(α) = 0 we have h(x) ∈ IQ (α) = irrQ (α) so that

h(x) = irrQ (α)q(x)

for some q(x) ∈ Q[x]. Let β be a conjugate of α over Q. Then β is also a root of
irrQ (α). Hence h(β) = 0 and so β is also an algebraic integer.

We recall that a monic polynomial f (x) = x n + a1 x n−1 + · · · + an−1 x + an ∈

Z[x] is said to be p-Eisenstein with respect to the prime p if

p | a1 , . . . , p | an−1 , p | an , p 2 an .

Eisenstein’s irreducibility criterion asserts that if f (x) is p-Eisenstein for some

prime p then f (x) is irreducible in Z[x].
√ √
Example 5.3.1 Let α = 3 2 − 3 4. Then
√ √ √ √
α 3 = 2 − 6 2 + 6 4 − 4 = −2 − 6( 2 − 4) = −2 − 6α
3 3 3 3

so that α is a root of the monic cubic polynomial x 3 + 6x + 2 ∈ Z[x] and is thus an

algebraic integer. As x 3 + 6x + 2 is 2-Eisenstein it is irreducible in Z[x]. Hence

irrQ (α) = x 3 + 6x + 2.

The other two roots of irrQ (α) are

√ √ √ √
α = ω 2 − ω2 4, α = ω2 2 − ω 4,
3 3 3 3

where ω is a complex cube root of unity. Thus α and α are also algebraic integers.

Theorem 5.3.2 If α is an algebraic integer then

irrQ (α) ∈ Z[x].

Proof: Let the conjugates of the algebraic integer α over Q be α1 = α, α2 , . . . , αn .

Then

irrQ (α) = (x − α1 )(x − α2 ) · · · (x − αn )

= x n − (α1 + α2 + · · · + αn )x n−1 + (α1 α2 + · · · + αn−1 αn )x n−2
+ · · · + (−1)n α1 α2 · · · αn .

As irrQ (α) ∈ Q[x], we have

α1 + · · · + αn ∈ Q,
α1 α2 + · · · + αn−1 αn ∈ Q,
···
α1 α2 · · · αn ∈ Q.
CB609-05 CB609/Alaca & Williams August 7, 2003 16:40 Char Count= 0

5.3 Conjugates of an Algebraic Integer 93

But, by Theorem 5.3.1, α1 , . . . , αn are all algebraic integers. Hence, by Theorem

4.1.8,

α1 + · · · + αn , α1 α2 + · · · + αn−1 αn , . . . , α1 α2 · · · αn

are all algebraic integers. Since they are all rational, by Theorem 4.2.4 they must
in fact be ordinary integers. Hence irrQ (α) ∈ Z[x].

It is an immediate consequence of Theorem 5.3.2 that if α ∈ C satisfies a poly-

nomial of the form

x m + am−1 x m−1 + · · · + a1 x + a0 ∈ Z[x]

then the monic polynomial of least degree in Q[x] of which α is a root belongs to
Z[x].
We use Theorem 5.3.2 to prove the following result (compare Theorem 4.2.2).
√
Theorem
√ √ integral closure of A = Z + Z −3 in the field B =
5.3.3 The
Q( −3) = {a + b −3 | a, b ∈ Q} is
√
1 + −3
A =Z+Z
B
.
2

√ √
Proof: Let α ∈ Z + Z 1+ 2 −3 . Then α = m + n 1+ 2 −3 for some m, n ∈ Z.
Clearly α ∈ B. As α is a root of the monic polynomial

x 2 − (2m + n)x + (m 2 + mn + n 2 ) ∈ A[x],

√
α is integral over A and thus belongs to A B . Hence Z + Z 1+ 2 −3 ⊆ A B .
√
We now show that A B ⊆ Z + Z 1+ 2 −3 . Let α ∈ A B . Clearly α ∈ B so that
√
α = a + b −3 for some a, b ∈ Q. Thus α is a root of the monic polynomial

x 2 − 2ax + (a 2 + 3b2 ) ∈ Q[x].

The discriminant of this polynomial is

(2a)2 − 4(a 2 + 3b2 ) = −12b2

so that it is reducible in Q[x] if b = 0 and irreducible in Q[x] if b = 0. Hence

x − a, if b = 0,
irrQ (α) =
x − 2ax + (a + 3b ), if b = 0.
2 2 2

As α ∈ A B , α is integral over A and thus is a root of a monic polynomial

x n + α1 x n−1 + · · · + αn ∈ A[x].
CB609-05 CB609/Alaca & Williams August 7, 2003 16:40 Char Count= 0

94 Algebraic Extensions of a Field

√
For i = 1, 2, . . . , n we have αi ∈ A so that αi = ai + bi −3 for some ai , bi ∈ Z.
Thus α is a root of the monic polynomial

(x n + a1 x n−1 + · · · + an )2 + 3(b1 x n−1 + · · · + bn )2 ∈ Z[x].

Hence α is an algebraic integer and so, by Theorem 5.3.2, irrQ (α) ∈ Z[x], that is

a ∈ Z, if b = 0,
2a, a 2 + 3b2 ∈ Z, if b = 0.
√ √ √
In the former case α = a + b −3 = a = a + 0 1+ 2 −3 ∈ Z + Z 1+ 2 −3 . In
the latter case we have a = m/2 for some m ∈ Z. If m ∈ 2Z then a ∈ Z and b ∈ Z.
If m ∈ 2Z + 1 then 2b ∈ 2Z + 1. Hence, in both cases, we see that a = m/2 and
b = m/2 + n, where m, n ∈ Z. Thus
√ √
m m √ 1 + −3 1 + −3
α= + +n −3 = −n + (m + 2n) ∈Z+Z .
2 2 2 2
√
Hence A B ⊆ Z + Z 1+ 2 −3 .
This completes the proof that
√
1 + −3
AB = Z + Z .
2

5.4 Algebraic Integers in a Quadratic Field

In this section we determine the algebraic integers in a field Q(α) obtained by
adjoining a root α(∈ C) of an irreducible quadratic polynomial x 2 + ax + b ∈ Q[x]
to Q; that is, Q(α) is the smallest subfield of C containing both Q and α. We note
that α ∈
/ Q as x 2 + ax + b is irreducible in Q[x]. Clearly

a0 + a1 α + · · · + am α m
Q(α) = | m, n (nonnegative integers),
b0 + b1 α + · · · + bn α n

a0 , . . . , am , b0 , . . . , bn ∈ Q, b0 + b1 α + · · · + bn α = 0 .
n

As α 2 = −b − aα, we obtain recursively that α k = ck + dk α (k = 2, 3, . . .), where

ck , dk ∈ Q. Thus

e0 + e1 α
Q(α) = | e0 , e1 , f 0 , f 1 ∈ Q, ( f 0 , f 1 ) = (0, 0) .
f0 + f1α
As f 02 − a f 0 f 1 + b f 12 = 0 for ( f 0 , f 1 ) = (0, 0) and

e0 + e1 α e0 f 0 − ae0 f 1 + be1 f 1 e1 f 0 − e0 f 1
= + α,
f0 + f1α f 02 − a f 0 f 1 + b f 12 f 02 − a f 0 f 1 + b f 12
CB609-05 CB609/Alaca & Williams August 7, 2003 16:40 Char Count= 0

5.4 Algebraic Integers in a Quadratic Field 95

we deduce that

Q(α) = {x + yα | x, y ∈ Q},

where α 2 + aα + b = 0. The field Q(α) is called a quadratic field or a quadratic

extension of Q. Different quadratic polynomials, for example x 2 + x + 1 and x 2 +
6x + 12, can give rise to the same quadratic field K . Our next theorem gives a
unique way of representing a quadratic field.

Theorem 5.4.1 Let K be a quadratic field. Then there exists a unique squarefree
√
integer m such that K = Q( m).

Proof: Suppose that K = Q(α), where α is a root of the irreducible polynomial

x 2 + ax + b ∈ Q[x]. Then α = α1 or α2 , where
√ √
−a + a 2 − 4b −a − a 2 − 4b
α1 = , α2 = .
2 2
As

α1 + α2 = −a ∈ Q

we have

Q(α1 ) = Q(α2 )

so that
√
−a + a 2 − 4b √
K = Q(α) = Q(α1 ) = Q = Q( c),
2

where c = a 2 − 4b ∈ Q is not the square of a rational number as x 2 + ax + b is

irreducible in Q[x]. Now

c = p/q,

where p, q ∈ Z are such that

q > 0, ( p, q) = 1.

Let r 2 denote the largest square dividing pq. Then pq = r 2 m, where m is a square-
free integer (= 1) and
√
√
p √ √ √
K =Q c =Q = Q( pq) = Q( r 2 m) = Q(r m) = Q( m).
q
√
Now let n be another squarefree integer such that K = Q( n). Hence
√ √
Q( m) = Q( n)
CB609-05 CB609/Alaca & Williams August 7, 2003 16:40 Char Count= 0

96 Algebraic Extensions of a Field

and so
√ √
m=x+y n

for some x, y ∈ Q. Squaring we obtain

√
m = x 2 + ny 2 + 2x y n.

If x y = 0 then
√ m − x 2 − ny 2
n= ,
2x y
√
contradicting that n∈
/ Q as n is squarefree. Hence x y = 0. If y = 0 then
√
m = x,
√ √ √
contradicting that m ∈
/ Q as m is squarefree. Thus x = 0 and m = y n so that

m = y 2 n.

As m is squarefree, we deduce that y 2 = 1 so that m = n. Hence m is uniquely

determined by K .

√
We next determine the algebraic integers in the quadratic field K = Q( m) =
√
{a + b m | a, b ∈ Q}, where m is a squarefree integer. The set of algebraic integers
in K is denoted by O K .

Theorem 5.4.2 Let K be a quadratic field. Let m be the unique squarefree integer
√
such that K = Q( m). Then the set O K of algebraic integers in K is given by
 √
 Z + Z m, √ if m ≡ 1 (mod 4),
OK = 1+ m
Z + Z , if m ≡ 1 (mod 4).
2
√
checked that the elements of Z + Z m if m ≡ 1 √
Proof: It is√easily (mod 4) and of
1+ m
Z+Z 2
if m ≡ 1 (mod 4) are algebraic integers in K = Q( m). Thus
 √
 Z + Z m, √ if m ≡ 1 (mod 4),
OK ⊇ 1+ m
Z + Z , if m ≡ 1 (mod 4).
2
We complete the proof by showing the inclusion in the reverse direction. Let α ∈
√
O K . Then α ∈ K and so α = a + b m for some a, b ∈ Q. Thus α is a root of the
monic polynomial

x 2 − 2ax + (a 2 − mb2 ) ∈ Q[x].

CB609-05 CB609/Alaca & Williams August 7, 2003 16:40 Char Count= 0

5.4 Algebraic Integers in a Quadratic Field 97

The discriminant of this polynomial is

(2a)2 − 4(a 2 − mb2 ) = 4mb2

so that it is reducible in Q[x] if b = 0 and irreducible in Q[x] if b = 0. Hence

x − a, if b = 0,
irrQ (α) =
x − 2ax + (a − mb ), if b = 0.
2 2 2

As α is an algebraic integer, by Theorem 5.3.2 we have irrQ (α) ∈ Z[x] so that

a ∈ Z, if b = 0,
2a, a − mb ∈ Z, if b = 0.
2 2

√
If b = 0 we have α = a ∈ Z ⊂ Z + Z m. Now suppose that b = 0. If 2a ∈ 2Z
then a ∈ Z and so mb2 ∈ Z. Since m is squarefree we see that b ∈ Z. In this case
√ √
α = a + b m ∈ Z + Z m. If 2a ∈ 2Z + 1 then as 4(a 2 − mb2 ) ∈ Z we deduce
that 4mb2 ∈ Z. As m is squarefree we have 2b ∈ Z. If 2b ∈ 2Z then b ∈ Z and so

a 2 = (a 2 − mb2 ) + mb2 ∈ Z,

contradicting that 2a ∈ 2Z + 1. Hence 2b ∈ 2Z + 1. Thus a = (2u + 1)/2 and

b = (2v + 1)/2, where u, v ∈ Z. Then
1
a 2 − mb2 = ((2u + 1)2 − m(2v + 1)2 )
4
so that
m−1
= u 2 + u − m(v 2 + v) − (a 2 − mb2 ) ∈ Z.
4
Hence m ≡ 1 (mod 4) and
√ 2u + 1 2v + 1 √
α =a+b m = + m
2 √2
1+ m
= (u − v) + (2v + 1)
2
√
1+ m
∈Z+Z .
2
This completes the proof of the reverse inclusion and thus the proof of the
theorem.

√
The quadratic field K = Q( m), where m is a squarefree integer, is said to be
real if K ⊆ R and imaginary if K ⊆ R. Clearly K is real if m > 0 and imaginary
if m < 0. We close this section by determining the unit group U (O K ) when K is
an imaginary quadratic field.
CB609-05 CB609/Alaca & Williams August 7, 2003 16:40 Char Count= 0

98 Algebraic Extensions of a Field

Theorem 5.4.3 Let K be an imaginary quadratic field. Then

 √
 {±1, ±i} Z4 , if K = Q(√−1),
U (O K ) = {±1, ±ω, ±ω2 } Z6 , if K = Q( −3),

{±1} Z2 , otherwise,
√
where ω = (−1 + −3)/2.
√ √
Proof: If K = Q( −1) then, by Theorem 5.4.2, we have O K = Z + Z −1 =
Z + Zi, and U (O K ) = {±1, ±i} follows from Exercise 1 of Chapter 1.
√
√
If K = Q( −3) then, by Theorem 5.4.2, we have O K = Z + Z 1+ 2 −3 =
Z + Zω, and U (O K ) = {±1, ±ω, ±ω2 } follows√ from Exercise
√ 2 of Chapter 1.
If K is an imaginary quadratic field = Q( −1), Q( −3) then by Theorem
√
5.4.1 K = Q( m) for a unique negative, squarefree, integer m = −1, −3. If m ≡ 1
√
(mod 4) then O K = Z + Z m and U (O K ) = {±1} √by Exercise
3 of Chapter 1 as
1+ −3
m < −1. If m ≡ 1 (mod 4) then O K = Z + Z 2
by Theorem 5.4.2 and
U (O K ) = {±1} by Exercise 4 of Chapter 1 as m < −3.

5.5 Simple Extensions

Definition 5.5.1 (Simple extension) Let K be a subfield of C and let α ∈ C. Let

K (α) = F,
F
α∈F
K ⊆F ⊆C

where the intersection is taken over all subfields F of C, which contain both K and
α. The intersection is nonempty as C itself is such a field. Since the intersection of
subfields of C is again a subfield of C, K (α) is the smallest field containing both
K and α. We say that K (α) is formed from K by adjoining a single element α. A
subfield L of C for which there exists α ∈ C such that L = K (α) is called a simple
extension of K .

Clearly if α ∈ K then K (α) = K .

For K ⊆ C and α ∈ C let

b0 + b1 α + · · · + bk α k b0 , . . . , bk , c0 , . . . , ch ∈ K ,
L= | k, h ∈ N ∪ {0}, .
c0 + c1 α + · · · + ch α h c0 + c1 α + · · · + cn α h = 0
Then L is a subfield of C that contains both K and α. Moreover any subfield of C
containing both K and α must contain all the elements of L. Hence L is the smallest
subfield of C containing both K and α, so that L = K (α).
CB609-05 CB609/Alaca & Williams August 7, 2003 16:40 Char Count= 0

5.5 Simple Extensions 99

We are interested in simple extensions K (α) of K when α is algebraic over K

and the minimal polynomial of α over K has degree n. By the preceding remarks,
each element β of K (α) is of the form
f (α)
β= ,
g(α)
where

f (x) = b0 + b1 x + · · · + bk x k ∈ K [x],
g(x) = c0 + c1 x + · · · + ch x h ∈ K [x],

and

g(α) = 0.

This implies that irr K (α) g(x) and, since irr K (α) is irreducible in K [x], that

irr K (α), g(x) = K [x].

Thus we can find polynomials m(x), n(x) ∈ K [x] such that

m(x)irr K (α) + n(x)g(x) = 1.

As irr K (α) has the root α, we see that

n(α)g(α) = 1

so that
1
= n(α)
g(α)
and thus
f (α)
β= = f (α)n(α).
g(α)
Hence each element β of K (α) can be expressed as a polynomial in α with coeffi-
cients in K , say

β = d0 + d1 α + · · · + dl αl ,

where l is a nonnegative integer and d0 , d1 , . . . , dl ∈ K . Let

h(x) = d0 + d1 x + · · · + dl x l ∈ K [x],

so that β = h(α). As K is a field we can divide h(x) by irr K (α) to obtain polynomials
u(x) ∈ K [x] and v(x) ∈ K [x] such that

h(x) = u(x)irr K (α) + v(x), deg v(x) < deg(irr K (α)) = n.

CB609-05 CB609/Alaca & Williams August 7, 2003 16:40 Char Count= 0

100 Algebraic Extensions of a Field

Then, as irr K (α) has the root α, we have

h(α) = v(α),
and so
β = v(α).
Hence every element of K (α) is of the form
a0 + a1 α + · · · + an−1 α n−1 ,
where a0 , a1 , . . . , an−1 ∈ K and n = deg(irr K (α)). Thus we have proved the fol-
lowing result.

Theorem 5.5.1 Let K be a subfield of C. Let α ∈ C be algebraic over K . Let

n = deg(irr K (α)). Then
K (α) = {a0 + a1 α + · · · + an−1 α n−1 | a0 , . . . , an−1 ∈ K }.

Theorem 5.5.1 shows that K (α) can be viewed as an n-dimensional vector space
over K with basis {1, α, . . . , α n−1 }. The dimension n is called the degree of the
extension K (α) over K .

Definition 5.5.2 (Degree of the extension K (α) over K ) Let K be a subfield of C.

Let α ∈ C be algebraic over K of degree n (so that n = deg K (α) = deg(irr K (α))).
The degree of the extension K (α) over K , written [K (α) : K ], is defined by
[K (α) : K ] = n.

√
Example 5.5.1 Let m be a squarefree integer. Then m ∈ C is a root of the
√ √ √
polynomial x 2 − m ∈ Q[x]. Now x 2 − m = (x − m)(x + m), where ± m ∈ /
Q as m is squarefree, so that x − m is irreducible in Q[x], and thus
2

√
irrQ ( m) = x 2 − m.
By Theorem 5.5.1 we have
√ √
Q( m) = {a0 + a1 m | a0 , a1 ∈ Q}
and
√
[Q( m) : Q] = 2,
√
so that Q( m) is a quadratic extension of Q.

Example 5.5.2 Let

√ √ 1/3
α = (5 + 17)1/3 + (5 − 17) ∈ R.
CB609-05 CB609/Alaca & Williams August 7, 2003 16:40 Char Count= 0

5.5 Simple Extensions 101

Then
√ √ √
α 3 = (5 + 17) + 3(5 + 17)2/3 (5 − 17)1/3
√ √ √
+ 3(5 + 17)1/3 (5 − 17)2/3 + (5 − 17)
√ √ √ √
= 10 + 3(5 + 17)1/3 (5 − 17)1/3 ((5 + 17)1/3 + (5 − 17)1/3 )
√ √
= 10 + 3((5 + 17)(5 − 17))1/3 α
= 10 + 3 · 81/3 α
= 10 + 6α,

so that α is a root of the monic polynomial x 3 − 6x − 10 ∈ Z[x]. Hence α is

an algebraic integer. Moreover, as x 3 − 6x − 10 is 2-Eisenstein, it is irreducible.
Hence

irrQ (α) = x 3 − 6x − 10.

Thus, by Theorem 5.5.1,

Q(α) = {a0 + a1 α + a2 α 2 | a0 , a1 , a2 ∈ Q}

and

[Q(α) : Q] = deg(irrQ (α)) = 3,

so that Q(α) is a cubic extension of Q.

Example 5.5.3 Let p be a prime number and let ω = e2πi/ p ∈ C. Clearly

ω p = e2πi = 1, so that ω is a root of the monic polynomial x p − 1 ∈ Z[x]. Thus ω
is an algebraic integer. In Z[x] we have

x p − 1 = (x − 1)(x p−1 + x p−2 + · · · + x + 1).

As ω is not a root of x − 1, it must be a root of f p (x) = x p−1 + x p−2 + · · · + x + 1.

We show that f p (x) is irreducible in Z[x]. We have

(x + 1) p − 1 p p p
f p (x + 1) = = x p−1 + x p−2 + x p−3 + · · · + .
(x + 1) − 1 1 2 p−1

As p is a prime the coefficients pi (i = 1, 2, .

. . , p − 1) of f p (x + 1) are all di-
visible by p. Moreover, the constant term p−1 p
= p is not divisible by p 2 . Hence
f p (x + 1) is p-Eisenstein and therefore irreducible in Z[x]. Thus f p (x) is irre-
ducible in Z[x], and thus in Q[x], proving that

irrQ (ω) = x p−1 + x p−2 + · · · + x + 1

and

deg(irrQ (ω)) = p − 1.
CB609-05 CB609/Alaca & Williams August 7, 2003 16:40 Char Count= 0

102 Algebraic Extensions of a Field

Thus Q(ω) is an extension of Q of degree p − 1. A field such as Q(ω), which is

formed by adjoining a root of unity to Q, is called a cyclotomic field.

Definition 5.5.3 (Cyclotomic field) If K is a subfield of C such that K = Q(ω)

for some root of unity ω then K is called a cyclotomic field.

√ √
Example 5.5.4 The√quadratic field Q( −3) = {a + b −3 | a, b ∈ Q} is a cy-
clotomic field as Q( −3) = Q(ω), where ω is a complex cube root of unity.

5.6 Multiple Extensions

We now consider the field obtained by adjoining several elements α1 , . . . , αk ∈
C (k ≥ 2) to a subfield K of C. We denote this field by K (α1 , . . . , αk ). It is the
smallest subfield of C that contains both K and the αi ; that is, it is the intersection
of all the subfields of C containing both K and the αi . The field K (α1 , . . . , αk ) is
called a multiple extension of K . Clearly the order of α1 , . . . , αk does not matter.
The field K (α1 , . . . , αk ) can be regarded as the field obtained by a succession of k
single adjunctions, namely,
K (α1 , α2 ) = K (α1 )(α2 ),
K (α1 , α2 , α3 ) = K (α1 , α2 )(α3 ),
···
K (α1 , α2 , . . . , αk ) = K (α1 , α2 , . . . , αk−1 )(αk ).
When each αi (i = 1, 2, . . . , k) is algebraic over K , it is an important result that the
multiple extension K (α1 , . . . , αk ) is in fact a simple extension K (α) for a suitable
α ∈ C that is algebraic over K . We prove this in Theorem 5.6.2 after treating the
case k = 2 in Theorem 5.6.1.

Theorem 5.6.1 Let K be a subfield of C. Let α ∈ C and β ∈ C be algebraic over

K . Then there exists γ ∈ C that is algebraic over K such that
K (α, β) = K (γ ).

Proof: Let
p(x) = irr K (α), q(x) = irr K (β).
Then
p(x) = (x − α1 ) · · · (x − αm ) ∈ K [x],
where
α1 = α, α2 , . . . , αm
CB609-05 CB609/Alaca & Williams August 7, 2003 16:40 Char Count= 0

5.6 Multiple Extensions 103

are the conjugates of α over K , and

q(x) = (x − β1 ) · · · (x − βn ) ∈ K [x],
where
β1 = β, β2 , . . . , βn
are the conjugates of β over K . By Theorem 5.2.1 we know that the αi are distinct,
as are the β j . The set

αr − αs
S= | r, s = 1, . . . , m; t, u = 1, . . . , n; t = u
βt − βu
consists of a finite number of complex numbers. We choose a rational number c
different from all the members of S. With this choice the mn elements
αi + cβ j (i = 1, . . . , m; j = 1, . . . , n)
are all distinct. Let
γ = α1 + cβ1 = α + cβ
and set
K 1 = K (γ ).
We also let
p1 (x) = p(γ − cx) ∈ K 1 [x].
As
p1 (β) = p(γ − cβ) = p(α) = 0
and
q(β) = 0
we see that β is a common root of p1 (x) and q(x). We show next that these poly-
nomials have no other common roots. Let λ ∈ C be a common root of p1 (x) and
q(x) with λ = β. As λ is a root of q(x) different from β, we have λ = β j for some
j with 2 ≤ j ≤ n. Then, as
p(γ − cβ j ) = p1 (β j ) = 0,
γ − cβ j must be equal to one of α1 , . . . , αm , say αk . Hence
αk + cβ j = γ = α1 + cβ1
so that
α1 − αk
c= ,
β j − β1
CB609-05 CB609/Alaca & Williams August 7, 2003 16:40 Char Count= 0

104 Algebraic Extensions of a Field

contradicting the choice of c. Now let h(x) = irr K 1 (β). Then h(x) | p1 (x) and h(x) |
q(x). Since p1 (x) and q(x) have exactly one common root in C, we must have
deg h(x) = 1. Thus h(x) = x + δ for some δ ∈ K 1 . Now 0 = h(β) = β + δ so that
β = −δ ∈ K 1 . Then α = γ − cβ ∈ K 1 . This shows that
K (α, β) ⊆ K 1 = K (γ ).
Since γ = α + cβ ∈ K (α, β) we have
K (γ ) ⊆ K (α, β)
and thus
K (α, β) = K (γ ).

Theorem 5.6.2 Let K be a subfield of C. Let α1 , α2 , . . . , αn be algebraic over K .

Then there exists α ∈ C algebraic over K such that
K (α1 , α2 , . . . , αn ) = K (α).

Proof: The result is trivial if n = 1, so we may suppose that n ≥ 2. By Theo-

rem 5.6.1 there exists β2 ∈ C algebraic over K such that K (α1 , α2 ) = K (β2 ).
Again by Theorem 5.6.1 there exists β3 ∈ C algebraic over K such that
K (α1 , α2 , α3 ) = K (β2 , α3 ) = K (β3 ). Continuing in this way, we obtain a finite
sequence β2 , β3 , . . . , βn of complex numbers, each algebraic over K , such that
K (α1 , α2 , . . . , αn ) = K (β2 , α3 , . . . , αn )
= K (β3 , α4 , . . . , αn )
= ...
= K (βn−1 , αn )
= K (βn ).
If K is a subfield of C, and α ∈ C and β ∈ C are algebraic over K , the proof
of Theorem 5.6.1 shows how to find γ ∈ C algebraic over K such that K (α, β) =
K (γ ). We have only to find a rational number c such that the elements α + cβ
are all distinct as α ranges over the conjugates of α over K and β ranges over the
conjugates of β over K . Then K (α, β) = K (α + cβ). We illustrate this in the next
two examples.
√ √
√ √ Q( 2, 3) as a simple
Example 5.6.1√We express √ extension. √
The conjugates
√ of
2 over Q are 2 and − 2. The conjugates of 3 over Q are 3 and − 3. The
four numbers
√ √ √ √ √ √ √ √
2 + 3, 2 − 3, − 2 + 3, − 2 − 3
are all distinct, so by Theorem 5.6.1 we have
√ √ √ √
Q( 2, 3) = Q( 2 + 3).
CB609-05 CB609/Alaca & Williams August 7, 2003 16:40 Char Count= 0

5.6 Multiple Extensions 105

Set
√ √
α= 2+ 3 ∈ R.
Squaring α we obtain
√
α 2 = 5 + 2 6,
so that
√
α 2 − 5 = 2 6.
Squaring α 2 − 5 we get
α 4 − 10α 2 + 25 = 24.
Thus α is a root of the monic quartic polynomial
f (x) = x 4 − 10x 2 + 1 ∈ Z[x].
This shows that α is an algebraic integer.
We now show that f (x) is irreducible in Z[x] and thus in Q[x]. Since f (±1) =
−8 = 0, f (x) has no linear factors in Z[x]. Thus if f (x) factors in Z[x], it must
factor as a product of two quadratic polynomials in Z[x], say,
x 4 − 10x 2 + 1 = (x 2 + ax + b)(x 2 + cx + d),
where a, b, c, d ∈ Z. Equating coefficients of x 3 , x 2 , x, and 1, we obtain
a + c = 0,
b + ac + d = −10,
bc + ad = 0,
bd = 1.
From the first equation we have c = −a, so the second equation becomes
b + d + 10 = a 2 .
From the last equation we have b = d = ±1, so that b + d = ±2. Hence a 2 = 8
or 12, which is impossible. This proves that x 4 − 10x 2 + 1 is irreducible in Q[x]
and so
√ √
irrQ ( 2 + 3) = x 4 − 10x 2 + 1
and
√ √ √ √
[Q( 2, 3) : Q] = [Q( 2 + 3) : Q] = 4.
√ √
Example
√ √ 5.6.2 We
√ express Q( 3, 3 2) as
√ a simple
√ extension.
√ The√conjugates of
3 are 3 and − 3. The conjugates of 3 2 are 3 2, ω 3 2, and ω2 3 2, where ω is
CB609-05 CB609/Alaca & Williams August 7, 2003 16:40 Char Count= 0

106 Algebraic Extensions of a Field

a complex cube root of unity. The six numbers

√ √ √ √ √ √
3 + 2, − 3 + 2, 3 + ω 2,
3 3 3

√ √ √ √ √ √
− 3 + ω 2, 3 + ω2 2, − 3 + ω2 2
3 3 3

are all distinct, so by Theorem 5.6.1 we have

√ √ √ √
Q( 3, 2) = Q( 3 + 2).
3 3

We conclude this chapter by proving the very important fact that every element
of a simple extension K (α) of a subfield K of C, where α is algebraic over K , is
algebraic over K .

Theorem 5.6.3 Let K be a subfield of C. Let α ∈ C be algebraic over K . Then

every element β of K (α) is algebraic over K , and the degree of β over K is less
than or equal to the degree of α over K .

Proof: Let β ∈ K (α), where α is algebraic over K . Let n = deg(irr K (α)). By The-
orem 5.5.1 each of the powers β j , j = 0, 1, . . . , n, of β can be written in the
form

n−1
βj = a jk α k ,
k=0

where each a jk ∈ K . The homogeneous system of linear equations

n
a jk x j = 0, k = 0, 1, . . . , n − 1,
j=0

has a solution (x0 , x1 , . . . , xn ) ∈ K n+1 with not all of the x j equal to zero, as the
number of unknowns is greater than the number of equations. Then

n
n
n−1
n−1
n
xjβ j = xj a jk α k = αk a jk x j = 0,
j=0 j=0 k=0 k=0 j=0

proving that β is algebraic over K and that the degree of β over K is less than or
equal to the degree of α over K .

Exercises
1. Prove that the set I K (α) defined in Section 5.1 is an ideal.
2. Prove that x 4 + 1
√is irreducible in Q[x] (see Example
√ 5.1.1).
3. Prove that x 2 − 2x + 1 is irreducible in Q( 2)[x] (see Example 5.1.2).
CB609-05 CB609/Alaca & Williams August 7, 2003 16:40 Char Count= 0

Exercises 107

4. Determine

1+i
irrQ(i) √
2
and

1+i
irrQ(√−2) √ .
2
√ √
5. Prove that [Q(√ 3 + 3 2) :√Q] = 6 (see Example 5.6.2).
6. Prove that Q(√2, i)√= Q( 2√ + i). √
7. Prove that Q( 2, i 2) = Q( 2 + i 2).
8. Find the minimal polynomial of 21/3 + ω over Q(21/3 ), where ω is a complex cube root
of unity.
9. Determine α ∈ C such that
√ √ √
Q( 2, 3, 5) = Q(α).
10. Prove that
√ √ √
[Q( 2, 3, 5) : Q] = 8.
11. Determine the conjugates of 31/3 − 32/3 .
12. Let θ ∈ C be a root of x 3 + 11x + 4 = 0. Prove that [Q(θ ) : Q] = 3.
13. Prove that (−θ + θ 2 )/2 is an algebraic integer in K = Q(θ ), where θ 3 + 11θ − 4 = 0.
√
14. Let θ ∈ C be a root of x 5 + x + 1 = 0. If θ ∈ Q( −3), what is irrQ θ ?
15. Let ω = e2πi/5 . Prove that

1 √ √
ω= 5 − 1 + i 10 + 2 5 .
4
√ √
16. Let ω = e2πi/5 . Show that 5 ∈ Q(ω) by expressing 5 in the form
√
5 = aω + bω2 + cω3 + dω4
for suitable integers a, b, c, d.
17. Prove that

1 √ √
i 10 + 2 5 + 2i 10 − 2 5
2
is an algebraic integer in Q(e2πi/5 ).
18. Determine the conjugates of
121/5 + 541/5 − 1441/5 + 6481/5
over Q.
√
19. Let m be a squarefree integer ≡ 1 (mod 4). Let A =Z+Z m and
√
B = Q( m). Prove that
√
1+ m
A =Z+Z
B
.
2
20. Let θ be a nonreal algebraic number. Prove that the complex conjugate θ̄ of θ is one of
the conjugates of θ over Q.
CB609-05 CB609/Alaca & Williams August 7, 2003 16:40 Char Count= 0

108 Algebraic Extensions of a Field

a 2 −4c
21. Let p be an odd prime. Let a and c be integers with a ≡ 1 (mod 2) and p
= −1.
Prove that x + ax + px + c is irreducible in Z[x].
4 2

22. Use Exercise 21 to prove that [Q(θ) : Q] = 4, where θ is a root of x 4 + 7x 2 + 5x +

4 = 0.

Suggested Reading
1. E. R. Scheinerman, When close enough is close enough, American Mathematical
Monthly 107 (2000), 489–499.
A technique is presented for proving identities involving algebraic integers numerically. For ex-
ample Shanks’s identity [2]

√ √ √ √ √
5 + 22 + 2 5 = 11 + 2 29 + 16 − 2 29 + 2 55 − 10 29
can be proved using this technique.
2. D. Shanks, Incredible identities, Fibonacci Quarterly 12 (1974), 271, 280.

Biographies
1. H. C. Williams, Daniel Shanks (1917–1996), Notices of the American Mathematical
Society 44 (1997), 813–816.
2. H. C. Williams, Daniel Shanks (1917–1996), Mathematics of Computation 66 (1997),
929–934.
CB609-06 CB609/Alaca & Williams August 27, 2003 16:53 Char Count= 0

6
Algebraic Number Fields

6.1 Algebraic Number Fields

An algebraic number field is a field K that is obtained from the field of rational
numbers Q by adjoining a finite number of algebraic numbers.

Definition 6.1.1 (Algebraic number field) An algebraic number field is a subfield

of C of the form Q(α1 , . . . , αn ), where α1 , . . . , αn are algebraic numbers.
√ √ √ √
Example 6.1.1 Q( 2, 3, 7); Q( 7
1 + i, θ), where θ is a root of the polyno-
√ √ √ √
mial x − x + 1; and Q( 1 + 2 + 1 − 2, 53 + 3 5) are all examples of
5 3 3

algebraic number fields.

By Theorem 5.6.2 an algebraic number field can be obtained by adjoining a

single algebraic number θ to Q.

Theorem 6.1.1 If K is an algebraic number field then there exists an algebraic

number θ such that K = Q(θ ).

Proof: This is the special case K = Q of Theorem 5.6.2.

In fact the algebraic number θ in Theorem 6.1.1 can always be taken to be an

algebraic integer.

Theorem 6.1.2 If K is an algebraic number field then there is an algebraic integer

θ such that K = Q(θ ).

Proof: Let K be an algebraic number field. By Theorem 6.1.1 there is an algebraic

number φ such that K = Q(φ). By Theorem 4.2.6 we have φ = θ/b, where θ is an
algebraic integer and b is a nonzero rational integer. Thus

K = Q(φ) = Q(θ/b) = Q(θ ).

109
CB609-06 CB609/Alaca & Williams August 27, 2003 16:53 Char Count= 0

110 Algebraic Number Fields

√ √
Example 6.1.2
√ We√ show that the algebraic number field Q( 2, 3) is Q(θ),
where θ = √ 2 +√ 3. This was done in a different way in Example 5.6.1. Clearly
Q(θ) ⊆ Q( 2, 3). As
√ 9 1
2 = − θ + θ 3 ∈ Q(θ)
2 2
and
√ 11 1
3 = θ − θ 3 ∈ Q(θ),
2 2
√ √ √ √ √ √ √ √
we see that Q( 2, 3) ⊆ Q( 2 + 3). Hence Q( 2, 3) = Q( 2 + 3).

The form of the elements in an algebraic number field follows immediately from
Theorem 5.5.1. We have

Theorem 6.1.3 Let K = Q(θ ) be an algebraic number field, where θ is an algebraic

number. Let the degree of the polynomial irrQ (θ) be n. Then every element of K is
expressible uniquely in the form
c0 + c1 θ + · · · + cn−1 θ n−1 ,
where c0 , c1 , . . . , cn−1 ∈ Q, and every such quantity c0 + c1 θ + · · · + cn−1 θ n−1
(c0 , c1 , . . . , cn−1 ∈ Q) belongs to K .

Clearly K is an n-dimensional vector space over Q and the degree of K over Q

is n. K is called a quadratic field if n = 2, a cubic field if n = 3, a quartic field if
n = 4, and a quintic field if n = 5.

Definition 6.1.2 (The set O K ) The set of all algebraic integers that lie in the
algebraic number field K is denoted by O K ; that is,
OK = ∩ K .

Theorem 6.1.4 Let K be an algebraic number field. Then O K is an integral domain.

Proof: By Theorem 4.1.8 we know that is an integral domain (⊆ C). Hence, as

K (⊆ C) is a field, O K = ∩ K is an integral domain.

Definition 6.1.3 (Ring of integers of an algebraic number field) O K is called

the ring of integers of the algebraic number field K .

Example 6.1.3 In Theorem 5.4.2 we determined the ring of integers O K of a

quadratic field K . Taking m = −1 and m = −3 in Theorem 5.4.2, we see that
√
OQ(√−1) = Z + Z −1
CB609-06 CB609/Alaca & Williams August 27, 2003 16:53 Char Count= 0

6.1 Algebraic Number Fields 111

and
√
1+ −3
OQ(√−3) = Z + Z .
2

√
Thus the Gaussian domain √ Z + Z −1 (Example 1.1.2) is the ring of inte-
gers √Q( −1) and the Eisenstein domain Z + Zω = Z +
of √the quadratic field
−1+ −3
Z 2
= Z + Z 1+ 2 −3 (Example 1.1.3) is the ring of integers of the
√
quadratic field Q( −3).
When K is not a quadratic field, it is a more difficult problem to determine O K .
We determine O K for some algebraic number fields of degree > 2 in Chapter 7.
Indeed it is an area of current research to determine O K explicitly for certain
classes of algebraic number fields K . See the references at the end of Chapter 7 in
this connection.

Theorem 6.1.5 If K is an algebraic number field then the quotient field of O K is

Proof: Let F denote the quotient field of O K , and let α ∈ F. Then α = b/c, where
b ∈ O K and c ∈ O K with c
= 0. As O K ⊆ K we have b ∈ K and c ∈ K so that,
as K is a field, α = b/c ∈ K . Hence F ⊆ K .
Now let α ∈ K . By Theorem 4.2.6 we have α = b/c, where b is an algebraic
integer and c is a nonzero rational integer. Clearly b = αc ∈ K so b ∈ O K . Thus
α = b/c ∈ F. Hence K ⊆ F.
This proves that F = K , so the quotient field of O K is K .

Theorem 6.1.6 If K is an algebraic number field then O K is integrally closed.

Proof: By Theorem 6.1.5, the quotient field of O K is K . Let β ∈ K be integral

over O K . As O K is integral over Z, by Theorem 4.1.11 β is integral over Z, that
is, β is an algebraic integer in K . Hence β ∈ O K . This proves that O K is integrally
closed.

For an algebraic number field K , we showed in Theorem 6.1.2 that there is an

algebraic integer θ such that K = Q(θ). As θ ∈ K and θ ∈ we see that θ ∈ ∩
K = O K . We now wish to show that θ can be taken from any given nonzero ideal of
O K (Theorem 6.1.8). To prove this we make use of the next result (Theorem 6.1.7),
which asserts that if I is a nonzero ideal of O K then I ∩ Z always contains a nonzero
integer.

Theorem 6.1.7 Let K be an algebraic number field. Then every nonzero ideal in
O K contains a nonzero rational integer.
CB609-06 CB609/Alaca & Williams August 27, 2003 16:53 Char Count= 0

112 Algebraic Number Fields

Proof: Let I
= {0} be an ideal in O K . Choose α ∈ I with α
= 0. As α ∈ I ⊆ O K , α
is an algebraic integer. Let irrQ (α) = x n + b1 x n−1 + · · · + bn . We show that bn
= 0.
If n = 1 then irrQ (α) = x + b1 so that α + b1 = 0. Hence b1 = −α
= 0. If n ≥ 2
then bn
= 0 as irrQ (α) is irreducible in Q[x] by Theorem 5.1.1. By Theorem 5.3.2
we know that irrQ (α) ∈ Z[x] as α is an algebraic integer. Hence b1 , . . . , bn ∈ Z.
Thus bn = −α n − b1 α n−1 − · · · − b1 α ∈ I . Hence bn is a nonzero rational integer
in I .

Theorem 6.1.8 Let K be an algebraic number field. Let I be a nonzero ideal of

O K . Then there exists γ ∈ I such that K = Q(γ ).

Proof: By Theorem 6.1.2 there exists θ ∈ O K such that K = Q(θ ). By Theorem

6.1.7 there exists c ∈ Z ∩ I with c
= 0. Set γ = cθ . As θ ∈ O K and c ∈ I we
have γ ∈ I . Moreover, as c ∈ Z \ {0}, we have K = Q(θ) = Q(cθ) = Q(γ ), where
γ ∈ I.

6.2 Conjugate Fields of an Algebraic Number Field

Let K be an algebraic number field. In this section we begin by √
determining the
number of monomorphisms σ : K → C. For example, if K = Q( 2) then
√ √
σ1 (x + y 2) = x + y 2 (x, y ∈ Q)
and
√ √
σ2 (x + y 2) = x − y 2 (x, y ∈ Q)
are two monomorphisms from K to C.

Theorem 6.2.1 Let K be an algebraic number field of degree n over Q. Then there
are exactly n distinct monomorphisms σk : K → C (k = 1, . . . , n).

Proof: By Theorem 6.1.1 there exists an algebraic number θ ∈ K such that K =

Q(θ ). Let p(x) = irrQ (θ ). Then
deg p(x) = deg (irrQ (θ)) = [Q(θ) : Q] = n,
so that θ has n distinct conjugates over Q (Theorem 5.2.1), say θ1 = θ, θ2 , . . . , θn ,
and
p(x) = (x − θ1 )(x − θ2 ) · · · (x − θn ).
By Theorem 6.1.3 each element α of K can be expressed uniquely in the form α =
a0 + a1 θ + · · · + an−1 θ n−1 , where a0 , a1 , . . . , an−1 ∈ Q, so, for k = 1, 2, . . . , n,
we can define
σk : K → C
CB609-06 CB609/Alaca & Williams August 27, 2003 16:53 Char Count= 0

6.2 Conjugate Fields of an Algebraic Number Field 113

by
σk (a0 + a1 θ + · · · + an−1 θ n−1 ) = a0 + a1 θk + · · · + an−1 θkn−1 .
We show that σk (k = 1, 2, . . . , n) is a field homomorphism.
First we show that σk (k = 1, 2, . . . , n) is additive. Let α, β ∈ K . Then
α = a0 + a1 θ + · · · + an−1 θ n−1
and
β = b0 + b1 θ + · · · + bn−1 θ n−1 ,
where a0 , a1 , . . . , an−1 , b0 , b1 , . . . , bn−1 ∈ Q. Hence
α + β = (a0 + b0 ) + (a1 + b1 )θ + · · · + (an−1 + bn−1 )θ n−1
and so
σk (α + β) = (a0 + b0 ) + (a1 + b1 )θk + · · · + (an−1 + bn−1 )θkn−1
= (a0 + a1 θk + · · · + an−1 θkn−1 ) + (b0 + b1 θk + · · · + bn−1 θkn−1 )
= σk (α) + σk (β).
Thus σk is additive.
Next we show that σk (k = 1, 2, . . . , n) is multiplicative. With the same notation,
we let
f (x) = a0 + a1 x + · · · + an−1 x n−1 ∈ Q[x]
and
g(x) = b0 + b1 x + · · · + bn−1 x n−1 ∈ Q[x]
so that
f (θ) = α, g(θ) = β.
Dividing f (x)g(x) by p(x) in Q[x], we obtain a quotient q(x) ∈ Q[x] and a re-
mainder r (x) ∈ Q[x] such that
f (x)g(x) = p(x)q(x) + r (x), deg r (x) < deg p(x) = n.
Hence, as p(θ) = 0, we have
αβ = f (θ )g(θ ) = p(θ )q(θ) + r (θ) = r (θ).
Thus, as p(θk ) = 0, we have
σk (αβ) = σk (r (θ)) = r (θk ) = p(θk )q(θk ) + r (θk ) = f (θk )g(θk ) = σk (α)σk (β),
so that σk is multiplicative.
Hence we have shown that σk (k = 1, 2, . . . , n) is a homomorphism.
CB609-06 CB609/Alaca & Williams August 27, 2003 16:53 Char Count= 0

114 Algebraic Number Fields

We now show that σk (k = 1, 2, . . . , n) is injective so that it is a monomor-

phism. Suppose α = a0 + a1 θ + · · · + an−1 θ n−1 ∈ K and β = b0 + b1 θ + · · · +
bn−1 θ n−1 ∈ K are such that σk (α) = σk (β). Then we have
a0 + a1 θk + · · · + an−1 θkn−1 = b0 + b1 θk + · · · + bn−1 θkn−1
so that θk is a root of the polynomial
(a0 − b0 ) + (a1 − b1 )x + · · · + (an−1 − bn−1 )x n−1 ∈ Q[x]
of degree < n. As the deg(irrQ θk ) = deg p(x) = n, this polynomial must be the
zero polynomial so that
a0 − b0 = a1 − b1 = · · · = an−1 − bn−1 = 0;
that is,
a0 = b0 , a1 = b1 , . . . , an−1 = bn−1 ,
and so
α = a0 + a1 θ + · · · + an−1 θ n−1 = b0 + b1 θ + · · · + bn−1 θ n−1 = β,
proving that σk is injective.
Finally, let λ : K → C be a monomorphism. Then
p(λ(θ)) = λ( p(θ )) = λ(0) = 0
so that
λ(θ ) = θk
for some k ∈ {1, 2, . . . , n}. Thus
λ(θ) = σk (θ)
and so
λ(a0 + a1 θ + · · · + an−1 θ n−1 ) = a0 + a1 θk + · · · + an−1 θkn−1
= σk (a0 + a1 θ + · · · + an−1 θ n−1 )
for all a0 , a1 , . . . , an−1 ∈ Q, proving that
λ = σk .
Hence {σk | k = 1, 2, . . . , n} comprise all the monomorphisms from K to C.

For k = 1, 2, . . . , n, we have
range σk = σk (K )
= {σk (a0 + a1 θ + · · · + an−1 θ n−1 ) | a0 , a1 , . . . , an−1 ∈ Q}
= {a0 + a1 θk + · · · + an−1 θkn−1 | a0 , a1 , . . . , an−1 ∈ Q}
= Q(θk )
CB609-06 CB609/Alaca & Williams August 27, 2003 16:53 Char Count= 0

6.2 Conjugate Fields of an Algebraic Number Field 115

so that
σk : Q(θ) → Q(θk )
is an isomorphism. Hence all the fields Q(θk ) (k = 1, 2, . . . , n) are isomorphic.

Definition 6.2.1 (Conjugate fields of an algebraic number field) Let K be an

algebraic number field. Let θ be an algebraic number such that K = Q(θ). Let
θ1 = θ, θ2 , . . . , θn
be the conjugates of θ over Q. Then the fields
Q(θ1 ) = Q(θ) = K , Q(θ2 ), . . . , Q(θn )
are called the conjugate fields of K .

By the remarks preceding Definition 6.2.1 each of the conjugate fields of K is

isomorphic to K .
It appears from the definition that the conjugate fields of K may depend upon
the choice of algebraic number θ such that K = Q(θ). We show that this is in fact
not the case.

Theorem 6.2.2 Let K be an algebraic number field. Let θ be an algebraic number

such that K = Q(θ ). Let θ1 = θ, θ2 , . . . , θn be the conjugates of θ. Let φ be another
algebraic number such that K = Q(φ). Let c0 , c1 , . . . , cn−1 ∈ Q be such that
φ = c0 + c1 θ + · · · + cn−1 θ n−1 .
For k = 1, 2, . . . , n set
φk = c0 + c1 θk + · · · + cn−1 θkn−1
so that φ1 = φ. Then φ1 , φ2 , . . . , φn are the conjugates of φ over Q, and
Q(θk ) = Q(φk ), k = 1, 2, . . . , n.

Proof: Let

n
n
f (x) = (x − φk ) = (x − (c0 + c1 θk + · · · + cn−1 θkn−1 )) ∈ K 1 [x],
k=1 k=1

where K 1 is the algebraic number field given by

K 1 = Q(θ1 , θ2 , . . . , θn ).
Clearly Q ⊆ K ⊆ K 1 ⊆ C. The coefficients of f (x) are (up to sign) the elementary
symmetric polynomials in c0 + c1 θk + · · · + cn−1 θkn−1 (k = 1, 2, . . . , n) and so are
polynomials with rational coefficients in the elementary symmetric polynomials in
θ1 , θ2 , . . . , θn . Since θ1 + θ2 + · · · + θn , θ1 θ2 + · · · + θn−1 θn , . . . , θ1 θ2 · · · θn are
CB609-06 CB609/Alaca & Williams August 27, 2003 16:53 Char Count= 0

116 Algebraic Number Fields

(up to sign) the coefficients of irrQ (θ) ∈ Q[x], they are all rational numbers, and so
the coefficients of f (x) are all rational. Hence f (x) ∈ Q[x]. As f (φ) = 0 we have
irrQ (φ) | f (x), say f (x) = irrQ (φ)g(x), where g(x) ∈ Q[x]. Then

n = deg f (x) = deg(irrQ (φ)g(x)) = deg(irrQ (φ)) + deg g(x).

Now deg(irrQ (φ)) = [Q(φ) : Q] = [K : Q] = n, so that deg g(x) = 0; that is,

g(x) ∈ Q, say g(x) = c. Since f (x) = c irrQ (φ) and both f (x) and irrQ (φ) are
monic polynomials of degree n, we have c = 1. Thus f (x) = irrQ (φ). Hence
φ1 , φ2 , . . . , φn are the conjugates of φ over Q.
Finally, for k = 1, 2, . . . , n, we have

Q(φk ) = Q(c0 + c1 θk + · · · + cn−1 θkn−1 ) ⊆ Q(θk )

and

[Q(φk ) : Q] = [Q(θk ) : Q] (= n)

so that

Q(φk ) = Q(θk ), k = 1, 2, . . . , n.

6.3 The Field Polynomial of an Element of an Algebraic Number Field

Let K be an algebraic number field of degree n over Q. Let θ ∈ K be such that
K = Q(θ ). Let θ1 = θ, θ2 , . . . , θn be the conjugates of θ over Q.
For α ∈ K there exist unique rational numbers c0 , c1 , . . . , cn−1 such that

α = c0 + c1 θ + · · · + cn−1 θ n−1

(see Theorem 6.1.3). For k = 1, 2, . . . , n we set

αk = c0 + c1 θk + · · · + cn−1 θkn−1 ∈ Q(θk ).

Definition 6.3.1 (Complete set of conjugates of α relative to K ) The set of al-

gebraic numbers {α1 = α, α2 , . . . , αn } is called a complete set of conjugates of α
relative to K . More briefly they are called the “K -conjugates of α” or the “conju-
gates of α relative to K .”
√ √
Example 6.3.1 Let K = Q(θ), where θ = 2+ 3. From Example 5.6.1 we see
that

irrQ (θ) = x 4 − 10x 2 + 1.

As
√ √ √ √ √ √ √ √
x 4 − 10x 2 + 1 = (x − 2 − 3)(x − 2 + 3)(x + 2 − 3)(x + 2 + 3)
CB609-06 CB609/Alaca & Williams August 27, 2003 16:53 Char Count= 0

6.3 Field Polynomial 117

the conjugates of θ are

√ √ √ √ √ √ √ √
θ1 = 2 + 3, θ2 = 2 − 3, θ3 = − 2 + 3, θ4 = − 2 − 3.
√ √ √ √ √ √
Let α = 2 3 so that α ∈ Q( 3) ⊂ Q( 2, 3) = Q( 2 + 3) = Q(θ) = K
(Example 5.6.1). Hence α = a + bθ + cθ 2 + dθ 3 for some a, b, c, d ∈ Q. Thus
√ √ √ √ √ √ √
2 3 = a + b( 2 + 3) + c( 2 + 3)2 + d( 2 + 3)3
√ √ √
= (a + 5c) + (b + 11d) 2 + (b + 9d) 3 + 2c 6.

Hence

a + 5c = 0, b + 11d = 0, b + 9d = 2, 2c = 0,

so that

a = 0, b = 11, c = 0, d = −1,

giving

α = 11θ − θ 3 .

The K -conjugates of α are

√
α1 = α = 11θ − θ 3 = 2 3,
√
α2 = 11θ2 − θ23 = −2 3,
√
α3 = 11θ3 − θ33 = 2 3,
√
α4 = 11θ4 − θ43 = −2 3.

Thus the complete set of conjugates of α relative to K is α, −α, α, −α.

The conjugates of α relative to K are obtained from α by applying the monomor-

phisms σk : K → C (k = 1, 2, . . . , n) to α. Clearly σk (α) = αk (k = 1, 2, . . . , n)
and αk ∈ Q(θk ) (k = 1, 2, . . . , n). It can be shown that the conjugates of α relative
to K do not depend on the choice of θ such that K = Q(θ) (Exercise 1 of this
Chapter).

Definition 6.3.2 (Field polynomial of α over K ) Let K be an algebraic number

field of degree n. Let α ∈ K . Let α1 = α, α2 , . . . , αn be the K -conjugates of α. Then
the field polynomial of α over K is the polynomial

n
fld K (α) = (x − αk ).
k=1

Clearly fld K (α) ∈ C[x]. However, much more is true as the next theorem shows.
CB609-06 CB609/Alaca & Williams August 27, 2003 16:53 Char Count= 0

118 Algebraic Number Fields

Theorem 6.3.1 Let K be an algebraic number field of degree n. Let α ∈ K . Then

fld K (α) ∈ Q[x].

Proof: Let θ ∈ K be such that K = Q(θ). We have deg(irrQ (θ)) = [Q(θ ) : Q] =

[K : Q] = n. As α ∈ K , by Theorem 6.1.3 there exist c0 , c1 , . . . , cn−1 ∈ Q such
that

α = c0 + c1 θ + · · · + cn−1 θ n−1 .

The K -conjugates of α are α1 = α, α2 , . . . , αn , where

αk = c0 + c1 θk + · · · + cn−1 θkn−1 , k = 1, 2, . . . , n.

The field polynomial of α over K is

n
n
fld K (α) = (x − αk ) = (x − (c0 + c1 θk + · · · + cn−1 θkn−1 )).
k=1 k=1

Clearly fld K (α) ∈ K 1 [x], where K 1 = Q(θ1 , . . . , θn ). Arguing as in the proof of

Theorem 6.2.2, we deduce that the coefficients of fld K (α) are polynomials with
rational coefficients in the elementary symmetric polynomials in θ1 , . . . , θn and so
belong to Q. Hence fld K (α) ∈ Q[x].

Example 6.3.2 The cubic polynomial x 3 + 11x + 4 ∈ Z[x] is irreducible. Let its
three roots be θ1 = θ, θ2 , and θ3 . One of these roots is real and the other two
are nonreal and complex conjugates of one another (Exercise 2 of this Chapter).
Let K = Q(θ ) so that [K : Q] = deg(irrQ (θ)) = deg(x 3 + 11x + 4) = 3. Let α =
(θ + θ 2 )/2 ∈ K . We determine fld K (α). We have

(θ1 + θ12 ) (θ2 + θ22 ) (θ3 + θ32 )
fld K (α) = x − x− x−
2 2 2
= x + a2 x + a1 x + a0 ,
3 2
CB609-06 CB609/Alaca & Williams August 27, 2003 16:53 Char Count= 0

6.3 Field Polynomial 119

where
(θ1 + θ12 ) (θ2 + θ22 ) (θ3 + θ32 )
a2 = − − −
2 2 2
1 1 2
= − (θ1 + θ2 + θ3 ) − (θ1 + θ22 + θ32 ),
2 2
1
a1 = ((θ1 + θ1 )(θ2 + θ2 ) + (θ2 + θ22 )(θ3 + θ32 ) + (θ3 + θ32 )(θ1 + θ12 ))
2 2
4
1
= ((θ1 θ2 + θ2 θ3 + θ3 θ1 ) + (θ1 θ22 + θ12 θ2 + θ2 θ32 + θ22 θ3 + θ3 θ12 + θ32 θ1 )
4
+ (θ12 θ22 + θ22 θ32 + θ32 θ12 )),
1
a0 = − (θ1 + θ12 )(θ2 + θ22 )(θ3 + θ32 )
8
1
= − θ1 θ2 θ3 (1 + θ1 )(1 + θ2 )(1 + θ3 )
8
1
= − θ1 θ2 θ3 (1 + (θ1 + θ2 + θ3 ) + (θ1 θ2 + θ2 θ3 + θ3 θ1 ) + θ1 θ2 θ3 ).
8
Now
x 3 + 11x + 4 = (x − θ1 )(x − θ2 )(x − θ3 ),
so that
θ1 + θ2 + θ3 = 0,
θ1 θ2 + θ2 θ3 + θ3 θ1 = 11,
θ1 θ2 θ3 = −4.
Hence
θ12 + θ22 + θ32 = (θ1 + θ2 + θ3 )2 − 2(θ1 θ2 + θ2 θ3 + θ3 θ1 ) = −22,
θ12 θ22 + θ22 θ32 + θ32 θ12 = (θ1 θ2 + θ2 θ3 + θ3 θ1 )2 − 2θ1 θ2 θ3 (θ1 + θ2 + θ3 ) = 121,
θ1 θ22 + θ12 θ2 + θ2 θ32 + θ22 θ3 + θ3 θ12 + θ32 θ1
= θ1 θ2 (θ2 + θ1 ) + θ2 θ3 (θ3 + θ2 ) + θ3 θ1 (θ1 + θ3 )
= −3θ1 θ2 θ3 = 12, as θ1 + θ2 + θ3 = 0,
so that
a2 = 11, a1 = 36, a0 = 4.
Hence
fld K (α) = x 3 + 11x 2 + 36x + 4,
showing that α ∈ O K .

In the next theorem we relate the field polynomial of α over K to the minimal
polynomial of α over Q.
CB609-06 CB609/Alaca & Williams August 27, 2003 16:53 Char Count= 0

120 Algebraic Number Fields

Theorem 6.3.2 Let K be an algebraic number field of degree n. Let α ∈ K . Then

fld K (α) = (irrQ (α))s ,

where s is the positive integer

n
s= .
deg(irrQ (α))

Proof: Let {α1 = α, α2 , . . . , αn } be a complete set of conjugates of α relative

to K . Then

n
fld K (α) = (x − αk ) ∈ Q[x]
k=1

by Theorem 6.3.1. As fld K (α) has α as a root, we have

irrQ (α) | fld K (α)

in Q[x]. Hence, as Q[x] is a unique factorization domain, we have

fld K (α) = (irrQ (α))s h(x),

where h(x) is a monic polynomial of Q[x], which is not divisible by the irreducible
polynomial irrQ (α), and s is a positive integer. Suppose that h(x) is a nonconstant
polynomial. Then h(αk ) = 0 for some k ∈ {1, 2, . . . , n}.
Now choose θ ∈ K such that K = Q(θ ). Let θ1 = θ, θ2 , . . . , θn be the conjugates
of θ over Q. As α ∈ K there exists a polynomial

r (x) = a0 + a1 x + · · · + an−1 x n−1 ∈ Q[x]

such that α = r (θ). Thus α j = r (θ j ) for j ∈ {1, 2, . . . , n}.

Next let

g(x) = h(r (x)) ∈ Q[x].

Then g(θk ) = h(r (θk )) = h(αk ) = 0. Thus g(x) is a multiple of irrQ (θk ) = irrQ (θ) ∈
Q[x]. Hence g(θ j ) = 0 for j = 1, 2, . . . , n. In particular g(θ) = 0. Thus h(α) =
h(r (θ)) = g(θ) = 0. Hence h(x) is a multiple of irrQ (α) in Q[x], contradicting that
h(x) is not divisible by irrQ (α).
We have shown that h(x) is a constant polynomial; that is, h(x) = c, c ∈ Q. But
h(x) is monic so c = 1. Thus

fld K (α) = (irrQ (α))s

as asserted. Comparing degrees of the polynomials in this equation, we see that

n = deg(fld K (α)) = s deg(irrQ (α))

CB609-06 CB609/Alaca & Williams August 27, 2003 16:53 Char Count= 0

6.3 Field Polynomial 121

so that
n
s= .
deg(irrQ (α))
Theorem 6.3.2 tells us that the conjugates of α with respect to K are the roots of
irrQ (α) in C each repeated s = n/deg(irrQ (α)) times.

Theorem 6.3.3 Let K be an algebraic number field. Let α ∈ O K . Then the

K -conjugates of α are algebraic integers.

Proof: Let α ∈ O K . Then, by Theorem 5.3.2, we have

irrQ (α) ∈ Z[x],
and so by Theorem 6.3.2
fld K (α) ∈ Z[x].
Thus the K -conjugates of α being the roots of a monic polynomial with rational
integer coefficients are algebraic integers.

Suppose α ∈ Q, then
αk = σk (α) = α, k = 1, 2, . . . , n,
so all of the K -conjugates of α are equal. Conversely, if all the K -conjugates of α
are equal then
fld K (α) = (x − α)n .
Hence, by Theorem 6.3.2, we have
(irrQ (α))s = (x − α)n .
But the roots of irrQ (α) are all distinct (Theorem 5.2.1) so that
irrQ (α) = x − α, s = n.
As irrQ (α) ∈ Q[x] we deduce that α ∈ Q. Hence we have shown the following
result.

Theorem 6.3.4 Let K be an algebraic number field. Let α ∈ K . Then all the
K -conjugates of α are equal if and only if α ∈ Q.

If the K -conjugates of α are all distinct then fld K (α) is a product of distinct linear
factors and so by Theorem 6.3.2 we have s = 1 and irrQ (α) = fld K (α). Hence
[Q(α) : Q] = deg(irrQ (α)) = deg(fld K (α)) = n = [K : Q].
CB609-06 CB609/Alaca & Williams August 27, 2003 16:53 Char Count= 0

122 Algebraic Number Fields

Since Q(α) ⊆ K we deduce that K = Q(α). Conversely, if K = Q(α) then

deg(irrQ (α)) = [K : Q] = n, so that by Theorem 6.3.2 s = 1 and fld K (α) =
irrQ (α). Hence the K -conjugates of α are distinct. We have proved the following
theorem.

Theorem 6.3.5 Let K be an algebraic number field. Let α ∈ K . Then all the K -
conjugates of α are distinct if and only if K = Q(α).

Let K = Q(θ ) be an algebraic number field of degree n. Let θ1 = θ, θ2 , . . . , θn

be the conjugates of θ over Q. Using the preceding ideas it is easy to show
that if there are exactly m distinct fields among the conjugate fields Q(θ1 ) =
K , Q(θ2 ), . . . , Q(θn ) then m divides n and each distinct field occurs n/m times
(Exercise 3 of this Chapter). If m = 1 so that Q(θ1 ) = · · · = Q(θn ) = K , the field
K is said to be a normal or Galois extension of Q.
√ √ √ √ √ √
Example √ Let K = Q( 2, 3) = Q( 2 + 3). The conjugates of 2 + 3
√ 6.3.3
are ± 2 ± 3 and the conjugate fields of K all coincide with K as
√ √ √ √
Q(± 2 ± 3) = Q( 2 + 3) = K .

Thus K is a normal field.

√ √
Example 6.3.4 Let K = Q( 3 2) so that K ⊆ R. The conjugates of 3 2 are
√ √ √
2, ω 2, ω2 2,
3 3 3

where ω and ω2 are the two complex cube roots of unity, since
√ √ √ √
irrQ ( 2) = x 3 − 2 = (x − 2)(x − ω 2)(x − ω2 2).
3 3 3 3

The conjugate fields of K are

√ √ √
K 1 = Q( 2) = K , K 2 = Q(ω 2), K 3 = Q(ω2 2).
3 3 3

Clearly as K 1 is a real field, and K 2 , K 3 are not, we√have K 1

= K 2 , K 1
= K 3 .
We show that K 2
= K 3 . Suppose K 2 = K 3 . Then ω2 3 2 ∈ K 2 and so there exist
a, b, c ∈ Q such that
√ √ √
ω2 2 = a + bω 2 + c(ω 2)2 .
3 3 3

Taking complex conjugates, we obtain as ω̄ = ω2

√ √ √
ω 2 = a + bω2 2 + cω( 2)2 .
3 3 3

Subtracting we deduce that

√ √ √
(ω2 − ω) 2 = −b(ω2 − ω) 2 + c(ω2 − ω)( 2)2 ,
3 3 3
CB609-06 CB609/Alaca & Williams August 27, 2003 16:53 Char Count= 0

6.4 Discriminant of a Set of Elements 123

so that
√ √ √
2 = −b 2 + c( 2)2 .
3 3 3

Hence
√
1 + b = c 2.
3

√
Since 3
2∈
/ Q we must have 1 + b = c = 0, so that
√ √
ω2 2 = a − ω 2.
3 3

Thus
√
(ω2 + ω) 2 = a;
3

that is (as ω2 + ω = −1),

√
2 = −a ∈ Q,
3

a contradiction. √ √
Hence all the conjugate fields of Q( 3 2) are distinct, and Q( 3 2) is not a normal
field.
√ √
Example 6.3.5 Let K = Q( 4 2) so that K ⊆ R. The conjugates of 4 2 are
√ √ √ √
2, i 2, − 2, − i 2,
4 4 4 4

as
√ √ √ √ √
irrQ ( 2) = x 4 − 2 = (x − 2)(x − i 2)(x + 2)(x + i 2).
4 4 4 4 4

The conjugate fields of K are

√
Q( 2) = K ,
4

√
Q(i 2) = L (say),
4

√ √
Q(− 2) = Q( 2) = K ,
4 4

√ √
Q(−i 2) = Q(i 2) = L .
4 4

Clearly K
= L as K is a real field and L is a nonreal
√ field.
Hence there are two distinct conjugate fields. Q( 2) is not a normal field.
4

6.4 The Discriminant of a Set of Elements in an Algebraic Number Field

Let K be an algebraic number field of degree n. Let ω1 , ω2 , . . . , ωn be any n
elements of K . An important quantity defined in terms of ω1 , ω2 , . . . , ωn and their
conjugates relative to K is the discriminant D(ω1 , . . . , ωn ). As we shall see the
discriminant has some very nice properties. For example, D(ω1 , . . . , ωn ) is always a
rational number, which is nonzero if and only if ω1 , . . . , ωn are linearly independent
CB609-06 CB609/Alaca & Williams August 27, 2003 16:53 Char Count= 0

124 Algebraic Number Fields

over Q. Moreover, if ω1 , . . . , ωn are all algebraic integers then D(ω1 , . . . , ωn ) is a

rational integer.

Definition 6.4.1 (Discriminant of n elements in an algebraic number field of

degree n) Let K be an algebraic number field of degree n. Let ω1 , . . . , ωn be
n elements of the field K . Let σk (k = 1, 2, . . . , n) denote the n distinct mono-
morphisms : K −→ C. For i = 1, . . . , n let
ωi(1) = σ1 (ωi ) = ωi , ωi(2) = σ2 (ωi ), . . . , ωi(n) = σn (ωi )
denote the conjugates of ωi relative to K . Then the discriminant of {ω1 , . . . , ωn } is
(1)
(1) 2
ω1 ω (1)
· · · ω
(2) 2 n
ω ω2(2) · · · ωn(2)
1
D(ω1 , . . . , ωn ) = . .. .. .
.. . · · · .

ω(n) ω(n) · · · ω(n)
1 2 n

√ √
Example 6.4.1 Let K = Q( 2, 3) and choose
√ √ √ √
ω1 = 1, ω2 = 2, ω3 = 3, ω4 = 2 + 3.
By Example 5.6.1 we know that K is a quartic field. The four monomor-
phisms : K −→ C are given by
√ √ √ √ √ √
σ1 (a + b 2 + c 3 + d 6) = a + b 2 + c 3 + d 6,
√ √ √ √ √ √
σ2 (a + b 2 + c 3 + d 6) = a + b 2 − c 3 − d 6,
√ √ √ √ √ √
σ3 (a + b 2 + c 3 + d 6) = a − b 2 + c 3 − d 6,
√ √ √ √ √ √
σ4 (a + b 2 + c 3 + d 6) = a − b 2 − c 3 + d 6,
where a, b, c, d ∈ Q. Hence
√ √ √ √ 2
1 +
√2 √3 √ 2 √3
√ √ √ √
1 √2 −√3 √2 − √3 .
D(1, 2, 3, 2 + 3) =
1 −√2 −√2 + √3
√3
1 − 2 − 3 − 2 − 3
As the fourth column of the determinant is the sum of the second and third columns,
we deduce that
√ √ √ √
D(1, 2, 3, 2 + 3) = 0.

We can now define the discriminant D(α) of an element α of an algebraic number

field.

Definition 6.4.2 (Discriminant D(α) of an element α) Let K be an algebraic

number field of degree n. Let α ∈ K . Then we define the discriminant D(α) of
CB609-06 CB609/Alaca & Williams August 27, 2003 16:53 Char Count= 0

6.4 Discriminant of a Set of Elements 125

α by

D(α) = D(1, α, α 2 , . . . , α n−1 ).

Theorem 6.4.1 Let K be an algebraic number field of degree n. Let α ∈ K . Then

D(α) = (α (i) − α ( j) )2 ,
1≤i< j≤n

where α (1) = α, α (2) , . . . , α (n) are the conjugates of α with respect to K .

Proof: If z 1 , z 2 , . . . , z n are complex numbers we have the value of the determinant

n−1
z1 z 1n−2 ··· z1 1
n−1
z z 2n−2 ··· z2 1
2
..
. .. .. .. = (z i − z j )
. ··· . . 1≤i< j≤n
z n−1 z nn−2 ··· zn 1
n

(see, for example, [3, pp. 17–18]). Interchanging columns 1 and n, columns 2 and
n − 1, etc., we obtain the evaluation of the Vandermonde determinant

1 z 1 · · · z 1n−2 z 1n−1

1 z 2 · · · z n−2 z n−1
2 2
.. .. . . = (−1)[n/2] (z i − z j )
. . ··· .. ..
1≤i< j≤n
1 z n · · · z n−2 z n−1
n n
n(n−1)
= (−1) 2 (z i − z j )
1≤i< j≤n

= (z j − z i ),
1≤i< j≤n

n n(n − 1)
≡ (mod 2).
2 2
Hence for any α ∈ K we have

1 α α2 ··· α n−1

1 α (2) (α (2) )
2
··· (α (2) )
n−1

.. .. .. .. = (α ( j) − α (i) )
. . . ··· .
1≤i< j≤n
1 α (n) (n)
(α )
2
··· (α (n) )n−1
CB609-06 CB609/Alaca & Williams August 27, 2003 16:53 Char Count= 0

126 Algebraic Number Fields

so that
 2

D(α) =  (α ( j) − α (i) ) =
2
(α (i) − α ( j) ) .
1≤i< j≤n 1≤i< j≤n

Definition 6.4.3 (Discriminant of a polynomial) Let

f (x) = an x n + an−1 x n−1 + · · · + a1 x + a0 ∈ C[x],

where n ∈ N and an
= 0. Let x1 , . . . , xn ∈ C be the roots of f (x). The discriminant
of f (x) is the quantity

disc( f (x)) = an2n−2 (xi − x j )2 ∈ C.
1≤i< j≤n

Clearly f (x) has a repeated root if and only if disc( f (x)) = 0. The discriminant is
a 2n−2
times a symmetric polynomial in x1 , . . . , xn . The degree of disc( f (x)) in each
xi is 2(n − 1). Thus when disc( f (x)) is expressed as a function of a0 , a1 , . . . , an ,
it consists of terms
k n
an−1 k1 an−2 k2 a0
Can 2n−2
··· ,
an an an
where

k1 + k2 + · · · + kn ≤ 2n − 2,

so that disc( f (x)) is a polynomial in the coefficients of f (x).

Theorem 6.4.2 Let K be an algebraic number field of degree n. Let α ∈ K . Then

D(α) = disc(fld K (α)).

Proof: Let α (1) = α, α (2) , . . . , α (n) be the K -conjugates of α. Then the roots of
fld K (α) are α (1) , . . . , α (n) . Hence, by Definition 6.4.3 and Theorem 6.4.1, we have
2
disc(fld K (α)) = (α (i) − α ( j) ) = D(α).
1≤i< j≤n
√ √
Example 6.4.2 Let K = Q( 3 2) and choose α = 3 2. Then the conjugates of α
are
√ √ √
α1 = 2, α2 = ω 2, α3 = ω2 2,
3 3 3
CB609-06 CB609/Alaca & Williams August 27, 2003 16:53 Char Count= 0

6.4 Discriminant of a Set of Elements 127

√
where ω = (−1 + −3)/2. Then
√
α1 − α2 = (1 − ω) 2,
3

√
α2 − α3 = ω(1 − ω) 2,
3

√
α3 − α1 = ω2 (1 − ω) 2,
3

so that
(α1 − α2 )(α2 − α3 )(α3 − α1 ) = (1 − ω)3 2.
Now
(1 − ω)3 = 1 − 3ω + 3ω2 − ω3 = −3(ω − ω2 ),
so that
(1 − ω)6 = 32 (ω − ω2 )2 = 32 (ω2 + ω − 2) = −33 .
Hence
D(α) = ((α1 − α2 )(α2 − α3 )(α3 − α1 ))2 = (1 − ω)6 22 = −22 · 33 .

Theorem 6.4.3 Let K be an algebraic number field of degree n. Let α ∈ K . Then

K = Q(α) if and only if D(α)
= 0.

Proof: We have by Theorems 6.3.5 and 6.4.1

K = Q(α) ⇐⇒ K -conjugates of α are distinct ⇐⇒ D(α)
= 0.

Theorem 6.4.4 Let K be an algebraic number field of degree n.

(a) If ω1 , . . . , ωn ∈ K then
D(ω1 , . . . , ωn ) ∈ Q.

(b) If ω1 , . . . , ωn ∈ O K then
D(ω1 , . . . , ωn ) ∈ Z.

Proof: (a) By Theorem 6.1.1 we have K = Q(θ) for some θ ∈ K . Then, for i =
1, 2, . . . , n, we have
ωi = c0 i + c1 i θ + · · · + cn−1 i θ n−1 ,
where c0 i , . . . , cn−1 i ∈ Q. Hence, for j = 1, 2, . . . , n, we have
( j)
ωi = c0 i + c1 i θ j + · · · + cn−1 i θ n−1
j , (6.4.1)
CB609-06 CB609/Alaca & Williams August 27, 2003 16:53 Char Count= 0

128 Algebraic Number Fields

where θ1 = θ, θ2 , . . . , θn are the conjugates of θ over Q and ωi(1) , . . . , ωi(n) are the
K -conjugates of ωi (i = 1, 2, . . . , n). Using the expressions (6.4.1) in Definition
6.4.1, we see that any permutation of the conjugates of θ leaves D(ω1 , . . . , ωn )
invariant as it merely causes a permutation of the rows of the matrix of which
D(ω1 , . . . , ωn ) is the square of the determinant. Hence D(ω1 , . . . , ωn ) is a sym-
metric function of the roots of the polynomial
(x − θ1 )(x − θ2 ) · · · (x − θn ) = x n + an−1 x n−1 + · · · + a0 ,

where a0 , a1 , . . . , an−1 ∈ Q. By the symmetric function theorem, D(ω1 , . . . , ωn )

is a polynomial in the coefficients a0 , a1 , . . . , an−1 and hence a rational number.
(b) If ω1 , . . . , ωn ∈ O K then D(ω1 , . . . , ωn ), being obtained from them and
their conjugates by a series of additions and multiplications, is also in O K . Since
D(ω1 , . . . , ωn ) ∈ Q, by Theorem 4.2.4 we have D(ω1 , . . . , ωn ) ∈ Z.
(c) If the set {ω1 , . . . , ωn } is linearly dependent over Q, then there exist rational
numbers c1 , . . . , cn not all zero such that
c1 ω1 + · · · + cn ωn = 0.

Applying each monomorphism σk (k = 1, . . . , n) to this equation, we obtain the

following homogeneous system of n linear equations in the n quantities c1 , . . . , cn :


 c1 ω1(1) + · · · + cn ωn(1) = 0,


c1 ω1(2) + · · · + cn ωn(2) = 0,

 ···

 c ω(n) + · · · + c ω(n) = 0.
1 1 n n

As this system has a nontrivial solution (c1 , . . . , cn )

= (0, . . . , 0) ∈ Qn , its deter-
minant must be zero. Hence
D(ω1 , . . . , ωn ) = 0.

Now suppose that the set {ω1 , . . . , ωn } is linearly independent over Q. Then
{ω1 , . . . , ωn } is a basis for the vector space K over the field Q. In particular as
1, θ, . . . , θ n−1 ∈ K there exist rational numbers ci j (i, j = 1, . . . , n) such that


 1 = c11 ω1 + · · · + c1n ωn ,

θ = c21 ω1 + · · · + c2n ωn ,

 ···
 n−1
θ = cn1 ω1 + · · · + cnn ωn .
Hence
D(θ ) = D(1, θ, . . . , θ n−1 ) = |det(ci j )|2 D(ω1 , . . . , ωn ).

As K = Q(θ), by Theorem 6.4.3 we know that D(θ)

= 0 so that

D(ω1 , . . . , ωn )
= 0.
CB609-06 CB609/Alaca & Williams August 27, 2003 16:53 Char Count= 0

6.5 Basis of an Ideal 129

6.5 Basis of an Ideal

We now use our knowledge of the properties of the discriminant to show that every
ideal in the ring O K of integers of an algebraic number field K has a finite basis
considered as an Abelian group, that is, as a Z-module. Thus O K is Noetherian.
We first prove a preliminary result.

Theorem 6.5.1 Let K be an algebraic number field with [K : Q] = n. Let I be a

nonzero ideal in O K . Then there exist η1 , . . . , ηn ∈ I such that

D(η1 , . . . , ηn )
= 0.

Proof: By Theorem 6.1.2 we have K = Q(θ ) for some θ ∈ O K . By Theorem 6.4.3

D(θ)
= 0. Further, by Theorem 6.1.7, as I is a nonzero ideal of O K , there exists
c ∈ I ∩ Z with c
= 0. Hence, as I is an ideal of O K ,

η1 = c, η2 = c θ, . . . , ηn = c θ n−1 ∈ I

and

D(η1 , . . . , ηn ) = D(c, cθ, . . . , cθ n−1 ) = c2n D(1, θ, . . . , θ n−1 ) = c2n D(θ)

= 0.

We are now in a position to prove that every ideal of the ring of integers of an
algebraic number field has a finite basis.

Theorem 6.5.2 Let K be an algebraic number field of degree n. Let I be a nonzero

ideal of O K . Then there exist elements η1 , . . . , ηn of I such that every element α of
I can be expressed uniquely in the form

α = x1 η1 + · · · + xn ηn ,

where x1 , . . . , xn ∈ Z.

Proof: As I is a nonzero ideal of O K , by Theorem 6.5.1 there exists a set

{η1 , . . . , ηn } of elements of I such that D(η1 , . . . , ηn )
= 0. By Theorem 6.4.4
D(η1 , . . . , ηn ) ∈ Z so that |D(η1 , . . . , ηn )| is a positive integer. Let

S = {|D(η1 , . . . , ηn )| : η1 , . . . , ηn ∈ I, D(η1 , . . . , ηn )
= 0}.

Clearly S is a nonempty set of positive integers and thus contains a least member,
say |D(η1 , . . . , ηn )|, η1 , . . . , ηn ∈ I. As D(η1 , . . . , ηn )
= 0, by Theorem 6.4.4(c)
{η1 , . . . , ηn } is a basis for the vector space K over Q. Let α ∈ I . Then there exist
unique rational numbers x1 , . . . , xn such that

α = x1 η1 + · · · + xn ηn .
CB609-06 CB609/Alaca & Williams August 27, 2003 16:53 Char Count= 0

130 Algebraic Number Fields

Suppose at least one of the xi is not an integer. By permuting η1 , . . . , ηn , if necessary,

we may suppose that x1 ∈ / Z. Then there is a unique integer l such that

l < x1 < l + 1.

Set

γ = α − lη1 .

As α ∈ I and η1 ∈ I we see that γ ∈ I . Moreover,

γ = (x1 − l)η1 + x2 η2 + · · · + xn ηn .

Applying each monomorphism σk (k = 1, 2, . . . , n) to this equation, we obtain

 (1)

 γ = (x1 − l)η1(1) + x2 η2(1) + · · · + xn ηn(1) ,

 (2)
γ = (x1 − l)η1(2) + x2 η2(2) + · · · + xn ηn(2) ,

 ···

 (n)
γ = (x1 − l)η1(n) + x2 η2(n) + · · · + xn ηn(n) ,

where γ (1) = γ , γ (2) , . . . , γ (n) are the K -conjugates of γ and ηi(1) = ηi ,

ηi(2) , . . . , ηi(n) are the K -conjugates of ηi (i = 1, 2, . . . , n). By Cramer’s rule, we
deduce that
(1)
γ η2(1) · · · ηn(1)

γ (2) η(2) · · · η(2)
2 n
. .. ..
.. . ··· .

γ (n) η(n) · · · η(n)
2 n
x1 − l = (1) .
η1 η (1)
· · · η (1)
(2) 2 n
η η2(2) · · · ηn(2)
1
. .. ..
.. . ··· .

η(n) η(n) · · · η(n)
1 2 n

Hence
D(γ , η2 , . . . , ηn )
(x1 − l)2 =
D(η1 , η2 , . . . , ηn )
so that

0 < |D(γ , η2 , . . . , ηn )| = (x1 − l)2 |D(η1 , η2 , . . . , ηn )| < |D(η1 , η2 , . . . , ηn )|.

This contradicts the minimality of |D(η1 , η2 , . . . , ηn )|. Hence all the xi are integers
and each element α ∈ I can be expressed uniquely in the form α = x1 η1 + · · · +
xn ηn .
CB609-06 CB609/Alaca & Williams August 27, 2003 16:53 Char Count= 0

6.5 Basis of an Ideal 131

Clearly, as η1 , . . . , ηn ∈ I and I is an ideal, we have

Zη1 + · · · + Zηn = {k1 η1 + · · · + kn ηn | k1 , . . . , kn ∈ Z} ⊆ I

and Theorem 6.5.2 tells us that

I ⊆ Zη1 + · · · + Zηn

so that

I = Zη1 + · · · + Zηn ,

showing that I is a finitely generated Z-module.

We now use Theorem 3.5.3 and Theorem 6.5.2 to show that the ring of integers
of an algebraic number field is Noetherian.

