M.

Ram Murty
Jody Esmonde

Problems in
Algebraic Number Theory
Second Edition
M. Ram Murty Jody Esmonde
Department of Mathematics and Statistics Graduate School of Education
Queen’s University University of California at Berkeley
Kingston, Ontario K7L 3N6 Berkeley, CA 94720
Canada USA
[email protected] [email protected]

Editorial Board
S. Axler F.W. Gehring K.A. Ribet
Mathematics Department Mathematics Department Mathematics Department
San Francisco State East Hall University of California,
University University of Michigan Berkeley
San Francisco, CA 94132 Ann Arbor, MI 48109 Berkeley, CA 94720-3840
USA USA USA
[email protected] [email protected]. [email protected]
umich.edu

Mathematics Subject Classification (2000): 11Rxx 11-01

Library of Congress Cataloging-in-Publication Data

Esmonde, Jody
Problems in algebraic number theory / Jody Esmonde, M. Ram Murty.—2nd ed.
p. cm. — (Graduate texts in mathematics ; 190)
Includes bibliographical references and index.
ISBN 0-387-22182-4 (alk. paper)
1. Algebraic number theory—Problems, exercises, etc. I. Murty, Maruti Ram. II. Title.
III. Series.
QA247.E76 2004
512.7′4—dc22 2004052213

ISBN 0-387-22182-4 Printed on acid-free paper.

 2005 Springer Science+Business Media, Inc.

All rights reserved. This work may not be translated or copied in whole or in part without the
written permission of the publisher (Springer Science+Business Media, Inc., 233 Spring Street, New
York, NY 10013, USA), except for brief excerpts in connection with reviews or scholarly analysis.
Use in connection with any form of information storage and retrieval, electronic adaptation, com-
puter software, or by similar or dissimilar methodology now known or hereafter developed is for-
bidden.
The use in this publication of trade names, trademarks, service marks, and similar terms, even if
they are not identified as such, is not to be taken as an expression of opinion as to whether or not
they are subject to proprietary rights.

Printed in the United States of America. (MP)

9 8 7 6 5 4 3 2 1 SPIN 10950647

springeronline.com
It is practice first and knowledge afterwards.
Vivekananda
Preface to the Second
Edition

Since arts are more easily learnt by examples than precepts, I have
thought fit to adjoin the solutions of the following problems.
Isaac Newton, in Universal Arithmetick
Learning is a mysterious process. No one can say what the precise rules
of learning are. However, it is an agreed upon fact that the study of good
examples plays a fundamental role in learning. With respect to mathemat-
ics, it is well-known that problem-solving helps one acquire routine skills in
how and when to apply a theorem. It also allows one to discover nuances of
the theory and leads one to ask further questions that suggest new avenues
of research. This principle resonates with the famous aphorism of Lichten-
berg, “What you have been obliged to discover by yourself leaves a path in
your mind which you can use again when the need arises.”
This book grew out of various courses given at Queen’s University be-
tween 1996 and 2004. In the short span of a semester, it is difficult to cover
enough material to give students the confidence that they have mastered
some portion of the subject. Consequently, I have found that a problem-
solving format is the best way to deal with this challenge. The salient
features of the theory are presented in class along with a few examples, and
then the students are expected to teach themselves the finer aspects of the
theory through worked examples.
This is a revised and expanded version of “Problems in Algebraic Num-
ber Theory” originally published by Springer-Verlag as GTM 190. The
new edition has an extra chapter on density theorems. It introduces the
reader to the magnificent interplay between algebraic methods and analytic
methods that has come to be a dominant theme of number theory.
I would like to thank Alina Cojocaru, Wentang Kuo, Yu-Ru Liu, Stephen
Miller, Kumar Murty, Yiannis Petridis and Mike Roth for their corrections
and comments on the first edition as well as their feedback on the new
material.

Kingston, Ontario Ram Murty

March 2004

vii
Preface to the First
Edition

It is said that Ramanujan taught himself mathematics by systematically

working through 6000 problems1 of Carr’s Synopsis of Elementary Results
in Pure and Applied Mathematics. Freeman Dyson in his Disturbing the
Universe describes the mathematical days of his youth when he spent his
summer months working through hundreds of problems in differential equa-
tions. If we look back at our own mathematical development, we can certify
that problem solving plays an important role in the training of the research
mind. In fact, it would not be an exaggeration to say that the ability to
do research is essentially the art of asking the “right” questions. I suppose
Pólya summarized this in his famous dictum: if you can’t solve a problem,
then there is an easier problem you can’t solve – find it!
This book is a collection of about 500 problems in algebraic number
theory. They are systematically arranged to reveal the evolution of concepts
and ideas of the subject. All of the problems are completely solved and
no doubt, the solutions may not all be the “optimal” ones. However, we
feel that the exposition facilitates independent study. Indeed, any student
with the usual background of undergraduate algebra should be able to
work through these problems on his/her own. It is our conviction that
the knowledge gained by such a journey is more valuable than an abstract
“Bourbaki-style” treatment of the subject.
How does one do research? This is a question that is on the mind of
every graduate student. It is best answered by quoting Pólya and Szegö:
“General rules which could prescribe in detail the most useful discipline
of thought are not known to us. Even if such rules could be formulated,
they would not be very useful. Rather than knowing the correct rules of
thought theoretically, one must have them assimilated into one’s flesh and
blood ready for instant and instinctive use. Therefore, for the schooling of
one’s powers of thought only the practice of thinking is really useful. The
1 Actually, Carr’s Synopsis is not a problem book. It is a collection of theorems used

by students to prepare themselves for the Cambridge Tripos. Ramanujan made it famous
by using it as a problem book.

ix
x Preface

independent solving of challenging problems will aid the reader far more
than aphorisms.”
Asking how one does mathematical research is like asking how a com-
poser creates a masterpiece. No one really knows. However, it is clear
that some preparation, some form of training, is essential for it. Jacques
Hadamard, in his book The Mathematician’s Mind, proposes four stages
in the process of creation: preparation, incubation, illumination, and ver-
ification. The preparation is the conscious attention and hard work on a
problem. This conscious attention sets in motion an unconscious mecha-
nism that searches for a solution. Henri Poincaré compared ideas to atoms
that are set in motion by continued thought. The dance of these ideas in the
crucible of the mind leads to certain “stable combinations” that give rise
to flashes of illumination, which is the third stage. Finally, one must verify
the flash of insight, formulate it precisely, and subject it to the standards
of mathematical rigor.
This book arose when a student approached me for a reading course on
algebraic number theory. I had been thinking of writing a problem book on
algebraic number theory and I took the occasion to carry out an experiment.
I told the student to round up more students who may be interested and so
she recruited eight more. Each student would be responsible for one chapter
of the book. I lectured for about an hour a week stating and sketching the
solution of each problem. The student was then to fill in the details, add
ten more problems and solutions, and then typeset it into TEX. Chapters 1
to 8 arose in this fashion. Chapters 9 and 10 as well as the supplementary
problems were added afterward by the instructor.
Some of these problems are easy and straightforward. Some of them
are difficult. However, they have been arranged with a didactic purpose.
It is hoped that the book is suitable for independent study. From this
perspective, the book can be described as a first course in algebraic number
theory and can be completed in one semester.
Our approach in this book is based on the principle that questions focus
the mind. Indeed, quest and question are cognates. In our quest for truth,
for understanding, we seem to have only one method. That is the Socratic
method of asking questions and then refining them. Grappling with such
problems and questions, the mind is strengthened. It is this exercise of the
mind that is the goal of this book, its raison d’être. If even one individual
benefits from our endeavor, we will feel amply rewarded.

Kingston, Ontario Ram Murty

August 1998
Acknowledgments

We would like to thank the students who helped us in the writing of this
book: Kayo Shichino, Ian Stewart, Bridget Gilbride, Albert Chau, Sindi
Sabourin, Tai Huy Ha, Adam Van Tuyl and Satya Mohit.
We would also like to thank NSERC for financial support of this project
as well as Queen’s University for providing a congenial atmosphere for this
task.

J.E.
M.R.M.
August 1998

xi
Contents

Preface to the Second Edition vii

Preface to the First Edition ix

Acknowledgments xi

I Problems
1 Elementary Number Theory 3
1.1 Integers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Applications of Unique Factorization . . . . . . . . . . . . . 8
1.3 The ABC Conjecture . . . . . . . . . . . . . . . . . . . . . 9
1.4 Supplementary Problems . . . . . . . . . . . . . . . . . . . . 10

2 Euclidean Rings 13
2.1 Preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.2 Gaussian Integers . . . . . . . . . . . . . . . . . . . . . . . . 17
2.3 Eisenstein Integers . . . . . . . . . . . . . . . . . . . . . . . 17
2.4 Some Further Examples . . . . . . . . . . . . . . . . . . . . 21
2.5 Supplementary Problems . . . . . . . . . . . . . . . . . . . . 25

3 Algebraic Numbers and Integers 27

3.1 Basic Concepts . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.2 Liouville’s Theorem and Generalizations . . . . . . . . . . . 29
3.3 Algebraic Number Fields . . . . . . . . . . . . . . . . . . . . 32
3.4 Supplementary Problems . . . . . . . . . . . . . . . . . . . . 38

4 Integral Bases 41
4.1 The Norm and the Trace . . . . . . . . . . . . . . . . . . . . 41
4.2 Existence of an Integral Basis . . . . . . . . . . . . . . . . . 43
4.3 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.4 Ideals in OK . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
4.5 Supplementary Problems . . . . . . . . . . . . . . . . . . . . 50

xiii
xiv CONTENTS

5 Dedekind Domains 53
5.1 Integral Closure . . . . . . . . . . . . . . . . . . . . . . . . . 53
5.2 Characterizing Dedekind Domains . . . . . . . . . . . . . . 55
5.3 Fractional Ideals and Unique Factorization . . . . . . . . . . 57
5.4 Dedekind’s Theorem . . . . . . . . . . . . . . . . . . . . . . 63
5.5 Factorization in OK . . . . . . . . . . . . . . . . . . . . . . 65
5.6 Supplementary Problems . . . . . . . . . . . . . . . . . . . . 66

6 The Ideal Class Group 69

6.1 Elementary Results . . . . . . . . . . . . . . . . . . . . . . . 69
6.2 Finiteness of the Ideal Class Group . . . . . . . . . . . . . . 71
6.3 Diophantine Equations . . . . . . . . . . . . . . . . . . . . . 73
6.4 Exponents of Ideal Class Groups . . . . . . . . . . . . . . . 75
6.5 Supplementary Problems . . . . . . . . . . . . . . . . . . . . 76

7 Quadratic Reciprocity 81
7.1 Preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . . 81
7.2 Gauss Sums . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
7.3 The Law of Quadratic Reciprocity . . . . . . . . . . . . . . 86
7.4 Quadratic Fields . . . . . . . . . . . . . . . . . . . . . . . . 88
7.5 Primes in Special Progressions . . . . . . . . . . . . . . . . 91
7.6 Supplementary Problems . . . . . . . . . . . . . . . . . . . . 94

8 The Structure of Units 99

8.1 Dirichlet’s Unit Theorem . . . . . . . . . . . . . . . . . . . 99
8.2 Units in Real Quadratic Fields . . . . . . . . . . . . . . . . 108
8.3 Supplementary Problems . . . . . . . . . . . . . . . . . . . . 115

9 Higher Reciprocity Laws 117

9.1 Cubic Reciprocity . . . . . . . . . . . . . . . . . . . . . . . 117
9.2 Eisenstein Reciprocity . . . . . . . . . . . . . . . . . . . . . 122
9.3 Supplementary Problems . . . . . . . . . . . . . . . . . . . . 125

10 Analytic Methods 127

10.1 The Riemann and Dedekind Zeta Functions . . . . . . . . . 127
10.2 Zeta Functions of Quadratic Fields . . . . . . . . . . . . . . 130
10.3 Dirichlet’s L-Functions . . . . . . . . . . . . . . . . . . . . . 133
10.4 Primes in Arithmetic Progressions . . . . . . . . . . . . . . 134
10.5 Supplementary Problems . . . . . . . . . . . . . . . . . . . . 136

11 Density Theorems 139

11.1 Counting Ideals in a Fixed Ideal Class . . . . . . . . . . . . 139
11.2 Distribution of Prime Ideals . . . . . . . . . . . . . . . . . . 146
11.3 The Chebotarev density theorem . . . . . . . . . . . . . . . 150
11.4 Supplementary Problems . . . . . . . . . . . . . . . . . . . . 153
CONTENTS xv

II Solutions

1 Elementary Number Theory 159

1.1 Integers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
1.2 Applications of Unique Factorization . . . . . . . . . . . . . 166
1.3 The ABC Conjecture . . . . . . . . . . . . . . . . . . . . . 170
1.4 Supplementary Problems . . . . . . . . . . . . . . . . . . . . 173

2 Euclidean Rings 179

2.1 Preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . . 179
2.2 Gaussian Integers . . . . . . . . . . . . . . . . . . . . . . . . 181
2.3 Eisenstein Integers . . . . . . . . . . . . . . . . . . . . . . . 185
2.4 Some Further Examples . . . . . . . . . . . . . . . . . . . . 187
2.5 Supplementary Problems . . . . . . . . . . . . . . . . . . . . 189

3 Algebraic Numbers and Integers 197

3.1 Basic Concepts . . . . . . . . . . . . . . . . . . . . . . . . . 197
3.2 Liouville’s Theorem and Generalizations . . . . . . . . . . . 198
3.3 Algebraic Number Fields . . . . . . . . . . . . . . . . . . . . 199
3.4 Supplementary Problems . . . . . . . . . . . . . . . . . . . . 202

4 Integral Bases 207

4.1 The Norm and the Trace . . . . . . . . . . . . . . . . . . . . 207
4.2 Existence of an Integral Basis . . . . . . . . . . . . . . . . . 208
4.3 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
4.4 Ideals in OK . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
4.5 Supplementary Problems . . . . . . . . . . . . . . . . . . . . 214

5 Dedekind Domains 227

5.1 Integral Closure . . . . . . . . . . . . . . . . . . . . . . . . . 227
5.2 Characterizing Dedekind Domains . . . . . . . . . . . . . . 228
5.3 Fractional Ideals and Unique Factorization . . . . . . . . . . 229
5.4 Dedekind’s Theorem . . . . . . . . . . . . . . . . . . . . . . 233
5.5 Factorization in OK . . . . . . . . . . . . . . . . . . . . . . 234
5.6 Supplementary Problems . . . . . . . . . . . . . . . . . . . . 235

6 The Ideal Class Group 245

6.1 Elementary Results . . . . . . . . . . . . . . . . . . . . . . . 245
6.2 Finiteness of the Ideal Class Group . . . . . . . . . . . . . . 245
6.3 Diophantine Equations . . . . . . . . . . . . . . . . . . . . . 247
6.4 Exponents of Ideal Class Groups . . . . . . . . . . . . . . . 248
6.5 Supplementary Problems . . . . . . . . . . . . . . . . . . . . 250
xvi CONTENTS

7 Quadratic Reciprocity 263

7.1 Preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . . 263
7.2 Gauss Sums . . . . . . . . . . . . . . . . . . . . . . . . . . . 266
7.3 The Law of Quadratic Reciprocity . . . . . . . . . . . . . . 267
7.4 Quadratic Fields . . . . . . . . . . . . . . . . . . . . . . . . 270
7.5 Primes in Special Progressions . . . . . . . . . . . . . . . . 270
7.6 Supplementary Problems . . . . . . . . . . . . . . . . . . . . 272

8 The Structure of Units 279

8.1 Dirichlet’s Unit Theorem . . . . . . . . . . . . . . . . . . . 279
8.2 Units in Real Quadratic Fields . . . . . . . . . . . . . . . . 284
8.3 Supplementary Problems . . . . . . . . . . . . . . . . . . . . 291

9 Higher Reciprocity Laws 299

9.1 Cubic Reciprocity . . . . . . . . . . . . . . . . . . . . . . . 299
9.2 Eisenstein Reciprocity . . . . . . . . . . . . . . . . . . . . . 303
9.3 Supplementary Problems . . . . . . . . . . . . . . . . . . . . 308

10 Analytic Methods 313

10.1 The Riemann and Dedekind Zeta Functions . . . . . . . . . 313
10.2 Zeta Functions of Quadratic Fields . . . . . . . . . . . . . . 316
10.3 Dirichlet’s L-Functions . . . . . . . . . . . . . . . . . . . . . 320
10.4 Primes in Arithmetic Progressions . . . . . . . . . . . . . . 322
10.5 Supplementary Problems . . . . . . . . . . . . . . . . . . . . 324

11 Density Theorems 333

11.1 Counting Ideals in a Fixed Ideal Class . . . . . . . . . . . . 333
11.2 Distribution of Prime Ideals . . . . . . . . . . . . . . . . . . 337
11.3 The Chebotarev density theorem . . . . . . . . . . . . . . . 340
11.4 Supplementary Problems . . . . . . . . . . . . . . . . . . . . 341

Bibliography 347

Index 349
 Part I

Problems
Chapter 1

Elementary Number
Theory

1.1 Integers
The nineteenth century mathematician Leopold Kronecker wrote that “all
results of the profoundest mathematical investigation must ultimately be
expressible in the simple form of properties of integers.” It is perhaps this
feeling that made him say “God made the integers, all the rest is the work
of humanity” [B, pp. 466 and 477].
In this section, we will state some properties of integers. Primes, which
are integers with exactly two positive divisors, are very important in number
theory. Let Z represent the set of integers.

Theorem 1.1.1 If a, b are relatively prime, then we can find integers x, y

such that ax + by = 1.
Proof. We write a = bq + r by the Euclidean algorithm, and since a, b
are relatively prime we know r != 0 so 0 < r < |b|. We see that b, r are
relatively prime, or their common factor would have to divide a as well.
So, b = rq1 + r1 with 0 < r1 < |r|. We can then write r = r1 q2 + r2 , and
continuing in this fashion, we will eventually arrive at rk = 1 for some k.
Working backward, we see that 1 = ax + by for some x, y ∈ Z. !
Remark. It is convenient to observe that
! " ! "! " ! "! "! "
a q 1 b q 1 q1 1 r
= =
b 1 0 r 1 0 1 0 r1
! "! " ! "! "
q 1 q1 1 qk 1 rk−1
= ···
1 0 1 0 1 0 rk
! "
r
= A k−1 ,
rk

3
4 CHAPTER 1. ELEMENTARY NUMBER THEORY

(say). Notice det A = ±1 and A−1 has integer entries whose bottom row
gives x, y ∈ Z such that ax + by = 1.

Theorem 1.1.2 Every positive integer greater than 1 has a prime divisor.
Proof. Suppose that there is a positive integer having no prime divisors.
Since the set of positive integers with no prime divisors is nonempty, there
is a least positive integer n with no prime divisors. Since n divides itself,
n is not prime. Hence we can write n = ab with 1 < a < n and 1 < b < n.
Since a < n, a must have a prime divisor. But any divisor of a is also
a divisor of n, so n must have a prime divisor. This is a contradiction.
Therefore every positive integer has at least one prime divisor. !

Theorem 1.1.3 There are infinitely many primes.

Proof. Suppose that there are only finitely many primes, that is, suppose
that there is no prime greater than n where n is an integer. Consider the
integer a = n! + 1 where n ≥ 1. By Theorem 1.1.2, a has at least one prime
divisor, which we denote by p. If p ≤ n, then p | n! and p | (a − n!) = 1.
This is impossible. Hence p > n. Therefore we can see that there is a prime
greater than n for every positive integer n. Hence there are infinitely many
primes. !

Theorem 1.1.4 If p is prime and p | ab, then p | a or p | b.

Proof. Suppose that p is prime and p | ab where a and b are integers. If
p does not divide a, then a and p are coprime. Then ∃x, y ∈ Z such that
ax + py = 1. Then we have abx + pby = b and pby = b − abx. Hence
p | b − abx. Thus p | b. Similarly, if p does not divide b, we see that p | a. !

Theorem 1.1.5 Z has unique factorization.

Proof.
Existence. Suppose that there is an integer greater than 1 which can-
not be written as a product of primes. Then there exists a least integer
m with such a property. Since m is not a prime, m has a positive divi-
sor d such that m = de where e is an integer and 1 < d < m, 1 < e < m.
Since m is the least integer which cannot be written as a product of primes,
we can write d and e as products of primes such that d = p1 p2 · · · pr and
e = q1 q2 · · · qs . Hence m = de = p1 p2 · · · pr · q1 q2 · · · qs . This contradicts
our assumption about m. Hence all integers can be written as products of
primes.
Uniqueness. Suppose that an integer a is written as
a = p1 · · · pr = q 1 · · · q s ,
where pi and qj are primes for 1 ≤ i ≤ r, 1 ≤ j ≤ s. Then p1 | q1 · · · qs ,
so there exists qj such that p1 | qj for some j. Without loss of generality,
1.1. INTEGERS 5

we can let qj be q1 . Since p1 is a prime, we see that p1 = q1 . Dividing

p1 · · · pr = q1 · · · qs by p1 = q1 , we have p2 · · · pr = q2 · · · qs . Similarly there
exists qj such that p2 | qj for some j. Let qj be q2 . Then q2 = p2 . Hence
there exist q1 , . . . , qr such that pi = qj for 1 ≤ i ≤ r and r ≤ s. If r < s,
then we see that 1 = qr+1 · · · qs . This is impossible. Hence r = s. Therefore
the factorization is unique. !

Example 1.1.6 Show that

1 1
S =1+ + ··· +
2 n
is not an integer for n > 1.

Solution. Let k ∈ Z be the highest power of 2 less than n, so that 2k ≤

n < 2k+1 . Let m be the least common multiple of 1, 2, . . . , n excepting 2k .
Then
m m
mS = m + + ··· + .
2 n
Each of the numbers on the right-hand side of this equation are integers,
except for m/2k . If m/2k were an integer, then 2k would have to divide the
least common multiple of the number 1, 2, . . . , 2k − 1, 2k + 1, . . . , n, which
it does not. So mS is not an integer, which implies that S cannot be an
integer.

Exercise 1.1.7 Show that

1 1 1
1+ + + ··· +
3 5 2n − 1
is not an integer for n > 1.

We can use the same method to prove the following more general result.

Exercise 1.1.8 Let a1 , . . . , an for n ≥ 2 be nonzero integers. Suppose there is a

prime p and positive integer h such that ph | ai for some i and ph does not divide
aj for all j #= i.
Then show that
1 1
S= + ··· +
a1 an
is not an integer.

Exercise 1.1.9 Prove that if n is a composite integer, then n has a prime factor
√
not exceeding n.

Exercise 1.1.10 Show that if the smallest prime factor p of the positive integer
√
n exceeds 3 n, then n/p must be prime or 1.
6 CHAPTER 1. ELEMENTARY NUMBER THEORY

!Exercise
" 1.1.11 Let p be prime. Show that each of the binomial coefficients
p
k
, 1 ≤ k ≤ p − 1, is divisible by p.

Exercise 1.1.12 Prove that if p is an odd prime, then 2p−1 ≡ 1 (mod p).

Exercise 1.1.13 Prove Fermat’s little Theorem: If a, p ∈ Z with p a prime, and

p ! a, prove that ap−1 ≡ 1 (mod p).

For any integer n we define φ(n) to be the number of positive integers

less than n which are coprime to n. This is known as the Euler φ-function.

Theorem 1.1.14 Given a, n ∈ Z, aφ(n) ≡ 1 (mod n) when gcd(a, n) = 1.

This is a theorem due to Euler.
Proof. The case where n is prime is clearly a special case of Fermat’s
little Theorem. The argument is basically the same as that of the alternate
solution to Exercise 1.1.13.
Consider the ring Z/nZ. If a, n are coprime, then a is a unit in this
ring. The units form a multiplicative group of order φ(n), and so clearly
aφ(n) = 1. Thus, aφ(n) ≡ 1 (mod n). !

Exercise 1.1.15 Show that n | φ(an − 1) for any a > n.

Exercise 1.1.16 Show that n ! 2n − 1 for any natural number n > 1.

Exercise 1.1.17 Show that

φ(n) #$ 1
%
= 1−
n p
p|n

by interpreting the left-hand side as the probability that a random number chosen
from 1 ≤ a ≤ n is coprime to n.

Exercise 1.1.18 Show that φ is multiplicative (i.e., φ(mn) = φ(m)φ(n) when

gcd(m, n) = 1) and φ(pα ) = pα−1 (p − 1) for p prime.

Exercise 1.1.19 Find the last two digits of 31000 .

Exercise 1.1.20 Find the last two digits of 21000 .

Let π(x) be the number of primes less than or equal to x. The prime
number theorem asserts that
x
π(x) ∼
log x
as x → ∞. This was first proved in 1896, independently by J. Hadamard
and Ch. de la Vallée Poussin.
We will not prove the prime number theorem here, but derive various
estimates for π(x) by elementary methods.
1.1. INTEGERS 7

Exercise 1.1.21 Let pk denote the kth prime. Prove that

pk+1 ≤ p1 p2 · · · pk + 1.

Exercise 1.1.22 Show that

k
pk < 22 ,
where pk denotes the kth prime.

Exercise 1.1.23 Prove that π(x) ≥ log(log x).

Exercise 1.1.24 By observing that any natural number can be written as sr2
with s squarefree, show that
√
x ≤ 2π(x) .
Deduce that
log x
π(x) ≥ .
2 log 2

&
Exercise 1.1.25 Let ψ(x) = pα ≤x log p where the summation is over prime
powers pα ≤ x.
(i) For 0 ≤ x ≤ 1, show that x(1 − x) ≤ 41 . Deduce that
' 1
1
xn (1 − x)n dx ≤
0 4n

for every natural number n.

(1
(ii) Show that eψ(2n+1) 0 xn (1 − x)n dx is a positive integer. Deduce that

ψ(2n + 1) ≥ 2n log 2.

(iii) Prove that ψ(x) ≥ 12 x log 2 for x ≥ 6. Deduce that

x log 2
π(x) ≥
2 log x

for x ≥ 6.

Exercise 1.1.26 By observing that

)* +
# )
) 2n
p ) ,
) n
n<p≤2n

show that
9x log 2
π(x) ≤
log x
for every integer x ≥ 2.
8 CHAPTER 1. ELEMENTARY NUMBER THEORY

1.2 Applications of Unique Factorization

We begin this section with a discussion of nontrivial solutions to Diophan-

tine equations of the form xl + y m = z n . Nontrivial solutions are those for
which xyz != 0 and (x, y) = (x, z) = (y, z) = 1.

Exercise 1.2.1 Suppose that a, b, c ∈ Z. If ab = c2 and (a, b) = 1, then show

that a = d2 and b = e2 for some d, e ∈ Z. More generally, if ab = cg then a = dg
and b = eg for some d, e ∈ Z.

Exercise 1.2.2 Solve the equation x2 + y 2 = z 2 where x, y, and z are integers

and (x, y) = (y, z) = (x, z) = 1.

Exercise 1.2.3 Show that x4 +y 4 = z 2 has no nontrivial solution. Hence deduce,

with Fermat, that x4 + y 4 = z 4 has no nontrivial solution.

Exercise 1.2.4 Show that x4 − y 4 = z 2 has no nontrivial solution.

Exercise 1.2.5 Prove that if f (x) ∈ Z[x], then f (x) ≡ 0 (mod p) is solvable for
infinitely many primes p.

Exercise 1.2.6 Let q be prime. Show that there are infinitely many primes p so
that p ≡ 1 (mod q).

We will next discuss integers of the form Fn = 22 + 1, which are called

n

the Fermat numbers. Fermat made the conjecture that these integers are
all primes. Indeed, F0 = 3, F1 = 5, F2 = 17, F3 = 257, and F4 = 65537
5
are primes but unfortunately, F5 = 22 + 1 is divisible by 641, and so F5 is
composite. It is unknown if Fn represents infinitely many primes. It is also
unknown if Fn is infinitely often composite.

Exercise 1.2.7 Show that Fn divides Fm − 2 if n is less than m, and from this
deduce that Fn and Fm are relatively prime if m #= n.

n
Exercise 1.2.8 Consider the nth Fermat number Fn = 22 +1. Prove that every
prime divisor of Fn is of the form 2n+1 k + 1.

α
Exercise 1.2.9 Given a natural number n, let n = pα 1
1 · · · pk
k
be its unique
factorization as a product of prime powers. We define the squarefree part of n,
denoted S(n), to be the product of the primes pi for which αi = 1. Let f (x) ∈ Z[x]
be nonconstant and monic. Show that lim inf S(f (n)) is unbounded as n ranges
over the integers.
1.3. THE ABC CONJECTURE 9

1.3 The ABC Conjecture

αk
Given a natural number n, let n = pα 1 · · · pk be its unique factorization
1

as a product of prime powers. Define the radical of n, denoted rad(n), to

be the product p1 · · · pk .
In 1980, Masser and Oesterlé formulated the following conjecture. Sup-
pose we have three mutually coprime integers A, B, C satisfying A+B = C.
Given any ε > 0, it is conjectured that there is a constant κ(ε) such that
# $ # $1+ε
max |A|, |B|, |C| ≤ κ(ε) rad(ABC) .

This is called the ABC Conjecture.

Exercise 1.3.1 Assuming the ABC Conjecture, show that if xyz #= 0 and xn +
y n = z n for three mutually coprime integers x, y, and z, then n is bounded.

[The assertion xn + y n = z n for n ≥ 3 implies xyz = 0 is the celebrated

Fermat’s Last Theorem conjectured in 1637 by the French mathematician
Pierre de Fermat (1601–1665). After a succession of attacks beginning
with Euler, Dirichlet, Legendre, Lamé, and Kummer, and culminating in
the work of Frey, Serre, Ribet, and Wiles, the situation is now resolved, as
of 1995. The ABC Conjecture is however still open.]

Exercise 1.3.2 Let p be an odd prime. Suppose that 2n ≡ 1 (mod p) and

2n #≡ 1 (mod p2 ). Show that 2d #≡ 1 (mod p2 ) where d is the order of 2 (mod p).

Exercise 1.3.3 Assuming the ABC Conjecture, show that there are infinitely
many primes p such that 2p−1 #≡ 1 (mod p2 ).

Exercise 1.3.4 Show that the number of primes p ≤ x for which

2p−1 #≡ 1 (mod p2 )

is ( log x/ log log x, assuming the ABC Conjecture.

In 1909, Wieferich proved that if p is a prime satisfying

2p−1 !≡ 1 (mod p2 ),

then the equation xp +y p = z p has no nontrivial integral solutions satisfying

p ! xyz. It is still unknown without assuming ABC if there are infinitely
many primes p such that 2p−1 !≡ 1 (mod p2 ). (See also Exercise 9.2.15.)
A natural number n is called squarefull (or powerfull) if for every prime
p | n we have p2 | n. In 1976 Erdös [Er] conjectured that we cannot have
three consecutive squarefull natural numbers.

Exercise 1.3.5 Show that if the Erdös conjecture above is true, then there are
infinitely many primes p such that 2p−1 #≡ 1 (mod p2 ).
10 CHAPTER 1. ELEMENTARY NUMBER THEORY

Exercise 1.3.6 Assuming the ABC Conjecture, prove that there are only finitely
many n such that n − 1, n, n + 1 are squarefull.

Exercise 1.3.7 Suppose that a and b are odd positive integers satisfying

rad(an − 2) = rad(bn − 2)

for every natural number n. Assuming ABC, prove that a = b. (This problem is
due to H. Kisilevsky.)

1.4 Supplementary Problems

Exercise 1.4.1 Show that every proper ideal of Z is of the form nZ for some
integer n.

Exercise 1.4.2 An ideal I is called prime if ab ∈ I implies a ∈ I or b ∈ I. Prove

that every prime ideal of Z is of the form pZ for some prime integer p.

Exercise 1.4.3 Prove that if the number of prime Fermat numbers is finite, then
the number of primes of the form 2n + 1 is finite.

Exercise 1.4.4 If n > 1 and an − 1 is prime, prove that a = 2 and n is prime.

Exercise 1.4.5 An integer is called perfect if it is the sum of its divisors. Show
that if 2n − 1 is prime, then 2n−1 (2n − 1) is perfect.

Exercise 1.4.6 Prove that if p is an odd prime, any prime divisor of 2p − 1 is of

the form 2kp + 1, with k a positive integer.

Exercise 1.4.7 Show that there are no integer solutions to the equation x4 −y 4 =
2z 2 .

Exercise 1.4.8 Let p be an odd prime number. Show that the numerator of
1 1 1
1+ + + ··· +
2 3 p−1

is divisible by p.

Exercise 1.4.9 Let p be an odd prime number greater than 3. Show that the
numerator of
1 1 1
1 + + + ··· +
2 3 p−1
is divisible by p2 .

Exercise 1.4.10 (Wilson’s Theorem) Show that n > 1 is prime if and only
if n divides (n − 1)! + 1.
1.4. SUPPLEMENTARY PROBLEMS 11

Exercise 1.4.11 For each n > 1, let Q be the product of all numbers a < n
which are coprime to n. Show that Q ≡ ±1 (mod n).

Exercise 1.4.12 In the previous exercise, show that Q ≡ 1 (mod n) whenever

n is odd and has at least two prime factors.

Exercise 1.4.13 Use Exercises 1.2.7 and 1.2.8 to show that there are infinitely
many primes ≡ 1 (mod 2r ) for any given r.

Exercise 1.4.14 Suppose p is an odd prime such that 2p + 1 = q is also prime.

Show that the equation
xp + 2y p + 5z p = 0
has no solutions in integers.

Exercise 1.4.15 If x and y are coprime integers, show that if

xp + y p
(x + y) and
x+y
have a common prime factor, it must be p.

Exercise 1.4.16 (Sophie Germain’s Trick) Let p be a prime such that 2p +

1 = q > 3 is also prime. Show that
xp + y p + z p = 0
has no integral solutions with p ! xyz.

Exercise 1.4.17 Assuming ABC, show that there are only finitely many con-
secutive cubefull numbers.

Exercise 1.4.18 Show that ,1

= +∞,
p
p
where the summation is over prime numbers.

Exercise 1.4.19 (Bertrand’s Postulate) (a) If a0 ≥ a1 ≥ a2 ≥ · · · is a de-

creasing sequence of real numbers tending to 0, show that
∞
,
(−1)n an ≤ a0 − a1 + a2 .
n=0
&
(b) Let T (x) = n≤x ψ(x/n), where ψ(x) is defined as in Exercise 1.1.25. Show
that
T (x) = x log x − x + O(log x).
(c) Show that
-x. , -x.
T (x) − 2T = (−1)n−1 ψ = (log 2)x + O(log x).
2 n
n≤x

Deduce that -x.

ψ(x) − ψ ≥ 13 (log 2)x + O(log x).
2
Chapter 2

Euclidean Rings

2.1 Preliminaries
We can discuss the concept of divisibility for any commutative ring R with
identity. Indeed, if a, b ∈ R, we will write a | b (a divides b) if there exists
some c ∈ R such that ac = b. Any divisor of 1 is called a unit. We will
say that a and b are associates and write a ∼ b if there exists a unit u ∈ R
such that a = bu. It is easy to verify that ∼ is an equivalence relation.
Further, if R is an integral domain and we have a, b != 0 with a | b and
b | a, then a and b must be associates, for then ∃c, d ∈ R such that ac = b
and bd = a, which implies that bdc = b. Since we are in an integral domain,
dc = 1, and d, c are units.
We will say that a ∈ R is irreducible if for any factorization a = bc, one
of b or c is a unit.

Example 2.1.1 Let R be an integral domain. Suppose there is a map

n : R → N such that:
(i) n(ab) = n(a)n(b) ∀a, b ∈ R; and
(ii) n(a) = 1 if and only if a is a unit.
We call such a map a norm map, with n(a) the norm of a. Show that every
element of R can be written as a product of irreducible elements.
Solution. Suppose b is an element of R. We proceed by induction on the
norm of b. If b is irreducible, then we have nothing to prove, so assume that
b is an element of R which is not irreducible. Then we can write b = ac
where neither a nor c is a unit. By condition (i),
n(b) = n(ac) = n(a)n(c)
and since a, c are not units, then by condition (ii), n(a) < n(b) and n(c) <
n(b).

13
14 CHAPTER 2. EUCLIDEAN RINGS

If a, c are irreducible, then we are finished. If not, their norms are

smaller than the norm of b, and so by induction we can write them as
products of irreducibles, thus finding an irreducible decomposition of b.
√
Exercise 2.1.2 Let D be squarefree. Consider R = Z[ D]. Show that every
element of R can be written as a product of irreducible elements.
√ √ √
Exercise 2.1.3 Let R = Z[ −5]. Show that 2, 3, 1 + −5, and 1 − −5 are
irreducible in R, and that they are not associates.
√ √
We now observe that 6 = 2 · 3 = (1 + −5)(1 − −5), so that R does
not have unique factorization into irreducibles.
We will say that R, an integral domain, is a unique factorization domain
if:

(i) every element of R can be written as a product of irreducibles; and

(ii) this factorization is essentially unique in the sense that if a = π1 · · · πr

and a = τ1 · · · τs , then r = s and after a suitable permutation, πi ∼ τi .

Exercise 2.1.4 Let R be a domain satisfying (i) above. Show that (ii) is equiv-
alent to (ii$ ): if π is irreducible and π divides ab, then π | a or π | b.

An ideal I ⊆ R is called principal if it can be generated by a single

element of R. A domain R is then called a principal ideal domain if every
ideal of R is principal.

Exercise 2.1.5 Show that if π is an irreducible element of a principal ideal

domain, then (π) is a maximal ideal, (where (x) denotes the ideal generated by
the element x).

Theorem 2.1.6 If R is a principal ideal domain, then R is a unique fac-

torization domain.

Proof. Let S be the set of elements of R that cannot be written as a

product of irreducibles. If S is nonempty, take a1 ∈ S. Then a1 is not
irreducible, so we can write a1 = a2 b2 where a2 , b2 are not units. Then
(a1 ) " (a2 ) and (a1 ) " (b2 ). If both a2 , b2 ∈
/ S, then we can write a1 as
a product of irreducibles, so we assume that a2 ∈ S. We can inductively
proceed until we arrive at an infinite chain of ideals,

(a1 ) " (a2 ) " (a3 ) " · · · " (an ) " · · · .

%∞
Now consider I = i=1 (ai ). This is an ideal of R, and because R is a
principal ideal domain, I = (α) for some α ∈ R. Since α ∈ I, α ∈ (an ) for
some n, but then (an ) = (an+1 ). From this contradiction, we conclude that
the set S must be empty, so we know that if R is a principal ideal domain,
2.1. PRELIMINARIES 15

every element of R satisfies the first condition for a unique factorization

domain.
Next we would like to show that if we have an irreducible element π,
and π | ab for a, b ∈ R, then π | a or π | b. If π ! a, then the ideal (a, π) = R,
so ∃x, y such that

ax + πy = 1,
⇒ abx + πby = b.

Since π | abx and π | πby then π | b, as desired. By Exercise 2.1.4, we have

shown that R is a unique factorization domain. !

The following theorem describes an important class of principal ideal

domains:

Theorem 2.1.7 If R is a domain with a map φ : R → N, and given

a, b ∈ R, ∃q, r ∈ R such that a = bq + r with r = 0 or φ(r) < φ(b), we
call R a Euclidean domain. If a ring R is Euclidean, it is a principal ideal
domain.

Proof. Given an ideal I ⊆ R, take an element a of I such that φ(a) is

minimal among elements of I. Then given b ∈ I, we can find q, r ∈ R such
that b = qa + r where r = 0 or φ(r) < φ(a). But then r = b − qa, and so
r ∈ I, and φ(a) is minimal among the norms of elements of I. So r = 0,
and given any element b of I, b = qa for some q ∈ R. Therefore a is a
generator for I, and R is a principal ideal domain. !

Exercise 2.1.8 If F is a field, prove that F [x], the ring of polynomials in x with
coefficients in F , is Euclidean.

The following result, called Gauss’ lemma, allows us to relate factoriza-

tion of polynomials in Z[x] with the factorization in Q[x]. More generally,
if R is a unique factorization domain and K is its field of fractions, we will
relate factorization of polynomials in R[x] with that in K[x].

Theorem 2.1.9 If R is a unique factorization domain, and f (x) ∈ R[x],

define the content of f to be the gcd of the coefficients of f , denoted by
C(f ). For f (x), g(x) ∈ R[x], C(f g) = C(f )C(g).

Proof. Consider two polynomials f, g ∈ R[x], with C(f ) = c and C(g) = d.

Then we can write

f (x) = ca0 + ca1 x + · · · + can xn

and
g(x) = db0 + db1 x + · · · + dbm xm ,
16 CHAPTER 2. EUCLIDEAN RINGS

where c, d, ai , bj ∈ R, an , bm != 0. We define a primitive polynomial to be a

polynomial f such that C(f ) = 1. Then f = cf ∗ where f ∗ = a0 +a1 x+· · ·+
an xn , a primitive polynomial, and g = dg ∗ , with g ∗ a primitive polynomial.
Since f g = cf ∗ dg ∗ = cd(f ∗ g ∗ ), it will suffice to prove that the product of
two primitive polynomials is again primitive.
Let
f ∗ g ∗ = k0 + k1 x + · · · + km+n xm+n ,
and assume that this polynomial is not primitive. Then all the coefficients
ki are divisible by some π ∈ R, with π irreducible. Since f ∗ and g ∗ are
primitive, we know that there is at least one coefficient in each of f ∗ and
g ∗ that is not divisible by π. We let ai and bj be the first such coefficients
in f ∗ and g ∗ , respectively.
Now,

ki+j = (a0 bi+j + · · · + ai−1 bj+1 ) + ai bj + (ai+1 bj−1 + · · · + ai+j b0 ).

We know that ki+j , a0 , a1 , . . . , ai−1 , b0 , b1 , . . . , bj−1 are all divisible by π, so

ai bj must also be divisible by π. Since π is irreducible, then π | ai or π | bj ,
but we chose these elements specifically because they were not divisible by
π. This contradiction proves that our polynomial f ∗ g ∗ must be primitive.
Then f g = c df ∗ g ∗ where f ∗ g ∗ is a primitive polynomial, thus proving
that C(f g) = cd = C(f )C(g). !

Theorem 2.1.10 If R is a unique factorization domain, then R[x] is a

unique factorization domain.

Proof. Let k be the set of all elements a/b, where a, b ∈ R, and b != 0,

such that a/b = c/d if ad − bc = 0. It is easily verified that k is a field; we
call k the fraction field of R. Let us examine the polynomial ring k[x]. We
showed in Exercise 2.1.8 that k[x] is a Euclidean domain, and we showed in
Theorem 2.1.7 that all Euclidean domains are unique factorization domains.
We shall use these facts later.
First notice that given any nonzero polynomial f (x) ∈ k[x], we can write
this polynomial uniquely (up to multiplication by a unit) as f (x) = cf ∗ (x),
where f ∗ (x) ∈ R[x] and f ∗ (x) is primitive. We do this by first finding a
common denominator for all the coefficients of f and factoring this out. If
we denote this constant by β, then we can write f = f $ /β, where f $ ∈ R[x].
We then find the content of f $ (which we will denote by α), and factor this
out as well. We let α/β = c and write f = cf ∗ , noting that f ∗ is primitive.
We must prove the uniqueness of this expression of f . If

f (x) = cf ∗ (x) = df $ (x),

with both f ∗ (x) and f $ (x) primitive, then we can write

f $ (x) = (c/d)f ∗ (x) = (a/b)f ∗ (x),

2.2. GAUSSIAN INTEGERS 17

where gcd(a, b) = 1. Since the coefficients of f $ (x) are elements of R, then

b | aγi for all i, where γi are the coefficients of f ∗ . But since gcd(a, b) = 1,
then b | γi for all i. Since f ∗ is a primitive polynomial, then b must be
a unit of R. Similarly, we can write f ∗ (x) = (b/a)f $ (x), and by the same
argument as above, a must be a unit as well. This shows that f ∗ (x) ∼ f $ (x).
Let us suppose that we have a polynomial f (x) ∈ R[x]. Then we can
factor this polynomial as f (x) = g(x)h(x), with g(x), h(x) ∈ k[x] (be-
cause k[x] is a unique factorization domain). We can also write cf ∗ (x) =
d1 g ∗ (x)d2 h∗ (x), where g ∗ (x), h∗ (x) ∈ R[x], and g ∗ (x), h∗ (x) are primitive.
We showed above that the polynomial g ∗ (x)h∗ (x) is primitive, and we know
that this decomposition f (x) = cf ∗ (x) is unique. Therefore we can write
f ∗ (x) = g ∗ (x)h∗ (x) and thus f (x) = cg ∗ (x)h∗ (x). But both f (x) and
f ∗ (x) = g ∗ (x)h∗ (x) have coefficients in R, and f ∗ (x) is primitive. So c
must be an element of R.
Thus, when we factored f (x) ∈ k[x], the two factors were also in R[x].
By induction, if we decompose f into all its irreducible components in k[x],
each of the factors will be in R[x], and we know that this decomposition
will be essentially unique because k[x] is a unique factorization domain.
This shows that R[x] is a unique factorization domain. !

2.2 Gaussian Integers

√
Let Z[i] = {a + bi | a, b ∈ Z, i = −1}. This ring is often called the ring of
Gaussian integers.

Exercise 2.2.1 Show that Z[i] is Euclidean.

Exercise 2.2.2 Prove that if p is a positive prime, then there is an element

x ∈ Fp := Z/pZ such that x2 ≡ −1 (mod p) if and only if either p = 2 or p ≡ 1
(mod 4). (Hint: Use Wilson’s theorem, Exercise 1.4.10.)

Exercise 2.2.3 Find all integer solutions to y 2 + 1 = x3 with x, y #= 0.

Exercise 2.2.4 If π is an element of R such that when π | ab with a, b ∈ R, then

π | a or π | b, then we say that π is prime. What are the primes of Z[i]?

Exercise 2.2.5 A positive integer a is the sum of two squares if and only if
a = b2 c where c is not divisible by any positive prime p ≡ 3 (mod 4).

2.3 Eisenstein Integers

√
Let ρ = (−1 + −3)/2. Notice that ρ2 + ρ + 1 = 0, and ρ3 = 1. Notice
also that ρ2 = ρ. Define the Eisenstein integers as the set Z[ρ] = {a + bρ :
a, b ∈ Z}. Notice that Z[ρ] is closed under complex conjugation.
18 CHAPTER 2. EUCLIDEAN RINGS

Exercise 2.3.1 Show that Z[ρ] is a ring.

Exercise 2.3.2 (a) Show that Z[ρ] is Euclidean.

(b) Show that the only units in Z[ρ] are ±1, ±ρ, and ±ρ2 .

Notice that (x2 + x + 1)(x − 1) = x3 − 1 and that we have

(x − ρ)(x − ρ) = (x − ρ)(x − ρ2 ) = x2 + x + 1

so that

(1 − ρ)(1 − ρ2 ) = 3 = (1 + ρ)(1 − ρ)2 = −ρ2 (1 − ρ)2 .

Exercise 2.3.3 Let λ = 1 − ρ. Show that λ is irreducible, so we have a factor-

ization of 3 (unique up to unit).

Exercise 2.3.4 Show that Z[ρ]/(λ) has order 3.

We can apply the arithmetic of Z[ρ] to solve x3 +y 3 +z 3 = 0 for integers

x, y, z. In fact we can show that α3 + β 3 + γ 3 = 0 for α, β, γ ∈ Z[ρ] has no
nontrivial solutions (i.e., where none of the variables is zero).

Example 2.3.5 Let λ = 1 − ρ, θ ∈ Z[ρ]. Show that if λ does not divide θ,

then θ3 ≡ ±1 (mod λ4 ). Deduce that if α, β, γ are coprime to λ, then the
equation α3 + β 3 + γ 3 = 0 has no nontrivial solutions.
Solution. From the previous problem, we know that if λ does not divide
θ then θ ≡ ±1 (mod λ). Set ξ = θ or −θ so that ξ ≡ 1 (mod λ). We write
ξ as 1 + dλ. Then

±(θ3 ∓ 1) = ξ3 − 1
= (ξ − 1)(ξ − ρ)(ξ − ρ2 )
= (dλ)(dλ + 1 − ρ)(1 + dλ − ρ2 )
= dλ(dλ + λ)(dλ − λρ2 )
= λ3 d(d + 1)(d − ρ2 ).

Since ρ2 ≡ 1 (mod λ), then (d − ρ2 ) ≡ (d − 1) (mod λ). We know from

the preceding problem that λ divides one of d, d − 1, and d + 1, so we
may conclude that ξ 3 − 1 ≡ 0 (mod λ4 ), so ξ 3 ≡ 1 (mod λ4 ) and θ ≡ ±1
(mod λ4 ). We can now deduce that no solution to α3 + β 3 + γ 3 = 0 is
possible with α, β, and γ coprime to λ, by considering this equation mod
λ4 . Indeed, if such a solution were possible, then somehow the equation

±1 ± 1 ± 1 ≡ 0 (mod λ4 )

could be satisfied. The left side of this congruence gives ±1 or ±3; certainly
±1 is not congruent to 0 (mod λ4 ) since λ4 is not a unit. Also, ±3 is not
2.3. EISENSTEIN INTEGERS 19

congruent to 0 (mod λ4 ) since λ2 is an associate of 3, and thus λ4 is not.

Thus, there is no solution to α3 + β 3 + γ 3 = 0 if α, β, γ are coprime to λ.

Hence if there is a solution to the equation of the previous example, one

of α, β, γ is divisible by λ. Say γ = λn δ, (δ, λ) = 1. We get α3 +β 3 +δ 3 λ3n =
0, δ, α, β coprime to λ.

Theorem 2.3.6 Consider the more general

α3 + β 3 + ελ3n δ 3 = 0 (2.1)
for a unit ε. Any solution for δ, α, β coprime to λ must have n ≥ 2, but if
(2.1) can be solved with n = m, it can be solved for n = m − 1. Thus, there
are no solutions to the above equation with δ, α, β coprime to λ.
Proof. We know that n ≥ 1 from Example 2.3.5. Considering the equation
mod λ4 , we get that ±1±1±ελ3n ≡ 0 (mod λ4 ). There are two possibilities:
if λ3n ≡ ±2 (mod λ4 ), then certainly n cannot exceed 1; but if n = 1, then
our congruence implies that λ | 2 which is not true. The other possibility
is that λ3n ≡ 0 (mod λ4 ), from which it follows that n ≥ 2.
We may rewrite (2.1) as
−ελ3n δ 3 = α3 + β 3
= (α + β)(α + ρβ)(α + ρ2 β).
We will write these last three factors as A1 , A2 , and A3 for convenience.
We can see that λ6 divides the left side of this equation, since n ≥ 2. Thus
λ6 | A1 A2 A3 , and λ2 | Ai for some i. Notice that
A1 − A2 = λβ,
A1 − A3 = λβρ2 ,
and
A2 − A3 = λβρ.
Since λ divides one of the Ai , it divides them all, since it divides their
differences. Notice, though, that λ2 does not divide any of these differences,
since λ does not divide β by assumption. Thus, the Ai are inequivalent
mod λ2 , and only one of the Ai is divisible by λ2 . Since our equation is
unchanged if we replace β with ρβ or ρ2 β, then without loss of generality
we may assume that λ2 | A1 . In fact, we know that
λ3n−2 | A1 .
Now we write
B1 = A1 /λ,
B2 = A2 /λ,
B3 = A3 /λ.
20 CHAPTER 2. EUCLIDEAN RINGS

We notice that these Bi are pairwise coprime, since if for some prime p, we
had p | B1 and p | B2 , then necessarily we would have

p | B1 − B2 = β

and
p | λB1 + B2 − B1 = α.
This is only possible for a unit p since gcd(α, β) = 1. Similarly, we can
verify that the remaining pairs of Bi are coprime. Since λ3n−2 | A1 , we
have λ3n−3 | B1 . So we may rewrite (2.1) as

−ελ3n−3 δ 3 = B1 B2 B3 .

From this equation we can see that each of the Bi is an associate of a cube,
since they are relatively prime, and we write

B1 = e1 λ3n−3 C13 ,
B2 = e2 C23 ,
B3 = e3 C33 ,

for units ei , and pairwise coprime Ci . Now recall that

A1 = α + β,
A2 = α + ρβ,
A3 = α + ρ2 β.

From these equations we have that

ρ2 A3 + ρA2 + A1 = α(ρ2 + ρ + 1) + β(ρ2 + ρ + 1)

= 0

so we have that
0 = ρ2 λB3 + ρλB2 + λB1
and
0 = ρ2 B3 + ρB2 + B1 .
We can then deduce that

ρ2 e3 C33 + ρe2 C23 + e1 λ3n−3 C13 = 0

so we can find units e4 , e5 so that

C33 + e4 C23 + e5 λ3n−3 C13 = 0.

Considering this equation mod λ3 , and recalling that n ≥ 2, we get that

±1 ± e4 ≡ 0 (mod λ3 ) so e4 = ∓1, and we rewrite our equation as

C33 + (∓C2 )3 + e5 λ3(n−1) C13 = 0.

2.4. SOME FURTHER EXAMPLES 21

This is an equation of the same type as (2.1), so we can conclude that if

there exists a solution for (2.1) with n = m, then there exists a solution
with n = m − 1.
This establishes by descent that no nontrivial solution to (2.1) is possible
in Z[ρ]. !

2.4 Some Further Examples

Example 2.4.1 Solve the equation y 2 + 4 = x3 for integers x, y.

Solution. We first consider the case where y is even. It follows that x must
also be even, which implies that x3 ≡ 0 (mod 8). Now, y is congruent to
0 or 2 (mod 4). If y ≡ 0 (mod 4), then y 2 + 4 ≡ 4 (mod 8), so we can
rule out this case. However, if y ≡ 2 (mod 4), then y 2 + 4 ≡ 0 (mod 8).
Writing y = 2Y with Y odd, and x = 2X, we have 4Y 2 + 4 = 8X 3 , so that

Y 2 + 1 = 2X 3

and
(Y + i)(Y − i) = 2X 3 = (1 + i)(1 − i)X 3 .

We note that Y 2 + 1 ≡ 2 (mod 4) and so X 3 is odd. Now,

(Y + i)(Y − i)
X3 =
(1 + i)(1 − i)
! "! "
1+Y 1−Y 1+Y 1−Y
= + i − i
2 2 2 2
! "2 ! "2
1+Y 1−Y
= + .
2 2

We shall write this last sum as a2 + b2 . Since Y is odd, a and b are integers.
Notice also that a + b = 1 so that gcd(a, b) = 1. We now have that

X 3 = (a + bi)(a − bi).

We would like to establish that (a+bi) and (a−bi) are relatively prime. We
assume there exists some nonunit d such that d | (a + bi) and d | (a − bi).
But then d | [(a + bi) + (a − bi)] = 2a and d | (a + bi) − (a − bi) = 2bi. Since
gcd(a, b) = 1, then d | 2, and thus d must have even norm. But then it is
impossible that d | (a + bi) since the norm of (a + bi) is a2 + b2 = X 3 which
is odd. Thus (a + bi) and (a − bi) are relatively prime, and each is therefore
a cube, since Z[i] is a unique factorization domain. We write

a + bi = (s + ti)3 = s3 − 3st2 + (3s2 t − t3 )i.

22 CHAPTER 2. EUCLIDEAN RINGS

Comparing real and imaginary parts yields

a = s3 − 3st2 ,
b = 3s2 t − t3 .

Adding these two equations yields a+b = s3 −3st2 +3s2 t−t3 . But a+b = 1,
so we have

1 = s3 − 3st2 + 3s2 t − t3
= (s − t)(s2 + 4st + t2 ).

Now, s, t ∈ Z so (s − t) = ±1 and (s2 + 4st + t2 ) = ±1. Subtracting the

second equation from the square of the first we find that −6st = 0 or 2.
Since s and t are integers, we rule out the case −6st = 2 and deduce that
either s = 0 or t = 0. Thus either a = 1, b = 0 or a = 0, b = 1. It follows
that Y = ±1, so the only solutions in Z to the given equation with y even
are x = 2, y = ±2.
Next, we consider the case where y is odd. We write x3 = (y+2i)(y−2i).
We can deduce that (y + 2i) and (y − 2i) are relatively prime since if d
divided both, d would divide both their sum and their difference, i.e., we
would have d | 2y and d | 4i. But then d would have even norm, and since
y is odd, (y + 2i) has odd norm; thus d does not divide (y + 2i). Hence,
(y + 2i) is a cube; we write

y + 2i = (q + ri)3 = q 3 − 3qr2 + (3q 2 r − r3 )i.

Comparing real and imaginary parts we have that 2 = 3q 2 r − r3 so that

r | 2, and the only values r could thus take are ±1 and ±2. We get that
the only possible pairs (q, r) we can have are (1, 1), (−1, 1), (1, −2), and
(−1, −2). Solving for y, and excluding the cases where y is even, we find
that x = 5, y = ±11 is the only possible solution when y is odd.
√
Exercise 2.4.2 Show that Z[ −2] is Euclidean.

Exercise 2.4.3 Solve y 2 + 2 = x3 for x, y ∈ Z.

Example 2.4.4 Solve y 2 + 1 = xp for an odd prime p, and x, y ∈ Z.

Solution. Notice that the equation y 2 + 1 = x3 from an earlier problem is
a special case of the equation given here. To analyze the solutions of this
equation, we first observe that for odd y, y 2 ≡ 1 (mod 4). Thus x would
need to be even, but then if we consider the equation mod 4 we find that it
cannot be satisfied; y 2 + 1 ≡ 2 (mod 4), while xp ≡ 0 (mod 4). Thus y is
even; it is easy to see that x must be odd. If y = 0, then x = 1 is a solution
for all p. We call this solution a trivial solution; we proceed to investigate
solutions other than the trivial one. Now we write our equation as

(y + i)(y − i) = xp .
2.4. SOME FURTHER EXAMPLES 23

If y != 0, then we note that if some divisor δ were to divide both (y + i) and

(y − i), then it would divide 2i; if δ is not a unit, then δ will thus divide
2, and also y, since y is even. But then it is impossible that δ also divide
y + i since i is a unit. We conclude that (y + i) and (y − i) are relatively
prime when y != 0. Thus (y + i) and (y − i) are both pth powers, and we
may write
(y + i) = e(a + bi)p
for some unit e and integers a and b. We have analyzed the units of Z[i];
they are all powers of i, so we write

(y + i) = ik (a + bi)p .

Now,
(y − i) = (y + i) = (−i)k (a − bi)p .
Thus

(y + i)(y − i) = ik (a + bi)p (−i)k (a − bi)p

= (a2 + b2 )p
= xp ,

and it follows that x = (a2 + b2 ). We know that x is odd, so one of a and b

is even (but not both). We now have that

(y + i) − (y − i) = 2i
= ik (a + bi)p − (−i)k (a − bi)p .

We consider two cases separately:

Case 1. k is odd.
In this case we use the binomial theorem to determine the imaginary
parts of both sides of the above equation. We get

2 = Im[(i)k ((a + bi)p + (a − bi)p )]

  
*p ! " * p ! "
p p
= Im (i)k  ap−j (bi)j + ap−j (−bi)j 
j=0
j j=0
j
* ! "
p
= 2(−1)(k−1)/2 ap−j (b)j (−1)j/2 .
even j,
j
0≤j<p

Thus ! "
* p
(k−1)/2
1 = (−1) p−j
a (b) (−1)
j j/2
.
even j,
j
0≤j<p
24 CHAPTER 2. EUCLIDEAN RINGS

Since a divides every term on the right-hand side of this equation, then
a | 1 and a = ±1. We observed previously that only one of a, b is odd; thus
b is even. We now substitute a = ±1 into the last equation above to get
* ! "
p
±1 = (b)j (−1)j/2
even j,
j
0≤j<p
! " ! " ! "
p p p
= 1 − b2 + b4 − · · · ± bp−1 .
2 4 p−1
If the sign of 1 on the left-hand side of this equality were negative, we would
have that b2 | 2; b is even and in particular b != ±1, so this is impossible.
Thus
! " ! " ! "
p p p
0 = −b2 + b4 − · · · ± bp−1
2 4 p−1
! " ! " ! "
p p p
= − + b2 − · · · ± bp−3 .
2 4 p−1
# $
Now we notice that #2 $| b, so 2 | p2 . If p ≡ 3 (mod 4), then we are
finished because
#p$ 2 ! 2 . Suppose in fact
p
that 2q is the largest power of
2 #dividing
$ 2 . We # shall
$ show that 2
q+1
will then divide every term in
2 p p−3 p
b 4 − ··· ± b p−1 , and this will establish that no b #will $
satisfy our
equation. We consider one of these terms given by (b)j−2 pj , for an even
#p$
value of j; we rewrite this as b2m−2 2m (we are not concerned with the
sign of the term). We see that
! " ! "
p p−2 (p)(p − 1)
=
2m 2m − 2 2m(2m − 1)
! "! "
p−2 p 1
= ,
2m − 2 2 m(2m − 1)
so we are considering a term
! "! "
p−2 p b2m−2
.
2m − 2 2 m(2m − 1)
# $
Now, 2q | p2 by assumption. Recall that b is even; thus 22m−2 | b2m−2 .
Now m ≥ 2; it is easy to see then that 2m − 2 ≥ m, so 22m−2 does not
divide m. Thus when we reduce the fraction
b2m−2
m(2m − 1)
to lowest terms, the numerator is even and the denominator is odd. There-
fore, ! "! "
p−2 p b2m−2
2(2 ) |
q
.
2m − 2 2 m(2m − 1)
2.5. SUPPLEMENTARY PROBLEMS 25
# $ # p $ p−3
Thus 2q+1 divides every term in b2 p4 − · · · ± p−1 b and we deduce that
no value of b can satisfy our equation.
Case 2. k is even.
This case is almost identical to the first case; we mention only the
relevant differences. When we expand

(y + i) − (y − i) = 2i = ik (a + bi)p − (−i)k (a − bi)p

and consider imaginary parts, we get

* ! "
p
1 = (−1)k/2 ap−j (b)j (−1)(j−1)/2 .
j
odd j,
0<j≤p

We are able to deduce that b = ±1; substituting we get the equation

* ! "
(j−1)/2 p
±1 = a (b) (−1)
p−j j
j
odd j,
0<j≤p
! " ! " ! "
p2 4 p p
= 1−a +a − ··· ± ap−1 ,
2 4 p−1

which we can see is identical to the equation we arrived at in Case 1, with b

replaced by a. Thus we can reproduce the proof of Case 1, with b replaced
by a, to establish that there are no nontrivial solutions with k even. We
conclude that the equation y 2 + 1 = xp has no nontrivial solution with
x, y ∈ Z.
√
Exercise 2.4.5 Show that Z[ 2] is Euclidean.
√ √
Exercise 2.4.6 Let ε = 1+ 2. Write εn = un +vn 2. Show that u2n −2vn2 = ±1.
√ √
√ 2] such that 1 < η < 1 √
Exercise 2.4.7 Show that there is no unit η in Z[ + 2.
Deduce that every unit (greater than zero) of Z[ 2] is a power of ε = 1 + 2.

2.5 Supplementary Problems

√
Exercise 2.5.1 Show that R = Z[(1 + −7)/2] is Euclidean.
√
Exercise 2.5.2 Show that Z[(1 + −11)/2] is Euclidean.

Exercise 2.5.3 Find all integer solutions to the equation x2 + 11 = y 3 .

√
Exercise 2.5.4 Prove that Z[ 3] is Euclidean.
√
Exercise 2.5.5 Prove that Z[ 6] is Euclidean.
26 CHAPTER 2. EUCLIDEAN RINGS

√
Exercise 2.5.6 Show that Z[(1 + −19)/2] is not Euclidean for the norm map.
√
Exercise 2.5.7 Prove that Z[ −10] is not a unique factorization domain.
√
Exercise 2.5.8 Show that there are only finitely many rings Z[ d] with d ≡ 2
or 3 (mod 4) which are norm Euclidean.

Exercise 2.5.9 Find all integer solutions of y 2 = x3 + 1.

Exercise 2.5.10 Let x1 , ..., xn be indeterminates. Evaluate the determinant of

the n × n matrix whose (i, j)-th entry is xj−1
i . (This is called the Vandermonde
determinant.)
Chapter 3

Algebraic Numbers and

Integers

3.1 Basic Concepts

A number α ∈ C is called an algebraic number if there exists a polynomial
f (x) = an xn + · · · + a0 such that a0 , . . . , an , not all zero, are in Q and
f (α) = 0. If α is the root of a monic polynomial with coefficients in Z,
we say that α is an algebraic integer . Clearly all algebraic integers are
algebraic numbers. However, the converse is false.
√
Example 3.1.1 Show that 2/3 is an algebraic number but not an alge-
braic integer.
Solution.
√ Consider the polynomial
√ f (x) = 9x2 − 2, which is in Q[x]. Since
f ( 2/3) = 0,√ we know that 2/3 is an algebraic number.
Assume 2/3 is an algebraic integer. Then there exists a monic √ poly-
nomial in Z[x], say g(x) = xn + bn−1 xn−1 + · · · + b0 , which has α = 2/3
as a root. So
/ √ 0n / √ 0n−1
2 2
g(α) = + bn−1 + · · · + b0 = 0,
3 3
√ √
⇒ ( 2)n + bn−1 ( 2)n−1 (3) + · · · + b0 (3)n = 0.
√
If i is odd, ( 2)i is not an integer. So we can separate our equation into
two smaller equations:
* √ i √ *
bi 2 3n−i = 0 ⇒ 2 bi 2(i−1)/2 3n−i = 0
i odd i odd

and * √ i
bi 2 3n−i = 0
i even

27
28 CHAPTER 3. ALGEBRAIC NUMBERS AND INTEGERS

for i = 0, . . . , n. Since 3 | 0, each sum above must be divisible by 3. In

particular, because each summand containing bi , i != n, has a factor of 3, 3
divides the summand containing bn = 1. This tells us that 3 | 2(n−1)/2 if n
is odd, and 3 | 2n/2√if n is even. In either case, this is false and hence we
can conclude that 2/3 is not an algebraic integer.

Exercise 3.1.2 Show that if r ∈ Q is an algebraic integer, then r ∈ Z.

Exercise 3.1.3 Show that if 4 | (d + 1), then

√
−1 ± −d
2
is an algebraic integer.

Theorem 3.1.4 Let α be an algebraic number. There exists a unique poly-

nomial p(x) in Q[x] which is monic, irreducible and of smallest degree, such
that p(α) = 0. Furthermore, if f (x) ∈ Q[x] and f (α) = 0, then p(x) | f (x).
Proof. Consider the set of all polynomials in Q[x] for which α is a root
and pick one of smallest degree, say p(x). If p(x) is not irreducible, it can
be written as a product of two lower degree polynomials in Q[x]: p(x) =
a(x)b(x). However, p(α) = a(α)b(α) = 0 and since C is an integral domain,
either a(α) = 0 or b(α) = 0. But this contradicts the minimality of p(x),
so p(x) must be irreducible.
Suppose there were two such polynomials, p(x) and q(x). By the division
algorithm,
p(x) = a(x)q(x) + r(x),
where a(x), r(x) ∈ Q[x], and either deg(r) = 0 or deg(r) < deg(q). But
p(α) = a(α)q(α) + r(α) = 0 and q(α) = 0 together imply that r(α) = 0.
Because p(x) and q(x) are the smallest degree polynomials with α as a root,
r = 0. So p(x) = a(x)q(x) and a(x) ∈ Q∗ (the set of all nonzero elements
of Q), since deg(p) = deg(q). Thus p(x) is unique up to a constant and so
we may suppose its leading coefficient is 1.
Now suppose f (x) is a polynomial in Q[x] such that f (α) = 0. If p(x)
does not divide f (x) then, since p(x) is irreducible, gcd(p(x), f (x)) = 1. So
we can find a(x), b(x) ∈ Q[x] such that a(x)p(x) + b(x)f (x) = 1. However,
putting x = α yields a contradiction. Thus, p(x) | f (x). !

The degree of p(x) is called the degree of α and is denoted deg(α); p(x)
is called the minimal polynomial of α.
Complex numbers which are not algebraic are called transcendental .
Well before an example of a transcendental number was known, mathe-
maticians were assured of their existence.

Example 3.1.5 Show that the set of algebraic numbers is countable (and
hence the set of transcendental numbers is uncountable).
3.2. LIOUVILLE’S THEOREM AND GENERALIZATIONS 29

Solution. All polynomials in Q[x] have a finite number of roots. The set
of rational numbers, Q, is countable and so the set Q[x] is also countable.
The set of algebraic numbers is the set of all roots of a countable number of
polynomials, each with a finite number of roots. Hence the set of algebraic
numbers is countable.
Since algebraic numbers and transcendental numbers partition the set
of complex numbers, C, which is uncountable, it follows that the set of
transcendental numbers is uncountable.
√
Exercise 3.1.6 Find the minimal polynomial of n where n is a squarefree
integer.
√
Exercise 3.1.7 Find the minimal polynomial of 2/3.

3.2 Liouville’s Theorem and Generalizations

In 1853, Liouville showed that algebraic numbers cannot be too well ap-
proximated by rationals.

Theorem 3.2.1 (Liouville) Given α, a real algebraic number of degree

n != 1, there is a positive constant c = c(α) such that for all rational
numbers p/q, (p, q) = 1 and q > 0, the inequality
1 1
1 1
1α − p 1 > c(α)
1 q1 qn

holds.

Proof. Let f (x) = an xn + an−1 xn−1 + · · · + a0 be ∈ Z[x] whose degree

equals that of α and for which α is a root. (So deg(f ) ≥ 2). Notice that
1 ! "1 1 ! "1
1 1 1 1
1f (α) − f p 1 = 1f p 1
1 q 1 1 q 1
1 ! " ! "n−1 1
1 p
n
p 1
1 1
= 1an + an−1 + · · · + a0 1
1 q q 1
1 1
1 an pn + an−1 pn−1 q + · · · + a0 q n 1
= 11 1
1
qn
1
≥ .
qn
If α = α1 , ..., αn are the roots of f , let M be the maximum of the values
|αi |, 1 ≤ i ≤ n. If |p/q| is greater than 2M , then
1 1
1 1
1α − p 1 ≥ M ≥ M .
1 q1 qn
30 CHAPTER 3. ALGEBRAIC NUMBERS AND INTEGERS

If |p/q| ≤ M , then 1 1
1 p 11
1αi − ≤ 3M
1 q1
so that
1 1
1
1α − p 11 1 1
1 1 ≥ 2n ≥ .
q |an |q n
j=2 |αj − p/q| |an |(3M )n−1 q n

Hence, the theorem holds with

! "
1
c(α) = min M, .
|an |(3M )n−1
!

Using this theorem, Liouville was able to give specific examples of tran-
scendental numbers.

Example 3.2.2 Show that

*∞
1
n=0
10 n!

is transcendental.

Solution. Suppose not, and call the sum α. Look at the partial sum

*k
1 pk
= ,
n=0
10 n! qk

with qk = 10k! . Thus,

1 1 1 ∞ 1
1 1 1 * 1 1
1α − pk 1 = 11 1
1
1 qk 1 1 10n! 1
n=k+1
! "k+2 ! "(k+2)(k+3)
1 1 1
= + + + ···
10(k+1)! 10(k+1)! 10(k+1)!
3 4
1 1 1
≤ 1 + 2 + 3 + ···
10(k+1)! 10 10
! "
1
= S,
10(k+1)!

where S = 1 + 1/102 + 1/103 + · · · , an infinite geometric series which has

a finite sum. So 1 ∞ 1
1 * 1 1 S S
1 1
1 1≤ = k+1 .
1 10n! 1 10(k+1)! qk
n=k+1
3.2. LIOUVILLE’S THEOREM AND GENERALIZATIONS 31

If α were algebraic of degree of n then, by Liouville’s theorem, there exists

a constant c(α) such that
1 1
1 1
1α − pk 1 ≥ c(α) ,
1 qk 1 qkn
so we have
S c(α)
≥ n .
qkk+1 qk
However, we can choose k to be as large as we want to obtain a contradic-
tion. So α is transcendental.

5∞It is−n!
easy to see that this argument can be generalized to show that
n=0 a is transcendental for all positive integers a. We will prove this
fact in the Supplementary Exercises for this chapter.

In 1873, Hermite showed the number e is transcendental and in 1882,

Lindemann proved the transcendency of π. In fact, he showed more gen-
erally that for an algebraic number α, eα is transcendental. This implies
that π is transcendental since eπi = −1.
In 1909, Thue was able to improve Liouville’s inequality. He proved
that if α is algebraic of degree n, then there exists a constant c(α) so that
for all p/q ∈ Q, 1 1
1 1
1α − p 1 ≥ c(α) .
1 q 1 q n/2+1
This theorem has immediate Diophantine applications.

Example 3.2.3 Let f (x, y) be an irreducible polynomial of binary form of

degree n ≥ 3. Assuming Thue’s theorem, show that f (x, y) = m for any
fixed m ∈ Z∗ has only finitely many solutions.
Solution. Suppose f (x, y) = m has infinitely many solutions, and write it
in the form
n
6
f (x, y) = (x − αi y) = m,
i=1
where αi is an algebraic number of degree ≥ 3 ∀i = 1, . . . , n.
Without loss of generality, we can suppose that for an infinite number
of pairs (x, y), we have
1 1 1 1
1x 1 1 1
1 − α1 1 ≤ 1 x − αi 1 for i = 2, . . . , n.
1y 1 1y 1

Further, by the triangle inequality,

1 1 !1 1 1 1"
1 − αi 1 ≥ 1 1 x − αi 1 + 1 x − α1 1
1x 1 1 1 1 1
1y 1 2 y 1 1 1 y 1
1
≥ |αi − α1 | for i = 2, . . . , n.
2
32 CHAPTER 3. ALGEBRAIC NUMBERS AND INTEGERS

Hence,
1 1 1 1
1x 1 1x 1
|f (x, y)| = |y | 1 − α1 1 · · · 1 − αn 11 ,
1 n1 1
y y
1 1
1 x 1
|m| ≥ k|y|n 11 − α1 11 ,
y
1 1
|m| 1x 1
n
≥ 1 − α1 11 ,
1
k|y| y

where
n
6
1
k= |αi − α1 |.
2n−1 i=2

However, by Thue’s theorem, this implies

c m 1 m(ck)−1
≤ ⇔ ≤ .
y n/2+1 ky n y n/2+1 yn

However, for n ≥ 3, this holds for only finitely many (x, y), contradicting
our assumption. Thus f (x, y) has only finitely many solutions.

Over a long series of improvements upon Liouville’s theorem, in 1955

Roth was able to show the inequality can be strengthened to
1 1
1 1
1α − p 1 ≥ c(α, ε) ,
1 q1 q 2+ε

for any ε > 0. This improved inequality gives us a new family of transcen-
dental numbers.
&∞ n
Exercise 3.2.4 Show that n=1 2−3 is transcendental.
&∞
Exercise 3.2.5 Show that, in fact, n=1 2−f (n) is transcendental whenever

f (n + 1)
lim > 2.
n→∞ f (n)

3.3 Algebraic Number Fields

The theory of algebraic number fields is vast and rich. We will collect below
the rudimentary facts of the theory. We begin with

Example 3.3.1 Let α be an algebraic number and define

Q[α] = {f (α) : f ∈ Q[x]},

a subring of C. Show that Q[α] is a field.

3.3. ALGEBRAIC NUMBER FIELDS 33

Solution. Let f be the minimal polynomial of α, and consider the map

φ : Q[x] → Q[α] such that
n
* n
*
i
ai x → ai αi .
i=0 i=0

Notice that
n
* m
*
φ(g) + φ(h) = ai αi + bi αi = φ(g + h)
i=0 i=0

and
/ n 0 m 
* * *
φ(g)φ(h) = ai αi  bj α j  = ai bj αi+j = φ(gh).
i=0 j=0 0≤i+j≤n+m

So φ is a homomorphism. Furthermore, it is clear that ker φ = (f ), the

ideal generated by f (see Theorem 3.1.4). Thus, by the ring homomorphism
theorems,
Q[x]/(f ) 1 Q[α].
Let g be a polynomial in Q[x] such that f does not divide g. From Chapter
2, we know that Q[x] is a Euclidean domain and is therefore also a PID.
We also learned in Chapter 2 that the ideal generated by any irreducible
element in a PID is a maximal ideal. Since f is irreducible, Q[x]/(f ) is a
field and so Q[α] is a field, as desired.
From now on, we will denote Q[α] by Q(α).

A field K ⊆ C is called an algebraic number field if its dimension over

Q is finite. The dimension of K over Q is called the degree of K and is
denoted [K : Q]. Notice that if α is an algebraic number of degree n, then
Q(α) is an algebraic number field of degree n over Q.
Let α and β be algebraic numbers. Q(α, β) is a field since it is the
intersection of all the subfields of C containing Q, α, and β. The intersection
of a finite number of subfields in a fixed field is again a field.

Theorem 3.3.2 (Theorem of the Primitive Element) If α and β are

algebraic numbers, then ∃ θ, an algebraic number, such that Q(α, β) = Q(θ).
Proof. Let f be the minimal polynomial of α and let g be the minimal
polynomial of β. We want to show that we can find λ ∈ Q such that
θ = α + λβ and Q(α, β) = Q(θ). We will denote Q(θ) by L. Clearly
L = Q(θ) ⊆ Q(α, β).
Define φ(x) = f (θ − λx) ∈ L[x]. Notice that φ(β) = f (θ − λβ) =
f (α) = 0. So β is a root of φ. Choose λ ∈ Q in such a way that β is the
only common root of φ and g. This can be done since only a finite number
34 CHAPTER 3. ALGEBRAIC NUMBERS AND INTEGERS

of choices of λ are thus ruled out. So gcd(φ(x), g(x)) = c(x − β), c ∈ C∗ .

Then c(x − β) ∈ L[x] which implies that c, cβ ∈ L, and so β ∈ L.
Now, θ = α + λβ ∈ L which means that α ∈ L. So Q(α, β) ⊆ L = Q(θ).
Thus, we have the desired equality: Q(α, β) = Q(θ). !
This theorem can be generalized quite easily using induction: for a set
α1 , . . . , αn of algebraic numbers, there exists an algebraic number θ such
that Q(α1 , . . . , αn ) = Q(θ). Therefore, any algebraic number field K is
Q(θ) for some algebraic number θ.

Exercise 3.3.3 Let α be an algebraic number and let p(x) be its minimal poly-
nomial. Show that p(x) has no repeated roots.

The roots of the minimal polynomial p(x) of α are called the conjugate
roots or conjugates of α. Thus, if n is the degree of p(x), then α has n
conjugates.

Exercise 3.3.4 Let α, β be algebraic numbers such that β is conjugate to α.

Show that β and α have the same minimal polynomial.

If θ = θ(1) and θ(2) , . . . , θ(n) are the conjugates of θ, then Q(θ(i) ), for
i = 2, . . . , n, is called a conjugate field to Q(θ). Further, the maps θ → θ(i)
are monomorphisms of K = Q(θ) → Q(θ(i) ) (referred to as embeddings of
K into C).
We can partition the conjugates of θ into real roots and nonreal roots
(called complex roots).
K is called a normal extension (or Galois extension) of Q if all the
conjugate fields of K are identical √ to K. For example, any quadratic exten-
sion√ of Q is normal. √ However, Q( 3
2) is not
√ since the two conjugate fields
2 3
Q(ρ 2) and Q(ρ 2) are distinct from Q( 3 2). (Here ρ is a primitive cube
3

root of unity.)
We also define the normal closure of any field K as the extension K̃
of smallest degree containing all the conjugate fields of K. Clearly this is
well-defined for if there were two such fields, K̃1 and K̃2 , then K̃1 ∩ K̃2
would have the same property and have smaller √ degree if K̃
√1 != K̃2 . In the
above example, the normal closure of Q( 3 2) is clearly Q( 3 2, ρ).

Example 3.3.5 Show that Liouville’s theorem holds for α where α is a

complex algebraic number of degree n ≥ 2.

Solution. First we note that if α is algebraic, then so is α (the complex

conjugate of α), since they satisfy the same minimal polynomial. Also,
every element in an algebraic number field is algebraic, since if the field
Q(γ) has degree n over Q, then for any β ∈ Q(γ) the elements 1, β, . . . , β n
are surely linearly dependent. This implies that α + α = 2 Re(α) and
α − α = 2i Im(α) are algebraic, since they are both in the field Q(α, α).
3.3. ALGEBRAIC NUMBER FIELDS 35

We can apply Liouville’s theorem to Re(α) to get a constant c =

c(Re(α)) such that 1 1
1 1
1Re(α) − p 1 ≥ c ,
1 q 1 qm
where Re(α) has degree m. Now,
1 1 7! "2
1 p 11 p
1α − = Re(α) − + (Im(α))2
1 q1 q
1 1
1 p 1
≥ 11Re(α) − 11
q
c
≥ .
qm

To prove the result, it remains only to show that if the degree of α is n,

then the degree of Re(α) ≤ n. Consider the polynomial
n
6
g(x) = [2x − (α(i) + α(i) )],
i=1

where α = α(1) , α(2) , . . . , α(n) are the algebraic conjugates of α. Certainly

Re(α) satisfies this equation, so we must verify that its coefficients are in
Q.
To prove this, we need some Galois Theory. Let f be the minimal poly-
nomial of α over Q, and let F be the splitting field of this polynomial (i.e.,
the normal closure of Q(α)). Recall that f is also the minimal polynomial
of α, and so F contains α(i) and α(i) for i = 1, . . . , n. Consider the Galois
group of F , that is, all automorphisms of F leaving Q fixed. These auto-
morphisms permute the roots of f , which are simply the conjugates of α.
It is easy to see that the coefficients of g(x) will remain unchanged under
a permutation of the α(i) ’s, and so they must lie in the fixed field of the
Galois group, which is Q.
Since Re(α) satisfies a polynomial with coefficients in Q of degree n, it
follows that the minimal polynomial of Re(α) must divide this polynomial,
and so have degree less than or equal to n. This proves Liouville’s theorem
for complex algebraic numbers.

Exercise 3.3.6 Let K = Q(θ) be of degree n over Q. Let ω1 , . . . , ωn be a basis

(j)
of K as a vector space over Q. Show that the matrix Ω = (ωi ) is invertible.

Exercise 3.3.7 Let α be an algebraic number. Show that there exists m ∈ Z

such that mα is an algebraic integer.

Exercise 3.3.8 Show that Z[x] is not (a) Euclidean or (b) a PID.
36 CHAPTER 3. ALGEBRAIC NUMBERS AND INTEGERS

The following theorem gives several characterizations of algebraic inte-

gers. Of these, (c) and (d) are the most useful for they supply us with an
immediate tool to test whether a given number is an algebraic integer or
not.

Theorem 3.3.9 Prove that the following statements are equivalent:

(a) α is an algebraic integer.
(b) The minimal polynomial of α is monic ∈ Z[x].
(c) Z[α] is a finitely generated Z-module.
(d) ∃ a finitely generated Z-submodule M != {0} of C such that αM ⊆ M .
Proof. (a) ⇒ (b) Let f (x) be a monic polynomial in Z[x], f (α) = 0. Let
φ(x) be the minimal polynomial of α.
Recall the definition of primitive polynomials given in Chapter 2: a
polynomial f (x) = an xn + · · · + a0 ∈ Z[x] is said to be primitive if the gcd
of the coefficients of f is 1. In particular, a monic polynomial is primitive.
By Theorem 3.1.4, we know f (x) = φ(x)ψ(x), for some ψ(x) ∈ Q[x]. By
the proof of Theorem 2.1.10, we know we can write
a
φ(x) = φ1 (x), φ1 (x) primitive, a, b ∈ Z, φ1 (x) ∈ Z[x],
b
c
ψ(x) = ψ1 (x), ψ1 (x) primitive, c, d ∈ Z, ψ1 (x) ∈ Z[x].
d
So b df (x) = acφ1 (x)ψ1 (x). But by Gauss’ lemma (see Theorem 2.1.9),
φ1 (x)ψ1 (x) is primitive, and f (x) is primitive, so bd = ±ac and f (x) =
±φ1 (x)ψ1 (x). Thus the leading term of both φ1 (x) and ψ1 (x) is ±1. Fur-
ther, φ(α) = 0 ⇒ φ1 (α) = 0. So in fact φ(x) = ±φ1 (x) which is monic in
Z[x].
(b) ⇒ (c) Let φ(x) = xn + an−1 xn−1 + · · · + a0 ∈ Z[x] be the minimal
polynomial of α. Recall Z[α] = {f (α) : f (x) ∈ Z[x]}. In order to prove (c),
it is enough to find a finite basis for Z[α].
Claim: {1, α, . . . , αn−1 } generates Z[α] (as a Z-module).
Proof: It suffices to show that αN , for any N ∈ Z+ , is a linear combi-
nation of {1, α, . . . , αn−1 } with coefficients in Z. We proceed inductively.
Clearly this holds for N ≤ n−1. For N ≥ n, suppose this holds ∀αj , j < N .

αN = αN −n αn
= αN −n [−(a0 + a1 α + · · · + an−1 αn−1 )]
= (−αN −n a0 )1 + (−αN −n a1 )α + · · · + (−αN −n an−1 )αn−1 .

By our inductive hypothesis, −αN −n ai ∈ Z[α] ∀i = 0, 1, . . . , n − 1.

Then Z[α] is a Z-module generated by {1, α, . . . , αn−1 }.
3.3. ALGEBRAIC NUMBER FIELDS 37

(c) ⇒ (d) Let M = Z[α]. Clearly αZ[α] ⊆ Z[α].

(d) ⇒ (a) Let x1 , . . . , xr generate M over Z. So M ⊆ Zx1 + · · · + Zxr .
By assumption αxi ∈ M5 ∀i = 1, . . . , r. It follows that there exists a set of
n
cij ∈ Z such that αxi = j=1 cij xj ∀i = 1, . . . , r. Let C = (cij ). Then
   
x1 x1
 ..   .. 
C  .  = α . ,
xr xr
 
x1
(C − αI)  ...  = 0.
 
⇔
xr
Since not all of x1 , . . . , xr can vanish, det(C − αI) = 0. In other words,
1 1
1c11 − x c12 ··· c1n 11
1
1 c21 c22 − x · · · c2n 11
1
1 .. .. .. 1 = 0 when x = α.
1 . . . 11
1
1 cn1 cn2 · · · cnn − x1
This is a polynomial equation in Z[x] of degree n whose leading coefficient
is (−1)n . Take
:
det(C − xI) for n even,
f (x) =
− det(C − xI) for n odd.
Then f (x) is a monic polynomial in Z[x] such that f (α) = 0. Thus α is an
algebraic integer. !

Example 3.3.10 Let K be an algebraic number field. Let OK be the set

of all algebraic integers in K. Show that OK is a ring.
Solution. From the above theorem, we know that for α, β, algebraic
integers, Z[α], Z[β] are finitely generated Z-modules. Thus M = Z[α, β] is
also a finitely generated Z-module. Moreover,
(α ± β)M ⊆ M,
and
(αβ)M ⊆ M.
So α ± β and αβ are algebraic integers; i.e., α ± β and αβ are in OK . So
OK is a ring.

Exercise 3.3.11 Let f (x) = xn + an−1 xn−1 + · · · + a1 x + a0 , and assume that

for p prime p | ai for 0 ≤ i < k and p2 ! a0 . Show that f (x) has an irreducible
factor of degree at least k. (The case k = n is referred to as Eisenstein’s criterion
for irreducibility.)
38 CHAPTER 3. ALGEBRAIC NUMBERS AND INTEGERS

Exercise 3.3.12 Show that f (x) = x5 + x4 + 3x3 + 9x2 + 3 is irreducible over

Q.

3.4 Supplementary Problems

Exercise 3.4.1 Show that
,∞
1
a n!
n=0

is transcendental for a ∈ Z, a ≥ 2.

Exercise 3.4.2 Show that

,∞
1
a 3n
n=1

is transcendental for a ∈ Z, a ≥ 2.