Theorem 6.5.3 Let K be an algebraic number field. Then O K is a Noetherian

domain.

Proof: Z and O K are integral domains with Z ⊆ O K . Z is a Noetherian domain

(Example 3.1.3). O K = 1 is a finitely generated Z-module by Theorem 6.5.2.
Hence by Theorem 3.5.3 O K is a Noetherian domain.
Alternatively we can avoid the use of Theorem 3.5.3 by arguing as follows. Let I
be an ideal of O K . If I = {0} then I = 0 is finitely generated. If I
= {0} then I is
finitely generated by Theorem 6.5.2. Hence every ideal of O K is finitely generated
and thus the domain O K is Noetherian.

Definition 6.5.1 (Basis of an ideal) Let K be an algebraic number field of degree

n. Let I be a nonzero ideal of O K . If {η1 , . . . , ηn } is a set of elements of I such that
every element α ∈ I can be expressed uniquely in the form

α = x1 η1 + · · · + xn ηn (x1 , . . . , xn ∈ Z)

then {η1 , . . . , ηn } is called a basis for the ideal I .

As the representation of each element α of a nonzero ideal I by a basis

{η1 , . . . , ηn } of I is unique, the basis elements η1 , . . . , ηn are linearly indepen-
dent over Q.
By Theorem 6.5.2 every ideal of the ring of integers of an algebraic number field
possesses a basis. Our next result enables us to recognize when a set of elements
{λ1 , . . . , λn } of an ideal in the ring of integers of an algebraic number field is a
basis for the ideal.

Theorem 6.5.4 Let K be an algebraic number field of degree n. Let I be a nonzero

ideal of O K .
CB609-06 CB609/Alaca & Williams August 27, 2003 16:53 Char Count= 0

132 Algebraic Number Fields

(a) Let {η1 , . . . , ηn } and {λ1 , . . . , λn } be two bases for I . Then

D(η1 , . . . , ηn ) = D(λ1 , . . . , λn )

and

n
ηi = ci j λ j , i = 1, 2, . . . , n,
j=1

where ci j (i, j = 1, 2, . . . , n) are rational integers such that

det(ci j ) = ±1.

(b) Let {η1 , . . . , ηn } be a basis for I and let λ1 , . . . , λn ∈ I be such that

D(λ1 , . . . , λn ) = D(η1 , . . . , ηn ).

Then {λ1 , . . . , λn } is a basis for I .

Proof: (a) As {λ1 , . . . , λn } is a basis for I , we have

I = Zλ1 + · · · + Zλn .

Since η1 , . . . , ηn ∈ I there exist ci j ∈ Z (i, j = 1, 2, . . . , n) such that

n
ηi = ci j λ j , i = 1, 2, . . . , n. (6.5.1)
j=1

As {η1 , . . . , ηn } is a basis for I , we have

I = Zη1 + · · · + Zηn .

Since λ1 , . . . , λn ∈ I there exist di j ∈ Z (i, j = 1, 2, . . . , n) such that

n
λj = d jk ηk , j = 1, 2, . . . , n.
k=1

Thus, for i = 1, 2, . . . , n, we have

 

n
n
n n
ηi = ci j d jk ηk =  ci j d jk  ηk .
j=1 k=1 k=1 j=1

As {η1 , . . . , ηn } is a basis for I , η1 , . . . , ηn are linearly independent over Q, so that

n
1, if i = k,
ci j d jk =
0, if i
= k.
j=1

We define the n × n matrices C and D by

C = [ci j ] , D = [di j ],
CB609-06 CB609/Alaca & Williams August 27, 2003 16:53 Char Count= 0

6.5 Basis of an Ideal 133

so that C and D have rational integer entries. Then

C D = In ,

where In is the n × n identity matrix. Thus

det(C)det(D) = det(C D) = det(In ) = 1.

But det(C), det(D) ∈ Z so

det(C) = det(D) = ±1.

From (6.5.1) we have

D(η1 , . . . , ηn ) = (det(ci j ))2 D(λ1 , . . . , λn ) = (det(C))2 D(λ1 , . . . , λn )

so that

D(η1 , . . . , ηn ) = (±1)2 D(λ1 , . . . , λn ) = D(λ1 , . . . , λn ).

This completes the proof of part (a).

(b) As {η1 , . . . , ηn } is a basis for I and λ1 , . . . , λn ∈ I , there exist di j ∈ Z (i, j =
1, 2, . . . , n) such that

n
λi = di j η j , i = 1, 2, . . . , n.
j=1

Hence

D(λ1 , . . . , λn ) = (det(di j ))2 D(η1 , . . . , ηn ).

As D(λ1 , . . . , λn ) = D(η1 , . . . , ηn ) we deduce that

(det(di j ))2 = 1

so that

det(di j ) = ±1.

Thus the matrix D = (di j ) has an inverse D −1 = C = (ci j ) all of whose entries are
integers, and

n
ηj = ci j λ j (i = 1, 2, . . . , n).
j=1

Let α ∈ I . Then, as {η1 , . . . , ηn } is a basis for I , there exist a1 , . . . , an ∈ Z such

that

n
α= ai ηi .
i=1
CB609-06 CB609/Alaca & Williams August 27, 2003 16:53 Char Count= 0

134 Algebraic Number Fields

Hence
n

n
n
n
α= ai ci j λ j = ai ci j λj,
i=1 j=1 j=1 i=1

n
where each i=1 ai ci j ∈ Z ( j = 1, 2, . . . , n). This proves that every element α of
I can be expressed in the form

α = b1 λ1 + · · · + bn λn

for some integers b1 , . . . , bn .

Now suppose that α can be expressed in more than one way in this form, say,

α = b1 λ1 + · · · + bn λn = b1 λ1 + · · · + bn λn ,

where b1 , . . . , bn , b1 , . . . , bn ∈ Z. Hence

e1 λ1 + · · · + en λn = 0,

where ei = bi − bi ∈ Z (i = 1, 2, . . . , n). If at least one of the ei is nonzero then

λ1 , . . . , λn are linearly dependent over Q and so by Theorem 6.4.4(c) we have

D(λ1 , . . . , λn ) = 0.

Hence

D(η1 , . . . , ηn ) = 0,

so that η1 , . . . , ηn are linearly dependent over Q, contradicting that {η1 , . . . , ηn }

is a basis for I . Hence ei = 0 (i = 1, 2, . . . , n) and so bi = bi (i = 1, 2, . . . , n),
establishing that α is uniquely expressible in the form b1 λ1 + · · · + bn λn with
b1 , . . . , bn ∈ Z.
This completes the proof that {λ1 , . . . , λn } is a basis for I .

√ √ √
Example 6.5.1 Let K = Q( 7) so that O K = Z + Z 7 = {a + b 7 |√a, b ∈ Z}
by Theorem 5.4.2. Let I be the principal ideal of O K generated by 2 + 7. Then
√ √
I = {(a + b 7)(2 + 7) | a, b ∈ Z}
√ √
= {a(2 + 7) + b(7 + 2 7) | a, b ∈ Z}
√ √
= (2 + 7)Z + (7 + 2 7)Z,
√ √
so that {2 + 7, 7 + 2 7} is a basis for I . However, a little more effort yields
a“simpler” basis, that is, one having a rational integer as one of the basis elements.
CB609-06 CB609/Alaca & Williams August 27, 2003 16:53 Char Count= 0

6.5 Basis of an Ideal 135

We have
√ √
I = {(a + b 7)(2 + 7) | a, b ∈ Z}
√
= {(2a + 7b) + (a + 2b) 7 | a, b ∈ Z}
√
= {(2(c − 2b) + 7b) + c 7 | b, c ∈ Z}
√
= {3b + c(2 + 7) | b, c ∈ Z}
√
= 3Z + (2 + 7)Z,
√
showing that {3, 2 + 7} is a basis for I .

If {η1 , . . . , ηn } and {λ1 , . . . , λn } are two bases for the same nonzero ideal of the
ring of integers of an algebraic number field then we know by Theorem 6.5.4 that

D(η1 , . . . , ηn ) = D(λ1 , . . . , λn ).

Hence we can make the following definition.

Definition 6.5.2 (Discriminant of an ideal) Let K be an algebraic number field

of degree n. Let I be a nonzero ideal of O K . Let {η1 , . . . , ηn } be a basis of I . Then
the discriminant D(I ) of the ideal I is the nonzero integer given by

D(I ) = D(η1 , . . . , ηn ).

Example√6.5.2 We determine the discriminant of the ideal I in Example 6.5.1. As

{3, 2 + 7} is a basis for I , we have
√
D(I ) = D(3, 2 + 7)
√
3 2 + 7 2
= √
3 2 − 7
√ √
= (3(2 − 7) − 3(2 + 7))2
√
= (−6 7)2
= 252.

Next we consider bases of ideals in the ring of integers of a quadratic field.

Theorem 6.5.5 Let K be a quadratic field. Let m be the unique squarefree integer
√
such that K = Q( m).

(a) m
≡ 1 (mod 4). Let a, b, c ∈ Z with a
= 0 and c
= 0. Then
√ √
{a, b + c m} is a basis for the ideal a, b + c m

if and only if
c | a, c | b, ac | b2 − mc2 . (6.5.2)
CB609-06 CB609/Alaca & Williams August 27, 2003 16:53 Char Count= 0

136 Algebraic Number Fields

(b) m ≡ 1 (mod 4). Let a, b, c ∈ Z with a

= 0, c
= 0, and b ≡ c (mod 2). Then
√ √
b+c m b+c m
{a, } is a basis for the ideal a,
2 2
if and only if

c | a, c | b, 4ac | b2 − mc2 . (6.5.3)

Proof: (a) Suppose first that (6.5.2) holds. Then there are integers x, y, z such that

a = cx, b = cy, b2 − mc2 = acz.

Hence

bx − ay = 0, by − az = mc.
√
Let α ∈ I = a, b + c m. Then there exist θ ∈ O K and φ ∈ O K such that
√
α = θa + φ(b + c m).

As θ, φ ∈ O K and m
≡ 1 (mod 4), by Theorem 5.4.2 there exist integers r, s, t, u
such that
√ √
θ = r + s m, φ = t + u m.

Hence
√ √ √
α = (r + s m)a + (t + u m)(b + c m)
√
= (ra + tb + umc) + (sa + tc + ub) m
√
= s(bx − ay) + (ra + tb + u(by − az)) + (scx + tc + ucy) m
√
= (r − sy − uz)a + (t + sx + uy)(b + c m),
√
proving that {a, b + c m} is a basis for I .
√ √
Conversely, suppose that {a, b + c m} is a basis for the ideal I = a, b + c m.
√ √ √
As ma ∈ I and m(b + c m) ∈ I there exist integers x, y, u, v such that
√ √
ma = xa + y(b + c m),
√ √ √
m(b + c m) = ua + v(b + c m).
√
Equating coefficients of 1 and m, we obtain

xa + yb = 0,
yc = a,
ua + vb = cm,
vc = b.
CB609-06 CB609/Alaca & Williams August 27, 2003 16:53 Char Count= 0

6.6 Prime Ideals in Rings of Integers 137

From the second and fourth equations, we see that c | a and c | b respectively. From
the third and fourth equations, we obtain

uac + b2 = c2 m

so that

ac | b2 − mc2 .

(b) This case can be treated similarly to part (a).

6.6 Prime Ideals in Rings of Integers

In Theorem 1.5.6 we saw that a maximal ideal of an integral domain D is always
a prime ideal. We noted that the converse is not always true but that it is true in a
principal ideal domain. In this section we show the important result that a prime
ideal is always maximal in the ring of integers of an algebraic number field.

Theorem 6.6.1 Let P be a prime ideal of the ring O K of integers of an algebraic

number field K . Then P is a maximal ideal of O K .

Proof: Suppose that the assertion of the theorem is false. Then there exists a prime
ideal P1 of O K that is not a maximal ideal. Let S be the set of all proper ideals of
O K that strictly contain P1 . As P1 is not a maximal ideal, S is a nonempty set. By
Theorem 6.5.3 O K is a Noetherian domain. Hence, by Theorem 3.1.3, S contains
a maximal element; that is, there is a maximal ideal P2 such that

P1 ⊂ P2 ⊂ O K .

By Theorem 1.5.6 P2 is a prime ideal. Since every nonzero ideal in O K contains

a nonzero rational integer (Theorem 6.1.7) we see that P1 ∩ Z
= {0}. Hence, by
Theorem 1.6.2, P1 ∩ Z is a prime ideal of Z. But Z is a principal ideal domain
(Theorem 1.4.1) so P1 ∩ Z = p for some p ∈ Z. By Theorem 1.5.4 p is a prime.
Thus

p = P1 ∩ Z ⊆ P2 ∩ Z ⊆ Z.

Now P2 ∩ Z
= Z as 1 ∈
/ P2 , so as p is a maximal ideal of Z (Theorem 1.5.7), we
have

P1 ∩ Z = P2 ∩ Z = p.

As P1 ⊂ P2 there exists α ∈ P2 with α ∈ / P1 . Since α ∈ O K there exist a positive

integer k and a0 , . . . , ak−1 ∈ Z such that

α k + ak−1 α k−1 + · · · + a1 α + a0 = 0,
CB609-06 CB609/Alaca & Williams August 27, 2003 16:53 Char Count= 0

138 Algebraic Number Fields

and so

α k + ak−1 α k−1 + · · · + a1 α + a0 ∈ P1 .

Let l be the least positive integer for which there exist b0 , . . . , bl−1 ∈ Z such that

αl + bl−1 αl−1 + · · · + b1 α + b0 ∈ P1 . (6.6.1)

Now, as α ∈ P2 , we have

αl + bl−1 αl−1 + · · · + b1 α = α(αl−1 + bl−1 αl−2 + · · · + b1 ) ∈ P2 .

Hence, as P1 ⊂ P2 and P2 is an ideal,

b0 = (αl + · · · + b1 α + b0 ) − (αl + · · · + b1 α) ∈ P2 .

But b0 ∈ Z so

b0 ∈ P2 ∩ Z = P1 ∩ Z

and thus b0 ∈ P1 . From (6.6.1) we deduce that

αl + bl−1 αl−1 + · · · + b1 α ∈ P1 .

If l = 1 then α ∈ P1 , contradicting α ∈
/ P1 . Hence l ≥ 2 and

α(αl−1 + · · · + b1 ) ∈ P1 .

Since P1 is a prime ideal and α ∈

/ P1 we deduce that

αl−1 + · · · + b1 ∈ P1 ,

contradicting the minimality of l since l − 1 is a positive integer as l ≥ 2.

Exercises
1. Let K be an algebraic number field of degree n. Let θ ∈ K be such that K = Q(θ ).
Let θ1 = θ, θ2 , . . . , θn be the conjugates of θ over Q. Let α ∈ K so there exist unique
rational numbers c0 , c1 , . . . , cn−1 such that

α = c0 + c1 θ + · · · + cn−1 θ n−1 .

For k = 1, 2, . . . , n let

αk = c0 + c1 θk + · · · + cn−1 θkn−1

so that α1 = α. Prove that the set of conjugates {α1 , α2 , . . . , αn } of α relative to K does

not depend on the choice of θ .
2. Prove that the cubic equation x 3 + ax + b = 0, where a, b ∈ R, has three distinct
real roots if −4a 3 − 27b2 > 0, one real and two nonreal complex conjugate roots if
−4a 3 − 27b2 < 0, and at least two equal real roots if −4a 3 − 27b2 = 0.
CB609-06 CB609/Alaca & Williams August 27, 2003 16:53 Char Count= 0

Exercises 139

3. Let K = Q(θ ) be an algebraic number field of degree n. Let θ1 = θ, θ2 , . . . , θn be

the conjugates of θ over Q. Suppose that there are exactly m distinct fields among
Q(θ1 ), . . . , Q(θn ). Prove that m | n and each field occurs n/m times.
√
4. Let m be a squarefree integer. Let K = Q( m). Prove that
√ √
σ1 (x + y m) = x + y m (x, y ∈ Q)
and
√ √
σ2 (x + y m) = x − y m (x, y ∈ Q)
are the only monomorphisms from K to C.
√
5. Let m be a cubefree integer. Let K = Q( 3 m). Determine all the monomorphisms from
K to C.
√ √
6. Let θ = 3 1 + i + 3 1 − i. Determine all the monomorphisms from Q(θ ) to C.
7. Let θ be a root of the equation x 6 + 2x 2 + 2 = 0. Let K = Q(θ). How many distinct
elements are there in the complete set of conjugates of α = θ 2 + θ 4 relative to K ?
8. Let θ be a root of the equation x 3 + 2x + 2 = 0. Let K = Q(θ) and α = θ − θ 2 .
Determine the field polynomial of α over K .
9. Let K = Q(θ ), where θ 3 −√4θ + 2 = 0. Let α = θ + √ θ 2 ∈ K . Determine D(α).
10. Find a basis for the ideal 5 in O K , where√K = Q( √5). √
11. Determine the discriminant of the ideal 5 + 2, 7 + 2 2 in O K , where K = Q( 2).
12. Let m be a squarefree integer ≡ 2 (mod 4). Prove that
√ √
2, m = 2Z + mZ.
13. Let m be a squarefree integer ≡ 3 (mod 4). Prove that
√ √
2, 1 + m = 2Z + (1 + m)Z.
14. Let m be a squarefree integer ≡ 1 (mod 4). Is
√ √
1+ m (1 + m)
2, = 2Z + Z?
2 2
15. Prove that the discriminant D of the cubic polynomial x 3 + ax 2 + bx + c ∈ Z[x] is
D = a 2 b2 − 4b3 − 4a 3 c − 27c2 + 18abc.
Deduce that D ≡ 0 or 1 (mod 4).
16. Prove that the discriminant of the quartic polynomial x 4 + ax 2 + bx + c ∈ Z[x] is
D = 16a 4 c − 4a 3 b2 − 128a 2 c2 + 144ab2 c − 27b4 + 256c3 .
Deduce that D ≡ 0 or 1 (mod 4).
17. Let K be an algebraic number field. Let α ∈ K . Let α be a conjugate of α relative to
K . Prove that D(α) = D(α ).
18. Let K be an algebraic number field. Let α ∈ K . Let β be a conjugate of α relative to
K . Prove that fld K (α) = fld K (β).
19. Let K be an algebraic number field. Let α, β ∈ K be such that fld K (α) = fld K (β). Prove
that α and β are conjugates relative to K .
20. Let K = Q(θ), where θ 3 + 4θ − 2 = 0. Is K = Q(θ + θ 2 )?
CB609-06 CB609/Alaca & Williams August 27, 2003 16:53 Char Count= 0

140 Algebraic Number Fields

21. Let K = Q(θ ), where θ 4 − 4θ 2 + 8 = 0. Find a rational number c such that Q(θ +
cθ 3 )
= K .
22. Prove that the discriminant of the trinomial polynomial x n + ax + b ∈ Z[x], where n
is an integer ≥ 2, is
(−1)(n−1)(n−2)/2 (n − 1)n−1 a n + (−1)n(n−1)/2 n n bn−1 .
23. Prove that the discriminant of the trinomial polynomial x n + ax r + b ∈ Z[x], where n
and r are integers satisfying n > r ≥ 1 and (n, r ) = 1, is
(−1)(n−1)(n−2)/2 (n − r )n−r r r a n br −1 + (−1)n(n−1)/2 n n bn−1 .

Suggested Reading
1. E. T. Bell, Gauss and the early development of algebraic numbers, National Mathematics
Magazine 18 (1944), 188–204, 219–233.
Bell provides a very readable account of the early development of algebraic numbers.
2. R. L. Goodstein, The discriminant of a certain polynomial, Mathematical Gazette 53
(1969), 60–61.
The formula for the discriminant of x n + ax r + b is derived
3. L. Mirsky, An Introduction to Linear Algebra, Oxford University Press, London 1972.
The evaluation of the Vandermonde determinant is carried out on pages 17 and 18.
4. D. W. Masser, The discriminants of special equations, Mathematical Gazette 50 (1966),
158–160.
The formula for the discriminant of x n + ax + b is derived.

Biographies
1. The website

http://www-groups.dcs.st-and.ac.uk/˜history/

has a biography of A.-T. Vandermonde (1735–1796). Nowhere in his four mathematical

papers does the so-called Vandermonde determinant appear!
CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

7
Integral Bases

7.1 Integral Basis of an Algebraic Number Field

A basis of the principal ideal of the ring O K of integers of an algebraic number
field K generated by 1, that is, O K itself, is called an integral basis for K .

Definition 7.1.1 (Integral basis of an algebraic number field) Let K be an alge-

braic number field. A basis for O K is called an integral basis for K .

In view of this definition the following theorem, which gives an integral basis
for a quadratic field, is just a restatement of Theorem 5.4.2.

Theorem 7.1.1 Let K be a quadratic field. Let m be the unique squarefree integer
√ √
suchthat K√=Q( m). Then {1, m} is an integral basis for K if m ≡ 1 (mod 4)
and 1, 1+2 m is an integral basis for K if m ≡ 1 (mod 4).

If {η1 , . . . , ηn } and {λ1 , . . . , λn } are two integral bases for an algebraic num-
ber field K then Theorem 6.5.4 shows that D(η1 , . . . , ηn ) = D(λ1 , . . . , λn ), and
that if {η1 , . . . , ηn } is an integral basis for K and λ1 , . . . , λn ∈ O K are such that
D(λ1 , . . . , λn ) = D(η1 , . . . , ηn ) then {λ1 , . . . , λn } is also an integral basis for K .
We can therefore make the following definition.

Definition 7.1.2 (Discriminant of an algebraic number field) Let K be an alge-

braic number field of degree n. Let {η1 , . . . , ηn } be an integral basis for K . Then
D(η1 , . . . , ηn ) is called the discriminant of K and is denoted by d(K ).

Clearly if K is an algebraic number field of degree n and λ1 , . . . , λn ∈ O K are

such that D(λ1 , . . . , λn ) = d(K ), then {λ1 , . . . , λn } is an integral basis for K .
We determine the discriminant of a quadratic field.

141
CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

142 Integral Bases

Theorem 7.1.2 Let K be a quadratic field. Let m be the unique squarefree integer
√
such that K = Q( m). Then the discriminant d(K ) of K is given by

4m, if m ≡ 1 (mod 4),
d(K ) =
m, if m ≡ 1 (mod 4).

Proof: We appeal to Theorem 7.1.1. If m ≡ 1 (mod 4), an integral basis for K is

√
{1, m} so that
√ 2
1 m √
d(K ) = √ = (−2 m)2 = 4m.
1 − m
√
If m ≡ 1 (mod 4), an integral basis for K is 1, 1+2 m so that
√ 2
1+ m
1 √
2√
d(K ) = = (− m)2 = m.
1− m
1
2
√ √ √
Since d(K ) = m or 2 m the next theorem follows immediately from The-
orem 7.1.2.
√
Theorem 7.1.3 Let K be a quadratic field. Then K = Q( d(K )).

We note that the quadratic field K is a real field if and only if d(K ) > 0.
Next we define the norm of an ideal in the ring of integers of an algebraic number
field.

Definition 7.1.3 (Norm of an ideal) Let K be an algebraic number field of degree

n. Let I be a nonzero ideal of O K . Then the norm of the ideal I , written N (I ), is
the positive integer defined by

D(I )
N (I ) = .
d(K )

We now justify that N (I ) is indeed a positive integer.

Theorem 7.1.4 Let K be an algebraic number field of degree n. Let I be a nonzero

ideal of O K . Then the norm N (I ) of the ideal I is a positive integer.

Proof: Let {η1 , . . . , ηn } be a basis for I and let {ω1 , . . . , ωn } be an integral basis
for K . As η1 , . . . , ηn ∈ O K there exist ci j (i, j = 1, . . . , n) ∈ Z such that

n
ηi = ci j ω j , i = 1, . . . , n.
j=1
CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

7.1 Integral Basis of an Algebraic Number Field 143

Hence

D(η1 , . . . , ηn ) = (det(ci j ))2 D(ω1 , . . . , ωn )

so that

D(I ) = det(ci j ))2 d(K ).

Since D(I ) = 0 we have det(ci j ) = 0 so that

D(I )
N (I ) = = |det(ci j )|
d(K )

is a positive integer.

√
Example
√ 7.1.1 Let K = Q( −5).
√ Let I be the ideal of O K generated by 2 and
1 + −5; that is, I = 2, 1 + −5 . We determine √ the norm N (I ) of the ideal I .
First we find a basis for I . As O K = {x + y −5 | x, y ∈ Z} we have
√ √ √
I = {2(a + b −5) + (1 + −5)(c + d −5) | a, b, c, d ∈ Z}
√
= {(2a + c − 5d) + (2b + c + d) −5 | a, b, c, d ∈ Z}
√
= {(2a + (y − 2b − d) − 5d) + y −5 | a, b, d, y ∈ Z}
√
= {2(a − b − 3d) + y + y −5 | a, b, d, y ∈ Z}
√
= {2x + y + y −5 | x, y ∈ Z}
√
= {2x + (1 + −5)y | x, y ∈ Z},

√
so that {2, 1 + −5} is a basis for I . Hence
√ 2
√ 2 1 + √−5 √
D(I ) = D(2, 1 + −5) = = (−4 −5)2 = −80.
2 1 − −5

By Theorem 7.1.2 we have

d(K ) = 4(−5) = −20.

Hence

D(I ) −80 √
N (I ) = = = 4 = 2.
d(K ) −20

√
Example 7.1.2 Let K = Q( m), where m is a squarefree integer with m ≡
√ √
1 (mod 4), so that O K = {x + y m | x, y ∈ Z}. Let α = a + b m ∈ O K . We
CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

144 Integral Bases

determine the norm of the principal ideal α . We have

√
α = a + b m
√ √
= {(x + y m)(a + b m) | x, y ∈ Z}
√ √
= {x(a + b m) + y(bm + a m) | x, y ∈ Z}
√ √
so that {a + b m, bm + a m} is a basis for α . Hence
√ √
a + b m bm + a m 2
D(α ) = √ √
a − b m bm − a m
√ √ √ √
= ((a + b m)(bm − a m) − (a − b m)(bm + a m))2 .

Recalling the identity

(A + B)(C − D) − (A − B)(C + D) = 2(BC − AD),

we see that

D(α ) = 22 m(a 2 − mb2 )2 .

Now d(K ) = 4m (Theorem 7.1.2) so that

22 m(a 2 − mb2 )2
N (α ) = = |a 2 − mb2 |.
4m

We now use Theorem 6.5.5 to determine the norms of a wide class of ideals in a
quadratic field.

Theorem 7.1.5 Let K be a quadratic field. Let m be the unique squarefree integer
√
such that K = Q( m).

(a) m ≡ 1 (mod 4). Let a, b, c ∈ Z be such that

a = 0, c = 0, c | a, c | b, ac | b2 − mc2 .

Then
√
N (a, b + c m ) = |ac|.

(b) m ≡ 1 (mod 4). Let a, b, c ∈ Z be such that

a = 0, c = 0, b ≡ c (mod 2), c | a, c | b, 4ac | b2 − mc2 .

Then
√
b+c m
N (a, ) = |ac|.
2
CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

7.1 Integral Basis of an Algebraic Number Field 145

√ √
Proof: (a) By Theorem 6.5.5(a) {a, b + c m} is a basis for the ideal a, b + c m .
Hence
√ √
D(a, b + c m ) = D(a, b + c m)
√
a b + c m 2
= √
a b−c m
= 4a 2 c2 m.

As m ≡ 1 (mod 4) we have

d(K ) = 4m,

by Theorem 7.1.2. Thus

√
4a 2 c2 m
N (a, b + c m ) = = |ac|.
4m
√
(b) By Theorem 6.5.5(b) we see that a, b+c2 m is a basis for the ideal
√
a, b+c2 m
. Hence
√ √
b+c m b+c m
D a, = D a,
2 2
√ 2
b+c m
a
√
= 2 = a 2 c2 m.
b−c m
a
2
As m ≡ 1 (mod 4) we have

d(K ) = m,

by Theorem 7.1.2. Thus

√ 2 2
b+c m a c m
N a, = = |ac|.
2 m
We remark that Example 7.1.1 is the special case a = 2, b = c = 1, m = −5
of Theorem 7.1.5(a).
The next theorem determines the norm of a principal ideal in the ring of integers
of an arbitrary algebraic number field, which is generated by a rational integer.

Theorem 7.1.6 Let K be an algebraic number field of degree n. Let c be a nonzero

rational integer. Then the norm of the principal ideal c of O K generated by c is

N (c ) = |c|n .
CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

146 Integral Bases

Proof: Let {ω1 , . . . , ωn } be an integral basis for K . Then

α ∈ c ⇐⇒ α = cβ for a unique β ∈ O K
⇐⇒ α = c(x1 ω1 + · · · + xn ωn ) for unique x1 , . . . , xn ∈ Z
⇐⇒ α = x1 (cω1 ) + · · · + xn (cωn ) for unique x1 , . . . , xn ∈ Z.

This shows that {cω1 , . . . , cωn } is a basis for the principal ideal c . Hence

D(c ) = D(cω1 , . . . , cωn ) = c2n D(ω1 , . . . , ωn ) = c2n d(K )

so that

D(c ) √ 2n
N (c ) = = c = |c|n .
d(K )

If K is an algebraic number field of degree n then by Theorem 6.1.2 there exists

θ ∈ O K such that K = Q(θ). By Theorem 6.4.3 we have D(θ) = 0. Let ω1 , . . . , ωn
be an integral basis for K . Then there exist ci j (i, j = 1, . . . , n) ∈ Z such that


 1 = c11 ω1 + · · · + c1n ωn ,

θ = c21 ω1 + · · · + c2n ωn ,

 ···
 n−1
θ = cn1 ω1 + · · · + cnn ωn .

Hence

D(θ ) = D(1, θ, . . . , θ n−1 ) = (det(ci j ))2 D(ω1 , . . . , ωn ) = |det(ci j )|2 d(K ),

showing that

D(θ) = m 2 d(K )

for some positive integer m(= |det(ci j )|). The positive integer m is called the index
of θ.

Definition 7.1.4 (Index of θ) Let K be an algebraic number field. Let θ ∈ O K be

such that K = Q(θ ). Then the index of θ, written ind θ, is the positive integer given
by

D(θ) = (ind θ )2 d(K ).

Theorem 7.1.7 Let K be an algebraic number field of degree n. Let θ ∈ O K be

such that K = Q(θ). Then {1, θ, θ 2 , . . . , θ n−1 } is an integral basis for K if and
only if ind θ = 1.
CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

7.1 Integral Basis of an Algebraic Number Field 147

Proof: We have
{1, θ, θ 2 , . . . , θ n−1 } is an integral basis for K

⇐⇒ D(1, θ, . . . , θ n−1 ) = d(K ) (Theorem 6.5.4)

⇐⇒ D(θ ) = d(K )
⇐⇒ ind θ = 1.

Clearly if D(θ) is squarefree then ind θ = 1 so that, by Theorem 7.1.7,

{1, θ, . . . , θ n−1 } is an integral basis.

Theorem 7.1.8 Let K be an algebraic number field of degree n. Let θ ∈ O K be

such that K = Q(θ). If D(θ) is squarefree then {1, θ, . . . , θ n−1 } is an integral basis
for K .

To apply Theorem 7.1.8 in a particular example we need to calculate D(θ ). The

following result is often useful in this connection.

Theorem 7.1.9 Let θ be an algebraic number of degree n. Let θ1 = θ, θ2 , . . . , θn

be the conjugates of θ over Q, that is, the roots of f (x) = irrQ (θ). Then

n
D(θ ) = (−1)n(n−1)/2 f (θi ).
i=1

Proof: We have

n
f (x) = irrQ (θ) = (x − θi ).
i=1

Differentiating f (x) using the product rule, we obtain

n
n
f (x) = (x − θ j )
i=1 j = 1
j = i

so that

n

f (θi ) = (θi − θ j ), i = 1, 2, . . . , n.
j =1
j = i
CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

148 Integral Bases

Hence

n
n
n

f (θi ) = (θi − θ j )
i=1 i=1 j = 1
j = i

= (θi − θ j ) (θi − θ j )
1≤i< j≤n 1≤ j<i≤n

= (−1)n(n−1)/2 (θi − θ j ) (θ j − θi )
1≤i< j≤n 1≤ j<i≤n

= (−1)n(n−1)/2 (θi − θ j ) (θi − θ j )
1≤i< j≤n 1≤i< j≤n

= (−1)n(n−1)/2 (θi − θ j )2
1≤i< j≤n
= (−1) n(n−1)/2
D(θ),
by Theorem 6.4.1.

Clearly with the notation of Theorem 7.1.9, we have

n
disc(irrQ θ ) = (θi − θ j )2 = (−1)n(n−1)/2 f (θi ) = D(θ),
1≤i< j≤n i=1

in agreement with Theorem 6.4.2.

We apply Theorem 7.1.9 in the case when θ is an algebraic integer of degree 3.

Theorem 7.1.10 Let a, b be integers such that x 3 + ax + b ∈ Z[x] is irreducible.

Let θ ∈ C be a root of x 3 + ax + b so that K = Q(θ) is a cubic field and θ ∈ O K .
Then
D(θ) = −4a 3 − 27b2 .

Proof: Let f (x) = irrQ (θ) = x 3 + ax + b. Let θ1 = θ, θ2 , θ3 be the conjugates of

θ over Q so that
(x − θ1 )(x − θ2 )(x − θ3 ) = x 3 + ax + b.
Equating coefficients we obtain
θ1 + θ2 + θ3 = 0,
θ1 θ2 + θ2 θ3 + θ3 θ1 = a,
θ1 θ2 θ3 = −b.
Now
f (x) = 3x 2 + a
CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

7.1 Integral Basis of an Algebraic Number Field 149

so that

f (θ1 ) f (θ2 ) f (θ3 ) = (3θ12 + a)(3θ22 + a)(3θ32 + a)

= a 3 + 3a 2 (θ12 + θ22 + θ32 ) + 9a(θ12 θ22 + θ22 θ32 + θ32 θ12 )
+ 27θ12 θ22 θ32 .

Next we observe that

θ12 + θ22 + θ32 = (θ1 + θ2 + θ3 )2 − 2(θ1 θ2 + θ2 θ3 + θ3 θ1 ) = −2a,

θ12 θ22 + θ22 θ32 + θ32 θ12 = (θ1 θ2 + θ2 θ3 + θ3 θ1 )2 − 2θ1 θ2 θ3 (θ1 + θ2 + θ3 ) = a 2 ,
θ12 θ22 θ32 = (θ1 θ2 θ3 )2 = b2 ,

f (θ1 ) f (θ2 ) f (θ3 ) = a 3 + 3a 2 (−2a) + 9a(a 2 ) + 27b2 = 4a 3 + 27b2 .

Hence by Theorem 7.1.9 we obtain

f (θ1 ) f (θ2 ) f (θ3 ) = −4a 3 − 27b2 ,

3·2
D(θ) = (−1) 2

as asserted.

In the next example we find an integral basis for a particular cubic field using
Theorems 7.1.8 and 7.1.10.

Example 7.1.3 Let K = Q(θ ), where θ is a root of x 3 + x + 1. The cubic poly-

nomial x 3 + x + 1 is irreducible in Z[x] so that [K : Q] = 3. Also θ ∈ O K . By
Theorem 7.1.10 we have

D(θ) = −4(1)3 − 27(1)2 = −31.

As −31 is squarefree, by Theorem 7.1.8 {1, θ, θ 2 } is an integral basis for K .

More generally we have the following result.

Theorem 7.1.11 Let a, b be integers such that x 3 + ax + b ∈ Z[x] is irreducible

and −4a 3 − 27b2 is squarefree. Let θ be a root of x 3 + ax + b. Then {1, θ, θ 2 } is
an integral basis for the cubic field Q(θ ).

Other values of a and b satisfying the conditions of Theorem 7.1.11 are

(a, b) = (−1, −1), (2, 1), (4, 1), (−1, 3), and (5, 3).

Similarly to Theorem 7.1.10 we can use Theorem 7.1.9 to prove the following
result.
CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

150 Integral Bases

Theorem 7.1.12 Let a, b be integers such that x 4 + ax + b ∈ Z[x] is irreducible.

Let θ be a root of x 4 + ax + b so that K = Q(θ ) is a quartic field and θ ∈ O K .
Then

D(θ) = −27a 4 + 256b3 .

Appealing to Theorems 7.1.8 and 7.1.12 we obtain

Theorem 7.1.13 Let a, b be integers such that x 4 + ax + b ∈ Z[x] is irreducible

and −27a 4 + 256b3 is squarefree. Let θ be a root of x 4 + ax + b. Then {1, θ, θ 2 , θ 3 }
is an integral basis for the quartic field Q(θ ).

The quartic polynomial x 4 + x + 1 is irreducible and has discriminant −27(1)4 +

256(1)3 = 229, which is prime. Hence, by Theorem 7.1.13, the quartic field Q(θ ),
where θ 4 + θ + 1 = 0, has {1, θ, θ 2 , θ 3 } as an integral basis.
If we take K to be the cubic field Q(θ), where θ 3 − 2 = 0, then D(θ) = −108 =
−3 · 62 and Theorem 7.1.11 is not applicable. As D(θ)/d(K ) is a perfect square
(= (ind θ)2 ), we have
D(θ )
= 1, 4, 9, or 36
d(K )
so that

d(K ) = −108, −27, −12, or − 3,

and further information is required to determine which case actually occurs.

In some cases the following result first proved by Ludwig Stickelberger (1850–
1936) in 1897 is useful (see [18]).

Theorem 7.1.14 Let K be an algebraic number field. Then

d(K ) ≡ 0 or 1 (mod 4).

Proof: Let {ω1 , ω2 , . . . , ωn } be an integral basis for K . Let ωi(1) = ωi , ωi(2) , . . . , ωi(n)
be the K -conjugates of ωi (i = 1, 2, . . . , n). In the expansion of the determinant
(1)
ω1 (1) (1)
(2) ω2(2) · · · ωn
ω ω2 · · · ωn(2)
1
. . ..
.. .. ··· .

ω(n) ω(n) · · · ω(n)
1 2 n

there are n! terms, half of which occur with positive signs and half with negative
signs. Let the sum of those with positive signs be λ and those with negative signs
CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

7.1 Integral Basis of an Algebraic Number Field 151

µ so that
( j)
det(ωi ) = λ − µ.

Set

A = λ + µ, B = λµ.

Then
( j)
d(K ) = (det(ωi ))2 = (λ − µ)2 = (λ + µ)2 − 4λµ = A2 − 4B.
( j)
As ωi ∈ O K (i = 1, 2, . . . , n), by Theorem 6.3.3 each ωi (i, j = 1, 2, . . . , n) ∈ .
Hence λ, µ ∈ so

A ∈ , B ∈ .

Let θ ∈ O K be such that K = Q(θ). Let θ1 = θ, θ2 , . . . , θn be the conjugates of θ

over Q. If we express each ω j as a polynomial in θ with rational coefficients, A
becomes a symmetric function of θ1 , . . . , θn with rational coefficients, and so

A ∈ Q.

Hence A ∈ ∩ Q = Z. Then
A2 − d(K )
B= ∈ Q,
4
so that B ∈ ∩ Q = Z. Finally, as A, B ∈ Z, we have

d(K ) = A2 − 4B ≡ 0 or 1 (mod 4).

The next example illustrates the use of Theorem 7.1.14 to determine the discrim-
inant of an algebraic number field K = Q(θ) when D(θ ) is not squarefree.

Example 7.1.4 The cubic polynomial x 3 − x − 2 ∈ Z[x] is irreducible. Let θ be a

root of x 3 − x − 2 and set K = Q(θ ) so that [K : Q] = 3. By Theorem 7.1.10 we
have

D(θ ) = −4(−1)3 − 27(−2)2 = −104 = −26 · 22 .

Since D(θ )/d(K ) must be a square in Z, we have

D(θ )
= 1 or 4
d(K )
so that

d(K ) = −104 or −26.

CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

152 Integral Bases

But by Theorem 7.1.14, d(K ) ≡ 0 or 1 (mod 4) so that d(K ) = −26. Hence

d(K ) = −104, and {1, θ, θ 2 } is an integral basis for K .

The following result is a straightforward generalization of Example 7.1.4.

Theorem 7.1.15 Let a, b be integers such that x 3 + ax + b ∈ Z is irreducible and

−4a 3 − 27b2 = 4m,

where m is a squarefree integer ≡ 2 or 3 (mod 4). Let θ be a root of x 3 + ax + b.

Then {1, θ, θ 2 } is an integral basis for the cubic field Q(θ).

In the next example we give a cubic field Q(θ ) for which {1, θ, θ 2 } is not an
integral basis.

Example 7.1.5 The cubic polynomial x 3 + 11x + 4 ∈ Z[x] is irreducible. Let θ

be a root of x 3 + 11x + 4. Set K = Q(θ ) so that K is a cubic field. By Theorem
7.1.10 we have

D(θ) = −4(11)3 − 27(4)2 = −5756 = −1439 · 22 ,

where 1439 is prime. As D(θ )/d(K ) is the square of an integer, we have

D(θ)
= 1 or 4
d(K )
so that

d(K ) = −4 · 1439 or −1439.

The first of these is ≡ 0 (mod 4) and the second is ≡ 1 (mod 4), so we cannot
use Theorem 7.1.14 to distinguish between them. We recall from Example 6.3.2 that
(θ + θ 2 )/2 is an integer of K as it is a root of the polynomial x 3 + 11x 2 + 36x + 4 ∈
Z[x]. Hence {1, θ, θ 2 } is not an integral basis for K . Thus

d(K ) = D(1, θ, θ 2 ) = D(θ) = −1439 · 22 ,

so d(K ) = −1439. Since

1 0 0 2
θ + θ2
D 1, θ, = 0 1 0 D(1, θ, θ 2 )
2 0 1 1
2 2
1 1
= D(θ ) = (−5756) = −1439,
4 4

1, θ, (θ + θ 2 )/2 is an integral basis for Q(θ).
CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

7.1 Integral Basis of an Algebraic Number Field 153

In the next example Theorem 7.1.14 is not sufficient to distinguish between the
possible values of the discriminant and we have to carry out a more detailed analysis.
√ √
Example 7.1.6 Let θ = 3 2 and set K = Q(θ) = Q( 3 2). Since θ is a root of the
irreducible polynomial x 3 − 2 ∈ Z[x], we have irrQ (θ) = x 3 − 2 and [K : Q] =
deg(irrQ (θ )) = 3. By Theorem 7.1.10 we have

D(θ) = −4(0)3 − 27(−2)2 = −108 = −22 · 33 .

As D(θ )/d(K ) is a perfect square, we must have

D(θ)
= 12 , 22 , 32 , or 62
d(K )
so that

d(K ) = −108, −27, −12, or − 3.

Each of these possibilities is congruent to 0 or 1 modulo 4, so we cannot use

Theorem 7.1.14 to distinguish among them. We proceed instead by showing that if
x1 + x2 θ + x3 θ 2 ∈ O K , where x1 , x2 , x3 ∈ Q, then x1 , x2 , x3 ∈ Z so that {1, θ, θ 2 }
is an integral basis for K and d(K ) = −108. Clearly Z + Zθ + Zθ 2 ⊆ O K , so we
wish to show that O K ⊆ Z + Zθ + Zθ 2 .
Let α ∈ O K . Then α ∈ K and thus there exist x1 , x2 , x3 ∈ Q such that

α = x1 + x2 θ + x3 θ 2 .

The K -conjugates of α are


 α = x1 + x2 θ + x3 θ 2 ,
α = x1 + x2 ωθ + x3 ω2 θ 2 ,

α = x1 + x2 ω2 θ + x3 ωθ 2 ,

where ω is a complex cube root of unity. Hence, as 1 + ω + ω2 = 0, we have


 α + α + α = 3x1 ,
θ 2 (α + ω2 α + ωα ) = 6x2 ,

θ(α + ωα + ω2 α ) = 6x3 .

As α ∈ O K , by Theorem6.3.3
√
we have α, α , α ∈ ; as θ ∈ O K we have θ, θ 2 ∈ ;
and as ω, ω2 ∈ Z + Z 1+ 2 −3 = OQ(√−3) we have ω, ω2 ∈ . Thus

α + α + α , θ 2 (α + ω2 α + ωα ), θ(α + ωα + ω2 α ) ∈

so that

3x1 ∈ ∩ Q = Z, 6x2 ∈ ∩ Q = Z, 6x3 ∈ ∩ Q = Z.

CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

154 Integral Bases

Set

yi = 6xi ∈ Z, i = 1, 2, 3,

so that

6α = y1 + y2 θ + y3 θ 2 . (7.1.1)

Before proceeding we note the following simple result. Suppose θ | n in O K ,

where n ∈ Z. Then n = θw for w ∈ O K . Thus n 3 = θ 3 w 3 = 2w 3 . Now w3 =
n 3 /2 ∈ Q and w 3 ∈ O K ⊆ so that w 3 ∈ Q ∩ = Z. Thus 2 | n 3 in Z. But 2
is a prime, so 2 | n in Z. We have shown that

θ | n in O K =⇒ 2 | n in Z.

We also note that θ, θ 2 , and θ 3 divide 2 in O K as 2 = θ 3 . Using these results in

yi = 2z i , i = 1, 2, 3,

so that

3α = z 1 + z 2 θ + z 3 θ 2 , z 1 , z 2 , z 3 ∈ Z.

If z 2 = z 3 = 0 then 3α = z 1 so that α = z 1 /3 ∈ Q. But α ∈ O K ⊆ so that α ∈

Q ∩ = Z. Hence α ∈ Z + Zθ + Zθ 2 as required. If (z 2 , z 3 ) = (0, 0) then α =
1
(z + z 2 θ + z 3 θ 2 ) ∈
3 1
/ Q since deg(irrQ (θ)) = 3. Hence
1
Q(α) = Q( (z 1 + z 2 θ + z 3 θ 2 )) = Q.
3
Now α ∈ Q(θ ) so Q(α) ⊆ Q(θ) and thus [Q(α) : Q] | [Q(θ ) : Q] = 3; that is,
[Q(α) : Q] = 1 or 3. But Q(α) = Q so [Q(α) : Q] = 1. Hence [Q(α) : Q] = 3.
Thus the minimal polynomial of α over Q is of degree 3. Now α is a root of

x 3 + c1 x 2 + c2 x + c3 ∈ Q[x],

where

c1 = −(α + α + α ) = −z 1 ,
1
c2 = αα + α α + α α = (z 12 − 2z 2 z 3 ),
3
−1 3
c3 = −αα α = (z + 2z 23 + 4z 33 − 6z 1 z 2 z 3 ).
27 1
Hence

irrQ (α) = x 3 + c1 x 2 + c2 x + c3 .
CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

7.1 Integral Basis of an Algebraic Number Field 155

Since α ∈ we must have x 3 + c1 x 2 + c2 x + c3 ∈ Z[x] so that c1 , c2 , c3 ∈ Z; that

is,

z 12 − 2z 2 z 3 ≡ 0 (mod 3), (7.1.2)

z 13 + 2z 23 + 4z 33 − 6z 1 z 2 z 3 ≡ 0 (mod 27). (7.1.3)

Suppose that at least one of z 1 , z 2 , z 3 is not divisible by 3. Then (7.1.2) and (7.1.3)
show that 3 does not divide any of z 1 , z 2 , z 3 . From (7.1.3) we have

z 13 + 2z 23 + 4z 33 ≡ 0 (mod 3).

As z 3 ≡ z (mod 3) for any integer z, we have

z 1 + 2z 2 + z 3 ≡ 0 (mod 3).

Thus, as 3 z 1 , z 2 , z 3 , we must have

(z 1 , z 2 , z 3 ) ≡ (1, 2, 1) or (2, 1, 2) (mod 3)

so that

z 2 ≡ 2z 1 (mod 3), z 3 ≡ z 1 (mod 3).

Define integers t and u by

z 2 = 2z 1 + 3t, z 3 = z 1 + 3u.

Then

z 13 + 2z 23 + 4z 33 − 6z 1 z 2 z 3
= z 13 + 2(2z 1 + 3t)3 + 4(z 1 + 3u)3 − 6z 1 (2z 1 + 3t)(z 1 + 3u)
= 9z 13 + 54(t z 12 + 2t 2 z 1 + t 3 + 2u 2 z 1 + 2u 3 − tuz 1 )
≡ 9z 13 (mod 27)
≡ 0 (mod 27),

as 3 z 1 , contradicting (7.1.3). Hence z 1 ≡ z 2 ≡ z 3 ≡ 0 (mod 3). Thus we can

define integers w1 , w2 , w3 by

z i = 3wi , i = 1, 2, 3.

Then

α = w1 + w2 θ + w3 θ 2 ∈ Z + Zθ + Zθ 2 ,

proving O K ⊆ Z + Zθ + Zθ 2 as required.
CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

156 Integral Bases

√
We have shown that {1, θ, θ 2 } is an integral basis for K = Q(θ) = Q( 3 2), so
that
√ √
2 2
1 3
2 ( 3
2)
√ √ √
d(Q( 2)) = 1 ω√ 2 ω2 (√3 2)2 = −108.
3 3

1 ω2 3 2 ω( 3 2)2

Richard Dedekind (1831–1916) [5] determined an integral basis for the cubic
√
field Q( 3 m) (with m a cubefree integer) in 1900 (see Theorem 7.3.2).
We conclude
√ this
√ section with the determination of an integral basis for the quartic
field Q( −1 + 2).

√ √ √ √
Example 7.1.7 Let K be the quartic field Q( −1 + 2) = Q( −1, 2). We
show that

√ √ 1 √ √
O K = Z + Z −1 + Z 2 + Z ( 2 + −2)
2

and

d(K ) = 256.

It is easy to check that

√ √ 1 √ √
Z + Z −1 + Z 2 + Z ( 2 + −2) ⊆ O K ,
2

so we have to prove that

√ √ 1 √ √
O K ⊆ Z + Z −1 + Z 2 + Z ( 2 + −2) .
2
√ √ √
Let θ ∈ O K . The subfields of K are Q, Q( −1), Q( 2),and Q( −2). If
√ √ √ √ √
θ ∈ Q then θ ∈ Z ⊂ Z + Z −1 + Z 2 + Z 12 ( 2 + −2) . If θ ∈ Q( −1)
√ √ √ √ √
then θ ∈ OQ(√−1) = Z + Z −1 ⊂ Z + Z −1 + Z 2 + Z 12 ( 2 + −2) .
√ √ √ √
If θ ∈ Q( 2) then θ ∈ OQ(√2) = Z + Z 2 ⊂ Z + Z −1 + Z 2 +
√ √ √ √
Z 12 ( 2 + −2) . If θ ∈ Q( −2) then θ ∈ OQ(√−2) = Z + Z −2 ⊂
√ √ √ √
Z + Z −1 + Z 2 + Z 12 ( 2 + −2) . Hence we may suppose that θ
does not belong to any of the subfields of K . As θ ∈ K we have
√ √ √
θ = a0 + a1 −1 + a2 2 + a3 −2, a0 , a1 , a2 , a3 ∈ Q.
CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

7.1 Integral Basis of an Algebraic Number Field 157

The conjugates of θ over Q are

√ √ √
θ
= a0 + a1 −1 + a2 2 + a3 −2,
√ √ √
θ
= a0 − a1 −1 + a2 2 − a3 −2,
√ √ √
θ
= a0 + a1 −1 − a2 2 − a3 −2,
√ √ √
θ
= a0 − a1 −1 − a2 2 + a3 −2.
√ √
Then θ + θ = 2a0 + 2a2 2,
√ θ + θ√ = 2a0 +√2a1 −1, √ θ + θ =

2a0 + 2a3 −2 must be integers of Q( 2), Q( −1), Q( −2) respec-

tively. Hence 2a0 , 2a1 , 2a2 , 2a3 ∈ Z. Define integers bi by bi = 2ai (i = 0, 1, 2, 3)
so that
1 √ √ √
θ = (b0 + b1 −1 + b2 2 + b3 −2).
2
Set

c = b02 + 2b32 ∈ Z,
d = b02 + b12 − 2b22 − 2b32 ∈ Z,
e = b0 b3 − b1 b2 ∈ Z,

so that θ is a root of

d b0 d d 2 + 8e2
f (x) = x − 2b0 x + c +
4 3
x + −2b3 e −
2
x+ ∈ Q[x].
2 2 16
As θ is of degree 4 over Q (since it does not belong to any of the subfields of K ) the
polynomial f (x) must be the minimal polynomial of θ over Q. Hence, as θ ∈ O K ,
we have f (x) ∈ Z[x], and so d/2 ∈ Z and (d 2 + 8e2 )/16 ∈ Z. Hence

d ≡ 0 (mod 2), d 2 + 8e2 ≡ 0 (mod 16).

From these congruences we deduce that

d ≡ 0 (mod 4), e ≡ 0 (mod 2).

Hence

b02 + b12 − 2b22 − 2b32 ≡ 0 (mod 4) (7.1.4)

and

b0 b3 − b1 b2 ≡ 0 (mod 2). (7.1.5)

If b0 or b1 is odd from (7.1.4) we see that the other is odd as well. Then
from (7.1.5) we deduce that b2 ≡ b3 (mod 2), and (7.1.4) gives the contradiction
2 ≡ 0 (mod 4). Thus

b0 ≡ b1 ≡ 0 (mod 2) and b2 ≡ b3 (mod 2).

CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

158 Integral Bases

Hence we can define integers c0 , c1 , c2 , c3 by

b0 = 2c0 , b1 = 2c1 , b2 = 2c2 + c3 , b3 = c3 .

Then
1 √ √ √
θ = (b0 + b1 −1 + b2 2 + b3 −2)
2
√ √ 1 √ √
= c0 + c1 −1 + c2 2 + c3 ( 2 + −2)
2

√ √ 1 √ √
∈ Z + Z −1 + Z 2 + Z ( 2 + −2)
2
as required.√ √ √ √
Thus {1, −1, 2, 12 ( 2 + −2)} is an integral basis for K and
2
√ √ 1 √ √
1 −1 2 ( 2 + −2)
2

√ √ 1 √ √
1 − −1 2 ( 2 − −2)

d(K ) = 2
√ √
= 256.

1 √ √ 1
−1 − 2 (− 2 − −2)
2
√ √ √ √
1
1 − −1 − 2 (− 2 + −2)
2

√ √ √ √
An integral basis for the quartic field Q( m + n) = Q( m, n), where m
and n are distinct squarefree integers, was determined by K. S. Williams [19] in
1970.

Definition 7.1.5 (Monogenic number field) Let K be an algebraic number field of

degree n. If there exists an element θ ∈ O K such that {1, θ, . . . , θ n−1 } is an integral
basis for K then K is said to be monogenic and the integral basis {1, θ, . . . , θ n−1 }
is called a power basis for K .

Clearly every quadratic field is monogenic. The cubic fields in Examples 7.1.4
and 7.1.6 are monogenic. Dedekind showed in 1878 that not every algebraic number
field is monogenic by proving that the cubic field
K = Q(θ), θ 3 − θ 2 − 2θ − 8 = 0,

is not monogenic (see [4]).

√ √
Example 7.1.8 We show that the quartic field K = Q( −1, 2) considered in
Example 7.1.7 is monogenic. Let
√ √
2+i 2
θ= ,
2
CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

7.1 Integral Basis of an Algebraic Number Field 159

so that
1 √ √
θ 2 = i, θ 3 = (− 2 + i 2), θ 4 = −1.
2
Then, by Example 7.1.7, we have

√ 1 √ √
O K = Z + Zi + Z 2 + Z ( 2 + i 2)
2
= Z + Zθ 2 + Z(θ − θ 3 ) + Zθ
= Z + Zθ + Zθ 2 + Zθ 3 ,
so that K is monogenic with power basis {1, θ, θ 2 , θ 3 }.

We conclude this section with a simple upper bound for the absolute value of the
discriminant of an algebraic number field as well as a theorem giving the sign of
the discriminant.

Theorem 7.1.16 Let K be an algebraic number field of degree n. Let λ1 , . . . , λn ∈

O K be such that D(λ1 , . . . , λn ) = 0. Then
|d(K )| ≤ |D(λ1 , . . . , λn )|.
Moreover, if D(λ1 , . . . , λn ) is squarefree then {λ1 , . . . , λn } is an integral basis for
OK .

Proof: Let {η1 , . . . , ηn } be an integral basis for K . Then there exist ci j (i, j =
1, 2, . . . , n) ∈ Z such that


 λ1 = c11 η1 + · · · + c1n ηn ,

λ2 = c21 η1 + · · · + c2n ηn ,

 ···

λn = cn1 η1 + · · · + cnn ηn .
Hence
D(λ1 , . . . , λn ) = (det ci j )2 D(η1 , . . . , ηn ) = (det ci j )2 d(K ).
As D(λ1 , . . . , λn ) = 0 we see that det (ci j ) = 0. Thus, as det (ci j ) ∈ Z, we have
(det ci j )2 ≥ 1 and so
|D(λ1 , . . . , λn )| ≥ |d(K )|.
If D(λ1 , . . . , λn ) is squarefree then from
D(λ1 , . . . , λn ) = (det ci j )2 d(K ),
we deduce that det ci j = ±1. Hence D(λ1 , . . . , λn ) = d(K ), proving that
{λ1 , . . . , λn } is an integral basis for K .

The next theorem is due to Alexander Brill (1842–1935) (see [3]).

CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

160 Integral Bases

Theorem 7.1.17 Let K be an algebraic number field of degree n. Let θ ∈ O K be

such that K = Q(θ). Let θ1 = θ, θ2 , . . . , θn be the conjugates of θ. Let r be the
number of θ1 , . . . , θn that are real. Then

sgn (d(K )) = (−1)(n−r )/2 .

Proof: As r is the number of θ1 , . . . , θn that are real, the number of θ1 , . . . , θn that

are nonreal is n − r . Since the nonreal conjugates occur in complex conjugate pairs,
n − r is even, say n − r = 2s, so n = r + 2s. Now let {ω1 , . . . , ωn } be an integral
( j)
basis for K . Let ωk ( j = 1, 2, . . . , n) be the conjugates of ωk (k = 1, 2, . . . , n).
( j) ( j)
Then d(K ) = det (ωk )2 . Set det (ωk ) = A + i B with A, B ∈ R. Since the change
of i into −i in this determinant is equivalent to the interchange of s pairs of rows,
( j)
we have A − i B = (−1)s det (ωk ). Hence A − i B = (−1)s (A + i B). If s is even
( j)
then A − i B = A + i B so B = 0 and d(K ) = det (ωk )2 = A2 is positive. If s is
( j)
odd then A − i B = −(A + i B) so A = 0 and d(K ) = det (ωk )2 = (i B)2 = −B 2
is negative. Hence sgn (d(K )) = (−1)s = (−1)(n−r )/2 .

7.2 Minimal Integers

Let K be an algebraic number field of degree n. Let θ ∈ O K be such that K = Q(θ ).
Then every α ∈ O K can be expressed in the form

α = a0 + a1 θ + · · · + an−1 θ n−1 , (7.2.1)

where a0 , a1 , . . . , an−1 are rational numbers uniquely determined by α and θ. If

k ∈ {1, 2, . . . , n − 1} is such that

ak = 0, ak+1 = · · · = an−1 = 0

so that

α = a0 + a1 θ + · · · + ak θ k

then α is called an integer of degree k in θ . If a1 = a2 = · · · = an−1 = 0 so that

α = a0 then α is called an integer of degree 0 in θ. The integers of degree 0 in θ
are precisely the rational integers.
We are going to show that the denominators of the a j are bounded. Of course if
{1, θ, . . . , θ n−1 } is a power basis for K then all the denominators are equal to 1.
First we prove the following result.

Theorem 7.2.1 Let K be an algebraic number field of degree n. Let ω1 , . . . , ωn ∈

O K be such that

D(ω1 , . . . , ωn ) = 0.
CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

7.2 Minimal Integers 161

Then for each α ∈ O K there exist unique rational integers x1 , . . . , xn such that

n
xj
α= ωj
j=1
D(ω1 , . . . , ωn )

and

D(ω1 , . . . , ωn ) | x 2j , j = 1, 2, . . . , n.

Proof: As ω1 , . . . , ωn ∈ O K , by Theorem 6.4.4(b) D(ω1 , . . . , ωn ) is a rational in-

teger, which is nonzero by assumption. Further, as D(ω1 , . . . , ωn ) = 0, by Theorem
6.4.4(c) ω1 , . . . , ωn are linearly independent over Q and thus form a basis for K
over Q. Hence there exist unique rational numbers y1 , . . . , yn such that

n
α= yjωj. (7.2.2)
j=1

Let σ1 (= 1), σ2 , . . . , σn be the n monomorphisms: K −→ C. Applying these to

(7.2.2), we obtain

n
σk (α) = y j σk (ω j ), k = 1, 2, . . . , n. (7.2.3)
j=1

Regarding (7.2.3) as a system of n linear equations in the n unknowns y1 , . . . , yn ,

we obtain by Cramer’s rule

σ1 (ω1 ) · · · σ1 (ω j−1 ) σ1 (α) σ1 (ω j+1 ) · · · σ1 (ωn )

··· ···

σn (ω1 ) · · · σn (ω j−1 ) σn (α) σn (ω j+1 ) · · · σn (ωn )
yj =
det(σi (ω j ))
for j = 1, 2, . . . , n. Hence
2
σ1 (ω1 ) · · · σ1 (ω j−1 ) σ1 (α) σ1 (ω j+1 ) · · · σ1 (ωn )

y j D(ω1 , . . . , ωn ) =
2
··· ···

σn (ω1 ) · · · σn (ω j−1 ) σn (α) σn (ω j+1 ) · · · σn (ωn )

is an algebraic integer for j = 1, 2, . . . , n and since y 2j D(ω1 , . . . , ωn ) ∈ Q we

deduce that y 2j D(ω1 , . . . , ωn ) ∈ Z. Set y j = r j /s j , where r j ∈ Z, s j ∈ N, and
(r j , s j ) = 1. Then
r 2j
D(ω1 , . . . , ωn ) ∈ Z, j = 1, 2, . . . , n.
s 2j
As (r j , s j ) = 1 we deduce that

s 2j | D(ω1 , . . . , ωn ), j = 1, 2, . . . , n.
CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

162 Integral Bases

Let
rj
x j = y j D(ω1 , . . . , ωn ) = D(ω1 , . . . , ωn ) ∈ Z, j = 1, 2, . . . , n.
sj
Then, from (7.2.2), we obtain

n
xj
α= ωj.
j=1
D(ω1 , . . . , ωn )

Finally, we observe that

x 2j D(ω1 , . . . , ωn )
= r 2j ∈ Z,
D(ω1 , . . . , ωn ) s 2j
so that
D(ω1 , . . . , ωn ) | x 2j , j = 1, 2, . . . , n.
This completes the proof of the theorem.

We note that if D(ω1 , . . . , ωn ) is squarefree then by Theorem 7.2.1 we have

D(ω1 , . . . , ωn ) | x j , j = 1, 2, . . . , n.
Hence, by Theorem 7.2.1, for each α ∈ O K there exist unique rational integers
a j = x j /D(ω1 , . . . , ωn ) such that

n
α= ajωj,
j=1

proving that {ω1 , . . . , ωn } is an integral basis for K , a result that we have seen
before in Theorem 7.1.16.
We now use Theorem 7.2.1 to bound the denominators of the a j in (7.2.1).

Theorem 7.2.2 Let K be an algebraic number field of degree n. Let θ ∈ O K be such

that K = Q(θ). Let α ∈ O K . Then there exist unique rational numbers r j /s j ( j =
1, 2, . . . , n) with (r j , s j ) = 1 and s j > 0 such that

n
rj
α= θ j−1
j=1
sj

and
1 ≤ s j ≤ |D(θ )|, s 2j | D(θ).

Proof: As θ ∈ O K we have
1, θ, θ 2 , . . . , θ n−1 ∈ O K ,
CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

7.2 Minimal Integers 163

so by Theorem 6.4.4(b)

D(θ) = D(1, θ, θ 2 , . . . , θ n−1 ) ∈ Z.

Further, as K = Q(θ ), by Theorem 6.4.3 we have

D(θ) = 0.

Then by Theorem 7.2.1 there exist unique rational integers x1 , . . . , xn such that
n
x j j−1
α= θ
j=1
D(θ)

and

D(θ) | x 2j , j = 1, 2, . . . , n.

For j = 1, 2, . . . , n we define coprime integers r j and s j (> 0) by

sgn (D(θ ))x j |D(θ)|
rj = , sj = ,
(x j , D(θ)) (x j , D(θ))
so that
rj xj rj n
= , α= θ j−1 ,
sj D(θ) s
j=1 j

and

1 ≤ s j ≤ |D(θ )|.

Finally, for j = 1, 2, . . . , n we have

r 2j x 2j
2
D(θ) = ∈ Z,
sj D(θ)
so that as (r j , s j ) = 1

s 2j | D(θ).

Theorem 7.2.2 enables us to define the concept of a “minimal integer of degree

k in θ .” Let K be an algebraic number field of degree n. Fix θ ∈ O K such that
K = Q(θ). For k ∈ {0, 1, 2, . . . , n − 1} define the set Sk by

Sk = {ak ∈ Q | a0 + a1 θ + · · · + ak θ k ∈ O K
for some a0 , a1 , . . . , ak−1 ∈ Q}. (7.2.4)

Clearly

S0 = Z
CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

164 Integral Bases

and
Sk ⊇ Z, k = 1, 2, . . . , n − 1.

By Theorem 7.2.2 the denominators of the rational numbers in Sk are bounded.

Hence Sk has a least positive element ak∗ . Clearly a0∗ = 1.

Definition 7.2.1 (Minimal integer of degree k in θ) With the preceding notation

any integer of K that is of the form
a0 + a1 θ + · · · + ak−1 θ k−1 + ak∗ θ k ,

where a0 , a1 , . . . , ak−1 ∈ Q, is called a minimal integer of degree k in θ.

The next theorem gives the structure of the set Sk .

Theorem 7.2.3 With the preceding notation

Sk = ak∗ Z.

Proof: Let a ∈ Sk . Let m be the least positive integer such that

ma ∈ Z, mak∗ ∈ N.

By the division algorithm there exist q ∈ Z and r ∈ Z such that

ma = qmak∗ + r, 0 ≤ r < mak∗ .

Hence
r r
a = qak∗ + , 0≤ < ak∗ .
m m
As a ∈ Sk there exist b0 , b1 , . . . , bk−1 ∈ Q such that
b0 + b1 θ + · · · + bk−1 θ k−1 + aθ k ∈ O K .

Similarly, as ak∗ ∈ Sk , there exist c0 , c1 , . . . , ck−1 ∈ Q such that

c0 + c1 θ + · · · + ck−1 θ k−1 + ak∗ θ k ∈ O K .

Then
r k
(b0 − qc0 ) + (b1 − qc1 )θ + · · · + (bk−1 − qck−1 )θ k−1 + θ ∈ OK ,
m
so that
r
∈ Sk .
m
If 0 < r/m < ak∗ this contradicts the minimality of ak∗ . Hence r/m = 0 and a =
qak∗ , proving Sk = ak∗ Z.

The next result gives the form of ak∗ .

CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

7.2 Minimal Integers 165

Theorem 7.2.4 For k = 0, 1, 2, . . . , n − 1

1
ak∗ =
dk
for some dk ∈ N.

Proof: We prove the assertion by induction on k ∈ {0, 1, 2, . . . , n − 1}. The result

is true for k = 0 as
1
a0∗ = 1 =
d0
with d0 = 1. Assume now that
1
ak∗ = , dk ∈ N, k = 0, 1, . . . , l − 1,
dk
∗
where 1 ≤ l ≤ n − 1. By the definition of al−1 and the inductive hypothesis there
exist rational numbers a0 , a1 , . . . , al−2 such that
1
a0 + a1 θ + · · · + al−2 θ l−2 + θ l−1 ∈ O K .
dl−1
Then
1
a0 θ + a1 θ 2 + · · · + al−2 θ l−1 + θ l ∈ OK .
dl−1
Hence
1
∈ Sl .
dl−1
Thus, by Theorem 7.2.3, there exists m ∈ Z such that
1
= al∗ m.
dl−1
This proves that
1
al∗ =
dl
with dl = mdl−1 . This completes the inductive step and the result follows by the
principle of mathematical induction.

The next theorem shows that each dk−1 (k = 1, 2, . . . , n − 1) divides its succes-
sor dk .

Theorem 7.2.5 For k = 1, 2, . . . , n − 1

dk−1 | dk .
CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

166 Integral Bases

Proof: Let k ∈ {1, 2, . . . , n − 1}. Exactly as in the proof of Theorem 7.2.4 we

deduce that
∗
ak−1 ∈ Sk = ak∗ Z

so that
∗
ak−1 = mak∗

for some m ∈ Z. Hence, by Theorem 7.2.4, we have

1 1
=m
dk−1 dk
so that

dk = mdk−1 ,

proving

dk−1 | dk , k = 1, 2, . . . , n − 1.

The next theorem gives the form of an integer of degree k (k = 0, 1, 2, . . . ,

n − 1) in θ. As an immediate consequence we obtain the form of a minimal integer
of degree k in θ .

Theorem 7.2.6 If α is an integer of degree k in θ then there exist

a0 , a1 , . . . , ak ∈ Z such that
a0 + a1 θ + · · · + ak−1 θ k−1 + ak θ k
α= .
dk
In particular if α is a minimal integer of degree k in θ then there exist
a0 , a1 , . . . , ak−1 ∈ Z such that

a0 + a1 θ + · · · + ak−1 θ k−1 + θ k
α= .
dk

Proof: We prove the assertion by induction on k ∈ {0, 1, 2, . . . , n − 1}. Let α be

an integer of degree 0 in θ. Then α = a0 for some a0 ∈ Z. But d0 = 1 so that
α = a0 /d0 is of the asserted form and the result is true for k = 0.
Assume now that all integers of degree up to l − 1 in θ are of the specified form,
where l ∈ {1, 2, . . . , n − 1}. Let α be any integer of degree l in θ . By Theorems
7.2.3 and 7.2.4 there exist r0 , r1 , . . . , rl−1 ∈ Q and al ∈ Z such that
al l
α = r0 + r1 θ + · · · + rl−1 θ l−1 + θ.
dl
CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

7.2 Minimal Integers 167

Let β be a minimal integer in θ of degree l − 1. By the minimality of β and the

inductive hypothesis, there exist s0 , s1 , . . . , sl−2 ∈ Z such that
s0 + s1 θ + · · · + sl−2 θ l−2 + θ l−1
β= .
dl−1
By Theorem 7.2.5 we have dl−1 | dl so that
dl
α − al θβ ∈ O K .
dl−1
Thus

dl dl r j − al s j−1
l−1
r0 + θj
dl−1 j=1
dl−1

is an integer of degree l − 1 in θ . Hence, by the inductive hypothesis, there exist

c0 , c1 , . . . , cl−1 ∈ Z such that

dl dl r j − al s j−1
l−1 cj l−1
r0 + θj = θ j.
dl−1 j=1
dl−1 j=0
dl−1

Equating coefficients we obtain

c0
r0 = ,
dl
al s j−1 + c j
rj = , j = 1, 2, . . . , l − 1.
dl
Define integers a0 , a1 , . . . , al−1 by

a0 = c0 , a j = al s j−1 + c j , j = 1, 2, . . . , l − 1.

Then
aj
rj = , j = 0, 1, . . . , l − 1,
dl
and
a0 + a1 θ + · · · + al−1 θ l−1 + al θ l
α= .
dl
This completes the inductive step and the theorem follows by the principle of
mathematical induction and (for the second part) Theorem 7.2.4.

We now come to the main theorem of this section. We show that if αk (k =

0, 1, . . . , n − 1) is a minimal integer in θ of degree k then {α0 , α1 , . . . , αn−1 } is an
integral basis for K = Q(θ ).
CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

168 Integral Bases

Theorem 7.2.7 Let K be an algebraic number field of degree n. Let θ ∈ O K be

such that K = Q(θ). For k = 0, 1, 2, . . . , n − 1 let αk be a minimal integer in θ of
degree k. Then {α0 , α1 , . . . , αn−1 } is an integral basis for K .

Proof: In any integral basis for K = Q(θ) at least one of the basis elements must
be of the form a0 + a1 θ + · · · + an−1 θ n−1 (a0 , a1 , . . . , an−1 ∈ Q) with an−1 = 0;
otherwise the integral basis could not represent θ n−1 . Replacing the basis element
by its negative, if necessary, we may suppose that an−1 > 0. We choose an integral
basis {ω1 , . . . , ωn } for K with

ωn = a0 + a1 θ + · · · + an−1 θ n−1 , an−1 > 0, an−1 least.

Let k ∈ {1, 2, . . . , n − 1} and suppose that

ωk = b0 + b1 θ + · · · + bn−1 θ n−1 (b0 , . . . , bn−1 ∈ Q).

Replacing ωk by −ωk if necessary we may suppose that bn−1 ≥ 0. Let m be the

unique nonnegative integer such that
bn−1 bn−1
−1<m ≤ .
an−1 an−1
Then

0 ≤ bn−1 − man−1 < an−1 .

If bn−1 − man−1 = 0 we set

ωk = ωk − mωn−1 .

Then {ω1 , . . . , ωk−1 , ωk , ωk+1 , . . . , ωn } is an integral basis for K . This contradicts
the minimality of an−1 as the coefficient of θ n−1 in ωk is bn−1 − man−1 , which
is positive and strictly less than an−1 . Hence bn−1 − man−1 = 0, so that bn−1 is a
rational integral multiple of an−1 . Thus there exist rational integers m 1 , . . . , m n−2
such that ω1 = ω1 − m 1 ωn , ω2 = ω1 − m 2 ωn , . . . , ωn−1
= ωn−1 − m n−1 ωn are

integers of degrees at most n − 2 in θ. Moreover, {ω1 , . . . , ωn−1 , ωn } is an integral
basis for K . Among all integral bases {ω1 , . . . , ωn } for which ω1 , . . . , ωn−1 are
integers of degree at most n − 2 in θ, we choose one for which the coefficient of
θ n−2 is positive and minimal, and we continue our construction until we arrive at
an integral basis α0 , α1 , α2 , . . . , αn−1 , where each αi is of degree i in θ. Let

i
αi = aik θ k , ai k ∈ Q.
k=0

Then

d(K ) = D(α1 , . . . , αn ) = (a00 a11 · · · an−1 n−1 )2 D(θ).

CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

7.2 Minimal Integers 169

For i = 0, 1, . . . , n − 1 let βi be a minimal integer of degree i. Then

D(β0 , . . . , βn−1 ) = (a0∗ a1∗ · · · an−1
∗
)2 D(θ).
By Theorem 7.1.16 we have
|d(K )| ≤ |D(β0 , . . . , βn−1 )|
so that
a00 a11 · · · an−1 n−1 ≤ a0∗ a1∗ · · · an−1
∗

and thus
a00 a11 an−1 n−1
∗ · ∗ ··· ∗ ≤ 1.
a0 a1 an−1
As aii ∈ Si (i = 0, 1, . . . , n − 1), by Theorem 7.2.3 each aii /ai∗ (i =
0, 1, . . . , n − 1) is a positive integer and so
aii = ai∗ , i = 0, 1, . . . , n − 1.
Thus each αi (i = 0, 1, . . . , n − 1) is a minimal integer of degree i in θ .

Theorem 7.2.7 gives a method of finding an integral basis for an algebraic number
field of degree n. We have only to find a minimal integer of each degree up to n − 1.
This is illustrated for some cubic fields in the next section.
Our final theorem of this section gives some further useful information about the
denominators of minimal integers.

Theorem 7.2.8 Let K be an algebraic number field of degree n. Let θ ∈ O K be

such that K = Q(θ). For k = 0, 1, 2, . . . , n − 1 let
ak0 + ak1 θ + · · · + akk−1 θ k−1 + θ k
αk = (ak0 , . . . , akk−1 ∈ Z)
dk
be a minimal integer in θ of degree k, so that α0 = d0 = 1. Then
d0 d1 · · · dn−1 = ind θ
and
di2(n−i) | D(θ ), i = 0, 1, . . . , n − 1.

Proof: By Theorem 7.2.7 {α0 , α1 , . . . , αn−1 } is an integral basis for K . Hence

D(α0 , α1 , . . . , αn−1 ) = d(K ).
However,
D(θ ) (ind θ)2 d(K )
D(α0 , α1 , . . . , αn−1 ) = = .
(d0 · · · dn−1 )2 (d0 · · · dn−1 )2
CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

170 Integral Bases

Hence, as d0 , . . . , dn−1 , ind θ are positive integers, we have

ind θ = d0 d1 · · · dn−1 .

Further
D(θ)
= d(K ) ∈ Z
(d0 · · · dn−1 )2
so that
(d0 · · · dn−1 )2 | D(θ).

For i = 0, 1, . . . , n − 1 we have by Theorem 7.2.5

di | di+1 | · · · | dn−1

so that

di2(n−i) | D(θ).

7.3 Some Integral Bases in Cubic Fields

In this section we use Theorems 7.2.7 and 7.2.8 to find integral bases for some
cubic fields. The following elementary theorem will be very helpful in connection
with the calculations.

Theorem 7.3.1 Let θ be a root of the cubic equation x 3 + ax + b = 0 (a, b ∈ Q).

Then y0 + y1 θ + y2 θ 2 (y0 , y1 , y2 ∈ Q) is a root of the cubic equation x 3 + Ax 2 +
Bx + C = 0, where
A = −3y0 + 2ay2 ,
B = 3y02 + ay12 + a 2 y22 − 4ay0 y2 + 3by1 y2 ,
C = −y03 + by13 − b2 y23 − ay0 y12 − a 2 y0 y22 + 2ay02 y2 + aby1 y22 − 3by0 y1 y2 .

Proof: Let θ, θ , θ ∈ C be the three roots of the cubic equation x 3 + ax + b = 0

so that
θ + θ + θ = 0,
θθ + θ θ + θ θ = a,
θθ θ = −b.

Then
2 2
θ 2 + θ + θ = (θ + θ + θ )2 − 2(θθ + θ θ + θ θ) = −2a,

2 2 2 2
θ 2 θ + θ θ + θ θ 2 = (θθ + θ θ + θ θ)2 − 2θθ θ (θ + θ + θ ) = a 2 ,
CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

7.3 Some Integral Bases in Cubic Fields 171

2 2 2 2
θθ + θ 2 θ + θ θ + θ 2 θ + θ θ + θ θ = θθ (θ + θ ) + θθ (θ + θ )
+ θ θ (θ + θ ) = −3θ θ θ = 3b.
Now set
α = y0 + y1 θ + y2 θ 2 ,
2
α = y0 + y1 θ + y2 θ ,
2
α = y0 + y1 θ + y2 θ .
Then
2 2
α + α + α = 3y0 + y1 (θ + θ + θ ) + y2 (θ 2 + θ + θ ) = 3y0 − 2ay2 ,

2 2 2 2
αα + α α + α α = 3y02 + y12 (θ θ + θ θ + θ θ) + y22 (θ 2 θ + θ θ + θ θ 2 )
2 2
+ 2y0 y1 (θ + θ + θ ) + 2y0 y2 (θ 2 + θ + θ )
2 2 2 2 2
+ y1 y2 (θ θ + θ 2 θ + θθ + θ 2 θ + θ θ + θ θ )
= 3y02 + ay12 + a 2 y22 − 4ay0 y2 + 3by1 y2
and
αα α = y03 + y13 θ θ θ + y23 (θ θ θ )2 + y0 y12 (θθ + θ θ + θ θ)
2 2 2 2
+ y0 y22 (θ 2 θ + θ θ + θ θ 2 ) + y02 y1 (θ + θ + θ )
2 2
+ y02 y2 (θ 2 + θ + θ ) + y12 y2 θθ θ (θ + θ + θ )
+ y1 y22 θ θ θ (θθ + θ θ + θ θ)
2 2 2 2
+ y0 y1 y2 (θ θ + θ 2 θ + θθ + θ 2 θ + θ θ + θ θ )
= y03 − by13 + b2 y23 + ay0 y12 + a 2 y0 y22 − 2ay02 y2 − aby1 y22 + 3by0 y1 y2 .
The result now follows as α = y0 + y1 θ + y2 θ 2 is a root of
(x − α)(x − α )(x − α ) = x 3 − (α + α + α )x 2
+ (αα + αα + α α )x − αα α .

Example 7.3.1 Let θ be a root of θ 3 − 3θ + 9 = 0. In the notation of Theorem

7.3.1 we have a = −3, b = 9. We determine the polynomial of which θ 2 /3 is a
root. We have y0 = y1 = 0, y2 = 1/3. Then, by Theorem 7.3.1, we obtain
1
A = 2(−3) = −2,
3
1
B = (−3)2 2 = 1,
3
2 1
C = −9 3 = −3,
3
and θ 2 /3 is a root of x 3 − 2x 2 + x − 3 = 0. This shows that θ 2 /3 is an algebraic
integer of Q(θ). In this case it is easy to check that θ 2 /3 is a root of x 3 − 2x 2 +
CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

172 Integral Bases

x − 3 = 0 directly. Let α = θ 2 /3. Then θ 2 = 3α. Hence

81 = (−9)2 = (θ 3 − 3θ )2 = θ 6 − 6θ 4 + 9θ 2 = 27α 3 − 54α 2 + 27α,

so that α 3 − 2α 2 + α − 3 = 0.

Example 7.3.2 Let θ be a root of the cubic equation x 3 − x + 4 = 0. Here a =

−1, b = 4. We consider α = 12 θ + 12 θ 2 , so that y0 = 0, y1 = 1/2, y2 = 1/2. Then,
by Theorem 7.3.1, we obtain
1
A = 2(−1) = −1,
2
1
B = (−1 + 1 + 12) = 3,
4
1
C = (4 − 16 − 4) = −2,
8
so that α is a root of x 3 − x 2 + 3x − 2 = 0. This proves that (θ + θ 2 )/2 is an
integer of Q(θ).

Example 7.3.3 Let θ be a root of the cubic equation x 3 + 11x + 4 = 0. Here

a = 11, b = 4. We consider α = 12 θ + 12 θ 2 , so that y0 = 0, y1 = 1/2, y2 = 1/2.
Then, by Theorem 7.3.1, we obtain

1
A = 2 · 11 · = 11,
2
1 1 1 1
B = 11 · 2
+ 112 · 2 + 3 · 4 · · = 36,
2 2 2 2
1 1 1
C = 4 · 3 − 4 · 3 + 4 · 11 · 3 = 4,
2
2 2 2
so that α is a root of the cubic equation x 3 + 11x 2 + 36x + 4 = 0 and thus an
integer of Q(θ) (see Example 6.3.2).

Example 7.3.4 Let θ be a root of the cubic equation x 3 − 21x − 236 = 0. Here
a = −21, b = −236. We consider α = (1 + θ)/3, so that y0 = y1 = 1/3, y2 = 0.
By Theorem 7.3.1 we obtain

1
A = −3 = −1,
3
1 1
B = 3 · 2 − 21 · 2 = −2,
3 3
1 1 1 −216
C = − 3 − 236 3 + 21 3 = = −8,
3 3 3 27
so that α is a root of the equation x 3 − x 2 − 2x − 8 = 0. Hence (1 + θ)/3 is an
integer of the cubic field Q(θ).
CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

7.3 Some Integral Bases in Cubic Fields 173

Example 7.3.5 Let θ be a root of x 3 − 21x − 236 = 0. Here a = −21, b = −236.