Exercise 3.4.3 Show that

∞
, 1
n=1
af (n)
is transcendental when
f (n + 1)
lim > 2.
n→∞ f (n)

Exercise 3.4.4 Prove that f (x) = x6 + 7x5 − 12x3 + 6x + 2 is irreducible over

Q.

Exercise 3.4.5 Using Thue’s theorem, show that f (x, y) = x6 +7x5 y −12x3 y 3 +
6xy 5 + 8y 6 = m has only a finite number of solutions for m ∈ Z∗ .

Exercise 3.4.6 Let ζm be a primitive mth root of unity. Show that

# i j
(ζm − ζm ) = (−1)m−1 mm .
0≤i,j≤m−1
i&=j

Exercise 3.4.7 Let # i

φm (x) = (x − ζm )
1≤i≤m
(i,m)=1

denote the mth cyclotomic polynomial. Prove that

#
xm − 1 = φd (x).
d|m

Exercise 3.4.8 Show that φm (x) ∈ Z[x].

Exercise 3.4.9 Show that φm (x) is irreducible in Q[x] for every m ≥ 1.

3.4. SUPPLEMENTARY PROBLEMS 39

Exercise 3.4.10 Let I be a subset of the positive integers ≤ m which are coprime
to m. Set # i
f (x) = (x − ζm ).
i∈I

Suppose that f (ζm ) = 0 and f (ζm

p
)#= 0 for some prime p. Show that p | m. (This
observation gives an alternative proof for the irreducibility of φm (x).)

Exercise 3.4.11 Consider the equation x3 + 3x2 y + xy 2 + y 3 = m. Using Thue’s

theorem, deduce that there are only finitely many integral solutions to this equa-
tion.

Exercise 3.4.12 Assume that n is an odd integer, n ≥ 3. Show that xn +y n = m

has only finitely many integral solutions.

Exercise 3.4.13 Let ζm denote a primitive mth root of unity. Show that Q(ζm )
is normal over Q.

Exercise 3.4.14 Let a be squarefree and greater than 1, and let p be prime.
Show that the normal closure of Q(a1/p ) is Q(a1/p , ζp ).
Chapter 4

Integral Bases

In this chapter, we look more closely at the algebraic structure of OK , the

ring of integers of an algebraic number field K. In particular, we show that
OK is always a finitely generated Z-module admitting a Q-basis for K as a
generating set (where K is viewed as a Q-vector space). We will define the
trace and norm of an element of any number field. We will also define an
important invariant of a number field called the discriminant which arises
in many calculations within the number field. Finally, ideals in the ring of
integers of a number field will be briefly discussed at the end of the chapter.

4.1 The Norm and the Trace

We begin by defining two important rational numbers associated with an
element of an algebraic number field K. Recall that if K is an algebraic
number field, then K can be viewed as a finite-dimensional vector space over
Q. Then if α ∈ K, the map from K to K defined by Φα : v → αv defines a
linear operator on K. We define the trace of α by TrK (α) := Tr(Φα ) and
the norm of α by NK (α) := det(Φα ) (where Tr and det are the usual trace
and determinant of a linear map). We sometimes also use the notation
TrK/Q for TrK and NK/Q for NK .
Thus, to find TrK (α), we choose any Q-basis ω1 , ω2 , . . . , ωn of K and
write
*
αωi = aij ωj ∀ i,

so TrK (α) = TrA and NK (α) = det A where A is the matrix (aij ).

Lemma 4.1.1 If K is an algebraic number field of degree n over Q, and

α ∈ OK its ring of integers, then TrK (α) and NK (α) are in Z.

41
42 CHAPTER 4. INTEGRAL BASES

5n
Proof. We begin by writing αωi = j=1 aij ωj ∀i. Then we have

n
*
(k) (k)
α(k) ωi = aij ωj ∀i, k,
j=1

where α(k) is the kth conjugate of α. We rewrite the above by introducing

the Kronecker delta function to get
n
* n
*
(j) (k)
δjk α(j) ωi = aij ωj ,
j=1 j=1

;
0 if i != j,
where δij = Now, if we define the matrices
1 if i = j.

(j)
A0 = (α(i) δij ), Ω = (ωi ), A = (aij ),

the preceding statement tells us that ΩA0 = AΩ or A0 = Ω−1 AΩ, so we

conclude that TrA = TrA0 and det A = det A0 . But TrA0 is just the sum
of the conjugates of α and is thus (up to sign) the coefficient of the xn−1
term in the minimal polynomial for α; similarly, det A0 is just the product
of the conjugates of α and is thus equal (up to sign) to the constant term
in the minimal polynomial for α. Thus TrK (α) and NK (α) are in Z. !

Exercise 4.1.2 Let K = Q(i). Show that i ∈ OK and verify that TrK (i) and
NK (i) are integers.

√
Exercise 4.1.3 Determine the algebraic integers of K = Q( −5).

Given an algebraic number field K and ω1 , ω2 , . . . , ωn a Q-basis for K,

consider5the correspondence from K to Mn (Q) given by α → (aij ) where
αωi = aij ωj . This is readily seen to give a homomorphism from K to
Mn (Q). From this we can deduce that TrK (·) is in fact a Q-linear map
from K to Q.

Lemma 4.1.4 The bilinear pairing given by B(x, y) : K × K → Q such

that (x, y) → TrK (xy) is nondegenerate.

Proof. We recall that if V is a finite-dimensional vector space over a field

F with basis e1 , e2 , . . . , en and B : V × V → F is a bilinear map, we can
associate a matrix to B as follows. Write
*
v = ai ei with ai ∈ F,
*
u = bi ei with bi ∈ F.
4.2. EXISTENCE OF AN INTEGRAL BASIS 43

Then
*
B(v, u) = B(ai ei , u)
i
*
= ai B(ei , u)
i
*
= ai bj B(ei , ej )
i,j

and we associate to B the matrix (B(ei , ej )). B is said to be nondegenerate

if the matrix associated to it is nonsingular. This definition is independent
of the choice of basis (see Exercise 4.1.5 below).
Now, if ω1 , ω2 , . . . , ωn is a Q-basis for K, then the matrix associated to
B(x, y) with respect to this basis is just

(B(ωi , ωj )) = (TrK (ωi ωj )),

5 (k) (k)
but TrK (ωi ωj ) = ωi ωj and thus we see that

(B(ωi , ωj )) = ΩΩT ,

where Ω is nonsingular because ω1 , ω2 , . . . , ωn form a basis for K. Thus

B(x, y) is indeed nondegenerate. !

Exercise 4.1.5 Show that the definition of nondegeneracy above is independent

of the choice of basis.

4.2 Existence of an Integral Basis

Let K be an algebraic number field of degree n over Q, and OK its ring of
integers. We say that ω1 , ω2 , . . . , ωn is an integral basis for K if ωi ∈ OK
for all i, and OK = Zω1 + Zω2 + · · · + Zωn .

Exercise 4.2.1 Show that ∃ω1∗ , ω2∗ , . . . , ωn∗ ∈ K such that

OK ⊆ Zω1∗ + Zω2∗ + · · · + Zωn∗ .

Theorem 4.2.2 Let α1 , α2 , . . . , αn be a set of generators for a finitely

generated Z-module M , and let N be a submodule.
(a) ∃β1 , β2 , . . . , βm in N with m ≤ n such that

N = Zβ1 + Zβ2 + · · · + Zβm

5
and βi = j≥i pij αj with 1 ≤ i ≤ m and pij ∈ Z.
(b) If m = n, then [M : N ] = p11 p22 · · · pnn .
44 CHAPTER 4. INTEGRAL BASES

Proof. (a) We will proceed by induction on the number of generators of

a Z-module. This is trivial when n = 0. We can assume that we have
proved the above statement to be true for all Z-modules with n − 1 or
fewer generators, and proceed to prove it for n. We define M $ to be the
submodule generated by α2 , α3 , . . . , αn over Z, and define N $ to be N ∩M $ .
Now, if n = 1, then M $ = 0 and there is nothing to prove. If N = N $ , then
the statement is true by our induction hypothesis.
So we assume that N != N $ and consider A, the set of all integers k
such that ∃k2 , k3 , . . . , kn with kα1 + k2 α2 + · · · + kn αn ∈ N . Since N is a
submodule, we deduce that A is a subgroup of Z. All additive subgroups
of Z are of the form mZ for some integer m, and so A = k11 Z for some k11 .
Then let β1 = k11 α1 + k12 α2 + · · · + k1n αn ∈ N . If we have some α ∈ N ,
then
*n
α= hi αi ,
i=1
with hi ∈ Z and h1 ∈ A so h1 = ak11 . Therefore, α − aβ1 ∈ N $ . By the
induction hypothesis, there exist
*
βi = kij αj ,
j≥i

i = 2, 3 . . . , m, which generate N $ over Z and which satisfy all the conditions

above. It is clear that adding β1 to this list gives us a set of generators of
N. 5
(b) Consider α, an arbitrary element of M . Then α = ci αi . Recalling
that *
βi = pij αj ,
j≥i
5 $
we write c1 = p11 q1 + r1 , with 0 ≤ r1 < p11 . Then α − q1 β1 = ci αi
where 0 ≤ c$1 < p11 . Note that α ≡ α − q1 β1 (mod N ). Next we write
c$2 = p22 q2 + r2 , where 0 ≤ r2 < p22 , and note that
α ≡ α − q1 β1 − q2 β2 (mod N ).
It is clear by induction
5 that we can continue this process to arrive at an
expression α$ = ki αi with 0 ≤ ki < pii and α ≡ 5 α$ (mod N ). 5
It remains only to show that if we have α = ci αi and β = di αi
where ci != di for at least one i and 0 ≤ ci , di < pii , then α and β are
distinct mod N . Suppose that this is not true, and that
* *
ci αi ≡ di αi (mod N ),
where 5
ci != di for at least one i. Suppose ci = di for i < r and cr != dr .
Then (ci − di )αi ∈ N , so
 
* * * *
(ci − di )αi = ki βi = ki  pij αj  .
i≥r i≥r i≥r j≥i
4.2. EXISTENCE OF AN INTEGRAL BASIS 45

Since cr , dr are both less than prr , we have cr = dr , a contradiction. Thus,

each coset in M/N has a unique representative
*
α= ci αi ,

with 0 ≤ ci < pii , and there are p11 p22 · · · pnn of them. So [M : N ] =
p11 p22 · · · pnn . !

Exercise 4.2.3 Show that OK has an integral basis.

Exercise 4.2.4 Show that det(Tr(ωi ωj )) is independent of the choice of integral

basis.

We are justified now in making the following definition. If K is an

algebraic number field of degree n over Q, define the discriminant of K as
(j)
dK := det(ωi )2 ,

where ω1 , ω2 , . . . , ωn is an integral basis for K.

Exercise 4.2.5 Show that the discriminant is well-defined. In other words, show
that given ω1 , ω2 , . . . , ωn and θ1 , θ2 , . . . , θn , two integral bases for K, we get the
same discriminant for K.

We can generalize the notion of a discriminant for arbitrary elements of

K. Let K/Q be an algebraic number field, a finite extension of Q of degree
n. Let σ1 , σ2 , . . . , σn be the embeddings of K. For a1 , a2 , . . . , an ∈ K we
< =2
can define dK/Q (a1 , . . . , an ) = det(σi (aj )) .

Exercise 4.2.6 Show that

#! "2
dK/Q (1, a, . . . , an−1 ) = σi (a) − σj (a) .
i>j

We denote dK/Q (1, a, . . . , a n−1

) by dK/Q (a).
&
Exercise 4.2.7 Suppose that ui = n j=1 aij vj with aij ∈ Q, vj ∈ K. Show that
! "2
dK/Q (u1 , u2 , . . . , un ) = det(aij ) dK/Q (v1 , v2 , . . . , vn ).

For a module M with submodule N , we can define the index of N in

M to be the number of elements in M/N , and denote this by [M : N ].
Suppose α is an algebraic integer of degree n, generating a field K. We
define the index of α to be the index of Z + Zα + · · · + Zαn−1 in OK .

Exercise 4.2.8 Let a1 , a2 , . . . , an ∈ OK be linearly independent over Q. Let

N = Za1 + Za2 + · · · + Zan and m = [OK : N ]. Prove that

dK/Q (a1 , a2 , . . . , an ) = m2 dK .
46 CHAPTER 4. INTEGRAL BASES

4.3 Examples
Example 4.3.1 Suppose that the minimal polynomial of α is Eisensteinian
with respect to a prime p, i.e., α is a root of the polynomial

xn + an−1 xn−1 + · · · + a1 x + a0 ,

where p | ai , 0 ≤ i ≤ n − 1 and p2 ! a0 . Show that the index of α is not

divisible by p.
Solution. Let M = Z + Zα + · · · + Zαn−1 . First observe that since

αn + an−1 αn−1 + · · · + a1 α + a0 = 0,

then αn /p ∈ M ⊆ OK . Also, |NK (α)| = a0 !≡ 0 (mod p2 ).

We will proceed by contradiction. Suppose p | [OK : M ]. Then there
is an element of order p in the group OK /M , meaning ∃ξ ∈ OK such that
ξ !∈ M but pξ ∈ M . Then

pξ = b0 + b1 α + · · · + bn−1 αn−1 ,

where not all the bi are divisible by p, for otherwise ξ ∈ M . Let j be the
least index such that p ! bj . Then
! "
b0 b1 bj−1 j−1
η = ξ− + α + ··· + α
p p p
bj j bj+1 j+1 bn
= α + α + · · · + αn
p p p
is in OK , since both ξ and
b0 b1 bn
+ α + · · · + αj−1
p p p
are in OK .
If η ∈ OK , then of course ηαn−j−1 is also in OK , and
bj n−1 αn
ηαn−j−1 = α + (bj+1 + bj+2 α + · · · + bn αn−j−2 ).
p p

Since both αn /p and (bj+1 +bj+2 α+· · ·+bn αn−j−2 ) are in OK , we conclude
that (bj αn−1 )/p ∈ OK .
We know from Lemma 4.1.1 that the norm of an algebraic integer is
always a rational integer, so
! "
bj n−1 bnj NK (α)n−1
NK α =
p pn
bnj an−1
0
=
pn
4.3. EXAMPLES 47

must be an integer. But p does not divide bj , and p2 does not divide a0 , so
this is impossible. This proves that we do not have an element of order p,
and thus p ! [OK : M ].

Exercise 4.3.2 Let m ∈ Z, α ∈ OK . Prove that dK/Q (α + m) = dK/Q (α).

Exercise 4.3.3 Let α be an algebraic integer, and let f (x) be the minimal poly-
n /
nomial of α. If f has degree n, show that dK/Q (α) = (−1)( 2 ) n (i)
i=1 f (α ).
(

√
Example 4.3.4 Let K = Q( D) with D a squarefree integer. Find an
integral basis for OK .
√
Solution. An arbitrary element α of K is of the form α = r1 + r√ 2 D with
r1 , r2 ∈ Q. Since [K : Q] = 2, α has only one conjugate: r1 − r2 D. From
Lemma 4.1.1 we know that if α is an algebraic integer, then TrK (α) = 2r1
and
√ √
NK (α) = (r1 + r2 D)(r1 − r2 D)
= r12 − Dr22

are both integers. We note also that since α satisfies the monic polynomial
x2 − 2r1 x + r12 − Dr22 , if TrK (α) and NK (α) are integers, then α is an
algebraic integer. If 2r1 ∈ Z where r1 ∈ Q, then the denominator of r1 can
be at most 2. We also need r12 −Dr22 to be an integer, so the denominator of
r2 can be no more than 2. Then let r1 = g1 /2, r2 = g2 /2, where g1 , g2 ∈ Z.
The second condition amounts to
g12 − Dg22
∈ Z,
4
which means that g12 − Dg22 ≡ 0 (mod 4), or g12 ≡ Dg22 (mod 4).
We will discuss two cases:
Case 1. D ≡ 1 (mod 4).
If D ≡ 1 (mod 4), and g12 ≡ Dg22 (mod√4), then g1 and g2 are either
both
√ even or both odd. So if α = r1 + r2 D is an algebraic integer of
Q( D), then either r1 and r2 are both integers, or they are both fractions
with denominator 2. √
We recall from Chapter 3 that if 4 | (−D + 1), then
√ (1 + D)/2 is an
algebraic integer. This suggests that we use 1, (1 + D)/2 as a basis; it is
clear from the discussion above that this is in fact an integral basis.
Case 2. D ≡ 2, 3 (mod 4).
If g12 ≡ Dg 2
√ 2 (mod 4), then both g1 and g2 must be even. Then a basis
for OK is 1, D; again it is clear that this is an integral basis.
√
Exercise 4.3.5 √ If D ≡ 1 (mod 4), show that every integer of Q( D) can be
written as (a + b D)/2 where a ≡ b (mod 2).
48 CHAPTER 4. INTEGRAL BASES

Example 4.3.6 Let K = Q(α) where α = r1/3 , r = ab2 ∈ Z where ab is

squarefree. If 3 | r, assume that 3 | a, 3 ! b. Find an integral basis for K.

Solution. The minimal polynomial of α is f (x) = x3 − r, and α’s con-

jugates are α, ωα, and ω 2 α where ω is a primitive cube root of unity. By
Exercise 4.3.3,
63
dK/Q (α) = − f $ (α(i) ) = −33 r2 .
i=1
< =
So −33 r2 = m2 dK where m = OK : Z + Zα + Zα2 . We note that f (x)
is Eisensteinian for every prime divisor of a so by Example 4.3.1 if p | a,
p ! m. Thus if 3 | a, 27a2 | dK , and if 3 ! a, then 3a2 | dK .
We now consider β = α2 /b, which is a root of the polynomial x3 − a2 b.
This polynomial is Eisensteinian for any prime which divides b. Therefore
b2 | dK . We conclude that dK = −3n (ab)2 where n = 3 if 3 | r and n = 1 or
3 otherwise. We will consider three cases: r !≡ 1, 8 (mod 9), r ≡ 1 (mod 9)
and r ≡ 8 (mod 9).
Case 1. If r !≡ 1, 8 (mod 9), then r3 !≡ r (mod 9).
Then the polynomial (x + r)3 − r is Eisensteinian with respect to the
prime 3. A root of this polynomial is α − r and dK/Q (α − r) = dK/Q (α) =
−27r2 . This implies that 3 ! m and so m = b.
We can choose as our integral basis 1, α, α2 /b, all of which are algebraic
integers. We verify that this is an integral basis by checking the index of
Z + Zα + Zα2 in Z + Zα + Zα2 /b, which is clearly b. Thus

α2
OK = Z + Zα + Z .
b

Case 2. If r ≡ 1 (mod 9), then c = (1 + α + α2 )/3 is an algebraic

integer.
In fact, since TrK (α) = TrK (α2 ) = 0, then TrK (c) = 1 ∈ Z and

NK (1 + α + α2 ) NK (α3 − 1) (r − 1)2
NK (c) = = =
27 27NK (α − 1) 27

because the minimal polynomial for 1 − α is x3 + 3x2 + 3x + 1 − r. The

other coefficient for the minimal polynomial of c is (1 − r)/3 which is an
integer. If c is in OK , then OK /(Z + Zα + Zα2 ) has an element of order
3, and so 3 | m. Then dK = −3(ab)2 , so m = 3b. We will choose as our
integral basis α, α2 /b, c, noting that since

α2
1 = 3c − α − b ,
b
α = 0 + α + 0,
α2
α2 = 0+0+b ,
b
4.4. IDEALS IN OK 49

then Theorem 4.2.2 tells us that the index of Z+Zα+Zα2 in Zα+Zα2 /b+Zc
is 3b. Therefore
α2 1 + α + α2
OK = Zα + Z +Z .
b 3
Case 3. If r ≡ 8 (mod 9), consider d = (1 − α + α2 )/3.
This is an algebraic integer. TrK (d) = 1, NK (d) = (1 + r)2 /27 ∈ Z, and
the remaining coefficient for the minimal polynomial of d is (1 + r)/3 ∈ Z.
By the same reasoning as above, we conclude that 3 | m and so m = 3b.
We choose α, α2 /b, d as an integral basis, noting that

α2
1 = 3d + α − b ,
b
α = 0 + α + 0,
α2
α2 = 0+0+b ,
b
so that the index of Z + Zα + Zα2 in Zα + Zα2 /b + Zd is 3b. We conclude
that
α2 1 − α + α2
OK = Zα + Z +Z .
b b
Exercise 4.3.7 Let ζ be any primitive pth root of unity, and K = Q(ζ). Show
that 1, ζ, . . . , ζ p−2 form an integral basis of K.

4.4 Ideals in OK
At this point, we have shown that OK is indeed much like Z in its algebraic
structure. It turns out that we are only halfway to the final step in our
generalization of an integer in a number field. We may think of the ideals
in OK as the most general integers in K, and we remark that when this
set of ideals is endowed with the usual operations of ideal addition and
multiplication, we recover an arithmetic most like that of Z. We prove now
several properties of the ideals in OK .

Exercise 4.4.1 Let a be a nonzero ideal of OK . Show that a ∩ Z #= {0}.

Exercise 4.4.2 Show that a has an integral basis.

Exercise 4.4.3 Show that if a is a nonzero ideal in OK , then a has finite index
in OK .

Exercise 4.4.4 Show that every nonzero prime ideal in OK contains exactly one
integer prime.

We define the norm of a nonzero ideal in OK to be its index in OK . We

will denote the norm of an ideal by N (a).
50 CHAPTER 4. INTEGRAL BASES

Exercise 4.4.5 Let a be an integral ideal with basis α1 , . . . , αn . Show that

(j)
[det(αi )]2 = (N a)2 dK .

4.5 Supplementary Problems

Exercise 4.5.1 Let K be an algebraic number field. Show that dK ∈ Z.

Exercise 4.5.2 Let K/Q be an algebraic number field of degree n. Show that
dK ≡ 0 or 1 (mod 4). This is known as Stickelberger’s criterion.

Exercise 4.5.3 Let f (x) = xn + an−1 xn−1 + · · · + a1 x + a0 with (ai ∈ Z) be

the minimal polynomial of θ. Let K = Q(θ). If for each prime p such that
p2 | dK/Q (θ) we have f (x) Eisensteinian with respect to p, show that OK = Z[θ].

Exercise 4.5.4 If the minimal polynomial of α is f (x) = xn + ax + b, show that

for K = Q(α),
n ! "
dK/Q (α) = (−1)( 2 ) nn bn−1 + an (1 − n)n−1 .

Exercise 4.5.5 Determine an integral basis for K = Q(θ) where θ3 + 2θ + 1 = 0.

Exercise 4.5.6 (Dedekind) Let K = Q(θ) where θ3 − θ2 − 2θ − 8 = 0.

(a) Show that f (x) = x3 − x2 − 2x − 8 is irreducible over Q.
(b) Consider β = (θ2 + θ)/2. Show that β 3 − 3β 2 − 10β − 8 = 0. Hence β is
integral.
(c) Show that dK/Q (θ) = −4(503), and dK/Q (1, θ, β) = −503. Deduce that 1, θ, β
is a Z-basis of OK .
(d) Show that every integer x of K has an even discriminant.
(e) Deduce that OK has no integral basis of the form Z[α].

Exercise 4.5.7 Let m = pa , with p prime and K = Q(ζm ). Show that

(1 − ζm )ϕ(m) = pOK .

Exercise 4.5.8 Let m = pa , with p prime, and K = Q(ζm ). Show that

(−1)ϕ(m)/2 mϕ(m)
dK/Q (ζm ) = .
pm/p

ϕ(m)−1
Exercise 4.5.9 Let m = pa , with p prime. Show that {1, ζm , . . . , ζm } is
an integral basis for the ring of integers of K = Q(ζm ).

Exercise 4.5.10 Let K = Q(ζm ) where m = pa . Show that

(−1)ϕ(m)/2 mϕ(m)
dK = .
pm/p
4.5. SUPPLEMENTARY PROBLEMS 51

Exercise 4.5.11 Show that Z[ζn + ζn−1 ] is the ring of integers of Q(ζn + ζn−1 ),
where ζn denotes a primitive nth root of unity, and n = pα .

Exercise 4.5.12 Let K and L be algebraic number fields of degree m and n,

respectively, over Q. Let d = gcd(dK , dL ). Show that if [KL : Q] = mn, then
OKL ⊆ 1/dOK OL .

Exercise 4.5.13 Let K and L be algebraic number fields of degree m and n,

respectively, with gcd(dK , dL ) = 1. If {α1 , . . . , αm } is an integral basis of OK and
{β1 , . . . , βn } is an integral basis of OL , show that OKL has an integral basis {αi βj }
given that [KL : Q] = mn. (In a later chapter, we will see that gcd(dK , dL ) = 1
implies that [KL : Q] = mn.)
√ √
Exercise 4.5.14 Find an integral basis for Q( 2, −3).

√
Exercise 4.5.15 Let p and q be distinct primes ≡ 1 (mod 4). Let K = Q( p),
√ √ √
L = Q( q). Find a Z-basis for Q( p, q).

Exercise 4.5.16 Let K be an algebraic number field of degree n over Q. Let

a1 , . . . , an ∈ OK be linearly independent over Q. Set

∆ = dK/Q (a1 , . . . , an ).

Show that if α ∈ OK , then ∆α ∈ Z[a1 , . . . , an ].

Exercise 4.5.17 (Explicit Construction of Integral Bases) Suppose K is

an algebraic number field of degree n over Q. Let a1 , . . . , an ∈ OK be linearly
independent over Q and set

∆ = dK/Q (a1 , . . . , an ).

For each i, choose the least natural number dii so that for some dij ∈ Z, the
number
,i
wi = ∆−1 dij aj ∈ OK .
j=1

Show that w1 , . . . , wn is an integral basis of OK .

Exercise 4.5.18 If K is an algebraic number field of degree n over Q and

a1 , . . . , an ∈ OK are linearly independent over Q, then there is an integral basis
w1 , . . . , wn of OK such that

aj = cj1 w1 + · · · + cjj wj ,

cij ∈ Z, j = 1, . . . , n.

Exercise 4.5.19 If Q ⊆ K ⊆ L and K, L are algebraic number fields, show that

dK | dL .
52 CHAPTER 4. INTEGRAL BASES

Exercise 4.5.20 (The Sign of the Discriminant) Suppose K is a number

field with r1 real embeddings and 2r2 complex embeddings so that
r1 + 2r2 = [K : Q] = n
(say). Show that dK has sign (−1)r2 .

Exercise 4.5.21 Show that only finitely many imaginary quadratic fields K are
Euclidean.
√
Exercise 4.5.22 Show that Z[(1 + −19)/2] is not Euclidean. (Recall that in
Exercise 2.5.6 we showed this ring is not Euclidean for the norm map.)

Exercise 4.5.23 (a) Let A = (aij ) be an m × m matrix, B = (bij ) an n × n

matrix. We define the (Kronecker) tensor product A ⊗ B to be the mn × mn
matrix obtained as
 
Ab11 Ab12 · · · Ab1n
 Ab21 Ab22 · · · Ab2n 
 
 . .. ..  ,
 .. . . 
Abn1 Abn2 · · · Abnn
where each block Abij has the form
 
a11 bij a12 bij ··· a1m bij
 a21 bij a22 bij ··· a2m bij 
 
 . .. ..  .
 .. . . 
am1 bij am2 bij ··· amm bij
If C and D are m × m and n × n matrices, respectively, show that
(A ⊗ B)(C ⊗ D) = (AC) ⊗ (BD).

(b) Prove that det(A ⊗ B) = (det A)n (det B)m .

Exercise 4.5.24 Let K and L be algebraic number fields of degree m and n,

respectively, with gcd(dK , dL ) = 1. Show that
dKL = dn m
K · dL .

If we set
log |dM |
δ(M ) = ,
[M : Q]
deduce that δ(KL) = δ(K) + δ(L) whenever gcd(dK , dL ) = 1.

Exercise 4.5.25 Let ζm denote a primitive mth root of unity and let K =
Q(ζm ). Show that OK = Z[ζm ] and
(−1)φ(m)/2 mϕ(m)
dK = / φ(m)/(p−1)
.
p|m p

Exercise 4.5.26 Let K be an algebraic number field. Suppose that θ ∈ OK is

such that dK/Q (θ) is squarefree. Show that OK = Z[θ].
Chapter 5

Dedekind Domains

5.1 Integral Closure

The notion of a Dedekind domain is the concept we need in order to estab-
lish the unique factorization of ideals as a product of prime ideals. En route
to this goal, we will also meet the fundamental idea of a Noetherian ring.
It turns out that Dedekind domains can be studied in the wider context of
Noetherian rings. Even though a theory of factorization of ideals can also
be established for Noetherian rings, we do not pursue it here.

Exercise 5.1.1 Show that a nonzero commutative ring R with identity is a field
if and only if it has no nontrivial ideals.

Theorem 5.1.2 Let R be a commutative ring with identity. Then:

(a) m is a maximal ideal if and only if R/m is a field.

(b) ℘ is a prime ideal if and only if R/℘ is an integral domain.

(c) Let a and b be ideals of R. If ℘ is a prime ideal containing ab, then

℘ ⊇ a or ℘ ⊇ b.

(d) If ℘ is a prime ideal containing the product a1 a2 · · · ar of r ideals of R,

then ℘ ⊇ ai , for some i.

Proof. (a) By the correspondence between ideals of R containing m and

ideals of R/m, R/m has a nontrivial ideal if and only if there is an ideal a
of R strictly between m and R. Thus,

m is maximal,
⇔ R/m has no nontrivial ideals,
⇔ R/m is a field.

53
54 CHAPTER 5. DEDEKIND DOMAINS

(b) ℘ is a prime ideal

⇔ ab ∈ ℘ ⇒ a ∈ ℘ or b ∈ ℘,
⇔ ab + ℘ = 0 + ℘ in R/℘ ⇒ a + ℘ = 0 + ℘ or b + ℘ = 0 + ℘ in R/℘,
⇔ R/℘ has no zero-divisors,
⇔ R/℘ is an integral domain.

(c) Suppose that ℘ ⊇ ab, ℘ !⊇ a. Let a ∈ a, a ∈ / ℘. We know that ab ∈ ℘

for all b ∈ b since ab ⊆ ℘. But, a ∈ / ℘. Thus, b ∈ ℘ for all b ∈ b, since ℘ is
prime. Thus, b ⊆ ℘.
(d) (By induction on r). The base case r = 1 is trivial. Suppose r > 1
and ℘ ⊇ a1 a2 · · · ar . Then from part (c), ℘ ⊇ a1 a2 · · · ar−1 or ℘ ⊇ ar . If
℘ ⊇ a1 a2 · · · ar−1 , then the induction hypothesis implies that ℘ ⊇ ai for
some i ∈ {1, . . . , r − 1}. In either case, ℘ ⊇ ai for some i ∈ {1, . . . , r}. !

Exercise 5.1.3 Show that a finite integral domain is a field.

Exercise 5.1.4 Show that every nonzero prime ideal ℘ of OK is maximal.

Let R be an integral domain. We can always find a field containing R.

As an example of such a field, take Q(R) := {[a, b] : a, b ∈ R, b != 0} such
that we identify elements [a, b] and [c, d] if ad − bc = 0. We define addition
and multiplication on Q(R) by the following rules: [a, b] · [c, d] = [ac, bd]
and [a, b] + [c, d] = [ad + bc, bd].
We can show that this makes Q(R) into a commutative ring with [a, b] ·
[b, a] = 1, for a, b != 0, so that any nonzero element is invertible (i.e., Q(R)
is a field). It contains R in the sense that the map taking a to [a, 1] is a
one-to-one homomorphism from R into Q(R). The field Q(R) is called the
quotient field of R. We will usually write a/b rather than [a, b].
For any field L containing R, we say that α ∈ L is integral over R if α
satisfies a monic polynomial equation f (α) = 0 with f (x) ∈ R[x].
R is said to be integrally closed if every element in the quotient field of
R which is integral over R, already lies in R.

Exercise 5.1.5 Show that every unique factorization domain is integrally closed.

Theorem 5.1.6 For α ∈ C, the following are equivalent:

(1) α is integral over OK ;
(2) OK [α] is a finitely generated OK -module;
(3) There is a finitely generated OK -module M ⊆ C such that αM ⊆ M.
Proof. (1) ⇒ (2) Let α ∈ C be integral over OK . Say α satisfies a monic
polynomial of degree n over OK . Then

OK [α] = OK + OK α + OK α2 + · · · + OK αn−1
5.2. CHARACTERIZING DEDEKIND DOMAINS 55

and so is a finitely generated OK -module.

(2) ⇒ (3) Certainly, αOK [α] ⊆ OK [α], so if OK [α] is a finitely generated
OK -module, then (3) is satisfied with M = OK [α].
(3) ⇒ (1) Let u1 , u2 , . . . , un generate M as an OK -module. Then αui ∈
M for all i = 1, 2, . . . , n since αM ⊆ M. Let
n
*
αui = aij uj
j=1

for i ∈ {1, 2, . . . , n}, aij ∈ OK . Let A = (aij ), B = αIn − A = (bij ). Then

n
* n
*
bij uj = (αδij − aij )uij
j=1 j=1
n
*
= αui − aij uj
j=1
= 0 for all i.
T T
Thus, B(u1 , u2 , . . . , un ) = (0, 0, . . . , 0) . But
T T
(0, 0, . . . , 0) != (u1 , u2 , . . . , un ) ∈ Cn .

Thus, det(B) = 0. But, the determinant of B is a monic polynomial in

OK [α], so α is integral over OK . !

Note that this theorem, and its proof, were exactly the same as Theo-
rem 3.3.9, with OK replacing Z.

Theorem 5.1.7 OK is integrally closed.

Proof. If α ∈ K is integral over OK , then let

M = OK u1 + · · · + OK un , αM ⊆ M.

Let OK5= Zv1 +· · ·+Zvm , where {v1 , . . . , vm } is a basis for K over Q. Then
m 5n
M = i=1 j=1 Zvi uj is a finitely generated Z-module with αM ⊆ M , so
α is integral over Z. By definition, α ∈ OK . !

5.2 Characterizing Dedekind Domains

A ring is called Noetherian if every ascending chain a1 ⊆ a2 ⊆ a3 ⊆ · · · of
ideals terminates, i.e., if there exists n such that an = an+k for all k ≥ 0.

Exercise 5.2.1 If a " b are ideals of OK , show that N (a) > N (b).
56 CHAPTER 5. DEDEKIND DOMAINS

Exercise 5.2.2 Show that OK is Noetherian.

Theorem 5.2.3 For any commutative ring R with identity, the following
are equivalent:

(1) R is Noetherian;

(2) every nonempty set of ideals contains a maximal element; and

(3) every ideal of R is finitely generated.

Proof. (1) ⇒ (2) Suppose that S is a nonempty set of ideals of R that

does not contain a maximal element. Let a1 ∈ S. a1 is not maximal in
S, so there is an a2 ∈ S with a1 # a2 . a2 is not a maximal element of S,
so there exists an a3 ∈ S with a1 # a2 # a3 . Continuing in this way, we
find an infinite ascending chain of ideals of R. This contradicts R being
Noetherian, so every nonempty set of ideals contains a maximal element.
(2) ⇒ (3) Let b be an ideal of R. Let A be the set of ideals contained
in b which are finitely generated. A is nonempty, since (0) ∈ A. Thus, A
has a maximal element, say a = (x1 , . . . , xn ). If a != b, then ∃x ∈ b \ a.
But then a + (x) = (x1 , x2 , . . . , xn , x) is a larger finitely generated ideal
contained in b, contradicting the maximality of b. Thus, b = a, so b is
finitely generated. Thus, every ideal of R is finitely generated.
(3) ⇒ (1) %Let a1 ⊆ a2 ⊆ a3 ⊆ · · · be an ascending chain of ideals of
∞
R. Then a = i=1 ai is also an ideal of R, and so is finitely generated, say
a = (x1 , . . . , xn ). Then x1 ∈ ai1 , . . . , xn ∈ ain . Let m = max(i1 , . . . , in ).
Then a ⊆ am , so a = am . Thus, am = am+1 = · · · , and the chain does
terminate. Thus, R is Noetherian. !

Thus, we have proved that:

(1) OK is integrally closed;

(2) every nonzero prime ideal of OK is maximal; and

(3) OK is Noetherian.

A commutative integral domain which satisfies these three conditions

is called a Dedekind domain. We have thus seen that OK is a Dedekind
domain.

Exercise 5.2.4 Show that any principal ideal domain is a Dedekind domain.

√
Exercise 5.2.5 Show that Z[ −5] is a Dedekind domain, but not a principal
ideal domain.
5.3. FRACTIONAL IDEALS AND UNIQUE FACTORIZATION 57

5.3 Fractional Ideals and Unique Factoriza-

tion
Our next goal is to show that, in OK , every ideal can be written as a product
of prime ideals uniquely. In fact, this is true in any Dedekind domain.
A fractional ideal A of OK is an OK -module contained in K such that
there exists m ∈ Z with mA ⊆ OK . Of course, any ideal of OK is a
fractional ideal by taking m = 1.

Exercise 5.3.1 Show that any fractional ideal is finitely generated as an OK -

module.

Exercise 5.3.2 Show that the sum and product of two fractional ideals are again
fractional ideals.

Lemma 5.3.3 Any proper ideal of OK contains a product of nonzero prime

ideals.

Proof. Let S be the set of all proper ideals of OK that do not contain a
product of prime ideals. We need to show that S is empty. If not, then
since OK is Noetherian, S has a maximal element, say a. Then, a is not
prime since a ∈ S, so there exist a, b ∈ OK , with ab ∈ a, a ∈ / a, b ∈/ a.
Then, (a, a) $ a, (a, b) $ a. Thus, (a, a) ∈ / S, (a, b) ∈ / S, by the maximality
of a.
Thus, (a, a) ⊇ ℘1 · · · ℘r and (a, b) ⊇ ℘$1 · · · ℘$s , with the ℘i and ℘$j non-
zero prime ideals. But ab ∈ a, so (a, ab) = a.
Thus, a = (a, ab) ⊇ (a, a)(a, b) ⊇ ℘1 · · · ℘r ℘$1 · · · ℘$s . Therefore a contains
a product of prime ideals. This contradicts a being in S, so S must actually
be empty.
Thus, any proper ideal of OK contains a product of nonzero prime ideals.
!

Lemma 5.3.4 Let ℘ be a prime ideal of OK . There exists z ∈ K, z ∈

/ OK ,
such that z℘ ⊆ OK .

Proof. Take x ∈ ℘. From the previous lemma, (x) = xOK contains a

product of prime ideals. Let r be the least integer such that (x) contains
a product of r prime ideals, and say (x) ⊇ ℘1 · · · ℘r , with the ℘i nonzero
prime ideals.
Since ℘ ⊇ ℘1 · · · ℘r , ℘ ⊇ ℘i , for some i, from Theorem 5.1.2 (d). With-
out loss of generality, we can assume that i = 1, so ℘ ⊇ ℘1 . But ℘1 is a
nonzero prime ideal of OK , and so is maximal. Thus, ℘ = ℘1 .
58 CHAPTER 5. DEDEKIND DOMAINS

Now, ℘2 · · · ℘r % (x), since r was chosen to be minimal. Choose an

element b ∈ ℘2 · · · ℘r , b ∈
/ xOK . Then

bx−1 ℘ = bx−1 ℘1 ⊆ (℘2 · · · ℘r )(x−1 ℘1 )

= x−1 (℘1 · · · ℘r )
⊆ x−1 xOK
= OK .

Put z = bx−1 . Then z℘ ⊆ OK . Now, if z were in OK , we would have

bx ∈ OK , and so b ∈ xOK . But this is not the case, so z ∈
−1
/ OK . Thus,
we have found z ∈ K, z ∈
/ OK with z℘ ⊆ OK . !

Let ℘ be a prime ideal. Define

℘−1 = {x ∈ K : x℘ ⊆ OK }.

Lemma 5.3.4 implies, in particular, that ℘−1 != OK .

Theorem 5.3.5 Let ℘ be a prime ideal of OK . Then ℘−1 is a fractional

ideal and ℘℘−1 = OK .

Proof. It is easily seen that ℘−1 is an OK -module. Now, ℘ ∩ Z != (0), from

Exercise 4.4.1, so let n ∈ ℘ ∩ Z, n != 0. Then, n℘−1 ⊆ ℘℘−1 ⊆ OK , by
definition. Thus, ℘−1 is a fractional ideal.
It remains to show that ℘℘−1 = OK . Since 1 ∈ ℘−1 , ℘ ⊆ ℘℘−1 ⊆ OK .
℘℘ is an ideal of OK , since it is an OK -module contained in OK . But ℘
−1

is maximal, so either ℘℘−1 = OK , in which case we are done, or ℘℘−1 = ℘.

Suppose that ℘℘−1 = ℘. Then x℘ ⊆ ℘ ∀x ∈ ℘−1 . Since ℘ is a finitely
generated Z-module (from Exercise 4.4.2), x ∈ OK for all x ∈ ℘−1 , by
Theorem 5.1.6. Thus, ℘−1 ⊆ OK . But 1 ∈ ℘−1 , so ℘−1 = OK . From the
comments above, and by the previous lemma, we know this is not true, so
℘℘−1 != ℘. Thus, ℘℘−1 = OK . !

Theorem 5.3.6 Any ideal of OK can be written as a product of prime

ideals uniquely.

Proof.
Existence. Let S be the set of ideals of OK that cannot be written as
a product of prime ideals. If S is nonempty, then S has a maximal element,
since OK is Noetherian. Let a be a maximal element of S. Then a ⊆ ℘ for
some maximal ideal ℘, since OK is Noetherian. Recall that every maximal
ideal of OK is prime. Since a ∈ S, a != ℘ and therefore a is not prime.
Consider ℘−1 a. ℘−1 a ⊂ ℘−1 ℘ = OK . Since a # ℘,

℘−1 a # ℘−1 ℘ = OK ,
5.3. FRACTIONAL IDEALS AND UNIQUE FACTORIZATION 59

since for any x ∈ ℘\a,

℘−1 x ⊆ ℘−1 a ⇒ x ∈ ℘℘−1 a = OK a = a,

which is not true. Thus, ℘−1 a is a proper ideal of OK , and contains a

properly since ℘−1 contains OK properly. Thus, ℘−1 a ∈ / S, since a is a
maximal element of S. Thus, ℘−1 a = ℘1 · · · ℘r , for some prime ideals ℘i .
Then, ℘℘−1 a = ℘℘1 · · · ℘r , so a = ℘℘−1 a = ℘℘1 · · · ℘r . But then a ∈ / S, a
contradiction.
Thus, S is empty, so every ideal of OK can be written as a product of
prime ideals.
Uniqueness. Suppose that a = ℘1 · · · ℘r = ℘$1 · · · ℘$s are two factoriza-
tions of a as a product of prime ideals.
Then, ℘$1 ⊇ ℘$1 · · · ℘$s = ℘1 · · · ℘r , so ℘$1 ⊇ ℘i , for some i, say ℘$1 ⊇ ℘1 .
But ℘1 is maximal, so ℘$1 = ℘1 . Thus, multiplying both sides by (℘$1 )−1
and cancelling using (℘$1 )−1 ℘$1 = OK , we obtain

℘$2 · · · ℘$s = ℘2 · · · ℘r .

Thus, continuing in this way, we see that r = s and the primes are unique
up to reordering. !

It is possible to show that any integral domain in which every non-zero

ideal can be factored as a product of prime ideals is necessarily a Dedekind
domain. We refer the reader to p. 82 of [Mat] for the details of the proof.
This fact gives us an interesting characterization of Dedekind domains.

When ℘ and ℘$ are prime ideals, we will write ℘/℘$ for (℘$ )−1 ℘. We
will also write
℘1 ℘2 · · · ℘r
℘$1 ℘$2 · · · ℘$s
to mean (℘$1 )−1 (℘$2 )−1 · · · (℘$s )−1 ℘1 ℘2 · · · ℘r .

Exercise 5.3.7 Show that any fractional ideal A can be written uniquely in the
form
℘1 . . . ℘ r
,
℘(1 . . . ℘(s
where the ℘i and ℘(j may be repeated, but no ℘i = ℘(j .

Exercise 5.3.8 Show that, given any fractional ideal A =

# 0 in K, there exists a
fractional ideal A−1 such that AA−1 = OK .

For a and b ideals of OK , we say a divides b (denoted a | b), if a ⊇ b.

Exercise 5.3.9 Show that if a and b are ideals of OK , then b | a if and only if
there is an ideal c of OK with a = bc.

Define d to be the greatest common divisor of a, b if:

60 CHAPTER 5. DEDEKIND DOMAINS

(i) d | a and d | b; and

(ii) e | a and e | b ⇒ e | d.
Denote d by gcd(a, b).
Similarly, define m to be the least common multiple of a, b if:
(i) a | m and b | m; and
(ii) a | n and b | n ⇒ m | n.
Denote m by lcm(a, b).
In the next two exercises, we establish the existence (and uniqueness)
of the gcd and lcm of two ideals of OK .
2r 2r
Let a = i=1 ℘ei i , b = i=1 ℘fi i , with ei , fi ∈ Z≥0 .
/r min(ei ,fi )
Exercise 5.3.10 Show that gcd(a, b) = a + b = i=1 ℘i .
/r max(ei ,fi )
Exercise 5.3.11 Show that lcm(a, b) = a ∩ b = i=1 ℘i .

Exercise 5.3.12 Suppose a, b, c are ideals of OK . Show that if ab = cg and

gcd(a, b) = 1, then a = dg and b = eg for some ideals d and e of OK . (This
generalizes Exercise 1.2.1.)

Theorem 5.3.13 (Chinese Remainder Theorem) (a) Let a, b be ide-

als so that gcd(a, b) = 1, i.e., a + b = OK . Given a, b ∈ OK , we can
solve

x ≡ a (mod a),
x ≡ b (mod b).

(b) Let ℘1 , . . . , ℘r be r distinct prime ideals in OK . Given ai ∈ OK , ei ∈

Z>0 , ∃x such that x ≡ ai (mod ℘ei i ) for all i ∈ {1, . . . , r}.
Proof. (a) Since a + b = OK , ∃x1 ∈ a, x2 ∈ b with x1 + x2 = 1. Let
x = bx1 + ax2 ≡ ax2 (mod a). But, x2 = 1 − x1 ≡ 1 (mod a). Thus, we
have found an x such that x ≡ a (mod a). Similarly, x ≡ b (mod b).
(b) We proceed by induction on r. If r = 1, there is nothing to show.
Suppose r > 1, and that we can solve x ≡ ai (mod ℘ei i ) for i = 1, . . . , r − 1,
say a ≡ ai (mod ℘ei i ) for i = 1, . . . , r − 1. From part (a), we can solve
e
x ≡ a (mod ℘1e1 · · · ℘r−1
r−1
),
x ≡ ar (mod ℘r ).
er

e
Then x − ai ∈ ℘e11 · · · ℘r−1
r−1
, x ≡ ar (mod ℘err ). Thus, x − ai ∈ ℘ei i ∀i, i.e.,
ei
x ≡ ai (mod ℘i ) ∀i. !

We define the order of a in ℘ by ord℘ (a) = t if ℘t | a and ℘t+1 ! a.

5.3. FRACTIONAL IDEALS AND UNIQUE FACTORIZATION 61

Exercise 5.3.14 Show that ord℘ (ab) = ord℘ (a) + ord℘ (b), where ℘ is a prime
ideal.

Exercise 5.3.15 Show that, for α #= 0 in OK , N ((α)) = |NK (α)|.

2r
Theorem 5.3.16 (a) If a = i=1 ℘ei i , then
r
6
N (a) = N (℘ei i ).
i=1

(b) OK /℘ 1 ℘e−1 /℘e , and

N (℘e ) = (N (℘))e

for any integer e ≥ 0.

Proof. (a) Consider the map
r
φ : OK −→ ⊕ (OK /℘ei i ),
i=1
x −→ (x1 , . . . , xr ),

where xi ≡ x (mod ℘ei i ).

The function φ is surjective by the Chinese Remainder Theorem, and φ
is a homomorphism since each of the r components x −→ xi (mod ℘ei i ) is
a homomorphism. >r 2r
Next, we show by induction that i=1 ℘ei i = i=1 ℘ei i . The base case
r = 1 is trivial. Suppose r > 1, and that the result is true for numbers
smaller than r.

r
/r−1 0
? ?
℘ei i = lcm ℘ei i , ℘err
i=1 i=1
/r−1 0
6
= lcm ℘ei i , ℘err
i=1
r
6
= ℘ei i .
i=1
>r 2r
Thus, ker(φ) = i=1 ℘ei i = i=1 ℘ei i , which implies that

OK /a 1 ⊕(OK /℘ei i ).
2r
Hence, N (a) = i=1 N (℘ei i ).
(b) Since ℘e # ℘e−1 , we can find an element α ∈ ℘e−1 /℘e , so that
ord℘ (α) = e − 1. Then ℘e ⊆ (α) + ℘e ⊆ ℘e−1 . So ℘e−1 | (α) + ℘e . But
62 CHAPTER 5. DEDEKIND DOMAINS

(α) + ℘e != ℘e , so ℘e−1 = (α) + ℘e , by unique factorization. Define the map

φ : OK −→ ℘e−1 /℘e by φ(γ) = γα + ℘e . This is clearly a homomorphism
and is surjective since ℘e−1 = (α) + ℘e .
Now,

γ ∈ ker(φ) ⇔ γα ∈ ℘e
⇔ ord℘ (γα) ≥ e
⇔ ord℘ (γ) + ord℘ (α) ≥ e
⇔ ord℘ (γ) + e − 1 ≥ e
⇔ ord℘ (γ) ≥ 1
⇔ γ ∈ ℘.

Thus, OK /℘ 1 ℘e−1 /℘e . Also,

(OK /℘e )/(℘e−1 /℘e ) 1 OK /℘e−1

since the map from OK /℘e to OK /℘e−1 taking x + ℘e to x + ℘e−1 is a

surjective homomorphism with kernel ℘e−1 /℘e . Thus,

N (℘e ) = |OK /℘e | = |OK /℘e−1 | |℘e−1 /℘e |

= N (℘e−1 )N (℘)
= N (℘)e−1 N (℘) by the induction hypothesis
= N (℘)e . !

Thus, the norm function is multiplicative. Also, we can extend the

definition of norm to fractional ideals, in the following way. Since any
fractional ideal can be written uniquely in the form ab−1 where a, b are
ideals of OK , we can put

N (a)
N (ab−1 ) = .
N (b)

Let OK = Zω1 + · · · + Zωn . Then, if p is a prime number, we have

pOK = Zpω1 + · · · + Zpωn , and so N ((p)) = pn , where n = [K : Q].

Exercise 5.3.17 If we write pOK as its prime factorization,

e
pOK = ℘e11 · · · ℘gg ,
&g
show that N (℘i ) is a power of p and that if N (℘i ) = pfi i , i=1 ei fi = n.
5.4. DEDEKIND’S THEOREM 63

5.4 Dedekind’s Theorem

The number ei found in the previous exercise is called the ramification
degree of ℘i . (We sometimes write e℘i for ei .) We say the prime number p
ramifies in K if some ei ≥ 2. If all the fi ’s are 1, we say p splits completely.
Our next goal is to show Dedekind’s Theorem: If p is a prime number
that ramifies in K, then p | dK . Recall that dK = det(Tr(ωi ωj )), where
ω1 , . . . , ωn is any integral basis for OK .
Let D−1 = {x ∈ K : Tr(xOK ) ⊆ Z}.

Exercise 5.4.1 Show that D −1 is a fractional ideal of K and find an integral

basis.

Exercise 5.4.2 Let D be the fractional ideal inverse of D −1 . We call D the

different of K. Show that D is an ideal of OK .

Theorem 5.4.3 Let D be the different of an algebraic number field K.

Then N (D) = |dK |.

Proof. For some m > 0, mD−1 is an ideal of OK . Now,

mD−1 = Zmω1∗ + · · · + Zmωn∗ .

Let
n
*
mωi∗ = aij ωj ,
j=1

so
n
* aij
ωi∗ = ωj ,
j=1
m

and
n
*
ωi = bij ωj∗ .
j=1

Thus, (bij ) is the matrix inverse of (aij /m). But

/ n
0
*
Tr(ωi ωj ) = Tr bik ωk∗ ωj
k=1
n
*
= bik Tr(ωk∗ ωj )
k=1
= bij .

Thus, det(bij ) = dK .
64 CHAPTER 5. DEDEKIND DOMAINS

However, by Exercise 4.2.8, since mD−1 is an ideal of OK with integral

basis mω1∗ , . . . , mωn∗ , we know that

dK/Q (mω1∗ , . . . , mωn∗ ) = N (mD−1 )2 dK

and from Exercise 4.2.7 we have

dK/Q (mω1∗ , . . . , mωn∗ ) = (det(aij ))2 dK ,

which shows that

|det(aij )| = N (mD−1 ) = mn N (D−1 ),

and thus
aij
|det( )| = N (D−1 ) = N (D)−1 .
m
Hence, |dK | = |det(bij )| = | det(aij /m)|−1 = N (D). !

Theorem 5.4.4 Let p ∈ Z be prime, ℘ ⊆ OK , a prime ideal and D the

different of K. If ℘e | (p), then ℘e−1 | D.

Proof. We may assume that e is the highest power 5nof ℘ dividing (p). So let
(p) = ℘e a, gcd(a, ℘) = 1. Let x ∈ ℘a. Then x = i=1 pi ai , pi ∈ ℘, ai ∈ a.
Hence,
*n
p
x ≡ ppi api (mod p),
i=1

and
n
* m m
pm
x ≡ ppi api (mod p).
i=1
m m m
For sufficiently large m, ppi ∈ ℘e , so xp ∈ ℘e and thus, xp ∈ ℘e a = (p).
m
Therefore, Tr(xp ) ∈ pZ, which implies that
m
(Tr(x))p ∈ pZ
⇒ Tr(x) ∈ pZ
⇒ Tr(p−1 ℘a) ⊆ Z
⇒ p−1 ℘a ⊆ D−1
⇒ Dp−1 ℘a ⊆ DD−1 = OK
⇒ D ⊆ p℘−1 a−1 = ℘e a℘−1 a−1 = ℘e−1
⇒ ℘e−1 | D. !

Exercise 5.4.5 Show that if p is ramified, p | dK .

Dedekind also proved that if p | dK , then p ramifies. We do not prove

this here. In the Supplementary Problems, some special cases are derived.
5.5. FACTORIZATION IN OK 65

5.5 Factorization in OK
The following theorem gives an important connection between factoring
polynomials mod p and factoring ideals in number fields:

Theorem 5.5.1 Suppose that there is a θ ∈ K such that OK = Z[θ]. Let

f (x) be the minimal polynomial of θ over Z[x]. Let p be a rational prime,
and suppose
f (x) ≡ f1 (x)e1 · · · fg (x)eg (mod p),
e
where each fi (x) is irreducible in Fp [x]. Then pOK = ℘e11 · · · ℘gg where
℘i = (p, fi (θ)) are prime ideals, with N (℘i ) = pdeg fi .
Proof. We first note that (p, f1 (θ))e1 · · · (p, fg (θ))eg ⊆ pOK . Thus it suf-
fices to show that (p, fi (θ)) is a prime ideal of norm pdi where di is the
degree of fi .
Now, since fi (x) is irreducible over Fp , then Fp [x]/(fi (x)) is a field.
Also,

Z[x]/(p) 1 Fp [x], ⇒ Z[x]/(p, fi (x)) 1 Fp [x]/(fi (x)),

and so Z[x]/(p, fi (x)) is a field.

Consider the map ϕ : Z[x] → Z[θ]/(p, fi (θ)). Clearly

(p, fi (x)) ⊆ ker(ϕ) = {n(x) : n(θ) ∈ (p, fi (θ))}.

If n(x) ∈ ker(ϕ), we can divide by fi (x) to get

n(x) = q(x)fi (x) + ri (x), deg(ri ) < deg(fi ).

We assume that ri is nonzero, for otherwise the result is trivial. Since

n(θ) ∈ (p, fi (θ)), then ri (θ) ∈ (p, fi (θ)), so ri (θ) = pa(θ) + fi (θ)b(θ). Here
we have used the fact that OK = Z[θ].
Now define the polynomial h(x) = ri (x) − pa(x) − fi (x)b(x). Since
h(θ) = 0 and f is the minimal polynomial of θ, then h(x) = g(x)f (x) for
some polynomial g(x) ∈ Z[x]. We conclude that ri (x) = pã(x) + fi (x)b̃(x)
for some ã(x), b̃(x) ∈ Z[x]. Therefore ri (x) ∈ (p, fi (x)).
Thus,
Z[θ]/(p, fi (θ)) 1 Z[x]/(p, fi (x)) 1 Fp [x]/(fi (x))
and is therefore a field. Hence, (p, fi (θ)) is a maximal ideal and is therefore
prime.
e" e"
Now, let e$i be the ramification index of ℘i , so that pOK = ℘11 · · · ℘gg ,
and let di = [OK /℘i : Z/p]. Clearly di is the degree of the polynomial fi (x).
Since f (θ) = 0, and since f (x) − f1 (x)e1 · · · fg (x)eg ∈ pZ[x], it follows that
f1 (θ)e1 · · · fg (θ)eg ∈ pOK = pZ[θ]. Also, ℘ei i ⊂ pOK + (fi (θ)ei ) and so
e" e"
℘e11 · · · ℘egg ⊆ pOK = ℘11 · · · ℘gg .
66 CHAPTER 5. DEDEKIND DOMAINS

Therefore, ei ≥ e$i for all i. But

* *
ei di = deg f = [K : Q] = e$i di .

Thus, ei = e$i for all i. !

Exercise 5.5.2 If in6 the previous

7 theorem we do not assume that OK = Z[θ]
but instead that p ! OK : Z[θ] , show that the same result holds.

Exercise 5.5.3 Suppose that f (x) in the previous exercise is Eisensteinian with
respect to the prime p. Show that p ramifies totally in K. That is, pOK = (θ)n
where n = [K : Q].

Exercise 5.5.4 Show that (p) = (1 − ζp )p−1 when K = Q(ζp ).

5.6 Supplementary Problems

Exercise 5.6.1 Show that if a ring R is a Dedekind domain and a unique fac-
torization domain, then it is a principal ideal domain.

Exercise 5.6.2 Using

√ Theorem 5.5.1, find a prime ideal factorization of 5OK
and 7OK in Z[(1 + −3)/2].

Exercise 5.6.3 Find a prime ideal factorization of (2), (5), (11) in Z[i].
√
Exercise 5.6.4 Compute the different D of K = Q( −2).
√
Exercise 5.6.5 Compute the different D of K = Q( −3).

Exercise 5.6.6 Let K = Q(α) be an algebraic number field of degree n over Q.

Suppose OK = Z[α] and that f (x) is the minimal polynomial of α. Write

f (x) = (x − α)(b0 + b1 x + · · · + bn−1 xn−1 ), bi ∈ O K .

Prove that the dual basis to 1, α, . . . , αn−1 is

b0 bn−1
,... , ( .
f ( (α) f (α)

Deduce that
1
D −1 = (Zb0 + · · · + Zbn−1 ).
f ( (α)

Exercise 5.6.7 Let K = Q(α) be of degree n over Q. Suppose that OK = Z[α].

Prove that D = (f ( (α)).

Exercise 5.6.8 Compute the different D of Q[ζp ] where ζp is a primitive pth

root of unity.
5.6. SUPPLEMENTARY PROBLEMS 67

Exercise 5.6.9 Let p be a prime, p ! m, and a ∈ Z. Show that p | φm (a)

if and only if the order of a (mod p) is n. (Here φm (x) is the mth cyclotomic
polynomial.)

Exercise 5.6.10 Suppose p ! m is prime. Show that p | φm (a) for some a ∈ Z if

and only if p ≡ 1 (mod m). Deduce from Exercise 1.2.5 that there are infinitely
many primes congruent to 1 (mod m).

Exercise 5.6.11 Show that p ! m splits completely in Q(ζm ) if and only if p ≡ 1

(mod m).

Exercise 5.6.12 Let p be prime and let a be squarefree and coprime to p. Set
θ = a1/p and consider K = Q(θ). Show that OK = Z[θ] if and only if ap−1 #≡ 1
(mod p2 ).

Exercise 5.6.13 Suppose that K = Q(θ) and OK = Z[θ]. Show that if p | dK ,

p ramifies.

Exercise 5.6.14 Let K = Q(θ) and suppose that p | dK/Q (θ), p2 ! dK/Q (θ).
Show that p | dK and p ramifies in K.

Exercise 5.6.15 Let K be an algebraic number field of discriminant √ dK . Show

that the normal closure of K contains a quadratic field of the form Q( dK ).

Exercise 5.6.16 Show that if p ramifies in K, then it ramifies in each of the

conjugate fields of K. Deduce that if p ramifies in the normal closure of K, then
it ramifies in K.

Exercise 5.6.17 Deduce the following special case of Dedekind’s theorem: if

p2m+1 /dK show that p ramifies in K.

Exercise 5.6.18
√ Determine the prime ideal factorization of (7), (29), and (31)
in K = Q( 3 2).

Exercise 5.6.19 If L/K is a finite extension of algebraic number field, we can

view L as a finite dimensional vector space over K. If α ∈ L, the map v 0→ αv is
a linear mapping and one can define, as before, the relative norm NL/K (α) and
relative trace T rL/K (α) as the determinant and trace, respectively, of this linear
map. If α ∈ OL , show that T rL/K (α) and NL/K (α) lie in OK .

Exercise 5.6.20 If K ⊆ L ⊆ M are finite extensions of algebraic number fields,

show that NM/K (α) = NL/K (NM/L (α)) and T rM/K (α) = T rL/K (T rM/L (α)) for
any α ∈ M . (We refer to this as the transitivity property of the norm and trace
map, respectively.)

Exercise 5.6.21 Let L/K be a finite extension of algebraic number fields. Show
that the map
T rL/K : L × L → K
is non-degenerate.
68 CHAPTER 5. DEDEKIND DOMAINS

Exercise 5.6.22 Let L/K be a finite extension of algebraic number fields. Let
a be a finitely generated OK -module contained in L. The set
−1
DL/K (a) = {x ∈ L : T rL/K (xa) ⊆ OK }

is called codifferent of a over K. If a #= 0, show that DL/K

−1
(a) is a finitely generated
OK -module. Thus, it is a fractional ideal of L.

Exercise 5.6.23 If in the previous exercise a is an ideal of OL , show that the

fractional ideal inverse, denoted DL/K (a) of DL/K
−1
(a) is an integral ideal of OL .
(We call DL/K (a) the different of a over K. In the case a is OL , we call it the
relative different of L/K and denote it by DL/K .)

Exercise 5.6.24 Let K ⊆ L ⊆ M be algebraic number fields of finite degree

over the rationals. Show that

DM/K = DM/L (DL/K OM ).

Exercise 5.6.25 Let L/K be a finite extension of algebraic number fields. We

define the relative discriminant of L/K, denoted dL/K as NL/K (DL/K ). This is
an integral ideal of OK . If K ⊆ L ⊆ M are as in Exercise 5.6.24, show that
[M :L]
dM/K = dL/K NL/K (dM/L ).

Exercise 5.6.26 Let L/K be a finite extension of algebraic number fields. Sup-
pose that OL = OK [α] for some α ∈ L. If f (x) is the minimal polynomial of α
over OK , show that DL/K = (f ( (α)).

Exercise 5.6.27 Let K1 , K2 be algebraic number fields of finite degree over K.

If L/K is the compositum of K1 /K and K2 /K, show that the set of prime ideals
dividing dL/K and dK1 /K dK2 /K are the same.

Exercise 5.6.28 Let L/K be a finite extension of algebraic number fields. If L̃

denotes the normal closure, show that a prime p of OK is unramified in L if and
only if it is unramified in L̃.
Chapter 6

The Ideal Class Group

This chapter mainly discusses the concept of the ideal class group, and some
of its applications to Diophantine equations. We will prove that the ideal
class group of an algebraic number field is finite, and establish some related
results.
As in all other chapters, we shall let K be an algebraic number field
with degree n over Q, and let OK be the ring of algebraic integers in K.

6.1 Elementary Results

This section serves as preparation and introduction to the remainder of the
chapter. We start by a number of standard results.
Recall that the ring OK is Euclidean if given α ∈ K, ∃β ∈ OK such that
|N (α − β)| < 1. Indeed, given θ, γ ∈ OK , the fact that there exist q, r ∈ OK
with r = θ − qγ and
|N (r)| < |N (γ)|
is equivalent to the fact that there exists q ∈ OK such that

|N (θ/γ − q)| < 1.

Let α = θ/γ, let β = q, and we have

|N (α − β)| < 1.

In general, OK is not Euclidean, but the following results always hold:

Lemma 6.1.1 There is a constant HK such that given α ∈ K, ∃β ∈ OK ,

and a non-zero integer t, with |t| ≤ HK , such that

|N (tα − β)| < 1.

69
70 CHAPTER 6. THE IDEAL CLASS GROUP

Proof. Let {ω1 , ω2 , . . . , ωn } be an integral basis of OK . Given any α ∈ K,

there exists m ∈ Z such that mα ∈ OK , so α can be written as
n
*
α= ci ωi ,
i=1

with ci ∈ Q for all i = 1, 2, . . . , n.

Let L be a natural number. Partition the interval [0, 1] into L parts,
each of length 1/L. This induces a subdivision of [0, 1]n into Ln subcubes.
Consider the map ϕ : αZ −→ [0, 1]n defined by
ϕ
tα −→ ({tc1 }, {tc2 }, . . . , {tcn }),
where t ∈ Z, and {a} denotes the fractional part of a ∈ R. Let t run from
0 to Ln (the number of subcubes in [0, 1]n ). The number of choices for t is
then Ln + 1, which is one more than the number of subcubes. There must
be two distinct values of t, say t1 and t2 , so that t1 α and t2 α get mapped
to the same subcube of [0, 1]n . Let
n
*
β= ([t1 ci ] − [t2 ci ])ωi ,
i=1

where [a] denotes the integer part of a ∈ R. Then,

n
*
(t1 − t2 )α − β = ({t1 ci } − {t2 ci })ωi .
i=1

Let t = t1 − t2 , then
1 / 01
1 *n 1
1 1
|N (tα − β)| = 1N ({t1 ci } − {t2 ci })ωi 1 .
1 1
i=1

Since |({t1 ci } − {t2 ci })| ≤ 1/L, we then have

n
/ n 0
1 6 * (j)
|N (tα − β)| ≤ n |ω | ,
L j=1 i=1 i

(j) 2n 5n (j)
where ωi is the jth conjugate of ωi . If we take Ln > j=1 ( i=1 |ωi |) =
HK (say), then
|N (tα − β)| < 1.
Furthermore, since 0 ≤ t1 , t2 ≤ Ln , we have |t| ≤ Ln . Thus, if we choose
1/n
L = HK , we are done. !

Let us call HK as defined above the Hurwitz constant, since the lemma
is due to A. Hurwitz.

Exercise 6.1.2 Show that given α, β ∈ OK , there exist t ∈ Z, |t| ≤ HK , and

w ∈ OK so that |N (αt − βw)| < |N (β)|.
6.2. FINITENESS OF THE IDEAL CLASS GROUP 71

6.2 Finiteness of the Ideal Class Group

The concept of the ideal class group arose from Dedekind’s work in estab-
lishing the unique factorization theory for ideals in the ring of algebraic
integers of a number field. Our main aim of this section is to prove that the
ideal class group is finite. We start by introducing an equivalence relation
on ideals.
We proved in Exercise 5.3.7 that any fractional ideal A can be written
uniquely in the form
℘1 . . . ℘s
A= $ ,
℘1 . . . ℘$r
where the ℘i , ℘$i are primes in OK , and no ℘i is a ℘$j (recall that we write
1/℘ = ℘−1 ). In particular, we can always write any fractional ideal A in
the form
b
A= ,
c
where b, c are two integral ideals.
Two fractional ideals A and B in K are said to be equivalent if there
exist α, β ∈ OK such that (α)A = (β)B. In this case, we write A ∼ B.
Notice that if OK is a principal ideal domain then any two ideals are
equivalent.

Exercise 6.2.1 Show that the relation ∼ defined above is an equivalence rela-
tion.

Theorem 6.2.2 There exists a constant CK such that every ideal a ⊆ OK

is equivalent to an ideal b ⊆ OK with N (b) ≤ CK .
Proof. Suppose a is an ideal of OK . Let β ∈ a be a non-zero element such
that |N (β)| is minimal.
For each α ∈ a, by Exercise 6.1.2, we can find t ∈ Z, |t| ≤ HK , and
w ∈ OK such that
|N (tα − wβ)| < |N (β)|.
Moreover, since α, β ∈ a, so tα − wβ ∈ a; and therefore, by the minimality
of |N (β)|, we must have tα = wβ. Thus, we have shown that for any α ∈ a,
there exist t ∈ Z, |t| ≤ HK , and w ∈ OK such that tα = wβ.
Let 6
M= t,
|t|≤HK

and we have M a ⊆ (β). This means that (β) divides (M )a, and so
(M )a = (β)b,
for some ideal b ⊆ OK .
Observe that β ∈ a, so M β ∈ (β)b, and hence (M ) ⊆ b. This implies
that |N (b)| ≤ N ((M )) = CK . Hence, a ∼ b, and CK = N ((M )) satisfies
the requirements. !
72 CHAPTER 6. THE IDEAL CLASS GROUP

Exercise 6.2.3 Show that each equivalence class of ideals has an integral ideal
representative.

Exercise 6.2.4 Prove that for any integer x > 0, the number of integral ideals
a ⊆ OK for which N (a) ≤ x is finite.

Theorem 6.2.5 The number of equivalence classes of ideals is finite.

Proof. By Exercise 6.2.3, each equivalence class of ideals can be rep-

resented by an integral ideal. This integral ideal, by Theorem 6.2.2, is
equivalent to another integral ideal with norm less than or equal to a given
constant CK . Apply Exercise 6.2.4, and we are done. !

As we did in the proof of Exercise 6.2.3, it is sufficient to consider only

integral representatives when dealing with equivalence classes of ideals.
Let H be the set of all the equivalence classes of ideals of K. Given C1
and C2 in H, we define the product of C1 and C2 to be the equivalence class
of AB, where A and B are two representatives of C1 and C2 , respectively.

Exercise 6.2.6 Show that the product defined above is well defined, and that H
together with this product form a group, of which the equivalence class containing
the principal ideals is the identity element.

Theorem 6.2.5 and Exercise 6.2.6 give rise to the notion of class number.
Given an algebraic number field K, we denote by h(K) the cardinality of
the group of equivalence classes of ideals (h(K) = |H|), and call it the class
number of the field K. The group of equivalence classes of ideals is called
the ideal class group.
With the establishment of the ideal class group, the result in Theorem
6.2.2 can be improved as follows:

Exercise 6.2.7 Show that the constant CK in Theorem 6.2.2 could be taken to
be the greatest integer less than or equal to HK , the Hurwitz constant.

The improvement on the bound enables us to determine the class num-

ber of many algebraic number fields. We demonstrate this by looking at
the following example:
√
Example 6.2.8 Show that the class number of K = Q( −5) is 2.
√
Solution. We proved in Exercise 4.1.3 that the integers in K are Z[ −5],
so that √
(1) (1)
ω1 = 1, ω2 = −5,
(2) (2) √
ω1 = 1, ω2 = − −5,
6.3. DIOPHANTINE EQUATIONS 73
√
and the Hurwitz constant is (1 + 5)2 = 10.45 · · · . Thus, CK = 10. This
implies that every equivalence class of ideals C ∈ H has an integral repre-
sentative a such that N (a) ≤ 10. a has a factorization into a product of
primes, say,
a = ℘1 ℘2 . . . ℘m ,
where ℘i is prime in OK for all i = 1, . . . , m.
Consider ℘1 . There exists, by Exercise 4.4.4, a unique prime number
p ∈ Z such that p ∈ ℘1 . This implies that ℘1 is in the factorization of
(p) into 2product of primes in OK . Thus, N (℘1 ) is a power of p. Since
m
N (a) = i=1 N (℘i ), and N (a) ≤ 10, we deduce that N (℘i ) ≤ 10 for all i.
And so, in particular, N (℘1 ) ≤ 10. Therefore, p ≤ 10. Thus, p could be 2,
3, 5, or 7. √
For p = 2, 3, 5, and 7, (p) factors in Z[ −5] as follows:
√ √
(2) = (2, 1 + −5)(2, 1 − −5),
√ √
(3) = (3, 1 + −5)(3, 1 − −5),
√ √
(7) = (7, 3 + −5)(7, 3 − −5),

and √
(5) = ( −5)2 .
√ √ √
Thus,
√ ℘1 can√ only be (2,√1 + −5), √(2, 1 − −5), (3, 1 + −5), (3, 1 −
−5), (7, 3 + −5), (7, 3 − −5), or ( −5). The same conclusion
√ holds for
any ℘i for i = 2, . . . , m. Moreover, it can be seen that ( −5) is principal,
and all the others are not principal (by taking the norms), but are pairwise
equivalent by the following relations:
√ √
(2, 1 + −5) = (2, 1 − −5),
√ √ √
(3, 1 + −5)(1 − −5) = (3)(2, 1 − −5),
√ √ √
(3, 1 − −5)(1 + −5) = (3)(2, 1 + −5),
√ √ √
(7, 3 + −5)(3 − −5) = (7)(2, 1 − −5),
√ √ √
(7, 3 − −5)(3 + −5) = (7)(2, 1 + −5).

Therefore, a is equivalent to either the class of principal ideals or the class

of those primes listed above. √
Hence, the class number of K = Q( −5) is 2, and the problem is solved.
In the supplementary problems we will derive a sharper constant than CK .

6.3 Diophantine Equations

In this section, we look at the equation

x2 + k = y 3 , (6.2)
74 CHAPTER 6. THE IDEAL CLASS GROUP

which was first introduced by Bachet in 1621, and has played a fundamental
role in the development of number theory. When k = 2, the only integral
solutions to this equation are given by y = 3 (see Exercise 2.4.3); and this
result is due to Fermat. It is known that the equation has no integral
solution for many different values of k. There are various methods for
discussing integral solutions of equation (6.2). We √
shall present, here, the
one that uses applications of the quadratic field Q( −k), and the concept
of ideal class group. This method is usually referred to as Minkowski’s
method. We start with a simple case, when k = 5.

Example 6.3.1 Show that the equation x2 + 5 = y 3 has no integral solu-

tion.
Solution. Observe that if y is even, then x is odd, and x2 +5 ≡ 0 (mod 4),
and hence x2 ≡ 3 (mod 4), which is a contradiction. Therefore, y is odd.
Also, if a prime p | (x, y), then p | 5, so p = 5; and hence, by dividing both
sides of the equation by 5, we end up with 1 ≡ 0 (mod 5), which is absurd.
Thus, x and y are coprime.
Suppose now that (x, y) is an integral solution to the given equation.
We consider the factorization
√ √
(x + −5)(x − −5) = y 3 , (6.3)
√
in the ring of integers Z[ −5]. √ √
Suppose a prime ℘ divides the gcd of (x + −5) and (x − −5) (which
implies ℘ divides (y)). Then ℘ divides (2x). Also, since y is odd, ℘ does
not divide (2). Thus, ℘ divides (x).
√ This is a contradiction
√ to the fact that
x and y are coprime. Hence, (x + −5) and (x − −5) are coprime ideals.
This and equation (6.3) ensure (by Exercise 5.3.12) that
√ √
(x + −5) = a3 and (x − −5) = b3 ,

for some ideals a and b. √

Since the class number of Q( −5) was found in Example 6.2.8 to be 2,
c2 is principal for any ideal c. Thus, since a3 and b3 are principal,√we deduce
that a and b are also principal. Moreover, since the units of Q( −5) are 1
and −1, which are both cubes, we conclude that
√ √
x + −5 = (a + b −5)3 ,

for some integers a and b. This implies that

1 = b(3a2 − 5b2 ).

It is easy to see that b | 1, so b = ±1. Therefore, 3a2 − 5 = ±1. Both cases

lead to contradiction with the fact that a ∈ Z.
Hence, the given equation does not have an integral solution.
6.4. EXPONENTS OF IDEAL CLASS GROUPS 75

The discussion for many, but by no means all, values of k goes through
without any great change. For instance, one can show that when k = 13
and k = 17, the only integral solutions to equation (6.2) are given by y = 17
and y = 5234, respectively.
We now turn to a more general result.

Exercise 6.3.2 Let k > 0 be a squarefree positive integer. Suppose that k ≡ 1, 2

(mod 4), and k does not have the form k = 3a2 ± 1 for an integer a. Consider
the equation

x2 + k = y 3 . (6.4)
√
Show that if 3 does not divide the class number of Q( −k), then this equation
has no integral solution.

6.4 Exponents of Ideal Class Groups

The study of class groups of quadratic fields is a fascinating one with
many conjectures and few results. For instance, it was proved in 1966
by H. Stark and A. Baker (independently) that there are exactly √ nine
imaginary quadratic fields of class number one. They are Q( −d) with
d = 1, 2, 3, 7, 11, 19, 43, 67, 163.
By combining Dirichlet’s class number formula (see Chapter 10, Exercise
10.5.12) with √ analytic results due √ to Siegel, one can show that the class
number of Q( −d) grows like d. More precisely, if h(−d) denotes the
class number,
log h(−d) ∼ 12 log d
as d → ∞.
The study of the growth of class numbers of real quadratic fields is
more complicated. For example, it is a classical conjecture of Gauss that
there are infinitely many real quadratic fields of class number 1. Related to
the average behaviour of class numbers of real quadratic fields, C. Hooley
formulated some interesting conjectures in 1984.
Around the same time, Cohen and Lenstra formulated general conjec-
tures about the distribution of class groups of quadratic fields. A particular
case of these conjectures is illustrated by the following. Let p be prime != 2.
They predict that the probability that p divides the order of the class group
of an imaginary quadratic field is
∞ !
6 "
1
1− 1− .
i=1
pi

A similar conjecture is made in the real quadratic case.

These conjectures suggest that as a first step, it might be worthwhile
to investigate the exponents of class groups of imaginary quadratic fields.
76 CHAPTER 6. THE IDEAL CLASS GROUP

A similar analysis for the real quadratic fields is more difficult and is post-
poned to Exercise 8.3.17 in Chapter 8.

Exercise 6.4.1 Fix a positive integer g > 1. Suppose that n is odd, greater
√ than
1, and ng − 1 = d is squarefree. Show that the ideal class group of Q( −d) has
an element of order g.

Exercise 6.4.2 Let g be odd and greater than 1.√If d = 3g − x2 is squarefree

with x odd and satisfying x2 < 3g /2, show that Q( −d) has an element of order
g in the class group.

Exercise 6.4.3 Let g be odd. Let N be the number of squarefree integers of

the form 3g − x2 , x odd, 0 < x2 < 3g /2. For g sufficiently large, show that
N ( 3g/2 . Deduce that there are infinitely many imaginary quadratic fields
whose class number is divisible by g.

6.5 Supplementary Problems

√
Exercise 6.5.1 Show that the class number of K = Q( −19) is 1.

We define the volume of a domain C ⊆ Rn to be

@
vol(C) = χ(x)dx
C

where χ(x) is the characteristic function of C:

;
1, x ∈ C,
χ(x) =
0, x !∈ C.

Exercise 6.5.2 (Siegel) Let C be a symmetric, bounded domain in Rn . (That

is, C is bounded and if x ∈ C so is −x.) If vol(C) > 1, then there are two distinct
points P, Q ∈ C such that P − Q is a lattice point.

Exercise 6.5.3 If C is any convex, bounded, symmetric domain of volume > 2n ,

show that C contains a non-zero lattice point. (C is said to be convex if x, y ∈ C
implies λx + (1 − λ)y ∈ C for 0 ≤ λ ≤ 1.)

Exercise 6.5.4 Show in the previous question if the volume ≥ 2n , the result is
still valid, if C is closed.

Exercise 6.5.5 Show that there exist bounded, symmetric convex domains with
volume < 2n that do not contain a lattice point.
6.5. SUPPLEMENTARY PROBLEMS 77

Exercise 6.5.6 (Minkowski) For x = (x1 , . . . , xn ), let

n
,
Li (x) = aij xj , 1 ≤ i ≤ n,
j=1

be n linear forms with real coefficients. Let C be the domain defined by

|Li (x)| ≤ λi , 1 ≤ i ≤ n.

Show that if λ1 · · · λn ≥ |det A| where A = (aij ), then C contains a nonzero

lattice point.

Exercise 6.5.7 Suppose that among the n linear forms above, Li (x), 1 ≤ i ≤ r1
are real (i.e., aij ∈ R), and 2r2 are not real (i.e., some aij may be nonreal).
Further assume that

Lr1 +r2 +j = Lr1 +j , 1 ≤ j ≤ r2 .

That is,
n
,
Lr1 +r2 +j (x) = ar1 +j,k xk , 1 ≤ j ≤ r2 .
k=1

Now let C be the convex, bounded symmetric domain defined by

|Li (x)| ≤ λi , 1 ≤ i ≤ n,

with λr1 +j = λr1 +r2 +j , 1 ≤ j ≤ r2 . Show that if λ1 · · · λn ≥ |det A|, then C

contains a nonzero lattice point.

Exercise 6.5.8 Using the previous result, deduce that if K is an algebraic num-
ber field
8with discriminant dK , then every ideal class contains an ideal b satisfying
N b ≤ |dK |.

Exercise 6.5.9 Let Xt consist of points

(x1 , . . . , xr , y1 , z1 , . . . , ys , zs )

in Rr+2s where the coordinates satisfy

8 8
|x1 | + · · · + |xr | + 2 y12 + z12 + · · · + 2 ys2 + zs2 < t.

Show that Xt is a bounded, convex, symmetric domain.

Exercise 6.5.10 In the previous question, show that the volume of Xt is

2r−s π s tn
,
n!
where n = r + 2s.
78 CHAPTER 6. THE IDEAL CLASS GROUP

Exercise 6.5.11 Let C be a bounded, symmetric, convex domain in Rn . Let

a1 , . . . , an be linearly independent vectors in Rn . Let A be the n × n matrix
whose rows are the ai ’s. If

vol(C) > 2n |det A|,

show that there exist rational integers x1 , . . . , xn (not all zero) such that

x1 a1 + · · · + xn an ∈ C.

Exercise 6.5.12 (Minkowski’s Bound) Let K be an algebraic number field

of degree n over Q. Show that each ideal class contains an ideal a satisfying
$ %r
n! 4 2
Na ≤ n |dK |1/2 ,
n π

where r2 is the number of pairs of complex embeddings of K, and dK is the

discriminant.

Exercise 6.5.13 Show that if K #= Q, then |dK | > 1. Thus, by Dedekind’s

theorem, in any nontrivial extension of K, some prime ramifies.

Exercise 6.5.14 If K and L are algebraic number fields such that dK and dL
are coprime, show that K ∩ L = Q. Deduce that

[KL : Q] = [K : Q][L : Q].

√
Exercise 6.5.15 Use Minkowski’s bound to show that Q( 5) has class number
1.
√
Exercise 6.5.16 Using Minkowski’s bound, show that Q( −5) has class number
2.
√ √
Exercise
√ 6.5.17 Compute the class numbers of the fields Q( 2), Q( 3), and
Q( 13).
√
Exercise 6.5.18 Compute the class number of Q( 17).
√
Exercise 6.5.19 Compute the class number of Q( 6).
√ √ √ √
Exercise 6.5.20 Show that the fields Q( −1), Q( −2), Q( −3), and Q( −7)
each have class number 1.

Exercise 6.5.21 Let K be an algebraic number field of degree n over Q. Prove

that
- π .n $ nn %2
|dK | ≥ .
4 n!

Exercise 6.5.22 Show that |dK | → ∞ as n → ∞ in the preceding question.

6.5. SUPPLEMENTARY PROBLEMS 79

Exercise 6.5.23 (Hermite) Show that there are only finitely many algebraic
number fields with a given discriminant.

Exercise 6.5.24 Let p be a prime ≡ 11 (mod 12). If p > 3n , show that the
√
ideal class group of Q( −p) has an element of order greater than n.

Exercise 6.5.25 Let K = Q(α) where α is a root of the polynomial f (x) =

x5 − x + 1. Prove that Q(α) has class number 1.
√
Exercise 6.5.26 Determine the class number of Q( 14).

Exercise 6.5.27 If K is an algebraic number field of finite degree over Q with

dK squarefree, show that K has no non-trivial subfields.
Chapter 7

Quadratic Reciprocity

The equation x2 ≡ a (mod p), where p is some prime, provides the starting
point for our discussion on quadratic reciprocity. We can ask whether there
exist solutions to the above equation. If yes, how do these solutions depend
upon a? upon p? Gauss developed the theory of quadratic reciprocity to
answer these questions. His solution is today called the Law of Quadratic
Reciprocity. Gauss, however, christened his result Theorema Auruem, the
Golden Theorem.
In this chapter, we will be examining this interesting facet of number
theory. We will begin with some of the basic properties of reciprocity. We
will then take a brief trip into the realm of Gauss sums, which will provide
us with the necessary tools to prove the Law of Quadratic Reciprocity.
Finally, once we have developed this Golden Theorem, we will show its
usefulness in the study of quadratic fields, as well as primes in certain
arithmetic progressions.

7.1 Preliminaries
In this section, we would like to search for solutions to equations of the
form x2 ≡ a (mod p), where p is prime. We will discover that quadratic
reciprocity gives us a means to determine if any solution exists.
In order to appreciate the usefulness of quadratic reciprocity, let us
consider how we would tackle the congruence

x2 ≡ −1 (mod 5).

The naive method would be to take all the residue classes in (Z/5Z) and
square them. We would get 02 ≡ 0, 12 ≡ 1, 22 ≡ 4, 32 ≡ 4, and 42 ≡ 1.
Since 4 ≡ −1 (mod 5), we have found two solutions to the above equation,
namely 2 and 3. This brute force method works well for small primes but
becomes impractical once the size of the numbers gets too large. Thus

81
82 CHAPTER 7. QUADRATIC RECIPROCITY

it would be nice to have a more accessible method to determine solutions.

The following exercise shows us a way to determine if there exists a solution
when p is fixed. However, determining solutions to the congruence is still
a difficult problem.

Exercise 7.1.1 Let p be a prime and a #= 0. Show that x2 ≡ a (mod p) has a

solution if and only if a(p−1)/2 ≡ 1 (mod p).

Notice that Exercise 7.1.1 merely provides us with a means of deter-

mining whether a solution exists and gives us no information on how to
actually find a square root of a (mod p).
Exercise 7.1.1 works very well for a fixed p. Suppose, however, we wish
to fix a and vary p. What happens in this case? This question motivates
the remainder of our discussion on quadratic reciprocity.
Definition. The Legendre symbol (a/p), with p prime, is defined as follows:

2
! " 
a  1 if x ≡ a (mod p) has a solution,
= −1 if x2 ≡ a (mod p) has no solution,
p 

0 if p | a.

If (a/p) = 1, we say that a is a quadratic residue mod p. If (a/p) = −1, a

is said to be a quadratic nonresidue mod p.

Exercise 7.1.2 Using Wilson’s theorem and the congruence

k(p − k) ≡ −k2 (mod p),

compute (−1/p) for all primes p.

Remark. One of the interesting results of this exercise is that we can now
determine which finite fields Fp , for p prime, have an element that acts like
√
−1. For example, if p = 5, then p ≡ 1 (mod 4), and so, (−1/p) = 1.
So there exists an element a ∈ F5 such that a2 = −1. However, 7 ≡ 3
(mod 4), so F7 can have no element that is the square root of −1.
Before going any further, we will determine some properties of the Leg-
endre symbol.

Exercise 7.1.3 Show that

$ %
a
a(p−1)/2 ≡ (mod p).
p

Exercise 7.1.4 Show that

$ % $ %$ %
ab a b
= .
p p p

Exercise 7.1.5 If a ≡ b (mod p), then (a/p) = (b/p).

7.1. PRELIMINARIES 83

Exercises 7.1.3 to 7.1.5 give some of the basic properties of the Legen-
dre symbol that we will exploit throughout the remainder of this chapter.
Notice that Exercise 7.1.4 shows us that the product of two residues mod p
is again a residue mod p. As well, the product of two quadratic nonresidues
mod p is a quadratic residue mod p. However, a residue mod p multiplied
by a nonresidue mod p is a nonresidue mod p.

Theorem 7.1.6 For all odd primes p,

! " ;
2 1 if p ≡ ±1 (mod 8),
=
p −1 if p ≡ 3, 5 (mod 8).
√
Proof. To exhibit this result, we will work in the field Q(i), where i = −1.
Notice that the ring of integers of this field is Z[i]. We wish to find when
there exist solutions to x2 ≡ 2 (mod p). We will make use of Exercise 7.1.1
which tells us there exists a solution if and only if 2(p−1)/2 ≡ 1 (mod p).
Working in Z[i], we observe that

(1 + i)2 = 1 + 2i + i2 = 2i.

Also, for p prime,

! " ! "
p p 2
(1 + i) = 1 +
p
i+ i + · · · + ip .
1 2

Considering the above equation mod pZ[i], we get

(1 + i)p ≡ 1 + ip (mod pZ[i]).

But we also observe that

(1 + i)p = (1 + i)(1 + i)p−1

= (1 + i)((1 + i)2 )(p−1)/2
= (1 + i)(2i)(p−1)/2
= i(p−1)/2 (1 + i)2(p−1)/2 .

So,

i(p−1)/2 (1 + i)2(p−1)/2 ≡ 1 + ip (mod pZ[i]). (7.1)

We now consider the various possiblities for p (mod 8).

If p ≡ 1 (mod 8), then ip = i. As well, i(p−1)/2 = 1. So, equation (7.1)
becomes
(1 + i) ≡ (1 + i)2(p−1)/2 (mod pZ[i]),
which implies
1 ≡ 2(p−1)/2 (mod pZ[i]).
84 CHAPTER 7. QUADRATIC RECIPROCITY

So, 1 ≡ 2(p−1)/2 (mod p), and thus (2/p) = 1 by Exercise 7.1.1.

If p ≡ −1 (mod 8), then ip = −i. As well, i(p−1)/2 = −i. So, (7.1)
becomes
(1 − i) ≡ −i(1 + i)2(p−1)/2 (mod pZ[i]),
which implies
1 ≡ 2(p−1)/2 (mod pZ[i]).
Again, we have 1 ≡ 2(p−1)/2 (mod p), and thus (2/p)=1.
If p ≡ 3 (mod 8), then ip = −i. Also, i(p−1)/2 = i. So, the above
equation becomes

(1 − i) ≡ i(1 + i)2(p−1)/2 (mod pZ[i]),

−i(1 + i) ≡ i(1 + i)2(p−1)/2 (mod pZ[i]),
(p−1)/2
−1 ≡ 2 (mod pZ[i]).

Since 1 !≡ 2(p−1)/2 (mod p), (2/p) = −1.

Finally, if p ≡ 5 (mod 8), then ip = i. As well, i(p−1)/2 = −1. From
this, it follows that

(1 + i) ≡ −1(1 + i)2(p−1)/2 (mod pZ[i]),

−1 ≡ 2(p−1)/2 (mod pZ[i]).

Hence, (2/p) = −1, thus completing the proof. !

The above result can be restated as
! "
2 2
= (−1)(p −1)/8
p
for odd primes p.

Exercise 7.1.7 Show that the number of quadratic residues mod p is equal to
the number of quadratic nonresidues mod p.
&p−1
Exercise 7.1.8 Show that a=1 (a/p) = 0 for any fixed prime p.

The proof of the Law of Quadratic Reciprocity to be given does not

originate with Gauss, but is of a later date. The proof, however, makes
use of Gauss sums, and as a result, we will make a brief detour to describe
these functions.

7.2 Gauss Sums

Definition. Let p be a prime and let ζp be a primitive pth root of unity.
We define the Gauss Sum as follows:
* !a"
S= ζ a,
p p
a mod p
7.2. GAUSS SUMS 85

where (a/p) is the Legendre symbol.

This sum has some interesting properties. We explore some of them

below.

Theorem 7.2.1 For S as defined above,

! "
2 −1
S = p.
p

Proof. From the definition of a Gauss sum, we have

  
* !a" * !b"
S2 =  ζ a  ζ b .
p p p p
a mod p b mod p

By applying Exercise 7.1.4, we can simplify the above to get

* ! ab "
2
S = ζpa+b .
p
a,b

We now make a substitution by letting b = ca, where (c, p) = 1. Thus,

* * ! a2 c "
2
S = ζpa(1+c) .
p
(a,p)=1 (c,p)=1

Again, using Exercise 7.1.4, we get

* * !c"
2
S = ζ a(1+c)
p p
(a,p)=1 (c,p)=1
 
* !c" *
=  ζpa(1+c)  .
p
(c,p)=1 (a,p)=1

Observe that (1 + c, p) = 1 or (1 + c, p) = p. Since (c, p) = 1, the second

case will only happen if c = p − 1. But then, if (1 + c, p) = 1, we will have
*
ζpa(1+c) = ζp(1+c) + ζp2(1+c) + · · · + ζp(p−1)(1+c) = −1.
(a,p)=1

But (1 + c, p) = p implies that

*
ζpa(1+c) = 1 + 12 + · · · + 1p−1 = p − 1.
(a,p)=1
86 CHAPTER 7. QUADRATIC RECIPROCITY

Thus
 
* !c" *
S2 =  ζpa(1+c) 
p
(c,p)=1 (a,p)=1
* !c" !
p−1
"
= (−1) + (p − 1)
p p
1≤c≤p−2
* ! c " ! −1 "
= (−1) + (p − 1).
p p
1≤c≤p−2

But ! " ! " ! "

* c * c −1
= − .
p p p
1≤c≤p−2 1≤c≤p−1

From Exercise 7.1.8, we know that the first term on the right-hand side
must be equal to 0. So,
* ! c " ! −1 "
2
S = (−1) + (p − 1)
p p
1≤c≤p−2
3 ! "4 ! "
−1 −1
= (−1) − + (p − 1)
p p
! " ! "
−1 −1
= + (p − 1)
p p
! "
−1
= p.
p
E F
But now we have shown the desired result, namely, S 2 = −1
p p. !

In the next exercise, we are going to prove an important identity that

we will utilize in proving the law of quadratic reciprocity.

Exercise 7.2.2 Show that

$ %
q
Sq ≡ S (mod q),
p

where q and p are odd primes.

7.3 The Law of Quadratic Reciprocity

We are now in a position to prove the Theorema Auruem, which we do in
this section. We also demonstrate how to use this beautiful result.
7.3. THE LAW OF QUADRATIC RECIPROCITY 87

Theorem 7.3.1 (Law of Quadratic Reciprocity) Let p and q be odd

primes. Then ! " ! "
p q p−1 q−1
= (−1) 2 · 2 .
q p

Proof. From Exercise 7.2.2, we have

! "
q
q
S ≡ S (mod q).
p

Thus, cancelling out an S from both sides will give us

! "
q
S q−1
≡ (mod q).
p

Since q is odd, q − 1 must be divisible by 2. So

3 ! "4(q−1)/2
−1
S q−1 = (S 2 )(q−1)/2 = p .
p

The last equality follows from Theorem 7.2.1. Thus,

! " 3 ! "4(q−1)/2
q −1
≡ p (mod q).
p p

From Exercise 7.1.2, (−1/p) = (−1)(p−1)/2 . We substitute this into the

above equation to get
! "
q q−1 p−1 q−1
≡ p 2 (−1) 2 · 2 (mod q).
p

Exercise 7.1.3 tells us that p(q−1)/2 ≡ (p/q) (mod q). So,

! " ! "
q p p−1 q−1
≡ (−1) 2 · 2 (mod q).
p q

But both sides only take on the value ±1, and since q ≥ 3, the congruence
can be replaced by an equals sign. This gives us
! " ! "
p q p−1 q−1
= (−1) 2 · 2 .
q p
!
With this result, we can answer the question we asked at the beginning
of this chapter. That is, if we fix some a, for what primes p will x2 ≡ a
(mod p) have a solution? Expressed in terms of the Legendre symbol, we
want to know for which p will (a/p) = 1. We know from Exercise 7.1.4 that
88 CHAPTER 7. QUADRATIC RECIPROCITY

the Legendre symbol is multiplicative. So, we can factor a as a = q1e1 · · · qnen .

Thus ! " ! " e1 ! " en
a q1 qn
= ··· .
p p p
So the question is reduced to evaluating (q/p) for each prime q. Note that
we already know how to solve (−1/p) and (2/p). Thus, all that needs to be
done is to evaluate (q/p) where q is an odd prime. The next exercise helps
us to determine this.

Exercise 7.3.2 Let q be an odd prime. Prove:

(a) If q ≡ 1 (mod 4), then q is a quadratic residue mod p if and only if p ≡ r
(mod q), where r is a quadratic residue mod q.
(b) If q ≡ 3 (mod 4), then q is a quadratic residue mod p if and only if p ≡ ±b2
(mod 4q), where b is an odd integer prime to q.

The next exercise will demonstrate how to use Exercise 7.3.2 to compute
(q/p) in the special cases q = 5, 7.
- . - .
5 7
Exercise 7.3.3 Compute p
and p
.

7.4 Quadratic Fields

In this section, we will focus on quadratic fields, that is, all algebraic number
fields K such that [K : Q] = √ 2. It can be shown that all quadratic extensions
can be written as K = Q( d), where d is some squarefree integer.
With every algebraic number field comes an associated ring of integers,
OK . Suppose that p ∈ Z, and p is prime. We can let pOK be the ideal of
OK generated by p. Since OK is a Dedekind domain (see Chapter 5), every
ideal can be written as a product of prime ideals, i.e., pOK = ℘e11 · · · ℘err .
However, because K is a quadratic extension,

p2 = N (pOK ) = N (℘1 )e1 · · · N (℘r )er .

So N (℘) = p, or N (℘) = p2 . But then, we have three possibilities:

(1) pOK = ℘℘$ , where ℘ != ℘$ ;

(2) pOK = ℘2 ; and

(3) pOK = ℘.

If (1) is true, we say that p splits. When case (2) occurs, we say that
p ramifies. Finally, if (3) occurs, we say that p is inert, i.e., it stays prime.
In the next exercises, we will see that we can determine which case occurs
by using quadratic reciprocity.
7.4. QUADRATIC FIELDS 89

√
Exercise 7.4.1 Find the discriminant of K = Q( d) when:
(a) d ≡ 2, 3 (mod 4); and
(b) d ≡ 1 (mod 4).

Remark. From the above exercise it follows √ that if m = dK is the dis-

criminant of a quadratic field K, then 1, (m + m)/2 will always form an
integral basis.

Theorem 7.4.2 Assume p is an odd prime. Then (d/p) = 1 if and only if

pOK = ℘℘$ , where ℘ != ℘$ , and ℘ prime.
Proof. ⇒ From
√ our assumption,√we have a2 ≡ d (mod p) for some a. Let
℘ = (p, a + d) and ℘$ = (p, a − d).
We claim that pOK = ℘℘$ .
√ √
℘℘$ = (p, a + d)(p, a − d)
√ √
= (p2 , p(a + d), p(a − d), a2 − d)
√ √
= (p)(p, a + d, a − d, (a2 − d)/p)
= (p).
The last equality holds because 2a and p are both elements √ of the
√ second
ideal. But (2a, p) = 1. From this, it follows that 1 ∈ (p, a+ d, a− d, (a2 −
d)/p). It is clear that ℘ != ℘$ because if they were equal, then 2a and p would
be in ℘, from which it follows that ℘ = OK , which is false. Since the norm
of pOK is p2 , N (℘) must divide p2 . Since ℘ != (1), N (℘) != 1. Also, it
cannot be p2 because then N (℘$ ) = 1, which is false. So, both ℘ and ℘$
have norm p, and thus, they must be prime. √
⇐ In the comments after Exercise 7.4.1 we noted that {1, (m + m)/2}
always forms an integral basis of OK where m = d if d ≡ 1 (mod 4) and
m = 4d if d ≡ 2, 3 (mod 4). Since√ ℘℘$ = pOK , there must exist a ∈ ℘, but
a !∈ pOK . So, a = x + y(m + m)/2, where x, y ∈ Z, but p does not divide
both x and y. Now, consider aOK , the ideal generated by a. We can write
aOK = ℘q, q ⊆ OK . Now, taking the norms of both sides, we discover that
N (℘) = p must divide
1E 1
1 ym F2 y 2 m 11
N (aOK ) = 11 x + − .
2 4 1
So, (2x + ym)2 ≡ y 2 m (mod p). If p | y, then p | (2x + ym)2 . But then
p | 2x, and since p is odd, p | x. This contradicts the fact that p did not
divide both x and y. So p does not divide y, and since Z/pZ is a field,
(2x + ym)2
≡m (mod p).
y2
But then we have found some z such that z 2 ≡ m (mod p). Since m = d
or m = 4d, then (d/p) = 1. !
90 CHAPTER 7. QUADRATIC RECIPROCITY

Exercise 7.4.3 Assume that p is an odd prime. Show that (d/p) = 0 if and only
if pOK = ℘2 , where ℘ is prime.

Exercise 7.4.4 Assume p is an odd prime. Then (d/p) = −1 if and only if

pOK = ℘, where ℘ is prime.

What we have shown is a method for determining what happens to an

odd prime in a quadratic field that utilizes the Legendre symbol. We have
yet to answer what happens to p if p = 2. An analogous result holds for
this case.

Theorem 7.4.5 Suppose p = 2. Then:

(a) 2OK = ℘2 , ℘ prime if and only if 2 | dK ;
(b) 2OK = ℘℘$ , ℘ prime if and only if d ≡ 1 (mod 8) and 2 ! dK ; and
(c) 2OK = ℘, ℘ prime if and only if d ≡ 5 (mod 8) and 2 ! dK .
dK , then d ≡ 2, 3 (mod 4). If d ≡ 2 (mod 4), then we
Proof. (a) ⇐ If 2 | √
claim that (2) = (2, d)2 . Note that
√ √ √
(2, d)2 = (4, 2 d, d) = (2)(2, d, d/2).

Since d is squarefree, then 2 and d/2 are relatively

√ prime and thus the
second ideal above is actually OK . So (2) = (2, √d)2 .
If d ≡ 3 (mod 4) we claim that (2) = (2, 1 + d)2 , since
! "
√ √ √ √ 1+d √
(2, 1 + d)2 = (4, 2 + 2 d, 1 + d + 2 d) = (2) 2, 1 + d, + d .
2
√ √
Now we note that 1 + d and (1 + d)/2 + d are relatively prime, and so
the second ideal is OK .
⇒ We consider dK , which we know is congruent to either 0 or 1 mod
4. Suppose
√ that dK ≡ 1 (mod 4). Then OK is generated as a Z-module by 2
1, (1 + d)/2. There √ must exist some element a in ℘ which is not in ℘ .
So a = m + n(1 + d)/2 where we can assume that m and n are either 0
or 1, since for any α ∈ OK , a + 2α is in ℘ but not in ℘2 .
Now, if n = 0, then m != 0 because otherwise a = 0 and is obviously in
(2). But if m = 1, then a = 1 and a ∈ / ℘. So, n = 1 and m = 0 or 1. We
know that a2 ∈ (2), and
/ √ 02
2 1+ d
a = m+
2
√
2 d−1 1+ d
= m + + (2m + 1) ∈ (2).
4 2
But 2m + 1 is odd and so a2 ∈
/ (2), and we have arrived at a contradiction.
7.5. PRIMES IN SPECIAL PROGRESSIONS 91

We conclude that dK ≡ 0 (mod 4), and so clearly 2 | dK .

(b) ⇐ Suppose that d ≡ 1 (mod 8). Then √ clearly from√the previous
problem 2 ! dK . We claim that (2) = (2, (1 + d)/2)(2, (1 − d)/2). Note
that
/ √ 0/ √ 0 / √ √ 0
1+ d 1− d 1+ d 1− d 1−d
2, 2, = (2) 2, , , .
2 2 2 2 8
√ √
But the second ideal is just OK since it contains 1 = (1+ d)/2+(1− d)/2.
⇒ Now suppose that (2) splits in OK . We know from part (a) that
d ≡ 1 (mod 4). If (2) = ℘℘$ , then N (℘) = 2. There exists √ an element a
which is in ℘ but not in ℘℘$ = (2). Then a = m + n(1 + d)/2 where not
both m, n are even. Therefore 2 divides the norm of the ideal generated by
a, and 1 1
1 (2m + n)2 n2 d 11
N ((a)) = |N (a)| = 11 − .
4 4 1
So (2m+n)2 ≡ n2 d (mod 8). We know that 2 ! d. So suppose that n is even,
and further suppose that n = 2n1 where n1 is odd. Then 2 | (m+n1 )2 +n21 d,
and since 2 ! n21 d, then 2 ! (m + n1 )2 , which implies that m is even. But we
assumed that not both m and n were even. Now suppose that 4 | n. Then
4 | (2m + n) and so m is even, a contradiction. Then n must be odd, and
we can find an integer n2 such that nn2 ≡ 1 (mod 8).
Then d ≡ n22 (2m+n)2 (mod 8), and since 2 ! d, we conclude n2 (2m+n)
is odd, and d ≡ 1 (mod 8), as desired.
(c) Just as in Exercise 7.4.4, this follows directly from parts (a) and (b),
since if 2 ! dK and d !≡ 1 (mod 8), then d ≡ 5 (mod 8). We know that (2)
cannot split or ramify in this case, so it must remain inert. !

7.5 Primes in Special Progressions

Another interesting application of quadratic reciprocity is that it can be
used to show there exist infinitely many primes in certain arithmetic pro-
gressions. In the next two exercises, we imitate Euclid’s proof for the exis-
tence of an infinite number of primes to show that there are infinitely many
primes in the following two arithmetic progressions, 4k + 1 and 8k + 7.

Exercise 7.5.1 Show that there are infinitely many primes of the form 4k + 1.

Exercise 7.5.2 Show that there are infinitely many primes of the form 8k + 7.

The results we have just derived are just a special case of a theorem
proved by Dirichlet. Dirichlet proved that if l and k are coprime inte-
gers, then there must exist an infinite number of primes p such that p ≡ l
(mod k). What is interesting about these two exercises, however, is the fact
that we used a proof similar to Euclid’s proof for the existence of an infinite
92 CHAPTER 7. QUADRATIC RECIPROCITY

number of primes. An obvious question to ask is whether questions about

all arithmetic progressions can be solved in a similar fashion.
The answer, sadly, is no. However, not all is lost. It can be shown that if
l2 ≡ 1 (mod k), then we can apply a Euclid-type proof to show there exist
an infinite number of primes p such that p ≡ l (mod k). (See Schur [S].
For instance, Exercises 1.2.6 and 5.6.10 give Euclid-type proofs for p ≡ 1
(mod k) using cyclotomic polynomials.) Surprisingly, the converse of this
statement is true as well. The proof is not difficult, but involves some
Galois Theory. It is due to Murty [Mu].
We can restate our two previous exercises as follows:
(1) Are there infinitely many primes p such that p ≡ 1 (mod 4)?
(2) p ≡ 7 (mod 8)?
From what we have just discussed, we observe that we can indeed apply
a Euclid-type proof since 12 ≡ 1 (mod 4) and 72 ≡ 1 (mod 8).

Exercise 7.5.3 Show that p ≡ 4 (mod 5) for infinitely many primes p.

In their paper [BL], Bateman and Low show that if l is an integer

relatively prime to 24, then there are infinitely many primes p such that
p ≡ l (mod 24). Their proof makes use of the interesting fact that every
integer l relatively prime to 24 has the property l2 ≡ 1 (mod 24). (All the
integers relatively prime to 24 are 1, 5, 7, 11, 13, 17, 19, and 23. A quick
mental calculation will show you that the statement is true.) Because of
this fact, they can use a proof similar to Euclid’s.
Their proof relies on the ability to “cook up” a specific polynomial
f (x) ∈ Z[x]. This polynomial is created in such a way so that we can use
quadratric reciprocity. Notice that in our exercises there is also some poly-
nomial sitting in the background. In Exercise 7.5.1, we used the polynomial
f (x) = 4x2 + 1. In Exercise 7.5.2, f (x) = 16x2 − 2 was used, and finally,
in the previous exercise, f (x) = 25x2 − 5. Not all the polynomials used
are as simple as the ones we used. The next example uses a fourth degree
polynomial.

Example 7.5.4 Show there are an infinite number of primes in the arith-
metic progession 15k + 4.

Solution. Since 42 ≡ 1 (mod 15), we can use a Euclid-type proof. We will

start with a couple of observations about the polynomial

f (x) = x4 − x3 + 2x2 + x + 1.
7.5. PRIMES IN SPECIAL PROGRESSIONS 93

First, we note that it can be factored in the following three ways:

E x F2
f (x) = x2 −
− 1 + 15 2
4 x , (7.2)
2
E x F2
f (x) = −x2 + − 12 + 34 (x + 1)2 , (7.3)
2
E x 3 F2 5
2
f (x) = −x + − 2 − 4 (x − 1)2 . (7.4)
2
We note by (7.2) that if p divides f (x), then −15 is a quadratic residue
mod p. By quadratic reciprocity,
EpF EpF
= 1.
3 5
So, there will be a solution only if (p/3) = 1 and (p/5) = 1 or if they
both equal −1. The first case will happen if p ≡ 1 (mod 3) and p ≡ 1, 4
(mod 5). The second happens if p ≡ 2 (mod 3) and p ≡ 2, 3 (mod 5). So,

EpF   1 if p ≡ 1, 2, 4, 8 (mod 15),
= −1 if p ≡ 7, 11, 13 (mod 15),
15 

0 otherwise.

From equation (7.3), we see that (−3/p) = 1. Using Exercise 7.3.2,


! " 
3  1 if p ≡ 1, 11 (mod 12),
= −1 if p ≡ 5, 7 (mod 12),
p 

0 otherwise.

So, since we already know what (−1/p) is, we find that


!
−3
"  1 if p ≡ 1, 7 (mod 12),
= −1 if p ≡ 5, 11 (mod 12),
p 

0 otherwise.

Finally, equation (7.4) tells us that (5/p) = 1. But we know this only
happens when p ≡ 1, 4 (mod 5).
When we combine all these results, we find that any prime divisor of
f (x) must be congruent to either 1 (mod 15) or 4 (mod 15).
We can now begin the Euclid-type proof. Suppose that there were only
a finite number of primes such that p ≡ 4 (mod 15). Let p1 , . . . , pn be
these primes. We now consider the integer d = f (15p1 p2 · · · pn + 1). From
what we have just said, d is divisible by some prime p such that p ≡ 1, 4
(mod 5). Not all the prime divisors have the form p ≡ 1 (mod 5). This
follows from the fact that d = f (15p1 · · · pn +1) = 15p1 · · · pn g(p1 · · · pn )+4,
where g(x) is some polynomial. So, there is a divisor p such that p ≡ 4
94 CHAPTER 7. QUADRATIC RECIPROCITY

(mod 15). But p cannot be in p1 , p2 , . . . , pn because when they divide d,

they leave a remainder of 4. This gives us the needed contradiction.

One item that we did not discuss is how to derive a polynomial that we
can use in a Euclid-style proof. One method involves a little ingenuity and
some luck. By playing around with some equations, you may happen upon
such a polynomial. Murty, on the other hand, describes [Mu] an explicit
construction for these polynomials. Though interesting in their own right,
we will refrain from going into any detail about these polynomials.

7.6 Supplementary Problems

Exercise 7.6.1 Compute (11/p).

Exercise 7.6.2 Show that (−3/p) = 1 if and only if p ≡ 1 (mod 3).

Exercise 7.6.3 If p ≡ 1 (mod 3), prove that there are integers a, b such that
p = a2 − ab + b2 .

Exercise 7.6.4 If p ≡ ±1 (mod 8), show that there are integers a, b such that
a2 − 2b2 = ±p.

Exercise 7.6.5 If p ≡ ±1 (mod 5), show that there are integers a, b such that
a2 + ab − b2 = ±p.

Exercise 7.6.6 Let p be a prime greater than 3. Show that:

(a) (−2/p) = 1 if and only if p ≡ 1, 3 (mod 8);
(b) (3/p) = 1 if and only if p ≡ 1, 11 (mod 12);
(c) (−3/p) = 1 if and only if p ≡ 1 (mod 6);
(d) (6/p) = 1 if and only if p ≡ 1, 5, 19, 23 (mod 24); and
(e) (−6/p) = 1 if and only if p ≡ 1, 5, 7, 11 (mod 24).

Exercise 7.6.7 If p is a prime dividing n8 − n4 + 1, show that p is coprime to

n2 , n3 + n, and n3 − n. Deduce that there are integers a, b, c such that

an2 ≡ 1 (mod p),

3
b(n + n) ≡ 1 (mod p),
c(n3 − n) ≡ 1 (mod p).

Exercise 7.6.8 Let the notation be as in Exercise 7.6.7 above.

(a) Observe that x8 − x4 + 1 = (x4 − 1)2 + (x2 )2 . Deduce that

(an4 − a)2 + 1 ≡ 0 (mod p).

7.6. SUPPLEMENTARY PROBLEMS 95

(b) Observe that x8 − x4 + 1 = (x4 + x2 + 1)2 − 2(x3 + x)2 . Deduce that

(bn4 + bn2 + b)2 − 2 ≡ 0 (mod p).

(c) Observe that x8 − x4 + 1 = (x4 − x2 + 1)2 + 2(x3 − x)2 . Deduce that

(cn4 − cn2 + c)2 + 2 ≡ 0 (mod p).

(d) From x8 − x4 + 1 = (x4 + 1)2 − 3(x2 )2 , deduce that

(an4 + a)2 ≡ 3 (mod p).

(e) From x8 − x4 + 1 = (x4 − 12 )2 + 3( 12 )2 , deduce that

(2n4 − 1)2 ≡ −3 (mod p).

(f) From x8 − x4 + 1 = (x4 + 3x2 + 1)2 − 6(x3 + x)2 , deduce that

(bn4 + 3bn2 + b)2 ≡ 6 (mod p).

(g) From x8 − x4 + 1 = (x4 − 3x2 + 1)2 + 6(x3 − x)2 , deduce that

(cn4 − 3cn2 + c)2 ≡ −6 (mod p).

Exercise 7.6.9 From Exercises 7.6.7 and 7.6.8, deduce that any prime divisor p
of n8 − n4 + 1 satisfies
$ % $ % $ % $ % $ % $ %
−1 2 −2 3 6 −6
= = = = = = 1.
p p p p p p
Deduce that p ≡ 1 (mod 24). Prove that there are infinitely many primes p ≡ 1
(mod 24).

[Exercises 7.6.7, 7.6.8, 7.6.9 were suggested by a paper of P. Bateman

and M.E. Low, Prime Numbers in Arithmetic Progressions with Difference
24, Amer. Math. Monthly, 72 (1965), 139–143.]

Exercise 7.6.10 Show that the number of solutions of the congruence

x2 + y 2 ≡ 1 (mod p),

with 0 < x < p, 0 < y < p, (p an odd prime) is even if and only if p ≡ ±3
(mod 8).

Exercise 7.6.11 If p is a prime such that p − 1 = 4q with q prime, show that 2

is a primitive root mod p.

Exercise 7.6.12 (The Jacobi Symbol) Let Q be a positive odd number. We

can write Q = q1 q2 · · · qs where the qi are odd primes, not necessarily distinct.
Define the Jacobi symbol
$ % # s $ %
a a
= .
Q j=1
qi

If Q and Q( are odd and positive, show that:

96 CHAPTER 7. QUADRATIC RECIPROCITY

(a) (a/Q)(a/Q( ) = (a/QQ( ).

(b) (a/Q)(a( /Q) = (aa( /Q).
(c) (a/Q) = (a( /Q) if a ≡ a( (mod Q).

Exercise 7.6.13 If Q is odd and positive, show that

$ %
−1
= (−1)(Q−1)/2 .
Q
2
Exercise 7.6.14 If Q is odd and positive, show that (2/Q) = (−1)(Q −1)/8
.

Exercise 7.6.15 (Reciprocity Law for the Jacobi Symbol) Let P and Q
be odd, positive, and coprime. Show that
$ %$ %
P Q P −1 Q−1
= (−1) 2 · 2 .
Q P

Exercise 7.6.16 (The Kronecker Symbol) We can define (a/n) for any in-
teger a ≡ 0 or 1 (mod 4), as follows. Define

-a. $ a %   0 if a ≡ 0 (mod 4),
= = 1 if a ≡ 1 (mod 8),
2 −2 

−1 if a ≡ 5 (mod 8).

For general n, write n = 2c n1 , with n1 odd, and define

- a . - a .c $ a %
= ,
n 2 n1
where (a/2) is defined as above and (a/n1 ) is the Jacobi symbol.
Show that if d is the discriminant of a quadratic field, and n, m are positive
integers, then $ % $ %
d d
= for n ≡ m (mod d)
n m
and $ % $ %
d d
= sgn d for n ≡ −m (mod d).
n m

Exercise 7.6.17 If p is an odd prime show that the least positive quadratic
√
nonresidue is less than p + 1.
(It is a famous conjecture of Vinogradov that the least quadratic non-residue
mod p is O(pε ) for any ε > 0.)

Exercise 7.6.18 Show that x4 ≡ 25 (mod 1013) has no solution.

Exercise 7.6.19 Show that x4 ≡ 25 (mod p) has no solution if p is a prime

congruent to 13 or 17 (mod 20).

Exercise 7.6.20 If p is a prime congruent to 13 or 17 (mod 20), show that

x4 + py 4 = 25z 4 has no solutions in integers.
7.6. SUPPLEMENTARY PROBLEMS 97

√
Exercise 7.6.21 Compute the class number of Q( 33).
√
Exercise 7.6.22 Compute the class number of Q( 21).
√
Exercise 7.6.23 Show that Q( −11) has class number 1.
√
Exercise 7.6.24 Show that Q( −15) has class number 2.
√
Exercise 7.6.25 Show that Q( −31) has class number 3.
Chapter 8

The Structure of Units

8.1 Dirichlet’s Unit Theorem

Let K be a number field and OK its ring of integers. An element α ∈ OK
is called a unit if ∃β ∈ OK such that αβ = 1. Evidently, the set of all units
in OK forms a multiplicative subgroup of K ∗ , which we will call the unit
group of K.
In this chapter, we will prove the following fundamental theorem, which
gives an almost complete description of the structure of the unit group of
K, for any number field K.

Theorem (Dirichlet’s Unit Theorem) Let UK be the unit group of

K. Let n = [K : Q] and write n = r1 + 2r2 , where, as usual, r1 and 2r2 are,
respectively, the number of real and nonreal embeddings of K in C. Then
there exist fundamental units ε1 , . . . , εr , where r = r1 + r2 − 1, such that
every unit ε ∈ UK can be written uniquely in the form

ε = ζεn1 1 · · · εnr r ,

where n1 , . . . , nr ∈ Z and ζ is a root of unity in OK . More precisely, if

WK is the subgroup of UK consisting of roots of unity, then WK is finite
and cyclic and UK 1 WK × Zr .

Definition. α ∈ OK is called a root of unity if ∃m ∈ N such that αm = 1.

Exercise 8.1.1 (a) Let K be an algebraic number field. Show that there are
only finitely many roots of unity in K.
(b) Show, similarly, that for any positive constant c, there are only finitely many
α ∈ OK for which |α(i) | ≤ c for all i.

If α is an algebraic integer all of whose conjugates lie on the unit cir-

cle, then α must be a root of unity by the argument in (a). Indeed, the

99
100 CHAPTER 8. THE STRUCTURE OF UNITS

polynomials
n
6
fα,h (x) = (x − α(i)h )
i=1
(i)h
cannot all be distinct since |α | = 1. If fα,h (x) is identical with fα,k (x)
where h < k (say), then the roots must coincide. If αh = αk , then α
is a root of unity and we are done. If not, after a suitable relabelling
we may suppose that α(1)h = α(2)k , α(2)h = α(3)k , . . . , α(n−1)h = α(n)k ,
α(n)h = α(1)k . Therefore,
2
α(1)h = α(2)kh
n n−1
= α(3)k hn−2
= · · · = α(1)k
n

so that again, α is a root of unity.

This is a classical result due to Kronecker.

Exercise 8.1.2 Show that WK , the group of roots of unity in K, is cyclic, of

even order.

Definition.

(i) An (additive) subgroup Γ of Rm is called discrete if any bounded

subset of Rm contains only finitely many elements of Γ.

(ii) Let {γ1 , . . . , γr } be a linearly independent set of vectors in Rm (so

that r ≤ m). The additive subgroup of Rm generated by γ1 , . . . , γr is
called a lattice of dimension r, generated by γ1 , . . . , γr .

Theorem 8.1.3 Any discrete subgroup Γ of Rm is a lattice.

Proof. We prove this by induction on m:

In the trivial case, where Γ = (0), Γ is a lattice of dimension 0. We will
thus, heretofore, assume that Γ != (0).
Suppose first that m = 1, so that Γ ⊆ R.
Let α be a nonzero element of Γ and let A = {λ ∈ R : λα ∈ Γ}. By
hypothesis, the set {γ ∈ Γ : |γ| ≤ |α|} is finite. Then A ∩ [−1, 1] is finite
and contains a least positive element 0 < µ ≤ 1.
Let β = µα and suppose that λβ ∈ Γ, with λ ∈ R. Then

λβ − [λ]β = (λ − [λ])β = (λ − [λ])µα ∈ Γ,

which, by the minimality of µ, implies that λ = [λ], i.e., λ ∈ Z which

implies that Γ = Zβ is a lattice of dimension 1.
Now, suppose that m > 1.
Let {v1 , . . . , vk } be a maximal linearly independent subset of Γ (so that
Γ ⊆ Rv1 +· · ·+Rvk ), let V be the subspace of Rm spanned by {v1 , . . . , vk−1 }
and let Γ0 = Γ ∩ V . Then Γ0 is a discrete subgroup of V 1 Rk−1 (as vector
spaces) so, by the induction hypothesis, is a lattice. That is, there are
8.1. DIRICHLET’S UNIT THEOREM 101

linearly independent vectors w1 , . . . , wl ∈ V such that Γ0 = Zw1 +· · ·+Zwl

and, since v1 , . . . , vk−1 ∈ Γ0 , we must have l = k − 1.
Evidently, {w1 , . . . , wk−1 , wk := vk } is also a maximal linearly indepen-
dent subset of Γ (since span{v1 , . . . , vk−1 } = span{w1 , . . . , wk−1 } = V ).
Let
; k G
*
T = ai wi ∈ Γ : 0 ≤ ai < 1 for 1 ≤ i ≤ k − 1 and 0 ≤ ak ≤ 1 .
i=1

T is bounded, hence finite, by hypothesis. We may therefore choose

5k element x ∈ T with smallest nonzero coefficient ak of wk , say x =
an
i=1 bi wi . Since bk != 0, the set {w1 , . . . , wk−1 , x} is linearly independent.
Moreover, for any γ ∈ Γ, writing γ = c1 w1 + · · · + ck−1 wk−1 + ck x, we see
that there are integers d1 , ..., dk−1 so that
k−1
*
γ $ = γ − [ck ]x − di wi ∈ T.
i=1

Since the coefficient of wk in γ $ is (ck − [ck ])bk < bk , by the minimality of

bk , we must have that ck − [ck ] = 0 so that ck ∈ Z and γ $ ∈ Γ0 ⇒ γ ∈
Γ0 + Zx ⇒ Γ = Γ0 + Zx = Zw1 + · · · + Zwk−1 + Zx is a lattice of dimension
k. !

Below, we will develop the proof of Dirichlet’s Unit Theorem.

Let σ1 , . . . , σr1 , σr1 +1 , σ r1 +1 , . . . , σr1 +r2 , σ r1 +r2 be the real and complex
conjugate embeddings of K in C. Let E = {k ∈ Z : 1 ≤ k ≤ r1 + r2 }. For
k ∈ E, set ;
k if k ≤ r1 ,
k=
k + r2 if k > r1 .
If A ⊆ E, set A = {k : k ∈ A}. Note that E∪E = {k ∈ Z : 1 ≤ k ≤ r1 +2r2 }
and that, if E = A ∪ B is a partition of E, then E ∪ E = (A ∪ A) ∪ (B ∪ B)
is a partition of E ∪ E.

Lemma 8.1.4 (a) Let m, n ∈ Z with 0 < m ≤ n and let ∆ = (dij ) ∈

Mn×m (R). For any integer t > 1, there is a nonzero x = (x1 , . . . , xn ) ∈
Zn with each |xi | ≤ t such that, if y = x∆ = (y1 , . . . , ym ) ∈ Rm , then
each |yi | ≤ ct1−n/m , where c is a constant depending only on the matrix
∆.
(b) Let E = A ∪ B be a partition of E and let m = |A ∪ A|, n = r1 + 2r2 =
[K : Q]. Then there is a constant c, depending only on K, such that,
for t sufficiently large, ∃α ∈ OK such that
c1−m t1−n/m ≤ |α(k) | ≤ ct1−n/m for k ∈ A,
c−m t ≤ |α(k) | ≤ t for k ∈ B.
102 CHAPTER 8. THE STRUCTURE OF UNITS

5n
Proof. (a) Let δ = max1≤j≤m i=1 |dij |. Then, for

0 != x = (x1 , . . . , xn ) ∈ Zn≥0

with each |xi | ≤ t, 1 n 1

1* 1
1 1
|yj | = 1 xi dij 1 ≤ δt.
1 1
i=1

Consider the cube [−δt, δt] ∈ Rm . Let h be an integer ≥ 1 and divide

m

the given cube into hm equal subcubes so that each will have side length
2δt/h. Now, for each x = (x1 , . . . , xn ) ∈ [0, t]n ∩ Zn , y = (y1 , . . . , ym ) ∈
[−δt, δt]m which means that there are (t + 1)n such points y ∈ [−δt, δt]m .
Thus, if hm < (t + 1)n , then two of the points must lie in the same subcube.
That is, for some x$ != x$$ , we have that, if y = (x$ − x$$ )∆ = (y1 , . . . , ym ),
then each |yi | ≤ 2δt/h.
Since t > 1 and n/m ≥ 1, (t + 1)n/m > tn/m + 1 so there exists an
integer h with tn/m < h < (t + 1)n/m (in particular, hm < (t + 1)n ). Then
|yi | ≤ 2δt/h < 2δt1−n/m for each i.
(b) Let {ω1 , . . . , ωn } ⊂ OK be linearly independent over Q and suppose
5n 5n (k)
that (x1 , . . . , xn ) ∈ Zn . If α = j=1 xj ωj , we have α(k) = i=1 xi ωi .
Let k1 , . . . , ku be the elements of A with ki = ki and let l1 , . . . , lv be the
elements of A with li != li , so that m = u + 2v. Let
 (kj )

ωi for 1 ≤ j ≤ u,
dij = Re ωi(lj ) for u < j ≤ u + v,

 (l )
Im ωi j for u + v < j ≤ 2u + v = m,

and let ∆ = (dij ). By (a), there is a nonzero x = (x1 , .5 . . , xn ) ∈ Zn

n
with each |xi | ≤ t such that, if y = x∆, then each |yj | = | i=1 xi dij | ≤
1−n/m
Ct , for some constant C.
For 1 ≤ j ≤ u,
n
* n
* (kj )
yj = xi dij = xi ωi = α(kj ) ,
i=1 i=1

for u < j < u + v, yj = Re α(lj ) and for u + v < j < u + 2v, yj =

Im α(lj ) ⇒ α(lj ) = yj + yj ⇒ |α(lj ) | ≤ 2Ct1−n/m = ct1−n/m . Therefore, for
any k ∈ A ∪ A, |α(k) | ≤ ct1−n/m .
If k ∈ B ∪ B, then each |xi | ≤ t, and therefore
1 1
1* n 1 *n
(k)
1 1
|α | = 1 1 xj ωj 11 ≤ t |ωj | = δt.
1 j=1 1 j=1

Choosing t0 = δ, we see that |α(k) | ≤ t, for all t ≥ t0 .

8.1. DIRICHLET’S UNIT THEOREM 103

On the other hand, for h ∈ B,

n
6
1 ≤ |NK (α)| = |α(l) |
l=1
6 6
= |α(k) | |α(l) | ≤ cm (t1−n/m )m |α(h) |tn−m−1
k∈A∪A l∈B∪B
(h) −m
⇒ |α | ≥ tc .
Similarly, for j ∈ A,
1 ≤ |NK (α)| ≤ |α(j) |(ct1−n/m )m−1 tn−m = |α(j) |cm−1 tm/n−1
⇒ |α(j) | ≥ c1−m t1−n/m . !

Lemma 8.1.5 Let E = A ∪ B be a proper partition of E.

(a) There exists a sequence of nonzero integers {αv } ⊆ OK such that
(k)
|αv(k) | > |αv+1 | for k ∈ A,
(k)
|αv(k) | < |αv+1 | for k ∈ B,
and |NK (αv )| ≤ cm , where c is a positive constant depending only on
K and m = |A ∪ A|.
(b) There exists a unit ε with |ε(k) | < 1, for k ∈ A and |ε(k) | > 1, for
k ∈ B.
Proof. (a) Let t1 be an integer greater than 1 and let {tv } be the sequence
defined recursively by the relation tv+1 = M tv for all v ≥ 1, where M is a
positive constant that will be suitably chosen. By Lemma 8.1.4, for each
v, ∃αv ∈ OK such that
c1−m t1−n/m
v ≤ |αv(k) | ≤ ct1−n/m
v for k ∈ A,
c−m tv ≤ |αv(k) | ≤ tv for k ∈ B.
Now, let κ = min{1, n/m − 1} and choose M such that M κ > cm so that
both M > cm and M n/m−1 > cm . Then, if k ∈ A,
! "1−n/m
tv+1 1−n/m (k)
|αv(k) | ≥ c−m+1 t1−n/m
v = c−m+1
> ctv+1 ≥ |αv+1 |
M
and if k ∈ B, then
tv+1 (k)
|αv(k) | ≤ tv = < c−m tv+1 ≤ |αv+1 |.
M
Also,
6 6
|NK (αv )| = |αv(i) | |αv(j) | ≤ (ct1−n/m
v )m tn−m
v = cm .
i∈A∪A j∈B∪B
104 CHAPTER 8. THE STRUCTURE OF UNITS

(b) Let {αv } ⊆ OK be the sequence of algebraic integers in (a). Define

the sequence of principal ideals Av = (αv ). Then N (Av ) = NK (αv ) ≤ cm
(where N (Av ) := #OK /Av ).
Since there are only finitely many integral ideals Av of bounded norm,
∃µ ∈ N such that, for some v > µ, Av = Aµ which means that αµ = εαv ,
for some unit ε. We conclude that
1 1 ;
1 α(k) 1 < 1 for k ∈ A,
(k) 1 µ 1
|ε | = 1 (k) 1
1 αv 1 > 1 for k ∈ B.
!

Theorem 8.1.6 (Dirichlet’s Unit Theorem) Let UK be the unit group

of K. Let n = [K : Q] and write n = r1 + 2r2 , where, as usual, r1 and
2r2 are, respectively, the number of real and nonreal embeddings of K in C.
Then there exist fundamental units ε1 , . . . , εr , where r = r1 + r2 − 1, such
that every unit ε ∈ UK can be written uniquely in the form

ε = ζεn1 1 · · · εnr r ,

where n1 , . . . , nr ∈ Z and ζ is a root of unity in OK . More precisely, if

WK is the subgroup of UK consisting of roots of unity, then WK is finite
and cyclic and UK 1 WK × Zr .
Proof. Let UK be the unit group of K and consider the homomorphism

f : UK → Rr
ε 9→ (log |ε(1) |, . . . , log |ε(r) |).

We will show that:

(a) ker f = WK ; and
(b) Im f = Γ is a lattice of dim r in Rr .
a) Suppose that ε ∈ ker f , i.e.,

|ε(1) | = · · · = |ε(r1 +r2 −1) | = 1,

(r1 +r2 +1)
⇒ |ε | = · · · = |ε(r1 +2r2 −1) | = 1.

But, since ε ∈ UK ,
n
6
1 = |NK (ε)| = |ε(i) | = |ε(r1 +r2 ) ||ε(r1 +2r2 ) | = |ε(r1 +r2 ) |2 ,
i=1

⇒ |ε(r1 +r2 ) | = |ε(r1 +2r2 ) | = 1.

By Exercise 8.1.1, the number of ε ∈ OK such that |ε(i) | ≤ 1 for all i is

finite. ε must, therefore, have finite order in UK , i.e, εk = 1, for some
positive integer k and so ε ∈ WK .
8.1. DIRICHLET’S UNIT THEOREM 105

(b) If −M < log |ε(i) | < M , for i = 1, . . . , r, then e−M < |ε(i) | < eM
for all i !∈ S = {r1 + r2 , r1 + 2r2 }. But
|NK (ε)| r1 +2(r2 −1)
|ε(r1 +r2 ) |2 = 2 i|
< eM < eM n .
i*∈S |ε

Thus, we have that each |ε(i) | < eM n/2 . By Exercise 8.1.1, there are only
finitely many ε for which this inequality holds. Therefore, any bounded
region in Rm contains only finitely many points of Γ so, by Theorem 8.1.3,
Γ is a lattice of dimension t ≤ r.
By Lemma 8.1.5 (b), we can find for each 1 ≤ i ≤ r, a unit 6i such that
(i) (j)
|6i | > 1 and |6i | < 1 for j != i and 1 ≤ j ≤ r. Let xi be the image of
6i under the map f . We claim that x1 , ..., xr are linearly independent. For
suppose that
c1 x1 + · · · + cr xr = 0,
with the ci ’s not all zero. We may suppose without loss of generality that
c1 > 0 and c1 ≥ cj for 1 ≤ j ≤ r. Then,
r
* r
*
(i) (i)
0= ci log |61 | ≥ c1 log |61 |,
i=1 i=1

so that
r
* (i)
log |61 | ≤ 0.
i=1
Now the product of the conjugates of 61 has absolute value 1. By our choice
(i)
of 61 , we see that |61 | < 1 and we deduce that
r
* (i)
log |61 | > 0,
i=1

which is a contradiction. Thus, UK 1 WK ×Γ, and as shown above, Γ 1 Zr .

!

Exercise 8.1.7 (a) Let Γ be a lattice of dimension n in Rn and suppose that

{v1 , . . . , vn } and {w1 , . . . , wn } are two bases for Γ over Z. Let V and W
be the n × n matrices with rows consisting of the vi ’s and wi ’s, respectively.
Show that |det V | = |det W |. Thus, we can unambiguously define the volume
of the lattice Γ, vol(Γ) = the absolute value of the determinant of the matrix
formed by taking, as its rows, any basis for Γ over Z.
(b) Let ε1 , . . . , εr be a fundamental system of units for a number field K. Show
(i)
that the regulator of K, RK = |det(log |εj |)|, is independent of the choice of
ε1 , . . . , ε r .
√
If ε is a fundamental unit in Q( d), then so are −ε, ε−1 , −ε−1 . Sub-
ject to the constraint ε√> 1, ε is uniquely determined and is called the
fundamental unit of Q( d).
106 CHAPTER 8. THE STRUCTURE OF UNITS

√
Exercise 8.1.8 (a) Show that, for any real quadratic field K = Q( d), where d
is a positive squarefree integer, UK 3 Z/2Z×Z. That is, there is a fundamen-
tal unit ε ∈ UK such that UK = {±εk : k ∈ Z}. Conclude that the equation
x2 − dy 2 = 1 (erroneously dubbed Pell’s equation) has infinitely many integer
solutions for d ≡ 2, 3 mod 4 and that the equation x2 − dy 2 = 4 has infinitely
many integer solutions for d ≡ 1 mod 4.
(b) Let d ≡ 2, 3 (mod 4). Let b be the smallest positive
√ integer such that one of
db2 ± 1 is a square, say a2 , a > 0. Then a + b d is a unit. Show that it is the
fundamental
√ √ unit. Using this algorithm, determine the fundamental units of
Q( 2), Q( 3).
√
(c) Devise a similar algorithm to compute the fundamental √ unit in Q( d), for
d ≡ 1 (mod 4). Determine the fundamental unit of Q( 5).
√
Exercise 8.1.9 (a) For an imaginary quadratic field K = Q( −d) (d a positive,
squarefree integer), show that

Z/4Z for d = 1,

UK 3 Z/6Z for d = 3,


Z/2Z otherwise.

(b) Show that UK is finite ⇔ K = Q or K is an imaginary quadratic field.

(c) Show that, if there exists an embedding of K in R, then WK 3 {±1} 3 Z/2Z.
Conclude that, in particular, this is the case if [K : Q] is odd.

Theorem 8.1.10 (a) Let ζm = e2πi/m , K = Q(ζm ). If m is even, the only

roots of unity in K are the mth roots of unity, so that WK 1 Z/mZ.
If m is odd, the only ones are the 2mth roots of unity, so that WK 1
Z/2mZ.

(b) Suppose that [K : Q] = 4. Then WK is one of the six groups Z/2lZ, 1 ≤

l ≤ 6. If, furthermore, K has no real embedding, then UK 1 WK × Z.

(c) Let K = Q(ζp ), p an odd prime. For any unit ε ∈ UK , ε = ζpk u, for
some real unit u ∈ UK ∩ R, k ∈ Z/pZ.

(d) Let K be as in (c) and let L = Q(ζp + ζp−1 ). Then L = K ∩ R and

conclude that UK = :ζp ; × UL .

(e) For K = Q(ζ5 ),

UK = {±ζ5k εl : k ∈ Z/5Z, l ∈ Z},

√ √
where ε = (1 + 5)/2 is the fundamental unit of Q( 5).
(m+1)/2
Proof. (a) If m is odd, then ζ2m = −ζ2m m+1
= −ζm which implies that
Q(ζm ) = Q(ζ2m ). It will, therefore, suffice to establish the statement for m
even. Suppose that θ ∈ Q(ζm ) is a primitive kth root of unity, k ! m. Then
8.1. DIRICHLET’S UNIT THEOREM 107

Q(ζm ) contains a primitive rth root of unity, where r = lcm(k, m) > m.

Then Q(ζr ) ⊆ Q(ζm )
⇒ ϕ(r) = [Q(ζr ) : Q] ≤ [Q(ζm ) : Q] = ϕ(m)
(where ϕ denotes the Euler phi-function). But m is even and m properly
divides r implies that ϕ(m) properly divides ϕ(r), so that, in particular,
ϕ(m) < ϕ(r), a contradiction. Thus, the mth roots of unity are the only
roots of unity in Q(ζm ).
(b) WK is cyclic, generated by an rth root of unity, for some even r ≥ 2.
Q(ζr ) ⊆ K means that 4 = [K : Q] ≥ ϕ(r). A straightforward computation
shows that
ϕ(r) ≤ 4 ⇒ r ∈ {2, 4, 6, 8, 10, 12}.
If K has no real embedding, then r1 = 0, r2 = 2 ⇒ r = r1 + r2 − 1 = 1. By
Dirichlet’s theorem, UK 1 WK × Z.
(c) Let ε ∈ UK . Then 1 = |(ε/ε)(i) | = |ε(i) /ε(i) |, for i = 1, . . . , n = [K :
Q]. By the remark in the solution to Exercise 8.1.1, ε/ε is a root of unity
in K and so ε/ε = ±ζpk , for some k.
5p−2
Since OK = Z[ζp ], we may write ε = i=0 ai ζpi , each ai ∈ Z. Then
/p−2 0p p−2
* *
εp = ai ζpi ≡ ai (mod p).
i=0 i=0
5p−2
Since ε = i=0 ai ζ −i ,
p−2
*
εp ≡ ai ≡ εp (mod p).
i=0

If ε = −ζpk ε, then εp ≡ −εp (mod p). This implies that 2εp ≡ 0 (mod p)
and so εp ≡ 0 (mod p). In other words, εp ∈ (p), a contradiction, since ε
is a unit.
Thus ε/ε = ζpk , for some k. Let r ∈ Z such that 2r ≡ k (mod p) and let
ε1 = ζp−r ε. Then ε1 = ζpr ε = ζpr−k ε = ζp−r ε = ε1 ⇒ ε1 ∈ R and ε = ζpr ε1 .
(d) Let α = ζp + ζp−1 . Then α = 2 cos(2π/p) ∈ R so L = Q(α) ⊆ K ∩ R.
But since [K : K ∩ R] = 2 and ζp2 − αζp + 1 = 0 so that [K : Q(α)] ≤ 2, we
have that L = K ∩ R. Thus, ε ∈ UK , meaning ε = ±ζpk ε1 , for some ε1 ∈ L,
so that
UK = :ζp ; × UL .
√
(e) It remains only to show that Q(ζ5 +ζ5−1 ) = Q( 5). Let α = ζ5 +ζ5−1 .
ζ54 + ζ53 + ζ52 + ζ5 + 1 = 0,
⇒ ζ52 + ζ5 + 1 + ζ5−1 + ζ5−2 = 0,
or α2 +
√ α − 1 = 0. Since α = 2 cos(2π/5) >
√ 0, this implies that α =
(−1 + 5)/2 and we conclude that Q(α) = Q( 5). !
108 CHAPTER 8. THE STRUCTURE OF UNITS

Exercise 8.1.11 Let [K : Q] = 3 and suppose that K has only one real embed-
ding. Then, by Exercise 8.1.9 (c), WK = {±1} implies that UK = {±uk : k ∈ Z},
where u > 1 is the fundamental unit in K.
(a) Let u, ρeiθ , ρe−iθ be the Q-conjugates of u. Show that u = ρ−2 and that
dK/Q (u) = −4 sin2 θ(ρ3 + ρ−3 − 2 cos θ)2 .
(b) Show that |dK/Q (u)| < 4(u3 + u−3 + 6).
(c) Conclude that u3 > d/4 − 6 − u−3 > d/4 − 7, where d = |dK |.
√
3
Exercise 8.1.12 Let α = 2, K = Q(α). Given that dK = −108:
(a) Show that, if u is the fundamental unit in K, u3 > 20.
(b) Show that β = (α − 1)−1 = α2 + α + 1 is a unit, 1 < β < u2 . Conclude that
β = u.

Exercise 8.1.13 (a) Show that, if α ∈ K is a root of a monic polynomial f ∈

Z[x] and f (r) = ±1, for some r ∈ Z, then α − r is a unit in K.
√ 2
(b) Using the fact that if K = Q( 3 m), then dK = −27m √ , for any cubefree
integer m, determine the fundamental unit in K = Q( 3 7).
√
(c) Determine the fundamental unit in K = Q( 3 3).

8.2 Units in Real Quadratic Fields

In Exercise 8.1.8, we developed a simple algorithm with which we can deter-
mine the fundamental unit of a real quadratic field. However, this algorithm
is extremely inefficient and, moreover, there is no way of determining the
number of steps it will take to terminate. In this section, we develop a more
efficient and more enlightening algorithm, using continued fractions.

Definition.

(i) A finite continued fraction is an expression of the form

1
a0 + ,
1
a1 +
1
a2 + · · · +
1
an−1 +
an
where each ai ∈ R and ai ≥ 0 for 1 ≤ i ≤ n. We use the notation
[a0 , . . . , an ] to denote the above expression.

(ii) [a0 , . . . , an ] is called a simple continued fraction if a1 , . . . , an ∈ Z.

(iii) The continued fraction Ck = [a0 , . . . , ak ], 0 ≤ k ≤ n, is called the kth

convergent of [a0 , . . . , an ].
8.2. UNITS IN REAL QUADRATIC FIELDS 109

Evidently, a finite simple continued fraction represents a rational num-

ber. Conversely, using the Euclidean algorithm, one can show that every
rational number can be expressed as a finite simple continued fraction.

Exercise 8.2.1 (a) Consider the continued fraction [a0 , . . . , an ]. Define the se-
quences p0 , . . . , pn and q0 , . . . , qn recursively as follows:

p0 = a0 , q0 = 1,
p1 = a0 a1 + 1, q1 = a 1 ,
pk = ak pk−1 + pk−2 , qk = ak qk−1 + qk−2 ,

for k ≥ 2. Show that the kth convergent Ck = pk /qk .

(b) Show that pk qk−1 − pk−1 qk = (−1)k−1 , for k ≥ 1.
(c) Derive the identities
(−1)k−1
Ck − Ck−1 = ,
qk qk−1
for 1 ≤ k ≤ n, and
ak (−1)k
Ck − Ck−2 = ,
qk qk−2
for 2 ≤ k ≤ n.
(d) Show that
C1 > C3 > C5 > · · · ,
C0 < C2 < C4 < · · · ,
and that every odd-numbered convergent C2j+1 , j ≥ 0, is greater than every
even-numbered convergent C2k , k ≥ 0.

Remark. By (b), we can conclude that if [a0 , . . . , an ] is a simple continued

fraction, then the integers pk , qk are relatively prime.
It is also useful to note for k ≥ 1,
! "! " ! " ! "
a0 1 a1 1 a 1 pk pk−1
··· k = ,
1 0 1 0 1 0 qk qk−1

which is easily proved by induction. Thus, the convergents can be easily

retrieved by matrix multiplication.

Exercise 8.2.2 Let {ai }i≥0 be an infinite sequence of integers with ai ≥ 0 for
i ≥ 1 and let Ck = [a0 , . . . , ak ]. Show that the sequence {Ck } converges.

Definition. We define the continued fraction [a0 , a1 , . . . ] to be the limit as

k → ∞ of its kth convergent Ck .

[a0 , a1 , . . . ] = lim Ck .
k→∞
110 CHAPTER 8. THE STRUCTURE OF UNITS

Exercise 8.2.3 Let α = α0 be an irrational real number greater than 0. Define

the sequence {ai }i≥0 recursively as follows:

1
ak = [αk ], αk+1 = .
αk − ak
Show that α = [a0 , a1 , . . . ] is a representation of α as a simple continued fraction.

By Exercise 8.2.3, it is evident that every real number α has an expres-

sion as a simple continued fraction. We can also show that the representa-
tion of an irrational number as a simple continued fraction is unique. From
now on, we will call the representation of α as a simple continued fraction
simply the continued fraction of α.

Theorem 8.2.4 (a) Let α be an irrational number and let Cj = pj /qj ,

for j ∈ N, be the convergents of the simple continued fraction of α. If
r, s ∈ Z with s > 0 and k is a positive integer such that

|sα − r| < |qk α − pk |,

then s ≥ qk+1 .

(b) If α is an irrational number and r/s is a rational number in lowest

terms, s > 0, such that
1
|α − r/s| < ,
2s2
then r/s is a convergent of the continued fraction of α.

Proof. (a) Suppose, on the contrary, that 1 ≤ s < qk+1 . For each k ≥ 0,
consider the system of linear equations

pk x + pk+1 y = r,
qk x + qk+1 y = s.

Using Gaussian elimination, we easily find that

(pk+1 qk − pk qk+1 )y = rqk − spk ,

(pk qk+1 − pk+1 qk )x = rqk+1 − spk+1 .

By Exercise 8.2.1 (b), pk+1 qk − pk qk+1 = (−1)k so pk qk+1 − pk+1 qk =

(−1)k+1 . Thus, the unique solution to this system is given by

x = (−1)k (spk+1 − rqk+1 ),

y = (−1)k (rqk − spk ).

We will show that x and y are nonzero and have opposite signs.
8.2. UNITS IN REAL QUADRATIC FIELDS 111

If x = 0, then
r pk+1
=
s qk+1

and since (pk+1 , qk+1 ) = 1, this implies that qk+1 |s, and so qk+1 ≤ s,
contradicting our hypothesis.
If y = 0, then r = pk x, s = qk x, so that

|sα − r| = |x| · |qk α − pk | ≥ |qk α − pk |,

again contradicting our hypothesis.

Suppose now that y < 0. Then since qk x = s − qk+1 y and each qj ≥
0, x > 0. On the other hand, if y > 0, then qk+1 y ≥ qk+1 > s so qk x =
s − qk+1 y < 0 and x < 0.
By Exercise 8.2.1 (d), if k is even,

pk pk+1
<α< ,
qk qk+1

while, if k is odd,
pk+1 pk
<α< .
qk+1 qk

Thus, in either case, qk α − pk and qk+1 α − pk+1 have opposite signs so that
x(qk α − pk ) and y(qk+1 α − pk+1 ) have the same sign.

⇒ |sα − r| = |(qk x + qk+1 y)α − (pk x + pk+1 y)|

= |x(qk α − pk ) + y(qk+1 α − pk+1 )|
≥ |x| · |qk α − pk | + |y| · |qk+1 α − pk+1 |
≥ |x| · |qk α − pk | ≥ |qk α − pk |,

a contradiction, thus establishing our assertion.

(b) Suppose that r/s is not a convergent of the continued fraction of α,
i.e., r/s != pn /qn for all n. Let k be the largest nonnegative integer such
that s ≥ qk . (Since s ≥ q0 = 1 and qk → ∞ as k → ∞, we know that such
a k exists.) Then qk ≤ s ≤ qk+1 and by (a),

1
|qk α − pk | ≤ |sα − r| = s|α − r/s| < ,
1 1 2s
1
1 pk 11 1
⇒ 1α − qk 1 < 2sqk .
112 CHAPTER 8. THE STRUCTURE OF UNITS

Since r/s != pk /qk , |spk − rqk | ≥ 1,

1 1
1 |spk − rqk | 1 pk
1 r 11
≤ = 1 − 1
sqk sqk qk s
1 1
1 pk r 1
1
= 1 − + α − α11
qk s
1 1
1
1 pk 11 11 r 11
≤ 1α − 1 + 1α − 1
qk s
1 1
< + 2.
2sqk 2s
This would imply that
1 1
< 2,
2sqk 2s
so qk > s, a contradiction. !

Exercise 8.2.5√ Let d be a positive integer, not a perfect square. Show that, if
|x2 −dy 2 | <√ d for positive integers x, y, then x/y is a convergent of the continued
fraction of d.

Definition. A simple continued fraction is called periodic with period k if

there exist positive integers N, k such that an = an+k for all n ≥ N . We
denote such a continued fraction by [a0 , . . . , aN −1 , aN , aN +1 , . . . , aN +k−1 ].

Exercise 8.2.6 Let α be a quadratic irrational (i.e, the minimal polynomial of

the real number α over Q has degree 2). Show that there are integers P0 , Q0 , d
such that √
P0 + d
α= with Q0 |(d − P02 ).
Q0
Recursively define
√
Pk + d
αk = ,
Qk
ak = [αk ],
Pk+1 = ak Qk − Pk ,
2
d − Pk+1
Qk+1 = ,
Qk
for k = 0, 1, 2, . . . . Show that [a0 , a1 , a2 , . . . ] is the simple continued fraction of
α.

Exercise 8.2.7 Show that the simple continued fraction expansion of a quadratic
irrational α is periodic.

√ Show that, if d is a positive integer but not a perfect square,

Exercise 8.2.8
and α = α0 = d, then

p2k−1 − dqk−1
2
= (−1)k Qk ,
8.2. UNITS IN REAL QUADRATIC FIELDS 113

for all k ≥ 1, where pk /qk is the kth convergent of the continued fraction of α
and Qk is as defined in Exercise 8.2.6.
√
Let n be the period of the continued fraction of d. We can show,
using properties of purely periodic continued fractions, that n is the smallest
positive integer such that Qn = 1 (so that Qj != 0, for 0 < j < n) and that
Qn != −1 for all n. In particular, this implies that (−1)k Qk = ±1 if and
only if n|k. For the sake of brevity, we omit the proof of this fact.
√
Theorem 8.2.9 Let n be the period of the continued fraction of d.
(a) All integer solutions to the equation x2 − dy 2 = ±1 are given by
√ √
x + y d = ±(pn−1 + qn−1 d)l : l ∈ Z,

√ pn−1 /qn−1 is the (n − 1)th convergent of the continued fraction

where
of d.
√
(b) If d is squarefree, √
d ≡ 2, 3 (mod 4), then pn−1 + qn−1 d is the funda-
mental unit of Q( d).
(c) The equation x2 − dy 2 = −1 has an integer solution if and only if n is
odd.
(d) If d has a prime divisor p ≡ 3 (mod 4), then the equation x2 −dy 2 = −1
has no integer solution.
Proof. (a) For any solution (x, y) to the given equation,
√ √
(x + dy)−1 = ±(x − dy).

Therefore, one of ±(a, ±b) is a solution to x2 − dy 2 = ±1 if and only if each

of the four pairs is a solution. It will, thus, suffice to show that all positive
solutions are given by
√ √
x + y d = (pn−1 + qn−1 d)m : m > 0.

By Exercise 8.2.5, if x2 − dy 2 = ±1, then x = pk−1 , y = qk−1 , for some k.

2
By Exercise 8.2.8, p2k−1 − dqk−1 = (−1)k Qk = ±1 ⇒ Qk = ±1 and, by our
Remark, this implies that n|k. Since

pn−1 < p2n−1 < · · · and qn−1 < q2n−1 < · · · ,

we have that, in particular, the least positive solution to the given equation
is x1 = pn−1 , y1 = qn−1 . We√will now show √ that all positive solutions
(xm , ym ) are given by xm + ym d√ = (x1 + y1 d)√m , m > 0.
Taking Q-conjugates, xm − ym d = (x1 − y1 d)m
√ √
(xm + ym d)(xm − ym d) = (x21 − dy12 )m = (±1)m = ±1,
114 CHAPTER 8. THE STRUCTURE OF UNITS

so that (xm , ym ) is indeed a solution. Evidently, x1 < xm , y1 < ym , so that

(xm , ym ) is a positive solution.
Now, suppose that (X, Y ) is a positive solution that is not one of the
(xm , ym ). Then ∃ an integer κ ≥ 0 such that
√ √ √
(x1 + y1 d)κ < X + Y d < (x1 + y1 d)κ+1 ,

or √ √ √
1 < (x1 + y1 d)−κ (X + Y d) < x1 + y1 d.
√ √
But x21 − dy12 = ±1 which implies that (x1 + y1 d)−κ = [±(x1 − y1 d)]κ .
Define the integers s, t such that
√ √ √ √ √
s + t d = (x1 + y1 d)−κ (X + Y d) = ±(x1 − y1 d)κ (X + Y d).

Then
√ √ √ √
s2 − dt2 = [±(x1 − y1 d)κ (X + Y d)][±(x1 + y1 d)κ (X − Y d)]
= X 2 − dY 2 = ±1.

Thus, (s, t) is a solution to the given equation with

√ √
1 < s + t d < x1 + y1 d.

Also, √ √ √
0 < (x1 + y1 d)−1 < (s + t d)−1 < 1 < s + t d.
But this implies that
√ √ √ √
2s = s + t d ± [±(s − t d)] = s + t d ± (s + t d)−1 > 0,
√ √ √
2t d = s + t d ∓ [±(s − t d)] > 0,

√ is a positive√solution. By hypothesis, then, s ≥ x1 , t ≥ y1 and,

and so (s, t)
since s + t d < x1 + y1 √ d, we have a contradiction.
(b) Since pn−1 + qn−1 d > 1, this follows immediately from part (a).
(c) x2 − dy 2 = −1 ⇒ x = pk−1 , y = qk−1 for some k, by Exercise 8.2.5.
But p2k−1 − dqk−1
2
= (−1)k Qk = −1 if and only if n|k and k is odd. Clearly
then, a solution exists if and only if n is odd.
(d) x2 − dy 2 = −1 implies that x2 ≡ −1( mod p) for all p|d. But, for
p ≡ 3 mod 4, this congruence has no solutions. !
√ √
Exercise 8.2.10 (a) Find the simple continued fractions of 6, 23.
√
(b) Using
√ Theorem 8.2.9 (c), compute the fundamental unit in both Q( 6) and
Q( 23).
√
Exercise 8.2.11 (a) Show that [d, 2d] is the continued fraction of d2 + 1.
8.3. SUPPLEMENTARY PROBLEMS 115

(b) Conclude that,√if d2 + 1 is squarefree,

√ d ≡ 1, 3 (mod 4), then the fundamen-
tal√unit of√Q( d2 + √ 1) is d + d2 + 1. Compute the fundamental unit of

Q( 2), Q( 10), Q( 26).

√
(c) Show that the continued fraction of d2 + 2 is [d, d, 2d].
√
(d) Conclude that,
√ if d2 +2 is squarefree, then the fundamental unit √
of Q( d2√+ 2)
2
is d
√ + 1 +√ d d2 + 2. Compute the fundamental unit in Q( 3), Q( 11),
Q( 51), Q( 66).

8.3 Supplementary Problems

√
Exercise √8.3.1 If n2 − 1 is squarefree, show that n + n2 − 1 is the fundamental
unit of Q( n2 − 1).

Exercise 8.3.2 Determine the units of an imaginary quadratic field from first
principles.

Exercise
√ 8.3.3 Suppose that 22n √
+ 1 = dy 2 with d √
squarefree.
√ Show that 2 +
n

y d is the fundamental unit of Q( d), whenever Q( d) #= Q( 5).

√
Exercise 8.3.4 (a) Determine the continued√fraction expansion of 51 and use
it to obtain the fundamental unit ε of Q( 51).
√
(b) Prove from first principles that all units of Q( 51) are given by εn , n ∈ Z.

Exercise 8.3.5 Determine a unit #= ±1 in the ring of integers of Q(θ) where

θ3 + 6θ + 8 = 0.

Exercise 8.3.6 Let p be an odd prime > 3 and supose that it does not divide
the class number of Q(ζp ). Show that
xp + y p + z p = 0
is impossible for integers x, y, z such that p ! xyz.

Exercise 8.3.7 Let K be a quadratic field of discriminant d. Let P0 denote

the group of principal fractional ideals αOK with α ∈ K satisfying NK (α) > 0.
The quotient group H0 of all nonzero fractional ideals modulo P0 is called the
restricted class group of K. Show that H0 is a subgroup of the ideal class group
H of K and [H : H0 ] ≤ 2.

Exercise 8.3.8 Given an ideal a of a quadratic field K, let a( denote the conju-
gate ideal. If K has discriminant d, write
|d| = pα1
1 p2 · · · pt

where p1 = 2, α1 = 0, 2, or 3 and p2 , . . . , pt are distinct odd primes. If we write

pi OK = ℘2i show that for any ideal a of OK satisfying a = a( we can write
a = r℘a1 1 · · · ℘at t ,
r > 0, ai = 0, 1 uniquely.
116 CHAPTER 8. THE STRUCTURE OF UNITS

Exercise 8.3.9 An ideal class C of H0 is said to be ambiguous if C 2 = 1 in H0 .

Show that any ambiguous ideal class is equivalent (in the restricted sense) to one
of the at most 2t ideal classes
℘a1 1 · · · ℘at t , ai = 0, 1.

Exercise 8.3.10 With the notation as in the previous two questions, show that
there is exactly one relation of the form
℘a1 1 · · · ℘at t = ρOK , NK (ρ) > 0,
&t
with ai = 0 or 1, i=1 ai > 0.

Exercise 8.3.11 Let K be a quadratic field of discriminant d. Show that the

number of ambiguous ideal classes is 2t−1 where t is the number of distinct primes
dividing d. Deduce that 2t−1 divides the order of the class group.

Exercise 8.3.12 If K is a quadratic field of discriminant d and class number 1,

show that d is prime or d = 4 or 8.

Exercise 8.3.13 If a real quadratic field K has odd class number, show that K
has a unit of norm −1.
√ √
Exercise 8.3.14 Show that 15 + 4 14 is the fundamental unit of Q( 14).
√
Exercise 8.3.15 In Chapter 6 we showed that Z[ 14] is a PID√(principal ideal
domain). Assume the following hypothesis: given α, β ∈ Z[ 14], such that
gcd(α, β) √
= 1, there is a prime π ≡ α (mod β) for which the fundamental √unit
ε = 15 + 4 14 generates the coprime residue classes (mod π). Show that Z[ 14]
is Euclidean.
√
It is now known that Z[ 14] is Euclidean and is the main theorem of
the doctoral thesis of Harper [Ha]. The hypothesis of the previous exer-
cise is still unknown however and is true if the Riemann hypothesis holds
for Dedekind zeta functions of number fields (see Chapter 10). The hy-
pothesis in the question should be viewed as a number field version of a
classical conjecture of Artin on primitive roots. Previously the classifica-
tion of Euclidean rings of algebraic integers relied on some number field
generalization of the Artin primitive root conjecture. But recently, Harper
and Murty [HM] have found new techniques which circumvent the need of
such a hypothesis in such questions. No doubt, these techniques will have
further applications.

Exercise 8.3.16 Let d = a2 + 1. Show that if |u2 − dv 2 | =

# 0, 1 for integers u, v,
then √
|u2 − dv 2 | > d.

2g
Exercise 8.3.17 Suppose that n √is odd, n ≥ 5, and that n +1 = d is squarefree.
Show that the class group of Q( d) has an element of order 2g.
Chapter 9

Higher Reciprocity Laws

9.1 Cubic Reciprocity

√
Let ρ = (−1 + −3)/2 be as in Chapter 2, and let Z[ρ] be the ring of
Eisenstein integers. Recall that Z[ρ] is a Euclidean domain and hence a
PID. We set N (a + bρ) = a2 − ab + b2 which is the Euclidean norm as
proved in Section 2.3.

Exercise 9.1.1 If π is a prime of Z[ρ], show that N (π) is a rational prime or the
square of a rational prime.

Exercise 9.1.2 If π ∈ Z[ρ] is such that N (π) = p, a rational prime, show that π
is a prime of Z[ρ].

Exercise 9.1.3 If p is a rational prime congruent to 2 (mod 3), show that p is

prime in Z[ρ]. If p ≡ 1 (mod 3), show that p = ππ where π is prime in Z[ρ].

Exercise 9.1.4 Let π be a prime of Z[ρ]. Show that αN (π)−1 ≡ 1 (mod π) for
all α ∈ Z[ρ] which are coprime to π.

Exercise 9.1.5 Let π be a prime not associated to (1 − ρ). First show that
3 | N (π) − 1. If (α, π) = 1, show that there is a unique integer m = 0, 1, or 2
such that
α(N (π)−1)/3 ≡ ρm (mod π).

Let N (π) != 3. We define the cubic residue character of α (mod π) by

the symbol (α/π)3 as follows:

(i) (α/π)3 = 0 if π | α;

(ii) α(N (π)−1)/3 ≡ (α/π)3 (mod π) where (α/π)3 is the unique cube root
of unity determined by the previous exercise.

117
118 CHAPTER 9. HIGHER RECIPROCITY LAWS

Exercise 9.1.6 Show that:

(a) (α/π)3 = 1 if and only if x3 ≡ α (mod π) is solvable in Z[ρ];
(b) (αβ/π)3 = (α/π)3 (β/π)3 ; and
(c) if α ≡ β (mod π), then (α/π)3 = (β/π)3 .

Let us now define the cubic character χπ (α) = (α/π)3 .

Exercise 9.1.7 Show that:

(a) χπ (α) = χπ (α)2 = χπ (α2 ); and
(b) χπ (α) = χπ (α).

Exercise 9.1.8 If q ≡ 2 (mod 3), show that χq (α) = χq (α2 ) and χq (n) = 1 if n
is a rational integer coprime to q.

This exercise shows that any rational integer is a cubic residue mod q.
If π is prime in Z[ρ], we say π is primary if π ≡ 2 (mod 3). Therefore if
q ≡ 2 (mod 3), then q is primary in Z[ρ]. If π = a + bρ, then this means
a ≡ 2 (mod 3) and b ≡ 0 (mod 3).

Exercise 9.1.9 Let N (π) = p ≡ 1 (mod 3). Among the associates of π, show
there is a unique one which is primary.

We can now state the law of cubic reciprocity: let π1 , π2 be primary.

Suppose N (π1 ), N (π2 ) != 3 and N (π1 ) != N (π2 ). Then

χπ1 (π2 ) = χπ2 (π1 ).

To prove the law of cubic reciprocity, we will introduce Jacobi sums

and more general Gauss sums than the ones used in Chapter 7. Let Fp
denote the finite field of p elements. A multiplicative character on Fp is a
homomorphism χ : F× p → C . The Legendre symbol (a/p) is an example
×

of such a character. Another example is the trivial character χ0 defined by

χ0 (a) = 1 for all a ∈ F×p . It is useful to extend the definition of χ to all of
Fp . We set χ(0) = 0 for χ != χ0 and χ0 (0) = 1.
For a ∈ F×p , define the Gauss sum

*
ga (χ) = χ(t)ζ at ,
t∈Fp

where ζ = e2πi/p is a primitive pth root of unity. We also write g(χ) for
g1 (χ).

√
Theorem 9.1.10 If χ != χ0 , then |g(χ)| = p.
9.1. CUBIC RECIPROCITY 119

Proof. We first observe that a != 0 and χ != χ0 together imply that

ga (χ) = χ(a−1 )g(χ) because
*
χ(a)ga (χ) = χ(a) χ(t)ζ at
t∈Fp
*
= χ(at)ζ at
t∈Fp

= g(χ).

With our conventions that χ0 (0) = 1, we see for a != 0,

*
ga (χ0 ) = ζ at = 0,
t∈Fp

since this is just the sum of the pth roots of unity. Finally, g0 (χ) = 0 if
χ != χ0 and g0 (χ0 ) = p.
Now, by our first observation,
*
ga (χ)ga (χ) = |g(χ)|2 (p − 1).
a∈Fp

On the other hand,

* * * *
ga (χ)ga (χ) = χ(s)χ(t) ζ as−at .
a∈Fp s∈Fp t∈Fp a∈Fp

If s != t, the innermost sum is zero, being the sum of all the pth roots of
unity. If s = t, the sum is p. Hence |g(χ)|2 = p. !

Let χ1 , χ2 , . . . , χr be characters of Fp . A Jacobi sum is defined by

*
J(χ1 , . . . , χr ) = χ1 (t1 ) · · · χr (tr ),
t1 +···+tr =1

where the summation is over all solutions of t1 + · · · + tr = 1 in Fp . The

relationship between Gauss sums and Jacobi sums is given by the following
exercise.

Exercise 9.1.11 If χ1 , . . . , χr are nontrivial and the product χ1 · · · χr is also

nontrivial, prove that g(χ1 ) · · · g(χr ) = J(χ1 , . . . , χr )g(χ1 · · · χr ).

Exercise 9.1.12 If χ1 , . . . , χr are nontrivial, and χ1 · · · χr is trivial, show that

g(χ1 ) · · · g(χr ) = χr (−1)pJ(χ1 , . . . , χr−1 ).

We are now ready to prove the cubic reciprocity law. It will be conve-
nient to work in the ring Ω of all algebraic integers.
120 CHAPTER 9. HIGHER RECIPROCITY LAWS

Lemma 9.1.13 Let π be a prime of Z[ρ] such that N (π) = p ≡ 1 (mod 3).
The character χπ introduced above can be viewed as a character of the finite
field Z[ρ]/(π) of p elements. J(χπ , χπ ) = π.

Proof. If χ is any cubic character, Exercise 9.1.12 shows that g(χ)3 =

pJ(χ, χ) since χ(−1) = 1. We can write J(χ, χ) = a + bρ for some a, b ∈ Z.
But
/ 03
*
3
g(χ) = χ(t)ζ t

t
*
≡ χ3 (t)ζ 3t (mod 3Ω)
t
*
≡ ζ 3t (mod 3Ω)
t*=0
≡ −1 (mod 3Ω).

Therefore, a + bρ ≡ −1 (mod 3Ω). In a similar way,

g(χ)3 ≡ a + bρ ≡ −1 (mod 3Ω).

√
Thus, b −3 ≡ 0 (mod 3Ω) which means −3b2 /9 is an algebraic integer
and by Exercise 3.1.2, it is an ordinary integer. Thus, b ≡ 0 (mod 3) and
a ≡ −1 (mod 3). Also, from Exercise 9.1.12 and Theorem 9.1.10, it is clear
that |J(χ, χ)|2 = p = J(χ, χ)J(χ, χ). Therefore, J(χ, χ) is a primary prime
of norm p. Set J(χπ , χπ ) = π $ . Since ππ = p = π $ π $ , we have π | π $ or
π | π $ . We want to eliminate the latter possibility.
By definition,
*
J(χπ , χπ ) = χπ (t)χπ (1 − t)
t
*
≡ t(p−1)/3 (1 − t)(p−1)/3 (mod π).
t

The polynomial x(p−1)/3 (1 − x)(p−1)/3 has degree 32 (p − 1) < p − 1. Let g

be a primitive root (mod π). Then

* p−2
*
tj = g aj ≡ 0 (mod π)
t a=0

if g j !≡ 1 (mod π), which is the case since j < p − 1. Thus, J(χπ , χπ ) ≡ 0

(mod π). Therefore π | π $ , as desired. !

Exercise 9.1.14 Show that g(χ)3 = pπ.

9.1. CUBIC RECIPROCITY 121

Lemma 9.1.15 Let π1 = q ≡ 2 (mod 3) and π2 = π be a prime of Z[ρ] of

norm p. Then χπ (q) = χq (π). In other words,
EqF ! "
π
= .
π 3 q 3

Proof. Let χπ = χ, and consider the Jacobi sum J(χ, . . . , χ) with q terms.
Since 3 | q + 1, we have, by Exercise 9.1.12, g(χ)q+1 = pJ(χ, . . . , χ). By
Exercise 9.1.14, g(χ)3 = pπ so that

g(χ)q+1 = (pπ)(q+1)/3 .

Recall that *
J(χ, . . . , χ) = χ(t1 ) · · · χ(tq ),

where the sum is over all t1 , . . . , tq ∈ Z/pZ such that t1 + · · · + tq = 1. The

term in which t1 = · · · = tq satisfies qt1 = 1 and so χ(q)χ(t1 ) = 1. Raising
both sides to the qth power and noting that q ≡ 2 (mod 3) gives

χ(q)2 χ(t1 )q = 1

and so χ(t1 )q = χ(q). Therefore, the “diagonal” term which corresponds

to t1 = · · · = tq has the value χ(q). If not all the ti are equal, then we
get q different q-tuples from a cyclic permutation of (t1 , . . . , tq ). Thus
J(χ, . . . , χ) ≡ χ(q) (mod q).
Hence (pπ)(q+1)/3 ≡ pχ(q) (mod q) so that

p(q−2)/3 π (q+1)/3 ≡ χ(q) (mod q).

We raise both sides of this congruence to the (q − 1)st power (recalling that
q − 1 ≡ 1 (mod 3)):
2
p(q−2)(q−1)/3 π (q −1)/3
≡ χ(q)q−1 ≡ χ(q) (mod q).

By Fermat’s little Theorem, pq−1 ≡ 1 (mod q). Therefore,

2
χπ (q) ≡ π (q −1)/3
≡ χq (π) (mod q)

so that χπ (q) = χq (π) as desired. !

Theorem 9.1.16 Let π1 and π2 be two primary primes of Z[ρ], of norms

p1 , p2 , respectively. Then χπ1 (π2 ) = χπ2 (π1 ). In other words,
! " ! "
π2 π1
= .
π1 3 π2 3
122 CHAPTER 9. HIGHER RECIPROCITY LAWS

Proof. To begin, let γ1 = π1 , γ2 = π2 . Then p1 = π1 γ1 , p2 = π2 γ2 , and

p1 , p2 ≡ 1 (mod 3). Now,
g(χγ1 )p2 = J(χγ1 , . . . , χγ1 )g(χpγ21 )
by Exercise 9.1.11. (There are p2 characters in the Jacobi sum.) Since
p2 ≡ 1 (mod 3), χpγ21 = χγ1 . Thus,
< =(p2 −1)/3
g(χγ1 )3 = J(χγ1 , . . . , χγ1 ).
As before, isolating the “diagonal” term in the Jacobi sum and observing
that the contribution from the other terms is congruent to 0 (mod p2 ), we
find
2
J(χγ1 , . . . , χγ1 ) ≡ χγ1 (p−1
2 ) ≡ χγ1 (p2 ) (mod p2 ).
By Exercise 9.1.14, g(χγ1 )3 = p1 γ1 so that
(p1 γ1 )(p2 −1)/3 ≡ χγ1 (p22 ) (mod p2 ).
Hence χπ2 (p1 γ1 ) = χγ1 (p22 ). Similarly, χπ1 (p2 π2 ) = χπ2 (p21 ). Now, by Ex-
ercise 9.1.7, χγ1 (p22 ) = χπ1 (p2 ). Thus
χπ2 (p1 γ1 ) = χπ1 (p2 ),
χπ1 (p2 γ2 ) = χπ2 (p1 ).
Therefore
χπ1 (π2 )χπ2 (p1 γ1 ) = χπ1 (π2 p2 )
= χπ2 (p21 )
= χπ2 (p1 π1 γ1 )
= χπ2 (π1 )χπ2 (p1 γ1 ),
which gives χπ1 (π2 ) = χπ2 (π1 ), as desired. !

Exercise 9.1.17 Let π be a prime of Z[ρ]. Show that x3 ≡ 2 (mod π) has a

solution if and only if π ≡ 1 (mod 2).

9.2 Eisenstein Reciprocity

The Eisenstein reciprocity law generalizes both the laws of quadratic and
cubic reciprocity. In 1850, Eisenstein published the proof of this general-
ization by using the (then) new language of ideal numbers due to Kummer.
We do not prove this law here but content ourselves with understanding
its formulation and applying it to Fermat’s Last Theorem, in a particular
instance. We begin with the definition of the power residue symbol.
Let m be a positive integer. Then, it is known that Z[ζm ] is the ring
of integers of Q(ζm ). In the case m is prime, this was proved in Chapter
4 (Exercise 4.3.7). The general case is deduced from Exercises 4.5.9 and
4.5.13. Let ℘ be a prime ideal of Z[ζm ] not containing m. Let q be its norm.
9.2. EISENSTEIN RECIPROCITY 123

2
Exercise 9.2.1 Show that q ≡ 1 (mod m) and that 1, ζm , ζm m−1
, . . . , ζm are dis-
tinct coset representatives mod ℘.

Exercise 9.2.2 Let α ∈ Z[ζm ], α #∈ ℘. Show that there is a unique integer i

(modulo m) such that
α(q−1)/m ≡ ζm i
(mod ℘).

We can now define the power residue symbol. For α ∈ Z[ζm ], and ℘ a
prime ideal not containing m, define (α/℘)m as:
(i) (α/℘)m = 0 if α ∈ ℘; and
(ii) if α !∈ ℘, (α/℘)m is the unique mth root of unity such that
! "
α
α(N (℘)−1)/m ≡ (mod m)
℘ m
as determined by Exercise 9.2.2.

Exercise 9.2.3 Show that:

(a) (α/℘)m = 1 if and only if xm ≡ α (mod ℘) is solvable in Z[ζm ].
(b) for all α ∈ Z[ζm ], α(N (℘)−1)/m ≡ (α/℘)m (mod ℘).
(c) (αβ/℘)m = (α/℘)m (β/℘)m .
(d) if α ≡ β (mod ℘), then (α/℘)m = (β/℘)m .

Exercise 9.2.4 If ℘ is a prime ideal of Z[ζm ] not containing m, show that

$ %
ζm (N (℘)−1)/m
= ζm .
℘ m

We will now extend the definition of (α/℘)m . Let a = ℘1 ℘2 · · · ℘r be the

prime ideal decomposition of a. Suppose a is coprime to m. For α ∈ Z[ζm ],
define
EαF 6r ! "
α
= .
a m i=1 ℘i m
If β ∈ Z[ζm ] is coprime to m, define
! " ! "
α α
= .
β m (β) m

Exercise 9.2.5 Suppose a and b are ideals coprime to (m). Show that:
(a) (αβ/a)m = (α/a)m (β/a)m ;
(b) (α/ab)m = (α/a)m (β/b)m ; and
(c) if α is prime to a and xm ≡ α (mod a) is solvable in Z[ζm ], then (α/a)m = 1.

Exercise 9.2.6 Show that the converse of (c) in the previous exercise is not
necessarily true.
124 CHAPTER 9. HIGHER RECIPROCITY LAWS

Now let 7 be an odd prime number. Recall that by Exercise 5.5.4, we

have (7) = (1 − ζ) ))−1 in Z[ζ) ] and (1 − ζ) ) is a prime ideal of degree 1.
We will say that α ∈ Z[ζ) ] is primary if it is prime to 7 and congruent to
a rational integer modulo (1 − ζ) )2 . In the case 7 = 3, we required α ≡ 2
(mod 3) which is the same as α ≡ 2 (mod (1 − ζ3 )2 ). So this new definition
is weaker but will suffice for our purpose.