We consider α = (−2 − θ + θ 2 )/18, so that
−2 −1 1
y0 = , y1 = , y2 = .
18 18 18
By Theorem 7.3.1 we obtain

1 −36
A= (6 − 42) = = −2,
18 18
1 972
B = 2 (12 − 21 + 441 − 168 + 708) = = 3,
18 324
1
C = 3 (8 + 236 − 55696 − 42 + 882 − 168 − 4956 + 1416)
18
−58320
= = −10,
5832
so that α is a root of the cubic equation x 3 − 2x 2 + 3x − 10 = 0. Hence (−2 −
θ + θ 2 )/18 is an integer of the cubic field Q(θ).

In the next four examples we use Theorems 7.2.7 and 7.2.8 to give integral bases
for the following cubic fields:

Q(θ ), θ 3 − 3θ + 9 = 0 (Example 7.3.6),

Q(θ), θ 3 − θ + 4 = 0 (Example 7.3.7),
Q(θ ), θ 3 + 11θ + 4 = 0 (Example 7.3.8),
Q(θ), θ 3 − 21θ − 236 = 0 (Example 7.3.9).

Example 7.3.6 Let K = Q(θ ), θ 3 − 3θ + 9 = 0. The polynomial x 3 − 3x + 9 ∈

Z[x] is irreducible, so K is a cubic field (n = 3). By Theorem 7.1.10 we have

D(θ ) = −4(−3)3 − 27 · 92 = −2079 = −33 · 7 · 11.

Let d1 be the denominator of a minimal integer in θ of degree 1. By Theorem 7.2.8

we see that d12(3−1) | D(θ), that is, d14 | −33 · 7 · 11, so that d1 = 1. Hence θ is a
minimal integer of degree 1. Let d2 be the denominator of a minimal integer of
degree 2. By Theorem 7.2.8 we have d22(3−2) | D(θ), that is, d22 | −33 · 7 · 11, so that
d2 = 1 or 3. But it was shown in Example 7.3.1 that θ 2 /3 is an integer of K . Hence
d2 = 3 and θ 2 /3 is a minimal integer in θ of degree 2. Then, by Theorem 7.2.7,
we deduce that {1, θ, θ 2 /3} is an integral basis for K . By Theorem 7.2.8 we have
ind θ = d0 d1 d2 = 3. Thus
D(θ ) −33 · 7 · 11
d(K ) = = = −3 · 7 · 11 = −231.
(ind θ )2 32
CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

174 Integral Bases

Example 7.3.7 Let K = Q(θ), θ 3 − θ + 4 = 0. The polynomial x 3 − x + 4 ∈

Z[x] is irreducible, so K is a cubic field (n = 3). By Theorem 7.1.10 we have
D(θ) = −4(−1)3 − 27 · 42 = −22 · 107.
Let d1 be the denominator of a minimal integer in θ of degree 1. By Theorem 7.2.8 we
see that d12(3−1) | D(θ), that is, d14 | −22 · 107. But 107 is a prime, so d1 = 1. Hence
θ is a minimal integer of degree 1. Let d2 be the denominator of a minimal integer
in θ of degree 2. By Theorem 7.2.8 we have d22(3−2) | D(θ), that is, d22 | −22 · 107,
showing that d2 = 1 or 2. Now 12 θ + 12 θ 2 is an integer of K (Example
7.3.2) so
it
is a minimal integer in θ of degree 2. Hence by Theorem 7.2.7 1, θ, (θ + θ )/2 2

is an integral basis for K . By Theorem 7.2.8 we have ind θ = d0 d1 d2 = 2. Thus

D(θ) −22 · 107
d(K ) = = = −107.
(ind θ)2 22

Example 7.3.8 Let K = Q(θ), θ 3 + 11θ + 4 = 0. By Theorem 7.1.10 D(θ) =

−4 · 113 − 27 · 42 = −5756 = −22 · 1439, where 1439 is a prime. By Theorem
7.2.8 we have d14 | −22 · 1439 and d22 | −22 · 1439, so that d1 = 1 and d2 = 1
7.3.3 (θ +2 θ )/2
is an integer of K , so that d2 = 2. Thus by
2
or 2. By Example
Theorem 7.2.7 1, θ, (θ + θ )/2 is an integral basis for K . By Theorem 7.2.8
we have ind θ = d0 d1 d2 = 2. Finally,
D(θ) −22 · 1439
d(K ) = = = −1439.
(ind θ)2 22

Example 7.3.9 Let K = Q(θ), θ 3 − 21θ − 236 = 0. By Theorem 7.1.10 D(θ)

= −4(−21)3 − 27(−236)2 = 37044 − 1503792 = −1466748 = −22 · 36 · 503,
where 503 is a prime. By Theorem 7.2.8 we have d14 | −22 · 36 · 503, so that d1 = 1
or 3. In Example 7.3.4 it was shown that (1 + θ)/3 is an integer of K , so we must
have d1 = 3. By Example 7.3.5 (−2 − θ + θ 2 )/18 is an integer of K . Thus 18 | d2 ,
say d2 = 2 · 32 · m, where m ∈ N. By Theorem 7.2.8 we have
−22 · 36 · 503 = D(θ) = d(K )(d0 d1 d2 )2 = 22 · 36 · m 2 d(K ),
so that m = 1, d(K ) = −503, and

1 + θ −2 − θ + θ 2
1, ,
3 18
is an integral basis for K .

Definition 7.3.1 (Pure cubic field) A cubic field K is said to be pure if there exists
√
a rational integer m, which is not a perfect cube, such that K = Q( 3 m).
√
In Example 7.1.6 we found an integral basis for the pure cubic field Q( 3 2).
CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

7.3 Some Integral Bases in Cubic Fields 175

Example 7.3.10 We show that the cubic field K given by

K = Q(θ ), θ 3 + 6θ + 2 = 0, θ ∈ R,
√
is the pure cubic field Q( 3 2). Clearly −1 < θ < 0. Because the√
function x − 2/x
increases monotonically from −1 to 0 as√x varies from −2 to − 2, there exists a
unique real number a with −2 < a < − 2 such that
2
θ =a− .
a
Then

2 3 2
a− +6 a− +2=0
a a
so that
8
a3 − + 2 = 0.
a3
Hence
a 3 = 2 or − 4.
As a < 0 we must have a 3 = −4; that is,
a = −22/3 .
Thus
θ = 21/3 − 22/3 .
This shows that
K = Q(θ) ⊆ Q(21/3 ).
Further,
θ 2 = −4 + 2 · 21/3 + 22/3
so that
θ + θ 2 = −4 + 3 · 21/3 .
Hence
4 1 1
21/3 = + θ + θ 2,
3 3 3
proving that
Q(21/3 ) ⊆ Q(θ) = K .
This completes the proof that K = Q(21/3 ).
CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

176 Integral Bases

As
4 + θ + θ 2 2/3 4 − 2θ + θ 2
21/3 = , 2 = ,
3 3
and {1, 21/3 , 22/3 } is an integral basis for K = Q(21/3 ), we see that

4 + θ + θ 2 4 − 2θ + θ 2
1, ,
3 3
is an integral basis for K = Q(θ). Since
4 + θ + θ 2 4 − 2θ + θ 2
− =θ
3 3
and
4 + θ + θ2 1 + θ + θ2
−1= ,
3 3
a simpler integral basis is

1 + θ + θ2
1, θ, .
3
√
We now give an integral basis for the pure cubic field Q( 3 m). As we have already
mentioned this basis was first given by Dedekind [5] in 1900.

Theorem 7.3.2 Let m be a cubefree integer. Set m = hk 2 , where h is squarefree, so

that k is squarefree and (h, k) = 1. Set θ = m 1/3 and K = Q(θ). Then an integral
basis for K is

θ2
1, θ, , if m 2 ≡ 1 (mod 9),
k

k 2 ± k 2θ + θ 2
1, θ, , if m ≡ ±1 (mod 9).
3k
The discriminant d(K ) of K is given by

−27h 2 k 2 , if m 2 ≡ 1 (mod 9),
d(K ) =
−3h 2 k 2 , if m ≡ ±1 (mod 9).

We leave the proof of Theorem 7.3.2 as an exercise (Exercise 6). From Theorem
7.3.2 we obtain Table 1.
If K is a pure cubic field given in the form K = Q(θ), θ 3 + aθ + b = 0, a, b ∈
Z, it is known that −4a 3 − 27b2 = −3c2 for some positive integer c (in Exam-
ple 7.3.10 we have a = 6, b = 2, c = 18), and an integral basis for K has been
given by Spearman and Williams [15].
CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

7.4 Index and Minimal Index 177

Table 1. √Integral bases and discriminants for

Q( 3 k), 2 ≤ k ≤ 20, k cubefree
√
k Integral basis (θ = 3
k) Discriminant

2 {1, θ, θ 2 } −108 = −22 · 33

3 {1, θ, θ 2 } −243 = −35
5 {1, θ, θ 2 } −675 = −33 · 52
6 {1, θ, θ 2 } −972 = −22 · 35
7 {1, θ, θ 2 } −1323 = −33 · 72
10 {1, θ, (1 + θ + θ 2 )/3} −300 = −22 · 3 · 52
11 {1, θ, θ 2 } −3267 = −33 · 112
12 {1, θ, θ 2 /2} −972 = −22 · 35
13 {1, θ, θ 2 } −4563 = −33 · 132
14 {1, θ, θ 2 } −5292 = −22 · 33 · 72
15 {1, θ, θ 2 } −6075 = −35 · 52
17 {1, θ, (1 − θ + θ 2 )/3} −867 = −3 · 172
19 {1, θ, (1 + θ + θ 2 )/3} −1083 = −3 · 192
20 {1, θ, θ 2 /2} −2700 = −22 · 33 · 52
√ √ √ √ √
Note:
√ Q( 3 4) =√Q( 3 2), √Q( 3 9) = Q( 3 3), Q( 3 16) =
Q( 2), and Q( 18) = Q( 12).
3 3 3

The discriminant of an arbitrary cubic field K = Q(θ), θ 3 + aθ + b = 0, was

obtained by Llorente and Nart [12] in 1983, and an integral basis was first given by
Alaca [1].
We conclude this section by √ mentioning that Funakura [8] has given an integral
basis for a pure quartic field Q( 4 k), where k ∈ Z is such that x 4 − k is irreducible
over Q. Appealing to his results, we obtain Tables 2 and 3.

√
Table 2. Integral bases and discriminants for Q( 4 k),
x 4 − k irreducible in Q[x], 2 ≤ k ≤ 10
√
k Integral basis (θ = 4
k) Discriminant

2 {1, θ, θ 2 , θ 3 } −2048 = −211

3 {1, θ, θ 2 , θ 3 } −6912 = −28 · 33
5 {1, θ, (1 + θ 2 )/2, (θ + θ 3 )/2} −2000 = −24 · 53
6 {1, θ, θ 2 , θ 3 } −55296 = −211 · 33
7 {1, θ, θ 2 , θ 3 } −87808 = −28 · 73
10 {1, θ, θ 2 , θ 3 } −256000 = −211 · 53
√ √
Note: Q( 4 8) = Q( 4 2).
CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

178 Integral Bases

√
Table 3. Integral bases and discriminants for Q( 4 −k),
x 4 + k irreducible in Q[x], 1 ≤ k ≤ 10
√
k Integral basis (θ = 4
−k, arg θ = π/4) Discriminant

1 {1, θ, θ 2 , θ 3 } 256 = 28
2 {1, θ, θ 2 , θ 3 } 2048 = 211
3 {1, θ, (1 + θ 2 )/2, (θ + θ 3 )/2} 432 = 24 · 33
5 {1, θ, θ 2 , θ 3 } 32000 = 28 · 53
6 {1, θ, θ 2 , θ 3 } 55296 = 211 · 33
7 {1, θ, (1 + θ 2 )/2, (1 + θ + θ 2 + θ 3 )/4} 1372 = 22 · 73
9 {1, θ, θ 2 /3, θ 3 /3} 2304 = 28 · 32
10 {1, θ, θ 2 , θ 3 } 256000 = 211 · 53
√ √ √
Note: Q( 4 −4) = Q(i) and Q( 4 −8) = Q( 4 −2).

7.4 Index and Minimal Index of an Algebraic Number Field

Let K be an algebraic number field of degree n over Q. An element α ∈ O K is
called a generator of K if K = Q(α). By Theorem 6.4.3 α is a generator of K if
and only if D(α) = 0. For a generator α of K , the index of α is the positive integer
ind α given by
D(α) = (ind α)2 d(K )
(see Definition 7.1.4). We now define the index i(K ) and minimal index m(K ) of
the field K .

Definition 7.4.1 (Index of a field) The index of K is

i(K ) = gcd {ind α | α a generator of K }.

Definition 7.4.2 (Minimal index of a field) The minimal index of K is

m(K ) = min {ind α | α a generator of K }.

Clearly
i(K ) | m(K ). (7.4.1)

Theorem 7.4.1 Let K be an algebraic number field. Then m(K ) = 1 if and only if
K possesses a power basis.

Proof: Suppose m(K ) = 1. Then there exists a generator α of K such that

ind α = 1. Hence D(1, α, . . . , α n−1 ) = D(α) = (ind α)2 d(K ) = d(K ) so that
CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

7.4 Index and Minimal Index 179

{1, α, . . . , α n−1 } is an integral basis for K . Hence K possesses a power basis.

Conversely, suppose K possesses a power basis, say {1, α, . . . , α n−1 }. Then
{1, α, . . . , α n−1 } is an integral basis for K and so
D(1, α, . . . , α n−1 ) = d(K ).
But
D(1, α, . . . , α n−1 ) = D(α) = (ind α)2 d(K ),
so ind α = 1 and hence m(K ) = 1.

From (7.4.1) and Theorem 7.4.1 we obtain

Theorem 7.4.2 Let K be an algebraic number field such that K possesses a power
basis. Then i(K ) = 1.

In Example 7.4.4 we give an algebraic number field K for which i(K ) = 1 but
K does not possess a power basis. This shows that the converse of Theorem 7.4.2
is not true. Theorem 7.4.2 gives a convenient way of establishing that an algebraic
number field does not have a power basis; all we have to do is to show that i(K ) ≥ 2.
In the next theorem we determine the index and minimal index of a quadratic
field directly from their definitions.

Theorem 7.4.3 Let K be a quadratic field. Then i(K ) = m(K ) = 1.

Proof: As K is a quadratic field, by Theorem 5.4.1 there exists a unique squarefree

√
integer m such that K = Q( m). √
First we suppose that m ≡ 1 (mod 4) so that 1, 1+2 m is an integral basis
for K (Theorem
√
5.4.2) and d(K ) = m (Theorem 7.1.2). Let α ∈ O K . Then α =
1+ m
a+b 2
for some a, b ∈ Z. Now
√ 2
1+ m
1 a+b
2 √
D(α) = √ = (−b m)2 = b2 m,
1 1− m
a+b
2
so that
D(α) = 0 if and only if b = 0.
Thus α is a generator of K if and only if b = 0. Further,

D(α) b2 m
ind α = = = |b|,
d(K ) m
CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

180 Integral Bases

so that

i(K ) = gcd {|b| | b ∈ Z, b = 0} = 1

and

m(K ) = min {|b| | b ∈ Z, b = 0} = 1.

√
Next we suppose that m ≡ 2 or 3 (mod 4), so that {1, m} is an integral basis
for K (Theorem 5.4.2) and d(K ) = 4m (Theorem 7.1.2). Let α ∈ O K . Then α =
√
a + b m for some a, b ∈ Z. Now
√ 2
1 a + b m √
D(α) = √ = (−2b m)2 = 4b2 m,
1 a−b m

so that

D(α) = 0 if and only if b = 0.

Thus α is a generator of K if and only if b = 0. Further,

D(α) 4mb2
ind α = = = |b|,
d(K ) 4m

so that

i(K ) = gcd {|b| | b ∈ Z, b = 0} = 1

and

m(K ) = min {|b| | b ∈ Z, b = 0} = 1.

Of course we could have argued that a quadratic field clearly has a power basis
so that by Theorem 7.4.1 m(K ) = 1 and then by (7.4.1) i(K ) = 1.
In the next four examples we determine i(K ) and m(K ) for some cubic fields K .

Example 7.4.1 We determine the index i(K ) and the minimal index m(K ) of the
cubic field K = Q(θ), where θ is a root of f (x) = x 3 − 3x + 9. Let θ and θ
be the other two roots of f (x), so that x 3 −
3x + 9 = (x − θ )(x − θ )(x − θ ).
By Example 7.3.6 we know that 1, θ, θ 2 /3 is an integral basis for K , D(θ) =
−33 · 7 · 11, and d(K ) = −3 · 7 · 11. Let α ∈ O K . Then α = a + bθ + cθ 2 /3 for
CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

7.4 Index and Minimal Index 181

some a, b, c ∈ Z. The conjugates of α are

θ2
α = a + bθ + c ,
3
2

θ
α = a + bθ + c ,
3
2

θ
α = a + bθ + c .
3
Hence, as θ + θ + θ = 0, we have
c c
α − α = (θ − θ ) b + (θ + θ ) = (θ − θ ) b − θ ,
3 3c
c
α − α = (θ − θ ) b + (θ + θ ) = (θ − θ ) b − θ ,

3 3c
c
α − α = (θ − θ ) b + (θ + θ ) = (θ − θ ) b − θ .
3 3
Thus, by Theorem 6.4.1,

D(α) = (α − α )2 (α − α )2 (α − α )2
c c c
= (θ − θ )2 (θ − θ )2 (θ − θ )2 (b − θ)2 (b − θ )2 (b − θ )2
3 3 3
2
c 3 3b
= D(θ) f
3 c
2
bc2 c3
= −3 · 7 · 11 b −
3 3
+
3 3
= −3 · 7 · 11 (3b − bc + c ) .
3 2 3 2

Then

D(α) −3 · 7 · 11(3b3 − bc2 + c3 )2
ind α = = = |3b3 − bc2 + c3 |.
d(K ) −3 · 7 · 11
Hence

m(K ) = min {|3b3 − bc2 + c3 | | b, c ∈ Z, 3b3 − bc2 + c3 = 0} = 1

3b3 − bc2 + c3 = 1 for (b, c) = (1, −1).

By (7.4.1) i(K ) = 1. As m(K ) = 1, K has a power basis by Theorem 7.4.1. Now

2
θ2 θ2 θ2 θ4 θ2 θ2 θ2
D 1, , = D 1, , = D 1, , − θ = D 1, , −θ
3 3 3 9 3 3 3
2
θ
= D 1, θ, = d(K ),
3
CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

182 Integral Bases

so
2
θ2 θ2
1, ,
3 3

is a power basis for K . This is easily seen directly as

2
θ2 θ2 θ2
a + bθ + c = a + (b + c) − b
3 3 3
for all a, b, c ∈ Z.

Example 7.4.2 We show that m(K ) = i(K ) = 1 for the cubic field K = Q(θ ),
where θ is a root of f (x) = x 3 − x + 4.
Let θ and θ be the other two roots of f (x) so that

f (x) = x 3 − x + 4 = (x − θ)(x − θ )(x − θ ).

By Example 7.3.7 we know that 1, θ, θ +2 θ is an integral basis for K and d(K ) =
2

−107. Let α ∈ O K . Then α = a + bθ + c θ +2 θ for some a, b, c ∈ Z. Exactly as
2

in Example 7.4.1 we find that

D(α) = −107(2b3 + 3b2 c + bc2 + c3 )2 .

Then

D(α) −107(2b3 + 3b2 c + bc2 + c3 )2
ind α = = = |2b3 + 3b2 c + bc2 + c3 |.
d(K ) −107
Hence

m(K ) = min {|2b3 + 3b2 c + bc2 + c3 | | b, c ∈ Z, 2b3 + 3b2 c + bc2 + c3 = 0} = 1

2b3 + 3b2 c + bc2 + c3 = 1 for (b, c) = (−1, 1).

Then, by (7.4.1), i(K ) = 1.

As m(K ) = 1, K has a power basis by Theorem 7.4.1. Now, as θ 3 = θ − 4, θ 4 =
θ 2 − 4θ , we obtain
2
θ + θ2 θ + θ2 θ + θ2 θ2 − θ
D 1, , = D 1, , −2
2 2 2 2

θ + θ2 θ2 − θ θ + θ2
= D 1, , = D 1, ,θ
2 2 2
θ + θ2
= D(1, θ, ) = d(K ),
2
CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

7.4 Index and Minimal Index 183

so
2
θ + θ2 θ + θ2
1, ,
2 2

is a power basis for K .

Example 7.4.3 We show that

m(K ) = i(K ) = 2 for the cubic field K = Q(θ ), θ 3 − 21θ − 236 = 0.

Let θ and θ be the other two roots of f (x) = x 3 − 21x − 236 so that

f (x) = x 3 − 21x − 236 = (x − θ)(x − θ )(x − θ ). (7.4.2)

By Example 7.3.9 we know that 1, 1 +3 θ , − 2 −18θ + θ is an integral basis for K and
2

d(K ) = −503. Let α ∈ O K . Then

(1 + θ) −2 − θ + θ 2
α =a+b +c
3 18

for some a, b, c ∈ Z. The other conjugates of α are

2

(1 + θ ) −2 − θ + θ
α = a + b +c
3 18

and
2

(1 + θ ) −2 − θ + θ
α =a+b +c .
3 18

From (7.4.2) we deduce that θ + θ + θ = 0. Thus,

2
θ 2 − θ = (θ − θ )(θ + θ ) = −(θ − θ )θ ,

and we obtain

b c c
α − α = (θ − θ ) − − θ ,
3 18 18

and similarly

b c c b c c
α − α = (θ − θ ) − − θ , α − α = (θ − θ ) − − θ .
3 18 18 3 18 18
CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

184 Integral Bases

Hence, appealing to (7.4.2), we obtain, as D(θ) = −22 · 36 · 503 from Example

7.3.9,
2
D(α) = (α − α )(α − α )(α − α )
2
2 c 6 6b
= (θ − θ )(θ − θ )(θ − θ ) f − 1
186 c
2
3
c6 6b 6b
= D(θ) 6 12 − 1 − 21 − 1 − 236
23 c c
−22 · 36 · 503
3 2
= (6b − c)3
− 21c 2
(6b − c) − 236c
26 312
503
= − 4 6 {216b3 − 108b2 c − 108bc2 − 216c3 }2
2 ·3
= −503{2b3 − b2 c − bc2 − 2c3 }2 .
Then

D(α) −503(2b3 − b2 c − bc2 − 2c3 )2
ind α = = = |2b3 − b2 c − bc2 − 2c3 |.
d(K ) −503
Now
2b3 − b2 c − bc2 − 2c3 ≡ 0 (mod 2)
for all b, c ∈ Z, so that
ind α ≡ 0 (mod 2)
for all α ∈ O K . But
2b3 − b2 c − bc2 − 2c3 = 2
for (b, c) = (1, 0), so that
m(K ) = min {|2b3 − b2 c − bc2 − 2c3 | | b, c ∈ Z, 2b3 − b2 c − bc2 − 2c3 = 0} = 2
and
i(K ) = gcd {|2b3 − b2 c − bc2 − 2c3 | | b, c ∈ Z, 2b3 − b2 c − bc2 − 2c3 = 0} = 2.
As m(K ) = 2, K does not possess a power basis by Theorem 7.4.1.

Dedekind [4] gave in 1878 the first example of an algebraic number field without
a power basis, namely, the cubic field L given by
L = Q(φ), φ 3 − φ 2 − 2φ − 8 = 0.
The field L is in fact the same field as the field K = Q(θ), θ 3 − 21θ − 236 = 0,
in Example 7.4.3, as θ and φ are related by θ = 3φ − 1.
CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

7.4 Index and Minimal Index 185

The next example gives a cubic field K for which

i(K ) = 1, m(K ) = 2, K does not possess a power basis,

which shows that the converse of Theorem 7.4.2 does not hold.
√
Example 7.4.4 Let K = Q( 3 175). An integral basis for K is given by
{1, 1751/3 , 2451/3 } and d(K ) = −33 · 52 · 72 (see Theorem 7.3.2). Let α ∈ O K .
Then there exist a, b, c ∈ Z such that

α = a + b1751/3 + c2451/3 .

The other conjugates of α are

α = a + bω1751/3 + cω2 2451/3 ,

α = a + bω2 1751/3 + cω2451/3 ,

where ω is a complex cube root of unity. As

1 + ω + ω2 = 0, ω3 = 1,

we obtain

α − α = (1 − ω)(b1751/3 − cω2 2451/3 ),

α − α = (1 − ω2 )(b1751/3 − cω2451/3 ),
α − α = (ω − ω2 )(b1751/3 − c2451/3 ).

Hence

D(α) = (α − α )2 (α − α )2 (α − α )2
= {(1 − ω)(1 − ω2 )(ω − ω2 )}2 (175b3 − 245c3 )2
= −27(175b3 − 245c3 )2 = −33 · 52 · 72 (5b3 − 7c3 )2 .

Then

D(α) −33 · 52 · 72 (5b3 − 7c3 )2
ind α = = = |5b3 − 7c3 |.
d(K ) −33 · 52 · 72

Thus

i(K ) = gcd {|5b3 − 7c3 | | b, c ∈ Z, 5b3 − 7c3 = 0}

and

m(K ) = min {|5b3 − 7c3 | | b, c ∈ Z, 5b3 − 7c3 = 0}.

Since |5 · 13 − 7 · 13 | = 2 and |5 · 13 − 7 · 03 | = 5 we see that i(K ) = 1 and

m(K ) = 1 or 2. Suppose m(K ) = 1. Then there exist integers B and C such that
CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

186 Integral Bases

5B 3 − 7C 3 = ±1. Thus 5B 3 ≡ ±1 (mod 7), so B 3 ≡ ±3 (mod 7). But this is im-

possible as the only cubes modulo 7 are 0, ±1. Hence m(K ) = 2. By Theorem 7.4.1
K does not possess a power basis.

Llorente and Nart [12, Theorem 4, p. 585] have given a necessary and sufficient
condition for a cubic field to have index 2.

Definition 7.4.3 (Inessential discriminant divisor) Let K be an algebraic number

field of degree n over Q. A prime p is called an inessential discriminant divisor or
common index divisor if p | ind α for every generator α of K .

The inessential discriminant divisors of an algebraic number field K are precisely

the prime factors of the index i(K ). Example 7.4.3 shows that the only inessential
discriminant divisor of the cubic field Q(θ), θ 3 − 21θ − 236 = 0, is the prime 2.
Indeed the set of inessential discriminant divisors of a cubic field is either the empty
set φ or {2}. This is a special case of the general result due to Zyliński [20] that
a prime p can be an inessential discriminant divisor of an algebraic number field
of degree n only if p < n. Thus when K is a quartic field the set of inessential
discriminant divisors is φ, {2}, {3}, or {2, 3}.

7.5 Integral Basis of a Cyclotomic Field

Let m be a positive integer. The number of positive integers less than or equal to
m that are coprime with m is denoted by φ(m). The arithmetic function φ(m) is
called Euler’s phi function. Let ζm be any primitive mth root of unity. There are
φ(m) primitive mth roots of unity, namely ζmr , r = 1, 2, . . . , m, (r, m) = 1. Let
K m = Q(ζm ). It is easy to show that K m = Q(ζmr ) for any r ∈ {1, 2, . . . , m} with
(r, m) = 1, so that K m is independent of the primitive mth root of unity chosen.
The field K m is called the mth cyclotomic field. For odd m the fields K m and
K 2m coincide as −ζm is a primitive 2mth root of unity. Clearly ζm is a root of the
polynomial

m
f m (x) = (x − ζmr ).
r =1
(r, m) = 1

It is known that f m (x) ∈ Z[x] and that f m (x) is irreducible, so that

irrQ (ζm ) = f m (x).

Moreover, the degree of f m (x) is φ(m) so that

[K m : Q] = φ(m).
CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

7.5 Integral Basis of a Cyclotomic Field 187

The smallest field containing both K m and K n is K [m,n] , where [m, n] denotes the
least common multiple of m and n. Also, K m ∩ K n = K (m,n) . If m ≡ 2 (mod 4) then
K m ⊆ K n holds if and only if m | n. Thus if m and n are distinct and not congruent
to 2 (mod 4) the cyclotomic fields K m and K n are distinct.
The next theorem gives an integral basis for K m as well as a formula for the
discriminant d(K m ).

Theorem 7.5.1 Let m be a positive integer. Let ζm be a primitive mth root of

unity. Let K m denote the cyclotomic field Q(ζm ). Then {1, ζm , ζm2 , . . . , ζmφ(m)−1 } is
an integral basis for K m . Further,
φ(m) m φ(m)
d(K m ) = (−1) 2 φ(m) ,
p p−1
p|m

where the product is over all primes p dividing m.

We refer the reader to Narkiewicz [13, Theorem 4.10, p. 169] for a proof of this
theorem.
Taking m = 3, 4, 5, 8 in Theorem 7.5.1, we obtain d(K 3 ) = −3, d(K 4 ) =
−4, d(K√ 5 ) = 125, d(K 8 ) =√256. The first two of these are familar to us as
K 3 = Q( −3) and √ K4 = √ Q( −1) are quadratic fields. The fourth equality is
√also
known to us as 2 ( 2 + −2) is a primitive eighth root of unity, so K 8 = Q( 2 +
1
√ √ √ √ √
−2) = Q( 2, −2) = Q( 2, i) and we showed that d(Q( 2, i)) = 256 in
Example 7.1.7.

Example 7.5.1 We show that

√
K 5 = Q i 10 + 2 5 .

Let β be the primitive fifth root of unity, e2πi/5 , so that

2π 2π
β = e2πi/5 = cos + i sin = c + is,
5 5
where
2π 2π
c = cos , s = sin .
5 5
Then
1 = β 5 = (c + is)5 = (c5 − 10c3 s 2 + 5cs 4 ) + i(5c4 s − 10c2 s 3 + s 5 ).
Hence
5c4 s − 10c2 s 3 + s 5 = 0.
CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

188 Integral Bases

As s = 0 and c2 + s 2 = 1 we obtain
5c4 − 10c2 (1 − c2 ) + (1 − c2 )2 = 0,
that is,
16c4 − 12c2 + 1 = 0,
so that
√
3± 5
c = . 2
8
√
Now c ≈ 0.3, c2 ≈ 0.09, (3 − 5)/8 ≈ 0.09, so
√
3− 5
c =
2
8
and
√ √ √
3− 5 6−2 5 5−1
c= = = .
8 16 4
Hence
√ 2 √
5 − 1 10 + 2 5
s =1−c =1−
2 2
= ,
4 16
so
√
10 + 2 5
s= .
4
We have shown that

2πi 1 √ √
β=e 5 = 5 − 1 + i 10 + 2 5 .
4
Squaring we obtain

4πi 1 √ √
β =e
2 5 = − 5−1+i 10 − 2 5 ,
4
as
√
√ 5−1 √
10 − 2 5 = 10 + 2 5.
2
Further,

12 √ √
β = β̄ =
3
− 5 − 1 − i 10 − 2 5 ,
4

1 √ √
β 4 = β̄ = 5 − 1 − i 10 + 2 5 .
4
CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

Exercises 189

Hence

1 √
β − β4 = i 10 + 2 5,
2
so

√
Q i 10 + 2 5 = Q(2β − 2β 4 ) ⊆ Q(β).

Also,

1 √ √ 2
β= −12 + 2 i 10 + 2 5 − i 10 + 2 5 ,
8
so

√
Q(β) ⊆ Q i 10 + 2 5 .

This shows that

√
K 5 = Q i 10 + 2 5

and

√
d Q i 10 + 2 5 = d(K 5 ) = 125.

√
Integral bases for quartic fields like Q(i 10 + 2 5), which contain a quadratic
subfield, have been given by Huard, Spearman, and Williams [10].
The final theorem of this chapter is immediate from Definition 7.1.5 and Theorem
7.5.1.

Theorem 7.5.2 The cyclotomic field K m = Q(ζm ) is monogenic for every positive
integer m.

Exercises
1. Let D denote the discriminant of

f (x) = x n + an−1 x n−1 + · · · + a1 x + a0 ∈ Z[x].

Prove that

D ≡ 0 or 1 (mod 4).
√ √
2. Using
√ the method of Example 7.1.6, prove
√ that {1, 3 3, ( 3 3)2 } is an integral basis for
Q( 3 3). What is the discriminant of Q( 3 3)?
CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

190 Integral Bases

3. Prove that
√ √ √ √ √
1 + 5 1 + 13 1 + 5 + 13 + 65
1, , ,
2 2 4
√ √
√ for K = Q( 5, 13). What is d(K )?
is an integral√basis
4. Let K = Q( 5, 13). Use Exercise 3 to prove that

1 √ √ √
O K = { (x + y 5 + z 13 + w 65) | x, y, z, w ∈ Z,
4
x ≡ y ≡ z ≡ w (mod 2), x − y − z + w ≡ 0 (mod 4)}.

5. Let K = Q(θ ), where θ 3 − 9θ − 6 = 0. Prove that {1, θ, θ 2 } is an integral basis for

K and that d(K ) = 23 · 35 .
6. Prove Theorem 7.3.2.
7. Let
K = Q(θ),
where θ 3 − 6θ + 36 = 0. Prove that θ 2 /6 ∈ O K . Show that
1, θ, θ /6 is an integral basis for K and that d(K ) = −22 · 3 · 79.
2

8. Let K = Q(θ ), where θ3 − 3θ + 56 = 0. Prove that (θ − 1)/3 ∈ O K and (θ 2 + θ −

2)/9 ∈ O K . Show that 1, θ −3 1 , θ +9θ − 2 is an integral basis for K and that d(K ) =
2

−22 · 29.
9. Let K 1 = Q(θ1 ), where θ13 + 27θ1 + 240 = 0, and K 2 = Q(θ2 ), where θ23 + 27θ2 +
72 = 0. Prove that d(K 1 ) = d(K 2 ) = −35 . Is K 1 = K 2 ?
10. Let K = Q(θ),√ where θ 4 − 17θ 2 − 34θ − 17 = 0. Prove that d(K ) = 173 .
11. Let K = Q( 2). Prove that d(K ) = −211 .
4

12. Prove from first principles that K = Q(θ ), θ 3 + 30θ + 90 = 0, θ ∈ R, is a pure cubic
field, and express K in the form K = Q(m 1/3 ) for some cubefree integer m.
13. Let K = Q(θ ), where θ 3 − 4θ + 2 = 0. Prove that {1, θ, θ 2 } is an integral basis for K
and that d(K ) = 22 ·37. √
14. Prove that K 5 = Q(i 5 + 2 5).
15. If p is an odd prime prove that
p−1
[Q(e2πi/ p + e−2πi/ p ) : Q] = .
2
16. Suppose that x 3 + ax + b ∈ Z[x] is irreducible. Prove that K = Q(θ ), θ 3 + aθ + b =
0, θ ∈ R, is a pure cubic field if and only if −4a 3 − 27b2 = −3c2 for some positive
integer c.
17. Let K be an algebraic number field. Let L be a conjugate field of K . Prove that
d(K ) = d(L).
18. Let K be an algebraic number field. Let σ be a monomorphism : K −→ C. Let L be the
conjugate field σ (K ). Let {ω1 , . . . , ωn } be an integral basis for K . Is {σ (ω1 ), . . . , σ (ωn )}
an integral basis for L?
19. Determine an integral basis for

K = Q(θ), θ 3 + 30θ + 15 = 0.

20. If K and L are algebraic number fields with K ⊆ L prove that d(K ) | d(L).
CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

where y0 , y1 , . . . , yn−1 ∈ Q. Prove that

y j D(θ ) ∈ Z, j = 0, 1, . . . , n − 1.

22. Let K be an algebraic number field of degree n. Is it possible to find λ1 , . . . , λn ∈ O K

such that D(λ1 , . . . , λn ) = −d(K )?
23. Let K be an algebraic number field. Prove from first principles that an integral basis
for K can always be chosen to include 1.
24. Prove that
√ √
√
3 1 + 3 10 + ( 3 10)2
1, 10,
3
√
is an integral basis
√ for
√ Q( 3 √10) using Theorem 7.2.7. √
25. Prove that {1, 4 2, ( 4 2)2 , ( 4 2)3 } is an integral basis for Q( 4 2) using the ideas of
Example 7.1.6.
26. Prove that
√ √ √
√ 1 + ( 4 5)2 4 5 + ( 4 5)3
,
4
1, 5,
2 2
√
is an integral basis for Q( 4 5) using Theorem 7.2.7.
27. Use Brill’s theorem to show that
φ(m)
sgn(d(K m )) = (−1) 2 .

28. Let m be a positive integer. Let ζm be a primitive mth root of unity. What is sgn(d(Q(ζm +
ζm−1 )))?

Suggested Reading
1. Ş. Alaca, p-integral bases of a cubic field, Proceedings of the American Mathematical
Society 126 (1998), 1949–1953.
A p-integral basis of a cubic field K is determined for each rational prime p, and then an integral
basis of K and the discriminant d(K ) of K are obtained from its p-integral bases.
2. Ş. Alaca, p-integral bases of algebraic number fields, Utilitas Mathematica 56 (1999),
97–106.
The properties of p-integral bases of an algebraic number field K are developed and used to
show how an integral basis of K can be obtained from its p-integral bases.
3. A. Brill, Ueber die Discriminante, Mathematische Annalen 12 (1877), 87–89.
This is the original paper of Brill giving the sign of the discriminant of an algebraic number field.
CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

192 Integral Bases

4. R. Dedekind, Über den Zusammenhang zwischen der Theorie der Ideale und der Theo-
rie der höheren Kongruenzen, Abh. Kgl. Ges. Wiss. Göttingen 23 (1878), 1–23. (Gesam-
melte Mathematische Werke I, pp. 202–232, Vieweg, Wiesbaden, 1930.)
It is shown that the cubic field K = Q(θ), θ 3 − θ 2 − 2θ − 8 = 0, does not have a power
basis.
5. R. Dedekind, Über die Anzahl der Idealklassen in reinen kubischen Zahlkörpern, Jour-
nal für die reine und angewandte Mathematik 121 (1900), 40–123. (Gesammelte Math-
ematische Werke II, pp. 148–233, Vieweg, Wiesbaden, 1931.)
√
An integral basis is given for the pure cubic field Q( 3 m).
6. D. S. Dummit and H. Kisilevsky, Indices in cyclic cubic fields, in Zassenhaus, H. (ed.),
Number Theory and Algebra, Collected Papers Dedicated to Henry B. Mann, Arnold
E. Ross and Olga Taussky-Todd, pp. 29–42, Academic Press, New York, 1979.
It is shown that infinitely many cyclic cubic fields have a power basis.
7. H. T. Engstrom, On the common index divisors of an algebraic field, Transactions of
the American Mathematical Society 32 (1930), 223–237.
The basic properties of the index of an algebraic number field are given.
8. T. Funakura, On integral bases of pure quartic fields, Mathematical Journal of Okayama
University 26 (1984), 27–41.
An explicit integral basis is given for a pure quartic field.
9. M.-N. Gras, Sur les corps cubiques cycliques dont l’anneau des entiers est monogène,
Annales Scientifiques de l’Université de Besançon, Mathematics, 1973, 26 pp.
Necessary and sufficient conditions are given for a cyclic cubic field to have a power basis.
It can be deduced from these that infinitely many cyclic cubic fields do not have a power
basis.
10. J. G. Huard, B. K. Spearman, and K. S. Williams, Integral bases for quartic fields with
quadratic subfields, Journal of Number Theory 51 (1995), 87–102.
√
Let L be√ quartic
field √ with quadratic √subfield Q( c), where √
c is a squarefree integer. Then
L = Q( c, a + b c), where a + b c is not a square in Q( c). The discriminant of L and
an integral basis for L are determined explicitly.
11. R. H. Hudson and K. S. Williams, The integers of a cyclic quartic field, Rocky Mountain
Journal of Mathematics 20 (1990), 145–150.
An explicit integral basis is given for a quartic field with Galois group Z4 .
12. P. Llorente and E. Nart, Effective determination of the rational primes in a cubic field,
Proceedings of the American Mathematical Society 87 (1983), 579–585.
A necessary and sufficient condition is given for a cubic field to have index 2.
13. W. Narkiewicz, Elementary and Analytic Theory of Algebraic Numbers, Springer-
Verlag, Berlin, 1990.
The principal properties of cyclotomic fields are summarized in Theorem 4.10, p. 169.
14. B. K. Spearman and K. S. Williams, The conductor of a cyclic quartic field, Publica-
tiones Mathematicae 48 (1996), 13–43.
An explicit formula is given for the discriminant of a cyclic quartic field Q(θ), where θ 4 + Aθ 2 +
Bθ + C = 0.
15. B. K. Spearman and K. S. Williams, An explicit integral basis for a pure cubic field,
Far East Journal of Mathematical Sciences 6 (1998), 1–14.
An explicit integral basis is given for a pure cubic field K = Q(θ), θ 3 + aθ + b = 0.
CB609-07 CB609/Alaca & Williams August 7, 2003 16:45 Char Count= 0

Biographies 193

16. B. K. Spearman and K. S. Williams, Cubic fields with a power basis, Rocky Mountain
Journal of Mathematics 31 (2001), 1103–1109.
It is shown that there exist infinitely many cubic fields L with a power basis such that the splitting
field M of L contains a given quadratic field K .
17. B. K. Spearman and K. S. Williams, Cubic fields with index 2, Monatshefte für Math-
ematik 134 (2002), 331–336.
Let d be a squarefree integer with d = 1 allowed. If d ≡ 1 (mod 8) it√is shown that there do
not exist any cubic fields with index 2 whose splitting field contains Q( d). If d ≡ 1 (mod 8) it
√ exist infinitely many cubic fields K with i(K ) = m(K ) = 2 whose splitting
is shown that there
field contains Q( d).
18. L. Stickelberger, Über eine neue Eigenschaft der Diskriminanten algebraischer
Zahlkörper, International Mathematische Kongress, Zürich, 1897, 182–193.
It is shown that d(K ) ≡ 0 or 1 (mod 4) for an algebraic number field K .
19. K. S. Williams, Integers of biquadratic fields, Canadian Mathematical Bulletin 13
(1970), 519–526.
√ √ √ √
This paper gives an explicit integral basis for the quartic field Q( m + n) = Q( m, n),
where m and n are distinct squarefree integers.
20. E. Zyliński, Zur Theorie der ausserwesentlichen Discriminantenteiler algebraischer
Körper, Mathematische Annalen 73 (1913), 273–274.
It is shown that a prime p can only be an inessential discriminant divisor of an algebraic number
field of degree n if p < n.

Biographies
1. E. T. Bell, Men of Mathematics, Simon and Schuster, New York, 1937.
Chapter 27 is devoted to Ernst Kummer (1810–1893) and Richard Dedekind (1831–1916).
2. R. A. Mollin, Algebraic Number Theory, Chapman and Hall/CRC Press, London/Boca
Raton, Florida, 1999.
A brief biography of Ludwig Stickelberger (1850–1936) is given on page 43.
3. The website

http://www-groups.dcs.st-and.ac.uk/˜history/

has biographies of Alexander Brill (1842–1935) and Richard Dedekind (1831–1916).

CB609-08 CB609/Alaca & Williams August 7, 2003 16:49 Char Count= 0

8
Dedekind Domains

8.1 Dedekind Domains

In Chapter 6 it was shown that the ring of algebraic integers O K of an algebraic
number field K has the following three properties:
O K is a Noetherian domain (Theorem 6.5.3),
O K is integrally closed (Theorem 6.1.6), and
each prime ideal P of O K is a maximal ideal (Theorem 6.6.1).

An integral domain with these properties is called a Dedekind domain after Richard
Dedekind, the creator of the modern theory of ideals.

Definition 8.1.1 (Dedekind domain) An integral domain D that satisfies the fol-
lowing three properties:
D is a Noetherian domain, (8.1.1)
D is integrally closed, and (8.1.2)
each prime ideal of D is a maximal ideal, (8.1.3)
is called a Dedekind domain.

In view of the remarks before Definition 8.1.1, we have

Theorem 8.1.1 Let K be an algebraic number field. Let O K be the ring of integers
of K . Then O K is a Dedekind domain.

The next theorem gives another class of integral domains that are Dedekind
domains.

Theorem 8.1.2 Let D be a principal ideal domain. Then D is a Dedekind domain.

Proof: Let D be a principal ideal domain. By Theorem 3.1.2 D is a Noetherian

domain, so (8.1.1) holds. By Theorem 3.3.1 D is a unique factorization domain and

194
CB609-08 CB609/Alaca & Williams August 7, 2003 16:49 Char Count= 0

8.2 Ideals in a Dedekind Domain 195

thus, by Theorem 4.2.5, D is integrally closed, so (8.1.2) holds. By Theorem 1.5.7

each prime ideal of D is maximal so that (8.1.3) holds. Hence D is a Dedekind
domain.

Our main objective in this chapter is to show that every ideal I (= 0, 1) of a
Dedekind domain can be expressed uniquely as a product of prime ideals. We also
show that every ideal of a Dedekind domain is generated by at most two elements.

8.2 Ideals in a Dedekind Domain

The first step toward our objective of proving that in a Dedekind domain every
proper ideal is a product of prime ideals is to show that every such ideal contains a
product of prime ideals. This is actually true in a Noetherian domain.

Theorem 8.2.1 In a Noetherian domain every nonzero ideal contains a product of

one or more prime ideals.

Proof: Suppose that D is a Noetherian domain that possesses at least one nonzero
ideal that does not contain a product of one or more prime ideals. Let S be the set
of all such ideals. By assumption S is not empty. As D is Noetherian, by Theorem
3.1.3 S contains a (nonzero) ideal A maximal with respect to the property of not
containing a product of one or more prime ideals. Clearly A itself is not a prime
ideal. Hence, by Theorem 1.6.1, there exist ideals B and C such that
BC ⊆ A, B ⊆ A, C ⊆ A.
Define the ideals B1 and C1 of D by
B1 = A + B, C1 = A + C.
Clearly
A ⊂ B1 , A ⊂ C1 ,
so that B1 ∈ S, C1 ∈ S. Hence there exist prime ideals P1 , . . . , Pk such that
B1 ⊇ P1 · · · Ph , C1 ⊇ Ph+1 · · · Pk .
But
B1 C1 = (A + B)(A + C) ⊆ A,
so
A ⊇ P1 · · · Pk ,
contradicting that A ∈ S.
CB609-08 CB609/Alaca & Williams August 7, 2003 16:49 Char Count= 0

196 Dedekind Domains

As a Dedekind domain is a Noetherian domain, the next theorem is an immediate

consequence of Theorem 8.2.1.

Theorem 8.2.2 In a Dedekind domain every nonzero ideal contains a product of

one or more prime ideals.

Our next step is to obtain an inverse of a prime ideal P in a Dedekind domain.

To do this, we extend the notion of an “ideal” to that of a “fractional ideal.”

Definition 8.2.1 (Fractional ideal) Let D be an integral domain. Let K be the

quotient field of D. A nonempty subset A of K with the following three properties:
(i) α ∈ A, β ∈ A =⇒ α + β ∈ A,
(ii) α ∈ A, r ∈ D =⇒ r α ∈ A, and
(iii) there exists γ ∈ D with γ = 0 such that γ A ⊆ D

is called a fractional ideal of D.

Condition (iii) means that the elements of a fractional ideal have γ as a “common
denominator.”

Example 8.2.1 Let

n
A= |n∈Z
25
so that A is a nonempty subset of Q. Clearly A has properties (i) and (ii). Also,
25A = Z so that (iii) holds. Hence A is a fractional ideal of Z.

Example 8.2.2 Let

n
A= | n ∈ Z, m ∈ N ∪ {0} .
5m
Clearly A is a nonempty subset of Q having properties (i) and (ii). However, there
is no nonzero integer k such that k A ⊆ Z, so (iii) does not hold. Thus A is not a
fractional ideal of Z.

A fractional ideal of D that is a subset of D is clearly an ideal of D in the ordinary

sense. Moreover, an ideal of D is a fractional ideal of D that is a subset of D. We
often refer to the ideals of D in the ordinary sense as integral ideals. If A is a
fractional ideal of D and γ is a common denominator for A then γ A is an integral
ideal of D.
It follows immediately from Definition 8.2.1 that if A is a fractional ideal of D
then
1
A = I,
γ
CB609-08 CB609/Alaca & Williams August 7, 2003 16:49 Char Count= 0

8.2 Ideals in a Dedekind Domain 197

where γ ∈ D \ {0} and I is an integral ideal of D. This representation is not unique

as
1
A= (δ I )
γδ
for any δ ∈ D \ {0}.
If D is a Noetherian domain each integral ideal I of D is finitely generated.
Hence

I = α1 , . . . , αk ,

for some α1 , . . . , αk ∈ D, and thus

1 1 α1 αk
A= I = α1 , . . . , αk = , . . . , ;
γ γ γ γ
that is, every fractional ideal A of D is also finitely generated.
It is easily verified that if A and B are fractional ideals of D so are A + B and
AB. We note that if γ and δ are common denominators for A and B respectively,
then γ δ is a common denominator for both A + B and AB.

Definition 8.2.2 (The set P̃ for a prime ideal P) Let D be an integral domain
and let K be the quotient field of D. For each prime ideal P of D we define the set
P̃ by

P̃ = {α ∈ K : α P ⊆ D}.

Theorem 8.2.3 Let D be an integral domain and let P be a prime ideal of D. Then
P̃ is a fractional ideal of D.

Proof: If α ∈ P̃ and β ∈ P̃ then α P ⊆ D and β P ⊆ D. Hence (α + β)P ⊆ α P +

β P ⊆ D, so that α + β ∈ P̃.
If α ∈ P̃ and r ∈ D then α P ⊆ D and thus r α P ⊆ D, so that r α ∈ P̃.
Take π ∈ P \ {0}. For any α ∈ P̃ we have α P ⊆ D so that in particular απ ∈ D.
Hence π P̃ ⊆ D.
Thus the three properties in Definition 8.2.1 hold, showing that P̃ is a fractional
ideal of D.

Theorem 8.2.4 Let D be a Dedekind domain. Let P be a prime ideal of D. Then

P P̃ = D.

Proof: We first show that

P P̃ = D or P P̃ = P.
CB609-08 CB609/Alaca & Williams August 7, 2003 16:49 Char Count= 0

198 Dedekind Domains

As P̃ and P are both fractional ideals of D so is P P̃. Clearly P P̃ ⊆ D so that

P P̃ is an integral ideal of D. As 1 ∈ P̃ we have P ⊆ P P̃. Since P is a prime ideal
and D is a Dedekind domain, P is a maximal ideal. Thus P P̃ = P or P P̃ = D.
Next we show that D ⊂ P̃. If α ∈ D then α P ⊆ D so that α ∈ P̃. Hence D ⊆ P̃.
To prove that D ⊂ P̃ we show that P̃ contains an element γ of K that does not lie in
D. Let β ∈ P \ {0}. By Theorem 8.2.2 there exist prime ideals P1 , . . . , Pk (k ≥ 1)
with
β ⊇ P1 · · · Pk .

We choose k to be the least positive integer for which such an inclusion holds. Since
P1 · · · Pk ⊆ β ⊆ P

and P is a prime ideal, we have

Pi ⊆ P, for some i ∈ {1, 2, . . . , k}.

Relabeling P1 as Pi and Pi as P1 , if necessary, we may suppose that

P1 ⊆ P.

But as D is a Dedekind domain, P1 is a maximal ideal, and so

P1 = P.

We now consider two cases according as k = 1 or k ≥ 2. If k = 1 then

P = P1 = β.

As β = 0 we can define γ = 1/β ∈ K . Suppose γ ∈ D. Then β is a unit of D and

P = β = D, contradicting that P is a prime ideal. Hence γ ∈ D. Also,
1
γ P = β = 1 = D,
β
so that γ ∈ P̃. Hence γ ∈ P̃ \ D in this case. If k ≥ 2 then by the minimality of k
we have
P2 · · · Pk ⊆ β.

Hence there exists δ ∈ P2 · · · Pk but δ ∈ β. As β = 0 we can define γ = δ/β ∈ K .

As δ ∈ β, we see that γ = δ/β ∈ D. However,
Pδ = P1 δ ⊆ P1 · · · Pk ⊆ β,

Pγ = Pδ/β ⊆ D

and thus γ ∈ P̃. Hence γ ∈ P̃ \ D in this case. This completes the proof that
D ⊂ P̃.
CB609-08 CB609/Alaca & Williams August 7, 2003 16:49 Char Count= 0

8.2 Ideals in a Dedekind Domain 199

Finally, we show that

P P̃ = D.

Recall that we have shown that P P̃ = P or P P̃ = D. We show that P P̃ = P, so

that we must have P P̃ = D. Suppose that P P̃ = P. We show that P̃ is closed under
multiplication. Let α ∈ P̃ and β ∈ P̃. Then α P ⊆ P P̃ = P and β P ⊆ P P̃ = P.
Thus
αβ P ⊆ α P ⊆ P,

showing that αβ ∈ P̃. Hence P̃ is closed under multiplication. This proves that P̃ is
an integral domain, which strictly contains D. As D is a Noetherian domain, all its
ideals (integral or fractional) are finitely generated. Hence P̃ is a finitely generated
fractional ideal of D. Thus P̃ is a finitely generated D-module. Hence, by the
remark following Theorem 4.1.4, P̃ is integral over D. However, D is integrally
closed in its quotient field (since D is a Dedekind domain) so that D = P̃. This
contradicts that P̃ ⊃ D. Hence P P̃ = D.

Example 8.2.3 Let

√
D = Z + Z 6.
√
As D is the ring of integers of K = Q( 6), D is a Dedekind domain with quotient
field K . Let
√
P = 2, 6.
√
It is easily checked that P is a prime ideal of D with P = 2Z + Z 6. Then
P̃ = {α ∈ K | α P ⊆ D}
√ √ √ √
= {x + y 6 | x, y ∈ Q, (x + y 6)2, 6 ⊆ Z + Z 6}
√ √ √ √ √ √
= {x + y 6 | x, y ∈ Q, 2(x + y 6) ∈ Z + Z 6, (x + y 6) 6 ∈ Z + Z 6}
√
= {x + y 6 | 2x ∈ Z, 2y ∈ Z, x ∈ Z, 6y ∈ Z}
√
= {x + y 6 | x ∈ Z, 2y ∈ Z}
n√
= {m + 6 | m, n ∈ Z}
2 √
2m + n 6
={ | m, n ∈ Z}
2
1 √
= {2m + n 6 | m, n ∈ Z}
2
1 √
= (2Z + Z 6)
2
1
= P.
2
CB609-08 CB609/Alaca & Williams August 7, 2003 16:49 Char Count= 0

200 Dedekind Domains

8.3 Factorization into Prime Ideals

We now use Theorem 8.2.4 to prove the fundamental property of a Dedekind domain
D, namely, that every proper integral ideal of D can be expressed uniquely (up to
order) as a product of prime ideals.

Theorem 8.3.1 If D is a Dedekind domain every integral ideal (= 0, D) is a

product of prime ideals and this factorization is unique in the sense that if
P1 P2 · · · Pk = Q 1 Q 2 · · · Q l ,
where the Pi and Q j are prime ideals, then k = l and after relabeling (if necessary)
Pi = Q i , i = 1, 2, . . . , k.

Proof: Suppose there exist integral ideals (= 0, D) of D that are not products of
prime ideals. As D is a Dedekind domain, it is Noetherian, and so by the maximal
principle (Theorem 3.1.3) there is an ideal A(= 0, D) of D maximal with respect
to the property of not being a product of prime ideals. By Theorem 8.2.1 there exist
prime ideals P1 , . . . , Pk (k ≥ 1) of D such that
P1 · · · Pk ⊆ A.
Let k be the smallest positive integer for which such a product exists. If k = 1 then
P1 ⊆ A ⊂ D. As P1 is a prime ideal, it is a maximal ideal since D is a Dedekind
domain. Thus A = P1 . This is impossible as A is not a product of prime ideals.
Hence k ≥ 2. By Theorem 8.2.4 we have P̃ 1 P1 = D so that
P̃ 1 P1 P2 · · · Pk = D P2 · · · Pk .
Hence
P̃ 1 A ⊇ P̃ 1 P1 · · · Pk = P2 · · · Pk .
From the proof of Theorem 8.2.4 we have D ⊂ P̃ 1 so that A ⊆ P̃ 1 A. If A = P̃ 1 A
then
A ⊇ P2 · · · Pk ,
which contradicts the minimality of k as k − 1 ≥ 1. Hence A ⊂ P̃ 1 A. Since P̃ 1 A
is an ideal of D, by the maximality property of A, we have
P̃ 1 A = Q 2 · · · Q h
for prime ideals Q 2 , . . . , Q h . Then
A = AD = A P̃ 1 P1 = P1 Q 2 · · · Q h
is also a product of prime ideals, which contradicts the way A was chosen. Hence
every ideal (= 0, D) of D is a product of prime ideals.
CB609-08 CB609/Alaca & Williams August 7, 2003 16:49 Char Count= 0

8.3 Factorization into Prime Ideals 201

Suppose now that factorization of ideals as products of prime ideals is not always
unique. By the maximal principle we may choose B to be an ideal (= 0, D)
maximal with respect to the property of having at least two distinct factorizations
as the product of prime ideals, say,
B = P1 · · · Pk = Q 1 · · · Q l ,
where P1 , . . . , Pk , Q 1 , . . . Q l are prime ideals. Then, as
P1 · · · Pk ⊆ Q 1 ,
and Q 1 is a prime ideal, by Theorem 1.6.1 we have
Pi ⊆ Q 1
for some i ∈ {1, 2, . . . , k}. Relabeling P1 as Pi and vice versa, we may suppose
that
P1 ⊆ Q 1 .
Since P1 is a prime ideal, it is a maximal ideal as D is a Dedekind domain, and thus
P1 = Q 1 .
Therefore
B P̃ 1 = P̃ 1 P1 P2 · · · Pk = P2 · · · Pk
and
B P̃ 1 = B Q̃ 1 = Q̃ 1 Q 1 · · · Q h = Q 2 · · · Q h .
If B P̃ 1 = B then B P̃ 1 P1 = B P1 , so B = B P1 . Define the fractional ideal B̃ of D
by
B̃ = P̃ 1 · · · P̃ k .
Then
B B̃ = P1 · · · Pk P̃ 1 · · · P̃ k = P1 P̃ 1 · · · Pk P̃ k = D
so that
D = B B̃ = B P̃ 1 B̃ = P1 ,
which is false as P1 (being a prime ideal) is a proper ideal of D. Hence B P̃ 1 = B.
As D ⊂ P̃ 1 we have B ⊆ B P̃ 1 . But B P̃ 1 = B, so we must have
B ⊂ B P̃ 1 .
Since B P̃ 1 is an ideal of D strictly containing B, by the maximality of B,
B P̃ 1 has exactly one factorization as a product of prime ideals. Thus from
B P̃ 1 = P2 · · · Pk = Q 2 · · · Q h we deduce that k − 1 = h − 1 (that is, k = h) and
CB609-08 CB609/Alaca & Williams August 7, 2003 16:49 Char Count= 0

202 Dedekind Domains

after relabeling, we obtain Pi = Q i (i = 2, . . . , k). This implies that the two fac-
torizations of B into prime ideals are the same, which is a contradiction.
This completes the proof of the theorem.

Theorem 8.3.2 Let K be an algebraic number field. Then every proper integral
ideal of O K can be expressed uniquely up to order as a product of prime ideals.

Proof: This follows immediately from Theorems 8.1.1 and 8.3.1.

Example 8.3.1 Let

√
D = {a + b −5 | a, b ∈ Z}.
√
As D = O K , where K = Q( −5), D is a Dedekind domain. D is not a unique
factorization domain as
√ √
6 = 2 · 3 = (1 + −5)(1 − −5),
√ √
where 2, 3, 1 + −5, and 1 − −5 are nonassociated irreducibles of D. We show
how the use of prime ideals restores unique factorization. We let
√
P = 2, 1 + −5,
√
P1 = 3, 1 + −5,
√
P2 = 3, 1 − −5.

Then
√ √ √
P = 2, 1 + −5 = 2, 1 + −5, 2 − (1 + −5)
√ √ √ √
= 2, 1 + −5, 1 − −5 = 2, 2 − (1 − −5), 1 − −5
√
= 2, 1 − −5,
√ √ √
P 2 = 2, 1 + −52 = 2, 1 + −52, 1 − −5
√ √
= 4, 2(1 + −5), 2(1 − −5), 6
√ √
= 22, 1 + −5, 1 − −5, 3
= 21 = 2,
√ √
P1 P2 = 3, 1 + −53, 1 − −5
√ √
= 9, 3(1 + −5), 3(1 − −5), 6
√ √
= 33, 1 + −5, 1 − −5, 2
= 31 = 3,
√ √
P P1 = 2, 1 + −53, 1 + −5
√ √ √
= 6, 2(1 + −5), 3(1 + −5), (1 + −5)2
√ √ √
= 1 + −51 − −5, 2, 3, 1 + −5
√ √
= 1 + −51 = 1 + −5,
CB609-08 CB609/Alaca & Williams August 7, 2003 16:49 Char Count= 0

8.3 Factorization into Prime Ideals 203

√ √
P P2 = 2, 1 + −53, 1 − −5
√ √
= 2, 1 − −53, 1 − −5
√ √ √
= 6, 2(1 − −5), 3(1 − −5), (1 − −5)2
√ √ √
= 1 − −51 + −5, 2, 3, 1 − −5
√
= 1 − −51
√
= 1 − −5.

Hence
√ √
2 = P 2 , 3 = P1 P2 , 1 + −5 = P P1 , 1 − −5 = P P2 ,

and
√ √
6 = 23 = 1 + −51 − −5 = P 2 P1 P2 .

It is known from Exercises 20 and 21 of Chapter 1 that P, P1 , and P2 are distinct

prime ideals.

If A is a proper integral ideal of a Dedekind domain D then Theorem 8.3.1 tells

us that we can express A uniquely (apart from order) in the form

A = Q1 · · · Qh ,

where Q 1 , . . . , Q h are prime ideals. Let P1 , . . . , Pn denote the distinct prime ide-
als among Q 1 , . . . , Q h . Suppose that Pi (i = 1, 2, . . . , n) occurs ai times among
Q 1 , . . . , Q h so that each ai ≥ 1 (i = 1, 2, . . . , n) and a1 + a2 + · · · + an = h. Then

A = P1a1 · · · Pnan ,

where a1 , . . . , an are positive integers. Clearly this representation of A is unique.

We extend the factorization A = P1a1 · · · Pnan to allow the possibility A = 1 = D
by taking a1 = · · · = an = 0 in this case; that is, D = 1 is regarded as the unique
empty product of prime ideals. With this convention every nonzero integral ideal
of a Dedekind domain can be expressed uniquely as a product of powers of prime
ideals.
Let A and B be nonzero integral ideals of a Dedekind domain D. Then AB is a
nonzero integral ideal of D. Let P1 , . . . , Pn denote the distinct prime ideals of D
that occur in the prime ideal factorizartions of at least one of A, B, and AB. Then

n
n
n
A= Piai , B= Pibi , AB = Pici ,
i=1 i=1 i=1
CB609-08 CB609/Alaca & Williams August 7, 2003 16:49 Char Count= 0

204 Dedekind Domains

where we have grouped together all equal prime ideal factors so that ai , bi , ci (i =
1, 2, . . . , n) are nonnegative integers. Hence

n
n
n
n
Pici = AB = Piai Pibi = Piai +bi ,
i=1 i=1 i=1 i=1

so that by Theorem 8.3.1

ci = ai + bi , i = 1, 2, . . . , n.
n n n
Hence, if A = i=1 Piai and B = i=1 Pibi , then AB = i=1 Piai +bi .

Definition 8.3.1 (Divisibility of integral ideals) Let D be a Dedekind domain. Let

A and B be nonzero integral ideals of D. We say that A divides B, written A | B,
if there exists an integral ideal C of D such that B = AC.
n n
If A = i=1 Piai and B = i=1 Pibi , where P1 , . . . , Pn are distinct prime ideals
and a1 , . . . , an , b1 , . . . , bn are nonnegative integers, then
A | B ⇐⇒ ai ≤ bi , i = 1, 2, . . . , n.

We now wish to extend the representation of integral ideals as products of prime

ideals to fractional ideals; in this case negative as well as zero and positive exponents
of the prime ideals will occur.
Let A be a nonzero fractional ideal of the Dedekind domain D. Let α ∈ D \ {0}
and β ∈ D \ {0} be any two common denominators for A. Then
αA = B, βA = C,

where B and C are nonzero integral ideals of D. Suppose that

n
n
α = Piri , B= Pisi ,
i=1 i=1

n
n
β = Piti , C = Piu i ,
i=1 i=1

where P1 , . . . , Pn are distinct prime ideals and ri , si , ti , u i (i = 1, 2, . . . , n) are

nonnegative integers. Then as
αC = α(βA) = β(αA) = βB

we have

n
n
Piri +u i = Pisi +ti ,
i=1 i=1

so that by Theorem 8.3.1

ri + u i = si + ti , i = 1, 2, . . . , n.
CB609-08 CB609/Alaca & Williams August 7, 2003 16:49 Char Count= 0

8.3 Factorization into Prime Ideals 205

Hence we can define the prime ideal factorization of the fractional ideal A to be

n
A= Pisi −ri
i=1

and this definition is a valid one since it is independent of the choice of common
denominator of A. With this notation, as P P̃ = 1 for any prime ideal P of D, we
have
P̃ = P −1 .
If P1 , . . . , Pn are prime ideals such that

n
n
Piai = Pibi ,
i=1 i=1

where ai , bi (i = 1, 2, . . . , n) are integers (positive, negative, or zero), then mul-

n
tiplying both sides by i=1 PiM , where M is an integer such that ai + M > 0
and bi + M > 0 for all i, and appealing to Theorem 8.3.1, we deduce that
ai + M = bi + M (i = 1, 2, . . . , n), that is, ai = bi (i = 1, 2, . . . , n). Hence the
representation of a nonzero fractional ideal as a product of prime ideals is unique.
The following result is now clear.

Theorem 8.3.3 The set of all nonzero integral and fractional ideals of a Dedekind
domain D forms an Abelian group with respect to multiplication. The identity
n
element of the group is 1 = D and the inverse of A = i=1 Piai , where P1 , . . . , Pn
are distinct prime ideals and a1 , . . . , an are integers (positive, negative, or zero),
is

n
A−1 = Pi−ai .
i=1

Theorem 8.3.4 Let K be an algebraic number field. Let O K be the ring of integers
of K . Then the set of all nonzero integral and fractional ideals of O K forms an
Abelian group I (K ) with respect to multiplication.

Proof: This follows immediately from Theorems 8.1.1 and 8.3.3.

Example 8.3.2 With the notation of Example 8.2.3 we have

√
P = 2, 6
and
1
P̃ = P.
2
CB609-08 CB609/Alaca & Williams August 7, 2003 16:49 Char Count= 0

206 Dedekind Domains

Thus
1
P −1 = P.
2
We check this directly. We have

1 1 1 √
P P = P 2 = 2, 62
2 2 2
1 √ √
= 4, 2 6, 6 = 2, 6, 3 = 1,
2
√
as 1 = 3 − 2 ∈ 2, 6, 3. This shows that P −1 = 12 P.
√
Example
√ 8.3.3 Let D = Z + Z 6. We determine the inverse A−1 of the ideal
A = 6 of D, illustrating the ideas of this section. Let
√ √
P = 2, 6, Q = 3, 6.

P and Q are distinct prime ideals of D such that

√ √ √ √ √
P Q = 6, 2 6, 3 6, 6 = 6 6, 2, 3 = 6 = A.

Thus A = P Q is the prime ideal factorization of A and so

A−1 = P −1 Q −1 ,

where
1
P −1 = P̃ = P (Example 8.3.2)
2
and
1
Q −1 = Q̃ = Q.
3
Thus

−1 1 1 1 1
A = P Q = P Q = A.
2 3 6 6
This is clear as A( 16 A) = 16 A2 = 16 6 = 1.

8.4 Order of an Ideal with Respect to a Prime Ideal

Let A be a nonzero fractional or integral ideal of a Dedekind domain D. Then A
can be written uniquely in the form

n
A= Piai ,
i=1
CB609-08 CB609/Alaca & Williams August 7, 2003 16:49 Char Count= 0

8.4 Order of an Ideal 207

where the Pi are distinct prime ideals and the ai are integers (positive, negative, or
zero).

Definition 8.4.1 (Order of an ideal with respect to a prime ideal) With the
preceding notation, the order of the nonzero ideal A of the Dedekind domain D
with respect to the prime ideal Pi (i = 1, 2, . . . , n), written ord Pi (A), is defined
by

ord Pi (A) = ai .

For any prime ideal P = P1 , . . . , Pn we define

ord P (A) = 0.

Clearly ord P (1) = 0 and ord P (P k ) = k for all prime ideals P.

√ √
Example 8.4.1 Let D = Z + Z 6. Let B be the ideal 12, 6 6. Then, with the
notation of Example 8.3.3, we have
√ √ √ √
B = 12, 6 6 = 62, 6 = 62 2, 6
= A2 P = (P Q)2 P = P 3 Q 2 ,

so that

ord P (B) = 3, ord Q (B) = 2.

We now extend the concept of divisibility from integral ideals (Definition 8.3.1)
to fractional ideals.

Definition 8.4.2 (Divisibility of fractional ideals) Let D be a Dedekind domain.

Let A and B be nonzero fractional ideals of D. We say that A divides B, written
A | B, if there exists an integral ideal C of D such that B = AC.

Clearly if A and B are nonzero fractional or integral ideals of a Dedekind domain,

we have

A | B ⇐⇒ ord P (A) ≤ ord P (B) for all prime ideals P.

√
Example 8.4.2 Let D = Z + Z 6. Let A and B be the fractional ideals of D
given by
√
3 6 3√
A= , , B = 3, 6.
2 2 2
CB609-08 CB609/Alaca & Williams August 7, 2003 16:49 Char Count= 0

208 Dedekind Domains

√ √
We show that A | B. Let P = 2, 6 and Q = 3, 6. P and Q are distinct prime
ideals of D (Example 8.3.3). We have, as P 2 = 2 and Q 2 = 3,
1 √ 1
A = 3, 6 = Q = P −2 Q
2 2
and
3 √
B = 2, 6 = P −1 Q 2 .
2
Since ord P A = −2 < −1 = ord P B and ord Q A = 1 < 2 = ord Q B, we see that
A | B. Indeed B = AC with C = P Q.

The next theorem gives a necessary and sufficient condition for an ideal A to
divide an ideal B. It is usually remembered as “To contain is to divide.”

Theorem 8.4.1 Let D be a Dedekind domain. Let A and B be nonzero integral or

fractional ideals of D. Then

A | B if and only if A ⊇ B.

Proof: As A is a nonzero integral or fractional ideal of D, A−1 is a nonzero integral

or fractional ideal of D. Thus B A−1 is a nonzero integral or fractional ideal of D.
Then

A ⊇ B ⇐⇒ A A−1 ⊇ B A−1
⇐⇒ D ⊇ B A−1
⇐⇒ B A−1 is an integral ideal of D
⇐⇒ B A−1 = C for some integral ideal C of D
⇐⇒ B = AC for some integral ideal C of D
⇐⇒ A | B.

The two basic properties of the function ord P (A) are given in the next theorem.

Theorem 8.4.2 Let D be a Dedekind domain. Let P be a prime ideal of D. Let A

and B be nonzero integral or fractional ideals of D. Then

(a) ord P (AB) = ord P (A) + ord P (B),

(b) ord P (A + B) = min(ord P (A), ord P (B)).

Proof: (a) We have

A= P ord P (A) , B = P ord P (B) ,
P P
CB609-08 CB609/Alaca & Williams August 7, 2003 16:49 Char Count= 0

8.4 Order of an Ideal 209

where the products are taken over all prime ideals P of D, so that

P ord P (AB) = AB = P ord P (A)+ord P (B) .
P P

Of course only finitely many of the exponents in the products are nonzero. Hence,
by the uniqueness property, we have

ord P (AB) = ord P (A) + ord P (B)

for all prime ideals P of D.

(b) Set C = A + B. As A and B are nonzero ideals, so is C. Then

AC −1 + BC −1 = (A + B)C −1 = CC −1 = D.

Hence AC −1 ⊆ AC −1 + BC −1 = D and BC −1 ⊆ AC −1 + BC −1 = D. So AC −1
and BC −1 are both integral ideals of D. Suppose AC −1 ⊆ P and BC −1 ⊆ P. Then

D = AC −1 + BC −1 ⊆ P + P = P,

which is impossible. Thus either AC −1 ⊆ P or BC −1 ⊆ P; that is, by Theorem

8.4.1, P AC −1 or P BC −1 , so

min(ord P (AC −1 ), ord P (BC −1 )) = 0.

Finally, by part (a), we obtain

min(ord P (A), ord P (B)) = ord P (C) = ord P (A + B).

We next define the order of a nonzero element with respect to a prime ideal.

Definition 8.4.3 (Order of a nonzero element with respect to a prime ideal) Let
D be a Dedekind domain with quotient field K . For α ∈ K , α = 0, we define

ord P (α) = ord P (α)

for any prime ideal P of D.

The next theorem allows us to recognize when an element α belongs to an ideal

A of D in terms of the orders of α and A with respect to prime ideals P.

Theorem 8.4.3 Let D be a Dedekind domain with quotient field K . Let A be a

nonzero ideal of D. Let α ∈ K , α = 0. Then

α ∈ A if and only if ord P (α) ≥ ord P (A) for all prime ideals P of D.
CB609-08 CB609/Alaca & Williams August 7, 2003 16:49 Char Count= 0

210 Dedekind Domains

Proof: We have

α ∈ A ⇐⇒ α ⊆ A
⇐⇒ A | α (by Theorem 8.4.1)
⇐⇒ ord P (A) ≤ ord P (α) for all prime ideals P of D
⇐⇒ ord P (α) ≥ ord P (A) for all prime ideals P of D.

The next theorem gives the basic properties of the order of an element with
respect to a prime ideal.

Theorem 8.4.4 Let D be a Dedekind domain with quotient field K . Let P be a

prime ideal of D.

(a) For α ∈ K ∗ and β ∈ K ∗

ord P (αβ) = ord P (α) + ord P (β).

(b) For α, β, α + β ∈ K ∗ ,
ord P (α + β) ≥ min(ord P (α), ord P (β)).

(c) If α, β, α + β ∈ K ∗ are such that ord P (α) = ord P (β) then

ord P (α + β) = min(ord P (α), ord P (β)).

Proof: (a) We have for any prime ideal P of D

ord P (αβ) = ord P (αβ)

= ord P (αβ)
= ord P (α) + ord P (β)
= ord P (α) + ord P (β).

(b) As α + β ∈ α + β we have by Theorems 8.4.2(b) and 8.4.3

ord P (α + β) ≥ ord P (α + β)

= min(ord P (α), ord P (β))
= min(ord P (α), ord P (β)).

(c) Without loss of generality we may suppose that

ord P (α) > ord P (β).

Then by part (b) we have

ord P (α + β) ≥ ord P (β).

CB609-08 CB609/Alaca & Williams August 7, 2003 16:49 Char Count= 0

8.4 Order of an Ideal 211

Thus

ord P (β) = ord P ((α + β) − α)

≥ min(ord P (α + β), ord P (α))
= ord P (α + β) (as ord P (α) > ord P (β))
≥ ord P (β).

Hence

ord P (α + β) = ord P (β) = min(ord P (α), ord P (β)).

Example 8.4.3 We give a simple example to show that if ord P (α) = ord P (β)
then ord P (α + β) may actually be larger than ord P (α). Take D = Z, α = 1, β =
4, P = 5. Then

ord P (α) = 0, ord P (β) = 0, ord P (α + β) = 1.

Theorem 8.4.5 Let D be a Dedekind domain with quotient field K . Given any finite
set of prime ideals P1 , . . . , Pk of D and a corresponding set of integers a1 , . . . , ak
then there exists α ∈ K such that

ord Pi (α) = ai , i = 1, 2, . . . , k,

and

ord P (α) ≥ 0, for any prime ideal P = P1 , . . . , Pk .

Proof: As

k
k
P1a1 Piai +1 | P1a1 +1 Piai +1 ,
i=2 i=2

by Theorem 8.4.1 we have

k
k
P1a1 Piai +1 ⊇ P1a1 +1 Piai +1 .
i=2 i=2

By the uniqueness property we have

k
k
P1a1 Piai +1 = P1a1 +1 Piai +1 ,
i=2 i=2

so that

k
k
P1a1 Piai +1 ⊃ P1a1 +1 Piai +1 .
i=2 i=2
CB609-08 CB609/Alaca & Williams August 7, 2003 16:49 Char Count= 0

212 Dedekind Domains

Hence there exists

k
k
α1 ∈ P1a1 Piai +1 , α1 ∈
/ P1a1 +1 Piai +1 .
i=2 i=2

Thus
ord P1 (α1 ) = a1
and
ord Pi (α1 ) ≥ ai + 1 for i = 1.
Similarly we can define α j ∈ K for j = 2, . . . , k such that
ord P j (α j ) = a j
and
ord Pi (α j ) ≥ ai + 1 for i = j.
Now set
α = α1 + α2 + · · · + αk ∈ K .
Then, by Theorem 8.4.4(b), we have
ord P1 (α2 + · · · + αk ) ≥ min(ord P1 (α2 ), . . . , ord P1 (αk )) ≥ a1 + 1 > ord P1 (α1 ),
so that
ord P1 (α) = ord P1 (α1 + (α2 + · · · + αk ))
= min(ord P1 (α1 ), ord P1 (α2 + · · · + αk ))
= ord P1 (α1 ),
that is,
ord P1 (α) = a1 .
Similarly,
ord P j (α) = a j , j = 2, . . . , k.
Finally, for P = P1 , . . . , Pk we have
ord P (αi ) ≥ 0, i = 1, 2, . . . , k,
so that
ord P (α) ≥ 0.

If D is a Dedekind domain with quotient field K , A is a nonzero fractional or

integral ideal of D, and a, b, c ∈ A, then we write
a ≡ b (mod A) if and only if A | a − b.
CB609-08 CB609/Alaca & Williams August 7, 2003 16:49 Char Count= 0

8.4 Order of an Ideal 213

We observe that

A | a − b ⇐⇒ a − b ⊆ A ⇐⇒ a − b ∈ A ⇐⇒ a + A = b + A.

The properties

a ≡ a (mod A),
a ≡ b (mod A) =⇒ b ≡ a (mod A),
a ≡ b (mod A), b ≡ c (mod A) =⇒ a ≡ c (mod A),
a ≡ b (mod A) =⇒ ac ≡ bc (mod A)

are easily proved.

Theorem 8.4.6 (Chinese remainder theorem) Let D be a Dedekind domain.

(a) Let P1 , . . . , Pk be distinct prime ideals in D. Let a1 , . . . , ak be positive integers. Let

α1 , . . . , αk be elements of D. Then there exists α ∈ D such that
α ≡ αi (mod Piai ), i = 1, 2, . . . , k.

(b) Let I1 , . . . , Ik be pairwise relatively prime ideals of D. Let α1 , . . . , αk be elements of

D. Then there exists α ∈ D such that
α ≡ αi (mod Ii ), i = 1, 2, . . . , k.

Proof: (a) Consider the ideal

Q 1 = P1a1 + P2a2 · · · Pkak

of D. Suppose P is a prime ideal such that

P | Q1.

Now

P1a1 ⊆ Q 1 , P2a2 · · · Pkak ⊆ Q 1 ,

so, by Theorem 8.4.1, we have

Q 1 | P1a1 , Q 1 | P2a2 · · · Pkak .

Hence

P | P1a1 , P | P2a2 · · · Pkak .

From the first of these we deduce that P | P1 so that P = P1 . Hence

P1 | P2a2 · · · Pkak ,
CB609-08 CB609/Alaca & Williams August 7, 2003 16:49 Char Count= 0

214 Dedekind Domains

which contradicts that P1 is a prime ideal distinct from P2 , . . . , Pk . Thus there is

no prime ideal dividing Q 1 . Hence Q 1 = D; that is

P1a1 + P2a2 · · · Pkak = 1.

Hence there exist x1 ∈ P1a1 and y1 ∈ P2a2 · · · Pkak such that

x1 + y1 = 1.

Thus

y1 ≡ 1 (mod P1a1 ), y1 ≡ 0 (mod Piai ), i = 2, . . . , k.

Similarly, for j = 2, . . . , k we can find y j such that

a
y j ≡ 1 (mod P j j ), y j ≡ 0 (mod Piai ), i = j.

Now let

α = α1 y1 + · · · + αk yk ∈ D.

Then as α2 y2 + · · · + αk yk ∈ P1a1 we have

α ≡ α1 y1 ≡ α1 (mod P1a1 ).

Similarly,
a
α ≡ α j (mod P j j ), j = 2, . . . , k.

This completes the proof of part (a).

(b) Part (b) follows from part (a) by observing that any congruence of the form
x ≡ α (mod I ) is equivalent to the system of congruences x ≡ α (mod Piai ), i =
1, 2, . . . , r , where I = P1a1 · · · Prar .

Example 8.4.4 Let D = Z[x]. D is not a Dedekind domain as the prime ideal x
is not a maximal ideal (Example 1.5.6). Consider the pair of congruences

α ≡ 0 (mod 2),
α ≡ 1 (mod x).

The moduli 2 and x are distinct prime ideals. However, the congruences are not
simultaneously solvable in D, since any solution of α ≡ 0 (mod 2) has an even
constant term, whereas any solution of α ≡ 1 (mod x) has a constant term equal
to 1. This shows that the Chinese remainder theorem does not necessarily hold in
an integral domain that is not a Dedekind domain.
CB609-08 CB609/Alaca & Williams August 7, 2003 16:49 Char Count= 0

8.5 Generators of Ideals in a Dedekind Domain 215

8.5 Generators of Ideals in a Dedekind Domain

In this section we show that every fractional or integral ideal of a Dedekind domain
is generated by at most two elements.

Theorem 8.5.1 Let D be a Dedekind domain. Let A be a fractional or integral

ideal of D. Then A is generated by at most two elements.

Proof: If A = {0} then A = 0, and if A = D then A = 1, so that we may

suppose that A = {0}, D. Let β ∈ A, β = 0, and β = unit. Then β ⊆ A so that
A | β. Hence there exists a nonzero integral ideal B of D such that
β = AB.
Let P1 , . . . , Pn be the set of distinct prime ideals for which either
ord Pi (A) = 0 or ord Pi (AB) = 0 (or both).
This set is nonempty as A = D. By Theorem 8.4.5 there exists α ∈ K (the quotient
field of D) such that
ord Pi (α) = ord Pi (A), i = 1, 2, . . . , n,
ord P (α) ≥ 0, P = P1 , . . . , Pn .
For P = P1 , . . . , Pn we have ord P (A) = 0 so that
ord P (α) ≥ ord P (A) for all prime ideals P.
Hence
α ∈ A.
For i = 1, 2, . . . , n we have
ord Pi (A) = min(ord Pi (A), ord Pi (AB)) (as B is an integral ideal)
= min(ord Pi (α), ord Pi (AB))
= min(ord Pi (α), ord Pi (AB))
= ord Pi (α + AB),
by Theorem 8.4.2(b). For P = P1 , . . . , Pn we have ord P (A) = ord P (AB) = 0, so
that
ord P (A) = min(ord P (α), ord P (AB))
= min(ord P (α), ord P (AB))
= ord P (α + AB)
by Theorem 8.4.2(b). Hence
ord P (A) = ord P (α + AB) for all prime ideals P.
CB609-08 CB609/Alaca & Williams August 7, 2003 16:49 Char Count= 0

216 Dedekind Domains

Hence

A = α + AB.