Exercise 9.2.7 If α ∈ Z[ζ* ] is coprime to 1, show that there is an integer c ∈ Z

(unique mod 1) such that ζ*c α is primary.

We can now state the Eisenstein reciprocity law: let 7 be an odd prime,
a ∈ Z prime to 7 and let α ∈ Z[ζ) ] be primary. If α and a are coprime, then
EαF EaF
= .
a ) α )
We will now apply this to establish the theorems of Wieferich and Furt-
wangler on Fermat’s Last Theorem: let 7 be an odd prime and suppose
x) + y ) + z ) = 0 for three mutually coprime integers x, y, z with 7 ! xyz.
(This is the so-called first case.) We let ζ = ζ) be a primitive 7th root of
unity and factor the above equation as

(x + y)(x + ζy) · · · (x + ζ )−1 y) = (−z)) .

Exercise 9.2.8 With notation as above, show that (x + ζ i y) and (x + ζ j y) are

coprime in Z[ζ* ] whenever i #= j, 0 ≤ i, j < 1.

Exercise 9.2.9 Show that the ideals (x + ζ i y) are perfect 1th powers.

Exercise 9.2.10 Consider the element

α = (x + y)*−2 (x + ζy).

Show that:
(a) the ideal (α) is a perfect 1th power.
(b) α ≡ 1 − uλ (mod λ2 ) where u = (x + y)*−2 y.

Exercise 9.2.11 Show that ζ −u α is primary.

Exercise 9.2.12 Use Eisenstein reciprocity to show that if x* + y * + z * = 0 has

a solution in integers, 1 ! xyz, then for any p | y, (ζ/p)−u
* = 1. (Hint: Evaluate
(p/ζ −u α)* .)

Exercise 9.2.13 Show that if

x* + y * + z * = 0

has a solution in integers, l ! xyz, then for any p | xyz, (ζ/p)−u

* = 1.
9.3. SUPPLEMENTARY PROBLEMS 125

Exercise 9.2.14 Show that (ζ/p)−u

* = 1 implies that p*−1 ≡ 1 (mod 12 ).

Exercise 9.2.15 If 1 is an odd prime and

x* + y * + z * = 0

for some integers x, y, z coprime to 1, then show that p*−1 ≡ 1 (mod 12 ) for every
p | xyz. Deduce that 2*−1 ≡ 1 (mod 12 ).

The congruence 2)−1 ≡ 1 (mod 72 ) was first established by Wieferich in

1909 as a necessary condition in the first case of Fermat’s Last Theorem.
The only primes less than 3 × 109 satisfying this congruence are 1093 and
3511 as a quick computer calculation shows. It is not known if there are
infinitely many such primes. (See also Exercise 1.3.4.)

9.3 Supplementary Problems

Exercise 9.3.1 Show that there are infinitely many primes p such that (2/p) =
−1.

Exercise 9.3.2 Let a be a nonsquare integer greater than 1. Show that there
are infinitely many primes p such that (a/p) = −1.

Exercise 9.3.3 Suppose that x2 ≡ a (mod p) has a solution for all but finitely
many primes. Show that a is a perfect square.

Exercise 9.3.4 Let K be a quadratic extension of Q. Show that there are in-
finitely many primes which do not split completely in K.

Exercise 9.3.5 Suppose that a is an integer coprime to the odd prime q. If

xq ≡ a (mod p) has a solution for all but finitely many primes, show that a is a
perfect qth power. (This generalizes the penultimate exercise.)

Exercise 9.3.6 Let p ≡ 1 (mod 3). Show that there are integers A and B such
that
4p = A2 + 27B 2 .
A and B are unique up to sign.

Exercise 9.3.7 Let p ≡ 1 (mod 3). Show that x3 ≡ 2 (mod p) has a solution if
and only if p = C 2 + 27D2 for some integers C, D.

Exercise 9.3.8 Show that the equation

x3 − 2y 3 = 23z m

has no integer solutions with gcd(x, y, z) = 1.

Chapter 10

Analytic Methods

10.1 The Riemann and Dedekind Zeta Func-

tions
The Riemann zeta function ζ(s) is defined initially for Re(s) > 1 as the
infinite series
∞
* 1
ζ(s) = .
n=1
ns

Exercise 10.1.1 Show that for Re(s) > 1,

#$ 1
%−1
ζ(s) = 1− s ,
p
p

where the product is over prime numbers p.

Exercise 10.1.2 Let K be an algebraic number field and OK its ring of integers.
The Dedekind zeta function ζK (s) is defined for Re(s) > 1 as the infinite series
, 1
ζK (s) = ,
a
(N a)s

where the sum is over all ideals of OK . Show that the infinite series is absolutely
convergent for Re(s) > 1.

Exercise 10.1.3 Prove that for Re(s) > 1,

#$ 1
%−1
ζK (s) = 1− .
℘
(N ℘)s

127
128 CHAPTER 10. ANALYTIC METHODS

5 10.1.4 Let {aδm }m=1 be a sequence of complex numbers, and let

Theorem ∞

A(x) = m≤x am = O(x ), for some δ ≥ 0. Then

*∞
am
m=1
ms

converges for Re(s) > δ and in this half-plane we have

*∞ @ ∞
am A(x) dx
s
=s
m=1
m 1 xs+1

for Re(s) > 1.

Proof. We write
*M M
*
am # $
= A(m) − A(m − 1) m−s
m=1
ms m=1
M
* −1
= A(M )M −s + A(m){m−s − (m + 1)−s }.
m=1

Since @ m+1
dx
m−s − (m + 1)−s = s ,
m xs+1
we get
*M @ M
am A(M ) A(x) dx
s
= s
+ s .
m=1
m M 1 xs+1
For Re(s) > δ, we find
A(M )
lim = 0,
M →∞ M s

since A(x) = O(xδ ). Hence, the partial sums converge for Re(s) > δ and
we have @ ∞
*∞
am A(x) dx
s
= s
m=1
m 1 xs+1
in this half-plane. !

Exercise 10.1.5 Show that (s − 1)ζ(s) can be extended analytically for Re(s) >
0.

Exercise 10.1.6 Evaluate

lim (s − 1)ζ(s).
s→1

Example 10.1.7 Let K = Q(i). Show that (s − 1)ζK (s) extends to an

analytic function for Re(s) > 12 .
10.1. THE RIEMANN AND DEDEKIND ZETA FUNCTIONS 129

Solution. Since every ideal a of OK is principal, we can write a = (a + ib)

for some integers a, b. Moreover, since

a = (a + ib) = (−a − ib) = (−a + ib) = (a − ib)

we can choose a, b to be both positive. In this way, we can associate with

each ideal a a unique lattice point (a, b), a ≥ 0, b ≥ 0. Conversely, to each
such lattice point (a, b) we can associate the ideal a = (a + ib). Moreover,
N a = a2 + b2 . Thus, if we write
* 1 *∞
an
ζK (s) = s
=
a
N a n=1
ns

we find that *
A(x) = an
n≤x

is equal to the number of lattice points lying in the positive quadrant defined
by the circle a2 + b2 ≤ x. We will call such a lattice point (a, b) internal
if (a + 1)2 + (b + 1)2 ≤ x. Otherwise, we will call it a boundary lattice
point. Let I be the number of internal lattice points, and B the number of
boundary lattice points. Then
π
I≤ x ≤ I + B.
4
Any boundary point (a, b) is contained in the annulus
√ √ √ √
( x − 2)2 ≤ a2 + b2 ≤ ( x + 2)2

and an upper bound for B is provided by the area of the annulus. This is
easily seen to be
√ √ √ √ √
π( x + 2)2 − π( x − 2)2 = O( x).
√
Thus A(x) = πx/4 + O( x). By Theorem 10.1.4, we deduce that
@ ∞ @ ∞
π dx E(x)
ζK (s) = s + s dx,
4 1 x s
1 xs+1
√
where E(x) = O( x), so that the latter integral converges for Re(s) > 12 .
Thus @ ∞
π E(x)
(s − 1)ζK (s) = s + s(s − 1) dx
4 1 xs+1
is analytic for Re(s) > 12 .

Exercise 10.1.8 For K = Q(i), evaluate

lim (s − 1)ζK (s).

s→1+
130 CHAPTER 10. ANALYTIC METHODS

Exercise 10.1.9 Show that the number of integers (a, b) with a > 0 satisfying
a2 + Db2 ≤ x is
πx √
√ + O( x).
2 D
√
Exercise 10.1.10 Suppose K = Q( −D) where D > 0 and −D #≡ 1 (mod 4)
and OK has class number 1. Show that (s − 1)ζK (s) extends analytically to
Re(s) > 12 and find
lim (s − 1)ζK (s).
s→1

(Note that there are only finitely many such fields.)

10.2 Zeta Functions of Quadratic Fields

In this section, we will derive the analytic continuation of zeta functions of
quadratic fields to the region Re(s) > 21 .
√
Exercise 10.2.1 Let K = Q( d) with d squarefree, and denote by an the num-
ber of ideals in OK of norm n. Show that an is multiplicative. (That is, prove
that if (n, m) = 1, then anm = an am .)

- .
Exercise 10.2.2 Show that for an odd prime p, ap = 1 + d
p
.

√
Exercise 10.2.3 Let
- d.K be the discriminant of K = Q( d). Show that for all

primes p, ap = 1 + dpK .

Exercise 10.2.4 Show that for all primes p,

α $
, % , $ dK %
dK
a pα = = .
pj δ
j=1 δ|pα

Exercise 10.2.5 Prove that

, $ dK %
an = .
δ
δ|n

Exercise 10.2.6 Let dK be-the .discriminant of the quadratic field K. Show that
there is an n > 0 such that dnK = −1.

Exercise 10.2.7 Show that

) )
) $ %)
) , dK )
) ) ≤ |dK |.
) )
)n≤x n )
10.2. ZETA FUNCTIONS OF QUADRATIC FIELDS 131

Theorem 10.2.8 (Dirichlet’s Hyperbola Method) Let

* EnF
f (n) = g(δ)h
δ
δ|n

and define
*
G(x) = g(n),
n≤x
*
H(x) = h(n).
n≤x

Then for any y > 0,

* * ExF * ExF ! "
x
f (n) = g(δ)H + h(δ)G − G(y)H .
δ δ y
n≤x δ≤y δ< x
y

Proof. We have
* *
f (n) = g(δ)h(e)
n≤x δe≤x
* *
= g(δ)h(e) + g(δ)h(e)
δe≤x δe≤x
δ≤y δ>y
* ExF H ExF * I
= g(δ)H h(e) G + − G(y)
δ e
δ≤y e≤ x
y

* ExF * ExF ! "

x
= g(δ)H + h(e)G − G(y)H
δ x e y
δ≤y e≤ y

as desired. !

Example 10.2.9 Let K be a quadratic field, and an the number of ideals

of norm n in OK . Show that
* √
an = cx + O( x),
n≤x

where
∞ !
* "
dK 1
c= .
δ δ
δ=1

Solution. By Exercise 10.2.5,

* ! dK "
an =
δ
δ|n
132 CHAPTER 10. ANALYTIC METHODS
#d $
so that
√ we can apply Theorem 10.2.8 with g(δ) = δ and h(δ) = 1,
K

y = x. We get
* * ! dK " J x K * ExF √ √
an = + G − G( x)[ x].
√ δ δ √ δ
n≤x δ≤ x δ< x

By Exercise 10.2.7, |G(x)| ≤ |dK |. Hence

* * ! dK " J x K √
an = + O( x).
√ δ δ
n≤x δ≤ x

Now [x/δ] = x/δ + O(1) so that

* * ! dK " x √
an = + O( x).
√ δ δ
n≤x δ≤ x

Finally,
* ! dK " 1 * ∞ ! "
dK 1 * ! dK " 1
= −
√ δ δ δ δ √ δ δ
δ≤ x δ=1 δ> x
and by Theorem 10.1.4 we see that
∞ !
* "
dK 1
c=
δ δ
δ=1

converges and
* ! dK " 1 !
1
"
=O √ .
√ δ δ x
δ> x
Therefore * √
an = cx + O( x).
n≤x

Exercise 10.2.10 If K is a quadratic field, show that (s − 1)ζK (s) extends to

an analytic function for Re(s) > 12 .

Dedekind conjectured in 1877 that (s − 1)ζK (s) extends to an entire

function for all s ∈ C and this was proved by Hecke in 1917 for all algebraic
number fields K. Moreover, Hecke established a functional equation for
ζK (s) and proved that
2r1 (2π)r2 hK RK
lim (s − 1)ζK (s) = L ,
s→1 w |dK |
where r1 is the number of real embeddings of K, r2 is the number of complex
embeddings, hK is the class number of K, dK is the discriminant of K, w
is the number of roots of unity in K, and RK is the regulator defined
as the determinant of the r × r matrix (log |σi (εj )|), and ε1 , . . . , εr are a
system of fundamental units, σ1 , . . . , σr , σr+1 , σr1 +1 , . . . , σr1 +r2 are the n
embeddings of K into C.
10.3. DIRICHLET’S L-FUNCTIONS 133

10.3 Dirichlet’s L-Functions

Let m be a natural number and χ a Dirichlet character mod m. That is, χ
is a homomorphism
χ : (Z/mZ)∗ → C∗ .
We extend the definition of χ to all natural numbers by setting
;
χ (a mod m) if (a, m) = 1,
χ(a) =
0 otherwise.
Now define the Dirichlet L-function:
*∞
χ(n)
L(s, χ) = .
n=1
ns

Exercise 10.3.1 Show that L(s, χ) converges absolutely for Re(s) > 1.

Exercise 10.3.2 Prove that

) )
) )
), )
) χ(n))) ≤ m.
)
)n≤x )

Exercise 10.3.3 If χ is nontrivial, show that L(s, χ) extends to an analytic

function for Re(s) > 0.

Exercise 10.3.4 For Re(s) > 1, show that

#$ χ(p)
%−1
L(s, χ) = 1− s .
p
p

Exercise 10.3.5 Show that

=
, ϕ(m) if a ≡ b (mod m),
χ(a)χ(b) =
χ mod m
0 otherwise.

Exercise 10.3.6 For Re(s) > 1, show that

, , 1
log L(s, χ) = ϕ(m) .
npns
χ mod m pn ≡1 mod m

Exercise 10.3.7 For Re(s) > 1, show that

, , 1
χ(a) log L(s, χ) = ϕ(m) .
npns
χ mod m pn ≡a mod m

Exercise 10.3.8 Let K = Q(ζm ). Set

#
f (s) = L(s, χ).
χ

Show that ζK (s)/f (s) is analytic for Re(s) > 12 .

134 CHAPTER 10. ANALYTIC METHODS

10.4 Primes in Arithmetic Progressions

In this section we will establish the infinitude of primes p ≡ a (mod m) for
any a coprime to m.

Lemma 10.4.1 Let {an } be a sequence of nonnegative numbers. There

exists a σ0 ∈ R (possibly infinite) such that

*∞
an
f (s) =
n=1
ns

converges for σ > σ0 and diverges for σ < σ0 . Moreover, if s ∈ C, with

Re(s) > σ0 , then the series converges uniformly in Re(s) ≥ σ0 + δ for any
δ > 0 and
*∞
an (log n)k
f (k) (s) = (−1)k
n=1
ns

for Re(s)
5∞> σ0 . (σ0 is called the abscissa of convergence of the (Dirichlet)
series n=1 an n−s .)

Proof. If there is no real value of s for which the series converges, we

take σ0 = ∞. Therefore, suppose there is some real s0 for which the
series converges. Clearly by the comparison test, the series converges for
Re(s) > s0 since the coefficients are nonnegative. Now let σ0 be the infimum
of all real s0 for which the series converges. The uniform convergence in
Re(s) ≥ σ0 + δ for any δ > 0 is now immediate. Because of this, we can
differentiate term by term to calculate f (k) (s) for Re(s) > σ0 . !

Theorem 10.4.2 Let an ≥ 0 be a sequence of nonnegative numbers. Then

*∞
an
f (s) =
n=1
ns

defines a holomorphic function in Re(s) > σ0 and s = σ0 is a singular point

of f (s). (Here σ0 is the abscissa of convergence of the Dirichlet series.)

Proof. By the previous lemma, it is clear that f (s) is holomorphic in

Re(s) > σ0 . If f is not singular at s = σ0 , then there is a disk

D = {s : |s − σ1 | < δ},

where σ1 > σ0 such that |σ0 − σ1 | < δ and a holomorphic function g in D

such that g(s) = f (s) for Re(s) > σ0 , s ∈ D. By Taylor’s formula
∞
* ∞
*
g (k) (σ1 ) f (k) (σ1 )
g(s) = (s − σ1 )k = (s − σ1 )k
k! k!
k=0 k=0
10.4. PRIMES IN ARITHMETIC PROGRESSIONS 135

since g(s) = f (s) for s in a neighborhood of σ1 . Thus, the series

∞
* (−1)k f (k) (σ1 )
(σ1 − s)k
k!
k=0

converges absolutely for any s ∈ D. By the lemma, we can write this series
as the double series
∞
* ∞
(σ1 − s)k * an (log n)k
.
k! n=1
n σ1
k=0

If σ1 − δ < s < σ1 , this convergent double series consists of nonnegative

terms and we may interchange the summations to find
*∞ ∞ ∞
an * (σ1 − s)k (log n)k * an
σ1
= < ∞.
n=1
n k! n=1
ns
k=0

Since σ1 − δ < σ0 < σ1 , this is a contradiction for s = σ0 . Therefore, the

abscissa of convergence is a singular point of f (s). !

Exercise 10.4.3 With the notation as in Section 10.3, write

# ,∞
cn
f (s) = L(s, χ) = s
.
χ n=1
n

Show that cn ≥ 0.

Exercise 10.4.4 With notation as in the previous exercise, show that

,∞
cn
ns
n=1

diverges for s = 1/ϕ(m).

Theorem 10.4.5 Let L(s, χ) be defined as above. Then L(1, χ) != 0 for

χ != χ0 .
Proof. By the previous exercise, the abscissa of convergence of
6 *∞
cn
f (s) = L(s, χ) =
χ n=1
ns

is greater than or equal to 1/ϕ(m). If for some χ != χ0 we have L(1, χ) = 0,

then f (s) is holomorphic at s = 1 since the zero of L(s, χ) cancels the
simple pole at s = 1 of
6! 1
"
L(s, χ0 ) = ζ(s) 1− s .
p
p|m
136 CHAPTER 10. ANALYTIC METHODS

By Exercise 10.3.3, each L(s, χ) extends to an analytic function for Re(s) >
0. By Exercise 10.1.5, ζ(s) (and hence L(s, χ0 )) is analytic for Re(s) > 0,
s != 1. Thus, f (s) is analytic for Re(s) > 0. By Theorem 10.4.2, the
abscissa of convergence of the Dirichlet series
*∞
cn
n=1
ns

is not in Re(s) > 0 which contradicts the divergence of the series at s =

1/ϕ(m). !

Exercise 10.4.6 Show that

, 1
= +∞.
p
p≡1 (mod m)

Exercise 10.4.7 Show that if gcd(a, m) = 1, then

, 1
= +∞.
p
p≡a (mod m)

10.5 Supplementary Problems

Exercise 10.5.1 Define for each character χ (mod m) the Gauss sum
,
g(χ) = χ(a)e2πia/m .
a (mod m)

If (n, m) = 1, show that

,
χ(n)g(χ) = χ(b)e2πibn/m .
b (mod m)

√
Exercise 10.5.2 Show that |g(χ)| = m.

Exercise 10.5.3 Establish the Pólya–Vinogradov inequality:

) )
) )
), )
) χ(n) ) ≤ 1 m1/2 (1 + log m)
) ) 2
)n≤x )

for any nontrivial character χ (mod m).

Exercise 10.5.4 Let p be prime. Let χ be a character mod p. Show that there
is an a ≤ p1/2 (1 + log p) such that χ(a) =
# 1.

Exercise 10.5.5 Show that if χ is a nontrivial character mod m, then

, χ(n) $√ %
m log m
L(1, χ) = +O .
n u
n≤u
10.5. SUPPLEMENTARY PROBLEMS 137

Exercise 10.5.6 Let D be a bounded open set in R2 and let N (x) denote the
number of lattice points in xD. Show that
N (x)
lim = vol(D).
x→∞ x2

Exercise 10.5.7 Let K be an algebraic number field, and C an ideal class of K.

Let N (x, C) be the number of nonzero ideals of OK belonging to C with norm
≤ x. Fix an integral ideal b in C −1 . Show that N (x, C) is the number of nonzero
principal ideals (α) with α ∈ b with |NK (α)| ≤ xN (b).

Exercise 10.5.8 Let K be an imaginary quadratic field, C an ideal class of OK ,

and dK the discriminant of K. Prove that
N (x, C) 2π
lim = 8 ,
x→∞ x w |dK |
where w is the number of roots of unity in K.

Exercise 10.5.9 Let K be a real quadratic field with discriminant dK , and fun-
damental unit ε. Let C be an ideal class of OK . Show that
N (x, C) 2 log ε
lim = √ ,
x→∞ x dK
where N (x, C) denotes the number of integral ideals of norm ≤ x lying in the
class C.

Exercise 10.5.10 Let K be an imaginary quadratic field. Let N (x; K) denote

the number of integral ideals of norm ≤ x. Show that
N (x; K) 2πh
lim = 8 ,
x→∞ x w |dK |
where h denotes the class number of K.

Exercise 10.5.11 Let K be a real quadratic field. Let N (x; K) denote the
number of integral ideals of norm ≤ x. Show that
N (x; K) 2h log ε
lim = 8 ,
x→∞ x |dK |
where h is the class number of K.

Exercise 10.5.12 (Dirichlet’s Class Number Formula) Suppose that K is

a quadratic field with discriminant dK . Show that
 2πh
$ % 
 √ if dK < 0,
∞
, dK 1  w |dK |
=
n n 

n=1  2h√ log ε if dK > 0,
|dK |

where h denotes the class number of K.

138 CHAPTER 10. ANALYTIC METHODS

Exercise 10.5.13 Let d be squarefree and positive.

√ Using√
Dirichlet’s class num-
ber formula, prove that the class number of Q( −d) is O( d log d).

Exercise 10.5.14 Let d be squarefree and positive.

√ Using √Dirichlet’s class num-
ber formula, prove that the class number h of Q( d) is O( d).

Exercise 10.5.15 With ψ(x) defined (as in Chapter 1) by

,
ψ(x) = log p,
pα ≤x

prove that for Re(s) > 1,

' ∞
ζ( ψ(x)
− (s) = s dx.
ζ 1 xs+1

Exercise 10.5.16 If for any ε > 0,

ψ(x) = x + O(x1/2+ε ),

show that ζ(s) #= 0 for Re(s) > 12 .

1
A famous hypothesis of Riemann asserts that ζ(s) != 0 for Re(s) > 2
and this is still (as of 2004) unresolved.
Chapter 11

Density Theorems

Given an algebraic number field K, we may ask how the ideals are dis-
tributed in the ideal classes. We may ask the same of the distribution of
prime ideals. It turns out that in both cases, they are equidistributed in the
sense of probability. For many reasons, it has been customary to view the
latter set of results as generalizations of the celebrated theorem of Dirichlet
about primes in arithmetic progressions.

11.1 Counting Ideals in a Fixed Ideal Class

As usual, let K be a fixed algebraic number field of degree n over Q, and
denote by N (x; K) the number of ideals of OK with norm ≤ x. For an ideal
class C, let N (x, C) be the number of ideals in C with norm ≤ x. Clearly,
*
N (x; K) = N (x, C)
C

where the summation is over all ideal classes of OK . Let us fix an ideal b
in C −1 and note that if a is an ideal in C with norm ≤ x, then ab = (α)
with α ∈ b and |N (α)| ≤ xN (b). Conversely, if α ∈ b and |N (α)| ≤ xN (b),
then a = (α)b−1 is an integral ideal in C with norm ≤ x. Thus, N (x, C)
is the number of principal ideals (α) contained in b with norm less than or
equal to xN (b).
If β1 , ..., βn is an integral basis of b, then we may write

α = x1 β1 + · · · + xn βn

for some integers x1 , ..., xn . Thus, N (x, C) is the number of such α’s (up to
associates), with |N (α)| ≤ xN (b). We will now try to extract a single ele-
ment from the set of such associates by means of inequalities. Let 61 , ..., 6r
be a system of fundamental units (with r = r1 + r2 − 1 as in Theorem
8.1.6). Recall that it is customary (as we did in Chapter 8) to order our

139
140 CHAPTER 11. DENSITY THEOREMS

embeddings K → K (i) in such a way that for 1 ≤ i ≤ r1 , K (i) are real,

and K (i) = K (i+r2 ) for r1 + 1 ≤ i ≤ r1 + r2 . We keep this convention
throughout this discussion. By Exercise 8.1.7, the regulator RK is non-zero
and we may find real numbers c1 , ..., cr such that
r
* E F
(i)
cj log |6j | = log |α(i) ||N (α)|−1/n , 1 ≤ i ≤ r.
j=1

Following Hecke [He], we will call the cj ’s the exponents of α. We now want
to show that this equation also holds for i = r + 1. Setting ei = 1 if K (i) is
real, and ei = 2 if K (i) is non-real, we see that
r+1
* E F
ei log |α(i) ||N (α)|−1/n = 0,
i=1

because
|α(1) · · · α(n) | = |N (α)|.
Also,
r+1
* (i)
ei log |6j | = 0.
i=1

Consequently,
r
* E F
(r+1)
cj log |6j | = log |α(r+1) ||N (α)|−1/n ,
j=1

as desired. Thus, this equation holds for all α(i) , 1 ≤ i ≤ n. (Why?) By

Theorem 8.1.6, every unit u of OK has the form

ζ6n1 1 · · · 6nr r

where ζ is a root of unity in K and the ni ’s are rational integers. Thus,

any associate uα of α has exponents

c1 + n1 , ..., cr + nr .

Therefore, each α has an associate with a set of exponents satisfying the

inequalities
0 ≤ cj < 1, j = 1, 2, ..., r.
If w denotes the number of roots of unity in K, we see then that wN (x, C)
is equal to the number of rational integers (x1 , ..., xn ) not all zero, satisfying
the following conditions:

α = x1 β1 + · · · + xn βn ;
11.1. COUNTING IDEALS IN A FIXED IDEAL CLASS 141

|N (α)| = |α(1) · · · α(n) | ≤ xN (b)

and for 1 ≤ i ≤ n,
E F *r
(i)
log |α(i) ||N (α)|−1/n = cj log |6j |, 0 ≤ cj < 1, j = 1, ..., r.
j=1

In this way, we reduce the problem to counting lattice points in a region

of the Euclidean space Rn . This region can be described as follows. We
choose arbitrary real values for the xj and “define” the set of numbers
n
* (i)
α(i) = xj βj .
j=1

Corresponding to this set of numbers there is a uniquely determined set of

“exponents” c1 , ...., cr provided the xj ’s do not lie on the subspace defined
by α(i) = 0 for some i satisfying 1 ≤ i ≤ n. Thus, if we include α(i) != 0
to the above set of inequalities, these inequalities describe a well-defined
region Bx (say) in Rn . That is, if we put
(i)
α(i) = x1 β1 + · · · + xn βn(i) , 1 ≤ i ≤ n,

and
N (α) = α(1) · · · α(n) ,
then Bx is the set of n-tuples (x1 , ..., xn ) ∈ Rn satisfying

|α(1) · · · α(n) | ≤ xN (b)

and either α(i) = 0 for some i or for all 1 ≤ i ≤ n, we have

E F *r
(i)
log |α(i) ||N (α)|−1/n = cj log |6j |,
j=1

with 0 ≤ cj < 1, j = 1, ..., r.

Exercise 11.1.1 Show that Bx is a bounded region in Rn .

Exercise 11.1.2 Show that tB1 = Btn for any t > 0.

Exercise 11.1.3 Show that N (x, C) = O(x). Deduce that N (x; K) = O(x).

The idea now is to approximate the number of lattice points satisfying

the above inequalities by the volume of Bx . This we do below. As will be
seen, the calculation is an exercise in multivariable calculus.
Before we begin, it is important to note that for some δ > 0 and t = x1/n ,
we have
vol(B(t−δ)n ) ≤ wN (x, C) ≤ vol(B(t+δ)n ).
142 CHAPTER 11. DENSITY THEOREMS

To see this, we may associate each lattice point lying inside Bx with an ap-
propriate translate of the standard unit cube, namely [0, 1]n . Each translate
lying entirely within Bx contributes 1 to the volume of Bx . The error term
arises from the cubes intersecting with the boundary. In view of Exer-
cise 11.1.2, it is intuitively clear (see also Exercise 11.1.12 below) that by
enlarging the region by some fixed quantity and reducing the region by a
fixed quantity δ in the way indicated, the above inequalities are assured.
Thus,
(x1/n − δ)n vol(B1 ) ≤ wN (x, C) ≤ (x1/n + δ)n vol(B1 )
so that
1
wN (x, C) = vol(B1 )x + O(x1− n ).
The essential feature of the theorem below is that this volume is indepen-
dent of the ideal class under consideration.

Theorem 11.1.4 (Dedekind)

2r1 (2π)r2 RK
vol(B1 ) = L .
|dK |

(i)
Proof. Let M be the maximal value of | log |6j || for j = 1, ..., r. We first
complete the domain Bx by adding the points of the space lying in the
subspace α(i) = 0 for some i and that also satisfy the inequalities

|α(j) | ≤ erM (xN (b))1/n , j = 1, 2, ..., n.

Since these subspaces are of lower dimension, their contribution to the

volume is negligible. We denote the completed space by Bx∗ . If we now
change variables and put xj = yj x1/n , we see that the volume is equal to
@ @
x ··· dy1 · · · dyn = J (say).
B1∗

Now B1∗ is the domain described by

n
* (i)
α(i) = yj βj , 1≤i≤n
j=1

and
0 < |α(1) · · · α(n) | ≤ N (b),
so that there exist cj ’s for 1 ≤ j ≤ r satisfying 0 ≤ cj < 1 and
E F *r
(i)
log |α(i) ||N (α)|−1/n = cj log |6j |, 1≤i≤n
j=1
11.1. COUNTING IDEALS IN A FIXED IDEAL CLASS 143

or
|α(i) | ≤ erM (N (b))1/n , 1≤i≤n
and at least one α(i) = 0. As noted, the region defined by these latter
conditions are manifolds of lower dimension and thus make no contribution
to the n-fold integral and thus, these conditions may be omitted in the
evaluation of J. To evaluate the integral, we change variables:
n
* (i)
ui := α(i) = yj βj , 1 ≤ i ≤ r1 ,
j=1

n
*
√ (i)
ui + ui+r2 −1 := yj βj , r1 + 1 ≤ i ≤ r1 + r2 .
j=1

Thus, with our convention concerning the ordering of embeddings,

n
/ (i) (i+r )
0
* βj + βj 2
ui = yj ,
j=1
2

n
/ (i) (i+r2 )
0
* βj − βj
ui+r2 = yj √ ,
j=1
2 −1
for r1 + 1 ≤ i ≤ r1 + r2 . The absolute value of the Jacobian for this change
of variables is easily computed to be
L
2−r2 N (b) |dK |.

Hence, @ @
2r2
vol(B1∗ ) = L ··· du1 · · · dun ,
N (b) |dK | B̃1∗

where B̃1∗ is the image of B1∗ under the change of variables. The variables
u1 , ..., ur1 may take one of two signs and so if we insist ui ≥ 0 for i = 1, ..., r1 ,
we must multiply our volume (with this additional constraint) by a factor
of 2r1 . We now shift to polar co-ordinates. Put

ρj = uj 1 ≤ j ≤ r1

and
ρj cos θj = uj , ρj sin θj = uj+r2 , r1 + 1 ≤ j ≤ r1 + r2 ,
with 0 ≤ θj < 2π and ρj ≥ 0. The Jacobian of this transformation is easily
seen to be
ρr1 +1 · · · ρr1 +r2 .
Thus,
@ @
2r1 +r2 (2π)r2
vol(B1∗ ) = L ··· ρr1 +1 · · · ρr1 +r2 dρ1 · · · dρr1 +r2
N (b) |dK | C1∗
144 CHAPTER 11. DENSITY THEOREMS

where C1∗ is the domain described by:

+r2
r16
e
0≤ ρj j ≤ N (b)
j=1

 −1/n
r
6 r
*
e (i)
log ρi  ρj j  = cj log |6j |
j=1 j=1

for 1 ≤ i ≤ r1 + r2 . (Recall that ei = 1 for 1 ≤ i ≤ r1 and 2 otherwise.)

We make (yet) another change of variables. Put
e
τj = ρj j , 1 ≤ j ≤ r1 + r2 .

The Jacobian of this transformation is easily seen to be

2−r2 ρ−1 −1
r1 +1 · · · ρr1 +r2

so that the integral becomes

@ @
2r1 (2π)r2
L ··· dτ1 · · · dτr1 +r2
N (b) |dK | D1∗

where the region D1∗ is described by the conditions

τ1 · · · τr1 +r2 ≤ N (b), τi > 0,

*r
ei (i)
log τi = log(τ1 · · · τr1 +r2 ) + ei cj log |6j |.
n j=1

We make one final change of variables: write the ci ’s in terms of the τi ’s

and put
u = τ1 · · · τr+1 .
The Jacobian of this transformation is now seen to be RK and the final
result is
@ @ 1 @ 1
2r1 (2π)r2 RK N (b) 2r1 (2π)r2 RK
vol(B1 ) = L du ··· dc1 · · · dcr = L
N (b) |dK | 0 0 0 |dK |

which completes the proof. !

By our remarks preceding the statement of Theorem 11.1.4, we imme-
diately deduce:

Theorem 11.1.5 (Weber)

2r1 (2π)r2 RK 1
N (x, C) = L x + O(x1− n ).
w |dK |
11.1. COUNTING IDEALS IN A FIXED IDEAL CLASS 145

If N (x; K) is the number of integral ideals of K with norm ≤ x, then

2r1 (2π)r2 hK RK 1
N (x; K) = L x + O(x1− n ),
w |dK |

where hK denotes the class number of K.

Following Hecke, we define the ideal class zeta function as

* 1
ζ(s, C) = .
N (a)s
a∈C

Note that the Dedekind zeta function defined in the previous chapter may
now be written as *
ζK (s) = ζ(s, C)
C

where the summation is over all ideal classes.

1
Exercise 11.1.6 Prove that ζ(s, C) extends to the region 5(s) > 1 − n
except
for a simple pole at s = 1 with residue

2r1 (2π)r2 RK
8 .
w |dK |

Deduce that ζK (s) extends to 5(s) > 1 − n1 except for a simple pole at s = 1
with residue
2r1 (2π)r2 hK RK
ρK := 8 ,
w |dK |
where hK denotes the class number of K. (This is usually called the analytic
class number formula.)

Exercise 11.1.7 Prove that there are infinitely many prime ideals ℘ in OK which
are of degree 1.

Exercise 11.1.8 Prove that the number of prime ideals ℘ of degree ≥ 2 and
with norm ≤ x is O(x1/2 log x).

Exercise 11.1.9 Let µ be defined on integral ideals a of OK as follows. µ(OK ) =

1, and if a is divisible by the square of a prime ideal, we set µ(a) = 0. Otherwise,
we let µ(a) = (−1)k when a is the product of k distinct prime ideals. Show that
,
µ(b) = 0
b|a

unless a = OK .
146 CHAPTER 11. DENSITY THEOREMS

Exercise 11.1.10 Prove that the number of ideals of OK of odd norm ≤ x is

#$ 1
%
1
ρK x 1− + O(x1− n ),
N (℘)
℘|2

where the product is over non-zero prime ideals ℘ of OK dividing 2OK .

Exercise 11.1.11 Let A(x) be the number of ideals of OK of even norm ≤ x

and B(x) of odd norm ≤ x. Show that

A(x)
lim =1
x→∞ B(x)

if and only if K = Q or K is a quadratic field in which 2 ramifies.

Exercise 11.1.12 With notation as in the discussion preceding Theorem 11.1.4,

let Vx denote the set of n-tuples (x1 , ..., xn ) satisfying

|α(1) · · · α(n) | ≤ xN (b).

Let t = x1/n . Show that there is a δ > 0 such that for each lattice point P
contained in V(t−δ)n , all the points contained in the translate of the standard
cube by P belong to Vx .

11.2 Distribution of Prime Ideals

Let H be the ideal class group of K. Following Hecke, we define for each
character
χ : H → C∗
the Hecke L-function
* χ(a)
L(s, χ) := ,
a
N (a)s

where χ(a) is simply χ(C) if a belongs to the ideal class C. If χ is the trivial
character χ0 , note that L(s, χ0 ) = ζK (s), the Dedekind zeta function. Since
H is a finite abelian group of order hK , its character group is also finite of
order hK and so, in this way, we have attached hK L-functions to K.

Exercise 11.2.1 Show that L(s, χ) converges absolutely for 5(s) > 1 and that

#$ χ(℘)
%−1
L(s, χ) = 1− ,
℘
N (℘)s

in this region. Deduce that L(s, χ) #= 0 for 5(s) > 1.

11.2. DISTRIBUTION OF PRIME IDEALS 147

Exercise 11.2.2 If χ is not the trivial character, show that

,
χ(C) = 0
C

where the summation is over the ideal classes C of H.

Exercise 11.2.3 If C1 and C2 are distinct ideal classes, show that

,
χ(C1 )χ(C2 ) = 0.
χ

If C1 = C2 , show that the sum is hK . (This is analogous to Exercise 10.3.5.)

From Theorem 11.1.5, we obtain:

Theorem 11.2.4 Let n be the degree of K/Q. If χ != χ0 , then L(s, χ)

extends analytically to <(s) > 1 − n1 .

Proof. By Theorem 11.1.5, we have

* * * 1
χ(a) = χ(C)N (x, C) = O(x1− n ),
C a∈C,N (a)≤x C

since (by Exercise 11.2.2) *

χ(C) = 0.
C
!
In 1917, Hecke proved that L(s, χ) extends to an entire function if χ !=
χ0 , and satisfies a suitable functional equation relating L(1 − s, χ) with
L(s, χ). In the case χ = χ0 , he showed that ζK (s) extends meromorphically
to the entire complex plane, with only a simple pole at s = 1. Moreover, it
satisfies the functional equation

ξK (s) = ξK (1 − s),

where /L 0s
|dK |
ξK (s) := Γ(s/2)r1 Γ(s)r2 ζK (s)
2r2 π n/2
with Γ(s) denoting the Γ-function. Recall that this is defined by
@ ∞
Γ(s) = e−t ts−1 dt
0

for <(s) > 0 and can be extended meromorphically to the entire complex
plane via the functional equation

Γ(s + 1) = sΓ(s).
148 CHAPTER 11. DENSITY THEOREMS

Our goal is to show that each ideal class contains infinitely many prime
ideals. This is analogous to Dirichlet’s theorem about primes in arithmetic
progressions. Indeed, as we will indicate later, the result is more than an
analogue. It is a generalization that includes the celebrated theorem of
Dirichlet.

Exercise 11.2.5 Let C be an ideal class of OK . For 5(s) > 1, show that
, , 1
χ(C) log L(s, χ) = hK
χ ℘m ∈C
mN (℘)ms

where the first summation is over the characters of the ideal class group and the
second summation is over all prime ideals ℘ of OK and natural numbers m such
that ℘m ∈ C.

We now proceed as we did in Chapter 10, Section 4. For the sum on

the right hand side in the previous exercise, we separate the contribution
from n = 1 and n ≥ 2. The latter part is shown to converge for <(s) > 1/2
(see 11.2.6 below). Thus, if we can show that L(1, χ) != 0 for χ != χ0 , we
may conclude as in Exercise 10.4.7 that
* 1
= +∞.
N (℘)
℘∈C

Exercise 11.2.6 Show that

, 1
mN (℘)ms
n≥2,℘m ∈C

converges for 5(s) > 1/2.

Exercise 11.2.7 If χ2 #= χ0 show that L(1, χ) #= 0.

This gives us a fairly self-contained proof of the infinitude of prime ide-

als in a fixed ideal class in the case hK is odd, for in that case, there are
no characters of order 2 in the character group. A genuine difficulty arises
in trying to show L(1, χ) != 0 for χ real. Historically, this was first cir-
cumvented using class field theory (and in most treatments, many authors
still follow this route). A somewhat easier argument allows us to deduce
the non-vanishing from the relatively simpler assertion that for real valued
characters χ, L(s, χ) admits an analytic continuation to <(s) ≥ 1/2. Then,
we may consider the function
* r(a)
f (s) = := ζK (s)L(s, χ)
a
N (a)s

which is easily seen to be a Dirichlet series with non-negative coefficients.

Moreover, it is easily verified that
r(b2 ) ≥ 1.
11.2. DISTRIBUTION OF PRIME IDEALS 149

If L(1, χ) = 0, then f (s) is regular at s = 1 and by Theorem 10.4.2, is

analytic for <(s) > σ0 , where σ0 is the abscissa of convergence of f (s).
Now * 1
lim f (s) ≥ lim = +∞.
s→ 21+ 1+
s→ 2 N (b)2s
b

Thus, σ0 ≥ 1/2. However, the assumption that L(s, χ) admits an analytic

continuation to <(s) ≥ 1/2 implies that f (s) is analytic for <(s) ≥ 1/2, a
contradiction. Thus, L(1, χ) != 0.
In the above argument, we only used the fact that we can continue
L(s, χ) to the real line segment [1/2, 1]. It seems unrealistic to expect any
refinement of this argument unless we use the fact that we are dealing with
a quadratic character in some fundamental way. Indeed, if we consider the
series
ζ(2s)
g(s) = ,
ζ(s)
then it is easy to see that the Dirichlet coefficients of g(s) are ±1. Moreover,
g(s) has a zero at s = 1 and ζ(s)g(s) = ζ(2s) has non-negative coefficients.
However, it does not admit an analytic continuation to the line segment
[1/2, 1] as it has a simple pole at s = 1/2.
So far, we have been able to extend the results of Chapter 10 to show
the infinitude of prime ideals in a fixed ideal class. It is possible to refine
these results by introducing the notion of Dirichlet density. We say that a
set of prime ideals S of prime ideals of OK has Dirichlet density δ if
5
℘∈S 1/N (℘)
s
lim = δ.
s→1+ log ζK (s)

Clearly, any set of prime ideals with a positive Dirichlet density is infinite.

Exercise 11.2.8 Let C be a fixed ideal class in OK . Show that the set of prime
ideals ℘ ∈ C has Dirichlet density 1/hK .

Exercise 11.2.9 Let m be a natural number and (a, m) = 1. Show that the set
of primes p ≡ a(mod m) has Dirichlet density 1/φ(m).

Exercise 11.2.10 Show that the set of primes p which can be written as a2 +5b2
is 1/4.

By using more sophisticated methods, it is possible to obtain asymptotic

formulas for the number of prime ideals lying in a given ideal class. Indeed,
using standard Tauberian theory, one can show that the number of prime
ideals ℘ in a given ideal class with norm ≤ x is asymptotic to
1 x
,
hK log x
150 CHAPTER 11. DENSITY THEOREMS

as x tends to infinity.
It is possible to go further. Let f0 be an ideal of OK and f∞ a subset
of real embeddings of K. We write formally f = f0 f∞ and define the f-ideal
class group as follows. We define an equivalence relation on the set of ideals
coprime to f0 by declaring that two ideals a and b are equivalent if

(α)a = (β)b

for some α, β ∈ OK coprime to f0 , α −β ∈ f0 and σ(α/β) > 0 for all σ ∈ f∞ .

The set of equivalence classes turns out to be finite and can be given the
structure of an abelian group, which we denote by H(f) and call the f-ideal
class group. In the case that f0 = OK and f∞ is the empty set, this group
is the usual ideal class group. If f∞ consists of all the real embeddings of
the given field, we call H(f) the ray class group (mod f0 ). One may now
define L-functions (following Hecke) attached to characters of these groups.
Proceeding as above, the theory can be developed to deduce the expected
density theorems. Indeed, for a fixed f-ideal class C, the set of prime ideals
℘ lying in C has Dirichlet density 1/|H(f)|. It is possible to derive even an
asymptotic formula for the number of such prime ideals ℘ ∈ C with norm
≤ x of the form
1 x
∼
|H(f)| log x
as x tends to infinity.

Exercise 11.2.11 Show that if K = Q, the principal ray class group mod m is
isomorphic to (Z/mZ)∗ .

The previous exercise realizes the coprime residue classes mod m as a

ray class group. In this way, the theorem of Hecke generalizes the classical
theorem of Dirichlet about the uniform distribution of prime numbers in
arithmetic progressions.

11.3 The Chebotarev density theorem

Let K/k be a Galois extension of algebraic number fields with Gal(K/k) =
G. Recall that if ℘ is a prime ideal then so is σ(℘) for any σ ∈ G. For
each prime ideal p of k, we have (analogous to the situation in Chapter 5)
a factorization
pOK = ℘e11 · · · ℘err .
If we apply a Galois automorphism σ to both sides of this equality, we get

pOK = σ(℘1 )e1 · · · σ(℘r )er .

By uniqueness of factorization, we deduce that G acts on the set of prime

ideals ℘1 , ..., ℘r that lie above a fixed prime ideal p.
11.3. THE CHEBOTAREV DENSITY THEOREM 151

Exercise 11.3.1 Show that the action of the Galois group on the set of prime
ideals lying above a fixed prime of k is a transitive action.

The decomposition group of ℘, denoted D℘ , is the subgroup of G con-

sisting of elements σ satisfying σ(℘) = ℘. The inertia group of ℘, denoted
I℘ , is the subgroup of elements σ satisfying

σ(x) ≡ x(mod ℘) ∀x ∈ OK .

It is easily seen that I℘ is a normal subgroup of D℘ . The quotient D℘ /I℘

can be shown to be a cyclic group canonically isomorphic to the Galois
group of the finite field OK /℘ viewed as an extension of Ok /p. Thus,
there is an element (well-defined modulo I℘ ), denoted σ℘ , whose image in
Gal((OK /℘)/(Ok /p)) is the mapping

x 9→ xN (p) .

We call σ℘ the Frobenius automorphism of ℘. For p unramified in K, one

can show easily that as ℘ ranges over the prime ideals above p, the σ℘
comprise a conjugacy class of G. This conjugacy class is denoted σp and is
called the Artin symbol of p.
Now fix a conjugacy class C of G. The Chebotarev density theorem
states the following.

Theorem 11.3.2 (Chebotarev) Let K/k be a finite Galois extension of

algebraic number fields with Galois group G. If C is a conjugacy class of
G, the prime ideals p of Ok with σp ∈ C has Dirichlet density |C|/|G|.
Thus, the Artin symbols are equidistributed in the conjugacy classes with
the expected probability.
A prime ideal p of k is said to split completely in K if

pOK = ℘1 · · · ℘n

where n = [K : k]. This is equivalent to the assertion that the Artin symbol
σp is equal to 1. Thus, from Chebotarev’s density theorem, we immediately
deduce:

Theorem 11.3.3 The set of prime ideals p which split completely in K

has Dirichlet density 1/[K : k].

Exercise 11.3.4 By taking k = Q and K = Q(ζm ), deduce from Chebotarev’s

theorem the infinitude of primes in a given arithmetic progression a (mod m)
with (a, m) = 1.
√
Exercise 11.3.5 If k = Q and K = Q( D), deduce from Chebotarev’s theorem
that the set of primes p with Legendre symbol (D/p) = 1 is of Dirichlet density
1/2.
152 CHAPTER 11. DENSITY THEOREMS

Exercise 11.3.6 If f (x) ∈ Z[x] is an irreducible normal polynomial of degree n

(that is, its splitting field has degree n over Q), then show that the set of primes
p for which f (x) ≡ 0 (mod p) has a solution is of Dirichlet density 1/n.

Exercise 11.3.7 If f (x) ∈ Z[x] is an irreducible polynomial of degree n > 1,

show that the set of primes p for which f (x) ≡ 0 (mod p) has a solution has
Dirichlet density < 1.

Exercise 11.3.8 Let q be prime. Show that the set of primes p for which p ≡ 1
(mod q) and
p−1
2 q ≡ 1(mod p),
has Dirichlet density 1/q(q − 1).

Exercise 11.3.9 If a natural number n is a square mod p for a set of primes p

which has Dirichlet density 1, show that n must be a square.

Let K/k be a finite Galois extension of algebraic number fields with

Galois group G as above. Let V be a finite dimensional vector space over
C. Suppose we have a representation

ρ : G → GL(V )

where GL(V ) denotes the group of invertible linear transformations of V

into itself. E. Artin defined an L-function attached to ρ by setting:
6 # $−1
L(s, ρ; K/k) = det 1 − ρ(σ℘ )N (p)−s |V I℘
p

where the product is over all prime ideals p of k and ℘ is any prime ideal of
K lying above p, which is well-defined modulo the inertia group I℘ . Thus
taking the characteristic polynomial of ρ(σ℘ ) acting on the subspace V I℘ ,
which is the subspace of V fixed by I℘ , we get a well-defined factor for each
prime ideal p. The product over all prime ideals p is easily seen to converge
absolutely for <(s) > 1 (why?). As these L-functions play a central role
in number theory, we will briefly give a description of results pertaining to
them and indicate some of the open problems of the area. The reader may
find it useful to have some basic knowledge of the character theory of finite
groups as explained for instance in [Se].
The celebrated Artin’s conjecture predicts that if ρ is a non-trivial irre-
ducible representation, then L(s, ρ; K/k) extends to an entire function of
s. If ρ is one-dimensional, then Artin proved his famous reciprocity law by
showing that in this case, his L-function coincides with Hecke’s L-function
attached to a suitable generalized ideal class group of k. This theorem is
so-called since it entails all of the classical reciprocity laws including the
law of quadratic reciprocity. Subsequently, R. Brauer proved that for any ρ
L(s, ρ; K/k) extends to a meromorphic function for all s ∈ C. He did this by
proving an induction theorem which is really a statement about irreducible
11.4. SUPPLEMENTARY PROBLEMS 153

characters of finite groups. More precisely, if χ is an irreducible character

of G, then Brauer’s theorem states that there are nilpotent subgroups Hi of
G and one dimensional characters ψi of H so that for some set of integers
ni , we have *
χ= ni IndG
Hi ψi
i

where IndG H ψ indicates the character induced from H to G by ψ.

To see how Brauer’s theorem implies the meromorphy of Artin L-series,
it is convenient to modify our notation slightly by writing L(s, χ; K/k) for
L(s, ρ; K/k) with χ(g) = tr ρ(g). As is evident, the definition of an Artin
L-series attached to ρ depends only on the character χ of ρ. With this
convention, it is easy to verify that

L(s, χ1 + χ2 ; K/k) = L(s, χ1 ; K/k)L(s, χ2 ; K/k)

and that
L(s, IndG
H ψ, K/k) = L(s, ψ; K/K )
H

where K H indicates the subfield of K fixed by H. Thus, by Brauer’s

theorem, we may write
6
L(s, χ; K/k) = L(s, IndG
Hi ψi ; K/k) .
ni

By the invariance of Artin L-series under induction, we obtain

6
L(s, χ; K/k) = L(s, ψi ; K/K Hi )ni .
i

Now, by Artin’s reciprocity law, each of the L-functions appearing in the

product is a Hecke L-function, which by Hecke’s theorem is known to be
entire. In this way, we get the meromorphic continuation of L(s, χ; K/k). It
is one of the aims of the Langlands program to prove Artin’s conjecture. The
celebrated Langlands-Tunnell theorem says that when ρ is 2-dimensional
with solvable image, then Artin’s conjecture is true. This theorem played
a pivotal role in the work of Wiles resolving Fermat’s last theorem.

11.4 Supplementary Problems

Following a suggestion of Kumar Murty (see [FM]), we indicate in the sup-
plementary problems (11.4.1 to 11.4.10) below how Artin L-series may be
used to give a proof of Chebotarev’s theorem using the techniques developed
in this chapter and Chapter 10. A modest background in the representation
theory of finite groups would be useful. For instance, the first three chap-
ters of [Se] should be sufficient background. The reader may also consult
[La] for an alternative (and more classical) approach.
154 CHAPTER 11. DENSITY THEOREMS

Exercise 11.4.1 Let G be a finite group and for each subgroup H of G and each
irreducible character ψ of H define aH (ψ, χ) by
,
IndG
H ψ = aH (ψ, χ)χ
χ

where the summation is over irreducible characters χ of G. For each χ, let Aχ

be the vector (aH (ψ, χ)) as H varies over all cyclic subgroups of G and ψ varies
over all irreducible characters of H. Show that the Aχ ’s are linearly independent
over Q.

Exercise 11.4.2 Let G be a finite group with t irreducible characters. By the

previous exercise, choose a set of cyclic subgroups Hi and characters ψi of Hi
so that the t × t matrix (aHi (ψi , χ)) is non-singular. By inverting this matrix,
show that any character χ of G can be written as a rational linear combination
of characters of the form IndGHi ψi with Hi cyclic and ψi one-dimensional. (This
result is usually called Artin’s character theorem and is weaker than Brauer’s
induction theorem.)

Exercise 11.4.3 Deduce from the previous exercise that some positive integer
power of the Artin L-function L(s, χ; K/k) attached to an irreducible character
χ admits a meromorphic continuation to 5(s) = 1.

Exercise 11.4.4 If K/k is a finite Galois extension of algebraic number fields

with group G, show that
#
ζK (s) = L(s, χ; K/k)χ(1) ,
χ

where the product is over all irreducible characters χ of G.

Exercise 11.4.5 Fix a complex number s0 ∈ C with 5(s0 ) ≥ 1 and any finite
Galois extension K/k with Galois group G. For each subgroup H of G define the
Heilbronn character θH by
,
θH (g) = n(H, χ)χ(g)
χ

where the summation is over all irreducible characters χ of H and n(H, χ) is the
order of the pole of L(s, χ; K/K H ) at s = s0 . By Exercise 11.4.3, the order is a
rational number. Show that θG |H = θH .

Exercise 11.4.6 Show that θG (1) equals the order at s = s0 of the Dedekind
zeta function ζK (s).

Exercise 11.4.7 Show that

,
n(G, χ)2 ≤ (ords=s0 ζK (s))2 .
χ
11.4. SUPPLEMENTARY PROBLEMS 155

Exercise 11.4.8 For any irreducible non-trivial character χ, deduce that

L(s, χ; K/k)

admits an analytic continuation to s = 1 and that L(1, χ; K/k) #= 0.

Exercise 11.4.9 Fix a conjugacy class C in G = Gal(K/k) and choose gC ∈ C.

Show that
, 1 |C| ,
= χ(gC ) log L(s, χ; K/k).
m N (p)ms |G| χ
n,p, σp ∈C

Exercise 11.4.10 Show that

&
p, σp ∈C 1/N (p)s |C|
lim =
s→1+ log ζk (s) |G|

which is Chebotarev’s theorem.

Exercise 11.4.11 Show that ζK (s)/ζk (s) is entire. (This is called the Brauer-
Aramata theorem.)

Exercise 11.4.12 (Stark) Let K/k be a finite Galois extension of algebraic num-
ber fields. If ζK (s) has a simple zero at s = s0 , then L(s, χ; K/k) is analytic at
s = s0 for every irreducible character χ of Gal(K/k).

Exercise 11.4.13 (Foote-K. Murty) For any irreducible character χ of Gal(K/k),

show that
L(s, χ; K/k)ζK (s)
is analytic for s #= 1.

Exercise 11.4.14 If K/k is solvable, show that

,
n(G, χ)2 ≤ (ords=s0 ζK (s)/ζk (s))2 .
χ&=1
 Part II

Solutions
Chapter 1

Elementary Number
Theory

1.1 Integers
Exercise 1.1.7 Show that
1 1 1
1+ + + ··· +
3 5 2n − 1

is not an integer for n > 1.

Solution. Let S = 1 + 13 + 15 + · · · + 2n−11

. We can find an integer k such
that 3 ≤ 2n − 1 < 3
k k+1
. Define m to be the least common multiple of all
the numbers 3, 5, . . . , 2n − 1 except for 3k . Then
m m m
mS = m + + + ··· + .
3 5 2n − 1

Each of the numbers on the right side of this equation is an integer, except
for m/3k . If m/3k were an integer, then there would be some integer b such
that m = 3k b, but 3k does not divide 3, 5, . . . , 3k − 2, 3k + 2, . . . , 2n − 1
so it cannot divide their least common multiple. Therefore mS is not an
integer, and clearly neither is S.

Exercise 1.1.8 Let a1 , . . . , an for n ≥ 2 be nonzero integers. Suppose there is a

prime p and positive integer h such that ph | ai for some i and ph does not divide
aj for all j #= i.
Then show that
1 1
S= + ··· +
a1 an
is not an integer.

159
160 CHAPTER 1. ELEMENTARY NUMBER THEORY

Solution. Let h be the maximum power of p dividing ai . We use the

notation ph =ai to mean that ph | ai but ph+1 ! ai . Let m be the least
common multiple of a1 , . . . , ai−1 , ai /ph , ai+1 , . . . , an . Then
m m m m
mS = + ··· + + + ··· + .
a1 ai−1 ai an

We see that m/aj is an integer for j = 1, 2, . . . , i − 1, i + 1, . . . , n.

However, ai does not divide m, since if it did then ph would clearly have to
divide m, which means we can find a b ∈ Z such that m = ph b. Since ph
does not divide aj for j = 1, . . . , i − 1, i + 1, . . . , n, it does not divide their
least common multiple. Hence m/ai is not an integer, which implies that
mS is not an integer. Therefore
1 1
S= + ··· +
a1 an
is not an integer.

Exercise 1.1.9 Prove that if n is a composite integer, then n has a prime factor
√
not exceeding n.

Solution. Since n is composite, we can write√ n = ab where a √ and b are

integers with
√ √ 1 < a ≤ b < n. We have a ≤ n since otherwise n<a≤b
and ab > n n = n. Now, a certainly has a prime divisor, and any prime
divisor of a is also a prime
√ divisor of n. Hence n has a prime factor which
is less than or equal to n.

Exercise 1.1.10 Show that if the smallest prime factor p of the positive integer
√
n exceeds 3 n, then n/p must be prime or 1.

Suppose that the smallest prime factor p of the positive integer

Solution. √
n exceeds 3 n. Then p > n1/3 . Hence L n/p < n2/3 . If n/p is composite,
n/p has
L a prime factor not exceeding n/p by Exercise 1.1.9. We see
that n/p < n1/3 . A prime factor of n/p is also that of n, and so we
have found a prime factor which is smaller than n1/3 , which contradicts
our assumption. Therefore n/p is a prime or 1.

!Exercise
" 1.1.11 Let p be prime. Show that each of the binomial coefficients
p
k
, 1 ≤ k ≤ p − 1, is divisible by p.

Solution. Since ! "

p p!
=
k k! (p − k)!
we see that the numerator is divisible by p, and the denominator is not, for
1 ≤ k ≤ p − 1. The result is now evident.

Exercise 1.1.12 Prove that if p is an odd prime, then 2p−1 ≡ 1 (mod p).
1.1. INTEGERS 161

Solution.
p−1 ! "
* p
2 = (1 + 1)
p p
= 1+ +1
k
k=1
≡ 1+1 (mod p)

by the previous exercise.

Exercise 1.1.13 Prove Fermat’s little Theorem: If a, p ∈ Z with p a prime, and

p ! a, prove that ap−1 ≡ 1 (mod p).

Solution. We can apply induction. For instance,

p−1 ! "
* p
3p = (1 + 2)p = 1+ 2k + 2p
k
k=1
≡ 1 + 2p (mod p)

since the binomial coefficients are divisible by p. By the previous exercise

2p ≡ 2 (mod p) and so we find that

3p ≡ 3 (mod p).

Alternate Solution. We consider the field Z/pZ, obtained by taking

congruences mod p. Let a denote the class of a (mod p). If p ! a, then
a !≡ 0 (mod p), and so a is a unit in the field Z/pZ. The units of this field
form a multiplicative group G of order p − 1. By elementary group theory,
a|G| = ap−1 = 1, which means that ap−1 ≡ 1 (mod p).

Exercise 1.1.15 Show that n | φ(an − 1) for any a > n.

Solution. an ≡ 1 (mod an − 1) and n is the smallest power of a with this

property. Thus, a has order n (mod an − 1). Therefore, n | φ(an − 1).

Exercise 1.1.16 Show that n ! 2n − 1 for any natural number n > 1.

Solution. Let us suppose the set of n > 1 such that 2n ≡ 1 (mod n) is

nonempty. By the well-ordering principle, there is a least such number, call
it n0 . Then
2n0 ≡ 1 (mod n0 ).
By Euler’s theorem,
2φ(n0 ) ≡ 1 (mod n0 ).
Let d = (n0 , φ(n0 )). By the Euclidean algorithm, we can find integers x
and y so that
n0 x + φ(n0 )y = d.
162 CHAPTER 1. ELEMENTARY NUMBER THEORY

Thus, 2d ≡ 1 (mod n0 ). If d > 1, this gives 2d ≡ 1 (mod d) contradicting

the minimality of n0 . Thus, d = 1 and we get

2≡1 (mod n0 )

which is also a contradiction.

Exercise 1.1.17 Show that

φ(n) #$ 1
%
= 1−
n p
p|n

by interpreting the left-hand side as the probability that a random number chosen
from 1 ≤ a ≤ n is coprime to n.

Solution. The probability that a number chosen from 1 ≤ a ≤ n is

coprime to n is clearly φ(n)/n. On the other hand, this is tantamount to
insisting that our number is not divisible by any prime divisors of n, which
is represented by the right-hand side of the formula.

Exercise 1.1.18 Show that φ is multiplicative (i.e., φ(mn) = φ(m)φ(n) when

gcd(m, n) = 1) and φ(pα ) = pα−1 (p − 1) for p prime.

Solution. By the previous exercise, it is clear that φ is multiplicative.

When n = pα , we find
! "
1
φ(pα ) = pα 1 − = pα−1 (p − 1).
p

Exercise 1.1.19 Find the last two digits of 31000 .

Solution. We find the residue class that 31000 belongs to in Z/100Z. This
is the same as finding the last two digits. By Euler’s theorem, 340 ≡ 1
(mod 100), since

φ(100) = φ(4)φ(25) = 2(20) = 40.

Therefore,
31000 = (340 )25 ≡ 1 (mod 100).
The last two digits are 01.

Exercise 1.1.20 Find the last two digits of 21000 .

Solution. We need to find the residue class of 21000 in Z/100Z. Since 2

is not coprime to 100, we cannot apply Euler’s theorem as in the previous
exercise. However, we have

x = 21000 ≡ 1 (mod 25),

1.1. INTEGERS 163

x = 21000 ≡ 0 (mod 4).

We determine which residue classes z (mod 100) satisfy z ≡ 1 (mod 25)
and z ≡ 0 (mod 4).
The last condition means z = 4k. We solve 4k ≡ 1 (mod 25). Thus,
6(4k) ≡ 6 (mod 25) so that k ≡ −6 (mod 25). That is, k ≡ 19 (mod 25).
Hence, z = (19)4 = 76. This means

21000 ≡ 76 (mod 100).

The last two digits are 76.

Exercise 1.1.21 Let pk denote the kth prime. Prove that

pk+1 ≤ p1 p2 · · · pk + 1.

Solution. We see that the number p1 p2 · · · pk +1 is coprime to p1 , p2 , . . . , pk

and either must be prime, or divisible by a prime different from p1 , . . . , pk .
Thus,
pk+1 ≤ p1 p2 · · · pk + 1.

Exercise 1.1.22 Show that k

pk < 22 ,
where pk denotes the kth prime.

Solution. From the preceding exercise, we know that pk+1 ≤ p1 · · · pk + 1.

1 2
Now we have p1 < 22 and p2 < 22 . Suppose that pk < 22 is true for
k

2 < k ≤ n.
Then

pn+1 ≤ p1 p2 · · · p n + 1
1 2
< 22 22 · · · 22 + 1
n

= 22 −2 + 1
n+1

< 22
n+1
.

Hence pn < 22 is true for n ≥ 1.

n

Exercise 1.1.23 Prove that π(x) ≥ log(log x).

Solution. From the previous exercise, we have that pn < 22 for n ≥ 1.

n

Hence we can see that π(22 ) ≥ n.

n

n−1 n
For x > 2, choose an integer n so that ee < x ≤ ee . Then

log x ≤ en log e = en

and
log(log x) ≤ n log e = n.
164 CHAPTER 1. ELEMENTARY NUMBER THEORY

If n > 2, then en−1 > 2n .

n−1
π(x) ≥ π(ee )
≥ π(e2 )
n

≥ π(22 )
n

≥ n
≥ log(log x).

This proves the result.

Exercise 1.1.24 By observing that any natural number can be written as sr2
with s squarefree, show that √
x ≤ 2π(x) .
Deduce that
log x
π(x) ≥ .
2 log 2
Solution. For any set of primes S define fS (x) to be the number of integers
n such that 1 ≤ n ≤ x with γ(n) ⊂ S where γ(n) is the set of primes
dividing n. Suppose that S is a finite set with t elements. Write such an
n in√the form n = r2 s with s squarefree. Since 1 ≤ r2 s ≤ x, we see that
r ≤ x and there are at most 2t choices for s corresponding
√ to the various
subsets of S since s is squarefree. Thus fS (x) ≤ 2t x.
Put π(x) = m so that pm+1 > x. If S = {p1 , . . . , pm }, then fS (x) = x.
Then √ √
x ≤ 2m x = 2π(x) x.
√
Thus x ≤ 2π(x) and hence 12 log x ≤ π(x) log 2, or equivalently,

log x
π(x) ≥ .
2 log 2
&
Exercise 1.1.25 Let ψ(x) = pα ≤x log p where the summation is over prime
powers pα ≤ x.
(i) For 0 ≤ x ≤ 1, show that x(1 − x) ≤ 14 . Deduce that
' 1
1
xn (1 − x)n dx ≤ n
0 4
for every natural number n.
(1 n
(ii) Show that eψ(2n+1) 0
x (1−x)n dx is a positive integer. Deduce that ψ(2n+
1) ≥ 2n log 2.
(iii) Prove that ψ(x) ≥ 12 x log 2 for x ≥ 6. Deduce that

x log 2
π(x) ≥
2 log x
for x ≥ 6.
1.1. INTEGERS 165

Solution. Clearly 4x2 − 4x + 1 = (2x − 1)2 ≥ 0 so (i) is now immediate.

M1
The integral 0 xn (1 − x)n dx consists of a sum of rational numbers whose
denominators are less than 2n + 1. Since lcm(1, 2, . . . , 2n + 1) = eψ(2n+1) ,
we find @ 1
eψ(2n+1) xn (1 − x)n dx
0

is a positive integer. Thus, eψ(2n+1) ≥ 22n . This proves (ii).

For (iii), choose n so that 2n − 1 ≤ x < 2n + 1. Then, by (ii),

ψ(x) ≥ ψ(2n − 1) ≥ (2n − 2) log 2 > (x − 3) log 2.

For x ≥ 6, x − 3 > x/2 so that ψ(x) > x log 2/2. Since ψ(x) ≤ π(x) log x,
we deduce that
x log 2
π(x) ≥
2 log x
for x ≥ 6.

Exercise 1.1.26 By observing that

)* +
# )
) 2n
p ) ,
) n
n<p≤2n

show that
9x log 2
π(x) ≤
log x
for every integer x ≥ 2.

Solution. Since 1! "

6 1 2n
p 11 ,
n
n<p≤2n

we deduce that *
log p ≤ 2n log 2
n<p≤2n

because ! "
2n
≤ 22n .
n
Therefore
θ(2n) − θ(n) ≤ 2n log 2,
where *
θ(n) = log p.
p≤n

An easy induction shows that θ(2r ) ≤ 2r+1 log 2 for every positive integer
r. given an integer x ≥ 2, determine r so that

2r ≤ x < 2r+1 .
166 CHAPTER 1. ELEMENTARY NUMBER THEORY

Then
θ(x) ≤ θ(2r+1 ) ≤ 2r+2 log 2 ≤ 4x log 2.
We deduce, in particular,
*
log p ≤ 4x log 2,
√
x<p≤x

so that #1 $# √ $
2 log x π(x) − π( x) ≤ 4x log 2.
This means
√ 8x log 2
π(x) − π( x) ≤
log x
and
8x log 2 √ 9x log 2
π(x) ≤ + x≤
log x log x
because
√ x log 2
x≤
log x
for x ≥ 10, as is easily checked by examining the graph of
√
f (x) = x log 2 − log x.

For x ≤ 10, the inequality is verified directly.

1.2 Applications of Unique Factorization

Exercise 1.2.1 Suppose that a, b, c ∈ Z. If ab = c2 and (a, b) = 1, then show
that a = d2 and b = e2 for some d, e ∈ Z. More generally, if ab = cg then a = dg
and b = eg for some d, e ∈ Z.
β1 β2
Solution. Write a = pα 1 α2
1 p2 · · · pr and b = q1 q2 · · · qs where pi and
αr βs

qj are primes for 1 ≤ i ≤ r and 1 ≤ j ≤ s and pi != qj for any i, j since

(a, b) = 1.
β1
ab = (p1α1 · · · pα
r )(q1 · · · qs )
r βs

= c2
= p2γ 1 2γr
1 · · · pr · q12θ1 · · · qs2θs ,

where c = pγ11 · · · pγr r q1θ1 · · · qsθs .

By unique factorization, αi = 2γi and βj = 2θj for 1 ≤ i ≤ r and
1 ≤ j ≤ s. Hence we can write a = p2γ 1 2γr
1 · · · pr and b = q12θ1 · · · qs2θs .
2 2
Hence ∃d, e ∈ Z such that a = d and b = e where d = pγ11 · · · pγr r and
e = q1θ1 · · · qsθs .
The argument for gth powers is identical.
1.2. APPLICATIONS OF UNIQUE FACTORIZATION 167

Exercise 1.2.2 Solve the equation x2 + y 2 = z 2 where x, y, and z are integers

and (x, y) = (y, z) = (x, z) = 1.

Solution. Assume that x and y are odd. Then both x2 ≡ 1 (mod 4) and
y 2 ≡ 1 (mod 4). Hence z 2 ≡ 2 (mod 4). But there is no z ∈ Z satisfying
z 2 ≡ 2 (mod 4), so one of x or y is even.
Without loss of generality, suppose x is even and y is odd. Then z is
odd. We have x2 = z 2 − y 2 , so

x2 z2 − y2
= ,
4 4
E x F2 (z + y) (z − y)
⇒ = .
2 2 2
Since (x, y) = (y, z) = (x, z) = 1, we see that ((z + y)/2, (z − y)/2) = 1. By
Exercise 1.2.1, there exist a, b ∈ Z such that (z + y)/2 = a2 and (z − y)/2 =
b2 . Hence we have the two equations z + y = 2a2 and z − y = 2b2 .
Thus the solution is x = 2ab, y = a2 −b2 , and z = a2 +b2 where (a, b) = 1
and a and b have opposite parity since y and z are odd. Conversely, any
such triple (x, y, z) gives rise to a solution.

Exercise 1.2.3 Show that x4 +y 4 = z 2 has no nontrivial solution. Hence deduce,

with Fermat, that x4 + y 4 = z 4 has no nontrivial solution.

Solution. Suppose that x4 + y 4 = z 2 has a nontrivial solution. Take |z| to

be minimal. By Exercise 1.2.2, we can write

x2 = 2ab, (1.1)
2 2 2
y = b −a , (1.2)
z = b2 + a2 , (1.3)

with (x, y) = 1 and a and b having opposite parity.

Suppose that b is even. Then we see that

y 2 = b2 − a2 ≡ −1 ≡ 3 (mod 4).

This is impossible. Hence a is even. Then ∃c ∈ Z such that a = 2c and

(c, b) = 1. Then x2 = 2 · 2bc = 4bc. Since (b, c) = 1, b and c are perfect
squares by Exercise 1.2.1. Hence ∃m, n ∈ Z such that b = m2 , c = n2
where (m, n) = 1. By (1.2), we see that y 2 = b2 − a2 = m4 − 4n4 . Hence
(2n2 )2 + y 2 = (m2 )2 and (2n2 , y) = (y, m2 ) = (2n2 , m2 ) = 1.
By Exercise 1.2.2, 2n2 = 2αβ, y = β 2 − α2 , and m2 = α2 + β 2 where
(α, β) = 1 and α and β have opposite parity. Thus we can see that n2 = αβ.
Hence by Exercise 1.2.1, ∃p, q ∈ Z such that α = p2 and β = q 2 . Hence we
have m2 = p4 + q 4 . This is a solution of the equation x4 + y 4 = z 2 . But
m < b < |z| since m2 = b < b2 + a2 = z. This is a contradiction to the
minimality of |z|. Therefore x4 + y 4 = z 2 has no nontrivial solution.
168 CHAPTER 1. ELEMENTARY NUMBER THEORY

Now suppose that x4 + y 4 = z 4 has a nontrivial solution. This would

imply that x4 + y 4 = t2 where t = z 2 has a nontrivial solution. But we
proved above that this is impossible, so x4 + y 4 = z 4 has no nontrivial
solution.

Exercise 1.2.4 Show that x4 − y 4 = z 2 has no nontrivial solution.

Solution. Suppose that x4 − y 4 = (x2 + y 2 )(x2 − y 2 ) = z 2 has a nontrivial

solution, and choose the solution such that |x| is minimal. If x is even,
then both y and z must be odd (since x, y, z are coprime). But then we can
rewrite the equation as (x2 )2 = z 2 + (y 2 )2 and we know from Exercise 1.2.2
that this equation has no solutions for x even. So x is odd.
Suppose that y is odd. We again write the equation as (x2 )2 = z 2 +(y 2 )2 ,
and we see that by Exercise 1.2.2 we can write
z = 2ab, y 2 = a2 − b2 , x2 = a2 + b2 ,
for relatively prime integers a, b. Now,
a4 − b4 = (a2 + b2 )(a2 − b2 ) = x2 y 2 = (xy)2 ,
4 4 2
√ have found another solution to the equation x − y = z . But
and we
a < a + b = x, contradicting our assumption about x. We conclude
2 2

that there are no solutions for y odd.

Now suppose that y is even. Then we can use Exercise 1.2.2 and write
y 2 = 2cd, z = c2 − d2 , x2 = c2 + d2 ,
where c, d are coprime and of opposite parity. Without loss, we assume that
c is even, d odd. But then we have (2c, d) = 1, and we can use Exercise 1.2.1
which says that we can find integers s, t such that 2c = s2 , d = t2 . In fact,
s is even so we can write s = 2u, and thus c = 2u2 . Therefore we can write
x2 = c2 + d2 = (2u2 )2 + (t2 )2 . We now deduce that we can find integers v, w
such that 2u2 = 2vw, t2 = v 2 − w2 , x = v 2 + w2 . Since u2 = vw, we can
write v = a2 , w√= b2 . But looking back, we see that t2 = v 2 − w2 = a4 − b4 ,
and since a = v < v 2 + w2 = x, which is a contradiction. So x4 − y 4 = z 2
has no nontrivial solutions.

Exercise 1.2.5 Prove that if f (x) ∈ Z[x], then f (x) ≡ 0 (mod p) is solvable for
infinitely many primes p.

Solution. We will call p a prime divisor of f if p | f (n) for some n. Clearly

f always has a prime divisor. Hence it suffices to show that f has infinitely
many prime divisors. Suppose that f has only finitely many prime divisors.
Let f (x) = an xn + an−1 xn−1 + · · · + a1 x + a0 and let p1 , . . . , pk be the
prime divisors of f . For simplicity, we will write m = p1 · · · pk . Then
f (a0 m) = an (a0 m)n + · · · + a1 (a0 m) + a0
= a0 (an a0 n−1 mn + an−1 a0 n−2 mn−1 + · · · + a1 m + 1).
1.2. APPLICATIONS OF UNIQUE FACTORIZATION 169

Let g(x) = an a0 n−1 xn + an−1 a0 n−2 xn−1 + · · · + a1 x + 1. Then we can see

that (pi , g(m)) = 1 for 1 ≤ i ≤ k. Hence g(m) has a prime divisor different
from pi for 1 ≤ i ≤ k. The prime divisor of g(m) is also that of f (a0 m).
Hence we can see that there is a new prime divisor of f different from pi for
1 ≤ i ≤ k. This is a contradiction. Therefore f has infinitely many prime
divisors.

Exercise 1.2.6 Let q be prime. Show that there are infinitely many primes p so
that p ≡ 1 (mod q).

Solution. Let us consider the polynomial

xq − 1
f (x) = = 1 + x + · · · + xq−1 ,
x−1
and suppose that p is a prime divisor of f (x). Then xq ≡ 1 (mod p)
for some x. Let x0 be an integer such that f (x0 ) ≡ 0 (mod p). Then
x0 q ≡ 1 (mod p). If x0 is not congruent to 1 (mod p), then q is the order
of x0 (mod p) since q is a prime. Consider the multiplicative group G =
{1, . . . , p − 1}. We see that x0 ∈ G. Since q is the order of x0 (mod p),
we can see that q | (p − 1). Hence p − 1 ≡ 0 (mod q) and hence p ≡ 1
(mod q). If x0 ≡ 1 (mod p), then 1 + x0 + · · · + xq−1
0 ≡ 0 (mod p) means
p = q. Therefore, any prime divisor of f is either q or ≡ 1 (mod q).
By Exercise 1.2.5, there are infinitely many primes p such that f (x) ≡ 0
(mod p) is solvable since f (x) ∈ Z[x]. We conclude that there are infinitely
many primes p such that p ≡ 1 (mod q).

Exercise 1.2.7 Show that Fn divides Fm − 2 if n is less than m, and from this
deduce that Fn and Fm are relatively prime if m #= n.

Solution. Write m = n + k where k is a nonzero positive integer. Then

Fm − 2 Fn+k − 2
=
Fn Fn
22
n+k
−1
=
22 n + 1
(22 )2 − 1
n k

=
22 n + 1
t2 − 1
k

= t2 −1 − t2 −2 + · · · − 1,
k k
=
t+1

where t = 22 . Hence Fn divides Fm − 2.

n

Let d be the greatest common divisor of Fn and Fm . Then d | Fn , so

d | Fm − 2 and d | Fm . Hence d | 2 and hence d = 1 or 2. Since Fn and Fm
are odd, d = 1. Thus (Fn , Fm ) = 1. Therefore Fn and Fm are relatively
prime if m != n.
170 CHAPTER 1. ELEMENTARY NUMBER THEORY

n
Exercise 1.2.8 Consider the nth Fermat number Fn = 22 +1. Prove that every
prime divisor of Fn is of the form 2n+1 k + 1.

Solution. Let p be a prime divisor of 22 + 1. Then 22 ≡ −1 (mod p),

n n

so (22 )2 ≡ 1 (mod p). Hence we have 22

n n+1
≡ 1 (mod p). We will show
that the order of 2 (mod p) is 2n+1 . Let x = ordp 2. Since x | 2n+1 we
can write x = 2m where m is an integer and 1 ≤ m ≤ n + 1. Hence for all
n ≥ m, 22 ≡ 1 (mod p). But by assumption 22 ≡ −1 (mod p), which
n n

implies that 22 ≡ 1 (mod p) holds only if m ≥ n + 1. We now consider

m

the multiplicative group G = {1, . . . , p − 1}. We must have 2n+1 | p − 1

since ordp 2 = 2n+1 and the order of G is p − 1. Therefore we can write
p − 1 = 2n+1 k where k is an integer, and we conclude that p = 2n+1 k + 1.
α
Exercise 1.2.9 Given a natural number n, let n = pα 1
1 · · · pk
k
be its unique
factorization as a product of prime powers. We define the squarefree part of n,
denoted S(n), to be the product of the primes pi for which αi = 1. Let f (x) ∈ Z[x]
be nonconstant and monic. Show that lim inf S(f (n)) is unbounded as n ranges
over the integers.

Solution. By Exercise 1.2.5, we know that f has infinitely many prime

divisors. Let p be such a prime and suppose f (x0 ) ≡ 0 (mod p). Observe
that f (x0 + p) ≡ f (x0 ) + pf $ (xo ) (mod p2 ). We define the discriminant of
a monic polynomial f to be
6
(ri − rj )2 ,
i>j

where r1 , . . . , rn are the roots of f . If p | f $ (x0 ), then p would divide the

discriminant of f . (Why? see Exercise 4.3.3.) Choosing p sufficiently large,
we may assume this does not happen. In either case, we deduce that the
squarefree part of f (x0 ) is divisible by p or the squarefree part of f (x0 + p)
is divisible by p. If S(f (n)) were bounded, we have derived a contradiction.

1.3 The ABC Conjecture

Exercise 1.3.1 Assuming the ABC Conjecture, show that if xyz #= 0 and xn +
y n = z n for three mutually coprime integers x, y, and z, then n is bounded.
# $
Solution. First observe that max |x|, |y|, |z| > 1 for otherwise we have
xyz = 0. By the ABC Conjecture, we have
# # $1+ε
max |x|n , |y|n , |z|n ) ≤ κ(ε) rad(xyz) .
# $
Without any loss of generality, suppose that max |x|, |y|, |z| = |z|. We de-
duce that |z|n ≤ κ(ε)|z|3+3ε . Since |z| > 1, we conclude that n is bounded.

Exercise 1.3.2 Let p be an odd prime. Suppose that 2n ≡ 1 (mod p) and

2n #≡ 1 (mod p2 ). Show that 2d #≡ 1 (mod p2 ) where d is the order of 2 (mod p).
1.3. THE ABC CONJECTURE 171

Solution. Since 2n ≡ 1 (mod p), we must have d | n. Write n = de. If

2d = 1 + kp and p | k, then
2n = 2de = (1 + kp)e
≡ 1 + kpe (mod p2 )
≡ 1 (mod p2 ),
a contradiction. This proves the result.

Exercise 1.3.3 Assuming the ABC Conjecture, show that there are infinitely
many primes p such that 2p−1 #≡ 1 (mod p2 ).

Solution. Let us write 2n − 1 = un vn where un is the squarefree part of

2n − 1 and vn is the squarefull (or powerfull) part of 2n − 1. (Recall that a
number N is called squarefull (powerfull) if for every prime q | N we have
q 2 | N . Thus for any number N , N/S(N ) is squarefull (or powerfull) with
S(N ) the squarefree part of N .) Therefore (un , vn ) = 1.
We begin by showing that if p | un , then 2p−1 !≡ 1 (mod p2 ). Indeed, we
know that p | 2n −1 and p2 ! 2n −1. (As defined earlier in this chapter, pα =n
means that pα | n but pα+1 ! n. In this case, we would write p=2n − 1.) By
Exercise 1.3.2, p2 ! 2d − 1 where d is the order of 2 (mod p). Now d | (p − 1)
by the little theorem of Fermat and Lagrange’s theorem. Write df = p − 1.
Then 2d = 1 + kp with p ! k so that 2p−1 = (1 + kp)f ≡ 1 + kf p (mod p2 ).
Since f | p − 1 and p ! k, we find that 2p−1 !≡ 1 (mod p2 ) for every prime p
dividing un .
Now suppose there are only finitely many such primes p. Since un
is squarefree, this implies that un is bounded. Now consider the ABC
equation:
(2n − 1) + 1 = 2n .
The ABC Conjecture implies that
# $1+ε
2n ≤ κ(ε) 2 rad(2n − 1) .
1/2
But rad(2n − 1) ≤ un vn and so
# $1+ε
un vn = 2n − 1 < 2n ≤ κ(ε) 2un vn1/2 .
Since un is bounded, this implies that vn is bounded, and hence n is
bounded, a contradiction.
Remark. This is due to J. Silverman [Sil] who also obtains, assuming the
ABC Conjecture, that the number of primes p ≤ x for which 2p−1 !≡ 1
(mod p2 ) is > log x.

Exercise 1.3.4 Show that the number of primes p ≤ x for which

2p−1 #≡ 1 (mod p2 )
is ( log x/ log log x, assuming the ABC Conjecture.
172 CHAPTER 1. ELEMENTARY NUMBER THEORY

Solution. If for any n, we have un = 1, then the ABC Conjecture implies,

as above, that n is bounded. Thus for n sufficiently large (say n > N ),
un > 1. For each prime q satisfying N < q ≤ (log x)/ log 2, we have uq > 1.
Moreover, for any two distinct primes q1 and q2 , gcd(2q1 − 1, 2q2 − 1) = 1
because p | 2q1 − 1 and p | 2q2 − 1 implies that the order of 2 (mod p)
divides q1 and q2 and so it divides their gcd, which is 1. This implies that
p | 1, which is a contradiction.
Thus, for every prime p | uq , we find 2p−1 !≡ 1 (mod p2 ). In addition,
the u$q s are mutually coprime. In this way we obtain
! " log x
log x log 2 log 2
π ≥ # log x $ by Exercise 1.1.25
log 2 2 log log 2
log x
=
2(log log x − log log 2)
log x
> ,
log log x
primes p < x such that 2p−1 !≡ 2 (mod p2 ).

Exercise 1.3.5 Show that if the Erdös conjecture above is true, then there are
infinitely many primes p such that 2p−1 #≡ 1 (mod p2 ).
2
Solution. Suppose for p > p0 that 2p−1 ≡ 1 (mod p2 ). Let t = p≤p0 p.
Then 6
φ(t) = (p − 1).
p≤p0

Now consider the sequence cn = 2 − 1. We claim that cn is powerfull.

ntφ(t)

Indeed if 2 < p ≤ p0 , then by Euler’s theorem p2 | cn . If p > p0 , and

p | cn , then p2 | vntφ(t) by the argument in Exercise 1.3.3. Thus p2 | cn and
so cn is squarefull. For n even, say n = 2k, we deduce that both 2ktφ(t) − 1
and 2ktφ(t) + 1 are powerfull. But then, so is 2ktφ(t) , contrary to the Erdös
conjecture.

Exercise 1.3.6 Assuming the ABC Conjecture, prove that there are only finitely
many n such that n − 1, n, n + 1 are squarefull.

Solution. If n − 1, n, n + 1 are squarefull, consider the ABC equation

(n2 − 1) + 1 = n2 .
Then,
# $1+ε
n2 ≤ κ(ε) rad(n2 (n2 − 1))
# √ √ $1+ε
≤ κ(ε) n1/2 n − 1 n + 1
since n, n − 1, and n + 1 are all squarefull. The inequality implies that n
is bounded. (This is due to A. Granville.)
1.4. SUPPLEMENTARY PROBLEMS 173

Exercise 1.3.7 Suppose that a and b are odd positive integers satisfying

rad(an − 2) = rad(bn − 2)

for every natural number n. Assuming ABC, prove that a = b. (This problem is
due to H. Kisilevsky.)

Solution. Suppose without loss that a < b. Hence log b > log a so we can
choose ε > 0 so that log b > (1 + ε) log a. Now apply the ABC Conjecture
to the equation (bn − 2) + 2 = bn . Then
# $1+ε
bn ≤ κ(ε) 2b rad(bn − 2)
# $1+ε
≤ κ(ε) 2b rad(an − 2)
# $1+ε
≤ κ(ε) 2ban .

Taking nth roots and letting n → ∞ gives log b ≤ (1 + ε) log a, which is a

contradiction. This completes the proof.

Of course, we may consider the equation rad(an − c) = rad(bn − c)

for a fixed integer c coprime to a and b. The above argument applies in
this context as well. Recently, R. Schoof and C. Corrales-Rodrigáñez [Sc]
established this result in the special case c = 1 without assuming ABC.
It is also worth observing that we do not need the equation

rad(an − 2) = rad(bn − 2)

satisfied for all natural numbers n, but just an infinite subsequence.

1.4 Supplementary Problems

Exercise 1.4.1 Show that every proper ideal of Z is of the form nZ for some
integer n.

Solution. Suppose there is an ideal I for which this is not true. Then
show that there exist elements a, b ∈ I such that gcd(a, b) = 1.