Finally,

A = α + β = α, β.

Exercises
1. Let D be a Dedekind domain. Let A and B be integral ideals of D with A = D, B = D.
Prove from first principles that AB = D.
2. Let D be a Dedekind domain. Let A be a nonzero integral ideal of D. Let P be a prime
ideal of D. If P does not divide A prove that ord P (A) = 0.
√
3. Determine all fractional ideals of Z + Z −1.
√
4. Find all ideals in Z + Z −6 that contain 6.
5. Let D be a principal ideal domain with quotient field K . Prove that every fractional
ideal of D is of the form {dα | d ∈ D} for some α ∈ K .
6. Let K be an algebraic number field of degree n. Let a be a nonzero rational integer.
Prove that a belongs to at most a n integral ideals of O K .
√ √ √
7. Show that 3, 1 + 2 −5 | 1 + 2 −5 in O K , where K = Q( −5). Determine an
integral ideal A such that
√ √
1 + 2 −5 = 3, 1 + 2 −5A.
√ √
8. Determine the fractional ideal 3, 1 + 2 −5−1 of O K , where K = Q( −5).
9. Let K be an algebraic number field. Let I be an integral ideal of O K . Let a ∈ I . Prove
that there exists an integral ideal I of O K such that a = I I .
10. Let K be an algebraic number field. Let I be a nonzero integral ideal of O K . Let α ∈ K
have the following property:
a ∈ I =⇒ aα ∈ I.
Prove that α ∈ O K .
√ √ √ √
11. Let I be the ideal of Z + Z −5 generated by 1 + −5, 3 + −5, and 19 + 9 −5.
√
Determine α, β ∈ Z + Z −5 such that I = α, β.
12. Let I and J be nonzero integral ideals of a Dedekind domain D. Let P1 , . . . , Pk be the
distinct prime ideals dividing either I or J (or both) so that
I = P1a1 · · · Pkak , J = P1b1 · · · Pkbk ,
for nonnegative integers a1 , . . . , ak , b1 , . . . , bk . The greatest common divisor gcd (I, J )
and the least common multiple lcm (I, J ) of I and J are defined by
gcd(I, J ) = P1min(a1 ,b1 ) · · · Pkmin(ak ,bk ) ,
lcm(I, J ) = P1max(a1 ,b1 ) · · · Pkmax(ak ,bk ) .
Prove that
gcd (I, J ) = I + J
CB609-08 CB609/Alaca & Williams August 7, 2003 16:49 Char Count= 0

9.1 Norm of an Integral Ideal

We have already defined (Definition 7.1.3) the norm N (A) of a nonzero integral
ideal A in the ring O K of integers of an algebraic number field K by

D(A)
N (A) = , (9.1.1)
d(K )
where D(A) is the discriminant of the ideal A (Definition 6.5.2) and d(K ) is the
discriminant of the field K (Definition 7.1.2). Two main results of this chapter are
the following:
N (A) = card(O K /A), (9.1.2)
where O K /A is the factor ring of O K by A, and
N (AB) = N (A)N (B) (9.1.3)
for any two nonzero integral ideals A and B of O K .
We require a couple of preliminary results to establish (9.1.2).

Theorem 9.1.1 Let C be an n × n matrix with rational integer entries. Then, by

applying to C a finite sequence of elementary operations of the types
(1) interchange of two rows or two columns,
(2) addition of an integral multiple of one row (or column) to another row (or column),
we can transform C into a diagonal matrix
 
d1 0 · · · 0
 0 d2 · · · 0 
 
 . .. ..  ,
 .. . ··· . 
0 0 · · · dn
where the integers d1 , d2 , . . . , dn are such that
|det C| = |d1 | · · · |dn |.

218
CB609-09 CB609/Alaca & Williams August 7, 2003 16:51 Char Count= 0

9.1 Norm of an Integral Ideal 219

Proof: The proof is by induction on the size n of the matrix C. If n = 1 the result
is clearly true. Now suppose that it is true for all (n − 1) × (n − 1) matrices with
rational integer entries. Let C be a given n × n matrix with rational integer entries.
If C is the zero matrix there is nothing to prove, so we may suppose that C = On .
Let k denote any one of the nonzero entries in C. By means of elementary operations
of type 1 we can transform C into a matrix B = (bi j ) in which b11 = k. If k does
not divide all the remaining entries in the first row and first column we can find an
integer j (2 ≤ j ≤ n) such that b1 j or b j1 = qk + r with 0 < r < |k|; then by an
elementary operation of type 2, subtracting q times the first row or column from the
jth row or column, we obtain a matrix with an entry r < |k|. Applying elementary
operations of type 1 we can move r to the (1, 1) position and repeat the process.
After a finite number of operations we obtain a matrix in which the (1, 1) entry
divides all the entries in the first row and column. Thus by means of a finite number
of operations of types 1 and 2 we can transform the matrix into one of the form
 
d1 0 ··· 0
 
0 
 
 
 . (n − 1) × (n − 1)  .
 . . submatrix with 
 
 integer entries 
0
Applying the inductive hypothesis to the (n − 1) × (n − 1) submatrix we finally
get a matrix of the required diagonal type.
Clearly, elementary operations of types 1 and 2 at most change the sign of the
determinant so that
|det C| = |d1 | · · · |dn |.

It can be shown that d1 , d2 , . . . , dn in Theorem 9.1.1 can be arranged to satisfy

d1 | d2 | · · · | dn , in which case the matrix is said to be in Smith normal form.

Theorem 9.1.2 Let G be a free Abelian group with n generators ω1 , . . . , ωn , so

that each element of G is uniquely expressible as
x1 ω1 + · · · + xn ωn , x1 , . . . , xn ∈ Z.
Let H be the subgroup of G generated by the n elements η1 , . . . , ηn so that
H = {y1 η1 + · · · + yn ηn | y1 , . . . , yn ∈ Z}.
As each ηi ∈ H ⊆ G we have

n
ηi = ci j ω j , i = 1, 2, . . . , n,
j=1
CB609-09 CB609/Alaca & Williams August 7, 2003 16:51 Char Count= 0

220 Norms of Ideals

where each ci j ∈ Z. Let C be the n × n matrix whose (i, j) entry is ci j . Then

|det C|, if det C = 0,
[G : H ] =
∞, if det C = 0.

Proof: We wish to transform the matrix C into the form given in Theorem 9.1.1 by
means of elementary operations of types 1 and 2.
An elementary operation of type 1, that is, interchanging rows or columns of C,
corresponds to rearranging the order of the generators η1 , . . . , ηn of H or ω1 , . . . , ωn
of G, and it so leaves [G : H ] unchanged.
The elementary operation of type 2, which adds k times the ith row to the lth row,
corresponds to replacing ci j by ci j + kcl j ( j = 1, 2, . . . , n) and hence replaces

n
ηi = ci j ω j
j=1

by

n
ηi + kηl = (ci j + kcl j )ω j .
j=1

But it is clear that {η1 , . . . , ηn } and {η1 , . . . , ηi + kηl , . . . , ηl , . . . , ηn } generate the

same subgroup so that again [G : H ] is unaltered.
Finally, the elementary operation of type 2, which adds k times the lth column
to the jth column, corresponds to replacing ci j by ci j + kcil (i = 1, 2, . . . , n) and
thus is equivalent to replacing the generators ω1 , . . . , ωn of G by the equivalent set
{ω1 , . . . , ωl − kω j , . . . , ωn } since

n
ηi = ci h ωh
h=1
n
= ci h ωh + ci j ω j + cil ωl
h=1
h = j, l

n
= ci h ωh + (ci j + kcil )ω j + cil (ωl − kω j )
h=1
h = j, l

n
= ci h ωh ,
h=1

where

ci h , h = j,
ci h =
ci j + kcil , h = j,
CB609-09 CB609/Alaca & Williams August 7, 2003 16:51 Char Count= 0

9.1 Norm of an Integral Ideal 221

and

ωh , h= l,
ωh =
ωl − kω j , h = l.
Thus [G : H ] remains unchanged.
Hence transforming C into diag(d1 , . . . , dn ), where |d1 | · · · |dn | = |det C|, by
elementary operations as in Theorem 9.1.1, we obtain a set of generators for G,
namely,
G = ω1 , . . . , ωn ,
such that
H = d1 ω1 , . . . , dn ωn .
Clearly
x1 ω1 + · · · + xn ωn ∈ H
⇐⇒ x1 ω1 + · · · + xn ωn = y1 d1 ω1 + · · · + yn dn ωn
for some yi ∈ Z (i = 1, 2, . . . , n)
⇐⇒ xi = yi di (i = 1, 2, . . . , n)
⇐⇒ di | xi (i = 1, 2, . . . , n).
Suppose now that det C = 0. Hence d1 · · · dn = 0 so that each di = 0 (i =
1, 2, . . . , n). Then a complete set of coset representatives for G modulo H is
{x1 ω1 + · · · + xn ωn | x1 = 0, 1, . . . , |d1 | − 1; . . . ; xn = 0, 1, . . . , |dn | − 1},
and thus
[G : H ] = |d1 | · · · |dn | = |det C|.
Finally, suppose that det C = 0. Hence d1 · · · dn = 0 so that di = 0 for some
i ∈ {1, 2, . . . , n}. Then kωi + H (k = 0, 1, 2, . . .) are distinct cosets of H in G so
that [G : H ] = ∞.

We can now prove (9.1.2).

Theorem 9.1.3 Let K be an algebraic number field with [K : Q] = n. Let O K be

the ring of integers of K . Let A be a nonzero integral ideal of O K . Then
N (A) = card(O K /A).

Proof: Let {η1 , . . . , ηn } be a basis for A and {ω1 , . . . , ωn } an integral basis for K .
Then
n
ηi = ci j ω j , i = 1, 2, . . . , n,
j=1
CB609-09 CB609/Alaca & Williams August 7, 2003 16:51 Char Count= 0

222 Norms of Ideals

for ci j ∈ Z (i, j = 1, 2, . . . , n). Thus

D(A) D(η1 , . . . , ηn )
N (A) = = = |det(ci j )| = [O K : A] = card(O K /A),
d(K ) D(ω1 , . . . , ωn )
by Theorem 9.1.2.

9.2 Norm and Trace of an Element

If K is an algebraic number field of degree n and α is an element of K then there
are two very important quantities associated with α, namely,
σ1 (α) + σ2 (α) + · · · + σn (α) and σ1 (α)σ2 (α) · · · σn (α),
where
σk : K −→ C, k = 1, 2, . . . , n,
are the n distinct monomorphisms from K to C. These quantities are called the
trace and norm of α respectively.

Definition 9.2.1 (Norm and trace of an element) Let K be an algebraic number

field of degree n. Let α ∈ K . Let α1 = α, α2 , . . . , αn be the K -conjugates of α. Then
the trace of α is denoted by tr(α) and is defined by
tr(α) = α1 + α2 + · · · + αn ,
and the norm of α is denoted by N (α) and is defined by
N (α) = α1 α2 · · · αn .

If α ∈ Q then by Theorem 6.3.4 we know that all the K -conjugates of α are all
equal to α. Hence for α ∈ Q we have
tr(α) = α + · · · + α = nα
and
N (α) = α · · · α = α n .
√
If K is a quadratic field then K = Q( m) for some squarefree integer m.
√
Let α ∈ K . Then α = r + s m for some r, s ∈ Q. The K -conjugates of α are
√ √
α = r + s m and α = r − s m. The trace of α is
tr(α) = α + α = 2r
and the norm of α is
N (α) = αα = r 2 − s 2 m.
Recalling Definition 2.2.1 we observe that φm (α) = |N (α)|.
CB609-09 CB609/Alaca & Williams August 7, 2003 16:51 Char Count= 0

9.2 Norm and Trace of an Element 223

From Definitions 6.3.2 and 9.2.1, we see that

fld K (α) = x n − tr(α)x n−1 + · · · + (−1)n N (α).
In particular when K is a quadratic field (so that n = 2) we have
fld K (α) = x 2 − tr(α)x + N (α).
From Theorem 6.3.1 we deduce that for an arbitrary algebraic number field K
tr(α) ∈ Q and N (α) ∈ Q.
Further, if α ∈ O K then, by Theorem 6.3.3, α1 = α, α2 , . . . , αn are algebraic inte-
gers so that fld(α) ∈ Z[x] and thus
tr(α) ∈ Z and N (α) ∈ Z.
In the next theorem we show that the trace is additive and the norm is multiplica-
tive.

Theorem 9.2.1 Let K be an algebraic number field of degree n. Let α, β ∈ K .

Then
tr(α + β) = tr(α) + tr(β)
and
N (αβ) = N (α)N (β).

Proof: Let σk : K −→ C (k = 1, 2, . . . , n) be the n distinct monomorphisms from

K to C. Then

n n
tr(α + β) = σk (α + β) = (σk (α) + σk (β))
k=1 k=1

n
n
= σk (α) + σk (β) = tr(α) + tr(β)
k=1 k=1
and

n
N (αβ) = σk (αβ) = (σk (α)σk (β))

k=1

k=1

n
n
= σk (α) σk (β) = N (α)N (β).
k=1 k=1

The next theorem tells us about the norm of a unit.

Theorem 9.2.2 Let K be an algebraic number field of degree n.

(a) If α is a unit of O K then N (α) = ±1.
(b) If α ∈ O K and N (α) = ±1 then α is a unit of O K .
CB609-09 CB609/Alaca & Williams August 7, 2003 16:51 Char Count= 0

224 Norms of Ideals

Proof: (a) Let α ∈ O K be a unit. Then there exists β ∈ O K such that αβ = 1.

Taking norms we obtain by Theorem 9.2.1

N (α)N (β) = N (αβ) = N (1) = 1.

As N (α) ∈ Z and N (β) ∈ Z we deduce that N (α) = ±1.

(b) Let α ∈ O K be such that N (α) = ±1. Let σk : K −→ C (k = 1, 2, . . . , n)
be the n distinct monomorphisms from K to C with σ1 = 1. Then

n
σk (α) = N (α) = ±1.
k=1

Set

n
β=± σk (α),
k=2

so that

αβ = 1.

As α ∈ O K , each σk (α) ∈ O K (k = 1, 2, . . . , n) by Theorem 6.3.3, so that β ∈ O K .

Hence α is a unit of O K .

The next theorem is often useful in showing that an algebraic integer is an

irreducible.

Theorem 9.2.3 Let K be an algebraic number field. If α ∈ O K is such that

N (α) = ± p,

where p is a rational prime, then α is an irreducible.

Proof: Suppose that α ∈ O K is such that N (α) = ± p, where p is a prime. Clearly

α = 0 as N (0) = 0. Moreover, α is not a unit as the norm of a unit is ±1 by Theorem
9.2.2. Thus if α is not irreducible then there exist nonzero, nonunit elements β and
γ of O K such that

α = βγ .

Then, by Theorem 9.2.1, we have

± p = N (α) = N (βγ ) = N (β)N (γ ).

As N (β) ∈ Z, N (γ ) ∈ Z, and p is a prime, we must have

N (β) or N (γ ) = ±1.
CB609-09 CB609/Alaca & Williams August 7, 2003 16:51 Char Count= 0

9.2 Norm and Trace of an Element 225

Hence, by Theorem 9.2.2(b), β or γ is a unit, which is a contradiction. This proves

that α is irreducible.

We emphasize
√ that the converse
√ of Theorem 9.2.3 is not true. To see this take
K =√ Q( −5) and √ α = 1 + −5. Then α is an irreducible in O K but N (α) =
(1 + −5)(1 − −5) = 6 is not prime.
In the next theorem we use Theorems 9.2.2 and 9.2.3 to give a condition that,
when satisfied by an element α of a cubic field, guarantees that α is a unit, and we
give a similar condition for α to be an irreducible.

Theorem 9.2.4 Let a and b be integers such that the cubic polynomial x 3 + ax + b
is irreducible in Z[x]. Let

K = Q(θ ), where θ 3 + aθ + b = 0,

so that K is a cubic field.

(a) If r, s, t are integers such that

r 3 − bs 3 + b2 t 3 + ar s 2 + a 2r t 2 − 2ar 2 t − abst 2 + 3br st = ±1

then r + sθ + tθ 2 is a unit in O K .
(b) If r, s, t are integers such that

r 3 − bs 3 + b2 t 3 + ar s 2 + a 2r t 2 − 2ar 2 t − abst 2 + 3br st = ± p,

where p is a prime then r + sθ + tθ 2 is an irreducible in O K .

Proof: Let θ, θ , θ be the roots of x 3 + ax + b = 0 so that

θ + θ + θ = 0,
θ θ + θ θ + θ θ = a,
θθ θ = −b.

Then
2 2
θ 2 + θ + θ = (θ + θ + θ )2 − 2(θθ + θ θ + θ θ) = −2a,
2 2 2 2
θ 2 θ + θ θ + θ θ 2 = (θ θ + θ θ + θ θ)2 − 2θθ θ (θ + θ + θ ) = a 2 ,
2 2 2 2
θθ + θ 2 θ + θ θ + θ 2 θ + θ θ + θ θ
= (θ + θ )θ θ + (θ + θ )θθ + (θ + θ )θ θ
= −θ θ θ − θ θ θ − θθ θ = −3θθ θ = 3b.
CB609-09 CB609/Alaca & Williams August 7, 2003 16:51 Char Count= 0

226 Norms of Ideals

Hence
2 2
N (r + sθ + tθ 2 ) = (r + sθ + tθ 2 )(r + sθ + tθ )(r + sθ + tθ )
= r 3 + s 3 θθ θ + t 3 (θθ θ )2 + r s 2 (θθ + θ θ + θ θ)
2 2 2 2
+ r t 2 (θ 2 θ + θ θ + θ θ 2 ) + r 2 s(θ + θ + θ )
2 2
+ r 2 t(θ 2 + θ + θ ) + s 2 tθθ θ (θ + θ + θ )
+ st 2 θθ θ (θθ + θ θ + θ θ)
2 2 2 2
+ r st(θθ + θ 2 θ + θθ + θ 2 θ + θ θ + θ θ )
= r 3 − bs 3 + b2 t 3 + ar s 2 + a 2r t 2 − 2ar 2 t − abst 2 + 3br st.

The assertions of the theorem now follow from Theorems 9.2.2 and 9.2.3.

Example 9.2.1 Let K be the cubic field given by K = Q(θ), where

θ 3 − 4θ + 2 = 0. We show that θ − 1 is a unit of O K . This is the special case

a = −4, b = 2, r = −1, s = 1, t = 0

of Theorem 9.2.4(a) as

r 3 − bs 3 + b2 t 3 + ar s 2 + a 2r t 2 − 2ar 2 t − abst 2 + 3br st

= r 3 − bs 3 + ar s 2
= −1 − 2 + 4 = 1.

Similarly, we can show that 2θ − 1 is a unit of O K since in this case

a = −4, b = 2, r = −1, s = 2, t = 0

and

r 3 − bs 3 + b2 t 3 + ar s 2 + a 2r t 2 − 2ar 2 t − abst 2 + 3br st

= r 3 − bs 3 + ar s 2
= −1 − 16 + 16 = −1.

In the next theorem we relate N (α) and N (α).

Theorem 9.2.5 Let K be an algebraic number field of degree n. Let O K be the ring
of integers of K . Let α ∈ O K . Then

N (α) = |N (α)|.

Proof: Let {ω1 , . . . , ωn } be an integral basis for K . Then {αω1 , . . . , αωn } is a

minimal basis for the principal ideal α. Hence
CB609-09 CB609/Alaca & Williams August 7, 2003 16:51 Char Count= 0

9.2 Norm and Trace of an Element 227

2
σ1 (αω1 ) ··· σ1 (αωn )

σ2 (αω1 ) ··· σ2 (αωn )

D(α) = .. .. ,
. ··· .

σn (αω1 ) ··· σn (αωn )
where σk : K −→ C (k = 1, 2, . . . , n) are the n distinct monomorphisms from K
to C. Thus

σ1 (α)σ1 (ω1 ) · · · σ1 (α)σ1 (ωn ) 2

σ2 (α)σ2 (ω1 ) · · · σ2 (α)σ2 (ωn )

D(α) = . .
.
. ··· .
.

σn (α)σn (ω1 ) · · · σn (α)σn (ωn )

2
σ1 (ω1 ) ··· σ1 (ωn )

σ (ω ) ··· σ2 (ωn )
2 2 1
= (σ1 (α)σ2 (α) · · · σn (α)) . ..
.. ··· .

σn (ω1 ) ··· σn (ωn )
= N (α)2 d(K ),

so that

D(α)
N (α) = = N (α)2 = |N (α)|.
d(K )

We see that Example 7.1.2 and Theorem 7.1.6 are special cases of Theorem 9.2.5.
We also observe that if α ∈ O K , where K is an algebraic number field, then

card(O K /α) = N (α) = |N (α)|.

Next we determine the norm of the principal ideal α in terms of the constant term
of the minimal polynomial of α over Q.

Theorem 9.2.6 Let K be an algebraic number field of degree n. Let α ∈ K . Let

irrQ (α) = x m + bm−1 x m−1 + · · · + b0 ∈ Q[x].

Then

N (α) = |b0 |n/m .

Proof: By Theorem 6.3.2 we know that m | n. Let α1 = α, α2 , . . . , αm be the roots

of irrQ (α) so that

x m + bm−1 x m−1 + · · · + b0 = (x − α1 )(x − α2 ) · · · (x − αm )

CB609-09 CB609/Alaca & Williams August 7, 2003 16:51 Char Count= 0

228 Norms of Ideals

and thus
b0 = (−1)m α1 α2 · · · αm .
Again by Theorem 6.3.2 a complete set of conjugates of α is
α1 , . . . , α 1 , α2 , . . . , α 2 , . . . , α m , . . . , α m ,
where each αi is repeated n/m times. Hence
n/m n/m
N (α) = α1 α2 · · · αmn/m = (α1 α2 · · · αm )n/m
n/m
= ((−1)m b0 )n/m = (−1)n b0
and thus by Theorem 9.2.5
n/m
N (α) = |N (α)| = |(−1)n b0 | = |b0 |n/m .

√ √ √
Example 9.2.2
√ Let K = Q( 2 + 3). We determine N ( 2). The minimal poly-
nomial of 2 over Q is x 2 − 2. Hence in the notation of Theorem 9.2.6 we have
n = 4, m = 2, b1 = 0, b0 = −2. Thus by Theorem 9.2.6 we obtain
√
N ( 2) = | − 2|4/2 = 22 = 4.

9.3 Norm of a Product of Ideals

In this section we prove the multiplicative property (9.1.3) of norms of ideals.
We will need the following result, the proof of which closely resembles that of
Theorem 8.5.1.

Theorem 9.3.1 Let D be a Dedekind domain. Let A be a fractional or integral

ideal of D with A = 0, 1. Let B be an integral ideal of D with B =
0, 1.
Then there exists γ ∈ A such that
A = γ + AB.

Proof: Let P1 , . . . , Pn be the set of distinct prime ideals for which either
ord Pi (A) = 0 or ord Pi (AB) = 0 (or both).
This set is nonempty as A = D. By Theorem 8.4.5 we can find an element γ of the
quotient field of D such that
ord Pi (γ ) = ord Pi (A), i = 1, 2, . . . , n,
ord P (γ ) ≥ 0, P = P1 , . . . , Pn .
Thus
ord P (γ ) ≥ ord P (A) for all prime ideals P,
CB609-09 CB609/Alaca & Williams August 7, 2003 16:51 Char Count= 0

9.3 Norm of a Product of Ideals 229

and so
γ ∈ A.
Now for i = 1, 2, . . . , n we have
ord Pi (γ + AB) = min(ord Pi (γ ), ord Pi (AB))
= min(ord Pi (γ ), ord Pi (AB))
= min(ord Pi (A), ord Pi (AB))
= ord Pi (A),
as B is an integral ideal. For a prime ideal P = P1 , . . . , Pn we have ord P (A) =
ord P (AB) = 0 so that
ord P (γ + AB) = min(ord P (γ ), ord P (AB))
= min(ord P (γ ), 0)
=0
= ord P (A).
Hence
ord P (γ + AB) = ord P (A)
for all prime ideals P, and so
A = γ + AB.

Theorem 9.3.2 Let K be an algebraic number field. Let A and B be nonzero

integral ideals in D = O K . Then
N (AB) = N (A)N (B).

Proof: If A or B = D then the result is trivially true as N (D) = card(D/D) = 1.

Hence we may assume that A = D and B = D. Let k = N (A) and l = N (B). Then,
by Theorem 9.1.3, the ring D/A has k elements, say,
α1 + A, . . . , αk + A.
Also, D/B has l elements, say,
β1 + B, . . . , βl + B.
By Theorem 9.3.1 there is an element γ of A such that
A = γ + AB.
If γ = 0 then A = AB so that B = D, contradicting B = D. Hence γ = 0. Let
δ ∈ D. Then there is a unique integer i (1 ≤ i ≤ k) such that
δ ≡ αi (mod A).
CB609-09 CB609/Alaca & Williams August 7, 2003 16:51 Char Count= 0

230 Norms of Ideals

Clearly,
δ − αi ∈ A = γ + AB
so there exist σ ∈ D and τ ∈ AB such that
δ − αi = σ γ + τ.
Similarly, there is a unique integer j (1 ≤ j ≤ l) such that
σ ≡ β j (mod B),
that is,
σ − β j ∈ B.
As γ ∈ A we have
(σ − β j )γ ∈ AB.
Hence
δ = αi + σ γ + τ = αi + β j γ + (σ − β j )γ + τ ≡ αi + β j γ (mod AB).
This shows that the set of kl elements αi + β j γ + AB (i = 1, . . . , k; j = 1, . . . , l)
is a complete set of representatives of D/AB. We must still show that they are
distinct. Suppose
αi + β j γ + AB = α p + βq γ + AB.
Then
αi + β j γ ≡ α p + βq γ (mod AB)
and thus
αi − α p ≡ (βq − β j )γ (mod AB).
But γ ∈ A so
αi − α p ∈ A.
Thus i = p and
β j γ ≡ βq γ (mod AB).
Hence
(β j − βq )γ ∈ AB.
h bi
Now let B = i=1 Pi (bi > 0) be the prime ideal decomposition of B. Then
ord Pi (A) = ord Pi (γ + AB)
= min(ord Pi (γ ), ord Pi (AB))
= min(ord Pi (γ ), ord Pi (A) + ord Pi (B))
CB609-09 CB609/Alaca & Williams August 7, 2003 16:51 Char Count= 0

9.4 Norm of a Fractional Ideal 231

and it follows that

ord Pi (A) = ord Pi (γ ), i = 1, . . . , h. (9.3.1)
If β j − βq = 0 then (β j − βq )γ is a nonzero element of AB so that for i = 1, . . . , h
we have
ord Pi ((β j − βq )γ ) ≥ ord Pi (AB)
and thus
ord Pi (β j − βq ) + ord Pi (γ ) ≥ ord Pi (A) + ord Pi (B).
Then, appealing to (9.3.1), we deduce that
ord Pi (β j − βq ) ≥ ord Pi (B), i = 1, . . . , h,
which shows that
β j − βq ∈ B,
and hence j = q, contradicting β j − βq = 0. This proves that β j − βq = 0 so that
j = q. Hence {αi + β j γ + AB | i = 1, . . . , k; j = 1, . . . , l} is a complete set of
distinct representatives of D/AB and so
N (AB) = card(D/AB) = kl = N (A)N (B).

9.4 Norm of a Fractional Ideal

The multiplicative property of the norm (Theorem 9.3.2) allows us to extend the
definition of the norm of an integral ideal (of the ring of integers of an algebraic
number field) to the norm of a fractional ideal.

Definition 9.4.1 (Norm of a fractional ideal) Let K be an algebraic number field.

Let O K be its ring of integers. Let A be a nonzero fractional ideal of O K . Then there
exists a nonzero integral ideal I of O K and a nonzero element α of O K such that
1
I.A=
α
We define the norm N (A) of the fractional ideal A by
N (I )
N (A) = ,
N (α)
where N (I ), N (α) are the norms of the integral ideals I and α.

Definition 9.4.1 is valid for if I and J are nonzero integral ideals of O K and α
and β are nonzero elements of O K such that
1 1
A= I = J
α β
CB609-09 CB609/Alaca & Williams August 7, 2003 16:51 Char Count= 0

232 Norms of Ideals

then
β I = α J,
so that we have the equal products of integral ideals
βI = αJ,
and thus by Theorem 9.3.2
N (β)N (I ) = N (βI ) = N (αJ ) = N (α)N (J ),
so that
N (I ) N (J )
= .
N (α) N (β)
When the fractional ideal A of Definition 9.4.1 is actually an integral ideal the two
definitions of the norm coincide.
The next theorem shows that the multiplicative property of the norm carries over
to fractional ideals.

Theorem 9.4.1 Let K be an algebraic number field. Let O K be its ring of integers.
Let A and B be nonzero fractional ideals of O K . Then
N (AB) = N (A)N (B).

Proof: As A and B are nonzero fractional ideals of O K there exist nonzero integral
ideals I and J of O K and nonzero elements α and β of O K such that
1 1
A= I, B = J.
α β
Then
1
AB = IJ
αβ
so that
N (I J )
N (AB) = (Definition 9.4.1)
N (αβ)
N (I J )
= (Definition 1.6.2)
N (αβ)
N (I )N (J )
= (Theorem 9.3.2)
N (α)N (β)
N (I ) N (J )
= ·
N (α) N (β)
= N (A)N (B).
√ √
Example 9.4.1 Let K = Q( 6) so that O K = {a + b 6 | a, b ∈ Z}. Let A be the
fractional ideal of O K given by
CB609-09 CB609/Alaca & Williams August 7, 2003 16:51 Char Count= 0

Exercises 233

1√
A = 1, 6.
2
Then
1
A= I,
2
√
where I is the integral ideal 2, 6. Now
√ √ √
I 2 = 2, 62 = 4, 2 6, 6 = 22, 6, 3 = 2
√
(as 1 = 3 − 2 ∈ 2, 6, 3) so that
N (I )2 = N (I 2 ) = N (2) = 22
and thus
N (I ) = 2.
Hence
N (I ) 2 1
N (A) = N ( 12 I ) = = 2 = .
N (2) 2 2

Exercises
1. Let p be a prime such that p ≡ 3 or 5 (mod 8). Prove that there does not exist an element
α ∈ OQ(√ p) such that N (α) = 2.
2. Let p be a prime such that p ≡ 5 or 7 (mod 8). Prove that there does not exist an element
α ∈ OQ(√ p) such that N (α) = −2.
3. Let K be an algebraic number field and O K its ring of integers. If I is an integral ideal
of O K such that N (I ) is a prime, then I is a prime ideal.
4. Let K be an algebraic number field and O K its ring of integers. If I is a nonzero integral
ideal of O K , prove that I | N (I ).
5. Let K be an algebraic number field. Let n be a given positive integer. Prove that there
are only finitely many integral ideals I of O K such that N (I ) = n.
6. Let K = Q(θ ), where θ 3 − θ − 1 = 0. Prove that 23, 3 − θ is a prime ideal in O K .
√ √
7. Let K = Q( −23). Let I = 2, 12 (1 + −23).
(a) Prove that N (I ) = 2.
√
(b) Prove that I 3 = (−3 + −23)/2.
(c) Use each of (a) and (b) to prove that I is not a principal ideal.
8. Let K be an algebraic number field and O K its ring of integers. Let I be an integral
ideal of O K such that N (I ) = |N (a)| for some a ∈ I . Prove that I = a.
9. Let K be an algebraic number field and O K its ring of integers. Let P be a prime ideal
of O K . Prove that G = {a + P | a ∈ O K , a ∈ P} is a cyclic group with respect to
multiplication. What is the order of G?
10. Let K be an algebraic number field and O K its ring of integers. Let P be a prime ideal
of O K . Prove that P ∩ Z = p for some prime p.
CB609-09 CB609/Alaca & Williams August 7, 2003 16:51 Char Count= 0

234 Norms of Ideals

11. Let K be an algebraic number field and O K its ring of integers. Show that
a p ≡ a (mod P) ⇐⇒ a ≡ m (mod P) for some m ∈ Z,
where P ∩ Z = p (see Exercise 10).
12. Determine the fractional ideals of Z + Zi.
13. Let K be an algebraic number field and O K its ring of integers. Let m ∈ Z \ {0}. Prove
that there exist only finitely many integral ideals of O K to which m belongs.
√
14. Find all the ideals of Z + Z√−5 that contain 6.
15. Find all the ideals of Z + Z 2 with norm 12.
16. Determine the set of positive integers that are not norms of ideals of Z + Zi.
17. Let K be an algebraic number field. Let O K be its ring of integers. Let P be a prime
ideal of O K . Let a ∈ O K be such that P a. Prove that
a N (P) − 1 ≡ 0 (mod P).
18. Give an example of an algebraic number field K and an integral ideal I = a, b, c of
O K such that I = a, b, a, c, b, c.
19. Determine all complex quadratic fields K for which O K possesses elements of norm
38 and trace 11.
20. Let K be an algebraic number field. Let α, β ∈ O K \ {0}. Prove that N (α)tr(β/α) ∈ Z.
√ √ √
21. Solve 3x ≡ 5 (mod A) in Z + Z −5, where A = 3 −5, 10 + 10 −5.
22. Let K be an algebraic number field. Let I be an integral ideal of O K . Let m be the least
positive integer in I . Prove that m | N (I ).
23. Let K be an algebraic number field. Let I be an integral ideal of O K such that pq | N (I ),
where p and q are distinct primes. Prove that I is not a prime ideal.
24. Let K be a quadratic field. Let α ∈ O K be such that |N (α)| = ab, where a and b are
coprime positive integers. Prove that
a, αb, α = α.
25. Prove that d1 , d2 , . . . , dn in Theorem 9.1.1 can be arranged to satisfy d1 | d2 | · · · | dn .
26. Let K be an algebraic number field. Prove that O K is a principal ideal domain if and
only if for every pair (α, β) ∈ O K × O K such that
α = 0, β = 0, β α, |N (α)| ≥ |N (β)|,
there exist γ ∈ O K and δ ∈ O K such that
0 < |N (αγ − βδ)| < |N (β)|.

Suggested Reading
1. P. B. Bhattacharya, S. K. Jain, and S. R. Nagpaul, Basic Abstract Algebra, second edition,
Cambridge University Press, Cambridge, United Kingdom, 1994.
Chapter 20 is devoted to the Smith normal form of a matrix over a principal ideal domain.
2. C. C. MacDuffee, An Introduction to Abstract Algebra, Wiley, New York, 1956.
Section 105 discusses the Smith normal form of a matrix and mentions that this form is named for
H. J. S. Smith.
CB609-09 CB609/Alaca & Williams August 7, 2003 16:51 Char Count= 0

Biographies 235

3. I. N. Stewart and D. O. Tall, Algebraic Number Theory, second edition, Chapman and
Hall, London, 1987.
Section 1.6 contains a very readable discussion of free Abelian groups.

Biographies
1. H. J. S. Smith, Report on the Theory of Numbers, Chelsea, New York, 1964.
This book contains a biographical sketch of the Irish mathematician Henry John Stephen Smith
(1826–1883).
2. The website

http://www-groups.dcs.st-and.ac.uk/˜history/

contains a biography of H. J. S. Smith.

CB609-10 CB609/Alaca & Williams August 7, 2003 16:53 Char Count= 0

10
Factoring Primes in a Number Field

10.1 Norm of a Prime Ideal

We begin by showing that each prime ideal in the ring of integers of an algebraic
number field is associated with a unique rational prime.

Theorem 10.1.1 Let K be an algebraic number field. Let P be a prime ideal of

O K . Then there exists a unique rational prime p such that

P | p.

Proof: As P is a prime ideal in O K , P ∩ Z is a prime ideal in Z (Theorem 1.6.2).

Hence, by Theorems 1.4.1 and 1.5.4, we have

P ∩ Z = p

for some rational prime p. Thus

P ⊇ p

and so by Theorem 8.4.1 we have

P | p.

Suppose q is another rational prime such that

P | q.

Then P ⊇ p and P ⊇ q so that

P ⊇ p, q.

As p and q are distinct primes we have gcd( p, q) = 1 so that there are integers
a and b such that ap + bq = 1. Hence 1 ∈ p, q ⊆ P. Thus O K ⊆ P, which is
impossible.
Hence the prime p is uniquely determined by P | p.

236
CB609-10 CB609/Alaca & Williams August 7, 2003 16:53 Char Count= 0

10.1 Norm of a Prime Ideal 237

The rational prime p in Theorem 10.1.1 is called the prime lying below P as
P ⊇ p. Given a rational prime p, any prime ideal P such that P | p is said to
be a prime ideal lying over p.
Next we relate the norm of the prime ideal P to the prime p lying below P.

Theorem 10.1.2 Let K be an algebraic number field with [K : Q] = n. Let P be

a prime ideal of O K . Let p be the rational prime lying below P. Then
N (P) = p f
for some integer f ∈ {1, 2, . . . , n}.

Proof: As p lies below P we have P | p. Hence p = P Q for some integral

ideal Q of O K . By Theorem 9.3.2 we have
N ( p) = N (P Q) = N (P)N (Q).
As the K -conjugates of p comprise p repeated n times (Theorem 6.3.4), we have
N ( p) = p n ,
so that, by Theorem 9.2.5,
N ( p) = |N ( p)| = p n .
Hence we have
p n = N (P)N (Q)
so that
N (P) = p f
for some f ∈ {1, 2, . . . , n}.

Definition 10.1.1 (Inertial degree) Let K be an algebraic number field with

[K : Q] = n. Let p be the rational prime lying below P. Then the positive integer
f such that
N (P) = p f
is called the inertial degree of P in O K and is denoted by f K (P).

From Theorems 9.1.3 and 10.1.2 we see that

card(O K /P) = p f
so that O K /P is a finite field with p f elements. Consider the elements
a + P (a ∈ Z)
CB609-10 CB609/Alaca & Williams August 7, 2003 16:53 Char Count= 0

238 Factoring Primes in a Number Field

of O K /P. If a, a ∈ Z are such that a ≡ a (mod p) then p | a − a so that p |

a − a . But P | p so P | a − a and thus a ≡ a (mod P), that is, a + P =
a + P. Conversely suppose that a + P = a + P (a, a ∈ Z). Then a − a ∈ P,
a − a ⊆ P, and P | a − a , so

a − a = P Q

for some integral ideal Q of O K . Taking norms we obtain

|a − a |n = N (a − a ) = N (P Q) = N (P)N (Q) = p f N (Q),

so that p | (a − a )n and thus, as p is a prime, p | a − a and so a ≡ a (mod p).

We have shown that for a, a ∈ Z

a + P = a + P ⇐⇒ a ≡ a (mod p).

Hence the cosets a + P (a ∈ {0, 1, . . . , p − 1}) are distinct and the prime field of
O K /P is

F = {a + P | a = 0, 1, . . . , p − 1} Z/ p.

The inertial degree f = f K (P) is given by

f K (P) = [O K /P : F] = [O K /P : Z/ p].

If f K (P) = 1 then [O K /P : F] = 1 so that O K /P = F; that is,

O K /P = {a + P | a = 0, 1, . . . , p − 1}.

Let α ∈ O K . Then α + P ∈ O K /P so that α + P = a + P for some a ∈

{0, 1, . . . , p − 1}. Hence α ≡ a (mod P) for some a ∈ Z, when f K (P) = 1.

Theorem 10.1.3 Let K be an algebraic number field with [K : Q] = n. Let p be

a rational prime. Suppose that the principal ideal p factors in O K in the form
e
p = P1e1 · · · Pg g , (10.1.1)

where P1 , . . . , Pg are distinct prime ideals of O K and e1 , . . . , eg are positive in-

tegers. Suppose that f i is the inertial degree of Pi (i = 1, 2, . . . , g) in K , that is,
f i = f K (Pi ). Then

e1 f 1 + · · · + eg f g = n.

Proof: As f i is the inertial degree of Pi in K , we have N (Pi ) = p fi (i = 1, 2, . . . , g)

so that
CB609-10 CB609/Alaca & Williams August 7, 2003 16:53 Char Count= 0

10.1 Norm of a Prime Ideal 239

e
p n = N ( p) = N (P1e1 · · · Pg g )
= N (P1 )e1 · · · N (Pg )eg
e1 eg
= ( p f1 ) · · · ( p f g )
= p e1 f1 +···+eg f g .
Hence
e1 f 1 + · · · + eg f g = n.

It follows immediately from Theorem 10.1.3 that

ei ∈ {1, 2, . . . , n}, f i ∈ {1, 2, . . . , n}, i = 1, 2, . . . , g, (10.1.2)
and that
P n | p (for some prime ideal P) =⇒ p = P n . (10.1.3)

Definition 10.1.2 (Decomposition number) With the notation of Theorem 10.1.3,

the positive integer g defined in (10.1.1) is called the decomposition number of
p in K and is written g K ( p).

Theorem 10.1.4 Let K be an algebraic number field of degree n. Let p be a rational

prime. Then
g K ( p) ≤ n.

Proof: With the notation of Theorem 10.1.3, as e1 , . . . , eg , f 1 , . . . , f g are positive

integers, we deduce that
n = e1 f 1 + · · · + eg f g ≥ 1 + · · · + 1 = g.

Thus in an algebraic number field of degree n, the principal ideal p (with p a

rational prime) cannot split into a product of more than n distinct prime powers.

Definition 10.1.3 (Ramification index) Let K be an algebraic number field of

degree n. Let P be a prime ideal of O K . Let p be a rational prime lying below P.
Then the unique positive integer e such that
P e | p, P e+1 p
is called the ramification index of P in K and is written e K (P). From (10.1.2) we
see that e K (P) ≤ n.

In the notation of Theorem 10.1.3 we have

e K (Pi ) = ei , f K (Pi ) = f i , g K ( p) = g.
CB609-10 CB609/Alaca & Williams August 7, 2003 16:53 Char Count= 0

240 Factoring Primes in a Number Field

Definition 10.1.4 (Ramification) Let K be an algebraic number field of degree n.

Let p be a rational prime. Let P1 , . . . , Pg be the prime ideals of O K lying above p.
Then
e
p = P1e1 · · · Pg g ,

where

ei = e K (Pi ), i = 1, 2, . . . , g.

If ei > 1 for some i ∈ {1, 2, . . . , g} then p is said to ramify in K . If ei = 1 for

i = 1, 2, . . . , g then p is said to be unramified in K .

The following theorem of Dedekind, which we shall not prove here, enables us
to recognize when a rational prime p ramifies in an algebraic number field K .

Theorem 10.1.5 (Dedekind) Let K be an algebraic number field. Then the rational
prime p ramifies in K if and only if p | d(K ).

In the next section we examine the factorization of a rational prime p into prime
ideals in O K , when K is a quadratic field.
We conclude this section by proving the following simple but useful result.

Theorem 10.1.6 Let K be an algebraic number field. Let I (= 0) be an ideal of
OK .

(a) If N (I ) = p, where p is a prime, then I is a prime ideal.

(b) N (I ) ∈ I .

Proof: (a) Clearly I = 1 as N (I ) = 1. By Theorem 8.3.1 I is a product of prime

ideals. If there exist prime ideals P and Q such that P Q divides I (where P and
Q may or may not be distinct) then I = P Q A for some integral ideal A of O K .
Hence, by Theorem 9.4.1, we have

p = N (I ) = N (P Q A) = N (P)N (Q)N (A),

which contradicts that p is a rational prime as N (P), N (Q), and N (A) are positive
integers with N (P) > 1 and N (Q) > 1. Thus I is a prime ideal.
(b) By Theorem 9.1.3 we have

N (I ) = card (O K /I ).

Hence

N (I )(x + I ) = 0 + I, for all x ∈ O K ;

CB609-10 CB609/Alaca & Williams August 7, 2003 16:53 Char Count= 0

10.2 Factoring Primes in a Quadratic Field 241

that is,
N (I )x ∈ I, for all x ∈ O K .
Taking x = 1 ∈ O K we obtain
N (I ) ∈ I
as asserted.

From Theorems 10.1.6(b) and 8.4.1 we deduce that

N (I ) ⊆ I
and
I | N (I )
for all ideals I of the ring of integers of an algebraic number field.

10.2 Factoring Primes in a Quadratic Field

Let p be a rational prime and let K be a quadratic field. By Theorem 10.1.4 we
have
g = g K ( p) ≤ 2
so that
g = 1 or 2.
If g = 2 by Theorem 10.1.3 we have
e1 f 1 + e2 f 2 = 2
so that
e1 = f 1 = e2 = f 2 = 1.
If g = 1 we have
e1 f 1 = 2
so that
(e1 , f 1 ) = (2, 1) or (1, 2).
Thus in the case of a quadratic field there are just three possibilities:
(i) g = 2, e1 = f 1 = e2 = f 2 = 1,
(ii) g = 1, e1 = 2, f 1 = 1,
(iii) g = 1, e1 = 1, f 1 = 2.
CB609-10 CB609/Alaca & Williams August 7, 2003 16:53 Char Count= 0

242 Factoring Primes in a Number Field

In other words,
(i) p = P1 P2 , N (P1 ) = N (P2 ) = p, P1 = P2 ,
(ii) p = P 2 , N (P) = p,
(iii) p = P, N (P) = p 2 ,
where P1 , P2 , P denote prime ideals of O K . In case (i) we say that p splits in K ,
in case (ii) that p ramifies in K , and in case (iii) that p is inert (or remains prime)
in K . In cases (i) and (iii) p is unramified in K .
Our next theorem gives necessary and sufficient conditions for each of (i), (ii),
(iii) to occur. As usual mp denotes the Legendre symbol of the integer m modulo
the odd prime p.

Theorem 10.2.1 Let K be a quadratic field so that there exists a squarefree integer
√
m such that K = Q( m). Let p be a rational prime.

(a) If p > 2, m
p
= 1 or p = 2, m ≡ 1 (mod 8) then
p = P1 P2 ,
where P1 and P2 are distinct prime ideals with N (P1 ) = N (P2 ) = p.
(b) If p > 2, p | m or p = 2, m ≡ 2 or 3 (mod 4) then
p = P 2 ,
where P isa prime
ideal with N (P) = p.
(c) If p > 2, p = −1 or p = 2, m ≡ 5 (mod 8) then
m

p is a prime ideal of O K .

Proof: As m is squarefree we have m ≡ 0 (mod 4) so that m ≡ 1, 2, or 3 (mod 4).

cases.
We consider seven
(i): p > 2, mp = 1. As mp = 1 there exists a ∈ Z such that a 2 ≡ m (mod p).
√ √
As p m we have p a. Let P1 = p, a + m and P2 = p, a − m.
We show first that P1 = P2 . Suppose on the contrary that P1 = P2 . Then
√ √
2a = (a + m) + (a − m) ∈ P1 .
But 2a ∈ Z so
2a ∈ P1 ∩ Z = p.
Hence p | 2a. This is impossible as p is odd and p a.
Next we show that p = P1 P2 . We have
√ √
P1 P2 = p, a + m p, a − m
√ √
= p 2 , p(a + m), p(a − m), a 2 − m
√ √
= p p, a + m, a − m, (a 2 − m)/ p
= pI,
CB609-10 CB609/Alaca & Williams August 7, 2003 16:53 Char Count= 0

10.2 Factoring Primes in a Quadratic Field 243

where I is the ideal

√ √
I = p, a + m, a − m, (a 2 − m)/ p.
As gcd(2a, p) = 1 there exist integers x and y such that
x p + y(2a) = 1.
Thus
√ √
1 = x p + y(a + m) + y(a − m) ∈ I
and so I = 1.This
proves that p = P1 P2 , showing that this case falls under (a).
(ii): p > 2, p = −1. Suppose that
m

p = P1 P2 (P1 = P2 ) or P12 .
√
In each case we have N (P1 ) = p so that f K (P1 ) = 1. Hence, as m ∈ O K , there
exists a ∈ Z such that
√
m ≡ a (mod P1 )
and so
m ≡ a 2 (mod P1 ).
But, as m ∈ Z, a 2 ∈ Z and P1 | p, we must have
m ≡ a 2 (mod p),

contradicting m
p
= −1. Thus p is a prime ideal in O K , and this case falls
under (c).
√
(iii): p > 2, p | m. Set P = p, m. Then
√ √ √
P 2 = p, m p, m = p 2 , p m, m = pI,
where I is the ideal
√
I = p, m, m/ p.
As m is squarefree, we have gcd( p, m/ p) = 1, so that there exist integers x and y
such that
x p + y(m/ p) = 1.
Hence
1 = x p + y(m/ p) ∈ I,
so I = 1, that is, p = P 2 , and this case falls under (b).
√
(iv): p = 2, m ≡ 2 (mod 4). Set P = 2, m. Then
√ √ √
P 2 = 2, m2, m = 4, 2 m, m = 2I,
CB609-10 CB609/Alaca & Williams August 7, 2003 16:53 Char Count= 0

244 Factoring Primes in a Number Field

where I is the ideal

√
I = 2, m, m/2.
As m/2 is odd, say m/2 = 2k + 1, then
1 = (−k)2 + (1)m/2 ∈ I,
so I = 1, and P 2 = 2. This case falls under (b).
√ √
(v): p = 2, m ≡ 3 (mod 4). Let P = 2, 1 + m. As 1 − m = 2 − (1 +
√ √
m) we see that P = 2, 1 − m. Then
√ √
P 2 = 2, 1 + m2, 1 − m
√ √
= 4, 2(1 + m), 2(1 − m), 1 − m
= 2I,
where I is the ideal
√ √
I = 2, 1 + m, 1 − m, (1 − m)/2.
As m ≡ 3 (mod 4), (1 − m)/2 is an odd integer, and I = 1. Hence 2 = P 2
and this case falls under (b).

(vi): p = 2, m ≡ 1 (mod 8). Let

√ √
1+ m 1− m
P1 = 2, and P2 = 2, .
2 2
Thus
√ √
1+ m 1− m 1−m
P1 P2 = 4, 2 ,2 , = 2I,
2 2 4
where I is the ideal
√ √
1+ m 1− m 1−m
I = 2, , , .
2 2 8
Now
√ √
1+ m 1− m
1= + ∈I
2 2
so that I = 1 and thus 2 = P1 P2 . If P1 = P2 then
√ √
1+ m 1− m
1= + ∈ P1 ,
2 2
which is impossible, so P1 = P2 , and this case falls under (a).

(vii): p = 2, m ≡ 5 (mod 8). Suppose that 2 = P1 P2 (P1 = P2 ) or P12 . Then

√
in both cases N (P1 ) = 2 so that f K (P1 ) = 1. Hence, as (1 + m)/2 ∈ O K , there
CB609-10 CB609/Alaca & Williams August 7, 2003 16:53 Char Count= 0

10.2 Factoring Primes in a Quadratic Field 245

exists a rational integer a such that

√
1+ m
≡ a (mod P1 ).
2
Thus
√ √
1− m 1+ m
=1− ≡ 1 − a (mod P1 )
2 2
and so
√ √
1−m 1+ m 1− m
= ≡ a(1 − a) (mod P1 ).
4 2 2
Since (1 − m)/4 ∈ Z, a(1 − a) ∈ Z, and P1 | 2, we have
1−m
≡ a(1 − a) (mod 2).
4
Thus
1−m
≡ 0 (mod 2)
4
so that m ≡ 1 (mod 8), contradicting m ≡ 5 (mod 8). Hence 2 is inert in O K and
this case falls under (c).

From the proof of Theorem 10.2.1 we see that we can express the factorizations
of the principal ideal p (with p prime) into prime ideals of O K , where K is the
√
quadratic field Q( m) (m squarefree), as follows:


 2, √ √ if m ≡ 5 (mod 8),

2, 2 (1 + m)2, 2 (1 − m), if m ≡ 1 (mod 8),
1 1
2 = √

 2, 1 + m2 , if m ≡ 3 (mod 4),
 √ 2
2, m , if m ≡ 2 (mod 4),
and for p > 2




p, if p m and x 2 ≡ m (mod p)


 √ √
is insolvable,
p = p, x + m p, x − m if p m and x 2 ≡ m (mod p)



 is solvable,
 √
p, m2 , if p | m.
Recalling that for a squarefree integer m

√ m, if m ≡ 1 (mod 4),
d(Q( m)) =
4m, if m ≡ 2, 3 (mod 4),
CB609-10 CB609/Alaca & Williams August 7, 2003 16:53 Char Count= 0

246 Factoring Primes in a Number Field

we see from Theorem 10.2.1 that

√ √
p ramifies in Q( m) ⇐⇒ p | d(Q( m)),
in agreement with Theorem 10.1.5.
We next simplify the statement of Theorem 10.2.1 by introducting the Kronecker
symbol, which is an extension of the Legendre symbol from an odd prime to the
prime 2.

Definition 10.2.1 (Kronecker symbol) d be a nonsquare integer with d ≡

Let
d
0 or 1 (mod 4). The Kronecker symbol 2 is defined by

 0, if d ≡ 0 (mod 4),
d
= 1, if d ≡ 1 (mod 8),
2 
−1, if d ≡ 5 (mod 8).

Thus −42
= 0, 17
2
= 1, −3
2
= −1, and −6
2
is not defined. Making use
of the Kronecker symbol, Theorem 10.2.1 can be reformulated as follows:

Theorem 10.2.2 Let K be a quadratic field. Let d = d(K ). Let p be a rational

prime. Then

d
(i) p splits ⇐⇒ = 1,
p

d
(ii) p ramifies ⇐⇒ = 0,
p

d
(iii) p is inert ⇐⇒ = −1,
p

where dp is the Legendre symbol for p > 2 and the Kronecker symbol for p = 2.

Example 10.2.1
√
(a) 11 is inert in Q( −163) as

−163 2
= = −1.
11 11
√
(b) 23 is inert in Q( 37) as

37 14 2 7
= = = (+1)(−1) = −1.
23 23 23 23
√
(c) 2 ramifies in Q( 7) as

28
= 0.
2
√
Indeed 2 = 2, 1 + 72 .
CB609-10 CB609/Alaca & Williams August 7, 2003 16:53 Char Count= 0

10.2 Factoring Primes in a Quadratic Field 247

√
(d) 3 splits in Q( 7) as

28 1
= = 1.
3 3
Indeed as x 2 ≡ 7 (mod 3) has the solution x = 1, we have
√ √
3 = 3, 1 + 73, 1 − 7.
√ √
(e) 7 ramifies in Q( 7) as 287
= 0. Indeed 7 = 72 .

√
Let K be the quadratic field Q( m), where m is a squarefree integer. There are
exactly two monomorphisms : K −→ C, namely, 1 and σ given by
√ √
1(a + b m) = a + b m,
√ √
σ (a + b m) = a − b m,
for all a, b ∈ Q. Let I be an ideal of O K . By Theorem 8.5.1 we know that I is
generated by at most two elements. Hence I = α or I = α, β and we define the
conjugate ideal σ (I ) of I by
σ (I ) = σ (α) or σ (α), σ (β)
respectively. It is customary to write α for σ (α) and
√ I for √σ (I ). Clearly

(α ) =√ σ (α) =√α so that (I ) = I . Thus if I = 2 + 3, 1 − 2 3 then I =
2

2 − 3, 1 + 2 3. We note that

(α + β) = σ (α + β) = σ (α) + σ (β) = α + β
and
(αβ) = σ (αβ) = σ (α)σ (β) = α β
for all α and β in K . From these properties it is easy to show that if I =
α1 , . . . , αn then I = α1 , . . . , αn .
The basic property of conjugate ideals is given in the next theorem.

Theorem 10.2.3 Let I and J be ideals of the ring of integers of a quadratic field
K . Then
(I J ) = I J .

Proof: As α = α, α we may suppose that both I and J are generated by two
elements, say,
I = α, β, J = γ , δ.
Then
I J = α, βγ , δ = αγ , βγ , αδ, βδ
CB609-10 CB609/Alaca & Williams August 7, 2003 16:53 Char Count= 0

248 Factoring Primes in a Number Field

so that
(I J ) = (αγ ) , (βγ ) , (αδ) , (βδ) = α γ , β γ , α δ , β δ
= α , β γ , δ = I J
as asserted.

If I is an ideal of the ring of integers of an arbitrary algebraic number field K ,

we have already observed that I | N (I ). In the case of quadratic fields, we can
say something stronger.

Theorem 10.2.4 Let K be a quadratic field. Let I be an ideal of O K . Then

N (I ) = I I .

Proof: The assertion of the theorem is trivial if I = 0 or 1, so we may suppose
that I = 0, 1. Let
b1 br
I = P1a1 P1 · · · Prar Pr Q c11 · · · Q cs s R1d1 · · · Rtdt
be the prime ideal decomposition of I , where P1 , . . . , Pr are distinct prime ideals
such that
P P = p, N (P) = N (P ) = p, P = P ,
Q 1 , . . . , Q s are distinct prime ideals such that
Q = Q = q, N (Q) = q 2 ,
R1 , . . . , Rt are distinct prime ideals such that
R = R , R 2 = r , N (R) = r,
and p, q, r denote rational primes. Then, by Theorem 10.2.3, we have
a1 ar
I = P1 P1b1 · · · Pr Prbr Q c11 · · · Q cs s R1d1 · · · Rtdt .
Hence
a1 +b1 ar +br
I I = P1a1 +b1 P1 · · · Prar +br Pr Q 2c 1 2cs 2d1
1 · · · Q s R1 · · · Rt
2dt

= p1 a1 +b1 · · · pr ar +br q1 2c1 · · · qs 2cs r1 d1 · · · rt dt
= p1a1 +b1 · · · prar +br q12a · · · qs2cs r1d1 · · · rtdt .
Further, by Theorem 9.3.2, we have
N (I ) = N (P1 )a1 N (P1 )b1 · · · N (Pr )ar N (Pr )br N (Q 1 )c1 · · ·
N (Q s )cs N (R1 )d1 · · · N (Rt )dt
= p1a1 p1b1 · · · prar prbr q12c1 · · · qs2cs r1d1 · · · rtdt .
CB609-10 CB609/Alaca & Williams August 7, 2003 16:53 Char Count= 0

10.3 Factoring Primes in a Monogenic Number Field 249

Thus
I I = N (I )
as asserted.

By Theorem 10.2.4 we have for a nonzero ideal I of O K

N (I ) = I I = I (I ) = N (I )
so that
N (I ) = N (I )
for some unit ∈ O K . But N (I ) and N (I ) are both positive integers so that = +1
and thus
N (I ) = N (I ).
This result is trivially true if I = 0.
We close this section by noting that (in the notation of Theorem 10.2.2): If dp

= 1 then
p = PP for some prime ideal P with P = P , N (P) = N (P ) = p;
if dp = 0 then p = P 2 for some prime ideal P with P = P , N (P) = p; and

if dp = −1 then p = P for some prime ideal P with P = P , N (P) = p 2 .
An ideal I such that I = I is called self-conjugate.

10.3 Factoring Primes in a Monogenic Number Field

Let K be an algebraic number field. Recall that K is said to be monogenic (Definition
7.1.5) if there exists θ ∈ O K such that
O K = Z + Zθ + · · · + Zθ n−1 ,
where [K : Q] = n. The next theorem shows how to factor p (with p a rational
prime) into prime ideals in a monogenic number field. It was originally proved by
Dedekind [3] in 1878.

Theorem 10.3.1 Let K = Q(θ ) be an algebraic number field of degree n such that
O K = Z + Zθ + · · · + Zθ n−1 .
Let p be a rational prime. Let
f (x) = irrQ θ ∈ Z[x].
−
Let denote the natural map : Z[x] → Z p [x], where Z p = Z/ pZ. Let
f¯ (x) = g1 (x)e1 · · · gr (x)er ,
CB609-10 CB609/Alaca & Williams August 7, 2003 16:53 Char Count= 0

250 Factoring Primes in a Number Field

where g1 (x), . . . , gr (x) are distinct monic irreducible polynomials in Z p [x] and
e1 , . . . , er are positive integers. For i = 1, 2, . . . , r let f i (x) be any monic polyno-
mial of Z[x] such that f¯ i = gi . Set

Pi = p, f i (θ), i = 1, 2, . . . , r.

Then P1 , . . . , Pr are distinct prime ideals of O K with

p = P1e1 · · · Prer

and

N (Pi ) = p deg fi , i = 1, 2, . . . , r.

Proof: For i = 1, 2, . . . , r let θi be a root of gi in a suitable extension field of Z p .

This extension field is the finite field Z p [θi ] Z p [x]/gi (x). Let νi : Z[θ] →
Z p [θi ] be the surjective homomorphism given by

νi (h(θ)) = h̄(θi ).

Then

Z[θ]/ker νi νi (Z[θ]) = Z p [θi ]

is a field, so that ker νi is a prime ideal of Z[θ] = O K .

Clearly

νi ( p) = 0, νi ( f i (θ)) = f¯i (θi ) = gi (θi ) = 0,

so that

p ∈ ker νi , f i (θ) ∈ ker νi ,

and thus

p, f i (θ) ⊆ ker νi .

If g(θ ) ∈ ker νi then

ḡ(θi ) = νi (g(θ )) = 0

so that gi (x) | ḡ(x) in Z p [x]. Thus

ḡ(x) = f¯i (x)h̄(x) for some h̄ ∈ Z p [x].

Hence (g − f i h)(x) ∈ Z[x] has coefficients that are divisible by p so that

g(θ ) = (g(θ ) − f i (θ)h(θ )) + f i (θ)h(θ)

∈ p + f i (θ)
= p, f i (θ),
CB609-10 CB609/Alaca & Williams August 7, 2003 16:53 Char Count= 0

10.3 Factoring Primes in a Monogenic Number Field 251

proving

ker νi ⊆ p, f i (θ).

We have shown that

Pi = p, f i (θ ) = ker νi , i = 1, 2, . . . , r.

Thus each Pi (i = 1, 2, . . . , r ) is a prime ideal of O K .

Next we show that the prime ideals Pi (i = 1, 2, . . . , r ) are distinct. Suppose that
Pi = P j for some i, j ∈ {1, 2, . . . , r }. Then p, f i (θ) = p, f j (θ). Hence

f j (θ ) = pg(θ) + f i (θ)h(θ)

for some g(x), h(x) ∈ Z[x]. Applying νi we obtain

g j (θi ) = f¯ j (θi ) = f¯i (θi )h̄(θi ) = gi (θi )h̄(θi ) = 0,

so that gi (x) | g j (x) in Z p [x]. Hence

g j (x) = gi (x)l(x)

for some l(x) ∈ Z p [x]. As gi (x) and g j (x) are both monic polynomials, which are
irreducible in Z p [x], we have l(x) = 1 so that gi (x) = g j (x) and thus i = j.
We show next that

p = P1e1 · · · Prer .

For any ideals A, B1 , B2 we have

(A + B1 )(A + B2 ) ⊆ A + B1 B2 ,

so that

P1e1 · · · Prer = p, f 1 (θ)e1 · · · p, fr (θ)er

= ( p + f 1 (θ))e1 · · · ( p + fr (θ))er
⊆ p + f 1 (θ)e1 · · · fr (θ)er
= p + f 1 (θ)e1 · · · fr (θ)er
= p + f (θ)
= p

and so

p | P1e1 · · · Prer .

Now Pi = p, f i (θ ) ⊇ p, so

Pi | p, i = 1, 2, . . . , r.
CB609-10 CB609/Alaca & Williams August 7, 2003 16:53 Char Count= 0

252 Factoring Primes in a Number Field

Hence

p = P1k1 · · · Prkr ,

where

ki ∈ {1, 2, . . . , ei }, i = 1, 2, . . . , r. (10.3.1)

Now

O K /Pi = Z[θ ]/Pi = Z[θ]/ker νi νi (Z[θ]) = Z p [θi ],

so that

N (Pi ) = card(O K /Pi ) = card(Z p [θi ]) = p di ,

where

di = deg gi = deg f¯i .

Hence we have

p n = N ( p) = N (P1k1 · · · Prkr )

= N (P1 )k1 · · · N (Pr )kr
= ( p d1 )k1 · · · ( p dr )kr
= p d1 k1 +···+dr kr ,

so that

d1 k1 + · · · + dr kr = n. (10.3.2)

Comparing degrees in

f¯(x) = f¯1 (x)e1 · · · f¯r (x)er

we obtain

d1 e1 + · · · + dr er = n. (10.3.3)

From (10.3.1)–(10.3.3) we deduce that

ki = ei , i = 1, 2, . . . , r,

so that

p = P1e1 · · · Prer ,

as asserted.
Finally, we observe that
f¯i
N (Pi ) = p di = p deg = p deg fi , i = 1, 2, . . . , r.
CB609-10 CB609/Alaca & Williams August 7, 2003 16:53 Char Count= 0

10.4 Some Factorizations in Cubic Fields 253

Theorem 10.3.1 relates the factorization of a monic irreducible polynomial

f (x) ∈ Z[x] modulo a prime p to the factorization of p into prime ideals in the
algebraic number field K defined by a root of f (x) when K is monogenic. Primes
p for which the congruence f (x) ≡ 0 (mod p) is solvable, so that f (x) has at least
one linear factor modulo p, are called prime divisors of f and the set of prime
divisors of f is denoted by P( f ). Thus

P(x 2 + 1) = {2, p (prime) ≡ 1 (mod 4)}.

The set P( f ) is discussed in the beautiful article by Gerst and Brillhart [4].
If f (x) factors modulo p into a product of distinct linear factors, we say that
f (x) splits completely modulo p. The set of all primes p such that f (x) splits
completely modulo p is denoted by Spl( f ). This set is discussed by Wyman in his
classic article [7]. Thus for example

Spl(x 3 − 31x + 62) = { p (prime > 2) ≡ 1, 2, 4, 8, 15, 16, 23, 27, 29, 30
(mod 31)}

(see [5]).
The next section will be devoted to numerical examples illustrating Theorem
10.3.1.

10.4 Some Factorizations in Cubic Fields

Example
√ 10.4.1 √We factor 5 as a product of prime ideals in O K , where K =
Q( 2). Set θ = 3 2. We have seen in Example 7.1.6 that {1, θ, θ 2 } is an integral
3

basis for K = Q(θ) so that K is monogenic. The minimal polynomial of θ over Q

is x 3 − 2. We have

x 3 − 2 = (x + 2)(x 2 + 3x + 4) (mod 5),

where x + 2 and x 2 + 3x + 4 are irreducible (mod 5). Hence, by Theorem 10.3.1,

we have

5 = P Q,

where

P = 5, θ + 2, Q = 5, θ 2 + 3θ + 4

are distinct prime ideals with

N (P) = 5, N (Q) = 52 = 25.

As a check on the calculation in Example 10.4.1 we compute P Q directly.

CB609-10 CB609/Alaca & Williams August 7, 2003 16:53 Char Count= 0

254 Factoring Primes in a Number Field

We have

P Q = 5, θ + 25, θ 2 + 3θ + 4
= 25, 5(θ + 2), 5(θ 2 + 3θ + 4), θ 3 + 5θ 2 + 10θ + 8
= 25, 5(θ + 2), 5(θ 2 + 3θ + 4), 5θ 2 + 10θ + 10
= 55, θ + 2, θ 2 + 3θ + 4, θ 2 + 2θ + 2
= 5

1 = 1 · 5 + (2θ + 2)(θ + 2) − 2(θ 2 + 3θ + 4).

Example 10.4.2 Let K = Q(θ), where θ 3 − 9θ − 6 = 0. It is known that [K :

Q] = 3, {1, θ, θ 2 } is an integral basis for K , and d(K ) = 23 · 35 (see Ex-
ercise 5 of Chapter 7). As 2 | d(K ) and 3 | d(K ), both 2 and 3 ramify in
K by Dedekind’s theorem (Theorem 10.1.5). We determine their prime ideal
decompositions.
We have

x 3 − 9x − 6 ≡ x(x + 1)2 (mod 2)

so that by Theorem 10.3.1 we have

2 = P Q 2 ,

where

P = 2, θ, Q = 2, θ + 1

are distinct prime ideals with N (P) = N (Q) = 2. In fact P and Q are both prin-
cipal ideals as we now show. From

θ 3 − 9θ − 6 = 0,

we deduce that

(θ + 1)3 − 3(θ + 1)2 − 6(θ + 1) + 2 = 0,

so that

θ + 1 | 2.

Hence

Q = 2, θ + 1 = θ + 1
CB609-10 CB609/Alaca & Williams August 7, 2003 16:53 Char Count= 0

10.4 Some Factorizations in Cubic Fields 255

and

P = 2Q −2
2
=
(θ + 1)2
6(θ + 1) + 3(θ + 1)2 − (θ + 1)3
=
(θ + 1)2
2
= 3 + 3 − (θ + 1)
θ +1
= 3(6 + 3(θ + 1) − (θ + 1)2 ) + 2 − θ
= 26 + 2θ − 3θ 2 .

As a check on this calculation we note that

P Q 2 = 26 + 2θ − 3θ 2 1 + θ2

= (26 + 2θ − 3θ 2 )(1 + 2θ + θ 2 )
= 26 + 54θ + 27θ 2 − 4θ 3 − 3θ 4
= 2,

as θ 3 = 9θ + 6 and θ 4 = 9θ 2 + 6θ.
Turning to the prime 3 we have

x 3 − 9x − 6 ≡ x 3 (mod 3),

so that by Theorem 10.3.1 we obtain

3 = R 3 ,

where R = 3, θ is a prime ideal with N (R) = 3. We show that R is a principal

ideal. We have

P R = 2, θ3, θ = 6, 2θ, 3θ, θ 2 = 6, θ = θ 3 − 9θ, θ = θ,

so that

R = θP −1
= θ Q 2 (P Q 2 )−1
θ (θ + 1)2
=
2
θ 3 + 2θ 2 + θ 2θ 2 + 10θ + 6
= =
2 2
= 3 + 5θ + θ 2 .
CB609-10 CB609/Alaca & Williams August 7, 2003 16:53 Char Count= 0

256 Factoring Primes in a Number Field

We now verify directly that R 3 = 3. We have

(3 + 5θ + θ 2 )2 = 9 + 30θ + 31θ 2 + 10θ 3 + θ 4

= 69 + 126θ + 40θ 2 ,

(3 + 5θ + θ 2 )3 = (3 + 5θ + θ 2 )(69 + 126θ + 40θ 2 )

= 207 + 723θ + 819θ 2 + 326θ 3 + 40θ 4
= 2163 + 3897θ + 1179θ 2
= 3(721 + 1299θ + 393θ 2 ).

To complete the verification of R 3 = 3 we must show that 721 + 1299θ + 393θ 2
is a unit of O K = Z + Zθ + Zθ 2 so that

R 3 = 3 + 5θ + θ 2 3 = 3721 + 1299θ + 393θ 2 = 3.

We do this by seeking a, b, c ∈ Z such that

(721 + 1299θ + 393θ 2 )(a + bθ + cθ 2 ) = 1.

Multiplying out the left-hand side, replacing θ 3 and θ 4 by 9θ + 6 and 9θ 2 + 6θ

respectively, and equating the coefficients of 1, θ, θ 2 , we are led to the three linear
equations in a, b, c :

721a + 2358b + 7794c = 1,

1299a + 4258b + 14049c = 0,
393a + 1299b + 4258c = 0.

Using the program MAPLE we find that

a = −119087, b = −9885, c = 14007,

which is easily checked directly. Hence

(721 + 1299θ + 393θ 2 )−1 = −119087 − 9885θ + 14007θ 2 ,

so that 721 + 1299θ + 393θ 2 is a unit. Alternatively, we could have used Theorem
9.2.4.

This example was considered in [2: p. 230]. However, the value of N (θ + 1) is

given there incorrectly as −4 (see Eq. (7.16)). Its correct value is

N (θ + 1) = (θ + 1)(θ + 1)(θ + 1)
= θ θ θ + (θθ + θ θ + θ θ) + (θ + θ + θ ) + 1
=6−9+0+1
= −2,
CB609-10 CB609/Alaca & Williams August 7, 2003 16:53 Char Count= 0

10.5 Factoring Primes in an Arbitrary Number Field 257

where θ, θ , θ are the roots of x 3 − 9x − 6. The factorization into prime ideals of

the principal ideals considered in [2] are
2 = PQ 2 , 3 = R 3 , θ = PR, θ + 1 = Q, θ − 1 = QS,
where
P = 2, θ = 26 + 2θ − 3θ 2 , Q = 2, θ + 1 = θ + 1,
R = 3, θ = 3 + 5θ + θ 2 , S = 7 + θ − θ 2 .

10.5 Factoring Primes in an Arbitrary Number Field

Theorem 10.3.1 was actually proved by Dedekind in the following slightly stronger
form. For all but at most a finite number of primes, Theorem 10.5.1 gives the
factorization of a prime into prime ideals in an arbitrary algebraic number field.

Theorem 10.5.1 Let K = Q(θ ) be an algebraic number field with θ ∈ O K . Let p

be a rational prime. Let
f (x) = irrQ (θ) ∈ Z[x].
Let ¯ denote the natural map : Z[x] −→ Z p [x], where Z p = Z/ pZ. Let
f¯(x) = g1 (x)e1 · · · gr (x)er ,
where g1 (x), . . . , gr (x) are distinct monic irreducible polynomials in Z p [x] and
e1 , . . . , er are positive integers. For i = 1, 2, . . . , r let f i (x) be any monic polyno-
mial of Z[x] such that f¯i = gi . Set
Pi = p, f i (θ), i = 1, 2, . . . , r.
If ind(θ) ≡ 0 (mod p) then P1 , . . . , Pr are distinct prime ideals of O K with
p = P1e1 · · · Prer
and
N (Pi ) = p deg fi , i = 1, 2, . . . , r.

We leave the proof of Theorem 10.5.1 as an exercise (Exercise 2) since it can

be modeled on the proof of Theorem 10.3.1. We note that Theorem 10.3.1 is the
special case ind θ = 1 of Theorem 10.5.1.

field Q(θ), where θ − θ + 4 = 0. An integral

3
Example 10.5.1
Let K be the cubic
basis for K is 1, θ, (θ + θ )/2 (Example 7.3.7). This basis is not a power basis
2

and at this stage we do not know whether K is monogenic or not. As ind θ = 2

we can apply Theorem 10.5.1 to obtain the prime ideal factorization of any prime
CB609-10 CB609/Alaca & Williams August 7, 2003 16:53 Char Count= 0

258 Factoring Primes in a Number Field

p = 2 in O K . The prime 107 ramifies in O K as d(K ) = −107 (Example 7.3.7) and

we make use of Theorem 10.5.1 to find its precise prime decomposition. We have

x 3 − x + 4 ≡ (x − 6)2 (x + 12) (mod 107),

so that

107 = P 2 Q,

where P and Q are the distinct prime ideals given by

P = 107, θ − 6, N (P) = 107,

and

Q = 107, θ + 12, N (Q) = 107.

We next show that K is monogenic. Let α = (θ + θ 2 )/2. Then

2
θ + θ2 θ 2 + 2θ 3 + θ 4
α =
2
=
2 4
θ 2 + 2(θ − 4) + (θ 2 − 4θ ) θ + θ2
= = −2 − θ + ,
4 2
so that

θ = −2 + α − α 2 .

Hence

θ + θ2
O K = Z + Zθ + Z = Z + Z(−2 + α − α 2 ) + Zα = Z + Zα + Zα 2 ,
2
proving that {1, α, α 2 } is an integral basis for K . This basis is clearly a power
basis, so K is monogenic.
As K is monogenic we can apply Theorem 10.3.1 to factor the prime 2 in O K .
By Example 7.3.2 we know that α = (θ + θ 2 )/2 is a root of x 3 − x 2 + 3x − 2 = 0.
Thus K = Q(α), where α 3 − α 2 + 3α − 2 = 0. Now

x 3 − x 2 + 3x − 2 ≡ x(x 2 + x + 1) (mod 2),

where x 2 + x + 1 is irreducible (mod 2), so by Theorem 10.3.1 we have

2 = P1 Q 1 ,

where P1 and Q 1 are the distinct prime ideals given by

θ + θ2
P1 = 2, α = 2, , N (P) = 2,
2
Q 1 = 2, 1 + α + α 2 = 2, −1 + θ 2 , N (Q) = 4.
CB609-10 CB609/Alaca & Williams August 7, 2003 16:53 Char Count= 0

10.5 Factoring Primes in an Arbitrary Number Field 259

If we had in error applied Theorem 10.3.1 directly to the prime 2, we would have
obtained the incorrect factorization
2 = 2, θ 2, 1 + θ2 ,
as
x 3 − x + 4 ≡ x(x + 1)2 (mod 2),
showing that the condition p ind(θ) is essential in Theorem 10.5.1.

If K is an algebraic number field of degree n and p is a rational prime such that

p = P n for some prime ideal P of O K , then we say that p is completely ramified
in K . If K = Q(θ ) with θ ∈ O K and the polynomial irrQ θ is p-Eisenstein then p
completely ramifies.

Theorem 10.5.2 Let K = Q(θ) be an algebraic number field of degree n with

θ ∈ O K . Let x n + an−1 x n−1 + · · · + a1 x + a0 ∈ Z[x] be the minimal polynomial
of θ over Q. If p is a prime such that
p || a0 , p | a1 , . . . , p | an−1
then
p = P n
in O K for some prime ideal P.

Proof: Let P be a prime ideal of O K that divides p. As

θ n = −an−1 θ n−1 − · · · − a1 θ − a0 ,
and each of a0 , a1 , . . . , an−1 is divisible by p, we see that
P | θn .
As P is a prime ideal, we deduce that
P | θ .
Thus we can define positive integers r and s by
P r || p, P s || θ.
Then, from
a0 + θ n = −an−1 θ n−1 − · · · − a1 θ,
as each of a1 , . . . , an−1 is divisible by p, we obtain
P r +1 | a0 + θ n = a0 + θn .
CB609-10 CB609/Alaca & Williams August 7, 2003 16:53 Char Count= 0

260 Factoring Primes in a Number Field

But p || a0 , so P r || a0 and thus P r || θn . Hence r = sn and so P sn || p. Thus

P n | p and by (10.1.3) we have p = P n as asserted.

Example 10.5.2 Let K be the cubic field Q(θ), where θ 3 − 2θ + 2 = 0. As irrQ θ =

x 3 − 2x + 2 is 2-Eisenstein, by Theorem 10.5.2, we have 2 = P 3 for some prime
ideal P. Indeed P = 2, θ as
2, θ 3 = 8, 4θ, 2θ 2 , θ 3 = 8, 4θ, 2θ 2 , 2θ − 2
= 24, 2θ, θ 2 , θ − 1 = 2 = P 3 ,
as 1 = θ 2 − (θ + 1)(θ − 1) ∈ 4, 2θ, θ 2 , θ − 1.

10.6 Factoring Primes in a Cyclotomic Field

Let m be a positive integer and let ζm be a primitive mth root of unity. The cyclotomic
field Q(ζm ) is denoted by K m . We give (without proof) the decomposition of a
rational prime p into prime ideals in O K m .

Theorem 10.6.1 Let m = pr m 1 , where r ∈ N ∪ {0}, m 1 ∈ N, and p m 1 . Let h

be the least positive integer such that p h ≡ 1 (mod m 1 ). Then h | φ(m 1 ) and
p = (P1 P2 · · · Pφ(m 1 )/ h )φ( p ) ,
r

where P1 , P2 , . . . , Pφ(m 1 )/ h are distinct prime ideals with

N (Pi ) = p h , i = 1, 2, . . . , φ(m 1 )/ h.

We refer the reader to Mann’s book [6] for a proof of this theorem.

Example 10.6.1 We determine the prime ideal decomposition of 3 in O K 9 . Here

p = 3, m = 9, φ(m) = 6, r = 2, m 1 = 1, and h = 1 so that by Theorem 10.6.1
3 = P 6 ,
where P is a prime ideal with N (P) = 3.

Example 10.6.2 We determine the prime ideal decomposition of 2 in O K 5 . Here

p = 2, m = 5, r = 0, m 1 = 5, φ(m 1 ) = 4, φ( pr ) = 1, and h = 4 so that by
Theorem 10.6.1
2 = P,
where P is a prime ideal with N (P) = 24 .

Example 10.6.3 We determine the prime ideal decomposition of 2 in O K 7 . Here

p = 2, m = 7, r = 0, m 1 = 7, φ(m 1 ) = 6, φ( pr ) = 1, and h = 3 so that by
CB609-10 CB609/Alaca & Williams August 7, 2003 16:53 Char Count= 0

Exercises 261

Theorem 10.6.1
2 = P1 P2 ,
where P1 and P2 are distinct prime ideals with
N (P1 ) = N (P2 ) = 23 .
By Theorem 7.5.2 the cyclotomic field K 7 is monogenic so we can apply Theorem
10.3.1 to obtain P1 and P2 explicitly. We have
x7 − 1
irrQ (ζ7 ) = = x6 + x5 + x4 + x3 + x2 + x + 1
x −1
and
x 6 + x 5 + x 4 + x 3 + x 2 + x + 1 ≡ (x 3 + x + 1)(x 3 + x 2 + 1) (mod 2),
so that
P1 = 2, 1 + ζ7 + ζ73 , P2 = 2, 1 + ζ72 + ζ73 .

Exercises
1. In Example 10.4.2 show that 721 + 1299θ + 393θ 2 is a unit by finding its norm.
2. Factor 2 into prime ideals in OQ(√47) .
3. Factor 6 into prime ideals in OQ(√366) .
4. Factor 2 into prime ideals in OQ( √3 2) .
5. Factor 2 into prime ideals in OQ( √3 3) .
6. Factor 2 into prime ideals in OQ(√2+√−1) .
7. Prove√(10.1.2) and (10.1.3).
8. Is Q( 3 10) monogenic?
9. Modify the proof
√ of Theorem 10.3.1 to prove Theorem 10.5.1. √
10. Let K = Q( 3). In O K we have 3 = P 3 , where P = 3 3 is a prime ideal of norm
3

3. Are there any rational primes p = 3 such that p = Q 3 in O K for some prime ideal
Q? √
11. Determine all rational primes p that ramify in Q( 3 6) together with their prime ideal
factorizations. √ √
12. Determine the prime ideal decomposition of the prime 47 in Q( 2, 3).
13. Let K = Q(θ), where θ 3 − θ + 4 = 0. The ideal I = 2, θ is principal in O K . Find a
generator of I .
14. Factor 5 into prime ideals in O K 5 .
15. Factor 3 into prime ideals in O K 7 .
16. As ζm is a unit of OQ(ζm ) , we know that N (ζm ) = ±1. Show that the + sign holds.
17. Prove that 1 + ζm + ζm2 + · · · + ζmk−1 is a unit of OQ(ζm ) if k is a positive integer coprime
with m.
18. Let K 1 and K 2 be algebraic number fields. Suppose that the prime p is totally ramified
in O K 1 and unramified in O K 2 . Prove that K 1 ∩ K 2 = Q.
CB609-10 CB609/Alaca & Williams August 7, 2003 16:53 Char Count= 0

262 Factoring Primes in a Number Field

√
19. Let K = Q(θ ), where θ 3 − θ − 1 = 0. Prove that θ ∈ Q(θ ).
20. Let Q(θ1 ) and Q(θ2 ) be algebraic number fields. Prove that

[Q(θ1 , θ2 ) : Q] ≤ [Q(θ1 ) : Q][Q(θ2 ) : Q].

21. If [Q(θ1 ) : Q] and [Q(θ2 ) : Q] are coprime, prove that equality holds in the inequality
in Exercise 20.
22. Let p be an odd prime. Prove that

j −j
(1 − ζ p )(1 − ζ p )
, j = 1, 2, . . . , p − 1,
(1 − ζ p )(1 − ζ p−1 )
are real units of OQ(ζ p ) .
23. Let

α = 1 + ζ23
2
+ ζ23
4
+ ζ23
5
+ ζ23
6
+ ζ23
10
+ ζ23
11

and

β = 1 + ζ23 + ζ23
5
+ ζ23
6
+ ζ23
7
+ ζ23
9
+ ζ23
11
.

Prove that 2 is not a prime in OQ(ζ23 ) by considering the divisibility of αβ by 2.

24. Prove that 2 is an irreducible in OQ(ζ23 ) .
25. What can you deduce from Exercises 23 and 24 about OQ(ζ23 ) ?

Suggested Reading
1. G. Bachman, The decomposition of a rational prime ideal in cyclotomic fields, American
Mathematical Monthly 73 (1966), 494–497.
An alternate proof of the way a rational prime ideal decomposes in the ring of integers of a
cyclotomic field is given
2. Z. I. Borevich and I. R. Shafarevich, Number Theory, Academic Press, New York and
London, 1966.
Example 10.4.2 is based upon Example 2, p. 230.
3. R. Dedekind, Über den Zusammenhang zwischen der Theorie der Ideale und der Theorie
der höheren Kongruenzen, Abh. Kgl. Ges. Wiss. Göttingen 23 (1878), 1–23. (Gesam-
melte Mathematische Werke I, pp. 202–232, Vieweg, Wiesbaden, 1930.)
Theorem 10.3.1 is Theorem 1 on pages 212 and 213 of Dedekind’s Collected Papers.
4. I. Gerst and J. Brillhart, On the prime divisors of polynomials, American Mathematical
Monthly 78 (1971), 250–260.
The set of all primes for which an irreducible polynomial has at least one linear factor (mod p) is
considered.
5. J. G. Huard, B. K. Spearman, and K. S. Williams, The primes for which an abelian cubic
polynomial splits, Tokyo Journal of Mathematics 17 (1994), 467–478.
Let x 3 + ax + b ∈ Z[x] be an irreducible abelian cubic polynomial. Explicit integers a1 , . . . , an , m
are determined such that x 3 + ax + b ≡ 0 (mod p) has three solutions ⇐⇒ p ≡ a1 , . . . , an
(mod m) except for finitely many primes p.
CB609-10 CB609/Alaca & Williams August 7, 2003 16:53 Char Count= 0

6. H. B. Mann, Introduction to Algebraic Number Theory, Ohio State University Press,

Columbus, Ohio, 1955.
For a proof of Theorem 10.6.1, see Theorems 8.7 and 8.8 in Chapter 8.
7. B. F. Wyman, What is a reciprocity law?, American Mathematical Monthly 79 (1972),
571–586.
The set of primes p for which an irreducible polynomial factors (mod p) into a product of linear
factors is discussed.
CB609-11 CB609/Alaca & Williams August 7, 2003 16:58 Char Count= 0

11
Units in Real Quadratic Fields

√
11.1 The Units of Z + Z 2
In Theorem 5.4.3 we determined the unit group U (O K ) for an imaginary quadratic
field K . The objective of this chapter is to determine the structure of the unit group
U (O K ) for an arbitrary real quadratic field K . We show that

U (O K ) Z2 × Z

(see Theorems 11.5.1 and 11.5.2). This is accomplished by showing that there exists
a unit in O K such that every unit is of the form ± n (n ∈ Z). We show further
that there exists a unique unit > 1 of O K with this property. This unit is called the
fundamental unit of O K (or of K ). In Section 11.6 we show how continued fractions
can be used to determine the fundamental unit. In Chapter 13 we prove Dirichlet’s
unit theorem, which gives the structure of U (O K ) for an arbitrary algebraic number
field K .
To illustrate some of the ideas that will be involved, we begin by determining
√
U (OQ(√2) ) = U (Z + Z 2).

√ √
Theorem 11.1.1 All the units of Z + Z 2 are given by ±(1 + 2)n (n ∈ Z), so
that
√
U (Z + Z 2) Z2 × Z.
√
Proof: We begin by showing that there does not exist a unit λ of Z + Z 2 satisfying
√
1 < λ < 1 + 2. (11.1.1)

Suppose on the contrary that such a unit λ exists having property (11.1.1).
√ By
Theorem 6.2.1 there are exactly two monomorphisms σ1 and σ2 : Q( 2) → C.
These monomorphisms are given by
√ √ √ √
σ1 (x + y 2) = x + y 2, σ2 (x + y 2) = x − y 2,

264
CB609-11 CB609/Alaca & Williams August 7, 2003 16:58 Char Count= 0

√
11.1 The Units of Z + Z 2 265
for all x, y ∈ Q. As λ is a unit we have λ | 1 so that
√
1 = λµ for some µ ∈ Z + Z 2. (11.1.2)

Set λ = σ2 (λ) and µ = σ2 (µ). Applying σ2 to (11.1.2) we obtain

1 = σ2 (1) = σ2 (λµ) = σ2 (λ)σ2 (µ) = λ µ .

Hence

1 = (λλ )(µµ ).

But λλ ∈ Z and µµ ∈ Z so that

λλ = ±1.

We consider two cases: (i) λλ = 1 and (ii) λλ = −1.

Case (i): λλ = 1. In this case by (11.1.1) we have

√ 1
2−1= √ < λ < 1
1+ 2
so that
√ √
2 < λ + λ < 2 + 2

and thus
1 λ + λ 1
0.7 < √ < < 1 + √ < 1.8.
2 2 2
As (λ + λ )/2 ∈ Z we must have (λ + λ )/2 = 1. From λλ = 1 and λ + λ = 2
we deduce that λ = λ = 1, contradicting λ > 1.

Case (ii): λλ = −1. In this case by (11.1.1) we have

√
−1 < λ < 1 − 2

so that

0 < λ + λ < 2;

that is,
λ + λ
0< < 1.
2
This is a contradiction as (λ + λ )/2 ∈ Z. √
This
√ completes the proof that there are no units of Z + Z 2 between 1 and
1 + 2.
CB609-11 CB609/Alaca & Williams August 7, 2003 16:58 Char Count= 0

266 Units in Real Quadratic Fields

√
Now let η be√any unit > 1. Since there is no unit between 1 and 1 + 2 we must
have η ≥ 1 + 2. Then there exists a unique positive integer n such that
√ √
(1 + 2)n ≤ η < (1 + 2)n+1 .

Thus
√ −n √
1 ≤ η(1 + 2) < 1 + 2.
√ √
As η(1 + −2)−n is a unit of Z + Z 2, we have
√
η = (1 + 2)n , n ∈ N. (11.1.3)

If η is a unit with 0 < η < 1 then 1/η is a unit with 1/η > 1. Hence, from (11.1.3)
we have
1 √
= (1 + 2)n
η
for some n ∈ N, so that
√
η = (1 + 2)−n , n ∈ N.

If η is a unit with −1 < η < 0 then −1/η is a unit with −1/η > 1. Hence, by
(11.1.3), there exists n ∈ N such that
−1 √
= (1 + 2)n ,
η
so that
√ −n
η = −(1 + 2) , n ∈ N.

If η is a unit with η < −1 then −η is a unit with −η > 1. Hence, by (11.1.3),

there exists n ∈ N such that
√
−η = (1 + 2)n ,

so that
√ n
η = −(1 + 2) , n ∈ N.

Clearly
√
±1 = ±(1 + 2)0 .

Hence every unit η is given by

√
η = ±(1 + 2)k , k ∈ Z.

This completes the proof that

√
U (Z + Z 2) Z2 × Z.
CB609-11 CB609/Alaca & Williams August 7, 2003 16:58 Char Count= 0

11.2 The Equation x 2 − my 2 = 1 267

11.2 The Equation x 2 − my 2 = 1

In this section we show that there exist integers x and y with (x, y) = (±1, 0) such
that x 2 − my 2 = 1, where m is a positive integer that is not a perfect square. This
result tells us that
√
x + y m (= ±1) ∈ U (O K ),
√
where K = Q( m).
Euler (1707–1783) attributed to the English mathematician John Pell (1611–
1685) a method of solving the equation x 2 − my 2 = 1 in integers x and y. Thus
the equation has become known as the Pell equation. However, this method had
been found by another English mathematician, William Brouncker (1620–1684),
in a series of letters (1657–1658) to Pierre Fermat (1601–1665). Lagrange (1736–
1813) was the first mathematician to prove that the equation x 2 − my 2 = 1 has
infinitely many solutions in integers x and y.

Theorem 11.2.1 Let m be a positive integer that is not a perfect square. Then there
exist integers x and y with (x, y) = (±1, 0) such that

x 2 − my 2 = 1.

Proof: Let N be a positive integer. We show first that there exist integers x and y
such that
√ 1
0 < |x − y m| < , 0 < y ≤ N . (11.2.1)
N
We divide the interval 0 < x ≤ 1 into N subintervals r/N < x ≤ (r + 1)/N , r =
0, 1, . . . , N − 1, each of the same length 1/N . For i = 0, 1, . . . , N we define the
integers xi and yi by
√
xi = [i m] + 1, yi = i.

Now
√ √ √
[i m] ≤ i m < [i m] + 1

so that
√
xi − 1 ≤ yi m < xi ;

that is,
√
0 < xi − yi m ≤ 1, i = 0, 1, . . . , N .
√
Thus we have N + 1 numbers xi − yi m lying in the interval 0 < x ≤ 1. Hence
at least two of these numbers lie in the same subinterval (r/N , (r + 1)/N ]; that is,
CB609-11 CB609/Alaca & Williams August 7, 2003 16:58 Char Count= 0

268 Units in Real Quadratic Fields

there exist integers i and j with i = j, 0 ≤ i ≤ N , 0 ≤ j ≤ N such that

√
r/N < xi − yi m ≤ (r + 1)/N

and
√
r/N < x j − y j m ≤ (r + 1)/N .

Interchanging i and j, if necessary, we may suppose that

√ √
xi − yi m ≥ x j − y j m,

so that
√ √ 1
0 ≤ (xi − yi m) − (x j − y j m) < .
N
We note that yi − y j = i − j = 0. We define the integers x and y by

(xi − x j , yi − y j ), if yi − y j > 0,
(x, y) =
(x j − xi , y j − yi ), if yi − y j < 0.
Thus
−1 √ 1
< x − y m < , y > 0,
N N
and

y = |y| = |yi − y j | = |i − j| ≤ N .

This completes the proof of (11.2.1).

Next we show that there exist infinitely many pairs of integers (x, y) with y = 0
such that
√
0 < |x 2 − my 2 | < 1 + 2 m. (11.2.2)

Let N1 be any positive integer. By (11.2.1) there exist integers x1 and y1 such that
√ 1
0 < |x1 − y1 m| < , 0 < y1 ≤ N1 .
N1
√
Now let N2 be any positive integer > 1/|x1 − y1 m|. By (11.2.1) there exist inte-
gers x2 and y2 such that
√ 1
0 < |x2 − y2 m| < , 0 < y2 ≤ N2 .
N2
Nr , xr , yr (r = 1, 2, . . . , k − 1), we choose
Continuing in this way, after obtaining √
Nk to be any integer > 1/|xk−1 − yk−1 m| and integers xk and yk (> 0) such that
√ 1
0 < |xk − yk m| < , 0 < yk ≤ Nk .
Nk
CB609-11 CB609/Alaca & Williams August 7, 2003 16:58 Char Count= 0

11.2 The Equation x 2 − my 2 = 1 269

Clearly
√ 1 √ 1 √
0 < |xk − yk m| < < |xk−1 − yk−1 m| < < |xk−2 − yk−2 m|
Nk Nk−1
√ 1 √
< · · · < |x2 − y2 m| < < |x1 − y1 m|
N2
so that (xk , yk ) (k = 1, 2, . . .) is an infinite sequence of pairs of integers satisfying
√ 1 1
0 < |xk − yk m| < ≤ .
Nk yk
Hence, as m is not a perfect square and yk > 0, we have
√ √ √ 1 √
0 < |xk + yk m| ≤ |xk − yk m| + 2yk m < + 2yk m.
yk
Then, as
√ √
|xk2 − myk2 | = |xk − yk m||xk + yk m|,
we deduce that
1 1 √ 1 √ √
0 < |xk2 − myk2 | < ( + 2yk m) = 2 + 2 m ≤ 1 + 2 m,
yk yk yk
for k = 1, 2, . . . , proving that there are infinitely many pairs of integers (x, y) with
y > 0 satisfying (11.2.2), namely, (x, y) = (xk , yk ) (k = 1, 2, . . .).
√
From (11.2.2) we see that there is an integer t with 0 < |t| < 1 + 2 m for which
the equation
x 2 − my 2 = t (11.2.3)
has infinitely many distinct solutions in integers x and y. Replacing x by −x and
y by −y if necessary we see that (11.2.3) has infinitely many distinct solutions in
positive integers x and y. Let t be such that (11.2.3) has infinitely many solutions
in positive integers x and y, with |t| minimal. As m is not a perfect square we see
that |t| > 0. We show next that (t, y) = 1 for infinitely many (indeed, for all but
finitely many) of these solutions. Suppose on the contrary that (t, y) > 1. As t has
only finitely many prime factors, there exists at least one prime divisor p of t for
which Eq. (11.2.3) has infinitely many solutions in integers x and y with p | y. For
each such solution we have p | x so that p 2 | t and we conclude that the equation
x 2 − my 2 = t/ p 2
has infinitely many solutions in positive integers x and y, contradicting that |t| is
the least such integer with this property.
Let (x, y) be one of the solutions of (11.2.3) in positive integers so that (t, y) = 1.
Let u be the unique integer such that
uy ≡ 1 (mod t), 0 < u < |t|.
CB609-11 CB609/Alaca & Williams August 7, 2003 16:58 Char Count= 0

270 Units in Real Quadratic Fields

Since there are |t| residue classes modulo t, we can find two such solutions, say
(x1 , y1 ) and (x2 , y2 ), such that

u 1 x1 ≡ u 2 x2 (mod t).

Then we have
√ √ √
x1 + m y1 (x1 + m y1 )(x2 − m y2 )
√ =
x2 + m y2 x22 − my22
√
(x1 x2 − my1 y2 ) + m(x2 y1 − x1 y2 )
=
√ t
= x + m y,

where

x = (x1 x2 − my1 y2 )/t, y = (x2 y1 − x1 y2 )/t.

Clearly x ∈ Q and y ∈ Q. We show that x ∈ Z and y ∈ Z. We have

u 1 u 2 (x2 y1 − y2 x1 ) = (u 1 y1 )(u 2 x2 ) − (u 2 y2 )(u 1 x1 )

≡ u 2 x2 − u 1 x1
≡ 0 (mod t).

Now (u 1 u 2 , t) = 1 so that

x2 y1 − y2 x1 ≡ 0 (mod t),

proving that y ∈ Z. Similarly, we have (as u 1 x1 ≡ u 2 x2 (mod t) and u 1 y1 ≡

u 2 y2 (mod t))

u 1 u 2 (x1 x2 − my1 y2 ) = (u 1 x1 )(u 2 x2 ) − m(u 1 y1 )(u 2 y2 )

≡ (u 1 x1 )2 − m(u 1 y1 )2
= u 21 (x12 − my12 )
= u 21 t
≡ 0 (mod t),

so that as (u 1 u 2 , t) = 1 we have

x1 x2 − my1 y2 ≡ 0 (mod t),

proving that x ∈ Z. Hence

√ √ √
x1 + m y1 = (x2 + m y2 )(x + m y)
CB609-11 CB609/Alaca & Williams August 7, 2003 16:58 Char Count= 0

11.3 Units of Norm 1 271

and so

t = x12 − my12 = (x22 − my22 )(x 2 − my 2 ) = t(x 2 − my 2 )

so that

x 2 − my 2 = 1.

Now if (x, y) = (±1, 0) then

√ √
x1 + m y1 = ±(x2 + m y2 )

so that

(x1 , y1 ) = ±(x2 , y2 ).

But x1 > 0 and x2 > 0, so (x1 , y1 ) = (x2 , y2 ), contradicting that (x1 , y1 ) and (x2 , y2 )
are distinct solutions of (11.2.3).
Hence we have shown the existence of a pair of integers (x, y) = (±1, 0) such
that x 2 − my 2 = 1.

11.3 Units of Norm 1

Let m be a positive squarefree integer. Theorem 11.2.1 tells us that there exist
√
positive integers x and y such that x 2 − my 2 = 1. Hence λ = x + y m is a unit
√
of O K , where K = Q( m), such that λ > 1 and N (λ) = 1. Since λn → ∞ as
n → ∞, O K has infinitely many units of norm 1, namely {λn | n ∈ Z}. All of these
√
units are of the form u + v m, where u and v are integers such that u 2 − mv 2 = 1.
√
However, when m ≡ 1 (mod 4), there may be units in O K√of the form (u + v m)/2,
where u and v are both odd integers. For example (3 + 5)/2 is a unit of norm
√ 1 in
√ √
OQ( 5) . In contrast, OQ( 17) does not contain any units of the form (u + v 17)/2,
where u and v are both odd integers, since u 2 − 17v 2 = ±4 cannot hold modulo 8
for odd integers u and v.
√ √
Let λ = x + y m be a unit of O K (K = Q( m)) of norm 1 with x and y
both integers or possibly in the case m ≡ 1 (mod 4) both halves of odd integers.
We now show how the signs of x and y determine to which of the four intervals
(−∞, −1), (−1, 0), (0, 1), or (1, ∞) λ belongs.

Theorem 11.3.1 Let m be a positive squarefree integer. Let x and y both be integers
or both halves of odd integers such that x 2 − my 2 = 1. Then
√
x + y m > 1 ⇐⇒ x > 0, y > 0, (11.3.1)
√
0 < x + y m < 1 ⇐⇒ x > 0, y < 0, (11.3.2)
√
−1 < x + y m < 0 ⇐⇒ x < 0, y > 0, (11.3.3)
√
x + y m < −1 ⇐⇒ x < 0, y < 0. (11.3.4)
CB609-11 CB609/Alaca & Williams August 7, 2003 16:58 Char Count= 0

272 Units in Real Quadratic Fields

Proof: First we prove (11.3.1). We have

√ √
1 1 √ 1+ m 1+ 2
x > 0, y > 0 =⇒ x ≥ , y ≥ =⇒ x + y m ≥ ≥ > 1.
2 2 2 2
√ √
Conversely, as (x + y m)(x − y m) = x 2 − my 2 = 1, we have
√ √
x + y m > 1 =⇒ 0 < x − y m < 1
 1 √ √ 1

 x = ((x + y m) + (x − y m)) > > 0,
=⇒ 2 2
 1 √ √ 1−1
 y = √ ((x + y m) − (x − y m)) > √ = 0.
2 m 2 m
This proves (11.3.1).
Next we prove (11.3.2). We have
1 1
x > 0, y < 0 =⇒ x > 0, −y > 0 =⇒ x ≥ , − y ≥
√ 2 √ 2
√ 1+ m 1+ 2
=⇒ x − y m ≥ ≥ >1
√ 2 2
=⇒ 0 < x + y m < 1,
√ √
as (x − y m)(x + y m) = 1. Conversely,
√ √
0 < x + y m < 1 =⇒ x − y m > 1
 1 √ √ 1

 x = ((x + y m) + (x − y m)) > > 0,
=⇒ 2 2
 1 √ √ 1−1
 y = √ ((x + y m) − (x − y m)) < √ = 0.
2 m 2 m
This proves (11.3.2).
Finally, (11.3.4) follows from (11.3.1) and (11.3.3) follows from (11.3.2) by
changing x to −x and y to −y.

Definition 11.3.1 (Fundamental unit of norm 1) Let m be a positive squarefree

integer. Let

Sm = {(x, y) | x ∈ N, y ∈ N}, if m ≡ 2, 3 (mod 4),

and
x y
Sm = {( , ) | x ∈ N, y ∈ N, x ≡ y (mod 2)}, if m ≡ 1 (mod 4).
2 2
Let (a, b) ∈ Sm be the solution of a 2 − mb2 = 1 for which a is least. (Theorem
√
11.2.1 guarantees that (a, b) exists.) Let = a + b m so that is a unit of OQ(√m)
of norm 1. The unit is called the fundamental unit of norm 1 of OQ(√m) .
CB609-11 CB609/Alaca & Williams August 7, 2003 16:58 Char Count= 0

11.3 Units of Norm 1 273

We note that
√ √
≥ 1 + m ≥ 1 + 2, if m ≡ 2, 3 (mod 4),
√ √
1+ m 1+ 5
≥ ≥ , if m ≡ 1 (mod 4),
2 2
so that

> 1.

Our next theorem shows how the units of norm 1 in OQ(√m) are related to the
fundamental unit of norm 1.

Theorem 11.3.2 Let m be a positive squarefree integer. Let be the fundamental

unit of norm 1 of OQ(√m) . Then

(a) is the smallest unit in OQ(√m) of norm 1 that is greater than 1,

(b) every unit in OQ(√m) of norm 1 is of the form ± n for some integer n, and
(c) if τ is a unit of norm 1 in OQ(√m) such that τ > 1 and every unit in OQ(√m) of norm 1
is of the form ±τ k for some integer k then τ = .

Proof: (a) As is the fundamental unit of OQ(√m) of norm 1, we have by Definition

11.3.1
√
= a + b m, (a, b) ∈ Sm , a 2 − mb2 = 1, a least.

Suppose that 1 is a unit of OQ(√m) of norm 1 with 1 < 1 < . Then, by Theorems
5.4.2 and 11.3.1, we have
√
1 = a1 + b1 m, (a1 , b1 ) ∈ Sm , a12 − mb12 = 1.

By the minimality of a we have

a < a1

so that
a2 − 1 a2 − 1
b2 = < 1 = b12 ,
m m
and thus

b < b1 .

Hence
√ √
= a + b m < a1 + b1 m = 1 ,

contradicting 1 < . Thus no such unit 1 exists, proving that is the smallest unit
of OQ(√m) of norm 1 that is greater than 1.
CB609-11 CB609/Alaca & Williams August 7, 2003 16:58 Char Count= 0

274 Units in Real Quadratic Fields

(b) Let η be a unit of OQ(√m) of norm 1. Let η∗ be the unit of OQ(√m) of norm 1
defined by


 η, if η ≥ 1,



 1/η, if 0 < η < 1,
η∗ = (11.3.5)

 −1/η, if − 1 < η < 0,




−η, if η ≤ −1,
so that
η∗ ≥ 1.
Let k be the unique nonnegative integer such that
k ≤ η∗ < k+1 .
Then η∗ −k is a unit of OQ(√m) of norm 1 satisfying
1 ≤ η∗ −k < .
By part (a) there is no unit in OQ(√m) of norm 1 strictly between 1 and . Hence
η∗ −k = 1
and so
η∗ = k .
Then, from (11.3.5), we obtain
η = ± n
for some choice of sign and some integer n.
(c) By assumption we have
= ±τ l ,
for some integer l, and by part (b) we have
τ = ± n ,
for some integer n. Hence
= ±(± n )l = ± ln
so that
ln−1 = ±1
and thus
2(ln−1) = 1.
CB609-11 CB609/Alaca & Williams August 7, 2003 16:58 Char Count= 0

11.4 Units of Norm −1 275

√
If ln − 1 = 0 then is a root of unity in OQ(√m) . But Q( m) is a real field so the
only roots of unity in OQ(√m) are ±1. Hence = ±1, contradicting > 1. Thus
ln − 1 = 0 and so
l = n = ±1,
showing that
τ = ± or ± −1 .
Since τ > 1 and > 1 we deduce that
τ = .

11.4 Units of Norm −1

Let m be a positive squarefree integer. We have already observed that the ring
√
OQ(√m) of integers of the real quadratic field Q( m) may or may not contain units
√
of norm −1. Indeed OQ(√2) has units such as 1 + 2 of norm −1 whereas OQ(√3)
does not contain any units of norm −1. We suppose that OQ(√m) contains units of
norm −1 and show that there exists a unique unit σ > 1 in OQ(√m) of norm −1
such that all units in OQ(√m) of norm −1 are given by ±σ 2k+1 (k = 0, ±1, ±2, . . .)
and all units in OQ(√m) of norm 1 are given by ±σ 2k (k = 0, ±1, ±2, . . .).

Theorem 11.4.1 Let m be a positive squarefree integer. Suppose that OQ(√m) con-
tains units of norm −1. Then there exists a unique unit σ > 1 of norm −1 in OQ(√m)
such that every unit in OQ(√m) is of the form ±σ n for some integer n.

Proof: Let ρ be a unit in OQ(√m) of norm −1. Let ρ denote its conjugate. Then
ρρ = N (ρ) = −1
so that
ρ 2 ρ = 1.
2

Thus ρ 2 is a unit of OQ(√m) of norm 1. Hence, by Theorem 11.3.2(b), we have

ρ 2 = ± n ,
for some integer n, where is the fundamental unit of OQ(√m) of norm 1. Clearly
ρ 2 > 0 and n > 0 so that
ρ2 = n .
If n is even, say n = 2k, then
ρ 2 = 2k
CB609-11 CB609/Alaca & Williams August 7, 2003 16:58 Char Count= 0

276 Units in Real Quadratic Fields

so that
ρ = ± k .
Hence
N (ρ) = N (± k ) = N ()k = 1,
contradicting N (ρ) = −1. Thus n must be odd, say n = 2l + 1, and so
ρ 2 = 2l+1 .
Hence
= (ρ −l )2 .
Set
σ = ρ −l
so that σ is a unit of norm −1 such that
= σ 2.
If µ is a unit of OQ(√m) of norm −1 then µρ −1 is a unit of OQ(√m) of norm 1 and
thus by Theorem 11.3.2(b)
µρ −1 = ± k
for some k ∈ Z. Hence, as ρ = l σ and = σ 2 , we deduce that
µ = ± k ρ = ± k+l σ = ±σ 2(k+l)+1 .
However, if µ is a unit of OQ(√m) of norm 1 then by Theorem 11.3.2(b)
µ = ± k
for some k ∈ Z. Hence, as = σ 2 , we deduce that
µ = ± k = ±σ 2k .
Thus every unit of OQ(√m) is of the form
±σ n (n ∈ Z).
Note that n even gives the units of norm 1 and n odd the units of norm −1.
Replacing σ by 1/σ if 0 < σ < 1, by −1/σ if −1 < σ < 0, and by −σ if
σ < −1, we may suppose that σ > 1. We show that σ is uniquely determined: For
suppose σ and τ are two units > 1 of norm −1 in OQ(√m) such that every unit is
of the form ±σ n (n ∈ Z) and of the form ±τ q (q ∈ Z). Then there exist integers k
and l such that
σ = ±τ k , τ = ±σ l .
CB609-11 CB609/Alaca & Williams August 7, 2003 16:58 Char Count= 0

11.4 Units of Norm −1 277

Hence
σ = ±σ kl
and so
σ 2 = σ 2kl ,
giving
σ 2(kl−1) = 1.
Suppose kl − 1 = 0. Then σ is a root of unity. But OQ(√m) being a real field contains
no roots of unity except ±1. Thus σ = ±1, a contradiction. Hence kl − 1 = 0 and
so
k = l = ±1.
Thus
σ = ±τ or ± τ −1 .
But σ > 1 and τ > 1 so
σ = τ,
proving that σ is unique.

Definition 11.4.1 (Fundamental unit of norm −1) Let m be a positive squarefree

integer such that OQ(√m) contains units of norm −1. The unique unit σ > 1 of
norm −1 such that every unit in OQ(√m) is of the form ±σ n (n ∈ Z) is called the
fundamental unit of OQ(√m) of norm −1.

We next relate the fundamental unit of norm 1 and the fundamental unit σ of
norm −1 when OQ(√m) contains units of norm −1.

Theorem 11.4.2 Let m be a positive squarefree integer such that OQ(√m) contains
units of norm −1. Then the fundamental unit of norm 1 and the fundamental unit
σ of norm −1 are related by
= σ 2.

Proof: By Theorem 11.4.1 we have = ±σ k for some k ∈ Z. As > 1 and σ > 1

the plus sign must hold so that
= σk
for some k ∈ Z. Then
1 = N () = N (σ k ) = N (σ )k = (−1)k ,
CB609-11 CB609/Alaca & Williams August 7, 2003 16:58 Char Count= 0

278 Units in Real Quadratic Fields

so that k is even, say k = 2g, g ∈ Z. Hence

= σ 2g . (11.4.1)
Now
N (σ 2 ) = N (σ )2 = (−1)2 = 1
so that σ 2 is a unit of norm 1 and thus, by Theorem 11.3.2(b), we have σ 2 = ± l
for some l ∈ Z. As σ > 1 and > 1 the plus sign must hold so that
σ 2 = l (11.4.2)
for some l ∈ Z. From (11.4.1) and (11.4.2) we deduce that
= gl
so that
gl−1 = 1.
As is not a root of unity, we deduce that gl − 1 = 0, so that
= σ 2 or σ −2 .
As > 1 and σ > 1 we have
= σ2
as asserted.

11.5 The Fundamental Unit

Theorems 11.3.2 and 11.4.1 show that all the units of OQ(√m) are given by ± n (n ∈
Z) or by ±σ n (n ∈ Z) depending on whether OQ(√m) has only units of norm 1 or
not. This enables us to define the “fundamental unit” of OQ(√m) .

Definition 11.5.1 (Fundamental unit) Let m be a positive squarefree integer. The

fundamental unit η of OQ(√m) is defined to be σ if OQ(√m) contains units of norm
−1 and to be otherwise. We note that η > 1.

By Theorems 11.3.2 and 11.4.1 we have

Theorem 11.5.1 Let m be a positive squarefree integer. Then every unit of OQ(√m)
is of the form ±ηn (n ∈ Z), where η is the fundamental unit of OQ(√m) . If OQ(√m)
contains units of norm −1 these are given by ±ηn with n odd and the ones of norm
1 by ±ηn with n even.

From Theorem 11.5.1 we have immediately

CB609-11 CB609/Alaca & Williams August 7, 2003 16:58 Char Count= 0

11.5 The Fundamental Unit 279

Theorem 11.5.2 Let K be a real quadratic field. Then

U (O K ) Z2 × Z.

The following analogue of Theorem 11.3.2(a) is a simple consequence of The-

orem 11.5.1.

Theorem 11.5.3 Let K be a real quadratic field. The fundamental unit of O K is

the smallest unit of O K greater than 1.

Proof: Let η be the fundamental unit of O K and suppose that there exists a unit θ
of O K with

1 < θ < η.

By Theorem 11.5.1 we have

θ = ±ηn

for some n ∈ Z. As θ and η are both positive, the positive sign must hold and we
have

θ = ηn .

If n ≥ 1 then

θ = ηn ≥ η,

contradicting θ < η. If n ≤ 0 then

θ = ηn ≤ 1,

contradicting θ > 1. Hence no such θ can exist, proving that η is the smallest unit
greater than 1.

Before proceeding to find the norm of the fundamental unit η of OQ(√m) for
certain special values of m, we present in Table 4 the values of , σ , and η for
squarefree positive integers m < 40.
We next determine the norm of the fundamental unit of OQ(√m) when m is an
odd prime p. First we consider the case p ≡ 1 (mod 4).

Theorem 11.5.4 Let p be a prime with p ≡ 1 (mod 4). Then the fundamental unit
of OQ(√ p) has norm −1.

We give two proofs of this theorem, the first due to Hilbert and the second due
to Peter Gustav Lejeune Dirichlet (1805–1859).
CB609-11 CB609/Alaca & Williams August 7, 2003 16:58 Char Count= 0

280 Units in Real Quadratic Fields

Table 4. Fundamental units of OQ(√m) , 2 ≤ m < 40, m squarefree

Fundamental unit of Fundamental unit of

norm 1 norm −1 Fundamental unit Norm
m () (σ ) (η) N (η)
√ √ √
2 3 + 2√ 2 1+ 2 1 + √2 −1
3 2 +√ 3 √ 2 +√ 3 1
5 (3 + √ 5)/2 (1 + 5)/2 (1 + √ 5)/2 −1
6 5 + 2√6 5 + 2√6 1
7 8 + 3√7 √ 8 + 3√ 7 1
10 19 + 6√10 3+ 10 3 + √10 −1
11 10 + 3√ 11 √ 10 +√3 11 1
13 (11 + 3 √13)/2 (3 + 13)/2 (3 + 13)/2
√ −1
14 15 + 4√ 14 15 + 4√ 14 1
15 4+ √ 15 √ 4 + √15 1
17 33 + 8 √ 17 4+ 17 4 + 17 √ −1
19 170 +√39 19 170 +√39 19 1
21 (5 + 21)/2√ (5 + 21)/2
√ 1
22 197 + 42√ 22 197 + 42 √ 22 1
23 24 + 5 √23 √ 24 + 5√ 23 1
26 51 + 10√ 26 5 +√ 26 5 +√ 26 −1
29 (27 + 5 √29)/2 (5 + 29)/2 (5 + 29)/2
√ −1
30 11 + 2 30 √ 11 + 2 30√ 1
31 1520 + 273 √ 31 1520 + 273√ 31 1
33 23 + 4√33 23 + 4√33 1
34 35 + 6
√ 34 35 + 6√ 34 1
35 6+ √ 35 √ 6 + √35 1
37 73 + 12√ 37 6+ 37 6 + √37 −1
38 37 + 6√38 37 + 6√38 1
39 25 + 4 39 25 + 4 39 1

First proof: Suppose that the fundamental unit η of OQ(√ p) has norm 1. Then

N (η) = ηη = 1.

As η > 1 we have 0 < η < 1 so that 1 + η = 0. Hence

1+η
η= .
1 + η

Let m be the largest positive integer such that

m | 1 + η, m | 1 + η
CB609-11 CB609/Alaca & Williams August 7, 2003 16:58 Char Count= 0

11.5 The Fundamental Unit 281

in OQ(√ p) . Set
1+η
γ = ∈ OQ(√ p)
m
and
1 + η
γ = ∈ OQ(√ p)
m
so that
γ
η=
γ
and

k | γ , k | γ , k ∈ N =⇒ k = 1. (11.5.1)

Now

γ = ηγ ,

where η is a unit of OQ(√ p) , so that

γ = γ . (11.5.2)

Let Q be any prime ideal of OQ(√ p) such that

Q | γ . (11.5.3)

Then, by (11.5.2), we have

Q | γ . (11.5.4)

Taking conjugate ideals in (11.5.4), we obtain

Q | γ . (11.5.5)

As Q is a prime ideal, and the discriminant of OQ(√ p) is p, by Theorem 10.2.2 we

have

p
Q = Q = q, where q is a rational prime with = −1,
q
or

p
Q=
Q , Q Q = q, where q is a rational prime with = 1,
q
or

p
Q = Q , Q = q, where q is a rational prime with
2
= 0.
q
CB609-11 CB609/Alaca & Williams August 7, 2003 16:58 Char Count= 0

282 Units in Real Quadratic Fields

In the first case, from (11.5.3) and (11.5.4), we deduce that q | γ , q | γ , con-
tradicting (11.5.1).
In the second case we have by (11.5.3) and (11.5.5) as Q and Q are distinct
prime ideals

q = Q Q | γ = γ .

Hence q | γ and q | γ , contradicting (11.5.1).

√ √
In the third case we have q = p and Q = p. Hence p is the only prime
ideal that can divide γ . Thus
√
γ = p j

for some nonnegative integer j. If j ≥ 2 then p | γ and p | γ , contradicting

(11.5.1). Hence j = 0 or 1. If j = 0 then

γ = 1

and so

γ =λ

for some unit λ of OQ(√ p) . As λλ = ±1 we have

γ λ λ2
η= = = = ±λ2 ,
γ λ λλ
contradicting that η is the fundamental unit of OQ(√ p) . If j = 1 then
√
γ = p

so that
√
γ =λ p

for some unit λ of OQ(√ p) . Hence

√
γ λ p −λ −λ2
η= = √ = = = ∓λ2 ,
γ λ (− p) λ λλ
again contradicting that η is the fundamental unit of OQ(√ p) .
This completes the proof that the fundamental unit of OQ(√ p) ( p (prime) ≡ 1
(mod 4)) must have norm −1.

Second proof: Suppose that the fundamental unit η of OQ(√ p) has norm 1. Then
√
x+y p
η= ,
2
CB609-11 CB609/Alaca & Williams August 7, 2003 16:58 Char Count= 0

11.5 The Fundamental Unit 283

where x and y are positive integers such that

x > 2, x ≡ y (mod 2), x 2 − py 2 = 4.
We first treat the case when x ≡ y ≡ 0 (mod 2). Set x = 2X, y = 2Y so that X
and Y are positive integers such that
√
η = X + Y p, X 2 − pY 2 = 1, X > 1.
As p ≡ 1 (mod 4), X is odd and Y is even, so that X 2−1 , X +1
2
, and Y
2
are positive
integers such that
2
(X − 1) (X + 1) Y
· =p .
2 2 2
As X 2+1 − X 2−1 = 1, the integers X −1
2
and X +1
2
are coprime. Since p | X −1
2
· X +1
2
either p | X 2−1 or p | X 2+1 .
If p | X 2−1 then
2
(X − 1) (X + 1) Y
· =
2p 2 2
and thus there are positive coprime integers A and B such that
X −1 X +1 Y
= A2 , = B 2, = AB,
2p 2 2
so that
X = 2 p A2 + 1 = 2B 2 − 1, Y = 2AB.
Hence
B 2 − p A2 = 1
√
so that B + A p is a unit of norm 1 in OQ(√ p) . Now
1 ≤ B ≤ B 2 ≤ 2B 2 − 1 = X
and
1 ≤ A < 2AB = Y
so that
√ √
1 < B + A p < X + Y p = η,
contradicting that η is the fundamental unit of OQ(√ p) .
If p | X 2+1 then
2
(X − 1) (X + 1) Y
· =
2 2p 2
CB609-11 CB609/Alaca & Williams August 7, 2003 16:58 Char Count= 0

284 Units in Real Quadratic Fields

and thus there are integers A and B such that

X −1 X +1 Y
= A2 , = B 2, = AB,
2 2p 2
so that
X = 2A2 + 1 = 2 p B 2 − 1, Y = 2AB.
Hence
A2 − p B 2 = −1
√
so that A + B p is a unit of norm −1 in OQ(√ p) , contradicting that all the units of
OQ(√ p) have norm 1.
Now we turn to the case when x ≡ y ≡ 1 (mod 2). Reducing x 2 − py 2 = 4
modulo 8 we see that p ≡ 5 (mod 8). Now x − 2, x + 2, and y are positive odd
integers such that
(x − 2)(x + 2) = py 2 .
As (x + 2) − (x − 2) = 4 the integers x − 2 and x + 2 are coprime. Since
p | (x − 2)(x + 2) either p | x − 2 or p | x + 2.
If p | x − 2 then
(x − 2)
· (x + 2) = y 2
p
and there exist positive coprime odd integers A and B such that
x −2
= A2 , x + 2 = B 2 , y = AB,
p
so that
x = p A2 + 2 = B 2 − 2, y = AB.
Hence
B 2 − p A2 = 4
√
so that (B + A p)/2 is a unit of norm 1 in OQ(√ p) . From B 2 − p A2 = 4 we see
that B = 1. Thus, as B is odd and positive, we must have B ≥ 3. Hence
1 < B < B2 − 2 = x
and
1 ≤ A < AB = y
so that
√ √
B+A p x+y p
1< < ,
2 2
CB609-11 CB609/Alaca & Williams August 7, 2003 16:58 Char Count= 0

11.5 The Fundamental Unit 285

√
contradicting that (x + y p)/2 is the fundamental unit of norm 1 in OQ(√ p) by
Theorem 11.3.2(a).
If p | x + 2 then
(x + 2)
(x − 2) = y2
p
and thus there are positive coprime odd integers A and B such that
x +2
x − 2 = A2 , = B 2 , y = AB,
p
so that
x = A2 + 2 = p B 2 − 2, y = AB.
Hence
A2 − p B 2 = −4
√
so that (A + B p)/2 is a unit of norm −1 in OQ(√ p) , contradicting that the fun-
damental unit of OQ(√ p) has norm 1.
This completes the proof of Theorem 11.5.4 using Dirichlet’s method.

When m is a prime p ≡ 3 (mod 4) the fundamental unit of OQ(√ p) has norm 1.

This is a special case of the following theorem.

Theorem 11.5.5 Let m be a positive squarefree integer. If there exists a prime q ≡ 3

(mod 4) dividing m then the fundamental unit of OQ(√m) has norm 1.

Proof: Suppose that the fundamental unit η of OQ(√m) has norm −1. Then
√
x+y m
η= ,
2
where x and y are integers such that

x ≡ y ≡ 0 (mod 2), if m ≡ 2, 3 (mod 4),
x ≡ y (mod 2), if m ≡ 1 (mod 4),
and
x 2 − my 2
= N (η) = −1.
4
Hence
x 2 − my 2 = −4.
As q | m we deduce that
x 2 ≡ − 4 (mod q).
CB609-11 CB609/Alaca & Williams August 7, 2003 16:58 Char Count= 0

286 Units in Real Quadratic Fields

Thus

−1 −4
= =1
q q

so that q ≡ 1 (mod 4), contradicting that q ≡ 3 (mod 4). This proves that η must
have norm 1.

The following two theorems can be proved using Dirichlet’s method in a similar
manner to the second proof of Theorem 11.5.4.

Theorem 11.5.6 Let p be a prime with p ≡ 5 (mod 8). Then the fundamental unit
of OQ(√2 p) has norm −1.

Theorem 11.5.7 Let p and q be distinct primes such that

p
p ≡ q ≡ 1 (mod 4), = −1.
q

Then the fundamental unit of OQ(√ pq) has norm −1.

Up to this point we have said almost nothing about calculating the fundamental
unit η of OQ(√m) for a particular value of m. We address this problem in the next
section.

11.6 Calculating the Fundamental Unit

Let m be a positive squarefree integer. The standard method of calculating the
fundamental unit η of OQ(√m) is by means of the continued fraction expansion of
√
m. We assume that the reader is familiar with the basic properties of continued
fractions as found for example in Chapter 7 of the book on elementary number
theory by Niven, Zuckerman, and Montgomery [2]. We just recall the basic facts
that we shall need and refer the reader to [2] for proofs.
Given a positive squarefree integer m, we define a sequence α0 , α1 , α2 , . . . of
real numbers by
√
α0 = m (11.6.1)

and
1
αn+1 = , n = 0, 1, 2, . . . . (11.6.2)
αn − [αn ]
CB609-11 CB609/Alaca & Williams August 7, 2003 16:58 Char Count= 0

11.6 Calculating the Fundamental Unit 287

Example 11.6.1 If m = 31 we find that

√
α0 = 31,
√
1 1 5 + 31
α1 = =√ = ,
α0 − [α0 ] 31 − 5 6
√
1 1 1 + 31
α2 = = √ = ,
α1 − [α1 ] 5 + 31 5
−1
6 √
1 1 4 + 31
α3 = = √ = ,
α2 − [α2 ] 1 + 31 3
−1
5 √
1 1 5 + 31
α4 = = √ = ,
α3 − [α3 ] 4 + 31 2
−3
3 √
1 1 5 + 31
α5 = = √ = ,
α4 − [α4 ] 5 + 31 3
−5
2 √
1 1 4 + 31
α6 = = √ = ,
α5 − [α5 ] 5 + 31 5
−3
3 √
1 1 1 + 31
α7 = = √ = ,
α6 − [α6 ] 4 + 31 6
−1
5 √
1 1 5 + 31
α8 = = √ = ,
α7 − [α7 ] 1 + 31 1
−1
6 √
1 1 5 + 31
α9 = = √ = = α1 ,
α8 − [α8 ] 5 + 31 − 10 6
α10 = α2 , α11 = α3 , . . . .

Clearly each αn > 1 and

√
Pn + m
αn = , n = 0, 1, 2, . . . , (11.6.3)
Qn
where P0 = 0, Q 0 = 1, and Pn , Q n are positive integers for n ≥ 1. Moreover,
it is known that there exists a positive integer l such that αl+1 = α1 . It follows
from (11.6.2) that αl+2 = α2 , αl+3 = α3 , . . . so that the sequence {αn }n≥1 is purely
periodic. We set

an = [αn ], n = 0, 1, 2, . . . , (11.6.4)
CB609-11 CB609/Alaca & Williams August 7, 2003 16:58 Char Count= 0

288 Units in Real Quadratic Fields

so that {an }n≥1 is a sequence of positive integers. Since {αn }n≥1 is a purely periodic
sequence so is the sequence {an }n≥1 .

Example 11.6.1 (continued) For m = 31

{an }n≥0 = {5, 1, 1, 3, 5, 3, 1, 1, 10, 1, 1, 3, 5, 3, 1, 1, 10, . . .}.

Next we define two further sequences of integers {h n }n≥−1 and {kn }n≥−1 by

h −1 = 1, h 0 = a0 , h n = an h n−1 + h n−2 , n = 1, 2, . . . (11.6.5)

and

k−1 = 0, k0 = 1, kn = an kn−1 + kn−2 , n = 1, 2, . . . . (11.6.6)

All of the h n and kn are positive except for k−1 = 0.

Example 11.6.1 (continued) For m = 31

{h n }n≥−1 = {1, 5, 6, 11, 39, 206, 657, 863, 1520, 16063, 17583, . . .},
{kn }n≥−1 = {0, 1, 1, 2, 7, 37, 118, 155, 273, 2885, 3158, . . .}.

It is easily shown that (h n , kn ) = 1 and

hn 1
= a0 + , n = 0, 1, 2, . . . . (11.6.7)
kn 1
a1 +
. 1
a2 + . . +
1
an−1 +
an

To save space we abbreviate the fraction on the right-hand side of (11.6.7) by the
space-saving flat notation [a0 , a1 , a2 , . . . , an ] so that (11.6.7) becomes

hn
= [a0 , a1 , a2 , . . . , an ], n = 0, 1, 2, . . . . (11.6.8)
kn
√
It is known that limn→∞ h n /kn exists and is equal to m.
CB609-11 CB609/Alaca & Williams August 7, 2003 16:58 Char Count= 0

11.6 Calculating the Fundamental Unit 289

Example 11.6.1 (continued) For m = 31 we have to seven decimal places

h0 h1 6 h2 11 h3 39
= 5, = = 6, = = 5.5, = = 5.5714285,
k0 k1 1 k2 2 k3 7
h4 206 h5 657 h6 863
= = 5.5675675, = = 5.5677966, = = 5.5677419,
k4 37 k5 118 k6 155
h7 1520 h8 16063
= = 5.5677655, = = 5.5677642,
k7 273 k8 2885
h9 17583
= = 5.5677644,
k9 3158
√
and 31 = 5.5677643 . . . .
√
We write [a0 , a1 , a2 , . . .] for limn→∞ [a0 , a1 , a2 , . . . , an ], and we say that m
has the infinite continued fraction expansion
√
m = [a0 , a1 , a2 , . . .]. (11.6.9)
The convergents of the infinite continued fraction [a0 , a1 , a2 , . . .] are the rational
numbers h n /kn (n = 0, 1, 2, . . .). As we have already mentioned, the sequence
{αn }n≥1 is purely periodic. Let l be the least positive integer such that αl+n = αn
(equivalently, Pl+n = Pn , Q l+n = Q n ) for all positive integers √ n. The integer l is
called the period of the continued fraction expansion of d. Then
√
m = [a0 , a1 , . . . , al , a1 , . . . , al , a1 , . . . , al , . . .]
and we abbreviate this by
√
m = [a0 , a1 , . . . , al ]. (11.6.10)

Example 11.6.1 (continued) For m = 31 we have l = 8 and

√
31 = [5, 1, 1, 3, 5, 3, 1, 1, 10, 1, 1, 3, 5, 3, 1, 1, 10, 1, . . .]
= [5, 1, 1, 3, 5, 3, 1, 1, 10].
√
Clearly a0 = [ m] and it is further known that al = 2a0 . Putting
√ √
Pn + m Pn−1 + m
αn = , αn−1 = ,
Qn Q n−1
and [αn−1 ] = an−1 in
1
αn = (n ≥ 1),
αn−1 − [αn−1 ]
we obtain
√
Pn + m Q n−1
= √ .
Qn (Pn−1 − an−1 Q n−1 ) + m
CB609-11 CB609/Alaca & Williams August 7, 2003 16:58 Char Count= 0

290 Units in Real Quadratic Fields

Cross-multiplying and equating coefficients we see that

Pn (Pn−1 − an−1 Q n−1 ) + m = Q n Q n−1 ,
Pn + Pn−1 − an−1 Q n−1 = 0.
√
Thus αn = (Pn + m)/Q n (n ≥ 1) is determined recursively by
Pn = −Pn−1 + an−1 Q n−1 ,
m − Pn2
Qn = ,
Q
n−1 √
Pn + m
an = ,
Qn
√
with P0 = 0, Q 0 = 1, a0 = [ m].
The central result that allows us to determine the fundamental unit of a real
quadratic field is the following theorem.

Theorem 11.6.1 Let m be a positive squarefree integer. Let h n /kn (n = 0, 1, 2, . . .)

√
be the convergents of the infinite continued fraction expansion of m. Let l be the
period of the expansion.
If l is even then x 2 − my 2 = −1 has no solutions in integers x and y and the
solution of x 2 − my 2 = 1 in positive integers x and y with x least is (x, y) =
(h l−1 , kl−1 ).
If l is odd then x 2 − my 2 = −1 has solutions in integers x and y and the solution
of x 2 − my 2 = −1 in positive integers x and y with x least is given by (x, y) =
(h l−1 , kl−1 ).

If m ≡ 2, m ≡ 3 (mod 4), or m ≡ 1 (mod 8) all the units of OQ(√m) are of the

√
form x + y m with x and y integers, so that by Theorems 11.5.3 and 11.6.1 we
see that the fundamental unit η of OQ(√m) is given by
√
η = h l−1 + kl−1 m, N (η) = (−1)l .
If m ≡ 5 (mod 8) there may or may not be units of OQ(√m) of the form 12 (x +
√ √
y m) with x and y odd integers. If there are no such units then η ∈ Z + Z m and,
as in the previous case, we have
√
η = h l−1 + kl−1 m, N (η) = (−1)l .
√
If there are such units then η ∈ Z + Z m and it can be shown that η3 ∈ Z +
√ √
Z m. In this case η3 = x + y m, where x and y are positive integers satisfying
x 2 − my 2 = ±1 with x least so that by Theorems 11.5.3 and 11.6.1
√
η3 = h l−1 + kl−1 m, N (η) = (−1)l .
√
If η = (A + B m)/2, where A and B are odd positive integers, then
√ 3
A+B m √
= h l−1 + kl−1 m,
2
CB609-11 CB609/Alaca & Williams August 7, 2003 16:58 Char Count= 0

11.6 Calculating the Fundamental Unit 291

and so
A3 + 3AB 2 m = 8h l−1 ,
3A2 B + B 3 m = 8kl−1 .
Hence
1/3
A | h l−1 , 1 ≤ A < 2h l−1
and
1/3
kl−1
B | kl−1 , 1 ≤ B < 2 .
m
This gives the following algorithm for determining the fundamental unit η of
OQ(√m) for a positive squarefree integer m.

Step 1: h −1 = 1, k−1 = 0.
√ √
P0 = 0, Q 0 = 1, a0 = [ m], h 0 = [ m], k0 = 1.

Step 2: Determine Pn , Q n , an , h n , kn (n = 1, 2, . . .) recursively by means of

Pn = −Pn−1 + an−1 Q n−1 , n = 1, 2, . . . ,
m − Pn2
Qn = , n = 1, 2, . . . ,
Q
n−1 √
Pn + m
an = , n = 1, 2, . . . ,
Qn
h n = an h n−1 + h n−2 , n = 1, 2, . . . ,
kn = an kn−1 + kn−2 , n = 1, 2, . . . .
Stop at the first integer N > 1 such that
PN = P1 , Q N = Q 1 .

Step 3: l = N − 1.

Step 4: If m ≡ 2 (mod 4), m ≡ 3 (mod 4), or m ≡ 1 (mod 8) then

√
η = h l−1 + kl−1 m, N (η) = (−1)l .

Step 5: If m ≡ 5 (mod 8) determine all positive odd divisors A of h l−1 less than
1/3
2h l−1 and all positive odd divisors B of kl−1 less than 2 (kl−1 /m)1/3 . If for some
pair (A, B) we have
A3 + 3AB 2 m = 8h l−1 , 3A2 B + B 3 m = 8kl−1 ,
then
√
A+B m
η= , N (η) = (−1)l ;
2
CB609-11 CB609/Alaca & Williams August 7, 2003 16:58 Char Count= 0

292 Units in Real Quadratic Fields

otherwise
√
η = h l−1 + kl−1 m, N (η) = (−1)l .
We present several examples.

Example 11.6.1 (continued) m = 31 ≡ 3 (mod 4). Starting with

h −1 = 1, k−1 = 0, P0 = 0, Q 0 = 1, a0 = 5, h 0 = 5, k0 = 1,
we obtain successively the values of
Pn , Q n , an , h n , kn , n = 1, 2, . . . ,
as in Step 2.
n Pn Q n an hn kn
−1 1 0
0 0 1 5 5 1
1 5 6 1 6 1
2 1 5 1 11 2
3 4 3 3 39 7
4 5 2 5 206 37
5 5 3 3 657 118
6 4 5 1 863 155
7 1 6 1 1520 273
8 5 1 10 16063 2885
9 5 6 1 17583 3158
As
P9 = P1 = 5, Q 9 = Q 1 = 6,
we see that
N = 9, l = N − 1 = 8, h l−1 = h 7 = 1520, kl−1 = k7 = 273,
√ √
η = h l−1 + kl−1 m = 1520 + 273 31, N (η) = (−1)l = 1.
√
The fundamental unit of OQ(√31) is 1520 + 273 31 of norm 1.

Example 11.6.2 m = 41 ≡ 1 (mod 8). Here

n Pn Qn an hn kn
−1 1 0
0 0 1 6 6 1
1 6 5 2 13 2
2 4 5 2 32 5
3 6 1 12 397 62
4 6 5 2 826 129
CB609-11 CB609/Alaca & Williams August 7, 2003 16:58 Char Count= 0

11.6 Calculating the Fundamental Unit 293

As
P4 = P1 = 6, Q 4 = Q 1 = 5,
we have
N = 4, l = N − 1 = 3, h l−1 = h 2 = 32, kl−1 = k2 = 5,
√ √
η = h l−1 + kl−1 m = 32 + 5 41, N (η) = (−1)l = −1.
√
The fundamental unit of O Q(√41) is 32 + 5 41 of norm −1.

Example 11.6.3 m = 82 ≡ 2 (mod 4). Here

n Pn Qn an hn kn
−1 1 0
0 0 1 9 9 1
1 9 1 18 163 18
2 9 1 18 2943 325
As
P2 = P1 = 9, Q 2 = Q 1 = 1,
we have
N = 2, l = N − 1 = 1, h l−1 = h 0 = 9, kl−1 = k0 = 1,
√ √
η = h l−1 + kl−1 m = 9 + 82, N (η) = (−1)l = −1.
√
The fundamental unit of OQ(√82) is 9 + 82 of norm −1.

Example 11.6.4 m = 13 ≡ 5 (mod 8). Here

n Pn Qn an hn kn
−1 1 0
0 0 1 3 3 1
1 3 4 1 4 1
2 1 3 1 7 2
3 2 3 1 11 3
4 1 4 1 18 5
5 3 1 6 119 33
6 3 4 1 137 38
As
P6 = P1 = 3, Q 6 = Q 1 = 4,
we deduce that
N = 6, l = N − 1 = 5, h l−1 = h 4 = 18, kl−1 = k4 = 5.
CB609-11 CB609/Alaca & Williams August 7, 2003 16:58 Char Count= 0

294 Units in Real Quadratic Fields

Next,
1/3
A odd, A | h l−1 , 1 ≤ A < 2h l−1 =⇒ A | 9, 1 ≤ A < 5.3 =⇒ A = 1 or 3,
1/3
kl−1
B odd, B | kl−1 , 1 ≤ B < 2 =⇒ B | 5, 1 ≤ B < 1.5 =⇒ B = 1.
m
Of the pairs (A, B) = (1, 1), (3, 1) only the latter satisfies the pair of equations
A3 + 39AB 2 = 144, 3A2 B + 13B 3 = 40.
Hence the fundamental unit η(>1) of O Q(√13) is
√
3 + 13
η= , N (η) = −1.
2

Example 11.6.5 m = 37 ≡ 5 (mod 8). Here

n Pn Qn an hn kn
−1 1 0
0 0 1 6 6 1
1 6 1 12 73 12
2 6 1 12 882 145
As
P2 = P1 = 6, Q 2 = Q 1 = 1,
we have
N = 2, l = N − 1 = 1, h l−1 = h 0 = 6, kl−1 = k0 = 1.
Clearly the pair of equations
A3 + 111AB 2 = 48, 3A2 B + 37B 3 = 8
has no solutions in positive integers, and so the fundamental unit η (>1) of
OQ(√37) is
√
η = 6 + 37, N (η) = −1.

11.7 The Equation x 2 − my 2 = N

Let m be a positive squarefree integer. The following theorem from the theory
of continued fractions assists us in finding the solutions (if any) of the equation
√
x 2 − my 2 = N when N is a nonzero integer satisfying |N | < m.

Theorem 11.7.1 Let m be a positive nonsquare integer. Let {h n }n≥−1 and {kn }n≥−1
be defined as in (11.6.5) and (11.6.6). Let gn = h 2n − mkn2 , n = −1, 0, 1, . . . . Let
CB609-11 CB609/Alaca & Williams August 7, 2003 16:58 Char Count= 0

11.7 The Equation x 2 − my 2 = N 295

√
l be the period of the continued fraction expansion of m. Let N be an integer
√
satisfying 0 < |N | < m. Then the equation x 2 − my 2 = N is solvable in coprime
integers x and y if and only if N = gr for some r ∈ {0, 1, 2, . . . , sl − 1}, where
s = 1 if l is even and s = 2 if l is odd in which case a solution is (x, y) = (h r , kr ).
√
Example 11.7.1 We choose m = 31. Here 31 = 5.567 . . . . Thus the integers N
√
satisfying 0 < |N | < m are

N = −5, −4, −3, −2, −1, 1, 2, 3, 4, 5.

From Example 11.6.1 we have l = 8, so that s = 1 and sl − 1 = 7, and the values

of gn (n = 0, 1, . . . , 7) are given in the following table:

n hn kn gn = h 2n − 31kn2
0 5 1 −6
1 6 1 5
2 11 2 −3
3 39 7 2
4 206 37 −3
5 657 118 5
6 863 155 −6
7 1520 273 1
√
By Theorem 11.7.1 the equation x 2 − 31y 2 = N (0 < |N | < 31) is solvable in
coprime integers x and y for

N = −3, 1, 2, 5

and is not solvable in coprime integers for

N = −5, −4, −2, −1, 3, 4.

Example 11.7.2 We saw in Example 11.7.1 that the equation x 2 − 31y 2 = −3

has the solutions (x, y) = (11, 2) and (206, 37). We now determine all solutions of
x 2 − 31y 2 = −3 in integers x and y. (Notice that if x and y are integers satisfying
x 2 − 31y 2 = −3 then x and y are necessarily coprime as −3 is squarefree.) Let
(x, y) ∈ Z2 be a solution of x 2 − 31y 2 = −3. Thus, as (x, y) = (11, 2) is a solution
of this equation, we have in OQ(√31)
√ √ √ √
x + y 31x − y 31 = 11 + 2 3111 − 2 31.
√ √
As 11 + 2 31 and 11 − 2 31 are prime ideals of OQ(√31) by Theorem 10.1.6(a)
since
√ √
N (11 + 2 31) = |N (11 + 2 31)| = | − 3| = 3 (a prime),
CB609-11 CB609/Alaca & Williams August 7, 2003 16:58 Char Count= 0

296 Units in Real Quadratic Fields

appealing to Theorem 8.3.2 we see that

√ √
x + y 31 = 11 ± 2 31.

Hence by Theorem 1.3.1 we obtain

√ √
x + y 31 = u(11 ± 2 31),
√
where u ∈ U (OQ(√31) ). The fundamental unit of OQ(√31) is 1520 + 273 31 (Ex-
ample 11.6.1) so that by Theorem 11.5.1
√
u = ±(1520 + 273 31)n

for some n ∈ Z. Hence all the solutions of x 2 − 31y 2 = −3 are given by

√ √ √
x + y 31 = ±(1520 + 273 31)n (11 ± 2 31), n = 0, ±1, ±2, . . . .

It is easily checked that these are solutions of x 2 − 31y 2 = −3. In particular the
solution (x, y) = (206, 37) is given by
√ √
−(1520 + 273 31)(11 − 2 31).

Example 11.7.3 We answer the question “Is the equation x 2 − 41y 2 = 2 solvable
in integers√x and y?”
As 2 < 41 we can apply Theorem 11.7.1. From Example 11.6.2 we have l = 3,
so that s = 2 and sl − 1 = 5, and the values of gn (n = 0, 1, 2, 3, 4, 5) are as
follows:

n hn kn gn
0 6 1 −5
1 13 2 5
2 32 5 −1
3 397 62 5
4 826 129 −5
5 2049 320 1

As 2 = gn (n = 0, 1, 2, 3, 4, 5) the equation x 2 − 41y 2 = 2 has no solution in co-

prime integers x and y and thus (as 2 is squarefree) no solution in integers x
and y.

Example 11.7.4 We determine all the solutions in integers x and y of the equation
x 2 − 10y 2 = 10. In this case we cannot apply Theorem 11.7.1 directly as m = N =
√
10 and |N | < m. Thus we proceed differently.
Let (x, y) ∈ Z2 be a solution of x 2 − 10y 2 = 10. Then 10 | x 2 and as 10 is
squarefree we deduce that 10 | x. Setting x = 10z in the equation we obtain
CB609-11 CB609/Alaca & Williams August 7, 2003 16:58 Char Count= 0

Exercises 297
√
y 2 − 10z 2 = −1. Thus y + z 10 is a unit of OQ(√10) of norm −1. As the fun-
√
damental unit of OQ(√10) is 3 + 10 (of norm −1), we have by Theorem 11.4.1
√ √
y + z 10 = ±(3 + 10)2n+1

for some n ∈ Z. Hence all the solutions of x 2 − 10y 2 = 10 are given by

√ √ √
x + y 10 = ±(3 + 10)2n+1 10, n ∈ Z.

It is easily verified that these are solutions of x 2 − 10y 2 = 10.

Exercises
1. Prove Theorem 11.5.6.
2. Prove Theorem 11.5.7.
3. Determine the fundamental unit of OQ(√11) .
4. Determine the fundamental unit of OQ(√17) .
√
5. Prove that the fundamental unit of OQ(√94) is η = 2143295 + 221064 94.
√
6. Prove that the fundamental unit of OQ(√163) is η = 64080026 + 5019135 163.
√
7. Prove that the fundamental unit of OQ(√165) is η = (13 + 165)/2.
√
8. Show that √1790 = [42, 3, 4, 8, 4, 3, 84].
9. Show that 925 = [30, 2, 2, 2, 2, 60]. √
10. Determine the length of the continued fraction expansion of 850.
11. Determine the norm of the fundamental unit of OQ(√1378) .
12. Let m be a positive squarefree integer such that OQ(√m) contains units of norm −1. Let
σ be the fundamental unit of OQ(√m) of norm −1. Prove that σ is the smallest unit > 1
of norm −1 in OQ(√m) .
13. Let η be the fundamental unit of OQ(√134) . Determine α ∈ OQ(√134) such that 2η = α 2 .
√
14. Let p be a prime ≡ 3 (mod 8). Let t + u p be the fundamental unit of OQ(√ p) , which
necessarily is of norm 1. Starting from t 2 − pu 2 = 1, and using Dirichlet’s method of
proving Theorem 11.5.4, prove that the equation x 2 − py 2 = −2 is solvable in integers
x and y.
15. Let p be a prime ≡ 7 (mod 8). Prove that the equation x 2 − py 2 = 2 is solvable in
integers x and y.
16. Let p be a prime ≡ 9 (mod 16) for which the congruence x 4 ≡ 2 (mod p) is insolvable.
Prove that the norm of the fundamental unit of OQ(√2 p) is −1.

17. Let p and q be distinct primes with p ≡ q ≡ 1 (mod 4) and qp = 1 (so that qp = 1
by the law of quadratic reciprocity). Suppose that the congruences x 4 ≡ p (mod q) and
y 4 ≡ q (mod p) are insolvable. Prove that the norm of the fundamental unit of OQ(√ pq)
is −1.
18. Is the equation x 2 − 82y 2 = 2 solvable in integers x and y?
19. Determine all solutions of x 2 − 96y 2 = 161.
CB609-11 CB609/Alaca & Williams August 7, 2003 16:58 Char Count= 0

298 Units in Real Quadratic Fields

20. Prove that all solutions of x 2 − 10y 2 = 10 (see Example 11.7.1) are given recursively
by ±(xk , yk ), where
xk+1 = 19xk + 60yk , yk+1 = 6xk + 19yk , k = 0, ±1, ±2, . . . ,
and x0 = 10, y0 = 3.
21. Let m be a positive integer such that m − 1 and m are not perfect squares but 4m + 1
is a perfect square. Prove that the equation x 2 − my 2 = −1 is insolvable in integers x
and y.

Suggested Reading
1. H. W. Lenstra Jr., Solving the Pell equation, Notices of the American Mathematical
Society 49 (2002), 182–192.
This up-to-date article describes the use of smooth numbers to solve Pell’s equation x 2 − my 2 = 1.
2. I. Niven, H. S. Zuckerman, and H. L. Montgomery, An Introduction to the Theory of
Numbers, fifth edition, Wiley, New York, 1991.
Chapter 7 contains a comprehensive treatment of continued fractions.
3. W. Patz, Tafel der Regelmässigen Kettenbrüche und ihrer Vollständigen Quotienten für
die Quadratwurzeln aus der Natürlichen Zahlen von 1–10000, Akademie-Verlag, Berlin,
1955.
√
This book gives the continued fraction expansions of m for all nonsquares m up to 10,000.

Biographies
1. E. T. Bell, Men of Mathematics, Simon and Schuster, New York, 1937.
Chapters 9 and 10 are devoted to Leonhard Euler (1707–1783) and Joseph-Louis Lagrange (1736–
1813) respectively.
2. J. J. Burckhardt, Leonhard Euler, 1707–1783, Mathematics Magazine 56 (1983), 262–
273.
A brief overview of the work of Euler is given.
3. H. Davenport, Dirichlet, Mathematical Gazette 43 (1959), 268–269.
A short biography of Dirichlet is given.
4. H. Koch, Gustav Peter Lejeune Dirichlet, in H. G. W. Begehr, H. Koch, J. Kramer,
N. Schappacher, and E. J. Thiele (Eds.), Mathematics in Berlin, 33–40, Birkhäuser
Verlag, Berlin, 1998.
The book in which this biography of Dirichlet is included describes the many facets of Berlin’s role
in mathematics. It includes biographies of many famous mathematicians connected with Berlin.
5. The website

http://www-groups.dcs.st-and.ac.uk/˜history/

has biographies of William Brouncker, Lejeune Dirichlet, Leonhard Euler, Joseph-Louis

Lagrange, and John Pell.
CB609-12 CB609/Alaca & Williams August 7, 2003 17:0 Char Count= 0

12
The Ideal Class Group

12.1 Ideal Class Group

We have already seen that the nonzero integral and fractional ideals of the ring O K
of integers of an algebraic number field K form a group I (K ) under multiplication
(Theorem 8.3.4). The principal ideals in I (K ) are of the form α = {r α | r ∈ O K }
for some α ∈ K ∗ and they form a subgroup P(K ) of I (K ) as
αβ−1 = αβ −1 ∈ P(K ).

The group I (K ) is an Abelian group so P(K ) is a normal subgroup of I (K ) and

the factor group I (K )/P(K ) is well defined and Abelian.

Definition 12.1.1 (Ideal class group) Let K be an algebraic number field. Let
I (K ) be the group of nonzero fractional and integral ideals of O K . Let P(K ) be the
subgroup of principal ideals of I (K ). Then the factor group I (K )/P(K ) is called
the ideal class group of K and is denoted by H (K ).

It is an important result that H (K ) is always a finite group. This is proved in

Section 12.5 as a consequence of some theorems of Hermann Minkowski (1864–
1909) in the geometry of numbers.

Definition 12.1.2 (Class number) Let K be an algebraic number field. The order
of the ideal class group H (K ) is called the class number of K and is denoted by
h(K ).

If two nonzero ideals A and B of O K are in the same class of H (K ) =

I (K )/P(K ), we say that they are equivalent and write A ∼ B. Clearly
A ∼ B ⇐⇒ A P(K ) = B P(K )
⇐⇒ A−1 B ∈ P(K )
⇐⇒ A−1 B = α for some α ∈ K ∗
⇐⇒ B = Aα for some α ∈ K ∗
⇐⇒ aA = bB for some a, b ∈ O K \ {0}.

299
CB609-12 CB609/Alaca & Williams August 7, 2003 17:0 Char Count= 0

300 The Ideal Class Group

√
Example 12.1.1 In Q( −5) we have
√ √
2, 1 − −5 ∼ 3, 1 + −5

as
√ √
32, 1 − −5 = 6, 3(1 − −5)
√ √ √
= (1 + −5)(1 − −5), 3(1 − −5)
√ √
= 1 − −51 + −5, 3.

Theorem 12.1.1 Let K be an algebraic number field. Then

h(K ) = 1 ⇐⇒ O K is a principal ideal domain

⇐⇒ O K is a unique factorization domain.

Proof: If h(K ) = 1 then [I (K ) : P(K )] = card(I (K )/P(K )) = card H (K ) =

h(K ) = 1 so that P(K ) = I (K ). Hence every ideal of O K is principal and so O K
is a principal ideal domain, and thus a unique factorization domain, by Theorem
3.3.1.
Conversely, if O K is a unique factorization domain, since it is a Dedekind domain,
it is a principal ideal domain (Exercise 13 of Chapter 8). Hence every ideal of O K
is principal and so I (K ) = P(K ) and thus

h(K ) = card H (K ) = card(I (K )/P(K )) = [I (K ) : P(K )] = 1.

Leonard Carlitz (1907–1999) has shown that h(K ) = 1 or 2 if and only if when-
ever a nonzero nonunit α ∈ O K can be written α = uπ1 · · · πs = u π1 · · · πt with
u, u units and π1 , . . . , πs , π1 , . . . , πt prime elements of O K then s = t [2].
In the next three sections we prove three theorems of Minkowski from which we
can deduce that the class number is always finite.

12.2 Minkowski’s Translate Theorem

Let R denote the vector space of all n-tuples (x1 , x2 , . . . , xn ) with x1 , x2 , . . . , xn ∈
n

R. We let Zn be the subset of Rn given by

Zn = {(x1 , . . . , xn ) ∈ Rn | x1 , . . . , xn ∈ Z}.

The elements of Zn are called lattice points and Zn is called a lattice. Clearly Zn is
a group under addition. For α = (a1 , . . . , an ) ∈ Rn we set

||α|| = max |ai | (∈ R).

1≤i≤n
CB609-12 CB609/Alaca & Williams August 7, 2003 17:0 Char Count= 0

12.2 Minkowski’s Translate Theorem 301

Definition 12.2.1 (Translate) If S is a subset of Rn and α ∈ Rn we let

Sα = {α + β | β ∈ S}.

The set Sα (⊆ Rn ) is called a translate of S in Rn .

Clearly S0 = S, where 0 = (0, . . . , 0).

Definition 12.2.2 (Magnification) If S is a subset of Rn and a ∈ R+ we let

aS = {aβ | β ∈ S}.

The set aS (⊆ Rn ) is called a magnification of S in Rn .

Definition 12.2.3 (Bounded set) A subset S of Rn is said to be bounded if there

exists B ∈ R+ such that

||α|| ≤ B for all α ∈ S.

Definition 12.2.4 (Closed set) Let α ∈ Rn . Let r ∈ R+ . The set

{β ∈ Rn | ||α − β|| < r }

is called a neighborhood of α. The point α is called a limit point of the subset S of

Rn if every neighborhood of α contains a point β = α such that β ∈ S. The set S
is said to be closed if every limit point of S is a point of S.

Definition 12.2.5 (Convex set) A subset S of Rn is said to be convex if

tβ + (1 − t)γ ∈ S for all β, γ ∈ S and all t ∈ R with 0 ≤ t ≤ 1.

Clearly S is convex if it contains the line segment joining β and γ for all points
β and γ in S.

Definition 12.2.6 (Convex body) A closed, bounded, convex subset of Rn is called

a convex body.

If S is a convex body so are Sα (α ∈ Rn ) and aS (a ∈ R+ ). Moreover, aS ⊆ bS

if 0 < a < b.
A theorem of Minkowski asserts that every convex body S has a volume, which
we denote by V (S) and which has the following properties:

(i) 0 ≤ V (S) < ∞,

(ii) if Si (i = 1, 2, . . . , k) are disjoint convex bodies with Si ⊆ S (i = 1, 2, . . . , k) then
k
i=1 V (Si ) ≤ V (S),
CB609-12 CB609/Alaca & Williams August 7, 2003 17:0 Char Count= 0

302 The Ideal Class Group

(iii) V (S) = V (Sα ) for all α ∈ Rn ,

(iv) V (aS) = a n V (S) for all a ∈ R+ .
If S1 , . . . , Sk are disjoint convex bodies we define

k
V (S1 ∪ S2 ∪ · · · ∪ Sk ) = V (Si ).
i=1

The volume V (S) of a convex body S is defined by means of the multiple integral

V (S) = ··· d x1 · · · d xn .
S

Definition 12.2.7 (Hypercube Ht ) The hypercube Ht (t ∈ R+ ) in Rn is defined by

Ht = {β ∈ Rn | ||β|| ≤ t}.

It is easy to check that the hypercube Ht is a convex body. Its volume is given by
t t t n
V (Ht ) = ··· dβ1 · · · dβn = dβ = (2t)n .
β1 =−t βn =−t β=−t

Theorem 12.2.1 (Minkowski’s translate theorem) Let S be a convex body in

Rn that contains the origin 0 = (0, 0, . . . , 0). If V (S) ≥ 1 then for at least one
α ∈ Zn \ {0}
S ∩ Sα = φ.

Proof: We first treat the case V (S) > 1. Suppose on the contrary that
S ∩ Sα = φ for all α ∈ Zn \ {0}.
First we prove that
Sβ ∩ Sγ = φ for all β, γ ∈ Zn , β = γ . (12.2.1)
Let x ∈ Sβ ∩ Sγ . Then x ∈ Sβ and x ∈ Sγ . Hence x − γ ∈ Sβ−γ and x − γ ∈ S.
Thus x − γ ∈ Sβ−γ ∩ S, contradicting our assumption. This proves (12.2.1).
Now let N be an arbitrary positive integer, and let
T = {α ∈ Zn | ||α|| ≤ N }.
Clearly
card T = (2N + 1)n .
Our second step is to show that

V Sα = (2N + 1)n V (S). (12.2.2)
α∈T
CB609-12 CB609/Alaca & Williams August 7, 2003 17:0 Char Count= 0

12.2 Minkowski’s Translate Theorem 303

We have

V Sα = V (Sα ) (by (12.2.1))
α∈T α∈T

= V (S)
α∈T
= V (S)card T
= (2N + 1)n V (S),
as asserted.
As S is a bounded set, we can define the diameter d ∈ R+ of S by
d = max ||s1 − s2 ||.
s1 ,s2 ∈S

We let H be the hypercube HN +d , that is,

H = HN +d = {β ∈ Rn | ||β|| ≤ N + d}.
Clearly
V (H ) = (2N + 2d)n .
Our third step is to show that
Sα ⊆ H for all α ∈ T. (12.2.3)
Let α ∈ T and β ∈ Sα . Then β = α + s, where s ∈ S. As α ∈ T we have ||α|| ≤ N .
As 0 ∈ S we have α = α + 0 ∈ Sα so that s = β − α, where α, β ∈ Sα . Thus
||s|| = ||β − α|| ≤ max ||t1 − t2 || = max ||(α + s1 ) − (α + s2 )||
t1 ,t2 ∈Sα s1 ,s2 ∈S
= max ||s1 − s2 || = d.
s1 ,s2 ∈S

Hence
||β|| = ||α + s|| ≤ ||α|| + ||s|| ≤ N + d
so that β ∈ H . Hence Sα ⊆ H for α ∈ T .
We are now in a position to complete the proof. From (12.2.3) we have

Sα ⊆ H
α∈T

so that

V Sα ≤ V (H )
α∈T

and thus
(2N + 1)n V (S) ≤ (2N + 2d)n .
CB609-12 CB609/Alaca & Williams August 7, 2003 17:0 Char Count= 0

304 The Ideal Class Group

Therefore
 d n
n 1+
2N + 2d  N 
V (S) ≤ = .
2N + 1 1 
1+
2N
As d and n are fixed, letting N → ∞ we obtain V (S) ≤ 1, contradicting V (S) > 1.
Hence there is at least one α ∈ Zn \ {0} with S ∩ Sα = φ.
We now turn to the case V (S) = 1. Let k be a positive integer. We consider the
convex body

1
S(k) = 1 + S
k
obtained by magnifying the convex body S by a factor 1 + 1/k. As S contains 0 so
does S(k). The volume of S(k) satisfies

1 1 n 1 n
V (S(k)) = V 1+ S = 1+ V (S) = 1 + > 1.
k k k
Thus the first part of the theorem applies to the convex body S(k). Hence there
exists a translate (S(k))αk (αk ∈ Zn \ {0}) of S(k) such that

S(k) ∩ (S(k))αk = φ.

Let xk ∈ S(k) ∩ (S(k))αk .

Next we consider the set

Ak = {βk ∈ Zn \ {0} | xk ∈ S(k) ∩ (S(k))βk }.

Clearly Ak = φ as αk ∈ Ak . For βk ∈ Ak we have

xk ∈ S(k) ⊆ S(1)

and

xk ∈ (S(k))βk so xk − βk ∈ S(k) ⊆ S(1).

Let d1 denote the diameter of S(1). Then

||βk || = ||xk − (xk − βk )|| ≤ ||xk || + ||xk − βk ||

= ||xk − 0|| + ||(xk − βk ) − 0||
≤ d1 + d1
= 2d1 ,

as 0 ∈ S(1). Thus βk lies in a bounded set, namely,

βk ∈ H2d1 ,
CB609-12 CB609/Alaca & Williams August 7, 2003 17:0 Char Count= 0

12.3 Minkowski’s Convex Body Theorem 305

so that

Ak ⊆ H2d1 .

But

Ak ⊆ Zn

Ak ⊆ H2d1 ∩ Zn .

Thus every Ak is contained in the finite set H2d1 ∩ Zn . Hence we can find a subse-
quence of the sequence {xk } for which the corresponding βk ∈ Ak is constant, say,
equal to α. After relabeling we may therefore assume that

xk ∈ S(k) ∩ (S(k))α .

As each xk ∈ S(1), the infinite sequence {xk } is bounded. Hence, by the Bolzano–
Weierstrass theorem, the sequence {xk } has at least one limit point, say x. Let l be
an arbitrary positive integer. We have

xl ∈ S(l),
xl+1 ∈ S(l + 1) ⊂ S(l)
xl+2 ∈ S(l + 2) ⊂ S(l + 1) ⊂ S(l),
···

so that the infinite sequence {xk }k≥l lies in S(l). As x is a limit point of {xk }k≥l and
S(l) is closed, we deduce that x ∈ S(l) for every positive integer l. Thus
∞ ∞
1
x∈ S(l) = 1+ S = S.
l=1 l=1
l

Similarly,
∞

x∈ (S(l))α = Sα .
l=1

Hence

x ∈ S ∩ Sα

so that S ∩ Sα = φ for α ∈ Zn \ {0}.

12.3 Minkowski’s Convex Body Theorem

Minkowski’s famous convex body theorem asserts that if the volume of a convex
body, which is symmetrical about the origin, is large enough then the convex body
CB609-12 CB609/Alaca & Williams August 7, 2003 17:0 Char Count= 0

306 The Ideal Class Group

must contain at least one lattice point different from the origin. We begin by making
the notion “symmetrical about the origin” precise.

Definition 12.3.1 (Centrally symmetric set) A subset S of Rn is said to be cen-

trally symmetric if

−α ∈ S for all α ∈ S.

We note that a subset S of Rn , which is both centrally symmetric and convex, must
contain the origin 0 = (0, 0, . . . , 0). To see this take any α ∈ S. As S is centrally
symmetric, we have −α ∈ S. Then, as S is convex, we have 12 (α) + 12 (−α) ∈ S,
that is, 0 ∈ S.
We now use Minkowski’s translate theorem to prove his convex body theorem.

Theorem 12.3.1 (Minkowski’s convex body theorem) Let S (⊆ Rn ) be a centrally

symmetric convex body of volume V (S) ≥ 2n . Then S contains a lattice point = 0.

Proof: Let T be the magnification of S given by

1 1
T =
S = { α | α ∈ S}.
2 2
As S is a centrally symmetric convex body so is T . As T is both centrally symmetric
and convex it contains 0. Moreover,

V (T ) = V ( 12 S) = 1
2n
V (S) ≥ 1,

so that by Minkowski’s translate theorem (Theorem 12.2.1) there exists α ∈ Zn \

{0} such that

T ∩ Tα = φ.

Let x ∈ T ∩ Tα . Then x ∈ Tα and thus x − α ∈ T . Since T is centrally symmetric,

we have α − x = −(x − α) ∈ T . Hence, as x ∈ T and T is convex, we have
1 1
(α − x) + x ∈ T,
2 2
so that α/2 ∈ T and thus α ∈ S. This proves that S contains the lattice point
α = 0.

12.4 Minkowski’s Linear Forms Theorem

Let n be a positive integer and let r and s be nonnegative integers such that r + 2s =
n. Let a jk ( j, k = 1, 2, . . . , n) be n 2 complex numbers with det a jk = 0 and

a jk ∈ R for j = 1, 2, . . . , r ; k = 1, 2, . . . , n, (12.4.1)
CB609-12 CB609/Alaca & Williams August 7, 2003 17:0 Char Count= 0

12.4 Minkowski’s Linear Forms Theorem 307

for each j = r + 1, . . . , n there exists k j ∈ {1, 2, . . . , n}

such that a jk j ∈ C \ R, (12.4.2)

and

a j+s k = a jk for j = r + 1, . . . , r + s; k = 1, 2, . . . , n. (12.4.3)

Here z̄ denotes the complex conjugate of the complex number z. Set

n
L j = L j (x) = a jk xk , j = 1, 2, . . . , n, (12.4.4)
k=1

so that L 1 , . . . , L n are linear forms such that L 1 , . . . , L r are real,

L r +1 , . . . , L n=r +2s are nonreal, and L r +s+1 = L r +1 , . . . , L r +2s = L r +s .