Exercise 1.4.2 An ideal I is called prime if ab ∈ I implies a ∈ I or b ∈ I. Prove

that every prime ideal of Z is of the form pZ for some prime integer p.

Solution. If I is an ideal, then it is of the form nZ for some integer n by

the previous question. Then ab ∈ I implies that n | ab. But then since I is
prime, either a ∈ I or b ∈ I, so n | a or n | b, implying that n is prime.

Exercise 1.4.3 Prove that if the number of prime Fermat numbers is finite, then
the number of primes of the form 2n + 1 is finite.
174 CHAPTER 1. ELEMENTARY NUMBER THEORY

Solution. Consider primes of the form 2n + 1. If n has an odd factor, say

n = rs with r odd, then 2rs + 1 is divisible by 2s + 1, and is therefore not
prime.

Exercise 1.4.4 If n > 1 and an − 1 is prime, prove that a = 2 and n is prime.

Solution. If a > 2, then an − 1 is divisible by a − 1. So assume a = 2.

Then if n has a factor, say k, then 2k − 1 | 2n − 1. Therefore if an − 1 is
prime, a = 2 and n is prime. Numbers of this form are called Mersenne
numbers.

Exercise 1.4.6 Prove that if p is an odd prime, any prime divisor of 2p − 1 is of

the form 2kp + 1, with k a positive integer.

Solution. Suppose q is a prime divisor of 2p − 1. Then q must be odd. We

note that 2p ≡ 1 (mod q) and also, by Fermat’s little Theorem, 2q−1 ≡ 1
(mod q). Then p | (q − 1) since p is prime. Then q ≡ 1 (mod p) so
q = mp + 1 but since q is odd, m = 2k for some k, and so q = 2kp + 1.

Exercise 1.4.7 Show that there are no integer solutions to the equation x4 −y 4 =
2z 2 .

Solution. We will consider only solutions with gcd(x, y) = 1, since any

common factor of x, y will also divide z. Therefore any solution to this
equation with gcd(x, y) != 1 will lead to a solution with gcd(x, y) = 1.
We notice that since the right-hand side of the equation is even, x and y
are either both even or both odd. Since x, y are coprime, they must be odd.
Then x4 − y 4 ≡ 0 (mod 4) and so z is even. We can factor the equation as

(x2 + y 2 )(x2 − y 2 ) = 2z 2 .

We note that x2 + y 2 ≡ 2 (mod 4) and x2 − y 2 ≡ 0 (mod 4), so (x2 + y 2 )/2

is odd. Now we rewrite our equation as
! 2 "
x + y2
(x2 − y 2 ) = z 2 .
2

If there is an integer δ such that δ | (x2 + y 2 )/2 and δ | x2 − y 2 , then δ | 2x2

and δ | 2y 2 . But x, y are relatively prime and so δ | 2. We know that
2 ! (x2 + y 2 )/2 so δ = 1 and our two factors are coprime.
This implies

x2 + y 2 = 2a2 ,
x2 − y 2 = 4b2 ,

since x2 − y 2 ≡ 0 (mod 4). We now factor this second equation above as

! "! "
x+y x−y
= b2 .
2 2
1.4. SUPPLEMENTARY PROBLEMS 175

It is easy to see that the two factors are coprime, and so we can write

x+y = 2c2 ,
x−y = 2d2 .

Now we notice that we have

(x + y)2 + (x − y)2 = 2x2 + 2y 2 = 2(x2 + y 2 ).

Given our expressions above, this translates into the equation

4c4 + 4d4 = 4a2 ,

but we know that x4 + y 4 = z 2 has no solutions in Z. Thus, the given

equation has no solution.

Exercise 1.4.8 Let p be an odd prime number. Show that the numerator of
1 1 1
1+ + + ··· +
2 3 p−1
is divisible by p.

Solution. Look at the sum modulo p.

Exercise 1.4.9 Let p be an odd prime number greater than 3. Show that the
numerator of
1 1 1
1 + + + ··· +
2 3 p−1
is divisible by p2 .

Solution. Pair up 1/i and 1/(p − i) and consider the sum mod p.

Exercise 1.4.10 (Wilson’s Theorem) Show that n > 1 is prime if and only
if n divides (n − 1)! + 1.

Solution. When n is prime, consider (n − 1)! (mod n) by pairing each

residue class with its multiplicative inverse.

Exercise 1.4.11 For each n > 1, let Q be the product of all numbers a < n
which are coprime to n. Show that Q ≡ ±1 (mod n).

Solution. Q is clearly congruent to the product of elements of order 2.

Now pair up a and (n − a).

Exercise 1.4.12 In the previous exercise, show that Q ≡ 1 (mod n) whenever

n is odd and has at least two prime factors.

Solution. Clearly Q ≡ (−1)s (mod n) where 2s is the number of elements

a satisfying a2 ≡ 1 (mod n). Use the Chinese Remainder Theorem (see
Exercise 5.3.13) to determine s.
176 CHAPTER 1. ELEMENTARY NUMBER THEORY

Exercise 1.4.13 Use Exercises 1.2.7 and 1.2.8 to show that there are infinitely
many primes ≡ 1 (mod 2r ) for any given r.

Solution. If p | Fn , then p ≡ 1 (mod 2n+1 ). For each n ≥ r, we have

p ≡ 1 (mod 2r ). By Exercise 1.2.7 these primes are all distinct.

Exercise 1.4.14 Suppose p is an odd prime such that 2p + 1 = q is also prime.

Show that the equation
xp + 2y p + 5z p = 0
has no solutions in integers.

Solution. Consider the equation mod q. Then xp ≡ ±1 or 0 (mod q).

Exercise 1.4.15 If x and y are coprime integers, show that if

xp + y p
(x + y) and
x+y
have a common prime factor, it must be p.

Solution. Suppose a prime q is a common factor. Then

xp + y p
0≡ ≡ xp−1 − xp−2 y + · · · + y p−1 ≡ pxp−1 (mod q).
x+y

Exercise 1.4.16 (Sophie Germain’s trick) Let p be a prime such that 2p +

1 = q > 3 is also prime. Show that
xp + y p + z p = 0
has no integral solutions with p ! xyz.

Solution. By the previous exercise, x + y = ap and (xp + y p )/(x + y) = cp

for some integers a and c. By symmetry, y + z = bp , x + z = dp . If
q ! xyz, then xp + y p + z p ≡ 0 (mod q) is impossible since ±1 ± 1 ± 1 ≡ 0
(mod q) is impossible. Now suppose q | xyz. If q | x, then q ! yz so that
2x + y + z ≡ bp ≡ ap + dp (mod q) which again is impossible if a, b, and
d are coprime to q. Thus one of these must be divisible by q. It is easy to
see that this must be b. Thus, y + z ≡ 0 (mod q). Since
yp + zp
y+z
is also a pth power, tp (say), we obtain the congruence
tp ≡ py p−1 (mod q).
Since q does not divide t, we deduce that
py p−1 ≡ ±1 (mod q).
Also, x + y = ap implies y ≡ ap (mod q) so that y is a pth power mod q
which is coprime to q. Thus, p ≡ ±1 (mod q), a contradiction.
1.4. SUPPLEMENTARY PROBLEMS 177

Exercise 1.4.17 Assuming ABC, show that there are only finitely many con-
secutive cubefull numbers.

Solution. If n − 1 and n are cubefull, then apply ABC to n − (n − 1) = 1.

Exercise 1.4.18 Show that ,1

= +∞,
p
p

where the summation is over prime numbers.

Solution. Clearly,

* 1 6! 1
"−1 6 !
1 1
"
≤ 1− = 1 + + 2 + ···
n p p p
n≤x p≤x p≤x

since every natural number n ≤ x can be written as a product of primes

p ≤ x. Now take logs. Then
 
*1 * 1
+ O(1) ≥ log  .
p n
p≤x n≤x

Since the harmonic series diverges, the result follows.

Exercise 1.4.19 (Bertrand’s Postulate) (a) If a0 ≥ a1 ≥ a2 ≥ · · · is a de-

creasing sequence of real numbers tending to 0, show that
∞
,
(−1)n an ≤ a0 − a1 + a2 .
n=0

&
(b) Let T (x) = n≤x ψ(x/n), where ψ(x) is defined as in Exercise 1.1.25. Show
that
T (x) = x log x − x + O(log x).
(c) Show that
-x. , -x.
T (x) − 2T = (−1)n−1 ψ = (log 2)x + O(log x).
2 n
n≤x

Deduce that -x.

ψ(x) − ψ ≥ 13 (log 2)x + O(log x).
2
Solution. From
∞
*
(−1)n an = a0 − (a1 − a2 ) − (a3 − a4 ) − · · · ,
n=0

(a) is immediate.
178 CHAPTER 1. ELEMENTARY NUMBER THEORY

To see (b), observe that

 
* * * * ExF
log n =  log p = ψ = T (x).
α
m
n≤x n≤x p |n m≤x

By comparing areas,
* @ x
log n = (log t) dt + O(log x)
n≤x 1

implies (b).
The first part of (c) is now clear. Since ψ(x/n) is a decreasing function
of n, we apply (a) to get
ExF ExF
ψ(x) − ψ +ψ ≥ (log 2)x + O(log x).
2 3
By Exercise 1.1.14, ψ(x) ≤ 2x log 2. Therefore,
ExF
ψ(x) − ψ ≥ 13 (log 2)x + O(log x).
2
Hence, there is a prime between x/2 and x for x sufficiently large.
(This simple proof is due to S. Ramanujan. We can deduce ψ(x) ≤
2x log 2 directly from (a) and (b) without using the solution to Exercise
1.1.26.)
Chapter 2

Euclidean Rings

2.1 Preliminaries
√
Exercise 2.1.2 Let D be squarefree. Consider R = Z[ D]. Show that every
element of R can be written as a product of irreducible elements.
√
Solution. We define a map n : R → N such that for a + b D ∈ R,
√
n(a + b D) = |a2 − Db2 |.

We must check that this map satisfies conditions (i) and (ii) from the pre-
vious example. √ √
(i) For a + b D, c + d D ∈ R,
√ √ √
n[(a + b D)(c + d D)] = n[(ac + bdD) + (ad + bc) D]
= |(ac + bdD)2 − (ad + bc)2 D|
= |(a2 − b2 D)(c2 − d2 D)|
√ √
= n(a + b D)n(c + d D),

so condition (i) is satisfied.

√ √
(ii) If r = a + b D is a unit in R, then ∃s = c + d D ∈ R such that
rs = 1. But by condition (i), since 1 = n(1), 1 = n(r)n(s). Since our map
n only takes on values in the positive integers, then n(r) = n(s) = 1 for all
units of R. The converse is clear.
Since we have found a map n which satisfies the conditions of Exam-
ple 2.1.1, we can deduce that every element of R can be written as a product
of irreducible elements.

√ √ √
Exercise 2.1.3 Let R = Z[ −5]. Show that 2, 3, 1 + −5, and 1 − −5 are
irreducible in R, and that they are not associates.

179
180 CHAPTER 2. EUCLIDEAN RINGS

√
Solution. We define a map n : R → N such that n(a + b −5) = a2 + 5b2 .
If 2 is not irreducible, then there are elements r, s ∈ R such that rs = 2,
with r, s not units. But then n(r)n(s) = n(2) = 4, and since r, s are not
units, it must be that n(r) = n(s) = 2. Then we must find integers a, b
such that a2 +5b2 = 2, which is clearly impossible, so 2 must be irreducible.
If 3 is not irreducible then we can find r, s ∈ R with rs = 3, and r, s
not units. Since n(3) = 9, we must have n(r) = n(s) = 3. But by the same
argument √ as above, we see that this is impossible.
n(1√+ −5) = 6. The only proper divisors of 6 are 2 and 3, and so
if 1 + √ −5 is not irreducible, then we can find r ∈ R, r not a unit and
r | (1 + −5) with either n(r) √ = 2 or n(r) = 3. But we showed √ above that
this √
is not possible, so 1 + −5 is irreducible. Since n(1 − −5) = 6, then
1 − −5 must also be irreducible.
If two elements of R are associates, then they must have the same norm,
a fact which√follows immediately from the condition that all units have norm
1. If a + b −5 is a unit, then a2 +√5b2 = 1. This will only occur √ when
a = ±1, and so the only units of Z[ −5] are 1 and −1. Of 2, 3, 1 ± √−5,
we see that√ the only two which could possibly be associates are 1 + −5
and √1 − −5 because they have the√same norm. However, if√we multiply
1 + −5 by either of the units of Z[ −5], we will not get 1 − −5, and so
they cannot be associates. √ √
We conclude that 2, 3, 1 + −5 and 1 − −5 are all irreducible and are
not associates.

Exercise 2.1.4 Let R be a domain satisfying (i) above. Show that (ii) is equiv-
alent to (ii$ ): if π is irreducible and π divides ab, then π | a or π | b.

Solution. Suppose R satisfies both (i) and (ii) above. Let π ∈ R be an

irreducible element, and suppose that π | ab, where

a = τ1 τ 2 · · · τ r ,

b = γ1 γ2 · · · γs ,
and τi , γj are irreducible.
We know that π | ab = τ1 · · · τr γ1 · · · γs , so it follows that ab = πλ1 · · · λn
where each λi is irreducible. By condition (ii), π ∼ τi for some i, or π ∼ γj
for some j. Thus, if π | ab, then π | a or π | b.
Now suppose that R is a domain satisfying conditions (i) and (ii0 ) above,
and suppose that we have an element a which has two different factoriza-
tions into irreducibles: a = τ1 · · · τr and a = π1 · · · πs . Consider τ1 . We
know that τ1 | a, and so τ1 | π1 · · · πs . By (ii0 ) we know that τ1 | πi for
some i, and since both are irreducible, they must be associates.
We can now remove both τ1 and πi from our factorization of a. We
next consider τ2 . Following the same process, we can pair up τ2 with its
associate, and we can continue to do this until we have paired up each of
2.2. GAUSSIAN INTEGERS 181

the irreducible factors τi with its associate πj . It is clear that if we continue

in this fashion, we must have r = s.

Exercise 2.1.5 Show that if π is an irreducible element of a principal ideal

domain, then (π) is a maximal ideal (where (x) denotes the ideal generated by
the element x).

Solution. We define gcd(a, b), the greatest common divisor of a, b ∈ R, to

be an element d such that the ideal (a, b) equals the ideal (d). It is unique
up to units. For a unique factorization domain, this definition coincides
with the usual one. We note that d must divide both a and b since they
are in the ideal (d).
If π is irreducible, we consider the ideal (π, α) where α is any element
not in (π). Since α is not a multiple of π, and π is irreducible, then the
only common divisors of α and π will be units. Then gcd(π, α) = 1. In
other words, the ideal generated by π is a maximal ideal.

Exercise 2.1.8 If F is a field, prove that F [x], the ring of polynomials in x with
coefficients in F , is Euclidean.

Solution. We define a map φ : F [x] → N such that for f ∈ F [x], φ(f ) =

deg(f ). Now consider any two polynomials f (x), g(x) ∈ F [x].
If deg(g) > deg(f ), then we can certainly write f (x) = 0 · g(x) + f (x),
which satisfies the Euclidean condition. Then we can assume that m =
deg(g) ≤ deg(f ) = n, and write f (x) = a0 + a1 x + · · · + an xn , and g(x) =
b0 + b1 x + b2 x2 + · · · + bm xm , where an , bm != 0 and m ≤ n. We proceed
by induction on the degree of f . That is, we will prove by induction on
the degree of f that we can write f (x) = q(x)g(x) + r(x), where r = 0 or
deg(r) < deg(g).
Define a new polynomial

h(x) = an b−1
m g(x)x
n−m
.

Observe that the leading term of h(x) is an b−1 m n−m

m bm x x = an xn which
is the leading term of f (x), so that if f1 (x) = f (x) − h(x), either f1 (x) = 0
or deg(f1 ) < deg(f ). The theorem holds for f1 (x), so we can write f1 (x) =
f (x) − h(x) = q(x)g(x) + r(x), where r = 0 or deg(r) < deg(g) = m. Now
f (x) = q(x)g(x)+h(x)+r(x) and since h(x) is a multiple of g(x), the result
follows.

2.2 Gaussian Integers

Exercise 2.2.1 Show that Z[i] is Euclidean.

Solution. We define a map φ : Z[i] → N such that φ(a + bi) = a2 + b2 .

Now, given any two elements of Z[i], say α = a + bi and γ = c + di, can we
find q, r ∈ Z such that a+bi = q(c+di)+r, where r = 0 or φ(r) < φ(c+di)?
182 CHAPTER 2. EUCLIDEAN RINGS

Since we cannot divide α and γ in the ring Z[i], we move temporarily

to the ring Q[i] = {r + si | r, s ∈ Q}. In this ring,

α (a + bi)
=
γ (c + di)
(ac + bd) (bc − ad)
= + 2 i
(c2 + d2 ) (c + d2 )
= r + si,

with r, s ∈ Q. We can now choose m, n ∈ Z such that |r − m| ≤ 1/2, and

|s − n| ≤ 1/2. We set q = m + ni. Then q ∈ Z[i], and α = qγ + r for some
suitable r, with

φ(r) = φ(α − qγ)

= φ(α/γ − q)φ(γ)
= [(r − m)2 + (s − n)2 ]φ(γ)
#1 1$
≤ 4 + 4 φ(γ)
1
= 2 φ(γ)
< φ(γ).

We have shown that our map φ satisfies the properties specified above, and
so Z[i] is Euclidean.

Exercise 2.2.2 Prove that if p is a positive prime, then there is an element

x ∈ Fp := Z/pZ such that x2 ≡ −1 (mod p) if and only if either p = 2 or p ≡ 1
(mod 4). (Hint: Use Wilson’s theorem, Exercise 1.4.10.)

Solution. If p = 2, then 1 ≡ −1 (mod 2), so 12 = 1 ≡ −1 (mod 2). Hence

we can take x = 1. Conversely, if 1 ≡ −1 (mod p), we can see that p = 2
since 1 = ap − 1 for some integer a which implies ap = 2.
We will show that in any field Fp where 1 is not congruent to −1
(mod p), x2 ≡ −1 (mod p) for an element x if and only if x has order
4 in the group of units of the field. Suppose that x2 ≡ −1 (mod p). Then
the first four powers of x are x, −1, −x, 1. Hence x has order 4.
Conversely, suppose that x has order 4. Then, x4 = (x2 )2 ≡ 1 (mod p),
so (x2 )2 − 1 ≡ 0 (mod p). Hence (x2 + 1)(x2 − 1) ≡ 0 (mod p). Since x is
an element of a field, x2 + 1 ≡ 0 (mod p) or x2 − 1 ≡ 0 (mod p). However,
if x2 − 1 ≡ 0 (mod p), x has order 2. Hence x2 ≡ −1 (mod p).
If p != 2, then Fp is a field where 1 is not congruent to −1 (mod p).
Hence the existence of an element x such that x2 ≡ −1 (mod p) is equiv-
alent to the existence of an element of order 4 in the group of units of Fp .
Let Up be the group of units of Fp . Then |Up | = p − 1, and since the order
of any element divides the order of the group, if Up has an element of order
4, then we have 4 | p − 1 or, equivalently, p ≡ 1 (mod 4). Conversely, if we
2.2. GAUSSIAN INTEGERS 183

suppose that p ≡ 1 (mod 4), then we can write p = 4k + 1 where k is an

integer and Up is a cyclic group of order p − 1 = 4k. If g is a generator of
Up , then g has order 4k. So g k has order 4. Hence we can see that if 1 is
not congruent to −1 mod p, x2 ≡ −1 (mod p) occurs if and only if x has
order 4 in the group of units of the field, which occurs if and only if p ≡ 1
(mod 4).
Alternate Solution: Wilson’s theorem gives

(p − 1)! ≡ −1 (mod p).

We can pair up k and (p − k) in the product above so that

k(p − k) ≡ −k 2 (mod p)

implies ! "
(p−1)/2 p−1 2
(−1) ! ≡ −1 (mod p).
2
Thus, if p ≡ 1 (mod 4), there is an x ∈ Fp so that x2 ≡ −1 (mod p).
The converse follows from Fermat’s little Theorem:

1 ≡ (x2 )(p−1)/2 ≡ (−1)(p−1)/2 (mod p)

so that (p − 1)/2 is even. That is, p ≡ 1 (mod 4).

We will provide another alternative proof of this fact in Chapter 7, using
quadratic residues.

Exercise 2.2.3 Find all integer solutions to y 2 + 1 = x3 with x, y #= 0.

Solution. If x is even, then x3 ≡ 0 (mod 8), which implies in turn that

y 2 ≡ 7 (mod 8). However, if y ≡ 1, 3, 5, 7 (mod 8), then y 2 ≡ 1 (mod 8).
So x must be odd, and y even.
We can factor this equation in the ring Z[i] to obtain (y + i)(y − i) = x3 .
If ∃δ such that δ | (y + i) and δ | (y − i), then δ | 2i, which implies that
δ | 2. But this would mean that x is divisible by 2, which we know is not
true. Therefore, we know that (y + i) and (y − i) are relatively prime in
Z[i], and that they must both be cubes.
We know that we can write y + i = e1 (a + bi)3 and y − i = e2 (c + di)3
where a, b, c, d ∈ Z and e1 , e2 are units in Z[i]. However, the only units of
Z[i] are ±1 and ±i, and these are all cubes, so without loss, assume that
e1 = e2 = 1.
Next, we expand our factorization for y + i to get

y + i = a3 + 3a2 bi + 3ab2 − b3 i.

Comparing the imaginary parts, we get 1 = 3a2 b − b3 = b(3a2 − b2 ), with

a, b ∈ Z. The only integers which multiply together to give 1 are ±1, so
184 CHAPTER 2. EUCLIDEAN RINGS

we know that b = ±1. If b = 1, then we have 1 = 3a2 − 1, implying

3a2 = 2, which has no integer solutions, so b != 1. If b = −1, then we get
1 = −3a2 + 1, and so 3a2 = 0, and a must be 0. However, if we substitute
a = 0 back into our original equation for y + i, we find that y = 0, which
we did not allow, and so b != −1.
We conclude that the equation y 2 + 1 = x3 has no integer solutions with
x, y != 0.

Exercise 2.2.4 If π is an element of R such that when π | ab with a, b ∈ R, then

π | a or π | b, then we say that π is prime. What are the primes of Z[i]?

Solution. Given π = (a+bi) ∈ Z[i], we define the complex conjugate of π to

be the element π = (a−bi). We note that n(π) = a2 +b2 = ππ, and so given
any prime π in Z[i], we know that π divides n(π). Using this information,
we observe that we can find all the Gaussian primes by examining the
irreducible factors of the primes of Z. For, let n(π) = p1 p2 · · · pk be the
prime decomposition of n(π). We know that π | n(π), so π | pi for some i.
If π | pi and π | pj , with pi != pj , then π | 1, since gcd(pi , pj ) = 1. But then
π would be a unit, and thus not irreducible. So, by examining all of the
divisors of the primes in Z, we will discover all of the primes of Z[i], once
and only once each.
We let π be a prime in Z[i], and p the prime in Z such that π | p. By
the properties of the map n, n(π) | n(p) = p2 , so n(π) = p or n(π) = p2 . If
we let π = a + bi, then a2 + b2 = p, or a2 + b2 = p2 .
All the primes of Z are congruent to 1, 2 or 3 (mod p), and we will
examine these cases separately.
Case 1. p ≡ 3 (mod 4).
We just proved that if π = a + bi is prime, then either a2 + b2 = p, or
a2 + b2 = p2 for some integer prime p. Let us assume that the first of these
possibilities is true. We know that p is odd, so one of a, b is even. Let us
say that a is even, and b odd, so that a = 2x and b = 2y + 1 for some
x, y ∈ Z. Then

a2 + b2 = 4x2 + 4y 2 + 4y + 1
= 4(x2 + y 2 + y) + 1
≡ 1 (mod 4).

Since we had assumed that p ≡ 3 (mod 4), we have a contradiction. So

a2 + b2 = p2 , which means that n(π) = n(p), and so p and π must be
associates.
Therefore, primes in Z that are congruent to 3 (mod 4) and their asso-
ciates are prime in the ring Z[i].
Case 2. p ≡ 2 (mod 4).
There is, of course, only one such integer prime: 2. Assume we have
a prime π which divides 2. Since 2 = (1 + i)(1 − i), then π | (1 + i) or
2.3. EISENSTEIN INTEGERS 185

π | (1 − i). But n(1 + i) = n(1 − i) = 2, and it is easy to show that (1 + i)

and (1 − i) are irreducible in Z[i] and so they are prime. So, if π | 2, then
π ∼ (1 + i) or π ∼ (1 − i).
Case 3. p ≡ 1 (mod 4).
We recall Wilson’s Theorem, Exercise 1.4.10, which states that if p is a
prime, then (p − 1)! ≡ −1 (mod p). We will in fact be using a corollary of
this theorem, which states that if p is a prime number of the form 1 + 4m,
then p | (n2 + 1), where (2m)! = n. (We can also apply Exercise 2.2.2.)
If p | (n2 + 1) = (n + i)(n − i) and π | p, then π | (n + i) or π | (n − i).
If p were to divide (n ± i), then p | n and p | 1, which is clearly not the
case since p is not a unit. We conclude that p and π are not associates,
so n(π) != n(p), which implies that n(π) = a2 + b2 = p. Thus, if p ≡ 1
(mod 4), then p does not remain prime in Z[i]. We can deduce that if
π = a ± bi and a2 + b2 = p, then π is prime in Z[i].

Exercise 2.2.5 A positive integer a is the sum of two squares if and only if
a = b2 c where c is not divisible by any positive prime p ≡ 3 (mod 4).

Solution. Suppose that a is the sum of two squares. Let a = s2 + t2 and

let (s, t) = b. Then a = (bx)2 + (by)2 = b2 (x2 + y 2 ) where (x, y) = 1. Let
c = x2 + y 2 . Then we have a = b2 c where c is the sum of two relatively
prime squares.
By Exercise 2.2.2, c is not divisible by any prime p ≡ 3 (mod 4). In
fact, suppose that p | x2 + y 2 . Then, x2 + y 2 ≡ 0 (mod p), x2 ≡ −y 2
(mod p) and so ((y −1 )2 · x2 ) = (y −1 · x)2 ≡ −1 (mod p). By Exercise 2.2.2,
either p = 2 or p ≡ 1 (mod 4). Hence c is not divisible by any prime p ≡ 3
(mod 4).
Now suppose that we have an integer a which we can write as a = b2 c,
and suppose that c is not divisible by any positive prime p ≡ 3 (mod 4).
Then c is a product of primes each of which, by Exercise 2.2.4, is a sum of
two squares. Then b2 = n(b) and c = n(t + ri) where t and r are integers.
Then

b2 · c = n(b(t + ri))
= n(bt + bri)
= b 2 · t2 + b 2 · r 2
= (bt)2 + (br)2 .

Hence a = b2 c is written as the sum of two squares.

2.3 Eisenstein Integers

Exercise 2.3.1 Show that Z[ρ] is a ring.
186 CHAPTER 2. EUCLIDEAN RINGS

Solution. First observe that Z[ρ] is a subset of the complex numbers, so

associativity, distributivity, and commutativity are immediate for addition
and multiplication. Also, 0, 1 ∈ Z[ρ], so we have additive and multiplicative
identities. If a + bρ ∈ Z[ρ], then −a − bρ ∈ Z[ρ], so we have additive
inverses. It remains to verify closure under addition and multiplication; if
a, b, c, d ∈ Z, then (a + bρ) + (c + dρ) = (a + c) + (b + d)ρ ∈ Z[ρ], so we have
closure under addition. Also (a+bρ)(c+dρ) = ac+(ad+bc)ρ+bdρ2 . We will
therefore have closure under multiplication if ρ2 ∈ Z[ρ]. But ρ2 = −1 − ρ,
so Z[ρ] is a commutative ring with unit.

Exercise 2.3.2 (a) Show that Z[ρ] is Euclidean.

(b) Show that the only units in Z[ρ] are ±1, ±ρ, and ±ρ2 .

Solution. (a) Define φ : Z[ρ] → N so that φ(a + bρ) = a2 − ab + b2 = αα

for α ∈ Z[ρ]. We consider α, β ∈ Z[ρ], β != 0. We have

α αβ
= .
β ββ

Now ββ ∈ Z and αβ ∈ Z[ρ], so

αβ
= s + tρ
ββ

for some s, t ∈ Q. We set m and n to be the integers closest to s and t,

respectively, i.e., choose m and n so |m − s| ≤ 1/2 and |n − t| ≤ 1/2. We
set q = m + nρ. Now,
! "
α
φ −q = (s − m)2 − (s − m)(t − n) + (t − n)2
β
≤ 14 + 1
4 + 1
4
< 1.

So, writing r = α − qβ, then if r != 0, φ(r) = φ(β)φ(α/β − q) < φ(β), and

with the map φ, for any α, β ∈ Z[ρ], we can write α = qβ + r where r = 0
or φ(r) < φ(β). Thus, Z[ρ] is Euclidean.
(b) Observe that φ is a multiplicative map into the natural numbers, so
that if η is a unit of Z[ρ], then φ(η) = 1. We thus see immediately that
±1, ±ρ, ±ρ2 are all units (it is easy to see that they are distinct). Suppose
η = a+bρ were a unit of Z[ρ]. Then a2 −ab+b2 = 1 and (2a−b)2 +3b2 = 4.
From this equation it is clear that b = 0 or ±1 are the only possible integer
values b could take, since 3b2 must be less than 4. To each solution for b
there are two corresponding solutions for a, and thus at most six distinct
pairs (a, b) in total. By the pigeonhole principle the list given above includes
all the units of Z[ρ].
2.4. SOME FURTHER EXAMPLES 187

Exercise 2.3.3 Let λ = 1 − ρ. Show that λ is irreducible, so we have a factor-

ization of 3 (unique up to unit).

Solution. We have that φ(λ) = 3. If d | λ, then φ(d) | 3, i.e., φ(d) = 1 or

3, so d is either a unit or an associate of λ, and λ is irreducible.

Exercise 2.3.4 Show that Z[ρ]/(λ) has order 3.

Solution. Suppose α ∈ Z[ρ], so that α = a + bρ for integers a, b. Then

α = a + b − b(1 − ρ) = a + b − bλ ≡ a + b (mod λ). Considered mod 3,
a + b could have residues 0, 1, or 2. Since λ | 3 (see Exercise 2.3.3, above),
then α will have one of these residues mod λ. Since φ(λ) does not divide
φ(1) = 1, or φ(2) = 4, none of these classes are equivalent mod λ, and so
we have three distinct residue classes, which we may denote by 0 and ±1.

2.4 Some Further Examples

√
Exercise 2.4.2 Show that Z[ −2] is Euclidean.
√ √
Solution. We√define a norm φ : Z[ −2] → N by φ(a + b −2) = a2 + 2b2 .
For α, β ∈ Z[ −2], we √ consider α/β =√αβ/ββ. Notice that ββ = φ(β) √ so
ββ ∈ Z. Also, β ∈ Z[ −2], so αβ ∈ Z[ −2], so α/β = αβ/ββ = c + d −2
for some c, d ∈ Q. We choose m and n as the closest integers to√c and d,
i.e., so that |m − c| ≤ 1/2 and |n − d| ≤ 1/2. We write q = m + n −2. We
have that φ(α/β − q) = (c − m)2 + 2(d − n)2 ≤ 1/4 + 1/2 < 1. So we write
α = qβ + r and r = α √− qβ. If r != 0, then φ(r) = φ(β)φ(α/β − q) < φ(β).
We conclude that Z[ −2] is Euclidean.

Exercise 2.4.3 Solve y 2 + 2 = x3 for x, y ∈ Z.

√ √
Solution. Write (y + −2)(y − −2) = x3 . If y were even, then x would
be also, but if x is even, then x3 ≡ 0 (mod 8) whereas
√ 8 does not divide
√
y 2 + 2. So y and x are both odd. Observe that (y + −2) and (y − √−2)
are relatively prime, since if d divided both, then d would divide 2 −2
and would
√ thus have even norm, which is not possible since y is odd.
√ Thus
(y + −2) is a cube multiplied by a unit. The only units of Z[ −2] are
1 and −1, which are both cubes. Without loss, assume that the unit in
question is 1. We write
√ √
(y + −2) = (a + b −2)3
√
= a3 − 6ab2 + (3a2 b − 2b3 ) −2.

Comparing real and imaginary parts, we find that

y = a3 − 6ab2
188 CHAPTER 2. EUCLIDEAN RINGS

and

1 = (3a2 b − 2b3 )
= b(3a2 − 2b2 ).

Thus, b | 1 so b = ±1. It follows that a = ±1. Substituting into the

equation for y, we find that y = ±5. Thus, the only solution to the given
equation is x = 3, y = ±5.
√
Exercise 2.4.5 Show that Z[ 2] is Euclidean.
√ √
Solution. We √ define a norm φ : Z[ √2] → N by φ(a + b 2) = |a2 − 2b2 |.
Let α, β ∈ Z[ 2]. We write β = c + d 2 and consider
√
α α(c − d 2)
= √ .
β β(c − d 2)
√ √ √ √
Notice that√|β(c−d √2)| = φ(β) so β(c−d 2) ∈ Z. Also, (c−d 2) ∈ Z[ 2],
so α(c − d 2) ∈ Z[ 2], so
√
α α(c − d 2) √
= √ =t+u 2
β β(c − d 2)
for some t, u ∈ Q. We choose m and n as the closest integers to t√and u,
i.e. so that |m − t| ≤ 1/2 and |n − u| ≤ 1/2. We write q = m + n 2. We
have that

φ(α/β − q) = | (t − m)2 − 2(u − n)2 |

≤ | (t − m)2 | + | 2(u − n)2 |
1 1
≤ 4 + 2
< 1.

We write α = qβ + r, so r =√α − qβ. If r != 0, then φ(r) = φ(β)φ(α/β − q) <

φ(β). We conclude that Z[ 2] is Euclidean.
√ √
Exercise 2.4.6 Let ε = 1+ 2. Write εn = un +vn 2. Show that u2n −2vn2 = ±1.

Solution. Since φ is multiplicative and we have φ(ε) = | − 1|, then

φ(εn ) = |(−1)n | = |u2n − 2vn2 | = 1.

This gives infinitely many solutions to x2√− 2y 2 = ±1. It is easy to see that
all of these solutions are distinct: ε ∈ Z[ 2] and ε > 1 so εn+1 > εn for all
positive n.
√ √
Exercise 2.4.7 Show that there is no unit η in Z[
√ 2] such that 1 < η < 1 √
+ 2.
Deduce that every unit (greater than zero) of Z[ 2] is a power of ε = 1 + 2.
2.5. SUPPLEMENTARY PROBLEMS 189

Solution. Since −1 is a unit, for any unit ξ, −ξ is also a unit, and negative
and positive units are in√one-to-one correspondence;√we shall only consider
the positive units
√ of Z[ √2]. We write η as a + b 2. Since√ η is a unit,
φ(η) =√(a + b √ 2)(a − b 2) = ±1. By assumption (a + b 2) > 1 and
|(a + b 2)(a − b 2)| = 1, so it follows that
√
−1 < (a − b 2) < 1.
√ √
Also, by assumption, 1 < a + b 2 < 1 + 2. So, adding these two inequal-
ities gives √
0 < 2a < 2 + 2.
Since a ∈ Z this implies that a = 1. Notice now that there is no integer b
such that √ √
1 < 1 + b 2 < 1 + 2.
If any unit, ψ, did exist which was not some power of√ε, then by our
Euclidean algorithm√we would be able√ tok+1divide by (1 + 2)k , where k is
chosen so that (1 + 2) < ψ √
k
< (1 + 2) and this would produce √a new
unit ψ $ where 1 < ψ $ <
√ 1 + 2. So the only positive units of Z[ 2] are
those of the form (1 + 2)n ; there are infinitely many.

2.5 Supplementary Problems

√
Exercise 2.5.1 Show that R = Z[(1 + −7)/2] is Euclidean.

Solution. Given α, β ∈ R, we want to find γ, δ ∈ R such that α = βγ + δ,

with N (δ) < N (β). This is equivalent to showing that we can find a γ with
N (α/β − γ) < 1. √ √
Now, α/β = x + y −7 with x, y ∈ Q. Let γ = (u + v −7)/2 with
u, v ∈ Z and u ≡ v (mod 2). We want
! ! √ "" E
√ u + v −7 u F2 E v F2
N x + y −7 − = x− +7 y− <1
2 2 2
or, equivalently,
(2x − u)2 + 7(2y − v)2 < 4.
First consider 2y. Choose for v either [2y] or [2y] + 1, so that 2y − v ≤ 1/2.
Now choose for u either [2x] or [2x] + 1, whichever has the same parity as
v. Then 2x − u ≤ 1. Then

(2x − u)2 + 7(2y − v)2 ≤ 1 + 7

4 = 11
4 < 4.
√
We have found a γ which works, and proved that Z[(1 + −7)/2] is Eu-
clidean.
√
Exercise 2.5.2 Show that Z[(1 + −11)/2] is Euclidean.
190 CHAPTER 2. EUCLIDEAN RINGS

Solution. Proceed as in Exercise 2.5.1. Given √ α, β, we wish to find γ

such that
√ N (α/β − γ) < 1. Let α/β = x + y −11, x, y ∈ Q, and γ =
(u + v −11)/2 with u, v ∈ Z and u ≡ v (mod 2). We want
! ! √ ""
√ u + v −11
N x + y −11 − < 1,
2
or
(2x − u)2 + 11(2y − v)2 < 4.
As in the previous exercise, choose v first to be the integer which is closest
to 2y, and then choose u to be the integer closest to 2x which also has the
same parity as v. Then (2x − u) ≤ 1 and (2y − v) ≤ 1/2, so

(2x − u)2 + 11(2y − v)2 ≤ 1 + 11

4 = 15
4 < 4.
√
Therefore Z[(1 + −11)/2] is Euclidean.

Exercise 2.5.3 Find all integer solutions to the equation x2 + 11 = y 3 .

√
Solution. In the ring Z[(1 + −11)/2], we can factor the equation as
√ √
(x − −11)(x + −11) = y 3 .
√ √
Now, suppose that δ | (x − −11) √ and δ | (x + −11) (which implies
that δ | y). Then
√ δ | 2x and δ | 2 −11 which means that δ | 2 because
otherwise, δ | −11, meaning that 11 | x and 11 | y, which we can see is
not true by considering congruences mod 112 . Then δ = 1 or 2, since 2 has
no factorization in this ring. We will consider these cases separately.
Case 1. δ = 1.
Then the two factors of y 3 are coprime and we can write
! √ "3
√ a + b −11
(x + −11) = ε ,
2
√
where a, b ∈ Z and a ≡ b (mod 2). Since the units of Z[(1 + −11)/2] are
±1, which are cubes, then we can bring the unit inside the brackets and
rewrite the above without ε. We have
√ √ √ √
8(x + −11) = (a + b −11)3 = a3 + 3ab2 −11 − 33ab2 − 11b3 −11

and so, comparing real and imaginary parts, we get

8x = a3 − 33ab2 = a(a2 − 33b2 ),

8 = 3a2 b − 11b3 = b(3a2 − 11b2 ).

This implies that b | 8 and so we have 8 possibilities: b = ±1, ±2, ±4, ±8.
Substituting these back into the equations to find a, x, and y, and remem-
bering that a ≡ b (mod 2) and that a, x, y ∈ Z will give all solutions to the
equation.
2.5. SUPPLEMENTARY PROBLEMS 191

Case 2. δ = 2.
If δ = 2, then y is even and x is odd. We can write y = 2y1 , which gives
the equation ! √ "! √ "
x + −11 x − −11
= 2y13 .
2 2
Since 2 divides the right-hand side of this equation, it must divide the
left-hand side, so 1! √ "
1 x + −11
1
2 1
2
or 1! √ "
1 x − −11
1
2 1 .
2
However, since x is odd, 2 divides neither of the factors above. We conclude
that δ != 2, and thus we found all the solutions to the equation in our
discussion of Case 1.
√
Exercise 2.5.4 Prove that Z[ 3] is Euclidean.
√ √
Solution. Given α, β ∈ Z[ 3] we want to find γ, δ ∈ Z[ 3] such that
α = βγ + δ, with N (δ) < N (β). Put √ another way, we want to show
√ that
N (α/β − γ) < 1. Let α/β = x + y 3, x, y ∈ Q. Let γ = u + v 3, with
u, v ∈ Z.
Now, N (α/β − γ) = |(x − u)2 − 3(y − v)2 |. This will be maximized when
(x − u) is small and (y − v) is large. Choose for u and v the closest integers
to x and y, respectively. Then the minimum value for (x − u) is 0, while
the maximum value for (y − v) is 1/2. Then N (α/β − γ) ≤ |− 3/4| < 1.
The conclusion follows.
√
Exercise 2.5.5 Prove that Z[ 6] is Euclidean.
√
Solution. Assume √ that Z[√ 6] is not Euclidean. This means that
√ there √ is
at least one x + y 6 ∈ Q( 6) such that there is no γ = u + v 6 ∈ Z[ 6]
such that |(x − u)2 − 6(y − v)2 | < 1. Without loss, we can suppose that
0 ≤ x ≤ 1/2, and 0 ≤ y ≤ 1/2. We assert that there exist such a pair (x, y)
such that
(x − u)2 ≥ 1 + 6(y − v)2 ,
or
6(y − v)2 ≥ 1 + (x − u)2 ,
for every u, v ∈ Z. In particular, we will use the following inequalities:

either (a) x2 ≥ 1 + 6y 2 or (b) 6y 2 ≥ 1 + x2 , (2.1)

2 2 2 2
either (a) (1 − x) ≥ 1 + 6y or (b) 6y ≥ 1 + (1 − x) , (2.2)
either (a) (1 + x)2 ≥ 1 + 6y 2 or (b) 2
6y ≥ 1 + (1 + x) .2
(2.3)
192 CHAPTER 2. EUCLIDEAN RINGS

If x = y = 0, then both first inequalities fail, so we can rule out this case.
Next, we look at the first two inequalities on the left. Since x2 , (1 − x)2 ≤ 1
and 1+6y 2 ≥ 1 and x, y are not both 0, these two inequalities fail so (2.1 (b))
and (2.2 (b)) must be true. Now consider (2.3 (a)). If (1 + x)2 ≥ 1 + 6y 2
and 6y 2 ≥ 1 + (1 − x)2 as we just showed, then
(1 + x)2 ≥ 1 + 6y 2 ≥ 2 + (1 − x)2
which implies that 4x ≥ 2 and since x ≤ 1/2, we conclude that x = 1/2.
Substituting this into the previous inequalities, we get that
9
4 ≥ 1 + 6y 2 ≥ 94 ,
so 6y 2 = 54 . Let y = r/s with gcd(r, s) = 1. We now have that 24r2 = 5s2 .
Since r ! s, then r2 | 5, so r = 1. But then 24 = 5s2 , a contradiction.
Therefore, (2.3 (b)) is true, which implies that
6y 2 ≥ 1 + (1 + x)2 ≥ 2.
However, since y ≤ 1/2, 6y 2 ≥ 2 implies that√6 ≥ 8, a contradiction. Then
neither (2.3 (a)) nor (2.3 (b)) are true, so Z[ 6] must be Euclidean.
√
Exercise 2.5.6 Show that Z[(1 + −19)/2] is not Euclidean for the norm map.

Solution. If a ring R is Euclidean, then given any α, β ∈ R we can

find δ, γ such that α = βγ + δ with δ = 0 or N (δ) < N (β). Another
way of describing this condition is to say that given any β ∈ R, we can
find a representative for each nonzero residue class of R/(β) such that the
representative has norm √ less than the norm of β. We will try to find an
element of R = Z[(1 + −19)/2] for which this is not true.
Consider β = 2. N (2) = 4. We want to find all other elements of R
with norm strictly less than 4.
! √ "
a + b −19 a2 + 19b2
N = < 4,
2 4
⇒ a2 + 19b2 < 16.
First note that if b > 0, there are no solutions to this inequality. For b = 0,
we can have a = 0, ±2, since a ≡ b (mod 2). Thus, there are just three
elements with norm less than 4. However, there are more than three√ residue
classes of R/(2) (check this!). Therefore, the ring R = Z[(1 + −19)/2] is
non-Euclidean with respect to the norm map.
√
Exercise 2.5.7 Prove that Z[ −10] is not a unique factorization domain.
√ √
Solution. Consider the elements 2 + −10, 2 − −10, 2, 7. Show that they
are all irreducible and are not associates. Then note that
√ √
(2 + −10)(2 − −10) = 14,
2 · 7 = 14.
2.5. SUPPLEMENTARY PROBLEMS 193

√
Exercise 2.5.8 Show that there are only finitely many rings Z[ d] with d ≡ 2
or 3 (mod 4) which are norm Euclidean.
√ √
Solution. If Z[ d]√is Euclidean for the norm map, then for any δ ∈ Q( d),
we can find α ∈ Z[ d] such that

|N (δ − α)| < 1.
√ √
Write δ = r + s d, α = a + b d, a, b ∈ Z, r, s ∈ Q. Then

|(r − a)2 − d(s − b)2 | < 1.

In particular, take r = 0, s = t/d where t is an integer to be chosen later.

Then 1 ! "2 1
1 t 11
1 2
1a − d b − 1<1
1 d 1
so that |(bd − t)2 − da2 | < d. Since (bd − t)2 − da2 ≡ t2 (mod d), there are
integers x and z such that

z 2 − dx2 ≡ t2 (mod d),

with |z 2 − dx2 | < d.

In case d ≡ 3 (mod 4), we choose an odd integer t such that

5d < t2 < 6d,

which we can do if d is sufficiently large. Then z 2 − dx2 = t2 − 5d or t2 − 6d.

Then one of the equations

z 2 − t2 = d(x2 − 5)

or
z 2 − t2 = d(x2 − 6)
is true. We consider this modulo 8. Then t2 ≡ 1 (mod 8) since t is odd.
Also, x2 , z 2 ≡ 0, 1, or 4 (mod 8) and d ≡ 3 or 7 (mod 8). We are easily led
to t2 − z 2 ≡ 0, 1, or 5 (mod 8). This means

d(x2 − 5) ≡ 5, 4, or 1 (mod 8)

or
d(x2 − 6) ≡ 6, 5, 2, or 1 (mod 8).
All of these congruences are impossible. In case d ≡ 2 (mod 4), we choose
t odd satisfying 2d < t2 < 3d and proceed as above.
(The case d ≡ 1 (mod 4) is more difficult and has been handled by
Heilbronn who was the first to show that there are only finitely many real
quadratic fields which are norm-Euclidean.)
A more general and analogous result for imaginary quadratic fields will
be proved in Exercise 4.5.21 in Chapter 4.
194 CHAPTER 2. EUCLIDEAN RINGS

Exercise 2.5.9 Find all integer solutions of y 2 = x3 + 1.

Solution. We will determine all integer solutions of y 2 − 1 = x3 . From

(y − 1)(y + 1) = x3 , we see that if (y − 1, y + 1) = 1, then y − 1 = u3 ,
y + 1 = v 3 (say). Thus,

2 = v 3 − u3 = (v − u)(v 2 + vu + u2 )

from which we deduce that

v − u = ±1, v 2 + vu + u2 = ±2

or
v − u = ±2, v 2 + vu + u2 = ±1.
This gives rise to four cases. The only case that leads to a solution is
v − u = 2 and v 2 + vu + u2 = 1. This yields the solution (x, y) = (−1, 0).
Now suppose (y − 1, y + 1) = 2. This gives rise to two cases

y − 1 = 2u3 , y + 1 = 4v 3 and y − 1 = 4u3 , y + 1 = 2v 3 .

In the first case, we are led to u3 + 1 = 2v 3 and in the second case, we

get 2u3 + 1 = v 3 . As −1 is a cube, both equations are covered if we can
determine all integer solutions of

x3 + y 3 = 2z 3 .

We will use a “descent” argument to determine all coprime solutions.

To this end, we consider the ring of Eisenstein integers Z[ρ] where ρ2 +
ρ + 1 = 0. We recall a few facts about this ring. It is well-known that this
is a Euclidean domain for the norm map given by

N (a + bρ) = a2 + ab + b2 .

Its unit group is {±1, ±ρ, ±ρ2 }. It is also easily checked that 1, ρ, ρ2 repre-
sent all the distinct coprime residue classes modulo 2Z[ρ]. We see that the
cube of every coprime residue class is 1 (modulo 2). If u is a unit ≡ 1 (mod
2), then u = ±1. Now we claim that any coprime solution (x, y, z) of

x3 + y 3 = 2uz 3

in Z[ρ] satisfies N (xyz) = 1. Suppose not. Let (x, y, z) be such that N (xyz)
is minimal and ≥ 2. We may let

A = x + y, B = ρx + ρ2 y, C = ρ2 x + ρy

so that
ABC = 2uz 3 , A + B + C = 0.
2.5. SUPPLEMENTARY PROBLEMS 195

Let d = (A, B, C) so that the above equation becomes

ABC E z F3
= 2u .
d d d d
Now 2 is an irreducible element in Z[ρ] and A/d, B/d, C/d are mutually
coprime (as their sum is zero) so it can divide only one of them, say C/d
without any loss of generality. Thus, we may write

A/d = u1 α3 , B/d = u2 β 3 , C/d = −2u3 γ 3 ,

with u1 , u2 , u3 units. Also, αβγ != 0 for otherwise, z = 0 and x = ±y,

which are not coprime solutions. Hence,

u1 α3 + u2 β 3 = 2u3 γ 3 ,

and dividing by the unit u1 gives the equation

α3 + u$ β 3 = 2u4 γ 3

for some units u$ and u4 . Observe that (β, 2) = 1 for otherwise, 2|α and 2|γ
which implies that α, β, γ are not coprime, a contradiction. Reducing the
above equation mod 2 shows that u$ is a cube mod 2, and by our remark
above u$ must be ±1. Thus u$ is a cube and we have

α3 + β 3 = 2uγ 3 .

Notice that by our choice of (x, y, z)

N (xyz)3 ≤ N (αβγ)3 = N (ABC/d3 ) = N (z)3 /N (d)3

which means that N (xyd)3 ≤ 1. Thus, x, y, d are units. Hence, x3 = ±1

and y 3 = ±1 and z is also a unit. Thus, N (xyz) = 1 contrary to our choice.
This proves our claim.
Therefore, the only solution for u3 + 1 = 2v 3 is u3 = ±1. This leads
to the solutions (x, y) = (2, 3), (1, 0) for the equation y 2 − 1 = x3 . In the
other case of 2u3 + 1 = v 3 , we have v = ±1 which leads to (0, 1), (2, −3).
We get a final set of five integer solutions for y 2 − 1 = x3 .

Exercise 2.5.10 Let x1 , ..., xn be indeterminates. Evaluate the determinant of

the n × n matrix whose (i, j)-th entry is xj−1
i . (This is called the Vandermonde
determinant.)

Solution. Let V (x1 , ..., xn ) denote the value of the determinant. If we fix
x2 , ..., xn , we may view the determinant as a polynomial in x1 of degree
n − 1. Since the determinant is zero if x1 = xi for i ≥ 2, the roots of the
polynomial are x2 , ..., xn . It is easy to see that the leading coefficient is

(−1)n−1 V (x2 , ..., xn )

196 CHAPTER 2. EUCLIDEAN RINGS

so that the determinant is

n
6
(−1)n−1 V (x2 , ..., xn ) (x1 − xj ).
j=2

By induction, we see that

6
V (x1 , ..., xn ) = (−1)( 2 )
n
(xi − xj ).
j>i
Chapter 3

Algebraic Numbers and

Integers

3.1 Basic Concepts

Exercise 3.1.2 Show that if r ∈ Q is an algebraic integer, then r ∈ Z.

Solution. Let r = c/d, (c, d) = 1, be an algebraic integer. Then r is the

root of a monic polynomial in Z[x], say f (x) = xn + bn−1 xn−1 + · · · + b0 .
So
E c Fn E c Fn−1
f (r) = + bn−1 + · · · + b0 = 0
d d
⇔ cn + bn−1 cn−1 d + · · · + b0 dn = 0.

This implies that d | cn , which is true only when d = ±1. So r = ±c ∈ Z.

Exercise 3.1.3 Show that if 4 | (d + 1), then

√
−1 ± −d
2
is an algebraic integer.

Solution. Consider the monic polynomial

d+1
x2 + x + ∈ Z[x]
4
when 4 | d + 1. The roots of this polynomial, which by definition are
algebraic integers, are
N
√
−1 ± 1 − 4 d+1
4 −1 ± −d
x= = .
2 2

197
198 CHAPTER 3. ALGEBRAIC NUMBERS AND INTEGERS

√
Exercise 3.1.6 Find the minimal polynomial of n where n is a squarefree
integer.

Solution. If n = 1, the minimal √ polynomial is x − 1. If n != 1, then

x2 − n is irreducible and has n as a root. Thus, the minimal √ polynomial
is either linear or quadratic. If it is linear, we obtain that √n is rational, a
contradiction. Thus, x2 − n is the minimal polynomial of n when n != 1.
√
Exercise 3.1.7 Find the minimal polynomial of 2/3.
√ √
Solution. It is x2 − 2/9 since 2/3 is a root, and 2/3 is not rational.

3.2 Liouville’s Theorem and Generalizations

&∞ n
Exercise 3.2.4 Show that n=1 2−3 is transcendental.

Solution. Suppose that

*∞
1
α=
n=1
2 3n

is algebraic. We proceed as in Example 3.2.2 and consider the partial sum:

*k
1 pk
= ,
n=1
2 3n qk

with qk = 23 . As before,
k

1 1 11 ∞ 1
1 p k 1 1 * 1 11 S
1α − 1 = 1 1≤ .
1 qk 1 1 23n 1 23k+1
n=k+1

But since α is algebraic, by Roth’s theorem we have the inequality

S c(α)
≥ 2+ε .
qk3 qk

But again we can choose k to be as large as we want, and so for ε sufficiently

small, this inequality does not hold. Thus, α is transcendental.
&∞
Exercise 3.2.5 Show that, in fact, n=1 2−f (n) is transcendental when

f (n + 1)
lim > 2.
n→∞ f (n)

Solution. Suppose
∞
* 1
α=
n=1
2f (n)
3.3. ALGEBRAIC NUMBER FIELDS 199

is algebraic. Following the same argument as above, we get the inequalities

1 1
S 1 pk 11 c(α)
1
≥ 1α − 1 ≥ 2+ε ,
qk+1 qk qk

where qk = 2f (k) . Now, for k sufficiently large,

f (k + 1)
>2+δ ⇒ f (k + 1) > (2 + δ)f (k).
f (k)
So, for large k,

qk+1 2f (k+1) (1+δ)

= f (k) > 2(1+δ)f (k) = qk
qk 2

which implies that qk+1 > qk2+δ . By Roth’s theorem, we can deduce that

c(α) S
≤ ,
qk2+ε qk+1
c(α) S
⇒ ≤ ,
qk2+ε qk2+δ
S ε
⇒ qkδ ≤ q .
c(α) k

As qk → ∞, this implies δ ≤ ε, a contradiction for ε < δ/2 (say).

3.3 Algebraic Number Fields

Exercise 3.3.3 Let α be an algebraic number and let p(x) be its minimal poly-
nomial. Show that p(x) has no repeated roots.

Solution. Suppose α is a repeated root of p(x). Then we can write

p(x) = (x − α)2 g(x),

for some polynomial g(x) ∈ C[x], and

p$ (x) = 2(x − α)g(x) + (x − α)2 g $ (x).

So p$ (α) = 0 and from Theorem 3.1.4, p(x) | p$ (x). But deg(p$ ) < deg(p),
and we have a contradiction. If β is a repeated root of p(x), then by the
following exercise, β has the same minimal polynomial and repeating the
above argument with β leads to a contradiction. Thus p has no repeated
roots.

Exercise 3.3.4 Let α, β be algebraic numbers such that β is conjugate to α.

Show that β and α have the same minimal polynomial.
200 CHAPTER 3. ALGEBRAIC NUMBERS AND INTEGERS

Solution. Let p(x) be the minimal polynomial of α, and let q(x) be the
minimal polynomial of β. By the definition of conjugate roots, β is a
common root of p(x) and q(x).
Using the division algorithm, we can write p(x) = a(x)q(x) + r(x) for
some a(x), r(x) ∈ Q[x] and either r = 0 or deg(r) < deg(q). But

p(β) = a(β)q(β) + r(β) = 0

and q(β) = 0 so r(β) must also be 0. Since q is the minimal polynomial for
β, r = 0. Thus p(x) = a(x)q(x), but, by Theorem 3.1.4, p is irreducible,
and both p(x) and q(x) are monic, so p(x) = q(x).

Exercise 3.3.6 Let K = Q(θ) be of degree n over Q. Let ω1 , . . . , ωn be a basis

(j)
of K as a vector space over Q. Show that the matrix Ω = (ωi ) is invertible.

Solution.  
(2) (n)
ω1 ω1 ··· ω1
 (2) (n) 
 ω2 ω2 ··· ω2 
Ω=
 .. .. .. .
 . . . 
(2) (n)
ωn ωn ··· ωn
Since θ is an algebraic number of degree n, α1 = 1, α2 = θ, . . . , αn = θn−1
(j)
also forms a basis for K over Q. Let A = (αi ). Then,
1 1
1 1 1 ··· 1 11
1
1 θ
1 θ(2) ··· θ(n) 11
det A = 1 . .. .. 1
1 .. . . 11
1
1θn−1 θ(2)n−1 · · · θ(n)n−1 1

which is the Vandermonde determinant. So A is invertible. Further,

n
*
(j)
ωi = (bik αk )(j)
k=1
*n
(j)
= bik αk ,
k=1

where 1 ≤ i, j ≤ n, and bik ∈ Q.

Since the set {ω1 , . . . , ωn }, as well as the set {α1 , . . . , αn }, are linearly
independent sets, it follows that both the rows and columns of the matrix
B = (bik ) are linearly independent. Hence B is invertible and Ω = BA
and from elementary linear algebra, det Ω = det B det A != 0. Thus Ω is
invertible.

Exercise 3.3.7 Let α be an algebraic number. Show that there exists m ∈ Z

such that mα is an algebraic integer.
3.3. ALGEBRAIC NUMBER FIELDS 201

Solution. Let p(x) ∈ Q[x] be the minimal polynomial of α. So,

p(α) = αn + an−1 αn−1 + · · · + a1 α + a0 = 0.
Choose m ∈ Z so that ma0 , ma1 , . . . , man−1 are all integers. Now,
mn αn + mn an−1 αn−1 + · · · + mn a1 α + mn a0 = 0,
⇔ (mα) + man−1 (mα)
n n−1
+ ··· + m n−1
a1 (mα) + m a0n
= 0.
Let g(x) = xn + man−1 xn−1 + · · · + mn−1 a1 x + mn a0 , then g(x) is a monic
polynomial in Z[x] and g(mα) = 0. Thus mα is an algebraic integer.

Exercise 3.3.8 Show that Z[x] is not (a) Euclidean or (b) a PID.

Solution. (a) Consider the elements 2, x ∈ Z[x]. Clearly there is no way

to write x = a(x)2 + r(x) where both the conditions (i) a(x), r(x) ∈ Z[x]
and (ii) deg(r) < deg(2) = 0 or r = 0 are satisfied.
(b) Again consider the two polynomials 2, x ∈ Z[x]. Clearly (x) !⊆ (2)
and (2) !⊆ (x). Now, if (x, 2) = (α) for some α ∈ Z[x], then α | 2 and α | x.
But if α | 2, then α ∈ Z, so α = ±1 or α = ±2. However, ±1 ∈ / (x, 2)
and ±2 ! x. So the ideal generated by x and 2 is not generated by a single
element in Z[x]. Z[x] is not a PID.

Exercise 3.3.11 Let f (x) = xn + an−1 xn−1 + · · · + a1 x + a0 , and assume that

for p prime p | ai for 0 ≤ i < k and p2 ! a0 . Show that f (x) has an irreducible
factor of degree at least k. (The case k = n is referred to as Eisenstein’s criterion
for irreducibility.)

Solution. We will prove this by induction on n, the degree of f (x). The

case when n = 1 is trivial, so let us assume that the above statement is
true for any polynomial of degree less than n.
If f (x) is irreducible, there is nothing to prove, so assume that f (x) is
not irreducible. Then we can write
f (x) = g(x)h(x)
= (b0 + b1 x + · · · + br xr )(c0 + c1 x + · · · + ct xt ).
Since p | a0 and p2 ! a0 , and a0 = b0 c0 , we deduce p | b0 or p | c0 but not
both.
Suppose, without loss of generality, that p | b0 . We next consider a1 =
b0 c1 + b1 c0 . Since p | a1 and p | b0 , but p ! c0 , then p | b1 . Continuing in
this fashion, we get that p | bi for 0 ≤ i < k. If r < k, then we can factor
out a p from each of the coefficients in g(x), but this is absurd since then
p would divide every coefficient in f (x), and f is monic. Therefore k ≤ r,
p | bi , 0 ≤ i < k, p2 ! b0 . Also br ct = 1 implies that g(x) or −g(x) is monic.
In any event, we have another polynomial which satisfies the conditions set
out above but which has degree less than n. Thus, by induction, g(x) has
an irreducible factor of degree greater than or equal to k, and this factor is
also an irreducible factor of f (x).
202 CHAPTER 3. ALGEBRAIC NUMBERS AND INTEGERS

Exercise 3.3.12 Show that f (x) = x5 + x4 + 3x3 + 9x2 + 3 is irreducible over

Q.

Solution. By applying Exercise 3.3.11 to f (x) with p = 3, we deduce

that if f (x) is not irreducible, then we can factor it into the product of
a polynomial of degree 4 and a polynomial of degree 1, and so f (x) has
a rational root. However, we showed in Exercise 3.1.2 that if r ∈ Q is
an algebraic integer, then r ∈ Z. Thus, f (x) must have an integral root,
and this root must divide the constant term which is 3. The only choices
are then ±1, ±3, and it is easy to check that these are not roots of the
polynomial in question.
We conclude that f (x) is irreducible since it has no rational root.

3.4 Supplementary Problems

Exercise 3.4.1 Show that
,∞
1
a n!
n=0
is transcendental for a ∈ Z, a ≥ 2.

Solution. Suppose it is algebraic, and call the sum α. Look at the partial
sum
* k
1 pk
αk = n!
= ,
n=0
a qk
with qk = ak! . Then
1 1
1 *
∞
1 11
1
|α − αk | = 1 1
1 an! 1
n=k+1
1
≤ M,
a(k+1)!
where ! "2
1 1
M =1+ + + ··· ,
ak+2 ak+2
an infinite geometric series with a finite sum. Thus,
1 1
1 * ∞
1 11 M
1
1 n! 1 ≤ k+1 .
1 a 1 qk
n=k+1

If α is algebraic of degree n, then Liouville’s theorem tells us that we can

find a constant c(α) such that
1 1
M 1 1
≥ 1α − pk 1 ≥ c(α) .
k+1
qk 1 qk 1 qkn
However, we can choose k as large as we wish to obtain a contradiction.
3.4. SUPPLEMENTARY PROBLEMS 203

Exercise 3.4.2 Show that

,∞
1
a 3n
n=1

is transcendental for a ∈ Z, a ≥ 2.

Solution. Suppose that

*∞
1
α=
n=1
a3n

is algebraic. Consider the partial sum

*k
1 pk
αk = 3n
= ,
n=1
a qk

with qk = a3 . We have
k

1 1 11 ∞ 1
1 1 * 1 11
1α − pk 1 = 11 n 1 ≤
S
,
1 qk 1 1 3
a 1 a k+13
n=k+1

where S = 1 + 1/a + 1/a2 + · · · . Then by Roth’s theorem,

S c(α, ε)
≥ 2+ε .
qk3 qk

But, we can choose k to be as large as we want to produce a contradiction.

Exercise 3.4.3 Show that

∞
, 1
n=1
af (n)
is transcendental when
f (n + 1)
lim > 2.
n→∞ f (n)

Solution. Suppose that

∞
* 1
α=
n=1
af (n)

is algebraic. Following the same argument as in the previous exercise, we

get 1 1
S 1 pk 11 c(α)
1
≥ 1α − 1 ≥ 2+ε ,
qk+1 qk qk
where qk = af (k) . For k sufficiently large,

f (k + 1)
> 2 + δ,
f (k)
204 CHAPTER 3. ALGEBRAIC NUMBERS AND INTEGERS

and so
qk+1 af (k+1) (1+δ)
= f (k) > a(1+δ)f (k) = qk .
qk a
This implies that qk+1 > qk2+δ . By Roth’s theorem, we can deduce that
c(α) S c(α) S
2+ε ≤ q ⇒ 2+ε ≤ 2+δ ,
qk k+1 qk qk
S ε
⇒ qkδ ≤ q .
c(α) k
As qk → ∞, we find δ ≤ ε, a contradiction for sufficiently small ε.

Exercise 3.4.4 Prove that f (x) = x6 + 7x5 − 12x3 + 6x + 2 is irreducible over

Q.

Solution. By Exercise 3.3.11, since 2 | ai for 0 ≤ i < 5 then f (x) has

a factor of degree at least 5. This means that the polynomial is either
irreducible or it has a rational root. We showed earlier that if a polynomial
in Z[x] has a rational root, then the root is actually an integer. We also
know that any roots of a polynomial will divide its constant term, which in
this case is 2. It suffices to check that ±1, ±2, are not roots to deduce that
f (x) is irreducible.

Exercise 3.4.5 Using Thue’s theorem, show that f (x, y) = x6 +7x5 y −12x3 y 3 +
6xy 5 + 8y 6 = m has only a finite number of solutions for m ∈ Z∗ .

Solution. Use the previous exercise to prove that the polynomial is irre-
ducible. The result follows from Thue’s theorem and Example 3.2.3.

Exercise 3.4.6 Let ζm be a primitive mth root of unity. Show that

# i j
(ζm − ζm ) = (−1)m−1 mm .
0≤i,j≤m−1
i&=j

Solution. Since
m−1
6
xm − 1 = (x − ζm
i
),
i=0
we see that the constant term is
m−1
6
(−1)m i
ζm = −1.
i=0

Differentiating x − 1 above via the product rule, and setting x = ζm

m j
, we
see that
m−1
6
mζmj(m−1)
= (ζm
j i
− ζm ).
i=0
i*=j

Taking the product over j gives the result.

3.4. SUPPLEMENTARY PROBLEMS 205

Exercise 3.4.7 Let # i

φm (x) = (x − ζm )
1≤i≤m
(i,m)=1

denote the mth cyclotomic polynomial. Prove that

#
xm − 1 = φd (x).
d|m

Solution. Every mth root of unity is a primitive dth root of unity for some
d | m. Conversely, every dth root of unity is also an mth root of unity for
d | m. The result is now immediate.

Exercise 3.4.8 Show that φm (x) ∈ Z[x].

Solution. We induct on m. For m = 1, this is clear. Suppose we have

proved it true for φr (x) with r < m. Then setting
6
v(x) = φd (x),
d|m
d<m

we have by induction v(x) ∈ Z[x]. Since v(x) is monic, and v(x) | (xm − 1),
we find by long division that (xm − 1)/v(x) = φm (x) ∈ Z[x].

Exercise 3.4.9 Show that φm (x) is irreducible in Q[x] for every m ≥ 1.

Solution. Let f (x) be the minimal polynomial of ζm and suppose φm (x) =

f (x)g(x) with f (x), g(x) ∈ Q[x]. By Gauss’ lemma (see Theorem 2.1.9) we
may suppose that f (x), g(x) ∈ Z[x]. Let p be coprime to m. Then ζm p
is
again a primitive mth root of unity. Thus

f (ζm
p
)g(ζm
p
) = 0.

Suppose f (ζmp
) != 0. Then g(ζm p
) = 0. Since g(xp ) ≡ g(x)p (mod p) we
deduce that g(x) and f (x) have a common root in Fp , a contradiction since
xm − 1 has no multiple roots in Fp . Thus, f (ζm p
) = 0 for any (p, m) = 1.
It follows that f (ζm ) = 0 for any (i, m) = 1. Therefore deg(f ) = ϕ(m) =
i

deg(φm ).

Exercise 3.4.10 Let I be a subset of the positive integers ≤ m which are coprime
to m. Set # i
f (x) = (x − ζm ).
i∈I

Suppose that f (ζm ) = 0 andf (ζm

p
) #= 0 for some prime p. Show that p | m. (This
observation gives an alternative proof for the irreducibility of φm (x).)
206 CHAPTER 3. ALGEBRAIC NUMBERS AND INTEGERS

Solution. Let K = Q(ζm ). Then f (ζmp

) divides
6
(ζmp
− ζmi
)
0≤i≤m−1

in the ring OK . Hence NK/Q (f (ζm

p
)) divides m by Exercise 3.4.6 above.
Since
f (x)p − f (xp ) ∈ pZ[x],
we see upon setting x = ζm that p | f (ζm
p
) and hence p | m, as desired.

Exercise 3.4.11 Consider the equation x3 + 3x2 y + xy 2 + y 3 = m. Using Thue’s

theorem, deduce that there are only finitely many integral solutions to this equa-
tion.

Solution. To use Thue’s theorem, we must show that f (x, y) = x3 +

3x2 + xy 2 + y 3 is irreducible. If this polynomial is reducible, then so is
the polynomial f (x, 1) = x3 + 3x2 + x + 1. However, since f (x, 1) has
degree 3, then if it reduces it will have a factor of degree 1, a rational root.
We have already shown that all rational roots of a monic polynomial in
Z[x] are actually integers, and all roots must divide the constant term of
a polynomial. The only possibilities for such a root are x = ±1. A quick
calculation shows that neither of these two are in fact a root of f (x, 1), and
so f (x, 1) is irreducible, implying that f (x, y) is irreducible. We can now
apply the results of Example 3.2.3.

Exercise 3.4.12 Assume that n is an odd integer, n ≥ 3. Show that xn +y n = m

has only finitely many integral solutions.

Solution. If (x0 , y0 ) is a solution to xn + y n = m, then (x0 + y0 ) | m, since

xn + y n = (x + y)(xn−1 − xn−2y + · · · + y n−1 ).
Suppose |x| ≥ m. Then the distance between x and the nearest nth power
will be greater than m, and x cannot satisfy the above equation. We then
have a bound on the size of x along with the constraint that x+y | m. There
can only be a finite number of pairs which satisfy these two constraints.

Exercise 3.4.13 Let ζm denote a primitive mth root of unity. Show that Q(ζm )
is normal over Q.

Solution. ζm is a root of the mth cyclotomic polynomial, which we

have shown to be irreducible. Thus, the conjugate fields are Q(ζm
j
) where
(j, m) = 1 and these are identical with Q(ζm ).

Exercise 3.4.14 Let a be squarefree and greater than 1, and let p be prime.
Show that the normal closure of Q(a1/p ) is Q(a1/p , ζp ).

Solution. The polynomial xp − a is irreducible (by Eisenstein’s criterion).

The conjugates of a1/p are ζpj a1/p . If K is the normal closure of Q(a1/p ), it
must contain all the pth roots of unity. The result is now immediate.
Chapter 4

Integral Bases

4.1 The Norm and the Trace

Exercise 4.1.2 Let K = Q(i). Show that i ∈ OK and verify that TrK (i) and
NK (i) are integers.

Solution. We know that i is a root of the irreducible polynomial x2 + 1,

and so its conjugates are i, −i.
Thus, TrK (i) = i − i = 0 ∈ Z and NK (i) = i(−i) = 1 ∈ Z.
√
Exercise 4.1.3 Determine the algebraic integers of K = Q( −5).
√
Solution. We first note that √ 1, −5 form a Q-basis for K. Thus any
α ∈ K looks like α = r1 + r2 −5 with r1 , r2 √ ∈ Q. Since [K : √
Q] = 2, we
can deduce that the conjugates of α are r1 + r2 −5 and r1 − r2 −5. Then
TrK (α) = 2r1 and
√ √
NK (α) = (r1 + r2 −5)(r1 − r2 −5)
= r12 + 5r22 .

By Lemma 4.1.1, if α ∈ OK , then the trace and norm are integers. Also, α
is a root of the monic polynomial x2 − 2r1 x + r12 + 5r22 which is in Z[x]
√ when
the trace and norm are integers. We conclude that for α = r1 + r2 −5 to
be in OK , it is necessary and sufficient that 2r1 and r12 + 5r22 be integers.
This implies that r1 has a denominator at most 2, which forces the same for
r2 . Then by setting r1 = g1 /2 and r2 = g2 /2 we must have (g12 +5g22 )/4 ∈ Z
or, equivalently, g12 + 5g22 ≡ 0 (mod 4). Thus, as all squares are congruent
to 0 or 1 (mod 4), we conclude that g1 and g2 are√themselves even, and
thus r1 , r2 ∈ Z. We conclude then that OK = Z + Z −5.

Exercise 4.1.5 Show that the definition of nondegeneracy above is independent

of the choice of basis.

207
208 CHAPTER 4. INTEGRAL BASES

Solution. If f1 , . . . , fn is another basis and A = (B(fi , fj )), then

A = P T BP,

where P is the change of basis matrix from e1 , . . . , en to f1 , . . . , fn . Since

P is nonsingular, det A != 0 if and only if det B != 0.

4.2 Existence of an Integral Basis

Exercise 4.2.1 Show that ∃ω1∗ , ω2∗ , . . . , ωn∗ ∈ K such that

OK ⊆ Zω1∗ + Zω2∗ + · · · + Zωn∗ .

Solution. Let ω1 , ω2 , . . . , ωn be a Q-basis for K, and recall from Ex-

ercise 3.3.7 that for any α ∈ K there is a nonzero integer m such that
mα ∈ OK . Thus we can assume that ω1 , ω2 , . . . , ωn are in OK . Now, as
the bilinear pairing B(x, y) defined previously was nondegenerate, we can
find a5dual basis ω1∗ , ω2∗ , . . . , ωn∗ satisfying B(ωi , ωj∗ ) = δij . If we write
ωj∗ = ckj ωk we have

δij = TrK (ωi ωj ∗ )

*
= TrK (ωi ckj ωk )
*
= ckj TrK (ωi ωk ).

(j)
If we introduce now the matrices, C = (cij ), Ω = (ωi ), then the above
becomes
In = ΩΩT C ⇒ C −1 = ΩΩT .
We conclude that C is nonsingular and that ω1 ∗ , ω2 ∗ , . . . , ωn ∗ forms a Q-
basis for K.
Let α be an arbitrary element of OK . We write
n
*
α= aj ωj∗ with aj ∈ Q
j=1

so
n
*
αωi = aj ωi ωj∗ ∀i,
j=1

and *
TrK (αωi ) = aj TrK (ωi ωj∗ ) = ai ∀i.
But αωi ∈ OK implies the left-hand side above is in Z, and thus ai ∈ Z for
all i. It follows then that OK ⊆ Zω1∗ + Zω2∗ + · · · + Zωn∗ .

Exercise 4.2.3 Show that OK has an integral basis.

4.2. EXISTENCE OF AN INTEGRAL BASIS 209

Solution. We apply the results of Theorem 4.2.2 with M = Zω1∗ + Zω2∗ +

· · · + Zωn∗ and N = OK . It follows directly from the theorem that there
exist ω1 , ω2 , . . . , ωn ∈ OK such that OK = Zω1 + Zω2 + · · · + Zωn .

Exercise 4.2.4 Show that det(Tr(ωi ωj )) is independent of the choice of integral

basis.

Solution. Let ω1 , ω2 , . . . , ωn and θ1 , θ2 , . . . , θn be two distinct integral

bases for an algebraic number field K. We can write
n
*
ωi = cij θj ,
j=1
*n
θi = dij ωj ,
j=1

for all i, where cij and dij are all integers. Then (cij ) and (cij )−1 both have
entries in Z. So det(cij ), det(cij )−1 ∈ Z, meaning that det(cij ) = ±1.
Then
// 0/ 00
* *
Tr(ωi ωj ) = Tr cil θl cjm θm
l m
*
= cil cjm Tr (θl θm ) .
l,m

(j) (j)
Now if we define Ω = (ωi ), C = (cij ), Θ = (θi ), then we can write
the above as the matrix equation ΩT Ω = C(ΘT Θ)C T from which it follows
that the determinants of Ω and Θ are equal, up to sign. Hence, det(ΘT Θ) =
det(ΩT Ω).

Exercise 4.2.5 Show that the discriminant is well-defined. In other words, show
that given ω1 , ω2 , . . . , ωn and θ1 , θ2 , . . . , θn , two integral bases for K, we get the
same discriminant for K.

Solution. Just as above, we have ΩT Ω = C(ΘT Θ)C T for some matrix C

with determinant ±1. Then dK = (det Ω)2 = (det Θ)2 (det C)2 = (det Θ)2 .
This proves that the discriminant does not depend upon the choice of in-
tegral basis.

Exercise 4.2.6 Show that

#! "2
dK/Q (1, a, . . . , an−1 ) = σi (a) − σj (a) .
i>j

We denote dK/Q (1, a, . . . , an−1 ) by dK/Q (a).

210 CHAPTER 4. INTEGRAL BASES

Solution. First#we note$ that σi (a) takes a to its ith conjugate, a(i) . Define
the matrix Ω = σi (aj ) . Then it is easy to see that
 
1 a ··· an−1
. .. .. 
Ω =  .. . . ,
1 a(n) ··· a(n)n−1
which is a Vandermonde matrix, and so
6# $ 6# $
det Ω = a(i) − a(j) = σi (a) − σj (a) .
i>j i>j

It follows that
< =2
dK/Q (a) = det(σi (aj ))
6# $2
= σi (a) − σj (a) .
i<j

&n
Exercise 4.2.7 Suppose that ui = j=1 aij vj with aij ∈ Q, vj ∈ K. Show that
! "2
dK/Q (u1 , u2 , . . . , un ) = det(aij ) dK/Q (v1 , v2 , . . . , vn ).
< =2
Solution. By definition, dK/Q (u1 , u2 , . . . , un ) = det(σi (uj )) .
/ n 0 n
* *
σi (uj ) = σi ajk vk = ajk σi (vk ).
k=1 k=1

If we define the matrices U = (σi (uj )), A = (aij ), V = (σi (vj )), then it is
clear that U = V AT and so (det U )2 = (det V AT )2 , and we get the desired
result:
dK/Q (u1 , u2 , . . . , un ) = (det(aij ))2 dK/Q (v1 , v2 , . . . , vn ).

Exercise 4.2.8 Let a1 , a2 , . . . , an ∈ OK be linearly independent over Q. Let

N = Za1 + Za2 + · · · + Zan and m = [OK : N ]. Prove that
dK/Q (a1 , a2 , . . . , an ) = m2 dK .

Solution. Let α1 , α2 , . . . , αn be an integral basis of O5 K . Theorem 4.2.2

says that N has a basis β1 , β2 , . . . , βn such that βi = j≥i pij αj . Then
from Exercise 4.2.7,
# $2
dK/Q (β1 , . . . , βn ) = det(pij ) dK/Q (α1 , . . . , αn )
= m2 dK .
Reasoning as in Exercise 4.2.5, we deduce
dK/Q (β1 , . . . , βn ) = dK/Q (a1 , . . . , an ),
which proves the result.
4.3. EXAMPLES 211

4.3 Examples
Exercise 4.3.2 Let m ∈ Z, α ∈ OK . Prove that dK/Q (α + m) = dK/Q (α).
2
Solution. By definition, dK/Q (α) = i<j (α(i) − α(j) )2 . We note that the
ith conjugate of α + m is simply α(i) + m, and so
6# $2
dK/Q (α + m) = α(i) + m − (α(j) + m)
i<j
6# $2
= α(i) − α(j)
i<j
= dK/Q (α),

as desired.

Exercise 4.3.3 Let α be an algebraic integer, and let f (x) be the minimal poly-
n /
nomial of α. If f has degree n, show that dK/Q (α) = (−1)( 2 ) n (i)
i=1 f (α ).
(

Solution. Let f (x) be the minimal

2n polynomial of α. Then if α(1) , . . . , α(n)
(k)
are the conjugates of α, f (x) = k=1 (x − α ). Then
n
* f (x)
f $ (x) =
k=1
(x − α(k) )

and 6
f $ (α(i) ) = (α(i) − α(k) ).
k*=i

Therefore
n
6 n 6
6
(i)
# $
f (α )
$
= α(i) − α(k)
i=1 i=1 k*=i
6J # $2 K
= − α(i) − α(k)
i<k

(−1)( 2 ) dK/Q (α).

n
=
√
Exercise 4.3.5 √ If D ≡ 1 (mod 4), show that every integer of Q( D) can be
written as (a + b D)/2 where a ≡ b (mod 2).
√
Solution. By Example 4.3.4, an integral basis is given by 1, (1 + D)/2.
Thus every integer is of the form
/ √ 0 √ √
1+ D (2c + d) + d D a+b D
c+d = = .
2 2 2

Then we see that a = 2c + d, b = d satisfies a ≡ b (mod 2).

212 CHAPTER 4. INTEGRAL BASES

Conversely, if a ≡ b (mod 2), writing d = b and a = 2c + d for some c,

we find √ / √ 0
a+b D 1+ D
=c+d
2 2
√
is an integer of Q( D).

Exercise 4.3.7 Let ζ be any primitive pth root of unity, and K = Q(ζ). Show
that 1, ζ, . . . , ζ p−2 form an integral basis of K.

Solution. The minimal polynomial of ζ is the pth cyclotomic polynomial,

xp − 1
Φ(x) = = 1 + x2 + · · · + xp−1 .
x−1
We want to show that this is irreducible. Consider instead the polynomial
F (x) = Φ(x + 1). Clearly F will be irreducible over Q if and only if Φ is.
! " ! "
(x + 1)p − 1 p p−3 p
F (x) = =x p−1
+ px p−2
+ x + ··· + x + p.
x 2 p−2
This is Eisensteinian with respect to p and so F (x) (and thus Φ(x)) is
irreducible. The conjugates of ζ are ζ, ζ 2 , . . . , ζ p−1 . We can deduce that
[K : Q] = p − 1.
Now,
pxp−1 (x − 1) − (xp − 1)
Φ$ (x) =
(x − 1)2
pxp−1 − (1 + x + · · · + xp−1 )
= ,
x−1
and so
pζ −k
Φ$ (ζ k ) = .
ζk − 1
Using Exercise 4.3.3, we can compute
p−1
6 pζ −k
dK/Q (ζ) = ±
ζk − 1
k=1
1
= ±pp−1 2p−1
k=1 (ζ − 1)
k

1
= ±pp−1 2p−1
k=1 (1 − ζk)
= ±pp−2 ,
2p−1
since k=1 (1 − ζ k ) = Φ(1) = p. We know that dK/Q (ζ) = pp−2 = m2 dK
and also that p ! m because F , the minimal polynomial for ζ − 1, is p-
Eisensteinian, and dK/Q (ζ − 1) = dK/Q (ζ). Then m must be 1, meaning
that OK = Z[1, ζ, . . . , ζ p−2 ].
4.4. IDEALS IN OK 213

4.4 Ideals in OK
Exercise 4.4.1 Let a be a nonzero ideal of OK . Show that a ∩ Z #= {0}.

Solution. Let α be a nonzero algebraic integer in a satisfying the minimal

polynomial xr + ar−1 xr−1 + · · · + a0 = 0 with ai ∈ Z ∀i and a0 not zero.
Then a0 = −(αr + · · · + a1 α). The left-hand side of this equation is in Z,
while the right-hand side is in a.

Exercise 4.4.2 Show that a has an integral basis.

Solution. Let a be an ideal of OK , and let ω1 , ω2 , . . . , ωn be an integral

basis for OK . Note that for any ωi in OK , a0 ωi = −(αr + · · · + a1 α)ωi ∈ a.
Therefore a has finite index in OK and a ⊆ OK = Zω1 + Zω2 + · · · + Zωn
has maximal rank. Then since a is a submodule of OK , by Theorem 4.2.2
there exists an integral basis for a.

Exercise 4.4.3 Show that if a is a nonzero ideal in OK , then a has finite index
in OK .

Solution. Surely, if OK = Zω1 + Zω2 + · · · + Zωn , then by the preceding

two exercises we can pick a rational integer a such that

aOK = aZω1 + aZω2 + · · · + aZωn ⊂ a ⊂ OK .

But aOK obviously has index an in OK . Thus, the index of a in OK must

be finite.

Exercise 4.4.4 Show that every nonzero prime ideal in OK contains exactly one
integer prime.

Solution. If ℘ is a prime ideal of OK , then certainly it contains an integer,

from Exercise 4.4.1. By the definition of a prime ideal, if ab ∈ ℘, either
a ∈ ℘ or b ∈ ℘. So ℘ must contain some rational prime. Now, if ℘
contained two distinct rational primes p, q, say, then it would necessarily
contain their greatest common denominator which is 1. But this contradicts
the assumption of nontriviality. So every prime ideal of OK contains exactly
one integer prime.

Exercise 4.4.5 Let a be an integral ideal with basis α1 , . . . , αn . Show that

(j)
[det(αi )]2 = (N a)2 dK .

Solution. Since a is a submodule of index N a in OK , this is immediate

from Exercise 4.2.8.
214 CHAPTER 4. INTEGRAL BASES

4.5 Supplementary Problems

Exercise 4.5.1 Let K be an algebraic number field. Show that dK ∈ Z.

Solution. By definition
(j) # $
dK = det(ωi )2 = det Tr(ωi ωj ) ,

where ω1 , . . . , ωn is an integral basis of OK . Since Tr(ωi ωj ) ∈ Z, the

determinant is an integer.

Exercise 4.5.2 Let K/Q be an algebraic number field of degree n. Show that
dK ≡ 0 or 1 (mod 4). This is known as Stickelberger’s criterion.

Solution. Let ω1 , . . . , ωn be an integral basis of OK . By definition,

# $2
dK = det σi (ωj ) ,

where σ1 , . . . , σn are the distinct embeddings of K into Q. Now write

det(σi (ωj )) = P − N,

where P is the contribution arising from the even permutations and N the
odd permutations in the definition of the determinant. Then

dK = (P − N )2 = (P + N )2 − 4P N.

Since σi (P + N ) = P + N , and σi (P N ) = P N we see that P + N and

P N are integers. Reducing mod 4 gives the result.

Exercise 4.5.3 Let f (x) = xn + an−1 xn−1 + · · · + a1 x + a0 with ai ∈ Z be

the minimal polynomial of θ. Let K = Q(θ). If for each prime p such that
p2 | dK/Q (θ) we have f (x) Eisensteinian with respect to p, show that OK = Z[θ].

Solution. By Example 4.3.1, the index of θ is not divisible by p for any

prime p satisfying p2 | dK/Q (θ). By Exercise 4.2.8,

dK/Q (θ) = m2 dK ,
< =
where m = OK : Z[θ] . Hence m = 1.

Exercise 4.5.4 If the minimal polynomial of α is f (x) = xn + ax + b, show that

for K = Q(α),
n ! "
dK/Q (α) = (−1)( 2 ) nn bn−1 + an (1 − n)n−1 .
4.5. SUPPLEMENTARY PROBLEMS 215

Solution. By Exercise 4.3.3,

n
6
dK/Q (α) = (−1)( 2 )
n
f $ (α(i) ),
i=1

where α(1) , . . . , α(n) are the conjugates of α. Now

f $ (x) = nxn−1 + a
1
= (nxn + ax)
x
so that
(−n(aα(i) + b) + aα(i) )
f $ (α(i) ) = .
α(i)
Hence
n
6 n
6
f $ (α(i) ) = (−1)n b−1 (a(1 − n)α(i) − nb)
i=1 i=1
! "
nb
= b−1 an (1 − n)n f .
a(1 − n)

Exercise 4.5.5 Determine an integral basis for K = Q(θ) where θ3 + 2θ + 1 = 0.

Solution. By applying the previous exercise, the discriminant of θ is −59,

which is squarefree. Therefore OK = Z[θ].

Exercise 4.5.6 (Dedekind) Let K = Q(θ) where θ3 − θ2 − 2θ − 8 = 0.