Theorem 12.4.1 (Minkowski’s linear forms theorem) Let δ1 , . . . , δn be n positive

real numbers such that
s
2
δ1 · · · δn ≥ |det(a jk )|, (12.4.5)
π
where the a jk are defined in (12.4.1)–(12.4.3), and

δ j = δ j+s , j = r + 1, . . . , r + s. (12.4.6)

Then there exist integers y1 , . . . , yn , not all zero, such that

n

a jk yk ≤ δ j , j = 1, 2, . . . , n.

k=1

Proof: We define n real linear forms M j = M j (x) ( j = 1, 2, . . . , n) in terms of the

L j (x) ( j = 1, 2, . . . , n) by


 L , j = 1, 2, . . . , r,
 j
M j = 12 (L j + L j+s ), j = r + 1, . . . , r + s, (12.4.7)


1
2i
(L j−s − L j ), j = r + s + 1, . . . , n.

∂M
From (12.4.7) we see that the n × n matrix ∂ L kj is given by
 
Ir Or,s Or,s
∂Mj  
=
 Os,r
1
I
2 s
1
2 s
I ,

∂ Lk
1 −1
Os,r I
2i s 2i s
I
where Il denotes the l × l identity matrix and Ol,m denotes the l × m zero
matrix.

∂M
Adding the (r + s + k)th column to the (r + k)th column in the matrix ∂ L kj for
CB609-12 CB609/Alaca & Williams August 7, 2003 17:0 Char Count= 0

308 The Ideal Class Group

k = 1, 2, . . . , s, we obtain the upper triangular matrix

 
Ir Or,s Or,s
 
A=  Os,r Is 1
I ,
2 s 

Os,r Os −1 I
2i s

where Os = Os,s . The determinant of a triangular matrix is the product of its diag-
onal entries so that
s
−1 s i
detA = = .
2i 2

As the elementary column operations used to obtain A do not change the value of
the determinant of the matrix, we have
s
∂Mj i
det = detA = .
∂ Lk 2

∂M
The quantity det ∂ L kj is the Jacobian of the M j with respect to the L k so that
s
∂(M1 , . . . , Mn ) i
= .
∂(L 1 , . . . , L n ) 2

From (12.4.4) we deduce that

∂L j
= a jk , j, k = 1, 2, . . . , n,
∂ xk
so that

∂(L 1 , . . . , L n ) ∂L j
= det = det(a jk ).
∂(x1 , . . . , xn ) ∂ xk

Hence
s
∂(M1 , . . . , Mn ) ∂(M1 , . . . , Mn ) ∂(L 1 , . . . , L n ) i
= = det(a jk ),
∂(x1 , . . . , xn ) ∂(L 1 , . . . , L n ) ∂(x1 , . . . , xn ) 2

so that

∂(x1 , . . . , xn ) 2s
=
∂(M , . . . , M ) |det(a )| . (12.4.8)
1 n jk

Now let S be the subset of Rn given by

S = {x ∈ Rn | |L j (x)| ≤ δ j , j = 1, 2, . . . , r + s}.
CB609-12 CB609/Alaca & Williams August 7, 2003 17:0 Char Count= 0

12.4 Minkowski’s Linear Forms Theorem 309

It is easily checked that S is a centrally symmetric convex body. The volume of S

is given by

V (S) = ··· d x1 · · · d xr +s .
|L 1 (x)| ≤ δ1
.
.
.
|L r +s (x)| ≤ δr +s .

Now, as
|L r + j (x)| = |L r + j (x)| = |L r +s+ j (x)|, j = 1, 2, . . . , s,
and
δr + j = δr +s+ j , j = 1, 2, . . . , s,
we deduce that
|L r + j (x)| ≤ δr + j ⇐⇒ |L r + j (x)|2 ≤ δr2+ j
⇐⇒ |L r + j (x)L r +s+ j (x)|2 ≤ δr2+ j
⇐⇒ |(Mr + j (x) + i Mr +s+ j (x))(Mr + j (x) − i Mr +s+ j (x))|2 ≤ δr2+ j
⇐⇒ Mr + j (x)2 + Mr +s+ j (x)2 ≤ δr2+ j
for j = 1, 2, . . . , s. Hence

V (S) = ··· d x1 · · · d xn .
|M1 (x)| ≤ δ1
.
.
.
|Mr (x)| ≤ δr
Mr +1 (x)2 + Mr +s+1 (x)2 ≤ δr2+1
.
.
.
Mr +s (x)2 + Mn (x)2 ≤ δr2+s

Making the change of variable

M j = M j (x), j = 1, 2, . . . , n,
in the integral, we obtain

∂(x1 , . . . , xn )
V (S) = ···
∂(M , . . . , M ) d M1 · · · d Mn .
1 n
|M1 | ≤ δ1
.
.
.
|Mr | ≤ δr
Mr +1 + Mr2+s+1 ≤ δr2+1
2

.
.
.
Mr2+s + Mn2 ≤ δr2+s
CB609-12 CB609/Alaca & Williams August 7, 2003 17:0 Char Count= 0

310 The Ideal Class Group

Appealing to (12.4.8), we have

2s
V (S) = ··· d M1 · · · d Mn .
|det(a jk )|
|M1 | ≤ δ1
.
.
.
|Mr | ≤ δr
Mr +1 + Mr2+s+1 ≤ δr2+1
2

.
.
.
Mr2+s + Mn2 ≤ δr2+s

Expressing the integral in terms of repeated integrals, we deduce that

 
s
2 s r δ j  
V (S) = dMj  ··· d Mr +k d Mr +s+k  .
|det(a jk )| j=1 −δ j k=1
Mr2+k + Mr2+s+k ≤ δr2+k

Now
δ
d M = 2δ
−δ

and

··· d M d N = area of a circle of radius δ = πδ 2 ,
M 2 + N 2 ≤ δ2

so that
2s r s
V (S) = 2δ j πδr2+k
|det(a jk )| j=1 k=1

Hence, by Minkowski’s convex body theorem (Theorem 12.3.1), S contains a lattice

point y = (y1 , . . . , yn ) = (0, . . . , 0). Thus
|L j (y)| ≤ δ j , j = 1, 2, . . . , r + s,

from which the asserted result follows by (12.4.3), (12.4.4), and (12.4.6).
CB609-12 CB609/Alaca & Williams August 7, 2003 17:0 Char Count= 0

12.5 Finiteness of the Ideal Class Group 311

12.5 Finiteness of the Ideal Class Group

In this section we use Minkowski’s linear forms theorem to show that every class
in the ideal class group H (K ) of an algebraic number field K contains an integral
ideal of O K with norm less than a certain bound, called the Minkowski bound,
that depends only on the degree of the field K and the discriminant of K . For a
particular algebraic number field K these ideas give a method of determining the
ideal class group H (K ).

Theorem 12.5.1 Let K = Q(θ) be an algebraic number field of degree n = r + 2s,

where θ has r real conjugates and s pairs of nonreal complex conjugates. Let A
be an integral or fractional ideal of O K . Then there exists an element α (= 0) ∈ A
such that
s
2
|N (α)| ≤ N (A) |d(K )|.
π

Proof: Let θ1 , θ2 , . . . , θn be the conjugates of θ . We reorder θ1 , θ2 , . . . , θn in such

a way that θ1 , θ2 , . . . , θr ∈ R and θr +1 , θr +2 , . . . , θn ∈ C \ R. As the complex con-
jugate of any conjugate of θ is also a conjugate of θ (Exercise 20 of Chapter 5), we
can further order θr +1 , θr +2 , . . . , θn so that θr +s+1 = θr +1 , . . . , θn = θr +2s = θr +s ,
where r + 2s = n. Let σ1 , . . . , σn be the n monomorphisms : K → C chosen so
that σi (θ ) = θi . Hence σr +s+t = σr +t (t = 1, . . . , s).
Let {α1 , . . . , αn } be a basis for A. We define n linear forms L j (x) ( j =
1, 2, . . . , n) by

n
L j (x) = σ j (αk )xk .
k=1

These forms satisfy (12.4.1)–(12.4.4) with a jk = σ j (αk ) ( j, k = 1, 2, . . . , n). More-

over,

|det(a jk )| = |det(σ j (αk ))| = |D(A)| = N (A) |d(K )| = 0.

Let
s/n
2
δj = N (A)1/n |d(K )|1/2n , j = 1, 2, . . . , n.
π

Then
s s
2 2
δ 1 · · · δn = N (A)|d(K )| =
1/2
|det(a jk )|
π π
CB609-12 CB609/Alaca & Williams August 7, 2003 17:0 Char Count= 0

312 The Ideal Class Group

so, by Minkowski’s linear forms theorem (Theorem 12.4.1), there exist integers
y1 , . . . , yn , not all zero, such that
s/n
2
|L j (y)| ≤ N (A)1/n |d(K )|1/2n , j = 1, 2, . . . , n.
π
Choose m ∈ {1, 2, . . . , n} such that σm = 1, where 1 denotes the identity monomor-
phism from K to K . Set

n
n
α = L m (y) = σm (αk )yk = αk yk ,
k=1 k=1

so that α ∈ A and α = 0. The conjugates of α are

n
σ j (α) = σ j (αk )yk = L j (y), j = 1, 2, . . . , n.
k=1

Hence
s/n
2
|σ j (α)| ≤ N (A)1/n |d(K )|1/2n , j = 1, 2, . . . , n,
π
and so
s
2
|N (α)| = |σ1 (α) · · · σn (α)| ≤ N (A)|d(K )|1/2
π
as asserted.

Theorem 12.5.2 Let K = Q(θ) be an algebraic number field of degree n = r + 2s,

where θ has r real conjugates and s pairs of nonreal complex conjugates. Let
C ∈ H (K ). Then C contains an integral ideal B = 0 with
s
2
N (B) ≤ |d(K )|.
π

Proof: Let A be an ideal in the class C −1 of H (K ). Then A−1 ∈ C. By Theorem

12.5.1 there exists α (= 0) ∈ A such that
s
2
|N (α)| ≤ N (A) |d(K )|.
π
As α ∈ A we have α ⊆ A so that, by Theorem 8.4.1, A | α. Hence B = αA−1
is an integral ideal of O K . As α = 0 we have B = 0. Also, B ∈ C as A−1 ∈ C.
Finally,
N (B) = N (αA−1 ) = N (α)N (A−1 )
s
−1 2
= |N (α)|N (A) ≤ |d(K )|.
π
CB609-12 CB609/Alaca & Williams August 7, 2003 17:0 Char Count= 0

12.5 Finiteness of the Ideal Class Group 313

We next establish that there are only finitely many integral ideals in the ring of
integers of an algebraic number field having a given norm.

Theorem 12.5.3 Let K be an algebraic number field. Let k be a positive integer.

There are only finitely many integral ideals A of O K with N (A) = k.

Proof: Let A be an integral ideal of O K with N (A) = k. By Theorem 9.1.3

card (O K /A) = N (A) = k.

Hence

k + A = k(1 + A) = 0 + A

so that k ∈ A. Thus k ⊆ A and so A | k. By Theorem 8.3.1 there exist distinct
prime ideals P1 , . . . , Pr and nonnegative integers a1 , . . . , ar such that

k = P1a1 · · · Prar .

Hence, as A | k, we have

A = P1c1 · · · Prcr ,

where

ci ∈ {0, 1, . . . , ai } for i = 1, 2, . . . , r.

Thus there are at most

(a1 + 1)(a2 + 1) · · · (ar + 1)

possibilities for A.

We now use Theorems 12.5.2 and 12.5.3 to show that the ideal class group of an
algebraic number field is finite.

Theorem 12.5.4 Let K be an algebraic number field. Then the ideal class group
H (K ) of K is a finite group (so that the class number h(K ) = card H (K ) is finite).

Proof: By Theorem
s √12.5.3 there are only finitely many integral ideals B of O K
with N (B) ≤ π2 |d(K )|. It follows from Theorem 12.5.2 that each ideal class
is represented by an integral ideal of O K from a finite set. Thus there are only
finitely many ideal classes. Hence H (K ) is a finite group and h(K ) is finite.

The quantity on the right-hand side of the inequality in Theorem 12.5.2 is called
the Minkowski bound for the number field K .
CB609-12 CB609/Alaca & Williams August 7, 2003 17:0 Char Count= 0

314 The Ideal Class Group

Definition 12.5.1 (Minkowski bound) Let K = Q(θ) be an algebraic number

field of degree n. Let r denote the number of real conjugates of θ and s the number
of complex conjugate pairs of nonreal conjugates of θ so that r + 2s = n. The
Minkowski bound for K is denoted by M K and is given by
s
2
MK = |d(K )|.
π

The significance of the Minkowski bound M K is that Theorem 12.5.2 guarantees

that every ideal class of K contains a nonzero integral ideal with norm less than or
equal to M K . In fact by more detailed reasoning it can be shown that every ideal
class contains a nonzero integral ideal with norm less than or equal to
s
4 n!
|d(K )|, (12.5.1)
π nn
but we will not prove this here. As the norm of a nonzero integral ideal is at least
1, we have the inequality
π s n n
|d(K )| ≥ . (12.5.2)
4 n!
nn
Now s ≤ n
2
and n!
≥ 2n−1 so for n ≥ 2 we have
π n/2 1 π
|d(K )| ≥ 2n−1 = π n/2 ≥ > 1
4 2 2
so that
|d(K )| > 1 for K = Q.

12.6 Algorithm to Determine the Ideal Class Group

The results of the previous section give us a method of determining all the ideal
classes of a given algebraic number field K . To determine representatives of the
ideal classes, we need only look at the integral ideals of O K with norm less than
or equal to the Minkowski bound M K . If A is such an ideal then N (P) ≤ M K for
every prime ideal P dividing A. Now N (P) = p f for some rational prime p and
some positive integer f so the prime ideals occurring in the prime factorizations of
the various integral ideals A are all factors of rational primes p ≤ M K . Thus if we
take each rational prime p ≤ M K , determine the prime ideal factorization of p
in O K , and form all possible products of the prime ideal factors of these various
rational primes that yield ideals with norm ≤ M K then we are sure to have at least
one representative of every ideal class.
In particular, if every rational prime ≤ M K factors into a product of prime ideals
of O K , each of which is a principal ideal, then K has class number h(K ) = 1. For
in this case every ideal of the type described here will also be principal.
CB609-12 CB609/Alaca & Williams August 7, 2003 17:0 Char Count= 0

12.6 Algorithm to Determine the Ideal Class Group 315

Algorithm to find the ideal class group H (K ) of an algebraic number field K :

Input. Algebraic number field K = Q(θ ).

Step 1. Determine n = [K : Q].
Step 2. Determine r the number of real conjugates of θ . Then s = 12 (n − r ).
Step 3. Determine d(K ).
Step 4. Compute the Minkowski bound M K = (2/π )s |d(K )|.
Step 5. Determine all rational primes p ≤ M K .
Step 6. Determine the prime ideal factorization of each principal ideal p in O K with p as
in Step 5.
Step 7. Determine all products of these prime ideals having norm ≤ M K .
Step 8. Determine the generators of H (K ) from the classes of these products.
Output. H (K ).

We illustrate this algorithm by finding the ideal class group of several algebraic
number fields.
We denote the class containing the ideal A by [A] and the class of principal ideals
by 1.

√
Example 12.6.1 We show that K = Q( −19) has class number h(K ) = 1. Here

n = 2, r = 0, s = 1, d(K ) = −19.

The Minkowski bound is

s
2 2√ 2
MK = |d(K )| = 19 < · 5 < 4,
π π 3

so that the primes p ≤ M K are p = 2 and 3. As

−19 −19
= = −1,
2 3

the principal ideals 2 and 3 are both prime ideals in O
√K . This is the situation
described just before the algorithm
√ and so h(K ) = h(Q( −19)) = 1. Hence the
1+ −19
√
ring of integers Z + Z 2
of Q( −19) is a principal ideal domain and thus
a unique factorization domain.

√
Example 12.6.2 We show that K = Q( −163) has class number h(K ) = 1. Here

n = 2, r = 0, s = 1, d(K ) = −163.

The Minkowski bound is

s
2 2 √ 2
MK = |d(K )| = 163 < · 13 < 9,
π π 3
CB609-12 CB609/Alaca & Williams August 7, 2003 17:0 Char Count= 0

316 The Ideal Class Group

so that the primes p ≤ M K are p = 2, 3, 5, and 7. As

−163 −163 −163 −163
= = = = −1,
2 3 5 7
the principal
√ ideals 2, 3, 5, 7 are all prime ideals √ h(K ) =
√ in O K . Hence
1+ −163
h(Q( −163)) = 1. Thus the ring of integers Z + Z 2
of Q( −163) is a
principal ideal domain and so a unique factorization domain.
√
Example 12.6.3 We show that K = Q( 23) has class number h(K ) = 1. Here

n = 2, r = 2, s = 0, d(K ) = 92.

The Minkowski bound is

s
2 √
MK = |d(K )| = 92 < 10,
π
so that the primes p ≤ M K are p = 2, 3, 5, and 7. As

92 −1
= = −1,
3 3
3 is a prime ideal in O K . As

92 2
= = −1,
5 5
5 is also a prime ideal in O K . As

92
= 0,
2
2 ramifies in O K . Indeed
√ 2
2 = 2, 1 + 23 .
√
The prime ideal 2, 1 + 23 is principal as
√ √ √ √
2, 1 + 23 = 2, 5 + 23 (as 5 + 23 = 2 · 2 + (1 + 23))
√ √
= 5 + 23 (as 5 + 23 | 2).

Finally, as

92 1
= = 1,
7 7
7 splits in O K . We have
√ √
7 = 7, 3 + 237, 3 − 23.
CB609-12 CB609/Alaca & Williams August 7, 2003 17:0 Char Count= 0

12.6 Algorithm to Determine the Ideal Class Group 317

√
The prime ideal 7, 3 + 23 is principal as
√ √ √ √
7, 3 + 23 = 7, 4 − 23 (as 4 − 23 = 7 − (3 + 23))
√ √
= 4 − 23 (as 4 − 23 | 7).
√ √
Similarly, 7, 3 − 23 = 4 + 23.
Hence every prime ideal in O K dividing a rational prime ≤ M K is principal and
so O K is a principal ideal domain, that is,
√
h(Q( 23)) = 1.

Before continuing we expand on that part√of the calculation in Example 11.6.3

that shows that the prime ideal P = 2, 1 + 23 is principal. We must find α ∈ P
such that P = α. By Exercise 8 of Chapter 9 it suffices to find an α ∈ P √such
that |N (α)| = N (P)√= 2, equivalently, N (α) = ±2. As α is an integer of Q( 23)
we have α = x + y 23 (x, y ∈ Z) so we wish to solve x 2 − 23y 2 = ±2. This can
be done by means of Theorem√11.7.1. However, in this √ case x = 5, y = 1 is an
obvious
√ solution and α = 5 + 23 = (2)2 + (1)(1 + 23) ∈ P. Thus P = α =
5 + 23.

Example 12.6.4 We show that

√
H (Q( −14)) Z4 .
Here
√
K = Q( −14), n = 2, r = 0, s = 1, d(K ) = −56.
The Minkowski bound is
s
2 2 √ 1√
MK = |d(K )| = 56 < 224 < 5.
π π 3

The rational primes p ≤ M K are p = 2 and 3. As

−56 −56
= 0 and = 1,
2 3
we have
2 = P 2 , 3 = P1 P2 ,
where the prime ideals P, P1 , P2 are given by
√ √ √
P = 2, −14, P1 = 3, 1 + −14, P2 = 3, 1 − −14.
The norms of these ideals are
N (P) = 2, N (P1 ) = N (P2 ) = 3.
CB609-12 CB609/Alaca & Williams August 7, 2003 17:0 Char Count= 0

318 The Ideal Class Group

√ √ √
Clearly
√ 2 − −14 ∈ P and 2
√ − −14 ∈ P1 . Hence
√ 2 − −14 ⊆ P and 2 −
−14 ⊆ P1 . Thus P | 2 − −14 and √ P1 | 2 − −14. As P and P1 are dis-
tinct prime ideals, we have P P1 | 2 − −14. Hence there exists an integral ideal
B of O K such that
√
2 − −14 = P P1 B.

Taking norms we obtain

√
18 = N (2 − −14) = N (P)N (P1 )N (B) = 6N (B),

so that N (B) = 3. Hence B = P1 or P2 . If B = P2 then

√
2 − −14 = P P1 P2 = P3,
√
so that 3 | 2 − −14, which is impossible. Hence B = P1 and
√
2 − −14 = P P12 .

Thus
√
[P][P1 ]2 = [P P12 ] = [2 − −14] = 1.

As [P]2 = [P 2 ] = 1 we deduce that [P] = [P1 ]2 and [P1 ]4 = 1. Also, [P1 ][P2 ] =
[P1 P2 ] = 1 = [P1 ]4 so that [P2 ] = [P1 ]3 . Thus 1, [P1 ], [P1 ]2 , and [P1 ]3 comprise
all the ideal classes. We show that these four ideal classes are in fact distinct. We
√ that [P1 ] = 1. Suppose that [P1 ] √= 1. Then [P] = 1 so that the
2 2
do this by proving
ideal P = 2, −14 is principal, say P = x + y −14, where x, y ∈ Z. Then
√ √
2 = N (2, −14) = N (x + y −14) = x 2 + 14y 2 ,
√
which is impossible. This proves that H (Q( −14)) is a cyclic group of order 4
generated by the class of P1 .

It is often useful when determining the ideal class group of an algebraic number
field K = Q(θ ) to calculate N (k + θ) for k = 0, 1, 2, . . . and use those values that
only involve the primes p ≤ M K to find relations among the ideal classes.
This is illustrated in the next example.

Example 12.6.5 We show that

√
H (Q( −65)) Z2 × Z4 .

In this example we have

√
K = Q( −65), n = 2, r = 0, s = 1, d(K ) = −260.
CB609-12 CB609/Alaca & Williams August 7, 2003 17:0 Char Count= 0

12.6 Algorithm to Determine the Ideal Class Group 319

The Minkowski bound is

s
2 2 √
MK = |d(K )| = 260 < 11.
π π
The primes p ≤ M K are p = 2, 3, 5, and 7. As

−260 −260 1
= 0, = = 1,
2 3 3

−260 −260 −1
= 0, = = −1,
5 7 7
we have

2 = P12 , 3 = Q 1 Q 2 , 5 = P22 , 7 = prime ideal,

where
√ √
P1 = 2, 1 + −65, Q 1 = 3, 1 + −65,
√ √
Q 2 = 3, 1 − −65, P2 = 5, −65

are distinct prime ideals. Thus

[P1 ]2 = 1, [Q 1 ][Q 2 ] = 1, [P2 ]2 = 1.

√
Next we calculate the values of N (k + −65) for k = 0, 1, 2, . . . , retaining those
that only involve the primes 2, 3, and 5, until we have enough values to find all the
relations between [P1 ], [Q 1 ], [Q 2 ], and [P2 ]. The first relevant value is
√
N (4 + −65) = 81 = 34 .

We have
√ √
4+ −65 = 3 + (1 + −65) ∈ Q 1 ,

so that
√
4 + −65 ⊆ Q 1

and thus
√
Q 1 | 4 + −65.
√
Suppose that Q 2 | 4 + −65; then
√
3 = Q 1 Q 2 | 4 + −65,
√
which is impossible. Hence Q 2 4 + −65. Let r be the unique positive integer
such that
√
Q r1 || 4 + −65.
CB609-12 CB609/Alaca & Williams August 7, 2003 17:0 Char Count= 0

320 The Ideal Class Group

Then
√
4 + −65 = Q r1 B

for some integral ideal B of O K with Q 1 B and Q 2 B. As Q 1 B and Q 2 B

we see that N (B) is not a power of 3. Taking norms we obtain
√
34 = 81 = N (4 + −65) = N (Q r1 B) = N (Q 1 )r N (B) = 3r N (B),

so that r = 4 and N (B) = 1. Hence B = 1 and

√
4 + −65 = Q 41 ,

showing that

[Q 1 ]4 = 1.

Then, from [Q 1 ][Q 2 ] = 1 = [Q 1 ]4 , we deduce that

[Q 2 ] = [Q 1 ]3 .

Next we find that

√
N (5 + −65) = 90 = 2 · 32 · 5.

We have
√ √
5+ −65 = 2(2) + 1(1 + −65) ∈ P1 ,
√ √
5 + −65 = 2(3) − 1(1 − −65) ∈ Q 2 ,
√ √
5 + −65 = 1(5) + 1( −65) ∈ P2 ,

so that
√ √ √
P1 | 5 + −65, Q 2 | 5 + −65, P2 | 5 + −65.

Hence
√
5 + −65 = P1r Q s2 P2t B,

for positive integers r, s, t and an integral ideal B of O K with P1 B, Q 2 B, P2

B. We show next that Q 1 B. Suppose that Q 1 | B. Then
√
3 = Q 1 Q 2 | 5 + −65,

which is impossible. Hence Q 1 B. Since Q 1 B and Q 2 B the norm of B cannot

be a power of 3. Taking norms we obtain
√
2 · 32 · 5 = 90 = N (5 + −65) = 2r · 3s · 5t N (B),
CB609-12 CB609/Alaca & Williams August 7, 2003 17:0 Char Count= 0

12.6 Algorithm to Determine the Ideal Class Group 321

so that r = 1, t = 1, and N (B) = 32−s . As N (B) is not a power of 3, we have

s = 2, N (B) = 1, B = 1. Hence
√
5 + −65 = P1 Q 22 P2 .
Thus
[P1 ][Q 2 ]2 [P2 ] = 1.
Hence
[P2 ] = [P1 ][Q 2 ]2 [P2 ]2 = [P1 ][Q 1 ]6 = [P1 ][Q 1 ]2 .
Thus all the ideal classes of K lie among
1, [Q 1 ], [Q 1 ]2 , [Q 1 ]3 , [P1 ], [P1 ][Q 1 ], [P1 ][Q 1 ]2 , [P1 ][Q 1 ]3 .
We claim that all these classes are distinct. We show first that
1, [Q 1 ], [Q 1 ]2 , [Q 1 ]3
are distinct. As [Q 1 ]4 = 1 it suffices to show that [Q 1 ]2 = 1. If [Q 1 ]2 = 1 then Q 21
is a principal ideal, say,
√
Q 21 = x + y −65, x, y ∈ Z.
Taking norms we deduce that
9 = x 2 + 65y 2 ,
so that x = ±3, y = 0. Then
Q 21 = 3 = Q 1 Q 2
and so Q 1 = Q 2 , contradicting that Q 1 and Q 2 are distinct ideals.
√ H = {1, [Q 1 ], [Q 1 ] , [Q 1 ] } is a subgroup of order 4 of G =
2 3
Thus
H (Q( −65)), so by Lagrange’s theorem, 4 | |G|. If |G| < 8 then |G| = 4 and so
G = H . Hence [P1 ] = 1, [Q 1 ], [Q 1 ]2 , or [Q 1 ]3 . Now [P1 ] = 1 as 2 = x 2 + 65y 2
for integers x and y. Thus ord[P1 ] = 2. As ord[Q 1 ] = ord[Q√ 1 ] = 4 we must have
3

[P1 ] = [Q 1 ] . Hence [P1 Q 1 ] = 1 so that P Q 1 = x + y −65 for integers x and

2 2 2

y. Then
18 = N (P1 Q 21 ) = x 2 + 65y 2 ,
which is impossible. This proves that
√
H (Q( −65)) = {1, [Q 1 ], [Q 1 ]2 , [Q 1 ]3 , [P1 ], [P1 ][Q 1 ], [P1 ][Q 1 ]2 , [P1 ][Q 1 ]3 },
where
[Q 1 ]4 = [P1 ]2 = 1,
CB609-12 CB609/Alaca & Williams August 7, 2003 17:0 Char Count= 0

322 The Ideal Class Group

Table √
5. Nontrivial ideal class groups
H (Q( k)), −30 < k < 0, k squarefree
√
k H (Q( k))
√
−5 {1, A} Z2 , A = [2, 1√+ −5], A2 = 1
−6 {1, A} Z2 , A = [2, √−6], A2 = 1
−10 {1, A} Z2 , A = [2, −10], √ A2 = 1
−13 {1, A} Z2 , A = [2, 1 + −13],√ A2 = 1
−14 {1, A, A2 , √
A3 } Z4 , A = [3, 1 +√ −14],
A = [2, −14], A3 = [3,√1 − −14], A4 = 1
2

−15 {1, A} Z2 , A = [2, 12 (3 + −15)], √ A =1

−17 {1, A, A , A } √
2 3
Z4 , A = [3, 1 + −17], √
A2 = [2, 1 + −17], A3 = [3, 1 − √ −17], A4 = 1
−21 {1, A, B, √AB} Z2 × Z2 , A = [2, √1 + −21],
B = [3, −21], AB = [5, 3 + −21],
A2 = B 2 = (AB)2 = 1 √
−22 {1, A} Z2 , A = [2, −22], A√2 = 1
−23 {1, A, A2 } Z3 ,√A = [2, 12 (1 + −23)],
A2 = [2, 12 (1 − −23)], A3 = 1 √
−26 {1, A, A2 , A3 , √
A4 , A5 } Z6 , A = √[5, 2 + −26],
A2 = [3, 1 − √−26], A3 = [2, −26], √
A4 = [3, 1 + −26], A5 = [5, 2 − −26], √ A =1
6

−29 {1, A, A , A , √
2 3
A , A } Z6 , A = [3, √
4 5
1 + −29],
A2 = [5, 1 + √−29], A3 = [2, 1 +√ −29],
A4 = [5, 1 − −29], A5 = [3, 1 −√ −29], A6 = 1
−30 {1, A, B, √AB} Z2 × Z2 , A √ = [2, −30],
B = [3, −30], AB = [5, −30],
A2 = B 2 = (AB)2 = 1

Note: Excluded fields have class number 1.

so that
√
H (Q( −65)) Z4 × Z2 .

Using the method illustrated in Examples 12.6.1–12.6.5 we can construct Tables

5 and 6 of class groups.
For a quadratic field K , Dirichlet has given an explicit formula for h(K ). We
refer the reader to [1, p. 342] for a proof.

Theorem 12.6.1 Let K be a quadratic field of discriminant d. Then

|d|−1
−w(d) d
h(K ) = r , if d < 0,
2|d| r =1 r
CB609-12 CB609/Alaca & Williams August 7, 2003 17:0 Char Count= 0

12.6 Algorithm to Determine the Ideal Class Group 323

Table √
6. Nontrivial ideal class groups
H (Q( k)), 2 ≤ k < 100, k squarefree
√
k H (Q( k))
√
10 {1, A} Z2 , A = [2, 10], √ A = 21
2

15 {1, A} Z2 , A = [2, 1√+ 15], A = 1

26 {1, A} Z2 , A = [2, √26], A2 = 1
30 {1, A} Z2 , A = [2, 30], √ A = 21
2

34 {1, A} Z2 , A = [3, 1 + √34], A = 1

35 {1, A} Z2 , A = [2, 1 + √35], A2 = 1
39 {1, A} Z2 , A = [2, 1√+ 39], A2 = 1
42 {1, A} Z2 , A = [2, √42], A2 = 1
51 {1, A} Z2 , A = [3, 51], √ A = 21
2

55 {1, A} Z2 , A = [2, 1√+ 55], A = 1

58 {1, A} Z2 , A = [2, √58], A2 = 1
65 {1, A} Z2 , A = [5, √65], A2 = 1
66 {1, A} Z2 , A = [3, √66], A2 = 1
70 {1, A} Z2 , A = [2, √70], A2 = 1
74 {1, A} Z2 , A = [2, √74], A2 = 1
78 {1, A} Z2 , A = [2, 78], √ A2 = 1
79 {1, A, A } Z√
2
3 , A = [3, 1 + 79],
A2 = [3, 1 − 79], A3 = 1 √
82 {1, A, A2 , √
A3 } Z4 , A = [3, 2 √
+ 82],
A2 = [2, 82], A3 = [3, 2 − 82],
A4 = 1 √
85 {1, A} Z2 , A = [5, 85], √ A = 21
2

87 {1, A} Z2 , A = [2, 1 + √87], A = 1

91 {1, A} Z2 , A = [2, 1 + √91], A2 = 1
95 {1, A} Z2 , A = [2, 1 + 95], A2 = 1

Note: Excluded fields have class number 1.

and

−1 d πr
h(K ) = log sin , if d > 0.
log η 1≤r <d/2 r d

Here w(d) denotes the number of roots of unity in OQ(√d) (d < 0) so that

 6, if d = −3,
w(d) = 4, if d = −4,

2, if d < −4,

where dn (n ∈ N) is the Kronecker symbol and η is the fundamental unit of
OQ(√d) (d > 0).
CB609-12 CB609/Alaca & Williams August 7, 2003 17:0 Char Count= 0

324 The Ideal Class Group

√
Example 12.6.6 We use Dirichlet’s formula to show that h(Q( −15)) = 2. Here
d = −15, w(d) = 2, and Theorem 12.6.1 gives

14
√ 1 −15
h(Q( −15)) = − r
15 r =1 r
−1
= (1(1) + 2(1) + 3(0) + 4(1) + 5(0) + 6(0) + 7(−1) + 8(1)
15
+ 9(0) + 10(0) + 11(−1) + 12(0) + 13(−1) + 14(−1))
−1
= (1 + 2 + 4 − 7 + 8 − 11 − 13 − 14)
15
−1
= (−30) = 2.
15

√
√ Dirichlet’s formula to show that h(Q( 5)) = 1. In this
Example 12.6.7 We use
case d = 5, η = (1 + 5)/2, and Theorem 12.6.1 gives

2

√ −1 5 πr
h(Q( 5)) = √ log sin
1+ 5 r =1
r 5
log
2

−1 π 2π
= √ log sin − log sin
1+ 5 5 5
log
2

1 2π π
= √ log sin − log sin
1+ 5 5 5
log
2
√ √
1 10 + 2 5 10 − 2 5
= √ log − log
1+ 5 4 4
log
2
√
3+ 5
√ log
1 10 + 2 5 2
= √ log √ = √ = 1.
1+ 5 10 − 2 5 1+ 5
2 log 2 log
2 2

√
Tables 7 and 8, which give the class numbers of quadratic fields Q( k) with k
squarefree between −195 and 197, can be constructed using Dirichlet’s formula.
We conclude this section by determining the ideal class group for two cubic fields
and a quartic field.
CB609-12 CB609/Alaca & Williams August 7, 2003 17:0 Char Count= 0

12.6 Algorithm to Determine the Ideal Class Group 325

Table 7. Class
√ numbers of imaginary quadratic fields
K = Q( k), − 195 ≤ k < 0, k squarefree

k h(K ) k h(K ) k h(K ) k h(K ) k h(K )

−1 1 −38 6 −78 4 −115 2 −158 8
−2 1 −39 4 −79 5 −118 6 −159 10
−3 1 −41 8 −82 4 −119 10 −161 16
−5 2 −42 4 −83 3 −122 10 −163 1
−6 2 −43 1 −85 4 −123 2 −165 8
−7 1 −46 4 −86 10 −127 5 −166 10
−10 2 −47 5 −87 6 −129 12 −167 11
−11 1 −51 2 −89 12 −130 4 −170 12
−13 2 −53 6 −91 2 −131 5 −173 14
−14 4 −55 4 −93 4 −133 4 −174 12
−15 2 −57 4 −94 8 −134 14 −177 4
−17 4 −58 2 −95 8 −137 8 −178 8
−19 1 −59 3 −97 4 −138 8 −179 5
−21 4 −61 6 −101 14 −139 3 −181 10
−22 2 −62 8 −102 4 −141 8 −182 12
−23 3 −65 8 −103 5 −142 4 −183 8
−26 6 −66 8 −105 8 −143 10 −185 16
−29 6 −67 1 −106 6 −145 8 −186 12
−30 4 −69 8 −107 3 −146 16 −187 2
−31 3 −70 4 −109 6 −149 14 −190 4
−33 4 −71 7 −110 12 −151 7 −191 13
−34 4 −73 4 −111 8 −154 8 −193 4
−35 2 −74 10 −113 8 −155 4 −194 20
−37 2 −77 8 −114 8 −157 6 −195 4

Example 12.6.8 We show that H (K ) is trivial for the cubic field K = Q(θ), where
θ 3 + θ + 1 = 0 (see Example 7.1.3). Thus O K is a principal ideal domain and so
is a unique factorization domain. Here

D(θ) = −4 · 13 − 27 · 12 = −31

is negative so that x 3 + x + 1 = 0 has one real root and two nonreal roots. Thus

r = 1, s = 1.

As D(θ) is squarefree, K = Z + Zθ + Zθ 2 and

d(K ) = D(θ) = −31.

The Minkowski bound is

s
2 2 √ 2
MK = |d(K )| = 31 < · 6 = 4,
π π 3
so that the primes p ≤ M K are p = 2 and 3. The polynomial x 3 + x + 1 is
CB609-12 CB609/Alaca & Williams August 7, 2003 17:0 Char Count= 0

326 The Ideal Class Group

Table 8. Class
√ numbers of real quadratic fields
K = Q( k), 0 < k ≤ 197, k squarefree

k h(K ) k h(K ) k h(K ) k h(K ) k h(K )

2 1 39 2 79 3 118 1 159 2
3 1 41 1 82 4 119 2 161 1
5 1 42 2 83 1 122 2 163 1
6 1 43 1 85 2 123 2 165 2
7 1 46 1 86 1 127 1 166 1
10 2 47 1 87 2 129 1 167 1
11 1 51 2 89 1 130 4 170 4
13 1 53 1 91 2 131 1 173 1
14 1 55 2 93 1 133 1 174 2
15 2 57 1 94 1 134 1 177 1
17 1 58 2 95 2 137 1 178 2
19 1 59 1 97 1 138 2 179 1
21 1 61 1 101 1 139 1 181 1
22 1 62 1 102 2 141 1 182 2
23 1 65 2 103 1 142 3 183 2
26 2 66 2 105 2 143 2 185 2
29 1 67 1 106 2 145 4 186 2
30 2 69 1 107 1 146 2 187 2
31 1 70 2 109 1 149 1 190 2
33 1 71 1 110 2 151 1 191 1
34 2 73 1 111 2 154 2 193 1
35 2 74 2 113 1 155 2 194 2
37 1 77 1 114 2 157 1 195 4
38 1 78 2 115 2 158 1 197 1
√
Gauss conjectured that Q( k) has class number 1 for infinitely many
squarefree k ∈ N. It is still not known whether this conjecture is true or
false.

irreducible (mod 2), so that the principal ideal 2 is prime in O K . The factor-
ization of x 3 + x + 1 into irreducibles (mod 3) is
x 3 + x + 1 ≡ (x − 1)(x 2 + x − 1) (mod 3),
so that by Theorem 10.3.1 the factorization of 3 into prime ideals in O K is
3 = P Q,
where
P = 3, θ − 1, N (P) = 3,
Q = 3, θ 2 + θ − 1, N (Q) = 32 .
Now
(θ − 1)3 + 3(θ − 1)2 + 4(θ − 1) + 3 = 0,
CB609-12 CB609/Alaca & Williams August 7, 2003 17:0 Char Count= 0

12.6 Algorithm to Determine the Ideal Class Group 327

so that

θ − 1 | 3,

and thus

P = 3, θ − 1 = θ − 1.

Further,
3
= −4 − 3(θ − 1) − (θ − 1)2 = −2 − θ − θ 2
θ −1
so that
3
Q = 3P −1 = 3θ − 1−1 =
θ −1
= −2 − θ − θ 2 = 2 + θ + θ 2 .

Hence all the prime ideals dividing the principal ideals p ( p (prime) ≤ M K ) are
principal so that the ideal class group H (K ) is trivial.
√ √
Example
√ 12.6.9 We show that H (Q( 3
2)) is trivial. Let θ = 3
2 and K = Q(θ) =
Q( 3 2). Clearly, irrQ θ = x 3 − 2, which has one real root (namely θ) and two
nonreal roots (namely ωθ and ω2 θ, where ω is a complex cube root of unity). Thus

r = 1, s = 1.

It was shown in Example 7.1.6 that {1, θ, θ 2 } is an integral basis for K and d(K ) =
−108. The Minkowski bound is
s
2 2√ 2 21
MK = |d(K )| = 108 < · = 7.
π π 3 2
Thus the primes p ≤ M K are p = 2, 3, and 5. Clearly,

2 = P 3 ,

where P = θ is a principal prime ideal of norm 2. Also,

3 = Q 3 ,

where Q = θ + 1 is a principal ideal of norm 3. This is clear as

Q 3 = θ + 13 = (θ + 1)3 = θ 3 + 3θ 2 + 3θ + 1 = 3 + 3θ + 3θ 2

= 3(1 + θ + θ 2 ) = 3,

since 1 + θ + θ 2 is a unit of O K as

(1 + θ + θ 2 )(−1 + θ ) = −1 + θ 3 = 1.
CB609-12 CB609/Alaca & Williams August 7, 2003 17:0 Char Count= 0

328 The Ideal Class Group

Finally, as
x 3 − 2 = (x + 2)(x 2 − 2x − 1) (mod 5),
by Theorem 10.3.1 we have
5 = P Q,
where P and Q are distinct prime ideals with
P = 5, 2 + θ , N (P) = 5,
Q = 5, −1 − 2θ + θ 2 , N (Q) = 52 .
Now
5 = 4 + 1 = θ 6 + 1 = (θ 2 + 1)(θ 4 − θ 2 + 1) = (θ 2 + 1)(1 + 2θ − θ 2 ),
so that 1 + 2θ − θ 2 | 5 and thus
Q = 1 + 2θ − θ 2
and
P = 5Q −1 = 51 + 2θ − θ 2 −1
= 5(1 + 2θ − θ 2 )−1 = 5(1 + 2θ − θ 2 )−1
5
= 1 + θ 2 .
1 + 2θ − θ 2
√
Since√all the prime factors of 2, 3, and 5 are principal, H (Q( 3 2)) is trivial. Hence
h(Q( 3 2)) = 1.
√
Q( 3 k) for cubefree
The class numbers of √ √ positive
√ integers
√k up to 101 are given
in Table 9. Note that Q( 3 −k) = Q( 3 k) and Q( k 2 ) = Q( 3 k).
3

Example√12.6.10 We show that the ideal class group H (K ) of the quartic field
K = Q( 2 + i) is trivial. We have already observed that K is a cyclotomic field,
namely, K = Q(ζ8 ) = K 8 . Thus, by Theorem 7.5.2, K is a monogenic
√ field.
√ Indeed
O K = Z + Zζ8 + Zζ8 + Zζ8 , by Theorem 7.5.1. Set θ = ζ8 = ( 2 + −2)/2.
2 3

The minimal√ polynomial

√ of θ is irrQ θ = x 4 + 1, which has four nonreal roots,
namely, 12 (± 2 ± i 2). Thus r = 0, s = 2. It was shown in Example 7.1.7 that
d(K ) = 256. Thus the Minkowski bound M K satisfies
s 2
2 2 √ 64
MK = |d(K )| = 256 = 2 < 7.
π π π
Hence the primes p ≤ M K are p = 2, 3, and 5.
The factorization of x 4 + 1 into irreducible polynomials modulo 2 is given by
x 4 + 1 ≡ (x + 1)4 (mod 2)
CB609-12 CB609/Alaca & Williams August 7, 2003 17:0 Char Count= 0

12.6 Algorithm to Determine the Ideal Class Group 329

√Table 9. Class numbers of

Q( 3 k), 2 ≤ k ≤ 101, k cubefree
√ √ √
k h(Q( 3 k)) k h(Q( 3 k)) k h(Q( 3 k))
2 1 37 3 69 1
3 1 38 3 70 9
5 1 39 6 71 1
6 1 41 1 73 3
7 3 42 3 74 3
10 1 43 12 76 6
11 2 44 1 77 3
12 1 45 1 78 3
13 3 46 1 79 6
14 3 47 2 82 1
15 2 51 3 83 2
17 1 52 3 84 3
19 3 53 1 85 3
20 3 55 1 86 9
21 3 57 6 87 1
22 3 58 6 89 2
23 1 59 1 90 3
26 3 60 3 91 9
28 3 61 6 92 3
29 1 62 3 93 3
30 3 63 6 94 3
31 3 65 18 95 3
33 1 66 6 97 3
34 3 67 6 99 1
35 3 68 3 101 2

so that by Theorem 10.3.1 the factorization of the principal ideal 2 into prime
ideals in O K is

2 = P14 ,

where

P1 = 2, 1 + θ , N (P1 ) = 2.

Now

(1 + θ)4 − 4(1 + θ )3 + 6(1 + θ )2 − 4(1 + θ ) + 2 = 0,

so that 1 + θ | 2. Hence

P1 = 2, 1 + θ = 1 + θ

is principal.
CB609-12 CB609/Alaca & Williams August 7, 2003 17:0 Char Count= 0

330 The Ideal Class Group

The factorization of x 4 + 1 into irreducible polynomials modulo 3 is given by

x 4 + 1 ≡ (x 2 + x − 1)(x 2 − x − 1) (mod 3),

so that by Theorem 10.3.1 the factorization of the principal ideal 3 into prime
ideals in O K is

3 = P2 P3 ,

where

P2 = 3, 1 − θ − θ 2 , P3 = 3, 1 + θ − θ 2 , N (P2 ) = N (P3 ) = 32 , P2 = P3 .

As θ 4 + 1 = 0 we see that θ | 1, so that θ is a unit of O K . Further, as

(1 − θ − θ 2 )(1 + θ − θ 2 ) = −3θ 2 ,

we deduce that

1 − θ − θ 2 | 3 and 1 + θ − θ 2 | 3,

so that the ideals

P2 = 1 − θ − θ 2 , P3 = 1 + θ − θ 2

are principal.
The factorization of x 4 + 1 into irreducible polynomials modulo 5 is given by

x 4 + 1 ≡ (x 2 + 2)(x 2 − 2) (mod 5),

so that by Theorem 10.3.1 the factorization of the principal ideal 5 into prime
ideals in O K is

5 = P4 P5 ,

where

P4 = 5, 2 + θ 2 , P5 = 5, −2 + θ 2 , N (P4 ) = N (P5 ) = 52 , P4 = P5 .

Now

(2 + θ 2 )(−2 + θ 2 ) = −4 + θ 4 = −5,

so that

2 + θ 2 | 5, − 2 + θ 2 | 5.

Hence the ideals

P4 = 2 + θ 2 , P5 = −2 + θ 2

are principal.
CB609-12 CB609/Alaca & Williams August 7, 2003 17:0 Char Count= 0

12.7 Applications to Binary Quadratic Forms 331

Table 10. Class numbers of

cyclotomic fields K m , 3 ≤ m ≤ 45,
m ≡ 2 (mod 4)

m h(K m ) m h(K m ) m h(K m )

3 1 17 1 32 1
4 1 19 1 33 1
5 1 20 1 35 1
7 1 21 1 36 1
8 1 23 3 37 37
9 1 24 1 39 2
11 1 25 1 40 1
12 1 27 1 41 121
13 1 28 1 43 211
15 1 29 8 44 1
16 1 31 9 45 1

Note: K 2n = K n for n odd.

We have shown that all the prime ideals dividing the principal ideals
p, where p is a prime ≤ M K , are principal so that the ideal class group H (K ) is
trivial. Hence h(K 8 ) = 1.

We conclude this section with a short table of class numbers of cyclotomic fields
(Table 10).

12.7 Applications to Binary Quadratic Forms

√
Let m be a squarefree integer. Let K be the quadratic field Q( m). The discriminant
d(K ) is given by

d(K ) = 22δ m, (12.7.1)

where
!
0, if m ≡ 1 (mod 4),
δ= (12.7.2)
1, if m ≡ 2 or 3 (mod 4).
Let p be an odd prime such that

d(K )
= 1. (12.7.3)
p

We observe that (12.7.3) is equivalent to m
p
= 1 as p = 2. Then

p = P1 P2 ,
CB609-12 CB609/Alaca & Williams August 7, 2003 17:0 Char Count= 0

332 The Ideal Class Group

where P1 and P2 are distinct conjugate prime ideals of O K . Let h denote the class
number h(K ). Then

p h = ph = P1h P2h .

As [P1 ] ∈ H (K ) and card(H (K )) = h, we have

[P1h ] = [P1 ]h = 1.

Thus P1h is a principal ideal, say,

√
x+y m
P1h = ,
21−δ
where x and y are rational integers with x ≡ y (mod 2) if m ≡ 1 (mod 4). As P2h
is the conjugate ideal of P1h , we have
√
x−y m
P2 =
h
,
21−δ
so that
x 2 − my 2
ph = .
41−δ
Thus

x 2 − my 2
ph = θ
41−δ
for some θ ∈ U (O K ). But
41−δ p h
θ= ∈Q
x 2 − my 2
so that θ ∈ U (Z) = {±1}. Hence

41−δ p h = ±(x 2 − my 2 ), (12.7.4)

showing that 41−δ p h is represented by one or both of the binary quadratic forms
x 2 − my 2 and −x 2 + my 2 .
We prove that
!
1, if m ≡ 2 or 3 (mod 4),
(x, y) = (12.7.5)
1 or 2, if m ≡ 1 (mod 4).
Suppose that q is an odd prime with q | (x, y). Then x = q x1 and y = qy1 for
integers x1 and y1 with x1 ≡ y1 (mod 2) if m ≡ 1 (mod 4). From (12.7.4) we deduce
that q 2 | 41−δ p h . As q = 2 we must have q = p. Thus
√ √
x1 + y1 m x1 + y1 m
P1 = p
h
= P1 P2 ,
21−δ 21−δ
CB609-12 CB609/Alaca & Williams August 7, 2003 17:0 Char Count= 0

12.7 Applications to Binary Quadratic Forms 333

so that P2 | P1h , contradicting that P1 and P2 are distinct prime ideals. Hence there
are no odd primes dividing (x, y) and so

(x, y) = 2w

for some nonnegative integer w. From (12.7.4) we deduce that 22w | 22(1−δ) so that
0 ≤ w ≤ 1 − δ. If m ≡ 2 or 3 (mod 4) then δ = 1 and w = 0. If m ≡ 1 (mod 4)
then δ = 0 and w = 0 or 1.
If m is negative then x 2 − my 2 > 0, so the plus sign holds in (12.7.4). If m is
positive and there exist integers T and U such that T 2 − mU 2 = −1 then

−(x 2 − my 2 ) = (T 2 − mU 2 )(x 2 − my 2 ) = (T x + mU y)2 − m(T y + U x)2

and the plus sign holds in (12.7.4).

If m ≡ 1 (mod 8) then δ = 0 (by (12.7.2)) and (12.7.4) gives

x 2 − y 2 ≡ x 2 − my 2 = ±4 p h ≡ 4 (mod 8),

so that x ≡ y ≡ 0 (mod 2). Setting x = 2u, y = 2v (u, v ∈ Z), we obtain from

(12.7.4)

p h = ±(u 2 − mv 2 ).

From (12.7.5) we deduce that (u, v) = 1.

If m ≡ 5 (mod 8) then δ = 0 (by (12.7.2)) and x ≡ y (mod 2). Set

x = v + 2u, y = v,

where u, v ∈ Z. Then (12.7.4) becomes

±4 p h = x 2 − my 2 = (v + 2u)2 − mv 2 = 4u 2 + 4uv + (1 − m)v 2

so that

p h = ±(u 2 + uv + ( 1−m
4
)v 2 ).

From (12.7.5) we have

(2u, v) = (v + 2u, v) = (x, y) = 1 or 2,

so that (u, v) = 1 or 2. But (u, v)2 | p h , so that (u, v) = p t for some nonnegative
integer t. Hence t = 0 and (u, v) = 1. We have proved the following result.

Theorem
12.7.1 Let m be a squarefree integer. Let p be an odd prime with
√
m
p
= 1. Let h denote the class number of the quadratic field Q( m).
CB609-12 CB609/Alaca & Williams August 7, 2003 17:0 Char Count= 0

334 The Ideal Class Group

If m is negative or m is positive and there are integers T and U such that

T 2 − mU 2 = −1 then there exist coprime integers u and v such that

 u 2 − mv 2 , if m ≡ 1 (mod 8) or m ≡ 2, 3 (mod 4),
p =
h
1
 u 2 + uv + (1 − m)v 2 , if m ≡ 5 (mod 8).
4
Otherwise there exist coprime integers u and v such that

 u 2 − mv 2 or − u 2 + mv 2 , if m ≡ 1 (mod 8) or m ≡ 2, 3 (mod 4),
p =
h
1 1
 u 2 + uv + (1 − m)v 2 or − u 2 − uv − (1 − m)v 2 , if m ≡ 5 (mod 8).
4 4

The reader should compare this theorem with Theorems 1.4.4 and 1.4.5.
In the opposite direction to Theorem 12.7.1 we have the following simple result.

Theorem 12.7.2 Let ax 2 + bx y + cy 2 be an integral binary quadratic form of

discriminant
d. Let p be an odd prime with p a. Let k be a positive integer.
If p = −1 then there do not exist coprime integers u and v such that
d

p k = au 2 + buv + cv 2 . (12.7.6)

Proof: Suppose on the contrary that d
p
= −1 and there are coprime integers u
and v satisfying (12.7.6). Then, as d = b − 4ac, we have
2

4ap k = (2au + bv)2 − dv 2 . (12.7.7)

From (12.7.7) we see that
p | v =⇒ p | 2au + bv =⇒ p | u, as p 2a,
contradicting that (u, v) = 1. Hence p v. Then there exists an integer w such that
vw ≡ 1 (mod p). Thus
((2au + bv)w)2 ≡ dv 2 w2 ≡ d (mod p),

so that dp = 0 or 1, contradicting that dp = −1. This proves that no such
integers u and v exist.

In the next three examples we apply Theorems 12.7.1 and 12.7.2 in the cases
m = −1, −2, and −3. We recover Theorems 2.5.1, 2.5.2, and 2.5.3 respectively
(see Exercises 12, 14, and 16 of Chapter 2).
√
Example
12.7.1 m = −1. Here h = h(Q( −1)) = 1. If p is an odd prime with
−1
p
= 1, by Theorem 12.7.1 there exist (coprime) integers u and v such that
p = u 2 + v 2 . Conversely, by Theorem 12.7.2, if there exist integers u and v such
CB609-12 CB609/Alaca & Williams August 7, 2003 17:0 Char Count= 0

12.7 Applications to Binary Quadratic Forms 335

that p = u 2 + v 2 (so that (u, v) = 1) then −4 = 0 or 1, that is, −1 = 1 as
p p

p = 2. Since −1p
= 1 ⇐⇒ p ≡ 1 (mod 4), and 2 = 12 + 12 , we deduce that for
a prime p

p = u 2 + v 2 ⇐⇒ p = 2 or p ≡ 1 (mod 4)

(see Theorem 2.5.1 and Exercise 12 of Chapter 2).

√
Example
12.7.2 m = −2. Here h = h(Q( −2)) = 1. Let p be an odd prime. If
−2
p
= 1, then, by Theorem 12.7.1, there exist (coprime) integers u and v such
that p = u 2 + 2v 2 . Conversely,
by Theorem
12.7.2, if there
exist
integers u and
−8 −2
v such that p = u + 2v then p = 0 or 1, so that p = 1 as p is odd.
2 2

Since −2p
= 1 ⇐⇒ p ≡ 1, 3 (mod 8), and 2 = 02 + 2 · 12 , we deduce that for a
prime p

p = u 2 + 2v 2 ⇐⇒ p = 2 or p ≡ 1, 3 (mod 8)

(see Theorem 2.5.2 and Exercise 14 of Chapter 2).

√
Example
12.7.3 m = −3. Here h = h(Q( −3)) = 1. Let p be an odd prime. If
−3
p
= 1, then, by Theorem 12.7.1, there exist (coprime) integers u and v such
that p = u 2 + uv + v 2 . Conversely, by
Theorem
12.7.2, if there exist integers
u and
v such that p = u + uv + v then p = 0 or 1, so that p = 3 or −3
2 2 −3
= 1.
p

Since −3p
= 1 ⇐⇒ p ≡ 1 (mod 3), 2 = u 2 + uv + v 2 , and 3 = 12 + 1 · 1 + 12 ,
we have for a prime p

p = u 2 + uv + v 2 ⇐⇒ p = 3 or p ≡ 1 (mod 3)

(see Theorem 2.5.3 and Exercise 16 of Chapter 2).

√
We now give an example with h(Q( m)) > 1.
√
Example
12.7.4 m = −5. Here h = h(Q( −5)) = 2. If p is an odd prime with
−5
p
= 1, by Theorem 12.7.1 there exist coprime integers u and v such that p 2 =
u 2 + 5v 2 . Conversely, by Theorem 12.7.2, if there exist coprime
integers u and v
such that p = u + 5v , where p is an odd prime, then −5
2 2 2
p
= 0 or 1. Hence
p = 5 or p ≡ 1, 3, 7, or 9 (mod 20). Clearly 22 , 52 = u 2 + 5v 2 with (u, v) = 1, so
that for a prime p

p 2 = u 2 + 5v 2 , (u, v) = 1 ⇐⇒ p ≡ 1, 3, 7, 9 (mod 20).

CB609-12 CB609/Alaca & Williams August 7, 2003 17:0 Char Count= 0

336 The Ideal Class Group

We note that

292 = 112 + 5 · 122 and 29 = 32 + 5 · 22 ,

32 = 22 + 5 · 12 but 3 = u 2 + 5v 2 .

We return to the representability of p by the form u 2 + 5v 2 in Example 12.7.5.

Theorem 12.7.1 is concerned with the representability of p h by a binary quadratic

form of discriminant m or 4m. But what about the representability of p itself by
such a form? To tackle this problem we must use our knowledge of the generators of
√
the ideal class group of Q( m). We illustrate the ideas involved with two examples.
√
Example 12.7.5 The ideal class group of K = Q( −5) is
√
H (K ) = H (Q( −5)) = {1, [Q]} Z2 ,

where [Q] is the class of the ideal Q given by

√
Q = 2, 1 + −5, Q 2 = 2, N (Q) = 2

(see Table 5). Let p = 2, 5 be a prime such that −5
p
= 1, so that

p = P1 P2 ,

where P1 and P2 are distinct conjugate prime ideals of O K . Hence√[P1 ] = 1 or

[P1 ] = [Q]. In the first case P1 is a principal ideal, say, P1 = x + y −5, where
x, y ∈ Z, so that
√ √
p = N (P1 ) = N (x + y −5) = |N (x + y −5)| = x 2 + 5y 2 .

In the second case [P1 Q] = [P1 ][Q] √

= [Q]2 = [Q 2 ] = [2] = 1, so that P1 Q is
a principal ideal, say, P1 Q = x + y −5, where x, y ∈ Z, and
√
2 p = N (P1 )N (Q) = N (P1 Q) = N (x + y −5) = x 2 + 5y 2 .

Hence we have shown that for a prime p = 2, 5

−5
= 1 =⇒ p or 2 p = x 2 + 5y 2 for integers x and y.
p

Conversely, if p = 2, 5 is a prime with p or 2 p = x 2 + 5y 2 for some integers x

and y then p y so that yz ≡ 1 (mod p) for some integer z and thus, as x 2 ≡ −5y 2
(mod p), we have

(x z)2 ≡ − 5(yz)2 ≡ −5 (mod p),

so that −5
p
= 0 or 1. As p =
5 we have −5
p
= 1. Hence, for a prime p = 2, 5,
CB609-12 CB609/Alaca & Williams August 7, 2003 17:0 Char Count= 0

12.7 Applications to Binary Quadratic Forms 337

we have shown that

−5
p or 2 p = x + 5y ⇐⇒
2 2
= 1.
p

Now suppose that p = x 2 + 5y 2 . Working modulo 4, we obtain

−1
p = x 2 + 5y 2 =⇒ p ≡ x 2 + y 2 (mod 4) =⇒ p ≡ 1 (mod 4) =⇒ =1
p
and modulo 5, we obtain

p
p = x + 5y =⇒ p ≡ x (mod 5) =⇒ p ≡ 1, 4 (mod 5) =⇒
2 2 2
= 1.
5

However, if 2 p = x 2 + 5y 2 then modulo 8 we deduce

−1
2 p = x + 5y =⇒ 2 p ≡ 6 (mod 8) =⇒ p ≡ 3 (mod 4) =⇒
2 2
= −1
p
and modulo 5 we get

2 p = x 2 + 5y 2 =⇒ 2 p ≡ x 2 (mod 5) =⇒ 2 p ≡ 1 or 4 (mod 5)

p
=⇒ p ≡ 2, 3 (mod 5) =⇒ = −1.
5
By the law of quadratic reciprocity, we have

−1 p −1 5 −5
= = = 1.
p 5 p p p
Hence, for a prime p = 2, 5, we have proved that

−1 p
p = x + 5y ⇐⇒
2 2
= = 1,
p 5

−1 p
2 p = x + 5y ⇐⇒
2 2
= = −1.
p 5

When p is a prime such that

−1 p
= = −1
p 5
(for example a prime p ≡ 3 (mod 20)) 2 p is represented by a binary quadratic form
of discriminant −20, namely x 2 + 5y 2 . But what about the representation of p itself
by a binary quadratic form of discriminant −20? It cannot be the form x 2 + 5y 2
but maybe there is some other form of discriminant −20 that represents p. We show
that this is indeed the case and that the form can be taken to be 2x 2 + 2x y + 3y 2 .
CB609-12 CB609/Alaca & Williams August 7, 2003 17:0 Char Count= 0

338 The Ideal Class Group

Suppose that p = 2, 5 is a prime such that

−1 p
= = −1.
p 5

Then there exist integers y and z such that 2 p = z 2 + 5y 2 . Clearly z ≡ y (mod 2).
Thus we can define an integer x by z = y + 2x. Then

2 p = (y + 2x)2 + 5y 2 = 4x 2 + 4x y + 6y 2

so that

p = 2x 2 + 2x y + 3y 2 .

The form 2x 2 + 2x y + 3y 2 has discriminant = 22 − 4 · 2 · 3 = −20. Conversely, if

p = 2x 2 + 2x y + 3y 2 for integers x and y, then 2 p = X 2 + 5Y 2 with X = 2x + y
and Y = y. Hence p satisfies

−1 p
= = −1.
p 5
We have shown that for a prime p = 2, 5

−1 p
p = x 2 + 5y 2 ⇐⇒ = = 1 ⇐⇒ p ≡ 1, 9 (mod 20),
p 5

−1 p
p = 2x + 2x y + 3y ⇐⇒
2 2
= = −1 ⇐⇒ p ≡ 3, 7 (mod 20).
p 5
This result provides a refinement of Example 12.7.4. Note that

p = x 2 + 5y 2 =⇒ p 2 = u 2 + 5v 2 , (u, v) = 1,
with u = x 2 − 5y 2 , v = 2x y,

and

p = 2x 2 + 2x y + 3y 2 =⇒ p 2 = u 2 + 5v 2 , (u, v) = 1,
with u = 2x 2 + 2x y − 2y 2 , v = 2x y + y 2 .

We conclude by noting that the prime 2 is represented by 2x 2 + 2x y + 3y 2 but

not by x 2 + 5y 2 and that 5 is represented by x 2 + 5y 2 but not by 2x 2 + 2x y + 3y 2 .
√
Example 12.7.6 The ideal class group of K = Q( −21) is
√
H (K ) = H (Q( −21)) = {1, [A], [B], [A][B]} Z2 × Z2 ,

where [A] is the class of the ideal A given by

√
A = 2, 1 + −21, A2 = 2, N (A) = 2,
CB609-12 CB609/Alaca & Williams August 7, 2003 17:0 Char Count= 0

12.7 Applications to Binary Quadratic Forms 339

[B] is the class of the ideal B given by

√
B = 3, −21, B 2 = 3, N (B) = 3,

and
√
[A][B] = [AB] = [5, 3 + −21], [A]2 = [B]2 = [AB]2 = 1

(see Table 5).

Let p = 2, 3, 7 be a prime such that −21
p
= 1. By the law of quadratic reci-
procity we have

−21 p p −1
=
p 3 7 p
so that

p p −1
, , = (1, 1, 1), (−1, 1, −1), (1, −1, −1), or (−1, −1, 1).
3 7 p

As −21 p
= 1 we have

p = P1 P2 ,

where P1 and P2 are distinct conjugate prime ideals in O K . Hence

[P1 ] = 1, [A], [B], or [AB].

√
If [P1 ] = 1 then P1 is a principal ideal, say, P1 = x + y −21, where x, y ∈ Z,
so that
√
p = N (P1 ) = N (x + y −21) = x 2 + 21y 2 .

If [P1 ] = [A] then [A

√ P1 ] = [A][P1 ] = [A] = 1, so that A P1 is a principal ideal,
2

say, A P1 = x + y −21, where x, y ∈ Z, so that

√
2 p = N (A P1 ) = N (x + y −21) = x 2 + 21y 2 .

Similarly, if [P1 ] = [B] we find that 3 p = x 2 + 21y 2 (x, y ∈ Z) and if [P1 ] = [AB]
then 5 p = x 2 + 21y 2 (x, y ∈ Z).
Next if p = x 2 + 21y 2 then we have
2 2
p x + 21y 2 x
= = = 1,
3 3 3
2
p x + 21y 2 x2
= = = 1,
7 7 7

−1
= 1, as p ≡ x 2 + y 2 ≡ 1 (mod 4).
p
CB609-12 CB609/Alaca & Williams August 7, 2003 17:0 Char Count= 0

340 The Ideal Class Group

Similarly,

p p −1
2 p = x + 21y =⇒
2 2
= −1, = 1, = −1,
3 7 p

p p −1
3 p = x 2 + 21y 2 =⇒ = 1, = −1, = −1,
3 7 p

p p −1
5 p = x + 21y =⇒
2 2
= −1, = −1, = 1.
3 7 p

−21
However, if p = 2, 3, 7 is a prime such that p
= −1 then p, 2 p, 3 p, 5 p =
x 2 + 21y 2 , so we have proved the following: If p = 2, 3, 7 is a prime then

p p −1
p = x 2 + 21y 2 ⇐⇒ = 1, = 1, = 1,
3 7 p

p p −1
2 p = x + 21y ⇐⇒
2 2
= −1, = 1, = −1,
3 7 p

p p −1
3 p = x + 21y ⇐⇒
2 2
= 1, = −1, = −1,
3 7 p

p p −1
5 p = x + 21y =⇒
2 2
= −1, = −1, = 1.
3 7 p

Easy calculations show that

2 p = x 2 + 21y 2 ⇐⇒ p = 2u 2 + 2uv + 11v 2 ,
3 p = x 2 + 21y 2 ⇐⇒ p = 3u 2 + 7v 2 ,
5 p = x 2 + 21y 2 ⇐⇒ p = 5u 2 + 4uv + 5v 2 .

Thus, for a prime p = 2, 3, 7, we have

p = x 2 + 21y 2 ⇐⇒ p = 1, 25, 37 (mod 84),
p = 2x 2 + 2x y + 11y 2 ⇐⇒ p = 11, 23, 71 (mod 84),
p = 3x 2 + 7y 2 ⇐⇒ p = 19, 31, 55 (mod 84),
p = 5x 2 + 4x y + 5y 2 ⇐⇒ p = 5, 17, 41 (mod 84).

We leave it to the reader to determine which forms represent the primes 2, 3, and 7.

These examples suggest a theorem of the following type: If ax 2 + bx y + cy 2 is a

form of discriminant D then there exist positive integers s, a1 , . . . , as , m (depending
on a, b, c) such that for a prime p = 2 not dividing D
p = ax 2 + bx y + cy 2 ⇐⇒ p ≡ a1 , . . . , as (mod m).
CB609-12 CB609/Alaca & Williams August 7, 2003 17:0 Char Count= 0

Exercises 341

However, such a result does not hold for every form ax 2 + bx y + cy 2 . This is proved
in [6], where it is shown that every arithmetic progression either contains no primes
of the form x 2 + 14y 2 or contains primes of both forms x 2 + 14y 2 and 2x 2 + 7y 2 ,
proving that congruences cannot be used to distinguish the representability of a
prime by x 2 + 14y 2 from that by 2x 2 + 7y 2 . By the methods used in Examples
12.7.5 and 12.7.6 we can prove that for a prime p = 2, 7

p 2
p = x + 14y or 2x + 7y ⇐⇒
2 2 2 2
= =1
7 p
⇐⇒ p ≡ 1, 9, 15, 23, 25, 39 (mod 56)
and

p 2
p = 3x + 2x y + 5y ⇐⇒
2 2
= = −1
7 p
⇐⇒ p ≡ 3, 5, 13, 19, 27, 45 (mod 56).
Muskat [5] has shown how to distinguish the representations p = x 2 + 14y 2
and p = 2x 2 + 7y 2 as follows. Let p be a prime with p ≡ 1, 9, 15, 23, 25,
39 (mod 56). Then, as in Example 12.7.1, we can show that p = u 2 + 7v 2 for some
integers u and v. If p ≡ 1 (mod 8) then u is odd and v ≡ 0 (mod 4), and replacing
u by −u if necessary we may suppose that u ≡ 1 (mod 4); if p ≡ 7 (mod 8) then
u ≡ 0 (mod 4) and v is odd, and replacing v by −v if necessary we may suppose
that v ≡ 1 (mod 4). Thus in both cases we have
2 p + u + v ≡ 3 (mod 4)
and Muskat has proved that
p = x 2 + 14y 2 ⇐⇒ 2 p + u + v ≡ 3 (mod 8),
p = 2x 2 + 7y 2 ⇐⇒ 2 p + u + v ≡ 7 (mod 8).

Exercises
√ √
1. Prove that H (Q( −6)) = {1, [2, −6]} Z2 .
√
2. Prove that h(Q( −7)) = 1.
√
3. Prove that h(Q( −11)) = 1.
√
4. Prove that H (Q( −13)) Z2 .
√
5. Prove that H (Q( −15)) Z2 .
√
6. Prove that H (Q( −17)) Z4 .
√
7. Prove that H (Q( −23)) Z3 .
√
8. Prove that H (Q( −26)) Z6 .
√
9. Prove that H (Q( −30)) Z2 × Z2 .
√
10. Prove that H (Q(√ −47)) Z5 .
11. Prove that h(Q(√6)) = 1.
12. Prove that h(Q( 10)) = 2.
CB609-12 CB609/Alaca & Williams August 7, 2003 17:0 Char Count= 0

342 The Ideal Class Group

√ √
13. Prove that H (Q( 15)) = {1, [2, 1 + 15]} Z2 .
14. Let K = Q(θ ), where
√ θ 3 − 4θ + 2 = 0. Prove that h(K ) = 1.
15. Prove that h(Q(√3)) = 1.
3

16. 3
Prove that h(Q( √ 5)) = 1.
17. Determine h(Q( 4 2)).
18. Let p be a prime = 2, 5. Prove that

p = x 2 + 10y 2 ⇐⇒ p ≡ 1, 9, 11, 19 (mod 40),

p = 2x 2 + 5y 2 ⇐⇒ p ≡ 7, 13, 23, 37 (mod 40).

19. Let p be a prime = 3, 13. Prove that

p p
p = 2x + x y + 5y ⇐⇒
2
= 2
= −1.
3 13
20. Let p be a prime = 3, 5. Prove that

p p
p = x + x y + 4y ⇐⇒
2 2
= = 1,
3 5

p p
p = 2x 2 + x y + 2y 2 ⇐⇒ = = −1.
3 5
21. Determine exactly which primes p are represented by x 2 + x y + 5y 2 .
22. Determine exactly which primes p are represented by x 2 + x y + 11y 2 .
23. Let p be a prime = 2, 31. Prove that

2 p
p = x 2 + 62y 2 , 2x 2 + 31y 2 , or 7x 2 + 2x y + 9y 2 ⇐⇒ = =1
p 31
and

2 p
p = 3x + 2x y + 21y or 11x + 4x y + 6y ⇐⇒
2 2 2
= 2
= −1.
p 31
24. Let K be a quadratic field. Let I be an ideal of O K . If I 2 is a principal ideal prove that
I is a equivalent to its conjugate ideal I .
25. Let K be an imaginary quadratic field with discriminant d < −4. Use Dirichlet’s class
number formula to prove that
1 d
h(K ) = .
d r
2− |d|
1≤r 2
2
26. Let p be a prime ≡ 3 (mod 4). Use Dirichlet’s class number formula to prove that
√
h(Q( − p)) ≡ 1 (mod 2).
√
27. Let K = Q( n), where n is a squarefree integer > 1 with n ≡ 1 (mod 4). Let m be a
positive integer dividing n. Prove that
√
√ m+ n
m, n = m,
2
in O K .
CB609-12 CB609/Alaca & Williams August 7, 2003 17:0 Char Count= 0

Biographies 343
√
28. Let p be a prime with p ≡ 3 (mod 4). It is known that h(Q( p)) is odd. Use this fact
to prove that there exist integers a and b such that
a 2 − pb2 = (−1)( p+1)/4 2.
√
[Hint: Consider the ideal 2, 1 + p.]

Suggested Reading
1. Z. I. Borevich and I. R. Shafarevich, Number Theory, Academic Press, New York and
London, 1966.
Dirichlet’s formula for the class number of a quadratic field is proved in Chapter 5.
2. L. Carlitz, A characterization of algebraic number fields with class number two, Pro-
ceedings of the American Mathematical Society 11 (1960), 391–392.
It is proved that an algebraic number field K has class number h(K ) ≤ 2 if and only if whenever
a nonzero, nonunit α ∈ O K can be written as α = uπ1 · · · πs = u π1 · · · πt with u, u units and
π1 , . . . , πs , π1 , . . . , πt are primes in O K then s = t.
3. D. A. Marcus, Number Fields, Springer-Verlag, New York, Heidelberg, Berlin, 1977.
Chapter 5 contains a proof of (12.5.1).
4. J. M. Masley and H. L. Montgomery, Cyclotomic fields with unique factorization, Journal
für die reine und angewandte Mathematik 286/287 (1976), 248–256.
The authors prove that there are precisely 29 distinct cyclotomic fields K m (m ≡ 2 (mod 4)) with
h(K m ) = 1, namely those given by
m = 3, 4, 5, 7, 8, 9, 11, 12, 13, 15, 16, 17, 19, 20, 21, 24, 25, 27, 28, 32, 33, 35, 36,
40, 44, 45, 48, 60, 84.

5. J. B. Muskat, On simultaneous representations of primes by binary quadratic forms,

Journal of Number Theory 19 (1984), 263–282.
It is shown how the representability of primes by the forms x 2 + 14y 2 and 2x 2 + 7y 2 can be
distinguished.
6. B. K. Spearman and K. S. Williams, Representing primes by binary quadratic forms,
American Mathematical Monthly 99 (1992), 423–426.
It is shown that the representability of primes by the forms x 2 + 14y 2 and 2x 2 + 7y 2 cannot be
decided by congruence considerations alone.
7. H. M. Stark, A complete determination of the complex quadratic fields of class number
one, Michigan Mathematical Journal 14 (1967), 1–27.
√
The author shows that h(Q( k)) = 1, where k is a negative squarefree integer, if and only if
k = −1, −2, −3, −7, −11, −19, −43, −67, −163.

Biographies
1. J. V. Brawley, In memoriam: Leonard Carlitz (1907–1999), Finite Fields and Applica-
tions 6 (2000), 203–206.
A brief biography of Carlitz is given.
2. F. T. Howard, In memoriam—Leonard Carlitz, Fibonacci Quarterly 38 (2000), 316.
Another brief biography of Carlitz is given.
CB609-13 CB609/Alaca & Williams August 7, 2003 17:4 Char Count= 0

13
Dirichlet’s Unit Theorem

13.1 Valuations of an Element of a Number Field

Let K be an algebraic number field of degree n ≥ 2 over Q. Let {σ1 , . . . , σn } be
the set of all monomorphisms : K → C. If σi (K ) ⊆ R we say that σi is a real
embedding; otherwise σi is said to be a complex embedding. As usual ᾱ denotes
the complex conjugate of α ∈ C. We define for all α ∈ K

σ̄i (α) = σi (α).

Since complex conjugation is an automorphism of C, σ̄i is a monomorphism : K →

C. Hence σ̄i = σ j for some j. Now σi = σ̄i if and only if σi is real, and σ̄¯i =
σi so that complex monomorphisms occur as conjugate pairs. We enumerate the
monomorphisms in such a way that σ1 , . . . , σr are real, σr +1 , . . . , σr +s are complex,
and σr +s+1 = σr +1 , . . . , σn = σr +2s = σr +s . The conjugate fields of K are K (i) =
σi (K ), i = 1, 2, . . . , n. The r conjugate fields K (1) , . . . , K (r ) are real and the n − r
fields K (r +1) , . . . , K (n) are nonreal with K (r +s+1) = K (r +1) , . . . , K (n) = K (r +2s) =
K (r +s) . We note that

n = r + 2s (13.1.1)

and
1 n
r + s ≥ (r + 2s) = ≥ 1.
2 2
If s = 0 then all the conjugate fields of K are real and K is said to be a totally
real field. If r = 0 then K and all its conjugate fields are nonreal and K is said to be
a totally complex or totally imaginary field. If K is a normal field then K is either
totally real or totally complex, since all the conjugate fields of K coincide.

Example 13.1.1 The cubic polynomial x 3 − 6x + 2 ∈ Z[x] is 2-Eisenstein and has

discriminant −4(−6)3 − 27(2)2 = 756 > 0. Thus it is irreducible and has three real

344
CB609-13 CB609/Alaca & Williams August 7, 2003 17:4 Char Count= 0

13.1 Valuations of an Element of a Number Field 345

roots. Thus the cubic field

K = Q(θ), where θ 3 − 6θ + 2 = 0,

is a totally real field.

√
Example
√ 13.1.2
√ The√field Q( √2 + i) is totally
√ complex as the conjugates of
2 + i are 2 + i, 2 − i, − 2 + i, − 2 − i.

We next define the valuations of an element of an algebraic number field.

Definition 13.1.1 (Valuations of a field element) For a ∈ K we define

βi (a) = |σi (a)|, i = 1, 2, . . . , r + s.

The r + s quantities βi (a) (i = 1, 2, . . . , r + s) are called the valuations of a.

Clearly σi (a) ∈ K (i) and [K (i) : Q] = [K : Q] = n, so that σi (a) is an algebraic

number of degree at most n. Thus σi (a) is also an algebraic number of degree at
most n. Hence σi (a)σi (a) is an algebraic number of degree at most n 2 . This proves
that each valuation

βi (a) = |σi (a)| = σi (a)σi (a)

is a nonnegative real algebraic number of degree at most 2n 2 .

√
Example 13.1.3 Let K = Q( 3 2). Here n = 3, r = 1, s = 1. The three monomor-
phisms : K −→ C are given by
√ √ √ √
σ1 (a + b 2 + c( 2)2 ) = a + b 2 + c( 2)2 ,
3 3 3 3

√ √ √ √
σ2 (a + b 2 + c( 2)2 ) = a + bω 2 + cω2 ( 2)2 ,
3 3 3 3

√ √ √ √
σ3 (a + b 2 + c( 2)2 ) = a + bω2 2 + cω( 2)2 ,
3 3 3 3

√
where ω = e2πi/3 = (−1 + i 3)/2 is a complex cube root of unity, so that σ3 = σ2 .
Then
√ √ √ √ √ √ √ √
β1 ( 2 + ( 2)2 ) = |σ1 ( 2 + ( 2)2 )| = | 2 + ( 2)2 | = 2 + ( 2)2
3 3 3 3 3 3 3 3
CB609-13 CB609/Alaca & Williams August 7, 2003 17:4 Char Count= 0

346 Dirichlet’s Unit Theorem

and
√ √ √ √ √ √ 2
β2 ( 2 + ( 2)2 ) = |σ + = |ω + ω 2) |
3 3 3 3 2 3 2 3
( 2 ( 2) )| 2 (
2
√
−1 + i √3 √ −1 − i 3 √

= 2+
3
( 2)2
3

2 2
√
−1 √ √ 3 √ √

= ( 2 + ( 2)2 ) + i ( 2 − ( 2)2 )
3 3 3 3

2 2
√ √ √ √
( 3 2 + ( 3 2)2 )2 + 3( 3 2 − ( 3 2)2 )2
=
4
√ √
= −2 + 2 2 + ( 2)2 .
3 3

In the next section we develop the properties of the valuations of an element of

an algebraic number field K . Using these properties we prove later in the chapter
the famous theorem of Dirichlet concerning the units of O K .

Theorem 13.1.1 (Dirichlet’s unit theorem) Let K be an algebraic number field

of degree n. Let r be the number of real conjugate fields of K and 2s the number of
complex conjugate fields of K so that r and s satisfy (13.1.1). Then O K contains
r + s − 1 units 1 , . . . , r +s−1 such that each unit of O K can be expressed uniquely
nr +s−1
in the form ρ1n 1 · · · r +s−1 , where ρ is a root of unity in O K and n 1 , . . . , n r +s−1 are
integers.

13.2 Properties of Valuations

In this section we develop the properties of valuations that we shall need to prove
Dirichlet’s unit theorem. We fix once and for all an integral basis {ω1 , . . . , ωn } for
K . If a ∈ O K the coordinates of a are the uniquely determined rational integers
c1 , . . . , cn given by

a = c 1 ω1 + · · · + c n ωn .

We set

M = max |σi (ω j )| (13.2.1)

1≤i, j≤n

and

D = det(σi (ω j )). (13.2.2)

As {ω1 , . . . , ωn } is an integral basis for K , we have

D 2 = d(K )
CB609-13 CB609/Alaca & Williams August 7, 2003 17:4 Char Count= 0

13.2 Properties of Valuations 347

so that
|D| = |d(K )|1/2 (13.2.3)
and
D = 0. (13.2.4)

Lemma 13.2.1 If m ∈ Z then

βi (m) = |m|, i = 1, 2, . . . , r + s.

Proof: For i = 1, 2, . . . , r + s, each σi : K → C is a monomorphism so that

σi (a) = a for all a ∈ Q. Hence for m ∈ Z we have
βi (m) = |σi (m)| = |m|, i = 1, 2, . . . , r + s.

Lemma 13.2.2 If a, b ∈ K then

βi (a)βi (b) = βi (ab), i = 1, 2, . . . , r + s.

Proof: For i = 1, 2, . . . , r + s and a, b ∈ K we have

βi (ab) = |σi (ab)| = |σi (a)σi (b)| = |σi (a)||σi (b)| = βi (a)βi (b).

Lemma 13.2.3 If a ∈ O K is such that its coordinates ci (i = 1, 2, . . . , n) satisfy

|ci | ≤ C then
βi (a) ≤ nC M, i = 1, 2, . . . , r + s.

Proof: We have
a = c 1 ω1 + · · · + c n ωn
so that for i = 1, 2, . . . , n
σi (a) = c1 σi (ω1 ) + · · · + cn σi (ωn )
and thus for i = 1, 2, . . . , r + s we have
βi (a) = |σi (a)| = |c1 σi (ω1 ) + · · · + cn σi (ωn )|
≤ |c1 ||σi (ω1 )| + · · · + |cn ||σi (ωn )|
≤ CM + ··· + CM
= nC M.