(a) Show that f (x) = x3 − x2 − 2x − 8 is irreducible over Q.
(b) Consider β = (θ2 + θ)/2. Show that β 3 − 3β 2 − 10β − 8 = 0. Hence β is
integral.
(c) Show that dK/Q (θ) = −4(503), and dK/Q (1, θ, β) = −503. Deduce that 1, θ, β
is a Z-basis of OK .
(d) Show that every integer x of K has an even discriminant.
(e) Deduce that OK has no integral basis of the form Z[α].

Solution. Note that if (a) is not true, then f has a linear factor and by
the rational root theorem, this factor must be of the form x − a where a | 8.
A systematic check rules out this possibility. (b) can be checked directly.
(c) This is easy to deduce from the formula

dK/Q (1, θ, β) = 14 dK/Q (θ)

as a simple computation shows.

216 CHAPTER 4. INTEGRAL BASES

For (d), write x = A + Bθ + Cβ, A, B, C ∈ Z. Since

β2 = 6 + 2θ + 3β,
2
θ = 2β − θ,
θβ = 2β + 4,

we find

x2 = (a2 + 6C 2 + 8BC) + θ(2C 2 − B 2 + 2AB) + β(2B 2 + 3C 2 + 2AC + 4BC)

so that
dK/Q (1, x, x2 ) ≡ −503(BC)2 (3C + B)2 (mod 2),
which is an even number in all cases.
By (d), dK/Q (α) is even and hence is not equal to −503, which proves
(e).

Exercise 4.5.7 Let m = pa , with p prime and K = Q(ζm ). Show that

(1 − ζm )ϕ(m) = pOK .

Solution. First note that

xm − 1 6
= (x − ζm
b
)
xm/p − 1 1≤b<m
(b,m)=1

so that taking the limit as x goes to 1 of both sides gives

6
p = (1 − ζmb
)
1≤b<m
(b,m)=1
6 1 − ζm
b
= (1 − ζm )ϕ(m) .
1 − ζm
1≤b<m
(b,m)=1

This latter quantity is a unit since

1 − ζm 1 − ζm
ab
= = 1 + ζm
b
+ · · · + ζm
b(a−1)
1 − ζm
b 1 − ζm
b

for any a satisfying ab ≡ 1 (mod m).

Exercise 4.5.8 Let m = pa , with p prime, and K = Q(ζm ). Show that

(−1)ϕ(m)/2 mϕ(m)
dK/Q (ζm ) = .
pm/p
4.5. SUPPLEMENTARY PROBLEMS 217

Solution. We need to compute the Vandermonde determinant given by

6
(−1)ϕ(m)/2 (ζm
a b
− ζm ).
1≤a,b<m
(a,m)=1
(b,m)=1
a*=b

Let 6
θ= (ζm − ζm
b
).
1≤b<m
(b,m)=1

Clearly θ = φ$m (ζm ) and NK/Q (θ) is the discriminant we seek. Since

xm − 1
φm (x) = ,
xm/p − 1
we find
m−1
mζm
φ$m (ζm ) = m/p
,
ζm −1
and the norm of this element is
mm
m/p
.
NK/Q (ζm − 1)

m/p
Because η = ζm is a primitive pth root of unity,

NK/Q (ζm
m/p
− 1) = NQ(η)/Q (η − 1)m/p .

In Exercise 4.3.7, we saw that NQ(η)/Q (η − 1) = p.

ϕ(m)−1
Exercise 4.5.9 Let m = pa , with p prime. Show that {1, ζm , . . . , ζm } is
an integral basis for the ring of integers of K = Q(ζm ).

Solution. Clearly Z[ζm ] ⊆ OK . We want to prove the reverse inclusion.

Let λ = 1 − ζm . Since λj = (1 − ζm )j ∈ Z[ζm ] and ζm j
= (1 − λ)j ∈ Z[ζm ],
we see that Z[ζm ] = Z[λ]. Thus, it suffices to show Z[λ] ⊇ OK . Let α ∈ OK
and write
ϕ(m)−1 ϕ(m)−1
* *
α= aj λj = j
bj ζm , aj , bj ∈ Q.
j=0 j=0

It suffices to show aj ∈ Z for j = 0, 1, . . . , ϕ(m) − 1. Let σc (ζm ) = ζm

c
.
Then
ϕ(m)−1 ϕ(m)−1
* *
σc (α) = bj σc (ζm )j = cj
bj ζm
j=0 j=0
218 CHAPTER 4. INTEGRAL BASES

for each (c, m) = 1. We solve for bj using Cramer’s rule. Moreover, by the
previous question, we see that
cj
bj = cj ∈ Z.
pdj

Thus, aj has at most a power of p in its denominator. Let n be the least

nonnegative integer such that pn aj ∈ Z for j = 0, 1, . . . , ϕ(m) − 1. Suppose
n > 0. Let k be the smallest nonnegative integer such that p does not
divide pn ak . Then
ar λr ∈ pOK = λϕ(m) OK
(by the penultimate question) for r = 0, 1, . . . , k − 1. Since α is an integer,
pn α ∈ pOK = λϕ(m) OK so that
ϕ(m)−1
*
pn aj λj−k ∈ λOK
j=k

so that pn ak ∈ λOK ∩ Z = pZ, a contradiction.

Exercise 4.5.10 Let K = Q(ζm ) where m = pa . Show that

(−1)ϕ(m)/2 mϕ(m)
dK = .
pm/p

Solution. This is immediate from the previous two questions.

Exercise 4.5.11 Show that Z[ζn + ζn−1 ] is the ring of integers of Q(ζn + ζn−1 ),
where ζn denotes a primitive nth root of unity, and n = pα .

Solution. Suppose α = a0 + a1 (ζn + ζn−1 ) + · · · + aN (ζn + ζn−1 )N is an

algebraic integer with N ≤ 12 φ(n) − 1 and the ai ∈ Q. By subtracting
those terms with ai ∈ Z, we may suppose aN !∈ Z. Multiplying by ζnN and
expanding the result as a polynomial in ζn , we find that

ζnN α = aN + · · · + aN ζn2N

is an algebraic integer in Q(ζn ). Therefore, it lies in Z[ζn ]. Since

2N ≤ φ(n) − 2 ≤ φ(n) − 1,

we conclude aN ∈ Z, contrary to our assumption above. (This is also true

for arbitrary n by applying Exercise 4.5.25.)

Exercise 4.5.12 Let K and L be algebraic number fields of degree m and n,

respectively, over Q. Let d = gcd(dK , dL ). Show that if [KL : Q] = mn, then
OKL ⊆ 1/dOK OL .
4.5. SUPPLEMENTARY PROBLEMS 219

Solution. Let {α1 , . . . , αm } be a Z-basis for OK and let {β1 , . . . , βn } be

a Z-basis for OL . Then αi βj , 1 ≤ i ≤ m, 1 ≤ j ≤ n, is a Q-basis for KL
over Q since [KL : Q] = mn. Any ω ∈ OKL can therefore be written as
* mij
ω= αi βj ,
i,j
r

where r, mij ∈ Z and gcd(r, gcd(mij )) = 1. It suffices to show that r | dK

and by symmetry r | dL so that r | d. Since [KL : Q] = mn, every
embedding σ of K into C can be extended to KL acting trivially on L.
Hence * mij
σ(ω) = σ(αi )βj .
i,j
r
5
Set xi = j mij βj /r. We then obtain m equations
m
*
σ(αi )xi = σ(ω),
i=1

one for each σ : K 8→ C. We solve for xi by Cramer’s rule: xi = γi /δ where

δ = det(σ(αi )). Since δ 2 = dK we find
n
* δ 2 mij
δγi = βj ∈ OK
j=1
r

since δ and each of γi are algebraic integers. Hence dK mij /r are all integers.
It follows that r divides all dK mij . Since gcd(r, gcd(mij )) = 1, we deduce
r | dK .

Exercise 4.5.13 Let K and L be algebraic number fields of degree m and n,

respectively, with gcd(dK , dL ) = 1. If {α1 , . . . , αm } is an integral basis of OK and
{β1 , . . . , βn } is an integral basis of OL , show that OKL has an integral basis {αi βj }
given that [KL : Q] = mn. (In a later chapter, we will see that gcd(dK , dL ) = 1
implies that [KL : Q] = mn.)

Solution. This is immediate from the previous question.

√ √
Exercise 4.5.14 Find an integral basis for Q( 2, −3).
√ √
Solution. If K = Q( 2), L = Q( −3), then dK = 8, dL = −3 which
are√coprime.
√ By the previous question, a Z-basis for the ring of integers of
Q( 2, −3) is given by
: √ ! √ "O
√ 1 + −3 √ 1 + −3
1, 2, , 2 .
2 2
√
Exercise 4.5.15 Let p and q be distinct primes ≡ 1 (mod 4). Let K = Q( p),
√ √ √
L = Q( q). Find a Z-basis for Q( p, q).
220 CHAPTER 4. INTEGRAL BASES

Solution. We have dK = p, dL = q which are coprime. Now invoke the

penultimate question to deduce that
: √ √ ! √ "! √ "O
1+ p 1+ q 1+ p 1+ q
1, , ,
2 2 2 2
√ √
is a Z-basis for the ring of integers of Q( p, q).

Exercise 4.5.16 Let K be an algebraic number field of degree n over Q. Let

a1 , . . . , an ∈ OK be linearly independent over Q. Set

∆ = dK/Q (a1 , . . . , an ).

Show that if α ∈ OK , then ∆α ∈ Z[a1 , . . . , an ].

Solution. Write α = c1 a1 + · · · + cn an for some rational numbers ci . By

taking conjugates, we get a system of n equations and we can solve for ci
using Cramer’s rule. Thus cj = Aj D/∆ where D2 = ∆ and it is easy to
see that Aj D is an algebraic integer. Therefore ∆cj is an algebraic integer
lying in Q so ∆cj ∈ Z, as required.

Exercise 4.5.17 (Explicit Construction of Integral Bases) Suppose K is

an algebraic number field of degree n over Q. Let a1 , . . . , an ∈ OK be linearly
independent over Q and set

∆ = dK/Q (a1 , . . . , an ).

For each i, choose the least natural number dii so that for some dij ∈ Z, the
number
,i
wi = ∆−1 dij aj ∈ OK .
j=1

Show that w1 , . . . , wn is an integral basis of OK .

Solution. First observe that there are integers cij so that

i
*
∆−1 cij aj ∈ OK
j=1

(e.g., cij = ∆). Clearly w1 , . . . , wn are linearly independent over Q because

dK/Q (w1 , . . . , wn ) = ∆−n (d11 · · · dnn )2 dK/Q (a1 , . . . , an )

by Exercise 4.2.7, and the right-hand side is nonzero. Observe now that if
α ∈ OK can be written as

α = ∆−1 (c1 a1 + · · · + cj aj )

for some j, then djj | cj . Indeed, write cj = sdjj + r, 0 ≤ r < djj , so that
# $
α − swj = ∆−1 (c1 − dj1 )a1 + · · · + raj ∈ OK ,
4.5. SUPPLEMENTARY PROBLEMS 221

contrary to our choice of wj if r != 0.

We now show by induction on j that every number of OK of the form

∆−1 (x1 a1 + · · · + xj aj )

with xi ∈ Z lies in Z[w1 , . . . , wn ]. For j = 1, there is nothing to prove

because then d11 | x1 and we are done. Assume that we have proved it for
j < k. Then suppose

y = ∆−1 (x1 a1 + · · · + xk ak ) ∈ OK

with xi ∈ Z. Then dkk | xk so that for some integer t,

y − twk = ∆−1 (x$1 a1 + · · · + x$k−1 ak−1 ) ∈ OK .

By induction, the right-hand side lies in Z[w1 , . . . , wn ] and so does y. For

j = n, this means that every number of OK of the form ∆−1 (x1 a1 + · · · +
xn an ) with xi ∈ Z lies in Z[w1 , . . . , wn ]. But by the previous exercise,
every α ∈ OK can be so expressed.

Exercise 4.5.18 If K is an algebraic number field of degree n over Q and

a1 , . . . , an ∈ OK are linearly independent over Q, then there is an integral basis
w1 , . . . , wn of OK such that

aj = cj1 w1 + · · · + cjj wj ,

cij ∈ Z, j = 1, . . . , n.

Solution. We take w1 , . . . , wn as constructed in the previous exercise.

This is an integral basis. Solving for aj and noting that the matrix (dij ) is
lower triangular, we see that each aj can be written as above. Moreover,
the cij ∈ Z since w1 , . . . , wn is an integral basis by construction.

Exercise 4.5.19 If Q ⊆ K ⊆ L and K, L are algebraic number fields, show that

dK | dL .

Solution. Let [K : Q] = m, [L : K] = n. Let a1 , . . . , am be an integral

basis of K. Extend this to a basis of L (viewed as a vector space over Q) so
that a1 , . . . , amn is linearly independent over Q. By Exercise 3.3.7, we may
suppose each ai is an algebraic integer. By the previous exercise, there is
an integral basis w1 , . . . , wmn of OL such that

aj = cj1 w1 + · · · + cjj wj , cij ∈ Z.

Since the matrix (cij ) is triangular, it is easy to see that w1 , . . . , wm lie in

K. Because the wi are algebraic integers, and a1 , . . . , am is an integral basis
of K, it follows that w1 , . . . , wm is an integral basis of K. Now write down
the definition of the discriminant of L. Let σ1 , . . . , σmn be the embeddings
(i)
of L into C such that σi (wj ) = wj for 1 ≤ j ≤ m.
222 CHAPTER 4. INTEGRAL BASES

We order the σ1 , . . . , σmn so that σi (x) = σi" (x) for i ≡ i$ (mod m)

and x ∈ K. Then
# $2
dL = det σi (wj )
1 (1) (m) (1) (m)
1
(m) 12
1w ··· w1 w1 ··· w1 ··· w1
1 1 1
1 . 1
1 .. 1
1 1
1 (1) (m) (1) (m) (m) 1
1 wm ··· wm wm ··· wm ··· wm 1
= 1 (1) (m) (m+1) (2m) (mn) 1
1wm+1 · · · wm+1 wm+1 ··· wm+1 ··· wm+1 11
1
1 .. 1
1 . 1
1 (1) (m) (m+1)
1
(mn) 1
1w ··· wmn wmn ··· w
mn mn

1 (1) (m)
12
1w ··· w1 0 ··· 0 1
1 1 1
1 . 1
1 .. 1
1 1
1 (1) (m) 1
1 wm ··· wm 0 ··· 0 1
= 1 (1) (m) (m+1) (1) (mn) (m) 1
1wm+1 ··· wm+1 wm+1 − wm+1 ··· wm+1 − wm+1 11
1
1 .. 1
1 . 1
1 (1) (m) (m+1) (1) (mn)
1
(m) 1
1w ··· wmn wmn − wmn ··· w −w
mn mn mn
= dK · a,

where a is an algebraic integer. Since dL /dK is a rational number, we

deduce that a is a rational integer.

Exercise 4.5.20 (The Sign of the Discriminant) Suppose K is a number

field with r1 real embeddings and 2r2 complex embeddings so that

r1 + 2r2 = [K : Q] = n

(say). Show that dK has sign (−1)r2 .

Solution. Let ω1 , . . . , ωn be an integral basis of K. Then

# $2
dK = det σi (ωj ) ,

where σ1 , . . . , σn are the embeddings of K. Clearly

# $ # $
det σi (ωj ) = (−1)r2 det σi (ωj )

since complex conjugation #interchanges $ r2 rows.

#If r 2 is$ even, then det σ i (ω j ) is real so that dK > 0. If r2 is odd,
det σi (ωj ) is purely imaginary so that dK < 0.

Exercise 4.5.21 Show that only finitely many imaginary quadratic fields K are
Euclidean.
4.5. SUPPLEMENTARY PROBLEMS 223

Solution. If ψ is a Euclidean algorithm for OK , then let α0 ∈ OK be such

that ψ(α0 ) is the minimum nonzero value of ψ. Then, every residue class
mod α0 is represented by 0 or an element α ∈ OK such that ψ(α) = 0.
Thus α0 is a unit. In an imaginary quadratic field, there are only finitely
many units. If dK != −3, −4, the units are ±1. Thus, if dK != −3, −4, we
find NK/Q (α0 ) ≤ 3, which implies that α0 = ±1. In particular, if α1 is such
that ψ(α1 ) = min ψ(α) where the minimum ranges over ψ(α) != ψ(α0 ),
then α1 is not a unit, ψ(α1 ) > ψ(α0 ) so that not every residue class mod
α1 can be represented by a class containing an element whose ψ-value is
smaller than ψ(α1 ).
√
Exercise 4.5.22 Show that Z[(1 + −19)/2] is not Euclidean. (Recall that in
Exercise 2.5.6 we showed this ring is not Euclidean for the norm map.)

Solution. The argument of the previous exercise shows that not all residue
classes mod 2 and mod 3 are represented by elements of smaller ψ-value.

Exercise 4.5.23 (a) Let A = (aij ) be an m × m matrix, B = (bij ) an n × n

matrix. We define the (Kronecker) tensor product A ⊗ B to be the mn × mn
matrix obtained as
 
Ab11 Ab12 · · · Ab1n
 Ab21 Ab22 · · · Ab2n 
 
 . .. ..  ,
 .. . . 
Abn1 Abn2 · · · Abnn

where each block Abij has the form

 
a11 bij a12 bij ··· a1m bij
 a21 bij a22 bij ··· a2m bij 
 
 . .. ..  .
 .. . . 
am1 bij am2 bij ··· amm bij

If C and D are m × m and n × n matrices, respectively, show that

(A ⊗ B)(C ⊗ D) = (AC) ⊗ (BD).

(b) Prove that det(A ⊗ B) = (det A)n (det B)m .

Solution. Part (a) is a straightforward matrix multiplication computation.

For part (b), we use linear algebra to find a matrix U such that U −1 BU is
upper triangular:
 
c11 c12 · · · c1n
 0 c22 · · · c2n 
 
U −1 BU =  . .. ..  .
 .. . . 
0 0 · · · cnn
224 CHAPTER 4. INTEGRAL BASES

Then det B = c11 c22 · · · cnn . Also, by (a),

(I ⊗ U )−1 (A ⊗ B)(I ⊗ U ) = A ⊗ (U −1 BU )

which is  
Ac11 Ac12 ··· Ac1n
 0 Ac22 ··· Ac2n 
 
 .. .. ..  .
 . . . 
0 0 ··· Acnn
Again, by linear algebra, we see that
n
6
det(A ⊗ B) = det(Acii )
i=1
6n
= (cm
ii det A)
i=1
= (det B)m (det A)n ,

as desired.

Exercise 4.5.24 Let K and L be algebraic number fields of degree m and n,

respectively, with gcd(dK , dL ) = 1. Show that

dKL = dn m
K · dL .

If we set
log |dM |
δ(M ) =
[M : Q]
deduce that δ(KL) = δ(K) + δ(L) whenever gcd(dK , dL ) = 1.

Solution. By a previous exercise, OKL has integral basis {αi βj } where

α1 , . . . , αm is an integral basis of OK and β1 , . . . , βn is an integral basis of
OL . Let
# $
A = TrK/Q (αi αj ) ,
# $
B = TrL/Q (βi βj ) .

Then it is easily verified that the discriminant of KL/Q is det(A ⊗ B).

L . The second part of the question is
By the previous exercise, this is dnK dm
immediate upon taking logarithms.

Exercise 4.5.25 Let ζm denote a primitive mth root of unity and let K =
Q(ζm ). Show that OK = Z[ζm ] and

(−1)φ(m)/2 mϕ(m)
dK = / φ(m)/(p−1)
.
p|m p
4.5. SUPPLEMENTARY PROBLEMS 225

Solution. Factor 6
m= pα .
pα -m

Since the discriminants of Q(ζpα ) for pα =m are coprime, we have by the

previous exercise (and Supplementary Exercise 4.5.8)

log |dK | * ! 1
"
= α− log p.
φ(m) α
p−1
p -m

The sign of the determinant is (−1)r2 = (−1)ϕ(m)/2 . The fact that OK =

Z[ζm ] follows by an induction argument and Exercise 4.5.13.

Exercise 4.5.26 Let K be an algebraic number field. Suppose that θ ∈ OK is

such that dK/Q (θ) is squarefree. Show that OK = Z[θ].
< =
Solution. Let m = OK : Z[θ] . Then, by Exercise 4.2.8,

dK/Q (θ) = m2 dK .

If dK/Q (θ) is squarefree, m = 1.

Chapter 5

Dedekind Domains

5.1 Integral Closure

Exercise 5.1.1 Show that a nonzero commutative ring R with identity is a field
if and only if it has no nontrivial ideals.

Solution. If x ∈ R, x != 0, is a nonunit, then 1 ∈/ (x), so (x) is a nontrivial

ideal.
Suppose that R has a nontrivial ideal a. Let x ∈ a, x != 0. Then (x) ⊆ a.
If x is a unit, then 1 ∈ (x) ⊆ a, so a = R, a contradiction. Thus, x is not a
unit, so R is not a field.

Exercise 5.1.3 Show that a finite integral domain is a field.

Solution. Let R be a finite integral domain. Let x1 , x2 , . . . , xn be the

elements of R. Suppose that xi xj = xi xk , for some xi != 0.
Then xi (xj − xk ) = 0. Since R is an integral domain, xj = xk , so j = k.
Thus, for any xi != 0,

{xi x1 , xi x2 , . . . , xi xn } = {x1 , x2 , . . . , xn }.

Since 1 ∈ R, there exists xj such that xi xj = 1. Therefore, xi is invertible.

Thus all nonzero elements are invertible, so R is a field.

Exercise 5.1.4 Show that every nonzero prime ideal ℘ of OK is maximal.

Solution. OK /℘ is finite from Exercise 4.4.3 and it is an integral domain

from Theorem 5.1.2 (b). Thus, Exercise 5.1.3 shows that OK /℘ is a field,
which in turn implies that ℘ is a maximal ideal of OK .

Exercise 5.1.5 Show that every unique factorization domain is integrally closed.

227
228 CHAPTER 5. DEDEKIND DOMAINS

Solution. Let R be a unique factorization domain, α ∈ Q(R).

Then α = a/b, for some a, b ∈ R with (a, b) = 1. If α is integral over R,
then we have a polynomial equation
αn + cn−1 αn−1 + · · · + c0 = 0, ci ∈ R.
Thus, multiplying through by bn and isolating an , we have
an = −b(cn−1 an−1 + · · · + c1 abn−2 + c0 bn−1 ).
Thus, b|an . But, (a, b) = 1. By unique factorization, (an , b) = 1. Therefore,
b is a unit in R, so a/b ∈ R.
Thus, R is integrally closed.

5.2 Characterizing Dedekind Domains

Exercise 5.2.1 If a " b are ideals of OK , show that N (a) > N (b).

Solution. Define a map f : OK /a −→ OK /b by f (x + a) = x + b. If

x + a = y + a, then x − y ∈ a ⊆ b, so x + b = y + b. Thus, f is well-defined.
The function f is not one-to-one since for any y ∈ b\a, f (y + a) = 0,
but y + a != 0. It is onto since x + b = f (x + a). So we have a map from
the finite set OK /a to the finite set OK /b which is onto but not one-to-one.
Thus |OK /a| > |OK /b|, i.e., N (a) > N (b).

Exercise 5.2.2 Show that OK is Noetherian.

Solution. Suppose that a1 # a2 # a3 # · · · is an ascending chain of

ideals which does not terminate. Then N (a1 ) > N (a2 ) > N (a3 ) > · · · but
N (ai ) is finite and positive for all i, so such a strictly decreasing sequence
of positive integers must stop. Thus, the ascending chain of ideals must
terminate.

Exercise 5.2.4 Show that any principal ideal domain is a Dedekind domain.

Solution. Let R be a principal ideal domain. R is Noetherian since every

ideal is finitely generated. R is integrally closed since any principal ideal
domain is a unique factorization domain, and so is integrally closed by
Exercise 5.1.5.
Let (p) != 0 be a prime ideal and (x) ⊇ (p). Then, p ∈ (x), so p = xy,
for some y ∈ R. Thus, xy ∈ (p), so x ∈ (p) or y ∈ (p). If x ∈ (p), then
(x) = (p), and if y ∈ (p), then y = pq, for some q ∈ R. This would imply
that p = xy = xqp and so xq = 1, since R is an integral domain. Thus,
(x) = R. Therefore, (p) is maximal.
Thus, R is a Dedekind domain.
√
Exercise 5.2.5 Show that Z[ −5] is a Dedekind domain, but not a principal
ideal domain.
5.3. FRACTIONAL IDEALS AND UNIQUE FACTORIZATION 229

√
Solution. Z[ −5] is not a unique factorization
√ √ domain as was seen in
Chapter 2 by taking 6 = 2 × 3 = (1 + −5)(1 − −5), and so cannot be a
principal ideal domain.
To see that it is a Dedekind domain, it is enough to show that √ it is
the set of algebraic integers of the algebraic number field K =√Q( −5).
However, we have already proved this, in Exercise 4.1.3. So Z[ −5] is a
Dedekind domain.

5.3 Fractional Ideals and Unique Factoriza-

tion
Exercise 5.3.1 Show that any fractional ideal is finitely generated as an OK -
module.

Solution. Let A be a fractional ideal of OK . Choose m ∈ Z such that

mA ⊆ OK . Since A is an OK -module, mA is an OK -module contained
in OK and so is an ideal of OK . Since OK is Noetherian, mA is finitely
generated as an ideal. If mA is generated as an ideal by {a1 , . . . , an },
then A is generated by {m−1 a1 , . . . , m−1 an } as an OK -module. Thus, A
is finitely generated as an OK -module.

Exercise 5.3.2 Show that the sum and product of two fractional ideals are again
fractional ideals.

Solution. Let A and B be fractional ideals. Since A and B are both

OK -modules, so are their sum and product.
Let mA ⊆ OK , nB ⊆ OK with m, n ∈ Z. Then

mn(AB) = (mA)(nB)
⊆ OK ,

so AB is a fractional ideal. Also,

mn(A + B) = n(mA) + m(nB)

⊆ nOK + mOK
⊆ OK ,

so A + B is a fractional ideal.

Exercise 5.3.7 Show that any fractional ideal A can be written uniquely in the
form
℘1 . . . ℘ r
,
℘(1 . . . ℘(s
where the ℘i and ℘(j may be repeated, but no ℘i = ℘(j .
230 CHAPTER 5. DEDEKIND DOMAINS

Solution. Choose a nonzero element c ∈ Z such that b := cA ⊆ OK . Let

(c) = m1 · · · ms , b = n1 · · · nt , the mi , nj prime.
Then (c)A = b, so
n1 · · · nt
A= ,
m1 · · · ms
and cancelling the primes on the numerator that equal some prime on the
denominator, we have that mi != nj ∀i, j. Also, if

a1 · · · a v
A= ,
b1 · · · bw

with no ai = bj , then

a1 · · · av m1 · · · ms = b1 · · · bw n1 · · · nt .

By unique factorization and the fact that no bi is an aj and no mi is an nj ,

the bi ’s must coincide up to reordering with the mj ’s and the ai ’s with the
nj ’s, and so the factorization is unique.

Exercise 5.3.8 Show that, given any fractional ideal A =

# 0 in K, there exists a
fractional ideal A−1 such that AA−1 = OK .

Solution. Let
℘1 · · · ℘r
A= .
℘$1 · · · ℘$s
Then
℘$1 · · · ℘$s
A−1 =
℘1 · · · ℘r

is a fractional ideal with AA−1 = OK .

Exercise 5.3.9 Show that if a and b are ideals of OK , then b | a if and only if
there is an ideal c of OK with a = bc.

Solution. If b ⊇ a, then c := ab−1 ⊆ bb−1 = OK . Thus, a = bc, with c an

ideal of OK .
If a = bc with c ⊆ OK , then a = bc ⊆ b.

/r min(ei ,fi )
Exercise 5.3.10 Show that gcd(a, b) = a + b = i=1 ℘i .

Solution. a ⊆ a + b, b ⊆ a + b, so a + b | a, a + b | b.
If e | a and e | b, then a ⊆ e, b ⊆ e, so that a + b ⊆ e, i.e., e | a + b.
Therefore, a + b = gcd(a, b).
5.3. FRACTIONAL IDEALS AND UNIQUE FACTORIZATION 231

2r min(ei ,fi )
Let d = i=1 ℘i , and let min(ei , fi ) = ai ,
r
6
a = ℘ei i
i=1
6r r
6
= ℘ei i −ai ℘ai i
i=1 i=1
6r
= ℘ei i −ai d, ei − ai ≥ 0 ∀i.
i=1

Thus, d ⊇ a which implies that d | a. Similarly, d | b.

Suppose that e | a and e | b. Let
r
6
e= ℘ki i , ki ≥ 0,
i=1

be the unique factorization of e as a product of prime ideals. Suppose

ki > ei for some i ∈ {1, . . . , r}. We know that ℘ki i | e and e | a, so ℘ki i | a,
i.e., ℘ki i ⊇ ℘e11 · · · ℘err . Thus,
e e
(℘−1 ei k i
i ) ℘i ⊇ ℘e11 · · · ℘i−1
i−1 i+1
℘i+1 · · · ℘err ,
e e
℘i ⊇ ℘ki i −ei ⊇ ℘e11 · · · ℘i−1
i−1 i+1
℘i+1 · · · ℘err ,
℘i ⊇ ℘j ,
for some j != i, and so ℘i = ℘j , since ℘j is maximal. But this is a contra-
diction, so ki ≤ ei for all i, and every prime occurring in e must occur in a.
Similarly for b, so ki ≤ min(ei , fi ). Thus, e | d, so d = gcd(a, b).
/r max(ei ,fi )
Exercise 5.3.11 Show that lcm(a, b) = a ∩ b = i=1 ℘i .

Solution. a ⊇ a ∩ b, b ⊇ a ∩ b, so a | a ∩ b, b | a ∩ b. Suppose that a | e

and b | e. Then e ⊆ a, e ⊆ b, so e ⊆ a ∩ b. Thus, a ∩ b = lcm(a, b). Let
2r max(ei ,fi )
m = i=1 ℘i and let
r
6 s
6
e= ℘ki i (℘$j )tj ,
i=1 j=1

where ℘1 , . . . , ℘r , ℘$1 , . . . , ℘$s are distinct prime ideals. Suppose ki < ei for
some i,
e ⊆ a,
⇒ ℘k11 · · · ℘kr r (℘$1 )t1 · · · (℘$s )ts ⊆ ℘e11 · · · ℘err ,
ki−1 ki+1
⇒ ℘k11 · · · ℘i−1 ℘i+1 · · · ℘kr r (℘$1 )t1 · · · (℘$s )ts ⊆ ℘e11 · · · ℘ei i −ki · · · ℘err
⊆ ℘ei i −ki
⊆ ℘i .
232 CHAPTER 5. DEDEKIND DOMAINS

Thus, ℘i ⊇ ℘j , for some j = ! i, or ℘i = ℘$j , for some j. Neither is true, so

ki ≥ ei . Similarly, ki ≥ fi . Thus, m | e, so m = lcm(a, b).

Exercise 5.3.12 Suppose a, b, c are ideals of OK . Show that if ab = cg and

gcd(a, b) = 1, then a = dg and b = eg for some ideals d and e of OK . (This
generalizes Exercise 1.2.1.)

Solution. We factor uniquely into prime ideals:

a = ℘e11 · · · ℘err

and
b = (℘$1 )f1 · · · (℘$t )ft
where ℘1 , . . . , ℘r , ℘$1 , . . . , ℘$t are distinct prime ideals since gcd(a, b) = 1.
Now let c = ℘a1 1 · · · ℘ar r (℘$1 )b1 · · · (℘$t )bt . Since ab = cg , we must have

ei = ai g, 1 ≤ i ≤ r,
fi = bi g, 1 ≤ i ≤ t,

by unique factorization. Thus a = dg and b = eg with

d = ℘a1 1 · · · ℘rar ,
e = (℘$1 )a1 · · · (℘$t )at ,

as desired.

Exercise 5.3.14 Show that ord℘ (ab) = ord℘ (a) + ord℘ (b), where ℘ is a prime
ideal.

Solution. From a previous exercise, a = ℘t a1 and b = ℘s b1 , where ℘ & a1 ,

and ℘ & b1 . Thus, ab = ℘t+s a1 b1 , so ℘s+t | ab. If ℘s+t+1 | ab, then
ab = ℘s+t+1 c, so ℘c = a1 b1 .
Thus, ℘ ⊇ a1 b1 , so ℘ ⊇ a1 or ℘ ⊇ b1 , since ℘ is prime. This is a
contradiction, so ord℘ (ab) = t + s = ord℘ (a) + ord℘ (b).

Exercise 5.3.15 Show that, for α #= 0 in OK , N ((α)) = |NK (α)|.

5n
Solution. Let OK = Zω1 + · · · + Zωn . There exist αi = j=1 pij ωj , with
pii > 0, pij ∈ Z such that (α) = Zα1 + · · · + Zαn and N ((α)) = p11 · · · pnn ,
from Theorem 4.2.2.
We also know
5n that (α) = Zαω1 + · · · + Zαωn . Now, NK (α) = det(cij )
where αωi = j=1 cij ωj . And, if C = (cij ), R = (rij ), P = (pij ), then

(αω1 , . . . , αωn )T = C(ω1 , . . . , ωn )T

= R(α1 , . . . , αn )T
= RP (ω1 , . . . , ωn )T .
5.4. DEDEKIND’S THEOREM 233

5n
where αωi = j=i rji αj , rji ∈ Z. Therefore, by definition,

NK (α) = det(RP ) = det(R) det(P ).

R and R−1 have integer entries, since {αωi } and {αi } are both Z-bases
for (α). Thus, det(R) = ±1. So, det(C) = ±det(P ), det(P ) ≥ 0. Thus,
|det(C)| = det(P ).
Thus, |NK (α)| = |det(C)| = det(P ) = N ((α)).

Exercise 5.3.17 If we write pOK as its prime factorization,

e
pOK = ℘e11 · · · ℘gg ,
&g
show that N (℘i ) is a power of p and that if N (℘i ) = pfi i , i=1 ei fi = n.
e
Solution. Since ℘ei i + ℘j j = OK for i != j,

OK /pOK 1 OK /℘e11 ⊕ · · · ⊕ OK /℘egg ,

by the Chinese Remainder Theorem. Therefore

pn = N (℘e11 ) · · · N (℘egg )
= N (℘1 )e1 · · · N (℘g )eg .

Thus, N (℘i ) = pfi , for some positive integer fi , and n = e1 f1 + · · · + eg fg .

5.4 Dedekind’s Theorem

Exercise 5.4.1 Show that D −1 is a fractional ideal of K and find an integral
basis.

Solution. D−1 is an OK -module since if x ∈ OK and y ∈ D−1 , then

xy ∈ D−1 because
Tr(xyOK ) ⊆ Tr(yOK ) ⊆ Z.
Thus, OK D−1 ⊆ D−1 .
Now, let {ω1 , . . . , ωn } be an integral basis of OK . There is a dual basis
(see Exercise 4.2.1) {ω1∗ , . . . , ωn∗ } such that Tr(ωi ωj∗ ) = δij , 1 ≤ i, j ≤ n.
Now Tr(ωi∗ ωj ) ∈ Z for all ωj , ωi∗ , so Zω1∗ + · · · + Zωn∗ ⊆ D−1 .
We claim that D−1 = Zω1∗ + · · · + Zωn∗ . Let x ∈ D−1 . Then
n
*
x= ai ωi∗ , ai ∈ Q,
i=1

since {ω1∗ , . . . , ωn∗ } is a Q-basis for K. Then

/ r 0
*
Tr(xωj ) = Tr ai ωi ωj = aj ∈ Z.
∗

i=1
234 CHAPTER 5. DEDEKIND DOMAINS

Therefore x ∈ Zω1∗ + · · · + Zωn∗ . Therefore D−1 = Zω1∗ + · · · + Zωn∗ .

2n for each ωi there
Since ∗
is an ai ∈ Z such that ai ωi∗ ∈ OK , if we let
m = i=1 ai , then mD ⊆ OK . Thus, D−1 is a fractional ideal.
−1

Exercise 5.4.2 Let D be the fractional ideal inverse of D −1 . We call D the

different of K. Show that D is an ideal of OK .

Solution. D is certainly a fractional ideal of OK and DD−1 = OK . But,

from Lemma 4.1.1, 1 ∈ D−1 . Thus, D ⊆ DD−1 = OK . Thus, D is an ideal
of OK .

Exercise 5.4.5 Show that if p is ramified, p | dK .

Solution. Since p is ramified, e℘ > 1 for some prime ideal ℘ containing p.

Thus, ℘ | D, from the previous theorem, say ℘a = D. From the multiplica-
tivity of the norm function, we have N (℘)N (a) = N (D), so N (℘) | N (D).
Thus, pf℘ | dK , and so p | dK .

5.5 Factorization in OK
Exercise 5.5.2 If in6 the previous
7 theorem we do not assume that OK = Z[θ]
but instead that p ! OK : Z[θ] , show that the same result holds.

Solution. Let m be the index of Z[θ] in OK . Then for α ∈ OK , mα ∈ Z[θ].

In other words, given any α, we may write mα = b0 + b1 θ + b2 θ2 + · · · +
bn−1 θn−1 . Consider this expression mod p. Since m is coprime to p there
is an m$ such that m$ m ≡ 1 (mod p). Then

α ≡ b0 m$ + b1 m$ θ + · · · + bn−1 m$ θn−1 (mod p).

Thus, OK ≡ Z[θ] (mod p).

In the proof of the previous exercise, we only used the fact that OK =
Z[θ] at one point. This was when we wrote that ri (θ) = pa(θ) + fi (θ)b(θ).
We now note that we simply need that ri (θ) ≡ pa(θ) + fi (θ) (mod p). The
proof will follow through in the same way, and we deduce that (p, fi (θ)) is
a prime ideal of Z[θ]. However, since

OK /(p) 1 Z[θ]/(p),

then
OK /(p, fi (θ)) 1 Z[θ]/(p, fi (θ)).
The rest of the proof will be identical to what was written above.

Exercise 5.5.3 Suppose that f (x) in the previous exercise is Eisensteinian with
respect to the prime p. Show that p ramifies totally in K. That is, pOK = (θ)n
where n = [K : Q].
5.6. SUPPLEMENTARY PROBLEMS 235
< =
Solution. By Example 4.3.1, we know that p ! OK : Z[θ] . Moreover,
f (x) ≡ xn (mod p). The result is now immediate from Exercise 5.5.2.

Exercise 5.5.4 Show that (p) = (1 − ζp )p−1 when K = Q(ζp ).

Solution. The minimal polynomial of ζp is Φp (x), the pth cyclotomic

polynomial. Recall that f (x) = Φp (x + 1) is p-Eisensteinian, and that
ζp − 1 is a root. Since Q(ζp ) = Q(1 − ζp ), and Z[ζp ] = Z[1 − ζp ], the
previous exercise tells us that (p) = (1 − ζp )p−1 .

5.6 Supplementary Problems

Exercise 5.6.1 Show that if a ring R is a Dedekind domain and a unique fac-
torization domain, then it is a principal ideal domain.

Solution. Consider an arbitrary prime ideal I of R. Since R is a Dedekind

domain, I is finitely generated and we can write I = (a1 , . . . , an ) for some
set of generators a1 , . . . , an . In a unique factorization domain, every pair
of elements has a gcd, and so d = gcd(a1 , . . . , an ) exists. Then (d) =
(a1 , . . . , an ) = I, thus proving that R is a principal ideal domain.

Exercise 5.6.2 Using

√ Theorem 5.5.1, find a prime ideal factorization of 5OK
and 7OK in Z[(1 + −3)/2].

Solution. We now consider f (x) (mod 7). We have

x2 − x + 1 ≡ x2 + 6x + 1 ≡ (x + 2)(x + 4) (mod 7)

so 7 splits and its factorization is

! √ "! √ "
5 + −3 9 + −3
(7) = 7, 7, .
2 2

Exercise 5.6.3 Find a prime ideal factorization of (2), (5), (11) in Z[i].

Solution. The minimal polynomial of i is x2 + 1. We consider it first mod

2.
x2 + 1 ≡ (x + 1)2 (mod 2)
so (2) = (2, i + 1)2 = (i + 1)2 since 2i = (i + 1)2 .

x2 + 1 ≡ x2 − 4 ≡ (x + 2)(x − 2) (mod 5)

so (5) = (5, i + 2)(5, i − 2) = (i + 2)(i − 2) since (2 + i)(2 − i) = 5.

Finally, we consider f (x) (mod 11). Since if the polynomial reduces it
must have an integral root, we check all the possibilities mod 11 and deduce
that it is in fact irreducible. Thus, 11 stays prime in Z[i].
236 CHAPTER 5. DEDEKIND DOMAINS

√
Exercise 5.6.4 Compute the different D of K = Q( −2).

Solution. By Theorem 5.4.3, we know√ that N (D) = |dK | = 8. Also, we

showed in Chapter 2 that OK = Z[√ −2] is Euclidean and thus a principal
ideal domain. Then D = (a + b −2) for some a, b ∈ Z and N (D) =
a2 + 2b2 = 8. The only solution in integers is a = 0, b = ±2, so
√ √
D = (2 −2) = (−2 −2).
√
Exercise 5.6.5 Compute the different D of K = Q( −3).

Solution. As in the previous exercise, we first observe that N (D) = |dK | =

3 and since the integers of this ring are the Eisenstein integers, which form
a Euclidean ring, Z[ρ] is a principal ideal domain and D = (a + bρ) for some
a, b ∈ Z. We must find all solutions to the equation

N (D) = 3 = a2 − ab + b2 .

Since this is equivalent to (2a − b)2 + 3b2 = 12, we note that |b| cannot be
greater than 2. Checking all possibilities, we find all the elements of Z[ρ]
of norm 3: 2 + ρ, −1 + ρ, −2 − ρ, 1 − ρ, 1 + 2ρ and −1 − 2ρ. Some further
checking reveals that these six elements are all associates, and so they each
generate the same principal ideal. Thus, D = (2 + ρ).

Exercise 5.6.6 Let K = Q(α) be an algebraic number field of degree n over Q.

Suppose OK = Z[α] and that f (x) is the minimal polynomial of α. Write

f (x) = (x − α)(b0 + b1 x + · · · bn−1 xn−1 ), bi ∈ O K .

Prove that the dual basis to 1, α, . . . , αn−1 is

b0 bn−1
,... , ( .
f ( (α) f (α)

Deduce that
1
D −1 = (Zb0 + · · · + Zbn−1 ).
f ( (α)

Solution. Let α1 = α, α2 , . . . , αn be the n distinct roots of f (x). We

would like to show that
*n
f (x) αr
· $ i = xr
i=1
x − αi f (αi )

for 0 ≤ r ≤ n − 1. Define the polynomial

*n
f (x) αr
gr (x) = · $ i − xr .
i=1
x − αi f (αi )
5.6. SUPPLEMENTARY PROBLEMS 237

Consider gr (α1 ). Note that f (α1 )/(α1 − αi ) = 0 for all i except i = 1. Also,
! "
f (x)
= f $ (α1 ).
x − α1 x=α1

Thus, gr (α1 ) = 0, and similarly, gr (αi ) = 0 for 1 ≤ i ≤ n. Since

deg(gr (x)) ≤ n − 1, it can have at most n − 1 roots. As we found n
distinct roots, gr (x) must be identically zero.
For a polynomial h(x) = c0 + c1 x + · · · + cm xm ∈ K[x], we define the
trace of h(x) to be
m
*
Tr(h(x)) = Tr(ci )xi ∈ Q[x].
i=0

Since α1 , . . . , αn are all the conjugates of α, it is clear that

! " n
*
f (x)αr f (x)αir
Tr = = xr .
(x − α)f $ (α) i=1
(x − αi )f $ (αi )

But,
! " n−1
* ! "
f (x)αr bi α r
Tr = Tr xi = xr .
(x − α)f $ (α) i=0
f $ (α)
Thus, ! "
bi α r
Tr =0
f $ (α)
unless i = r, in which case the trace is 1. Recall that if ω1 , . . . , ωn is a basis,
its dual basis ω1∗ , . . . , ωn∗ is characterized by Tr(ωi ωj∗ ) = δij , the Kronecker
delta function. Thus, we have found a dual basis to 1, α, . . . , αn−1 , and it
is
b0 bn−1
,... , $ .
f $ (α) f (α)
By Exercise 5.4.1,
1
D−1 = (Zb0 + · · · + Zbn−1 ).
f $ (α)

Exercise 5.6.7 Let K = Q(α) be of degree n over Q. Suppose that OK = Z[α].

Prove that D = (f ( (α)).

Solution. Let f (x) be the minimal polynomial of α, and let

f (x) = (x − α)(b0 + · · · + bn−1 xn−1 ).

Since f (x) is monic, bn−1 = 1. Also, an−1 = bn−2 − αbn−1 which means
that bn−2 = an−1 + α, where an−1 is an integer.
238 CHAPTER 5. DEDEKIND DOMAINS

We know from the previous exercise that

1
D−1 = (Zb0 + · · · + Zbn−1 ).
f $ (α)

Since bn−1 = 1, Z ⊆ Zb0 + · · · + Zbn−1 . Since bn−2 = an−1 + α, we can

deduce that α ∈ Zb0 + · · · + Zbn−1 , and by considering the expressions for
ai , 0 ≤ i ≤ n, we see that in fact αi ∈ Zb0 + · · · + Zbn−1 for 1 ≤ i ≤ n − 1.
Thus,
Z[α] ⊆ Zb0 + · · · + Zbn−1 ⊆ Z[α],
and so we have equality. Thus,
1
D−1 = OK
f $ (α)

and so D = (f $ (α)).

Exercise 5.6.8 Compute the different D of Q[ζp ] where ζp is a primitive pth

root of unity.

Solution. We can apply the results of the previous exercise to get D =

(f $ (ζp )).

xp − 1
f (x) = = xp−1 + xp−2 + · · · + x + 1,
x−1
pxp−1 (x − 1) − (xp − 1)
f $ (x) = ,
(x − 1)2
pζp−1
f $ (ζp ) = .
ζp − 1

Since ζp−1 is a unit, we find

! "
p
D= .
ζp − 1

From Exercise 5.5.4 we know that (p) = (1−ζp )p−1 , and so D = (1−ζp )p−2 .

Exercise 5.6.9 Let p be a prime, p ! m, and a ∈ Z. Show that p | φm (a)

if and only if the order of a (mod p) is n. (Here φm (x) is the mth cyclotomic
polynomial.)
2
Solution. Since xm − 1 = d|m φd (x), we have am ≡ 1 (mod p). Let k be
the order of a (mod p). Then k | m. If k < m, then
6
ak − 1 = φd (a) ≡ 0 (mod p)
d|k
5.6. SUPPLEMENTARY PROBLEMS 239

so that φd (a) ≡ 0 (mod p) for some d | k. Then

am − 1 = φm (a)φd (a)( other factors ) ≡ 0 (mod p2 ).

Since φm (a + p) ≡ φm (a) (mod p) and similarly for φd (a), we also have

(a + p)m ≡ 1 (mod p2 ). But then (a + p)m = am + mam−1 p (mod p2 ) so
that mam−1 ≡ 0 (mod p), a contradiction.
Conversely, suppose that a has order m so that am ≡ 1 (mod p). Then
φd (a) ≡ 0 (mod p) for some d | m. If d < m, then the order of a (mod p)
would be less than m.

Exercise 5.6.10 Suppose p ! m is prime. Show that p | φm (a) for some a ∈ Z if

and only if p ≡ 1 (mod m). Deduce from Exercise 1.2.5 that there are infinitely
many primes congruent to 1 (mod m).

Solution. If p | φm (a), by the previous exercise the order of a (mod p) is

m so that m | p − 1.
Conversely, if p ≡ 1 (mod m), there is an element a of order m (mod p)
because (Z/pZ)∗ is cyclic. Again by the previous exercise p | φm (a).
If there are only finitely many primes p1 , . . . , pr (say) that are congruent
to 1 (mod m), then setting a = (p1 · · · pr )m we examine the prime divisors
of φm (a). Observe that the identity
6
xm − 1 = φd (x)
d|m

implies that φm (0) = ±1. Thus, the constant term of φm (x) is ±1 so that
φm (a) is coprime to a and hence coprime to m. (If φm (a) = ±1, one can
replace a by any suitable power of a, so that |φm (a)| > 1.)
By what we have proved, any prime divisor p of φm (a) coprime to m
must be congruent to 1 (mod m). The prime p is distinct from p1 , . . . , pr .

Exercise 5.6.11 Show that p ! m splits completely in Q(ζm ) if and only if p ≡ 1

(mod m).

Solution. Observe that φm (x) has a root mod p if and only if p ≡ 1

(mod m) by the previous exercise. But then if it has one root a it has ϕ(m)
roots because m | (p − 1) and so (Z/pZ)∗ has a cyclic subgroup of order m.
Thus, φm (x) splits completely if and only if p ≡ 1 (mod m).

Exercise 5.6.12 Let p be prime and let a be squarefree and coprime to p. Set
θ = a1/p and consider K = Q(θ). Show that OK = Z[θ] if and only if ap−1 #≡ 1
(mod p2 ).

Solution. Assume that OK = Z[θ]. We will show that ap−1 !≡ 1 (mod p2 ).

By Theorem 5.5.1,
pOK = ℘p
240 CHAPTER 5. DEDEKIND DOMAINS

since Q(θ) has degree p over Q. Moreover,

℘ = (p, θ − a).

Also, (θ − a) ∈ ℘ and (θ − a) !∈ ℘2 so that

(θ − a) = ℘a

for some ideal a. Taking norms, we find |N (θ − a)| = pN a and (N a, p) = 1.

(θ − a) is a root of (x + a)p − a and this polynomial is irreducible since
Q(θ − a) = Q(θ) has degree p over Q. Hence

N (θ − a) = ap − a = pN a

so that ap !≡ a (mod p2 ).
Conversely, suppose that ap !≡ a (mod p2 ). Then the polynomial

(x + a)p − a
< =
is Eisenstein with respect to the prime p. Therefore p ! OK< : Z[θ − a]=
by Example 4.3.1. But Z[θ − a] = Z[θ] so we deduce that p ! OK : Z[θ] .
In addition, xp − a is Eisenstein with respect
< to every
= prime divisor of a.
Again, by Example 4.3.1, we deduce that OK : Z[θ] is coprime to a. By
Exercises 4.3.3 and 4.2.8,
< =2
dK/Q (θ) = (−1)(2) pp ap−1 = OK : Z[θ] · dK .
p

Since the index of θ in OK is coprime to both p and a, it must equal 1.

Thus OK = Z[θ].

Exercise 5.6.13 Suppose that K = Q(θ) and OK = Z[θ]. Show that if p | dK ,

p ramifies.

Solution. We will use the result of Theorem 5.5.1. Let f (x) be the minimal
polynomial of Z[θ]. Suppose that p | dK , and

f (x) ≡ f1 (x)e1 · · · fg (x)eg (mod p).

2
Since p | dK/Q (θ) = (θi − θj )2 , then θi ≡ θj in Fp for some i != j. Thus,
f has multiple roots in Fp . Hence, one of the ei ’s is greater than 1.

Exercise 5.6.14 Let K = Q(θ) and suppose that p | dK/Q (θ), p2 ! dK/Q (θ).
Show that p | dK and p ramifies in K.
< =
Solution. Recall that dK/Q (θ) = m2 dK where m = OK : Z[θ] . Clearly,
since p | dK/Q (θ) but p2 ! dK/Q (θ), p | dK , and p ! m. We can now apply the
result of Exercise 5.5.2. Using the same argument as in 5.6.13, we deduce
that f (x) has a multiple root mod p and so p ramifies in K.
5.6. SUPPLEMENTARY PROBLEMS 241

Exercise 5.6.15 Let K be an algebraic number field of discriminant √ dK . Show

that the normal closure of K contains a quadratic field of the form Q( dK ).

Solution. Let K̃ be the normal closure of K, ω1 , . . . , ωn an integral basis

of K, σ1 , . . . , σn the distinct embeddings of K into C. Then
# $2
dK = det σi (ωj ) ∈ Z.
√ # $
Thus dK = det σi (ωj ) ∈ K̃.

Exercise 5.6.16 Show that if p ramifies in K, then it ramifies in each of the

conjugate fields of K. Deduce that if p ramifies in the normal closure of K, then
it ramifies in K.

Solution. Since each embedding σi : K → K (i) is an isomorphism of fields,

any factorization of
(p) = ℘e11 · · · ℘egg
(i)
takes each prime ideal ℘j into a conjugate prime ideal ℘j . If some ej > 1
then in each conjugate field, p ramifies. The second part is straightforward
upon intersecting with K.

Exercise 5.6.17 Deduce the following special case of Dedekind’s theorem: if

p2m+1 /dK show that p ramifies in K.
√
Solution. By the penultimate exercise, p ramifies in Q( dK ) and hence
in the normal closure. By the previous exercise, p ramifies in K.

Exercise 5.6.18
√ Determine the prime ideal factorization of (7), (29), and (31)
in K = Q( 3 2).

Solution. By Example 4.3.6, OK = Z[21/3 ]. We may apply Theorem 5.5.1.

Since x3 − 2 is irreducible mod 7, 7OK is prime in OK . Since

x3 − 2 ≡ (x + 3)(x2 − 3x + 9) (mod 29)

and the quadratic factor is irreducible, we get

29OK = ℘1 ℘2

where deg ℘1 = 1, deg ℘2 = 2 and ℘1 , ℘2 are prime ideals. Finally,

x3 − 2 ≡ (x − 4)(x − 7)(x + 11) (mod 31)

so that 31OK splits completely in K.

Exercise 5.6.19 If L/K is a finite extension of algebraic number field, we can

view L as a finite dimensional vector space over K. If α ∈ L, the map v 0→ αv is
a linear mapping and one can define, as before, the relative norm NL/K (α) and
relative trace T rL/K (α) as the determinant and trace, respectively, of this linear
map. If α ∈ OL , show that T rL/K (α) and NL/K (α) lie in OK .
242 CHAPTER 5. DEDEKIND DOMAINS

Solution. By taking a basis ω1 , ..., ωn of L over K and repeating the

argument of Lemma 4.1.1, the result follows immediately.

Exercise 5.6.20 If K ⊆ L ⊆ M are finite extensions of algebraic number fields,

show that NM/K (α) = NL/K (NM/L (α)) and T rM/K (α) = T rL/K (T rM/L (α)) for
any α ∈ M . (We refer to this as the transitivity property of the norm and trace
map, respectively.)

Solution. Fix an algebraic closure M of M . Let σ1 , ..., σm be the distinct

embeddings of L into M which are equal to the identity on K. By field
theory, we can extend these to embeddings of M into M . Of these, let
η1 , ..., ηn be the ones trivial on L. If σ is an arbitrary embedding of M into
M which is trivial on K, then as σ is also an embedding of L into M , it
must be σj for some j. Thus, σj−1 σ fixes L and so must be an ηi for some
i. Thus, every embedding of M is of the form σj ◦ ηi so that
6 6
NM/K (α) = σj (ηi (α)) = σj (NM/L (α)) = NL/K (NM/L (α)),
i,j j

as desired.

Exercise 5.6.21 Let L/K be a finite extension of algebraic number fields. Show
that the map
T rL/K : L × L → K
is non-degenerate.

Solution. This follows from an argument analogous to the proof of Lemma

4.1.4.

Exercise 5.6.22 Let L/K be a finite extension of algebraic number fields. Let
a be a finitely generated OK -module contained in L. The set
−1
DL/K (a) = {x ∈ L : T rL/K (xa) ⊆ OK }

is called codifferent of a over K. If a #= 0, show that DL/K

−1
(a) is a finitely generated
OK -module. Thus, it is a fractional ideal of L.
−1
Solution. The fact that DL/K (a) is an OK -module is clear. To see that it
is finitely generated, we take an OK -basis of a and repeat the argument in
Exercise 5.4.1 to deduce the result.

Exercise 5.6.23 If in the previous exercise a is an ideal of OL , show that the

fractional ideal inverse, denoted DL/K (a) of DL/K
−1
(a) is an integral ideal of OL .
(We call DL/K (a) the different of a over K. In the case a is OL , we call it the
relative different of L/K and denote it by DL/K .)
−1 −1
Solution. We have DL/K (a)DL/K (a) = OL and 1 ∈ DL/K (a) so that
−1
DL/K (a) ⊆ DL/K (a)DL/K (a) ⊆ OL so that DL/K (a) is an integral ideal of
OL .
5.6. SUPPLEMENTARY PROBLEMS 243

Exercise 5.6.24 Let K ⊆ L ⊆ M be algebraic number fields of finite degree

over the rationals. Show that

DM/K = DM/L (DL/K OM ).

−1
Solution. We have x ∈ DM/L if and only if T rM/L (xOM ) ⊆ OL which is
equivalent to
−1 −1 −1
DL/K T rM/L (xOM ) ⊆ DL/K OL iff T rL/K (DL/K T rM/L (xOM )) ⊆ OK

−1
which by transitivity of the trace is equivalent to T rM/K (xDL/K ) ⊆ OK .
That is, we must have
−1 −1
xDL/K ⊆ DM/K

which is true if and only if

−1
x ∈ DL/K DM/K ,

which means
−1 −1
DM/L = DL/K DM/K ,

which gives the required result.

Exercise 5.6.25 Let L/K be a finite extension of algebraic number fields. We

define the relative discriminant of L/K, denoted dL/K as NL/K (DL/K ). This is
an integral ideal of OK . If K ⊆ L ⊆ M are as in Exercise 5.6.24, show that
[M :L]
dM/K = dL/K NL/K (dM/L ).

Solution. By the transitivity of the norm map and by Exercise 5.6.24, we

have
NM/K (DM/K ) = NM/K (DM/L DL/K ) =
[M :L]
NL/K (NM/L (DL/K DM/L )) = NL/K (DL/K NM/L (DM/L ))

which gives the result. We remark here that Dedekind’s theorem concerning
ramification extends to relative extensions L/K. More precisely, a prime
ideal p of OK is said to ramify in L if there is a prime ideal ℘ of OL such
that ℘2 |pOL . One can show that p ramifies in L if and only if p|dL/K . The
easy part of this assertion that if p is ramified then p|dL/K can be proved
following the argument of Exercise 5.4.5. The converse requires further
theory of relative differents. We refer the interested reader to [N].

Exercise 5.6.26 Let L/K be a finite extension of algebraic number fields. Sup-
pose that OL = OK [α] for some α ∈ L. If f (x) is the minimal polynomial of α
over OK , show that DL/K = (f ( (α)).
244 CHAPTER 5. DEDEKIND DOMAINS

Solution. This result is identical to Exercises 5.6.6 and 5.6.7. More gen-
erally, one can show the following. For each θ ∈ OL which generates L over
K, let f (x) be its minimal polynomial over OK . Define δL/K (θ) = f $ (θ).
Then DL/K is the ideal generated by the elements δL/K (θ) as θ ranges over
such elements. We refer the interested reader to [N].

Exercise 5.6.27 Let K1 , K2 be algebraic number fields of finite degree over K.

If L/K is the compositum of K1 /K and K2 /K, show that the set of prime ideals
dividing dL/K and dK1 /K dK2 /K are the same.

Solution. By Exercise 5.6.25, we see that every prime ideal dividing

dK1 /K dK2 /K

also divides dL/K . Suppose now that p is a prime ideal of OK which divides
dL/K but not dK1 /K . We have to show that p divides dK2 /K . By the defini-
tion of the relative discriminant, there is a prime ideal ℘ of OL lying above
p which divides the different DL/K . This ideal cannot divide DK1 /K OL
for this would imply that p divides dK1 /K , contrary to assumption. Since
DL/K = DL/K1 DK1 /K , we deduce that ℘ divides DL/K1 . Now let α ∈ OK2
so that α generates K2 over K. Let f (x) be its minimal polynomial over
K1 and g(x) its minimal polynomial over K. (We have assumed that we
have fixed a common algebraic closure which contains K1 and K2 .) Then
L = K1 (α) and g(x) = f (x)h(x) for some polynomial h over K1 . Hence,
g $ (x) = f $ (x)h(x)+f (x)h$ (x) which implies g $ (α) = f $ (α)h(α). Thus, g $ (α)
is in the ideal generated by f $ (α). By the remark in the solution of Exercise
5.6.26, we deduce that f $ (α) ∈ DL/K ⊆ ℘. Therefore, g $ (α) ∈ ℘. The same
remark enables us to deduce that g $ (α) ∈ DK2 /K implying that p divides
dK2 /K .

Exercise 5.6.28 Let L/K be a finite extension of algebraic number fields. If L̃

denotes the normal closure, show that a prime p of OK is unramified in L if and
only if it is unramified in L̃.

Solution. If we apply the preceding exercise to the compositum of the

conjugate fields of L, the result is immediate.
Chapter 6

The Ideal Class Group

6.1 Elementary Results

Exercise 6.1.2 Show that given α, β ∈ OK , there exist t ∈ Z, |t| ≤ HK , and
w ∈ OK so that |N (αt − βw)| < |N (β)|.

Solution. If we apply Lemma 6.1.1 with α replaced by α/β, we conclude

that there exist t ∈ Z, |t| ≤ HK , and w ∈ OK such that
|N (tα/β − w)| < 1.
This implies |N (tα − wβ)| < |N (β)|.

6.2 Finiteness of the Ideal Class Group

Exercise 6.2.1 Show that the relation ∼ defined above is an equivalence rela-
tion.

Solution. It is trivial that A ∼ A, and if A ∼ B then B ∼ A, for any

ideals A and B. Suppose now that A ∼ B, and B ∼ C. That is, there exist
α, β, γ, θ ∈ OK such that (α)A = (β)B, and (γ)B = (θ)C. It is now easily
seen that (αγ)A = (βθ)C. Thus, A ∼ B and B ∼ C imply A ∼ C.
Hence, ∼ is an equivalence relation.

Exercise 6.2.3 Show that each equivalence class of ideals has an integral ideal
representative.

Solution. Suppose A is a fractional ideal in K. Let A = b/c, with b, c ⊆

OK .
We know from Exercise 4.4.1 that c ∩ Z != {0}, so there exists 0 != t ∈ Z
such that t ∈ c. Thus, c ⊇ (t) = tOK , and so c divides (t). This implies
that there exists an integral ideal e ⊆ OK such that
ce = (t). (6.1)

245
246 CHAPTER 6. THE IDEAL CLASS GROUP

We now have
b ceb
(t)A = (t) = = eb ⊆ OK .
c c
Thus, A ∼ be ⊆ OK , and the result is proved.

Exercise 6.2.4 Prove that for any integer x > 0, the number of integral ideals
a ⊆ OK for which N (a) ≤ x is finite.

Solution. Since the norm is multiplicative and takes values > 1 on prime
ideals, and since integral ideals have unique factorization, it is sufficient to
prove that there are only a finite number of prime ideals ℘ with N (℘) ≤ x.
Now, any prime ℘ contains exactly one prime p ∈ Z, as shown in Exer-
cise 4.4.4. Thus, ℘ occurs in the factorization of (p) ⊆ OK into prime ideals.
Since N (℘) ≥ 2, we have N (℘) = pt for some t ≥ 1. This implies 2 there are
s
at most n possibilities for such2s ℘, since the factorization (p) = ai
i=1 ℘i
implies that p = N ((p)) = i=1 N (℘i ) leading to s ≤ n. Moreover,
n ai

p ≤ N (℘) ≤ x. This proves the exercise.

Exercise 6.2.6 Show that the product defined above is well-defined, and that H
together with this product form a group, of which the equivalence class containing
the principal ideals is the identity element.

Solution. To show that the product defined above is well-defined we only

need to show that if A1 ∼ B1 and A2 ∼ B2 , then A1 A2 ∼ B1 B2 . Indeed, by
definition, there exist α1 , α2 , β1 , β2 ∈ OK such that (α1 )A1 = (β1 )B1 and
(α2 )A2 = (β2 )B2 . Therefore

(α1 α2 )A1 A2 = (β1 β2 )B1 B2 .

Thus, A1 A2 ∼ B1 B2 .
Now, it is easy to check that H with the product defined above is closed,
associative, commutative, and has the class of principal ideals as the iden-
tity element. Thus, to finish the exercise, we need to show that each element
of H does have an inverse. Suppose C is an arbitrary element of H. Let
a ⊆ OK be a representative of C (we showed in Exercise 6.2.3 that every
equivalence class of ideals contains an integral representative). If we pro-
ceed as we did when deriving equation (6.1), we conclude that there exists
an integral ideal b such that ab is principal. It then follows immediately
that the class containing b is the inverse of C.

Exercise 6.2.7 Show that the constant CK in Theorem 6.2.2 could be taken to
be the greatest integer less than or equal to HK , the Hurwitz constant.

Solution. As in Lemma 6.1.1, let {ω1 , ω2 , . . . , ωn } be an integral basis of

OK . Let C be a given class of ideals. We denote by C −1 the inverse class of
6.3. DIOPHANTINE EQUATIONS 247

C in H. Let a be an integral representative of C −1 . Consider the following

set
; n
G
*
1/n
S = s ∈ OK |s = mi ωi , mi ∈ Z, 0 ≤ mi < (N (a)) +1 .
i=1

Then |S| ≥ N (a) + 1. Since N (a) = [OK : a], we can find distinct a, b ∈ S
such that a ≡ b (mod a). Thus, (a − b) ⊆ a. This implies that there exists
an integral ideal b such that
5n(a − b) = ab. It is easy to observe that b ∈ C.
We may write a − b = i=1 pi ωi . Since a, b ∈ S, |pi | ≤ (N (a))1/n + 1,
and so we have
1 n ! n "1
16 * (j) 11
|N (a − b)| = 1 1 pi ω i 1
j=1 i=1
n
/ n 0
6 * (j)
≤ |pi ||ωi |
j=1 i=1
n
/ n 0
< =n 6 * (j)
≤ (N (a))1/n + 1 |ωi |
j=1 i=1
< =n
≤ (N (a))1/n + 1 HK .

We also know that since (a − b) = ab, |N (a − b)| = N (a)N (b). Thus,

< =n
N (b) ≤ 1 + (N (a))−1/n HK .

However, observe that we can always replace a by the ideal ca, in the same
equivalence class, for< any c ∈ OK \{0},
=n and with |N (c)| arbitrarily large; we
can therefore make 1 + (N (a))−1/n arbitrarily close to 1.
Thus, every equivalence class C has an integral representative b with
N (b) ≤ HK . This implies that every ideal is equivalent to another integral
ideal with norm less than or equal to HK .

6.3 Diophantine Equations

Exercise 6.3.2 Let k > 0 be a squarefree positive integer. Suppose that k ≡ 1, 2
(mod 4), and k does not have the form k = 3a2 ± 1 for an integer a. Consider
the equation

x2 + k = y 3 . (6.4)
√
Show that if 3 does not divide the class number of Q( −k), then this equation
has no integral solution.

Solution. Similar to what was done in Example 6.3.1, y must be odd

(consider congruences modulo 4). Also, if a prime p | (x, y), then p | k;
248 CHAPTER 6. THE IDEAL CLASS GROUP

and hence, since k is squarefree (so, in particular, k is not divisible by

p2 ), by dividing both sides of the equation (6.4) by p, we end up having a
contradiction modulo p. Thus, x and y are coprime.
Suppose now that (x, y) is an integral solution to equation (6.4). As
given, k√≡ 1, 2 (mod √ 4), so −k ≡ 3, 2 (mod 4). Thus, the integers in
K = Q( −k) are Z[ −k]. We consider the factorization
√ √
(x + −k)(x − −k) = y 3 , (6.5)
√
in the ring of integers Z[ −k].
As√in Example 6.3.1,
√ suppose a prime ℘ divides the gcd of the ideals
(x + −k) and (x − −k) (which implies ℘ divides (y)). Then ℘ divides
(2x). Since y is odd, ℘ does not divide (2). Thus, ℘ divides √ (x). This
contradicts
√ the fact that x and y are coprime. Hence, (x + −k) and
(x − −k) are coprime. Equation (6.5) now implies that
√ √
(x + −k) = a3 and (x − −k) = b3 ,

for some ideals a and b.

Let h(K) be the class number of the field K, then ch(K) is principal for
any ideal c. As given, 3 ! h(K), so (3, h(K)) = 1. Thus, since a3 and b3 are
principal, a and b are also principal. We must have
√ √
(x + −k) = ε(a + b −k)3 , (6.6)
√
for some integers a,√
b, and a unit ε ∈ Z[ −k].√
Let ε = x1 + x2 −k. Then, since α ∈ Z[ −k] is a unit if and only if
N (α) = ±1, we have

x21 + kx22 = ±1. (6.7)

As given k > 0 and k is square-free, so k > 1. Thus, equation (6.7) implies

x2 = 0 and x1 = ±1. Hence, ε = ±1, and in equation (6.6) it could be
absorbed into the cube. We have
√ √
(x + −k) = (a + b −k)3 .

This implies 1 = b(3a2 − kb2 ). It is clear that b | 1, so b = ±1. Both cases

lead to either k = 3a2 + 1 or k = 3a2 − 1, which violates the hypothesis.
Hence, we conclude that equation (6.4) does not have an integral solu-
tion.

6.4 Exponents of Ideal Class Groups

Exercise 6.4.1 Fix a positive integer g > 1. Suppose that n is odd, greater
√ than
1 and ng − 1 = d is squarefree. Show that the ideal class group of Q( −d) has
an element of order g.
6.4. EXPONENTS OF IDEAL CLASS GROUPS 249

Solution. Since
√ d is even√ and squarefree, d ≡ 2 (mod 4). The ring of
integers of Q( −d) is Z[ −d]. We have the ideal factorization:
√ √
(n)g = (ng ) = (1 + d) = (1 + −d)(1 − −d).
√ √
The ideals (1 + −d) and (1 − −d) are √ coprime since
√ n is odd. Thus
by Theorem 5.3.13, each of the ideals (1 + −d), (1 − −d) must be gth
powers. Thus
√
ag = (1 + −d),
√
(a$ )g = (1 − −d),

with aa$ = (n). Hence a has √ order dividing g in the class group.
Suppose am = (u + v −d) for some u, v ∈ Z. Note that v cannot
be zero for otherwise am = (u) implies that (a$ )m = (u) so that (u) =
gcd(am , (a$ )m ), contrary to gcd(a, a$ ) = 1. Therefore
√ v != 0.
Now take norms of the equation am = (u + v −d) to obtain

nm = u2 + v 2 d ≥ d = ng − 1.

If m ≤ g − 1, we get ng−1 ≥ ng − 1 which implies that 1 ≥ ng−1 (n − 1) ≥ 2,

a contradiction. √
Therefore ag = (1 + −d) and am is not principal for any √ m < g. Thus
there is an element of order g in the ideal class group of Q( −d).

Exercise 6.4.2 Let g be odd and greater than 1.√If d = 3g − x2 is squarefree

with x odd and satisfying x2 < 3g /2, show that Q( −d) has an element of order
g in the class group.
√
Solution.
√ Observe that d ≡ 2 (mod 4) so the ring of integers of Q( −d)
is Z[ −d]. The factorization
√ √
3g = (x + −d)(x − −d)
√ √ √
shows that 3 splits in Q( −d), as the ideals (x + −d) and (x − −d) are
coprime. Thus
(3) = ℘1 ℘$1 .
We must have √
(x + −d) = ℘g1 .
Therefore, the√ order of ℘m1 in the ideal class group is a divisor of g. If
2 2
℘m
1 = (u + v −d), then 3 = u + v d. If v != 0, we deduce 3m ≥ d > 3g /2
which is a contradiction if m ≤ g − 1. Either ℘1 has order g or v = 0. In
the latter case, we get u2 = 3m , a contradiction since m is odd.

Exercise 6.4.3 Let g be odd. Let N be the number of squarefree integers of

the form 3g − x2 , x odd, 0 < x2 < 3g /2. For g sufficiently large, show that
N ( 3g/2 . Deduce that there are infinitely many imaginary quadratic fields
whose class number is divisible by g.
250 CHAPTER 6. THE IDEAL CLASS GROUP

Solution. The number of integers under consideration is

1
√ 3g/2 + O(1).
2 2
From these, we will remove any number divisible by the square of a prime.
Since g is odd, x2 ≡ 3g (mod 4) implies that x2 ≡ −1 (mod 4) has a
solution. This is a contradiction. Therefore 4 ! 3g − x2 . If 3 | x, then
3 | 3g − x2 so we remove such numbers. Their count is
1
√ 3g/2 + O(1).
6 2
If p is odd and greater than 3, the number of 3g − x2 divisible by p2 is at
most
3g/2
√ + O(1).
p2 2
Thus,  

 ! " 

3g/2 1 1 * 1 3 g/2 
N≥ √ − − +O
2  2 6 p2 g 

 p2 <3g 
p≥5

by using Exercise 1.1.26. Now

* 1 ∞
* 1 # $ # $
≤ = 14 − 15 + 15 − 61 + · · ·
p2 n=5
n(n − 1)
p≥5
1
= 4.

Since 12 − 14 − 61 = 12
1
, we see N > 3g/2 . By the previous exercise, each of
these values gives rise to a distinct quadratic field whose class group has
exponent divisible by g. By applying this result for powers of g we deduce
that there are infinitely many imaginary quadratic fields of class number
divisible by g.
(This argument is due to Ankeny and Chowla.)

6.5 Supplementary Problems

√
Exercise 6.5.1 Show that the class number of K = Q( −19) is 1.
√
Solution. We know that 1, (1 + −19)/2 forms an integral basis. We then
write
√
(1) (1) 1 + −19
ω1 = 1, ω2 = ,
2
√
(2) (2) 1 − −19
ω1 = 1, ω2 = ,
2
6.5. SUPPLEMENTARY PROBLEMS 251

and use this to find the Hurwitz constant

2
/ 2 0
6 * (j)
HK = |ωi |
j=1 i=1
! 1 √ 1" ! 1 √ 1"
1 1 + −19 1 1 1 − −19 1
= 1
1+1 1 1
1+1 1
2 1 2 1
= 13.53 · · · .

Just as in Example 6.2.8, we examine all the primes p ≤ 13 to determine

the prime ideals with N (℘) ≤√13. The primes in question are 2, 3, 5, 7, 11,
and 13. They factor in Z[(1 + −19)/2] as follows: 2, 3, and 13 stay prime,
and
! √ "! √ "
1 + −19 1 − −19
5 = ,
2 2
! √ "! √ "
3 + −19 3 − −19
7 = ,
2 2
! √ "! √ "
5 + −19 5 − −19
11 = .
2 2

These are all principal ideals√and thus are all equivalent. This shows that
the class number of K = Q( −19) is 1.

Exercise 6.5.2 (Siegel) Let C be a symmetric, bounded domain in Rn . (That

is, C is bounded and if x ∈ C so is −x.) If vol(C) > 1, then there are two distinct
points P, Q ∈ C such that P − Q is a lattice point.

Solution. Let ϕ(x) = 1 or 0 according as x ∈ C or not. Then set

*
ψ(x) = ϕ(x + γ).
γ∈Zn

Clearly, ψ(x) is bounded and integrable. thus

@ @ *
ψ(x) dx = ϕ(x + γ) dx
Rn /Zn Rn /Zn γ∈Zn
* @
= ϕ(x + γ) dx
γ∈Zn Rn /Zn
* @
= ϕ(x) dx
γ∈Zn γ+Rn /Zn
@
= ϕ(x) dx
Rn
= vol(C) > 1.
252 CHAPTER 6. THE IDEAL CLASS GROUP

Since ψ(x) takes only integer values, we must have ψ(x) ≥ 2 for some x.
Therefore, there are two distinct points P +γ, P +γ $ in C so their difference
is a lattice point.

Exercise 6.5.3 If C is any convex, bounded, symmetric domain of volume > 2n ,

show that C contains a non-zero lattice point. (C is said to be convex if x, y ∈ C
implies λx + (1 − λ)y ∈ C for 0 ≤ λ ≤ 1.)

Solution. By the previous question, the bounded symmetric domain 12 C

contains two distinct points 12 P and 12 Q such that 12 P − 12 Q is a lattice
point, because
# $ vol(C)
vol 12 C = > 1.
2n
Since C is convex,
0 != γ = 12 P − 12 Q ∈ C

as P, Q ∈ C. This is a nonzero lattice point in C.

Exercise 6.5.4 Show in the previous question if the volume ≥ 2n , the result is
still valid, if C is closed.

Solution. We can enlarge our domain by ε to create Cε of volume > 2n .

For each ε, Cε contains a lattice point. Since

lim Cε = C,
ε→0

C also contains a lattice point (perhaps on the boundary).

Exercise 6.5.5 Show that there exist bounded, symmetric convex domains with
volume < 2n that do not contain a lattice point.

Solution. Consider −1 < xi < 1, 1 ≤ i ≤ n. This hypercube has volume

2n and the only lattice point it contains is 0.

Exercise 6.5.6 (Minkowski) For x = (x1 , . . . , xn ), let

n
,
Li (x) = aij xj , 1 ≤ i ≤ n,
j=1

be n linear forms with real coefficients. Let C be the domain defined by

|Li (x)| ≤ λi , 1 ≤ i ≤ n.

Show that if λ1 · · · λn ≥ |det A| where A = (aij ), then C contains a nonzero

lattice point.
6.5. SUPPLEMENTARY PROBLEMS 253

Solution. Clearly C is convex, bounded, and symmetric. We want to

compute the volume of C/2,
@ @
··· dx1 · · · dxn .
|Li (x)|≤λi /2
1≤i≤n

We make a linear change of variables: y = Ax. Then

dy1 · · · dyn = (det A) dx1 · · · dxn ,

so we get
λ1 · · · λn
vol(C/2) = ≥1
|det A|
from which the result follows because C is closed.

Exercise 6.5.7 Suppose that among the n linear forms above, Li (x), 1 ≤ i ≤ r1
are real (i.e., aij ∈ R), and 2r2 are not real (i.e., some aij may be nonreal).
Further assume that

Lr1 +r2 +j = Lr1 +j , 1 ≤ j ≤ r2 .

That is,
n
,
Lr1 +r2 +j (x) = ar1 +j,k xk , 1 ≤ j ≤ r2 .
k=1

Now let C be the convex, bounded symmetric domain defined by

|Li (x)| ≤ λi , 1 ≤ i ≤ n,

with λr1 +j = λr1 +r2 +j , 1 ≤ j ≤ r2 . Show that if λ1 · · · λn ≥ |det A|, then C

contains a nonzero lattice point.

Solution. We replace the nonreal linear forms by real ones and apply the
previous result. Set

Lr1 +j + Lr1 +r2 +j

L$r1 +j =
2
and
Lr1 +j − Lr1 +r2 +j
L$$r1 +j = .
2
Then L$r1 +j , L$$r1 +j are linear forms. Clearly, if

λr1 +j
|L$r1 +j | ≤ √ ,
2
λr1 +j
|L$$r1 +j | ≤ √ ,
2
254 CHAPTER 6. THE IDEAL CLASS GROUP

then
|Lr1 +j | ≤ λr1 +j ,
so we replace |Lr1 +j | ≤ λr1 +j , |Lr1 +r2 +j | ≤ λr1 +j with L$r1 +j and L$$r1 +j
satisfying the inequalities above. We deduce by the results established in
the previous questions, that this domain contains a nonzero lattice point
provided
λ1 · · · λn 2−r2
> 1,
|det A$ |
where A$ is the appropriately modified matrix. A simple linear algebra
computation shows det A$ = 2−r2 det A.

Exercise 6.5.8 Using the previous result, deduce that if K is an algebraic num-
ber field
8with discriminant dK , then every ideal class contains an ideal b satisfying
N b ≤ |dK |.

Solution. Let a be any integral ideal and α1 , . . . , αn an integral basis of

a. Consider the linear forms
n
* (i)
Li (x) = αj xj
j=1

and the bounded symmetric convex domain defined by

|Li (x)| ≤ |∆|1/n ,

(i)
where |∆| = |det(αj )|. By the previous question, the system has a non-
trivial integral solution, (x1 , . . . , xn ). Let

ω = x1 α1 + · · · + xn αn ∈ a.

Then (ω) ⊆ a so that for some ideal b, ab = (ω). But

|N (ω)| = |N a||N b| ≤ |∆|

by construction. Also,
|∆|2 = (N a)2 |dK |
L
by Exercise 4.4.5. Hence, |N b| ≤ |dK |. Given any ideal a we have L found
an ideal b in the inverse class whose norm is less than or equal to |dK |.

Exercise 6.5.9 Let Xt consist of points

(x1 , . . . , xr , y1 , z1 , . . . , ys , zs )

in Rr+2s where the coordinates satisfy

8 8
|x1 | + · · · + |xr | + 2 y12 + z12 + · · · + 2 ys2 + zs2 < t.

Show that Xt is a bounded, convex, symmetric domain.

6.5. SUPPLEMENTARY PROBLEMS 255

Solution. The fact that Xt is bounded and symmetric is clear. To see

convexity, let
P = (a1 , . . . , ar , b1 , c1 , . . . , bs , cs )
and
Q = (d1 , . . . , dr , e1 , f1 , . . . , es , fs )
be points of Xt . We must show that λP + µQ ∈ Xt whenever λ, µ ≥ 0 and
λ + µ = 1. Clearly
|λai + µdi | ≤ λ|ai | + µ|di |.
Also
L L L
(λbi + µei )2 + (λci + µfi )2 ≤ λ b2i + c2i + µ e2i + fi2 ,
as is easily verified. From these inequalities, it follows that λP + µQ ∈ Xt
so that Xt is convex.

Exercise 6.5.10 In the previous question, show that the volume of Xt is

2r−s π s tn
,
n!
where n = r + 2s.

Solution. We begin by making a change of variables to polar coordinates:

2yj = ρj cos θj , 2xj = ρj sin θj , 4 dyj dzj = ρj dρj dθj so that integrating
over xi ≥ 0 for 1 ≤ i ≤ r gives
@
vol(Xt ) = 2 · 2r −2s
ρ1 · · · ρs dx1 · · · dxr dρ1 · · · dρs dθ1 · · · dθs
@
= 2r 2−2s (2π)s ρ1 · · · ρs dx1 · · · dxr dρ1 · · · dρs ,
Yt

where
Yt = {(x1 , . . . , xr , ρ1 , . . . , ρs ) : xi , ρj ≥ 0, x1 + · · · + xr + ρ1 + · · · + ρs ≤ t}.
Let fr,s (t) denote the value of the above integral. By changing variables, it
is clear that
@ 1
fr,s (1) = fr−1,s (1 − x1 ) dx1
0
@ 1
= fr−1,s (1) xr−1+2s
1 dx1
0
1
= fr−1,s (1).
r + 2s
Proceeding inductively, we get
(2s)!
fr,s (1) = f0,s (1).
(r + 2s)!
256 CHAPTER 6. THE IDEAL CLASS GROUP

Now we evaluate f0,s (1) by integrating with respect to ρs first:

@ 1
f0,s (1) = ρs f0,s−1 (1 − ρs ) dρs
0
@ 1
= f0,s−1 (1) ρs (1 − ρs )2s−2 dρs
0
f0,s−1 (1)
= .
2s(2s − 1)

Again, proceeding inductively, we find f0,s (1) = 1/(2s)! so that

1 1
fr,s (1) = = .
(r + 2s)! n!
This completes the proof.

Exercise 6.5.11 Let C be a bounded, symmetric, convex domain in Rn . Let

a1 , . . . , an be linearly independent vectors in Rn . Let A be the n × n matrix
whose rows are the ai ’s. If

vol(C) > 2n |det A|,

show that there exist rational integers x1 , . . . , xn (not all zero) such that

x1 a1 + · · · + xn an ∈ C.

Solution. Consider the set D of all (x1 , . . . , xn ) ∈ Rn such that

x1 a1 + · · · + xn an ∈ C.

It is easily seen that D is bounded, symmetric, and convex because C is.

Moreover, D = A−1 C so that by linear algebra,

vol(D) = vol(C)(|det A|)−1 .

Thus, if vol(D) > 2n , then D contains a lattice point (x1 , . . . , xn ) != 0 such

that x1 a1 + · · · + xn an ∈ C. But vol(D) > 2n is equivalent to

vol(C) > 2n |det A|,

as desired.

Exercise 6.5.12 (Minkowski’s Bound) Let K be an algebraic number field

of degree n over Q. Show that each ideal class contains an ideal a satisfying
$ %r
n! 4 2
Na ≤ n |dK |1/2 ,
n π
where r2 is the number of pairs of complex embeddings of K, and dK is the
discriminant.
6.5. SUPPLEMENTARY PROBLEMS 257

Solution. Given any ideal b, let ω1 , . . . , ωn be a basis of b. Let

#
ai = σ1 (ωi ), . . . , σr1 (ωi ), Re(σr1 +1 (ωi )),
$
Im(σr1 +1 (ωi )), . . . , Re(σr1 +r2 (ωi )), Im(σr1 +r2 (ωi )) ∈ Rn .

Then the ai are linearly independent vectors in Rn . Consider the bounded

symmetric convex domain Xt defined in Exercise 6.5.9 above (with r =
r1 , s = r2 ). By Exercise 6.5.10 above, the volume of Xt is

2r1 −r2 π r2 tn
.
n!
If t is chosen so that this volume is greater than 2n |det A|, then Xt contains
a lattice point (x1 , . . . , xn ) so that

0 != x1 a1 + · · · + xn an ∈ Xt .

Let us set α = x1 ω1 + · · · + xn ωn ∈ b. By the arithmetic mean – geometric

mean inequality, we find
t
|N (α)|1/n <
n
so that ! "n
t
|N (α)| < .
n
Moreover, det A = 2−r2 |N (b)||dK |1/2 and
! "r 2
tn 2n |det A| 4
= r1 −r2 r2 = |N b||dK |1/2 .
n! 2 π π

Thus, there is an α ∈ b, α != 0 such that

! "r
n! 4 2
|N (α)| ≤ n |N b||dK |1/2 .
n π

Write (α) = ab for some ideal a. Then

! "r
n! 4 2
Na ≤ n |dK |1/2 ,
n π

as desired.

Exercise 6.5.13 Show that if K #= Q, then |dK | > 1. Thus, by Dedekind’s

theorem, in any nontrivial extension of K, some prime ramifies.

Solution. Since N a ≥ 1, we have by the Minkowski bound,

! "r
n! 4 2
1≤ n |dK |1/2
n π
258 CHAPTER 6. THE IDEAL CLASS GROUP

so that
nn E π Fr 2 nn E π Fn/2
|dK |1/2 ≥ ≥ = cn ,
n! 4 n! 4
say. Then
cn+1 E π F1/2 ! 1
"n
= 1+ ,
cn 4 n
which is greater than 1 for every positive n. Hence cn+1 > cn . We have
c2 > 1 so that |dK | > 1, if n ≥ 2.

Exercise 6.5.14 If K and L are algebraic number fields such that dK and dL
are coprime, show that K ∩ L = Q. Deduce that

[KL : Q] = [K : Q][L : Q].

Solution. If M = K ∩ L, then by a result of Chapter 4, dM | dK and

dM | dL . Since dK and dL are coprime, dM = 1. But then, by the previous
exercise, M = Q. We have

[KL : Q] = [KL : K][K : Q].

Let L = Q(θ) and g its minimal polynomial over Q. If [KL : K] < [L : Q],
then the minimal polynomial h of θ over K divides g and has degree smaller
than that of g. Thus the coefficients of h generate a proper extension T
(say) of Q which is necessarily contained in K. Hence, dT |dK . If we let L̃
be the normal closure of L over Q, then h ∈ L̃[x]. We now need to use the
fact that primes which ramify in L are the same as the ones that ramify in
L̃ (see Exercise 5.6.28). Since T is contained in L̃, we see that dT |dL̃ and
by the quoted fact, we deduce that dL and dK have a common prime factor
if dT > 1, which is contrary to hypothesis. Thus, dT = 1 and by 6.5.13 we
deduce T = Q, a contradiction.
√
Exercise 6.5.15 Using Minkowski’s bound, show that Q( 5) has class number
1.
√
Solution. The discriminant of Q( 5) is 5 and the Minkowski bound is
√
2! √ 5
5= = 1.11 . . . .
22 2
√
The only ideal of norm less than 5/2 is the trivial ideal which is principal.
√
Exercise 6.5.16 Using Minkowski’s bound, show that Q( −5) has class number
2.
√
Solution. The discriminant of Q( −5) is −20 and the Minkowski bound
is
2√ 4
20 = (2.236 . . . ) = 2.84 . . . .
π π
6.5. SUPPLEMENTARY PROBLEMS 259

We need to look at ideals of norm 2.√There is only one ideal of norm 2 and
by Exercise 5.2.5 we know that Z[ −5] is not a principal ideal domain.
Hence the class number must be 2.
√ √
Exercise
√ 6.5.17 Compute the class numbers of the fields Q( 2), Q( 3), and
Q( 13).

Solution. The discriminants of these fields are 8, 12, and 13 respectively.

The Minkowski bound is
1
L √
2 |dK | < 12 13 = 1.802 . . . .

The only ideal of norm less than 1.8 is the trivial ideal, which is principal, so
the class number is 1. (Recall that
√ in Exercises √ 2.4.5 and 2.5.4 we showed
that the ring of integers of Q( 2) and Q( 3) are Euclidean and hence
PIDs. So that the class number is 1 for each of these was already known
to us from Chapter 2.)
√
Exercise 6.5.18 Compute the class number of Q( 17).
√
Solution. The discriminant of Q( 17) is 17 and the Minkowski bound is
1
√
2 17 = 2.06 . . . .

We need to consider ideals of norm 2. Since

√ √
9 − 17 3 − 17 3 + 17
−2 = = · ,
4 2 2
√ √
2 splits and the principal ideals ((3 + 17)/2) and ((3 − 17)/2) are the
only ones of norm 2. Therefore, the class number is 1.
√
Exercise 6.5.19 Compute the class number of Q( 6).

Solution. The discriminant is 24 and the Minkowski bound is

1
√ √
2 24 = 6 = 2.44 . . . ,
√
2 ramifies in Q( 6). Moreover,
√ √
−2 = (2 − 6)(2 + 6)
√ √ √
so that the ideal (2 − 6) is the only one of norm 2 since (2 + 6)/(2 − 6)
is a unit. Thus, the class number is 1.
√ √ √ √
Exercise 6.5.20 Show that the fields Q( −1), Q( −2), Q( −3), and Q( −7)
each have class number 1.
260 CHAPTER 6. THE IDEAL CLASS GROUP

Solution. The Minkowski bound for an imaginary quadratic field K is

2L
|dK |.
π
The given fields have discriminants equal to −4, −8, −3, −7, respectively.
Since √
2√ 4 2
8= = 1.80 . . . ,
π π
we deduce that every ideal is principal.

Exercise 6.5.21 Let K be an algebraic number field of degree n over Q. Prove

that
- π .n $ nn %2
|dK | ≥ .
4 n!
Solution. This follows directly from Minkowski’s bound.

Exercise 6.5.22 Show that |dK | → ∞ as n → ∞ in the preceding question.

Solution. From integral calculus,

log n! = n log n − n + O(log n),
so that E πF
log |dK | ≥ 2 + log n + O(log n).
4
Exercise 6.5.23 (Hermite) Show that there are only finitely many algebraic
number fields with a given discriminant.

Solution. We give a brief hint of the proof. From the preceding question,
the degree n of K is bounded. By Minkowski’s theorem, we can find an
element α != 0 in OK so that
L
|α(1) | ≤ |dK |, |α(i) | < 1, i = 2, . . . , r.
We must show α generates K, but this is not difficult. With these inequal-
ities, the coefficients of the minimal polynomial of α are bounded. Since
the coefficients are integers, there are only finitely many such polynomials.

Exercise 6.5.24 Let p be a prime ≡ 11 (mod 12). If p > 3n , show that the
√
ideal class group of Q( −p) has an element of order greater than n.
2
Solution. Since
√ −p ≡ 1 (mod 3), x ≡ −p (mod 3) has a solution and so
3 splits in Q( −3). Write
(3) = ℘1 ℘$1 .
We claim the order of ℘1 in the ideal class group is at least n. If not, ℘m
1
√
is principal and equals (u + v −p) (say) for some m < n. Taking norms,
we see
3m ≡ u2 + pv 2
6.5. SUPPLEMENTARY PROBLEMS 261

has a solution. If v != 0, this is a contradiction since p > 3n . If v = 0, we

get m is even, so that
m/2
1 = (u) = (3
℘m ) = ℘1 (℘$1 )m/2 ,
m/2

contradicting unique factorization.

Exercise 6.5.25 Let K = Q(α) where α is a root of the polynomial f (x) =

x5 − x + 1. Prove that Q(α) has class number 1.

Solution. Since f (x) is a polynomial of the type described in Exer-

cise 4.5.4, we deduce immediately that dK/Q (α) = 55 −44 = 2869 = 19·151.
Since dK/Q (α) is squarefree, Exercise 4.5.26 tells us that OK = Z[α] and
dK = 2869. A quick look at the graph of f (x) = x5 − x + 1 shows that
r1 = 1 and r2 = 2.
Using Minkowski’s bound, we find that every ideal class must contain an
ideal a of norm strictly less than 4. Therefore we must look at the numbers
2 and 3 to see how they factor in this ring. Let ℘ be an ideal such that
N ℘ = 2 or 3. Then ℘ is prime, because OK /℘ is a field. Recall that to find
℘ we consider f (x) mod p. If
f (x) ≡ f1e1 (x) · · · fgeg (x) (mod p),
e
then pOK factors as ℘e11 · · · ℘gg with N ℘i = pfi where fi is the degree of
fi (x).
First, suppose we have an ideal ℘ with N ℘ = 2. Then ℘ must appear in
the prime factorization of 2OK , and so in the factorization of f (x) (mod 2),
there is a linear factor. However, it is easy to see that x5 − x + 1 has no
linear factor mod 2, and so there are no ideals of norm 2. By a similar
argument, we see that x5 − x + 1 must have a linear factor mod 3, if there
exists an ideal ℘ with norm 3. Again, it is clear that x5 − x + 1 can have
no linear factor, and we conclude that there are no ideals of norm 3 in OK .
Thus, the only ideal of norm less than 4 is the trivial ideal, and we conclude
that OK has class number 1.
√
Exercise 6.5.26 Determine the class number of Q( 14).

Solution. The Minkowski bound in this case is

2! √ √
14 · 4 = 14 = 3.74 . . . .
22
Then we must check ideals of norm less than or equal to 3. Therefore we
must look at 2OK and 3OK to see how they factor in this ring. Since 3 is
inert, there are no ideals of norm 3. However, 2 ramifies as
√ √
(2) = (4 + 14)(4 − 14) = ℘2 ,
√
where ℘ = (4 √ + 14). This is a principal ideal, and we conclude that all
ideals of Q( 14) are principal and so the class number is 1.
262 CHAPTER 6. THE IDEAL CLASS GROUP

Exercise 6.5.27 If K is an algebraic number field of finite degree over Q with

dK squarefree, show that K has no non-trivial subfields.

Solution. By Exercise 6.5.13 and 5.6.25, any proper subfield would intro-
duce a power into the discriminant of K.
Chapter 7

Quadratic Reciprocity

7.1 Preliminaries
Exercise 7.1.1 Let p be a prime and a = # 0. Show that x2 ≡ a (mod p) has a
(p−1)/2
solution if and only if a ≡ 1 (mod p).

Solution. ⇒ Suppose that x2 ≡ a (mod p) has a solution. Let x0 be this

solution, i.e., x20 ≡ a (mod p). But then,

a(p−1)/2 ≡ (x20 )(p−1)/2 ≡ xp−1

0 ≡1 (mod p).

The last congruence follows from Fermat’s Little Theorem.

⇐ We begin by noting that a !≡ 0 (mod p). So a (mod p) can be viewed
as an element of (Z/pZ)× , the units of (Z/pZ). Since (Z/pZ)× is a cyclic
group, there exists some generator g such that :g; = (Z/pZ)× . So, a = g k ,
where 1 ≤ k ≤ p − 1. From our hypothesis,

a(p−1)/2 ≡ g k(p−1)/2 ≡ 1 (mod p).

Because the order of g is p − 1, p − 1|k(p − 1)/2. But this implies that 2|k.
So k = 2k $ . So, we can write a (mod p) as
" "
a ≡ g k ≡ g 2k ≡ (g k )2 (mod p).

Hence a is a square mod p, completing the proof.

Exercise 7.1.2 Using Wilson’s theorem and the congruence

k(p − k) ≡ −k2 (mod p)

compute (−1/p) for all primes p.

263
264 CHAPTER 7. QUADRATIC RECIPROCITY

Solution. The case when p = 2 is trivial since every odd number is con-
gruent to 1 (mod 2). Then we will assume that p is an odd prime. To
begin, we recall Wilson’s theorem (proved in Exercise 1.4.10) which states
that for any prime p, we have (p − 1)! ≡ −1 (mod p). We note that (p − 1)!
can be expressed as
! " ! ! ""
p−1 p−1 p−1
(p − 1)! = 1 · 2 · 3 · · · · p− · p− −1 · · · (p − 1).
2 2 2

Thus, when we mod out by p on both sides, we get

! " ! ! ""
p−1 p−1 p−1
−1 ≡ 1 · 2 · 3 · · · · − · − −1 · · · (−1)
2 2 2
3! " 42
(p−1)/2 p−1
≡ (−1) ! (mod p).
2

If p ≡ 1 (mod 4), then it follows that (p − 1)/2 = 2a for some integer

a. Hence, from the above identity,
3! " 42 3! " 42
p−1 p−1
(−1)(p−1)/2 ! ≡ ! ≡ −1 (mod p).
2 2

So, −1 is a quadratic residue mod p if p ≡ 1 (mod 4).

If p ≡ 3 (mod 4), we find that (p − 1)/2 is odd. If x2 ≡ −1 (mod p),
then by Exercise 7.1.1, (−1)(p−1)/2 ≡ 1 (mod p). But since p ≡ 3 (mod 4),
we know that (−1)(p−1)/2 ≡ −1 (mod p). So, there can be no solutions to
x2 ≡ −1 (mod p) if p ≡ 3 (mod 4).
Thus ! " ;
−1 1 if p ≡ 1 (mod 4),
=
p −1 if p ≡ 3 (mod 4).
Finally, we observe that we can encode this information more compactly
as ! "
−1
= (−1)(p−1)/2 .
p

Exercise 7.1.3 Show that

$ %
(p−1)/2 a
a ≡ (mod p).
p

Solution. If p | a, then the conclusion is trivial. So, suppose p does not

divide a. By Fermat’s Little Theorem, ap−1 ≡ 1 (mod p). We can factor
this statement as

ap−1 − 1 ≡ (a(p−1)/2 − 1)(a(p−1)/2 + 1) ≡ 0 (mod p).

Thus, a(p−1)/2 ≡ ±1 (mod p). We will consider each case separately.

7.1. PRELIMINARIES 265

If a(p−1)/2 ≡ 1 (mod p), then by Exercise 7.1.1, there exists a solution

to the equation x2 ≡ a (mod p). But this implies (a/p) = 1.
If a(p−1)/2 ≡ −1 (mod p), then Exercise 7.1.1 tells us there is no solution
to the equation x2 ≡ a (mod p). So, (a/p) = −1. We conclude that
! "
a
a(p−1)/2 ≡ (mod p).
p

Exercise 7.1.4 Show that

$ % $ %$ %
ab a b
= .
p p p

Solution. We will use Exercise 7.1.3 to prove this result. Thus,

! "
ab
≡ (ab)(p−1)/2 (mod p).
p

Similarly, ! "
a
≡ a(p−1)/2 (mod p),
p
and ! "
b
≡ b(p−1)/2 (mod p).
p
But then
! " ! "! "
ab (p−1)/2 (p−1)/2 (p−1)/2 a b
≡ (ab) =a b ≡ (mod p).
p p p

Thus ! " ! "! "

ab a b
≡ (mod p).
p p p
But because the Legendre symbol only takes on the values ±1, we can
rewrite this statement as
! " ! "! "
ab a b
= ,
p p p

which is what we wished to show.

Exercise 7.1.5 If a ≡ b (mod p), then

$ % $ %
a b
= .
p p

Solution. This is clear from the definition of the Legendre symbol. If a is

a quadratic residue mod p then so is b. The same is true if a is a quadratic
nonresidue.
266 CHAPTER 7. QUADRATIC RECIPROCITY

Exercise 7.1.7 Show that the number of quadratic residues mod p is equal to
the number of quadratic nonresidues mod p.

Solution. The equation a(p−1)/2 ≡ 1 (mod p) will have (p−1)/2 solutions.

This can be deduced from the fact that (Z/pZ)× is a cyclic group with some
generator g, where g is a (p − 1)st root of unity. So, for any even power of
g, i.e., g 2k , we will have
(g 2k )(p−1)/2 ≡ (g p−1 )k ≡ 1 (mod p).
For any odd power,
(g 2k+1 )(p−1)/2 ≡ (g (p−1)/2 ) !≡ 1 (mod p).
The last congruence holds since g is a (p−1)st root of unity. Thus, half of the
elements of (Z/pZ)× will correspond to some even power of g, and hence,
a(p−1)/2 ≡ 1 (mod p). But this in turn implies that there are (p − 1)/2
elements such that (a/p) = 1. Since there are (p − 1) residues mod p, there
are (p − 1) − (p − 1)/2 = (p − 1)/2 residues that are not squares. But now
we have that the number of quadratic residues mod p and the number of
quadratic nonresidues mod p are equal.

Exercise 7.1.8 Show that

p−1 $ %
, a
=0
a=1
p
for any fixed prime p.

Solution. From Exercise 7.1.7, the number of residues equals the number
of nonresidues. So, there are (p − 1)/2 residues, and (p − 1)/2 nonresidues.
Thus
p−1 ! "
* a p−1 p−1
= (1) + (−1) = 0.
a=1
p 2 2

7.2 Gauss Sums

Exercise 7.2.2 Show that
$ %
q q
S ≡ S (mod q),
p
where q and p are odd primes.

Solution. Let K = Q(ζq ), where ζq is a primitive qth root of unity. Let

R = OK be its ring of integers. So
 q
* !a"
Sq =  ζ a
p p
a mod p
* ! a "q
≡ ζpaq (mod qR).
p
a mod p
7.3. THE LAW OF QUADRATIC RECIPROCITY 267

This follows from the fact that (x1 + · · · + xn )q ≡ xq1 + · · · + xqn (mod q).
Also, because q is an odd prime, and because (a/p) only takes on the values
±1, we have (a/p)q = (a/p). Hence
* !a"
q
S ≡ ζ aq (mod qR).
p p
a mod p

However, ! " ! "! 2" ! 2"

a a q aq
= = .
p p p p
So, it follows that
* ! aq 2 "
S q
≡ ζpaq (mod qR)
p
a mod p
! " * ! "
q aq
≡ ζpaq (mod qR).
p p
a mod p

But as a runs through all the residue classes mod p, so will aq. Thus,
! "
q
Sq ≡ S (mod qR).
p
From this, it follows that
! "
q
q
S ≡ S (mod q),
p
completing the proof.

7.3 The Law of Quadratic Reciprocity

Exercise 7.3.2 Let q be an odd prime. Prove:
(a) If q ≡ 1 (mod 4), then q is a quadratic residue mod p if and only if p ≡ r
(mod q), where r is a quadratic residue mod q.
(b) If q ≡ 3 (mod 4), then q is a quadratic residue mod p if and only if p ≡ ±b2
(mod 4q), where b is an odd integer prime to q.

Solution. (a) We begin by rewriting the result of Theorem 7.3.1 in the

following equivalent form:
! "! "
p q p−1 q−1
= (−1) 2 · 2 .
q p
Since q ≡ 1 (mod 4), (q − 1)/2 is even, so we will have
! "! "
p q
= 1.
q p
268 CHAPTER 7. QUADRATIC RECIPROCITY

From this it follows that either (p/q) = 1 = (q/p), or (p/q) = −1 = (q/p).

We can now prove the statement.
⇒ If (q/p) = 1, then (p/q) = 1. So, by Exercise 7.1.5, p ≡ r (mod q),
where r is a quadratic residue mod q.
⇐ Suppose p ≡ r (mod q), and (r/q) = 1. But then (p/q) = (r/q), and
thus, (p/q) = 1. Since (q/p) = (p/q) by quadratic reciprocity, we must also
have that (q/p) = 1, which implies that q is a quadratic residue mod p.
This completes the proof of (a).
(b) Suppose q ≡ 3 (mod 4). Then, quadratic reciprocity gives us
! " ! "
q p−1 p
= (−1) 2 .
p q
⇒ Suppose (q/p) = 1. Then, we have two cases:
Case 1. (−1)(p−1)/2 = −1 and (p/q) = −1.
Case 2. (−1)(p−1)/2 = 1 and (p/q) = 1.
Case 1. First we note that (−1)(p−1)/2 = −1 implies that p ≡ 3
(mod 4). Because q ≡ 3 (mod 4), we have from an earlier exercise that
! " ! "
−1 p
= = −1.
q q
But then, we find that
! " ! "! 2" ! 2"
−1 −1 b −b
= = .
q q q q
Hence (p/q) = (−b2 /q), and p ≡ −b2 (mod q). We can suppose that b is
odd. If not, we can replace it with b$ = b + q, which is odd. Since b is odd,
b = 2n + 1. Thus b2 = 4n2 + 4n + 1, and so −b2 ≡ 3 (mod 4). Since we
already deduced that p ≡ 3 (mod 4), we have p ≡ −b2 (mod 4). Because
p ≡ −b2 (mod q), we conclude that p ≡ −b2 (mod 4q).
Case 2. (−1)(p−1)/2 = 1 implies that p ≡ 1 (mod 4). Also, (p/q) =
(b /q). Assume b is odd for the same reason given above. So p ≡ b2
2

(mod q). Because b is odd, we have b = 2n + 1, which means that b2 ≡ 1

(mod 4). But p ≡ 1 (mod 4), so clearly p ≡ b2 (mod 4). Since we also
know that p ≡ b2 (mod q), we conclude that p ≡ b2 (mod 4q).
⇐ Suppose we have p ≡ ±b2 (mod 4q), where b is coprime to q. We
will examine each case, p ≡ b2 (mod 4q) and p ≡ −b2 (mod 4q) separately.
If p ≡ b2 (mod 4q), we have p ≡ b2 ≡ 1 (mod 4), and p ≡ b2 (mod q).
But p ≡ 1 (mod 4) implies that (−1)(p−1)/2 = 1. Since p ≡ b2 (mod q) it
follows that ! " ! 2"
p b
=1= .
q q
Hence ! " ! "
q p−1 p
= (−1) 2 = 1 · 1 = 1,
p q
7.3. THE LAW OF QUADRATIC RECIPROCITY 269

so q is a quadratic residue mod p.

If p ≡ −b2 (mod 4q), we have both that p ≡ −b2 ≡ 3 (mod 4), and
p ≡ −b2 (mod q). So, (−1)(p−1)/2 = −1 from the fact that p ≡ 3 (mod 4).
From p ≡ −b2 (mod q), we deduce that
! " ! 2" ! "! 2" ! "
p −b −1 b −1
= = = .
q q q q q

Since q ≡ 3 (mod 4), we know that (−1/q) = −1. So, (p/q) = −1. Thus,
! " ! "
q p
= (−1)(p−1)/2 = (−1)(−1) = 1.
p q

It now follows that q is a quadratic residue mod p. This completes the

proof.

Exercise 7.3.3 Compute (5/p) and (7/p).

Solution. We will first compute (5/p). Since 5 ≡ 1 (mod 4), we can use
part (a) of Exercise 7.3.2. So (5/p) = 1 if and only if p ≡ r (mod 5), where
r is a quadratic residue mod 5. It is easy to determine which r are quadratic
residues mod 5; 12 ≡ 1, 22 ≡ 4, 32 ≡ 4, 42 ≡ 1. So, 1 and 4 are quadratic
residues mod 5, while 2 and 3 are not. Thus

 1 if p ≡ 1, 4 (mod 5),
! " 
5
= −1 if p ≡ 2, 3 (mod 5),
p 

0 if p = 5.

Now, we will find (7/p). Since 7 ≡ 3 (mod 4), we must use part (b) of
Exercise 7.3.2. So, we have to compute all the residues mod 28 of all the
squares of odd integers prime to 7. Some calculation reveals that

12 , 132 , 152 , 272 ≡ 1 (mod 28),

32 , 112 , 172 , 252 ≡ 9 (mod 28),

and
52 , 92 , 172 , 232 ≡ 25 (mod 28).

Thus 
 1 if p ≡ ±1, ±9, ±25 (mod 28),
! " 
7
= −1 if p ≡ ±5, ±11, ±13 (mod 28),
p 

0 if p = 7.
270 CHAPTER 7. QUADRATIC RECIPROCITY

7.4 Quadratic Fields

√
Exercise 7.4.1 Find the discriminant of K = Q( d) when:
(a) d ≡ 2, 3 (mod 4); and
(b) d ≡ 1 (mod 4).
√
Solution. (a) If d ≡ 2, 3 (mod 4), then ω1 = 1, ω2 = d forms an integral
basis for OK . Then an easy calculation shows that√dK = 4d.
(b) For d ≡ 1 (mod 4), then ω1 = 1, ω2 = (1 + d)/2 forms an integral
basis. Then dK = d.
(See also Example 4.3.4.)

Exercise 7.4.3 Assume that p is an odd prime. Show that (d/p) = 0 if and only
if pOK = ℘2 , where ℘ is prime.
√
Solution. ⇒ We claim that pOK = (p, d)2 . Notice that
√ √ √
(p, d)2 = (p2 , p d, d) = (p)(p, d, d/p).
√
Because d√is squarefree, d/p and p are relatively prime. So, (p, d, d/p) = 1.
Since (p, d) is a prime ideal (for the same reason given above), we have
shown that pOK ramifies. √
⇐ Once again, let m be the discriminant of K = Q( d). Since √ pOK
ramifies we can find some a ∈ ℘, but a !∈ pOK . So, a = x + y(m + m)/2.
Since a2 ∈ pOK , we get

(2x + ym)2 + my 2 (2x + ym) √

+ 2y m ∈ pOK .
4 4
Thus, p | (2x + ym)2 + my 2 and p | y(2x + ym). If p | y, then p | 2x. Since p
is odd, p | x. But then a ∈ pOK . This is a contradiction. So, p | (2x + ym)
and p | my 2 . Thus, p | m. But this means that (d/p) = 0, since p is an odd
prime.

Exercise 7.4.4 Assume p is an odd prime. Then (d/p) = −1 if and only if

pOK = ℘, where ℘ is prime.

Solution. This follows immediately from Theorem 7.4.3 and Exercise 7.4.3.
If (d/p) = −1, then we know that pOK does not split, nor does it ramify.
So pOK must stay inert. Conversely, if pOK is inert, the only possible value
for (d/p) is −1.

7.5 Primes in Special Progressions

Exercise 7.5.1 Show that there are infinitely many primes of the form 4k + 1.
7.5. PRIMES IN SPECIAL PROGRESSIONS 271

Solution. Suppose there are only a finite number of primes of this form.
Let p1 , p2 , . . . , pn be these primes. Let d = (2p1 p2 · · · pn )2 + 1. Let p be
a prime that divides this number. So d ≡ 0 (mod p), which implies that
(−1/p) = 1.
From Exercise 7.1.2, we know that this only occurs if p ≡ 1 (mod 4).
So, p ∈ {p1 , p2 , . . . , pn }. But for any pi ∈ {p1 , p2 , . . . , pn }, pi does not
divide d by construction. So p !∈ {p1 , p2 , . . . , pn }. Thus, our assumption is
incorrect, and so, there must exist an infinite number of primes of the form
4k + 1.

Exercise 7.5.2 Show that there are infinitely many primes of the form 8k + 7.

Solution. Suppose the statement is false, that is, there exist only a finite
number of primes of the form 8k + 7. Let p1 , p2 , . . . , pn be these primes.
Construct the following integer, d = (4p1 p2 · · · pn )2 − 2. Let p be any
prime that divides this number. But then (4p1 p2 · · · pn )2 ≡ 2 (mod p), so
(2/p) = 1. From Theorem 7.1.6, we can deduce that p ≡ ±1 (mod 8).
We claim that all the odd primes that divide d cannot have the form
8k + 1. We observe that 2 | (4p1 p2 · · · pn )2 − 2. So, any odd prime that
divides d must divide 8(p1 p2 · · · pn )2 − 1. If all primes were of the form
8k + 1, we would have

8(p1 p2 · · · pn )2 − 1 = (8k1 + 1)e1 · (8k2 + 1)e2 · · · (8km + 1)em .

But now consider this equation mod 8. We find that −1 ≡ 1 (mod 8),
which is clearly false. So, there must be at least one odd prime p of the
form 8k + 7 that divides d.
So, p ∈ {p1 , p2 , . . . , pn }. But p cannot be in {p1 , p2 , . . . , pn } since every
pi leaves a remainder of −2 when dividing d. So, {p1 , p2 , . . . , pn } does not
contain all the primes of the form 8k + 7. But we assumed that it did.
We have arrived at a contradiction. Therefore, there must be an infinite
number of primes of the form 8k + 7.

Exercise 7.5.3 Show that p ≡ 4 (mod 5) for infinitely many primes p.

Solution. From the preceding comments, we know that we can use a

Euclid-type proof to prove this assertion since 42 ≡ 1 (mod 5). So, suppose
there exists only a finite number of such primes, say p1 , p2 , . . . , pn . Consider
the integer 25(p1 p2 · · · pn )2 − 5. Then for any prime divisor not equal to 5,
we will have (5/p) = 1. From Exercise 7.3.3, we know that this will only
occur if p ≡ 1, 4 (mod 5). Since p != 5, then p | 5(p1 · · · pn )2 − 1. So, all the
prime divisors cannot be congruent to 1 (mod 5), because if this was true,
−1 ≡ 1 (mod 5) which is clearly false. So, there is some p ≡ 4 (mod 5)
that divides 5(p1 · · · pn )2 − 1. But p is not any of the p1 , . . . , pn since none
of these numbers divide 5(p1 · · · pn )2 − 1. This gives us a contradiction.
272 CHAPTER 7. QUADRATIC RECIPROCITY

7.6 Supplementary Problems

Exercise 7.6.1 Compute (11/p).

Solution. Since 11 ≡ 3 (mod 4), we use Exercise 7.3.2 (b) which says that
11 is a quadratic residue mod p if and only if p ≡ ±b2 (mod 44) where b
is an odd integer prime to p. If we compute b2 (mod 44) for all possible b,
we get

12 , 212 , 232 , 432 ≡ 1 (mod 44),

32 , 192 , 252 , 412 ≡ 9 (mod 44),
52 , 172 , 272 , 392 ≡ 25 (mod 44),
2 2 2 2
7 , 15 , 29 , 37 ≡ 5 (mod 44),
92 , 132 , 312 , 352 ≡ 37 (mod 44).

It is now easy to determine the primes for which 11 is a quadratic residue

by applying Exercise 7.3.2.

Exercise 7.6.3 If p ≡ 1 (mod 3), prove that there are integers a, b such that
p = a2 − ab + b2 .
√ √
Solution.
√ Let ρ = (1+ −3)/2 and consider Q( −3). The ring of integers
of Q( −3) is Z[ρ]. Since p ≡ 1 (mod 3), x2 +3 ≡ 0 (mod p) has a solution.
Hence p splits in Z[ρ]. Now use the fact that Z[ρ] is Euclidean.

Exercise 7.6.4 If p ≡ ±1 (mod 8), show that there are integers a, b such that
a2 − 2b2 = ±p.
√
Solution. Consider the Euclidean ring Z[ 2].

Exercise 7.6.5 If p ≡ ±1 (mod 5), show that there are integers a, b such that
a2 + ab − b2 = ±p.
√
Solution. Let ω = (1 + 5)/2 and consider Z[ω].

Exercise 7.6.10 Show that the number of solutions of the congruence

x2 + y 2 ≡ 1 (mod p),

0 < x < p, 0 < y < p (p an odd prime), is even if and only if p ≡ ±3 (mod 8).

Solution. Pair up the solutions, (x, y) with (y, x). The number of solutions
is even unless (x, x) is a solution which means that 2 is a square mod p.

Exercise 7.6.11 If p is a prime such that p − 1 = 4q with q prime, show that 2

is a primitive root mod p.
7.6. SUPPLEMENTARY PROBLEMS 273

Solution. The only possible orders of 2 (mod p) are 1, 2, 4, q, 2q, or 4q.

Since p > q, the orders 1, 2, 4 are impossible. If the order is q or 2q, then
2 is a quadratic residue mod p. However, p ≡ 5 (mod 8).

Exercise 7.6.12 (The Jacobi Symbol) Let Q be a positive odd number. We

can write Q = q1 q2 · · · qs where the qi are odd primes, not necessarily distinct.
Define the Jacobi symbol
$ % # s $ %
a a
= .
Q j=1
qi

If Q and Q( are odd and positive, show that:

(a) (a/Q)(a/Q( ) = (a/QQ( ).
(b) (a/Q)(a( /Q) = (aa( /Q).
(c) (a/Q) = (a( /Q) if a ≡ a( (mod Q).

Solution. All of these are evident from the properties of the Legendre
symbol. For (c), note that a ≡ a$ (mod Q) implies a ≡ a$ (mod qi ) for
i = 1, 2, . . . , s.

Exercise 7.6.13 If Q is odd and positive, show that

$ %
−1
= (−1)(Q−1)/2 .
Q

Solution.
! " 6 s ! " 6 s !s
−1 −1
= = (−1)(qj −1)/2 = (−1) j=1 (qj −1)/2 .
Q j=1
qj j=1

Now, if a and b are odd, observe that

! "
ab − 1 a−1 b−1 (a − 1)(b − 1)
− + = ≡0 (mod 2).
2 2 2 2

Hence
a−1 b−1 ab − 1
+ ≡ (mod 2).
2 2 2
Applying this observation repeatedly in our context gives
! "
−1
= (−1)(Q−1)/2 .
Q
2
Exercise 7.6.14 If Q is odd and positive, show that (2/Q) = (−1)(Q −1)/8
.

Solution. If a and b are odd, then

! 2 "
a2 b2 − 1 a − 1 b2 − 1 (a2 − 1)(b2 − 1)
− + = ≡0 (mod 8)
8 8 8 8
274 CHAPTER 7. QUADRATIC RECIPROCITY

so we have
a2 − 1 b2 − 1 a2 b2 − 1
+ ≡ (mod 8).
8 8 8
Again applying this repeatedly in our context gives the result.

Exercise 7.6.15 (Reciprocity Law for the Jacobi Symbol) If P and Q are
odd, positive, and coprime, show that
$ %$ %
P Q P −1 Q−1
= (−1) 2 · 2 .
Q P
2r 2s
Solution. Write P = i=1 pi and Q = j=1 qj . Then
! " s !
6 " r !
s 6
6 "
P P pi
= =
Q j=1 j q j=1 i=1
qj
6s 6r ! "
qj pi −1 qj −1
= (−1) 2 · 2

j=1 i=1
pi
! " !
Q pi −1 qj −1
2 · 2
= (−1) i,j
P
by the reciprocity law for the Legendre symbol. But, as noted in the pre-
vious exercises,
* r
pi − 1 P −1
≡ (mod 2)
i=1
2 2
and
*s
qj − 1 Q−1
≡ (mod 2),
j=1
2 2

which completes the proof.

Exercise 7.6.16 (The Kronecker Symbol) We can define (a/n) for any in-
teger a ≡ 0 or 1 (mod 4), as follows. Define

-a. $ a %   0 if a ≡ 0 (mod 4),
= = 1 if a ≡ 1 (mod 8),
2 −2 

−1 if a ≡ 5 (mod 8).

For general n, write n = 2c n1 , with n1 odd, and define

- a . - a .c $ a %
= ,
n 2 n1
where (a/2) is defined as above and (a/n1 ) is the Jacobi symbol.
Show that if d is the discriminant of a quadratic field, and n, m are positive
integers, then $ % $ %
d d
= for n ≡ m (mod d)
n m
7.6. SUPPLEMENTARY PROBLEMS 275

and $ % $ %
d d
= sgn d for n ≡ −m (mod d).
n m

Solution. Let d = 2a d$ , n = 2b n$ , m = 2c m$ with d$ , n$ , m$ odd. If a > 0,

the case b > 0 is trivial since then c > 0 and both the symbols are zero. So
suppose b = c = 0. Then
! " ! a $ " ! "a ! $ " EnF
d 2 d 2 d 2 "
= = = (−1)a(n −1)/8 $ (−1)(n−1)(d −1)/4
n n n n d

and similarly
! " ! a $" EmF
d 2 d 2 "
= = (−1)a(m −1)/8 $ (−1)(m−1)(d −1)/2 .
m m d

Since 4 | d, the first factors coincide for m and n. The same is true for the
other factors in the case n ≡ m (mod d). But if n ≡ −m (mod d), they
differ by sgn d$ which is sgn d.
In the case a = 0, we note d ≡ 1 (mod 4). Then
! " ! " ! " b ! " ! "b ! "
d d d d 2 d
= = =
n 2b n $ 2 n$ d n$

since (d/2) = (2/d) for d ≡ 1 (mod 4). Thus (d/n) = (n/d) and (−1/d) =
sgn d. Therefore
! " ! "
d d
= for m, n > 0, m ≡ n (mod d)
m n

and ! " E F ! " EmF ! d "

d n −m
= = = sgn d = sgn d
n d d d m
for n ≡ −m (mod d).

Exercise 7.6.17 If p is an odd prime show that the least positive quadratic
√
nonresidue is less than p + 1.
(It is a famous conjecture of Vinogradov that the least quadratic nonresidue
mod p is O(pε ) for any ε > 0.)

Solution. Let n be the least positive quadratic nonresidue and m the least
such that mn > p, so that n(m − 1) ≤ p. Since p is prime, n(m − 1) < p <
mn. Now mn − p < n so that
! " ! " ! "
mn − p mn m
1= = =− .
p p p

Therefore m ≥ n, so that (n − 1)2 < n(n − 1) < p.

276 CHAPTER 7. QUADRATIC RECIPROCITY

Exercise 7.6.18 Show that x4 ≡ 25 (mod 1013) has no solution.

Solution. First observe that 1013 is prime. If x4 ≡ 25 (mod 1013) had a

solution x0 , then x20 ≡ ±5 (mod 1013). However,
! " ! " ! " ! "
±5 5 1013 3
= = = = −1
1013 1013 5 5
so the congruence has no solutions.

Exercise 7.6.19 Show that x4 ≡ 25 (mod p) has no solution if p is a prime

congruent to 13 or 17 (mod 20).

Solution. If the congruence has a solution, then

! " ! " E F ! "
±5 5 p ±3
1= = = = = −1,
p p 5 5
a contradiction.

Exercise 7.6.20 If p is a prime congruent to 13 or 17 (mod 20), show that

x4 + py 4 = 25z 4 has no solutions in integers.

Solution. We may suppose that gcd(x, y, z) = 1, because otherwise we

can cancel the common factor. Also gcd(p, z) = 1 for otherwise p | x and
p | y. Now reduce the equation mod p. We have a solution to x4 ≡ 25z 4
(mod p) so that by the previous question,
! " ! "
±5z 2 ±5
1= = = −1,
p p
a contradiction.
√
Exercise 7.6.21 Compute the class number of Q( 33).

Solution. The discriminant is 33 and the Minkowski bound is

1
√
2 33 = 2.87 . . . .

Since 33 ≡ 1 (mod 8), 2 splits as a product of two ideals each of norm 2.

Moreover, √ √
33 − 5 33 + 5
2= ·
2 2
and so the principal ideals
/√ 0 /√ 0
33 − 5 33 + 5
and
2 2

are the only ones of norm 2. Hence the class number is 1.

7.6. SUPPLEMENTARY PROBLEMS 277

√
Exercise 7.6.22 Compute the class number of Q( 21).

Solution. The discriminant is 21 and the Minkowski √ bound is 2.29 . . . .

However, 21 ≡ 5 (mod 8) so that 2 is inert in Q( 21). Therefore, there
are no ideals of norm 2 and the class number is 1.
√
Exercise 7.6.23 Show that Q( −11) has class number 1.

Solution. The field has discriminant −11 and Minkowski’s bound is

2√
11 = 2.11 . . . .
π
We must examine ideals of√ norm 2. Since −11 ≡ 5 (mod 8), by Theo-
rem 7.4.5, 2 is inert in Q( −11), so that there are no ideals of norm 2.
Hence the class number is 1.
√
Exercise 7.6.24 Show that Q( −15) has class number 2.

Solution. The field has discriminant −15 and Minkowski’s bound is

2√
15 = 2.26 . . . .
π
Since −15 ≡ 1 (mod 8), by Theorem 7.4.5, 2 splits as a product of two
ideals ℘, ℘$ each of norm 2. If ℘ were principal, then
! √ "
u + v −15
℘=
2

for integers u, v. However, 8 = u2 + 15v 2 has no solution. Thus ℘ is not

principal and the class number is 2.
√
Exercise 7.6.25 Show that Q( −31) has class number 3.

Solution. The discriminant is −31 and the Minkowski bound is

2√
31 = 3.26 . . . .
π
So we must consider ideals of norm less than or equal to 3.
Since −31 ≡ 1 (mod 8), 2 splits as ℘2 · ℘$2 (say). Since
! "
−31
= −1,
3

3 is inert so there are no ideals of norm 3. Moreover, neither ℘2 or ℘$2 are

principal since we cannot solve

8 = u2 + 31v 2 .
278 CHAPTER 7. QUADRATIC RECIPROCITY

Since ! √ "! √ "

12 + (31) · 12 1 − −31 1 + −31
8= =
4 2 2
√ √
and ((1− −31)/2), ((1+ −31)/2) are coprime, ℘32 and (℘$2 )3 are principal.
℘2 cannot be principal since 16 = u2 + 31v 2 has no solution except (u, v) =
(±4, 0) in which case

℘22 = (4) = (2)2 = ℘22 (℘$2 )2 ,

which implies (℘$2 )2 = 1, a contradiction. Thus, as each ideal class contains

either (1), ℘2 or ℘$2 and because ℘22 is inequivalent to (1) or ℘2 we must
have ℘22 ∼ ℘$2 . Thus, ℘32 ∼ (1). Thus, the class number is 3.
Chapter 8

The Structure of Units

8.1 Dirichlet’s Unit Theorem

Exercise 8.1.1 (a) Show that there are only finitely many roots of unity in K.
(b) Show, similarly, that for any positive constant c, there are only finitely many
α ∈ OK for which |α(i) | ≤ c for all i.

Solution. (a) Suppose that αm = 1. Then α ∈ OK , |α|m = 1 ⇒ |α| = 1

and, if σ1 , . . . , σn are the distinct embeddings of K in C, then, for each
α(i) = σi (α), we have that σi (α)m = σi (αm ) = 1 ⇒ |α(i) | = 1 for i =
1, . . . , n.
The characteristic polynomial of α is
n
6
fα (x) = (x − α(i) ) = xn + an−1 xn−1 + · · · + a0 ∈ Z[x].
i=1

Now,
an−j = (−1)j sj (α(1) , . . . , α(n) ),
where sj (α(1) , . . . , α(n) ) is the jth symmetric function in the α(i) , i.e., the
sum of all products of the α(i) , taken j at a time. This implies that
! "
n
|an−j | ≤ ≤ n!.
j
Thus, since the aj ’s are bounded, there are only finitely many choices for
the coefficients of the characteristic polynomial of a root of unity α ∈ K
and, hence, only finitely many such roots of unity.
(b) Suppose that α ∈ OK such that |α(i) | ≤ c for all i = 1, . . . , n. As in
(a), let
n
6
fα (x) = (x − α(i) ) = xn + an−1 xn−1 + · · · + a0 ∈ Z[x]
i=1

279
280 CHAPTER 8. THE STRUCTURE OF UNITS

be the characteristic polynomial of α. Then

! "
n
|an−j | = |sj (α(1) , . . . , α(n) )| ≤ cj ≤ cn n!.
j
Thus, the coefficients of the characteristic polynomial of such an α in OK
are bounded. There are, therefore, only finitely many such α.

Exercise 8.1.2 Show that WK , the group of roots of unity in K, is cyclic, of

even order.
q
Solution. Let α1 , . . . , αl be the roots of unity in K. For j = 1, . . . , l, αj j =
1 for some qj which implies that αj = e2πipj /qj , for some 0 ≤ pj ≤ qj − 1.
2l
Let q0 = i=1 qj . Then, clearly, each αi ∈ :e2πi/q0 ; so WK is a subgroup of
the cyclic group :e2πi/q0 ; and is, thus, cyclic. Moreover, since {±1} ⊆ WK ,
WK has even order.

Exercise 8.1.7 (a) Let Γ be a lattice of dimension n in Rn and suppose that

{v1 , . . . , vn } and {w1 , . . . , wn } are two bases for Γ over Z. Let V and W
be the n × n matrices with rows consisting of the vi ’s and wi ’s, respectively.
Show that |det V | = |det W |. Thus, we can unambiguously define the volume
of the lattice Γ, vol(Γ) = the absolute value of the determinant of the matrix
formed by taking, as its rows, any basis for Γ over Z.
(b) Let ε1 , . . . , εr be a fundamental system of units for a number field K. Show
(i)
that the regulator of K, RK = |det(log |εj |)|, is independent of the choice of
ε1 , . . . , ε r .

Solution. (a) Since {w1 , . . . , wn } is a Z-basis for Γ,5

we can express each
n
vi as a Z-linear combination of the wj ’s, say vi = j=1 aij wj . Setting
A = (aij ), we have that V = AW . Since {v1 , . . . , vn } is also an integral
basis for Γ, the matrix A is invertible, A ∈ GLn (Z) ⇒ det A ∈ Z∗ = {±1}.
Thus, |det V | = |det A||det W | = |det W |.
(b) As before, define
f : UK → Rr ,
ε 9 → (log |ε(1) |, . . . , log |ε(r) |).
Im f is a lattice of dimension r in Rr , with Z-basis {f (ε1 ), . . . , f (εr )},
for any system of fundamental units, ε1 , . . . , εr . By definition, RK is the
absolute value of the determinant of the matrix formed by taking, as its
rows, the f (ε(i) )’s. By (a) then, RK is independent of the particular system
of fundamental units.
√
Exercise 8.1.8 (a) Show that, for any real quadratic field K = Q( d), where d
is a positive squarefree integer, UK 3 Z/2Z×Z. That is, there is a fundamen-
tal unit ε ∈ UK such that UK = {±εk : k ∈ Z}. Conclude that the equation
x2 − dy 2 = 1 (erroneously dubbed Pell’s equation) has infinitely many integer
solutions for d ≡ 2, 3 mod 4 and that the equation x2 − dy 2 = 4 has infinitely
many integer solutions for d ≡ 1 mod 4.
8.1. DIRICHLET’S UNIT THEOREM 281

(b) Let d ≡ 2, 3 (mod 4). Let b be the smallest positive

√ integer such that one of
db2 ± 1 is a square, say a2 , a > 0. Then a + b d is a unit. Show that it is the
fundamental
√ √ unit. Using this algorithm, determine the fundamental units of
Q( 2), Q( 3).
√
(c) Devise a similar algorithm to compute the fundamental √ unit in Q( d), for
d ≡ 1 (mod 4). Determine the fundamental unit of Q( 5).

Solution. (a) Since K ⊆ R, the only roots of unity in K are {±1}, so

WK = {±1}. Moreover, since there are r1 = 2 real and 2r2 = 0 nonreal
embeddings of K in C, by Dirichlet’s theorem, we have that UK 1 WK ×Z 1
Z/2Z × Z. √ √
Suppose that d ≡ 2, 3 (mod 4), so that OK = Z[ d]. If ε = a + b d ∈
2
UK , then
√ √
NK (ε) = (a + b d)(a − b d) = a2 − db2 = 1.
√ 2
i.e., each ε = a + b d ∈ UK yields a solution (a, b) ∈ Z2 to the equation
2 2 2
x − dy = 1. Since UK 1 Z is infinite, there are infinitely many such
solutions.
Suppose, now, that d ≡ 1 (mod 4), so that
; √ G
a+b d
OK = : a, b ∈ Z, a ≡ b (mod 2) .
2
√ 2
If ε = (a + b d)/2 ∈ UK , then NK (ε) = (a2 − db2 )/4 = 1 ⇒ a2 − db2 = 4.
2
Since UK is infinite and each of its elements yields an integral solution to
x2 − dy 2 = 4, this equation has infinitely many √ solutions x, y ∈ Z.
√
(b) We have that db2 ± 1 = a√ 2
so NK (a + b d)√= ±1 and a + b d ∈ UK .
Also, a, b > 0 means that a + b d >√1 so a + b d = εk , for some k ≥ √1,
where ε is the fundamental unit in Q( √d). If k > 1,√then write ε = α+β d,
α, β > 0. It is easy to see that a + b d = (α + β d)k implies that α < a
and β√< b. But dβ 2 ± 1 = α2 , contradicting the minimality of b. Therefore,
a + b d is,√ in fact, the fundamental √unit. √
In Q( 2), 2(1)2 − √ 1 = 12 ⇒ 1 + 2 is the fundamental unit. In Q( 3),
3(1)2 + 1 = 22 ⇒ 2 + 3 is the fundamental unit.
(c) Let d ≡ 1 (mod 4). The same argument as in (b) shows that, if
2
b is the smallest positive
√ integer such that one of db ± 4 √ is a square,√say
2
a , a > 0, then (a + b √ d)/2 is the fundamental unit in Q( d). In Q( 5),
5(1)2 − 4 = 12 ⇒ (1 + 5)/2 is the fundamental unit.
√
Exercise 8.1.9 (a) For an imaginary quadratic field K = Q( −d) (d a positive,
squarefree integer), show that

Z/4Z for d = 1,

UK 3 Z/6Z for d = 3,


Z/2Z otherwise.
282 CHAPTER 8. THE STRUCTURE OF UNITS

(b) Show that UK is finite ⇔ K = Q or K is an imaginary quadratic field.

(c) Show that, if there exists an embedding of K in R, then WK 3 {±1} 3 Z/2Z.
Conclude that, in particular, this is the case if [K : Q] is odd.
√
Solution.
√ (a) Suppose that −d ≡ 2, 3 (mod 4) so that OK = Z[ −d]. Let
a + b −d ∈ UK . Then a2 + db2 = 1 (since a2 , b2 , d ≥ 0, a2 + db2 != −1). If
b = 0, then a = ±1. If b != 0, then a2 + db2 ≥ d; thus,

d>1 ⇒ UK = {±1} 1 Z/2Z.

For d = 1, a2 + db2 = a2 + b2 = 1 ⇒ (a, b) ∈ {(±1, 0), (0, ±1)}

⇒ UK = {±1, ±i} = :i; 1 Z/4Z.

√
Suppose now that −d ≡ 1 (mod 4). Then (a + b −d)/2 ∈ UK ⇒
a2 + db2 = 4. If d > 4, then the only solutions to this equation are (a, b) =
(±2, 0)
⇒ UK 1 Z/2Z.

If d = 3, then a2 + db2 = a2 + 3b2 = 4. Then (a, b) ∈ {(±2, 0), (±1, ±1)},

so that
UK = {±ζ3 , ±ζ32 , ±1} = :−ζ3 ; 1 Z/6Z
√
(where ζ3 = (−1 + −3)/2).
(b) We have already shown that for K a quadratic imaginary field, UK
is finite. Now, suppose that UK is finite. Then r1 + r2 = 1 which implies
that either r1 = 1, r2 = 0 so that [K : Q] = r1 + 2r2 = 1 and K = Q
or r1 = 0, r2 = 1 and [K : Q] = 2 and K !⊆ R, so that K is quadratic
imaginary.
(c) If there exists an embedding σ : K 8→ R, then K 1 σ(K) and in
particular WK 1 Wσ(K) ⊆ R. Since the only real roots of unity are {±1},
we must have WK = σ −1 ({±1}) = {±1}. In particular, if [K : Q] = r1 +2r2
is odd, then r1 is odd and, therefore, ≥ 1.

Exercise 8.1.11 Let [K : Q] = 3 and suppose that K has only one real embed-
ding. Then, by Exercise 8.1.8 (c), WK = {±1} implies that UK = {±uk : k ∈ Z},
where u > 1 is the fundamental unit in K.

(a) Let u, ρeiθ , ρe−iθ be the Q-conjugates of u. Show that u = ρ−2 and that
dK/Q (u) = −4 sin2 θ(ρ3 + ρ−3 − 2 cos θ)2 .
(b) Show that |dK/Q (u)| < 4(u3 + u−3 + 6).
(c) Conclude that u3 > d/4 − 6 − u−3 > d/4 − 7, where d = |dK |.
8.1. DIRICHLET’S UNIT THEOREM 283

Solution. (a) NK (u) = uρ2 = ±1, so u = ±ρ−2 , but u > 1 implies that
u = ρ−2 ,
6
dK/Q (u) = (u(r) − u(s) )2
1≤r<s≤3

= (ρ−2 − ρeiθ )2 (ρ−2 − ρe−iθ )2 (ρeiθ − ρe−iθ )2

# −4 $2 # $2
= ρ − ρ(eiθ + e−iθ ) + ρ2 ρ(eiθ − e−iθ )
# $2 # $2
= ρ−4 − ρ(2 cos θ) + ρ2 ρ(2i sin θ)
= −4 sin2 θ(ρ3 + ρ−3 − 2 cos θ)2 .

(b)
|dK/Q (u)| = 4 sin2 θ(ρ3 + ρ−3 − 2 cos θ)2 .
Now set x = ρ3 + ρ−3 , c = cos θ and consider

f (x) = (1 − c2 )(x − 2c)2 − x2 .

This function attains a maximum when x = −2(1−c2 )/c and this maximum
is 4(1 − c2 ) ≤ 4. Therefore

|dK/Q (u)| ≤ 4(ρ3 + ρ−3 )2 + 16

= 4(u3 + u−3 + 6).

(c) Since d = |dK | < |dK/Q (u)|, we have

d d
u3 > − 6 − u−3 > − 7.
4 4
√
3
Exercise 8.1.12 Let α = 2, K = Q(α). Given that dK = −108:
(a) Show that, if u is the fundamental unit in K, u3 > 20.
(b) Show that β = (α − 1)−1 = α2 + α + 1 is a unit, 1 < β < u2 . Conclude that
β = u.

Solution. (a) By Exercise 8.1.11, u3 > 108/4 − 7 = 20 so u2 > 202/3 > 7.

(b) Computation shows that 17 < α−1 < 1 and therefore 1 < (α−1)−1 <
7 < u2 . Since β is a power of u, this power must be 1. Therefore β = u is
the fundamental unit in K.

Exercise 8.1.13 (a) Show that, if α ∈ K is a root of a monic polynomial f ∈

Z[x] and f (r) = ±1, for some r ∈ Z, then α − r is a unit in K.
√ 2
(b) Using the fact that if K = Q( 3 m), then dK = −27m √ , for any cubefree
3
integer m, determine the fundamental unit in K = Q( 7).
√
(c) Determine the fundamental unit in K = Q( 3 3).
284 CHAPTER 8. THE STRUCTURE OF UNITS

Solution. (a) Let g(x) = f (x + r). Then g(x) is a monic polynomial in

Z[x] with constant term g(0) = f (r) = ±1. Since g(α − r) = 0, the minimal
polynomial of α − r ∈ OK divides g(x) and, thus, has constant term ±1.
Therefore NK (α − r) = ±1 so α − r ∈ UK .
(b) dK = −27 · 72 implies that u3 > 27 · 72 /4 − 7 > 323 and so u2 >
(323)2/3 > 47 by Exercise 8.1.11. √
Let f (x) = x3 −√7 and note that f (2) = 1. Then
√
3
7 − 2 ∈ UK . Also,
2
1/u √ < 1/47 < 2 − 7 < 1 implies that 1√< (2 − 7) < u2 . Therefore,
3 3 −1

(2 − 3 7)−1 is the√fundamental unit of Q( 3 7).

(c) Let α = 3 3, K = Q(α). dK = −27 · 32 = −35 . We observe
that α2 is a root of f (x) = x3 − 9 and f (2) = −1 so α2 − 2 ∈ UK .
1
We have u3 > 35 /4 − 7 > 53 and thus u2 > 14. Since 14 < α2 − 1 <
2 2 2
1, 1 < (α − 2) < 14 < u . Thus, (α − 2) is the fundamental unit of
−1 −1

K = Q(α).

8.2 Units in Real Quadratic Fields

Exercise 8.2.1 (a) Consider the continued fraction [a0 , . . . , an ]. Define the se-
quences p0 , . . . , pn and q0 , . . . , qn recursively as follows:

p0 = a0 , q0 = 1,
p1 = a0 a1 + 1, q1 = a 1 ,
pk = ak pk−1 + pk−2 , qk = ak qk−1 + qk−2 ,

for k ≥ 2. Show that the kth convergent Ck = pk /qk .

(b) Show that pk qk−1 − pk−1 qk = (−1)k−1 , for k ≥ 1.
(c) Derive the identities
(−1)k−1
Ck − Ck−1 = ,
qk qk−1
for 1 ≤ k ≤ n, and
ak (−1)k
Ck − Ck−2 = ,
qk qk−2
for 2 ≤ k ≤ n.
(d) Show that
C1 > C3 > C5 > · · · ,
C0 < C2 < C4 < · · · ,
and that every odd-numbered convergent C2j+1 , j ≥ 0, is greater than every
even-numbered convergent C2k , k ≥ 0.

Solution. (a) We prove this by induction on k.

For k = 0, C0 = [a0 ] = p0 /q0 . For k = 1,

1 a0 a1 + 1 p1
C1 = [a0 , a1 ] = a0 + = = .
a1 a1 q1
8.2. UNITS IN REAL QUADRATIC FIELDS 285

For k > 1, suppose that

pk ak pk−1 + pk−2
Ck = = .
qk ak qk−1 + qk−2
Since pk−1 , pk−2 , qk−1 , qk−2 depend only on a0 , . . . , ak−1 ,
3 4
1
Ck+1 = a0 , a1 , . . . , ak−1 , ak +
ak+1
# 1
$
ak + ak+1 pk−1 + pk−2
= # 1
$
ak + ak+1 qk−1 + qk−2
ak+1 (ak pk−1 + pk−2 ) + pk−1
=
ak+1 (ak qk−1 + qk−2 ) + qk−1
ak+1 pk + pk−1 pk+1
= = .
ak+1 qk + qk−1 qk+1
(b) Again, we apply induction on k. For k = 1,
p1 q0 − p0 q1 = (a0 a1 + 1) · 1 − a0 a1 = 1.
For k ≥ 1,
pk+1 qk − pk qk+1 = (ak+1 pk + pk−1 )qk − pk (ak+1 qk + qk−1 )
= pk−1 qk − pk qk−1 = −(−1)k−1 = (−1)k ,
by our induction hypothesis.
(c) By (b),
pk qk−1 − qk pk−1 = (−1)k−1 .
Dividing by qk qk−1 , we obtain the first identity. Now,
pk pk−2 pk qk−2 − pk−2 qk
Ck − Ck−2 = − = .
qk qk−2 qk qk−2
But
pk qk−2 − pk−2 qk = (ak pk−1 + pk−2 )qk−2 − pk−2 (ak qk−1 + qk−2 )
= ak (pk−1 qk−2 − pk−2 qk−1 ) = ak (−1)k−2 ,
establishing the second identity.
(d) By (c),
ak (−1)k
Ck − Ck−2 = .
qk qk−2
Thus, Ck < Ck−2 , for k odd and Ck > Ck−2 , for k even. In addition,
(−1)2m−1
C2m − C2m−1 = <0 ⇒ C2m−1 > C2m ,
q2m q2m−1
⇒ C2k < C2(j+k+1) < C2(j+k)+1 < C2j+1 .
286 CHAPTER 8. THE STRUCTURE OF UNITS

Exercise 8.2.2 Let {ai }i≥0 be an infinite sequence of integers with ai ≥ 0 for
i ≥ 1 and let Ck = [a0 , . . . , ak ]. Show that the sequence {Ck } converges.

Solution. By Exercise 8.2.1 (d), we have

C1 > C3 > C5 > · · · .

Moreover, each C2j+1 > a0 so that the sequence {C2j+1 }j≥0 is decreasing
and bounded from below and is, thus, convergent, say limj→∞ C2j+1 = α1 .
Also,
C0 < C2 < C4 < · · ·

and C2j < C2k+1 for all j, k ≥ 0. In particular, each C2j < C1 . The
sequence {C2j }j≥0 is increasing and bounded from above and, therefore,
also converges, say limj→∞ C2j = α2 . We will show that α1 = α2 .
Since each ai ≥ 1, q0 , q1 ≥ 1, we easily see, by induction on k, that
qk = ak qk−1 + qk−2 ≥ 2k − 3. By Exercise 8.2.1 (c),

1 1
C2j+1 − C2j = ≤ → 0,
q2j+1 q2j (4j − 1)(4j − 3)

as j → ∞. Thus, both sequences converge to the same limit α = α1 = α2

and
lim Cj = α.
j→∞

Exercise 8.2.3 Let α = α0 be an irrational real number greater than 0. Define

the sequence {ai }i≥0 recursively as follows:

1
ak = [αk ], αk+1 = .
αk − ak

Show that α = [a0 , a1 , . . . ] is a representation of α as a simple continued fraction.

Solution. By induction on k, we easily see that each αk is irrational.

Therefore αk+1 > 1 which means that ak+1 ≥ 1 so that [a0 , a1 , . . . ] is a
simple continued fraction. Also,

1
α = α0 = [α0 ] + (α0 − [α0 ]) = a0 +
α1
= [a0 , α1 ] = [a0 , a1 , α2 ] = · · · = [a0 , a1 , . . . , ak , αk+1 ],

for all k. By Exercise 8.2.1 (a),

αk+1 pk + pk−1
α=
αk+1 qk + qk−1
8.2. UNITS IN REAL QUADRATIC FIELDS 287

so that
1 1
1 αk+1 pk + pk−1 pk 11
1
|α − Ck | = 1 − 1
αk+1 qk + qk−1 qk
1 1
1 −(pk qk−1 − pk−1 qk ) 1
1
= 1 1
(αk+1 qk + qk−1 )qk 1
1 1
1
1 1 1
1
= 1
(αk+1 qk + qk−1 )qk 1
1 1
< ≤ →0
qk2 (2k − 3)2
as k → ∞. Thus,
α = lim Ck = [a0 , a1 , . . . ].
k→∞

Exercise 8.2.5√ Let d be a positive integer, not a perfect square. Show that, if
|x2 −dy 2 | <√ d for positive integers x, y, then x/y is a convergent of the continued
fraction of d.
√
Solution. Suppose first that 0 < x2 − dy 2 < d. Then
√ √ √
(x + y d)(x − y d) > 0 ⇒ x > y d,
1 1 √
1√ 1 x √ x−y d
⇒ 1 d − x1 = − d=
1 y1 y y
2 2
x − dy x2 − dy 2 1
= √ < √ < 2.
y(x + y d) y(2y d) 2y
Thus,
√ by Theorem 8.2.4 (b), x/y is a convergent of the continued fraction
of d. √
Similarly, if − d < x2 − dy 2 < 0, then
1 1 x
0 < y 2 − x2 < √ ⇒ y>√ ,
d d d
1 1 √
1 1 1 1 y − x/ d
⇒ 1√ − y 1 =
y
−√ =
1 d x1 x d x
2 2
y − x /d y 2 − x2 /d 1
= √ < √ < 2.
x(y + x/ d) 2x / d
2 2x
√
Thus y/x is a convergent of the continued fraction of 1/ d.
Let α be any irrational number. Then α = [a0 , a1 , . . . ] implies 1/α =
[0, a0 , a1 , . . . ]. We therefore have that the (k + 1)th convergent of the con-
tinued fraction of 1/α is the reciprocal of the kth convergent of α, for all
k ≥ 0.
Using this fact, we √ find that, as before, x/y is a convergent of the
continued fraction of d.
288 CHAPTER 8. THE STRUCTURE OF UNITS

Exercise 8.2.6 Let α be a quadratic irrational (i.e, the minimal polynomial of

the real number α over Q has degree 2). Show that there are integers P0 , Q0 , d
such that √
P0 + d
α=
Q0
with Q0 |(d − P02 ). Recursively define
√
Pk + d
αk = ,
Qk
ak = [αk ],
Pk+1 = ak Qk − Pk ,
2
d − Pk+1
Qk+1 = ,
Qk
for k = 0, 1, 2, . . . . Show that [a0 , a1 , a2 , . . . ] is the simple continued fraction of
α.

Solution. There exist a, b, e, f ∈ Z, e, f > 0, e not a perfect square, such

that √ L
a+b e af + eb2 f 2
α= =
f f2
and, evidently, f 2 |(a2 f 2 − eb2 f 2 ). Set P0 = af, Q0 = f 2 , d = eb2 f 2 . This
sequence is well-defined, since d is not a perfect square ⇒ Qk != 0 for all k.
1
By Exercise 8.2.3, it will suffice to show that αk+1 = αk −a k
for all k.
√
Pk + d
αk − ak = − ak
Qk
√
d − (ak Qk − Pk )
=
Qk
√
d − Pk+1
=
Qk
2
d − Pk+1
= √
Qk ( d + Pk+1 )
Qk Qk+1
= √
Qk ( d + Pk+1 )
1
= .
αk+1

Exercise 8.2.7 Show that the simple continued fraction expansion of a quadratic
irrational α is periodic.

Solution. By Exercise 8.2.6, we may write

√
P0 + d
α=
Q0
8.2. UNITS IN REAL QUADRATIC FIELDS 289

with Q0 |(P02 − d). Setting

√
Pk + d
αk = ,
Qk
ak = [αk ],
Pk+1 = ak Qk − Pk ,
2
d − Pk+1
Qk+1 = ,
Qk

we have

P0 = 0, P1 = d,
Q0 = 1, Q1 = 1,
L L
α0 = d2 + 1, α1 = d + d2 + 1,
a0 = d, a1 = 2d.

and α = [a0 , a1 , . . . ]. Now,

αk pk−1 + pk−2
α=
αk qk−1 + qk−2

and if α$ denotes the Q-conjugate of α,

! "
α$ pk−1 + pk−2 qk−2 α$ − Ck−2
α = k$
$
⇒ αk$ =− .
αk qk−1 + qk−2 qk−1 α$ − Ck−1

Since Ck−1 , Ck−2 → α as k → ∞,

α$ − Ck−2
→ 1.
α$ − Ck−1
√
Therefore, αk$ < 0, and, since αk > 0, αk − αk$ = 2 d/Qk > 0, for all
2
sufficiently large k. We also have Qk Qk+1 = d − Pk+1 so

2
Qk ≤ Qk Qk+1 = d − Pk+1 ≤d

and
2
Pk+1 ≤ d − Qk ≤ d

for sufficiently large k. Thus there are only finitely many possible values
for Pk , Qk and we conclude that there exist integers i < j such that Pi =
Pj , Qi = Qj . Then ai = aj and, since the ai are defined recursively, we
have
α = [a0 , a1 , . . . , ai−1 , ai , . . . , aj−1 ].
290 CHAPTER 8. THE STRUCTURE OF UNITS

√ Show that, if d is a positive integer but not a perfect square,

Exercise 8.2.8
and α = α0 = d, then
p2k−1 − dqk−1
2
= (−1)k Qk ,
for all k ≥ 1, where pk /qk is the kth convergent of the continued fraction of α
and Qk is as defined in Exercise 8.2.6.
√
Solution. By inspection, p20 − dq02 = [ d]2 − d = −Q1 . Now, suppose that
k ≥ 2. Writing
√ αk pk−1 + pk−2
d = α0 = [a0 , a1 , . . . , ak−1 , αk ] =
αk qk−1 + qk−2
√
and recalling that αk = (Pk + d)/Qk , we have
√
√ (Pk + d)pk−1 + Qk pk−2
d= √ ,
(Pk + d)qk−1 + Qk qk−2
√ √
⇒ dqk−1 + (Pk qk−1 + Qk qk−2 ) d = Pk pk−1 + Qk pk−2 + pk−1 d.
√
Equating coefficients in Q( d), we have
dqk−1 = Pk pk−1 + Qk pk−2
and
pk−1 = Pk qk−1 + Qk qk−2 .
Computation yields
p2k−1 − dqk−1
2
= (pk−1 qk−2 − pk−2 qk−1 )Qk = (−1)k Qk .
√ √
Exercise 8.2.10 (a) Find the simple continued fractions of 6, 23.
√
√ Theorem 8.2.9 (c), compute the fundamental unit in both Q( 6) and
(b) Using
Q( 23).
√
Solution. (a) Using notation of previous exercises, setting α = α0 = 6,
we have
P1 = 2,
P0 = 0, P2 = 2,
Q1 = 2,
Q0 = 1, √ Q2 = 1,
√ 2+ 6 √
α0 = 6, α1 = α2 = 2 + 6,
2
a0 = 2, a2 = 4.
a1 = 2,
√
Thus, the period of the continued fraction of α√is 2 ⇒ 6 = [a0 , a1 , a2 ] =
[2, 2, 4]. Applying
√ the same procedure, we find 23 = [4, 1, 3, 1, 8].
(b) For 6, C1 = p1 /q1 √ = [a0 , a1 ] =√a0 + 1/a1 = 2 + 1/2 = 5/2. Thus,
the fundamental
√ unit in Q( 6) is 5 + 2 6.
√For 23, C3 =
√ [4, 1, 3, 1] = 24/5. Therefore the fundamental unit in
Q( 23) is 24 + 5 23.
8.3. SUPPLEMENTARY PROBLEMS 291

√
Exercise 8.2.11 (a) Show that [d, 2d] is the continued fraction of d2 + 1.
2
(b) Conclude that,√if d + 1 is squarefree,
√ d ≡ 1, 3 (mod 4), then the fundamen-
tal√unit of√Q( d2 + √ 1) is d + d + 1. Compute the fundamental unit of
2

Q( 2), Q( 10), Q( 26).

√
(c) Show that the continued fraction of d2 + 2 is [d, d, 2d].
√
(d) Conclude that,
√ if d2 +2 is squarefree, then the fundamental unit √
of Q( d2√+ 2)
2
is d
√ + 1 + d d√+ 2. Compute the fundamental unit in Q( 3), Q( 11),
2

Q( 51), and Q( 66).

Solution.
√ (a) Observing that d2 < d2 + √
1 < (d + 1)2 for all d > 0, we see
that [ d2 + 1] = d and setting α = α0 = d2 + 1, we have
P0 = 0, P1 = d,
Q0 = 1, Q1 = 1,
L L
α0 = d2 + 1, α1 = d + d2 + 1,
a0 = d, a1 = 2d.
√
This implies
√ that the period of the continued fraction of d + 1 is 1.
2

Therefore d + 1 = [a0 , a1 ] = [d, 2d].

2

(b) d ≡ 1, 3 (mod 4) and thus d2 + 1 ≡ √2 (mod 4). Thus, if√d2 + 1 is

squarefree,
√ then the fundamental unit of Q( d2 + 1) is p0 + q0 d2 + 1 =
d + d + 1.
2
√
(c) Observing that d2√< d2 +2 < (d+1)2 for all d ≥ 1 we get [ d2 + 2] =
d and setting α = α0 = d2 + 2, we have

P0 = 0, P1 = d, P2 = d,
Q0 = 1, Q1 = 2, Q2 = 1,
L √ L
d+ d2 + 2
α0 = d2 + 2, α1 = α2 = d + d2 + 2,
2
a0 = d, a1 = d, a2 = 2d.
√
Therefore the period of the continued fraction of d2 + 2 is 2, so
L
d2 + 2 = [a0 , a1 , a2 ] = [d, d, 2d]
and thus
p1 1 d2 + 1
=d+ = .
q1 d d
(d) For√all d, d2 +2 ≡ 2, 3 (mod
√ 4) so, if d is squarefree,
√ the fundamental
unit in Q( d2 + 2) is p1 + q1 d2 + 2 = d2 + 1 + d d2 + 2.

8.3 Supplementary Problems

√
Exercise √8.3.1 If n2 − 1 is squarefree, show that n + n2 − 1 is the fundamental
unit of Q( n2 − 1).
292 CHAPTER 8. THE STRUCTURE OF UNITS
√
Solution. The continued fraction of n2 − 1 is [n − 1, 1, 2(n − 1)], so that
the first two convergents are (n − 1)/1 and n/1. Now apply Theorem 8.2.9.

Exercise 8.3.2 Determine the units of an imaginary quadratic field from first
principles.
√
Solution. We have already determined the units of Q(i) and Q( −3).
2
They are ±1, ±i and ±1, ±ρ, √ ±ρ , respectively. Now we determine the
√ field Q( −d) where −d ≡ 2, 3 (mod 4). All units are
units of an arbitrary
of the form a + b −d with a2 + db2 = 1. It is easy to see that since d ≥ 2,
this has no solution except√for a = ±1, b = 0. If −d ≡ 1 (mod 4), then units
will be of the form (a + b −d)/2 where a ≡ b (mod 2) and a2 + db2 = 4.
Since we already know the units for d = 3, then d ≥ 7 and once again, the
only solution is a = ±1, b = 0.

Exercise
√ 8.3.3 Suppose that 22n √
+ 1 = dy 2 with d √
squarefree.
√ Show that 2 +
n

y d is the fundamental unit of Q( d), whenever Q( d) #= Q( 5).

√
Solution. Suppose not. Let (a + b d)/2 be the fundamental unit. Then
/ √ 0j
√ a+b d
2 +y d=
n
2

for some integer j. If an odd prime p divides j, we can write

/ √ 0p
√ u+v d
2 +y d=
n
,
2
√
where (u + v d)/2 is again a unit. Hence
*! p "
2n+p = v 2k dk up−2k = uA,
2k
k

say. Clearly, (u, A) = 1 so we must have either u = 1, A = 2n+p or

u = 2n+p , A = 1. The latter case is impossible since

A = pv p−1 d(p−1)/2 > 1.

Hence u = 1. But then dv 2 = 5 or −3. The former case is√ruled out by

the hypothesis and the latter is impossible. Thus, if 2n + y d is not the
fundamental unit, it must be the square of a unit. But then, this means
that √ √ √
2n+4 + 4y d = (u + v d)2 = (u2 + dv 2 ) + 2uv d
and this case is also easily ruled out.
√
Exercise 8.3.4 (a) Determine the continued√fraction expansion of 51 and use
it to obtain the fundamental unit ε of Q( 51).
8.3. SUPPLEMENTARY PROBLEMS 293

√
(b) Prove from first principles that all units of Q( 51) are given by εn , n ∈ Z.

Solution. See Exercise 8.2.11.

Exercise 8.3.5 Determine a unit #= ±1 in the ring of integers of Q(θ) where

θ3 + 6θ + 8 = 0.

Solution. Consider η = 1 + θ.

Exercise 8.3.6 Let p be an odd prime > 3 and suppose that it does not divide
the class number of Q(ζp ). Show that

xp + y p + z p = 0

is impossible for integers x, y, z such that p ! xyz.

Solution. We factor
p−1
6
xp + y p = (x + ζpi y) = −z p .
i=0

Since p ! xyz, the terms (x + ζpi y) are mutually coprime for 0 ≤ i ≤ p − 1.

Moreover, viewing the equation as an ideal equation gives

(x + ζpi y) = api

for some ideal ai . Since p does not divide the class number, ai itself must
be principal. Therefore
(x + ζpi y) = εαp ,
where ε is a unit in Q(ζp ) and α is an integer of Q(ζp ).
By Exercise 4.3.7, Z[ζp ] is the ring of integers of Q(ζp ) and

1, ζp , ζp2 , . . . , ζpp−2

is an integral basis. Therefore, we can express

α = a0 + a1 ζp + · · · + ap−2 ζpp−2 ,

so that
αp ≡ ap0 + · · · + app−2 ≡ a (mod p),
where a = a0 +· · ·+ap−2 , by a simple application of Fermat’s little Theorem
(Exercise 1.1.13). Also, we may write ε = ζps η where 0 ≤ s < p and η is a
real unit, by Theorem 8.1.10. Hence, for i = 1,

x + ζp y ≡ ζps β (mod p),

where β is a real integer of Q(ζp ). Also, by complex conjugation,

x + ζp−1 y ≡ ζp−s β (mod p).

294 CHAPTER 8. THE STRUCTURE OF UNITS

Then
ζp−s (x + ζp y) − ζps (x + ζp−1 y) ≡ 0 (mod p).
Since 1, ζp , . . . , ζpp−2 is an integral basis for Q(ζp ),

c0 + c1 ζp + · · · + cp−2 ζpp−2 ≡ 0 (mod p)

holds if and only if

c0 ≡ c1 ≡ · · · ≡ cp−2 ≡ 0 (mod p).

Write ζpp−s (x + ζp y) − (ζps x + ζps−1 y) ≡ 0 (mod p). If 2 < s ≤ p − 2 and

s != (p + 1)/2, the powers of ζp are all different and less than p − 1 in the
above congruence. Hence x ≡ y ≡ 0 (mod p), a contradiction.
If s = (p + 1)/2, then x + ζp y ≡ ζp x + y (mod p) so that x ≡ y (mod p).
Similarly, x ≡ z (mod p) so that

xp + y p + z p ≡ 3xp ≡ 0 (mod p).

Since p > 3, we get p | x, a contradiction.

If s = 0, then x + ζp y − (x + ζp−1 y) ≡ 0 (mod p) so that y ≡ 0 (mod p),
a contradiction.
If s = 1, ζp−1 x + y ≡ ζp x + y (mod p) and so p | x, a contradiction.
If s = 2, x + ζp y ≡ ζp4 x + ζp3 y and so x ≡ y ≡ 0 (mod p), again a
contradiction.
Finally, if s = p − 1,

ζp (x + ζp y) ≡ ζp−1 x + ζp−2 y (mod p),

that is,
ζp3 (x + ζp y) ≡ ζp x + y (mod p),
which gives p | x (since p ≥ 5), again a contradiction.

Exercise 8.3.7 Let K be a quadratic field of discriminant d. Let P0 denote

the group of principal fractional ideals αOK with α ∈ K satisfying NK (α) > 0.
The quotient group H0 of all nonzero fractional ideals modulo P0 is called the
restricted class group of K. Show that H0 is a subgroup of the ideal class group
H of K and [H : H0 ] ≤ 2.

Solution. If d < 0, the norm of any nonzero element is greater than 0 and
so the notions of restricted class group and ideal class group coincide. If
d > 0, then H = √ H0 if and only if there exists a unit in K of norm −1.
This
√ is because dOK is in the same coset of OK (mod√P0 ) if and only if
d = εα with NK (α) > 0, and ε is a unit. Since NK ( d) < 0, this can
happen if and only if ε is a unit of norm −1.
8.3. SUPPLEMENTARY PROBLEMS 295

Exercise 8.3.8 Given an ideal a of a quadratic field K, let a( denote the conju-
gate ideal. If K has discriminant d, write

|d| = pα1
1 p2 · · · pt ,

where p1 = 2, α1 = 0, 2, or 3 and p2 , . . . , pt are distinct odd primes. If we write

pi OK = ℘2i show that for any ideal a of OK satisfying a = a( we can write

a = r℘a1 1 · · · ℘at t ,

r > 0, ai = 0, 1 uniquely.

Solution. We first factor a as a product of prime ideals. The fact that

a = a$ implies that if a prime ideal ℘ divides a, so does ℘$ . If ℘ is inert then
it is principal and generated by a rational prime. If ℘ splits, then ℘ and ℘$
(which both occur in a with the same multiplicity) can be paired to give
again a principal ideal generated by a rational prime. Only the ramified
prime ideals cannot be so paired. Since ℘2i = pi OK , we see immediately
that a has a factorization of the form described above. If we have another
factorization
a = r1 ℘ b 1 · · · ℘ b t , r1 > 0, bi = 0, 1,
then taking norms we obtain r2 pa1 1 · · · pat t = r12 pb11 · · · ptbt so that ai ≡ bi
(mod 2). Since ai = 0 or 1, and bi = 0 or 1, this means ai = bi . Hence
r = r1 .

Exercise 8.3.9 An ideal class C of H0 is said to be ambiguous if C 2 = 1 in H0 .

Show that any ambiguous ideal class is equivalent (in the restricted sense) to one
of the at most 2t ideal classes

℘a1 1 · · · ℘at t , ai = 0, 1.

Solution. Let a be an ideal lying in an ambiguous class. Then a2 = (α)

with NK (α) > 0. But we have aa$ = (NK (a)). Therefore

aa$ = (NK (a)/α)(α) = (NK (a)/α)a2

so that a = a$ . By the previous question, a can be written as

r℘a1 1 · · · ℘at t ,

r > 0, ai = 0, 1. In the restricted class group, these form at most 2t ideal

classes.

Exercise 8.3.10 With the notation as in the previous two questions, show that
there is exactly one relation of the form

℘a1 1 · · · ℘at t = ρOK , NK (ρ) > 0,

&t
with ai = 0 or 1, i=1 ai > 0.
296 CHAPTER 8. THE STRUCTURE OF UNITS

Solution. We first √ show there is at least one such relation. If d < 0,

then ℘1 · · · ℘t = ( d). If d > 0, let ε be a generator of √ all the totally
positive√units of K (i.e., units ε >√0 and ε$ > 0). Set µ = d(1 − ε). Then
µ$ = − d(1 − ε$ ) so that εµ$ = − d(ε − 1) = µ. Therefore (µ) = (µ$ ) and
by the penultimate question, we have the result.
To establish uniqueness, consider first the case d < 0. If

℘a1 1 · · · ℘at t = ρOK ,

NK (ρ) > 0,
√
then (ρ) = (ρ$ ) since √
℘$i = ℘i . Therefore
√ ρ = b d for some b ∈ Q. The
other two cases of Q( −1) are Q( −3) are similarly dealt with. If now
d > 0, then ρ = ηρ$ with η a totally positive unit. We write η = εm for
some m. With µ defined as above, notice that ρ/µm = ρ$ /(µ$ )m so that
ρ/µm = r ∈ Q. Hence (ρ) = (µ)m . Since (µ) has order 2 in the (restricted)
class group we are done.

Exercise 8.3.11 Let K be a quadratic field of discriminant d. Show that the

number of ambiguous ideal classes is 2t−1 where t is the number of distinct primes
dividing d. Deduce that 2t−1 divides the order of the class group.

Solution. This is now immediate from the previous two exercises. Since
the restricted class group has index 1 or 2 in the ideal class group, the
divisibility assertion follows.

Exercise 8.3.12 If K is a quadratic field of discriminant d and class number 1,

show that d is prime or d = 4 or 8.

Solution. If d has t distinct prime divisors, then 2t−1 divides the class
number. Thus t ≤ 1. Since the discriminant is either squarefree or four
times a squarefree number, the result is now clear.

Exercise 8.3.13 If a real quadratic field K has odd class number, show that K
has a unit of norm −1.

Solution. Since K has odd class number, H = H0 . This means there is a

unit of norm −1.
√ √
Exercise 8.3.14 Show that 15 + 4 14 is the fundamental unit of Q( 14).
√
Solution. The continued fraction development of 14 is

[3, 1, 2, 1, 6]

and the convergents are easily computed:

3 4 11 15
, , , ,... .
1 1 3 4
√
By Theorem 8.2.9, we find that 15 + 4 14 is the fundamental unit.
8.3. SUPPLEMENTARY PROBLEMS 297

√
Exercise 8.3.15 In Chapter 6 we showed that Z[ 14] is a PID√(principal ideal
domain). Assume the following hypothesis: given α, β ∈ Z[ 14], such that
gcd(α, β) √
= 1, there is a prime π ≡ α (mod β) for which the fundamental √unit
ε = 15 + 4 14 generates the coprime residue classes (mod π). Show that Z[ 14]
is Euclidean.

√ We define the Euclidean algorithm inductively as follows. For

Solution.
u ∈ Z[ 14], define ϕ(u) = 1. If ε generates the residue classes (mod π),
where π is prime, define ϕ(π) = 2. If β is a prime for which ε does not
generate
√ the residue classes (mod β), define ϕ(β) = 3. Now for any γ ∈
Z[ 14], factor
γ = π1e1 · · · πrer β1f1 · · · βsfs ,
where ϕ(πi ) = 2, ϕ(βi ) = 3. Define

ϕ(γ) = 2(e1 + · · · + er ) + 3(f1 + · · · + fs ).

An easy induction argument using the Chinese Remainder Theorem (The-

orem 5.3.13) shows that ϕ is a Euclidean algorithm.

Exercise 8.3.16 Let d = a2 + 1. Show that if |u2 − dv 2 | =

# 0, 1 for integers u, v,
then √
|u2 − dv 2 | > d.
√
Solution. The continued fraction expansion of d is

[a, 2a, 2a, . . . ]

√
and the convergents pk /qk always
√ satisfy p2k − dqk2 = ±1. If |u2 − dv 2 | < d,
then u/v is a convergent of d, by Exercise 8.2.5. Since |u2 − dv 2 | = ! 0, 1,
we are done.
2g
Exercise 8.3.17 Suppose that n √is odd, n ≥ 5, and that n +1 = d is squarefree.
Show that the class group of Q( d) has an element of order 2g.
2g
√ √
Solution. We√ have (n) =√( d − 1)( d + 1). Since n is odd, each of the
ideals (−1 + d) and (1 + d) must be coprime. By Exercise 5.3.12, each
of them must be a 2gth power. Therefore
√
a2g = ( d − 1)

and √
(a$ )2g = ( d + 1).
2g
√ to g. Observe that n +1
We claim that a has order greater than or equal √ ≡
2 (mod 4) because n is odd. Therefore 1, d is an integral basis of Q( d).
If am were principal, then
√
am = (u + v d)
298 CHAPTER 8. THE STRUCTURE OF UNITS

implies that
nm = |u2 − dv 2 |,
√
which by the previous exercise is either 0, 1 or > d. The former cannot
hold since n ≥ 5. Thus,
√ L
nm > d = n2g + 1 > ng .

Hence, m > g. Since m | 2g, we must have that a has order 2g.
It is conjectured that there are infinitely many squarefree numbers of
the form n2g + 1. Thus, this argument does not establish that there are
infinitely many real quadratic fields whose class number is divisible by g.
However, by a simple modification of this argument, we can derive such a
result. We leave it as an exercise to the interested reader.
Chapter 9

Higher Reciprocity Laws

9.1 Cubic Reciprocity

Exercise 9.1.1 If π is a prime of Z[ρ], show that N (π) is a rational prime or the
square of a rational prime.

Solution. Let N (π) = n > 1. Then ππ = n. Now n is a product of

rational prime divisors. Since π is prime, π | p for some rational prime p.
Write p = πγ. Then N (p) = N (π)N (γ) = p2 . Thus, either N (π) = p or
N (π) = p2 .

Exercise 9.1.2 If π ∈ Z[ρ] is such that N (π) = p, a rational prime, show that π
is a prime of Z[ρ].

Solution. If π factored in Z[ρ], then π = αβ and p = N (π) = N (α)N (β)

which implies that N (α) = 1 or N (β) = 1 so that π cannot be factored
nontrivially in Z[ρ].

Exercise 9.1.3 If p is a rational prime congruent to 2 (mod 3), show that p is

prime in Z[ρ]. If p ≡ 1 (mod 3), show that p = ππ where π is prime in Z[ρ].

Solution. Let p ≡ 2 (mod 3) be a rational prime. If p = πγ, with

N (γ), N (π) > 1, then p2 = N (π)N (γ) implies that N (π) = p and N (γ) = p.
Writing π = a + bρ, we find p = N (π) = a2 − ab + b2 so that

4p = 4a2 − 4ab + 4b2 = (2a − b)2 + 3b2 .

Hence p ≡ (2a − b)2 (mod 3), a contradiction since 2 is not a square mod
3.
Finally, if p ≡ 1 (mod 3), then by quadratic reciprocity:
! " ! "! " E F ! "
−3 −1 3 p 1
= = = =1
p p p 3 3

299
300 CHAPTER 9. HIGHER RECIPROCITY LAWS

so that x2 ≡ −3 (mod p) has a √solution. √Hence py = x2 + 3 for some

x, y ∈ Z. Therefore p divides (x+ −3)(x− −3) = (x+1+2ρ)(x−1−2ρ).
If p were prime in Z[ρ], it would divide one of these two factors, which is
not the case. Thus p = αβ for some α, β ∈ Z[ρ] and N (α) > 1, N (β) > 1.
Hence N (α) = p so that αα = p. Moreover, α is prime by Exercise 9.1.2.
Recall that in Section 2.3 we found that 3 = −ρ2 (1 − ρ)2 and (1 − ρ) is
irreducible, so that 3 is not a prime in Z[ρ].

Exercise 9.1.4 Let π be a prime of Z[ρ]. Show that αN (π)−1 ≡ 1 (mod π) for
all α ∈ Z[ρ] which are coprime to π.

Solution. Since π is prime, the ideal (π) is prime. Hence Z[ρ]/(π) is

a field, containing N (π) elements. Its multiplicative group, consisting of
classes coprime to π, has N (π) − 1 elements. Thus, by Lagrange’s theorem,
the result is immediate.

Exercise 9.1.5 Let π be a prime not associated to (1 − ρ). First show that
3 | N (π) − 1. If (α, π) = 1, show that there is a unique integer m = 0, 1 or 2 such
that
α(N (π)−1)/3 ≡ ρm (mod π).

Solution. By Exercise 9.1.3, we know that π | (αN (π)−1 − 1). By Exer-

cise 9.1.4, we know N (π) ≡ 1 (mod 3). Thus, we can write β = α(N (π)−1)/3
and observe that

β 3 − 1 = (β − 1)(β − ρ)(β − ρ2 ).

Since π is prime and divides β 3 − 1, it must divide one of the three factors
on the right. If π divides at least two factors, then π | (1 − ρ) which means
π is an associate of 1 − ρ, contrary to assumption. Thus, β ≡ 1, ρ, or ρ2
(mod π) as desired.

Exercise 9.1.6 Show that:

(a) (α/π)3 = 1 if and only if x3 ≡ α (mod π) is solvable in Z[ρ];
(b) (αβ/π)3 = (α/π)3 (β/π)3 ; and
(c) If α ≡ β (mod π), then (α/π)3 = (β/π)3 .

Solution. Clearly if x3 ≡ α (mod π) has a solution, then by Exercise 9.1.4,

α(N (π)−1)/3 ≡ 1 (mod π) so that (α/π)3 = 1. For the converse, let g be a
primitive root of Z[ρ]/(π). Then, writing α = g r we find g (rN (π)−1)/3 ≡ 1
(mod π) so that

r(N (π) − 1)
≡0 (mod N (π) − 1).
3
Hence 3 | r, and α is a cube mod π. That is, x3 ≡ α (mod π) has a solution.
This proves (a).
9.1. CUBIC RECIPROCITY 301

For (b),
! "
αβ
≡ (αβ)(N (π)−1)/3 ≡ α(N (π)−1)/3 β (N (π)−1)/3
π 3
EαF !β "
≡ (mod π).
π 3 π 3

For (c), if α ≡ β (mod π), then

EαF ! "
β
≡ α(N (π)−1)/3 ≡ β (N (π)−1)/3 ≡ (mod π).
π 3 π 3

Exercise 9.1.7 Show that:

(a) χπ (α) = χπ (α)2 = χπ (α2 ); and
(b) χπ (α) = χπ (α).

Solution. χπ (α) is by definition one of 1, ρ, or ρ2 so that (a) is immediate.

For (b), observe that

α(N (π)−1)/3 ≡ χπ (α) (mod π)

implies
α(N (π)−1)/3 ≡ χπ (α) (mod π)
on the one hand. On the other hand,

α(N (π)−1)/3 ≡ χπ (α) (mod π)

by definition. Part (b) is now immediate.

Exercise 9.1.8 If q ≡ 2 (mod 3), show that χq (α) = χq (α2 ) and χq (n) = 1 if n
is a rational integer coprime to q.