Lemma 13.2.3 tells us that the integers of K with bounded coordinates have
bounded valuations.
CB609-13 CB609/Alaca & Williams August 7, 2003 17:4 Char Count= 0

348 Dirichlet’s Unit Theorem

Lemma 13.2.4 If a ∈ O K is such that

βi (a) ≤ L , i = 1, 2, . . . , r + s,
then the coordinates ci (i = 1, 2, . . . , n) of a satisfy
n!L M n−1
|ci | ≤ , i = 1, 2, . . . , n.
|d(K )|1/2

Proof: We have
a = c1 ω1 + · · · + cn ωn
so that
σi (a) = c1 σi (ω1 ) + · · · + cn σi (ωn ), i = 1, 2, . . . , n.
Hence, by Cramer’s rule, we have
Ni
ci = , i = 1, 2, . . . , n, (13.2.5)
D
where the determinant D is defined in (13.2.2) and the determinant Ni is
formed from D by replacing the ith column by the column consisting of
σ1 (a), σ2 (a), . . . , σn (a). Expanding Ni by its ith column we obtain
n
Ni = σk (a)(−1)k+i k ,
k=1

where each k is an (n − 1) × (n − 1) determinant whose entries ∈ {σ p (ωq ) |

p, q = 1, 2, . . . , n}. As |σ p (ωq )| ≤ M for all p, q ∈ {1, 2, . . . , n}, we see that
|k | ≤ (n − 1)!M n−1 ,
so that for i = 1, 2, . . . , n

n
|Ni | ≤ βk (a)|k | ≤ Ln!M n−1 . (13.2.6)
k=1

Finally, from (13.2.5), (13.2.6), and (13.2.3), we deduce that

|Ni | n!L M n−1
|ci | = ≤ , i = 1, 2, . . . , n.
|D| |d(K )|1/2

Lemma 13.2.4 tells us that the integers of K with bounded valuations have
bounded coordinates. The next lemma is an immediate consequence of this fact.

Lemma 13.2.5 There are only finitely many a ∈ O K , all of whose valuations
βi (a) (i = 1, 2, . . . , r + s) lie below a given limit.

Proof: Let a ∈ O K be such that βi (a) ≤ L , i = 1, 2, . . . , r + s. Then, by Lemma

13.2.4, we have
a = c1 ω1 + · · · + cn ωn ,
CB609-13 CB609/Alaca & Williams August 7, 2003 17:4 Char Count= 0

13.2 Properties of Valuations 349

where each ci ∈ Z and |ci | ≤ n!L M n−1 /|d(K )|1/2 . The number of possible choices
for each ci is

n!L M n−1
2 + 1,
|d(K )|1/2
so the number of a ∈ O K with βi (a) ≤ L (i = 1, 2, . . . , r + s) is at most

n
n!L M n−1
2 +1 .
|d(K )|1/2
Lemma 13.2.6 Let a ∈ K . Then
r +s

N (a) = βi (a)di ,
i=1

where

1, i = 1, . . . , r
di = (13.2.7)
2, i = r + 1, . . . , r + s.

Proof: We have
N (a) = |N (a)|
n
=| σi (a)|
ri=1
+s +2s
r

= σi (a) σi (a)

i=r +s+1
ri=1
+s r +s

= σi (a) σi+s (a)

i=r +1
ri=1
+s r +s

= σi (a) σ̄i (a)

i=r +1
i=1
r r +s

= σi (a) σi (a)σi (a)

i=r +1
i=1
r r +s

= σi (a) |σi (a)|2

i=1 i=r +1

r r +s

= |σi (a)| |σi (a)|2
i=1 i=r +1
r +s

= βi (a)di ,
i=1

where di is given by (13.2.7).

CB609-13 CB609/Alaca & Williams August 7, 2003 17:4 Char Count= 0

350 Dirichlet’s Unit Theorem

Lemma 13.2.7 If is a unit of O K then

r +s

βi ()di = 1,
i=1

where di is defined in (13.2.7).

Proof: This result follows immediately from Lemma 13.2.6 as N () = N (1) =

Lemma 13.2.8 Let pi , qi (i = 1, 2, . . . , r + s) be rational numbers such that

0 ≤ pi < qi , i = 1, 2, . . . , r + s.

Then there exists a ∈ K such that

pi < βi (a) < qi , i = 1, 2, . . . , r + s.

Proof: For i = 1, 2, . . . , r + s choose h i = 12 ( pi + qi ) ∈ Q so that pi < h i < qi .

For i = r + s + 1, . . . , r + 2s set

h i = h i−s .

Thus h i is defined for i = 1, 2, . . . , n and h r +s+ j = h r + j for j = 1, 2, . . . , s.

Consider the system of n linear equations in the n unknowns b1 , . . . , bn
given by

b1 σi (ω1 ) + · · · + bn σi (ωn ) = h i , i = 1, 2, . . . , n. (13.2.8)

All the constant terms in this system are real. The determinant of the coefficient
matrix of this system is D = 0 (see (13.2.2)), so that the system has a unique
solution (b1 , . . . , bn ) ∈ Cn . The first r equations in the system (13.2.8) have real
coefficients and the last n − r = 2s equations occur in complex conjugate pairs.
Hence (b1 , . . . , bn ) ∈ Rn .
Now let

qi − pi
δ = min ,
1≤i≤r +s 2Mn
so that
qi − pi
0<δ≤ , i = 1, 2, . . . , r + s. (13.2.9)
2Mn
Next choose ci ∈ Q such that

|bi − ci | < δ, i = 1, 2, . . . , n.
CB609-13 CB609/Alaca & Williams August 7, 2003 17:4 Char Count= 0

13.2 Properties of Valuations 351

Set
a = c 1 ω1 + · · · + c n ωn ∈ K .
Then
σi (a) = c1 σi (ω1 ) + · · · + cn σi (ωn ), i = 1, 2, . . . , n,
so that
σi (a) − h i = (c1 − b1 )σi (ω1 ) + · · · + (cn − bn )σi (ωn ), i = 1, 2, . . . , n.
Hence for i = 1, 2, . . . , r + s we have
|σi (a) − h i | ≤ |c1 − b1 ||σi (ω1 )| + · · · + |cn − bn ||σi (ωn )|
≤ M(|c1 − b1 | + · · · + |cn − bn |)
< Mnδ
qi − pi
≤ ,
2
so that
(qi − pi ) (qi − pi )
hi − < |σi (a)| < h i + , i = 1, 2, . . . , r + s,
2 2
that is,
pi < βi (a) < qi , i = 1, 2, . . . , r + s.

Lemma 13.2.9 Let k be a positive integer. Let A be a nonzero (integral or fractional)

ideal of O K with N (A) ≤ k n . Then A contains an element a = 0 with βi (a) ≤
n Mk (i = 1, 2, . . . , r + s).

Proof: First we consider the case when A is an integral ideal. Let

S = {b ∈ O K | b = b1 ω1 + · · · + bn ωn , b1 , . . . , bn ∈ {0, 1, 2, . . . , k}}.
Clearly card S = (k + 1)n > k n ≥ N (A), so that there exist b ∈ S, b ∈ S, b =
b such that
b ≡ b (mod A).
Set a = b − b so that a = 0, a ∈ O K , and a ≡ 0 (mod A). The latter condition
is equivalent to a ∈ A. The coordinates a1 , . . . , an of a satisfy
|ai | = |bi − bi | ≤ k, i = 1, 2, . . . , n,
where b1 , . . . , bn are the coordinates of b and b1 , . . . , bn the coordinates of b , so
that by Lemma 13.2.3
βi (a) ≤ nk M, i = 1, 2, . . . , r + s.
CB609-13 CB609/Alaca & Williams August 7, 2003 17:4 Char Count= 0

352 Dirichlet’s Unit Theorem

Now we treat the case when A is a fractional ideal of O K satisfying N (A) ≤ k n .

Let γ be a common denominator for A. Let γ1 = γ , γ2 , . . . , γn be the conjugates
of γ so that N (γ ) = γ1 · · · γn ∈ Z \ {0} and N (γ )A = γ2 · · · γn (γ A) is an integral
ideal of O K . Set m = |N (γ )|. Then m is a positive integer such that B = m A is an
integral ideal of O K with

N (B) = N (m A) = m n N (A) ≤ (mk)n .

By the previous case there exists b(= 0) ∈ B such that

βi (b) ≤ mkn M, i = 1, 2, . . . , r + s.

As b ∈ B there exists a ∈ A such that b = ma. Clearly a = 0 and

b βi (b) mkn M
βi (a) = βi = ≤ = kn M, i = 1, 2, . . . , r + s,
m βi (m) m
by Lemmas 13.2.1 and 13.2.2.

Lemma 13.2.10 There exists a fixed bound B > 0 such that for each number a ∈ K
with 12 < N (a) ≤ 1 there exists a unit ∈ O K such that

β j (a) ≤ B, j = 1, 2, . . . , r + s.

Proof: Let a ∈ K satisfy 12 < N (a) ≤ 1. Set I = a so that 12 < N (I ) ≤ 1. Let
S be the set of all such distinct principal ideals I . By Lemma 13.2.9, for each
I = a in S there exists b(= 0) ∈ I such that

βi (b) ≤ n M, i = 1, 2, . . . , r + s.

As b ∈ a we have b = qa for some q(= 0) ∈ O K . Then

N (q)
< N (q)N (I ) = N (q)N (a) = N (qa)
2
r +s

= N (qa) = N (b) = βi (b)di
i=1
r +s

≤ (n M)di = (n M)n
i=1

so that

N (q) < 2(n M)n .

Hence among the principal ideals q there are only finitely many different ones,
say,

q1 , . . . , qt .
CB609-13 CB609/Alaca & Williams August 7, 2003 17:4 Char Count= 0

13.2 Properties of Valuations 353

Thus each q = q j for some unit of O K and some j ∈ {1, 2, . . . , t}. Set
l= max βi (q −1
j ).
i = 1, . . . , r + s
j = 1, . . . , t

Then, for i = 1, . . . , r + s and j = 1, . . . , t, we have

1 = βi (1) = βi (q j q −1 −1
j ) = βi (q j )βi (q j ) ≤ lβi (q j ).

Thus
βi (a) ≤ lβi (q j )βi (a) = lβi (q j a) = lβi (qa) = lβi (b) ≤ ln M,
that is βi (a) ≤ B (i = 1, 2, . . . , r + s) with B = ln M.

Lemma 13.2.11 For each j ∈ {1, 2, . . . , r + s − 1} there exists a ∈ K with 1

2
<
N (a) ≤ 1 such that
βi (a) > B, i = 1, 2, . . . , r + s, i = j.

Proof: If r + s = 1 there is nothing to do so we may suppose that r + s ≥ 2. By

Lemma 13.2.8 there exists a ∈ K such that
B < βi (a) < 21/n B, i = 1, 2, . . . , r + s, i = j,
1 1
n−1
< β j (a) < 1 , j ∈ {1, 2, . . . , r },
2B 1−
2 n B n−1
1 1
n−2 < β j (a) < n−2 , j ∈ {r + 1, . . . , r + s − 1}.
22−n B 2
1 1 1
22 B 2
Clearly,
βi (a) > B, i = 1, 2, . . . , r + s, i = j.
If j ∈ {1, 2, . . . , r } then
1 2 s B r +2s−1 1
β1 (a) · · · βr (a)βr +1 (a)2 · · · βr +s (a)2 > B r −1 n−1
B = n−1
=
2B 2B 2
and
r −1 1 2s
β1 (a) · · · βr (a)βr +1 (a)2 · · · βr +s (a)2 < 21/n B 21/n B
1− n1
2 B n−1
r +2s−1
2 n B r +2s−1
= 1 = 1,
21− n B n−1
so that by Lemma 13.2.6
1
< N (a) < 1.
2
CB609-13 CB609/Alaca & Williams August 7, 2003 17:4 Char Count= 0

354 Dirichlet’s Unit Theorem

If j ∈ {r + 1, . . . , r + s − 1} then

2

2 s−1 1
β1 (a) · · · βr (a)βr +1 (a) · · · βr +s (a) > B
2 2 r
B n−2
21/2 B 2

B r +2s−2 1
= n−2
=
2B 2
and

2
r 2/n 2 s−1 1
β1 (a) · · · βr (a)βr +1 (a)2 · · · βr +s (a)2 < 21/n B 2 B
22−n B
1 1 n−2
2
r +2s−2
2 n B r +2s−2
= 2 = 1,
21− n B n−2
so that by Lemma 13.2.6
1
< N (a) < 1.
2
Lemma 13.2.12 For each j ∈ {1, 2, . . . , r + s − 1} there exists a unit j ∈ O K
such that
βi ( j ) < 1, i = 1, 2, . . . , r + s, i = j,
β j ( j ) > 1.

Proof: If r + s = 1 there is nothing to do so we may suppose that r + s ≥ 2.

By Lemma 13.2.11, for each j ∈ {1, 2, . . . , r + s − 1}, there exists a ∈ K with
1
2
N (a) ≤ 1 such that
βi (a) > B, i = 1, 2, . . . , r + s, i = j.
By Lemma 13.2.10 there exists a unit j ∈ O K such that
βi ( j a) ≤ B, i = 1, 2, . . . , r + s.
Hence by Lemma 13.2.2
βi ( j a) B
βi ( j ) = < = 1, i = 1, 2, . . . , r + s, i = j.
βi (a) B
Then, by Lemma 13.2.7, we obtain
r +s

1= βi ( j )di < β j ( j )d j
i=1

so that β j ( j ) > 1.

Definition 13.2.1 (Independent units) Let K be an algebraic number field. Let

1 , . . . , k (k ≥ 1) be units of O K . The units 1 , . . . , k are said to be independent
CB609-13 CB609/Alaca & Williams August 7, 2003 17:4 Char Count= 0

13.2 Properties of Valuations 355

if and only if
1r1 · · · krk = 1 (r1 , . . . , rk ∈ Z) =⇒ r1 = · · · = rk = 0.

Our next lemma shows that the units 1 , 2 , . . . , r +s−1 constructed in Lemma
13.2.12 are independent.

Lemma 13.2.13 The units 1 , . . . , r +s−1 of Lemma 13.2.12 are independent.

Proof: If r + s = 1 there is nothing to do so we may suppose that r + s ≥ 2. We

suppose that there exist integers ρ1 , . . . , ρr +s−1 , not all zero, such that
r +s−1
ρ
j j = 1. (13.2.10)
j=1

As
r +s−1
−ρ j
j =1
j=1

we can replace (ρ1 , . . . , ρr +s−1 ) by (−ρ1 , . . . , −ρr +s−1 ), if necessary, to ensure that
at least one of ρ1 , . . . , ρr +s−1 is positive. Relabeling 1 , . . . , r +s−1 , if necessary,
we may suppose that ρ1 , . . . , ρk (k ≥ 1) are positive and ρk+1 , . . . , ρr +s−1 are
nonpositive. From the valuations β1 , . . . , βk we form the product
β = β1d1 · · · βkdk
and from the remaining valuations we form
β = βk+1
d d
k+1
· · · βr +s
r +s
.
By Lemmas 13.2.1 and 13.2.2 we have
β(1) = β (1) = 1, β(a)β(b) = β(ab), β (a)β (b) = β (ab), (13.2.11)
for all a and b in K . By Lemma 13.2.7 we obtain
β()β () = 1,
and thus
β () = β −1 () (13.2.12)
for every unit of O K . For j = 1, 2, . . . , k we have
β ( j ) = βk+1
d d
k+1
· · · βr +s
r +s
( j ) = βk+1 ( j )dk+1 · · · βr +s ( j )dr +s ,
so that
β ( j ) < 1, j = 1, 2, . . . , k,
CB609-13 CB609/Alaca & Williams August 7, 2003 17:4 Char Count= 0

356 Dirichlet’s Unit Theorem

by Lemma 13.2.12. For j = k + 1, . . . , r + s − 1 we have

β( j ) = β1d1 · · · βkdk ( j ) = β1 ( j )d1 · · · βk ( j )dk ,

so that

β( j ) < 1, j = k + 1, . . . , r + s − 1,

by Lemma 13.12.12. Next, by (13.2.10) and (13.2.11), we obtain

 
r +s−1
ρj r +s−1

1 = β (1) = β  j  = β ( j )ρ j ,
j=1 j=1

so that by (13.2.12)

k r +s−1

β ( j )ρ j β( j )−ρ j = 1.
j=1 j=k+1

However, all of the factors on the left-hand side are ≤ 1, and the first k of them are
< 1. This is the required contradiction.

For j = 1, 2, . . . , r + s − 1 we set

a j = β j ( j ), (13.2.13)

so that by Lemma 13.2.12 we have

a j > 1, j = 1, 2, . . . , r + s − 1. (13.2.14)

Lemma 13.2.14 For each unit ∈ O K with

βν () ≤ 1, ν = 1, 2, . . . , r + s − 1,

there exist integers ρ1 , . . . , ρr +s−1 such that the unit

ρ ρ
η = 1 1 · · · r +s−1
r +s−1

satisfies

1 < βν (η) ≤ aν , ν = 1, 2, . . . , r + s − 1, βr +s (η) ≤ 1.

Proof: If r + s = 1 then by Lemma 13.2.7 we have

βr +s (η)dr +s = 1

for any unit η of O K , so that

βr +s (η) = 1,
CB609-13 CB609/Alaca & Williams August 7, 2003 17:4 Char Count= 0

13.2 Properties of Valuations 357

and this is all we require in this case. Thus we may suppose that r + s ≥ 2. Let
be a fixed unit of O K satisfying

βν () ≤ 1, ν = 1, 2, . . . , r + s − 1.

We consider all units η of O K of the form

k
η = 1k1 · · · r +s−1
r +s−1

with

kν ≥ 0, ν = 1, 2, . . . , r + s − 1,

and

βν (η) ≤ aν , ν = 1, 2, . . . , r + s − 1.

We note that is such a unit in view of (13.2.14). For these units we have by Lemma
13.2.12
k
βr +s (η) = βr +s (1k1 · · · r +s−1
r +s−1
)
r +s−1

= βr +s () βr +s (i )ki
i=1
≤ βr +s (),

so that all the valuations βν (η) (ν = 1, 2, . . . , r + s) are bounded. Thus, by Lemma

13.2.5, there are only finitely many η of the type considered. Among these finitely
many η, we choose η to be such that βr +s (η) is least. For this η we must have

1 < βν (η), ν = 1, 2, . . . , r + s − 1.

Otherwise, for some ν0 ∈ {1, 2, . . . , r + s − 1} we have

1 ≥ βν0 (η).

Then, for ν = 1, 2, . . . , r + s − 1, ν = ν0 , we have

βν (ν0 η) = βν (ν0 )βν (η) < βν (η) ≤ aν ;

for ν = ν0 we have

βν0 (ν0 η) = βν0 (ν0 )βν0 (η) ≤ βν0 (ν0 ) = aν0 ;

and for ν = r + s we have

βr +s (ν0 η) = βr +s (ν0 )βr +s (η) < βr +s (η),

contradicting the minimality of η.

CB609-13 CB609/Alaca & Williams August 7, 2003 17:4 Char Count= 0

358 Dirichlet’s Unit Theorem

Lemma 13.2.15 There exists a unit

σ
0 = 1σ1 · · · r +s−1
r +s−1
∈ OK
with
1 < βν (0 ), ν = 1, 2, . . . , r + s − 1.

Proof: This is the special case = 1 of Lemma 13.2.14.

Lemma 13.2.16 For each unit ∈ O K there exist integers τ1 , . . . , τr +s−1 such that
the unit
τ
η = 1τ1 · · · r +s−1
r +s−1

satisfies
1 < βν (η) ≤ aν , ν = 1, 2, . . . , r + s − 1,
and
βr +s (η) ≤ 1.

Proof: The case r + s = 1 follows as in the proof of Lemma 13.2.14. Thus we may
suppose that r + s ≥ 2. Let be a unit of O K . Set
X= max βν ().
1≤ν≤r +s−1
σ
By Lemma 13.2.15 there exists a unit 0 = 1σ1 · · · r +s−1
r +s−1
of O K satisfying
1 < βν (0 ), ν = 1, 2, . . . , r + s − 1.
Set
Y = min βν (0 ),
1≤ν≤r +s−1

so that
Y > 1.
We may choose k ∈ N so that
Y k ≥ X.
Then
βν (0 )k ≥ βν (), ν = 1, 2, . . . , r + s − 1.
Hence the unit λ = 0−k of O K satisfies
βν ()
βν (λ) = βν (0−k ) = ≤ 1, ν = 1, 2, . . . , r + s − 1.
βν (0 )k
CB609-13 CB609/Alaca & Williams August 7, 2003 17:4 Char Count= 0

13.3 Proof of Dirichlet’s Unit Theorem 359

Thus, by Lemma 13.2.14, there exist integers ρ1 , . . . , ρr +s−1 such that the unit
ρ ρ
η = λ1 1 · · · r +s−1
r +s−1

satisfies
1 < βν (η) ≤ aν , ν = 1, 2, . . . , r + s − 1.

We observe that
ρ ρ τ
η = 0−k 1 1 · · · r +s−1
r +s−1
= 1τ1 · · · r +s−1
r +s−1

with
τ j = ρ j − kσ j , j = 1, 2, . . . , r + s − 1.

Finally, as βν (η) > 1 for ν = 1, 2, . . . , r + s − 1, we deduce from Lemma 13.2.7

that
βr +s (η) ≤ 1.

Lemma 13.2.17 There exist finitely many units η1 , . . . , ηh of O K such that every
unit of O K is of the form
ρ ρ
= η j 1 1 · · · r +s−1
r +s−1

for some j ∈ {1, 2, . . . , h} and some ρ1 , . . . , ρr +s−1 ∈ Z.

Proof: By Lemma 13.2.16 each unit of O K can be expressed in the form

−τ
= η1−τ1 · · · r +s−1
r +s−1

for some integers τ1 , . . . , τr +s−1 and some unit η ∈ O K satisfying

βν (η) ≤ aν , ν = 1, 2, . . . , r + s − 1,
βr +s (η) ≤ 1.

Hence, by Lemma 13.2.5, there are only finitely many such η, say η1 , . . . , ηh . Thus
ρ ρ
= η j 1 1 · · · r +s−1
r +s−1

for some j ∈ {1, 2, . . . , h} and some integers ρ1 , . . . , ρr +s−1 .

We are now in a position to complete the proof of Dirichlet’s unit theorem in the
next section.

13.3 Proof of Dirichlet’s Unit Theorem

By Lemma 13.2.17 the unit group U (O K ) is generated by the units

1 , . . . , r +s−1 , η1 , . . . , ηh ,
CB609-13 CB609/Alaca & Williams August 7, 2003 17:4 Char Count= 0

360 Dirichlet’s Unit Theorem

that is,

U (O K ) = 1 , . . . , r +s−1 , η1 , . . . , ηh .

Let H be the subgroup of U (O K ) given by

H = 1 , . . . , r +s−1 .

By Lemma 13.2.17 there are h distinct cosets of H in U (O K ), so that the factor

group U (O K )/H has order h. Hence for ∈ U (O K ) we have

( H )h = H,

so that

h ∈ H.

Thus for each ∈ U (O K ) there exist a1 , . . . , ar +s−1 ∈ Z such that

a
h = 1a1 · · · r +s−1
r +s−1
.

Let λ1 , . . . , λm be m units of O K , where m ≥ r + s. By the previous observation

there exist integers a11 , . . . , a1 r +s−1 , . . . , am1 , . . . , am r +s−1 such that
a
λ1h = 1a11 · · · r +s−1
1 r +s−1
,
···
am r +s−1
λm = 1am1 · · · r +s−1
h
.

Consider the homogeneous system of r + s − 1 integral linear equations in the m

unknowns x1 , . . . , xm :

a11 x1 + · · · + am1 xm = 0,
···
a1 r +s−1 x1 + · · · + am r +s−1 xm = 0.

As m > r + s − 1 this system has a solution (x1 , . . . , xm ) ∈ Qm with

(x1 , . . . , xm ) = (0, . . . , 0). Multiplying each xi by the least common multiple of
the denominators of x1 , . . . , xm , we may suppose that (x1 , . . . , xm ) ∈ Zm . Then
a1 r +s−1 hx1 am r +s−1 hxm
λ1hx1 · · · λm
hxm
= 1a11 · · · r +s−1 · · · 1am1 · · · r +s−1
x1 +···+am r +s−1 xm )
= 1h(a11 x1 +···+am1 xm ) · · · r +s−1
h(a1 r +s−1

= 10 · · · r0+s−1
= 1.

This proves that any m units of O K with m ≥ r + s are not independent. Therefore
there are no more than r + s − 1 independent units in O K . But by Lemma 13.2.13
the r + s − 1 units 1 , . . . , r +s−1 are independent. Hence, by the main theorem of
CB609-13 CB609/Alaca & Williams August 7, 2003 17:4 Char Count= 0

13.4 Fundamental System of Units 361

finitely generated Abelian groups, U (O K ) is the direct product of cyclic groups,

r + s − 1 of which have infinite order and the remaining ones finite order. The
elements of a cyclic group of finite order however are roots of unity. This proves
xr +s−1
that every unit of O K can be expressed in the form η1x1 · · · r +s−1 , where η is a
root of unity and x1 , . . . , xr +s−1 ∈ Z. Finally, we show that this representation is
unique. Suppose
x y y
η1x1 · · · r +s−1
r +s−1
= θ1 1 · · · r +s−1
r +s−1
,

where η, θ are roots of unity and x1 , . . . , xr +s−1 , y1 , . . . , yr +s−1 ∈ Z. Then

y −x1 −xr +s−1
ηθ −1 = 1 1
y
· · · r +s−1
r +s−1
.

As η, θ are roots of unity so is ηθ −1 and thus there is a positive integer k such that
−1 k
ηθ = 1. Hence
k(y1 −x1 ) k(y −xr +s−1 )
1 · · · r +s−1
r +s−1
= 1.

As 1 , . . . , r +s−1 are independent we deduce that

k(y1 − x1 ) = · · · = k(yr +s−1 − xr +s−1 ) = 0

so that

x1 = y1 , . . . , xr +s−1 = yr +s−1 ,

and thus η = θ.

13.4 Fundamental System of Units

Definition 13.4.1 (Fundamental system of units) Let K be an algebraic number
field of degree n with r real embeddings and 2s complex embeddings (so that
r + 2s = n). If 1 , . . . , r +s−1 are r + s − 1 units of O K such that

1 , . . . , r +s−1 are independent (13.4.1)

and

every unit of O K can be expressed in the form

ar +s−1
= η1a1 · · · r +s−1 , where η is a root of unity in K , (13.4.2)

then {1 , . . . , r +s−1 } is called a fundamental system of units of O K .

Theorem 13.1.1 guarantees that O K always possesses a fundamental system

of units for any algebraic number field K . By the final argument in the proof of
Dirichlet’s unit theorem, we see that the representation (13.4.2) of a unit in terms
of a fundamental system of units is unique.
CB609-13 CB609/Alaca & Williams August 7, 2003 17:4 Char Count= 0

362 Dirichlet’s Unit Theorem

If r + s = 1 the fundamental system of units is empty and every unit of O K is a

root of unity. Our next theorem tells us for which fields K this occurs.

Theorem 13.4.1 Let K be an algebraic number field. Then every unit in O K is a

root of unity if and only if K = Q or K is an imaginary quadratic field.

Proof: Let K be of degree n. Let r be the number of real embeddings of K and 2s

the number of complex embeddings so that r + 2s = n. By Theorem 13.1.1 every
unit in O K is a root of unity

⇐⇒ r + s = 1
⇐⇒ r = 1, s = 0 or r = 0, s = 1
⇐⇒ K = Q or K = imaginary quadratic field.

This theorem is consistent with what we already know, namely,

U (OQ ) = U (Z) = {±1}

and for m squarefree and negative (Theorem 5.4.3)


 {±1, ± i}, if m = −1,
√
U (OQ( m) ) = {±1, ± ω, ± ω2 }, if m = −3,

{±1}, otherwise,
where ω is a complex cube root of unity.
If r + s = 2, a fundamental system of units consists of exactly one unit.

Definition 13.4.2 (Fundamental unit) Let K be an algebraic number field with

r + s = 2. Then any unit ∈ O K such that {} is a fundamental system of units for
O K is called a fundamental unit.

If is a fundamental unit of O K then every unit of O K is expressible uniquely

in the form η k , where η is a root of unity in O K and k ∈ Z. Moreover, if and 1
are two fundamental units for O K , then either 1 = λ or 1 = λ −1 for some root
of unity λ in O K .
Our next theorem tells us exactly which fields possess a fundamental unit.

Theorem 13.4.2 Let K be an algebraic number field. Then K possesses a funda-

mental unit if and only if K is a real quadratic field, a cubic field with exactly one
real embedding, or a totally imaginary quartic field.

Proof: Let K be of degree n. Let r be the number of real embeddings of K and

2s the number of complex embeddings, so that r + 2s = n. By Theorem 13.1.1
CB609-13 CB609/Alaca & Williams August 7, 2003 17:4 Char Count= 0

13.5 Roots of Unity 363

U (O K ) possesses a fundamental unit

⇐⇒ r + s = 2
⇐⇒ r = 2, s = 0, n = 2
or
r = 1, s = 1, n = 3
or
r = 0, s = 2, n = 4
⇐⇒ K = real quadratic field
or
K = cubic field with one real embedding
or
K = totally imaginary quartic field.

13.5 Roots of Unity

The theorems of this section give us some information about which roots of unity
can belong to the ring of integers O K of an algebraic number field K .
We recall that if ζk is a primitive kth root of unity then
[Q(ζk ) : Q] = φ(k), (13.5.1)
where Euler’s phi function φ is defined by

φ(k) = number of integers m satisfying (13.5.2)

1 ≤ m ≤ k with (m, k) = 1.
It is shown in texts on elementary number theory (see, for example, [3, Theorem
2.19, p. 69]) that φ is multiplicative; that is,
φ(kl) = φ(k)φ(l) (13.5.3)
whenever k and l are coprime positive integers. If p is a prime there are pa − 1
positive integers less than pa (a ≥ 1) of which pa−1 − 1 are multiples of p and the
remainder coprime with p. Hence
φ( pa ) = ( pa − 1) − ( pa−1 − 1) = pa − pa−1 = pa−1 ( p − 1). (13.5.4)
Thus if k = p1a1 · · · prar is the factorization of k into powers of distinct primes
p1 , . . . , pr then by (13.5.3) and (13.5.4) we deduce that
φ(k) = p1a1 −1 · · · prar −1 ( p1 − 1) · · · ( pr − 1). (13.5.5)
Using the prime power decompositions of the positive integers up to 40 in conjunc-
tion with (13.5.5), we obtain the following table of values of φ(k), k = 1, 2, . . . , 40.
CB609-13 CB609/Alaca & Williams August 7, 2003 17:4 Char Count= 0

364 Dirichlet’s Unit Theorem

k φ(k) k φ(k) k φ(k) k φ(k)

1 1 11 10 21 12 31 30
2 1 12 4 22 10 32 16
3 2 13 12 23 22 33 20
4 2 14 6 24 8 34 16
5 4 15 8 25 20 35 24
6 2 16 8 26 12 36 12
7 6 17 16 27 18 37 36
8 4 18 6 28 12 38 18
9 6 19 18 29 28 39 24
10 4 20 8 30 8 40 16

In the next six lemmas we prove elementary results about φ(k) that will be used
in the proofs of our theorems giving information about the roots of unity in the ring
of integers of an algebraic number field (Theorems 13.5.1–13.5.4).

Lemma 13.5.1 For all positive integers n

n
φ(n) ≥ .
2

Proof: If p is an odd prime and k is a positive integer then

p − 1 ≥ p 1/2 , if k = 1,
φ( p ) =
k
p k−1 ( p − 1) > p k−1 ≥ p k/2 , if k ≥ 2,

so that

φ( p k ) ≥ pk .

Hence, if N is an odd positive integer, as φ is multiplicative we have

√
φ(N ) ≥ N.

Let n be a positive integer. Set n = 2α N , where α is a nonnegative integer and N

is an odd positive integer. If α = 0 or 1 then

√ n
φ(n) = φ(N ) ≥ N≥
2
CB609-13 CB609/Alaca & Williams August 7, 2003 17:4 Char Count= 0

13.5 Roots of Unity 365

and if α ≥ 2
√ √
φ(n) = 2α−1 φ(N ) ≥ 2α/2 φ(N ) ≥ 2α/2 N = n
so that

n
φ(n) ≥
2
for all positive integers n.

Lemma 13.5.2 Let n be a positive integer. If φ(k) ≤ n then k ≤ 2n 2 .

Proof: By Lemma 13.5.1 we have

k
≤ φ(k) ≤ n
2
so that
k ≤ 2n 2 .

Lemma 13.5.3 φ(k) = 1 if and only if k = 1, 2.

Proof: By Lemma 13.5.2

φ(k) = 1 =⇒ k ≤ 2
and since φ(1) = φ(2) = 1 the result follows.

Lemma 13.5.4 φ(k) = 2 if and only if k = 3, 4, 6.

Proof: By Lemma 13.5.2

φ(k) = 2 =⇒ k ≤ 8
and since φ(1) = 1, φ(2) = 1, φ(3) = 2, φ(4) = 2, φ(5) = 4, φ(6) = 2, φ(7) =
6, and φ(8) = 4 the result follows.

Lemma 13.5.5 φ(k) = 4 if and only if k = 5, 8, 10, 12.

Proof: By Lemma 13.5.2

φ(k) = 4 =⇒ k ≤ 32
and the result follows by appealing to the table of values of the Euler phi function
preceding Lemma 13.5.1.
CB609-13 CB609/Alaca & Williams August 7, 2003 17:4 Char Count= 0

366 Dirichlet’s Unit Theorem

Lemma 13.5.6 If n ≥ 3 then φ(n) is even.

Proof: If n ≥ 3 then either there exists an odd prime p dividing n or n = 2α with

α ≥ 2. In the former case n = p β n 1 , where β ≥ 1 and n 1 is not divisible by p, so
that
φ(n) = p β−1 ( p − 1)φ(n 1 ) ≡ 0 (mod 2),

as 2 | p − 1. In the latter case

φ(n) = 2α−1 ≡ 0 (mod 2),

as α ≥ 2.

We now use Lemma 13.5.2 to show that the ring of integers of an algebraic
number field can only contain finitely many roots of unity.

Theorem 13.5.1 Let K be an algebraic number field. Then O K contains only

finitely many roots of unity.

Proof: Let [K : Q] = n. Let ζk be a primitive kth root of unity in O K . Then ζk ∈ K

so that
Q(ζk ) ⊆ K ,

and thus
[Q(ζk ) : Q] ≤ [K : Q],

that is,
φ(k) ≤ n.

Hence, by Lemma 13.5.2, we have

k ∈ {1, 2, . . . , 2n 2 },

proving that there are only finitely many roots of unity in O K .

If K has odd degree then we can say exactly which roots of unity are in O K .

Theorem 13.5.2 Let K be an algebraic number field of odd degree n. Then the
only roots of unity in O K are ±1.

Proof: Let ζk be a primitive kth root of unity in O K . Then Q(ζk ) ⊆ K and so

[Q(ζk ) : Q] | [K : Q], that is, φ(k) | n. But n is odd so that φ(k) is odd. By Lemma
13.5.6, we must have k ≤ 2, that is, k = 1, 2. Clearly ζ1 = 1 and ζ2 = −1 belong
in O K .

Taking n = 3 in Theorem 13.5.2 we obtain immediately the following result.

CB609-13 CB609/Alaca & Williams August 7, 2003 17:4 Char Count= 0

13.5 Roots of Unity 367

Theorem 13.5.3 The only roots of unity in the ring of integers of a cubic field are
±1.

The situation is much more complicated when n is even. We just determine the
roots of a unity in the ring of integers of a quartic field.

Theorem 13.5.4 Let K be a quartic field. Then the only possible roots of
unity = ±1 in O K are

ζ3 , ζ4 , ζ5 , ζ6 , ζ8 , ζ10 , ζ12 ,

and their powers. Moreover,

√
ζ3 ∈ O K ⇐⇒ K ⊇ Q( −3),
√
ζ4 ∈ O K ⇐⇒ K ⊇ Q( −1),

√
ζ5 ∈ O K ⇐⇒ K = Q( −10 − 2 5),
√
ζ6 ∈ O K ⇐⇒ K ⊇ Q( −3),
√ √
ζ8 ∈ O K ⇐⇒ K = Q( 2, −1),

√
ζ10 ∈ O K ⇐⇒ K = Q( −10 − 2 5),
√ √
ζ12 ∈ O K ⇐⇒ K = Q( 3, −1).

Proof: Let ζk be a primitive kth root of unity in O K . Then Q(ζk ) ⊆ K and thus
[Q(ζk ) : Q] | [K : Q], that is, φ(k) | 4. Hence φ(k) = 1, 2, or 4. Thus, by Lem-
mas 13.5.3–13.5.5, we have k = 1, 2, 3, 4, 5, 6, 8, 10, or 12. Hence the only possi-
ble roots of unity in O K are ζ1 = 1, ζ2 = −1, ζ3 , ζ4 , ζ5 , ζ6 , ζ8 , ζ10 , ζ12 , and their
powers. √ √
As e2πi/3 = (−1 + i 3)/2, we have Q(ζ3 ) = Q(e2πi/3 ) = Q( −3), so that
√
ζ3 ∈ O K ⇐⇒ K ⊇ Q(ζ3 ) ⇐⇒ K ⊇ Q( −3).
√
Similarly,
√ we can show that ζ4 ∈ O K ⇐⇒ K ⊇ Q( −1) and ζ6 ∈ O K ⇐⇒ K ⊇
Q( −3).
Next, as

1 √ √
e 2πi/5
= ( 5−1+i 10 + 2 5),
4
we have

√
Q(ζ5 ) = Q(e 2πi/5
) = Q( −10 − 2 5),
CB609-13 CB609/Alaca & Williams August 7, 2003 17:4 Char Count= 0

368 Dirichlet’s Unit Theorem

so that
ζ5 ∈ O K ⇐⇒ K ⊇ Q(ζ5 )
⇐⇒ K = Q(ζ5 ) (as [K : Q] = [Q(ζ5 ) : Q] = 4)

√
⇐⇒ K = Q( −10 − 2 5).
Similarly, we can show that

√
ζ10 ∈ O K ⇐⇒ K = Q( −10 − 2 5).
Finally,
1+i
e2πi/8 = √ ,
2
so that

1+i √ √
Q(ζ8 ) = Q(e 2πi/8
)=Q √ = Q( 2, −1).
2
Thus
ζ8 ∈ O K ⇐⇒ K ⊇ Q(ζ8 )
⇐⇒ K = Q(ζ8 ) (as [K : Q] = [Q(ζ8 ) : Q] = 4)
√ √
⇐⇒ K = Q( 2, −1).
The corresponding result for ζ12 can be shown similarly.

Example 13.5.1 There are infinitely many quartic fields K such that ζ4 ∈ O K .
Let
P = {2, 3, 5, 7, 11, 13, 17, . . .}
be the set of prime numbers. It is a theorem going back to Euclid that P is an infinite
set. For p ∈ P let
√ √
K p = Q( −1, p).
An easy calculation shows that
[K p : Q] = 4 for all p ∈ P.
Moreover, the only quadratic subfields of K p ( p ∈ P) are
√ √ √
Q( −1), Q( p), Q( − p).
√
Let p, q ∈ P with p√ = q. Then Q( q) is a quadratic subfield of K q but not of
√ √ √
K p as Q( q) = Q( −1), Q( p), Q( − p). Hence K p = K q . This shows that
{K p | p ∈ P} is an infinite set of distinct quartic fields. The ring of integers of each
K p ( p ∈ P) contains ζ4 by Theorem 13.5.4.
CB609-13 CB609/Alaca & Williams August 7, 2003 17:4 Char Count= 0

13.6 Fundamental Units in Cubic Fields 369

Example 13.5.2 There are infinitely many quartic fields K such that the only roots
of unity in their rings of integers O K are ±1. Let

P3,4 = {3, 7, 11, 19, 23, 31, . . .}

be the set of prime numbers ≡ 3 (mod 4). For p ∈ P3,4 let

√
θ p = 1 + p, K p = Q(θ p ).

Clearly θ p is a root of the polynomial

f p (x) = x 4 − 2x 2 + (1 − p) ∈ Z[x].

As 2 || 1 − p for p ∈ P3,4 , f p (x) is 2-Eisenstein and thus irreducible in Z[x]. Hence

f p (x) is the minimal polynomial of θ p over Q and so

[K p : Q] = 4.
√
If q ∈ P3,4 is such that q = p then an easy calculation shows that q ∈ / K p . But
√
q ∈ K q so that K p = K q . Hence, as there are infinitely many primes p ≡ 3
(mod 4), {K p | p ∈ P3,4 } is an infinite set of distinct quartic fields. As K p ⊆ R,
none of the roots of unity ζk , k ∈ {3, 4, 5, 6, 8, 10, 12}, belongs to O K p . Thus, by
Theorem 13.5.4, the only roots of unity in O K p are ±1.

13.6 Fundamental Units in Cubic Fields

Let K be a cubic field with exactly one real embedding. By Theorem 13.4.2 we
know that K possesses a fundamental unit η. Suppose further that K is a real field.
Then η ∈ R. By Theorem 13.5.3 the only roots of unity in K are ±1. Hence the only
fundamental units are ±η and ±η−1 . Exactly one of these four units is greater than
√
1. Thus K has √ a unique fundamental unit η > 1. We determine η for K = Q( 3 2)
and K = Q( 3 3). The main tool is Theorem 13.6.3, which gives a lower bound for
the fundamental unit in terms of the discriminant of the field K .
We first prove two elementary inequalities needed in the proof of Theorem 13.6.3.

Lemma 13.6.1 For all x ∈ R and all θ ∈ R

sin2 θ (x − 2 cos θ)2 < x 2 + 4.

Proof: For all θ ∈ R we have

1 − sin2 θ cos2 θ − sin4 θ = 1 − sin2 θ(cos2 θ + sin2 θ)

= 1 − sin2 θ = cos2 θ ≥ 0
CB609-13 CB609/Alaca & Williams August 7, 2003 17:4 Char Count= 0

370 Dirichlet’s Unit Theorem

with equality if and only if θ = (2k + 1)π/2, k ∈ Z. Thus, for all x ∈ R and all
θ ∈ R, we have

(x cos θ + 2 sin2 θ)2 + 4(1 − sin2 θ cos2 θ − sin4 θ) > 0 (13.6.1)

(2k + 1)π (2k + 1)π

x cos + 2 sin 2
= 2.
2 2
Expanding the square in (13.6.1), we obtain

x 2 cos2 θ + 4x sin2 θ cos θ + 4 − 4 sin2 θ cos2 θ > 0,

so that

−4x sin2 θ cos θ + 4 sin2 θ cos2 θ < x 2 cos2 θ + 4.

Thus

sin2 θ (x − 2 cos θ)2 = x 2 sin2 θ − 4x sin2 θ cos θ + 4 sin2 θ cos2 θ

< x 2 sin2 θ + x 2 cos2 θ + 4 = x 2 + 4.

Lemma 13.6.2 For x ≥ 33

2
x x 15
−3 −1> − .
8 8 4

Proof: We have
x 2 x

15 2 3 x 27

3 33 27

9
−3 − − = − ≥ − = > 1,
8 8 4 4 4 4 4 4 4 8
so that
x 2

x 15 2
−3 −1> −
8 8 4
and thus
2
x x 15
−3 −1> − .
8 8 4

Theorem 13.6.1 Let K be a real cubic field with two complex embeddings. Let
η > 1 be the fundamental unit of O K . If |d(K )| ≥ 33 then
|d(K )| − 27
η3 > .
4
CB609-13 CB609/Alaca & Williams August 7, 2003 17:4 Char Count= 0

13.6 Fundamental Units in Cubic Fields 371

Proof: Let η, ρeiθ , ρe−iθ be the conjugates of η, where ρ ∈ R+ . Then, as η is a

unit, we have N (η) = ±1 so that (recalling Definition 9.2.1)
ηρeiθ ρe−iθ = ±1,
that is,
ηρ 2 = ±1.
As η > 0 and ρ 2 > 0 we must have ηρ 2 = 1 so that
η = ρ −2 .
Next we determine D(η). We have
2
1 η η2
D(η) = D(1, η, η ) = 1
2
ρeiθ ρ 2 e2iθ .
1 ρe−iθ ρ 2 e−2iθ
Next
1 η η2 1 ρ −2 ρ −4
1 ρeiθ ρ e = 0 ρe − ρ −2
2 2iθ iθ
ρ e − ρ −4
2 2iθ

1 ρe−iθ ρ 2 e−2iθ 0 ρe−iθ − ρ −2 ρ 2 e−2iθ − ρ −4

= (ρeiθ − ρ −2 )(ρ 2 e−2iθ − ρ −4 ) − (ρe−iθ − ρ −2 )(ρ 2 e2iθ − ρ −4 )
= ρ 3 e−iθ − e−2iθ − ρ −3 eiθ + ρ −6 − ρ 3 eiθ + e2iθ + ρ −3 e−iθ − ρ −6
= −ρ 3 (eiθ − e−iθ ) + (e2iθ − e−2iθ ) − ρ −3 (eiθ − e−iθ )
= −ρ 3 2i sin θ + 2i sin 2θ − ρ −3 2i sin θ
= −2i sin θ(ρ 3 + ρ −3 − 2 cos θ),
so that
D(η) = −4 sin2 θ(ρ 3 + ρ −3 − 2 cos θ)2 .
Hence, by Lemma 13.6.1, we obtain
|D(η)| = 4 sin2 θ(ρ 3 + ρ −3 − 2 cos θ)2
< 4((ρ 3 + ρ −3 )2 + 4)
= 4(ρ 6 + ρ −6 + 6)
= 4(η3 + η−3 + 6).
Now, as K is a cubic field, by Theorem 7.1.16 we have
|d(K )| ≤ |D(α, β, γ )|
for any α, β, γ ∈ O K with D(α, β, γ ) = 0, so that in particular we have
|d(K )| ≤ |D(1, η, η2 )| = |D(η)|.
CB609-13 CB609/Alaca & Williams August 7, 2003 17:4 Char Count= 0

372 Dirichlet’s Unit Theorem

Hence
|d(K )| < 4(η3 + η−3 + 6).
Thus
|d(K )|
η3 + η−3 > −6
4
and so

|d(K )|
η −
6
− 6 η3 + 1 > 0.
4
Completing the square, we obtain, as |d(K )| ≥ 33,

2
2

|d(K )| |d(K )| |d(K )| 15 2

η −
3
−3 > −3 −1> − ,
8 8 8 4
so that

3
η − |d(K )| − 3 > |d(K )| − 15 .
8 8 4
If

|d(K )|
η − 3
−3 <0
8
then

|d(K )| |d(K )| 15
− 3 − η3 > −
8 8 4
so that
15 3
η3 < −3= ,
4 4
contradicting η > 1. Hence

|d(K )|
η − 3
−3 ≥0
8
so that

|d(K )| |d(K )| 15
η −
3
−3 > − ,
8 8 4
which gives
|d(K )| − 27
η3 > .
4
We next use Theorem
√ 13.6.1 to determine the fundamental unit > 1 of the ring
of integers of Q( 3 2).
CB609-13 CB609/Alaca & Williams August 7, 2003 17:4 Char Count= 0

13.6 Fundamental Units in Cubic Fields 373

√ √
Theorem 13.6.2 The fundamental unit > 1 of OQ( √3 2) is 1 + 3 2 + ( 3 2)2 .
√
Proof: We set α = 3
2 and K = Q(α). The ring of integers of K is
O K = Z + Zα + Zα 2 ,
and the discriminant of K is
d(K ) = −108
(see Table 1). For the field K , we have n = 3, r = 1, s = 1, so, by Theorems
13.4.2 and 13.5.3, K possesses a unique fundamental unit η > 1. We show that
η = 1 + α + α2.
Set
u = 1 + α + α2 ∈ OK .
Clearly, as 0 < α < 2 and 73 = 343 < 400 = 202 , we have
1 < u < 1 + 2 + 4 = 7 < 202/3 .
Moreover, u is a unit of O K as
1 1 −1 + α −1 + α −1 + α
= = = =
u 1+α+α 2 (1 + α + α )(−1 + α)
2 −1 + α 3 −1 + 2
= −1 + α ∈ O K .
Appealing to Theorem 13.6.1, we obtain, as |d(K )| = 108 > 33,
108 − 27 81
η3 > = > 20
4 4
so that
η > 201/3 .
Hence we have shown that
1 < u < η2 . (13.6.2)
Since η is a fundamental unit of O K , and the only roots of unity in O K are ±1, we
have by Dirichlet’s unit theorem
u = ±ηk , for some k ∈ Z. (13.6.3)
As η > 1 we deduce from (13.6.2) and (13.6.3) that the plus sign holds in (13.6.3)
and k = 1; that is, u = η as asserted.

In the next theorem we determine

√ in a similar manner the fundamental unit > 1
of the ring of integers of Q( 3 3).
CB609-13 CB609/Alaca & Williams August 7, 2003 17:4 Char Count= 0

374 Dirichlet’s Unit Theorem

√ √
Theorem 13.6.3 The fundamental unit of OQ( √3 3) is 4 + 3 3 3 + 2( 3 3)2 .
√
Proof: We set α = 3
3 and K = Q(α). The ring of integers of K is
O K = Z + Zα + Zα 2
and the discriminant of K is
d(K ) = −243
(see Table 1). For the field K we have n = 3, r = 1, s = 1, so, by Theorems 13.4.2
and 13.5.3, O K possesses a unique fundamental unit η > 1. We show that
η = 4 + 3α + 2α 2 .
Set
u = 4 + 3α + 2α 2 ∈ O K .
As 0 < α < 3/2 we have
3 3
1 < u < 4 + 3( ) + 2( )2 = 13.
2 2
Next we show that u is a unit of O K . We seek r, s, t ∈ Z such that
(4 + 3α + 2α 2 )(r + sα + tα 2 ) = 1.
Multiplying out the left-hand side and making use of the relations α 3 = 3, α 4 = 3α,
we obtain
(4r + 6s + 9t) + (3r + 4s + 6t)α + (2r + 3s + 4t)α 2 = 1,
so that
4r + 6s + 9t = 1,
3r + 4s + 6t = 0,
2r + 3s + 4t = 0.
Solving this system of linear equations, we obtain
r = −2, s = 0, t = 1.
Hence
1
= −2 + α 2 ∈ OK ,
4 + 3α + 2α 2
so that 4 + 3α + 2α 2 ∈ U (O K ). By Theorem 13.6.1 we have

243 − 27
η3 > = 54,
4
CB609-13 CB609/Alaca & Williams August 7, 2003 17:4 Char Count= 0

13.6 Fundamental Units in Cubic Fields 375

Table 11. Fundamental unit (> 1) of

√
Q( 3 m) for a few values of m ∈ N
√
m Fundamental unit > 1 of Q( 3 m)
√ √ 2
2 1 + 3√2 + ( 3 2)
√ 2
3 4 + 3 3 3√+ 2( 3 3)√
5 41 + 24 3√5 + 14( 3√5)2
6 109 +√60 3 6√ + 33( 3 6)2
7 4 + 2 3 7√+ ( 3 7)2 √
11 89 + 40√ 3
11 + 18(√3 11)2
12 55 + 24 12 + 21
3
2
( 3 12)2

so that

η6 > 542 = 2916 > 2744 = 143

and thus

η2 > 14.

Hence we have shown that

1 < u < η2 .

Then, exactly as in the proof of Theorem 13.6.2, we deduce that u = η.

√
Table 11 gives the fundamental unit > 1 for a few pure cubic fields Q( 3 m), m ∈
N.
Table 12 gives the fundamental unit of the first thirty cubic fields K with exactly
one real embedding arranged in order of increasing |d(K )|.

Table 13 gives a fundamental system of units {1 , 2 } of O K for the first thirty

cubic fields K = Q(θ ) having three real embeddings arranged in order of increasing
d(K ).
In Chapter 14 we make use of our knowledge of the units of O K , where K = Q(θ),
θ 3 − 4θ + 2 = 0, to determine all the solutions in integers of the equation

y(y + 1) = x(x + 1)(x + 2).

In the next example we determine a fundamental system of units for O K .

Example 13.6.1 The polynomial x 3 − 4x + 2 ∈ Z[x] is 2-Eisenstein and so is

irreducible. Hence K = Q(θ), where θ 3 − 4θ + 2 = 0, is a cubic field. The
CB609-13 CB609/Alaca & Williams August 7, 2003 17:4 Char Count= 0

376 Dirichlet’s Unit Theorem

Table 12. Fundamental unit of cubic fields K with

exactly one real embedding and −268 ≤ d(K ) < 0

d(K ) K = Q(θ ) Fundamental unit

−23 x3 + x2 − 1 θ + θ2
−31 x3 − x2 − 1 θ
−44 x − x2 − x − 1
3
θ
−59 x 3 + 2x − 1 2 + θ2
−76 x 3 − 2x − 2 1+θ
−83 x3 − x2 + x − 2 1 + θ2
−87 x 3 + x 2 + 2x − 1 2 + θ + θ2
−104 x3 − x − 2 1 + θ + θ2
−107 x − x 2 + 3x − 2
3
3 + θ2
−108 x3 − 2 1 + θ + θ2
−116 x − x2 − 2
3
1 + θ + θ2
−135 x 3 + 3x − 1 3 + θ2
−139 x + x2 + x − 2
3
3 + 2θ + θ 2
−140 x 3 + 2x − 2 3 + θ + θ2
−152 x 3 − x 2 − 2x − 2 1 + θ + θ2
−172 x3 + x2 − x − 3 2 + 2θ + θ 2
−175 x 3 − x 2 + 2x − 3 2 + θ2
−199 x 3 − x 2 + 4x − 1 4 − θ + θ2
−200 x 3 + x 2 + 2x − 2 9 + 5θ + 3θ 2
−204 x3 − x2 + x − 3 4 + θ + 2θ 2
−211 x 3 − 2x − 3 2 + 2θ + θ 2
−212 x − x 2 + 4x − 2
3
15 − 2θ + 4θ 2
−216 x 3 + 3x − 2 17 + 3θ + 5θ 2
−231 x3 + x2 − 3 2 + 2θ + θ 2
−239 x3 − x − 3 2 + 2θ + θ 2
−243 x3 − 3 4 + 3θ + 2θ 2
−244 x + x 2 − 4x − 6
3
5 + 6θ + 2θ 2
−247 x3 + x − 3 2 + θ + θ2
−255 x3 − x2 − 3 2 + θ + θ2
−268 x + x 2 − 3x − 5
3
3 + 3θ + θ 2

discriminant of x 3 − 4x + 2 is

−4(−4)3 − 27(2)2 = 148 = 22 · 37,

which is positive, so that K has three real embeddings. Furthermore, we have

d(K ) = 148 or 37.

But the smallest discriminant of a cubic field with three real embeddings is 49
(see Table 13). Hence d(K ) = 37 and so d(K ) = 148. Thus K must be the third
field listed in Table 13. Hence K = Q(φ), where φ 3 + φ 2 − 3φ − 1 = 0. The
CB609-13 CB609/Alaca & Williams August 7, 2003 17:4 Char Count= 0

13.6 Fundamental Units in Cubic Fields 377

Table 13. Units of totally real cubic fields K with 0 < d(K) ≤ 1101

d(K ) K = Q(θ) 1 2

49 x 3 + x 2 − 2x − 1 −1 + θ + θ 2 2 − θ2
81 x 3 − 3x − 1 2 + θ − θ2 −θ
148 x + x 2 − 3x − 1
3
θ 2 − θ2
169 x 3 − x 2 − 4x − 1 2 + 2θ − θ 2 −θ
229 x 3 − 4x − 1 θ 2+θ
257 x 3 − 5x − 3 4 + θ − θ2 5 + θ − θ2
316 x + x 2 − 4x − 2
3
−3 + θ + θ 2 −5 + θ + θ 2
321 x 3 + x 2 − 4x − 1 −θ −1 + 2θ + θ 2
361 x 3 + x 2 − 6x − 7 4 + θ − θ2 5 − θ2
404 x 3 − x 2 − 5x − 1 −θ 1 − θ − θ2
469 x 3 + x 2 − 5x − 4 −1 − θ −1 + 2θ + θ 2
473 x 3 − 5x − 1 −θ −2 − θ
564 x + x 2 − 5x − 3
3
−2 + θ −1 − θ + θ 2
568 x 3 − x 2 − 6x − 2 −5 − θ + θ 2 −7 − 4θ + 2θ 2
621 x 3 − 6x − 3 −2 − θ 1 + 2θ
697 x − x 2 − 8x − 5
3
6 + 2θ − θ 2 7 + 2θ − θ 2
733 x 3 + x 2 − 7x − 8 1+θ −5 − 2θ
756 x 3 − 6x − 2 5 − θ2 11 + θ − 2θ 2
761 x 3 − x 2 − 6x − 1 θ 2+θ
785 x 3 + x 2 − 6x − 5 1+θ −4 + θ + θ 2
788 x 3 − x 2 − 7x − 3 2+θ −1 − 2θ
837 x 3 − 6x − 1 −θ −3 − 6θ − 2θ 2
892 x + x 2 − 8x − 10
3
3 + θ − θ2 1 + 3θ + θ 2
940 x 3 − 7x − 4 −11 − 2θ + 2θ 2 −3 + θ + θ 2
961 x + x 2 − 10x − 8
3
−1 + 2θ + 2θ 2 3 + 4θ − 2θ 2
985 x 3 + x 2 − 6x − 1 θ −2 + θ
993 x 3 + x 2 − 6x − 3 5 − θ − θ2 5 − θ2
1016 x 3 + x 2 − 6x − 2 7 − θ − θ2 −11 − θ + θ 2
1076 x 3 − 8x − 6 1+θ −7 − 3θ
1101 x 3 + x 2 − 9x − 12 5 + 2θ − θ 2 −7 − 4θ + 2θ 2

relationship between θ and φ is given by

1
φ= .
θ −1

From Table 13 we see that a fundamental system of units for O K is

{φ, 2 − φ 2 }.

In terms of θ, we deduce that

{θ − 1, 2θ 2 − 4θ + 1}
CB609-13 CB609/Alaca & Williams August 7, 2003 17:4 Char Count= 0

378 Dirichlet’s Unit Theorem

Table 14. Fundamental unit of some

√
pure quartic fields Q( 4 −m)
√
m Fundamental unit of Q(θ ), θ = 4
−m

1 θ − θ2 + θ3
2 −1 + θ 2 − θ 3
5 −2 + 2θ − θ 2
6 1 + 4θ − 4θ 2 + 2θ 3
10 −27 + 12θ − θ 2 − 3θ 3

is a fundamental system of units for O K . Finally, as

(2θ − 1)(2θ 2 − 4θ + 1) = 4θ 3 − 10θ 2 + 6θ − 1

= θ(θ 3 − 4θ + 2) − (θ 4 − 4θ 3 + 6θ 2 − 4θ + 1)
= −(θ − 1)4 ,

we see that

{θ − 1, 2θ − 1}

is a fundamental system of units for O K .

√
Table 14 gives a fundamental unit for a few pure quartic fields Q( 4 −m), m ∈ N.
Such fields are totally imaginary quartic fields.

13.7 Regulator
Let {1 , . . . , r +s−1 } and {1 , . . . , r +s−1 } be any two fundamental systems of units
for the ring of integers O K of an algebraic number field K . As {1 , . . . , r +s−1 } is
a fundamental system of units, we have

j = ζ b j 1 1 j · · · r +s−1
a a
r +s−1 j
, j = 1, 2, . . . , r + s − 1, (13.7.1)

where ζ is a root of unity in K and ai j , b j ∈ Z. Similarly, as {1 , . . . , r +s−1 } is

also a fundamental system of units, we have

a1 j a
j = ρ b j 1 · · · r +s−1
r +s−1 j
, j = 1, 2, . . . , r + s − 1, (13.7.2)
CB609-13 CB609/Alaca & Williams August 7, 2003 17:4 Char Count= 0

13.7 Regulator 379

where ai j , bj ∈ Z. Hence, for j = 1, 2, . . . , r + s − 1, we have by (13.7.1) and

(13.7.2)
r +s−1 r +s−1
r +s−1 a
ak j
a k j
j = ζ b j k = ζ b j ζ bk l lk
k=1 k=1 l=1
r +s−1 r +s−1
r +s−1
ak j bk
= ζ bj+ l
alk ak j
k=1 k=1
.
l=1

By the uniqueness of the representation of units, we have

+s−1
r
1, if l = j,
alk ak j = (13.7.3)
0, if l = j.
k=1

Next we define the (r + s − 1) × (r + s − 1) matrices A and A by

A = [ai j ], A = [ai j ].

From (13.7.3) we see that

A A = Ir +s−1 .

Thus

det A · det A = det (A A) = det Ir +s−1 = 1.

As the matrices A and A have integral entries, both det A and det A are integers
so that det A = det A = ±1 and hence

|det A| = |det A | = 1. (13.7.4)

Let σk (k = 1, 2, . . . , n) be the n = [K : Q] distinct monomorphisms : K → C with

σ1 , . . . , σr real, σr +1 , . . . , σr +s complex, and σr +s+1 = σr +1 , . . . , σn = σr +2s =
σr +s . For j, k = 1, 2, . . . , r + s − 1 we have
r +s−1 r +s−1
al j

σk ( j ) = σk ζ bj
l = σk (ζ )bj
σk (l )al j
l=1 l=1

so that

r +s−1
r +s−1
r +s−1

|σk ( j )| = σk (ζ )b j σk (l )al j = |σk (ζ )|b j |σk (l )|al j = |σk (l )|al j

l=1 l=1 l=1

and thus
+s−1
r
log |σk ( j )| = al j log |σk (l )|. (13.7.5)
l=1
CB609-13 CB609/Alaca & Williams August 7, 2003 17:4 Char Count= 0

380 Dirichlet’s Unit Theorem

Let E and E denote the (r + s − 1) × (r + s − 1) matrices [log |σi ( j )|] and

[log |σi ( j )|] respectively. Then, from (12.7.5), we deduce that

E = AE,

and so

det E = det (AE) = det A · det E.

Finally,

|det E | = |det A · det E| = |det A||det E| = |det E|,

by (13.7.4).
We have shown that the nonnegative real number

|det (log |σi ( j )|)|

is independent of the choice of fundamental system of units {1 , . . . , r +s−1 } of

O K . We can therefore introduce the following concept.

Definition 13.7.1 (Regulator) Let K be an algebraic number field of degree n

over Q. Let r be the number of real embeddings of K and 2s the number of non-
real embeddings of K so that n = r + 2s. Let σi (i = 1, 2, . . . , n) be the n dis-
tinct monomorphisms : K → C with σ1 , . . . , σr real, σr +1 , . . . , σr +s complex, and
σr +s+1 = σr +1 , . . . , σr +2s = σr +s . Let {1 , . . . , r +s−1 } be any fundamental system
of units of O K . Let E denote the (r + s − 1) × (r + s − 1) matrix whose entry in
the (i, j) place is

log |σi ( j )|, i, j = 1, 2, . . . , r + s − 1.

Then the nonnegative real number

R(K ) = |det E|

is called the regulator of K .

If K is either Q or an imaginary quadratic field then r + s − 1 = 0 and the set

comprising a fundamental system of units of O K is empty. In this case we understand
R(K ) to be zero. Otherwise R(K ) > 0. We now determine the regulator of a real
quadratic field.

Theorem 13.7.1 Let K be a real quadratic field. Then

R(K ) = log η,

where η is the fundamental unit (> 1) of K .

CB609-13 CB609/Alaca & Williams August 7, 2003 17:4 Char Count= 0

13.7 Regulator 381

Proof: As K is a real quadratic field we have n = r = 2 and s = 0 so that r + s −

1 = 1. Thus a fundamental system of units of O K is {η} and

R(K ) = |det (log |η|)| = |log |η|| = |log η| = log η,

as η > 1 ensures log η > 0.

√ √
Example 13.7.1 The fundamental
√ unit of O K , where K = Q( 2), is η = 1 + 2
so that R(K ) = log(1 + 2).

Example 13.7.2 Let K be the cubic field given by

K = Q(θ), θ 3 − 4θ + 2 = 0.

The discriminant of the polynomial x 3 − 4x + 2 is positive as

−4(−4)3 − 27(2)2 = 148,

so that the three roots θ, θ , θ of x 3 − 4x + 2 = 0 are all real. Hence

n = r = 3, s = 0, r + s − 1 = 2.

Thus a fundamental system of units of O K comprises two units, and it was shown
in Example 13.6.1 that these can be taken to be θ − 1 and 2θ − 1.
We choose the roots θ, θ , θ of x 3 − 4x + 2 = 0 so that θ < θ < θ . Thus

θ −2.2143, θ 0.5391, θ 1.6751.

Then
|θ − 1| 3.2143, log |θ − 1| 1.1676,
|2θ − 1| 5.4286, log |2θ − 1| 1.6916,
|θ − 1| 0.4609, log |θ − 1| −0.7745,
|2θ − 1| 0.0782, log |2θ − 1| −2.5484,
|θ − 1| 0.6751, log |θ − 1| −0.3928,
|2θ − 1| 2.3502, log |2θ − 1| 0.8545.

Hence

log |θ − 1| log |θ − 1|
R(K ) = |det |
log |2θ − 1| log |2θ − 1|
= |log |θ − 1|log |2θ − 1| − log |θ − 1|log |2θ − 1||
|(1.1676)(−2.5484) + (0.7745)(1.6916)|
| − 2.9755 + 1.3101|
= | − 1.6654| = 1.6654.
CB609-13 CB609/Alaca & Williams August 7, 2003 17:4 Char Count= 0

382 Dirichlet’s Unit Theorem

It should be noted that we also have

log |θ − 1| log |θ − 1|
R(K ) = |det |
log |2θ − 1| log |2θ − 1|
= |log |θ − 1|log |2θ − 1| − log |θ − 1|log |2θ − 1||
|(1.1676)(0.8545) + (0.3928)(1.6916)|
|0.9977 + 0.6644|
= 1.6621
and

log|θ − 1| log|θ − 1|
R(K ) = |det |
log|2θ − 1| log|2θ − 1|
= |log|θ − 1|log|2θ − 1| − log|θ − 1|log|2θ − 1||
| − (0.7745)(0.8545) − (0.3928)(2.5484)|
| − 0.6618 − 1.0010|
= |1.6628| = 1.6628.
We close by remarking that some authors use a slightly different definition of
the regulator.

Exercises
1. Prove that there do not exist a, b, c ∈ Z such that
√ √ √ √
−2 + 2 2 + ( 2)2 = a + b 2 + c( 2)2
3 3 3 3

(see Example 13.1.3).

2. Show that if θ is a root of a monic polynomial f (x) ∈ Z[x] and n ∈ Z is such that
f (n) = ±1 then θ − n ∈ U (Q(θ )).
3. Prove that x 3 − 2x − 2 has only one real root θ and that θ satisfies
1.7 < θ < 1.8.
Use Theorem 13.6.1 to show that the fundamental unit (> 1) of OQ(θ ) is 1 + θ .
4. Prove that x 3 − x − 2 has only one real root θ and that θ satisfies
1.5 < θ < 1.6.
Use Theorem 13.6.1 to prove that 1 + θ + θ 2 is the fundamental unit (> 1) of OQ(θ ) .
5. Prove that x 3 − x 2 + x − 2 has only one real root θ and that θ satisfies
1.3 < θ < 1.4.
Use Theorem 13.6.1 √ to show√that the fundamental unit (> 1) of OQ(θ ) is 1 + θ 2 .
6. Prove that 41 + 24 5 + 14( 3 5)2 is the fundamental unit (> 1) of OQ( √3 5) .
3

√ √
7. Prove that 109 + 60 3 6 + 33( 3 6)2 is the fundamental unit (> 1) of OQ( √3 6) .
√ √
8. Prove that 4 + 2 3 7 + ( 3 7)2 is the fundamental unit (> 1) of OQ( √3 7) .
CB609-13 CB609/Alaca & Williams August 7, 2003 17:4 Char Count= 0

is a totally real field. Prove that −1 + θ + θ 2 and 2 − θ 2 are independent units of K .

10. Prove that
√ √ √
1 + 2 34 − ( 34)2 and 35 − 6( 34)2
4 4 4

√
are independent units in Q( 4 34).
11. Prove that

√ √ √ 1 √ √
1 + 2, 2 + 3, 2+ 6
2
√ √
is a fundamental system of units for O K , where K = Q( 2, 3).
12. Let K be an algebraic number field such that U (O K ) contains a nonreal root of unity.
Prove that N (α) > 0 for every α ∈ K \ {0}.
13. Let K = Q(ζm ), where ζm is a primitive mth root of unity, m ≥ 3. Determine r and s
for K .
14. Let K be a totally imaginary quartic field containing a real quadratic field k. By Theorem
13.4.2 K possesses a fundamental unit. Give conditions under which this fundamental
unit can be taken to be the fundamental unit of k.
15. Prove that
√ √
{1 + 2, 1 + ( 2)2 }
4 4

√
is a fundamental system of units for O K , where K = Q( 4 2).

Suggested Reading
1. H. Cohen, A Course in Computational Number Theory, Springer-Verlag, Berlin,
Heidelberg, New York, 1996.
This book describes 148 algorithms that are fundamental for number theoretic computations.
Algorithms 4.9.9 and 4.9.10 calculate the roots of unity in the ring of integers of an arbitrary
algebraic number field. Algorithm 6.5.8 computes a fundamental system of units.
2. C. Levesque, Systemes fondamentaux d’unites de certains composes de deux corps
quadratiques, I, Canadian Journal of Mathematics 33 (1981), 937–945.
√ √
The author determines a fundamental system of units for certain quartic fields Q( m, n), where
m and n are positive integers.
3. I. Niven, H. S. Zuckerman, and H. L. Montgomery, An Introduction to the Theory of
Numbers, fifth edition, Wiley, New York, 1991.
A proof that Euler’s phi function φ(n) is multiplicative can be found on page 69.
4. B. L. van der Waerden, Ein Logarithmenfreier Beweis des Dirichletschen Einheiten-
satzes, Abhandlungen aus dem Mathematischen Seminar der Universität Hamburg 6
(1928), 259–262.
The proof of Dirichlet’s unit theorem given in this chapter is based upon the approach in this paper.
CB609-13 CB609/Alaca & Williams August 7, 2003 17:4 Char Count= 0

384 Dirichlet’s Unit Theorem

Biographies
1. G. Frei, Bartel Leendert van der Waerden, Historia Mathematica 20 (1993), 5–11.
A brief biography of van der Waerden (1903–1996) is given.
2. G. Frei, J. Top, and L. Walling, A short biography of B. L. van der Waerden, Nieuw
Archief voor Wiskunde 12 (1994), 137–144.
A well-written biography of van der Waerden is given.
CB609-14 CB609/Alaca & Williams August 7, 2003 17:6 Char Count= 0

14
Applications to Diophantine Equations

An equation that is to be solved in integers is called a Diophantine equation in

honor of Diophantus (ca. 200–ca. 284), who proposed in his chief work Arithmetic
many problems to be solved in rational numbers or integers. Not much is known
about Diophantus. He lived in Alexandria, probably was not a Greek, and likely
did most of his work during the latter half of the third century.
In this chapter we apply algebraic number theory to solve some Diophantine
equations. We will be principally interested in the Diophantine equation
y 2 = x 3 + k,
where k is a given integer. This equation is often called Bachet’s equation, after
the French mathematician Claude Gaspard Bachet de Méziriac (1581–1638), who
showed how to find solutions of y 2 = x 3 − 2 in rationals x and y from the solution
(x, y) = (3, 5). In 1917 Axel Thue (1863–1922) showed that for any given nonzero
integer k, Bachet’s equation has at most finitely many solutions in integers x and
y. Deep estimates from transcendental number theory give bounds for the sizes
of the solutions x and y. Hence the problem of finding all solutions in integers
of Bachet’s equation for a given nonzero integer k is reduced to a finite search.
In Section 14.1 we use elementary congruence considerations to give classes of k
for which Bachet’s equation has no solutions in integers. In Section 14.2 we use
the arithmetic of quadratic fields to determine all the solutions in integers (if any)
of Bachet’s equation for certain classes of k. In particular when k = −2 we show
that (x, y) = (3, ±5) are the only solutions in integers of y 2 = x 3 − 2, a result first
stated by Fermat. In Section 14.3 we find all the solutions in integers x and y of
the equation y(y + 1) = x(x + 1)(x + 2).

14.1 Insolvability of y 2 = x 3 + k Using Congruence Considerations

In this section, using only simple congruence arguments, we give four classes of
integers k for which Bachet’s equation y 2 = x 3 + k has no solutions in integers x
and y.

385
CB609-14 CB609/Alaca & Williams August 7, 2003 17:6 Char Count= 0

386 Applications to Diophantine Equations

Theorem 14.1.1 Let M and N be integers such that

M ≡ 3 (mod 4), N ≡ 2 (mod 4),

p (prime) | N /2 =⇒ p ≡ 1 (mod 4).

Set

k = M 3 − N 2.

Then the equation y 2 = x 3 + k has no solutions in integers x and y.

Proof: Suppose that (x, y) ∈ Z2 is a solution of y 2 = x 3 + k. As k ≡ −1 (mod 4)

we have

y 2 ≡ x 3 − 1 (mod 4). (14.1.1)

Now y 2 ≡ 0 or 1 (mod 4) for every integer y, so (14.1.1) cannot be satisfied if x is

even or x ≡ 3 (mod 4). Hence we must have x ≡ 1 (mod 4). As k = M 3 − N 2 we
see that

y 2 + N 2 = x 3 + M 3 = (x + M)(x 2 − M x + M 2 ). (14.1.2)

Since x ≡ 1 (mod 4) and M ≡ 3 (mod 4) we deduce that

x 2 − M x + M 2 ≡ 3 (mod 4). (14.1.3)

Hence x 2 − M x + M 2 is odd and (14.1.3) shows that it has at least one prime factor
p ≡ 3 (mod 4). Thus y 2 ≡ −N 2 (mod p). By assumption p N . Hence
2
−1 −N 2 y
= = = 1,
p p p

contradicting p ≡ 3 (mod 4). This proves that the Diophantine equation y 2 =

x 3 + k has no solutions

The following table gives some values of k (with |k| < 100) covered by Theorem
14.1.1.

M −1 15 3 3
N 2 58 2 10
k −5 11 23 −73

Theorem 14.1.2 Let M and N be integers such that

M ≡ 2 (mod 4), N ≡ 1 (mod 2),

p (prime) | N =⇒ p ≡ 1 (mod 4).
CB609-14 CB609/Alaca & Williams August 7, 2003 17:6 Char Count= 0

14.1 Insolvability of y 2 = x 3 + k Using Congruences 387

Set
k = M 3 − N 2.
Then the equation y 2 = x 3 + k has no solutions in integers x and y.

Proof: Suppose that (x, y) ∈ Z2 is a solution of y 2 = x 3 + k. Considering the equa-

tion modulo 4, we obtain
y 2 ≡ x 3 − 1 (mod 4).
Hence, as in the proof of Theorem 14.1.1, we must have x ≡ 1 (mod 4). As k =
M 3 − N 2 we have
y 2 + N 2 = x 3 + M 3 = (x + M)(x 2 − M x + M 2 ).
Since x ≡ 1 (mod 4) and M ≡ 2 (mod 4) we obtain
x 2 − M x + M 2 ≡ 3 (mod 4).
The rest of the proof is the same as that of Theorem 14.1.1.

The following table gives some values of k (|k| < 100) for which Theorem
14.1.2 applies.

M 2 −2 2 −2 6 14 6
N 1 1 5 5 13 53 17
k 7 −9 −17 −33 47 −65 −73

Theorem 14.1.3 Let M and N be integers such that

M ≡ 4, 6 (mod 8), N ≡ 1 (mod 2),
p (prime) | N =⇒ p ≡ ±1 (mod 8).
Set
k = M 3 + 2N 2 .
Then the equation y 2 = x 3 + k has no solutions in integers x and y.

Proof: Suppose that (x, y) ∈ Z2 is a solution of y 2 = x 3 + k. As k = M 3 + 2N 2 ≡

2 (mod 4) we have
y 2 ≡ x 3 + 2 (mod 4).
Hence x
≡ 0 (mod 2), x
≡ 1 (mod 4), and so x ≡ 3 (mod 4). Next
y 2 − 2N 2 = x 3 + M 3 = (x + M)(x 2 − M x + M 2 ).
CB609-14 CB609/Alaca & Williams August 7, 2003 17:6 Char Count= 0

388 Applications to Diophantine Equations

If x ≡ 3 (mod 8) then

x 2 − M x + M 2 ≡ 1 − 3M + M 2 ≡ ±3 (mod 8).

Hence x 2 − M x + M 2 is odd and at least one of its prime factors p is ≡ ±3 (mod 8).
Thus p N , y 2 ≡ 2N 2 (mod p), and so
2 2
2 2N y
= = = 1,
p p p
contradicting p ≡ ±3 (mod 8).
If x ≡ 7 (mod 8) then

x + M ≡ 7 + M ≡ ±3 (mod 8).

Hence x + M is odd and at least one of its prime factors p is ≡ ±3 (mod 8). Thus
p N , y 2 ≡ 2N 2 (mod p), and so
2 2
2 2N y
= = = 1,
p p p
contradicting p ≡ ±3 (mod 8).
This proves the insolvability of y 2 = x 3 + k in integers x and y.

The following table gives some values of k (|k| < 100) for which Theorem 14.1.3
applies.

M −2 −4 −10 −4 4 −2
N 1 7 23 1 1 7
k −6 34 58 −62 66 90

Theorem 14.1.4 Let M and N be integers such that

M ≡ 4 (mod 8), N ≡ 1 (mod 2),

p (prime) | N =⇒ p ≡ 1, 3 (mod 8).

Set

k = M 3 − 2N 2 .

Then the equation y 2 = x 3 + k has no solutions in integers x and y.

Proof: Suppose that (x, y) ∈ Z2 is a solution of y 2 = x 3 + k. As k = M 3 − 2N 2 ≡

2 (mod 4) we have

y 2 ≡ x 3 + 2 (mod 4).
CB609-14 CB609/Alaca & Williams August 7, 2003 17:6 Char Count= 0

14.2 Solving y 2 = x 3 + k Using Algebraic Numbers 389

Hence x
≡ 0 (mod 2), x
≡ 1 (mod 4), and so x ≡ 3 (mod 4). Further, as k ≡ −2
(mod 8), we have

y 2 ≡ x 3 − 2 (mod 8)

so that x
≡ 7 (mod 8). Hence x ≡ 3 (mod 8). Next

y 2 + 2N 2 = x 3 + M 3 = (x + M)(x 2 − M x + M 2 ).

As x ≡ 3 (mod 8) and M ≡ 4 (mod 8), we see that x + M ≡ 7 (mod 8). Hence

x + M is odd and has at least one prime factor p ≡ 5 or 7 (mod 8). Thus
p N , y 2 ≡ −2N 2 (mod p), and so
2
−2 −2N 2 y
= = = 1,
p p p

contradicting p ≡ 5 or 7 (mod 8).

The following table gives some values of k (|k| < 100) for which Theorem 14.1.4
applies.

M 4 4 −4 −4 4
N 3 1 1 3 9
k 46 62 −66 −82 −98

14.2 Solving y 2 = x 3 + k Using Algebraic Numbers

In this section we make use of results from algebraic number theory to determine
all the solutions in integers x and y of y 2 = x 3 + k for certain classes of integers k.
The principal results that we use are the following two theorems.

Theorem 14.2.1 Let D be a Dedekind domain. Let A, B, C be nonzero integral

ideals of D such that A and B are coprime and

AB = C n ,

where n is a positive integer. Then there exist ideals A1 and B1 of D such that

A = An1 , B = B1n , C = A1 B1 .

Proof: As D is a Dedekind domain, every nonzero integral ideal of D can be

expressed uniquely as a product of prime ideals (Theorem 8.3.1). Thus

C = P1a1 · · · Prar ,
CB609-14 CB609/Alaca & Williams August 7, 2003 17:6 Char Count= 0

390 Applications to Diophantine Equations

where P1 , . . . , Pr are r (≥ 0) distinct prime ideals and a1 , . . . , ar are positive inte-

gers. Hence
AB = P1na1 · · · Prnar .
As A and B are coprime ideals, each prime power Pinai (i = 1, . . . , r ) divides either
A or B but not both. Hence by relabeling if necesssary we have
na
A = P1na1 · · · Psnas , B = Ps+1s+1 · · · Prnar ,
a
for some integer s with 0 ≤ s ≤ r . Set A1 = P1a1 · · · Psas and B1 = Ps+1
s+1
· · · Prar ;
then A = An1 , B = B1n , and C = A1 B1 as asserted.

As the ring of integers of an algebraic number field is a Dedekind domain (The-

orem 8.1.1), Theorem 14.2.1 applies in this case.

Theorem 14.2.2 Let K be an algebraic number field. Let h denote the class number
of K . Let A be an integral ideal of O K such that Ak is a principal ideal for some
positive integer k coprime with h. Then A is a principal ideal.

Proof: Let [A] denote the class of A. As the order of H (K ) is h, we have [A]h = I
so that Ah is a principal ideal. Since (h, k) = 1 there exist integers r and s such that
r h + sk = 1. Then, as Ak is a principal ideal, so is
r k s
A = Ar h+sk = Ah A
as asserted.

We now sketch the ideas involved in using Theorems 14.2.1 and 14.2.2 to obtain
classes of rational integers k for which we can find the solutions (if any) of the
Diophantine equation y 2 = x 3 + k.
We begin by supposing that the equation y 2 = x 3 + k has a solution in integers
x and y, so that
√ √
x 3 = (y + k)(y − k),
√ √ √
where y + k and y − k are integers of the quadratic field K = Q( k). √ We
assume that k is squarefree and that k ≡ 2 or 3 (mod 4) so that O K = Z + Z √ k.
(The latter condition avoids 2’s in the denominators of the integers of K = Q( k).)
Passing to ideals, we obtain
√ √
x3 = y + ky − k.
√ √
If the values of k are chosen so that the principal ideals y + k and y − k
are coprime, then we can deduce from Theorem 14.2.1 that
√
y + k = A3
CB609-14 CB609/Alaca & Williams August 7, 2003 17:6 Char Count= 0

14.2 Solving y 2 = x 3 + k Using Algebraic Numbers 391

for some ideal A of O K . Further, if the class number of K is not divisible by 3, we
know by Theorem 14.2.2 that A is a principal ideal, say
√
A = a + b k

for some integers a and b. Thus

√ √ √
y + k = a + b k3 = (a + b k)3

and so by Theorem 1.3.1

√ √
y+ k = (a + b k)3 ,

where is a unit of O K . Two cases arise depending on whether k is negative

or positive. If k is negative then there are only finitely many possibilities for .
Indeed if k
= −1 then = ±1 and if k = √ −1 then = ±1, ± i (Theorem 5.4.3).
Since cubes can be absorbed into (a + b k)3 , and −1 = (−1)3 , i = (−i)3 , and
− i = i 3 , the equation becomes
√ √
y + k = (a + b k)3 .
√
Equating coefficients of k, we obtain

1 = 3a 2 b + kb3 = b(3a 2 + kb2 ),

so that b = ±1. It is now an easy matter to determine the possibilities for a, and
then the solutions x, y in integers of y 2 = x 3 + k (see Theorem 14.2.3). If k is
√there are infinitely many possibilities for . Indeed = ±η , where
l
positive then
η = T + U k (> 1) is the fundamental unit of O K and l ∈ Z√(Theorem 11.5.1).
Absorbing the cubes −1 = (−1)3 and ζ 3m = (ζ m )3 into (a + b k)3 we see that we
have only to examine the three equations
√ √
y + k = (a + b k)3 ,
√ √
y + k = η(a + b k)3 ,

and
√ √
y+ k = η2 (a + b k)3 .

The first of these equations can be treated as in the case k < 0. For the other two
equations it is convenient to impose congruence conditions on k, T , and U to
ensure that they do not have any solutions. This is illustrated in Theorem 14.2.4. It
should be noted that absorbed cubes must be taken into account when seeking all
solutions of y 2 = x 3 + k.
CB609-14 CB609/Alaca & Williams August 7, 2003 17:6 Char Count= 0

392 Applications to Diophantine Equations

Theorem 14.2.3 Let k be an integer such that

k < −1,
k is squarefree,
k ≡ 2, 3 (mod 4),
√
h(Q( k))
≡ 0 (mod 3).

(a) If there exists an integer a such that

k = 1 − 3a 2
then the only solutions in integers of y 2 = x 3 + k are
x = 4a 2 − 1, y = ±(3a − 8a 3 ).
(b) If there exists an integer a such that
k = −1 − 3a 2
then the only solutions in integers of y 2 = x 3 + k are
x = 4a 2 + 1, y = ±(3a + 8a 3 ).
(c) If k
= ±1 − 3a 2 for any integer a then y 2 = x 3 + k has no solutions in integers x
and y.

Proof: We suppose that y 2 = x 3 + k has a solution in integers x and y and show

that either case (a) or case (b) holds. (We note that in case (a) k ≡ 1 (mod 3) and
in case (b) k ≡ 2 (mod 3), so that cases (a) and (b) are exclusive.)
First we show that x ≡ 1 (mod 2). As y 2 ≡ 0, 1 (mod 4) and k ≡ 2, 3 (mod 4),
we see that x 3 = y 2 − k ≡ 1, 2, 3 (mod 4). But x 3
≡ 2 (mod 4) so x ≡ 1 (mod 2).
Next we prove that (x, k) = 1. Suppose not. Then there exists a prime p such
that p | x and p | k. As k is squarefree we have p || k. Hence p || x 3 + k and so
p || y 2 , a contradiction.
From x ≡ 1 (mod 2) and (x, k) = 1 we deduce that (x, 2k) = 1 so that there are
integers l and m such that
lx + m(2k) = 1. (14.2.1)
√
Now let K = Q( k) so that K is an imaginary √ quadratic field. As
k ≡ 2, 3 (mod 4) the ring O K of integers
√ of K is
√ {u + v k | u, v ∈ Z}. We now
show that the principal ideals y + k and y − k of O K are coprime. Suppose
not. Then there exists a prime ideal P such that
√ √
P | y + k, P | y − k.
Hence
√ √
y+ k ∈ P, y − k ∈ P.
CB609-14 CB609/Alaca & Williams August 7, 2003 17:6 Char Count= 0

14.2 Solving y 2 = x 3 + k Using Algebraic Numbers 393

Thus
√ √ √
2 k = (y + k) − (y − k) ∈ P

and so
√ √
2k = k(2 k) ∈ P. (14.2.2)

Now
√ √
y + ky − k = y 2 − k = x 3 = x3

so that

P | x3 .

As P is a prime ideal, we deduce that

P | x.

Thus

x ∈ P. (14.2.3)

From (14.2.1)–(14.2.3), we see √that 1 ∈ P, contradicting

√ that P is a prime ideal.
We√have shown √ that y + k and y − k are coprime ideals of O K with
y + ky − k = x . As K is an algebraic number field, O K is a Dedekind
3

domain by Theorem 8.1.1, and thus by Theorem 14.2.1 there exists an ideal A of
O K such that
√
y + k = A3 .
√
Thus A3 is a principal ideal and, as h(Q( k))
= 0 (mod 3), by Theorem 14.2.2, A
is a principal ideal, say,
√
A = a + b k,

where a, b ∈ Z. Hence
√ √ √
y + k = a + b k3 = (a + b k)3 .

By Theorem 1.3.1 there exists a unit ∈ O K such that

√ √
y + k = (a + b k)3 . (14.2.4)

As k < −1 and k ≡ 2, 3 (mod 4) by Theorem 5.4.3 we have = ±1. Taking con-

jugates in (14.2.4), we obtain
√ √
y − k = (a − b k)3 . (14.2.5)
CB609-14 CB609/Alaca & Williams August 7, 2003 17:6 Char Count= 0

394 Applications to Diophantine Equations

Thus
√ √ √ √
x 3 = y 2 − k = (y + k)(y − k) = (a + b k)3 (a − b k)3
√ √ 3
= ((a + b k)(a − b k))3 = a 2 − kb2
so that
x = a 2 − kb2 . (14.2.6)
Adding and subtracting (14.2.4) and (14.2.5), we obtain
√ √
2y = ((a + b k)3 + (a − b k)3 )
and
√ √ √
2 k = ((a + b k)3 − (a − b k)3 ),
so that
y = (a 3 + 3kab2 ), 1 = (3a 2 b + kb3 ).
From 1 = b(3a 2 + kb2 ) we see that b = ±1, so that b = ±. If b = then
x = a 2 − k, y = (a 3 + 3ka), 1 = 3a 2 + k,
so that
k = 1 − 3a 2
and
x = 4a 2 − 1, y = ±(3a − 8a 3 ).
Clearly
2
x 3 + k = (4a 2 − 1)3 + (1 − 3a 2 ) = 64a 6 − 48a 4 + 9a 2 = 8a 3 − 3a = y 2 .
If b = − then
x = a 2 − k, y = (a 3 + 3ka), 1 = −3a 2 − k,
so that
k = −1 − 3a 2
and
x = 4a 2 + 1, y = ±(3a + 8a 3 ).
Clearly
3 2
x 3 + k = 4a 2 + 1 − 1 − 3a 2 = 64a 6 + 48a 4 + 9a 2 = 8a 3 + 3a = y 2 .
This completes the proof of the theorem
CB609-14 CB609/Alaca & Williams August 7, 2003 17:6 Char Count= 0

14.2 Solving y 2 = x 3 + k Using Algebraic Numbers 395

√ k = −2 = 1 − 3 · 1 satisfies the conditions of The-

2
Example 14.2.1 The integer
orem 14.2.3(a) as h(Q( −2)) = 1, so that the only solutions in integers of the
equation y 2 = x 3 − 2 are (x, y) = (3, ±5). This result was first stated by Fermat.

The values of k in the range −200 < k < −2 that√ satisfy the conditions of The-
√ k = −74 = 1 − 3 · 5 (h(Q( −74)) = 10) and k = −146 =
2
orem 14.2.3(a) are
1 − 3 · 72 (h(Q( −146)) = 16). Hence, by Theorem 14.2.3(a), the only solutions
in integers of y 2 = x 3 − 74 are (x, y) = (99, ±985) and the only solutions in inte-
gers to y 2 = x 3 − 146 are (x, y) = (195, ±2723).

Example 14.2.2 The smallest integer k in absolute value √

satisfying the conditions
of Theorem 14.2.3(b) is k = −13 = −1 − 3 · 22 as h(Q( −13)) = 2. Hence, by
Theorem 14.2.3(b), the only solutions in integers x and y of the equation y 2 =
x 3 − 13 are (x, y) = (17, ±70). In the range −200 < k < −1 there is only one
√ the conditions of Theorem 14.2.3(b), namely, k =
other value of k that satisfies
−193 = −1 − 3 · 82 (h(Q( −193)) = 4). The only solutions in integers of the
equation y 2 = x 3 − 193 are (x, y) = (257, ±4120).

Example√14.2.3 The integer k = −5 satisfies the conditions of Theorem 14.2.3(c)

as h(Q( −5)) = 2. Hence the equation y = x 3 − 5 has no solutions in integers.
2

We note that k = −5 was also covered by Theorem 14.1.1.

Similarly, we find that y 2 = x 3 + k is not solvable in integers x and y for

k = −6, −10, −14, −17, −21, −22.

In the next theorem we find a result similar to that of Theorem 14.2.3(c) in the
case when k is positive.

Theorem 14.2.4 Let k be an integer such that

k > 0,
k is squarefree,
k ≡ 2, 3 (mod 4),
√
h(Q( k))
≡ 0 (mod 3).
√ √
Let T + U k be the fundamental unit of K = Q( k) of norm 1. If

k ≡ 4 (mod 9), U ≡ 0 (mod 9)

k ≡ 7 (mod 9), U ≡ ±3 (mod 9)

CB609-14 CB609/Alaca & Williams August 7, 2003 17:6 Char Count= 0

396 Applications to Diophantine Equations

k ≡ 4 (mod 7), U ≡ 0 (mod 7)

then the equation y 2 = x 3 + k has no solutions in integers x and y.

Proof: Exactly as in the proof of Theorem 14.2.3 we obtain

√ √
y + k = (a + b k)3 ,

where is a unit of O K . Let η be the fundamental unit of O K so that

= ±ηl

for some l ∈√Z. As the cubes −1√= (−1)3 and η√3m = (ηm )3 can be absorbed into the
cube (a + b k)3 , we have y + k = (a + b k)3 , where = 1, η, or η2 . Further,
as η = η3 /η2 and η2 = η3 /η, we have
√ √ 1 1
y+ k = (a + b k)3 , where ∈ {1, η, } or {1, 2 , η2 }.
η η
We choose ∈ {1, η, 1/η} if η has norm 1 and ∈ {1, 1/η2 , η2 } if η has norm −1.
Thus in both cases
√ √
∈ {1, T + U k, T − U k},
√
where T + U k is√ the fundamental unit (> 1) of O K of norm 1. If = 1, equating
the coefficients of k we obtain 1 = 3a 2 b + kb3 , so that b | 1 and thus b = ±1.
√ ±1 = b = 3a b + kb = 3a + k ≥ k > 1, a contradiction. Thus = T ±
2 2 4 2
Hence
U k. Then
√ √ √
y + k = (T ± U k)(a + b k)3
√ √
= (T ± U k)((a 3 + 3kab2 ) + (3a 2 b + kb3 ) k)
= (T (a 3 + 3kab2 ) ± U k(3a 2 b + kb3 ))
√
+ (T (3a 2 b + kb3 ) ± U (a 3 + 3kab2 )) k

so that

1 = T (3a 2 b + kb3 ) ± U (a 3 + 3kab2 ). (14.2.7)

Case (i): k ≡ 4 (mod 9), U ≡ 0 (mod 9). As U ≡ 0 (mod 9), from T 2 − kU 2 =

1 we obtain T ≡ ±1 (mod 81), say

T ≡ (mod 81), = ±1.

Then from (14.2.7) modulo 9, we deduce that

1 ≡ (3a 2 b + 4b3 ) (mod 9). (14.2.8)

CB609-14 CB609/Alaca & Williams August 7, 2003 17:6 Char Count= 0

14.2 Solving y 2 = x 3 + k Using Algebraic Numbers 397

Clearly this congruence implies that b
≡ 0 (mod 3). Hence b ≡ ±1 (mod 3), say
b ≡ λ (mod 3), λ = ±1.
Thus
b3 ≡ λ (mod 9).
Then from (14.2.8) we deduce that
1 ≡ λ(3a 2 + 4) (mod 9),
so that
3a 2 + 4 ≡ λ ≡ ±1 (mod 9),
giving
3a 2 ≡ 4 or 6 (mod 9),
both of which are impossible.

Case (ii): k ≡ 7 (mod 9), U ≡ ±3 (mod 9). In this case we have U 2 ≡ 0

(mod 9). Then from T 2 − kU 2 = 1 we deduce that T 2 ≡ 1 (mod 9), so that
T ≡ (mod 9), = ±1.
Next, from (14.2.7) modulo 3, we obtain
1 ≡ b3 (mod 3),
so that
b ≡ (mod 3), b3 ≡ (mod 9).
Then from (14.2.7) modulo 9 we have
1 ≡ 3a 2 + 7 ± 3a 3 (mod 9).
Clearly this implies a
≡ 0 (mod 3), so a ≡ ±1 (mod 3), a 2 ≡ 1 (mod 3), and
a 3 ≡ a (mod 3). Hence
1 ≡ 1 ± 3a (mod 9),
giving a ≡ 0 (mod 3), a contradiction.

Case (iii): k ≡ 4 (mod 7), U ≡ 0 (mod 7). From

√ √ √
y + k = (T ± U k)(a + b k)3
we deduce that
√ √ √
y− k = (T ∓ U k)(a − b k)3 ,
CB609-14 CB609/Alaca & Williams August 7, 2003 17:6 Char Count= 0

398 Applications to Diophantine Equations

so that
√ √
x 3 = y 2 − k = (y + k)(y − k)
√ √ √ √
= (T ± U k)(a + b k)3 (T ∓ U k)(a − b k)3
3
= (T 2 − kU 2 ) a 2 − kb2
3
= a 2 − kb2
and hence
x = a 2 − kb2 .
Now
x 3 ≡ 0, 1, 6 (mod 7)
and
y 2 ≡ 0, 1, 2, 4 (mod 7),
so
y 2 − x 3 = k ≡ 4 (mod 7)
gives
y 2 ≡ 4 (mod 7), x 3 ≡ 0 (mod 7).
Thus
x ≡ 0 (mod 7)
and so
a 2 − 4b2 ≡ 0 (mod 7);
that is,
a ≡ ± 2b (mod 7).
From U ≡ 0 (mod 7) and T 2 − kU 2 = 1 we deduce that
T 2 ≡ 1 (mod 49)
so that
T ≡ ± 1 (mod 49).
Then from (14.2.7) we obtain
1 ≡ ± 2b3 (mod 7),
which is impossible.
CB609-14 CB609/Alaca & Williams August 7, 2003 17:6 Char Count= 0

14.2 Solving y 2 = x 3 + k Using Algebraic Numbers 399

This completes the proof that y 2 = x 3 + k is insolvable in integers x and y in all

three cases.

Example 14.2.4 We choose √ k = 58 so that k ≡ 2 (mod 4) and √ k≡

√ In this case h(Q( 58)) = 2 and the fundamental unit of Q( 58) is
4 (mod 9).
99 + 13 58 of norm −1. Thus the fundamental unit of norm 1 is
√ √
(99 + 13 58)2 = 19603 + 2574 58

so that

U = 2574 ≡ 0 (mod 9).

Thus, by Theorem 14.2.4, the equation y 2 = x 3 + 58 is not solvable in integers x

and y. We note that this equation is also covered by Theorem 14.1.3.

Example 14.2.5 We √ choose k = 7. Here k ≡ 3 (mod 4)√ and k ≡

7 (mod 9).√Also, h(Q( 7)) = 1 and the fundamental unit of Q( 7) of norm
1 is 8 + 3 7 so that U = 3 ≡ 3 (mod 9). Hence, by Theorem 14.2.4, the equation
y 2 = x 3 + 7 has no solutions in integers x and y. This equation is also covered by
Theorem 14.1.2.

Example 14.2.6 We √choose k = 158. Here k ≡ 2 (mod 4) and √ k≡

4 (mod 7). Also, h(Q(√158)) = 1 and the fundamental unit of Q( 158) of
norm 1 is 7743 + 616 158 so that U = 616 ≡ 0 (mod 7). Thus, by Theorem
14.2.4, the equation y 2 = x 3 + 158 has no solutions in integers x and y.
√
We conclude this section by giving an example where h(Q( k)) ≡
0 (mod 3).

Theorem 14.2.5 The equation

y 2 = x 3 − 31 (14.2.9)

has no solutions in integers x and y.

Proof: Suppose that y 2 = x 3 − 31 has a solution in integers x and y.

First we note that 31 y, for if 31 | y then 31 | x and so 312 | x 3 − y 2 = 31, a
contradiction.
Next we show that x must be even. Suppose that x is odd. If
x ≡ 1 (mod 4) then x 3 ≡ 1 (mod 4) so that y 2 ≡ 2 (mod 4), which is impos-
sible. If x ≡ 3 (mod 4) then x 2 + 3x + 9 ≡ 3 (mod 4). Also x 2 + 3x + 9 > 1.
CB609-14 CB609/Alaca & Williams August 7, 2003 17:6 Char Count= 0

400 Applications to Diophantine Equations

Hence x 2 + 3x + 9 has a prime factor p ≡ 3 (mod 4). Now

y 2 + 4 = x 3 − 27 = (x − 3)(x 2 + 3x + 9),
so that y 2 + 4 ≡ 0 (mod p), which is impossible. This proves that x is even and y
is odd. √ √
An integral basis for K = Q( −31) is 1, 1 + 2 −31 . The prime ideal factor-
ization of 2 in O K is given by
√ √
3 + −31 3 − −31
2 = 2, 2, (14.2.10)
2 2
√
(see Theorem 10.2.1). We show that 2, 3 + 2 −31 is not a principal ideal. Suppose on
√
the contrary that 2, 3 + 2 −31 is a principal ideal. Then there exist rational integers
a and b such that
√ √
3 + −31 1 + −31
2, = a + b .
2 2
Taking norms we obtain
√ √
3 + −31 1 + −31
2 = N 2, = N a + b
2 2
√
2a + b + b −31 (2a + b)2 + 31b2
= | N( ) |=
2 4
so that
(2a + b)2 + 31b2 = 8,
which is clearly impossible.
Next, appealing to (14.2.9) and (14.2.10), we deduce that
√ √ √ √
y + −31 y − −31 3 + −31 3 − −31 x 3
= 2, 2, . (14.2.11)
2 2 2 2 2
√ √
We show that the two ideals y + 2 −31 and y − 2 −31 are coprime. If not, then there
exists a prime ideal P such that
√ √
y + −31 y − −31
P| , P | .
2 2
Then
√ √
y + −31 y − −31
∈ P, ∈ P,
2 2
so
√ √
√ y + −31 y − −31
−31 = − ∈ P.
2 2
CB609-14 CB609/Alaca & Williams August 7, 2003 17:6 Char Count= 0

14.3 The Equation y(y + 1) = x(x + 1)(x + 2) 401

Thus
√
P | −31.
√
But P and −31 are both prime ideals so that
√
P = −31.

Hence
√
√ y + −31
−31 |
2

y+
√
−31 √
so that 2
∈ −31. This shows that there exist integers u and v such
that
√ √
y+ −31 √ u + v −31
= −31 .
2 2

√ u =
1 and y =√−31v,
Hence
contradicting 31 y. This proves that the ideals
y + −31 y − −31
2
and 2
are coprime. Thus, replacing y by −y if necessary,
we see from (14.2.11) that there exists an ideal A of O K such that
 √ √

 y + −31 3 + −31 3

 = 2, A ,

 2 2


 √ √
y − −31 3 − −31 3 (14.2.12)

 = 2, Ā ,

 2 2




 x = A Ā,
2
√
where Ā denotes the conjugate ideal of A. Since h(Q( −31)) = 3 the ideal A3
is principal.
√ Then, from the first equality in (14.2.12), we deduce that the ideal
3 + −31
2, 2
is principal, a contradiction. This completes the proof that the equation
y 2 = x 3 − 31 has no solutions in integers x and y.

We conclude this section by giving two short tables (Tables 15 and 16) of solutions
of y 2 = x 3 + k.

14.3 The Diophantine Equation

y(y + 1) = x(x + 1)(x + 2)
In this section we use the arithmetic of the cubic field

K = Q(θ), θ 3 − 4θ + 2 = 0, (14.3.1)
CB609-14 CB609/Alaca & Williams August 7, 2003 17:6 Char Count= 0

402 Applications to Diophantine Equations

Table 15. Solutions (x, y) ∈ Z2 of

y 2 = x 3 + k, − 20 ≤ k < 0

k Solutions (x, y) of y 2 = x 3 + k
−1 (1, 0)
−2 (3, ±5)
−3 insolvable
−4 (2, ±2), (5, ±11)
−5 insolvable
−6 insolvable
−7 (2, ±1), (32, ±181)
−8 (2, 0)
−9 insolvable
−10 insolvable
−11 (3, ±4), (15, ±58)
−12 insolvable
−13 (17, ±70)
−14 insolvable
−15 (4, ±7)
−16 insolvable
−17 insolvable
−18 (3, ±3)
−19 (7, ±18)
−20 (6, ±14)

to determine all the solutions in integers x and y of the equation

y(y + 1) = x(x + 1)(x + 2); (14.3.2)

that is, we determine all those integers that are simultaneously a product of two
consecutive integers and a product of three consecutive integers. This problem was
proposed by Edgar Emerson to Burton W. Jones (1902–1983) and was first solved
by Louis J. Mordell (1888–1972) in 1963. We follow the solution given by Mordell
in his paper [6].
We need the following facts about the field K and its ring of integers O K :

O K = Z + Zθ + Zθ 2 , (14.3.3)
O K is a unique factorization domain, (14.3.4)
a fundamental system of units of O K is{, η},
where = θ − 1 and η = 2θ − 1. (14.3.5)

Result (14.3.3) is Exercise 13 of Chapter 7. Result (14.3.4) is Exercise 14 of Chapter

12. For result (14.3.5) see Example 13.6.1. By Dirichlet’s unit theorem every unit
of O K is given by ± m ηn for integers m and n.
CB609-14 CB609/Alaca & Williams August 7, 2003 17:6 Char Count= 0

14.3 The Equation y(y + 1) = x(x + 1)(x + 2) 403

Table 16. Solutions (x, y) ∈ Z2 of

y 2 = x 3 + k, 0 < k ≤ 20

k Solutions (x, y) of y 2 = x 3 + k
1 (−1, 0), (0, ±1), (2, ±3)
2 (−1, ±1)
3 (1, ±2)
4 (0, ±2)
5 (−1, ±2)
6 insolvable
7 insolvable
8 (−2, 0), (1, ±3), (2, ±4), (46, ±312)
9 (−2, ±1), (0, ±3), (3, ±6), (6, ±15), (40, ±253)
10 (−1, ±3)
11 insolvable
12 (−2, ±2), (13, ±47)
13 insolvable
14 insolvable
15 (1, ±4), (109, ±1138)
16 (0, ±4)
17 (−2, ±3), (−1, ±4), (2, ±5), (4, ±9), (8, ±23),
(43, ±282), (52, ±375), (5234, ±378661)
18 (7, ±19)
19 (5, ±12)
20 insolvable

If we set

X = 2x + 2, Y = 2y + 1, (14.3.6)

the equation (14.3.2) becomes

2Y 2 = X 3 − 4X + 2. (14.3.7)

Clearly any solution of (14.3.7) must have X even and Y odd. We will show that
the only solutions of (14.3.7) are

(X, Y ) = (−2, ±1), (0, ±1), (2, ±1), (4, ±5), (12, ±29).

Thus all the solutions of (14.3.2) are

(x, y) = (0, 0), (0, −1), (−1, 0), (−1, −1), (−2, 0), (−2, −1), (1, 2),
(1, −3), (5, 14), (5, −15).

This proves that the only integers that are simultaneously a product of two consec-
utive integers as well as a product of three consecutive integers are 0, 6, and 210.
CB609-14 CB609/Alaca & Williams August 7, 2003 17:6 Char Count= 0

404 Applications to Diophantine Equations

All such solutions are given by

0 = (−1)(0) = (0)(1) = (−2)(−1)(0) = (−1)(0)(1) = (0)(1)(2),
6 = (2)(3) = (−3)(−2) = (1)(2)(3),
210 = (14)(15) = (−15)(−14) = (5)(6)(7).
We let θ, θ , θ ∈ C be the three roots of x 3 − 4x + 2 = 0 so that

 θ + θ + θ = 0,
θθ + θ θ + θ θ = −4,

(14.3.8)

θθ θ = −2.
We need a number of lemmas.

Lemma 14.3.1 θ is a prime in O K .

Proof: From (14.3.8) we deduce that |N (θ)| = |θθ θ | = 2, which is a rational

prime, so that θ is a prime in O K .

Lemma 14.3.2 4θ − 3 is a prime in O K .

Proof: We have by (14.3.8)

N (4θ − 3) = (4θ − 3)(4θ − 3)(4θ − 3)
= 64θ θ θ − 48(θθ + θ θ + θ θ) + 36(θ + θ + θ ) − 27
= 64(−2) − 48(−4) + 36(0) − 27
= −128 + 192 − 27
= 37,
which is a rational prime, so that 4θ − 3 is a prime in O K .

Lemma 14.3.3 2 = ρθ 3 , where ρ ∈ U (O K ).

Proof: From θ 3 − 4θ + 2 = 0 we deduce that

θ3
= 2θ − 1 ∈ O K .
2
Further,

θ3 N (θ)3 (−2)3
N = = = −1.
2 8 8
Hence
θ3
∈ U (O K ).
2
CB609-14 CB609/Alaca & Williams August 7, 2003 17:6 Char Count= 0

14.3 The Equation y(y + 1) = x(x + 1)(x + 2) 405

Thus
2
∈ U (O K )
θ3
and so
2
=ρ
θ3
for some ρ ∈ U (O K ).

Lemma 14.3.4 If (X, Y ) ∈ Z2 is a solution of (14.3.7) then

(X − θ)(X 2 + θ X + (θ 2 − 4)) = ρθ 3 Y 2 .

Proof: We have by (14.3.1), (14.3.7), and Lemma 14.3.3

(X − θ)(X 2 + θ X + (θ 2 − 4)) = X 3 − 4X − θ 3 + 4θ
= X 3 − 4X + 2
= 2Y 2
= ρθ 3 Y 2 .

Lemma 14.3.5 The only possible primes in O K dividing both X − θ and X 2 +

θ X + (θ 2 − 4) are θ and 4θ − 3.

Proof: Let π be a prime of O K dividing both X − θ and X 2 + θ X + (θ 2 − 4). Then

π divides

(X 2 + θ X + (θ 2 − 4)) − (X + 2θ)(X − θ)
3θ 3 − 4θ 3(4θ − 2) − 4θ 8θ − 6
= 3θ 2 − 4 = = =
θ θ θ
2
= (4θ − 3) = ρ(4θ − 3)θ 2 ,
θ
by (14.3.1) and Lemma 14.3.3. As ρ is a unit this shows that the only possibilities
for π are π = θ and π = 4θ − 3.

Lemma 14.3.6 θ is a common factor of X − θ and X 2 + X θ + (θ 2 − 4) such that

θ2 X − θ.

Proof: By Lemma 14.3.3 we have θ 3 | 2 in O K . Hence, as X is even, we deduce that

We are now ready to prove the main result of this section.

CB609-14 CB609/Alaca & Williams August 7, 2003 17:6 Char Count= 0

406 Applications to Diophantine Equations

Theorem 14.3.1 The solutions in integers X and Y of the equation

2Y 2 = X 3 − 4X + 2
are
(X, Y ) = (−2, ±1), (0, ±1), (2, ±1), (4, ±5), (12, ±29).

Proof: We define the nonnegative integer n by

(4θ − 3)n | X − θ, (4θ − 3)n+1 X − θ. (14.3.9)
Then, as O K is a unique factorization domain, from (14.3.3), (14.3.5), Lemma
14.3.4, Lemma 14.3.5, and (14.3.9), we deduce that
X − θ = ±θ (4θ − 3)n l ηm (a + bθ + cθ 2 )2 (14.3.10)
for
some integers
2 l, m, a, b, c. By absorbing squares into the square
a + bθ + cθ 2 , we may rewrite (14.3.10) as
2
X − θ = ±θ(4θ − 3) N L η M A + Bθ + Cθ 2 , (14.3.11)
where
L , M, N ∈ {0, 1}. (14.3.12)
Taking norms of both sides of (14.3.11), we obtain
X 3 − 4X + 2 = ±2 · 37 N · Z 2 (14.3.13)
for some Z ∈ Z. As X is even we may set X = 2X 1 , where X 1 ∈ Z, in (14.3.13),
and obtain
4X 13 − 4X 1 + 1 = ±37 N · Z 2 . (14.3.14)
Reducing (14.3.14) modulo 8, we obtain
1 ≡ ±5 N (mod 8),
showing that N
= 1. Hence by (14.3.12) we have N = 0. Thus (14.3.11) becomes
2
X − θ = ±θ L η M A + Bθ + Cθ 2 , (14.3.15)
where L , M ∈ {0, 1}. Expanding the square in (14.3.15) and making use of θ 3 =
4θ − 2, we obtain
X − θ = ±θ L η M ((A2 − 4BC) + (2AB + 8BC − 2C 2 )θ
+ (2AC + B 2 + 4C 2 )θ 2 ). (14.3.16)
We now consider each of the four possibilities
(L , M) = (0, 0), (0, 1), (1, 0), and (1, 1).
CB609-14 CB609/Alaca & Williams August 7, 2003 17:6 Char Count= 0

14.3 The Equation y(y + 1) = x(x + 1)(x + 2) 407

(i): (L , M) = (0, 0). In this case (14.3.16) becomes

X − θ = ±θ((A2 − 4BC) + (2AB + 8BC − 2C 2 )θ

+ (2AC + B 2 + 4C 2 )θ 2 ). (14.3.17)

Using θ 3 = 4θ − 2, and equating terms in 1, θ, and θ 2 on both sides of (14.3.7),

we obtain

4AC + 2B 2 + 8C 2 = ∓X, (14.3.18)

A2 − 4BC + 4B 2 + 8AC + 16C 2 = ∓1, (14.3.19)
AB + 4BC − C 2 = 0. (14.3.20)

Taking the Eq. (14.3.19) modulo 4, we see that the plus sign holds. Then (14.3.18)–
(14.3.20) can be written as

X = 4AC + 2B 2 + 8C 2 , (14.3.21)
(A + 4C)2 + 4B 2 − 4BC = 1, (14.3.22)
B(A + 4C) = C 2 . (14.3.23)

When B = 0, (14.3.23) gives C = 0. Then from (14.3.21) we obtain X = 0. If

B
= 0, from (14.3.22) and (14.2.23) we deduce that
C4
+ 4B 2 − 4BC = 1. (14.3.24)
B2
Now for all x ∈ R we have

(x − B)2 ((x + B)2 + 2B 2 ) ≥ 0.

Hence

x 4 ≥ 4B 3 x − 3B 4

and so
x4
+ 4B 2 − 4Bx ≥ B 2 .
B2
Taking x = C in this inequality, and appealing to (14.3.24), we deduce that

1 ≥ B 2,

so that B = ±1. Then from (14.3.24) we have

C 4 + 4 ∓ 4C = 1,

so that C = ±1. Hence from (14.3.23) we obtain A = C 2 /B − 4C = ∓3. Finally,

from (14.3.21) we obtain

X = 2 − 12 + 8 = −2.
CB609-14 CB609/Alaca & Williams August 7, 2003 17:6 Char Count= 0

408 Applications to Diophantine Equations

(ii): (L , M) = (0, 1). In this case (14.3.15) becomes

2
X − θ = ±θ(2θ − 1) A + Bθ + Cθ 2 .
Multiplying by θ, and absorbing θ 2 into the square, we obtain with a slight change
of notation
∓(X θ − θ 2 ) = (1 − 2θ)((A2 − 4BC) + (2AB + 8BC − 2C 2 )θ
+ (2AC + B 2 + 4C 2 )θ 2 ),
and so equating coefficients of 1, θ, θ 2 we have
0 = A2 − 4BC + 4(2AC + B 2 + 4C 2 ), (14.3.25)
∓X = 2AB + 8BC − 2C − 2(A − 4BC) − 8(2AC + B + 4C ), (14.3.26)
2 2 2 2

±1 = 2AC + B 2 + 4C 2 − 2(2AB + 8BC − 2C 2 ). (14.3.27)

From (14.3.25) we see that A is even and from (14.3.27) that B is odd. Then,
considering (14.3.27) modulo 4, we deduce that the + sign holds in (14.3.27).
Hence the − sign holds in (14.3.26). Thus (14.3.25)–(14.3.27) become
0 = (A + 4C)2 + 4B(B − C), (14.3.28)
1 = A(2C − 4B) + B 2 − 16BC + 8C 2 , (14.3.29)
X = 2A2 + 8B 2 + 34C 2 − 2AB + 16AC − 16BC. (14.3.30)
Suppose first that C = 2B. Then from (14.3.29) we obtain B 2 = 1. Since solu-
tions (A, B, C) and (−A, −B, −C) give the same value for X , we need only
take B = 1. Hence C = 2. Then from (13.3.28) we obtain (A + 8)2 = 4 so that
A = −6, −10. Then from (14.3.30) with (A, B, C) = (−6, 1, 2) we obtain X = 4
and with (A, B, C) = (−10, 1, 2) we obtain X = 12.
Suppose next that C
= 2B. Then from (14.3.29) we have
B 2 − 16BC + 8C 2 − 1
A= . (14.3.31)
4B − 2C
Thus
B2 − 1
A + 4C = . (14.3.32)
4B − 2C
Then, from (14.3.28), we deduce that
2 2
B −1
+ 4B(B − C) = 0,
4B − 2C
so that
B 4 − 2B 2 + 1 + 16B(2B − C)2 (B − C) = 0. (14.3.33)
This shows that B | 1, so that B = ±1. Then (14.3.33) gives (as C
= 2B) C = B =
±1. Next from (14.3.32) we obtain A = −4C = ∓4 and finally from (14.3.30)
CB609-14 CB609/Alaca & Williams August 7, 2003 17:6 Char Count= 0

14.3 The Equation y(y + 1) = x(x + 1)(x + 2) 409

X = 2. All solutions of 2Y 2 = X 3 − 4X + 2 have now been found.

(iii): (L , M) = (1, 0). In this case (14.3.16) becomes

∓(X − θ ) = (θ − θ 2 )((A2 − 4BC) + (2AB + 8BC − 2C 2 )θ

+(2AC + B 2 + 4C 2 )θ 2 ).

Equating coefficients of θ and θ 2 , we obtain

±1 = (A2 − 4BC) + 6(2AC + B 2 + 4C 2 ) − 4(2AB + 8BC − 2C 2 ),

0 = (2AB + 8BC − 2C 2 ) − (A2 − 4BC) − 4(2AC + B 2 + 4C 2 ).

The first equation shows that A is odd and the second that A is even. This case
cannot occur.
(iv): (L , M) = (1, 1). In this case (14.3.15) becomes
2
±(X − θ) = θ(1 − θ )(1 − 2θ) A + Bθ + Cθ 2 .

On multiplying by θ, and absorbing θ 2 into the square, we obtain with a slight

change of notation

±θ(X − θ) = (1 − 3θ + 2θ 2 )((A2 − 4BC) + (2AB + 8BC − 2C 2 )θ

+ (2AC + B 2 + 4C 2 )θ 2 ).

Equating coefficients of 1 and θ 2 , we obtain

0 = (A2 − 4BC) + 6(2AC + B 2 + 4C 2 ) − 4(2AB + 8BC − 2C 2 ),

∓1 = 9(2AC + B 2 + 4C 2 ) − 3(2AB + 8BC − 2C 2 ) + 2(A2 − 4BC).

The first equation shows that A is even and then that B is even since
6B 2 ≡ 0 (mod 4). The second equation shows that B is odd. This case cannot
occur.
This completes the proof of the theorem.

As an immediate consequence of Theorem 14.3.7 we have the main result of this

section.

Theorem 14.3.2 The only solutions in integers x and y of the equation

y(y + 1) = x(x + 1)(x + 2)

are

(x, y) = (0, 0), (0, −1), (−1, 0), (−1, −1), (−2, 0), (−2, −1), (1, 2), (1, −3),
(5, 14), (5, −15).
CB609-14 CB609/Alaca & Williams August 7, 2003 17:6 Char Count= 0

410 Applications to Diophantine Equations

Proof: This follows immediately from Theorem 14.3.1 by using the transformation
(14.3.6).

Exercises
1. Determine all integers k in the range |k| < 200 to which Theorem 14.1.1 applies.
2. Determine all integers k in the range |k| < 200 to which Theorem 14.1.2 applies.
3. Determine all integers k in the range |k| < 200 to which Theorem 14.1.3 applies.
4. Determine all integers k in the range |k| < 200 to which Theorem 14.1.4 applies.
5. Let M and N be integers such that
M ≡ 4 (mod 8), N ≡ 1 (mod 2),
p (prime) | N =⇒ p ≡ 1 or 3 (mod 8).
Set
k = M 3 − 2N 2 .
Prove that the equation y 2 = x 3 + k has no solutions in integers x and y.
6. Determine all integers k in the range |k| < 200 to which the result of Exercise 5 applies.
7. Let M and N be integers such that
M ≡ 3 (mod 4), N ≡ ±2 (mod 6),
p (prime) | N =⇒ p ≡ ±1 (mod 12).
Set
k = M 3 + 3N 2 .
Prove that the equation y 2 = x 3 + k has no solutions in integers x and y.
8. Determine all integers k in the range |k| < 200 to which the result of Exercise 7 applies.
9. Formulate and prove a result analogous to that of Exercise 7 when k has the form
M 3 − 3N 2 .
10. Determine all integers k in the range |k| < 200 to which the result of Exercise 9 applies.
11. Prove that the equation
y 2 = x 3 + 45
has no solutions in integers x and y.
12. Determine a class of integers k containing k = 45 for which the equation
y2 = x 3 + k
has no solutions in integers x and y.
13. Let M and N be integers such that
M ≡ 2 (mod 6), N ≡ ±1 (mod 6),
p (prime) | M =⇒ p ≡ 2 (mod 3).
Set
k = 4M 3 − 3N 2 .
Prove that the equation y 2 = x 3 + k has no solutions in integers x and y.
CB609-14 CB609/Alaca & Williams August 7, 2003 17:6 Char Count= 0

Biographies 411

14. Determine all integers k in the range |k| < 200 to which the result of Exercise 13
applies.
15. Show that the condition M ≡ 2 (mod 6) can be replaced by M ≡ 0 (mod 6), M
= 0
in Exercise 13 without affecting the result.
16. Determine all integers k in the range |k| < 200 to which the result of Exercise 15
applies.
17. Formulate and prove an analogous result to that of Exercise 13 for k of the form
4M 3 + 3N 2 .
18. Prove that y 2 = x 3 + 13 has no solutions in integers.
19. Prove that y 2 = x 3 + 51 has no solutions in integers.

Suggested Reading
1. W. W. Rouse Ball, A Short Account of the History of Mathematics, Dover, New York,
1960.
A brief discussion of Bachet’s work is given on pages 305 and 306.
2. E. Brown and B. T. Myers, Elliptic curves from Mordell to Diophantus and back, Amer-
ican Mathematical Monthly 109 (2002), 639–649.
In this beautifully written article the authors discuss the number of integer points (x, y) on the
elliptic curve y 2 = x 3 − x + m 2 , where m is a nonnegative integer.
3. H. M. Edgar, Classes of equations of the type y 2 = x 3 + k having no rational solutions,
Nagoya Mathematical Journal 28 (1966), 49–58.
Conditions are given for the equation y 2 = x 3 + k to have no rational solutions.
4. T. Heath, A History of Greek Mathematics, Volume 1: From Thales to Euclid, Volume 2:
From Aristarchus to Diophantus, Dover, New York, 1981.
Chapter 20 in Volume 2 contains an interesting discussion of the work of Diophantus including
his methods for finding solutions in integers of equations of degrees 1, 2, and 3.
5. H. London and R. Finkelstein, On Mordell’s Equation y 2 − k = x 3 , Bowling Green State
University Press, Bowling Green, Ohio, 1973.
The authors provide a comprehensive treatment of the equation y 2 = x 3 + k with many references.
6. L. J. Mordell, On the integer solutions of y(y + 1) = x(x + 1)(x + 2), Pacific Journal
of Mathematics 13 (1963), 1347–1351.
Section 13.3 is based on this beautifully written paper of Mordell.
7. L. J. Mordell, Diophantine Equations, Academic Press, London and New York, 1969.
Mordell’s book is a very readable standard reference text on Diophantine equations.

Biographies
1. V. Bjerknes, Axel Thue, Nordisk Matematisk Tidskrift 4 (1922), 33–46.
A biography of Axel Thue is given.
2. J. W. S. Cassels, L. J. Mordell, Bulletin of the London Mathematical Society 6 (1974),
69–96.
A biography of L. J. Mordell (1888–1972) is given.
3. H. Davenport, L. J. Mordell, Acta Arithmetica 9 (1964), 3–12.
Another biography of L. J. Mordell is presented.
CB609-14 CB609/Alaca & Williams August 7, 2003 17:6 Char Count= 0

412 Applications to Diophantine Equations

4. L. J. Mordell, Reminiscences of an octogenarian mathematician, American Mathemat-

ical Monthly 78 (1971), 952–961.
This article is based on a talk given to the Fellows of St. John’s College, Cambridge, in 1968.
5. The website

http://www-groups.dcs.st-and.ac.uk/˜history/

has biographies of Bachet, Diophantus, Mordell, and Thue.

CB609-15 CB609/Alaca & Williams August 7, 2003 17:11 Char Count= 0

List of Definitions

Definition Term defined Page

1.1.1 Integral domain 1
1.1.2 Divisor 3
1.1.3 Unit 4
1.1.4 Associate 5
1.2.1 Irreducible 6
1.2.2 Prime 6
1.3.1 Ideal 8
1.3.2 Principal ideal 9
1.3.3 Proper ideal 9
1.4.1 Principal ideal domain 10
1.4.2 Greatest common divisor 11
1.5.1 Maximal ideal 16
1.5.2 Prime ideal 18
1.6.1 Sum of ideals 21
1.6.2 Product of ideals 21
2.1.1 Euclidean function 27
2.1.2 Euclidean domain 28
2.2.1 Function φm 30
2.3.1 Universal side divisor 44
2.4.1 Almost Euclidean function 46
2.4.2 Almost Euclidean domain 47
3.1.1 Ascending chain of ideals 54
3.1.2 Terminating ascending chain 54
3.1.3 Ascending chain condition 55
3.1.4 Noetherian domain 55
3.1.5 Maximal condition 56
3.2.1 Factorization domain 59
3.3.1 Unique factorization domain 61
3.4.1 R-action 65

413
CB609-15 CB609/Alaca & Williams August 7, 2003 17:11 Char Count= 0

414 List of Definitions

Definition Term defined Page

3.4.2 R-module 65
3.4.3 Submodule 66
3.4.4 Submodule generated by a set 66
3.4.5 Finitely generated module 66
3.4.6 Factor module 66
3.4.7 Module homomorphism 67
3.5.1 Noetherian module 67
4.1.1 Element integral over a domain 74
4.1.2 Algebraic integer 74
4.1.3 Element algebraic over a field 75
4.1.4 Algebraic number 75
4.1.5 Domain integral over a subdomain 75
4.2.1 Integral closure 81
4.2.2 Integrally closed domain 84
5.1.1 Minimal polynomial of α over K 89
5.1.2 Degree of α over K 89
5.2.1 Conjugates of α over K 90
5.5.1 Simple extension 98
5.5.2 Degree of the extension K (α) over K 100
5.5.3 Cyclotomic field 102
6.1.1 Algebraic number field 109
6.1.2 The set O K 110
6.1.3 Ring of integers of an algebraic number field 110
6.2.1 Conjugate fields of an algebraic number field 115
6.3.1 Complete set of conjugates of α relative to K 116
6.3.2 Field polynomial of α over K 117
6.4.1 Discriminant of n elements in an algebraic number field 124
of degree n
6.4.2 Discriminant D(α) of an element α 124
6.4.3 Discriminant of a polynomial 126
6.5.1 Basis of an ideal 131
6.5.2 Discriminant of an ideal 135
7.1.1 Integral basis of an algebraic number field 141
7.1.2 Discriminant of an algebraic number field 141
7.1.3 Norm of an ideal 142
7.1.4 Index of θ 146
7.1.5 Monogenic number field 158
7.2.1 Minimal integer of degree k in θ 164
7.3.1 Pure cubic field 174
7.4.1 Index of a field 178
CB609-15 CB609/Alaca & Williams August 7, 2003 17:11 Char Count= 0

List of Definitions 415

Definition Term defined Page

7.4.2 Minimal index of a field 178
7.4.3 Inessential discriminant divisor 186
8.1.1 Dedekind domain 194
8.2.1 Fractional ideal 196
8.2.2 The set P̃ for a prime ideal P 197
8.3.1 Divisibility of integral ideals 204
8.4.1 Order of an ideal with respect to a prime ideal 207
8.4.2 Divisibility of fractional ideals 207
8.4.3 Order of a nonzero element with respect to a prime ideal 209
9.2.1 Norm and trace of an element 222
9.4.1 Norm of a fractional ideal 231
10.1.1 Inertial degree 237
10.1.2 Decomposition number 239
10.1.3 Ramification index 239
10.1.4 Ramification 240
10.2.1 Kronecker symbol 246
11.3.1 Fundamental unit of norm 1 272
11.4.1 Fundamental unit of norm −1 277
11.5.1 Fundamental unit 278
12.1.1 Ideal class group 299
12.1.2 Class number 299
12.2.1 Translate 301
12.2.2 Magnification 301
12.2.3 Bounded set 301
12.2.4 Closed set 301
12.2.5 Convex set 301
12.2.6 Convex body 301
12.2.7 Hypercube Ht 302
12.3.1 Centrally symmetric set 306
12.5.1 Minkowski bound 314
13.1.1 Valuations of a field element 345
13.2.1 Independent units 354
13.4.1 Fundamental system of units 361
13.4.2 Fundamental unit 362
13.7.1 Regulator 380
CB609-15 CB609/Alaca & Williams August 7, 2003 17:11 Char Count= 0

416
CB609-15 CB609/Alaca & Williams August 7, 2003 17:11 Char Count= 0

Location of Theorems

Theorem Page Theorem Page Theorem Page

1.1.1 4 2.2.8 35 3.5.1 67
1.2.1 7 2.3.1 37 3.5.2 69
1.2.2 7 2.3.2 38 3.5.3 69
1.3.1 9 2.3.3 39 3.5.4 70
1.4.1 10 2.3.4 40 3.5.5 70
1.4.2 10 2.3.5 41 4.1.1 76
1.4.3 11 2.3.6 44 4.1.2 76
1.4.4 12 2.3.7 45 4.1.3 77
1.4.5 14 2.3.8 45 4.1.4 78
1.5.1 16 2.4.1 47 4.1.5 79
1.5.2 17 2.5.1 48 4.1.6 80
1.5.3 18 2.5.2 48 4.1.7 80
1.5.4 19 2.5.3 48 4.1.8 80
1.5.5 20 2.5.4 48 4.1.9 80
1.5.6 20 2.5.5 49 4.1.10 80
1.5.7 20 2.5.6 49 4.1.11 81
1.6.1 22 2.5.7 49 4.1.12 81
1.6.2 22 2.5.8 49 4.2.1 82
2.1.1 28 3.1.1 55 4.2.2 82
2.1.2 29 3.1.2 56 4.2.3 83
2.1.3 29 3.1.3 56 4.2.4 84
2.2.1 30 3.2.1 58 4.2.5 84
2.2.2 32 3.2.2 59 4.2.6 85
2.2.3 33 3.2.3 60 5.1.1 89
2.2.4 34 3.3.1 61 5.2.1 91
2.2.5 34 3.3.2 62 5.3.1 91
2.2.6 35 3.3.3 64 5.3.2 92
2.2.7 35 3.3.4 64 5.3.3 93

417
CB609-15 CB609/Alaca & Williams August 7, 2003 17:11 Char Count= 0

418 Location of Theorems

Theorem Page Theorem Page Theorem Page

5.4.1 95 7.1.9 147 8.5.1 215
5.4.2 96 7.1.10 148 9.1.1 218
5.4.3 98 7.1.11 149 9.1.2 219
5.5.1 100 7.1.12 150 9.1.3 221
5.6.1 102 7.1.13 150 9.2.1 223
5.6.2 104 7.1.14 150 9.2.2 223
5.6.3 106 7.1.15 152 9.2.3 224
6.1.1 109 7.1.16 159 9.2.4 225
6.1.2 109 7.1.17 160 9.2.5 226
6.1.3 110 7.2.1 160 9.2.6 227
6.1.4 110 7.2.2 162 9.3.1 228
6.1.5 111 7.2.3 164 9.3.2 229
6.1.6 111 7.2.4 165 9.4.1 232
6.1.7 111 7.2.5 165 10.1.1 236
6.1.8 112 7.2.6 166 10.1.2 237
6.2.1 112 7.2.7 168 10.1.3 238
6.2.2 115 7.2.8 169 10.1.4 239
6.3.1 118 7.3.1 170 10.1.5 240
6.3.2 120 7.3.2 176 10.1.6 240
6.3.3 121 7.4.1 178 10.2.1 242
6.3.4 121 7.4.2 179 10.2.2 246
6.3.5 122 7.4.3 179 10.2.3 247
6.4.1 125 7.5.1 187 10.2.4 248
6.4.2 126 7.5.2 189 10.3.1 249
6.4.3 127 8.1.1 194 10.5.1 257
6.4.4 127 8.1.2 194 10.5.2 259
6.5.1 129 8.2.1 195 10.6.1 260
6.5.2 129 8.2.2 196 11.1.1 264
6.5.3 131 8.2.3 197 11.2.1 267
6.5.4 131 8.2.4 197 11.3.1 271
6.5.5 135 8.3.1 200 11.3.2 273
6.6.1 137 8.3.2 202 11.4.1 275
7.1.1 141 8.3.3 205 11.4.2 277
7.1.2 142 8.3.4 205 11.5.1 278
7.1.3 142 8.4.1 208 11.5.2 279
7.1.4 142 8.4.2 208 11.5.3 279
7.1.5 144 8.4.3 209 11.5.4 279
7.1.6 145 8.4.4 210 11.5.5 285
7.1.7 146 8.4.5 211 11.5.6 286
7.1.8 147 8.4.6 213 11.5.7 286
CB609-15 CB609/Alaca & Williams August 7, 2003 17:11 Char Count= 0

Location of Theorems 419

Theorem Page Theorem Page Theorem Page

11.6.1 290 12.7.2 334 14.1.1 386
11.7.1 294 13.1.1 346 14.1.2 386
12.1.1 300 13.4.1 362 14.1.3 387
12.2.1 302 13.4.2 362 14.1.4 388
12.3.1 306 13.5.1 366 14.2.1 389
12.4.1 307 13.5.2 366 14.2.2 390
12.5.1 311 13.5.3 367 14.2.3 392
12.5.2 312 13.5.4 367 14.2.4 395
12.5.3 313 13.6.1 371 14.2.5 399
12.5.4 313 13.6.2 373 14.3.1 406
12.6.1 322 13.6.3 374 14.3.2 409
12.7.1 333 13.7.1 380
CB609-15 CB609/Alaca & Williams August 7, 2003 17:11 Char Count= 0

420
CB609-15 CB609/Alaca & Williams August 7, 2003 17:11 Char Count= 0

Location of Lemmas

Lemma Page Lemma Page

2.2.1 30 13.2.16 358
13.2.1 347 13.2.17 359
13.2.2 347 13.5.1 364
13.2.3 347 13.5.2 365
13.2.4 348 13.5.3 365
13.2.5 348 13.5.4 365
13.2.6 349 13.5.5 365
13.2.7 350 13.5.6 366
13.2.8 350 13.6.1 369
13.2.9 351 13.6.2 370
13.2.10 352 14.3.1 404
13.2.11 353 14.3.2 404
13.2.12 354 14.3.3 404
13.2.13 355 14.3.4 405
13.2.14 356 14.3.5 405
13.2.15 358 14.3.6 405

421
CB609-15 CB609/Alaca & Williams August 7, 2003 17:11 Char Count= 0

422
CB609-15 CB609/Alaca & Williams August 7, 2003 17:11 Char Count= 0

Bibliography

[1] Z. I. Borevich and I. R. Shafarevich, Number Theory, Academic Press, New York and
London, 1966.
[2] H. Cohen, A Course in Computational Algebraic Number Theory, Springer-Verlag,
Berlin/Heidelberg/New York, 1996.
[3] H. Cohn, Advanced Number Theory, Dover, New York, 1980.
[4] H. Cohn, A Classical Invitation to Algebraic Numbers and Class Fields,
Springer-Verlag, New York/Heidelberg/Berlin, 1978.
[5] D. A. Cox, Primes of the Form x 2 + ny 2 , Wiley, New York, 1989.
[6] R. Dedekind, Theory of Algebraic Integers, Cambridge University Press, Cambridge,
UK, 1996.
[7] J. Esmonde and M. Ram Murty, Problems in Algebraic Number Theory,
Springer-Verlag, New York, 1999.
[8] G. J. Janusz, Algebraic Number Fields, Second Edition, Graduate Studies in
Mathematics Volume 7, American Mathematical Society, Providence, Rhode Island,
1996.
[9] H. Koch, Number Theory: Algebraic Numbers and Functions, Graduate Studies in
Mathematics Volume 24, American Mathematical Society, Providence, Rhode Island,
2000.
[10] S. Lang, Algebraic Number Theory, Springer-Verlag, New York, 1986.
[11] R. L. Long, Algebraic Number Theory, Dekker, New York, 1977.
[12] H. B. Mann, Introduction to Algebraic Number Theory, Ohio State University Press,
Columbus, Ohio, 1955.
[13] D. A. Marcus, Number Fields, Springer-Verlag, New York/Heidelberg/Berlin,
1977.
[14] R. A. Mollin, Algebraic Number Theory, Chapman and Hall/CRC Press,
London/Boca Raton, Florida, 1999.
[15] R. Narasimhan, S. Raghavan, S. S. Rangachari, and S. Lal, Algebraic Number
Theory, Tata Institute of Fundamental Research, Bombay, India, 1966.
[16] W. Narkiewicz, Elementary and Analytic Theory of Algebraic Numbers,
Springer-Verlag, Berlin/Heidelberg/New York, 1989.
[17] T. Ono, An Introduction to Algebraic Number Theory, Plenum, New York, 1990.
[18] M. E. Pohst, Computational Algebraic Number Theory, Birkhäuser Verlag,
Basel/Boston/Berlin, 1993.
[19] M. Pohst and H. Zassenhaus, Algorithmic Algebraic Number Theory, Cambridge
University Press, Cambridge, UK, 1989.
[20] H. Pollard and H. G. Diamond, The Theory of Algebraic Numbers, Mathematical
Association of America, Washington, DC, 1975.

423
CB609-15 CB609/Alaca & Williams August 7, 2003 17:11 Char Count= 0

424 Bibliography

[21] P. Ribenboim, Classical Theory of Algebraic Numbers, Springer-Verlag, New York,

2001.
[22] P. Samuel, Algebraic Theory of Numbers, Kershaw, London, 1972.
[23] I. Stewart and D. Tall, Algebraic Number Theory and Fermat’s Last Theorem, A. K.
Peters, Natick, Massachusetts, 2002.
[24] E. Weiss, Algebraic Number Theory, McGraw-Hill, New York, 1963.
[25] H. Weyl, Algebraic Theory of Numbers, Princeton University Press, Princeton, New
Jersey, 1940.
CB609-IND CB609/Alaca & Williams August 7, 2003 17:14 Char Count= 0

Index

Abel, N. H., 4, 26 Cámpoli, O. A., 47, 51

Abelian group, 4, 65 cancellation law, 3
Alaca, Ş., 177, 191 Carlitz, L., 300, 343
Aleksandrov, P. S., 73 Cassels, J. W. S., 411
algebraic integer, 74 centrally symmetric set, 306
algebraic number, 75 Chatland, H., 35, 51
algebraic number field, 109 Chinese remainder theorem, 213, 217
algebraic over a field, 75 Clark, D. A., 46, 51
almost Euclidean domain, 46 class number, 299, 313
almost Euclidean function, 46 closed set, 301
Arpaia, P. J., 51 Cohen, H., 383
ascending chain condition, 55 Cohn, P. M., 25, 72
ascending chain of ideals, 54 common denominator of a fractional ideal, 196
associate, 5 common index divisor, 186
Atterton, T. W., 87 complete set of conjugates, 116
conjugate fields, 115
Bachet’s equation, 385 conjugate ideal, 247
Bachet, C., 385, 412 conjugates of α relative to K , 116
Bachman, G., 262 conjugates of an element algebraic over a field,
Ball, W. W. R., 411 90
Barner, K., 53 content of a polynomial, 83
Barnes, E. S., 34, 51 continued fraction, 264, 298
basis of an ideal, 131 convex body, 301
basis of an ideal in a quadratic field, convex set, 301
135 Cox, D. A., 25, 51
Begehr, H. G. W., 298 cubic domain, 2
Behrbohm, H., 34 cubic field, 110
Bell, E. T., 25, 140, 193, 298 cubic reciprocity, 1
Berg, E., 34 cyclotomic field, 102, 186, 343
Bhattacharya, P. B., 234
binary quadratic form, 47, 331, 336 Davenport, H., 35, 51, 53, 298, 411
biquadratic reciprocity, 1, 25 decomposition number, 239
Bjerknes, V., 411 Dedekind domain, 194
Bolzano–Weierstrass theorem, 305 Dedekind, R., 84, 87, 156, 176, 184, 192–194, 240,
Borevich, Z. I., 262, 343 249, 257, 262
bounded set, 301 degree of α over K , 89
Bourbaki, N., 87 degree of an element algebraic over a field, 89
Brauer, A. T., 34, 53 degree of an extension, 100
Brawley, J. V., 343 degree of the extension K (α) over K , 100
Brill, A., 159, 191, 193 diameter, 303
Brillhart, J., 253, 262 Dick, A., 73
Brouncker, W., 267, 298 Dickson, L. E., 34, 53
Brown, E., 411 Diophantus, 385, 412
Bühler, W. K., 25 Dirichlet’s unit theorem, 346, 383
Burckhardt, J. J., 298 Dirichlet, P. G. L., 25, 279, 298, 322, 346

425
CB609-IND CB609/Alaca & Williams August 7, 2003 17:14 Char Count= 0

426 Index

discriminant, 123, 141 Harper, M., 46, 52

discriminant of a cyclotomic field, 187 Hartley, B., 73
discriminant of a field, 141 Hawkes, T. O., 73
discriminant of a polynomial, 126, 140 Heath, T., 411
discriminant of a quadratic field, 141 Heath-Brown, D. R., 48, 52
discriminant of a set of elements, 124 Heilbronn, H. A., 34, 53
discriminant of an element, 124 Hilbert basis theorem, 57, 73
discriminant of an ideal, 135 Hilbert, D., 57, 73, 279
distributive law, 22 Hoffman, P., 53
divisibility of fractional ideals, 207 Hofreiter, N., 34
divisibility of integral ideals, 204 Howard, F. T., 217, 343
divisor, 3 Hua, L. K., 35
Dubois, D. W., 52 Huard, J. G., 189, 192, 262
Dummit, D. S., 72, 192 Hudson, R. H., 192
hypercube, 302
Edgar, H. M., 411
Eisenstein domain, 1 ideal, 8
Eisenstein integer, 1 ideal class group, 299, 311
Eisenstein, G., 1, 25, 26 independent units, 354
Eisenstein’s irreducibility criterion, 92 index of θ, 146
elementary operations, 218 index of a cubic field, 186
Emerson, E., 402 index of a field, 178
Engstrom, H. T., 192 inertial degree, 237
equivalence class, 5 inessential discriminant divisor, 186
equivalence relation, 5 Inkeri, K., 35
Erdös, P., 34, 53 integer of degree k in θ, 160
Euclid, 368 integers of a quadratic field, 96
Euclidean algorithm, 29 integral basis, 141
Euclidean domain, 27, 28 integral basis of a cyclotomic field,
Euclidean function, 27, 28 187
Euler’s phi function, 186, 383 integral closure, 81, 87
Euler, L., 267, 298 integral domain, 1
integral ideal, 196, 218
factor module, 66 integral over a domain, 74
factorization domain, 59, 60 integrally closed, 84
Fermat, P., 53, 267, 385 integrally closed domain, 84
field, 1 irreducible, 6, 7
field of quotients, 3
field polynomial, 117 Jackson, T., 48, 52
finitely generated module, 66 Jacobson, N., 73
Finkelstein, R., 411 Jain, S. K., 234
Foote, R. M., 72 Jodeit, M. A., 28, 52
fractional ideal, 196 Jones, B. W., 402
Fraleigh, J. B., 3, 25
free Abelian group, 219, 235 K -conjugates of α, 116
Frei, G., 384 Keston, J. F., 35
Funakura, T., 177, 192 Kimberling, C. H., 73
fundamental system of units, 361, 383 Kinkade, L., 25
fundamental unit, 264, 278, 373, 375 Kisilevsky, H., 192
fundamental unit of norm −1, 277 Ko, C., 35
fundamental unit of norm 1, 272 Koch, H., 298
Kramer, J., 298
Galois extension, 122 Kronecker symbol, 246
Gauss, C. F., 1, 25, 26, 83, 140 Kummer, E., 193
Gaussian domain, 1, 73
Gaussian integer, 1 Lagrange, J.-L., 267, 298
generator of a field, 178 lattice, 300
Gerst, I., 253, 262 lattice points, 300
Girard–Fermat theorem, 48, 52, 53 Legendre symbol, 48, 242, 246
Goodstein, R. L., 140 Lenstra, H. W., 298
Gras, M.-N., 192 Levesque, C., 383
greatest common divisor, 11, 29, 64 limit point, 301
Greene’s theorem, 64 Llorente, P., 177, 186, 192
Greene, J., 47, 52, 72 London, H., 411
CB609-IND CB609/Alaca & Williams August 7, 2003 17:14 Char Count= 0

Index 427

MacDuffee, C. C., 234 prime ideal, 18

magnification, 301 prime ideal lying over a rational prime, 237
Mahoney, M. S., 53 prime lying below a prime ideal, 237
Mann, H. B., 192, 263 primitive polynomial, 83
Marcus, D. A., 343 principal ideal, 9
Masley, J. M., 343 principal ideal domain, 10, 27, 29, 30, 56, 60, 61
Masser, D. W., 140 principal ideal ring, 25
maximal condition, 56 product of ideals, 21
maximal ideal, 16 proper ideal, 9
Min, S. H., 35 pure cubic field, 174
minimal index of a field, 178
minimal integer, 164 quadratic domain, 2
minimal integer of degree k in θ , 163, 164 quadratic field, 110
minimal polynomial of an element algebraic over a quartic field, 110
field, 89 quotient, 27, 28
Minkowski bound, 311, 313, 314 quotient field, 3, 111
Minkowski’s convex body theorem, 305, 306
Minkowski’s linear forms theorem, 306, 307 Rédei, L., 35, 51
Minkowski’s translate theorem, 302 ramification, 240
Minkowski, H., 299 ramification index, 239
Mirsky, L., 140 reducible, 6
module, 64 reflexive property, 3, 5
module homomorphism, 67 regulator, 380
module isomorphism, 67 Reid, C., 73
Mollin, R. A., 193 remainder, 27, 28
monogenic number field, 158, 249 Remak, R., 35, 53
monomorphism, 112 Rhai, T.-S., 28, 52
Montgomery, H. L., 298, 343, 383 ring of integers of an algebraic number field, 110
Mordell, L. J., 402, 411, 412 Rogers, C. A, 53
Motzkin, Th., 52 Rogers, K., 52
multiple extension, 102 Ross, A. E., 192
Muskat, J. B., 343 Rowe, D. E., 25
Myers, B. T., 411 Rudin, W., 73

Nagpaul, S. R., 234 Samuel, P., 46, 52, 73, 217

Narkiewicz, W., 187, 192 Schappacher, N., 298
Nart, E., 177, 186, 192 Scheinerman, E. R., 108
neighborhood, 301 Schuster, L., 35
Niven, I., 298, 383 self-conjugate ideal, 249
Noether, E., 54, 73 Shafarevich, I. R., 262, 343
Noetherian domain, 54, 55, 57, 58 Shanks, D., 108
Noetherian module, 67 Sheh, W. T., 35
Noetherian ring, 55 simple extension, 98
norm of a fractional ideal, 231 Singh, S., 52
norm of a prime ideal, 237 Smith normal form, 219, 234
norm of a principal ideal, 145 Smith, H. J. S., 234, 235
norm of an element, 222 smooth numbers, 298
norm of an ideal, 142, 218 Spearman, B. K., 176, 189, 192, 193, 262, 343
norm-Euclidean, 30, 32 Stark, H. M., 343
normal extension, 122 Steger, A., 52
normal field, 344 Steinitz, E., 84, 87
Stewart, I. N., 235
Oppenheim, A., 35 Stickelberger, L., 193
order of an element with respect to a prime ideal, 209 strictly ascending chain, 54
order of an ideal with respect to a prime ideal, 207 Stubhaug, A., 26
Ore, O., 25, 26 subdomain, 2
submodule, 65, 66
p-Eisenstein, 92 submodule generated by a set, 66
Patz, W., 298 sum of ideals, 21
Pell’s equation, 298 Swinnerton-Dyer, H. P. F., 35, 51
Pell, J., 267, 298 symmetric property, 5
Perron, O., 35, 53
power basis, 158, 178 Tall, D. O., 235
prime, 5, 6 Taussky-Todd, O., 192
CB609-IND CB609/Alaca & Williams August 7, 2003 17:14 Char Count= 0

428 Index

terminating ascending chain, 54 Vandermonde determinant, 125, 140

Thiele, E. J., 298 Vandermonde, A.-T., 140
Thue, A., 385, 411, 412 Varouchas, Y., 48, 52
to contain is to divide, 208 vector space, 64
Top, J., 384
totally complex field, 344 Wagner, J., 25
totally imaginary field, 344 Walling, L., 384
totally real field, 344 Waterhouse, W. C., 25
trace of an element, 222 Weil, A., 25
transitive property, 3, 5 Weyl, H., 73
translate, 301 Williams, H. C., 108
Williams, K. S., 48, 52, 53, 158, 176, 189, 192, 193,
unique factorization domain, 61, 72 262, 343
unit, 4 Wilson, J. C., 53
unit group, 97, 264 Wyman, B. F., 253, 263
units of norm −1, 275
units of norm 1, 271 Zagier, D., 48, 53
universal side divisor, 44 Zariski, O., 73, 217
Zassenhaus, H., 192
valuations of an element, 345 Zuckerman, H. S., 298, 383
van der Waerden, B. L., 73, 383, 384 Zyliński, E., 186, 193

Awesome Polynomials For Mathematics Competitions Look Inside
No ratings yet
Awesome Polynomials For Mathematics Competitions Look Inside
12 pages
Algebraic Number Theory (Franz Lemmermeyer
No ratings yet
Algebraic Number Theory (Franz Lemmermeyer
91 pages
Matrix Assignment
No ratings yet
Matrix Assignment
10 pages
(Lecture Notes) Andrei Jorza-Math 5c - Introduction To Abstract Algebra, Spring 2012-2013 - Solutions To Some Problems in Dummit & Foote (2013)
No ratings yet
(Lecture Notes) Andrei Jorza-Math 5c - Introduction To Abstract Algebra, Spring 2012-2013 - Solutions To Some Problems in Dummit & Foote (2013)
30 pages
4) Number Theory (J. Hunter)
No ratings yet
4) Number Theory (J. Hunter)
160 pages
2010 ARML Advanced Number Theory
No ratings yet
2010 ARML Advanced Number Theory
2 pages
Isi Book Solutions
No ratings yet
Isi Book Solutions
71 pages
MATH
No ratings yet
MATH
1 page
117 Polynomial Problems From Amsp Toc
No ratings yet
117 Polynomial Problems From Amsp Toc
3 pages
Problems and Results On 3-Chromatic Hypergraphs An
No ratings yet
Problems and Results On 3-Chromatic Hypergraphs An
20 pages
Elementary Number Theory
No ratings yet
Elementary Number Theory
14 pages
Number Theory Muzammil Tanveer
No ratings yet
Number Theory Muzammil Tanveer
71 pages
Inclusion Exclusion Principle
No ratings yet
Inclusion Exclusion Principle
8 pages
Algebra Can Be Fun by Yakov I. Perelman, G. Yankovsky (Z-Lib - Org) - Text
100% (1)
Algebra Can Be Fun by Yakov I. Perelman, G. Yankovsky (Z-Lib - Org) - Text
234 pages
M.SC Mathematics Semester I.
No ratings yet
M.SC Mathematics Semester I.
13 pages
Logical Reasoning
No ratings yet
Logical Reasoning
8 pages
117 Polynomial Problems From Amsp Toc
0% (1)
117 Polynomial Problems From Amsp Toc
3 pages
Introductory Modular Arithmetic in Mathematical Competitions - Lecture Notes
No ratings yet
Introductory Modular Arithmetic in Mathematical Competitions - Lecture Notes
12 pages
Proof of The ABC Conjecture by P S Thul
No ratings yet
Proof of The ABC Conjecture by P S Thul
92 pages
Algorithmic Number Theory (Web Draft, 2002) - S. Arun-Kumar
100% (1)
Algorithmic Number Theory (Web Draft, 2002) - S. Arun-Kumar
200 pages
KVPY Report: Analytic Number Theory
No ratings yet
KVPY Report: Analytic Number Theory
8 pages
Bijections
No ratings yet
Bijections
5 pages
En Enumeration Notes
No ratings yet
En Enumeration Notes
115 pages
Introduction of Sets: Ets and Unctions
No ratings yet
Introduction of Sets: Ets and Unctions
0 pages
Ps 552 Lectures
No ratings yet
Ps 552 Lectures
70 pages
(Bolyai Society Mathematical Studies) Ervin Gyori, Gyula O.H. Katona, László Lovász - Horizons of Combinatorics (2008, Springer)
No ratings yet
(Bolyai Society Mathematical Studies) Ervin Gyori, Gyula O.H. Katona, László Lovász - Horizons of Combinatorics (2008, Springer)
274 pages
Mathematical Plums - Hosberger
No ratings yet
Mathematical Plums - Hosberger
190 pages
Maaatom2 2
No ratings yet
Maaatom2 2
50 pages
Analytic Number Theory
No ratings yet
Analytic Number Theory
14 pages
1983 AIME Problems
No ratings yet
1983 AIME Problems
17 pages
Recurrent Sequences: Key Results, Applications, and Problems Series: Problem Books in Mathematics
No ratings yet
Recurrent Sequences: Key Results, Applications, and Problems Series: Problem Books in Mathematics
1 page
Beamer Presentation For The P Adic Integers
No ratings yet
Beamer Presentation For The P Adic Integers
71 pages
Cnditional Probability
No ratings yet
Cnditional Probability
5 pages
Properties of Square Numbers
No ratings yet
Properties of Square Numbers
1 page
Enumerative Combinatorics
No ratings yet
Enumerative Combinatorics
505 pages
Stair Step Math
No ratings yet
Stair Step Math
84 pages
Basic Properties of Numbers
No ratings yet
Basic Properties of Numbers
5 pages
Inter 1
No ratings yet
Inter 1
181 pages
Programme Project Report (PPR) of Master of Science (Mathematics) M.Sc. (Mathematics)
No ratings yet
Programme Project Report (PPR) of Master of Science (Mathematics) M.Sc. (Mathematics)
23 pages
Jozsef Wildt 2014
No ratings yet
Jozsef Wildt 2014
7 pages
ZPZ Corrected - Alison Miller - MOP 2011 PDF
No ratings yet
ZPZ Corrected - Alison Miller - MOP 2011 PDF
5 pages
The Contest Probelm Book IV
No ratings yet
The Contest Probelm Book IV
201 pages
Invariants
No ratings yet
Invariants
2 pages
111 KDD
100% (1)
111 KDD
3 pages
Herstein Topics in Algebra Solution 5 3
No ratings yet
Herstein Topics in Algebra Solution 5 3
9 pages
Rearrangement Inequality
0% (1)
Rearrangement Inequality
3 pages
2021 Y11 MM12 June Exam P1 - Vce Methods Unit 1
No ratings yet
2021 Y11 MM12 June Exam P1 - Vce Methods Unit 1
8 pages
Some Results On Cubic Residues
No ratings yet
Some Results On Cubic Residues
5 pages
ICPC Asia West Finals Statements
No ratings yet
ICPC Asia West Finals Statements
22 pages
Donald R. Woods Notes On Introductory Combinatorics
No ratings yet
Donald R. Woods Notes On Introductory Combinatorics
200 pages
Vedic Maths: Quant
No ratings yet
Vedic Maths: Quant
15 pages
Algebra Maths Book
No ratings yet
Algebra Maths Book
430 pages
Sitomer Conrad MassPoints Crux Apr1976
No ratings yet
Sitomer Conrad MassPoints Crux Apr1976
35 pages
BCW Hall 2
No ratings yet
BCW Hall 2
8 pages
Number Theory
100% (3)
Number Theory
140 pages
Matthew Morrow - Algebraic Number Theory
No ratings yet
Matthew Morrow - Algebraic Number Theory
63 pages
Larry Joel Goldstein - Analytic Number Theory
No ratings yet
Larry Joel Goldstein - Analytic Number Theory
300 pages
Algebraic Number Theory
No ratings yet
Algebraic Number Theory
83 pages
Pub Factorization Unique and Otherwise
No ratings yet
Pub Factorization Unique and Otherwise
262 pages
@iitwale On Telegram: Decucted Portion Mathematics Code - 041 Class Ix
No ratings yet
@iitwale On Telegram: Decucted Portion Mathematics Code - 041 Class Ix
2 pages
(Ebook PDF) Elementary & Intermediate Algebra 4th Edition by Michael Sullivan III Download
No ratings yet
(Ebook PDF) Elementary & Intermediate Algebra 4th Edition by Michael Sullivan III Download
54 pages
Summative Assessment 2 Pre Cal
No ratings yet
Summative Assessment 2 Pre Cal
1 page
MuPAD Users Guide
No ratings yet
MuPAD Users Guide
1,004 pages
MA101 College Algebra (Course Outline)
No ratings yet
MA101 College Algebra (Course Outline)
2 pages
Experiment 6 - Linear Systems, Regression, Curve Fitting, and Interpolation
No ratings yet
Experiment 6 - Linear Systems, Regression, Curve Fitting, and Interpolation
24 pages
ST Quadratic Part 1
No ratings yet
ST Quadratic Part 1
2 pages
Chapter 04
No ratings yet
Chapter 04
31 pages
XI Chemistry Study Materials Class XI 281022 221029 203855 1 Removed
No ratings yet
XI Chemistry Study Materials Class XI 281022 221029 203855 1 Removed
6 pages
XII Maths Important Concept and Question
No ratings yet
XII Maths Important Concept and Question
73 pages
Prealgebra 2nd Edition by Julie Miller Molly ONeill and Nancy Hyde Solution Manual
100% (48)
Prealgebra 2nd Edition by Julie Miller Molly ONeill and Nancy Hyde Solution Manual
141 pages
Ma2111 Impartant
No ratings yet
Ma2111 Impartant
13 pages
Math 10 Peta 3rd Quarter
No ratings yet
Math 10 Peta 3rd Quarter
2 pages
MATH1061 Practice Exam
No ratings yet
MATH1061 Practice Exam
35 pages
Polynomial Handling Using NumPy
No ratings yet
Polynomial Handling Using NumPy
3 pages
Integration Formula
No ratings yet
Integration Formula
5 pages
Inverse Proportion Web
No ratings yet
Inverse Proportion Web
9 pages
Lecture 17: General Quantum Errors CSS Codes: CPSC 519/619: Quantum Computation John Watrous, University of Calgary
No ratings yet
Lecture 17: General Quantum Errors CSS Codes: CPSC 519/619: Quantum Computation John Watrous, University of Calgary
7 pages
Putnam 2012
No ratings yet
Putnam 2012
1 page
5th HCF& LCM - WS-9
No ratings yet
5th HCF& LCM - WS-9
2 pages
Matlab Implementation of FEM For Elasticity
No ratings yet
Matlab Implementation of FEM For Elasticity
25 pages
Assignment - Matrics - Math 23-24
No ratings yet
Assignment - Matrics - Math 23-24
3 pages
7.4h Solving Equations Algebraic Fractions TES
No ratings yet
7.4h Solving Equations Algebraic Fractions TES
6 pages
Espinosa Mark Joven D
No ratings yet
Espinosa Mark Joven D
4 pages
Jr. Maths 1A 2024 AP
No ratings yet
Jr. Maths 1A 2024 AP
15 pages
Basiccalculus q1 Module1 Week1
No ratings yet
Basiccalculus q1 Module1 Week1
14 pages
Name: Teacher: Date: Score:: Fractions Worksheets
No ratings yet
Name: Teacher: Date: Score:: Fractions Worksheets
2 pages
Radicals
No ratings yet
Radicals
7 pages
Quadratics, Inequalities and Simultaneous Equations (H) - Unlocked
No ratings yet
Quadratics, Inequalities and Simultaneous Equations (H) - Unlocked
15 pages
10 Mathematics Ncert Ch03 Pair of Linear Equations in Two Variables Ex 3.1
No ratings yet
10 Mathematics Ncert Ch03 Pair of Linear Equations in Two Variables Ex 3.1
4 pages

Introductory Algebraic Number Theory

Uploaded by

Introductory Algebraic Number Theory

Uploaded by

CB609-driver CB609/Alaca & Williams August 27, 2003 17:1 Char Count= 0

This page intentionally left blank

INTRODUCTORY ALGEBRAIC NUMBER THEORY

Kenneth S. Williams is Professor Emeritus and Distinguished Research Professor of Mathematics at

INTRODUCTORY ALGEBRAIC NUMBER

Cambridge University Press

© Saban Alaca and Kenneth S. Williams 2004

This publication is in copyright. Subject to statutory exception and to the provision of

First published in print format 2004

isbn-13 978-0-511-16494-1 eBook (EBL)

isbn-13 978-0-521-83250-2 hardback

isbn-13 978-0-521-54011-7 paperback

List of Tables page xi

8.3 Factorization into Prime Ideals 200

Suggested Reading 343

It is planned to provide solutions to selected questions, as well as corrections to

to the class of 2000–1 (Yaroslav Bezverkhnyev, Joanne Charlebois, Colette Haley,

1.1 Integral Domains

Definition 1.1.1 (Integral domain) An integral domain is a commutative ring that

Example 1.1.2 Z + Zi = {a + bi | a, b ∈ Z} is an integral domain. The ele-

Example 1.1.3 Z + Zω = {a +√ bω | a, b ∈ Z}, where ω is the complex cube

a quadratic domain. If k is a nonzero integer such that k 2 divides m then

Example 1.1.10 Z + Zθ + Zθ 2 = {a + bθ + cθ 2 | a, b, c ∈ Z}, where θ is a root

1.1 Integral Domains 3

Properties of an Integral Domain

1 = 1 · 1 (as 1 is an identity) = 1 (as 1 is an identity).

(b) D possesses a left cancellation law, that is,

as well as a right cancellation law

ac = bc, c = 0 =⇒ a = b (a, b, c ∈ D).

Example 1.1.12 1 + i | 2 in Z + Zi as 2 = (1 + i)(1 − i).

Example 1.1.13 x 2 + x + 1 | x 4 + x 2 + 1 in Z[x] as x 4 + x 2 + 1 = (x 2 + x + 1)

Example 1.1.14 (1 − ω)2 | 3 in Z + Zω as 3 = (1 − ω)2 (1 + ω) (see Example

(c) a | b and a | c implies a | xb + yc for any x ∈ D and y ∈ D.

Definition 1.1.3 (Unit) An element a of an integral domain D is called a unit if

Theorem 1.1.1 If D is an integral domain then U (D) is an Abelian group with

Proof: U (D) is closed under multiplication by property (d). Multiplication of el-

Example 1.1.18 Let Zn denote the cyclic group of order n.

1.2 Irreducibles and Primes 5

We remark that in Chapter 11 we will show that

Definition 1.1.4 (Associate) Two nonzero elements a and b of an integral domain

1.2 Irreducibles and Primes

Definition 1.2.1 (Irreducible) A nonzero, nonunit element a of an integral domain

Example 1.2.1 2 is irreducible in Z, for if 2 = ab with a ∈ Z and b ∈ Z then

Hence we see that

(a, b) = (±1, 0) or (±2, 0)

Definition 1.2.2 (Prime) A nonzero, nonunit element p of an integral domain D

Example 1.2.4 2 is a prime in Z. Suppose 2 | ab, where a, b ∈ Z, so that ab is

1.2 Irreducibles and Primes 7

Example 1.2.6 1 + i is a prime in Z + Zi. To show this, suppose that 1 + i |

(a + bi)(c + di) = (1 + i)(x + yi).

Taking the modulus of both sides of this equation, we obtain

As 2 is a prime in Z, we have either 2 | a 2 + b2 or 2 | c2 + d 2 . Interchanging a + bi

a + bi = 2(r + si) = (1 + i)((r + s) + (−r + s)i),

so that 1 + i | a + bi. In the latter case a = 2r + 1 and b = 2s + 1, where r and

a + bi = 2(r + si) + (1 + i) = (1 + i)((r + s + 1) + (−r + s)i),

so that 1 + i | a + bi. Hence 1 + i is a prime in Z + Zi.

Theorem 1.2.1 In any integral domain D a prime is irreducible.

Proof: Let p ∈ D be a prime and suppose that p = ab, where a, b ∈ D. As ab =

The converse of Theorem 1.2.1

Every quadratic polynomial in D[X ] having roots in the quotient

Then every irreducible in D is prime.

Let r = ab/ p ∈ D, and consider the quadratic polynomial

f (X ) = p(X − a/ p)(X − b/ p).

in D[X ]. Then cd = p. As p is irreducible, one of c and d is a unit of D, say d, so

Definition 1.3.1 (Ideal) An ideal I of an integral domain D is a nonempty subset

It is clear that if a1 , . . . , an ∈ I then r1 a1 + · · · + rn an ∈ I for all r1 , . . . , rn ∈ D.

Example 1.3.1 If {a1 , . . . , an } is a set of elements of the integral domain D then

is an ideal of D, which we denote by a1 , . . . , an .

Definition 1.3.2 (Principal ideal) An ideal I of an integral domain D is called a

If D is an integral domain the principal ideal a generated by a ∈ D is just the

Definition 1.3.3 (Proper ideal) An ideal I of an integral domain D is called a

Thus a proper ideal of an integral domain D is an ideal I such that {0} ⊂ I ⊂ D.

Example 1.3.2 For any positive integer k, the set

Example 1.3.3 Let

Example 1.3.4 Let

Theorem 1.3.1 Let D be an integral domain and let a, b ∈ D ∗ = D \ {0}. Then

1 = 1 · 1 (as 1 is an identity) = 1 (as 1 is an identity).

is an ideal of D, which we denote by a1 , . . . , an .

If D is an integral domain the principal ideal a generated by a ∈ D is just the

Proof: Let I be an ideal of Z. If I = {0} then I = 0 is a principal ideal. Thus we

so that y isan odd integer satisfying

Example 1.5.1 The ideal x 2 + 1 is maximal in R[x]. To show this, assume

Example 1.5.2 5 is not a maximal ideal of Z + Zi as

so that J = 1 = D. This proves that I is maximal.

so that (b + I )−1 exists and is equal to y + I .

Example 1.5.4 The principal ideal I = x 2 + 1 is not a prime ideal of C[x] as

Example 1.5.5 The ideal I = 1 + i is a prime ideal of Z + Zi. To see this,

Proof: Suppose that a is a prime ideal of D. Let b, c ∈ D be such that a | bc so

a = 0 so bd = 1. Thus b is a unit and J = b = D ⊃ J , a contradiction. Hence I

Further, if I = i and J = j are principal ideals, then I J = i j. We leave it to

m + n = d.

Let a, b ∈ D be such that ab ∈ P. Set A = a, B = b so that AB = ab =

16. Prove that x is a prime ideal of Z[x].