Solution. Since q = q,

χq (α) = χq (α) = χq (α) = χq (α2 )

by the previous exercise. Also,

χq (n) = χq (n) = χq (n2 ) = χq (n)2 .

Since χq (n) != 0, we deduce χq (n) = 1.

Exercise 9.1.9 Let N (π) = p ≡ 1 (mod 3). Among the associates of π, show
there is a unique one which is primary.
302 CHAPTER 9. HIGHER RECIPROCITY LAWS

Solution. Write π = a + bρ. All the associates of π can be written down:

a + bρ, −b + (a − b)ρ, (b − a) − aρ, −a − bρ, b + (b − a)ρ, (a − b) + aρ. Since
p = a2 − ab + b2 , not both a and b are divisible by 3.
If b ≡ 0 (mod 3), then π ≡ 1 (mod 3) implies a ≡ 1 (mod 3) so that
−a ≡ 2 (mod 3) and hence −a − bρ is the primary. If a ≡ 0 (mod 3) then
one of (b − a) − aρ or (a − b) + aρ is primary. If both a and b are coprime
to 3, then we must have a ≡ b (mod 3), for otherwise 3 | p, contrary to
assumption. Thus a ≡ b ≡ ±1 (mod 3), so that one of b + (b − a)ρ or
−b + (a − b)ρ is primary, as desired.

Exercise 9.1.11 If χ1 , . . . , χr are nontrivial and the product χ1 · · · χr is also

nontrivial prove that g(χ1 ) · · · g(χr ) = J(χ1 , . . . , χr )g(χ1 · · · χr ).

Solution. Define ψ : Fp →5C by ψ(t) = ζ t . Then ψ(t1 + t2 ) = ψ(t1 )ψ(t2 )

and we can write g(χ) = t χ(t)ψ(t). This is just for notational conve-
nience. Now
/ 0 / 0
* *
g(χ1 ) · · · g(χr ) = χ1 (t1 )ψ(t1 ) · · · χr (tr )ψ(tr )
t1 tr
/ 0
* *
= ψ(s) χ1 (t1 ) · · · χr (tr ) .
s t1 +···+tr =s

If s != 0, writing ti = sui , the inner sum becomes

(χ1 · · · χr )(s)J(χ1 , . . . , χr ).
If s = 0, then the inner sum is
*
χ1 (t1 ) · · · χr (tr ) = 0
t1 +···+tr =0

since t1 , . . . , tr−1 can be chosen arbitrarily so that tr = −t1 − · · · − tr−1 ,

and each of the sums corresponding to t1 , . . . , tr−1 is zero since χ1 , . . . , χr
are nontrivial. This completes the proof.

Exercise 9.1.12 If χ1 , . . . , χr are nontrivial, and χ1 · · · χr is trivial, show that

g(χ1 ) · · · g(χr ) = χr (−1)pJ(χ1 , . . . , χr−1 ).
Solution. By the previous exercise,
g(χ1 ) · · · g(χr−1 ) = J(χ1 , . . . , χr−1 )g(χ1 · · · χr−1 ).
Multiplying both sides of the equation by g(χr ) leads us to evaluate
g(χr )g(χ1 · · · χr−1 ).
However, (χ1 · · · χr−1 )χr = χ0 means that
g(χ1 · · · χr−1 ) = g(χr ) = χr (−1)p
by the proof of Theorem 9.1.10.
9.2. EISENSTEIN RECIPROCITY 303

Exercise 9.1.14 Show that g(χ)3 = pπ.

Solution. By Exercise 9.1.12, g(χ)3 = pJ(χ, χ) and by the previous exer-

cise, J(χ, χ) = π.

Exercise 9.1.17 Let π be a prime of Z[ρ]. Show that x3 ≡ 2 (mod π) has a

solution if and only if π ≡ 1 (mod 2).

Solution. We first observe that x3 ≡ 2 (mod π) is solvable if and only if

x3 ≡ 2 (mod π $ ) is solvable for any associate of π. Thus we may assume
that π is primary.
If π = q is a rational prime, then χq (2) = χ2 (q) = χ2 (1) = 1 so that 2
is a cubic residue for all such primes. If π = a + bρ is primary, by cubic
reciprocity, χπ (2) = χ2 (π). The norm of (2) is 4 and

π = π (4−1)/3 ≡ χ2 (π) (mod 2).

Thus χπ (2) = χ2 (π) = 1 if and only if π ≡ 1 (mod 2).

9.2 Eisenstein Reciprocity

2
Exercise 9.2.1 Show that q ≡ 1 (mod m) and that 1, ζm , ζm m−1
, . . . , ζm are dis-
tinct coset representatives mod ℘.

Solution. Observe that

m−1
6
xm − 1
= 1 + x + · · · + xm−1 = (x − ζm
i
).
x−1 i=1

Putting x = 1 in this identity gives

m−1
6
m= (1 − ζm
i
).
i=1

If ζm
i j
≡ ζm (mod ℘) (say), then ζm
j−i
≡ 1 (mod ℘) so that m ≡ 0 (mod ℘),
contrary to m !∈ ℘. Thus 1, ζm , . . . , ζm
m−1
are distinct mod ℘. Moreover,
the cosets#they represent
$∗ form a multiplicative subgroup of Z[ζm ]/℘ of order
m. Since Z[ζm ]/℘ has order q − 1 = N (℘) − 1, we must have m | q − 1.

Exercise 9.2.2 Let α ∈ Z[ζm ], α #∈ ℘. Show that there is a unique integer i

(modulo m) such that
α(q−1)/m ≡ ζm i
(mod ℘).
# $∗
Solution. Since Z[ζm ]/℘ has q−1 elements, we have αq−1 ≡ 1 (mod ℘).
Thus
m−1
6
(α(q−1)/m − ζm
i
) ≡ 0 (mod ℘).
i=1
304 CHAPTER 9. HIGHER RECIPROCITY LAWS

Since ℘ is a prime ideal, there is an integer i, 0 ≤ i < m, such that

α(q−1)/m ≡ ζm
i
(mod ℘).

This i is unique by Exercise 9.2.1.

Exercise 9.2.3 Show that:

(a) (α/℘)m = 1 if and only if xm ≡ α (mod ℘) is solvable in Z[ζm ];
N (℘)−1
(b) for all α ∈ Z[ζm ], α m ≡ (α/℘)m (mod ℘);
(c) (αβ/℘)m = (α/℘)m (β/℘)m ; and
(d) if α ≡ β (mod ℘), then (α/℘)m = (β/℘)m .

Solution. If xm ≡ α (mod ℘) has a solution, then

α(N (℘)−1)/m ≡ xN (℘)−1 ≡ 1 (mod ℘)

by the analogue of Fermat’s

# little
$∗ Theorem. Thus, (α/℘)m = 1. For the
converse, we know that Z[ζm ]/℘ is cyclic, being the multiplicative group
of a finite field. Let g be a generator, and set α = g r . If (α/℘) = 1, then
α(q−1)/m ≡ 1 (mod ℘). Hence g r(q−1)/m ≡ 1 (mod ℘). Since g has order
q − 1, we must have
q−1
r ≡0 (mod q − 1).
m
Hence, m | r, so that α is an mth power. This proves (a).
For (b), we need only note the case α ≡ 0 (mod ℘) which is clear. Parts
(c) and (d) are proved exactly as in the case of the cubic residue symbol in
Exercise 9.1.6.

Exercise 9.2.4 If ℘ is a prime ideal of Z[ζm ] not containing m show that

$ %
ζm (N (℘)−1)/m
= ζm .
℘ m

Solution. By definition,
! "
ζm (N (℘)−1)/m
≡ ζm (mod ℘).
℘ m
(N (℘)−1)/m
Since both (ζm /℘)m and ζm are mth roots of unity and by Exer-
cise 9.2.1, distinct roots represent distinct classes, we must have
! "
ζm (N (℘)−1)/m
= ζm .
℘ m

Exercise 9.2.5 Suppose a and b are ideals coprime to (m). Show that:
9.2. EISENSTEIN RECIPROCITY 305

(a) (αβ/a)m = (α/a)m (β/a)m ;

(b) (α/ab)m = (α/a)m (β/b)m ; and
(c) if α is prime to a and xm ≡ α (mod a) is solvable in Z[ζm ], then (α/a)m = 1.

Solution. Parts (a) and (b) are immediate from Exercise 9.2.3. For (c), we
note that xm ≡ α (mod ℘) has a solution for every prime ideal ℘ dividing
a. Thus by Exercise 9.2.3 (a), (α/℘)m = 1 for every prime ideal ℘ dividing
a. thus, (α/a)m = 1.

Exercise 9.2.6 Show that the converse of (c) in the previous exercise is not
necessarily true.

Solution. Choose two distinct prime ideals ℘1 , ℘2 coprime to (m). The

map

Z[ζm ]/℘1→ Z[ζm ]/℘1 ,

β (mod ℘1 ) 9→ β (N (℘1 )−1)/m (mod ℘1 ),

is a homomorphism with kernel consisting of the subgroup of mth powers

by Exercise 9.2.3 (a). This subgroup has size (N (℘1 ) − 1)/m. Hence the
image is the subgroup of mth roots of unity. Thus, we can find β, γ so that

β (N (℘1 )−1)/m ≡ ζm (mod ℘1 ),

(N (℘2 )−1)/m
γ ≡ m−1
ζm (mod ℘2 ).

By the Chinese Remainder Theorem (Theorem 5.3.13), we can find

α ≡ β (mod ℘1 ),
α ≡ γ (mod ℘2 ).

Then (α/℘1 )m = (β/℘1 )m = ζm , (α/℘2 )m = (γ/℘2 )m = ζm m−1

, and there-
fore (α/℘1 ℘2 )m = 1. But then x ≡ α (mod ℘1 ℘2 ) has no solution because
m

neither xm ≡ α (mod ℘1 ) nor xm ≡ α (mod ℘2 ) has a solution.

Exercise 9.2.7 If α ∈ Z[ζ* ] is coprime to 1, show that there is an integer c ∈ Z

(unique mod 1) such that ζ*c α is primary.

Solution. Let λ = 1 − ζ) . Since the prime ideal (λ) has degree 1, there
is an a ∈ Z such that α ≡ a (mod λ). Hence (α − a)/λ ≡ b (mod λ). So
we can write α ≡ a + bλ (mod λ2 ). Since ζ) = 1 − λ, we have ζ)c ≡ 1 − cλ
(mod λ2 ). Thus,

ζ)c α ≡ (1 − cλ)(a + bλ) ≡ a + (b − ac)λ (mod λ2 ).

Now, (a, 7) = 1 for otherwise λ | α contrary to assumption. Choose c

so that ac ≡ b (mod λ). Then ζ)c α ≡ a (mod λ2 ). Moreover c is clearly
unique mod 7.
306 CHAPTER 9. HIGHER RECIPROCITY LAWS

Exercise 9.2.8 With notation as above, show that (x + ζ i y) and (x + ζ j y) are

coprime in Z[ζ* ] whenever i #= j, 0 ≤ i, j < 1.

Solution. Suppose a is an ideal of Z[ζ) ] which contains both (x + ζ i y) and

(x + ζ j y). Then (ζ j − ζ i )y ∈ a and

(ζ j − ζ i )x = ζ j (x + ζ i y) − ζ i (x + ζ j y) ∈ a.

Since x and y are coprime, we deduce that (ζ j − ζ i ) ∈ a. By Exercise 4.3.7

or by 4.5.9, we deduce that a and (7) are not coprime. Since (7) is totally
ramified and (7) = (1 − ζ))−1 we see that (1 − ζ) = λ ∈ a. Thus (z) ∈ (λ)
which implies 7 | z contrary to assumption.

Exercise 9.2.9 Show that the ideals (x + ζ i y) are perfect 1th powers.

Solution. Since the ideals (x + ζ i y), 1 ≤ i ≤ 7 − 1, are mutually coprime,

and their product is an 7th power, each ideal must be an 7th power (see
Exercise 5.3.12).

Exercise 9.2.10 Consider the element

α = (x + y)*−2 (x + ζy).

Show that:
(a) the ideal (α) is a perfect 1th power.
(b) α ≡ 1 − uλ (mod λ2 ) where u = (x + y)*−2 y.

Solution. (a) is immediate from the previous exercise. To prove (b),

observe that x + ζy = x + y − λy. Thus

α = (x + y))−1 − λy(x + y))−2 = (x + y))−1 − λu.

Now x) + y ) + z ) ≡ x + y + z (mod 7), by Fermat’s little Theorem. If

7 | (x + y), then 7 | z, contrary to assumption. Therefore

(x + y))−1 ≡ 1 (mod 7)

since 7 ! (x + y). Since (7) = (λ))−1 we find α ≡ 1 − uλ (mod λ2 ) which

gives (b).

Exercise 9.2.11 Show that ζ −u α is primary.

Solution. We have

ζ −u α = (1 − λ)−u α ≡ (1 + uλ)(1 − uλ) (mod λ2 )

so that ζ −u α ≡ 1 (mod λ2 ). Hence, ζ −u α is primary.

9.2. EISENSTEIN RECIPROCITY 307

Exercise 9.2.12 Use Eisenstein reciprocity to show that if x* + y * + z * = 0 has

a solution in integers, 1 ! xyz, then for any p | y, (ζ/p)−u
* = 1. (Hint: Evaluate
(p/ζ −u α)* .)

Solution. By Exercise 9.2.11, ζ −u α is primary so by Eisenstein reciprocity,

! " ! " ! "−u ! "
p ζ −u α ζ α
= = .
ζ −u α ) p ) p ) p )

Now (ζ −u α) = (α) is an 7th power by Exercise 9.2.10 (a). So the left-hand

side of the above equation is 1. To evaluate (α/p)) , note that

α ≡ (x + y))−1 (mod p),

since p | y. Thus, since p is primary,

! " ! " ! "
α (x + y))−1 p
= = ,
p ) p ) (x + y))−1 )

again by Eisenstein reciprocity. By Exercise 9.2.9, (x + y) is an 7th power

of an ideal so that (α/p)) = 1. Therefore (ζ/p)−u
) = 1 for every p | y.

Exercise 9.2.13 Show that if

x* + y * + z * = 0

has a solution in integers, l ! xyz, then for any p | xyz, (ζ/p)*−u = 1.

Solution. We proved this for p | y in the previous exercise. Since the

equation is symmetric in x, y, z the same applies for p | x or p | z.

Exercise 9.2.14 Show that (ζ/p)−u

* = 1 implies that p*−1 ≡ 1 (mod 12 ).

Solution. Let us factor (p) = ℘1 · · · ℘g as a product of prime ideals. We

know that N (℘i ) = pf , and that gf = 7 − 1. Thus, by Exercises 9.2.4 and
9.2.5,
! " 6g ! " 6g
ζ ζ
ζ (N (℘i )−1)/) = ζ g(p −1)/) .
f
= =
p ) i=1 ℘i ) i=1

Since (ζ/p)u) = 1, we have

pf − 1
ug ≡0 (mod 7).
7

Moreover, (g, 7) = 1, (u, 7) = 1, so that pf ≡ 1 (mod 72 ). Since f | 7 − 1,

we deduce that p)−1 ≡ 1 (mod 72 ).
Exercise 9.2.14 is a famous result of Furtwangler proved in 1912.
308 CHAPTER 9. HIGHER RECIPROCITY LAWS

Exercise 9.2.15 If 1 is an odd prime and

x* + y * + z * = 0

for some integers x, y, z coprime to 1, then show that p*−1 ≡ 1 (mod 12 ) for every
p | xyz. Deduce that 2*−1 ≡ 1 (mod 12 ).

Solution. By Exercise 9.2.14, the first assertion is immediate. For the

second part, observe that at least one of x, y, z must be even.

9.3 Supplementary Problems

Exercise 9.3.1 Show that there are infinitely many primes p such that (2/p) =
−1.

Solution. We know that (2/p) = −1 if and only if p ≡ ±3 (mod 8).

Suppose there are only finitely many such primes q1 , . . . , qk (say) excluding
3. Consider the number b = 8q1 q2 · · · qk + 3. By construction b is not
divisible by any qi . Moreover, b ≡ 3 (mod 8) so that (2/b) = −1. Let b =
p1 · · · pm be the prime decomposition of b with pi not necessarily distinct.
Since ! " 6 m ! "
2 2
−1 = =
b i=1
pi

we must have (2/pi ) = −1 for some i. Since pi is distinct from q1 , . . . , qk ,

this is a contradiction.

Exercise 9.3.2 Let a be a nonsquare integer greater than 1. Show that there
are infinitely many primes p such that (a/p) = −1.

Solution. Without loss of generality, we may suppose a is squarefree and

greater than 2 by the previous exercise. Suppose there are only finitely
many primes q1 , . . . , qk , (say) such that (a/qi ) = −1. (This set could pos-
sibly be empty.) Write a = 2e r1 · · · rm for the prime factorization of a with
e = 0 or 1 and ri odd and distinct. By the Chinese Remainder Theorem,
we can find a solution to the simultaneous congruences

x ≡ 1 (mod qi ), 1 ≤ i ≤ k,
x ≡ 1 (mod 8),
x ≡ 1 (mod ri ), 1 ≤ i ≤ m − 1,
x ≡ c (mod rm ),

with c any nonresidue mod rm . (It is here that we are assuming a has at
least one odd prime divisor.) Let b be a solution greater than 1 and write
b = p1 · · · pt as its prime decomposition with pi not necessarily distinct.
9.3. SUPPLEMENTARY PROBLEMS 309

Since b ≡ 1 (mod 8), (2/b) = 1. Also, by the quadratic reciprocity law for
the Jacobi symbol, (ri /b) = (b/ri ). Thus
EaF ! "e 6
m E F ! "
2 ri c
= = = −1.
b b i=1 b rm

Also,
EaF t !
6 "
a
= .
b i=1
pi

Therefore, (a/pi ) = −1 for some pi . Moreover, pi | b and b is coprime to

q1 , . . . , qk by construction. This is a contradiction.

Exercise 9.3.3 Suppose that x2 ≡ a (mod p) has a solution for all but finitely
many primes. Show that a is a perfect square.

Solution. Write a = b2 c with c squarefree. Then (c/p) = 1 for all but

finitely many primes. By the previous exercise, this is not possible if c > 1.
Thus, c = 1.

Exercise 9.3.4 Let K be a quadratic extension of Q. Show that there are in-
finitely many primes which do not split completely in K.
√
Solution. Let K = Q( D), with D squarefree and greater than 1. By
the previous question, there are infinitely many primes p such that x2 ≡
D (mod p) has no solution. Hence (D/p) = −1. By the theory of the
Kronecker symbol, we deduce that p does not split in K.

Exercise 9.3.5 Suppose that a is an integer coprime to the odd prime q. If

xq ≡ a (mod p) has a solution for all but finitely many primes, show that a is a
perfect qth power. (This generalizes the penultimate exercise.)

Solution. We must show that if a is not a qth power, then there are
infinitely many primes p such that xq ≡ a (mod p) has no solution. We
will work in the field K = Q(ζq ). Write

(a) = ℘e11 · · · ℘enn ,

where the ℘i are distinct prime ideals of OK . We claim that q ! ei for some
i. To see this, let pi = ℘i ∩ Z. Since (q, a) = 1 we have q != pi for any i, so
each pi is unramified in OK . Thus

ordpi a = ord℘i (a) = ei .

If q | ei for all i, then a would be a qth power. So we may suppose q ! en .

Now let q1 , . . . , qk be a finite set of prime ideals different from ℘1 , . . . , ℘n
310 CHAPTER 9. HIGHER RECIPROCITY LAWS

and (1 − ζq ). By the Chinese Remainder Theorem (Exercise 5.3.13), we can

find an x ∈ OK satisfying the simultaneous congruences

x ≡ 1 (mod qi ), 1 ≤ i ≤ k,
x ≡ 1 (mod q),
x ≡ 1 (mod ℘j ), 1 ≤ j ≤ n − 1,
x ≡ c (mod ℘n ),

where c is chosen so that (c/℘n ) = ζq . Let b be a solution greater than 1.

Then b ≡ 1 (mod q) and hence is primary. Hence
EaF ! "
b
=
b q a q

by the Eisenstein reciprocity law. Now

! " n !
6 " ei
b b
= = ζqen != 1,
a q i=1 ℘i q

since q ! en . On the other hand, factoring b into a product of prime ideals

and using the multiplicativity of the qth power residue symbol, we deduce
that ! "
a
!= 1
℘ q
for some prime ideal ℘ dividing b. Hence xq ≡ a (mod ℘) has no solution.
Since b is coprime to the given set q1 , . . . , qk of prime ideals (possibly
empty), we can produce inductively infinitely many prime ideals ℘ such
that xq ≡ a (mod ℘) has no solution. A fortiori, there are infinitely many
primes p such that xq ≡ a (mod p) has no solution.

Exercise 9.3.6 Let p ≡ 1 (mod 3). Show that there are integers A and B such
that
4p = A2 + 27B 2 .
A and B are unique up to sign.

Solution. We work in the ring of Eisenstein integers Z[ρ]. Since p ≡ 1

(mod 3), p splits as ππ in Z[ρ]. Writing π = a + bρ, we see that

p = a2 − ab + b2 .

Thus,

4p = (2a − b)2 + 3b2 = (2b − a)2 + 3a2 = (a + b)2 + 3(a − b)2 .

A simple case by case examination (mod 3) shows that one of a, b or a − b

is divisible by 3 because p ≡ 1 (mod 3). The uniqueness is evident from
the uniqueness of p = a2 − ab + b2 .
9.3. SUPPLEMENTARY PROBLEMS 311

Exercise 9.3.7 Let p ≡ 1 (mod 3). Show that x3 ≡ 2 (mod p) has a solution if
and only if p = C 2 + 27D2 for some integers C, D.

Solution. If x3 ≡ 2 (mod p) has a solution, so does x3 ≡ 2 (mod π) where

p = ππ is the factorization of p in Z[ρ]. By Exercise 9.1.17, π ≡ 1 (mod 2),
and we write π = a + bρ. By the previous exercise, we can assume

p = a2 − ab + b2

with 3 | b (without loss of generality). Thus, π ≡ 1 (mod 2) implies that

a ≡ 1 (mod 2), b ≡ 0 (mod 2). Writing

4p = A2 + 27B 2 ,

we have A = 2a−b, B = b/3. Since B is even, so is A. Thus, p = C 2 +27D2 .

Conversely, if p = C 2 + 27D2 , then 4p = (2C)2 + 27(2D)2 . By uniqueness,
B = ±2D, by the previous exercise. Thus B is even and so is b. Therefore
π = a + bρ ≡ 1 (mod 2). By Exercise 9.1.7, x3 ≡ 2 (mod π) has a solution
in Z[ρ]. Since Z[ρ]/(π) has p elements, we can find an integer y ≡ x
(mod π). Thus y 3 ≡ 2 (mod π). But then y 3 ≡ 2 (mod π) so that y 3 ≡ 2
(mod p).

Exercise 9.3.8 Show that the equation

x3 − 2y 3 = 23z m

has no integer solutions with gcd(x, y, z) = 1.

Solution. Reduce the equation mod 23 to find x3 ≡ 2y 3 (mod 23). If

23 ! y, 2 is a cubic residue (mod 23). By the previous exercise, we can
write
23 = C 2 + 27D2
which is not possible. If 23 | y, then 23 | x and then 23 | gcd(x, y, z).
Chapter 10

Analytic Methods

10.1 The Riemann and Dedekind Zeta Func-

tions
Exercise 10.1.1 Show that for Re(s) > 1,

#$ 1
%−1
ζ(s) = 1− s ,
p
p

where the product is over prime numbers p.

Solution. Since every natural number can be factored uniquely as a prod-

uct of prime powers, it is clear that when we expand the product
6! 1 1 1
"
1 + s + 2s + 3s + · · ·
p
p p p

the term 1/ns occurs exactly once. The assertion is now evident.

Exercise 10.1.2 Let K be an algebraic number field and OK its ring of integers.
The Dedekind zeta function ζK (s) is defined for Re(s) > 1 as the infinite series
, 1
ζK (s) = ,
a
(N a)s

where the sum is over all ideals of OK . Show that the infinite series is absolutely
convergent for Re(s) > 1.

Solution. For any s with Re(s) > 1, it suffices to show that the partial
sums
* 1
(N a)s
N a≤x

313
314 CHAPTER 10. ANALYTIC METHODS

are bounded. Indeed, since any ideal can be expressed as a product of

powers of prime ideals, it is evident that
* 1 6 ! 1 1
"
≤ 1 + + + · · · ,
(N a)σ (N ℘)σ (N ℘)2σ
N a≤x N ℘≤x

where σ = Re(s) > 1. Hence

* 1 6 ! 1
"−1
≤ 1 − .
(N a)σ (N ℘)σ
N a≤x N ℘≤x

For each prime ideal ℘, we have a unique prime number p such that N (℘) =
pf for some integer f . Moreover, there are at most [K : Q] prime ideals
corresponding to the same prime p. In fact, they are determined
5g from the
e
factorization pOK = ℘e11 · · · ℘gg and by Exercise 5.3.17, i=1 ei fi = [K : Q]
where N ℘i = pfi . Since ei ≥ 1 and fi ≥ 1, we find g ≤ [K : Q]. Hence

* 1 6! 1
"−[K:Q]
≤ 1− σ .
(N a)σ p
N a≤x p≤x

2
Since the product p (1 − p )
−σ −1
converges absolutely for σ > 1, the result
follows.

Exercise 10.1.3 Prove that for Re(s) > 1,

#$ 1
%−1
ζK (s) = 1− .
℘
(N ℘)s

Solution. Since every ideal a can be written uniquely as a product of

prime ideals (see Theorem 5.3.6), we find that when the product
6! 1 1
"
1+ + + · · ·
℘
(N ℘)s (N ℘)2s

is expanded, 1/(N a)s occurs exactly once for each ideal a of OK .

Exercise 10.1.5 Show that (s − 1)ζ(s) can be extended analytically for Re(s) >
0.

Solution. We apply Theorem 10.1.4 with am = 1. Then A(x) = [x], the

greatest integer less than or equal to x. Thus,

*∞ @ ∞
1 [x] dx
s
=s .
n=1
n 1 xs+1
10.1. THE RIEMANN AND DEDEKIND ZETA FUNCTIONS 315

Writing [x] = x − {x} we obtain for Re(s) > 1,

@ ∞
s {x} dx
ζ(s) = −s .
s−1 1 xs+1
Since {x} is bounded by 1, the latter integral converges for Re(s) > 0.
Thus, (s − 1)ζ(s) is analytic for Re(s) > 0.

Exercise 10.1.6 Evaluate

lim (s − 1)ζ(s).
s→1

Solution. From the previous exercise we have

@ ∞
{x} dx
lim (s − 1)ζ(s) = lim s − lim s(s − 1)
s→1 s→1 s→1 1 xs+1
and the latter limit is zero since the integral is bounded. Hence, the desired
limit is 1.

Exercise 10.1.8 For K = Q(i), evaluate

lim (s − 1)ζK (s).
s→1+

Solution. Clearly, this limit is π/4.

Exercise 10.1.9 Show that the number of integers (a, b) with a > 0 satisfying
a2 + Db2 ≤ x is
πx √
√ + O( x).
2 D
√
Solution. Corresponding to each such (a, b) we associate (a, Db) which
lies inside the circle√u2 + v 2 ≤ x. We now count these “lattice” points.
We will call (a, Db) internal if (a + 1)2 + D(b + 1)2 ≤ x. Otherwise,
call it a boundary lattice point. Let I be the number of internal lattice
points, and
√ B the number of boundary lattice points. Each lattice point
has area D. Thus √ √
π
DI ≤ x ≤ D(I + B)
2
since in our count a > 0 and b ∈ Z. A little reflection shows that any
boundary point is contained in the annulus
#√ √ $2 #√ √ $2
x − D + 1 ≤ u2 + v 2 ≤ x+ D+1
√
which has area O( xD). Thus
√ √
DB = O( xD)
√
and we get B = O( x). Thus,
πx √
I = √ + O( x).
2 D
316 CHAPTER 10. ANALYTIC METHODS

√
Exercise 10.1.10 Suppose K = Q( −D) where D > 0 and −D #≡ 1 (mod 4)
and OK has class number 1. Show that (s − 1)ζK (s) extends analytically to
Re(s) > 12 and find
lim (s − 1)ζK (s).
s→1

(Note that there are only finitely many such fields.)

√
Solution. Each ideal of OK is principal, of the form (a + b D). We may
choose a > 0. Thus
* *∞
1 an
ζK (s) = = ,
a>0
(a2 + Db2 )s
n=1
ns
b∈Z

where an is the number of solutions of a2 + Db2 = n with a > 0, b ∈ Z. By

Theorem 10.1.4, we have
@ ∞
A(x)
ζK (s) = s dx,
1 xs+1
5
where A(x) = n≤x an . By the previous exercise,
πx √
A(x) = √ + O( x)
2 D
so that @ ∞
πs E(x) dx
ζK (s) = √ +s ,
2 D(s − 1) 1 xs+1
√ 1
where E(x) = O( x). The latter integral converges for Re(s) > 2. This
gives the desired analytic continuation. Moreover,
π π
lim (s − 1)ζK (s) = √ = L .
s→1 2 D |dK |

(In the next section, we will establish a similar result for any quadratic field
K.)

10.2 Zeta Functions of Quadratic Fields

√
Exercise 10.2.1 Let K = Q( d) with d squarefree, and denote by an the num-
ber of ideals in OK of norm n. Show that an is multiplicative. (That is, prove
that if (n, m) = 1, then anm = an am .)

Solution. Let a be an ideal of norm n and let

k
6
n= pα
i
i

i=1
10.2. ZETA FUNCTIONS OF QUADRATIC FIELDS 317

be the unique factorization of n into distinct prime powers. Then by unique

factorization of ideals,
6 s
e
a= ℘j j .
j=1

We see immediately that

k
6
an = apαi .
i
i=1
This implies that an is multiplicative.

Exercise 10.2.2 Show that for an odd prime p, ap = 1 + (d/p).

√
Solution. By Theorem 7.4.2, we see that an odd prime p splits in Q( d)
if and only if (d/p) = 1, in which case there are two ideals of norm p. If
p does not split, there are no ideals of norm p. Finally, if p ramifies, then
(d/p) = 0 and there is only one ideal of norm p, by Exercise 7.4.3.
√
Exercise 10.2.3 Let dK be the discriminant of K = Q( d). Show that for all
primes p, ap = 1 + (dK /p).

Solution. Since dK = d or 4d, the result is clear for odd primes p from
the previous exercise. We therefore need only consider p = 2. If 2 | dK ,
then by Theorem 7.4.5, 2 ramifies and there is only one ideal of norm 2. If
2 ! dK , then
! " ;
dK 1 if dK ≡ 1 (mod 8),
=
2 −1 if dK ≡ 5 (mod 8),
by the definition of the Kronecker symbol. The result is now immediate
from Theorem 7.4.5.

Exercise 10.2.4 Show that for all primes p,

,α $ % ,$ %
dK dK
a pα = = .
j=1
pj α
δ
δ|p

Solution. The norm of any prime ideal is either p or p2 , the latter occurring
if and only if (d/p) = −1. Thus for α = 2, the formula is established and
for α = 1, the previous exercise applies. If (d/p) = −1, then clearly apα = 0
if α is odd and if α is even, then there is only one ideal of norm pα . If p
splits, then any ideal of norm pα must be of the form

℘j (℘$ )α−j

for some j, where pOK = ℘℘$ . It is now clear that apα = α + 1 which is
the sum
*α ! "
dK
.
j=0
pj
318 CHAPTER 10. ANALYTIC METHODS

Finally, if p ramifies, there is only one ideal of norm pα , namely ℘α where

pOK = ℘2 . This completes the proof.

Exercise 10.2.5 Prove that

, $ dK %
an = .
δ
δ|n

Solution. Since an is multiplicative,

 
6 * α ! "
 d K 
an =
α j=0
p
p -n

by the previous exercise. The result is now immediate upon expanding the
product.

Exercise 10.2.6 Let dK be the discriminant of the quadratic field K. Show that
there is an n > 0 such that (dK /n) = −1.

Solution. We know that dK ≡ 0 or 1 (mod 4). If dK ≡ 1 (mod 4), then

for any odd n we have
! " ! "
dK n
= .
n |dK |
Let |dK | = pa where p is an odd prime. Since dK is squarefree, p ! a. Let
u be a quadratic nonresidue modulo p. We can find an odd n ≡ u (mod p)
and n ≡ 1 (mod 2a) by the Chinese Remainder Theorem. Then
! " ! "E F ! "! "
dK n n u 1
= = = −1,
n p a p a

as desired. If dK is even, let dK = d1 d2 where d1 is 4 or 8 and d2 is an odd

discriminant. Then ! " ! "! "
dK d1 d2
=
n n n
by definition of the Kronecker symbol. Since d1 = 4 or 8, it is easy to find
an a such that (d1 /a) = −1. Choose n ≡ a (mod d1 ), n ≡ 1 (mod d2 ).
Then (dK /n) = −1, as desired.

Exercise 10.2.7 Show that

) )
) $ %)
) , dK )
) ) ≤ |dK |.
) )
)n≤x n )
10.2. ZETA FUNCTIONS OF QUADRATIC FIELDS 319

Solution. Let ! "

* dK
S= .
n
n mod |dK |
(n,dK )=1

Choose n0 such that (n0 , |dK |) = 1 and (dK /n0 ) = −1. (This is possible
by the previous exercise.) Then
! " * ! "
dK dK
S= .
n0 nn0
n mod |dK |
(n,|dK |)=1

As n ranges over residue classes mod |dK | so does nn0 . Hence

! "
dK
−S = S=S
n0

so that S = 0.
Now define v by
v|dK | ≤ x < (v + 1)|dK |.

Then,
* ! dK " * !
dK
"
=
n n
n≤x |dK |v≤n<x

since  
* ! dK " * v * ! "
dK 
= 
n n
n<|dK |v j=1 (j−1)|dK |<n<j|dK |

and the inner sum is zero because it is equal to S. Thus

1 1
1 ! "1
1 * dK 1
1 1 ≤ |dK |.
1 n 11
1 n≤x

Exercise 10.2.10 If K is a quadratic field, show that (s − 1)ζK (s) extends to

an analytic function for Re(s) > 12 .

Solution. By Theorem 10.1.4,

@ ∞
cs E(x) dx
ζK (s) = +s ,
s−1 1 xs+1
√
where E(x) = O( x). The integral therefore converges for Re(s) > 12 .
320 CHAPTER 10. ANALYTIC METHODS

10.3 Dirichlet’s L-Functions

Exercise 10.3.1 Show that L(s, χ) converges absolutely for Re(s) > 1.

Solution. Since |χ(n)| ≤ 1, we have

1 1
1* ∞
χ(n) 11 * 1
∞
1
1 1 ≤ ,
1
n=1
ns 1 n=1 nσ

where σ = Re(s). The latter series converges absolutely for Re(s) > 1.

Exercise 10.3.2 Prove that

) )
) )
), )
) χ(n)) ≤ m.
) )
)n≤x )

Solution. If χ is nontrivial, there is an a (mod m) coprime to m such that

χ(a) != 1. Then
* * *
χ(b) = χ(ab) = χ(a) χ(b).
b mod m b mod m b mod m
(b,m)=1 (b,m)=1 (b,m)=1

Hence *
χ(b) = 0.
b mod m
(b,m)=1

Now, partition the interval [1, x] into subintervals of length m and suppose
that km < x ≤ (k + 1)m. Then
* * *
χ(n) = χ(n) + χ(n).
n≤x n≤km km<n≤x

The first sum on the right-hand side is zero and the second sum is bounded
by m .

Exercise 10.3.3 If χ is nontrivial, show that L(s, χ) extends to an analytic

function for Re(s) > 0.

Solution. By Theorem 10.1.4, this is now immediate.

Exercise 10.3.4 For Re(s) > 1, show that

#$ χ(p)
%−1
L(s, χ) = 1− s .
p
p
10.3. DIRICHLET’S L-FUNCTIONS 321

Solution. Since χ is completely multiplicative,

6! χ(p) χ(p2 )
"
L(s, χ) = 1 + s + 2s + · · ·
p
p p
6 ! "
χ(p) χ(p)2
= 1+ s + + · · ·
p
p ps
6! χ(p)
"−1
= 1− s .
p
p

Exercise 10.3.5 Show that

=
, ϕ(m) if a ≡ b (mod m),
χ(a)χ(b) =
χ mod m
0 otherwise.

Solution. If a ≡ b (mod m), the result is clear. If a !≡ b (mod m), let ψ

be a character such that ψ(a) != ψ(b). Then
* *
ψ(ba−1 )χ(ba−1 ) = (ψχ)(ba−1 )
χ mod m χ mod m
*
= χ(ba−1 ),
χ mod m

because as χ ranges over characters mod m, so does ψχ. But

# $ *
1 − ψ(ba−1 ) χ(ba−1 ) = 0
χ mod m

so the result follows.

Exercise 10.3.6 For Re(s) > 1, show that

, , 1
log L(s, χ) = ϕ(m) .
npns
χ mod m pn ≡1 mod m

Solution. By Exercises 10.3.3 and 10.3.5, we find

* * 1 *
log L(s, χ) = χ(pn )
npns
χ mod m n,p χ mod m
* 1
= ϕ(m) .
npns
pn ≡1( mod m)

Exercise 10.3.7 For Re(s) > 1, show that

, , 1
χ(a) log L(s, χ) = ϕ(m) .
npns
χ mod m pn ≡a mod m
322 CHAPTER 10. ANALYTIC METHODS

Solution. By Exercise 10.3.3,

* * 1 *
χ(a) log L(s, χ) = χ(a)χ(pn )
npns
χ mod m n,p χ mod m

since the series converges absolutely in Re(s) > 1. By Exercise 10.3.5, the
inner sum on the right-hand side is ϕ(m) when pn ≡ a (mod m) and zero
otherwise. The result is now immediate.

Exercise 10.3.8 Let K = Q(ζm ). Set

#
f (s) = L(s, χ).
χ

Show that ζK (s)/f (s) is analytic for Re(s) > 12 .

Solution. The primes that split completely in Q(ζm ) are those primes
p ≡ 1 (mod m). Thus, because there are φ(m) ideals of norm p for p ≡ 1
(mod m),

6! 1
"−1
ζK (s) = 1−
℘
(N ℘)s
6 ! "−φ(m)
1
= 1− s g(s),
p
p≡1 mod m

where g(s) is analytic for Re(s) > 12 . By Exercise 10.3.6

6 6 ! "−φ(m)
1
L(s, χ) = 1− h(s),
ps
χ p≡1 mod m

where h(s) is analytic and nonzero for Re(s) > 12 (Why?).

Thus, ζK (s)/f (s) is analytic for Re(s) > 12 . This gives the analytic
continuation of ζK (s) for Re(s) > 12 . We can in fact show that
6
ζK (s) = L(s, χ).
χ

10.4 Primes in Arithmetic Progressions

Exercise 10.4.3 With the notation as in Section 10.3, write
# ,∞
cn
f (s) = L(s, χ) = s
.
χ n=1
n

Show that cn ≥ 0.
10.4. PRIMES IN ARITHMETIC PROGRESSIONS 323

Solution. This is immediate from Exercise 10.3.6 because

 
* 1
f (s) = expφ(m) .
npns
n p ≡1 mod m

Exercise 10.4.4 With notation as in the previous exercise, show that

,∞
cn
ns
n=1

diverges for s = 1/φ(m).

Solution. By Euler’s theorem (Theorem 1.1.14), pφ(m) ≡ 1 (mod m) for

prime p ! m. Thus,
 
# $ * 1
f 1/φ(m) = expφ(m) n/φ(m)

pn ≡1mod m
np
 
*1
≥ exp .
p
p!m

Since *1
= +∞,
p
p

we are done (by Exercise 1.4.18).

Exercise 10.4.6 Show that

, 1
= +∞.
p
p≡1 mod m

Solution. By Exercise 10.3.6,

* 1 1 *
ns
= log L(s, χ).
np φ(m)
pn ≡1 mod m χ mod m

As s → 1+ , we see that
* 1
= +∞
p
p≡1 mod m

because L(1, χ) != 0 for χ != χ0 and

 
6! "
1 
lim log L(s, χ0 ) = lim+ logζ(s) 1− s = +∞.
s→1+ s→1 p
p|m
324 CHAPTER 10. ANALYTIC METHODS

Note that * 1 * 1
≤ < ∞.
npn p
p(p − 1)
n≥2
p

Exercise 10.4.7 Show that if gcd(a, m) = 1, then

, 1
= +∞.
p
p≡a mod m

Solution. By Exercise 10.3.7,

* 1 1 *
= χ(a) log L(s, χ).
npns φ(m)
pn ≡a mod m χ mod m

As s → 1+ , we see that
* 1
= +∞
p
p≡a mod m

again because L(1, χ) != 0 for χ != χ0 and

lim log L(s, χ0 ) = +∞.

s→1+

Hence there are infinitely many primes in any given coprime residue class
mod m.

10.5 Supplementary Problems

Exercise 10.5.1 Define for each character χ (mod m) the Gauss sum
,
g(χ) = χ(a)e2πia/m .
a mod m

If (n, m) = 1, show that

,
χ(n)g(χ) = χ(b)e2πibn/m .
b mod m

Solution.
*
χ(n)g(χ) = χ(n)χ(a)e2πia/m
a mod m
*
= χ(b)e2πibn/m
b mod m

upon setting a = bn in the first sum. Observe that as a ranges over coprime
residue classes (mod m), so does bn since (n, m) = 1.
10.5. SUPPLEMENTARY PROBLEMS 325

√
Exercise 10.5.2 Show that |g(χ)| = m.

Solution.
1 12
m−1
* *1
m−1
1 *
1
2 2 2πibn/m 1
|χ(n)| |g(χ)| = 1 χ(b)e 1
1 1
n=1 n=0 b mod m
* m−1
*
= χ(b1 )χ(b2 ) e2πi(b1 −b2 )n/m .
b1 ,b2 n=0

The last sum is the sum of a geometric progression and is 0 unless b1 ≡ b2

(mod m) in which case it is m. Thus,
φ(m)|g(χ)|2 = mφ(m)
from which the result follows.

Exercise 10.5.3 Establish the Pólya–Vinogradov inequality:

) )
) )
), )
)
) χ(n))) ≤ 12 m1/2 (1 + log m)
)n≤x )

for any nontrivial character χ (mod m).

Solution. By the two previous exercises, we get

/ 0
* * *
2πibn/m
χ(n) = g(χ)−1
χ(b)e .
n≤x n≤x b mod m

Interchanging summation gives

* e2πib([x]+1)/m − 1
e2πibn/m = ,
n≤x
e2πib/m − 1

provided b !≡ 0 (mod m). Observe that the numerator is bounded by 2 and

the denominator can be written as
πb
eπib/m (eπib/m − e−πib/m ) = 2ieπib/m sin
m
so that for b !≡ 0 (mod m)
1 1
1 1
1 * 2πibn/m 1 1
1 e 1≤
1 1 |sin πb | .
1n≤x 1 m

π(m−b)
m | = |sin
Since |sin πb m |, we may suppose b ≤ m/2. In that case,
1 1 ! "
1sin πb 1 ≥ 2 πb
1 1
1 m1 π m
326 CHAPTER 10. ANALYTIC METHODS

because sin x ≥ 2x/π for 0 ≤ x ≤ π/2, as is seen by looking at the graph

of sin x. Therefore
1 1
1 1
1* 1 * m
1 χ(n)1 ≤ |g(χ|−1
1 1 2b
1n≤x 1 b mod m
√
m
≤ log m.
2
The last inequality follows by noting that
* 1 @ m
dt
≤1+ = 1 + log m.
b 1 t
b≤m

Exercise 10.5.4 Let p be prime. Let χ be a character mod p. Show that there
is an a ≤ p1/2 (1 + log p) such that χ(a) #= 1.

Solution. If each a ≤ p1/2 (1 + log p) = u (say) satisfies χ(a) = 1, then

*
χ(a) = u.
a≤u

By the Pólya–Vinogradov inequality, the left-hand side is ≤ 12 p1/2 (1+log p),

which is a contradiction.

Exercise 10.5.5 Show that if χ is a nontrivial character mod m, then

, χ(n) $√ %
m log m
L(1, χ) = +O .
n u
n≤u

Solution. By Theorem 10.1.4,

* χ(n) @ ∞ A(x) dx
= ,
n>u
n 1 x2

where ;
0 if x < u,
A(x) = 5
u<n≤x χ(n) if x ≥ u.
√
By Pólya–Vinogradov, A(x) = O( m log m) and so
* χ(n) @ ∞ A(x) dx !√ "
m log m
= = O .
n>u
n u x2 u

Therefore
*∞ * χ(n) !√ "
χ(n) m log m
L(1, χ) = = +O .
n=1
n n u
n≤u
10.5. SUPPLEMENTARY PROBLEMS 327

Exercise 10.5.6 Let D be a bounded open set in R2 and let N (x) denote the
number of lattice points in xD. Show that
N (x)
lim = vol(D).
x→∞ x2
Solution. Without loss of generality, we may translate our region by a
lattice point to the first quadrant. A lattice point (u, v) ∈ xD if and only if
(u/x, v/x) ∈ D. This suggests we partition the plane into squares of length
1/x with sides parallel to the coordinate axes. This then partitions D into
small squares each of area 1/x2 . The number of lattice points of xD is then
clearly the number of “interior squares.” For this partition of the region, we
write down the lower and upper Riemann sums. Let Ix denote the number
of “interior” squares, and Bx the number of “boundary” squares. Then, by
the definition of the Riemann integral
Ix Ix + Bx
≤ vol(D) ≤
x2 x2
so that
Ix Ix + Bx
lim 2
= lim = vol(D).
x→∞ x x→∞ x2
On the other hand,
Ix ≤ N (x) ≤ Ix + Bx .
Thus,
N (x)
lim = vol(D).
x→∞ x2
Exercise 10.5.7 Let K be an algebraic number field, and C an ideal class of K.
Let N (x, C) be the number of nonzero ideals of OK belonging to C with norm
≤ x. Fix an integral ideal b in C −1 . Show that N (x, C) is the number of nonzero
principal ideals (α) with α ∈ b with |NK (α)| ≤ xN (b).

Solution. For any a ∈ C, ab = (α) so that (α) ⊆ b. Moreover NK (α) =

N (a)N (b). Conversely, if α ∈ b, let a = b−1 (α) which is an integral ideal
of norm ≤ x.

Exercise 10.5.8 Let K be an imaginary quadratic field, C an ideal class of OK ,

and dK the discriminant of K. Prove that
N (x, C) 2π
lim = 8 ,
x→∞ x w |dK |
where w is the number of roots of unity in K.

Solution. By the previous exercise, wN (x, C) is the number of integers

α ∈ b with 0 < |NK (α)| < x|N (b)| where b ∈ C −1 is fixed. Let β1 , β2 be an
integral basis of b, and let β1$ , β2$ be the conjugates of β1 , β2 , respectively.
Define
D = {(u, v) ∈ R2 : 0 < |uβ1 + vβ2 |2 < 1}.
328 CHAPTER 10. ANALYTIC METHODS

Clearly D is bounded√ (as is easily verified). Then wN (x, C) is the number

of lattice points in xD so that (by the penultimate exercise)

wN (x, C)
lim = vol(D).
x→∞ x
Set u1 = Re(uβ1 + vβ2 ), u2 = Im(uβ1 + vβ2 ) so that
@@
vol(D) = du dv
|uβ1 +vβ2 |2 <1
@@
2
= L · du1 du2
N b |dK |
u21 +u22 <1
2π
= L .
N b |dK |

Exercise 10.5.9 Let K be a real quadratic field with discriminant dK , and fun-
damental unit ε. Let C be an ideal class of OK . Show that

N (x, C) 2 log ε
lim = √ ,
x→∞ x dK

where N (x, C) denotes the number of integral ideals of norm ≤ x lying in the
class C.

Solution. As in the previous two exercises, we fix an ideal b ∈ C −1 and we

are reduced to counting the number (α), with (NK (α)) ≤ xN b. Therefore,
fix an integral basis β1 , β2 of b, β1$ , β2$ denote the conjugates, respectively.
Notice that we have infinitely many choices of α since (α) = (εm α) for any
integer m. Our first step is to isolate only one generator. Since α/α$ is a
unit of norm ±1, there is an integer m so that
1α1
1 1
−2m log ε ≤ log 1 $ 1 < (−2m + 2) log ε.
α
Thus, setting ω = εm α, we find
1 1
1 ω 1
0 ≤ log 11 1 < log ε.
|NK (ω) | 1
1/2

If ω1 , ω2 are associated elements of b satisfying the same inequality, then

the fact that ω1 = ηω2 for some unit η gives 1 ≤ |η| < ε. Thus, η = ±1
because ε is a fundamental unit. Therefore 2N (x, C) is the number of ω ∈ b
such that 0 < |NK (ω)| < (N b)x,
1 1
1 ω 1
1
0 ≤ log 1 1 < log ε.
|NK (ω)|1/2 1
10.5. SUPPLEMENTARY PROBLEMS 329

Now consider
; G
2
0 < |uβ11 + vβ2 ||uβ1$ + vβ2$ | < 1,1
D= (u, v) ∈ R : 1 1 +vβ2 1
0 < log 1 |uβ1 +vβ2uβ
|1/2 |uβ " +vβ " |1/2 1
< log ε.
1 2

Proceeding as in the previous question gives

2N (x, C) 4 log ε
lim =L .
x→∞ x |dK |

Exercise 10.5.10 Let K be an imaginary quadratic field. Let N (x; K) denote

the number of integral ideals of norm ≤ x. Show that
N (x; K) 2πh
lim = 8 ,
x→∞ x w |dK |
where h denotes the class number of K.

Solution. Clearly, *
N (x; K) = N (x, C),
C

where the (finite) sum is over the ideal classes of K. Thus,

N (x; K) * N (x, C)
lim = lim
x→∞ x x→∞ x
C

and by the penultimate question,

N (x, C) 2π
lim = L .
x→∞ x w |dK |

Exercise 10.5.11 Let K be a real quadratic field. Let N (x, K) denote the
number of integral ideals of norm ≤ x. Show that
N (x; K) 2h log ε
lim = 8 ,
x→∞ x |dK |
where h is the class number of K.

Solution. This follows exactly as in the previous question except that we

invoke the corresponding limit of N (x, C) for the real quadratic case.

Exercise 10.5.12 (Dirichlet’s Class Number Formula) Suppose that K is

a quadratic field with discriminant dK . Show that

,∞ $ %  √ 2πh
if dK < 0,
dK 1 w |d |
= 2h log K
n=1
n n √ ε
if dK > 0,
|dK |

where h denotes the class number of K.

330 CHAPTER 10. ANALYTIC METHODS

Solution.√By Example 10.2.9, the number of integral ideals of norm ≤ x

is cx + O( x) where
∞ !
* "
dK 1
c= .
n=1
n n

Thus,
∞ ! "
N (x, K) * dK 1
lim = .
x→∞ x n=1
n n

Comparing this limit with the previous two questions gives the desired
result.

Exercise 10.5.13 Let d be squarefree and positive.

√ Using√
Dirichlet’s class num-
ber formula, prove that the class number of Q( −d) is O( d log d).
√
Solution. Let D be the discriminant of Q( −d). By Dirichlet’s class
number formula,
*∞ ! "
2πh D 1
L = .
w |D| n=1 n n
Since |D| = |d| or 4|d|, it suffices to prove that

*∞ ! "
D 1
= O(log d).
n=1
n n

By a previous exercise (Exercise 10.5.5), we have for any u ≥ 1,

∞ ! " /√ 0
* D 1 * !D" 1 d log d
= +O ,
n=1
n n n n u
n≤u

which
√ was derived using the Pólya-Vinogradov inequality. Choosing u =
d, and noting that
* !D" 1
= O(log d),
√ n n
n≤ d

we obtain the result.

Exercise 10.5.14 Let d be squarefree and positive.

√ Using √Dirichlet’s class num-
ber formula, prove that the class number h of Q( d) is O( d).
√
Solution. Let D be the discriminant of Q( d). By Dirichlet’s class number
formula
∞ ! "
2h log ε * D 1
L = ,
|D| n=1
n n
10.5. SUPPLEMENTARY PROBLEMS 331
√
where ε is the fundamental unit of Q( d). Since
√
√ a+b d
ε=a+b d or
2
for integers a, b and a2 − db2 = ±1 or ±4 we deduce that

log ε > log d.

Estimating the infinite series as in the previous question, the result is now
immediate.

Exercise 10.5.15 With ψ(x) defined (as in Chapter 1) by

,
ψ(x) = log p,
pα ≤x

prove that for Re(s) > 1,

' ∞
ζ( ψ(x)
− (s) = s dx.
ζ 1 xs+1
Solution. Taking logs of the identity
6! 1
"−1
ζ(s) = 1− s
p
p

we differentiate to obtain
*∞
ζ$ Λ(n)
− (s) = ,
ζ n=1
ns

where Λ(n) = log p if n = pα and 0 otherwise. It is clear that

*
ψ(x) = Λ(n)
n≤x

and so the result now follows by Theorem 10.1.4.

Exercise 10.5.16 If for any ε > 0,

ψ(x) = x + O(x1/2+ε ),

show that ζ(s) #= 0 for Re(s) > 12 .

Solution. If the given estimate holds, we obtain an analytic continuation

of
ζ$
− (s)
ζ
for Re(s) > 12 , apart from a simple pole at s = 1. Thus ζ(s) has no zeros
in Re(s) > 12 .
Chapter 11

Density Theorems

11.1 Counting Ideals in a Fixed Ideal Class

Exercise 11.1.1 Show that Bx is a bounded region in Rn .

Solution. Since the integral basis β1 , ..., βn is linearly independent over Q,

(j)
dK/Q (β1 , ..., βn ) = [det(βi )]2 != 0.

Thus the linear map

φ(x1 , ..., xn ) = (α(1) , ..., α(n) )

(i)
is invertible. Let M be the largest of the values | log |6j || for 1 ≤ i, j ≤ r.
Then for each element of Bx , we have from the second relation defining Bx ,

|α(i) | ≤ erM (N (b)x)1/n ,

holds for 1 ≤ i ≤ n. Therefore the image of φ is a bounded set and thus

the inverse image is bounded.

Exercise 11.1.2 Show that tB1 = Btn for any t > 0.

Solution. There are two conditions defining Bx . The second one involving
units is invariant under the homogenous change of variables. The first
inequality gets multiplied by tn .

Exercise 11.1.3 Show that N (x, C) = O(x). Deduce that N (x; K) = O(x).

Solution. We may write the xi ’s in terms of the α(i) ’s by inverting

the transformation matrix, as noted in Exercise 11.1.1. As the α(i) ’s are
O(x1/n ), we deduce that the xi ’s are O(x1/n ). The result is now immediate.

333
334 CHAPTER 11. DENSITY THEOREMS

1
Exercise 11.1.6 Prove that ζ(s, C) extends to the region 5(s) > 1 − n
except
for a simple pole at s = 1 with residue

2r1 (2π)r2 RK
8 .
w |dK |

Deduce that ζK (s) extends to 5(s) > 1 − n1 except for a simple pole at s = 1
with residue
2r1 (2π)r2 hK RK
ρK := 8 ,
w |dK |
where hK denotes the class number of K.

Solution. If we let αK = ρK /hK , and consider the Dirichlet series

*∞
am
f (s) = s
:= ζ(s, C) − αK ζ(s),
m=1
m

then by Theorem 11.1.5, we have

* 1
am = O(x1− n ).
m≤x

By Theorem 10.1.4, f (s) converges for <(s) > 1 − n1 . As ζ(s) has a simple
pole at s = 1 with residue 1, the latter assertions are immediate.

Exercise 11.1.7 Prove that there are infinitely many prime ideals in OK which
are of degree 1.

Solution. We have
* 1 * 1
log ζK (s) = + .
℘
N (℘) s nN (℘)ns
n≥2,℘

The second sum is easily seen to converge for <(s) > 1/2. The first sum
can be separated into two parts, one over primes of first degree and the
other over primes of degree ≥ 2. Again, the second sum converges for
<(s) > 1/2. If the first sum consisted of only finitely many terms, the right
hand side would tend to a finite limit as s → 1+ , which is not the case as
the Dedekind zeta function has a simple pole at s = 1.

Exercise 11.1.8 Prove that the number of prime ideals ℘ of degree ≥ 2 and
with norm ≤ x is O(x1/2 log x).

Solution. If ℘ is a prime ideal of degree r ≥ 2, then N (℘) = pr ≤ x

implies r ≤ (log x)/ log 2 and p ≤ x1/2 . For each prime, there are a bounded
number of prime ideals in K above p. Thus, the final estimate is obtained
by counting the number of possible pairs (p, r) and this is O(x1/2 log x).
11.1. COUNTING IDEALS IN A FIXED IDEAL CLASS 335

Exercise 11.1.9 Let µ be defined on integral ideals a of OK as follows. µ(OK ) =

1, and if a is divisible by the square of a prime ideal, we set µ(a) = 0. Otherwise,
we let µ(a) = (−1)k when a is the product of k distinct prime ideals. Show that
,
µ(b) = 0
b|a

unless a = OK .

Solution. Clearly, the function

*
f (a) = µ(b)
b|a

is multiplicative and so, it suffices to evaluate it on prime ideals. But this

is clearly zero.

Exercise 11.1.10 Prove that the number of ideals of OK of odd norm ≤ x is

#$ 1
%
1
ρK x 1− + O(x1− n ),
N (℘)
℘|2

where the product is over prime ideals ℘ of OK dividing 2OK .

Solution. By the previous exercise, the number of such ideals is

* *
µ(b)
N (a)≤x b|(a,2)

since an ideal has odd norm if and only if it has no prime ideal divisor
above 2. Interchanging summation and using Theorem 11.1.5, we obtain
the result.

Exercise 11.1.11 Let A(x) be the number of ideals of OK of even norm ≤ x

and B(x) of odd norm ≤ x. Show that
A(x)
lim =1
x→∞ B(x)
if and only if K = Q or K is a quadratic field in which 2 ramifies.

Solution. From the previous exercise, we see that

6! 1
"
1
B(x) = 1− ρK x + O(x1− n ).
N (℘)
℘|2

We see that A(x) ∼ B(x) if and only if A(x) + B(x) ∼ 2B(x). That is, if
and only if ! "
1 6 1
= 1− .
2 N (℘)
℘|2
336 CHAPTER 11. DENSITY THEOREMS

Let ℘1 , ..., ℘t be the prime ideals above 2 with norms 2m1 , ..., 2mt respec-
tively. The above equation implies
t ! "
1 6 2mi − 1
= .
2 i=1 2mi

In other words, we have

t
6
2m1 +···+mt −1 = (2mi − 1).
i=1

The right hand side is a product of odd numbers and so the only way the
left hand side can be odd is if m1 + · · · + mt = 1 which means that there is
only one prime ideal above 2 and it has norm 2. This can only happen if
K = Q or if K is a quadratic field in which 2 ramifies.

Exercise 11.1.12 With notation as in the discussion preceding Theorem 11.1.4,

let Vx denote the set of n-tuples (x1 , ..., xn ) satisfying

|α(1) · · · α(n) | ≤ xN (b).

Let t = x1/n . Show that there is a δ > 0 (independent of x) such that for each
lattice point P contained in V(t−δ)n , all the points contained in the translate of
the standard unit cube by P belong to Vx .

Solution. We fix δ > 0 and choose it appropriately later. We may write

the norm form α(1) · · · α(n) as
*
ai1 ,...,in xi11 · · · xinn
i1 ,...,in

where the summation is over all positive integers i1 , ..., in such that i1 +
· · · + in = n and the ai1 ,...,in ’s are rational integers. If P = (u1 , ..., un ),
then any point contained in the translate of the standard unit cube by P
is of the form (u1 + t1 , ..., un + tn ) with ti ’s bounded by 1. Thus, by the
solution of Exercise 11.1.1, we deduce that the norm of any such point is
* n−1
ai1 ,...,in ui11 · · · uinn + O(x n ).
i1 ,...,in

This has absolute norm

n−1 n−1 n−1
≤ (t − δ)n + O(x n ) ≤ x − nδx n + O(x n )≤x

if we choose δ sufficiently large so that the negative sign dominates. (This

result is important to make the intuitive argument preceding Theorem
11.1.4 rigorous. Indeed, a similar argument shows that there is a δ > 0
so that for any lattice point P contained in Vx , the entire translate of the
11.2. DISTRIBUTION OF PRIME IDEALS 337

standard unit cube by P is also contained in V(t+δ)n . If U is the group

generated by the fundamental units, there is a natural action of U on Rn
which preserves the absolute value of the norm form. Then Bx can be
described as the set of orbits under this action. That is, Bx = Vx /U so
that from the containments established above, we deduce the result stated
before Theorem 11.1.4.)

11.2 Distribution of Prime Ideals

Exercise 11.2.1 Show that L(s, χ) converges absolutely for 5(s) > 1 and that

#$ χ(℘)
%−1
L(s, χ) = 1− ,
℘
N (℘)s

in this region. Deduce that L(s, χ) #= 0 for 5(s) > 1.

Solution. We have by multiplicativity of χ,

6! χ(℘)
"−1
L(s, χ) = 1−
℘
N (℘)s

and the product converges absolutely for <(s) > 1 if and only if
* 1
℘
N (℘)s

converges in this region, which is certainly the case as there are only a
bounded number of prime ideals above a given prime p. The non-vanishing
is also clear.

Exercise 11.2.2 If χ is not the trivial character, show that

,
χ(C) = 0
C

where the summation is over the ideal classes C of H.

Solution. If χ is not the trivial character, there is a C0 such that χ(C0 ) !=

1. Thus,
* *
χ(CC0 ) = χ(C0 ) χ(C),
C C

where we have used the fact that as C runs over elements of the ideal class
group, so does CC0 . The result is now immediate.
338 CHAPTER 11. DENSITY THEOREMS

Exercise 11.2.3 If C1 and C2 are distinct ideal classes, show that

,
χ(C1 )χ(C2 ) = 0.
χ

If C1 = C2 , show that the sum is hK . (This is analogous to Exercise 10.3.5.)

Solution. If C1 = C2 , the result is clear. We consider the sum

*
χ(A)
χ

for A != 1. We can then take a non-trivial character ψ of the subgroup

generated by A and extend this character to the full ideal class group in
the usual way. Then,
* *
(χψ)(A) = ψ(A) χ(A)
χ χ

and as before, the result is now evident.

Exercise 11.2.5 Let C be an ideal class of OK . For 5(s) > 1, show that
, , 1
χ(C) log L(s, χ) = hK
χ ℘m ∈C
mN (℘)ms

where the first summation is over the characters of the ideal class group and the
second summation is over all prime ideals ℘ of OK and natural numbers m such
that ℘m ∈ C.

Solution. In the left hand side, we insert the series for log L(s, χ). By
interchanging the summation and using the orthogonality relations estab-
lished in the Exercise 11.2.3, we obtain the desired result.

Exercise 11.2.6 Show that

, 1
mN (℘)ms
n≥2,℘m ∈C

converges for 5(s) > 1/2.

Solution. As noted earlier, the number of prime ideals above a fixed prime
p is at most the degree of the number field. Thus, the result is clear from
the fact that * 1
mpms
m≥2,p

converges for <(s) > 1/2.

Exercise 11.2.7 If χ2 #= χ0 show that L(1, χ) #= 0.

11.2. DISTRIBUTION OF PRIME IDEALS 339

Solution. We use the classical inequality

3 + 4 cos θ + cos 2θ ≥ 0,

as follows. We write
χ(℘) = eiθp
so that for real σ > 1,
# $
< 3 log ζK (σ) + 4 log L(σ, χ) + log L(σ, χ2 )
* 1
= (3 + 4 cos θp + cos 2θp ) ≥ 0.
m,℘
mN (℘)σm

Hence,
|ζK (σ)3 L(σ, χ)4 L(σ, χ2 )| ≥ 1.
If L(1, χ) = 0, the left hand side of this inequality tends to zero as σ → 1+ ,
which is a contradiction.

Exercise 11.2.8 Let C be a fixed ideal class in OK . Show that the set of prime
ideals ℘ ∈ C has Dirichlet density 1/hK .

Solution. We have by the orthogonality relation,

* * 1
χ(C) log L(s, χ) = hK + O(1)
χ
N (℘)s
℘∈C

as s → 1+ . Since L(1, χ) != 0, we may take limits of the left hand side

as s → 1+ and obtain a bounded quantity from the non-trivial characters.
Since L(s, χ0 ) = ζK (s), we deduce immediately that
5
℘∈C 1/N (℘)
s
1
lim = ,
s→1+ log ζK (s) hK
as desired.

Exercise 11.2.9 Let m be a natural number and (a, m) = 1. Show that the set
of primes p ≡ a(mod m) has Dirichlet density 1/φ(m).

Solution. This is immediate from Exercise 10.3.7.

Exercise 11.2.10 Show that the set of primes p which can be written as a2 +5b2
has Dirichlet density 1/4.
√
Solution. We have already seen that the class number of Q( −5) is 2.√The
set of prime ideals lying in the principal class are of the form (a + b −5)
and have norm a2 + 5b2 . By Hecke’s theorem, the Dirichlet density of these
prime ideals is 1/2 and taking into account that there are two ideals of
norm p in the principal class gives us the final density of 1/4.
340 CHAPTER 11. DENSITY THEOREMS

Exercise 11.2.11 Show that if K = Q, the principal ray class group mod m is
isomorphic to (Z/mZ)∗ .

Solution. The elements of the principal ray class group are the ideals (α)
modulo m with a totally positive generator and (α, m) = 1. The result is
now clear.

11.3 The Chebotarev density theorem

Exercise 11.3.1 Show that action of the Galois group on the set of prime ideals
lying above a fixed prime p of k is a transitive action.

Solution. Suppose not. Take a prime ideal ℘ which is not in the Galois
orbit of ℘i (say) lying above the prime ideal p. By the Chinese remainder
theorem (Theorem 5.3.13), we may find an element x ∈ ℘ and x − 1 ∈ σ(℘i )
for all σ in the Galois group. But then, NK/k (x) is an integer of Ok which
on one hand is divisible by ℘ and on the other coprime to p, a contradiction.

Exercise 11.3.4 By taking k = Q and K = Q(ζm ), deduce from Chebotarev’s

theorem the infinitude of primes in a given arithmetic progression a (mod m)
with (a, m) = 1.

Solution. The Galois group consists of automorphisms τa satisfying

τa (ζm ) = ζm
a
.

Comparing this with the action of the Frobenius automorphism of p, we see

that σp = τa where p ≡ a (mod p). By Chebotarev, the Dirichlet density
of primes p for which σp = τa is 1/φ(m).
√
Exercise 11.3.5 If k = Q and K = Q( D), deduce from Chebotarev’s theorem
that the set of primes p with Legendre symbol (D/p) = 1 is 1/2.

Solution. By Theorem 7.4.2, we see that these are precisely the set of
primes which split completely in K and by Chebotarev, the density of such
primes is 1/2.

Exercise 11.3.6 If f (x) ∈ Z[x] is an irreducible normal polynomial of degree n

(that is, its splitting field has degree n over Q), then show that the set of primes
p for which f (x) ≡ 0 (mod p) has a solution is of Dirichlet density 1/n.

Solution. By Theorem 5.5.1 and Exercise 5.5.2, we see that the set of
primes p for which f (x) ≡ 0 (mod p) has a solution coincides with the set
of primes p which split completely in the field obtained by adjoining a root
of f . By our assumption, this is a Galois extension of degree n and to say
p splits completely is equivalent to saying that σp = 1. By Chebotarev, the
Dirichlet density of such primes is 1/n.
11.4. SUPPLEMENTARY PROBLEMS 341

Exercise 11.3.7 If f (x) ∈ Z[x] is an irreducible polynomial of degree n > 1,

show that the set of primes p for which f (x) ≡ 0 (mod p) has a solution has
Dirichlet density < 1.

Solution. Let K be the splitting field of f over Q with Galois group G.

Let H be the subgroup corresponding to the field obtained by adjoining a
root of f to Q. It is not difficult to see that f (x) ≡ 0 (mod p) has a solution
if and only if the Artin symbol σp lies in some conjugate of H. This is a set
stable under conjugation. If we take into account that the identity element
is common to all the conjugate subgroups of H, we obtain

| ∪g∈G gHg −1 | ≤ [G : H](|H| − 1) + 1 = |G| + 1 − [G : H] < |G|

if [G : H] = n > 1, which is the case.

Exercise 11.3.8 Let q be prime. Show that the set of primes p for which p ≡ 1
(mod q) and
p−1
2 q ≡ 1(mod p),
has Dirichlet density 1/q(q − 1).

Solution. The second condition happens if and only if xq ≡ 2 (mod p) has

a solution and together with p ≡ 1 (mod p), the conditions are equivalent
to saying p splits completely in the field Q(ζq , 21/q ). As this field has degree
q(q − 1), the result now follows from Chebotarev’s theorem.

Exercise 11.3.9 If a natural number n is a square mod p for a set of primes p

which has Dirichlet density 1, show that n must be a square.
√
Solution. If n is not a square, the field Q( n) is quadratic over Q and by
Chebotarev, the density of primes for which n is not a square is 1/2. (This
shows that we can assert the conclusion of the theorem if the set of primes
p for which n is a square has density > 1/2.)

11.4 Supplementary Problems

Exercise 11.4.1 Let G be a finite group and for each subgroup H of G and each
irreducible character ψ of H, define aH (ψ, χ) by
,
IndG
H ψ = aH (ψ, χ)χ
χ

where the summation is over irreducible characters χ of G. For each χ, let Aχ

be the vector (aH (ψ, χ)) as H varies over all cyclic subgroups of G and ψ varies
over all irreducible characters of H. Show that the Aχ ’s are linearly independent
over Q.
342 CHAPTER 11. DENSITY THEOREMS

Solution. If *
cχ Aχ = 0
χ

for some integers cχ , then by Frobenius reciprocity, the character

*
φ= cχ χ
χ

restricts to the zero character on every cyclic subgroup. By the linear

independence of characters (or equivalently, by the orthogonality relations),
each cχ is equal to zero.

Exercise 11.4.2 Let G be a finite group with t irreducible characters. By the

previous exercise, choose a set of cyclic subgroups Hi and characters ψi of Hi
so that the t × t matrix (aHi (ψi , χ)) is non-singular. By inverting this matrix,
show that any character χ of G can be written as a rational linear combination
of characters of the form IndG
Hi ψi , with Hi cyclic and ψi one-dimensional. (This
result is usually called Artin’s character theorem and is weaker than Brauer’s
induction theorem.)

Solution. Since the row rank of a matrix is equal to the column rank, it
is clear that we can choose a set of such Hi ’s and ψi ’s. Thus,
*
IndG
Hi ψi = aHi (ψi , χ)χ.
χ

Moreover,
aH (ψ, χ) = (IndG
H ψ, χ)

are all non-negative integers. Thus, the inverse matrix consists of rational
entries.

Exercise 11.4.3 Deduce from the previous exercise that some positive integer
power of the Artin L-function L(s, χ; K/k) attached to an irreducible character
χ admits a meromorphic continuation to 5(s) = 1.

Solution. By the previous exercise, we may write L(s, χ; K/k) as a product

of functions of the form L(s, ψi , K/K Hi )mi with mi ’s rational numbers. By
the Artin reciprocity law, each L(s, ψi , K/K Hi ) coincides with L(s, χi ) with
χi a Hecke character of K Hi . As L(s, χi ) has a meromorphic continuation
to <(s) = 1, the result follows.

Exercise 11.4.4 If K/k is a finite Galois extension of algebraic number fields

with group G, show that
#
ζK (s) = L(s, χ; K/k)χ(1) ,
χ

where the product is over all irreducible characters χ of G.

11.4. SUPPLEMENTARY PROBLEMS 343

Solution. Since the right hand side represents the L-function attached to
the regular representation, which is IndG
1 1, we have that it is equal to

L(s, 1, K/K) = ζK (s),

by the invariance property under induction of Artin L-series.

Exercise 11.4.5 Fix a complex number s0 ∈ C with 5(s0 ) ≥ 1 and any finite
Galois extension K/k with Galois group G. For each subgroup H of G define the
Heilbronn character θH by
,
θH (g) = n(H, χ)χ(g)
χ

where the summation is over all irreducible characters χ of H and n(H, χ) is the
order of L(s, χ; K/K H ) at s = s0 . (By Exercise 11.4.3, the order is a rational
number.) Show that θG |H = θH .

Solution. We have *
θG |H = n(G, χ)χ|H .
χ

But *
χ|H = (χ|H , ψ)ψ
ψ

where the sum is over irreducible characters ψ of H. Thus,

/ 0
* *
θG |H = n(G, χ)(χ|H , ψ) ψ.
ψ χ

By Frobenius reciprocity,

(χ|H , ψ) = (χ, IndG

H ψ)

so that the inner sum is

*
n(G, χ)(χ, IndG
H ψ).
χ

This is equal to n(H, ψ) since

6
L(s, χ, K/k)(χ,IndH ψ) .
G
L(s, ψ, K/K H ) = L(s, IndG
H ψ, K/k) =
χ

Thus, *
θG |H = n(H, ψ)ψ = θH .
ψ

Exercise 11.4.6 Show that θG (1) equals the order at s = s0 of the Dedekind
zeta function ζK (s).
344 CHAPTER 11. DENSITY THEOREMS

Solution. We have *
θG (1) = n(G, χ)χ(1)
χ

which by Exercise 11.4.4 is the order of ζK (s) at s = s0 .

Exercise 11.4.7 Show that

,
n(G, χ)2 ≤ (ords=s0 ζK (s))2 .
χ

Solution. We compute (θG , θG ) using the orthogonality relations to obtain

*
n(G, χ)2 .
χ

On the other hand,

1 *
(θG , θG ) = |θG (g)|2 .
|G|
g∈G

By Exercise 11.4.5, θG (g) = θ/g0 (g). But if H is abelian, n(H, ψ) ≥ 0 by

Artin’s reciprocity law and so for h ∈ H,
*
|θH (h)| ≤ n(H, ψ)|ψ(h)|.
ψ

Thus,
*
|θH (h)| ≤ n(H, ψ) = θH (1),
ψ

which by Exercise 11.4.6 is the order of the Dedekind zeta function ζK (s)
at s = s0 . The result is now immediate.

Exercise 11.4.8 For any irreducible non-trivial character χ, deduce that

L(s, χ; K/k)

admits an analytic continuation to s = 1 and that L(1, χ; K/k) #= 0.

Solution. Since the Dedekind zeta function has a simple pole at s = 1, we

see that the previous exercise applied to the point s0 = 1 implies
*
n(G, χ)2 ≤ 0
χ*=1

because n(G, 1) = 1. Hence, n(G, χ) = 0 for any χ != 1. By Exercise 11.4.3

we are done.
11.4. SUPPLEMENTARY PROBLEMS 345

Exercise 11.4.9 Fix a conjugacy class C in G = Gal(K/k) and choose gC ∈ C.

Show that
, 1 |C| ,
= χ(gC ) log L(s, χ; K/k).
m N (p)ms |G| χ
m,p, σp ∈C

Solution. This is an immediate consequence of the orthogonality relations.

Exercise 11.4.10 Show that

&
p, σp ∈C 1/N (p)s |C|
lim =
s→1+ log ζk (s) |G|
which is Chebotarev’s theorem.

Solution. We take the limit as s → 1+ in both sides of the equation of the

previous exercise. Observe that by Exercise 11.4.8, the limit as s → 1+ of
log L(s, χ; K/k) for χ != 1 is finite. Since L(s, 1; K/k) = ζk (s), the result is
now immediate.

Exercise 11.4.11 Show that ζK (s)/ζk (s) is entire. (This is called the Brauer-
Aramata theorem.)

Solution. By Exercise 11.4.7, we have

n(G, 1)2 ≤ (ords=s0 ζK (s))2 .
But n(G, 1) = ords=s0 ζk (s) and as both ζK (s) and ζk (s) are regular every-
where except at s = 1, we deduce that
ords=s0 ζk (s) ≤ ords=s0 ζK (s),
for s0 != 1. But for s0 = 1, this inequality is also true. The result is now
immediate. (It is a famous conjecture of Dedekind that if K is an arbitrary
extension of k (not necessarily Galois), then ζK (s)/ζk (s) is always entire.
This exercise shows that the conjecture is true in the Galois case. The
result is also known if K is contained in a solvable extension of k.)

Exercise 11.4.12 (Stark) Let K/k be a finite Galois extension of algebraic num-
ber fields. If ζK (s) has a simple zero at s = s0 , then L(s, χ; K/k) is analytic at
s = s0 for every irreducible character χ of Gal(K/k).

Solution. If ζK (s) has a simple zero at s = s0 , then

*
n(G, χ)2 ≤ 1.
χ

By the meromorphy of Artin L-series, we have that each n(G, χ) is an

integer. The inequality implies that for at most one χ, we have |n(G, χ)| =
1. If χ is non-abelian, then the factorization of ζK (s) as in Exercise 11.4.4
gives a contradiction for the corresponding L-function introduces a pole or
zero of order greater than 1. Hence χ is abelian, but in this case the result
is known by Artin reciprocity and Hecke’s theorem.
346 CHAPTER 11. DENSITY THEOREMS

Exercise 11.4.13 (Foote- K. Murty) For any irreducible character χ of Gal(K/k),

show that
L(s, χ, K/k)ζK (s)
is analytic for s #= 1.

Solution. This is immediate from the inequality

|n(G, χ)| ≤ ords=s0 ζK (s)

for s0 != 1.

Exercise 11.4.14 If K/k is solvable, show that

,
n(G, χ)2 ≤ (ords=s0 ζK (s)/ζk (s))2 .
χ&=1

Solution. Let f = θG − n(G, 1)1 and note that

*
(f, f ) = n(G, χ)2 .
χ*=1

Now by Exercise 11.4.5

*
f (g) = θ/g0 (g) − n(G, 1) = n(:g;, 1) − n(G, 1) + n(:g;, ψ)ψ(g).
ψ*=1

The subfield fixed by :g; is a subfield of K and we know Dedekind’s con-

jecture for this extension. Thus,

n(:g;, 1) − n(G, 1) ≥ 0.

Therefore,
*
|f (g)| ≤ n(:g;, 1) − n(G, 1) + n(:g;, ψ) = ords=s0 (ζK (s)/ζk (s)),
ψ*=1

from which the inequality follows.

Bibliography

[B] Bell, E.T. Men of Mathematics. Simon and Schuster, New York,
1937.

[BL] Bateman, P. and Low, M.E. Prime numbers in arithmetic progres-

sion with difference 24. Amer. Math. Monthly 72 (1965), 139–143.

[Er] Erdös, P. Problems and results on consecutive integers. Eureka 38

(1975/6), 3–8.

[FM] R. Foote and V. Kumar Murty, Zeros and Poles of Artin L-series,
Math. Proc. Camb. Phil. Soc., 105 (1989), 5-11.

[Ga] Gauss, C.F. Disquisitiones Arithmeticae. Springer-Verlag, New

York, 1986. (Translated by Arthur A. Clarke, revised by William
C. Waterhouse.)
√
[Ha] Harper, M., Z[ 14] is Euclidean, Canadian Journal of Mathematics,
56(1), (2004), 55–70.

[He] Hecke, E., Lectures on the Theory of Algebraic Numbers, GTM 77,
Springer-Verlag, 1981.

[HM] Harper, M., and Murty, R., Euclidean rings of algebraic integers,
Canadian Journal of Mathematics, 56(1), (2004), 71-76.

[IR2] Ireland, K. and Rosen, M. A Classical Introduction to Modern Num-

ber Theory, 2nd Ed. Springer-Verlag, New York, 1990.

[La] Lang, S., Algebraic Number Theory, Springer-Verlag, 1986.

[L-K] Loo-Keng, Hua, Introduction to Number Theory. Springer-Verlag,

Berlin, 1982.

[Ma] Marcus, D. Number Fields. Springer-Verlag, New York, 1977.

[Mat] Matsumura, H., Commutative Ring Theory, Cambridge University

Press, 1989.

347
348 BIBLIOGRAPHY

[Mu] Murty, R. Primes in certain arithmetic progressions. J. Madras Uni-

versity, Section B 51 (1988), 161–169.
[Mu2] Murty, R., Problems in Analytic Number Theory, GTM/RIM 206,
Springer-Verlag, 2001.
[Na] Narasimhan, R. et al. Algebraic Number Theory. Commercial Print-
ing Press, Bombay.

[N] Narkiewicz, W., Elementary and analytic theory of algebraic num-

bers, Second edition, Springer-Verlag, Berlin, PWN - Polish Scien-
tific Publishers, Warsaw, 1990.

[Ro] Rosen, K. Elementary Number Theory and its Applications.

Addison-Wesley, Murray Hill, 1988.
[S] Schur, I. Über die Existenz unendlich vieler Primzahlen in einigen
speziellen arithmetischen Progressionen. S-B Berlin Math. Ges., 11
(1912), 40–50, appendix to Archiv der Math. und Phys. (3), vol. 20
(1912–1913). (See also his Collected Papers, Springer-Verlag, New
York, 1973.)
[Sc] Schoof, R. and Corrales-Rodrigáñez, C. The support problem and
its elliptic analogue. J. Number Theory 64 (1997), no. 2, 276–290.

[Se] J.-P. Serre, Linear Representations of Finite Groups, GTM 42,

Springer-Verlag, 1977.
[Sil] Silverman, J. Wieferich’s criterion and the ABC conjecture. J. Num-
ber Theory 30 (1988), no. 2, 226–237.

[St] Stewart, I. and Tall, D. Algebraic Number Theory. Chapman and

Hall, London, 1979.
Index

ABC Conjecture, 9 codifferent, 68, 242

L-function, 133 conjugate
OK , 37 complex, 184
φ-function, 6 of an algebraic integer, 34
π(x), 6 continued fraction
finite, 108
abscissa of convergence, 134 simple, 108
algebraic integer, 27 convergent, 108
conjugates, 34 Corrales-Rodrigáñez, C., 173
degree of, 28 cubic
minimal polynomial, 28 character, 118
algebraic number, 27 residue character, 117
algebraic number field, 33 cubic reciprocity, law of, 118, 119
degree of, 33 cyclotomic field
ambiguous ideal class, 116, 295 discriminant, 52, 224
Ankeny, 250 integral basis, 49, 52, 212, 224
Artin character theorem, 342 cyclotomic polynomial, 38, 205
Artin reciprocity law, 152
Artin symbol, 151 de la Vallée Poussin, Ch., 6
Artin’s conjecture, 152
decomposition group, 151
associate, 13
Dedekind domain, 56
Bateman, P., 92 Dedekind zeta function, 127, 146,
Bertrand’s Postulate, 11, 177 313
Brauer induction theorem, 152 Dedekind’s conjecture, 345
Brauer-Aramata theorem, 155, 345 Dedekind’s theorem, 63
different, 63, 68, 234, 242
character Diophantine equations
cubic, 118 x2 + y 2 = z 2 , 8, 167
cubic residue, 117 x2 + 11 = y 3 , 25, 190
Chebotarev density theorem, 151 x2 + 5 = y 3 , 74
Chinese Remainder Theorem, 60, x2 + k = y 3 , 75, 247
175 x3 + y 3 + z 3 = 0, 18
Chowla, 250 x4 + y 4 = z 4 , 8, 167
class number, 72 x4 + y 4 = z 2 , 8, 167
formula, 137, 329 x4 − y 4 = 2z 2 , 10, 174
class number formula, 145 x4 − y 4 = z 2 , 8, 168

349
350 INDEX

y 2 + 1 = xp , 22 greatest common divisor, 59

y 2 + 1 = x3 , 17, 183
y 2 + 2 = x3 , 22, 187 Hadamard, J., 6
y 2 + 4 = x3 , 21 Hecke, 132
using Thue’s theorem, 31 Hecke L-function, 146
Dirichlet, 91 Hecke, E., 146
L-function, 133 Hermite, 31, 79, 260
character, 133 Hurwitz constant, 70
class number formula, 137, 329 hyperbola method, 131
hyperbola method, 131
Unit Theorem, 99, 104 ideal
Dirichlet density, 149 ambiguous, 116, 295
discriminant, 45 class group, 72
of a cyclotomic field, 52, 224 fractional, 57
of a quadratic field, 89, 270 prime, 10, 173
sign of, 52, 222 principal, 14
index
Eisenstein criterion, 37, 46, 201 of an algebraic integer, 45
Eisenstein integers, 17 integer
Eisenstein reciprocity law, 122, 124 algebraic, 27
embedding, 34 perfect, 10
Erdös, P., 9 powerfull, 9, 171
Euclidean domain, 15, 35, 201 radical of, 9
quadratic, 17, 18, 22, 25, 52, squarefree, 8, 170
181, 186–189, 222 squarefull, 9, 171
Euler, 6, 172 integral basis, 43
φ-function, 6 construction of, 51, 220
cubic field, 48
Fermat, 8, 167 cyclotomic field, 49, 212
Last Theorem, 8, 9, 167 quadratic field, 47
n = 3, 18 irreducible, 13
first case, 124, 125
little Theorem, 6, 161, 171, Jacobi sum, 119
174 Jacobi symbol, 95, 273
number, 8, 10, 173 reciprocity law, 96, 274
fractional ideal, 57
Frobenius automorphism, 151 Kisilevsky, H., 10, 173
fundamental units, 99, 104 Kronecker, 100
Furtwangler, 124 Kronecker symbol, 96, 274

Galois extension, 34 Lagrange, 171

Gamma function, 147 Langlands program, 153
Gauss sum, 84, 118, 136, 324 least common multiple, 60
Gauss’ lemma, 15, 36 Legendre symbol, 82
Germain, Sophie, 11, 176 Lindemann, 31
Granville, A., 172 Liouville’s theorem, 29, 34
INDEX 351

Low, M.E., 92 prime decomposition, 88

units of, 106, 280, 281
Masser, 9 zeta functions, 130
Mersenne numbers, 174 quadratic reciprocity, law of, 87
minimal polynomial, 28
Minkowski’s bound, 78, 256 radical of an integer, 9
Murty, M.R., 94 Ramanujan, 178
ramification, 243
Noetherian, 55 ramification degree, 63
norm ray class group, 150
of an algebraic integer, 41 regulator, 140
of an ideal, 49 relative different, 68, 242
norm map, 13, 67, 242 relative discriminant, 68, 243
normal closure, 34 relative norm, 67, 241
normal extension, 34 relative trace, 67, 241
number Riemann hypothesis, 138
algebraic, 27 Riemann zeta function, 127
Fermat, 8, 10, 173 Roth, 32
Mersenne, 174
perfect, 10 Schoof, 173
transcendental, 28, 38, 202 Silverman, 171
split completely, 151
Oesterlé, 9 squarefree, 8, 170
squarefull, 9, 171
Pólya–Vinogradov inequality, 136, Stickelberger’s criterion, 50, 214
325
perfect number, 10 tensor product, 52, 223
power residue symbol, 123 Thue, 31, 38, 39, 204, 206
powerfull, 9, 171 trace map, 67, 242
primary algebraic integer, 118, 124 trace of an algebraic integer, 41
prime number theorem, 6 transcendental number, 28, 38, 202
primes in arithmetic progressions,
91, 134 unique factorization
primitive element, theorem of, 33 fractional ideals, 59, 229
primitive polynomial, 16 ideals in OK , 58
principal ideal, 14 integers, 4
principal ideal domain, 14, 35, 201 unique factorization domain, 14
unit, 13
quadratic
nonresidue, 82 Vandermonde determinant, 26, 195,
residue, 82 200, 210
quadratic field, 88 volume
discriminant, 89, 270 of a domain, 76
Euclidean, 17, 18, 22, 25, 52, of a lattice, 105, 280
181, 186–189, 222
integral basis, 47 Wieferich, 9, 124
352 INDEX

Wilson’s theorem, 10, 175, 185

zeta function
Dedekind, 127, 313
Riemann, 127